elife00776.xml

<?xml version="1.0" encoding="UTF-8"?><!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Archiving and Interchange DTD v1.1d1 20130915//EN"  "JATS-archivearticle1.dtd"><article article-type="research-article" dtd-version="1.1d1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><front><journal-meta><journal-id journal-id-type="nlm-ta">elife</journal-id><journal-id journal-id-type="hwp">eLife</journal-id><journal-id journal-id-type="publisher-id">eLife</journal-id><journal-title-group><journal-title>eLife</journal-title></journal-title-group><issn publication-format="electronic">2050-084X</issn><publisher><publisher-name>eLife Sciences Publications, Ltd</publisher-name></publisher></journal-meta><article-meta><article-id pub-id-type="publisher-id">00776</article-id><article-id pub-id-type="doi">10.7554/eLife.00776</article-id><article-categories><subj-group subj-group-type="display-channel"><subject>Research article</subject></subj-group><subj-group subj-group-type="heading"><subject>Genomics and evolutionary biology</subject></subj-group><subj-group subj-group-type="heading"><subject>Plant biology</subject></subj-group></article-categories><title-group><article-title>Cytoplasmic genetic variation and extensive cytonuclear interactions influence natural variation in the metabolome</article-title></title-group><contrib-group><contrib contrib-type="author" id="author-4739"><name><surname>Joseph</surname><given-names>Bindu</given-names></name><xref ref-type="aff" rid="aff1"/><xref ref-type="other" rid="par-1"/><xref ref-type="fn" rid="con1"/><xref ref-type="fn" rid="conf2"/></contrib><contrib contrib-type="author" id="author-4740"><name><surname>Corwin</surname><given-names>Jason A</given-names></name><xref ref-type="aff" rid="aff1"/><xref ref-type="other" rid="par-1"/><xref ref-type="fn" rid="con2"/><xref ref-type="fn" rid="conf2"/></contrib><contrib contrib-type="author" id="author-4741"><name><surname>Li</surname><given-names>Baohua</given-names></name><xref ref-type="aff" rid="aff1"/><xref ref-type="other" rid="par-1"/><xref ref-type="fn" rid="con3"/><xref ref-type="fn" rid="conf2"/></contrib><contrib contrib-type="author" id="author-4742"><name><surname>Atwell</surname><given-names>Suzi</given-names></name><xref ref-type="aff" rid="aff1"/><xref ref-type="other" rid="par-1"/><xref ref-type="fn" rid="con4"/><xref ref-type="fn" rid="conf2"/></contrib><contrib contrib-type="author" corresp="yes" id="author-1122"><name><surname>Kliebenstein</surname><given-names>Daniel J</given-names></name><xref ref-type="aff" rid="aff1"/><xref ref-type="aff" rid="aff2"/><xref ref-type="corresp" rid="cor1">*</xref><xref ref-type="other" rid="par-1"/><xref ref-type="fn" rid="con5"/><xref ref-type="fn" rid="conf1"/></contrib><aff id="aff1"><institution content-type="dept">Department of Plant Sciences</institution>, <institution>University of California, Davis</institution>, <addr-line><named-content content-type="city">Davis</named-content></addr-line>, <country>United States</country></aff><aff id="aff2"><institution content-type="dept">DynaMo Center of Excellence</institution>, <institution>University of Copenhagen</institution>, <addr-line><named-content content-type="city">Frederiksberg</named-content></addr-line>, <country>Denmark</country></aff></contrib-group><contrib-group content-type="section"><contrib contrib-type="editor"><name><surname>Weigel</surname><given-names>Detlef</given-names></name><role>Reviewing editor</role><aff><institution>Max Planck Institute for Developmental Biology</institution>, <country>Germany</country></aff></contrib></contrib-group><author-notes><corresp id="cor1"><label>*</label>For correspondence: <email>kliebenstein@ucdavis.edu</email></corresp></author-notes><pub-date date-type="pub" publication-format="electronic"><day>08</day><month>10</month><year>2013</year></pub-date><pub-date pub-type="collection"><year>2013</year></pub-date><volume>2</volume><elocation-id>e00776</elocation-id><history><date date-type="received"><day>26</day><month>03</month><year>2013</year></date><date date-type="accepted"><day>03</day><month>09</month><year>2013</year></date></history><permissions><copyright-statement>© 2013, Joseph et al</copyright-statement><copyright-year>2013</copyright-year><copyright-holder>Joseph et al</copyright-holder><license xlink:href="http://creativecommons.org/licenses/by/3.0/"><license-p>This article is distributed under the terms of the <ext-link ext-link-type="uri" xlink:href="http://creativecommons.org/licenses/by/3.0/">Creative Commons Attribution License</ext-link>, which permits unrestricted use and redistribution provided that the original author and source are credited.</license-p></license></permissions><self-uri content-type="pdf" xlink:href="elife00776.pdf"/><abstract><object-id pub-id-type="doi">10.7554/eLife.00776.001</object-id><p>Understanding genome to phenotype linkages has been greatly enabled by genomic sequencing. However, most genome analysis is typically confined to the nuclear genome. We conducted a metabolomic QTL analysis on a reciprocal RIL population structured to examine how variation in the organelle genomes affects phenotypic variation. This showed that the cytoplasmic variation had effects similar to, if not larger than, the largest individual nuclear locus. Inclusion of cytoplasmic variation into the genetic model greatly increased the explained phenotypic variation. Cytoplasmic genetic variation was a central hub in the epistatic network controlling the plant metabolome. This epistatic influence manifested such that the cytoplasmic background could alter or hide pairwise epistasis between nuclear loci. Thus, cytoplasmic genetic variation plays a central role in controlling natural variation in metabolomic networks. This suggests that cytoplasmic genomes must be included in any future analysis of natural variation.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.001">http://dx.doi.org/10.7554/eLife.00776.001</ext-link></p></abstract><abstract abstract-type="executive-summary"><object-id pub-id-type="doi">10.7554/eLife.00776.002</object-id><title>eLife digest</title><p>The vast majority of genes in plant and animal cells are located on chromosomes within the nucleus. However, cells also contain a small number of genes outside the nucleus in cellular organelles such as the mitochondria, which generate energy, and the chloroplasts, which carry out photosynthesis. All these non-nuclear genes comprise the organellar genome.</p><p>When trying to explain how variation in genes leads to differences in the characteristics of animals and plants, geneticists have historically paid most attention to the genes inside the nucleus. However, more recent work has shown that variation in the organellar genome can also contribute to differences between individuals, although the relative contribution of organellar genes versus nuclear genes remains unclear.</p><p>Now, Joseph et al. have performed the first large-scale analysis of how variation in the organellar genome affects the characteristics (or phenotype) of the plant model organism, Arabidopsis. The study examined the degree to which variation in each of roughly 13,000 nuclear genes and 200 organellar genes affected the levels of thousands of metabolites inside cells.</p><p>This metabolomics analysis revealed that variation in the organellar genome contributed to variation in the levels of more than 80% of the metabolites studied. Organellar genes also helped to regulate the effect of nuclear genes. This combination of direct and indirect influences helps to explain how a small number of organellar genes can have a disproportionately large effect on phenotype.</p><p>The work of Joseph et al. suggests that the role of the organellar genome has been significantly underestimated to date, and that geneticists should consider variation in both the nuclear and organellar genome when attempting to determine how genes affect phenotype.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.002">http://dx.doi.org/10.7554/eLife.00776.002</ext-link></p></abstract><kwd-group kwd-group-type="author-keywords"><title>Author keywords</title><kwd>genomic variation</kwd><kwd>maternal genetics</kwd><kwd>organellar variation</kwd></kwd-group><kwd-group kwd-group-type="research-organism"><title>Research organism</title><kwd>Arabidopsis</kwd></kwd-group><funding-group><award-group id="par-1"><funding-source><institution-wrap><institution>National Science Foundation</institution></institution-wrap></funding-source><award-id>DBI0820580</award-id><principal-award-recipient><name><surname>Joseph</surname><given-names>Bindu</given-names></name><name><surname>Corwin</surname><given-names>Jason A</given-names></name><name><surname>Li</surname><given-names>Baohua</given-names></name><name><surname>Atwell</surname><given-names>Suzi</given-names></name><name><surname>Kliebenstein</surname><given-names>Daniel J</given-names></name></principal-award-recipient></award-group><funding-statement>The funder had no role in study design, data collection and interpretation, or the decision to submit the work for publication.</funding-statement></funding-group><custom-meta-group><custom-meta><meta-name>elife-xml-version</meta-name><meta-value>2</meta-value></custom-meta><custom-meta specific-use="meta-only"><meta-name>Author impact statement</meta-name><meta-value>Genes located within cellular organelles impact phenotype more than expected due to their interactions with nuclear genes.</meta-value></custom-meta></custom-meta-group></article-meta></front><body><sec id="s1" sec-type="intro"><title>Introduction</title><p>A central goal of modern biology is to understand how the integration of gene functions across a genome lead to the individual’s specific phenotype. A key facet to this effort is to develop models that would allow directly inferring a species phenotypic variation from its genetic variation. This goal of mathematically linking genetic to phenotypic variation is central to all studies of quantitative genetics ranging from human genetics to plant breeding to ecology and has led to massive genome resequencing projects focused on developing the genomic databases to allow these studies (<xref ref-type="bibr" rid="bib45">Liti et al., 2009</xref>; <xref ref-type="bibr" rid="bib2">Altshuler et al., 2010</xref>; <xref ref-type="bibr" rid="bib9">Cao et al., 2011</xref>). However, in most quantitative genomics studies, genome analysis is largely confined to the nuclear genome with much less attention paid to the organellar genome. This is in contrast to the central role that the organellar genome plays in controlling organismal metabolism and function and the developing body of literature suggesting that organellar genomic variation can modulate the effects of nuclear genomic variation.</p><p>Genomic variation in human organelles has been linked to several severe diseases (<xref ref-type="bibr" rid="bib73">Wallace et al., 1988</xref>; <xref ref-type="bibr" rid="bib70">Taylor and Turnbull, 2005</xref>; <xref ref-type="bibr" rid="bib63">Schon et al., 2012</xref>). However, these organellar variants are typically rare with large phenotypic consequences such that they can be followed using simple maternal inheritance studies without consideration for quantitative variation in the nuclear genome (<xref ref-type="bibr" rid="bib63">Schon et al., 2012</xref>). More recently, quantitative studies on human diseases suggest that genetic variation in organellar genomes modify the quantitative effect of nuclear loci and disease phenotypes (<xref ref-type="bibr" rid="bib7">Battersby and Shoubridge, 2001</xref>; <xref ref-type="bibr" rid="bib50">McRae et al., 2008</xref>; <xref ref-type="bibr" rid="bib64">Schon and Przedborski, 2011</xref>). Additionally, the use of structured populations in mice, yeast, and birds have shown that cytoplasmic genome variation can influence high order phenotypes including fitness, cognition, and biomass (<xref ref-type="bibr" rid="bib59">Roubertoux et al., 2003</xref>; <xref ref-type="bibr" rid="bib80">Zeyl et al., 2005</xref>; <xref ref-type="bibr" rid="bib53">Park et al., 2006</xref>; <xref ref-type="bibr" rid="bib16">Dimitrov et al., 2009</xref>). Within ecological studies, mitochondrial variation has also been shown to alter fitness and create hybrid isolation in a variety of invertebrate species (<xref ref-type="bibr" rid="bib76">Willett and Burton, 2004</xref>; <xref ref-type="bibr" rid="bib72">Wade and Goodnight, 2006</xref>; <xref ref-type="bibr" rid="bib18">Dowling et al., 2007</xref>, <xref ref-type="bibr" rid="bib19">2010</xref>; <xref ref-type="bibr" rid="bib77">Wolf, 2009</xref>; <xref ref-type="bibr" rid="bib75">Willett, 2012</xref>). However, these studies didn’t directly interrogate the interaction of the nuclear and cytoplasmic genomes for quantitative traits or test the breadth of phenotypes affected.</p><p>In plants, genetic variation in mitochondria is also linked with large qualitative phenotypes, such as cytoplasmic male sterility (<xref ref-type="bibr" rid="bib30">Hanson, 1991</xref>; <xref ref-type="bibr" rid="bib62">Schnable and Wise, 1998</xref>). Plant breeding has a long history of using diallele crosses to test for the presence of maternal effects (cytoplasmic genetic variation) on a phenotype and this was recently extended to a small reciprocal F2 family structure to show that the cytoplasmic effects could have significant impacts on plant height in maize (<xref ref-type="bibr" rid="bib68">Tang et al., 2013</xref>). In rice, agronomic traits have also been shown to be influenced by interactions between cytoplasmic and nuclear genomes but the specific loci were not identified (<xref ref-type="bibr" rid="bib69">Tao et al., 2004</xref>). However, these reciprocal F2 populations typically have generated the impression that cytoplasmic effects on phenotypic variation are quite small in plants possibly because of an inability to account for interactions between the nuclear and cytoplasmic genomes (<xref ref-type="bibr" rid="bib65">Singh, 1965</xref>; <xref ref-type="bibr" rid="bib15">Crane and Nyquist, 1967</xref>; <xref ref-type="bibr" rid="bib20">Eenink and Garretsen, 1980</xref>; <xref ref-type="bibr" rid="bib51">Miura et al., 1997</xref>; <xref ref-type="bibr" rid="bib54">Primomo et al., 2002</xref>). In contrast to the previous estimates of small effects, genomic sequencing within Arabidopsis has shown the presence of considerable genetic polymorphism in both the plastidic and mitochondrial genomes suggesting the potential for broad phenotypic consequences (<xref ref-type="bibr" rid="bib52">Moison et al., 2010</xref>).</p><p>The above studies have shown that cytoplasmic genome variation can influence phenotypic variation. Additionally, genes underlying qualitative interactions between the nucleus and cytoplasm leading to cytoplasmic male sterility and interspecific isolation have been identified. There are, however, numerous open questions remaining about how cytoplasmic variation influences quantitative phenotypic variation. What is the level of phenotypic variation influenced by genetic variation in the cytoplasm in comparison to individual nuclear loci? How much epistatic interaction is there between genetic variation in nuclear loci and the cytoplasmic genomes? What is the breadth of phenotypes that might be influenced by cytoplasmic variation?</p><p>To begin answering these questions, we utilized metabolomics to investigate how genetic variation in the cytoplasmic and nuclear genomes interacts to control metabolome variation in the reciprocal Arabidopsis Kas × Tsu recombinant inbred line (RIL) population (<xref ref-type="bibr" rid="bib48">McKay et al., 2008</xref>; <xref ref-type="bibr" rid="bib33">Juenger et al., 2010</xref>). We focused on metabolomics because it is cost effective for large sample numbers and because plant organelles, mitochondria and plastids, are central to the function of plant primary metabolism, as well as many specialized metabolites in both energy generation and biosynthesis (<xref ref-type="bibr" rid="bib25">Fiehn et al., 2000</xref>; <xref ref-type="bibr" rid="bib58">Roessner et al., 2001</xref>; <xref ref-type="bibr" rid="bib24">Fiehn, 2002</xref>). Within Arabidopsis, there are nearly 13,000 genes in the nuclear genome predicted to be involved in metabolic processes of which ∼3000 may be targeted to the mitochondria and the plastid (<xref ref-type="bibr" rid="bib4">Arabidopsis genome initiative, 2000</xref>). This included complete biosynthetic pathways for numerous amino acids and key energy production processes like respiration and photosynthesis. Aiding these processes are 88 total predicted genes in the plastidic genome and 121 in the mitochondrial genome; most of which function to facilitate the metabolic processes that occur in organelles using both nuclear and cytoplasmic-encoded proteins. Thus, variation in the organellar genomes could directly influence the function of any of nuclear gene functioning within the organelle (<xref ref-type="bibr" rid="bib22">Etterson et al., 2007</xref>; <xref ref-type="bibr" rid="bib67">Tang et al., 2007</xref>; <xref ref-type="bibr" rid="bib77">Wolf, 2009</xref>). This suggests that metabolism is an ideal phenotype to use for testing how cytoplasmic variation can influence phenotypic variation.</p><p>We measured the metabolome within the Kas × Tsu RIL population to test how cytoplasmic variation can affect quantitative variation within the metabolome. This population was generated from a reciprocal cross and approximately half of the resulting lines carry Kas organelles while the other half carry Tsu organelles allowing for explicit analysis of the influence of cytoplasmic genetic variation (<xref ref-type="bibr" rid="bib48">McKay et al., 2008</xref>; <xref ref-type="bibr" rid="bib33">Juenger et al., 2010</xref>). Analysis of the Kas × Tsu RIL population metabolome showed that genetic variation in the organelles influenced the accumulation of over 80% of the detectable metabolites. In contrast to previous observations suggesting that the cytoplasm has only small effects, phenotypic changes associated with cytoplasmic variation were as large and often larger than that found for individual nuclear loci (<xref ref-type="bibr" rid="bib65">Singh, 1965</xref>; <xref ref-type="bibr" rid="bib15">Crane and Nyquist, 1967</xref>; <xref ref-type="bibr" rid="bib20">Eenink and Garretsen, 1980</xref>; <xref ref-type="bibr" rid="bib51">Miura et al., 1997</xref>; <xref ref-type="bibr" rid="bib54">Primomo et al., 2002</xref>). In addition, the cytoplasm was found to be a central hub in the epistatic network controlling natural variation in plant metabolism. This centrality led to the cytoplasmic background displaying the unexpected ability to hide or alter epistatic interactions between nuclear loci. Thus, genetic variation in the cytoplasmic organelles has widespread and large quantitative effects on natural phenotypic variation and can influence the link between nuclear loci.</p></sec><sec id="s2" sec-type="results"><title>Results</title><sec id="s2-1"><title>Comparative metabolome genetics across populations</title><p>To partition phenotypic variance between the effects of the nuclear and cytoplasmic genomes upon the <italic>A. thaliana</italic> metabolome, we identified and measured metabolite levels, using non-targeted GC-TOF-MS, in leaf tissues of 316 lines of the Kas × Tsu <italic>A. thaliana</italic> RIL population harvested with fourfold replication across two separate experiments. Within the 316 lines for this population, 136 have the Kas cytoplasm and 180 have the Tsu cytoplasm. A total of 2435 metabolites were identified in over 25% of the RILs with 215 of these being known compounds and 2220 being unidentified compounds. Of these 2435 metabolites, 559 were identified in both experiments (161 known and 398 unknown) while the rest were specific to one of the two experiments.</p><p>Classical breeding studies utilize reciprocal crosses with linear modeling to partition heritability between the nuclear and cytoplasmic genomes (<xref ref-type="bibr" rid="bib65">Singh, 1965</xref>; <xref ref-type="bibr" rid="bib15">Crane and Nyquist, 1967</xref>; <xref ref-type="bibr" rid="bib20">Eenink and Garretsen, 1980</xref>; <xref ref-type="bibr" rid="bib51">Miura et al., 1997</xref>; <xref ref-type="bibr" rid="bib54">Primomo et al., 2002</xref>). In line with the standard heritability calculations for classical breeding experiments with reciprocal crosses, we used a linear modeling approach to approximate the level of heritability in this population that can be statistically ascribed to the nuclear and cytoplasmic genome. The line heritability model simply partitions the RILs into two subpopulations based on the maternal parent and tests the level of reproducibility that is controlled by differences between the lines and between the subpopulations. Using the line heritability model, we focused on the 559 metabolites found in both experiments to estimate the broad sense heritability of the metabolome. This analysis showed that 361 metabolites had a significant (p&lt;0.01) line effect, of which 334 metabolites showed significant heritability based on the maternal parent subpopulation. This maternal effect was quite small with an average of 1.6 ± 0.1% (range 15–0%) (<xref ref-type="fig" rid="fig1">Figure 1</xref> and <xref ref-type="supplementary-material" rid="SD1-data">Figure 1—source data 1</xref>). In comparison, 77 metabolites had significant nuclear broad sense heritability with an average of 21.3 ± 0.3% (range 58–2%) (<xref ref-type="fig" rid="fig1">Figure 1</xref> and <xref ref-type="supplementary-material" rid="SD1-data">Figure 1—source data 1</xref>). This is a similar level of variation to that found in a previous analysis of metabolomic variation in Arabidopsis (<xref ref-type="bibr" rid="bib60">Rowe et al., 2008</xref>; <xref ref-type="bibr" rid="bib13">Chan et al., 2010</xref>). The combined variance of the genetic components, both nuclear and cytoplasmic, explained approximately 19 times as much of the variance as the combination of experiment and block (<xref ref-type="supplementary-material" rid="SD1-data">Figure 1—source data 1</xref>) (<xref ref-type="bibr" rid="bib60">Rowe et al., 2008</xref>; <xref ref-type="bibr" rid="bib13">Chan et al., 2010</xref>). This raises an apparent disparity between the nuclear genome having high heritability that is significant for a low fraction of metabolites (<xref ref-type="supplementary-material" rid="SD1-data">Figure 1—source data 1</xref>). In contrast, the cytoplasmic effects have relatively low average heritability but significantly impact most metabolites (<xref ref-type="supplementary-material" rid="SD1-data">Figure 1—source data 1</xref>). The most likely explanation for this disparity is the large difference in the degrees of freedom for the two factors, nuclear vs cytoplasmic used for the significance test. When conducting the ANOVA test, there are only two maternal subpopulations, Kas or Tsu, whereas there are 316 different nuclear genomes/RILs, and as such, the F value for the same level of variance will be very different in the two tests (<xref ref-type="bibr" rid="bib61">Rowe and Kliebenstein, 2008</xref>). Thus, variation in most metabolites identifies a significant effect of cytoplasmic genetic variation, even though this effect appears to be smaller than that for the nuclear genome in the standard line heritability estimation approach.<fig id="fig1" position="float"><object-id pub-id-type="doi">10.7554/eLife.00776.003</object-id><label>Figure 1.</label><caption><title>Line estimation of heritability in nuclear and organellar genomes.</title><p>We compared the estimated metabolite heritability’s due to nuclear (solid line) and organelle (dashed line) variation across the lines of the Kas × Tsu (black) RIL populations. Shown are frequency plots of heritability and for each class, the bin size is 5% for the frequency plots.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.003">http://dx.doi.org/10.7554/eLife.00776.003</ext-link></p><p><supplementary-material id="SD1-data"><object-id pub-id-type="doi">10.7554/eLife.00776.004</object-id><label>Figure 1—source data 1.</label><caption><title>Heritability.</title><p>Results of the ANOVA model (Model I) for metabolic variance in the experiment are shown by metabolite.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.004">http://dx.doi.org/10.7554/eLife.00776.004</ext-link></p></caption><media mime-subtype="xlsx" mimetype="application" xlink:href="elife00776s001.xlsx"/></supplementary-material></p><p><supplementary-material id="SD2-data"><object-id pub-id-type="doi">10.7554/eLife.00776.005</object-id><label>Figure 1—source data 2.</label><caption><title>Means.</title><p>All metabolic mean accumulation is shown by line and subpopulation.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.005">http://dx.doi.org/10.7554/eLife.00776.005</ext-link></p></caption><media mime-subtype="xlsx" mimetype="application" xlink:href="elife00776s002.xlsx"/></supplementary-material></p></caption><graphic xlink:href="elife00776f001"/></fig></p></sec><sec id="s2-2"><title>Single locus analysis of nuclear variation and the plant metabolome</title><p>We next moved beyond the classical line heritability approach and directly test if cytoplasmic variation may have similar effects to individual nuclear loci. We mapped QTLs in the Kas × Tsu RILs controlling the mean accumulation of all 2435 measured metabolites (215 known and 2220 unidentified compounds) using 1069 markers with a median spacing of 0.35 cM (<xref ref-type="supplementary-material" rid="SD2-data">Figure 1—source data 2</xref>) (<xref ref-type="bibr" rid="bib31">Jansen, 1994</xref>; <xref ref-type="bibr" rid="bib79">Zeng et al., 1999</xref>; <xref ref-type="bibr" rid="bib8">Broman et al., 2003</xref>). This identified 2974 QTLs effecting the accumulation of 1822 metabolites with an average of 1.22 ± 0.02 QTLs per metabolite (<xref ref-type="supplementary-material" rid="SD3-data">Figure 2—source data 1</xref>). These QTLs predominantly partitioned into 14 metabolomic QTL hotspots that were unequally distributed across the genome (<xref ref-type="fig" rid="fig2">Figure 2</xref>). Chromosome IV had five detectable hotspots while chromosomes I and II had only a single significant hotspot (<xref ref-type="fig" rid="fig2">Figure 2</xref>). The hotspots had an average LOD interval size of approximately 5 cM that on average contains a similar number of genes as found within the combined mitochondrial and plastidic genomes.<fig id="fig2" position="float"><object-id pub-id-type="doi">10.7554/eLife.00776.006</object-id><label>Figure 2.</label><caption><title>Genetic architecture of metabolite QTLs across the Kas × Tsu genome.</title><p>(<bold>A</bold>) The number of metabolites for which a QTL was detected within a 5 cM sliding window is plotted against the genetic location of the metabolite QTLs in cM. The permuted threshold (p=0.05) for detection of a significant metabolite hotspot is 54 QTLs. The graph is scaled to match part (<bold>B</bold>). Hotspots are labeled above the respective locus with the chromosome and cM. (<bold>B</bold>) Heat map showing the location and effect of significant QTLs detected for average metabolite accumulation across the five chromosomes. Red indicates a positive effect of the Kas allele, while green indicates a positive effect of the Tsu allele. Vertical white lines separate the chromosomes (I to V from left to right). Clustering on the left is based on the absolute Pearson correlation of QTL effects across all significant loci for each metabolite. Only metabolites with two or more QTLs were plotted.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.006">http://dx.doi.org/10.7554/eLife.00776.006</ext-link></p><p><supplementary-material id="SD3-data"><object-id pub-id-type="doi">10.7554/eLife.00776.007</object-id><label>Figure 2—source data 1.</label><caption><title>QTL Lists.</title><p>The location and effect of all metabolic QTLs identified from the CIM model are presented.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.007">http://dx.doi.org/10.7554/eLife.00776.007</ext-link></p></caption><media mime-subtype="xlsx" mimetype="application" xlink:href="elife00776s003.xlsx"/></supplementary-material></p></caption><graphic xlink:href="elife00776f002"/></fig></p><p>To compare nuclear and cytoplasmic genetic variation effects upon metabolism, we utilized these metabolomic QTL hotspots to develop an additive model. This model uses the genetic marker at the center of each metabolomic QTL hotspot as separate terms and also incorporates the cytoplasmic variation as an additional term equivalent to each nuclear locus. We used this additive model to directly test all QTL hotspot-metabolite linkages and obtain the mean effect of variation at each hotspot on the metabolome (<xref ref-type="supplementary-material" rid="SD4-data">Figure 3—source data 1</xref>, <xref ref-type="fig" rid="fig3s1 fig3s2 fig3s3 fig3s4 fig3s5 fig3s6 fig3s7 fig3s8 fig3s9 fig3s10 fig3s11 fig3s12">Figure 3—Figure supplements 1–12</xref>). For all hotspots, the primary metabolites altered by variation at each hotspot were distributed across the primary metabolism pathways with no obvious strong qualitative network specific loci as previously found (<xref ref-type="fig" rid="fig3">Figure 3</xref>) (<xref ref-type="bibr" rid="bib60">Rowe et al., 2008</xref>).<fig-group><fig id="fig3" position="float"><object-id pub-id-type="doi">10.7554/eLife.00776.008</object-id><label>Figure 3.</label><caption><title>Metabolomic consequence of variation at nuclear loci.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at nuclear loci. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the cytoplasmic genome and gray boxes are metabolites that were not detected. The two loci shown are those that had the most metabolites affected within the metabolic map. All other nuclear loci are plotted in <xref ref-type="fig" rid="fig3s1 fig3s10 fig3s11 fig3s12 fig3s2 fig3s3 fig3s4 fig3s5 fig3s6 fig3s7 fig3s8 fig3s9">Figure 3—figure supplements 1–12</xref>. (<bold>A</bold>) Estimated allele effects of the M.V.59 hotspot. (<bold>B</bold>) Estimated allele effects of the M.I.83 hotspot.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.008">http://dx.doi.org/10.7554/eLife.00776.008</ext-link></p><p><supplementary-material id="SD4-data"><object-id pub-id-type="doi">10.7554/eLife.00776.009</object-id><label>Figure 3—source data 1.</label><caption><title>Single marker ANOVA.</title><p>p values for main effect model of metabolic QTL variance using hotspot markers</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.009">http://dx.doi.org/10.7554/eLife.00776.009</ext-link></p></caption><media mime-subtype="xlsx" mimetype="application" xlink:href="elife00776s004.xlsx"/></supplementary-material></p></caption><graphic xlink:href="elife00776f003"/></fig><fig id="fig3s1" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.010</object-id><label>Figure 3—figure supplement 1.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.II.91.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.010">http://dx.doi.org/10.7554/eLife.00776.010</ext-link></p></caption><graphic xlink:href="elife00776fs001"/></fig><fig id="fig3s2" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.011</object-id><label>Figure 3—figure supplement 2.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.III.51.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.011">http://dx.doi.org/10.7554/eLife.00776.011</ext-link></p></caption><graphic xlink:href="elife00776fs002"/></fig><fig id="fig3s3" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.012</object-id><label>Figure 3—figure supplement 3.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.III.64.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.012">http://dx.doi.org/10.7554/eLife.00776.012</ext-link></p></caption><graphic xlink:href="elife00776fs003"/></fig><fig id="fig3s4" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.013</object-id><label>Figure 3—figure supplement 4.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.IV.3.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.013">http://dx.doi.org/10.7554/eLife.00776.013</ext-link></p></caption><graphic xlink:href="elife00776fs004"/></fig><fig id="fig3s5" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.014</object-id><label>Figure 3—figure supplement 5.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.IV.17.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.014">http://dx.doi.org/10.7554/eLife.00776.014</ext-link></p></caption><graphic xlink:href="elife00776fs005"/></fig><fig id="fig3s6" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.015</object-id><label>Figure 3—figure supplement 6.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.IV.23.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.015">http://dx.doi.org/10.7554/eLife.00776.015</ext-link></p></caption><graphic xlink:href="elife00776fs006"/></fig><fig id="fig3s7" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.016</object-id><label>Figure 3—figure supplement 7.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.IV.51.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.016">http://dx.doi.org/10.7554/eLife.00776.016</ext-link></p></caption><graphic xlink:href="elife00776fs007"/></fig><fig id="fig3s8" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.017</object-id><label>Figure 3—figure supplement 8.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.IV.72.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.017">http://dx.doi.org/10.7554/eLife.00776.017</ext-link></p></caption><graphic xlink:href="elife00776fs008"/></fig><fig id="fig3s9" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.018</object-id><label>Figure 3—figure supplement 9.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.IV.82.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.018">http://dx.doi.org/10.7554/eLife.00776.018</ext-link></p></caption><graphic xlink:href="elife00776fs009"/></fig><fig id="fig3s10" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.019</object-id><label>Figure 3—figure supplement 10.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.V.45.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.019">http://dx.doi.org/10.7554/eLife.00776.019</ext-link></p></caption><graphic xlink:href="elife00776fs010"/></fig><fig id="fig3s11" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.020</object-id><label>Figure 3—figure supplement 11.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.V.82.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.020">http://dx.doi.org/10.7554/eLife.00776.020</ext-link></p></caption><graphic xlink:href="elife00776fs011"/></fig><fig id="fig3s12" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.021</object-id><label>Figure 3—figure supplement 12.</label><caption><title>Metabolomic consequence of variation at nuclear locus M.V.94.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation at each nuclear hotspot locus. A red box shows increased metabolite accumulation when the line contains the Kas allele at the nuclear locus while green shows increased metabolite accumulation when the line contains the Tsu allele at the nuclear locus. White boxes are metabolites that were detected but not significantly influenced by the locus and gray boxes are metabolites that were not detected. The specific hotspot metabolomic QTL locus is listed at the top of the page using the nomenclature shown in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.021">http://dx.doi.org/10.7554/eLife.00776.021</ext-link></p></caption><graphic xlink:href="elife00776fs012"/></fig></fig-group></p></sec><sec id="s2-3"><title>The comparative effects of cytoplasmic genome variation to nuclear variation</title><p>We next proceeded to use the additive model to directly compare the role of cytoplasmic genome variation to that of individual nuclear loci in controlling metabolome variation (<xref ref-type="supplementary-material" rid="SD4-data">Figure 3—source data 1</xref>). In comparison to the broader question of cytoplasmic heritability tested in the line heritability model (<xref ref-type="supplementary-material" rid="SD1-data">Figure 1—source data 1</xref>), this model allows us to ask the more specific question of how the cytoplasmic variation compares to individual nuclear loci in altering metabolite accumulation. The additive model showed that variation in the cytoplasmic genome significantly altered variation in 1755 out of the 2435 detected metabolites. This is in comparison the nuclear loci that only affected 298 metabolites on average (range 438–158). In addition to affecting more metabolites than nuclear loci, the average effect size of the cytoplasmic genome on metabolite variation was higher than that for the nuclear loci (<xref ref-type="fig" rid="fig4">Figure 4</xref>).<fig-group><fig id="fig4" position="float"><object-id pub-id-type="doi">10.7554/eLife.00776.022</object-id><label>Figure 4.</label><caption><title>Comparison of estimated QTL allele effects between nuclear and cytoplasmic genetic variation.</title><p>The distribution of percent allele effects are shown for all metabolite/loci combinations with positive being the Tsu allele increases the metabolite concentration in comparison to the Kas allele. Solid black line shows the allele effects for all nuclear genomic loci while the dashed line shows cytoplasmic genetic variation allele effects.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.022">http://dx.doi.org/10.7554/eLife.00776.022</ext-link></p></caption><graphic xlink:href="elife00776f004"/></fig><fig id="fig4s1" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.023</object-id><label>Figure 4—figure supplement 1.</label><caption><title>Main effect estimations.</title><p>Using pairwise epistasis model run for separately for all metabolites, we identified all significant main effect markers and, epistatic interaction terms. For each metabolite, we identified each significant term and estimated the genetic variance controlled by this term by taking the ratio of the Type III sums of squares (SS) to the total sums of squares (total model SS + Residual SS) for that term. In this figure, we present the distribution of the per marker and per interaction heritability for each marker/interaction based on metabolites that are significant for that marker/interaction. The two plots titled ‘main’ and ‘main trim’ that show the heritability solely attributable to the individual main effect nuclear and cytoplasmic loci. ‘main’ shows the full plot while ‘main_trim’ shows the distribution trimmed to improve the resolution on the nuclear markers. The plot entitled ‘cytonuclear’ shows the distribution of the heritability attributed to all significant interactions between the cytoplasmic genome and the listed nuclear locus. For instance, the plot of M.I.83 in this graph shows the distribution of heritability across all metabolites significant for the cytoplasm × M.I.83 cytonuclear interaction.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.023">http://dx.doi.org/10.7554/eLife.00776.023</ext-link></p></caption><graphic xlink:href="elife00776fs013"/></fig><fig id="fig4s2" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.024</object-id><label>Figure 4—figure supplement 2.</label><caption><title>Distribution of heritability of pairwise interaction of nuclear loci M.II.91 and M.I.83.</title><p>Using the pairwise epistasis model analysis separately for all metabolites, we identified all significant main effect markers, individual genetic loci, and interactions, epistatic terms. For each metabolite, we identified each significant term and estimated the genetic variance controlled by this term by taking the ratio of the Type III sums of squares (SS) to the total sums of squares (total model SS + Residual SS) for that term from Model III. In this figure, we present the distribution of the per marker and per interaction heritability for each marker/interaction for the metabolites that are significant for that marker/interaction. These plots are titled with specific nuclear locus show the distribution of heritability controlled by interactions between the titled marker and all the other loci in the model. Again this is only for interaction terms that were significant in the individual metabolite models.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.024">http://dx.doi.org/10.7554/eLife.00776.024</ext-link></p></caption><graphic xlink:href="elife00776fs014"/></fig><fig id="fig4s3" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.025</object-id><label>Figure 4—figure supplement 3.</label><caption><title>Distribution of heritability of pairwise interaction of nuclear loci M.III.64 and M.III.51.</title><p>Using the pairwise epistasis model analysis separately for all metabolites, we identified all significant main effect markers, individual genetic loci, and interactions, epistatic terms. For each metabolite, we identified each significant term and estimated the genetic variance controlled by this term by taking the ratio of the Type III sums of squares (SS) to the total sums of squares (total model SS + Residual SS) for that term from Model III. In this figure, we present the distribution of the per marker and per interaction heritability for each marker/interaction for the metabolites that are significant for that marker/interaction. These plots are titled with specific nuclear locus show the distribution of heritability controlled by interactions between the titled marker and all the other loci in the model. Again this is only for interaction terms that were significant in the individual metabolite models.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.025">http://dx.doi.org/10.7554/eLife.00776.025</ext-link></p></caption><graphic xlink:href="elife00776fs015"/></fig><fig id="fig4s4" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.026</object-id><label>Figure 4—figure supplement 4.</label><caption><title>Distribution of heritability of pairwise interaction of nuclear loci M.IV.23 and M.IV.17.</title><p>Using the pairwise epistasis model analysis separately for all metabolites, we identified all significant main effect markers, individual genetic loci, and interactions, epistatic terms. For each metabolite, we identified each significant term and estimated the genetic variance controlled by this term by taking the ratio of the Type III sums of squares (SS) to the total sums of squares (total model SS + Residual SS) for that term from Model III. In this figure, we present the distribution of the per marker and per interaction heritability for each marker/interaction for the metabolites that are significant for that marker/interaction. These plots are titled with specific nuclear locus show the distribution of heritability controlled by interactions between the titled marker and all the other loci in the model. Again this is only for interaction terms that were significant in the individual metabolite models.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.026">http://dx.doi.org/10.7554/eLife.00776.026</ext-link></p></caption><graphic xlink:href="elife00776fs016"/></fig><fig id="fig4s5" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.027</object-id><label>Figure 4—figure supplement 5.</label><caption><title>Distribution of heritability of pairwise interaction of nuclear loci M.IV.51 and M.IV.23.</title><p>Using the pairwise epistasis model analysis separately for all metabolites, we identified all significant main effect markers, individual genetic loci, and interactions, epistatic terms. For each metabolite, we identified each significant term and estimated the genetic variance controlled by this term by taking the ratio of the Type III sums of squares (SS) to the total sums of squares (total model SS + Residual SS) for that term from Model III. In this figure, we present the distribution of the per marker and per interaction heritability for each marker/interaction for the metabolites that are significant for that marker/interaction. These plots are titled with specific nuclear locus show the distribution of heritability controlled by interactions between the titled marker and all the other loci in the model. Again this is only for interaction terms that were significant in the individual metabolite models.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.027">http://dx.doi.org/10.7554/eLife.00776.027</ext-link></p></caption><graphic xlink:href="elife00776fs017"/></fig><fig id="fig4s6" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.028</object-id><label>Figure 4—figure supplement 6.</label><caption><title>Distribution of heritability of pairwise interaction of nuclear loci M.IV.82 and M.IV.72.</title><p>Using the pairwise epistasis model analysis separately for all metabolites, we identified all significant main effect markers, individual genetic loci, and interactions, epistatic terms. For each metabolite, we identified each significant term and estimated the genetic variance controlled by this term by taking the ratio of the Type III sums of squares (SS) to the total sums of squares (total model SS + Residual SS) for that term from Model III. In this figure, we present the distribution of the per marker and per interaction heritability for each marker/interaction for the metabolites that are significant for that marker/interaction. These plots are titled with specific nuclear locus show the distribution of heritability controlled by interactions between the titled marker and all the other loci in the model. Again this is only for interaction terms that were significant in the individual metabolite models.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.028">http://dx.doi.org/10.7554/eLife.00776.028</ext-link></p></caption><graphic xlink:href="elife00776fs018"/></fig><fig id="fig4s7" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.029</object-id><label>Figure 4—figure supplement 7.</label><caption><title>Distribution of heritability of pairwise interaction of nuclear loci M.V.59 and M.V.45.</title><p>Using the pairwise epistasis model analysis separately for all metabolites, we identified all significant main effect markers, individual genetic loci, and interactions, epistatic terms. For each metabolite, we identified each significant term and estimated the genetic variance controlled by this term by taking the ratio of the Type III sums of squares (SS) to the total sums of squares (total model SS + Residual SS) for that term from Model III. In this figure, we present the distribution of the per marker and per interaction heritability for each marker/interaction for the metabolites that are significant for that marker/interaction. These plots are titled with specific nuclear locus show the distribution of heritability controlled by interactions between the titled marker and all the other loci in the model. Again this is only for interaction terms that were significant in the individual metabolite models.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.029">http://dx.doi.org/10.7554/eLife.00776.029</ext-link></p></caption><graphic xlink:href="elife00776fs019"/></fig><fig id="fig4s8" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.030</object-id><label>Figure 4—figure supplement 8.</label><caption><title>Distribution of heritability of pairwise interaction of nuclear loci M.V.94 and M.V.82.</title><p>Using the pairwise epistasis model analysis separately for all metabolites, we identified all significant main effect markers, individual genetic loci, and interactions, epistatic terms. For each metabolite, we identified each significant term and estimated the genetic variance controlled by this term by taking the ratio of the Type III sums of squares (SS) to the total sums of squares (total model SS + Residual SS) for that term from Model III. In this figure, we present the distribution of the per marker and per interaction heritability for each marker/interaction for the metabolites that are significant for that marker/interaction. These plots are titled with specific nuclear locus show the distribution of heritability controlled by interactions between the titled marker and all the other loci in the model. Again this is only for interaction terms that were significant in the individual metabolite models.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.030">http://dx.doi.org/10.7554/eLife.00776.030</ext-link></p></caption><graphic xlink:href="elife00776fs020"/></fig></fig-group></p><p>Using the additive model, we generated per locus estimates of heritability that showed the cytoplasmic genome explained twice the variance as any individual nuclear locus on average (<xref ref-type="fig" rid="fig4s1 fig4s2 fig4s3 fig4s4 fig4s5 fig4s6 fig4s7 fig4s8">Figure 4—figure supplements 1–8</xref>). Interestingly, the number of genes within a 5 cM region surrounding the nuclear hotspots is comparable to the number of genes present within the combined organellar genomes, which suggests that this is not simply a matter of genetic potential. Additionally, the Kas allele of both the nuclear and cytoplasmic loci were more frequently associated with increasing metabolite concentration leading to a change in the entire metabolic network within this population based on the cytoplasmic genome (<xref ref-type="fig" rid="fig4 fig5">Figures 4 and 5</xref>). Thus, the line heritability approach to partitioning genetic variation dramatically underestimated the effect of cytoplasmic genetic variation and cytoplasmic variation has larger effects than any individual nuclear locus even though there are a similar potential number of causative genes.<fig id="fig5" position="float"><object-id pub-id-type="doi">10.7554/eLife.00776.031</object-id><label>Figure 5.</label><caption><title>Metabolomic consequences of cytoplasmic genomic variation.</title><p>A map of central metabolism was created in cytoscape and this was used to plot the estimated allele effect of genetic variation in the cytoplasmic genomes using the reciprocal sub-populations. Colors are as given in <xref ref-type="fig" rid="fig3">Figure 3</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.031">http://dx.doi.org/10.7554/eLife.00776.031</ext-link></p><p><supplementary-material id="SD5-data"><object-id pub-id-type="doi">10.7554/eLife.00776.032</object-id><label>Figure 5—source data 1.</label><caption><title>Genetic polymorphisms between the Kas and Tsu Organelles.</title><p>AGI represents the code for each gene with position being from the +1 of transcription for that gene. The specific base at each site in Col-0, Kas and Tsu are shown.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.032">http://dx.doi.org/10.7554/eLife.00776.032</ext-link></p></caption><media mime-subtype="xlsx" mimetype="application" xlink:href="elife00776s005.xlsx"/></supplementary-material></p></caption><graphic xlink:href="elife00776f005"/></fig></p></sec><sec id="s2-4"><title>Cytoplasmic polymorphisms</title><p>To identify potential causal polymorphisms between the Kas and Tsu organelles, we obtained short-read sequencing data and identified SNPs in the organellar genomes. This analysis showed that there were polymorphisms spread out across a large number of genes including regulatory genes, such as <italic>rpoC2</italic> in the plastid, several unknown genes, and key energy genes, such as Rubisco large subunit. The most striking changes were the large number of polymorphisms within the NADH deyhdrogenases in both the mitochondria and the plastid (<xref ref-type="supplementary-material" rid="SD5-data">Figure 5—source data 1</xref>). Specifically in the mitochondria, 32 of the 96 polymorphisms were within the NADH dehydrogenase 7 and a further 13 polymorphisms in four other NADH dehydrogenase complex genes. This is vastly more polymorphisms than would be expected by random chance suggesting a change in NADH and related metabolism between the Kas and Tsu organelles (hyper-geometric test, p&lt;0.001). This agrees with the widespread metabolic consequences of the cytoplasmic genetic variation on processes that require NADH like glutamine synthesis, lipid metabolism, and any other process that utilizes NADH requiring cytochromes P450. However, within this population the mitochondrial and plastidic genomes are perfectly co-inherited making it impossible to resolve the individual effects of the mitochondrial and plastidic genomes (<xref ref-type="bibr" rid="bib32">Joseph et al., 2013</xref>). Additionally, it isn’t possible to ascribe metabolites to specific subcellular compartments using a whole tissue extract as most metabolites, even those synthesized in specific compartments, often accumulate in most compartments within a plant cell (<xref ref-type="bibr" rid="bib40">Krueger et al., 2011</xref>). Thus, genetic variation in the cytoplasm can have significant effects on modulating global plant metabolism requiring further work to identify the specific mechanistic causes.</p></sec><sec id="s2-5"><title>Cytonuclear epistasis shapes the plant metabolome</title><p>Previous work on natural variation in the plant metabolome had shown extensive two-way and three-way epistasis between nuclear loci but could not test for the presence of epistatic interactions between nuclear and cytoplasmic loci (<xref ref-type="bibr" rid="bib60">Rowe et al., 2008</xref>). To test for cytonuclear epistatic interactions, we used a pairwise epistasis model that directly tested all pairwise combinations of the 14 nuclear QTL and cytoplasm against all metabolites while including all terms from the additive model (<xref ref-type="supplementary-material" rid="SD6-data">Figure 6—source data 1</xref>, <xref ref-type="supplementary-material" rid="SD7-data">2</xref>). This pairwise epistasis model only uses 120 of the 314 available degrees of freedom (38%) leaving the majority of the degrees of freedom for the residual variance, which suggests that we are not over fitting the model. We did not extend this analysis to a full genome survey of all possible loci because these surveys do not account for existing main effect loci. After multiple testing adjustments, we only plotted pairwise interactions that significantly affected at least 10% of the metabolites to present a conservative image of the interaction network (<xref ref-type="fig" rid="fig6">Figure 6</xref>).<fig id="fig6" position="float"><object-id pub-id-type="doi">10.7554/eLife.00776.033</object-id><label>Figure 6.</label><caption><title>Epistatic networks of metabolism.</title><p>All epistatic interactions between the cytoplasmic genomic variation and nuclear loci (as labeled in <xref ref-type="fig" rid="fig2">Figure 2</xref>) are plotted as lines connecting the main effect loci as nodes. The size of the main effect nodes represents the fraction of total metabolites affected by the given locus. The color of the lines show the fraction of metabolites linked with this interaction, light blue 10–15% of metabolites, dark blue 15–20% of metabolites.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.033">http://dx.doi.org/10.7554/eLife.00776.033</ext-link></p><p><supplementary-material id="SD6-data"><object-id pub-id-type="doi">10.7554/eLife.00776.034</object-id><label>Figure 6—source data 1.</label><caption><title>Pairwise ANOVA p values.</title><p>p values for main effect and interaction terms using hotspot markers.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.034">http://dx.doi.org/10.7554/eLife.00776.034</ext-link></p></caption><media mime-subtype="xlsx" mimetype="application" xlink:href="elife00776s006.xlsx"/></supplementary-material></p><p><supplementary-material id="SD7-data"><object-id pub-id-type="doi">10.7554/eLife.00776.035</object-id><label>Figure 6—source data 2.</label><caption><title>Pairwise ANOVA sums of squares.</title><p>The type III sums of squares from the model run in <xref ref-type="supplementary-material" rid="SD6-data">Figure 6—source data 1</xref>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.035">http://dx.doi.org/10.7554/eLife.00776.035</ext-link></p></caption><media mime-subtype="xlsx" mimetype="application" xlink:href="elife00776s007.xlsx"/></supplementary-material></p></caption><graphic xlink:href="elife00776f006"/></fig></p><p>Importantly, the cytoplasmic locus was one of the two nodes with the highest number of significant pairwise epistatic interactions connecting to 6 of the 14 nuclear loci. This level of interaction is higher than the average of the network degree distribution (3.5 ± 0.46 SE) (<xref ref-type="fig" rid="fig6">Figure 6</xref>). The M.I.83 nuclear locus had the highest level of interactions involving 7 of the other 13 nuclear loci but did not interact directly with the cytoplasmic variation (<xref ref-type="fig" rid="fig6">Figure 6</xref>). Thus, there is strong pairwise epistasis between the cytoplasmic and nuclear genetic loci. Further, the percent of genetic variation controlled by interactions of two nuclear loci was of similar scale to that involving a cytonuclear interaction (<xref ref-type="fig" rid="fig4s1 fig4s2 fig4s3 fig4s4 fig4s5 fig4s6 fig4s7 fig4s8">Figure 4—figure supplements 1–8</xref>).</p></sec><sec id="s2-6"><title>Cytoplasmic genetic variation has larger phenotypic effects on metabolite epistasis than the nuclear partner</title><p>The large number of significant epistatic interactions between nuclear and cytoplasmic genetic variation forms a cohesive network but the network approach does not let us visualize or directly quantify the effects of each locus in an epistatic pair. Thus, we wanted to develop a better approach to visualize and compare epistatic effects (<xref ref-type="bibr" rid="bib10">Carlborg and Haley, 2004</xref>; <xref ref-type="bibr" rid="bib11">Carlborg et al., 2006</xref>; <xref ref-type="bibr" rid="bib3">Alvarez-Castro and Carlborg, 2007</xref>). We utilized the center of mass concept from physics to generate a description of epistatic effect (<xref ref-type="bibr" rid="bib29">Gartenhaus and Schwartz, 1957</xref>). In center of mass calculations across two dimensions, each individual objects mass is described in its x,y coordinates. For describing epistasis, we converted the x,y-cartography such that the alleles for one QTL with Tsu being -1 and Kas being 1 are along the x-axis. Similarly, along the y-axis are the alleles at the second QTL thus creating a calculable genotypic cartography (<xref ref-type="fig" rid="fig7">Figure 7A</xref>). This effectively replaces the physical distance metric in the center of mass calculation with a genetic distance metric such that they represent a scaled allelic effect. We then replace the individual objects mass the in center of mass calculations with the unscaled average phenotypic value for each of the homozygous genotypic combinations of the two QTLs allowing a genetic centroid of the phenotype to be estimated for each metabolite and plotted (<xref ref-type="fig" rid="fig7">Figure 7A</xref>). This allows a single plot to compare the magnitude of epistasis and the allelic direction across multiple metabolites (Please see the materials and methods for a more detailed description of this approach and <xref ref-type="fig" rid="fig7s1">Figure 7—figure supplement 1</xref> for a representative model plot).<fig-group><fig id="fig7" position="float"><object-id pub-id-type="doi">10.7554/eLife.00776.036</object-id><label>Figure 7.</label><caption><title>Different epistatic patterns across genotypic combinations.</title><p>The center of mass calculations were used to estimate the phenotypic center for each metabolite that was significantly affected by the given combination of loci. The hexbin plots show the distribution of phenotypic centers for all significant metabolites. The number of metabolites per hexbin are shown in the legend to the right of each graph. All other significant epistatic pair combinations are plotted in <xref ref-type="fig" rid="fig7s2">Figure 7—figure supplement 2</xref>. (<bold>A</bold>) Metabolites that are additive for both the cytoplasmic loci and M.IV.23. (<bold>B</bold>) Metabolites that are epistatically affected by an interaction between M.IV.23 and the cytoplasmic genotypes. (<bold>C</bold>) Metabolites that are epistatically affected by an interaction between M.I.83 and the cytoplasmic genotypes. (<bold>D</bold>) Metabolites that are epistatically affected by an interaction between M.I.83 and M.IV.23 genotypes.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.036">http://dx.doi.org/10.7554/eLife.00776.036</ext-link></p></caption><graphic xlink:href="elife00776f007"/></fig><fig id="fig7s1" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.037</object-id><label>Figure 7—figure supplement 1.</label><caption><title>Descriptive model of the epistatic center plot.</title><p>This is a representative plot showing where the specific modeled metabolites, as presented in the table, are position on the center of phenotype plot. Letters label each modeled metabolite on the plot. The allele at each locus is positioned as the Kas allele at 1 and Tsu allele at −1.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.037">http://dx.doi.org/10.7554/eLife.00776.037</ext-link></p></caption><graphic xlink:href="elife00776fs021"/></fig><fig id="fig7s2" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.038</object-id><label>Figure 7—figure supplement 2.</label><caption><title>Hexbin plots show the distribution of phenotypic centers for cytoplasmic and nuclear pairwise epistatic interactions.</title><p>Only interactions significantly affecting 10% or more of the metabolites are shown. Each interaction is listed at the top of the page and the number of metabolite per hexbin is shown in the legend to the right of each graph. Center of mass calculations were used to estimate the phenotypic center for each metabolite that was significantly affected by the given epistasis interaction. The hexbin plots show the distribution of phenotypic centers for all significant metabolites.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.038">http://dx.doi.org/10.7554/eLife.00776.038</ext-link></p></caption><graphic xlink:href="elife00776fs022"/></fig></fig-group></p><p>To compare the role of each locus in a pairwise epistatic interaction, we plotted the phenotypic center for only the metabolites that showed significant epistasis for each pairwise interaction in our pairwise marker model analysis (<xref ref-type="fig" rid="fig7">Figure 7</xref> and <xref ref-type="fig" rid="fig7s2">Figure 7—figure supplement 2</xref>). For all cytonuclear epistatic interactions, the axis representing cytoplasmic genetic variation showed a wider phenotypic range than the axis representing variation in the nuclear locus (<xref ref-type="fig" rid="fig7">Figure 7B,C</xref> and <xref ref-type="fig" rid="fig7s2">Figure 7—figure supplement 2</xref>). Additionally, the range of phenotypic effects in epistasis between two nuclear loci was typically smaller than that between nuclear and cytoplasmic genome (<xref ref-type="fig" rid="fig7s2">Figure 7—figure supplement 2</xref>). Thus, variation within the cytoplasmic locus has a larger epistatic effect upon metabolomic variation than genetic variation in the nuclear locus further suggesting that genetic variation in the cytoplasm has larger than recognized phenotypic consequences.</p></sec><sec id="s2-7"><title>Quantifying the model improvement by including cytoplasmic variation</title><p>The above analysis showed that the cytoplasmic genetic variation plays an important role in mediating variation within the plant metabolome. To quantify how much the inclusion of the cytoplasm improved the genetic models in describing metabolomic variation, we analyzed our additive and pairwise epistasis model models with and without the cytoplasmic terms. For each metabolite for each model, we estimated the percent of variance explained by the significant terms in the model (<xref ref-type="fig" rid="fig8">Figure 8</xref>). This showed that including the cytoplasmic genome in the additive model nearly doubled the average variance compared to additive model containing only the nuclear loci. Adding the pairwise epistatic terms to the additive model without the cytoplasmic genome nearly tripled the fraction of metabolite variance explained by the genetic model without over-fitting the model (<xref ref-type="fig" rid="fig8">Figure 8</xref>). Including the cytoplasmic genome into the pairwise model shifted the average model genetic variance from 38% to 53% (<xref ref-type="fig" rid="fig8">Figure 8</xref>). Thus, the inclusion of the cytoplasmic term into the pairwise model explained as much (15%), if not more, variance than could be explained from the simple additive model built with only nuclear loci (13%) (<xref ref-type="fig" rid="fig8">Figure 8</xref>). This shows that the cytoplasmic genetic variation is critical to being able to fully describe metabolomic variation within Arabidopsis.<fig id="fig8" position="float"><object-id pub-id-type="doi">10.7554/eLife.00776.039</object-id><label>Figure 8.</label><caption><title>Distribution of estimated variance between main effects and epistatic interactions.</title><p>The cytoplasmic term was added or dropped from the different statistical models to compare the total variance explained by each model for each metabolite. Dotted lines show estimated variance using solely the main effect loci without interactions (additive model), solid lines show the distribution of estimated variance across the metabolites using the pairwise epistasis model while the dashed lines are the results for the three-way epistasis model including the most prevalent three-way interactions as indicated by the epistatic network. The gray lines show the models with only the nuclear loci while the black lines show the model with the nuclear and cytoplasmic loci. For the frequency plot, bin size is 0.025 r<sup>2</sup>.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.039">http://dx.doi.org/10.7554/eLife.00776.039</ext-link></p></caption><graphic xlink:href="elife00776f008"/></fig></p></sec><sec id="s2-8"><title>Cytonuclear epistasis alters nuclear epistatic relationships</title><p>The epistatic network diagram showed that there were several instances of triangles where two nuclear loci showing pairwise epistasis also interacted with the cytoplasmic variation (<xref ref-type="fig" rid="fig6">Figure 6</xref>). This connectivity between the cytoplasmic genetic variation and nuclear QTL epistatic pairs led us to question if genetic variation in the cytoplasmic genomes could alter epistatic interactions between nuclear loci. To test if cytoplasmic genetic variation could alter nuclear epistasis, we identified all pairs of nuclear loci that interacted with each other and the cytoplasmic genome (<xref ref-type="fig" rid="fig6">Figure 6</xref>). These 10 three-way interactions were then included in the pairwise epistatic model to generate a three-way interaction model (<xref ref-type="supplementary-material" rid="SD6-data SD7-data">Figure 6—source data 1,2</xref>). We found that including these 10 new three-way interactions significantly increased the total variance explained by the model for the distribution of metabolites while only using less than 42% of the available degrees of freedom (<xref ref-type="fig" rid="fig8">Figure 8</xref>, <italic>t</italic>-test p&lt;0.001).</p><p>To examine how cytoplasmic genetic variation influences nuclear epistasis we visualized all significant three-way interactions using the phenotypic center approach, (<xref ref-type="fig" rid="fig9s1 fig9s2">Figure 9—figure supplements 1–3</xref>). This showed that the cytoplasmic variation enhanced or even changed nuclear epistasis. Within the significant Cytoplasm × M.I.83 × M.IV.23 epistatic interaction, we identified larger effects of the nuclear loci than had been possible in the pairwise model (<xref ref-type="fig" rid="fig7 fig9">Figures 7 and 9</xref>, <xref ref-type="fig" rid="fig9s1 fig9s2">Figure 9—figure supplements 1–3</xref>). For example pyruvate and metabolite 227710 have quite large nuclear epistatic interactions that are only visible in one or the other cytoplasmic background (<xref ref-type="fig" rid="fig9">Figure 9B,C</xref>). Neither metabolite had a large nuclear epistasis due to comparisons averaging across the different cytoplasmic backgrounds (<xref ref-type="fig" rid="fig7">Figure 7</xref> vs <xref ref-type="fig" rid="fig9">Figure 9</xref>). In contrast, salicylic acid had different patterns of nuclear epistasis in the two cytoplasmic genetic backgrounds (<xref ref-type="fig" rid="fig9">Figure 9D</xref>). A similar pattern of the cytoplasmic variation altering nuclear epistasis was seen for other three-way interactions (<xref ref-type="fig" rid="fig9s4 fig9s5 fig9s6 fig9s7">Figure 9—figure supplements 4–7</xref>).<fig-group><fig id="fig9" position="float"><object-id pub-id-type="doi">10.7554/eLife.00776.040</object-id><label>Figure 9.</label><caption><title>The cytoplasmic background alters nuclear epistatic interactions.</title><p>The analysis of the M.IV.23 × M.I.83 × cytoplasmic three-way epistatic interaction is shown for all significantly affected metabolites. All other metabolite distributions for three-way epistatic combinations are plotted in <xref ref-type="fig" rid="fig9s1 fig9s2">Figure 9—figure supplements 1–3</xref>. (<bold>A</bold>) The center of mass calculations were used to estimate the phenotypic center for each metabolite that shows a significant three-way epistasis with the M.IV.23, M.I.83 and cytoplasmic genetic variation. All significant metabolites are plotted as unique points. The specific metabolites boxed and labeled show the location of the metabolites shown in parts, B, C and D respectively. For each locus, the Kas allele is plotted at 1 while Tsu is −1. (<bold>B</bold>) Effect of the M.IV.23 × M.I.83 × cytoplasmic epistasis upon the accumulation of unknown 227710. Average and standard error are shown. (<bold>C</bold>) Effect of the M.IV.23 × M.I.83 × cytoplasmic epistasis upon the accumulation of pyruvate. Average and standard error are shown. (<bold>D</bold>) Effect of the M.IV.23 × M.I.83 × cytoplasmic epistasis upon the accumulation of salicylic acid. Average and standard error are shown.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.040">http://dx.doi.org/10.7554/eLife.00776.040</ext-link></p></caption><graphic xlink:href="elife00776f009"/></fig><fig id="fig9s1" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.041</object-id><label>Figure 9—figure supplement 1.</label><caption><title>Distribution of phenotypic centers of metabolites significantly affected by M.I.83 × M.IV.3 × cytoplasmic three-way epistatic interaction.</title><p>The page shows the distribution of phenotypic centers of metabolites significantly affected by three way epistatic interactions. Of the 10 three-way interactions involving cytoplasmic loci that were tested, only the interactions showing significant epistasis for 10% or more metabolites are shown. All significant metabolites are plotted as unique points. For each locus, the Kas allele is plotted at 1 while Tsu is −1.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.041">http://dx.doi.org/10.7554/eLife.00776.041</ext-link></p></caption><graphic xlink:href="elife00776fs023"/></fig><fig id="fig9s2" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.042</object-id><label>Figure 9—figure supplement 2.</label><caption><title>Distribution of phenotypic centers of metabolites significantly affected by M.I.83 × M.IV.17 × cytoplasmic three-way epistatic interaction.</title><p>The page shows the distribution of phenotypic centers of metabolites significantly affected by three-way epistatic interactions. Of the 10 three-way interactions involving cytoplasmic loci that were tested, only the interactions showing significant epistasis for 10% or more metabolites are shown. All significant metabolites are plotted as unique points. For each locus, the Kas allele is plotted at 1 while Tsu is −1.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.042">http://dx.doi.org/10.7554/eLife.00776.042</ext-link></p></caption><graphic xlink:href="elife00776fs024"/></fig><fig id="fig9s3" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.043</object-id><label>Figure 9—figure supplement 3.</label><caption><title>Distribution of phenotypic centers of metabolites significantly affected by M.I.83 × M.V.82 × cytoplasmic three-way epistatic interaction.</title><p>For all metabolites that were epistatically affected by M.I.83 × M.IV.23 × cytoplasmic three-way interaction, hexbin plot showing change in the distribution of phenotypic centers corresponding interaction of nuclear loci M.I.83 × M.IV.23 as compared to the distribution shown in <xref ref-type="fig" rid="fig7">Figure 7D</xref>. The number of metabolite per hexbin is shown in the legend to the right of the graph. Center of mass calculations were used to estimate the phenotypic center for each metabolite that was significantly affected by the given interaction. The hexbin plots show the distribution of phenotypic centers for all significant metabolites.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.043">http://dx.doi.org/10.7554/eLife.00776.043</ext-link></p></caption><graphic xlink:href="elife00776fs025"/></fig><fig id="fig9s4" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.044</object-id><label>Figure 9—figure supplement 4.</label><caption><title>Distribution of phenotypic centers of metabolites significantly affected by M.I.83 × M.IV.23 × cytoplasmic three-way epistatic interaction.</title><p>For all metabolites that were epistatically affected by M.I.83 × M.IV.23 × cytoplasmic three-way interaction, hexbin plot showing change in the distribution of phenotypic centers corresponding interaction of nuclear loci M.I.83 × M.IV.23 as compared to the distribution shown in <xref ref-type="fig" rid="fig7">Figure 7D</xref>. The number of metabolite per hexbin is shown in the legend to the right of the graph. Center of mass calculations were used to estimate the phenotypic center for each metabolite that was significantly affected by the given interaction. The hexbin plots show the distribution of phenotypic centers for all significant metabolites.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.044">http://dx.doi.org/10.7554/eLife.00776.044</ext-link></p></caption><graphic xlink:href="elife00776fs026"/></fig><fig id="fig9s5" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.045</object-id><label>Figure 9—figure supplement 5.</label><caption><title>Movement of phenotypic centers of metabolites significantly affected by M.I.83 × M.IV.23 × cytoplasmic three-way epistatic interaction.</title><p>For metabolites epistatically affected by M.I.83 × M.IV.23 × cytoplasmic three-way interaction, we plotted the center of phenotypes corresponding to interaction of nuclear loci M.I.83 × M.IV.23 as estimated using the three-way epistasis and pairwise epistasis models. For each metabolite, open diamonds indicate the center of phenotype estimated using the pairwise epistasis model while closed diamonds indicate the estimated location using the three-way epistasis model. For each metabolite the pairwise and three-way center of phenotypes are connected by a line to indicate the change in location.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.045">http://dx.doi.org/10.7554/eLife.00776.045</ext-link></p></caption><graphic xlink:href="elife00776fs027"/></fig><fig id="fig9s6" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.046</object-id><label>Figure 9—figure supplement 6.</label><caption><title>Long range movement of phenotypic centers of metabolites significantly affected by M.I.83 × M.IV.23 × cytoplasmic three-way epistatic interaction.</title><p>For metabolites epistatically affected by M.I.83 × M.IV.23 × cytoplasmic three-way interaction, we plotted the center of phenotypes corresponding to interaction of nuclear loci M.I.83 × M.IV.23 as estimated using the three-way epistasis and pairwise epistasis models. For each metabolite, open diamonds indicate the center of phenotype estimated using the pairwise model while closed diamonds indicate the estimated location using the three-way model. For each metabolite the pairwise and three-way center of phenotypes are connected by a line to indicate the change in location. Only those metabolites whose centers of phenotypes are located farther than 0.1 from the origin are provided to increase visual resolution.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.046">http://dx.doi.org/10.7554/eLife.00776.046</ext-link></p></caption><graphic xlink:href="elife00776fs028"/></fig><fig id="fig9s7" position="float" specific-use="child-fig"><object-id pub-id-type="doi">10.7554/eLife.00776.047</object-id><label>Figure 9—figure supplement 7.</label><caption><title>Distribution of phenotypic center of metabolites significantly affected by M.I.83 × M.IV.23 × cytoplasmic three-way epistatic interaction.</title><p>For metabolites epistatically affected by M.I.83 × M.IV.23 × cytoplasmic three-way interaction, we measured the change in distance from the (1,1) point to test if the inclusion of the cytoplasmic genotype has a uniform shift upon the phenotype of the affected metabolites. The histogram with the gray dashed border shows the frequency distribution of metabolites that decreased the distance from the point (1,1) when comparing the three-way to pairwise position. The histogram with black solid border shows the frequency distribution of metabolites that increased the distance. The respective mean and median for each class are shown in the respective lines.</p><p><bold>DOI:</bold> <ext-link ext-link-type="doi" xlink:href="10.7554/eLife.00776.047">http://dx.doi.org/10.7554/eLife.00776.047</ext-link></p></caption><graphic xlink:href="elife00776fs029"/></fig></fig-group></p></sec></sec><sec id="s3" sec-type="discussion"><title>Discussion</title><sec id="s3-1"><title>Cytoplasmic genetic variation has a significant contribution to metabolic variation</title><p>Even though maternal contribution to complex phenotypes has long been suspected, most previous genomic surveys of natural variation using transcriptomics or metabolomics have either not had the capacity to directly assess the influence of cytoplasmic genetic variation or just ignore it completely. Thus, we used the reciprocal Kas × Tsu Arabidopsis RIL population to directly quantify the role of cytoplasmic genetic variation in quantitative variation of metabolomics traits. Our analysis showed that the cytoplasmic genome variation affected the phenotypic variation for over 80% of metabolites. The affected metabolites included key parts of central metabolism and some of the most specialized metabolites (<xref ref-type="fig" rid="fig1 fig5 fig6">Figures 1, 5 and 6</xref>). Interestingly, the combined variance of the nuclear genome was larger than the variance due to cytoplasmic genome (<xref ref-type="fig" rid="fig1">Figure 1</xref>). In contrast, cytoplasmic genetic variation affected more metabolites than any of the individual nuclear loci typically with larger effects (<xref ref-type="fig" rid="fig3 fig4 fig5">Figures 3–5</xref> and <xref ref-type="fig" rid="fig4s1 fig4s2 fig4s3 fig4s4 fig4s5">Figure 4—figure supplements 1–5</xref>). This is true even though the average nuclear locus spanned approximately the same number of genes as the organellar genomes combined.</p><p>We also identified a high level of significant epistatic interaction between cytoplasmic genetic variation and nuclear genetic variation, as cytonuclear epistasis (<xref ref-type="bibr" rid="bib22">Etterson et al., 2007</xref>; <xref ref-type="bibr" rid="bib67">Tang et al., 2007</xref>; <xref ref-type="bibr" rid="bib77">Wolf, 2009</xref>). This cytonuclear epistasis explained as much, if not more, metabolic variation than the combined additive effects of the nuclear loci. In addition, the cytoplasmic genome controlled larger phenotypic changes than the nuclear locus within a cytonuclear interaction (<xref ref-type="fig" rid="fig7">Figure 7</xref>). This interaction between the cytoplasmic and nuclear loci was further extended to three-way interactions and showed that the cytoplasmic background could hide or alter the interaction between two nuclear loci (<xref ref-type="fig" rid="fig9">Figure 9</xref>). Thus, the cytoplasmic genetic background plays a key role in determining how natural variation within nuclear loci will function.</p></sec><sec id="s3-2"><title>What is the mechanistic basis of cytoplasmic genetic variation?</title><p>The effect of the cytoplasmic genetic variation was spread out across nearly all of primary metabolism making it impossible to ascribe a causal link with solely plastidic or mitochondrial genes (<xref ref-type="fig" rid="fig5">Figure 5</xref>). Similarly analysis of genomic variation in the two organelles found a large number of SNPs that could be affecting numerous genes within the organelles with an observed bias towards genes involved in the NADH dehydrogenase complex (<xref ref-type="supplementary-material" rid="SD5-data">Figure 5—source data 1</xref>). Thus, organellar genomes could be playing a role in the cytoplasmic genetic influence on metabolite variation, but given their coinheritance, it is not possible to resolve the direct causal polymorphisms. Separating the two genomes to resolve their relative roles would require identifying rare individuals where the two organelles show bi-parental inheritance instead of the uni-parental/maternal mode of inheritance typical for Arabidopsis (<xref ref-type="bibr" rid="bib5">Azhagiri and Maliga, 2007</xref>).</p><p>The plant cytoplasmic genomes contain only about 1% of the number of genes as found within the nuclear genome (<xref ref-type="bibr" rid="bib4">Arabidopsis genome initiative, 2000</xref>). Yet genetic variation in this small fraction of genes has a large consequence upon the plants metabolic variation (<xref ref-type="fig" rid="fig5">Figure 5</xref>). This could be from genetic polymorphisms in organellar genes that are central to plant metabolism, such as those central to photosynthesis or NADH synthesis (<xref ref-type="supplementary-material" rid="SD5-data">Figure 5—source data 1</xref>). Alternatively, the polymorphisms could be within the organellar genes that are required to facilitate the function of the several thousand nuclear genes whose protein products are transported into the organelles and function there (<xref ref-type="bibr" rid="bib1">Ajjawi et al., 2010</xref>). Finally, it is possible that the polymorphisms in the organelle impact retrograde signaling pathways, thus altering the function of the nuclear regulatory mechanisms (<xref ref-type="bibr" rid="bib71">Vinti et al., 2000</xref>; <xref ref-type="bibr" rid="bib43">Larkin et al., 2003</xref>; <xref ref-type="bibr" rid="bib21">Estavillo et al., 2011</xref>; <xref ref-type="bibr" rid="bib78">Xiao et al., 2012</xref>). Identifying the specific causal polymorphisms within the cytoplasmic genomes will require direct whole genome sequencing of the organellar genomes from all available genotypes to query the extent of natural variation in the organellar genome (<xref ref-type="bibr" rid="bib9">Cao et al., 2011</xref>). An alternative may be to identify the causal genes underlying the nuclear loci involved in three-way interactions with the cytoplasm to triangulate the identity of the cytoplasmic gene.</p></sec><sec id="s3-3" sec-type="conclusion"><title>Conclusion</title><p>In this work, we report a first genomic survey of how genetic variation within cytoplasmic genomes influences metabolomic variation. Cytoplasmic genetic variation alters metabolite variation with effects that are equivalent to, if not greater than, individual nuclear loci. More importantly, the cytoplasmic background significantly influences the ability to detect epistasis between nuclear loci. Thus cytoplasmic genomes should be included in any future analysis of natural variation, either by being included as a genotype in GWA studies or by designing future populations as reciprocal populations to allow for direct analysis of the cytoplasmic genomic variation in controlling the phenotype. This inclusion will allow direct assessment of how cytoplasmic genomic variation influences other phenotypic classes, such as transcriptomics or broader physiological phenotypes. Natural genetic variation in the organellar genomes while frequently ignored will have to be kept at the front of future experimental approaches designed to understand the evolution and genetic architecture of organismal phenotypes.</p></sec></sec><sec id="s4" sec-type="materials|methods"><title>Materials and methods</title><sec id="s4-1"><title>Growth of the Kas × Tsu RIL population</title><p>Seeds of the 341 lines of the Kas × Tsu recombinant inbred population were obtained from the Arabidopsis Biological Resource Center (ABRC, Columbus OH, USA) (<xref ref-type="bibr" rid="bib34">Juenger et al., 2006</xref>, <xref ref-type="bibr" rid="bib33">2010</xref>; <xref ref-type="bibr" rid="bib48">McKay et al., 2008</xref>). We grew a total of four to five plants per line, split into two randomized complete blocks per experiment with two independent experiments separated by approximately 3 months. This provides four independent replicates per genotype. We grew plants in large planting trays with 156 individual wells (b × w × h: 30 × 25 × 100 mm), filled with standard potting soil (Sunshine Mix #1, Sun Gro Horticulture, Bellevue WA). Prior to sowing, we imbibed seeds in distilled water and cold stratified them at 4°C for 4 days. We placed approximately 3–5 seeds of a single genotype in the center of a well and covered the trays with a transparent plastic hood to retain humidity during germination. Plants from both reciprocal subpopulations where intermixed in the block design to allow direct statistical testing of cytoplasmic effects. After 1 week, we removed the transparent hoods and surplus plants to leave one seedling per well. We watered plants twice a week with nutrient-enriched water (0.5% N-P-K fertilizer in a 2-1-2 ratio, Grow More 4-18-38, Grow More Inc., Gardena CA) and kept them in a climate-controlled chamber at 22°C and a day/night cycle of 10 hr/14 hr. These plants and experiment were the same as previously described (<xref ref-type="bibr" rid="bib32">Joseph et al., 2013</xref>).</p></sec><sec id="s4-2"><title>Metabolomics analysis</title><p>31 days after sowing, we harvested plants for metabolomics analysis. One leaf from the first fully mature adult leaf pair of each plant was removed and ground in extraction solution as previously described (<xref ref-type="bibr" rid="bib60">Rowe et al., 2008</xref>; <xref ref-type="bibr" rid="bib13">Chan et al., 2010</xref>). Metabolite identity was determined by comparing retention time and mass to the 2007 UC Davis Genome Center Metabolomics Facility metabolites database (<ext-link ext-link-type="uri" xlink:href="http://fiehnlab.ucdavis.edu/Metabolite-Library-2007">http://fiehnlab.ucdavis.edu/Metabolite-Library-2007</ext-link>; <xref ref-type="bibr" rid="bib26">Fiehn et al., 2005</xref>). Mixed samples were run approximately every 20 samples to optimize the peak identification and quantification algorithms and to control for variation in the detection as previously described (<xref ref-type="bibr" rid="bib27">Fiehn et al., 2008</xref>; <xref ref-type="bibr" rid="bib23">Fernie et al., 2011</xref>). The ion count values were used as a surrogate for metabolite abundance. Metabolite abundance was median normalized prior to analysis to account for any technical variation between samples. A separate leaf was extracted for glucosinolates and analyzed by HPLC according to previously described methods with the results reported elsewhere (<xref ref-type="bibr" rid="bib35">Kliebenstein et al., 2001a</xref>, <xref ref-type="bibr" rid="bib36">b</xref>). From the analysis of glucosinolates, a set of 10 lines with aberrant or genetically impossible glucosinolate profiles based on the known parentage were removed from the analysis, leaving a total 316 lines.</p></sec><sec id="s4-3"><title>Estimation of heritability</title><p>All RIL lines were represented in every block in both experiments creating a perfectly balanced randomized complete block design. All phenotypic data was used to calculate estimates of broad-sense heritability (H) for each phenotype as H = σ<sup>2</sup><sub>g</sub>/σ<sup>2</sup><sub>p</sub> , where σ<sup>2</sup><sub>g</sub> was estimated for both the RIL genotypes and cytoplasmic genotypes and σ<sup>2</sup><sub>p</sub> was the total phenotypic variance for a trait (<xref ref-type="bibr" rid="bib46">Liu, 1998</xref>). The ANOVA model (Line heritability Model) for each metabolite phenotype in each line (y<sub>gmeb</sub>) was: <inline-formula><mml:math id="inf1"><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mi mathvariant="italic">gceb</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mo> </mml:mo><mml:mi>μ</mml:mi><mml:mo>+</mml:mo><mml:msub><mml:mi>C</mml:mi><mml:mi>c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>G</mml:mi><mml:mi>g</mml:mi></mml:msub><mml:mrow><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>C</mml:mi><mml:mi>c</mml:mi></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow></mml:mrow><mml:mo>+</mml:mo><mml:msub><mml:mi>E</mml:mi><mml:mi>e</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>B</mml:mi><mml:mi>b</mml:mi></mml:msub><mml:mrow><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>E</mml:mi><mml:mi>e</mml:mi></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow></mml:mrow><mml:mo>+</mml:mo><mml:msub><mml:mi>C</mml:mi><mml:mi>c</mml:mi></mml:msub><mml:mo>×</mml:mo><mml:mo> </mml:mo><mml:msub><mml:mi>E</mml:mi><mml:mi>e</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mo> </mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mi mathvariant="italic">gceb</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, where <italic>c</italic> = the Kas or Tsu cytoplasm; <italic>g</italic> = the 1…316 for the 316 RILs, e = experiment 1 or 2 and b = block 1…8 nested within experiment. This allowed cytoplasmic effects to be directly tested in the C term and each RIL genotype (G) nested within the appropriate cytoplasmic class, either Kas or Tsu. Experiment and block nested within experiment were treated as random terms within the model to better parse the variation. All resulting variance estimates, p values and heritability terms are presented (<xref ref-type="supplementary-material" rid="SD1-data">Figure 1—source data 1</xref>). σ<sup>2</sup><sub>g</sub> for RIL was pulled from the <italic>G</italic><sub><italic>g</italic></sub><italic>(C</italic><sub><italic>c</italic></sub><italic>)</italic> term while σ<sup>2</sup><sub>g</sub> for cytoplasmic variation was pulled directly from the <italic>C</italic><sub><italic>c</italic></sub><italic>M</italic><sub><italic>m</italic></sub> term. We used mean values for the RILs for further analysis as we had a randomized complete block design with no missing lines. Additionally, means and LSmeans were correlated with an average r<sup>2</sup> of 0.96 for the 159 known metabolites present in both experiments.</p></sec><sec id="s4-4"><title>QTL analysis</title><p>We the previously reported genetic map for these lines of the Kas × Tsu RIL population (<xref ref-type="bibr" rid="bib48">McKay et al., 2008</xref>; <xref ref-type="bibr" rid="bib32">Joseph et al., 2013</xref>). To detect metabolite QTLs, we used the average phenotype per RIL across all experiments (<xref ref-type="supplementary-material" rid="SD2-data">Figure 1—source data 2</xref>) (<xref ref-type="bibr" rid="bib6">Basten et al., 1999</xref>; <xref ref-type="bibr" rid="bib79">Zeng et al., 1999</xref>; <xref ref-type="bibr" rid="bib74">Wang et al., 2006</xref>). For QTL detection, composite interval mapping (CIM) was implemented using cim function in R/qtl package with a 10 cM window. Forward regression was used to identify three cofactors per trait. The declaration of statistically significant QTLs was based on permutation-derived empirical thresholds using 1000 permutations for each mapped trait. QTLs with a LOD score above 2 were considered significant for further analysis (<xref ref-type="bibr" rid="bib14">Churchill and Doerge, 1994</xref>; <xref ref-type="bibr" rid="bib17">Doerge and Churchill, 1996</xref>). Composite interval mapping to assign significance based on the underlying trait distribution is robust at handling normal or near normal trait distributions (<xref ref-type="bibr" rid="bib56">Rebai, 1997</xref>), as found for most of our phenotypes. The define.peak function implemented in R/eqtl package was used to identify the peak location and one-LOD interval of each significant QTL for each trait (<xref ref-type="bibr" rid="bib74">Wang et al., 2006</xref>). The effectscan function in R/qtl package was used to estimate the QTL additive effect (<xref ref-type="bibr" rid="bib55">R Development Core Team, 2012</xref>). Allelic effects for each significant QTL are presented as percent effect, by estimating <inline-formula><mml:math id="inf2"><mml:mrow><mml:mrow><mml:mrow><mml:mo>[</mml:mo><mml:mrow><mml:msub><mml:mrow><mml:mover accent="true"><mml:mi>x</mml:mi><mml:mo>¯</mml:mo></mml:mover></mml:mrow><mml:mi mathvariant="italic">Tsu</mml:mi></mml:msub><mml:mo>−</mml:mo><mml:msub><mml:mrow><mml:mover accent="true"><mml:mi>x</mml:mi><mml:mo>¯</mml:mo></mml:mover></mml:mrow><mml:mi mathvariant="italic">Kas</mml:mi></mml:msub></mml:mrow><mml:mo>]</mml:mo></mml:mrow></mml:mrow><mml:mo>/</mml:mo><mml:msub><mml:mrow><mml:mover accent="true"><mml:mi>x</mml:mi><mml:mo>¯</mml:mo></mml:mover></mml:mrow><mml:mi mathvariant="italic">RIL</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> for each significant main effect marker.</p><p>QTL clusters were identified using a QTL summation approach where the position of each QTL for each trait was plotted on the chromosome by placing a 1 at the peak of the QTL. This was then used to sum the number of traits that had a detected QTL at a given position using a 5 cM sliding window across the genome (<xref ref-type="bibr" rid="bib38">Kliebenstein et al., 2006</xref>). The QTL clusters identified defined genetic positions that were named respective to their phenotypic class and genetic positions with a prefix indicating the phenotype followed by the chromosome number and the cM position. For example, M.I.83 indicates a metabolomics QTL hotspot on chromosome I at 83 cM. The QTLs detected at the previously characterized and cloned glucosinolate AOP locus lies underneath the M.IV.17 metabolomics hotspot (<xref ref-type="bibr" rid="bib47">Magrath et al., 1994</xref>; <xref ref-type="bibr" rid="bib37">Kliebenstein et al., 2001c</xref>, <xref ref-type="bibr" rid="bib35">a</xref>; <xref ref-type="bibr" rid="bib39">Kroymann et al., 2003</xref>).</p></sec><sec id="s4-5"><title>Additive ANOVA model</title><p>To directly test the additive effect of each identified QTL cluster we used an ANOVA model containing the markers most closely associated with each of the significant QTL clusters as individual main effect terms. For each metabolite the average accumulation in lines of genotype <italic>g</italic> at marker <italic>m</italic> was shown as <italic>y</italic><sub><italic>gm</italic></sub>. The model (Additive Model) for each metabolite in each line (y<sub>gm</sub>) was: <inline-formula><mml:math id="inf3"><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>g</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mo> </mml:mo><mml:mi>μ</mml:mi><mml:mo>+</mml:mo><mml:mo> </mml:mo><mml:munderover><mml:mstyle displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>g</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mn>2</mml:mn></mml:munderover><mml:munderover><mml:mstyle displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>m</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>m</mml:mi></mml:munderover><mml:msub><mml:mi>M</mml:mi><mml:mrow><mml:mi>g</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:mo> </mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>g</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula>, where <italic>g</italic> = Kas(1) or Tsu(2); <italic>m</italic> = 1, …,14. The main effect of the markers was denoted as <italic>M</italic> involving 15 markers (m). The cytoplasmic genome was included as an additional marker to test for cytoplasmic genome effects. We tested all metabolites with the appropriate model using lm function implemented in the R/car package, which returned all p values, Type III sums-of-squares for the complete model and each main effect. QTL main-effect estimates (in terms of allelic substitution values) were estimated for each marker (<xref ref-type="bibr" rid="bib28">Fox and Weisberg, 2011</xref>; <xref ref-type="bibr" rid="bib55">R Development Core Team, 2012</xref>). There is no significant single marker or pairwise segregation distortion in this population indicating that the model is balanced for all markers (<xref ref-type="bibr" rid="bib48">McKay et al., 2008</xref>).</p></sec><sec id="s4-6"><title>QTL epistasis analysis</title><p>To test directly for epistatic interactions between the detected QTLs, we conducted an ANOVA using the pairwise epistasis model. We used this pairwise epistasis model per metabolite because we had previous evidence that RIL populations have a significant false negative QTL detection issue and wanted to be inclusive of all possible significant loci (<xref ref-type="bibr" rid="bib12">Chan et al., 2011</xref>). Within the model, we tested all possible pairwise interactions between the markers. For each phenotype, the average value in the RILs of genotype <italic>g</italic> at marker <italic>m</italic> was shown as <italic>y</italic><sub><italic>gm</italic></sub>. The model (Pairwise epistasis model) for each metabolite in each line (y<sub>gm</sub>) was: <inline-formula><mml:math id="inf4"><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>g</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mo> </mml:mo><mml:mi>μ</mml:mi><mml:mo>+</mml:mo><mml:mo> </mml:mo><mml:munderover><mml:mstyle displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>g</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mn>2</mml:mn></mml:munderover><mml:munderover><mml:mstyle displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>m</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>m</mml:mi></mml:munderover><mml:msub><mml:mi>M</mml:mi><mml:mrow><mml:mi>g</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:mo> </mml:mo><mml:munderover><mml:mstyle displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>g</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mn>2</mml:mn></mml:munderover><mml:munderover><mml:mstyle displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>m</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>m</mml:mi></mml:munderover><mml:munderover><mml:mstyle displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>n</mml:mi><mml:mo>=</mml:mo><mml:mi>m</mml:mi><mml:mo>+</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>m</mml:mi></mml:munderover><mml:msub><mml:mi>M</mml:mi><mml:mrow><mml:mi>g</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub><mml:msub><mml:mi>M</mml:mi><mml:mrow><mml:mi>g</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:mo> </mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>g</mml:mi><mml:mi>m</mml:mi><mml:mi>n</mml:mi></mml:mrow></mml:msub><mml:mo> </mml:mo></mml:mrow></mml:math></inline-formula> where <italic>g</italic> = Kas(1) or Tsu(2); <italic>m</italic> = 1, …,14 and n was the identity of the second marker for an interaction. The main effect of the markers was denoted as <italic>M</italic> having a model involving 15 markers. The cytoplasmic genome was included as an additional single-locus marker to test for interactions between the cytoplasmic and nuclear genomes. p values, Type III sums-of-squares for the complete model and each individual term and QTL pairwise-effect estimates in terms of allelic substitution values were obtained as described for additive model ANOVA (<xref ref-type="bibr" rid="bib28">Fox and Weisberg, 2011</xref>; <xref ref-type="bibr" rid="bib55">R Development Core Team, 2012</xref>). Significance values were corrected for multiple testing within a model using FDR (&lt;0.05). The main effect and epistatic interactions of the loci were visualized using cytoscape.v2.8.3 with interactions significant for less than 10% of the phenotypes were excluded from the network analysis (<xref ref-type="bibr" rid="bib60">Rowe et al., 2008</xref>; <xref ref-type="bibr" rid="bib66">Smoot et al., 2011</xref>). The 10% threshold was chosen as an additional multiple testing correction to provide a more conservative image of the network. There are no pairwise locus segregation distortions within this population showing that the genotypes in this analysis are balanced (<xref ref-type="bibr" rid="bib48">McKay et al., 2008</xref>). The same style of model was run to test for specific three-way interactions by including specific three-way terms as indicated (three-way epistasis model).</p></sec><sec id="s4-7"><title>Plotting the epistatic center of phenotype</title><p>To plot the epistatic effect of QTLs upon a set of metabolites, we utilized the center of mass calculations. To do this we transitioned the physical distance metric in center of mass to a genetic distance metric for mapping the center of phenotype. In this, we classify one locus of each interaction as being on the x-axis and the other locus being positioned on the y-axis. On each axis, the allelic value of each specific genetic locus is plotted in relation to the heterozygote. For instance, in a two QTL situation the x-axis would be the alleles of QTL1 with the Kas having an allelic value of 1 and the Tsu having an allelic value of −1 while the other axis would be the same for QTL2 there by positioning each of the four homozygous genotypic class in one of the four quadrants. For each metabolite, the center of phenotype was calculated using center of mass calculations as <inline-formula><mml:math id="inf5"><mml:mrow><mml:msub><mml:mi>x</mml:mi><mml:mi mathvariant="italic">geno</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mfrac><mml:mrow><mml:msubsup><mml:mstyle displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>N</mml:mi></mml:msubsup><mml:msub><mml:mi>p</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:msub><mml:mi>x</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow><mml:mrow><mml:msubsup><mml:mstyle displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>N</mml:mi></mml:msubsup><mml:msub><mml:mi>p</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="inf6"><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mi mathvariant="italic">geno</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mfrac><mml:mrow><mml:msubsup><mml:mstyle displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>N</mml:mi></mml:msubsup><mml:msub><mml:mi>p</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:msub><mml:mi>y</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow><mml:mrow><mml:msubsup><mml:mstyle displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>N</mml:mi></mml:msubsup><mml:msub><mml:mi>p</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mrow></mml:math></inline-formula> where <italic>p</italic> is the average un-scaled phenotypic value for each of the four homozygous genotypic class and <italic>x</italic> is the x-coordinate of the corresponding genotypic class and <italic>y</italic> is the y-coordinate of corresponding genotypic class. The center of phenotype (x<sub>geno</sub>,y<sub>geno</sub>) of all the metabolites significant for an interaction were plotted to visualize the distribution centers of phenotype for each pairwise interaction.</p></sec><sec id="s4-8"><title>Sequence diversity between Kas and Tsu Organelles</title><p>Kas-1 and Tsu-1 reads were obtained from NCBI (<ext-link ext-link-type="uri" xlink:href="http://www.ncbi.nlm.nih.gov/">www.ncbi.nlm.nih.gov/</ext-link>, accessions SRX246466 and SRX246442 respectively) and aligned using Bowtie2 (<xref ref-type="bibr" rid="bib42">Langmead et al., 2009</xref>; <xref ref-type="bibr" rid="bib41">Langmead and Salzberg, 2012</xref>) to the Col-0 mitochondrial and chloroplast references obtained from TAIR (<ext-link ext-link-type="uri" xlink:href="http://www.arabidopsis.org/">www.arabidopsis.org/</ext-link>) (<xref ref-type="bibr" rid="bib57">Rhee et al., 2003</xref>). Aligned sequence reads were subsequently processed using SAMtools (<xref ref-type="bibr" rid="bib44">Li et al., 2009</xref>), Picard (<ext-link ext-link-type="uri" xlink:href="http://picard.sourceforge.net">http://picard.sourceforge.net</ext-link>) and the Genome Analysis Toolkit (GATK) (<xref ref-type="bibr" rid="bib49">McKenna et al., 2010</xref>). SNP discovery between Kas-1 and Tsu-1 was carried out using the UnifiedGenotyper package of GATK (<xref ref-type="supplementary-material" rid="SD5-data">Figure 5—source data 1</xref>).</p></sec></sec></body><back><sec sec-type="additional-information"><title>Additional information</title><fn-group content-type="competing-interest"><title>Competing interests</title><fn fn-type="conflict" id="conf1"><p>DJK: Reviewing editor, <italic>eLife</italic>.</p></fn><fn fn-type="conflict" id="conf2"><p>The other authors declare that no competing interests exist.</p></fn></fn-group><fn-group content-type="author-contribution"><title>Author contributions</title><fn fn-type="con" id="con1"><p>BJ, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article</p></fn><fn fn-type="con" id="con2"><p>JAC, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article</p></fn><fn fn-type="con" id="con3"><p>BL, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article</p></fn><fn fn-type="con" id="con4"><p>SA, Acquisition of data, Analysis and interpretation of data</p></fn><fn fn-type="con" id="con5"><p>DJK, Conception and design, Analysis and interpretation of data, Drafting or revising the article</p></fn></fn-group></sec><ref-list><title>References</title><ref id="bib1"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Ajjawi</surname><given-names>I</given-names></name><name><surname>Lu</surname><given-names>Y</given-names></name><name><surname>Savage</surname><given-names>LJ</given-names></name><name><surname>Bell</surname><given-names>SM</given-names></name><name><surname>Last</surname><given-names>RL</given-names></name></person-group><year>2010</year><article-title>Large-scale reverse genetics in Arabidopsis: case studies from the Chloroplast 2010 Project</article-title><source>Plant Physiol</source><volume>152</volume><fpage>529</fpage><lpage>40</lpage><pub-id pub-id-type="doi">10.1104/pp.109.148494</pub-id></element-citation></ref><ref id="bib2"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Altshuler</surname><given-names>DL</given-names></name><name><surname>Durbin</surname><given-names>RM</given-names></name><name><surname>Abecasis</surname><given-names>GR</given-names></name><name><surname>Bentley</surname><given-names>DR</given-names></name><name><surname>Chakravarti</surname><given-names>A</given-names></name><name><surname>Clark</surname><given-names>AG</given-names></name><etal/></person-group><year>2010</year><article-title>A map of human genome variation from population-scale sequencing</article-title><source>Nature</source><volume>467</volume><fpage>1061</fpage><lpage>73</lpage><pub-id pub-id-type="doi">10.1038/nature09534</pub-id></element-citation></ref><ref id="bib3"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Alvarez-Castro</surname><given-names>JM</given-names></name><name><surname>Carlborg</surname><given-names>O</given-names></name></person-group><year>2007</year><article-title>A unified model for functional and statistical epistasis and its application in quantitative trait loci analysis</article-title><source>Genetics</source><volume>176</volume><fpage>1151</fpage><lpage>67</lpage><pub-id pub-id-type="doi">10.1534/genetics.106.067348</pub-id></element-citation></ref><ref id="bib4"><element-citation publication-type="journal"><collab>Arabidopsis genome initiative</collab><year>2000</year><article-title>Analysis of the genome sequence of the flowering plant <italic>Arabidopsis thaliana</italic></article-title><source>Nature</source><volume>14</volume><fpage>796</fpage><lpage>815</lpage><pub-id pub-id-type="doi">10.1038/35048692</pub-id></element-citation></ref><ref id="bib5"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Azhagiri</surname><given-names>AK</given-names></name><name><surname>Maliga</surname><given-names>P</given-names></name></person-group><year>2007</year><article-title>Exceptional paternal inheritance of plastids in Arabidopsis suggests that low-frequency leakage of plastids via pollen may be universal in plants</article-title><source>Plant J</source><volume>52</volume><fpage>817</fpage><lpage>23</lpage><pub-id pub-id-type="doi">10.1111/j.1365-313X.2007.03278.x</pub-id></element-citation></ref><ref id="bib6"><element-citation publication-type="book"><person-group person-group-type="author"><name><surname>Basten</surname><given-names>CJ</given-names></name><name><surname>Weir</surname><given-names>BS</given-names></name><name><surname>Zeng</surname><given-names>Z-B</given-names></name></person-group><year>1999</year><source>QTL Cartographer, version 1.13</source><publisher-loc>Raleigh, NC</publisher-loc><publisher-name>Department of Statistics, North Carolina State University</publisher-name></element-citation></ref><ref id="bib7"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Battersby</surname><given-names>BJ</given-names></name><name><surname>Shoubridge</surname><given-names>EA</given-names></name></person-group><year>2001</year><article-title>Selection of a mtDNA sequence variant in hepatocytes of heteroplasmic mice is not due to differences in respiratory chain function or efficiency of replication</article-title><source>Hum Mol Genet</source><volume>10</volume><fpage>2469</fpage><lpage>79</lpage><pub-id pub-id-type="doi">10.1093/hmg/10.22.2469</pub-id></element-citation></ref><ref id="bib8"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Broman</surname><given-names>KW</given-names></name><name><surname>Wu</surname><given-names>H</given-names></name><name><surname>Sen</surname><given-names>Ś.</given-names></name><name><surname>Churchill</surname><given-names>GA</given-names></name></person-group><year>2003</year><article-title>R/qtl: QTL mapping in experimental crosses</article-title><source>Bioinformatics</source><volume>19</volume><fpage>889</fpage><lpage>90</lpage><pub-id pub-id-type="doi">10.1093/bioinformatics/btg112</pub-id></element-citation></ref><ref id="bib9"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Cao</surname><given-names>J</given-names></name><name><surname>Schneeberger</surname><given-names>K</given-names></name><name><surname>Ossowski</surname><given-names>S</given-names></name><name><surname>Guenther</surname><given-names>T</given-names></name><name><surname>Bender</surname><given-names>S</given-names></name><name><surname>Fitz</surname><given-names>J</given-names></name><etal/></person-group><year>2011</year><article-title>Whole-genome sequencing of multiple <italic>Arabidopsis thaliana</italic> populations</article-title><source>Nat Genet</source><volume>43</volume><fpage>956</fpage><lpage>60</lpage><pub-id pub-id-type="doi">10.1038/ng.911</pub-id></element-citation></ref><ref id="bib10"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Carlborg</surname><given-names>O</given-names></name><name><surname>Haley</surname><given-names>CS</given-names></name></person-group><year>2004</year><article-title>Epistasis: too often neglected in complex trait studies?</article-title><source>Nat Rev Gen</source><volume>5</volume><fpage>614</fpage><lpage>18</lpage><pub-id pub-id-type="doi">10.1038/nrg1407</pub-id></element-citation></ref><ref id="bib11"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Carlborg</surname><given-names>O</given-names></name><name><surname>Jacobsson</surname><given-names>L</given-names></name><name><surname>Ahgren</surname><given-names>P</given-names></name><name><surname>Siegel</surname><given-names>P</given-names></name><name><surname>Andersson</surname><given-names>L</given-names></name></person-group><year>2006</year><article-title>Epistasis and the release of genetic variation during long-term selection</article-title><source>Nat Genet</source><volume>38</volume><fpage>418</fpage><lpage>20</lpage><pub-id pub-id-type="doi">10.1038/ng1761</pub-id></element-citation></ref><ref id="bib12"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Chan</surname><given-names>EK</given-names></name><name><surname>Rowe</surname><given-names>HC</given-names></name><name><surname>Corwin</surname><given-names>JA</given-names></name><name><surname>Joseph</surname><given-names>B</given-names></name><name><surname>Kliebenstein</surname><given-names>DJ</given-names></name></person-group><year>2011</year><article-title>Combining genome-wide association mapping and transcriptional networks to identify novel genes controlling glucosinolates in <italic>Arabidopsis thaliana</italic></article-title><source>PLOS Biol</source><volume>9</volume><fpage>e1001125</fpage><pub-id pub-id-type="doi">10.1371/journal.pbio.1001125</pub-id></element-citation></ref><ref id="bib13"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Chan</surname><given-names>EK</given-names></name><name><surname>Rowe</surname><given-names>HC</given-names></name><name><surname>Hansen</surname><given-names>BG</given-names></name><name><surname>Kliebenstein</surname><given-names>DJ</given-names></name></person-group><year>2010</year><article-title>The complex genetic architecture of the metabolome</article-title><source>PLOS Genet</source><volume>6</volume><fpage>e1001198</fpage><pub-id pub-id-type="doi">10.1371/journal.pgen.1001198</pub-id></element-citation></ref><ref id="bib14"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Churchill</surname><given-names>GA</given-names></name><name><surname>Doerge</surname><given-names>RW</given-names></name></person-group><year>1994</year><article-title>Empirical threshold values for quantitative trait mapping</article-title><source>Genetics</source><volume>138</volume><fpage>963</fpage><lpage>71</lpage></element-citation></ref><ref id="bib15"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Crane</surname><given-names>PL</given-names></name><name><surname>Nyquist</surname><given-names>WE</given-names></name></person-group><year>1967</year><article-title>Effects of different gene-cytoplasm systems on quantitative characters in reciprocal F2 crosses of maize (<italic>Zea mays</italic> L.)</article-title><source>Crop Sci</source><volume>7</volume><fpage>376</fpage><lpage>8</lpage><pub-id pub-id-type="doi">10.2135/cropsci1967.0011183X000700040030x</pub-id></element-citation></ref><ref id="bib16"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Dimitrov</surname><given-names>LN</given-names></name><name><surname>Brem</surname><given-names>RB</given-names></name><name><surname>Kruglyak</surname><given-names>L</given-names></name><name><surname>Gottschling</surname><given-names>DE</given-names></name></person-group><year>2009</year><article-title>Polymorphisms in multiple genes contribute to the spontaneous mitochondrial genome instability of <italic>Saccharomyces cerevisiae</italic> S288C strains</article-title><source>Genetics</source><volume>183</volume><fpage>365</fpage><lpage>83</lpage><pub-id pub-id-type="doi">10.1534/genetics.109.104497</pub-id></element-citation></ref><ref id="bib17"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Doerge</surname><given-names>RW</given-names></name><name><surname>Churchill</surname><given-names>GA</given-names></name></person-group><year>1996</year><article-title>Permutation tests for multiple loci affecting a quantitative character</article-title><source>Genetics</source><volume>142</volume><fpage>285</fpage><lpage>94</lpage></element-citation></ref><ref id="bib18"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Dowling</surname><given-names>DK</given-names></name><name><surname>Friberg</surname><given-names>U</given-names></name><name><surname>Hailer</surname><given-names>F</given-names></name><name><surname>Arnqvist</surname><given-names>G</given-names></name></person-group><year>2007</year><article-title>Intergenomic epistasis for fitness: within-population interactions between cytoplasmic and nuclear genes in <italic>Drosophila melanogaster</italic></article-title><source>Genetics</source><volume>175</volume><fpage>235</fpage><lpage>44</lpage><pub-id pub-id-type="doi">10.1534/genetics.105.052050</pub-id></element-citation></ref><ref id="bib19"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Dowling</surname><given-names>DK</given-names></name><name><surname>Meerupati</surname><given-names>T</given-names></name><name><surname>Arnqvist</surname><given-names>G</given-names></name></person-group><year>2010</year><article-title>Cytonuclear interactions and the economics of mating in seed beetles</article-title><source>Am Nat</source><volume>176</volume><fpage>131</fpage><lpage>40</lpage><pub-id pub-id-type="doi">10.1086/653671</pub-id></element-citation></ref><ref id="bib20"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Eenink</surname><given-names>AH</given-names></name><name><surname>Garretsen</surname><given-names>F</given-names></name></person-group><year>1980</year><article-title>Research on the inheritance of fasciation in lettuce (<italic>Lactuca sativa</italic> L.)</article-title><source>Euphytica</source><volume>29</volume><fpage>653</fpage><lpage>60</lpage><pub-id pub-id-type="doi">10.1007/BF00023213</pub-id></element-citation></ref><ref id="bib21"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Estavillo</surname><given-names>GM</given-names></name><name><surname>Crisp</surname><given-names>PA</given-names></name><name><surname>Pornsiriwong</surname><given-names>W</given-names></name><name><surname>Wirtz</surname><given-names>M</given-names></name><name><surname>Collinge</surname><given-names>D</given-names></name><name><surname>Carrie</surname><given-names>C</given-names></name><etal/></person-group><year>2011</year><article-title>Evidence for a SAL1-PAP chloroplast retrograde pathway that functions in drought and high light signaling in Arabidopsis</article-title><source>Plant Cell</source><volume>23</volume><fpage>3992</fpage><lpage>4012</lpage><pub-id pub-id-type="doi">10.1105/tpc.111.091033</pub-id></element-citation></ref><ref id="bib22"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Etterson</surname><given-names>JR</given-names></name><name><surname>Keller</surname><given-names>SR</given-names></name><name><surname>Galloway</surname><given-names>LF</given-names></name></person-group><year>2007</year><article-title>Epistatic and cytonuclear interactions govern outbreeding depression in the autotetraploid <italic>Campanulastrum americanum</italic></article-title><source>Evolution</source><volume>61</volume><fpage>2671</fpage><lpage>83</lpage><pub-id pub-id-type="doi">10.1111/j.1558-5646.2007.00234.x</pub-id></element-citation></ref><ref id="bib23"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Fernie</surname><given-names>AR</given-names></name><name><surname>Aharoni</surname><given-names>A</given-names></name><name><surname>Willmitzer</surname><given-names>L</given-names></name><name><surname>Stitt</surname><given-names>M</given-names></name><name><surname>Tohge</surname><given-names>T</given-names></name><name><surname>Kopka</surname><given-names>J</given-names></name><etal/></person-group><year>2011</year><article-title>Recommendations for reporting metabolite data</article-title><source>Plant Cell</source><volume>23</volume><fpage>2477</fpage><lpage>82</lpage><pub-id pub-id-type="doi">10.1105/tpc.111.086272</pub-id></element-citation></ref><ref id="bib24"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Fiehn</surname><given-names>O</given-names></name></person-group><year>2002</year><article-title>Metabolomics - the link between genotypes and phenotypes</article-title><source>Plant Mol Biol</source><volume>48</volume><fpage>155</fpage><lpage>71</lpage><pub-id pub-id-type="doi">10.1023/A:1013713905833</pub-id></element-citation></ref><ref id="bib25"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Fiehn</surname><given-names>O</given-names></name><name><surname>Kopka</surname><given-names>J</given-names></name><name><surname>Dormann</surname><given-names>P</given-names></name><name><surname>Altmann</surname><given-names>T</given-names></name><name><surname>Trethewey</surname><given-names>RN</given-names></name><name><surname>Willmitzer</surname><given-names>L</given-names></name></person-group><year>2000</year><article-title>Metabolite profiling for plant functional genomics</article-title><source>Nat Biotechnol</source><volume>18</volume><fpage>1157</fpage><lpage>61</lpage><pub-id pub-id-type="doi">10.1038/81137</pub-id></element-citation></ref><ref id="bib26"><element-citation publication-type="book"><person-group person-group-type="author"><name><surname>Fiehn</surname><given-names>O</given-names></name><name><surname>Wohlgemuth</surname><given-names>G</given-names></name><name><surname>Scholz</surname><given-names>M</given-names></name></person-group><year>2005</year><article-title>Setup and annotation of metabolomic experiments by integrating biological and mass spectrometric metadata</article-title><person-group person-group-type="editor"><name><surname>Ludäscher</surname><given-names>B</given-names></name></person-group><source>Data integration in the life sciences, Proceedings</source><publisher-loc>Heidelberg</publisher-loc><publisher-name>Springer-Verlag</publisher-name><fpage>224</fpage><lpage>39</lpage></element-citation></ref><ref id="bib27"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Fiehn</surname><given-names>O</given-names></name><name><surname>Wohlgemuth</surname><given-names>G</given-names></name><name><surname>Scholz</surname><given-names>M</given-names></name><name><surname>Kind</surname><given-names>T</given-names></name><name><surname>Lee</surname><given-names>DY</given-names></name><name><surname>Lu</surname><given-names>Y</given-names></name><etal/></person-group><year>2008</year><article-title>Quality control for plant metabolomics: reporting MSI-compliant studies</article-title><source>Plant J</source><volume>53</volume><fpage>691</fpage><lpage>704</lpage><pub-id pub-id-type="doi">10.1111/j.1365-313X.2007.03387.x</pub-id></element-citation></ref><ref id="bib28"><element-citation publication-type="book"><person-group person-group-type="author"><name><surname>Fox</surname><given-names>J</given-names></name><name><surname>Weisberg</surname><given-names>S</given-names></name></person-group><year>2011</year><source>An R companion to applied regression</source><publisher-loc>Thousand Oaks, CA</publisher-loc><publisher-name>SAGE</publisher-name></element-citation></ref><ref id="bib29"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Gartenhaus</surname><given-names>S</given-names></name><name><surname>Schwartz</surname><given-names>C</given-names></name></person-group><year>1957</year><article-title>Center-of-mass motion in many-particle systems</article-title><source>Phys Rev</source><volume>108</volume><fpage>482</fpage><lpage>90</lpage><pub-id pub-id-type="doi">10.1103/PhysRev.108.482</pub-id></element-citation></ref><ref id="bib30"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Hanson</surname><given-names>MR</given-names></name></person-group><year>1991</year><article-title>Plant mitochondrial mutations and male-sterility</article-title><source>Annu Rev Genet</source><volume>25</volume><fpage>461</fpage><lpage>86</lpage><pub-id pub-id-type="doi">10.1146/annurev.ge.25.120191.002333</pub-id></element-citation></ref><ref id="bib31"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Jansen</surname><given-names>RC</given-names></name></person-group><year>1994</year><article-title>Controlling the type-I and type-II errors in mapping quantitative trait loci</article-title><source>Genetics</source><volume>138</volume><fpage>871</fpage><lpage>81</lpage></element-citation></ref><ref id="bib32"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Joseph</surname><given-names>B</given-names></name><name><surname>Corwin</surname><given-names>JA</given-names></name><name><surname>Zuest</surname><given-names>T</given-names></name><name><surname>Li</surname><given-names>B</given-names></name><name><surname>Iravani</surname><given-names>M</given-names></name><name><surname>Schaepman-Strub</surname><given-names>G</given-names></name><etal/></person-group><year>2013</year><article-title>Hierarchical nuclear and cytoplasmic genetic architectures for plant growth and defense</article-title><source>Plant Cell</source><volume>25</volume><fpage>1929</fpage><lpage>45</lpage><pub-id pub-id-type="doi">10.1105/tpc.113.112615</pub-id></element-citation></ref><ref id="bib33"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Juenger</surname><given-names>TE</given-names></name><name><surname>Sen</surname><given-names>S</given-names></name><name><surname>Bray</surname><given-names>E</given-names></name><name><surname>Stahl</surname><given-names>E</given-names></name><name><surname>Wayne</surname><given-names>T</given-names></name><name><surname>McKay</surname><given-names>J</given-names></name><etal/></person-group><year>2010</year><article-title>Exploring genetic and expression differences between physiologically extreme ecotypes: comparative genomic hybridization and gene expression studies of Kas-1 and Tsu-1 accessions of <italic>Arabidopsis thaliana</italic></article-title><source>Plant Cell Environ</source><volume>33</volume><fpage>1268</fpage><lpage>84</lpage><pub-id pub-id-type="doi">10.1111/j.1365-3040.2010.02146.x</pub-id></element-citation></ref><ref id="bib34"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Juenger</surname><given-names>TE</given-names></name><name><surname>Wayne</surname><given-names>T</given-names></name><name><surname>Boles</surname><given-names>S</given-names></name><name><surname>Symonds</surname><given-names>VV</given-names></name><name><surname>McKay</surname><given-names>J</given-names></name><name><surname>Coughlan</surname><given-names>SJ</given-names></name></person-group><year>2006</year><article-title>Natural genetic variation in whole-genome expression in <italic>Arabidopsis thaliana</italic>: the impact of physiological QTL introgression</article-title><source>Mol Ecol</source><volume>15</volume><fpage>1351</fpage><lpage>65</lpage><pub-id pub-id-type="doi">10.1111/j.1365-294X.2006.02774.x</pub-id></element-citation></ref><ref id="bib35"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Kliebenstein</surname><given-names>DJ</given-names></name><name><surname>Gershenzon</surname><given-names>J</given-names></name><name><surname>Mitchell-Olds</surname><given-names>T</given-names></name></person-group><year>2001a</year><article-title>Comparative quantitative trait loci mapping of aliphatic, indolic and benzylic glucosinolate production in <italic>Arabidopsis thaliana</italic> leaves and seeds</article-title><source>Genetics</source><volume>159</volume><fpage>359</fpage><lpage>70</lpage></element-citation></ref><ref id="bib36"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Kliebenstein</surname><given-names>DJ</given-names></name><name><surname>Kroymann</surname><given-names>J</given-names></name><name><surname>Brown</surname><given-names>P</given-names></name><name><surname>Figuth</surname><given-names>A</given-names></name><name><surname>Pedersen</surname><given-names>D</given-names></name><name><surname>Gershenzon</surname><given-names>J</given-names></name><etal/></person-group><year>2001b</year><article-title>Genetic control of natural variation in <italic>Arabidopsis thaliana</italic> glucosinolate accumulation</article-title><source>Plant Physiol</source><volume>126</volume><fpage>811</fpage><lpage>25</lpage><pub-id pub-id-type="doi">10.1104/pp.126.2.811</pub-id></element-citation></ref><ref id="bib37"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Kliebenstein</surname><given-names>DJ</given-names></name><name><surname>Lambrix</surname><given-names>VM</given-names></name><name><surname>Reichelt</surname><given-names>M</given-names></name><name><surname>Gershenzon</surname><given-names>J</given-names></name><name><surname>Mitchell-Olds</surname><given-names>T</given-names></name></person-group><year>2001c</year><article-title>Gene duplication and the diversification of secondary metabolism: side chain modification of glucosinolates in <italic>Arabidopsis thaliana</italic></article-title><source>Plant Cell</source><volume>13</volume><fpage>681</fpage><lpage>93</lpage><pub-id pub-id-type="doi">10.1105/tpc.13.3.681</pub-id></element-citation></ref><ref id="bib38"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Kliebenstein</surname><given-names>DJ</given-names></name><name><surname>West</surname><given-names>MA</given-names></name><name><surname>van Leeuwen</surname><given-names>H</given-names></name><name><surname>Loudet</surname><given-names>O</given-names></name><name><surname>Doerge</surname><given-names>RW</given-names></name><name><surname>St Clair</surname><given-names>DA</given-names></name></person-group><year>2006</year><article-title>Identification of QTLs controlling gene expression networks defined a priori</article-title><source>BMC Bioinformatics</source><volume>7</volume><fpage>308</fpage><pub-id pub-id-type="doi">10.1186/1471-2105-7-308</pub-id></element-citation></ref><ref id="bib39"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Kroymann</surname><given-names>J</given-names></name><name><surname>Donnerhacke</surname><given-names>S</given-names></name><name><surname>Schnabelrauch</surname><given-names>D</given-names></name><name><surname>Mitchell-Olds</surname><given-names>T</given-names></name></person-group><year>2003</year><article-title>Evolutionary dynamics of an Arabidopsis insect resistance quantitative trait locus</article-title><source>Proc Natl Acad Sci USA</source><volume>100</volume><fpage>14587</fpage><lpage>92</lpage><pub-id pub-id-type="doi">10.1073/pnas.1734046100</pub-id></element-citation></ref><ref id="bib40"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Krueger</surname><given-names>S</given-names></name><name><surname>Giavalisco</surname><given-names>P</given-names></name><name><surname>Krall</surname><given-names>L</given-names></name><name><surname>Steinhauser</surname><given-names>MC</given-names></name><name><surname>Bussis</surname><given-names>D</given-names></name><name><surname>Usadel</surname><given-names>B</given-names></name><etal/></person-group><year>2011</year><article-title>A Topological map of the compartmentalized <italic>Arabidopsis thaliana</italic> leaf metabolome</article-title><source>PLOS ONE</source><volume>6</volume><fpage>e17806</fpage><pub-id pub-id-type="doi">10.1371/journal.pone.0017806</pub-id></element-citation></ref><ref id="bib41"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Langmead</surname><given-names>B</given-names></name><name><surname>Salzberg</surname><given-names>SL</given-names></name></person-group><year>2012</year><article-title>Fast gapped-read alignment with Bowtie 2</article-title><source>Nat Methods</source><volume>9</volume><fpage>354</fpage><lpage>57</lpage><pub-id pub-id-type="doi">10.1038/nmeth.1923</pub-id></element-citation></ref><ref id="bib42"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Langmead</surname><given-names>B</given-names></name><name><surname>Trapnell</surname><given-names>C</given-names></name><name><surname>Pop</surname><given-names>M</given-names></name><name><surname>Salzberg</surname><given-names>SL</given-names></name></person-group><year>2009</year><article-title>Ultrafast and memory-efficient alignment of short DNA sequences to the human genome</article-title><source>Genome Biol</source><volume>10</volume><fpage>R25</fpage><pub-id pub-id-type="doi">10.1186/gb-2009-10-3-r25</pub-id></element-citation></ref><ref id="bib43"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Larkin</surname><given-names>RM</given-names></name><name><surname>Alonso</surname><given-names>JM</given-names></name><name><surname>Ecker</surname><given-names>JR</given-names></name><name><surname>Chory</surname><given-names>J</given-names></name></person-group><year>2003</year><article-title>GUN4, a regulator of chlorophyll synthesis and intracellular signaling</article-title><source>Science</source><volume>299</volume><fpage>902</fpage><lpage>6</lpage><pub-id pub-id-type="doi">10.1126/science.1079978</pub-id></element-citation></ref><ref id="bib44"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Li</surname><given-names>H</given-names></name><name><surname>Handsaker</surname><given-names>B</given-names></name><name><surname>Wysoker</surname><given-names>A</given-names></name><name><surname>Fennell</surname><given-names>T</given-names></name><name><surname>Ruan</surname><given-names>J</given-names></name><name><surname>Homer</surname><given-names>N</given-names></name><etal/></person-group><year>2009</year><article-title>The Sequence alignment/map (SAM) format and SAMtools</article-title><source>Bioinformatics</source><volume>25</volume><fpage>2078</fpage><lpage>9</lpage><pub-id pub-id-type="doi">10.1093/bioinformatics/btp352</pub-id></element-citation></ref><ref id="bib45"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Liti</surname><given-names>G</given-names></name><name><surname>Carter</surname><given-names>DM</given-names></name><name><surname>Moses</surname><given-names>AM</given-names></name><name><surname>Warringer</surname><given-names>J</given-names></name><name><surname>Parts</surname><given-names>L</given-names></name><name><surname>James</surname><given-names>SA</given-names></name><etal/></person-group><year>2009</year><article-title>Population genomics of domestic and wild yeasts</article-title><source>Nature</source><volume>458</volume><fpage>337</fpage><lpage>41</lpage><pub-id pub-id-type="doi">10.1038/nature07743</pub-id></element-citation></ref><ref id="bib46"><element-citation publication-type="book"><person-group person-group-type="author"><name><surname>Liu</surname><given-names>B-H</given-names></name></person-group><year>1998</year><source>Statistical genomics: linkage, mapping and QTL analysis</source><publisher-loc>Boca Raton, Florida</publisher-loc><publisher-name>CRC Press</publisher-name></element-citation></ref><ref id="bib47"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Magrath</surname><given-names>R</given-names></name><name><surname>Bano</surname><given-names>F</given-names></name><name><surname>Morgner</surname><given-names>M</given-names></name><name><surname>Parkin</surname><given-names>I</given-names></name><name><surname>Sharpe</surname><given-names>A</given-names></name><name><surname>Lister</surname><given-names>C</given-names></name><etal/></person-group><year>1994</year><article-title>Genetics of aliphatic glucosinolates. I. Side chain elongation in <italic>Brassica napus</italic> and <italic>Arabidopsis thaliana</italic></article-title><source>Heredity</source><volume>72</volume><fpage>290</fpage><lpage>9</lpage><pub-id pub-id-type="doi">10.1038/hdy.1994.39</pub-id></element-citation></ref><ref id="bib48"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>McKay</surname><given-names>JK</given-names></name><name><surname>Richards</surname><given-names>JH</given-names></name><name><surname>Nemali</surname><given-names>KS</given-names></name><name><surname>Sen</surname><given-names>S</given-names></name><name><surname>Mitchell-Olds</surname><given-names>T</given-names></name><name><surname>Boles</surname><given-names>S</given-names></name><etal/></person-group><year>2008</year><article-title>Genetics of drought adaptation in <italic>Arabidopsis thaliana</italic> II. QTL analysis of a new mapping population Kas-1 x Tsu-1</article-title><source>Evolution</source><volume>62</volume><fpage>3014</fpage><lpage>26</lpage><pub-id pub-id-type="doi">10.1111/j.1558-5646.2008.00474.x</pub-id></element-citation></ref><ref id="bib49"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>McKenna</surname><given-names>A</given-names></name><name><surname>Hanna</surname><given-names>M</given-names></name><name><surname>Banks</surname><given-names>E</given-names></name><name><surname>Sivachenko</surname><given-names>A</given-names></name><name><surname>Cibulskis</surname><given-names>K</given-names></name><name><surname>Kernytsky</surname><given-names>A</given-names></name><etal/></person-group><year>2010</year><article-title>The genome analysis toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data</article-title><source>Genome Res</source><volume>20</volume><fpage>1297</fpage><lpage>303</lpage><pub-id pub-id-type="doi">10.1101/gr.107524.110</pub-id></element-citation></ref><ref id="bib50"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>McRae</surname><given-names>AF</given-names></name><name><surname>Byrne</surname><given-names>EM</given-names></name><name><surname>Zhao</surname><given-names>ZZ</given-names></name><name><surname>Montgomery</surname><given-names>GW</given-names></name><name><surname>Visscher</surname><given-names>PM</given-names></name></person-group><year>2008</year><article-title>Power and SNP tagging in whole mitochondrial genome association studies</article-title><source>Genome Res</source><volume>18</volume><fpage>911</fpage><lpage>7</lpage><pub-id pub-id-type="doi">10.1101/gr.074872.107</pub-id></element-citation></ref><ref id="bib51"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Miura</surname><given-names>H</given-names></name><name><surname>Fukuda</surname><given-names>Y</given-names></name><name><surname>Sawada</surname><given-names>S</given-names></name></person-group><year>1997</year><article-title>Expression of dormancy in diallel F1 and F2 seed of wheat ripened under controlled environment</article-title><source>J Genet Breed</source><volume>51</volume><fpage>195</fpage><lpage>200</lpage></element-citation></ref><ref id="bib52"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Moison</surname><given-names>M</given-names></name><name><surname>Roux</surname><given-names>F</given-names></name><name><surname>Quadrado</surname><given-names>M</given-names></name><name><surname>Duval</surname><given-names>R</given-names></name><name><surname>Ekovich</surname><given-names>M</given-names></name><name><surname>Le</surname><given-names>D-H</given-names></name><etal/></person-group><year>2010</year><article-title>Cytoplasmic phylogeny and evidence of cyto-nuclear co-adaptation in <italic>Arabidopsis thaliana</italic></article-title><source>Plant J</source><volume>63</volume><fpage>728</fpage><lpage>38</lpage><pub-id pub-id-type="doi">10.1111/j.1365-313X.2010.04275.x</pub-id></element-citation></ref><ref id="bib53"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Park</surname><given-names>HB</given-names></name><name><surname>Jacobsson</surname><given-names>L</given-names></name><name><surname>Wahlberg</surname><given-names>P</given-names></name><name><surname>Siegel</surname><given-names>PB</given-names></name><name><surname>Andersson</surname><given-names>L</given-names></name></person-group><year>2006</year><article-title>QTL analysis of body composition and metabolic traits in an intercross between chicken lines divergently selected for growth</article-title><source>Physiol Genomics</source><volume>25</volume><fpage>216</fpage><lpage>23</lpage><pub-id pub-id-type="doi">10.1152/physiolgenomics.00113.2005</pub-id></element-citation></ref><ref id="bib54"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Primomo</surname><given-names>VS</given-names></name><name><surname>Falk</surname><given-names>DE</given-names></name><name><surname>Ablett</surname><given-names>GR</given-names></name><name><surname>Tanner</surname><given-names>JW</given-names></name><name><surname>Rajcan</surname><given-names>I</given-names></name></person-group><year>2002</year><article-title>Inheritance and interaction of low palmitic and low linolenic soybean</article-title><source>Crop Sci</source><volume>42</volume><fpage>31</fpage><lpage>6</lpage><pub-id pub-id-type="doi">10.2135/cropsci2002.0031</pub-id></element-citation></ref><ref id="bib55"><element-citation publication-type="book"><collab>R Development Core Team</collab><year>2012</year><source>R: a language and environment for statistical computing</source><publisher-loc>Vienna</publisher-loc><publisher-name>R Foundation for Statistical Computing</publisher-name></element-citation></ref><ref id="bib56"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Rebai</surname><given-names>A</given-names></name></person-group><year>1997</year><article-title>Comparison of methods for regression interval mapping in QTL analysis with non-normal traits</article-title><source>Genet Rec</source><volume>69</volume><fpage>69</fpage><lpage>74</lpage><pub-id pub-id-type="doi">10.1017/S0016672396002558</pub-id></element-citation></ref><ref id="bib57"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Rhee</surname><given-names>SY</given-names></name><name><surname>Beavis</surname><given-names>W</given-names></name><name><surname>Berardini</surname><given-names>TZ</given-names></name><name><surname>Chen</surname><given-names>GH</given-names></name><name><surname>Dixon</surname><given-names>D</given-names></name><name><surname>Doyle</surname><given-names>A</given-names></name><etal/></person-group><year>2003</year><article-title>The Arabidopsis information resource (TAIR): a model organism database providing a centralized, curated gateway to Arabidopsis biology, research materials and community</article-title><source>Nucleic Acids Res</source><volume>31</volume><fpage>224</fpage><lpage>8</lpage><pub-id pub-id-type="doi">10.1093/nar/gkg076</pub-id></element-citation></ref><ref id="bib58"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Roessner</surname><given-names>U</given-names></name><name><surname>Luedemann</surname><given-names>A</given-names></name><name><surname>Brust</surname><given-names>D</given-names></name><name><surname>Fiehn</surname><given-names>O</given-names></name><name><surname>Linke</surname><given-names>T</given-names></name><name><surname>Willmitzer</surname><given-names>L</given-names></name><etal/></person-group><year>2001</year><article-title>Metabolic profiling allows comprehensive phenotyping of genetically or environmentally modified plant systems</article-title><source>Plant Cell</source><volume>13</volume><fpage>11</fpage><lpage>29</lpage><pub-id pub-id-type="doi">10.1105/tpc.13.1.11</pub-id></element-citation></ref><ref id="bib59"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Roubertoux</surname><given-names>PL</given-names></name><name><surname>Sluyter</surname><given-names>F</given-names></name><name><surname>Carlier</surname><given-names>M</given-names></name><name><surname>Marcet</surname><given-names>B</given-names></name><name><surname>Maarouf-Veray</surname><given-names>F</given-names></name><name><surname>Cherif</surname><given-names>C</given-names></name><etal/></person-group><year>2003</year><article-title>Mitochondrial DNA modifies cognition in interaction with the nuclear genome and age in mice</article-title><source>Nat Genet</source><volume>35</volume><fpage>65</fpage><lpage>9</lpage><pub-id pub-id-type="doi">10.1038/ng1230</pub-id></element-citation></ref><ref id="bib60"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Rowe</surname><given-names>HC</given-names></name><name><surname>Hansen</surname><given-names>BG</given-names></name><name><surname>Halkier</surname><given-names>BA</given-names></name><name><surname>Kliebenstein</surname><given-names>DJ</given-names></name></person-group><year>2008</year><article-title>Biochemical networks and epistasis shape the <italic>Arabidopsis thaliana</italic> metabolome</article-title><source>Plant Cell</source><volume>20</volume><fpage>1199</fpage><lpage>216</lpage><pub-id pub-id-type="doi">10.1105/tpc.108.058131</pub-id></element-citation></ref><ref id="bib61"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Rowe</surname><given-names>HC</given-names></name><name><surname>Kliebenstein</surname><given-names>DJ</given-names></name></person-group><year>2008</year><article-title>Complex genetics control natural variation in <italic>Arabidopsis thaliana</italic> resistance to <italic>Botrytis cinerea</italic></article-title><source>Genetics</source><volume>180</volume><fpage>2237</fpage><lpage>50</lpage><pub-id pub-id-type="doi">10.1534/genetics.108.091439</pub-id></element-citation></ref><ref id="bib62"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Schnable</surname><given-names>PS</given-names></name><name><surname>Wise</surname><given-names>RP</given-names></name></person-group><year>1998</year><article-title>The molecular basis of cytoplasmic male sterility and fertility restoration</article-title><source>Trends Plant Sci</source><volume>3</volume><fpage>175</fpage><lpage>80</lpage><pub-id pub-id-type="doi">10.1016/S1360-1385(98)01235-7</pub-id></element-citation></ref><ref id="bib63"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Schon</surname><given-names>EA</given-names></name><name><surname>DiMauro</surname><given-names>S</given-names></name><name><surname>Hirano</surname><given-names>M</given-names></name></person-group><year>2012</year><article-title>Human mitochondrial DNA: roles of inherited and somatic mutations</article-title><source>Nat Rev Genet</source><volume>13</volume><fpage>878</fpage><lpage>90</lpage><pub-id pub-id-type="doi">10.1038/nrg3275</pub-id></element-citation></ref><ref id="bib64"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Schon</surname><given-names>EA</given-names></name><name><surname>Przedborski</surname><given-names>S</given-names></name></person-group><year>2011</year><article-title>Mitochondria: the next (neurode) generation</article-title><source>Neuron</source><volume>70</volume><fpage>1033</fpage><lpage>53</lpage><pub-id pub-id-type="doi">10.1016/j.neuron.2011.06.003</pub-id></element-citation></ref><ref id="bib65"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Singh</surname><given-names>M</given-names></name></person-group><year>1965</year><article-title>Cytoplasmic effects on some agronomic characters in backcross maize hybrids</article-title><source>Indian J Genet Plant Breed</source><volume>25</volume><fpage>198</fpage><lpage>207</lpage></element-citation></ref><ref id="bib66"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Smoot</surname><given-names>M</given-names></name><name><surname>Ono</surname><given-names>K</given-names></name><name><surname>Ruscheinski</surname><given-names>J</given-names></name><name><surname>Wang</surname><given-names>P-L</given-names></name><name><surname>Ideker</surname><given-names>T</given-names></name></person-group><year>2011</year><article-title>Cytoscape 2.8: new features for data integration and network visualization</article-title><source>Bioinformatics</source><volume>27</volume><fpage>431</fpage><lpage>2</lpage><pub-id pub-id-type="doi">10.1093/bioinformatics/btq675</pub-id></element-citation></ref><ref id="bib67"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Tang</surname><given-names>Z</given-names></name><name><surname>Wang</surname><given-names>X</given-names></name><name><surname>Hu</surname><given-names>Z</given-names></name><name><surname>Yang</surname><given-names>Z</given-names></name><name><surname>Xu</surname><given-names>C</given-names></name></person-group><year>2007</year><article-title>Genetic dissection of cytonuclear epistasis in line crosses</article-title><source>Genetics</source><volume>177</volume><fpage>669</fpage><lpage>72</lpage><pub-id pub-id-type="doi">10.1534/genetics.107.074302</pub-id></element-citation></ref><ref id="bib68"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Tang</surname><given-names>Z</given-names></name><name><surname>Yang</surname><given-names>Z</given-names></name><name><surname>Hu</surname><given-names>Z</given-names></name><name><surname>Zhang</surname><given-names>D</given-names></name><name><surname>Lu</surname><given-names>X</given-names></name><name><surname>Deng</surname><given-names>D</given-names></name><etal/></person-group><year>2013</year><article-title>Cytonuclear epistatic quantitative trait locus mapping for plant height and ear height in maize</article-title><source>Mol Breed</source><volume>31</volume><fpage>1</fpage><lpage>14</lpage><pub-id pub-id-type="doi">10.1007/s11032-012-9762-3</pub-id></element-citation></ref><ref id="bib69"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Tao</surname><given-names>DY</given-names></name><name><surname>Hu</surname><given-names>FY</given-names></name><name><surname>Yang</surname><given-names>JY</given-names></name><name><surname>Yang</surname><given-names>GF</given-names></name><name><surname>Yang</surname><given-names>YQ</given-names></name><name><surname>Xu</surname><given-names>P</given-names></name><etal/></person-group><year>2004</year><article-title>Cytoplasm and cytoplasm-nucleus interactions affect agronomic traits in japonica rice</article-title><source>Euphytica</source><volume>135</volume><fpage>129</fpage><lpage>34</lpage><pub-id pub-id-type="doi">10.1023/B:EUPH.0000009548.81467.73</pub-id></element-citation></ref><ref id="bib70"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Taylor</surname><given-names>RW</given-names></name><name><surname>Turnbull</surname><given-names>DM</given-names></name></person-group><year>2005</year><article-title>Mitochondrial DNA mutations in human disease</article-title><source>Nat Rev Genet</source><volume>6</volume><fpage>389</fpage><lpage>402</lpage><pub-id pub-id-type="doi">10.1038/nrg1606</pub-id></element-citation></ref><ref id="bib71"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Vinti</surname><given-names>G</given-names></name><name><surname>Hills</surname><given-names>A</given-names></name><name><surname>Campbell</surname><given-names>S</given-names></name><name><surname>Bowyer</surname><given-names>JR</given-names></name><name><surname>Mochizuki</surname><given-names>N</given-names></name><name><surname>Chory</surname><given-names>J</given-names></name><etal/></person-group><year>2000</year><article-title>Interactions between hy1 and gun mutants of Arabidopsis, and their implications for plastid/nuclear signalling</article-title><source>Plant J</source><volume>24</volume><fpage>883</fpage><lpage>94</lpage><pub-id pub-id-type="doi">10.1046/j.1365-313x.2000.00936.x</pub-id></element-citation></ref><ref id="bib72"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Wade</surname><given-names>MJ</given-names></name><name><surname>Goodnight</surname><given-names>CJ</given-names></name></person-group><year>2006</year><article-title>Cyto-nuclear epistasis: two-locus random genetic drift in hermaphroditic and dioecious species</article-title><source>Evolution</source><volume>60</volume><fpage>643</fpage><lpage>59</lpage><pub-id pub-id-type="doi">10.1111/j.0014-3820.2006.tb01146.x</pub-id></element-citation></ref><ref id="bib73"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Wallace</surname><given-names>DC</given-names></name><name><surname>Singh</surname><given-names>G</given-names></name><name><surname>Lott</surname><given-names>MT</given-names></name><name><surname>Hodge</surname><given-names>JA</given-names></name><name><surname>Schurr</surname><given-names>TG</given-names></name><name><surname>Lezza</surname><given-names>AMS</given-names></name><etal/></person-group><year>1988</year><article-title>Mitochondrial DNA mutation associated with Leber’s hereditary optic neuropathy</article-title><source>Science</source><volume>242</volume><fpage>1427</fpage><lpage>30</lpage><pub-id pub-id-type="doi">10.1126/science.3201231</pub-id></element-citation></ref><ref id="bib74"><element-citation publication-type="book"><person-group person-group-type="author"><name><surname>Wang</surname><given-names>S</given-names></name><name><surname>Basten</surname><given-names>CJ</given-names></name><name><surname>Zeng</surname><given-names>Z-B</given-names></name></person-group><year>2006</year><source>Windows QTL Cartographer 2.5</source><publisher-loc>Raleigh, NC</publisher-loc><publisher-name>Department of Statistics, North Carolina State University</publisher-name></element-citation></ref><ref id="bib75"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Willett</surname><given-names>CS</given-names></name></person-group><year>2012</year><article-title>Quantifying the elevation of mitochondrial DNA evolutionary substitution rates over nuclear rates in the intertidal copepod <italic>Tigriopus californicus</italic></article-title><source>J Mol Evol</source><volume>74</volume><fpage>310</fpage><lpage>8</lpage><pub-id pub-id-type="doi">10.1007/s00239-012-9508-1</pub-id></element-citation></ref><ref id="bib76"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Willett</surname><given-names>CS</given-names></name><name><surname>Burton</surname><given-names>RS</given-names></name></person-group><year>2004</year><article-title>Evolution of interacting proteins in the mitochondrial electron transport system in a marine copepod</article-title><source>Mol Biol Evol</source><volume>21</volume><fpage>443</fpage><lpage>53</lpage><pub-id pub-id-type="doi">10.1093/molbev/msh031</pub-id></element-citation></ref><ref id="bib77"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Wolf</surname><given-names>JB</given-names></name></person-group><year>2009</year><article-title>Cytonuclear interactions can favor the evolution of genomic imprinting</article-title><source>Evolution</source><volume>63</volume><fpage>1364</fpage><lpage>71</lpage><pub-id pub-id-type="doi">10.1111/j.1558-5646.2009.00632.x</pub-id></element-citation></ref><ref id="bib78"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Xiao</surname><given-names>Y</given-names></name><name><surname>Savchenko</surname><given-names>T</given-names></name><name><surname>Baidoo</surname><given-names>EEK</given-names></name><name><surname>Chehab</surname><given-names>WE</given-names></name><name><surname>Hayden</surname><given-names>DM</given-names></name><name><surname>Tolstikov</surname><given-names>V</given-names></name><etal/></person-group><year>2012</year><article-title>Retrograde signaling by the plastidial metabolite MEcPP regulates expression of nuclear stress-response genes</article-title><source>Cell</source><volume>149</volume><fpage>1525</fpage><lpage>35</lpage><pub-id pub-id-type="doi">10.1016/j.cell.2012.04.038</pub-id></element-citation></ref><ref id="bib79"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Zeng</surname><given-names>Z-B</given-names></name><name><surname>Kao</surname><given-names>C-H</given-names></name><name><surname>Basten</surname><given-names>CJ</given-names></name></person-group><year>1999</year><article-title>Estimating the genetic architecture of quantitative traits</article-title><source>Genet Res</source><volume>75</volume><fpage>345</fpage><lpage>55</lpage></element-citation></ref><ref id="bib80"><element-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Zeyl</surname><given-names>C</given-names></name><name><surname>Andreson</surname><given-names>B</given-names></name><name><surname>Weninck</surname><given-names>E</given-names></name></person-group><year>2005</year><article-title>Nuclear-mitochondrial epistasis for fitness in <italic>Saccharomyces cerevisiae</italic></article-title><source>Evolution</source><volume>59</volume><fpage>910</fpage><lpage>4</lpage><pub-id pub-id-type="doi">10.1111/j.0014-3820.2005.tb01764.x</pub-id></element-citation></ref></ref-list></back><sub-article article-type="article-commentary" id="SA1"><front-stub><article-id pub-id-type="doi">10.7554/eLife.00776.048</article-id><title-group><article-title>Decision letter</article-title></title-group><contrib-group content-type="section"><contrib contrib-type="editor"><name><surname>Weigel</surname><given-names>Detlef</given-names></name><role>Reviewing editor</role><aff><institution>Max Planck Institute for Developmental Biology</institution>, <country>Germany</country></aff></contrib></contrib-group></front-stub><body><boxed-text><p>eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see <ext-link ext-link-type="uri" xlink:href="http://elife.elifesciences.org/review-process">review process</ext-link>). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.</p></boxed-text><p>Thank you for sending your work entitled “Cytosolic genetic variation and extensive cytonuclear interactions control natural variation in the metabolome” for consideration at <italic>eLife</italic>. Your article has been favorably peer reviewed by Deputy editor Detlef Weigel and another reviewer.</p><p>The two reviewers discussed their comments before we reached this decision, and they have assembled the following comments to help you prepare a revised submission.</p><p>The main point of this paper, that cytoplasmic factors are likely to play an important role in phenotypic variation and should not be ignored, is important, and well substantiated in this paper. To support their claim with many phenotypes, the authors measured hundreds of metabolites in two <italic>A. thaliana</italic> recombinant inbred line populations, both from the same two accessions, but with one set having inherited the cytoplasm from one accession, and the other from the other accession. The authors convincingly show that inherited differences in the cytoplasm – almost certainly mutations in the mitochondrial and chloroplast genomes – have measurable and non-negligible effects on metabolites, and that there is substantial epistatic interaction between organellar and nuclear genomes.</p><p>That said, several claims seem overstated. Starting with the title, “control” is clearly too strong. First, the authors should spell out what fraction of the genetic variation is attributable to cytoplasmic variants, what fraction is due to C×G interactions, and how does this compare to straight G effects? Second, how representative the findings of the paper are of other phenotypes remains unknown, and the authors should be more circumspect when extending their claims to phenotypes that might have no or only a very indirect connection to metabolites. Furthermore, both reviewers found the paper hard to follow – the presentation needs to be greatly simplified, so that the important points come across better.</p><p>Specific criticisms:</p><p>1) The authors’ explanation for the average heritability due to the organellar genome being much lower than the one coming from the nuclear genomes seems reasonable. However, both reviewers had a hard time trying to understand why there are apparently no metabolites that show cytoplasmic heritability greater than 0.15. Two different models were used to estimate nuclear and cytoplasmic heritability, but why? Along these lines, how do cytoplasmic and nuclear heritability compare for the metabolites significant for both? And on a related note, please explain why significant effects of the nuclear genome were estimated by partitioning the variance across 316 RIL genomes. These are not independent: surely you want the estimates produced by linkage mapping? Finally, that the complete cytoplasmic genome has a larger effect than individual nuclear loci (<xref ref-type="fig" rid="fig6">Figure 6</xref>) does not seem surprising, given that the mtDNA and cpDNA represent contain dozens of genes.</p><p>2) A substantial criticism is that the authors do not tell us anything about sequence differences in the organellar genomes of the RIL parents. The short read data are available, so it would not be too much of a burden to determine how many coding variants there are, and how the distribution of coding variants compares between biosynthetic genes in the organellar and nuclear genomes. We note that the Discussion states that differences have been found in the organellar genomes.</p><p>3) You discuss whether RILs have less phenotypic variation because they have less genetic variation. You write “A common concern of RIL populations is that they only sample two alleles per locus and thus might be limited in their genetic variation relative to species-wide diversity (Kim et al., 2007; Nordborg and Weigel, 2008).” This may or may not be common concern, but it is not the chief one, in our opinion, and it is also not the point made in the papers cited. As far as one can tell, Kim et al. (2007) does not discuss this issue at all, and Nordborg and Weigel (2008) make a completely different point, namely that RILs have greater power to detect rare alleles of large effect that “may not be important from an evolutionary perspective”. This is central to your paper because of the very real possibility that the observed QTL hotspots (which mostly differ from another RIL population) are due to rare alleles of large effect, which may well be deleterious. This does not invalidate the results, but changes their interpretation, in particular with respect to their evolutionary significance.</p><p>4) There is little consideration of existing knowledge of the contribution of organellar biosynthesis steps to metabolic pathways. An important question is whether metabolites that are known to be made in organelles are more likely to be affected by variation in the organellar genomes than those that are not.</p><p>Finally, the authors make the important point that the organellar genotypes would add greatly to GWA studies. We agree. Since the authors have metabolome data for 96 accessions (<xref ref-type="bibr" rid="bib13">Chan et al., 2010</xref>), one wonders how much effort it would be to download existing read data, identify polymorphisms in the mitochondrial and chloroplast genomes, and include them in the GWA analysis. (A major unknown is what the distribution of shared polymorphisms in the organellar genomes is, since GWA studies rely on linkage having been broken up by recombination.) Such an analysis would greatly elevate the study and the authors are strongly encouraged to consider this very seriously.</p><p>Overall, the clarity of the manuscript should be improved considerably. The text is replete with jargon and technical terms that are not properly introduced. That the models used are not always clearly stated also contributes to the paper being hard to read. The main model(s) need to be in the main text, so it is clear where the results come from.</p></body></sub-article><sub-article article-type="reply" id="SA2"><front-stub><article-id pub-id-type="doi">10.7554/eLife.00776.049</article-id><title-group><article-title>Author response</article-title></title-group></front-stub><body><p><italic>Starting with the title, “control” is clearly too strong</italic>.</p><p>We have changed the title to reflect influence rather than control. Additionally, we have worked to better caveat the work throughout to reflect the link to quantitative analysis of organelles in a genomic phenotype.</p><p><italic>First, the authors should spell out what fraction of the genetic variation is attributable to cytoplasmic variants, what fraction is due to C×G interactions, and how does this compare to straight G effects</italic>?</p><p>We have worked to make this clearer by dissecting all variance components into the individual main effect (nuclear v cytoplasmic) and epistatic (cytonuclear v nuclear x nuclear) and included this in the description to help with the interpretation of our results. This is also specifically shown in Figure 6–figure supplement 1 where we parse genetic variation to each specific term of main effect or interaction for all metabolites.</p><p><italic>Second, how representative the findings of the paper are of other phenotypes remains unknown, and the authors should be more circumspect when extending their claims to phenotypes that might have no or only a very indirect connection to metabolites</italic>.</p><p>We have worked to better lay out that while this manuscript focuses on metabolomic variation, further work is required to test how this extends to other phenotypes. As a respectful point, we find it hard to imagine a phenotype that has no, or even an indirect connection to metabolites given the fundamental basis that metabolism plays in providing the energy and structure for any other phenotype to properly occur.</p><p><italic>Furthermore, both reviewers found the paper hard to followthe presentation needs to be greatly simplified, so that the important points come across better</italic>.</p><p>We have worked to improve the writing and hope that this has been achieved.</p><p><italic>Specific criticisms</italic>:</p><p><italic>1) The authors’ explanation for the average heritability due to the organellar genome being much lower than the one coming from the nuclear genomes seems reasonable. However, both reviewers had a hard time trying to understand why there are apparently no metabolites that show cytoplasmic heritability greater than 0.15. Two different models were used to estimate nuclear and cytoplasmic heritability, but why? Along these lines, how do cytoplasmic and nuclear heritability compare for the metabolites significant for both? And on a related note, please explain why significant effects of the nuclear genome were estimated by partitioning the variance across 316 RIL genomes. These are not independent: surely you want the estimates produced by linkage mapping? Finally, that the complete cytoplasmic genome has a larger effect than individual nuclear loci (</italic><xref ref-type="fig" rid="fig6"><italic>Figure 6</italic></xref><italic>) does not seem surprising, given that the mtDNA and cpDNA represent contain dozens of genes</italic>.</p><p>We are slightly confused by the first question as the analysis in question used a single model to estimate nuclear and cytoplasmic heritability that is shown in <xref ref-type="supplementary-material" rid="SD1-data">Figure 1—source data 1</xref>. This model uses the standard approach to estimate heritability in a RIL population where you use the proportion of line variation of the total variation to estimate genetic reproducibility. In a non-reciprocal RIL population the line variation accounts for the variance across all the lines in the population. We did this classical approach to provide a common entry point to the community into this analysis as well as a common linkage to most previous RIL analysis. We have also introduced a description saying that the second model is asking a different question than the per line heritability question asked by the model in <xref ref-type="supplementary-material" rid="SD1-data">Figure 1—source data 1</xref>.</p><p>We are unsure as to why no metabolites had a cytoplasmic heritability above 0.15 using this analysis but this is what the data showed and by doing the classical approach we can now show that this number is misleading based upon the ensuing linkage mapping. To also help with the analysis, we have provided locus and interaction specific estimates of heritability (per locus/interaction R2) that is now discussed in the section on epistatic interactions. This analysis is shown in Figure 6–figure supplement 1. This analysis shows that if we parse the nuclear genetic variation into individual nuclear hotspot variation, the cytoplasmic heritability is actually greater on average than any specific nuclear hotspot.</p><p>Finally, we have also mentioned in the Discussion that the organellar genomes have roughly the same number of genes as an individual nuclear locus because given the genetic resolution of recombination these loci still represent a region of &gt;100 genes.</p><p><italic>2) A substantial criticism is that the authors do not tell us anything about sequence differences in the organellar genomes of the RIL parents. The short read data are available, so it would not be too much of a burden to determine how many coding variants there are, and how the distribution of coding variants compares between biosynthetic genes in the organellar and nuclear genomes. We note that the Discussion states that differences have been found in the organellar genomes</italic>.</p><p>We have obtained suitable short reads and provided a list of polymorphisms between the Kas and Tsu mitochondria and organelle as Figure 7–source data 1. This has been added into the section on cytoplasmic effects along with a commentary on how the changes are focused on NADH related metabolic genes, which would explain the widespread consequences in the metabolic grid of the organellar genetic variation because NADH is central to nitrate assimilation, the GOGAT cycle, lipid metabolism, and any process involving a cytochrome P450.</p><p><italic>3) You discuss whether RILs have less phenotypic variation because they have less genetic variation. You write “A common concern of RIL populations is that they only sample two alleles per locus and thus might be limited in their genetic variation relative to species-wide diversity (Kim et al., 2007; Nordborg and Weigel, 2008).” This may or may not be common concern, but it is not the chief one, in our opinion, and it is also not the point made in the papers cited. As far as one can tell, Kim et al. (2007) does not discuss this issue at all, and Nordborg and Weigel (2008) make a completely different point, namely that RILs have greater power to detect rare alleles of large effect that “may not be important from an evolutionary perspective”. This is central to your paper because of the very real possibility that the observed QTL hotspots (which mostly differ from another RIL population) are due to rare alleles of large effect, which may well be deleterious. This does not invalidate the results, but changes their interpretation, in particular with respect to their evolutionary significance</italic>.</p><p>We have worked to better clarify this section to reflect both the power and rarity concern. The specific paragraph in question now attempts to show that while the RIL population may be susceptible to the rare allele issues, we have similar if not greater phenotypic variation than the accessions, meaning that we actually have better power to detect loci and interactions.</p><p><italic>4) There is little consideration of existing knowledge of the contribution of organellar biosynthesis steps to metabolic pathways. An important question is whether metabolites that are known to be made in organelles are more likely to be affected by variation in the organellar genomes than those that are not</italic>.</p><p>We have added in a commentary in the first results section on the organellar variation of how most metabolites, even those specifically synthesized in a single compartment, are then transported throughout the plant cell even from one organelle to the other. Additionally, a number of metabolites are made in multiple compartments including the organelles and the cytosol, as such ascribing individual compartmental function is not as straightforward as is required for a hypergeometric test or the like. Additionally, our extraction protocol cannot differentiate between the different pools within a cell. Given that nearly all primary metabolites are influenced by the organellar genomic variation this analysis doesn’t have the power to refine the mechanistic basis of this change.</p><p><italic>Finally, the authors make the important point that the organellar genotypes would add greatly to GWA studies. We agree. Since the authors have metabolome data for 96 accessions (</italic><xref ref-type="bibr" rid="bib13"><italic>Chan et al., 2010</italic></xref><italic>), one wonders how much effort it would be to download existing read data, identify polymorphisms in the mitochondrial and chloroplast genomes, and include them in the GWA analysis. (A major unknown is what the distribution of shared polymorphisms in the organellar genomes is, since GWA studies rely on linkage having been broken up by recombination.) Such an analysis would greatly elevate the study and the authors are strongly encouraged to consider this very seriously</italic>.</p><p>We agree that this would be nice, but for several reasons we feel that this is more appropriate for a follow-up study. The first is technical in that the available short read data were largely made with protocols to enrich the nuclear genomic DNA at the cost of the organellar sequences and as such, we had to identify specific Kas and Tsu samples prepped with a different protocol. The other available short read archive sequences did not provide us sufficient coverage on the organellar genomes to call polymorphisms. The second reason is genetic in that all published literature on the Arabidopsis organellar genetic variation using targeted gene analysis shows that the organelles do not show any evidence of independent assortment and instead they behave in a more clonal fashion which greatly complicates any potential for GWA as the entire organelles are essentially within complete LD. The final reason is for clarity, as the reviewers have rightly noted, we need to improve the manuscript’s clarity and adding a completely additional section on accessions and the organelles would require the equivalent of a full manuscript’s worth of analysis to assess the link between the organelle and population structure, LD, and then finally GWA with metabolites. Thus, we feel that this request is an excellent foundation for another complete manuscript, but it would not help to clarify this manuscript mechanistically, genetically, or editorially.</p><p><italic>Overall, the clarity of the manuscript should be improved considerably. The text is replete with jargon and technical terms that are not properly introduced. That the models used are not always clearly stated also contributes to the paper being hard to read. The main model(s) need to be in the main text, so it is clear where the results come from</italic>.</p><p>We have worked throughout to better clarify the text and more explicitly link to the proper model.</p><p>[Editors’ note: after re-review, the authors edited the article further to clarify the primary message and removed sections of the article relating to comparative analysis to other populations.]</p></body></sub-article></article>