Genome-wide copy number variation drives adaptive evolution in Macaca mulatta populations

Abstract

Copy number variation (CNV) represents a significant contributor to intraspecific genomic diversity, yet remains under-characterized in rhesus macaque (Macaca mulatta) populations. In this study, we performed whole-genome sequencing of 8 M. m. lasiotis individuals and integrated these with 6 subspecies-representative genomes, yielding a comparative dataset of 60 rhesus macaque genomes. Utilizing a read-depth-based approach for CNV detection, our analysis identified ∼59.6 Mb (2.2% of the reference genome) exhibiting CNV. We established a comprehensive CNV map consisting of 1,272 nonoverlapping CNV regions (CNVRs) across the M. mulatta genome. Hierarchical clustering analysis revealed discordance between CNVR-derived population structure and SNV-based phylogenetic relationships among Chinese subspecies. Comparative genomic analysis identified highly differentiated CNVRs between Indian and Chinese populations enriched for genes associated with immune response and xenobiotic metabolism. Subspecies-specific CNVRs (su_CNVRs) showed functional enrichment in pathways related to environmental adaptation and phenotypic variation. Notably, M. m. tcheliensis exhibited higher gene copies in METRNL and FBP1/FBP2 genes than other subspecies, potentially helping it adapt to cold environments and enhance gluconeogenesis capacity under food scarcity in the cold season. Furthermore, we identified 6 body size-associated genes with considerable CNV among different subspecies, highlighting CNV-mediated regulation of body size in M. mulatta subspecies. This investigation not only elucidates the evolutionary significance of CNVs in shaping genomic architecture and adaptive traits in M. mulatta, but also establishes a foundational resource for future population genetics studies utilizing structural variation markers.

Graphical Abstract

Graphical Abstract.

Genomes vary from one another in multifarious ways, and this comprehensive genetic variation underpins the heritability of animal phenotype and their adaptations to the environment. Among various genomic variations, copy number variations (CNVs) are the dominant class of genomic structure variation referring to the duplications or deletions of DNA sequences with length >50 bp and even within a few Mb. Genome sequencing studies have demonstrated that CNVs account for the majority of variable nucleotide positions across individual genomes (Conrad et al. 2010). CNVs located in or adjacent to gene regions can affect gene expression through gene dose effect, location effect, changing gene structure or other mechanisms (Chiang et al. 2017). Some CNVs not only regulate individual genes directly but may also reshape the functional architecture of the genome by driving the expansion or contraction of gene families (Thornton and DeSalle 2000; Yilmaz et al. 2021). CNVs can further lead to phenotypic, physiological, and adaptive differences, which have been functionally validated in various systems (Henkel et al. 2019; Guo et al. 2021; Huang et al. 2021; Yilmaz et al. 2021; Pos et al. 2021; Shi et al. 2023). For instance, increasing copy number of SPATA31 enables humans to resist UV damage and adapt to a diurnal lifestyle (Bekpen et al. 2017). Fluorescence in situ hybridization (FISH) experiments have demonstrated that CNV of the AMY1 gene is directly associated with dietary adaptation in humans (Perry et al. 2007). In cattle, CNVs have been shown to arise independently across breeds and contribute to breed-specific traits, as validated by quantitative fluorescent polymerase chain reaction (qPCR) and FISH analyses (Liu et al. 2010).

Rhesus macaque (Macaca mulatta) is one of the most widely distributed nonhuman primates in Old World Monkeys and also a widely used model animal in biomedical research. It is native to forest but also found coexisting with humans in more than 11 countries of South or Southeast Asia. The species first diverged into Indian macaque (M. m. mulatta) and Chinese macaque 162,000 years ago (Hernandez et al. 2007). The spread of Chinese macaque thereafter formed 6 subspecies: M. m. siamica, M. m. brevicaudus, M. m. littoralis, M. m. lasiotis, M. m. vestita, and M. m. tcheliensis (Jiang et al. 1991). Different subspecies exhibit high levels of phenotypic diversity on the morphology, physiology and genetics, and vary in the adaptive characteristics making them an excellent model for evolutionary adaption research. Chinese macaques have larger body size than Indian macaques (Clarke and O’Neil 1999), and subspecies inhabiting high altitude, such as M. m. tcheliensis (about 8.8 kg in males) and M. m. lasiotis (about 8.0 kg in males) are heavier than subspecies inhabiting low altitude, such as M. m. brevicaudus (about 6.3 kg) and M. m. siamica (about 6.7 kg) (Zhang et al. 2016). When infected by simian immunodeficiency virus (SIV), Chinese macaques had more CD4 positive cell and higher viral load than Indian macaques, thus can better simulate the progression of human immunodeficiency virus (HIV) infection (Liao et al. 2012). Despite many studies on this species, the genetic mechanism underlying the high phenotypic diversity and environmental adaptation among subspecies of rhesus macaque, especially among different Chinese subspecies, has not been illustrated.

Most studies of macaque genetic variations have focused on single-nucleotide variants (SNVs) (Hernandez et al. 2007; Xue et al. 2016; Zhong et al. 2016; Bimber et al. 2017; Liu et al. 2018). However, CNVs among rhesus macaque populations, as well as the relevance of this diversity to phenotypic difference and their adaptation to environment, remains understudied. Lee et al. analyzed CNVs based on genome of 10 Indian macaque individuals using array-based comparative genomic hybridization (aCGH) method, only 123 CNVs loci were detected (Lee et al. 2008). With similar method, Gschwind et al. (2017) detected CNVs in the genome of 21 crab-eating macaque (M. fascicularis), generating 2,951 CNVs on average per individual. While using genome-wide human SNP array, Ng et al. (2015) identified CNVs between human and India/Chinese rhesus macaque, and found CNV regions between each of the 2 rhesus macaques and human were different, indicating differentiation of CNVs between the 2 macaques. With the development of genome sequencing technique, a read-depth method was applied to identify CNVs in the genome. For instance, Li et al. investigated interspecific CNVs based on genome of 9 macaque species (Li et al. 2020), observing that CNVs intersected with genes related to nutrient metabolism, xenobiotic/drug metabolism and immune function. Using genome sequencing of 32 rhesus macaques, Thomas et al. (2021) uncovered genomic patterns of CNVs with an average of 3,646 CNVs per individual, 88% of the CNVs were deletion. Brasó-Vives et al. (2020) constructed a genome-wide CNV map for rhesus macaque from genome of 198 individuals, demonstrating that about one-eighth of the reference genome is composed of copy number variable regions or fixed duplications. However, genome-wide CNVs of rhesus macaque, especially at the population scale from different subspecies, remain uncharacterized. To reveal the genome-wide CNVs in rhesus macaque populations, we sequenced genome of 8 M. m. lasiotis individuals and combined public available genome data representing 6 subspecies of rhesus macaque, yielding a total of 60 genomes for analysis. First, we identified CNV regions (CNVRs) shared by multiple individuals and subspecies (Mm_CNVRs) to construct a genome-wide CNV map in rhesus macaque. Secondly, we performed pairwise comparisons using 3 approaches: the Vst, the Kruskal–Wallis test, and the Mann–Whitney U test to identify highly differentiated CNVRs among different subspecies. Thirdly, we also identified CNVRs that were specific to each subspecies (su_CNVRs). Our goal was to characterize the patterns and diversity of CNVs among rhesus macaque subspecies and to identify candidate CNVs potentially associated with complex phenotypes or environmental adaptations. This study also will facilitate the selection of suitable rhesus macaque subspecies for disease model, thus expand their applications in biomedical research in future.

Materials and methods

Animal samples and sequencing

In 2021, blood samples of 8 individuals of M. m. lasiotuis (Supplementary Table S1) were collected using sterile venipuncture into EDTA anticoagulant tubes from Sichuan Greenhouse Biotechnology Co., Ltd. (Meishan, Sichuan). According to the record, all the individuals were obtained from wild populations in Mabian County, Sichuan Province, China. DNA was isolated with a commercially available DNeasy Blood and Tissue Kit (Qiagen, Germany), then was sequenced using Illumina HiSeq 2 500 technology with a 150 bp paired-end strategy (Beijing Berry Hekang Biotechnology Co., Ltd.). In addition, blood samples of M. m. littoralis, M. m. siamica and M. m. brevicaudus (Supplementary Table S2) were provided by Dr Ming Li (Institute of zoology, Chinese academy of science), while muscle samples of M. m. tcheliensis (Supplementary Table S2) were from Dr Jundong Tian (Zhengzhou University, Henan Province). These samples were used for qPCR validation of copy numbers of selected genes.

Sequence alignment to reference genome

We downloaded 52 genome data of rhesus macaque from NCBI Sequence Read Archive (SRA; https://www.ncbi.nlm.nih.gov/sra) (Liu et al. 2018). They were from 6 subspecies of rhesus macaque, including 10 individuals of M. m. mulatta, 10 M. m. siamica, 10 M. m. lasiotis, 5 M. m. tcheliensis, 5 M. m. brevicaudus, and 12 M. m. littoralis (Supplementary Table S3). Combining genome sequences of M. m. lasiotis generated by this study, we analyzed a total of 60 genome sequences representing 6 subspecies of M. mulatta. The genome sequences were examined using the FastQC software (http://www.bioinformatics.babraham.ac.uk/projects/fastqc). Then, the adapters of sequencing reads and low-quality reads were removed by Trimmomatic-0.36 (Bolger et al. 2014), with setting the parameters as MINLEN = 30, LEADING = 3, TRAILING = 3, and SLIDINGWINDOW = 4:15. The clean reads were aligned to the reference genome (Mmul_10, GCF_003339765.1) by BWA, using the following alignment parameters: mem-t 4-k 32–M (Li and Durbin 2009). Finally, PCR duplicates that could affect the CNVs analysis were removed by using Picard-2.9.1 (https://broadinstitute.github.io/picard/) with Mark-duplicates.

CNVs calling and CNVRs detection

We used 10 rhesus macaque genomes with relatively high sequencing coverage, selected from different subspecies, to evaluate the performance of 2 CNV detection strategies: CNVkit and CNVnator. Various detection parameters were tested for each method, and their performance was comprehensively assessed based on the number of overlapping CNVs, average CNV length, and the false positive rate (calculated as the number of shared CNVs between the 2 tools divided by the total number of CNVs detected by each tool). Ultimately, we selected CNVnator v0.4 for our analysis (Abyzov et al. 2011). As suggested by Abyzov et al. (2011), the sequencing coverage should be inversely proportional to the optimal bin size. They also empirically observed that the ratio of the average read-depth (RD) signal to its standard deviation at the optimal bin size is typically around 4 to 5. Therefore, to avoid the bias caused by sequencing depth of different individuals, for genomes with sequencing coverage 10×–20×, we used a 300 bp window for CNVs calling, while for genome data with high coverage, we used a 100 bp window for CNVs calling. The CNVs calls were filtered with P-value <0.01, q0 < 0.5, size >1 kb, and RD >1.4 or RD <0.6. The CNVs screened under this condition were used as the final identified CNVs collection for each individual.

Copy number variation regions (CNVRs) of different subspecies were obtained using “CNV_overlap.py” script on GitHub (https://github.com/bjtrost/TCAG-WGS-CNV-workflow) (Trost et al. 2018). CNVs shared by 3 or more individuals of 1 subspecies were considered to be CNVR of the subspecies (Pierce et al. 2018; Seol et al. 2019). We merged all CNVRs of each subspecies using “merge” option of BEDtools (Quinlan and Hall 2010). CNVR shared by 2 or more subspecies was defined as M. mulatta CNVRs (Mm_CNVRs). CNVR only found in 1 subspecies was defined as subspecies specific CNVRs (su_CNVRs) (Supplementary Figure S1).

To verify the effectiveness of the identified Mm_CNVRs, a dataset of structural variants (SVs) generated from genome of different subspecies of Chinese rhesus macaques, which were sequenced by long-reads platform of the PacBio CLR, were utilized for comparison. SVs were called using Sniffles (v1.0.12) (Sedlazeck et al. 2018), resulting in a nonredundant SVs dataset (unpublished data). We examined the overlapping CNVs between Mm_CNVRs and SV dataset, using BEDtools (Quinlan and Hall 2010) to extract the intersection, with an overlap threshold of ≥1 kb.

CNVRs annotation and functional enrichment

Genes on the autosomes of Mmul_10 were downloaded from NCBI. CNVRs annotation analysis was performed using “window” option of BEDtools (Quinlan and Hall 2010) to find genes intersecting with CNVRs and 1 kb flanking sequences. We retained CNVRs intersected with protein-coding genes for subsequent analysis. At the same time, we used the R package RIdeogram (Hao et al. 2020) to draw a CNVs map of rhesus macaque genome based on the site and annotation information of Mm_CNVRs.

The DAVID database (https://david.ncifcrf.gov/) was used for the gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) functional enrichment analysis (Huang et al. 2009), and the significance threshold was set to Fisher exact P-value <0.05.

Differential CNVRs among subspecies

We applied 3 methods to investigate differential CNVRs among subspecies, including the clustering analysis, the V_ST statistic and the Kruskal–Wallis (K–W) test (Kruskal and Wallis 1952). The clustering analysis was performed using R package pvclust (Suzuki and Shimodaira 2006) based on all CNVRs (including Mm_CNVRs and su_CNVRs). We assigned either 1 or 0 to each individual according to the presence or absence of 1 CNVR. Approximately unbiased (AU) and bootstrap probability (BP) values were obtained based on the unweighted pair-group average (UPGMA) of 10,000 multiscale bootstraps resamples to indicate the significance of clustering (Suzuki and Shimodaira 2006). Next, we used Vst statistic to detect CNVRs with high differentiation between subspecies. For CNVRs annotated to intersect protein-coding genes, we used the “-genotype” option of CNVnator to calculate the copy number of CNVRs in each individual. The Vst value was calculated as (Vt − Vs)/Vt, where Vt is estimated among all individuals between 2 populations and Vs is the average variance within each population (Redon et al. 2006). Finally, K–W test was performed to detect CNVRs intersecting with protein-coding genes that were significantly different in copy number between subspecies. The significance level was set as P-value <0.01. To examine which subspecies pair was significant differentiation, the Mann–Whitney U test (Mann and Whitney 1947) was further used for CNVRs that showed significant differences in the K–W test with P-value <0.01. Highly differentiated CNVRs shared by all subspecies, but its CNV type or copy number change amplitude is different among subspecies. We then performed enrichment analysis focusing on terms potentially related to environmental adaptation in rhesus macaques. The genes associated with these terms were considered candidate genes with adaptive significance.

Validation of CNVs with qPCR

To validate highly differential CNVR genes identified by our computational analysis, the real-time qPCR was performed to examine copy numbers of 4 CNVR genes (ARID1B, FBP1, IRS2, and NOG) and a single-copy gene HBB based on DNA from different subspecies of Chinese macaque (Supplementary Table S2). The 4 CNVR genes varied on copy numbers among subspecies, potentially associated with environmental adaptation or phenotype variation. We used NCBI Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/) to online design primers. And all the primers are available in Supplementary Table S4.

DNA concentrations were examined using a microspectrophotometer (Onedrop 1000+) to keep similar DNA concentrations in all samples. Using a M. m. tcheliensis individual as control, the relative quantitative △△CT method was used to calculate the copy number of the target product in each experimental sample, and the copy number was compared with that obtained by sequencing data analysis. The copy number calculation formula of qPCR is as follows: CN = 2 × 2^−ΔΔCT (Li et al. 2020). We used TransStart Green qPCR SuperMix and a LightCycler 480 II instrument for qPCR, and the system was set as: 0.2 μL template DNA, 0.4 μL primer, 5 μL 2× Taq Pro Universal SYBR qPCR Master Mix (Novocam Biotechnology Co., LTD.) and 4.4 μL ddH₂O. The qPCR reaction conditions were performed according to the instructions for 2× Taq Pro Universal SYBR qPCR Master Mix. Thermal cycling conditions consisted of 1 cycle of pre-denaturation 3 min at 95 °C; denaturation by 40 cycles of 5 s at 95 °C, annealing 30 s at 60 °C; the solution curve parameters are 15 s at 95 °C, 60 s at 60 °C, and 15 s at 95 °C. All qPCR experiments were triplicate.

Results

Resequencing and CNVs detection

The whole-genome sequencing of 8 M. m. lasiotis individuals generated an average of 66.99 Gb of raw data per sample, with an average sequencing depth of 22.55×. After quality control, the clean data per sample averaged 66.16 Gb, with a mean alignment rate of 99.7% to the Mmul_10 reference genome (Supplementary Table S5). The downloaded macaque genome data was conducted the same process and showed a coverage ranging from 6.9× to 44× (Supplementary Table S3). In combination, we jointly analyzed 60 macaque genomes from various regions of China (Figure 1) and India for CNV calling. We compared 2 CNV calling strategies, CNVkit and CNVnator, using genome data from 10 individuals representing different subspecies. Our results showed that CNVnator exhibited a significantly lower false positive rate (10%) compared with CNVkit (76%) on the test data (Supplementary Table S6). Therefore, we selected CNVnator for the final CNV calling and applied stringent filtering thresholds. We detected credible CNVs with an average number of 1,817 in the 60 genomes, ranging from a minimum of 1,066 to a maximum of 6,383 CNVs. The averaged cumulative length of CNVs was 59.6 Mb accounting for 2.2% of genome. Among the individuals, YD8 of Indian macaque had the shortest CNV accumulation length (30,100 kb), and CX10 of M. m. lasiotis had the longest accumulation length (369,850 kb). Most individuals had more copy number deletion than duplication, with the highest value of deletion/duplication in HN3 individuals (6.1:1) (Supplementary Table S7). The results indicated different individuals varied considerably on the CNVs number, cumulative length and type. We analyzed the CNVs among individuals, categorizing the counts of rare (individuals <3), major (3 ≤ individuals ≤ total), and shared (present in all individuals) CNVs for each subspecies. Notably, M. m. mulatta exhibited the most CNVs in all categories, with 11,288 rare |CNVs, 2,919 major CNVs, and 317 shared CNVs. In contrast, M. m. tcheliensis had the least CNVs in both the rare CNVs (3,581) and major (400) categories, and M. m. lasiotis had the least shared CNVs (84). We focused on CNVRs that had fixed within population, which we define as those shared by 3 or more individuals of a given subspecies (Supplementary Table S8).

Figure 1.

Map showing the geographic origins of the samples. Colors indicate the collection sites corresponding to each rhesus macaque subspecies. N represents the number of individuals sampled at each location. The map was generated based on the standard map (Review Number: GS [2024] 0650) downloaded from the National Platform for Common GeoSpatial Information Services (https://www.tianditu.gov.cn/), with no modifications to the base map boundaries. Digital elevation model (DEM) data were obtained from https://www.gscloud.cn/, and the map was produced using ArcGIS 10.0 software.

Mm_CNVRs distribution pattern and gene annotation

We merged CNVRs occurring in more than 2 subspecies to obtain rhesus macaque CNVRs (Mm_CNVRs). A total of 1,272 Mm_CNVRs were identified including 980 copy number deletion (DEL), 148 duplication (DUP), and 144 complex variation regions (both with DEL and DUP) (Supplementary Table S9). Among them, 68.95% (877/1,272) of the Mm_CNVRs intersected with the SV dataset from different macaque subspecies generated based on long-read sequencing (Supplementary Figure S2), indicating the reliability of our identification methods. Mm_CNVRs had a cumulative length of 85762.1 kb, accounting for 3.19% of the macaque genome (Table 1). The length of Mm_CNVRs ranged from 1.8 kb to 2215.3 kb, with about 51% (651/1,272) Mm_CNVRs in a length <20 kb, and CNVRs with longer length were very rare. Most DEL (89.8%, 880/980) showed a length <80 kb, and most DUP (89.2%, 132/148) had a length <60 kb. DUP had less number and shorter length than DEL in Mm_CNVRs (Figure 2A). Calculating cumulative length of Mm_CNVRs on chromosomes, we found that Mm_CNVRs distributed unevenly on the chromosomes, with a notable enrichment on chr19 accounting for 15.05% of the chromosome (Figure 2B).This distribution pattern is similar to that observed in other macaque species (Brasó-Vives et al. 2020; Li et al. 2020). Annotation of Mm_CNVRs found 734/1,272 (57.7%) Mm_CNVRs and their flanking sequences intersected with 1,356 genes, including protein-coding genes, lncRNA, tRNA, miRNA, rRNA, and other noncoding RNA genes. Of which, 639 Mm_CNVRs intersected with 1,021 protein-coding genes (Figure 2C). Finally, we constructed a genome-wide CNVs map (Figure 2D) labeling all DELs and DUPs on the corresponding position of chromosomes, and Mm_CNVRs intersecting with protein-coding genes were also labeled.

Table 1.

Statistics on the number and cumulative length of Mm_CNVR and subspecies CNVR.

Subspecies	Number of CNVR	Number of DEL	Number of DUP	Number of BOTH	Total length (bp)	Proportion of genome covered by CNVRs (%)
M. m. mulatta	3,236	2,687	536	13	40,308,900	1.50%
M. m. siamica	1,202	986	205	11	80,265,300	2.99%
M. m. lasiotis	1,621	1,285	291	45	64,497,900	2.40%
M. m. tcheliensis	576	370	205	1	37,050,600	1.38%
M. m. brevicaudus	656	523	130	3	32,281,200	1.20%
M. m. littoralis	1,201	964	223	14	52,372,500	1.95%
Mm_CNVR	1,272	980	148	144	85,762,100	3.19%

Figure 2.

(A) Distribution of 3 types of CNVR in the range of 0–400 kb length. (B) The ratio of the cumulative length of Mm_CNVR to autosome length. (C) The proportion of different types under Mm_CNVR gene annotation and the number of types are shown in the figure. (D) The comprehensive CNVs map of M. mulatta genome.

Differential CNVRs among subspecies

By merging CNVs of individuals in 1 subspecies, we obtained a total of 8,492 subspecies CNVRs in the 6 subspecies of rhesus macaques, with Indian macaque harboring the highest number (3,236) and M. m. tcheliensis the least number (576), respectively. The cumulative lengths of subspecies CNVRs were accounted for 1.20% (M. m. brevicaudus) to 2.99% (M. m. siamica) of the genome (Table 1). The number of DELs was much more than that of DUPs in all subspecies (DEL/DUP ranged 1.8–6.6). The Indian macaque had the highest number of DELs and DUPs compared with the 5 Chinese subspecies (Table 1).

A hierarchical cluster analysis based on the all CNVRs was conducted to exhibit genetic differentiation among subspecies (Figure 3A). The 60 individuals were grouped into 2 distinct clusters (the Indian cluster and the Chinese cluster), indicating the high genetic differentiation between them. Within the Chinese cluster, individuals from 1 subspecies grouped together forming 5 clusters. M. m. tcheliensis, the subspecies inhibiting the highest latitude among the Chinese macaques, diverged from other subspecies firstly, followed by the M. m. brevicaudus, the subspecies inhibiting the lowest latitude. The latest divergence occurred between M. m. littoralis and M. m. lasiotis. Compared with previous reported SNV phylogenetic tree on the 5 Chinese macaque subspecies (Liu et al. 2018), relationships among the Chinese subspecies based on CNVs is not consistent with those derived from SNVs.

Figure 3.

(A) Hierarchical clustering of all individuals according to the presence or absence of CNVR, each color box represents a macaque subspecies, from left to right: M. m. tcheliensis, M. m. brevicaudus, M. m. siamica, M. m. littoralis, M. m. lasiotis, and M. m. mulatta). (B) Mean values of pairwise Vst analysis of M. mulatta subspecies (in descending order of mean. (C) Manhattan plot of pairwise Vst analysis of 5 Chinese macaque subspecies, a point represents the Vst value obtained from a pairwise Vst analysis, and the red dotted line represents the screening threshold 0.56. (D) The number of Mm_CNVRs showing extremely significant differences was identified through pairwise nonparametric tests among subspecies.

In order to evaluate differentiation of gene regions among different subspecies, a pairwise Vst analysis was performed based on the 639 CNVRs intersecting with protein-coding genes. The average level of differentiation was quantified by calculating the mean of Vst between any 2 subspecies (Figure 3B). The Vst results were generally consistent with those of cluster analysis. Vst values between Indian macaque and different Chinese subspecies were relatively higher than that between 2 Chinese subspecies. The highest Vst value was 0.34 between M. m. mulatta and M. m. littoralis, while the lowest Vst was only 0.09 between M. m. siamica and M. m. lasiotis. Within Chinese subspecies, M. m. tcheliensis showed relatively higher Vst values to M. m. littoralis, M. m. brevicaudus, and M. m. siamica, indicating high differentiation of M. m. tcheliensis to other Chinese subspecies. Vst values between M. m. brevicaudus and M. m. littoralis, between M. m. siamica and M. m. lasiotis were the smallest.

Highly differential CNVRs among subspecies

We performed the K–W test on the Indian and Chinese macaques to identify highly differential CNVRs between them (P-value < 0.01). A total of 266 highly differential CNVRs were found, intersecting with 408 protein-coding genes (Supplementary Table S10). Enrichment analysis revealed these genes were predominantly associated with 8 immune-related GO terms and a drug metabolism-cytochrome P450 (mcc00982) KEGG pathway (Supplementary Figure S3), including immune-related genes such as AP3D1, AP3B1, RTD1A, RTD1B, RTD1C, IL15, ELANE, and drug metabolism related genes: GSTM2, GSTM1, UGT2B33, and CYP2C93. Indian macaques generally had more copies of these genes, for instance, CYP2C93 showed the biggest difference on median copy number with 5.5 in Indian versus 3.5 in Chinese macaques (Table 2).

Table 2.

Copy numbers of immune genes and drug metabolism genes in the 2 populations.

CNVR loci	Immune genes	Drug metabolism genes	Indian macaque (mean)	Chinese macaque (mean)	Indian macaque (median)	Chinese macaque (median)
19:1974301-2023200	AP3D1	—	1.8	1.4	1.8	1.3
6:75086001-75088600	AP3B1	—	1.6	0.8	1.9	0.8
6:75109901-75115500	AP3B1	—	1.7	1.4	2.0	1.2
8:7699501-7731000	RTD1C, RTD1B, RTD1A	—	2.0	2.8	2.0	2.7
5:139580101-139590300	IL15	—	1.6	0.4	2.0	0.1
19:586801-622500	ELANE	—	1.8	1.4	1.8	1.3
1:114448201-114466200	—	GSTM1, GSTM2	2.0	2.7	2.1	2.6
5:64610401-64809600	—	UGT2B33	1.6	1.1	1.7	1.2
9:95425301-95448300	—	CYP2C93	5.2	3.5	5.5	3.5

We further explored the highly differential CNVRs among the 5 Chinese subspecies by excluding India macaque for comparison to reduce the bias due to its high differentiation from Chinese subspecies. First, the pairwise Vst values were calculated to identify top 5% CNVRs (Vst value > 0.56) (Figure 3C;  Supplementary Table S10). Among them, M. m. tcheliensis vs M. m. littoralis, M. m. tcheliensis vs M. m. brevicaudus and M. m. tcheliensis vs M. m. siamica had a much more number of highly differential CNVRs than other pairwise subspecies (Supplementary Table S11). This indicates the highly differentiation of M. m. tcheliensis from the other Chinese subspecies. In contrast, no significant CNVRs were found in M. m. lasiotis vs M. m. siamica and M. m. lasiotis vs M. m. littoralis, and only one in M. m. lasiotis vs M. m. brevicaudus. Secondly, we performed pairwise comparisons using the Mann–Whitney U test on the CNVRs identified from the K–W test results (P-value <0.01) (Figure 3D). We found 39.59% (253/639) of CNVRs significantly differed between at least 2 subspecies (P-value < 0.01), intersecting with 455 protein-coding genes (Supplementary Table S10). M. m. tcheliensis vs M. m. littoralis had the highest number of highly differential CNVRs, while there were less number between M. m. brevicaudus and other subspecies.

Functional enrichment of the highly differential CNVR genes

We performed enrichment analysis on the highly differential CNVRs to understand their potential biological significance. The major enriched terms were grouped into 5 functional categories, and the number of genes intersecting with highly differential CNVRs in each category was quantified (Supplementary Figure S4). The major CNVRs were primarily enriched in the following pathways or functional terms (Supplementary Figure S5).

Gluconeogenesis and glucose metabolism

The highly differential CNVRs between M. m. tcheliensis and other subspecies (M. m. littoralis and M. m. siamica) were significantly enriched in gluconeogenesis (GO:0006094) and carbohydrate metabolism-related terms. We identified 2 genes of FBP1 and FBP2 were located within a highly differential CNVR at a length of 55,199 bp. Copy number of the CNVR was plotted exhibiting the highest copies of both genes in M. m. tcheliensis than in other Chinese subspecies (Figure 4A).

Figure 4.

Copy number distribution of 4 highly differentiated genes in each subspecies. (A) Gene FBP1, FBP2; (B) Gene NOG; (C) Gene TAS2R46; (D) Gene METRNL.

Ontogeny

The highly differential CNVRs between M. m. tcheliensis and other Chinese subspecies, and between M. m. siamica and M. m. lasiotis were enriched in somite development (GO:0061053), spinal cord development (GO: 0021510), lung morphogenesis (GO:0060425). These terms included genes such as NOG, SRF, SOX11. In addition, the highly differential CNVRs between M. m. littoralis and M. m. siamica, between M. m. littoralis and M. m. lasiotis enriched in thyroid hormone synthesis (mcc04918), osteoclast differentiation (mcc04380), negative regulation of osteoblast differentiation (GO: 0045668), bone mineralization (GO:0030282), which included genes of ACP5, TRAF2, LILRA4, NFKB2, HSPA5, SLC5A5, PDIA4, SEMA4D, NOG, SBNO2, IFITM5. In particular, we demonstrated variation of the NOG gene among the subspecies, with the highest gene copy in the M. m. tcheliensis and less copies in M. m. brevicaudus and M. m. siamica (Figure 4B).

Environmental adaptation

The highly differential CNVRs were enriched in bitter taste perception (GO:0001580), bitter receptor activity (GO:0033038), taste transduction (mcc04742) pathways, which included the TAS2R46 gene, a gene participating chemical stimuli detection. Copy number of TAS2R46 gene showed high variation among subspecies, which might be an adaptation of different Chinese subspecies to different food in their habitats. In M. m. tcheliensis TAS2R46 had the lowest copy among the 5 subspecies, and showed no polymorphism among 10 individuals indicating its fixation of the TAS2R46 loci in this subspecies (Figure 4C). In addition, the highly differential CNVRs enriched in pathways such as response to cold (GO:0009409), positive regulation of brown adipocyte differentiation (GO:0090336) and brown adipocyte differentiation (GO:0050873), which included a cold adaptation related gene METRNL. METRNL is associated with energy expenditure during exercise or after exposure to cold, and also associated with metabolic adaptation in cold environments. The median copy number of METRNL was highest in M. m. tcheliensis, the subspecies inhibiting the highest latitude among all subspecies. While subspecies inhibiting the warmer regions such as M. m. brevicaudus, M. m. siamica had the lowest copy of METRNL gene (Figure 4D).

To explore whether body size differences among macaque subspecies is related to CNVs, we searched body size-related genes in Mm_CNVRs, and identified 14 CNVRs overlapping with previously reported body size related genes(He et al. 2019) including 5 genes with significant copy number difference among the 5 subspecies (U test, P < 0.01; Table 3). Median copy numbers of such genes were plotted in Figure 5. Notably, all the 5 genes had the highest average copy numbers in M. m. tcheliensis who had the biggest body size among the 5 subspecies, and 3 genes (ARID1B, IGF1R, and IRS2) showed the least copy numbers in M. m. brevicaudus and M. m. siamica, both subspecies had relatively smaller body size compared with other subspecies (Figure 5).

Table 3.

Body genes with extremely significant differences among Chinese macaque subspecies (P-value < 0.01, P-values of differential genes are shown in bold).

Gene names	Pairwise comparisons
	M. m. siamica vs M. m. lasiotis	M. m. siamica vs M. m. tcheliensis	M. m. siamica vs M. m. brevicaudus	M. m. siamica vs M. m. littoralis	M. m. lasiotis vs M. m. brevicaudus	M. m. lasiotis vs M. m. littoralis	M. m. tcheliensis vs M. m. littoralis	M. m. brevicaudus vs M. m. littoralis
ARID1B	0.006	0.002	0.701	0.839	0.012	0.002	0.002	0.548
EPHA5	0.006	0.004	0.009	0.001	0.453	0.442	0.670	0.098
IGF1R	0.003	0.002	1.000	0.033	0.016	0.061	0.002	0.023
BRF1	0.531	0.203	0.065	0.005	0.154	0.006	0.007	0.956
IRS2	0.002	0.003	0.618	0.001	0.009	0.181	0.002	0.004

Figure 5.

Copy number distribution of 5 body genes in each subspecies. (A) Gene ARID1B; (B) gene EPHA5; (C) gene IGF1R; (D) gene BRF1; (E) gene IRS2.

Subspecies specific CNVRs

We also identified a total of 2,306 subspecies specific CNVRs (su_CNVRs), varied from 56 to 1,635 in the 6 subspecies (Table 4; Supplementary Table S9). The cumulative length of su_CNVRs ranged from 1532.1 kb (M. m. littoralis) to 13455.9 kb (M. m. siamica). Although M. m. mulatta had much more su_CNVRs than other subspecies, most M. m. mulatta su_CNVRs were very short in length. There was more DEL than DUP in su_CNVRs except in M. m. tcheliensis who had more DUP (56) than DEL (22) in su_CNVRs. Annotation of su_CNVRs revealed that 31.82% (M. m. lasiotis) to 67.86% (M. m. littoralis) su_CNVRs were annotated in the protein-coding regions (Supplementary Table S12). Functional enrichment of the su_CNVRs in each subspecies was showed (Supplementary Figure S6).

Table 4.

Statistics of su_CNVR of each subspecies.

Subspecies	No. of su_CNVR	No. of DEL	No. of DUP	No. of BOTH	Average (bp)	Median (bp)	Total length (bp)
M. m. mulatta	1,635	1,509	125	1	3,128.9	2,100	5,115,900
M. m. siamica	152	138	14	0	89,370.4	40,950	13,584,300
M. m. lasiotis	286	240	46	0	11,851.1	7,800	3,389,400
M. m. tcheliensis	78	22	56	0	61,230.8	24,000	4,776,000
M. m. brevicaudus	56	52	4	0	26,689.3	15,450	1,494,600
M. m. littoralis	99	89	10	0	15,506.1	11,100	1,535,100

M. m. mulatta su_CNVRs enriched in 20 nervous system-related pathways and 4 immune response-related pathways, involving 60 nervous system-related genes and 7 immune response-related genes (Supplementary Table S13). Among the genes, the TPPP located in a M. m. mulatta specific DUP. M. m. littoralis su_CNVRs enriched in 2 terms related to thyroid hormone secretion (mcc04919, GO: 0046966), and MED1, MED16 and CCND1 genes intersected with the subspecies specific DELs. M. m. lasiotis su_CNVRs enriched in 2 immune cell receptor signaling pathways (mcc04662 and mcc04660), 2 nervous system-related pathways (mcc04360 and GO:0010975), and the differentiation of osteoclasts pathway (mcc04380). M. m. siamica su_CNVRs mainly enriched in nervous system-related terms (GO:0021695, GO:0010976, GO:0007268, GO:0001755, and mcc04360), including genes related the development of the nervous system and the regulation of neural signal transmission. These genes all intersected with the subspecies-specific deletion regions.

M. m. tcheliensis su_CNVRs significantly enriched in the terms of osteoclast differentiation (GO: 0001649) and growth factor activity (GO: 0008083). Two genes of GDF10 and GDF2 were completely covered by a subspecies-specific DUP at a length of 234.6 kb. No enriched entry was significant in M. m. brevicaudus su_CNVRs; however, we identified a shared enriched entry, the SMAD protein signal transduction (GO: 0060395) both in the su_CNVRs of M. m. brevicaudus and M. m. tcheliensis. The SMAD term contained many GDF (growth differentiation factor) family genes, including GDF2 and GDF10 that intersected with a DUP specific in M. m. tcheliensis (the largest body-sized subspecies), and GDF1 gene that intersected with a DEL specific in M. m. brevicaudus (the smallest body-sized subspecies). Variation on the copy number of these GDF genes between the 2 subspecies was illustrated in Supplementary Figure S7.

qPCR validation

Four highly differential CNVR genes that had potentially biological significance were selected for qPCR validation. The qPCR results of NOG gene copy number were highly consistent with that of bioinformatics analysis (Figure 6D). While copy numbers of the ARID1B, FBP1, and IRS2 genes in the 5 subspecies showed similar trend between the qPCR and bioinformatics analysis (Figure 6A–C), with relatively higher copy numbers in the M. m. tcheliensis than in other subspecies. There were 2 possible reasons leading to the difference. First, the individuals used for bioinformatics analysis and qPCR experiment were not same. Previous study indicated that gene copy numbers showed polymorphic among populations, even among individuals in the same population (Pierce et al. 2018; Li et al. 2020; Nyangiri et al. 2020). Secondly, qPCR for estimation of gene copy is relatively quantitative, which estimates gene copy referring to the internal control gene, the HBB. Although the exact copy number of genes could not be determined, the trend of gene copy numbers among different subspecies should be similar between qPCR and bioinformatics analysis.

Figure 6.

Comparison of copy number distributions determined by real-time quantitative PCR and bioinformatics analysis. Blue represents the average copy number of genes within each subspecies based on NGS data, and red represents the average copy number within each subspecies validated by qPCR experiments. (A) Gene ARID1B; (B) gene FBP1; (C) gene IRS2; (D) gene NOG.

Discussion

Characterization of genome-wide CNVs in rhesus macaque

While CNVs represent ubiquitous genomic features contributing substantially to population diversity, their functional significance and evolutionary trajectories remain under-characterized in rhesus macaques, particularly compared with well-documented SNVs. This study, based on whole-genome sequencing data from 60 individuals representing 6 different subspecies (including India macaque and 5 subspecies of Chinese macaque), provides a comprehensive characterization of CNVs. Our analysis reveals that CNVs span an average cumulative length of 59.6 Mbp, constituting 2.2% of the rhesus macaque reference genome. This genomic proportion aligns with previous findings from a 198-individual cohort (Brasó-Vives et al. 2020) and represents ∼7-fold greater than that of SNPs (8.5 Mbp) (Liu et al. (2018). Additionally, we identified a total of 1,272 macaque CNV regions (Mm_CNVRs) integrating CNVs across 60 individuals, enabling us to construct a comprehensive CNV map. Consistent with findings in other species (Redon et al. 2006; Stafuzza et al. 2019; Shi et al. 2023), our study observed a significantly higher number of DEL events than DUP events, and Mm_CNVRs showed preferentially occurred in genic regions, which aligns with findings in the human genome (Conrad et al. 2010) and other macaque species (Lee et al. 2008; Li et al. 2020).

Identification of CNVs from genome has long been challenged despite the varied CNV calling strategies. This suggests that the sensitivity for detecting, specifically DEL and DUP, may differ between aCGH-based methods and read-depth-based approaches. This variation in sensitivity could explain the significantly higher number of DELs compared with DUPs identified in our study, which contrasts with the findings reported by Lee et al. (2008) and Ng et al. (2015). Although CNV detection based on whole-genome sequencing performed better than array-based methods (Zhou et al. 2018), its accuracy, sensitivity, and breakpoint resolution are still influenced by sequencing depth, with higher depth generally improving sensitivity and accuracy (Ahmad et al. 2023). Due to the lack of enough genomic data from different macaque subspecies, the genomes of some Chinese subspecies were only available at ∼10× sequencing depth, which may pose certain limitations for CNV calling. To improve detection accuracy, we applied different window strategies for genomes with different sequencing depths, following the recommendations of previous studies (Abyzov et al. 2011). This approach may reduce the impact of sequencing depth differences on CNV detection, as the total length of CNVs did not show significant variation across samples with different sequencing depths. However, it is worth noting that such strategies may also introduce new biases, which warrant further investigation in future studies. Additionally, we implemented stringent filtering criteria and assessed individual CNV overlaps. Notably, the majority of identified Mm_CNVRs (68.95%) can be validated in the SV dataset based on long-read sequencing platform (unpublished data), further supporting the reliability of our findings.

Although short-read whole-genome sequencing provides a practical and widely adopted method for CNV detection, particularly in estimating absolute copy numbers, it has certain limitations in breakpoint resolution due to local sequence complexity and variations in sequencing depth (Escaramis et al. 2015). To improve the resolution and accuracy of CNV identification, especially in repetitive or structurally complex genomic regions, future studies incorporating long-read sequencing for CNV detection in rhesus macaques will facilitate a more comprehensive analysis of genomic structural variations (Lavrichenko et al. 2021). This approach will help advance our understanding of CNV diversity and its evolutionary significance in rhesus macaques.

Genetic divergence among subspecies based on CNVs

Phylogenetic relationships among macaque subspecies have been investigated based on genome-wide SNVs, revealing high genetic differentiation between Indian and Chinese macaque (Hernandez et al. 2007), while within Chinses macaques, M. m. siamica diverged firstly from other subspecies, and divergence between M. m. tcheliensis and M. m. brevicaudus occurred most recently (Liu et al. 2018). We estimated genetic differentiation using CNVRs based hierarchical clustering. The high differentiation between Indian and Chinese macaques was supported. However, relationships among Chinese subspecies were inconsistent with the SNVs-based phylogenetic tree reported by Liu et al. (2018). M. m. tcheliensis the subspecies distribution the northernmost showed the greatest divergence from other subspecies, followed by the M. m. brevicaudus, the subspecies distribution the southernmost (Figure 3A). Vst analysis also supported the hierarchical cluster result (Figure 3B), with relatively higher Vst values in M. m. tcheliensis than other Chinese subspecies. The results may imply that CNVs are less informative as phylogentic markers compared with SNVs. This may stem from fundamental differences in the nature and resolution of these 2 types of genetic variation. Firstly, SNVs are generally more abundant and stable across the genome, while CNVs often result from structural genomic rearrangements and may be subject to stronger selective pressures or recurrent mutation events (Freeman et al. 2006; Redon et al. 2006). Environmental stimulation can change copy number of genes interacting with environment, thereby accelerating adaptation to the environment (Hull et al. (2017). Numerous studies demonstrated CNVs as an important genetic mechanism for environmental adaptation in humans, rhesus macaques, domestic animals and plants (Narang et al. 2014; Prunier et al. 2017; Antunes de Lemos et al. 2018; Li et al. 2020). Therefore, the CNVR based differentiation among macaque subspecies in our study probably reflects adaptation of subspecies to environment rather to evolutionary relationships. Secondly, different types of CNVs (duplications and deletions) are generated through distinct molecular mechanisms and occur at various genomic locations, making them under different selective pressures. As a result, the phylogenetic signals conveyed by different CNV types can vary substantially. A previous study based on 236 individuals from 125 human populations found that duplications were less correlated with phylogenetic relationships and exhibited stronger signatures of selection than deletions (Sudmant et al. 2015). In the future, analyses based on larger-scale genomic datasets are needed to further explore differences on the phylogenetic information between these 2 types of CNV, as well as their potential utility as genetic markers.

Highly differential CNVRs underlie immunological differentiation between Chinese and Indian macaques

Our comparative analysis revealed highly differentiated CNVRs between Indian and Chinese rhesus macaque populations, with functional enrichment in genes related to immune regulation and xenobiotic metabolism. Previous study proved that variations in gene copy number linked with immune competence in macaques, for instance, Chinese macaques exhibited elevated CCL3L1 copy numbers compared with Indian macaques, which had caused the delayed SIV disease progression in Chinese macaques (Degenhardt et al. 2009). Intriguingly, we observed the highly differential CNVR gene AP3B1—a key modulator of viral release of HIV (Azevedo et al. 2009)—with significantly lower copy numbers in Chinese populations (0.8) versus Indian macaques (1.9; P < 0.01), suggesting potential difference in lentiviral defense mechanisms. We hypothesize that this pattern may reflect natural selection driven by regional pathogen pressures. Indian macaques typically inhabit warmer and more humid tropical rainforest environments, where exposure to a broader and more diverse array of pathogenic bacteria and viruses is likely. The increased copy number of immune-related genes in Indian populations may enhance immune responses and provide a selective advantage under such pathogen-rich conditions. In addition, 4 drug-metabolism enzyme genes (GSTM2, GSTM1, UGT2B33, and CYP2C93), were found intersecting with highly differential CNVRs between Indian and Chinese macaque. Drug metabolism genes exhibit high polymorphisms and can significantly affect drug metabolic levels and drug half-life. CYP genes encode cytochrome P450 oxidase, and variations on the CYP genes can cause differences in drug metabolism between species or individuals (Zanger and Schwab 2013). Human individuals with more than 2 copies of CYP2D6 gene metabolized substrates more rapidly (He et al. 2011). Given the high variation on the copy numbers of immunity genes and drug-metabolism genes between Indian macaque and Chinese macaque, the possible impacts on the results should be taken into account when they were applied to biomedical studies.

Contributions of CNVs to local adaptation in Chinese macaque subspecies

Genes intersecting with environmental factors are frequently subjected to natural selection resulting in CNVs as adaptive response to environmental shifts. These CNVs are suggested to critically influence phenotypic diversity and adaptive traits (Guo et al. 2021). To investigate CNV contributions to local adaptation among Chinese macaque subspecies, we conducted comprehensive analysis to identify highly differential CNVRs and subspecies specific CNVRs. Notably, many of these CNVRs intersected with genes function on environmental adaptation and phenotypic variation. Copy numbers of several such differential CNV genes were further validated using qPCR. In particular, 1 highly differential CNVR contained 2 gluconeogenesis related genes, FBP1 and FBP2, which exhibited elevated copy numbers in M. m. tcheliensis than in other subspecies (Figure 4A). FBP genes are highly expressed in the liver and muscle of vertebrates, encoding rate-limiting enzymes in gluconeogenesis, a process of converting noncarbohydrates into glucose or glycogen (Dzugaj 2006). Gluconeogenesis ensures body's blood sugar levels remain at normal levels, while it can be greatly enhanced during long-term starvation. Zebrafish study indicated that FBP1 expression was significantly upregulated during starvation (Drew et al. 2008). M. m. tcheliensis inhabits the Taihang Mountains in China and is the highest latitudinal distribution among subspecies habitats. It often suffers extreme winter temperatures as low as −14 °C (Tian et al. 2013) and prolonged seasonal food scarcity with reliance on low-energy, high-fiber vegetation (Lu et al. 2007). Such conditions create sustained metabolic challenges during winter and early spring, necessitating efficient gluconeogenic capacity for blood glucose maintenance. Previous study indicated both FBP1 and FBP2 genes were under positive selection in M. m. tcheliensis (Liu et al. 2018). Our study provides adaptive mechanism in M. m. tcheliensis by elevating copy number of FBP1 and FBP2 genes to enhance gluconeogenesis, enabling survival on minimal food intake while sustaining essential glucose levels.

Moreover, we identified METRNL (meteorin-like) gene associated with cold environment adaptation, within a highly differentiated CNVRs. METRNL gene had the highest copy number in the M. m. tcheliensis (distributed at the highest latitude), whereas subspecies inhabiting warmer regions (M. m. brevicaudus and M. m. siamica) showed the lowest copy number (Figure 4D). METRNL protein is a circulating factor functionally induced in muscle after exercise and in adipose tissue on cold exposure. Elevated METRNL levels could enhance energy expenditure, improve glucose tolerance, and activate beige fat thermogenesis-related genes in murine models, and confirmed that METRNL is essentially required for cold-induced adaptive thermogenesis (Rao et al. 2014). The observed intersubspecies variation in METRNL gene copy number (Figure 4D) potentially drives metabolic adaptations to thermal stress across macaque populations likely by modulating circulating METRNL protein levels.

Body size of different Chinese macaque subspecies is accordance with Bergmann's rule, endothermic species in colder climates typically exhibit larger body sizes to minimize heat loss (Bergmann 1847). The northernmost subspecies M. m. tcheliensis demonstrates the largest body size, followed by the southwestern M. m. siamica, and the southernmost M. m. brevicaudus the smallest, while genetic mechanism of the adaptive trait remains uncovered. Importantly, our study evidenced that CNVs had high association with body size variation of macaque subspecies. We identified 6 genes exhibiting highly CNV differentiation across subspecies (NOG, ARID1B, IGF1R, EPHA5, BRF1, and IRS2). The largest-bodied M. m. tcheliensis showed the highest mean copy numbers for all 6 genes, whereas smaller subspecies (M. m. brevicaudus and M. m. siamica) displayed the lowest copies for 4 genes. Our study implies a dosage pattern, higher copies of these genes linked with bigger body size and lower gene copies with smaller body size in Chinese subspecies. Among the identified CNV genes, NOG gene participate multiple pathways on skeletal development and somite formation. It encodes antagonists of bone morphogenetic proteins (BMPs) critical for skeletal development (Hogan 1996). Mice with a homozygous deletion of NOG showed developmental disorders such as severe skeletal dysplasia, short bones, and loss of bone and multiple joints (Brunet et al. 1998). ARID1B is a novel reported short stature gene, functional loss of ARID1B may lead to Coffin–Siris syndrome with small body size, while missense mutation of ARID1B can cause the nonsyndromic based short stature (Yu et al. 2015). Moreover, knockdown of ARID1B in zebrafish embryos significantly reduced body length (Liu et al. 2020). The median copy number of ARID1B in M. m. tcheliensis is about 1.2, while it is no more than 0.2 in M. m. brevicaudus, M. m. siamica, and M. m. littoralis. IGF1R encodes the insulin-like growth factor 1 receptor, it is a tyrosine kinase that phosphorylates various proteins, including the insulin receptor substrates encoded by IRS1 and IRS2. Mice knocked out of the IGF1R are 50% smaller than normal mice (LeRoith and Accili 2008). IRS2 generally cooperates with IRS1 they play an important role in maintaining blood glucose homeostasis, coordinating skeletal muscle growth and development (Dong et al. 2006; Long et al. 2011). Our study suggests CNV-mediated regulation of body size is a dosage pattern in macaque subspecies. Computational simulation of effects of genic CNVs demonstrated that increment in 1 or 2 copies had large effects on overall expression due to regulatory feedbacks (Mileyko et al. 2008; Chiang et al. 2017). Elevation of copy number in these genes may enhance expression of transcripts and protein through dose effect, as the copy numbers of exons or adjacent regulatory elements are increased, which potentially drives larger body size, and has a positive effect on reducing body heat loss improving tolerance to nutrient consumption in winter. However, the lack of subspecies-specific transcriptomic data limits our ability to assess the functional consequences of these CNVs. Future studies integrating gene expression data and conducting functional assays at the cellular level or in alternative model organisms will be necessary to validate the regulatory or phenotypic effects of candidate CNVs.

Furthermore, our analysis revealed numerous subspecies-specific CNV regions (su_CNVRs) with potential functional significance in driving phenotypic divergence. Notable, among the 6 subspecies only M. m. tcheliensis has more DUP su_CNVRs than DEL, indicating its unique genetic background. One M. m. tcheliensis specific DUP covering 2 GDF genes (GDF10 and GDF2) may be involved in its biggest body size. GDF10 is a regulatory factor for cell growth and differentiation in embryonic and adult tissues, and also is the most important members of BMPs (Hino et al. 2004). It regulates bone morphogenesis (Cunningham et al. 1995; Takao et al. 1996), mutations on the GDF10 lead to body size variation among cattle populations (Adoligbe et al. 2012). And GDF2 synergistically interacts with BMP2 to coordinate limb development. Conversely, we found a deletion specifically occurred in the GDF1 gene of M. m. brevicaudus, who is smallest in body size. GDF1 is an endogenous factor, and expression of GDF1 is significantly down-regulated in the brown adipocytes of both genetic and acquired obesity mice (Yuki et al. 2016). The deletion in the GDF1 gene of M. m. brevicaudus may inhibit the expression of GDF1 leading to a slender shape in the subspecies.

In conclusion, this study characterizes the genome-wide CNV landscape of Macaca mulatta based on genomes of 60 individuals across 6 subspecies. We identified 1,272 nonredundant CNV regions, covering ∼59.6 Mb (2.2%) of the reference genome. The CNV-based hierarchical clustering showed discordance with SNV-derived phylogenetic relationships among Chinese subspecies. The highly differentiated CNVRs among different subspecies were identified, notably many of these CNVRs intersected genes associated with environmental adaptation in different subspecies, including FBP1, FBP2, NOG, TAS2R46, and METRNL. In addition, we identified 5 body size related genes with significant copy number differences across 5 Chinese subspecies: ARID1B, IGF1R, EPHA5, BRF1, and IRS2, suggesting potential links of CNV with body size of rhesus macaque. Finally, the number of subspecies specific CNVRs identified across the 6 subspecies ranged from 56 to 1,635. Collectively, our findings underscore the genomic variations of CNVs and their relevance in macaque adaptation, which provides a valuable resource for future population and evolutionary genomic research in primates.

Authors’ contributions

J.L. led the project. Y.L., Z.F., and J.L. designed the study. Y.L. and L.Z. performed bioinformatics analysis. Y.L. wrote the manuscript. J.L., L.Z., J.X., and M.L. revised the article. All authors read and approved the final version of the manuscript. Y.L. and L.Z. contributed equally to this work.

Funding

This research was funded by the National Natural Science Foundation of China (Grants No. 32171607 and 32371696).

Ethics statement

All samples were obtained in accordance with Chinese regulations for the implementation of protection of terrestrial wild animals (State Council Decree [1992] No. 13) and all laboratory work was approved by the Guidelines for Care and Use of Laboratory Animals and the Ethics Committee of Sichuan University (Chengdu, China) (Permit No.: 20210308003).

Data availability

The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (Chen et al. 2021) in National Genomics Data Center (CNCB-NGDC Members and Partners 2022), China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA: CRA013624) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa.

Supplementary material

Supplementary material can be found at https://academic.oup.com/cz.