Edinburgh Research Explorer Signatures of purifying selection and site-specific positive selection on the mitochondrial DNA of dromedary camels (Camelus

31 The two species of the Old World Camelini tribe, dromedary and Bactrian camels, show 32 superior adaptability to the different environmental conditions they populate, e.g. desert, 33 mountains and coastal areas, which might be associated with adaptive variations on their 34 mitochondrial DNA. Here, we investigate signatures of natural selection in the 13-35 mitochondrial protein-coding genes of different dromedary camel populations from the 36 Arabian Peninsula, Africa and southwest Asia. The full mitogenome sequences of 42 37 dromedaries, 38 domestic Bactrian, 29 wild Bactrian camels and 31 samples representing the 38 New World Lamini tribe reveal species-wise genetic distinction among Camelidae family 39 species, with no evidence of geographic distinction among dromedary camels. We observe 40 gene-wide signals of adaptive divergence between the Old World and New World camels, with 41 evidence of purifying selection among Old World camel species. Upon comparing the different 42 Camelidae tribes, 27 amino acid substitutions across ten mtDNA protein-coding genes were 43 found to be under positive selection, in which, 24 codons were defined to be under positive 44 adaptive divergence between Old World and New World camels. Seven codons belonging to 45 three genes demonstrated positive selection in dromedary lineage. A total of 89 codons were 46 found to be under positive selection in Camelidae family based on investigating the impact of 47 amino acid replacement on the physiochemical properties of proteins, including equilibrium 48 constant and surrounding hydrophobicity. These mtDNA variants under positive selection in 49 the Camelidae family might be associated with their adaptation to their contrasting 50 environments.

The camel species of these two tribes populate contrasting environmental niches.Dromedary camels, also known as Arabian camels, are mainly distributed throughout hot arid regions of Africa, Arabian Peninsula and southwest Asia, while some populations inhabit the mountains and coast of the west, southwest and southeast of the Arabian Peninsula (Almathen et al., 2018).Domestic and wild Bactrian camels are widespread in central Asia, including Mongolia, China and Kazakhstan, throughout cold plains and rocky mountains reaching altitudes up to 4000m (Burger et al., 2019).Camels of the Lamini tribe populate the high-altitude mountains of South America which can reach 7000m in height (Wu et al., 2014).
These contrasting environmental conditions might be associated with different adaptations associated with energy production and thermoregulation.For example, under water stress, dromedary camels show adaptive heterothermy, during which their body temperature fluctuates by 6 o C -8 o C, between 34 o C to more than 40 o C, to avoid water loss through perspiration during high temperatures (Tibary and El Allali, 2020).While the environmental niche occupied by Lamini species is mainly characterized by high altitude hypoxia, for which Lamini camels exhibit superior adaptability (Beall, 2007).
The mitochondria are involved in producing ATP molecules as an energy source and generating heat to maintain body temperature (Blier et al., 2001).This organelle carries its own genetic material comprising 13 protein-coding genes (ND1, ND2, ND3, ND4, ND4L, ND5, ND6, CYTB, COX-1, COX-2, COX-3, ATP6 and ATP8), 22 tRNA and 2 rRNA genes.The encoded proteins are part of the electron transport chain (Complex I -Complex V) embedded in the inner-membrane of each mitochondrion, which have been found to be targets of selective pressures to meet these two primary physiological functions (Blier et al., 2001).
The analysis of mtDNA protein-encoding genes has been shown to be effective at identifying footprints of purifying and adaptive (positive) selection in a range of species.Ruiz-Pesini et al. (2004) report that in humans living within the arctic zone, several mitochondrial genes exhibit signatures of adaptive selection, including ND2, ND4 and ATP6.Signatures of positive selection in arctic dwelling hare have also been identified in mitochondrial genes, including ATP8, CYTB and ND4 (Melo-Ferreira et al., 2014).To date, analyses of the mitogenomes of Old World camels has largely concerned diversity analyses of partial mtDNA sequences, which indicates limited phylogeographic structuring among the dromedary camels in Asia and Africa (Almathen et al., 2016, Alaqeely et al., 2021).A study by Mohandesan et al. (2017) analyzed the full mitogenome of the three Old World camel species, revealing signals of gene-wide purifying selection including positive selection in 18 sites or "codons" among the three species.
However, only a limited number of samples (≤ 10) were analysed for each species.
Here, we investigated the mtDNA protein-coding genes of dromedary camels (n = 42) distributed along Africa, Arabian Peninsula and southwest Asia for signatures of selection at gene-wide and site-specific levels.Publicly available full mitogenome sequences of camels representing domestic Bactrian (n = 38), wild Bactrian (n = 29) and New World Lamini (n = 31) were also included for comparison.

Whole genome sequence data from the Camelidae family
Genomic DNA of seven dromedary camels, which included two from Kuwait and five from Saudi Arabia, was extracted from 5ml blood samples using the DNeasy ® Blood and Tissue kit (Qiagen).Whole-genome sequencing of DNA was performed on the Illumina Hiseq 2000 platform using 150 bp paired-end libraries with 270 bp insert size.Sequencing of the Kuwaiti samples was performed by the BGI in China, and at Macrogen in South Korea for the Saudi samples (Table S1).
Single Nucleotide Polymorphisms (SNPs) were called using the HaplotypeCaller algorithm implemented in Genome Analysis Toolkit (GATK) v4.1 (McKenna et al., 2010).Quality control filtering criteria were applied on the per-sample variants using the VariantFilteration algorithm in GATK.This included: (1) excluding variants with low quality by depth (QD) (QD < 2); (2) excluding variants with root mean square of mapping quality for all reads of a site less than 40 (MQ < 40.0); (3) excluding variants with base quality score less than 30 (QUAL < 30); (4) excluding variants with high probability of allele-specific strand bias between forward and reverse strand (FS > 60); (5) excluding variants with bias in mapping quality between the reads supporting the reference and alternative alleles (MQRankSum < -12.5); and (6) excluding variants with bias in the position of the alternative allele towards the ends of the reads (ReadPosRankSum <-8).SNPs with depth of coverage ranging between two reads and three standard deviation (SD) from the mean depth of coverage across samples were retained.Fasta files of the full mtDNA sequences of each sample were generated using FastaAlternateReferenceMaker algorithm in GATK v4.1.(File S1)
We identified the best-fitting DNA evolution model according to the Bayesian Information Criterion (BIC) implemented in MEGA software which was the GTR (General Time Reversible) model with gamma correction parameter = 0.56.A neighbor-joining distance tree was constructed using MEGA, employing 1000 bootstrap replications to assess the mtDNA relationship among the camel species.ggtree package (Yu, 2020) for R software version 4.2 (R Core Team, 2022) was used to plot the tree.

Gene-wide signatures of selection analyses
Signatures of selection were investigated on the 13 mtDNA protein-coding genes at a gene-

Site-specific signatures of selection analyses
Signatures of selection at the site-specific level were identified along the 13 mtDNA proteincoding genes using the site models implemented in the CODEML package of Phylogenetic Analysis by Maximum Likelihood (PAML) v4.9 (Yang, 1997, Yang, 2007, Yang et al., 2000).
Several different models implemented in the CODEML package were evaluated.The one ratio (M0) model allows a single ω for all codons.The nearly neutral model (M1a) assumes two classes of codons: one with 0 ≤ ω0 < 1 and a proportion of codons p0, while the second class assumes ω1 = 1 and a proportion of codons p1 = 1 − p0.Model M2a (positive selection) is an extension of the M1a model with a third class that allows for ω2 > 1 and a proportion of codons p2= 1 − p1 − p0.Model M3 (discrete) uses by default three discrete classes to model the heterogeneity of ω between codons.The M7 and M8 models assume a beta distribution of ω over codons with two beta function parameters (p and q).The M7 model does not allow for codons under positive selection by constraining ω to be in the interval (0, 1).In contrast, the M8 model allows for codons with ω1 > 1 with a proportion of codons p1.After calculating the log likelihood value (L) of each model fitted to our data, likelihood ratio tests (LRTs) were conducted between the M1a-M2a and M7-M8 models to test for positive selection, and between the M0-M3 models to test for variable selection pressures among codons.The statistic for each LRT is defined as twice the log likelihood difference between two models (2ΔL).This statistic was compared to a χ 2 distribution with a degree of freedom (d.f.) equal to the difference in the number of parameters between the two models (d.f.= 2 for M1a-M2a and M7-M8, and d.f.= 4 for M0-M3).A Bayes Empirical Bayes (BEB) approach was used to identify codons under positive selection.Sites with BEB posterior probability > 0.5 were considered to be under positive selection, with a value > 0.95 considered as extreme selection (Yang et al., 2005).The site model analyses were conducted at the three different levels described previously (1: dromedary with domestic and wild Bactrian; 2: dromedary with New World Lamini; 3: Old World camels with New World Lamini).
Branch-site models of positive selection, implemented in the CODEML package (Zhang et al., 2005), were also tested on the mtDNA protein-coding genes to detect codons under positive selection in a specific lineage, the "foreground lineage", but which remained neutral or under purifying selection in the other lineages, the "background lineages".The dromedary camels were set as the foreground lineage, while the other camel populations were set as background lineages.On the branch-site models, Model A assumes positive selection by defining four classes of codons.Class 0 includes codons with 0 ≤ ω0 < 1 throughout the tree.Class 1 includes codons with ω1 = 1 throughout the tree.Class 2a includes codons with 0 ≤ ω0 < 1 in the background lineages, and ω2 ≥ 1 in the foreground lineage.Class 2b includes codons with ω1 = 1 in the background lineages, and ω2 ≥ 1 in the foreground lineages.The null model, which does not assume positive selection, is the same as Model A, but fixes ω2 = 1.As in the site models, the log likelihood value (L) was calculated for each model fitted to our data, and LRTs were conducted between Model A and the null model to test for positive selection.The calculated log likelihood statistic for the LRT was compared to a χ 2 distribution (d.f.= 1).A BEB approach was used to identify codons under positive selection as in the site models.

Camelidae full mtDNA diversity statistics and molecular phylogeny
The alignment of the 140 full mtDNA sequences returned 3,520 variable sites, of which 191 were singletons and 3,329 were parsimony-informative.Higher nucleotide and haplotype diversities were calculated for Lamini samples (Hd = 0.998 and π = 0.0254) than for those of the Camelini tribe (Table 1).Among the Camelini tribe, relatively lower nucleotide diversity and higher haplotype diversity values were observed in dromedaries (π = 0.00109 and Hd = 0.997) than in domestic Bactrian (π = 0.00124 and Hd = 0.989) and wild Bactrian camels (π = 0.00115 and Hd = 0.781).Negative Tajima's D and Fu and Li's F statistics were calculated for dromedary camels (Tajima's D = -2.24,Fu and Li's F = -2.87)and domestic Bactrian camels (Tajima's D = -1.223,Fu and Li's F = -2.42),which deviated significantly from neutrality in dromedaries.Positive Tajima's D and Fu and Li's F statistics were calculated for wild Bactrian camels (Tajima's D = 2.655, Fu and Li's F = 2.067) and Lamini camels (Tajima's D = 0.167, Fu and Li's F = 0.524), which deviated significantly from neutrality in the wild Bactrian camels.The neighbour-joining phylogenetic tree revealed species-wise mtDNA distinction among the Camelidae family, separate dromedary, Bactrian and Lamini camels from each other.Within the Lamini tribe, the Vicuna camels were differentiated from the other New World camels, while among the two-humped Old World camels we observed differentiation between the domestic and wild Bactrian camels.No geographic-genetic distinction was observed among the different dromedary camel populations (Fig. 1).  2 and Table S4).Comparing dromedary with Lamini camels revealed significant signals of positive selection in ND5, ATP6 and COX3 (Table 2 and Table S4).Comparing Old World camels with Lamini camels showed significant signals of positive selection in ND1, ND4L, ND5, ATP6 and COX3 (Table 2 and Table S4).The ratio of non-synonymous substitutions per non-synonymous sites (Dn) to the synonymous substitutions per synonymous sites (Ds) (ω ratio) was less than one, indicating purifying selection, on all the mtDNA protein-coding genes at the different analyses.Among the thirteen genes, ATP8 showed the highest ω ratio, whilst, COX1 and COX2 showed the lowest ω ratios (Fig. S1 and Table S5).

Site-specific signatures of selection
The LRT results from the M0-M3 models, which tests for variable selection pressures among codons, returned significant signals of selection both for dromedary with Lamini, and Old World with New World Lamini camels, in which a proportion of these sites had evolved under positive selection with ω value larger than 1 (Table S6).Twenty-seven amino acid substitutions in ten mtDNA protein-coding genes were identified to be under positive selection by Bayes Empirical Bayes (BEB) analyses.This included seven codons among dromedary with domestic and wild Bactrian, 21 codons among dromedary with Lamini, and 24 codons among Old World camels with New World Lamini camels (Table S7).Two of these codons are in genes for which positive selection models (M8 and M2a) fit the data significantly better than neutral models (Table 3).These are COX2 (dromedary with Lamini, M8 and M2a) and ND6 (Old World with New World Lamini camels, M8) (Table S7).Of particular note is codon 138 in COX2 which, among dromedary with Lamini and Old World camels with New World Lamini, is under extreme selection based on the BEB posterior probability value.Based on the BEB posterior probability values of the branch-site model with dromedary camels as the foreground lineage, a total of seven codons within three genes were found to be under positive selection.These include codons 34 and 66 in ATP8, codons 1 and 51 in ATP6, and codons 101, 116 and 133 in ND6.The three codons identified in ND6 exhibit intra-species variation among the dromedaries.None of the genes carrying these codons have a significantly better fit to the positive selection model (Model A) than the neutral null model (Table 4 & Table S8).We next sought to investigate the magnitude of the impact of amino acid replacement on the physiochemical properties of the proteins.Using TreeSAAP we identified 89 codons in nine genes to be under positive selection.These included eight codons in ND1, eleven codons in ND2, two codons in ATP6, five codons in COX3, seven codons in ND3, seven codons in ND4, twenty-seven codons in ND5, ten codons in ND6, and twelve codons in CYTB genes.Of these codons, 71 were found to be under positive selection between dromedaries and Lamini.
Positive selection was also observed in 26 and 25 codons between dromedaries and domestic Bactrian, and dromedaries and wild Bactrian camels, respectively.In particular, two amino acid replacements (codon 9 in ND4 and codon 101 in ND6) were found to be polymorphic among dromedaries, differentiating the two Omani camels and a single Kenyan dromedary from the other dromedaries.Out of the 31 physiochemical properties considered by TreeSAAP, we identified significant changes in six properties: the equilibrium constant (ionization COOH) (pKʹ); power to be at the middle of alpha helix (αm); long range non-bonded energy (El); solvent accessible reduction ratio (Rₐ); surrounding hydrophobicity (Hp); and alpha helical tendencies (Pα) (Table S9).Asia (Burger et al., 2019).These results are in agreement with those reported in Camelini species by Mohandesan et al. (2017).Positive Tajima's D and Fu and Li's F statistics were observed for New World Lamini species, which might also be related to the confined distribution of these camels in South America (Fan et al., 2020).

Discussion
The phylogenetic analysis revealed inter-species genetic distinction between Camelini and Lamini tribes.Within the Camelini tribe, dromedary camels separated from Bactrian camels, which also show genetic distinction between domestic and the wild-type.However, neither continental-wise nor population-wise genetic distinction was observed among the different dromedary camel populations included in the study.The observed inter-species genetic distinction between the Camelini and Lamini tribes is likely resulted from their divergence about 16.3 Mya, which was followed by the later migration of the Camelini species to Eurasia, about 6.5 -7.5 Mya, and the Lamini tribe to South America about 3 Mya (Burger et al., 2019).
The later divergence between the dromedary and Bactrian camels, about 4.4 Mya, and the different geographical distribution of these two species may explain the mitochondrial genetic distinction observed between them.Such separation has also been observed previously at autosomal level using whole genome sequence data (Ming et al., 2020, Bahbahani andAlmathen, 2022).The genetic distinction between domestic and wild Bactrian camels can likely be attributed to the 1.1 Mya divergence between them (Burger et al., 2019).Genetic separation is also observed between the wild vicuna species and the other Lamini species, with Lama and Vicugna diverging around 2-3 Mya (Fan et al., 2020).
The lack of genetic distinction among the different dromedary camel populations may be due to the historical purpose of this well-adapted species in transporting goods and people between Asia and Africa, and throughout the Arabian Peninsula.Dromedary camels were initially used in transportation along the "incense road" connecting the southern to the northern parts of Arabian Peninsula.Dromedaries were used to transport goods between Africa and Asia during the 1 st millennium Before Common era (BC) via the Islands of Socotra.In parallel to transporting goods, dromedaries were used to transfer people from Africa and different parts of the Arabian Peninsula to Makkah during the annual Pilgrimage (Wilson, 1998).All of these practices were associated with high interbreeding and gene flow between the dromedary camels along the region, which have also been observed using whole genome sequence data (Bahbahani and Almathen, 2022).Current breeding practices of local camel owners, which rely on random mating between camels, further enhance gene flow between dromedary populations.
Although the ω ratios indicate signals of purifying selection on all mtDNA protein-coding genes at the varying levels of analysis, significant signals of positive selection were identified on ND1, ND4L, ND5, ATP6 and COX3 upon comparing Old World camels with New World Lamini.This indicates adaptive divergence between these two tribes, which might be attributed to the different environmental conditions they inhabit.The various species of Camelini populate the desert and mountainous areas of Asia and Africa, adapting to arid and semi-arid conditions including tolerance to high temperature, and scarcity of food and water.In contrast, Lamini species inhabit the high altitudes of south America, reaching more than 7000 m, and thus face environmental challenges characterized with high altitude such as hypoxia (Beall, 2007, Wu et al., 2014).These environmental challenges present significant pressure on mtDNA genes, for example to adapt to low oxygen concentration in the environment given the role of mitochondria in oxygen consumption and production of ATP molecules.Upon comparing the two domestic Old World camel species (dromedary and Bactrian), significant signals of purifying selection were defined in six mtDNA protein-coding genes: ND1, COX1, ND4, ND5, ND6, and CYTB, among which ND1, ND5 and ND6 have previously been identified to be under purifying selection (Mohandesan et al., 2017).Here, purifying selection would act to eliminate deleterious alleles that negatively affect the functionality of mitochondria.
The MK, NI and ω ratio analyses are considered conservative approaches to detect signatures of selection given that mutations affecting specific codons in a gene that can be masked by gene-wide purifying selection pressures (Crandall et al., 1999).The site-model BEB analysis identified a total of 21 and 24 codons to be under positive selection, potentially associated with adaptive divergence between dromedaries or Old World camels and New World camel species, respectively.Seven codons under positive selection were identified when comparing dromedary with domestic and wild Bactrian, indicative of adaptive divergence between these two Old World camel species, potentially related to the different environmental niches they occupy.Interestingly, seven codons identified by the branch-site model BEB analysis were found to be under positive selection in dromedaries, indicating that these codons might be associated with adaptation to desert environment in dromedaries.Among these sites, codon 1 in ATP6 was found to be under positive selection in dromedaries using the branch-site models by Mohandesan et al. (2017).Although all of these codons were defined to be under positive selection based on BEB posterior probability, none of the genes fit the signature of selection alternative model (Model A) better than the neutral null model, which was also observed by Mohandesan et al. (2017).
Several codon substitutions were found to be under positive selection by the TreeSAAP software between dromedary and Lamini, and between dromedary and Bactrain.This approach is considered more sensitive than ω ratio-based approaches since a single biochemically adaptive physiochemical change can be too weak to be identified by gene-based approaches where conservation scores are high (Hughes, 2007, McClellan, 2013).We observed magnitude impacts on six physiochemical properties of proteins: the equilibrium constant (ionization COOH), power to be at the middle of alpha helix, long range non-bonded energy, solvent accessible reduction ratio, surrounding hydrophobicity, and alpha helical tendencies.Changes in these properties can impact protein functions in several ways.An increase in alpha helical tendency (Pα) may lead to long and rigid alpha helix, which renders interactions with amino acid motifs more difficult.Decreasing this property makes the amino acid more flexible to an open alpha helix, increasing the likelihood of amino acid interactions within a protein (Burkin et al., 2000).Changes of the equilibrium constant (ionization COOH) (pKʹ) and solvent accessible reduction ratio (Rₐ) impacts amino acid water solubility, rending it more hydrophilic or hydrophobic, and is considered an important change when dehydrated, increasing longevity by reducing reactive oxygen species (ROS) production (Beckstead et al., 2009).Increasing the surrounding hydrophobicity (Hp) makes the surrounding area of the amino acid site hydrophobic (Mohandesan et al., 2017).While changes in long-range non-bonded energy (El) impacts amino acid interactions and may contribute of protein structure change.
The signals of selections characterized here were investigated based on gene-wide and sitespecific analyses.These analyses have taken the advantage of including large number of samples in this study from the different species of the Camelidae family.These analyses can be further improved upon including camel samples from more diverse geographical locations with specific phenotypic traits, such as racing performance and milk production.Such phenotypes may explain the polymorphic codons within the dromedary lineage, defined by the branch-site models and TreeSAAP, in ND6 and ND4 genes.

Conclusion
In this study signatures of selection were investigated in the mitochondrial protein-coding genes of different dromedary camel populations by comparing to Bactrian and Lamini camels, revealing signals of adaptive divergence between Old World and New World camel species.
Gene-level purifying selection was identified among the Old World camels, and signals of positive selection were identified at specific codons in Old World camels.These codons provide a link between the mtDNA evolution in such camel species and their adaptation to the diverse environmental conditions they populate.
Table S9.TreeSAAP positive selection sites based on the impact on amino acids physiochemical properties changes.Pairwise comparisons between dromedary and Bactrian, dromedary and wild Bactrian and dromedary and Lamini camels.The equilibrium constant (ionization COOH) (pKʹ), power to be at the middle of alpha helix (αm), long range nonbonded energy (E1), solvent accessible reduction ratio (Rₐ), surrounding hydrophobicity (Hp), and alpha helical tendencies (Pα).
wide level using three approaches: McDonald-Kreitman (MK); neutrality index (NI); and ω ratio tests using DnaSP.The MK test measures the ratio of nonsynonymous (Pn) to synonymous (Ps) substitutions within species as compared to the ratio of nonsynonymous (Kn) to synonymous (Ks) substitutions between species.A Kn/Ks > Pn/Ps indicates a signal of positive selection, while a Kn/Ks < Pn/Ps and Kn/Ks = Pn/Ps are signals of purifying selection and neutrality, respectively (McDonald and Kreitman, 1991).NI calculates the ratio of polymorphic sites within species to the fixed sites between species at nonsynonymous and synonymous substitutions (NI=(Pn * Ks)/(Kn * Ps)).Whereby NI = 1 equates no selection, NI > 1 indicates purifying selection, while NI < 1 indicates positive selection.The ω ratio is the ratio of non-synonymous substitutions per non-synonymous sites (Dn) to the synonymous substitutions per synonymous sites (Ds).A Dn/Ds ratio > 1 indicates positive selection, while a Dn/Ds ratio < 1 indicates purifying selection, and Dn/Ds = 1 indicates no selection.Twotailed Fisher's exact test P-values were computed to define statistically significant signals for the MK and NI tests.Each of these analyses were conducted at three levels: (1) comparing dromedary camels (n = 42) to domestic and wild Bactrian (n = 67); (2) comparing the dromedary camels (n = 42) to the New World Lamini camels (n = 31); and (3) comparing Old World camels (dromedary and Bactrian) (n = 109) to the New World Lamini camels (n = 31).
Codons under positive selection were also determined using TreeSAAP v3.2(Woolley et al., 2003).TreeSAAP identifies selective influences based on 31 structural and biochemical properties.It accounts for the impact magnitude of amino acid changes on physiochemical properties of the protein, and tests if the observed degree of amino acid substitution deviates from the neutral expectation.If the change in magnitude ≥ 6 and the goodness-of-fit test Pvalue < 0.001, it is considered a strong indication of positive selection in the physiochemical property tested.For estimating the significance of changes, the software outputs two main values: category value and statistical z-score value (equivalent to P-value).The category values were numbered from (1 to 8); 1 is the most conservative amino acid category and 8 is the most radical value.Category values ≥ 6 and z-score value ≥ 3.09 are considered as significant signals of positive selection.A phylogenetic tree for each mitochondrial protein-coding gene in the different species of the Camelidae family was constructed using MEGA and used as an input for TreeSAAP.

Fig. 1 .
Fig. 1.Neighbor joining phylogenetic tree of the Camelidae family based on the full mitogenome.
By analyzing the full mitogenomes of the Camelidae family species we have identified signatures of positive adaptive divergence between Camelini and Lamini tribes.Within Camelini species, signals of gene-wide purifying selection and site-specific positive selection were also observed that might be associated with their adaptation to the different environmental conditions they are populating.Signals of past population expansion and/or positive selection in dromedary camels can be observed based on the significant negative Tajima's D and Fi and Lu's F statistics.These signals might relate to historical population growth and the continental-wise distribution of dromedaries throughout Asia and Africa.While significant positive values in wild Bactrian and non-significant negative values in domestic Bactrian camels potentially relates to the restricted distribution of wild Bactrian camels, being found only in a protected area in the Mongolian Gobi Desert, and the limited distribution of the domestic Bactrian camels in Central

Table 1 : Haplotype and nucleotide diversities, Tajima's D and Fu and Li's F statistics for dromedary, domestic Bactrian, wild Bactrian and Lamini camels included in the study.
*Significant values with P-value < 0.05

Table 2 : Types of selection signals defined on the mitochondrial protein-coding genes based on McDonald-Kreitman test and Neutrality index on the three different analyses levels.
1The P-values are the significant values of the two-tailed fisher's exact test.values with asterisk are significant with significant level of *P<0.05,**P<0.01,***P<0.001.

Table 4 : Sites identified by branch-site models, based on Bayes Empirical Bayes (BEB), as potentially under selection. Gene Codon Amino acid in Foreground lineage Amino acid in Background lineage BEB P-value Posterior probability ATP8
P-values are the significant values of the positive selection alternative and neutral null models likelihood ratio test of the branch-site models.