The 1-humped camel (Camelus dromedarius), referred to as the Arabian or dromedary camel, was domesticated approximately 4,000 yr ago in the southern Arabian Peninsula, possibly in present day Yemen and Oman (Schwartz and Dioli, 1992; Wilson, 1998). Its introduction into east Africa is thought to have occurred through the Horn of Africa, via the Suez canal, approximately 3,500 yr ago (Tefera and Getachew, 2012). According to the Food and Agriculture Organization of the United Nations (FAO), eastern Africa is known as the heartland of camel production; out of 25.89 million camels worldwide, 7.0 million, 4.25 million, and 2.40 million camels are found in Somalia, Sudan, and Ethiopia, respectively (FAO, 2011). These animals are integral, vital socioeconomic cornerstones and support the survival of millions of people in the semidry and arid zones of Asia and Africa (Epstein, 1971). The ability to survive and reproduce under prolonged water shortage, poor quality feed, and high heat load (Schroter et al., 1989; Dahlborn et al., 1992) is due to their unique morphological, behavioral, and physiological adaptations (Wilson, 1998; Ouajd and Kamel, 2009).

Knowledge about existing morphological, physiological (production profile), and genetic diversity of any animal resource is an essential prerequisite to establish effective utilization and conservation programs. The Domestic Animal Diversity Information System (DAD-IS) database and FAO recognize 97 camel breeds worldwide, based on distribution and specific characteristics (http://www.fao.org/dad-is/). Likewise, in eastern Africa, the classification is generally based on the ethnic groups and geographical distribution of the pastoral communities that own them (Tefera and Gebreah, 2001). The FAO officially recognizes 5 breeds in Ethiopia (Somali/Ogaden, Ethiopian Dromedary, Afar, Anfi, and Borena); however, pastoralists of the region recognize additional breeds (http://www.fao.org/dad-is/).

Camels are adapted to the lowlands of Ethiopia, contributing greatly to the food security and economic states of pastoralists (Eyasu, 2007; Mehari et al., 2007). They provide milk and meat of high nutritional and medicinal value (Yagil, 2004; Mehari et al., 2007; Al-Haj and Al-Kanhal, 2010) as well as transportation. They also serve as a means of investment and social prestige (Farah et al., 2007; Gwida et al., 2012). The regional states Somali, Afar, and Oromia cover 61 to 65% of the total Ethiopian land area and possess almost the entire camel population of the country (Abebe, 2001). In the face of climate change, camels are expected to become the most important domestic livestock in terms of rural food security of northern Africa (Salem et al., 2011).

Despite the socioeconomic importance at the household and national levels, the camel represents one of the least researched domestic animal species in Ethiopia (Mehari et al., 2007; Sirak, 2012). Consequently, the country does not have a national camel breeding program, registry, or functional camel development strategy. Information regarding existing diversity and potential productivity is incomplete for Ethiopian camels.

Camel breeds are officially recognized by phenotypic and genetic characterization methods (FAO, 2011). Despite observable phenotypic diversity and reported productivity variations, there is no record of genetic characterization for Ethiopian camels. Therefore, the goals of this study were to investigate morphometric and genetic variation so that to better understand the diversity of camel breeds in eastern Ethiopia.

Samples for this study were collected from 19 localities located in pastoral and livestock areas of Somali, Afar, and Oromia regional states of Ethiopia. Pastoralists, belonging to different communities, recognize their camels as specific breeds; however, as these locally recognized breeds are not confirmed by genetic data, they will henceforth be referred to as ecotypes to avoid confusion. Number and distribution of samples, assigned ecotypes, and number of samples used in separate analyses are available in Table 1 and Fig. 1. Five hundred thirty-five individuals belonging to the 8 ecotypes were subjected to morphological characterization; all data were collected from living individuals.

Between April and May of 2014, qualitative and quantitative characteristics (e.g., herd size and history of specific camels) were recorded for individuals of the 8 ecotypes following FAO (2011) phenotypic standards. Structured questionnaires were administered to camel owners and general information was gathered on pedigree, husbandry, classifications, productivity, and tolerance of camels. Morphologic data also were collected on general characteristics and descriptions of camels such as body color, hair length, pelage, tail length, ear size, orientation and hump size, and udder size. After morphologic characteristics were taken, ear tissue samples were collected from unrelated animals of both sexes. An approximately 2 cm² area was shaved with sterile surgical blades, disinfected with 70% alcohol, and then cut with a sterile blade and collected into 50-mL screw-capped sample bottles containing 10 mL of lysis buffer (Longmire et al., 1997) and shipped at ambient temperature to Texas Tech University for analysis.

Three exomorphic characters hump circumference (HC), hump length (HL), and ear length (EL) were the most diagnostic variables separating the 8 ecotypes. Means, standard deviations, and ranges of all measurements are provided in Table 5. The highest percentage of variation was explained by the first 2 components in the PCA (PC1 59.5% and PC2 14.7%) based on 17 exomorphic characteristics for 525 individuals. The bivariate plot of PC1 and PC2 showed a clear separation in size of the small ecotype (Afar) and samples from the remaining 7 ecotypes (Figure 2) . The remaining 7 ecotypes were larger, substantially overlapped with one another, and were segregated from Afar.

Discriminant functions (DF) and loadings of variables on each conical axis are summarized in Table 3. Variables with the highest loadings and the first conical axis were related to overall size. Within the DFA, the variables distance between eyes, EL, WOC, and HC separated the ecotypes based on size. These measurements displayed mainly positive scores on DF1 with the percentage of variation in DF1 being 37%, and DF2 being 28.2% with clear distinction of Afar from all other ecotypes (Fig. 3).

Both DFA and PCA shared 2 variables: EL and HC. These variables were averaged for the 7 ecotypes (EL: 12.1 and HC: 148.9) and Afar (EL: 11.5 and HC: 81.6), showing high disparity between Afar and the remaining ecotypes. Geographical information was not included when analyzing the individuals because the driving force for separation of ecotypes is thought to be the isolation of pastoralists rather than the movement of domesticated camels.

Cyt-b sequences (n = 31) were added to reference sequences from GenBank (n = 4). Not all of the 31 cyt-b sequences were complete (1,143 bp); the first 22 bp were not amplified for these individuals, reducing the total sequence length to 1,121 bp. Excluding outgroup individuals, nucleotide composition analyzed by MEGA 6 (Tamura et al., 2013) for the individuals in this study was A = 29%, C = 27.7%, T = 28.4%, and G = 14.9%. Of the 1,121 nucleotides, 25 were variable, 1,097 were conserved, and 9 were considered phylogenetically informative. Using Popart version 1.7 minimum spanning network analysis no unique haplotype for the 8 ecotypes was recovered (Leigh and Bryant, 2015) (Fig. 4). In addition, overlays of haplotypes into the distribution of localities did not support isolation of populations by distance.

Bayesian inference analysis produced a topology (Fig. 5) that failed to recover supported monophyletic groups among samples of Ethiopian ecotypes. In Fig. 5, supported nodes were predominantly terminal and without apparent structure with regard to ecotype or geographic origin. Genetic distance between clades I and II was 1.0%, genetic distance within clade I was 0.7% and within clade II was 0.3%.

In the microsatellite analysis, Structure Harvester determined the most appropriate value of K to be 2 across all runs. Two individuals from 2 different ecotypes (Jigjiga and Liben) had complete association (posterior probability of 1.0) to cluster 1. Only 2 individuals from the ecotype Ayden showed high association to cluster 2 with a posterior probability of 0.898. No individual was completely unique to cluster 2 (Fig. 6).

The morphological analysis failed to recover either the 5 defined breeds or the 8 ecotypes and instead identified only 2 significant groups: 1) Afar and 2) the remainder of the ecotypes examined in this study. Afar and Borena are the only FAO-reported breeds included in the study. The 7 remaining clustered ecotypes contain the Borena ecotype separated from the Afar ecotype. These 2 groups appear to be split based on overall size, with one group including the smallest camel ecotype (Afar) and the 7 averaged-sized ecotypes making up the other.

As previously noted, the camels of Ethiopia are separated into 5 recognized breeds (Afar, Borena, Ethiopian Dromedary this sounds including all camels in Ethiopia and Somali/Ogaden based on their production profile and phenotypic characteristics (FAO, 2011). The distribution of these recognized breeds includes the lowland areas of the county, which overlap with the study area. When comparing C. dromedarius in Ethiopia, which is best suited for the semiarid dry climate of Africa, to the genetic study of camels worldwide, it was expected that low genetic diversity between the breeds/ecotypes would be present. No data of this study were comparable to previous studies (Mburu et al., 2003; Ji et al., 2009; Mishra et al., 2009; Mahmoud et al., 2012; Nouairia et al., 2015).

Although these groups are clearly differentiated morphologically, genetic differentiation was not supported. Results of the Bayesian analysis of the cyt-b gene fail to recover monophyletic relationships between the ecotypes. In tandem, the microsatellite analysis resulted in a K = 2, identifying a potential split in the genetic lineages. However, the programs limitations need to be accounted for in that it cannot determine a K = 1 as a result of the methodology. With the high amounts of admixture from both populations in the structure results, it should be assumed K = 1 is the most plausible estimate of number of identifiable clusters.

The Structure results, when plotted with the geographical locations of the sampling localities, did not show a definite pattern. There seems to be a pattern of no genetic structure among the samples; however, the genotyping data were too limited. In general, individuals with high posterior probability to cluster 1 (Jigjiga, Issa, and Afar) were geographically located in the northern portion of the study region whereas ecotypes with high probability of belonging to cluster 2 were more southern (Borana, Ayden, and Hoor); however, 2 ecotypes (Kerreyu and Liben) did not follow this trend.

Given the currently available data, there is insufficient molecular evidence to recognize the 8 ecotypes as genetically distinct groups; however, the morphological data are capable of separating the Afar ecotype (technically a recognized breed) from the other ecotypes. It is likely that the differences in production profiles and other phenotypic characteristics reported by pastoralists are the result of intense artificial selection for these traits, as previously noted in cattle (Hayes et al., 2009). If these traits are controlled by a few loci, the remainder of the genome may have experienced little selective pressure. It is possible that with a more thorough sampling scheme (both genomic and organismal in scale), some level of genetic differentiation may be detected. A more detailed investigation of the molecular variability of pastoralist-recognized ecotypes, and the phenotypic expression of that variation, has the potential to increase understanding of domesticated animal breeds. More research is needed to address the intraspecies relationships of the Ethiopian camel.