Functional genomics analysis reveals the evolutionary adaptation and demographic history of pygmy lorises

Significance

Although strepsirrhines occupy a key node in primate phylogeny, our knowledge of this group remains quite limited. Here, we integrate comparative genomics analyses and functional assays to reveal the genetic underpinnings of evolutionary adaptation in the pygmy loris, a unique strepsirrhines group. We identified a series of genes that have contributed to certain distinctive adaptive traits of the pygmy loris, namely PITRM1 (low metabolic rate), MYOF (slow movement), and PER2 (hibernation). Our findings serve to deepen our understanding of the adaptive evolution of the strepsirrhines, and may also provide useful information for future studies of human disorders related to abnormal metabolism, skeletal muscle development, and circadian rhythms.

Abstract

Lorises are a group of globally threatened strepsirrhine primates that exhibit many unusual physiological and behavioral features, including a low metabolic rate, slow movement, and hibernation. Here, we assembled a chromosome-level genome sequence of the pygmy loris (Xanthonycticebus pygmaeus) and resequenced whole genomes from 50 pygmy lorises and 6 Bengal slow lorises (Nycticebus bengalensis). We found that many gene families involved in detoxification have been specifically expanded in the pygmy loris, including the GSTA gene family, with many newly derived copies functioning specifically in the liver. We detected many genes displaying evolutionary convergence between pygmy loris and koala, including PITRM1. Significant decreases in PITRM1 enzymatic activity in these two species may have contributed to their characteristic low rate of metabolism. We also detected many evolutionarily convergent genes and positively selected genes in the pygmy loris that are involved in muscle development. Functional assays demonstrated the decreased ability of one positively selected gene, MYOF, to up-regulate the fast-type muscle fiber, consistent with the lower proportion of fast-twitch muscle fibers in the pygmy loris. The protein product of another positively selected gene in the pygmy loris, PER2, exhibited weaker binding to the key circadian core protein CRY, a finding that may be related to this species’ unusual circadian rhythm. Finally, population genomics analysis revealed that these two extant loris species, which coexist in the same habitat, have exhibited an inverse relationship in terms of their demography over the past 1 million years, implying strong interspecies competition after speciation.

Strepsirrhines occupy a key node in primate phylogeny, and possess characteristics that are considered plesiomorphic to those of the haplorhines (monkeys, apes, and humans) (Fig. 1A and SI Appendix, Fig S1A). Although some strepsirrhine genomes have been sequenced (1–4), high-quality genome sequences from this suborder are still rare, and this has impeded studies on the adaptive mechanisms underlying strepsirrhine evolution. Slow lorises (genus Nycticebus) are a group of globally threatened strepsirrhines that mainly inhabit the forests of Southeast Asia. Nine species are now recognized; they are all globally threatened (5). Notably, a recent study proposed pygmy slow loris (Nycticebus pygmaeus) as a new genus that differs from the other Nycticebus species from morphological, behavioral, karyotypical, and genetic data, and suggest a common name of pygmy lorises (Xanthonycticebus pygmaeus) to distinguish it from other slow lorises (6).

Fig. 1.

Study schematic. (A) Phylogenetic tree of primates and the crucial genes contribute to the specific phenotypes of pygmy loris. (B) Karyotype of pygmy loris, and location of crucial genes at different chromosomes.

Pygmy lorises have evolved certain distinctive adaptive traits that are rarely seen in other primates (7). For example, pygmy lorises have a much lower metabolic rate relative to other eutherian species of similar body mass, which is thought to be related to the high concentrations of toxins and digestion inhibitors in their largely exudate diet (8–11). In addition, pygmy lorises display slow movement and locomotion among the strepsirrhines, and are characterized by a muscle fiber composition that differs from the general condition in mammals (12), which is characterized by a predominance of slow-twitch muscle fibers. These fibers are slow to contract and consume less energy than the white muscle fibers found in most primates, but are better suited to tasks that require endurance, such as sustained locomotion (13–15). Pygmy lorises are the few primates that harbor toxins (16). The toxin in pygmy lorises is activated when oils exuded from a brachial gland are combined with saliva (17). Pygmy loris venom is known to function in ectoparasite control and as a protection against predators (16). It is one of the rare lineages that use venom for intraspecific competition (18), and can cause death in other slow lorises and small mammals as well as anaphylactic shock in humans (16). This notwithstanding, pygmy lorises display resistance to their own toxins (19). In addition, the lorises have the capacity to hibernate, a very rare phenomenon in primates (20, 21). Although these features are highly unusual for a primate, the precise mechanisms underlying their particular characteristics remain largely unknown.

Here, we assemble a chromosome-level genome sequence of the pygmy loris (X. pygmaeus) and resequence 50 whole genomes from pygmy lorises and 6 whole genomes from Bengal slow lorises (BSL, Nycticebus bengalensis). The genome sequence has enabled us to obtain unprecedented insights into the unique biology of the pygmy lorises, without having to harm or disturb an animal of conservation concern. Moreover, revealing the molecular genetics that underpin these unique primates may provide insights relevant to human genetic disease (22, 23), including disorders of metabolism, skeletal muscle development, and circadian rhythms. The availability of the genome sequence should also empower a holistic, scientifically grounded approach to loris conservation.

Results

Pygmy Loris Chromosome-Level Assembly by Long-Read Sequencing.

We set out to study the genetics and genomics of the pygmy loris and to explore the underlying evolutionary mechanisms of this highly unusual group of primates. We first sequenced the genome of a female pygmy loris using single-molecule real-time sequencing technology (PacBio Sequel System) (Dataset S1, Tables S1–S3). From the long-read sequences, we assembled the pygmy loris genome into 2,844 contigs, with a contig N50 of 4.73 Mb. An additional 60.82 Gb of Illumina-based short reads were generated to correct sequencing errors, and 254-Gb clean high-throughput chromosome conformation capture (Hi-C) data were used to order the contigs (SI Appendix, Fig. S1B). The final assembly was 2.78 Gb distributed across 25 (2n = 50) chromosome-level pseudomolecules with a scaffold N50 of 129.9 Mb (Fig. 1B and Dataset S1, Table S4). The pygmy loris genome showed high quality and contiguity relative to other sequenced strepsirrhine genomes (1–4). For example, compared with the mouse lemur (Microcebus murinus) genome (1), the pygmy loris genome contained fewer scaffolds (7,679 versus 2,844, respectively) and a higher scaffold N50 (103.22 Mb vs. 129.9 Mb). Compared with the bushbaby (Otolemur garnettii), the pygmy loris genome had 9,278-fold better contiguity and a 9.4-fold higher scaffold N50 (13.85 Mb vs. 129.9 Mb, respectively). Gene annotation identified a total of 26,200 protein-coding genes in the pygmy loris genome (SI Appendix, Fig. S1C and Dataset S1, Table S5); this is somewhat higher than that seen in other primates (4, 24, 25) and may well result from false positives due to the use of ab initio predictions (26). Based on the well-annotated protein-coding genes, we performed comparative genomics, including analyses of the expansion/contraction of gene families, positive selection, and convergent evolution, in an attempt to reveal some of the potential genetic mechanisms underlying adaptive evolution in pygmy lorises.

Enhanced Detoxifying Ability Revealed by Comparative Genomics.

Contrary to popular belief that pygmy lorises are largely frugivorous, pygmy lorises feed regularly on exudates that they obtain through active gouging, but are also capable of consuming insect prey (27) that contain toxins or digestive deterrents that are suspected to have negative effects on other mammals (8). For example, several of the secondary metabolic products that are commonly consumed by pygmy lorises contain high levels of heterocyclic compounds, such as phenolics and terpenoids (8). The sequencing of the pygmy loris genome revealed substantial expansion of various gene families involved in the binding and metabolism of heterocyclic compounds (SI Appendix, Fig. S1D and Dataset S1, Table S6). Considering the abundance of heterocyclic compounds that exist as secondary metabolites of the pygmy loris diet (8), we propose that these expanded gene families may enhance the ability of pygmy loris to detoxify secondary metabolites in their diet.

By integrating transcriptome data from eight pygmy loris organs (i.e., heart, liver, spleen, stomach, kidney, brain, muscle, and tongue), we identified a total of 29 expanded gene families that are highly and specifically expressed in the liver, a vital organ for xenobiotic detoxification and metabolism (SI Appendix, Fig. S1E and Dataset S1, Table S7). One of the most prominent of these gene families was represented by amplified copies of the glutathione S-transferase α (GSTA) gene (Fig. 2A). This gene encodes an enzyme that functions in phase II detoxification of electrophilic compounds (Fig. 2B), including carcinogens, therapeutic drugs, environmental toxins, and products of oxidative stress, in conjugation with glutathione (28). We identified 14 copies of the GSTA gene family in the pygmy loris haploid genome, considerably more than in other mammals, including humans, macaques, and mice (Fig. 2A). Because the 14 GSTA copies locate to different chromosomes in the pygmy loris genome (Fig. 1B), they may not be due to tandem duplication. Transcriptome analysis across species revealed that the expression of GSTA in pygmy loris liver and stomach was significantly higher than that in humans and mice (Fig. 2C and Dataset S1, Table S8). Thus, based on the genome and transcriptome results, we propose that the GSTA gene family enhances the ability of pygmy lorises to detoxify secondary metabolites from their diet. Interestingly, we found that the GSTA family is also significantly expanded in small-eared galago (O. garnettii), a species of bushbabies with a diet consisting of 50% animal matter (mainly invertebrates, such as beetles, orthopterans, and centipedes) and 50% fruits (29, 30). Thus, expansion of the GSTA gene family may also have helped the small-eared galago to detoxify secondary metabolites in its diet.

Fig. 2.

Functional genomics study of pygmy loris. (A) Unrooted neighbor-joining tree showing the GSTA gene families across species, with 14 copies in pygmy loris and relatively few copies in other species. Red stars mark two GSTA genes showing high expression levels in pygmy loris liver. (B) FMO5 and GSTA are involved in liver detoxification. (C) Expression levels (FPKM) of GSTA in six tissues from humans, mice, and pygmy lorises are presented. (D) Expression levels (FPKM) of FMO5 in eight tissues from pygmy loris. (E) The alignment of mammalian PITRM1 amino acid sequences was suggestive of three convergent substitutions between koala and pygmy loris. Amino acids unique to pygmy loris and koala are shown in red; other amino acids at these positions are shown in blue. (F) Reaction progress curves depicting sample fluorescence (RFU) as a function of time(s) for peptide hydrolysis assays conducted using Mca-KLVFFAEDK(Dnp)-OH, an established target of PITRM1 activity. (G) V_max and K_m values of human and pygmy loris PITRM1 enzymes.

We used the codeml package as implemented in PAML (11) to detect branch-specific positive selection in pygmy lorises, and identified a total of 610 genes displaying signals of positive selection (Dataset S1, Table S9). From the transcriptome data, we also identified 453 genes specifically and highly expressed in the pygmy loris liver compared with other tissues of this species (Dataset S1, Table S10), hereafter termed liver-specific expressed genes. Among the liver-specific expressed genes, four (FMO5, TRPC5, GDPD4, and ABCC9) exhibited positive selection in the pygmy loris lineage (SI Appendix, Fig.S1F). Of these, flavin-containing monooxygenase 5 (FMO5) plays an important role in the phase I detoxification and oxygenation of a variety of xenobiotics (Fig. 2B) (31). Compared with FMO5 expression in mouse and human (SI Appendix, Fig.S1G), our transcriptome analysis showed high expression of positively selected FMO5 in the pygmy loris liver, suggesting a role in detoxification (Fig. 2D).

Convergent Evolution of PITRM1 between Pygmy Loris and Koala May Have Contributed to Their Characteristic Low Metabolic Rate.

Because of the high concentrations of toxic secondary compounds they encounter and their low calorific diet, pygmy lorises must maintain a low basal metabolic rate in order to detoxify the secondary compounds and to conserve energy (8, 9). This strategy is also employed by koalas (32). For this reason, we sought evidence for convergent evolution between the pygmy loris and koala, and identified 226 genes containing amino acid substitutions that showed parallel evolutionary patterns specifically between the two species (Dataset S1, Table S11). Of note, the PITRM1 gene displayed three convergent amino acid substitutions in both the pygmy loris and koala (i.e., K633R, T697S, and D925E), which were conserved in the nine other mammals screened (Fig. 2E). Two of these substitutions (K633R, T697S) are also present in the Sunda slow loris (SI Appendix, Fig.S2A), which is a species of slow loris that also exhibits a low metabolic rate like pygmy loris (8). Our result suggests a similar adaptive evolutionary mechanism at the genetic level between Sunda slow loris and pygmy loris.

PITRM1 is a highly conserved zinc metalloprotease, also known as presequence peptidase (PreP), which plays an important role in metabolism (33). Knockdown of PITRM1 in mice is reported to decrease basal metabolism and oxygen consumption (34). To test the potential impact of the observed convergent changes in PITRM1 on metabolic function in pygmy loris, we studied the functional consequences of the amino acid substitutions identified in PITRM1 (i.e., K633R, T697S, and D925E). Our activation kinetics model showed that pygmy loris PITRM1 exhibited a twofold decrease in enzyme activity relative to that in humans (Fig. 2 F and G), suggesting that the substitutions in the pygmy loris PITRM1 sequence contribute to a decrease in basal metabolic rate, similar to PITRM1 knockdown effect in mice (34).

Adaptive Evolution of Genes Involved in Muscle Development.

Pygmy lorises display slow movement and locomotion compared with many other mammals and the composition of skeletal muscle fiber in pygmy lorises differs from that of other primates (13, 15). Notably, pygmy lorises show a predominance of red (slow-twitch) muscle fibers, which are slow to contract and consume less energy than the white muscle fibers found in most primates (13–15). To explore the genetic mechanisms underlying the special skeletal muscle of pygmy lorises, we used our genomic sequence data to identify amino acid substitutions that correlated with muscle fiber development. Three genes (MYOF, CHRNA1, and HMCN2) that are known to play important roles in muscle fiber development were found to be under positive selection in pygmy lorises (P < 0.05 by branch site model in PAML) (Fig. 3A) (35). MYOF, which encodes the myoferlin protein, is associated with the plasma membrane (36). In humans, dysfunction of MYOF can cause abnormal muscle morphology and skeletal muscle fiber degeneration, which affects both proximal and distal muscles (37). Twelve of 24 amino acid positions in MYOF that display positive selection in pygmy lorises are also present in Sunda slow lorises (SI Appendix, Fig. S2B), which also exhibit slow movement like pygmy lorises. To explore the functional consequences of positive selection on MYOF, we performed a functional assay by transfecting mouse myoblasts with pygmy loris MYOF, human MYOF, as well as an empty plasmid as a control. The assay revealed that the ability to up-regulate the expression of the MyHC2a (i.e., MYH2) and MyHC2x (i.e., MYH1) genes, which encode fast twitch myosin motor proteins, was weakened in the pygmy loris MYOF as compared to human MYOF (Fig. 3B). This finding is consistent with the lower proportion of fast-twitch muscle fibers found in pygmy lorises.

Fig. 3.

Genetic mechanism underlying muscle evolution and abnormal circadian rhythm in pygmy loris identified by functional genomics. (A) Positive selection of MYOF, and alignment of protein sequence surrounding the positively selected site (P285A). (B) Pygmy loris MYOF gene significantly inhibited the expression of the gene encoding the fast-twitch muscle fiber, as compared to the human MYOF gene. **P < 0.01, *P < 0.05. (C) Convergent evolution of the NEB protein between pygmy loris and two-toed sloth (C. hoffmanni). The alignment of mammalian NEB amino acid sequences is presented. Amino acids unique to pygmy loris and sloth are shown in green; alternative amino acids at these positions in other mammalian NEB orthologs are shown in orange. (D) Alignment of mammalian amino acid sequences of period domain of PER2. Candidate amino acid substitution (S1271A) in pygmy loris is shown in blue; other amino acids at these positions are shown in red. (E) Three-dimensional structure model (PDB ID: 6OF7) (110) shows that the S1271A substitution in PER2 is located within the binding domain of the PER2-CRY complex. The gray shading represents PER2, whereas the light-yellow shading represents CRY1. The site of this substitution is depicted in red. (F) Coimmunoprecipitation assay identified the weakened binding ability of pygmy loris PER2 with CRY1, as compared with human. WT is human Per2; SA is S1271A mutated human Per2; Loris is Per2 of pygmy loris.

We also performed convergent evolution analysis between the pygmy loris and the two-toed sloth (Choloepus hoffmanni), a mammalian species well known for its slow movement. A total of 20 protein-coding genes showing signals of convergent evolution were identified (Dataset S1, Table S12). One of these genes, NEB, encodes nebulin, an actin-binding protein that is localized to the thin filament of the sarcomeres in skeletal muscle (38). Mutation of NEB is related to muscle weakness and nemaline myopathy in humans (39). Our results show that NEB was highly expressed in skeletal muscle in pygmy loris (SI Appendix, Fig. S1H). Furthermore, we identified seven nonsynonymous amino acid substitutions that were common to pygmy loris and sloth, which occurred at sites that were conserved in nine other mammals (Fig. 3C). Given its important role in muscle, we propose that these convergent substitutions in the NEB gene may well be related to slow movement in both pygmy loris and sloth. Interestingly, we found that five of the seven convergent amino acid positions also display substitutions in the Sunda slow loris (SI Appendix, Fig. S2C).

Genetic Mechanism Underlying Abnormal Circadian Rhythm in Pygmy Lorises.

Low temperatures during the winter months can result in high energetic costs for thermoregulation in pygmy lorises (40). During the cold periods, pygmy lorises can enter a state of torpor (i.e., short hibernation), which can last up to 63 h (21). This ability to hibernate is unique in primates outside of Madagascar and is in all likelihood associated with changes in circadian rhythmicity (23). Here, we identified five genes (PER2, ENOX2, PTGER3, PROX1, and SIX3) that are involved in the regulation of circadian rhythm and that have also been under positive selection in pygmy lorises. Genes in the period (PER) gene family encode components of the circadian clock, which regulates daily rhythms of locomotor activity, metabolism, and behavior (41). PER2 is expressed in a circadian pattern in the suprachiasmatic nucleus, the primary circadian pacemaker in the mammalian brain (42). Mutations in PER2 can shorten, extend, or even disrupt circadian rhythms (43). PER2 forms complexes with cryptochromes (CRY), which are required for photic entrainment of circadian rhythm and can negatively regulate the CLOCK/BMAL1-dependent transactivation of clock and clock-controlled genes (44). Here, we found multiple substitutions in pygmy loris PER2, of which S1271A was predicted to affect the binding domain between PER2 and CRY based on three-dimensional structure simulation (Fig. 3 D and E). As expected, a functional coimmunoprecipitation assay confirmed that the ability to bind with CRY was weakened in both pygmy loris PER2 and S1271A-mutated human PER2 compared to human PER2 (35.181 ± 3.867 in pygmy loris; 63.269 ± 1.20 in human) (Fig. 3F). In conclusion, our findings suggest that the binding of PER2 to CRY1 has been weakened in pygmy lorises, and this may be associated with the unusual circadian rhythmicity in pygmy lorises.

Population Genomics of Lorises.

Lorises are listed as either Vulnerable, Endangered, or Critically Endangered by the International Union for Conservation of Nature (IUCN). Genomic variation is important for inferring demographic history and helpful for conservation management (45, 46). Here, we resequenced the whole genomes of 50 pygmy lorises and six BSLs (N. bengalensis) (Dataset S1, Table S13), which have an overlapping geographic distribution (Fig. 4A). By mapping to the reference genome, we identified 36.3 million high-quality single nucleotide polymorphisms (SNPs) across the two species. A number of analyses, including phylogenetic trees based on nuclear genomes and mtDNA (SI Appendix, Fig. S3 A and B), principal component analysis (PCA) (SI Appendix, Fig. S4A), and population structure by ADMIXTURE (SI Appendix, Fig.S4B), supported separation of the two species. Historical fluctuations in effective population size (Ne) reconstructed by the pairwise sequential Markovian coalescent (PSMC) model (47) (Fig. 4B and SI Appendix, Fig.S4C) and relative cross-coalescence rate by the multiple sequentially Markovian coalescent (MSMC) model (Fig. 4C) showed that pygmy lorises and BSLs diverged more than millions of years ago. Consistently, Nekaris and Nijman (6) reported that there are marked differences between the pygmy loris and the other slow loris (Nycticebus), and proposed pygmy loris as a new genus (X. pygmaeus). In addition, pygmy lorises and BSLs have exhibited an inverse relationship in terms of the change in effective population size over the past 1 million y, which may have been caused by spatial competition for the same ecological niche and high-level habitat sharing in Southeast Asia (Fig. 4A). The effective population sizes of the two species have been dramatically reduced over the last 60 kyr (Fig. 4B), suggesting that the reduction in biological diversity has been ongoing for a considerable time, possibly due to human migration or the cold climate during the last glaciation.

Fig. 4.

Evolutionary history of lorises. (A) Geographic distribution of X. pygmaeus and N. bengalensis. Loris pictures are copyrighted by Stephen D. Nash/IUCN/SSC Primate Specialist Group, and are used with their permissions in this study. (B) Population demographic history inferred from sequences of X. pygmaeus and N. bengalensis using the PSMC method. (C) Divergence time between X. pygmaeus and N. bengalensis inferred by means of MSMC method.

Because of high-level habitat sharing, we next investigated the pattern of genomic differentiation between pygmy lorises and BSLs with regard to sympatric speciation. Genomic divergence between pygmy lorises and BSLs was most pronounced at the ends of the chromosomes (i.e., the telomeric regions) (SI Appendix, Fig.S4D). This is similar to previous observations on the genomic landscape of species divergence in birds (48) and is probably due to the conserved role of telomeres during meiosis, which cluster on the nuclear envelope to form a “telomere bouquet,” enabling chromosome movements to promote homologous synapsis. Our findings suggest that rapid evolution of the telomeric region may have driven speciation in lorises.

Discussion

Our knowledge of strepsirrhines, an extant group of primates occupying a key node in primate phylogeny, still remains limited. Previous studies of strepsirrhines have been largely confined to macroscopic exploration (13, 23). Here, we present a chromosome-scale genome assembly for the pygmy lorises (Xanthonycticebus gen. nov.). Comparative genomics combined with functional arrays disclosed the rapid evolution of PITRM1, MYOF, and PER2, associated with the low metabolic rate, slow movement, and unusual circadian rhythm of pygmy lorises, respectively.

Pygmy lorises have a much lower metabolic rate relative to other eutherian species of similar body mass, which is thought to be related to the high concentrations of toxins and digestion inhibitors in their largely exudate diet (8–11). Our findings show that the GSTA gene family expanded significantly in the pygmy lorises, which is likely to be capable of enhanced detoxification of secondary metabolites from their diet. While pygmy loris venom is harmful to other animals, pygmy lorises are resistant to their venom. Moreover, there is also evidence that pygmy lorises are resistant to toxins of snakes (19), probably due to the enhanced detoxification properties of the pygmy lorises. We found that pygmy loris PITRM1 exhibited a decrease in enzyme activity relative to that in humans. PITRM1 is a highly conserved zinc metalloprotease that plays an important role in metabolism (33). Knockdown of PITRM1 in mice is reported to decrease basal metabolism and oxygen consumption (34). Therefore, a decrease of PITRM1 is likely to be functionally associated with a low metabolic rate.

In addition to pygmy lorises displaying slow movement and locomotion in relation to other strepsirrhines, they have a muscle fiber composition that differs from the general condition evident in mammals (12), being characterized by a predominance of slow-twitch muscle fibers. These fibers are slow to contract and consume less energy than the white muscle fibers found in most primates; in this respect, they are better suited to tasks that require endurance, such as sustained locomotion (13–15). We noted rapid evolution of MYOF in pygmy loris, and convergent evolution of the NEB gene between pygmy loris and sloth, which may well be responsible for the unique muscle fiber composition in pygmy loris.

Finally, it should be emphasized that slow lorises as well as pygmy lorises are a highly threatened group of Asian mammals, the threats being primarily due to forest loss and illegal trade (49). Genetic rescue through translocation and the introduction of variation into imperiled populations has been proposed as a powerful means to preserve, or even increase, the genetic diversity and hence the evolutionary potential of endangered species (50–52). Based on our large-scale population genomics analysis, we offer insights into both the historic and contemporary population dynamics of lorises, which provide a theoretical framework to support conservation efforts pertaining to this very special strepsirrhine species.

Methods

Sampling and DNA and RNA Extraction.

The skeletal muscle samples from an adult female pygmy loris used for de novo genome assembly were provided by the Animal Branch of the Germplasm Bank of Wild Species, Kunming Institute of Zoology, Chinese Academy of Sciences. Eight different tissues—including liver, spleen, stomach, kidney, brain, muscle, heart, and tongue—were used for transcriptome analysis. Sample collection was performed in accordance with the methods approved by the Institutional Animal Care and Use Committee at the Kunming Institute of Zoology (ID: SMKX-20200118-06). Genomic DNA was extracted from skeletal muscle using a DNeasy Tissue Kit (Qiagen) according to the manufacturer’s recommended instructions. The quality and quantity of the extracted DNA were measured using a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific), Qubit 2.0 fluorometer (Thermo Fisher Scientific), and agarose gel electrophoresis. Total RNA was extracted from the eight pygmy loris tissues using a RNeasy Plus Universal Kit (Qiagen). The spectrophotometer and fluorometer were used to measure the quality and quantity of extracted RNA, respectively.

PacBio Genome Sequencing and Assembly.

Single-molecule sequencing was performed by the PacBio Sequel System. After filtering out adaptor sequences, low-quality reads, and duplicate reads, a total of 136.61 Gb of PacBio sequence data were generated. Read N50 was ∼14,136 bp, the longest read length was 169,536 bp, and the average length was 8,729 bp. We next used Canu (53) to assemble the genome with the PacBio reads and obtained a 2.89-Gb genome with a contig N50 of 1.51 Mb. We used WTDBG (https://github.com/ruanjue/wtdbg) to assemble the pygmy loris genome, which was 2.82 Gb in size with a contig N50 of 3.84 Mb. The Canu and WTDBG assembly results were merged and optimized using Quickmerge (54). We further generated 60.82 Gb of short-read sequences by Illumina HiSEq. 2000 and polished the assembly with Pilon v1.22 (parameters “–mindepth 10 –changes –threads 4 –fix bases.”) (55). The final pygmy loris genome was 2.89 Gb in size, with a contig N50 of 4.73 Mb.

Hi-C Library Preparation and Sequencing.

Hi-C libraries were prepared as described previously (56) using pygmy loris skeletal muscle tissue from the above individual. Briefly, under adsorption of avidin magnetic beads, DNA with biotin was captured, and the Hi-C library was constructed through a series of steps, including the terminal repair of DNA fragments, joint connection, evaluation of PCR amplification, and library purification.

Qubit 2.0 was used for preliminary quantification. An Agilent 2100 bioanalyzer was used to check the insert size of the library. Libraries that passed inspection were sequenced using the Illumina HiSeq PE150 platform according to their effective concentrations. Thus, a total of 254-Gb sequence data were obtained.

The original pygmy loris genome sequence draft was 2.89 Gb in length, with a contig N50 of 4.74 Mb and 3,858 contigs in total. After Hi-C–assisted genome assembly, the genome size was 2.78 Gb, contig N50 was 4.7 Mb, super scaffold N50 was 129.9 Mb, and anchor rate of contig length was 96.30%.

Evaluation of Assembly Results.

Genome completeness was assessed by three methods: 1) Benchmarking universal single-copy orthologs (BUSCO) evaluation, resulting in a BUSCO-estimated pygmy loris genome integrity of 89.06% (complete BUSCOs/total BUSCOs) (Dataset S1, Table S14) (57); 2) core eukaryotic genes mapping approach (CEGMA) evaluation (58), whereby assembly completeness was assessed by mapping 248 conserved core eukaryotic genes (CGEs) and 244 highly conserved CGEs to the genome (Dataset S1, Table S15); and 3) comparative analysis with second-generation reads, with 98.21% of the reference genome mapped to the short sequences (59).

Repeat Annotation.

Because of the relatively low level of conservation of repeat sequences between species, the prediction of repeat sequences for the pygmy loris required the construction of a specific repeat sequence database. Therefore, based on structure and de novo predictions, we used LTR_FINDER v1.05 (60), MITE-Hunter (61), RepeatScout v1.0.5 (62), and PILER-DF v2.4 (63) to construct a repeat sequence database for the pygmy loris genome. PASTE Classifier (64) was used to classify the database, which was then merged with the Repbase (65) as a final repeat sequence database. We next used RepeatMasker v4.0.6 (66) to predict repeat sequences based on the established repeat sequence database, yielding ∼1.54 Gb of repeat sequence.

Genome Annotation.

Protein-coding genes were predicted using ab initio prediction, homologous species prediction, and UniGene prediction. GENSCAN (67), Augustus v2.4 (68), GlimmerHMM v3.0.4 (69), GeneID v1.4 (70), and SNAP v2006-07-28 (71) were used for ab initio prediction; GeMoMa v1.3.1 (72) was used for predictions based on related primate species, including Callithrix jacchus, Rhinopithecus bieti, Homo sapiens, and Macaca mulatta; PASA v2.0.2 (73) was used to predict UniGene sequences based on full-length transcriptome data without reference assembly. EvidenceModeler (v2012-06-25) was used to integrate all predicted genes of different strategies (74) with parameters ‘Mode: STANDARD, S-ratio: 1.13 score > 1000’ and the following weight values: PROTEIN OTHER 50, PROTEIN GeMoMa 50, TRANSCRIPT assembler-PASA 50; ABINITIO PREDICTION Genscan 0.3, ABINITIO PREDICTION Augustus 0.3, ABINITIO PREDICTION GlimmerHMM 0.3, ABINITIO PREDICTION SNAP 0.3, ABINITIO_PREDICTION GeneID 0.3, and OTHER PREDICTION OTHER 100. PASA was used to modify the final gene models (73). Finally, we obtained 26,200 genes (Dataset S1, Table S6). Annotation of the predicted genes was performed by BLASTing the gene (and predicted protein) sequences against several nucleotide and protein sequence databases, including COG (75), Kyoto Encyclopedia of Genes and Genomes (76), NCBI-NR, and Swiss-Prot (77), with an E value cutoff of 1e⁻⁵ used to assess orthology.

Based on the Rfam (78) and miRBase (79) databases, we used Infernal v1.1 (80) to predict ribosomal RNA (rRNA) and microRNA (miRNA) sequences and used tRNAscan-SE v1.3.1 (81) to identify transfer RNA (tRNA). We annotated 1,457 genes encoding long noncoding RNAs (lncRNAs), which were subdivided into 581 miRNAs, 539 rRNAs, and 337 tRNAs.

Phylogenetic Tree Construction and Evolution of Gene Families.

Coding sequences (CDS) of 11 species (human, chimpanzee, gorilla, orangutan, gibbon, rhesus macaque, marmoset, tree shrew, tarsier, mouse lemur, bush baby) were downloaded from the Ensembl genome browser and, together with that of the pygmy loris, were used to construct gene families. Briefly, we removed CDS with fewer than 30 encoded amino acids, premature stop codons, and nontriplet lengths. TreeFam v7.0 (www.treefam.org/) (82) was used to produce gene family category. The longest translation form was chosen to represent each gene. We obtained 32,250 gene families and 116 single-copy orthologs. CDS from each single-copy family were aligned by MUSCLE v3.7 (https://www.ebi.ac.uk/Tools/msa/muscle/) (83). Then, RAxML was applied to these sequence sets to build phylogenetic trees (84). MCMCTREE in PAML v4.4 (35) was used to infer divergence times with approximate likelihood calculations using the 116 single-copy gene families (84). Expansion and contraction of gene families in the pygmy loris were determined by CAFÉ v2.1 (85). A phylogenetic tree based on the RAxML results was used in CAFÉ to infer changes in gene family size (86).

Positively Selected Genes.

The 8,538 single-copy orthologous genes among human, chimpanzee, macaque, marmoset, mouse lemur, pygmy loris, mouse, and horse were identified by TreeFam v7.0 (www.treefam.org/) (82). The single-copy orthologous gene sequences across the nine species were aligned by MUSCLE (v3.8.31) (https://drive5.com/muscle5/), and low-quality aligned regions were further trimmed by Gblocks (v0.91b) with default parameters. The aligned genes with CDS lengths <100 bp were removed for downstream evolutionary analyses. Based on a reliable constructed species-guided tree topology, the branch-site model in PAML (v4.4) (abacus.gene.ucl.ac.uk/software/paml.html) and the likelihood rate test were utilized to detect positively selected genes in the pygmy loris based on the single-copy orthologous genes with P < 0.05 (χ² test, corrected for multiple testing by the false discovery rate). The Bayes empirical algorithm was applied to calculate the posterior probabilities for inferred positively selected sites.

Convergent Evolution.

The one-to-one orthologous genes among human, chimpanzee, macaque, marmoset, mouse lemur, mouse, horse, cow, opossum, pygmy loris, and koala (or sloth) were identified by TreeFam (82). The single-copy orthologous gene sequences across the 11 species were aligned by prank (v3.8.31) (87), and low-quality aligned regions were further trimmed by Gblocks (v0.91b) with default parameters (88). The aligned genes with CDS lengths <100 bp were removed for downstream analyses. The ancestral amino acid sequences were reconstructed for orthologs using the codeml program in the PAML package (v4.7) (35) under the accepted phylogenetic tree. Convergent sites were identified where amino acid positions were identical in the pygmy loris and koala (or sloth) but were different from that of their respective most recent common ancestor. Then a PERL program that was developed by our laboratory was used to calculate the number of parallel amino acid replacements for a specified pair of branches (89). If the posterior probability of the reconstructed ancestral amino acid site was more than 90%, then they were retained and their states were deemed to be reliable.

Transcriptome Analysis.

A paired-end sequencing library was constructed from poly (A)+ RNA, as described in the Illumina manual, and sequenced on the Illumina HisEq. 2000 sequencing platform. For each sample, ∼5 Gb of data were generated, which were deposited in the Genome Sequence Archive (GSA) database (https://ngdc.cncb.ac.cn/gsa/) under accession ID CRA003461 for transcriptomes. RNA-sequencing reads were aligned with HISAT2 v2.0.3 (90), and gene-expression levels were quantified with Cufflinks 2.2.1 (91). FPKM (fragments per kilobase per million) values of a tissue that were twofold higher than in other tissues and had an adjusted P < 0.05 were identified as tissue-specific highly expressed genes.

Measurement of PITRM1 Enzyme Activity.

Total RNA from HEK293T cells was synthesized in order to clone the human PITRM1 CDS. The PITRM1 CDS of pygmy loris was then derived by in vitro mutagenesis from the human PITRM1 CDS (K642R, T707S, D963E) on the basis that the slow loris and human differ in terms of three PRTIM1 substitutions (K642R, T707S, D963E). The human and pygmy loris PITRM1 CDSs were flanked with XhoI and BamHI restriction sites at their 5′ and 3′ ends via PCR using primers 5′-tgctcgagATGTGGCGCTGCGGCGGGCGGCA and 3′-ggatccTCATTGGATGATCCAGGATGGGTC. The CDSs were then cloned into the pTXB-6xHis expression vector to express 6xHis fusion proteins. Plasmids were transduced into Escherichia coli BL21 (DE3) competent cells (Sangon), monoclone cultured to an OD (optical density) 0.6. Proteins were purified using a His-tag Protein Purification Kit (Beyotime #P2226) following the manufacturer’s instructions. Proteins were stored in elution buffer (50% glycerol, NaCl 800 mM, Tris⋅HCl 25 mM, DTT 1 mM) at −80 °C and determined using the Bradford protein concentration assay (ThermoFisher #23200). The Aβ-derivative fluorogenic substrate Mca-KLVFFAEDK(Dnp)-OH was synthesized as described previously (92). Fluorescence resonance energy transfer assays for human PITRM1 and slow loris PITRM1 were performed as described previously (92). Briefly, each assay contained 2 nM enzyme and 10 μM substrate in 50 mM potassium phosphate buffer, pH7.0. The enzymatic assay was performed using a BioTek Synergy HT plate reader (excitation: 350 nm, emission: 405 nm; BioTek).

Functional Validation of MYOF in Unusual Pygmy Loris Musculature.

Neonatal male mice (within 48 h after birth) were killed by cervical dislocation and disinfected with 70% ethanol, after which muscle tissue was taken for experimentation. Muscle tissues were cut into pieces and digested with collagenase at 37 °C for 60 min. Cell suspensions were successively passed through 100-, 70-, and 40-μm cell filters, and centrifuged at 300 × g for 5 min. The precipitates were resuspended with complete medium and inoculated into six-well plates for 30 min at 37 °C. After the supernatant had been transferred to the new well and repeated three consecutive times, it was placed in Petri dishes coated with poly-d-lysine and cultured at 37 °C. The medium was changed every 2 d.

The plasmids CMV-MCS-polyA-EF1A-zsGreen-sv40-puromycin with slow loris and human MYOF coding sequences were transfected. At the logarithmic phase, the mixture was added to the culture dish with LIPO2000 transfection reagent according to the manufacturer’s protocol. After incubation for 12 h, fresh complete medium was replaced. The transfection efficiency was observed the next day. qPCR was used to detect the relative expression levels of MyHC2a, MyHC2b, MyHC2x, and MyHC-slow before and after transfection.

Per2 Functional Assays.

PER2(Human)-3×HA, PER2(Human, S1155A)-3×HA, and PER2(Loris)-3×HA were digested with BamHI and XhoI restriction enzymes. CRY1(Human)-3×FLAG was digested with KpnI and NotI restriction enzymes. PER2(Human)-3×HA, PER2(Human, S1155A)-3×HA, PER2(Loris)-3×HA, and CRY1(Human)-3×FLAG were individually subcloned into the pCDNA3.1 vector.

The HEK293Tells were collected and lysed in RIPA buffer containing protease inhibitors (mixture). In total, 5% of the cell extracts were retained for input, while the remainder were incubated with mouse IgG (ABclonal, #AC005) and anti-FLAG antibodies (ABclonal, #AE004) overnight at 4 °C. For immunoprecipitation of protein complexes, cell extracts were precleared with protein-A/G beads (88803# Thermo Scientific) and incubated with the indicated antibodies for 2 h at 4 °C. The beads were washed three times with RIPA buffer, and the bound proteins were eluted by boiling in 1× loading buffer and subjected to immunoblotting with the indicated antibodies.

Resequencing the Pygmy Loris and BSL Genomes.

We resequenced the whole genomes of 56 individuals, including 50 pygmy loris and 6 BSL based on the existing sample information we have (Dataset S1, Table S13). The majority of slow lorises came from Yunnan province, a southwest region of China. Only two of the slow lorises came from Vietnam. All of the slow lorises in this study were sampled from the wild and taken after natural death; a few samples came from zoos. Samples were stored in the −80 °C refrigerator. Combining information from 16S rRNA and COI (cytochrome c oxidase subunit I) genes, as well as morphological feature of the slow loris, we distinguished the BSL and pygmy loris species. The muscle and skin tissue of slow loris were used for genome resequencing. In total, 10 μg genomic DNA, prepared by the standard CTAB (cetyltrimethylammonium bromide) extraction protocol, were used to construct libraries with a 350-bp insert size. The sequence libraries were constructed according to the Illumina library preparation pipeline on the Illumina HisEq. 4000 platform, generating 150-bp paired-end reads.

All clean short reads were aligned against the assembled pygmy loris reference genome using BWA-MEM v0.7.12 (59). The aligned BAM files were sorted, and PCR-duplicated reads were removed using SAMtools v1.3.1 (93). RealignerTargetCreator and IndelRealigner in the Genome Analysis Toolkit (GATK) v3.7.0 (94) were used for local realignment around indels. SNPs were genotyped using GenotypeGVCFs from the GATK package with default parameters after calling GVCFs (genomic VCFs) in HaplotypeCaller. Finally, we applied the following criteria to all SNPs for hard filtering: QUAL < 30, QD < 2.0, MQ < 40.0, MQRankSum < −12.5, ReadPosRankSum < −8.0, and SOR > 4.0. Genotypes were imputed and phased with BEAGLE v4.1 (95).

Population Genetic Analysis.

A neighbor-joining tree was constructed using the APE (96) package from the pair-wise identical-by-state distance matrix calculated using PLINK 1.9 (97) and based on autosomal SNPs. After pruning the SNPs based on linkage disequilibrium using PLINK, we performed genetic structure clustering of autosomal SNPs using ADMIXTURE (98) and PCA with smartpca in EIGENSOFT v5.0.2 (99).

Demographic Analysis.

To reconstruct the demographic history of slow lorises, we applied the PSMC method based on high-depth (>20×) sequenced individuals. The parameters were set to “-N25 -t15 -r5 -p 4 + 25x+4 + 6”. We applied 100 bootstraps by randomly sampling and replacing 5-Mb fragments of the consensus sequence. To infer more recent demographic history and divergence time between BSLs and pygmy lorises, we applied MSMC v2.0 (100) to multiple individuals based on default parameters. MSMC input was created from all site VCF files genotyped from the GVCF files. MSMC was run on a four-haplotype model by randomly selecting individuals. For all demographic analyses, the neutral mutation rate (μ) was set to 2.2 × 10⁻⁸ mutations per site per generation (101) and generation time (g) was set to 5 y.

Population Divergence.

We used the genetic differentiation index F_ST and absolute divergence Dxy to identify highly differentiated regions for the BSL and pygmy loris species. We used vcftools v0.1.12 (102) to calculate average F_ST and θπ with 50-kb sliding windows and a step size of 25 kb. Dxy was calculated using an in-house python script. We defined potentially selective windows with the top 5% of values for F_ST and Dxy, and two-tailed top 5% of values for the θπ-ratio. The outlier windows were annotated using snpEff (103), and gene ontology (GO) functional enrichment analysis was performed using the topGO package v2.40.0 (104) with Fisher’s exact test.

Mitochondrial Genome Analysis.

The Perl script NOVOPlasty v2.7.2 (105) was used to assemble individual mitochondrial genomes from paired-end short reads. The complete pygmy loris and BSL mitochondrial genomes (GenBank accession nos. KX397281 and MG515246) were used as seed and reference sequences for pygmy loris and BSL mitochondrial genome assembly, respectively. The k-mer parameter was set to 35, unless assembly failure occurred, in which case it was reduced to 23 to 31. Only complete circularized mitochondrial genomes were kept for downstream analysis. We downloaded all published complete mitochondrial genomes for slow lorises, with the slender loris being used as the outgroup, from GenBank. Clustal Omega v1.2.0 (106) was applied for multiple sequence alignment after rotating the mitochondrial sequences using Cyclic DNA Sequence Aligner (107). Finally, we built the maximum-likelihood tree based on whole mitochondrial sequences using the Tamura-Nei model (108) in MEGA v7 (109).

Data, Materials, and Software Availability

Genome assemblies, DNA-sequencing data, and RNA-sequencing data have been deposited into the GSA database, https://ngdc.cncb.ac.cn/gsa/ (Project no. PRJCA003786) (111). The de novo genome of pygmy lorises is under accession ID GWHBCHX00000000. The Resequenced genomes of pygmy lorises and BSLs are under accession ID CRA003477. Transcriptome data were deposited in the GSA database (accession ID CRA003461).