Genetic structure of the date palm (Phoenix dactylifera) in the Old World reveals a strong differentiation between eastern and western populations

Abstract

Background and Aims Date palms (Phoenix dactylifera, Arecaceae) are of great economic and ecological value to the oasis agriculture of arid and semi-arid areas. However, despite the availability of a large date palm germplasm spreading from the Atlantic shores to Southern Asia, improvement of the species is being hampered by a lack of information on global genetic diversity and population structure. In order to contribute to the varietal improvement of date palms and to provide new insights on the influence of geographic origins and human activity on the genetic structure of the date palm, this study analysed the diversity of the species.

Methods Genetic diversity levels and population genetic structure were investigated through the genotyping of a collection of 295 date palm accessions ranging from Mauritania to Pakistan using a set of 18 simple sequence repeat (SSR) markers and a plastid minisatellite.

Key Results Using a Bayesian clustering approach, the date palm genotypes can be structured into two different gene pools: the first, termed the Eastern pool, consists of accessions from Asia and Djibouti, whilst the second, termed the Western pool, consists of accessions from Africa. These results confirm the existence of two ancient gene pools that have contributed to the current date palm diversity. The presence of admixed genotypes is also noted, which points at gene flows between eastern and western origins, mostly from east to west, following a human-mediated diffusion of the species.

Conclusions This study assesses the distribution and level of genetic diversity of accessible date palm resources, provides new insights on the geographic origins and genetic history of the cultivated component of this species, and confirms the existence of at least two domestication origins. Furthermore, the strong genetic structure clearly established here is a prerequisite for any breeding programme exploiting the effective polymorphism related to each gene pool.

INTRODUCTION

During the plant domestication process, the combined actions of breeding, selection, migration and admixture have given rise to cultivated populations genetically and phenotypically distinct from the ancestral gene pools (Grassi et al., 2003; Doebley et al., 2006; Arroyo et al., 2006). During this process, humans have notably selected traits linked to productivity, fruit quality and fertility (Zohary et al., 2012). Knowledge of the population genetics and domestication history of cultivated species is of great importance for the genetic improvement of crops relying on the effective conservation and use of the germplasm represented by the agrobiodiversity and the wild relative populations.

The date palm, Phoenix dactylifera, belongs to the Arecaceae, and is the emblematic species of the oasis agriculture. These trees produce dates with high nutritional value and maintain fertile areas of life in deserts. They are said to have a thousand uses, are highly symbolic for Muslim, Christian and Jewish religions, and have accompanied the development of early human societies. The historical distribution area extends from Mauritania in the west to Pakistan in the east and north-western India (Pintaud et al., 2013). The date palm is also present in sub-Saharan Africa and has been introduced in California, Peru, Australia and other countries (Barrow, 1998). The date palm is one of the earliest cultivated fruit trees and one of the first plants pollinated by humans (Zohary and Spiegel-Roy, 1975). Sexual propagation of date palm has been carried out since the Neolithic (Tengberg, 2003), but the resulting progeny often has unpredictable characteristics because of the high heterozygosity rate of the species. Thus, by maintaining the genetic integrity of date palm cultivars, vegetative propagation of offshoots has been set up in palm groves mainly to preserve the organoleptic traits of fruits which reduce the genetic diversity of the germplasm in the cultivated areas. Moreover, offshoots are produced in limited numbers during the date palm’s life span (Zaid and de Wet, 2002; Bouguedoura, 2012).

Date palm is currently the main crop of the arid and semi-arid countries of North Africa and the Middle East. World date fruit production reaches almost 8 million tons, every year generating millions of US dollars in benefit to local and national economies (FAOstat). However, date palm production has shifted from traditional cultivation in rich and diverse agrosystems to intensive monocultures (Jain et al., 2011). This development has led to severe genetic erosion, with the loss of cultivars and the overall impoverishment of date palm agrobiodiversity (Jain et al., 2011). Furthermore, date palm groves undergo biotic and abiotic stress, which has made it essential to set up breeding programmes to select tolerant varieties and enrich the germplasm.

The conservation of date palm genetic resources, hence, has become a critical issue for the development of date palm production and food security in desert and semi-desert areas. A global evaluation of the genetic diversity of current date palm accessions is therefore required, along with plans for the preservation of worldwide date palm germplasm.

Genetic studies using simple sequence repeat (SSR) markers were first carried out for the analysis of the genetic diversity of Phoenix dactylifera in Tunisia (Zehdi et al., 2004, 2012) and, since then, several studies have focused on the date palm genetic diversity in Sudan (Elshibli and Korpelainen, 2008), Oman (Al-Ruqaishi et al., 2008), Qatar (Ahmed and Al-Qaradawi, 2009; Elmeer et al., 2011), Iraq (Khierallah et al., 2011), Iran (Arabnezhad et al., 2012) and recently in the UAE (Chaluvadi et al., 2014). However, all these studies were based on a relatively small number of accessions centred on countries, and consequently are not representative of the overall date palm genetic diversity. A wider analysis of the date palm diversity is required to unravel the genetic relationships between the geographic groups distributed in the Old World from west to east, and to identify the potential backgrounds of genetic diversity useful for breeding programmes and for the selection of adaptability traits to biotic and abiotic stress. Furthermore, the location of centres of date palm domestication still remains unclear. Recently, the existence of an eastern domestication centre has been shown, but the question of the existence of other possible origins has not been fully resolved to date (Tengberg, 2012). Moreover, according to the domestication syndrome criteria of Meyer et al. (2012), some morphological characters, that are likely to have evolved from the ancestral wild date palm populations to the selected cultivars, correspond essentially to differences in branching, fruit and seed character. These criteria correspond largely to the traits generally selected in fruit crops (Doebley et al., 2006), but remain hypothetical since wild date palm populations have not been characterized yet (Pintaud et al., 2013).

In order to study the genetic structure and the genetic diversity in date palm, we have assessed the polymorphism of SSR markers in date palm accessions to reveal the molecular basis of genetic diversity and the relationships among a large sample of date palm accessions over a wide geographic distribution. We used nuclear microsatellite markers to examine the genetic diversity and to provide a description of the genetic structure that could be used to select genotype samples appropriate for further genetic association studies and for building a genetic core collection. We also analysed a chloroplast minisatellite (CpSSR), which is a haploid marker with a lower mutation rate than the nuclear SSRs (Provan et al., 1999), and thus evolve differently. Indeed, being maternally inherited, the chloroplast genome provides us with information about the maternal origin of the plant.

MATERIALS AND METHODS

Plant material

The sample included in this study consisted of 295 Phoenix dactylifera genotypes. The plants were collected from the geographical distribution of the species to cover the greatest possible genetic diversity, including accessions from traditional western (from Mauritania to Egypt) and eastern (from Djibouti to Pakistan) cultivation areas (Fig. 1; Supplementary Data Table S1).

Fig. 1.

Origin of the 295 date palm accessions classified into ten groups and two regions: I and II as defined according to their spatial and genetic proximity. Region I, Mauritania, Algerian, Morocco, Tunisia and Egypt; region II, Djibouti, Oman, the UAE, Iraq and Pakistan. The colours correspond to the genetic clusters defined by STRUCTURE analysis, with cluster I in green and cluster II in red.

Ten groups were defined according to the geographic origin of the studied date palm accessions. The number of accessions per group ranged from 14 for the Mauritanian group to 50 for the Algerian and Pakistan groups (Fig. 1).

DNA preparation

Total cellular DNA was extracted from young, silica gel-dried or lyophilized leaves of the 295 accessions using the TissueLyser and the DNeasy Plant Mini Kit (Qiagen SA, Courtaboeuf, France) or the PureLink Plant Total DNA Purification Kit (Invitrogen) according to the manufacturer’s protocol. After purification, DNA concentrations were determined using a GeneQuant spectrometer (Amersham Pharmacia Biotech, France). The quality was checked by agarose minigel electrophoresis (Sambrook et al., 1989). The resulting DNA solutions were stored at −20 °C.

Amplification and genotyping

The 18 SSR loci selected for this study included dinucleotide repeats from an SSR genomic library (Billotte et al., 2004) and from introns of genes coding for transcription factors (Ludeña et al., 2011), as well as tri-/hexanucleotide repeats from coding sequences identified by in silico mining (Aberlenc et al., 2014) in the whole date palm genome sequence (Al Dous et al., 2011). New primers for some of these loci were designed in the present study (Table1, Supplementary Data Table S2). In addition to the 18 nuclear SSR loci, the plastid dodecanucleotide minisatellite identified in the intergenic spacer psbZ-trnfM (Henderson et al., 2006) was genotyped for all date palm accessions to reveal the repeat number (three or four), defining the chlorotype (Occidental and Oriental, respectively) in date palm (Pintaud et al., 2010, 2013). This repeat number is linked to other mutations in the chloroplast genome (Ballardini et al., 2013).

Table 1.

Genetic diversity obtained at 18 SSR loci in the 295 date palm accessions

Marker	Reference	NA	NA,P	Major allele frequency	NG	PIC	Ho	He	FIS
mPdCIR010	Billotte et al. (2004)	15	6	0·260	45	0·821	0·815	0·787	–0·035
mPdCIR015	Billotte et al. (2004)	13	6	0·341	35	0·775	0·750	0·747	−0·005*
mPdCIR016	Billotte et al. (2004)	5	3	0·436	11	0·612	0·580	0·589	0·016
mPdCIR025	Billotte et al. (2004)	16	5	0·300	47	0·781	0·793	0·722	−0·098
mPdCIR032	Billotte et al. (2004)	13	5	0·342	36	0·774	0·758	0·743	−0·021
mPdCIR035	Billotte et al. (2004)	11	3	0·519	24	0·603	0·527	0·609	0·135***
mPdCIR057	Billotte et al. (2004)	10	4	0·639	26	0·536	0·492	0·469	−0·049
mPdCIR063	Billotte et al. (2004)	15	4	0·347	32	0·726	0·637	0·647	0·016*
mPdCIR078	Billotte et al. (2004)	27	5	0·158	94	0·888	0·806	0·810	0·005
mPdCIR085	Billotte et al. (2004)	20	7	0·266	68	0·848	0·787	0·786	−0·001
mPdIRD013	Aberlenc-Bertossi et al. (2014)	3	1	0·986	3	0·027	0·240	0·230	−0·040
mPdIRD031	Aberlenc-Bertossi et al. (2014)	4	3	0·863	7	0·232	0·240	0·225	−0·064
mPdIRD033	Aberlenc-Bertossi et al. (2014)	4	2	0·880	7	0·210	0·202	0·206	0·021
mPdIRD040	Aberlenc-Bertossi et al. (2014)	5	3	0·651	10	0·438	0·485	0·451	−0·076
PdAG1-ssr	Ludeña et al. (2011)	33	5	0·210	92	0·874	0·802	0·812	0·012
PdAP3-ssr-F4	Accession number: KC188337	13	6	0·224	47	0·839	0·795	0·744	−0·068
PdCUC3-ssr1	Accession number: HM622273	2	1	0·992	2	0·017	0·200	0·190	−0·050
PdCUC3-ssr2	Accession number: HM622273	28	7	0·185	82	0·882	0·769	0·810	0·052*
Overall		13·17	4·22	0·478	37·11	0·605	0·571	0·569	−0·014

N_G, number of genotypes per locus; N_A, number of alleles per locus; N_A,P, number of alleles with a frequency >5 %; PIC polymorphic information content; H_e expected heterozygosity; H_o observed heterozygosity; F_IS, fixation index values.

Exact test significant at *P < 0·05, ***P < 0·001.

Amplification reactions were performed in a final volume of 20 μL containing 15 ng of template DNA, 10× reaction buffer, 5 pmol of each forward and reverse primer, 0·2 mm of each deoxynucleotide, 2 mm MgCl₂ and 1 U of Taq polymerase (Sigma, USA). The forward primers were 5' labelled with one of three fluorescent compounds (6-FAM, NED or HEX) to enable analysis with automated sequencers. PCR was carried out using an Eppendorf Mastercycler pro vapo protect (AG, Hamburg, Germany) thermocycler. After 5 min at 94 °C, 35 cycles were performed with 30 s at 94 °C, 60 s at the annealing temperature (depending on the locus) and 30 s at 72 °C, followed by a final extension step of 5 min at 72 °C. Amplified products were run on an ABI 3130XL Genetic Analyzer (Applied Biosystems, USA). Analysis and allele size scoring were performed using the GeneMapper V3.7 software (Applied Biosystems).

Genetic diversity analyses

PowerMarker 3.25 (Liu and Muse, 2005) was used to estimate the total number of alleles (N_A), the number of alleles with a frequency higher than 5 % (N_A,P) major allele frequency, total number of genotypes (N_G) and the polymorphic information content (PIC) at each locus. The GenAlEx 6.5 program (Peakall and Smouse 2006, 2012) was used to calculate the observed (H_o) and the expected (H_e) heterozygosities. The inbreeding coefficient (F_IS) and the genetic differentiation (F_ST) were computed according to the formula of Weir and Cockerham (1984) using GENEPOP 4.0 (Raymond and Rousset, 1995). GENEPOP software was also used to estimate linkage disequilibrium. χ² tests were used to determine the significance of the linkage disequilibrium (Weir, 1996). In addition, the multilocus characterization of the sample was tested by the use of Ohta’s parameters (Ohta, 1982). D_IT², D_IS², D_ST², D′_IS² and D′_ST² are the total variance of linkage disequilibrium, the components due to linkage within groups, allelic differentiation between pairwise loci, gametic differentiation between pairwise loci and linkage disequilibrium in the total sample, respectively.

Genetic structure analysis

To identify the population structure of the date palm collection, we employed a model-based clustering algorithm implemented in the computer program STRUCTURE version 2.2 (Pritchard et al., 2000). This algorithm identifies the optimal number of clusters (K) with different allele frequencies and assigns portions of individual genotypes to these clusters. It assumes Hardy–Weinberg equilibrium and linkage equilibrium within clusters. The STRUCTURE algorithm was run without previous information about the geographic origin of the accessions, using a model with admixture and correlated allele frequencies, and with ten independent replicate runs for each K value (K ranging from 1 to 10). For each run, we used a burn-in period of 10 000 iterations, and a post-burn-in simulation length of 1 000 000. The most probable number of clusters was assigned by using the run with the maximum likelihood, which was validated with an ad hoc quantity based on the second-order rate of change in the log probability of data between different K values (Evanno et al., 2005). To obtain optimal alignment of the independent runs, the CLUMPP version 1.1 software (Jakobsson and Rosenberg, 2007) was used with greedy algorithms, 10 000 random input orders and 10 000 repeats, to calculate the average pairwise similarity (H') of runs. The output obtained was used as input by the cluster DISTRUCT version 1.1 visualization program (Rosenberg, 2004).

Date palm genetic relationships between populations

The GenAlEx 6.5 program (Peakall and Smouse, 2006, 2012) was used to calculate the number of alleles per locus and per group (N_A) as well as the number of alleles with a frequency higher than 5 % (N_A,P). PowerMarker 3.25 (Liu and Muse, 2005) was used to estimate the observed (H_o) and the expected (H_e) heterozygosities, fixation index values (F_IS) and the allelic richness of each group.

To generate hierarchical classifications, we calculated the shared allele distance (D_AS) (Chakraborty and Jin, 1993) among the ten pre-defined populations. We used the obtained distance matrix to construct the dendrogram using the Neighbor–Joining (NJ) algorithm. Bootstrap values were computed over 10 000 replications with the PowerMarker V3.25 software (Liu and Muse, 2005).

Multivariate analysis was performed to examine the relationship between accessions, using discriminant analysis of principal components (DAPC) (Jombart et al., 2010), carried out on the allele frequency matrix using the R software (R Core Team, 2012).

The genetic differentiation generated by the 18 loci was estimated by calculating F_ST according to the formula of Weir and Cockerham (1984) using GENEPOP 4.0, and Fisher’s method (Raymond and Rousset, 1995) was applied to test the significance of pairwise F_ST values. The analysis of molecular variance (AMOVA) implemented in the GenAlEx 6.5 program (Peakall and Smouse, 2006, 2012) was conducted to estimate the hierarchical differentiation at two levels: a country of origin level and a region level, distinguishing two geographic regions: an eastern region and a western region. The F_ST significance was determined by running 10 000 permutations.

RESULTS

SSR polymorphism

A total of 237 alleles were identified across the 18 nuclear SSR loci studied, varying from two (PdCUC3-ssr1) to 33 (PdAG1-ssr) per locus (Table 1). The number of alleles per locus with a frequency >5 % ranged from one (PdCUC3-ssr1 and mPdIRD013) to seven (mPdCIR85 and PdCUC3-ssr2). The average number of alleles per locus was 13·17, but dropped to 4·22 when rare alleles (i.e. with a frequency of <5 %) were removed. The major allele showed a highly variable frequency from one locus to another, ranging from 0·158 (mPdCIR078) to 0·992 (PdCUC3-ssr1) (Supplementary Data Fig. S1). The number of genotypes per locus varied from two (PdCUC3-ssr1) to 94 (mPdCIR078), with an average of 37·11 and with a total number of 668. The average PIC value for the 18 loci was 0·605, and the most informative locus was mPdCIR78 (0·888). The Nei’s genetic diversity ranged from 0·190 (PdCUC3-ssr1) to 0·812 (PdAG1-ssr) with an average of 0·569 (Table 1), suggesting that the examined date palm germplasm has high levels of polymorphism. Three of the 18 SSR loci displayed a significant heterozygosity deficit (P < 0·01) (mPdCIR035 and mPdCIR063 loci and PdCUC3-ssr2) and the mPdCIR015 locus displayed a significant heterozygosity excess (P < 0·01).

Variance components of linkage disequilibrium

A total of 153 two-locus pairs were analysed to estimate variance components of linkage disequilibrium. The total variance of the disequilibrium D_IT² was of 0·0287. D′_IS² (0·0255) was larger than D′_ST² (0·0001); in addition, the D_ST² (0·0015) value was larger than D_IS² (0·0006), signifying that a greater proportion of total variance in disequilibrium resulted from deviation among sub-groups rather than from within groups. Out of 153 pairs of loci, 52 significant non-random associations were detected for SSR loci combinations assessed by the χ² test (Supplementary Data Table S3). As shown in Table 2, non-random association of the SSR alleles at particular variable loci was mainly caused by limited migration and random process or genetic drift. In fact, the dual relationship I (D_IS² < D_ST²; D′_IS² > D′_ST²) was only found for 11 loci pairs. However, the dual relationship II (D_IS² > D_ST²; D′_IS² > D′_ST²), characteristic of non-systematic disequilibrium, was found for all other pairs.

Table 2.

Dual relationships and average values of significant disequilibrium coefficients for 32 pairs of SSR loci

Dual relationships	No. of locus pairs	Ohta coefficient
		DIS2	D′IS2	DST2	D′ST2	D′IT2
I. DIS2 < DST2; D′IS2 > D′ST2	41	0·0005	0·0285	0·0017	0·0001	0·0319
II. DIS2 > DST2; D′IS2 > D′ST2	10	0·0010	0·0145	0·0008	0·0001	0·0173
All	52	0·0005	0·0255	0·0015	0·0001	0·0028

Model-based Bayesian clustering analysis

The model-based Bayesian clustering approach implemented in STRUCTURE (Pritchard et al., 2000) was used to investigate the genetic structure of date palm according to the models with two clusters (K = 2) to six clusters (K = 6). The change rate in the log likelihood between successive K values (ΔK) (Evanno et al., 2005) revealed a first level of clustering at K = 2 for the studied date palm accessions (ΔK = 1428·11) (Fig. 2; Supplementary Data Fig. S2).

Fig. 2.

Inferred population structure for K = 2 to K = 6 as the presumed number of sub-populations within the date palm collection, including 295 accessions. The bar plot chart, generated by DISTRUCT, depicts classifications with the highest probability among assumed clusters in the date palm groups. Each individual is represented by a vertical bar, partitioned into coloured segments representing the proportion of the individual’s genome in the K clusters. The date palm groups are separated by a black line.

Based on the average pairwise similarity of individual assignments across runs (H′) generated by CLUMPP for the ten STRUCTURE runs, the highest similarity coefficient (H′) was observed for K = 2 (H′ = 0·997), indicating the stability of the results for this model (Fig. 2).

At K = 2, the date palm accessions were differentiated into two geographic clusters, the first one named the Western cluster consisted of accessions from North Africa except those from Djibouti, and the second cluster named the Eastern cluster consisted of the accessions originated from Asia and those from Djibouti (Fig. 2). At K = 3, the accessions from Djibouti were separated from those collected in Asia and an admixture was observed in the Egyptian group. At K = 4, a fourth cluster was identified, including the Mauritanian and Algerian accessions.

At K = 5, the Egyptian accessions were separated from the other western accessions. At K = 6, the genetic structure of date palm within two main gene pools was not modified since the accessions of the sixth cluster were not consistently assigned and hence no meaningful additional cluster was observed (Fig. 2).

It is important to note that the Tunisian, Omanian and UAE groups showed minimal admixture signal and were clearly maintained for all K values. This allowed us to consider that the Tunisian group is the most characteristic of the Western cluster diversity, whereas the Omanian and the UAE groups are the best representatives of the Eastern cluster.

Genetic diversity of the chloroplast minisatellite

Two haplotypes (242 and 254 bp), corresponding, respectively to three and four repetitions of the 12 bp minisatellite, were identified in the date palm accessions studied, corresponding to the two chlorotypes previously identified (Pintaud et al., 2013). Their frequencies were significantly different between the two geographic regions (G-tests, P < 0·05). The Oriental chlorotype (254 bp) was present at a very high frequency in eastern date palms (0·965). A total of 139 individuals out of 144 displayed this chlorotype and only five accessions had the Occidental chlorotype. The African date palms presented the two chlorotypes; however, the 242 bp haplotype (the Occidental chlorotype) is slightly more abundant, with an allele frequency of 0·529 (80 samples out of 151), with a maximum value obtained for the Tunisian group (Fig. 3). It is important to note that only the Occidental chlorotype was found in the Tunisian samples collected on the islands of Kerkennah and Djerba, representing spontaneous populations (Ben Salah, 2010; Pintaud et al., 2013).

Fig. 3.

Frequencies of the occidental (red column charts) and oriental (green column charts) chlorotypes.

Genetic structure and diversity between date palm geographic groups

The genetic diversity levels within the ten defined geographic groups measured by the estimators differed. In the western groups, Egypt showed the highest expected heterozygosity values, with 0·610, while the Mauritanian group had the lowest, with 0·476 (Table 3). The observed heterozygosity was the highest for Algeria (0·616) and the lowest for Mauritania (0·460). As the number of alleles observed in a group is highly dependent on the sample size, the allelic richness was computed for each group and region (Table 3). The highest allelic richness was detected for the Egyptian group, with 5·904, while the lowest value was noted in Mauritania (4·166). When the eastern and western regions were considered, the number of private alleles was higher in the eastern region, whereas the values of allelic richness were equivalent between the two regions (Table 3).

Table 3.

Genetic diversity within geographic date palm groups and regions

Population	Western region						Eastern region
	Maur 14	Mc 21	Al 50	Tn 38	Eg 28	All 151	Dj 18	Om 34	UAE 24	Iq 18	Pak 50	All 144
NA	4·167	5·944	7·111	5·833	6·944	9·000	5·833	5·611	5·722	5·167	8·111	10·722
NA,P	3·000	4·000	4·278	3·833	4·667	4·167	3·778	3·722	3·889	3·722	3·944	3·833
He	0·476	0·598	0·590	0·580	0·610	0·614	0·564	0·539	0·539	0·577	0·593	0·588
Ho	0·460	0·606	0·616	0·579	0·566	0·582	0·559	0·570	0·583	0·590	0·583	0·578
FIS	0·070	0·011	−0·033*	0·014	0·089***	0·052	0·038*	−0·043	−0·060	0·007	0·027	0·008
Ar	4·166	5·423	5·422	4·783	5·904	5·140	5·343	4·676	5·037	4·818	5·629	5·100

Maur, Mauritania; Mc, Morocco; Al, Algeria; Tn, Tunisia; Eg, Egypt; Dj, Djibouti; Ir, Iraq; Om, Oman; Pak, Pakistan.

N_A, number of alleles per locus; N_A,P, number of alleles with a frequency >5 %; H_e expected heterozygosity; H_o observed heterozygosity; F_IS, fixation index values; A_r, allelic richness.

Exact test significant at *P < 0·05, ***P < 0·001.

The defined groups displayed substantial genetic differentiation since the average F_ST value was 0·107, ranging from 0·004 for pairwise comparisons between the Oman and UAE groups to 0·152 between the Oman and Tunisian groups (Table 4). All pairwise F_ST values were significant at P < 10^–4 (Table 4), except for those observed between the UAE and Oman and between Iraq and Pakistan. The same result was obtained with D_AS distances, where the highest distances were obtained between Tunisia and the two eastern groups: Oman and the UAE (Table 4).

Table 4.

Geographic group pairwise comparisons

	Maur	Mc	Al	Tn	Eg	Dj	Om	UAE	Iq	Pak
Maur	–	0·101	0·093	0·148	0·140	0·157	0·182	0·170	0·195	0·173
Mc	0·066*	–	0·064	0·057	0·084	0·165	0·201	0·213	0·180	0·169
Al	0·059*	0·036*	–	0·072	0·097	0·140	0·190	0·183	0·162	0·165
Tn	0·093*	0·031*	0·037*	–	0·124	0·257	0·276	0·275	0·254	0·251
Eg	0·085*	0·040*	0·054*	0·065*	–	0·115	0·191	0·187	0·133	0·137
Dj	0·102*	0·087*	0·079*	0·136*	0·062*	–	0·112	0·122	0·066	0·065
Om	0·116*	0·114*	0·105*	0·152*	0·104*	0·070*	–	0· 040	0·042	0·037
UAE	0·118*	0·119*	0·105*	0·151*	0·104*	0·080*	0·004	–	0·065	0·040
Iq	0·119*	0·089*	0·085*	0·129*	0·067*	0·036*	0·042*	0·046*	–	0·011
Pak	0·104*	0·087*	0·089*	0·132*	0·072*	0·035*	0·023*	0·028*	0·006	–

D_AS, genetic distances (above diagonal) and F_ST (below diagonal); global F_ST = 0·107.

*P > 10^–4.

Maur, Mauritania; Mc, Morocco; Al, Algeria; Tn, Tunisia; Eg, Egypt; Dj, Djibouti; Ir, Iraq; Om, Oman; Pak, Pakistan.

Genetic relationships among the pre-defined date palm groups were also assessed based on D_AS genetic distances and the NJ algorithm (Fig. 4). According to the bootstrap values, the ten date palm groups were classified into two clusters: cluster I, including Mauritania, Morocco, Algeria, Tunisia and Egypt and corresponding to the Western cluster, was clearly distinguished from cluster II (or Eastern cluster), including Djibouti, Oman, Iraq, the UAE and Pakistan groups, by a high bootstrap support of 80 % (Fig. 4).

Fig. 4.

NJ clustering of geographic groups based on D_AS genetic distance values, as well as the distribution of the genetic clusters within each of them. The colours correspond to genetic clusters defined by the STRUCTURE analysis, as reported in Fig. 1 with cluster I in green and cluster II in red (pie charts). The numbers next to the nodes indicate bootstrap support percentages in 1000 pseudoreplicates.

The DAPC performed on allelic frequencies and with a proportion of conserved variance of 0·85 showed that the first axis differentiated the eastern from the western geographic groups (Fig. 5), and the second axis differentiated Mauritania and Egypt from the other western groups and Djibouti from the eastern ones. The observed group distribution reflected their geographic location; Egypt and Djibouti were found at an intermediate position between the two eastern and western regions.

Fig. 5.

Discriminant analysis of principal components (DAPC) on SSR data of genotypes. The colours of the groups correspond to the geographic groups. Axes 1 and 2 explain 83 % of the total variance (shaded vertical bars in the eigenvalue histogram). Maur, Mauritania; Mc, Morocco; Al, Algeria; Tn, Tunisia; Eg, Egypt; Dj, Djibouti; Ir, Iraq; Om, Oman; Pak, Pakistan.

Based on the AMOVA, the genetic variance was about 7 % among these two defined regions and 4 % among date palm population groups (Table 5). The majority of total genetic variance (89 %) was due to the variation within groups.

Table 5.

Analysis of molecular variance (AMOVA) for the clusters of date palm accessions based on 18 SSR markers

Source of variation	d.f.	Sum of squares	Estimated variability	Percentage variance
Among regions	1	138·471	0·398	7*
Among groups	8	145·974	0·230	4*
Within groups	580	3049·454	5·258	89*
Total	589	3333·900	5·886

The region level considered the Western and Eastern clusters; the group level included the ten geographic date palm groups.

*P < 0·001 based on 999 permutations.

The number of accessions assigned to their geographic group (ten countries) was 204 (69·15 %) and the number of accessions assigned to their region (western or eastern) was 286 (96·95 %) (Table 6).

Table 6.

Pattern of assigned accessions per geographic group and region

Region	Group	Total	Assigned group		Assigned region
			Self-group	Other group	Self-region	Other region
Western	Mauritania	14	11	3	14	0
	Morocco	21	11	10	19	2
	Algeria	50	39	11	49	1
	Tunisia	38	36	2	38	0
	Egypt	28	23	5	23	5
Eastern	Djibouti	18	11	7	18	0
	Oman	34	24	10	34	0
	UAE	24	13	11	24	0
	Iraq	18	11	7	17	1
	Pakistan	50	25	25	50	0
	Total (%)	295	204 (69·15)	91 (30·85)	286 (96·95)	9 (3·05)

DISCUSSION

Date palm germplasm is structured into two gene pools

The current study used both nuclear SSR and chloroplast microsatellite genotyping to analyse the diversity and genetic structure of date palm accessions collected in ten countries from Mauritania to Pakistan.

A total of 295 accessions were analysed using 18 nuclear SSRs loci which revealed >200 alleles (approx. 13 alleles per locus). This result highlights a high degree of genetic diversity within the studied date palm sampling and shows an average allele per locus significantly higher than that obtained in several previous studies focusing on a single country, such as Tunisia (9·71) (Zehdi et al., 2012), Iraq (8·54) (Khierallah et al., 2011) or Qatar (7·7) (Elmeer et al., 2011).

The analyses of the genetic structure of the date palm population, using the Bayesian clustering approach, NJ hierarchical classification and DACP, are clearly consistent with a geographical structuring into two clusters. The first one, named the Eastern pool, consists of the date palm accessions from Djibouti, Oman, Iraq, the UAE and Pakistan groups, and the second, named the Western pool, consists of the remaining accessions from Africa, including Mauritania, Morocco, Algeria, Tunisia and Egypt groups. This geographic divergence is highly significant (F_ST = 0·077, P = 0·001). Furthermore, the assignation percentage of accessions reached 69·15 % when the ten pre-defined groups were considered and it increased up to 96·95 % when the western and eastern regions were considered. In addition, D_AS distances showed the highest values when comparing Western vs. Eastern groups, in particular Tunisia vs. Oman or UAE.

These results are consistent with those of Cherif et al. (2013), that identified an east–west structure in male-specific microsatellite alleles using a set of male and female date palm genotypes representative of the diversity of the species. The distribution of Y haplotypes in western and eastern haplogroups allowed two male ancestral paternal lineages to be traced accounting for all known Y diversity in date palm (Cherif et al., 2013).

Al-Ruqaishi et al. (2008) also used microsatellite markers to analyse the genetic diversity among genotypes of date palm from Oman, Bahrain and Iraq, and one Moroccan genotype Medjool that appeared highly distinct from the Middle East accessions.

Several genetic analyses performed on date palm previously suggested the existence of eastern and western gene pools (Pintaud et al., 2013). Elshibli and Korpelainen (2008) compared the genetic diversity of date palm from Sudan and Morocco and showed a significant differentiation between them. Additionally, Zehdi et al. (2012) showed that the eastern group was genetically different from the Tunisian groups. Arabnezhad et al. (2012) obtained the same result using a new set of SSR markers in a different date palm collection from Iran, Iraq and North Africa.

Similar outcomes were reported in genetic structuring analyses of other Mediterranean crops. Indeed, numerous genetic studies on olive trees reported a genetic differentiation between western and eastern Mediterranean areas (Besnard et al., 2002, 2007, 2013; Breton et al., 2006; 2009 Sarri et al., 2006; Haouane et al., 2011) and showed the structuring of the Mediterranean olive germplasm into three main gene pools, strongly matching three distinct geographic areas, i.e. western, central and eastern Mediterranean regions. The genetic structure of cultivated grapevine revealed three main genetic groups: cultivars from western regions, from the Balkans and Eastern Europe, and from Eastern Mediterranean, Caucasus, Middle and Far East countries (Bacilieri et al., 2013). In apricots, the Mediterranean germplasm was also structured into three main gene pools: Irano-Caucasian, North Mediterranean Basin and South Mediterranean Basin (Bourguiba et al., 2012). All these studies indicate that the main Mediterranean crops have more than one cradle of genetic diversification, being strongly linked to geographic distribution. In line with these conclusions, our findings in date palm revealed a deep genetic structuring arising from two geographic gene pools.

Genetic history of cultivated date palm

The currently accepted hypothesis for date palm domestication history is a single origin around the Persian Gulf, followed by the diffusion of the cultivated date palms toward the west up to Morocco and the east up to India (Munier, 1973). In the present study, however, the nuclear SSR markers have shown that the western accessions strongly differed from the eastern ones. Assuming a single eastern origin, this pattern of diversity would be observed only in the case of a strong bottleneck during expansion toward the west. This would have led to an important genetic drift and therefore a loss or a fixation of alleles in the western group that therefore would have displayed a very different allelic pattern from that of the eastern group. However, the diversity levels observed in the Eastern and Western clusters appear similar, questioning the accuracy of the single origin hypothesis, in accordance with previous morphometric data (Terral et al., 2012).

The strong geographic structure between the Eastern and Western pools highlighted here suggests another hypothesis, which is the existence of two cultivation origins, one in the east and one in the west. The Persian Gulf might indeed not have been the only domestication centre for date palms. North Africa could also have been a domestication centre, either independent (primary domestication) or after the introduction of genotypes from the Middle East and their crossing with local date palms (secondary domestication). The origin of this structuration is, however, imperfectly known due to the still limited information about distribution and genetic structure of wild date palm populations in the two diversification centres (Gros-Balthazard, 2012).

The plastid data provide additional insights on this structure. The two previously identified chlorotypes, Oriental and Occidental (Henderson et al., 2006; Pintaud et al., 2010, 2013), were also recovered in the present study. Investigations on the chloroplast genome have shown that these two haplotypes underlie a strong genetic differentiation. Indeed, sequence alignment of approx. 50 kb of an Occidental chloroplast from Elche, Spain and an Oriental chloroplast (cultivar Khalass) from the Arabic peninsula revealed 18 SNPs along with a number of other mutations, including microsatellites, indels and small inversions (Scarcelli et al., 2011), while the comparison of two complete sequences of Oriental chloroplast genomes, of cultivars Khalass and Aseel, resulted in only three SNP (Kahn et al., 2012).

The Eastern cluster showed a predominance of the Oriental chlorotype with a mean of 96·5 % and it reached 100 % in the Djiboutian, UAE and Omanian groups. The Western cluster, however, revealed both haplotypes but with a predominance of the Occidental chlorotype and with a maximum value in the Tunisian group. The most parsimonious hypothesis is that each cluster first had its own haplotype before population movements and anthropogenic actions resulted in the dissemination of the Oriental chlorotype in the Western cluster, and vice versa to a lesser extent. The high level of genetic diversity and the coexistence of the two chlorotypes in the Western cluster on the other hand indicate that the diffusion of accessions from the eastern to the western region was more important.

The absence of the Occidental chlorotype in the Djiboutian, Omanian and UAE groups, as well as the absence of the Oriental chlorotype in the Tunisian island’s samples strengthens this hypothesis. However, this pattern could also be observed in the case of a strong reduction of population size leading to genetic drift and/or selection of one or the other chlorotype. Therefore, the identification of wild populations and their genotyping is necessary to test these hypotheses.

While the study of the nuclear genome using SSR markers revealed that this component is highly structured between western and eastern date palms, the previously identified chlorotypes (Occidental and Oriental) were not always related to the region, particularly in the west. Consequently, some groups present a western nuclear genome associated with the Oriental chlorotype, which is particularly the case in Mauritania and Algeria. The chloroplast provides information concerning the maternal origin of the plant. This pattern could be observed if female genotypes were initially imported from the eastern to the western regions and were subsequently pollinated by local males (displaying the Occidental chlorotype) over generations, leading to new populations with a predominantly western nuclear genome and Oriental chlorotype. This would be in accordance with the fact that date palm cultivation has always been based on the selection of female accessions.

In the STRUCTURE analysis based on microsatellites, we have noted the presence of admix genotypes (33 out of the 295) (self-population assignment <0·85). These genotypes could represent hybrids between eastern and western plants and are significant in the border area between Asia and Africa. This hypothesis has been confirmed by the presence of the two chlorotypes with similar frequencies in the Egyptian samples. This can be explained by the fact that Egypt is a strategic area crossed by most nomads and pilgrims who imported date fruits from the Middle East to the Maghreb and vice versa. However, the high frequency of the Oriental chlorotype in Mauritanian and Algerian date palms could also be the result of recent introduction or import of accessions.

Conclusions

The results presented here show that date palm displays high genetic diversity and that the genetic variation is geographically structured. Indeed, nuclear SSRs have confirmed the existence of two pools named Eastern and Western. Eastern accessions are substantially different from the Western ones, suggesting that they each have their own autochthonous origin. Therefore, this strong differentiation confirms the existence of at least two domestication events. Northern Africa would also have been either a primary or a secondary domestication centre.

The present study gives a detailed picture of date palm genetic diversity across its traditional cultivation area in the Old World. This information can be used to establish core collections useful in breeding strategies for agronomically interesting traits such as fruit quality or resistance to biotic and abiotic stress through genome-wide association studies.

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxfordjournals.org and consist of the following. Table S1: date palm germplasm accessions, region, code, sex, origin and chlorotype. Table S2: characteristics of the 18 microsatellite and the plastid minisatellite markers. Table S3: analysis of significant linkage disequilibrium. Figure S1: histogram illustrating the frequency distributions of alleles for each microsatellite among groups. Figure S2: ΔK plotted against K values.