Tweet
Background
The Sahelian Fulani are the largest nomadic pastoral ethnic group. Their origins are still largely unknown and their Eurasian genetic component is usually explained by recent admixture events with northern African groups. However, it has also been proposed that Fulani may be the descendants of ancient groups settled in the Sahara during its last Green phase (12000-5000 BP), as also suggested by Y chromosome results.
We produced 23 high-coverage (30 ×) whole genomes from Fulani individuals from 8 Sahelian countries, plus 17 samples from other African groups and 3 Europeans as controls, for a total of 43 new whole genome sequences. These data have been compared with 814 published modern whole genomes and analyzed together with relevant published ancient individuals (for a total of > 1800 samples). These analyses showed that the non-sub-Saharan genetic ancestry component of Fulani cannot be only explained by recent admixture events, but it could be shaped at least in part by older events by events more ancient than previously reported, possibly tracing its origin to the last Green Sahara.
Conclusions
According to our results, Fulani may be the descendants of Saharan cattle herders settled in that area during the last Green Sahara. The exact ancestry composition of such ghost Saharan population(s) cannot be completely unveiled from modern genomes only, but the joint analysis with the available African ancient samples suggested a similarity between ancient Saharans and Late Neolithic Moroccans.
Background
About one third of the African continent is occupied by the Sahara desert that spans from the Atlantic coast to the Red Sea and represents the widest hot desert on Earth. At the north, the Sahara is bordered by the Mediterranean Sea, while on the south the Sahelian belt separates the desert from tropical forests further south. Because of its geographic position and ecological features, the Sahel is characterized by a intermediate environment mainly composed of savannah and grasslands. The northern part of the Sahelian belt shows a semi-arid climate, where no extensive agriculture is possible, while the more humid southern area allows agricultural practices. However, pastoralist groups can exploit the pastures in the northern Sahel and then move southwards during the driest months. So, the Sahel represents a natural contact area between the pastoralist and agricultural lifestyle, represented by groups speaking languages from three out of four African linguistic families, namely Afro-Asiatic, Nilo-Saharan and Niger-Congo [1].
Among the people living in the Sahel, the Fulani, speaking a language belonging to the Niger-Congo linguistic family, represent an interesting case. They are the largest pastoral ethnic group in the world, with an estimated population of 20-40 million people settled in a broad area covering 18 African countries, from the Atlantic coast to the lake Chad basin and further east to the Blue Nile region in Sudan [2]. Fulani are historically nomadic pastoralists, although most of them have now adopted a sedentary lifestyle based on farming. Nowadays, most Fulani live in western Africa, where the first evidence of their presence traces back to the XI century, when they settled on the Fouta Djallon highlands in central Guinea. Later, they moved eastward along the Niger river and they arrived at the lake Chad region in the XV century [3,4]. Despite their recent history being relatively known, their origins are still a matter of debate. Over time, it has been proposed that their ancestors could be ancient Egyptians, Nubians, Persians, Jews, Arabs, Ethiopians or western Africans; however, the most widely accepted theories trace the Fulani origins back to ancient northern Africans or near eastern populations [5–8]. In this context, it has also been proposed that the Fulani may be the descendants of ancient Saharan populations [1]. Indeed, the Sahara has not always been as harsh as today: between 12,000 and 5,000 before present (BP), it was a lush and fertile environment occupied by savannah, forests and a wide hydrogeographic network of rivers and lakes. This phase, generally known as “Green Sahara”, was just the last of several alternating arid and humid phases that have characterized the geological history of this region [9,10]. This region was inhabited by different human groups with peculiar material cultures during the last Green Sahara period, as suggested by several paleoanthropological and archeological findings [9], including several Saharan rock paintings as the rock art in the Tassili-n-Ajjer plateau in the Algerian Sahara, dated to 8,000 BP. Interestingly, this painting represents cattle and pastoralism rituals very similar to the ceremonies still practiced by present-day Fulani [1,11], suggesting a possible link between these people and the ancient inhabitants of the Sahara.
The ancient Saharan hypothesis about the Fulani origins seems also to be supported by some analyses of the human Y chromosome: Fulani show a high proportion of a Y lineage, i.e. E1b-M2/Z15939, dating back to the last Green Sahara. This lineage is observed at lower frequencies also in western and northern Africa, while it is absent in other sub-Saharan regions, pointing to a Green Saharan origin in the Fulani ancestors [12]. However, other genetic data seem to give contrasting results. Considering the other human uniparental system, i.e. the mitochondrial DNA (mtDNA), Fulani harbor a non negligible proportion of Eurasian U5b and H1cb haplotypes, possibly arrived from a northern African source [13,14]. As for the autosomal DNA, genome-wide studies have shown that Fulani people display both a western African component and a northern African/Eurasian one, but the estimates of such components differ across different studies and sub-groups [15–19]. In this context, the most striking result has been obtained from the lactase locus: Fulani show a high proportion of the lactase persistence allele T-13910 [20], that is typical of the European populations [21–23]. This finding has been explained suggesting that the European lactase persistence haplotype was first introduced in northern Africa and then in the Fulani population by recent admixture with a northern African population about 2000 years ago [15]. Similarly, the presence of European ancestry components on the autosomal or uniparental portions of the Fulani genome is usually explained by recent admixture between a Western or Central African source and an “Eurasian” source dating back to the last two millennia [18].
[...]
Results
Whole genome diversity and population structure of the Fulani people We selected 23 Fulani samples from 8 countries in the Sahelian belt. To this group of samples, 17 samples from other African populations and 3 European individuals have been added, for a total of 43 samples, representing 14 different populations, 9 African linguistic sub-families and 13 sampling countries, analyzed by high coverage (> 30 ×) whole genome sequencing (Table 1).
[...]
First, we performed a principal component analysis (PCA, Fig. 2): in general, most of the variability can be explained by geographic location and ethnic group affiliations, with the first PC separating the southern African Khoe-San speaking individuals and some central African Rainforest hunter-gatherers from all the others, while the second PC separates Africans and non Africans, according to previous studies [18,30]
[...]
Looking at the PCA in more detail, we can observe that three groups of samples form long stretches from central/western African cluster to northern African and Eurasian clusters: 1) the eastern African samples (shades of green); 2) the African admixed (grey); 3) the Fulani individuals (red/orange triangles). The eastern African group is the most dispersed along both PC1 and PC2, with people from the Great Lakes region (dark green) nearer to the Bantu groups (including Luhya) and people from the Horn (light green) shifted towards the Middle East/Europe along PC2, consistently with their known Eurasiatic ancestry component [41]. The second group is composed by the two African admixed populations, African Caribbean in Barbados (ACB) and African in Southwest US (ASW) from the 1000 Genomes project [29]: consistently with their known history related to the trans-Atlantic slave trade, these samples occupy a long stretch along PC2, with some individuals tightly clustering with western/central African people and some others shifted towards the Europeans, while their variation is very low along PC1 [42]. Finally, the third group is represented by the Fulani people (orange/red triangles). Differently from the two previous groups, the Fulani cluster is more homogeneous along both PCs, occupying a defined region in the PCA space, between western/central Africa and the non-African regions along PC2. Along the PC1, the position of the Fulani cluster is above the line of the northern African sample, differently from eastern African and African admixed groups that lie below. This observed difference between the Fulani and the other two groups with a well-known Eurasian component may suggest a different source of the non-sub-Saharan ancestry in the Fulani genomes, in addition to differences between the groups contributing to their sub-Saharan ancestry. Despite the Fulani general homogeneity, it is worth noting that not all the individuals fall in the Fulani cluster: indeed, some of them fall in the western/central African groups or among the African-admixed samples, suggesting that admixture events between Fulani and different neighboring populations may have shaped their genetic pool at some extent [15,43].
We further investigate the population structure of the Fulani people compared to the other African and non-African groups by performing an admixture analysis [44], for K from 2 to 10 (Additional File 3, Supplementary Figure S1). For the sake of clarity, hereafter we refer to the genetic ancestry components with the name of the regions/populations mainly representing them. At K=2, the African (blue) and non-African (orange) ancestry components are clearly separated, while K=3 distinguishes the African hunter-gatherer ancestry (yellow), as previously observed [45]. At K=4, when the cross validation error minimizes, we can see a northern African/Middle Eastern ancestry component (light blue), different from the European one: interestingly, this component is observed both in Fulani and in the eastern African groups, that have a known admixture history with the northern Africa/Middle East [46,41]. On the other hand, this ancestry component is absent in the African Admixed ACB and ASW [29], which instead show a non-African ancestry of European origin. At K=5, we can observe an eastern African ancestry component (dark blue), different from the northern African/Middle Eastern one and present in the groups from eastern Africa. At K=6, the northern African ancestry component (purple) is distinguished from the Middle Eastern one: among the Fulani, this new component replaces the old one and represents the second most frequent component (37% on average). Higher Ks further define the Fulani sub-Saharan ancestry component (violet), that ca be defined as a Senegambian component since it is representative (~ 100%) of these people (i.e. Wolof, Mende and Mandenka), while it decreases south- and eastward. This admixture analysis further confirms the presence of two subgroups among our Fulani samples, clearly differentiated by their ancestry composition. On the basis of both the PCA and admixture results, we then subset the Fulani in two groups: hereafter, FulaniA and FulaniB will refer respectively to the samples within the main Fulani PCA/admixture cluster or outside it (Additional File 1, Supplementary Table S1).
Figure 3. Admixture analysis on ancient and modern dataset for K=5 and K=6. Each bar represents a different individual, each color a different genetic ancestry component. Only a subset of individuals has been represented, selecting only relevant groups and randomly subsampling groups larger than 25 (see Additional File 3, Supplementary Figure S2 for the full version of the admixture plot).
The same components are also observed in the FulaniB and in both the African Admixed groups, but at drastically different proportions, suggesting that these groups experienced different evolutionary trajectories. At K=6 (second lowest cross validation error), we observe that the non-sub-Saharan ancestry component in FulaniA is mainly characterized by the new orange component. Interestingly, this component is typical of north-western African individuals older than 8000 BP (more specifically, Taforalt and IAM individuals) [47,48] and can be also observed in other Sahelian/Eastern African groups, while it is virtually absent from African Admixed and other sub-Saharan populations (Fig. 3). On the other hand, the Iranian Neolithic blue ancestry component is not represented anymore among FulaniA. Higher Ks confirmed this pattern where FulaniA share a non-negligible portion of their non sub-Saharan ancestry with ancient northern African people. As for the FulaniA sub-Saharan ancestry component (yellow), this analysis confirmed the link with the Wolof and Mende groups already highlighted by the admixture based on modern data only. We further explored the link between Fulani (A and B) and ancient people by performing D-statistics in the form D(Fulani, sub-Saharan Africa, Eurasia/North Africa, Chimpanzee), where a) Fulani were alternatively represented by FulaniA or FulaniB; b) excluding the Southern Africa and Northern European regions from sub-Saharan Africa and Eurasia/North Africa respectively; c) testing both FulaniA and FulaniB also as sub-Saharan groups and d) using Chimpanzee as outgroup. Overall, the Eurasia/North Africa pool tended to share more alleles with both Fulani clusters rather than central/western African people and less compared to eastern African populations (both ancient and modern). However, when focusing on the relationships between the two Fulani clusters and the ancient north-western African individuals in the form D (FulaniA,FulaniB, ancient North-western Africa, Chimpanzee), we observed that FulaniA were closer to all the 3 ancient north-western African groups than FulaniB (Additional File 1, Table S3 and Additional File 3, Figure S3). When analyzed by qpWave/qpAdm framework with two putative sources, both Fulani groups were modeled by an ancient sample from Cameroon (about 7000 BP) and the Early Neolithic Moroccans (about 6000 BP), with Iberomaurusian-like component being higher among FulaniA (Additional File 1, Table S3A). In addition, FulaniA were also successfully modeled by modern Berbers and modern Sahelian people. When 3 putative sources were tested, we observed that FulaniA can be only modeled by the same two previous ancient sources plus the eastern Sahelian Agew at approximately the same proportions, while FulaniB could be successfully modeled by two sub-Saharan sources and a northern African one, with the latter accounting for no more than 27% (Additional File 1, Table S3B). Finally, we also performed 10 admixture graphs varying the number of admixture edges from 2 to 5 and retained the three graphs characterized by the lowest score, testing only the FulaniA. For all the inferred graphs FulaniA were modeled as composed of two main ancestries. The first ancestry (between 60% and 70%) is close to Western African groups related to Mende from Sierra Leone (MSL). This group also harbored a minor proportion from a ghost population which stems as an outgroup of all the other populations analyzed, as already suggested in previous studies [49]. The second ancestry (more than 30%) was derived from an unsampled group close to Iberomaurusian samples, modeled as a sister group of ancient Northern Africans, or alternatively, as an outgroup of all the “Eurasian-ancestry” enriched groups (Additional File 3, Supplementary Figure S4). This may suggest that, at least part of the non-sub-Saharan ancestry in present day Fulani might have an ancient origin, and is not exclusively derived by migrations in the last two millennia.
Patterns of shared haplotypes and runs of homozygosity
To infer patterns of shared ancestry between Fulani groups and other populations, we identified and analyzed the identity-by-descent (IBD) segments longer than 2 cM in our modern dataset (Additional File1, Table S5). FulaniA shared more IBD segments with themselves, while FulaniB shared more IBD with FulaniA and themselves. Interestingly, we could observe a general pattern differentiating FulaniA and FulaniB, with the former sharing more segments with Sahelian, eastern African, northern African and Eurasian populations, while the latter shared more IBD with western-central African groups. When comparing short (< 5 cM) and long (> 5 cM) IBD segments, we could observe that this pattern was mainly driven by the shortest IBD segments, suggesting that it formed in more ancient time, while differences between the two Fulani clusters are more nuanced for the long IBD segments. We further explored the links between Fulani and neighboring sub-Saharan people by performing D-statistics in the form D(Sahel, West Africa, Fulani, Chimpanzee), where Fulani could be alternatively FulaniA or FulaniB (Additional file 1, Supplementary Table S6). Overall, both Fulani clusters were characterized by the same general pattern, although FulaniA showed a slight trend toward Sahelian groups speaking Nilo-Saharan languages, with the notable exception of the Wolof from Western Africa, in line with IBD results. We then investigated the patterns of runs of homozygosity (ROH) to shed light on the demographic history of the Fulani (Additional File 1, Supplementary Table S7). The two Fulani clusters showed differences, with FulaniB being more similar to other western and central African groups both in terms of number of ROHs and their total length (Fig.4).
Figure 4: Analysis of runs of homozygosity (ROH). Each boxplot summarizes the total length of ROH segments by population group (see Additional File 1, Supplementary Table S1). Observations have been defined as outlier if they were 1.5 times the interquartile range less than the first quartile or greater than the third quartile. Population groups with a sample size less than 3 have not been plotted.
On the other hand, FulaniA showed more ROHs compared to their neighboring groups, with a pattern between Sahelian/central African populations on one hand and northern African populations on the other. We further explored the ROH pattern subsetting the Fulani clusters in Sahelian or western African groups and we observed that FulaniA and FulaniB individuals from western Africa showed very similar ROH patterns (Additional File 3, Supplementary Figure S5A). On the contrary, Sahelian FulaniA showed a higher number of ROHs; when analyzing the short (< 1.5 Mb) and long (> 1.5 Mb) ROHs separately (Additional File 3, Supplementary Figure S5B-C), we observed that this pattern was mainly driven by the long segments. On the other hand, the short ROH segments recreated the clear differentiation between FulaniA and FulaniB, regardless of their sampling region, with the former showing the number and the mean sum of ROHs more similar to eastern Africans rather than western/central African groups.
A Green Sahara story
Most recent studies based on the genome-wide variability suggested that the origin of the non-sub-Saharan ancestry of Fulani traces back to relatively recent times (1800-300 BP), after admixture events between a western African and northern African/southern European source [15,18,37,38]. Here, we analyzed the Fulani genomes in the frame of the worldwide genetic variability considering both modern and ancient individuals. In this context, the analysis of aDNA allowed us to characterize the non-sub-Saharan ancestry component in the Fulani in more detail, shedding new light on its origin. By comparing the Fulani genetic diversity with those of the large number of ancient individuals currently available [40] through an admixture analysis (K=6, Fig. 3), we observed that their non-sub-Saharan ancestry is characterized by a component observed in extremely high proportions (virtually 100%) also in ancient Morroccans older than 8000 BP [47,50], and an azure Levantine component. We could also see a non-negligible fraction of the WHG red component, as in the Late Neolithic Moroccan individuals, possibly arrived from Iberia to the Maghreb [47]. On the other hand, Fulani lacked the Iran Neolithic/CHG blue component, which was observed instead in modern northern Africans, Middle Easterners and Europeans. This pattern of ancestry components can be hardly explained solely by recent admixture between western African and northern African groups. Indeed, in this case, we would have expected to observe also the Iran Neolithic/CHG blue component in the Fulani, since this component is present in all the modern northern African, European and Middle Eastern samples. So, the admixture event(s) that forged the non-sub-Saharan ancestry in Fulani should have occurred before the arrival of this component in the areas to the north of the Sahara. It is challenging to obtain exact time estimates for such events [51], considering that they were probably followed by later admixture events involving the same or very similar source populations [15,18,37,38]. However, considering the results of the admixture analysis (K=6), we can try to define broad time windows. Indeed, considering the presence of the azure (Levantine) and red (WHG) components in Fulani, in addition to the orange (Iberomaurusian) one, and the absence of the blue (Iran Neolithic/CHG) component, we can postulate that their non-sub-Saharan component dates back to about 8000-7000 years ago from a source population similar (except for the blue component) to the Late Neolithic Moroccans (dated about 5000 BCE, considering their radiocarbon time estimates) (Additional 2, Supplementary Table S2). In addition, the qpAdm analysis also points to a period corresponding to 7000 BP considering the radiocarbon time estimates of the two ancient sources that successfully modelled both FulaniA and FulaniB (Additional 2, Supplementary Table S4). Intriguingly, this time window corresponds to the period of the last Green Sahara.
The Sahara has played and still plays an important role in the population history of the African continent. Currently, it acts as a strong geographic barrier because of its arid environment; however, during its humid phases (called Green Sahara), it became a fertile and lush land inhabited by different species, including humans [10]. The last of these fertile periods occurred approximately between 12000 and 5000 BP, during the Holocene climatic optimum, and the archaeological findings from that time are characterized by different material cultures [9], testifying the rich diversity of the populations in that region. Indeed, the fertile environment of the Sahara, with extensive hydrogeological networks connecting rivers and lakes, may have promoted extensive contacts between the different groups settled in this area. The last Green Sahara was also characterized by social and cultural changes, such as the spreading of pastoralism
and farming, both developed locally or arrived from the Middle East to northeastern Africa between 7000 and 6000 years ago [1,52,53]. In this scenario, the formation of the non-sub-Saharan component in Fulani should not be interpreted as a simple admixture event between a western African source group and a northern African one. On the contrary, we should consider that many different groups inhabited the Saharan area at that time [53,54] and the favorable climatic conditions could have allowed a relatively continuous gene flow between them (Fig. 5), as also suggested by the presence of a western African ancestry component in the Iberomaurusians [50] and in addition to the main northeast-to-northwest and northeast-to-east routes [48,55,56].
Figure 5: Movements of main non-sub-Saharan ancestry components in the Green Sahara 8000-7000 BP and a possible scenario for the formation of the Green Saharan population(s). The ancestry components are the same represented in the admixture with modern and ancient individuals for the same populations (Fig.3 and Additional File 3, Supplementary Figure S2). Faded arrows represent possible movements across the Green Sahara.
These Saharan people may have actually been the ancestors of present-day Sahelian people, including Fulani. This is also supported by our admixture analysis, showing that also the Neolithic pastoralists from Kenya and the modern central Sahelian groups showed the orange and azure components at non-negligible proportions (Fig. 3 and Additional File 3, Supplementary Figure S2). Moreover, the absence of the blue component in the modern Nilo-Saharan groups from both central and eastern Sahel further confirms this hypothesis, since it has been proposed that the Nilo-Saharan languages were much more widespread in the Green Sahara region during the last humid phase, to be replaced by Afro-Asiatic languages only in more recent times (about 1500 years ago) [9,53]. In this context, it is also worth noting that a recent whole genome study further confirmed the presence of a shared ancestry between Fulani and Afro-Asiatic speakers from Eastern Africa [37]. Interestingly, it has also been proposed that the Green Sahara area (more precisely the lake Chad basin, occupied by the Megalake Chad at the time [9]) was also the homeland of the Niger-Congo (i.e. the language currently spoken by Fulani) about 7000 years ago, intriguingly corresponding to the possible time estimate of the Fulani non-sub-Saharan components. This language then spread westwards (and eastwards to a lesser extent) replacing the pre-existing languages [57].
The exact genetic ancestry composition of the “ghost” Saharan populations cannot be assessed without ancient individuals from that area at that time; however, on the basis of the ancient data currently available, we can propose that they were genetically similar to the Late Neolithic Moroccans here analyzed, although this group already shows the Iranian Neolithic/CHG component that was probably more diluted further south, while the yellow western African component was possibly present at higher proportions (Fig.5).
With the end of the Green Sahara and the subsequent gradual desertification of that region, the Saharan groups moved westwards, eastwards, southwards or northwards, as also suggested by the Y chromosome data [12]. This phenomenon was not abrupt and was slightly longer in the central Sahel rather than in the east depending on the local hydrogeological conditions; according to the archaeological evidence, the changing climate conditions led to changes in the socio-economic organization of the different groups. In particular, it has been suggested that cattle pastoralism became the predominant form of subsistence at that time since it was a more reliable source of food, leading to the establishment of a pan-Saharan cattle cult, as testified by rock paintings and rituals that spread westwards with the pastoral groups in search of new pastures [54,58]. This scenario may also reconcile the apparent contrasting evidence from different genetic systems linking Fulani with sub-Saharan, northern African, European or eastern African groups [12,15,17,37,38,43,59–61]; indeed, if we assume the past presence of Green Saharan populations with extensive contacts, it is not surprising that these groups could share genetic affinities. With the Sahara desertification and the following different population dynamics, the genetic differentiation between the different groups may have led to the loss or maintenance of different genetic variants because of genetic drift and/or selection.
All this bulk of data seems to point to a Green Sahara origin of the Fulani with regard to their non-sub-Saharan component. Their ancestors were possibly Saharan cattle herders that moved westwards in response to the changing climate and then mixed with local people. Later, the peculiar Fulani lifestyle, historically characterized by nomadism and endogamy [59,62], and their population size dramatic decrease [37] could have prevented the dilution of this ancient “Green Saharan” ancestry component that can be still identified.
Finally, both Fulani clusters show the same ancestry components, pointing to a common ancestry of the two groups, although at different proportions, suggesting different population dynamics.
https://www.biorxiv.org/content/10.1...569v1.full.pdf
The Sahelian Fulani are the largest nomadic pastoral ethnic group. Their origins are still largely unknown and their Eurasian genetic component is usually explained by recent admixture events with northern African groups. However, it has also been proposed that Fulani may be the descendants of ancient groups settled in the Sahara during its last Green phase (12000-5000 BP), as also suggested by Y chromosome results.
We produced 23 high-coverage (30 ×) whole genomes from Fulani individuals from 8 Sahelian countries, plus 17 samples from other African groups and 3 Europeans as controls, for a total of 43 new whole genome sequences. These data have been compared with 814 published modern whole genomes and analyzed together with relevant published ancient individuals (for a total of > 1800 samples). These analyses showed that the non-sub-Saharan genetic ancestry component of Fulani cannot be only explained by recent admixture events, but it could be shaped at least in part by older events by events more ancient than previously reported, possibly tracing its origin to the last Green Sahara.
Conclusions
According to our results, Fulani may be the descendants of Saharan cattle herders settled in that area during the last Green Sahara. The exact ancestry composition of such ghost Saharan population(s) cannot be completely unveiled from modern genomes only, but the joint analysis with the available African ancient samples suggested a similarity between ancient Saharans and Late Neolithic Moroccans.
Background
About one third of the African continent is occupied by the Sahara desert that spans from the Atlantic coast to the Red Sea and represents the widest hot desert on Earth. At the north, the Sahara is bordered by the Mediterranean Sea, while on the south the Sahelian belt separates the desert from tropical forests further south. Because of its geographic position and ecological features, the Sahel is characterized by a intermediate environment mainly composed of savannah and grasslands. The northern part of the Sahelian belt shows a semi-arid climate, where no extensive agriculture is possible, while the more humid southern area allows agricultural practices. However, pastoralist groups can exploit the pastures in the northern Sahel and then move southwards during the driest months. So, the Sahel represents a natural contact area between the pastoralist and agricultural lifestyle, represented by groups speaking languages from three out of four African linguistic families, namely Afro-Asiatic, Nilo-Saharan and Niger-Congo [1].
Among the people living in the Sahel, the Fulani, speaking a language belonging to the Niger-Congo linguistic family, represent an interesting case. They are the largest pastoral ethnic group in the world, with an estimated population of 20-40 million people settled in a broad area covering 18 African countries, from the Atlantic coast to the lake Chad basin and further east to the Blue Nile region in Sudan [2]. Fulani are historically nomadic pastoralists, although most of them have now adopted a sedentary lifestyle based on farming. Nowadays, most Fulani live in western Africa, where the first evidence of their presence traces back to the XI century, when they settled on the Fouta Djallon highlands in central Guinea. Later, they moved eastward along the Niger river and they arrived at the lake Chad region in the XV century [3,4]. Despite their recent history being relatively known, their origins are still a matter of debate. Over time, it has been proposed that their ancestors could be ancient Egyptians, Nubians, Persians, Jews, Arabs, Ethiopians or western Africans; however, the most widely accepted theories trace the Fulani origins back to ancient northern Africans or near eastern populations [5–8]. In this context, it has also been proposed that the Fulani may be the descendants of ancient Saharan populations [1]. Indeed, the Sahara has not always been as harsh as today: between 12,000 and 5,000 before present (BP), it was a lush and fertile environment occupied by savannah, forests and a wide hydrogeographic network of rivers and lakes. This phase, generally known as “Green Sahara”, was just the last of several alternating arid and humid phases that have characterized the geological history of this region [9,10]. This region was inhabited by different human groups with peculiar material cultures during the last Green Sahara period, as suggested by several paleoanthropological and archeological findings [9], including several Saharan rock paintings as the rock art in the Tassili-n-Ajjer plateau in the Algerian Sahara, dated to 8,000 BP. Interestingly, this painting represents cattle and pastoralism rituals very similar to the ceremonies still practiced by present-day Fulani [1,11], suggesting a possible link between these people and the ancient inhabitants of the Sahara.
The ancient Saharan hypothesis about the Fulani origins seems also to be supported by some analyses of the human Y chromosome: Fulani show a high proportion of a Y lineage, i.e. E1b-M2/Z15939, dating back to the last Green Sahara. This lineage is observed at lower frequencies also in western and northern Africa, while it is absent in other sub-Saharan regions, pointing to a Green Saharan origin in the Fulani ancestors [12]. However, other genetic data seem to give contrasting results. Considering the other human uniparental system, i.e. the mitochondrial DNA (mtDNA), Fulani harbor a non negligible proportion of Eurasian U5b and H1cb haplotypes, possibly arrived from a northern African source [13,14]. As for the autosomal DNA, genome-wide studies have shown that Fulani people display both a western African component and a northern African/Eurasian one, but the estimates of such components differ across different studies and sub-groups [15–19]. In this context, the most striking result has been obtained from the lactase locus: Fulani show a high proportion of the lactase persistence allele T-13910 [20], that is typical of the European populations [21–23]. This finding has been explained suggesting that the European lactase persistence haplotype was first introduced in northern Africa and then in the Fulani population by recent admixture with a northern African population about 2000 years ago [15]. Similarly, the presence of European ancestry components on the autosomal or uniparental portions of the Fulani genome is usually explained by recent admixture between a Western or Central African source and an “Eurasian” source dating back to the last two millennia [18].
[...]
Results
Whole genome diversity and population structure of the Fulani people We selected 23 Fulani samples from 8 countries in the Sahelian belt. To this group of samples, 17 samples from other African populations and 3 European individuals have been added, for a total of 43 samples, representing 14 different populations, 9 African linguistic sub-families and 13 sampling countries, analyzed by high coverage (> 30 ×) whole genome sequencing (Table 1).
[...]
First, we performed a principal component analysis (PCA, Fig. 2): in general, most of the variability can be explained by geographic location and ethnic group affiliations, with the first PC separating the southern African Khoe-San speaking individuals and some central African Rainforest hunter-gatherers from all the others, while the second PC separates Africans and non Africans, according to previous studies [18,30]
[...]
Looking at the PCA in more detail, we can observe that three groups of samples form long stretches from central/western African cluster to northern African and Eurasian clusters: 1) the eastern African samples (shades of green); 2) the African admixed (grey); 3) the Fulani individuals (red/orange triangles). The eastern African group is the most dispersed along both PC1 and PC2, with people from the Great Lakes region (dark green) nearer to the Bantu groups (including Luhya) and people from the Horn (light green) shifted towards the Middle East/Europe along PC2, consistently with their known Eurasiatic ancestry component [41]. The second group is composed by the two African admixed populations, African Caribbean in Barbados (ACB) and African in Southwest US (ASW) from the 1000 Genomes project [29]: consistently with their known history related to the trans-Atlantic slave trade, these samples occupy a long stretch along PC2, with some individuals tightly clustering with western/central African people and some others shifted towards the Europeans, while their variation is very low along PC1 [42]. Finally, the third group is represented by the Fulani people (orange/red triangles). Differently from the two previous groups, the Fulani cluster is more homogeneous along both PCs, occupying a defined region in the PCA space, between western/central Africa and the non-African regions along PC2. Along the PC1, the position of the Fulani cluster is above the line of the northern African sample, differently from eastern African and African admixed groups that lie below. This observed difference between the Fulani and the other two groups with a well-known Eurasian component may suggest a different source of the non-sub-Saharan ancestry in the Fulani genomes, in addition to differences between the groups contributing to their sub-Saharan ancestry. Despite the Fulani general homogeneity, it is worth noting that not all the individuals fall in the Fulani cluster: indeed, some of them fall in the western/central African groups or among the African-admixed samples, suggesting that admixture events between Fulani and different neighboring populations may have shaped their genetic pool at some extent [15,43].
We further investigate the population structure of the Fulani people compared to the other African and non-African groups by performing an admixture analysis [44], for K from 2 to 10 (Additional File 3, Supplementary Figure S1). For the sake of clarity, hereafter we refer to the genetic ancestry components with the name of the regions/populations mainly representing them. At K=2, the African (blue) and non-African (orange) ancestry components are clearly separated, while K=3 distinguishes the African hunter-gatherer ancestry (yellow), as previously observed [45]. At K=4, when the cross validation error minimizes, we can see a northern African/Middle Eastern ancestry component (light blue), different from the European one: interestingly, this component is observed both in Fulani and in the eastern African groups, that have a known admixture history with the northern Africa/Middle East [46,41]. On the other hand, this ancestry component is absent in the African Admixed ACB and ASW [29], which instead show a non-African ancestry of European origin. At K=5, we can observe an eastern African ancestry component (dark blue), different from the northern African/Middle Eastern one and present in the groups from eastern Africa. At K=6, the northern African ancestry component (purple) is distinguished from the Middle Eastern one: among the Fulani, this new component replaces the old one and represents the second most frequent component (37% on average). Higher Ks further define the Fulani sub-Saharan ancestry component (violet), that ca be defined as a Senegambian component since it is representative (~ 100%) of these people (i.e. Wolof, Mende and Mandenka), while it decreases south- and eastward. This admixture analysis further confirms the presence of two subgroups among our Fulani samples, clearly differentiated by their ancestry composition. On the basis of both the PCA and admixture results, we then subset the Fulani in two groups: hereafter, FulaniA and FulaniB will refer respectively to the samples within the main Fulani PCA/admixture cluster or outside it (Additional File 1, Supplementary Table S1).
Figure 3. Admixture analysis on ancient and modern dataset for K=5 and K=6. Each bar represents a different individual, each color a different genetic ancestry component. Only a subset of individuals has been represented, selecting only relevant groups and randomly subsampling groups larger than 25 (see Additional File 3, Supplementary Figure S2 for the full version of the admixture plot).
The same components are also observed in the FulaniB and in both the African Admixed groups, but at drastically different proportions, suggesting that these groups experienced different evolutionary trajectories. At K=6 (second lowest cross validation error), we observe that the non-sub-Saharan ancestry component in FulaniA is mainly characterized by the new orange component. Interestingly, this component is typical of north-western African individuals older than 8000 BP (more specifically, Taforalt and IAM individuals) [47,48] and can be also observed in other Sahelian/Eastern African groups, while it is virtually absent from African Admixed and other sub-Saharan populations (Fig. 3). On the other hand, the Iranian Neolithic blue ancestry component is not represented anymore among FulaniA. Higher Ks confirmed this pattern where FulaniA share a non-negligible portion of their non sub-Saharan ancestry with ancient northern African people. As for the FulaniA sub-Saharan ancestry component (yellow), this analysis confirmed the link with the Wolof and Mende groups already highlighted by the admixture based on modern data only. We further explored the link between Fulani (A and B) and ancient people by performing D-statistics in the form D(Fulani, sub-Saharan Africa, Eurasia/North Africa, Chimpanzee), where a) Fulani were alternatively represented by FulaniA or FulaniB; b) excluding the Southern Africa and Northern European regions from sub-Saharan Africa and Eurasia/North Africa respectively; c) testing both FulaniA and FulaniB also as sub-Saharan groups and d) using Chimpanzee as outgroup. Overall, the Eurasia/North Africa pool tended to share more alleles with both Fulani clusters rather than central/western African people and less compared to eastern African populations (both ancient and modern). However, when focusing on the relationships between the two Fulani clusters and the ancient north-western African individuals in the form D (FulaniA,FulaniB, ancient North-western Africa, Chimpanzee), we observed that FulaniA were closer to all the 3 ancient north-western African groups than FulaniB (Additional File 1, Table S3 and Additional File 3, Figure S3). When analyzed by qpWave/qpAdm framework with two putative sources, both Fulani groups were modeled by an ancient sample from Cameroon (about 7000 BP) and the Early Neolithic Moroccans (about 6000 BP), with Iberomaurusian-like component being higher among FulaniA (Additional File 1, Table S3A). In addition, FulaniA were also successfully modeled by modern Berbers and modern Sahelian people. When 3 putative sources were tested, we observed that FulaniA can be only modeled by the same two previous ancient sources plus the eastern Sahelian Agew at approximately the same proportions, while FulaniB could be successfully modeled by two sub-Saharan sources and a northern African one, with the latter accounting for no more than 27% (Additional File 1, Table S3B). Finally, we also performed 10 admixture graphs varying the number of admixture edges from 2 to 5 and retained the three graphs characterized by the lowest score, testing only the FulaniA. For all the inferred graphs FulaniA were modeled as composed of two main ancestries. The first ancestry (between 60% and 70%) is close to Western African groups related to Mende from Sierra Leone (MSL). This group also harbored a minor proportion from a ghost population which stems as an outgroup of all the other populations analyzed, as already suggested in previous studies [49]. The second ancestry (more than 30%) was derived from an unsampled group close to Iberomaurusian samples, modeled as a sister group of ancient Northern Africans, or alternatively, as an outgroup of all the “Eurasian-ancestry” enriched groups (Additional File 3, Supplementary Figure S4). This may suggest that, at least part of the non-sub-Saharan ancestry in present day Fulani might have an ancient origin, and is not exclusively derived by migrations in the last two millennia.
Patterns of shared haplotypes and runs of homozygosity
To infer patterns of shared ancestry between Fulani groups and other populations, we identified and analyzed the identity-by-descent (IBD) segments longer than 2 cM in our modern dataset (Additional File1, Table S5). FulaniA shared more IBD segments with themselves, while FulaniB shared more IBD with FulaniA and themselves. Interestingly, we could observe a general pattern differentiating FulaniA and FulaniB, with the former sharing more segments with Sahelian, eastern African, northern African and Eurasian populations, while the latter shared more IBD with western-central African groups. When comparing short (< 5 cM) and long (> 5 cM) IBD segments, we could observe that this pattern was mainly driven by the shortest IBD segments, suggesting that it formed in more ancient time, while differences between the two Fulani clusters are more nuanced for the long IBD segments. We further explored the links between Fulani and neighboring sub-Saharan people by performing D-statistics in the form D(Sahel, West Africa, Fulani, Chimpanzee), where Fulani could be alternatively FulaniA or FulaniB (Additional file 1, Supplementary Table S6). Overall, both Fulani clusters were characterized by the same general pattern, although FulaniA showed a slight trend toward Sahelian groups speaking Nilo-Saharan languages, with the notable exception of the Wolof from Western Africa, in line with IBD results. We then investigated the patterns of runs of homozygosity (ROH) to shed light on the demographic history of the Fulani (Additional File 1, Supplementary Table S7). The two Fulani clusters showed differences, with FulaniB being more similar to other western and central African groups both in terms of number of ROHs and their total length (Fig.4).
Figure 4: Analysis of runs of homozygosity (ROH). Each boxplot summarizes the total length of ROH segments by population group (see Additional File 1, Supplementary Table S1). Observations have been defined as outlier if they were 1.5 times the interquartile range less than the first quartile or greater than the third quartile. Population groups with a sample size less than 3 have not been plotted.
On the other hand, FulaniA showed more ROHs compared to their neighboring groups, with a pattern between Sahelian/central African populations on one hand and northern African populations on the other. We further explored the ROH pattern subsetting the Fulani clusters in Sahelian or western African groups and we observed that FulaniA and FulaniB individuals from western Africa showed very similar ROH patterns (Additional File 3, Supplementary Figure S5A). On the contrary, Sahelian FulaniA showed a higher number of ROHs; when analyzing the short (< 1.5 Mb) and long (> 1.5 Mb) ROHs separately (Additional File 3, Supplementary Figure S5B-C), we observed that this pattern was mainly driven by the long segments. On the other hand, the short ROH segments recreated the clear differentiation between FulaniA and FulaniB, regardless of their sampling region, with the former showing the number and the mean sum of ROHs more similar to eastern Africans rather than western/central African groups.
A Green Sahara story
Most recent studies based on the genome-wide variability suggested that the origin of the non-sub-Saharan ancestry of Fulani traces back to relatively recent times (1800-300 BP), after admixture events between a western African and northern African/southern European source [15,18,37,38]. Here, we analyzed the Fulani genomes in the frame of the worldwide genetic variability considering both modern and ancient individuals. In this context, the analysis of aDNA allowed us to characterize the non-sub-Saharan ancestry component in the Fulani in more detail, shedding new light on its origin. By comparing the Fulani genetic diversity with those of the large number of ancient individuals currently available [40] through an admixture analysis (K=6, Fig. 3), we observed that their non-sub-Saharan ancestry is characterized by a component observed in extremely high proportions (virtually 100%) also in ancient Morroccans older than 8000 BP [47,50], and an azure Levantine component. We could also see a non-negligible fraction of the WHG red component, as in the Late Neolithic Moroccan individuals, possibly arrived from Iberia to the Maghreb [47]. On the other hand, Fulani lacked the Iran Neolithic/CHG blue component, which was observed instead in modern northern Africans, Middle Easterners and Europeans. This pattern of ancestry components can be hardly explained solely by recent admixture between western African and northern African groups. Indeed, in this case, we would have expected to observe also the Iran Neolithic/CHG blue component in the Fulani, since this component is present in all the modern northern African, European and Middle Eastern samples. So, the admixture event(s) that forged the non-sub-Saharan ancestry in Fulani should have occurred before the arrival of this component in the areas to the north of the Sahara. It is challenging to obtain exact time estimates for such events [51], considering that they were probably followed by later admixture events involving the same or very similar source populations [15,18,37,38]. However, considering the results of the admixture analysis (K=6), we can try to define broad time windows. Indeed, considering the presence of the azure (Levantine) and red (WHG) components in Fulani, in addition to the orange (Iberomaurusian) one, and the absence of the blue (Iran Neolithic/CHG) component, we can postulate that their non-sub-Saharan component dates back to about 8000-7000 years ago from a source population similar (except for the blue component) to the Late Neolithic Moroccans (dated about 5000 BCE, considering their radiocarbon time estimates) (Additional 2, Supplementary Table S2). In addition, the qpAdm analysis also points to a period corresponding to 7000 BP considering the radiocarbon time estimates of the two ancient sources that successfully modelled both FulaniA and FulaniB (Additional 2, Supplementary Table S4). Intriguingly, this time window corresponds to the period of the last Green Sahara.
The Sahara has played and still plays an important role in the population history of the African continent. Currently, it acts as a strong geographic barrier because of its arid environment; however, during its humid phases (called Green Sahara), it became a fertile and lush land inhabited by different species, including humans [10]. The last of these fertile periods occurred approximately between 12000 and 5000 BP, during the Holocene climatic optimum, and the archaeological findings from that time are characterized by different material cultures [9], testifying the rich diversity of the populations in that region. Indeed, the fertile environment of the Sahara, with extensive hydrogeological networks connecting rivers and lakes, may have promoted extensive contacts between the different groups settled in this area. The last Green Sahara was also characterized by social and cultural changes, such as the spreading of pastoralism
and farming, both developed locally or arrived from the Middle East to northeastern Africa between 7000 and 6000 years ago [1,52,53]. In this scenario, the formation of the non-sub-Saharan component in Fulani should not be interpreted as a simple admixture event between a western African source group and a northern African one. On the contrary, we should consider that many different groups inhabited the Saharan area at that time [53,54] and the favorable climatic conditions could have allowed a relatively continuous gene flow between them (Fig. 5), as also suggested by the presence of a western African ancestry component in the Iberomaurusians [50] and in addition to the main northeast-to-northwest and northeast-to-east routes [48,55,56].
Figure 5: Movements of main non-sub-Saharan ancestry components in the Green Sahara 8000-7000 BP and a possible scenario for the formation of the Green Saharan population(s). The ancestry components are the same represented in the admixture with modern and ancient individuals for the same populations (Fig.3 and Additional File 3, Supplementary Figure S2). Faded arrows represent possible movements across the Green Sahara.
These Saharan people may have actually been the ancestors of present-day Sahelian people, including Fulani. This is also supported by our admixture analysis, showing that also the Neolithic pastoralists from Kenya and the modern central Sahelian groups showed the orange and azure components at non-negligible proportions (Fig. 3 and Additional File 3, Supplementary Figure S2). Moreover, the absence of the blue component in the modern Nilo-Saharan groups from both central and eastern Sahel further confirms this hypothesis, since it has been proposed that the Nilo-Saharan languages were much more widespread in the Green Sahara region during the last humid phase, to be replaced by Afro-Asiatic languages only in more recent times (about 1500 years ago) [9,53]. In this context, it is also worth noting that a recent whole genome study further confirmed the presence of a shared ancestry between Fulani and Afro-Asiatic speakers from Eastern Africa [37]. Interestingly, it has also been proposed that the Green Sahara area (more precisely the lake Chad basin, occupied by the Megalake Chad at the time [9]) was also the homeland of the Niger-Congo (i.e. the language currently spoken by Fulani) about 7000 years ago, intriguingly corresponding to the possible time estimate of the Fulani non-sub-Saharan components. This language then spread westwards (and eastwards to a lesser extent) replacing the pre-existing languages [57].
The exact genetic ancestry composition of the “ghost” Saharan populations cannot be assessed without ancient individuals from that area at that time; however, on the basis of the ancient data currently available, we can propose that they were genetically similar to the Late Neolithic Moroccans here analyzed, although this group already shows the Iranian Neolithic/CHG component that was probably more diluted further south, while the yellow western African component was possibly present at higher proportions (Fig.5).
With the end of the Green Sahara and the subsequent gradual desertification of that region, the Saharan groups moved westwards, eastwards, southwards or northwards, as also suggested by the Y chromosome data [12]. This phenomenon was not abrupt and was slightly longer in the central Sahel rather than in the east depending on the local hydrogeological conditions; according to the archaeological evidence, the changing climate conditions led to changes in the socio-economic organization of the different groups. In particular, it has been suggested that cattle pastoralism became the predominant form of subsistence at that time since it was a more reliable source of food, leading to the establishment of a pan-Saharan cattle cult, as testified by rock paintings and rituals that spread westwards with the pastoral groups in search of new pastures [54,58]. This scenario may also reconcile the apparent contrasting evidence from different genetic systems linking Fulani with sub-Saharan, northern African, European or eastern African groups [12,15,17,37,38,43,59–61]; indeed, if we assume the past presence of Green Saharan populations with extensive contacts, it is not surprising that these groups could share genetic affinities. With the Sahara desertification and the following different population dynamics, the genetic differentiation between the different groups may have led to the loss or maintenance of different genetic variants because of genetic drift and/or selection.
All this bulk of data seems to point to a Green Sahara origin of the Fulani with regard to their non-sub-Saharan component. Their ancestors were possibly Saharan cattle herders that moved westwards in response to the changing climate and then mixed with local people. Later, the peculiar Fulani lifestyle, historically characterized by nomadism and endogamy [59,62], and their population size dramatic decrease [37] could have prevented the dilution of this ancient “Green Saharan” ancestry component that can be still identified.
Finally, both Fulani clusters show the same ancestry components, pointing to a common ancestry of the two groups, although at different proportions, suggesting different population dynamics.
https://www.biorxiv.org/content/10.1...569v1.full.pdf