https://orcid.org/0000-0001-8914-8631
Figures
Abstract
African trypanosomosis (AT), caused by protozoan parasites of the genus Trypanosoma, has plagued the African continent for centuries, affecting both humans and animals. Its principal vector, tsetse flies, can be found across sub-Saharan Africa. Vector control represents an efficient way to reduce the burden of AT. In Mozambique, control campaigns reshaped tsetse fly distribution to what it is today, with four species presently found: Glossina brevipalpis, G. pallidipes, G. morsitans and, G. austeni. Additionally, G. brevipalpis can be observed in two National parks, Gorongosa National Park in the Centre and Maputo National Park in the South, with an 840 km wide tsetse-free zone between them. In order to improve our knowledge on the genetic diversity in these populations, and their probable isolation, we undertook a population genetics study with 11 microsatellite loci. We found that these two zones behave as strongly isolated subpopulations, only exchanging a few individuals per year. To explain this finding, we suggest the existence of undocumented pocket populations between the two parks, or, in the absence of these, the accidental translocation of tsetse flies during human-driven animal transportation. We suggest that translocation through human-driven animal movement should be explored in future studies investigating Glossina populations. If eradication were to be attempted, re-invasion of the tsetse via motorized human transport should be considered in conjunction with the exploration of other sites within a 30 km radius to validate that no sources of re-invasion exist around these parks.
Author summary
African trypanosomoses (AT), is caused by single celled parasites of the genus Trypanosoma. These have plagued the African continent for centuries, affecting both humans and animals. Its principal vector, tsetse flies, can be found across sub-Saharan Africa. Vector control represents an efficient way to reduce the burden of AT. In Mozambique, control campaigns reshaped tsetse fly distribution to what it is today, with four species presently found: Glossina brevipalpis, G. pallidipes, G. morsitans and, G. austeni. Additionally, G. brevipalpis can be observed in two National parks, Gorongosa National Park in the Centre and Maputo National Park in the South, with an 840 km wide tsetse-free zone between them. In the present work, we have analysed the genetic variation of this tsetse fly within and across these two parks. We found that these two zones behave as strongly isolated subpopulations, only exchanging a few individuals per year. Within parks, average dispersal appeared to reach 30 km per generation. To explain this finding, we suggest the existence of undocumented pocket populations between the two parks. In the absence of such pockets, the most probable explanation is the accidental translocation of tsetse flies during human-driven animal transportation. If eradication were to be attempted, re-invasion of the tsetse via motorized human transport should be considered in conjunction with the exploration of other sites within a 30 km radius to validate that no sources of re-invasion exist around these parks.
Citation: Brito DRA, Ségard A, Mulandane FC, Cossa NV, Mucache HN, Ravel S, et al. (2025) Unravelling genetic differentiation between Glossina brevipalpis populations from two distant National Parks in Mozambique. PLoS Negl Trop Dis 19(5): e0012953. https://doi.org/10.1371/journal.pntd.0012953
Editor: Brian L. Weiss, Yale School of Public Health, UNITED STATES OF AMERICA
Received: February 27, 2025; Accepted: May 12, 2025; Published: May 30, 2025
Copyright: © 2025 Brito et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting Information files.
Funding: This project has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement n°101000467, acronym ‘’COMBAT’’ (Controlling and Progressively Minimizing the Burden of Animal Trypanosomosis). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
African trypanosomosis (AT), caused by protozoan parasites of the genus Trypanosoma, has plagued the African continent for centuries, affecting both humans and animals, causing a heavy economic burden [1–4]. These blood parasites have a diverse range of mammalian hosts [1]. Their primary biological vectors are the hematophagous flies known as tsetse (Glossina spp). Control of tsetse flies has been conducted over the last and present century so as to reduce the burden caused by both human (HAT) and animal (AAT) African trypanosomosis [3,5,6].
Glossina species are classified into three main groups (Fusca, Palpalis and Morsitans). These presently include a total of 31 species/subspecies adapted to specific habitats [3,5,7]. The Fusca group corresponds to forest-dwelling flies that are mostly found in western-central Africa, except for Glossina longipennis and G. brevipalpis, which are located in eastern and southern Africa [3,5]. Considering the case of Mozambique, four species of tsetse flies are found in the country, namely G. austeni, G. morsitans and G. pallidipes from the Morsitans group and G. brevipalpis of the Fusca group [8–10]. In this country, the southern fly belt consists of two species (G. austeni and G. brevipalpis), while the central and northern belts are composed of the four tsetse species [8–10].
In the late 1890s, tsetse fly populations suffered a significant reduction, partially due to the occurrence of a rinderpest pandemic that drastically diminished the populations of both wild and domestic ungulates, particularly in southern Africa, including Mozambique [5,11]. Tsetse flies survived in small pockets, allowing for a subsequent reoccurrence and expansion of the extremely reduced Glossina populations once the presence of game and cattle was re-established [5,11]. Nevertheless, in South Africa and Mozambique, this event was followed, between 1920 and 1970, by important vector control programs against all tsetse fly species [12–14]. These control campaigns reshaped the tsetse distribution to what it is today, with the dominance of G. brevipalpis and G. austeni in several zones of Southern Africa, and in particular in Mozambique [11–13,15–17].
Glossina brevipalpis has been detected in Kenya, Rwanda, Tanzania, Zambia, Mozambique and South Africa [18]. The distribution of G. brevipalpis is limited by its habitat (forests with bush clumps for its breeding sites) and climate. It is specifically influenced by temperature (16 °C to 32 °C) and relative humidity (≥ 70% of relative humidity) [3,14,15]. In Mozambique, the southern provinces, Gaza and Inhambane, have remained tsetse-free for the last 120 years, without evidence of tsetse fly reinvasion. This separates the central population of G. brevipalpis, mainly in the Gorongosa National Park, from the southern population of the same species, primarily situated in the Maputo National Park [18] (Fig 1).
Traps contained in each site and their precise GPS coordinates can be seen in S1 Table. (This map was produced from base map imported from The Humanitarian Data Exchange - https://data.humdata.org/dataset/cod-ab-moz).
One very efficient strategy to reduce trypanosomosis incidence is vector control [19]. In order to design the most efficient control strategy, understanding the dynamics of the existing populations of Glossina is important [20]. It is not only essential to know the existing species and their distribution within a country, but also to understand the genetic structure and gene flow between their populations [20,21]. The use of polymorphic genetic markers and population genetic tools can provide key information on the structure of targeted populations, reproductive strategies, gene flow, population sizes and dispersal distances. It can thus lead to designing the best strategies for control [20,21]. For this purpose, microsatellites have been extensively used for tsetse fly population genetics over the last two decades [21–25].
The present paper aims to assess the genetic diversity and gene flow within and between the central and southern Mozambique populations of G. brevipalpis located in the Gorongosa and Maputo National parks respectively.
Methods
Ethics statement
All applicable international, national, and/or institutional guidelines for the care and use of animals were followed and all procedures performed in studies involving animals were in accordance with the ethical standards of Biotechnology Centre – Eduardo Mondlane University and the practice at which the study was conducted (Animal Research Ethics Committee of the Biotechnology Centre of the Eduardo Mondlane University (CEPA-CBUEM); approval number: CEPA-CBUEM 05/2023). The sampling procedures reported herein were authorized by the respective park authorities: For Gorongosa National Park the Department of Scientific Services of Gorongosa National Park issued the permit PNG/DSCi/C231/2022; For Maputo National Park: permit 03/01/2022 issued by the National Administration for the Conservation Areas.
Study sites and sample collection
Gorongosa National Park (GNP) is located in the central region of Mozambique. It was originally established as a hunting reserve in the 1920s and now includes 3,770 km2 of protected land, which is composed of a mix of forest and savannah landscapes, with a high diversity of fauna and flora. This area includes four species of tsetse flies, namely G. pallidipes, G. morsitans, G. brevipalpis and G. austeni. About 840 km to the south, in the most southern point of the country, there is a tsetse hot spot that partially overlaps with the Maputo National Park (MNP). The park was established in 1960 as a national reserve to protect the extensive population of elephants found in the region. Today, it covers a total of 1,718 km2, comprising various ecosystems including forests that harbour two species of tsetse flies, G. brevipalpis and G. austeni. Between these two parks, including northern Maputo province, Gaza and Inhambane provinces, a putatively tsetse-free area has existed since the rinderpest pandemic of 1896 [12]. For this reason, these two parks were chosen as sampling areas to study gene flow of G. brevipalpis populations between the south and central regions of Mozambique.
Six hundred and seventy-nine (679) G. brevipalpis specimens were collected in GNP (342) and MNP (337) using 49 traps (36 H-traps [26] and 8 NGUs traps [27], enhanced with odour attractants (1-octen-3-ol and acetone) [26–28]). We deployed H traps in the two parks to capture G. brevipalpis. However, we also deployed NGU traps to capture other Glossina species for another study, which caught some G. brevipalpis individuals used in the present study. All flies were captured between June 14 and July 22 2022 corresponding to less than one tsetse fly generation [7]. Twenty-nine (29) traps were deployed in GNP and 20 in MNP (Fig 1). In each park, we selected four different sites in which two to six traps were placed. Neighbouring traps covered an average distance of 272 m with a minimal distance of 150 m and a maximum of 516 m (Fig 1). All traps were georeferenced using the eTrex 10 GPS (Garmin Ltd., USA) (See S1 Table). The overall sampling transects, i.e., the distance between the most extreme traps, reached 25 km in GNP and 14 km in MNP. Tsetse flies were identified using morphological criteria [29].
DNA extraction and genotyping
Four hundred and two (402) G. brevipalpis (GNP – 202 and MNP – 200) were selected for genotyping based on the location and condition of specimens (S1 Table). DNA was extracted from 3 legs per individual using the chelex method [30] after which the DNA was diluted for amplification. We used 11 microsatellite primers developed by Gstöttenmayer et al., (2023) [31] to genotype individual flies collected from both parks: Gb5, Gb28, Gb35, Gb48, Gb66, Gb70, Gb72, Gb73, Gb92, Gb158, and Gb165. To incorporate the dyes into the PCR product, microsatellite forward primers contained the M13 adapter sequence, and amplification was done with M13 primers attached to one of these four dyes: VIC, NED, PET and FAM. PCRs were carried out in a 20 µL reaction with 1x PCR Buffer, 0.2 mM DNTPs, 0.08 µM forward primer, 0.1 µM reverse primer, 0.1 µM M13 primer with dye, 0.5 U Taq polymerase and 10 µL of diluted DNA. The PCR cycling conditions were as follows: 95 ○C for 3 mins and then 10 cycles of 95 ○C for 30 s, Ta (58/59 ○C) + 5 ○C for 30 s, 72 ○C for 1 min dropping 0.5 ○C each cycle, then 30 cycles of 95 ○C for 30 s, Ta (58/59 ○C) for 30 s, 72 ○C for 1 min, with final elongation step at 72 ○C for 5 min. The PCR products were pooled by four (the different dyes) and resolved on an ABI 3500XL sequencer (Thermo Fisher Scientific, USA). Allele calling was done using GeneMapper Software v6.1 [32] and the size standard GS600LIZ (Thermo Fisher Scientific, USA).
Data analysis
Genotypic data were formatted into the appropriate file types using Create v1.37 [33]. Analyses were done using 10 autosomal loci developed by Gstöttenmayer et al., (2023) [31] (S2 Table). As Gb70 is sex-linked [31], it was removed from our dataset.
Quality testing of the sampling
To describe the population genetic structure of G. brevipalpis, we used Wright’s F-statistics [34]: FIS, which measures the effect of deviation from random mating within subpopulations on inbreeding, and FST, which measures the effect of subdivision. These were respectively estimated with Weir and Cockerham’s (1984) [35] unbiased estimators (θ and f). Their significant deviation from 0 was tested with permutations of alleles between individuals within subsamples (for FIS) and individuals between subsamples (for subdivision). In each case, the statistic used was the FIS estimator or the natural logarithm of the maximum likelihood ratio (G) [36], respectively. We also tested linkage disequilibrium between each pair of loci with the G-based randomization test over all subsamples [37]. For each test, we implemented 10,000 permutations. Estimates and testing were undertaken with Fstat v2.9.4 [38].
A Wahlund effect occurs when individuals belonging to subpopulations with different allele frequencies are admixed into the same subsample. This was first used to explain heterozygote deficits as compared to Hardy-Weinberg expectations following this phenomenon [39]. Nonetheless, Wahlund effects can also minimize subdivision measures, and increase linkage disequilibrium between loci. To determine the levels at which G. brevipalpis from Mozambique were subdivided, we used the Wahlund effect detection technique of Goudet et al., (1994) [40]. For this, we considered four sampling designs, depending on the level considered: trap, location, park, and all flies together (All). If a Wahlund effect occurs at one level, a significant change should be observed. We thus computed FIS-trap, FIS-location, FIS-park, and FIS-All, and compared them using the Wilcoxon signed rank test [41] for paired data using R-commander package v2.9-5 (Rcmdr) [42,43] for R v4.4.2 [44], with the alternative hypotheses: FIS-trap < FIS-location < FIS-park < FIS-All. We also compared FST’s measured at these levels (except All). For those, we undertook the same test but with a reverse alternative hypothesis. We finally compared the proportions of locus pairs in significant LD with one-sided Fisher exact tests [45] with R v4.4.2 (command fisher.test). Because of the non-independent test series undertaken for each parameter compared, we needed to adjust the obtained p-values with the Benjamini and Yekutieli (BY) procedure [46], with the command “p.adjust” in R v4.4.2.
Quality testing of the loci
We first checked the statistical independence between each locus pair with the G-based randomization test at the BY level of significance. We also examined the deviations from expected genotypic frequencies with FIS estimates and testing as described above. We additionally computed 95% confidence intervals with 5,000 bootstraps over loci for the average of FIS and FST with Fstat v2.9.4. To obtain 95% CI of FIS for each locus, we used 5,000 bootstraps of individuals in each subsample for each locus with Genetix v4.05.2 [47]. We computed the average of FIS and its 95% CI for each locus obtained with Genetix v4.05.2 weighted by the number of visible genotypes and the local genetic diversity as estimated by Nei’s unbiased estimator (HS) estimated with Fstat v2.9.4. We estimated the variation of FIS and FST across loci with the standard error (SE(FIS) and SE(FST)) obtained by jackknives over loci with Fstat v2.9.4. The presence of null alleles was assessed with several criteria as described elsewhere [48,49]. In the case of null alleles, the ratio ,a positive correlation is expected between FIS and FST, and between the number of missing data (Nmissing) and FIS. Correlations were assessed and their significance tested with one-sided Spearman’s rank correlation using Rcmdr v2.9-5. Null allele frequencies (pnulls) were then estimated with the EM algorithm [50] using FreeNA [51]. For this, we recoded missing data as homozygotes for allele 999 as advised [51], only for loci for which missing data indeed corresponded to true null homozygotes (see Results section). Finally, we undertook the regression FIS ~ pnulls with its 95% CI of bootstraps and computed its determination coefficient and the intercept, which should correspond to the value of FIS in the absence of null alleles (FIS-0).
Analyses of population subdivision and gene flow
We estimated subdivision with FreeNA and the ENA algorithm to correct for null alleles (FST-ENA). For this, we recoded missing genotypes suspected to correspond to true null homozygotes as 999999 as recommended [51]. Microsatellite loci generally display a high level of polymorphism, leading FST to reflect mutation and immigration together. To correct for this excess of polymorphism, we used the FST’ approach [52,53]. The maximum possible FST (FST-max) was computed with Fstat v2.9.4 after allele recoding by RecodeData v0.1 [53], which was used to compute . Meirmans and Hedrick proposed an alternative correction (GST“) [54] which is expected to be more accurate. Nevertheless, GST” neither allows correction for null alleles nor the estimate of 95% CIs with available programs.
We estimated the probable number of immigrant flies exchanged between the two parks as , assuming an infinite Island model (n = infinite), or
, assuming a two-island model (n = 2) [55].
Effective population sizes
We estimated effective population sizes with five methods: the heterozygote excess method (Hex) from De Meeûs and Noûs, (2023) [56]; the linkage disequilibrium method (LD) [57] corrected for missing data [58]; the Coancestry method (CoA) [59]; the one locus and two loci correlation method (1L2L) [60]; and the sibship frequency method (Sib) [61]. For Hex we estimated Ne from the published formula in a spreadsheet program from FIS values obtained in each subsample for each locus and averaged across loci as recommended [56]. For LD and CoA, we used NeEstimator v2.1 [62]; for 1L2L we used Estim v1.2 [60]; and finally, for Sib, we used Colony [63]. We also kept the minimum (min) and maximum (max) values and averaged all estimates across methods by weighting those with the number of usable figures obtained, as recommended [64]. In case of unusable values (i.e., “Infinite”), and in order to compare effective population sizes between the two parks, we retrieved the 95% CI outputted by the softwares used (bootstraps over individuals for Hex, 1L2L, and Sib, jackknives for CoA and parametric for LD). As suggested by Waples [65], we then replaced all “Infinite” by a very big figure (here we chose 10 times the maximum observed one, 30000), and computed harmonic means of all Ne and 95%CI across methods and overall.
Time of isolation or gene flow between the two parks
We used the averaged effective population size estimate to compute the number of generations of the split between GNP and MNP subpopulations, assuming total isolation, with the formula (e.g., Hedrick (2005), equation 9.13a, p 502 [52]). We also extracted the immigration rate (m) from Nem estimated with n = 2 or n = infinite and used it to compute the average dispersal distances covered by tsetse flies per generation, assuming rare transportation between the two parks, as
and its 95% CI.
Isolation by distance, sex biased dispersal between traps and bottleneck within each park
We undertook isolation by distance analyses within each park separately (to avoid a two-points regression), and to confirm the free circulation of tsetse flies in these two areas. We used Rousset’s model [55] in two dimensions, where FR is Rousset’s genetic distance between two traps, a is the intercept, b is the slope of the corresponding regression and ln(Dgeo) is the natural logarithm of the geographic distance between the two traps. Rousset showed that, in case of isolation by distance b is the reverse of the neighbourhood (number of connected individuals, Nb = 1/b). To compute FR, we used the ENA algorithm [51] to correct for the effect of null alleles [66]. Between each pair of traps we computed FST-ENA and its 95%CI with 5000 bootstraps over loci with FreeNA [51], and computed geographic distances with the package geosphere [67] for R (command distGeo). We then computed FR = FST-ENA/(1-FST-ENA) and its 95%CI and regressed those against ln(Dgeo) to obtain the slope of isolation by distance and its 95%CI.
A genetic signature of sex biased dispersal can only be detected for very big differences between genders, e.g., when one sex is highly philopatric and the other disperse a lot [68]. This thus applies to weakly subdivided populations. We randomized the gender of each fly 10,000 times to test for the significant difference of the corrected assignment index, AIc, its variance, vAIc, and FST. We did these computations and corresponding tests (two-sided) with HierFstat [69] for R v4.4.2.
In order to help the interpretation of results, we also tried to detect a genetic signature of a past bottleneck that would correspond to the rinderpest pandemic that occurred in the years 1898 – 1901, or the massive vector control that was undertaken in the years 1948 – 1970. We implemented this test with the software Bottleneck [70]. As recommended (De Meeûs et al., (2021), p108-109 [71]), a significant bottleneck signature was recognised when the Wilcoxon test outputted significant p-values for the infinite allele model (IAM) and the two-phase mutation model (TPM), at least. We obtained a global p-value across the two parks with the generalized binomial procedure [72], using k’ = 2 as recommended [73], with the software MultiTest v1.2 [37].
Bayesian clustering and neighbour-joining trees
Additionally, we used the Bayesian clustering method implemented by the software STRUCTURE v2.3.4 [74], with a 5,000 burning period and 50,000 MCMC iterations, a number of clusters from 1 to 4 and 10 replicates. We then looked for the optimal partition with the method proposed by Evanno et al.,‘s (2005) [75] with the online facility StructureSelector [76]. We hoped that this approach would help elucidate the pattern of dispersal between the two parks through the detection of probable immigrants or individuals that probably descended from immigrant parents a few generations ago.
To validate the result produced by STRUCTURE v2.3.4, we used a Neighbour-Joining (NJ) approach. We computed Cavalli-Sforza and Edward’s chord distance [77] corrected for null alleles [51] between all individuals of G. brevipalpis from GNP and MNP. The Genetic distance matrix was then used to build an NJTree with MEGA X [78]. Another two trees were built using only individuals with complete genotypes (no missing data). With this alternative dataset, we also used PopTree [79], which uses Nei’s DA genetic distance [80], with 1000 bootstraps. All trees were edited in Interactive Tree of Life (iTOL) v6 [81].
Results
Of the 402 tsetse flies submitted to genotyping, 396 were successfully genotyped (i.e., with more than 4 loci amplified). These 396 genotypes (196 GNP and 200 MNP) were used for further analyses and are presented in S2 Table.
Quality testing of the sampling
Significance only occurred for comparisons with FIS and LD, and only between “All” (all traps of the two parks pooled into a single subsample) and each of the three other sampling designs (all pBY < 0.005). This means that migration of G. brevipalpis is free within the sampling areas in each park, i.e., across 25 km in GNP and 14 km in MNP. Consequently, for subsequent analyses, the level of subpopulation used was the park (i.e., Gorongosa and Maputo).
Quality testing of the loci
We found five locus pairs in significant LD, one of which (Gb5 and Gb66) remained significant after BY correction (pBY = 0.02). There was a significant heterozygote deficit: FIS = 0.166 in 95% CI = [0.084, 0.256] (p-value < 0.0002). It seemed that null alleles explained these figures well. Indeed, rSE ≈ 3, the correlation between FIS and FST was positive and significant (0.6848, p-value = 0.0175). Nevertheless, the correlation between Nmissing and FIS was negative (ρ = -0.0788, p-value = 0.5943) because of an excess of missing genotypes at seven loci. Indeed, it can be seen from Fig 2, that the correspondence of missing genotypes with homozygotes for null alleles may be true for only three loci (Gb35, Gb73, and Gb165). Consequently, for null allele frequency estimates, we recoded missing data as 999999 only for these three loci. The resulting regression FIS ~ pnulls provided a coefficient of determination (R2) below 0.9 (R2 = 0.8584, FIS-0 = 0.0316). This came from two outlier loci displaying too small FIS when compared to the corresponding null allele frequency: loci Gb73 and Gb165. We thus chose to recode missing data for these two loci as “000000” because most blanks probably corresponded to other amplification failures than null alleles (hence only Gb35 remained 999999 for missing genotypes). With this new dataset, we obtained a very good adjustment (Fig 3). The model indeed explained almost 100% of FIS variation across loci, with a negative intercept (FIS-0 = - 0.0043 in 95% CI = [- 0.0704, 0.059]), compatible with pangamic populations of average effective population size Ne = 116 in 95% Confidence Interval (CI) = [7, Infinite] [56].
The average (straight line) and its 95% CI (dotted lines) are represented. Loci for which missing data probably corresponded to null homozygotes only were used for the regression and are indicated with black crosses. Other loci are indicated by their names and empty circles with their 95% CI (black dashes).
The average equation (straight line) and its determining coefficient (R2) and of the 95% CI (dotted lines) are represented.
To conclude, two loci, Gb5 and Gb66, are in statistical linkage and thus may introduce redundancy in multi-locus-based tests and parameter estimates. Locus Gb5 displayed a significant heterozygote deficit FIS = 0.0674 in 95% CI = [0.0082, 0.1244], while locus Gb66 displayed a non-significant excess: FIS = -0.0058 in 95% CI = [-0.0636, 0.0506]. Subdivision at these two loci (0.032 and 0.016 respectively) appeared significantly smaller than the average FST = 0.066 in 95%CI = [0.038, 0.099]. A glance at https://genome.ucsc.edu/ for G. brevipalpis draft genome, and a BLAT [82] of primer sequences of these two loci revealed that these two loci were not found in the same scaffold. It is thus highly unlikely that these sequences are close to each other. The statistical linkage we found may, therefore, be a consequence of selective events driving these two markers correlatively. We thus chose to exclude loci Gb5 and Gb66 from further analyses and kept coding missing genotypes as null homozygotes (999999) for Gb35 only as regard to subdivision estimates with FreeNA.
With the eight remaining loci (Gb28, Gb35, Gb48, Gb72, Gb73, Gb92, Gb158, and Gb165), no locus pair remained significant LD after BY correction (all pBY = 1). There was a significant heterozygote deficit FIS = 0.205 in 95% CI = [0.114, 0.299] (two-sided p-value < 0.0002). Additionally, we observed a higher FIS = 0.276 in MNP than in GNP (FIS = 0.132) (p-value = 0.0039, Wilcoxon signed rank test). This probably came from a difference in null allele frequencies, which appeared to be much higher in MNP than in GNP (0.1164 and 0.0531 respectively, p-value = 0.0039, Wilcoxon signed rank test). Alternatively, genetic diversity did not vary significantly between MNP and GNP (HS = 0.7398 and 0.7868, respectively, p-value = 0.9512).
Population subdivision and gene flow
We observed a highly significant subdivision between the two parks (p-value < 0.0001), with FST-ENA’ = 0.2693 in 95% CI = [0.1631, 0.3964]. With these figures, we could estimate the number of immigrants exchanged between the two subpopulations Nem = 0.70 in 95% CI = [1.36, 0.39] individuals per generation with n = infinite, and Nem = 0.35 in 95% CI = [0.20, 0.68] with n = 2.
Effective population sizes
For the sake of consistency in Hex results, and to get less “Infinite” results with that method, we used the averaged FIS from Genetix v 4.05 computations and its 95%CI for each locus and each park. We observed two “infinite” estimates of effective population sizes in MNP (for heterozygote excess and CoAncesty methods), which could not be taken into account for the average. Consequently, the following estimates probably corresponded to underestimations of the real average effective population size, especially so for MNP. The average effective population size was Ne = 844 in minmax = [676, 1012]. GNP displayed a smaller Ne = 629 in minmax = [21, 2068] compared to MNP with Ne = 1202 in minmax = [110, 3328], but the significance of this difference could not be tested. Nevertheless, for the three methods for which MNP was not “infinite”, it always gave a higher Ne than the one in GNP. As can be seen from the Fig 4, harmonic means confirmed this trend, though 95%CI largely overlapped. The arithmetic average appeared significantly higher than what was computed by Gstöttenmayer et al., (2023) [31]. The intercept of the regression of FIS ~ pnulls and harmonic means, also provided significantly smaller figures (Fig 4). This may mean that the real Ne lies between 200 and 1000 individuals (Fig 4).
We present results obtained with arithmetic means (ARI) of Ne (empty circles) and minimum and maximum values (black dashes) (minmax), with harmonic means (HAR) of Ne (empty circles) and 95% confidence intervals (95%CI) and the result obtained from the intercept of the regression FIS ~ pnulls (FIS-0, “infinite upper bound not shown) and its 95%CI for the present study. Arithmetic mean and minmax of values obtained in Gstöttenmayer et al., (2023) [31] (Gsto) are presented for comparison. The ordinates are in log scale.
Time of isolation or gene flow between the two parks
With the isolation hypothesis, and with these Ne values, we computed that the subpopulations from the two parks needed to have split 86 years ago with a 95% CI = [47, 138], and with a minmax = [38, 166]. In terms of dates, this would lead to years 1936 in 95% CI = [1884, 1975], and the oldest date going back to 1856. In case of still ongoing exchanges of immigrants, we computed that this would have required an average dispersal distance of less than 1 km per generation for n = 2, and between 1 and 2 km per generation for n = infinite.
Isolation by distance, sex biased dispersal between traps and bottleneck within each park
In GNP b = -0.0013 in 95%CI = [-0.0061, 0.001], and in MNP b = -0.0009 in 95%CI = [-0.0010, -0.001], hence a total absence of isolation by distance.
We could not detect any significant genetic signature of sex-biased dispersal. Parameters provided values in different directions and the smallest p-value = 0.0566.
There was no obvious signature of a bottleneck, as p-value = 0.0365 for the IAM model and p-value > 0.25 (maximum possible with k = 2) for the TPM and SMM models.
Bayesian clustering and neighbour-joining trees
Bayesian clustering expectedly produced an optimal partition with two clusters, but with multiple miss-assignments (Fig 5). Upon further inspection, these miss-assigned individuals seemed to contain several missing genotypes. Indeed, a correlation test between the probability of belonging to the park of origin and the number of missing data in each individual fly happened to be highly significant (ρ = - 0.1804, p-value = 0.0002). We thus reran STRUCTURE keeping individuals with the fewest number of missing data. Exploring data with individuals with at least less than 40% missing genotypes still provided a significant signature (ρ = -0.1808, p-value = 0.0002). With 30% missing data, the correlation was not significant anymore (ρ = -0.0633, p-value = 0.136). We thus concluded that miss-assigned individuals with more than two missing data (i.e., more than 25% of the Multilocus genotype) rather corresponded to errors, while other misplaced individuals (with a 75% complete genotype at least), probably corresponded to individuals that inherited genes from recent immigrants or even represented recently introduced individuals (if with more than 90% genomic match). This was confirmed by the fact that even individuals with complete genotypes could be miss-assigned (S1 Fig). Individuals with no or a single missing genotype and with more than 90% assignment to the wrong park would correspond to recently introduced individuals (all in GNP). Other miss-assigned individuals would then correspond to individuals that inherited genes from more or less recent immigrants (14 in GNP and 6 in MNP). We also ran STRUCTURE with only one or no missing data allowed (S1 Fig). Miss-assigned individuals can still be met, though “recent” immigrants cannot be detected anymore with complete genotypes.
Information on the top indicate the number of missing genotypes for the individuals that were the least assigned to their park of origin, with a threshold of 0.9, and displaying less than three missing genotypes. Individuals in bold blue are more than 90% miss-assigned and assimilated to new immigrants from MNP in GNP. Numbers in black are interpreted as introgressed with alien alleles inherited from some immigrant ancestors in the past.
A Neighbour-joining approach on all individuals corrected for null alleles (S2 Fig) or on individuals with complete genotypes (S3 Fig), confirmed that some miss-assigned individuals were indeed genetically close to individuals from the other park.
Discussion
During this study, we faced many genotyping problems, with a substantial number of missing genotypes at several loci. Most missing genotypes were due to PCR failures and not to null alleles, as Gb35 was the only locus for which missing data corresponded to null homozygotes. Fortunately, this did not alter most of our results as we could demonstrate that heterozygote deficits were fully explained by null alleles. We also could compute genetic differentiation accordingly. Nevertheless, during the Bayesian clustering process, such missing data significantly altered the probability of assignment of the most affected individuals to their population. Indeed, only the dataset with individuals with less than 25% of missing genotypes (i.e., one or two missing loci) did not produce a significant effect of missing genotypes on probabilities of assignment. It means that the algorithm applied by this software used missing genotypes as informative to build the clusters.
Effective population sizes appeared substantial (Ne = 844 in minmax = [676, 1012]), though probably underestimated. These appeared much larger than previously reported (Ne = 76 in minmax = [16, 192]) [31], to what was predicted by the intercept of the regression FIS ~ pnulls or if we use Waples’ method. One hundred probably represents an extreme lower limit, while 1000 probably corresponds to the upper one.
Subdivision between parks was highly significant. With a two-subpopulation model, this subdivision appeared compatible with the number of immigrants Nem = 0.35 in 95% CI = [0.20, 0.68] exchanged per generation. Translated in years, this would mean 2 in 95% CI = [1, 4] individuals per year, exchanged between the two parks, or a total separation dating to 1936 on average with a time window between 1858 and 1966. This would be in line with both the rinderpest pandemic (1898–1901) and the last vector control campaign, which was implemented between 1948 and 1970 in the area [5,12–14,83]. Nevertheless, no signature of any bottleneck could be assessed. This does not confirm a drastic decrease in tsetse populations following any of these two events. This may be due to immigration events from other areas not included in our study, which may have brought enough exogenous alleles to wipe out any bottleneck signature. According to published data from the last 30 years [18], no G. brevipalpis subpopulations were reported that could link GNP and MNP in a step-by-step model. However, considering that very limited data are available from that region, this lack of reporting does not exclude the possibility that such pockets might exist and could be discovered in the future. Historical records identified a G. brevipalpis population pocket in Mutamba River valley near Inhambane city on the coast [84]. To this day, this area has favourable conditions for G. brevipalpis [15], but no recent reports indicate the presence of this species. A thorough survey of these historical sites needs to be conducted. Accordingly, Bayesian clustering did not support that total isolation between the two subpopulations occurred around 1946, nor that step-by-step dispersal could take place between undocumented subpopulations connecting the two parks. Indeed, the most probable interpretation of our results is the direct exchange of (very) rare immigrants, i.e., four per year on average at most. This was confirmed by using a Neighbour-joining approach, which confirmed that some miss-assigned individuals were indeed genetically close to individuals from the other park.
Effective population size and immigration could be translated into no more than 2 km distance dispersal per generation between the two parks, which appeared much smaller than the 24 km that these flies are able to travel within a park. The difference in dispersal distance within and between the parks illustrates how low survival is for tsetse flies travelling from GNP to MNP (or back). The maximum dispersal distance ever reported for a tsetse fly was 25 km in 24 days for G. tachinoides, after a mark-release-recapture study [85], and around 40 km/generation for G. palpalis gambiensis with population genetics data [86]. Consequently, it appears highly unlikely that immigrants could have crossed the 840 km separating the two parks on their own. The motorized anthropochory of these flies may be an alternative explanation for the flow of individuals between these two extremely distant populations.
We know that the recent introduction of wild animals occurred from one of these two parks to the other using trucks or airplanes that may have represented shelter, food and easy transport for G. brevipalpis [87,88]. For example, in 2019 waterbucks, warthogs, and oribi were transported to MNP from GNP and more recently, in 2021, a pack of African wild dogs were translocated from the buffer zone of MNP to GNP (M. Stalmans, personal communication, April 16, 2025). An informal inquiry at the National Directorate of Livestock Development within the Ministry of Agriculture, Environment and Fisheries provided evidence of livestock transportation (cattle and goats) between the two provinces (Maputo and Sofala) where the parks are located (J. Fabbri, personal communication, April 16, 2025). This can be related to long, though much smaller, range dispersal distances that were recently documented in other tsetse fly species [22,86,89]. Therefore, we advise considering that such a phenomenon may account for reinvasions over very long distances.
Conclusion
The most relevant result obtained during this work is that, despite the existence of two distant and supposedly isolated populations, there is strong evidence indicating the exchange of rare individuals. This could be due to undiscovered pocket populations between the two parks, if so, a detailed entomological survey of possible habitats between the two parks needs to be conducted. A more probable explanation of our findings is that tsetse flies may have been moved between parks via motorized human transport means. Such passive dispersal should thus be investigated more thoroughly in future studies. We also recommend that a similar study be carried out on G. austeni populations in Mozambique that show a similar distribution pattern to that of G. brevipalpis. Such a study might further elucidate the tsetse dynamics between the centre and south of the country and assist in the development of efficient control strategies. If the eradication of the southern tsetse belt in MNP was attempted, the possibility of re-invasion of the tsetse via motorized human transport means needs being taken in to account as well as the exploration of other sites within a 30 km radius to validate that no sources of re-invasion exist around this park.
Supporting information
S1 Table. Trapping sites with GPS coordinates, the number of captured Glossina brevipalpis and number of genotyped tsetse at Gorongosa National Park and Maputo National Park in Mozambique.
https://doi.org/10.1371/journal.pntd.0012953.s001
(XLSX)
S2 Table. Dataset of tsetse genotypes.
Three hundred and ninety-six (396) Glossina brevipalpis genotypes collected at Gorongosa National Park (196) and Maputo National Parks (200) in Mozambique.
https://doi.org/10.1371/journal.pntd.0012953.s002
(TXT)
S3 Table. The 10 polymorphic microsatellite loci used for genotyping Mozambican Glossina brevipalpis and associated genetic population parameters.
Sample size (genotypes per locus), number of alleles, allele size range, Write’s fixation index in individuals as compared to subsamples (FIS), expected heterozygosity (Hexp), Nei’s unbiased expected heterozygosity (HS), and observed heterozygosity (HO). *Loci removed from final analyses.
https://doi.org/10.1371/journal.pntd.0012953.s003
(XLSX)
S1 Fig. Structure results on Glossina brevipalpis from Gorongosa National Park (GNP) and Maputo National Park (MNP) in Mozambique for individuals with no more than one missing genotype out of eight (top) or without missing genotype (bottom).
Numbers on the top indicate the number of missing genotypes for individuals assigned to their park of origin with less than 50% probability. Average probabilities of assignment to the park of origin was 0.9234 and 0.7811 for One missing and No missing datasets, respectively.
https://doi.org/10.1371/journal.pntd.0012953.s004
(TIF)
S2 Fig. A Neighbour-Joining tree on all individuals of Glossina brevipalpis from Gorongosa National Park and Maputo National Park in Mozambique.
The tree is based on Cavalli-Sforza and Edward’s chord distance corrected for null alleles, built with MEGA X and edited with iTOL V6.
https://doi.org/10.1371/journal.pntd.0012953.s005
(PDF)
S3 Fig. Neighbour-Joining trees of Glossina brevipalpis from Gorongosa National Park (orange) and Maputo National Park (blue) in Mozambique for individuals with complete genotypes.
These trees are based on Cavalli-Sforza and Edward’s chord distance corrected for null alleles (left), or on Nei’s DA genetic distance (right). These were built with MEGA X and PopTree, respectively, and edited in iTOL V6. Probabilities in black are the assignment of corresponding individuals to their park of origin.
https://doi.org/10.1371/journal.pntd.0012953.s006
(TIF)
Acknowledgments
We would like to acknowledge the Biotechnology Centre of Eduardo Mondlane University and INTERTRYP IRD/CIRAD for using their laboratories and for their continued partnership throughout this study. We would also like to acknowledge Keila Zandamela and Edmilson Philimone for their assistance and dedication during this project. Authors would like to thank Marc Stalmans of Gorongosa National Park and José Fabbri of the National Directorate of Livestock Development (the Ministry of Agriculture, Environment and Fisheries, Mozambique) for providing information about the transport of livestock and wildlife within the country.
More details about the COMBAT project can be found in Boulangé et al. [90].
References
- 1. Magez S, Esteban J, Torres P, Oh S, Radwanska M, Bruschi F. Salivarian Trypanosomes Have Adopted Intricate Host-Pathogen Interaction Mechanisms That Ensure Survival in Plain Sight of the Adaptive Immune System. Pathogens. 2021;10.
- 2. Morrison LJ, Vezza L, Rowan T, Hope JC. Animal African Trypanosomiasis: Time to Increase Focus on Clinically Relevant Parasite and Host Species. Trends Parasitol. 2016;32(8):599–607. pmid:27167665
- 3.
Mulligan HW, Potts WH. The African trypanosomiases. London: Allen and Unwin; 1970.
- 4. Food and Agriculture Organization. The disease | Programme Against African Trypanosomosis (PAAT). 2019 [cited 2023 February 7. ]. https://www.fao.org/paat/the-programme/the-disease/en/
- 5.
Ford J. The role of the trypanosomiases in African ecology: a study of the tsetse fly problem. Oxford, UK: Oxford University Press; 1971.
- 6.
Food and Agriculture Organization, World Health Organization. Vector control and the elimination of gambiense human African trypanosomiasis (HAT) - Joint FAO/WHO Virtual Expert Meeting, 5-6 October 2021. FAO, WHO; 2022. https://doi.org/10.4060/cc0178en
- 7. Krafsur ES. Tsetse flies: genetics, evolution, and role as vectors. Infect Genet Evol. 2009;9(1):124–41. pmid:18992846
- 8. Rodrigues CM, Garcia HA, Rodrigues AC, Costa-Martins AG, Pereira CL, Pereira DL, et al. New insights from Gorongosa National Park and Niassa National Reserve of Mozambique increasing the genetic diversity of Trypanosoma vivax and Trypanosoma vivax-like in tsetse flies, wild ungulates and livestock from East Africa. Parasit Vectors. 2017;10(1):337. pmid:28716154
- 9. Sigauque I, Van den Bossche P, Moiana M, Jamal S, Neves L. The distribution of tsetse (Diptera: Glossinidae) and bovine trypanosomosis in the Matutuine District, Maputo Province, Mozambique. Onderstepoort J Vet Res. 2000;67(3):167–72. pmid:11131117
- 10. Specht EJK. Prevalence of bovine trypanosomosis in Central Mozambique from 2002 to 2005. Onderstepoort J Vet Res. 2008;75(1):73–81. pmid:18575067
- 11.
Spinage CA. African Ecology: Benchmarks and Historical Perspectives. Berlin, Heidelberg: Springer Berlin Heidelberg; 2012. https://doi.org/10.1007/978-3-642-22872-8
- 12. Dias JAT dos S. The status of the tsetse fly in Mozambique before 1896. South Afr J Sci. 1962;:243–7.
- 13. Dias JAT dos S. Aspectos epidemiológicos da tripanosomíase rhodesiense, em Moçambique. An Inst Med Trop (Lisb). 1952;9.
- 14. Hornby HE. Report on the tsetse-fly problems of Maputo. An Inst Med Trop (Lisb). 1947;4:313–471.
- 15. de Beer CJ, Dicko AH, Ntshangase J, Moyaba P, Taioe MO, Mulandane FC, et al. A distribution model for Glossina brevipalpis and Glossina austeni in Southern Mozambique, Eswatini and South Africa for enhanced area-wide integrated pest management approaches. PLoS Negl Trop Dis. 2021;15(11):e0009989. pmid:34843478
- 16. Du Toit R. Trypanosomiasis in Zululand and the control of tsetse flies by chemical means. Onderstepoort J Vet Res. 1954;26:317–87.
- 17.
Wint W. Mapping historical tsetse fly distributions in southern Africa. 2001. http://ergodd.zoo.ox.ac.uk
- 18.
Cecchi G, Paone M, de Gier J, Zhao W. The continental atlas of the distribution of tsetse flies in Africa. Rome: FAO; 2024. https://doi.org/10.4060/cd2022en
- 19. Courtin F, Camara M, Rayaisse J-B, Kagbadouno M, Dama E, Camara O, et al. Reducing Human-Tsetse Contact Significantly Enhances the Efficacy of Sleeping Sickness Active Screening Campaigns: A Promising Result in the Context of Elimination. PLoS Negl Trop Dis. 2015;9(8):e0003727. pmid:26267667
- 20. Diall O, Cecchi G, Wanda G, Argilés-Herrero R, Vreysen MJB, Cattoli G, et al. Developing a Progressive Control Pathway for African Animal Trypanosomosis. Trends in Parasitology. 2017;33(7):499–509.
- 21. Solano P, Ravel S, de Meeûs T. How can tsetse population genetics contribute to African trypanosomiasis control?. Trends Parasitol. 2010;26(5):255–63. pmid:20202905
- 22. Ravel S, Ségard A, Mollo BG, Mahamat MH, Argiles-Herrero R, Bouyer J, et al. Limited impact of vector control on the population genetic structure of Glossina fuscipes fuscipes from the sleeping sickness focus of Maro, Chad. Parasite. 2024;31:13. pmid:38450719
- 23. Ravel S, Séré M, Manangwa O, Kagbadouno M, Mahamat MH, Shereni W, et al. Developing and quality testing of microsatellite loci for four species of Glossina. Infect Genet Evol. 2020;85:104515. pmid:32861909
- 24. Ravel S, de Meeus T, Dujardin JP, Zézé DG, Gooding RH, Dusfour I, et al. The tsetse fly Glossina palpalis palpalis is composed of several genetically differentiated small populations in the sleeping sickness focus of Bonon, Côte d’Ivoire. Infect Genet Evol. 2007;7(1):116–25. pmid:16890499
- 25. Solano P, Ravel S, Bouyer J, Camara M, Kagbadouno MS, Dyer N, et al. The population structure of Glossina palpalis gambiensis from island and continental locations in Coastal Guinea. PLoS Negl Trop Dis. 2009;3(3):e392. pmid:19290038
- 26. Kappmeier K. A newly developed odour-bated “H trap” for the live collection of. Onderstepoort J Vet Res. 2000;15.
- 27.
Kuzoe FAS, Schofiled CJ. Strategic Review of Traps and Targets for Tsetse and African Trypanosomiasis Control. World Health Organization; 2004. https://apps.who.int/iris/handle/10665/69013
- 28. Kappmeier K, Nevill EM. Evaluation of conventional odour attractants for Glossina brevipalpis and Glossina austeni (Diptera: Glossinidae) in South Africa. Onderstepoort J Vet Res. 1999;66(4):307–16. pmid:10689703
- 29.
Pollock JN. Training Manual for Tsetse Control Personnel Volume 1. FAO. 1992.
- 30. Walsh PS, Metzger DA, Higuchi R. Chelex 100 as a medium for simple extraction of DNA for PCR-based typing from forensic material. Biotechniques. 1991;10(4):506–13. pmid:1867860
- 31. Gstöttenmayer F, Moyaba P, Rodriguez M, Mulandane FC, Mucache HN, Neves L, et al. Development and characterization of microsatellite markers for the tsetse species Glossina brevipalpis and preliminary population genetics analyses. Parasite. 2023;30:34. pmid:37712836
- 32.
Thermo Fisher Scientific. Applied Biosystems GeneMapper Software 6. USA: Thermo Fisher Scientific; 2020.
- 33. Coombs JA, Letcher BH, Nislow KH. create: a software to create input files from diploid genotypic data for 52 genetic software programs. Mol Ecol Resour. 2008;8(3):578–80. pmid:21585837
- 34. Wright S. The interpretation of population structure by f-statistics with special regard to systems of mating. Evolution. 1965;19(3):395–420.
- 35.
Weir BS, Cockerham CC. Estimating F-Statistics for the Analysis of Population Structure. 1984.
- 36. Goudet J, Raymond M, de Meeüs T, Rousset F. Testing differentiation in diploid populations. Genetics. 1996;144(4):1933–40. pmid:8978076
- 37. De Meeûs T, Guégan J-F, Teriokhin AT. MultiTest V.1.2, a program to binomially combine independent tests and performance comparison with other related methods on proportional data. BMC Bioinformatics. 2009;10:443. pmid:20030807
- 38.
Goudet J. FSTAT (version 2.9.4), a program (for Windows 95 and above) to estimate and test population genetics parameters. 2003.
- 39. Wahlund S. Zusammensetzung von populationen und korrelationserscheinungen vom standpunkt der vererbungslehre aus betrachtet. Hereditas. 1928;11(1):65–106.
- 40. Goudet J, De Meeus T, Day AJ, Gliddon CJ. The different levels of population structuring of the dogwhelk, Nucella lapillus, along the south Devon coast. Genet Evol Aquat Org. 1994;81–95.
- 41. Wilcoxon F. Individual Comparisons by Ranking Methods. Biometrics Bulletin. 1945;1(6):80.
- 42. Fox J. Extending the R Commander by “plug-in” packages. R News. 2007;7:46–52.
- 43. Fox J. TheRCommander: A Basic-Statistics Graphical User Interface toR. J Stat Soft. 2005;14(9).
- 44.
R Core Team. R: A language and environment for statistical computing. Vienna: R Foundation for Statistical Computing; 2024.
- 45.
Fisher RA. Statistical Methods for Research Workers. In: Kotz S, Johnson NL. Breakthroughs in Statistics: Methodology and Distribution. New York, NY: Springer; 1992. 66–70. https://doi.org/10.1007/978-1-4612-4380-9_6
- 46. Benjamini Y, Yekutieli D. The control of the false discovery rate in multiple testing under dependency. Ann Statist. 2001;29(4).
- 47.
Belkhir K, Borsa P, Chikhi L, Raufaste N, Bonhomme F. Genetix 4.05, logiciel sous Windows TM pour la génétique des populations. Université de Montpellier II; 2004.
- 48. De Meeûs T. Revisiting FIS, FST, Wahlund Effects, and Null Alleles. J Hered. 2018;109(4):446–56. pmid:29165594
- 49. Manangwa O, De Meeûs T, Grébaut P, Ségard A, Byamungu M, Ravel S. Detecting Wahlund effects together with amplification problems: Cryptic species, null alleles and short allele dominance in Glossina pallidipes populations from Tanzania. Mol Ecol Resour. 2019;19(3):757–72. pmid:30615304
- 50. Dempster AP, Laird NM, Rubin DB. Maximum Likelihood from Incomplete Data Via the EM Algorithm. Journal of the Royal Statistical Society Series B: Statistical Methodology. 1977;39(1):1–22.
- 51. Chapuis M-P, Estoup A. Microsatellite null alleles and estimation of population differentiation. Mol Biol Evol. 2007;24(3):621–31. pmid:17150975
- 52. Hedrick PW. A standardized genetic differentiation measure. Evolution. 2005;59(8):1633–8.
- 53. Meirmans PG. Using the AMOVA framework to estimate a standardized genetic differentiation measure. Evolution. 2006;60(11):2399–402. pmid:17236430
- 54. Meirmans PG, Hedrick PW. Assessing population structure: F(ST) and related measures. Mol Ecol Resour. 2011;11(1):5–18. pmid:21429096
- 55. Rousset F. Genetic differentiation and estimation of gene flow from F-statistics under isolation by distance. Genetics. 1997;145(4):1219–28. pmid:9093870
- 56. De Meeûs T, Noûs C. A new and almost perfectly accurate approximation of the eigenvalue effective population size of a dioecious population: comparisons with other estimates and detailed proofs. Peer Community Journal. 2023;3:e51.
- 57. Waples RS. A bias correction for estimates of effective population size based on linkage disequilibrium at unlinked gene loci*. Conserv Genet. 2006;7(2):167–84.
- 58. Peel D, Waples RS, Macbeth GM, Do C, Ovenden JR. Accounting for missing data in the estimation of contemporary genetic effective population size (N(e) ). Mol Ecol Resour. 2013;13(2):243–53. pmid:23280157
- 59. Nomura T. Estimation of effective number of breeders from molecular coancestry of single cohort sample. Evol Appl. 2008;1(3):462–74. pmid:25567728
- 60. Vitalis R, Couvet D. Estimation of effective population size and migration rate from one- and two-locus identity measures. Genetics. 2001;157(2):911–25. pmid:11157007
- 61. Wang J. Sibship Reconstruction From Genetic Data With Typing Errors. Genetics. 2004;166(4):1963–79.
- 62. Do C, Waples RS, Peel D, Macbeth GM, Tillett BJ, Ovenden JR. NeEstimator v2: re-implementation of software for the estimation of contemporary effective population size (Ne ) from genetic data. Mol Ecol Resour. 2014;14(1):209–14. pmid:23992227
- 63. Jones OR, Wang J. COLONY: a program for parentage and sibship inference from multilocus genotype data. Mol Ecol Resour. 2010;10(3):551–5. pmid:21565056
- 64. De Meeûs T, Noûs C. A simple procedure to detect, test for the presence of stuttering, and cure stuttered data with spreadsheet programs. Peer Community Journal. 2022;2.
- 65. Waples RS. Practical application of the linkage disequilibrium method for estimating contemporary effective population size: A review. Mol Ecol Resour. 2024;24(1):e13879. pmid:37873672
- 66. Séré M, Thévenon S, Belem AMG, De Meeûs T. Comparison of different genetic distances to test isolation by distance between populations. Heredity (Edinb). 2017;119(2):55–63. pmid:28537571
- 67.
Hijmans RJ, Karney C, Williams E, Vennes C. Package “geosphere”: Spherical Trigonomertry. 2019.
- 68. Goudet J, Perrin N, Waser P. Tests for sex-biased dispersal using bi-parentally inherited genetic markers. Mol Ecol. 2002;11(6):1103–14. pmid:12030985
- 69. Goudet J. Hierfstat, a package for R to compute and test hierarchical F‐statistics. Molecular Ecology Notes. 2004;5(1):184–6.
- 70. Cornuet JM, Luikart G. Description and power analysis of two tests for detecting recent population bottlenecks from allele frequency data. Genetics. 1996;144(4):2001–14. pmid:8978083
- 71. De Meeûs T, Chan CT, Ludwig JM, Tsao JI, Patel J, Bhagatwala J, et al. Deceptive combined effects of short allele dominance and stuttering: an example with Ixodes scapularis, the main vector of Lyme disease in the U.S.A. Peer Community Journal. 2021;1.
- 72. Teriokhin AT, de Meeûs T, Guégan JF. On the power of some binomial modifications of the Bonferroni multiple test. Zh Obshch Biol. 2007;68(5):332–40. pmid:18038646
- 73. De Meeûs T. Statistical decision from k test series with particular focus on population genetics tools: a DIY notice. Infect Genet Evol. 2014;22:91–3. pmid:24444592
- 74. Pritchard JK, Stephens M, Donnelly P. Inference of population structure using multilocus genotype data. Genetics. 2000;155(2):945–59. pmid:10835412
- 75. Evanno G, Regnaut S, Goudet J. Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Mol Ecol. 2005;14(8):2611–20. pmid:15969739
- 76. Li Y-L, Liu J-X. StructureSelector: A web-based software to select and visualize the optimal number of clusters using multiple methods. Mol Ecol Resour. 2018;18(1):176–7. pmid:28921901
- 77. Cavalli-Sforza LL, Edwards AWF. Phylogenetic analysis: models and estimation procedures. Evolution. 1967;21(3):550–70. pmid:28563688
- 78. Kumar S, Stecher G, Li M, Knyaz C, Tamura K. MEGA X: Molecular Evolutionary Genetics Analysis across Computing Platforms. Mol Biol Evol. 2018;35(6):1547–9. pmid:29722887
- 79. Takezaki N, Nei M, Tamura K. POPTREE2: Software for constructing population trees from allele frequency data and computing other population statistics with Windows interface. Mol Biol Evol. 2010;27(4):747–52. pmid:20022889
- 80. Nei M, Tajima F, Tateno Y. Accuracy of estimated phylogenetic trees from molecular data. II. Gene frequency data. J Mol Evol. 1983;19(2):153–70. pmid:6571220
- 81. Letunic I, Bork P. Interactive Tree of Life (iTOL) v6: recent updates to the phylogenetic tree display and annotation tool. Nucleic Acids Res. 2024;52(W1):W78–82. pmid:38613393
- 82. Kent WJ. BLAT--the BLAST-like alignment tool. Genome Res. 2002;12(4):656–64. pmid:11932250
- 83. Silva MA de A. Missão de combate às tripanossomíases da Província de Moçambique = Tsetse and trypanosomiasis department of the Province of Mozambique. An Inst Med Trop. 1958;15(Número Espec Dedic Aos VI Congr Med Trop E Palud):241–61.
- 84.
Missão de Combate às Tripanosomíases. Relatório Anual de 1949. Missão de Combate às Tripanossomoses; 1950.
- 85. Cuisance D, Février J, Déjardin J, Filledier J. Dispersion linéaire de Glossina palpalis gambiensis et de Glossina tachinoides dans une galerie forestière en zone soudano-guinéenne (Burkina-Faso). Rev Elev Médecine Vét Pays Trop. 1985.
- 86. Kagbadouno MS, Séré M, Ségard A, Camara AD, Camara M, Bucheton B, et al. Population genetics of Glossina palpalis gambiensis in the sleeping sickness focus of Boffa (Guinea) before and after eight years of vector control: no effect of control despite a significant decrease of human exposure to the disease. Peer Community Journal. 2024;4:e21.
- 87. Parque Nacional de Maputo. Reserva Especial de Maputo recebe animais de Gorongosa. In: Parque Nacional de Maputo. 2019 [cited 28 Jan 2025. ]. https://parquemaputo.gov.mz/reserva-especial-de-maputo/reserva-especial-de-maputo-recebe-animais-de-gorongosa/
- 88. MWA. A Mission to Save African Wild Dogs in Maputo National Park. In: MWA. 2023 [cited 16 Jan 2025. ]. https://www.mwa.co.mz/post/a-mission-to-save-african-wild-dogs-in-maputo-national-park
- 89. Ravel S, Mahamat MH, Ségard A, Argilés-Herrero R, Bouyer J, Rayaisse J-B, et al. Population genetics of Glossina fuscipes fuscipes from southern Chad. Peer Community Journal. 2023;3:e31.
- 90. Boulangé A, Lejon V, Berthier D, Thévenon S, Gimonneau G, Desquesnes M, et al. The COMBAT project: controlling and progressively minimizing the burden of vector-borne animal trypanosomosis in Africa. Open Res Eur. 2022;2:67. pmid:37645305