NMR Metabolomics of Planktonic and Biofilm Modes of Growth in

Oct 5, 2007 - Bo Zhang , Robert Powers ... Fabienne Faÿ , Isabelle Linossier , David Carteau , Alexandra Dheilly , Alla Silkina , Karine Vallée-Réh...
0 downloads 0 Views 502KB Size
Download PDF
Anal. Chem. 2007, 79, 8037-8045

NMR Metabolomics of Planktonic and Biofilm Modes of Growth in Pseudomonas aeruginosa Erica L. Gjersing,† Julie L. Herberg,* Joanne Horn,‡ Charlene M. Schaldach, and Robert S. Maxwell

Lawrence Livermore National Laboratory, 7000 East Avenue, Livermore, California 94550

Bacteria often reside in communities where the cells have secreted sticky, polymeric compounds that allow them to attach to surfaces. This sessile lifestyle, referred to as a biofilm, affords the cells within these communities a tolerance of antibiotics and antimicrobial treatments. Biofilms of the bacterium Pseudomonas aeruginosa have been implicated in cystic fibrosis and are capable of colonizing medical implant devices, such as heart valves and catheters, where treatment of the infection often requires the removal of the infected device. This mode of growth is in stark contrast to planktonic, free floating cells, which are more easily eradicated with antibiotics. The mechanisms contributing to a biofilm’s tenacity and a planktonic cell’s susceptibility are just beginning to be explored. In this study, we have used a metabolomic approach employing nuclear magnetic resonance (NMR) techniques to study the metabolic distinctions between these two modes of growth in P. aeruginosa. Onedimensional 1H NMR spectra of fresh growth medium were compared with spent medium supernatants from batch and chemostat planktonic and biofilms generated in continual flow system culture. In addition, 1H highresolution magic angle spinning NMR techniques were employed to collect 1H NMR spectra of the corresponding cells. Principal component analysis and spectral comparisons revealed that the overall metabolism of planktonic and biofilm modes of growth appeared similar for the spent media, while the planktonic and biofilm cells displayed marked differences. To determine the robustness of this technique, we prepared cell samples under slightly different preparation methods. Both techniques showed similar results. These feasibility studies show that there exist chemical differences between planktonic and biofilm cells; however, in order to identify these metabolomic differences, more extensive studies would have to be performed, including 1H-1H total correlated spectroscopy. Bacteria, such as Pseudomonas aeruginosa, often reside in communities called biofilms that are attached to surfaces and * To whom correspondence should be addressed. E-mail: [email protected]. Phone: (925) 422-5900. Fax: 925-422-3160. † Current address: The Department of Chemical Engineering, University of California Davis, Davis, CA 95616. ‡ Current address: Microchip Biotechnologies Inc., 6693 Sierra La., Suite F, Dublin, CA 94568. 10.1021/ac070800t CCC: $37.00 Published on Web 10/05/2007

© 2007 American Chemical Society

encase themselves in a polysaccharide matrix. P. aeruginosa biofilms have been implicated in cystic fibrosis and nosocomial infections where they are of major concern due to their tolerance of antibiotics.1,2 An understanding of the mechanisms involved in a biofilm’s ability to withstand an antimicrobial agent is needed to develop new methods of treatment.3 Comparisons of biofilm cells to their planktonic counterparts have already begun to elucidate the many factors involved in a biofilm’s ability to withstand antimicrobials. So far, significant proteomic differences have been discovered for Pseudomonas putida, when this bacteria is grown in suspended cultures versus a sessile mode,4 and throughout the development of a P. aeruginosa biofilm.5 In addition, changes in expression have been observed in genes involvedinmotility,quorumsensing,andpolysaccharideproduction.6-8 The metabolome, defined as the inventory of all the metabolites in a biological system,9 is akin to the proteome and genome. Genetic regulation causes changes in protein expression, which should at least partly be reflected in the complement of cellular metabolites.10 Metabolome data are useful for a variety of applications, all with the ultimate goal of understanding cause and effect processes within biological systems. As a functional genomics tool, metabolite concentrations have been used to expose phenotypes in yeast mutants11 and plants12 that, based on growth rates and fluxes, otherwise appeared to be identical. Metabolome analysis has also proven useful in measuring metabolite responses to stress such as disease in abalone,13 and growth rate14 and culture (1) Stewart, P. S.; Costerton, J. W. Lancet 2001, 358, 135-138. (2) Costerton, J. W.; Stewart, P. S.; Greenburg, P. E. Science 1999, 284, 13181322. (3) Davies, D. Nat. Rev. Drug Discovery 2003, 2, 114-122. (4) Sauer, K.; Camper, A. K. J. Bacteriol. 2001, 183, 6579-6589. (5) Sauer, K.; Camper, A. K.; Ehrlich, G. D.; Costerton, J. W.; Davies, D. G. J. Bacteriol. 2002, 184. (6) Whiteley, M.; Bangera, M. G.; Bumgarner, R. E.; Parsek, M. R.; Teitzel, G. M.; Lory, S.; Greenburg, P. E. Nature 2001, 413, 860-864. (7) Davey, M. E.; O’Toole, G. A. Microbiol. Mol. BioL. Rev. 2000, 64, 846847. (8) O’Toole, G. A.; Kaplan, H.; Kolter, R. Annu. Rev. Microbiol. 2000, 54, 4979. (9) Fiehn, O. Plant Mol. Biol. 2002, 48, 155-171. (10) Fiehn, O.; Kloska, S.; Altmann, T. Curr. Opin. Biotechnol. 2001, 12, 8226. (11) Raamsdonk, L. M.; Teusink, B.; Broadhurst, D.; Zhang, N.; Hayes, A.; Walsh, M. C.; Berden, J. A.; Brindle, K. M.; Kell, D. B.; Rowland, J. J.; Westerhoff, H. V.; van Dam, K.; Oliver, S. G. Nat. Biotechnol. 2001, 19, 45-50. (12) Weeckwerth, W.; Ehlers Loureiro, M.; Wenzel, K.; Fiehn, O. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 7809-7814. (13) Viant, M. R.; Rosenblum, E. S.; Tjeerdema, R. S. Environ. Sci. Technol. 2003, 37, 4982-4989. (14) Tweeddale, H.; Notley-McRobb, L.; Ferenci, T. J. Bacteriol. 1998, 180, 5109-5116.

Analytical Chemistry, Vol. 79, No. 21, November 1, 2007 8037

densities15 in Escherichia coli. Biological samples used for metabolomics have included a variety of biofluids such as urine, serum, blood, or bile for human and animal subjects,16,17 while plant and microorganism studies have been conducted on cells,18,19 cell extracts,20 or growth medium supernatants.21 Due to the diversity of metabolites and the range of concentrations, however, identifying all of the components of the metabolome, even for simple microorganisms, has proven difficult. Nuclear magnetic resonance (NMR) spectroscopy is a technique that can provide overall profiles of all the metabolite species within a crude sample.20 Pattern recognition techniques, such as principal component analysis (PCA), are often employed in order to gather useful information from the NMR spectra because such a wide variety of chemical species (i.e., NMR peaks) are present.22-25 The methods for such analysis are well developed in the drug discovery field, where bodily fluids from both humans and animals are routinely screened in order to develop methods for disease diagnosis.26,16 For studies on microorganisms, the application of PCA to liquid 1H NMR spectra has been used to distinguish between different strains of Bacillus cereus.27 Spectral comparison alone, without the use of PCA, has also found use in identifying antibiotic-resistant strains of Staphylococcus aureus28 and in discriminating between different species of bacteria.29-31 NMR spectra obtained using techniques such as high-resolution magic angle spinning (HRMAS), which are useful for studying semisolid samples such as cells, can also be used in metabolomic studies. HRMAS NMR involves spinning samples around their own axis at speeds between 4 and 12 kHz and at an angle of 54.7° relative to the external magnetic field in order to reduce line widths by averaging out chemical shift anisotropy, magnetic susceptibility, and dipolar coupling.32 Comparisons of 1H HRMAS NMR spectra have been used to distinguish between healthy and diseased human tissue samples33,34 and to determine chemical compositions of algal cells.18,19,35 Combined with PCA analysis, 1H (15) Lui, X.; NG, C.; Ferenci, T. J. Bacteriol. 2000, 182, 4158-4164. (16) Nicholson, J. K.; Wilson, I. D. Prog. Nucl. Magn. Reson. Spectrosc. 1989, 21, 449-501. (17) Lindon, J. C.; Nicholson, J. K.; Holmes, E.; Everett, J. R. Concepts Magn. Reson. 2000, 12, 289-320. (18) Broberg, A.; Kenne, L. Anal. Biochem. 2000, 284, 367-374. (19) Chauton, M. S.; Storseth, T. R.; Johnsen, G. J. Appl. Phycol. 2003, 15, 533542. (20) Fan, T. W. M. Prog. Nucl. Magn. Reson. Spectrosc. 1996, 28, 161-219. (21) Bubb, W. A.; Wright, L. C.; Cagney, M.; Santangelo, R. T.; Sorrell, T. C.; Kuchel, P. W. Magn. Reson. Med. 1999, 42, 442-453. (22) Lindo, J. C.; Holmes, E.; Nicholson, J. K. Proc. Nucl. Magn. Reson. 2001, 39, 1-40. (23) Daykin, C. A.; Van Duynhoven, J. P. M.; Groenewegen, A.; Dachtler, M.; Van Amelsvoort, J. M. M.; Mulder, T. P. J. J. Agric. Food Chem. 2005, 53, 1428-1434. (24) Ward, J. L.; Harris, C.; Lewis, J.; Beale, M. H. Phytochemistry 2003, 62, 949-957. (25) Holmes, E.; Nicholson, J. K.; Nicholls, A. W.; Lindo, J. C.; Connor, S. C.; Polley, S.; Connelly, J. Chemom. Intell. Lab. Syst, 2003, 44, 245-255. (26) Griffin, J. L. Curr. Opin. Chem. Biol. 2003, 7, 648-654. (27) Bundy, J. G.; Willey, T. L.; Castell, R. S.; Ellar, D. J.; Brindle, K. M. FEMS Microbiol. Lett. 2005, 242, 127-136. (28) Ohara, T.; Y., I.; Itoh, K.; Tetsuka, T. J. Infect. 2001, 43, 116-121. (29) Garg, M.; Misra, M. K.; Chawla, S.; Prasad, K. N.; Roy, R.; Gupta, R. K. Eur. J. Clin. Invest. 2003, 33, 518-524. (30) Bourne, R.; Himmelreich, U.; Sharma, A.; Mountforn, C.; Sorrell, T. J. Clin. Microbiol. 2001, 39, 2916-2923. (31) Delpassand, E. S.; Chari, M. V.; Stager, C. E.; Morrisett, J. D.; Ford, J. J.; Romazi, M. J. Clin. Microbiol. 1995, 33, 1258-1262. (32) Fukushima, E.; Roeder, S. B. W. Experimental Pulse NMR: A Nuts and Bolts Approach; Addison-Wesley Publishing Co., Inc.: Reading, MA, 1981.

8038

Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

HRMAS NMR spectra have been used to investigate the spectral information contained in different types of NMR experiments (Carr-Purcell-Meiboom-Gill, diffusion editing, and J-resolved spectroscopy) on liver tissues36 and to distinguish between types of intestinal tissue.37 The work presented here demonstrates the use of NMR-based metabolomic profiling techniques to characterize crude samples of planktonic and biofilm samples of the bacteria P. aeruginosa. Traditional liquid-state NMR spectroscopy was used to characterize the medium in which the cells were cultured and to determine differences in excreted metabolites and alterations in medium components. One-dimensional 1H NMR spectra of fresh growth medium were compared with spent medium supernatants from batch planktonic (which refers to bacterial growth in a closed system with a static source of nutrients where cells are suspended), chemostat planktonic (which refers to bacteria that are grown in a continuously fed suspended culture), and biofilms generated in a continual flow system (continuously fed and attached to a surface). In addition, 1H HRMAS NMR techniques were utilized to determine the chemical composition of the cells themselves for the two modes of growth. Finally, PCA was employed to investigate the statistical significance of the differences between collected 1H NMR spectra. To determine the robustness of this method, we repeated these experiments twice and used slightly different preparation methods on the cells. MATERIALS AND METHODS Inoculation Culture. P. aeruginosa strain PAO1 was grown from freezer culture on Luria-Bertani (LB)38 agar plates, which had an initial pH of 7.2. Cultures used for inoculating the chemostat and biofilm reactors were grown from a single colony inoculated in 50 mL of LB medium38 (22 °C, 18 h, with agitation). Planktonic Cultures. Two methods were used to grow planktonic P. aeruginosa strain PAO1 cultures. Batch cultures were generated by growth in 50 mL of LB medium (37 °C, 16 h, with agitation), after which samples were withdrawn for analysis. Alternatively, planktonic cultures were also grown in continual feed chemostats. Continuous planktontic cultures were grown in chemostats in LB medium (22 °C). Fresh LB medium was pumped into the 100-mL reactor volume at a rate of 0.41 mL/min (residence time, 4 h). Samples were harvested after the reactors operated under these conditions for 48 h. Biofilm Culture. To cultivate biofilm samples, a 3-mL volume of a planktonic culture was injected into a 1-m length of silicon tubing (internal diameter of 3/32 in.; Cole-Palmer) and allowed to incubate under static conditions for 4 h. Fresh LB medium was then pumped through unidirectionally (0.15 mL/min) for 48 h. (33) Cheng, L. L.; Ma, M. J.; Becerra, L.; Ptak, T.; Tracey, I.; Lackner, A.; Gonzalez, R. G. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 5408-6413. (34) Moka, D.; Vorreuther, R.; Schicha, H.; Spraul, M.; Humpfer, E.; Lipinski, M.; Foxall, P. J. D.; Nicholson, J. K.; Lindo, J. C. J. Pharma. Biomed. Anal. 1998, 17, 125-132. (35) Storseth, T. R.; Hansen, K.; Skjermo, J.; Kranes, J. Carbohydr. Res. 2004, 339, 421-424. (36) Wang, Y.; Bollard, M. E.; Keun, H.; Antti, H.; Beckonert, O.; Ebbels, T. M.; Lindo, J. C.; Holmes, E.; Tang, H.; Nickolson, J. K. Anal. Biochem. 2003, 323, 26-32. (37) Wang, Y.; Tang, H.; Holmes, E.; Lindo, J. C.; Turini, M. E.; Sprenger, N.; Bergonzelli, G.; Fay, L. B.; Kochhar, S.; Nicholson, J. K. J. Proteome Res. 2005, 4, 1324-1329. (38) Sambrook, J.; Fritsch, E. F.; Maniatis, T. Molecular Cloning; Cold Spring Harbor Laboratory Press: Cold Spring Harbor, NY, 1989.

The flow rate for this culture was 0.15 mL/min, and it had a laminar flow profile and a residence time of 20 min (less than double the time of the planktonic culture), which ensured that no planktonic bacteria would be retained in the system. Both inoculation and flow-through cultivation were carried out at the temperature of the respective planktonic culture, either batch (37 °C) or chemostat (22 °C), which the biofilm sample was to be compared. A medium break tube was placed upstream from the tubing to prevent contamination of the medium fluid reservoir. Biomass was harvested in a manner similar to that described by Sauer and Camper4 for protein analysis, whereby the tubing was pinched to dislodge the biofilm and then purged with filtered air and the sample collected in a centrifuge tube. Sample Designations. Biofilm samples were prepared from three biofilm bioreactors, and each reactor was divided equally into three sections of tubing with an approximate volume of 1 mL. The first section was designated as the section closest to the fresh medium inlet, the second being from the middle of the tubing, and the third most proximal to the outlet. For the analysis presented here, the samples will be referred to as BF for biofilm, followed by the reactor number and then the section number (e.g., a sample taken from the first reactor, first section (proximal to the medium inlet) is designated as BF1-1, second section BF1-2, and third section (distal to the medium inlet) BF1-3). Six planktonic cultures (three batch and three chemostat) were used for this analysis, and two samples were collected from each culture. Planktonic samples are referred to by the culture number first followed by the sample number, with a PL-B indicating batch planktonic or PL-C indicating chemostat planktonic, preceding. Thus, the first sample from the first batch culture will be PL-B1-1, and the second sample is PL-B-1-2. The sample designation scheme implemented for the supernatant samples was also used for the associated cell samples. Liquid Sample Preparation for 1H NMR Spectroscopy. Fresh LB medium samples were prepared by aseptically extracting a 0.5-mL aliquot of sterile medium and placing it in an NMR tube. To prepare spent medium supernatant samples, 25 mL of planktonic culture or 1-mL aliquots extracted from the tubular biofilm cultures were centrifuged (4085g, 5 min, 5 °C) to pellet cellular material. A 0.5-mL aliquot of the liquid supernatant was aseptically extracted from the centrifuge tube and placed in an NMR tube. All samples were then frozen (-20 °C) until NMR experiments were performed. In accordance with general NMR procedures, prior to acquiring the spectra, 0.5 mL of D2O + 0.2% volume tetramethylsilane (TMS) was added to the NMR tube for locking and referencing. 1H HRMAS NMR Sample Preparation. After liquid samples had been withdrawn, the remaining liquid supernatant was removed from the centrifuge tube, 0.5 mL of D2O was added, and the tube was vortexed for ∼5 s. The samples were then centrifuged in a microfuge (5 × 103 rpm, 5 min), and the supernatant was removed. This washing procedure was repeated two more times to remove any remaining medium from the cells. Cell pellets were either kept in the centrifuge tubes and refrigerated (5 °C) until the 1H HRMAS NMR experiments could be performed (∼3 weeks), or cell pellets were transferred to sterile freezer vials and immediately frozen (-70 °C). Immediately prior to each 1H HRMAS NMR experiment, a 4-mm HRMAS rotor (Bruker Biospin) was filled with the refrigerator-stored cells. Since the cell samples had been washed in D2O, there was enough residual for deuterium locking. Cell samples were streaked on

LB agar plates prior to and after running the 1H HRMAS NMR experiments, and the cells were determined to be viable in all instances. Frozen samples were lyophilized overnight and, immediately prior to each 1H HRMAS NMR experiment, 0.0039 ( 0.0001 g of the dehydrated cellular material and 200 µL of D2O were mixed and transferred to a 4-mm HRMAS rotor (Bruker Biospin). TMS was not added to any cell samples. Although refrigerator storage may have resulted in alterations of the observed chemical signatures, all samples were treated identically so relative chemical differences could still be discerned. Freezing and lyophilization were performed on separate samples to eliminate any potential aritifacts caused by storing these samples in a refrigerator. Liquid Supernatant 1H NMR Experiments. All 1H NMR spectra were acquired at 500.0894 MHz on a Bruker Avance spectrometer at ambient temperature. A 5-mm Bruker BBO probe was used to acquire liquid supernatant 1H NMR spectra. Water suppression was accomplished using a presaturation pulse sequence. For each spectrum, 64k data points were collected for 128 transients over a spectral width of 6775 Hz. A relaxation time delay of 3 s was used. All spectra were manually phase corrected and referenced to TMS at 0 ppm. 1H HRMAS NMR Experiments. 1H HRMAS NMR spectra of the cells were also acquired at 500.0894 MHz on a Bruker Avance spectrometer at ambient temperature. Spectra were acquired with a SB BL4 probe with z-gradients while spinning at 5 kHz. Presaturation water suppression was used to minimize the large water signal. For each sample, 64k data points were collected with 1000 scans and a recycle delay of 5 s. The spectra were manually phase corrected and externally referenced to TMS. Data Processing and Principal Component Analysis. The same data analysis was performed for both the liquid supernatant and 1H HRMAS spectra. The spectral region from 0 to 10 ppm was divided into equal “buckets” of 0.04 ppm using AMIX software (version 3.0.1, Bruker Biospin). The region from 4.5 to 6 ppm was set equal to zero to ensure that the variability of the suppressed water signal would not interfere with the statistical analysis. This analysis is shown in Figure 1. The integrated intensities were saved as an ASCII file and loaded as matrices into MatLab (version 7.0.4, MathWorks). The MatLab Statistics Toolbox (version 7.0.4, MathWorks) was used to perform principal component analysis. All PCA and spectral plots were generated using MatLab. RESULTS Spent Media: Biofilm and Batch Planktonic Cultures. To determine whether different cellular growth conditions could be distinguished by 1H NMR analysis of the resulting spent medium, we compared fresh LB growth medium with spent medium from planktonic and biofilm-grown cultures. Initially, samples were generated from batch planktonic cultures and continuous biofilm reactors, and these were frozen until they could be analyzed. 1H NMR spectrum of biofilm sample BF2-2 is presented along with planktonic sample PL-B-1-1 in Figure 2. In this figure, both 1H NMR spectra are compared directly with LB medium. The TMS reference peak at 0 ppm is shown in both the batch planktonic and LB medium spectra but is not present in the biofilm supernatant. Only one biofilm sample, BF3-3, displayed the Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

8039

Figure 1. Conversion of NMR spectra to “bucket” data for PCA. The region containing the suppressed water signal from 4.5 to 6 ppm is set equal to zero for all the samples to remove the suppressed water signal.

Figure 3. PCA score plot for LB media and batch planktonic and biofilm supernatant samples. The first principal component (x-axis) separates the LB media from the spent supernatants, while the second principal component (y-axis) distinguishes the groupings of the batch planktonic and biofilm supernatants.

Figure 2. Comparison of liquid 1H NMR spectra from fresh LB media to spent media from batch planktonic and biofilm cultures for the first run. Biofilm sample BF2-2 is presented along with batch planktonic sample PL1-1. (Top) Overlay of the 1H NMR spectra from fresh LB media and spent biofilm supernatant. (Bottom) 1H NMR spectra from fresh LB media compared to spent planktonic supernatant.

reference TMS peak at 0 ppm, but the intensity of this peak was much smaller than for the batch planktonic and LB medium spectra. All the samples were prepared with the same solution of D2O + TMS and handled in exactly the same manner, so it appears likely that the TMS had decomposed in the biofilm supernatant. The initial pH of the LB growth media was measured to be 7.2. The final pH after growth of both the planktonic and 8040 Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

biofilm samples was tested and found to be in the range of 7-8. The final pH was likely altered due to metabolic activity and would be expected to be different between planktonic and biofilm samples, as well as within the biofilm itself.39 Thus, some of the observed metabolomic difference may well be due to alterations in pH. The contribution of pH to differences in the metabolome is of interest and a future area of research, but beyond the scope of the current research. Due to pH ranging from 7 to 8 in the biofilm samples, it is not possible that strong acids or bases reacted with the TMS. This leaves two possibilities: (1) a strong oxidizing agent was present in the biofilm supernatant or (2) a reaction occurred due to the presence of some silicone compound that leached into liquid from the silicone tubing used to cultivate the biofilm samples. To validate that the groups of spectra were all statistically different, PCA was performed. The score plot, which is shown in Figure 3, demonstrates that the samples from the LB medium and batch planktonic and biofilm supernatants are all contained within separate groupings. The first principal component (PC), which accounts for 45% of the variability in the data, separates the fresh LB medium from the spent supernatant samples. The second PC, which describes an additional 33% of the variability, isolates the batch planktonic samples from the LB and biofilm samples. An additional PC, not shown, describes 12% of the variation, so that the 90% of the variation in the data is described by three principal components. Region expansions of 1H NMR spectra from batch planktonic and biofilm spent culture supernatants showed that most of the (39) Vroom, J. M.; et al. Appl. Environ. Microbiol. 1999, 65, 3502.

Figure 5. PCA score plot for LB media and chemostat planktonic and biofilm supernatant samples for the second run. The first principal component (x-axis) separates the LB media from the spent supernatants while the second principal component (y-axis) distinguishes the groupings of the chemostat planktonic and biofilm supernatants.

Figure 4. Region expansions for biofilm and batch planktonic supernatant 1H NMR spectra. Very few differences can be seen in the location of peaks; the main differences are seen in the intensity of the peaks. (Top) Expansion of the spectral region from 0.5 to 2 ppm. (Middle) Expansion of the spectral region from 2 to 4 ppm. (Bottom) Expansion of the spectral region from 6.5 to 8 ppm.

variation between the biofilm and batch planktonic 1H NMR spectra occurs due to higher peak intensities in the planktonic supernatant. This is shown in Figure 4. Specifically, cell-generated acetate at 1.92 ppm20 appears to have a stronger intensity and, therefore, higher concentrations, in the planktonic culture. This higher intensity may well be due to the batch culture technique used to generate planktonic samples allowing cell products to accumulate in the medium. In contrast, biofilm samples were grown in a flow-through system where cell products could be either flushed away or diluted with fresh medium. There does not appear to be any detectable new or novel peaks present in either of the samples relative to each other, with the exception of the TMS reference peak.

Spent Media: Biofilm and Chemostat Planktonic Cultures. To eliminate any potential variability introduced by batch versus continual feed modes of planktonic growth, and to better isolate metabolite differences solely owing to planktonic/biofilm processes, a second set of experiments utilized chemostat-grown continuous planktonic cultures and compared these to continual feed biofilm reactors. Again, after samples were collected, they were frozen until analyzed. PCA was performed on the 1H NMR spectra of supernatant samples, and its shows that the samples from the LB media and chemostat planktonic and biofilm supernatants are all contained within separate groupings, shown in Figure 5. The first PC, which accounts for 71.2% of the variability in the data, separates the LB media from the supernatant samples. The second PC, which describes an additional 20.5% of the variability, isolates the planktonic samples from the LB and biofilm samples; 90% of the variation in the samples is described by these two components. Not surprisingly, these results demonstrate that spent medium is most dissimilar from fresh medium. However, when supernatant samples of only the spent media from the biofilm and chemostat planktonic cultures were compared alone, a clear separation of the two modes of growth along the first principal component, which describes 75.2% of the variability, was observed. The second PC accounts for an additional 11.1% of the variability and represents the differences within each of the groupings, shown in Figure 6. These data further support the contention that differences in 1H NMR spectra are due solely to biofilm versus planktonic modes of growth, since they are apparent even when both types of cultures were grown in continual feed mode, and thus, artifacts due to batch verses planktonic growth were eliminated. Peaks that are greater than zero in the loading plot in Figure 6 are of higher intensity or, since equal volumes were used for these studies, higher concentration in the planktonic supernatant. Conversely, those peaks which appear in the negative region of the loading plot have higher concentrations in the biofilm supernatant. The peaks at 0.73-0.97, 1.05, 1.65, 1.71, and 1.89 ppm are the spectral peaks, which stand out from the noise level (having a loading greater than 0.1) that are greater in the planktonic culture and lower in the biofilm culture. The peaks, which appear at values less than -0.1, have higher concentrations in the biofilm supernatant are at 0.99, 1.03, 2.13, 2.35, 2.41, 3.05, Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

8041

Figure 6. (Top) PCA score plot for chemostat planktonic and biofilm supernatant samples. (Bottom) Loading plot of the first principal component. Peaks greater than zero have a higher concentration in the planktonic supernatant; peaks less than zero have a higher concentration in the biofilm supernatant.

and 3.27 ppm. Since the temperature and reference peaks were tightly controlled, pH changes could account for some observed shifts in the NMR spectra. Slight pH differences within the biofilms, and between biofilms and planktonic cultures, are difficult to determine under any circumstances, because of the very small spatial and area scales involved in biofilm structure. 1H NMR metabolomic methods can provide insight into these changes, which would in turn shed more light on biofilm function and interactions. Presently, however, it is not known if all of the observed shifts are solely due to pH changes. Further understanding of the contribution of pH alterations on metabolomic chemistry would clearly require deconvoluting slight shifts in pH and metabolomic activity. Even if all the observed shifts were due to alterations in pH, however, these differences would still be valid indicators of alterations in the respective metabolomes. 1H HRMAS NMR of Cells. Biofilm and batch planktonic cultures were stored at 5 °C. While the supernatant from cultures were analyzed by liquid-state NMR spectroscopy, cells were investigated with 1H HRMAS solid-state NMR spectroscopy. One cell sample was run from each of the three sections of the three biofilm reactors for a total of nine biofilm samples for samples that were stored at 5 °C. Three batch planktonic cell samples (also stored at 5 °C), corresponding to each of the supernatant samples, were investigated. One of the principal concerns when running 1H HRMAS NMR on whole cells was sample preparation and storage. Previous studies have shown that experimental conditions, such as magnet shimming and sample preparation, can effect the results of a 8042

Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

principal component analysis for liquid samples.40 Preliminary studies conducted in our laboratory indicated that the 1H HRMAS NMR spectra are also very sensitive to alterations in magnet conditions. Initially, 1H NMR spectra were collected a week or two apart from each other, and despite efforts to normalize experimental conditions by adjusting the magnet shimming, PCA analysis revealed groupings based on the date that the spectra were acquired rather than the sample type. To further define this potential artifact in the data, the cell samples were collected and stored over the course of two weeks. The 1H NMR spectra for all stored samples were then acquired over the course of a few days, without extreme changes in the shimming. These results provided a more reliable analysis and are presented here. Figure 7 displays representative 1H HRMAS NMR spectra from batch planktonic culture PLB-3 and biofilm sample BF2-3 (both stored at 5 °C). The expanded regions, the middle and bottom plots, show that significant qualitative differences, not just peak intensities (as observed in the liquid NMR spectra), are found between the two spectra. These data clearly show that the planktonic and biofilm cell samples display distinctly different 1H HRMAS NMR peaks. The bottom plot of Figure 7 shows that the planktonic batch culture displays peaks at 6.59 and 7.87 ppm that are not found in the biofilm cells. The center plot of Figure 7 displays many areas where the planktonic sample has higher peak intensity, most notably at 1.74 , 3.02 , 3.25 and from 3.5 to 4 ppm. However, the peak at 2.33 ppm has greater intensity in the biofilm spectra. The NMR peak at 2.33 ppm is most likely the acetyl group of alginate,41 which is produced in large quantities in P. aeruginosa biofilms. The original reference for this peak is at 2.13 ppm. The shift from 2.33 to 2.13 ppm is most likely due to the fact that, for our work, spectra were acquired at 25 °C while the reference was at 92 °C. There are additional resonance peaks between 3.73 and 5.36 ppm reported by the reference. However, because of the water suppression scheme, which suppresses the broad water signal and protons that are exchanging with the water, these peaks were not observed in the data presented here. The PCA analysis of the 1H HRMAS NMR data in Figure 8 shows that the score plot of the cell samples (refrigerator-stored batch planktonic and biofilm samples) divides into two distinct and widely separated groupings based on the first PC, which describes 60% of the variation within the data set. The second PC, which accounts for 15% of the variation within the data, describes the wide variation in the biofilm samples. This analysis also reveals separation of the spectra of one of the biofilm reactors, BF1 circled in light gray, from the other two. The separation is most likely due to the fact that reactors 2 and 3 were grown in parallel and BF1 was grown a week prior to the others, demonstrating that growth variability can be measured using this technique. 1H HRMAS NMR of Cells: Frozen and Lyophilized Biofilm and Chemostat Planktonic Cultures. A second separate set of biofilm reactors was run from which the extracted samples were immediately frozen at -70 °C to eliminate any potential artifacts associated with storage at 5 °C. These were compared to chemostat-grown cells also prepared by freezing and lyophilization. (40) Defernez, M.; Colquhoun, I. J. Phytochemistry 2003, 62, 1009-1017. (41) Skjak-Braek, G.; Grasdalen, H.; Larsen, B. Carbohydr. Res. 1986, 154, 239250.

Figure 8. PCA score plot for 1H HRMAS NMR spectra of biofilm and batch planktonic cells. Clear separation of two types of samples is shown. In addition, one of the biofilm reactors, BF1, is separated from the other two shown circled in light gray.

second PC, which accounts for 12.5% of the variation within the data, describes the wide variation in the biofilm samples. The loading plot shown in the bottom of Figure 10 provides a more quantitative view of the differences between the two types of cells: the majority of the peaks in the spectra are in the negative region of the loading plot, which is a reflection of the lower peak intensities seen for the biofilm samples in Figure 9.

Figure 7. Comparison of representative 1H HRMAS NMR spectra of batch planktonic and biofilm cell samples. Many differences in the location and number of peaks can be seen throughout the spectra. (Top) Full spectra showing the variable region of water suppression that must be removed for PCA. (Middle) Expansion of the region from 0 to 4 ppm. (Bottom) Expansion of the region from 6.5 to 8.5 ppm.

Two samples from each of these three biofilm reactors and each of the three planktonic chemostats were investigated. Figure 9 displays representative 1H HRMAS NMR spectra from chemostat planktonic cells compared to biofilm cells. Most noticeable is the fact that the peak intensities of the biofilm spectra are significantly less than those in the planktonic culture indicating that, on a mass basis, the biofilm cells are less metabolite rich than the chemostat grown cells. The PCA analysis of the 1H HRMAS NMR data in Figure 10 shows that the score plot of the cell samples divides into two distinct and widely separated groupings based on the first PC, which describes 80.5% of the variation within the data set. The

DISCUSSION Extracellular Metabolic Profiles. In previous studies comparing biofilm and planktonic cells, chemostats have been employed to ensure that all of the cells experience the same environment. Using that method, the biofilm is cultivated on a surface inside a reactor, which also contains suspended cells. In the present work, we wanted to analyze the extracellular environment of the two growth modes, which required separate cultivations. In the first set of experiments, spent media from batch planktonic and continuous feed-generated biofilm cultures were compared. The separate cultures showed no qualitative differences in metabolites, but the quantities produced varied, with generally higher levels occurring in the batch planktonic supernatants. The higher levels of metabolites in the batch planktonic culture were most probably the result of accumulation of metabolic end products due to batch mode of growth, whereas the biofilm medium was continually refreshed so end products could not accumulate. It was noted, however, that, despite these confounding issues, the supernatants of biofilm and batch planktonic cultures could be readily distinguished by PCA (Figure 3). To better discern metabolic differences between planktonic and biofilm modes of growth as reflected by the composition of extracellular growth medium, planktonic cells were grown in (continual feed) chemostats and compared to biofilm cells grown under similar continuous culture conditions. Under these conditions, it was revealed that some metabolic components were present in greater concentrations in the planktonic culture, while others were more concentrated in the biofilm-derived samples (Figure 4). Once again, PCA analysis was successful in distinguishing the differences between the two growth modes (Figure 5) The spectra for the two types of cells appear markedly similar despite the batch versus continuous feed modes of cultivation. From this analysis, it appears that, at the current degree of resolution, the overall metabolism in a biofilm system is quite Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

8043

Figure 10. (Top) PCA score plot for chemostat planktonic and biofilm cell samples for the second run. (Bottom) Loading plot of the first principal component.

Figure 9. Comparison of representative 1H HRMAS NMR spectra of chemostat planktonic and biofilm cell samples. Many differences in the location and number of peaks can be seen throughout the spectra. (Top) Full spectra showing the variable region of water suppression that must be removed for PCA. (Middle) Expansion of the region from 0 to 2 and 2-4 ppm. (Bottom) Expansion of the region from 6 to 9 ppm.

similar to that of planktonic cultures. This result might be expected since it is known that cells have various, stratified growth rates throughout the depth of a biofilm. Cells at the biofilm-liquid interface cover the most surface area and have the highest growth rates and most active metabolism.3 Therefore, it is likely that it is the metabolic products of these cells which dominate the secreted metabolite signatures that are reflected in the NMR spectra. Previous studies have shown that these surface cells are as susceptible to antimicrobial treatment as planktonic cells.42,43 This 8044 Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

combined evidence indicates that the cells on the biofilm-liquid interface are utilizing the same metabolic pathways as suspended cells. Intracellular Metabolic Profiles. Both qualitative and quantitative differences between planktonic and biofilm modes of growth were observed when the cells themselves were studied by 1H HRMAS NMR; this was the case whether biofilm cells were compared to batch- or chemstat-grown planktonic cells and also was apparent whether cells were stored or immediately frozen and lyophilized. The 1-D 1H NMR cell spectra presented here are far too complex to identify all of the individual compounds within the two systems; however, they do demonstrate the distinct nature of the two types of cells. This was demonstrated using PCA analysis to determine significant differences between spectral types; biofilm and planktonic cells clearly grouped separately even when culturing techniques were comparable and storage and sample preparation issues were eliminated. In one case using stored biofilm cells, cells derived from a single reactor could even be distinguished from the other biofilm reactor trials, demonstrating the potential sensitivity of this technique to reveal subtle metabolic variability between different cultures ostensibly grown under the same conditions (Figure 8). Lower spectral peak intensities were clearly apparent in cells derived from biofilms; this became apparent even when potential artifacts arising from batch versus chemostat planktonic growth were eliminated. This could well be the result of cells located closer to the substrate operating at lower metabolic rates, as has been reported previously.9 Thus, for a given mass of harvested (42) Fux, C. A.; Costerton, J. W.; Stewart, P. S.; Stoodley, P. Trends Microbiol. 2005, 13, 34-40. (43) Spoering, A. L.; Lewis, K. J. Bacteriol. 2001, 183, 6746-6751.

cells, the lower metabolic rate of those beneath the biofilm surface would tend to depress peak intensities for a given sample. Most interesting, perhaps, is the observation of different peaks apparent in biofilm and planktonic cultures; this invites the possibility of using this type of analytical approach to identify specific metabolites and the causative metabolic processes from which they originate. In the case of cells stored at 5 °C, the cells were most probably under stress at the time the 1H NMR spectra were collected. Investigating the stress response of cells to environmental stimuli is one goal of metabolomics, and this study presents just one of the thousands of possible instances that could be investigated. Subsequent trials, however, attempted to eliminate this factor by immediate freezing and freeze-drying cells to best preserve the state they were in when harvested. In either case, the results demonstrated that 1H HRMAS NMR is capable of distinguishing multiple relative differences between cells grown under different conditions using a single analytical technique. The PCA analysis of the 1H HRMAS NMR spectra clearly shows the statistical significance of the differences between the planktonic and biofilm cells, with 60 and 80.5% of the difference described by the first PC (Figures 8 and 10). The second PC, which describes the spread in the biofilm samples, is also quite significant at 15 and 12.5% (Figures 8 and 10). This result is most likely indicative of the varied physiological states within a biofilm, from anaerobic at the center of clusters to the highly aerobic growth rate zone at the liquid interface. In contrast, the planktonic samples, which are assumed to be homogeneous, form a tighter grouping and lend further credence to the conjecture that planktonic cells are generally uniform in terms of metabolism. It was anticipated that the variation in the biofilm samples would be distinguished based on the reactor that the samples were grown in and by the section of biofilm reactor. Since the biofilm growth setup is essentially a plug flow reactor, the beginning sections of tubing should have higher concentrations of nutrients while the later sections should have higher concentrations of metabolites. However, no such groupings were apparent in the cell or supernatant sample PCAs. The obvious answer to the lack of clear separation between the different sections of biofilm

reactors would be to assume that significant mixing occurs during the harvesting procedure, which homogenizes the samples. However, if this were the case, all the reactors should cluster together much more readily than is apparent by either the supernatant or cell data. Therefore, these results seem to reflect true heterogeneity, both along the length of the biofilm reactor system and between the reactors, so tracking the location of biofilm samples within a reactor was abandoned for subsequent trials. A more in-depth study of the causes and conditions involved in this heterogeneity should be undertaken in the future. This study has shown that in order to conduct NMR-based metabolomic studies on biofilm systems, 1H NMR of the cell contents rather than that of the supernatant will have to be employed. This work has also shown the utility of employing 1H HRMAS NMR to analyze metabolic signatures within cells and make relative comparisons between cells grown under varying conditions. The true utility of this metabolomic approach will become more evident when combined with proteomic and genomic data. However, due to the vast variation in biofilm systems under different growth conditions, this type of work will most certainly have to be performed on a single, well-defined system. The work presented in this paper represents one small piece of a much larger puzzle that will eventually form a picture of the metabolic responses of biofilm and planktonic cells to environmental factors. ACKNOWLEDGMENT Work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under contract W-7405-ENG-48. The project (05-ERD-026) was funded by the Laboratory Directed Research and Development Program at LLNL. The authors thank Sarah Chinn for NMR support, Josh Ulloa for laboratory assistance, and Dan Ohman for providing the P. aeruginosa strain used in these studies.

Received for review April 20, 2007. Accepted August 15, 2007. AC070800T

Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

8045