**3. Results**

A period of 2.3 years was covered by six sampling campaigns targeting eight drinking water sites, i.e., DWDS of eight local hospitals, representing di fferent geographic regions of the West Bank (Supplementary Figure S1). The six sampling campaigns targeted the main seasons in Palestine, i.e., spring (March to May), summer (June to August), and autumn (October to December). These seasons were sampled twice from 2012 to 2014. Data on sampling per site and sampling campaign are given in detail in Zayed [21].

The occurrence of *Legionella* species, with emphasis on *L. pneumophila,* in the drinking water of the West Bank was determined by cultivation and molecular approaches. To gain insight into potential environmental drivers, a set of physico-chemical and microbiological parameters were determined. Sampling included bulk water of the DWDS and biofilm analyzed by culture- and PCR-based methods using environmental DNA extracted directly from the sample material.

#### *3.1. Physico-Chemical and Microbiological Characteristics of the Drinking Water*

The drinking water supplied to the hospitals in the West Bank was mostly derived from groundwater either by wells or springs and stored in a reservoir of each hospital. The supplied drinking water was characterized by high hardness (210–350 mg/<sup>L</sup> CaCO3-equivalents), high bicarbonate level (170–250 mg/L), high conductivity (400–1000 μS/cm), and a high content of total dissolved solids (TDS) (260–470 mg/L). It contained variable amounts of chloride (20–110 mg/L) and sulfate (10–40 mg/L). On-site analyses of the drinking water retrieved from the hospitals' DWDS showed that a high Mg concentration (21–40 mg/L) in addition to the high Ca concentration (75–100 mg/L) contributed to the high hardness. All water components showed a high regional variability) [21].

In hospital drinking water, pH ranged from 7.6 to 8.4, with an average pH per hospital between 7.6 and 8.0. Temperature of the cold water ranged from 18 to 26 ◦C and only rarely reached values higher than 27 ◦C (<9% of total sampling). On average (mean value for all samplings per hospital) cold water temperature per hospital ranged between 21 and 25 ◦C. Hot water temperature ranged from 30 to 70 ◦C and had a mean temperature per hospital between 39 ◦C and 52 ◦C. Chlorine ranged from 0.1 to 1 mg/<sup>L</sup> and was only rarely (<7% of total sampling) elevated above these values. On average per hospital chlorine ranged between 0.2 and 0.7 mg/L. Heterotrophic plate counts (HPC) were assessed at 22 and 37 ◦C. HPC at 37 ◦C varied from 3 × 10<sup>2</sup> to 4 × 10<sup>5</sup> CFU/L, with averages per hospital ranging from 1.5 × 10<sup>4</sup> to 1.6 × 105. At 22 ◦C, HPC were about one order of magnitude lower, with an average per hospital of 1.4 × 10<sup>3</sup> to 6.3 × 104.

#### *3.2. Abundance of L. pneumophila in Water and Biofilm as Assessed by Cultivation and PCR-Based Methods*

An overview on the *Legionella* abundances in the bulk water and biofilm is given in Table 1. For water samples, *Legionella* plate counts were mostly below detection level, with only three sampling sites out of eight where *Legionella* were detected in a range of 8 to 148 CFU/L (mean value per site), and only one site with more frequent observation of *Legionella* in summer and autumn (hospital F). In biofilm, culturable *Legionella* were detected at all sampling sites. On average, *Legionella* positive swabs per sampling site were 15% ranging from 3 to 30% (mean value per site) (Table 1).

PCR-based detection of *Legionella* spp. and *L. pneumophila* showed a higher fraction of *Legionella*-positive samples. In water samples, at seven out of eight sampling sites *L. pneumophila* was detected with an average detection rate per sampling site of 46% and a range from 0 to 100%. In biofilm samples, *Legionella* spp. was regularly detected at all sites at an average detection rate of 68% ranging from 40 to 93% per site (Table 1.).

For water samples, *L. pneumophila*-specific PCR was more sensitive than plate counts. The observations by culture and PCR were consistent in a way that whenever plate counts were observed, PCR gave positive results, whereas a large set of PCR-positive samples did not show any plate counts (Supplementary Figure S2A).

Culture based detection of *L. pneumophila* in biofilms was much more successful than in water samples. *Legionella* plate counts from water samples were only positive when about half or more of the biofilm swabs were positive for *L. pneumophila* cultivation (Supplementary Figure S2B).

In general, the detection of non-*L. pneumophila* colonies on the agar media used for *Legionella* spp. counts was very rare as assessed by species-specific PCR. The number of *Legionella* spp. counts can therefore be approximately addressed as *L. pneumophila* counts.

#### *3.3. Comparison of Abundance of Culturable Legionella in Hot and Cold Water*

Hot water was available at five of the eight sampling sites except for spring 2013. Hot water temperature ranged from 30 to 70 ◦C and had a mean value around 45 ◦C [21]. At the five sites, sampling was achieved in parallel for hot and cold water. There was no significant difference observed for the *Legionella* plate counts (Supplementary Figure S3). These counts were usually below detection limit in hot water as in cold water. Only for site F (Bethlehem), there was an increased level of *Legionella* counts in summer 2013 and 2014: in cold water 467 CFU/L in 2013 and 421 CFU/L in 2014; in hot water plate counts were similar (508 CFU/L) in 2013, while in 2014 no *Legionella* were detected. Thus, a low level of culturable *Legionella spp*. was observed also for the hot water in this sampling comparison.

#### *3.4. Seasonal Dynamics of L. pneumophila in Biofilm and Water*

*L. pneumophila* showed an increase in biofilm samples from spring to autumn across all sampling sites. Both culture- and PCR-based methods showed this tendency (Figure 1). For PCR-based detection, the percentage of positive swabs was in general higher than for culture-based detection. Culture-based methods and *L. pneumophila*-specific PCR-based methods showed a good correlation (r2 = 0.83, Figure 1A; correlation with *Legionella* genus-specific PCR: r2 = 0.78). The relative increase was higher for culture from spring to autumn (increase from about 10 to 20% of positive swabs, Figure 1C) compared to the PCR-based detection (increase from about 60 to 80% of positive swabs, Figure 1B). The comparison of culture-based and PCR-based detection of *L. pneumophila* indicated an increase of culturability of *L. pneumophila* from biofilm from spring to autumn (and the respective exposure to higher temperature) from about 15% in spring to about 27% in autumn as shown by the respective ratios (Figure 1D).

In contrast, *L. pneumophila* abundance in bulk water showed a maximum in summer as detected by cultivation and, as a tendency, by *L. pneumophila*-specific PCR (Figure 2B,C). Detection by PCR was more sensitive than plate counts; the correlation between PCR detection and plate counts (CFU/L) in a seasonal comparison was r2 = 0.69 (Figure 2A).

On a seasonal basis, there was no correlation between biofilm and water samples neither by culture nor by PCR-based detection; this is consistent with the observation of different seasonal maxima for biofilm and water. In summary, the seasonal dynamics of the *Legionella* abundance was not strongly pronounced, especially not as detected by PCR.

**Figure 1.** Seasonal variation of abundances of *L. pneumophila* in biofilm of eight sampling sites (mean values) of the West Bank sampled from 2012 to 2014. (**A**) Swabs positive by culture vs. swabs positive by *L. pneumophila*-specific (L1-primer) PCR (*p* < 0.05), (**B**) swabs positive by *L. pneumophila*-specific (L1-primer) PCR vs. seasons (NS: Not Significant), and (**C**) swabs positive by culture vs. seasons (*p* < 0.05). (**D**) Ratio of swabs positive by culture vs. swabs positive by *L. pneumophila*-specific PCR vs. seasons (*p* < 0.05) (*n* = 45, mean values *n* = 6).

**Figure 2.** Seasonal variation of abundances of *L. pneumophila* in water of eight sampling sites (mean values) of the West Bank sampled from 2012 to 2014. (**A**) *Legionella* plate counts vs. water samples positive by *L. pneumophila*-specific (L1-primer) PCR (*p* < 0.05). (**B**) Water samples positive *by L. pneumophila*-specific (L1-primer) PCR vs. seasons (NS, not significant). (**C**) *Legionella* plate counts of water samples vs. seasons (*n* = 45, mean values *n* = 6) (*p* < 0.05).

#### *3.5. Influence of Physico-Chemical and Bacteriological Parameters on Legionella Abundance in Water and Biofilm*

For the assessment of the relationship between *Legionella* abundance in water and biofilm with bacteriological and physico-chemical parameters in water, these parameters were pairwise compared and displayed in a correlation matrix (Supplementary Table S1). Eighteen quantitatively determined parameters were used to define the physico-chemical background of the sampling sites. Heterotrophic plate counts incubated at 25 ◦C and 37 ◦C were used as general bacteriological parameters. Culturable *Legionella* counts were used for water and biofilm. PCR-based detection of *L. pneumophila* and the *Legionella* genus were added for biofilm swabs. For the clonal level of *L. pneumophila*, the incidence

of the seven most abundant MLVA-genotypes (Gt) and four clonal complexes (VACC) were included in the correlation matrix.

Correlation analyses of the abundance of culturable *L. pneumophila* ("*Legionella* count") in water and biofilm vs. the physico-chemical parameters showed a correlation with the magnesium concentration and the Ca/Mg ratio in the sampled tap water, but not with any other of the analyzed physico-chemical parameters (Supplementary Table S1). Surprisingly, there was no significant correlation with chlorine concentration or temperature. Abundance of culturable *Legionella* in water and biofilm were correlated; furthermore, both parameters were correlated with occurrence of specific *L. pneumophila* genotypes and clonal complexes (VACC). The magnesium concentration in water showed a relationship not only with the abundance of *Legionella* in water and biofilm, but also with some genotypes and clonal complexes. The findings of the correlation analysis are elaborated in more detail in the following paragraphs.

As shown in Figure 3, there was a tight correlation between the Ca/Mg ratio and the Mg concentration for the whole data set (Figure 3A) as well as for the mean of the eight sampling sites (Figure 3B). There was no correlation between Ca and Mg; Ca varied between 103 and 78 mg/<sup>L</sup> with no correlation with the Mg concentration.

**Figure 3.** (**A**) Correlation between the Ca/Mg ratio and the Mg concentration for the total data set (*n* = 45) and (**B**) for the mean values of the eight sampling sites (*n* = 45, mean values *n* = 8) *p* < 0.001.

For the water samples (Figure 4), the plate counts of *L. pneumophila* showed a negative correlation with increasing Mg concentrations. This was observed for the overall set of samples (regression not shown) as well as for the mean for each sampling site (Figure 4A). For both data sets, the correlation could be best described by a power function with a correlation coefficient of r2 = 0.54 and 0.78 (0.68 for the linear correlation, Figure 4A) respectively. Figure 4B shows the correlation with the Ca/Mg ratio that is best described by an exponential function (r2 = 0.63). Abundance of culturable *L. pneumophila* increased with increasing Ca/Mg ratio as expected from the above described (Figure 3) relationship of Mg with the Ca/Mg ratio. *L. pneumophila*-specific PCR of water samples showed the same tendency, i.e., a negative correlation with increasing Mg concentrations, but was not significant (Figure 4C).

**Figure 4.** Correlation of *L. pneumophila* abundance with Mg and the Ca/Mg ratio of the water samples as detected by plate counts (**A**,**B**) (*p* < 0.05) and *L. pneumophila* specific PCR (**C**) (*n* = 45, mean values *n* = 8) (NS: not significant).

For the biofilm samples, a correlation between the magnesium concentration and the PCR-based detection of *Legionella* spp. and *L. pneumophila* was observed as shown for the mean of each sampling site in Figure 5. By contrast, culture-based analyses of swabs did not show a clear trend. As for the water samples, increasing Mg concentrations yielded a lower percentage of positive biofilm samples (swabs), both for *Legionella* genus-specific PCR (Figure 5A) and *L. pneumophila*-specific PCR (Figure 5B).

**Figure 5.** (**A**) PCR-based detection of the genus *Legionella* (*p* < 0.05) and (**B**) *L. pneumophila* in biofilm swabs vs. magnesium concentration in bulk water (*p* < 0.05) (*n* = 45, mean values *n* = 8).

For culture-based and PCR-based detection of *L. pneumophila* in water and biofilm, correlations with the Ca/Mg ratio yielded similar correlation coefficients as with Mg, but with an increase of *Legionella* abundance with an increasing Ca/Mg ratio as expected from the correlation between Mg and the Ca/Mg ratio (see Figure 4B and Supplementary Table S1).

#### *3.6. Prevalence and Biogeography of L. pneumophila Genotypes Across the West Bank*

During the six sampling campaigns from 2012 to 2014, 180 strains were obtained and successfully genotyped by MLVA-8(12) using 13 loci resulting in 27 different genotypes (Table 2). Twelve genotypes were represented by three strains up to a maximum of 74 strains per genotype. The remaining 15 genotypes were represented by two or one strains. Details on the results of the MLVA analysis and the clonal structure of the *L. pneumophila* population of the West Bank are given by Zayed et al. [20].

The 27 genotypes were affiliated with four VNTR clonal complexes (VACC 1, 2, 5, and 11) indicating the genetic relatedness among the respective genotypes. Seventeen of the MLVA-genotypes are affiliated with nine different sequence types (ST), meaning that some of the MLVA-genotypes pertained to the same ST. In the following analyses, MLVA genotyping will be used as a classification scheme and as a basis to study the ecology of *L. pneumophila* on a clonal level.

Though sampling was achieved in the same way for all hospital DWDS, the yield of isolates per site was rather variable and in accordance with the isolation success from water and biofilm (Table S2). Most isolates were retrieved from biofilm swabs (175 strains) and only a minor fraction from water samples (five strains in total, three from Bethlehem (sampling site F), one from Jenin (sampling site A) and Hebron (sampling site G)). The genotypes isolated from water were among the most abundant genotypes in general, and of high relevance for the respective sampling site (Table 2). For a comparison, 15 isolates retrieved during the same sampling period from biofilm samples of the Al Quds-University were added and indicated for the respective genotypes (Table 2), sampling sites (Supplementary Table S2), and the biogeographic distribution (Figure 6).

The isolates comprised serogroups (Sg) 1, 6, 8, 10, and 2–14 (Table 2). Sg 1 comprised most of the isolates (62%) and a total of seven MLVA-genotypes. Sg 6 was the second most important serogroup (30%), comprising the largest diversity with 11 MLVA-genotypes [20].


**Table 2.** MLVA-Genotype composition of isolated *L. pneumophila* strains and their a ffiliation with VNTR clonal complexes (VACC), sequence types (ST), and serogroup (Sg).

NA: not available; \*, ST was assessed for strains of the same MLVA-8(12) genotype, and not directly for the West Bank strains; e, ST was estimated from the MLVA-8 pattern; W, genotype retrieved from water, in brackets the site of isolation is indicated; AQ, contains strains retrieved from biofilm of the AQU.

The biogeographic distribution of the strains according to their MLVA-genotype and clonal VNTR complex (VACC) is indicated in Figure 6. On the genotype (Gt) level, Figure 6A shows a genotype pattern that varied on the regional level. In the north of the West Bank, Gt 4(17) was highly prevalent. In the south, the pattern showed high divergence from site to site. For example, site G is dominated by Gt 6(18) that was not retrieved from any other site; similarly, site F showed a high prevalence of Gt 10(141) that was endemic for this site. Furthermore, on the level of the clonal complexes the regional variability was well pronounced (Figure 6B). In general, there was a high prevalence of VACC1 in the West Bank, except for sites E, F, and H in the south. Site F showed a high prevalence of VACC11, site E for VACC 2, and site H for VACC 5.

The richness, i.e., the number of *L. pneumophila* MLVA-genotypes, varied from 2 to 7 per sampling site, with a mean of 4.5 genotypes for the eight sampling sites. The ratio of number of genotypes vs. number of strains retrieved per sampling site was added as an indicator of the "genotype diversity" ranging from 0.11 to 0.67 and a mean of 0.30 (Supplementary Table S2). There was no significant correlation (r2 = 0.28) between the number of strains and the number of genotypes retrieved per sampling site. In contrast, correlation analysis revealed a negative correlation (r2 = 0.53) between the average percentage of biofilm swabs positive for *Legionella* culture per sampling site and the "genotype diversity", i.e., sites with low *Legionella* incidence on biofilm swabs showed a high diversity compared to a low diversity in case of high *Legionella* incidence (Supplementary Figure S4).

**Figure 6.** Biogeographic distribution of the *L. pneumophila* strains according to their MLVA-8(12) genotype (**A**) and their clonal complex (VACC) (**B**). In Figure 6A, the respective MLVA-8(12) genotype is indicated in the following way; genotype MLVA-8 plus genotype MLVA-12 in brackets, e.g., the genotype MLVA-8(12) 4(17) is indicated as "4(17)".

#### *3.7. Environmental Factors Correlating with Genotype Abundance and Composition*

The variable pattern of genotype prevalence and composition across the West Bank raised the question concerning the influencing environmental factors. There was a broad set of physico-chemical parameters recorded or obtained from the water authorities of the West Bank [21]. The eight quantitively measured parameters that showed differences between the sampling sites were used for detailed statistical analysis, i.e., PCA and cluster analysis. The selection criterion for the genotypes to be included in these analyses was the number of strains available per genotype, i.e., only genotypes represented by at least three isolates were included. This selection resulted in ten genotypes subjected to PCA and cluster analyses.

Three groups of MLVA-genotypes could be separated by PCA and cluster analysis based on the eight parameters, i.e., concentrations of chloride, sulfate, Ca, Mg, total dissolved solids (TDS), turbidity, Ca/Mg-ratio ("Ratio"), and *Legionella* plate counts (Figure 7). Cluster analyses revealed three groups for the ten genotypes (Figure 7A). By PCA, eight of the ten genotypes were assigned to three groups, with two genotypes (Gt 9(92) and Gt 63(83)) being close the PCA-group B1 comprising Gt 4(17) and Gt 13(72) (Figure 7B). These four genotypes (Gt 9(92), 63(83), 4(17), and 13(72)) were included in one group (cluster group B) by cluster analyses. For further considerations, genotypes Gt 9(92) and Gt 63(83) were combined with Gt 4(17) and Gt 13(72) in group B based on the cluster results, and because they can be considered to live in a similar environment with respect to the chosen parameters.

**Figure 7.** (**A**) Dendrogram showing group average hierarchical clustering of MLVA-8(12) genotypes. (**B)** Principal component analysis (PCA) of MLVA-8(12) genotypes showing the effect of biological and physicochemical parameters. Light green ellipses represent the log normal distributions of principal component values for genotype groups. Genotype groups were suggested as pertaining to the same niche. The resulting niches are A (green), niche B (red), niche C (blue). Niches A and C are consistent with the calculated grouping A and C; the calculated group B1 was enlarged to include two more genotypes in a suggested group B resulting in an enlarged niche B. Niches are generally considered as larger than the calculated genotype groups. Legend: Ratio, Ca:Mg ratio; Leg. count, *Legionella* plate counts in water samples; TDS, total dissolved solids; Turb, turbidity; Ca, Calcium; Mg, Magnesium; SO4, Sulfate.

The environment of these three groups could be characterized by the respective parameters in summary and assigns distinct niches to the three groups. The characterization of the three groups by the individual parameters is given in Table 3. While a comprehensive and stable distinction is shown for the sum of the parameters, many of the individual parameters also allow a distinction between two or three groups.



**Legend:** Independent *t*-test: NS, not significant; \*, *p* ≤ 0.05; \*\*, *p* ≤ 0.01; \*\*\*, *p* ≤ 0.001.

The three groups of genotypes were considered to co-occur in their respective environment as described by the above-mentioned parameters. The respective habitats as described by these environmental parameters for the three groups of genotypes could be considered as their "niche". This means that to each niche three to four genotypes were assigned. More genotypes may be sharing these niches, but they were not included in the analysis due to the low number of strains per genotype.
