Novel positive interactions between invasive species: widespread mammals disperse a non-native woody species

Novel ecosystems often emerge from the co-occurrence of species with no shared evolutionary history, leading to unique and potentially synergistic interactions. Positive interactions between non-native species can be crucial for successful invasions. In this study, we investigate the novel interaction between two invasive mammals (Axis axis and Sus scrofa) and their role as dispersers of the invasive tree species Gleditsia triacanthos.

We assessed this dispersal mechanism at a local scale in Entre Ríos, Argentina, by feces sampling. We analyzed the seed content in feces and performed germination experiments to compare them to non-consumed seeds. Additionally, using global occurrence data, we mapped the co-occurrence of both mammal species with G. triacanthos to identify regions of potential synergistic interactions.

We estimated a density of 1917 feces per hectare for A. axis and 267 for S. scrofa. Of the feces sampled, 56% of A. axis and 37% of S. scrofa feces contained at least one G. triacanthos seed. The average number of seeds per feces was 6.8 for deer and 2 for wild boar. Germination experiments confirmed the germination capacity of seeds consumed by both mammals. Global co-occurrence analysis revealed widespread spatial overlap, with stronger co-occurrence between G. triacanthos and S. scrofa.

Synthesis and applications. Our study provides empirical evidence that management plans should adopt an integrated approach. The movement of invasive animals significantly influences propagule pressure and given their broad distribution and widespread use of the landscape, merely reducing plant density is insufficient. These animals ensure the dispersal of the few invasive plants that may remain. Therefore, management strategies should not only focus on controlling plant density but also on regulating animal populations, especially those that facilitate the spread of non-native species. Such integrated management is crucial for preventing the reinforcement of invasive dynamics and improving long-term control efforts.

Methods
Study site and species

Field sampling was conducted at “Potrero San Lorenzo” ranch, an agricultural, forestry, and conservation establishment located in Gualeguaychú, Entre Ríos Province, Argentina (33°00'03"S 58°15'35"W). The climate of the study site is temperate and subhumid, with an average annual temperature ranging between 16.7°C and 18.5°C and an average annual precipitation of 1098 mm (Hydraulic Directorate of Entre Ríos, Official Network of Automatic Weather Stations 2009-2021). The natural vegetation is characterized by grassland physiognomy with open woodlands dominated by native tree species such as Vachellia caven, Neltuma affinis, Celtis ehrenbergiana, Schinus molle, and Scutia buxifolia (Sabattini et al., 2015). Recently, we characterized the woody vegetation structure at the study site, including both native and non-native species (see Supplementary Material 1). The results indicate that native species dominate in terms of adult tree abundance, and that G. triacanthos is still relatively scarce at the adult stage. However, G. triacanthos exhibited a significantly higher sapling-to-adult ratio compared to native species. Regarding the native fauna, notable species include Pseudalopex gymnocercus (the pampas fox), Cerdocyon thous (crab-eating fox), Hydrochoerus hydrochaeris (capybara), Myocastor coypus (coypu), and Rhea americana (greater rhea) (Sabattini et al., 2015).

Gleditsia triacanthos is the most dominant of several non-native woody species invading the region. This species exhibits various characteristics that confer invasive potential, including rapid diameter and height growth, a short juvenile period (5–7 years), high seed production, and a high germination rate (Blair, 1990; Fernandez et al., 2017). Flowering occurs between late spring and summer, while fruiting and seed maturation takes place from autumn to late winter (Ruiz et al., 2022). The main seed dispersal vectors are watercourses (Gantes et al., 2011; Dana et al., 2022) and fauna that consume the fruits due to their high protein and fiber content (Blair, 1990; Ferreras et al., 2015). The seeds possess a hard coat that prevents imbibition; therefore, scarification is required for germination to occur (Ferreras et al., 2015). Seed scarification can take place in the digestive tract of animals that consume fruits or through physicochemical processes occurring in the soil (Blair, 1990; Ferreras et al., 2015).

Additionally, the wildlife of the region is dominated by non-native species such as A. axis (axis deer) and S. scrofa (wild boar) (Gürtler et al., 2018; Shalom et al., 2025). The A. axis is native to Asia, specifically the Indian subcontinent (Tellarini et al., 2019), while S. scrofa originates from Eurasia (Barrios-García & Ballari, 2012). Both species were introduced to Argentina in the early 20th century for hunting purposes (Daciuk, 1978; Navas, 1987). Due to their high reproductive rates (Gethöffer et al., 2007; Herrero et al., 2008), they have successfully expanded their range and are now found across much of Argentina (Tellarini et al., 2019; Ballari & Barrios-García, 2022). Although S. scrofa is omnivorous and A. axis herbivorous, both species have generalist diets and habits, making them among the most invasive animal species worldwide (Johnsingh & Sankar, 1991; Schley & Roper, 2003; Cuevas et al., 2010). Their invasion has significant ecological consequences, as they impact native fauna and flora as well as habitat structure (Barrios-García & Ballari, 2012; Tellarini et al., 2019; Shalom et al., 2025).

Assessing potential local interactions (Objectives 1 and 2)

Feces sampling was conducted in the winter of 2023 using three fixed transects. Each transect measured 50 m in length, starting from a fixed point at the boundary fence and extending into the woodland interior. All feces samples present within 2 m on either side of the transect were recorded and collected, covering a total sampled area of 200 m² per transect. Samples were stored in sealed plastic bags for preservation and subsequent processing. Each sample was labeled to document its collection location and relevant metadata. The identification of feces was carried out using published references (Ballari, 2013 for wild boar, Calfayan, 2021 for axis deer, and Birochio, 2008 for pampas fox) and was corroborated by independent identifications made by the park rangers at the study site. The density and proportion of feces from different species were then estimated. As A. axis and S. scrofa were the only species whose feces contained G. triacanthos seeds, the methods below focus specifically on these two mammals.

All feces samples of A. axis and S. scrofa were analyzed at the laboratory of the Instituto de Investigaciones Fisiológicas y Ecológicas Vinculadas a la Agricultura (IFEVA) to identify and quantify tree seeds. This analysis provided an estimate of the average number of seeds per fecal sample. First, each sample was manually disassembled to check for the presence of seeds. Then, the number of seeds was recorded, distinguishing between G. triacanthos and "other species" as a secondary category.

The germination of G. triacanthos and V. caven seeds (the only “other species” found) was evaluated under four treatments: seeds consumed by axis deer, seeds consumed by wild boar, manually scarified seeds (unconsumed), and control seeds (unconsumed and unscarified). The consumed seeds were collected from feces samples, while the manually scarified and control seeds were obtained from tree pods at the study site. Scarification was performed manually by sanding the seeds with 80-grit sandpaper (Aranda et al., 2015).

The germination experiment was conducted under controlled conditions in germination chambers at 18°C in darkness (Tognetti et al., 2019). Seeds were placed in sterilized Petri dishes lined with moistened filter paper using distilled water. For the consumed seed treatments, one Petri dish was used per feces sample and tree species, meaning the number of seeds varied depending on the content of each sample. For the unconsumed seed treatments (manually scarified and control seeds) ten Petri dishes containing ten seeds each were used for each tree species. In total, 726 seeds of G. triacanthos were placed to germinate in 91 Petri dishes, and 713 seeds of V. caven in 100 dishes.

Over a period of 20 days (Tognetti et al., 2019), seeds were checked every 3–4 days for germination status (e.g., germinated, imbibed, rotted). During this period, seeds were kept moist by regularly spraying them with distilled water. Intact seeds from the consumed seed treatments (those showing no signs of imbibition or rotting) were subsequently scarified to assess germination capacity. This allowed us to evaluate whether seeds that had passed through the animal digestive tract failed to germinate due to damage or insufficient scarification (Ferreras et al., 2015). Germination monitoring and conditions for these consumed and subsequently scarified seeds were identical to those of the other treatments.

Statistical analyses

All statistical analyses were conducted using R (R Core Team, 2024). We first evaluated the distribution of the data using the “fitdist” function from the fitdistrplus package (Delignette-Muller & Dutang, 2015). Count variables (i.e., number of feces per hectare and number of seeds per feces) were best fitted by a negative binomial distribution. In these cases, we used generalized linear mixed models (GLMMs) fitted with the “glmer.nb” function from the lme4 package (Bates et al., 2015). The probability of finding seeds in feces, which naturally follows a binomial distribution, was analyzed using GLMMs with a binomial family and a logit link function (via the “glmer” function in lme4; Bates et al., 2015). Germination experiment results were analyzed using a generalized linear model (GLM; “glm” function) with a binomial distribution and logit link. In this case, the response variable was the proportion of germinated seeds relative to the total number of seeds in each Petri dish.

The fixed and random structure of each model was defined according to the design of the corresponding experiment. To analyze the number of feces per hectare, we included animal species and distance from the edge as fixed effects, and modeled distance nested within transect as a random effect. For the analysis of seed presence in feces, we included animal species and tree species as fixed effects, and feces ID nested within transect as a random effect. The model analyzing the number of seeds per feces included the same fixed and random effects as the previous one but was restricted to samples containing at least one seed (i.e., zeros were excluded). The germination models included treatment and tree species as fixed effects.

Finally, for all models we assessed model assumptions. Diagnostics included checks for overdispersion (in binomial models), residual uniformity, and overall fit using simulation-based residuals from the DHARMa package (Hartig, 2022). We then used the “Anova” function (car package, Fox et al., 2012) to obtain type-II analysis-of-variance tables, and extracted predicted means and standard errors with the “predict” function to generate the result figures. All figures were created using the ggplot2 package (Wickham et al., 2016).

Global analysis of invasive species co-occurrence (Objective 3)

We obtained occurrence records of G. triacanthos (honeylocust), A. axis (axis deer) and S. scrofa (wild boar) from the Global Biodiversity Information Facility (https://www.gbif.org/occurrence). Then, the separate datasets with 120786 occurrence records of G. triacanthos, 5042 of A. axis and 987828 of S. scrofa were compiled in pairs (G. triacanthos - A. axis and G. triacanthos - S. scrofa) and the records with missing or invalid geographic coordinates were removed. To process the spatial data, we converted occurrence records into simple feature objects (sf package; Pebesma, 2018) using the WGS 84 coordinate reference system (EPSG: 4326).

Then, a spatial grid was generated over the global extent, using a grid cell size of 1°, to analyze species co-occurrence. For each grid cell, we counted the number of occurrences of each species. We assigned a presence/absence status for each species in each grid cell and then each grid cell was classified as: Co-occurrence (G. triacanthos and A. axis or S. scrofa present), only G. triacanthos present*,* only A. axis or S. scrofa present*,* or neither species present. A map for each pair (G. triacanthos – A. axis and G. triacanthos – S. scrofa), as well as a map showing the triple co-occurrence (G. triacanthos, A. axis, and S. scrofa), was generated using the ggplot2 package (Wickham et al., 2016). Basemap layers were sourced from the rnaturalearth and rnaturalearthdata packages (South et al., 2017), displaying country and oceans boundaries for context.

Finally, to quantify the degree of spatial overlap between species at a global scale, we used the CooccurrenceAffinity package (Mainali & Slud, 2025) to calculate different co-occurrence indices. For each pair of species (G. triacanthos - A. axis and G. triacanthos - S. scrofa), Jaccard and Sørensen-Dice indices were calculated. These are the most commonly used metrics in species co-occurrence studies, and values close to zero indicate strong negative associations, while values near one suggest strong positive associations. In light of a recent review highlighting the limitations of these conventional indices, we also employed Alpha (α), a statistical parameter that quantifies affinity between two entities based on their co-occurrence patterns (Mainali et al., 2022). Unlike conventional indices, α is not restricted to a fixed range, and negative values indicate avoidance or negative association, values near zero suggest independence, and positive values reflect increasing levels of co-occurrence or positive association (Mainali et al., 2022).