bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
1 GENETIC BASES OF APOSEMATIC TRAITS: INSIGHTS FROM THE SKIN
2 TRANSCRIPTIONAL PROFILES OF Oophaga POISON FROGS
3
4
5 Andrés Posso-Terranova1-2*, José Andres1-3.
6
7
8 1. Department of Biology. University of Saskatchewan. 112 Science Pl. Saskatoon. SK.
9 Canada.
10 2. Universidad Nacional de Colombia sede Palmira. A.A. 237. Palmira. Colombia.
11 3. Department of Ecology and Evolution. Cornell University. Corson Hall. Ithaca. NY.
12 14853.
13
14 *Corresponding author
15 Email: [email protected]
16
17 All authors contributed equally to this work. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
18 ABSTRACT
19 Aposematic organisms advertise their defensive toxins to predators using a variety
20 of warning signals, including bright coloration. While most Neotropical poison frogs
21 (Dendrobatidae) rely on crypsis to avoid predators, Oophaga poison frogs from South
22 America advertise their chemical defenses, a complex mix of diet-derived alkaloids, by
23 using conspicuous hues. The present study aimed to characterize the skin transcriptomic
24 profiles of the South American clade of Oophaga poison frogs (O. anchicayensis, O.
25 solanensis, O. lehmanni and O. sylvatica). Our analyses showed very similar
26 transcriptomic profiles for these closely related species in terms of functional annotation
27 and relative abundance of gene ontology terms expressed. Analyses of expression profiles
28 of Oophaga and available skin transcriptomes of cryptic anurans allowed us to propose
29 possible mechanisms for the active sequestration of alkaloid-based chemical defenses and
30 to highlight some genes that may be potentially involved in resistance mechanisms to
31 avoid self-intoxication and skin coloration. In doing so, we provide an important
32 molecular resource for the study of warning signals that will facilitate the assembly and
33 annotation of future poison frog genomes.
34
35 KEYWORDS
36 Dendrobatids, RNA sequencing, transcriptomes, gene ontology, candidate
37 genes. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
38 INTRODUCTION
39 Aposematic organisms advertise their defensive toxins to predators using a variety
40 of warning signals [1-3]. Among aposematic species, the poison frogs (Dendrobatidae)
41 from the tropical rain forests of Central and South America represent one of the most
42 spectacular examples of warning coloration. While the majority of dendrobatids rely on
43 crypsis to avoid predators, some members of this family are both brightly colored and
44 chemically defended [4]. From a genetic point of view, these aposematic defenses can be
45 defined as a complex phenotype resulting from the integration (i.e. covariation) of
46 different genetic elements related to conspicuousness, bold behavior, unpalatability, diet
47 specialization, etc. While a long history of research has been devoted to understand the
48 genetics of warning coloration in arthropods, particularly in Heliconius butterflies [5, 6],
49 the molecular underpinnings of aposematism in vertebrates, particularly the mechanisms
50 whereby individuals become toxic (or distasteful) remain mostly unknown. In this study,
51 our primary aim was to characterize the skin transcriptomic profiles of Oophaga species
52 as a first attempt to shed light on the molecular genetic bases of aposematic components
53 in poison frogs: the ability to sequester alkaloid-based chemical defenses, warning
54 coloration, and the resistance mechanisms to avoid self-intoxication.
55 The frogs of this complex inhabit the lowland Pacific rainforests of the
56 Colombian and Ecuadorian Chocó. Preliminary molecular data from nuclear and
57 mitochondrial markers showed a similar genetic background among lineages in contrast
58 to an extraordinary diversity of morphotypes [7]. Individuals from different allopatric
59 lineages can be relatively homogeneous, striped, or spotted, and their colors range from
60 bright red, to orange and yellow (Fig 1) [8]. These polymorphic coloration patterns serve bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
61 as a warning signal of their chemical defenses [9], a complex mix of diet-derived
62 alkaloids secreted by the dermal glands [10, 11]. Although these chemicals may involve
63 metabolism and transport through other tissues, the final accumulation of toxins as well
64 as the production of pigmentary cells is performed in the skin tissue of these organisms
65 even during the adult stage [12-14]. Thus, we expected that the genes, pathways, and/ or
66 gene networks potentially associated with coloration, alkaloid metabolism, transport and
67 storage, to be expressed in the skin of these lineages.
68 Although the lineages of the Oophaga studied here display a wide variety of
69 warning signals, most of them share a black background coloration (Fig 1). In adult frogs,
70 this melanistic coloration is controlled by the lack of iridophores and xanthophores in the
71 dermal chromatophore unit, as well as the size and dispersion pattern of the melanin-
72 containing organelles (melanosomes) within the dermal melanophores [15]. Thus, we
73 also hypothesized that in adult individuals, genes involved in the amount, size and
74 distribution of the melanosomes should be differentially expressed in the skin. Similarly,
75 because the secreted alkaloids of chemically-defended dendrobatids are known to disrupt
76 the normal ion-channel activity and to alter the neurotransmitter-receptor binding
77 capacity of nerve and muscle cells [16-18], we hypothesized that the fraction of genes
78 differentially expressed in the skin of poison frogs (when compared to non-aposematic
79 species) is also enriched in genes related to alkaloid sequestration, transport and
80 autotoxicity avoidance. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
81 Fig 1. Hierarchical clustering (bootstrap=999) and heatmap contrasting the 82 differentially expressed unigenes from skin tissue between Oophaga and cryptic species. 83 Color within the heatmap represent the log2 fold change values. Dark blue indicates 84 upregulation while a lighter coloration indicates downregulation. Rows of the heatmap 85 correspond to the differentially expressed unigenes (n=1,931) detected in this study. 86 bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
87 To test these hypotheses, we first assembled the skin transcriptomes of five
88 different Harlequin poison frogs species: O. histrionica [19], O. lehmanni [20], O.
89 sylvatica, [21] and the recently described O. anchicayensis and O. solanensis [22]. Then,
90 we compared the transcription profiles of these aposematic species with that of a diverse
91 group of cryptic anurans. Functional annotation of unigenes showing a phylogenetic
92 signal (i.e. highly represented and differentially expressed in Oophaga poison frogs)
93 allowed us to propose plausible metabolic pathways (and candidate genes) related with
94 toxicity and color. Overall, this study provides an important molecular resource for the
95 study of aposematism within poison frogs and will facilitate the future assembling and
96 annotation of the complex dendrobatid genomes [23].
97
98 METHODS
99 Library preparation and sequencing
100 Due the conservation status of these frogs (CR, EN, VU; IUCN) and the
101 restrictions imposed by the Ethics Committee, we were allowed to euthanize a very
102 limited number of samples per species. Seven individuals (O. lehmanni and O.
103 anchicayensis, n=2 per each species; O. histrionica, O. sylvatica, and O. solanensis, n=1;
104 each species) were euthanized in the field with benzocaine gel at 5% by direct application
105 into the mouth cavity. Individual skin samples were taken and stored at -80 °C in
106 RNAlater® (Life Technologies, Carlbad, CA). RNA extractions were performed using
107 Tri-Reagent (Molecular Research Center, Cincinnati, OH). After quality check (Agilent
108 Bioanalyzer Agilent Technologies, Wilmington, DE) samples were used to generate
109 RNA-seq libraries using the Illumina Truseq RNA Sample Prep protocol (Illumina, San bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
110 Diego, CA). Libraries were cleaned using AMPure XP and sequenced on a single
111 Illumina HiSeq2000 lane (TruSeq SBS v. 3) as follows: four libraries (O. anchicayensis,
112 O. histrionica, O. lehmanni and O. sylvatica) were sequenced using single-end (150-bp)
113 while the remaining three libraries (O. anchicayensis, O. solanensis and O. lehmanni)
114 were run in a single lane of a paired-end module (100-bp, x 2). All animal procedures
115 were approved by the Ethics Committee of Universidad Nacional de Colombia (Acta
116 No.03, July 22nd, 2015) and were conducted based on the NIH Guide for the Principles of
117 Animal Care. Sampling was conducted according to the research permits granted by
118 Autoridad Nacional de Licencias Ambientales ANLA, Resolución 0255 del 12 de Marzo
119 de 2014 and Resolución 1482 del 20 de Noviembre de 2015. Sequencing was carried out
120 by the Cornell University’s BioResource Center. Datasets including raw RNA sequence
121 reads are available from the authors upon request.
122 Transcriptome assemblies and functional annotation
123 Initial read quality trimming, filtering and removal of adapters was performed
124 using FLEXBAR v.2.5 [24]. All retained reads were ≥ 50bp with an average quality of ≥
125 30 and less than two uncalled (ambiguous) bases (S1 Table). For each library, a de novo
126 transcriptome was assembled using TRINITY v.2.1.0 [25] with default settings (kmer=25,
127 minimum contig length=48) and keeping only the longest transcript per cluster for
128 subsequent analyses. Varying these parameters (i.e. default settings) did not result in
129 assemblies with longer N50 values. To account for differences in the quality and
130 sequencing strategy among individual transcriptomes, we constructed a composite de
131 novo reference transcriptome. Briefly, we followed [26] and combined 20% randomly
132 selected reads (>33 million in total) from each of the four single-end libraries (O. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
133 anchicayensis, O. histrionica, O. lehmanni and O. sylvatica). Then, a TRINITY assembly
134 was performed using a minimum contig length of 350 bp. To filter out highly similar
135 contigs that may potentially represent alternatively spliced transcripts, we implemented
136 the error correction module of iAssembler v1.3.2 [27] with default parameters (maximum
137 length of end clips=30 bp, minimum overlap length=40 bp, minimum percent sequence
138 identity=95%).
139 Gene annotation was conducted using a sequential BLASTX search to both, the
140 available Xenopus transcriptomes and the NCBI non-redundant (nr) database. The
141 composite transcripts were first compared with that of the Xenopus databases retaining
142 annotations with E-values ≤ 10-5. Unannotated contigs were then submitted for BLASTX
143 to the nr protein database for possible identification. Gene ontology (GO) annotation and
144 term mapping was performed using Blast2Go v 5.2.0 with default significance cutoffs
145 [28].
146 To estimate the completeness of each transcriptome, we implemented the
147 expected gene content of Benchmarking Universal Single-Copy Orthologs (BUSCO v3)
148 [29] as implemented in gVolante v1.1.0 [30]. This widely used metrics for transcriptome
149 assembly [31] assesses the quality of any given transcriptome by estimating the
150 completeness of core vertebrate genes predicted to be ubiquitous in eukaryotes. The pre-
151 processing of the reads and all BLASTX analyses were run in the Bugaboo Dell Xeon
152 cluster of the western Canada's WestGrid computer facilities (www.westgrid.ca).
153 Transcriptome profiles
154 Analyses on the distribution of reads and the relative abundance of different
155 transcripts within each individual sample (i.e. lineage) were carried out by mapping bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
156 RNA-seq reads back to the composite de novo assembly. In order to identify highly
157 represented unigenes, we first implemented eXpress (probabilistic assignment of
158 ambiguously mapping sequenced fragments) [32] to estimate the effective number of
159 reads (ER) that mapped to the contigs in the reference transcriptome after adjusting for
160 read number and length biases. Then, to compare the proportion of reads that mapped to a
161 transcript in each Oophaga RNA-library, we estimated the number of transcripts per
162 million (TPM), a measure of RNA abundance that allows the comparison between
163 samples (sum of all TPMs in each sample are the same) [33]. We visually explored for
164 highly represented unigenes in each RNA-library by using dispersion plots of the ER
165 values and percentile plots of the TPM distribution. Additionally, we performed a
166 Principal Component Analysis (PCA) of the reference contigs dataset and using each
167 library TPMs values as independent variables and selected the overall top expressed
168 unigenes (2%).
169 To further characterize the skin transcriptome profile of Oophaga poison frogs,
170 we compared the transcription levels of annotated unigenes in this clade with those
171 observed in several cryptic (i.e. non-aposematic) anurans, including a toad (Bufo
172 gargarizans), and five frog species (Pelophylax nigromaculatus, Polypedates
173 megacephalus, Rana catesbeiana, Rana sylvatica, and Xenophrys sangzhiensis, S2 Table)
174 [34, 35]. RNA-seq data from these species was obtained from the Sequence Read Archive
175 platform (SRA) and represented the only skin-based RNA-seq available to us for
176 comparison purposes. Although transcriptomic datasets are available for
177 phylogenetically closer and cryptic species (i.e Colostethus) [36], these datasets were not
178 included in our analysis because they are not derived from skin tissue and may bias our bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
179 analyses. Read quality, contamination screening, duplicate removal, and quality trimming
180 steps were performed in FLEXBAR v.2.5 as previously described. The filtered raw reads
181 from individual libraries (Oophaga, n=7; cryptic species n=6) were mapped to our
182 composite de novo reference transcriptome using Bowtie2 v2.3.4.1 [37]. To accommodate
183 for sequence divergence among the taxa included in our study (≈75–98 similarity), we
184 followed [38] and ran the alignment algorithm allowing for soft clipping (--local) and
185 with a maximum penalty value of three (--mp 3) . Then, we implemented eXpress [32] to
186 estimate the effective number of reads (ER) that mapped to the contigs in the reference
187 transcriptome after adjusting for read number and length biases. Only unigenes with an
188 average of >10 ER/library were included in the differential expression analyses [39].
189 To identify genes showing a phylogenetic signal in their expression levels and to
190 estimate fold changes in expression levels between aposematic (i.e. Oophaga) and cryptic
191 species, we followed [40] and implemented the R package DEseq [41] to select unigenes
192 with an adjusted p-value of <0.05. To visualize patterns of expression, we constructed a
193 multidimensional scaling plot (Euclidean distances) of the ER data using Past v. 3.18
194 [42]. A heatmap of the expression differences between Oophaga and cryptic anuran
195 lineages (ngenes=1,931) was generated using the ‘heatmap.2’ function of the ‘gplots’
196 package, with Euclidean distances and complete linkage for clustering (bootstrap=999)
197 [43]. BLAST analysis and GO term mapping for the 1,931 unigenes included in this
198 analysis was performed as described above.
199
200 RESULTS
201 Transcriptome assemblies and functional annotation bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
202 Illumina sequencing produced an average number of reads per sample of 118.3
203 million for paired-end and 37.7 million for single-end libraries. After a stringent read
204 trimming involving the removal of low-quality sequences, duplicated reads and reads
205 containing adapter/primer sequences, we retained over 81% of the initial sequencing data.
206 Paired-end libraries produced a fairly consistent number of reads (S1 Table).
207 After the removal/merging of highly similar contigs (paired-end:7,5% - 9,6%;
208 single-end: 6.0% - 13.4%), the 150-bp based transcriptomes recovered a large number of
209 contigs ranging from 35,287 (O. anchicayensis) to 107,381 (O. sylvatica). Paired-reads
210 libraries (100-bp; 2X) generated transcriptomes with a relatively similar number of
211 contigs (40,398 for O. solanensis and 60,494 for O. lehmanni). N50 values and average
212 transcript length (AL) were lower for paired-end libraries (N50= 498-561bp; AL=
213 441.78±379 - 475.93±436bp) than for assemblies produced with single-end libraries
214 (N50=667-1579bp; AL=538.9±1001bp - 668.6±819 bp) (Table 1). Assembled and
215 annotated transcriptomes are available from the authors upon request. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
216 Table 1. Summary statistics and quality assessment estimators of the eight de novo assembled transcriptomes generated in this study 217 (Oophaga species n=7; composite reference n=1). 218 219 Paired-end libraries Single-end libraries
O. solanensis O. anchicayensis O. lehmanni O. sylvatica O. lehmanni O. anchicayensis O. histrionica Reference
22.23 32.06 35.7 93.9 77.3 23.16 85.97 31.6 File size (Mbytes) 43,690 63,520 66,937 123,916 117,082 37,552 111,823 31,498 # contigs # contigs after filtration 40,398 57,516 60,494 107,381 104,523 35,287 96,892 31,498 (removal/merging of highly similar contigs) 19,853,219 28,061,593 31,856,993 81,266,530 66,367,430 20,238,091 74,768,159 31,968,628 # bases 17,164,733 23,756,585 26,287,312 58,224,462 50,845,151 15,278,290 53,582,406 31,968,628 # bases after filtration 7.5% 9.5% 9.6% 13.3% 10.7% 6.0% 13.4% 0.0% Merged contigs Average transcript 454.41± 405 441.78± 379 475.93 ± 436 655.8 ± 799 566.8± 648 538.9 ± 1579 668.6 ± 819 1,014.91 ± 948.43 length (AL) 6,115 7,915 8,400 15,102 12,906 54,265 15,136 16,041 Maximum length 528 498 561 1079 809 667 1,101 1,316 N50 46.5% 47.7% 50.8% 42.1% 41.6% 53.4% 44.5% 62.6% Contigs with BLAST hits 220 bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
221 Comparative BLAST analysis indicated that our transcriptomes recovered a
222 significant proportion of the X. laevis and X. tropicalis transcriptomes (50,592 and 41,042
223 unigenes respectively). Paired-end transcriptomes showed a lower number of significant
224 BLAST hits (18,263 – 29,670) than single-end libraries (18,526 – 40,651) (S1 Fig). In
225 general, individual de-novo transcriptomes were of good quality and recovered 51-79%
226 of the core eukaryotic universal single-copy orthologs [29] (S3 Table). The final de novo
227 composite skin transcriptome reconstructed from all the RNA reads of the Oophaga
228 species yielded 31,498 contigs greater than 350 bp. The overall assembly incorporated
229 94% of all initial reads and the level of fragmentation was low with half of all base pairs
230 clustered into contigs of 1,316 bp in length or greater. The maximum contig length was
231 16,041 bp and the AL was 1014.91 ± 948.43 (S2A Fig). Nucleotide-based BLAST
232 analyses (BLASTX) revealed that ~63% of the contigs (n=19,732) show significant
233 similarities with either annotated gene products and/or known protein domains (E-value ≤
234 10−5) (S2B Fig) and only a small fraction of unigenes (3.4%) showed significant
235 homology to the same annotated transcript. The highest BLAST scores were found with X.
236 tropicalis (74%), X. laevis (2%) and the green sea turtle Chelonia mydas (0.5%). The
237 composite skin transcriptome was of higher quality when compared with individual
238 assemblies, having 88.54% of the ortholog count of BUSCO [29] (S3 Table).
239 Gene Ontology (GO) assignments were used to classify the functions of the
240 predicted unigenes based on contigs with significant BLASTX (E-value ≤ 10−5). Based on
241 GO level II, unigenes were assigned to 27 biological processes (BP), 22 cell components
242 (CC) and 15 molecular functions (MF) (S2C Fig). Some unigenes were associated with
243 multiple GO annotations because a single sequence may be annotated in any or all bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
244 categories, giving more GO annotations than sequences annotated [44]. Within BP, ~47%
245 of the annotations were assigned to basic cellular, metabolic processes and biological
246 regulation. The remaining unigenes were involved in a broad range of BP such as
247 response to stimulus (7%), response to stress (7%), localization (6%), developmental
248 process (5%), signal transduction (4%), biogenesis (4%), immune response (2%),
249 reproductive process (1.7%) and cellular adhesion (1%). Within the CC category, other
250 than the essential cell constituents (44%), the membrane components were highly
251 represented in the transcriptome (19%). Within MF, most of the unigenes were assigned
252 to binding and catalytic activities (72%) followed by transporter activity (9%), molecular
253 transducers (7%) and molecular function regulators (4%) (S2C Fig).
254 Transcriptome profiles
255 Dispersion plots of ER values in Oophaga RNA libraries showed a common
256 pattern of over-represented unigenes across RNA-libraries. That is, the same group of
257 unigenes showed similar patterns of over-representation no matter RNA library’s origin
258 (i.e. different Oophaga species) or composition (paired vs. single-read libraries) (S3 Fig).
259 The TPM represents a measure of RNA abundance and hence, it provides a general
260 overview of gene expression levels in a particular sample. Interestingly, TPM
261 distribution plots (percentiles) showed a distinct spike in expression levels towards
262 percentile 98% for all our RNA-libraries (S4A Fig). After the conservative filtration of
263 the 2% top expressed unigenes, we selected a subset of 1,437 unigenes with particular
264 higher levels of expression (higher TPM values and PCA outliers, S4B Fig) in Oophaga
265 skin tissue. Homology BLAST analysis of these unigenes revealed that 76% (n=1,092)
266 show significant similarities with annotated gene products and/or known protein domains bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
267 distributed mainly across amphibian (frogs) species (X. tropicalis=30%, X. laevis=14%
268 and Rana catesbeiana=4%).
269 After adjusting for library size and length bias, the maximum effective number of
270 reads (ER) in Oophaga species ranged from 2,461 (O. anchicayensis single-end) to 5,813
271 (O. sylvatica – single-end) while in the cryptic species varied from 82,509 (Rana
272 catesbeiana) to 452,018 (Bufo gargarizans) (higher coverage in cryptic species, S2
273 Table). This indicated that after the optimization of our alignment parameters, a
274 considerable amount of reads from each RNA-seq library actually mapped to our
275 composite de novo transcriptome. After the removal of contigs with low expression
276 profiles (<10 read counts), our RNA-seq dataset was composed of reads mapping to
277 20,771 genes. Multidimensional scaling plots (stress<0.05) revealed a clear separation by
278 groups (Oophaga vs. cryptic species, p<0.001 Bonferroni corrected) with higher variation
279 among the cryptic species, as expected given the broad taxonomic range of the members
280 of this group (Fig 2).
281 About 9% of the highly expressed genes (n= 1,931) also showed a phylogenetic
282 signal (i.e. significant fold changes in expression levels between Oophaga and all other
283 anurans). Most of these genes (n=1,656; 85.8%) show higher levels of expression in
284 Oophaga. The clustering analysis (Fig 1) revealed that the toad (B. gargarizans) has the
285 closest transcription profile to Oophaga species. One possible explanation is the
286 presence of alkaloid-based toxins in its skin glands [45].
287 bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
288 Fig 2. Multidimensional scaling (MDS) plot of 13 RNA-sequencing samples 289 from skin tissue of Oophaga (n=7, red dots) and cryptic species (n=6, blue dots). RNA- 290 seq data shown here represent the effective read counts (ER) for 31,498 skin-specific 291 contigs. Color ellipses represent a 95% confidence interval. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
292 Homology BLAST analyses revealed that 65% (n=1,252) show significant
293 similarities with annotated gene products and/or known protein domains (S4 Table; S5A
294 Fig) distributed mainly across amphibian (frogs) species (X. tropicalis=32%, X.
295 laevis=19% and Rana catesbeiana=4%) (S5B Fig). Based on GO mapping level II, these
296 unigenes were assigned to 24 BP, 20 CC and 13 MF (S5C Fig). Multilevel GO term
297 classification assigned the highly expressed unigenes to 16 BP, 5 CC and 9 MF (S5C
298 Fig). Within BP, 13% of unigenes were associated with oxidation-reduction processes,
299 followed by response to stimulus (10%) and regulation of macromolecule metabolic
300 processes (7.3%). Within CC, the integral components of the cell membrane were highly
301 represented (39%), followed by protein complex (27%) and nuclear components (20%).
302 Within MF, most of the differentially expressed unigenes were assigned to metal ion
303 binding activity (21%), followed by protein binding activity and transmembrane
304 transporters (17%), oxidoreductase activity (12%) and structural molecule activity (10%)
305 (S5C Fig). The full functional annotation of these genes allowed us to propose plausible
306 mechanisms that may be involved in the ability to sequester alkaloid-based chemical
307 defenses, warning coloration and the auto-resistance mechanisms to avoid self-
308 intoxication (S5 Table).
309
310 DISCUSSION
311 Studies focusing on highly represented and differential gene expression in related
312 lineages have provided valuable insights into the molecular underpinnings of phenotypic
313 variation [46, 47]. These analyses rely on the premises that lineages show contrasting
314 phenotypes of interest, and that most differentially expressed genes relate to the trait of bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
315 interest rather than the overall of genomic differentiation between lineages. Here, we
316 examined shared patterns of gene expression in a small clade of closely related
317 aposematic frogs (Oophaga species). Because all members of this monophyletic group
318 show contrasting hues and accumulate alkaloid-derived toxins in their dermal glands [48,
319 49], we assumed that the genes related with these traits are likely to 1) be highly
320 expressed in the skin of Oophaga poison frogs and 2) showed contrasting expression
321 patterns between aposematic and cryptic and less (or not) poisonous species. Using
322 expression levels in a multivariate framework, we first identified ortholog genes that are
323 highly represented in Oophaga skin tissue and then, we compared with those
324 differentially expressed in the skin of contrasting groups and that show similar patterns of
325 expression among Oophaga species. We finally used a functional annotation analyses to
326 highlight a series of genes that are possibly associated with a series of aposematic traits.
327 The characterization of our skin transcriptomic profiles allowed us to identify
328 genes that are potentially associated with alkaloid transportation and resistance to
329 autotoxicity in poison frogs (Dendrobatidae). Because the alkaloids secreted by these
330 frogs are structurally very similar to those present in plants [18, 50-52], it is reasonable to
331 assume that the transportation and accumulation of alkaloids in these frogs may be
332 carried out by similar systems to those described in plants in which alkaloids are
333 transferred from source to sink organs [53-55]. If so, at least, two alternative membrane
334 mechanisms may help to explain the transport and storage of unmodified diet-derived
335 alkaloid by the specialized cells of skin secretory glands. First, alkaloids in a lipophilic
336 state may freely pass through the cell membranes by simple diffusion and accumulate in
337 acidic secretory lysosomes if they become protonated to form hydrophilic cations. This bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
338 ion-trap mechanism is not energy-dependent and does not necessarily require the
339 expression of any transporters. Alternatively, the transportation of alkaloids may be
340 managed by proton-antiport carrier systems in an energy-requiring manner [56-58].
341 Under this alternative model, diet derived alkaloids may be taken up by ABC transporters
342 [59, 60] and may accumulate in the secretory lysosomes of the skin gland cells by a
343 cation exchanger antiporter (CAX) system, dependent on the pH gradient generated by
344 “vacuolar” type ATPases (V-H+-ATPases) and/or pyrophosphatases (V-H+-PPases). To
345 find some support for these two potential models that might explain the accumulation and
346 posterior secretion of dietary alkaloids, we investigated the common transcriptional
347 profile across Oophaga species. For each of these species, our differential expression and
348 transcript abundance analysis revealed that at least 15 homologs of different type II CAX
349 are highly expressed in skin tissue (S5 Table). A result suggesting the existence of an
350 active membrane transportation-accumulation mechanism of alkaloids in harlequin
351 poison frogs. Consistent with this hypothesis, we also found putative ABC transporters
352 and at least one V-H+-ATPase on the skin transcriptome. All together, these results
353 suggest that active lysosomal exocytosis may play a key role in the secretion of alkaloids
354 in these frogs.
355 The accumulation of toxic compounds implies that organisms must avoid self-
356 intoxication (auto-resistance). While the membrane transportation mechanisms described
357 above reduce auto-toxicity by compartmentalizing the sequestered alkaloids, other non-
358 alternative excluding mechanisms are likely to contribute to auto-toxicity resistance. In
359 harlequin poison frogs, the major toxic components present in skin tissue are
360 histrionicoxins (HTX) [51]. These alkaloids are known to cause temporary paralysis or bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
361 even death by inactivating or blocking voltage-gate ion channels [61]. Resistance to this
362 type of cytotoxic compounds usually arises through either, an increased expression level
363 of P450 enzymes (CYPs) that metabolize the toxin, or through target insensitivity via
364 mutations that reduce the toxin's ability to bind the ion channel itself [62]. Our analyses
365 revealed that five CYP homologs are amongst the most highly expressed genes in the
366 skin of harlequin poison frogs (S5 Table). This result suggests that the oxidative
367 biotransformation of lipophilic alkaloids to hydrophilic compounds [63] is an important
368 auto resistance mechanism in these frogs. Although the overproduction of the CYP-
369 enzymes has been interpreted as evidence for metabolic detoxification [64], there is an
370 alternative explanation for the high expression levels of CYP proteins in skin of harlequin
371 poison frogs. While many of the alkaloids sequestered in the wild are accumulated
372 unchanged in the dermal glands, at least two species of dendrobatids, stereo-selectively
373 hydroxylate the toxic pumiliotoxin PTX-(+)-251D to convert it into a much more potent
374 toxin, the allopumiliotoxin aPTX-(+)-267A [17]. Both PTX-(+)-251D and aPTX-(+)-
375 267A have been isolated from the skin of O. histrionica, and CYPs are known to be
376 involved in the stereo-selective hydroxylation of alkaloids [65]. Thus, it is also possible
377 that the CYPs expressed in the skin of harlequin poison frogs may enhance the
378 antipredator potency of ingested PTXs.
379 Despite the major role that detoxification enzymes may have in the resistance to
380 diet acquired toxins, mutations conferring constitutive resistance to alkaloids are also
381 likely to be involved in auto-resistance in chemically-defended dendrobatids. In insects,
382 single amino-acid substitution in the voltage-gated sodium channels (Na+K+-ATPases)
383 are responsible for resistance to host-plant phytochemicals [66-68]. In amphibians, the bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
384 only documented example of this this type of target insensitivity is from distantly-related
385 lineages of alkaloid-defended frogs, 24 species of Neotropical dendrobatids and the
386 aposematic Madagascar frog Mantella aurantiaca [69]. In all of these cases, they carried
387 different amino acid replacements in the inner pore of the voltage-gated sodium channel
388 Nav1.4 that are not found in other frog [70]. Our analyses revealed eight genes encoding
389 voltage-gated ion channel proteins (VGIC), one of which (8966), encodes the gamma-1
390 subunit of a sodium channel.
391 Aposematism in harlequin frogs is a complex phenotype that results from the
392 integration of different elements including unpalatability and conspicuousness. Thus,
393 another of our goals was to identify genes potentially related with warning coloration.
394 Many aposematic organisms, including the Oophaga poison frogs studied here exhibit
395 color patterns that show strong color and/or luminance contrast such between black and
396 red, orange or yellow [15]. It has been proposed that such patterning increases aversion
397 learning as well as conspicuousness and distinctiveness from palatable prey [71-76]. In
398 anurans, coloration relates to both the structure and pigment composition of the dermal
399 chromatophore units in which three types of pigment cells (xanthophores, iridophores and
400 melanophores) are laid one on another [77]. In dark/black skin areas of harlequin poison
401 frogs, no xanthophores or iridophores are present and the distal fingers of the
402 melanophores are filled with melanin granules (melanosomes) obscuring the dermis [15].
403 Thus, genes involved in the amount, size and distribution of the melanosomes are likely
404 to play a significant role in the coloration pattern of these frogs.
405 In one of the few detailed studies of coloration in anurans [78], authors suggest a
406 common origin for of all the pigment granules found in the cells of the chromatophore: a bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
407 primordial organelle (vesicle) derived from the rough endoplasmic reticulum (RER).
408 According to this model, in the formation of melanosomes, the pre-melanosomes are
409 derived from cisternae of the RER, which then fuse with vesicles derived from the Golgi
410 complex containing tyrosinase enzymes [12, 78-81]. A highly expressed gene identified
411 here (tyrosinase regulator, 25062) may contribute to regulate this mechanism, which in
412 turn might translate into hue differences. Interestingly, among Oophaga species, we have
413 found populations that are characterized by light-brown background coloration as oppose
414 to black [15, 22]. One tantalizing possibility is that this phenotypic difference is indeed
415 associated with the expression differences of these tyrosinase enzymes.
416 Obviously, there are other molecular and cellular mechanisms that might be
417 associated to the difference between light and dark background coloration. In fishes and
418 amphibians, dark hues are known to be produced by the interaction between high levels
419 of melanocyte-stimulating hormone (α-MSH) and several variants of its transmembrane
420 receptor (MC1R) through the dispersion of melanosomes within the melanophore (by
421 increasing cAMP intracellular levels) [82, 83]. Thus, it is possible that structural or
422 expression differences in the MC1R might contribute to dark phenotypes in Oophaga
423 frogs. While there is a strong evidence that different mutations at MC1R cause either
424 light or dark phenotypes in many mammals, birds and reptiles [84-88], the only two
425 studies conducted in frogs are inconclusive [89, 90]. A detailed inspection of the coding
426 sequences recovered for this gene revealed the presence of different length isoforms,
427 making MC1R a promising candidate gene candidate to explain the differences in
428 background coloration in poison frogs [15]. Finally, our study revealed another two
429 highly and differentially expressed genes (14447-Melanoregulin-; 2038-Melanocortin bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
430 phosphoprotein), which may also contribute to dark hues. In this case, the predicted
431 products of these genes are key proteins that mediate the melanosome transport and
432 distribution in epidermal cells through the formation of a tripartite protein complex [91].
433 The disruption of the transport protein complex results in pigmentary dilution and lighter
434 phenotypes by the clustering of melanosomes around the nucleus [92].
435 Overall, this study demonstrates the utility of using RNA-sequencing with non-
436 model organisms to identify loci that may be of adaptive importance. Altogether, these
437 data enabled us to provide a first global study of Oophaga poison frogs transcriptomes
438 and propose potential mechanisms for alkaloid sequestration, auto-resistance to toxic
439 compounds, and variation in coloration and patterns. It is important to recognized that
440 due the sampling restrictions imposed to us (see methods), we are most likely missing
441 some of the genes responsible for these traits. It is also probable that the genes associated
442 with coloration, alkaloid metabolism, transport and storage are differentially expressed
443 not only in skin, but other tissues in the organism. Despite the potential caveats, other
444 molecular studies have shown that some of the genes identified in this study are indeed
445 potentially related with aposematic traits in dendrobatids [15, 70, 93].
446 Despite the critical impact of genetic basis of coloration in the evolution and
447 diversification of aposematic phenotypes, the genetic architecture of coloration in poison
448 frogs (Dendrobatidae) remains virtually unexplored. Here, we report probably the first
449 RNA-seq study in Oophaga poison frogs, a model system for understanding the
450 relationship between toxicity, diet and coloration. The skin-expressed genes that we have
451 identified here provide initial working hypotheses to further unravel the molecular
452 genetics mechanisms to sequester alkaloid-based chemical defenses, warning coloration bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
453 and the auto-resistance mechanisms to avoid self-intoxication. Hence, further analysis
454 aiming to compare the amino acid substitutions in the expressed cDNAs will provide
455 insights about the structure and function of each of these genes. Finally, our comparative
456 transcriptome data provide an important new resource to better understand the evolution
457 of warning signals in nature.
458
459 Acknowledgements
460 We would like to thank the former members of Rick Harrison’s lab (requiescat in
461 pace) at Cornell University for their valuable discussions and comments on early versions
462 of the manuscript, which was greatly improved by the thoughtful comments of Dr.
463 Andrew Crawford at Universidad de los Andes (Colombia). This study was funded by
464 COLCIENCIAS (Departamento Administrativo de Ciencia, Tecnología e Innovación,
465 Colombia, 529-2011) grant to A.P.-T. and an NSERC-Discovery grant to J.A.A. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
466 REFERENCES
467 1. Brown J. The evolution of parental care, aposematism and color diversity in 468 Neotropical poison frogs. Evol Ecol. 2013;27(4):825-9. doi: 10.1007/s10682-013-9642-2. 469 2. Cummings M, Crothers L. Interacting selection diversifies warning signals in a 470 polytypic frog: an examination with the strawberry poison frog. Evol Ecol. 471 2013;27(4):693-710. doi: 10.1007/s10682-013-9648-9. 472 3. Harvey PH, Bull JJ, Pemberton M, Paxton RJ. The evolution of aposematic 473 coloration in distasteful prey: a family model. The American Naturalist. 474 1982;119(5):710-9. doi: 10.2307/2461187. 475 4. Grant T, Frost DR, Caldwell JP, Gagliardo RON, Haddad CFB, Kok PJR, et al. 476 Phylogenetic systematics of dart-poison frogs and their relatives (Amphibia: 477 Athesphatanura: Dendrobatidae). Bull Am Mus Nat Hist. 2006:1-262. doi: 10.1206/0003- 478 0090(2006)299[1:psodfa]2.0.co;2. 479 5. Langham GM, Benkman C. Specialized avian predators repeatedly attack novel 480 color morphs of Heliconius butterflies. Evolution. 2004;58(12):2783-7. 481 6. Ramos RR, Freitas A. Population biology and wing color variation in Heliconius 482 erato phyllis (Nymphalidae). J Lepid Soc. 1999;53(1):11-21. 483 7. Medina I, Wang IJ, Salazar C, Amézquita A. Hybridization promotes color 484 polymorphism in the aposematic harlequin poison frog, Oophaga histrionica. Ecol Evol. 485 2013;3(13):4388-400. doi: 10.1002/ece3.794. 486 8. Posso-Terranova A, Andrés JA. Complex niche divergence underlies lineage 487 diversification in Oophaga poison frogs. J Biogeogr. 2016;43(10):2002-15. doi: 488 10.1111/jbi.12799. 489 9. Maan ME, Cummings ME. Poison frog colors are honest signals of toxicity, 490 particularly for bird predators. The American Naturalist. 2012;179(1):E1-E14. 491 10. Saporito R, Donnelly M, Garraffo H, Spande T, Daly J. Geographic and seasonal 492 variation in alkaloid-based chemical defenses of Dendrobates pumilio from Bocas del 493 Toro, Panama. J Chem Ecol. 2006;32(4):795-814. doi: 10.1007/s10886-006-9034-y. 494 11. Saporito R, Donnelly M, Spande T, Garraffo HM. A review of chemical ecology 495 in poison frogs. Chemoecology. 2012;22(3):159-68. doi: 10.1007/s00049-011-0088-0. 496 12. Bagnara JT. Development of the spot pattern in the leopard frog. J Exp Zool. 497 1982;224(2):283-7. doi: 10.1002/jez.1402240219. 498 13. Wolnicka-Glubisz A, Pecio A, Podkowa D, Kolodziejczyk LM, Plonka PM. 499 Pheomelanin in the skin of Hymenochirus boettgeri (Amphibia: Anura: Pipidae). Exp 500 Dermatol. 2012;21(7):537-40. doi: doi:10.1111/j.1600-0625.2012.01511.x. 501 14. Daly JW. The chemistry of poisons in amphibian skin. Proceedings of the 502 National Academy of Sciences. 1995;92(1):9-13. doi: 10.1073/pnas.92.1.9. 503 15. Posso-Terranova A, Andrés JA. Diversification and convergence of aposematic 504 phenotypes: truncated receptors and cellular arrangements mediate rapid evolution of 505 coloration in harlequin poison frogs. Evolution. 2017;71(11):2677-92. 506 16. Daly JW, Brown GB, Mensah-Dwumah M, Myers CW. Classification of skin 507 alkaloids from neotropical poison-dart frogs (Dendrobatidae). Toxicon. 1978;16(2):163- 508 88. doi: http://dx.doi.org/10.1016/0041-0101(78)90036-3. 509 17. Daly JW, Garraffo HM, Spande TF, Clark VC, Ma J, Ziffer H, et al. Evidence for 510 an enantioselective pumiliotoxin 7-hydroxylase in dendrobatid poison frogs of the genus bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
511 Dendrobates. Proceedings of the National Academy of Sciences. 2003;100(19):11092-7. 512 doi: 10.1073/pnas.1834430100. 513 18. Daly JW, Kaneko T, Wilham J, Garraffo HM, Spande TF, Espinosa A, et al. 514 Bioactive alkaloids of frog skin: Combinatorial bioprospecting reveals that pumiliotoxins 515 have an arthropod source. Proceedings of the National Academy of Sciences. 516 2002;99(22):13996-4001. doi: 10.1073/pnas.222551599. 517 19. Berthold AA. Ueber verschiedene neue oder seltene reptilien aus Neu-Granada 518 und crustaceen aus China. Nachr Ges Wiss Göttingen. 1846;3(3):43. 519 20. Myers CW, Daly JW. Preliminary evaluation of skin toxins and vocalizations in 520 taxonomic and evolutionary studies of poison-dart frogs (Dendrobatidae). Bull Am Mus 521 Nat Hist. 1976;(157):175-262. PubMed PMID: 4238030. 522 21. Funkhouser J. New frogs from Ecuador and southwestern Colombia. Zoologica. 523 1956;41:73-80. 524 22. Posso-Terranova A, Andrés J. Multivariate species boundaries and conservation 525 of harlequin poison frogs. Mol Ecol. 2018;27(17):3432-51. doi: 10.1111/mec.14803. 526 23. Rogers RL, Summers K, Wu Y, Guo C, Zheng J, Xun X, et al. Genomic takeover 527 by transposable elements in the strawberry poison frog. Mol Biol Evol. 528 2018;35(12):2913-27. doi: 10.1093/molbev/msy185. 529 24. Dodt M, Roehr J, Ahmed R, Dieterich C. FLEXBAR—Flexible barcode and 530 adapter processing for next-generation sequencing platforms. Biology. 2012;1(3):895- 531 905. PubMed PMID: doi:10.3390/biology1030895. 532 25. Haas BJ, Papanicolaou A, Yassour M, Grabherr M, Blood PD, Bowden J, et al. 533 De novo transcript sequence reconstruction from RNA-seq using the Trinity platform for 534 reference generation and analysis. Nat Protocols. 2013;8(8):1494-512. doi: 535 10.1038/nprot.2013.084 536 http://www.nature.com/nprot/journal/v8/n8/abs/nprot.2013.084.html#supplementary- 537 information. 538 26. Gould B, McCouch S, Geber M. De Novo transcriptome assembly and 539 identification of gene candidates for rapid evolution of soil Al tolerance in Anthoxanthum 540 odoratum at the long-term park grass experiment. PLoS One. 2015;10(7):e0124424. doi: 541 10.1371/journal.pone.0124424. 542 27. Zheng Y, Zhao L, Gao J, Fei Z. iAssembler: a package for de novo assembly of 543 Roche-454/Sanger transcriptome sequences. BMC Bioinformatics. 2011;12(1):1-8. doi: 544 10.1186/1471-2105-12-453. 545 28. Conesa A, Götz S, García-Gómez JM, Terol J, Talón M, Robles M. Blast2GO: a 546 universal tool for annotation, visualization and analysis in functional genomics research. 547 Bioinformatics. 2005;21(18):3674-6. doi: 10.1093/bioinformatics/bti610. 548 29. Waterhouse RM, Seppey M, Simão FA, Manni M, Ioannidis P, Klioutchnikov G, 549 et al. BUSCO Applications from quality assessments to gene prediction and 550 phylogenomics. Mol Biol Evol. 2018;35(3):543-8. doi: 10.1093/molbev/msx319. 551 30. Nishimura O, Hara Y, Kuraku S. gVolante for standardizing completeness 552 assessment of genome and transcriptome assemblies. Bioinformatics. 2017;33(22):3635- 553 7. doi: 10.1093/bioinformatics/btx445. 554 31. Holding ML, Margres MJ, Mason AJ, Parkinson CL, Rokyta DR. Evaluating the 555 performance of De Novo assembly methods for venom-gland transcriptomics. Toxins 556 (Basel). 2018;10(6):249. PubMed PMID: doi:10.3390/toxins10060249. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
557 32. Roberts A, Pachter L. Streaming fragment assignment for real-time analysis of 558 sequencing experiments. Nat Meth. 2013;10(1):71-3. doi: 10.1038/nmeth.2251 559 http://www.nature.com/nmeth/journal/v10/n1/abs/nmeth.2251.html#supplementary- 560 information. 561 33. Wagner GP, Kin K, Lynch VJ. Measurement of mRNA abundance using RNA- 562 seq data: RPKM measure is inconsistent among samples. Theory Biosci. 563 2012;131(4):281-5. doi: 10.1007/s12064-012-0162-3. 564 34. Huang L, Li J, Anboukaria H, Luo Z, Zhao M, Wu H. Comparative transcriptome 565 analyses of seven anurans reveal functions and adaptations of amphibian skin. Sci Rep. 566 2016;6:24069. doi: 10.1038/srep24069 567 https://www.nature.com/articles/srep24069#supplementary-information. 568 35. Eskew EA, Shock BC, LaDouceur EEB, Keel K, Miller MR, Foley JE, et al. Gene 569 expression differs in susceptible and resistant amphibians exposed to Batrachochytrium 570 dendrobatidis. Royal Society Open Science. 2018;5(2):170910. doi: 571 10.1098/rsos.170910. 572 36. Santos JC, Tarvin RD, O'Connell LA, Blackburn DC, Coloma LA. Diversity 573 within diversity: Parasite species richness in poison frogs assessed by transcriptomics. 574 Mol Phylogenet Evol. 2018;125:40-50. doi: https://doi.org/10.1016/j.ympev.2018.03.015. 575 37. Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat 576 Methods. 2012;9:357. doi: 10.1038/nmeth.1923 577 https://www.nature.com/articles/nmeth.1923#supplementary-information. 578 38. Nicholls SM, Aubrey W, de Grave K, Schietgat L, Creevey CJ, Clare A. 579 Probabilistic recovery of cryptic haplotypes from metagenomic data. bioRxiv. 2017. doi: 580 10.1101/117838. 581 39. Nanoth Vellichirammal N, Zera AJ, Schilder RJ, Wehrkamp C, Riethoven J-JM, 582 Brisson JA. De Novo trascriptome assembly from fat body and flight muscles transcripts 583 to identify morph-specific gene expression profiles in Gryllus firmus. PLoS One. 584 2014;9(1):e82129. doi: 10.1371/journal.pone.0082129. 585 40. Schurch NJ, Schofield P, Gierliński M, Cole C, Sherstnev A, Singh V, et al. How 586 many biological replicates are needed in an RNA-seq experiment and which differential 587 expression tool should you use? RNA. 2016;22(6):839-51. doi: 10.1261/rna.053959.115. 588 41. Anders S, Huber W. Differential expression analysis for sequence count data. 589 Genome Biol. 2010;11(10):R106. doi: 10.1186/gb-2010-11-10-r106. 590 42. Hammer Ø, Harper D, Ryan P. PAST: Paleontological statistics software package 591 for education and data analysis. . Palaeontolia Electronica. 2001;4(1):9. 592 43. Warnes GR, Bolker B, Bonebakker L, Gentleman R, Liaw WHA, Lumley T, et al. 593 gplots: Various R programming tools for plotting data. R package version 301. 2016;See 594 https://CRAN.R-project.org/package=gplots. 595 44. Xie H, Wasserman A, Levine Z, Novik A, Grebinskiy V, Shoshan A, et al. Large- 596 scale protein annotation through gene ontology. Genome Res. 2002;12(5):785-94. doi: 597 10.1101/gr.86902. PubMed PMID: PMC186564. 598 45. Qi F, Li A, Inagaki Y, Kokudo N, Tamura S, Nakata M, et al. Antitumor activity 599 of extracts and compounds from the skin of the toad bufo Bufo gargarizans Cantor. Int 600 Immunopharmacol. 2011;11(3):342-9. doi: https://doi.org/10.1016/j.intimp.2010.12.007. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
601 46. Stranger BE, Forrest MS, Dunning M, Ingle CE, Beazley C, Thorne N, et al. 602 Relative impact of nucleotide and copy number variation on gene expression phenotypes. 603 Science. 2007;315(5813):848-53. doi: 10.1126/science.1136678. 604 47. Gamazon ER, Wheeler HE, Shah KP, Mozaffari SV, Aquino-Michaels K, Carroll 605 RJ, et al. A gene-based association method for mapping traits using reference 606 transcriptome data. Nat Genet. 2015;47(9):1091-8. doi: 10.1038/ng.3367 607 http://www.nature.com/ng/journal/v47/n9/abs/ng.3367.html#supplementary-information. 608 48. Silverstone PA. A revision of the poison-arrow frogs of the genus Dendrobates 609 Wagler. The American Naturalist. 1975;21:1-55. 610 49. Silverstone PA. Observations on the Behavior and Ecology of a Colombian 611 Poison-Arrow Frog, the Kõkoé-Pá (Dendrobates histrionicus Berthold). Herpetologica. 612 1973;29(4):295-301. doi: 10.2307/3891565. 613 50. Daly JW, Martin Garraffo H, Spande TF, Jaramillo C, Stanley Rand A. Dietary 614 source for skin alkaloids of poison frogs (Dendrobatidae). J Chem Ecol. 1994;20(4):943- 615 55. doi: 10.1007/bf02059589. 616 51. Daly JW, Spande TF, Garraffo HM. Alkaloids from Amphibian Skin: A 617 Tabulation of Over Eight-Hundred Compounds. J Nat Prod. 2005;68(10):1556-75. doi: 618 10.1021/np0580560. 619 52. Daly JW, Ware N, Saporito RA, Spande TF, Garraffo HM. N- 620 Methyldecahydroquinolines: An unexpected class of alkaloids from Amazonian poison 621 frogs (Dendrobatidae). J Nat Prod. 2009;72(6):1110-4. doi: 10.1021/np900094v. 622 53. Hashimoto T, Yamada Y. Alkaloid biogenesis: molecular aspects. Annu Rev 623 Plant Biol. 1994;45(1):257-85. 624 54. Hartmann T. Chemical ecology of pyrrolizidine alkaloids. Planta. 625 1999;207(4):483-95. 626 55. Yazaki K. Transporters of secondary metabolites. Curr Opin Plant Biol. 627 2005;8(3):301-7. doi: https://doi.org/10.1016/j.pbi.2005.03.011. 628 56. Carqueijeiro I, Noronha H, Duarte P, Gerós H, Sottomayor M. Vacuolar transport 629 of the medicinal alkaloids from Catharanthus roseus is mediated by a proton-driven 630 antiport. Plant Physiol. 2013;162(3):1486-96. 631 57. Otani M, Shitan N, Sakai K, Martinoia E, Sato F, Yazaki K. Characterization of 632 vacuolar transport of the endogenous alkaloid berberine in Coptis japonica. Plant 633 Physiol. 2005;138(4):1939-46. doi: 10.1104/pp.105.064352. 634 58. Shitan N, Yazaki K. Accumulation and membrane transport of plant alkaloids. 635 Curr Pharm Biotechnol. 2007;8(4):244-52. doi: 636 http://dx.doi.org/10.2174/138920107781387429. 637 59. Sakai K, Shitan N, Sato F, Ueda K, Yazaki K. Characterization of berberine 638 transport into Coptis japonica cells and the involvement of ABC protein. J Exp Bot. 639 2002;53(376):1879-86. doi: 10.1093/jxb/erf052. 640 60. Shitan N, Bazin I, Dan K, Obata K, Kigawa K, Ueda K, et al. Involvement of 641 CjMDR1, a plant multidrug-resistance-type ATP-binding cassette protein, in alkaloid 642 transport in Coptis japonica. Proceedings of the National Academy of Sciences. 643 2003;100(2):751-6. doi: 10.1073/pnas.0134257100. 644 61. Daly JW, McNeal ET, Overman LE, Ellison DH. A new class of cardiotonic 645 agents: structure-activity correlations for natural and synthetic analogs of the alkaloid bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
646 pumiliotoxin B (8-hydroxy-8-methyl-6-alkylidene-1-azabicyclo[4.3.0]nonanes). J Med 647 Chem. 1985;28(4):482-6. doi: 10.1021/jm00382a017. 648 62. Chaudhary A, Willett KL. Inhibition of human cytochrome CYP1 enzymes by 649 flavonoids of St. John's wort. Toxicology. 2006;217(2–3):194-205. doi: 650 http://dx.doi.org/10.1016/j.tox.2005.09.010. 651 63. Sigel A, Sigel H, Sigel RK. The ubiquitous roles of cytochrome P450 proteins: 652 metal ions in life sciences: John Wiley & Sons; 2007. 653 64. Santos JC, Tarvin RD, O’Connell LA. A review of chemical defense in poison 654 frogs (Dendrobatidae): ecology, pharmacokinetics, and autoresistance. In: Schulte AB, 655 Goodwin ET, Ferkin HM, editors. Chemical Signals in Vertebrates 13. Cham: Springer 656 International Publishing; 2016. p. 305-37. 657 65. Giddings L-A, Liscombe DK, Hamilton JP, Childs KL, DellaPenna D, Buell CR, 658 et al. A stereoselective hydroxylation step of alkaloid biosynthesis by a unique 659 cytochrome P450 in Catharanthus roseus. J Biol Chem. 2011;286(19):16751-7. doi: 660 10.1074/jbc.M111.225383. 661 66. Dobler S, Dalla S, Wagschal V, Agrawal AA. Community-wide convergent 662 evolution in insect adaptation to toxic cardenolides by substitutions in the Na,K-ATPase. 663 Proceedings of the National Academy of Sciences. 2012;109(32):13040-5. doi: 664 10.1073/pnas.1202111109. 665 67. Labeyrie E, Dobler S. Molecular adaptation of Chrysochus leaf beetles to toxic 666 compounds in their food plants. Mol Biol Evol. 2004;21(2):218-21. doi: 667 10.1093/molbev/msg240. 668 68. Aardema ML, Zhen Y, Andolfatto P. The evolution of cardenolide-resistant forms 669 of Na+,K+-ATPase in Danainae butterflies. Mol Ecol. 2012;21(2):340-9. doi: 670 10.1111/j.1365-294X.2011.05379.x. 671 69. Clark VC, Raxworthy CJ, Rakotomalala V, Sierwald P, Fisher BL. Convergent 672 evolution of chemical defense in poison frogs and arthropod prey between Madagascar 673 and the Neotropics. Proceedings of the National Academy of Sciences. 674 2005;102(33):11617-22. doi: 10.1073/pnas.0503502102. 675 70. Tarvin RD, Santos JC, O'Connell LA, Zakon HH, Cannatella DC. Convergent 676 substitutions in a sodium channel suggest multiple origins of toxin resistance in poison 677 frogs. Mol Biol Evol. 2016;33(4)(4):1068-81. doi: 10.1093/molbev/msv350. 678 71. Gamberale-Stille G, Guilford T. Contrast versus colour in aposematic signals. 679 Anim Behav. 2003;65(5):1021-6. doi: http://dx.doi.org/10.1006/anbe.2003.2098. 680 72. Bowdish TI, Bultman TL. Visual cues used by mantids in learning aversion to 681 aposematically colored prey. Am Midl Nat. 1993:215-22. 682 73. Sherratt T, Beatty C. The evolution of warning signals as reliable indicators of 683 prey defense. The American Naturalist. 2003;162:377-89. 684 74. Prudic KL, Skemp AK, Papaj DR. Aposematic coloration, luminance contrast, 685 and the benefits of conspicuousness. Behav Ecol. 2007;18(1):41-6. doi: 686 10.1093/beheco/arl046. 687 75. Aronsson M, Gamberale-Stille G. Evidence of signaling benefits to contrasting 688 internal color boundaries in warning coloration. Behav Ecol. 2012;24(2)(2):349-54. doi: 689 10.1093/beheco/ars170. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
690 76. Aronsson M, Gamberale-Stille G. Importance of internal pattern contrast and 691 contrast against the background in aposematic signals. Behav Ecol. 2009;20(6):1356-62. 692 doi: 10.1093/beheco/arp141. 693 77. Bagnara JT, Taylor JD, Hadley ME. The dermal chromatophore unit. The Journal 694 of Cell Biology. 1968;38(1):67-79. doi: 10.1083/jcb.38.1.67. 695 78. Bagnara J, Taylor J. Differences in pigment-containing organelles between color 696 forms of the red-backed salamander, Plethodon cinereus. Z Zellforsch Mikrosk Anat. 697 1970;106(3):412-7. doi: 10.1007/bf00335782. 698 79. Bagnara JT, Frost SK, Matsumoto J. On the development of pigment patterns in 699 mmphibians. Am Zool. 1978;18(2):301-12. doi: 10.1093/icb/18.2.301. 700 80. Bagnara JT, Taylor JD, Prota G. Color changes, unusual melanosomes, and a new 701 pigment from leaf frogs. Science. 1973;182(4116):1034-5. doi: 10.2307/1737830. 702 81. Palumbo A, D Cosmo A, Gesualdo I, Vincent J. Subcellular localization and 703 function of melanogenic enzymes in the ink gland of Sepia officinalis. Biochem J. 704 1997;323(3):749-56. 705 82. Sugimoto M. Morphological color changes in fish: regulation of pigment cell 706 density and morphology. Microsc Res Tech. 2002;58(6):496-503. doi: 707 10.1002/jemt.10168. 708 83. Logan DW, Burn SF, Jackson IJ. Regulation of pigmentation in zebrafish 709 melanophores. Pigment Cell Res. 2006;19(3):206-13. doi: 10.1111/j.1600- 710 0749.2006.00307.x. 711 84. Baião PC, Parker PG. Evolution of the melanocortin-1 receptor (MC1R) in 712 boobies and gannets (Aves, Suliformes). J Hered. 2012;103(3):322-9. doi: 713 10.1093/jhered/esr151. 714 85. Everts RE, Rothuizen J, van Oost BA. Identification of a premature stop codon in 715 the melanocyte-stimulating hormone receptor gene (MC1R) in Labrador and Golden 716 retrievers with yellow coat colour. Anim Genet. 2000;31(3):194-9. doi: 10.1046/j.1365- 717 2052.2000.00639.x. 718 86. Gangoso L, Grande JM, Ducrest AL, Figuerola J, Bortolotti GR, Andres JA, et al. 719 MC1R-dependent, melanin-based colour polymorphism is associated with cell-mediated 720 response in the Eleonora’s falcon. J Evol Biol. 2011;24(9):2055-63. doi: 10.1111/j.1420- 721 9101.2011.02336.x. 722 87. Gross JB, Borowsky R, Tabin CJ. A novel role for MC1R in the parallel evolution 723 of depigmentation in independent populations of the Cavefish Astyanax mexicanus. PLoS 724 Genet. 2009;5(1):e1000326. doi: 10.1371/journal.pgen.1000326. 725 88. Corso J, Gonçalves G, Thales dF. Sequence variation in the melanocortin-1 726 receptor (MC1R) pigmentation gene and its role in the cryptic coloration of two South 727 American sand lizards. Genet Mol Biol. 2012;35(1):81-7. 728 89. Herczeg G, Matsuba C, Merilä J. Sequence variation in the melanocortin-1 729 receptor gene (Mc1r) does not explain variation in the degree of melanism in a 730 widespread amphibian. Ann Zool Fennici. 2010;47:37-45. 731 90. Chikako M. Geographic variations of melanocortine 1 receptor gene (MC1R) in 732 the common frog (Rana temporaria) in Northern Europe. Amphibia-Reptilia. 733 2012;33:105-11. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
734 91. Ohbayashi N, Maruta Y, Ishida M, Fukuda M. Melanoregulin regulates retrograde 735 melanosome transport through interaction with the RILP–p150 glued complex in 736 melanocytes. J Cell Sci. 2012;125(6):1508-18. doi: 10.1242/jcs.094185. 737 92. Van Gele M, Dynoodt P, Lambert J. Griscelli syndrome: a model system to study 738 vesicular trafficking. Pigment Cell & Melanoma Research. 2009;22(3):268-82. doi: 739 10.1111/j.1755-148X.2009.00558.x. 740 93. Stuckert AMM, Moore E, Coyle KP, Davison I, MacManes MD, Roberts R, et al. 741 Variation in pigmentation gene expression is associated with distinct aposematic color 742 morphs in the poison frog Dendrobates auratus. BMC Evol Biol. 2019;19(1):85. doi: 743 10.1186/s12862-019-1410-7. 744 bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
745 SUPPORTING INFORMATION
746 SUPPORTING TABLES
747 S1 Table. Number of reads and library type (i.e paired vs. single-end) for each 748 individual RNA-seq experiment. 749
Raw number of Reads after Remaining % Library Individual library reads filtering of reads type O. solanensis 128,599,724 105,179,065 81% Paired O. anchicayensis 124,232,804 110,217,128 88% reads* O. lehmanni 102,225,680 91,493,848 89%
O. sylvatica 48,929,329 45,147,654 92%
O. lehmanni 43,633,080 41,086,379 94% Single reads O. anchicayensis) 7,998,636 7,173,285 89%
O. histrionica 50,204,617 47,376,912 94%
Reference transcriptome N/A 33,895,474 100% Single reads
*. Number of reads correspond to the combined value of both ends libraries. 750 bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
751 S2 Table. RNA-sequence datasets from cryptic species used for the differential 752 expression analysis of this work. 753 Species Experiment (SRA*) RNA-seq library Total bases Date available Bufo gargarizans SRX2640691 Paired-end (150 b) 19.2 Gbp 2017-03-15 Pelophylax nigromaculatus SRX2640690 Paired-end (150 b) 18.0 Gbp 2017-03-15 Polypedates megacephalus SRX1720193 Paired-end (100 b) 6.3 Gbp 2016-04-25 Rana catesbeiana SRX2988982 Paired-end (100 b) 9.2 Gbp 2017-07-07 Rana sylvatica SRX2989023 Paired-end (100 b) 8.9 Gbp 2017-07-07 Xenophrys sangzhiensis SRX1720191 Paired-end (100 b) 6.8 Gbp 2016-04-25 * Sequence Read Archive accession number from Genbank. 754 755 756 S3 Table. Quality assessment report for each individual and the composite reference 757 transcriptome based on the protein count of the core eukaryotic genes (BUSCO, see 758 methods in the main article) 759
Busco v.2 mapping Transcriptome Library Complete Partial Complete + Partial O. solanensis Paired 20.60% 30.47% 51.07% O. anchicayensis Paired 24.46% 37.77% 62.23% O. lehmanni Paired 29.61% 40.77% 70.38% O. sylvatica Single 45.92% 33.48% 79.40% O. lehmanni Single 33.91% 39.91% 73.82% O. anchicayensis Single 9.87% 30.47% 40.34% O. histrionica Single 39.06% 37.77% 76.83% Reference assembly Single 45.92% 42.62% 88.54% 760
761 S4 Table. Summary statistics of the differentially expressed unigenes (see methods) and 762 their BLAST annotations.
Differentially With Without expressed unigenes annotation annotation
Up-regulated 1656 (85.6%) 996 (51.6%) 660 (34.2%) Down-regulated 275 (14.2%) 256 (13.3%) 19 (1%) Total 1931 (100%) 1252 (64.9%) 679 (35.2%) 763 bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
764 S5 Table. Transcripts with potential important functions in the alkaloid sequestration system, auto-resistance to toxic compounds and 765 variation in coloration in harlequin poison frogs. The corresponding composite Oophaga transcriptome is available from the authors 766 upon request. 767 Length Transcript ID Fold change padj BLAST description GO Names list (bp) C:cell; P:gonad development; P:thyroid gland development; P:luteinizing hormone secretion; P:thyroid hormone generation; P:cellular response to hormone 14256 779.11 0.00 Transmembrane protein 41B 1253 stimulus; P:follicle-stimulating hormone secretion; P:negative
regulation of organ growth; C:extracellular region; F:hormone activity 26699 517.76 0.00 Transmembrane protein 181 1079 C:extracellular space 14644 297.83 0.00 Transmembrane protein 116-like 1743 14399 192.77 0.01 Transmembrane protein 26 1039 6476 174.07 0.00 Transmembrane protein 1690 F:structural constituent of eye lens F:chondroitin sulfate binding; C:nucleus; F:identical protein binding; F:calcium-dependent protein binding; C:cytosol; C:chromaffin granule membrane; 15428 105.09 0.02 Transmembrane protein 220 697 P:regulation of transcription,
DNA-templated; F:heparin transportation Alkaloid binding; P:negative regulation of cellular process; F:carbohydrate binding; C:apical plasma membrane 8646 100.29 0.01 Transmembrane protein 69 936 20904 66.89 0.04 Transmembrane protein 186-like 615 bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
P:'de novo' pyrimidine nucleobase biosynthetic process; F:orotidine- 19857 53.30 0.05 Transmembrane protein 70, mitochondrial 532 5'-phosphate decarboxylase activity; P:'de novo' UMP biosynthetic process 24171 45.71 0.01 Transmembrane protein 125-like 4098 Keratinocyte-associated Transmembrane protein 2- 19808 37.86 0.04 1043 like P:negative regulation of potassium ion transmembrane transport; P:positive regulation of myofibroblast contraction; P:response to salt; P:negative regulation of platelet aggregation; P:response to progesterone; 16724 27.39 0.02 Transmembrane protein 206 1589 P:positive regulation of vasoconstriction; P:negative regulation of gastrin-induced gastric acid secretion; C:neuronal cell body; P:negative regulation of insulin secretion; F:prostaglandin …. 24522 25.75 0.00 Transmembrane protease serine 13-like 2292 F:oxidoreductase activity P:response to dehydroepiandrosterone; F:gamma-catenin binding; C:extracellular space; P:response 25621 7.42 0.05 Cation-transporting ATPase 13A4 469 to 11-deoxycorticosterone; C:membrane; P:metabolic process; P:response to progesterone; P:response to estradiol; C:desmosome; F:hydrolase activity F:calcium ion binding; 9463 -6.16 0.04 Lysosomal-trafficking regulator 850 P:keratinization 768 bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
Length Transcript ID Fold change padj BLAST description GO Names list (bp) F:calcium ion binding; 15350 175.29 0.00 EF-hand calcium-binding domain-containing 10-like 436 C:cellular_component P:single-organism cellular localization; P:cell cycle process; F:binding; P:organelle organization; C:intracellular organelle part; P:single-organism transport; P:establishment of protein localization; C:endosome; C:membrane-bounded vesicle; 16951 95.46 0.01 S100 calcium binding protein Z 2161 P:vesicle-mediated transport; site insensitivity site
P:regulation of cellular process; - C:membrane; C:cytoplasmic vesicle; P:establishment of localization in cell; P:cellular protein localization; P:establishment of organelle localization C:extracellular space; F:carboxypeptidase activity; 1216 63.08 0.03 Extracellular calcium-sensing receptor-like 1214 P:bradykinin catabolic process; P:proteolysis; P:response to
glucocorticoid target resistance: - P:ion transport; P:regulation of membrane potential; C:integral component of membrane; F:extracellular ligand-gated ion Potassium Voltage-gated channel subfamily E channel activity; P:synaptic 18107 4239.03 0.00 1864 member 3-like transmission; C:synapse; F:binding; P:regulation of biological process; P:neurological
system process; P:response to auto Alkaloid stimulus; C:plasma membrane part bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
F:high voltage-gated calcium channel activity; F:protein C- terminus binding; C:voltage-gated calcium channel complex; 16614 9.90 0.04 MGC114646 protein 1514 P:calcium ion transmembrane transport; C:cytosol; P:positive regulation of cytosolic calcium ion concentration 907 -6.46 0.01 Voltage-dependent anion channel 2 990 376 -10.41 0.00 Voltage-dependent anion channel 388 C:sarcoplasmic reticulum; C:voltage-gated calcium channel complex; F:calcium channel regulator activity; F:voltage-gated calcium channel activity; F:mating pheromone activity; P:pheromone- dependent signal transduction Sodium channel, voltage-dependent, gamma subunit 8966 102.76 0.01 640 involved in conjugation with 1 cellular fusion; C:T-tubule; C:extracellular exosome; F:metal ion binding; P:mating; P:regulation of ion transmembrane transport; P:calcium ion transmembrane transport; P:regulation of calcium ion transport 769 bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
Fold change padj BLAST description Length (bp) GO Names list
420.10 0.04 Cytochrome P450-like TBP 408 36.47 0.01 Cytochrome P450, family 4, subfamily 458 F:oxygen binding; P:oxygen transport; C:voltage-gated F, polypeptide 22 potassium channel complex; F:delayed rectifier potassium channel activity; F:A-type (transient
outward) potassium channel activity; P:protein homooligomerization; C:neuronal cell body; P:membrane repolarization; F:oxygen transporter activity; F:ion channel binding; F:heme binding; C:hemoglobin complex; C:dendrite; C:caveola; C:perinuclear endoplasmic reticulum; F:iron ion
binding; P:cellular response to BMP stimulus; C:sarcolemma; P:regulation of ion transmembrane transport; P:potassium ion export
30.72 0.02 Cytochrome P450 2F2-like 1201 metabolic resistance: - 17.97 0.00 Cytochrome P450 2D26-like 1851 P:sucrose biosynthetic process; C:chromosome;
F:binding detoxification 6.93 0.04 Cytochrome P450 2K1-like 429 C:nucleus; F:DNA binding; P:cytoskeleton organization; F:protein heterodimerization activity; F:calcium ion binding; C:nucleosome; F:calcium- dependent phospholipid binding
auto Alkaloid
bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
Fold change Padj BLAST description Length (bp) GO Names list 273.95 0.00 Melanoregulin-like protein 1131 25.44 0.03 Melanocortin receptor binding NECAP 2070 P:regulation of appetite; C:secretory granule; Coloration endocytosis associated 2 C:peroxisomal matrix; P:positive regulation of transcription from RNA polymerase II promoter; F:hormone activity; P:regulation of blood pressure; C:extracellular space; P:neuropeptide signaling pathway; P:glucose homeostasis; F:type 1 melanocortin receptor binding; P:cellular pigmentation; P:regulation of corticosterone secretion; F:type 3 melanocortin receptor binding; F:type 4 melanocortin receptor binding; P:regulation of glycogen metabolic process; P:negative regulation of tumor necrosis factor production 12.71 0.05 Melanocortin m-phase phosphoprotein 2732 P:regulation of nucleobase-containing compound 8 metabolic process; P:G-protein coupled receptor signaling pathway; P:single-multicellular organism process; F:peptide binding; F:melanocortin receptor activity; F:protein binding; P:cellular macromolecule metabolic process; F:carbohydrate derivative binding; C:intracellular membrane-bounded organelle; P:homeostatic process; C:membrane; P:regulation of gene expression; C:macromolecular complex 13.52 0.00 Tyrosine kinase, non-receptor 2 1981 F:protein binding; C:proton-transporting ATP synthase complex; P:single-organism cellular process; P:nucleobase-containing compound metabolic process; P:transport; C:mitochondrial inner membrane; F:hydrolase activity
770 bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
771 SUPPORTING FIGURES
772 S1 Fig. Pie charts representing the number and proportion of contigs with significant 773 BLAST hits (E<1.0E-5) for each individual transcriptome from Oophaga species. 774 Transcriptomes obtained using paired-ends RNA-seq libraries are represented in (A) and 775 those obtained with single-end libraries in (A). bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
776 S2 Fig. A) Contig length distribution of the de-novo composite reference transcriptome 777 of Oophaga species from skin tissue. B) Pie chart representing the number and proportion 778 of contigs with significant BLAST hits (E<1.0E-5) in the reference Oophaga 779 transcriptome. C) Gene ontology (GO) categories distribution (level II) for the annotated 780 unigenes in the Oophaga reference transcriptome. bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
781 S3 Fig. Distribution plots of the effective number of reads (ER) from Oophaga RNA-seq 782 experiments that mapped to the composite reference transcriptome contigs (n=31,498). bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
783 S4 Fig. a) Percentile plot of the estimated transcripts per million (TPM) in the composite 784 reference transcriptome as calculated based on raw RNA reads from individual libraries. 785 Color names are equivalent to those in tendency lines. b) Principal component analyses 786 (PCA) plot of the reference contigs dataset using TPMs values as independent variables. 787 Red dots represent the selected highly represented unigenes (2% of the total contigs; n= 788 1,437) while dark dots represent the remaining ones (n=30,061) bioRxiv preprint doi: https://doi.org/10.1101/706655; this version posted July 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
789 S5 Fig. A) Pie chart representing the total number of differentially expressed unigenes 790 and the proportion with significant BLAST hits (E<1.0E-5). B) BLAST hit species 791 distribution of highly expressed unigenes. C) Gene ontology (GO) categories distribution 792 (multilevel) for the annotated highly expressed unigenes.