Biofilm Formation among Stenotrophomonas maltophilia Isolates Has Clinical Relevance: The ANSELM Prospective Multicenter Study

The ability to form biofilms is a recognized trait of Stenotrophomonas maltophilia, but the extent of its clinical relevance is still unclear. The present multicenter prospective study (ANSELM) aims at investigating the association between biofilm formation and clinical outcomes of S. maltophilia infections. One hundred and nine isolates were collected from various geographical origins and stratified according to their clinical relevance. Biofilm formation was evaluated by the microtiter plate assay and correlated with microbiological and clinical data from the associated strains. Antibiotic susceptibility of the planktonic cells was tested by the disk diffusion technique, while antibiotic activity against mature biofilms was spectrophotometrically assessed. Most strains (91.7%) were able to form biofilm, although bloodborne strains produced biofilm amounts significantly higher than strains causing hospital- rather than community-acquired infections, and those recognized as “definite” pathogens. Biofilm formation efficiency was positively correlated with mechanical ventilation (p = 0.032), whereas a negative relationship was found with antibiotic resistance (r2 = 0.107; p < 0.001), specifically in the case of the pathogenic strains. Mature S. maltophilia biofilms were markedly more resistant (up to 128 times) to cotrimoxazole and levofloxacin compared with their planktonic counterparts, especially in the case of bloodborne strains. Our findings indicate that biofilm formation by S. maltophilia is obviously a contributing factor in the pathogenesis of infections, especially in deep ones, thus warranting additional studies with larger cohort of patients and isolates.


Introduction
Stenotrophomonas maltophilia is a globally emerging multidrug-resistant Gram-negative pathogen. It has a propensity to cause a plethora of opportunistic infections in humans, mainly associated with the respiratory tract [1].
Recently, a handful of studies have reported the frequency and characterization of biofilm-producing S. maltophilia strains prospectively isolated from several hospitals worldwide. Overall, the findings revealed that biofilm formation is highly conserved in S. maltophilia and occurs with relevant efficacy leading to high biomass amount [2,5,[13][14][15][16][17][18][19][20][21]. However, these studies were aimed at investigating the phenotypic and genotypic characteristics or virulence traits, without providing evidence for the relationship between biofilm formation and the clinical course of diseases.
In the present study, 109 S. maltophilia clinical isolates were collected during a multicenter prospective cohort study involving five European hospitals and were evaluated for biofilm formation, antibiotic resistance, and genetic heterogeneity. In addition, the efficiency of biofilm formation and the biofilm resistance to commonly used antibiotics were cross-referenced both with microbiological and clinical data aimed at determining possible relationships. The knowledge gained from these results may contribute to the design of novel diagnostic and therapeutic interventions to prevent and/or cure biofilm-associated infections caused by S. maltophilia.

Materials and Methods
This study (ANSELM, clinicAl sigNificance of StEnotrophomonas maLtophilia biofilM) was approved by the Ethics Committee of "G. d'Annunzio" University of Chieti-Pescara, Italy (permission number 1864, 11 December 2018).

Bacterial Strains
A total of 109 S. maltophilia isolates were collected from different clinical specimens at selected institutions in five European countries between February-July 2019.
One isolate per patient was included in the study. Species identification was performed using a miniaturized BD BBL Crystal system (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) or an automated system (Vitek 2 or Vitek MS system, bioMérieux SA, F-69280 Marcy l'Etoile, France; BD Phoenix 100, Becton, Dickinson and Company). Each isolate was stored at −80 • C in a Microbank™ cryogenic system (Biolife Italiana, Milan, Italy) until use when it was plated three times onto Müller-Hinton Agar (MHA; Oxoid SpA, Garbagnate M.se, Milan, Italy) to recover the original phenotypic traits.

Microbiological and Clinical Data
For each strain, the following microbiological and clinical data were collected by each Center participating in the study: (i) age and gender of the patient; (ii) isolation site; (iii) community-acquired infection (if it occurred at least 48 h prior to hospitalization) or hospital-acquired infection (if it occurred at least 48 h after admission); (iv) ward, if hospital-acquired infection; (v) clinical presentation; (vi) clinical diagnosis; vii) underlying comorbidities; (viii) risk factors; (ix) clinical outcome; and (x) antibiotic therapy, indicating drugs, dose, administration route, and duration of both "empiric" therapy (started before S. maltophilia isolation) and "targeted" therapy (that is a drug administrated based on the results of antimicrobial susceptibility testing).
The etiological role of each strain was defined according to the CDC guidelines [22] and were as follows: (i) "definite pathogen", if the patient had symptoms and signs of infection at the site of isolation and no other pathogen was isolated from that site; (ii) "probable pathogen", if the patient had symptoms and signs of infection at the site of isolation but the culture yielded polymicrobial growth; (iii) "possible pathogen", if the signs and symptoms of infection were evident but not clearly related to the site of isolation; (iv) "nonpathogen", if there was no evidence of infection at the time of isolation.

Standardized Inoculum Preparation
A standardized inoculum was prepared according to the intended use. For biofilm formation assay and microscopic observation, several colonies were grown overnight at 37 • C onto MHA, resuspended in 5 mL Trypticase Soy Broth (TSB; Oxoid SpA), and then incubated at 37 • C under agitation (130 rpm). After overnight incubation, the broth culture was corrected with sterile TSB to an optical density measured at 550 nm (OD 550 ) of 1.0 (corresponding to 1-2 × 10 9 CFU/mL), and then diluted 1:100 (v/v) in TSB to achieve a final inoculum concentration of 1-2 × 10 7 CFU/mL.
Inoculum size and purity were checked by a viable cell count on MHA.

Microtiter Plate (MTP) Assay for Biofilm Quantification
Biofilm biomass-including both adherent bacteria and extracellular polymeric substance (EPS)-was measured by crystal violet MTP assay. In brief, standardized inoculum (200 µL) was aseptically added to each well of a 96-well polystyrene tissue culture plate (Becton, Dickinson and Company) and incubated aerobically at 37 • C under static conditions. Wells containing TSB only were considered as controls. At the end of the incubation, spent medium was discarded and each well was washed twice with PBS (pH 7.2) (Sigma-Aldrich, Milan, Italy) to remove non-adherent cells. Biofilm samples were fixed by incubating plates at 60 • C for 1 h, and then stained for 5 min with Hucker-modified crystal violet (200 µL) [23]. After the plates were air-dried, crystal violet was extracted by exposure for 15 min to 200 µL of 33% glacial acetic acid (Sigma-Aldrich), and the biofilm biomass was then assessed by measuring the optical density at 492 nm (OD 492 ) (Sunrise, Tecan, Milan, Italy). According to the criteria proposed by Stepanović et al. [24], each strain was classified for biofilm formation efficiency as follows: (i) non-producer (OD ≤ ODc); weak-producer [ODc < OD ≤ (2 × ODc)]; moderate-producer [(2 × ODc) < OD ≤ (4 × ODc)]; and strongproducer (OD > 4 × ODc). Cut-off value (ODc) was defined as OD (mean negative control) + 3 × standard deviations.
MIC values for SXT and LVX were obtained by the broth microdilution technique, according to the CLSI guidelines [25]. MBC was evaluated in duplicate by culture; a 10 µL media from wells showing no visible growth at MIC determination were inoculated onto MHA. Following incubation at 37 • C for 24 h, an MBC value was defined as the minimum antibiotic concentration able to eradicate 99.9% of the starting inoculum.
Escherichia coli ATCC25922 and Pseudomonas aeruginosa ATCC27853 reference control strains were assessed in parallel for quality control.

Antibiotic Activity Against Biofilm Formation and Mature Biofilm
The Minimum Biofilm Inhibitory Concentration (MBIC) and the Minimum Biofilm Eradication Concentration (MBEC) values of LVX and SXT were assessed against a S. maltophilia biofilm. An aliquot (200 µL) of the standardized inoculum was added into each well of a 96-well polystyrene, flat bottom, tissue culture-treated microtiter (Iwaki, Bibby srl; Milan, Italy) and aerobically incubated at 37 • C for 24 h under static conditions. Biofilm samples were washed once with PBS (200 µL) and then exposed to each tested antibiotic at several concentrations previously prepared in CAMHB.
After a 20 h-exposure to antibiotics, biofilm samples were washed twice with sterile PBS (200 µL) and then TSB (200 µL) was added to each well to verify their effect on the biofilm. After 6 h-and 24 h-incubation at 37 • C, OD 620 of broth supernatant was spectrophotometrically assessed. MBIC was defined as the lowest antibiotic concentration allowing a regrowth of ≤10% compared to the positive (unexposed) control well readings (representing at least a 1-Log growth difference), whereas MBEC was defined as the lowest antibiotic concentration causing biofilm eradication (i.e., no growth, like negative controls exposed to DMSO).

Evaluation of Biofilms by Microscopic Analysis
The ultrastructure of the biofilm produced by selectedrepresentative S. maltophilia strains was evaluated by both Scanning Electron Microscopy (SEM) and Confocal Laser Scanning Microscopy (CLSM).
SEM analysis: biofilm sample grown (37 • C, 24 h, static) on a polystyrene coupon accommodated in a 6-well microtiter (Becton, Dickinson and Company) was washed twice in PBS, then fixed overnight in a mixture of 2% paraformaldehyde (Electron Microscopy Sciences, Rome, Italy) + 2% glutaraldehyde (Sigma-Aldrich) in 0.15 M sodium cacodylate buffer (pH 7.4; Honeywell Fluka, Milan, Italy), with the addition of alcian blue 0.1% (8GX; Sigma-Aldrich) aimed at detecting EPS. Samples were post-fixed for 90 min in 1% OsO 4 (Electron Microscopy Sciences) in 0.15 M cacodylate buffer, dehydrated in ascending ethanol series, stained en-bloc with 2% alcoholic uranyl acetate for 60 min and rinsed in 100% ethanol. Samples were then gold coated in a Desk Sputter Coater 1 (PVD, Teheran, Iran), and finally analysed with the Phenom Pro (Thermo Fisher Scientific, Monza, Italy) under "high-vacuum" modality and with an operating voltage of 15 kV.
Representative images were acquired during both SEM (ProSuite software Thermo Fisher Scientific) and CLSM (ZEN 2.3 SP1 software, Carl Zeiss, ver. 14.0) observations and processed for display using Photoshop (Adobe Systems Inc., San Jose, CA, USA) software.

Statistical Analysis
Each experiment was carried out at least in triplicate and repeated on two different occasions (n ≥ 6). The distribution of results was assessed using a D'Agostino-Pearson normality test, and then the differences in the biofilm biomass (OD 492 ) were evaluated accordingly: (i) using a Kruskal-Wallis test followed by a Dunn's multiple comparisons post-test (among more than 2 groups) or a Mann-Whitney test (between 2 groups), in case datasets did not pass the normality test; (ii) by an unpaired-t test, in cases of normally distributed datasets. The statistical significance of differences between variables was calculated by a Chi-square test. The relationship between biofilm biomass and antibiotic resistance level was assessed by linear regression analysis. MBIC/MIC and MBEC/MBC values were considered statistically significant if >2. The significance level was set at p < 0.05.
In reviewing the 76 putative pathogenic strains (i.e., those classified as "definite", "probable", or "possible" pathogen), the most common isolation site was airways followed by blood (59.2% vs. 30.3%, respectively; p < 0.001), whereas in the strains with a "definite" pathogen role, the most frequent isolation site was blood (60.8%; p < 0.01 vs. other groups). The infection was acquired with a comparable prevalence in community and hospital settings (65.8% vs. 63.5%, respectively). The previous administration of antibiotics was a risk factor, being significantly more prevalent among putative pathogenic strains compared to the non-pathogenic ones (70.9% vs. 29.1%, respectively; p < 0.0001). The administration of antibiotics cured the infection in most of cases (65.6% vs. 34.4%, respectively for cleared and uncleared outcomes; p < 0.001) and the same trend was observed for infections caused by strains classified as "definite" (69.6% vs. 26.0%, respectively; p < 0.001) and "probable" (44.4% vs. 26.7%, respectively; p < 0.05).

The Biofilm Forming Ability Is Highly Preserved in S. maltophilia, Although Strains with a Definite Etiological Role, Particularly Those From Blood, Show a Higher Efficiency
Biofilm forming ability in clinical S. maltophilia strains, as indicated by the results of MTP assay performed, is highly preserved (Figure 1). Namely, a great majority of the strains tested were recognized as biofilm producers (100 out of 109, 91.7%). However, the high variability observed in biofilm biomass values (OD 492 range: 0.150-3.089; coefficient of variation: 77.7%) suggests the existence of streaking strain-to-strain differences in the biofilm formation efficiency (Figure 1).  The biofilm formation capability was comparable between non-pathogen strains and those with a putative pathogenic role (93.9% vs. 90.8%, respectively), as indicated by the lack of significant differences in the median value of biofilm biomass ( Figure 2A). The finding was the same for three groups with ascribed pathogenic relevance ( Figure 2C) as well as for the strains originating from different isolation sites ( Figure 2E). When the efficiency of biofilm formation was estimated through the categorization of the strains into four classes, ranging from non-producer to strong producer, a similar distribution of the classes was noted among both putative pathogen and non-pathogen strains ( Figure 2B). Most strains were classified as "strong-producers" (p < 0.0001 vs. other classes), followed by "moderate-", "weak-", and "non-producers" ( Figure 2B). However, it is worth noting that all strains with a "definite" etiologic role were able to form biofilm and most of them were classified as "strong-producers", at a proportion significantly higher than that found for strains with "probable" and "possible" pathogenic role (73.9% vs. 42.4% vs. 25%, respectively; p < 0.05) ( Figure 2D). Further, the "strong-producer" phenotype was most prevalent among strains isolated from blood (78.3%; p < 0.0001 vs. other groups), followed by non-CF airway (50.0%), wound (40.0%), and CF airways (32.2%) ( Figure 2F).
Biofilm levels were also correlated to the setting where infection was acquired (hospital vs. community), previous administration of antibiotic therapy, administration of targeted antibiotic therapy and clinical outcome, and no significant differences in the median biofilm mass were observed ( Figure 3A,C,E,G). Categorization of the strains into four biofilm classes revealed the predominance of the "strong producers, regardless of the precipitating factor analysed ( Figure 3B,D,F,H). An important finding is that the strains able to form a higher biofilm amount were significantly more prevalent among HAI than CAI strains (60.6 vs. 33.3%, respectively for HAI and CAI; p < 0.05) ( Figure 3B).
With regard to other factors possibly contributing to strain-to strain differences in biofilm formation, the "strong-producer" phenotype was significantly associated with mechanical ventilation (p = 0.032). Ten out of 12 (83.4%) strains recovered from respiratory specimens were classified as "strong-producers", and one (8.3%) each for "moderate" and "non-producer" phenotypes.  (ii) pathogen (definite, probable, possible) or non-pathogen according to the CDC guidelines [22].
A decreasing trend in biofilm biomass median values was observed among multistrain lineages (ST4 > ST26 > ST28 > ST321) but remained below the level of significance (data not shown).

Crystal Violet Assay Is Highly Predictive in the Quantitative Analysis of Biofilm Formation.
To evaluate the predictive value of the MTP colorimetric assay used in this study to measure biofilm biomass formed by S. maltophilia, representative strains of strong-and weak-producer biofilm classes were comparatively evaluated by SEM and CLSM analyses. The biofilm formed by the "strong-producer" STMA_CZ_44 and BG10 strains was qualitatively and quantitatively more complex than that formed by the STMA_CZ_41 strain categorized as a "weak-producer" (Figures 5 and 6). The biofilm formed by STMA_CZ_44 and BG10 strains affects almost all the contact surfaces and comprised numerous cell clusters embedded into an amorphous EPS matrix ( Figures 5 and 6) whose composition was predominantly polysaccharidic, as shown by carbohydrate-binding Concanavalin-A stain at CLSM analysis ( Figure 6). In contrast, the biofilm formed by the "weak producer" STMA_CZ_41 showed less area coverage, cellularity, and EPS amount. Only three lineages were isolated from multiple countries: ST26 (Serbia and Spain), ST28 (Spain and Germany) and ST321 (Czech Republic and Spain) (Figure 4). It is interesting to note that STs isolated from the same country showed a low degree of similarity in the nucleotide sequence of MLST genes ( Figures S1-S7).

Crystal Violet Assay Is Highly Predictive in the Quantitative Analysis of Biofilm Formation
To evaluate the predictive value of the MTP colorimetric assay used in this study to measure biofilm biomass formed by S. maltophilia, representative strains of strongand weak-producer biofilm classes were comparatively evaluated by SEM and CLSM analyses. The biofilm formed by the "strong-producer" STMA_CZ_44 and BG10 strains was qualitatively and quantitatively more complex than that formed by the STMA_CZ_41 strain categorized as a "weak-producer" (Figures 5 and 6). The biofilm formed by STMA_CZ_44 and BG10 strains affects almost all the contact surfaces and comprised numerous cell clusters embedded into an amorphous EPS matrix (Figures 5 and 6) whose composition was predominantly polysaccharidic, as shown by carbohydrate-binding Concanavalin-A stain at CLSM analysis ( Figure 6). In contrast, the biofilm formed by the "weak producer" STMA_CZ_41 showed less area coverage, cellularity, and EPS amount.  . Ultrastructure analysis of S. maltophilia biofilm: confocal laser scanning microscopy. The biofilm was allowed to form in a μ-Dish for 24 h at 37 °C by S. maltophilia BG10 and STMA_CZ_41 strains, respectively representative of "strong" and "weak" biofilm producer classes. Biofilm sample was then stained using BacLight Live/Dead kit: Syto-9 tags live cells (green fluorescence); propidium iodide tags dead cells (red fluorescence); and Concanavalin-A tags extracellular polymeric substance (blue fluorescence). Merged: co-localization of Syto-9, propidium iodide and Concanavalin-A. Magnification: 1000×. Figure 6. Ultrastructure analysis of S. maltophilia biofilm: confocal laser scanning microscopy. The biofilm was allowed to form in a µ-Dish for 24 h at 37 • C by S. maltophilia BG10 and STMA_CZ_41 strains, respectively representative of "strong" and "weak" biofilm producer classes. Biofilm sample was then stained using BacLight Live/Dead kit: Syto-9 tags live cells (green fluorescence); propidium iodide tags dead cells (red fluorescence); and Concanavalin-A tags extracellular polymeric substance (blue fluorescence). Merged: co-localization of Syto-9, propidium iodide and Concanavalin-A. Magnification: 1000×.

The Antibiotic Resistance Level Is Higher Among Pathogenic Strains
The in vitro activity of seven antibiotics was evaluated by disk diffusion, and the results are shown in Figure S8. In the S. maltophilia strain population as a whole, the activity of minocycline and chloramphenicol (susceptibility rate of 96.3% and 90.7%, respectively) was significantly higher compared to the other antibiotics tested (at least p < 0.05). Ceftazidime had the lowest activity and was only effective against 47.3% of the strains. The same pattern of susceptibility rates was observed within groups of "non-pathogen" and pathogenic strains ( Figure S8). However, when the strains with a definite pathogenic role were considered separately, their susceptibility rates to all antibiotics tested were similarly high, ranging from 78.3% to 100%, except for ceftazidime (susceptibility rate: 69.6%; p < 0.05). It is worth noting that the levofloxacin susceptibility rate of pathogenic strains (95.7%) was significantly higher compared to that seen with non-pathogenic ones (72.7%; p < 0.05).
Next, the level of antibiotic-resistance was assessed by referring to both the number of resistances shown by each strain, and the multi-resistance phenotypes defined according to Magiorakos et al. [26]. The prevalence of the strains that showed no resistances was significantly lower among non-pathogenic strains than those with a definite pathogenic role (21.2% vs. 47.8%, respectively; p < 0.05) (data not shown). Non-MDR strains (range: 84.8-95.7%) were significantly more prevalent (p < 0.0001) compared to both MDR and XDR phenotypes, regardless of the group considered (data not shown).

The Efficiency of Biofilm Formation Is Affected by Antibiotic Resistance in Pathogenic Strains Only
We evaluated the existence of a relationship between biofilm formation efficiency and antibiotic resistance, and the results are summarized in Figures 7 and 8.
The efficiency of biofilm formation was further correlated with the level of antibiotic resistance, measured as the number of antibiotic-resistances showed by each strain (Figure 8). We established that the increasing number of resistances in pathogenic strains is inversely related to the median amount of biofilm they produce (median OD 492 : 0.660 vs. 0.471 and 0.275, respectively for strains showing no resistance, two resistances and three resistances; at least p < 0.05) ( Figure 8A). A linear regression analysis confirmed this trend since a significant negative relationship (r 2 = 0.107; p < 0.001) between the amount of biofilm formed and the antibiotic resistance level was found ( Figure 8C). In contrast, no significant correlation between biofilm production and level of antibiotic resistance was found among non-pathogen strains ( Figure 8B).
The frequency of biofilm classes was not dependent on the MDR phenotype (data not shown). However, a specific trend among pathogenic strains was observed where non-MDR phenotype was significantly associated with a higher average biofilm amount compared to the MDR strains (median OD 492 : 0.574 vs. 0.216, respectively; p < 0.01) ( Figure 8D).

The Preformed S. maltophilia Biofilm Is Highly Resistant to Both Cotrimoxazole and Levofloxacin
The susceptibility of mature biofilms formed by a selected set of strains representative of different biofilm formation efficiency, pathogenetic roles and sources was evaluated in vitro by measuring the MBIC and MBEC values. Cotrimoxazole and levofloxacin were selected because considered as "first-line" antibiotics for the treatment of S. maltophilia infection; the results are graphed in Figures S10-S13 and summarized in Tables 3 and 4.
Overall, the planktonic-to-biofilm lifestyle transition significantly increased S. maltophilia resistance to cotrimoxazole and levofloxacin, regarding biofilm formation and mature biofilm. In particular, cotrimoxazole MBIC/MIC values increased significantly for all the strains tested (MBIC/MIC range: 32 to 256), while MBEC/MBC value increase ranging from >2 to >64 was noted in 93.7% of the strains tested (Figures S10 and S11, Table 3).
The inhibitory effect of levofloxacin against biofilm formation was observed in 11 out of 16 (68.7%) strains and the MBIC/MIC ratios ranged from 4 to >1024, whereas MBEC/MBC ratios ranged from 8 to >128 suggesting that, in all the strains tested, there was an increased resistance to biofilm formation ( Figures S12 and S13, Table 4).

Discussion
The formation of biofilms by bacteria on inert surfaces has been extensively studied, and there appears to be a direct relationship between the capability of the organisms to form a biofilm and their pathogenicity. For instance, E. coli isolates causing prostatitis [27] or pyelonephritis [28,29] show a higher efficiency in biofilm formation than isolates from patients with cystitis or those without a pathogenic role. Similarly, P. aeruginosa isolates from the infected lungs of patients with chronic obstructive pulmonary disease tend to produce more biofilm than those from blood cultures [30].
Previous data on biofilm formation by clinical S. maltophilia isolates are limited. In addition, most of the studies have reported retrospectively on the frequency and characterization of biofilm-producing S. maltophilia from collected strains [10,[13][14][15]31,32] without giving any information about the relationship with clinical infection outcomes.
To the best of our knowledge, this is the first prospective, multicenter study investigating the association between the biofilm formation capability of S. maltophilia strains and the clinical outcomes of infections they cause, by using a representative collection of isolates stratified according to their clinical relevance. The overall results of our study strongly support a role of biofilm formation as contributing factor in the pathogenesis of S. maltophilia infections.
In agreement with previous studies [15][16][17][18]33], our findings showed that the ability to form biofilm is highly preserved in clinical S. maltophilia isolates, albeit with significant interstrain variability. Biofilm formation was noted among both the pathogenic and nonpathogenic strains, which suggests its essential role for bacterial persistence regardless of the colonization sites. However, the ability to form a higher amount of biofilm was significantly more prevalent among strains causing hospital-rather than communityacquired infections. In addition, strains with a "definite" etiological role were recognized as the most efficient at forming biofilms; namely, the proportion of "strong-producers" was significantly higher in this category as compared to the proportions recorded among strains associated with a "probable" or "possible" pathogenic role.
Further, microscopic observations indicated that clinically relevant strains form structurally complex mature biofilms within 24 h, suggesting that their transmission may be facilitated by the fast development of biofilms on clinical equipment [16].
Overall, these findings strongly suggest that the ability to produce biofilm can be considered a determinant of virulence for S. maltophilia, and support the role played by biofilm formation in establishing an infection.
When the efficiency of biofilm formation was correlated with the type of sample, an obvious trend emerged. The isolates from blood showed the highest capacity for biofilm formation, while those isolated from airways of CF patients produced the lowest amounts of biofilm. Our findings are consistent with an increased adherence to human bladder HTB-9 cells by blood-derived S. maltophilia strains compared to those from environmental or urinary sources [17].
The evidence that biofilm production by clinical strains is affected by the site of isolation might imply that regulation of the genes encoding biofilm components is modulated by environmental influences. In support of this, we found previously a significant association between a CF-associated genotype and strong biofilm formation [5], while Willsey et al. [34] observed recently that several biofilm-associated genes were induced differentially in synthetic CF sputum medium and it corresponded to an increase in aggregation and biofilm formation.
The MLST analysis performed on a representative subset revealed that all the STs were represented by one or two isolates, except for ST26 and ST28. These findings confirm those from previous studies [2,17,33,35], where S. maltophilia strains also showed a high heterogeneity.
Some of the STs found in this study, such as ST4, ST24, and ST77 are epidemic in that they are widely spread over the world, isolated both from human and animal sample types, as reported in PubMLST [36]. S. maltophilia ST28, which we found both in Spain and Germany, also circulates in Korea and the USA isolated from human samples [36]. Conversely, several STs are endemic for each country, as in the case of ST211 and ST26 (USA), ST174 (Tunisia), ST186 (Italy), ST295 (China), ST172 (Brazil), and ST371 (Australia) all isolated from human samples [36]. Other lineages we found in the present study (ST93, ST249, ST265 and ST319) are novel since they were not reported to date in PubMLST [36].
Although clonal lineages identified in our study were not equally effective in forming biofilm, it is worth noting that clonally related isolates from different laboratories did not share the same effectiveness to form biofilms. This suggests, once again, that biofilm formation is not a sequence type-specific feature, and that its expression varies substantially under different conditions. Further studies are, however, needed to confirm this hypothesis.
In search of factors possibly correlated with biofilm formation by clinical S. maltophilia strains, the efficiency in forming biofilm was also stratified both on demographic and clinical data. We revealed a significant association between strong biofilm formation and mechanical ventilation. This is concordant with previous studies indicating that S. maltophilia causes ventilator-acquired pneumonia with a mortality rate of up to 15% [37], probably related to the bacterium's ability to grow as a biofilm on the surfaces of endotracheal tubes and tracheostomy cannula [38,39]. Further studies are warranted to evaluate whether an augmented biofilm production is expected to correlate also with the presence of other types of indwelling medical devices, such as urinary catheters and CVC.
There was no correlation between the degree of biofilm formation and other risk factors evaluated as well as with clinical outcome. This could be due to the relevant values dispersion within each group and/or to the existence of other underlying factors.
The ability of bacterial cells to transfer genes horizontally is generally enhanced within biofilm communities, thereby facilitating the spread of antibiotic resistance [40]. However, in the present study we found an inverse relationship between antibiotic resistance and biofilm formation efficiency, specifically in the case of the pathogenic strains and for some antibiotics. This has already been reported for other bacteria [41,42]. Cepas et al. [41] observed a stronger biofilm formation in P. aeruginosa isolates susceptible to ciprofloxacin than in their resistant counterparts, while acquisition of quinolone resistance led to a decrease in biofilm production in both Salmonella typhimurium and uropathogenic E. coli [27,43]. In another recent study it has also been observed that strong biofilm-forming Acinetobacter baumannii isolates exhibited low-level resistance to gentamicin, minocycline, and ceftazidime [42]. The noteworthy result obtained in our study is that inverse relationship between antibiotic resistance and biofilm formation capacity was specific for pathogenic S. maltophilia strains only.
The pathogenic strains susceptible to colistin, ceftazidime, levofloxacin, and ticarcillinclavulanic acid produced significantly more biofilm than the resistant counterparts did, whereas no significant differences were found among non-pathogenic strains. This was further confirmed by the finding that the biofilm amount formed by S. maltophilia significantly decreases as the number of resistances to antibiotics increased, specifically among the pathogenic strains. Indeed, pathogenic S. maltophilia MDR isolates formed a lower average biofilm amount compared to non-MDR isolates.
It might therefore be speculated that the biofilm-forming S. maltophilia strains rely less frequently on intrinsic antimicrobial resistance for survival. A possible explanation is that isolates able to form abundant biofilms are not as dependent as weakly-or non-producers on the biologically costly expression of planktonic antimicrobial resistance to survive in an environment, such as hospital settings [44,45]. Consequently, S. maltophilia strains with an improved ability to form a biofilm could not have been selected under antibiotic pressure.
The use of trimethoprim-sulfamethoxazole (cotrimoxazole) or levofloxacin resulted in high cure rates against monomicrobial S. maltophilia infections, although a trend toward resistance selection with levofloxacin was observed [19]. We compared the in vitro susceptibility of planktonic and biofilm cells of selected isolates to these two antibiotics by measuring MIC and MBC values, and MBIC and MBEC values, respectively. The inhibitory effect on biofilm formation was higher for levofloxacin than cotrimoxazole (MBIC/MIC values up to 256 and >1024, respectively). This finding confirms the efficacy of fluoroquinolones in affecting S. maltophilia adhesion to polystyrene and, consequently, their clinical relevance for the prevention of biofilm-related infections [12].
In a way similar to other biofilm-producing bacteria, mature S. maltophilia biofilms were markedly more resistant to cotrimoxazole and levofloxacin compared with their planktonic counterparts. The level of resistance to levofloxacin displayed by mature biofilms formed by the isolates recovered from blood was the highest recorded. The bactericidal activity against a preformed biofilm was compromised in a comparable way, as indicated by the MBEC/MBC values up to 128 for both antibiotics. However, it is worth noting that activity of levofloxacin against mature biofilms formed by bloodborne isolates resulted significantly affected compared to cotrimoxazole.
Overall, these data indicate that biofilm-grown S. maltophilia cells express an increased resistance to antimicrobial agents compared with planktonic cells, regardless of the microbiologic and clinical features of the strain (i.e., biofilm formation efficiency, pathogenetic role and source). The increased resistance of bacteria embedded in the biofilm might result in treatment failure and a recurrent infection in patients infected with biofilm-forming isolates. Clinicians should therefore manage infections caused by biofilm-forming isolates very carefully. However, no standard therapeutic guidelines for the management of S. maltophilia biofilm-associated infections are currently available.
The main strength of our study relies on its prospective design, along with the inclusion of a large number of patients from multiple geographically distanced medical centers, which avoid introducing bias or minimizing generalizability. Additionally, the included patients met specific criteria for infection, namely those proposed by the CDC [22], and were one-blind reviewed to limit overstating clinical outcomes.
However, some limitations should be noted. First, we used a simplified laboratory assay for quantifying biofilm formation in vitro, namely the well-described and frequently published MTP method. Although microscopic analysis demonstrated that MTP assay is highly indicative of the biofilm amount formed on polystyrene, there is increasing evidence that in vitro assays do not necessarily accurately represent the in vivo biofilmforming ability of the isolate [46,47]. Assessing biofilm formation on microtiter plates under selected growth conditions may indeed not reflect the circumstances and gene expression during an infection in a living host. Yet, there is no standard laboratory protocol to assess in vivo biofilm formation.
Second, S. maltophilia can be involved in polymicrobial infections, mainly with P. aeruginosa and Staphylococcus aureus [44] which may influence the clinical outcome. During polymicrobial infections, P. aeruginosa elicited significantly higher S. maltophilia load in murine bronchoalveolar lavages and lung tissue [48]. In addition, when grown in mixed biofilm with S. maltophilia, over-expression of alginate by P. aeruginosa might be responsible for the protection of S. maltophilia against antibiotics [3]. According to our data, co-infecting pathogens were present in a minority of cases (<10%). In addition, we considered the patient cured only when all infecting bacteria were eradicated, and only these patients were included in the final analysis.

Conclusions
The highly conserved ability to form biofilm showed by S. maltophilia clinical strains, along with the significant increase in resistance to antibiotics during the planktonic-tobiofilm transition, clearly indicate that the capability for biofilm development may enable S. maltophilia to maintain its ecological niches as commensal microorganisms and that it can be a major virulence factor with important clinical repercussions.
Indeed, this study demonstrated that biofilm formation could be critical in the pathogenesis of S. maltophilia deep infections and revealed several practical implications worthy of future study. First, the strong association between biofilm formation and the site of infection makes testing for a biofilm an important tool in discriminating true bacterial infection from contamination. Second, as high-level biofilm formation seems to be fundamental to invasive S. maltophilia infections, tailoring diagnostic procedures to improve detection and quantification of biofilms should be considered. Third, since the biofilm-embedded cells exhibited a much higher antibiotic resistance than planktonic counterparts, the antibiotic susceptibilities of biofilm-forming S. maltophilia isolates should be interpreted cautiously. Furthermore, the testing of antibiotic susceptibility would conceivably be more accurately performed in a biofilm state.
The lack of correlation of S. maltophilia infection outcome with in vitro biofilm formation might be due to the limitations of the study-design and/or indicate that biofilm formation is one factor that, in combination with others, contributes to the development of human infections. Further studies with larger cohorts of patients and isolates are required to improve our understanding of clinical impact of S. maltophilia's ability to form biofilms.

Supplementary Materials:
The following are available online at https://www.mdpi.com/2076-260 7/9/1/49/s1, Figure S1: Phylogenetic tree of S. maltophilia strains based on atpD MLST profiles; Figure S2: Phylogenetic tree of S. maltophilia strains based on gapA MLST profiles; Figure S3: Phylogenetic tree of S. maltophilia strains based on guaA MLST profiles; Figure S4: Phylogenetic tree of S. maltophilia strains based on mutM MLST profiles; Figure S5: Phylogenetic tree of S. maltophilia strains based on nuoD MLST profiles; Figure S6: Phylogenetic tree of S. maltophilia strains based on ppsA MLST profiles; Figure S7: Phylogenetic tree of S. maltophilia strains based on recA MLST profiles; Figure S8: In vitro susceptibility to antibiotics by S. maltophilia; Figure S9: Biofilm formation by S. maltophilia non-pathogenic strains: stratification on the susceptibility phenotype; Figure   Informed Consent Statement: All subjects gave their informed consent for inclusion before they participated in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy restrictions.