Diversity and ice nucleation activity of Pseudomonas syringae in drone-based water samples from eight lakes in Austria

Lake locations and sampling

Freshwater sampling was conducted at eight different lakes in Austria in June 2018, Altauseer See (ALT), Grundlsee (GRU), Toplitzsee (TOP), Vorderer Gosausee (GOS), Gosaulacke (GOL), Hinterer Gosausee (HIN), Ossiacher See (OSS) and Wörthersee (WOR), (Fig. 1) (Benson et al., 2019). Sampling details and design of the Drone Water SamplEr (DOWSE) were previously described in Benson et al. (2019). Briefly, seven of the eight lakes were sampled with a drone 1, 25, and 50 m from the shoreline. Samples were collected from a 30 × 50 m grid of nine collection sites in the lake. Due to the narrow width of GOL, this lake was sampled at 1 and 25 m from shore in four locations to form a 45 × 25 m grid of eight sample collections (Benson et al., 2019). Lake area (km²) and elevation are listed in Table 1.

Lake sampling permissions and safety of drone operations

Permissions to sample the lakes were granted by the Austrian Federal Forests AG, DI Martin Heinz Stürmer, on 3 April 2018. A formal field collection permit was not required for this work. Sites for drone operations were carefully selected to be minimally intrusive to people in the area during the time of sampling. The drones used as part of this work were registered with the Federal Aviation Administration (FAA). The UAS pilot for the missions reported in this manuscript was a certified FAA Remote Pilot under Part 107, Certificate Number 4038906.

Collection and processing of microorganisms

Lake samples from Benson et al. (2019) were processed for storage on the same day of collection. Briefly, 100 mL from each location was filtered through a 0.2 µm single use Pall filter funnel (#28143-542; VWR International, Radnor, PA, USA). Filters were transferred to 15 mL tubes and stored at 4 °C. All samples were shipped on ice to VA, USA for culturing, storage, and downstream analyses. Details concerning the culturing of microorganisms for this study are described in Benson et al. (2019). Briefly, KBC agar plates were used to select for bacteria in the genus Pseudomonas and TSA agar plates were used to grow culturable bacteria (including Pseudomonas). While not all bacteria collected from the environment are culturable in a laboratory setting, we chose an agar plate composition that had been tested with precipitation samples. Previous studies explored various agar plate compositions (including TSA) with regard to the culturability of environmental rain (Failor et al., 2017), and simulated rain samples (Hanlon et al., 2017).

Sequence analysis and identification ( cts and 16S)

Pure strains were subjected to sequence analyses of the citrate synthase (cts) housekeeping gene (for strains on the semi-selective medium, KBC) or 16S rDNA (for strains cultured on TSA). Berge et al. (2014) used multi locus sequence typing (MLST) analysis of 216 P. syringae strains to identify 23 clades and describe 13 phylogroups in the P. syringae genetic complex. These authors showed that the cts housekeeping gene alone was 97% accurate in predicting phylogeny. Prior to sequencing, a 5 min. template boil was performed in 100 µL of water with a 1 mm toothpick stab of frozen stock or from a fresh colony streak plate. A 25 µL PCR reaction was carried out with 2 µL of template from the quick boil tube. The cts gene was amplified using GoTaq® Green Master Mix (Promega M712) with primers ctsFor (5′ CCC GTC GAG CTG CCA ATW CTG A 3′) and ctsRev (5′ ATC TCG CAC GGS GTR TTG AAC ATC 3′). The cts gene sequence data was used to confirm that strains belong to the P. syringae complex and to determine the phylogroups within the complex. Conditions of the cts PCR reaction were previously described in Pietsch, Vinatzer & Schmale (2017). PCR products were confirmed on an 1% TBE agarose gel with SYBRTM Safe DNA Gel Stain (Invitrogen by Thermo Fisher Scientific, Waltham, MA, USA). Reactions were cleaned enzymatically with rSAP and ExoI and Sanger sequencing reactions were performed by Eton Biosciences (AB1 3730 × 1 DNA Sequencer; Eton Biosciences, Durham, NC, USA). Sanger generated sequences were trimmed (using Phred Version 0.020425.c; http://bozeman.mbt.washington.edu/phrap.docs/phred.html) and subsequently queried against the NCBI nr (non-redundant) database. The same PCR parameters were used to amplify a portion of the 16S rDNA gene. The forward and reverse primers used were 518 F (5′ CCA GCA GCC GCG GTA ATA CG 3′) and 1491 R (5′ ATC GGY TAC CTT GTT ACG ACT TC 3′) (Turner et al., 1999). The identification of culturable bacteria was determined to the classification level of genus by using trimmed sequences as described above and sequences were submitted to GenBank.

Phylogenetic analysis of P. syringae complex strains in lakes

To specifically determine which of the strains isolated from KBC medium were within the P. syringae complex and to situate them phylogenetically, we compared, via BLAST, the cts sequences for these strains with the overall NCBI data base which contains the full diversity of this bacterium reported by Berge et al. (2014). Strains for which the cts sequence did not have at least 90% sequence similarity with that of strains in the P. syringae complex were eliminated from subsequent analysis thereby facilitating alignment of the cts sequences along a greater number of bases than if all strains were included in the analysis. Four additional strains were eliminated because we could not sequence the cts gene to obtain a sufficient number of bases (>300). The resulting numbers of P. syringae complex sequences for each lake were 10 for ALT, 30 for GRU, 37 for TOP, 39 for GOS, 14 for GOL, 18 for HIN, 39 for OSS and 1 for WOR. The phylogenetic relationships among strains was determined as previously reported (Berge et al., 2014) based on 311 base pairs of the retained sequences and for 34 reference strains via the neighbor-joining algorithm in MEGA7 with 1,000 bootstrap replicates. The total number of confirmed P. syringae complex strains from each lake is shown in Table 2.

Ice-nucleation activity of confirmed P. syringae complex strains

Confirmation of positive ice+ Pseudomonas samples that were tentative P. syringae (271 strains) was verified in three independent ice-nucleation assays and reported in Table S1. Positive ice+ strains that were part of the P. syringae complex are listed in Table 2. The assays were performed on a boat made with PARAFILM® M (Sigma P6543, 20 in. × 50 ft) floating in a Lauda Alpha RA 12 (LCKD 4908) cooling bath (08075; LAUDA-Brinkmann, LP, Delran, NJ, USA) with ethylene glycol coolant fluid (Air gas RAD64000246). This parafilm boat-based assay has been described previously (Pietsch et al., 2016; Garcia et al., 2019). Pure cultures of bacteria were transferred from culture plates with a sterile toothpick to 140 µL of DI water into a 96-well plate. Plates were vortexed for 30 s and incubated at 4 °C for 1 h. Twelve-microliter droplets of the microbial suspension were loaded in duplicate at −6 °C. The temperature of the cooling bath was decreased to −7 °C, held for 2 min, then decreased to −8 °C. After a final incubation time of 10 min at −8 °C, all drops that froze were considered to be INA+ samples. The freezing temperature was recorded for each drop on the cryofloat.

Analysis of microbial richness and similarity

16S rDNA sequences were analyzed from a total of 492 pure cultures of bacteria isolated from TSA. Richness and Shannon diversity were characterized at all levels of taxa for 492 strains cultured on general enriched TSA agar. Variation in richness and diversity with distance to shorelines and lakes levels were analyzed with mixed effect models with lake treated as a random effect. Random effects were tested using likelihood ratio tests whereas fixed effects were tested using F-tests with Satterthwaite adjusted degrees of freedom.

Population analyses

Variation in the composition of genera of bacteria among and within lakes was analyzed using non-metric multidimensional scaling (NMDS) (Shepard, 1962a, 1962b; Kruskal, 1964a, 1964b) and PERMutational Multivariate ANOVA (PERMANOVA; Anderson, 2001). Dissimilarity matrices for NMDS and PERMANOVA were created using Bray-Curtis dissimilarity (Bray & Curtis, 1957). The resulting 2D NMDS ordination was adequate for drawing inferences (stress = 0.14). All analyses were performed using the R statistical environment (R core team 2019) with heavy reliance on the packages vegan (Oksanen et al., 2019), lme4 (Bates et al., 2015), lmerTest (Kuznetsova, Brockhoff & Christensen, 2017), and asbio (Aho, 2023).

Culturing and identification of strains in the P. syringae complex

A total of 415 strains recovered from KBC were identified as Pseudomonas. Sequences were submitted to GenBank and assigned accession numbers (MW857572–MW857985 and MW892633). Based on specific sequence criteria relative to the cts gene, 271 strains were defined as tentative P. syringae based on significant sequence homology. These strains were then analyzed using a Pseudomonas specific database to determine if they were part of the P. syringae complex defined by Berge et al. (2014). A total of 188 strains were associated with the P. syringae complex and represented 32 haplotypes in eight phylogroups (Fig. 2). Workflow is shown in Fig. S1.

The numbers of tentative P. syringae sequences for each lake were 26 for ALT, 53 for GRU, 53 for TOP, 41 for GOS, 24 for GOL, 21 for HIN, 49 for OSS and 4 for WOR for a total of 271 strains. A phylogenetic analysis of these strains showed that 188 strains grouped with P. syringae as part of the P. syringae complex, and the remaining 83 grouped with reference strain PAO1 for P. fluorescens. The 188 strains in the P. syringae complex represented 32 haplotypes in phylogroups PG01, PG02, PG07, PG09, PG10, PG13, PG14 and PG15 of the P. syringae complex, for the portion of the cts gene that was sequenced. Among the 32 haplotypes, 11 were common to at least two lakes and five lakes had a total of 21 unique haplotypes, i.e., haplotypes represented by only one strain in one lake (Fig. 2).

The structure and size of the P. syringae complex phylogroups varied among lakes, and we found few correlations among the factors measured here. For example, neither the concentration nor the abundance of any of the haplotypes or phylogroups were significantly correlated with lake altitude (Spearman’s rank correlation, p < 0.05). PG01 and PG02 were the only phylogroups whose fractions in the population were significantly correlated with total concentrations of P. syringae complex strains in lakes; they increased as total concentrations increased (Spearman’s rank correlation, p < 0.05).

Ice nucleation activity of P. syringae strains

The numbers of tentative Pseudomonas strains that were tested for ice nucleation activity at −8 °C (and % that froze) in the droplet freezing assay for each lake are included in Table S1. There were 26 for ALT (12%), 50 for GRU (38%), 44 for TOP (18%), 40 for GOS (35%), 23 for GOL (26%), 20 for HIN (65%), 44 for OSS (30%) and 4 for WOR (0%) (Table S1). The frequency of tentative Pseudomonas strains that were ice nucleation active (at −8 °C) varied among lakes (0 to 65%) and was significantly positively correlated with the fraction of phylogroup 02 (PG02) strains in the population but with no other phylogroup component nor with the total abundance of tentative Pseudomonas strains in the lake (Spearman’s Rank correlation, p < 0.01).

Microbial richness and similarity among the eight lakes

Thirty-five unique genera of culturable bacteria were identified from the non-selective TSA plates across the eight lakes (7.34 × 10⁴ cfu/mL mean of eight lakes). These strains were assigned accession numbers MN490845 to MN491336. Over forty percent of the strains (43.5% (214/492)) were associated with the genus Pseudomonas. Population richness (based on genera of bacteria) and Shannon diversity were analyzed for the set of 492 bacteria identified from the eight lakes. The highest richness was detected in TOP and OSS lakes, while the lowest richness was reported in GOS and GOL (Fig. 3). Pseudomonas was the most abundant genus, and was present in all eight lakes at a mean percentage of 43.5% (214/492). The percentage of Pseudomonas ranged from 23% in HIN, to 64% in GOS. ALT, GRU, GOS, GOL and OSS had 50% or more Pseudomonas in the strains analyzed (Fig. 4). The Shannon diversity analysis confirmed TOP and OSS to have the highest diversity. The analysis also showed more diversity in HIN relative to ALT, whereas the richness data did not distinguish a difference between HIN and ALT.

Population analysis of culturable bacteria

A PERMANOVA of samples within lakes indicated that bacterial assemblages varied strongly by lake (R2 = 0.58, F_2,5 = 2.81, p = 0.001, 999 permutations). Bacterial assemblages did not vary with distance to lake shoreline (R2 = 0.04, F_2,5 = 0.73, p = 0.745, 999 permutations).

An NMDS analysis showed relationships of bacterial assemblages within and among lakes (Fig. 4). Particularly distinct were assemblages of TOP (lower right), HIN (center right) and OSS (top). In the 2D ordination for genus, the vector fitting p-values less than or equal to 0.004 (p ≤ 0.004) showed the strongest association to the ordination as marked by arrows in Fig. 5. ALT, GRU, GOS, GOL, and WOR were all distinguished by high levels of Pseudomonas (p ≤ 0.001). WOR also showed a correlation to Exiguobacterium (p ≤ 0.036). TOP was indicated by Janthinobacterium (p ≤ 0.001), while OSS was indicated by Pedobacter (p ≤ 0.001) and Brevundimonas (p ≤ 0.003). HIN was correlated to two genera, Flavobacterium (p ≤ 0.015) and Massilia (p ≤ 0.016).