Dietary Niche and Growth Rate of the Nonnative Tubenose Goby (Proterorhinus semilunaris) in the Lake Superior Basin

Abstract

The tubenose goby (Proterorhinus semilunaris) entered the Great Lakes in the 1990s via ballast water, but remains poorly studied within North America, making it difficult to predict its effects on native ecosystems. Dietary breadth and somatic growth rate have important ramifications for survival, competitiveness, and dispersal ability of a fish species, and thereby its ecological impact. We studied diet and growth of age-0 tubenose goby within the St. Louis River, a tributary to Lake Superior that contains the largest population within the Lake Superior basin. We sampled tubenose gobies from shallow, vegetated habitat during summer and fall. Stomach contents were identified and weighed to measure fullness and dietary breadth between seasons and several locations. We aged fish based on otolith daily increments to model somatic growth. Diet was dominated by isopods and amphipods, and dietary breadth was low and not significantly different between locations and seasons. Tubenose goby diet strongly overlapped with that of tadpole madtom (Noturus gyrinus), a native, demersal species. We tested several candidate growth models; the Gompertz Growth Function was the most parsimonious model among those examined. The model demonstrates that tubenose goby obtains a small maximum size and is short-lived. We conclude that tubenose goby presents a unique risk to the Great Lakes and other freshwater bodies because their life history is typical of invasive species, their diet overlaps with native fish, and because they occupy shallow, vegetated habitat which functions as both nursery and foraging habitat for many native fishes.

Introduction

Invasive species are among the most significant ecological threats in recent history (Burr et al., 1996; Kohler and Courtenay, 1986), particularly in the Laurentian Great Lakes (hereafter, Great Lakes). As of 2016, over 180 aquatic nonnative species have been documented in the Great Lakes, and the introduction rate is estimated as 1.3–1.8 new species per year (Sturtevant et al., 2016). Many of these nonnative species have caused extensive economic and ecological damage, such as declines in biodiversity and shifts in primary productivity (Mills et al., 1993; Pagnucco et al., 2015; Vanderploeg et al., 2002). The Great Lakes display high connectivity and have multiple vectors of introduction for new species, such as through ballast water from international shipping vessels or via the Saint Lawrence Seaway (Mills et al., 1993; O’Malia et al., 2018). Among the most recent nonnative fishes to be introduced to the Great Lakes is tubenose goby (Proterorhinus semilunaris), a small (<130 mm) demersal fish from the Black and Caspian Seas of Europe.

Tubenose goby is one of 24 neogobiine species from the Ponto-Caspian region; the subfamily includes four genera (Neilson and Stepien, 2009a). Several other neogobiine fish – monkey goby (Neogobius fluviatilis), racer goby (N. gymnotrachelus), bighead goby (N. kessleri), and round goby (N. melanostomus) – have successfully invaded freshwater systems in eastern and central Europe. To date, only round goby and tubenose goby have established in the Great Lakes. Round goby demonstrates a generalist diet and rapid rate of growth (Johnson et al., 2008; Kornis et al., 2017); furthermore, it reproduces quickly and interacts aggressively with other species (Bergstrom and Mensinger, 2009). These traits have been attributed as primary causes for its success (MacInnis and Corkum, 2000), and it competes with several native, demersal species for food and habitat resources, particularly mottled sculpin (Cottus bairdi), logperch (Percina caprodes), johnny darter (Etheostoma nigrum) and northern madtom (Noturus stigmosus) (French and Jude, 2001; Lauer et al. 2004). Round goby also predates upon native fauna (Jude et al., 1995) and consumes the eggs of native species, including smallmouth bass (Micropterus dolomieu), walleye (Sander vitreus), and lake sturgeon (Acipenser fulvescens) (Corkum et al., 2004; Steinhart et al., 2004; Roseman et al., 2006).

Tubenose goby was first discovered in 1990 in Lake St. Clair (Jude et al. 1992) and subsequently spread to Lake Erie, Lake Superior, Lake Huron, and Lake Ontario (United States Geological Survey, 2018). It likely was transported from the lower Great Lakes to Lake Superior by shipping vessels (Dopazo et al. 2008; Hensler and Jude, 2007). Like round goby, tubenose goby larvae exhibit diel vertical migration, which facilitates uptake by shipping vessels into ballast water for subsequent dispersal (Hensler and Jude, 2007; Janáč et al., 2013). A resident population exists at the mouth of the St. Louis River (Grant et al., 2012; Kocovsky et al., 2011), a tributary to Lake Superior at its southwest corner. The St. Louis River population is the largest in the Lake Superior basin and possesses low genetic variability (Neilson and Stepien, 2009b), indicating this population likely was introduced from a single event (Kocovsky et al., 2011). Among tubenose goby populations throughout the Great Lakes, however, many exhibit high genetic diversity, suggesting multiple founding sources (Stepien and Tumeo, 2006).

Wide dietary breadth is a characteristic of most invasive vertebrates (Sakai et al., 2001), and is a top predictor of success and range during the establishment phase of invasive fish species (García-Berthou, 2007; Pettitt-Wade et al., 2015). Predation by invasive fish can also have cascading trophic consequences, leading to the decrease of benthic invertebrates or zooplankton (Gallardo et al., 2015; Lederer et al., 2008) and result in competition with native fishes (Pyke, 2008; Strayer, 2010). Generally, Ponto-Caspian gobiids (racer goby, monkey goby, and round goby) have a broad diet and generalist feeding strategy in invaded areas of Central Europe (Grabowska and Przybylski, 2015). A broad diet and aggressive feeding strategy are also key aspects of successful round goby dispersal in the Great Lakes (Leino and Mensinger, 2016). Studies characterizing the diet of tubenose goby are primarily from European waterbodies. Adámek et al. (2010) found that tubenose goby diet was dominated by Chironomidae larvae and Isopoda, with minor representation of various zooplankton. Other invertebrate groups, including insects (Trichoptera, Ephemeroptera), are of varying importance and tubenose goby diet is generally characterized as opportunistic (Adámek et al., 2010; Vašek et al., 2014; Všetičková et al., 2014). Within the Great Lakes, tubenose goby consume small insects and crustaceans (Kocovsky et al., 2011) as well as round goby eggs (French and Jude, 2001). Tubenose goby exhibit diet plasticity between seasons and locations, but fish eggs have not been identified as a significant diet component (Vašek et al., 2014; Všetičková et al., 2014). Their diet overlaps with rainbow darter (Etheostoma caeruleum) and northern madtom (French and Jude, 2001); tubenose goby may also compete with mottled sculpin, logperch, and johnny darter for food resources. No diet studies have been conducted for this species within Lake Superior, although isotopic research has suggested that tubenose goby shows dependence on specific prey species elsewhere within the Great Lakes (Pettitt-Wade et al., 2015).

In addition to diet, rapid somatic growth rate is another important factor for the successful establishment of nonnative fishes in the Great Lakes (García-Berthou, 2007) and is characteristic of most successful invasive species (Sakai et al., 2001). Despite extensive research examining the growth rates for the closely-related round goby (Gruľa et al., 2012; MacInnis and Corkum, 2000; Thompson and Simon, 2015), to date there have been no growth rate estimates for tubenose goby. In European waterbodies, researchers examined tubenose goby cohort structure; however, these analyses were based on extrapolating age from size, in which age estimates of 1+ year were assigned to tubenose gobies >40 mm standard length (Janáč et al., 2012; Všetičková et al., 2014). Estimates of maximum age vary, ranging from 2 years (Všetičková et al., 2014) to 4 or 5 years (Leslie et al., 2002).

The purpose of this study was to characterize tubenose goby diet composition and growth in an invaded waterbody. We hypothesized that tubenose goby feeds opportunistically and has a high likelihood of dietary competition with native fishes. Furthermore, we hypothesized that tubenose goby would display high spatial and seasonal diet plasticity. Finally, we hypothesized that tubenose goby would demonstrate rapid somatic growth and small size (with corresponding rapid maturity), which are traits often found in successful invasive species (Sakai et al., 2001). To test our hypotheses, we regularly sampled tubenose goby and associated native fishes from multiple coastal wetlands within the St. Louis River, which is within the Lake Superior basin. We subsequently analyzed stomach contents of tubenose goby and an abundant, native, demersal species - tadpole madtom (Noturus gyrinus) - and measured tubenose goby growth rate based on daily otolith increments.

Methods

Study and Site Locations

The St. Louis River estuary (SLRE) is a drowned river mouth - coastal wetland complex that is situated at the mouth of the second-largest tributary to Lake Superior. Within the SLRE is the Duluth-Superior harbor, the largest port by cargo tonnage in the Great Lakes (O’Malia et al., 2018). The SLRE is home to over 40 resident fish species (Peterson et al., 2011) and is also an “invasion hotspot” due to a high amount of shipping traffic (Grigorovich et al., 2003). In 2011, it was home to at least 10 non-native species of fish (Peterson et al., 2011). The SLRE is over 4,850 hectares in size and contains many wetlands that provide essential nursery areas for a variety of fish species (Breneman et al., 2000; Hoffman et al., 2010; Peterson et al., 2011). Habitats that occur within the estuary include shallow flats (<3 m; either vegetated or unvegetated), as well as dredged and natural channels (Leino and Mensinger, 2016). Bottom substrates are both soft (clay, sand, and organic material) and hard (rip-rap and sheet-pile). In the lower estuary, around the harbor, the shoreline is extensively developed, whereas in the upper estuary, much of the shoreline remains unmodified (Leino and Mensinger, 2016). Clarity is generally low throughout the estuary owing to high concentrations of colored dissolved organic matter and local inputs and resuspension of clay (Peterson et al., 2011).

We selected several locations within the SLRE for preliminary sampling based on habitat and prior occurrence data (Ramage, 2017). Tubenose goby were first discovered in the SLRE in 2001 (United States Geological Survey, 2018). By 2006, they had become abundant and widespread in littoral habitat (Peterson et al., 2011), where they dominate young-of-year fish assemblages (Ramage, 2017). Tubenose goby have been shown to associate strongly with shallow areas of abundant and diverse submergent vegetation within the SLRE (Ramage, 2017). Based on preliminary sampling, we selected two locations, Indian Point Campground (IPC) and Lower Pokegama (LPK), for our study where we found consistent occurrence of tubenose goby. IPC and LPK are both sheltered bays that are influenced by tributaries (Kingsbury Creek and Little Pokegama River, respectively), have substrates composed of organic material and clay, and have shallow (<3 m) heavily-vegetated flats dominated by yellow pond lily (Nuphar variegata), coontail (Ceratophyllum demersum), pondweed (Potamogeton spp.), milfoil (Myriophyllum spp.), water celery (Vallisneria americana), and waterweed (Elodea spp.).

Field Sampling

We sampled fish by beach seine (7.6 m long by 1.2 m tall, 0.48 cm mesh; bag: 1.2 m deep by 1.8 m wide by 1.2 m tall). We sampled 20 m transects in shallow (<1 m) flats with dense, submergent vegetation for sampling because this habitat has a high density of tubenose goby (Ramage, 2017). We used an active sampling gear to minimize digestion of prey items, but recognize that beach seine efficiency can be affected by sampling in vegetation; past research has shown increased capture efficiencies to be associated with higher macrophytic biomass (Pierce et al., 1990). Summer sampling (July 30, 2018 - September 7, 2018) began after the emergence of visible aquatic vegetation and concluded prior to the first frost; fall sampling (October 1, 2018 - October 17, 2018) occurred after the first frost of the year and concluded prior to complete senescence of aquatic vegetation. We sampled each of our two sites on three different dates within each season, and during the morning for diet consistency (08:00–12:00). On each sampling date, our goal was to collect a minimum of 15 individuals per site; to accomplish this, we generally sampled 3–8 transects per site (with the exception of one sampling date and site, for which only 1 transect was necessary to obtain sufficient specimens). We allowed at least five days to pass between sampling dates at the same site to allow for fish recolonization and to ensure diet independence among sampling events. We sorted and tallied seine contents by species (sunfish, Lepomis spp., were not differentiated); tubenose goby and tadpole madtom were euthanized via MS-222 and retained. Non-benthic species were not retained due to differing foraging preferences. We made a small incision in the body cavity of each euthanized fish and preserved specimens in 95% ethanol (EtOH). We did not collect mottled sculpin (n = 0), johnny darter (n = 1), logperch (n = 0), or round goby (n = 0) in sufficient numbers for dietary comparisons.

Water characteristics and habitat data were collected for each transect. At each location, at a depth of 0.6 m, we deployed one HOBO Pendant (UA-002–08) logger to continuously measure surface water temperature (hourly). We collected surface water quality and habitat data at the beginning of each transect. Turbidity (NTU), pH, dissolved oxygen (mg/L and % saturation), temperature (°C), and specific conductivity (μS/cm) were measured at 0.5 m depth (Hydrolab HL4 multiparameter sonde, OTT Hydromet, Loveland, CO; calibrated prior to sampling). At each transect, visible aquatic vegetation cover was assigned a value ranging from 0 (no vegetation present) to 4 (dense; plants cover most of the sampling area), and we noted dominant taxa present. We sampled substrate via PONAR dredge and assigned a primary class from gravel, sand, silt, clay or organic composition.

Gut Content Processing

In the laboratory, we measured whole preserved fish total length (TL; ± 0.5 mm) and wet weight (± 0.001 g). All tubenose gobies >20 mm TL (n = 143) were included for diet and growth analyses and all tadpole madtoms (n = 36) were included for diet analyses. For both tubenose goby and tadpole madtom diet analysis, we removed and weighed each cardiac stomach (± 0.001 g) prior to and following stomach content removal, and calculated eviscerated fish weight by subtracting the weight of the full stomach from that of the undissected fish. Prior to weighing either fish or stomachs, we removed surface water by blotting and air drying (30 s.). We calculated the wet weight of stomach contents for each fish by subtracting the weight of the empty stomach from that of the undissected stomach. To account for length and weight preservation bias (Leslie and Moore, 1986), we conducted a preservation experiment: ten tubenose gobies (28.0–59.0 mm TL) were collected and measured for total length (± 0.5 mm) and weight (± 0.001 g) prior to preservation, then preserved in 95% EtOH for 32 days, and remeasured. Tubenose gobies retained 95.68% (± 1.44% SD) of their original length and 67.38% (± 4.29% SD) of original weight after preservation. Calculated fresh length and weight data were used for growth analyses, while preserved length and weight were used for the Gut Fullness Index (described below). We determined sex of tubenose gobies by visual gonadal examination (n = 142; Guellard et al., 2015; Valová et al., 2015).

We assessed diet composition proportion by dry weight. Stomach contents of tubenose goby or tadpole madtom were composited for analysis to create a single site- and date-specific sample; the number of individuals per sample ranged from 7 to 15. We sorted the stomach contents into 11 groups, consistent with European literature (Všetičková et al., 2014): Annelida, Cladocera/Copepoda, benthic macrocrustaceans (hereafter Amphipoda, Isopoda, and Ostracoda), Ephemeroptera, Trichoptera, Chironomidae larvae, Chironomidae pupae, other Diptera, terrestrial insects, fish eggs and fry, ‘other’ invertebrates, and detritus (not included in analysis). Composited stomach contents were dried at 55°C for 24 hours prior to weighing (± 0.001 mg). To evaluate the thoroughness of diet characterization, we examined prey accumulation curves for both tadpole madtom (n = 36) and a subset of tubenose gobies (n = 98). We used EstimateS software (Colwell, 2013) to estimate Chao 1, a nonparametric, abundance-based species richness estimator, and its associated 95% confidence interval (CI). A small CI (relative to Chao 1) indicates adequate sampling of common diet items (i.e., the species accumulation curve is approaching an asymptote).

Fragments of isopod and amphipods (the dominant prey items) were DNA-sequenced to determine species. DNA was extracted using a DNeasy Blood and Tissue Kit and eluted with 100 μL of sterile water. Mitochondrial DNA genetic marker COI locus (B/R5) was PCR amplified with forward and reverse primers; PCR thermal cycling was then applied, and products were dual indexed and then metabarcoded on the Illumina Miseq sequencing platform (US EPA Office of Research and Development laboratory, Cincinnati, OH). Species identification was determined by matching unknown sequences to public reference sequences in the Barcode of Life Database (http://www.boldsystems.org/; Oct. 24, 2019), with an identification threshold of ≥98% match.

Gut Content Analysis

For each fish, we calculated a gut fullness index (GFI): GFI = 10⁴ * (w/W_evi), the ratio between preserved wet diet weight (w; calculated by subtracting the weight of the empty gut lining from the weight of the full gut) and eviscerated preserved fish weight (W_evi; calculated by subtracting the weight of the full gut from the weight of the undissected fish). To account for unequal variance, Welch’s robust two-sample t-test was used to compare GFI between groups (species, season, site, size). Fish with empty guts were included in the analysis (TNG n = 5, MAD n = 3). We first used a standard power transformation of GFI by 0.6 to meet assumptions of approximately normal distribution (Shapiro-Wilk p > 0.05). For analysis, we sorted tubenose goby into two size categories based on existing literature: either <50 mm TL (small) or ≥50.0 mm TL (large; Janáč et al., 2012; Valová et al., 2015; Všetičková et al., 2014). For all tests, an α = 0.05 was used.

Based on date- and location-specific composite data, we calculated percent representation of each dietary item (%W_i) as a relative percent dry weight (Hyslop, 1980): %W_i = 100 * (W_i/∑W_i), where W_i is the weight of diet item i from all tracts and ∑W_i represents the total weight of all dietary items. Using the same data, we used Levins’ Index (B) to quantify dietary niche breadth (Levins, 1968): B = 1/∑p_i², where p_i represents the proportion of fish diet (pooled for each sampled date-site combination) of diet category i. This index produces a value between 1 and n (for this study, n = 11 categories). Low values indicate a low level of dietary generalism (diet being dominated by few prey categories), whereas high values indicate a high level of generalism and a more diverse diet. We tested for differences in B between groups (site and season) with a nonparametric Kruskal-Wallis test to account for unequal variance. Finally, we used Schoener’s Index (S) to quantify dietary overlap between species, seasons, and locations (Wallace, 1981): S = 1 – 0.5 ∑ |p_xi – p_yi|, where p_xi and p_yi represent the proportions by weight of prey category i in the diets of species x and y. Values range from 0 (complete dissimilarity) to 1 (identical composition of diet). Values of S > 0.6 are considered to indicate biological significance and likely food competition between species (Vašek et al., 2014; Wallace, 1981) or, alternatively, high similarity of diets between sites or seasons.

Size, Age, Growth, and Mortality

We used a power function to model the length-weight relationship for tubenose goby: W = a(L^b)(exp^ɛ), where W = average weight, a and b are constants, L = average length, and exp^ε is the multiplicative error term, which was used because we found increasing variance in weight with respect to an increase in length. We also transformed this model to a linear form.

We removed sagittae and lapilli pairs of otoliths from tubenose goby for aging by daily increments. Round goby and tubenose goby are closely related species with similar otolith structure (Neilson and Stepien, 2009b), and the use of daily increments via age-0 round goby has been validated (MacInnis, 1997). We used lapilli for aging because they provided more discernible increments than sagittal otoliths, which concurred with the findings of MacInnis and Corkum (2000) with respect to round goby. We polished and examined otoliths from a random subset of tubenose goby (n = 58) that encompassed the size range of fish sampled. We affixed lapilli to a glass slide, polished them with lapping paper (0.01–3.0 micron) to reveal daily increments, coated the polished otolith in immersion oil, and photographed it under transmitted light microscopy for aging. Two readers independently assigned an age to each otolith by counting increments from the nucleus to the outer edge. If the ages agreed within 5 days, the estimates were retained and averaged to assign a final age (n = 22); those individuals that differed >5 days (n = 36) were read again to determine an age by consensus (within 5 days), and then similarly averaged for age assignment. Fish that could not be reconciled between readers (n = 6) were removed from the dataset, and the remainder used for the analysis (n = 52).

The von Bertalanffy Growth Function (VBGF) and the Gompertz Growth Function are two of the most commonly used models for illustrating somatic fish growth, and both functions utilize length-at-age data. The VBGF curve typically exhibits an initial period of fast growth that approaches an asymptote; although extremely popular for illustrating fish growth, this growth function is most often used for adult fish. Therefore, it was also compared to Gompertz and linear regression functions. The Gompertz Growth Function is sigmoid shaped and often used to describe larval or juvenile fish growth; the produced curve is suited for fish that demonstrate low initial growth rates. We also utilized a simple linear regression to model growth, as has been done elsewhere to model larval fish growth that may have not yet reached an inflection point (Steinhart et al., 2004; Oyadomari and Auer, 2008). Owing to the absence of age-1 fish, models should be viewed as illustrating growth during the period in which fish were captured, and parameters should be interpreted cautiously outside of this period.

We fit both the von Bertalanffy Growth Function and the Gompertz Growth Function to model tubenose goby length-at-age using the FSA and nlstools packages in Rstudio Version 1.1.463 (Ogle et al., 2018). Residual plots suggested an increase in variability with increasing fitted values (i.e., growth depensation, or heteroscedasticity). Therefore, we used a multiplicative error structure for the growth models. We obtained starting values for the VBGF using polynomial regression and then applied them via nonlinear least squares to fit the model. We then used typical nonlinear estimation techniques with the data on the natural log scale to estimate model parameters: $L_t=L_{\infty }\left(1-ex{p}^{-K\left(t-t_0\right)}\right)ex{p}^{\epsilon }$, where L_t = average length-at-age t (days), L_∞ = asymptotic average length, K = the exponential rate of approach to the asymptotic length (“Brody Growth coefficient”), t₀ = a model artifact that represents the “average age when the average length is equal to zero,” and ε = the model error. To fit the Gompertz Growth Function, we found starting values by iteratively superimposing a curve of the function at chosen parameter values onto a scatterplot of the data, and then applied these using nonlinear least squares to parameterize the model. We used typical nonlinear estimation techniques with the data on the natural log scale to estimate model parameters: $L_t=L_{\infty }exp\left(-ex{p}^{-g_i\left(t-t_i\right)}\right)ex{p}^{\epsilon }$, where L_t = average length-at-age t (days), L_∞ = asymptotic average length, g_i = the instantaneous growth rate at the inflection point, t_i = the age (days) at the inflection point, and ε = the model error. We calculated 95% confidence intervals via bootstrapping (1000 iterations). Given the duration of the study and the nature of only capturing age-0 fish, maximum size is difficult to estimate and L_∞ should thus be treated with discretion for both models.

We used a simple linear regression to estimate daily instantaneous growth, as has been used elsewhere for age-0 fish (Steinhart et al., 2004; Oyadomari and Auer, 2008): L_t = (age at hatching) + G * (days) ε, where L_t = average length-at-time and G = overall daily growth rate (mm/day). To account for heteroscedasticity, we performed a simple linear regression on the natural log-transformed data: ln(L_t) = ln (a) + b * ln (days) + ε, where L_t = average length-at-time (days), ln(a) = the intercept, and b = slope.

We allowed the three models (VBGF, Gompertz, linear) to compete using Akaike’s Information Criterion (AIC) in RStudio (Version 1.1.463); this model selection approach balances the goodness of fit with model simplicity for selection of the most favorable model, indicated by the lowest AIC value.

We estimated tubenose goby instantaneous mortality by catch curve analysis. For each fish with length data (n = 209), we estimated its age using the Gompertz growth function, and then summed the catch data by five-day age bins. We calculated instantaneous total mortality (Z) by applying linear regression to all age classes older than and including the age with the maximum catch (Ogle, 2016): ln(C_t) = ln(vN₀) – Zt, where C_t = catch-at-age, v = a constant proportion of the population that is “vulnerable” to the fishery, N₀ = initial population size, and Z = instantaneous total mortality. We converted Z to a daily mortality rate (D) via the equation D = 1 – exp^−Z; daily survival rate was calculated as 1 – D.

Hatch dates were estimated from the subset of 52 aged fish. Age was subtracted from the capture date to determine the hatch date for each fish. The instantaneous mortality rate was applied to estimate frequencies for the entire year class. This mortality-corrected frequency was corrected for effort by dividing by the number of seines pulled on that capture-date. Hatch date frequencies were subsequently binned by five-day increments and converted to percent frequency. This provided an illustration of hatch dates from aged fish that was corrected for both mortality and effort.

Results

Fish Assemblage

We captured a total of 3,592 fish at the two sites, of which 216 were tubenose goby (5.99% of total fish; Fig. 1). Other common fish groups included sunfishes (bluegill, pumpkinseed) and black crappie (81.49% combined), golden shiner (3.48%), yellow perch (3.06%), and spottail shiner (2.53%). Tubenose goby was the second most common fish and the most common demersal species. Tadpole madtom comprised 1.03% of total fish (n = 37; one individual was not used for diet analysis due to damage while sampling), while only one johnny darter and one bullhead (Ameiurus sps.) were sampled. Other sparsely represented fishes (all <1.5%) included rock bass (Ambloplites rupestris), largemouth bass (Micropterus salmoides), northern pike (Esox lucius), walleye (Sander vitreus), common carp (Cyprinus carpio), and white sucker (Catostomus commersonii). No round gobies were captured.

Figure 1.

Fish community per season at Indian Point Campground (IPC) and Lower Pokegama (LPK). Percentages were calculated by number of individuals. Major groups included sunfish and crappie (Sunfish), golden shiner (GOS), tadpole madtom (TAM), tubenose goby (TNG), yellow perch (YEP), and Other (rock bass, largemouth bass, northern pike, walleye, spottail shiner, common carp, Johnny darter, and bullhead spp).

Diet Between Species, Seasons, and Sites

The most common prey item consumed by tubenose goby was benthic macrocrustacea, which composed 94.62% of the diet by dry weight (Table 1). We analyzed a subset of samples (n = 5 that included 58 fish) and found that among the benthic macrocrustacea, Isopoda comprised 65.07% of the dry weight, and Amphipoda and Ostracoda comprised 32.44% and 0.06% respectively. Based on CO1 sequencing of a subsample of benthic macrocrustacea tissue from stomach samples, we identified the amphipod species Gammarus fasciatus, as well as the isopod genus Caecidotea. Ephemeroptera, Trichoptera, Chironomidae larvae, and Cladocera/Copepoda were also present, although at lower percentages than benthic macrocrustacea. Chironomidae pupae (1 individual; 0.15% of total diet weight) and other Diptera (1 individual; 0.16% of total diet weight) were the least commonly consumed prey items, whereas fish eggs or fry, terrestrial insects, Annelida, and other invertebrates were not found. Detritus (sand, wood) was not included in diet proportions; detritus was uncommon and composed only 0.24% of total stomach material. The Levins’ Index (B) was 1.12 (of a maximum of 11), and average Gut Fullness Index was 194.41 (range: 0.00–656.93, Table 2). We did find a significant difference between large $\left(\widehat{\mathrm{y}}=160.50\right)$ and small $(\widehat{\mathrm{y}}=205.30)$ tubenose goby GFI (Welch’s t-test; p = 0.03, df = 140.83).

Table 1.

Diet composition (% dry weight) for prey categories by season, site (IPC = Indian Point Campground, LPK = Lower Pokegama), and species (TNG = Tubenose Goby, MAD = Tadpole Madtom). “Total” of each species refers to combined composite samples from each species. Detritus (sand, wood) was relatively minor, considered incidentally ingested, and not included in analysis.

Prey	Summer TNG	Fall TNG	IPC TNG	LPK TNG	Total TNG	Total MAD
Benthic Macrocrustacea	86.32	97.65	93.92	95.29	94.62	84.36
Copepoda/Cladocera	3.48	0.36	1.78	0.63	1.19	6.26
Ephemeroptera	5.06	0.82	1.45	2.44	1.96	1.52
Trichoptera	3.10	0.56	1.76	0.74	1.24	3.84
Chironomidae Larvae	1.43	0.40	0.79	0.58	0.68	3.93
Chironomidae Pupae	0.00	0.20	0.30	0.00	0.15	0.08
Other Diptera	0.61	0.00	0.00	0.32	0.16	0.00

Table 2.

Levins’ Index (B; mean ± SD), Schoener’s Index (S), Gut Fullness Index (GFI) and transformed Gut Fullness Index (GFI^0.6) for tadpole madtom and tubenose goby (TNG). P-values for Levin’s Index were calculated from replicates of site-date samples (n = 6 per category per site and season). Sites included Indian Point Campground (IPC) and Lower Pokegama (LPK). GFI (Q₂ = median, Q₁ = 1^st quartile, Q₃ = 3^rd quartile) was calculated for each fish as a proportion of total gut content weight to eviscerated fish weight (EtOH preserved; wet weight; ± 0.001 grams). Empty guts were included in GFI analysis & detritus was not included in diet weights.

		Tubenose Goby	Tadpole Madtom	Summer TNG	Fall TNG	IPC TNG	LPK TNG
Levins’ (B)		1.12	1.39	1.36±0.22	1.05±0.07	1.25±0.20	1.16±0.26
B p-value		insufficient n		0.01		0.26
Schoener’s (S)		0.891		0.885		0.972
GFI	Q2	162.8	218.4	203.4	154.0	159.6	165.7
	Q1	88.5	93.6	94.6	88.0	85.4	102.1
	Q3	271.3	520.8	346.8	191.4	229.0	292.7
GFI0.6	Q2	21.2	25.3	24.3	20.5	21.0	21.5
	Q1	14.7	15.2	15.3	14.7	14.4	16.0
	Q3	28.9	42.7	33.4	23.4	26.1	30.2
GFI0.6 p-value		0.08		0.02		0.17
n fish		144	36	68	76	66	78

Tadpole madtom displayed similar diet proportions as tubenose goby: benthic macrocrustacea composed the majority of the diet (84.36%), while Ephemeroptera, Trichoptera, Chironomidae larvae, and Cladocera/Copepoda all contributed smaller amounts (Table 1). The Levins’ Index was 1.39, indicating tadpole madtom has a broader diet niche than tubenose goby, but there were too few samples to test for a difference. Diet overlap between tadpole madtom and tubenose goby was high (S = 0.891), and the average GFI was not significantly different between species (Welch’s t-test; p = 0.08, df = 47.79). The prey accumulation curves for both species approached an asymptote of 6 prey taxa with relatively small associated 95% Upper Boundary Confidence Intervals (goby = 7.58, tadpole madtom = 7.14), indicating adequate sampling of common prey.

Tubenose goby summer and fall diets were primarily composed of benthic macrorustacea, and Schoener’s Index indicated a high degree of seasonal overlap (S = 0.885; Table 2). However, benthic macrorustacea increased from 86.32% of diet in summer to 97.65% in fall, with an accompanying decrease in percentage of other prey categories (Table 1). Levins’ Index differed between summer (B = 1.36 ± 0.22, n = 6) and fall (B = 1.05 ± 0.07, n = 6; Kruskal-Wallis test: p = 0.01, df = 1), indicating a greater dietary breadth in the summer. Furthermore, GFI decreased from summer to fall (Welch’s t-test; p = 0.02, df = 121.93).

Indian Point Campground (IPC) and Lower Pokegama (LPK) displayed high similarity and overlap in diet between sites (S = 0.972). Benthic macrocrustacea composed >93% of the diet at both locations (Table 1). Levins’ Index values did not differ between IPC (B = 1.25 ± 0.20, n = 6) and LPK (B = 1.16 ± 0.26, n = 6; Kruskal-Wallis test: p = 0.2, df = 1). The average GFI was not significantly different between IPC $(\widehat{\mathrm{y}}=174.26)$ and LPK ($\widehat{\mathrm{y}}=211.50$; Welch’s ttest: p = 0.17, df = 141.4).

Demographics, Growth, and Mortality

Tubenose goby total length ranged from 11.0 mm to 74.0 mm, with a median size of 39.0 mm TL (n = 209; Table 3). No fish were found that were aged one year or older. Of the fish with visible gonads, 59.86% were female and 40.14% were male (Fig. 2). Average size was significantly different between summer ($\widehat{\mathrm{y}}=30.6\pm 10.7$; n = 133) and fall ($\widehat{\mathrm{y}}=51.9\pm 7.7$; n = 76) via Welch’s t-test (p < 0.001, df = 196.07). Fresh wet weight ranged from 0.012–4.711 g with a median of 0.530 g. Age of fish >20 mm ranged from 34 to 111 days (n = 52). Based on the hatch date analysis (Fig. 3), peak hatch occurred during mid-June, with a median hatch date of June 26. The hatch dates spanned nearly two months, from June 1 to July 26.

Table 3.

Mean ± standard deviation (SD) fresh total length (TL), wet weight, total number of tubenose goby (n; tubenose goby = TNG) sampled, and number of individuals that were used for sex, size (small = 20–50 mm TL, large = >50 mm TL), and site (IPC = Indian Point Campground, LPK = Lower Pokegama).

Season	TL (mm)	Weight (g)	TNG	Sex		Size		Site
				Male	Female	Small	Large	IPC	LPK
Summer	30.6 ±10.7 (n = 133)	0.36 ±0.46 (n = 133)	133	36	30	57	9	26	42
Fall	51.9 ±7.7 (n = 76)	1.53 ±0.92 (n = 76)	83	21	55	40	36	40	36
Total	38.3 ±14.1 (n = 209)	0.79±0.87 (n = 209)	216	57	85	97	45	66	78

Figure 2.

Size (total length) frequency of tubenose gobies captured in summer (top; n = 66) and fall (bottom; n = 76). Only fish that were sexed (>20 mm) were used for analysis.

Figure 3.

Hatch dates by percent for tubenose goby based on a subset of gobies (n = 52). Hatch dates are corrected for mortality and effort. Median adjusted hatch date was June 17 and ranged from June 1-July 26, 2018.

Fresh total length (± 0.5 mm) was strongly correlated with wet weight (± 0.001 g) of tubenose goby (n = 209). We found a significant linear relationship (df = 205, r² = 0.99, p < 0.001) described by ln (W) = −12.724 (± 0.08 SE) + 3.297 (± 0.023 SE) ln(L) + ε, indicating allometric growth.

The best-fit von Bertalanffy Growth Function is L_t = 109.90 (1 – exp ^{−0.01(t – 13.72)}) exp ^ε. 95% confidence limits for each coefficient include: asymptotic length (L_∞ = 77.72–641.95), Brody growth coefficient (K = 0.001–0.02), and age-at-length zero (t₀ = 0.77–21.30).

The best-fit Gompertz Growth Function is L_t = 78.84exp (−exp ^{–0.03(t – 43.77)}) exp ^ε. 95% Confidence limits for each coefficient include: asymptotic length (67.97–101.58), rate of maturity per day (0.02–0.04), and the number of days of maximal growth rate (39.00–54.62).

The simple linear regression of overall daily growth rate (Oyadomari and Auer, 2008) is L_t = 4.06 + 0.618t where L_t = length-at-age (days), 4.06 = size at hatching (mm), 0.618 = the overall daily growth rate (mm/day), and t = the age of the fish in days (r² = 0.84; n = 52). However, because of heteroscedasticity within the length-age data, a linear regression was also performed on the natural log-transformed data to account for unequal variance, and parameterized as ln(L_t) = 2.82 + 0.014 ln(t) + ɛ where L_t = length at time (days), 2.82 = the intercept, and 0.014 = the slope (r² = 0.82; n = 52).

The three natural log-transformed models were compared using Akaike’s Information Criterion (AIC) and were selected in the order of Gompertz (−75.83), VBGF (−72.28), and natural log-linear regression (−49.75).

The catch-curve analysis yielded an instantaneous mortality rate of 0.036 (±0.003 SE), or daily mortality rate (D) of 3.50% per day (Fig. 5). The relationship was ln (C_t) = 4.73 – 0.036t (p < 0.001, r² = 0.86).

Figure 5.

Catch curve for tubenose goby sampled between July and October, 2018 at two sites in the St Louis River estuary, Lake Superior. Instantaneous mortality is represented by the slope of the regression (C_t = 4.73 – 0.036t (r² = 0.86) equating to a daily mortality rate of 3.5%.

Discussion

To our knowledge, this is one of few studies conducted within the Great Lakes to characterize the diet of age-0 tubenose goby and its potential for diet interactions with a native demersal fish, and the first study to estimate somatic growth. The study had three major findings. First, age-0 tubenose goby had low dietary breadth in the St. Louis River estuary, primarily consuming amphipods and isopods, and the finding was consistent between two locations within our study area and from summer through early fall. Our results demonstrate considerable diet composition differences between fish sampled from the non-native range and those sampled in the native, European habitat. Second, the diet overlapped strongly with the diet of native tadpole madtom, demonstrating the potential to compete with a native, demersal species. Although more research is needed to assess possible competition with other species, several benthic fishes such as darters and sculpins consume similar prey as madtoms and tubenose gobies (French and Jude, 2001). Third, both instantaneous growth and mortality were high, and the estimate of maximum size was small (either 109.9 mm or 78.8 mm TL). This discussion will examine the findings and implications of our findings with respect to tubenose goby diet and growth, both relative to the St. Louis River estuary and the Laurentian Great Lakes.

Diet

Tubenose goby diet was exclusively composed of invertebrates, and the majority of the biomass was benthic macrocrustacea (Isopoda and Amphipoda). We note, however, that diet composition results can be biased by variable digestion rates: hard tissues (i.e., Crustacea exoskeleton and Chironomidae heads) may persist in guts longer than soft-tissues (i.e., eggs and zooplankton). Benthic macrocrustacea, especially Amphipoda, have been found to also be important in tubenose goby diets in the St. Clair River in Michigan (French and Jude, 2001). Annelida (Hirudinea and Oligochaeta; 0% of diet), Chironomidae larvae, Trichoptera and Ephemeroptera were by comparison much less common in stomach contents, despite being relatively common within the St. Louis River estuary area compared to benthic macrocrustacea (Breneman et al., 2000) and being observed in tubenose goby diets in other areas. Studies from Europe have found consumption of benthic macrocrustacea by tubenose goby (Shorygin, 1939), but usually of lesser importance than Chironomidae (Adámek et al., 2010, 2007), Trichoptera, Ephemeroptera (Vašek et al., 2014), or zooplankton (Všetičková et al., 2014). Round goby and tubenose goby often occupy overlapping and rocky habitat in European systems (Mikl et al., 2017; Vašek et al., 2014); however, within the Great Lakes, tubenose goby prefers shallow, vegetated habitat in coastal wetlands (Peterson et al., 2011; Ramage, 2017), whereas round goby prefers rocky habitat or rip-rap (Leino and Mensinger, 2016; Ray and Corkum, 2001). Thus, the diet composition difference between European and Great lakes tubenose goby populations may arise from differences in regional habitat preference.

Diet overlap of tubenose goby was high between summer and fall, as well as between locations within the St. Louis River estuary. This similarity between sites does not support specialist tendencies or diet plasticity of tubenose goby, which contrasts with European literature that establishes a strong locational effect, indicating high feeding opportunism (Všetičková et al., 2014). Although tubenose goby diets were dominated by benthic macrorustacea, we lack prey density data to determine prey selectivity. Both study locations were similar in habitat and possibly in prey availability. In the Great Lakes, several species of Amphipoda and Isopoda (including several species of Gammaridae and Assellidae which were diet items identified by DNA) are highly abundant within high-macrophyte cover habitat (Tall et al., 2008).

Diet characteristics of tubenose goby indicate that there is strong potential for competitive interactions with native species. Tubenose goby and tadpole madtom both had a low Levins’ Index, resulting from a relatively narrow prey field. The high value of Schoener’s Index between tadpole madtom and tubenose goby (S = 0.891) indicates that dietary competition is likely between species, although this likelihood may be reduced if these species display different foraging strategies (French and Jude, 2001). The lack of significant difference in gut fullness between species may indicate that food availability is not yet a limiting factor or that neither species displays a clear competitive advantage. Slimy sculpin (Cottus cognatus), spoonhead sculpin (C. ricei), logperch, johnny darter, bullheads and channel catfish (Ictalurus punctatus) are other native, benthic species in the SLRE (Leino and Mensinger, 2016). Only tadpole madtom was compared to tubenose goby in this study and more research is needed to thoroughly assess diet overlap between tubenose goby and other benthic species; however, several other benthic species are primarily insectivores and have been suggested to show dietary overlap (Kocovsky et al., 2011). Sculpins, darters, and logperch have all been suggested to be vulnerable to resource competition from tubenose goby within the Great Lakes (Jude et al., 1995; French and Jude, 2001; Kocovsky et al., 2011; Ramage, 2017), but their lack of detection in this study suggests that this interaction may be reduced in the St. Louis River owing to different habitat preferences or life history strategies, or that these native species have been displaced already; it is possible that the presence of tubenose goby has contributed to the lack of detection of similar benthic species in their habitat. Slimy sculpin abundance has been at historically low levels in the SLRE since 1989 (Leino and Mensinger, 2016), and although logperch and johnny darters have been variably abundant in the soft substrate of the SLRE in past years, they have been recorded to prefer less-vegetated and sandier habitat than tubenose goby in the Great Lakes (Lane et al., 1996). Fish species likely display differing susceptibility to sampling via beach seine, and higher proportions of benthic species have been associated with lower total capture efficiency. Furthermore among benthic fish, smaller individuals have escaped seines more easily (Pierce et al., 1990). Ultimately, the ecological importance of the diet overlap between tubenose goby and tadpole madtom is difficult to discern owing to both lack of prey selectivity information (i.e., these diet items may be abundant in wetland habitat) and the need for better taxonomic resolution.

Growth

Tubenose goby growth was allometric (b = 3.297), and the growth coefficient was similar to round goby (b = 3.261; MacInnis, 1997). Nevertheless, several aspects of growth and life-history differed between the results of this study and published literature. European literature (Janáč et al., 2012; Všetičková et al., 2014) has assigned fish of >40 mm SL (approximately 50 mm TL) an age of 1+; however, based on our analysis, tubenose goby of that size are age 0+. This finding merits further investigation because the difference in cohort structure between the European literature and our results is substantial, and has important ecological implications in both native and non-native waterbodies. Based on our age estimates, the corresponding hatch dates indicated a protracted spawning period which ranged from June 2-July 29, consistent with previous findings from the Great Lakes (Leslie et al., 2002) and European literature that indicates protracted spawning (Janáč et al., 2013), and corroborating our age estimates. Unlike the multiple spawning periods within a season that may be exhibited by some populations of round goby, we are not aware of any research in which a population of tubenose goby has spawned multiple times within a season.

The lack of age 1+ fish found during our study suggests that maximum lifespan within the SLRE is less than in some other systems, where it has been estimated to be as high as 4–5 years (Leslie et al., 2002). However, our findings do concur with other research that has elsewhere found the tubenose goby to show an annual life cycle and attainment of sexual maturity within the first year of life (Valová et al., 2015; Saç, 2019). The short lifespan is consistent with our finding of high mortality of age-0 fish. Given the large amount of interagency sampling across different habitats that regularly occurs within the harbor without producing tubenose goby, it does not appear to be a result of capture bias; however, there is a possible caveat in that very little sampling occurs during the spring when adult tubenose goby are actively spawning. In annual surveys in the St. Louis River, tubenose goby are not captured by electrofishing or bottom trawling, and are generally confined to shallow, vegetated habitat (Peterson et al., 2011), which is the habitat we sampled. Sampling in habitat >2 m depth within the estuary via trawl has been conducted by multiple agencies (United States Geological Survey; United States Environmental Protection Agency; 1854 Treaty Authority) but has not historically produced large numbers of tubenose goby (Leino and Mensinger, 2016; Peterson et al., 2011). Seasonal migrations and predation have been suggested in other systems as causes for a decrease in winter abundance of this species (Kocovsky et al., 2011; Valová et al., 2015). Given the persistence of tubenose goby in the St. Louis River estuary despite a scarcity of age 1+ fish, it is possible that tubenose goby reaches sexual maturity within one year; indeed, visible gonads were distinguished in tubenose goby of 2–3 months old. A short life-span and rapid maturity is consistent with a small asymptotic size, either 78.84 mm (Gompertz) or 109.90 mm (VBGF). However, the L_∞ parameter is meaningful only in populations with a sufficiently low mortality to enable fish to reach maximum size (Francis, 1988), and thus this parameter may be underestimated by the models.

The size at hatch (4.06 mm, ± 2.71 SE) is similar to reported appearances of recently-hatched tubenose goby (5–6 mm) in the St. Clair River (Leslie et al., 2002). Based on the transformed linear growth function, age-0 tubenose goby displays an instantaneous growth rate (0.62 mm/day) consistent with age-0 specimens of other wetland species within the Great Lakes, including yellow perch (0.72 mm/day; Ney and Smith, 1975) and smallmouth bass (0.58–1.20 mm/day; Steinhart et al., 2004). However, such comparisons should be treated cautiously because our estimate includes the latter part of the growing season, when growth had slowed relative to summer. This period likely represents the maximum growth rate for tubenose goby in the St. Louis River estuary, as growth generally slows for temperate fishes in the fall and winter.

Conclusions

The tubenose goby reaches a small maximum size, grows rapidly, experiences high mortality, and likely reaches sexual maturity within the first year of life. Many of these attributes are important factors for successful establishment of an introduced species (Sakai et al., 2001; García-Berthou, 2007), and suggests that tubenose goby has a high chance of establishment if introduced to new, suitable water bodies. Vegetation corridors have been hypothesized to act as avenues of dispersal for this species (Kocovsky et al., 2011) and the lack of such vegetation in many areas of the Great Lakes may explain the limited dispersal of this species from areas of initial introduction.

Thus far, no substantial negative interaction between tubenose goby and native fishes has been recorded in the Great Lakes; however, our study suggests greater attention should be paid to interactions during the spring reproductive season (including possible egg predation) and within coastal wetland habitat. Fisheries managers may wish to consider the potential effect of an invasive benthic fish that prefers shallow, coastal wetland areas, which are vital for recruitment and production of many commercially or recreationally important fish species throughout the Great Lakes (Sierszen et al., 2012; Trebitz and Hoffman, 2015). In many areas of the Great Lakes, including the St. Louis River estuary, restoration of coastal wetland areas is a priority for multiple agencies and stakeholders. Such restoration projects are typically evaluated for their effects on the dispersal of invasive terrestrial and aquatic plants; indeed, strategies have historically been employed to control invasive plants in Great Lakes wetlands (Lishawa et al., 2015). However, the ecology of invasive fish is rarely a significant factor in the decision-making process of habitat modification (Jude and Deboe, 1996). Although the benefits of increased connectivity may outweigh the risk of tubenose goby expansion, facilitation of tubenose goby dispersal – and the resulting trophic effects of dietary competition and potential egg predation – should be considered by agencies as a possible unintended consequence of increasing shallow wetland habitat connectivity. Pre- and post-restoration monitoring can help to determine the effects of such projects on the dispersal and spread of tubenose goby.

Figure 4.

Gompertz (solid line), von Bertalanffy (dashed line), and Linear (dotted line) Growth Functions for age-0 tubenose goby. Gompertz is represented by L_t = 78.84e (−e ^{−0.03(t – 43.77)})e ^ε, von Bertalanffy is represented by L_t = 109.90 (1 – e ^{−0.01(t – 13.72)})e ^ε, and Linear is represented by L_t = 4.06 + 0.618t.