Transcriptomic changes following recent natural hybridization and allopolyploidy in the salt marsh species Spartina × townsendii and Spartina anglica (Poaceae)


Author for correspondence:
Malika Ainouche
Tel: +33(0)223235111


  • Allopolyploidy results from two events: the merger of divergent genomes and genome duplication. Both events have important functional consequences for the evolution and adaptation of newly formed allopolyploid species. In spite of the significant progress made in recent years, few studies have decoupled the effects of hybridization from genome duplication in the observed patterns of expression changes accompanying allopolyploidy in natural conditions.
  • We used Agilent rice oligomicroarrays to explore gene expression changes following allopolyploidy in Spartina that includes a classic example of recent allopolyploid speciation: S. anglica formed during the 19th century following genome duplication of the hybrid S. × townsendii.
  • Our data indicate important, but different, effects of hybridization and genome duplication in the expression patterns of the hybrid and allopolyploid. Deviation from parental additivity was most important following hybridization and was accompanied by maternal expression dominance, although transgressively expressed genes were also encountered. Maternal dominance was attenuated following genome duplication in S. anglica, but this species exhibits an increased number of transgressively overexpressed genes.
  • These results reflect the decoupled effects of the ‘genomic shock’ following hybridization and genome redundancy on the genetic, epigenetic and regulatory mechanisms characterizing transcriptomic evolution in allopolyploids.


In recent years, significant advances have been made in our understanding of the functional consequences of reticulate evolution (resulting from interspecific hybridization) and polyploidy (resulting from whole genome duplication) (reviewed in Osborn et al., 2003; Adams & Wendel, 2005; Chen & Ni, 2006; Chen, 2007; Adams, 2007; Doyle et al., 2008; Hegarty & Hiscock, 2008; Soltis & Soltis, 2009; Wendel, 2000). These two processes may occur independently or consecutively in the contexts of homoploid hybrid speciation, autopolyploid or allopolyploid speciation (Gross & Rieseberg, 2005; Wendel & Doyle, 2005), and are now recognized as central diversifying forces in plants and most eukaryotic lineages.

The palaeopolyploid Arabidopsis thaliana has allowed the investigation of the long-term effects of duplicate gene expression (Blanc & Wolfe, 2004; Chapman & Burke, 2006; Thomas et al., 2006; Ha et al., 2009), and experimentally resynthesized hybrids and allopolyploids have revealed the immediate consequences of genome merger and polyploidy, and can be compared with their more or less recent natural counterpart (Adams et al., 2004; Wang et al., 2006b; Flagel et al., 2008; Rapp et al., 2009). The integration of these patterns over a broad evolutionary timescale has suggested that short-term processes affecting the expression of redundant genes duplicated by polyploidy (homeologues) have a critical impact on the differential long-term loss or retention of the homeologues (Freeling, 2009).

An important component of the expression changes discovered in allopolyploids is the biased expression of the parental homeologous genes, which was first detected using cDNA amplified fragment length polymorphism (AFLP) and single-strand conformational polymorphism app-roaches (Lee & Chen, 2001; Kashkush et al., 2002; Adams et al., 2003, 2004; Tate et al., 2006). This phenomenon was particularly well analysed in Gossypium allotetraploids (Flagel et al., 2008, 2009; Hovav et al., 2008; Chaudhary et al., 2009), as Udall et al. (2006) developed a new microarray strategy differentiating homeologous genes. Following these studies, subfunctionalization through homeologue expression partitioning in different organs or different developmental stages is now recognized as an important immediate outcome of allopolyploidy with critical consequences on metabolic plasticity and associated phenotypic changes (Chen, 2007).

Microarrays also allowed the examination of the global levels of expression changes at a larger genomic scale in hybrids or allopolyploids by comparing the total expression level of the allopolyploid with that of its parents and/or the mid-parental expression value (Hegarty et al., 2005; Wang et al., 2006b; Gaeta et al., 2009; Pumphrey et al., 2009; Rapp et al., 2009). The analyses indicated that hybrid or allopolyploid speciation is accompanied by nonadditive expression patterns that reflect either parental expression dominance, in which gene expression is up- or downregulated to the level of one of the parents (Wang et al., 2006b; Rapp et al., 2009), or transgressive patterns (Hegarty et al., 2009).

In natural allopolyploids, the expression changes reflect the superimposed effects of hybridization (reuniting divergent regulatory networks), genome duplication (entailing functional redundancy and altered dosage) and subsequent long-term evolution of the allopolyploid lineage in various selective environmental contexts. Therefore, it is difficult to elucidate the respective impacts of these important evolutionary processes on a single system. Studies attempting to distinguish these phenomena have involved a combination of synthetic F1 hybrids that have been compared with synthetic and/or natural allopolyploids, for example in Senecio (Hegarty et al., 2006) or Gossypium (Flagel et al., 2008; Flagel & Wendel, 2010). These studies, although performed in different technical contexts (see Discussion), provide the first clues on the respective roles of hybridization and genome duplication in allopolyploid transcriptome evolution.

The genus Spartina (Poaceae), which is composed of perennial plants colonizing salt marshes, is one of the few systems in which both the F1 hybrid and the resulting allopolyploid are alive in natural conditions; the parental species can be identified and are available for analysis. Moreover, the natural hybridization and polyploidization events are recent as they happened during the end of the 19th century, thus allowing the exploration of the immediate consequences of allopolyploidy (Ainouche et al., 2004a). The homoploid F1 hybrid Spartina × townsendii was formed in southern England following hybridization between the introduced hexaploid (2n = 62) east American species Spartina alterniflora Loiseleur (as female parent) and the European hexaploid species Spartina maritima (Curtis) Fernald (2n = 60). Genome duplication in S. × townsendii gave rise to Spartina anglica C.E. Hubb., a highly fertile and vigorous dodecaploid (2= 122–124; Marchant, 1963), which has rapidly expanded in range and been introduced on several continents. Notable features of the neopolyploid are its larger ecological range compared with the parental species and its ability to rapidly colonize a vacant niche as pioneer species down the shore, where it is able to tolerate several hours of tidal submersion, making this species highly invasive (Ainouche et al., 2009).

The genetic context of allopolyploid species’ formation in Western Europe (i.e. genetic diversity of the parents, hybrid and allopolyploid) has been well documented (Baumel et al., 2001, 2002a, 2003; Yannic et al., 2004), as have the phylogenetic relationships and genetic divergence of the parents (Baumel et al., 2002b; Ainouche et al., 2004b). A strong genetic bottleneck has precluded the formation of the allopolyploid as a result of limited diversity of the parental species and ‘unique origin’ at the hybridization site. The populations of S. anglica analysed through the species’ range are composed of a major genotype that is identical to S. × townsendii and that exhibits additivity of the parental patterns (Ainouche et al., 2004a and references therein). No major structural changes seem to have affected the homeologous genomes reunited in S. anglica (Baumel et al., 2002a; Salmon et al., 2005; Parisod et al., 2009), whereas consistent methylation alterations were recorded in the hybrid, which were transmitted to the allopolyploid (Salmon et al., 2005; Parisod et al., 2009).

The aim of this study was to explore transcriptomic changes in the Spartina system, which represents an ideal situation to differentiate the effects of hybridization from the effects of genome duplication that occur immediately after natural allopolyploid speciation. As few genomic resources were available in Spartina, we used rice Agilent oligomicroarrays, which allowed the detection of repeatable expression differences between the parental species S. maritima and S. alterniflora (H. Chelaifa et al., unpublished). More specifically, we asked the following questions. (1) What is the extent of the ‘transcriptomic shock’ induced by interspecific hybridization in S. × townsendii? (2) What is the effect of genome doubling in the neoallopolyploid S. anglica? (3) Are there common features of transcriptome evolution following allopolyploid speciation emerging from the systems analysed to date?

Materials and Methods

Plant materials and RNA samples

As mentioned above, the parents S. maritima and S. alterniflora and the allopolyploid S. anglica have low interindividual genetic diversity in Western Europe. Thus, weak genetic diversity is expected among individuals within and between populations, which expand by predominantly clonal propagation. In Brittany (France), we collected the parental species S. maritima (Tascon Island, Morbihan) and S. alterniflora (Landerneau, Finistère) and the allopolyploid S. anglica (Roscoff, Finistère). The hybrid S. × townsendii was sampled in Hythe (UK) at the original hybridization site. Several individuals were collected at each site. Plants were transplanted in the glasshouse (at Rennes University) and were grown in the same conditions (30 cm3 daily watered pots containing a 2 : 1 mixture of soil and sand, respectively) under a day temperature of 20–22°C and night temperature of 14–16°C. The ploidy level was checked for each plant using flow cytometry. After 20 days of acclimatization, 1–2 g of young leaves per plant were collected from five individuals per species (originating from the same population), immediately frozen in liquid nitrogen and stored at −80°C.

Total RNA was isolated using Trizol reagent (Sigma) according to the manufacturer’s recommendations. Each RNA sample was quantified using a Nanodrop® ND 1000 Spectrophotometer (Nanodrop Technologies, Thermo Fisher Scientific Inc.), and the RNA quality was checked on RNA 6000 Nanochips using a Bioanalyzer 2100 (Agilent Technologies, Palo Alto, CA, USA). After processing, RNA was stored at −80°C.

Microarray experiment design and RNA hybridizations

The microarray experiments were performed at the Transcriptomic Platform BiogenOUEST® (Campus de Villejean, University of Rennes 1, Rennes, France). Twenty slides (five replicates per species) were hybridized on a 44 K rice Agilent array (Agilent G2519F), which contains 43 803 60-mer probes covering c. 21 509 genes of the rice genome. Microarray slides were washed and scanned with an Agilent scanner (G2566AA), according to the standard protocol of the manufacturer. Information from probe features was extracted from microarray scan images using Agilent Feature Extraction software v.9.5.1 (Agilent Technologies, USA). Expression data were submitted to the National Center for Biotechnology Information’s Gene Expression Omnibus repository, and are available under the accession number GSE18961.

Microarray data analysis

As our analyses were performed in the context of cross-species’ hybridization, the reproducibility of the hybridization patterns was carefully examined. The ANOVA per-gene analysis (using R software, Lucent Technologies, USA) was used to estimate the number of Spartina genes that hybridized with rice oligos in all replicates.

All signals (feature extraction files) were imported into the GeneSpring 10 software package (Agilent). Only data flagged as ‘present’ from the five replicates were subjected to normalization. A per-chip normalization to the 50th percentile and a per-gene normalization to the median were adopted. Data were analysed using variance analysis (< 0.05) with a Benjamini–Hochberg multiple test correction to identify significantly up- and downregulated genes. To mimic parental additivity in the hybrid and the allopolyploid, the ‘in silico’ mid-parent value (MPV) was calculated by averaging the values obtained for the parents across all replicates.

Gene ontology associated with significantly differentially expressed genes was also employed to identify functional groups using GeneSpring 10 software (Agilent).

Quantitative PCR

Microarray results were confirmed using real-time quantitative reverse transcriptase-polymerase chain reaction (RT-PCR) for a selection of 10 identified genes for which the corresponding oligos were found to be differentially expressed on the microarray: 18S ribosomal RNA gene, 28S ribosomal RNA gene, vacuolar ATPase B gene, poly(A)-binding protein, OsPAG1, actin gene, ribulose-phosphate 3-epimerase, NADPH oxidase, malate synthase-like family protein and phosphoenolpyruvate carboxykinase. Primers (Table 1) were designed in Spartina from expressed sequence tag (EST) sequences obtained in S. maritima (H. Chelaifa et al., unpublished).

Table 1.   Primers used for the quantitative reverse transcriptase-polymerase chain reaction (RT-PCR) analysis
Coding regions (Spartina maritima ESTs)Forward primer (5′–3′)Reverse primer (5′–3′)
  1. EST, expressed sequence tag.

  2. *Primers from Christin et al. (2009).


To confirm the expression patterns obtained from the microarray analysis, three replicates from two RNA samples (total of six replicates) were analysed for each species using quantitative PCR with a 7900HT Fast Real-Time PCR System (Applied Biosystems Inc., USA); 100 ng of total RNA from each sample were reverse transcribed with oligo(dT)20 primers using a Superscript III first-strand cDNA synthesis kit (Invitrogen), and were diluted 20 times with sterile distilled water. The absence of genomic DNA contamination in the RNA samples was confirmed using RNA RT, amplified without reverse transcriptase; 2 μl of each diluted cDNA were used per reaction with Applied Biosystems SYBR Green TaqMan® MIX and specific forward and reverse primers (Table 1). Primers were designed using Primer 3 software ( and selected to yield amplicons of 180–200 bp. PCR conditions were as follows: 50°C for 2 min, 95°C for 10 min, followed by 40 cycles of 95°C for 30 s, 60°C for 1 min and 72°C for 60 s. Primer pairs that produced dimers were eliminated by melting curve analysis. All data replicates were normalized using the glyceraldehyde-3-phosphate dehydrogenase gene amplified with the primer sequences AGAGTGCCTCGTCAAGGAGA (F) and CTCCCAAGCAATCCTCATGT (R) from Baisakh et al. (2006).


Microarray data quality and hybridization signals

Although high stringency in hybridization conditions and long (60-mer) probes limit the potential bias induced by cross-species’ hybridization of Spartina RNA on the rice microarray, we examined carefully the reproducibility of the hybridization level in each species. The ANOVA per-gene analysis (using R software) indicated that RNA from S. maritima, S. alterniflora, S. × townsendii and S. anglica hybridized with 13 573, 13 993, 13 210 and 13 535 rice genes of the array, respectively, and exhibited a stable and repeatable expression level. This represents c. 70% of the 21 510 rice genes of the microarrays. This proportion is consistent with the high coding sequence conservation (averaging 90%) estimated between rice and Spartina (H. Chelaifa et al., unpublished). Recent studies have shown that cross-genus microarray hybridization is both feasible and informative in plants (Hammond et al., 2006; Davey et al., 2009; Gaeta et al., 2009), although the rates of hybridization may vary according to biological systems, divergence rates and the technology employed. For example, over 83% of maize oligos cross-hybridized with cDNAs from Andropogon gerardii (Travers et al., 2007). Pachycladon species hybridized with an A. thaliana microarray with c. 76% hybridization (Voelckel et al., 2008).

Our experimental design and statistical analyses using Genespring 10.0 (Agilent) allowed the identification of reproducible expression differences among the five species. After data filtration with Genespring 10.0 (Agilent) (only flags ‘present’ were taken into account), we were able to compare the expression of a total of 10 744 genes that were expressed in at least one Spartina species.

Gene expression comparison between the parents S. maritima and S. alterniflora

Comparisons of the two parents identified 1247 genes that exhibited significant expression differences (< 0.05), which represents c. 11.6% transcriptomic divergence when considering the genes that hybridized with Spartina RNA on the array (10 744). Most of the differentially expressed genes were overexpressed in S. alterniflora (958 genes) relative to S. maritima (257 genes), with more than two-fold differences. Annotation of the differentially expressed genes between S. maritima and S. alterniflora, according to biological process, revealed that genes potentially involved in development and cellular growth were more represented (15.8%) than those involved in other functions. It was also observed that the genes involved in the metabolism of proteins, carbohydrate and lipid systems, cytoplasm organization, and biogenesis and transcription systems were differentially expressed in the leaves of these species. The large difference in the expression of developmental and cellular growth genes, highly upregulated in S. alterniflora and downregulated in S. maritima, is consistent with their different morphology and growth (H. Chelaifa et al., unpublished).

Gene expression in the homoploid hybrid S. × townsendii

Comparison of the expression patterns of S. × townsendii with the MPV was performed for 10 744 genes that hybridized on the array in the parental species and the hybrid, and that passed all criteria of normalization and flag filter. This analysis provided a list of 689 genes (6.4%) that showed significant differences in expression (< 0.05). Most of these genes (89%), whose expression was altered in the hybrid, exhibited differential expression between the parents S. maritima and S. alterniflora.

When comparing S. × townsendii with its parents, 1235 genes (11.5%) were found to be differentially expressed between the hybrid and S. maritima. By contrast, only 338 genes (2.9%) were differentially expressed between the hybrid and S. alterniflora. This four-fold difference (11.5% and 2.9%) suggests expression dominance of the maternal parent S. alterniflora in the hybrid. Sixty-five genes exhibited transgressive patterns, that are differentially expressed between the hybrid and both parents. Among these, 40 genes were upregulated and 25 were downregulated in the hybrid (Fig. 1).

Figure 1.

 Venn diagram of genes with significant differential expression between the hybrid Spartina × townsendii (H) and the parental species (paternal S. maritima and maternal S. alterniflora) with regard to up- and downregulation.

Following the analytical method developed by Rapp et al. (2009), 14 expected (additive or nonadditive) expression patterns of the hybrid relative to the parents were distinguished (Fig. 2). Additive parental patterns in the hybrid included cases in which the parents had either similar or different expression patterns, and where the hybrid had intermediate expression, being either statistically different or not from the parents. An important fraction (92%) of the expressed genes in the leaves of Spartina exhibited additive expression patterns in the hybrids. Among the genes that were differentially expressed between the parents, the most frequent situation was when genes were upregulated in S. alterniflora and downregulated in S. maritima, with an intermediate expression level in the hybrid, which was not statistically different from both parents (Fig. 2).

Figure 2.

 Expression patterns in the hybrid (H), (paternal) Spartina maritima and (maternal) S. alterniflora. n represents the corresponding number of genes in each condition. Double bars indicate a significant expression difference between the hybrid and the parents in the additive patterns.

Among the genes that were not additively expressed in the hybrid, we distinguished those that exhibited transgressive patterns (i.e. overexpressed or underexpressed compared with the parents) from those displaying parental dominance in the hybrid (Fig. 2).The number of genes that exhibited parental expression dominance appeared to be relatively important, S. alterniflora being the more frequently dominant, specifically for upregulated genes (= 573, Fig. 2).

Gene expression patterns in the allopolyploid S. anglica

The effects of genome duplication on the transcriptome of S. anglica were evaluated by comparing the allopolyploid with S. × townsendii, with the in silico MPV and with the parental species S. alterniflora and S. maritima. When comparing the expression patterns of S. anglica and S. × townsendii, we found 497 genes (4.6%) that exhibited significant expression differences (< 0.05). The majority of these genes (310) were overexpressed in the allopolyploid. We found that 391 (3.6%) genes were differentially expressed between S. anglica and the MPV, suggesting that the deviation from parental expression additivity is less pronounced in the allopolyploid than in the F1 hybrid (6.4%).

Spartina anglica exhibited 656 differentially expressed genes compared with S. maritima (6.1%), with 569 genes overexpressed in S. anglica. When we compared the allopolyploid with S. alterniflora, we found that 496 (4.6%) genes were differentially expressed. Among these genes (Fig. 3), 118 were transgressive in S. anglica (i.e. significantly exceeding the expression levels of both parental species; Fig. 3). The majority (101) of these transgressive genes were upregulated in the allopolyploid. Forty of these genes were also transgressive and upregulated in the F1 hybrid. Our data indicate that these transgressive genes show greater expression levels in the allopolyploid than in the hybrid.

Figure 3.

 Venn diagram of genes with significant differential expression between the parents (paternal Spartina maritima and maternal S. alterniflora) and the allopolyploid S. anglica (Allo) with regard to up- and downregulation.

An important fraction (88.2%) of the genes that hybridized on the array exhibited additive parental expression patterns in the allopolyploid (Fig. 4). No particular parental expression dominance was observed (Fig. 4), S. alterniflora being only slightly more frequently dominant than S. maritima.

Figure 4.

 Expression patterns in the allopolyploid Spartina anglica (Allo) and the parents (paternal) S. maritima and (maternal) S. alterniflora. n represents the corresponding number of genes in each condition. Double bars indicate a significant expression difference between the allopolyploid and the parents in the additive patterns

The transcriptomic differences among the parents, the hybrid and the allopolyploid are summarized in Fig. 5. This figure illustrates the following: (1) the relatively important transcriptomic divergence between the parents (11.6%); (2) the changes resulting from hybridization (6.4% difference between the F1 hybrid and the expected expression additivity estimated by in silico MPV); (3) the expression dominance of the maternal parent (S. alterniflora) in the F1 hybrid; (4) the additional expression changes (4.6%) entailed by genome doubling in S. anglica. Genome duplication appears to have a balancing effect with regard to expression divergence with the parents, as genome doubling in S. anglica decreased the expression divergence with S. maritima (which was higher in S. × townsendii) and increased the expression divergence with S. alterniflora (which was lower in S. × townsendii). Thus, polyploidization per se had a global buffering effect on the initial nonadditive expression levels observed following hybridization, although the nonadditively expressed genes in S. anglica were significantly overexpressed compared with the parental species and the F1 hybrid.

Figure 5.

 Transcriptomic differences among the five Spartina species, with the percentage of genes differentially expressed between the hybrid and the allopolyploid with the maternal parent S. alterniflora (in yellow) and male parent S. maritima (in blue). The underlined value indicates genes differentially expressed between the hybrid and the in silico mid-parent value (MPV). Black lines represent natural genealogical relationships, including the divergence of S. maritima and S. alterniflora from a common hexaploid ancestor, hybridization between S. maritima and S. alterniflora, and genome duplication (thicker line) of the hybrid S. × townsendii.

Biological significance of genes differentially expressed between Spartina species

The gene ontology (GO) of the differentially expressed genes was compared with the GO of all genes present on the array with a P value cut-off of 0.1. GO terms were assigned to all genes on the array for biological process. The functional category of genes for which expression was affected (i.e. differentially expressed) by hybridization and genome duplication was identified by comparing the hybrid and the MPV, on the one hand, and the hybrid and the allopolyploid, on the other (Fig. 6).

Figure 6.

 Functional distribution of differentially expressed genes among gene ontology biological process-annotated genes: (a) between the hybrid Spartina × townsendii and the mid-parent value (MPV); (b) between the hybrid S. × townsendii and the allopolyploid S. anglica.

Genes affected by hybridization are involved in various functions according to annotation by biological process (Fig. 6). Electron transporters, growth and cellular development and nucleotide metabolism are more represented than the other groups, and represent more than 50% of the totality of the genes. Cytoplasm organization (15.7%) and transcription factors (16%) were also largely represented, including WRKY transcription factor 50 (AK109578, = 0.00126), general transcription factor TFIIB (AK068941, = 0.00201) and bZIP transcription factor (AK070887, = 0.02130).

We attempted to identify the functional category of the genes exhibiting transgressive expression patterns in the hybrid. Although a specific function could not be assigned for most of these genes, we were able to identify the long terminal repeat-retrotransposon (LTR) Skipper (AK062058), the adh1 gene (AK110720) and a guanosine triphosphate (GTP)-binding protein (RACDP AK06779) as upregulated in S. × townsendii. A large proportion of the genes repressed in the hybrid (28%) were involved in carbohydrate metabolism.

Annotation by biological process of the genes exhibiting differential expression between the hybrid and the allopolyploid revealed that, again, genes potentially involved in development and cellular growth (44.7%) were significantly more represented (Fig. 6). It was also notable that transporters were highly represented (18.4%). Genes involved in carbohydrate and lipid metabolism, cytoplasm organization and as transcription factors were also differentially expressed in the leaves of the hybrid S. × townsendii and the allopolyploid S. anglica. This distribution of functional categories (abundance of growth and cellular development functions) in the comparison between the F1 hybrid and the allopolyploid was similar to the distribution of genes differentially expressed between the parental species S. maritima and S. alterniflora.

The largest proportion of genes that displayed expression activation in the allopolyploid S. anglica was involved in catalytic activities (27.7%). Genes involved in binding lipid protein (6.7%) and in protein biosynthesis (5.2%) were also represented. Only one important transcription factor was activated: TATA-binding protein TBP2 (AK061103).

Among the repressed genes in the allopolyploid, the largest proportion affected chloroplast genes (29.4%), including the chlororoplast 4.5S, 5S, 16S and 23S (AK060185) and the chloroplast rpoC2 (AK062380) genes. Two mitochondrial genes, mitochondrial 26S ribosomal RNA (AK061547) and a mitochondrial NADH dehydrogenase (AK064531), were also repressed in S. anglica.

Quantitative RT-PCR

Quantitative RT-PCR was employed to confirm the gene expression changes indicated by the microarray data. Seven genes exhibiting different expression patterns in the microarray were analysed, and the expression levels were compared with those obtained with the rice microarrays (Fig. 7). This comparison revealed similar expression trends (Fig. 7). For example, the ribulose-phosphate 3-epimerase, OsPAG1 and NADPH genes displayed the same level of expression in both quantitative RT-PCR and microarrays: they were differentially expressed in S. maritima and S. alterniflora, highly expressed in the hybrid compared with the parents and exhibited highest expression in S. anglica. Poly(A)-binding protein, 18S ribosomal gene and actin genes also showed the same expression patterns with quantitative RT-PCR and microarrays: they were differentially expressed between the parents, their expression in the hybrid was downregulated compared with S. alterniflora and S. anglica exhibited the highest expression. More variable expression between quantitative RT-PCR and microarrays was observed in the vacuolar ATPase. For all genes and all species, expression levels appeared to be significantly higher when estimated with quantitative RT-PCR than with microarrays. Although this is a frequently observed phenomenon in comparisons reported in the literature (because of different sensitivities of the techniques), these differences may be accentuated by the cross-species’ hybridization procedure used here.

Figure 7.

 Comparison of the microarray (dark grey bars) and quantitative PCR (light grey bars) expression patterns for seven genes. Broken lines indicate the expression changes between species using quantitative PCR data, and full lines indicate the expression changes using microarray data.


Despite the number of genes masked by stringent statistical analyses and the effect of cross-species’ hybridization, our results demonstrate that the rice oligomicroarray can be used to identify repeatable gene expression differences in Spartina species. The high stringency of the hybridization conditions and the long (60-mer) oligoprobes limit the potential bias resulting from mismatches (Oshlack et al., 2007), and increase the reliability of the heterologous microarray approach employed in this study. The relatively high coding sequence conservation between rice and Spartina and among Spartina species (averaging 90% and 97% identity, respectively; H. Chelaifa et al., unpublished) also limits the influence of mismatches on microarray hybridization that could potentially affect the detection of expression changes between species. Comparisons of the expression patterns in the leaves of the natural hybrid S. × townsendii and its parents revealed that genome merger had immediate and important consequences. The main observation was the high level of expression changes between the hybrid and its paternal parent S. maritima (11.4%). The level of these changes was equivalent to the proportion of differentially expressed genes (11.6%) between the parental species S. maritima and S. alterniflora, which diverged from a common hexaploid ancestor and exhibited 1–5% nucleotide divergence (H. Chelaifa et al., unpublished). There were also clearly differences of gene expression between the natural allododecaploid S. anglica and the F1 hybrid (4.6%). These results suggest that allopolyploid speciation was accompanied by two distinct phases of transcriptomic evolution, involving significant alterations following hybridization (in the hexaploid homoploid hybrid S. × townsendii) followed by additional changes in the allododecaploid. Only a few allopolyploid monocotyledonous species have been studied to date with regard to transcriptomic evolution, apart from wheat (e.g. He et al., 2003; Pumphrey et al., 2009), in which, by contrast with Spartina, present-day species have undergone human-selective pressure.

Nonadditivity and maternal dominance in the hybrid S. × townsendii

We evaluated the effects of interspecific hybridization on gene expression by comparing the natural hybrid S. × townsendii with the in silico MPV. About 6.4% of the analysed genes were nonadditively expressed compared with the parents S. maritima and S. alterniflora. These results suggest consistent changes induced by the merger of differentiated genomes. Gene misexpression resulting from the reunion of divergent regulatory networks may underlie reproductive isolation mechanisms in interspecific hybrids (Renaut et al., 2009). ‘Transcriptome shock’, the extended view of the ‘genome shock’ concept (McClintock, 1984), appears to be a common feature of the interspecific hybrids examined to date, even though the results were obtained in different genetic and technical contexts. In plants, using long (70-mer) oligonucleotide microarrays, Wang et al. (2006b) detected 5.2–5.6% deviations from parental expression additivity in leaves from hybrids between the tetraploids A. thaliana and A. arenosa. Hegarty et al. (2005) employed anonymous cDNA microarrays, which revealed that 5–7% genes of the array were differentially expressed in flowers in the interspecific triploid hybrid Senecio × baxteri resulting from hybridization between the diploid Senecio squalidus and the tetraploid Senecio vulgaris. Helianthus EST-based cDNA microarrays allowed the detection of expression changes between the homoploid interspecific hybrid Helianthus deserticola and its parental diploid species H. annuus and H. petiolaris (Lai et al., 2006), in which 12.8% of the genes on the array exhibited differential expression across seedlings from the three species. Flagel et al. (2008) employed a microarray technology using homeologue (i.e. parental)-specific probe sets, and showed that 24% of the genes exhibited biased expression patterns (unequal parental contributions) in a synthetic interspecific hybrid between the diploid Gossypium arboretum and G. raimondii.

An important proportion of the nonadditively expressed genes in S. × townsendii exhibited similar expression patterns to the maternal parent S. alterniflora. This phenomenon, termed ‘expression dominance’, was thoroughly described by Rapp et al. (2009) in two experimentally resynthesized allotetraploids (colchicine doubled hybrids) from the genus Gossypium, in which the maternal parent was alternatively ‘dominant’ or ‘recessive’ depending on the specific genomic combination. Expression dominance was also observed in synthetic hybrids between tetraploid A. thaliana and tetraploid A. arenosa (Wang et al., 2006b; Chen, 2007), in which most of the genes expressed at higher levels in A. thaliana than in A. arenosa were repressed to the level of A. arenosa in the hybrids. In S. × townsendii, expression dominance affects genes that are either upregulated or downregulated in S. alterniflora relative to S. maritima. The maternal expression dominance was unexpected, as it contrasts with the analyses performed at the genomic (DNA) level, in which the maternal genome from S. alterniflora seems to be more prone to genetic changes, as examined with AFLP (Salmon et al., 2005) and sequence-specific amplification polymorphism (SSAP) (Parisod et al., 2009) markers.

Not all the nonadditively expressed genes were the result of parental dominance, as 60 genes were also found to exhibit transgressive patterns in S. × townsendii. Extreme phenotypes relative to parents are frequently observed in interspecific hybrids (Gross & Rieseberg, 2005). Transgressive expression patterns were found to be correlated with adaptation to xeric conditions in the homoploid hybrid Helianthus deserticola (Lai et al., 2006). Equal numbers of transgressive genes were upregulated or downregulated relative to the parents, whereas, in our study, most transgressive genes (40/60) were upregulated. No particular functional category was over-represented among the genes that were nonadditively expressed in S. × townsendii, as also encountered in the hybrid Senecio × baxteri (Hegarty et al., 2006). Transgressive expression patterns were also reported in both the natural and experimentally resynthesized homoploid hybrid Senecio squalidus occurring in the UK (Hegarty et al., 2009), in which most nonadditive genes were upregulated compared with MPV.

Gene expression changes in hybrids involve various mechanisms that are being elucidated. The reunion of divergent genomes in the same nucleus results in the interaction of diverged regulatory hierarchies (Riddle & Birchler, 2003). Interacting cis- and trans-regulatory factors have been shown to be involved in expression changes of synthetic hybrids (Wittkopp et al., 2004; Stupar & Springer, 2006; Zhuang & Adams, 2007; Chaudhary et al., 2009; Tirosh et al., 2009). These changes may have important phenotypic impact (e.g. flowering time variation in Arabidopsis; Wang et al., 2006a).

Epigenetic regulation represents an important mechanism altering expression following hybridization. Important methylation changes have been reported in S. × townsendii (Salmon et al., 2005). Transposable elements are usually targets for epigenetic repression which may be altered following genomic stress, such as hybridization. In this study, we also identified a long terminal repeat-retrotransposon Skipper among the transgressively expressed genes in S. × townsendii, which is consistent with previous findings reporting the activation of transposable elements following hybridization (Kashkush et al., 2002). This transcriptional activation does not obviously entail transposition bursts (Madlung et al., 2005; Parisod et al., 2010), but may alter the expression of adjacent genes (Kashkush et al., 2003). In Arabidopsis, interspecific hybridization resulted in the activation of the paternal copies of the normally silenced ATHILA transposable element (Josefsson et al., 2006) which was associated with hybrid inviability; this activation was found to be sensitive to parental genome dosage. Interestingly, in Spartina, most methylation alterations following hybridization were found in regions flanking transposable elements (Parisod et al., 2009).

The multiple molecular mechanisms involved in the expression changes occurring in hybrids play a central role in hybrid vigour or heterosis (Birchler et al., 2003; Swanson-Wagner et al., 2006; Li et al., 2009). Although sterile, S. × townsendii exhibits particular vigour for vegetative traits (Marchant, 1967), and a recent survey in southern England has revealed that this F1 hybrid still fares well at its site of origin, where it co-occurs with the allopolyploid S. anglica (Renny-Byfield et al., 2010). Despite its aggressive ecological habit, S. anglica has not outcompeted S. × townsendii at the hybridization site after more than 100 yr.

Hybridization vs genome duplication-induced expression changes

Hybridization is considered to play a major role in the expression changes exhibited by allopolyploidy (Albertin et al., 2006; Hegarty et al., 2006, 2008; Adams, 2007), whereas moderate changes seem to be associated with autopolyploidy (Stupar & Springer, 2006; Wang et al., 2006b). Contrasting with previous investigations at the DNA level, highlighting the high genomic similarity between the F1 hybrid and the allopolyploid (Baumel et al., 2002a,b; Ainouche et al., 2004a, 2009; Salmon et al., 2005), chromosome doubling appears to have an important effect on the transcriptome of S. anglica. Epigenetic changes (methylation alteration at CpG sites) were found about five times more frequently following hybridization than after genome doubling and subsequent generation at both random loci (Salmon et al., 2005) and in the vicinity of transposable elements (Parisod et al., 2009), where S. anglica inherited almost all of the parental additive and nonadditive patterns encountered in the F1 hybrid. Our study indicates that, in addition to hybridization, genome duplication and subsequent generations caused 4.6% of the gene expression changes between the hybrid S. × townsendii and the allopolyploid.

It is now well established that allopolyploidy is accompanied by both additive and nonadditive patterns of parental gene expression affecting various organs and functions in different plant genera, such as Arabidopsis (Lee & Chen, 2001; Wang et al., 2006a; Ni et al., 2009), Brassica (Chen & Pikaard, 1997; Gaeta et al., 2007, 2009), Gossypium (Adams et al., 2003; Udall et al., 2006; Flagel et al., 2008; Hovav et al., 2008; Chaudhary et al., 2009; Rapp et al., 2009), Triticum (He et al., 2003; Pumphrey et al., 2009), Senecio (Hegarty et al., 2005, 2006) and Tragopogon (Tate et al., 2006), although the patterns and intensity of these changes, together with the associated mechanisms, vary among the models investigated to date (reviewed in Osborn et al., 2003; Adams, 2007; Chen, 2007; Doyle et al., 2008). It should be noted that not all studies have dissociated the respective effects of hybridization and genome duplication. Moreover, most of the knowledge gained on the short-term consequences of allopolyploidy on gene expression reported in the literature is based on experimentally resynthesized polyploids that are eventually compared with their natural counterpart. In most cases, naturally occurring F1 hybrids were not available, thus limiting the analysis to experimental conditions. To our knowledge, our study is the first to perform a comparison between the actual F1 hybrid and its immediately derived allopolyploid that formed and survived in natural conditions. Contrasting results have been reported in two studies that compared the global gene expression patterns between synthetic F1 hybrids and allopolyploids: Hegarty et al. (2006) found that genome duplication in the allohexaploid Senecio cambrensis had a ‘calming (reducing) effect’ on the altered patterns of gene expression detected in the synthetic triploid F1 hybrid Senecio × baxteri mentioned above. In natural and synthetic allotetraploid cotton, Flagel et al. (2008) showed that homeologue-biased expression resulting from the unequal contribution of the parental genomes to the total transcriptome was essentially a result of genome merger, but that additional changes (amplification of expression bias) appeared in the allopolyploids that were not instigated by genome merger. In Spartina, the deviation from parental expression additivity (as estimated by MPV) was more pronounced in the F1 hybrid S. × townsendii (6.4%) than in the allopolyploid S. anglica, suggesting a similar trend to that observed in Senecio. However, genome duplication also entailed expression changes in S. anglica (discussed in next section), as found in Gossypium.

Overexpression following genome duplication in the recent allododecaploid S. anglica

The maternal expression dominance encountered in S. × townsendii (for 573 genes) appears to be attenuated in S. anglica (126 genes). However, a remarkable feature of the effect of genome duplication in S. anglica is the increased number of transgressively expressed genes (compared with the F1 hybrid) that are overexpressed, indicating that genome duplication also had an immediate effect on transcriptome evolution. This overexpression trend revealed in S. anglica is interesting and contrasts with previous observations in which gene silencing was the main phenomenon characterizing the alteration of expression following polyploidy (Adams & Wendel, 2004; Adams, 2007; Chen, 2007). Wang et al. (2004, 2006b) reported mainly transcript deactivation in Arabidopsis; Pumphrey et al. (2009) found 7% downregulated genes and 9% upregulated genes (compared with MPV) in seedling leaves of synthetic allohexaploid wheat. Overexpressed genes following genome duplication in Spartina were mainly involved in catalytic activity. Downregulated genes were also encountered in S. anglica and were mostly represented by cytoplasmic genes. Although the dosage of cytoplasmic genes is not affected by whole genome duplication, it is interesting to note that Hegarty et al. (2009) reported an altered expression of cytoplasmic genes (including overexpression of a fertility restorer-like gene) following interspecific hybridization in Senecio squalidus, and considered the potential impact of the misregulation of cytoplasmic genes on the cytoplasmic male sterility that affects interspecific hybrids.

Various genetic, epigenetic and regulatory mechanisms may account for the expression changes in polyploids. In recent Tragopogon allotetraploids, preferential expression of parental homeologues (detected using cDNA AFLP) was found to be correlated with a loss of parental genomic fragments (Tate et al., 2006). These changes were not observed in the experimentally resynthesized Tragopogon F1 hybrids, indicating that they arose following genome duplication. The allotetraploid Tragopogon species were formed reciprocally multiple times and appear to be karyotypically unstable regarding their parental contribution (Lim et al., 2007). Homeologue loss and gene silencing occur in independently formed allopolyploid populations (Buggs et al., 2009; Tate et al., 2009). Similar findings have been reported in Brassica napus, another noteworthy structurally unstable allotetraploid (Gaeta et al., 2007; Cifuentes et al., 2010; Gaeta & Pires, 2010). Genome-wide analysis of gene expression (using Arabidopsis microarrays) in the B. napus lines exhibiting different numbers of genetic changes at the S5:6 generation indicated, however, no global effect of genomic rearrangement (Gaeta et al., 2009). Spartina anglica is, by contrast, a genetically stable allopolyploid that exhibits identical sequence composition and methylation patterns to the F1 hybrid S. × townsendii (Baumel et al., 2002a; Ainouche et al., 2004a, 2009; Salmon et al., 2005; Stupar & Springer, 2006); therefore, the observed changes are not likely to be the result of either structural changes or massive methylation alterations, as found in allotetraploid cotton (Liu et al., 2001).

The observed expression changes in S. anglica reflect the functional consequences of the high genomic redundancy of the allododecaploid genome. As S. anglica is a perennial clonal species in which individuals may survive > 10 yr, one century of evolution may actually represent a limited number of generations elapsed since the time of species’ formation. Thus, it is likely that most of the observed changes represent immediate or very short-term responses to whole genome duplication, although we cannot rule out evolutionary events affecting the transcriptome subsequent to allopolyploid speciation, as reported in recent Tragopogon allopolyploids (Buggs et al., 2009, 2010). Following duplication, different gene categories can exhibit various expression patterns. It has been shown that most regulatory genes are dosage dependent, whereas housekeeping or enzymatic/ metabolic genes usually exhibit dominance or recessive patterns between allelic alternatives (Birchler et al., 2003). Dosage compensation is known to be a regulatory phenomenon reducing duplicate gene expression to a ‘diploid-like’ level (Birchler et al., 2005). A balance in regulatory processes, in which the components are sensitive to stoichiometric relationships, appears to play an important role in the long-term maintenance of duplicated genes during the evolution of polyploid species (Birchler & Veitia, 2007).

In conclusion, our analyses of the Spartina system have allowed us to analyse separately the short-term consequences of genome merger and duplication in a natural system during the allopolyploid speciation process.

Our data indicate important, but different, effects of both processes, reflecting the decoupled effects of hybridization, on the one hand, and genome redundancy, on the other, on the genetic, epigenetic and regulatory mechanisms that characterize transcriptomic evolution in allopolyploids. The changes are underestimated as the cross hybridization microarray procedure and the stringent data analysis might have masked actual expression alterations. Moreover, the global gene expression analysis does not allow us to distinguish the relative contributions of the homeologues to the transcriptome, or to differentiate cases in which similar levels of expression are attained either via equal expression of the homeologous parental genomes or via biased parental expression. Different homeologues may also have different binding affinities to the oligos when in competition with each other, which are not exhibited when they are not in competition. The new massive parallel sequencing technologies might offer promising perspectives for previously poorly investigated genomes of natural systems, such as Spartina, and for developing comparisons across various allopolyploids to provide a clear picture of the general trends characterizing transcriptome evolution during the polyploidy speciation process.


This work was supported by the French National Research Agency (ANR, Polyploidy and Biodiversity project) and the Centre National de la Recherche Scientifique (CNRS). H.C. was supported by a fellowship from the French Ministry of Education. Regis Bouvet is thanked for assistance during the microarray experiments. We are grateful to R. Rapp and J. Wendel for helpful exchanges on transcriptome analyses in polyploids, and to three anonymous reviewers for their constructive comments.