Transcriptomic Evidence for Reproductive Suppression in Male Peromyscus eremicus (Cactus Mouse) Subjected to Acute Dehydration

17 Understanding how organisms adapt to extreme environments is an outstanding question facing 18 evolutionary biologists. Research related to a specific example of adaptation, mammals in desert 19 environments, has focused on survival, while questions related to the reproductive effects of dehydration 20 have been largely ignored. Here, we explore the reproductive consequences of acute dehydration by 21 utilizing RNAseq data in the desert-specialized rodent, Peromyscus eremicus. Nine genes were 22 consistently differentially expressed between hydrated and dehydrated mice, a low number which aligns 23 with current perceptions of this species’ extreme desert specialization. However, these differentially 24 expressed genes include Insulin-like 3 (Insl3), a regulator of male fertility, as well as Slc45a3 and Slc38a5, 25 both of which interact with genes with important roles in reproductive function. Together, these findings 26 suggest that acute dehydration is linked to reproductive mitigation, a result which is unexpected in an 27 animal capable of surviving and successfully reproducing without available external water sources. 28


Introduction
analyses; however, we extend this genetic approach by shifting the focus from adaptions for survival to 93 include adaptations for reproductive success. 94 In nature, wild cactus mice are subjected to both acute and chronic dehydration, and understanding 95 the reproductive effects of dehydration stress is an initial step for fully characterizing the suite of 96 phenotypes enabling successful reproduction. Given that this species has evolved in southwestern United 97 States deserts, we predicted that neither acute nor chronic water stress, while physiologically demanding, 98 would be associated with reproductive suppression. To test these predictions, we leveraged previous 99 research that characterized the transcriptome of male P. eremicus reproductive tissues from functional and 100 comparative perspectives . We extend upon this work by performing 101 an RNAseq experiment to identify differentially expressed genes in testes between male P. eremicus 102 subjected to acute dehydration versus control (fully hydrated) animals in order to determine the impacts, 103 if any, on male fertility. We pursue this line of research on the effects of dehydration on reproduction in 104 cactus mouse in order to begin to address the need for studies focusing on adaptation related to 105 reproductive success in animals living in extreme, and changing, environments. Hampshire in conditions that mimic temperature and humidity levels in southwestern US deserts, as 113 described previously . Males and females are housed together, which 114 provides olfactory cues to support reproductive maturation. Males do not undergo seasonal testicular 115 atrophy, as indicated by successful reproduction throughout the year. The individuals used in this study 116 were deemed reproductively mature once they became scrotal. 117 Males that were provided with water ad libidum had free access to water prior to euthanasia, and 118 these individuals are labeled as WET mice in our analyses. Mice that were water deprived, which we refer 119 to as DRY mice, were weighed and then water deprived for ~72 hours directly prior to euthanasia. All 120 mice were weighed prior to sacrifice, and DRY mice were evaluated for weight loss during dehydration. 121 Individuals in the study were collected between September 2014 -April 2016. 122 Cactus mice were sacrificed via isoflurane overdose and decapitation in accordance with 123 University of New Hampshire Animal Care and Use Committee guidelines (protocol number 130902) and 124 guidelines established by the American Society of Mammalogists (Sikes et al., 2016). Trunk blood 125 samples were collected following decapitation for serum electrolyte analyses with an Abaxis Vetscan VS2 126 using critical care cartridges (Abaxis). The complete methodology and results of the electrolyte study, as 127 well as the reported measures of water consumption and weight loss due to dehydration are described fully 128 elsewhere . Rather, this study focused on differential gene expression between 129 the testes of 11 WET and 11 DRY mice. Testes were harvested within ten minutes of euthanasia, placed  Assembly of Testes Transcriptome 142 We assembled a testes transcriptome from a single reproductively mature male using the de novo 143 transcriptome protocol described previously (MacManes, 2016). The testes transcripts were assembled 144 with alternative methodologies utilizing several optimization procedures to produce a high-quality 145 transcriptome; however, the permutations of this assembly process are described extensively elsewhere 146 (MacManes, 2016; . The testes transcriptome we selected was 147 constructed as described below. The raw reads were error corrected using Rcorrector version 1.0.1 (Song 148 & Florea, 2015), then subjected to quality trimming (using a threshold of PHRED <2, as per MacManes 149 2014) and adapter removal using Skewer version 0.1.127 (Jiang et al, 2014). These reads were then 150 assembled in the de novo transcriptome assembler BinPacker version 1.0 (Liu et al., 2016). We also 151 reduced sequence redundancy to improve the assembly using the sequence clustering software CD-HIT-152 EST version 4.6 (Li & Godzik, 2006;Fu et al., 2012). We further optimized the assembly with Transrate 153 version 1.0.1 (Smith-Unna et al., 2015) by retaining only highly supported contigs. We then evaluated the 154 assembly's structural integrity with Transrate and assessed completeness using the vertebrata database in 155 BUSCO version 1.1b1 (Simão et al., 2015). We quasimapped the raw reads to the assembly with Salmon 156 version 0.7.2 (Patro, Duggal & Kingsford, 2015) to confirm that mapping rates were high. Finally, the  demonstrated that differential gene expression (DGE) analyses produce more accurate results than 165 differential transcript expression (DTE) analyses. Furthermore, the differential gene expression approach 166 is more appropriate than differential transcript expression for the scope of our research question, which is 167 true of many evolutionary genomic studies (Soneson et al., 2016). However, because both DTE and DGE 168 approaches are widespread in current literature, we deemed it important to confirm that these 169 methodologies yielded concordant results in the current study. analyzing DGE, and edgeR performed optimally within our sample size range. While edgeR is a widely 173 used statistical package for evaluating differential expression, we also confirmed our results with another 174 popular package, DESeq2 (Love, Huber & Anders, 2014), in order to validate our findings. 175 We performed differential expression analyses with three alternative methodologies. Two analyses with DESeq2 in order to corroborate our DGE results from edgeR. Thus, the results for all three 190 differential expression analyses were evaluated to determine the coincidence among the genes identified 191 as significantly different between the WET and DRY groups. These alternative differential expression 192 methods are described in detail below. 193 We quasimapped each of the 11 WET and 11 DRY sample read sets to the testes transcriptome 194 with Salmon version 0.7.2 to generate transcript count data. To perform the gene-level analysis in edgeR, 195 we constructed a gene ID to transcript ID mapping file, which was generated by a BLASTn (Altschul et  Next, we performed a transcript-level analysis using edgeR. To accomplish this, the Salmon-205 generated count data was imported into R and analyzed as was described above for the gene-level analysis 206 in edgeR. After determining which transcript IDs were differentially expressed, we identified the 207 corresponding genes using the gene ID to transcript ID matrix described previously. The significantly 208 expressed transcripts without corresponding gene matches were selected for an additional BLASTn search 209 in the NCBI non-redundant nucleotide database (http://blast.ncbi.nlm.nih.gov/Blast.cgi). However, these 210 results were not subjected to any additional analyses, because these matches were not consistent across all 211 three differential expression analyses. This list of BLASTn search matches is provided in supplementary 212 materials (DTEno-matchBLASTnSequences.md).

213
The third analysis used DESeq2 to conduct an additional gene-level test, using the same methods 214 as described for the previous gene-level analysis, with the exception that data were imported into an 215 alternative software package. We determined the significantly differentially expressed genes (p < 0.05) 216 based on normalized counts and using the Benjamini-Hochburg correction (Benjamini & Hochburg, 1995) 217 for multiple comparisons. We only retained genes with a -1 < log 2 fold change > 1 in order to filter genes 218 at a conservative threshold for differential expression based on our sample size .

219
This filtering was not necessary for either of the edgeR analyses because log 2 fold changes exceeded this 220 threshold for the differentially expressed genes and transcripts (-1.3 < log 2 fold change > 1.4, in all cases). 221 We also compared the log 2 fold change values (of treatment differences by mapped count) for each  The testes transcriptome was assembled from a 45.8 million paired read data set. Additionally, 232 there were 9-20 million paired reads for each of the 22 testes data sets used for the differential expression 233 analysis (Supplemental 1), yielding 304,466,486 reads total for this analysis. The raw reads are available 234 at the European Nucleotide Archive under study accession number PRJEB18655. All data files, including 235 the testes un-annotated transcriptome, the dammit annotated transcriptome, and the data generated by the 236 differential gene expression analysis (described below) are available on DropBox 237 (https://www.dropbox.com/sh/ffr9xrmjxj9md1m/AACpxjQNn-Jlf25qNdslfRSCa?dl=0). These files will 238 be posted to Dryad upon manuscript acceptance. All code for these analyses is posted on GitHub 239 (https://github.com/macmanes-lab/testesDGE).

241
The performance of multiple transcriptome assemblies was evaluated thoroughly, and the selected 242 optimized testes assembly met high quality and completeness standards, and it also contains relatively few 243 contigs and has high read mapping rates (Table 1). Therefore, this transcriptome was used for our 244 differential expression analyses. The transcriptome was also annotated, and the complete statistics for this 245 dammit annotation are provided in Table 2. 246

247
Salmon quasimapping rates of all read datasets to the assembly were sufficiently high (range: 248 81.46% -87.02%; Supplemental 1), indicating the successful generation of transcript count data for our 249 differential expression analyses. The exact test performed for our gene-level analysis in edgeR indicated 250 that fifteen genes reached statistical significance (after adjusting for multiple comparisons) for DGE 251 between the WET and DRY treatment groups (Figure 1). Specifically, seven genes were more highly 252 expressed in WET individuals, and eight genes were more highly expressed in DRY individuals (Table   253 3). 254 We also performed an alternative transcript-level analysis using the referenced transcriptome 255 mapped reads exclusively with edgeR. The exact test found 66 differentially expressed transcripts ( Figure   256 2), 45 of which were more highly expressed in the WET group, and 21 were more highly expressed in the 257 DRY group ( Table 4). 10 of these differentially expressed transcripts were consistent with differentially 258 expressed genes from the edgeR DGE analysis. In addition, the significantly expressed transcripts without 259 an Ensembl ID match (nine WET and nine DRY) were retrieved for performing an nt all species BLASTn 260 search (http://blast.ncbi.nlm.nih.gov/Blast.cgi), and these results are in the supplementary materials.

261
The gene-level analysis conducted in DESeq2 yielded 215 significantly differentially expressed 262 genes (Figure 3), 67 of which were more highly expressed in the WET group, while 148 were highly 263 expressed in the DRY group. However, only 20 of these genes remained when we filtered them with a -1 264 < log 2 fold change > 1 to retain genes with a conservative threshold difference between treatment groups.

265
This list of 20 genes yielded 16 genes more highly expressed in WET mice and four genes highly expressed in DRY mice (Table 5). Nine of these genes overlapped with those found to be significant in 267 the previous two edgeR analyses.

268
To evaluate the correlation of log 2 fold change results for each gene (Ensembl ID) from the two 269 DGE analyses (EdgeR and DESeq2), we performed a regression of these log values, and they were 270 significantly correlated (Figure 4: Adj-R 2 = 0.6596; F(1,14214) = 2.754x10 4 ; p < 2.2x10 -16 ). This further 271 demonstrates the concordance of the DGE analyses in these two software packages.

272
To evaluate the degree to which the three analyses produced concordant results, we generated a 273 list of genes which were found to be significantly differently expressed by treatment across all three 274 analyses (Supplemental 2). There were six genes that were consistently highly-expressed in the WET 275 group and three genes that were highly-expressed in the DRY group. The six highly-expressed WET genes nine genes were corroborated by multiple methodologies, we are confident that they are differentially 281 expressed between our treatments. Estimates of expression for these genes generated using the gene-level 282 edgeR analysis are plotted in Figure 5. 283 The significantly differently expressed genes were evaluated for gene function and chromosomal 284 location ( Table 6). These genes occur throughout the genome; namely, they are located on different 285 chromosomes. The diverse functions of each gene will be described below. In addition, we generated 286 STRING diagrams (string-db.org) to view the protein-protein interactions for each of these nine genes     This is the first study to evaluate the reproductive correlates of acute dehydration in a desert-354 specialized rodent. We analyzed differential gene expression levels for testes in male Peromyscus 355 eremicus (cactus mouse) in acute dehydration (DRY) versus a control group that was fully hydrated 356 (WET). Our results provide evidence suggesting that reproductive function is attenuated in acutely 357 dehydrated mice, which is surprising, given that this is not consistent with our understanding of P. 358 eremicus as a desert specialist. While acute dehydration is less common than chronic dehydration for 359 desert mammals, it is a selective force they must overcome. Indeed, throughout much of the described 360 range of the cactus mouse, rainfall events may occur several times per year. Cactus mice, and many other 361 rodents, are known to rehydrate during these rainfall events (MacManes, personal observation). Following 362 rehydration, cactus mice experience acute dehydration, followed by a steady state of chronic dehydration.  poorly-characterized, our findings that genes integral to sperm development and activation interact with 405 genes differentially expressed in acute dehydration provide strong evidence that, contrary to our 406 expectations, acute dehydration is linked to reproductive suppression in the cactus mouse.

407
In contrast to genes that are down-regulated in dehydration, the genes that were upregulated in the 408 DRY group are known to be responsible for water homeostasis and cellular growth. The significance of 409 Rin2 is notable, because this protein is an effector for Rab5, which as a GTPase involved in vasopressin- proliferation and organ size control (Pan, 2010). 418 Emerging from this work is a hypothesis related to the reproductive response to water stress in the 419 cactus mouse, and perhaps other desert animals. Specifically, we hypothesize that while acute dehydration 420 is related to reproductive suppression, chronic dehydration is not. Indeed, it is virtually oxymoronic to 421 suggest that chronic dehydration, which is the baseline condition in desert animals, has negative 422 consequences for reproductive success. Generating an integrative, systems-level understanding of the 423 response to dehydration is required for testing this hypothesis. While understanding the renal response to   In addition, because global climate change is predicted to shift habitats toward extremes in temperature, 450 salinity, and aridity, and to alter species ranges, an enhanced understanding of the reproductive 451 consequences of these changes, and of the potential for organisms to rapidly adapt, may enable us to 452 effectively conserve innumerable species facing dramatic habitat changes.       335T* is the dataset which was used to assemble the testes transcriptome; therefore, these reads were 562 not used for the differential expression analysis.