Phenotypic and Transcriptomic Analyses of Seven Clinical Stenotrophomonas maltophilia Isolates Identify a Small Set of Shared and Commonly Regulated Genes Involved in the Biofilm Lifestyle

Microorganisms living in a biofilm are much more tolerant to antibiotics and antimicrobial substances than planktonic cells are. Thus, the treatment of infections caused by microorganisms living in biofilms is extremely difficult. Nosocomial infections (among others) caused by S. maltophilia, particularly lung infection among CF patients, have increased in prevalence in recent years. The intrinsic multidrug resistance of S. maltophilia and the increased tolerance to antimicrobial agents of its biofilm cells make the treatment of S. maltophilia infection difficult. The significance of our research is based on understanding the common mechanisms involved in biofilm formation of different S. maltophilia isolates, understanding the diversity of biofilm architectures among strains of this species, and identifying the differently regulated processes in biofilm versus planktonic cells. These results will lay the foundation for the treatment of S. maltophilia biofilms.

among CF patients, have increased in prevalence in recent years. The intrinsic multidrug resistance of S. maltophilia and the increased tolerance to antimicrobial agents of its biofilm cells make the treatment of S. maltophilia infection difficult. The significance of our research is based on understanding the common mechanisms involved in biofilm formation of different S. maltophilia isolates, understanding the diversity of biofilm architectures among strains of this species, and identifying the differently regulated processes in biofilm versus planktonic cells. These results will lay the foundation for the treatment of S. maltophilia biofilms.
KEYWORDS Stenotrophomonas, transcriptome, biofilms I mmunocompromised patients suffer a high risk of nosocomial bacterial infections.
These infections are mostly caused by opportunistic bacterial pathogens. Stenotrophomonas maltophilia is a Gram-negative bacterium and belongs to the Gammaproteobacteria. The role of S. maltophilia as a causative agent of infection remains unclear, and its pathogenicity is not yet fully understood. Nonetheless, S. maltophilia is today recognized as a clinically relevant human opportunistic pathogen that has been associated with a series of infections, such as respiratory tract infection, surgical site infections, peritonitis, endocarditis, bacteremia, and implant-associated infections (1)(2)(3). S. maltophilia together with other major pathogens, such as Staphylococcus aureus, nontuberculous Mycobacterium, Pseudomonas aeruginosa, or Burkholderia cenocepacia, can contribute to severe pulmonary infections in cystic fibrosis (CF) patients (4,5). The prevalence of S. maltophilia in nosocomial infections, particularly lung infections among CF patients, has increased in the last years, and its intrinsic resistance to a broad spectrum of antibiotics makes this organism a relevant multidrug-resistant pathogen in hospitalized patients (1,6,7). S. maltophilia exhibits a high level of intraspecies genomic and phenotypic diversity (8)(9)(10). Analysis of the globally collected genomes of 1,305 isolates from 22 nations indicated that at least 23 phylogenetically distinct lineages can be observed within the S. maltophilia species (11).
S. maltophilia, like most bacterial pathogens, can adhere and form biofilms on host tissues, such as the respiratory tract and skin, as well as on abiotic surfaces, such as medical devices and implants, and it can cause severe infections (12)(13)(14). It is well known that bacterial biofilms provide a shield against antibiotic treatment as well as host defense systems and therefore play a major role in chronic infections (15,16). Recently, a few molecular keys have been affiliated with S. maltophilia biofilm formation. Among those is the quorum sensing system, which is mediated by the diffusible signal factor (DSF) (17,18), cell motility (19), and genes involved in lipopolysaccharide/ exopolysaccharide biosynthesis (20,21). In further studies, iron availability was shown to play an essential role in biofilm formation (22,23).
Within this framework and with respect to the observed genome diversity (11,24), we asked to what extent genomic diversity affects common phenotypes. Focusing mainly on the biofilm properties of S. maltophilia, we assayed biofilm phenotypes and proteolytic and virulence profiles of selected isolates and correlated these findings with deep RNA transcriptomic data.
Our data provide insights into the biofilm architecture and heterogeneity on the species level. Additional transcriptome data of seven clinical isolates identified major genetic loci differentially expressed in S. maltophilia biofilms. Our findings indicate that a rather small number of genes (9.43% Ϯ 1.36%) are involved in the different lifestyles of biofilm cells compared to planktonic cells.

RESULTS AND DISCUSSION
High phenotypic variability at a strain-specific level in S. maltophilia. (i) S. maltophilia biofilm formation capabilities vary strongly at a strain-specific level. Few studies have analyzed the biofilm formation and profiles of S. maltophilia (22,25). To broaden this approach, we have analyzed the capabilities of 300 clinical and environmental isolates from different European countries to form biofilms (see Table S1 in the supplemental material). Of the tested isolates, 261 were of clinical origin and 39 were of environmental origin. All strains were able to adhere and form biofilms on microtiter plates after 24 h of incubation. Based on the microtiter plate biofilm assay, 14.3% of all the investigated isolates formed strong biofilms, 77.0% formed moderate biofilms, and 8.7% formed weak biofilms (Fig. 1). Thus, biofilm formation varied greatly at a strain-specific level. The relative biofilm optical density (OD) values differed by a factor of almost 14. Interestingly, the commonly used model organism, isolate K279a, was just a moderate biofilm former. With a relative biofilm OD value of 0.211, it was categorized as the second-to-last biofilm former within the category of the moderate biofilm formers.
Notably, we did not observe a relationship between biofilm formation capabilities and the origin of isolation. This concerned environmental as well as clinical isolates. Furthermore, no correlation was observed with respect to the phylogenetic position of the isolates within 1 of the 23 recently discovered lineages of the S. maltophilia species (11).
(ii) High variability in biofilm architecture formed by S. maltophilia. Acknowledging the various biofilm forming abilities of individual S. maltophilia strains, we went on to systematically evaluate the architecture of the biofilms of clinical S. maltophilia. Therefore, 40 isolates were chosen from the categories weak, moderate, and strong biofilm formers based on the results of our static biofilm assay in microtiter plates (Fig.  1). This analysis further confirmed the high level of strain-specific heterogeneity. Each strain produced a very distinct biofilm architecture, with some showing clear signs of multicellularity (Fig. 2). The architectures observed were categorized into either flat, patchy, rough, or filamentous biofilms. Of the 40 analyzed strains, 16 isolates formed flat (40%), 8 isolates formed patchy (20%), 14 isolates formed rough (35%), and 2 isolates formed filamentous (5%) biofilms. The observed architectures did not differ significantly when strains were grown in flow cells or in static systems. The biofilms ranged in thickness from 2.9 m (PEG 13-80-49) to 63.6 m (PEG 13-25-38). Using a , and PEG 13-106-64 (K) are highlighted. The last two formed the strongest biofilms. B to D and F to I were included in the transcriptome and other analyses. K279a (E) was included as a control strain. The relative biofilm formation for 6 technical replicates is illustrated. Strains are positioned from low to strong biofilm formers. OD values of relative biofilm formation ranged from 0.064 to 0.88. Isolates with a relative biofilm OD of Յ0.2 were classified as weak biofilm formers (green), and all isolates with a relative biofilm OD of Ն0.5 were classified as strong biofilm formers (red). All isolates with a relative biofilm OD between 0.2 and 0.5 were classified as moderate biofilm formers (orange). Standard deviations ranged from 0.003 to 0.098. Strains employed in transcriptome and virulence analyses are indicated with asterisks. Coordinates of all isolates together with their metadata are listed in Table S1. LIVE/DEAD stain, the biofilm-grown cells appeared to be viable in most cases. However, a large number of cells of the strains PEG 13-101-78 and PEG 13-64-34 appeared to be dead ( Fig. 2A). No relationship between the biofilm architecture of the individual strains and their origin of isolation, their phylogenetic position within the species, or their classification as strong, moderate, or weak biofilm formers was detected. Remarkably, isolate 454 formed dense clusters of cells that appeared to have a rosette-like structure ( Fig. 2B and C).
Until now, only a limited number of studies have shown diversity in biofilm architectures among strains of the same bacterial species. A study on the biofilm architectures of 96 Listeria monocytogenes isolates showed variations in biofilm architectures, with a particular predominant morphotype, the honeycomb-like morphotype. However, just like in our observations, there was no correlation between the biofilm architectures and the genetic lineages of the different isolates (26,27). Furthermore, Hornischer et al. generated a reference database for Pseudomonas aeruginosa and reported on large strain-specific variations with respect to resistance, virulence, and biofilm architecture (https://bactome.helmholtz-hzi.de) (28).
Proteolytic and virulence profiles differ largely at a strain-specific level. Knowing that extracellular proteases are associated with virulence of pathogenic bacteria, including S. maltophilia (29)(30)(31), and that extracellular proteolytic activity has been associated with biofilm formation in some bacterial species (32), we further asked to what extent the extracellular proteolytic activities of the different isolates differ at a strain-specific level. Therefore, we profiled the extracellular proteolytic activities of several S. maltophilia strains. As assumed, the different strains displayed different levels of protease activity in planktonic as well as in biofilm cultures. Surprisingly, extracellular protease activity was up to 40 times higher in biofilm cultures than in planktonic cultures (Fig. 3A). However, no correlation between biofilm forming capabilities and the phylogenetic origin within the species was observed.
In the light of the above-described observations, we also asked to what extent S. maltophilia isolates differ in their strain-specific virulence patterns. Therefore, we investigated the virulence of seven isolates using the Galleria mellonella model. The 7 isolates were chosen based on their different biofilm architectures, ranging from flat to patchy and rough with filaments. In these tests, we showed that an infection dose of 1 ϫ 10 5 CFU/larva of the different strains caused a different mortality outcome in G. mellonella larvae. Strain 454 appeared to be most virulent, with only 37.5% of larvae surviving after a 72-h incubation period (Fig. 3B). In contrast, strain 677 showed the least virulence, with 100% survival of larvae after 72 h. The high heterogeneity in virulence of S. maltophilia isolates demonstrated in this study has also been shown in other studies (33,34). We did not observe a correlation between biofilm profiles, biofilm architectures of these isolates, and their degrees of virulence using the G. mellonella model. However, since we analyzed only seven isolates with respect to their virulence in the G. mellonella model, we cannot exclude that a larger analysis will show that other strains might be more virulent.
Global transcriptome sequencing (RNA-seq) indicates a set of 106 shared highly expressed genes in S. maltophilia biofilms and not more than 33 strainspecific genes. To obtain a comprehensive overview of all genes that could possibly play an important role in mature S. maltophilia biofilms, we analyzed the transcriptomes of the same seven clinical S. maltophilia isolates as employed in the Galleria mellonella model (Table 1 and Table S1). The isolates were chosen because they differed in their biofilm architectures (Fig. 2). We first analyzed the transcriptomes of 72-h-old biofilms to identify the most relevant biofilm-expressed genes. We then compared these data sets with transcriptomes of planktonic cultures of the same isolates. Using these comprehensive data sets, we asked three main questions. First, which are the most strongly expressed genes in mature biofilms in general, and what are the main metabolic routes in these biofilms? Second, which are the strain-specifically expressed genes that contribute to the highly heterogeneous biofilm architectures? Third, which are the regulated but commonly expressed genes under biofilm versus planktonic conditions?
(i) S. maltophilia strongly expressed genes in biofilms give insight into metabolism and life on surfaces. In general, each strain transcribed between 4,000 and 4,350 genes under biofilm conditions (Table S3). It is noteworthy that the genome sizes of the seven isolates varied slightly (Table 1). Using a nucleotide activities per kilobase of exon model per million mapped reads (NPKM) cutoff of 10, in general, 80.5% of all genes were transcribed to some extent under the biofilm conditions (Table S3).
To identify the most relevant and strongly expressed genes in biofilm-grown cells, we decided to focus on the top 250 expressed genes (excluding ribosomal proteins) in our analyses (Table S4). This arbitrarily chosen cutoff correlated in all strains with the expression level of the rpoD gene, which is used in many studies as a reference gene (35,36). This initial analysis included all strongly expressed genes, independent of whether these were regulated or constitutively expressed in biofilm versus planktonic cells. A Venn analysis revealed that 42% (106 genes) of the top 250 biofilm-expressed genes were commonly expressed among all isolates (Tables 2 and 3). Thus, the shared A more detailed functional analyses indicated that among the 106 shared genes, the largest fractions were assigned to metabolism (20.75%), to membrane proteins and transporters (19.81%), and to transcription and translation (15.09%) ( Fig. 4A and Table 2).
Surprisingly, all strains expressed a gene coding for the alcohol dehydrogenase AdhP at a high level, up to 38 times higher than the expression level of the housekeeping gene rpoD (Fig. 4B). In Escherichia coli, AdhP is probably involved in the production of propanol, or similar short-chain alcohols, under biofilm conditions (37). Since this enzyme belongs to the alcohol dehydrogenase superfamily, we speculate that it is further involved in the production of various short-chain alcohols in S. maltophilia. In this respect, it had not been reported that S. maltophilia produces primary or secondary alcohols. Therefore, we analyzed the supernatants of biofilm and planktonic cultures using 1 H nuclear magnetic resonance (NMR). The NMR data indicated that biofilm-grown cells produced micromolar amounts of 1-and 2-propanol as well as acetate together with several unknown compounds (Table 4 and Fig. S1). These short-chain alcohols were not observed in any of the planktonic cultures. The occurrence of the different alcohols, however, varied at a strain-specific level.
Furthermore, it is noteworthy that the genes coding for a Pal protein together with the YgbF protein were strongly transcribed in all seven characterized strains. The Pal protein is part of the Tol-Pal complex and is involved in cell envelope-related processes. It is involved in transport processes but also in release of outer membrane vesicles and cell septum formation (38) and has been associated with survival and pathogenesis in bacteria (39).
Among the other most relevant genes, many are involved in iron uptake, such as the gene coding for the biopolymer transporter ExbD. Moreover, all strains expressed genes of the ATP-dependent Clp protease complex, which is commonly involved in proteolysis and regulation of different metabolic processes. Interestingly, some studies have shown that the Clp protease complex also plays a role in biofilm formation and virulence in Listeria monocytogenes and Actinobacillus pleuropneumoniae (40,41).
The gene encoding the DNA transport competence protein ComEA was strongly expressed in all seven isolates. ComEA is essential for DNA transport and bacterial competence (42). We have recently identified the comEA locus as potentially involved in phenotypic heterogeneity in S. maltophilia (10). The observation that comEA is highly expressed may suggest a broader role in S. maltophilia biofilm formation (43).
(ii) Genes strongly expressed at a strain-specific level give the first clues on heterogeneous phenotypes. Altogether, 142 of the 250 most strongly expressed genes seemed to be strain-specifically expressed in only one of the seven analyzed  isolates. While strain PC240 expressed only five genes at a strain-specific level, 454, the most virulent strain (Fig. 3B), expressed 33 genes at a strain-specific level (Table 3). While we initially hoped that these data would give us clues about the different biofilm architectures, the data did not allow for a final conclusion on these highly diverse biofilm phenotypes. Nevertheless, some of the observed differential expression profiles may in part provide the first clues about the high strain-specific variability.
For instance, SKK55 biofilms and some others consisted of a remarkable number of multicellular structures, such as long filamentous cells (Fig. 2). The zapA gene was one of 30 strain-specific genes of SKK55 (Table 3), which is involved in cell division (44)(45)(46). The elevated expression of the mraZ gene has been linked to the occurrence of filamentous cells in E. coli and Mollicutes (47,48). The observation made in this study may explain the frequent occurrence of filamentous cells in S. maltophilia SKK55 biofilms.
(iii) Relatively few genes may play a role in matured biofilms of S. maltophilia. In addition to comparing the transcriptomes of the seven S. maltophilia isolates under biofilm conditions, we also compared biofilm and planktonic cells for each isolate (Fig.  S2). Generally, under planktonic conditions, 96.9% of all genes were transcribed (NPKM value Ͼ 0) (Table S5) and 3.0% were not transcribed in all seven strains. Remarkably, the  majority of the top 250 strongly expressed genes in biofilms were equally highly expressed in planktonic cultures. These were often essential genes involved in general and aerobic metabolism (Tables S4 and S6). Comparing the biofilm-expressed genes with the planktonic expressed genes in detail, we noted that just 9.43% Ϯ 1.36% of all genes were differentially regulated, consisting of 7.46% Ϯ 1.49% upregulated genes and 1.96% Ϯ 1.02% downregulated genes using log 2 fold change cutoffs of 2 and Ϫ2 and a P adjusted (P adj ) value of Յ0.05 ( Fig. 5 and Table 1). In general, between 240 and 421 genes were upregulated and between 34 and 154 genes were downregulated when we compared the planktonic with the biofilm cells (Table 1).
A pangenome analysis of all seven isolates identified the gene clusters present in the core genome (Fig. 6). The integration of the log 2 fold change values of all up-and downregulated genes into the pangenome analysis revealed that 68% Ϯ 8% of all regulated genes belong to the core genome. A Venn analysis of all gene clusters of the regulated genes generated by the pangenome analysis identified 52 genes that were commonly upregulated and 1 gene (5S rRNA) that was commonly downregulated in all isolates ( Table 5). The largest fraction of the commonly upregulated genes coded for hypothetical and uncharacterized proteins (30.8%), followed by genes involved in transcription and translation (17.3%) (Fig. 7). A remarkably high number of genes involved in iron acquisition (11.5%) were found to be upregulated in almost all seven analyzed S. maltophilia isolates ( Fig. 7 and Table S7). Most interestingly, a hypothetical gene which is the corresponding homologue of smlt2713 in strain K279a was the most strongly and differentially expressed gene in all strains, with a log 2 fold change of 10 to 12 (Fig. 6, region A, and Table S7). It carries a TbpB (transferrin-binding-protein) motif, implying that it is possibly involved in iron uptake. Notably, Tbp proteins have been associated with virulence in Actinobacillus pleuropneumoniae (49,50). It is known that iron plays a role in biofilms of Bacillus subtilis (51) and S. maltophilia (22) as well as boosts biofilm formation of Staphylococcus aureus (52), P. aeruginosa (53), and Campylobacter jejuni (54). Further, a sulfite reductase gene was also upregulated in biofilm cells of all seven isolates, implying that the ability to assimilate sulfate and sulfite for energy conservation appears to be present in this organism. Furthermore, genes involved in nitrate transport and reduction were upregulated in four of the seven tested isolates (454, 677, ICU331, and PEG 13-68-68 [ Fig. 6, region B]). This suggests an anaerobic mode of respiration in biofilm cells.

1-Propanol 2-Propanol Threonine Acetate Ethanol
Representative spectra of the different compounds in the supernatants are given in Fig. S1. X, detected/ verified; -, not detected/verified. Anaerobic respiration in biofilms has also been reported for P. aeruginosa (55,56). Surprisingly, many motility affiliated genes were transcribed at elevated levels (Table  S7). Domka and colleagues also demonstrated that the expression of some motilityrelated genes was increased in 7-to 24-h-old biofilms of Escherichia coli (57). Further studies have reported the involvement of flagella and pili in shaping the biofilm structures of P. aeruginosa and E. coli (58)(59)(60)(61). Thus, our findings may imply a similar function in S. maltophilia. Furthermore, all isolates carry three copies of the ax21 genes, whereas for the reference strain K279a the presence of just two copies is known (smtl0387 and smtl0187). While two copies appeared to be strongly transcribed in biofilm cells of almost all seven isolates, a comparison with the planktonic cells implied that they are differentially transcribed in a strain-specific manner. In four of the seven tested isolates, the transcription of one or both copies of the ax21 gene was upregulated in biofilm cells (Table S7). The expression of the ax21 copies seems to be dependent on quorum sensing via the DSF in S. maltophilia (62). Nonetheless, its direct role in the biofilm formation by S. maltophilia remains unclear.
Notably, all analyzed S. maltophilia isolates code for at least four different cytochrome c oxidases. Under hypoxic conditions in biofilms, some of these cytochrome c oxidases were upregulated in some of the 7 isolates (Table S7), and a common cytochrome b oxidase was commonly upregulated in all isolates (Table 5).
Overall, with a proportion of 1.96% Ϯ 1.02%, relatively few genes were downregulated in all seven investigated isolates. Generally, tRNAs and 5S and 16S rRNAs were downregulated in almost all analyzed isolates (Table S7). This indicates a reduced translation in biofilm cells in comparison to planktonic cells, which was already demonstrated, for instance, for Staphylococcus epidermidis, Clostridium perfringens, and Gardnerella vaginalis (63)(64)(65).
Furthermore, some genes of a not-yet-characterized operon stretching from NIPOL-PBK_01784 to NIPOLPBK_01792 (using the 454 nomenclature) were downregulated in all isolates. This operon codes for three hypothetical proteins, one membrane protein, one putative protease, two TldD proteins, and one MoxR-like-ATPase. TldD family proteins inhibit the DNA gyrase (66), and the MoxR-like-ATPase has chaperone-like functions (67). Moreover, a lysoplasmalogenase gene (NIPOLPBK_00344 in 454) was downregulated in four of the seven isolates. These enzymes form fatty aldehyde and glycerophosphoethanolamine or glycerophosphocholine by cleaving the vinyl ether bond of lysoplasmalogen (68,69). These findings may imply a novel role of fatty acid and glycerol derivatives in the planktonic lifestyle.
Finally, on average, 1.16% Ϯ 0.78% of all genes were strain-specifically upregulated and 0.56% Ϯ 0.37% downregulated in the seven studied isolates. The majority of these genes are affiliated with metabolism (Table S8).
Conclusions. Our study analyzed in detail a large number of S. maltophilia isolates with respect to biofilm forming capabilities, architecture, and metabolite production in biofilms, together with protease profiling and estimation of virulence in the G. mellonella model. Together, our data imply that within isolates of the species S. maltophilia Locus tags were derived from the 454 genome entry CP060027 and homologues retrieved from the corresponding genomes of the other isolates (see Table 1 for GenBank accession numbers). Genes without significant regulation (P adj Ͼ 0.05 [gray]), genes with P adj of Ͻ0.05 (blue), and significantly up-or downregulated genes (P adj Ͻ 0.05; log 2 fold change of greater than 2 or less than Ϫ2 [red]) are illustrated. For isolate 454, 338 genes were upregulated of a total gene count of 4,269 genes, while 67 genes were downregulated. Of a total of 4,716 genes, 398 genes were upregulated and 37 genes were downregulated in ICU331. For SKK55, 284 genes were upregulated and 34 genes were downregulated of a total of 4,296 genes. In 677, 421 of 4,433 genes were upregulated, while 116 were downregulated. For PC239, 240 genes were upregulated and 148 genes were downregulated of a total of 4,564 genes. Of a total of 4,564 genes, 262 genes were upregulated and 153 genes were downregulated in PC240. For PEG 13-68-68, 359 genes were upregulated and 57 genes were downregulated of a total of 4,094 genes. Normalized gene read counts were used to build volcano plots.
an unexpected high phenotypic diversity exists. While strain specificity is a major challenge in S. maltophilia clinical research, we nevertheless have identified a set of 106 commonly and strongly expressed genes in biofilms and a maximum of 33 strainspecific genes. Further, our data imply that, on average, 7.46% Ϯ 1.49% of all genes were upregulated in biofilms versus planktonic cells and, on average, 1.96% Ϯ 1.02% of all genes were downregulated.
The phenotypic and omics data generated in this study will provide a solid basis for further S. maltophilia biofilm studies and will help to reveal potential targets for the development of more effective drugs against this emerging pathogen.

MATERIALS AND METHODS
Bacterial strains, chemicals, and growth conditions. Table S1 summarizes the S. maltophilia clinical and environmental isolates used in this study together with their metadata.
S. maltophilia strains were routinely cultured in LB medium (10 g/liter of tryptone, 5 g/liter of yeast extract, and 5 g/liter of NaCl) at 28°C or 37°C if not otherwise stated.
S. maltophilia biofilm assays. (i) Static biofilm assay in microtiter plates. For analyses of static biofilms, S. maltophilia cells were grown in flat-bottomed microtiter plates (Nunc MicroWell, catalog no. 161093; Thermo Fisher Scientific, Waltham, MA) according to the method of Steinmann et al., with the following modifications (24). An overnight culture of S. maltophilia was adjusted to 4.0 ϫ 10 7 cells/ml in LB medium. A total of 200 l of the culture was pipetted in microtiter plates and incubated without shaking at 37°C for 24 h. After drying the biofilm, the adhered cells were stained with 50 l of 0.5% crystal violet solution per well for 5 min and were washed three times with water afterwards. The plates were dried for 30 min at 37°C before crystal violet was dissolved in 150 l of 33% acetic acid per well and the optical density at 595 nm (OD 595 ) of the dye was measured. Six technical replicates were done per strain or condition. To obtain the relative biofilm OD value, we first measured the total OD (growth) value of each well in our microtiter plate assay and then set the crystal violet value in ratio to this, that is, value OD 595 crystal violet/OD 600 growth ϭ relative biofilm OD value. Isolates with a relative biofilm OD of Յ0.2 were classified as weak biofilm formers, and isolates with a relative biofilm OD of Ն0.5 were classified as strong biofilm formers. All isolates between an OD of 0.2 and 0.5 were classified as moderate biofilm formers.
(ii) Cultivation of S. maltophilia biofilms in flow chambers or -slides. For analyses of the biofilm architecture, S. maltophilia isolates were cultivated in three-channel flow chambers (70) or eight-well -slide (ibiTreat, catalog no. 80826; ibidi USA Inc., Fitchburg, WI). All experiments were performed at 28°C with 10% LB medium. After 72 h, cells were stained using the LIVE/DEAD BacLight bacterial viability kit (Thermo Fisher Scientific, Waltham, MA) and the biofilm was analyzed by confocal microscopy.
Fluorescence imaging analysis of S. maltophilia biofilms. Visualization of flow chamber and -slide biofilms was performed using a confocal laser scanning microscope (CLSM) Axio Observer.Z1/7 LSM 800 with airyscan (Carl Zeiss Microscopy GmbH, Jena, Germany) and a C-Apochromat 63ϫ/1.20W Korr UV VisIR objective. The microscope settings for the different fluorescent dyes are shown in Table S2. The analysis of the CLSM images and three-dimensional reconstructions were done with ZEN software (version 2.3; Carl Zeiss Microscopy GmbH). Biofilm architecture was analyzed at least at three different positions for each strain, and one representative CLSM image was chosen. More detailed quantitative analyses of biofilm architecture, such as roughness and thickness, were done using BiofilmQ software version 0.1.4. (71) (https://drescherlab.org/data/biofilmQ/docs/usage/installation.html).
Galleria mellonella virulence assay. The virulence of seven S. maltophilia isolates was tested using the Galleria mellonella model (72,73). Final-instar larvae of G. mellonella were obtained from Biosystems Technology, Exeter, UK. Healthy larvae were identified by their cream color without dark discoloration and their vital movements. Prior to infection, the S. maltophilia isolates were grown in LB medium at 37°C to an OD 600 of 0.5. Cells were washed twice in phosphate-buffered saline (PBS) and adjusted to 1 ϫ 10 7 /ml. For each isolate, an inoculum of 1 ϫ 10 5 CFU/larva was then injected into the last proleg of the larvae. For each experiment, a total of 15 larvae were used for each isolate. A total of 15 larvae injected with 10 l of PBS each served as a control. The larvae were then placed on Whatman paper-lined petri dishes and incubated at 37°C. The larvae were observed at 24, 48, and 72 h. Larvae were considered dead when no movement and a complete dark discoloration were observed.
Assessment of the extracellular proteolytic activity of biofilm and planktonic cultures. For assessing the protease activity, we used the method as previously published by Steinmann et al., with minor modifications (24). For planktonic cultures, 30 ml of 10% LB (DWK Life Sciences GmbH, Westfalen, Germany) was inoculated with overnight cultures and placed to shake at 28°C to an OD 600 of 0.5.  (Table 5) were identified by a Venn analysis and classified by the function of their gene products.
Ten-milliliter aliquots of the cultures were pelleted at 4°C for 20 min. The supernatant was collected and sterile filtered (CA membrane, 0.2 m). For biofilm cultures, overnight cultures were adjusted to an OD 600 of 0.05 in 10% LB medium. The biofilm of four biological replicates was grown in 24-well microtiter plates (Nunc cell culture plate, catalog no. 142475; Thermo Fisher Scientific, Waltham, MA) at 28°C for 48 h. The supernatant was collected, centrifuged as described above, and sterile filtered. The protease activity was determined using the Molecular Probes EnzChek protease assay kit (Thermo Fisher Scientific). The protease activity of three technical replicates was measured in black BRANDplates microtiter plates with transparent bottoms (BRAND GmbH ϩ Co. KG, Wertheim, Germany). A total of 100 l of substrate, 90 l of digestion buffer, and 10 l of culture supernatant were pipetted into each well. Plates were incubated at 37°C and fluorescence was measured every 30 min in a Synergy HT plate reader (BioTek Instruments Inc., Winooski, VT) at 590/20-nm extinction and 645/40-nm emission wavelengths.
RNA-seq and data analysis. RNA-seq was done as previously published, with minor modifications (74). For the preparation of cell material for RNA-seq analyses, biofilm cells were grown in flow cells as described above at 28°C for 72 h using 10% LB medium. To avoid contamination, media were supplemented with 50 g/ml of ampicillin. Biofilms were grown in flow cells for 72 h and were washed out of the flow cells with a 20% stop mix (95% ethanol and 5% phenol) and pelleted at 4°C for 20 min. Planktonic cells were grown in 10% LB medium supplemented with 50 g/ml of ampicillin at 28°C and were pelleted at the exponential phase (OD 600 ϳ 0.5) at 4°C for 20 min. The pellets were frozen in liquid nitrogen for later analysis. Three biological replicates of each strain were prepared.
Harvested biofilm/planktonic cells were resuspended in 800 l of RLT buffer from the RNeasy minikit (Qiagen, Hilden, Germany) with ␤-mercaptoethanol (10 l/ml), and cell lysis was performed using a laboratory ball mill. Subsequently, 400 l of RLT buffer (RNeasy minikit) with ␤-mercaptoethanol (10 l/ ml) and 1,200 l of 96% [vol/vol] ethanol were added. For RNA isolation, the RNeasy minikit was used as recommended by the manufacturer, but instead of RW1 buffer, RWT buffer (Qiagen, Hilden, Germany) was used to also isolate RNAs smaller than 200 nucleotides (nt). To determine the RNA integrity number (RIN), the isolated RNA was run on an Agilent Bioanalyzer 2100 using an Agilent RNA 6000 Nano kit as recommended by the manufacturer (Agilent Technologies, Waldbronn, Germany). Remaining genomic DNA was removed by treating the samples with Turbo DNase (Thermo Fisher Scientific, Waltham, MA). Pan-Prokaryozes riboPOOL kit v1 (siTOOLS Biotech, Planegg/Martinsried, Germany) was used to reduce the amount of rRNA-derived sequences. For sequencing, the strand-specific cDNA libraries were constructed with a NEBNext Ultra II directional RNA library preparation kit for Illumina (New England BioLabs, Frankfurt am Main, Germany) using 50 ng of rRNA-depleted RNA and 8 PCR cycles. To assess the quality and size of the libraries, samples were run on an Agilent Bioanalyzer 2100 using an Agilent highsensitivity DNA kit as recommended by the manufacturer (Agilent Technologies, Waldbronn, Germany). The concentrations of the libraries were determined using the Qubit double-stranded DNA (dsDNA) HS assay kit as recommended by the manufacturer (Life Technologies GmbH, Darmstadt, Germany). Sequencing was performed by using the HiSeq4000 or HiSeq2500 instrument (Illumina Inc., San Diego, CA) using either the HiSeq 3000/4000 SR cluster kit or the HiSeq Rapid PE cluster kit v2 for cluster generation and the HiSeq 3000/4000 SBS kit (50 cycles) for sequencing in the single-end mode and running 1 ϫ 50 cycles. For quality filtering and removal of remaining adaptor sequences, Trimmomatic v-0.39 (75) and a cutoff phred-33 score of 15 were used. The mapping of the remaining sequences was performed with the Bowtie program (version 2) (76) using the implemented end-to-end mode, which requires that the entire reads align from one end to the other. Surviving reads were mapped against the reference genomes. NCBI GenBank numbers are listed at the end of the Materials and Methods section and in the footnote to Table 1. Reads per gene feature were counted using feature Counts v.2.0.0 (77). Subsequent analysis was performed in R (version 3.6.1) (78), including normalization of the reads with DeSeq2 (version 1.26.0) (79) using the default fold change-shrinkage algorithm. The rlog function with "blind" parameter set to TRUE within DeSeq2 was used for data transformation to generate principal-component analyses (PCAs), which were drawn with ggplot2 (version 3.2.1) (80), whereas the Enhanced Volcano packages (version 1.4.0) (81) were used for the creation of volcano plots.
Venn analysis. Proteins with an identity of Ͼ 90% were combined in one cluster. Venn Painter tool (version 1.2.0) (82) was used to construct the Venn diagram and to identify the commonly expressed or regulated genes between the analyzed isolates.
Pangenome analysis. Pangenome analysis and the illustration of the corresponding differential gene expression were conducted with Anvi'o 6.1 (83). Gene clusters were built with a minbit of 0.5. For the differential gene expression, gene clusters with a log 2 fold change of greater than 2 were illustrated as upregulated and gene clusters with a log 2 fold change of less than Ϫ2 were illustrated as downregulated.
Determination of metabolites via 1 H NMR analyses. For NMR analysis, the supernatant of biofilm cultures grown in flow cells for 72 h in 10% LB medium at 28°C was collected. Samples were mixed with 40 mM phosphate buffer (K 2 HPO 4 and KH 2 PO 4 in H 2 O) in a ratio of 1:9 (sample to buffer). A standard (trimethylsilylpropanoic acid [TMSP], 5 mM) was also added. The 1 H NMR samples were measured with an excitation sculpting sequence for water suppression with a 600-MHz spectrometer with 128 scans at 300 K. A total of 65,536 data points and a spectral width of 16.0221 ppm was acquired with the O1 at the signal of water at 2,819 Hz. All spectra were processed with Topspin version 4.0.8, applying an exponential multiplication of the free induction decay (FID) with a line broadening factor of 0.3 Hz.
Data availability. The raw reads of the 42 mRNA sequencing runs have been deposited in the SRA nucleotide archive under BioProject no. PRJNA646397 and correspond to the SRA accession numbers SRX8752168 to SRX8752209. Updated GenBank files of the genomes of the seven isolates included in the transcriptome analyses were submitted under the following identifiers (IDs) and accession numbers: