Results

Altogether, 533 tissue samples were collected (521 analyzed) within a period of 19 years, between 1999 and 2018 in four locations at the Puget Sound area, WA, USA. Detailed allele frequencies of all eight microsatellite loci are shown in Suppl. Table A.2. A two-tailed t-test between the Fst values (without and with ENA correction for null alleles) was non-significant (p=0.463), indicating that null alleles do not influence the results. The inbreeding coefficient (Fis) was highly significant for every population.

Local scale

Des Moines marina

In total, 190 tissue samples from seven sampling dates (years 1999-2018; Suppl. Table A.3) were analyzed. All eight microsatellite loci were polymorphic, summing up to 71 alleles. About one third of total alleles (23/71), appeared in only a single sampling period, eight private alleles were found in year 2018, three in each of the sampling years 1999, 2003 and 2013, and two in each of the sampling years 2001, 2005 and 2007. Null alleles were detected in all loci, with high occurrence (> 0.1) in PB-41 (0.284), BS-811 (0.255), PB-49 (0.229), BS-8 (0.18), PB-29 (0.175) and BS-9 (0.126). The genetic diversity indices are summarized in Table 1 and reveal fluctuating allelic richness and significant heterozygote deficiency, suggesting that the populations are not in Hardy-Weinberg equilibrium (HWE). The highest peak in allelic richness was in 2018, followed by another peak in 2003, both of which coinciding with the peaks in private allelic richness.
The Mantel test showed no significant correlation between Nei’s genetic distance and the time between the different sampling periods (Suppl. Table A.4). The Des Moines populations were overall significantly (p=0.001) differentiated (Table 2). Interestingly, the populations from 1999-2007 exceeded the 2013-2018 population’s Dest with 0.064 (p=0.001) and 0.024 (p=0.062), respectively. The estimated genetic drift (Fs’) was further stronger between 1999-2007 (0.076) than between 2013 and 2018 (0.048). Pairwise population differentiation (Fst) values showed that only the pairs 2005/2013, 2005/2018 and 2007/2013 were not significantly different (Suppl. Table A.5A). This lack of similarities between the years contradicts the BAPS analysis, which revealed only a single cluster. Yet, two clusters were elucidated by STRUCTURE (Fig. 2A). In 1999 and 2001, cluster 1 (red in Fig. 2A) was dominant, whereas the later years were more mixed, with some slight dominance of cluster 2 (yellow in Fig. 2A) in 2007 and 2013.

Shilshole marina

In total, 152 tissue samples were analyzed from Shilshole marina. Collections were performed seven times during 1999 -2018, but year 1999 was not included in the analysis as only a single colony was observed (not collected). It is also noted that year 2005 sampling yielded with just 13 samples, despite increased collection efforts. All microsatellites were polymorphic, representing however, only a single BS-8 allele in five of the sampling periods (2001, 2003, 2005, 2007 and 2018), and just one allele of BS-9 in 2005. A total of 63 microsatellite alleles were found, with 15 private alleles, occurring just in a single sampling period (Suppl. Table A.3). Four private alleles occurred in 2018, followed by three in years 2013 and 2007, two in years 2005 and 2003 and a single private allele in 2001. Null alleles were frequent (>0.1) in BS-811 (0.302), PB-41 (0.270), PB-49 (0.141), PB-29 (0.127), PBC-1 (0.126) and BS-9 (0.101). The summarized genetic indices (Table 1) reveal a gradual decline in gene diversity (He) and that the populations were not under the HWE. The highest value of allelic richness was reached in 2005, just slightly higher than in the two preceding sampling years, whereas the peak in private allelic richness was reached in 2007. Nei’s genetic distance and the time between the sampling periods are significantly correlated (Suppl. Table A.4). Each run in STRUCTURE resulted in four clusters (Fig. 2C). Cluster 1 (blue in Fig. 2C) was more dominant in 2001 and 2003, cluster 2 (pink) revealed same levels across all years, and clusters 3 (yellow) and 4 (red) were more dominant as from 2005. BAPS suggested two clusters, years 2001-2003 and 2005-2018, a construction which is in line with the STRUCTURE analysis. Pairwise Fst values suggest that samplings from years 2001 and 2003, are significantly different from years 2007, 2013 and 2018. However, also 2007 and 2013 samplings are significantly different (p<0.05) (Suppl. Table A.5B). These Fst values further validate the clustering of 2001 with 2003. There was a weak but significant overall differentiation of the populations (Dest) and a slight genetic drift (Fs’) in Shilshole (Table 2). Interestingly, Fs’ was in an order of magnitude lower in the period of 2001 and 2003 (0.004) compared to 2013-2018 (0.039). The lower genetic drift between years 2001 and 2003 is congruent with the clustering of these two periods.

Edmonds marina

Four sampling periods have yielded 101 tissue samples. In an additional sampling session (year 1999), no colony was found on the marina’s hard bottom shallow substrates. All microsatellite loci were polymorphic, yet, only a single allele was found in 2003 and 2018 in locus BS-9. Out of the 55 alleles (Suppl. Table A.3), 17 were private alleles, two private alleles in year 2003, and 5 private alleles in each of the other sampling years. Null alleles were frequent in PB-41 (0.351), BS-811 (0.279), PB-29 (0.172), BS-8 (0.166) and PB-49 (0.108). The genetic indices (Table 1) reveal an increase in allelic richness and gene diversity from 2003 to 2005, with a peak in allelic richness and private allelic richness in 2005. There was no correlation between Nei’s genetic distance and the time between the sampling periods (Suppl. Table A.4). Pairwise Fst values suggest a significant difference between years 2003/2005, 2003/2018 and 2005/2018 (p<0.05) (Suppl. Table A.5C). The weak Dest was significant (p=0.006) and the Fs’ moderate (Table 2). BAPS suggested two clusters, one for year 2003 and the second for 2005-2018. Nevertheless, the inconsistent number of K suggested by STRUCTURE proposes a very weak population structure. Two clusters were finally assigned due to constant, but not always tall Delta K peaks at K = 2 (Fig. 2D). The BAPS results were confirmed by STRUCTURE, with year 2003 being assigned in a single cluster, whereas the other three sampling dates failing to group into a specific cluster.

Shelton marina

Shelton’s marina was visited only during three sessions, yielding 78 tissue samples. All microsatellite loci were polymorphic, although in 1999, loci BS-8, BS-9 and PB-41, in 2003 locus BS-9 and in 2018 locus BS-8 were represented by just a single allele. Forty-six alleles were recorded, whereof 19 were private, e.g., found in a single sampling date (Suppl. Table A.3). Most private alleles occurred in 2003 (n=9), followed by 2018 (n=7) and 1999 (n=3). Null alleles were frequent in four loci (PB-41, PB-49, BS-811 and PB-29), ranging from 0.220 to 0.268. The genetic indices (Table 1) reveal peaks of allelic richness and private allelic richness in 2003. Nei’s genetic distance was not correlated to the time passed between the sampling events (Suppl. Table A.4). STRUCTURE suggested three clusters (Fig. 2B), however, no meaningful trend could be detected. BAPS, on the other hand, grouped the years 1999 and 2003 as one cluster, and 2018 as the second. The different populations were weakly (p=0.014) separated according to Dest (Table 2), and the Fs’ was moderate. Pairwise Fst showed a significant differentiation between 1999/2018 and 2003/2018 (Suppl. Table A.5D), which is in line with the clusters suggested by BAPS.

Regional scale - Puget Sound

Overall analysis

At the regional scale, 104 alleles were recorded during the 19-years sampling period. Nineteen private alleles were found in Des Moines (26.8% of local alleles), eight (17.4%) in Shelton, nine (14.5%) in Shilshole and eight (14.5%) in Edmonds. Analyses for the three most frequent alleles per locus (Fig. 3) revealed just a single allele in loci BS-8 and BS-9 (181bp and 194bp, respectively) dominant every year in each location, while loci BS-811, PB-29 and PBC-1 showed variable allele frequencies. In Edmonds, the third most frequent allele in locus BS-811 was highly frequent in 2003 and 2018, but absent in the years between. Only eight alleles were present in every sampled population, alleles 186bp and 189bp in BS-2, alleles 152bp and 156bp in PB-29, and alleles 199bp, 202bp, 205bp and 210bp in PBC-1. Some alleles reflect site specific distributions. For example, allele 178bp in microsatellite BS-2 was presented in all Shilshole and Shelton populations, but was absent from Des Moines and Edmonds. In contrast, allele 192bp in microsatellite BS-8 was present every sampling period in Des Moines and Edmonds, but completely absent from Shilshole and Shelton.
The Des Moines populations were almost twice as much differentiated (Dest) compared to the other three marinas (Table 2), a result further supported by the population structure Fst, revealing the highest values for Des Moines. For both indices, the differences among the four locations (Dest and Fst) were higher in the earlier period (1999-2007) than between 2013 and 2018. For the inbreeding (Fis) index, only Shelton revealed remarkable lower values. The overall genetic drift values were similar in Des Moines, Edmonds and Shelton, and less than half in Shilshole, while for allelic richness the highest numbers were assigned to Des Moines (Table 1), followed by Shilshole, Edmonds and Shelton. Des Moines also showed the highest He, also a measure of the evenness of the allelic frequencies, similar to the number of effective alleles (Brown & Weir, 1983). These high levels were followed by Edmonds, Shilshole and Shelton.
Analyzing all 20 populations simultaneously resulted in four BAPS individual clusters, where each location was assigned to a separate cluster (Suppl. Fig. A.1). STRUCTURE, on the other hand, suggested two clusters (Fig. 4A). The Des Moines samples were mainly assigned to cluster 1 (yellow in Fig. 4A), whereas Shilshole samples belonged primarily to cluster 2 (red in Fig. 4A), with Shelton’s and Edmonds’ individuals being mixed. Netstruct suggested three highly significant clusters (Fig. 4B), with Des Moines belonging mainly to cluster one (blue in Fig. 4B), Shilshole to cluster two (red in Fig. 4B) and Shelton to cluster three (green in Fig. 4B), whereas Edmonds seems to be well-mixed.
Gene flow between sites for the whole period studied was somewhat restricted (Suppl. Fig. A.2), as Shelton received gene flow just from Shilshole, and thus emerged as the most isolated site, with only 0.02 of the genetic material being received from other locations, compared to 0.06 in Edmonds and Shilshole and 0.07 in Des Moines. The strongest gene flow was from Shilshole to Edmonds (0.031), followed by Des Moines to Shilshole (0.023). Edmonds and Shelton were not connected at all, while there was a unidirectional connection from Shelton to Des Moines (0.022) and a bidirectional exchange with Shilshole. Comparing only the sites without yearly divisions shows highly significant pairwise Fst for all pairs (Suppl. Table A.5E).

Yearly analyses

Des Moines and Shilshole exhibited in most sampling dates a similar number of private alleles and generally the highest numbers (Table 3). Yet, in 2005, Edmonds had more private alleles than Des Moines and Shilshole combined. Interestingly, Shelton had much fewer private alleles than Des Moines in 1999 and all other sites in 2018. On a yearly basis, pairwise Fst between years for the whole Puget Sound area shows that all combinations except 2001/2003, 2003/2005 and 2007/2013 are significantly differentiated (Suppl. Table A.5F).

U.S. west coast scale

The analyses included the Puget Sound sites, and two Californian sites, Santa Cruz (Reem et al., 2013a) and Moss Landing (Karahan et al., 2016), involving the five shared microsatellite loci (BS-811, PB-29, PB-41, PB-49, PBC-1) used for analyses in these locations.

Overall

Allelic richness and expected heterozygosity in the Puget Sound populations were lower than in the two Californian populations (Table 4). Despite similar allelic richness and expected heterozygosity, Moss Landing and Santa Cruz populations were clustered to two different groups (orange and red in Fig. 5A) in STRUCTURE, and all four Puget Sound populations formed a third cluster (yellow in Fig. 5A). BAPS created three distinct clusters when considering the four Puget Sound sites as a single location, and no gene flow between any of the sites was discernible (Suppl. Fig. A.3). Three highly significant clusters were selected using Netstruct for the west coast analysis (Fig. 5B). Strikingly, the Santa Cruz populations clustered together to a single entity, whereas the Moss Landing populations were mostly associated with the remote Seattle populations, showing minimal similarities to the close Santa Cruz cluster.

Separate years

During four sampling years (1999, 2001, 2005, 2007) DNA samples were taken in at least a single Californian and two Puget Sound sites. Analyses on each year separately, revealed only a single genetic connection between California and Puget Sound (year 2007), a substantial gene flow (0.039) from Des Moines to Moss Landing, whereas Santa Cruz, Shilshole and Edmonds remained isolated (Fig. 6).

Global scale

For the analyses, we used 10 worldwide sites (Table 4) and five shared microsatellite loci (BS-811, PB-29, PB-41, PB-49, PBC-1). Locus BS-811 had the highest number of alleles in all locations (20-65; Table 4). When considering the total number of individuals analyzed, Puget Sound emerged as the region with the lowest number of alleles per individual on either locus and on the combined loci. This puts the whole Mediterranean area (Reem et al., 2017), Moss Landing (Karahan et al., 2016) and South America (Rachel Ben-Shlomo et al., 2010) as containing the highest numbers of allele per 100 colonies (N=57-66). The number of alleles in the Puget Sound was notoriously low (N=16), primarily on locus PBC-1, presenting just 61.5% of alleles as compared to the second lowest site, Moss Landing. The European and Mediterranean locations (without Scandinavia) showed the highest number of alleles in PB-29, PB-41 and PBC-1, and were also leading among the other two loci. The average expected heterozygosity among Puget Sound populations was by far the lowest across the globe, suggesting a limited evenness of the allele frequencies. Furthermore, the allelic richness was among the lowest in Puget Sound, whereas the highest values were observed in Israel and Santa Cruz.
The analysis on the frequent alleles (>0.1 in at least a single population; Suppl. Table A.6) revealed some disparities between worldwide sites. For example, all the frequent alleles on locus PB-29 in the U.S. west coast are under 160bp, while in the other regions at least one allele was >160bp. In the most polymorphic locus BS-811, most of the European, South American and New Zealand frequent alleles are <260bp in contrast to the North American (<300+bp). Locus PB-41 did not display different patterns among locations, except for Israel (Paz et al., 2003). While in most sites the majority of frequent alleles was below 180bp, most alleles in the Israeli populations were >180bp. For locus PB-49, allele sizes >240bp were just found in Santa Cruz (Reem et al., 2013a), the US east coast (Stoner, Ben-Shlomo, Rinkevich, & Weissman, 2002) and the Mediterranean Sea (Reem et al., 2017), while in Israel (Paz et al., 2003), Scandinavia (Reem, Mohanty, Katzir, & Rinkevich, 2013b) and South America (Rachel Ben-Shlomo et al., 2010), the majority of the frequent allele sizes was <220bp. Locus PBC-1 contained alleles >210bp everywhere except in the Puget Sound (this study), South America (Ben-Shlomo et al., 2010) and New Zealand (Ben-Shlomo et al., 2001). In the Mediterranean Sea (Reem et al., 2017), the Atlantic coasts of Europe (Ben-Shlomo, Paz, & Rinkevich, 2006) and in South America (Ben-Shlomo et al., 2010), the alleles are shorter, <190bp. In New Zealand (Ben-Shlomo et al., 2001), most PBC-1 alleles were between 190 and 200 bp, while in the Puget Sound (this study), Moss Landing (Karahan et al., 2016) and Israel (Paz et al., 2003), sizes of 201-210bp were most commonly observed.