Tweet
The Entwined African and Asian Genetic Roots of the Medieval Peoples of the Swahili Coast
Abstract
The peoples of the Swahili coast of eastern Africa established a literate urban culture by the second millennium CE. They traded across eastern Africa and the Indian Ocean and were among the first sub-Saharan practitioners of Islam. An open question has been the extent to which these early interactions between Africans and non-Africans were accompanied by genetic admixture. We report genome-wide ancient DNA from 80 individuals in five medieval and early modern (1300-1800 CE) coastal towns, as well as people from an inland town postdating 1650 CE. Over half of the ancestry of most coastal individuals came from African ancestors; these African ancestors were primarily female. A slightly smaller proportion of ancestry was from Asia. This Asian component was approximately eighty to ninety percent from Near Eastern males and ten to twenty percent from Indian females. Peoples of African and Asian origins began to mix by around 1000 CE, a time when archaeological evidence documents changes on the coast that are often interpreted as marking the large-scale adoption of Islam. Before roughly 1500 CE, the Near Eastern ancestry detected in the individuals was mainly Persian-related, consistent with the narrative of the Kilwa Chronicle, the oldest history told by the Swahili themselves. After this time, the sources of Near Eastern ancestry became increasingly Arabian, consistent with the archaeological and historical evidence of growing interactions between the Swahili coast and parts of southern Arabia. Subsequent interactions of Swahili coast peoples with other Asian and African groups further changed the ancestry of present-day peoples relative to the ancient individuals we sequenced, highlighting how Swahili genetic legacies can be more clearly understood with ancient DNA.
First millennium CE sites on the littoral, beginning in the 7th century CE, were part of a shared material culture and practice network across the eastern African region [3, 4] and were fully engaged in Indian Ocean trading networks [5]. The annual southeast monsoons from May to October allowed trading vessels to travel from India or the Arabian Peninsula to the eastern African coast, and the northeast monsoons from November to March enabled their return in the same year [5, 6]. Muslims were present on the coast from the 8th century CE, although probably in a minority [7]. A transition can be identified in the archaeology of the coast during the 11th century, with the establishment of new settlements and the elaboration of older ones with coral-built mosques and tombs, all the while remaining deeply connected to sites throughout eastern Africa [8]. This archaeological transition has been discussed as reflecting the majority adoption of Islam [3, 9]; it can also be seen in the distinction of coastal ceramics and material traditions from hinterland assemblages [10].
Because present-day people, identified as Swahili, may differ in their genetic ancestry compared to populations of the 11th to 18th centuries, the origins of earlier peoples of the Swahili culture cannot be reconstructed without the aid of ancient DNA.
A set of common oral traditions recorded along the coast relate the founding of coastal towns to the arrival of a group known as the Shirazi, referring to a region in Persia [15–17]. This Shirazi tradition was first put into writing in the Kilwa Chronicle, transcribed in Portuguese in the 16th century [18]
This study presents the first genetic analysis of the medieval people of the Swahili coast of eastern Africa. We generated ancient DNA data from the skeletal remains of individuals found at five coastal or island towns: Manda, Mtwapa, Kilwa, Songo Mnara, and Lindi. These samples date to the mid-second millennium 1300-1800 CE but provide insight into genetic admixture events of the medieval period spanning from the 10th-11th centuries CE onwards. We also generate DNA from the skeletal remains of individuals found at the site of Makwasinyi (∼1650-1950 CE), about 100 km inland from the southern Kenyan coast, which has been populated by people who were in cultural contact with coastal groups. We compare the newly reported data from the ancient individuals to that of present-day coastal Swahili speakers [28, 29] and to published data from diverse ancient and present-day eastern African [30, 31] and Eurasian groups [31–38].
[...]
We performed principal component analysis (PCA, Figure 1B and Extended Data Figure 3), computing the eigenvector axes using 1286 present-day Eurasian and African individuals. Many of the newly reported ancient individuals projected onto this modern variation form a cline, with one end overlapping with ancient and present-day African groups and the other falling between present-day Persian and Indian groups. This cline suggests that the individuals might be mixed in different proportions of two source populations, where each can have multiple ancestry components. However, some coastal individuals, particularly from Songo Mnara and Lindi, do not fall along this cline. Individuals who deviate from the cline cannot be described by this same two-way mixture model and must harbor alternative ancestral contributions. Similar patterns of ancestry composition are evident with unsupervised clustering using ADMIXTURE (Figure 1C, Extended Data Figure 1), which suggests sub-Saharan African-associated components, Near Eastern-associated components, and East Asian or Indian-associated components.
[...]
Male ancestors mainly from Persia; female ancestors mainly from Africa and India
We tested whether male and female ancestors (based on genetically determined sex) contributed equal proportions of the three ancestry components in the ancient individuals sampled at Mtwapa, Manda, and Kilwa (Table 1). We could not carry out the same analysis for Songo Mnara because there were no individuals who fit the three-way model and had data of sufficient quality to permit this analysis.
We first analyzed mitochondrial haplogroups transmitted entirely by females, restricting to non-first-degree relatives. All but one individual carries an L* haplogroup, which today is almost wholly restricted to sub-Saharan Africans [41–43]. The one individual with a non-L haplogroup is from Mtwapa and carries haplogroup M30d1, which is largely restricted to South Asia [42]. The mtDNA haplogroup distribution of sampled coastal individuals is consistent with female ancestry deriving almost entirely from African or Indian sources, with a higher proportion of African female ancestors. We also expanded our mtDNA analysis to include ancient coastal and island individuals with low coverage or who did not pass quality control on the autosomes but did have sufficient mtDNA coverage. With this larger group of individuals, all but two have L* haplogroup indicative of sub-Saharan Africa. In addition to the M30d1 haplogroup, common in South Asia, we identified a male individual with haplogroup R0+16189, which is common in Saudi Arabia and the Horn of Africa [44].
We also examined male-transmitted Y chromosome sequences, which tell a different story than female-transmitted mitochondrial DNA. Two of three non-first-degree related males from Manda belong to haplogroup J2, while the third belongs to G2; both haplogroups are characteristic of Near Eastern ancestry (plausibly Persian) and are largely absent in sub-Saharan Africa [45–47]. The Kilwa individual also carries a J2 haplogroup. Fourteen of the twenty male Mtwapa individuals belong to J-family haplogroups, and two belong to R1a haplogroups, all of which are typically non-African lineages. The four remaining individuals belong to E1 family haplogroups, characteristic of sub-Saharan Africa. The Y chromosome haplogroup distribution is consistent with the male ancestry of sampled individuals from Mtwapa deriving mostly, but not entirely, from the Near East. The Y chromosome haplogroup distribution in Mtwapa differs qualitatively from that in Manda and Kilwa in that it includes more J1’s than J2’s, but is still consistent with the male ancestry of sampled individuals from Mtwapa deriving mostly, but not entirely, from the Near East.
Finally, we compared ancestry estimates from chromosomes 1-22 (the autosomes), which equally reflect the contributions of females and males, to those from chromosome X, which occurs in two copies in females compared to one in males, and thus predominantly reflects the contributions of female ancestors. African ancestry estimates from the autosomes are significantly lower in the individuals from Mtwapa, Manda, and Kilwa than estimates made from chromosome X: the 95% confidence intervals are non-overlapping, and the absolute value of the Z-score for a difference between the two estimates is greater than 2. This implies that the African ancestry in medieval Swahili coast populations primarily came from females, consistent with the mitochondrial DNA and Y chromosome evidence (Table 1). Similarly, the Persian ancestry in all three groups primarily came from males (Table 1). Under the simplifying assumption that the mixture occurred in a single event over just a few generations, rather than over a period that spanned many generations, we estimate the proportion of African ancestry contributed by females: 100% in sampled individuals at Manda (95% confidence interval entirely above this value), 76-90% at Mtwapa, and 80-98% at Kilwa (Methods, Table 1). Using the same approach, we estimate the proportion of Persian ancestry contributed by males to be 100% in sampled individuals in all three locations.
All three ancestry components began mixing by about 1000 CE
We used the software DATES [48] to estimate when mixture occurred. We detect strong signals of linkage disequilibrium driven by a mixture between ancestral populations of sub-Saharan African and Asian ancestry, with 95% confidence intervals for the inferred dates of 938-1242 CE for Mtwapa, 629-1412 CE for Manda, and 614-1336 CE for Kilwa (Table 1 and Extended Data Figure 2). The uncertainty intervals of the three groups overlap from 938 to 1242 CE. The Songo Mnara individual that fits the three-way model does not have sufficient data to estimate a date of mixture, but since the other three date ranges overlap, it is reasonable to hypothesize that they derive from mixture in the same period. Our analysis suggests that already admixed males with both Indian and Persian ancestry were present along the coast by ∼1000 CE and mixed further with primarily female sub-Saharan Africans.
In recent centuries, the coast was impacted by Arabian (plausibly Omani) migration
For the earliest individuals in our study with a Near Eastern ancestry component (individuals from Manda, Kilwa, and the I19550 individual from Songo Mnara), there is no evidence of Arabian-associated ancestry that is not well proxied by Persian sources. However, later individuals can be modeled as having Arabian-related ancestry, such as those buried at Mtwapa with radiocarbon dates in the 16th to 18th centuries. We were not able to determine the exact source of the Arabian-associated ancestry in the Mtwapa group. However, we know that it is somewhere on the genetic gradient between Arabians and Persians. We are also not able to determine if there is added Persian-Arabian ancestry or only added Arabian ancestry since the significant Persian admixture from ∼1000 CE would mask any smaller or partial subsequent additions of Persian ancestry. In one of our analyses, this added Persian-Arabian ancestry source for Mtwapa is well proxied by groups bordering the Strait of Hormuz, which separates the Arabian Peninsula from present-day Iran. The Strait of Hormuz and the Swahili coast were under Omani control simultaneously (beginning at the end of the 17th century). While Mtwapa individuals have Arabian and Persian ancestries, we do not detect Arabian-associated ancestry when we apply a similar analysis to the Manda individuals who lived earlier than the Mtwapa individuals. Since Manda individuals lived sometime between the 15th and 17th centuries, they predate Omani political dominance along the Swahili coast. Further evidence for Arabian immigration affecting parts of the Swahili coast in the later period comes from two individuals from Songo Mnara (I19547 and I19549). Both date to after 1500 CE and can be modeled with Arabian-related ancestry in qpAdm (Figure 2A).
Relationship between modern and medieval peoples of the eastern African coast
We analyzed genome-wide data from 89 people previously reported by Brucato et al. After removing two outliers (according to the ADMIXTURE graph in Extended Data Figure 1) from analysis, we found that these present-day people who live in coastal areas of Kilifi, Lamu, and Mombasa in Kenya and who identify as Swahili [28] typically have a small genetic inheritance from ancient people who lived in nearby coastal areas (proxied by the individual buried at Kilwa) (Figure 3A and Extended Data Table 3, Supplementary qpAdm modeling). We infer point estimates in the three coastal populations of 79-86% Bantu-associated ancestry proxied by the Lindi individual (all standard errors about 3%); 2-15% Pastoral Iron Age-associated ancestry [30] (all standard errors about 3%); and 6-14% ancestry from medieval Swahili-associated ancestry (all standard errors about 4%) (Figure 3A).
We also examined mitochondrial and Y chromosome haplotype determinations published in the same study [28]. A total of 231 of 237 individuals with mitochondrial DNA haplogroup determinations (97%) were from the L and M1 lineages, which are widespread in Africa, consistent with almost all female-lineage ancestry being from Africa [49–52]. Of 107 Y chromosome haplogroups, 102 (95%) are typically African, which is very different from medieval coastal individuals from whom almost all Y chromosome haplogroups are Near Eastern associated (Table 1).
[...]
Unlike the Mtwapa and Manda individuals, Makwasinyi individuals have no evidence of any recent Asian ancestry according to qpAdm. Instead, they are well modeled as 21.3±1.2% Pastoral Neolithic-associated ancestry and 78.7±1.2% Bantu-associated ancestry (Figure 3A). Herders were present in eastern Africa from at least 3000 BCE but are not documented beyond northernmost inland Kenya until two millennia later [30]. They persisted after people who practiced farming spread from west to east Africa in the first millennium CE, often associated with the Bantu expansion [26].
The genetic findings are also important to interpret in light of the linguistic evidence. Kiswahili is a Bantu language, and given our finding that most of the female ancestry of individuals in this study derived from Africans, our results point to a scenario in which the children of immigrant men of Asian origin adopted the languages of their mothers, as is common in matrilocal cultures [55]. But Kiswahili also has non-African influences consistent with the genetic and archaeological findings and reflecting 1500 years of interaction with societies around the Indian Ocean rim. Persian loanwords contribute up to 3% of Kiswahili, but it is unclear if they derive directly from Persian or through adoption into other Indian Ocean languages [56]. Arabic loanwords are the single largest non-Bantu element in Kiswahili (16-20%) [56]. However, these words, which predominantly relate to religion, jurisprudence, trade, and maritime vocabulary, may be primarily due to incorporations after around 1500 CE [25].
https://www.biorxiv.org/content/10.1....499442v1.full
Abstract
The peoples of the Swahili coast of eastern Africa established a literate urban culture by the second millennium CE. They traded across eastern Africa and the Indian Ocean and were among the first sub-Saharan practitioners of Islam. An open question has been the extent to which these early interactions between Africans and non-Africans were accompanied by genetic admixture. We report genome-wide ancient DNA from 80 individuals in five medieval and early modern (1300-1800 CE) coastal towns, as well as people from an inland town postdating 1650 CE. Over half of the ancestry of most coastal individuals came from African ancestors; these African ancestors were primarily female. A slightly smaller proportion of ancestry was from Asia. This Asian component was approximately eighty to ninety percent from Near Eastern males and ten to twenty percent from Indian females. Peoples of African and Asian origins began to mix by around 1000 CE, a time when archaeological evidence documents changes on the coast that are often interpreted as marking the large-scale adoption of Islam. Before roughly 1500 CE, the Near Eastern ancestry detected in the individuals was mainly Persian-related, consistent with the narrative of the Kilwa Chronicle, the oldest history told by the Swahili themselves. After this time, the sources of Near Eastern ancestry became increasingly Arabian, consistent with the archaeological and historical evidence of growing interactions between the Swahili coast and parts of southern Arabia. Subsequent interactions of Swahili coast peoples with other Asian and African groups further changed the ancestry of present-day peoples relative to the ancient individuals we sequenced, highlighting how Swahili genetic legacies can be more clearly understood with ancient DNA.
First millennium CE sites on the littoral, beginning in the 7th century CE, were part of a shared material culture and practice network across the eastern African region [3, 4] and were fully engaged in Indian Ocean trading networks [5]. The annual southeast monsoons from May to October allowed trading vessels to travel from India or the Arabian Peninsula to the eastern African coast, and the northeast monsoons from November to March enabled their return in the same year [5, 6]. Muslims were present on the coast from the 8th century CE, although probably in a minority [7]. A transition can be identified in the archaeology of the coast during the 11th century, with the establishment of new settlements and the elaboration of older ones with coral-built mosques and tombs, all the while remaining deeply connected to sites throughout eastern Africa [8]. This archaeological transition has been discussed as reflecting the majority adoption of Islam [3, 9]; it can also be seen in the distinction of coastal ceramics and material traditions from hinterland assemblages [10].
Because present-day people, identified as Swahili, may differ in their genetic ancestry compared to populations of the 11th to 18th centuries, the origins of earlier peoples of the Swahili culture cannot be reconstructed without the aid of ancient DNA.
A set of common oral traditions recorded along the coast relate the founding of coastal towns to the arrival of a group known as the Shirazi, referring to a region in Persia [15–17]. This Shirazi tradition was first put into writing in the Kilwa Chronicle, transcribed in Portuguese in the 16th century [18]
This study presents the first genetic analysis of the medieval people of the Swahili coast of eastern Africa. We generated ancient DNA data from the skeletal remains of individuals found at five coastal or island towns: Manda, Mtwapa, Kilwa, Songo Mnara, and Lindi. These samples date to the mid-second millennium 1300-1800 CE but provide insight into genetic admixture events of the medieval period spanning from the 10th-11th centuries CE onwards. We also generate DNA from the skeletal remains of individuals found at the site of Makwasinyi (∼1650-1950 CE), about 100 km inland from the southern Kenyan coast, which has been populated by people who were in cultural contact with coastal groups. We compare the newly reported data from the ancient individuals to that of present-day coastal Swahili speakers [28, 29] and to published data from diverse ancient and present-day eastern African [30, 31] and Eurasian groups [31–38].
[...]
We performed principal component analysis (PCA, Figure 1B and Extended Data Figure 3), computing the eigenvector axes using 1286 present-day Eurasian and African individuals. Many of the newly reported ancient individuals projected onto this modern variation form a cline, with one end overlapping with ancient and present-day African groups and the other falling between present-day Persian and Indian groups. This cline suggests that the individuals might be mixed in different proportions of two source populations, where each can have multiple ancestry components. However, some coastal individuals, particularly from Songo Mnara and Lindi, do not fall along this cline. Individuals who deviate from the cline cannot be described by this same two-way mixture model and must harbor alternative ancestral contributions. Similar patterns of ancestry composition are evident with unsupervised clustering using ADMIXTURE (Figure 1C, Extended Data Figure 1), which suggests sub-Saharan African-associated components, Near Eastern-associated components, and East Asian or Indian-associated components.
[...]
Male ancestors mainly from Persia; female ancestors mainly from Africa and India
We tested whether male and female ancestors (based on genetically determined sex) contributed equal proportions of the three ancestry components in the ancient individuals sampled at Mtwapa, Manda, and Kilwa (Table 1). We could not carry out the same analysis for Songo Mnara because there were no individuals who fit the three-way model and had data of sufficient quality to permit this analysis.
We first analyzed mitochondrial haplogroups transmitted entirely by females, restricting to non-first-degree relatives. All but one individual carries an L* haplogroup, which today is almost wholly restricted to sub-Saharan Africans [41–43]. The one individual with a non-L haplogroup is from Mtwapa and carries haplogroup M30d1, which is largely restricted to South Asia [42]. The mtDNA haplogroup distribution of sampled coastal individuals is consistent with female ancestry deriving almost entirely from African or Indian sources, with a higher proportion of African female ancestors. We also expanded our mtDNA analysis to include ancient coastal and island individuals with low coverage or who did not pass quality control on the autosomes but did have sufficient mtDNA coverage. With this larger group of individuals, all but two have L* haplogroup indicative of sub-Saharan Africa. In addition to the M30d1 haplogroup, common in South Asia, we identified a male individual with haplogroup R0+16189, which is common in Saudi Arabia and the Horn of Africa [44].
We also examined male-transmitted Y chromosome sequences, which tell a different story than female-transmitted mitochondrial DNA. Two of three non-first-degree related males from Manda belong to haplogroup J2, while the third belongs to G2; both haplogroups are characteristic of Near Eastern ancestry (plausibly Persian) and are largely absent in sub-Saharan Africa [45–47]. The Kilwa individual also carries a J2 haplogroup. Fourteen of the twenty male Mtwapa individuals belong to J-family haplogroups, and two belong to R1a haplogroups, all of which are typically non-African lineages. The four remaining individuals belong to E1 family haplogroups, characteristic of sub-Saharan Africa. The Y chromosome haplogroup distribution is consistent with the male ancestry of sampled individuals from Mtwapa deriving mostly, but not entirely, from the Near East. The Y chromosome haplogroup distribution in Mtwapa differs qualitatively from that in Manda and Kilwa in that it includes more J1’s than J2’s, but is still consistent with the male ancestry of sampled individuals from Mtwapa deriving mostly, but not entirely, from the Near East.
Finally, we compared ancestry estimates from chromosomes 1-22 (the autosomes), which equally reflect the contributions of females and males, to those from chromosome X, which occurs in two copies in females compared to one in males, and thus predominantly reflects the contributions of female ancestors. African ancestry estimates from the autosomes are significantly lower in the individuals from Mtwapa, Manda, and Kilwa than estimates made from chromosome X: the 95% confidence intervals are non-overlapping, and the absolute value of the Z-score for a difference between the two estimates is greater than 2. This implies that the African ancestry in medieval Swahili coast populations primarily came from females, consistent with the mitochondrial DNA and Y chromosome evidence (Table 1). Similarly, the Persian ancestry in all three groups primarily came from males (Table 1). Under the simplifying assumption that the mixture occurred in a single event over just a few generations, rather than over a period that spanned many generations, we estimate the proportion of African ancestry contributed by females: 100% in sampled individuals at Manda (95% confidence interval entirely above this value), 76-90% at Mtwapa, and 80-98% at Kilwa (Methods, Table 1). Using the same approach, we estimate the proportion of Persian ancestry contributed by males to be 100% in sampled individuals in all three locations.
All three ancestry components began mixing by about 1000 CE
We used the software DATES [48] to estimate when mixture occurred. We detect strong signals of linkage disequilibrium driven by a mixture between ancestral populations of sub-Saharan African and Asian ancestry, with 95% confidence intervals for the inferred dates of 938-1242 CE for Mtwapa, 629-1412 CE for Manda, and 614-1336 CE for Kilwa (Table 1 and Extended Data Figure 2). The uncertainty intervals of the three groups overlap from 938 to 1242 CE. The Songo Mnara individual that fits the three-way model does not have sufficient data to estimate a date of mixture, but since the other three date ranges overlap, it is reasonable to hypothesize that they derive from mixture in the same period. Our analysis suggests that already admixed males with both Indian and Persian ancestry were present along the coast by ∼1000 CE and mixed further with primarily female sub-Saharan Africans.
In recent centuries, the coast was impacted by Arabian (plausibly Omani) migration
For the earliest individuals in our study with a Near Eastern ancestry component (individuals from Manda, Kilwa, and the I19550 individual from Songo Mnara), there is no evidence of Arabian-associated ancestry that is not well proxied by Persian sources. However, later individuals can be modeled as having Arabian-related ancestry, such as those buried at Mtwapa with radiocarbon dates in the 16th to 18th centuries. We were not able to determine the exact source of the Arabian-associated ancestry in the Mtwapa group. However, we know that it is somewhere on the genetic gradient between Arabians and Persians. We are also not able to determine if there is added Persian-Arabian ancestry or only added Arabian ancestry since the significant Persian admixture from ∼1000 CE would mask any smaller or partial subsequent additions of Persian ancestry. In one of our analyses, this added Persian-Arabian ancestry source for Mtwapa is well proxied by groups bordering the Strait of Hormuz, which separates the Arabian Peninsula from present-day Iran. The Strait of Hormuz and the Swahili coast were under Omani control simultaneously (beginning at the end of the 17th century). While Mtwapa individuals have Arabian and Persian ancestries, we do not detect Arabian-associated ancestry when we apply a similar analysis to the Manda individuals who lived earlier than the Mtwapa individuals. Since Manda individuals lived sometime between the 15th and 17th centuries, they predate Omani political dominance along the Swahili coast. Further evidence for Arabian immigration affecting parts of the Swahili coast in the later period comes from two individuals from Songo Mnara (I19547 and I19549). Both date to after 1500 CE and can be modeled with Arabian-related ancestry in qpAdm (Figure 2A).
Relationship between modern and medieval peoples of the eastern African coast
We analyzed genome-wide data from 89 people previously reported by Brucato et al. After removing two outliers (according to the ADMIXTURE graph in Extended Data Figure 1) from analysis, we found that these present-day people who live in coastal areas of Kilifi, Lamu, and Mombasa in Kenya and who identify as Swahili [28] typically have a small genetic inheritance from ancient people who lived in nearby coastal areas (proxied by the individual buried at Kilwa) (Figure 3A and Extended Data Table 3, Supplementary qpAdm modeling). We infer point estimates in the three coastal populations of 79-86% Bantu-associated ancestry proxied by the Lindi individual (all standard errors about 3%); 2-15% Pastoral Iron Age-associated ancestry [30] (all standard errors about 3%); and 6-14% ancestry from medieval Swahili-associated ancestry (all standard errors about 4%) (Figure 3A).
We also examined mitochondrial and Y chromosome haplotype determinations published in the same study [28]. A total of 231 of 237 individuals with mitochondrial DNA haplogroup determinations (97%) were from the L and M1 lineages, which are widespread in Africa, consistent with almost all female-lineage ancestry being from Africa [49–52]. Of 107 Y chromosome haplogroups, 102 (95%) are typically African, which is very different from medieval coastal individuals from whom almost all Y chromosome haplogroups are Near Eastern associated (Table 1).
[...]
Unlike the Mtwapa and Manda individuals, Makwasinyi individuals have no evidence of any recent Asian ancestry according to qpAdm. Instead, they are well modeled as 21.3±1.2% Pastoral Neolithic-associated ancestry and 78.7±1.2% Bantu-associated ancestry (Figure 3A). Herders were present in eastern Africa from at least 3000 BCE but are not documented beyond northernmost inland Kenya until two millennia later [30]. They persisted after people who practiced farming spread from west to east Africa in the first millennium CE, often associated with the Bantu expansion [26].
The genetic findings are also important to interpret in light of the linguistic evidence. Kiswahili is a Bantu language, and given our finding that most of the female ancestry of individuals in this study derived from Africans, our results point to a scenario in which the children of immigrant men of Asian origin adopted the languages of their mothers, as is common in matrilocal cultures [55]. But Kiswahili also has non-African influences consistent with the genetic and archaeological findings and reflecting 1500 years of interaction with societies around the Indian Ocean rim. Persian loanwords contribute up to 3% of Kiswahili, but it is unclear if they derive directly from Persian or through adoption into other Indian Ocean languages [56]. Arabic loanwords are the single largest non-Bantu element in Kiswahili (16-20%) [56]. However, these words, which predominantly relate to religion, jurisprudence, trade, and maritime vocabulary, may be primarily due to incorporations after around 1500 CE [25].
https://www.biorxiv.org/content/10.1....499442v1.full