Range expansion of an invasive goby in the River Tiber basin ( Central Italy ) : effects on the abundance and shelter occupancy of a congeneric native goby

Padogobius nigricans and Padogobius bonelli are two gobies native to Italy and characterized by an allopatric distribution: P. nigricans is endemic to the Tyrrhenian catchment of central Italy, while P. bonelli is endemic to the Adriatic catchment of Northern Italy. In the past two decades P. bonelli has successfully established in central Italy outside of its original area of distribution, coming into contact with populations of P. nigricans. The superiority of P. bonelli in competition for breeding sites is well documented by laboratory studies; however, little evidence from the field has been provided. In this study, the occurrence and abundance of the two gobies were investigated in the River Tiber basin (central Italy) to assess the impacts of P. bonelli expansion on P. nigricans populations. Moreover, to investigate nest interference by P. bonelli towards P. nigricans, shelter occupancy of the two Italian gobies was investigated in the field with respect to individual density. Sampling was conducted by electrofishing in 77 sites throughout the upper River Tiber basin and nest occupation was analyzed in 8 river stretches. Padogobius bonelli has further expanded its distribution in recent years, following both natural and human-mediated dispersal patterns. It is numerically dominant over P. nigricans and fewer P. nigricans juveniles were found in the invaded than the uninvaded areas. Depending on adult density, P. bonelli can occupy up to 97% of nest sites and can seize a significant percentage of shelters even at low adult abundance. The expansion of P. bonelli into newly colonized areas in recent years has increased concern for the conservation of P. nigricans. In the invaded areas, the survival of P. nigricans populations is mostly through migration of individuals from the uninvaded areas located upstream of weirs, allowing isolated populations of this vulnerable and endangered species to persist.


Introduction
Padogobius nigricans and Padogobius bonelli are two vicariant freshwater goby species native to Italy.Padogobius nigricans is endemic to the Tyrrhenian drainages of central Italy (Bianco 1995), while P. bonelli is endemic to northern Adriatic basins, from the river Vomano (Italy) to the Krka drainages (Croatia) (Kottelat and Freyof 2007) (Figure 1A).However, in the past two decades, P. bonelli has widely extended its distribution range in the Tyrrhenian basins of central Italy, where it was probably introduced unintentionally during restocking activities with species of fisheries interest from northern Italy (Bianco and Ketmayer 2001).The cooccurrence of P. nigricans and P. bonelli has been reported from several locations in central Italy (Zerunian 2002;Lorenzoni et al. 2006), including the River Tiber basin, the main watercourse of the Italian peninsula.Here P. bonelli was first noticed in when it was found in small tributaries of the River Tiber (Mearelli et al. 1996).From there, the species has colonized the upper and middle course of the River Tiber and has spread into several adjoining tributaries (Lorenzoni et al. 2007).Aggressive behaviour, broad tolerance to environmental alterations (Zerunian 2002) and high energy allocation to reproduction (Pompei et al. 2016a) have favoured the establishment of viable P. bonelli populations in the newly invaded ecosystems.Padogobius bonelli was assessed as a major threat to the survival of P. nigricans, which is listed as "vulnerable" according to the IUCN Red List of Threatened Species (Crivelli 2006) and the Red List of Italian Vertebrates (Rondinini et al. 2013).One of the most negative impacts ascribed to goby invasions is competition with native fishes which is more severe for those species that share similar ecological requirements (Grabowska et al. 2016).Padogobius nigricans and P. bonelli share an analogous ecological niche, a high degree of diet overlap (Pompei et al. 2014) and common reproductive strategies (Zerunian et al. 1988;Mecatti et al. 2010).In both species, females lay eggs on the ceiling of the shelter that is created under stones, and males are responsible for parental care (Mecatti et al. 2010).Padogobius nigricans only exhibits territoriality after laying eggs (Zerunian et al. 1988); in contrast, P. bonelli males defend a specific territory, driving out intruders (Torricelli et al. 1986).There are no differences in timing of the breeding season (Pompei et al. 2016a, b) and where the two species cohabit, P. bonelli can outcompete P. nigricans for spawning sites preventing its reproduction through aggressive behaviour (Mecatti et al. 2010).Interspecific antagonistic interactions may be the main cause of the decline of P. nigricans populations, along with pollution and habitat alterations (Crivelli 2006).
Nevertheless, no studies on interference competition in the field are available and the hypothesis of reproductive failure of P. nigricans caused by P. bonelli is based only on indirect evidence.In some river courses, rapid decline of P. nigricans was observed in coincidence with P. bonelli establishment (Nocita and Zerunian 2007); in other cases, a low number of young-of-the-year P. nigricans were found in the presence of P. bonelli (Pompei et al. 2016b).The main aims of the present study were to update the information on the abundance and distribution of the non-native P. bonelli in the upper River Tiber basin twenty years after its introduction.During monitoring, particular attention was given to sites not yet invaded by P. bonelli (i.e. with "intact" populations of P. nigricans).These sites were usually located upstream of weirs (small dams up to 3 m high that interrupt river continuity) (Pompei et al. 2016b).Thus the possible impact of P. bonelli on the native goby was assessed by comparing the densities and population structure of P. nigricans in invaded and uninvaded systems.We hypothesize that P. bonelli has further spread into new adjacent areas in the past few years, overcoming the native species and, in some cases, replacing it.We expected a reduction in young-of-the-year and an alteration of the population structure of P. nigricans in the presence of P. bonelli due to nesting interference.
We further hypothesize that uninvaded populations of P. nigricans, usually located upstream of small weirs, could support the conservation of the invaded populations downstream by migration of individuals, since downstream drifting of early stages is a common dispersal mechanism for some goby species (Hayden and Miner 2009;Brownscombe and Fox 2012;Janáč et al. 2013).To test this, the population structure of P. nigricans in the invaded sites was analyzed distinguishing between populations downstream of weirs (i.e. with an uninvaded population of P. nigricans upstream) and the other invaded populations located in streams without any weirs.Moreover, to gain insight the mechanism of competition for breeding territory in the field, the occupation of nest sites by the two species with respect to individual density was analyzed.Padogobius bonelli was expected to occupy the majority of the nests sites as well as bigger nests, because nest size is a predictor of male quality (Lindström 1988;Bisazza et al. 1989), and males of both species are likely to compete for larger stones which may allow mating with more than one female.Lastly, nest size and the number of eggs in the nest were analyzed and compared between the two species.

Distribution and abundance
The investigated area is situated in the upper and middle basins of the River Tiber, the third-longest river in Italy.The area was divided into four subbasins: three corresponding to those of the main tributaries of the River Tiber (Chiascio-Topino, Nestore and Paglia rivers), while the main course of the River Tiber and the minor tributaries were considered as a separate sub-basin, henceforth called Tiber (Figure 1B).
Sampling focused mainly in the Tiber, since P. bonelli was not found in the other sub-basins during previous monitoring (Lorenzoni et al. 2010).During 2014-2015 surveys, 43 stations located in the Tiber were sampled.These data have been integrated with the data collected from 34 sampling stations in the period 2011-2014 during the monitoring for the 1 st update of the Regional Fish Map, a project which aimed to monitor the fish fauna of watercourses of Umbria Region (Lorenzoni et al. 2010).Thus, 77 sampling stations in 36 river courses were considered in the analyses (Figure 1B).
Fish sampling was carried out by electrofishing (electroshocker model: ELT62II-GI, direct current, 300 V, 10-100 Hz) using the removal method (Moran 1951;Zippin 1958) in the low flow period (June-September).Sampling was conducted by the same staff to ensure consistent effort among sites.For each river, the length of the stretch to be sampled was defined as 10 times the width of the wetted channel.Length and width of the river sector examined were measured to obtain the area necessary to calculate density (individuals/m 2 ) of the two species.
Specimens were lightly anaesthetized with 2-phenoxyethanol to facilitate handling without harming individuals.For those specimens whose distinction on a morphological basis aroused doubts, a small (< 5 mg) caudal fin piece was collected and stored in absolute alcohol at −20 °C for further molecular identification through a specially designed PCR-RFLP protocol (see below).For each specimen, total length (TL) and weight (W) were measured to the nearest 0.1 cm and 0.1 g, respectively.A sample of scales was removed from dorso-lateral or ventrolateral rows of the caudal peduncle of each individual (Pompei et al. 2014) and stored in 33% ethanol.After the field operations, all individuals were released at the site of capture.
For age determination, fish scales were observed under a stereo microscope and the annuli were counted; two independent age determinations were made by two different operators.An additional age determination was carried out in cases of contrasting results.The microscopic scalimetric method was validated by means of length-frequency distribution (Bagenal and Tesh 1985).
The density of species (D, individuals/m 2 ) was quantified for each sampling site and also expressed as the relative percentage.A Wilcoxon Signed Rank test was then used to assess differences in densities between P. bonelli and P. nigricans in the sites the two species cohabitate (invaded sites).For P. nigricans, the densities found in the invaded and in uninvaded sites were compared by Mann-Whitney U test to test if the presence of the non-native goby had affected the abundance of the native one.
The age composition of P. nigricans populations was analyzed separately for invaded and uninvaded sites.Specimens were grouped into two categories: young-of-the-year (age 0+ years) and adults (age > 1+ years).A Chi-squared test was used to assess differences in the proportion of these two groups in the invaded and uninvaded areas.An analogous comparison was carried out for P. nigricans only in the invaded areas, considering the populations inhabiting the stations with weirs upstream and the populations located in streams without any weir.

Nest occupation
Eight nesting stations were chosen.Six were located in different tributaries of the upper River Tiber (Aggia1, Carpina1, Lanna, Resina, Soara and Vaschi rivers) where both gobies are found, and two (Aggia2 and Carpina2, upstream of the weirs) were located in uninvaded areas (Figure 2).These sites were chosen because they were comparable in terms of environmental characteristics (Lorenzoni et al. 2007): they had predominantly run morphology and the substrate mainly consisted of cobbles and stones.The average area sampled (± SD) was 210 ± 37 m 2 , water temperature was 18.4 ± 2.3 °C and mean depth was 0.38 ± 0.11 m.
Each station was sampled four times.The first surveys were carried out soon after sampling for densities of the two species; the other three nest surveys were conducted weekly in June 2015.Each site was searched for nest sites and every stone large enough to host a nest (approx > 50 cm 2 , Lugli et al. 1990) was turned over.Stones containing batches of eggs were photographed.Because it is impossible to determine the parent species from a nest, several samples of eggs from each nest were collected, put in absolute ethanol and stored at −20°C for molecular identification, following the same protocol as for the fin clips (see below).After the field operations, stones were carefully placed in their original locations, taking care not to damage the eggs.
Once each nest was assigned to the correct species, the number of nests occupied by each species was counted.Then, to check for species dominance, a Chi-Squared test was performed for each station, considering the number of nest of the two species as the observed values, while the expected values were estimated for each species as the number of nest expected on the basis of the relative percentage of the adults found in the same station.
The total surface areas of stones on which batches of eggs were found were measured (stone area) by examining the photos using ImageJ software ver.1.51 (https://imagej.nih.gov/ij/).The number of unhatched eggs in each nest was counted.Nests that contained hatched eggs were excluded (Marconato et al. 1989).
To test the hypothesis that P. bonelli can both occupy the bigger nests and produce more eggs per nest, a comparison between the two species in stone area and number of eggs in the nest was carried out using Wilcoxon Signed Rank tests.These analyses were conducted comparing the two species both within each of the six invaded stations and using the data of all the six stations pooled.
Moreover, to investigate whether the presence of the non-native species could have affected the shelter choice of the native one, the number of eggs in invaded and uninvaded areas were compared using a Mann-Whitney U test.These analyses were conducted using the pooled data of stone area and number of eggs in the invaded and uninvaded areas.A linear regression was used, for each of the two species, to assess whether the number of eggs laid in a nest was correlated with the size of the nest stone.

Distribution and abundance
A total of 2315 gobies (696 P. bonelli, 1619 P. nigricans) were caught at the 77 sites.
Padogobius bonelli was not found in the Nestore and Paglia sub-basins.In contrast, in the Chiascio-Topino basin three specimens of the non-native goby were found during the last survey in the northernmost station of the River Chiascio (Density P. bonelli = 0.006 individuals /m 2 ; Density P. nigricans = 0.24 individuals /m 2 ) (Figure 2B, Supplementary material Table S1).
Four local extinctions of P. nigricans were recorded (Figure 2B).In these sites, the native species was totally replaced by the non-native one.Only 10 uninvaded sites were found.Among them, two were located in the southern part of the Tiber sub-basin and the other eight were located upstream of weirs.In these 10 sites, the density of P. nigricans (median and inter-quartile range of density: 0.90, 0.31-1.77individuals/m 2 ) was significantly higher than the density reached in the 23 sites where P. bonelli is also present (Mann-Whitney U test: n = 43, Z = 3.685, p = 0.021).
When examining the age composition of P. nigricans in the invaded stations downstream of the weirs (i.e. with intact population of P. nigricans upstream) (N = 8) and invaded stations without weirs (N = 15), some interesting considerations emerged.Stations with intact populations of P. nigricans upstream had a more juveniles (N = 16, 16.5% of the total) than invaded stations without weirs (N = 7, 4.7% of the total) (χ 2 = 11.339,df = 1, p < 0.01; Figure 3C, D).

Nest occupation
A total of 313 nest sites were examined in 8 locations.
In rivers Aggia1 and Resina, the majority of the nests were occupied by P. bonelli.In both tributaries only 2 nests out of 77 and 33 nests, respectively, belonged to P. nigricans.In the River Resina the number of nests observed for the two species did not differ from the expected values (χ 2 = 0.185, df = 1, p = 0.999), while in the River Aggia, P. bonelli occupied a greater number of nests than the expected values based on species abundance (χ 2 = 12.501, df = 1, p < 0.01).In River Carpina1, P. bonelli also spawned in most of the nests (Table 1).In Vaschi1 the number of nests observed did not differ from the number of nest expected (χ 2 = 0.008, df = 1, p = 0.99), and the proportion of nests occupied by the two species exactly reflects the relative abundance of individuals (Table 1).Similarly, in Lanna the percentage of the nests are close to those of the individuals.Padogobius nigricans occupied 60% of the nests and accounted for 67% of individuals.In Soara1, P. nigricans occupied a slightly greater percentage of sites compared to P. bonelli, despite the higher abundance of adults (Table 1).
Padogobius nigricans occupied larger nests and produced more eggs per nest in the invaded areas (Table 2), even if the differences between the two species in the mean number of eggs laid were not significant (Table 2).For P. nigricans, no differences in nest stone areas nor the number of eggs per nest were found between invaded and uninvaded areas (Mann-Whitney U test: nest stone areas: n = 117, Z = 1.200, p = 0.230; no.eggs: n = 98, Z = 1.631, p = 0.103) (Table 3).For both species the number of eggs in the nests were positively correlated with stone area (P.bonelli: r 2 = 0.193, r = 0.439, p < 0.01; P. nigricans: r 2 = 0.392, r = 0.731, p < 0.01) (Figure 4).

Discussion
Large numbers of fish species have been introduced into the River Tiber basin in recent decades (Lorenzoni et al. 2006).Some have had damaging effects on native fish fauna (Giannetto et al. 2012).
The establishment of P. bonelli in the upper River Tiber and the further expansion into newly colonized areas has increased concern for the conservation of P. nigricans.Several alarming trends have emerged from the present research, which has implications for the survival of the native populations.The presence of P. bonelli is more concentrated in the upper part of the Tiber basin: its distribution and abundance are closely tied to the distance from the points of first report i.e., to the time-lapse from the first introductions of the species.On average, P. bonelli reached significantly higher densities than P. nigricans in the invaded areas.In the sites with a longer invasion history, it was often the numerically dominant species.Padogobius bonelli has also invaded new river stretches downstream, moving its range edge farther south.The southernmost station in which P. bonelli was found is located approximately 140 km of river length from the site of the first report in 1996.At this station the non-native species was rarest, as observed in invasion patterns of several other alien gobies (Raby et al. 2010;Gutowsky et al. 2011;Brownscombe and Fox 2012).Colonization to the south (downstream) seems to be a result of a natural expanding migration over a continuous area and highlights the ability of P. bonelli to disperse over a considerable distances in a relatively short time.Given their small body sizes (maximum length of males in the River Tiber basin: P. nigricans 9.2 cm; P. bonelli 7.7 cm) and benthic habits, gobies are usually considered species with reduced vagility (Miller 1990;Hayden and Miner 2009;Janáč et al. 2013) and do not exhibit strong swimming abilities (Bergstrom et al. 2008).However several studies have pointed out exceptional dispersal abilities of gobies during range expansion, especially in lakes and in downstream movements in the rivers (Steingraeber and Thiel 2000; Irons et al. 2006;Roche et al. 2013), while rates of expansion upstream were usually much slower (Bergstrom et al. 2008;Brownscombe and Fox 2012).Padogobius bonelli was also found in a newly isolated site of the Chiascio-Topino sub-basin, located more than 70 km from the confluence with the River Tiber.Therefore, the most probable origin of these gobies is an unintentional introduction during illegal stoking with uncontrolled material.Considering the invasion pathways of P. bonelli in the upper Tiber, it is very likely that this location will act as a new spread point which suggests that the non-native goby could spread further downstream, following a "stepping stone and diffusion" dispersal mode.We hypothesize rapid density increases in recently invaded areas due to availability of resources at initial low density and to the high reproductive potential of P. bonelli (Pompei et al. 2016a).This dispersal pattern, both natural and humanmediated, raises further concern for the conservation of the endemic goby.Padogobius bonelli seems to affect P. nigricans population structure, as well as its abundance.In the invaded areas, the presence of fewer juveniles than uninvaded areas supports the notion of reduced P. nigricans reproduction.Nesting interference is common amongst gobies, and several studies have reported that native fishes can be excluded from reproductive territory through aggression from alien gobies (Janssen and Jude 2001;Van Kessel et al. 2011;Kakareko et al. 2013;Jermacz et al. 2014;Grabowska et al. 2016).However, the influence of P. bonelli on shelter occupancy of P. nigricans was as strong as we expected only in some of the examined sites.In other cases, the proportion of nests was more aligned with the densities of the two species, and the prevalence of P. bonelli is less marked.This situation may depend on resource availability.Optimal shelter could potentially be a limiting resource, making any competition between the two gobies much stronger (Błońska et al. 2016).If shelters are limited, it's likely that the more aggressive P. bonelli is a stronger competitor, ultimately excluding P. nigricans from the reproductive habitat.In laboratory conditions, P. bonelli males directly compete with conspecifics for possession of larger nest sites, because females use nest size as a predictor of male quality (Lindström 1988;Bisazza et al. 1989).The nest is a substrate which may physically limit the size of eggs mass in these species that lay eggs in a single layer, and thus, the ability of a male to spawn clutches (Lindström 1988;Marconato et al. 1989).For both species, as in most gobiids (Lindström 1988;Marconato et al. 1989;Takahashi et al. 2001), the number of the eggs in the clutches depends on the lower surface area of the nest, thus it is likely that males compete for larger stones which allow them to mate with more than one female, thereby maximizing reproductive success (Marconato et al. 1989).Moreover, in fish species with exclusive paternal care, females usually favour large nests (Takahashi and Khoda 2002).They prefer to spawn with males whose nests already contain eggs (Forsgren et al. 1996) and large brood sizes may induce more paternal care (Coleman et al. 1985).The preliminary results of this study indicate that P. nigricans occupied on average the larger nests and produced more eggs per nest.Moreover, no differences were found in nest characteristics (size and number of eggs) of P. nigricans between the invaded and uninvaded areas.Thus P. bonelli seems to interfere not with the "quality" of P. nigricans nests, but rather with the "quantity".Moreover, shelter is not only an essential environmental resource during the spawning period for both species, it also plays significant roles in predator protection and refuge from stream flows (Allouche 2002).
Occupying the best sites is apparently not enough for the conservation of P. nigricans.In the invaded areas, the survival of P. nigricans populations is mostly through migration of individuals from uninvaded areas located upstream of weirs that prevented the spread of the non-native species.Passive movement of early ontogenetic stages may represent an important dispersal mechanism in gobies (Janáč et al. 2013).In this sense, small weirs have had a dual task: they facilitate the survival of intact populations of P. nigricans, and allow recolonization of invaded areas, preventing long-term inter-specific exclusion (Belkessam et al. 1997).River damming is one of the most damaging anthropogenic alterations for freshwater environments (Baxter 1977;Dynesius and Nilsson 1994;Franchi et al. 2014).Damming impedes free movement of fish, emphasizing the consequences of isolation (Lorenzoni et al. 2006).Nevertheless, the presence of such barriers substantially contributes to the preservation of isolated populations of vulnerable and endangered species; for this reason actions to restore connectivity between rivers must take into account the risk of spread of invasive species (Van Kessel et al. 2016).

Figure 3 .
Figure 3. Age frequency distribution of 1) Padogobius nigricans considering all the stations of the Tiber sub-basin divided into A) invaded and B) uninvaded areas; and 2) P. nigricans considering only the stations of the Tiber sub-basin invaded by P. bonelli divided into C) stations with weirs upstream and D) stations without weirs upstream.Number of specimens and number of stations considered in each graph are indicated.

Figure 4 .
Figure 4. Relationship between nest stone area (cm 2 ) and the number of eggs in the nests (N eggs) for A) Padogobius nigricans (N = 98) and B) P. bonelli (N = 157) in stations of the upper River Tiber.

Table 1 .
Abundance of Padogobius nigricans (Pn) and P. bonelli (Pb) in the 37 sites of the Tiber sub-basin sampled during 2014-2015 expressed as density (D ind m -2 ) and relative abundance (%), and number and percentage of nests of the two species collected in four sampling dates in the 8 sites investigated in 2015 for the analysis of nest occupation.The results of comparisons by Chi-squared test (χ 2 ) between the number of the nests and the abundances of the two species are provided.

Table 2 .
Descriptive statistics for nest stones area (cm 2 ) and number of eggs in the nest (N eggs) for Padogobius bonelli (Pb) and P. nigricans (Pn) in the 6 invaded sites investigated.The results of the comparison (Wilcoxon Signed Rank test) between the two species for each site and for the total sample (Total) is reported.

Table 3 .
Descriptive statistics of nest stone areas (cm 2 ) and number of eggs in the nest (N eggs) for Padogobius nigricans in the 2 uninvaded sampling stations.The results of comparison (Mann-Whitney U test) between pooled data for the invaded (Table2) and uninvaded areas are shown.