Biological invasions in international seaports: a case study of exotic rodents in Cotonou

Black rat (Rattus rattus), brown rat (Rattus norvegicus), and house mouse (Mus musculus) are known to be among the most common anthropophilic rodent species in cities worldwide. These species are responsible for the destruction of domestic and industrial materials, considerable damage to food stocks as well as zoonotic pathogens circulation and transmission to humans and animals. These invasive species have disseminated in all continents following human-mediated exchanges, especially maritime transports. In particular, seaports appear as privileged rats and mice’s entry points into new regions, thus making them international regulations’ priorities for rodent surveillance and management. Yet, studies on seaport rodents are rare; in particular, investigations on their genetic structure are almost inexistent, thus precluding science-guided interventions. In order to fill such a gap, our study focused on the population genetics of R. rattus, R. norvegicus and M. musculus in the Autonomous Port of Cotonou, Benin. Nine different sites were surveyed for three years. In total, 366 R. rattus, 188 R. norvegicus and 244 M. musculus were genotyped using 18 microsatellites, 16 microsatellites and 17 microsatellites, respectively. Our results show very well-structured genetic clusters in all three species as well as limited impacts of rodent control campaigns. Using comparisons with genotypes from other European, Asian and African countries, we suggest for the first time that settlement of newly introduced individuals may be a rare event. Implications in terms of management units and control and monitoring are discussed.


Introduction
Human activities contribute strongly to the displacement of species outside their original distribution area, thus leading to an increased number of biological invasions (Wilson et al. 2009;Blackburn et al. 2011;Gallardo and Aldridge 2013;Padayachee et al. 2017;Seebens et al. 2017).In particular, maritime trade plays a major role in the global spread of exotic plant and animal species around the world (Veale et al. 2021).Among them, the transport and involuntary introduction of commensal rodents and their associated ectoparasites and pathogens are particularly worrying (Aplin et al. 2011;Song et al. 2003Song et al. , 2014;;Bona 2020).The black rat Rattus rattus, the Norway rat Rattus norvegicus and the domestic mouse Mus musculus fall within the 100 most impactful invasive alien species (Lowe et al. 2000;www.iucnr edlist.org), ranking within the top 15 rodent pest taxa in the world (Capizzi et al. 2014).These species have invaded all the continents except Antarctica (Long 2003).
The intensification of trade has accelerated their spread to new territories with invasion modalities that remain largely uncontrolled (e.g., Dalecky et al. 2015;Berthier et al. 2016).Consequently, they have now reached a quasi-global distribution (St Clair 2011).All three species impact local biodiversity, leading to the decrease or extinction of endemic species (Wyatt et al. 2009;Harper and Bunbury 2015;Doherty et al. 2016).They cause damage to infrastructure and human activities (Garba et al. 2014b;Panti-May et al. 2017) and huge economic losses, essentially related to the destruction of food stocks (Dossou et al. 2020;Diagne et al. 2020aDiagne et al. , b, 2021)).They are also involved in the transport, maintenance, and transmission of many zoonotic agents (review in Meerburg et al. 2009;Colombe et al. 2019).Their tigh interaction with humans has made them privileged sources of infectious diseases, thus inducing the loss of hundreds thousands of lives annually (Colombe et al. 2019).In order to avoid rodent-mediated dissemination of zoonotic agents showing epidemic/pandemic potential, International Health Regulation (World Health Organization 2008) imposes signatory states to monitor and control invasive rodents at their entry points, especially within seaport areas.However, scientific data to guide rodent management actions in such habitats are very scarce.
Thanks to its international seaport in the city of Cotonou (officially "Autonomous Port of Cotonou", hereafter designed as to APC), the Republic of Benin is at the heart of intercontinental trade between West African coastal as well as hinterland countries and the rest of the world.As a consequence, the APC constitutes a particularly favourable site for exotic rodents introduction and subsequent spread towards landlocked countries such as Niger, Burkina-Faso, Mali and Chad.For instance, trade flows from Benin to Niger are thought to have facilitated the introduction of the black rat and the house mouse in large cities such as Niamey, which is more than 1,000 km away from the coast (Berthier et al. 2016;Hima et al. 2019).This latter species was probably brought in by ships, settled in the seaport and then transported to Niamey with the goods loaded on the trucks (Hima et al. 2019).
Rattus rattus, Rattus norvegicus and Mus musculus are abundant within the APC, representing more than 80% of the small mammals caught (Hima et al. 2019;Dossou et al. 2020;Badou et al. submitted).In the industrial storehouses, Norway rats and house mice damage an average of 3,450 tons of imported rice, for an estimated loss of 58 k€ per warehouse and per year (Dossou et al. 2020).They may also be responsible for the destruction of the electricity networks, including those of some very expensive unloading machines (APC, pers. comm.).Additionally, they may be involved in the introduction of allochthonous zoonotic pathogens, such as Seoul Orthohantavirus that are responsible for hemorrhagic fevers (Castel et al. 2021).However, whether introduction events are frequent or not remains unknown, thus greatly limiting our ability to evaluate future risks.
In order to deal with these sanitary, food security, and economic issues, APC authorities periodically hire service providers which organize anticoagulant-based rodent control campaigns within the seaport.However, a threeyear long survey of invasive rodent abundance in the APC suggests that deratization is rather ineffective in the shortterm, with similar population levels being observed within six months (Badou et al. submitted).Yet, the processes at work after rodent control operations that enable population recovery (i.e., endogenous reproduction vs. immigration) remain unknown, thus obscuring the best options for coping with rodent control through efficient, sustainable, and cost-effective methods.
By allowing indirect estimation of rodent origin and dispersal at the spatial and temporal scales that are relevant for surveillance and control, population genetics may help to address these issues.In this study, we describe the population genetic structures of R. rattus, R. norvegicus and M. musculus within the APC using species-specific sets of microsatellite markers, and we evaluate gene flows at a very fine spatial scale.Using longitudinal sampling, we investigate the effect of rodent control campaigns on the population genetic estimates from the three pest species.Finally, we rely on samples from foreign countries to mimic putative introduction events and examine whether our microsatellite panels allow us to identify newly disembarked individuals.

Study area and sample collection
Between 2017 and 2020, rodents were sampled each semester at 9 different sites of the Cotonou seaport using a standardized protocol (Garba et al. 2014a;Badou et al. submitted).Two sites were located within the artisanal fishing port outside the industrial seaport: APC1 (fresh fish stores) and APC2 (fresh fish market).Within the industrial seaport, trapping was carried out in restaurants and canteens for local workers (APC3), in the Beninese Society of Port Manutention (SOBEMAP) warehouses number 1 (APC4), 2 (APC5), 3 (APC6) and 4 (APC7), in the garage and mechanical workshops of SOBEMAP (APC8) as well as in the parking lot of SOBEMAP heavy (including offduty) gears (APC9) (Fig. 1).Details on each trapping sites are provided in Supplementary material 1.
At each session and in each site, locally-made wire mesh traps (LxWxH 30 × 10x10cm) and manufactured Sherman traps (Sherman Inc, USA; 23.2 × 7.3x8.8cm) were systematically used in conjunction so as to maximize the chances of capturing both small and large species (Garba et al. 2014a;Dalecky et al. 2015;Lucaccioni et al. 2016;Badou et al. submitted).A mixture of peanut butter and sardines in oil was used as bait.The traps were set for 3-4 consecutive nights and were collected each morning.Rodents captured were brought to the laboratory and treated within the same day.The exact GPS coordinates of each capture were systematically recorded.In total, six sampling sessions were conducted: session 1: September-October 2017; session 2: March-April 2018; session 3: September-October 2018; session 4: March-April 2019; session 5: September-October 2019; session 6: March-April 2020.Importantly, two anticoagulant-based rodent control campaigns were conducted in APC4 to 8 before our trapping sessions 3 and 4, respectively.

Ethical aspect and data sharing
This program was conducted within the framework of the research agreement between the Republic of Benin and the French National Institute for Sustainable Development (IRD) that was reapproved on the 6th April 2017, as well as the partnership agreement between IRD and the University of Abomey-Calavi (signed on the 30th September 2010 and renewed on the 3rd July 2019).All trapping sessions were authorized by the APC administrative authorities and no trap was set without the explicit oral consent of port staff or artisanal seaport site owners/ operators.
Rodents were captured and brought alive to the lab where they were treated in a respectful manner in accordance with the guidelines of the American Society of Mammalogists (Sikes and Gannon 2011), sedated, and then sacrificed by cervical dislocation as recommended by Mills et al. (1995).Access to and benefit-sharing of genetic resources produced during the course of the present study was authorized by the Benin national authorities following the Nagoya international protocol (permit 608/DGEFC/DCPRN/PF-APA/SA).Samples and associated data were deposited in the Small

Spatial patterns of population genetic structure
Analyses were performed for each species separately.They were done first on sessions pooled for each site.Unless specified, only samples from sites with at least 11 individuals for a given species were used.Deviation from Hardy-Weinberg equilibrium (HWE) for each site and each locus as well as linkage disequilibrium (LD) between each pair of loci at each site were tested using GENEPOP v.4.6 (Rousset 2008).Corrections for multiple tests were performed using the false discovery rate (FDR) approach (Benjamini and Hochberg 1995) implemented in the Bioconductor qvalue R package (Storey 2002).
FreeNa (Chapuis and Estoup 2007) was used to estimate the presence of null alleles in each dataset in order to test whether they may explain heterozygote deficiencies (see below).
The allelic richness (a r ) was calculated at each site using the rarefaction procedure in FSTAT v2.9.4 (Goudet 2003).The observed (H O ) and expected (H E ) heterozygosities were estimated using GeneClass2 (Piry et al. 2004) and the F IS (Weir and Cockerham 1984) using Genepop.
Genetic differentiation between sites was estimated by pairwise F ST estimates (Weir and Cockerham 1984) using FSTAT.A 95% confidence interval (CI) for mean F ST was generated by bootstrap resampling (10,000 permutations) across loci.
Under a model of isolation by distance (IBD), genetic distance between individuals is expected to increase with geographical distance (Rousset 1997).IBD was estimated using the regression method based on the expected linear relationship between genetic and geographic distances (Rousset 1997(Rousset , 2000) ) in GENEPOP.Mantel tests (10,000 permutations) were performed using the pairwise genetic differentiation estimator ê (Watts et al. 2007) calculated between all pairs of individuals genotyped and the logarithm of their Euclidean geographical distances.The minimum distance between sites (d = 100 m) was used as a threshold to exclude pairs of individuals from the same site.
Population genetic structure was then explored using the clustering approach implemented in STRU CTU RE V.2.3.4 (Pritchard et al. 2000) using all individuals genotyped in order to determine the number of homogeneous genetic clusters (K) present in each species-specific dataset.The analyses were performed with a model including admixture and correlated allele frequencies, for K ranging from 1 to 10.Each run included a burn-in phase of 200,000 iterations followed by 600,000 iterations.We performed 10 independent analyses for each K value.The number of genetic groups was inferred by the delta-K method applied to the log probabilities of data (Evanno et al. 2005).For R. rattus, fifteen individuals from four districts (Agla, N = 4; Ladji, N = 4, Saint Jean, N = 5 and Zongo, N = 2) of the city of Cotonou were also included in our analyses (Dossou et al. 2022).
Population genetic structure was also investigated through Discriminant Principal Component Analysis (DAPC) which can handle the absence of HW equilibrium (Jombart et al. 2010).Analyses were conducted using all individuals under the R software using the adegenet and devtools packages, for K = 1 to 10.The most likely number of genetic groups (K) was determined using the Bayesian Information Criterion (BIC; Lebarbier and Mary-Huard 2006), and the a priori criterion of a delta-BIC less than six (Kass and Raftery 1995).

Changes in population genetic structure through time
Genotype data for each species were then separated into three temporal datasets, regrouping sessions "before" (sessions 1 and 2), "during" (sessions 3 and 4) and "after" rodent control campaigns (sessions 5 and 6), respectively.Using STRU CTU RE, we investigated whether genetic clustering in "before" and "after" datasets was similar to that observed in the whole dataset for each species.We then used NeEstimator v2 (Do et al. 2014) to estimate effective population size (N E ) for each site or group of sites corresponding to a genetic cluster in the industrial seaport before, during (when sufficient sampling) and after control campaigns.These analyses were conducted in both deratised and non-deratised sites, using the linkage disequilibrium (LD) method with a threshold minimum allele frequency of 0.02 (i.e.Pcrit parameter) (Waples and Do 2008).Sites or groups of sites with less than 10 genotyped individuals were excluded from the analyses.Signals for bottleneck effects were looked for using the Stepwise Mutation Model (SMM) on BOTTLENECK V1.2 (Piry et al. 1999) in each species-specific dataset.

Detection of potentially newly introduced individuals
For each species, individuals from three other countries (one individual/country) were genotyped and added to the APC dataset.For Rattus rattus, additional individuals were from Dakar (Senegal), Franceville (Gabon) and Niamey (Niger).The fifteen individuals from Cotonou core city were also included in the analyses.For Rattus norvegicus, the additional individuals were from Lyon (France), Bamako (Mali) and Ratchaburi (Thailand).For Mus musculus, extra individuals were from Dakar (Senegal), Lyon (France) and Niamey (Niger).Factorial correspondence analyses were performed using Genetix v 4.05 (Belkhir et al. 1996(Belkhir et al. -2004) ) in order to evaluate if the additional individuals could be clearly identified as different from those trapped within the APC.

Rodent trapping success
Across the six sampling sessions, a total of 803 rodents (367 Rattus rattus, 189 Rattus norvegicus and 247 Mus musculus) were captured in the nine APC sites out of a total of 6,569 trap nights (Table 1 and Fig. 1).Rattus rattus was mostly trapped in the artisanal seaport (sites APC1 and APC2) and, to a lesser extent, in the industrial seaport (especially in APC3, then in APC8 and APC9, and more marginally in other sites; Table 1).Rattus norvegicus and M. musculus were seldom in the artisanal seaport (APC1 and 2) as well as in some sites of the industrial seaport (APC 3 and 9).However, these two species were largely dominant in SOBEMAP warehouses (APC4, APC5, APC6 and APC7) and garage (ACP8; Table 1).Trap success was lower in sessions 3 and 4 (8.43% and 5.78%), which were both concomitant to two rodent control campaigns (Table 1).Detailed analyses of spatio-temporal variations of species-specific abundances as well as of species assemblages are available in Badou et al. (submitted).
For R. norvegicus, linkage disequilibrium was significant for 9 out of the 600 tests performed (1.5%), and 5 of the significant values implicated the allele pair D1VKORC1-C and D1VKORC1-A, which were thus considered as linked.After having removed D1VKORC1-C locus, only 3 out of the 525 pairs of loci remained significantly linked (0.57%).The remaining loci were thus considered to be genetically independent, and were kept for subsequent analyses.For R. rattus and M. musculus, 21.13% (194 of 918) and 22.5% (153/680) of the LD tests were found significant (q-value < 0.05), respectively.Significant values concerned nearly all pairs of loci and all sites in both species.All loci were thus kept for subsequent analyses in these two species.
All loci were found at HW equilibrium for R. rattus, except D18R75.For R. norvegicus, all loci were found at HW equilibrium except D8Rat162.For M. musculus, seven loci (D2Mit456, D3Mit456, D4Mit241, D5Mit30, D6Mit373, D7Mit176 and D18Mit8) were not at HW equilibrium, exhibiting F IS values ranging (between -0.09 and 0.56).Mean null allele frequencies estimated with FreeNA for loci that were not at HW equilibrium in the three species were sufficiently low not to affect significantly estimates of genetic diversity and differentiation (Chapuis and Estoup 2007;Chapuis et al. 2008), except for D5Mit30 in M. musculus (mean null allele frequency = 0.24; F IS = 0.56), which was then excluded from the microsatellite panel.
Genetic analyses were thus conducted using eighteen microsatellites for R. rattus, fifteen microsatellites for R. norvegicus and sixteen microsatellites for M. musculus.
Genetic diversity estimates varied among sites for the three rodent species (Table 2).In R. rattus, mean values were of 4.62 ± 1.6 for allelic richness, of 0.60 ± 0.2 for expected heterozygosity and of 0.05 ± 0.02 for F IS .In R. norvegicus, mean values of genetic diversity estimates were quite similar to those of the former species, being of 4.02 ± 1.1 for allelic richness, of 0.58 ± 0.16 for expected heterozygosity, and of 0.04 ± 0.01 for F IS .In M. musculus, mean values were of 4.95 ± 1.2 for allelic richness, of 0.64 ± 0.14 for expected heterozygosity, and of 0.04 ± 0.02 for F IS .
In R. rattus, STRU CTU RE clearly identified three genetic groups (Fig. 2A): the highest delta-K value was for K = 3 (delta-K = 1,405.69),and was considerably higher than for K = 4 (delta-K = 16.96).In the DAPC analysis, two genetic groups were identified: the highest delta-BIC value was 33.21 for K = 2.However, the genetic groups observed at K = 3 were very similar to those obtained with STRU CTU RE, with APC1 and APC2 being distinct genetic groups, while a third group included the individuals from the industrial seaport (APC3-9).When STRU CTU RE is performed with the individuals from the core city of Cotonou, they also identified three genetic groups, with individuals from the city clustering with those from the industrial seaport.
In R. norvegicus, STRU CTU RE clearly identified two genetic groups (Fig. 2B): the highest delta-K was for K = 2 (delta-K = 681.08),and was considerably higher than for K = 3 (delta-K = 31.29).The DAPC highest delta-BIC value (9.06) was obtained for K = 2, and patterns were very similar to those obtained with STRU CTU RE: a first cluster gathered the individuals from the APC3, APC6, APC7 and APC8 sites, while the individuals from APC4 were assigned to a second cluster.The individuals from APC5 had a variable mixed inferred ancestry.
In M. musculus, STRU CTU RE clearly identified three genetic groups (Fig. 2C): the highest delta-K value was for K = 3 (delta-K = 236.93),and was considerably higher than for K = 4 (delta-K = 168.85).In the DAPC analysis, the highest delta-BIC value (12.01) was obtained for K = 2, while the delta-BIC value for K = 3 was 8.68.However, patterns for K = 3 were very similar to those obtained with STRU CTU RE, thus pointing towards three genetic clusters: individuals from APC4 and APC5 were mainly assigned to a first genetic pool; individuals from APC6 and APC7 formed a second group; and individuals from APC8 constituted a third group on their own.

Changes through time
Effective population sizes (N E ) were estimated for sites or groups of sites before, during and after rodent control campaigns in the industrial seaport for all three species, as well as for sites of the artisanal seaport at the same period of time for R. rattus.
In R. rattus, N E increased strongly from 11.3 (sessions 1 and 2, before control) to 34.2 (session 5 and 6, after control) in the group formed by APC4-8 sites where deratization occurred.Only two individuals were captured in these sites during sessions 3 and 4 which fell during control operations.N E values increased slightly in the same period for sites that were non-deratized (APC1: from 11.6 to 19.5, and ACP2: from 5.8 to 12.7).
Effective population size also rapidly increased after deratting campaigns in some genetic clusters for both R. norvegicus and M. musculus.Indeed, in the R. norvegicus genetic cluster from APC4-5 sites, N E increased from 19.5 (sessions 1 and 2) to 35.1 (sessions 5 and 6).In the same manner, in the M. musculus genetic cluster from APC6-7, effective population size increased from 21.2 (sessions 1 and 2) to 35.6 (sessions 5 and 6).In contrast, N E remained globally stable in APC6-8 (43.5 to 43) in R. norvegicus and in APC4-5 (8.8 to 7.4) in M. musculus.However, in APC8 N E has decreased from 24.7 to 8.6.
Significant signals for bottleneck effects were detected in nearly all sites for R. rattus (p < 0.01), except in APC2 (p = 0.06).In R. norvegicus a significant signal was found in only one site (APC6: p = 0.02).In M. musculus, significant signals were found in APC5 (p = 0.03), APC6

Detecting potential new migrants
In Rattus rattus, individuals sampled in APC were grouped with those sampled in the cities of Cotonou and Niamey, while this cluster was clearly distinct from both Franceville (Gabon) and Dakar (Senegal) individuals (Fig. 3a).In R. norvegicus, individuals sampled in APC formed a well-defined group from which the three individuals from Mali, France, and Thailand were very distinctly separated (Fig. 3b).In the same manner, M. musculus individuals from APC formed a group that was also clearly differentiated from mice from France, Dakar and Niamey, although the two individuals from Niamey did not group together (Fig. 3c).

Discussion
Human movements and migrations have played a leading role in the global spread of domestic rodents (Roberts 1991;Kovács 2012;Puckett et al. 2020).In particular, commercial exchanges (imports and/or exports) through maritime trade largely contribute to the long-distance transport and introduction of invasive rodent species in seaports around the world (Veale et al. 2021), from which they colonise entire continents.Among these species, R. rattus, R. norvegicus and M. musculus are currently spreading in Africa (Konečný et al. 2013;Dalecky et al. 2015;Berthier et al. 2016;Hima et al. 2019).Here, we investigated and compared for the first time the population genetic structures of these three species that coexist in an international seaport.Rodent sampling in 9 different sites of the port first indicated spatial segregation of the species: R. rattus was found in much higher numbers in the artisanal seaport, where the two other species did not occur, or occur only on very rare occasions.Rattus norvegicus and M. musculus co-occurred in the industrial seaport and dominated largely the trapping results in the warehouses located along the unloading docks (APC4-7).The spatial segregation observed between these three species in APC could be due to habitat preferences, interspecific competition and/or historical factors that are beyond the scope of the present paper, but discussed in Badou et al. (submitted).
Population genetic studies using microsatellite markers of commensal rodents in urban contexts were often conducted at larger spatial scales than that considered in our study.For instance, the rare studies conducted on R. rattus all focused on West African cities with sampling covering several different districts (Niamey, Niger ).The only study on urban M. musculus we are aware of was conducted in Dakar, Senegal (Stragier et al. 2022), and again was conducted at the scale of the city.To our knowledge, the only genetic study available on rodent populations from a seaport area and its surroundings was conducted in Paranagua, Brazil on 71 R.  2022).Differences in spatial scales and genetic markers used in all these studies make meaningful comparisons difficult.However, it is noteworthy that genetic diversity estimates observed for all three species within the APC were of the same order of magnitude as those obtained in other studies.This suggests similar introduction histories throughout the world (Gardner-Santana et al. 2009;Desvars-Larrive et al. 2019;Gatto-Almeida et al. 2022).
Our results suggest that new introductions of rodents are rare in the APC.Indeed, no outlier individual was detected by the FCA analysis in rodents sampled within the APC during our 3 year-long monitoring study.Also, allele numbers and allelic richness in R. rattus from Cotonou (a r = 4.41; Badou et al. 2021) and in the APC (a r = 4.62; this study) are quite similar, whereas one would expect these estimates to be higher in the seaport if individuals had been regularly (and, especially, recently) disembarked off moored ships and then persisted within the local black rat population.We were surprised by such patterns since the absence of signal reflecting new introduction events suggest that rats and mice do not descend frequently from the boats.Yet, considering the quasi-global distribution of these three species in the world, it is unlikely that this does not occur on a rather frequent basis.Moreover, we observed that anti-rat discs along mooring cables, though mandatory by IHR, were poorly and/or badly used by ships docking at APC: out of 13 vessels and 119 cables inspected, only 79 discs (69%) were indeed installed, many of which being badly installed; in total, no ship was perfectly protected with all the cables with well-laid discs (personal observations).Finally, the recent detection of Seoul virus in R. norvegicus captured in the APC in 2018 (Castel et al. 2021) suggested that the introduction of some migrant rats sometimes occurs.Those individuals could however have just transmitted their pathogens before dying, being unable to establish themselves and reproduce with resident rodents.Indeed, the non-detection of new introductions in genetic data may be explained by the presence of small mammal communities already onsite, which could render the settlement of newly immigrating individuals or species difficult.This hypothesis is supported by eco-ethological studies on commensal rodents, showing that introduced individuals may be rapidly identified as intruders by the residents, and subsequently rejected (e.g., in R. rattus : Ewer 1971;Granjon and Cheylan 1989;Barnett 1958; in R. norvegicus and M. musculus Berdoy and Drickamer 2007).If this is indeed the case, it is important to note that deratisation campaigns leading to the Although data are lacking to conclude definitely, it may be an additional argument in favour of a rigorous prevention strategy against rodent introduction through ships, rather than rodent control only.
Within APC, population genetic analyses showed very marked spatial patterns for the three species.This of course has important consequences in terms of rodent control since the genetic clusters observed should serve as a basis to guide deratization campaigns through the adequate definition of eradication units (Richardson et al. 2019).In R. norvegicus and M. musculus, two and three genetic clusters were observed within the industrial seaport, respectively.In R. rattus, two different groups were observed within the artisanal seaport, but only one within the industrial seaport.This pattern for the last species strongly contrasts with that observed using the same microsatellite markers at the whole Cotonou city scale, where only two genetic clusters associated with lower mean F ST (0.107) were detected in black rats (Badou et al. 2021).
Genetic structure may reflect genetic isolation by distance, which is significant in all species.Indeed, invasive pests that occupy a broad range of urban conditions (primarily constrained by access to water, food/rubbish and nesting/ burrowing sites) might be expected to exhibit spatial genetic patterns driven only by isolation by distance, thus reflecting spatial limitation of dispersal (Combs et al. 2018b).As such, the gradual pattern of differentiation observed between both sides of the industrial seaport in R. norvegicus (Fig. 2B) could suggest IBD at the scale of the port.IBD may also result from social behaviour of rodent species.Indeed, R. norvegicus (Combs et al. 2018a, b;Gardner-Santana et al. 2009) and M. musculus (Lippens et al. 2017) are known for strong social structure and very limited active dispersal.In M. musculus in particular, home ranges of a few tens of metres in commensal habitats (Pocock et al. 2005), and effective dispersal of only few hundred meters (Lippens et al. 2017) may explain the high F ST values observed in this species (within the industrial seaport: mean F ST = 0.09 for M. musculus; 0.06 for R. norvegicus; 0.02 for R. rattus).High genetic structure levels such as those observed within the APC in R. rattus (F ST > 0.16) and M. musculus (F ST > 0.13) between genetic groups, may also be explained by other factors.Indeed, they may reflect the presence of physical barriers to individual dispersal.Such barriers within the urban landscape have already been evidenced for Norway rats, such as major waterways in Baltimore (Gardner-Santana et al. 2009), Salvador (Kajdacsi et al. 2013) and New Orleans (Combs et al. 2018b), roadways in Salvador and Vancouver, or resource deserts in New York City (Combs et al. 2018b).Within the APC, however, we could not identify obvious physical barriers that could explain the observed marked genetic partitions.Finally, genetic structure may result from variations in effective population size, i.e. demographic barrier (Piry et al. 2004;Berthier et al. 2016;Richardson et al. 2017;Stragier et al. 2019).Such a demographic process may be especially relevant for commensal rodents whose natural dispersal distances (i.e.not human-mediated) are generally short ranged (< 1 km) and mainly prompted by the lack of feeding and harborage sites (Byers et al. 2019;Pocock et al. 2005).Control measures could also be another factor generating important gaps in densities and thus genetic discontinuities, such as for instance between the deratised industrial port (which may have been recolonized from the city) and the non-deratised artisanal port in R. rattus.
Some genetic signatures may suggest an effect of control measures on population structures.For instance, high LD values and significant tests for bottleneck signals were observed in both R. rattus and M. musculus.Increased linkage disequilibrium can indeed reflect bottleneck events (Slatkin 2008).However, similar results (i.e., LD disequilibrium and significant tests for bottleneck) were also observed for R. rattus in the non-deratized APC1 site, thus suggesting that other processes are at work.Also, high LD values may reflect the recent mixing of individuals from several subpopulations that have different allele frequencies (Slatkin 2008), which may also result from rodent control events.Rodent reinfestation after control operations may occur either by reproduction of non-poisoned or poison-resistant individuals, or by immigration of new individuals from neighbouring sites (Richardson et al. 2019).
Whatever the invasive rodent species considered here, our FCA analyses demonstrated that microsatellite markers makes it possible to detect recently introduced, hence expectedly genetically distant individuals.Indeed, for all three species, all individuals collected within the Autonomous Port of Cotonou over the three years of our study appeared much more similar to each other than to "allochthonous" ones.The only exception was APC and some Niamey black rat, which cluster together; but this supports the hypothesis that Niamey black rat originated from Benin seaport (Hima et al. 2019).These results thus highlight the interest of our panel of microsatellite loci as a valuable tool to trace back the geographic origin of house mice potentially introduced at APC.This should be useful to help seaport authorities to survey new rodent introductions in their facilities, thus allowing them to operationalize the IHR in terms of surveillance of maritime trade-associated rodent reservoirs introduction at entry points (World Health Organization, 2006).In the particular case of APC, this seems particularly realistic because of the recent implementation of a lab following an academic / seaport partnership initiative, and essentially dedicated to invasive terrestrial and marine species monitoring and management within Cotonou seaport ("Portuary Platform for Environmental Surveillance"; see https:// view.genial.ly/ 6059b b3f64 e78f0 d9cb3 2a7f).
With the high increasing rate of invasions (Pimentel et al. 2005;Seebens et al. 2017), there is a serious need for the development of methods for invasive species control (Brockerhoff et al. 2010;Glen et al. 2013;Thresher et al. 2014).Our trapping data suggested that populations of all three species rapidly recovered in terms of number of individuals (i.e., six and twelve months after control).In R. rattus, population genetic analyses integrating fifteen individuals from the core city showed that they grouped within the same genetic cluster than those from the industrial seaport.This pattern confirms the limited effect of the control measures deployed in the industrial seaport, at least on black rats which, otherwise, should result in higher genetic differentiation of the local populations (Gatto-Almeida et al. 2022).This adds to our previous results here above that poisoning campaigns alone (and as they stand) are insufficient to reach long-term decrease of invasive rodent abundances within the seaport area, and that accompanying measures are required.The latter could consist of environmental modifications towards less rodent-favoring habitats (e.g., rearrangement and physical protection of food stocks inside warehouses, rat-proofing of buildings, fueling of cracks and holes, setup of rat-proof grids along sewer networks, favoring of cats within the seaport, etc.).This would also reduce the dependency on toxicons such as anticoagulant rodenticides that may be dangerous for non-target species as well as to humans (Salas et al. 2000;Hendges et al. 2019;Kasiotis et al. 2021), and favor resistance evolution among target rodents (Hodroge et al. 2011;Buckle 2013).Such more suitable and long-term adapted measures have already been suggested in other settings (e.g., sewer networks: Channon et al. 2006;United States: Witmer and Shiels 2017;Brazilian slum: Hacker et al. 2016;Richardson et al. 2019).Unfortunately, we are not aware of such environmentally-based rodent management strategies in seaport habitats while their implementation in this type of close areas sounds quite feasible.Our study highlights the importance of monitoring of what descends from the ships to detect the eventual introduction of new invasive species (i.e.R. tanezumi; Bastos et al. 2011;Kosoy et al. 2015;Ringani et al. 2022) and the pathogens they could host.On the other hand, to monitor those that are carried on trucks to avoid propagations beyond the APC or even in the hinterland countries.

Fig. 1
Fig. 1 Trapping sites in the Autonomous Port of Cotonou (APC) on Google Earth©.Deratised and non-deratised sites are indicated in green and yellow, respectively

Table 1
Sampling of invasive rodents in nine sites of the Autonomous Port of Cotonou (APC) over six quaterly monitoring sessions.For each site and session, raw number of captures as well as an estimate of the trapping success (Ts = 100 *Nm / NTN, Nm being the total number of small mammals captured and NTN the total number of trap-nights) are provided.R. ra: Rattus rattus, : Berthier et al. 2016; Franceville, Gabon: Mangombi et al. 2016; Cotonou, Benin: Badou et al. 2021).Urban R. norvegicus genetics was investigated in a few cities in North and South America (i.e., Baltimore, USA: Gardner-Santana et al. 2009; Salvador, Brazil: Kajdacsi et al. 2013; New York City, USA: Combs et al. 2018a; Salvador, Brazil; New Orleans, USA; Vancouver, Canada; and New York City, USA: Combs et al. 2018b

Fig. 2
Fig. 2 Spatial genetic structure in the Autonomous Port of Cotonou using STRU CTU RE.A Results for R. rattus for K = 3; B R. norvegicus for K = 2; C M. musculus for K = 3.Each vertical line represents an individual and barplots indicate individual ancestry estimates.D Report on the APC map of species-specific genetic clusters

Fig. 3
Fig. 3 FCA analyses conducted on microsatellite data for a R. rattus from the APC and the cities of Cotonou, Franceville (Gabon), Dakar (Senegal) and Niamey (Niger); b R. norvegicus from APC and the

Table 3
Pairwise F ST values among sampled sites calculated for all loci in each species, using 18 microsatellites for Rattus rattus (A), 15 microsatellites for Rattus norvegicus (B), and 16 microsatellites for Mus musculus (C) (p = 0.007) and APC8 (p = 0.02), but not in other sites (APC4 and APC7).