Cheating the Locals: Invasive Mussels Steal and Benefit from the Cooling Effect of Indigenous Mussels

The indigenous South African mussel Perna perna gapes during periods of aerial exposure to maintain aerobic respiration. This behaviour has no effect on the body temperatures of isolated individuals, but when surrounded by conspecifics, beneficial cooling effects of gaping emerge. It is uncertain, however, whether the presence of the invasive mussel Mytilus galloprovincialis limits the ability of P. perna for collective thermoregulation. We investigated whether varying densities of P. perna and M. galloprovincialis influences the thermal properties of both natural and artificial mussel beds during periods of emersion. Using infrared thermography, body temperatures of P. perna within mixed artificial beds were shown to increase faster and reach higher temperatures than individuals in conspecific beds, indicating that the presence of M. galloprovincialis limits the group cooling effects of gaping. In contrast, body temperatures of M. galloprovincialis within mixed artificial mussel beds increased slower and exhibited lower temperatures than for individuals in beds comprised entirely of M. galloprovincialis. Interestingly, differences in bed temperatures and heating rates were largely dependent on the size of mussels, with beds comprised of larger individuals experiencing less thermal stress irrespective of species composition. The small-scale patterns of thermal stress detected within manipulated beds were not observed within naturally occurring mixed mussel beds. We propose that small-scale differences in topography, size-structure, mussel bed size and the presence of organisms encrusting the mussel shells mask the effects of gaping behaviour within natural mussel beds. Nevertheless, the results from our manipulative experiment indicate that the invasive species M. galloprovincialis steals thermal properties as well as resources from the indigenous mussel P. perna. This may have significant implications for predicting how the co-existence of these two species may change as global temperatures continue to rise.


Introduction
Thermoregulatory behaviour often plays a crucial role in improving the performance of ectotherms and maintaining viable populations in harsh and variable environments [1,2]. Behavioural thermoregulation is the primary method of thermal homeostasis for the many ectotherms that are incapable of thermogenesis [1], but to date, most of our knowledge of such behaviour comes from terrestrial reptiles, amphibians and insects [3,4] and to a lesser extent intertidal invertebrates [5][6][7][8]. These studies report diverse types of thermoregulatory behaviours, including posturing in lizards and gastropods [9][10][11], aggregation in snakes [12], marine snails [6], and mussels [13], habitat engineering in termites [14], active microhabitat selection in amphibians [15] and intertidal gastropods [5,6,16] and regional heterothermy in seastars [17]. In most cases, thermoregulatory behaviour has only been considered at the individual level or in the context of intraspecific interactions. Noticeably, much less is known about how the thermoregulatory behaviour of one species influences the thermal properties of another. Two intertidal mussels with different responses to air exposure dominate rocky intertidal communities on the south coast of South Africa: the indigenous mussel Perna perna and the invasive mussel Mytilus galloprovincialis. The two co-exist on shores with partial habitat segregation, M. galloprovincialis generally occurring higher on the shore, but with a region of overlap where the two occur in mixed beds [18]. During emersion, P. perna exhibits periodic closure and opening of the shell (gaping); this behaviour allows the maintenance of aerobic respiration, but increases levels of water loss and the risk of desiccation due to both evaporation and expulsion of water during valve closure. In contrast, M. galloprovincialis keeps its valves closed when exposed to air. This behaviour reduces the risk of desiccation, but decreases the efficiency of anaerobic respiration [19]. These different responses to aerial exposure, along with different byssal attachment strengths, are thought to contribute to the different vertical distributions of the two species [18,[20][21][22].
Recent laboratory experiments reveal that the gaping behaviour of P. perna is ineffective in reducing body temperatures of solitary mussels, but that when individuals are surrounded by conspecifics, evaporative cooling effects emerge for the group as a whole [23]. Consequently, increasing densities of M. galloprovincialis within P. perna beds may reduce the ability of individual P. perna to thermoregulate, whilst allowing M. galloprovincialis to 'steal' beneficial thermal properties from P. perna. This hypothesis is consistent with observations that, during periods of severe heat-stress, mortality rates of P. perna surrounded by non-gaping M. galloprovincialis in the field were higher than those of P. perna surrounded by conspecifics [23].
We investigated the thermal relationship between M. galloprovincialis and P. perna by testing the effect of varying densities of the two species on the thermal properties of both natural and artificial mussel beds during periods of aerial heat-stress of various intensities. Increasing densities of non-gaping M. galloprovincialis within the midshore region and in artificial mussel beds were expected to limit the ability of P. perna to thermoregulate and thus increase the amount of heat-stress experienced by both species of mussels. Conversely, increasing densities of gaping P. perna were expected to lower the thermal stress experienced by neighbouring M. galloprovincialis.

Study location
The study was undertaken on a moderately exposed rocky intertidal shore at Jongensfontein (34°25'12.26"S, 21°21'28.27"E) on the south coast of South Africa, where both species are abundant in the mid-shore region and display similar size class distributions (S1 Fig). The site has moderate topographic variability, with most of the shore comprising horizontal surfaces (<10°i nclination), with few crevices, pits, boulders or rock pools (Fig 1A), all of which can influence small-scale temperature variability [24]. The rocky substratum is comprised of granite and experiences a semi-diurnal tidal cycle with a tidal range of 1.5 to 2 metres. Field permits were acquired from the South African Department of Agriculture, Forestry and Fisheries (DAFF).

Natural variability
To investigate the effect of the proportions of P. perna and M. galloprovincialis on small-scale thermal properties of natural mussel beds, we took infrared (IR) images of 27 to 30 permanent 20 cm × 20 cm plots along a 50 m transect running parallel to the shore within the midshore region during the lowest daytime tidal cycles in the austral summer, between the 20 and 23 January 2015. Plots were haphazardly selected 1-2 m apart in areas with 100% cover of mussels on flat to slightly sloping (<10°inclination) rocky substrata (Fig 1B). High resolution digital images were used to estimate percent cover of P. perna and M. galloprovincialis, using the software package Coral Point Count with Excel Extensions (CPC E ) version 3.6 [25], which overlaid  a-b) Photos of mixed mussel beds within the midshore region at Jongensfontein, South Africa. The invasive mussel Mytilus galloprovincialis is easily distinguished from the indigenous mussel Perna perna by its blue colouration. (c-d) Experimental manipulations of mussel beds: artificial mussels beds were created by filling circular cages (20 cm diameter) with the desired size and density of the two mussel species (c). These were then placed high on the shore during a summer low-tide to simulate an extreme heat-event (d).
an evenly spaced matrix of 144 points on each image. Infrared images of each plot were taken using a FLUKE Ti25 IR camera twice every day during daytime aerial exposure, i.e. an hour before and after low-tide. Camera settings and specifications were comparable to other intertidal studies that have used IR thermography [26,27] with emissivity values set at 0.95, thermal sensitivities of 0.09°, and accuracies of ±2°C or ±2% of reading, whichever was greater. Maximum daily air temperatures recorded at a nearby (~10 km) coastal weather station ranged from 26.1°C to 31.3°C during the four-day sampling period (South African Weather Service 2015), producing a range of thermal conditions including extreme heat events within the rocky intertidal zone at the field site.
Mean P. perna and M. galloprovincialis surface temperatures for each plot were calculated using the single point measurement tool in the FLUKE SmartView 3.7.23 software package. Single pixels within each IR image representing the surface temperatures of 10 randomly selected P. perna and M. galloprovincialis individuals were selected and averaged to give a mean surface temperature for the two species within each plot (N.B. different individuals were chosen for each of the eight sampling events). Surface temperatures of intertidal molluscs measured by IR imagery have previously been shown to be highly correlated with internal body temperatures [28]. Within the present study, comparisons between shell surface temperatures of mussels measured by IR imagery and internal body temperatures measured with digital thermocouples (4 Channel Data Logging Thermometer 800024, SPER SCIENTIFIC Ltd.) also revealed highly significant positive correlations when assessed in the field (r 2 = 0.96, n = 122, p<0.001, y = 0.9722x+1.1239). Mussel shell surface temperature were shown to be a consistent proxy for mussel body temperature over a wide range of temperature values (from 22°C to 40°C). We therefore refer to shell temperatures measured by IR imagery as body temperatures throughout the rest of the manuscript.

Density manipulations
Artificial mussel beds were used to test experimentally whether M. galloprovincialis benefits from the gaping behaviour of P. perna and/or whether P. perna experiences increased heat-stress when interspersed with M. galloprovincialis in mixed mussel beds. These beds were constructed by removing individuals from the midshore region, cleaning the outer surface of encrusting organisms and placing them inside open circular cages (20 cm diameter) made of coarse plastic mesh ( Fig 1C; adapted from [23]). Cages were filled with either (i) 100% P. perna, (ii) 100% M. galloprovincialis, or (iii) 50% P. perna and 50% M. galloprovincialis (hereafter referred to as 50:50 mixed) with species interspersed haphazardly. This design was replicated using both large (6 to 7.5 cm shell length) and small individuals (4 to 5.5 cm shell length) to test whether the effect of species composition on thermal properties of mussel beds varies depending on the size of individuals. Once assembled, all cages (n = 3 per treatment due to limited number of large M. galloprovincialis) were submerged in a large open rock pool for 90 mins before simultaneously being removed at midday and interspersed across a large flat section of rocky substrata high on the shore, where they remained for 120 mins (Fig 1D). This rock pool was chosen because it was large and remained permanently open to the nearshore environment. Thus the water temperature experienced by mussels during this incubation period would have been equivalent to that experienced by submerged mussels on the rocky shore during high tide. Many individuals had reattached themselves to neighbouring mussels using their byssal threads whilst being submerged. This suggests mussels were responding well to being removed and were not physiologically stressed before the beginning of the experiment.
A combination of IR thermography and temperature data loggers was used to measure the temporal and spatial variability in heat-stress amongst treatments. Data loggers (iButton DS1923L) with a thermal resolution of 0.0625°C and an accuracy of 0.5°C were placed on top of each artificial mussel bed to measure changes in bed temperature at a sampling frequency of 2 mins over the course of the experiment. Temperatures recorded by these iButtons were found to be highly correlated with mean body temperatures of several individual mussels within artificial beds and therefore represent good proxies for measuring changes in bed (or habitat) temperatures (J.A. Lathlean, pers. obs). IR images of each artificial bed were taken every 5 mins, starting immediately after the cages were removed from the water, in order to assess both temporal and spatial variability in the body temperatures of multiple individuals within the same bed. Ten randomly selected P. perna and/or M. galloprovincialis individuals were selected within the centre of each cage to limit potential edge effects for body temperature measurement using the single point measurement tool in the FLUKE SmartView 3.7.23 software package. These ten measurements were averaged to give mean estimates of body temperatures of the two species within each bed.

Data analysis
Pearson correlation analysis was used to test whether the percent cover of P. perna within a plot influenced absolute temperatures in mean body temperatures of P. perna and M. galloprovincialis during the eight sampling events. A three-way ANOVA was used to test whether mean body temperatures of mussels differed between (i) species (orthogonal, fixed, 2 levels: P. perna or M. galloprovincialis), (ii) sampling dates (orthogonal, fixed, 4 levels: 20 th , 21 st , 22 nd , 23 rd January), and (iii) time of day (orthogonal, fixed, 2 levels: before and after low-tide). This was undertaken to assess whether mean body temperature of mussels differed between the eight sampling events due to the prevailing abiotic conditions. For the manipulative experiment, two-way ANOVAs were used to test whether density treatments (orthogonal, fixed, 3 levels: 100% P. perna, 100% M. galloprovincialis, 50:50 mixed) and the size of individuals (orthogonal, fixed, 2 levels: large or small) influenced bed or body temperatures of P. perna and M. galloprovincialis within artificial mussel beds. These two-way ANOVAs included separate tests for both absolute temperature and rates of change (degrees per min) for each of the response variables, both halfway through and at the end of the experimental period. Data were normally distributed and showed equal variances using the Shapiro-Wilks test for normality and Cochran's test, respectively.

Natural variability
Body temperatures of P. perna did not vary in response to differences in the percent cover of P. perna, irrespective of sampling date or whether measurements were taken before or after lowtide (Fig 2, Table 1). In addition, body temperatures of P. perna were not significantly different from those of M. galloprovincialis (Fig 3), irrespective of sampling event. Instead, body temperatures of both species appeared to be influenced by the timing of low-tide, sampling date and the time of day (Fig 3, Table 2). For example, irrespective of the percent cover of P. perna, body temperatures of both mussel species were greater (i) on days when the timing of low-tide was closer to midday, (ii) after low-tide compared to before, and (ii) on the 22 and 23 January 2015 when air temperatures were particularly high (Figs 2 and 3, Table 2). Body temperatures of individual mussels typically differed by 6°C to 8°C within a single 20 cm × 20 cm plot and at times by as much as 10.4°C and 11.8°C for M. galloprovincialis and P. perna respectively across the entire midshore region (Fig 3). Noticeably, this variability was highest during the hottest day of our sampling experiments, 23 January (Fig 3).

Density manipulations
Mean body temperatures of M. galloprovincialis within monospecific beds reached higher temperatures than P. perna placed within monospecific beds (Fig 4, Table 3). This was true irrespective of whether beds were comprised of small or large individuals (Table 3). Body temperatures of both M. galloprovincialis and P. perna were similar within mixed mussel beds and after 80 min of exposure body temperatures of M. galloprovincialis within mixed mussel beds were significantly cooler than their monospecific counterparts, and vice versa for P. perna within mixed mussel beds (Table 3). Body temperatures of mussels increased significantly faster within 100% M. galloprovincialis and mixed mussel beds compared to 100% P. perna (Table 3). Body temperatures also increased faster within beds comprised of large as opposed to small individuals (Table 3). However, there was no interaction between density treatment and size of individuals within artificial mussel beds (Table 3). Relationship between percent cover of P. perna and the mean surface temperatures of P. perna within naturally occuring midshore mussel beds during four consecutive day-time summer low-tides. Left column represents measurements taken approximately one hour before low-tide; right column represents measurements taken approximately one hour after low tide. As for body temperature, differences in bed temperatures and heating rates were largely dependent on the size of mussels within the bed (Fig 5, Table 4). In contrast to body temperatures, however, bed temperatures and heating rates were greater amongst mussel beds comprised of small individuals. Whilst not statistically significant, bed temperatures increased faster and were somewhat greater amongst cages filled with either 50% or 100% M. galloprovincialis than mussel beds comprised of 100% P. perna, the difference increasing with time ( Fig 5).

Discussion
Artificially manipulating the densities of two competing intertidal mussel species significantly reduced the capacity of the indigenous species P. perna for group thermoregulation during periods of heat-stress. Conversely, within mixed mussel beds, the invasive mussel M. galloprovincialis appeared to benefit from the evaporative cooling behaviour employed by P. perna since body temperatures of M. galloprovincialis in mixed beds remained significantly cooler than conspecifics in monospecific beds. The results also suggest that irrespective of species or thermoregulatory behaviour, mussel beds comprised of larger individuals are more effective at reducing thermal stress across the entire mussel bed. This supports current physiological principles that suggest the greater body mass of large individuals will act as a buffer against changes in abiotic conditions and help maintain relatively stable body temperatures [2]. However, it should be noted that the size of individual mussels within beds only affected bed temperature and not the body temperatures of individual mussels and that heating rates of body temperatures increased with an increase in the size of individual mussels within a bed. The cause of this discrepancy between mussel bed and mussel body temperatures is unknown.
The thermal relationship between the two species also supports previous laboratory and field-based studies showing that mussels within monospecific beds of P. perna are able to maintain lower body temperatures and show better survival than mussels within monospecific M. galloprovincialis beds [23]. The fact that M. galloprovincialis 'steals' the cooling effect produced by gaping P. perna may contribute to the ecological success of M. galloprovincialis, especially along the southern coast of South Africa. More importantly, however, it indicates how the response of a species to changes in ambient temperature can be shaped by indirect interactions with other species. Table 1. Pearson correlations coefficients (r) for comparisons between the percent cover of P. perna and absolute/ relative temperatures in surface temperatures of P. perna and M. galloprovincialis measured one hour before and one hour after low-tide (LT) across four consecutive days in summer. Bold font highlights significant results.  Unlike artificial beds, body temperatures of M. galloprovincialis and P. perna in natural mussel beds within the midshore region did not vary as a function of species composition. Instead, body temperatures of both species appeared to vary depending on the timing of lowtide, the time of day and differences in maximum air temperatures, which supports previous findings [29,30]. Temperature variation amongst plots, and individuals within plots, was also high and comparable to that found in other intertidal studies [31][32][33]. Differences in the effect of species composition between natural and artificial beds could be related to the fact that the size and structure of natural mussel beds is inherently more variable (e.g. mixed size classes, multi-layered) and distributed across a wider range of topographies than the artificial beds used in this study. Additionally, mussels in natural beds were often covered with encrusting flora and fauna whilst those in artificial beds were not. Thus, variability associated with smallscale differences in topography, size-structure, mussel bed size and the presence of encrusting organisms could have masked the effects of thermoregulatory behaviour within natural mussel beds.  Increasing temperatures associated with climate change are expected to favour positive biological interactions [34]. However, as average global temperatures continue to rise, we might expect the negative effects of M. galloprovincialis on P. perna to become stronger. For example, many intertidal communities are expected to undergo further 'coastal squeeze', which results from rising sea-levels forcing species higher up the shore and higher atmospheric temperatures  and heat-stress forcing species further down, increasing competition for primary substrata [35].
Consequently, due to reductions in the availability of primary substrata alone, we might expect increased levels of competition between P. perna and M. galloprovincialis, thus theoretically reducing the ability of P. perna to thermoregulate. This might not initially have negative impacts on populations of P. perna since small increases in body temperatures of ectothermic species may increase metabolism, growth and reproduction. However, if temperatures surpass a physiological threshold the benefits of increased body temperatures will soon become detrimental.

Conclusion
The results of our study highlight the complexity of interspecific interactions and how these can influence thermoregulation of one or more species. We show that small-scale changes in the relative abundances of an intertidal ectotherm can influence the thermal characteristics of another ectotherm through differences in thermoregulatory behaviour. This finding highlights the fact that the responses of species to changing environmental conditions need to be viewed within the context of both direct and indirect biological interactions.