A genetic glimpse of the Chinese straight-tusked elephants

Abstract

Straight-tusked elephants (genus: Palaeoloxodon) including their island dwarf forms are extinct enigmatic members of the Pleistocene megafauna and the most common Pleistocene elephants after the mammoths. Their taxonomic placement has been revised several times. Using palaeogenomic evidence, previous studies suggested that the European P. antiquus has a hybrid origin, but no molecular data have been retrieved from their Asian counterparts, leaving a gap in our knowledge of the global phylogeography and population dynamics of Palaeoloxodon. Here, we captured a high-quality complete mitogenome from a Pleistocene Elephantidae molar (CADG841) from Northern China, which was previously morphologically assigned to the genus Elephas (Asian elephant), and partial mitochondrial sequences (838 bp) of another Palaeoloxodon sp. specimen (CADG1074) from Northeastern China. We found that both Chinese specimens cluster with a 244 000-year-old P. antiquus (specimen name: WE) from Western Europe, suggesting that this clade may represent a population with a large spatial span across Eurasia. Based on the fossil record and the molecular dating of both the divergences of different Palaeoloxodon mitochondrial clades and previously determined hybridization events, we propose that this Eurasian-wide WE clade provides evidence for an earlier migration and/or another hybridization event that happened in the evolutionary history of straight-tusked elephants.

1. Introduction

The order Proboscidea was once widely distributed and species rich, but all that remains of it today are the three living elephant species [1]. Among the extinct members, Palaeoloxodon, the straight-tusked elephants, is a genus that left the richest fossil record in Eurasia second only to Mammuthus [2]. Eurasian Palaeoloxodon are thought to be derived from the African P. recki [3], which can be traced back to the early Pliocene and survived until the Middle Pleistocene [4]. After appearing in Eurasia around 0.78 million years ago (Ma) [5], Palaeoloxodon likely diversified into several species, including P. antiquus, P. huaihoensis, P. namadicus and P. naumanni in mainland Eurasia [6,7], as well as some dwarf species inhabiting the Mediterranean islands, such as P. cf. mnaidriensis [8,9]. Palaeoloxodon in Europe became extinct around 34 thousand years ago (Ka) [10], while its extinction time in China is in need of re-evaluation owing to insufficient direct dating information, although a suggestive lower limit is the end of the Late Pleistocene [11,12].

Based on morphological comparisons, Palaeoloxodon was once synonymized with Elephas (Asian elephants), sometimes considered a subgenus of Elephas [1,3]. However, mitogenomic evidence suggested that P. antiquus was a close relative of the extant African forest elephant (Loxodonta cyclotis) [13], while nuclear biparental evidence suggested a three-way hybrid descended from ancestral African elephants, woolly mammoths or Asian elephants, and modern African forest elephants [14]. These findings not only elucidate that Palaeoloxodon should be taxonomically separated from Elephas, but also emphasize the role of hybridization in shaping elephantid evolution. Despite this, the global phylogeographic structure of Palaeoloxodon species remains unknown.

Chinese Palaeoloxodon specimens are morphologically categorized into several different species. For instance, Liu proposed a subspecies (P. naumanni huaihoensis) based on an incomplete large skeleton found in Anhui Province [15], while Qi considered this taxon to encompass all Chinese Palaeoloxodon and elevated it to species level (P. huaihoensis) [7]. Others, however, argued for P. naumanni to be the only correct species name [11]. This issue became more complicated when Li et al. argued that in Northern China, P. namadicus represented an ancestral type while P. naumanni stood for a more derived type [16]. Molecular studies of Chinese Palaeoloxodon are obviously needed to provide additional evidence for species categorization.

In this study, we use ancient DNA methodology and next-generation sequencing to investigate putative Palaeoloxodon individuals across China. We performed phylogenetic analyses for two Chinese samples and estimated the divergence time between one well-preserved Chinese sample and other Loxodonta/Palaeoloxodon lineages. Our results offer new clues into the evolution of the straight-tusked elephants but also raise a number of questions that warrant further investigation.

2. Material and methods

We sampled six potential Palaeoloxodon fossils from China (figure 1 and electronic supplementary material, S1): one from Ruzhou County, Henan Province (CADG930, Henan Natural History Museum), one from Sihong County, Jiangsu Province (CADG984, Sihong Museum), three from Harbin City, Heilongjiang Province (CADG1037, CADG1072 and CADG1074, Daqing Museum), and one from Yangyuan County, Hebei Province (CADG841, Institute of Vertebrate Paleontology and Paleoanthropology). Among them, CADG841 was previously recognized as an Asian elephant and was radiocarbon dated to greater than 50 300 years ago [18], while the other samples were putatively identified as Palaeoloxodon.

Figure 1.

Fossil sites of P. antiquus, P. cf. mnaidriensis and Chinese Palaeoloxodon [9,13,17]. The map draft (chart no. GS(2016)1613) was downloaded from the National Administration of Surveying, Mapping and Geoinformation of China (http://bzdt.ch.mnr.gov.cn).

Detailed DNA extraction, library preparation, pre-sequencing, target-enrichment capture, sequencing and data processing procedures are provided in the electronic supplementary material. Read coverage across the reference was calculated using Qualimap v2.2.1 [19]. The ancient DNA damage pattern was investigated using mapDamage2 [20].

To determine the phylogenetic position of our Chinese Palaeoloxodon, 33 Elephantidae mitogenomes were downloaded from GenBank, including 14 woolly mammoths (M. primigenius), one Columbian mammoth (M. columbi), three Asian elephants (E. maximus), four European straight-tusked elephants (P. antiquus), one Mediterranean dwarf straight-tusked elephant (P. cf. mnaidriensis), seven African forest elephants (L. cyclotis) and three African savannah elephants (L. africana) (electronic supplementary material, table S1). The two newly obtained sequences (CADG841 and CADG1074) and the 33 downloaded sequences were aligned using MAFFT v7.471 [21] on the CIPRES platform [22]. The D-loop region was removed before further analysis using Bioedit v7.2.5 [23] due to possible high intergeneric variation and equivocal alignment. Using the above dataset, a maximum-likelihood (ML) tree was computed using RAxML-HPC v.8 [24] on the CIPRES portal by the following steps: the mitogenome was split into several alignments: protein-coding genes (first, second, third coding positions), transfer RNA (tRNA) and ribosomal RNA (rRNA), and then the best substitution model for each partition was determined using PartitionFinder2 [25] with the greedy search algorithm and linked branch lengths, and the output models were used to compute the ML tree.

The divergence times of different Elephantidae lineages were estimated using BEAST v1.10.4 [26]. To retain maximum length of homologous sequences in Bayesian estimation, we excluded CADG1074 from the dataset due to its short length. After sequence alignment and removal of the D-loop region, PartitionFinder2 was used to select the best substitution models, setting identical parameters as described for RAxML-HPC, except that the optimal substitution models were specified for BEAST (electronic supplementary material, table S2). Two sets of dates were combined to calculate the time of the most recent common ancestor of different Elephantidae lineages: (i) the radiocarbon and stratigraphic dates associated with downloaded mammoth and straight-tusked elephants (electronic supplementary material, table S1); (ii) the divergence age between Loxodonta and Elephas/Mammuthus (with a mean of 7.6 Ma and a standard deviation of 610 000 years with normal distribution) [27]. Assuming a strict molecular clock with proboscidean substitution rate [27], three tree prior models, i.e. constant size, Bayesian Skyline, and birth–death were individually tested, then the Markov chain Monte Carlo model was carried out with 30 million iterations under each tree prior model twice. We inspected the convergence of each parameter using Tracer v1.7 [28]. After discarding the first 10% states as burn-in, sufficient sampling was verified by the effective sample size values of all parameters exceeding 200. The maximum clade credibility tree was output using TreeAnnotator v1.8.4. The tree was visualized in an increasing branch order using FigTree v1.4.3 [29].

3. Results

(a) . Mitochondrial DNA of Chinese straight-tusked elephants

We obtained approximately 6.6 to 48.3 and 18.6 to 41.9 million paired-end reads for pre-sequencing libraries (electronic supplementary material, table S3) and captured libraries (electronic supplementary material, table S4), respectively. When testing different mitogenomic references, more reads were mapped to the P. antiquus WE individual (KY499558) than to both the P. antiquus NN individual (KY499555) and other Elephantidae species (electronic supplementary material, table S5) for the two best preserved samples (CADG841 and CADG1074). After specifying WE as the mapping reference, the final sequence length for CADG841 is 16 103 bp (35.5X), while for CADG1074 we could only reconstruct 838 bp (0.05X) (electronic supplementary material, figure S2 and table S6). The occurrence of DNA damage could not be tested for CADG1074 due to the insufficient number of mapped reads, but the length of the reads recovered was between 35 and 86 bp (mean length: 55 bp), consistent with what is expected for ancient DNA. The mean read length and misincorporation pattern of CADG841 were consistent with its antiquity (electronic supplementary material, figures S3 and S4). All other samples yielded nearly no endogenous DNA even after target enrichment (electronic supplementary material, table S3 and S4).

(b) . Phylogenetic analyses

Phylogenetic analyses suggest that both CADG841 and CADG1074 fall within the diversity of forest elephants (clade F) and form a cluster with the P. antiquus WE individual (244 000 years old) from Western Europe, which is robust when using different tree computing methods (figure 2a,b) or different tree prior models (figure 2b and electronic supplementary material, figure S5). Using Bayesian tip-dating and fossil calibration based on the complete mitogenomes, the coalescent time for mammoth and Asian elephants is around 6.79 Ma (95% CI: 5.88–7.68), while the posterior mean of all mitogenomes here is ca 7.25 Ma (95% CI: 6.41–8.15) (figure 2b and electronic supplementary material, table S7), both congruent with previous molecular studies [13], as well as the presence of the earliest Asian elephant and African elephant fossils suggesting 5.2–6.7 and 5.4–7.3 Ma for the above nodes, respectively [1]. The coalescent time for CADG841 and the WE individual was estimated at 0.39 Ma (95% CI: 0.30–0.49).

Figure 2.

Mitochondrial phylogenetic relationships of Elephantidae. (a) ML tree considering all sites. The node labels represent bootstrap values. (b) Bayesian maximum clade credibility tree based on complete mitogenomes using the constant size model as tree prior. CADG1074 was excluded here due to its short sequence length. The yellow dot indicates the node for time calibration. Blue node bars show the 95% highest posterior density estimates for node ages. The median posterior age estimates and the Bayesian posterior probabilities are shown together with ‘/’ above main nodes. Different clade names were kept consistent with previous studies [13,14,30].

4. Discussion

One of the main challenges in the study of Pleistocene elephants is to distinguish Elephas from Palaeoloxodon based on morphological traits. In this study, we addressed the genetic identity of two Elephas/Palaeoloxodon elephant samples via the sequencing of a complete mitogenome for one (CADG841) and a set of short mitochondrial fragments for the other (CADG1074). Phylogenetic analyses suggested both specimens fall within a clade containing P. antiquus from Western Europe. The lack of information for Palaeoloxodon in its eastern range makes the addition of our two samples valuable for understanding the biogeography of Palaeoloxodon.

A clustering relationship of our samples together with the WE straight-tusked elephant may indicate a genetically connected population across Eurasia. However, this molecular relationship could not be morphologically examined since CADG841 has an Elephas-like morphology while the WE specimen is an undiagnostic fragment that was assigned as P. antiquus according to the strata information [13]. It should be noted that no Palaeoloxodon species has been assumed occupying both Europe and China [2,6]. Thus, our results that are suggestive of a pan-Eurasian Palaeoloxodon population as well as the discrepancy between the molecular and the morphological identification of specimen CADG841 suggest that a thorough investigation of the morphology of all proposed Palaeoloxodon species is warranted.

According to the analyses of nuclear genomes by Palkopoulou et al. [14], the hybridization between the West African forest elephant and the NN Palaeoloxodon clade occurred later than 609 Ka (the upper bound of the intraspecies split time between the West and Central African forest elephants), while the earliest fossil record of Palaeoloxodon in Eurasia dates to about 780 Ka [5]. Taking into account that the hybridization could only have happened in Africa since there are no known Loxodonta specimens found outside of Africa [31,32], the above-mentioned dating gap suggests that there may have been at least another, earlier hybridization event corresponding to the earliest Eurasian members in the evolutionary history of Palaeoloxodon. In that case, Palaeoloxodon would have experienced at least two independent hybridization events, making it a partially polyphyletic taxon, at least on the genomic level. Moreover, at least two migration events of Palaeoloxodon to Eurasia would need to have occurred.

The above hypothesis about multiple hybridization and migration events is consistent with the detection of two distinct Palaeoloxodon clades in the mitochondrial tree, both nested within the diversity of Loxodonta cyclotis, the African forest elephant, although it has to be noted that molecular estimates have their own set of uncertainties. We estimated the divergence time of the different Palaeoloxodon clades at ca 1.22 Ma for the NN clade and circa 2.34 Ma for the WE clade, which are both substantially earlier than the upper bound of 609 Ka when the hybridization event that gave rise to the NN elephants supposedly occurred at the earliest. Although the mitochondrial divergence should predate the timing of hybridization, in both cases even the lower bound for the mitochondrial divergence predates the upper bound of hybridization considerable with 1.01 Ma for NN and 2.01 Ma for WE.

Given the notorious uncertainties of molecular dating, it is, however, also possible that the hybridization event resulting in the NN elephants may have taken place early enough to be also responsible for the origin of the first Palaeoloxodon specimens in Eurasia. Thus, when this single hybridization event took place, several different mitogenomic lineages that diverged long before would have passed from African forest elephants into the straight-tusked elephant population and then become fixed for the NN and WE lineages in different populations, respectively. It is currently not possible to distinguish between these two hypotheses as nuclear data from an individual carrying the WE-mitochondrial lineage would be necessary for doing so. Unfortunately, the biomolecular preservation of both the WE specimen and our best specimen CADG841 was insufficient for obtaining meaningful amounts of nuclear DNA.

In this study, we show that the WE clade of the genus Palaeoloxodon existed not only in Europe but also in China. The two detected clades (WE and NN) in the mitochondrial phylogenetic tree might indicate a single hybridization event between African forest elephants and the ancestors of straight-tusked elephants, followed by lineage sorting or alternatively be the result of separate hybridization and migration events, which would make the taxon Palaeoloxodon not only the result of hybridization but effectively also at least partially polyphyletic. However, we caution that mitochondrial DNA is a single genetic marker which might be greatly affected by incomplete lineage sorting and therefore does not necessarily reflect any hybridization event we mentioned here. Future nuclear studies on the African forest elephant ancestral components in the WE clade will be necessary to resolve this uncertainty and distinguish two hybridization hypotheses.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

The complete mitochondrial genome sequence of CADG841 newly generated in this study is available at GenBank under accession number OP765243. The final bam files of CADG1074 and CADG841 are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.8kprr4xsd [33].

The data are provided in the electronic supplementary material [34].

Authors' contributions

H.L.: conceptualization, data curation, formal analysis, visualization, writing—original draft, writing—review and editing; J.H.: conceptualization, data curation, writing—original draft, writing—review and editing; S.B.: data curation, writing—original draft, writing—review and editing; J.Y.: funding acquisition, project administration, resources, writing—review and editing; X.C.: resources, writing—review and editing; B.X.: investigation, writing—review and editing; S.S.: investigation, writing—review and editing; Z.D.: investigation, writing—review and editing; X.L.: funding acquisition, writing—original draft, writing—review and editing; M.H.: conceptualization, validation, visualization, writing—original draft, writing—review and editing; G.S.: conceptualization, funding acquisition, methodology, project administration, resources, supervision, validation, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was financially supported by the National Natural Science Foundation of China (grant no. 42172027). S.B. is funded by the German Research Foundation (grant no. 42555017). The funder had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.