**4. Discussion**

#### *4.1. Legionella Abundance in Water and Biofilm*

The number of culturable *L. pneumophila* was surprisingly low in the drinking water of the eight sampling sites in hospitals of the West Bank, with only three positive sites out of eight and a low average abundance (mean values for the three positive sites: 8–150 CFU/L). Hot water followed this trend and did not show higher concentrations than cold water. Using *L. pneumophila*-specific PCR, *L. pneumophila* was detected at all sampling sites except for site E. PCR-based detection was thus more sensitive with a broad range of site specific variability from 0 to 100% of water samples being positive on average per site. In biofilm, culturable *L. pneumophila* was detected at all sites, but with high sampling site specific variability. The incidence rate of culturable *L. pneumophila* in biofilm ranged on average per site from 3 to 30%, whereas *L. pneumophila*-specific PCR was positive on average for 40 to 93% of the biofilm samples per site. Due to the rare incidence of non-*L. pneumophila* colonies on agar media, *Legionella* counts on agar media will be addressed in the discussion as the culturable fraction of *L. pneumophila* in water and biofilm.

Culture-based detection was always lower compared to *L. pneumophila*-specific PCR detection in bulk water, but culture-based data were consistent with PCR-based data, i.e., PCR-based detection was always positive when cultivation was successful. This higher sensitivity of the PCR-based detection compared to cultivation was valid for water and biofilm, an observation often reported [29]. Bonetta et al. [30] suggested the viable-but-not-culturable state and the PCR detection of DNA of dead *Legionella* cells as reasons for the higher PCR detection.

Culture-based detection in biofilm was far more successful than in water. Culturable *L. pneumophila* was only detected in water when about 50% of the biofilm samples of the respective site and sampling were positive. This is consistent with observations in other studies showing that a major fraction of the microbial biomass in a DWDS is found in the biofilms attached to the pipe walls presumably due to improved nutrient conditions and shelter from stressing agents [31].

Compared to an annual study in a comparable climatic region, i.e., a water network at an university campus (Oranim) close to Haifa, Israel, much higher *Legionella* counts in drinking water were observed [13]. The level of culturable *Legionella* in water at Oranim campus was in the range of 10 to 5800 CFU/L, with more than 60% positive water samples. In the Oranim study, samples with culturable *L. pneumophila* in water usually showed culture positive biofilm samples, as it was the case in the West Bank study.

With respect to Mediterranean climate, there were several drinking water studies performed in buildings in Italy including hospitals and hot water systems [32,33]. Leoni et al. reported an incidence of 93.7% in hospital water in Bologna with average *Legionella* counts of 2400 CFU/L [32]. Borella et al. report for 60% of the water samples sampled in hotels across Italy a contamination of >1000 CFU/L, with 20% exceeding even 10,000 CFU/L [33]. Based on an extensive study of the drinking water supply system of a city in Northwestern France (Rennes), Sobral et al. (2011) reported an abundance of culturable *Legionella* of 50–200,000 CFU/L with a mean of 800 CFU/L for hot water systems [12].

Compared to the above-mentioned studies, the drinking water in the West Bank seemed to have a relatively low contamination with culturable *Legionella* despite a set of risk factors, such as frequent water supply interruption and storage in containers, raising the question for the drivers behind this unexpected observation.

#### *4.2. Seasonal Dynamics of L. pneumophila*

The seasonal dynamics of *L. pneumophila* abundance was not strongly pronounced and showed a di fferent pattern for water and biofilm. Abundance in water had a maximum in summer, and thus followed the temperature regime over the year. Abundance in biofilm increased from spring to autumn, indicating a maximum of *L. pneumophila* prevalence in biofilm in autumn. These seasonal patterns of water and biofilm were better observed by culture-based analyses than by *L. pneumophila*-specific PCR. The observation that in biofilm the ratio of culturable *L. pneumophila* vs. PCR-based detection approximately doubled from spring to autumn, may have contributed to the better detection of this seasonal trend by culture-based analysis. The weak seasonal dynamics of *Legionella* abundance may have been due to the site-dependent variability of the temperature regime. In addition, there was a lack of correlation between temperature and *Legionella* abundance by culture and PCR (Table S1).

In contrast, maxima in spring and summer of culturable *Legionella* for both water and biofilm were observed in the water network of a campus close to Haifa [13]. Reasons for these divergent observations might be manifold. A main di fference was the presence of water reservoirs in the West Bank hospitals. Depending on size and management, reservoirs may serve as a bu ffer for water temperature and thus may change seasonal e ffects on *L. pneumophila* in water and biofilm.

#### *4.3. Influence of Physico-Chemical and Bacteriological Factors on Legionella Abundance*

To determine factors that influence the prevalence of *L. pneumophila* at the di fferent sampling sites a set of physico-chemical and bacteriological parameters was analysed with respect to their relationship with *Legionella* abundance in water and biofilm of the eight sampling sites. In addition, the abundance of relevant *L. pneumophila* MLVA-genotypes was included in this analysis.

In terms of physico-chemical parameters, only the magnesium concentration showed a significant negative correlation with *Legionella* abundance in water and biofilm (Figures 4 and 5). Magnesium concentration showed a very close negative correlation with the Ca/Mg ratio but was not correlated with the calcium concentration. Moreover, the calcium concentration did not correlate with *Legionella* abundance. Due to this Mg vs. Ca/Mg relationship, *Legionella* abundance showed a positive correlation with the Ca/Mg ratio. Therefore, we hypothesize that either Mg, the Ca/Mg ratio or a factor closely related to Mg had an influence on *Legionella* abundance in the drinking water of the West Bank. In terms of bacteriological factors, there was no correlation of *Legionella* abundance in water and biofilm with heterotrophic plate counts, but with the prevalence of specific *L. pneumophila* genotypes and their clonal complexes. The magnesium concentration also showed a negative correlation with the respective specific genotypes and clonal complexes of *L. pneumophila* (Supplementary Table S1).

The magnesium concentrations for all sampling sites were rather high ranging from 21 to 40 mg/L. High magnesium concentrations were observed in a hydrological study of spring and ground water in the West Bank with especially high values in the Eastern part of the West Bank [34]. To the best of our knowledge, no study on *Legionella* abundance in drinking water is available that deals with drinking water with magnesium concentrations of this high level which is due to the specific geological conditions.

In general, an impact of magnesium on *Legionella* abundance has not ye<sup>t</sup> been demonstrated so far. There were large-scale investigations in residential drinking water distribution systems (DWDS) that included magnesium in the overall analysis of environmental drivers for *Legionella* abundance [35,36]. However, due to low magnesium concentrations in these DWDS (<3 mg/L), an e ffect of magnesium was not observed. On the other hand, some *L. pneumophila* studies in drinking water provide data on higher magnesium concentrations [32,33]. Borella et al. showed in the Italian hotel study that samples with no culturable *Legionella* had the highest Mg concentrations (mean 19 mg/L) [33]. Leoni et al. (2005) found in their study of buildings in Bologna based on the same public water supply, that absence of culturable *Legionella* significantly correlated with a higher Mg content of 16 mg/<sup>L</sup> vs. 11 mg/<sup>L</sup> for their presence [32]. The smaller the studied hot water systems were, the lower the *Legionella* counts presumably due to water softening devices in larger hot water systems. The observation of reduced *Legionella* abundance at sites with higher Mg concentrations may indicate that Mg may also have played a role at the Italian sites at low concentrations, in comparison to the West Bank study. However, the aspect of Mg impact on *Legionella* was not further studied, nor was there an analysis of *L. pneumophila* genotypes for these studies.

In summary, there are quite a few indications that Mg could play a role for suppression of *L. pneumophila* abundance. However, future more detailed studies including growth studies with Mg using a set of di fferent genotypes of *L. pneumophila* are needed to elucidate this aspect in more detail.

#### *4.4. Biogeography of L. pneumophila Genotype Prevalence*

A large set of 27 MLVA-genotypes was retrieved from 180 *L. pneumophila* isolates with most isolates obtained from biofilm due to the low abundance of culturable *L. pneumophila* strains in the bulk water. Only five strains were obtained from water, assigned to four di fferent genotypes (Table 2) that were frequently isolated from biofilm. As shown in detail by Zayed et al. [20], the *L. pneumophila* population had a high uniqueness, i.e., the major fraction of the MLVA-genotypes (20 out of 27 genotypes) has been described so far only for the West Bank. In addition, the distribution of genotypes among the four clonal clusters (VACC) indicates a high genetic diversity of the whole strain set.

The prevalence of the genotypes in the West Bank showed a site-specific regional diversity (Figure 6). Moreover, a large fraction of the genotypes unique for the West Bank occurred only in one site. Zayed et al. [20] suggested that the site-specific groundwater-based individual water sources may have contributed to the di fferences among the di fferent sites.

As an index of genotype diversity per sampling site, the ratio of the number of genotypes vs. the number of strains retrieved from the respective site was used. A correlation analysis showed that sites where high numbers of strains were isolated had a low genotype diversity compared to sites with low isolate numbers (Supplementary Table S2, Supplementary Figure S4). Only from sites with low genotype diversity were isolates obtained from water. This may indicate that higher numbers of *L. pneumophila* in the bulk water reduce the diversity in biofilm, or, that high *L. pneumophila* counts in water are often due to a rather restricted number of genotypes or even a single genotype. The observations in the water network near Haifa support the hypothesis that a single/few specific genotype(s) may cause high abundance in drinking water and biofilm [13].

On the other hand, this aspect could also have contributed to the obtained diversity of genotypes in the West Bank: the low abundance of *L. pneumophila* in the drinking water may have supported or conserved a larger diversity in the biofilms.

#### *4.5. Environmental Drivers of Genotype Consortia or Do L. pneumophila Genotypes Prefer Specific Niches?*

MLVA-genotypes with more than three representatives were analyzed by average hierarchical cluster analysis and PCA with respect to their co-occurrence and positioning in the frame of bacteriological and physico-chemical parameters. Both statistical analyses showed comparable groupings of the genotypes in three groups characterized by a distinct set of environmental parameters describing the niche of each group (Figure 7, Table 3).

The highest abundance of culturable *Legionella* in water was associated with group A, comprising two VACC11-genotypes (Gt 10(93), Gt 10(141)) and one VACC2-genotype (Gt 64(74)). This group was also characterized by low chloride and low magnesium concentrations compared to the other two groups. Group B comprised four genotypes: two of VACC1, the highly abundant Gt 4(17) and Gt 63(83), one genotype of VACC2 (Gt 13(72)), and one of VACC11 (Gt 9(93)). This group was distinct from groups A and C with respect to most parameters but was with most parameters in between the other two groups. Groups B and C had higher average Mg concentrations. This may indicate that these genotypes are more tolerant towards Mg. Group C comprised the VACC1-genotype Gt 6(18) and the VACC5-genotypes Gt 16(1) and Gt 40(47) that had a high prevalence in Hebron in the South of the West Bank. The environmental parameters of this group were characterized by high sulphate, chloride, and calcium concentrations and a resulting high content of TDS.

The impact of the environment in DWDS on *L. pneumophila* was usually analyzed with respect to the species level or to the serogroup (Sg) level [32,33]. Borella et al. (2005) showed evidence that hard water selected against Sg 1, but in favor of Sg 2-14 [33]. These observations are not supported by our study, where sites with the highest hardness were inhabited by group B and C, that both comprised members of Sg 1 and Sg 6 (Tables 2 and 3); group A inhabiting the softest water comprised only members of Sg 6. However, all of these sites of the West Bank had relatively hard water, and the high Mg concentrations may have had an additional impact.

On the MLVA-genotype level, Rodriguez-Martinez et al. showed for the first time-distinct niche preferences for *L. pneumophila* genotypes most pronounced with respect to temperature in a campus water network (Haifa, Israel) [13]. Niche preferences were previously shown for many aquatic bacterial species with a few studies tackling the subspecies or clonal level [37,38]. For *L. pneumophila*, this is the first assessment of niching of its genotypes as characterized by a set of environmental parameters in drinking water systems of a larger region such as the West Bank. We think that the information on niche preferences in combination with a genotyping at an adequate resolution (such as MLVA with 8 to 13 loci) could be helpful in order to better understand and model abundance of *L. pneumophila* in DWDS. These genotypic groups occupying environmental niches could be considered as ecotypes, i.e., a set of strains of a bacterial species inhabiting the same niche [38].

Though a broad set of environmental factors were analyzed for describing the niches of *L. pneumophila* genotypes, more studies including more conceivable factors are needed to complement the niche understanding, e.g., the abundance of protozoa and their species composition and their interaction with drinking water bacteria and especially the di fferent genotypes of *L. pneumophila* could be of high relevance [6]. As shown by Sharaby et al. [39], the interaction of *L. pneumophila* with protozoa may be genotype-dependent in addition to being temperature-dependent in a genotype-dependent way.

Studies on physiological and autecological traits of the relevant genotypes are considered crucial to allow better predictions and modeling [39]. Overall, we observed a correlation of specific genotypes with specific environmental niches which might be stemming from the di fferent types of local groundwater used as source water for the respective drinking water.

#### *4.6. Relevance of the Observed Genotypes for Human Health and Environmental Issues in the West Bank and Worldwide*

MLVA-genotyping of *L. pneumophila* can be considered as an economic genotyping method that has a good level of resolution for addressing clinical and environmental aspects [20]. MLVA-genotyping was shown to be comparable to sequenced-based typing (SBT) but has a higher resolution, which is especially of relevance for a better resolution of the highly important large STs such as ST1. Comparability and the increased resolution was shown for the method analyzing 8 loci (MLVA-8) as for the here used MLVA-8(12) using 13 loci [12,19,20,27] Using 13 loci increased the resolution compared to eight loci significantly; however, the used nomenclature allows a direct comparison of MLVA-8 with MLVA-8(12): the genotype indicated before brackets indicates the genotype according to MLVA-8, while the number in brackets refers to the genotype obtained from MLVA-12. Compared to SBT, MLVA subdivided strains from the West Bank of the larger STs into several MLVA-genotypes) [20].

Based on the InternationalMLVA data base and larger studies [12,16,27], the worldwide distribution and relevance for health and environment of the strains from the West Bank were estimated [20]. Most of the genotypes (20 out of 27) were considered as unique for the West Bank. Having high relevance worldwide as clinical and environmental MLVA-genotypes, Gt 4(17) and Gt 64(74) can be regarded as having MLVA-profiles identical to *L. pneumophila* strain Paris and strain Philadelphia-1, respectively. In Israel, Gt 6(18) played an eminent role as clinical and environmental genotype in addition to GT 4(17). For more details on the occurrence of the remaining genotypes not exclusive to the West Bank see Zayed et al. [20].

More specifically for the West Bank, some of the highly abundant MLVA-genotypes can be of special health relevance. PCR-based direct analysis of the sequence types in respiratory specimens of pneumonia patients of the West Bank revealed that genotypes belonging to ST1 and ST461 were present in half of the *L. pneumophila* positive clinical specimens [40]. We assume that the highly abundant representatives of these sequence types of our study are the relevant pathogenic genotypes (Table 2), meaning that Gt 4(17) may have been relevant for the detected ST1 infections, and the highly abundant genotypes in the South of the West Bank, i.e., Gt 9(92), Gt 10(93), and Gt 10(141), may have been responsible for the ST461 infections [18].

The highly abundant MLVA-genotypes Gt 4(17) and Gt 6(18) from the West Bank were also of high relevance in drinking water distribution systems and clinical isolates in Israel [13,14,39]. In contrast to observations in the West Bank, Gt 4(17) and Gt 6(18) had a high abundance in water and biofilm of a campus drinking water network in Northern Israel. The presence of Gt 4 was associated with average *Legionella* counts in water of 2500 CFU/L at an average water temperature of 20.6 ◦C. The presence of Gt 6 was associated with average *Legionella* counts in water of 240 CFU/L at an average water temperature of 27.9 ◦C. In the West Bank, Gt 4(17) was only detected in biofilm with no detection of culturable *Legionella* in bulk water; Gt 6(18) was endemic at one site, isolated regularly from biofilm and only once from water.

Rodriguez-Martinez et al. [13] concluded that the presence of Gt 4 could be considered as an indicator of high *Legionella* presence in drinking water, and they suggested Gt 4 as indicator genotype. Based on the observations in the West Bank where Gt 4 was very frequently observed in biofilms in the Northern part without co-occurrence of high *Legionella* counts in water, we sugges<sup>t</sup> that the presence in biofilm might be not an indicator for high *Legionella* counts in water [13]. Due to the worldwide occurrence of Gt 4 and the observations in Israel and the West Bank, we sugges<sup>t</sup> that Gt 4 may be regarded as an indicator of high *Legionella* abundance when showing up in the water phase. Furthermore, the presence of Gt 4 in biofilm might be considered as warning that, if conditions change, a "*Legionella* bloom" may be at risk.

#### *4.7. Relevance of the Findings of this Study for Drinking Water Management Strategies*

*L. pneumophila* is a major water-based pathogen, and its high abundance in drinking water is therefore a significant health risk. On the other hand, there are substantial uncertainties in the assessment of this risk [41,42]. Our study revealed that *L. pneumophila* populations in drinking water are composed of a set of genotypes sharing similar niche characteristics. Physico-chemical factors seemed to determine these niches and may also have shaped the *L. pneumophila* community, i.e., consortia of genotypes. The response of specific MLVA-genotypes to the environmental factors seemed to determine the abundance of *Legionella* in drinking water. Understanding the relationship between *L. pneumophila* genotypes and their environmental drivers might be crucial for understanding *L. pneumophila* abundance and the design of managemen<sup>t</sup> concepts. If there are generally important genotypes (such as Gt 4(17)) and environmental drivers (such as Mg) or if there is a high individuality between the drinking water sites, future studies have to elucidate for a broad set of climatic regions and in more detail. Furthermore, a broader set of conceivable niche-relevant parameters and environmental drivers should be included in future studies. The overall finding of our study is that a genotype-based ecology for understanding *L. pneumophila* abundance in artificial drinking water distribution systems can be considered as of high relevance for their managemen<sup>t</sup> and to this end also to human health. Therefore, we emphasize the need of a genotype-based ecology for *L. pneumophila* enabling the definition of niches for specific genotypes, their co-occurrence, and interactions.
