Patterns and Mechanisms of Sex Ratio Distortion in the Collaborative Cross Mouse Mapping Population

In species with single-locus, chromosome-based mechanisms of sex determination, the laws of segregation predict an equal ratio of females to males at birth. Here, we show that departures from this Mendelian expectation are commonplace in the 8-way recombinant inbred Collaborative Cross (CC) mouse population. More than one-third of CC strains exhibit significant sex ratio distortion (SRD) at wean, with twice as many male-biased than female-biased strains. We show that these pervasive sex biases persist across multiple breeding environments, are stable over time, are not fully mediated by maternal effects, and are not explained by sex-biased neonatal mortality. SRD exhibits a heritable component, but QTL mapping analyses and targeted investigations of sex determination genes fail to nominate any large effect loci. These findings, combined with the reported absence of sex ratio biases in the CC founder strains, suggest that SRD manifests from multilocus combinations of alleles only uncovered in recombined CC genomes. We speculate that the genetic shuffling of eight diverse parental genomes during the early CC breeding generations led to the decoupling of sex-linked drivers from their co-evolved suppressors, unleashing complex, multiallelic systems of sex chromosome drive. Consistent with this interpretation, we show that several CC strains exhibit copy number imbalances at co-evolved X-and Y-linked ampliconic genes that have been previously implicated in germline genetic conflict and SRD in house mice. Overall, our findings reveal the pervasiveness of SRD in the CC population and nominate the CC as a powerful resource for investigating sex chromosome genetic conflict in action. ARTICLE SUMMARY We compiled breeding records from The Collaborative Cross (CC) mouse mapping population to quantify the frequency and explore potential mechanisms of sex ratio distortion. Strikingly, more than one-third of CC strains yield significantly sex-biased litters. These sex biases are not mediated by environmental effects and are moderately heritable. We conclude that the widespread sex ratio distortion in the CC manifests from multilocus permutations of selfish sex-linked elements and suppressors that are only recovered in the recombinant CC strains.


156
Compilation of Collaborative Cross breeding records 157 Breeding records were obtained for 58 Collaborative Cross strains maintained in The Jackson 158 Laboratory's Repository between January 2016 and July 2019. Eleven strains sired <100 pups 159 surviving to wean age during this time frame and were excluded from further analyses. The data 160 recorded for each live-born litter include: strain name, unique dam and sire identifiers, dam date 161 of birth, sire date of birth, date mating established, litter birth date, litter size at birth, litter size at 162 wean, and the number of weaned males and females. All CC breeding data are provided in 163 Table S1. Strain-level summaries of these breeding data are provided in Table S2. 164 165 Breeding records for CC mice maintained at the University of North Carolina Chapel Hill (UNC) 166 were obtained from the UNC Systems Genetics website 167 (http://csbio.unc.edu/CCstatus/index.py?run=availableLines). These data are also made 168 available in Table S3. 169 170 Breeding records for 43 F1 crosses between distinct CC strains (i.e., CC-RIX crosses) were 171 kindly shared by colleagues at The Jackson Laboratory and are provided in Table S4.  172  173 For brevity, we exclude the laboratory code from CC strain names in figures and tables 174 throughout this manuscript. Note that in all cases of such ambiguity, we are referencing CC 175 lines maintained at the Jackson Laboratory. 176 177 Estimating sex ratios and survival statistics 178 For most analyses, CC strain sex ratios were calculated as the proportion of females at wean. 179 Sex ratios for CC lines maintained at UNC are presented in public data as the ratio of females to 180 males at wean. Comparisons of CC strain sex ratios between the JAX and UNC breeding 181 centers utilize JAX CC strain sex ratios calculated per this alternative definition. 182 183 Neonatal survival was approximated as the fraction of pups born to a given strain that survive to 184 wean. We acknowledge that this estimate is potentially imprecise, as it is often difficult to 185 accurately count pups at birth and pups that were cannibalized shortly after birth are likely 186 missed in these tallies. Litter size at birth was used as a proxy for the in utero survival rate. 187 Although litter size is shaped by a multitude of factors, strains with smaller litters may 188 experience higher rates of embryonic lethality than strains with larger litter sizes, all else being 189 equal. 190 191 To estimate survival during early in utero development and throughout the neonatal period, we 192 devised an ad hoc metric that combined litter size at birth and survival to wean. Specifically, we 193 computed the median litter size at birth and median birth-to-wean survival rate across all CC 194 strains. For a particular focal CC strain, we then computed the difference between the strain-195 specific litter size and the overall CC population-wide median litter size. Similarly, we calculated 196 the difference between the strain-specific survival rate and the overall median survival rate in 197 the CC population. We then summed these two values into a single measure of aggregate 198 strain-specific survival from conception to wean. 199 200 Linear modeling and statistical analyses of temporal changes in SRD 201 The sex of each weaned pup was coded as a binomial indicator and modeled as a function of 202 strain, birth month, and birth year using the glm function in RStudio (v. 1.3.1056). Post-hoc Wald 203 tests were used to determine whether any independent variables provide a significant 204 explanatory effect. R code to recapitulate these findings is available as a supplementary 205 document (cc_srd_analysis.R) on FigShare. 206 207 Analyses of maternal condition and male reproductive phenotypes 208 Maternal body mass and body fat percentage estimates for CC strains were obtained from the 209 McMullan1 and McMullan3 datasets in the Mouse Phenome Database (Table S5; (Bogue et al. 210 2020)). We also accessed male reproductive phenotype datasets for CC (Lazear, Shorter3, 211 Shorter4) and parental inbred strains (Odet1; (Odet et al. 2015)) via the Mouse Phenome 212 Database (Table S6). Spearman Rank correlations and Mann-Whitney U-tests were used to 213 assess relationships between phenotypes and SRD. R code to replicate analyses of maternal 214 condition (maternalCondition.R) and reproductive phenotypes (ReproPhenotypeAnalysis.R) is 215 available on FigShare. 216 217 Broad-sense heritability estimation 218 We treated the sex ratio estimated from weaned pups born to individual CC mating units as 219 independent, within-strain replicate phenotype measures. Mating units producing fewer than 30 220 pups over their breeding history were excluded due to the high uncertainty in calculated sex 221 ratios. We then fit a one-way ANOVA model (sex ratio ~ strain) to estimate the broad-sense 222 heritability (H 2 ) of the sex ratio at wean using the interclass correlation method ( To estimate the copy number of the X and Y-linked ampliconic genes Slx/Slxl1, Sly, Sstx, and 284 Ssty1/2, we first identified the reference coordinates of all annotated paralogs from these genes 285 using the Ensembl Paralogues feature. To discover any additional unannotated paralogs, we 286 blated each annotated paralog sequence against the mm10 reference, retaining full-length hits 287 with >90% sequence identity to the parent sequence. Genomic coordinates for all ampliconic 288 genes are provided in Two independent cross-sections from two of the three biological replicates were then scored for 326 the following testis histology phenotypes: the total number of tubules per cross-section, total 327 cross section area, the number of tubules with vacuoles, the number of tubules with eosinophilic 328 cells (a proxy for cell death), and the cross-sectional area of 50 representative tubules. Testis 329 cross-sections from the third CC032/GeniUncJ replicate exhibited an unusually high fraction 330 (>50%) of vacuolized tubules harboring no post-meiotic cells ( Figure S1).

390
Widespread Sex Ratio Distortion in the Mouse Collaborative Cross 391 We collated breeding records from the Collaborative Cross mouse colony maintained in The 392 Jackson Laboratory's Repository (Table S1). These records summarize the breeding 393 performance of 58 inbred CC strains organized into 3,890 independent mating units that 394 produced 54,034 pups between January 2016 and May 2019 (median = 874 pups per strain). 395 Eleven strains produced fewer than 100 pups during this time period and were excluded from 396 further analysis. Remarkably, 18 of the remaining 47 CC lines (38%) sired progeny with a 397 significant departure from the expected 1:1 sex ratio (uncorrected binomial P < 0.05; Figure 1).

398
Of these 18 strains, 12 are significantly male-biased, whereas six produce an excess of 399 females. This finding aligns with the significant, albeit slight, skew toward males in the 400 aggregate CC population breeding records (21,469 males versus 20,439 females; two-sided 401 binomial P = 4.99x10 -7 ). The most significantly female-biased strain, CC065/UncJ, produces 402 67.4% females (n = 264; Binomial Test P = 1.54x10 -8 ). CC032/GeniUncJ is the most male-403 biased strain, siring 68.6% males at wean (n = 927; P = 2.83x10 -30 ). 404 405 Statistical power to detect a significant departure from the expected Mendelian sex ratio is a 406 function of sample size. Many CC strains produced a modest number of pups over the survey 407 period, limiting our ability to detect weak SRD. Considering only those strains with >500 pups 408 (corresponding to ~60% power to detect a 45%:55% skew in the sex ratio; Figure S2), the 409 percentage of strains with significant SRD increases to 50%. We conclude that mild SRD is 410 pervasive in the CC reference mapping population, with a few strains showing extreme biases in 411 offspring sex ratios. Testing the Stability of SRD to Environmental Influences 428 Seasonal fluctuations in external temperatures can influence offspring sex in captive laboratory 429 house mouse populations (Drickamer 1990). To test for seasonal and larger-term temporal 430 effects on SRD in the CC, we modeled the sex of each weaned pup as a binomial outcome of 431 strain identity, birth month, and birth year. Neither birth month nor birth year provide significant 432 predictive power in this model (birth month, Wald Test P = 0.74; birth year, Wald Test P = 0.17). 433 These findings are recapitulated on a per strain basis: there is no evidence for variation in sex 434 ratio from month-to-month within strains (Fisher's Exact Test, P > 0.05; Table S11).

435
CC028/GeniUncJ and CC084/TauJ show slight variation in sex ratio from year-to-year ( Figure  436 S3; Fisher's Exact Test; P CC028 =0.0365 and P CC084 =0.0500), although these effects are modest 437 and do not remain significant after correcting for multiple testing. intervals. Sex ratios are estimated with high precision for some strain-year combinations, and 443 confidence intervals are masked by the plotting characters. Samples with significant male-and 444 female-sex biases are color coded blue and red, respectively. 445 446 447 To understand whether housing environment influences SRD in the CC, we next assessed the 448 concordance of JAX CC strain sex ratios with those of their sister-strain counterparts maintained 449 at an independent mouse facility at University of North Carolina (UNC) Chapel-Hill. Overall, 450 there is excellent concordance of the strain sex ratios between these two locations (Spearman's 451 Rho = 0.760, P = 1.97x10 -8 ; Figure 2). Notably, CC032 and CC065 are the most strongly male-452 and female-biased strains, respectively, regardless of facility. We next considered the possibility that random, non-genetic maternal effects influence offspring 487 sex in the CC. In the majority of CC strains, offspring sex does not vary from dam-to-dam within 488 a strain (Fisher's Exact Test; P>0.05; Table S12). We do observe slight fluctuations in sex ratio 489 across breeding dams in CC004/TauUncJ (P = 0.013), CC061/GeniUncJ (P = 0.008), and 490 CC068/TauUncJ (P = 0.033), although these effects are not significant after accounting for 491 multiple testing ( Figure S5; 42 tested strains, Bonferroni adjusted P = 0.0012). 492 Figure S5. Dam identity exerts a weak influence on offspring sex ratios in CC004/TauUncJ, 494 CC061/GeniUncJ, and CC068/TauUncJ. For each strain, the sex ratio of animals sired by each 495 dam (or pair of dams, in the case of trio matings) is plotted as the fraction of females at wean. 496 Dams producing significantly female-or male-biased litters are denoted by the red and blue 497 points, respectively. Error bars correspond to 95% confidence intervals calculated from the 498 binomial distribution. 499 500 501 Prior work has uncovered significant effects of parental age and litter parity number on 502 mammalian sex ratios (Huck et al. 1988). Parental age at litter birth, litter size, and the number 503 of litters born to each mating unit vary among CC strains (Table S2), prompting us to explore 504 whether these variables contribute to the observed SRD. We modeled the sex of each weaned 505 pup as a binomial outcome of strain identity and either dam or sire age. Parental age does not 506 offer significant explanatory power in this model (P > 0.05). Similarly, litter size and litter number 507 do not impact estimated sex ratios (P > 0.05). 508 509 Sex ratio distortion is independent of maternal genotype 510 Different maternal genotypes provide distinct uterine environments for early development and 511 could modify sex ratios in crosses involving sires from a common strain. We leveraged breeding 512 data from crosses between distinct CC strains (i.e., CC-RIX crosses) carried out by colleagues 513 at The Jackson Laboratory to address whether maternal genotype influences CC sex ratios. In 514 total, we surveyed data from 43 CC-RIX crosses profiling 24 different CC strains as sires (Table  515 S4). If sex ratio is strictly determined by the paternal transmission of X-versus Y-bearing sperm, 516 then the sex ratios of litters sired by males from different CC strains should be independent of 517 dam genotype. For each of the 24 CC sire strains included in this CC-RIX dataset, we asked 518 whether offspring sex varies as a function of dam strain identity. Although small sample sizes 519 limit our power (range: 13-103 progeny per CC-RIX cross; median = 31), we find no consistent 520 evidence for maternal genotype-dependence of sex ratios in the majority of the CC-RIX strains 521 tested (Fisher's Exact Test P > 0.05; Table S13; Figures S6, S7). Three CC strains are 522 exceptions, with marginal maternal genotype dependence on offspring sex ratios: 523 CC027/GeniUncJ (P = 5.2x10 -4 ), CC042/GeniUncJ (P = 0.038), and CC060/UncJ (P = 0.010) 524 ( Figure S6). Notably, CC027/GeniUncJ males, when mated to CC027/GeniUncJ or 525 CC011/UncJ females, produced sex-balanced litters, but yield female-biased litters in crosses to 526 CC037/TauUncJ dams and male-biased litters in crosses to CC002/UncJ dams ( Figure S6B). 527 Similarly, CC042/GeniUncJ males produce litters with a slight female bias in crosses with either 528 CC042/GeniUncJ or CC001/UncJ dams, and male-biased progeny in crosses to 529 CC005/TauUncJ females ( Figure S6A). However, we caution that only the maternal effects in 530 CC027/GeniUncJ remain significant after correction for multiple tests. Overall, these findings are 531 in broad agreement with the absence of significant maternal genotype effects on sex ratios in 532 the eight parental founder strains of the CC (Shorter et al. 2019a). 533 534 In summary, we find no evidence that season, housing environment, dam identity, parental age, 535 litter number, litter size, maternal condition, or maternal genotype systematically influence sex 536 ratios across the CC strains. Based on these findings, we conclude that the sex ratio of a given 537 CC strain is likely an intrinsic, biological property of that strain, rather than a plastic response to 538 environmental factors or mediated via parental effects. Evaluating sex differences in survival as a potential mechanism of sex ratio distortion 555 SRD can arise along a continuum of developmental timepoints ranging from differences in the 556 viability or fertilization efficiency of X-versus Y-bearing sperm to sex differences in post-birth 557 survival. If SRD stems from sex differences in survival during in utero development, more 558 extremely sex-biased strains should yield smaller litters. In contrast to this expectation, there is 559 no correlation between litter size and the absolute deviation from sex ratio parity in the CC 560 population ( Figure 3A; Spearman's Rho = -0.036, P = 0.819; analysis restricted to breeding 561 pairs only, to the exclusion of breeding trios and harem breeding units). Indeed, several sex-562 biased strains -including CC013/GeniUncJ, CC060/UncJ, CC006/TauUncJ, and CC001/UncJ -563 are among the most fecund CC lines. Similarly, sex differences in survival from birth to wean 564 are not correlated with SRD ( Figure 3B; Spearman's Rho = -0.0817, P = 0.601). 565 566 Mechanisms that contribute to increased rates of strain death in utero may also lead to 567 increased death rates in neonates. Consistent with this possibility, there is a significant positive 568 correlation between litter size and birth-to-wean survival rate; strains with larger litters at birth 569 have lower neonatal death rates (Figure 3C; Spearman's Rho = 0.426, P = 0.005). We 570 combined these two measures of survival during pre-and post-natal development into a single 571 statistic that summarizes strain variation in survival from conception to wean (see Methods). 572 Again, we find no significant correlation between this composite survival statistic and the 573 magnitude of SRD ( Figure 3D; Spearman's Rho = -0.0467, P = 0.766). 574 575 The absence of an overall association between survival in early development and SRD 576 suggests that mortality in early life does not provide a simple, unifying explanation for SRD in 577 the CC. However, it is noteworthy that two of the 18 significantly sex-biased strains are among 578 the 20% of CC strains with the lowest survival rates (CC026/GeniUncJ and CC040/TauUncJ; 579 Figure 3D). Sex differences in early development may contribute to SRD in certain strains, and 580 these findings motivate further work to dissect the developmental mechanisms of potential sex-581 specific mortality in these lines. Nonetheless, survival differences are unlikely to explain SRD in 582 the majority of sex-biased strains. 583 584 Figure 3. Correlations between neonatal survival, litter size, and sex ratio distortion. Sex ratio 585 distortion is not significantly correlated with average litter size (A) or birth to wean survival rate 586 (B). Average litter size at birth and survival to wean are positively correlated (C). The strength of 587 sex ratio distortion is not correlated with aggregated in utero and birth-to-wean survival. Points 588 corresponding to strains with significant female-and male-bias are color-coded red and blue, 589 respectively. 590 591 592 No Evidence for Single Locus Mechanisms of Sex Ratio Distortion 593 Our extensive analyses of possible non-genetic explanations for sex ratio variation in the CC 594 turned up no compelling explanations, suggesting that sex ratio variation likely carries a genetic 595 basis. We utilized sex ratio estimates from independent breeding units within each CC strain as 596 biological replicates to compute the relative proportion of variation in the sex ratio that is due to 597 within versus between strain differences (i.e., broad sense heritability, H 2 ; see Methods). 598 Despite considerable binomial noise in sex ratio estimates per breeding unit, the sex ratio is 599 modestly heritable in the CC (H 2 = 0.263). 600 601 To attempt to map genomic loci contributing to this heritable variation in SRD, we carried out a 602 genome-wide QTL scan in the CC. No autosomal or X-linked loci reached the genome-wide 603 threshold for significance ( Figure 4A). Similarly, we find no effect of the Y chromosome or 604 mitochondrial haplotype on SRD (one-way ANOVA, P > 0.05; Figures 4B and 4C) can lead to constitutively high (low) levels of Sox9 expression, providing a second mechanism 654 for sex reversal in mammals (Foster et al. 1994;Gonen et al. 2018). Interestingly, while we find 655 no evidence for duplication of Sox9 itself, we observe a ~1kb duplication and a ~2kb deletion 656 within the distal upstream Sox9 regulatory region that are specific to animals carrying the 657 NOD/ShiLtJ haplotype in these regions ( Figure S9). These structural mutations do not span any 658 annotated regulatory elements in the mm10 reference genome, but it is tempting to speculate 659 that one or both may function to maintain native Sox9 expression levels in the face of potentially 660 increased SRY dosage driven by the Sry-duplication present in this strain. 661 662 It is unlikely that the NOD/ShiLtJ-specific SVs documented here are associated with an 663 appreciable rate of sex reversal. NOD/ShiLtJ has served as a prominent mouse model of 664 autoimmune disorders for more than 40 years, with no cases of sex reversal documented in this 665 strain or in crosses involving this strain (including the CC), to our knowledge. In addition, Sry 666 duplications are relatively common in rodent systems, and are not generally associated with sex 667 reversal (Nagamine 1994;Lundrigan and Tucker 1997;Bullejos et al. 1999). Finally, half of the 668 sex-biased CC strains do not carry the NOD/ShiLtJ haplotype at either Sox9 or Sry (Table S14), 669 necessarily assigning causality of SRD to other mechanisms. In summary, although we find 670 novel structural rearrangements spanning the Sry sex determination gene and within the 671 putative regulatory region of its upstream signaling target, Sox9 (Figures S8 and S9), these 672 mutations seem unlikely, at face-value, to induce high rates of sex reversal and contribute to the 673 widespread SRD in the CC population. To address this possibility, we used publicly available whole genome sequences to estimate the 703 relative genomic copy number of these ampliconic genes in each realized CC strain ( fold across strains (range: 24-64 copies), with only 5 strains exhibiting estimated haploid copy 706 number states that fall outside the range delimited by the parental genomes (parental range: 28-707 60 haploid copies; Figure S10A). Across the CC population, Sly copy number ranges from 54-708 168. CC founder whole genome sequences were generated from female samples, barring 709 comparisons of Sly copy number status in the parental inbred and realized CC strains. Sstx copy number state of the inbred CC founder strains and the solid black line is an overall 717 trend line fit to the data using the method of least squares. Strains with significantly male-and 718 female-biased sex ratios are color-coded blue and red, respectively. 719 720 Overall, we find no correlation between the fraction of females at wean and Slx/Slxl1:Sly copy 721 number ratio (Spearman's Rho = 0.0968, P = 0.490; Figure 5A). We validated the genomic 722 read-depth Slx/Slxl1:Sly estimated copy number (CN) ratios using ddPCR assays in a 723 representative subset of CC strains (Table S16). There is excellent qualitative alignment 724 between these two orthogonal methods for copy number estimation (Figure 5B; Spearman's 725 Rho = 0.943, P = 0.0167), suggesting that the absence of a relationship between the 726 Slx/Slxl1:Sly ratio and SRD is unlikely due to technical errors. Despite the lack of an overall association between the Slx/Slxl1:Sly copy number ratio and sex 738 ratio distortion across CC strains, many sex biased strains do exhibit extreme amplicon copy 739 number ratios. In particular, CC006/TauUncJ has a relative excess of Sly copies relative to 740 Slx/Slxl1, consistent with the male bias in this strain. CC065/UncJ and CC058/UncJ, two 741 female-biased strains, exhibit a relative excess of Slx/Slxl1 compared to Sly, consistent with the 742 over-transmission of the X chromosome. However, there are clear exceptions to expected 743 trends. CC013/GeniUncJ has a moderately low Slx/Slxl1:Sly ratio, yet this strain is female-744 biased. CC002/UncJ and CC028/GeniUncJ have high Slx/Slxl1:Sly ratios, in contrast to 745 predictions given the male sex bias observed in these strains ( Figure 5A). whereas the mouse X-chromosome harbors several Spin1 gene clusters. The antagonistic 755 interactions of Ssty/Spin1 with Slx/Slxl1 and Sly prompted us to explore whether copy number 756 status at spindlin genes may factor into the complexity of SRD in the CC. 757 758 Spin1 and Ssty1/2 copy numbers span a 4.6-and 3.14-fold range in the CC population, 759 respectively (Spin1 range: 50-232 copies; Ssty1/2 range: 223-698 copies; Figure S10B). We 760 observe no significant relationship between the combined Spin1 and Ssty1/2 copy number and 761 sex ratio (Figure 5C; Spearman's Rho = -0.178, P = 0.203). Based on the known interactions 762 between spindlins and members of the Sycp3-like family, we reasoned that the ratios of 763 Slx/Slxl1 to spindlin CN and Sly to spindlin CN might be correlated with the degree of SRD. 764 These predictions are not upheld (Spearman rank correlation P > 0.05; Figure S11).

766
Overall, our findings uncover no global relationship between the copy number state of genes in 767 the Sycp3-like and spindlin gene families with SRD in the CC. However, the copy number state 768 of several CC lines accords with expectations under current models of SLX/SLXL1-SLY genetic 769 conflict, and we speculate that this established drive system may contribute to SRD in an 770 appreciable number of CC strains. Further, our genomic estimates of copy number for these 771 ampliconic genes may not accurately estimate the number of transcriptionally active genes in 772 each family. It is possible that a more widespread relationship between the copy number state 773 of these ampliconic genes and SRD is concealed by the inclusion of large numbers of non-774 expressed pseudogenes in our copy number tallies. Lastly, we cannot rule out 2019), a trend that may emerge from the differential death, motility, or fertilization capacity of 802 sperm bearing one sex chromosome relative to the other. It is widely acknowledged that many 803 CC lines are poor breeders, and it appears that in most cases, reproductive output is 804 constrained by male fertility (Shorter et al. 2017). While there is no significant relationship 805 between average litter size and SRD among CC strains (Figure 3A), we sought to assess the 806 relationship between SRD and more precise measures of male fertility. 807 808 We accessed publicly available reproductive phenotype datasets for several CC strains from the 809 Mouse Phenome Database (Bogue et al. 2018; Table S6). Although the limited number of 810 phenotyped CC strains effectively bars a rigorous statistical analysis, many sex-biased strains 811 do appear to have reduced fertility relative to sex-balanced strains. Sex-biased CC strains tend 812 to have lower testis weights than non-sex-biased CC strains, although this association is not 813 statistically significant (Figure 6A; Spearman's Rho = -0.382, P = 0.248). Similarly, despite the 814 lack of significant population-wide statistical correlations, several sex-biased strains -including 815 CC028/GeniUncJ, CC032/GeniUncJ, and CC040/ TauUncJexhibit low fractions of motile  816 sperm and low sperm density compared to strains that yield sex-balanced litters (Figure 6B,C). 817 We conclude that many sex-biased strains exhibit phenotypic signatures of reduced male 818 fertility.  Relative to other CC lines, CC032 males have low average testis weights (Figure 6A), low 838 sperm density (Figure 6B), and reduced motility ( Figure 6C). Histological analysis of testis 839 cross-sections reveal that CC032 males also have smaller seminiferous tubules than the 840 majority of the CC founder strains ( Figure 7A) and a higher fraction of tubules with vacuoles, 841 (Figures 7B-D). 842 843 Based on these phenotypic findings, we reasoned that targeted killing of X-bearing germ cells 844 could be a plausible explanation for the observed male bias in CC032. We used whole 845 chromosome painting to assess sex chromosome representation in mature sperm from this 846 strain. We observe equal numbers of X-and Y-bearing sperm (49.7% chrX-bearing sperm; 847 Binomial P = 0.854; Figures 7E and 7F)  Slx/Slxl1 copy number). As a consequence of this sex chromosome haplotype structure, the 863 ratio of Slx/Slxl1 to Sly in CC032 is lower than the CC-wide average ( Figure 5A) We scanned the genome of CC032 for potential structural mutations in other sex-linked 870 ampliconic genes that could, conceivably, amplify the strength of Slx/Slxl1:Sly-mediated SRD. 871 Strikingly, the CC032 genome shows a pronounced enrichment of reads mapping to chrXA.1 872 (chrX:3-6Mb; Figure S12) and chrXqA3 (chrX:30.5-35.5 Mb; Figure 8). These loci harbor 873 clusters of Spin1 and Spin2, both members of the spindlin gene family, as well as a large 874 number of genes in the Btbd35f family. These regions encompass several gaps on the mm10 875 mouse reference assembly and are present at variable copy number across the 8 CC founder 876 strains. However, the read depth profile of CC032 at these loci exceeds what is observed in any 877 of the 8 founder strains (Figure 8; Figure S12). Several other CC lines exhibit similar read 878 depth patterns across these two X-linked loci, but intriguingly, the strain haplotype origin of 879 these amplified regions is variable (Figure 8). CC032 harbors C57BL/6J ancestry across both 880 regions, but strains with 129S1/SvImJ, NOD/ShiLtJ, and PWK/PhJ-derived haplotypes exhibit 881 near identical read-depth signatures (Figure 8; Figure S12). This observation would seem to 882 rule out a single common founder effect and imply an incredible rate of structural instability at 883 these loci. However, due to the complex, ampliconic architecture of these loci, we cannot 884 definitively rule out the possibility that assembly, mapping, or genotyping errors have led to mis-885 assignment of strain haplotypes in these regions. an excess of males at wean (Figure 9a). Similarly, males from 12 DO lineages sire sex-biased 931 litters, with all but three producing an excess of males (Figure 9b). For both DO males and 932 females, the observed number of sex-biased families exceeds the ~9 families expected by 933 chance. Sample sizes within each DO lineage are modest (n=67-286; mean = 193), and we are 934 underpowered to detect slight departures from the expected sex ratio. Nonetheless, our findings 935 suggest that the 8-way genotypes associated with DO and CC mice are frequently associated 936 with SRD in both heterozygous and inbred states. 937 938 Despite an excess of both female and male DO lineages with significant SRD, there is no 939 correlation between the sex ratios of progeny sired by dams and sires from a given lineage 940 (Spearman's Rho = -0.12; P = 0.1094; Figure S13). Indeed, there are no cases where DO 941 males and DO females from the same lineage both have sex-biased litters ( Figure S13).