A Temporal Examination of the Planktonic and Biofilm Proteome of Whole Cell Pseudomonas aeruginosa PAO1 Using Quantitative Mass Spectrometry*

Chronic polymicrobial lung infections are the chief complication in patients with cystic fibrosis. The dominant pathogen in late-stage disease is Pseudomonas aeruginosa, which forms recalcitrant, structured communities known as biofilms. Many aspects of biofilm biology are poorly understood; consequently, effective treatment of these infections is limited, and cystic fibrosis remains fatal. Here we combined in-solution protein digestion of triplicate growth-matched samples with a high-performance mass spectrometry platform to provide the most comprehensive proteomic dataset known to date for whole cell P. aeruginosa PAO1 grown in biofilm cultures. Our analysis included protein–protein interaction networks and PseudoCAP functional information for unique and significantly modulated proteins at three different time points. Secondary analysis of a subgroup of proteins using extracted ion currents validated the spectral counting data of 1884 high-confidence proteins. In this paper we demonstrate a greater representation of proteins related to metabolism, DNA stability, and molecular activity in planktonically grown P. aeruginosa PAO1. In addition, several virulence-related proteins were increased during planktonic growth, including multiple proteins encoded by the pyoverdine locus, uncharacterized proteins with sequence similarity to mammalian cell entry protein, and a member of the hemagglutinin family of adhesins, HecA. Conversely, biofilm samples contained an uncharacterized protein with sequence similarity to an adhesion protein with self-association characteristics (AidA). Increased levels of several phenazine biosynthetic proteins, an uncharacterized protein with sequence similarity to a metallo-beta-lactamase, and lower levels of the drug target gyrA support the putative characteristics of in situ P. aeruginosa infections, including competitive fitness and antibiotic resistance. This quantitative whole cell approach advances the existing P. aeruginosa subproteomes and provides a framework for identifying and studying entire pathways critical to biofilm biology in this model pathogenic organism. The identification of novel protein targets could contribute to the development of much needed antimicrobial therapies to treat the chronic infections found in patients with cystic fibrosis.

Pseudomonas aeruginosa is a Gram-negative opportunistic pathogen that is a significant cause of morbidity and mortality in immunocompromised individuals, including those with cystic fibrosis (CF). 1 Late-stage polymicrobial infections in CF patients are dominated by P. aeruginosa (1). This highly adaptable bacterium contributes to excessive pulmonary inflammation, subsequent destruction of lung tissue, and decreased lung function (2)(3)(4). The ubiquity and persistence of this pathogen are due to a combination of virulence and resistance mechanisms that allow it to adapt to environmental changes and colonize the CF host (5). A key component of its adaptation and persistence is the transition from free-living (planktonic) bacteria into organized, surface-attached communities known as biofilms (6). Importantly, biofilms are a barrier to host defenses and reduce the efficacy of antimicrobial therapies (7,8). Attempts to understand the biochemical basis of biofilm development and resilience have consistently shown differences in protein composition in the model organism P. aeruginosa PAO1 (9 -13) (versus its planktonic counterparts). However, technical limitations of these studies have left much of the biofilm proteome unresolved. Accordingly, many aspects of biofilm structure and function are still poorly understood.
In this study, we aimed to establish a quantitative proteomic framework to describe P. aeruginosa cells as they progress through planktonic and biofilm lifestyles. To achieve this, we performed an in-solution digest of triplicate whole cell lysates from growth-condition-matched planktonic and biofilm cultures collected at three different time points. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) was performed using a high-performance quadrupole Orbitrap mass spectrometer (Q Exactive), which resulted in the high-confidence identification of 1884 proteins. We identified proteins unique to each growth mode and those that were differentially regulated and assigned functions according to the P. aeruginosa Community Annotation Project (PseudoCAP) (10). This work advances the growing body of P. aeruginosa planktonic subproteomes (11) and provides new knowledge about the influence of growth mode on the protein composition of both planktonic and biofilm P. aeruginosa over time. A better understanding of the key aspects of P. aeruginosa biofilm biology has the potential to reveal novel targets for antimicrobial therapies and vaccine development for CF patients (12).

EXPERIMENTAL PROCEDURES
Study Design-This study examined the protein content of whole cell samples collected from the biofilm and planktonic growth modes at three different time points (24, 48, and 96 h). Three replicate cultures for each growth mode and time point were inoculated with aliquots of the same overnight P. aeruginosa culture grown to early logarithmic phase. Growth conditions were closely matched for planktonic and biofilm samples, including media composition and volume, vessel material (glass), and aeration (static). Each of the six samples for a single time point (i.e. two growth modes with three biological replicates each) was carefully handled after harvesting to ensure equivalent postincubation processing. An in-solution digestion of the whole cell lysates was performed to remove sample bias introduced by the use of two-dimensional gel-based segregation and isolation (13). Protein contents of the digested lysates were separated and analyzed by means of LC-MS/MS. Bacterial Strains, Media, and Reagents-The parental strain for all studies was P. aeruginosa PAO1. All samples were inoculated with 5 ml of overnight culture normalized to an optical density at 600 nm (A 600 ) of 3.0. Planktonic samples were grown at 37°C in a static culture containing 400 ml of tryptic soy broth (BD, Franklin Lakes, NJ). Following even surface distribution of the inoculum, the biofilm samples were grown at 37°C on 400 ml of tryptic soy agar (BD) solidified in a glass dish (190 ϫ 100 mm; Corning, Tewksbury, MA). At each chosen time point (24,48, and 96 h) biofilm samples were scraped off the surface using a sterile scoopula and dispersed in 400 ml of fresh tryptic soy broth per sample. Triplicate cultures were normalized to an A 600 of 1.0 before further processing. All reagents, unless otherwise stated, were obtained from Sigma-Aldrich Canada Co. (Mississauga, ON, Canada).
Preparation of Whole Cell Samples-Equal volumes of normalized biological replicates were centrifuged in an Avanti J-E centrifuge (12,000 ϫ g, 10 min, 4°C) (Beckman Coulter, Pasadena, CA). The cell pellets were washed once in Tris-HCl (pH 8.3) and then frozen at Ϫ20°C. Cells were lysed using three rounds of liquid nitrogen freeze/ room temperature thaw followed by four 30-s rounds of sonication on ice (setting 3, Ultrasonic Processor XL, Misonix Inc., Farmingdale, NY) with 60 s of cooling time between sonication sessions. Cellular debris was removed by centrifugation in an Avanti J-E centrifuge (6000 ϫ g, 10 min, 4°C) (Beckman Coulter), and the supernatants were collected, treated with a protease inhibitor complex (Roche Diagnostics, Indianapolis, IN), and stored at Ϫ20°C for further processing.
Quantification and In-solution Digestion of Proteins-The concentration of protein in the lysates was measured using a Micro BCA protein assay kit as per the manufacturer's instructions (Thermo Fisher Scientific, Waltham, MA). Protein (15 g) was digested as previously described (14). Briefly, proteins were extracted using a denaturation buffer (6 mM urea/2 M thiourea in 10 mM HEPES, pH 8.0) and incubated at room temperature in reduction (10 mM dithiothreitol in a 50 mM ammonium bicarbonate buffer) and alkylation (55 mM iodoacetamide in 50 mM ammonium bicarbonate) buffers for 30 and 20 min, respectively. The digestion solutions were then treated with 0.3 g of Lys C enzyme per sample for 3 h and incubated overnight with 0.3 g of trypsin (Princeton Separations, Adelphia, NJ). The digestion was stopped by the addition of 40 l of 0.1% trifluoroacetic acid for every 100 l of digestion solution. Finally, the peptides were desalted and concentrated using MonoSpin TM C18 microcolumns according to the manufacturer's instructions (GL Sciences, Torrance, CA). In order to accommodate ambient-temperature shipping, the samples were lyophilized using a speed-vacuum concentrator (Savant Instruments, Holbrook, NY). The samples were reconstituted in 0.1% formic acid in water prior to analysis.
LC-MS/MS-Five microliters of reconstituted, digested protein was injected via online partial-loop into an EASY-Spray ES801 column (75 m ϫ 50 cm) containing PepMap RSLC C18 (2 m) stationary phase (Thermo Fisher Scientific). The sample was separated in reverse-phase mode on an EASY-nLC 1000 chromatography system (Thermo Fisher Scientific) using 0.1% formic acid as the mobile phase. A 120-min run was completed for each sample, including a pre-run equilibration and a post-run wash. Samples were run at 40°C, with an acetonitrile gradient from 0% to 30%, at a rate of 250 nL/min. Eluted peptides were pumped through an EASY-Spray integrated emitter (Thermo Fisher Scientific) for fed nano-electrospray ionization using a Q Exactive mass spectrometer. Fragmentation occurred in a nitrogen-filled higher-energy dissociation collision cell, after which the MS scans were acquired with an Orbitrap mass analyzer. Spectrum and peak list generation were performed using Q Exactive 2.2 and Xcalibur 2.2 (Thermo Fisher Scientific) with the following acquisition parameters: MS resolution 70,000 full width at half-maximum, MS/MS resolution 17,500 full width at half-maximum, target 1 ϫ 10 6 ions, 10 MS/MS scans per cycle, and 15 s dynamic exclusion. A spectral library was created using Proteome Discoverer 1.4 (Thermo Fisher Scientific). Sequest (XCorr Only) version 1.4.0.288 (Thermo Fisher Scientific) and X!Tandem version CYCLONE (2010.12.01.1) (The Global Proteome Machine Organization) were used to analyze the spectra using the UniProtKB (15) P. aeruginosa-ATCC15692 database (5564 entries) and the Pseudomonas Genome Database (10) (5560 entries). The tandem mass spectra were matched to amino acid sequences using 0.02-Da and 10.0-PPM fragment ion mass and parent ion tolerance, respectively, and the following variable modifications: deamidation of asparagine and glutamine, oxidation of methionine, carbamidomethyl alkylation of cysteine (for SEQUEST), Glu 3 pyro-Glu of the N terminus, ammonia loss of the N terminus, gln 3 pyro-Glu of the N terminus, deamidation of asparagine and glutamine, and oxidation of methionine and carbamidomethyl of cysteine (for X!Tandem). The information provided herein is compliant with the Minimum Information about a Proteomics Experiment Mass Spectrometry Informatics guidelines (16).
Initial Data Validation and Analysis-Peptide and protein identifications generated from the amino acid sequences were validated using Scaffold (version 4.0.5, Proteome Software Inc., Portland, OR). Pep-tides and proteins were accepted if they could be established at Ͼ95% probability using the Peptide (17) and Protein (18) Prophet algorithms, respectively, with Scaffold delta mass correction and at least two identified peptides. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. The false detection rate was determined by searching the MS/MS spectra against a nonsense database (i.e. versions of the databases described above with the sequences reversed). Statistical analysis of proteins detected in both planktonic and biofilm samples was completed using the average number of spectral counts (a.s.c.) (calculated from the triplicate samples), which has been shown to accurately reflect the relative abundance of a protein in a sample (19). Proteins with a.s.c. values of Ͻ4 were not included in the analysis because low-abundance proteins can fall outside the linear dynamic range of the relationship described above (19) and therefore were not deemed suitable for statistical testing. Statistical analysis of the semi-quantitative data was completed using an unpaired, two-tailed Student's t test with ␣ set at 0.001 to ensure a high level of significance. Welch's correction was applied to all calculations to add rigor and to account for the majority of the data showing unequal variance (i.e. statistically significant differences in standard deviations). Functional information for the highlighted proteins was gathered from the Pseudomonas Genome Database (10). The basic local alignment search tool (BLAST) (20) was used to search for sequence similarities to hypothetical or unknown proteins using UniProt accession numbers, the nonredundant protein sequences database, and the blastp algorithm. Protein-protein interaction networks were built using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) (version 9.05) with a medium confidence level (0.4) and all available prediction methods (21).
Quantification of Select Proteins Using Extracted Ion Chromatograms-A subgroup of proteins was chosen for manual interpretation and quantitation measurement in order to support the spectral count analysis. Briefly, a spectral library was generated from a representative run of each biological condition. Extracted ion currents (XICs) were measured for a minimum of two representative peptides per sample using Skyline (22) and Pinpoint (Thermo Fisher Scientific) software. The area under the curve (AUC) measurement, which represents the total signal of a peptide (23), was normalized to the AUC of combined XICs from two outer membrane proteins, BamB and OprL. These two proteins were chosen because they are core to the outer membrane structure. The data were tested using a two-tailed, unpaired Student's t test. Variances were assessed with an F-Test, and Welch's correction was applied when the data showed significantly different standard deviations.

RESULTS AND DISCUSSION
The objective of this study was to establish a quantitative proteomic framework to describe the similarities and differences of P. aeruginosa cells progressing through planktonic and biofilm lifestyles. Performing in-solution digestion allowed us to generate highly complex samples that could be analyzed with a high-performance MS platform. As a result, 1884 protein clusters were detected in the whole cell lysates with Ͼ95% probability and a minimum of two identified peptides (supplemental Table S1). This exceeds the results of previous P. aeruginosa PAO1 proteomic studies, including 991 proteins detected in the planktonic inner membrane (24) (with a minimum of one identified peptide), 338 in planktonic outer membrane vesicles (25), 395 in the planktonic periplasm (11), and ϳ700 from total cell extracts of biofilm and planktonically grown whole cells (9). Our false discovery rate was 0.1%, identifying three decoys within the detected proteins, all of which were found in only a single replicate with minimal spectral counts (Ͻ3). For our initial semi-quantitative analysis of the large dataset, we segregated the 163 proteins with average spectral counts of Ն4 that were (i) unique to planktonic growth bacteria (i.e. those that were detected in planktonic samples and absent in biofilm samples) (supplemental Table S2), (ii) unique to biofilm growth bacteria (i.e. those that were detected in biofilm samples and absent in planktonic samples) (supplemental Table S3), or (iii) modulated conserved protein clusters found in both planktonic and biofilm growth bacteria at significantly different levels (supplemental Table S4). The remaining proteins either were present in both growth modes with p values Ͼ 0.001 or had average spectral counts of Ͻ4 at all three time points. Although we did not consider data for proteins with a.s.c. Ͻ 4 for quantitative analysis (representing less than 20% of the dataset), this portion of the data still provides useful qualitative information about the whole cell proteome of PAO1.
Next, protein-protein interaction networks were identified for (i) proteins that were unique or significantly higher in planktonic cultures (Fig. 1) and (ii) proteins that were unique or significantly higher in biofilm cultures (Fig. 2). There were fewer connections between biofilm proteins than planktonic, suggesting a smaller number of biological interactions between the identified proteins in the biofilm cells. Conversely, as the protein networks are based on a combination of genomic context, high-throughput experiments, co-expression, and previous knowledge (21), it could also reflect a generalized lack of biofilm-sourced proteins due to the common practice of studying P. aeruginosa sourced from planktonically grown cultures. The latter scenario reiterates the importance of studies such as this one examining the key differences between planktonic-and biofilm-grown bacteria.
The cultures were grown and standardized to ensure that cell densities were closely matched at all time points. The average cell densities for the biofilm samples were recorded at A 600 (Ϯ S.D.) of 1.85 (Ϯ 0.06), 1.47 (Ϯ 0.15), and 2.4 (Ϯ 0.14) for 24, 48, and 96 h, respectively. Similarly, the average cell densities for the planktonic samples were recorded at A 600 of 1.94 (Ϯ 0.1), 1.2 (Ϯ 0.1), and 1.3 (Ϯ 0.1) for the same three time points. From this, we concluded that we had comparable samples that were representative of two populations of P. aeruginosa that differed only in their growth mode (free-living planktonic versus solid surface biofilm). Negative results obtained from immunoassays using specific anti-RNA polymerase antibodies to detect the presence of RNA polymerase in the cell-free supernatants allowed us to rule out cell lysis as a significant concern in the aged cultures (data not shown).
Proteins Unique to Planktonically Grown P. aeruginosa PAO1 Cells-From the 163 segregated proteins, we found 68 that were exclusively detected in planktonic samples at at least one time point (supplemental Table S2). The greatest number of assigned PseudoCAP functions belonged to the categories of "putative enzymes," "carbon compound metabolism," and "transcription, RNA processing, and degradation" at 48 h and "transport of small molecules" and "adaptation and protection" at 96 h (Fig. 3A). Three proteins were identified that were deemed unique to the planktonic samples at all time points examined: PA3195 (gapA), PA3450 (a probable antioxidant protein), and PA2462. PA2462 is a hypothetical protein found in the outer membrane and outer membrane vesicles (10) that has 98% sequence identity to the adhesin HecA. Interestingly, the PA2462 ion observed in MS increased in a progressive manner from 4.0 to 10.7 to 20.3 a.s.c. at the 24-h, 48-h, and 96-h time points, respectively. HecA mutants of the plant pathogen Erwinia chrysanthemi have reduced surface attachment, aggregate formation, and virulence (26). This gradual increase suggests that in the late-stage planktonic culture, in which the bacteria may be reaching saturation, a phenotypic change might be occurring that favors surface attachment, perhaps as a survival mechanism or as a prelude to pellicle/floc formation. Additionally, six proteins encoded by the pyoverdine locus were all identified as unique to the 96-h planktonic samples (Fig. 1). These proteins included PA2385 (pvdQ), PA2386 (pvdA), PA2397 (pvdE), PA2398 (fpvA), PA2402 (pvdI), and PA2424 (pvdL), suggesting higher levels of iron-scavenging capabilities via increased siderophore biosynthesis and transport (27) in later-stage planktonic cultures. This is consistent with previous studies that demonstrated pvd gene up-regulation in planktonic P. aeruginosa PA14 during growth in CF sputum (28) and mixedspecies biofilms (29). In contrast, pyoverdine deficiency (29) or pyoverdine-deficient subpopulations (30) within single-species P. aeruginosa biofilms have been noted. Moreover, an accumulation of pyoverdine-negative strains has been seen to increase as a function of colonization time in some CF patients (31). As pyoverdine-deficient PAO1 mutants display reduced virulence in multiple animal models of infection (32,33), it is possible that this siderophore is required during the transition from planktonic to biofilm (i.e. early colonization) but becomes less important in mature biofilms (i.e. established infection). This concept is supported by the observed levels in both growth modes studied within; however, further studies are required to determine the relative importance of this and other iron-scavenging compounds in planktonic and biofilm modes of growth.
Other proteins unique to the 96-h planktonic samples include PA0085 (hcp1), whose gene is encoded by the HSI-I virulence locus (34), and PA4689, which has 99% sequence FIG. 1. Protein interactions for unique and significantly increased proteins in planktonic samples. Interactions were determined using STRING 9.05 (21). One hundred and ninety-one interactions were observed among 83 proteins. Lines indicate known or predicted proteinprotein interactions, with thicker lines indicating higher levels of confidence. Legend (10): small red squares, proteins involved in DNA replication, recombination, modification, and repair (hupB, gyrA, uvrB); red dashed ovals, ribosomal proteins (rpsR, rpmD, rplX); small cyan squares, involved in cell envelope biogenesis (murG, PA4457 a.k.a. KdsD); cyan circles, secretion system component (hcp1, xcpQ). Note: gyrA and rplX are known drug targets (10). Other protein groups of interest include those involved in iron scavenging and transport (top left corner), oxidative phosphorylation (*), and sulfate metabolism (#). identity with mammalian cell entry (MCE) protein in P. aeruginosa. The MCE protein family is best studied in the pathogen Mycobacterium tuberculosis. It has been shown to be involved in the invasion of macrophages and other nonphagocytic cells, a key determinant of this pathogen's ability to evade host responses and cause disease (35).
Proteins Unique to P. aeruginosa PAO1 Cells Growing in a Biofilm-Fifty-eight proteins with a.s.c. Ն 4 were exclusively identified at at least one biofilm sample time point (supplemental Table S3). The greatest number of assigned Pseudo-CAP functions belonged to the categories of "putative enzymes," "secreted factors," and "transcriptional regulators" at 24 h, with "transport of small molecules" increased at all three time points (Fig. 3B). Four proteins unique to biofilm samples were detected at all three time points: PA1946 (rbsB, binding the protein component of ABC ribose transporter), PA3236 (a probable glycine betaine-binding protein), PA3923, and PA3922. PA3923 has 77% sequence identity to the adhesion AidA, whereas PA3922 only showed sequence similarity to other hypothetical proteins in a variety of Pseudomonas species. The consistent detection of PA3923 in the biofilm samples might be biologically relevant in light of studies in pathogenic Escherichia coli strains that showed that AidA was able to bind to itself and mediate autoaggregation and biofilm formation (36). Additional studies have shown that this adhesion can be inhibited by purified AidA and sodium deoxycholate (bile salt) (37). High levels of this protein in our sample could promote cell-cell interactions conducive to growth in a biofilm. In addition, because AidA can bind to itself and other related proteins (e.g. Antigen 43) (36), it could potentially play a role in influencing the composition of mixed-species infections, promoting aggregation to other members of the same or different species and thus maximizing virulence potential. Fittingly, an AidA mutant of another CF pathogen, Burkholderia cenocepacia, showed decreased accumulation and virulence in a nematode host (38). The observation that it is sensitive to environmental influences and can be inhibited by purified AidA makes it an attractive target for therapeutic intervention in CF infections. PA1899 (phzA2), a protein involved in the biosynthesis of phenazine, was exclusively found in the biofilm samples at 24 h. A related protein, PA0051 (phzH), a potential phenazinemodifying enzyme (10), had one of the highest a.s.c. values among all of the biofilm samples and was also unique to biofilm at the 24-h time point. Phenazines play a variety of roles in P. aeruginosa virulence, including inhibiting adher-

FIG. 2. Protein interactions for unique and significantly increased proteins in biofilm samples.
Interactions were determined using STRING 9.05 (21). Sixty interactions were observed among 72 proteins. Lines indicate known or predicted protein-protein interactions, with thicker lines indicating higher levels of confidence. Legend (10): large red circle, phenazine biosynthesis (phzM, phzH, phzA2); cyan squares, propanoate metabolism (PA1737, PA0130, PA3568, prpC); small red squares, transport of small molecules (PA4496/7, PA3236, rbsB); red dashed ovals, energy metabolism (exaA, adhC, PA3416/7). Other protein groups of interest include those involved in chemotaxis (*), nitrogen metabolism ( ‡), and tyrosine metabolism (#). ence and biofilm development of a competing organism (Candida albicans) (39), enhancing the release of pro-biofilm-forming extracellular DNA (40), and interfering with host respiratory epithelium via modulation of cellular functions and innate immune responses (for a review see Ref. 41). Interestingly, a study of the redundant operons phz1 and phz2 showed that phz2 was specifically required for phenazine production in P. aeruginosa PA14 biofilms (whereas both contribute to production in planktonic cultures), as well as lung colonization in a mouse model of infection (42), further supporting the influence of growth mode on this pathway. Although not exclusive to biofilm samples, a third protein involved in the phenazine biosynthetic process, PA420 (phzM), which converts phenazine-1-carboxylic acid to pyocyanin (43), was significantly increased at 48 h in biofilm samples. Increased abundance of multiple proteins in this pathway (Fig. 2) at several time points supports earlier work highlighting the importance of phenazines and their derivatives in biofilm biology and P. aeruginosa virulence.
Out of the six sample groups, the quorum sensing regulated protease LasA was detected only in the 24-h biofilm samples. This protein, a putative elastase (44), has been shown to facilitate the invasion of epithelial cells by P. aeruginosa PAO1 in vitro (45). In addition, LasA is able to lyse staphylococci (46,47) and therefore could potentially play a role in modulating the prominence of the CF pathogen Staphlococcus aureus in mixed-species biofilms. The detection of bacteriolytic proteins in the biofilm samples supports the clinical observation that P. aeruginosa is able to outcompete other CF pathogens and eventually become the dominant infecting bacteria in late-stage disease. The inability to detect LasA at the later time points could be related to the observation that some virulence mechanisms, such as drug resistance, can come at the cost of other mechanisms such as invasion (48). Accordingly, the uncharacterized protein PA2915 was detected only in 48-h biofilm samples. Further examination showed that this protein has orthologs in a variety of pathogens including S. aureus, Acinetobacter baumannii, and Legionella pneumophila (49); sequence analysis showed 99% similarity to a ␤-lactamase family protein. Production of ␤-lactamases by P. aeruginosa confers resistance to ␤-lactam antibiotics (50) and is a significant source of recalcitrance in biofilm-dominated chronic lung infections in CF (8).
Proteins Common to Both Planktonic and Biofilm P. aeruginosa PAO1 Cells-Thirty-seven proteins (a.s.c. Ն 4) were detected in both planktonic and biofilm samples at significantly different levels at at least one time point (p Յ 0.001; supplemental Table S4). This includes six proteins that were present at higher levels in biofilm samples at each of the 24-h, 48-h, and 96-h time points than planktonic samples at the same time points. Conversely, 1, 16, and 3 proteins were decreased at the 24-h, 48-h, and 96-h time points, respectively. The assigned PseudoCAP functions for the increased proteins showed an even distribution (Fig. 4A), whereas the assigned PseudoCAP functions for the decreased proteins were highest in the categories of "translation, post-translational modification, and degradation," "amino acid biosynthesis," and "energy metabolism" (Fig. 4B). This supports the general view of biofilms as slow-growing, metabolically lethargic communities. One of the modulated proteins, PA3692 (lptF), an outer membrane protein belonging to the OmpA protein family (10), was significantly increased in biofilm samples at 24 h, present at high levels at 48 h, and absent in the 96-h biofilm samples. A study by Damron et al. found increased levels of LptF in an alginate overproducing mutant (PAO1kinB::aacC1), as well as several CF isolates grown on a solid surface (51). The authors went on to show that this outer membrane lipoprotein (10) contributed to adhesion to lung epithelial cells in vitro (51). Another modulated protein of note is PA3168 (gyrA), DNA gyrase subunit A, which was significantly decreased in the 48-h biofilm samples. This protein is the target of fluoroquinolone antibiotics (52), and consequently mutations are often found in quinolone-resistant strains of P. aeruginosa (53). Studies that have examined the experimental adaptation of P. aeruginosa to conditions that mimic the CF lung have shown increased resistance with exposure to artificial sputum medium plus mucin that was further exacerbated with the addition of an antibiotic; wholegenome sequencing of the adapted pathogens revealed that 4 of the 24 evolved genotypes had gyrA mutations (54). Decreased levels of this protein in the biofilm samples align with its sensitivity to environmental influences, and would correspond with the putative increase in antibiotic resistance seen in bacteria growing as a part of a biofilm.
Quantification of Ion Currents for Select Proteins Supports Spectral Counting Analysis-Manual identification of the ion currents and quantitative measurement of extracted ion chromatograms was completed in order to support the spectral data discussed above. Measurement of XICs more accurately represents the quantity of a peptide in a sample (55). Consequently, we used this method to substantiate the results of spectral counting within our experiment. In general, the AUC measurements of the XICs showed strong concordance with the observations made from the spectral count analysis (Fig.  5). Increased accuracy of the AUC measurements of XIC (relative to spectral counting) allowed us to set ␣ at 0.05 in order to test our data for statistical significance. In doing so, we were able to confirm most of the significant time points identified in the previous analysis of the selected proteins (supplemental Table S4) and add several more. In general, the Pvd family of proteins retained their significant increase in the 96-h time point samples; however, the increased sensitivity of the XIC measurement allowed us to detect the presence of equal amounts of all four of the proteins in both growth modes (i.e. planktonic and biofilm) at the other two time points as well. These observations further support decreased ironscavenging potential in our late-stage biofilm samples. Hcp1 also showed confirmation of decreased levels in biofilm samples at 96 h, an observation that was extended to 48 h with the increased sensitivity of XIC measurement. PA2915 gained significance at 24 and 96 h in the biofilm samples (relative to 48 h only with a.s.c.), and significantly increased levels of the AidA-like protein (PA3923) were confirmed to be significantly higher at all three time points. Furthermore, the more detailed XIC analysis showed an interesting pattern at the 96-h time point. Two of the proteins, PhzA2 and PhzH, were confirmed to be higher in biofilm samples at the 24-h and 48-h time points; however, a more accurate reading of the MS data (via XIC) revealed that this pattern was reversed for both of these proteins at the 96-h time point (versus a.s.c. that were either zero or below threshold; supplemental Table S2). This supports previous observations of increased spectral counts for virulence-related proteins such as the Pvd-family, PA2462 (similar to HecA), hcp1, and PA4689 (similar to MCE) in our late-stage planktonic samples. This secondary evaluation supported our initial spectral counting analysis and added additional information about more subtle changes within the sample groups. Future experiments using complementary methods, such as selected reaction monitoring, have the potential to reveal additional complexities of the biofilm lifestyle.
Concluding Remarks-To the best of our knowledge, this study provides the most in-depth body of work to date ex-amining the influence of growth mode on the protein content of whole cell P. aeruginosa. In addition, it provides a unique view of both planktonic and biofilm proteomes as they transition through various stages of development. In general, the FIG. 5. Secondary validation of MS data using XIC measurement shows strong concordance with the initial analysis. A selection of proteins was manually validated and XICs were measured for further statistical analysis and confirmation of spectral counting data. y-axis values represent the AUC of the XIC. Asterisks indicate statistically significant differences (p Ͻ 0.05). Abbreviations: Bf, biofilm; Pt, planktonic.
profiles support many accepted notions of biofilm biology, including increased antibiotic resistance and decreased molecular and metabolic activity.
In addition to the above observations, this work identified multiple highly prominent hypothetical/unknown proteins that could play a significant role in biofilm structure or development. Further examination of the role of unique biofilm proteins with high sequence similarities to the adhesin AidA and a known metallo-␤-lactamase might provide novel targets for the treatment of refractory P. aeruginosa infections in CF patients. Of equal interest are proteins that were absent from the biofilm sample. The identification of a protein with 99% identity to MCE protein in mature planktonic samples warrants further investigation. In addition to MCE, other virulence proteins were prominent in the 96-h planktonic sample, including multiple proteins encoded by the pyoverdine locus and one protein encoded by the IAHP-related virulence locus; both of these paired with high levels of a protein with 98% sequence identity to the adhesin HecA. Further studies are required in order to determine whether this mature planktonic sample, with high virulence potential and surface-attachment abilities, represents a transitory P. aeruginosa population potentially involved in initial biofilm formation. The identification of two putative proteins with sequence similarity to known adhesions (HecA in planktonic samples and AidA in biofilm samples) reiterates the potential importance of these structures and also warrants further investigation.
The work presented here provides a framework and a reference collection for studying the proteins of entire pathways that may be integral to biofilm biology. This approach will become more relevant with the inevitable increase in multidrug-resistant strains of P. aeruginosa and the subsequent demand for multitarget antibiotic strategies.