Miniscule differences between the sex chromosomes in the giant genome of a salamander, Ambystoma mexicanum

In the Mexican axolotl (Ambystoma mexicanum) sex is known to be determined by a single Mendelian factor, yet the sex chromosomes of this model salamander do not exhibit morphological differentiation that is typical of many vertebrate taxa that possess a single sex-determining locus. Differentiated sex chromosomes are thought to evolve rapidly in the context of a Mendelian sex-determining gene and, therefore, undifferentiated chromosomes provide an exceptional opportunity to reconstruct early events in sex chromosome evolution. Whole chromosome sequencing, whole genome resequencing (48 individuals from a backcross of axolotl and tiger salamander) and in situ hybridization were used to identify a homomorphic chromosome that carries an A. mexicanum sex determining factor and identify sequences that are present only on the W chromosome. Altogether, these sequences cover ~300 kb, or roughly 1/100,000th of the ~32 Gb genome. Notably, these W-specific sequences also contain a recently duplicated copy of the ATRX gene: a known component of mammalian sex-determining pathways. This gene (designated ATRW) is one of the few functional (non-repetitive) genes in the chromosomal segment and maps to the tip of chromosome 9 near the marker E24C3, which was previously found to be linked to the sex-determining locus. These analyses provide highly predictive markers for diagnosing sex in A. mexicanum and identify ATRW as a strong candidate for the primary sex determining locus or alternately a strong candidate for a recently acquired, sexually antagonistic gene. AUTHOR SUMMARY Sex chromosomes are thought to follow fairly stereotypical evolutionary trajectories that result in differentiation of sex-specific chromosomes. In the salamander A. mexicanum (the axolotl), sex is determined by a single Mendelian locus, yet the sex chromosomes are essentially undifferentiated, suggesting that these sex chromosomes have recently acquired a sex locus and are in the early stages of differentiating. Although Mendelian sex determination was first reported for the axolotl more than 70 years ago, no sex-specific sequences have been identified for this important model species. Here, we apply new technologies and approaches to identify and validate a tiny region of female-specific DNA within the gigantic genome of the axolotl (1/100,000th of the genome). This region contains a limited number of genes, including a duplicate copy of the ATRX gene which, has been previously shown to contribute to mammalian sex determination. Our analyses suggest that this gene, which we refer to as ATRW, evolved from a recent duplication and presents a strong candidate for the primary sex determining factor of the axolotl, or alternately a recently evolved sexually antagonistic gene.


ABSTRACT 23
In the Mexican axolotl (Ambystoma mexicanum) sex is known to be determined by a 24 single Mendelian factor, yet the sex chromosomes of this model salamander do not 25 exhibit morphological differentiation that is typical of many vertebrate taxa that possess 26 a single sex-determining locus. Differentiated sex chromosomes are thought to evolve 27 rapidly in the context of a Mendelian sex-determining gene and, therefore, 28 undifferentiated chromosomes provide an exceptional opportunity to reconstruct early 29 events in sex chromosome evolution. Whole chromosome sequencing, whole genome 30 resequencing (48 individuals from a backcross of axolotl and tiger salamander) and in 31 situ hybridization were used to identify a homomorphic chromosome that carries an A. 32 mexicanum sex determining factor and identify sequences that are present only on the 33 W chromosome. Altogether, these sequences cover ~300 kb, or roughly 1/100,000 th of 34 the ~32 Gb genome. Notably, these W-specific sequences also contain a recently 35 duplicated copy of the ATRX gene: a known component of mammalian sex-determining 36 pathways. This gene (designated ATRW) is one of the few functional (non-repetitive) 37 genes in the chromosomal segment and maps to the tip of chromosome 9 near the 38 marker E24C3, which was previously found to be linked to the sex-determining locus. 39 These analyses provide highly predictive markers for diagnosing sex in A. mexicanum 40 and identify ATRW as a strong candidate for the primary sex determining locus or 41 alternately a strong candidate for a recently acquired, sexually antagonistic gene. In many species, sex is determined by the inheritance of highly differentiated 63 (heteromorphic) sex chromosomes, which have evolved independently many times 64 throughout the tree of life (1-3). Often these chromosomes differ dramatically in 65 morphology and gene content (4-6). In mammals, males have a large, gene rich X-66 chromosome and a degraded, gene poor Y-chromosome, while females have two X 67 chromosomes. In birds and many other eukaryotes, females are the heterogametic sex 68 with a large Z and smaller W chromosome, while males are homozygous, carrying two 69 Z chromosomes. Differentiated sex chromosomes are thought to arise through a 70 relatively stereotypical process that begins when a sex-determining gene arises on a 71 pair of homologous autosomes (5, 6). The acquisition of sexually antagonistic alleles, 72 alleles that benefit one sex and are detrimental to the other, favors the fixation of 73 mutational events that suppress recombination in the vicinity of the sex-determining 74 locus (7,8). Recombination suppression can lead to the accumulation of additional 75 sexually antagonistic mutations and repetitive elements, and over time this results in the 76 loss of nonessential parts of the Y or W chromosome, resulting in the formation of 77 heteromorphic sex chromosomes (9). 78 Unlike the majority of mammals and birds with stable sex-determining systems 79 and heteromorphic sex chromosomes, amphibians have undergone numerous 80 evolutionary transitions between XY and ZW-type mechanisms and may possess 81 morphologically indistinguishable (homomorphic) sex chromosomes, like those of the 82 axolotl (10-13). Homomorphic sex chromosomes are not altogether rare among 83 animals, with examples in fish (14), birds (15), reptiles (16) and amphibians (17). 84 Among most amphibians that have been investigated, homomorphy is prevalent (17-85 19). It has been suggested that a majority of salamanders have homomorphic sex 86 chromosomes (18,20), however, evidence for genetic sex determination in most 87 species is largely based on the observation of 1:1 sex ratios from clutches without 88 thorough demonstration of Mendelian inheritance. 89 Early developmental/genetic experiments revealed a ZW type sex-determining 90 mechanism for A. mexicanum (21)(22)(23). The first experiment to test for female 91 heterogamety involved sex reversal through implantation of a testis preprimordium from 92 a donor embryo to a host female embryo. The prospective ovary developed instead into 93 a functional testis. This sex-reversed male was then crossed with a normal female (24). 94 It was expected that if the female were homozygous for sex (XX), the offspring would all 95 be female. If the female were heterozygous for sex (ZW), however, the offspring would 96 have an approximate female to male ratio of 3:1. Two matings with the sex-reversed 97 animals produced a combined 26.1% males, consistent with the hypothesis that the 98 male was indeed a sex-reversed female with ZW chromosomes (21,24). Subsequent 99 studies showed normal sex ratios from matings with the F1 males and most of the F1 100 females, but several of the F1 females produced spawns of all females, suggesting they 101 carried the unusual WW genotype (24). 102 Following these foundational studies, early genetic mapping studies used cold 103 shock to inhibit meiosis II and assessed triploid phenotypes to estimate the frequencies 104 of equatorial separation and map distances between recessive mutations and their 105 linked centromeres (25). Based on these analyses, the sex determining locus was 106 predicted to occur near the end of an undefined chromosome (25) and later estimated 107 to be 59.1 cM distal to the centromere (essentially, freely recombining) (23). 108 Karyotypic analyses later indicated that the smallest chromosomes were 109 heteromorphic in Ambystoma species, suggesting that the smallest pair of 110 chromosomes carried the Mendelian sex determining factor in A. mexicanum (26) and 111 in the A. jeffersonianum species complex (27). However, more recent linkage mapping 112 studies indicated that sex was determined by a locus on one of the larger linkage 113 groups (26, 28), and chromosome sequencing studies have demonstrated that the 114 smallest chromosomes do not carry the sex determining region (29,30). Notably, 115 extensive cytogenetic studies performed by Callan (31), including the use of cold 116 treatments to add constrictions to chromosomes and examination of lampbrush 117 chromosomes from developing oocytes, revealed no features that could be associated 118 with differentiated sex chromosomes. These analyses not only indicated that the sex 119 chromosomes were apparently identical to one another, but also revealed that mitotic 120 chromosomes 9, 10 and 11 were essentially indistinguishable from one another (31). 121 More recently, meiotic mapping of polymorphisms within controlled crosses 122 localized the sex-determining region to the tip of Ambystoma LG9 (previously 123 designated LG5) distal to the marker E24C3 (29). These crosses included a mapping 124 panel that was generated by backcrossing female A. mexicanum/A. tigrinum hybrids 125 with male A. mexicanum. These crosses also revealed no difference in recombination 126 frequencies between the sexes. However, these studies were somewhat limited by the 127 fact that they did not sample large numbers of markers in close proximity to the sex 128 locus or W-specific sequences (29). Taken together, analyses of the Ambystoma sex 129 determination suggest that the sex chromosomes are largely undifferentiated and that, 130 presumably, the sex chromosomes arose recently within the tiger salamander species 131

complex. 132
To identify sex-linked (W-specific) regions in the undifferentiated sex 133 chromosomes of axolotl, we generated sequence reads for 48 individuals of known sex 134 that were derived from a backcross (A. mexicanum/A. tigrinum X A. mexicanum). These 135 reads were then aligned to an existing reference genome from a female axolotl (30, 32) 136 (www.ambystoma.org). Analyses of read coverage identified 152 putative W-linked 137 sequences, including two genes, an ATRX paralog and an ortholog of MAP2K3. The W-138 linked ATRX paralog, ATRW, is estimated to have duplicated within the last 20 million 139 years, providing an estimate of the possible origin of the sex-determining locus in the 140 tiger salamander species complex. In addition, we anticipate that these sex-linked 141 markers will be useful for identifying sex in juvenile axolotls within lab-reared 142 populations, where sex is an important covariate for experimental studies, including 143 studies of metamorphosis and regeneration (28,33). 144

Identification of the sex-bearing chromosomes by FISH 147
Previous studies have demonstrated that sex is linked to the marker E24C3, at a 148 distance of ~5.9 cM distal to the terminal marker on LG9 (29). Consistent with linkage 149 analyses, E24C3 was detected near the tip of an average-sized chromosome ( Figure  150 1). A second BAC corresponding to a marker from the opposite end of LG9 (E12A6) 151 localized to the opposite tip of the same chromosome, indicating that this chromosome 152 corresponds precisely to LG9 ( Figure 1). Notably, the BAC carrying E12A6 also cross-153 hybridized with the centromere of all chromosomes, a feature that could potentially be 154 useful in estimating distances of genes to their respective centromeres. 155 156 Laser capture, sequencing and assembly of the Z chromosome 157 In an attempt to increase the number of markers that could be associated with the sex 158 chromosome, we performed laser-capture sequencing on a chromosome corresponding 159 to LG9. This library was generated from a single dyad that was collected in a larger 160 series of studies on laser capture microscopy of axolotl chromosomes (34). The sex 161 chromosome library contained a total of ~143 M reads between 40 and 100 bp after 162 trimming and contained 995 reads that mapped to 23 distinct markers (transcripts) that 163 had been previously placed on LG9 (Figure 2). In total, this initial sequencing run 164 accounted for 40% of the markers that are known to exist on the linkage group, which 165 was considered strong evidence that this library sampled the sex chromosome. Given 166 this support, an additional lane of sequencing was performed, yielding ~936 M 167 additional reads (for a total of 1,078,893,614 reads). After trimming, ~542 M reads 168 remained. Alignment to human and bacterial genomes revealed that 1.7% and 0.1% of 169 trimmed reads aligned concordantly to the human genome and bacterial genomes, 170 respectively. These were considered contaminants and were removed from subsequent 171 analyses. Of the remaining reads, 68,844 aligned to 40 LG9 contigs representing 70% 172 of the known markers on LG9 ( Figure 2). An error-corrected assembly of these data 173 yielded a total of 1,232,131 scaffolds totaling 242.4 Mb with a scaffold N50 length of 295 174 bp, and contig N50 length of 126 bp. (Table 1: results from other chromosomes are 175 shown for comparison purposes). We also used this library to identify a set of scaffolds 176 from a recently published assembly of a male axolotl genome that could be assigned to 177 the Z chromosome on the basis of sequence coverage. This analysis yielded 2531 178 scaffolds spanning a total of 1.02 Gb (Supplementary Table 1 Figure 3B). While a ZW-type mechanism for sex 198 determination has been inferred for the newt (37), the exact chromosome that 199 determines sex is unknown and no candidate genes currently exist. 200 201

In silico identification of female-specific regions 202
To identify sex-specific regions of the genome, we aligned low coverage sequence data 203 from 26 males and 22 females to both the LG9 assembly and the first public draft 204 assembly of the axolotl genome (30, 32) (www.ambystoma.org). The draft assembly 205 was generated using a modified version of SparseAssembler (38) from 600 Gb of HiSeq 206 paired end reads and 640Gb of HiSeq mate pair reads. Sequencing data were 207 produced using DNA from a female axolotl, which should contain genomic regions from 208 both Z and W chromosomes. Notably, a recently published draft genome was generated 209 from a male and is not expected to represent W-specific regions (39). Males and 210 females used for re-sequencing efforts were drawn from a previously published meiotic 211 mapping panel, which was used in the initial mapping of the sex locus (29). Each 212 individual was sequenced to ~1X coverage with Illumina HiSeq short paired-end reads 213 (125bp) resulting in ~7.4 billion total male reads and 6.4 billion total female reads. The 214 ratio of female to male coverage was calculated across ~10.5M intervals covering ~19 215 Gb of the draft assembly. Genome-wide coverage ratios generally fell within a tight 216 distribution centered on equal coverage, after accounting for initial differences in 217 average depth of coverage ( Figure 4). Intervals were considered to be candidate sex-218 specific regions if enrichment scores [log2 (female coverage/adjusted male coverage)] 219 exceeded two. In total, these analyses identified only 201 candidate female-specific 220 intervals that were contained within 109 genomic scaffolds, with 20 genomic scaffolds 221 having 2 or more intervals (Supplementary Table 2). The combined size of these 222 intervals is approximately 300Kb or ~0.0094% of the genome. 47 intervals were 223 represented by zero male reads, and the average male coverage of male reads for 224 other intervals ranged from 0.002 to 8.63. 225

PCR validation of candidate regions 227
PCR primers were designed for all candidate scaffolds and subjected to initial PCR 228 validation using a panel of six females and six males (Supplementary Table 3). In total, 229 primers from 42 of the 109 scaffolds yielded specific amplicons in all females and no 230 amplicons from males and were considered sex-specific. The combined size of these 231 scaffolds is approximately 174Kb or ~0.0054% of the genome. Aside from the PCR 232 validated female-specific scaffolds, primers from one scaffold were present in all 233 females and one male, two were present in four females and no males, and four were 234 present in a subset of the animals with no specific trend toward one sex or the other. 235 Presumably these represent structural (insertion/deletion) variants that are segregating 236 within the lab population of A. mexicanum, perhaps representing tiger salamander (A. A. mexicanum (male) genome. These revealed that several predicted W-specific contigs 243 correspond to copies of repetitive elements with highly similar sequences elsewhere in 244 the genome, which appears to explain a majority of cases wherein primers yield 245 amplicons in both sexes or are polymorphic among males and females. 246 247

Identifying W-specific genes 248
To search for evidence of sex-specific genes, all 42 validated sex-specific scaffolds 249 were aligned (blastx) to the NCBI nonredundant protein database (41). In total, these 250 searches yielded alignments to 17 protein-coding genes (

271
The identification of a sex-linked ATRX homolog is notable as ATRX is known to 272 play major roles in sex determination in mammals and other vertebrates (45)(46)(47)(48). 273 Alignments between scaffold SuperContig_990642 and the autosomal ATRX homolog 274 revealed that two distinct ATRX homologs exist in axolotl ( Figure 5). Alignments 275 between ATRX and its sex-specific duplicate show polymorphisms in the ATRX gene 276 that are not present in sex-linked ATRX, characteristic of a hemizygously-inherited 277 duplication (Supplementary Figure 1). Henceforth, we refer to the conserved syntenic 278 homolog on LG2 as ATRX and the W-specific homolog as ATRW. A nucleotide 279 alignment between the axolotl ATRX and ATRW genes shows that the genes share 280 90% identity across 1089 aligned nucleotides, and as such it appears that the two 281 genes diverged relatively recently by transposition of a duplicate gene copy to the W 282 chromosome. To further test this idea and better define the timing of this duplication, 283 several trees were generated using ATRX homologs from multiple vertebrate taxa 284 ( Figure 6, Supplementary Figure 2). Based on these trees, we infer that a duplication 285 event gave rise to ATRW within Ambystoma, after divergence from its common 286 ancestor with newt (the two lineages shared a common ancestor ~151 MYA) (49). 287 Considering the degree of sequence divergence and the relative length of shared vs. 288 independent branches we estimate that the ATRW homolog may have arisen sometime 289 in the last 20 MY ( Figure 6B), a timing that roughly coincides with a major adaptive 290 radiation in the tiger salamander lineage (50, 51). 291 To shed further light on the evolution of ATRX and ATRW within the Ambystoma 292 lineage, we examined patterns of derived substitutions in ATRX and ATRW. Across the 293 251 bp alignment, 9 nucleotide substitutions can be attributed to ATRW since the 294 divergence of axolotl, and these are associated with changes in 2 amino acids. By 295 comparison, ATRX on LG2 shows only 1 nucleotide substitution since the duplication 296 event ( Figure 6). This suggests that ATRW may be evolving at a faster rate than ATRX, 297 in which case 20 MY may represent a substantial overestimate for the origin of the 298 duplication that gave rise to ATRW. 299 300 DISCUSSION 301

Sex chromosome evolution in the axolotl 302
The results from this study show that the homomorphic sex chromosomes of the axolotl 303 contain a small non-recombining region that is specific to the female W chromosome. 304 The female-specific sequence is estimated to be approximately 300Kb, or roughly 305 1/100,000 th of the enormous axolotl genome. It is not surprising that the differences in 306 recombination were not initially evident due to the physical size of the genome and 307 marker density in the Ambystoma meiotic map (29). With respect to the current 308 fragmented female genome assembly, it is still not possible to predict gene orders within 309 this region or locate possible inversions; however, the data are sufficient to identify 310 robust markers for sex and genes that exist in the non-recombining region. Of the few 311 protein-coding genes found within the validated sex-specific scaffolds, two appear to 312 represent non-repetitive coding sequences, including one that represents a relatively 313 recent duplication of the transcriptional regulator ATRX. 314 The ATRX gene is located in the non-recombining region of the X chromosome 315 in mammals. The gene encodes a chromatin remodeling protein that belongs to the 316 SWI/SNF family. It is linked to the rare recessive disorder, alpha-thalassemia X-linked 317 intellectual disability, which is characterized by severe intellectual disability, 318 developmental delays, craniofacial abnormalities, and genital anomalies in humans . In 319 some cases, a mutation in the ATRX gene can lead to female sex reversal due to early 320 testicular failure (52, 53). Gene expression studies performed in a marsupial and 321 eutherian showed that ATRX expression was highly conserved between the two 322 mammals and was necessary for the development of both male and female gonads 323 (48). Because ATRX is one of the few protein-coding genes present in the region of W-324 specific sequence and has been characterized in the sex differentiation of mammals, we 325 propose ATRW as a candidate sex gene for axolotl, or alternately a strong candidate for 326 an acquired, sexually antagonistic gene. 327

Reanalysis of expression data from recent published tissue-specific 328
transcriptomes showed expression of the ATRX gene (from LG2) in all major tissues 329 and developing embryos, however, they showed no evidence of expression of the 330 ATRW gene (54). The tissues represented in the study included whole limb segments, 331 blastemas from regenerating limbs, bone and cartilage, muscle, heart, blood vessel, gill, 332 embryos, testis, and notably, ovaries. It is not clear at what stage the ovarian tissue was 333 taken; however, the author suggests multiple ovaries were sequenced from an adult, 334 and multiple libraries exist for the tissue. It is possible that this sex-specific gene is 335 simply not highly expressed at this specific stage (or in the adult stage, in general) and 336 may only be expressed during early gonadogenesis. Similarly, W-linked genes in 337 chicken were unknown until RNAseq studies were performed prior to and during 338 gonadogenesis (55). 339 If ATRW is the primary sex-determining gene in axolotl, then the origin of this  Ongoing improvements to the Ambystoma genome assembly and development of 356 genome assemblies for other salamander taxa should improve our ability to assess 357 hypotheses related to the presence of homomorphic sex chromosomes (e.g. recent 358 evolution, high-turnover, and fountain of youth) (1, 17, 57-62). Additionally, recent 359 efforts to develop genetic tools for the axolotl model should facilitate functional analyses 360 that will be necessary to test whether ATRW is the primary sex-determining gene in 361 axolotl or elucidate its role as a sexually antagonistic factor (63, 64). Methods for 362 achieving targeted gene knockout and knock-ins have been developed in axolotl and 363 could be adapted to better assess the functionality of ATRW in axolotls (40,65,66). 364 365

Utility of sex-linked markers in axolotl research 366
Sex is an important biological variable in research, as it may contribute to variation in 367 experimental studies. Because axolotl is an important model for many areas of research 368 and has shown sex-specific effects, such as tail regeneration, it is important for 369 investigators to differentiate sex effects from other experimental variables (28). Until 370 now it was necessary to visualize the sex organs, utilize axolotls that had produced 371 gametes, or perform experiments in hybrid crosses that segregate markers at the linked 372 locus E24C3 in order to accurately determine sex in axolotls (29). However, many 373 experiments utilize juvenile animals that may not have completed gonadal differentiation 374 or maturation. With several robust markers for W-specific sequences in hand, it is now 375 possible to precisely differentiate sex of an axolotl with a simple PCR (67). These 376 markers will also positively impact axolotl husbandry, as individuals may be housed and 377 utilized in experiments accordingly. 378

Laser capture microdissection and amplification 381
Preparation of cells for metaphase spreads and laser capture were performed as 382 previously described (30). Briefly, fixed cells were spread on UV-treated 1.0mm 383 polyethylene naphthalate (PEN) membrane slides. Slides were inverted (membrane 384 side down) over a steam bath of distilled water for 7 seconds. Immediately after 385 steaming, 100 µl of the fixed cells were dropped across the middle of the slide 386 lengthwise. Each slide was subsequently placed in a steam chamber at ~35°C for 1 387 minute, then set on the hot plate for 5 minutes. After slides dried, chromosomes were 388 stained via immersion in freshly made Giemsa stain (Sigma-Aldrich GS500-500 ML: 389 0.4% Giemsa, 0.7 g/L KH2PO4, 1.0 g/L Na2HPO4) for 2 minutes, rinsed in 95% ethanol, 390 rinsed in distilled water, then allowed to dry in a desiccator until used. 391 The sex chromosome was captured using a Zeiss PALM Laser Microbeam 392 Microscope at 40X magnification as previously described (30)

Sequence analyses and assembly 410
Because amplified sequences contain a non-complex leader sequence corresponding 411 to the pseudorandom primers that are used for whole chromosome amplification, reads 412 were trimmed prior to further processing. Trimmomatic was used to remove leader 413 sequences derived from phiX and to trim any window of 40 nucleotides with quality 414 score lower than Q30 (68). Reads were then aligned to 945 model transcripts from the 415 Ambystoma linkage map (35) using the Burrows Wheeler Aligner with the single-end 416 mapping option and BWA-MEM algorithm (69). They were also aligned to several 417 bacterial genomes as well as the human reference genome using the paired-end 418 mapping option to identify exact matches for Bowtie 2 (70). Paired reads that mapped 419 concordantly to the human and bacterial genomes were considered potential 420 contaminants and removed. After trimming and removal of potential contaminants, the 421 reads were corrected with Blue (71) using female A. mexicanum whole genome shotgun 422 data (30) and assembled with SOAPdenovo2 (72). 423 To assign scaffolds from the whole genome assembly of a male axolotl genome 424 to the Z chromosome, error-corrected laser capture reads were aligned as paired-end 425 reads to the assembly with BWA-MEM and filtered to preserve only pairs with 426 concordant reads that map to the reference with no mismatches (69). For each scaffold 427 we calculated physical coverage (i.e. coverage by paired-end fragments: bedtools v. 428 2.27, genomeGoverageBed, option pc, (73)) and assigned scaffolds to the Z 429 chromosome if at least 5% of their bases were covered by reads from laser capture 430 sequencing. 431

FISH of sex-associated BAC E24C3 433
Fluorescent in situ hybridization of BACs to metaphase chromosome spreads were 434 performed as previously described (74,75)

. A Qiagen Large Construct kit (Qiagen 435
Science, 12462) was used to extract bacterial artificial chromosome (BAC) DNA for 436 E24C3 and E12A6, previously associated with sex (29). Probes for in situ hybridization 437 were labeled by nick-translation using direct fluorophores Cyanine 3-dUTP (Enzo Life 438 Sciences, ENZ-42501) or Fluorescein-12-dUTP (Thermo Scientific, R0101) as 439 described previously (74) and hybridization of BAC probes was performed as previously 440 described for axolotl chromosomes (40). 441 Phenol-chloroform extraction in 1.2X SSC was used to isolate repetitive DNA 442 fractions from female salamander tissue (76). DNA was denatured for 5 minutes at 443 120°C, re-associated at 60°C for 1 hour to obtain Cot DNA. Microtubes containing the 444 DNA were placed on ice for 2 minutes, then transferred to a bead bath at 42°C for 1 445 hour with 5X S1 nuclease buffer and S1 nuclease for a concentration of 100 units per 1 446 mg DNA. DNA was precipitated with 0.1 volume of 3M sodium acetate and 1 volume 447 isopropanol at room temperature, tubes were inverted several times and centrifuged at 448 14,000 rpm for 20 minutes at 4°C. DNA was washed with 70% ethanol, centrifuged at 449 14,000 rpm for 10 minutes at 4°C, air dried and solubilized in TE buffer. 450 451

Conservation and evolution of salamander chromosomes 452
To evaluate the sex chromosome assembly, we performed alignments between the sex 453 chromosome assembly and reference transcripts (V4: Sal-Site)(32) using megablast 454 (77) to identify genes that occur on the sex chromosome. These genes were then 455 aligned (tblastx) (78) to annotated protein coding genes from the chicken genome 456 assembly (Gallus_gallus-4.0). Annotated genes from scaffolds assigned on the basis of 457 read mapping were aligned (blastp) (78) to this set of annotated chicken genes. Those 458 with an alignment length of at least 50 amino acids and at least 60% identity were 459 considered potential homologs. 460 A similar approach was taken to identify the homologous newt linkage group to 461 assess potential sex candidate genes. Ambystoma reference transcripts from LG9 (V4) 462 were aligned (tblastx) (78) to the chicken genome assembly (41). Using the same 463 minimum thresholds as above, the potential homologs were then used to blast (tblastx) 464 (78) to the newt, Notophthalmus viridescens, reference transcripts (36). 465 466

Identification of female-specific regions 467
We applied depth of coverage analysis to identify single-copy regions in the assembly 468 that have approximately half of the modal coverage in females and 469 underrepresented/absent coverage in males. Reads were generated on an Illumina 470 HiSeq2000 (Hudson Alpha Institute for Biotechnology, Huntsville, Al.) from DNA that 471 was isolated via phenol-chloroform extraction (76) from 48 individuals that were drawn 472 from a previously described backcross mapping panel (42). The resulting reads were 473 aligned to the axolotl draft genome assembly using BWA-MEM (using default 474 parameters) followed by filtering of secondary alignments (samtools view -F2308) and 475 alignments clipped on both sides of the read. Merging of female and male bam files was 476 performed using Samtools merge (69, 79). 477 We used DifCover (https://github.com/timnat/DifCover) (80) to identify candidate shorter than 1Kb and contained fewer than 1000 valid bases (short scaffolds or intervals 493 that fall on the scaffold ends). These shorter intervals were filtered to exclude intervals 494 with fewer than 500 bases or fewer than 200 valid bases. 495 Scaffolds that were validated through PCR in a panel of 6 females and 6 males 496 were aligned to the V4 and V5 Ambystoma transcriptome assemblies in order to identify 497 the genes present on the W-specific portion of the sex chromosome. If a transcript 498 aligned to the scaffold with a percent identity higher than 95%, that transcript was 499 blasted (blastx) (78) to the NCBI nonredundant protein database to search for 500 homologous genes. 501 502

Primer design and PCR 503
Primers were designed within the sex candidate regions identified using Primer3 (81). 504 Each primer was 25-28 bp in length, with a target melting temperature of 60°C, 20-80% 505 GC content and 150-400 bp product sizes depending on the size of the region and 506 location of repeats (avoiding inclusion of repetitive sequence in primer and product). 507 Fragments were amplified using standard PCR conditions (150ng DNA, 50 ng of each 508 primer, 200 mM each dATP, dCTP, dGTP, dTTP; thermal cycling at 94°C for 4 minutes; 509 34 cycles of 94°C for 45 seconds, 55°C for 45 seconds, 72°C for 30 seconds; and 72°C 510 for 7 minutes). Reactions were tested on a panel of six males and six females to 511 validate sex specificity. Gel electrophoresis was performed and presence/absence was 512 recorded for each set of primers (Supplementary Figure 3). The scaffolds from which 513 primers were designed were considered female-specific if the primers yielded specific 514 amplicons in all six females and in no males. 515 Results from these data were used to develop a PCR based assay for 516 determining sex in axolotls at any stage of development. This method uses a primer pair 517 that amplifies a 219 bp DNA fragment in females and an internal control that yields a 518 486 bp DNA fragment in both sexes. This biplex PCR results in two bands (219 bp and 519 486 bp) for females and only the control band (486 bp) in males (67). 520 521

Phylogenetic Reconstruction 522
Homologene was used to collect putative homology groups from the ATRX genes in a 523 variety of eukaryotes (82). Sequence for axolotl ATRX was obtained from Ambystoma 524 reference transcripts, and the newt ATRX gene was obtained by aligning human ATRX 525 to the newt reference transcriptome (83). All sequences were aligned using MEGA7 526 (84) via MUSCLE (85). Sequences were trimmed to compare a conserved subregion of 527 the sequence that was present in all species, a string of 251 codons ( Figure 5). The gray shaded region shows the approximate timing of the ATRW duplication event. 622 The tiger salamander complex consists of 7 named species that occur in the same 623 monophyletic clade as A. californiense, A. mexicanum, and A. tigrinum (56, 91). This 624 tree was generated using Timetree (49) with modification to the position of A. 625 californiense based on Shaffer and McKnight (1996) and Shaffer et al. (2004). Gel electrophoresis of PCR for scaffolds determined to be sex-specific based on 659 computational analyses. Those that show presence in females only are denoted with an 660 asterisk and considered sex-specific. PCRs were tested on six females and six males, 661 and the associated lanes are denoted with ♀ and ♂, respectively. The first and last 662 lanes are labeled with "L" to denote 100 bp ladder. Numerical labels correspond to 663 primer information provided in Supplementary Table 3