Temporal correlation of population composition and environmental variables in the marine invader Ciona robusta

2.1 Study site and sampling

From May 2011 to June 2012, with the exception of January 2012, a total of 1,139 C. robusta specimens were collected on a monthly basis in three different locations of the Fusaro Lagoon, Casina Vanvitelliana (FCV), Foce di Mezza Chiaja (FFM), and Peschiera Romana (FPR) (Figure 1). The three sites differ in terms of substratum and distance from the Foce di Mezza Chiaja canal, the only channel allowing significant water exchange with the open sea (Figure 1). FCV is an 18th-century building connected to the mainland by a wooden bridge and located at ca. 750 m from Foce di Mezza Chiaja. Ciona robusta grows on the pillars of the bridge where it forms multi-generational aggregates at 0.5–3.0 m depth. FPR is an ancient Roman tufaceous ruin placed at ca. 550 m from the Foce di Mezza Chiaja. Here, adults are distributed over a vertical substratum of tuff bricks at 0.5–3.0 m depth. Finally, few 100 m from FPR and close to the Foce di Mezza Chiaja canal, C. robusta grows on vertical concrete rock docks at 0.5–1.5 m depth in the FFM site. Previous data indicate that a single population with high internal genetic diversity inhabits the Fusaro Lagoon (Affinito et al., 2015; Caputi et al., 2015; Sordino et al., 2008). Specimen sampling was carried out manually in randomly selected and unmanipulated 15 × 15 cm (0.44 m²) areas on artificial structures at ca. 0.5 m depth in each site. The collected specimens were promptly transported to the laboratory. There, a minimum of 20 individuals to be used for genetic analyses were immediately stored at −80°C. In addition, at least five individuals to be used for body measurements were fixed in 90% ethanol and, before biomass analysis, rinsed in running seawater to remove epibionts and any foreign body present on the surface of the tunic.

2.2 Ecological and physiological metrics

Environmental parameters (surface water temperature, salinity, dissolved oxygen, percentage of oxygen saturation, and turbidity) were measured monthly at each sampling site by immersing a multi-parametric probe (Multi Water Quality Checker U-50, Horiba) at 0.5 m depth from March 2011 to July 2012, with the exception of August 2011 (high seawater pollution), October 2011, and January 2012 (adverse weather conditions). For anatomical analyses, a total of 290 C. robusta individuals were haphazardly chosen at each collection time. Specimens were preserved in 90% ethanol upon arrival in the laboratory. For each individual, the tunic was manually separated from the body and the two parts were weighted with a Sartorius TE 1502S balance. Tunics and bodies were dehydrated at 105°C for 24 hr, and dry parts were weighed using a Sartorius R160P balance. To evaluate the current body condition (Labocha, Schutz, & Hayes, 2014) of the animals collected for the present study, we used the Condition Index (CI), defined as CI = DW_body/DW_total where DW stands for dry weight (Petersen, Schou, & Thor, 1995). According to Petersen et al. (1995), the threshold for a growing population has been set to a value of 0.46. Statistical analyses were performed in R (R Development Core Team, 2008).

2.3 Population genetics

2.4 Testing for differences among sampling sites

Given the proximity of the three sampling locations within the Fusaro Lagoon, we tested for significant differences among sites in terms of physico-chemical and genetic measures. Student's t test was performed on each of the recorded environmental variables. In addition, the microsatellite-based genetic composition of local C. robusta communities in the three sites was subject to principal coordinates analyses (PCoAs) and the Mantel test. To assess isolation by distance (IBD), we performed the Mantel test (Mantel, 1967) in the R package adegenet (Jombart, 2008), which compares the matrices of pairwise genetic and geographical distances (in kilometers). We tested the hypothesis of a correlation between these two distances in the Fusaro Lagoon. To assess spatial distance among populations, we calculated the shortest linear kilometric distance along the coast. The significance of the IBD pattern was determined with 999 permutations.

2.5 Statistical analyses

Statistical analyses on genetic data—principal coordinates analysis (PCoA) and canonical correspondence analysis (CCA)—were performed in R (R Development Core Team, 2008) using the package vegan (Oksanen et al., 2015). When performing CCAs, we tested for residual autocorrelation using the command mso in the R vegan package, which performs a permutation test for the spatial independence of residuals. A total of 1,000 permutations were performed for each dataset (environment, physiology, and trace metals). No indication of significant autocorrelation influencing the total variance was detected (data not shown). The partitioning of contingency coefficient for each of the three CCAs was analyzed in the R vegan package.

3.1 Pooling data from the three sampling sites

Specimen abundance varies across the lagoon in space and time (Supporting Information Table S1). For instance, C. robusta living at FCV and FFM sites is more prone to total depletion during seasonal die-offs than at FPR. We attribute the differences in temporal demographic structure to small-scale variation in anthropic pressure or to substratum type-specific responses. Environmental variables in the three sampling sites were not statistically different and were therefore ruled out as possible explanation (Supporting Information Tables S2 and S3). Similarly, no sign of statistically significant spatial separation was detected when analyzing the microsatellite dataset by means of principal coordinates analyses (PCoAs), a view further supported by the absence of isolation by distance as shown by using the Mantel test (Supporting Information Figure S1a,b). According to these findings, a single population of C. robusta inhabits the lagoon and, consequently, data from the three sampling locations were pooled together and used as monthly replicates (Tables 1 and 2).

3.2 Variation and correlation between population non-genetic characters and seawater conditions

The total abundance of C. robusta in the Fusaro Lagoon fluctuated considerably during the sampling time (Figure 2a). From spring to early autumn (May–October 2011), species abundance underwent a gradual and slow decline, until the population dropped to the minimum observed values from August to October 2011. Species abundance decreased between May and June in the two sampling years (2011 and 2012), suggesting a specific trend occurring in this season of the year. Rapid peak abundance was recorded in October–November 2011, promptly followed by a second major peak from February to May 2012 (Figure 2a). At the beginning of the observation period, the mean individual dry weight (mDW) showed two three-month periods of increment (May–July 2011 and August–October 2011), indicative of settlement and growth of consecutive generations (Figure 2b). The value of mDW decreased constantly from November 2011 to March 2012, followed by a further increase between April and June 2012, but lower than in the same months of 2011. The mean individual Condition Index (mCI) displayed a sinusoidal curve across the 0.46 cutoff value (Figure 2c) during the sampling period, with maximum values in May and November 2011 (0.538 and 0.563, respectively) and minimum values in February and June 2012 (0.360 and 0.370, respectively). Sea surface temperature (T) followed a typical seasonal curve (Figure 2d). On average, considering the period of March–July, year 2012 (mean T°C = 23.65, SD = 4.8) was warmer than 2011 (mean T°C = 21.04, SD = 6.4).

To evaluate the statistical inter-dependency between the variables herein measured, a cross-correlation matrix of the observed data was built by means of Pearson's r (Figure 3a). Population abundance significantly (p < 0.05) anticorrelated with mDW, suggesting reduced individual biomass in high-density aggregates; in turn, positive correlation (p < 0.05) was depicted between mDW and mWW. The only significant (p < 0.05) positive correlation between biological and environmental parameters was detected between mWW and salinity. Temperature showed a weak negative correlation with mCI (r = 0.40890214) (Figure 3a). The correlation between mCI and population abundance in the Fusaro Lagoon was maximum in November 2011 and April 2012 (Figure 3b). Additionally, values of mCI followed a U-shaped trend between May and October 2011, in concomitance with the progressive decline of species abundance (Figures 2a and 3b). This trend was not evident in 2012, maybe due to lack of summer samples. In February and June 2012, mCI correlation was peculiarly low in concomitance with a period of relatively high population abundance, whereas mCI was constantly around a 0.5 value in other months of the year 2012. However, it is of note that abundance estimates in February and June 2012 are the lowest in 2012, although higher than during most months of 2011 (Figures 2a and 3b).

3.3 Temporal changes in genetic composition

The sequencing of 80 ITS-2 fragments resulted in 16 distinct haplotypes, with an overall haplotype diversity of 0.827 and an overall nucleotide diversity of 0.05078 (Table 2). The expected heterozygosity value (H_e) decreased during the observed abundance decline period detected from May to October 2011, and remained relatively low until November 2011. The inbreeding coefficient F_is peaked in April and May in both sampling years, and decreased in winter months, with the exception of December 2011 (Table 2).

The ITS-2 median-joining network resulted in two sub-networks, named sub-network A and sub-network B, separated by 43 mutated positions and by two median vectors that indicate unsampled or extinct samples (Figure 4a). Sub-network A includes 11 strictly connected haplotypes (Figure 4a, right) sampled during the entire period under investigation, except May 2011 (Figure 4b). Sub-network B includes five loosely related haplotypes (Figure 4a, left); no haplotype belonging to the sub-network B was sampled from June to October 2011 and in April 2012 (Figure 4b, right).

To further estimate the small-scale temporal structure of C. robusta in the Fusaro Lagoon along the 13 months period, eight neutral microsatellite independent loci (Table 2) were employed. Neutrality of loci was tested using both the LOSITAN and the BAYESCAN approaches. LOSITAN putatively indicated Cin-5 and Cin-67 under balancing and positive selection, respectively. The BAYESCAN approach confirmed only Cin-67 being under moderate positive selection. The principal coordinates analysis (PCoA) performed using Cin-67 alone did not generate any clear difference compared with the analysis performed with all loci (data not shown). Evanno's method (Evanno et al., 2005) identified three distinct genetic clusters (red, blue, and green in Figure 5a). The temporal genetic structure described by the three clusters was highly variable in terms of individual assignment to each distinct cluster. A 0.8 was used to assert the assignation of each individual to any of the clusters (Table 3). Specimens in May 2011 were preferentially assigned to clusters 1 and 3. From June to August 2011, the majority of individuals in the analysis were assigned to cluster 2 and cluster 3, with few individuals assigned to cluster 1. No specimens were assigned to cluster 1 from June to October 2011 and in March 2012. Clusters 2 and 3 were represented by at least one specimen during the whole period of observation. The majority of individuals in May 2012 were assigned to cluster 3. Except for February 2012, when the three clusters were almost equally represented, dominance or co-dominance of one or two clusters was detected in all months. Unassigned individuals exceeded the amount of cluster-assigned ones during September–October 2011.

PCoA was applied to the samples in order to detect signatures of temporal differentiation between populations (Figure 5b). The two first axes of the PCoA account for 27.63% and 20.60% of the total variability, respectively. Nonetheless, no specific temporal grouping was shown by the PCoA except for monthly replicates (May and June) that grouped at the center of the plot. AMOVA of monthly samples indicated low differentiation among temporal populations and high differentiation within populations (Table 4).

3.4 Correlation of genetic lineages with physiological, physico-chemical, and toxicological data

Using data on accumulation of heavy metals in tunica and soma of C. robusta from our previous study (Arienzo et al., 2014), we correlated the frequency of individuals assigned to clusters 1–3 (or unassigned) with physico-chemical factors and concentration of heavy metals in tissues. Cluster 1 showed significant positive correlation with oxygen and anticorrelation with turbidity (Figure 6a, Supporting Information Table S4). Cluster 2 did not correlate positively with any environmental variables, but it anticorrelated with salinity (Figure 6a, Supporting Information Table S4). Cluster 3 showed a positive correlation with salinity and oxygen (Figure 6a, Supporting Information Table S4). Cluster 1 correlated with zinc concentration (soma), and cluster 3 with copper (tunic and soma) and zinc (tunic) (Figure 6b, Supporting Information Table S5). Cluster 2 showed no positive correlation, but it strongly anticorrelated with vanadium (soma and tunic) and zinc (tunic) (Figure 6b, Supporting Information Table S5). The within-cluster kinship was analyzed by means of the parameter relatedness (R) (Figure 6c). The three clusters behave in a significantly different way along the studied time span. With the exception of May 2011 (cluster 3), both clusters 2 and 3 showed high R (Figure 6c) from June to October 2011. The within-cluster R remained high until the end of 2011 for both clusters. Concerning cluster 1, R was high from December 2011 to April 2012 (despite no individuals assigned to cluster 1 were found in March 2012) and dropped during May–June 2012.

We interrogated the dataset for cluster–environment preferences by performing a canonical correspondence analysis (CCA) of the genetic, physiological, and environmental data herein collected, plus those from Arienzo et al. (2014). Although no clear results were found using environmental (Figure 7a) and non-genetic biological values (Figure 7b), three Evanno's identified clusters (Figure 5) behave differently with reference to accumulation of heavy metals, as expected from the R results (Figure 7c). In particular, cluster 1 showed a strong positive correlation with the presence of iron, barium, and aluminum metals in the tunic in May and December 2011 and in April 2012 (Figure 7c). Of note, cluster 1 reached its maximum abundance during the same months (Table 4). Cluster 2 shows strong correlation with aluminum in soma during summer 2011 (June and August) (Figure 7c). Additionally, Cluster 3 and unassigned samples show weak correlations with heavy metal concentration (Figure 7c). Finally, the partitioning of contingency coefficient for each of the three CCAs was analyzed. Results showed low (<40%) constrained values for the biological and environmental CCAs, and very high (87.75%) constrained value in the case of the trace metal CCA (Supporting Information Table S6).

4.1 Demography of C. robusta in the Fusaro Lagoon

No evidence of genetic differentiation was observed between sampling sites. This finding suggests that the distance between the sampling locations in the lagoon system falls within the range of passive dispersal of C. robusta eggs and larvae (Johannesson et al., 2018). Seasonal differences in abundance and biomass may be explained by varying habitat type, with slight differences in the influence of substratum and pollution on the survival of local settlements (Chase, Dijkstra, & Harris, 2016). The Peschiera Romana (FPR) collection site, where the species is always present, is relatively less disturbed by anthropic activities than FFM and FCV.

The annual trend is typical of warm waters, in which summer months limit the abundance of the species, maybe because of the species susceptibility to intense direct light (Jofré Madariaga, Rivadeneira, Tala, & Thiel, 2014). Using combined analysis of biometrics traits and environmental variables, wet weight resulted to be particularly vulnerable to changes in salinity, which may cause die-off of larger individuals and survival of smaller ones. Ciona robusta, whose optimal salinity in the Mediterranean Sea is 35‰, is known to reduce oxygen consumption rate to zero when salinity reaches 19‰ (Marin, Bressan, Beghi, & Brunetti, 1987; Shumway, 1978). In contrast, variation in dry weight (mDW) was not related to changes in the environmental parameters analyzed in this study, suggesting that other factors that operate in energy transfer and growth rate, such as food resources and competition for space, may influence this parameter in C. robusta. The decrease in mDW observed between October 2011 and March 2012 appears to be related to the increase in species abundance occurring in the same period, following in turn a typical yearly pattern picking up in late spring. As indicated by the Condition Index (CI), warm seawater conditions strongly limit population demography. Previously used in C. intestinalis (sensu Millar, 1953) as a suitable proxy for the energetic status of the analyzed population (Petersen et al., 1995), the CI was herein a reliable indicator of population health. Taking the various parameters together, the energetic variation from October to November 2011 (mCI = 0.563, abundance = 59.848, mDW = 2.663) is indicative of the birth of a new generation during months featuring favorable eco-physiological conditions (Toscano, 2014).

4.2 Population genetics

Although no temporal differentiation among samples was detected using either ITS-2 or microsatellites, both markers were able to identify the presence of multiple genetic pools: sub-networks A and B (ITS-2) and clusters 1–3 (microsatellites). Considering the temporal occurrence of these genetic pools, it is likely that sub-network B corresponds to cluster 1, whereas sub-network A corresponds to both clusters 2 and 3. However, as the individuals used for ITS-2 and microsatellite analyses were not the same, we are unable to clarify the issue at this stage. Mean relatedness of clusters 2 and 3 was constantly high from May 2011 to October of the same year, while becoming much more variable and relatively lower from November 2011 to June 2012. A likely explanation for the population persistence of C. robusta in the Fusaro Lagoon may be the sporadic chance recruitment from external populations or the presence of one or more unsampled pools of specimens internal to the lagoon. Of note, the number of individuals unassigned to any of the clusters suddenly doubled in November (from 23 to 46) (Table 4), evoking recruitment from unsampled internal or external sources. No C. robusta population inhabits close-by surroundings of the Fusaro Lagoon, and the closest known population (Castel Volturno) is located ca. 20 km north along the Tyrrhenian Sea coast (Caputi et al., 2015). However, the lagoon is used for mussel cultivation, making services for mussel growing, transport, and harvesting possible recruiting sources.

Recruits primarily belong to cluster 1 although not exclusively, and multiple recruitments may have occurred from November 2011 to June 2012. Being our sampling limited to a short time period (13-month), we are unable to decipher whether the abovementioned events persist over time and represent a recurring pattern of the C. robusta population in the Fusaro Lagoon. In a different scheme, the observed temporal change in genetic composition of C. robusta might reflect a process of sweepstakes reproductive success, in which a subset of the population contributes the majority of offspring to subsequent generations (Hedgecock & Pudovkin, 2011), as already suggested by the existence of site-specific haplotypes (Zhan et al., 2012). In broadcast-spawning marine invertebrates with large effective population size (Berná & Alvarez-Valin, 2014; Tsagkogeorga, Cahais, & Galtier, 2012) and low level of dispersal potential, this hypothesis deserves further investigation.

4.3 Tolerance to trace metals is increased in specific genotypes

Our results indicate that the three genetic clusters identified in the composition of the population of C. robusta respond differently to environmental variables. Interestingly, cluster 1 showed a positive and negative correlation with oxygen and turbidity, respectively. In contrast, the other two clusters correlated negatively with salinity and temperature (cluster 2) or positively with all environmental variables (cluster 3). These findings denote the coexistence of genetic lineages with different ecological profiles, likely representing genotypes with varying degrees of vulnerability to environmental factors. However, by CCAs, only trace metals may be defined as constrained explicatory variables of cluster behavior along time. Indeed, cluster 1, and to less extent cluster 2, strongly correlated with specific heavy metals accumulated in tunic and body. Of note, cluster 3 showed no correspondence with metal ions. Thus, the three clusters seem to possess distinct abilities to cope with environmental stressors in a way that maximizes fitness by ensuring species survival.

Few experimental works focused on the relationships between the genetic composition of tunicate populations and their response to stressors, and none of them analyzed the fine temporal scale of this subject. By studying the response of introduced versus native marine invertebrates to abiotic stressors at global scale, Lenz et al. (2011) reported that introduced organisms are usually more resistant to ecosystem disturbance than native ones. In ascidians, clear differences in susceptibility to stressors are observed during the life cycle, with immature stages being less tolerant than adults (Pineda et al., 2012). Herein, we provide evidence of distinct genetic pools within a confined population of the ascidian, C. robusta, and present correlative data suggesting that these lineages show differential responses to stressors.

Tolerance to trace metals may increase Ciona robusta survival in a polluted environment like the Fusaro Lagoon (Arienzo et al., 2014). It is also possible that bioaccumulation of trace metals is instrumental to a defense mechanism against natural predators. For example, high vanadium concentration in body and tunic could contribute in making ascidians unpalatable or toxic (Stoecker, 1980). It has been proved also that vanadium contents in C. robusta tissues are independent of the contents of the other metals (Arienzo et al., 2014), suggesting specific uptake and incorporation of vanadium. Our data suggest that specific genotypes may show different ability toward vanadium accumulation; in particular, cluster 2 strongly anticorrelated with vanadium concentration, whereas cluster 3 show positive correlation with vanadium in the tunic.