Orthologs definition and name assignment

To determine orthology among genes of both species, a Basic Local Alignment Search Tool for proteins (BLASTp - V. 2.12.0+) [13] analysis was performed among S. lycopersicum and S. pennellii proteome V. ITAG3.1 [14] and Spenn v2 [15], respectively using by default parameters and defining orthologs as the best bidirectional blast hit [7]. Additionally, to contrast the ortholog assignment, Orthofinder V.2.2.5 [16] was employed with default parameters.

Gene symbol for each tomato identifier were assigned by performing a single BLASTp analysis of S. lycopersicum against the Arabidopsis thaliana proteome (Araport11 version) [17] deriving each name from the best hit. Then, the orthologs of tomato in S. pennellii were assigned the same symbols.

RNA-seq data analysis

To analyze the transcriptomic patterns, we used RNA-seq experiments for S. lycopersicum and S. pennellii under control and drought conditions (11 days of drought treatments). Data was retrieved from the Sequence Read Archive database under project number PRJNA800740 [18]. Once it was downloaded, their sequencing quality was analyzed using FastQC V.0.11.9 [19], adapters were removed using TrimGalore V.0.6.7 [20], and gene expression quantification was performed using Kallisto V.0.46.2 [21] with default parameters and the same genome versions employed in ortholog detection (GenBank accession ID GCA_000188115 for S. lycopersicum and GCA_001406875 for S. pennellii). Then, to determine quantification by orthologs, we normalized the expression by transcript length in kilobases using equation 1 to test the effects of variations in the size of the ortholog in both species, similarly described by Eshel et al. [6]

With Q represents the discrete quantification value, E is the expression level, and L is the transcript length in kilobases. Additionally, the expression in its non-length normalized form was also tested in order provide a more comprehensive assessment of potential biases in the quantification process.

Expression models assignment

Expression models were fitted based on ortholog expression patterns in S. lycopersicum and S. pennellii under control and drought conditions. For this purpose, the expression was quantified and normalized by gene length (see RNA-seq Analysis) and differentially expressed ortholog genes were identified using DESeq2 [22], both intra and inter-species, with an adjusted p-value threshold of 0.05.

Then, a rule-based decision tree like algorithm was developed (S1 Fig., S1 File - decision tree and plots.ipynb) using differentially expressed genes and foldchange reported by DESeq2 to categorize ortholog patterns into the following models:

• • Stress-ready: Gene expression levels under control conditions in one species match those under drought conditions in the other species (i.e., transcript levels in one plant under drought match those of the other plant under control conditions). There are four cases for this criterion: (Case 1) Gene expression in S. pennellii increases under drought to match the levels of S. lycopersicum under control conditions; (Case 2) Gene expression in S. pennellii decreases under drought to match the levels of S. lycopersicum under control conditions; (Case 3) Gene expression in S. lycopersicum increases under drought to match the levels of S. pennellii under control conditions; (Case 4) Gene expression in S. lycopersicum decreases under drought to match the levels of S. pennellii under control conditions.
• • Shared response: Ortholog expression increases or decreases in both species under drought (Case 1: Upregulated or Case 2: Downregulated).
• • Unique response: Expression changes occur only in one species. Upregulated or downregulated in S. lycopersicum (Case 1 and 2, respectively) or upregulated or downregulated in S. pennellii (Case 3 and 4, respectively).
• • Opposite response: Expression changes occur in both species but in opposite directions. Case 1: Gene expression in S. lycopersicum is upregulated whilst it is downregulated in S. pennellii. Case 2: Gene expression in S. lycopersicum is downregulated whilst S. pennellii it is upregulated.
• • No response: Expression levels remain unchanged under all conditions, with basal levels differing between species (higher or lower in S. pennellii, Case 1 and 2, respectively) or being identical (Case 3).

GO term enrichment analysis

The Gene Ontology (GO) terms for tomatoes were obtained from the GOMAP project [23]. These terms were used to conduct an enrichment analysis using the GOATOOLS software V1.4.12 [24] through the find_enrichment script using default parameters for each case in expression models, assuming the same term for each ortholog. Finally, we filtered out GO terms that were significantly less concentrated (purified) in the study group compared to the general population and had a False Discovery Rate (FDR) lower than 0.05.

Additionally, in addition to the background comparison, a negative control was performed for each type of response using a random selection of genes, with the same number of genes as those identified for each response, which was carried out in triplicate.

Motif enrichment in promoters and Gene Regulatory Network modeling

Transcription Factors (TF) binding motifs for S. lycopersicum and S. pennellii were downloaded from PlantTFDB V.5.0 (https://planttfdb.gao-lab.org/) [25]. TFBM were scanned using the Find Individual Motif Occurrences (FIMO) V.5.4.1 [26] into promoters of unmasked genomes to retain the full sequence information (2Kb upstream of each gene starting coordinate) with q-value < 0.05. Due to the predictive nature of the information deposited in PlantTFDB, XSTREME V.5.4.1 [27] was employed to identify enriched sequence motifs. This approach enabled us to contrast and validate the findings obtained by FIMO.

To determine motif enrichment in the promoters of genes of interest, a one-tailed Fisher’s Exact Test was performed between the group of interest and all other gene groups (e.g., comparing orthologs in stress-ready Case 4 against genes in the shared, unique, opposite, and no response groups). The alternative hypothesis was accepted with a P value < 0.05, and no adjustment method was applied to avoid overlooking potentially important motifs. Additionally, as a negative control, the presence of random motifs was compared against the rest of the genes similarly to the negative control in the GO term enrichment analysis, which was carried out in triplicate.

Finally, gene regulatory networks for both species were developed using the Python version of GENIE3 [28], with the normalized gene counts from both control and drought conditions as input. A filtering step was then applied, retaining only the top 10% of regulatory factors based on their reported importance. Additionally, edges were kept only if the gene’s promoter contained the corresponding TFBM, to enhance network confidence. Once GRNs were obtained, to determine the similarity between them we compute the Jaccard index as described in equation 2 [29].

With E₁ and E₂ representing the sets of directed edges in both networks, and J denote the Jaccard Index, which quantifies the similarity between these GRNs (i.e., edges with the same node pairs and directions) relative to the total number of unique directed edges present in both networks.

Orthologous gene expression patterns in response to drought in S. lycopersicum and S. pennellii

To uncover differential gene expression patterns between the two species under control and drought conditions, it was first necessary to identify orthologous genes. Using bidirectional BLASTp analysis (Fig 1A, S2 Fig.) as previously described [6], we identified 21104 orthologous genes between S. lycopersicum and S. pennellii. Additionally, we identified 13325 and 23861 unique genes in S. lycopersicum and S. pennellii, respectively (S2A Fig. and S Table 1). The presence of these non-orthologous genes may be attributed to the high number of proteins with similar sequences within the genome, such as potential paralogs (S3 Fig. A) [30]. Furthermore, Orthofinder comparison identified fewer orthogroups by partitioning these putative paralogs into distinct groups that reflect one-to-one, one-to-many, or many-to-many relationships (S. Fig 3B), thus generating a complex challenge to determine differential expression and assigning gene expression models.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Top ten genes with higher variation under Stress ready (case 2) in S. pennellii.

https://doi.org/10.1371/journal.pone.0324724.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Comparative analysis of differential gene expression in response to drought between S. lycopersicum and S. pennellii.

(A) Bar chart showing the number of differentially expressed genes (DEGs) in response to drought in S. lycopersicum and S. pennellii. Upregulated genes are shown in red, and downregulated genes are shown in blue. S. lycopersicum exhibits a higher number of DEGs (6,442 total) compared to S. pennellii (4,049 total), with both species showing more upregulated genes than downregulated ones. (B) Volcano plots displaying the significance (−log10 adjusted p-value) versus fold change (log2 fold change) for DEGs in S. lycopersicum (middle) and S. pennellii (right). Red dots represent genes that meet both fold-change and statistical significance criteria, while blue dots indicate genes significant only by adjusted p-value. Green dots indicate genes with significant fold change but not statistically significant. Grey dots represent genes with no significant changes.

https://doi.org/10.1371/journal.pone.0324724.g002

After identifying putative orthologs between the two Solanum species, their expression levels were quantified using RNA-seq from Moreira et al. (17) under two conditions: control plants maintained under well-watered conditions and plants subjected to an 11-day drought treatment, in which watering was withheld until the soil reached 20% of field capacity. The Principal Component Analysis (PCA) revealed distinct expression patterns in orthologs, which were influenced not only by species differences, consistent with clade distances [31], but also by environmental conditions, such as drought and control treatments, and corroborated by measuring the Spearman correlation coefficients among replicates (S2B and S2C Fig.)

Furthermore, differential expression analysis showed that a larger number of orthologs are differentially expressed in S. lycopersicum compared to S. pennellii (6,442 versus 4,049, respectively) (Fig 2A). However, the proportion of upregulated and downregulated genes is consistent within each species, with more genes being downregulated overall and comparable levels of change observed between the two species (Fig 2B and 2C; S2 Table). Additionally, the analysis using data not normalized by gene length yielded similar results, although there were slight variations in p-values and log2foldchange (S3 Table). However, a one-to-one comparison of orthologous genes between the two species is required to distinguish shared and unique responses to drought.

Stress-ready gene expression highlights S. pennellii preadaptation to drought

To identify shared and unique patterns of ortholog expression, differential expression results were analyzed using a rule base algorithm to categorize each ortholog to an exclusive expression model (Fig 1B, S1 Fig., S1 and S2 Tables). Our findings indicate that the majority of orthologs show no significant changes in expression levels under drought, classifying them into the “No response” group. However, among the orthologs with differential expression, most are categorized as either unique or shared responses, comprising approximately 43% and 35% of the total, respectively (Fig 3).

A distinct subset of orthologs exhibits transcriptional preactivation or repression, termed stress-ready. These orthologs are defined as genes whose expression levels in one species under control conditions align with those in another species under stress conditions [6,7]. This group accounts for nearly 20% of all orthologous genes that alter their expression in response to drought, suggesting a form of molecular preadaptation. Fig 3 illustrates the four distinct cases of stress-ready gene expression patterns in orthologous genes between S. lycopersicum and S. pennellii under control (C) and drought (D) conditions. Case 1: Orthologs in S. pennellii under drought conditions reach the expression levels observed in S. lycopersicum under control conditions (217 genes, 2.8%). Case 2: Orthologs in S. pennellii maintain higher expression levels under control conditions, aligning with S. lycopersicum expression under drought conditions (230 genes, 3.0%). Case 3: Orthologs in S. lycopersicum under drought conditions reach the expression levels observed in S. pennellii under control conditions (570 genes, 7.4%). Case 4: Orthologs in S. lycopersicum under control conditions align with expression levels observed in S. pennellii under drought conditions (537 genes, 7.0%).

Overall, these results highlight the distinct gene expression responses between the two species, with stress-ready genes constitute nearly one-third of the unique responses under water deprivation, underscoring their importance in drought adaptation. Notably, S. pennellii appears to be preadapted to drought, as evidenced by the higher number of Case 3 and Case 4 orthologs, where S. lycopersicum orthologs align with the expression levels of S. pennellii. This suggests that S. pennellii has evolved molecular and regulatory mechanisms enabling it to maintain a stress-ready state, providing a significant advantage in surviving arid environments.

Enriched biological processes highlight distinct drought-response mechanisms in S. lycopersicum and S. pennellii

To gain insights into the biological functions underlying shared and exclusive gene expression patterns, including stress-ready genes, a Gene Ontology (GO) enrichment analysis was conducted (Fig 1B). Within the shared response group, orthologs that exhibited increased expression under drought in both species were enriched for genes associated with seed dormancy, water deprivation, abscisic acid (ABA) response, and oxidative stress, among other processes (S4 Table and S4 Fig). Conversely, orthologs with decreased expression under drought were associated with RNA metabolism, peptide biosynthesis, energy-related processes such as ATP and NADPH biosynthesis, photosynthesis (e.g., Photosystem II), and DNA replication, including regulation of the G2/M transition of the mitotic cell cycle (S4 Table and S4 Fig). These findings align with previous studies in S. lycopersicum, which identified similar GO categories as being affected by drought [10,11,32].

In contrast, unique responses showed distinct enrichments depending on the case. For Case 1 (genes induced by drought only in S. lycopersicum), enriched functions included mRNA processing, as seen in shared responses, along with histone and RNA modifications (S4 Table and S4 Fig). For Case 2 (genes repressed by drought only in S. lycopersicum), was primarily related to energy processes, including Photosystem II assembly and repair, NADPH regeneration, and amino acid biosynthesis (S4 Table and S4 Fig). Genes in Case 4 (induced by drought only in S. pennellii) were enriched for translation and protein synthesis processes (S4 Table and S4 Fig). Notably, no GO enrichment was found for Case 3 (genes induced by drought only in S. lycopersicum). These results highlight distinct drought-response strategies, were S. lycopersicum predominantly repress orthologs involved in energy-related processes, while S. pennellii uniquely upregulates translation and protein synthesis pathways, suggesting an adaptive mechanism that bolsters protein synthesis under drought stress.

For orthologs classified as stress-ready, significant GO enrichment was observed in Case 2 and 3. In Case 2, where S. pennellii orthologs are repressed under drought and match the expression levels of S. lycopersicum under control conditions, enriched functions included nucleosome assembly and DNA conformation changes, particularly genes encoding nucleosome histone core components (Table 1, Fig 4A, and S4 Table). For Case 3, where S. pennellii orthologs are preactivated and match the drought-induced expression of S. lycopersicum, showed enrichment in RNA processing, protein import into the nucleus, and transcription-related processes. Additionally, to verify whether the normalization process and the binary rules applied to the decision tree-like algorithm (e.g., differentially expressed vs. non-differentially expressed genes, positive vs. negative fold-change) effectively distinguish between different gene expression models, their expression patterns were examined (Fig 4B). Within this category, drought response genes that display these specific stress-ready expression patterns were identified (Fig 4C).

Finally, none of those processes were found to be enriched in the negative control (S5 Table), and thus, these findings suggest that stress-ready genes in S. pennellii are closely linked to gene expression regulation, playing a key role to plant preadaptation to arid environments. Additionally, the presence of drought-related genes within this group highlights a complex interaction between epigenetic regulation, gene expression, and mechanisms underlying adaptation to adverse environmental conditions (Table 2, Fig 4, and S4 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Top ten genes with higher variation under Stress ready (case 3) in S. lycopersicum.

https://doi.org/10.1371/journal.pone.0324724.t002

S. pennellii exhibits differential regulatory elements enriched in unique expression patterns

Since S. pennellii exhibits distinct gene expression patterns, including unique and stress-ready states that may contribute to its drought tolerance, transcription factors (TFs) potentially regulating these adaptations were examined. To do so, the FIMO tool was used to determine transcription factor binding motifs (TFBMs) within the promoters of genes grouped in each expression model in both species. A Fisher’s exact test was then applied to evaluate the statistical significance of motif enrichment in these responses compared to a genomic background (Fig 1C).

Our analysis revealed that specific TF families were enriched in distinct gene categories. For instance, TFBMs from the GATA family were enriched in the promoters of S. pennellii genes categorized under unique response 1. Specifically, three GATA TF motifs were significantly enriched in S. pennellii for this expression pattern, while no enrichment was observed in S. lycopersicum orthologs (Fig 5A and S5 Fig). GATA motifs were found in the promoters of 65 out of 1,293 orthologs in S. pennellii (Fig 5A and 5C). In contrast, two bZIP family motifs were exclusively enriched in S. lycopersicum under the unique response 3 category, with an additional bZIP member enriched in the unique response 4 category (Fig 5A and S5 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Distinct drought response patterns in S. pennellii highlight unique stress-ready and species-specific mechanisms.

Each line plot shows normalized expression levels for the two species under control (C) and drought (D) conditions. The total number of genes and percentage of the transcriptome for each case is indicated in parentheses. Stress-ready: Genes that are primed for drought in one or both species across four cases, with more pronounced responses in S. pennellii. Shared response: Genes exhibiting similar expression patterns under drought in both species across two cases. Unique response: Genes displaying species-specific expression patterns across four cases, particularly highlighting the unique responses of S. pennellii. Opposite response: Genes that show inverse expression changes between the two species under drought across two cases. No response: Genes that show minimal or no expression change under drought in each specie across three cases.

https://doi.org/10.1371/journal.pone.0324724.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Stress-ready orthologs exhibit distinct expression patterns, contributing to specific biological processes.

(A) GO term enrichment analysis for stress-ready orthologs in cases 2 and 3, shown on a green scale representing −log10(P value). (B) Mean standardized expression values (Z-scores) for the most enriched processes in the second (yellow) and third (green) cases of stress-ready responses. (C) VST transfiorrmed expression levels of two orthologs from Stress-ready case 2 (yellow bar) and two from case 3 (green bar). These orthologs were selected due to their role under drought reported in Arabidopsis thaliana orthologs.

https://doi.org/10.1371/journal.pone.0324724.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Differential motif enrichment between gene categories and tomato species.

(A) Heatmap showing the enrichment of transcription factor (TF) families in the promoters of genes categorized under stress-ready and unique response models in S. lycopersicum and S. pennellii. The x-axis represents different gene expression categories, and the y-axis represents TF families. Numbers in parentheses show the quantity of TFBM family members for each specie (S. lycopersicum and S. pennellii respectively) deposited in PlantTFDB. Numbers in cells indicate the number of distinct TFs within each family that are enriched in each gene category. The color gradient represents the number of different TFs, with darker shades of green indicating higher numbers of enriched TFs. (B) Bar chart showing the percentage of ERF and TCP TF motifs enriched in S. lycopersicum ortholog promoters. (C) Bar chart showing the percentage of ERF, MYB, GATA, and TCP TF motifs enriched in S. pennellii ortholog promoters. The numbers above bars indicate the number of enriched TF motifs relative to the total number of gene promoters analyzed for each category.

https://doi.org/10.1371/journal.pone.0324724.g005

For stress-ready orthologs, motifs from the TALE, DOF, and MIKS MADS TF families were exclusively enriched in S. pennellii under stress-ready case 3. Interestingly, Trihelix and MYB motifs were enriched in S. pennellii promoters under stress-ready case 4. Notably, the ERF family exhibited the highest number of enriched TF motifs, particularly in S. pennellii promoters under stress-ready case 3. ERF motifs were present in approximately 12% of the 570 ortholog promoters in S. lycopersicum, compared to 20% in S. pennellii (Fig 5B and Fig 5C), with De novo motif discovery also recovers ERF family motifs in S. pennellii (S6 Table). Finally, to validate the specificity of TFBM enrichment, Fisher’s exact test on triplicate sets of randomly selected promoters was performed. These analyses did reveal some inconsistent motif enrichments, appearing significantly only in one of the three replicates in each model of expression (S7 Table). Although these findings arise from in-silico methods, they underscore the consistency of the predictive data deposited in PlantTFDB and provide additional insight into the observed motif enrichments.

These results suggest that TFs from the ERF, GATA, MYB, and TCP families play pivotal roles in regulating drought-response gene expression. Promoters of S. pennellii genes exhibited significant enrichment across multiple unique and stress-ready states, with the ERF family in particular likely contributing to the enhanced stress-adaptive capacity of S. pennellii under drought conditions, with TFBM enrichments are likely not merely artifacts of random genomic sampling but potentially reflect genuine regulatory signatures.

ERF transcription factors contribute to the regulation of stress-ready genes

As stress-ready case 3 contains the highest number of stress-ready genes and exhibits specific enrichment of TF motif elements in S. pennellii, we further investigated the potential role of individual TFs from the ERF family in regulating genes within this category. Certain TF motifs were shared between the two species, including ESE3 and DEAR2 (Fig 6A). The ESE3 motif was found in the promoters of 58 S. lycopersicum genes and 29 S. pennellii genes, while the DEAR2 motif was detected in 14 S. lycopersicum and 9 S. pennellii promoters (S6 Fig.), where the higher number of promoters harboring these motifs in tomato could be related to the necessity of increased expression of these genes under drought conditions. Notably, DEAR2 followed the stress-ready Case 3 expression pattern in both species, as highlighted in red in Fig 6A.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. S. pennellii shows greater enrichment of individual ERF TF motifs in gene promoters compared to S. lycopersicum.

(A) Bar chart illustrating the number of genes with specific TFBMs in their promoters for both S. lycopersicum (blue) and S. pennellii (orange). S. pennellii has a higher number of genes with motifs associated with drought response, including SHN3, CRF4, and DEAR2, compared to S. lycopersicum. (B) Percentage of gene promoters containing TFBMs belonging to specific gene ontology (GO) categories overrepresented in the third category of stress-ready. The red dashed line at 50% highlights that S. pennellii consistently exceeds this threshold for drought-related motifs, suggesting a functional association. (C) High-confidence GRN illustrates ERF family members (represented by triangle nodes) regulating stress-ready gene orthologs (represented by rectangle nodes) in S. pennellii and S. lycopersicum (indicated by black and red edges, respectively). Node colors represent orange for stress-ready case 3, purple for stress-ready case 4, sky blue for no response case 3, yellow for no response case 2, and red for shared response case 1.

https://doi.org/10.1371/journal.pone.0324724.g006

Several ERF family TFs were exclusively identified in S. pennellii, including SHN3, DREB2C, Sopen11g018710, and CRF4, each targeting 10 or more genes. These TFs were associated with biological functions relevant to stress-ready Case 3, as indicated by GO term enrichment (Figs 4 and 6). Furthermore, nearly 50% or more of the genes containing these TF motifs were involved in RNA metabolic processes and transcriptional responses under drought (Fig 6B), highlighting the critical regulatory role of these TFs in controlling key biological processes.

To explore potential co-regulatory events and determine shared mechanisms between the two Solanum species, we constructed high-confidence gene regulatory networks (GRNs) for each species. We identified the top 10% of most relevant TF-target gene pairs using the Random Forest-based algorithm GENIE3 and integrated this with TFBM presence in promoters (Fig 1C). Four ERF members —Solyc11g045680/Sopen11g017570 (LEP), Solyc04g080910/Sopen04g034520, Solyc02g077360/Sopen02g022260, and Solyc11g042580/Sopen11g018710— were excluded due to their low expression levels lower than 10 raw reads counts [22] (S7 Fig). Although the resulting GRNs for S. pennellii and S. lycopersicum had a similar number of nodes and edges and contained many orthologous TFs (Table 3 and Cytoscape session available on GitHub at https://github.com/JMALab/stress-ready/blob/main/supp.%20file%201%20-%20network.cys), the number of shared regulatory edges was minimal. The Jaccard similarity coefficient was computed to be 0.044 prior to FIMO and 0.019 after its application, indicating that only 1.9% of the directed edges were shared between the two networks. Therefore, these results suggest that, in addition to the importance of each regulation of orthologous genes in each species, there is also a difference in the TFBMs present between the species.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Overview of resulting GRNs.

https://doi.org/10.1371/journal.pone.0324724.t003

Examining the expression of ERF family members enriched in one or both GRNs revealed that most fall into the “No response” category, suggesting that mechanisms beyond simple up- or downregulation of gene expression may drive their regulatory roles (Fig 6C). Moreover, distinct regulatory events were also observed between the two species. While most target genes were regulated by a single TF, co-regulation was evident for genes with two or more regulators. For example, DREB2C and Sopen10g030870 co-regulate Sopen03g037640, while SHN3 appears to function as a primary independent regulator (Fig 6C). These findings suggest that S. pennellii has evolved specialized regulatory mechanisms to control the stress-readiness of orthologous genes, enhancing its ability to tolerate drought conditions.