Ancient mitochondrial genomes from Chinese cave hyenas provide insights into the evolutionary history of the genus Crocuta

(a) Sampling

We collected two Late Pleistocene cave hyena skull fossils from the Songhua riverbed, Harbin, Heilongjiang Province, northeastern China (figure 1 and electronic supplementary material, figure S1). Two samples are stored in Zhaoyuan Museum, Daqing, Heilongjiang Province (ZYB:3149(CADG508) and ZYB:3148(CADG510)). One skull (CADG508) was crude, while another (CADG510) was covered by nitrocellulose lacquer, a chemical used for fossil preservation. To exclude dating interference from the lacquer, we selected the tooth root as dating material, which was embedded in the jaw and not affected by the chemical coating. Radiocarbon dating by accelerator mass spectrometry (AMS) was performed at the Beta Analytic Testing Laboratory in the US. The outer layers for both samples were removed before measurements. Calibrated by using the IntCal13 curve [14], CADG508 and CADG510 were dated to 20 427–20 052 years before present (calBP) (lab number: Beta-539399) and 20 450–20 055 calBP (lab number: Beta-539400) (electronic supplementary material, figure S2). Teeth without chemical coating from these samples were selected for further genetic investigations.

• Download figure
• Open in new tab
• Download PowerPoint

(b) DNA extraction and library preparation

DNA extraction and library preparation were performed in a dedicated ancient DNA laboratory at China University of Geosciences (Wuhan). Before extraction, we physically cracked the tooth and selected its inner part to exclude potential contamination on the surface. Approximately 200 mg of tooth powder of each sample was digested in 4.5 ml of EDTA (0.5M, pH = 8) and 0.06 ml of Proteinase K (20 mg ml⁻¹) for 16 h in a rotating hybridization oven at 37°C. After centrifugation at 7000 rpm for 10 min, we transferred the supernatant into an ultrafiltration tube (CentriconYM-10, Germany) and reduced it to 100 µl at 7000 rpm. Finally, DNA was purified using the QIAquick PCR Purification Kit (Qiagen, Germany) and eluted in 80 µl following the manufacturer's instructions.

A volume of 20 µl extract was used for double-stranded DNA library construction following the protocol described by Meyer & Kircher [15]. The fragmentation step was omitted as DNA from ancient samples is already highly degraded. NEB buffer 2 and BSA (New England Biolabs) were used in blunt-end repair and a 1 : 20 adapter diluted with Quick Ligase buffer (New England Biolabs) was mixed with templates in the Adapter Ligation Step. Additionally, Isothermal buffer (New England Biolabs) was chosen for Adapter Fill-in. Indexing PCR amplifications were performed using AccuPrime Pfx DNA Polymerase (Life Technologies) and dual primers (P5 and P7 for Illumina sequencing platform) under the following conditions: 95°C for 2 min and 17 cycles of 95°C for 15 s, 60°C for 30 s and 68°C for 30 s. For library pooling, the quality and concentration analyses were conducted with Qubit 2.0 (Invitrogen, USA) and an Agilent 2100 Bioanalyzer (Agilent, USA). All clean-up procedures were performed with the QIAquick PCR Purification Kit (Qiagen, Germany) and both extraction and library blank controls were set up to monitor potential contamination. The final sequencing was conducted on an Illumina HiSeqX10 platform at Annoroad Inc, China, producing approximately 205 million and 238 million 150 bp paired-end reads for CADG508 and CADG510, respectively.

(c) Data analysis

Illumina adapter sequences were trimmed from raw reads with Cutadapt v. 1.4.2 [16] and reads shorter than 30 bp were discarded. Flash v. 1.2.11 [17] was used to merge overlapping read pairs. The trimmed reads were mapped to the complete mitochondrial genome of a European cave hyena (GenBank accession number: JF894379) using the Burrows-Wheeler Aligner 0.6.2 [18] with the aln algorithm, seed turned off (-l 999), and otherwise default parameters. To test for endogenous nuclear DNA content, we also mapped the trimmed reads to a nuclear genome of the extant spotted hyena (GenBank accession number: GCA_008692635.1) using identical software and parameters. Sequences with a map quality score less than 30 were excluded by using ‘view’ and the alignment was sorted by 5’ mapping position using ‘sort’ in SAMtools v. 0.1.19 [19]. PCR duplicates generated during library amplification were removed using ‘rmdup’ in SAMtools. The final mitochondrial consensus sequence was generated based on maximum effective base depth [20] using ‘doFasta 3’ in ANGSD v. 0.916 [21], discarding mapped reads and bases with quality scores below 30. Read coverages across the reference were calculated using Qualimap v. 2.2.1 [22] and the authenticity of ancient DNA was tested using mapDamage2 [23], both with default parameters.

To identify the phylogenetic position of the two Chinese hyenas among the Family Hyaenidae, we retrieved 44 mitochondrial genomes from GenBank: 15 brown hyenas (P. brunnea) [24], one striped hyena (H. hyaena) [25], 23 cave/extant spotted hyenas (C. spp.) [6,25,26] (figure 1) and five aardwolves (P. cristata) [27] (electronic supplementary material, tables S1 and S2). Sequences were aligned with the software package Geneious Pro.5.3.4 [28] using the ClustalW algorithm [29]. The D-loop region was removed before further analysis due to its high interspecific variation and ambiguous alignment. We manually annotated the alignment for rRNA, tRNA, replication origin and protein-coding regions, to find suitable data partitions and their corresponding substitution models using PartitionFinder v. 1.1.1 [30] with the greedy search algorithm and linked branch lengths (electronic supplementary material, table S3). With four partitions defined, we then computed a maximum-likelihood (ML) tree using RAxML-HPC v.8 [31] on the CIPRES portal [32]. As RAxML allowed for only a single model of rate heterogeneity in partitioned analyses, we specified the GTRCAT substitution models for each partition, which approximates the GTR + G model suggested by PartitionFinder for most of the partitions (electronic supplementary material, table S3). In view of the deeply diverging evolutionary position of the genus Proteles in Hyaenidae indicated by previous studies [3,8,27], no extra outgroup was chosen for this analysis. For this ML analysis, we conducted 500 bootstrap replicates to assess the robustness of clade support.

Another set of partitions (n = 4) and relevant substitution models calculated using PartitionFinder were defined for BEAST analysis [33] with identical parameters, except for alternative substitution models that are compatible with this software (electronic supplementary material, table S3). Following this, we molecularly dated five sequences from individuals not directly radiocarbon dated (Ccsp015, Ccsp040, CC8, CC9 and GAO1; see electronic supplementary material table S2) using the BEAST v. 1.8.3 software package [33]. Firstly, the divergence age between Crocuta and Parahyaena/Hyaena from the fossil record was used to calibrate the phylogeny with a mean of 9.5 Ma (normal distribution, 95% interval of 9–10 Ma) [2,34]. Secondly, cave hyena specimens with direct dating ages were converted into ‘calibrated time before present’ (calBP) using the IntCal13 curve in Oxcal v. 4.4 online software [14,35]. The age of these samples was set as the mean of the calBP in tip dates (electronic supplementary material, table S2). Finally, we specified the priors of age as normal distribution ranging from 10 to 126 ka for sequences to be molecularly dated, since all cave hyenas included in this analysis were assumed to fall within the Late Pleistocene [6]. Meanwhile, we specified a strict molecular clock, and a constant population size (prior distribution: 1/X; initial value: 5.3 × 10⁵) in parameter settings. The Markov chain Monte Carlo (MCMC) chain ran for 70 million iterations with posterior samples drawn every 1000 steps. The convergence of the posterior was assessed by effective sample sizes greater than 200 using Tracer v. 1.5 [36] after a burn-in of 10% of all states. After obtaining molecular dating information (electronic supplementary material, table S2), we performed an additional BEAST run in which we included dates of all ancient sequences to estimate the molecular clock and the tMRCA of each lineage using parameters identical to those described above. Following this, we identified the maximum clade credibility (MCC) tree in TreeAnnotator v. 1.8.0 [33], and visualized and graphically edited the tree using FigTree v. 1.4.0 [37]. A median-joining network to investigate the phylogeographical relationships among Crocuta taxa was constructed using Network 10.0.0.0 [38] treating all gaps as ‘complete deletion’. To investigate the mitochondrial genetic diversity among Hyaenidae lineages, we computed within-group mean pairwise distances for cave hyenas, extant spotted hyenas, brown hyenas and aardwolves. This analysis excluded striped hyena due to the limited number of sequences available. To better explain the Network results, we further calculated the average pairwise distances for cave and extant spotted hyenas, assuming they belong to distinct lineages. Also, we defined five groups based on mitochondrial haplogroups/sub-haplogroups and computed the within-group mean distances for them to compare the genetic diversity for sub-lineages within Crocuta. All above distance calculations were performed using MEGA 7.0.14 [39], treating gaps and missing data as pairwise deletions to retain maximum homologous sequences. Based on the average mitochondrial mutation rate in Hyaenidae suggested by our previous Bayesian analysis, we specified the clock rate with a mean of 8.87 × 10⁻⁹ substitutions per site per year (normal distribution, 95% CI 8.01 × 10⁻⁹–9.76 × 10⁻⁹) to construct the maternal demographic histories for cave hyena, extant spotted hyena and all Crocuta separately using BEAST v. 1.8.3 [33]. The parameters were also the same as previously, except for the substitution model, for which we used HKY + G, as it approximates all best models for different partitions (electronic supplementary material, table S3) and was used to avoid over-parametrization. The final Bayesian skylines were reconstructed using Tracer v. 1.5 [36].

(a) Sequence recovery and DNA authenticity

After adapter trimming and short read removal, a total number of 2115 and 5350 unique reads successfully mapped to the cave hyena mitochondrial sequence (GenBank: JF894379), generating 6.6-fold and 16.6-fold coverages for CADG508 and CADG510, respectively (electronic supplementary material, figure S3). Our samples covered 90% and 93% (15 353 bp for CADG508 and 15 914 bp for CADG 510) of the complete mitochondrial genome (17 138 bp) (electronic supplementary material, table S4). For nuclear genome mapping, the results show that the endogenous DNA contents for both samples were less than 0.1% (electronic supplementary material, table S4) and we therefore did not perform additional analyses using the nuclear genomes. Length of most DNA fragments from our samples ranges from 40 to 60 bp (electronic supplementary material, figure S4) and both samples have significant C to T misincorporation at 5′ end of the reads and G to A misincorporation at 3′ end of the reads (electronic supplementary material, figure S5), indicating the DNA is not from modern contamination and thus authentic.

(b) Phylogenetic analyses

The phylogenetic trees computed using either RAxML-HPC or BEAST v. 1.8.3 both suggest identical topologies among genera in Hyaenidae, which are in line with a previous study [6], supporting an initial divergence of the aardwolf, followed by a subsequent divergence between cave/extant spotted and striped/brown hyenas (figure 2; electronic supplementary material, figure S6). Our result supports the previously determined close relationship between striped and brown hyenas. Within the genus Crocuta, our two individuals form a monophyletic clade within the Asian lineage (Haplogroup D) and are genetically closer to another individual discovered in China (Ccsp015) than to one in Russia (Ccsp041). We further find that, after splitting from the Asian lineage, several individuals from Europe (Haplogroup B) diverge first, followed by a pure African lineage (Haplogroup C) and a mixed lineage of European and African hyenas (Haplogroup A) (figure 2).

• Download figure
• Open in new tab
• Download PowerPoint

Assuming a 9.5 Ma calibration point for the Crocuta and Parahyaena/Hyaena divergence, our dated Bayesian phylogenetic analysis suggests the aardwolf diverged from the other three hyenas approximately 13.74 Ma (95% CI 12.37–15.17 Ma), quite close to previous molecular results which suggest 13.0 Ma (95% CI 10.1–16.4 Ma) [27]. The genera Hyaena and Parahyaena coalesce at 4.57 Ma (95% CI 4.08–5.09 Ma), which is compatible with the fossil records since the earliest putative Hyaena/Parahyaena ancestor was Ikelohyaena abronia dated to approximately 5.2 Ma [40] and a previous study specified 4.625 Ma for this node, estimated from integrating this ancestor with the two earliest Parahyaena fossil records [6]. The Bayesian analysis estimated the tMRCA of Asian Far East cave hyenas/European cave and extant spotted hyena mitochondrial genomes at 1.85 Ma (95% CI 1.62–2.09 Ma). The European haplogroup B diverged at 1.15 Ma (95% CI 0.99–1.31 Ma), while Haplogroups C and A diverged at approximately 0.83 Ma (95% CI 0.71–0.95 Ma). The Far East lineage, Haplogroup D, could be traced to a relatively young within haplogroup common ancestor of 131 ka (95% CI 96–169 ka) (figure 2).

(c) Network analysis

Using a median-joining network analysis of 25 Crocuta mitochondrial sequences with a total number of 688 variable nucleotide positions across a length of 15 455 bp, we find 19 unique haplotypes (figure 3). Only three haplotypes contained more than one individual. The topology of the Crocuta network is consistent with the Bayesian phylogeny (figure 2). Interestingly, three European haplotypes in Haplogroup B are all identical, despite coming from geographically and temporally different samples. Moreover, haplotypes of German individuals are included within both Haplogroups B and A. We also find both extant spotted hyenas and European cave hyenas within Haplogroup A. This phenomenon has been attributed to gene flow based on nuclear genome results [6]. Our samples came from one site in Northeastern China (figure 1) and had extremely close carbon dating ages (20 240 calBP for CADG508 and 20 253 calBP for CADG510), but belong to different haplotypes within Haplogroup D. Longer branches of the two new Chinese samples have not been detected in both phylogenetic tree (figure 2) and network (figure 3), indicating either no damaged bases have been introduced or the introduction of damaged bases is not much of an issue in the analyses, while we kept as many bases as possible by not including a minimum depth control during calling the mitochondrial consensus.

• Download figure
• Open in new tab
• Download PowerPoint

(d) Genetic distance comparisons

Average genetic distances for cave hyenas are twice that of extant spotted hyenas (1.965% compared with 0.923%), while those of aardwolves and brown hyenas are just 0.147% and 0.015%, respectively (table 1). Concentrating only on cave hyenas, the pairwise distances show that individuals from the Far East lineage (Haplogroup D) are genetically distinct from European cave hyenas by 2.7–3.5%, higher than any other pairwise comparison (electronic supplementary material, table S5). Furthermore, despite being treated as three identical haplotypes in the previous Network analysis, the three individuals from Haplogroup B (Ccsp014, Ccsp042 and Ccsp043) differ from each other at two nucleotide positions (0.013%). Similar discrepancies also occur among haplotypes that contain three individuals in Haplogroups A (CC8, CC9 and Ccsp040) and C (801, 815 and 793) (electronic supplementary material, tables S5 and S6). These variations are not seen in the haplotype network analysis, as gaps appeared in these variable positions in other sequences so they were lost by using the ‘complete deletion’ option from the sequence alignment in the network analysis. Mean genetic distances within haplogroups for cave hyenas suggest that Haplogroups A and D harbour similar levels of diversity (0.173% and 0.165%, respectively), while that for Haplogroup B is only 0.013% (table 1). However, this genetic pattern for Eurasian cave hyenas may be revised when more sequences are included.

(e) Demographic analyses

Through BSPs, we uncovered the female demographic histories for cave hyenas, extant spotted hyenas and a combined Crocuta dataset through time (figure 4). Beginning at approximately 270 ka, the maternal population size of cave hyenas underwent a slight decline, followed by an abrupt accelerated decline from approximately 60 ka. In figure 4, the node of the extinction of cave hyenas indicated by a dashed line was about 25–30 ka, while the radiocarbon dating information of our samples are around approximately 20 ka. We suppose this is a deviation between the expected value calculated by the BEAST software and the actual fossil record. By contrast, the maternal population size for extant spotted hyenas remained relatively stable during approximately 270 to approximately 100 ka. Its first decrease began about 100 ka and retained for approximately 40 000 years (from approx. 100 to approx. 60 ka), followed by a population recovery at approximately 30 ka. The second decline for extant spotted hyenas' population size accelerated around the beginning of the Holocene, which contained more frequent human activities than earlier. These differences between cave hyenas and extant spotted hyenas may reflect either the actual evolutionary events or the different sampling schemes we carried out (i.e. only a narrow range of sample ages is available for the late Pleistocene cave hyenas while only modern samples for spotted hyenas are included in our analyses). Combining cave and extant spotted hyenas revealed two declines during the Late Pleistocene: first at approximately 100 ka and second at approximately 20 ka. One transient stabilization period can be seen between approximately 50 ka and approximately 20 ka (figure 4).

• Download figure
• Open in new tab
• Download PowerPoint

(a) Evolutionary relationships and genetic diversity among Hyaenidae

We constructed phylogenetic trees with the mitochondrial genomes of representatives from the four genera within Hyaenidae (figure 1; electronic supplementary material, figure S6). Our results are concurrent with previous studies based on nuclear genes and morphological data [3,41]. However, these results are in contrast to analyses using complete cyt b gene sequences, which indicated Proteles is not in the deeply diverging position but is sister to Crocuta, which is then sister to a clade comprising Parahyaena and Hyaena [3]. These results have since been dismissed due to heterogeneity in base composition of the cyt b gene, which may incur some systematic error and thus fail to reveal the correct phylogeny [42]. Although the mitochondrial genome clarifies the intergeneric relationship of Hyaenidae, it does not suggest a reciprocally monophyletic relationship between cave and extant spotted hyenas, putatively due to introgression of the mitochondrial genome in these lineages [6]. It has been reported that spotted hyenas exhibit a social system similar to many primate societies unlike the other three species in the family Hyaenidae or other carnivores. They live in clans that vary in size and often contain many unrelated individuals [4], which makes gene flow among different populations more likely than in other Hyaenidae or carnivore species.

Although cave hyenas have become extinct, they harboured higher mitochondrial genetic diversity than extant spotted hyenas, brown hyenas or aardwolves (table 1). However, the aardwolves that we analysed here are from a single population, which probably does not reflect their global diversity. Despite this, aardwolves display medium levels of mitochondrial diversity compared to a number of mammalian species represented by individuals from more populations [27]. Since the mitochondrial genetic diversity of cave hyenas is approximately 13 times that of the aardwolves, cave hyenas may have been good at adapting to different environments across Eurasia. During the Pleistocene, complex gene flow between cave and extant spotted hyenas may have had some adaptive consequences, especially for extant spotted hyenas who received a number of cave hyena alleles linked to genes related to central nervous system function [6]. However, we are not sure if the gene flow from cave hyenas has introduced new mitochondrial lineages to extant spotted hyena and vice versa, or if there were replacements of other lineages. Additionally, we note that the least putatively introgressed cave hyena lineage (Haplogroup D), which includes only four individuals, shares similar levels of genetic diversity to that of one European lineage (Haplogroup A) (table 1). The pairwise distance for four Asian cave hyena ranges from 0.119% to 0.228% (electronic supplementary material, table S5), corresponding to four different haplotypes in our network analysis (figure 3). Therefore, we hypothesize that if more Asian samples were sequenced, there is a possibility we would find Haplogroup D to be even more diverse.

(b) Fossil records in China and Asian cave hyenas

In our Bayesian analysis, we observe that the early diverging Far East lineage containing our samples separates from the Crocuta ancestral node around 1.85 Ma (95% CI 1.62–2.09 Ma). However, as the paraphyletic relationship of cave and extant spotted hyenas are most likely to be the result of gene flow between Africa and Eurasia, more deeply diverging mitochondrial lineages supporting the initial divergence between cave and extant spotted hyena may have been lost. Therefore, this 1.85 Ma node may be related to the separation between European and Asian cave hyenas, rather than the divergence of cave and extant spotted hyenas. Moreover, according to the fossil record, the earliest Crocuta in Asia, C. honanensis, is from approximately 2 Ma [11], which is slightly older than the split of Haplogroup D from other Crocuta lineages, although taking the credibility interval of the molecular dating into account, the fossil record and the molecular dating are not contradictory. Crocuta honanensis is most parsimoniously explained as the evolutionary lineage leading to Eurasian cave hyenas. It is also reasonable, given the fossil record as well as mitochondrial and nuclear DNA data, to assume that European cave hyenas originated from Asia [6], as hyena fossils from Europe are much younger than those from Asia. However, the relationship between C. honanensis and the earliest European Crocuta is still unclear. Future palaeoproteomic analyses may have the power to provide insights into these questions by providing molecular data from much older specimens [43].

In China, several fossil sites have been reported to contain cave hyenas dated to the end of the Pleistocene or even early Holocene, such as Xiaonanhai fauna, Anyang City, Henan Province (13 075 ± 220 calBP) [44]; Shenxian cave, Lishui county, Jiangsu province (11 200 ± 1000 calBP) [45] and Shuanglong cave, Jinhua City, Zhejiang Province (7815 ± 385 calBP) [46]. However, the dating of these cave hyena records may need reconsideration since no direct dating is available for these Crocuta fossils. Moreover, some data were retrieved before 1980, and may not have used collagen making them less reliable than dates acquired more recently [47]. For example, a newly estimated ²³⁰Th/²³⁴U date for Shenxian cave is 109–7 ka. While the low boundary of 7 ka represents the latest layer of the cave, the actual date of the specimen excavated from this cave could be much older than previously thought [48]. Our two samples from Heilongjiang province were directly dated to 20 427–20 052 and 20 450–20 055 calBP (electronic supplementary material, figure S2), representing the latest directly dated record for Asian cave hyenas, thus rejecting a much earlier extinction of cave hyenas in the Far East region (approx. 40 ka) suggested by a previous study [47].

(c) Maternal demographic history of Crocuta

In our analysis (figure 4), the maternal population size of cave hyenas experienced a decline beginning approximately 270 ka. It has been reported that cave hyenas in Europe may not have been continuously present throughout the Middle and Late Pleistocene [47], while the amounts of Asian faunas containing cave hyenas decreased from the late Middle Pleistocene to the Late Pleistocene [49]. Therefore, both the fossil record and our genetic analyses support a demographic decline of cave hyenas during the Middle-Late Pleistocene transition. Comparatively, no such population contraction can be observed for extant spotted hyenas, congruent with reports of their fossils being well distributed in northern Africa during the Late Pleistocene [2,50]. Taking our analysis and the fossil records together, we suppose that these differences may reflect the true demographic histories of cave hyenas and spotted hyenas, rather than reflecting different sampling schemes, as mentioned previously.

When combining both spotted and extant cave hyena data, the population size of the genus Crocuta did not contract until approximately 100 ka. While it is problematic to combine data from two continents, the intermingled nature of African and Eurasian Crocuta, suggestive of several episodes of gene flow [6], provides some justification for this approach. Comparing the three BSPs suggests that the combined trajectory represents a more or less additive combination of the two individual graphs, which could argue for independent demographic histories during the investigated time frame of the last 400 000 years in this analysis. This is in line with the fact that the youngest divergence between an African and a Eurasian clade (the node separating subclades A1 and A2) dates to around 400 000 years ago.

With regard to the drivers behind the observed population fluctuations, we suggest that some global events, for example, climatic change or a decrease in the numbers of herbivores were the main drivers for demographic fluctuation of cave and extant spotted hyenas. Putatively due to climatic cooling, Asian cave hyenas appear to have moved to southern areas during the late Middle Pleistocene, for example, towards Taiwan and Thailand [49,51]. We assume that the Pleistocene African spotted hyenas may have been less influenced by cooling events and thus been able to maintain a relatively stable or even slightly increasing population size. Furthermore, based on effective population size uncovered in the genomes of a large consortium of ruminants [52], which could have been a vital food resource for Crocuta, Westbury et al. [6] suggested that cave and extant spotted hyenas may have sustained a similar decline during the Late Pleistocene, which has been supported by our analyses (decline since approx. 100 ka). However, previous PSMC findings suggested a continuous decline for extant spotted hyena until the current date [6], while maternal demographic results revealed a subsequent population recovery during approximately 75–35 ka. This phenomenon may have resulted from the limitation of our genetic marker, which reflects only the maternal evolutionary history.

Stuart & Lister [47] suggested extirpation of cave hyenas at 40 ka in Central Europe and the Russian Far East, and 31 ka in northwest and southern Europe. However, new direct dating has moved both time estimates closer to the present: a German sample (Ccsp043) was dated at approximately 23 798 calBP [6], while our samples from northeastern China were dated at approximately 20 240 and approximately 20 252 calBP (electronic supplementary material, table S4). The latest acceleration for cave hyenas' population decline was from approximately 60 ka and the time of cave hyenas’ extinction is pretty close to the end of the Last Glacial Maximum. Therefore, one factor that led to cave hyena's extinction possibly includes their physiological cold intolerance, which allowed them to sustain glacial periods. Another cause of cave hyenas' extinction may be attributed to competition for food and habitat with other carnivores as well as humans [47]. However, how these factors acted together and which could have been the major driver are still unclear. Our mitochondrial analysis together with the fossil record has provided clues in reconstructing the evolutionary history of the genus Crocuta. More evidence, particularly from the palaeogenomes of northern African and southern Asian samples, as well as combining results from molecular analyses with the latest insights of the fossil record, may further reveal the details of Crocuta evolutionary history and the mechanisms underlying their demographic fluctuation.