Thermophilic and non-thermophilic Campylobacter species Emits Distinct Volatile Organic Compounds in Different Culture Media and Growth Phases


 Bacteria emits a multitude of volatile organic compounds (VOCs) into the headspace as a mean of interactions with the environments, as well as intra- and interkingdom communication for survival and persistence in the nature and within their hosts. Campylobacter, which is often found in poultry and ruminants, has shown great persistence in aquatic environments, making it one of the world's most dangerous foodborne pathogens, killing thousands of people annually. In this study, the VOCs emitted by both thermophilic (C. jejuni, C. coli and C. lari) and non-thermophilic Campylobacter (C. fetus) of clinical concerns, impacted by nutrients composition (media) and growth phase were identified. Most thermophilic Campylobacter were shown to release volatile alcohols and ketones (1s,4R,7R,11R-1,3,4,7-Tetramethyltricyclo [5.3.1.0(4,11)] undec-2-en-8-one and Isophorone) during early stationary and stationary phases using active sampling with active charcoal adsorbent and GC-MS analysis. C. jejuni cultured in the Brain Heart Infusion had 1-Heptadecanol in its headspace gas, but not in Bolton Broth. The non-thermophilic C. fetus did not produce alcohols or ketones, but rather a variety of unidentified chemicals that will require further investigation in the future. Overall, PCA analysis revealed that the five Campylobacter strains studied created distinct volatilomes, allowing for future Campylobacter identification based on VOCs.


Introduction
The genus Campylobacter has been known for its challenging detection, isolation and identi cation in the laboratory, due to its physiological characteristics, such as non-glycolytic diet, low-oxygen requirement to grow (obligate microaerophile), and ability to enter in a coccoid form during non-favourable growth conditions, known as viable but non-culturable (VBNC) state 1,2,3 . However, despite the constraints of culturing this microorganism in vitro, Campylobacter spp. are ubiquitous in nature and colonise the intestinal tracts of many wild and domesticated animals 4,5 . These foodborne pathogens only require a small number of bacterial cells to cause human infection 5 , which is primarily transmitted through the consumption of poultry meat, particularly contaminated broiler chicken (Gallus gallus) 2,6 .
Campylobacter jejuni subsp. jejuni is the major species of the taxon and it is recognised as the leadingcause of human gastroenteritis worldwide 7,8,9 . The infection triggered by the pathogen is mainly characterised by mild and self-limiting symptoms, including watery to bloody diarrhoea, abdominal pain, fever, and vomiting 2,10 . However, severe sequalae may occur, such as Guillain-Barré syndrome, an autoimmune disease characterised by weakness of limbs and paralysis 2,10,11 . On the other hands, C. coli is the second most reported species of human campylobacteriosis. It is very similar to C. jejuni, and frequently isolated together with the C. jejuni from the contaminated poultry and faeces samples 12,13 . The third most clinically-signi cant Campylobacter species is C. lari. Although C. lari is infrequently associated with human infections, it has been isolated from patients with diarrhoea 14,12 . Another clinically important species is Campylobacter fetus subsp fetus that is an opportunistic human pathogen, that primarily affects patients with weakened immune system 15,12 .
The standard food safety testing is mainly based on culturing, polymerase chain reaction (PCR), and enzyme-linked immunosorbent assays (ELISA) approach that is time-consuming, labour-intensive, and prone to false negative results 3,16,17,18 . To meet the demand for an increasing number of food testing for global food trade, new method for Campylobacter detection in foods that is simpler, rapid, and highly sensitive is required urgently 10 . In this regards, a number of studies have demonstrated that bacteria produce distinctive volatiles organic compounds (VOCs) during culturing, create a single VOC-emission pattern that could potentially be developed into a unique metabolic biomarker for bacterial identi cation with high speci city, simplicity and in real-time 19,20,21,22 .
The bacterial VOCs are detected by means of gas chromatographic techniques, particularly the gas chromatography-mass spectrometry (GC-MS), which is generally preceded by a sample preparation 19,23 .
For instance, solid phase microextraction (SPME) has been largely used for sampling microbial VOCs 19,24 . However, owing to the lack of standardized methods, the nature of targeted sample matrix (i.e., the composition and the analytes' properties), the experimental requirements and constraints (i.e., costs and instrumental issues, etc.), select the most suitable sampling procedure to serve the purpose of VOC ngerprinting is challenging 25,23,26 . Nevertheless, sorption tube approach seems to be a suitable strategy that is widely adopted for gas sampling as it could retain a range of compatible VOCs at ambient temperature subjected to the interaction between the analytes and the sorbent material; the retained content could be desorbed for subsequent measurement 25,26 . This technique is simple, inexpensive, and more appropriate for thermally unstable compounds, when desorption via compatible solvent system is applied 27,28,24 .
Moreover, in contrast with the research attention that bacterial VOCs have drawn lately 19,29,30 , only few studies regarding to VOC-emissions of Campylobacter species have been published. The whole-cell fatty acids of Campylobacter fetus subsp. venerealis was investigated in a study that found hexadecanoic acid as the most signi cant volatile emitted by the bacterium 31 . Also, the pro le of VOCs emitted by C. jejuni in contaminated human stools and by C. jejuni and C. coli in chicken faeces was previously explored 32,33 . Additionally, the prevalence of alcohols were observed in broth inoculated with C. jejuni after 20 hours of incubation 34 . However, when subjected to the type of media, the growth phase, bacterial strain, and sampling setting, the VOCs emitted by a particular bacterial species could vary considerably 25,19,26,35 . Hence, this study attempted to investigate the variation of VOCs emitted by the four clinically signi cant species of Campylobacter (C. jejuni, C. coli, C. lari and C. fetus) in different culture media and growth phase. This research would be a cornerstone for further development of VOCbased sensing technology for detection of foodborne Campylobacter spp. in food.

Results
Identi cation of putative bacterial volatiles produced by Campylobacter Many peaks were detected during the GC-MS analysis of Campylobacter species. Figure 1 shows the similarity of the chromatograms of the bacterial samples and the blank media (negative control), demonstrating the complexity in identifying the putative biomarkers. Overall, the ve Campylobacter strains included in this study were found to produce a total of 38 to 68 identi ed VOCs during growth in either Bolton or Brain Heart Infusion broth; of these VOCs identi ed, only 15.7 to 41.3% remained after removing the VOCs found present in the blank samples (i.e. VOCs produced by the culture media) and 3.1 to 10.8% of these VOCs were detected in at least two biological replicates ( ltering criteria 2) ( Table 1). The VOCs that passed both the ltering criteria 1 and 2 were then labelled as the putative VOC biomarkers. a Identi ed VOC is compound that its identity has been con rmed by its Mass Spectrum (MS) and Retention Index (RI). b Criteria 1: VOC that is not found in the background (blank media). c Criteria 2: VOC that is detected in at least two biological replicates.
A total of 16 VOC biomarkers were identi ed to be associated with Campylobacter species (Figure 2; Supplementary data A1). These VOCs were found to belong to alcohol (3), ketone (2), ester (1), phenyl alkene (1), phenol (1) and unidenti ed (8) (Figure 1). Some of the putative biomarkers were labelled as "not identi ed" due to a low matching with the mass spectrum and retention index. However, these unidenti ed compounds were labelled according to their mass spectrum and calculated retention index for comparison between samples.
While both C. jejuni strains shared only 1 similar VOC biomarkers growing in Bolton broth, C. jejuni ATCC 29428 shared more VOC biomarkers with C. lari ATCC 35221: 3,7,, 1s,4R,7R, 3,4,11)] undec-2-en-8-one (ketone) and not identi ed 29 ( Figure 2). The VOC not identi ed 29 was in fact found to be emitted also by C. fetus ATCC 27374. Another ketone, isophorone was emitted by C. jejuni ATCC 33291 and C. coli 43478. It is worthy to note that C. coli did not share similarity with C. fetus and C. lari, in term of VOC biomarkers ( Figure 2).

Bacterial Growth
The growth curve of Campylobacter species cultured in both Bolton and BHI broth were determined by MPN enumeration approach at 0 h, 12 h, 24 h, and 48 h of incubation. The growth curve was then plotted ( Figure 3). All, but C. jejuni 33291 in BHI broth showed exponential growth in the rst 12 hours of incubation, indicating of the exponential growth phase; which was then followed by a signi cant deceleration in the growth rate at 12h to 24h and 48h, indicating the late exponential phase entering into early phase of stationary growth phase ( Figure 3). For C. jejuni 33291 cultivated in BHI broth, the exponential phase continued for another 12 hours before the growth rate started to slow down at 24 hours ( Figure 2d).

Volatile organic compounds emitted by Campylobacter species during different growth phases
Majority of the identi ed putative VOC-markers were detected after 12 hours of incubation, suggesting that the active production VOCs occur during the late exponential and early stationary phase of Campylobacter spp. under laboratory culturing conditions (Table 2). Generally, alcohol was detected in Campylobacter jejuni cultures at 24 hours and beyond; 1-Heptadecanol, which was only detected in C. jejuni ATCC 29428 BHI cultures was also detected after 12 hours of incubation (Table 2). Volatile ketones, 1s,4R,7R,11R-1, 3,4,11)] undec-2-en-8-one and Isophorone, were mainly detected in the stationary phase at 48 hours ( Table 2).

Discussion
The measurement results of bacterial VOCs depend greatly on the analytical methods adopted for ngerprinting, especially the gas sampling strategy. Although the charcoal-based matrix is a su ciently suitable adsorbent for a wide range of compounds 27 , one of the major drawbacks of using the charcoal adsorbent used in this study was the loss of volatiles, which could be caused by the solvent dilution factor, lesser a nity with polar compounds, and overload of the sorbent matrix 27,28,24 . These might explain the detection of less putative VOC-markers in this study as compared to other reported study that used different adsorbents 34 and technique such as thermal desorption 28,35 . However, the charcoal adsorbent has been proven to perform su ciently well to detect the most predominant bacterial volatiles that were commonly reported by other researchers, such as alcohols and ketones.
In this study, alcohols were the most commonly detected VOCs in Campylobacter spp., produced mainly in the early stationary and stationary growth phase ( Table 2). The detected alcohol-based VOCs included the fatty alcohols 3,7,11-trimethyl-3-dodecanol and 1-heptadecanol, as well as the non-cyclic alkene alcohol 1,8-nonadien-3-ol ( Figure 2). The probable pathways for bacterial synthesis of fatty alcohol and long-chain alcohols are through hydrogenation of methyl esters of fatty acids 36 , as well as by the β-or αoxidation of fatty acid by-products 37 . For instance, the production of fatty alcohols in E. coli begins with a thioesterase-mediated conversion of a fatty acyl-ACP (acyl carrier protein) to a free fatty acid. The free fatty acid is then converted to a fatty acyl-CoA by a fatty acyl-CoA synthase. The resulting fatty acyl-CoA can subsequently be metabolised via the β-oxidation route or reduced to its corresponding fatty alcohol by the generation of its corresponding fatty aldehyde in a NADPH-dependent fatty acyl-CoA reductasecatalysed process 38 . Additionally, 1,8-nonadien-3-ol has been reported to be emitted by Pseudomonas putida as an antimicrobial volatile against plant pathogens 39 , and commercial fermentation starters during food fermentation 40 . On the other hand, 1-heptadecanol was also found to be one of the antifungal volatiles emitted by Pseudomonas spp. isolated from canola and soybean 41 .
The two most predominantly isolated volatile ketones emitted by Campylobacter spp. in the current study were isophorone and 1s,4R,7R, 11R-1,3,4,7-tetramethyltricyclo [5.3.1.0(4,11)] undec-2-en-8-one ( Figure 2 and Table 2). The ketones were found to be emitted by the thermophilic Campylobacter (C. jejuni, C. coli and C. lari) in both Bolton broth and Brain Heart Infusion culture (Table 2). However, it is noteworthy to point out that ketones were detected generally after 12 hours of incubation at 24 and 48 hours, corresponding to the early phase of stationary and stationary growth phase of Campylobacter spp. in vitro (Table 2). Similarly, Reese et al. 35  The VOC pro ling of other clinically important Campylobacter species, such as C. coli, C. fetus subsp. fetus and C. lari, was a pioneering step towards the characterization of the volatilome of these pathogens. Isophorone and 3,5-bis(1,1-dimethylethyl)-phenol were discriminant volatiles emitted by C. coli., 3,7,11-trimethyl-3-dodecanol and 1s,4R,7R,11R-1, 3,4,7-tetramethyltricyclo [5.3.1.0(4,11)] undec-2-en-8-one were identi ed as biomarkers of C. lari. However, the pro le of C. fetus consisted of only notidenti ed compounds, that despite of their uncon rmed identity, these compounds were important to demonstrate the distinction of C. fetus among the other species. Nonetheless, future research should be able to widen the identi cation of the volatiles of these foodborne bacteria by employing more sensitive VOC-analytical methods.
The PCA analysis demonstrated clear distinguishable markers based on the VOCs pro le emitted by the various strains of Campylobacter spp. included in this work; as well as VOCs pro les that is growth phase dependent (Figure 4 and Figure 5). PCA has been used by numerous researchers to identify similarity and variation in the multivariable dataset, such as Milanowski and co-workers 51 that has demonstrated discrimination of salivary bacteria (Hafnia alvei, Pseudomonas luteola and Staphylococcus warneri) based on their volatiles' pro les; and discrimination of F. tularensis subsp. novicida, Bacillus anthracis Sterne and Bacillus anthracis Ames in the study by Reese and co-workers 35 .
The study of the variation in the VOCs emitted by the various clinically important Campylobacter spp. including Campylobacter jejuni subsp. jejuni in vitro via active sampling with an activated charcoal sorbent, revealed the culture-and growth phase-associated volatilome pro les of thermophilic and nonthermophilic Campylobacter spp. The ndings will provide an insight into the potential roles of bacterial VOCs in their growth and survival in various niches in environment and its host. The ndings may also provide a baseline to development of VOC-based detection method of the fastidious Campylobacter spp. Nevertheless, more in-depth studies are required to identify the unknown VOCs detected in this study. Also, it will be meaningful to investigate other in uencing factors, such as coculturing of Campylobacter spp. with other bacteria population to fully understand the mechanisms and role of bacterial VOCs.

Strains, Culture Media, and Laboratorial Maintenance
In this study, ve bacterial strains of Campylobacter genus were targeted: Campylobacter jejuni subsp. jejuni (ATCC 33291 and ATCC 29428), Campylobacter coli (ATCC 43478), Campylobacter fetus subsp. fetus (ATCC 27374), and Campylobacter lari (ATCC 35221). These strains were acquired from Microbiologics (Minnesota, USA). The strains of Campylobacter spp. were revived by inoculating the swab into sterile Bolton broth (BB; Oxoid, UK) supplemented with 5% of lysed horse blood. The lysed horse blood was prepared from fresh horse blood, which was purchased from the Faculty of Veterinary Medicine, Universiti Putra Malaysia. The fresh horse blood was subjected to multiple rounds of freezing and thawing until no separation into two layers was observed 53 . The inoculated media were incubated at 37 C for 48 h in a sealed universal bottle with 20% of headspace, creating the microaerophilic condition.
After 48 hours of incubation, the turbid broth was aliquoted into sterilized Eppendorf tubes (1.5 ml of volume) and spun down by centrifugation at 6,500 RPM (Rotation Per Minute) for 2 minutes. The supernatant was discarded and 0.5 ml of double strength Brain Heart Infusion broth (BHI; Himedia, India) and 0.5 ml glycerol were added into the vials, then it was thoroughly mixed by pipetting. The glycerol stocks were then divided into two parts to be maintained in the freezers at -20 C and -80 C.
The strains of Campylobacter were revived from glycerol stocks for the headspace-air sampling as follows: one hundred microliters were transferred from the glycerol vial to a sterile Bolton broth tube supplemented with 5% of lysed horse blood and incubated for 48 h at 37 C in a sealed universal bottle with 20% of headspace. After the growth, the inoculum was streaked onto modi ed Charcoal Cefoperazone Deoxycholate Agar (mCCDA; Oxoid, UK) in triplicate and incubated under microaerophilic atmosphere generated by Campygen 3.5 L (Oxoid, Basingstoke Hampshire, UK) at 37 C for 48 h. Five to ten colonies were then transferred to a test tube containing BB and gently mixed. After that, 1 ml of the bacterial suspension was pipetted into a 100 ml screw cap bottle (Duran, Mainz, Germany) containing 80 ml of culture media and leaving 20 ml of headspace volume for the headspace gas sampling and VOC detection.

Headspace-air Sampling And Experimental Conditions
Brie y the VOCs were collected by forcing the headspace gas in the bacteria culture bottle to pass through an ORBO activated coconut charcoal (100/50 mg) adsorbent tube (6 × 70 mm) (Merck, Darmstadt, Germany) at a xed rate by a pump. The adsorbent tube was then removed, sealed, and pending for solvent extraction and subsequent GC-MS analysis. However, preliminary tests were performed to adjust the volume of headspace-air that needed to pass through the adsorbent matrix to ascertain detectable signals for subsequent investigations. The apparatus used for the sampling was assembled inside a anaerobic jar with a lit candle to generate the microaerophilic condition. The sample bottle was equipped with a GL 45 screw cap with two ports (Duran, Mainz, Germany). Both inlet and outlet ports were connected by Tygon tubes (4.0 mm of internal diameter). The inlet tube was connected to a portable battery air pump DC-900 (Xilong, China) operated at a ow rate of approximately 2 L/min and the sorbent tube was inserted in the outlet tube on the other hand. The duration of the headspace-air sampling was according to the length of each incubation period.
The C. jejuni strains (ATCC 29428 and 33291) were inoculated individually in two different high-protein media: BB supplemented with 5% of lysed horse blood and double strength BHI; with initial concentration of approximately 10 4 MPN/ml. All the samples were incubated at 37 C and the headspace air was sampled during three time-points: 12 h, 24 h, and 48 h. The others Campylobacter species: C. coli, C. fetus subs. fetus and C. lari were inoculated only in BB supplemented with 5% of lysed horse blood and with initial bacterial concentration of approximately 10 5 , 10 3 , and10 3 MPN/ml, respectively. They were also incubated at 37 C with continuous active headspace-air sampling for 12 h, 24 h, and 48 h.

Solvent Desorption And Gc-ms Analysis
After the headspace air sampling, the sorbent tube was removed, and the sorbent content was transferred into a 1.5 ml vial. Then, 1 ml of carbon disul de (Merck, Darmstadt, Germany) was added into the vial for desorption of the VOCs. The vial was thoroughly shaken using a vortex mixer, and the extract was further transferred into another clean GC vial before sending for immediate GC-MS analysis 27,28 .
The carbon disul de extract was analysed under splitless mode using a GC-MS (Brand: Shimadzu; Model: QP2010 ULTRA 7890A GC/MS Agilent 5975) based on the following settings: a capillary column RTX-5 MS (length: 30 m, internal diameter: 0.25 mm, and lm thickness: 0.25 µm) was used; the temperature of the injection port was at 230°C; the oven was held at 40°C for 1 min, then the temperature was raised to 280°C at a rate of 6°C min−1, and the nal temperature was held for 1 min; a solvent cut of 3 min was performed and the length of the analysis was 25 min per sample.
For the samples containing BHI broth, the temperature program was further re ned as the preliminary results indicated the need to improve the data collection. In this case, the temperature of the injection port was increase to 250°C; similarly, the oven was held at 40°C for 1 min, then the temperature was raised to 250°C at a rate of 6°C min−1, and the nal temperature was held for 11 minutes; leading to a 46 min run time. For both settings, a full scan mode was used for the mass spectrometry analysis.
For referencing purposes, a series of n-alkanes standard (Merck, Darmstadt, Germany) was run under the exact same GC-MS operating conditions; the retention index calculated from the alkane runs is rather system independent and thus reproducible 54 .

Enumeration Of Bacterial Count
Before the incubation periods (12, 24, and 48 hours) and after the headspace air sampling 1 ml of the enrichment broth was collected from the sample for three-tube Most Probable Number (MPN) analysis. The aliquot was serially diluted using fresh BB with 5% of lysed horse blood in a 10-fold dilution until maximum 10 −9 , or depending on the expected growth phase, and incubated in triplicate at 37 C for 48 h under microaerophilic condition 53 .
The tubes were then checked for turbidity. The tubes that became turbid were plated on mCCDA and incubated for 48 h at 37 C under microaerophilic condition to con rm the presence of Campylobacter in the tubes. The concentration of bacteria cells in the tube (MPN/ml) was then calculated with a MPN calculator developed by the US governmental environmental protection agency (EPA) 55 . Also, the growth curve of Campylobacter species (MPN/ml) for all replicates throughout the experiment was then plotted.

Data Analysis
The VOCs were identi ed by taking both mass spectra and retention index into account. The mass spectrum of each compound was matched against the reference spectrum available in NIST mass spectral library, only those with high degree of similarity (>70%) were further considered 56 . In addition to library mapping, GC retention index (RI) provide complementary information to aid the identi cation of VOCs. The normalized retention time (non-isothermal Kovats retention index) was calculated according to the formula of Van Den Dool & Kratz 57 . For further identi cation, the calculated RIs were compared with the reference RIs of same kind of stationary phase; those with 5% or more of relative standard deviation were rejected 35 .
Moreover, two criteria were adopted here to decide the putative volatiles, that are most likely produced by the microorganism, with relative consistency (see Results, Table 1). First criterion, the compound is not present in the background (blank media) at the same timepoint (incubation period) of the bacterial samples, and second criterion, the compound was detected in at least two samples throughout the repetitions. Only ltered VOCs were considered for statistical comparison of bacterial strains and growth phases. The relative peak area of these selected compounds was subjected to PCA to explore the similarities/dissimilarities in VOCs patterns that might associated with the bacterial strains, the production of VOC during different growth phases and culture media. The calculation was performed using SAS JMP 15 software.  Figure 1 Comparison between chromatograms. Note: (a) C. jejuni 29428 after 24 hours of incubation and sampling, (b) negative control (blank Bolton Broth), and (c) overlap of chromatograms with the negative control peaks coloured in pink; the red arrows in part (c) represent the peaks that were not detected in the blank media.   (e) C. coli; (f) C. fetus; (g) C. lari; each point represents a bacterial sample with detected VOC-markers; non-coloured shape represents the blank media (negative control).

Figure 5
PCA score plots of all bacterial samples divided by each incubation period (12h, 24h, and 48h). Note: each point represents a bacterial sample with its detected VOC-markers; the strains absent in one of intervals means that no putative VOC-markers were identi ed in these strains in that interval.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryA1.pdf