Occurrence of Pseudomonas spp. in Raw Vegetables: Molecular and Phenotypical Analysis of Their Antimicrobial Resistance and Virulence-Related Traits

Pseudomonas is characterized by its great capacity to colonize different ecological niches, but also by its antimicrobial resistance and pathogenicity, causing human, animal, or plant diseases. Raw and undercooked food is a potential carrier of foodborne disease. The aim of this study was to determine the occurrence of Pseudomonas spp. among raw vegetables, analysing their antimicrobial resistance, virulence, and molecular typing. A total of 163 Pseudomonas spp. isolates (12 different species) were recovered from 77 of the 145 analysed samples (53.1%) and were classified into 139 different pulsed-field gel electrophoresis patterns. Low antimicrobial resistance levels, but one multidrug-resistant isolate, were found. Among the 37 recovered P. aeruginosa strains, 28 sequence-types and nine serotypes were detected. Eleven OprD patterns and an insertion sequence (ISPa1635) truncating the oprD gene of one imipenem-resistant strain were found. Ten virulotypes were observed, including four exoU-positive and thirty-one exoS-positive strains. The lasR gene was absent in three ST155 strains and was truncated by different insertion sequences (ISPre2, IS1411, and ISPst7) in other three strains. High biofilm, motility, pigment, elastase, and rhamnolipid production were detected. Our study demonstrated a low occurrence of P. aeruginosa (18%) and low antimicrobial resistance, but a high number of virulence-related traits in these P. aeruginosa strains, highlighting their pathological importance.


Introduction
Vegetables and fresh fruit are important products in a healthy diet. In recent years, the search for a good lifestyle has led to an increased consumption of fresh products. Nevertheless, vegetables have become increasingly recognised as potential carriers of foodborne diseases due to various contamination sources, such as dust, soil, manure, irrigation water, or wild animal faeces [1,2]. Moreover, fresh vegetables which are grown close to the soil, are often consumed raw, exposing consumers to the risk of infection [3]. Most foodborne diseases are not reported, and sometimes outbreaks may affect a wide number of people. Thus, there is a special interest to know the epidemiology and spread of foodborne pathogens that are adaptable to different environmental conditions [4]. Additionally, bacteria can develop antimicrobial resistance due to spontaneous mutations or acquisition of resistance mechanisms by horizontal gene transfer. Antimicrobial resistant isolates can also spread
The 163 isolates were classified into twelve different Pseudomonas species. P. putida (51 isolates), P. aeruginosa (50 isolates) and P. mendocina (32 isolates) were the most abundant ones and were isolated from 46, 26 and 29 samples, respectively (Table 1). Thirty samples harboured more than one different Pseudomonas species (Table S1). Table 1. Description of the twelve different Pseudomonas species isolated from raw vegetables, including the number of isolates and the samples where they were recovered.
None of the 163 isolates showed class A carbapenemase, metallo-beta-lactamase (MBL) or extended spectrum beta-lactamase (ESBL) phenotypes, whereas the AmpC inducible phenotype was detected in 31% of the isolates, which included all P. aeruginosa isolates and the multidrug resistant P. fluorescens isolate (Ps876) (Table S1).

Molecular Typing
One hundred and thirty-nine different pulsed-field gel electrophoresis (PFGE) patterns were detected among the 163 Pseudomonas spp. (Table S1), and regarding the 50 P. aeruginosa isolates, 36 PFGE patterns were observed (Table 2). Indistinguishable patterns were only detected among isolates from the same sample, selecting one strain for next steps; except for two P. aeruginosa isolates that showed the same PFGE pattern, but they were recovered from two different chard samples (Tables 2 and S1). One strain with different PFGE pattern and species per sample were included in further studies. Additionally, three P. putida isolates from the same cucumber were included because they showed the same PFGE profiles but different resistance phenotypes. After these criteria, 142 Pseudomonas spp. strains were chosen: 105 strains corresponding to Pseudomonas non-aeruginosa species and 37 P. aeruginosa strains (Table S1).
Twenty-eight different sequence types (ST) were determined among the 37 P. aeruginosa strains using Multilocus sequence typing (MLST) method ( Table 2). Eight of them (ST2416, ST2427-ST2432 and ST2448) were first described in this study and named by the MLST database. Seven ST were repeated more than once: ST155, ST274, ST982, ST1226, ST1228, ST2416 and ST2432 (Table 2 and Figure 1). The P. aeruginosa strains were distributed into three clusters (Cluster I, II and III) when a phylogenetic tree based on the MLST was obtained ( Figure 1). Cluster I was an outlier and only included the Ps760 strain (ST2448). Cluster II included the new high-risk clone ST155, whereas the cluster III possessed the intercontinental clones, ST253, ST274 and ST395. The new ST were distributed between both clusters.

Serotyping
Nine different serotypes were identified in these 37 P. aeruginosa strains, in addition to five non-agglutinable (Ps798, Ps839, Ps884, Ps892 and Ps913) and one poly-agglutinable (Ps796) strains. Serotype O:6 was the most predominant (40.5%), followed by O:5 (10.8%) and O:1 (8.1%) (  them (ST2416, ST2427-ST2432 and ST2448) were first described in this study and named by the MLST database. Seven ST were repeated more than once: ST155, ST274, ST982, ST1226, ST1228, ST2416 and ST2432 (Table 2 and Figure 1). The P. aeruginosa strains were distributed into three clusters (Cluster I, II and III) when a phylogenetic tree based on the MLST was obtained ( Figure 1). Cluster I was an outlier and only included the Ps760 strain (ST2448). Cluster II included the new high-risk clone ST155, whereas the cluster III possessed the intercontinental clones, ST253, ST274 and ST395. The new ST were distributed between both clusters.

Figure 1.
Maximum-likelihood phylogenetic tree based on the sequence type (MLST) and phenotypic characteristics of the 37 P. aeruginosa strains analysed in this study. The phylogenetic tree was obtained using the IQTREE v.1.6.1 [14] software. The iTol v.4 [15] program was used to visualise the phylogenetic tree and to perform the eight heatmaps. In order: biofilm biomass production (CV), bacterial metabolic activity inside the biofilm (FDA), pyocyanin and pyorubin production, elastase activity, swimming and swarming motility and rhamnolipids production. Legend values belong to the minimum and maximum data for each phenotypic assay ( Table 2). The three clusters (I, II and III) were marked with different colours: yellow (Cluster I), red (Cluster II) and blue (Cluster III).

Serotyping
Nine different serotypes were identified in these 37 P. aeruginosa strains, in addition to five non-agglutinable (Ps798, Ps839, Ps884, Ps892 and Ps913) and one poly-agglutinable (Ps796) strains. Serotype O:6 was the most predominant (40.5%), followed by O:5 (10.8%) and O:1 (8.1%) ( Table 2)  Maximum-likelihood phylogenetic tree based on the sequence type (MLST) and phenotypic characteristics of the 37 P. aeruginosa strains analysed in this study. The phylogenetic tree was obtained using the IQTREE v.1.6.1 [14] software. The iTol v.4 [15] program was used to visualise the phylogenetic tree and to perform the eight heatmaps. In order: biofilm biomass production (CV), bacterial metabolic activity inside the biofilm (FDA), pyocyanin and pyorubin production, elastase activity, swimming and swarming motility and rhamnolipids production. Legend values belong to the minimum and maximum data for each phenotypic assay ( Table 2). The three clusters (I, II and III) were marked with different colours: yellow (Cluster I), red (Cluster II) and blue (Cluster III).

Characterisation of Porin OprD
The oprD gene was amplified in all P. aeruginosa strains, and eleven amino acidic OprD profiles were distinguished (Tables 2 and S2). Only two strains had the same pattern as P. aeruginosa PAO1 (pattern A, wild type), and pattern B was the most frequently detected (23 strains). The deletion of two amino acids in the region from amino acid 372 to 383 of the loop 7 (loop L7-short), which encodes a protein OprD of 441 amino acids, was identified in 24 strains (64.8%) (patterns B and E) (Tables 2 and S2). Regarding the two imipenemresistant strains, pattern B was observed in Ps839 strain, and the Ps884 strain showed the oprD gene truncated by the insertion sequence ISPa1635 at nucleotide position 561 (pattern K). This insertion sequence belongs to the IS4 family, and this is the first description of ISPa1635 truncating oprD gene. Thus, the sequence was included in GenBank with the accession number MH050332.
A total of 11 P. aeruginosa strains (one per pattern) with different protein OprD profiles were selected to study their outer membrane proteins by SDS-PAGE. The OprD band was detected in all tested strains except in pattern K, which corresponded to the strain with ISPa1635 element truncating the oprD gene.

Virulence Patterns
The presence of virulence and QS genes was investigated in the 37 P. aeruginosa strains, and ten different virulence patterns were obtained (Tables 2 and S3). Regarding the T3SS, exoU gene was detected in 4 P. aeruginosa strains and exoS in 31 strains. Neither exoU nor exoS genes were amplified in two strains (pattern VII). Moreover, exoA and exoY genes did not amplify in one and three strains, respectively. The exoT, lasA, lasB, aprA, rhlAB, rhlC, rhlI, and rhlR genes were detected in all strains, whereas exlA gene was absent in all of them. The lasI and lasR genes, involved in the QS system, were not amplified in three strains (pattern IV) that belonged to ST155 (Ps764, Ps839, Ps892). In the remaining ST155 strain (Ps845), as well as in two more strains (Ps796, Ps852), the lasR amplicon sized higher than 2,000 bp, resulting in the first description of three insertion sequences (IS1411, ISPre2, and ISPst7, respectively) and truncating this gene (Table 2 and Figure 2). All lasR sequences were submitted in GenBank (accession number): IS1411 (MH050330), ISPre2 (MH050329), and ISPst7 (MH050331). Finally, the lasR gene of Ps893 strain sized lower than expected, 647 instead of 720 bp, due to the presence of a deletion at the beginning of the gene (Figure 2).  Table 2 and Figure 1 summarise the biofilm biomass production (CV) and the bacterial metabolic activity inside the biofilm (FDA) of the 37 P. aeruginosa strains. The 92% of P. aeruginosa strains displayed higher values for biomass biofilm production than P. aeruginosa PAO1. Ps883 and Ps796 strains showed the highest percentages (952% and 600%, respectively), whereas Ps893, Ps851 and Ps733 exhibited the lowest ones (56, 65 and 90%, respectively). For FDA assay, 73% of strains showed more metabolic activity than the control strain P. aeruginosa PAO1. Ps752, Ps796, Ps845, Ps852 and Ps913 were the highest producers (>700%), and Ps798 the lowest producer (26.2%). Likewise, strains that had an absent or truncated lasR gene showed high levels of biomass production and bacterial metabolic activity inside the biofilm, except for Ps893 that showed low levels in  Table 2 and Figure 1 summarise the biofilm biomass production (CV) and the bacterial metabolic activity inside the biofilm (FDA) of the 37 P. aeruginosa strains. The 92% of P. aeruginosa strains displayed higher values for biomass biofilm production than P. aeruginosa PAO1. Ps883 and Ps796 strains showed the highest percentages (952% and 600%, respectively), whereas Ps893, Ps851 and Ps733 exhibited the lowest ones (56, 65 and 90%, respectively). For FDA assay, 73% of strains showed more metabolic activity than the control strain P. aeruginosa PAO1. Ps752, Ps796, Ps845, Ps852 and Ps913 were the highest producers (>700%), and Ps798 the lowest producer (26.2%). Likewise, strains that had an absent or truncated lasR gene showed high levels of biomass production and bacterial metabolic activity inside the biofilm, except for Ps893 that showed low levels in comparison with reference P. aeruginosa PAO1 strain.

Motility
The different swarming and swimming patterns detected among the 37 P. aeruginosa strains are included in Figure S1. Analysing the swimming results, many strains belonging to the Cluster III showed higher motility than the remaining ones (Table 2 and Figure 1). In fact, 23 of 37 strains (62.1%) covered the entire Petri dish surface (5363.7 to 6400 mm 2 ) in swimming and swarming. Ps893 and Ps884 strains described the lowest swimming (21.6 and 25.2 mm 2 ) and swarming (30.0 and 38.3 mm 2 ) values (Table 2 and Figure 1). Ps796 showed a high value of swimming (5363.7 mm 2 ) but showed a low swarming value (97.2 mm 2 ). P. aeruginosa strains having an absent or truncated lasR gene showed medium and low levels of motility, except the Ps796 swimming (Table 2 and Figure 1). Ps892 (absent gene) and Ps893 (truncated gene) showed the lowest motility values. Considering the strains which carried an insertion sequence truncating the lasR gene (patterns IIa, IIb and VI), the swarming levels were lower than the swimming motility ones.

Elastase and Pigment Production
Results for quantification of pyorubin and pyocyanin production as well as elastase activity are summarised in Table 2 and Figure 1.
For pyorubin assay, 70.2% of strains showed high levels of production in comparison with P. aeruginosa PAO1. Besides, Ps845, Ps848 and Ps852 showed the highest levels of pyorubin. Conversely, 27% of strains were the lowest pyorubin producers, highlighting Ps733, Ps775 and Ps854 strains. For the pyocyanin values, 62% showed high levels of production, being Ps845 and Ps848 the most producers. Conversely, fourteen strains showed low levels in comparison with reference strain, and Ps883, Ps884 and Ps913 strains were the weakest producers. The four strains with an absent or truncated lasR gene showed low levels of pyocyanin production.
Regarding elastase activity assay, 83.7% of strains showed high levels of activity in comparison with P. aeruginosa PAO1. Ps846 and Ps855 were the most important elastase producers (Table 2 and Figure 1). Figure S2 shows some of the results obtained with the rhamnolipid assays in P. aeruginosa strains. The 94.6% of the strains showed halos ≥11 mm (the minimum value observed), many of them (10 strains) with halos ranging 19-20 mm of diameter (Table 2 and Figure 1). Ps851 and Ps884 did not produce rhamnolipids.

Discussion
Food and the environment have been described as reservoirs of bacteria harbouring antimicrobial resistance genes that could be transferred or mobilised into human pathogens. Moreover, the extensive use and even the misuse of antimicrobial agents in clinic, animal production, and agriculture could be a way to select and disseminate these resistant human pathogens [5], and among them Pseudomonas genus. Some reports showed different clinical cases caused by environmental Pseudomonas, such as P. mendocina, P. monteilii or P. putida [16][17][18][19], although P. aeruginosa is the most important pathogenic bacterium. In our study, the 53.1% of fresh vegetables were positive for Pseudomonas spp., and lettuce and chard were the most frequently contaminated vegetables, as well as in previous studies [12,13]. According to our results and previous reports [1,20,21], the vegetables cultivated in contact with the soil may be contaminated more easily with Pseudomonas coming from soil, fertilizers, manure or water used for irrigation. In contrast, the lowest presence of Pseudomonas spp. was among onion samples. This fact could be due to the layered structure of the onion and/or the bioactive compounds present in onions [22][23][24]. Previous studies revealed that fresh onions, even onion wastes, exhibited high antimicrobial activity against bacteria such as Escherichia coli, P. fluorescens and Bacillus cereus, among others.
Antibiotic susceptibility testing revealed that all Pseudomonas spp. were susceptible to aminoglycosides and fluoroquinolones, and they showed low resistance rates for aztreonam and carbapenems. Only two imipenem-resistant P. aeruginosa strains were detected, and none of them were an MBL producer. The imipenem-resistance of Ps884 strain was associated with the loss of function of its OprD porin due to the truncation of the oprD gene by the insertion sequence ISPa1635. The inactivation of this porin gene by insertion sequences has been deeply studied in clinical strains [25][26][27], but this is the first time that the ISPa1635 has been identified in a P. aeruginosa from food origin and truncating oprD gene. Conversely, the imipenem-resistant Ps839 strain showed the same amino acid changes detected in the OprD porin as those reported in carbapenem-susceptible P. aeruginosa isolates [28,29]. Thus, other resistance mechanisms such as active efflux pumps or AmpC hyperproduction could be involved in that phenotype.
Considering P. aeruginosa QS genes, rhlI and rhlR genes were amplified in all P. aeruginosa strains. However, lasI and lasR genes were not detected in three strains; the other three strains showed insertion sequences truncating the lasR gene (IS1411, ISPst7 and ISPre2), and one strain showed a short lasR gene leading the possibility of losing the QS function. Several reports have mentioned the frequency of lasI and lasR mutations, or the lack of lasR gene among clinical and environmental isolates to favour their adaptation or persistence [30][31][32]. They have also demonstrated that a mutation in lasR does not lead to virulence factors loss, due to the regulation mediated by the rhl system, taking control of the phenazines production or rhamnolipids synthesis [10]. In addition, there are two other QS mechanisms, the Pseudomonas quinolone signal (PQS) and the integrated quorum sensing (IQS) mechanism, able to replace, in many cases, the LasR function [6,10]. Regarding our phenotypic results, it is important to remark that a high percentage of analysed P. aeruginosa showed high levels of biofilm, pigments and rhamnolipids production, and elastase activity even those with an absent or truncated lasR gene; however, in these cases, the strains were not as mobile as the remaining strains. Nevertheless, the hypothesis of the action of other QS mechanisms could demonstrate the pathological importance of these P. aeruginosa strains.
In P. aeruginosa, the T3SS mechanism contributes to cytotoxicity and acute infections [6,7]. The exoU gene was detected in four strains, all of them situated in the same branch of the MLST cluster, including the ST253 [33] and the new one ST2427, and belonging to O:11 serotype or poly-agglutinable, as other reports [34]. Usually, the exoU gene is described in clinical isolates, but it has also been detected in environmental strains [29,35].
The pathogenicity and host adaptation of P. aeruginosa is associated with its worldwide dissemination and specific sequence types. Among all sequence types detected, unlike other reports regarding clinical isolates, the most important "high-risk clones" ST111, ST175 and ST235, were not found. However, intercontinental clones disseminated worldwide, newest high-risk clones, such as ST155 and ST244 [36,37], as well as the epidemic clone ST274 circulating in Spain [38,39], were detected. All of those, including ST252, ST253, and ST395 epidemic clones were also previously observed in clinical animal and environmental samples [36,[40][41][42]. However, none of the epidemic clones showed the same pathogen phenotype. In this case, the low motility but high biofilm production that possessed the STs from the second cluster, where ST155 clone was included. Conversely, the P. aeruginosa strains belonging to the third cluster exhibited higher motility, where high-risk clones ST253, ST274 and ST395 were englobed. However, there are some exceptions, such as P. aeruginosa strains with lacked or truncated lasR gene, Ps852 (ST267), Ps893 (ST395) or Ps764, Ps839, Ps845 and Ps892 (ST155). Further studies are needed to delve into the relationship between the genotype and phenotype of these epidemic clones because, to the best of our knowledge, this is the first time the analysis of this relationship is described in environmental samples.

Bacterial Isolates
One hundred and forty-five samples of raw vegetables were recovered from orchards (82 samples) and little markets (63 samples

Molecular Typing
The clonal relationship among the recovered isolates was determined by PFGE with SpeI restriction enzyme [28]. PFGE patterns were analysed by the Java program GelJ using the Dice coefficient [44].

Serotyping
P. aeruginosa strains were serotyped by slide agglutination according to the International Antigenic Typing Scheme (IATS), using 16 type O monovalent antisera specific for P. aeruginosa (O:1 to O:16) following the manufacturer's protocol (Bio-Rad, Temse, Belgium).

Characterisation of Porin OprD
Amino acid changes of the porin OprD were analysed by PCR and sequencing in all P. aeruginosa strains [29]. The mutations were determined by comparison with the sequence of the control strain P. aeruginosa PAO1 (GenBank accession number AE004091).
The outer membrane proteins (OMPs) of selected strains were stained with Coomassie Brilliant Blue and were visualised by SDS-PAGE (Sodium Dodecyl Sulphate-PolyAcrylamide Gel Electrophoresis) in a Bio-Rad Mini-Protean II apparatus (Bio-Rad, Temse, Belgium) as previously described [46]. P. aeruginosa PAO1 and PAO1 lacking-OprD were included as control strains.

Biofilm Quantification
The analysis of the total biofilm biomass was performed by crystal violet (CV) staining, and the bacterial metabolic activity inside the biofilm structure by fluorescein diacetate (FDA) assay among P. aeruginosa strains [47]. Both methods were carried out in flat-bottom microtiter 96-well plates after 24 h of bacterial incubation in Müeller-Hinton broth at 37 • C. For CV assay, 66% acetic acid and 10% CV were used, and the FDA working solution concentration was 0.1 mg/mL. Measures were performed using a POLARstar Omega microplate reader (BMG Labtech, Ortenberg, Germany). All assays were carried out in triplicate, including P. aeruginosa PAO1 as control.

Motility
Swarming and swimming motilities were studied in P. aeruginosa strains [47], placing 4 µL of bacterial suspension (1 × 10 9 cells in Luria-Bertani (LB) broth) on the middle of 0.5% (swarming) and 0.3% (swimming) LB agar plates and subsequent incubation at 37 • C overnight. The plates were imaged with Chemi Doc system (Bio-Rad, Temse, Belgium), and processed with Image Lab software (version 6.0.1, Bio-Rad). The entire plate area was 6400 mm 2 . All assays were performed in triplicate, including P. aeruginosa PAO1 as a control strain.

Elastase and Pigment Production
Bacteria were grown overnight in LB broth at 37 • C with shaking at 120 rpm. After centrifugation, 900 µL of supernatant was used to determine the elastase activity in P. aeruginosa strains by the Elastin Congo Red assay as previously described [48].
The chloroform extract method was used to quantify pyocyanin and pyorubin phenazines by measuring the absorbance of the corresponding solutions: the organic phase at 520 for pyocyanin, and the aqueous phase at 525 nm for pyorubin, using a POLARstar Omega microplate reader (BMG Labtech, Ortenberg, Germany) [49], and including P. aeruginosa PAO1 as control.

Rhamnolipids Detection
The detection of P. aeruginosa biosurfactant producers was carried out by the Cetyl Trimethylammonium Bromide-Methylene Blue (CTAB-MB) agar plates method [50][51][52]. One colony of each P. aeruginosa strain studied was inoculated in 3 mL of mineral salt medium (MSM) broth and was incubated at 35 • C and 130 rpm during 48 h. Following previous procedures, shallow wells of 6 mm were cut on the CTAB-MB agar plate surface. Twenty microliters of the inoculum were added into each well. The plates were incubated for 24-48 h at 35 • C, and then stored in the fridge for at least 24 h, to intensify the blue colour of the plates to facilitate the recognition of the rhamnolipids production. The halo was measured (mm). P. aeruginosa PAO1 was used as a positive control.

Conclusions
Pseudomonas strains contaminating fresh vegetables were found in this work, especially in lettuce and chard. A variety of different Pseudomonas species, including pathogenic to humans, such as P. aeruginosa, P. mendocina, P. monteilii or P. putida were detected. P. aeruginosa was recovered from 26 (17.9 %) of the vegetable samples and belonged to many different clones, comprising some international clones. Moreover, these strains showed low resistance to antibiotics but high presence of virulence-related traits, as high biofilm, pigments and rhamnolipids production. P. aeruginosa is an opportunistic human pathogen, and the food chain might be a source of transmission to humans. Identifying the natural reservoirs of this important pathogen and elucidating its molecular biology are crucial tasks in the pursuit of minimising its transmission. For all these reasons, the application of proper hygiene practices along the food production/supply chain is essential, not only for vegetable workers, but also for consumers.