When alien catfish meet—Resource overlap between the North AmericanIctalurus punctatusand immature EuropeanSilurus glanisin the Arno River (Italy)

Due to the increasing globalisation and ongoing introduction of alien species specifically regarding European freshwater ecosystems, native and already present alien species will be confronted with competitors with unknown outcomes. One such case is the situation of the European catfish Silurus glanis introduced in the Arno River (Central Italy), a species sought after by anglers, which is facing competition from the later introduced alien North American channel catfish Ictalurus punctatus. Large catfish species are highly valued among anglers, but their interspecific interactions and potential ecosystem‐level impacts are still poorly known. We used stomach contents and stable isotope analyses to study niche partitioning between these two alien catfish species, coexisting in the Arno River. The results suggest partial niche segregation, with immature S. glanis showing a narrower dietary and isotopic niche and a slightly higher trophic position than I. punctatus. Monitoring the catfish population sizes, trophic niches and effects on lower trophic levels are essential for future management and mitigation of their potential impacts on invaded freshwater ecosystems.

. This process has reshaped most freshwater communities. Although some introductions might have occurred accidentally, a significant proportion of species has been voluntarily introduced by anglers and sport fishing associations to improve recreational fishing.
One such species is the European or Wels catfish Silurus glanis (Linnaeus, 1758), a fish native to north-eastern Europe and Asia. It is known to be the largest freshwater Osteichthyes in Europe (max size 2.7 m and 130 kg; Stone, 2007;Boulêtreau & Santoul, 2016).
In the last decades, S. glanis experienced a drastic increase in population size and distribution due to angling activities (Copp et al., 2009;Cunico & Vitule, 2014). It was introduced in Italy in the 1990s (Balma, Delmastro, & Forneris, 1989) and was subsequently spread throughout the Italian peninsula with diverse impacts on native and also newly introduced species (Castaldelli et al., 2013). In the Arno River Basin, it was most likely introduced to Bilancino Lake and then spread downstream towards the Sieve River (a tributary of the Arno River; Gualtieri, Mecatti, & Cecchi, 2010). In the Florentine stretch of the Arno River, S. glanis has been reported by anglers since 2004 (Gualtieri & Mecatti, 2005). In the course of a few years, this species has become the target of many anglers. The increased investigation into invasive S. glanis revealed considerable positive economic (Cucherousset et al.., 2017;Rees et al., 2017) but negative environmental impacts (Castaldelli et al., 2013;Copp et al., 2009).
At the same time, the North American channel catfish Ictalurus punctatus (Rafinesque, 1818) was also introduced to Italy. However, information on pathways and means of introduction remain scarce (Haubrock, Balzani, Johović, Inghilesi, & Tricarico, 2018a;Haubrock et al., 2017;Ligas, 2008). In the Arno River, according to catches by local authorities and anglers, it is likely that I. punctatus was introduced near Pisa for the purpose of angling and was subsequently transported upriver (G. Castaldelli pers. comm.) as it is increasingly favoured by anglers due to its high reproduction rate and catchability (T. Busatto pers. comm.). In the inner-Florence stretch of the Arno River, it first appeared in 2004 (A. Nocita pers. comm.).
Catfish such as S. glanis have specific functions in ecosystems, affecting various ecological groups with their presence. S. glanis is known to be a generalist that is capable of rapidly adapting to new prey sources (Vejřík et al., 2017). However, data about the diet of immature (<60-80 cm) S. glanis from alien populations (but see Mecatti, 2005, andSyväranta et al., 2010, for a basic analysis of S. glanis' diet) and about interactions with other catfish species are scanty. Similarly, much more information is available about genetics and physiology of I. punctatus in regard to aquaculture usage, but most studies that investigated the behaviour of I. punctatus are considerably old, with a minority of studies focusing on its invasiveness and diet (Busbee, 1968;Haubrock, Balzani, Johović, et al., 2018a;Rosen, Schwalbe, Parizek, Holm, & Lowe, 1995).
Although these two catfish species are morphologically unequal, with S. glanis reaching larger sizes, they are highly tolerant to environmental factors and show high invasive capabilities (Copp et al., 2009;Hilge, 1985). Control activities on S. glanis have been carried out in countries such as Italy (see, e.g., Pascale et al., 2013; LIFE-Project LIFE03NAT/IT/000113 and LIFE15 NAT/IT/000989). Furthermore, these two species of catfish have different feeding strategies. With a fast expansion of the mouth while keeping opercular valves closed in the beginning to maximise the sucked volume, S. glanis imbibes prey inside its oral opening (Bruton, 1996). In contrast, the exact morphodynamic prey mechanism of I. punctatus has not been studied, but a combination of sucking and pursuing-and-biting is possible (Pavlov & Kasumyan, 2002). Moreover, length-weight relationships and thus growth rates might differ between both species. However, they are ecosystem and factor depending. Moreover, information for both I. punctatus in Europe as well as S. glanis from the Arno River has not yet been assessed and remains anecdotal. Furthermore, such information is needed to better understand the impact these species have on recipient ecosystems (Haubrock, Balzani, Johovic, et al., 2018b).
However, the diet of I. punctatus has been recently analysed by collecting stomach contents of specimens in the Arno River (Haubrock, Balzani, Johović, et al., 2018a;Haubrock et al., 2017). In contrast, recent and relevant data for S. glanis' feeding habit from Central Italy are preliminary or anecdotal (Gualtieri & Mecatti, 2005), making investigating facilitation processes or interfering competition processes difficult.
Interactions between alien top predators are very likely, because in fish communities, resource partitioning mainly occurs along the trophic dimensions, that is across trophic levels (Ross, 1986).
Stomach content analyses can provide a direct insight into recently consumed items and, consequently, feeding behaviour of different life stages, while stable isotope analysis (SIA) provides longterm dietary information. Both combined allow a reconstruction of the studied communities' trophic web and the investigation of niche partitioning (Boecklen, Yarnes, Cook, & James, 2011;Layman et al., 2012). This is possible because isotopic signatures relate with the trophic position (nitrogen, N) and with the main energy source (carbon, C) of a consumer (Layman et al., 2012). Additionally, SIA allows the estimation of trophic levels (Post, 2002) and quantification of feeding niches (Jackson, Parnell, Inger, & Bearhop, 2011;Newsome, Martinez del Rio, Bearhop, & Phillips, 2007;Swanson et al., 2015).
The degree of overlap among isotopic niches of different species can therein be assessed to evaluate the potential food competition between species (Balzani et al., 2016;Layman et al., 2012) and niche partitioning (Barbour et al., 2009;Werner, 1979).
Over the last years, many anglers and local authorities reported steadily decreasing catches of S. glanis (abundance and biomass) in the Arno River within Florence while simultaneously the amount of I. punctatus increased (Gherardi et al., 2008;Nocita, 2007;Nocita & Zerunian, 2007). Hence, the present work uses a combined approach of stomach content and stable isotope analyses from both co-occurring species (immature and mature I. punctatus, immature S. glanis) in order to investigate the degree of interspecific interactions (e.g. feeding competition or prey partitioning), thereby aiming to determine whether the presence of I. punctatus potentially affects the population of S. glanis.

| Sampling
Sampling was conducted twice a week, once in the morning (3-8 a.m.) and evening (4-11 p.m.) between May and June 2018, as these months are considered an ideal activity period for both species (samples were taken after the initial 2-month period of resumed activity, following the inactive winter months). These times were chosen because S. glanis shows its activity mostly during the night with peaks of activity before sunrise and after sunset (Carol, Zamora, & García-Berthou, 2007) while I. punctatus shows diurnal and nocturnal activity with highest feeding activity during dusk and dawn (Boujard & Leatherland, 1992). Moreover, February and March are typically periods of high water currents and strong rains, negatively affecting the sampling. The far west end of the inner-Florence river stretch was chosen due to its accessibility for fishing and the low human activity on both riversides ( Figure 1).
The sampling area was characterised by high turbidity and no aquatic vegetation with scarce vegetation on river banks during the first half of the sampling period due to seasonal patterns and anthropogenic disturbance. Fish were caught using standard fishing rods (2.20-3.90 m), 0.30-0.44 mm monofilament line and size 2-26 fishing hooks baited with a variety of food (maggots, worms, and freshly cut liver or baitfish) placed (a) on the bottom, (b) in the middle water and (c) below the surface. Caught specimens were euthanised (gill cuts) and put on ice before being transported to the laboratory for further processing. In total, 40 individuals of I. punctatus were sampled and the stage of maturity was set according to Haubrock, Balzani, Johovic, et al. (2018b) as 31.9 cm, resulting in 20 "immature I. punctatus" and 20 "mature I. punctatus." Additionally, 39 individuals of "immature S. glanis," recognised by the lack of mature and visible gonads and size ranges for immature individuals in this species identified by Copp et al. (2009), were sampled and are referred to as "S. glanis." Older and bigger S. glanis were not sampled due to (a) their rarity in the sampled stretch and (b) local pressure by anglers.

| Sample preparation
For all specimens, total length (TL; cm) was measured with an accuracy of 1 mm and body weight (F W ) with an accuracy of 1 g. Values F I G U R E 1 Sampling areas within the inner-city section of the Arno River in Tuscany (N43.7648903, E11.2697444) reported in the text are average plus or minus the standard deviation. Specimens were stored in a freezer (−20°C) until stomach contents were extracted for dietary analyses. For SIA, a sample of dorsal muscle tissue was taken from each specimen, preserved without any chemicals and stored in a freezer (−20°C). Samples were dried in an oven at 60°C for 48 hr and ground into a fine and homogenised powder with an agate mortar and pestle. For each sample, approximately 0.25 mg of powder was weighed and packed into a tin capsule for isotopic analyses. Samples were analysed with an elemental analyser (FlashEA 1112) connected to an isotope ratio mass spectrometer (Thermo Finnigan Delta Plus Advantage) at the National Research Council in Montelibretti, Rome.

| Growth and dietary analyses
As fish length affects fish weight, examination of the lengthweight relationship was achieved with a logarithmic transformation of fish weight and length. Applying a linear regression model, the slope of both linear models was used as a proxy to identify growth types (allometric vs. isometric; Karachle & Stergious, 2012). Thus, the species-specific growth was compared by testing for homogeneity of slopes in an analysis of covariance (ANCOVA; Dhillon & Fox, 2004) with the state "species" used as categorical independent variable. To compare the weight gained with increased length, weight was set as the dependent variable and length as covariate.
The weight of ingested food (ST W ) was expressed as the percental weight of the specimens total weight (F W ) following the approach of Hureau (1969): F I = (100*ST W )/F W , where F I is considered as the fullness index. Only cardiac, caecum and pyloric parts were considered without discarding contents of the intestine. The feeding intensity was calculated using the vacuity index (V I ), as the percentage of empty stomachs with respect to stomachs that contained prey items (Batistić et al., 2005). The diet breadth was estimated based on Levin's index formula (Whittaker, Levin, & Root, 1973): where B i is the standardised index of diet breadth for specimen i and p the sum of the squared proportion of each prey item of specimen I (Levins, 1968). For further analyses, the stomach content data were expressed as frequency of occurrence (F% = number of stomachs containing each food item in relation to total number of full stomachs) and abundance (N% = the number of individuals of each food item with respect to the total number of individuals). With these, the prominence value (PV) for each dietary component was estimated following the approach of Hickley, North, Muchiri, and Harper (1994): PV = N% * √ F% to estimate the diet overlap index α using the formula of Schoener (1986): being the number of food items, PV x i the prominence value of food item i in species x i and PV y i the prominence values of food item i in species y. The estimated index varies between 0, that is no overlap, and 1, when diets are identical (Wallace, 1981). Pianka's measure of diet-based niche overlap (Pianka, 1974) for species j and species k was calculated, where p ij and p ik are the proportions of the i th resource used by the j th and k th species.
Both estimates for diet overlap were used together, as Pianka's measure of niche overlap alone can be insufficient to identify a relation between niche overlap and competition, that is resource partitioning (Krebs, 1999). We utilised a permutational analysis of variance (PERMANOVA; S17 Bray-Curtis similarity, 3 fixed factor, three levels: "immature I. punctatus," "mature I. punctatus" and "S. glanis"; Type III (partial) sum of squares, unrestricted permutation of raw data, 1 dummy variable) and pairwise comparisons using the software PRIMER (Clarke & Warwick, 2001) to determine whether diets varied between species and life stages.

| Stable isotope analyses
The  Post, 2002). A primary producer was chosen rather than a primary consumer (e.g., molluscs) due to the lower variability of stable isotope values in plants (Bissattini, & Vignoli, 2017 ;McClelland, Valiela, & Michener, 1997)  δ 15 N and δ 15 C ranges (NR and CR) as well as the mean distance to the centroid (CD) are measurements of the extent of the δ 15 N/ δ 15 C bi-plot, reflecting the community niche width. The mean nearest neighbour distance (MNND) and the standard deviation of the nearest neighbour distance (SDNND) relate to the distance of species and indicated the degree of trophic redundancy. While Layman's metrics are usually employed in community-wide analyses, they can be used to compare species present within one ecosystem without consideration of the entire species community (Bissattini, & Vignoli, 2017;Jackson et al., 2011;Layman, 2007). Additionally, the corrected standard ellipse area (SEAc; considering 40% of central data points) and the corresponding 95% ellipse area (SEAb), which are convex hull areas encompassing a sampled population in a δ 15 N/δ 15 C bi-plot space and serve as measures of total isotopic niche space occupied, were calculated for both species and life stages for I. punctatus. With these, the degree of isotopic niche overlap (which lays between 0, i.e., no overlapping, and 1, completely overlapping) can be estimated and subsequently used as a quantitative measure for similarity within the diet of different species (Jackson et al., 2012). Calculations were computed using the R package SIBER (Jackson et al., 2011). Furthermore, the percentual isotopic niche overlap between groups was calculated using the formula of Stasko, Johnston, and Gunn (2015).
The directional pairwise probability of species to exert overlapping niche width was estimated for I. punctatus and S. glanis using the R-package "nicheROVER," which applies a Monte Carlo estimation (chain length: 10,000 steps) on the potential overlap. Thus, the package computes the directional pairwise probability of the niche of one species (not either species) overlapping onto the niche of another to quantify trophic relationships (Swanson et al., 2015).
Lastly, a PERMANOVA (1 fixed factor, three levels: "immature I. punctatus," "mature I. punctatus" and "S. glanis"; Type III (partial) sum of squares, unrestricted permutation of raw data) was performed to test for significant differences in δ 15 N and δ 13 C levels among life stages as well as species and to eventually determine whether groups and/or species are clearly distinguishable by either isotope.
Additionally, a canonical analysis of principal coordinates (CAP) for factors whose levels were identified by the PERMANOVA test to be significantly different was applied to identify how δ 15 N and δ 13 C contribute to differentiate the levels. Pearson correlations with CAP axis for each variable are reported. For all tests, the level of significance under which the null hypothesis was rejected is α = 0.05.  Figure 2).

| RE SULTS
Resultant, I. punctatus growth type can be seen as nearly isometric (similar length-weight growth), while "S. glanis" grow slower in weight than in length.

| Stable isotope analyses
With a trophic score of 3.3, S. glanis occupied a slightly higher trophic position than immature (3.1) and mature I. punctatus (2.9).
Silurus glanis, however, had a lower variability in Layman's metrics and lower N range compared to both immature and mature I. punctatus. In contrast, both I. punctatus showed a narrower C range (Table 3). The applied linear regression model for δ 15 N and total length of both species indicated a decreasing trophic position with increasing size that was significant for I. punctatus  Note. "0.00" indicates that the specific prey items in the fish's diet were considerably low; "na" indicates that respective prey item was not present.

F I G U R E 3 Linear regression between total length (TL) and δ 15 N and δ 13 C for Ictalurus punctatus (a, b) and Silurus glanis (c, d) from the inner-Florence section of the Arno River
The PERMANOVA main test confirmed significant differences among groups, and the post hoc comparisons showed that the three levels are clearly different (Table 5)

| D ISCUSS I ON
Interactions among alien fish species, especially with different life histories, are an understudied topic. More specifically, there is a lack of studies concerning trophic relationships among alien top predator's native to different geographic areas while co-occurring in a new region, making this, at least to our knowledge, the first study assessing this issue.
In this study, S. glanis was observed to mainly feed on small cyprinids and Dikerogammarus villosus, fitting to this species early developing piscivorous diet with generalist tendencies (Copp et al., 2009;Syväranta et al., 2010). Moreover, S. glanis was identified to be less of a generalist than I. punctatus, which showed a more diverse diet including insects, plant material, crayfish and fish that was highly variable among life stages. Additionally, all specimens of I. punctatus contained prey in their stomach, while almost half of S. glanis' stomachs were empty. The observed high number of empty stomachs in immature S. glanis was unexpected as especially smaller fish feed every day (Ricker, 1946). However, this result could be linked to the utilised sampling methods, inducing a lot of stress on hooked specimens and potentially resulting in vomiting of previously ingested prey, rather than on the time of sampling. Moreover, the length-toweight ratio of I. punctatus was higher than that of S. glanis and the fullness index estimated for I. punctatus was in all three cases (immature, mature and all specimens) more than twice than that measured for S. glanis. Concerning I. punctatus, omnivorous feeding with insectivorous tendency agrees with its diet and feeding behaviour from the native area (Busbee, 1968;. The high occurrence of plants (terrestrial as well as aquatic leaves) within the diet of sampled S. glanis is highly unusual (Carol, Benejam, Benito, & García-Berthou, 2009). However, alien species (and especially S. glanis) are known to adapt to new and easily accessible prey (Copp et al., 2009;Strayer, 2010). Abundant algae or plants could potentially be a response to either scarce animal prey or an adaptation to abundant vegetation as a replacement to avoid intra-as well as interspecific TA B L E 4 Estimated niche overlap for all groups using the maximum likelihood fitted standard ellipse area (SEAc) and the overlap between the corresponding 95% ellipse area (SEAb) competition (Copp et al., 2009;Davidson, Jennions, & Nicotra, 2011;Haubrock, Balzani, Johovic, et al., 2018b). Nonetheless, the used fullness index is potentially affected by physiological or morphological differences, for example differing weight-to-length ratios, but since both species belong to the order Siluriformes, these can be considered of secondary importance. However, these results indicate an advantageous adaptation to the invaded ecosystem, more efficient predation by I. punctatus or a comparably higher growth rate commonly found in majorly produced aquaculture fish species.

| Diet and niche overlap
Schoener's diet overlap between all groups was considerably high (α > 0.6), but while Pianka's diet-based niche overlap was similar between "mature I. punctatus" and "S. glanis," "Immature I. punctatus" and "S. glanis" showed a lower diet-based niche overlap value (0.36) with respect to Schoener's diet overlap value. This difference can be due to the "prominence value" in the used formula, weighing the importance of prey items and, thus, being affected by the sample size.
In contrast, SIA and the derived estimated ellipse overlay of SEAc indicated a higher similarity between niches occupied by "immature I. punctatus" and "S. glanis" than between "mature I. punctatus" and "S. glanis." Opposed to the narrow time-constrained information presented by stomach content analysis, SIA offers temporally integrated information on diet as well as, with SEAc and SEAb, occupied feeding items that can be affected by activity times or responses of species to being caught, while isotope analyses consider previously consumed and assimilated prey, thus a considerably longer time (Tarkan et al., 2018). Additionally, all three groups belong to approximately the same estimated trophic position as expected from I. punctatus and S. glanis (Gido & Franssen, 2007;Syväranta et al., 2010), but "S. glanis" had a slightly wider δ 13 C range than the two life stages of I. punctatus. This observation reflects the consumption of a greater variety of carbon sources. A study on S. glanis in France found that S. glanis was considerably enriched and variable in δ 13 C with respect to other fishes, explainable with the frequent predation on terrestrial birds and/or mammals (Syväranta et al., 2010). However, this study lacked large-bodied individuals. Nonetheless, it seems reasonable to expect similar results, as anglers' observations confirm the active predation on pigeons by large S. glanis in the Arno River (T. Busatto, pers. comm.; Cucherousset et al., 2012). Similarly, δ 15 N values tend to increase with the length of catfish (Syväranta et al., 2010), reflecting an increased piscivorous or animal diet. While this might be true for S. glanis (Syväranta et al., 2010), our results show generally lower mean δ 15 N values and lower trophic position of "mature I. punctatus" with respect to "immature I. punctatus." This can potentially be explained by the high level of eutrophication (Romero et al., 2013) and the natural accumulation of nutrients (i.e., dead matter) in the river bottom (Hansen & Kristensen, 1998;Hendrix et al., 1986). Immature I. punctatus are benthivorous and often detritivorous, while mature individuals exert increased offshore feeding (as indicated by the difference in δ 13 C values; Gido, Franssen, & Propst, 2006). Additionally, the lower mean δ 15 N values could be linked to a considerable competition for increasingly piscivorous mature I. punctatus with other species like mature S. glanis (Fry et al., 1999). Moreover, Fry et al. (1999) (Busbee, 1968;Rosen et al., 1995), while "S. glanis" showed a lower diet and diet-based niche overlap but a more similar isotopic niche occupation. Considering the estimation of higher probabilities for I. punctatus (immature and mature) to occur within the niche of each other as well as within the niche of S. glanis indicated that I. punctatus shows a higher plasticity in its niche occupation.
The high number of empty stomachs in S. glanis and high frequencies of certain prey items (e.g., detritus, plants, algae, small cyprinids) in the diet of I. punctatus result in an overall similar prey assortment for the diet analyses, while the associated energy intake and thus isotope assimilation might differ. However, the estimated overlap and thus potentially arising competitions (aside from competition for shelter) are only potential explanations for the observed decrease in S. glanis. Furthermore, it is possible that the feeding mechanism of "sucking-in" of small-sized S. glanis is less efficient compared to the more active predation mechanism used by I. punctatus. Also, Menzel (1945) observed that especially immature I. punctatus predate eggs and larvae or fish species, as they commonly share the same habitat, enabling this species to threaten entire populations in relative short time after its introduction (Lentsch, Converse, & Thompson, 1996;Marsh & Brooks, 1989;Nesler, 1995). In the case of the Arno River, I. punctatus is seemingly more efficient in its prey consumption and shows signs of an advantageous biology, making it better adapted to the non-native ecosystem that is slower-growing S. glanis of similar lengths. Furthermore, this hypothesis is supported by the wider trophic niche occupation and a less specialised diet paired with the observed higher fullness index as well as a higher length-weight ratio in I. punctatus.
The observed differences in stomach content analyses and SIA, based on the different information they provide, underline the opportunistic feeding behaviour of I. punctatus. Moreover, this opportunistic feeding behaviour leads to diet and niche overlap. Additionally, the results also highlight the potential for those two species that share a common habitat and overlap in their activity time to compete for resources in the case they become limited. Moreover, this observed niche overlap likely affects the more specialised S. glanis and could increase in times when resources are increasingly limited. Therefore, the potential of both alien catfish species to compete could be among the main reasons why a decrease in the abundance of mature S. glanis was observed after the introduction of I. punctatus, but more long-term studies covering the entire year while also including mature S. glanis will be needed.

| CON CLUS ION
The presence of the alien North American catfish I. punctatus was previously believed to have a negative impact on the population of S. glanis, indicated by the decreasing abundance of S. glanis and an increase in catchments of I. punctatus (Arno River: A. Nocita and T. Busatto pers. obs.; Po River: M. Fiedler, A. Battaglini and P. Haubrock pers. obs.). However, while results suggest partial niche segregation, they also indicate the possibility for competition that might be due to the potential for immature I. punctatus, which are the dominating life stage (Haubrock, Balzani, Johovic, et al., 2018b;Haubrock et al., 2017;Ligas, 2007), and immature S. glanis, a species that is more specialised in its niche occupation, to compete. Although both species are alien, S. glanis is a large body-sized fish (Copp et al., 2009) and cherished among "catch-and-release anglers" (Arlinghaus et al., 2007) as a trophy fish. However, with the increasing distribution of the fast-growing and comparably more generalist as well as opportunistic I. punctatus (Olden & Poff, 2005;Tucker & Hargreaves, 2004), the population of S. glanis could be negatively affected through competition. Further studies should focus on the degree of predation between big-sized Silurus glanis (>~70 cm) and the two life stages of the North American catfish under the consideration of differing age and length classes.

ACK N OWLED G EM ENTS
We want to thank Laura Buonanno (Florida State University, USA) and Lauren Tonelli (Invasive Species Centre, CA) for thorough proofreading of the manuscript. Funding was provided by the