A comparative genomics examination of desiccation tolerance and sensitivity in two sister grass species - pnas.org

Significance

This is a significant sister group contrast comparative study of the underpinning genomics and evolution of desiccation tolerance (DT), a critical trait in the evolution of land plants. Our results revealed that the DT grass Sporobolus stapfianus is transcriptionally primed to tolerate a dehydration/desiccation event and that the desiccation response in the DT S. stapfianus is distinct from the water stress response of the desiccation-sensitive Sporobolus pyramidalis. Our results also show that the desiccation response is largely unique, indicating a recent evolution of this trait within the angiosperms, and that inhibition of senescence during dehydration is likely critical in rendering a plant desiccation tolerant.

Abstract

Desiccation tolerance is an ancient and complex trait that spans all major lineages of life on earth. Although important in the evolution of land plants, the mechanisms that underlay this complex trait are poorly understood, especially for vegetative desiccation tolerance (VDT). The lack of suitable closely related plant models that offer a direct contrast between desiccation tolerance and sensitivity has hampered progress. We have assembled high-quality genomes for two closely related grasses, the desiccation-tolerant Sporobolus stapfianus and the desiccation-sensitive Sporobolus pyramidalis. Both species are complex polyploids; S. stapfianus is primarily tetraploid, and S. pyramidalis is primarily hexaploid. S. pyramidalis undergoes a major transcriptome remodeling event during initial exposure to dehydration, while S. stapfianus has a muted early response, with peak remodeling during the transition between 1.5 and 1.0 grams of water (gH2O) g−1 dry weight (dw). Functionally, the dehydration transcriptome of S. stapfianus is unrelated to that for S. pyramidalis. A comparative analysis of the transcriptomes of the hydrated controls for each species indicated that S. stapfianus is transcriptionally primed for desiccation. Cross-species comparative analyses indicated that VDT likely evolved from reprogramming of desiccation tolerance mechanisms that evolved in seeds and that the tolerance mechanism of S. stapfianus represents a recent evolution for VDT within the Chloridoideae. Orthogroup analyses of the significantly differentially abundant transcripts reconfirmed our present understanding of the response to dehydration, including the lack of an induction of senescence in resurrection angiosperms. The data also suggest that failure to maintain protein structure during dehydration is likely critical in rendering a plant desiccation sensitive.

Desiccation tolerance (DT) is a fundamental trait that is widespread and developed early in the evolution of the land plants (1, 2), and it is believed to have been critical in the colonization of the land by green algae (3). In tracheophytes, DT is generally limited to reproductive propagules, such as seeds and spores, while vegetative desiccation tolerance (VDT) occurs in only 0.086% of known vascular plant species (4). Our understanding of VDT (and its relationship to seed DT) has broadened with the recent expansion of whole-genome sequencing of resurrection plants, tracheophytes that can survive the desiccation of their vegetative tissues. Since the release of the Boea hygrometrica genome sequence (5), the genomes of four other resurrection angiosperms [Xerophyta schlecteri (6), Oropetium thomaeum (7, 8), Lindernia brevidens (9), and Eragrostis nindensis (10)], two lycophytes [Selaginella tamariscina (11) and Selaginella lepidophylla (12)], and the bryophyte Syntrichia caninervis (13) have been published. Apart from the obvious benefits of obtaining genomic resources for individual resurrection species, the establishment of a collection of resurrection plant genomes offered the possibility of the reconstruction of an ancestral genome of a desiccation-tolerant progenitor that would reveal a genomic signature (blueprint) that defines a common mechanism for DT. However, a genomic blueprint for DT has not emerged (4), which may be related to the small number of genomes available and limited phylogenetic sampling, that all tracheophytes possess desiccation-tolerant propagules (seeds or spores), which would obfuscate the comparative analyses, or that the origin of DT lies deep in the land plant phylogeny and is thus cryptic in the recent plant lineages. It may also be a combination of these possibilities or that there is no genomic blueprint for this fundamental trait. Although a genomic blueprint for DT has not been revealed, comparative studies have demonstrated that certain gene families, such as those for early light-inducible proteins (ELIPs) and late embryogenesis-abundant proteins, have expanded in species that exhibit VDT (6, 14, 15).

A corollary to the ancestral reconstruction approach to understanding the evolution of VDT and the genomic aspect of its phenotypic expression is the comparison of the genomes of closely related species that contrast the two extremes: sensitivity and tolerance. Such closely related contrasting species pairings are rare in resurrection plants, but this approach has been applied, albeit with species pairs that are not as close as would be ideal. The genomes and dehydration–rehydration transcriptomes of two resurrection eudicots within the Linderniaceae family (16), the desiccation-tolerant L. brevidens and the desiccation-sensitive (DS) Lindernia subracemosa, were sequenced and compared (9). The comparison revealed that at least in the Lindernia lineage, VDT evolved via a combination of gene duplications in gene families that are functionally associated with the desiccation response and a network-level rewiring of gene expression in vegetative tissue commonly associated with seed desiccation. More recently, a comparative analysis of two contrasting grass genomes along with their respective desiccation-related transcriptomes, the desiccation-tolerant E. nindensis and the related DS cereal Eragrostis tef, reinforced the potential role of gene duplications in the evolution of DT (10). Although there is still a significant phylogenetic distance between these two Eragrostis species (17), the comparative analysis and its extension to include other C4 grasses, including the desiccation-tolerant O. thomaeum, revealed chromatin restructuring and methylation patterns associated with down-regulated genes and specific seed-related orthologs whose expression is associated with VDT. The comparative transcriptome analyses indicated that genes having important roles in seed development and DT are broadly expressed under dehydration in both sensitive and tolerant species, with just a few genes uniquely expressed in the tolerant plants.

In this study, we have chosen two phylogenetically closely related C4 grasses, the homoiochlorophyllous desiccation-tolerant Sporobolus stapfianus and the DS Sporobolus pyramidalis, to develop detailed comparative genomic and transcriptomic analyses to further explore genomic inferences into the evolution of VDT. S. stapfianus and S. pyramidalis are members of the same clade, clade A, in the Sporobolus family of the Sporobolinae subtribe of the Chloridoid grasses (18). S. stapfianus has been the subject of many mechanistic studies of its DT phenotype (19, 20) and along with S. pyramidalis, the subject of a detailed comparative leaf metabolomics study that highlighted differences in the metabolic responses of the two species to dehydration (21). We constructed Hi-C–derived assemblies of the sequenced genomes for both species and conducted transcript profiling analyses for parallel reductions in water contents for both species as well as a full desiccation drying series for S. stapfianus. We performed a detailed comparative genomic analysis for the two species and extended the analysis to include other grass species, both desiccation tolerant and DS. Our results offer insights into the mechanism and evolution of VDT in the Chloridoid grasses.

Results

Generating Genomic Resources for Sporobolus Grasses.

One-step flow cytometric assays generated size estimates for each of the Sporobolus genomes. The haploid genome of S. stapfianus had an average of 1,385 pg of DNA per nucleus, which is approximately equal to a complete genome sequence of 1.354 Gbp, and the haploid genome of S. pyramidalis had an average of 1,867 pg of DNA per nucleus, which is ∼1.826 Gbp (Table 1). Draft genome assemblies were generated for each grass using Illumina whole-genome shotgun sequencing combined with Chicago and Hi-C proximity ligation (Materials and Methods). The final assemblies consisted of 11,574 scaffolds with an N50 of 19.4 Mb for S. stapfianus and 2,518 scaffolds with an N50 of 21.6 Mb for S. pyramidalis, with the longest scaffolds for both species greater than 60 Mbp. Despite their high contiguity, the assembled genomes are smaller than the estimated genome size, at 1.080 and 1.055 Gbp for S. stapfianus and S. pyramidalis, respectively. These differences between the estimated and assembled genome sizes are likely caused by collapsed homologous regions in these complex polyploid species as described in detail below. Both genomes have similar levels of repetitive elements, 39.7 and 41.3% for S. stapfianus and S. pyramidalis, respectively (Table 2), with almost identical distributions of known repeat families (SI Appendix, Table S1). Gypsy and Copia retrotransposons are the most predominant families of the known repeats at 36 and 10 to 12%, respectively, for the two genomes.

View this table:
Table 1.

Estimation of the genome size (1C value) using flow cytometry

View this table:
Table 2.

S. pyramidalis and S. stapfianus genome assemblies

The Sporobolus genomes were annotated using MAKER with a combination of RNASeq and PacBio Iso-Seq full-length transcripts as expressed sequence tag (EST) evidence and protein homology from other high-quality plant genomes. After filtering, the final annotations contained 52,208 and 51,207 gene models for S. stapfianus and S. pyramidalis, respectively (Table 2). Annotation completeness was assessed using Benchmarking Universal Single-Copy Orthologs (BUSCO) with the poales_odb10.2019–11-20 database of 4,896 conserved genes. The genome annotations recovered 93.5 and 92.4% of complete BUSCOs for S. stapfianus and S. pyramidalis, respectively, indicating that both genomes were well annotated and contained the vast majority of the coding portion of these two genomes (Table 3). Gene models were functionally annotated using a simplified maize–GAMER pipeline; 96% of genes were annotated with InterProScan domain/family information, and 66% were annotated with Gene Ontology (GO) descriptions for both genomes.

View this table:
Table 3.

Genome assemblies BUSCO v4 statistics vs. the grass (poales_odb10) dataset

Comparative Genomics of Chloridoid Grasses.

Sporobolus belongs to the Chloridoideae subfamily of grasses, a large and diverse group of predominantly C4 species with remarkable drought, heat, and salinity tolerance. The orphan grain crops finger millet and teff are found within Chloridoideae, as are several model desiccation-tolerant plants in the genera Oropetium, Eragrostis, Tripogon, and Sporobolus among others. Most of the surveyed Chloridoideae species (∼90%) are polyploid, including species from many of the aforementioned taxa. The availability of several high-quality chloridoid genomes facilitates detailed comparative genomic comparisons within these grasses. Macrosynteny between S. stapfianus and S. pyramidalis shows a clear 2:3 pattern, consistent with the tetraploid and hexaploid nature of these grasses, respectively (Fig. 1 and SI Appendix, Fig. S1). Comparisons with the closely related diploid chloridoid grass O. thomaeum also revealed 1:2 and 1:3 patterns of synteny for S. stapfianus and S. pyramidalis, respectively, supporting their polyploidy (Fig. 1 and SI Appendix, Fig. S2). Although neither Sporobolus genome is scaffolded into complete chromosomes, large 20-Mb+-sized scaffolds are highly collinear with the Oropetium genome with few structural large-scale rearrangements (SI Appendix, Fig. S2), which is consistent with the unusually high conservation of karyotype and collinearity observed among other chloridoid grass genomes (22).

Microsynteny within Chloridoideae grasses. A collinear region between O. thomaeum, S. stapfianus, and S. pyramidalis is highlighted, reflecting the ploidy of each species (diploid, tetraploid, and hexaploidy, respectively). Genes are shown in blue and green, and syntenic gene pairs are connected by gray lines.
" data-icon-position data-hide-link-title="0">Fig. 1.
Fig. 1.

Microsynteny within Chloridoideae grasses. A collinear region between O. thomaeum, S. stapfianus, and S. pyramidalis is highlighted, reflecting the ploidy of each species (diploid, tetraploid, and hexaploidy, respectively). Genes are shown in blue and green, and syntenic gene pairs are connected by gray lines.

Macrosyntenic analysis between the Sporobolus species and O. thomaeum exposed an overall more complex polyploid structure than the more straightforward tetraploid and hexaploid compositions (SI Appendix, Fig. S2). Roughly half the hexaploid S. pyramidalis genome has the expected 3:1 pattern of syntenic blocks compared with O. thomaeum, while 37% is only 2:1. The pattern is similar for tetraploid S. stapfianus, where ∼44% of syntenic blocks are 2:1 to O. thomaeum as expected and ∼42% of blocks are 1:1 (SI Appendix, Fig. S2). Similar assembly issues were observed in the tetraploid chloridoid grass E. nindensis, where one to four regions were assembled for each syntenic region in O. thomaeum (10). These discrepancies, combined with differences between the estimated and assembled genome sizes, suggest the Sporobolus genomes were partially collapsed during assembly in homologous regions. S. pyramidalis and S. stapfianus may be segmental allopolyploids with varying degrees of homology between chromosomes from separate subgenomes. Partial collapse during assembly would result in divergent homologous regions assembling separately and highly similar regions collapsing, which is supported by the observation that the ratio of assembled syntenic blocks is maintained across large syntenic blocks and whole chromosomes in O. thomaeum. For instance, two homologous regions are assembled in S. pyramidalis for chromosomes 3 and 4 from O. thomaeum, while three regions in S. pyramidalis were identified for most of chromosome 2 in O. thomaeum. Similar patterns were observed between S. stapfianus and O. thomaeum. To account for these issues related to polyploidy, syntenic gene pairs and orthogroups were used for downstream comparative genomics and transcriptomics analyses between the Sporobolus genomes and other chloridoid grasses.

Transcriptional Landscape of Desiccation in S. pyramidalis and S. stapfianus.

We generated RNASeq data from RNA isolated from leaf tissues at different stages of dehydration for both species (SI Appendix, Fig. S3). Differentially expressed genes were identified using edgeR (23), and the resulting gene lists were assigned to GO biological process categories enrichment using the Cytoscape (23) plugin Bingo (24). These analyses indicate that S. pyramidalis and S. stapfianus transcriptomes respond differently to dehydration and share few biological process adaptations during the drying process. When water content decreases from 3 to 2 grams of water (gH2O) g−1 dry weight (dw), S. pyramidalis exhibits a strong response with 11,978 statistically differentially abundant transcripts (SDATs), in contrast to the more moderate response of 1,776 SDATs in S. stapfianus (Fig. 2 A and B). A GO enrichment analysis of SDAT lists further demonstrates that during the 3 to 2 gH2O g−1 dw water content transition, few biological processes are shared between the two species (Fig. 2 C and D and SI Appendix, Fig. S4). Some biological process categories, including response to heat and response to reactive oxygen species, are common to both species (SI Appendix, Fig. S4). Moreover, while S. pyramidalis responds to the change in water content from 3 to 2 gH2O g−1 dw by modulating processes involving the ribosome and the cell wall, S. stapfianus initiates alterations in the abundance of transcripts that relate to the response to oxidative stress, response to water deficit, and protein refolding (SI Appendix, Fig. S4).

S. pyramidalis and S. stapfianus transcriptional landscape during desiccation/rehydration. (A and B) Bar plots of the numbers of differentially expressed genes (FDR ≤ 0.01) for S. pyramidalis (A) and S. stapfianus (B) from edgeR contrasts of sequential conditions; 2g corresponds to the contrast 2 vs. 3 gH2O g−1 dw, 1.5g corresponds to 1.5 vs. 2 gH2O g−1 dw, 1g corresponds to 1 vs. 1.5 gH2O g−1 dw, and so on. The last S. stapfianus contrast is 24 h after recovery irrigation vs. 3 gH2O g−1 dw. The numbers of up- and down-regulated genes are indicated at the top and bottom of each bar, respectively. The skull and bones icon indicates that S. pyramidalis is severely affected when at 1 gH2O g−1 dw and enters into senescence. (C and D) Graphs of enriched GO biological process categories in the contrast 2 vs. 3 gH2O g−1 dw for S. pyramidalis (C) and S. stapfianus (D). Nodes represent categories and edges represent the parent–child relationships in the ontology. Node identities and positions are identical in both graphs. Color is proportional to the ratio of increased abundance vs. decreased abundance transcripts in the category, with a green color indicating a ratio of more than one (a majority of increased abundance transcripts) and a magenta color indicating a ratio of less than one (a majority of decreased abundance transcripts). Category identifications and names are listed in SI Appendix, Fig. S4.
" data-icon-position data-hide-link-title="0">Fig. 2.
Fig. 2.

S. pyramidalis and S. stapfianus transcriptional landscape during desiccation/rehydration. (A and B) Bar plots of the numbers of differentially expressed genes (FDR ≤ 0.01) for S. pyramidalis (A) and S. stapfianus (B) from edgeR contrasts of sequential conditions; 2g corresponds to the contrast 2 vs. 3 gH2O g−1 dw, 1.5g corresponds to 1.5 vs. 2 gH2O g−1 dw, 1g corresponds to 1 vs. 1.5 gH2O g−1 dw, and so on. The last S. stapfianus contrast is 24 h after recovery irrigation vs. 3 gH2O g−1 dw. The numbers of up- and down-regulated genes are indicated at the top and bottom of each bar, respectively. The skull and bones icon indicates that S. pyramidalis is severely affected when at 1 gH2O g−1 dw and enters into senescence. (C and D) Graphs of enriched GO biological process categories in the contrast 2 vs. 3 gH2O g−1 dw for S. pyramidalis (C) and S. stapfianus (D). Nodes represent categories and edges represent the parent–child relationships in the ontology. Node identities and positions are identical in both graphs. Color is proportional to the ratio of increased abundance vs. decreased abundance transcripts in the category, with a green color indicating a ratio of more than one (a majority of increased abundance transcripts) and a magenta color indicating a ratio of less than one (a majority of decreased abundance transcripts). Category identifications and names are listed in SI Appendix, Fig. S4.

As dehydration advances from 2 to 1.5 gH2O g−1 dw in S. pyramidalis, the functional categories of SDATs remain relatively unchanged from that activated at the initial loss of water, and as it is undergoing senescence during the 1.5 to 1 gH2O g−1 dw transition, further acclimation appears unlikely. By contrast, S. stapfianus exhibits an increase to 3,730 SDATs during the 2 to 1.5 gH2O g−1 dw transition, but starting at the 1.5 to 1 gH2O g−1 dw transition, it initiates a major remodeling of its transcriptome (SI Appendix, Fig. S3), as indicated by a significant increase to 14,557 and 16,047 SDATs during these two transitions in water content, respectively (Fig. 2D). Global transcriptional remodeling continues during the 0.75 to 0.5 gH2O g−1 dw transition, albeit at a lower degree, with 8,146 SDATs (Fig. 2D). When desiccated S. stapfianus plants are rehydrated, another strong transcriptome reprogramming, with 27,280 SDATs 12 h after rehydration, is evident and shifts to a transcriptome functional expression profile more similar to that of the fully hydrated control (SI Appendix, Fig. S3). Although S. stapfianus appeared morphologically fully recovered after 24 h of rehydration, the transcriptional profile is not equivalent to that observed in leaves of plants with a water content of 3 gH2O g−1 dw (SI Appendix, Fig. S3), with 24,659 SDATs between the two conditions (Fig. 2B). Leaves from plants 24 h after rehydration have up-regulated SDATs classified in ribosome biogenesis GO categories and down-regulated SDATs in photosynthesis categories, as well as remnants of stress-responsive adaptations, including the response to water categories, and altered metabolism, suggested by the presence of glucose 6-phosphate, fructose 1,6-bisphosphate, and several other metabolism-related categories (SI Appendix, Fig. S4B).

Syntenic Ortholog Transcripts Comparison of the Hydrated State in S. pyramidalis and S. stapfianus.

To directly compare the transcriptomes for S. stapfianus and S. pyramidalis and identify differentially regulated transcripts that relate to the differences between the two species in the hydrated state prior to dehydration, we created a custom list of syntenic ortholog "genes" (Materials and Methods). Differential expression was accomplished using a contrast S. stapfianus vs. S. pyramidalis in edgeR (23), and the resultant syntenic ortholog gene lists were probed with GO enrichment as described previously for the intraspecies dehydration transcriptome analyses. The analyses demonstrate that S. stapfianus and S. pyramidalis have very different transcriptional landscapes under hydrated conditions that reflect functionally different priorities for each species. The S. stapfianus transcriptome significantly favors nitrogen, starch, and photosynthetic metabolic processes, whereas the S. pyramidalis transcriptome significantly favors processes involved in growth, primarily the biogenesis of cell wall components (SI Appendix, Fig. S5A). These differences are also reflected at the cellular component and molecular levels (SI Appendix, Fig. S5 B and C), with the majority of cellular functions related to the chloroplast and photosystems in S. stapfianus and the symplast, cytoskeleton, cell wall, and cell wall modification activities in S. pyramidalis.

Syntenic Ortholog Protein Comparisons of Desiccation Response in S. pyramidalis and S. stapfianus.

To further compare the response of S. pyramidalis and S. stapfianus to dehydration, we performed a proteomic analysis using young leaves at 3 and 1.5 gH2O g−1 dw and focused on proteins encoded by syntenic genes in a comparison of enriched GO biological process categories of accumulating and decreasing proteins in both water content conditions (SI Appendix, Fig. S6). At 1.5 gH2O g−1 dw, S. pyramidalis had increased accumulation of proteins that are almost exclusively involved in stress responses; S. stapfianus had increased accumulation of stress response proteins but also, accumulated proteins involved in the response to misfolded proteins and protein catabolism (SI Appendix, Fig. S6A), and it decreased the abundance of proteins involved in energy production (SI Appendix, Fig. S6B). The protein data demonstrate that, as observed for the transcriptomic profiles, S. pyramidalis and S. stapfianus follow predominantly different approaches of protein accumulation in their response to dehydration.

Expression Dynamics of DS and Desiccation-Tolerant Grasses.

To explore the evolution of VDT in the Chloridoideae subfamily of grasses, we made use of several high-quality genomes with similar dehydration expression datasets that were available for this group of grasses: the desiccation tolerant (S. stapfianus, O. thomaeum, and E. nindensis) and the DS (E. tef and S. pyramidalis). To facilitate comparisons between species with different ploidy, we clustered genes into syntenic orthologs using MCScan (25) and orthologous groups (orthogroups) using OrthoFinder (26) and compared expression patterns between genes in the same orthogroups. We identified 49,418 orthogroups from OrthoFinder containing 806,075 genes across 23 diverse land plant genomes and focused the subsequent analyses on orthogroups, orthologs, or syntenic gene pairs present in the genome of all chloridoid grasses.

We first surveyed the global expression profiles of the five Chloridoid grasses under well-watered, drought/desiccation, and rehydration conditions using transformed expression data of 19,267 shared syntenic orthologs across all species. We applied a dimensionality reduction on the resulting expression matrix through principal component analysis. The first two principal components collectively explain 62% of the variance and separate the expression datasets by species and stress (Fig. 3). Well-watered RNASeq samples are found in a single tight cluster of all five species, while desiccation and rehydration samples are found in dispersed but distinct clusters. Samples from dehydration and rehydration time courses in the DT species fall into two clusters, with E. nindensis and O. thomaeum samples intertwined in one cluster and S. stapfianus in the second. The dehydration samples from the two DS species (E. tef and S. pyramidalis) clustered together in a third distinct cluster. Samples of E. nindensis and O. thomaeum are separated by relative water content in principal component (PC)1 and by dehydration vs. rehydration in PC2, but interestingly, they are not delineated by species. Together, these results indicate that expression patterns are broadly conserved in leaf samples of all species but that dehydration and rehydration samples are distinct between the three lineages of DT species and their DS relatives.

Dimensional reduction of drought expression profiles across DS and DT Cloridoid grasses. Raw expression values for syntenic orthogroups were transformed by z score prior to principal component analysis. The first two principal components are plotted for the two DS Chloridoid grasses (E. tef and S. pyramidalis) and three tolerant grasses (E. nindensis, O. thomaeum, and S. stapfianus) with comparative expression datasets. Points are colored by species or hydration state as indicated in the key.
" data-icon-position data-hide-link-title="0">Fig. 3.
Fig. 3.

Dimensional reduction of drought expression profiles across DS and DT Cloridoid grasses. Raw expression values for syntenic orthogroups were transformed by z score prior to principal component analysis. The first two principal components are plotted for the two DS Chloridoid grasses (E. tef and S. pyramidalis) and three tolerant grasses (E. nindensis, O. thomaeum, and S. stapfianus) with comparative expression datasets. Points are colored by species or hydration state as indicated in the key.

The same leaf RNASeq data were analyzed in a pairwise fashion to identify genes with significantly increased transcript abundance under dehydrating conditions in all five species. These SDATs were clustered based on orthogroup using OrthoFinder (as described above) and compared between species. Orthogroups were used in this set of analyses as they contained more genes than the synteny-based analyses, and orthogroups have better resolution of recently duplicated genes. Across the five sequenced chloridoid grasses, the largest number of up-regulated orthogroups under dehydrating conditions was observed between the two Sporobolus species (Fig. 4), as expected since they are sister taxa. The second largest number of up-regulated orthogroups was shared between the two Sporobolus species and O. thomaeum (Fig. 4), which is consistent with their phylogenetic placement within the Chloridoideae. Many other orthogroups are up-regulated similarly in all five species (Fig. 4). The orthogroups uniquely up-regulated in all VDT species are enriched in 214 biological process GO terms (SI Appendix, Fig. S7). Highly enriched GO terms include ultraviolet UV light response, chlorophyll catabolism, reactive oxygen species (ROS) metabolism, seed dormancy maintenance by abscisic acid (ABA), and gene expression in response to heat stress, among others (SI Appendix, Fig. S7A), These GO terms are consistent with well-characterized processes related to DT. Other GO terms with a lower magnitude of enrichment include those related to lipids, osmoprotectant biosynthesis, high light response, energy metabolism, protein degradation, and ABA signaling (SI Appendix, Fig. S7 B and C). Seventy-one biological process GO terms were uniquely up-regulated in only the DS species (SI Appendix, Fig. S8). These included several terms related to salicylic acid as well as ethylene and ABA signaling, arabinose biosynthesis, cell wall biogenesis, and notably, leaf senescence, among others (SI Appendix, Fig. S4). We then asked whether any of the GO terms uniquely up-regulated in DT species would overlap with those uniquely down-regulated in DS species and vice versa (SI Appendix, Table S3). The GO term "protein folding" was uniquely up-regulated in DT and down-regulated in DS species. Across these five species, most seed-related orthogroups are up-regulated similarly (SI Appendix, Fig. S9). There are no seed orthogroups that are up-regulated in all three DT species without also being up-regulated in one or more DS species.

Venn diagram of up-regulated orthogroups across the five surveyed chloridoid grasses. The number of overlapping orthogroups with up-regulated expression under drought is shown for each comparison.
" data-icon-position data-hide-link-title="0">Fig. 4.
Fig. 4.

Venn diagram of up-regulated orthogroups across the five surveyed chloridoid grasses. The number of overlapping orthogroups with up-regulated expression under drought is shown for each comparison.

Expansion of ELIPs.

ELIPs have a conserved role in photoprotection during desiccation, and they have undergone massive tandem gene duplication in all sequenced resurrection plant genomes surveyed to date (14). We observed a similar duplication of ELIPs in the Sporobolus genomes (Fig. 5A). The S. stapfianus genome has 65 ELIPs in three tandem arrays, and the S. pyramidalis genome has 30 ELIPs in two tandem arrays (Fig. 5B). The largest array in S. stapfianus has 49 ELIPs compared with 17 in its corresponding homologous region, suggesting the duplications occurred after the divergence of the two S. stapfianus subgenomes. Both O. thomaeum and S. stapfianus have large tandem arrays of ELIPs, but the duplication events originated from different syntenic orthologs. The total number of ELIPs in S. pyramidalis is higher than some other desiccation-tolerant species, but when gene counts are normalized for ploidy, the ELIPs are within the range of other sensitive grasses.

ELIPs tandem duplication in S. stapfianus and ELIP gene abundance in leaf tissues. (A) Microsynteny of two ELIP tandem arrays is shown in S. stapfianus. ELIPs are shown in red, other genes are shown in gray, and syntenic homeologs between the scaffolds are denoted by gray connections. (B) The number of ELIPs in sequenced Chloridoideae grasses (E. tef, S. stapfianus, S. pyramidalis, E. coracana, O. thomaeum, and Z. mays) is plotted. The two desiccation-tolerant grasses are denoted in red. (C) Log2-transformed gene abundance (TPM) of the 30 ELIPs in S. pyramidalis and 65 ELIPs in S. stapfianus across each replicate of the leaf desiccation time courses.
" data-icon-position data-hide-link-title="0">Fig. 5.
Fig. 5.

ELIPs tandem duplication in S. stapfianus and ELIP gene abundance in leaf tissues. (A) Microsynteny of two ELIP tandem arrays is shown in S. stapfianus. ELIPs are shown in red, other genes are shown in gray, and syntenic homeologs between the scaffolds are denoted by gray connections. (B) The number of ELIPs in sequenced Chloridoideae grasses (E. tef, S. stapfianus, S. pyramidalis, E. coracana, O. thomaeum, and Z. mays) is plotted. The two desiccation-tolerant grasses are denoted in red. (C) Log2-transformed gene abundance (TPM) of the 30 ELIPs in S. pyramidalis and 65 ELIPs in S. stapfianus across each replicate of the leaf desiccation time courses.

ELIPs have little to no detectable expression in well-watered tissue, but they are highly induced in desiccating S. stapfianus leaf tissue after they reach 1.0 gH2O g−1 dw, and their expression continues 12 and 24 h postrehydration (Fig. 5). ELIPs are also up-regulated under drought in S. pyramidalis, and this occurs quickly in the dehydration process at 2.0 and 1.5 gH2O g−1 dw. However, their combined expression is less than S. stapfianus (Fig. 5C), similar to what has been observed in other DS grasses (14).

Discussion

The genomic resources we developed for the sister species S. stapfianus and S. pyramidalis offer a robust contrast that facilitates a strong comparison between a VDT and a DS grass species. The addition of the genomic resources from other resurrection grasses, O. thomaeum (8) and E. nindensis (10), broadens the comparison further into the Chloridoideae subfamily of grasses. The two genome assemblies revealed the complex mixed ploidy of these two grasses, with S. stapfianus primarily tetraploid and S. pyramidalis primarily hexaploid. The structural complexity of the two genomes likely contributed to the inability to assemble the genomes into chromosome-level contigs or to record sequenced genome sizes equivalent to those determined cytologically. The increase in ploidy between the two species probably occurred immediately after the divergence of the S. pyramidalis clade from the common ancestor of the two species (18). The assemblies did not reveal any genomic structural characteristics, with the exception perhaps of tandem arrays of ELIP genes (14), that could be attributed to the difference in VDT between the two species, which is consistent with the general observation that there is not a genomic "blueprint" for VDT in resurrection species (4). However, the assemblies did allow for a thorough comparative analysis, both structural and functional, of the gene space for each genome, and coupled with the in-depth transcriptome data, we were able to explore a detailed genomic assessment of the dehydration/desiccation responses within the Sporobolus sister species contrast.

The generation of transcriptomic and proteomic data for dehydrating young leaf tissue at specific water contents during a dry-down experiment such that the dehydration levels are survivable for both grasses provides a broad assessment of the stress response for each species. DS S. pyramidalis mounted a messenger RNA (mRNA)-level response to an initial drop in hydration as has been observed for the majority of dehydration-sensitive plants (27, 28). However, as dehydration to 1.5 gH2O g−1 dw was reached, the transcript abundance response declined dramatically (Fig. 2A), perhaps as the leaf water content reached a critical level for S. pyramidalis. The leaves of S. pyramidalis are wilted at 1.5 gH2O g−1 dw (21) but otherwise, appear undamaged, so it is tempting to speculate that the decline in the transcript abundance response may be related to wilting and perhaps, loss of turgor during wilting in S. pyramidalis. Although S. pyramidalis responds quickly to a loss of water, the early increased transcript abundance response appears to be focused on protein translational processes and transcripts common to heat and cold stress (SI Appendix, Fig. S4), and only later, as dehydration deepens, do transcripts associated with proline metabolism (osmoregulation) and redox proteins, common to water-deficit responses (27), accumulate. The early decline in transcripts involved in photosynthesis and cell wall homeostasis is also common to the dehydration response in most angiosperms (4). The later decline in transcripts that are associated with general biosynthetic processes is consistent with the general lack of a metabolic response to dehydration seen in metabolite profiling studies of S. pyramidalis at similar levels of water loss (21). Desiccation-tolerant S. stapfianus, in contrast, exhibited a significantly different qualitative transcriptional response to dehydration with a low-magnitude response in the early phase of dehydration. With the comparatively muted response and although there are some common transcript abundance responses between the two species, S. stapfianus clearly targets remodeling a completely different functional aspect of the transcriptome than does S. pyramidalis at similar water contents. Indeed, it appears that S. stapfianus targets the accumulation of transcripts that function more in stress-related activities unlike S. pyramidalis, which does not. The differences between the two transcriptional responses for the two species were unexpected as other studies have indicated that there was extensive overlap in functionality of the transcriptomes of both sensitive and tolerant grasses exposed to dehydration (10). Although there are a few common transcript abundance functional categories in the early response to dehydration in both species, it is clear that the overall transcriptome remodeling during dehydration is very different between them, as exemplified by the different dehydration thresholds for the accumulation ELIP transcripts.

For S. stapfianus, the primary remodeling of the transcriptome during dehydration appears to occur as the plants reach the 1.0- to 0.75-gH2O g−1 dw part of the drying curve, which appears to be a critical period in the desiccation response of all resurrection angiosperms studied so far (29) and concurs with early microarray data (30). In S. stapfianus, the transition from 1.0 to 0.75 gH2O g−1 dw occurs during leaf curling (19) and is likely at water contents just prior to and during a change in membrane fluidity that occurs as leaf water potentials approach −12 MPa (4). The functional aspects of the transcriptome remodeling during desiccation of S. stapfianus leaves have been documented previously and are in accord with the observation that transcript abundance is concordant with changes in metabolism associated with cellular protection aspects of DT (30). There was a dramatic alteration of the transcriptome upon rehydration of S. stapfianus leaves, which likely reflects the complex nature of the dehydration event. The magnitude of the change in the transcriptome, reflecting a change in abundance of at least half of the known transcripts, and the functional processes they represent indicate not only the stress incurred from the inrush of water and mechanical aspects of cellular expansion but also, the need to repair damage (from both desiccation and rehydration), reactivate energy metabolism, and reinstate the physiological integrity of the cells and tissues (4). The observation that transcripts encoding proteins involved in ribosome biogenesis are accumulated and those encoding proteins involved in photosynthesis have not recovered control levels at 24 h following rehydration highlights the extent of the impact that desiccation and rehydration have on plant cells and tissues even in DT plants. S. stapfianus requires between 48 and 72 h to regain the structural and physiological integrity seen in well-watered plants (19, 31).

The remodeling of the transcriptome in response to dehydration starts from two very different resting-state (fully hydrated) transcriptomes. Our functional analysis of the gene-level expression of the syntenic orthologs of the sister grasses, although somewhat confounded by the structural complexity of the two genomes, revealed that for S. stapfianus, the biosynthesis of starch and nitrogen compounds was perhaps a priority for young leaves under normal conditions, while for young leaves of S. pyramidalis, the priority appeared more focused on the construction of cell walls. Although somewhat speculative, the increase in nitrogen compounds, primarily amino acids from a combination of new synthesis and redistribution, was the focus of a recent study that demonstrated that these compounds are apparently used to fuel central metabolism or for other metabolic adjustments related to the acquisition of DT, such as osmoregulation (32). The differences in priorities are consistent with the changes in protein abundance from 3 to 1.5 gH2O g−1 dw. Although S. pyramidalis protein abundance changes did not reflect cell wall processes, perhaps due to the difficulty in extracting the majority of wall...