2.9 Demographic history analysis
The golden takin (TX-11, ~42X), Milu (Elaphurus davidianus , SRR5762659- SRR5762666, ~50X) (Chen et al. 2019) and Qinling snub-nosed monkey (SRR2017686, ~15X) (Yu et al. 2016) were selected for inferring the effective population size trajectory using the Pairwise Sequentially Markovian Coalescence (PSMC) model (v0.6.5-r67) (Li and Durbin 2011). The neutral mutation rate (μ) was estimated using r8s (v1.81) (Sanderson 2003). The remaining steps were performed as previously described (Yin et al. 2021).
Results
Chromosome-level de novo genome assembly of golden takin
To estimate the genome size of B. taxicolor , 124.37 Gb clean reads were used for k-mer analysis. The genome size was estimated at 2.79 Gb (k=17, Figure 1b ). We constructed a chromosome-level reference genome for B. taxicolor by PacBio HiFi combined with Hi-C sequencing technology (Figure 1c ). A total of 68.56 Gb HiFi reads (~23×) were generated by PacBio Sequel II platform (Table S1 ). We assembled 4671 contigs using hifasm software. The de novoassembled reference genome was about 2.95 Gb with a contig N50 of 68.05 Mb (Table 1 ), which had a longer contig N50 as compared to goat, sheep and cattle reference genomes (Table S2 ). Further, 344.7 Gb Hi-C reads (~117×) spanned to 4,512 scaffolds with a N50 of 101.27 Mb (Table 1 ), which anchored onto 25+XY chromosomes (Figure 1d ). The genome of takin had a good collinearity relationship with that of the goat (Saanen_v1) (Figure S1 ) and sheep (ARS-UI_Ramb_v2.0) (Figure S2 ). Based on the assembled genome, we evaluated its completeness with mammalia_odb9 database using BUSCO program. The completeness of assembled genome reached 94.2%, of which a single-copy reached 93.1% (Table S3 ).
Genome annotation
A total of 49.48% of the assembled genome was repeat sequences. Long interspersed nuclear elements (LINEs) accounted for 25.08%, which was the highest among all types of repeats, followed by short interspersed nuclear elements (SINEs, 9.92%), satellites (6.97%), long terminal repeats (LTRs, 4.33%), and DNA elements (1.97%) (Table S4 ). In addition, 554,037 simple sequence repeats (SSRs, 0.82%) were identified (Table S4 ). The dimers were the largest proportion (29.70%) of SSRs, followed by quadmers (14.72%), monomers (13.75%), pentamers (9.73%), hexamers (9.20%) and trimers (8.90%) (Table S5 ). 21,301 protein-coding genes were predicted by the homology-based method combined with RNA-seq data. The average gene length was 26,238 bp with 8.82 exons on average. 99.15% (20,282) and 87.69% (18,679) were functionally annotated in the Nr and InterPro databases, respectively.
Phylogenetic analysis at the mitochondrial and genomic level
The assembled mitochondrial genome was about 16,584 bp (GenBank accession No. OM237313), which shared 98.10%, 99.05% and 99.69% sequence similarity with Mishmi (GenBank accession No. NC_043930.1), Sichuan (GenBank accession No. NC_039686.1) and golden takin (GenBank accession No. KY399869.1 and KU361169.1), respectively. At the mitochondrial genome level, takin was closely related to goat branch as compared to muskox (Figure 2a ), which is consistent with previous studies (Feng et al. 2016, Kumar et al. 2019, Zhou et al. 2019).
To reveal the evolution of takin genome, we constructed the phylogenetic tree with representative subfamily Caprinae species. Interestingly, takin was most closely related to muskox (Figure 2b and Figure S3 ). Molecular dating using PAML MCMCtree showed that takin had split from the ancestor of sheep and goat at ~10.06 million years ago (Mya), while takin showed a split from the ancestor of muskox at ~8.26 Mya (Figure 2b ).
Chromosome evolution analysis
Using the sperm whale as outgroup, we reconstructed the ancestral karyotype (2n=60, Table S6 ) of the cattle, goat, sheep and takin. 559 conserved segments were identified and 30 ancestral chromosomes with a total length of ~2.60 Gb were predicted. Compared to the ancestral karyotype of bovid (2n=60), the karyotype of cattle (2n=60) was most conserved. The karyotype of goat (2n=60) occured only one translocation, however the karyotype of sheep (2n=54) occured three fusions and one translocation. The karyotype of takin decreased to 2n=52 after occurring four fusions and one translocation (Figure 3a ). Further, we constructed the collinear relationship among takin, sheep and cattle. There were three chromosome (Chr) fusions from cattle to sheep including Chr 1 and 3 to Chr 1, Chr 2 and 8 to Chr 2, and Chr 5 and 11 to Chr 3. In contrast, there were four fusions from cattle to takin including Chr 1 and 22 to Chr 1, Chr 2 and 25 to Chr 2, Chr 5 and 28 to Chr 5, and Chr 11 and 23 to Chr 11. Interestingly, the translocation occurred from bovine Chr 9 to takin’s Chr 14 and sheep’s Chr 9 (Figure 3b ).
Genetic diversity and population history of golden takins
Nine additional golden takins were carried out by genome resequencing, 442.36 Gb clean reads were generated (Table S7 ). In total of 3,328,260 SNPs were identified by GATK software. To examine the extent of inbreeding of golden takins, we estimated the θπ, heterozygosity and identified the runs of homozygosity (ROH). Golden takins displayed much lower single nucleotide variants (SNVs) (θπ=0.00028) as compared to Qinling giant pandas (θπ=0.00133). The average heterozygosity of nine golden takin genomes was about 0.00038, which was lower than domestic and most protected wild animals including Qinling giant pandas and snub-nosed monkeys, while was a little higher than that of snow leopards and cheetahs (Figure 4a ).
The heterozygosity of golden takin genomes and proportion of the genome in ROH were shown as Figure 4b. Based on the length of ROH > 1 Mb (eliminating the effects of linkage disequilibrium (LD)), the average of genomic inbreeding coefficient (FROH) was found to be 0.217 (Table S8 ). The number of short ROH (2.5-4.17 Mb) was more than 20, while the number of long ROH (6.25-10 Mb) was at least one in all of the nine golden takins (Figure 4c ). The result indicated that inbreeding occurred from 5th to 20th generations.
To investigate the demographic history of golden takins, the PMSC was performed. Effective population size dropped significantly from ~1 million years ago (the Xixiabangma Glaciation) for golden takins, Milu and snub-nosed monkeys. The golden takins declined more significantly than Milu and snub-nosed monkeys from ~100,000 years ago (Figure 4d ).
Discussion