Genetic Analyses of Saprolegnia Strains Isolated from Salmonid Fish of Different Geographic Origin Document the Connection between Pathogenicity and Molecular Diversity

Saprolegnia parasitica is recognized as one of the most important oomycetes pests of salmon and trout species. The amplified fragment length polymorphism (AFLP) and method sequence data of the internal transcribed spacer (ITS) were used to study the genetic diversity and relationships of Saprolegnia spp. collected from Canada, Chile, Japan, Norway and Scotland. AFLP analysis of 37 Saprolegnia spp. isolates using six primer combinations gave a total of 163 clear polymorphic bands. Bayesian cluster analysis using genetic similarity divided the isolates into three main groups, suggesting that there are genetic relationships among the isolates. The unweighted pair group method with arithmetic mean (UPGMA) and principal coordinate analysis (PCO) confirmed the pattern of the cluster analyses. ITS analyses of 48 Saprolegnia sequences resulted in five well-defined clades. Analysis of molecular variance (AMOVA) revealed greater variation within countries (91.01%) than among countries (8.99%). We were able to distinguish the Saprolegnia isolates according to their species, ability to produce oogonia with and without long spines on the cysts and their ability to or not to cause mortality in salmonids. AFLP markers and ITS sequencing data obtained in the study, were found to be an efficient tool to characterize the genetic diversity and relationships of Saprolegnia spp. The comparison of AFLP analysis and ITS sequence data using the Mantel test showed a very high and significant correlation (r2 = 0.8317).


Introduction
Saprolegniasis causes great damage and infection in fish in aquaculture and fish farms [1], and Saprolegnia parasitica is recognized as one of the most important oomycetes pests of salmon and trout species in Scandinavia, Chile, Japan, Canada and Scotland [2]. Thus, it causes losses of tens of millions of dollars in aquaculture businesses worldwide [1]. Saprolegnia spp. are generally termed "watermolds" and share common features with both fungi and algae [3]. All fish and ova in fresh water can possibly be infected by Saprolegnia spp., and the disease is termed saprolegniasis. Infected fish are easily recognized by the cotton-like white to greyish patches on the skin and gills [4]. During the last few decades there has been an increased focus on saprolegniasis in salmonid fish and a number of Saprolegnia outbreaks, and attempts to characterize Saprolegnia isolates have been reported [5,6]. Differences in pathogenicity have been proved between strains of a Saprolegnia species even within the same taxonomic grouping [7][8][9][10].
Traditionally, taxonomic characterization of Saprolegnia spp. has been based upon morphological and physiological characteristics [8,11], but in recent years, molecular methods J. Fungi 2021, 7, 713 2 of 13 have become useful tools when describing phylogeny, taxonomy and epidemiology [12][13][14][15][16]. Epidemiological studies of Saprolegnia spp. are particularly useful for identifying sources of infection, characterizing disease spread and improving disease management. However, molecular studies of Saprolegnia isolates are still limited [6], and an improved knowledge is important in order to reduce Saprolegnia outbreaks in the future.
Our aim was to investigate the genetic diversity and relationships of Saprolegnia spp. isolates collected from Canada, Chile, Japan, Norway and Scotland. Our hypotheses were that genetic diversity within and among these countries expresses broader genetic structures or, alternatively, that gene flow is high across all these countries, due to the trade and exchange of breeding materials. Furthermore, we tested if molecular markers (amplified fragment length polymorphism (AFLP) and sequence data of the internal transcribed spacer (ITS)) could distinguish the Saprolegnia isolates according to their species, their ability to produce oogonia with and without long spines on the cysts, and their ability to or not to cause mortality in salmonids. To achieve these goals, we investigated Saprolegnia spp. isolates (Table 1) using AFLP and ITS markers. We believe our study may provide novel insights into the genetics and biology of Saprolegnia spp., as well as indirect knowledge for disease management.

Saprolegnia Material
Thirty-seven Saprolegnia spp. strains were collected from salmonids in Canada (5), Chile (6), Norway (14), Scotland (9) and Japan (3) were investigated in this study (Table 1). The sampling procedure and all morphological, physiological and pathogenic characteristics were described by Stueland et al. [10]. Most strains were isolated from cultured Atlantic salmon (Salmo salar L.), ova, fry or brood stock suffering from saprolegniasis. However, the three Saprolegnia species from Japan were isolated from cultured sockeye salmon (Oncorhynchus nerka) and Coho salmon (Oncorhynchus kisutch) [7,9,[17][18][19]. One of the Norwegian Saprolegnia strains was isolated from brown trout (Salmo trutta) and one strain from common whitefish (Coregonus lavaretus) (Table 1). Twenty-five of the strains have previously been analyzed with respect to their morphological and physiological characteristics [10], and 9 of these strains have also been previously characterized with regard to their pathogenicity to Atlantic salmon [10].

Data Analysis of the AFLP
Data were recorded manually, and only clear polymorphic bands were scored for presence (1) or absence (0). The genetic similarity (GS) was estimated using the Dice coefficient, calculated as GS xy = 2a/(2a + b + c), where a is the number of bands present in both isolates, b is the number of bands present only in isolate x and c is the number of bands present only in isolate y [21]. The genetic similarity among the clones, based on the presence or absence of amplified fragments, was also calculated by Jaccard coefficients [22]. Both analyses resulted in the same clusters, and only the results obtained by the Dice coefficient are presented.
We measured the percentage of polymorphic bands in four countries, calculated by dividing the number of polymorphic bands at the country level by the total number of bands scored. The estimates were based on the isolates from four of the countries. Isolates from Japan were excluded from the analyses, since this country was represented by fewer than five isolates.
The matrix of similarity data was analyzed using the unweighted pair group method with arithmetic mean (UPGMA), as suggested by Sneath and Sokal (1973) [23]. UPGMA clustering was also carried out for all the isolates according to their country of origin, their ability to produce oogonia with and without spines, and their pathogenicity. We performed principal coordinate (PCO) analysis to classify and detect the structure of the relationships between the isolates with different countries of origin, differing ability to produce oogonia with and without spines, and differing pathogenicity. Statistical analyses and construction of dendrograms were performed using NTSYS-pc software version 2.1 [24].
Analyses of molecular variance (AMOVA) [25] was carried out using Arlequin software, version 2.000 [26]. The analysis was estimated according to the country of origin. Isolates from Japan were eliminated from the study because they were represented by fewer than five individuals. The genetic distances among Saprolegnia spp. isolates from Canada, Chile, Norway and Scotland were calculated using the pairwise genetic distance method [27]. The mean F ST was estimated in order to study genetic differentiation between countries. The significance of the F ST values was tested by 1000 permutations. These analyses were performed using Arlequin software, version 2.000 [26]. Gene flow was estimated by assuming Nm = (1/ FST − 1)/4 [28].

Sequencing of the ITS
The internal transcribed spacer 2 (ITS2) part of the nuclear ribosomal DNA was amplified with the primer pair ITS2/ITS4 according to White, Bruns, Lee and Taylor (1990) on a PTC-225 (Peltier Thermal Cycler, MJ Research, Waltham, MA, USA). PCR amplicons were purified with ExoSap IT (GE healthcare, Buckinghamshire, UK) according to the manufacturer's procedure and visualized on standard agarose gel to ensure the presence of single-band products. Both strands of the PCR amplicons were sequenced with the PCR primers using DYEnamic ET dye terminator chemistry (Amersham Biosciences, Chicago, IL, USA), purified on AutoSeq96 (Amersham Biosciences) plates, diluted with 10 µL of MQwater and subsequently analyzed on a MegaBace 1000 (Amersham Biosciences). Sequences were analyzed in Vector NTI Advanced 11 (Invitrogen, Waltham, MA, USA) and assembled in BioEdit 7.0.9.0 [29].

Cluster Analyses
The genetic structure of the Saprolegnia isolates were also investigated using the model-based Bayesian clustering approach of genetic admixture analysis (Structure 2.3.4 software) [33]. Simulations were performed with a dataset from K = 1 to K = 7. The method developed by Evanno et al. (2005) [34] was used to identify the number of genetically homogeneous clusters (K). We used a burn-in period of 100,000 runs and 500,000 MCMC runs to compute the probability of the data for estimating K. Among the seven independent runs, the one with the highest Ln Pr (X|K) value (log probability) was chosen and represented as bar plots. Bar plots of likelihoods and ∆K values were made with STRUCTURE HARVESTER [35].

Results
AFLP analyses of all 37 isolates using six primer combinations resulted in a total of 163 clear polymorphic bands ( Table 2). The number of polymorphic bands per primer combination ranged from 23 to 31 bands, with an average of 27.17 polymorphic bands per primer pair. With these six primer combinations, it was possible to uniquely distinguish among all 37 isolates. Genetic similarity for Saprolegnia isolates using Dice coefficients based on the AFLP data ranged from 0.0 to 1.0, with a mean of 0.41. The highest genetic similarity (1.0) was obtained between two isolates, one from Norway and the other from Scotland, while the lowest genetic similarity value (0.0) was observed between two isolates from Norway. Table 2. Nucleotide sequences of the selective primers used for AFLP analyses and the number of polymorphic bands resulting from each primer combination.

Primer Combination
EcoRI Primer 5 → 3 The percentage of polymorphic bands (PPB) in each group (i.e., country) was high, ranging from 29.45% to 84.05%, with a mean value of 55.52%. The Norwegian isolates showed the highest PPB and the Canadian showed the lowest PPB (Table 3). UPGMA clustering of the Saprolegnia isolates, based on their countries of origin, did not consistently reflect the geographic origin of the Saprolegnia isolates. Some Saprolegnia isolates from Norway clustered together with isolates from Scotland ( Figure 1). However, based on UPGMA, the Saprolegnia isolates were completely clustered according to their ability to produce oogonia with and without long spines on the cysts (Figure 1). We detected 15 specific markers that differentiated isolates that had the ability to produce oogonia with long spines from those without spines. UPGMA analysis of the Saprolegnia isolates according to their pathogenicity resulted in three clusters. All pathogenic isolates clustered together in one group, whereas the non-pathogenic isolates clustered together in two groups (Figure 1). ability to produce oogonia with and without long spines on the cysts (Figure 1). We detected 15 specific markers that differentiated isolates that had the ability to produce oogonia with long spines from those without spines. UPGMA analysis of the Saprolegnia isolates according to their pathogenicity resulted in three clusters. All pathogenic isolates clustered together in one group, whereas the non-pathogenic isolates clustered together in two groups (Figure 1).

Figure 1.
Dendrogram of isolates of Saprolegnia spp. as revealed by UPGMA cluster analysis of AFLP-based genetic similarity (Dice coefficient). The isolates are described by their geographic origin, as shown in Table 1.
The results of the PCO analyses supported the results of the UPGMA analysis. There was no distinguishable clustering pattern of Saprolegnia isolates from a certain country, but the isolates were completely grouped according to their ability to produce oogonia  Table 1. The results of the PCO analyses supported the results of the UPGMA analysis. There was no distinguishable clustering pattern of Saprolegnia isolates from a certain country, but the isolates were completely grouped according to their ability to produce oogonia with spines or without spines, and according to their ability to cause mortality or not in salmonids ( Figure 2). with spines or without spines, and according to their ability to cause mortality or not in salmonids ( Figure 2). In general, the genetic distances between the regions measured by pairwise differences were low (Table 4), the highest genetic distance (0.25198) was between Canada and Chile, and the lowest was between Norway and Scotland (0.04209). The AMOVA analyses based on the geographic origin of the isolates showed that most of the total genetic variability, i.e., 91.01%, was attributed to the variance within countries, while the among-country variance component only accounted for 8.99% (Table  5). A low proportion of the observed genetic differentiation can be explained by the level of the FST value (0.089), and the average estimated mean of gene flow (Nm) between the countries was relatively high (Nm = 2.559).  In general, the genetic distances between the regions measured by pairwise differences were low (Table 4), the highest genetic distance (0.25198) was between Canada and Chile, and the lowest was between Norway and Scotland (0.04209). The AMOVA analyses based on the geographic origin of the isolates showed that most of the total genetic variability, i.e., 91.01%, was attributed to the variance within countries, while the among-country variance component only accounted for 8.99% (Table 5). A low proportion of the observed genetic differentiation can be explained by the level of the F ST value (0.089), and the average estimated mean of gene flow (Nm) between the countries was relatively high (Nm = 2.559). As regards sequencing of the ITS, maximum parsimony analyses with a heuristic search gave the 30 most parsimonious trees (MPTs) with a length of 117 steps. Figure 3 shows a bootstrap tree where all bootstrap values above 50 are indicated on the representative branches. The phylogenetic analyses conducted on 48 partial ITS sequences of Saprolegnia resulted in five well-supported clades. The majority of strains, namely 26, formed a clade with S. salmonis and S. parasitica sequences retrieved from EMBL/GenBank. All isolates within this complex either produced cysts with long hairs (Table 1) or were not tested. A well supported sister clade of this S. parasitica complex consisted of two strains, and one of these was identified as S. hypogyna. The third clade consisted of strains identified as S. ferax, S.mixta, S. longicaulis and S. sp. nuchiae. Some of these different species have identical ITS haplotypes. The last two clades were well-supported and formed sister clades. Five isolates grouped with two strains previously identified as S. diclina and S. australis. The last two isolates grouped with one strain previously identified as S. cf. ferax. Strain S5 was received as a Saprolegnia sp. strains, but both the phylogenetic analysis and the EMBL/GenBank Blast search revealed that this strain clearly belongs to Leptolegnia. Based on the phylogenetic analyses, no geographic grouping could be observed. As regards sequencing of the ITS, maximum parsimony analyses with a heuristic search gave the 30 most parsimonious trees (MPTs) with a length of 117 steps. Figure 3 shows a bootstrap tree where all bootstrap values above 50 are indicated on the representative branches. The phylogenetic analyses conducted on 48 partial ITS sequences of Saprolegnia resulted in five well-supported clades. The majority of strains, namely 26, formed a clade with S. salmonis and S. parasitica sequences retrieved from EMBL/GenBank. All isolates within this complex either produced cysts with long hairs (Table 1) or were not tested. A well supported sister clade of this S. parasitica complex consisted of two strains, and one of these was identified as S. hypogyna. The third clade consisted of strains identified as S. ferax, S.mixta, S. longicaulis and S. sp. nuchiae. Some of these different species have identical ITS haplotypes. The last two clades were well-supported and formed sister clades. Five isolates grouped with two strains previously identified as S. diclina and S. australis. The last two isolates grouped with one strain previously identified as S. cf. ferax. Strain S5 was received as a Saprolegnia sp. strains, but both the phylogenetic analysis and the EMBL/GenBank Blast search revealed that this strain clearly belongs to Leptolegnia. Based on the phylogenetic analyses, no geographic grouping could be observed. Figure 3. Phylogram of isolates of Saprolegnia spp. as revealed by sequencing of ITS region 2. The isolates are described by their geographic origin, their ability to produce long spines on cysts and their pathogenicity to salmon (Table 1).  (Table 1).
Bayesian clustering analysis with STRUCTURE software assigned the 37 Saprolegnia strains to three different clusters. Structure analysis showed the maximum likelihood distribution L(K) of the real number of three groups (K = 3). This value was obtained using the value of ad hoc quantity (∆K) rather than maximum likelihood value L(K), as described by Evanno et al. (2005) [32]. Structure analysis clustered the Saprolegnia isolates into three main clusters (Figure 4) as the UPGMA and PCO groupings. Bayesian clustering analysis with STRUCTURE software assigned the 37 Saprolegnia strains to three different clusters. Structure analysis showed the maximum likelihood distribution L(K) of the real number of three groups (K = 3). This value was obtained using the value of ad hoc quantity (ΔK) rather than maximum likelihood value L(K), as described by Evanno et al. (2005) [32]. Structure analysis clustered the Saprolegnia isolates into three main clusters (Figure 4) as the UPGMA and PCO groupings.

Discussion
The present study provides, for the first time, a good indication that the different pathogenic isolates of Saprolegnia strains clustered at a molecular level. Thereby, we provide documentation that there is a connection between pathogenicity to Atlantic salmon and the molecular diversity of Saprolegnia strains. The pathogenic Saprolegnia strains included in this study that formed one genetic cluster also shared the same morphological characteristics, i.e., long hooked hairs on the secondary zoospore cysts. Several reports have tried to prove that the variability among Saprolegnia isolates is correlated to infectivity [7,8]. We have previously shown significant differences in pathogenicity among seven of the strains included in the present study, and concluded that initial growth rate of germinating cysts in pure water, together with the presence of long hooked hairs on the secondary cysts, may be an indicator of the pathogenicity of Saprolegnia strains to Atlantic salmon [10]. In this study, we have extended this also to include characterization at the genetic level, reassuring our initial findings. However, the number of strains tested in vivo in the present study was limited for animal welfare reasons as well as the high cost of performing in vivo pathogenicity testing. Thus, these findings should be followed by further investigations including in vivo studies.

Geographical Origin
The present study proved there is more genetic variation among Saprolegnia strains within each country than among the countries included in the study. UPGMA clustering, phylogenetic analyses, PCO and STRUCTURE analyses consistently reflected the geographic origin of the Saprolegnia isolates. Actually, the AMOVA showed that most of the total genetic variability was attributed to the variance within countries. These results suggest that Saprolegnia from the countries included in this study share much of the same genetic material, which was also supported by the average estimated mean of gene flow (Nm) among the countries, which was relatively high (Nm = 2.559). This is in contrast to Bangyeekhun et al. (2003) [13], who reported the genetic dissimilarity of pathogenic Saprolegnia from different geographic locations (Northern Europe, Southern Europe and USA) in their study. Factors to consider are the industrialized nature of aquaculture and the transfer of fish among the countries included in the present study. This is an important risk factor for pathogen transfer in general. All the Saprolegnia strains included in the AFLP analysis were sampled from Atlantic salmon-farming countries, with fish/eggs being originally exported from Norway. In this context, it is not surprising that there is more genetic variation among Saprolegnia strains within each country than among the countries included in the study. The high gene flow detected in the study may be due to the repeated Saprolegnia introduction events among these countries and a low level of sexual recombination over time. These results are in agreement with the result detected by Paul et al. (2018) [36].

Morphology
The long hairs on the zoospore cysts seem to be a characteristic dividing the Saprolegnia strains into different genetic clusters. Wiilloughby (1985) [37] stated that this characteristic of Saprolegnia strains isolated from fish is typical for S. parasitica. Beakes et al. (1994) [38] suggested that Saprolegnia strains isolated from infected fish that had the distinctive clusters of long spines on their secondary cysts fulfilled an updated species concept of Saprolegnia parasitica in function, if not in taxonomic context. However, the taxonomic status of Saprolegnia species isolated from fish has still not been resolved. Dieguez-Uribeondo et al. (2007) [6] stated that the species-level identification of parasitic isolates of Saprolegnia has, at best, proven problematic and, at worst, impossible. In the present study, all Saprolegnia strains with long hooked hairs on the secondary zoospore cysts grouped together into a single major cluster, suggesting that they form a separate taxon.

Phylogeny
The phylogenetic analyses of 48 ITS Saprolegnia sequences resulted in five well-defined clades with Leptolegnia as an outgroup. These well-defined clades are largely congruent with previous studies [6]. All strains of S. parasitica formed a well-defined clade consistent with Clade I in a previous study [6]. In the present study, the term S. parasitica has been used for all strains within this clade. All strains tested herein within the S. parasitica complex produced cysts with long hairs. The S. hypogyna strains formed a very well-supported sister clade (97%) and corresponded to Clade Ia in [6]. The AFLP results were, for the main part, congruent with these groupings, with the exception of S. hypogyna, which grouped within the S. parasitica complex. The third clade consisted of at least four Saprolegnia species. This highly supported clade (100%) corresponds to Clade II proposed by Dièguez-Uribeondo et al. (2007) [6] but contains at least an additional three species. The grouping of the last two clades was very well supported (bootstrap: 94%). One of these clades consists of a combination of S. diclina and S. australis strains. The other clade consists of three strains, one identified as S. cf. ferax. These isolates are well separated from S. ferax and are therefore most likely a different species. This grouping into two clades is not congruent with earlier studies [6], which separated these species into three clades, designated III, IV and V. They included a broader range of species and a higher number of isolates for each species than the present study, which may explain the few observed differences between the two studies.
Most of the species included in the present study have ITS haplotypes in EMBL/GenBank that are identical to those of other species of Saprolegnia. This illustrates a common problem with ITS and fungal and fungal-like organisms: the lack of resolution to distinguish closely related species. In other well-studied fungi of fungal-like genera, there has been a shift from ITS to other genetic markers (e.g., β-tubulin, translation elongation factor 1-α). To reveal the true phylogeny of the genus Saprolegnia, other genetic markers should be used.