Dispersal patterns significantly influence several population processes, including dynamics, spatial synchrony, and population structure (Gaines & McClenaghan 1980; Stenseth & Lidicker 1992; Andreassen et al. 2002). Both, the landscape structure and the innate dispersal capability of a species influence how the species disperse (Manel et al. 2003; Storfer et al. 2007; Manel et al. 2013; Balkenhol et al. 2017; Fletcher & Fortin 2018). Higher dispersal rates enhance population cohesiveness, thereby preventing isolation and inbreeding. Changes in dispersal rates may be attributed to factors such as resource competition and social interactions (Barnett 1963; Stenseth & Lidicker 1992; Andreassen et al. 2002). Therefore, understanding dispersal patterns, isolation, and gene flow is crucial for comprehending the spatio-temporal dynamics that shape animal population structure (Gaines & McClenaghan 1980; Stenseth & Lidicker 1992; Manel et al. 2003).

In the context of managing pest species, identifying the factors that influence dispersal and, consequently, population connectivity and gene flow can help us to understand how species are spatially distributed in terms of abundance and genetic diversity (Singleton et al. 1999; Mazzi & Dorn 2012). This also provides insights into the ability of pest species to re-infest areas after eradication. The knowledge of the ecology of a pest species is essential for designing effective prevention and control programs (Singleton et al. 1999; Piertney et al. 2016). For example, identifying landscape features that impede (i.e. barriers) or facilitate (i.e. corridors) the dispersal of a given species is necessary for understanding the temporal and spatial scale of different ecological processes, as well as for testing species-specific ecological hypotheses (Storfer et al. 2007; Balkenhol et al. 2017; Fletcher & Fortin 2018). Corridors are expected to enhance connectivity between suitable patches in fragmented landscapes, thereby mitigating the adverse genetic effects of isolation (Ramiadantsoa et al. 2015).

Despite the importance of dispersal, conventional ecological methods, such as capture-mark-recapture sampling, radio-telemetry, or spool and line studies, pose challenges. These methods are expensive and labor-intensive, due to efforts such as field marking and recapturing, and they may lack precision in dispersal studies (Spear et al. 2010; Hidalgo Mihart & Olivera Gómez 2011). Moreover, as dispersal events are frequently rare and may only occur once in an individual's lifetime, detecting them with these techniques can be particularly challenging (Heiberg et al. 2012; Whitmee & Orne 2012). For these reasons, advancements in population genomics have introduced innovative tools to population genetics research in the last decade. These advances have expanded our understanding of population structure and genetic diversity and have provided new ways to investigate species dispersal patterns (Holliday et al. 2018; Luikart et al. 2018; Maffey et al. 2022; Byers et al. 2021).

Population genomics is a field that aims to reveal the structure of populations and dispersal movements by examining gene flow and their spatial genetic variation (Balkenhol et al. 2017; Fletcher & Fortin 2018; Holliday et al. 2018; Luikart et al. 2018). The advent of Next Generation Sequencing (NGS) techniques, such as RAD-Seq (Restriction-site Associated DNA sequencing; Baird et al. 2008), enable the genotyping of thousands of Single Nucleotide Polymorphisms (SNPs) in organisms collected across a given landscape. This facilitates the investigation of the spatial distribution of genetic variation, enabling the inference of population processes and dispersal of a species in the recent past.

The Norway rat (Rattus norvegicus Berkenhout 1769, Rodentia: Muridae) is a ubiquitous pest, predominantly associated with human activity (Cavia et al. 2019; Cavia & Gómez Villafañe 2023). Originated in northern China and Mongolia, it disseminated through Asia and Europe via the silk and fur trade routes and arrived in America around 1775 as stowaways on merchant ships (Coto 1997; Puckett et al. 2016). Despite the substantial knowledge of various aspects of its biology and ecology, the full extent of its dispersal capacity remains unclear. Nevertheless, certain aspects of its daily movements are documented. For example, a maximum daily movement of 65 meters in rural areas and an average home range of 240 m² within poultry farm boundaries have been observed (Gómez Villafañe et al. 2008). Other studies have reported a mean daily activity area of 118 m² in pig and dairy farms (Montes de Oca et al. 2017). In contrast, day-to-day movement distances of up to 200 meters have been noted for urban rats in sewer systems (Heiberg et al. 2012). Regarding dispersal movements, it has been reported individuals covering distances ranging from 2–6 km and up to 11.5 km in urban areas (Gardner-Santana et al. 2009).

In rural areas, R. norvegicus poses a significant threat by consuming and contaminating the food of breeding animals, causing substantial damage (Meerburg et al. 2010). This leads to pre-harvest and post-harvest losses, particularly affecting cereal crops. Additionally, it also inflicts harm on human structures and stored food (Drummond 2001; Lambert et al. 2008; Meerburg et al. 2009; Brown et al. 2020). Moreover, is a primary transmitter and, in many cases, acts as a reservoir or vector for various diseases, including salmonellosis, leptospirosis, trichinosis, hemorrhagic fever with renal syndrome caused by hantavirus, bubonic plague, among others (Webster & Macdonald 1995; Coto 1997; Meerburg et al. 2009).

In Argentina, R. norvegicus is present in both rural and urban environments (Cavia et al. 2009; Vadell et al. 2010; Gómez Villafañe et al. 2013; Lovera et al. 2015; Cavia et al. 2019; Maroli & Gómez Villafañe 2021). In the north-eastern region of Buenos Aires province, it establishes itself in farms and towns, serving as favorable patches within an unfavorable matrix comprising grasslands, pastures and crop fields (Gómez Villafañe et al. 2008; Gómez Villafañe et al. 2013).

The harm inflicted by R. norvegicus on human health and agricultural activities likely stems from its high reproductive potential and close association with humans. Additionally, the implementation of costly short-term solutions, that are proven to be ineffective in the long term, contributes to ongoing issues (Macdonald et al. 1999; Lambert et al. 2008). When addressing these challenges, the concept of "Ecologically-based pest management" (Singleton et al. 1999) suggest using knowledge of the biology and ecology of pest species, encompassing factors such as dispersal patterns, habitat use, population structure, and reproductive behavior, to define management units (Piertney et al., 2016). Using this information as a decision-making tool can contribute to the formulation of more effective long-term management plans, thereby preventing re-infestations from neighboring areas (Mwanjabe & Leirs 1997). It is crucial to understand the specific spatial and environmental context that defines neighbor areas for each species and to determine the extent to which rats re-infest farms after local eradication.

Limited research has been conducted on the dispersal of R. norvegicus using molecular markers, such as microsatellite loci (Gardner-Santana et al. 2009; Kajdacsi et al. 2013; Berthier et al. 2016) and SNPs (Combs et al. 2018a; Combs et al. 2018b; Byers et al. 2020). These studies have predominantly focused on urban areas (Gardner-Santana et al. 2009; Kajdacsi et al. 2013; Berthier et al. 2016). Conversely, this kind of research in rural areas is scarce (i.e. Varudkar & Ramakrishnan 2015; Gómez Villafañe et al. 2019). Studies conducted in urban environments have unveiled a distinctive population structure characterized by clusters with limited gene flow between them (Scaltritti et al. 2025). This contrast with the pattern observed in rural areas, where there is no genetic subdivision and high levels of gene flow among farms (Varudkar & Ramakrishnan 2015; Gómez Villafañe et al. 2019).

This study aims to gain a deeper understanding and assess the level of genetic cohesiveness among R. norvegicus populations in a rural context. Additionally, this study aims to identify potential dispersal barriers and/or corridors for dispersal using a population genomic approach. We hypothesize that individual dispersal within the study area is constrained, reflecting a non-random movement pattern. Specifically, we anticipate that riverbanks serve as more effective corridors for facilitating the dispersal of individuals between favorable habitats compared to the borders of main roads.

Between winter 2016 and autumn 2018, specimens of R. norvegicus were sampled across 16 farms in a rural landscape covering 561,736 hectares (Online Resource 1). These farms were located along two perpendicular transects: One following the De la Cruz river, extending from Exaltación de la Cruz to San Andrés de Giles (from north-east to south-west), and the other transect following National Road 5 and then National Road 6, from San Andrés de Giles to Marcos Paz and General Las Heras, passing through Luján (from north-west to south-east) (Fig. 1b).

Specimens were captured using custom-designed live cage traps (15 × 16 × 31cm) suitable for rodents weighing over 30 grams. The traps were pre-baited with meat and carrot, and were strategically placed across various environments within each farm, including livestock sheds, silos where animal feed is stored, ditches, and other locations. Liver samples from each rat were dissected and stored in 100% ethanol for future DNA extraction.

Out of the total 222 individuals collected, two to ten individuals were selected (typically five) from each farm, resulting in a total of 93 individuals for genotyping. To ensure a broad representation of genetic diversity, specimens were deliberately chosen from traps located as far apart as possible within each farm, aiming to minimize the risk of consanguinity (Online Resource 1).

DNA was extracted from approximately 35 mg liver samples, using a salting-out protocol (Reiss et al. 1995). After digesting the tissue digestion with proteinase K, 20 µl of RNase A (PBL, Productos Bio-Lógicos®, 10 mg·ml-1) was added to each extract and incubated at 37°C for one hour to eliminate residual RNA. Subsequently, the DNA concentration and quality were assessed using a Qubit v2.0 fluorometer (Life Technologies) and a ND-1000 spectrophotometer (Nanodrop Technologies), respectively.

Extractions meeting the criteria of DNA integrity, as confirmed by inspection on a 1% agarose gel. These extracts contained at least 1 µg of high-quality genomic DNA per sample with a 260/280 index between 1.8–2 and a 260/230 index > 1.8. Then, a library of individually barcoded Restriction site-Associated DNA (Baird et al. 2008) was then prepared following the protocol described by Roesti et al. (2012; 2013), based on the methodology outlined in Hohenlohe et al. (2010). Briefly, the DNA samples were digested using the SbfI restriction enzyme. Then the adapter containing the barcode was bounded to the overhangs of the resulting fragments. All individuals were then pooled, and PCR amplification of the fragments was conducted using the forward primer embedded in the adapter sequence. PCR products falling within the range of 350–600 bp were gel-purified.

The libraries were subjected to single-end sequencing with 150 bp reads on a HiSeq4000 Illumina platform at the Genomics & Cell Characterization Core Facility (GC3F) of the University of Oregon, Eugene, USA.

The analysis of raw Illumina reads was conducted using the STACKS software (Catchen et al. 2013). Demultiplexing and filtering of low-quality reads (Phred score lower than 10) were performed with the process_radtags module of STACKS. One mismatch was allowed with both the barcode sequence and the adapter sequence. Subsequently, reads from all individuals were aligned to the R. norvegicus reference genome downloaded from NCBI (accession number JACYVU000000000, version JACYVU000000000.1; author: Howe K; unpublished) using Bowtie 2 (Langmead & Salzberg 2012). The alignments were then converted to BAM format using Samtools (Li et al. 2009). Next, we used the ref_map pipeline of STACKS with the default parameters on the RAD stacks, setting a minimum read depth of three. Loci with a minor allele frequency of at least 0.03 and a representation across all farms of at least a frequency of 0.5 were retained. This resulted in a multilocus matrix for further analyses. Additionally, individuals with more than 60% of missing data were removed.

After filtering, a total of 764 SNPs were retained from the analysis of 81 individuals, excluding those with more than 60% missing data out of the original 93. An average read depth of 3.86x per SNP per sample was obtained (Online Resource 2). Nucleotide diversity (P_i) ranged from 0.104 (E3) to 0.171 (G2). The mean observed heterozygosity (H_O) varied from 0.101 (M4) to 0.129 (M6) and was consistent with Hardy–Weinberg equilibrium. The number of private alleles (N_P) ranged from 0 to 3, with the highest count observed in L1, while the proportion of polymorphic loci (i.e., the proportion of polymorphic sites relative to the total number of variant sites) ranged from 0.148 (E3) to 0.461 (M1). Moreover, the inbreeding coefficient (F_IS) was close to zero for most farms, with M1 showing the highest value (0.111; Table 1).

The NJ tree revealed a scenario of low genetic differentiation among farms (Fig. 2a). Although genetic distances were low, a geographical trend emerged: farms in close geographical proximity tended to cluster together. Moreover, two distinct clusters were observed: the first one exclusively consisted of farms from Marcos Paz, General Las Heras and General Rodriguez (M3, M4, M5, M6, and M7), and the second one encompassed the remaining farms. This latter cluster was subdivided into two sub-clusters: the first brought together the three farms from Exaltación de la Cruz and one site from San Andrés de Giles (E1, E2, E3, and G3), while the second included the two farms from Luján with the remaining farms from San Andrés de Giles (L1, L2, G1, and G2).

Pairwise F_ST comparisons revealed that, on average, each farm differs significantly from approximately fifty percent of the other farms (Fig. 2b). Moreover, these differences tended to occur between farms that were distantly located. However, statistically significant differences were also observed between nearby farms (i.e. E3 with E1 and E2 or M6 with M5; Fig. 2b), which closely matches the groupings retrieved by the NJ analysis (Fig. 2a).

According to the PCA the first and second components explained 4.1% and 3.4% of the genetic variance, respectively. When individuals were ordered along these two components (Fig. 2c), the centroid of the groups of individuals from each farm showed a pattern similar to that observed in the NJ tree (Fig. 2a). On the right-hand side, we observed a grouping of the centroids of individuals from the farms from Marcos Paz, General Las Heras and General Rodríguez (M1 to M8), while on the left-hand side of the PCA, we found centroids of individuals of the rest of the farms that constituted the other main group in the NJ tree, albeit without the subdivision observed in that analysis.

After excluding the 19 individuals identified as consanguineous based on the output from COLONY FS families (Online Resource 3), the Bayesian assignment analysis indicated that K = 9 was the most likely number of genetic clusters, with K = 2 emerging as the second most probable scenario (Online Resource 4a). However, significant levels of admixture were observed in the majority of individuals with K = 9, who displayed contributions from nearly all clusters across their genomes (Online Resource 4b). Notably, for K = 2 (Online Resource 4c), individuals from the farms in Marcos Paz, General Las Heras and General Rodríguez appeared to be separated from the rest. Most individuals from these areas were assigned mainly to one of the two clusters, while individuals from the other farms (Luján, San Andrés de Giles, and Exaltación de la Cruz) were mainly assigned to the other cluster. However, all individuals showed to some extent some degree of mixing. This result aligns with what was observed in the NJ tree and with the trend observed in the PCA.

Spatial interpolation of individual genetic distances revealed differences in levels of differentiation between areas around the river and the road transects (Fig. 1b). A gradient of genetic differentiation was observed around the road, with higher values of differentiation around farms in San Andrés de Giles and lower values around farms in Marcos Paz, General Las Heras and General Rodríguez (Fig. 1b). Conversely, the rivers transect displayed higher levels of differentiation, particularly in Exaltación de la Cruz and San Andrés de Giles, compared to the broader genomic landscape. Moreover, according to the Mantel correlation test, a pattern of IBD was revealed along the road (r correlation value = 0.089; P-value = 0.028), though this pattern was not observed along the river (r correlation value = -0.051; P-value = 0.71) (Online Resource 5 a-b).

The kinship analysis identified 22 pairs of full-sibling or parent-offspring relationships. Based on estimated birth dates and collection sites, eight pairs were more likely full-siblings, eight were parent-offspring, and the remaining six could correspond to either type of relationship (Online Resources 1 and 3). Additionally, we detected nine pairs of HS across the study area (Online Resource 3). Most of these kinship pairs were concentrated along the road transect where most of the detected relatives were either from the same farm (0 km) or adjacent farms. While the proportion of HS pairs dropped sharply from null distance across this road transect, FS pairs exhibited a more gradual decrease (Fig. 3a-b). Accordingly, kinship frequency declined with increasing geographic distance (Online Resources 6 a-b; Online Resource 7). In contrast, no clear patterns were observed along the river transect due to the limited number of detected relatives: only three FS pairs (separated by 0, 28.1, and 35.97 km) and two HS pairs (both separated by 9.61 km) were found (Online Resource 3). Nevertheless, the decline in FS proportions followed a similar trend to that observed for the road (Fig. 3b). Notably, seven FS pairs and three HS pairs were captured at distances exceeding 10 km in both transects, including one FS and one HS pair separated by over 60 km along the road.

This study provides new insights into the population genomics of R. norvegicus sampled along two transects—one following a river and the other following a road—in a rural area. While numerous studies have explored the population structure and dispersal of R. norvegicus in urban landscapes, to the best of our knowledge, this study is the first to investigate this topic in a rural landscape using SNPs as genetic markers. Although there was low genetic differentiation between groups of individuals from different farms—indicating substantial gene flow—the distribution of genetic variability across the study area was not random. As expected, individuals from geographically close farms tended to be more genetically similar, suggesting that dispersal is spatially limited. This was especially evident along the road transect, where a pattern consistent with IBD was revealed: farms that were closer to each other showed greater genetic similarity and higher frequencies of close relatives, while this signal decreased with geographic distance. This pattern of IBD was not observed along the river transect, where a weaker gradient of genetic differentiation was expected (i.e. smoother correlation slope), based on the assumption that the river would act as a more effective dispersal corridor. These findings support our hypothesis that rat movement across the landscape is constrained and not entirely random. However, it also suggests that the spatial influence on genetic structure may vary depending on the type of landscape feature (i.e. watercourses or roads).

The IBD pattern suggests the presence of a natural dispersal mechanism across generations (i.e., active dispersal). Further analyses are necessary to understand how different environments are used for dispersal. However, we hypothesize that rats may establish colonies in farms or in other human structures that offer abundant resources but are surrounded by a less favorable landscape composed of grasslands, pastures, and crop fields that limit dispersal movements. Despite this, human structures are scattered across the rural landscape. Since farms are not widely separated along the road transect, rats may be able to migrate between neighboring farms and establish new colonies. This movement could lead successive generations to show increased genetic differentiation with greater geographic distance.

We also expected to observe IBD along the river transect; however, such a pattern was not observed. There are several possible explanations for this. First, the transect may not have been long enough to detect an IBD pattern, given that the watercourse and its borders likely facilitate dispersal more than the road. If dispersal along the river occurs with minimal constraint, genetic differentiation between individuals from different farms may be too weak to detect an IBD pattern at this scale. Second, the transect may not aligns perfectly with the dispersal corridor provided by the river. This could mean that effective dispersal distances do not correspond to measured geographic distances, potentially obscuring a genetic distance–geographic distance correlation. Third, geographical barriers along the river could limit gene flow. However, this hypothesis can be reasonably dismissed as we detected two pairs of full siblings between geographically distant sites (Table S2), F_ST values were low and neither different groups were detected. Fourth, it remains possible that rivers do not necessarily facilitate dispersal more than other features of the landscape. Finally, another possible explanation is that rats move along the river based on resource availability or social conflicts. They could move upstream and downstream more freely than they can along the road. Consequently, they may not need to settle in a new location after dispersal, as seen along the road, providing opportunities for further dispersal or return to the birthplace. This dispersal dynamic could lead to less discernible patterns across generations. This could explain why we only captured a few relatives along this transect, dispersed over different distances and not concentrated at distance zero (i.e., the same farm), as observed along the road. Rats along the watercourse may have even more opportunities to form new groups of individuals from different and distant birthplaces. Nevertheless, further studies analyzing diverse and more extended transects along rivers and streams are essential for a comprehensive understanding of dispersal along watercourses in rural areas.

In summary, although dispersal tends to be spatially limited, periods of resource scarcity may drive individuals to move across broader portions of the landscape. Alternatively, if resources are abundant, they might exhibit a tendency to remain at their birthplace (Davis et al. 1948). Therefore, differences in landscape structure and resource availability could account for the observed distinctions between transects.

Low values of genetic differentiation were observed between farms, indicating high levels of mixing among individuals collected from different farms. Besides, while the NJ, PCA, and STRUCTURE analysis for K = 2 show some differentiation of Marcos Paz farms from others, globally we cannot assert the existence of two well-defined groups. Furthermore, even in this scenario, the separation of Marcos Paz farms is not very pronounced, as no individual has been completely assigned to one of the two clusters; all showed to some extent a degree of mixing. In summary, although Marcos Paz farms tend to separate subtly from the rest, we cannot state the presence of a clearly defined population structure overall. Similarly, the absence of genetic subdivision and extensive gene flow in R. norvegicus was previously reported in a rural landscape using microsatellites as genetic markers. This study was based on a sampling scheme that only included farms from a single department (Exaltación de la Cruz) (Gómez Villafañe et al. 2019). Interestingly, Varudkar and Ramakrishnan (2015) compared the genetic structure and gene flow between two species of rats, the non-commensal Rattus satarae (Hinton 1918) and the commensal Rattus rattus (Linnaeus 1758), in rural (i.e., villages) and natural (i.e., pristine forests) environments in a mountain range in India using mitochondrial and microsatellite markers. The authors found high genetic partitioning and differentiation with low levels of gene flow and migration rates for the non-commensal R. satare, and high levels of admixture (i.e., no clustering), slight levels of genetic differentiation, and high migration rates and gene flow over long distances in the commensal species. These findings in R. rattus are in agreement with the results reported herein for the commensal R. norvegicus in a rural landscape.

Overall, our findings in rural areas as well as those reported by Gómez Villafañe et al. (2019) highlight a markedly different genetic structure compared to that reported in urban settings. Previous studies have shown strong genetic clustering, limited dispersal, and high levels of inbreeding among urban rat populations (Berthier et al. 2016, Byers et al. 2020, 2021, Combs et al. 2018, Scaltritti et al. 2025). These patterns are generally attributed to spatial constraints and abundant, localized resources (Himsworth et al. 2014; Byers et al. 2021). In contrast, the genetic patterns we observed suggest greater dispersal and admixture in rural populations, with correspondingly lower inbreeding. These differences likely reflect distinct ecological pressures and movement dynamics between environments (Barnett 1963; Gardner-Santana et al. 2009; Himsworth et al. 2014). Thus, our results highlight the importance of context-specific studies and caution against extrapolating genetic patterns across ecological systems, even within the same species, particularly when developing control or management strategies.

The observed IBD pattern along the road is likely the result of active dispersal across generations because passive dispersal would typically erase an IBD pattern if it were prevalent over natural movements. However, passive dispersal cannot be dismissed, as evidenced by the discovery of pairs of HS and FS at considerable distances (over 30 km, with two pairs even surpassing 60 km, Table S2). This exceeds the maximum reported dispersal distance for R. norvegicus (11.5 km; Gardner-Santana et al. 2009). These long-distance movements can be explained by human-mediated transport, such as trucks carrying food for breeding animals (Lack et al. 2012; Berthier et al. 2016; Gómez Villafañe et al. 2019). Rattus norvegicus is known to utilize human transportation vectors, like ships or trucks (Lack et al. 2012). While roadways may act as barriers to dispersal in urban contexts (Combs et al. 2018), in rural environments, they could facilitate the dispersal of rodent species through commercial transport, even across unfavorable environments (Berthier et al. 2016). Along the river, the predominant process is more difficult to determine, but active dispersal seems more likely, given that watercourses may facilitate dispersal movements, as evidenced in this study by the very low number of relatives found within the same farm, compared to the road transect.

In conclusion, our study highlights the role of both water courses and roads as corridors for the dispersal of R. norvegicus and serves as a foundational step for future research aimed at comprehensively addressing our initial hypothesis regarding the constraints on the movement of R. norvegicus across the study area. Further analyses focusing on the resistance of diverse landscape elements, such as grasslands, pastures, and crop fields, using varied cost matrices, are imperative. Identifying additional landscape features that may also serve as corridors will contribute to a holistic understanding of R. norvegicus dispersal in rural landscapes.

Trapping and handling were performed according to national and international animal care guidelines (Giannoni et al. 2003 and Sikes & Gannon. 2011). The captured individuals were humanely euthanized by cervical dislocation after prior anaesthetization with a mixture of ketamine and acepromazine maleate, following the procedures and protocols approved by the Argentine Law for Animal Care 14 346 and Ethics Committee for Research on Laboratory, Farms and Wild Animals from the National Council of Science and Technical Research (CONICET; resolution 1047, section 2, annex II).