Stress related epigenetic changes may explain opportunistic success in biological invasions in Antipode mussels

Different environmental factors could induce epigenetic changes, which are likely involved in the biological invasion process. Some of these factors are driven by humans as, for example, the pollution and deliberate or accidental introductions and others are due to natural conditions such as salinity. In this study, we have analysed the relationship between different stress factors: time in the new location, pollution and salinity with the methylation changes that could be involved in the invasive species tolerance to new environments. For this purpose, we have analysed two different mussels’ species, reciprocally introduced in antipode areas: the Mediterranean blue mussel Mytilus galloprovincialis and the New Zealand pygmy mussel Xenostrobus securis, widely recognized invaders outside their native distribution ranges. The demetylathion was higher in more stressed population, supporting the idea of epigenetic is involved in plasticity process. These results can open a new management protocols, using the epigenetic signals as potential pollution monitoring tool. We could use these epigenetic marks to recognise the invasive status in a population and determine potential biopollutants.

Epigenetics, or chemical signatures of DNA that can activate or silence genes (methylation 1,2 or histones (acethylation/de-acethylation 3,4 )), are rapid mechanisms of response of Eukaryotes to environmental challenges 5,6 . Perhaps the greatest challenge that a marine organism may undertake is to be transferred to a distant location mediated by a rapid vector, for example associated to human transport, much faster than the natural dispersal rate. During the travel, an organism may survive under adverse conditions [7][8][9][10] , and in the location of introduction it may need to adapt to a new environment that likely differs from the native one.
In some cases, translocated species are able to adapt easily to the recipient ecosystem, proliferate and may even become invasive [11][12][13] . The role of epigenetics in biological invasion processes, in particular methylation, has been discussed earlier 14 , but was deeply studied only in a few cases. Higher epigenetic than genetic variation reported in introduced populations of plants 15 and vertebrates 16 suggests that epigenetics may compensate for reduced genetic diversity in incursion areas, e.g. due to founder effects 17 . Decreased global methylation has been proposed as a mechanism for enhancing plasticity, thus tolerance to new conditions, in recent invertebrate invasion events 18 . Nevertheless, this phenomenon needs to be further explored, as it has enormous ecological implications. It could explain unpredictable invasive success of some species able to acquiring novel traits and tolerances in the recipient ecosystem. Niche shifts that facilitate expansion of invaders 19,20 could be at least partially explained through epigenetic mechanisms that are more rapid than genetic adaptation by natural selection. However, as long as the introduced population settles down in the recipient ecosystem and leaves the expansive phase, extra plasticity would no longer be necessary and the levels of methylation are expected to reverse to species' normal state, approaching those and be similar to those of native populations. This phenomenon would be reflected in increased methylation in older introduced populations versus more recent ones. In other words, generalized hypomethylation would be a signature of early (versus old) introductions 18 . Sampling sites, specimen collection and stress factors. Mussel specimens were sampled simultaneously from two sites of different environments in south Europe and two in New Zealand, where M. galloprovincialis and X. securis are respectively native to and introduced. The summarized information on the sampled populations is provided in Table 1. Two sites from Oceania, New Zealand (Nelson and Havelock, South Island) and two from south Europe (Aviles, Spain and nearby Vidourle, south France) were sampled for Mediterranean blue and pygmy mussels. Adult individuals of the two species were identified de visu and taken at random from the sampling sites (at least 15 per population and species). Specimens were preserved in absolute ethanol for further analysis. The sampling locations were chosen to account for difference in salinity preference of the considered species and pollution as environmental stress. Thus, Vidourle (France) is a protected coastal lagoon (included in Natura 2000 network) with brackish water, affected by seasonal inflows from Vidourle river and little anthropogenic influence. The levels of anthropogenic pressures in Vidourle can be compared to Havelock (Pelorus Sound, New Zealand), located in sounds at the mouth of two rivers far from industrial pollution and big urban centres. Both locations, as estuarine habitats, are subjected to eutrophication pressure and freshwater inputs (Havelock 49,50 ; Vidourle lagoon 51 ). On the other hand, Aviles (Spain) and Nelson (New Zealand) are both international commercial ports with busy shipping traffic, adjacent to regional urban centres and dominated by marine waters (with some flushing effect from local rivers). All sampling locations have low level of exposure to the open sea and resembling tidal ranges, therefore the effect of these natural stress factor was considered negligible and not accounted for in this study. General environmental information (temperature and salinity) for the sampling areas was collected from the online resources (e.g. National Agency of Meteorology, Global Sea Temperature website, ClimaTemps website, National Ports -Spanish ports at http://www.puertos.es/es-es), published national and regional reports and research papers.
Three different stress factors (invasion-related, natural and anthropogenically driven) have been considered. (i) the population age as a proxy for early invasion stress; (ii) salinity stress in relation to reported tolerance ranges for each species; and (iii) habitat degradation, represented by anthropogenic pressures assumed for each location. Each factor was scored as follows: • Early invasion stress (=population age). Scored 1-5 and relates to the invasion-naturalisation status of the population, where 1 corresponds to native population, 2 -to "old" introduction (>30 years) and therefore supposedly fully naturalized population, 3 -to 30-10-year introduction, 4 -to relatively recent introduction (10-5 years), and 5 -to very recent introductions (<5 years) ( Table 1). The introduction age was estimated from literature and official first reports of species detection whenever possible (X. securis in Europe: 2014 in Aviles versus 1992 in Vidourle). For M. galloprovincialis in New Zealand a score of 2 was considered, since, although non-native, the Atlantic lineage is suspected to be there for a long time, > 30 years at least 37,52 . • Salinity stress. Scored 1-4, since the range of differences among sites for this factor was comparatively shorter than for the introduction age.
For Xenostrobus securis, although tolerates salinities from 2 up to 30 or higher 42 , it prefers brackish waters (5 to 28 ppt 53 ) and the upper tolerance range of developing larvae is around 18 ppt. Therefore, salinity >30 ppt could be considered as suboptimal (i.e. stressful). For scoring this data, 1 correspond to places with salinity between 5 and 28 ppt., 2 -to places with salinity between 2 and 30 ppt., 3 -to places with salinity sometimes out of range, and 5 -to places with salinity out of range always. For M. galloprovincialis, also tolerant for a wide salinity range, significant salinity stress as measured from heart rate occurs below 22 ppt 31 . Thus <22 ppt was considered salinity stress for this species. For scoring this data, 1 correspond to places with salinity below to 22 ppt., 2 -to places with salinity below to 22 ppt. sometimes during the year, 3 -to places with salinity below to 22 ppt. most of the time, and 5 -to places with salinity below to 22 ppt. always. • Habitat degradation (=anthropogenic pressure). Scored between 1-4 based on pollution and other factors reported from the studied sampling sites (Table 1), where 1 correspond to protected area, 2 -to areas with continuous sediment loads around, 3 -to port areas with local industries such as forestries and farmlands, and 4 -to port areas with industrial big pollutions like heavy metals.
Mapping vulnerability is most commonly performed by combining multiple indicators into single indices of vulnerability for a given stressor under a given dimension, and then combining multiple indices in order to build an overall, relative estimate of vulnerability 54,55 . These "combinations" are usually simple arithmetic or geometric means, based on the Multiple Attribute Utility Theory (MAUT) that is widely used in economics, engineering, decision science, development studies and, to a lesser extent, social sciences [56][57][58][59] . In order to analyze the synergetic effect of considered stress factors on methylation patterns, additive and multiplicative aggregation of stressor scores was applied, following vulnerability assessment framework described by Hinkel (2011). This approach is widely used in the Indicator Based Vulnerability Assessment (IBVA), where aggregation approaches (e.g., weighted additive or multiplicative aggregation) based on MAUT are applied 60 .
Ethical statement. This study has been carried out on invertebrate species. Since they are reported as invasive in various regions, the materials and clothes employed in sampling and manipulation of individuals were carefully cleaned and disinfected to avoid further dispersion of these organisms or their propagules in the environment. This study adheres to the European Code of Conduct for Responsible Research. DNA analysis. Total DNA was extracted from the foot muscle of collected specimens using the E.Z.N.A Mollusc DNA kit (IOMEGA, Bio-Tek) following manufacturer's instructions. DNA samples were stored at 4 °C until analysis conducted. Aliquots were long-time stored at −20 °C. DNA quantification was carried out by fluorometric methodology with Qubit ® 2.0, and normalized to 100 ng/µl. MSAP (methylation-sensitive amplified polymorphism) methodology was applied to measure polymorphism in DNA methylation patterns. The global methylation was analysed following Díaz-Freije et al. 61 . This protocol was also described in detail in Ardura et al. 18 . Briefly, two aliquots of 100 ng DNA per sample were treated separately with EcoRI/HpaII and EcoRI/MspI. The enzymes (MspI and HpaII) recognize and cleave CCGG target sequences, depending on the methylated status in the inner and/or outer C 62 63 . PCR products were loaded with GeneScan GS-500 LIZ3130 size standard into an ABI Prism 3100 Genetic Analyzer (Applied Biosystem).
Fragment analysis and AFLP scoring were performed with GeneMapper v.4.0 software (Applied Biosystem). The following settings were applied for quality AFLP reading: analysis range, 50-500 base pairs (bp); minimum peak height, 50 relative fluorescence units; pass range for sizing quality: 0.75-1.0; maximum peak width: 1.5 bp; maximum peak height ratio: 1.8 (higher peaks were removed); normalization method: sum of signals. To confirm AFLP reproducibility five X. securis samples were analysed again following the same protocols. Data analysis. Individual and population MSAP profiles were analysed with the R package msap v.3.2.2 64 . The software yields a score matrix according to the methylation state, using the four patterns from presence-absence matrices obtained with the EcoRI-HpaII and EcoRI-MspI primer combinations (Type I to IV) described by Salmon et al. 62 .
Type IV variation could be due to high methylation status but also to mutations in restriction sites, thus real methylation state cannot be specified 62 . Following Fulnecek and Kovarik 65 , we have considered type II and III together as methylated loci. Finally, type I restriction sites are not methylated. The global methylation level was measured as the proportion of methylated loci (Types II and III) over the scorable loci (Types I, II and III), as in Nicotra et al. 66 and Ardura et al. 18 .
In methylation-susceptible loci (MSL) the observed proportion of methylated states across all samples exceeded a user-defined error rate-based threshold (ERT; 5% by default). The rest of loci were categorized as non-methylated (NML). Only polymorphic fragments with at least two occurrences of each state were used for subsequent analysis (Herrera & Bazaga 2010). MSL were used to assess epigenetic variation, and NML -to assess genetic variation since their banding pattern depends exclusively on changes in the sequence at the restriction target sites.
The R package msap v.3.2.2 64 and GenAlEx software 67,68 were employed for the following analyses. For both MSL and NML the amount of overall variation was estimated using the Shannon diversity index (I). Differences between Shannon's indices between MSL and NML were tested with the Wilcoxon rank sum test with continuity correction (W). The epigenetic (MSL) and genetic (NML) differentiation among populations and between pairs of populations were assessed by means of ɸ ST values, and principal coordinates analyses (PCoA) followed by analysis of molecular variance (AMOVA) 69 .
Software PAST 70 was employed to compare the mean proportion of methylated loci per individual between populations using conventional t-tests, after testing variance equality through F test. Analysis of residuals was performed and normality was assessed using Shapiro-Wilk test. A Principal Components Analysis (PCA) was carried out with the same software to identify the relative contribution of the different stressors to the dataset variance, after checking compliance of the necessary conditions in the dataset: applying correlation function and 0.7 eigenvalue cutoff. Correlation between methylation and stressors was calculated using Pearson's r after checking normality, with 0.05 significance threshold and applying Bonferroni correction whenever relevant. In the global analysis, taking each population as a unit, both MSL and NML differed significantly among the four studied populations of the two species, with significant Ф ST values for the two types of loci (Table 2). These values represent genetic differentiation between populations based on different alleles (or methylation patterns) in the loci analysed. The four ANOVAs (for the two species and locus type from methylation sensitiveness) were statistically significant, with p < 0.01 in all cases ( Table 2).

In
In terms of genetic differentiation (variation at NML), native-native and introduced-introduced pairwise Ф ST were both significant for X. securis, and the absolute value was greater between natives (Ф ST = 0.262 and 0.181 respectively, both with p << 0.001). For M. galloprovincialis only native-native pairwise Ф ST was significant (Ф ST = 0.036, p << 0.001), while introduced-introduced comparison was not (Ф ST = 0.013, P = 0.914). For M. galloprovincialis, differences among populations, although significant between native populations, did not separate completely the four populations for any type of loci. In PCoA analysis, more apparent differentiation of the Mediterranean (Vidourle) population was observed for MSL (Fig. 1a) while for NML the apparently more differentiated population was the Atlantic one (Aviles) (Fig. 1b). For X. securis, the populations were clearly separated based on MSL (Fig. 2a), while based on NML Aviles and Havelock populations were clearly apart but Nelson and Vidourle were more overlapped (Fig. 2b).
Given significant among-population differences in epigenetic variation in the reciprocally antipode mussels analysed here, the proportion of each type of methylation status for MSL loci was also analysed (  Table 2. Global statistical analysis of methylation sensitive loci (MSL) and non-methylated loci (NML) in the two mussel species analysed.   Table 4). The stressor scores were different in the analysed populations and species (Table 4), depending on the species and population characteristics as explained above. In the PCA, the components 1 and 2 together contributed with >76% variance ( Table 5). The relative contribution of stressors to the total variance was higher for population age in the component 1, salinity in the component 2, and anthropogenic pressure in the component 3 (Table 5), methylation being located opposite to the three stressors in the scatter plot of the principal component analysis (Fig. 3).
In the eight populations, no single stressor had significant correlation with global or mean individual methylation (data not shown). The sum of stressor scores for each population (Table 4) was negatively correlated with the mean individual methylation (r = −0.804, 6 d.f., P = 0.016) and the global methylation (r = −0.786, 6 d.f., P = 0.021), both significant after Bonferroni correction. For multiplicative model (Table 4), the correlation was also significant for both mean (r = −0.791, P = 0.019) and global methylation (r = −0.800, P = 0.017). Pairwise effects of stress factors were also assessed for summative and multiplicative models, applying Bonferroni for statistical significance. For multiplicative model, the results have shown significant correlation of salinity and anthropogenic pressure combination with mean methylation; salinity and population age combination with global methylation, and population age and anthropogenic pressure with either global and mean methylation    (data not shown). While for summative model, only the combinatory effects of anthropogenic pressure and population age with global methylation were significant (r = −0.783, P = 0.022).

Discussion
The results presented here reveal, for the first time, the evidence of epigenetic variation, through changes in DNA methylation, between two reciprocally introduced marine related species with similar biological traits, life style and invasive capacity. Our findings suggest that this variation can be explained by invasion-related stress factors including population age. Epigenetics being involved in the process of establishment in biological invasions had been demonstrated already 6,18,71 . Previous study from the same authors (Ardura et al. 18 ) hinted that the expected influence of environmental stressors on methylation could be detected in native populations without the "invasive" signature. For this reason, the idea of epigenome analysis of donor and introduced populations together with the population status assessment (early invasion phases) and environmental stress factors was challenged here.
The results of this new study supported that hypothesis, suggesting early invasion stage as a stressor, to be added to other environmental stressors as a trigger to epigenetic changes. Expansion into a new territory would require mobilization of physiological resources, and demethylation may facilitate the process by unblocking genes previously silenced in native (optimal) conditions. In molluscs, although previous studies have reported that highly expressed genes show the highest degree of methylation [72][73][74] reported a negative relationship between DNA methylation and expression of certain homeobox (hox) genes during embryonic development in C. gigas. This implies that DNA hypomethylation is potentially linked to the transcription of genes potentially involved in phenotypic plasticity and adaptation 75 . Thus, demethylation in response to exposure to pollutants was reported for example in humans 22 and fish 26 . Karan et al. 27 described hypomethylation associated to salinity stress in rice, and Lu et al. 76 reported changes in methylation patterns induced by salt treatment in cotton. In our study, single stressors did not correlate significantly with decreased methylation pattern at population level, but rather had a combinatory effect. The stress factors would be acting here mainly in a multiplicative way, thus suggesting a certain synergetic effect. In fact, the effect of co-exposure to different stressors is rarely summative, since their combined actions may synergistically multiply the negative effects on the organism or, conversely, one stressor may mask the effect of other stressor (see for example Manti & D' Arco 77 for a review). The same may apply to epigenetic responses to multiple stressors, as experimentally proven on different model and non-model species (reviewed inanalysed Vandegehuchte & Colin 78 ). However, the present study is observational and cause-effect associations cannot be proven from it. Experiments subjecting the two species to different levels and combinations of the stressors could be suggested for further empirical verification.
The levels of methylation and epigenetic variation, as well as their distribution in the genome are very diverse within and among phylogenetic groups, despite to be an ancient evolution mechanism 75 . It has been reported earlier, that epigenetic mechanisms may facilitate the establishment and spread of invasive species and important pests 79 (reviewed in Hawes et al. 80 ), and we could expect diverse levels of methylation signalling this. The global DNA methylation level reported in this study for the two analysed species (>50%, up to 70% in some cases) was considerably higher than the level reported earlier in other molluscs. For example, Zhikong scallop Chlamis farreri exhibited <30% global methylation in different tissues including muscle with dominant Type II (internal cytosine methylated) loci 81 . However, our results are not exceptional. Based on the current results, in both M. galloprovincialis and X. securis Type III methylation (hemimethylated loci) was more prevalent than Type II, and this is consistent with findings from recent epigenetic study of Octopus vulgaris 82 . Using the same method, high and variable methylation levels were estimated in other organisms: 70% in Hordeum vulgare 83   Ficopomatus enigmaticus 18 . This suggests that different organisms in general and mollusc species, in particular, likely exhibit different patterns of global methylation even at intraspecific level.
Huang et al. 83 obtained evidence that suggest that stress-induced DNA methylation variation can contribute to the rapid acclimatization to sudden environmental changes at the individual level, and the increased intrapopulation variation may maximize survival when invaders cross environmental barriers during biological invasions. In the studied mussels, demethylation was generally higher in populations subjected to greater stress levels. This would support the idea of greater plasticity favoured through epigenetic changes in invasive species undergoing adaptations to new environments. If proven true with further case studies and experiments, this discovery would have enormous implications for the management of invasive species and native pests. For example, eradication procedures involving physical or chemical treatment and causing sub-lethal stress to target organisms can potentially induce demethylation in the survivors, allowing for higher plasticity, and likely -more aggressive invasive behaviour. Therefore, if total extirpation of the unwanted population is impossible (or impractical), smooth ecosystem-based mitigation approaches should be considered. For example, suppression of invasive population by mechanical removal 84 , targeted harvesting of species (if applicable) or biocontrol 85 , complemented with enhancement of local biotic resistance by supporting native biodiversity 84 and improving the overall ecological status of the ecosystem 86 could be recommended.
As a last remark, epigenetics should be explored as a tool for determining invasiveness of non-indigenous species and assessing the potential associated risks. In a world of increased interchange of species across latitudes, it is impossible to prevent introductions completely. Therefore, along with minimising the risks of new incursions, by implementing national and international legislation initiatives (like IMO Ballast Water Convention, New Zealand Craft Risk Management Standard, ICES Code of Practice on the Introductions and Transfers of Marine Organisms), an effort should be put to enable better decisions about where to invest limited resources and get the best possible outcomes for biosecurity. Perhaps, epigenetics could assist in that, providing informed advice on the species with invasive potential enhanced by specific methylation patterns, and thus of the top priority for management and rapid response measures. Suarez-Ulloa et al. 87 = highlighted the importance of epigenetics as a potential tool for pollution monitoring, using marine invertebrates as model systems. This recommendation could be extended to marine biopollution by nuisance organisms. Ultimately, finding epigenetic markers specific to high-profile invaders could empower development of effective preventive frameworks and overall reduce the ecological threats caused by biopollution in the oceans worldwide.