Integrative Assessment of Reduced Listeria monocytogenes Susceptibility to Benzalkonium Chloride in Produce Processing Environments

ABSTRACT

For decades, quaternary ammonium compounds (QAC)-based sanitizers have been broadly used in food processing environments to control foodborne pathogens such as Listeria monocytogenes. Still, there is a lack of consensus on the likelihood and implication of reduced Listeria susceptibility to benzalkonium chloride (BC) that may emerge due to sublethal exposure to the sanitizers in food processing environments. With a focus on fresh produce processing, we attempted to fill multiple data and evidence gaps surrounding the debate. We determined a strong correlation between tolerance phenotypes and known genetic determinants of BC tolerance with an extensive set of fresh produce isolates. We assessed BC selection on L. monocytogenes through a large-scale and source-structured genomic survey of 25,083 publicly available L. monocytogenes genomes from diverse sources in the United States. With the consideration of processing environment constraints, we monitored the temporal onset and duration of adaptive BC tolerance in both tolerant and sensitive isolates. Finally, we examined residual BC concentrations throughout a fresh produce processing facility at different time points during daily operation. While genomic evidence supports elevated BC selection and the recommendation for sanitizer rotation in the general context of food processing environments, it also suggests a marked variation in the occurrence and potential impact of the selection among different commodities and sectors. For the processing of fresh fruits and vegetables, we conclude that properly sanitized and cleaned facilities are less affected by BC selection and unlikely to provide conditions that are conducive for the emergence of adaptive BC tolerance in L. monocytogenes.

IMPORTANCE Our study demonstrates an integrative approach to improve food safety assessment and control strategies in food processing environments through the collective leveraging of genomic surveys, laboratory assays, and processing facility sampling. In the example of assessing reduced Listeria susceptibility to a widely used sanitizer, this approach yielded multifaceted evidence that incorporates population genetic signals, experimental findings, and real-world constraints to help address a lasting debate of policy and practical importance.

INTRODUCTION

In the processing environments of ready-to-eat (RTE) foods, such as fruits and vegetables, Listeria monocytogenes is a food safety hazard of heightened concern. As a soil saprotroph, L. monocytogenes is widely distributed in natural environments where contamination of food and other materials often occurs. The tainted foodstuffs and materials can lead to reoccurring introduction and occasional persistence of the bacteria on food processing equipment and premises (1, 2). As an intracellular pathogen, L. monocytogenes can cause serious infections via contaminated foods, particularly threatening at-risk groups, including pregnant women, newborns, elderly people, and individuals with weakened immune systems. In the United States, there have been at least nine reported L. monocytogenes outbreaks linked to fresh produce since 2008, all of which were attributed to contamination in the processing or packing environment (3).

Food processing facilities rely on biocides to control L. monocytogenes and other harmful microorganisms. Quaternary ammonium compounds (QACs) such as benzalkonium chloride (BC) are among the most commonly used disinfectants or sanitizers in food processing environments. First registered with the U.S. Environmental Protection Agency (EPA) in 1947 (4), BC-containing products have been increasingly applied in a wide range of industrial, agricultural, clinical, and domestic settings (5). The legislation limits of BC concentrations for food producing plants, equipment, and utensils range between 200 and 400 ppm (4). Such widespread and prolonged use of BC-based disinfectants inevitably causes sublethal contact with various bacteria and consequent emergence of less-susceptible strains (5). Historical concerns over the decrease in microbial susceptibility to BC can be traced to the 1960s when clinical (6) and laboratory (7) observations of tolerant bacteria were reported. Since then, BC tolerance has been documented in a variety of microorganisms, including common foodborne pathogens such as Salmonella, Escherichia coli O157, Campylobacter, and L. monocytogenes (5).

Reduced susceptibility to BC may result from either individual or combined effects of (i) selection for inherently more tolerant bacteria (BC selection) and (ii) development of additional tolerance after adaptation to sublethal levels of BC (adaptive BC tolerance). In the example of L. monocytogenes in food processing environments, there is still a lack of consensus on the likelihood and implication of reduced BC susceptibility that may develop upon sublethal exposure to the sanitizer. Population-level evidence for BC selection linked to food processing is scarce. While multiple studies have shown the development of adaptive BC tolerance in L. monocytogenes under laboratory conditions (8–10), which can also cause reduced susceptibility to other disinfectants and antibiotics (11–13), the increased MICs of adapted L. monocytogenes strains were found to be still below the commonly applied concentration of a QAC-based disinfectant (14).

Rotational use of different sanitizers has been proposed to prevent the emergence of less-susceptible L. monocytogenes. Sanitizer rotation was recommended in the 2017 FDA draft guidance on L. monocytogenes control in RTE foods (15) and the 2016 document FSIS Sanitation Concerns in RTE Processing Environments (16). Despite the guidelines, the realistic impact of such tolerance on food industry remains debated. Arguments against the concerns over tolerance selection and/or adaptive tolerance development have been focused on their real-world likelihood and the practical efficacy of sanitizers in processing environments (17, 18).

Several factors have contributed to the debate. First, L. monocytogenes isolates are often classified as either tolerant (resistant) or sensitive (susceptible) to BC. (In this study, we follow the proposition by Bland et al. [19] to use the term “tolerance” instead of “resistance” to describe the reduced BC susceptibility of L. monocytogenes to QAC sanitizers; it remains debatable which term is more appropriate in the contexts of sanitizers [12] versus antibiotics [20].) The classification typically relied on phenotypic assays using isolates from individual sources, creating variations among studies. According to the MICs of 200 isolates from Norway and other European countries related to fish or fish processing environments of Norwegian origin, BC-tolerant isolates showing proton motive force-dependent efflux activity were defined as those having an MIC between 4 and 7 μg/mL (21). In another French survey of a similar scale on isolates from smoked fish production, BC-tolerant isolates had MICs from 7 to 15 μg/mL, with the tolerance mediated by a plasmid-independent efflux pump (two isolates) or a different mechanism (majority of tolerant isolates) (22). However, later discoveries of mobile genetic element-conferred, efflux pump-mediated BC tolerance, including plasmid-borne bcrABC (23), transposon-borne qacH (24), and genomic island-borne emrE (25), raised the MICs of tolerant isolates above 30 μg/mL. Second, while the development of adaptive BC tolerance has been experimentally observed in L. monocytogenes, the observations were mostly derived from adaptation conditions that did not realistically mimic BC exposure scenarios in processing environments. For example, laboratory tolerance was often induced through iterations of subculturing under otherwise optimal growth conditions, including nutrient broth and non-refrigeration temperatures (8–10, 12–14, 26). Temporal patterns of tolerance development, such as how soon it emerges and how long it lasts, and the possibility of the development under typical environmental constraints during food processing remain poorly investigated. Such information is highly relevant to designing and managing sanitation programs. Finally, as a prerequisite for developing adaptive BC tolerance, the occurrence of sanitizer residuals at sublethal levels to L. monocytogenes remains scarcely surveyed in processing environments, creating a substantial challenge for risk assessment and sanitation optimization.

In this study, we attempted to address the aforementioned data and evidence gaps toward an integrative risk assessment of reduced BC susceptibility in L. monocytogenes during produce processing. We substantiated the correlation between tolerance phenotypes and known genetic determinants of BC tolerance to facilitate the determination and prediction of tolerant and sensitive isolates. We performed a genomic survey on the prevalence and distribution of BC tolerance among 25,083 publicly available L. monocytogenes genomes from diverse sources in the United States. With the consideration of processing environment constraints, we refined the evaluation of adaptive response to BC by exploring the upper bound of adaptive BC tolerance and monitoring the temporal onset and duration of the tolerance in both tolerant and sensitive isolates. Finally, we examined residual BC levels throughout a fresh produce processing facility at different time points of daily operation to assess if the residual concentrations could be conducive to the emergence of adaptive BC tolerance in L. monocytogenes due to sublethal exposure to the sanitizer.

RESULTS

BC MICs in fresh produce isolates correlate with known resistance genes.

To establish the baselines of inherent BC MICs informed by the presence and absence of known BC tolerance genes, we assembled a set of 359 isolates representing diverse produce origins (22 commodities), different produce-related environments (packing houses and retailers), and the major genetic lineages of L. monocytogenes (three lineages and five serotypes) (see Table S1 and Fig. S1 in the supplemental material). These isolates were sampled between 1994 and 2019 in 26 U.S. states, Canada, Chile, Mexico, Peru, and Egypt, including 18 isolates linked to six vegetable or fruit commodities from five outbreaks (Table S1). The genomes of these isolates had been sequenced, allowing the selection of a substantial percentage of isolates (44.8%) that carry known BC tolerance genes, including the plasmid-borne bcrABC gene cassette (n = 154) (27) and chromosomally integrated qacH (n = 6) (24) and emrE (n = 1) (25). BC MICs of these isolates (n = 359) assumed a bimodal distribution (Fig. 1). All the tolerance gene-carrying isolates had MICs between 25 and 35 μg/mL, and nearly all the isolates (98.0%) devoid of any of the tolerance genes had an MIC of 5 μg/mL, except for a few showing an MIC of 10 μg/mL. The tolerance levels conferred by the tolerance genes are similar to those reported in previous studies, including 30 to 40 μg/mL for bcrABC (23) and 30 μg/mL for both qacH (24) and emrE (25). Therefore, despite the underrepresentation of isolates with qacH and emrE, MIC surveys from this and other studies correlating tolerance phenotype with genotype collectively show a clear distinction between inherently BC-sensitive (MIC ≤ 10 μg/mL) and BC-tolerant (MIC ≥ 25 μg/mL) isolates of L. monocytogenes.

FIG 1.

Distribution of BC MICs in 359 fresh produce isolates. “NA” indicates isolates that do not have any tolerance gene.

Prevalence of bcrABC correlates with prevalence of sequence types and varies among isolate source categories.

Thirty-one percent of the 25,083 surveyed isolates (see Materials and Methods) were predicted to be BC tolerant by carrying any of the surveyed tolerance genes, of which bcrABC is the predominant tolerance determinant, accounting for 93.6% of presumptive tolerant isolates (Table 1). Compared with bcrABC, other tolerance genes were rare, with only qacH exceeding 1% in prevalence (Table 1).

TABLE 1.

Prevalence of BC resistance determinants among publicly available L. monocytogenes genomes in the United States

Resistance determinant	% of isolates with determinant
	All isolates (n = 25,083)	BC-resistant isolates (n = 7,665)
bcrABC	28.60	93.58
qacH	1.76	5.75
emrC	0.04	0.13
emrE	0.14	0.44
qacC	<0.1	<0.1
qacA	<0.1	<0.1

We noticed an uneven distribution of bcrABC among lineages and sequence types (STs) using the seven-gene multilocus sequence typing (MLST) scheme by the Pasteur Institute (28) (Fig. 2). bcrABC was present in 17.9% and 45.2% of isolates in lineage I and lineage II, respectively, but absent from other lineages. The most prevalent ST in lineage I (ST5) and the five most prevalent STs in lineage II (ST321, ST7, ST155, ST9 and ST199) were all overrepresented by bcrABC-positive isolates; these six STs accounted for 78.2% of such isolates in the entire set (Fig. 2a). These STs all featured isolates from diverse sources, including vegetables and fruits, with environmental isolates being the predominant source category in each ST (Fig. 2b). While environmental isolates were found in 620 STs, these six STs contributed 42.8% of all isolates and 79.4% of bcrABC-positive isolates from the environmental category.

FIG 2.

Genomic survey of BC tolerance in publicly available L. monocytogenes genomes in the United States. (a) Minimum-spanning tree based on MLST of L. monocytogenes isolates marked by bcrABC and qacH prevalence in each ST. Major STs with high bcrABC prevalence are identified by ST numbers. (b) Minimum-spanning tree based on MLST of L. monocytogenes isolates marked by isolate source categories.

We further observed an association between high ST prevalence and high bcrABC prevalence. Within the entire isolate set, we ranked STs by size and collected isolates from larger to smaller STs until half of the set were sampled. The prevalence of bcrABC was significantly higher among isolates from the largest STs than the other half of isolates from smaller STs (Fig. 3a). The same observation was made in each source category (Fig. 3a). Notably, such association was mainly attributed to the six aforementioned prevalent STs in lineage I (ST5) and lineage II (ST321, ST7, ST155, ST9, and ST199) (Fig. 3a).

FIG 3.

Comparison of bcrABC prevalence by source categories and STs. The percentage above each bar is the bcrABC prevalence, and n is the number of isolates in this group. (a) bcrABC prevalence comparison between isolates from bigger STs and smaller STs by source categories. Isolates from the largest ST in lineage I (ST5) and the five largest STs in lineage II (ST321, ST155, ST199, ST7, and ST9) are shown in the hatched bars. (b) Comparison of bcrABC prevalence by major source categories (middle). Further comparison is expanded to subcategories under “environmental” (left) and “land food animals” (right). The left panel shows bcrABC prevalence in isolates belonging to major environmental source types. Different letters above the bar indicate significant difference in bcrABC prevalence by chi-square test (P < 0.05).

By source category, fruits and vegetables exhibited a relatively low level of bcrABC prevalence (18.6%), which is statistically similar to that of land food animals (22.8%), significantly lower than those of multi-ingredient food (48.6%), environmental (41.2%), and aquatic animals (35.3%), and only significantly higher than that of humans (7.2%) (Fig. 3b). When further compared with subcategories under land food animals, fruit and vegetable isolates ranked only above bovine and dairy isolates by source-specific bcrABC prevalence (Fig. 3b). Under the environmental category, subcategories associated with food processing environments (food contact surface, non-food-contact surface, and environmental swab or sponge) feature higher bcrABC prevalence than non-processing environments (water, water sediment, and soil) (Fig. 3b).

Temporal patterns of adaptive BC tolerance.

Because BC tolerance in U.S. isolates was predominantly caused by bcrABC, we focused on temporal characterization of the adaptive BC tolerance associated with this gene cassette. We observed both similarities and differences in the temporal dynamics of adaptive BC tolerance between isolates H7550 and H7550-Cd^s. The two isolates differed only by a plasmid that harbors bcrABC to confer BC tolerance (22). During sublethal BC exposure (half of the respective inherent MIC) under the growth-promoting condition, both isolates exhibited two phases of tolerance increase divided by a plateau of no MIC change (Fig. 4). Regardless of their inherent MICs, both isolates gained 5 μg/mL of MIC after the first phase and peaked with another increase of 15 μg/mL by the end of the second phase (Fig. 4). During the first phase, an increase of MIC emerged between 6 and 9 h in both isolates (Fig. 4). Under the growth-limiting conditions in 0.1% buffered peptone water at both 4°C (Fig. 5a and b) and 25°C (Fig. 5c and d), neither isolate developed any adaptive tolerance upon exposure to BC at 2.5 μg/mL (Fig. 5). At 25°C, both isolates did not survive after 6 h of exposure (Fig. 5c and d). Under the growth-limiting condition at 4°C and 25°C, the BC-tolerant isolate H7550 was not culturable after 3 h of exposure to BC at 15 μg/mL (half of the inherent tolerance of H7550). Therefore, H7550 was also adapted to BC at 2.5 μg/mL (half of the inherent tolerance of H7550-Cd^s).

FIG 4.

Temporal patterns of BC adaptive tolerance under the growth-promoting condition. The y axis on the left shows the MICs of the isolates (H7550-Cd^s, blue circles; H7550 orange circles). The y axis on the right shows BC concentrations in the cultures (solid horizontal lines). Two phases of tolerance development (P1 and P2) and the plateau in between are shaded. (a) Continuous monitoring of the BC MIC in H7550-Cd^s, first under sublethal exposure to BC (2.5 μg/mL, half of its intrinsic BC MIC) on nutrient culture at 30°C up to 96 h and then on subcultures without BC exposure. Four subcultures were monitored at 37°C, and each lasted 24 h (at 121, 145, 169, and 193 h). The exposure time after 24 h is not shown to scale. (b) Continuous monitoring of the BC MIC in H7550, first under sublethal exposure to BC (15 μg/mL, half of its intrinsic BC MIC) on nutrient culture at 30°C up to 96 h and then on subcultures without BC exposure. Four subcultures were monitored at 37°C, and each lasted 24 h (at 121, 145, 169, and 193 h). Exposure time after 24 h is not shown to scale.

FIG 5.

Temporal patterns of BC adaptive tolerance under the growth-limiting condition. The y axis on the left shows the MIC of the isolates (H7550-Cd^s, blue circles; H7550, orange circles). The y axis on the right shows BC concentrations in diluted buffered peptone water. Solid horizontal lines represent sublethal concentrations of BC (2.5 μg/mL). Empty circles represent unculturable cells. (a and b) Continuous monitoring of the BC MIC in H7550-Cd^s (a) and H7550 (b) in 0.1% buffered peptone water at 4°C up to 48 h; (c and d) continuous monitoring of the BC MIC in H7550-Cd^s (c) and H7550 (d) in 0.1% buffered peptone water at 25°C up to 48 h. The exposure time is not shown to scale.

Under the growth-promoting condition, in comparison to the synchronized MIC increase in the first phase, the development of additional tolerance took substantially longer in H7550 than in H7550-Cd^s (Fig. 4). The plateau after the first phase lasted 9 h in H7750 (9 to 18 h) compared with 3 h in H7550-Cd^s (9 to 12 h). Then it took 30 h for H7750 (18 to 48 h) to reach the maximum adaptive tolerance, compared with 6 h for H7550-Cd^s (12 to 18 h). Because of these prolonged adaptive responses in H7750, the full development of tolerance was more than three times longer in H7750 (72 h) than in H7750-Cd^s (21 h). Another notable difference between the two isolates occurred upon the removal of BC exposure after the full tolerance development. In H7550-Cd^s, the tolerance persisted in nutrient culture without external BC and throughout repeated subculturing (Fig. 4a), which confirmed previous reports that the acquired tolerance in BC-sensitive isolates was hereditary (29). In contrast, acquired tolerance in H7550 dropped substantially within 1 h after the removal of BC exposure (Fig. 4b). We confirmed that the plasmid carrying bcrABC was still present in H7550 after the removal of BC exposure (data not shown), suggesting that the waning of adaptive tolerance was not due to the loss of the plasmid.

In addition to H7550 and H7550-Cd^s, we examined adaptive BC tolerance in another three bcrABC-positive and BC-tolerant isolates and another three bcrABC-negative and BC-sensitive isolates using the same adaptation method. The additional isolates represented three major genetic lineages (Table S1) and exhibited similar temporal patterns of adaptive tolerance to H7550 or H7550-Cd^s (Fig. S2).

Residual BC levels in a produce processing facility.

We surveyed residual BC concentrations throughout a fresh produce processing facility in the Southeast United States that follows a daily regimen of sanitizing by QAC, production, and cleaning by water (see Materials and Methods). Right after QAC application, the vast majority of sampling locations reached BC concentrations that were multiple times higher (Table 2) than the upper bond of adaptive BC tolerance of L. monocytogenes at 50 μg/mL, which was obtained by adapting a BC-tolerant isolate under a growth-promoting condition (nutrient culture, 30°C). The exceptions were samples from the bottom shelf of the cart (below the detection limit on days 1 and 2), packing machine (20 and 30 μg/mL on day 1), and processing machine (60 and 70 μg/mL on day 1) (Table 2). These results suggest that the application of the QAC sanitizer that contained BC at 400 μg/mL should leave residual BC concentrations sufficient for effective control of L. monocytogenes at the sampled locations, even if the pathogen acquired maximum adaptive tolerance.

TABLE 2.

Residual BC concentrations in a produce processing facility

Sample locationa	BC concn (μg/mL) onb:
	Day 1		Day 2		Day 3
	Postapplication	Postrinse	Postapplication	Postrinse	Postapplication	Postrinse
Cart (bottom shelf, PR)	ND/ND	ND/ND	ND/ND	ND/ND	200/70	ND/ND
Packing machine (PR)	30/20	ND/ND	190/180	ND/ND	130/140	ND/ND
Line 1 conveyor belt (PR)	300/300	ND/ND	180/190	ND/ND	230/230	ND/ND
Floor drain (PR)	150/130	ND/ND	200/130	ND/ND	140/200	ND/ND
Line 2 small conveyor belt (PR)	230/250	ND/ND	300/290	ND/ND	300/240	ND/ND
Line 2 inclined conveyor belt (RR)	270/270	ND/ND	240/250	ND/ND	310/290	ND/ND
Line 2 sorting machine (RR)	170/160	ND/ND	370/380	ND/ND	280/270	ND/ND
Line 1 inclined conveyor belt (RR)	260/260	ND/ND	340/350	ND/ND	300/140	ND/ND
Line 2 large conveyor belt (PR)	190/200	ND/ND	210/200	ND/ND	150/140	ND/ND
Line 1 processing machine (PR)	60/70	ND/ND	240/270	ND/ND	270/260	ND/ND

^aPR, processing room; RR, raw product room.

^bResults are shown for replicate 1/replicate 2. ND, not detected (under 10 μg/mL).

Right after water rinse of equipment, no detectable presence of BC was found at all sampling locations on all 3 days (Table 2). The limit of detection was 10 μg/mL.

DISCUSSION

The strong correlation between known BC tolerance determinants, especially the plasmid-borne bcrABC gene cassette, and tolerance phenotypes of fresh produce isolates paved the way for evaluation of BC selection on L. monocytogenes through a large-scale genomic survey and suggested a need for differential assessment of adaptive BC tolerance in tolerant and sensitive strains.

The presence of bcrABC in different lineages and multiple phylogenetically distant STs can be explained by independent acquisition of the cassette, presumably through horizontal dissemination of the plasmid that harbors bcrABC. The similar sporadic presence of transposon-borne qacH was likely caused by horizontal gene transfer as well. Alternatively, the scattered phylogenetic distribution of bcrABC could result from independent loss of the plasmid in different STs whose common ancestor(s) had acquired the plasmid. Under both scenarios, the use of QACs as sanitizing agents may exert selective pressure on L. monocytogenes, causing the acquiring or maintaining of the bcrABC-harboring plasmid in different L. monoctytogenes populations.

The association between prevalent strains and high percentages of bcrABC-positive isolates across all major source categories is congruent with the assumption that bcrABC confers a selective advantage to commonly circulating strains under the widespread use of QACs. Some of these STs identified from U.S. isolates were also found prevalent and overrepresented by bcrABC-positive isolates in surveys in other countries, such as ST321 in Canada (30) and South Africa (31) and ST5 and ST9 in Canada (30), Norway (32), and Ireland (33).

By comparing bcrABC prevalence among source categories, we found evidence that BC selection is attributable to QAC application in food processing. The markedly higher bcrABC prevalence observed among isolates from food processing environments (food contact and non-food-contact surfaces, environmental swab, and sponge) than from sources more akin to natural environments (water and soil) (Fig. 3b) indicates possible selection for less-susceptible isolates caused by BC exposure in food processing environments. Consistent with our finding, recent surveys of L. monocytogenes genomes from foods and food manufacturing environments in the United States (4,969 isolates) (34) and Canada (1,279 isolates) (30) also found similarly high bcrABC prevalence (46.3% in the United States, 41.5% in Canada). Together, these results support the general recommendation by FDA and Food Safety and Inspection Service (FSIS) to rotate sanitizers in RTE processing environments.

Further comparison of bcrABC prevalence among food commodities indicates that the consequences of BC selection may vary among different agrifood sectors. A possible explanation for the relatively low bcrABC prevalence among fruit and vegetable isolates is that L. monocytogenes contamination of fresh produce, as well as the produce processing environments, originates primarily from sources closer to the organism’s natural habitats, such as soil and water, where bcrABC prevalence in L. monocytogenes is low (Fig. 3b). Furthermore, if BC selection in the produce processing environment is limited or negligible, isolates from such environments are presumably more diverse than would be expected if the use of QACs constitutes a population bottleneck. A recent study found that L. monocytogenes isolates from properly sanitized tree fruit packinghouses were not only predominantly bcrABC negative but also diverse and linked to transient contamination (35).

Notably, 494 isolates sampled from multiple facilities over three successive seasons in the tree fruit packinghouse study (35) were included in our genomic survey, and only four were found to harbor bcrABC. These mostly bcrABC-negative isolates belonged to the environmental subcategory of “environmental swab or sponge” (Fig. 3b), which nevertheless had an overall bcrABC prevalence as high as 52.1%. The contrast suggests that bcrABC prevalence in L. monocytogenes is also highly variable among the processing environments of certain commodities and sectors, consistent with the observed prevalence variation among commodities themselves (Fig. 3b). According to the extrapolation, isolates from meat processing environments, for example, should have a much higher bcrABC prevalence than isolates from produce processing environments. While bcrABC prevalence in processing environment isolates was difficult to determine for specific commodities due to the lack of commodity information for publicly available genomes, a survey of L. monocytogenes from meat processing in South Africa identified a relatively high bcrABC prevalence of 38% (31), comparable to what we observed from various meat commodities in the United States (Fig. 3b).

If different sectors are disparately affected by reduced BC susceptibility in L. monocytogenes, a question may be raised of whether sanitation programs, including rotational use of sanitizers, can be tailored to different sectors. Our genomic survey provides a promising tool to help address this question. National and international infrastructures for genomic surveillance of pathogens such as GenomeTrakr (36) supply large volumes of L. monocytogenes genomes from food commodities and their processing environments. The public data can be leveraged for timely and source-structured monitoring of sanitizer tolerance, similar to genomic monitoring of antibiotic resistance in the food supply chain (37, 38). Our survey results, while affected by incomplete information and sampling bias inherent to public genome data, suggest a need and feasibility for investigation of data-driven and targeted sanitation strategies in food processing environments.

Besides subjecting L. monocytogenes to BC selection, the bcrABC-encoding plasmid also accounts for distinct patterns of adaptive BC tolerance observed between tolerant and sensitive isolates. We revealed the distinction by continuously monitoring the MIC of each type of isolates over the entire course of extended BC exposure and after the removal of the exposure. Unlike previous studies that measured endpoint MICs following adaptation in single cultures (8, 9, 11) or through successive subcultures (10, 12, 13, 26), we characterized the complete temporal dynamics of tolerance development. Such characterization provides additional insights to assess QAC efficacy against adaptive tolerance, if any, under the realistic constraints of processing environments.

The upper bounds of adaptive tolerance derived from the optimal growth condition (Fig. 4) can be used for conservative risk assessment. The BC-sensitive isolate without bcrABC was able to develop maximum tolerance as much as 5 times its original MIC (from 5 μg/mL to 25 μg/mL) after 21 h of BC adaptation, which is temporally permissive during the 24-h interval between two consecutive QAC applications based on a once-a-day application regime. The BC-tolerant isolate harboring bcrABC required 72 h to achieve its full tolerance potential (MICs from 30 μg/mL to 50 μg/mL), whereas a 24-h adaptation during the application interval reached half of the potential (from 30 μg/mL to 40 μg/mL). Because the first 12 h of BC adaptation resulted in only a minor MIC increase (5 μg/mL) in both BC-sensitive and BC-tolerant isolates, reduction of the between-shift interval to 12 h can theoretically help prevent the bulk of tolerance development that took place after 12 h in the BC-sensitive isolate and after 18 h in the BC-tolerant isolate.

The distinct trajectories of adaptive tolerance after its full development in different types of isolates may also be factored into the control and monitoring of reduced BC susceptibility in L. monocytogenes. The fully developed adaptive BC tolerance in the BC-sensitive isolate (25 μg/mL) was close to the inherent BC resistance of the BC-tolerant isolate (30 μg/mL), did not require external BC to maintain, and showed stable inheritance by offspring cells. If adapted strains emerged, they could constitute a risk factor for subsequent horizontal acquisition of even higher tolerance. While our combined phenotypic and genotypic assays of an extensive collection of fresh produce isolates did not find any adapted BC-sensitive strain of elevated BC MIC, the possibility has yet to be ruled out. Preventive monitoring of the hereditary adaptive tolerance will be facilitated by the identification of its molecular determinant(s). It has been reported that mutations in an efflux pump regulator gene, fepR, are associated with adaptive BC tolerance in L. monoctytogenes (12, 26). The fepR mutations were observed after multiple rounds of serial passaging and selection of L. monocytogenes isolates in nutrient cultures with sublethal levels of BC (12, 26). No fepR mutation was found in any of the parent strains (26), and it is likely absent from any of the short-term-adapted (<24 h) isolates in single cultures in the current study (unpublished whole-genome sequencing data of parent and adapted cells). The adaptive and heritable BC tolerance we observed in BC-sensitive isolates is potentially mediated by unidentified determinant(s).

In comparison to tolerance development in BC-sensitive isolates, the adaptive response in BC-tolerant isolates was apparently delayed by the bcrABC-encoded efflux pump and likely mediated through a different, yet unknown mechanism, which was quickly reversed upon the removal of BC exposure (Fig. 4). Of note, adaptive antibiotic resistance that is transient and reversible can be mediated by different environmental and genetic factors that alter the phenotypic expression of the resistance as reviewed in reference 39.

The absence of adaptive tolerance in both BC-sensitive and BC-tolerant isolates under the growth-limiting condition (Fig. 5) may better inform the realistic likelihood of the adaptive response in processing facilities that maintain good sanitary standards. In food processing, food soil and debris are likely the only possible source of environmental nutrients to support bacterial growth. L. monocytogenes can grow on certain fruits and vegetables (40, 41). Elimination of produce debris and residues from processing environments by proper cleaning further subtracts another risk factor for the development of adaptive tolerance. The differential BC adaptation in response to different nutrient availability suggests that the adaptive mechanism is bacterial growth dependent. For example, genetic mutations are contingent upon DNA replication and more likely to accumulate over multiple rounds of cell multiplication during serial cultivation (12, 26).

Another critical processing environment constraint that determines the realistic likelihood of adaptive BC tolerance in L. monocytogenes is the availability of sublethal concentrations of BC in such environments. Our survey of the processing facility suggests that a properly maintained daily sanitation regime is unlikely to create conditions that are conducive to the development of adaptive BC tolerance. Because of the limited sensitivity of the commonly used BC measurement kit (10 μg/mL), it was possible that postrinse samples without detectable BC still contained low levels of the sanitizer that are sublethal to L. monocytogenes. However, these samples were unlikely to induce adaptive BC tolerance because the water samples would not support L. monocytogenes growth that is required for tolerance development (Fig. 5).

Although our facility survey thoroughly covered locations where liquid residues were visible and accessible, there might still be hidden sites that could provide harborage for sublethal levels of BC and/or food debris from routine cleaning, such as those that required disassembly of equipment to access. Further identification, investigation, and mitigation of these potential harborage sites are warranted.

In summary, by integrating a genomic survey, laboratory assays, and processing facility sampling, our study helps reconcile the debate over reduced BC susceptibility in L. monocytogenes linked to QAC application in food processing environments. For the processing of fresh fruits and vegetables, we conclude that properly sanitized and cleaned facilities may be less affected by BC selection and unlikely to provide conditions that are conducive for the emergence of adaptive BC tolerance in L. monocytogenes.

MATERIALS AND METHODS

Isolates and genomes.

L. monocytogenes isolates from produce or produce-related environments used for the BC MIC assay were sampled during a survey in 30 retail produce departments across seven U.S. states (42) or provided by the Center for Food Safety and Applied Nutrition, FDA. All of these isolates had been previously sequenced. Detailed information about these isolates, including genome accession numbers and isolation source, year, and location, is summarized in Table S1 in the supplemental. Isolate H7750 and its plasmid-cured derivative, H7550-Cd^s, were originally from the 1998 and 1999 multistate outbreak linked to contaminated hot dogs (27). A set of 25,083 publicly available L. monocytogenes genomes from the United States were collected to survey the distribution of known genetic determinants of BC tolerance. Genomes and associated metadata were downloaded from National Center for Biotechnology Information (NCBI). This set included all of the L. monocytogenes genomes deposited at NCBI as of 31 January 2022 that were sequenced by Illumina platforms (paired-end reads) and assembled by the NCBI SKESA assembler (43) to reach at least 2.5 Mb in the assembly. The genomes were sampled between 1931 and 2021 (70.8% after 2010) and from diverse sources, including humans (n = 7,327), environmental sources (n = 10,955), multi-ingredient food (n = 2,081), fruits and vegetables (n = 1,330), land food animals (n = 1,736), aquatic animals (n = 302), other animals (n = 235), and other sources (n = 309). Except for “other sources,” all of these source categories were deduced from publicly available descriptions of the isolation source associated with each genome according to the Interagency Food Safety Analytics Collaboration (IFSAC) Food Categorization Scheme (https://www.cdc.gov/foodsafety/ifsac/projects/food-categorization-scheme.html).

Determination of MICs.

Overnight cultures of isolates (H7550, H7550-Cd^s, and 359 isolates representing three major genetic lineages as noted in Table S1) in brain heart infusion broth (BHI) prepared from frozen stocks (–80°C) were transferred to Mueller-Hinton agar (MHA) with 5% defibrinated sheep blood (MHA-B) and incubated at 37°C for 72 h to enter the long-term survival (LTS) phase (44). Induced by prolonged nutritional starvation, the LTS state is more likely to represent the physiological state of persistent L. monocytogenes cells in produce processing environments where nutrients for bacterial growth are scarce. A high-throughput agar dilution assay was used for MIC determination (45). Specifically, for each isolate, two single colonies of LTS cells were picked and suspended in 200 μL of Mueller-Hinton broth (MHB) (~10⁹ CFU/mL), followed by serial dilutions in MHB to make a cell suspension at 10⁷ CFU/mL. Suspensions of 48 isolates were transferred to a 48-well plate. From the plate, 1 μL of each suspension was applied as spots by a 48-pin microplate replicator (model 140501, 1.5 by 1.5 mm) onto MHA plates with 2% defibrinated sheep blood (23) containing different concentrations of BC (0, 2.5, 5, 10, 15, 20, 25, 30, 35, 40, 45, and 50 μg/mL), with duplicate plates for each concentration. After 48 h of incubation at 30°C, MICs were determined as the lowest assessed BC concentrations that prevented visible growth of L. monocytogenes cells (27). The MICs of all 359 isolates from fresh produce or produce production environments (Table S1) were measured in three independent trials, each in duplicate. Isolates H7550 and H7550-Cd^s (H7550 cured of pLM80) (23) were used as BC-tolerant and BC-sensitive controls for determination of BC MIC.

Adaptation to sublethal concentrations of BC under the growth-promoting condition.

We determined the onset and duration of adaptive tolerance to BC by continuously monitoring the MICs of a BC-tolerant isolate, H7550 (lineage I, serotype 4b, ST6), and its BC-sensitive derivative, H7550-Cd^s, during and after sublethal exposure to BC. The continuous monitoring was performed by exposing L. monocytogenes cells to BC under the growth-promoting (nutrient culture at 30°C) condition. A third-generation quaternary ammonium compound (QAC) (alkyl distribution from C₈H₁₇ to C₁₆H₃₃) (Acros Organics, NJ), which is a mixture of alkyl dimethyl benzyl ammonium chloride and alkyl dimethyl ethylbenzyl ammonium chloride, was used to adapt L. monocytogenes. LTS cells of H7550 and H7550-Cd^s were prepared as described above. Ten microliters of cell suspension (~10⁶ CFU/mL) of each isolate was plated on MHA with 2% defibrinated sheep blood containing 2.5 μg/mL BC (half the intrinsic BC MIC of H7550-Cd^s) or 15 μg/mL BC (half the intrinsic BC MIC of H7550). The inoculated plates were incubated at 30°C for from 3 h up to 168 h. At each time point (3, 6, 9, 12, 15, 18, 21, 24, 48, 72, and 96 h), one plate was taken out of incubation to measure the MIC of the adapted isolate. Cells were collected by flushing the plate with 2 mL MHB along with gentle scrubbing by a spreader. About 1.5 mL of cell suspension was collected and then vortexed, from which 20 μL was taken for the determination of MIC as described above. After sublethal BC adaptation for 96 h, cells of each adapted isolate were harvested as described above. Twenty microliters of harvested cells was transferred to MHB without BC and incubated at 37°C for 24 h. The subculturing continued every 24 h for up to 4 days to confirm if BC adaptive tolerance was inheritable and contingent upon extracellular BC. The MIC of cells from each subculture was determined as described above.

Adaptation to sublethal concentrations of BC under growth-limiting conditions.

Adaptation of LTS cells of H7550 and H7550-Cd^s under growth-limiting conditions was performed similar to described above, except BC-containing 0.1% buffered peptone water instead of MHA was used for the adaptation at 25°C or 4°C.

Genomic survey of BC resistance genes.

Genome data were indexed with Colorid (https://github.com/hcdenbakker/colorid), an implementation of BItsliced Genomic Signature Index (BIGSI) (46). The presence or absence of BC tolerance genes, including bcrABC, qacH, qacC, qacA, emrE, emrC, mdrL, and lde was determined in each genome using the Colorid indices. The sequences of bcrABC, qacH, emrE, mdrL, and lde were retrieved from the Listeria database of the Pasteur Institute. The sequences of qacC (NCBI accession no. Z37964), qacA (NCBI accession no. X56628.1), and emrC (NCBI accession no. LT732640.1:1575-1961) were retrieved from NCBI.

Source categorization of publicly available L. monocytogenes genomes.

For source categorization of L. monocytogenes genomes, for each publicly available genome, the associated description of isolation source was retrieved from the NCBI Short Reads Archive (https://www.ncbi.nlm.nih.gov/sra) and parsed by LexMapr version 0.7.1 (https://github.com/cidgoh/LexMapr) according to the Interagency Food Safety Analytics Collaboration (IFSAC) Food Categorization Scheme (https://www.cdc.gov/foodsafety/ifsac/projects/food-categorization-scheme.html). Source categories deduced by LexMapr were further curated and consolidated if needed. Specifically, environmental sources without any specified animal or plant vehicle were categorized as “environmental.” Sources that were deduced as multi-ingredient by LexMapr were merged into a single category of “multi-ingredient food.” Fish, shellfish, and other aquatic animals were grouped as “aquatic animals.” All poultry, swine, bovine, dairy and meat sources were classified as “land food animals.” Genomes of unclassified or unknown sources were categorized into “other sources.”

Statistical analysis.

Comparison of bcrABC prevalence among source categories of L. monocytogenes was performed using chi-square tests, and post hoc analyses with Bonferroni correction were performed with the scipy (version 1.7.1) and statsmodels (0.13.0) Python packages.

Collection of sanitizer residues and measurement of BC concentrations.

We surveyed residual BC concentrations throughout a fresh produce processing facility in the Southeast United States. The facility processes fresh tomatoes, onions, jalapeno peppers, and cilantro by following a daily regimen that includes (i) early morning sanitizing using a QAC product (BC concentration at 400 μg/mL), (ii) production throughout the day, and (iii) evening cleaning of equipment and facility by tap water. The temperature of the facility was maintained at around 4°C. Residual sanitizer samples were collected at two time points from 10 different locations in the facility daily for 3 consecutive days. The time points include postapplication (early morning) and postrinse (evening). The sampling locations included a cart (bottom shelf), packing machine, conveyor belts (big, small, and inclined), floor drain, sorting machine, and processing machine. These locations were spread out across a raw product room and a processing room. The two rooms were connected by two processing lines. Two sampling locations associated with inclined conveyor belts from the two different lines in the raw product room were considered less accessible for sanitation and cleaning because of their locations in tight corners and the elevated heights of the belts. An ~10- to 15-mL sample of residual sanitizer was taken in duplicate at each sampling area by either transferring liquid with a 15-mL transfer pipette (VWR International Radnor, PA) or by wiping the area with a Nasco Speci-Sponge (Nasco Madison, WI) until saturated and squeezing the liquid into a 50-mL conical tube. Runoff liquid that dripped from a particular piece of processing equipment onto the floor underneath and near the equipment was sampled immediately after sanitation or cleaning. Liquid accumulation was directly sampled from the floor drain and the bottom shelf of a cart. Samples were transferred to a refrigerator within 1 to 2 h. Following the manufacturer’s instructions, residual sanitizer samples were titrated to determine BC concentration using the LaMotte 3043-DR-01 QAC kit (LaMotte Chestertown, MD), which has a range of 0 to 500 ppm and a sensitivity of 10 ppm.