To compare effects of acute UVR exposure on S. caninervis and S. ruralis, we grew isolates of both in a single white fluorescent light environment with no UVR. Shoots from an isolated clone of S. caninervis from southern Nevada, United States (Stark NV-107, United States, Nevada, Clark County, Newberry Mts, Christmas Tree Pass; UNLV) and an isolated clone of S. ruralis from Calgary, Alberta, Canada (Brinda 9108, Canada, Calgary, Bow River; UNLV) were cultivated in a growth chamber set to an 18-h photoperiod (18°C light and 8°C dark), at ca. 30 μmol m^–2 s^–1 PAR. The light source was 24″ F20 T12 GE Plant and Aquarium bulbs, warm tone, 3100 K, producing 750 lumens (GE Lighting, Boston, MA, United States). Cultures of a single genotype for each species were grown in lidded approximately 77 mm × 77 mm × 97 mm Magenta GA-7 plant culture boxes (bioWORLD, Dublin, OH, United States) from fragments on 1.2% agar made with an inorganic nutrient solution (Hoagland and Arnon, 1950).

After five months of growth without UVR exposure, the mature plants were subjected to a UVR exposure time series in triplicate. First, in order to filter out UV-C radiation, which is not present in solar radiation that reaches earth but may be in artificial UVR sources, culture box lids were replaced with 7.6 cm × 7.6 cm (3″ × 3″) acrylic windows, 3.175 mm (1/8 in) thick (Polycast Solacryl SUVT acrylic, Spartech, Maryland Heights, MO, United States), sealed to the culture boxes with wax film. These UV-transmitting windows transmit at least 90% across the visible and UV-A/B spectrum and then drop to near 0% transmittance near the boundary of UV-C (between 275 nm and 250 nm)¹. The sides of each culture box were wrapped in aluminum foil to ensure all light reaching plant cultures passed through the installed window. Prepared culture boxes were placed under four T8 reptile bulbs (ReptiSun 10.0 UVB, Zoo Med Laboratories Inc., San Luis Obispo, CA, United States).

Treatment light environment consisted of ca. 36 μmol m^–2 s^–1 broadband UV-AB radiation with a UV-B fluence rate of 0.36 mW cm^–2 and ca. 120 μmol m^–2 s^–1 PAR. This UV-B radiation level was selected to be within the approximate range of global irradiance (Pinedo et al., 2006), with intention to expose plants to a realistic level of UV-B radiation while ensuring the dose was sufficiently high to be detected by the plants. Broadband UVR and PAR were measured with LightScout UV and Quantum Sensors and the LightScout Sensor Reader (Spectrum Technologies, Aurora, IL, United States) while UV-B fluence was measured at several locations under the lamps with a handheld radiometer that was last calibrated in 2014 and independently evaluated in 2016 (SKU 430, Apollo Display Meter, Skye Instruments Ltd., Llandrindod Wells, United Kingdom). All sensors were covered with the same UV-transmitting acrylic as the samples prior to measurements and matched to the distances between plant, window, and light source. The first tissue collection was made after 10 min of exposure (T₁₀) and 30 min for the second (T₃₀). At the same time, control samples that were never subjected to a UVR treatment were collected (T₀). At collection, tissues were quickly snipped at the base in an effort to collect above-ground tissues and minimize agar collection, placed into 1.6 mL microcentrifuge tubes with a push-pin hole in the top, flash-frozen in liquid nitrogen, and stored at −80°C until further processing.

Frozen tissue was sent to Novogene (Novogene, Sacramento, CA, United States) for RNA extraction, library preparation, and transcriptome sequencing. There were three replicates made for each of the three treatments for both species, yielding 18 samples total. Total RNA was extracted with the Zymo Quick-RNA Plant Kit (Zymo Research, Irvine, CA, United States). RNA samples were processed according to standard Novogene protocols, including preliminary quality checks with gel electrophoresis followed by quantification and purity assessment with NanoDrop (Thermo Fisher Scientific, Waltham, MA, United States), and sample integrity assays with a Bioanalyzer 2100 (Agilent, Santa Clara, CA, United States). All 18 samples had an RNA integrity (RIN) score of at least 7. After quality checking procedures, oligo(dT) beads were used to enrich eukaryotic mRNA and rRNA was removed with the Illumina Ribo-Zero kit (Illumina, Inc., San Diego, CA, United States). RNA samples were then reverse-transcribed into double-stranded cDNA libraries by randomly fragmenting mRNA with fragmentation buffer, followed by adding random hexamers primer, a custom second-strand synthesis buffer (Illumina, Inc., San Diego, CA, United States), dNTPs, RNase H, and DNA polymerase I to initiate the second-strand synthesis. Next, after terminal repair and adaptor ligation, library preparation was completed with size selection and PCR enrichment. Library quality was assessed with Qubit 2.0 (Thermo Fisher Scientific, Waltham, MA, United States) to test preliminary library concentration, a Bioanalyzer 2100 (Agilent, Santa Clara, CA, United States) to test the insert size, and quantitative PCR to precisely quantify the library effective concentration size. Finally, libraries were sequenced on the 150 bp paired-end Illumina NovaSeq 6000 platform (Illumina, Inc., San Diego, CA, United States).

Transcriptomic data were first cleaned with Trimmomatic version 0.39 (Bolger et al., 2014) using a sliding window of four base pairs with a Phred quality score cutoff of 20, a minimum length of 20, and with a leading and trailing minimum of three. Bowtie2 (Langmead and Salzberg, 2012) and Tophat2 (Kim et al., 2013) were used to make indexes of the reference S. caninervis genome (Silva et al., 2020) for mapping and assembly. If multiple isoforms of a gene were detected in the transcriptome, they were binned and counted per parent gene. Transcripts with fewer than 10 read counts were discarded before downstream analyses. To account for overdispersion, transcript counts were transformed with the variance-stabilizing transformation (VST) prior to visualization with principal components analyses (PCAs) on all 18 samples together as well as on the nine samples from each species separately.

In order to quantify transcripts involved in the acute UVR response in S. ruralis and S. caninervis, reads were first mapped to the reference S. caninervis genome (Silva et al., 2020) with Tophat2 (Kim et al., 2013). Htseq-count version 0.9.1 (Anders et al., 2015) was used to estimate read counts per sample per gene and final analyses were performed in R (R Core Team, 2019) using DESeq2 (Love et al., 2014) to test for differential transcript abundance in plants at each point in the UVR treatment time series in each species. To identify sets of significant transcripts, P-values were first corrected with the Benjamini-Hochberg correction (Benjamini and Hochberg, 1995) to account for the false discovery rate (FDR) of multiple tests (Jafari and Ansari-Pour, 2019), followed by an adjusted P-value (P-adj) cut-off of 0.05 and an absolute value of the log2-fold change (LFC) minimum of 1. Sets of significantly increased or decreased transcripts at each time point in each species were compared and checked for overlap. Additionally, to test if distributions of Gene Ontology (GO; Gene Ontology Consortium, 2004) functional annotations were different for each set of significantly abundant transcripts from the overall pool of GO terms in the transcriptome, Fisher’s exact tests were performed in R for each of the GO categories (Biological Processes, Molecular Functions, and Cellular Components) using a modified version of the protocol described by Tribble et al. (2021). Additionally, individual GO terms that were enriched in each significantly differentially abundant transcript set relative to the whole transcriptome were detected using a maximum likelihood framework based on a binomial distribution as in Tribble et al. (2021). The script was modified to adjust P-values with the Benjamini-Hochberg FDR correction. Annotations were derived from the S. caninervis genome (Silva et al., 2020) and when annotations for multiple isoforms existed in the reference, the annotations (including GO terms) of the longest isoform were used.

Samples from the two study species were also analyzed together in a likelihood ratio test (LRT) with the reduced model of ∼species + time and a P-adj cut-off of 0.05 to test for transcripts that have different transcript abundance changes in the two species over the UVR treatment time course, controlling for differences at T₀. Next, VST transformed transcript counts from genes whose transcript abundance pattern differed significantly between the two species, identified via the LRT, were used in a cluster analysis to identify groups of genes that have similar patterns using the R package DEGreport (Pantano, 2020), with a minimum of five genes per cluster enforced. Transcript abundance Z-scores were calculated to scale and average replicates for plotting visualization.

All sets of significant transcripts were checked for the presence of two candidate gene families that were identified a priori: transcripts encoding early light-inducible proteins (ELIPs) and transcripts of late embryogenesis abundant genes (LEAs), using orthogroup lists in Silva et al. (2020). Both ELIPs and LEAs are involved in vegetative desiccation tolerance in S. caninervis as well as other resurrection plants (Costa et al., 2017; Van Buren et al., 2019; Silva et al., 2020) and were selected for the potential for cross-talk and cross-tolerance between desiccation tolerance and UVT. Furthermore, ELIPs function in photoprotection and in some plants the accumulation of ELIP transcripts is mediated by UVR8, the specialized UV-B receptor 8 (Hutin et al., 2003). Additionally, ELIPs, LEAs, and other genes with specific GO terms including: photosynthesis, response to light stimulus, response to stress, and photomorphogenesis were tested for differential transcript abundance as a group, relative to the distribution of differential abundance of all transcripts, following a modified version of the procedure described in Tribble et al. (2021). In brief, for each group of n genes (e.g., ELIPs), n transcripts were randomly sampled 10,000 from the pool of all assayed transcripts. The absolute LFC values of these sets of 10,000 samples were then compared to the distribution of absolute-value LFCs of the full assayed transcriptome and effect sizes were estimated with the Mann-Whitney U test (Mann and Whitney, 1947). Furthermore, number of significantly differentially abundant transcripts (DATs) after 10 and 30 min of UVR exposure in each species and in each gene group were tested against the number of significant transcripts in the 10,000 random samples to test if there were more or fewer than expected by chance.

The following libraries and tools were also used for analysis and visualization: tidyverse (Wickham et al., 2019), dplyr (Wickham et al., 2020), VennDiagram (Chen and Boutros, 2011), ggplot2 (Wickham, 2016), and the scripts in https://github.com/cmt2/bomTubers. Analysis code for this project is available on GitHub².

After quality filtering, a total of 905 M reads were recovered for all 18 samples; an average of about 50 million reads each. Syntrichia ruralis samples had an average read mapping rate of 51.2% while S. caninervis samples had an average read mapping rate 63.1% (per-sample mapping rates in Supplementary Table 1). In total, 6,851 genes were recovered in the transcriptome of S. ruralis and 6,868 genes in that of S. caninervis.

On the PCA of VST transcript counts of both species (Syntrichia ruralis and S. caninervis) at all three UVR treatment time points (T₀, T₁₀, and T₃₀), the two species separated strongly along the first PCA axis, which explained 93% of the variation (Figure 1). Syntrichia ruralis was tightly clustered while S. caninervis spread along PC2, which explained just 3% of the variation. There was no strong pattern of separation of UVR treatments on these two axes of the PCA, although UVR treatment did somewhat separate along PC2, but mostly due to variation among replicates within treatments.

In the PCA for S. ruralis alone, the three treatment groups separated slightly along axis PC1, which explained 43% of the variation (Figure 2A). However, T₀ spread widely across this axis and overlapped with T₁₀. There was some separation of T₁₀ and T₃₀ along PC2, which explained 23% of the variation, though again T₀ spread widely across this axis, especially in the same range as T₁₀. In the PCA of S. caninervis alone, PC1, which explained 45% of the variation, separated T₃₀ from the other two treatments (T₀ and T₁₀), though with some overlap with T₀. T₀ and T₁₀ were separated along PC2, which explained 20% of the variation (Figure 2B).

In S. ruralis, 18 transcripts were significantly differentially abundant between T₀ (no UVR exposure) and T₁₀ (10 min of UVR exposure), none of which were ELIPs or LEAs (Supplementary Table 2). Between T₀ and T₃₀, there were 38 transcripts significantly differentially abundant, which included two LEAs but no ELIPs (Supplementary Table 3). Seven individual transcripts were significantly differentially abundant in both the T₀ versus T₁₀ and the T₀ versus T₃₀ sets (Figure 3). Of the three GO categories (Biological Processes, Molecular Functions, and Cellular Components) in the significant transcript sets for T₁₀ and T₃₀ (30 min of UVR exposure) for S. ruralis, only those for Molecular Functions in transcripts significantly different at T₁₀ were significantly different from the overall distribution of GO terms (P = 0.0003).

In S. caninervis, of the 6,868 genes detected in the transcriptome, 10 were differentially abundant between T₀ and T₁₀, one of which was an ELIP and none of which were LEAs (Supplementary Table 4). When comparing T₀ and T₃₀, there were 126 genes that were significantly differentially abundant, which contained two ELIPs and four LEAs (Supplementary Table 5). Three transcripts were significantly differentially abundant in both the T₀ versus T₁₀ and the T₀ versus T₃₀ sets (Figure 4). Distribution of GO terms was not significantly different in any of the GO categories for significant transcript sets for T₁₀ or T₃₀. One GO term, GO:0016021 integral component of membrane, was significantly more abundant in the significant transcript set than in the overall transcriptome (P = 0.043).

While 18 transcripts were significantly differentially abundant at T₀ in S. ruralis and 10 were in S. caninervis, none of these were shared between the two (Figure 5A). In contrast, there were many more DATs in S. caninervis (126) than in S. ruralis (38) at T₃₀, and 13 transcripts were significant in both (Supplementary Table 6 and Figure 5B).

When analyzed together for a likelihood ratio test to identify transcripts that have a different abundance pattern at T₁₀ and T₃₀ while controlling for differences at T₀, a total of 6,859 transcripts were assayed. Of these, 69 were significantly different in the two species (Supplementary Table 7 and Figure 6). Cluster analyses of these transcripts identified six transcript clusters, each containing 6–19 transcripts (Supplementary Table 8 and Figure 7).

In S. ruralis, none of the hypothesized candidate groups had a significantly higher distribution of absolute-value LFCs than the entire transcriptome at T₃₀ (Supplementary Figure 1 and Figure 8A). One candidate group, LEAs, had more DATs at T₃₀ than expected by chance (Supplementary Figure 2 and Figure 8A). At T₁₀, ELIPs, LEAs, and transcripts with the GO term photosynthesis had a significantly higher distribution of absolute-value LFCs than the entire transcriptome (Supplementary Figures 3, 4) and transcripts with the GO term photosynthesis had more DATs than expected by chance (Supplementary Figure 5).

In S. caninervis, both LEAs and ELIPs had a significantly higher distribution of absolute-value LFCs than the entire transcriptome at T₃₀ (Supplementary Figure 6 and Figure 8B) and had more DATs than expected by chance (Supplementary Figure 7 and Figure 8B). At T₁₀, ELIPs had a significantly higher distribution of absolute-value LFCs than the entire transcriptome (Supplementary Figures 8, 9) and more DATs than expected by chance (Supplementary Figure 10).

At T₁₀, S. ruralis had 18 DATs compared to T₀. Putative functions of those genes included oxidation-reduction via peroxidases, photosynthetic monooxygenases, oxygen- and various metal-binding, and membrane and cell wall modifications (Supplementary Table 2). The differentially abundant transcript with the largest absolute-value LFC was of the gene Sc_g01013 – ribulose bisphosphate carboxylase small chain clone 512-like – a component of ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCo), which is essential for photosynthesis and photorespiration (Table 1). This transcript had a log2-fold increase of more than 4.5 with 10 min of UVR exposure. Similarly, Sc_g13199 – probable sodium/metabolite cotransporter BASS3, chloroplastic, an integral component of the chloroplast envelope – was increased. A relationship between UVR and photosynthesis is well-documented; most commonly UV-B radiation-induced impairment of photosynthesis has been observed (Vass, 1997; Sicora et al., 2006). While several components of photosynthetic machinery are susceptible to UVR-induced damage, one clear cause of photosynthetic impairment when exposed to intense or prolonged UVR is degradation of RuBisCo (Jordan et al., 1992; Nogués and Baker, 1995; Bischof et al., 2000). Along with other transcripts involved in oxidation-reduction processes, increased abundance of these transcripts may suggest ROS accumulation, oxidative stress, and photosynthetic machinery degradation from acute UVR exposure in S. ruralis. Interestingly, Sc_g01013 – ribulose bisphosphate carboxylase small chain clone 512-like – dropped drastically again at T₃₀ (Figure 3), suggesting this is a uniquely early-exposure response.

At T₃₀ the 38 DATs in S. ruralis (Supplementary Table 3) also included increases of several genes involved in oxidation-reduction processes and chloroplastic proteins, suggesting a continued response to UVR-induced ROS in the photosynthetic apparatus. The top-most differentially abundant transcript was Sc_g10872 – heavy metal-associated isoprenylated plant protein (HIPP) 28-like – a metallochaperone responsible for safe transport of metallic ions within the cell (Table 2). In Oryza sativa, HIPPs are also involved in cold and drought stress (de Abreu-Neto et al., 2013). Although it did not meet the significance threshold at T₁₀, abundance of this transcript increased consistently over the 30-min time series (Figure 3), suggesting a relationship between UVR exposure time and transcript abundance. Similarly, transcripts of Sc_g07931 – senescence/dehydration-associated protein At4g35985, chloroplastic-like, which is involved in cold and salt stress – increased at T₃₀. It is possible that in S. ruralis these genes are involved in UVR protection or that they are involved in something else entirely but that the response is UVR-induced in a case of genetic cross-talk (Sinclair et al., 2013).

In S. caninervis, the 10 transcripts significantly changed in abundance at T₁₀ contained several genes involved in the mitochondrial and other membranes (Supplementary Table 4), which are targets of ultraviolet radiation in plants (Murphy, 1983). Along with one significant transcript involved in oxidation-reduction, the high number of membrane-involved transcripts may be a sign of stress or simply an acclimatory response to protect membranes in the new environmental conditions. Indeed, the distinction between stress and regulatory acclimation is not easily defined (Hideg et al., 2013; Robson et al., 2019). The top-most differentially abundant transcript, which decreased with an LFC of −3.2, was Sc_g04894 – uncharacterized protein LOC112281062 – with only one GO term, response to water (Table 3). As this protein is uncharacterized, the role it is playing in UVR response can only be speculated upon, but given the high level of desiccation tolerance in this species, the functional annotation is intriguing. In fact, UVR has been implicated in desiccation tolerance in S. caninervis, where removal of natural levels of UVR from a field setting hindered recovery of photosynthetic efficiency from desiccation (Ekwealor et al., 2021). By T₃₀, abundance of this transcript had increased again to nearly an LFC of −1 (Figure 4).

Interestingly, many of the 126 transcripts differentially abundant for S. caninervis at T₃₀ were also for uncharacterized proteins (Supplementary Table 5 and Table 4). Many of these increased consistently with increased UVR exposure, though some absolute-value LFCs remained low at T₁₀ and increased in abundance only at T₃₀ (Figure 4). Furthermore, several of these transcripts were involved in oxidation-reduction and in membranes. Indeed, the GO term integral component of membrane was enriched in the DATs, relative to the transcriptome. UV-B radiation-induced ROS can cause oxidative damage to lipids, leading to lipid peroxidation, membrane permeability, and disruption of membrane integrity (Foyer et al., 1994; Dai et al., 1997; Alexieva et al., 2001). While the functions of these uncharacterized genes are not known, they may be involved in either protection or repair of cell or thylakoid membranes.

Syntrichia caninervis had one and two differentially abundant ELIP transcripts at T₀ and T₃₀, respectively, while S. ruralis had none at either time point. Additionally, the entire set of ELIPs was significantly increased and had a higher number of significantly differentially abundant ELIPs than expected by chance at T₃₀, suggesting a strong ELIP response with 30 min of UVR exposure in S. caninervis. Increased abundance with UVR may be due to perception of UV-B radiation in particular (Singh et al., 2014). Plants perceive UV-B radiation via the specialized UV-B receptor 8 (UVR8; Ulm and Nagy, 2005), which is remarkably conserved in structure and mechanism across land plants and green algae (Ulm and Nagy, 2005; Tilbrook et al., 2013; Jenkins, 2014). However, the regulatory role that UVR and the UVR8 receptor play may not be conserved (Robson et al., 2015). For example, a cytochrome P450 monooxygenase 98 (CYP98) enzyme is involved in biosynthesis of UV-absorbing pigments in some plants (Ehlting et al., 2006; Nelson and Werck-Reichhart, 2011), including phenolic components of the cuticle of P. patens, but acts in the first irreversible step in the biosynthesis of lignin in tracheophytes (Renault et al., 2017). In some plants UVR8 mediates the accumulation of transcripts encoding early light-inducible proteins (ELIPs; Singh et al., 2014), which function in photoprotection (Hutin et al., 2003) and desiccation tolerance in resurrection plants (Zeng, 2002; Oliver et al., 2004; Van Buren et al., 2019). While it is not yet known if or how ELIPs might be involved in UVT, their role in other forms of photoprotection and in desiccation tolerance suggest that this may also be an example of cross-talk in underlying pathways (Zeng, 2002; Hutin et al., 2003; Oliver et al., 2004; Van Buren et al., 2019).

In both S. ruralis and S. caninervis, late embryogenesis abundant (LEA) genes were more significantly differentially abundant at T₃₀ than expected by chance, while no LEAs were significantly differentially abundant at T₀ in either species. LEAs are involved in desiccation tolerance in S. caninervis as well as other resurrection plants (Costa et al., 2017; Van Buren et al., 2019; Silva et al., 2020). Moreover, some LEAs specifically function in protection against oxidative stress from ROS, even outside of drought conditions (Mowla et al., 2006). Both species had one significant group 3 LEA and one D-29 type (Sc_g05631 – late embryogenesis abundant protein, group 3-like, and Sc_g13874 – late embryogenesis abundant protein D-29), while S. caninervis additionally had a second group 3 type (Sc_g00779 – late embryogenesis abundant protein, group 3-like) and an additional group 14 LEA (Sc_g10451 – putative late embryogenesis abundant protein, LEA-14). Furthermore, the entire set of LEAs in S. caninervis were more differentially abundant than expected by chance at T₃₀. Given the role of LEAs in desiccation tolerance, these differences may reflect adaptations to different habitats where prolonged UVR is experienced in different hydration states.

Of the top 10 most species-specific transcript abundance responses with UVR exposure, two in particular had striking differences in response to UVR over the time series (Figure 6 and Table 4). The first of which, Sc_g08728 – Core-2/I-branching β-16-N-acetylglucosaminyltransferase family protein – is involved in biosynthesis of asparagine-attached glycans (N-glycans), which function in stress tolerance in A. thaliana (von Schaewen et al., 2008). This gene appears to be constitutively expressed in S. caninervis and remains in high abundance with or without UVR treatment, while abundance in S. ruralis is much lower with no UVR and drops drastically with UVR exposure. Sc_g14586 – pectinesterase 2-like – also shows an interesting pattern with UVR treatment in the two species. Pectinesterases are cell-wall-associated enzymes that have been implicated in many components of plant development (Phan et al., 2007). Transcript abundance is similar in the two species prior to UVR exposure, though higher in S. ruralis, but they respond to UVR treatment in opposite ways. In S. ruralis, transcript abundance increases at 10 min and decreases again slightly at 30 min. In S. caninervis, abundance decreases drastically then increases again. It is unclear what this gene could be doing in response to UVR, if anything, but this pattern nonetheless again demonstrates that these two species have different responses.

Six clusters of transcripts that have significantly different responses to 10 and 30 min of UVR exposure in the two species after controlling for differences at T₀ were identified (Supplementary Table 8). Cluster I, containing 13 transcripts, is characterized by an increase in abundance at T₁₀ and a plateau at T₃₀ in S. ruralis, but a decrease with increasing UVR time in S. caninervis (Figure 7). Many of these thirteen transcripts are for genes involved in the nucleus and in nucleic acid or protein binding. In fact, only nine of the thirteen genes have GO annotations at all, and six of them have at least one of the following GO Molecular Functions: nucleic acid binding, protein binding, DNA-binding transcription factor activity, and DNA binding; or a Cellular Components annotation of nucleus. Cluster V has a similar pattern in the two species and also has several GO annotations involving DNA binding and nuclear location. DNA absorbs wavelengths in the ultraviolet spectrum which can cause DNA breaks and lesions, and plants may respond to UVR-induced damage with DNA binding for repair (Pang and Hays, 1991; Tuteja et al., 2009; Gill et al., 2015). These patterns suggest the possibility that S. ruralis and S. caninervis may respond differently in terms of DNA protection when exposed to UVR. Cluster VI also has interesting patterns in the two species: high abundance with a small decrease at T₁₀ in S. ruralis, lower abundance with a large increase at T₃₀ in S. caninervis. Of the eight transcripts in this cluster, seven of them have GO annotations and all of them have a Cellular Components annotation of integral component of membrane, membrane, plasma membrane, or some combination of these. This may suggest some constitutive level of membrane-related transcript abundance in S. ruralis while S. caninervis responds with membrane-related transcripts only with increasing UVR exposure.

Mosses exposed to UVR while hydrated may utilize a more active response to UVR, while arid-adapted species may have passive protection. In this study we found little overlap in the transcriptomic response to broadband UVR between xeric S. caninervis and the more mesic S. ruralis. One explanation for the wide distribution of S. ruralis could be that it has a high degree of physiological plasticity and can acclimate to a variety of environmental conditions, but the transcriptomic response to acute UVR did not support this. In fact, S. caninervis had nearly twice as many DATs than S. ruralis, suggesting a more responsive and less constitutive response to UVR in the former. This result may be an indication that UVR radiation mediates other functions in S. caninervis, perhaps related to its habitat. In its natural desert habitat, this species experiences prolonged periods of desiccated quiescence seasonally, where it dissipates excess solar radiation as heat (Ekwealor et al., 2021). Since UV-B radiation varies substantially seasonally, relative to other regions of the spectrum (Robson et al., 2019), UV-B can act as a phenological cue for plants. For example, in Populus tremula interannual variation in UV-B radiation effects phenology and growth via UVR8-mediated ABA signaling with Flowering locus T (FT) genes (Strømme et al., 2015). The larger transcriptomic response in S. caninervis, along with the several differentially abundant uncharacterized protein transcripts, may suggest a phenological or other regulatory role of UVR in this species. Furthermore, the presence of three differentially abundant ELIPs in S. caninervis (but not S. ruralis) suggest a relationship between UVR and desiccation. It is possible that UVR is a phenological cue for seasonal desiccation in desert S. caninervis, though more research is needed to test this hypothesis. It will be interesting in the future to examine more genotypes in these species and their relatives from different habitat types, in a phylogenetic context, to explore how the expression patterns evolve.