Relationships of Late Pleistocene giant deer as revealed by Sinomegaceros mitogenomes from East Asia

Summary The giant deer, widespread in northern Eurasia during the Late Pleistocene, have been classified as western Megaloceros and eastern Sinomegaceros through morphological studies. While Megaloceros’s evolutionary history has been unveiled through mitogenomes, Sinomegaceros remains molecularly unexplored. Herein, we generated mitogenomes of giant deer from East Asia. We find that, in contrast to the morphological differences between Megaloceros and Sinomegaceros, they are mixed in the mitochondrial phylogeny, and Siberian specimens suggest a range contact or overlap between these two groups. Meanwhile, one deep divergent clade and another surviving until 20.1 thousand years ago (ka) were detected in northeastern China, the latter implying this area as a potential refugium during the Last Glacial Maximum (LGM). Moreover, stable isotope analyses indicate correlations between climate-introduced vegetation changes and giant deer extinction. Our study demonstrates the genetic relationship between eastern and western giant deer and explores the promoters of their extirpation in northern East Asia.


INTRODUCTION
The giant deer or megacerines (tribe Megacerini, family Cervidae) were among the emblematic mammals distributed widely in northern Eurasia in the Pleistocene.They became extinct in the Holocene under the dramatic effects of late Quaternary climate fluctuations. 1,2Megacerines have been suggested to be a model group for paleoecological studies, 3 yet their evolutionary history is still insufficiently explored.
The first appearance of undoubted megacerines in western Europe is from the Early Pleistocene (1.2 million years ago, Ma), 4,5 while the proposed appearance for Central Eurasia was much earlier (Late Pliocene). 3In contrast to the controversial status of giant deer taxonomy in pre-Pleistocene, and even into the Early and Middle Pleistocene in Europe, researchers agree on a morphology-based division between the Late Pleistocene Megaloceros giganteus of western Eurasia and an array of Pleistocene Sinomegaceros species from eastern Eurasia (Figure 1). 3,6This is based, among other features, on the extraordinarily broad and flattened brow (basal) tine, expanded in most Sinomegaceros species beyond the more modest flattening seen in Megaloceros and unique among living and fossil deer (Figure 2).Up to nine species of Sinomegaceros have been proposed, from Central Asia, China, and Japan 6 (five were recognized by Vislobokova 3 ).
In the western Eurasia group, M. giganteus, one of the largest known species in the subfamily Cervidae, has been a focus of previous studies.The species distribution extended from Ireland to central Siberia, and with an evolutionary history dating from 400 to about 8 thousand years ago (ka). 7,8With regard to its phylogenetic position, rather than indicating a closer phylogenetic relationship to the living red deer (Cervus elaphus/canadensis), 9 Lister et al. 5 and Hughes et al. 10 indicated a sister-group relationship of M. giganteus to the extant fallow deer ll OPEN ACCESS (Dama spp.) based on both morphological and molecular evidence.The cluster of M. giganteus and D. dama/mesopotamica was confirmed by Immel et al. 11 using two near-complete M. giganteus mitochondrial genomes.Recently, a decrease in the genetic diversity of this iconic species has been identified starting in Marine Isotope Stage (MIS) 3 (60-25 ka) and accelerating 22 ka during the Last Glacial Maximum (LGM) (26-19 ka), with three clades absent from the post-LGM genetic pool and two clades surviving into the Holocene. 2This is the first study that gave genetic insights into the population dynamics of the giant deer.
Analysis from morphological characters suggested that the eastern Eurasia cluster, Sinomegaceros, diverged from Megaloceros more than 1 Ma. 3,6Fossil records revealed that the range of Sinomegaceros covered continental areas from approximately 50 to 25 N, mainly in present-day eastern and northeastern China and the Japanese islands. 6,12In China, there are several well-known fossil species of Sinomegaceros from different stages of the Pleistocene, with Sinomegaceros konwanlinensis, S. pachyosteus, and S. ordosianus representing the Early, Middle, and Late Pleistocene, respectively. 13In Japan, Sinomegaceros yabei appeared from the second part of the Middle Pleistocene to the Pleistocene/Holocene boundary. 6,14n addition to the phylogenetic status of the giant deer, their extinction is also a focus of research.Unlike red deer (Cervus elaphus/hanglu/ canadensis), 15,16 moose (Alces alces), 17,18 and other extant megafauna species, 19,20 giant deer went extinct during the Holocene. 21There are various speculations about their disappearance.2][23][24] The most recent, mid-Holocene population of M. giganteus spanned from eastern Europe to western Siberia. 21owever, there is a lack of similar evidence for Sinomegaceros species.
Molecular studies have significant potential to illuminate the evolution and extinction of various species.Previous molecular studies have demonstrated that fossils from the Far East can provide valuable information for understanding the evolutionary history of Quaternary mammals, e.g., the cave hyena (Crocuta spp.), 25 horse (Equus dalianensis and E. przewalskii), 26 moose, 17 and red deer. 16Among megacerines, all available ancient DNA studies are devoted to Megaloceros, the western group of giant deer, while the eastern group, Sinomegaceros, remains unstudied.In this study, we generate Sinomegaceros mitogenomic data to provide the first genetic insights into the molecular evolution of the giant deer in East Asia and their relationship to western giant deer, and bring hypothesis to elucidate their regional extirpation in northeastern China.

Radiocarbon dating and ancient mitogenomes of giant deer in East Asia
Eight out of 16 Sinomegaceros specimens (Figure S1) were dated, ranging from 20,366 to 19,835 cal BP to beyond the limit of radiocarbon dating (>43,500 BP).For the specimen beyond the radiocarbon limit and the one not dated, we molecularly dated their ages to 102 ka (95% highest posterior density (HPD): 79-134 ka, CADG532) and 58 ka (95% HPD: 43-75 ka, CADG1199), respectively (Figure S2).Geographically, one specimen from Siberia, Russia (ARI38), one specimen from Japan (ARI135), and five specimens from northern and northeastern China , respectively, the latter now extending into southern Siberia based on our data.The possible overlapping region indicated in this study is circled by black dotted line.The triangles represent our samples in this study, while the circles represent samples from previous publications. 2,21The number of multiple samples found in the same location is numerically marked (N).The colored circles and triangles indicate samples that yielded aDNA.The black circles and triangles represent samples that did not yield aDNA. Pink, orange, purple, cyan, and blue circles stand for Megaloceros clades 1-5, while gray, red, and brown triangles stand for Sinomegaceros clades sino1-3, respectively.
(CADG496, CADG497, CADG532, CADG888, and CADG1199) were traced back to the pre-LGM, while two specimens from northeastern China (CADG1006 and ARI68) were dated to the LGM (Table 1 and detailed information see Table S1).
We obtained 617-3,932 unique reads for six out of 16 Late Pleistocene specimens.Six mitochondrial sequences of ancient giant deer individuals with mean coverages of 2.15-to 15.35-fold were generated.These sequences covered 83.45%-98.42% of the reference mitochondrial sequences (16,347 bp, GenBank: MW802558) (for detailed statistics see Table S2).

Phylogenetic analyses of mitogenomes
With addition of our Sinomegaceros individuals (Figure 3A), the phylogenetic topology is almost consistent between the maximum clade credibility (MCC) tree (Figure 3B) and the maximum-likelihood (ML) tree (Figure 3C).Additional to the five clades (clades 1-5) of M. giganteus revealed by Rey-Iglesia et al., 2 the Sinomegaceros individuals did not cluster together within one mitochondrial clade but formed three separate clades (clades sino1-3 in Figures 3B and 3C).Clade sino1 represents an earlier divergent branch than previous clade 1. Clade sino2 and sino3 were placed between clade 1 and clades 2-5 of Megaloceros, with either clade sino2 or clade sino3 sitting in the root Full details are given in Table S1.
position of clades 2-5 and the others, in the MCC tree and ML tree, respectively.In the MCC tree that shows timescales of divergences, the split event for clade sino1 occurred at approximately 599 ka (95% HPD: 489-722 ka).After the divergence of clade 1 at 349 ka (95% HPD: 286-422 ka), clade sino2 and sino3 diverged at approximately 173 (95% HPD: 140-209 ka) and 155 ka (95% HPD: 125-188 ka), respectively.As further divergence events occurred, clade 3, clade 4, and clade 5 appeared between approximately 107 and 97 ka, consistent with Rey-Iglesia et al. 2 Our median-joining network analyses (Figures 3D and S4) detected 327 mutation sites from the 14,670 bp homologous mitogenome of 41 giant deer individuals and divided them into 36 haplotypes, including 30 haplotypes for 35 Megaloceros and six haplotypes for six Sinomegaceros.Regarding the geographical distribution of specimens, all Megaloceros haplotypes show no distinct geographical pattern, while for Sinomegaceros, except for a Siberian specimen ARI38, five specimens all originate from northeastern China.It is worth mentioning that the haplotype represented by CADG496 has a significantly long mutation distance (112 mutation sites, marked with bold font in Figure 3D), relative to the distances between other haplotypes, even further than that between Sinomegaceros and Megaloceros.The haplotype represented by MW802546, a Megaloceros specimen from France that has been assigned to clade 5 in both the previous study 2 and our phylogenetic trees (Figures 3B and 3C), shows a relatively high mutation distance from other haplotypes of clade 5.
The average nucleotide diversity (Pi) and the slide window analysis of the Pi results show that Sinomegaceros has the highest Pi value, while the European population of Megaloceros has the lowest Pi value (Table 2; Figure S4).Tajima's D of Sinomegaceros and Megaloceros was calculated as À0.94905 and À0.11504, respectively.Considering the geographical distributions of Megaloceros, European population has a positive Tajima's D of 0.08502, while Ural and Siberian populations have a negative Tajima's D of À0.37357.

Population dynamics and isotope analysis
The Bayesian skyline plot (Figure 4A) shows that the effective population size of giant deer started to increase from the second half of MIS 5.After a stable period in MIS 4 and the onset of MIS 3, its population shrank until about 24 ka in the LGM.Then, there was a brief recovery in population size through 24-13 ka.In the Holocene, it kept stable until its extinction ($8 ka) in Central Asia.In Figure 4B, the d 13 Ccoll values remain stable, while d 15 Ncoll values show a significant decline during the LGM.

DISCUSSION
In this study, we present the first mitogenomic investigation of the eastern Sinomegaceros, including specimens of S. pachyosteus and S. ordosianus, and combine them with published mitogenomes from M. giganteus that have been divided into five clades. 2 In the phylogenetic trees, one Sinomegaceros clade (clade sino1), represented by S. pachyosteus CADG496, is in the basal position of all giant deer in the mitochondrial tree.Two other Sinomegaceros clades are placed paraphyletically between M. giganteus clades 1 and 2, the topology of the MCC and ML trees slightly differing in terms of which sino clade settles in the basal position of clades 2-5 and the others.Therefore, neither M. giganteus nor Sinomegaceros is monophyletic in the mitogenome trees.The median-joining network shows the same clustering pattern as the trees, with the haplotype represented by CADG496 having significantly more mutation sites and longer variation distance from both Megaloceros and other Sinomegaceros haplotypes.This implies that a clear boundary at the mitogenome level seemingly did not exist between Sinomegaceros and Megaloceros.However, mitochondrial data represent a maternally inherited single locus and does not reflect the full evolutionary history of populations or species.The same scenarios of interlaced mitochondrial lineages between two groups have also been found in other mammals.For example, the extinct Eurasian cave hyena was found to be intermixed with the extant African spotted hyena (Crocuta crocuta) in the mitochondrial phylogeny. 25,27,28The straight-tusked elephant (Palaeoloxodon spp.) was detected to fall within the mitogenetic diversity of extant forest elephants (Loxodonta cyclotis). 29,30As revealed by nuclear analyses in these two mammalian groups, 31,32 recovery of nuclear sequences from eastern and western giant deer would allow us to determine if the mitochondrial phylogeny indicates either gene flow between the two Sinomegaceros species and between the two giant deer genera, or incomplete lineage sorting, 33 i.e., that several different mitogenomic lineages that diverged long before became fixed in different ancestral populations.If gene flow between the two giant deer genera proves to be correct, it would suggest that the conclusion of van der Made and Tong 6 based on morphological evidence, that there was no evidence of interaction between western and eastern giant deer since their divergence, may need revision.Moreover, the position of clade sino1 (CADG496) might be the ''correct'' position for all Sinomegaceros, in the case that clades sino2 and sino3 had picked up genetic material from Megaloceros clades 2-5.It is too soon to consider any revisions to the current specific or generic taxonomy of these deer on the basis of our results.
Rey-Iglesia et al. concluded that two of five clades (clades 4 and 5) of M. giganteus recolonized central and northern Europe from southern refugia after the LGM. 2 This conclusion has been supported by the Tajima's D value of European Megaloceros population (À0.37357, see Table 2) in our study that shows a recent dispersal after bottleneck. 34In this study, we found representatives of a Sinomegaceros clade surviving during the LGM in northeastern China (Figures 3A-3C), since this clade (clade sino2) includes one specimen (ARI38) dated pre-LGM and two specimens (CADG1006 and ARI68) dated to the LGM (Table 1 and S2).Both of the other two newly detected clades contain only pre-LGM individuals, with one individual in clade sino1 AMS-dated to 34,793 cal BP and two in clade sino3 molecularly dated to 102 ka (95% HPD: 79-134 ka, CADG532) and 58 ka (95% HPD: 43-75 ka, CADG1199).Therefore, it is currently unknown whether these two clades survived later.The existence of megafaunal remains during the LGM is often a sign of refugia. 16,17,21,35Considering the Tajima's D value of Sinomegaceros (À0.94905), it is likely that East Asian Sinomegaceros experienced a recolonization like European Megaloceros.The Sinomegaceros population has the highest genetic diversity, and we suppose that the combined results in our study, i.e., the early-divergent pre-LGM clade (clade sino1), the existence of pre-LGM individuals in all eastern Asian clades, and clade sino2 that contains LGM Chinese specimens, indicate that northeastern China could have been an LGM refugium for giant deer.
A remarkable result of our study concerns the specimen from the Upper Paleolithic site of Kamenka, Buryatiya, in the Transbaikal area of Siberia near Ulan-Ude. 36This specimen (ARI38, see Figure S1), originally identified as M. giganteus, groups closely with the individual of S. ordosianus (ARI68) from Yushu (Jilin), northeast China (Figures 3A and 3B).The sampled specimen (ARI38), a partial humerus, was found in the same assemblage as a megacerine antler base (K-93A, see Figure S5), and the radiocarbon dates of the two specimens are close. 21The humerus shows several morphological features in common with M. giganteus compared to other deer genera, 5 confirming its megacerine identity.The antler base displays an extremely wide base of the dorsoventrally flattened brow tine, with the width of 71 mm.This is consistent with S. ordosianus and not with Late Pleistocene M. giganteus (cf.Lister 37 ). 6DNA content of the antler was unfortunately insufficient, but the humerus (ARI38) DNA together with the antler (K-93A) morphology indicate an extension of the range of Sinomegaceros (pink shaded area in Figure 1), likely across Mongolia, to the Baikal region, close to (or possibly overlapping with) the known range of M. giganteus (black dotted line circled area in Figure 1).Other specimens from Russia (in particular those in M. giganteus clade 1, placed between the Sinomegaceros clades in the tree) are from western Siberia, group with a western European individual (from Belgium), and are referable to M. giganteus.
Our Bayesian skyline plot shows a continuous decline in giant deer maternal population size in MIS 3, followed by a recovery in MIS 2 (Figure 4A), which suggests a dramatic effect of climate change on population dynamics of giant deer.In Europe, Siberia, and Japan, changes in vegetation and habitat caused by climate fluctuations are considered to be the main reasons for regional extirpations of giant deer. 2,14,21,22,38alynological studies have shown that the area of forest (mostly boreal coniferous forest) in northeastern China increased dramatically through 40 to 22 ka, 39 at the expense of open or semi-open habitats.This vegetational transition may also explain the decline of d 15 Ncoll values in Figure 4B, since the d 15 Ncoll values are positively related to the proportion of open area. 40Meanwhile, paleoenvironmental studies show lake expansion, a sign of relatively warm and humid climate, in northeastern China during the period 30 to 22 ka, followed by a lake retreat reflecting dry and cold climate after 22 ka. 41Consistent with this evidence, the fossil record of both the Mammuthus-Coelodonta fauna and humans in northeastern China show dramatic increase in density during the first half of the LGM (26-22 Ka), decreasing significantly during the coldest period from 22 to 18 ka. 42To sum up, northeastern China experienced climatic and vegetational changes before and after 22 ka and an extremely dry and cold climate in the second half of the LGM (22-18 ka).These changes may have severely reduced megafauna population sizes, including that of giant deer.Considering the arrival of humans in northern East Asia before the LGM, 43 possible anthropic contributions to the extirpation of giant deer in this region remain unclear and would require archaeological evidence.

Conclusions
The newly obtained six near-complete mitochondrial genomes of S. pachyosteus and S. ordosianus correspond to three mitochondrial clades, one of which represents the earliest divergent clade in all giant deer and the other two contained within the mitogenetic diversity of M. giganteus.Besides the two known clades of M. giganteus that survived the LGM, we find a clade of Sinomegaceros during the LGM, which implies that northeastern China could have been a refugium for eastern giant deer.Combined with stable isotope evidence, paleoenvironmental studies, and faunal fossil records, our estimation of the population sizes suggests that climate-induced vegetation shifts may have contributed to population decline of giant deer in northeastern China.

Limitations of the study
The failure to retrieve ancient DNA from specimens collected in Japan limited our ability to incorporate mitogenomes of giant deer across all of East Asia.Further sampling and successful obtaining of ancient DNA and radiocarbon dates from one of the three identified clades of Sinomegaceros that contain only one specimen will tell us if this clade may also survive the LGM in northeastern China.Nuclear genome data will detect the reason of the interlaced mitochondrial lineages of two morphologically distinct groups and further elucidate the evolutionary history of the giant deer.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: specimen was mixed with extraction buffer consisting of 3 mL EDTA (0.5 M, pH = 8) and 0.045 mL Proteinase K (20 mg/mL) in a 15 mL centrifuge tube.After overnight incubation in a rotating hybridization at 37 C, the mixture was centrifuged for 10 min at 7,000 rpm.The supernatant was transferred into an ultrafiltration centrifugal tube (10 kDa, Millipore) and centrifuged at 7,000 rpm until 100-150 mL left.The purification step was carried out using MinElute PCR Purification Kit (Qiagen), and finally 50 mL ancient DNA extract was obtained.
Multiple DNA double-stranded libraries was constructed for each specimen using 20 mL extract of each sample following the protocol. 63fter blunt-end repairing, adapter ligation, and fill-in steps, indexing PCR amplifications were performed using Q5 High-Fidelity DNA Polymerase (New England Biolabs) and dual primers (P5 and P7 for Illumina sequencing platform) under the following conditions: 120 s at 95 C and 17 cycles of 15 s denaturation at 95 C, 30 s annealing at 60 C and 30 s elongation at 68 C. Libraries concentration and fragments size distribution were quantified using Qubit 4.0 (Thermo Fisher Scientific) and 4145 TapeStation (Agilent Technologies).Next-generation sequencing was conducted on Illumina HiseqX platform in Annoroad Gene Technology Co., Ltd, Beijing, China.
For the samples processed at the Natural History Museum (UK) and the Natural History Museum of Denmark (ARI38, ARI68, ARI136 and ARI136), DNA extracts were USER treated prior to library build.Library build steps were the same as above, except for the index PCR, which was performed using AmpliTaq Gold DNA polymerase (Invitrogen).These libraries were sequenced on an Illumina HiSeq 2000 platform 80 bp single read at The National High-throughput DNA Sequencing Center, University of Copenhagen, Denmark.

Mitochondrial genome hybridization capture and sequencing
For the samples that yielded aDNA, hybridization capture was carried out (Table S2).All the capture experiments were processed in a modern molecular biology laboratory at CUG.The modern DNA extraction of red deer muscle for making bait was carried out using DNeasy Blood & Tissue Kit (Qiagen) following the instruction.Four paired primers (Table S3) were used for long-range PCR. 64The bait-making and mitochondrial genome hybridization capture steps followed Meyer and Kircher 63 and Fortes and Paijmans 65 with the minor modification that the AmpliTaq Gold Polymerase (Thermo Fisher Scientific) for post-capture amplification was replaced by Hieff Canace Plus High-Fidelity DNA Polymerase (Yeasen).The enriched libraries were purified with MinElute PCR Purification Kit (Qiagen) and then sequenced on Illumina HisqX platform in Annoroad Gene Technology Co., Ltd, Beijing, China.
For the samples processed at the Natural History Museum and the Natural History Museum of Denmark, hybridization capture was carried out using a Mybaits (MYcroarray) custom-design giant deer mitogenome array based on the consensus sequences from Immel et al. 11 The standard MYbaits v.3.0 protocol was applied with hybridization for 30 h at 55 C at all relevant steps.Post-capture amplification was performed using KAPA HiFi uracil+ premix (KAPA Biosystems) for 14 cycles following MYbaits v.3.0 protocol set up recommendations.Captured libraries were pooled equimolar and sequenced on an HiSeq 2000 80 bp single read.

Data processing
Sequencing raw reads were trimmed with fastp-0.22.0 47 and reads less than 30 bp were discarded.The trimmed reads were mapped against a complete mitochondrial genome of M. giganteus (GenBank: MW802558) using bwa-0.7.15 'aln' 48 with default options expect for disabled seed.SAMtools-1.3.1 49 was used to sort mapped reads and remove duplicates with options 'sort' and 'rmdup'.The final mitochondrial consensus sequence was produced using '-doFasta 2' in ANGSD-0.938, 50setting a minimum base depth of 2 (-setMinDepthInd 2) to avoid DNA damage and sequencing errors.Calculation of reads coverage and analyses of DNA damage were processed using Qualimap-2.2.1 51 and mapDamage2.0 52 with default parameters, respectively (Figures S6 and S7).

Phylogenetic analysis
To investigate the phylogenetic position of the eastern giant deer individuals, we carried out ML phylogenetic analysis with IQtree-1.6.12. 53he six newly obtained mitochondrial genomes were aligned with 35 ancient sequences of M. giganteus, four modern sequences of Dama and Rusa serving as outgroup (Table S4) using the Kglign 54 online offered by EMBL-EBI. 66Finally, 14,670 bp homologous sequences were obtained using Gblock-0.91b 55with default parameters for subsequent analyses.
To estimate the divergence time of different giant deer groups, we carried out BEAST analysis based on 14,670 bp homologous mitogenomes consisting of our six newly obtained sequences and 35 M giganteus sequences from previous publications 2,11 using BEAST-1.10.4. 44e determined the best suitable evolutionary model of 'HKY+I+G' for the mitogenome dataset using jModelTest-2.1.9. 56The mean value of clock rate was set to 1.65 3 10 À8 substitutions/site/year with a standard deviation of 0.01 based on Artiodactyla studies. 2,67The ages of samples with accurate date were used to calibrate the evolutionary rate as prior information in the tip-dating sets.The Markov chain Monte Carlo (MCMC) was run for 80 million generations, sampling every 1,000 generations.Tracer-1.6 57was used for checking the outfiles of BEAST analysis to ensure that all effective sample sizes (ESS) values are over 200.We finally constructed the MCC tree with TreeAnnotator-1.10.4 58 and modified it using FigTree-1.4.3. 59he maternal demographic history was constructed based on the same homologous mitogenomes of 41 giant deer using BEAST-1.10.4. 44e set a same clock rate of 1.65 3 10 À8 substitutions/site/year and evolutionary model of 'HKY+I+G'.Bayesian skyline plot was visualized with Tracer-1.5. 57

Figure 1 .
Figure 1.The geographical locations of megacerine individualsLight purple and pink shaded areas stand for the historical ranges of Megaloceros giganteus and Sinomegaceros spp., respectively, the latter now extending into southern Siberia based on our data.The possible overlapping region indicated in this study is circled by black dotted line.The triangles represent our samples in this study, while the circles represent samples from previous publications.2,21The number of multiple samples found in the same location is numerically marked (N).The colored circles and triangles indicate samples that yielded aDNA.The black circles and triangles represent samples that did not yield aDNA. Pink, orange, purple, cyan, and blue circles stand for Megaloceros clades 1-5, while gray, red, and brown triangles stand for Sinomegaceros clades sino1-3, respectively.

Figure 3 .
Figure 3. Geographical locations of the six ancient DNA containing samples and the maternal phylogeny of giant deer (A) Geographical locations of our six samples that yielded aDNA in this study.(B) MCC tree of all giant deer in BEAST based on 14,670 bp homologous mitogenome sequences.For each node, Bayesian posterior probabilities are shown at the branches.Blue node bars show 95% HPD of the divergence times.The locations and dating information are shown after the accession numbers or sample names.Our samples are marked with different colored triangles indicating clades sino1-3.The samples from publications are shown with different colored circles to indicate different clades referring to Rey-Iglesia et al. 2 (C) ML tree of 14,670 bp homologous mitogenome sequences of giant deer with Dama and Rusa serving as outgroup.Bootstrapping was performed with 1,000 replicates.The bootstrapping support values of each node are shown near the nodes.(D) Median-joining network of 41 giant deer individuals to show 36 haplotypes based on 327 mutation sites calculated with PopART.Black circles represent missing haplotypes.The numbers represent mutational steps between haplotypes.Haplotypes are colored corresponding to MCC tree and ML tree.35Megaloceros individuals were divided into 30 haplotypes, while six Sinomegaceros formed six haplotypes, respectively.The number with bold font indicates the significant far distance (112 mutation steps) of the haplotype represented by CADG496 to its nearest haplotype.The haplotypes' detailed information and Median-joining network based on regions can be found in FigureS3.

Figure 4 .
Figure 4. Population dynamics and stable isotope analyses of giant deer (A) Bayesian skyline plot based on 14,670 bp homologous mitogenome sequences of 35 Megaloceros samples from previous publications and six newly obtained Sinomegaceros individuals.The x axis is years before present (ka); the y axis stands for the estimated effective female population size (Ne).Dark blue line represents median value, and blue area is the 95% HPD limits.Pink and cyan shading represent MIS stages, and the dotted box stands for the LGM.(B) Trend of stable isotope values of five Sinomegaceros individuals (CADG496, CADG497, CADG532, CADG1006, and ARI68) from northeastern China with radiocarbon dating information.Blue and red triangles represent d 13 Ccoll and d 15 Ncoll values of specimens, respectively.Five specimens were marked with numbers 1-5.*The median age of CADG532 is a molecular age estimated using BEAST-1.10.4 44 (Figure S2).

Table 1 .
Radiocarbon dates and stable isotope data of Sinomegaceros

Table 2 .
Average Pi and Tajima's D of Sinomegaceros and Megaloceros

TABLE
d RESOURCE AVAILABILITY B Lead contact B Materials availability B Data and code availability d EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS d METHOD DETAILS B Ancient DNA extraction, amplification and sequencing B Mitochondrial genome hybridization capture and sequencing d QUANTIFICATION AND STATISTICAL ANALYSIS B Data processing