Differentiation of microbial species and strains in coculture biofilms by multivariate analysis of laser desorption postionization mass spectra Authors: Chhavi Bhardwaj, Yang Cui, Theresa Hofstetter, Suet Yi Liu, Hans C. Bernstein, Ross P. Carlson, Musahid Ahmed, and Luke Hanley. NOTICE: This is a postprint of an article that originally appeared in Analyston in 2013. DOI: 10.1039/c3an01389h. Bhardwaj C, Cui Y, Hofstetter T, Liu SY, Bernstein HC, Carlson RP, Ahmed M, Hanley L, "Differentiation of microbial species and strains in coculture biofilms by multivariate analysis of laser desorption postionization mass spectra," Analyst. 2013 Nov 21 138(22):6844-51. Made available through Montana State University’s ScholarWorks scholarworks.montana.edu Differentiation of microbial species and strains in coculture biofilms by multivar desorption postionization mas Chhavi Bhardwaj,a Yang Cui,a Theresa Ho Ber c c 7.87 e S) to a ro th ip diffe o l E. col s 2 stra ly d th n c e expe o ing a w re I- r Many microorganisms live within ed of monocultures or more commo bit complex interspecies relationship ng problematic medical or benecia he critical interactions occurring betw ng strains or species in a mixed cult lic potential as well as provide insig nd Saccharomyces cerevisiae are arg respectively, and serve as ideal ent study employs two distinct bina cal motif which revolves around ex se- oxidizing producer member, eith eed metabolic byproducts such as ac m's scavenger. Multivariate analysis can facilit act biological samples. Principal com ous mult ortho 12 b clas- strai ioniz Introductionivariate analysis methods. PC .e., gonal) variables called princ ecause it reduces variable dim sication and/or grouping o ns of microorganisms by PC ation mass spectra (MALDI- complex surface associated communities known as biolms.1 Biolms can be compris nly, exist as consortia of multiple interacting species.2,3 Consortial biolms can exhi s and dependencies that are subject of many ongoing investigations.4,5 Controlli l environmental multispecies biolms is challenging due to limited knowledge of t een the microorganisms organized within microscale structures. Spatially differentiati ure biolm can provide useful informa-tion, such as localization of available metabo ht into the competitive nature of the metabolic interactions.6,7 Escherichia coli a uably the best-studied prokaryotic and eukaryotic microorganisms, model systems for developing techniques to examine biolm interactions. The curr ry cocultures designed to mimic a naturally occurring producer-consumer ecologi changes of metabolites between microor-ganisms.5 Both cocultures have a gluco er an E. coli deletion mutant8 or the baker's yeast S. cerevisiae, which cross f etate or ethanol to a glucose-negative E. coli consumer strain which acts as the syste ate the processing of large data sets from the chemical analysis of biolms and other int ponent analysis (PCA) is the most commonly used and widely reported of the vari A decomposes data with correlated measurements into a new set of uncorrelated (instein, Ross P. Carlson, Musahid Ah to 10.5 eV vacuum ultraviolet (VUV) photon energies were us nalyze biofilms comprised of binary cultures of interacting mic tunable synchrotron and laser sources of VUV radiation. Princ rentiate species in Escherichia coli–Saccharomyces cerevisiae c i strains in a biofilm comprised of two interacting gene deletion in by no more than four gene deletions each out of approximate e E. coli strains into three distinct groups, two “pure” groups, a ocultures showed greater variance by PCA at 7.87 eV photon en ctation that the 7.87 eV photoionization selects a subset of low i ider range of analytes. These two VUV photon energies therefo MS constitute an additional experimental paipal components.9 PCA is id ensionality with a minimal l f experimental results to ext A of secondary ion mass MS).14,15iate analysis of laser s spectra fstetter,b Suet Yi Liu,b Hans C. meda and Luke Hanley*a d in laser desorption postionization mass spectrometry (LDPI-M organisms. The effect of photon energy was examined using bo al components analysis (PCA) was applied to the MS data to culture biofilms. PCA of LDPI-MS also differentiated individua trains, even though these strains differed from the wild type K-1 2000 genes. PCA treatment of 7.87 eV LDPI-MS data separate d a mixed region. Furthermore, the “pure” regions of the E. coli rgies compared to 10.5 eV radiation. This is consistent with the nization energy analytes while 10.5 eV is more inclusive, detect give different spreads via PCA and their respective use in LDP ameter to differentiate strains and species.eal for interrogating large data sets such as mass spectra,10– oss of information. For these reasons, PCA is oen used for ract useful information. Several studies have differentiated spectra (SIMS)12,13 and matrix assisted laser desorption sample preparation such as the addition of matrix, which can desorption laser (Spectra-Physics Explorer, Newport Corporation,obscure species in the low mass range. Both SIMS and MALDI- MS typically require thin, electrically conductive samples. Finally, ion suppression and local uctuations in ionization efficiency can complicate quantication in these and other methods in MS imaging.20 Some of the limitations of SIMS and MALDI-MS for the analysis of biolms and other biological samples can be over- come using laser desorption postionization mass spectrometry (LDPI-MS), which uses vacuum ultraviolet (VUV) radiation to induce a relatively ‘so’ single photon ionization of laser des- orbed neutrals.21,22 LDPI-MS, also referred to as two laser mass spectrometry (L2MS),23 does not require the addition of a matrix compound to enhance desorption. LDPI-MS is not sensitive to ion suppression since it detects desorbed neutrals and shows some advantages for quantication.24 Furthermore, LDPI-MS can readily analyze thick, electrically insulating samples.25 The ability of LDPI-MS to detect intact parent ions of molecular species depends upon the extent of energy transfer during the separate laser desorption and single photon ioni- zation steps. Prior LDPI-MS work showed that the higher VUV photon energies in the range of 7.8–12.5 eV improved sensitivity and produce intense molecular ion signal, but leads to forma- tion of fragment ions and background gas ions.26,27 Because the 10.5 eV photon energy appeared to provide the optimal balance between improved sensitivity and minimal fragmentation, a more robust laboratory source of 10.5 eV radiation was recently developed.25 The present study utilized 7.87 to 10.5 eV VUV photon energies from both tunable synchrotron and laser sources for LDPI-MS to analyze biolms comprised of binary cultures of interacting microorganisms. PCA was applied to the MS data to differentiate species in E. coli–S. cerevisiae biolms and to differentiate individual E. coli strains in a biolm comprised of two interacting gene deletion strains. Clear spatial separation of both species as well as both E. coli strains, based on their distinct metabolic states, was demonstrated by PCA of LDPI-MS of intact biolms. II. Experimental details A. Biolm growth and sample preparation Biolm culturing conditions were identical to those reported previously.25 The two E. coli K-12 deletion mutant strains (403G100-DptsGDptsMDglkDgcd expressing the reporter protein ds-tomato and 307G100-DaceADldhADfrdA expressing reporter protein citrine, are termed here the tomato and citrine strain, respectively) and their growth conditions were describedSIMS and MALDI-MS have been widely used for MS imaging, both with and without PCA, of biolms and other intact bio- logical samples.16–20 SIMS has the advantage of high spatial resolution. However, molecular information for many species is oen lost to fragmentation in SIMS, which readily detects low mass (m/z < 300) and atomic ions. MALDI-MS is less chemically destructive than SIMS and is oen used to detect molecular ions of species up to several kDa mass. However, MALDI-MS has lower spatial resolution than SIMS and requires extensiveIrvine, CA). This laser was operated at 2500 Hz repetition rate with a spot size of 30 mm diameter and laser desorption peak power density of 1 to 10 MW cm2. Tunable VUV synchrotron radiation in the range of 7.87–10.5 eV photon energy was additionally introduced into this instrument for single photon ionization of laser desorbed neutrals. The laser LDPI-MS at the University of Illinois at Chicago was also described previously.22,25 A 349 nm Nd:YLF laser (Spectra- Physics Explorer) was used for desorption at 10 or 100 Hz repetition rate which depended on the VUV ionization laser. The desorption laser beam was used at peak power density of 300 MW cm2 with a 50 mm diameter beam on the biolm. Laser postionization was carried out at two xed photon ener- gies: 7.87 and 10.5 eV. A 157.6 nm uorine laser (Optex Pro, Lambda Physik, Ft Lauderdale, FL) operating at 100 Hz was used for 7.87 eV postionization with a cross sectional area of 2 1 mm2 and an energy of 100 mJ per pulse. 10.5 eV laser pos- tionization was performed with a 118 nm beam generated by tripling the third harmonic of a Nd:YAG laser (355 nm, 20 mJ, 5 ns, Tempest, New Wave Research, Fremont, CA) in a Xe gas cell at 6.5 Torr pressure, as described previously.25 The resulting photoions were pulse extracted and analyzed using a 340 ppmpreviously.8 The yeast, S. cerevisiae, was grown using a similar protocol. The term monoculture refers to a biolm sample with only one species or strain. The term coculture refers to a binary culture of two genetically distinct microorganisms that have been grown together on the same substrate. The coculture biolms grown here were inoculated initially a few mm apart by two monocultures, then allowed to grow towards each other until they visually overlapped. LDPI-MS analysis of coculture samples was performed on three distinct regions: two spots comprised predominantly of one type of microbe (“pure” region at the outer edges of the coculture sample) and at the center of the sample where the two species visually overlapped (“mixed” region). All biolms used for the study were grown for an identical time period of 96 hours. All microbes were in stationary phase. Monoculture and coculture biolms were grown under identical experimental conditions, to eliminate any variation due to different growth phases. Biolms were grown on insulating polycarbonate membranes, the membranes were then adhered to a stainless steel plate with copper tape for introduction into vacuum for MS analysis. Biolms were also blotted onto stainless steel plates: the blots were then introduced into vacuum. B. LDPI-MS instrumentation LDPI-MS analysis was carried out on customized instruments utilizing either quasi-continuous synchrotron or pulsed laser VUV postionization sources. Details of the synchrotron LDPI-MS instrument have been reported previously.26–29 In brief, the synchrotron LDPI-MS was located on the Chemical Dynamics Beamline at the Advanced Light Source (Lawrence Berkeley National Laboratory, Berkeley, CA)30 and consisted of a commer- cial SIMS instrument (TOF.SIMS 5, ION-TOF Inc., Munster, Germany) modied by the addition of a 349 nm Nd:YLF pulsed sublimed gaseous neutrals, respectively. Prior control experi- ments with the 10.5 eV laser LDPI-MS established the presence of single photon ionization and ruled out both direct ions from laser desorption and photoelectron ionization effects due to the residual 355 nm beam used to generate VUV radiation.25 Photoelectron ionization was previously ruled out in the synchrotron LDPI-MS.27 The similarity of the spectra of 7.87 eV synchrotron and laser LDPI-MS ruled out photoelectron ioni- zation in the latter case. B. Analysis of intact biolms vs. blotted samples It is technically convenient to grow biolms on polymer membranes, but these insulating membranes can complicate analysis by MS imaging strategies. Membrane biolm analyses were attempted on the synchrotron LDPI-MS instrument, but reproducible MS signal was elusive due to what appeared to be excessive sample charging of these electrically insulating III. Results A. Effect of VUV photon energy on analysis of monoculture biolms Blots of yeast monoculture biolms were analyzed by synchro- tron LDPI-MS using 7.87, 8.5, 9.5 and 10.5 eV VUV photon energies with the results shown in Fig. 1. The synchrotron LDPI-MS of yeast monocultures showed that increasing photon energy resulted in more peaks in the mass spectra and higher overall signal intensity. 10.5 eV showed more peaks for molec- ular species as well as more fragments compared to the lower photon energies. However, fragmentation increased signi- cantly at photon energies between 10.5 and 15 eV (data not shown). Overall, 10.5 eV photon energy displayed an optimal balance between sensitivity and fragmentation, in agreement with prior experiments.26,27,35 Similar conclusions were reached from 7.87–10.5 eV synchrotron LDPI-MS of E. coli (tomatoA custom program was designed in-house for converting full mass spectra to integer m/z peak values similar to those used for electron impact MS database searching.32 No peak selections were made. Rather, the entire raw mass spectrum was aligned to integer m/z values, then zeroes were inserted as intensity values for peaks where no ion signal was observed above the assigned signal threshold of 35 mV. Ion intensities were normalized prior to PCA by setting the most intense peak to unity, thereby minimizing potential uctuations due to instrumental factors such as modulating detector efficiencies. Finally, ion intensities at each integer m/z 50–700 value for each sample were exported to a spreadsheet. A nonlinear iterative partial least squares algorithm (“princomp” function, MATLAB, The MathWorks Inc., Natick, MA) was used to perform PCA33,34 on the n  p data matrix where n represents the number of experiment trials for each strain and p represents the ion intensities written to the aforementioned spreadsheet for each integer m/z value. PCA was performed aer mean centering of the columns of the data matrix. Each data set used in the analysis was replicated at least three times and a minimum of 15 spectra of each microbe type was used to compose the original data matrix.mass accuracy reectron TOF which is similar to that described previously.25 The data acquisition soware was partially described elsewhere with respect to its use on a different instrument.31 The sample stage on the synchrotron LDPI-MS was rastered to analyze a 3 mm2 area for each biolm sample with 10 laser shots per spot before moving to an fresh spot for repeated analysis such that 1.2  105 laser shots were used for col- lecting each displayed synchrotron mass spectrum. The laser LDPI-MS utilized a sample stage that was continuously moving at a speed of 0.05 mm s1 while the desorption laser scanned over the sample for a total area of analysis of 1 mm2. Mass spectra were averaged over the entire analyzed area such that 4  104 laser shots were used to collect each displayed laser LDPI mass spectrum. C. Data acquisition and analysisstrain) monoculture blots (full mass spectra shown in ESI†). The mass spectral peaks observed for E. coli and yeast monocultures were tabulated with the VUV photon energies at which they were rst observed by synchrotron LDPI-MS (see ESI†). Prior synchrotron LDPI-MS studies26,27 led to the recent development of a rened laboratory VUV photoionization source that operates at 10.5 eV25 to complement the 7.87 eV uorine excimer laser source.22 These two laboratory laser VUV sources were used to collect 7.87 and 10.5 eV laser LDPI-MS data from yeast and E. colimonocultures. 10.5 eV laser LDPI-MS data for blotted monoculture biolms (yeast or E. coli) was collected and compared with the similar spectra collected by 10.5 eV synchrotron LDPI-MS (see ESI†). The synchrotron LDPI-MS data of the blotted samples showed overall higher S/N compared to that from laser LDPI-MS. The data traces labeled “Desorption Laser Only” and “Synchrotron Only” in Fig. 1 demonstrate the absence of any background signal arising from direct ionization via the desorption laser or VUV postionization of background or Fig. 1 7.87–10.5 eV synchrotron laser desorption postionization mass spectra (LDPI-MS) of blotted yeast monoculture biofilms. The bottom two traces show the controls performed with only synchrotron photoionization or desorption laser. membranes. However, intact membrane biolms could be analyzed directly in vacuum by laser LDPI-MS, as discussed below. Fig. 2 shows 10.5 eV laser LDPI mass spectra of blotted and intact biolm samples from yeast monocultures. There was 8  higher S/N for the intact membrane samples compared to the blotted biolms. Although the same peaks were observed for both samples, overall peak intensity was higher for membrane samples. Additionally, peaks at m/z 600–700 were not discern- ible in the blotted sample due to low S/N. Direct comparison was also made for membrane biolms analyzed with laser LDPI-MS vs. blotted biolms analyzed with synchrotron LDPI-MS: the different spectra showed similar peaks, albeit with differences in S/N and absolute signal intensity. The synchrotron MS data showed more peaks in the low mass (m/z < 300) region. Intact membrane biolms were not observed to delaminate in vacuum and maintained their structural integrity. However, the blotted biolms occasionally suffered from delamination in vacuum which adversely impacted the synchrotron LDPI-MS. from the limits of sample preparation: the clear boundaries of the microbial species were not visually apparent in the blotted samples, making it difficult to assign a given region to a specic species. Additionally, synchrotron LDPI-MS were acquired with desorption laser scanning over a relatively large, 3 mm2 area for each region, increasing the possibility of overlap of the different regions. By contrast, the laser LDPI-MS allowed analysis of a 1 mm2 area of an intact membrane biolm with relatively high S/N. Therefore, coculture E. coli and yeast biolms were studied using both the 7.87 and 10.5 eV laser photoionization sources, with Fig. 3 showing the latter result. The trace labeled “Mixed” in Fig. 3 refers to the region where the two species overlap. The “pure” regions of E. coli and yeast could be readily distinguished by visual examination of the intact membrane biolms and the mass spectra at both photon energies appeared similar to those of their respective monocultures. The mass spectra for the mixed region showed peaks from both of the species. 10.5 eV laser LDPI- MS showed more peaks in the high mass region (m/z > 400) compared to 7.87 eV data, indicating more molecular species samples (see ESI†). Given that these two microbes showedC. Coculture multispecies biolms 10.5 eV synchrotron LDPI-MS were recorded at three distinct regions of blotted E. coli and yeast coculture biolms (see ESI†). The “pure” E. coli and yeast regions of the biolms generated spectra that generally appeared similar to those from the cor- responding monocultures. However, synchrotron LDPI-MS of the mixed region of the coculture biolms tended to vary from one sample to the next. This variability might have resulted Fig. 2 10.5 eV laser LDPI-MS of yeast monoculture biofilms using two different sample preparation techniques. Spectra are normalized.The continuous synchrotron radiation readily ablated delami- nating pieces since the VUV beam passed very close to the top of the sample, leading to signal spikes near m/z 270 in the MS (see ESI†). Ablation signal was not observed in laser LDPI-MS since the pulsed VUV beam passed farther above the sample, missing the delaminated pieces of biolm. In any case, ablation would occur to a lesser extent by the pulsed VUV laser beam whose duty cycle was 106 at 7.87 eV and 107 at 10.5 eV (vs. near unity for the quasi-continuous synchrotron radiation).obvious mass spectral differences for the “pure” regions of each Fig. 3 10.5 eV laser LDPI-MS of coculture E. coli (tomato strain) and yeast multispecies membrane biofilm. “Mixed” indicates the region of overlap between the two species while other spectra correspond to “pure” regions of biofilms, as labeled in the inset photo of a typical coculture biofilm.were ionized at the higher photon energy (see ESI†). D. PCA of LDPI-MS for strain and species differentiation The results above demonstrated the capability of LDPI-MS to study monoculture biolms as well as multispecies coculture biolms (see also ESI†). The mass spectral data however, did not show much difference between the analyzed microbes. Additionally, the low mass region (m/z < 250) which is of great interest for metabolite study was too crowded to visually observe any obvious differences among strains. PCA was therefore used as a statistical tool to reduce the data size without signicant information loss and for better visualization of the data. The approach was optimized by rst applying PCA to the clearly distinct MS data of E. coli and yeast in a coculture biolm most to the separation of the mixed region from the two pure regions were m/z 54, 55, 59, 60, 67, 79, 84, 89, 105, 128 and 133. Similar analysis of the 7.87 eV data for the two pure regions indicated that peaks at m/z 55, 56, 67, 71, 78, 80, 88, 90, 104, 106, 127 and 129 contributed to the separation of the two strains along the rst principal component. PCA of 10.5 eV laser LDPI-MS data indicated that peaks at m/z 53, 65, 75, 77, 79, 81, 92, 94, 108, 216, 258 and 285 contributed to the separation of the two strains. However, it cannot be determined in these experiments which of these peaks are due to parent vs. fragment ions. PCA was also applied to the synchrotron LDPI-MS data, but no coherent results could be obtained from the analysis. One limitation of the synchrotron LDPI-MS data analysis was simply Fig. 6 Principal component 2 loadings plot for the entire 7.87 eV laser LDPI-MS data set of the coculture multistrain E. coli biofilm. species, it is not surprising that applying PCA to the data also readily distinguished the two species. The optimized PCA protocol was then applied to a more challenging system wherein a multistrain E. coli membrane biolm was examined. Both 7.87 eV and 10.5 eV laser LDPI-MS data of the genetically similar tomato and citrine E. coli strains were collected, with the former shown in Fig. 4 and the latter reported previously.25 Fig. 5a shows a plot of principal compo- nent 1 vs. principal component 2 for the 7.87 eV laser LDPI-MS data. While the mass spectra of the two strains in coculture samples showed only minor visual differences in the peak pattern, PCA treatment of the “pure” regions of the sample resulted in grouping the two strains separately. Furthermore, the mixed region was clearly distinguished from the two “pure” regions suggesting metabolic interactions resulted in altered physiologies. Fig. 5b is a scree plot9 which shows the variance of the entire data set with respect to the principal components. The scree plot indicates that 75% of the variance for the data is represented by its rst two principal components. In addi-Fig. 4 7.87 eV laser LDPI-MS of coculture E. coli (tomato and citrine) multistrain membrane biofilm. “Mixed” indicates the region of overlap between the two strains. tion, hierarchical cluster analysis was performed to the afore- mentioned 7.87 eV mass spectral data (see ESI†) that showed similar clustering of the strains to that obtained by PCA. Simi- larly, PCA was performed on the corresponding 10.5 eV MS data (see ESI†): it also distinguished the two “pure” regions of the biolm, with most variance along the rst principal component. Fig. 6 is the principal component 2 loadings plot, correlating which peaks in the mass spectra are most responsible for the differences seen by PCA between the mixed region and the two “pure” regions of the 7.87 eV laser LDPI-MS. In general, peaks with positive loadings along a given PC axis will show a higher relative intensity in samples with positive scores along the same PC axis (and negative loadings are similarly correlated with negative scores).12 Fig. 5a shows that the E. coli mixed region has a positive score while the two pure regions have negative scores along principal component 2. Therefore, positive peaks in Fig. 6 loadings plot for PC 2 correspond with the peaks that are more abundant in the mixed region (Fig. 5a). Fig. 6 and the biplot (not shown) indicate that the mass spectral peaks that contributed theFig. 5 (a) Principal component analysis of 7.87 eV laser LDPI-MS of the three different regions of a coculture multistrain E. coli (tomato and citrine strains) membrane biofilm. (b) Scree plot showing variance of the entire data set with respect to the principal components. biolms and can produce visually obvious boundaries between metabolism which can be cross-fed to the tomato scavenger strain as a substrate.8 The principal component 1 loadings plot of the 10.5 eV laser LDPI-MS showed that m/z 285 contributedthat too few spectra had been collected for a useful data matrix, a problem to be overcome by further experiments. IV. Discussion A. Distinguishing species, strains, and metabolic states by PCA of LDPI-MS data Different species as well as separate strains and interfacial regions of coculture biolms were distinguished by LDPI-MS detection of endogenous species followed by discrimination of the data by PCA. Additionally, hierarchical cluster analysis of the data showed that the strains and metabolic states of coculture biolms clustered separately, with the results corre- sponding with those for PCA (see ESI†). This is especially remarkable for the multistrain system because the strains differed from the wild type K-12 strain by no more than four gene deletions each out of approximately 2000 genes8 and visual examination of the mass spectral data showed barely any differences.25 PCA treatment of 7.87 eV LDPI-MS data separated the strains into three distinct groups (see Fig. 5a): two “pure” groups and a mixed region. Furthermore, the “pure” regions of the cocultures showed greater variance by PCA when analyzed by 7.87 eV photon energies than by 10.5 eV radiation. It is noteworthy that the E. coli mixed region (Fig. 5a) had a positive score along principal component 2 which resulted in a grouping outside of the pure tomato and citrine regions. These results indicate that the mixed region of the multi- strain system constituted a different collective metabolic prole than a simple summation of the pure metabolisms, possibly due to inter-strain interactions via metabolite or quorum sensing-like exchange. Use of multiple trials supports the claim of a distinct metabolic state in the mixed region that is not due to contaminants or other matrix effects. In particular, salt gradients are not expected within the multistrain biolms. Furthermore, salt effects are minimized by the detection of desorbed neutrals in LDPI-MS. LDPI-MS was also able to distinguish the bacteria from the yeast studied here both without and with the application of PCA, both signicant accomplishments. These results establish LDPI-MS as another MS tool available to probe local metabolic states and metabolite exchange in microbial systems.6 The principal component 2 loadings plot for the 7.87 eV laser LDPI-MS data (Fig. 6) revealed several mass spectral peaks that contributed the most to the separation of the strains. Exact mass measurements and/or tandem MS are required to conclusively assign peaks to any compound, so such assignments are generally not attempted here (see ESI†). Nevertheless, hypotheses are made regarding a few of the peaks contributing to the separation of the strains (Fig. 5 and 6). For example, the m/z 60 peak showed a higher relative intensity in the mixed region (Fig. 6) and contributed to the separation of the mixed region from the “pure” regions. This peak might have arisen from the molecular ion of acetate which is a common E. coli metabolic byproduct of glucose different species or strains. However, biolms grown on membranes were0.5–1mm thick and displayed topographical features that can hinder analysis by SIMS and MALDI-MS imaging. Prior LDPI-MS studies examined biolms grown using drip ow reactors,26 but coculture biolms grown by drip ow tend to lack clear boundaries between segregated species making analysis of pure cultures impossible without additional parallel experiments. Signicantly higher laser desorption energy was required to analyze intact membrane biolms compared to the blotted biolms. This might have resulted from the thickness of the membrane biolms and the effective absence of an imme- diately adjacent metal substrate which otherwise can undergo rapid heating to assist desorption.29,37 Synchrotron LDPI-MS was able to successfully analyze blots of biolms, which can also be analyzed by other MS imaging methods,16 but could not analyze intact membrane biolms. Laser LDPI-MS of intact membrane biolms showed overall higher signal intensity than from blotted samples. This could be due to the difference in desorption efficiency for the two samples. By contrast, blotting of the biolm resulted in loss of structural integrity and reduced sample volume that decreased the overall signal intensity. Furthermore, the efficiency of chemical transfer during blotting, including differential trans- fer, cannot be ruled out as a contributing factor.to the difference between the two E. coli strains. The m/z peak 285 could be attributed to the stearic acid parent ion, which is a common constituent of the E. coli cell walls and membranes.25 It is proposed that the two strains' metabolisms create locally different chemical microenvironments that inuence biolm composition. Both the synchrotron and laser LDPI-MS indicated that different classes of molecules can be targeted with different photon energies. New peaks appeared with increasing photon energy (see ESI†), as expected since many organic compounds have ionization energies in the 9–10 eV range and single photon ionization requires VUV photon energies in excess of a molecule's ionization energy.22,36 Comparison of the 7.87 and 10.5 eV data is consistent with the expectation that the lower photon energy selects a subset of low ionization energy ana- lytes while 10.5 eV is more inclusive, detecting across a wider range of analytes. These two VUV photon energies therefore give different spreads via PCA and they constitute an addi- tional experimental parameter to differentiate strains and species. Many metabolites which participate in cross- feeding or quorum sensing between interacting strains and species have relatively low molecular weights, so their molec- ular ions will be found in the low mass region. The lower mass region is of particular note in the 7.87 eV LDPI-MS of the E. coli cocultures. B. Comparison of sample preparation methods Thick, intact biolm samples grown on insulating membranes were successfully analyzed using laser LDPI-MS. Polycarbonate membranes provided a surface for robust growth of coculture displays submicron spatial resolution while atmospheric pres- both sufficiently robust that either could be implemented onC. Comparison of data analysis and instrumental methods Several studies have been reported which applied PCA to MALDI-MS data to obtain differentiation of species or strains as well as biomarker proling.14,15 However, it is noteworthy that differences between MALDI mass spectra due to ion suppres- sion, matrix peak intensities, and uctuating baseline might lead to unwarranted sample separation during PCA. By contrast, LDPI-MS can analyze the low mass region of coculture biolms free of ion suppression, without addition of any interfering matrix, and without any background subtraction. SIMS has the advantage of high spatial resolution and high sensitivity in the low mass region, both useful for studying metabolites and other small endogenous species. However, MS data for masses over m/z 300 can be difficult to observe in SIMS, limiting the application of SIMS to a reduced set of target molecules.12,13,18,38 The results presented here showed some LDPI-MS peaks up to m/z 1000, indicating its potential use for target species with a larger mass range than commonly avail- able to SIMS. Furthermore, LDPI-MS should induce less frag- mentation in molecular species than does SIMS, so LDPI-MS data should better represent the composition of the lower mass region. Nevertheless, the relatively low abundance of LDPI-MS peaks at m/z > 300 is attributed larger to fragmentation from energy transfer during laser desorption rather than single photon ionization.39,40 Relatively low species abundance and difficulty in desorbing some of the neutral species from intact biolm samples also likely contributed to the absence of higher mass peaks. A recent study described how LDPI-MS provides complementary chemical information to SIMS for the analysis of soil organic matter.40 Atmospheric pressure-based methods such as nano-desorp- tion electrospray ionization (nano-DESI)41,42 and laser ablation electrospray ionization (LAESI)19 allow for MS imaging of bio- lms without sample deformation or the loss of volatile species that is possible in vacuum. However, such electrospray ioniza- tion-based methods suffer from the same effects of ion suppression, ion–molecule reactions, and local uctuations in ionization efficiency that can complicate identication and quantication of species in SIMS.20 LDPI-MS seeks to overcome the shortcomings of direct ion formation by postionization of neutrals at pressures low enough to avoid subsequent ion- molecule reactions, albeit with the potential loss of volatile species prior to ionization. Unlike methods that detect direct ions, salt effects are minimized by the detection of desorbed neutrals in LDPI-MS. With its ability to analyze intact biological samples with reasonable spatial resolution, LDPI-MS can be used to study a range of biological samples such as infected medical implants, dental composites, or tooth samples. Spatial resolution here was limited to the mm-scale due to the relatively low signal that required collection of spectra from relatively large areas of the biolms. Prior work reported a spatial resolution of 50 mm is achievable with the laser LDPI-MS instrument as limited by mechanical vibrations and the low repetition rate of the Nd:YAG laser used to pump the 10.5 eV source.25 Synchrotron LDPI-MS has demonstrated a resolution of better than 5 mm on ancommercial instrumentation such as vacuum source-based MALDI-MS instruments, converting them into LDPI-MS instruments. The effect of varying delay time between the desorption and ionization processes was not explored in this study. However, prior work showed that sensitivity and resolution can be opti- mized in LDPI-MS at different mass ranges by varying delay times and other ion optical parameters, with more high mass peaks detected at longer delay times. A comparison of MS methods noted a lack of any single desorption/ionization technique that permits characterization of all components in crude oil samples.46 A similar statement can be made regarding the analysis of bio- lms, which combined with the data provided above supports the argument that LDPI-MS can make a valuable contribution to the analysis of these and other intact biological samples. Acknowledgements The authors acknowledge the assistance of Anjan Roy in developing the multivariate analysis and thank Gulsah Uygur for useful discussions. This work was supported by the National Institute of Biomedical Imaging and Bioengineering via grant EB006532. MA, TH, SL, and the Advanced Light Source were supported by the Director, Office of Energy Research, Office of Basic Energy Sciences, Chemical Sciences Division of the U.S. Department of Energy under contract no. DE-AC02-05CH11231. The contents of this manuscript are solely the responsibility of the authors and do not necessarily represent the official views of the National Institute of Biomedical Imaging and Bioengi- neering, the National Institutes of Health, or the Department of Energy. References 1 H. M. H. N. Bandara, J. Y. Y. Yau, R. M. Watt, L. J. Jin and L. P. Samaranayake, J. Med. Microbiol., 2009, 58, 1623–1631.sure-basedmethods typically allow spatial resolution in the tens of microns.18,20 The data presented here was acquired with two different LDPI-MS instruments. The synchrotron LDPI-MS instrument used tunable VUV for postionization that allowed exploration of the effect of increasing photon energies on mass spectral proles. The synchrotron LDPI-MS showed overall higher signal intensity and with appropriate sample preparation should allow for high resolution imaging of microbiological systems. Prior work on non-biolm samples demonstrated the utility of tuning across a wide and continuous range of photon energies.29,36 The 7.87 and 10.5 eV laser LDPI-MS allowed for in house data analysis and optimization as well as examination of electrically insulating samples such as membrane biolms. Furthermore, the 7.87 eV uorine laser and the 10.5 eV laser-based sources areorganic photoresist structure.29 LDPI-MS with femtosecond desorption lasers should permit micron scale resolution as well as depth proling of biological samples,43–45 the latter a strategy only denitively demonstrated on SIMS.20 By comparison, SIMS 2 C. J. Nobile, H. A. Schneider, J. E. Nett, D. C. Sheppard, S. G. Filler, D. R. Andes and A. P. Mitchell, Curr. Biol., 2008, 18, 1017–1024. 3 K. Brenner, L. You and F. H. Arnold, Trends Biotechnol., 2008, 26, 483–489. 4 J. Carlsson, Adv. Dental Res., 1997, 11, 75–80. 5 H. Bernstein and R. P. Carlson, Comput. Struct. Biotechnol. J., 2012, 3, e201210017. 27 M. T. M. Blaze, L. K. Takahashi, J. Zhou, M. Ahmed, G. L. Gasper, F. D. Pleticha and L. Hanley, Anal. Chem., 2011, 83, 4962–4969. 28 L. K. Takahashi, J. Zhou, K. R. Wilson, S. R. Leone and M. Ahmed, J. Phys. Chem. A, 2009, 113, 4035–4044. 29 O. Kostko, L. K. Takahashi and M. Ahmed, Chem.–Asian J., 2011, 6, 3066–3076. 30 P. A. Heimann, M. Koike, C. W. Hsu, D. Blank, X. M. Yang, 6 V. V. Phelan, W. T. Liu, K. Pogliano and P. C. Dorrestein, Nat. Chem. Biol., 2012, 8, 26–35. 7 B. Momeni, K. A. Brileya, M. W. Fields and W. Shou, eLife, 2013, 2, DOI: 10.7554/eLife.00230. 8 H. C. Bernstein, S. D. Paulson and R. P. Carlson, J. Biotechnol., 2012, 157, 159–166. 9 M. Otto, Chemometrics: Statistics and Computer Application in Analytical Chemistry, Wiley-VCH, New York, 2007. 10 B. J. Tyler, G. Rayal and D. G. Castner, Biomaterials, 2007, 28, 2412–2423. 11 B. Shrestha, J. M. Patt and A. Vertes, Anal. Chem., 2011, 83, 2947–2955. 12 D. J. Graham and D. G. Castner, Biointerph, 2012, 7, 49. 13 B. J. Tyler, S. Rangaranjan, J. Mo¨ller, A. Beumer and H. F. Arlinghaus, Appl. Surf. Sci., 2006, 252, 6712–6715. 14 P. Chen, Y. Lu and P. B. Harrington, Anal. Chem., 2008, 80, 1474–1481. 15 J. Qian, J. E. Cutler, R. B. Cole and Y. Cai, Anal. Bioanal. Chem., 2008, 392, 439–449. 16 J. D. Watrous and P. C. Dorrestein, Nat. Rev. Microbiol., 2011, 9, 683–694. 17 M. T. M. Blaze, B. Aydin, R. P. Carlson and L. Hanley, Analyst, 2012, 137, 5018–5025. 18 R. Trouillon, M. K. Passarelli, J. Wang, M. E. Kurczy and A. G. Ewing, Anal. Chem., 2012, 85, 522–542. 19 A. Vertes, V. Hitchins and K. S. Phillips, Anal. Chem., 2012, 84, 3858–3866. 20 J. S.FletcherandJ.C.Vickerman,Anal.Chem., 2013,85, 610–639. 21 L. Hanley and R. Zimmermann, Anal. Chem., 2009, 81, 4174– 4182. 22 A. Akhmetov, J. F. Moore, G. L. Gasper, P. J. Koin and L. Hanley, J. Mass Spectrom., 2010, 45, 137–145. 23 H. Sabbah, A. L. Morrow, A. E. Pomerantz and R. N. Zare, Energy Fuels, 2011, 25, 1597. 24 M. T. M. Blaze, A. Akhmetov, B. Aydin, P. D. Edirisinghe, G. Uygur and L. Hanley, Anal. Chem., 2012, 84, 9410–9415. 25 C. Bhardwaj, J. F. Moore, Y. Cui, G. L. Gasper, H. C. Bernstein, R. P. Carlson and L. Hanley, Anal. Bioanal. Chem., 2012, 405, 6969–6977. 26 G. L. Gasper, L. K. Takahashi, J. Zhou, M. Ahmed, J. F. Moore and L. Hanley, Anal. Chem., 2010, 82, 7472–7478.A. G. Suits, Y. T. Lee, M. Evans, C. Y. Ng, C. Flaim and H. A. Padmore, Rev. Sci. Instrum., 1997, 68, 1945–1951. 31 Y. Cui, J. F. Moore, S. Milasinovic, Y. Liu, R. J. Gordon and L. Hanley, Rev. Sci. Instrum., 2012, 83, 093702. 32 S. Stein, Anal. Chem., 2012, 84, 7274–7282. 33 E´. M. Timmins, S. A. Howell, B. K. Alsberg, W. C. Noble and R. Goodacre, J. Clin. Microbiol., 1998, 36, 367–374. 34 R. Goodacre, J. K. Heald and D. B. Kell, FEMS Microbiol. Lett., 1999, 176, 17–24. 35 G. L. Gasper, L. K. Takahashi, J. Zhou, M. Ahmed, J. F. Moore and L. Hanley, Nucl. Instrum. Methods Phys. Res., Sect. A, 2011, 649, 222–224. 36 L. K. Takahashi, J. Zhou, O. Kostko, A. Golan, S. R. Leone and M. Ahmed, J. Phys. Chem. A, 2011, 115, 3279– 3290. 37 L. Hanley, O. Kornienko, E. T. Ada, E. Fuoco and J. L. Trevor, J. Mass Spectrom., 1999, 34, 705–723. 38 B. Cillero-Pastor, G. Eijkel, A. Kiss, F. J. Blanco and R. M. A. Heeren, Anal. Chem., 2012, 84, 8909–8916. 39 G. Isaacman, K. R. Wilson, A. W. H. Chan, D. R. Worton, J. R. Kimmel, T. Nah, T. Hohaus, M. Gonin, J. H. Kroll, D. R. Worsnop and A. H. Goldstein, Anal. Chem., 2012, 84, 2335–2342. 40 S. Y. Liu, M. Kleber, L. K. Takahashi, P. Nico, M. Keiluweit and M. Ahmed, Anal. Chem., 2013, 85, 6100–6106. 41 J. Laskin, B. S. Heath, P. J. Roach, L. Cazares and O. J. Semmes, Anal. Chem., 2011, 84, 141–148. 42 C. M. Rath, T. Alexandrov, S. K. Higginbottom, J. Song, M. E. Milla, M. A. Fischbach, J. L. Sonnenburg and P. C. Dorrestein, Anal. Chem., 2012, 84, 9259–9267. 43 S. Milasinovic, Y. Liu, G. L. Gasper, Y. Zhao, J. L. Johnston, R. J. Gordon and L. Hanley, J. Vac. Sci. Technol., A, 2010, 28, 647–651. 44 S. Milasinovic, Y. Liu, C. Bhardwaj, M. T. M. Blaze, R. J. Gordon and L. Hanley, Anal. Chem., 2012, 84, 3945– 3951. 45 Y. Cui, C. Bhardwaj, S. Milasinovic, R. P. Carlson, R. J. Gordon and L. Hanley, ACS Appl. Mater. Interfaces, 2013, DOI: 10.1021/am4020633. 46 S. K. Panda, K.-J. Brockmann, T. Benter and W. Schrader, Rapid Commun. Mass Spectrom., 2011, 25, 2317–2326.