Do Species Mount Different Genetic Responses to Acute Drought Stress Across an Ecological Gradient?

Background: How do xerophytic species thrive in environments that experience extreme annual drought? Although critical to the survival of many species, the genetic responses to drought stress in non-model organisms are unknown. We investigated this question in Mentzelia section Bartonia (Loasaceae), which occurs throughout western North America, including arid lands. To better understand the genetic responses to drought stress among species that occur in different habitats, the gene expression levels of three species from the genus Mentzelia that occur across a precipitation gradient were compared. Two de novo reference transcriptomes were generated and annotated. Leaf and root tissues were collected from control and drought shocked plants and compared to one another for differential expression. A target-gene approach was also implemented to better understand how drought-related genes from model and crop species function in non-model systems. Results: When comparing the drought-shock treatment plants to their respective control plants, we identied 165 differentially expressed clusters across all three species. Differentially expressed genes including those associated with water movement, photosynthesis, and genes that delayed senescence. The transcriptome proling approach was coupled with a target genes approach that measured expression of 90 genes associated with drought tolerance in model organisms. Comparing differentially expressed genes with a log-fold values of 2 or greater between species and tissue types showed signicant differences in drought response. In pariwise comparisons, species that occurred in drier environments differentially expressed greater genes in leaves when drought shocked than those from wetter environments, but expression in the roots mostly produced opposite results. Conclusions: Arid-adapted species mounted greater genetic responses compared to the temperate species, with differences in leaf and root tissue responses. Target genes revealed differences and similarities between functional processes within Mentzelia and other plant groups. Differences in drought response depending on tissue type suggests that genetic and physiological responses occur within one tissue type and not the other, and tissue responses could be evolving at different rates. We determined that drought tolerance is enhanced through pathways of delayed senescence and that genetic responses were tissue and habitat specic.

similarities between functional processes within Mentzelia and other plant groups. Differences in drought response depending on tissue type suggests that genetic and physiological responses occur within one tissue type and not the other, and tissue responses could be evolving at different rates. We determined that drought tolerance is enhanced through pathways of delayed senescence and that genetic responses were tissue and habitat speci c.

Background
As sessile organisms, land plants have been coping with and adapting to drought stress since rst evolving onto land. Despite the potentially lethal consequences of drought stress, some plants only occur in arid regions that experience severe, annual drought. To cope with drought stress, plants have evolved multiple adaptive responses, which include morphological, metabolic, cellular, and physiological processes (Bartlett et al., 2016;El Ha d et al., 1998;Li et al., 2016). Understanding the physiological and molecular processes involved in drought-stress adaptation is crucial as climates become hotter and drier (Bartlett et al., 2016). Studying the physiological and evolutionary adaptations plants have evolved to maximize tness in arid environments will provide insight into what drives adaptive responses, as well as how vegetation will respond to climate change.
Although multiple studies have predicted ecosystem responses to instances of drought in model and agricultural species (Bartlett et al., 2016;El Ha d et al., 1998;Hoover et al., 2017;Wei et al., 2015;Xu et al., 2010), non-model species might exhibit different responses when experiencing drought, which includes variation in the order of physiological responses, as well as the functional gene groups that differentially express in response to drought stress (Hoover et al., 2017). When drought-stress signals are received, signal transduction leads to the induction of both physiological and metabolic processes (Le et al., 2011). Consequently, molecular and physiological responses are usually linked, in which a change in gene expression causes a physiological change (e.g., stomatal closure). How species adapt to drought stress at the gene-expression level, however, is less well-characterized than the physiological processes they underlie, and could vary, or even converge, across plant groups.
The expression of genes associated with stress regulation plays an important role in drought response . If plants regulate the impact of drought through molecular pathways, measuring the expression levels of genes that lead to physiological responses will illuminate how plants respond to stress in their natural environment. By discovering genes that are involved in drought response, future studies will be able to distinguish how adaptive evolution in conjunction with physiological plasticity facilitates or constrains drought responses (DeBiasse and Kelly, 2015). Although we have learned much from studies on how crop species and other model organisms respond to drought, little known about how the remaining species of plants are able to cope with drought stress on a genetic level, especially those in arid environments.
Plants have likely evolved numerous responses to mitigate drought stress, which includes genetic, physiological, and morphological adaptations. Geography can play a further role, in which species that occur in xeric environments might respond differently to drought stress than those that occur in mesic environments. Although studies that have examined drought stress responses have shed much light on how a species responds to drought stress, experimental comparative studies of species across an environmental gradient have the potential to further tease apart how species adapt to drought stress. If we compare, for example, a group of closely related species that occur in xeric and mesic environments, hypotheses can be formulated and tested to determine how species have adapted genetic responses to drought stress.
We formulated four hypotheses to explain the evolution of drought responses in a comparative framework. The rst hypothesis (HA-1) proposes that xerophytic species have a stronger genetic response when exposed to drought conditions than temperate species. If supported, we predict that xerophytic species would have a signi cant difference in gene expression, whether up or down-regulated, whereas temperate species would have no signi cant difference in gene expression. Hypothesis HA-1 suggests that the annual drought stress that xerophytic species experience has selected for genetic responses and/or selection has not favored adaptive responses at the genetic level in temperate species.
A second hypothesis (HA-2) posits that temperate species have a stronger response when exposed to drought conditions than xerophytic species. If supported, this would mean that temperate species have a signi cant difference in gene expression, whether the difference is in genes that have been up or down-regulated, whereas xerophytic species shows no signi cant difference in gene expression. If supported, this hypothesis would suggest that morphological adaptations in xerophytes mitigate stress caused by drought.
A third hypothesis (HA-3) proposes that all species have similarly strong responses when exposed to drought conditions, suggesting species that occur across an environmental graient have inherited genetic responses to drought from an ancestor that itself adapted to drought conditions. If supported, hypothesis HA-3 might explain why species of some clades have been successful in inhabiting arid environments.
Our nal hypothesis is a null hypothesis (Ho) which proposes that all species fail to mount a genetic response to drought, exhibiting negligible differences in gene regulation. The null hypothesis, if supported, would suggest that plants do not respond to drought at the molecular level, rather, they have other means, such as morphological adaptations.
Non-model plant systems might provide novel responses or mechanisms not identi ed in other modelplant systems, which could revolutionize the way crops are genetically modi ed for drought tolerance, in addition to adding to the knowledge of known genes and responses associated with drought. Because of research bias and potential undiscovered stress responses within plants, it is important to study species that are adapted across a wide range of habitats, including arid ecosystems. To test the above hypotheses, we subjected populations from three species in the genus Mentzelia L. (Loasaceae) that occur across an environmental gradient to a drought shock experiment and compared them to nondrought individuals. Transcriptomic expression levels were compared to determine the suites of traits that were differentially expressed in the drought shocked individuals.

Differential expression in roots and leaves
Among the 24 leaf and 24 root tissues sampled from three species, corset produced 64 (Figure 2).
The DE analyses with a false discovery rate P-value adjustment resulted in 165 DE cluster-transcripts across all species and tissue types, and among them, 140 were identi ed to homologs (Figure 3). The 25 remaining DE cluster transcripts were not assigned to homologous sequences and, therefore, not analyzed further. Mentzelia lifolia differentially expressed 103 transcripts in leaf and four in root tissues. Nine transcript-clusters were signi cantly up-regulated in M. lifolia leaf tissue and 94 were signi cantly down-regulated (Table S1). The most down-regulated genes were Adenylate cyclase proteins, which were reduced by a logFC of -8.4. Up-regulated transcript-clusters were categorized as endonuclease activity, wound response, membrane components, and nitrate reductase (NADH) activity. A differential expression analysis of root tissue from M. lifolia produced four down-regulated transcript-clusters (Table S2).
Mentzelia reverchonii differentially expressed two transcripts in leaf and 53 in root tissues. The DE analysis of leaf tissue resulted in two over-expressed transcript-clusters, each categorized as being involved in the oxidation reduction process (Table S3). There were 28 down-regulated and 25 upregulated transcript-clusters in the root tissue. The most highly up-regulated cluster was expressed with 7.0 logFC and was categorized as participating in ubiquitin-protein transferase activity and protein ubiquitination (Table S4).
Mentzelia speciosa differentially expressed three transcripts in leaf and none in root tissue. Two clusters were up-regulated and a single transcript-cluster was down-regulated (Table S5).
Species that occur in more xeric habitats have a stronger response than those that occur in less xeric Target gene expression analysis The target gene analysis included 90 target genes (Tables S6-S8). Because multiple transcript-clusters were identi ed as functions from the same process through GO annotating, multiple transcript-clusters associated with a single target gene. In total, 140 clusters matched to the targeted genes in M. lifolia leaf tissues and 44 out of 140 clusters were included in the edgeR exact test analysis and were not ltered out before the analysis. Using a target gene approach, we were able to compare co-expression of genes across species. From the 12 target genes that resulted in a logFC ≥ 2, we were able to compare how they are expressed similarly between the three species and tissue types. Results are summarized in Figure 5, Table S6, Figures S3-S10.

Discussion
We conducted a comparative transcriptomic study of Mentzelia to understand how plants have adapted to and will respond to drought stress across an environmental gradient. The xerophytic M. lifolia and semi-arid M. reverchonii had a greater response to drought exposure based on the number of signi cantly DE transcript-clusters. Using results from the FDR P-values, we conclude that the temperate M. speciosa responded less when experiencing drought shock than the arid and semi-arid adapted species, corresponding with the hypothesis that the xerophytic species will mount a greater molecular response than the temperate species. Although both arid species mounted a greater response than the temperate species, they showed opposite patterns in the tissues that responded. For M. lifolia, the mechanisms that allowed for drought tolerance occurred in the leaf tissue, whereas the genetic response in M. reverchonii was more apparent in roots than leaves. Each species could have a different reaction time to drought stress, and although one was responding strongly in leaf tissues, the other was in a phase of coping through root tissues. The response timing could have been affected by individual plant size according to Drake-Schultheis et al.(2020). Plant age affects water and carbon availability, resulting in younger and smaller plants being more susceptible to drought shock (Drake-Schultheis et al., 2020). The differences in response in tissues might suggest that the cascade of physiological events within leaves and roots are different between species.
Delayed senescence was differentially expressed in leaves and roots of M. lifolia and M. reverchonii. Mechanisms associated with delayed senescence, like jasmonic acid production, auxin regulation, and Serine/threonine-protein kinase CHK1-like protein, were identi ed among the DE transcript-clusters. The ability to delay processes resulting in cell death might be a response that has evolved through years of drought exposure. The historical severity and length of drought experienced by each species could be a factor that determines whether delayed senescence is a pathway that is utilized. Risking embolism and dehydration in order to delay senescence is an interesting process that should be further studied in xerophytes because it seems to play a major role in their adaptive response drought stress.
The three species in our experiment mounted similar, but not identical, genetic responses when exposed to drought. Many of the selected target genes were not found in our data set, however, because of ltering parameters and the large number of comparisons made during the statistical analysis, many could have been removed from the nal dataset. Many target genes experienced greater than two logFC, which is biologically signi cant and points to common molecular change that occur in response to stress across distantly related species. Although few target genes resulted in signi cant changes, we determined that genes that play a role in drought tolerance in other plant species are also acting within the plant systems of Mentzelia.
Mentzelia speciosa exhibited a genetic response to drought, but because it is exposed to drought stress less often, it might have responded less quickly to stress, suggesting a less e cient and effective response when compared to other species. If the process was rapid and failed quickly, the time of tissue collection could have missed the point with highest differential expression. Alternately, if the mechanisms triggered by drought in M. speciosa were slow to act, the time of tissue collection could have occurred before the genetic response.

Differentially expressed transcript-clusters in leaves
The greatest response to drought tolerance occurred in the leaves M. lifolia (Table S1), which is unsurprising given that leaves are the main site of evapotranspiration. The large response of downregulated transcript-clusters implies that metabolic processes are being shut down in order to conserve water and prevent tissue and cell damage. Transcript-clusters were categorized as transmembrane associated proteins, nucleic acid binding proteins, auxin regulation, rDNA transcription, endonuclease activity, or other enzymatic activities. Four transcript-clusters were identi ed as proteins involved in stress response, senescence, and wound response. Suppressing drought induced leaf senescence greatly increases drought tolerance in transgenic lines of Nicotiana tabacum (Rivero et al., 2007), and, therefore, down-regulation of senescence associated proteins to increased drought tolerability in M. lifolia is unsurprising. Rivero et al. (2007) determined that delaying senescence in transgenic lines of N. tabacum increased processes of reactive oxygen species scavenging, leading to extra protection for the photosynthetic apparatus that increased water use e ciency under drought stress. Jasmonic acid, a phytohormone responsible for signal transfer in response to senescence (Koo et al., 2013), was signi cantly down-regulated. Down-regulating jasmonic acid, a known senescence accelerator (Ke et al., 2015), would delay senescence, which appears to be a large component of how M. lifolia mitigates drought stress in leaves.
Many of the down-regulated transcript-clusters consisted of transcription factor proteins associated with auxin production and transport. Auxin response factors are associated with drought responses because of their role in hormonal response signaling as well as developmental and senescence processes (Rahman, 2013). When exposed to drought stress, developmental processes would most likely stop, inducing down-regulation of auxin. A single auxin related protein was up-regulated in leaf tissues as well. Ke et al. (2015) showed that when a transgenic line of poplar was designed to overproduce auxin, drought stress tolerance was increased. Auxin metabolism is monitored by many other metabolic pathways and plays a large role in overall plant homeostasis (Ke et al., 2015). Similar to the downregulation of senescence, auxin regulation may be delayed to prevent the occurrence of senescence related processes, but a single occurrence of up-regulation could be attributed to its role as a hormonal response signal or other developmental processes. Overall, leaf tissues in M. lifolia have a large response of down-regulation involved with senescence brought on by drought stress, and general metabolic processes, like photosynthesis, which would prevent water loss and cell death by attempting to maintain sustainable levels of homeostasis while regulating levels of reactive oxygen species (Miller et al., 2010).
Of the seven up-regulated transcript-clusters in the leaf tissue of M. lifolia, two were categorized as endonucleases. Endonucleases act to remove introns by breaking the phosphodiester bonds to produce functional mRNA (Xue et al., 2006). The up-regulation of endonucleases might be degrading mRNA (Schoenberg, 2011). Messenger RNA degradation prevents the translation of proteins involved in metabolic processes that could cause destabilization under stress. Two proteins identi ed as integral components of the cellular membrane were up-regulated, one speci cally identi ed as TIP1-1 aquaporin protein transmembrane transport, whereas the other is a form of glucosyltransferase involved in the accumulation of the yellow pigment crocetin during fruit development (Nagatoshi et al., 2012). Aquaporins primarily function to transport water through the cellular membrane, but Zhou et al. (2012) determined that the up-regulation of a PIP2 subgroup of aquaporin proteins enhanced drought tolerance in tobacco by increasing the ability to retain water, limit oxidation activity, and decrease the need for antioxidant activity. Aquaporins play an important role in drought stress; however, whether aquaporin proteins are up or down-regulated depends on the species and tissue (Santoni et al., 2003).
Mentzelia reverchonii leaf tissues up-regulated two photosynthetic enzymatic processes (Table S3). Ribulose bisphosphate carboxylase/oxygenase (RuBisCO) was up-regulated, suggesting that photosynthesis signaling increased because it cannot function, or it is deactivated when ratios of ATP/ADP become unfavorable due decreased photosynthesis (Crafts-Brandner and Salvucci, 2000). The metabolic responses typically associated with drought occur as a response to oxidative stress, rather than a direct response to water deprivation (Flexas et al., 2006). Because M. reverchonii has become accustomed to longer periods of drought stress, its initial response to water deprivation might be to increase carbon assimilation to prepare reserves for stress levels that are intolerable by increasing photosynthetic processes instead of immediately shutting down by reducing gas exchange through stomatal closure.
Mentzelia speciosa leaf tissues responded to drought stress by up-regulating polyubiquitin-like proteins (Table S5). The most common functional role of ubiquitin is the intracellular control of protein content and degradation (Sharma et al., 2016). Ubiquitin production under stress degrades non-drought stress proteins, which would impair the response to drought stress in an e cient way (Sharma et al., 2016).

Differentially expressed transcript-clusters in roots
Mentzelia lifolia root tissues up-regulated cellular respiration, while downregulating the production of secondary metabolites and the oxidation-reduction process (Table S2). When photosynthesis is slowed in the above-ground organs of the plant, carbon is allocated from the root's carbon sinks (Hasibeder et al., 2015). Although Hasibeder et al. (2015) determined that root respiration decreased under prolonged drought conditions in Trisetetum avescentis, the initial response of down-regulated photosynthesis might result in higher levels of respiration in roots (Flexas et al., 2006). Up-regulated respiration suggests that recovery efforts might occur to offset the decreased photosynthetic rate (Flexas et al., 2006).
Mentzelia reverchonii roots had a large response to the drought treatment (Table S4). Two of the 54 differentiated transcript-clusters were involved with ubiquitin activity. Increasing the expression of an enzyme responsible for protein degradation would play a direct role in expressed proteins and overall metabolic function (Sharma et al., 2016). Similar to M. lifolia leaves, auxin transport was up-regulated in a single transcript-cluster, suggesting that another function associated with auxin transport was increased, such as hormonal signaling (Rahman 2013).
Transcript-clusters up-regulated in M. reverchonii roots were associated with transmembrane transport, speci cally sodium-calcium transmembrane transport and cellulose synthase. Calcium transport plays a crucial role in drought, salinity stress signaling, and osmoregulation (Hu and Schmidhalter, 2005  The target genes NAC-56 and PGR5-like protein 1A were expressed in M. lifolia roots. NAC-56 in root tissues up-regulate target genes that aid in drought tolerance (Chung et al., 2018). No apical meristem transcription factors in transgenic-rice roots enhance drought tolerance by targeting genes responsible for changing root architecture (Chung et al., 2018). PGR5-like protein 1A, a thylakoid transmembrane protein, was downregulated. PGR5-like protein 1A plays a direct role in the photosynthetic cyclic electron ow that transport electrons to produce ATP (Hertle et al., 2013). Downregulating the ow of electron transport would decrease or shut down photosynthesis productivity.
Mentzelia reverchonii leaves down-regulated target genes of 9-cis-epoxycarotenoid dioxygenase NCED3, Aquaporin PIP2-1, and Guard cell S-type anion channel SLAC1, while roots up-regulated 9-cisepoxycarotenoid dioxygenase NCED3 and down-regulated Auxin transporter protein 1 ( Table 7), suggesting that the photosynthetic process, or at the very least stomata and transport channels, are being shut down. 9-cis-epoxycarotenoid dioxygenase is a key enzyme in ABA biosynthesis and is induced by drought stress to control the level of endogenous ABA produced (Iuchi et al., 2001). In M. reverchonii, leaves are downregulating, but roots are upregulating ABA. The down-regulation of plasma membrane intrinsic protein aquaporin might be due to it being a low expression aquaporin when constitutively expressed. The down-regulation of a negative regulator of guard cell anion like the R-type channel that responds rapidly to cystolic Ca2+ (Sah et al., 2016), and might be what is utilized in Mentzelia (Schroeder & Keller, 1992). An auxin transport protein was down-regulated in the roots of M. reverchonii. Downregulation of molecular and cellular components associated with senescence, like auxin, is regularly down-regulated in Mentzelia, and might be a driving factor in how multiple species tolerate drought stress.
Three target genes were DE in leaf and root tissues of M. speciosa (Table S7). Leaf tissues overexpressed Dehydration-responsive element-binding protein 2A, which is a gene element that helps regulate expression of genes utilized to cope with drought (Qin et al., 2008). Over expression within leaves might be the rst response of the cascade of mechanisms plants use to avoid damage from drought stress. A single target gene, NAC domain containing protein 52, was up-regulated in M. speciosa. Similar to M.
lifolia, NAC-56 in root tissues is a NAC transcription factor that up-regulates a group of target genes that aid in drought tolerance (Chung et al., 2018), and the up-regulation of NAC-56 might be to change root architecture to adapt to drought stress (Chung et al., 2018). The gene probable pectate lyase 8 (PLY8) was down-regulated within root tissues of M. speciosa. PLY8 is necessary for lateral root growth after inhibition by abscisic acid (Zhao et al., 2014). Down-regulation is probably attributed to inhibition through ABA, which is produced in response to drought stress in roots.

Tissue response between species
Considering the logFC ≥ 2 analysis, leaves and roots responded differently depending on species.
When leaf tissue response was compared between species, the species that occur in more xeric habitats mounted a stronger response than those in less xeric habitats (Figure 4). However, when comparing responses in root tissues, M. lifolia mounted the weakest response compared to the other two species. The genetic response, consequently, appears to be organ and environmental dependent. Although M. speciosa mounts a greater response to drought in root tissues compared to M. lifolia, overall M. speciosa responded less to drought shock compared to both M. reverchonii and M. lifolia.
The differences in tissue response suggests that drought response is tissue, species, and environment speci c, and selection pressures related to drought response might be acting on tissues in different ways and at different rates. Further research is needed to determine if selection is acting on leaf and root tissues separately. The cascade of physiological response and water regulation described by Bartlett et al. (2016) could be tissue speci c and instead of studying response as a function of the whole plant, future research should focus on speci c tissue responses. Manipulation of genes that play a role in roots or leaves independently might lead to novel pathways for genetic modi cation allowing for greater drought tolerance that is not only tissue speci c, but serves to enhance drought tolerance overall.

Conclusions
Roots and leaves respond and cope with drought through different pathways. Manipulation of genes that respond differently to drought in roots or leaves might lead to novel pathways for genetic modi cation, allowing for greater drought tolerance that is not only tissue speci c, but serves to enhance drought tolerance overall. Delayed senescence was a common mechanism used in both leaves and roots of M.
lifolia and M. reverchonii. The ability to delay cell death might be a response that has evolved through years of drought exposure. Risking embolism and dehydration in order to delay senescence is an interesting process that should be further studied in xerophytes because it seems to play a major role in their adaptation to drought stress. Mentzelia speciosa produced the weakest response to drought, which was unsurprising considering it was found in the most mesophytic environment compared to the other two xeric species , M. lifolia and M. reverchonii. The individuals of M. speciosa had been exposed to less frequent occurrences of drought, potentially leaving them without the evolved ability to respond in an e cient and timely way. We also observed differences in drought response depending on tissue type suggesting that species could have a mounted response within one tissue type and not the other, and tissue responses could be evolving at different rates. Our results suggest that, in additional to morphological evolution to limit drought stress, xerophytes have evolved a cascade of genetic responses that have tissue-speci c responses to mitigate drought stress through delayed senescence, decreased photosynthesis, and decreased water transport.

Methods
Species of Mentzelia occur across a wide environmental gradient, from southwestern North American deserts, to mesic habitats near the Continental Divide in the Rocky Mountains (Schenk, 2013a). Despite their ecological importance across western North America, and especially in drought-prone gypsum outcrops (Schenk, 2013 a, b), we do not understand the mechanisms behind their drought tolerance and success in xeric habitats. Three species of Mentzelia were sampled that occur across an environmental gradient, from desert to mesic ecosystems (Fig. 1). The xerophytic M. lifolia was sampled in the New Mexican Chihuahuan Desert. Mentzelia reverchonii, a semi-arid species, was collected in the Texas shortgrass prairies. The mesophytic M. speciosa was sampled in the central Rocky Mountains of Colorado. Natural populations of all three species were sampled because plants failed to grow in greenhouse conditions. The three sampled species belong to the same section within Mentzelia, section Bartonia, but are not each other's closest relatives (Schenk and Hufford, 2011). At the time of collection, the M. lifolia population in New Mexico received 2.16 cm of precipitation during July, and has a mean annual precipitation of 19.71 cm (Eischeid et al., 1995). The M. reverchonii population in Texas received 4.24 cm of precipitation for June, with a mean annual precipitation of 53.19 cm (Eischeid et al., 1995). The M. speciosa population in Colorado received 3.3 cm of precipitation in July, and has a mean annual precipitation of 55.60 cm (Eischeid et al., 1995). All habitats experienced average rates of rainfall for July according to precipitation data from the past 30 years. and four treatment plants that were developmentally identical (bolted from the rosette stage with owers present) were selected randomly while maximizing distance between them to avoid sampling closely related individuals. Four treatment plants were excavated with their roots intact and placed on the ground in full sun. We refer to this approach as a drought-shock treatment, which has been used in other studies to examine the response to drought in natural populations (e.g., Nyguyen et al., 2015; Rizhsky et al., 2002). While the drought-shock treatment was being conducted, four control plants were excavated with roots intact and sampled immediately to avoid sampling drought-stressed tissues. Control plants were excavated prior to leaf sampling to ensure that any wound response associated with the extraction from the ground would be identi ed in both the control and treatment plants, which would result in no differentially expressed (DE) genes associated with wounding after applying our bioinformatics pipeline (see below). Leaf and root tissues were collected in replicates of three, while ower, fruit, and stem tissues were collected from one individual per population for the purpose of generating a reference transcriptome. Mature leaf tissues with no insect damage were immediately placed in 1.5 mL of RNAlater TM RNA Stabilization Solution (Thermo Fisher Scienti c, Waltham, Massachusetts, U.S.A.) to preserve the RNA. Root tissues were sampled after the leaf tissues were sampled. Excess soil was removed by brushing off the roots, but much of the remaining soil washed away before RNA extraction.
Approximately 100 grams of tissue from the middle of the central tap root was sampled. The treatment plants were subjected to an hour of full sun exposure until leaves began to wilt and curl, indicating that the plants were experiencing acute drought stress. All tissue samples were placed in a -80°C freezer approximately 3-5 days after collection until the RNA was extracted. Vouchers for the collected population were pressed and deposited in the Georgia Southern University Herbarium (GAS).

RNA extraction
Tissue samples were thawed, removed from RNAlater, and placed into new tubes with 2.8 mm ceramic beads. Samples were frozen and homogenized with a Qiagen Tissuelyser II (Qiagen, Hilden, Germany) for one minute at 30 Hz. TRIzol extraction buffer (Thermo Fisher Scienti c) was added in aliquots of 1 mL to each sample, homogenized for nine additional minutes, and then incubated for ve minutes at room temperature. The phase separator, 1-bromo-3-chloropropane, was added in 100 μl aliquots, vortexed for 15 seconds, and incubated at room temperature for ve minutes. Samples were centrifuged at 12,000 g for 10 minutes. The upper aqueous phase was transferred to a new tube and 500 μl of chilled isopropanol was added. The samples were stored overnight in a -80°C freezer. After approximately 14 hours, the RNA samples were removed from the freezer and centrifuged at 12,000 g for 10 minutes. The supernatant was discarded, 1 mL of chilled 75% EtOH was added, and samples were centrifuged for ve minutes at 12,000 g. All supernatant was removed, and samples were air dried in a fume hood for 30 In order to ensure that each transcriptome was complete and an adequate representation of both species, HISAT2 v2.0.5 (Pertea et al., 2016) was used to map back the trimmed sequence reads from each sample. Average alignment percentages were calculated to ensure at least an 80% alignment rate average. Trinotate (https://trinotate.github.io/) was used to annotate the comprehensive and functional de novo transcriptome.

Expression analysis
Burrows-Wheeler Aligner (BWA) v0.7.13 ) determined transcript level abundance by mapping low divergence transcript sequences to a reference transcriptome. We employed the M. lifolia and M. speciosa reference transcriptomes to conduct reference-guided assemblies. Reference transcriptomes for the respective species were used as the target inputs, using the reference from the closely related M. speciosa for M. reverchonii (Schenk and Hufford, 2010). Each transcriptome was made into an FM-index to compress full text les for faster alignment rates. After the indices were made, the options for a mismatch penalty of 0.05, a gap open penalty of one, no output lower than 10, 20 threads, and "mem" option for local alignment of transcripts back to the reference were used for BWA analysis. SAMtools v1.3 ) was used to convert the SAM output les from BWA to sorted BAM les, which were inputted into Corset v1.07 (Davidson and Oshlack, 2014) that hierarchically grouped transcript contigs into clusters by shared reads and expression data. Counts of the number of transcripts included in each cluster were made and used as the input raw count data for differential expression analyses. The edgeR package (McCarthy et al., 2012; Robinson et al., 2010) was used for differential expression analysis in R (R Core Team, 2017). Transcript-cluster count les were read in by species and tissue type separately (e.g., root tissues of M. speciosa), with individuals grouped together by control or treatment. The DGEList function was used to create an object from the transcript cluster-counts for each species and tissue type individually. Transcript clusters with fewer than one transcript count per million in fewer than six of the eight individuals were discarded to reduce the number of rarely expressed genes that were not DE across all members of a group. Normalized factors were calculated to scale each library size. Common dispersion was calculated to maximize the negative binomial, conditional common-likelihood to estimate a common dispersion value across all genes. Tagwise dispersion was estimated with an empirical Bayes method based on weighted conditional maximum likelihood (McCarthy et al., 2012; Robinson et al., 2010). We used the exact test to determine differences in mean values between the two negative binomially distributed counts. A false discovery rate P-value adjustment was used to address multiple comparisons. We also compared DE clusters with log-fold changes of two or greater.
Differentially expressed clusters were annotated with Blast2Go (Conesa and Gotz, 2008) to identify their gene or protein name, along with a description and function.
Bivariate plots were generated in R, and differentially expressed genes with a log-fold values of 2 or greater were indicated. Differential expression values were estimated by comparing the treatment to the control for each species. A slope of one is expected if expression levels between species are identical, and we tested whether expression signi cantly deviated with the one-sample slope-test with the smatr package (Warton et al. 2012) in R. Statistical tests were conducted by constructing distributions for the test statistic following Taskinen et al. (2011).

Target gene approach
Complementary to the transciptome pro ling approach, we applied a target gene approach to search for expression patterns in drought associated genes. In comparison to approaches that strictly pro le the transcriptome for signi cantly DE genes, a targeted approach has the ability to determine the exact levels of differential expression (Marra et al., 2012) and whether genes associated with drought response are expressed at all. The target approach can determine commonalities in stress response across plants, allowing us to identify genes that commonly or uniquely respond to drought. We took advantage of previous studies to create a list of drought-tolerant genes and their sequences, then explicitly measured the response of the targeted genes to determine if Mentzelia responded similarly to drought stress.
We selected 90 genes from different gene families known to play a role in drought response (e.g., Chen  , as well as genes with GO terms related to or associated with drought tolerance from GenBank (Benson et al., 2005). The target genes from Arabidopsis, corn, soy bean, and sorghum were downloaded from GenBank. The entry name for each target gene was searched within sequence annotation reports from the reference transcriptomes to identify the associated transcript-cluster's ID. The edgeR results from the differential expression exact test for each species and tissue type was searched using the annotated genes or protein names to identify if the associated gene clusters and the level of gene expression for each target gene.

Declarations
Availability of data and materials The datasets generated and/or analysed during the current study are available in the NCBI GenBank repository: ncbi.nlm.nih.gov/genbank/. Raw sequence reads have been deposited in the Sequence Read Archive (Accession number: To be submitted while manuscript is in review), while the reference transcriptomes have been uploaded to the Transcriptome Shotgun Assembly Sequence Database through GenBank (Accession numbers: To be submitted while manuscript is in review).

Competing Interests
The authors declare that they have no competing interests.