Introduction

Biological invasions are a common phenomenon in many aquatic and terrestrial ecosystems worldwide and their frequency has been increasing for decades (Simberloff et al. 2013; Turbelin et al. 2017). The establishment and spread of non-native species can cause substantial economic damage and has the potential to alter the structure and functioning of the affected ecosystems (Carlton 2002; Bax et al. 2003; Simberloff 2011). The development of methods and legislative regulations that allow the successful control and management of bioinvasions therefore constitutes a pressing task for scientists, managers and politicians who are concerned with environmental policies and conservation ecology (Dahlstrom et al. 2011; Ojaveer et al. 2014). For this, however, it is indispensable to understand which abiotic and biotic factors determine invasion success in species that are introduced into new habitats. Among several species-specific characteristics, such as mobility, reproductive output and competitiveness, which could play a key role for invasion success, tolerance to stressful environmental conditions has been identified as a core quality (Lockwood et al. 2007; Catford et al. 2009; Van Kleunen et al. 2010).

Previous studies have reported that some non-native species are more stress tolerant than taxonomically related native species with corresponding ecological roles (Schneider and Helmuth 2007; Zardi et al. 2007; Schneider 2008; Lenz et al. 2011; Zerebecki and Sorte 2011; Bates et al. 2013; Lejeusne et al. 2014). Whether robustness, i.e. the capacity to survive adverse environmental conditions caused by heat, oxygen limitations or low salinities, in these species is an inherent trait that evolved in response to a harsh environment—long before they become invasive in other regions of the world—or whether stress tolerance was acquired during the invasion process remains unclear. The latter comprises the transport from the donor region into the recipient region, the release into the new environment after arrival as well as the establishment and spread in habitats from which the species was previously absent (Lockwood et al. 2007). The first mechanism would rather constitute a prerequisite for invasion success, while the second mechanism could take place during the invasion process and could influence its outcome. If the first is true, this would mean that some but not all species can become invasive, while in the second case any organism with a wide ecological niche and/or a broad tolerance range could become a potential invader—given that its robustness is selected for during the invasion process.

The assumption that robustness can be acquired during the invasion leads to the question of which mechanisms have the potential to enhance stress tolerance in individuals or populations of plants or animals and at which stage of the invasion process they occur? Worldwide biological invasions in the marine environment are facilitated by anthropogenic vectors that can transport species over long distances in short time intervals (Carlton and Ruiz 2005). Hence, the majority of successful range expansions of aquatic and terrestrial species that have been documented during the last decades was presumably initiated by human-facilitated, intentional or unintentional transport of organisms (Levine and D’Antonio 2003; Floerl et al. 2009; Hulme 2009; Anderson et al. 2015). This causality is well documented for marine systems in which shipping traffic, mainly by container vessels, plays a key role for the dispersal of marine life forms beyond their natural distribution limits (Seebens et al. 2013; Williams et al. 2013; Bailey 2015).

Vessels mainly offer two different options to transport algae or invertebrates from one sea area to another: their hulls and their ballast water tanks (Ruiz et al. 2000). Ship hulls, although most often coated with antifouling paints, represent a large solid settlement substratum for algae and sessile invertebrate species such as mussels (Drake and Lodge 2007; Davidson et al. 2009; Chapman et al. 2013; Chan et al. 2015). The latter colonize these surfaces, especially when the antifouling coating is old, damaged, or masked by other foulers, as larvae or post-larvae and this can occur in any environment in which these life stages are abundant. During their adult life, the animals remain on the ship hulls until they die or fall-off. During this time they often reach sexual maturity and can then release gametes into the water column (e.g. Coutts and Dodgshun 2007).

The second mode of vessel-based species transport is the uptake of marine life forms, from bacteria to fish, with the ballast water that regulates the buoyancy of ships (Endresen et al. 2004; Bailey 2015). Ballast water tanks are commonly emptied or filled when cargo is loaded or released—unless the loaded freight equals the released one—and, although the conditions inside these tanks are stressful, many organisms survive transport and will then be released during the next water exchange (Gollasch et al. 2000; Klein et al. 2010).

Both ways of transport, on the hull and inside the ballast water tanks, expose organisms to stressful abiotic conditions, but the prevailing stressors differ between the two environments (Wonham et al. 2001; Minchin and Gollasch 2003). During the voyage from one destination to another, the temperature inside ballast water tanks follows the sea surface temperature with a delay of 1–2 days (Gollasch et al. 2000). Furthermore, oxygen concentration in the tanks can decrease due to consumption by the organisms present, but can be re-increased due to mixing caused by movements of the ship (Gollasch et al. 2000), while salinity may change abruptly in case the ballast water is exchanged. Ballast water exchange at sea, i.e. > 200 nautical miles from shore, is now mandatory in many countries worldwide to reduce the amount of viable stages of marine life forms inside ballast water tanks before an over-sea vessel reaches its destination (Carney et al. 2017). In contrast to ballast water tanks, oxygen limitations are unlikely at the hulls, but the ambient salinity may change as well when the ship passes through estuaries, rivers or artificial waterways (Cohen 2006; Davidson et al. 2008). Heat stress can affect both environments, for instance, when the shipping route leads from temperate to tropical waters. Hence, during a ship voyage that starts and ends in a temperate environment but leads through tropical waters (see Kaluza et al. (2010) for examples), all organisms that are present on or inside the vessel and which originate from the temperate zone can experience temperature fluctuations of more than 20 °C (Hua and Hwang 2012). Consequently, during the passage, temperatures can be reached that constitute a severe heat stress for temperate organisms and this stress can persist for days to weeks (Gollasch et al. 2000). The stress may lead to partial mortality among the transported organisms that could select for those genotypes that exhibit the largest tolerance to elevated temperature. If this is the case, the frequency of tolerant genotypes in the transported population should increase during the voyage (Sakai et al. 2001). Hence, population robustness towards this stressor should be higher at the end of the journey than at its beginning.

However, so far, there are no empirical data that document such effect and therefore we designed a global study consisting of parallel experiments that mimicked heat stress during a transoceanic voyage in laboratory environments. As test organisms we used different bivalves of the family Mytilidae. These animals, due to their capacity to attach to solid surfaces, are frequently transported as hull fouling but also in ballast water tanks (e.g. Wonham 2004; Murray et al. 2011; Casoli et al. 2016; Huhn et al. 2016a). Our test organisms came from different latitudes and were therefore adapted to different thermal environments, ranging from cold-temperate regions (Finland) to the tropics (Brazil). We specifically tested the hypothesis that stress-induced mortality enhances population resistance, measured as the survival rate during a second stress event of the same kind. At the same time, we checked whether the origin of the species influences the outcome of this test.

Methods

Collection sites and test organisms

Parallel experiments were conducted at five different locations worldwide using the laboratory facilities of marine research institutes in Niterói (Brazil), Coquimbo (Chile), Tvärminne (Finland), Kiel (Germany) and Cascais (Portugal) from April to September 2012. At these sites, specimens from one locally abundant mytilid species were collected for experiments that tested whether the average robustness of a group of individuals can be enhanced by partial mortality that is induced by a preceding stress event of the same kind.

The species chosen for this study have a wide distributional range and often occur in large abundances. Furthermore, they have a life trait that makes them likely candidates for transport by human vectors: All mytilids form byssus threads which allow them to attach to solid surfaces such as ship hulls (Aguilera et al. 2017). Therefore, they are likely bioinvaders. The following paragraphs describe the sampling sites, sampling procedures and the way of transport to laboratories. More detailed information about these aspects is provided in Table 1.

Table 1 Information about test species collection, acclimation to laboratory and experimental conditions and experimental phases
Full size table

South-Western Atlantic, Brazil

The test species in Brazil was the intertidal brown mussel Perna perna. This species is considered native in West Africa and in India (Hicks et al. (2001), while its status on the coasts of South America is under discussion. While many authors consider it native in these regions, Souza et al. (2003), after the analysis of archeological records, suggested that the species was introduced from West Africa during the 16th century. More recently, P. perna invaded the Gulf of Mexico and by this showed its invasive potential (Hicks et al. 2001). For our study, individuals of the mussel were collected at the beach of Boa Viagem in the city of Niterói in Guanabara Bay (22°54′S, 43°07′W). The coastline here is mostly rocky with interspersed sand beaches and has a tidal range of 1.3 m. Sea surface temperature usually ranges between 15 and 27 °C and is lowest in July and highest in January. The highest temperature that was recorded in Guanabara Bay during the last 15 years was 33.4 °C (source for all maximum temperature information: http://orca.science.oregonstate.edu/2160.by.4320.8day.hdf.sst.modis.php). Salinity is typically between 32 and 35. In May 2012, individuals of P. perna, which were 30-50 mm in size (umbo to the posterior end of the shell), were collected from rocks by cutting their byssus with a scalpel. Mussels were transported to the laboratory in seawater filled containers within 20 min.

South-Eastern Pacific, Chile

The mytilid species in Chile was Semimytilus algosus, which is common along the shores of Northern-Central Chile where it colonizes rocks in inter- and subtidal habitats. It is native in these regions, but is also known as a successful marine invader from the coasts of Southern Africa (de Greef et al. 2013). We collected specimens in a size range of 20–30 mm in the Bay of Guayacan near Coquimbo (29°58′S, 71°21′W) by carefully detaching the animals from buoys by cutting their byssus. This was done in May and June 2012. Animals were transported to the nearby laboratory in dry conditions within 10 min. The Bay of Guayacan is fully marine with a salinity of 34, while the sea surface temperature in this area, which is influenced by coastal upwelling, takes minimum values around 13 °C in July and maximum values of about 20 °C in January (Table 1). The maximum temperature that was recorded in the Bay of Guayacan during the last 15 years was 21.4 °C.

Northern Baltic Sea proper, Finland

In Finland, the test species was the bay mussel Mytilus trossulus, which originates from the North Pacific, but can also be found at the east coast of North America (north of Gulf of Maine) and in parts of Northern Europe (McDonald et al. 1991, Väinölä and Strelkov 2011). However, it is not clear at what time it reached the regions which are located outside its native range (Väinölä and Strelkov 2011). Very recently, it has been found as an invasive species in the Strait of Magellan (Oyarzún et al. 2016). Specimens of this species were collected from subtidal rocks near Käringarna ca. 80 km east of Helsinki (Tvärminne, 59°50′N, 23°17′O) by the use of a bottom sleigh in April 2012. Sea surface temperature in the south-eastern Baltic Sea ranges from 0 °C in January/February to 22° in July/August. The highest temperature recorded during the last 15 years was 23.8 °C. The Baltic Sea is a semi-enclosed sea area with a mostly brackish water body that exhibits no tides. Salinity in Tvärminne is between 5.5 and 7.0. From the catch, we selected intact individuals in a size range of 7.5–12.5 mm and transported them to the laboratory submersed in seawater within 30 min.

Western Baltic Sea, Germany

In Germany, individuals of the blue mussel Mytilus edulis were collected in the Kiel Fjord (54°20′N, 10°09′E), Western Baltic, in September 2012. This mytilid species is native to the northern parts of the Northwest as well as the Northeast Atlantic (Seed 1992), but has been introduced to Argentina and Chile (Seed 1992) and the Mediterranean Sea (Casoli et al. 2016). The Kiel Fjord has sandy to muddy benthos habitats with boulders here and there that constitute the only natural hard substrata in the region. Sea surface temperatures in this south-western part of the Baltic Sea cover the same range as in Tvärminne, Finland (0–22 °C), while the maximum temperature recorded during the last 15 years was 23.2 °C. Salinity fluctuates between 15 and 20. The mussels (40–50 mm) were carefully detached from pilings in the shallow subtidal by cutting their byssus with a knife. After this they were transported to the laboratory in dry conditions within 30 min.

North-Eastern Atlantic, Portugal

In Portugal, the Mediterranean mussel Mytilus galloprovincialis was used as a test species. It is native to the Northeast Atlantic and the Mediterranean, but has invaded coastlines outside this range for instance in Australia, Chile, New Zealand, North America and South Africa (Toro et al. 2005; Elliott et al. 2008; Westfall and Gardner 2010). Individuals of M. galloprovincialis were collected in a rocky intertidal environment near Guia-Cascais (38°41′N, 9°27′W) in May and June 2012. In this region, the sea surface temperature of the Atlantic ranges from 13 °C in January/February to 21 °C in July/August and the highest temperature that was recorded here during the last 15 years was 21.5 °C. The water is fully marine with a salinity of 36, while the tidal amplitude along the rocky coastline is semidiurnal with tides that can range up to 4 m. After detaching the mussels (30–40 mm) from rocks in the intertidal by cutting their byssus threads, they were placed in buckets filled with seawater for transport. Transportation time to the laboratory did not exceed 30 min.

Husbandry conditions

After arrival in the respective laboratories, all mussels were cleaned from large-sized epibionts, such as hydrozoans, bryozoans, barnacles and macroalgae, by carefully brushing or scraping off the foulers. Furthermore, we documented the size of all individuals using calipers to measure the distance from the umbo to the posterior end of the shell. Prior to the start of the heat stress experiments, animals were acclimatized to laboratory conditions for time spans varying between 1 and 3 weeks (Table 1). During this period, mortality among individuals, which, after collection, ranged between 0 and  5% per day, dropped below 1% per day in all test species. At all sites, a group of individuals (i.e. Group C with n = 20–30) was kept under these conditions throughout the entire experimental phase without any further manipulation (Fig. 1). For acclimating and keeping the animals as well as for the heat stress experiments, we used indoor or outdoor tanks of different sizes (Table 2). These tanks received seawater either from a re-circulating water cycle that consisted of a reservoir and a filter unit (in Brazil and Portugal) or from a flow-through system supplied with water from the nearby sea (in Chile, Finland and Germany). The water inside the tanks was aerated by bubbling it with pressured air. During all phases of the studies, animals were continuously fed, while the type of food and the amounts fed differed between species. Most of the mytilids were fed with DT’s Premium Blend Live Marine Phytoplankton once a day (Table 2). This is a mixture of living microalgae that contains Nanochloropsis oculata, Phaeodactylum tricornutum and Chlorella sp. with a cell size of 2–20 μm at a concentration of approximately 2.5 × 106 cells ml−1 (M. Huhn, pers. comm.) (Table 2). Furthermore, we frequently monitored the water quality in the tanks by measuring nitrite, nitrate and ammonium (with chemical test kits or photometric procedures), oxygen concentrations (with hand-held oxymeters, e.g. VWR DO 200) and salinity (with refractometers, e.g. S/Mill-E by ATAGO). Faeces that accumulated in the tanks and dead animals were removed daily.

Fig. 1
figure 1

Schematic illustration of the experimental design that was adopted at all study sites. One group of mytilid test individuals experienced two heat stress events (Group A), while a second group only experienced one (Group B). A further group of animals was kept under the same conditions as Group A and B, but was not stressed at any time (Group C). The box indicates the time span during which mortality rates were recorded in all three experimental groups

Full size image
Table 2 Husbandry conditions during laboratory acclimation and the experiment
Full size table

Experimental design

At all sites we had three experimental groups: The first group, Group A, was exposed to heat stress two times, Group B was stressed only once and Group C was not exposed to heat stress at any time (Fig. 1). For the stress events, the water temperature inside the experimental units (see ‘Experimental set-up’ section) was increased in a stepwise manner until a previously defined target temperature was reached (Fig. 1). After the target temperature had been maintained for a site-specific time span (Table 1), heat stress was relaxed by lowering the temperature back to ambient conditions at the same rate as it had been increased. During the first stress event, which was applied only to Group A, we kept several individuals of the test animals together in one experimental unit due to space constraints. This first event was, at all sites, followed by a 14–18 days-long recovery phase (Table 1), during which the water temperature was brought back to the ambient level. The length of the first stress event varied between 17 and 28 days (Table 1) and this discrepancy in lengths was because the duration of stress events was adjusted to the robustness of the test animals (see ‘Pilot studies and heat stress regimes’). The following stress event, which was applied to both Group A and Group B, was in most cases of the same length as the first stress event (Table 1). Water temperature was increased in the same manner and to the same target temperature as for the first stress event, but we placed only one individual of the test species in each of the experimental units to ensure independency between replicates.

Our response variables were the mortality rates that occurred in Group A and Group B during the second stress event. To test whether the selection of individuals that were tolerant of heat stress increased the robustness of Group A towards a second stress event of the same kind, we compared mortality rates between Group A and Group B. At all sites, the individuals in Group B stem from the pool of animals that was collected at the beginning of the study, but they were only stressed once. Furthermore, they had spent the same time in the laboratory as the individuals that were stressed twice and were kept under exactly the same conditions. During the application of heat stress, all experimental containers were inspected daily to detect dead animals and to remove faeces. At each site, a further group of test individuals was kept under exactly the same conditions as the other experimental groups throughout the course of the study, but were not stressed at all (Group C, Fig. 1). This group served to assess the background mortality among the test animals.

Experimental set-up

During the heat stress events, animals were placed in small-sized experimental units, which were made of glass or plastic and had a volume of 0.3–0.7 l (Table 2). To manipulate the temperature inside the units, a part of the larger tanks, in which we acclimated the test animals, were now used as thermobaths. For this, they were filled with water, which we warmed with common electric aquaria heaters (e.g. Aquarium Heizer 300 Watt, EHEIM Jäger). The side walls of the tanks were insulated with Styrofoam panels to reduce heat loss. Submersed pumps ensured homogenous mixing of the water body inside the baths to avoid the formation of temperature gradients. Furthermore, we monitored the thermal regimes inside of the thermobaths every hour throughout the experiments using temperature loggers (HOBO Pendant® Temperature Data Logger). The smaller sized, single experimental units were placed in these thermobaths, but their interior was completely isolated from the surrounding water body to avoid any water exchange. The units were supplied with pressurized air that was led through a diffusor stone to keep oxygen concentrations inside the containers above stressful levels. The water inside the experimental units, which were not connected to any flow-through or re-circulating system, was renewed manually every day or every other day to maintain a good water quality. During the heat stress events, the water was replaced by seawater that was previously heated to the temperature that prevailed in the containers at the moment of exchange, while the units were re-connected to the seawater flow-through or the re-circulating system during the recovery phase (see below).

Pilot studies and heat stress regimes

The temperature regimes for the induction of heat stress and the duration of the stress events differed between species (Table 1). Both settings were adjusted in a way that allowed inducing a maximum mortality of 90% in Group A within a time span not longer than 4 weeks. It was our objective to reach a mortality rate that was substantial enough to select a group of robust individuals, but which would not reach 100% to leave replicates in Group A for the second stress event during which we compared tolerance towards heat stress between Group A and B. At all sites, we conducted pilot studies with various temperature regimes to identify the most suitable stress level for the induction of mortality during the first stress event (data not shown). At the same time, we took care that the chosen temperature levels did not exceed values that can be experienced by ship hull fouling organisms during a trans-equatorial passage, e.g. in the Pacific (Brown et al. 2015).

Statistical analyses

Survival data were visualized by Kaplan–Meier curves, while in case of proportional hazards Log-Rank tests or, in case of non-proportional hazards, Peto–Wilcoxon procedures allowed to test the null hypothesis that survival rates do not differ between groups of individuals that were previously exposed to heat stress (Group A) and those that were not (Group B). These tests were done for all 5 test species investigated in this study. All graphs and analyses were produced using the free statistical computing software R (R Development Core Team 2014) with the packages ‘survival’ and ‘ggplot2’.

Results

In three of the mytilid test species, no difference in robustness towards heat stress emerged between groups of individuals that survived a preceding heat challenge (Group A) and those that were not previously stressed (Group B). This was the case for the bay mussel Mytilus trossulus from Finland (Log-Rank Test: χ2 = 0, p = 0.93), the Mediterranean mussel M. galloprovincialis from Portugal (Peto-Wilcoxon Test: χ2 = 0, p = 0.83) and the brown mussel Perna perna from Brazil (Log-Rank Test: χ2 = 0.1, p = 0.81, Fig. 2a–c). In contrast to this, mussels of the species Semimytilus algosus, which was tested in Chile, showed an increased robustness towards heat when they survived a previous exposure to thermal stress. Mortality rates in this species were significantly lower among survivors of a preceding stress event than among non-stressed conspecifics (Log-Rank Test: χ2 = 8.6, p = 0.003, Fig. 2d), while median survival time was 10 days in the first and 12 days in the second group. The same picture was observed for individuals of the blue mussel M. edulis from the Western Baltic. In this species, survival under heat stress was again significantly higher among individuals that already lived through a preceding stress event than in the non-stressed reference group (Log-Rank Test: χ2 = 8.6, p = 0.003) (Fig. 2e). Median survival time was 15 days in the first and 25 days in the second group.

Fig. 2
figure 2

a–e Survival under elevated temperatures in groups of mytilids that previously experienced partial mortality due to heat stress (Group A, solid lines) and in groups that did not experience stress-induced mortality (Group B, hatched lines)

Full size image

The mortality among individuals that were not stressed at all (i.e. Group C) during the time stress event 2 was applied to Group A and Group B is shown in Table 1.

Discussion

In a global multi-sited experimental approach with five species of mytilids, we tested whether selection of tolerant individuals by stress-induced mortality enhances population robustness towards thermal stress. We observed this effect in two of the mussel species: Semimytilus algosus from Chile and Mytilus edulis from the Southwestern Baltic Sea, i.e. in 40% of the cases. The extent to which previously experienced stress enhanced population robustness, however, varied strongly between these two species. In S. algosus, the difference in the median time to death between experimental groups was 2 days and therefore relatively small compared to the effect size that was observed for M. edulis (i.e. 10 days). It is impossible to assess whether such a small effect is relevant for invasion success after transport or not. So far, no empirical data exist that would allow to tell how big an increase in robustness in a given group of individuals of a given species has to be to enhance invasiveness significantly. However, depending on the status of the gonads in the transported animals, the release of gametes can occur at any time after the arrival in the new habitat. The longer the arrivers survive in the new environment the more likely it is that they reproduce—given that other fertile conspecifics are present.

If our findings are representative for the whole taxonomic group of mussels and the type of stressor we investigated, they indicate that an increase in population robustness through the influence of adverse environmental conditions, for instance during ship transport, could be frequently expected (ca. 40% of all cases). Furthermore, the extent to which robustness is enhanced by this process can be substantial such as in the case of M. edulis. Although our results provide a strong hint about a potential increase in robustness in adult individuals during the first weeks after their arrival in a new habitat, they do not allow to make assumptions about the further fate of these individuals after the initial phase. Furthermore, they can, of course, not give information about whether their potential offspring will survive under the conditions at the site of arrival.

Every day, numerous marine species are transported on the hulls or in the ballast water tanks of oceangoing vessels (Carlton 1999). During transoceanic voyages that cover different climate zones, these organisms can experience temperature regimes that can deviate substantially from the thermal environment that they are adapted to. Especially in cases when voyages start in temperate zones and lead through equatorial regions (Kaluza et al. 2010), transported individuals can experience severe heat stress for days to weeks. With regard to both, the magnitude of increase in water temperature and the length of the stress phase (i.e. resident time in tropical and subtropical waters), many of the worldwide ship transport scenarios are comparable to the conditions we realized in our study (e.g. Gollasch et al. 2000; Briski et al. 2011; Hua and Hwang 2012; Sutherland and Levings 2013). Hence, the mechanism that we observed in two of our five test species, i.e. an increase in population robustness as a consequence of previously experienced environmental stress, may also be relevant for marine invertebrates that are transported by oceangoing vessels. This would mean that human-mediated transport of species has the potential to modify stress tolerance and therefore invasiveness in founder populations of the translocated organisms. Ship traffic is therefore not only an important vector that allows numerous marine organisms to reach areas from which they were previously absent; it can also increase their potential to successfully invade these habitats by enhancing their tolerance to adverse environmental conditions.

An important question in this context is which of the numerous abiotic pressures occurring during oceanic voyages can mediate a relevant increase in robustness, which, at a later stage of the invasion process, enables species to establish and spread in the new habitat? The most common stressors that marine invertebrates face in subtidal coastal ecosystems, naturally and through human activities, are hypoxia (Jewett et al. 2005; Diaz and Rosenberg 2008; Rabalais et al. 2010), heat stress (Ganning 1971; Smale and Wernberg 2013) and hyposalinity (Goodbody 1961; Braby and Somero 2006). So, any selection for tolerance towards these stressors should have the potential to increase species invasiveness. Here, it is important to notice that selection for heat tolerance can, at the same time, also increase the capacity to tolerate hypoxia and vice versa (Kültz 2005). This is because many of the physiological responses to abiotic stress are universal and they can therefore mediate tolerance to several different stressors, what is known as cross tolerance (Bayley et al. 2001; Kültz 2005; Sinclair et al. 2013). Hence, even if there is a discrepancy between the kind of stress experienced during transport and the adverse environmental conditions in the new habitat, a selection that occurred during the voyage may enhance the invasiveness in founder populations of introduced species.

Our experimental approach does not allow the unambiguous identification of the mechanism that was responsible for the change in stress tolerance that we observed in two of the five species we tested. Possibly, it was a change in the frequency of stress tolerant genotypes, which followed from partial mortality among the individuals of the test populations. Alternatively, it could have been a consequence of stress hardening. The latter is part of the phenotypic response of an organism to environmental stress, which, for instance, includes the synthesis of heat shock proteins that safeguard the three-dimensional structure of enzymes under elevated temperatures (Kültz 2005). Reversible phenotypic responses of this kind mediate stress tolerance that can prevail even after the stress has ceased. This is because heat stress proteins persist for short time spans. Furthermore, their synthesis is not stopped immediately after the stress ceased but may continue for some hours or days (Malmendal et al. 2006; Brun et al. 2008; Bahrndorff et al. 2009). However, we assume that the length of the recovery phase that we inserted in between the two stress events was sufficiently long to exclude hardening as a possible explanation for enhanced robustness in pre-stressed animals. Unfortunately, we were not able to test this by quantifying the abundance of heat-shock proteins in the animals. This would have been the most direct evidence to confirm that there was no carry-over of phenotypic acclimation to previously experienced stress. Another option would have been to let the recovery phase last for several months, but this was not possible due to time constraints.

Although we clearly showed that previously experienced stress accompanied by partial mortality among the test animals can influence the resistance of survivors to a new stress event of the same kind, we observed substantial differences in the response within the group of bivalves, which are difficult to understand. The majority of mytilids did not reveal any influence of stress-induced mortality on population resistance, while none of them showed the opposite effect, i.e. pre-stressed individuals were less robust than non-stressed conspecifics. However, the reasons for the absence of an observable effect may vary from species to species. In the case of the tropical P. perna, it could be that these animals are already very close to their upper temperature tolerance limit so that the differences between genotypes with regard to heat tolerance within the population are too small to produce any measurable effect under the experimental conditions we realized (Compton et al. 2007). However, this certainly does not apply to M. trossulus from Finland and to M. galloprovincialis from Portugal. These two species inhabit marine environments in which summer maximum water temperatures usually do not exceed 22 °C and they should therefore exhibit a higher flexibility to adapt to elevated temperatures. Furthermore, the two species that exhibited an effect, S. algosus from Chile and M. edulis from Germany, both come from thermal environments that are not very different from the ones in the Gulf of Finland or at the coast of Portugal. Also with regard to other variables, which may have been important for the outcome of this study, such as the level of mortality that was induced during the first stress event, there was no pronounced difference between M. trossulus and M. galloprovincials on the one hand and S. algosus and M. edulis on the other. We can therefore not offer a plausible explanation for the absence of a measurable effect in the first two species. This also means that the latitude or thermal environment from which a population of mussels stems is not a reliable predictor of whether stress-induced mortality can enhance its robustness towards heat stress.

The size of the effect that we observed was smaller in S. algosus than in M. edulis, while the stress that was applied in these two systems was very similar. In Chile, the ambient temperature was increased by 14 °C from 14 to 28 °C, while in Germany the change was about 12 °C when we increased the water temperature from 16 to 28 °C. However, S. algosus showed a sudden increase in mortality after 10 days and both groups of test animals, pre-stressed and non-stressed, reached > 80% mortality after 14 days. In case of M. edulis, the pre-stressed group stabilized at 40–50% mortality after 15 days of exposure to elevated temperatures, while in the group of non-stressed animals, the number of survivors continuously declined until the end of the experiment after 28 days. A possible reason for the larger effect size could be that the mussels that were kept in Germany received less food than the mytilids at the other sites and therefore had less energy available for stress compensation. In this case, the influence of the pre-stress may have been more dominant.

An enhanced tolerance towards stress is presumably of particular relevance during the early phases of biological invasions, when a relatively small number of transported specimens is released into an environment that may confront them with harsh abiotic conditions. Many important harbours worldwide are located in estuaries where ambient salinity is low and arriving marine species are immediately exposed to osmotic stress that may limit their survival, establishment and further dispersal (Miller et al. 2007; Ruiz et al. 2013). The more individuals that survive this filter, the more likely it is that the introduced species can establish a permanent population. Therefore, tolerance to stress has repeatedly been suggested as a key trait that determines the invasive success of aquatic but also terrestrial species (Lockwood et al. 2007; Catford et al. 2009; Van Kleunen et al. 2010). In this context, several studies have shown that marine invasive invertebrates are generally more tolerant to abiotic stressors, e.g. hypoxia, hyposalinity and heat stress, than taxonomically related and ecologically similar native species or species that have so far not been recognized as invaders (Schneider and Helmuth 2007; Zardi et al. 2007; Schneider 2008; Gröner et al. 2011; Lenz et al. 2011; Sareyka et al. 2011; Zerebecki and Sorte 2011; Bates et al. 2013; Lejeusne et al. 2014). It is therefore vitally important for our understanding of biological invasions to identify all mechanisms that have the potential to enhance the capacity of species to tolerate adverse conditions during transport or after arrival in the new habitat. These comprise (a) the release from other environmental pressures, such as predation, competition or parasitism (Torchin et al. 2001; Keane 2002; Clay 2003; Colautti et al. 2004), because this should increase the amount of metabolic energy that is available for the compensation of stress effects; (b) the ample availability of energy, e.g. in eutrophic environments that can help to mitigate stress effects by providing energy for, e.g., the synthesis of heat shock proteins (Labarta et al. 1997; Wendling et al. 2013; Jeno and Brokordt 2014; Huhn et al. 2016b), and (c) stress-induced mortality during transport that increases the mean tolerance to stress in a group of introduced animals or plants by increasing the frequency of robust genotypes (Sakai et al. 2001). Our study, although conducted in a laboratory environment and not in situ, hints at the relevance of the last process and points at the necessity to learn more about selection acting on organisms that are transported on ship hulls or in ballast water tanks. If vessels on trans-oceanic routes are not only vectors but also a selective barrier or training field for stress tolerance in invertebrates, this is a further strong point indicating that there is an urgent need for the prevention of hull fouling as well as ballast water contamination and the establishment of worldwide management regulations.