One light to rule them all: photodynamic inactivation of diverse Campylobacter

Abstract

Campylobacter, the leading cause of bacterial gastroenteritis worldwide, routinely contaminates UK poultry products, often destined for the kitchens of the general public who risk infection through improper handling. Alongside rising antimicrobial resistance (AMR), the sheer abundance of Campylobacter in the food chain is undermining current intervention strategies. A novel approach to reduce surface contamination on poultry products is violet-blue photodynamic inactivation (VB-PDI), which uses 405 nm visible light to activate endogenous photosensitizers and generate bactericidal intracellular levels of reactive oxygen species. We previously demonstrated VB-PDI is effective against reference strains of Campylobacter jejuni and determined the molecular underpinnings in detail. However, the cross-species efficacy, resistance potential and performance against AMR isolates remained uncharacterized. Here, we evaluate VB-PDI efficacy against 64 Campylobacter field isolates from UK poultry surveillance programmes, comprising three Campylobacter species of various antibiotic resistance profiles. All isolates were susceptible to VB-PDI to a similar level, demonstrating broad efficacy irrespective of species or phenotype. In vitro evolution assays did not result in any detectable increase in VB-PDI tolerance after fifteen repeat exposures which, together with the lack of any notably resistant isolates in our broad screen, indicates the risk of intrinsic or evolved resistance is minimal. Consequently, VB-PDI offers a promising and robust intervention strategy for reducing Campylobacter in the food chain, and here we discuss possible modes of implementation.

Introduction

Campylobacter is the most common agent of bacterial gastroenteritis worldwide, including the UK [1]. In 2024 and 2025 there were ~70,000 laboratory confirmed cases in the UK; a 55% rise from 2020 and the highest numbers on record [2]. It is generally accepted that, due to underreporting, the true number of infections is around tenfold higher, with a predicted £0.7 billion annual economic impact due to healthcare costs and productivity loss [3].

The primary pathogenic species responsible for human infections in 2024 were the thermophilic species Campylobacter jejuni (88.4%) and Campylobacter coli (11.2%), with a small number of cases attributed to Campylobacter fetus, Campylobacter lari and Campylobacter upsaliensis (0.4%) [2]. C. jejuni and C. coli are typically transmitted to humans via a zoonotic food-borne route and are primarily associated with poultry and pigs, respectively [4]. Approximately 70% of poultry flocks destined for the food chain, namely broiler chickens and turkeys, are colonized with Campylobacter in their intestinal tracts [5]. This prevalence is in part due to high-density poultry farming which promotes the spread of infection through flocks, although the ecology of Campylobacter in these environments is not well understood. Further downstream, slaughterhouse processing of broilers frequently operate in excess of 10,000 birds per hour. In some circumstances, this may lead to the surface of carcasses becoming contaminated with intestinal material through the process of evisceration [6]. Consequently, two-thirds of raw chicken sold by domestic retailers in the UK are contaminated, risking ingestion of viable Campylobacter following poor domestic hygiene and cooking practice, ultimately leading to human infections [3, 7, 8].

Once ingested, Campylobacter colonizes the mucosal layers of the small intestines, cecum and colon, with occasional invasion of host epithelial cells [9]. Infection typically results in moderate gastroenteritis, with the distinguishing symptom of haemorrhagic diarrhoea and resolves without specific treatment within a week [10]. However, up to 10% of cases are considered severe, requiring hospitalization and antibiotic treatment, particularly in high-risk groups, such as the immunocompromized [11]. Extra-intestinal spread is rare, however, post-infectious complications are not uncommon, most notably Guillain-Barré syndrome [12, 13].

Antimicrobial resistance (AMR) is a major public health concern, contributing to 4.95 million deaths in 2019 alone [14]. The sale of all antibiotics for use in food-producing animals in the UK has fallen dramatically since 2014, down by 77% in broilers and the use of high-priority critically important antimicrobials (HPCIAs), such as the fluoroquinolone (FQ) class antibiotic ciprofloxacin, has reduced by 99.8% to near zero across all poultry, with no HPCIAs used in broilers in 2024 [15, 16]. Despite this, we continue to see increasing rates of FQ resistance in UK poultry-associated Campylobacter, suggesting maintenance and even expansion of associated antimicrobial resistance genes (ARGs) despite an apparent lack of selective pressure [17–20]. Globally, Campylobacter MDR rates can exceed 90% in regions lacking effective antibiotic stewardship [21–23]. Despite this, global monitoring of AMR in Campylobacter remains limited, confounding efforts to manage its international spread.

Photodynamic inactivation (PDI) is an antimicrobial strategy that uses specific wavelengths of light to kill bacteria through the generation of intracellular reactive oxygen species (ROS) [24]. We previously demonstrated that C. jejuni is remarkably sensitive to PDI in the non-UV, violet-blue photodynamic inactivation (VB-PDI) at a maximum of 405 nm, which can be delivered to bacteria on a surface by a simple, low-power light-emitting diode (LED) [25]. We identified the molecular underpinnings of this sensitivity, ultimately attributed to two primary, compounding factors: first, the bacterium’s inherent sensitivity to ROS, in part due to a repertoire of essential, oxidation-liable metabolism components, such as the anaerobic-type iron-sulphur cluster enzyme pyruvate oxidoreductase (POR). Second, an exceptionally high abundance of the membrane-associated haem-precursor protoporphyrin IX, which alongside FMN and haem, acts as potent endogenous photosensitizers at 405 nm. These factors are collectively conserved across the Campylobacter genus owing to their indispensable central metabolic functions [25].

However, while the molecular mechanisms of VB-PDI have been largely elucidated, our previous study was limited to laboratory reference isolates of C. jejuni, leaving uncertainty in the broader efficacy of VB-PDI against diverse species and the variable phenotypes of field isolates that would ultimately represent the target of this promising intervention [26]. To address these concerns, we performed VB-PDI susceptibility assays on 64 UK poultry isolates of diverse genotypic and phenotypic composition. To understand the risk of resistance evolving against VB-PDI, we also performed laboratory evolution experiments by repeatedly exposing a field isolate of C. jejuni to consecutive, sub-lethal VB-PDI. Our results demonstrate the efficacy and sustainability of VB-PDI as a plausible intervention strategy to disrupt food chain transmission of Campylobacter to humans through surface decontamination.

Methods

Bioinformatic analyses

To assess the distribution of VB-PDI–associated genes, we scanned 6,658 Campylobacter genomes obtained from RefSeq, including contig, scaffold, chromosome and complete assemblies. Atypical genomes, metagenome-assembled genomes and large multi-isolate project datasets were excluded. Protein domain searches were performed using Pfam HMM profiles corresponding to domains implicated in VB-PDI function: PF01558.24 (Pyruvate ferredoxin/flavodoxin oxidoreductase), PF01855.24 (Pyruvate flavodoxin/ferredoxin oxidoreductase, thiamine diP-bdg), PF00034.26 (Cytochrome C), PF13442.11 (Cytochrome C oxidase, cbb3-type, subunit III), PF05201.20 (Glutamyl-tRNAGlu reductase, N-terminal domain), PF00762.24 (Ferrochelatase), PF01687.23 (Riboflavin kinase) and PF10087.14 (DUF2325). To verify the completeness and sensitivity of the searches, we included PF00521.26 (DNA gyrase/topoisomerase IV, subunit A) as a control domain expected to be present in all Campylobacter genomes. All searches were conducted using MacSyFinder v2 [27] (Table S1, available in the online version of this article).

Bacterial isolation and identification

All Campylobacter field isolates (n=64) used in this study were provided by the Animal and Plant Health Agency (APHA). These isolates were collected through various UK National Surveillance programmes between 2008 and 2024, to monitor AMR in Campylobacter species within poultry and on farm. A selection was made based on several criteria including: species, host species, sample type and clonal complex (CC) data. This included isolates for the Campylobacter species C. jejuni (67%, n=43/64), C. coli (22%, n=14/64) and C. lari (11%, n=7/64). Samples were either collected from poultry caecal samples (59%, n=38/64) or environmental sampling (41%, n=26/64), with isolates originating from broilers (44%, n=28/64), broiler farms (41%, n=26/64) and turkeys (16%, n=10/64). CC assignment was available for 50% (n=32/64) of isolates, predominantly the human-associated CC ST-21 (38%, n=24/64) (Table S2).

Campylobacter were isolated from samples using selective media, such as modified charcoal cefoperazone deoxycholate or Butzler agar, and then purified by subculture onto sheep blood agar (7% v/v) supplemented with 0.1% wt/v actidione. Campylobacter were identified to species level using MALDI-TOF MS. Confirmed Campylobacter isolates were stored in 10% v/v glycerol at −80 °C.

Bacterial culturing and media

Campylobacter isolates were grown in a MACS-VA500 incubator at 42 °C under microaerobic atmosphere conditions (10% v/v O₂; 10% v/v CO₂; 80% v/v N₂). C. jejuni NCTC 11168 was used throughout as a reference isolate and internal control. Campylobacter isolates were spread from glycerol stocks and grown on Columbia blood agar base (Sigma-Aldrich) containing 5% v/v defibrinated horse blood (E and O Laboratories Limited) and supplemented with vancomycin sulphate to 10 µg ml⁻¹ (CB agar). Revived cultures were incubated for 48 h before subculturing onto a fresh plate for ∼18 h before use in assays. c.f.u. counts were performed on Campylobacter blood-free selective agar (CS agar; Oxoid). Antibiotic disc diffusion assays were performed on Mueller-Hinton agar (Oxoid) supplemented with 5% v/v defibrinated horse blood and 20 mg l⁻¹ β-NAD (MHF agar).

VB-PDI susceptibility assay

VB light was generated by a Thorlabs M405L3 LED (Thorlabs) with a primary wavelength of 405 nm and maximum bandwidth of 20 nm. Power output in mW was calibrated using a power metre console (Thorlabs) 200 mm from the LED source. Dose energy was calculated in J cm⁻² from a function of power in W cm⁻² multiplied by exposure time in seconds. Freshly cultured cells (16–18 h-old CB plates) were diluted in PBS (Merck P4417) to an optical density at 600 nm of 0.1. 300 µl of the bacterial suspension was transferred to a cuvette and exposed for 20 min at either 0, 10, 20 or 30 mW. Post-exposure, the bacterial suspension was tenfold serial diluted in PBS to 10⁻⁷, and 5 µl of each dilution spot was plated in duplicate onto CS agar. Plates were incubated for 48 h and sample viability calculated in c.f.u. ml⁻¹. Reduction in viability was calculated as a log₁₀ reduction in c.f.u. ml⁻¹ relative to the 0 mW control. The technical limit of detection was 200 c.f.u. ml⁻¹; however, if undiluted spots had no identifiable colonies, they were recorded as 0. Temporal exposure assays were conducted in a similar way, except samples were taken from a continuously exposed sample at various timepoints, with power output calibrated to either 10 or 50 mW.

Antibiotic susceptibility testing

Antibiotic disc diffusion susceptibility testing was used for resistance screening of Campylobacter isolates following the EUCAST method for fastidious organisms. Briefly, freshly cultured cells (16–18 h-old CB plates) were diluted in PBS to an optical density at 600 nm of 0.1 and spread-plated using high-absorbency cotton swabs (Texwipe) onto MHF plates. The following antibiotic discs were stamped onto inoculated MHF agar using an antimicrobial susceptibility disc dispenser (Oxoid): Ciprofloxacin (5 µg); Tetracycline (30 µg); Erythromycin (15 µg); Gentamicin (10 µg); Chloramphenicol (30 µg) and Ertapenem (10 µg). Plates were incubated for 48 h and zones of inhibition were measured as a diameter in mm. In the case of ciprofloxacin, tetracycline and erythromycin, isolates were categorized as susceptible (S), susceptible at increased dose (I) or resistant (R) using EUCAST Clinical Breakpoints [28]. For the antibiotics where EUCAST breakpoints were unavailable (ertapenem, gentamicin and chloramphenicol), we defined breakpoints for the disc diffusion method by correlation to established broth microdilution MIC values [29] and by utilizing control isolates with broth microdilution-defined sensitive and resistant phenotypes. <R/≥S breakpoints in mm using the above method were: Ciprofloxacin, 26/50; Tetracycline, 30/30; Erythromycin, 20/20; Gentamicin, 20/20; Chloramphenicol, 15/15 and Ertapenem, 28/28.

Aerotolerance assay

Freshly cultured cells (16–18 h-old CB plates) were diluted in Mueller-Hinton broth (Oxoid) to an optical density at 600 nm of 0.1 and incubated aerobically at 37 °C in a 12-well culture plate with light shaking. Samples were taken immediately at 0 h and then at 12 and 24 h. Samples were serially diluted 10-fold in PBS to 10⁻⁶, and 5 µl of each dilution spot was plated in duplicate onto CB agar and incubated microaerobically for 24 h, after which colonies were counted. Aerotolerance was categorized into four classifications: hyper-aero-sensitive (HAS: no growth at 12 h); aero-sensitive (AS: growth at 12 h at 10⁻³ or lower dilutions; no growth at 24 h), aero-tolerant (AT: growth at 12 h at 10⁻⁴ or higher dilutions; no growth at 24 h); hyper-aero-tolerant (HAT: any growth at 24 h) [30, 31].

Laboratory evolution experiment

C. jejuni isolate 27 (C. jejuni; Table S3) was chosen as the ancestral isolate for the laboratory evolution experiment due to its average susceptibility to VB-PDI. The isolate was split into five concurrent lineages: two untreated control lineages (0 mW) and three VB-PDI-treated lineages (30 mW). For each exposure, cultures were diluted in PBS to an optical density at 600 nm of 0.1. Three hundred microlitres of the bacterial suspensions were transferred to a cuvette and exposed for 17 min at either 0 or 30 mW, corresponding to 0 and 30.6 J cm⁻², respectively. Following treatment, the 0 and 30 mW samples were diluted in PBS 1 : 100,000 and 1 : 100, respectively, to achieve an equivalent reduction in c.f.u. ml⁻¹. Two hundred microlitres of each sample was then spread onto CB agar and incubated for 48 h to yield around 200 distinct colonies. Following c.f.u. counts, all colonies were harvested and combined, then diluted in PBS to an optical density at 600 nm of 0.1 for the next round of exposure. The experiment continued for 15 rounds of exposure, with a glycerol stock of each lineage banked after each third exposure. Following completion, all banked lineage points and the ancestral isolate were revived and assayed by the standard VB-PDI susceptibility assay described above.

Statistical analyses

Unless otherwise stated, all significance values were calculated by a two-tailed, two-sample equal-variance t-test in Excel. P-values are reported as ns (≥0.05), * (<0.05), ** (<0.01) or *** (<0.001). For the temporal exposures, a linear regression model was performed in GraphPad Prism 10 using standard conditions (Table S4).

Results

Common zoonotic Campylobacter are VB-PDI susceptible

Following our previous work outlining the innate sensitivity of C. jejuni to VB-PDI, we hypothesized this technique would remain effective against all species of Campylobacter, as we assumed the endogenous photosensitizers responsible for sensitivity would be held in common [25]. To verify this, we conducted a bioinformatic screen of 6,658 Campylobacter genomes, representing 53 species, for the presence of genes required for photosensitizer biosynthesis and key ROS-liable enzymes responsible for sensitivity, finding them to indeed be ubiquitously present (Table S1). This suggests that all species of Campylobacter are likely to be vulnerable to VB-PDI.

Next, we performed VB-PDI susceptibility assays on 64 Campylobacter isolates derived from UK poultry farms, representing three species: C. jejuni (n=43), C. coli (n=14) and C. lari (n=7) (Fig. 1 and Table S3). Each isolate was exposed to three doses of 405 nm light (12, 24 and 36 J cm⁻²) and viability enumerated as a log₁₀ reduction in c.f.u. relative to an unexposed control. All isolates displayed dose-dependent susceptibility and collectively formed a normal distribution with a mean viability reduction of 3.86, 3.33 and 1.41 log₁₀ c.f.u. ml⁻¹ at 36, 24 and 12 J cm⁻², respectively (Fig. 1a). Susceptibility was not significantly different between the three species tested (Fig. 1b), and importantly, there was no individual isolate which was notably resistant, with at least a 2 log₁₀ c.f.u. ml⁻¹ reduction at 36 J cm⁻² in all cases. In fact, the isolates assayed here were significantly more sensitive to VB-PDI on average than C. jejuni NCTC 11168, the laboratory reference strain used in our previous study and as an internal control here (P=0.03, 0.002 and 0.003 at 12, 24 and 36 J cm⁻², respectively) (Fig. 1a) [25].

Fig. 1. VB-PDI susceptibility screening and cross-species efficacy: (a) Viability reduction of Campylobacter field isolates (n=64) following exposure to 12, 24 and 36 J cm⁻² 405 nm light. Black bars indicate median values. Red circles highlight the C. jejuni reference strain NCTC 11168. (b) No significant difference in VB-PDI susceptibility was detected between the three primary human enteritis-associated Campylobacter species: C. jejuni (n=43/64); C. coli (n=14/64); C. lari (n=7/64). Red dotted lines indicate the mean viability detection limit. ns=P≥0.05.

VB-PDI efficacy is not affected by aerotolerance

Given our previous work demonstrated the primary mechanism of VB-PDI inactivation of Campylobacter is through the generation of ROS, we reasoned that AT isolates, which can survive extended periods of aerobic stress, may show diminished susceptibility to VB-PDI through cross-resistance mechanisms. To assess this, we quantified aerotolerance to allocate our 64 isolates into four categories: HAS (n=14), AS (n=42), AT (n=5) and HAT (n=3) (Fig. 2a and Table S3). There was no significant difference between categories, indicating aerotolerance status does not influence the efficacy of VB-PDI.

Fig. 2. (a) VB-PDI susceptibility of isolates grouped by aerotolerance level: HAS (n=14); AS (n=42); AT (n=5) and HAT (n=3). Red dotted line indicates mean viability detection limit. ns=P≥0.05. (b) Temporal viability reduction of C. jejuni isolates exposed to low (10 mW; solid lines) and high (50 mW; dashed lines) power intensities. Loss of detectable viability is indicated by a black cross.

Inactivation by VB-PDI is linear and complete

To investigate how viability reduction proceeds over time, we exposed a subset of C. jejuni isolates representing the four quartiles of susceptibility, along with the reference isolate NCTC 11168, to low (10 mW) and high (50 mW) power intensities of VB-PDI (Fig. 2b). In all cases, reduction in c.f.u. increased linearly with exposure time (avg. R² 0.94, P<0.0001; Table S4). No viable bacteria were detected after 15 and 80 min at 50 and 10 mW, respectively, representing a reduction of >5 log₁₀ c.f.u. ml⁻¹ or 99.999%. We were unable to detect the presence of persister cells with our assay, which can sometimes occur with antibiotic exposure even at extremely high concentrations [32].

VB-PDI remains effective independent of AMR status

We screened all our isolates for susceptibility to the six veterinary and clinically relevant antibiotics recommended for harmonized monitoring of Campylobacter AMR in the EU: Ciprofloxacin (FQ); Tetracycline; Ertapenem (carbapenem); Erythromycin (macrolide); Gentamicin (aminoglycoside); Chloramphenicol (phenicol) [33] (Table S3). Ciprofloxacin and tetracycline resistance is very common in Campylobacter, with a corresponding resistance rate in our isolates of 61.5 and 55.4%, respectively, reflecting current trends [15]. No significant association was found between ciprofloxacin or tetracycline resistance and VB-PDI susceptibility (Fig. 3a, b). Ertapenem was introduced to monitoring programmes in 2022 and has seen a concerning expansion in the UK with rates as high as 53% in broiler-derived C. coli [15]. Resistance to ertapenem was present in 16.9% of our isolates, but likewise did not correlate with VB-PDI susceptibility (Fig. 3c). None of our isolates were resistant to chloramphenicol or gentamicin, and only one was resistant to erythromycin, preventing detailed analysis of these AMR phenotypes and possible co-resistance with VB-PDI. 18.5% of our isolates were MDR (resistance to ≥3 antimicrobial classes, including carbapenems), all of which, bar one, were ciprofloxacin, tetracycline, ertapenem resistant (Table S3). There was no significant difference in VB-PDI susceptibility between MDR, AMR or fully antibiotic susceptible isolates (Fig. S1). Taken together, these results demonstrate that the most common AMR phenotypes in Campylobacter do not provide cross-resistance to VB-PDI.

Fig. 3. VB-PDI remains effective independent of AMR status: VB-PDI susceptibility of isolates grouped by antibiotic susceptibility to (a) ciprofloxacin (susceptible, n=24; resistant, n=40), (b) tetracycline (susceptible, n=28; resistant, n=36) or (c) ertapenem (susceptible, n=53; resistant, n=11). Red dotted lines indicate the mean viability detection limit. S=susceptible. R=resistant. ns=P≥0.05.

The risk of evolved resistance to VB-PDI is minimal

For VB-PDI to be a sustainable intervention, the risk of resistance evolving must be low. We did not identify any isolates with a notably high innate resistance, with all 64 isolates reduced by more than 2 log₁₀ c.f.u. ml⁻¹ at 36 J cm⁻² (Fig. 1a). Using this value as a threshold, if such an innate phenotype exists in food chain populations of Campylobacter, it must be rare, occurring in <1.6% of isolates; a ‘low’ occurrence according to the EFSA scale [34]. To investigate whether we could induce resistance to VB-PDI through repeated selective pressure, we conducted a laboratory evolution experiment with C. jejuni field isolate 27, chosen for its average VB-PDI susceptibility profile (Table S3). We challenged three independent lineages of isolate 27 with 15 repeat exposures to VB-PDI at a dose resulting in a ~3 log₁₀ c.f.u. ml⁻¹ reduction, corresponding to a 0.1% survival rate of the population. Survivors were allowed to recover for 24 h before being pooled and exposed again. We included two replicate control lineages without VB-PDI exposure which were otherwise handled the same way and diluted following mock exposure to achieve an equivalent reduction in c.f.u. (Fig. S2). No significant change in susceptibility to VB-PDI was evident in any lineage compared to the ancestral isolate, suggesting resistance was unable to rapidly evolve in this isolate.

Discussion

The generation of ROS is the primary mechanism of VB-PDI and a key factor in Campylobacter’s innate susceptibility, as it is typically strictly microaerophilic and rapidly inactivated by aerobic stress. Indeed, some species have even been described as facultative anaerobes [35]. However, the emerging threat of AT thermophilic Campylobacter has been documented in recent years, with certain isolates able to withstand prolonged periods of aerobic stress and thus persist in the food chain, though to date the mechanisms underpinning this phenotype are poorly understood [36, 37]. Indeed, we identified 12.5% of our isolates as AT, all of which were C. coli isolated from broiler or turkey caeca, exemplifying their natural occurrence in poultry (Table S2-3). We hypothesized that AT isolates might manage oxidative stress more effectively and, therefore, demonstrate reduced susceptibility to VB-PDI. However, no such correlation was evident (Fig. 2a). This may reflect an inability of even highly AT Campylobacter to deploy sufficient ROS defences to survive the large burst of intracellular ROS generated by VB-PDI. This finding supports the sustainability of VB-PDI, as AT isolates are implicated in prolonged survival on poultry carcasses and, therefore, pose a greater risk of human infection [30]. However, the molecular basis of aerotolerance in Campylobacter is still unclear, and we acknowledge the low number of HAT isolates in our study (n=3) means further investigation is required if potential cross-resistance mechanisms are to be explored in detail.

As with aerotolerance, cross-resistance to VB-PDI by existing AMR phenotypes was plausible, as similar effects have been observed between certain disinfectants and antibiotics [38–40]. The ROS-based mechanism of VB-PDI does not have any direct overlap with the mechanism of action of antibiotics. However, there has been some evidence for the involvement of ROS in the bactericidal action of certain antibiotics, particularly the FQ, though this association remains controversial [41–44]. Nonetheless, we sought to clarify if there was any association between AMR status and VB-PDI susceptibility in Campylobacter. As anticipated, we found no such correlation, even in MDR isolates (Figs 3 and S1). We also considered that pre-exposure of Campylobacter to antibiotics, as will sometimes occur in poultry, may prime them to resist VB-PDI challenge through the induction of common stress resistance mechanisms. However, when we cultured a small selection of C. jejuni and C. coli isolates in the presence of sub-inhibitory ciprofloxacin, we found no significant reduction in efficacy, implying an absence of cross-resistance induction (Fig. 4). Taken together, this provides valuable evidence for the efficacy of VB-PDI as an alternative intervention which is unlikely to exacerbate the problem of AMR, as we have found no evidence for cross-resistance.

Fig. 4. VB-PDI susceptibility of ciprofloxacin-susceptible isolates cultured in the presence of no (0; control) or sub-inhibitory (0.0125 mg l⁻¹) ciprofloxacin. Cultures were challenged with 24 J cm⁻² and enumerated for log₁₀ reduction in viability compared to untreated controls.

Nonetheless, one of the main concerns with newly developed antimicrobials is the potential for resistance to rapidly develop, quickly rendering them ineffective. Unlike antibiotics which typically act on narrow targets, VB-PDI initiates a cascade of ROS-mediated damage with various cellular targets [25]. This multifaceted killing mechanism makes the rapid evolution of resistance unlikely, and we cannot envisage a plausible set of mutations which could confer high-level resistance. This is in stark contrast to ciprofloxacin and tetracycline resistance, which in Campylobacter can be readily acquired by a single point mutation in the chromosomally encoded gyrA gene, or horizontal acquisition of the tetO ARG, respectively, neither of which imposes a significant fitness cost, exemplified by their very common occurrence [45–47]. Persister sub-populations commonly observed following bactericidal antibiotic stress were not evident following VB-PDI for our chosen isolates, which may suggest a reduced probability of resistance development (Fig. 2b). While we acknowledge there has been some controversial evidence for a ‘viable but non-culturable’ state in Campylobacter which could be masking persister populations, we minimized this risk by promoting post-treatment recovery of sub-lethally damaged cells on enriching media under optimal atmospheric conditions. Despite extended incubations, we did not recover colonies beyond those surviving initial treatment and are therefore confident that the extent of oxidative damage inflicted by VB-PDI is sufficient to irreversibly inactivate the bacteria.

We considered whether VB-PDI is a stress that Campylobacter could naturally encounter in the environment when transitioning between hosts. According to the ASTM G173 reference spectrum, a 10 nm band centred at 405 nm contributes around 1% of total ground-level solar irradiance, equivalent to 1.13 mW cm⁻¹ [48]. Therefore, although an order of magnitude lower than the power intensities used in our study, with sufficient exposure time, direct sunlight could theoretically be bactericidal to Campylobacter. However, in contrast to aerotolerance which has presumably arisen in response to the extended aerobic stress likely encountered in the food chain environment, we did not identify any naturally occurring VB-PDI-resistant isolates in our screen. Nonetheless, we hypothesized that repeated exposure to inhibitory levels of VB-PDI could eventually promote evolution of resistance. We, therefore, performed a laboratory evolution experiment with repeated VB-PDI treatment over 15 consecutive exposures, yet found no evidence of evolved resistance (Fig. S2). Taken together, it appears the risk of resistance to VB-PDI evolving in Campylobacter through its use as an intervention in the food chain is unlikely, though further investigation is required to thoroughly discount this possibility.

Application of VB-PDI in the food chain

All Campylobacter assayed here could be reduced by at least 2 log₁₀ c.f.u. ml⁻¹ at a dose of 36 J cm⁻², and 44% of isolates were reduced by over 4 log₁₀ c.f.u. ml⁻¹ (Table S3). If a 2 log₁₀ c.f.u. g⁻¹ reduction could be achieved on contaminated broiler carcases, the public health risk of downstream infection could be reduced by as much as 90% [49]. The level of contamination on broiler carcases is variable, ranging from non-detectable to ‘highly contaminated’ if they exceed 3 log₁₀ c.f.u. g⁻¹ of skin [50]. Legislation requires abattoirs to measure contamination of broiler carcases using a process hygiene criteria: as of 2025, no more than 20% of sampled carcases can be highly contaminated [51]. When last assessed between 2012 and 2017, 26% (n=1755) of carcases leaving abattoirs in the UK were highly contaminated [52]. Therefore, effective interventions are needed across the primary production and processing stages of broiler meat if this target is to be met.

If VB-PDI could reduce contamination on carcases by 2 log₁₀ c.f.u. g⁻¹, then, based on the 2012–2017 data [52], the proportion of highly contaminated carcases would reduce from 26% to just 7%. At the retail level, the UK Food Standards Agency (FSA) set a target in 2015 for <7% of retail whole chicken being highly contaminated [53]. An independent survey of whole chicken in UK supermarkets in 2023 found that 10% were highly contaminated, demonstrating failure to attain the FSA target [54]. If a 2 log₁₀ c.f.u. g⁻¹ reduction were applied to these data, then only 2% of carcases would be highly contaminated. Either of these outcomes would represent a significant impact on the public health risk of Campylobacter infections. Furthermore, as VB-PDI uses low-power LEDs with a long lifespan and minimal maintenance, they provide a possible financial incentive to producers over other forms of processing interventions with higher operating costs. They may also represent an accessible option to low- and middle-income countries which often struggle with disease control in poor infrastructure, leading to high rates of infections [55].

However, although this study validates the broad efficacy and sustainability of VB-PDI in vitro, further work is needed before its implementation. The environmental and ecological complexity of viscera-contaminated poultry carcasses will differ considerably from the controlled, planktonic cultures used in our studies to date. For instance, Campylobacter persistence in abattoirs is often linked to biofilm formation which may shield cells from VB light penetration, particularly given these biofilms are frequently polymicrobial, and Campylobacter bound to poultry skin may have enhanced resilience and survival compared to those in liquid matrices [56, 57]. Trials of performance on naturally contaminated carcasses under industry-relevant settings are, therefore, essential to assess the efficacy of VB-PDI in the food chain.

Abbreviations

AMR

antimicrobial resistance

APHA

Animal and Plant Health Agency

ARG

antimicrobial resistance gene

AS

aero-sensitive

AT

aero-tolerant

CC

clonal complex

FQ

fluoroquinolone

FSA

Food Standards Agency

HAS

hyper-aero-sensitive

HAT

hyper-aero-tolerant

HPCIA

high-priority critically important antimicrobial

LED

light-emitting diode

PDI

photodynamic inactivation

POR

pyruvate oxidoreductase

ROS

reactive oxygen species

VB-PDI

violet-blue photodynamic inactivation