The Braveheart amphipod: a review of responses of invasive Dikerogammarus villosus to predation signals

Predator pressure is a fundamental force driving changes at all levels of the community structure. It may protect native ecosystems from alien species. Therefore, resistance to diverse predators resulting from a universal anti-predator strategy seems crucial for invasion success. We present a comprehensive review of the responses of an invasive amphipod Dikerogammarus villosus to sympatric and allopatric predator signals. We summarize diverse aspects of the gammarid anti-predator strategy, including predator identification, morphological and behavioural adaptations, effectiveness of shelter use and resistance to indirect predator effects. The response of D. villosus is independent of predator species (including totally allopatric taxa), which assures the high flexibility of its predator recognition system. It has a harder exoskeleton and better capability of utilizing shelters compared to other gammarids, resulting in relatively high resistance to predators. Therefore, it can use predator kairomones as indirect food signals (sharing the diet with the predator) and follow the predator scent. This resistance may allow D. villosus to reduce the costs of its physiological responses to predators and sustain growth in their presence. This might facilitate invasion success by increasing its competitive advantage.


INTRODUCTION
Dikerogammarus villosus (Sovinsky, 1894) is a gammarid of Ponto-Caspian origin, commonly regarded as one of the most invasive freshwater species in the world (DAISIE, 2009). In its native region, it lives in the lower courses of large rivers of the Black, Azov and Caspian Sea basins, and in limans formed at their outlets (Rewicz et al., 2014). It has spread in central and western Europe using the southern migration corridor through the Danube and Rhine rivers (Bij de Vaate et al., 2002), as well as the central corridor, through the Dnieper, Bug, Vistula and Elbe rivers (Grabowski, Bacela & Konopacka, 2007;Mastitsky & Makarevich, 2007). At present, it occupies the widest novel range (most of Europe, excluding the Iberian and Scandinavian Peninsulas, but including Great Britain) and reaches the highest abundances in invaded areas within the group of several invasive Ponto-Caspian gammarid species (Rewicz et al., 2014;Rewicz et al., 2017;Šidagyte et al., 2017;Gusev, Guseva & Sudnik, 2017).
In novel areas, D. villosus exerts a strong impact on local biota through several mechanisms (Gergs & Rothhaupt, 2015). Firstly, it is an omnivore with a tendency to animal food, efficiently preying on many invertebrate species (Krisp & Maier, 2005), including intra-guild predation on local amphipod species (Dick & Platvoet, 2000;MacNeil & Platvoet, 2005;Kinzler et al., 2009). D. villosus strongly prefers animal food over plants (Van Riel et al., 2006;Gergs & Rothhaupt, 2008b) and grows better on it (Gergs & Rothhaupt, 2008a). However, recent field studies have revealed that it can also act as a typical herbivore, consuming mainly plant food (Maazouzi et al., 2009;Hellmann et al., 2015;Koester, Bayer & Gergs, 2016). This points to its high plasticity and ability to use various available food resources (Mayer, Maas & Waloszek, 2012), depending on local circumstances, such as the community composition (Hellmann et al., 2017). Moreover, it efficiently competes with other gammarids, both native and other aliens, for food, shelters and optimum habitats (Dick, Platvoet & Kelly, 2002;Hesselschwerdt, Necker & Wantzen, 2008;Jermacz et al., 2015a). Competitive tensions and intra-guild predation are responsible for the reduction in the occupied ranges and abundances of native species, which are being outcompeted to less suitable habitats and sometimes even locally displaced (Dick, Platvoet & Kelly, 2002;Muskó et al., 2007;Hesselschwerdt, Necker & Wantzen, 2008;Platvoet et al., 2009). Furthermore, D. villosus can exert some more subtle effects on ecosystem functioning. For instance, being a less efficient shredder than other amphipods, displaced by its appearance, D. villosus may negatively affect food webs by reducing the numbers of organisms relying on shredded organic material (MacNeil et al., 2011). On the other hand, D. villosus does not respond to predation risk by reduction in feeding (Jermacz & Kobak, 2017); therefore, it can still be capable of shredding organic material when other amphipods suffer non-consumptive costs of predator pressure (Abjörnsson et al., 2000;Jermacz & Kobak, 2017;Richter et al., 2018).
The invasion success of D. villosus is regarded to result from several traits of its biology, including its fast growth rate, high fecundity, tolerance to wide ranges of abiotic factors, in particular raised salinity (Devin & Beisel, 2007), as well as high plasticity and omnivory (as reviewed by Grabowski, Bacela & Konopacka, 2007). Another trait contributing to its spread is the ability to adhere to various hard substrata and artificial materials, including boat hulls and diving equipment (Bacela-Spychalska et al., 2013). Although D. villosus is not well adapted to air exposure (Poznańska et al., 2013), under suitable conditions, e.g., hidden in a zebra mussel colony, individuals of this species can survive several days without water (Martens & Grabow, 2008), sufficient for successful transport to another water body.
Yet another important trait which can potentially affect the invasive potential of the species is an effective strategy of energy saving (Becker et al., 2016) observed also under predation risk (Jermacz et al., 2017;Jermacz & Kobak, 2017). In recent years, we have conducted a series of experimental studies on the reactions of D. villosus to predation cues and their potential implications for its functioning and invasiveness. In this review, we provide a synthesis of our research on these topics accompanied by the results of other authors on the biology of D. villosus and related amphipod species.

Survey methodology
To obtain a comprehensive set of literature reports on interactions between predators and amphipod prey, we conducted a literature survey in the Scopus database, using the following keywords: D. villosus or gammarid or amphipod combined with the following: anti-predator response or predator impact or anti-predator behaviour or predator defence or predator kairomone or predation risk or prey response. We found 67, 115 and 927 articles, respectively, meeting the criteria, 11 of which were of our own authorship. Among these articles were chosen key papers related to defence strategies exhibited by native and invasive amphipods.
Biological invasions add a new and interesting aspect to predator-prey interactions. In old systems, where predator and prey coevolve together for a long time, both are well adjusted to each other. The responses of prey species can be fine-tuned to specific predators (Wudkevich et al., 1997;Weber, 2003;Boeing, Ramcharan & Riessen, 2006), but also predator preying modes allow them to feed efficiently on available victims (Gliwicz, 2005). However, alien species, just transported to their novel locations, face completely new, unknown communities, containing new predators and new prey. In accordance with the Enemy Release Hypothesis (Torchin et al., 2003), as a consequence of the loss of natural parasites and predators, introduced species suffer lower pressure than native species. At the same time, local consumers may be unfamiliar with alien prey and unable to forage on them efficiently (Meijer et al., 2016). On the other hand, alien species are also not adapted to their new, potential predators which may prevent them from employing efficient anti-predation mechanisms and lead to an evolutionary trap: inefficient or even maladaptive responses or the lack of reactions to a danger (Salo et al., 2007;Zuharah & Lester, 2010).

Predator recognition by D. villosus
For a perfect invasive species, the mechanism of predator detection should be universal, enabling the recognition and subsequent response to a novel predator without a common evolutionary history. As a consequence of an improper identification of a predator signal, prey species are exposed to higher predation due to the lack of responses or maladaptive responses (Abjörnsson, Hansson & Brönmark, 2004;Banks & Dickman, 2007). Such a scenario was presented by Pennuto & Keppler (2008) who demonstrated that a native Gammarus fasciatus is able to avoid a narrower range of potential predators than an invasive Echinogammarus ischnus. Moreover, ineffective recognition of danger could result in costly defence reactions when the predation risk is low (Lima & Dill, 1990;Dunn, Dick & Hatcher, 2008) as was experimentally shown for Gammarus minus responding to a predatory fish Luxilus chrysocephalus (Wooster, 1998). Therefore, appropriate predation risk assessment is crucial for an adequate response and optimization of energy expenditure.
Laboratory experiments demonstrated the ability of D. villosus to recognize diverse fish predators, including bottom dwellers: the racer goby Babka gymnotrachelus (Jermacz et al., 2017;Jermacz, Dzierżyńska-Białończyk & Kobak, 2017), European bullhead Cottus gobio (Sornom et al., 2012) and spiny-cheek crayfish Orconectes limosus (Hesselschwerdt et al., 2009), as well as fish swimming in the water column: the Eurasian perch Perca fluviatilis, Amur sleeper Perccottus glenii (Ł Jermacz & J Kobak, pers. obs., 2015-2016) and red-bellied piranha Pygocentrus nattereri (Jermacz, Dzierżyńska-Białończyk & Kobak, 2017). Among these species, the goby, bullhead and perch have co-occurred with the gammarid in its home range, the Amur sleeper and crayfish were met several dozen years ago in its novel areas, whereas the piranha had no previous contact with D. villosus. The above-mentioned studies indicate a universal method of predator recognition exhibited by D. villosus, effective with regard to both native and novel predatory species. A situation when potential naïve prey recognizes and responds to a novel predator can be explained by several mechanisms. For example, conspecifics can be present in the predator diet, providing information about predation risk (Chivers & Smith, 1998), as was demonstrated for another invasive gammarid Pontogammarus robustoides (Jermacz, Dzierżyńska-Białończyk & Kobak, 2017). Moreover, the novel predator can be closely related to some native predators (Ferrari et al., 2007;Sih et al., 2010) and therefore release similar signals.
The avoidance reactions of D. villosus were studied by Jermacz, Dzierżyńska-Białończyk & Kobak (2017) in a flow-through Y-maze allowing gammarids to select an arm with or without the scent of predators fed on different diets. This study indicated that the avoidance of predators was induced in the presence of kairomones emitted by hungry predators (starving for 3 days), including totally allopatric, tropical P. nattereri (Jermacz, Dzierżyńska-Białończyk & Kobak, 2017). The avoidance response of D. villosus to another hungry predator, American spiny-cheek crayfish, was noted by Hesselschwerdt et al. (2009). Thus, the predator identification system of D. villosus seems to be independent of the presence of conspecifics in the predator's diet. Nevertheless, it should be noted that D. villosus did also recognize the predator diet and used it as an additional source of information about the predator status and current level of predation risk, though its responses to satiated predators did not include avoidance (see the chapter ''Positive response of D. villosus to the predation cue'' below) (Jermacz, Dzierżyńska-Białończyk & Kobak, 2017). Avoidance of a hungry predator, which is most determined to obtain food, and modifications of the responses to satiated predators suggest that D. villosus is capable of effective risk assessment and flexible responses, adjusted to the current situation. A similar relationship between the level of predator satiation and prey response was observed in the case of a water beetle Acilius sulcatus, responding only to hungry perch, but not to satiated fish (Abjörnsson et al., 1997).
The versatility of the predator detection mechanism of D. villosus could be related to the fact that active components of kairomones emitted by unrelated predators are often very similar (Von Elert & Pohnert, 2000). Therefore prey can react to diverse predators, including those which evolved in isolated ecosystems. In temperate European water bodies, fish usually have broad diet ranges and most of them feed on invertebrate food at least at particular life stages (Wootton, 1990;Gerking, 1994). Thus, a general response to hungry fish of particular size seems beneficial under such conditions. D. villosus is an invasive species characterized by a high dispersal rate. During the dispersal, the probability of meeting a novel predator is high, therefore species exhibiting universal defence mechanisms and/or the capability of quick adaptations are more likely to be successful invaders.

Anti-predator defence mechanisms of D. villosus
Site selection and shelter occupancy For a benthic organism, one of the most important elements of the anti-predator strategy is related to the optimal substratum choice. In general, prey survival rate increases with the level of substratum complexity and heterogeneity (Crowder & Cooper, 1982;Holomuzki & Hoyle, 1988;Czarnecka, 2016). Therefore, the distribution of benthic invertebrates depends on the bottom character (Czarnecka et al., 2009;Jermacz et al., 2015b) and their efficiency in using available substrata as shelters (Holomuzki & Hoyle, 1988;Kobak, Jermacz & Płąchocki, 2014;Kobak et al., 2016).
Compared to other gammarids, D. villosus is regarded as a sit-and-wait animal, spending most of its time in a shelter (Kinzler & Maier, 2006;Kley et al., 2009;Platvoet et al., 2009;Beggel et al., 2016). It prefers the substratum consisting of large gravel or stones (>6 cm in diameter), which provides it with suitable protection and enough empty space to move (Kley et al., 2009;Boets et al., 2010;Kobak, Jermacz & Dzierżyńska-Białończyk, 2015). Perhaps due to its low activity (Van Riel et al., 2007;Beggel et al., 2016), changes in shelter occupancy in the presence of predators observed in various studies are ambiguous. In the presence of benthivorous fish (European bullhead), D. villosus was observed to reduce its presence in the open field (i.e., outside shelters) and activity considerably, from ca. 55% under control conditions to only 20% of the total experimental time (Sornom et al., 2012). However, in other studies, the reduction in open field occupancy or activity in the presence of predatory fish was only slight, though significant (Beggel et al., 2016;Jermacz et al., 2015a;Jermacz & Kobak, 2017), or no response was observed at all Jermacz et al. (2017). These discrepancies might have resulted from the varying quality of shelters that could be occupied always or only in the presence of danger, as well as from the location of food. Sornom et al. (2012) found that in the presence of predators D. villosus decreased its activity and stayed more often in shelters made of holes in the solid substratum, whereas mesh shelters were always occupied irrespective of predator presence (>80% of the total time). In the studies by Jermacz et al. (2015a) and Jermacz & Kobak (2017) gammarids spent more than 95% of the total experimental time in gravel substratum under predator absence, which allowed for only a small, though still significant change in response to predators. Jermacz et al. (2017) andBeggel et al. (2016) found that gammarids spent most of their time in gravel shelters even in the absence of predators. Jermacz & Kobak (2017) observed gammarids to limit their occupation of the open space in the presence of predators when food was present in the direct vicinity of their shelters, whereas they kept exploring the unsheltered area in search of distant food sources. Thus, the presence of food can increase gammarid activity, which in turn can be reduced by the predation cue when the food is available at a short distance.
D. villosus often occurs on hard and complex substrata, difficult to access by predators. Stone substratum was found to offer it more protection against fish predation compared to Gammarus fossarum and G. pulex, but this advantage disappeared on sand (Kinzler & Maier, 2006). In the wild, D. villosus was often found associated with zebra mussel (Dreissena polymorpha) colonies (e.g., Devin et al., 2003;Boets et al., 2010). Kobak, Jermacz & Płąchocki (2014) demonstrated that living dreissenids provided D. villosus with the most effective shelter against fish predators (the racer goby and Amur sleeper), compared to stones, macrophytes and shell litter. It should be noted that this shelter was also useful against a species without a common evolutionary history and exhibiting a different feeding strategy than the sympatric gobies (the Amur sleeper). This study demonstrated the positive effect of dreissenids on prey survival only in the case of D. villosus, but not for other invasive (P. robustoides) and native (Gammarus fossarum) species. However, in contrast to our studies, Beekey, McCabe & Marsden (2004) showed that also native prey, including amphipods, experiences lower predation pressure in dreissenid beds.
Dreissenid beds can offer more effective protection than other substrata (Kobak, Jermacz & Płąchocki, 2014), to less active species, such as D. villosus (Kobak et al., 2016), spending most time in the shelter (Beggel et al., 2016;Jermacz & Kobak, 2017). This indicates that the presence of gregarious bivalves may promote the establishment of D. villosus. Dreissenid colonies, in contrast to other substratum types, form aggregations of objects bound with one another by byssal threads, hard to penetrate by fish (Kobak, Kakareko & Poznańska, 2010) which, in association with the high attachment ability of D. villosus compared to other gammarids (Bacela-Spychalska, 2016) may make a mussel bed a perfect shelter for this species. Moreover, the hard substratum which supplies not only shelter and clinging opportunity, but also food resources, such as a colony of D. polymorpha, seems to be an optimal habitat for the invasive gammarids and may allow them to limit their exploration activity (Jermacz & Kobak, 2017). Mussels provide both effective anti-predator protection (Beekey, McCabe & Marsden, 2004;McCabe et al., 2006;Kobak, Jermacz & Płąchocki, 2014) and valuable food resources, such as organic-rich pseudofaeces and macroinvertebrate prey of increased abundance (Gergs & Rothhaupt, 2008b;Kobak et al., 2016).
When shelters are limited (e.g., on sandy unvegetated nearshore bottoms in the wild), D. villosus exhibits an avoidance response to the predator cue, as shown by Hesselschwerdt et al. (2009) and Jermacz, Dzierżyńska-Białończyk & Kobak (2017) in a Y-maze. This response was observed in the presence of hungry predators (starving for 3 days), likely to pose the highest danger to their potential prey. Thus, in the absence of suitable shelters and the presence of a direct danger, a temporary increase in activity and active avoidance seems to be an optimum response. In natural conditions, such a response is likely to result in leaving the predator area or finding the nearest shelter, after which the activity is reduced as the predation risk decreases.

Aggregation forming
Shelter choice depends not only on the substratum quality but also the presence or absence of conspecifics and heterospecifics gammarids (Jermacz et al., 2015a;Jermacz et al., 2017). Laboratory experiments showed that D. villosus preferred shelters occupied by conspecifics over empty shelters and conspecifics located apart from shelters (Jermacz et al., 2017). Moreover, D. villosus exhibited a preference for shelters inhabited by conspecifics over those occupied by heterospecifics gammarids (P. robustoides), thus forming single-species aggregations (Jermacz et al., 2017). Nevertheless, in the presence of predators, the selectivity of gammarids was reduced and they grouped alike with conspecifics and heterospecifics. The choice of the substratum already inhabited by other prey individuals is an example of aggregation behaviour combined with the benefits of sheltered conditions. The main advantage of the aggregation strategy is a reduction in the individual risk of predation (Hamilton, 1971). On the other hand, the weakness of this strategy is the facilitation of detection by a predator, especially by species using vision for prey detection (Ioannou & Krause, 2008). However, when gammarids are aggregated under sheltered conditions, their detection seems to be difficult.
Notwithstanding the protective role of gammarid aggregations against predators, D. villosus did not increase the intensity of its grouping in shelters in the presence of predators (Sornom et al., 2012;Jermacz et al., 2017), in contrast to its relative, Pontogammarus robustoides (Jermacz et al., 2017). However, also in contrast to P. robustoides, D. villosus exposed to predation cues formed conspecific aggregations in open places, in the absence of shelters (Jermacz et al., 2017). The effectiveness of such a response as a protection against predators was demonstrated under laboratory conditions in which the racer goby avoided aggregated prey and consumed it less efficiently than singletons (Jermacz et al., 2017). This may be a consequence of the aforementioned clinging abilities of D. villosus (Bacela-Spychalska et al., 2013) and/or the hardness of its exoskeleton (Błońska et al., 2015), which are greater than those of other gammarids, such as G. fossarum or P. robustoides (Bacela-Spychalska et al., 2013) and increase predator handling costs, thus contributing to the resistance of such aggregations against predators. On the other hand, the easiest prey for predators were single inactive individuals (Jermacz et al., 2017), indicating that this state should be avoided by gammarids seeking protection against predation.

Depth selection
For an aquatic organism, the choice of an appropriate habitat is also related to water depth. Fish predation pressure at shallow nearshore locations can be significantly lower than at deeper sites (Gliwicz, Soń & Szynkarczyk, 2006;Perez et al., 2009). An experiment conducted by Kobak et al. (2017) in a 1 m deep tank with a depth gradient demonstrated that D. villosus in the presence of the racer goby relocated from the deepest zone, occupied preferentially under safe conditions, to the shallower bottom. Moreover, it also climbed upwards along the vertical tank walls and attached near the water surface (Kobak et al., 2017). Gobies are bottom-dwelling predators, rarely swimming freely in the water column; therefore the escape to the water column seems to be an effective response against them (Pinchuk et al., 2003). Our experimental results are reflected in field observations made in Lake Balaton occupied by Ponto-Caspian Gobiidae (Ferincz et al., 2016), where D. villosus occurs mainly on the stones near the water surface (Muskó et al., 2007).

External factors affecting gammarid responses to predators
The responses of D. villosus to predator cues are modified by environmental pollution, such as increased heavy metal concentration (Sornom et al., 2012). The gammarids exposed to the solution of 500 µg of cadmium per litre of water were observed to hide less often and be more active than the control individuals. Moreover, they no longer changed their behaviour in response to the presence of predators (Sornom et al., 2012).
Yet another potential factor that can potentially affect prey responses to predators is the presence of parasites, modifying host activity and it refuge use (Lafferty & Morris, 1996;Perrot-Minnot, Kaldonski & Cézilly, 2007;Kaldonski et al., 2008). D. villosus in European waters is parasitized by a microsporidian Cucumispora dikerogammari (Bacela-Spychalska et al., 2012). This parasite was found to affect the behaviour of its host, making it more active (Bacela-Spychalska, Rigaud & Wattier, 2014), which could increase the detection probability and in consequence the survival chance of infected prey. Activity increase could potentially expose parasitized individuals to predator attacks and reduce their defence capabilities, though at present no evidence exists for that and further studies are needed on this topic.
Finally, it should be noted that not all responses of D. villosus to predators can be considered as anti-predator defences ( Table 1). The predator diet can strongly modify the behaviour of gammarids and switch their responses from typical avoidance to even preference for predator scents (Jermacz, Dzierżyńska-Białończyk & Kobak, 2017). See the chapter ''Positive response of D. villosus to the predation cue'' below for the details.

D. villosus as prey
Prey selection is a universal process, in which predators must choose among prey that differ in density and defence strategy. To optimize their fitness, predators should select those prey species whose abundance is high and hunting cost is low (Emlen, 1966). Many variables can influence prey choice. Some of them are related to prey characteristics such as prey defence mechanisms, including behavioural (Andersson et al., 1986), morphological (Bollache et al., 2006, and physiological adaptations (Clinchy, Sheriff & Zanette, 2013) or to environmental factors, such as habitat structure, food-quantity and temperature (Crowder & Cooper, 1982). Effective predation also depends on predator hunting strategy and its flexibility (Grabowska et al., 2009).
Under experimental conditions D. villosus exhibited higher survival than other gammarids, including both native and invasive species, in the presence of diverse predators, such as the sympatric Ponto-Caspian gobies or the allopatric European bullhead and Amur sleeper (Kobak, Jermacz & Płąchocki, 2014;Błońska et al., 2015;Błońska et al., 2016;Beggel et al., 2016). Moreover, Kley et al. (2009) observed that the turbot (Lota lota) consumed fewer D. villosus compared to Gammarus roeselii. A similar result was shown by Błońska et al. (2015), who demonstrated that the racer goby always consumed preferentially native G. fossarum over D. villosus even if the gammarids were immobilized and unable to defend themselves. However, the goby did not exhibit any selectivity towards the waterborne chemical signals of native and invasive amphipods in a Y-maze (Błońska et al., 2015). On the other hand, Błońska et al. (2016) demonstrated that immobilized D. villosus and Table 1 Anti-predation mechanisms of D. villosus and other changes induced by the presence of predators.

Trait Comments References
Constitutive traits (not changing in the presence of predators, but potentially protective) Staying inactive in the shelter The species is less active than other gammarids Reduced consumption of food Shown when food had to be searched for, not shown when food was present directly in the shelter (2017) native G. fossarum were equally selected by other goby species (the round goby Neogobius melanostomus and the tubenose goby Proterorhinus semilunaris) and the European bullhead, whereas mobile D. villosus specimens were avoided, irrespective of the presence or absence of shelters. This indicates that the effective behavioural anti-predator responses of D. villosus determined its survival under the pressure of these predator species (Błońska et al., 2016). The coarse and complex substratum (gravel, stones and zebra mussel colonies) also improved the survival of D. villosus compared to fine substrata and other gammarid species (Kinzler & Maier, 2006;Kobak, Jermacz & Płąchocki, 2014). These results suggest that the mechanisms of the resistance of D. villosus to different predators may vary depending on their hunting mode, size and/or other traits. The effectiveness of goby predation on D. villosus was described in detail by Jermacz et al. (2017). They demonstrated that under particular conditions, for example when gammarids were active or aggregated, the percentage of successful gobiid attacks was lower than 25%. The predation efficiency exceeded 50% only in the case of single inactive gammarid individuals. Moreover, even when a fish already had a gammarid in its mouth, the prey was still able to escape without any visible damage. Such a low effectiveness of predation forces predatory species to multiply their effort to achieve the desired satiation level or choose alternative prey species if available. The necessity of feeding on prey generating high handling costs is unfavourable for the predator condition. For example, under laboratory conditions Błońska et al. (2015) demonstrated that gobiids fed with native G. fossarum or chironomid larvae grew significantly better than individuals forced to feed on D. villosus, and the latter group of fish exhibited weight loss after a 4-week exposure.

Jermacz & Kobak
These observations confirm that D. villosus is a comparatively poor food item for its potential predators, and is likely to be avoided in the presence of alternative prey species, which can make it relatively safe in the natural environment. Generally, amphipods are considered as one of the most important elements of the diet of many fish species (MacNeil, Dick & Elwood, 1999), however experimental results demonstrated that the role of D. villosus as food for the fish community could be significantly different from that of its native counterparts (Kobak, Jermacz & Płąchocki, 2014;Błońska et al., 2015;Błońska et al., 2016;Beggel et al., 2016), often replaced by the alien species (Dick & Platvoet, 2000;Dick, Platvoet & Kelly, 2002;Grabowski et al., 2006).

Positive response of D. villosus to the predation cue
In general, a chemical signal indicating predator presence induces a defence response responsible for the reduction in predation risk (Brönmark & Hansson, 2000;Ferrari, Wisenden & Chivers, 2010). However, in the case of omnivorous species, capable of feeding on predator faeces or their dead bodies, or partly sharing their diet, a predation signal does not always indicate only a danger and, as a consequence, does not always induce a defence response. Such a unique situation takes place in the case of D. villosus as it actively avoided the scent of hungry predators in a Y-maze, but did not exhibit an avoidance reaction to the predation cues emitted by predators fed with chironomids or other gammarids (including conspecifics) (Jermacz, Dzierżyńska-Białończyk & Kobak, 2017). On the contrary, it showed an active preference, moving towards the scent of satiated predators in a Y-maze (Jermacz, Dzierżyńska-Białończyk & Kobak, 2017). A similar response was induced by the presence of cues released by crushed conspecifics and other gammarid species. This reaction suggests that this omnivorous and cannibalistic species is able to use such a signal not as a predation cue, but as a source of information about the location of a feeding ground. As shown in the above sections of this review, D. villosus is characterized by an effective defence strategy (Kobak, Jermacz & Płąchocki, 2014;Błońska et al., 2015;Błońska et al., 2016;Jermacz et al., 2017), therefore being relatively safe in the presence of predators, especially when alternative prey items are available in the vicinity (Jermacz et al., 2015a;Błońska et al., 2016). In such a situation, D. villosus may follow a predator to feed on its faeces or sense wounded invertebrates as being its potential prey. A similar trade-off between predator avoidance and foraging was observed in the case of Gammarus pulex, which in the presence of food did not respond to the predation signal, contrary to the situation when it was exposed only to predator kairomones (Szokoli et al., 2015).

Costs of the anti-predator responses of D. villosus
Anti-predatory defences of prey organisms usually result in considerable energetic costs of the development of additional structures, selection of suboptimal habitats and/or decreased feeding due to the higher vigilance focused on predator detection (Hawlena & Schmitz, 2010;Sheriff & Thaler, 2014). This may result in growth reduction (Janssens & Stoks, 2013), weaker condition (Slos & Stoks, 2008) and/or finally even mortality (McCauley, Rowe & Fortin, 2011). The impact of the presence of predators on the feeding of D. villosus was checked by Jermacz & Kobak (2017). The gammarids considerably limited their feeding in the presence of predators (by 95% and 74% depending on the location of food, placed in the direct vicinity of shelters or away from them, respectively). Surprisingly, this response was even stronger than that of the related species P. robustoides (77% and 33%, respectively), though the latter seems to be more susceptible to predation pressure. On the other hand, no decrease in feeding was observed when single gammarids did not have to search for their food, having it available directly in their shelters. This shows that the aforementioned limitation of feeding in the open field resulted from the limited activity of gammarids (when food was located close to the shelter) or their increased vigilance in the open field (when food was distant from the shelter and no reduction in the search time was observed).
Nevertheless, the most important result of the study by Jermacz & Kobak (2017) was the demonstration that the growth rate of D. villosus supplied with food in their shelters (over a period of 2 weeks) was unaffected by the presence of predators. On the other hand, P. robustoides under the same conditions significantly reduced its growth rate by ca. 60% when exposed to predation cues (Jermacz & Kobak, 2017). Reduction in growth under predation risk was also observed in an amphipod Hyalella azteca, accompanying its induced morphological adaptations resulting in lower predator pressure (James & McClintock, 2017). This confirms the relatively high resistance of D. villosus to non-consumptive predator effects and shows that it may thrive in a good physiological condition under predatory pressure.
Resistance of D. villosus to predator non-consumptive effects was also confirmed by Richter et al. (2018), who did not observe any disturbance of gammarid feeding behaviour under the pressure of a benthivorous fish, the European bullhead (Cottus gobio). In contrast, another gammarid species (G. pulex) reduced its consumption in the presence of C. gobio kairomones (Abjörnsson et al., 2000). Lagrue, Besson & Lecerf (2015) have shown that the abundance of armoured detritivore prey did not decrease in the presence of predators, in contrast to that of non-armoured species. D. villosus is more armoured than other gammarids (Błońska et al., 2015), and thus the consequence of trade-offs between behavioural and morphological defences, such as the cost of the anti-predator responses of D. villosus seems to be less pronounced than that of other gammarids.

Ecological significance of the anti-predator strategy of D. villosus
We have shown that D. villosus is capable of flexible predator recognition (Jermacz, Dzierżyńska-Białończyk & Kobak, 2017) allowing the species to respond to both novel and known dangers. It seems unlikely that it may benefit from the naïvety of local predators in central and western Europe, as they are used to preying on native gammarids (MacNeil, Dick & Elwood, 1999;MacNeil, Elwood & Dick, 1999) and not very selective with regard to their benthic food, consuming also large quantities of alien amphipods (Rezsu & Specziár, 2006;Eckmann et al., 2008). Moreover, predators of Ponto-Caspian origin, sympatric to the gammarids, such as several species of gobiid fish, have also invaded the same regions and co-occur with D. villosus in most of its current range, and include it in their diet (Grabowska & Grabowski, 2005;Borza, Eros & Oertel, 2009;Brandner et al., 2013). Therefore, its ability to easily recognize potential dangers may be one of the traits facilitating its establishment in invaded areas.
The efficient defence mechanisms of D. villosus make this species relatively resistant to predation (Kobak, Jermacz & Płąchocki, 2014;Jermacz et al., 2017;Jermacz, Dzierżyńska-Białończyk & Kobak, 2017), which may help it in its competition with other gammarids (Jermacz et al., 2015a;Beggel et al., 2016). Other, less resistant and more often consumed species are preferentially removed from the environment by predators and must spend more energy and time on anti-predator vigilance, whereas D. villosus, as the least preferred potential food, may thrive in the presence of predators with no negative effect on its growth (Jermacz & Kobak, 2017). Moreover, its aggressive behaviour may force competing gammarid species to less suitable habitats (Platvoet et al., 2009;Jermacz et al., 2015a) or make them swim more often in the water column, which further exposes them to fish predation (e.g., Jermacz et al., 2015a;Beggel et al., 2016). In addition to direct intra-guild predation and competition for food, this displacement from shelter is likely to be another factor making D. villosus an efficient competitor displacing other species from the areas in which it appears. Negative interactions with D. villosus make other gammarids avoid the presence of the stronger competitor, increasing their migrations to new areas and switching to different habitats (Dick, Platvoet & Kelly, 2002;Hesselschwerdt, Necker & Wantzen, 2008;Platvoet et al., 2009;Jermacz et al., 2015a;Kobak et al., 2016).
Nevertheless, the nature of interactions between D. villosus and its related species is far more complex. The presence of predators does have an impact on D. villosus behaviour (Jermacz, Dzierżyńska-Białończyk & Kobak, 2017;Kobak et al., 2017;Jermacz et al., 2017) and may reduce its interspecific aggression, allowing the competing species to stay in its presence. Jermacz et al. (2015a) have demonstrated that another gammarid P. robustoides is easily displaced from habitats preferred by both species in a safe environment, but the presence of predatory fish changes the situation, allowing P. robustoides to stay in the area co-occupied by D. villosus. It is difficult to distinguish whether this is due to the reduction in D. villosus aggression or the higher substratum affinity of P. robustoides in the presence of predators (selecting the vicinity of the stronger competitor as the lesser evil), or both. Nevertheless, individuals of both species can take advantage of staying in a group and reduce the probability of a successful predator attack (Jermacz et al., 2017). This also shows how important it is to consider the effect of predators when studying competitive interactions between species because the consequences of competition in a predator-free situation, which is very unlikely in the wild, may be easily overestimated.
Moreover, as the reduction in the feeding rate of D. villosus in the presence of predators was observed by Jermacz & Kobak (2017), it is likely that the predatory impact of this gammarid on the local community can also be lower than expected on the basis of experiments conducted in fishless conditions. This confirms the results of the field studies in the River Rhine (Hellmann et al., 2015;Hellmann et al., 2017;Koester, Bayer & Gergs, 2016) and can explain their discrepancy with some laboratory experiments, indicating the strong predatory impact of D. villosus on invertebrates (Dick & Platvoet, 2000;MacNeil & Platvoet, 2005).
Unexpectedly, despite the high consumption of D. villosus commonly observed in the field (Kelleher et al., 1998;Grabowska & Grabowski, 2005;Eckmann et al., 2008;Borza, Eros & Oertel, 2009;Brandner et al., 2013;Czarnecka, Pilotto & Pusch, 2014), it was experimentally demonstrated that its dominance may in fact decrease the quality of food conditions for fish due to the higher difficulty of capturing and handling, leading to poor growth on a diet based on this species, compared to the diets consisting of native gammarids or chironomid larvae (Błońska et al., 2015). Thus, although fish feed on D. villosus in the areas invaded by this species, it seems they would have thrived much better if this invasion had not occurred and other gammarid species (usually displaced by the invader) had been available as alternative food (Błońska et al., 2015).

CONCLUSION
We have shown D. villosus as a species with efficient anti-predation mechanisms (both behavioural modifications and constitutive traits), relatively safe from predators and bearing lower costs of their non-consumptive effects (as indicated by its growth unaffected by the presence of fish), compared to related taxa. It can recognize sympatric and novel fish predators independent of their diet, though its precise responses are fine-tuned on the basis of food consumed by a predator, and can range from avoidance to preference. Sometimes D. villosus can even be attracted to a predator scent, probably utilizing their presence to locate potential food sources. Defence mechanisms of this species include activity reduction, aggregation and migration. In general, single immobile individuals outside the shelter are the most susceptible to predation. Therefore, threatened individuals try: (1) to stay in the shelter, at best co-occupied by other specimens; (2) if this is not possible, to move in search of a shelter; (3) if shelters are difficult to find, to aggregate with conspecifics, used as a substitute shelter; (4) if conspecifics are also difficult to locate (e.g., at a low density), to relocate to safer areas, e.g., away from the predator scent or to the shallower bottom. These traits are likely to give it a strong advantage in competition with similar species, both natives and other invaders, and contribute to its invasive potential. Moreover, we have demonstrated the strong importance of predator effects on interactions among gammarid species involving D. villosus, which cannot be neglected in future studies on this topic. It is likely that under predatory pressure the competitive impact of D. villosus on other gammarids as well as its predation on zoobenthos organisms are reduced, altering its impact on local communities.