Cytonuclear interactions affect adaptive traits of the annual plant Arabidopsis thaliana in the field
- aLaboratoire Evolution, Ecologie et Paléontologie, UMR CNRS 8198, Université de Lille, F-59655 Villeneuve d’Ascq Cedex, France;
- bInstitut National de la Recherche Agronomique (INRA), Laboratoire des Interactions Plantes-Microorganismes, UMR441, F-31326 Castanet-Tolosan, France;
- cCNRS, Laboratoire des Interactions Plantes-Microorganismes, UMR2594, F-31326 Castanet-Tolosan, France;
- dGénétique Quantitative et Evolution–Le Moulon, INRA, Université Paris-Sud, CNRS, AgroParisTech, Université Paris-Saclay, 91190 Gif-sur-Yvette, France;
- eMathematiques et Informatique Appliquées, AgroParisTech, INRA, Université Paris-Saclay, 75005 Paris, France;
- fInstitut Jean-Pierre Bourgin, INRA, AgroParisTech, CNRS, Université Paris-Saclay, F-78026 Versailles, France;
- gInstitute of Plant Sciences Paris-Saclay (IPS2), CNRS, INRA, Université Paris-Sud, Université Evry, Université Paris-Saclay, Batiment 630, F-91405 Orsay, France;
- hInstitute of Plant Sciences Paris-Saclay IPS2, Paris Diderot, Sorbonne Paris-Cité, Bâtiment 630, F-91405 Orsay, France
See allHide authors and affiliations
Edited by Maarten Koornneef, Max Planck Institute for Plant Breeding Research, Cologne, Germany, and approved February 18, 2016 (received for review October 20, 2015)

Significance
As the centers of photosynthesis and respiration, chloroplasts and mitochondria play a crucial role in energy metabolism. Nuclear and cytoplasmic genomes are known to be coadapted at the species level, because organelle metabolism relies on the proper interaction of organelle-encoded and nuclear-encoded proteins. We explored the extent of cytonuclear coadaptation at the intraspecific level in the classic model plant Arabidopsis thaliana: we measured in a field experiment 28 adaptive whole-organism traits on cytolines developed by substituting cytoplasmic genomes among natural strains. Our results indicate that interactions between nuclear and cytoplasmic genomes shape natural variation for most of the traits we studied, suggesting that these interactions can affect the evolutionary dynamics of natural populations of A. thaliana.
Abstract
Although the contribution of cytonuclear interactions to plant fitness variation is relatively well documented at the interspecific level, the prevalence of cytonuclear interactions at the intraspecific level remains poorly investigated. In this study, we set up a field experiment to explore the range of effects that cytonuclear interactions have on fitness-related traits in Arabidopsis thaliana. To do so, we created a unique series of 56 cytolines resulting from cytoplasmic substitutions among eight natural accessions reflecting within-species genetic diversity. An assessment of these cytolines and their parental lines scored for 28 adaptive whole-organism phenotypes showed that a large proportion of phenotypic traits (23 of 28) were affected by cytonuclear interactions. The effects of these interactions varied from slight but frequent across cytolines to strong in some specific parental pairs. Two parental pairs accounted for half of the significant pairwise interactions. In one parental pair, Ct-1/Sha, we observed symmetrical phenotypic responses between the two nuclear backgrounds when combined with specific cytoplasms, suggesting nuclear differentiation at loci involved in cytonuclear epistasis. In contrast, asymmetrical phenotypic responses were observed in another parental pair, Cvi-0/Sha. In the Cvi-0 nuclear background, fecundity and phenology-related traits were strongly affected by the Sha cytoplasm, leading to a modified reproductive strategy without penalizing total seed production. These results indicate that natural variation in cytoplasmic and nuclear genomes interact to shape integrative traits that contribute to adaptation, thereby suggesting that cytonuclear interactions can play a major role in the evolutionary dynamics of A. thaliana.
The genomes of eukaryotes originate from ancient endosymbiotic associations that eventually led to energy-harnessing organelles: mitochondria, common to all eukaryotes, and chloroplasts in the “green” lineage. The evolution of endosymbionts into cellular organelles was accompanied by massive gene loss, with a large proportion being transferred to the nucleus (1, 2). Nevertheless, mitochondria and chloroplasts retained a few (30–80) protein-encoding genes that play crucial roles in energy metabolism (respiration and photosynthesis). Mitochondrion and chloroplast metabolisms rely on the proper interaction of nuclear-encoded proteins and their counterparts encoded in the organelle genome. Consequently, the genes in nuclear and organellar compartments are expected to be coadapted (3).
Cytonuclear coadaptation has been demonstrated by altered phenotypes observed on interspecific exchanges of cytoplasm between related species in mammals (4), yeast (5), arthropods (6), and plants, whose interspecific crosses are frequently successful (7). These alterations affect organelle function and even the organism phenotype, indicating epistasis between nuclear and cytoplasmic genes. Although cytonuclear coadaptation is generally studied at the interspecific level, the existence of intraspecific genetic diversity in organelle genomes suggests a potential for genomic coadaptation within species. A few studies have reported phenotypic effects of intraspecific cytonuclear epistasis in nonplant species (8⇓⇓–11). In plants, many studies have focused on cytoplasmic male sterility (CMS), an impairment of pollen production governed by nucleo-mitochondrial interactions in some hermaphroditic species (12), in particular in crops and their relatives (13). The phenotypic effects of intraspecific cytonuclear epistasis other than CMS have been reported in only a limited number of plant systems (14⇓⇓–17), with evidence that cytoplasmic variation contributes to local adaptation (18, 19).
In recent years, several studies using reciprocal segregating populations of the model plant Arabidopsis thaliana have investigated the effect of cytonuclear epistasis on a number of laboratory-measured phenotypes such as the metabolome, defense chemistry and growth (17, 20, 21), water-use efficiency (22, 23), and seed germination (24, 25). Although some studies have reported significant effects of cytonuclear epistasis (17, 20, 21, 23, 25), others have found additive cytoplasmic effects but with weak or no cytonuclear epistasis (22). Each of these studies (with the exception of ref. 25) was, however, based on a single reciprocal cross between two natural accessions, thereby preventing the estimation of the prevalence of cytonuclear epistasis in this species. In addition, although these reports involve adaptive traits (26⇓⇓⇓–30), the investigation of the effect of cytonuclear epistasis on adaptive phenotypes in field conditions is, at best, scarce in A. thaliana.
Here, following the modern standards of ecological genomics (31), we explored the prevalence of cytonuclear interactions on adaptive whole-organism traits in the model plant A. thaliana in a field experiment. To do so, based on eight natural accessions of a core collection that covers a significant part of the species’ cytoplasmic and nuclear genetic diversity in A. thaliana (25, 32), we created eight series of seven cytolines. Cytolines are genotypes that combine the nuclear genome from one parent with the organelle genomes of another (33). We examined the cytolines and their parental accessions for effects of cytonuclear interactions on 28 field-measured traits related to germination, phenology, resource acquisition, plant architecture and seed dispersal, fecundity, and survival.
Results
To test the prevalence of cytonuclear interactions on fitness-relevant traits, we constructed relevant genetic resources in A. thaliana, hereafter named cytolines, by exchanging nuclear and cytoplasmic genomes between eight natural accessions, namely Blh-1, Bur-0, Ct-1, Cvi-0, Ita-0, Jea, Oy-0, and Sha (Fig. 1, SI Text, Fig. S1, and Table S1). These eight accessions amply cover the nuclear and cytoplasmic diversity found in A. thaliana (25, 32). Hereafter, a cytoline possessing the cytoplasm of accession A and the nucleus of accession B is designated as AcyBnuc. The genetic composition of the 56 cytolines was verified by genotyping (SI Text). Given the dense genotyping strategy used, the genomic regions affected by residual heterozygosity, if any, should be very limited in size. Bulks of seeds of each cytoline and parental accession were produced in controlled conditions to reduce maternal effects (SI Text, Fig. S2, and Table S2), and then used for conducting a phenotyping experiment in a field in northern France (Materials and Methods and SI Text), where adaptation based on nuclear genetic variation has been previously detected (28, 30). Each line was scored for a total of 28 phenotypic traits described as adaptive in A. thaliana (34⇓⇓–37) in a randomized complete block design (SI Text, Fig. S3, and Dataset S1). These phenotypic traits belong to six phenotype classes: germination, phenology, resource acquisition, architecture and seed dispersal, fecundity, and survival. Each trait was modeled separately using a mixed model. Depending on traits, the error variance was chosen to be either homogeneous or nucleus (i.e., heterogeneous) dependent (Table 1 and SI Text).
General strategy for the production of the cytolines. Diallele crosses were performed between eight natural accessions, and three backcrosses were performed with the male parent (nuclear donor). After dense genotyping, a plant with a fixed allele from the recurrent male parent at all markers was used as the cytoline founding mother. Details on the production and genotyping of cytolines are given in SI Text.
Global effects of nucleus, cytoplasm, and cytonuclear interactions on phenotype
Genetic composition of the plants throughout the cytoline construction. The 10 bars represent the five chromosome pairs of A. thaliana, with colors illustrating allelic origin (white: accession A, black: accession B). The background color of the rectangle stands for the origin of cytoplasm (orange: accession A, blue: accession B). After at least three backcrosses and genotyping, symbolized by the inversed triangle, the cytoline obtained possess the nuclear genome of parent B in the cytoplasmic background of parent A.
Identification numbers and genomic composition of cytolines
Experimental design for seed production. Each of the eight tables present in the growth chamber was divided in two parts. One plant per genotype was placed in each half of table. The positions were randomized in each block (colors). Genotypes are designed according to their accession name in the Versailles stock center (AV suffix for parental accessions, CV suffix for cytolines; Table S1). Empty positions were left empty or used to grow seeds from sister plants of some cytolines (not used in this study).
Postharvest storage of seeds used in the phenotyping experiment
Measure of rosette surface area (AREA). Rosette surface area (AREA) was measured using a nondestructive approach, by imaging each tray 28 d after sowing, and using a Canon digital camera (model EOS 500D). Each image of an array was submitted to a four-step treatment. In the first step, a photograph of each array was taken in the field. In the second step, the raw photographs were centered, detrapezoided, and resized using the software Adobe Photoshop CS3 Extented (version 10.0) and the software ImageMagick (version 7:6.6.2.6–1ubuntu4.2; Available from www.imagemagick.org/script/index.php). In the third step, plants were then manually selected from background using the software GIMP 2.8 (Available from www.gimp.org/). Background-free images were then cutted using a custom Perl script to obtain an individual background-free image for each plant of the experiment. In the fourth step, rosette surface area and rosette perimeter of each plant were then estimated using the ImageJ software (version 4.01; Universal Imaging). Rosette surface area was automatically estimated for each plant on pretreated photographs using the Area Set Measurement command, which first estimates the number of pixels defined by an object and then converts this number into a metric value of surface (cm2 in this study). Rosette perimeter was automatically estimated for each plant on pretreated photographs using the Perimeter Set Measurement command, which estimates the length of the outside boundary of the rosette image.
In agreement with the genetic diversity captured by the core collection of A. thaliana used in this study (32), all measured traits varied significantly across nuclear backgrounds (Table 1). The mean effect of the cytoplasm was significant for 9 of the 28 traits and was detected in each phenotype class, with the exception of fecundity (Table 1). More importantly, significant cytonuclear interactions were observed for most phenotypic traits (i.e., 23 of 28 traits) and for each phenotype class (Table 1). These observations are illustrated in Fig. 2 for each phenotype class with a representative trait showing significant nucleus, cytoplasm, and cytonuclear interaction effects (Table 1).
Cytoplasm origin has a variable influence according to the considered trait and nuclear host genome. For each trait, least-squares means (LSMs) are plotted by cytoplasm parent (horizontal axis). LSM values of cytolines with the same nucleus share the same symbol and are connected across cytoplasms. Values for the parental accessions are indicated in color. For better visual distinction, lines with a Blh-1, Bur-0, Ct-1, or Cvi-0 nucleus are plotted in the left panels, whereas lines with a Ita-0, Jea, Oy-0, or Sha nucleus are plotted in the right panels. (A) Percentage of germination 13 d after sowing (das). (B) Rosette surface area 28 d after sowing. (C) Bolting time. (D) Percentage of plants that completed their life cycle. Values were log-transformed for clearer visualization of the results. (E) Maximum height of the plant. (F) Percentage of aborted fruits. Values were log-transformed for a clearer visualization of the results. Because many plants of the lines with the Ita-0 nucleus died prematurely, postflowering traits were not analyzed for these lines.
Although no significant cytonuclear interaction was detected for total seed production, other individual traits related to fecundity (such as the number of seeds per fruit or the percentage of seeds produced on different types of branches) were significantly affected by cytonuclear interactions (Table 1). This observation suggests that cytolines of a given nuclear background produced the same number of seeds, but with contrasting reproductive strategies.
After testing for pairwise cytonuclear interactions in each parental pair for all quantitative traits (Table S3), several main features were discerned. First, we observed a large range of phenotypic effects among the 22 quantitative traits globally influenced by cytonuclear interactions. Although a significant global effect for cytonuclear interactions was detected for the three traits related to resource acquisition (surface area and perimeter of the rosette 28 d after sowing and diameter of the rosette at flowering) and for flowering interval (Table 1), no significant effect for pairwise cytonuclear interaction was detected for these traits in any specific parental pair (Table S3). Therefore, for these traits, cytoplasm substitution commonly led to phenotypic effects of minor intensity. In contrast, the significant global cytonuclear interactions detected for the remaining traits were mainly driven by pairwise cytonuclear interactions in one to four parental pairs, with the parental pairs Ct-1/Sha and Cvi-0/Sha accounting for 14 of the 28 observed significant pairwise cytonuclear interactions (Table S3). Second, significant asymmetrical phenotypic effects were observed between reciprocal cytoplasmic substitutions. For instance, the cytoline Ct-1cyShanuc bolted earlier than its parental accession Sha, whereas the cytoline ShacyCt-1nuc bolted at the same time as its parental accession Ct-1 (Fig. 3A). Asymmetrical effects of cytoplasm exchange were also observed for the pair Cvi-0/Sha. No differences were observed for any phenotypic trait scored in this study between the cytoline Cvi-0cyShanuc and its parental accession Sha. In contrast, in comparison with the parental accession Cvi-0, plants of the cytoline ShacyCvi-0nuc showed low fertility with a fruit abortion rate of 43.6% (Figs. 2F and 3B) and reduced fruit length in nonaborted fruits on the main stem (Fig. 3C), along with a longer reproductive period (Fig. 3D) and a greater maximum height (Fig. 3E). Finally, in some traits and for specific parental pairs, nuclear genotypes had contrasting effects when combined with the same cytoplasm (Fig. 4). For instance, Sha and Ct-1 nuclei had opposite and symmetrical effects on the percentage of seeds produced on the main stem (Fig. 4A) and on the height from the soil to the first fruit on the main stem (Fig. 4B) in their reciprocal cytoplasms (as shown by significant contrast for these traits; Table S3), but also in other cytoplasms, such as Ita-0.
Asymmetrical cytonuclear interactions observed in the Ct-1/Sha and Cvi-0/Sha pairs. Box plots of fitted values are presented for all genotypes with the two nuclei involved in the significant cytonuclear interaction. Values are plotted by cytoplasm donor in each group of genotypes sharing the same nucleus, color coded according to the nuclear donor. The color code is identical for the cytoplasm and the nucleus. (A) Bolting time for the Ct-1/Sha pair. (B) Percentage of aborted fruits for the Cvi-0/Sha pair. (C) Mean fruit length of fertilized fruits on the main stem for the Cvi-0/Sha pair. (D) Reproductive period for the Cvi-0/Sha pair. (E) Maximum height of the plant for the Cvi-0/Sha pair.
Symmetrical and opposite effects of Ct-1/Sha nuclear alleles depend on the cytoplasm they are combined with. LSMs are plotted for genotypes carrying either a Ct-1 (blue) or a Sha (red) nucleus, according to their cytoplasm parent (horizontal axis). (A) Percentage of seeds produced on the main stem. (B) Height from soil to the first fruit on the main stem. Both traits influence seed dispersal and both are significantly affected by the cytonuclear interactions in the Ct-1/Sha pair.
Cytonuclear interactions in specific pairs of parents
Discussion
We constructed a series of 56 intraspecific cytolines that substitute cytoplasms among eight accessions of the model species A. thaliana. This genetic material constitutes a valuable resource for further exploration of the effects and the prevalence of natural organelle genome variation and cytonuclear interactions on phenotypic adaptive traits in A. thaliana.
Alloplasmic lines resulting from cytoplasm exchanges at the interspecific level often demonstrate severely affected phenotypes due to incompatibilities accumulated in the genetic compartments of species of long-diverged lineages (3, 7). Our intraspecific cytolines allow exploration of both (i) the nonneutral cytoplasmic diversity at the intraspecific level and (ii) the proportion of intraspecific genetic diversity that is involved in the interaction between the organelle and nuclear genomes. This genetic resource is available to the scientific community through the Versailles Arabidopsis Stock Center (publiclines.versailles.inra.fr/), thereby facilitating its widespread use for analysis of cytonuclear interactions at all possible phenotypic scales. The diallele design underlying the cytolines described in this study provides access to the cytonuclear effects between all possible pairs of parents and potentially to their third-order interactions with the environment (11, 38). Cytonuclear interactions in A. thaliana have been also reported using reciprocal segregating populations (17, 39). Both types of genetic resources are complementary and valuable for testing the cytonuclear epistasis underlying the studied traits. Although segregating populations can help detect cytonuclear interactions that involve two or more nuclear loci in opposition to each other (15) (a feature that is clearly overlooked in cytolines), subtle phenotypic effects resulting from cytonuclear interactions can be masked due to transmission ratio distortion and interparental allelic epistasis (15), both of which are absent in cytolines.
In this study, in agreement with previous studies based on intraspecific cytolines in insects and yeast (9, 11), a greater proportion of phenotypic variance is likely explained by cytonuclear interactions than cytoplasmic effects alone. In addition, as previously observed in laboratory/greenhouse conditions (17, 20, 21, 23, 25), cytonuclear interactions in A. thaliana can affect a large proportion of adaptive whole-organism traits (>80%; Table 1) in field conditions. This observation suggests that variation in organelle function affects a large range of integrative traits, not only in controlled conditions, but also in more complex and ecologically realistic environments.
In other plant species, phenotypic evaluations of intraspecific cytolines are, at best, scarce. In maize, widespread phenotype effects of cytonuclear interactions have been observed in intrageneric, interspecific alloplasmic lines, but intraspecific cytolines based on cytoplasm exchange between subspecies of Zea mays are generally phenotypically indistinguishable from the parental cultivar (15). In that study, the absence of observable phenotypic effects of cytonuclear interactions at the intraspecific level may originate from the crossing design and the limited number of tested nuclei (one nucleus–seven Z. mays cytoplasm donors). In the present study, the number of parental lines, their coverage of natural genetic diversity, and the diallele crossing design not only revealed extensive effects of cytonuclear interactions in A. thaliana, but also led to observations that would have been overlooked in a specific, unique parental pair.
For instance, among the significant pairwise cytonuclear interactions observed for specific parental pairs, both asymmetrical and symmetrical responses were observed. Asymmetrical responses are observed when one cytoline has a clear differentiated phenotype, whereas the reciprocal cytoline behaves similarly to its nuclear parent. In addition to being a characteristic feature of CMS, asymmetrical responses to reciprocal cytonuclear exchange have been previously reported in plants, including at the interspecific level (7). Symmetrical responses have also been observed in the reciprocal exchange of genetic compartments, with nonparental combinations showing impaired phenotypes compared with the parental lines, e.g., longevity in intraspecific seed beetle (Acanthoscelides obtectus) cytolines (40) and fitness breakdown in reciprocal F2s of a copepod (Tigriopus californicus) (41, 42). In this latter case, nuclear-encoded cytochrome c and mitochondria-encoded subunits of cytochrome c oxidase have diverged between the two parental populations such that the interaction between the mismatched partners impairs complex IV activity (43). This example illustrates that allelic differences in nuclear genes coding for organellar proteins may have different outputs when their product interacts with organelle partners of different origin. Interestingly, the Ct-1/Sha parental pair illustrates both types of phenotypic responses (Figs. 3 and 4) and this combination affected more measured traits (8 of 28) than any other parental pair. In addition to bolting time, the other affected traits (Table S3) are all assumed to contribute to seed dispersal (35). A clear asymmetrical response was identified for bolting time (Fig. 3). In contrast, opposite symmetrical effects of the Sha and Ct-1 nuclear genomes were observed for the height from the soil to the first fruit on the main stem and for the percentage of seeds produced on the main stem: in a given cytoplasm, the phenotype of plants with the Ct-1 nucleus mirror those with the Sha nucleus (Fig. 4). This pattern may reflect nuclear polymorphisms that affect interactions between nuclear and organellar gene products. The next challenges will be to identify the genetic factors involved in these interactions and decipher the pathway from their molecular interaction to the whole-organism integrative traits (44, 45). This line of research will undoubtedly benefit from the vast information available on genetic polymorphisms for all but one (Ita-0) of the parental accessions used (1001 Genomes Project) (46).
The relatively high number of parental lines used in this study also allowed the identification of pairs of parents showing remarkable effects of disrupted cytonuclear coadaptation. For instance, the Cvi-0/Sha pair accounted for 6 of the 28 significant specific interactions (Table S3), all due to ShacyCvi-0nuc cytoline behavior. The Sha cytoplasm can induce CMS in other natural accessions (47); likewise, the ShacyCvi-0nuc cytoline is male-sterile, producing no or very few seeds in laboratory conditions (i.e., greenhouse and growth chamber), so that hand-pollination was necessary to produce the seeds used in this work (SI Text). Surprisingly, in our field experiment, the ShacyCvi-0nuc plants produced the same amount of seeds as Cvi-0 plants. Cross-pollination by neighboring plants is unlikely to account for their seed production because plants of the ShacyCvi-0nuc cytoline grown in the same common garden at the same period, but protected from foreign pollen by plastic tubes, also set seeds (Fig. S4). Although the total seed production of ShacyCvi-0nuc and Cvi-0 plants was comparable, the former had reduced numbers of seed per fruit and a higher percentage of fruit abortion, consistent with what is observed in laboratory growth conditions. However, they produced more flowers than Cvi-0 plants, due to their greater plant height and longer reproductive period, which compensated for their poor fertility. Hence, we predict that a cytoplasmic variant with a phenotype similar to that observed in the ShacyCvi-0nuc line has a much higher potential to produce progeny by outcrossing, without completely relying on cross-pollination for seed set. This compensatory system may have a significant impact on population adaptive response by transiently modifying the outcrossing rate (48) before the selection and eventual fixation of nuclear restorer(s) of fertility (49, 50). Consequently, the relative contribution of CMS to variation in the outcrossing rate among natural populations (e.g., from 0% to 20% in A. thaliana) (51) may be underestimated in highly selfing species.
Seed production of cytolines in a field located at the University of Lille. An experiment of 120 A. thaliana plants was organized in two blocks, each one being represented by 60 pots (9 × 9 × 9.5 cm, vol. ∼480 cm3; TEKU MQC) filled with damp standard culture soil (Huminsubstrat N3; Neuhaus). Each block is an independent randomization of 60 plants with one replicate per cytoline (n = 52) and one replicate per parental accession (n = 8). Experimental conditions in the frost-free greenhouse until the transport of pots outside to the field were similar to the conditions for the main experiment described in SI Text. (A) Overview of the experiment. (B) Close view of a ShacyCvi-0nuc plant protected by a plastic tube to avoid out pollination from neighboring conspecifics.
The next step to address the adaptive significance of cytonuclear interactions will involve the phenotyping of the cytolines in the native habitat of their parental accessions. In addition, because local populations of A. thaliana have shown substantial cytoplasmic polymorphism with the occurrence in the same locality of several cytotypes distributed across Eurasia (SI Text, Dataset S2, and Fig. S5) (52, 53), studying the evolutionary dynamics of natural populations of A. thaliana will benefit from considering the contribution of cytonuclear interactions.
Network of cytotypes found in the local metapopulation TOU. Each described cytotype is represented by a circle whose size is proportional to the number of individuals observed for this cytotype. Black dots represent hypothetical intermediates cytotypes that have not been observed in this study. The two red segments represent identical polymorphisms, whose distribution could lead to a reticulation of the network. Each segment between circles or dots represents one chloroplast or mitochondrial polymorphism. Capital red letters stand for cytotypes that were previously observed in ref. 2.
Materials and Methods
Plant Material: Creation and Seed Production of Cytolines.
A complete diallele cross was carried out between the eight selected natural accessions, followed by three backcrosses with the male parent and dense genotyping (Fig. 1 and Fig. S1). A detailed procedure of the production and genotyping of cytolines is given in SI Text. The genetic resources and the complete list of the genotyped markers for each cytoline are available on the Versailles Arabidopsis Stock Center website (publiclines.versailles.inra.fr/).
Field Experiment, Phenotype Characterization, and Data Analysis.
A field experiment of 2,700 A. thaliana plants was set up at the University of Lille 1 (northern France) following a randomized complete block design. Each plant was scored for a total of 28 phenotypic traits related to germination (n = 5), phenology (n = 4), resource acquisition (n = 3), architecture and seed dispersal (n = 5), fecundity (n = 10), and survival (n = 1). All traits were measured quantitatively with the exception of survival, which is a binary trait. Details of the field experiment and phenotype characterization are provided in SI Text.
Each trait was modeled using a mixed model described in detail in SI Text. In a first step, we assessed the impact of each factor in the model. To account for multiple testing, a Benjamini–Hochberg procedure (54) was performed within each term of the model, across the tested phenotypes to control false discovery rate (FDR) at nominal level 5%. In a second step, we conducted pairwise comparisons within the mixed model to identify pairs of parents that significantly contribute to cytonuclear interactions for each quantitative trait (SI Text). A global Benjamini–Hochberg adjustment of the P values was performed across pairs and traits to control FDR (nominal level: 5%).
SI Text
Construction and Genotyping of Cytolines.
The construction of the 56 cytolines is illustrated in Fig. 1. The genetic composition of the plants throughout the cytoline construction is illustrated in Fig. S1. Three recurrent backcrosses with the nuclear donor parent were realized on 54 from the 56 possible F1s of the diallele cross. F1 seeds from reciprocal crosses between Bur-0 and Ita-0 did not germinate whatever the conditions tested. To obtain the two concerned cytolines, we started backcrosses with the nuclear donor on bridge genotypes. Specifically, a plant from the cross Bur-0 × [Jea × Ita-0] F1 was backcrossed with Ita-0 for the Bur-0cyIta-0nuc combination; a plant from the cross [Ita-0 × Jea] F1 × Bur-0 was back-crossed with Bur-0 for the Ita-0cyBur-0nuc combination. Consequently, the genotyping of these cytolines was designed to discriminate Bur-0, Ita-0 and Jea alleles.
For each combination, 29 plants from the third backcross were genotyped with a set of 384 SNP markers (55), among which 134 on average were informative, according to the considered combination. To keep the distance between two genotyped positions below 3 Mb, and the distance between the last and first genotyped positions and the telomeres below 1 Mb, microsatellites were used in those intervals too large between informative SNPs. Among the 56 cytonuclear combinations, 42 cytolines were obtained at this stage. For the 14 remaining combinations, 24 or 48 plants from the selfing descent of plants chosen on their genotype were grown and genotyped at the position(s) were the mother plant was heterozygous to select the corresponding cytoline among the progenies fixed with the nuclear donor allele. The rule was sometimes relaxed in the centromeric regions where polymorphic markers were difficult to find and meiotic recombination is known to be rare. The number of markers used for each cytoline is available in Table S1. The complete list of the markers, with their positions in the genome, are available on the Versailles Arabidopsis Stock Center website (publiclines.versailles.inra.fr/).
The cytoplasm of cytolines was verified by sequencing intergenic chloroplastic regions as previously described (25).
Seed Production of Cytolines.
Seeds of the 56 cytolines and their eight parental accessions were produced in a large growth chamber (56 m2, light 16 h at 21 °C, dark 8 h at 18 °C). Sixteen plants of each genotype were grown. The plants were placed according to an experimental scheme designed to randomize the environmental heterogeneities in the chamber, known to be mainly due to border effects. One hundred twenty-eight plants were disposed on each of the eight tables in the chamber (Fig. S2). Plants were sown by groups of genotypes sharing the same nucleus. Seeds of the male sterile cytoline 39CV (ShacyCvi-0nuc) (47) were produced by hand pollination with pollen of the surrounding Cvi-0 plants. At the end of their life cycle (i.e., when siliques started to dry), plants were placed in a drying chamber and watering was stopped until complete drying of the plant. Plants were individually harvested and their seed production weighted. Bulks were made with equivalent amounts of seeds produced by each plant of a given genotype.
Four cytolines were still unsatisfactory at the start date of the field experiment, namely Blh-1cyShanuc, Bur-0cyIta-0nuc, Ct-1cyIta-0nuc, and Ct-1cyJeanuc, and were therefore not included in this study.
Field Experiment and Phenotypic Characterization.
An experiment of 2,700 A. thaliana plants was set up at the University of Lille 1 (North, France). The field experiment was organized in five blocks, each one being represented by nine arrays of 66 individual bottom-pierced wells (11 lines × 6 columns, Ø4 cm, vol. ∼38 cm3) (TEKU, JP 3050/66) filled with damp standard culture soil (Huminsubstrat N3; Neuhaus). Each block corresponded to an independent randomization of 540 plants with nine replicates per cytoline (n = 52) and nine replicates per parental accession (n = 8). In each block, the remaining 54 wells were left empty.
Five seeds were sown in each well on 11 March 2013 to mimic the spring seasonal germination cohort observed in natural populations of A. thaliana in the North of France. Germination was promoted by stratifying seeds four days at 4 °C in a cold chamber. After the stratification treatment, arrays were preventively treated against dark-winged fungus gnats (Vectobac; 8 mL/L) and placed for 28 d in a frost-free greenhouse that mimics outdoor conditions (no additional light or heating) but protects seeds from rainfall. To reduce microenvironmental variations, arrays were rotated daily in the cold chamber and in the greenhouse. Germination date and germination rate were monitored in the frost-free greenhouse during 13 d after the stratification treatment (see below). Wells were thinned to two seedlings and one seedling 14 and 20 d after the stratification treatment, respectively. Thereafter, and in the main text, the time after sowing is meant counted from the end of the stratification treatment.
Twenty-eight days after sowing, arrays were transported outside to a field located at the University of Lille 1. For each block, the nine arrays were organized according to a grid of three columns and three lines. Soil was tilled to allow arrays to be slightly buried, thereby facilitating root development. Plants were protected from herbivory by vertebrates and slugs as described in ref. 56.
Each plant was scored for a total of 28 phenotypic traits related to germination (n = 5), phenology (n = 4), resource acquisition (n = 3), architecture and seed dispersal (n = 5), fecundity (n = 10), and survival (n = 1) (Dataset S1):
Germination: Germination time (GERM) was measured as the number of days between sowing and the emergence of the first seedling. Using a phenological model integrating both photoperiod length and temperature (56), GERM was scaled in photothermal units (PTUs). Germination percentage was estimated 4, 5, 6, and 13 d after sowing (PGERM4, PGERM5, PGERM6, and PGERM13).
Phenology: Bolting time (BT), flowering interval (INT), and the reproductive period (RP) were scored as the time intervals between germination date and bolting date, between bolting date and flowering date, and between flowering date and date of maturation of the last fruit, respectively. BT, INT, and RP were scaled in PTU. By summing these three phenological traits, we estimated the length of the life cycle (LCYCLE).
Resource acquisition: rosette surface area (AREA expressed in cm2) and rosette perimeter (PERIM expressed in cm) were measured using a nondestructive approach 28 d after sowing (Fig. S3). At the start of flowering, the maximum diameter of the rosette measured at the nearest millimeter was used as a proxy for plant size (DIAM).
Architecture and seed dispersal: After maturation of the last fruit, the above-ground portion was harvested and stored at room temperature until further phenotyping. Plants were phenotyped for the following architectural and seed dispersal related traits: height from soil to the first fruit on the main stem (H1F), maximum height (HMAX), number of primary branches on the main stem (NPB), number of basal branches (NBB), and total number of branches (TOTB = NPB + NBB).
Fecundity: Because the number of seeds in a fruit is highly correlated with fruit length (56, 57), total seed production was approximated by total fruit length (FITTOT). Seed production is a good proxy for fecundity in a highly selfing annual species like A. thaliana (51). FITTOT was obtained by adding the fruit length produced on the main stem (FITSTEM), the primary branches on the main stem (FITPB), and the basal branches (FITBB). These estimates of fruit length were obtained by counting the number of fertilized fruits produced on each type of branches (FRUITSTEM, FRUITPB, and FRUITBB) and multiplying these counts by an estimate of their corresponding fruit (or silique) length (SILSTEM, SILPB and SILBB), estimated as the average of three representative fruits. We also calculated three ratios corresponding to the percentage of seeds produced by one branch type as a function of the total amount of seed produced: RSTEM = FITSTEM/FITTOT, RPB = FITPB/FITTOT, and RBB = FITBB/FITTOT. We also estimated the rate of fruit abortion (STERILITY) as the number of aborted fruits divided by the total number of fruits.
Survival: All plants that germinated but did not survive were counted as dead (SURVIVAL = 0). Harvested plants were counted as alive (SURVIVAL = 1).
Because all plants carrying a nucleus from Ita-0 were late flowering in this study, they were not able to complete their life cycle before summer heat. Consequently, postflowering traits were not measured on the Ita-0 parental accession, as well as on the cytolines with the Ita-0 nucleus.
Due to the absence of basal branches in most cytolines (Dataset S1), the traits related to basal branches (FITBB, FRUITBB, SILBB, and RBB) were not statistically analyzed in this study.
Data Analysis.
Model.
Each trait was modeled separately using the following mixed model (1):
As illustrated in Fig. 3, accounting for heterogeneous error variance was relevant for many traits, and significantly improved the accurate identification of parental pairs contributing to cytonuclear interactions. We also considered the inclusion of block × cytoplasm and block × cytoplasm × nucleus interactions in the model. However, adding these extra terms precluded the heterogeneous error variance assumption due to numerical instability [nonconvergence of the restricted maximum likelihood (ReML) procedure, inconsistent estimated effects, and/or degrees of freedom for parental pair contrasts]. Nonetheless, fitting the model with the additional interaction terms and homogeneous error variances for each trait led to results similar to those obtained with the mixed model (1): the same cytoplasm × nucleus interactions were detected across traits, except for the traits INT and FRUITPB (results not shown).
Because variation in rosette surface area and rosette perimeter may indirectly result from variation in germination time, the term GERM was also added as a covariate in the statistical model for the traits AREA and PERIM. All random effects were assumed to be Gaussian and independent, with a mean equal to 0. The line and column variances are
Inference was performed using ReML estimation, using the PROC MIXED procedure in SAS 9.1 (SAS Institute) for all traits with the exception of SURVIVAL, which was analyzed using the PROC GLIMMIX procedure in SAS 9.3. For all traits and genotypes, LSMs were computed.
Test for interactions in specific pairs of parents.
To identify pairs of parents that contribute to the nucleus × cytoplasm interaction, the hypothesis H0 {
Cytoplasmic Diversity in a Local Metapopulation of A. thaliana.
To evaluate cytoplasmic diversity in the local metapopulation TOU (37, 51), polymorphisms were analyzed (i) in four chloroplast intergenic regions (MatK-trnK, ndhC-trnV, rbcL-accD, ndhF-rpl32) for 24 individuals located in 11 stands (TOU-A1-41, TOU-C-2, TOU-D-5, TOU-E-7, TOU-F-1, TOU-I-2, TOU-P-6, TOU-Q-2, TOU-R-9, TOU-S-1, TOU-T-1, TOU-T-2, TOU-T-3, TOU-T-4, TOU-T-5, TOU-T-6, TOU-T-7, TOU-T-8, TOU-T-9, TOU-T-10, TOU-T-14, TOU-T-15, TOU-T-16, TOU-T-17) and (ii) in two mitochondrial regions (atp8-orf107c, ccmC) for 12 individuals (TOU-A1-41, TOU-C-2, TOU-D-5, TOU-E-7, TOU-F-1, TOU-I-2, TOU-P-6, TOU-Q-2, TOU-R-9, TOU-S-1, TOU-T-1, and TOU-T-8) (Dataset S2). Individuals TOU-T-2 to TOU-T-7, TOU-T-9, TOU-T-10, and TOU-T-14 to TOU-T-17 were assumed to carry the same cytoplasm as their sister plants with the same chlorotype. All chloroplast and mitochondrial polymorphisms were analyzed as described in ref. 25. The pairwise distances between the 11 stands range from 50 m to 1 km.
This analysis grouped TOU-A1-41, TOU-C-2, TOU-F-1, TOU-R-9, TOU-T-8, TOU-T-9, TOU-T-10, TOU-T-14, TOU-T-15, TOU-T-16, and TOU-T-17 in the previously described Z cytotype, whereas TOU-D-5 and TOU-S-1 were grouped in the previously described AA cytotype. These two cytotypes are very close to both the cytotype of the parental accession Jea and the cytotype Y, where fell TOU-P-6 (25). TOU-Q-2, TOU-T-1, TOU-T-2, TOU-T-3, TOU-T-4, TOU-T-5, TOU-T-6, TOU-T-7, and TOU-T-8 were grouped in the previously described BA cytotype. TOU-E-7 and TOU-I-2 correspond to cytoplasmic haplotypes that have not been described in a set of 95 worldwide accessions.
A cytoplasmic phylogenetic network of the TOU cytotypes was constructed using the same strategy as in (25) (Fig. S5).
Acknowledgments
We thank Cédric Glorieux for assistance during the common garden experiment and Richard Berthomé, Olivier Loudet, Hakim Mireau, and two anonymous referees for helpful comments on an earlier version of this manuscript. This work was funded by the Plant Biology and the Genetics and Plant Breeding Departments at Institut National de la Recherche Agronomique (Cytolignées pilotes-2010 and Cytoressources-2011) and by the Agence Nationale de la Recherche (ANR-12_ADAP-0004). This study was also supported by the LABEX (Laboratoire d'Excellence) TULIP (Towards a Unified Theory of Biotic Interactions: Role of Environmental Perturbations) (ANR-10-LABX-41; ANR-11-IDEX-0002-02). Institute Jean-Pierre Bourgin and Institute of Plant Sciences Paris-Saclay benefit from support from the LABEX Saclay Plant Sciences-SPS (ANR-10-LABX-0040-SPS).
Footnotes
- ↵1To whom correspondence should be addressed. Email: Francoise.Budar{at}versailles.inra.fr.
Author contributions: F.R., C.C., and F.B. designed research; F.R., E.B., E.W., L.B., S.D., C.C., and F.B. performed research; R.V. contributed new reagents/analytic tools; F.R., T.M.-H., E.W., M.-L.M.-M., C.C., and F.B. analyzed data; and F.R., T.M.-H., C.C., and F.B. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Data deposition: Genotyping data of the 56 cytolines are available on the Versailles Arabidopsis Stock Center website, publiclines.versailles.inra.fr/.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1520687113/-/DCSupplemental.
Freely available online through the PNAS open access option.
References
- ↵.
- Dyall SD,
- Brown MT,
- Johnson PJ
- ↵
- ↵
- ↵.
- Dey R,
- Barrientos A,
- Moraes CT
- ↵
- ↵
- ↵
- ↵
- ↵.
- Dowling DK,
- Friberg U,
- Hailer F,
- Arnqvist G
- ↵.
- Ellison CK,
- Burton RS
- ↵.
- Paliwal S,
- Fiumera AC,
- Fiumera HL
- ↵
- ↵
- ↵
- ↵.
- Allen JO
- ↵
- ↵.
- Joseph B, et al.
- ↵
- ↵
- ↵.
- Joseph B,
- Corwin JA,
- Li B,
- Atwell S,
- Kliebenstein DJ
- ↵
- ↵
- ↵.
- Lovell JT, et al.
- ↵.
- Alonso-Blanco C,
- Bentsink L,
- Hanhart CJ,
- Blankestijn-de Vries H,
- Koornneef M
- ↵
- ↵.
- Fournier-Level A, et al.
- ↵.
- Kerwin R, et al.
- ↵.
- Brachi B, et al.
- ↵
- ↵.
- Hancock AM, et al.
- ↵
- ↵
- ↵
- ↵.
- Cronk QCB,
- Whitthon J,
- Ree RH,
- Taylor IEP
- Reboud X, et al.
- ↵.
- Wender NJ,
- Polisetty CR,
- Donohue K
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵.
- Rawson PD,
- Burton RS
- ↵.
- Tang Z,
- Wang X,
- Hu Z,
- Yang Z,
- Xu C
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵.
- Picó FX,
- Méndez-Vigo B,
- Martínez-Zapater JM,
- Alonso-Blanco C
- ↵.
- Benjamini Y,
- Hochberg Y
- ↵
- ↵
- ↵.
- Roux F,
- Gasquez J,
- Reboud X
Citation Manager Formats
Article Classifications
- Biological Sciences
- Plant Biology