Surface analysis of bacterial systems using cryo‐X‐ray photoelectron spectroscopy

Surface analysis of biological systems using XPS often requires dehydration of the sample for it to be compatible with the ultrahigh vacuum of the spectrometer. However, if samples are frozen to liquid‐nitrogen temperature prior to and during analysis, water can be retained in the sample and the organization of the sample surface should be preserved to a higher degree than in desiccated samples. This article presents recent developments of cryo‐X‐ray photoelectron spectroscopy (cryo‐XPS) for analyses of hydrated biological samples at liquid nitrogen temperature. We describe experiments on bacterial cells, bacterial biofilms, and bacterial outer membrane vesicles using a variety of bacterial species. Differences and similarities in surface chemistry are monitored depending on growth in liquid culture, on culture plates, as well as in biofilms, and are discussed. Two data treatment methods providing decomposition of the C 1s spectra into lipid, polysaccharide, and peptide/peptidoglycan content are used and compared.


| INTRODUCTION
Bacterial interactions with their surrounding environment are mediated through their cell envelope, as well as other bacterial surface structures such as flagella and fimbriae. 1 These interactions are complex and rely on several factors including physicochemical interactions and specific molecular recognition mechanisms mediated through receptors. [2][3][4] To investigate these complex systems, analyses of the surface chemistry can be a useful tool in order to better understand and interpret processes involving the cell surface. For example, an increase in protein concentration at the surface of some bacteria, may indicate spore formation, 5 increased amounts of membrane proteins, or other types of adaptations. Differences in polysaccharide content may indicate changes in the lipopolysaccharide composition of the outer membrane in Gram-negative bacteria, 6 or be a result of altered expression of capsules or surface-bound extracellular substances. In this context, X-ray photoelectron spectroscopy (XPS) is an attractive analysis technique, as its surface sensitivity enables analyses of only the outermost part of the cell envelope of bacteria and/or surface appendages and coatings, 6 thereby enabling characterization of the cell envelope without influence from the cytoplasmic content of the bacterial cell. However, this is only valid if sample pretreatment methods keep the cells intact (i.e., does not rupture or beak them exposing their interior or causing building blocks to migrate in the cell-envelope structure).
Bacteria have several "lifestyles" (free-swimming planktonic cells, sessile cells, etc.) depending on factors such as stress levels, nutrient supply, and interactions with other (micro) organisms. Often these lifestyles can be correlated to metabolic differences. 7,8 Thus, planktonic cells cultivated in liquid broth may differ in surface composition compared with cells growing on a solid nutrient medium such as an nutrient agar plate or cells growing inside a biofilm. 9 Biofilms are coatings formed by bacterial cells embedded in a matrix of substances that they have secreted, so-called extracellular polymeric substances (EPS). 7 Bacteria inside biofilms display lower susceptibility to antibiotics and biocides as a result of several factors including changes in metabolic activity and that they are shielded from the environment by the EPS. 7,8 Bacterial cells, both in planktonic form and in biofilms, can release small parts of their cell membranes forming membrane vesicles (that also can be analyzed using XPS). These have been suggested to play a range of roles including cell-cell communication, nutrient uptake, scavenging of toxins, and as toxin-delivery vehicles. 10 In Gram-negative bacteria, these vesicles are called outer-membrane vesicles (OMVs) as they mainly originate from the outer membrane that buds and forms vesicles. 10,11 Also, Gram-positive bacteria form membrane vesicles, but the mechanism behind that process is less understood, although well documented. 11,12

| Short historic perspective on data treatment
Bacterial cells have been analyzed using XPS for several decades using freeze-dried or dehydrated samples to enable compatibility with highvacuum XPS systems. 13,14 These analyses have provided useful biochemical characterization of the cell envelope. The building blocks of the bacterial cell envelope mainly consist of different types of carbonbased substances, such as lipids, polysaccharides, proteins, and peptidoglycans, which are all contributing to the C 1s spectrum of a bacterial sample. In order to estimate the surface composition based on these substance groups, Rouxhet  In this paper, we describe how we have applied the method of cryo-XPS to different types of (hydrated and frozen) bacterial systems and discuss sample preparation and data treatment.  where EPS was removed, a colony from an agar plate was washed twice in 1 mL miliQ water and centrifuged at 4000 RPM for 5 minutes and the supernatant discarded. Thereafter, the bacterial pellet was resuspended in 100 μL miliQ water and the bacteria were immobilized on freshly cleaved mica as described above. AFM analysis was done in a multimode 8 using with ScanAsyst (Bruker) in peak force tapping mode in air.

| MATERIALS AND METHODS
Cryo-XPS analyses, data treatment, and sample pretreatment were done according to previously published procedures. 5,6,23 Analyses were performed on a Kratos Axis Ultra DLD electron spectrometer using a monochromatic Al Kα source operated at 150 W. Spectra were analyzed using a hybrid lens system with a magnetic lens. Angle between the incident X-ray photon beam and analyzer was 57.4 , photoelectron takeoff angle 90 , and collection angle of the analyzer ±15 . Maximum photoemission from sample was at 90 . The analysis area was 0.3 × 0.7 mm (slot). An analyzer pass energy of 160 and 20 eV was used for survey spectra and high-resolution spectra, respectively. The sample introduction chamber and analysis chamber were both pre-cooled to liquid nitrogen temperature before sample introduction and analysis, and were kept cold throughout the measurement. Sample charging during measurements was compensated using the built-in spectrometer charge-neutralizing system. The C 1s hydrocarbon peak at 285.0 eV was thereafter used for calibration of the binding energy scale. Chemical constituents giving rise to intensity in the C 1s peak were predicted using three mathematical components describing lipids, polysaccharides, and peptide (protein + peptidoglycan), here called spectral components, ie, independently from classical curve fitting of C 1s spectra using Gaussian-Lorentzian peak shapes. 6 For comparison, substance compositions were also calculated using Equations 2a to 2c after peak fitting using CasaXPS. Statistics were calculated using the Student's T test in Excel.
Adaptations in sample treatment for cryo-XPS measurements were made for each of the five sample groups: bacterial cells from liquid culture (1), culture plates (2 and 3), OMVs (4), and biofilms (5) as described below. All samples were measured at least as two biological replicates and all, except biofilms, were measured "fast-frozen", ie, were applied onto a sample holder at room temperature, frozen instantaneously on the sample stage in the sample introduction chamber of the spectrometer, and analyzed in hydrated frozen form, as previously described. 17,23 The sample was maintained at liquid nitrogen temperature throughout the analysis time, and the vacuum in the analysis chamber carefully was monitored (generally 3-5 × 10 −7 Pa).
Bacterial cells were analyzed both from liquid culture (1)  PBS was chosen to avoid introducing any C-containing substance that would "contaminate" the C 1s spectrum. (2) Bacteria analyzed directly from plates were collected from freshly cultivated plates using a loop and the collected biomass was directly applied onto the sample holder or (3) suspended in 2 mL PBS, vortexed, centrifuged, and the washed pellet used for XPS analyses. OMVs (4) were isolated from bacterial cultures using previously described methods. 21 On the day of analysis, the frozen sample was brought to the XPS in a sterile Petri dish on dry ice. A piece of the biofilm-covered glass was placed, still frozen, onto a sample holder precooled with liquid nitrogen. Thereafter, the sample holder was inserted into the analysis chamber as soon as possible (to limit water condensation on the cold metal). To take into account any spatial heterogeneities, three points were analyzed on each biofilm sample. After each analysis, sample and sample holder were sonicated in ethanol twice to clean, sterilize, and avoid sample contamination or carry-over.

| Planktonic cells
Bacterial cells grown in liquid culture or on a solid nutrient agar are sometimes both considered as planktonic cells, and may therefore display similar surface composition. 25 Consequently, XPS spectra of these two growth-forms may be similar, unless secretion of extracellular substances have taken place. To investigate this, we analyzed bacteria grown on culture plates and in liquid culture. We analyzed two different bacterial species: P. fluorescens (two reference strains DSM50090 and CIP69.13) and a Swedish river isolate (AH0123) of the species R. aquatilis. This river isolate was included as we observed that it produced very large quantities of EPS that was clearly visible by eye on culture plates as well as in liquid culture. This species is also known to have production of organic acids as well as extracellular polysaccharides of interest to industrial applications. [26][27][28] In cryo-SEM images, the secreted EPS is clearly observed as a layer in which the cells are dispersed ( Figure 1A,B). Using regular SEM, the EPS layer dehydrates, collapses, and forms peculiar features ( Figure 1C,D). AFM analyses showed that the EPS observed to cover the cells in the SEM analyses was removed to a large extent by washing, uncovering smooth rod-shaped bacteria with flagella ( Figure 1E,F).
Cryo-XPS analyses of planktonic cells of R. aquatilis showed differences in C 1s spectra between replicates of nonwashed bacteria, giving rise to large error bars in Figure 2. We hypothesize that these differences result from sample variations in the ratio of EPS to bacteria, which is very difficult to control during sample transfer from the  Table 2). As an example, the DSM50090 strain analyzed directly from plate gave a composition of 67 ± 3% peptide, 7 ± 2% lipid, and 26 ± 2% polysaccharide, whereas the rinsed showed 65 ± 4% peptide, 9 ± 2% lipid, and 26 ± 2% polysaccharide.

| Hydrophobicity
Hydrophobicity of the bacterial cell surface can be assumed to originate from nonpolar hydrocarbon functionalities in different types of macromolecules that build up the cell envelope. We previously compared data from cryo-XPS C 1s spectra to results from the traditional MATH assay (representing a partitioning of cells between an aqueous phase and an organic phase; Figure 3) for a range of planktonic bacterial cells grown on culture plates (E. coli lipopolysaccharide mutants). 21 This comparison showed that the ratio between nonpolar C (with bonds to other C and/or H) and total carbon in C 1s spectra follow the same trend as data obtained from the MATH assay for all strains except one (RN106). 21 Based on the results from the XPS analyses, the planktonic strains mentioned in this article increase in hydrophobicity in the order: River isolate R. aquatilis ≤ P. fluorescens (DSM50090) < ΔpJN2133 < P. fluorescens (CIP69.13) < ΔpelA = ΔwspF < PAO1 (Table 3). This comparison indicates that the first two strains were more hydrophilic than the previously described E. coli strains, whereas the PAO1 strains would be of similar hydrophobicity as the E. coli strains reported by Nakao et al. 21

| Bacterial biofilms
To study how differences in EPS influence XPS data of biofilms, we selected four different strains from P. aeruginosa, ie, the lab strain PAO1 (wild type, here labeled PAO1) and three mutants of PAO1 that were expected to have differences in EPS secretion. The ΔpelA mutant is deficient in the synthesis of an EPS polysaccharide involved in biofilm formation for nonmucoid strains (ie, does also not produce alginate) of P. aerugionsa. 19,29 The remaining two mutants were chosen as they express low (ΔpJN2133) or high (ΔwspF) levels of a signaling molecule (c-di-GMP) known to influence many pathways in the bacterial cell including EPS secretion as well as biofilm formation. 20 In order to differentiate between the signal from the bacterial cell and EPS, we compared biofilm data to data from planktonic cells grown on culture plates. A significant difference (p < 0.05) in peptide and lipid concentration was observed for three of the strains between the biofilms and the planktonic cells, ie, PAO1, ΔpelA, and ΔwspF ( Figure 4 and Table 2). However, the difference was less significant for the ΔpJN2133 strain has been reported to have low c-di-GMP and be a weak biofilm former, which could explain the closer resemblance between biofilms and planktonic cells for this strain. 20 The peptide component appears to be dominating at the biofilm surface whereas the amount of lipid-like substances is higher for cells grown on culture plates. Future studies, for example, using confocal microscopy and staining may explain these observed compositional changes at the surface of biofilms.

| Bacterial outer-membrane vesicles
Small membrane structures such as OMVs can be analyzed with cryo-XPS and their composition compared with that of their parent strain.
To this aim, we analyzed OMVs and parent bacteria from two E. coli strains. 21 In Figure 5, it can clearly be seen that the OMVs consist of relatively higher amounts of lipids (p < 0.05) for both strains. The

ΔflhD ΔwaaL
OMVs also showed reduced peptide content, which may represent a reduction in protein and/or peptidoglycan. This may either represent a reduction in outer membrane proteins caused by some sorting mechanisms during vesicle budding or that the budding leaves behind some of the peptidoglycan layer that is probed when the entire cell is analyzed. 6,10 It is worth noting, though, that peptidoglycan has been detected inside OMVs in some bacteria. 30 Previous work have shown that in order to separate the contribution of protein and peptidoglycan, XPS analyses can be combined with separate protein quantification using the Bradford assay. 30,31 These data suggest that cryo-XPS may be used as a tool to study compositional changes between the cell membrane and vesicles during vesicle genesis. F I G U R E 4 Average constituent concentration for the surface of biofilms grown for 12 and 15 h (average of six points distributed on two biological replicates (12 and 15 h), n = 6) as well as planktonic cells (from culture plates, two biological replicates, n = 2) for four P. aeruginosa strains. The prediction is based on fitting the C 1s peak using three spectral components representing peptide (protein and peptidoglycan), lipid, and polysaccharide 6 F I G U R E 5 Average constituent concentration based on C 1s spectra from E. coli bacteria and outer-membrane vesicles (OMVs). Significant difference between parent strain (p < 0.05) and OMV was found only for the lipid and polysaccharide component in the ΔflhD mutant and for the peptide and lipid components for the ΔflhD ΔwaaL double mutant shown above (n = 3 for both ΔflhD samples, n = 2 for both ΔflhD ΔwaaL samples). Data from model with spectral components described in the introduction, or the multivariate curve fitting approach that we developed and published in 2011. 5,6 To compare these two methods, we applied them side by side on the dataset presented in this article ( Figure 6). The trends confirm what we previously reported for a dataset originating from samples of Bacillus subtillis. 5,16 The peptide composition obtained from the spectral components is

| COMPARISON BETWEEN DATA-TREATMENT METHODS
generally higher than what is derived using Equation 2a to 2c ( Figure 6B,D). The opposite is observed for the lipid content, and the polysaccharide content is the most closely correlated between the two methods. However, Figure 6 shows that the relationship is far probably can account for some of the nonlinear correlation observed between these two data-treatment methods. The higher peptide content that is generally obtained using the spectral components may be due to carbon atoms with a neighboring N being more readily captured using the spectral components, than they would be from fitting the C 1s spectrum with single Gaussian-Lorentzian peaks at 285.0 and 286.5 eV and thereafter using Equations 2a to 2c. Due to the large overlap of the peptide and lipid components in both datatreatment method, it is not surprising that as the relative content of peptide increases, the lipid component decreases. The spectral peptide component has previously been shown to correlate well with the N content of the sample surface, and it was suggested that this could be used as an "independent" validation of the peptide composition at the surface. 6 Also, in this dataset, we observe that the content of N (represented by the N1s/C1s ratio) correlates to peptide content from both data-treatment methods. However, the peptide data seem to spread over a larger range giving rise to a more steep slope and the obtained R 2 was found to be higher for the method using spectral components (R 2 = 0.68 or 0.69 vs R 2 = 0.22 or 0.34 for planktonic and vesicle, or biofilm samples) than for Equations 2a to 2c ( Figure 6C,E).
The fact that these two data treatment methods assume slightly different "ideal" building blocks suggests that they are complementary but not directly comparable. In some bacterial systems, a number of substances may be present at the surface of the sample that are far from the ideal building blocks used in the two models, for example, extracellular DNA. These are not estimated in any of the models but may influence the obtained percentages from the two models differently. In this current dataset, we did not observe differences between the correlation data from biofilms and planktonic cells ( Figure 6).
However, this may well be the case for other systems. Thus, it may be advisable to use both data treatment methods in combination.

| CONCLUSIONS
Cryo-XPS can successfully be used to study the surface chemistry of a range of bacterial systems using a similar methodological approach.
We have shown that the C 1s spectrum can give a measure of the hydrophobicity of the bacterial cell surface, as well as information about surface composition of lipids, polysaccharides, and peptides (protein and/or peptidoglycan). From the methodological perspective, we observed that bacterial cells grown in liquid medium and on culture plates give rise to similar chemical composition, unless strains produce large amounts of EPS. Differences in surface composition were observed between planktonic cells, OMVs, and biofilms, showing that cryo-XPS can be a valuable tool to investigate, eg, production of EPS from bacterial cells, the composition of biofilm matrix in different types of biofilm, and membrane processes connected to OMV formation. Furthermore, we compared two published data treatment methods applicable to cryo-XPS data. We conclude that they are not correlated but give roughly similar estimates of surface composition.
In general, the obtained polysaccharide content corresponded fairly well between the data-treatment methods, but the peptide content was higher when using spectral components than when using Equations 2a to 2c. The opposite was found for the lipid content. Independent validation using the N 1s/C 1s ratio and the peptide component suggests a higher correlation with the spectral components compared with Equations 2a to 2c. However, the two methods are complementary and depending on research question one may be preferred over the other.
Cryo-XPS is a powerful method to study surface compositional changes of biological systems as it maintains water within the cell structure. This fast-frozen water plays an important role to preserve the structural organization of the bacterial sample enabling analyses at conditions more close to what is observed in ambient conditions. 5 Recently, near-ambient pressure-XPS has been applied to bacterial cells. 32 We hypothesize that this method and cryo-XPS should give comparable data. However, to investigate this, future work should be undertaken to show to what extent data from cryo-XPS and NAP-XPS overlap.