A total of 280 tea accessions were collected from eight African countries, including Kenya (183 samples), Rwanda (27), Tanzania (10), Malawi (20), South Africa (8), Cameroon (13), Nigeria (4), and Madagascar (15). Out of the total, 183 tea accessions sampled from Kenya represented the germplasm under cultivation in the entire East African region, i.e., Kenya, Uganda, Tanzania, Rwanda, and Burundi. The tea samples from Madagascar were collected from a deserted tea garden (located in Ambositra District) and morphologically identified as Camellia sinensis var. pubilimba Hung T. Chang, which occurs naturally in South China. The collection target was to cover the historic cultivation areas of southern and eastern Africa as well as the new entrants in western Africa. Details of all accessions are provided in Table S1. Healthy young leaves were sampled from randomly selected plants and immediately stored in silica-gel until DNA extraction. Total DNA was extracted following the cetyltrimethylammonium bromide (CTAB) method as described by Doyle and Doyle (1990).

Genotyping of the 280 tea accessions was carried out with a total of 23 highly polymorphic nSSRs employed in our previous study (Wambulwa et al., 2016). Visualization and sizing of the SSR fragments was performed using GENEMARKER v.4.0 (SoftGenetics LLC, State College, PA, USA). The genotyping procedure (amplification and fragment size determination) for all samples followed our previous study (Wambulwa et al., 2016). The nSSR data has been deposited at http://dx.doi.org/10.5061/dryad.3g02j in the Dryad Digital Repository.

Three fast-evolving plastid intergenic spacer regions (ndhF-rpl32, trnSGG-trnSr and trnSf1-trnGGG) were sequenced for 84 accessions selected based on results of the nSSR data (Figure S1, Table S1, highlighted in yellow). The 84 accessions represented the various genetic groups in the nSSR NJ tree and origins from the different countries. We also included some known hybrids together with their parents in order to untangle the genetic pedigrees of some specific accessions. Primer sequences for the three regions are provided in Table S2. PCR amplification was carried out on a GeneAmp PCR System 9700 thermal cycler (PerkinElmer, Foster City, CA, USA). The 20 μL PCR mix contained 2 μL of 10 × PCR buffer, 1.6 μL of 25 mM MgCl₂, 0.4 μL of 10 Mm dNTPs, 0.4 μL of each primer (forward and reverse), 0.15 μL of Taq polymerase (5 units/μL), 1 μL of 40–50 ng/μL of template DNA and 14.05 μL sterile double-distilled water. The PCR conditions included an initial denaturation at 94°C for 3 min, followed by 35 cycles of 30 s at 94°C for template denaturation, 30 s at 50°C for primer annealing, 50 s at 72°C for extension, and followed with an extension step of 10 min at 72°C. PCR products were purified using ExoSAP-IT (GE Healthcare Life Sciences, USA). Purified PCR products were sequenced in both directions with the same primers used for PCR and then analyzed on an ABI 3730xl DNA Sequencer (Applied Biosystems, Forster city, USA). All the cpDNA sequences were deposited in GenBank and the accession numbers are provided in Table S1.

No genotyping errors were detected in the nSSR dataset by MICROCHECKER. Only 0.34% missing data were present in the whole dataset due to consistent failures of PCR amplification. After rarefaction analysis, the mean number of alleles per locus across countries varied from 1.59 (TUGMS2 157) to 7.83 (Po9) (Table S3). Allelic richness ranged from 4.04 (Nigeria) to 6.6 (Cameroon). The overall genetic diversity (H_S) of the African tea accessions was 0.652. At the country level, Kenya exhibited the highest genetic diversity (H_S = 0.747), followed by Cameroon (0.716), and Rwanda (0.683), with South Africa having the lowest value (0.579). The fixation index (F) for the eight countries ranged from −0.17 (South Africa) to 0.112 (Kenya). The summary statistics of the genetic indices are summarized in Table 1.

The optimum value for K was 2, suggested by both L(K) and ΔK approaches (Figure S2). The genetic clustering of the 280 tea accessions is shown for K = 2, 3, and 4 (Figure 1). At K = 2, two genetic groups were defined, which corresponded to the C. sinensis var. sinensis and C. sinensis var. assamica groups found in our previous study (Wambulwa et al., 2016). Camellia sinensis var. sinensis (in red) included samples from Madagascar and some accessions from Kenya, while C. sinensis var. assamica (in green) contained most accessions from Kenya and those from the remaining six countries. Within both groups, many accessions possessed a mix of genetic material from both groups (representing “mosaic” genotypes). At K = 3, a new group (in blue) was separated from the C. sinensis var. assamica group, and included accessions mainly from Malawi, South Africa, and Kenya, while overall more accessions showed a “mosaic” genotype. At K = 4, some accessions from Kenya formed an additional group (in yellow) separated from the red group (C. sinensis var. sinensis; Figure 1). Geographical mapping of the genetic groups was based on the posterior probabilities at K = 3, because at this K-value, C. sinensis var. assamica (which constituted the majority of African tea) separated into two groups which are mainly distributed in southern and East Africa respectively (Figure 2).

The NJ tree showed three main clusters (Figure S1). The basal cluster (Cluster I) was comprised of seven accessions from Kenya (TRFK 91/1, TRFK 91/2, TRFK 306/1, TRFK 306/2, TRFK 306/3, TRFK 306/4, and TRFK 830/5). Cluster II included three sub-clusters that contained a total of eleven accessions from Malawi, Kenya, Nigeria, and South Africa. The pioneer accession for the introduction of the tea plant in Malawi (SFS 150) and its counterparts in Kenya and South Africa formed one of the sub-clusters. The third cluster (Cluster III) consisted of the remaining 262 accessions which were grouped into 34 polytomic sub-clusters.

The clustering behavior of tea accessions in the PCoA plot highlighted the distinction between the southern and East African gene pools, a pattern that was generally consistent with the NJ tree results (Figures S1, S3). Most accessions from Malawi and South Africa grouped into a single cluster including three accessions from Rwanda (35/49, 22/17, and K/32) and one from Tanzania (TRIT 201/82). The Kenyan accessions, TRFK 31/11 and TRFK 18/16, formed a single cluster together with 10 accessions from Rwanda. Five Tanzanian tea accessions grouped together with three Kenyan accessions (TRFK 430/90, TRFK 420/13, and TRFK 14/1). Some Rwandan accessions, e.g., RW/G539, RW/SR/2B1, and RW/SR/2B1/49, clustered separately with Kenyan accessions.

In fact, no clear country-specific genetic cluster was observed, except perhaps for Madagascar, albeit this including one accession from Cameroon (CL H81/22). In addition, in six cases accessions from different countries but with similar codes fell in the same subclusters: PC 108 (Malawi) and PC 108, 15/4PC 108 (South Africa); PC 81 (Malawi), 16/4PC 81 (South Africa), and AHP PC 81 (Kenya); SFS 150 (Malawi), 16/4SFS 150 (South Africa), and TRFCA SFS 150 (Kenya); BBK 35 (Kenya) and Var BB 35 (Nigeria); TRFK 6/8 (Kenya) and Var 68 (Nigeria); and SFS 204 (Malawi) and 16/4SFS 204 (South Africa).

In the NewHybrids analysis, a total of 22 accessions from Kenya were assigned P1 (parent 1), and 146 accessions assigned P2 (parent 2). The two parent assignations corresponded to C. sinensis var. sinensis and C. sinensis var. assamica, respectively, (Figure 3). For the remaining 112 accessions, a hybrid origin was suggested, including 70 F2s (second filial generation), one BC1 (backcross to parent 1), and 41 BC2s (backcross to parent 2), but no F1 (first filial generation) generation hybrid was detected. All 15 samples from Madagascar were assigned F2 hybrids status. Details of the distribution of the various parental and hybrid classes are provided in Tables S1, S4.

Genotyping and assessment of genetic diversity is an important prerequisite for the identification of core germplasm collections for conservation and optimizing crop improvement efforts (Fu, 2015). This is the first study to assess the genetic diversity of the tea plant across all the main tea-growing countries in Africa. We found that the genetic diversity of African tea (0.652) is much lower than that of Asia (0.86) calculated over 788 accessions from 14 countries, with both studies using 23 nSSR markers (Taniguchi et al., 2014). The higher diversity estimated for this region possibly results from the wider genetic sampling from the cradle of the tea plant in China and India (Hashimoto and Takashi, 1978; Hashimoto, 2001), which inherently possess higher levels of genetic diversity (Yao et al., 2012). Interestingly, our recent study found higher genetic variation among the East African tea accessions (0.717), probably because of the ex situ germplasm bank of tea plants in Kenya which contains accessions that are genetically highly disparate and assembled from various geographical origins (Wambulwa et al., 2016).

In a separate study, however, the Kenyan tea samples harbored the lowest genetic diversity when compared to Chinese and Japanese teas (Yao et al., 2008). The low variation could be attributed to the relatively short history of Kenyan tea plantations, the small sample size (only four accessions included) and the dominant markers used (ISSR). Sample size bias may be a potential confounding factor of genetic diversity estimates. In this study, we determined allelic richness based on rarefaction analysis, thus eliminating the problem of sample size disparity. Cameroon also showed the highest allelic richness (A_r = 6.6) despite its relatively small sample size, which possibly resulted from the diverse genetic origin of the tea cultivars in this country, as evidenced by the existence of five haplotypes including a private one. Unsurprisingly, among the African accessions, Kenya also exhibited a high genetic variation perhaps due to the diversity of geographic origins of the samples (Wambulwa et al., 2016).

Our results highlight the importance of the combination of differently inherited markers to clarify the breeding patterns of the tea plants in Africa. For example, TRFK 430/52, TRFK 430/63, and TRFK 430/90 are known to be hybrids between TRFCA SFS150 and EPK TN14-3 but without information on the direction of the cross. The haplotype analysis identified TRFCA SFS 150 to be the maternal parent, because the three hybrids shared its haplotype. In addition, Madagascan tea accessions could not be separated from C. sinensis var. sinensis accessions from Kenya by nSSR data. However, the cpDNA data could separate them into distinct haplotypes H3 and H9. This might indicate that the tea cultivars from Madagascar are of likely hybrid origin between C. sinensis var. sinensis (paternal) and C. sinensis var. pubilimba (maternal) native to South China, and probably introduced independently of those from mainland Africa. Accessions CL671 and China Hybrid Seedling also have different haplotypes but a similar nuclear genetic composition, implying that these accessions are also hybrids of BC or F2 (or later) generation with diverse maternal parents but probably sharing the same paternal parent, which was confirmed in our NewHybrids analysis.

Our results showed that African tea germplasm is genetically admixed and shared among countries, indicating that the tea accessions were often exchanged among different countries. For instance, the clustering patterns in the NJ tree suggested that the Rwandan accessions RW/35/49, RW/22/17, and RW/K/32 and the Tanzanian accession TRIT 201/82 might have been introduced from South Africa, while the Kenyan accessions TRFK 430/90, TRFK 420/13, and TRFK 14/1 possibly originated from Tanzania. The typically southern accessions of the SFS and PC series (with haplotype H2) are dominant in southern Africa and with low frequency in Kenya, Rwanda, Cameroon, and Nigeria, suggesting germplasm movement from the southern Africa region (Malawi and South Africa) northwards. Moreover, historical records showed that the southern breeding program (TRFCA) is the oldest in Africa (Anonymous, 1962). We therefore deduced that some of the pioneer germplasm or offspring thereof might have been moved northwards from southern Africa to East Africa and eventually to West Africa. Unique haplotypes were found in Cameroon, Kenya, and Madagascar, which suggested possibly independent germplasm exchanges and introductions from countries outside of Africa. Germplasm exchanges among countries may be an important strategy for tea or crop improvement in general. International exchange of crop germplasm is essential both for ensuring food security and for promoting research relations among the countries involved (Singh et al., 2012). In the face of the threats from climate change, interdependency among countries for genetic material of crop plants is expected to accelerate the development of new varieties for better adaptation to future climate challenges (FAO, 2015b). However, germplasm exchange should be exercised cautiously and aided by genetic fingerprinting methods, or otherwise problems of nominal redundancies in the accession databases may arise, exacerbated by the tendency of some countries to adopt different names for the same accession (Kisha and Cramer, 2011). In the present study, we did indeed find that accessions with similar names from different countries clustered together (e.g., SFS 150 from Malawi, 16/4SFS 150 from South Africa, and TRFCA SFS 150 from Kenya), suggesting that they are the same accession with slightly modified identifiers.

The NewHybrids analysis here showed similar results to our previous study on East African tea accessions (Wambulwa et al., 2016). Cultivars of C. sinensis var. assamica are the most popular and dominant tea plants in Africa. The popularity of C. sinensis var. assamica might be linked to its faster growth, ease of harvest and higher yields (Seurei, 1996). Accessions of C. sinensis var. sinensis on the other hand, are mainly held in germplasm banks as a breeding resource for hybridization, due to their lower yield because of the small leaves that are difficult to pluck. The introduction of machine harvesting, however, may result in commercialization of some of these cultivars for green and specialty tea processing in the future. Hybrid cultivars made up a high proportion of 62.5 (F2) and 36.6% (BC2) of the accessions. In these accessions, C. sinensis var. assamica contributed much more than C. sinensis var. sinensis as maternal parent. The distinct genetic group formed by the hybrids (Figure 1, in yellow) indicated that these cultivars are possibly fixed for some alleles linked to adaptation to local environments during their breeding history.

Southern Africa (TRFCA) and East Africa (TRFK) are recognized as the oldest breeding centers in Africa (Matheson and Bovill, 1950; Anonymous, 1962). Our results of both nSSR and cpDNA haplotype analysis basically reflect the activities of the two breeding centers on mainland Africa. Two distinct genetic groups were defined for C. sinensis var. assamica accessions, a southern group (in blue) and an eastern group (in green; Figures 2, 4). Although the haplotypes H1 and H2 differed only by a single mutation, H1 was associated with the pioneer selection TRFK 6/8 from East Africa, and H2 is preponderant in pioneer selections SFS 371, SFS 150, and SFS 204 from southern Africa (South Africa and Malawi; Figure 4). We concluded that TRFCA (southern Africa) and TRFK (East Africa) are the two main tea breeding centers for C. sinensis var. assamica in Africa, and have independently selected cultivars for specific traits. The occurrence of two distinct gene pools of C. sinensis var. assamica could be attributed to breeders preferences and the subsequent accumulation of divergently selected traits from distinct genetic resources to breed new preeminent cultivars.

Yield and quality are the universal traits of focus in tea breeding (Mondal, 2014). However, climate adaptation-associated traits might have played an important role in differentiating the accessions of the two breeding centers. Such traits include drought tolerance, reduced winter dormancy, hail/frost resistance, water logging tolerance, and cold hardiness. Southern Africa has a transition zone of sub-tropical and temperate climates, with desert or semi-arid regions centered in Namibia and Botswana, while eastern (equatorial) Africa has a tropical climate with a relatively high altitude and high annual precipitation. Recent studies on the tea plant have demonstrated that variations in climatic conditions can significantly affect profiles of secondary metabolite such as catechins and methylxanthines (Ahmed et al., 2014; Kowalsick et al., 2014; Larson, 2015) which directly influence tea quality. Going forward, it will be important for tea breeders to make use of niche modeling and climate data as a novel strategy for developing locally well adapted tea cultivars from divergent gene pools.

Camellia sinensis var. assamica is the main tea type under cultivation in Africa, a situation that may compromise the ability of the germplasm to adapt to biotic and abiotic challenges as a result of its relative genetic homogeneity. It is therefore important to maintain diverse genetic germplasm lineages. In the present study, several divergent haplotypes (H4, H5, H7, and H9) not derived from the tea plant were observed in the African germplasm, and may represent distinct gene pools, such as C. sinensis var. assamica in China that was recently defined as a new genetic entity (Meegahakumbura et al., 2016), and wild relatives of the tea plant. In fact, H4 represented a haplotype derived from C. irrawadiensis, native to Myanmar, which might have been introduced to breed for the presence of anthocyanin (purple) pigments in certain tea accessions (Kamunya et al., 2009). These accessions may be highly valuable resources for tea breeding in the future. Thus, it is very important to conserve such elite genetic stock as resources diversifying the African tea gene pool to be flexible to adjust to future directions in tea breeding. The phenomenon of differential selection by breeders is not limited to our study. A similar observation was made in maize (Van Heerwaarden et al., 2012), sorghum (Morris et al., 2013), wheat (Cavanagh et al., 2013), and rice (Xie et al., 2015) where accessions originating from different breeding programmes showed high differentiation. It has been shown that locally selected material collectively possesses a high diversity and potential for local adaptation (Ravigné et al., 2009), which will be very important for mitigating the adverse influence of climate change and other environmental abiotic and biotic challenges.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer LB and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review.