Introduction

Dispersal is usually sex-biased in mammals and birds, i.e., one sex disperses more frequently and/or further while the opposite sex remains and reproduces in its natal area [1–2]. This sex-biased dispersal pattern is usually consistent within and between closely related species [3–4]. However, local changes in ecological and demographic factors (e.g. intrinsic patch features such as food availability, size and isolation, population density, intra- and inter-specific interactions, sex ratio, relatedness) can alter an individual’s dispersal strategy [2, 5–7]. Individuals that otherwise would remain philopatric may disperse to seek reproductive opportunities or because the costs of dispersal surpass the benefits, individuals may therefore reproduce in their natal areas [4–5]. As a consequence, the dispersal pattern of the population, usually biased towards one sex, may become less obvious or reversed [2, 4, 8].

Since dispersal promotes genetic exchange between populations, changes in dispersal patterns are expected to influence a population’s genetic composition and conservation requirements [7, 9]. If gene flow increases, differentiated sub-populations may come into contact and exchange genes in contact zones, leading to potential genetic “swamping” and loss of local adaptations [8]. On the other hand, if gene flow is constrained, populations may become more structured, leading to a loss of genetic variation through drift and possibly a reduction of reproductive fitness if mating options become limited to kin resulting in inbreeding [8, 10].

Human activities may drive changes in population demography and connectivity, which can modify population dispersal patterns [2, 5]. Changes in sex-biased dispersal have been described for several taxa in populations inhabiting human-altered environments, including primates (e.g. Pan troglodytes verus, [11], Macaca cyclopis, [12], Ateles belzebuth, [13]). Moreover, female dispersal in Old World primates is frequently limited by the degree of overlap of a group’s home range with others [14]. In human-dominated environments, low population density and isolation of groups across the landscape may reduce the extent of overlap between home ranges, increasing the cost of dispersal for females. As a consequence, females may not disperse from their natal group, which in turn may increase genetic differentiation between sub-populations and lead to a less pronounced female-biased dispersal pattern [2, 8].

However, predicting the influence of human activities on dispersal is not straightforward. Its effects may depend on an individuals’ ability to move between populations and to reproduce, on the intensity of anthropogenic disturbance and on the species’ ability to cope with such disruption [15–16], which is often difficult to measure. Additionally, human activities may have contrasting effects on gene flow (e.g. decreasing, [17]; increasing, [18]) and on group composition [19–21]. As a result, these effects can easily be overlooked or mistaken for the characteristic pattern of a species, particularly when local human practices are illegal and concealed and when only one population is studied. Identifying how and whether anthropogenic activities influence dispersal may thus require comparison among populations subject to different levels of human disturbance [7, 9].

Here, we investigated population differences in dispersal behaviour in Guinea baboons (Papio papio). Our study was focused on two geographically distinct populations—southern Guinea-Bissau and Senegal—that inhabit environments with contrasting levels of human disturbance. Baboons from southern Guinea-Bissau inhabit a human-altered environment and have been extensively hunted, mainly for meat consumption [22–24]. The species’ distribution in Guinea-Bissau is highly fragmented, with few groups outside national parks [22, 25]. In contrast, the Niokolo Koba National Park population in Senegal, the geographic scale of which is comparable to the Guinea-Bissau study site, is not significantly affected by human activity and is continuously distributed within the park limits [26, 27].

Dispersal in Guinea baboons is thought to be female-biased. Previous studies have inferred long-term female-mediated gene flow across the species distribution [28], including the Senegalese [29] and the Guinea-Bissau population [23, 25], and on-going female-biased dispersal in Senegal [29]. Sex-biased dispersal in the last few generations has not been studied for the Guinea-Bissau population. However, results from our previous work suggested that the intensification of human activities might have modified the Guinea-Bissau population structure in the last thirty years (3–4 generations, at 12–15 years [30]). As the population in southern Guinea-Bissau is very fragmented, we initially hypothesized that populations inhabiting protected areas would be isolated, but evidence for on-going gene flow included: i) differentiated genetic units were found to be in contact, in particular in one protected area, where the location of genetic discontinuities and anthropogenic landscape features do not seem to be correlated, and ii) significant genetic similarity was found between samples distanced 115.5 km apart, with higher genetic dissimilarity found at 66 km corresponding to an area impacted by anthropogenic activities and low group density [25]. Thus, genetic structure within the Guinea-Bissau population suggested a recent increase in dispersal distances [25], an unexpected result for a species with female-biased dispersal and which may signal a recent change in sex-biased dispersal in the population.

Our aim in this study was to follow up our previous work by comparing patterns of sex-biased gene flow (direction and magnitude) at similar geographic scales between Guinea-Bissau and a less disturbed environment (in Senegal). By applying molecular tools to study sex-biased dispersal we can detect differences starting just a few generations ago and test its impact on population structure [31, 32]. While the philopatric sex is expected to show higher natal social group relatedness, display greater genetic differentiation among groups and thus higher population structure, the dispersing sex is expected to homogenize genetic differentiation across groups and thus display more similar allele frequencies across subpopulations [31, 32]. Given the higher degree of human disturbance across Guinea-Bissau when compared to Senegal, we expected to find differences in sex-specific population genetic structure between the two populations, which would indicate differences in the sex-biased dispersal. More specifically, we expected to find no difference in the degree of genetic structure between males and females in Guinea-Bissau (e.g. no sex-biased dispersal) and a higher degree of genetic structure in males when compared with females in Senegal (e.g. female-biased dispersal).

Methods

Study areas

The study encompasses two Guinea baboon populations—Guinea-Bissau (GB) and Senegal (SEN), separated by approximately 150 km (Fig 1). Previous analyses using samples collected across the entire distribution of the species found that GB and SEN constitute differentiated genetic clusters [33–34].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Study areas.

Top left: Black shading in overview map indicates the distribution of Guinea baboons (adapted from IUCN). Top right: Guinea-Bissau (GB—country area in light grey) and Senegal (SEN—country area in dark grey). Sampled sites indicated by black-dotted white circles. Bottom: Most important anthropogenic features (main roads, villages and towns) and sampling sites in Senegal (Niokolo Koba National Park—PNNK, sampling sites: 1—Gue Damantan, 2—Simenti, 3—Camp Lion, 4—Lingue Kountou, and 5—Niokolo) and in Guinea-Bissau (GB_B—Boé Natural Park, GB_Cufada—Cufada Lagoons Natural Park, and GB_Cantanhez—Cantanhez Woodlands Natural Park). The two main rivers (Corubal in Guinea-Bissau and the Gambia River in Senegal) are indicated by a grey line. The comparison of dispersal patterns between Senegal and Guinea-Bissau was done for the geographic scale of 66 km (the spatial scale between sampling sites in Senegal and between GB_Cantanhez and GB_Cufada in GB).

https://doi.org/10.1371/journal.pone.0194189.g001

In GB, the study area includes three geographically distinct localities—1) Cantanhez Woodlands National Park (GB_Cantanhez, north limit 11.389835° -14.810265° to south limit 11.010386° -15.234415°), 2) Cufada Lagoons Natural Park (GB_Cufada, north limit: 11.753801° -15.185281° to south limit: 11.627405° -14.903575°), and 3) Boé Natural Park (GB_Boé, north limit 12.206995° -14.005278° to south limit 11.825659° to -13.863576°). GB_Cantanhez and GB_Cufada are located in the southwest coastal part of GB and are separated roughly by 50 km. GB_Cantanhez (total area 106 767 ha) is located on a peninsula connected to mainland GB by a 25–30 km isthmus. GB_Cufada (a total area of 89,000 ha) is surrounded by the Corubal River in the north and Buba channel in the south and delimitated by two main towns—Buba and Fulacunda—in the southwest and in the northeast limits, respectively. GB_Boé is located in the southeast of the country and is separated from GB_Cantanhez and GB_Cufada roughly by 120 Km and 100 km, respectively. GB_Boé is delimitated in the west by the Corubal River and in the east by the Fefine River. In SEN, the study was carried out within the Niokolo Koba National Park (PNNK, north limit: 13.329490° -13.442568° and south limit: 12.644166° -13.094790°, a total area of 913,000 ha), across a maximum distance of 65 km between sampling sites.

The area between the GB’s Cantanhez and Cufada parks, which is comparable to the spatial scale among sampling sites in PNNK, has a higher degree of human disturbance (Fig 1). IBAP (Instituto da Biodiversidade e Áreas Protegidas, Institute for Biodiversity and Protected Areas), the national authority responsible for the management of protected areas in GB, has a set of rules in place that limit hunting practices and deforestation to a certain extent within the limits of protected areas. A considerable extension of the original habitat has been converted to crop areas outside protected areas in GB, with a variety of human infrastructure, including paved roads connecting the south of the country to the capital city and medium-size towns (Fig 1). Groups of baboons are rarely observed outside protected areas in GB [22]. Likewise, the populations of Guinea baboons outside protected areas in Senegal are thought to have suffered a widespread decline in the last thirty years [35–36] and fragmentation of the distribution is likely. To the best of our knowledge, the populations between GB_Boé and PNNK have not been recently surveyed.

Permission to work in the Guinea-Bissau protected areas (e.g. Cantanhez, Cufada, and Boé) was issued by IBAP. Permission to work in PNNK was issued by Direction des Parcs Nationaux and Ministère de l'Environnement et de la Protection de la Nature de la Republique du Senegal (Attestation 0383/24/03/2009 and 0373/10/3/2012).

Sampling

We reanalysed a microsatellite loci (STR) dataset generated by previous studies composed of 143 unique multi-locus genotypes (54 males and 89 females) sampled in GB’s protected areas (GB_Cantanhez N = 71, GB_Cufada N = 51, and GB_Boé N = 21 genotypes) [25, 37] and 165 unique genotypes (97 males and 68 females) sampled within PNNK area in SEN [29] (Fig 1; Table 1; Table A in S1 Appendix). Genotypes were obtained non-invasively from faecal samples, collected randomly from unhabituated and unidentified individuals in baboon home ranges, in foraging paths and sleeping sites (for details on sampling techniques, see [25, 29, 34]). During sample collection, the location of each sample was recorded using a GPS receiver.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Inference of dispersal patterns at multiple geographic scales for Guinea-Bissau and Senegal.

https://doi.org/10.1371/journal.pone.0194189.t001

In GB, sampling was carried out at 17 sampling sites within protected areas (average number of genotypes per sampling site = 8.4, varying between 3 and 21, the average distance between sampling sites = 64.9 km, ranging from 2.6 to 164.5 km, Figure A and Tables A and B in S1 Appendix). Sampling sites in Guinea-Bissau are unlikely to represent independent social groups. In Senegal, social parties may divide into smaller groups or merge into larger ones several times a day [38] and we assumed a similar social organization in Guinea-Bissau. Samples collected in GB at one site may represent groups of individuals of the same social unit (i.e., the individuals were foraging or sleeping together before defecating) but neighbouring sampling sites might not be independent (i.e., the groups might have slipped before or may have merged after the sampling moment).

In SEN, sampling was carried out at five sampling sites in an almost continuous design (average number of genotypes per sampling site = 33, varying between 108 and 11, the average distance between sampling sites = 18.1 km, varying between 4.1 and 65.0 km) (Table C in S1 Appendix). Most samples collected in SEN stem from the Simenti community, a troop under study by the Cognitive Ethology Laboratory of the German Primate Center (DPZ).

Results

Sex-biased gene flow at a broad scale in a human-dominated environment (in GB)

When STRUCTURE was run using the GB165 sub-set, K = 2 produced the highest modal value in the ΔK distribution, with the largest Log-likelihood and with highest posterior probability (posterior probability_K2 = 1) (Figures A and B in S6 Appendix). The plot of ranked averaged partial membership q for each individual to cluster 2 show a break between 0.71 and 0.75 and was nearly continuous between 0.75 and 1 (Figure C in S6 Appendix). Individuals were assigned to their respective genetic clusters if the averaged q > 0.75 and were considered admixed between clusters if 0 < q < 0.75. Males and females did not differ in genetic structure (Fig 2a and 2b; Figure B and Table A in S6 Appendix). For K = 2, 100% of males and 96% of females sampled at Cantanhez were assigned to cluster 1 (69 individuals, average q_{cluster 1} = 0.93, varying between 0.98 and 0.75) and two females were classified as admixed (Fig 2; Table A in S6 Appendix). In Boé, 62% of the males and 75% of the females were assigned to cluster 2 (average q_{cluster 2} = 0.86, varying between 0.95 and 0.78) or considered as admixed (Fig 2; Table A in S6 Appendix). Cufada was the only locality in which individuals were assigned to both genetic clusters; 41.2% of males and 77% of females were assigned to cluster 1 (average q_{cluster 1} = 0.91, varying between 0.97 and 0.78) and 18% of males and 12% of females were assigned to cluster 2 (average q_{cluster 2} = 0.81, varying between 0.93 and 0.75, Table A in S6 Appendix; Fig 2). Nevertheless, 41.2% of males and 12% of females were considered admixed between clusters at GB_Cufada (Table A in S6 Appendix; Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Population structure estimated in GB at 165 km.

a and b shows the proportion of males (a) and females (b) at each sampling site assigned to the two STRUCTURE clusters using the GB165 sub-set. Individual-based assignment to each cluster was confirmed if q > 0.75 (cluster 1 in red and cluster 2 in green). If 0 < q < 0.75, the individual was classified as admixed between clusters (yellow). Diagram size is proportional to sample size of each site. Numbers indicate the sampling sites at GB_Cufada where individuals assigned to cluster 2 were found; 1—Sr. Soares 1, 2—Bakar Contê, 3—Guebombol, 4—Bubatchingue. c shows the spatial autocorrelation correlogram of the multiclass tests between females (dark grey line) and males (black line); n _{pairwise comparisons}: [0–17 km[ = 1071, [17–34[ = 1183, [34–51[ = 824, [51–68[ = 899, [68–85[ = 189, [85–102[ = 22, [102–119[ = 78, [119–136[ = 656, [136–153[ = 300 and [153–170[ = 125. Males and females significantly differ in their autocorrelation pattern at the distance classes of [17–34 [(t²_[17–34[ = 9.1, P = 0.002) and [119–136 [(t²_[119–136[ = 4.15, P = 0.042).

https://doi.org/10.1371/journal.pone.0194189.g002

Significant positive spatial autocorrelation was found for males and females for pairwise comparisons between neighbouring groups (i.e., at [0–17 [km or for males only at [17–34 [km, Figure D in S6 Appendix; Fig 2c). Significant negative spatial autocorrelation was found for both sexes at distances between GB_Cantanhez and GB_Cufada (at [34–51 [km and [51–68 [km) and between GB_Boé, GB_Cufada and GB_Cantanhez (males: [119–136 [and [136–153 [and females: [136–153 [and [153–170 [km) (Figure D in S6 Appendix; Fig 2c). An exception to the negative spatial autocorrelation pattern for both sexes at large distance classes comprised pairwise comparisons between GB_Boé and the sampling site Bakar Contê at GB_Cufada ([85–102 [and [102–119 [km), at which females showed significant positive autocorrelation and males show a positive but not significant autocorrelation coefficient. Nevertheless, the n _{pairwise comparisons} for the distance class [85–102 km [is particularly low for both sexes (females n = 6 and males n = 16, Figure D in S6 Appendix; Fig 2c).

The spatial autocorrelation pattern differed significantly between the sexes (Total ω for data = 21.288, P = 0.003, Fig 2c) although significance was only found in two out of the ten distance classes considered. Males are significantly more similar than females between neighbouring groups within GB_Cufada and GB_Cantanhez areas but also between few groups from GB_Cufada and GB_Cantanhez (males_{[17–34 km[}: r = 0.04, P = 0.0002, females_{[17–34 km[}: r = - 0.003, P = 0.3, t²_[17–34[ = 9.1, P = 0.002). Females are significantly more similar than males at larger distances classes (i.e., between GB_Boé and GB_Cufada and GB_Cantanhez females _{[119–136 km[}: r = - 0.011 and males_{[119–136 km[}: r = - 0.041, t²_[119–136[ = 4.15, P = 0.042).

When the spatial autocorrelation analyses were conducted including the samples collected at GB_Cufada and GB_Cantanhez only and excluding GB_Boé, a pattern where both sexes display significant positive spatial autocorrelation at shorter distances and significant negative spatial autocorrelation at larger distances was revealed (Figure E in S6 Appendix). r intercepted zero at 24 km for males, and at 14 km for females (Figure E in S6 Appendix). The spatial autocorrelation pattern did not differ between the sexes (Total ω for data = 7.834, P = 0.1, Figure F in S6 Appendix) and significant differences were found only at the [16.5–33 [km distance class, as in the previous analyses, including GB_Boé (females are more significantly dissimilar than males, females_[16.5–33[: r = - 0.006, males_[16.5–33[: r = 0.006, P = 0.004). However, some differences were found from the spatial autocorrelation estimated using GB165 and including GB_Boé (S6 Appendix).

GENECLASS identified six individuals as first-generation migrants using the L_h/L_max statistic: one female sampled in Cantanhez with an origin in Cufada, one female sampled in Cufada with an origin in Cantanhez, two males sampled in Cufada with an origin in Boé), and two males sampled in Boé with an origin in Cufada. Of the six migrants, one male (sampled in Cufada in Sr. Soares 1 site with an origin in Boé) was identified assuming that the source population might not have been sampled. The first-generation migrant males from Boé sampled in GB_Cufada were assigned to cluster 2 in STRUCTURE analyses (q _cluster2 = 0.93 to 0.75), which is the genetic unit mostly represented in GB_Boé (see Table C in S6 Appendix for comparison between GENECLASS results and assignment using STRUCTURE).

Males and females differed in the distribution of assignment indexes in Cufada (two-tailed U-test, P = 0.001); mAIc was positive for females and negative for males (mAIc_females = 0.54 and mAIc_males = -1.16), 11 out of 13 male genotypes with no missing data displayed negative AIc values and males showed larger vAIc than females (vAIc_females = 0.23 and vAIc_males = 0.37) (Table B and Figures G, H and I in S6 Appendix). mAIc was not significant for GB_Cantanhez (S6 Appendix).

Comparative analyses at a 66 km scale between GB (human-dominated environment) and SEN (lower habitat fragmentation).

When STRUCTURE was run using GB66, K = 1 was the solution with the largest Log-likelihood and with highest posterior probability (posterior probability_K1 = 0.99) (Figure A in S7 Appendix).

For SEN66, K = 3 was the solution with the highest modal value in the ΔK distribution, with the largest Log-likelihood and with highest posterior probability (posterior probability_K3 = 1) (Figure A in S7 Appendix). When K = 3, SEN66 is broadly clustered in GD+SI+CL+LK (cluster 1) and NK (cluster 2), and seven individuals from SI form cluster 3 (six males and one female, average q_{cluster 3} = 0.86, q varying between 0.79 and 0.93) (Figure B in S7 Appendix). We suspected that the higher genetic similarity of the seven individuals forming cluster 3 biased the analysis. After removing those samples, we re-ran STRUCTURE using parameters set as before using N = 158 genotypes (67 females and 91 males). K = 2 was found to be the most probable clustering solution (highest modal value in the ΔK distribution, the largest Log-likelihood and with posterior probability_{K = 2} = 1) (Figures C and D in S7 Appendix). The plot of ranked averaged partial membership q of each individual to clusters was discontinuous between 0.78 and 0.64 for cluster 2 and was nearly continuous between 0.68 and 1 for cluster 1 (Figure E in S7 Appendix). We assigned individuals to the respective genetic clusters if the averaged q > 0.75 and considered admixed between clusters if 0 < q < 0.75.

Males in SEN66 showed a stronger genetic structure when compared to females (Figure D in S7 Appendix; Fig 3a and 3b). The sampling sites GD+SI+CL+LK comprise 100% of males assigned to cluster 1 (70 males, q_{cluster 1 average} = 0.96 varying between 0.99 and 0.77) and five admixed males, whereas NK harbour 100% of the males assigned to cluster 2 (q_{cluster 2 average} = 0.89 varying between 0.94 and 0.78) and two admixed males (Table A in S7 Appendix; Fig 3a and 3b). In contrast for females, NK harbour females assigned to both genetic units (cluster 1: one female, q _{cluster 1} = 0.95, cluster 2: two females, q _{cluster 2 average} = 0.83 varying between 0.87 and 0.79) and admixed individuals (Table A in S7 Appendix; Fig 3a and 3b).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Population structure for males and females in GB and SEN at 66 km.

a and b shows the proportion of males (a) and females (b) of each sampling site in Senegal (GD—Gue Damantan, SI—Simenti, CL—Camp Lion, LK—Lingue Kountou, and NK—Niokolo) assigned to each of two clusters identified by STRUCTURE (97 males and 68 females, assigned when q > 0.75 and considered admixed when 0 < q < 0.75). c, d, e, and f shows the spatial autocorrelation correlograms of the multiclass tests: c) GB females (FGB66, dashed grey line) vs. males (MGB66, dashed black line) (n _{pairwise comparisons}: [0–16.5[ = 788, [16.5–33[ = 993, [33–49.5[ = 758, [49.5–66[ = 828); d) SEN females (FSEN66, bold grey line) vs. males (MSEN66 bold black line) (n _{pairwise comparisons}: [0–16.5[ = 4468, [16.5–33[ = 809, [33–49.5[ = 131, [51–66[ = 1526); e) Males in GB (MGB66, dashed black line) vs. males in SEN (MSEN66, bold black line) (n _{pairwise comparisons}: [0–16.5[ = 3187, [16.5–33[ = 585, [33–49.5] = 203, [49.5–66[ = 1347), and f) Females in GB (FGB66, dashed grey line) vs. females in SEN (FSEN66, bold grey line) (n _{pairwise comparisons}: [0–16.5[ = 2069, [16.5–33[ = 1217, [34–49.5[ = 686, [49.5–66[ = 1007). The y-axis shows the autocorrelation coefficient r measuring genetic similarity (r>0) or dissimilarity (r<0) between pairs of individuals separated by four distance classes (X-axis: 0 to 49.5 km, starting point). Significant differences for individual distance classes are highlighted: *P < 0.05, **P < 0.001. The bootstrapped 95% error bars are shown. The correlagrams d and f are significantly different in the overall spatial autocorrelation pattern (d—Total ω for data = 20.21, P = 0.0001; f—Total ω for data = 11.1, P = 0.004).

https://doi.org/10.1371/journal.pone.0194189.g003

For GB, no difference was detected between males and females in the global spatial autocorrelation pattern in a spatial scale of 66 km or in every single distance classes considered (Total ω for data = 4.81, P = 0.30; single classes: t²_{[0–16.5 km[} = 0.38, P = 0.56; t²_{[16.5–33 km[} = 2.29, P = 0.17; t²_[33–49.5[ = 2.0, P = 0.16; t²_[49.5–66[ = 0.14, P = 0.74) (Fig 3c). For both sexes, significant positive spatial autocorrelation was found for at shorter distances (i.e., [0–16.5 [km for females and [16.5–33 [km for males) and significant negative spatial autocorrelation was found at distances representing pairwise comparisons between GB_Cantanhez and GB_Cufada ([33–49.5 [and [49.6–66 [km for males and [49.5–66 [km for females) (Figure F in S7 Appendix; Fig 3c).

In contrast for SEN66, males and females differed in the spatial autocorrelation pattern (Total ω for data = 20.21, P = 0.0001; Fig 3d). In SEN, both sexes displayed significant positive spatial autocorrelation at shorter distances classes (Figure F in S7 Appendix; Fig 3d) but males were significantly more similar than females at [16.5–33 [km (t²_{[16.5–33 km[} = 6.63, P = 0.01; males_{[16.5–33 km[}: r = 0.022, P = 0.0001; females _{[16.5–33 km[:} r = - 0.014, P = 0.008) and significantly more dissimilar at [33–66 [km (t²_{[33–49.5 km[} = 16.8, P = 0.0002, males_{[33–49.5 km[}: r = - 0.11, P = 0.0001, females_{[33–49.5 km[}: r = 0.011, P = 0.32; t² _{[49.5–66 km[} = 8.1, P = 0.004, males_{[49.5–66 km[}: r = - 0.039, P = 0.0001; females_{[49.5–66 km[}: r = - 0.008, P = 0.08) (Figure F in S7 Appendix; Fig 3d).

We compared the spatial autocorrelation of each sex between GB66 and SEN66. For males, the test for heterogeneity did not detect a difference in the overall spatial autocorrelation pattern and in three of the four single distance classes considered (Total ω for data = 5.6, P = 0.12, single classes: t²_{[0–16.5 km[} = 0.048, P = 0.84; t²_{[16.5–33 km[} = 0.085, P = 0.80; t²_[49.5–66[ = 0.70, P = 0.48) (Fig 3e). However, between 33 and 49.5 km, GB66 males were significantly more similar than SEN66 males, although not reaching positive autocorrelation (t²_[33–49.5[ = 6.04, P = 0.011, males GB_[33–49.5[: r = - 0.04 and males SEN_[33–49.5[: r = - 0.11). In contrast, the spatial correlogram was significantly different between females from GB66 and SEN66 (Total ω for data = 11.1, P = 0.004) (Fig 3f). When compared to SEN66 females, GB66 females were more similar at shorter distances (t²_{[0–16.5 km[}: = 9.85, P = 0.002, r_{females_GB} = 0.04, r_{females_SEN} = 0.006).

The mAIc statistic did not differ significantly between males and females in SEN.

Comparative analyses at a 26 km scale within the limits of protected areas in GB and SEN (i.e., lower habitat fragmentation).

In SEN26, females displayed significant positive autocorrelation in distance classes representing comparisons between samples collected in neighbouring groups (females_[0–8.66[: r = 0.004 P = 0.005). For GB, only females in Cantanhez26 displayed significant autocorrelation in distance classes considered; significant positive autocorrelation between neighbouring groups and significant negative autocorrelation at larger distances, which mainly includes pairwise comparisons from inside and outside the Peninsula of Cantanhez (Figure G in S7 Appendix).

Discussion

Primate dispersal patterns may change in populations inhabiting human-dominated habitats. We assessed the directionality and extent of sex-biased gene flow using molecular data to investigate dispersal patterns in Guinea baboons (Papio papio) at three different geographic scales and compared two closely related populations—Guinea-Bissau (fragmented distribution, human-dominated habitat) and Senegal (continuous distribution, protected area).

The genetic population structure estimated for a large geographical scale in GB, which included three geographical distinct sub-populations populations separated by a maximum distance linear distance of 165 km (GB_Cantanhez, GB_Cufada and GB_Boé), suggested a pattern of female-mediated gene flow across large distances (Table 1). This result is in agreement with the pattern estimated using maternally transmitted mitochondrial DNA at the same geographic scale [25]: with the most frequent mtDNA haplotypes shared between GB_Cantanhez, GB_Cufada and GB_Boé, the geographic distribution of haplotypes being uncorrelated with sampling locations and significant genetic similarity being detected at larger geographic distances. Female-biased gene flow was also found for Senegal and using data of the whole distribution of the species [28, 29].

Additionally, we found evidence for contact between genetically differentiated males at GB_Cufada (Table 1). GB_Cufada was previously found to be a contact zone between genetically differentiated units [25] but in this study, we identified males sampled at GB_Cufada that could be assigned with a high probability to the genetic unit mostly represented approximately 100 km apart using STRUCTURE or that had a negative AIc and thus had a low probability of being assigned to the population of origin. This result contrasts with GB_Cantanhez, where all the males were assigned to the same genetic unit and 64% displayed a positive AIc. Furthermore, although possibly limited by a small sample size in GB_Cufada and GB_Boé (i.e., N = 51 and 21 genotypes, respectively, when 50 individuals is the minimum sample size recommended, [62]), the analyses detecting first-generation migrants identified two males sampled in Cufada with an origin in Boé and two males sampled in Boé with an origin in Cufada. The immigrant male from Boé sampled in GB_Cufada and identified assuming a more restrictive migration rate harboured a mtDNA haplotype with a high frequency in GB_Boé [25] and was assigned to the genetic unit mostly represented in GB_Boé with a high probability (i.e., q > 0.90). Taken together, these results suggest male-mediated gene flow in southern Guinea-Bissau between GB_Cufada and Boé or a genetically close population (Table 1).

The result of male-mediated gene flow in southern Guinea-Bissau should be treated cautiously. Firstly, since faecal sampling comprised unidentified individuals, we cannot be sure that the putative male immigrants dispersed during their lifetime; some males could be pre-dispersal individuals, could be the offspring of immigrant lactating females [64] or admixed individuals originated from mating between immigrant females and resident males. For instance, 41% of the males sampled in GB_Cufada were classified as having admixed ancestry, which might explain the contrast in the proportion of males identified as immigrants in different analyses (e.g. the great majority of males at GB_Cufada displayed a negative AIc but a smaller proportion were assigned to the differentiated cluster 2 using STRUCTURE, Fig 2). Secondly, the analyses would benefit from a larger sample of males. Although non-invasive faecal sampling of unidentified and unobserved individuals was applied in Senegal and Guinea-Bissau, a lower proportion of males were sampled in GB_Cantanhez and GB_Cufada (0.34 and 0.33, respectively) when compared to Senegal and GB_Boé (0.59 and 0.62, respectively). Since ADO between studies did not vary significantly, high ADO is unlikely to result in a failure to detect male individuals in Guinea-Bissau. A more likely explanation is the negative demographic consequences of intensive hunting practices towards males in Guinea-Bissau. Hunters have stated during interviews that they target male baboons first, due to their larger body size and consequent higher economic value [37, 65]. Male carcasses are 60% more expensive than females and other primate species at Guinea-Bissau bushmeat markets [24]. However, a lack of quantitative data from Guinea-Bissau bushmeat markets prevents us from confirming of the male-directed hunting hypothesis. Thirdly, comparison between sex-specific genetic markers (i.e., mtDNA and Y-chromosome associated markers) and autosomal microsatellites loci allows for the investigation of male-specific population structure and to explicitly test male-biased gene flow in Guinea-Bissau. However, genetic variation of the Y chromosome seems to be low in the Guinea baboon and a set of variable Y-linked chromosomal microsatellite markers is not available [33]. Nevertheless, considering that we randomly sampled double the number of females in GB_Cufada than males (34 females and 17 males, S1 Appendix) and that females and males were assigned to cluster 2 in similar proportions (12% vs. 18%, respectively) in that locality, our results suggest that the population of baboons at GB_Cufada comprises a proportion of genetically differentiated males (Table 1).

Our study analysed intra-specific variation in dispersal patterns using individual-based molecular tools and a comparative framework, an approach that has been applied successfully in a variety of species and landscape contexts (e.g. [13, 66–67]). However, few studies have been able to analyse direct dispersal or have attempted to compare variation in dispersal for populations under different levels of anthropogenic pressure at comparable spatial scales [68]. Molecular tools can be a solution in such cases but a combination of factors related to sampling design (inclusion of pre-dispersers in the dataset, small sample size and a different geographic distance between sites) and differences between sexes in dispersal rates can confound the detection of sex-biased dispersal [31, 32].

Here, sampling strategies were comparable between Senegal and Guinea-Bissau with respect to sample size (e.g. NGB66 = 111 and NSEN = 165), genetic markers (S5 Appendix) and geographic scale(s) but methodological constraints related to the study of wild primates may limit the generalization of our conclusions to other populations. Firstly, the inclusion of a single site in an un-fragmented population (i.e., Senegal) limits the comparison of gene flow patterns between fragmented and non-fragmented populations to a maximum of 66 km with the consequent exclusion of genotypes from the GB165 sub-set, which in turn, may have an effect in the characterization of the patterns of sex-specific gene flow. Spatial autocorrelation for different sub-sets of samples (e.g. GB165, GB_Cufada and GB_Cantanhez only, and GB66) is slightly different, although the overall pattern remained the same (Figure E in S6 Appendix). Also, the genetic structure estimated using STRUCTURE in GB165 was not detected using the GB66 sub-set, possibly because of the lower number of genetically differentiated individuals which remained in the analyses using GB66 [69]. For instance, Guebombol sampling site was excluded from the comparative analyses because was distanced more than 66 km of sampling sites in GB_Cantanhez but was formed exclusively by males assigned to the genetic unit most frequently represented in GB_Boé (Fig 2a). These results suggest that using GB66 sub-set alone can lead to incomplete inference. However, it is worth highlighting that the IUCN’s Near threatened Guinea baboon suffered a widespread decline in the last thirty years [35, 36] and that the current conservation status of the populations outside protected areas is unknown [70]. Therefore, the likelihood of locating a second continuously distributed population of a comparably large range in Senegal (i.e., for 165 km) is low. Secondly, the statistical power of detecting a sex-bias in dispersal is limited by the inclusion of pre-dispersing individuals in both datasets [32]. However, a clear signal of sex-biased dispersal was not erased in Senegal and it is unlikely that the Guinea-Bissau population include a significantly greater number of pre-dispersing individuals than Senegal. Therefore, we conclude that, by contrasting the differences in rate and extent of sex-mediated gene flow in Guinea-Bissau and Senegal, we are likely to be highlighting an important biological phenomenon.

For the smaller spatial scale of 66 km, for which a comparison with Senegal was performed, we could not find evidence of sex-biased dispersal in Guinea-Bissau. Both males and females at GB66 display a spatial autocorrelation pattern expected for the philopatric sex (Table 1). These results contrast with the clear pattern of female-biased dispersal found for Senegal for the same geographic scale. When comparing the spatial autocorrelation of each sex between GB66 and SEN66, we could observe that (1) males of the two populations show a similar spatial genetic pattern, although SEN66 males show a stronger genetic structure at larger distances and (2) females in Guinea-Bissau are significantly more similar than females in Senegal at closer distances.

Restricted gene flow for males and females across human-dominated areas between GB_Cufada and GB_Cantanhez do not disagree with the pattern of female-biased mediated gene flow found at a broad scale in Guinea-Bissau and the evidence of contact between genetically differentiated males in GB_Cufada. In all the spatial autocorrelation analyses performed, the pattern of limited dispersal between GB_Cufada and GB_Cantanhez is clear for both sexes. SEN66 males show a stronger genetic structure than GB66 males but we inferred hindered male gene flow between GB_Cufada and GB_Cantanhez. In Senegal, males show a genetic structure which agrees with the geographic location of sampling sites (GD+SI+CL+LK and NK, Fig 3a) whereas we found that GB_Cufada is most likely formed by mixture of genetically differentiated males, a proportion of which is genetically similar to GB_Cantanhez males (Fig 2a). This result explains the similar genetic structure pattern between males in Guinea-Bissau and Senegal but weaker genetic structure for Guinea-Bissau males at larger distances when compared to Senegal males. Taken together and considering the pattern of historical female-mediated gene flow estimated by past studies [23, 25], our study suggests a modification in dispersal patterns for the Guinea-Bissau population from a female-biased dispersal pattern at a broad geographical scale to i) male-mediated gene flow towards GB_Cufada from a differentiated population genetically similar to GB_Boé, and ii) a lack of sex bias in dispersal at a smaller scale between GB_Cufada and GB_Cantanhez.

This apparent alteration in dispersal patterns can be explained by behavioral responses to fragmentation. Anthropogenic hunting and deforestation at southern Guinea-Bissau [22, 24, 25] may have altered population density, sex ratio, and age structure and/or induced defensive behaviors, such as avoidance of specific areas and altered ranging patterns [37], changes which potentially interfere with dispersal [2, 5]. Foraging and social grouping in Guinea baboons vary according to environmental contexts [26, 71] and, most likely, to anthropogenic environmental changes, given the importance of these variables in restricting the species occurrence [70]. Other empirical studies found exceptions to the pattern defined for the target species in anthropogenic environments (e.g. Pongo pygmaeus [72]; Ateles sp. [13]), including for primate species facing the same type of threats in Guinea-Bissau (Colobus polykomos and Piliocolobus temminckii [67] and Pan troglodytes verus [73]). However, considering that many socio-ecological features of the Guinea baboon remain unknown (Senegal: [26, 29, 38], Guinea-Bissau: [23, 25]) we cannot exclude the possibility that the species may display population-specific plasticity in dispersal and gene flow mediated by both sexes may not be atypical. Three possible non-exclusive scenarios to explain gene flow mediated by both sexes in southern Guinea-Bissau that could be tested:

1. Firstly, if we consider that the difference in the proportion of genetically differentiated individuals co-existing in Cufada between males and females is a sampling artefact, our results could be interpreted as if entire social groups moved to the Cufada Lagoons Natural Park. One reason for the displacement of entire social groups could be the high mortality risk associated to human disturbance [74, 75], which is probable outside protected areas in Guinea-Bissau [22].
2. Secondly, while Guinea-Bissau females might be displaying the same propensity for dispersal as Senegal females, males in Guinea-Bissau may have recently started to disperse towards GB_Cufada from GB_Boé or a genetically similar population. One reason could be a lower number of males at GB_Cufada, caused by a preference to first hunt male baboons due to a higher monetary return [24, 37, 75]. The consequent reduction in number of males could lower the competition for females and food resources and thus, induce male immigration towards the vacant home ranges (“vacuum effect” [76]; e.g. Panthera leo [77]). Although such an effect has not been reported for primate species, sex-specific life history strategies, unequal intra-sexual competition and different sensitivity to environmental factors between the sexes have been related to a sex-bias in dispersal for other species [5]. Furthermore, past studies have associated movements of male baboons towards areas with greater accessibility to cycling females to improve their reproductive opportunities. For instance, in some populations of Papio hamadryas, a species in which dispersal is described to be female-biased [78], males sometimes disperse to neighbouring olive baboon groups (Papio anubis) due to a higher success in acquiring females [79].
3. Finally, we need to consider that the differences in dispersal patterns between Guinea-Bissau and Senegal could have been accumulated during their evolutionary history. For species that have historically expanded their range, which could be the case of the Guinea baboon [80], both theoretical considerations and simulation experiments predict a greater abundance of individuals with higher dispersal capacity at the margins of the distribution (such as in Guinea-Bissau) compared with the core (as Senegal) (e.g. [81, 82]). The effect of spatial gradients of dispersal capacity may last for a considerable amount of time after the species is settled in the new range [82]. Thus, our results may also imply an intrinsic higher dispersal capacity of the Guinea-Bissau individuals as a response to changes in the environment, a characteristic that may have been selected by evolutionary mechanisms. Data on dispersal patterns from other marginal populations are needed to test this hypothesis.

Our results do not suggest an absence of gene flow but female dispersal between GB_Cufada and GB_Cantanhez seem to have been modestly or recently restricted when compared to the same geographic scale in Senegal. One first generation female migrant with an origin in GB_Cufada sampled in GB_Cantanhez was detected, which suggest current gene flow between those localities. However, that specific female was sampled in the marginal area of Cantanhez Woodlands National Park. Significant genetic dissimilarity between females from inside and outside the Peninsula of Cantanhez (Figure G in S7 Appendix) suggest lower levels of gene flow across the land bridge connecting the Peninsula to mainland Guinea-Bissau. Although levels of inbreeding do not seem to be particularly high, GB_Cantanhez show a tendency of lower levels of genetic diversity (Table C in S5 Appendix). Overall, these results suggest a modest level of isolation of the baboons’ population from the Peninsula of Cantanhez, which may have important implications for their conservation. Considering the long generation times of primate species, the genetic consequences of a recent restriction in gene flow between GB_Cufada and GB_Cantanhez could have just started to manifest.

Supporting information

Acknowledgments

Field assistants in Guinea-Bissau—M. Soares and M. Turé, park guides and guards, the DARI project, the NGOs CHIMBO and AD, the Guinea-Bissau governmental agencies Direção Geral de Florestas e Fauna and Instituto da Biodiversidade e Áreas Protegidas (IBAP) for samples exportation permits and logistical support during fieldwork, Instituto Nacional de Estudos e Pesquisas and Gabinete de Planificação Costeira for GIS maps, RAMA pharmacy and Café au lait for in-kind sponsorship, C. Schwarz, M. Nanqui, R. Dixe, C. Pereira and J. Huet for logistical support in Bissau, C. Roos (DPZ) and MWB lab members for lab work support, and J. Torres for map design. SEN Diréction des Parcs Nationaux and Ministère de l´Environnement et de la Protéction de la Nature de la République du Sénégal for permission to work and exporting samples from Senegal, the Conservator of the park Cdt. Samuel Diemé for support and cooperation, Anja Ebenau, Jacky Bassene, Samba Ciss, and Issakha N’Diaye for collecting samples.