In vitro discrimination of wound‐associated bacteria by volatile compound profiling using selected ion flow tube‐mass spectrometry

To determine if bacterial species responsible for clinically relevant wound infection produce specific volatile profiles that would allow their speciation.


Introduction
Infection of the wound bed is detrimental to healing, resulting in a failure of the wound to move through the phases of healing in a timely manner (Bowler et al. 2001;Werth en et al. 2010). This has been demonstrated experimentally in animal models (Gurjala et al. 2011;Pastar et al. 2013). Colonization and infection of the wound bed develops and changes over time and is dependent on body location; early colonizers are likely to originate from the patient's own skin flora and include coagulasenegative staphylococci and Staphylococcus aureus. Subsequently, Gram-negative rods derived from the gastrointestinal, oral and genitourinary mucosa, and from the local environment may begin to invade (Bowler et al. 2001). As wounds mature it is known that anaerobes will play a greater role, as the wound microenvironment is changed by autogenic succession (Bowler et al. 2001;Daeschlein 2013). Early diagnosis and management of clinically relevant wound infection is essential to avoid complications. Clinicians currently rely on clinical symptoms, signs and nonspecific laboratory tests for indicators of infection for early diagnosis. Wound and blood cultures are often utilized, but can take several days to be reported and interpretation in the context of wound infection can be difficult (Macgregor et al. 2008; Barajas-Nava et al. 2013;Blokhuis-Arkes et al. 2015). Burn wounds present additional challenges; the release of high levels of inflammatory mediators result from an altered baseline metabolic profile and a systemic response to injury seen in the burns patient (Greenhalgh et al. 2007;Jeschke and Herndon 2014). Standard indicators of infection are therefore difficult to apply and experienced burns specialists are required to attempt to identify subtle signs of infection without definitive point of care testing available (Greenhalgh et al. 2007). The World Health Organization (2014) estimates that 265 000 deaths worldwide occur annually as a direct result of burn wound injuries. Sepsis and other complications resulting from infection are the number one cause of mortality in patients with severe burns (Bowler et al. 2001;Church et al. 2006;Schultz et al. 2013;World Health Organization 2014;Saaiq et al. 2015). Loss of the protective skin barrier, often over a large area of the skins surface, results in a high risk of infection, complicated by an induced state of immunosuppression (Church et al. 2006;Schultz et al. 2013;Saaiq et al. 2015). A recent study based in a major UK regional burn centre identified Staph. aureus as the most common organism cultured from burn wound swabs, although several other organisms were also identified including b-haemolytic streptococcus, Pseudomonas aeruginosa, Escherichia coli and coagulase-negative staphylococci (Alrawi et al. 2014). Similarly, these organisms were also among the most commonly isolated from burn wounds in a study carried out in an American hospital, with the exception of b-haemolytic streptococcus (DiMuzio et al. 2014). A fast, noninvasive method for diagnosing wound infection would benefit patients by allowing clinicians to detect infection at an early stage, enabling appropriate treatment to be promptly administered, reducing the risk of further complications. Furthermore, a UK government-commissioned review (O'Neill 2015) highlights the need for rapid diagnostic tests as a key strategy in the battle against the rise of antimicrobial resistance. In particular, to facilitate a reduction in the use of empiric therapy by allowing treatment to be optimized quickly and to promote a transition from the use of broad-spectrum agents to targeted antibiotics. Introduction of a rapid test for infection, carried out in the clinic, would also aid the reduction of overuse by allowing the implementation of a strategy, whereby antibiotic therapy is not administered to patients without a positive bedside test to prove it is required. This may be of particular importance in the burns clinic where symptoms of inflammation are common, but not necessarily associated with the presence of infection.
Production of volatile compounds (VCs) occurs as a result of normal bacterial metabolism; a diverse and complex array of bacterial VCs have been identified, with up to 80 different compounds reportedly produced by a single bacterial species (Schulz and Dickschat 2007;Thorn and Greenman 2012). Characteristic VCs of certain micro-organisms have long been recognized by microbiologists, for example, the distinctive odour of indole from E. coli. Previous work has shown that using selected ion flow tube-mass spectrometry (SIFT-MS) analysis, it is possible to detect and quantify the different types and concentrations of VCs produced by a range of bacterial species in vitro (Thorn et al. 2011) and that by employing appropriate statistical techniques the characteristic VC profiles can be used to discriminate between bacterial species. In addition, SIFT-MS analysis has been used to investigate bacterial volatiles emanating from a variety of sample types including; blood culture samples (Allardyce et al. 2006a,b;Scotter et al. 2006), serum (Spooner et al. 2009), urine (Storer et al. 2011) and breath (Dummer et al. 2013;Gilchrist et al. 2013), as well as microorganisms prepared in liquid culture medium (Thorn et al. 2011;Shestivska et al. 2015).
SIFT-MS technology was originally developed to study the production of molecules that occur in cold interstellar clouds and is described extensively in the literature Span el et al. 2006;Span el 2011, 2015). Reagent ions (H 3 O + , NO + , O 2 + ) generated in a gas ion discharge source are selected by a quadrupole mass filter and injected into a fast flowing carrier gas (usually helium) in the reaction flow tube. Here, the sample gas is introduced via a heated sample inlet and chemical ionization occurs resulting in the production of characteristic product ions. Downstream reagent and product ions are separated and counted by a further quadrupole mass spectrometer and electron multiplier detector system. Absolute concentrations of trace gases can be quantified based on the ratios of ion count rates and the reaction rate constants determined by detailed studies of the reaction of VCs with the three precursor ions.
The main aim of this study is to detect and discriminate wound-associated bacteria, grown in vitro, using SIFT-MS coupled with multivariate data analysis. Bacterial species were cultured in complex media; (tryptone soya broth (TSB)) and in simulated wound fluid (SWF) to simulate wound similar conditions (Werth en et al. 2010). This approach was developed to determine whether the main bacterial species associated with wound infection produce characteristic volatile profiles, which could be used to develop a diagnostic tool for speciation. Wound dressing removal is associated with increased pain which can cause considerable patient distress, this in itself can contribute to delayed wound healing (Price et al. 2008;Upton et al. 2012). The ability to assess a wound for infection without removing the dressing, or by direct assessment of discarded dressing material to minimize additional interference with the wound, would be an advantage. With this in mind, the final aim of this study was to investigate the detection of species-specific profiles from sterile dressing material inoculated with wound-associated bacteria. This will help determine if future volatile analysis of real patient-discarded dressing material has the potential to be used to diagnose bacterial species present in the wound bed.

Preparation and maintenance of bacterial cultures
Bacterial cultures were maintained on beads (Microbank; Pro Lab Diagnostics, Canada) at À80°C, resuscitated as required on tryptone soya agar (Oxoid, UK) and incubated aerobically at 37°C, with the exception of Streptococcus pyogenes cultures, which were resuscitated on blood agar (Oxoid

Liquid broth cultures for headspace analysis
Overnight plate cultures were used to prepare a test suspension of each micro-organism in 10 ml TSB (Oxoid, UK) adjusted to an OD 620 nm of 0Á20. One millilitre of the test suspension was used to inoculate 9 ml of sterile TSB, aseptically dispensed into sterile 40-ml glass vials with a PTFE screw cap containing a silicone septum (Supelco, UK), resulting in a final starting OD 620 nm of 0Á02 in each vial. For each bacterial strain used, three separate vials were inoculated, and on each day the experiment was performed, three vials containing 10 ml sterile TSB were incubated as controls. Test and control vials were incubated aerobically at 37°C and 200 rev min À1 (Stuart S150 Orbital Incubator; Bibby Scientific, UK) for 5 h.

Liquid cultures (simulated wound fluid) for headspace analysis
Overnight plate cultures were used to prepare a test suspension of each micro-organism in 10 ml diluent containing 0Á85% NaCl with 0Á1% peptone (Oxoid) adjusted to an OD 620 nm of 0Á20. One millilitre of the test suspension was used to inoculate 40-ml glass vials (as specified previously), containing 5 ml fetal bovine serum (FBS) (Sigma Aldrich) and 4 ml sterile diluent, resulting in a SWF containing FBS and 0Á85% NaCl with 0Á1% peptone at a ratio of 1 : 1 by volume (Werth en et al. 2010) inoculated with the organism of interest at a final starting OD 620 nm of 0Á02 in each test vial. Test vials were prepared in triplicate and incubated for 5 h prior to analysis as described above.

Wound dressing cultures; simulated wound fluid
Overnight plate cultures were used to prepare test suspensions in 10 ml diluent containing 0Á85% NaCl with 0Á1% peptone (Oxoid) adjusted to an OD 620 nm of 0Á20. Ten millilitres of SWF was used to moisten a 10 9 10 cm nonadherent sterile wound dressing, composed of a cellulose pad laminated with a film of polyethylene, housed on a wire support in a sterile incubation chamber. One millilitre of bacterial test suspension was used to inoculate the dressing. Inoculated dressings were incubated in the sealed incubation chambers for 5 h at 37°C. Throughout the incubation period the dressing samples were slowly perfused with sterile SWF at a rate of approximately 2 ml h À1 via a sterile needle, to simulate the conditions of a moderately exuding dressed wound (50 ml in 24 h: Thomas et al. 1996). Following the incubation period, dressing samples were aseptically removed from the incubation chambers and transferred to individual 1-l gas sampling bags (Tedlar â ; Supelco) which were sealed and filled with synthetic air (20% oxygen, 80% nitrogen). The gas sampling bags were incubated for a further 1 h at 37°C to allow the equilibration of the headspace gases.

SIFT-MS analysis
Following incubation, samples were analysed by SIFT-MS (Profile 3; Instrument Science Limited, UK) in 'Full Mass Scan' mode using the H 3 O + precursor ion; each sample was analysed three times using repeat scans of 100 s, over a spectrum range of 10-200 m/z, to generate a total of nine scans per bacterial strain from three independent replicate samples. Samples were introduced to the instrument by piercing the silicone septum of the sample vial with a sterile needle attached to the SIFT-MS direct sampling inlet, samples were vented with a second sterile needle attached to a 0Á2 lmol l À1 syringe filter (Ministat; Sartorious Stedim Biotech, Germany) to allow the free flow of headspace gases. Similarly, dressing samples were analysed by piercing the silicone septum of the gas sampling bag with a sterile needle attached to the direct sampling inlet of the SIFT-MS instrument.

Data analysis
Mean count rates were calculated for each detected massto-charge ratio (m/z) product ion across the mass spectrum for each bacterial strain included in the analysis (n = 3). Where a product ion was detected in both the headspace of a bacterial culture and the uninoculated control, t-tests were performed to compare the count rate detected. Only product ions detected at significant levels (P < 0Á05) from the bacterial cultures compared to background control levels were included for further analysis. The mean product ion count rate of background volatiles from the uninoculated culture media were then subtracted from those detected in bacterial culture headspace, to determine the total count rate attributed to production of VCs in the bacterial culture. Any corrected negative values were disregarded (indicating a reduction significantly below the background levels), as were reagent ion peaks. A threshold detection signal of 10 counts per second (cps) was applied to identify discriminant VC product ion data. Further selection was required to identify discriminate product ions for the SWF data. To achieve this product, ions were disregarded if the 10 cps threshold was met by only a single strain of any species included in the study.
Ward's method of Hierarchical clustering was applied to the resultant data set. This is a data mining technique, whereby an algorithm is used to score the dissimilarity between cases in the data set and a matrix constructed. The dissimilarity matrix is visualized through the production of a dendrogram, in which the branching of the clusters groups the cases with the least dissimilarity, and the distance between the nodes indicates the relationship between the clusters. The data were subsequently transformed using principal component analysis (PCA), visualized by constructing plots of the principal components scores for each data set. All data were analysed using Microsoft Excel 2013 (Microsoft Corporation, Redmond, WA) and IBM SPSS Statistics ver. 20.0 (IBM Corporation, Armonk, NY). A correction for multiple comparisons was not performed. Using the SIFT-MS kinetics library database, which has been compiled by the instrument manufacturers through detailed studies of the reactions between neutral analyte compounds and the three reagent ions, it is possible to make preliminary identifications of the VCs based on the detected SIFT-MS product ions.

Results
Headspace analysis of wound-associated bacteria cultured in complex growth medium Multivariate analysis of the headspace count rates (log 10 ) of selected mass spectra product ions using Ward's method of Hierarchical clustering analysis is visualized by the dendrogram in Fig. 1. This shows successful discrimination between Pr. mirabilis, E. coli, Staph. aureus and Staph. epidermidis based on the profile of selected VC product ions detected. The data were transformed using PCA to reduce the selected variables to principal components. The first two principal components account for 36Á3 and 21Á7% of the variability within the data respectively (Fig. 2a). Plotting the scores generated by PCA enables the complex data set to be represented in twodimensional space and clearly visualizes the association of each bacterial strain with the derived principal components and shows that the strains of Pr. mirabilis and E. coli, occupy discrete regions of the two-dimensional plot. Staphylococcus aureus and Staph. epidermidis appear to occupy overlapping regions on this plot, however, when the third component is employed (accounting for 10Á1% of the total variance) to construct a three-dimensional plot, these two species separate along the z-axis (not shown). Figure 2(b) shows the loading of the eigenvectors representing 51 selected product ions based on the selection criteria described in the materials and methods. It should be noted that when the initial t-tests were performed, negative product ion count rate values were disregarded since these represent the consumption of substrates in the growth medium by the bacterial species, which was not the focus of this study. The 51 selected product ions constitute the original variables for the PCA shown in Fig. 2a. This eigenvector plot (Fig. 2b) shows the contribution of each of the original variables; product ion count rates, to the derived first two principal components and indicates the influence of these variables in determining the position of each test strain. It is important to note that both the absence and presence of product ions dictates the spatial location of the given bacterial strain within the principal component plot. For example, when considering Ps. aeruginosa and Strep. pyogenes within the principal component, plot the absence of the majority of detected product ions and the presence of product ions m/z 18 and 28 are responsible for the plotted location of these species. This data set shows no significant differentiation between Ps. aeruginosa and Strep. pyogenes based on the profile of selected VC product ions Journal of Applied Microbiology 123, 233--245 © 2017 The Society for Applied Microbiology detected following 5 h of incubation in TSB. This was predominantly due to the absence of product ions detected in the culture headspace, rather than the presence of numerous similar volatile product ions. Although a total of 51 product ions were selected for the analysis of all bacterial species, many of the 51 product ions were absent from the headspace of both Ps. aeruginosa and Strep. pyogenes strains and this is likely to be the reason for the inability to discriminate between these species. To further assess whether these species could be differentiated based on volatile analysis when cultured in TSB, independent samples were cultured for 24 h and the headspace gases were analysed by SIFT-MS and multivariate analysis performed on the data as previously described. Following product ion selection, hierarchical cluster analysis implementing Ward's method resulted in the production of a dendrogram showing the two species separated into discreet clusters, indicating differences between headspace VC product ions of Ps. aeruginosa and Strep. pyogenes detected following 24 h of incubation (Fig. 3).

Headspace analysis of wound-associated bacteria cultured in a simulated wound fluid
Hierarchical clustering analysis of selected product ions implementing Ward's method and using squared Euclidean distance as a measure of resemblance enabled discrimination of Pr. mirabilis, E. coli, Strep. pyogenes and Ps. aeruginosa at the species level and staphylococci at the genus level as visualized by the dendrogram shown in Fig. 4. Principal component analysis was utilized to transform this data set, the first two principal components account for 34Á3 and 19Á7% of the total variance (Fig. 5a). It can be seen by plotting the PCA scores that strains of E. coli and staphylococci each occupy specific discrete regions of the two-dimensional plot. Figure 5(b) shows the loading plot of 26 original variables (product ions) detected by SIFT-MS headspace analysis on the first two principal components, generated by the PCA represented in Fig. 5a. In Fig. 5a, strains of Pr. mirabilis, Ps. aeruginosa and Strep. pyogenes appear to occupy overlapping regions, however, when plotted in three dimensions employing the third principal component (accounting for 9Á5% of the total variance), strains of these species are separated along the z-axis (data not shown).
Wound dressing cultures; simulated wound fluid Figure 6 shows the SIFT-MS product ions detected following analysis of the headspace of wound dressing material inoculated with Ps. aeruginosa (PA2), E. coli (EC2) and Staph. aureus (SA3). The plot includes only the VCs detected at levels significantly higher than the uninoculated controls, following background subtraction. High background count rates across the mass spectral range  were detected from the uninoculated dressing material. However, a range of VC product ions were detected above the background count rates from the inoculated dressing samples. Interestingly, these differed from those detected in the headspace of the same species cultured in liquid medium, in both the number and the mass-tocharge ratio of product ions detected. Only 12 product ions were detected in the headspace of the wound dressing material and of these only six were common to the product ions detected in the headspace of bacteria cultured in SWF.

Discussion
The main study aim was to identify wound-associated bacterial species specific volatile profiles in vitro. The mass spectral profile of VC product ions derived from the wound-associated bacterial species included in this study varies between species. This also extends to some strains from a single species, as reported in previous studies (Thorn et al. 2011;Shestivska et al. 2015). This study has demonstrated that the culture conditions namely composition of culture media and duration of incubation, influence the range of product ions and count rates of the resultant VC mass spectra. The range and relative concentrations of bacterial VCs under varying physicochemical conditions has been previously observed in other research studies (O'Hara and Mayhew 2009;Chippendale et al. 2011;Dolch et al. 2012a;Dolch et al. 2012b). The range and count rates of product ions detected are greater in TSB, a complex nutrient-rich commercially available culture media, compared to a SWF containing peptone, NaCl and FBS. A comparison of the discriminant product ions produced when bacteria were grown in either TSB (Fig. 2) or SWF (Fig. 5), shows that there are clear differences as well as similarities. For example, m/z 28 is clearly important for the discrimination of Ps. aeruginosa. The differing nutrient sources in the culture media almost certainly results in the upregulation of altered bacterial metabolic pathways, resulting in   the production of different volatile metabolites (Audrain et al. 2015).
The results from this study demonstrate that, it is possible to discriminate wound-associated bacterial species based on the detected VC product ion profiles, whether cultures are grown within complex medium (TSB) or under wound similar conditions (in SWF). After 5 h of incubation it was possible to discriminate several species based on the profiles using Ward's method of hierarchical clustering analysis; however, it was not possible to discriminate between strains of Ps. aeruginosa and Strep. pyogenes. Similarly, when the transformed data set was plotted as principal components, Ps. aeruginosa and Strep. pyogenes occupy closely adjacent regions on the twodimensional plot. This is the result of the absence of product ions detected in the headspace of either species under these conditions, visualized by comparing the two plots in Fig. 2. Increasing the incubation period to 24 h resulted in an increase in the number and count rates of SIFT-MS product ions detected from Ps. aeruginosa. This included high count rates for the m/z 18 product ion in all strains and the m/z 28 product ion in two of the four strains analysed. Hierarchical clustering analysis of the data obtained after 24 h of incubation resulted in successful discrimination between the stains of these two species, based on the clustering shown in the resulting dendrogram. It is possible to make a preliminary identification of the VCs which correspond with the detected SIFT-MS product ions, using predetermined reaction rate constants. When using the H 3 O + reagent ion, m/z 18 and 28 product ions indicate the presence of protonated ammonia and hydrogen cyanide respectively Turner et al. 2006). It has been reported elsewhere that both ammonia and hydrogen cyanide production are characteristic of Ps. aeruginosa (Nawaz et al. 1991;Gilchrist et al. 2011). A recent study (Neerincx et al. 2015) which investigated the production of these volatiles over time in a number of Ps. aeruginosa strains determined    that both compounds began to rise above detectable levels only in older cultures, as they began to enter stationary phase. This would explain the presumptive detection of these important volatiles only after 24 h of culture within this study. It is highly likely that production of these particular volatiles will be important for identification of Ps. aeruginosa in vivo, where growth conditions result in a slow growing biofilm state. Indeed, elevated hydrogen cyanide concentrations in nose-exhaled breath have been identified using SIFT-MS as a potential biomarker of Ps. aeruginosa infection in adult cystic fibrosis patients (Gilchrist et al. 2013).
The results of this study successfully demonstrated that it is possible to discriminate wound-associated bacterial species based on a profile of selected SIFT-MS product ions when cultured under wound similar conditions. Using Ward's method of hierarchical clustering analysis to determine strain relatedness, four of the six bacterial species were successfully discriminated, based on the clustering shown in the resulting dendrogram. Using PCA and constructing a plot of the first three principal component scores resulted in similar discrimination of the bacterial species investigated to the hierarchical clustering analysis, indicating that either of these techniques maybe suitable for modelling this data set. However, the two species of staphylococci could not be discriminated from each other. Further analysis of the headspace of both species of staphylococci following 24 h of incubation in SWF, also failed to successfully discriminate these species, in contrast to staphylococci cultured in complex culture media (TSB). Future work will assess whether it is possible to discriminate between species of staphylococci based on volatile headspace analysis when these species are grown as biofilm rather than in liquid culture. A biofilm model will simulate more closely the real wound environment (James et al. 2008). The altered growth conditions will likely affect the metabolic profile of the organism and result in a change in the volatile metabolites produced.
This study also investigated the detection of speciesspecific profiles from sterile dressing material inoculated with wound-associated bacteria; continuously perfused with a SWF. A range of SIFT-MS product ions with count rates significantly greater than background control levels were detected from the bacterial species analysed. In addition, the different species resulted in the detection of different mass spectra profiles (Fig. 6). This indicates that volatile analysis of wound dressing material to identify infecting microbes may be possible. Further research is required to validate this approach, and a pilot study is MS peak (mass to charge ratio) Figure 6 Mass charge peaks detected at levels significantly greater (P < 0Á05) than control levels (SWF only) from dressing material inoculated with Pseudomonas aeruginosa PA2 (pale grey), Escherichia coli EC2 (dark grey) or Staphylococcus aureus SA3 (black), incubated for 5 h at 37°C while continuously perfused with sterile SWF.
now being undertaken to identify volatile product ion spectra from the gaseous headspace of patient's discarded wound dressings. Ultimately, it will be essential to identify clinically relevant infection as well as the causative organism. However, the ability to identify microbial VC product ions among the background derived from the dressing, demonstrates that assessment of a dressed wound or discarded dressing may be a valid approach. As previously discussed, the profile of the product ions of the species grown in wound dressing material was different to those detected from the same species grown using the same medium (SWF) in liquid culture. This suggests that the growth conditions as well as the medium have an influence over the metabolic state of the organism and therefore the array of volatiles produced. The bacterial cultures were exposed to the same total volume of SWF over the 5-h incubation period for both culture conditions. The differing traits of bacterial biofilms compared to their planktonic counterparts are now well documented (Cooper et al. 2014) and the development of biofilm within the wound dressing material is likely to be a contributing factor to the differing profiles identified. Biofilms occurring in the wound bed in vivo are not adhered to a solid surface, such as the fibres of dressing material, but the semi-solid structures that make up the tissue (Cooper et al. 2014). Development of a wound biofilm model based on a continuously perfused semisolid matrix would be an advantage for future research. This would allow volatile profiles of the wound-associated organisms to be investigated under conditions that closely simulate wound infection and would further facilitate the development of this approach to wound diagnostics. It is highly likely that unique product ion profiles are produced when these bacterial species cause infection in real wounds, not only as a result of the specific growth conditions but also due to interactions between the bacterial species and the host immune response.
Using the predetermined reaction rate constants, presumptive identification of the neutral analyte compounds which correspond to the detected SIFT-MS product ions is possible. However, it is important to be aware that the reaction of a single selected precursor ion with similar VCs or compound fragments can result in the production of the same product ions. For example, utilizing the H 3 O + precursor ion as used within this study, the production of 89 m/z product ion can result from the protonation of a number of different analytes with a molecular weight of 88 Da, including pentanol (C 5 H 12 O), butyric acid (C 3 H 7 COOH) and putrescine (C 4 H 12 N 2 ). Furthermore, if a mixture known to contain more than one of these compounds were analysed using the H 3 O + precursor ion in a multi-ion monitoring (MIM) mode (not used within this study) it would only be possible to determine the total partial pressure of all compounds resulting in the common product ion . In some cases it may be possible to rule out the presence of certain compounds based on the nature of the sample being analysed. However, pentanol (C 5 H 12 O), butyric acid (C 3 H 7 COOH) and putrescine (C 4 H 12 N 2 ) could all be produced as a result of bacterial metabolism, as products of fermentation and decarboxylation reactions (Schulz and Dickschat 2007;Thorn and Greenman 2012;Audrain et al. 2015). Ambiguity in the identification of neutral analyte compounds can potentially be overcome by the analysis of the sample using multiple precursor ions, usually H 3 O + and NO + , as the reaction of compounds with different precursor ions can result in the production of different product ions. The compounds may then be identified and quantified using the appropriate precursor ion reaction . This study has demonstrated that using SIFT-MS and multivariate statistical analysis it is possible to discriminate woundassociated bacterial species based on the profile of selected SIFT-MS product ions when these species are grown in complex culture media, both in TSB and SWF. Furthermore, our proof-of-principle experiment has confirmed that it is possible to detect VC product ions derived from bacteria when cultured within wound dressing material using a SWF. Future investigations must focus on identifying the compounds present in the headspace of bacterial culture by utilizing GC/MS for preliminary exploratory analysis and multiple precursor ions for SIFT-MS studies using MIM mode to accurately quantify the volatiles of interest identified.