ORIGINAL RESEARCH published: 13 May 2021 doi: 10.3389/fmicb.2021.598739 A Diffusion Model to Quantify Membrane Repair Process in Listeria monocytogenes Exposed to High Pressure Processing Based on Fluorescence Microscopy Data Bahareh Nikparvar 1 , Alicia Subires 2 , Marta Capellas 2 , Manuela Hernandez-Herrero 2 , Peter Crauwels 3 , Christian U. Riedel 3 and Nadav Bar 1* 1 Department of Chemical Engineering, Norwegian University of Science and Technology, Trondheim, Norway, 2 Department of Animal and Food Science, Autonomous University of Barcelona, Barcelona, Spain, 3 Department of Biology, Institute of Microbiology and Biotechnology, Ulm University, Ulm, Germany Edited by: Teresa Zotta, University of Basilicata, Italy The effects of environmental stresses on microorganisms have been well-studied, and Reviewed by: cellular responses to stresses such as heat, cold, acids, and salts have been extensively Hui Li, Chinese Academy of Agricultural discussed. Although high pressure processing (HPP) is becoming more popular as a Sciences, China preservation method in the food industry, the characteristics of the cellular damage Jose A. Egea, Spanish National Research Council, caused by high pressure are unclear, and the microbial response to this stress has Spain not yet been well-explored. We exposed the pathogen Listeria monocytogenes to HPP Didier Cabanes, (400 MPa, 8 min, 8◦ C) and found that the high pressure created plasma membrane University of Porto, Portugal pores. Using a common staining technique involving propidium iodide (PI) combined *Correspondence: Nadav Bar with high-frequency fluorescence microscopy, we monitored the rate of diffusion of PI nadi.bar@ntnu.no molecules into hundreds of bacterial cells through these pores on days 0, 1, 2, 3, and 4 after pressurization. We also developed a mathematical dynamic model based on mass Specialty section: This article was submitted to transfer and passive diffusion laws, calibrated using our microscopy experiments, to Food Microbiology, evaluate the response of bacteria to HPP. We found that the rate of diffusion of PI into the a section of the journal cells decreased over the 4 consecutive days after exposure to HPP, indicating repair of Frontiers in Microbiology the pressure-created membrane pores. The model suggested a temporal change in the Received: 25 August 2021 Accepted: 12 April 2021 size of pores until closure. To the best of our knowledge, this is the first time that pressure- Published: 13 May 2021 created membrane pores have been quantitatively described and shown to diminish with Citation: time. In addition, we found that the membrane repair rate in response to HPP was linear, Nikparvar B, Subires A, Capellas M, Hernandez-Herrero M, Crauwels P, and growth was temporarily arrested at the population level during the repair period. Riedel CU and Bar N (2021) A These results support the existence of a progressive repair process in some of the cells Diffusion Model to Quantify Membrane that take up PI, which can therefore be considered as being sub-lethally injured rather Repair Process in Listeria monocytogenes Exposed to High than dead. Hence, we showed that a subgroup of bacteria survived HPP and actively Pressure Processing Based on repaired their membrane pores. Fluorescence Microscopy Data. Front. Microbiol. 12:598739. Keywords: high pressure processing, Listeria monocytogenes, mathematical modeling, membrane damage, repair doi: 10.3389/fmicb.2021.598739 process, fluorescence microscopy Frontiers in Microbiology | www.frontiersin.org 1 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure 1. INTRODUCTION pressure treatment of 900 MPa, some Listeria (L.) monocytogenes cells remained viable and were able to recover at 14◦ C. Klotz et al. Bacteria in nature are exposed to various environmental stresses, (2010) examined the susceptibility of cell membranes in E. coli including changes in temperature or pH, radiation, antimicrobial to pressure-induced damage (500, 600, and 700 MPa) and found compounds, and osmotic pressure. Pressurization of bacteria that in a pressure-resistant strain, uptake of PI occurred only that normally grow in atmospheric conditions may trigger during exposure but not after pressure release, indicating that response mechanisms that enable them to adapt to the new the cells were able to reseal their leaky membranes. This again pressure condition and survive. Although the cell envelope supports the hypothesis of the presence of a recovery process and particularly the membrane structure have been reported to in sub-lethally injured cells, and adds to the evidence in the be susceptible to high pressure processing (HPP) (Pagán and literature that, in contrast to what has been traditionally assumed, Mackey, 2000; Winter and Jeworrek, 2009; Gänzle and Liu, cells that take up PI are not always dead (Shi et al., 2007; Davey 2015), the mechanisms that bacteria activate as a response to this and Hexley, 2011; Subires et al., 2013; Yang et al., 2015). stress are still largely unknown. Moreover, the potential existence Several authors have proposed simple models to of a membrane repair machinery in bacteria that responds to describe inactivation of pressure-treated bacteria and their pressure-induced damage has not been well-investigated. This growth behaviors (Koseki et al., 2007; Bover-Cid et al., phenomenon is of particular importance in the food industry, 2010; Hereu et al., 2014; Valdramidis et al., 2015; Rubio where exposure to HPP is used as a preservation method to et al., 2018). However, although a few of these studies inactivate foodborne bacteria. reported the existence of a repair process, they did not Release of low-molecular-weight metabolites, including identify the underlying mechanisms that allow bacteria nucleotides, amino acids, and inorganic ions, from bacterial to recover. cells exposed to different types of stress such as antibiotics or Here, we focused on foodborne pathogenic bacteria L. bacteriocins has been proposed as an indicator of membrane monocytogenes and the effects of HPP (400 MPa, 8 min, 8◦ C) damage (Lambert and Hammond, 1973; Gilbert et al., 1977; on its membrane. We developed a dynamic model that could Broxton et al., 1983; Zhen et al., 2013; Singh et al., 2016). Two estimate pressure-induced membrane damage over time. decades ago, high pressure stress had already been shown to cause Our predictions and findings suggest that sub-lethally injured deformation of the cell membrane and create surface pores (Ritz pressurized cells were able to repair their membranes by et al., 2001). Several authors demonstrated increased uptake resealing the surface pores after decompression. To the best of exogenous fluorescent molecules by pressurized cells. For of our knowledge, this is the first time that the repair example, membrane damage in Lactobacillus plantarum exposed of membrane pores following HPP has been shown in to high pressure was shown by staining cells with the fluorescent bacteria.The estimated time required for resealing the pore dye propidium iodide (PI) (Smelt et al., 1994). In the same work, area could be useful for the food industry to adjust the high leakage of ATP from these cells was observed, indicating a leaky pressure strategy applied (particularly by adjusting the pressure membrane (Smelt et al., 1994). Gänzle and Vogel (2001) showed strength and holding time) to design a more effective food changes in the kinetics of outer and cytoplasmic membrane preservation process. permeability in Escherichia (E.) coli after exposure to high pressure (300, 500, 600 MPa) by staining of treated cells with PI and 1-N-phenylnaphthylamine. Furthermore, a few studies have detected cellular proteins outside the cell after exposure 2. MATERIALS AND METHODS to high pressure (Smelt et al., 1994; Gänzle and Vogel, 2001; 2.1. Experiments Mañas and Mackey, 2004). One such example is the study of 2.1.1. Bacterial Strain and Culture Conditions Mañas and Mackey (2004), which detected intracellular proteins The L. monocytogenes Scott A strain used in this study was outside pressure-treated E. coli cells (200 MPa, 8 min), indicating provided by the Collection of Institut Pasteur (CIP 103575; membrane leakage. Paris, France) and selected based on its increased high pressure Several previous experiments showed that pressure-mediated resistance and widespread use as a reference strain in food damage in bacteria could be repairable such that the cells preservation technology testing (Alpas et al., 1999; Briers et al., could potentially grow after repair of the site of injury during 2011; Duru et al., 2020). Stock cultures grown in tryptone storage (Bozoglu et al., 2004; Jofré et al., 2010; Klotz et al., 2010). soya broth supplemented with 0.6% w/v yeast extract (TSBYE; Bozoglu et al. (2004) observed no colony formation in selective Oxoid/ThermoFisher Scientific, Hampshire, UK) were stored at or non-selective agar immediately after pressure treatment at 550 −80◦ C in glycerol (33% v/v). A loopful of the glycerol stock was MPa, suggesting that all cells were inactivated. However, they inoculated into 20 mL TSBYE and incubated at 37◦ C overnight detected growth in both selective and non-selective agar after 6 in a shaking (80 rpm) water bath. To prepare working cultures days at 4◦ C, and after 1 day at 22 and 30◦ C, probably due to a in the early stationary phase, the overnight culture was diluted recovery process. Jofré et al. (2010) reported that even after a high to 1:100 in 20 mL of fresh TSBYE and incubated under the same conditions for 18 h. The time to reach this growth phase Abbreviations: E. coli, Escherichia coli; FI, Fluorescence intensity; HPP, High was established by monitoring the optical density at 600 nm pressure processing; L. monocytogenes, Listeria monocytogenes; LOD, Limit of (Bioscreen C; Oy Growth Curves AB, Helsinki, Finland) in a detection; LOQ, Limit of quantification; PI, Propidium iodide. separate experiment. Frontiers in Microbiology | www.frontiersin.org 2 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure 2.1.2. High Pressure Processing (HPP) and a hybrid detector. For each sample and field, images were Early stationary phase cultures were transferred to 30-mL HDPE captured for 30 min at 1 frame per second (fps). The maximum bottles with screw caps (BNH0030PN, SciLabware Limited, achievable fluorescence intensity (FI) was obtained from the Stoke-on-Trent, UK), avoiding the presence of air bubbles. heat-treated sample, which was used every day as the reference Bottles had previously been sterilized with a 70% v/v ethanol to set the microscope gain parameter. solution overnight, rinsed three times with sterile distilled water, and dried at 60◦ C in an incubator. The possible presence of 2.1.4. Culture-Based Cell Counting ethanol residues after this procedure was ruled out, as growth of The number of viable L. monocytogenes cells was determined untreated samples was not reduced (data not shown). To prevent by the spread plate count method before exposure to HPP hazardous culture spills, caps were sealed with laboratory film, (untreated) and on days 0, 1, 2, 3, and 7 after the treatment (400 and bottles were vacuum packed in sterile plastic bags. Samples MPa, 8 min, 8 ◦ C). Samples were serially diluted at the designated were then cooled to 8◦ C in ice before HPP (approximately 30 time points in peptone saline solution (1 g/L neutralized min). bacteriological peptone [Oxoid/ThermoFisher Scientific] and 8.5 HPP was performed in a discontinuous isostatic press (Alstom g/L NaCl in water). Appropriate dilutions were spiral-plated ACB, Nantes, France) fitted with a 2-L pressure chamber (Eddy Jet; IUL, Barcelona, Spain) on the non-selective medium containing water. The temperature of the pressure chamber and tryptone soya agar supplemented with 0.6% (w/v) yeast extract transmission fluid was adjusted to the treatment temperature (Oxoid/ThermoFisher Scientific). Plates were then incubated at (8◦ C) using an external continuous cooling system. The samples 37◦ C for 48 h before colony counting. The limit of detection were then placed inside the chamber, allowed to re-equilibrate (LOD) and the limit of quantification (LOQ) (i.e., the lower limit to 8◦ C for 5 min, and pressurized at 400 MPa for 8 min. of acceptably accurate cell counts) of this method were 1.00 and The temperature was selected according to the common HPP 2.40 log CFU/mL, respectively. conducted in the industry (Muntean et al., 2016). The come-up time and the decompression time were both 2 min. Pressurized 2.2. Post-processing of Microscopy Data samples were kept in ice before being further processed for We used Fiji software, RRID:_SCR_002285 (an image-processing fluorescence microscopy analysis, and then stored at 8◦ C for package based on ImageJ), RRID:_SCR_003070, and MATLAB, the subsequent days of analysis. The storage temperature was RRID:_SCR_001622 to derive the mean red FI value for each cell selected to simulate the temperature deviations that occur in in every single frame of the image stacks (images were captured the cold chain/storage of refrigerated food products, i.e., abuse for 30 min at 1 fps as described above). On each consecutive day, temperature (Alvarez-Ordóñez et al., 2015). we monitored several fields of view (FOV) such that the total numbers of PI-positive cells (n) studied on days 0–4 were 118, 49, 2.1.3. Cell Preparation and Fluorescence Microscopy 21, 44, and 27, respectively. We analyzed a total of 318 bacteria in Analysis 30 FOV during the 4 days. The lowest and the average number of Pressure-treated cells were pelleted (13,000 g, 1 min) and bacteria analyzed per day were 20 and 45 cells, respectively. We resuspended in Dulbecco’s phosphate-buffered saline (DPBS, pH monitored on average 3 FOV per day and 10 bacteria per FOV. 7.4) immediately before microscopy analysis at 1 h (day 0), and One reason that the cell number was different on different on days 1, 2, 3, and 4 post-treatment. Sixty microliters of this days was that cells at the edges or those with improper orientation cell suspension was pipetted into the center of a glass-bottomed for image processing were ignored. Another reason could have microwell plate (35 mm petri dish, 14 mm microwell, coverglass been a reduction in the number of red cells (cells with PI No. 1.5; MatTek Corporation, Ashland, USA). To immobilize molecules inside) over the 4 days. cells and provide a stationary frame for real-time imaging, 200 The FI curves for each day were grouped into separate clusters µL of a 2% w/v low-gelling-temperature agarose (Sigma-Aldrich, based on the rate of PI diffusion into the cell, as the treated sample Saint Louis, USA) solution was carefully dispensed over the was not homogeneous in terms of pressure-induced damage. cell suspension drop, first surrounding and then covering it. This was because cells might react differently to high pressure; Approximately 2 mL of a staining solution containing 1.25 µM therefore, differences in the degree of damage and ability to repair PI (final concentration; 20 mM stock solution in DMSO, Life would lead to differences in the rate of change of FI and in the rate Technologies/ThermoFisher Scientific, Darmstadt, Germany) in of PI diffusion through pores. We used the k-means algorithm DPBS was carefully dispensed over the cell suspension and to cluster the FI curves, where the optimal number of clusters agarose mixture during real-time imaging. The delivery system (k) was determined using the elbow method (Kassambara, 2017). consisted of a syringe coupled to a plastic tube that was in The procedure was as the following: First, we considered the turn attached to a holed plate lid. Untreated and heat-treated last point of each intensity curve (i.e., FI value at t = 30 (80◦ C for 40 min) cell suspensions were prepared as described min) to create a database for each day. Then we used MATLAB, above and used as a negative and a positive control for PI RRID:_SCR_001622 to run the k-means algorithm for different staining, respectively. numbers of clusters of the database for each day: k = 1, , 2, ..., 10. Time-lapse image acquisition was performed with a Leica For each k, we calculated the sum of the squared distance (SSE) TCS SP5 (Leica Microsystems, Wetzlar, Germany) confocal between the centroid of a cluster and each member of that cluster. microscope using a 63x oil objective (NA 1.4), filters for PI We looked at the SSE as a function of the number of clusters and detection (excitation at 575–625 nm; emission at 660–710 nm), chose a number of clusters so that adding another cluster did not Frontiers in Microbiology | www.frontiersin.org 3 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure improve much better the SSE. The mean and standard deviation of the FI for each cluster of cells were calculated separately for each time point. To normalize the FI values and therefore make them comparable over the 4 days, on each consecutive day we divided the values obtained from the pressure-treated sample by the maximum intensity value achieved from the positive control sample on that day. This normalized FI value was then used to estimate the pore size. 2.3. Statistical Analysis We used MATLAB, RRID:_SCR_001622 to run analysis of variance (one-way ANOVA) and multi-comparison tests to investigate whether the estimated radius for the pore size on each day significantly differed from the values estimated for the other days. In the multi-comparison analysis, two clusters with specific mean values were considered to be significantly different if their intervals were disjoint, and not significantly different if their intervals overlapped. We used the Bonferroni method to calculate the intervals. Additionally, We ran the Bartlett test of the null hypothesis that the estimated radius over the 4 days comes from normal distributions with the same variance. We fitted a weighted least square model (regression model) FIGURE 1 | Morphology of pressure-treated cells. (A–D) TEM images showed that HPP could create membrane pores (Nikparvar et al., 2019). (A) Untreated to the estimated pore values on each day where weights were sample, (B–D) Pressure-treated sample, 400 MPa, 15 min. The black arrows determined according to the number of cells monitored each day. show the location of the pore area on the membrane. (E) A schematic model We used MATLAB, RRID:_SCR_001622 command “fitnlm” to fit showing the morphology of a created membrane pore area under high our linear regression model (y = ax + b, where a and b are the pressure. The geometry of the bacterium and its plasma membrane with fitting parameters) to the estimated values. thickness h is simplified as a rod shape. cin and cout are the concentration of fluorescent molecules inside and outside the cell, respectively, and Din , Dout , and Dpore are the diffusion coefficient of the fluorescent molecules inside, 2.4. A Computational Model to Describe outside, and within the pore area, respectively. Symbols are defined in Membrane Recovery Table 1. We developed a modified version of the model that Zarnitsyn et al. (2008) proposed for the transmembrane diffusion of small molecules through membrane wounds in human cells provided the size of the pore area is greater than the size of after sonication. Our proposed model was constructed based on the molecules. We assumed that the mechanism responsible for several assumptions. the movement of PI molecules was diffusion caused by random First, as stated in the literature (Pagán and Mackey, 2000; molecular motion. The rate of PI diffusion into the cell depends Winter and Jeworrek, 2009; Gänzle and Liu, 2015) and evidenced on the size of the pore area; therefore, it is possible to estimate by our previous work using flow cytometry (Nikparvar et al., the pore size by measuring the difference in FI resulting from the 2019), the bacterial membrane is one of the main structures in increased number of PI molecules bound to DNA. Based on the the cell that is damaged by HPP. We investigated the morphology mass balance of PI molecules, we can write: of the bacterial membrane after exposure to high pressure using transmission electron microscopy (TEM) and found that the membrane was damaged and became perforated under pressure dcin (Figures 1A–D). We assumed that the main damage occurred Vcell = Apore k(cout (t) − cin (t)), (1) dt in a single pore area. This assumption is valid for our model, because the estimated pore area in our work could be interpreted where Vcell is the cell volume (m3 ); cin and cout are the as an effective area (or total area) regardless of the number of intracellular and extracellular concentration of PI molecules pores. It is mathematically possible to estimate the number of (mol/m3 ), respectively; t is time (s); Apore is the pore area (m2 ); large pores by dividing the effective pore area by the size of the and k is the mass transfer coefficient (1/s), which is independent reference molecule (PI), which led us again to one large pore of the pore area and the cell volume but varies with the diffusion area (for all 4 consecutive days). A schematic geometry for the coefficient, D (m2 /s), of PI molecules. The expressions on the described membrane pore is shown in Figure 1E. left and right sides of the equation above (with the unit mol/s) Second, the membrane, which is otherwise impermeable represent the accumulation of PI molecules inside the cell and to fluorescent molecules such as PI, allows the diffusion of the rate of the transport into the cell through the pore, J (mol/s), these molecules within the pressure-induced membrane pore, respectively. If we assume that PI molecules were added to the Frontiers in Microbiology | www.frontiersin.org 4 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure sample at t = 0, cin (t = 0) = 0, and cout (t = 0) changes from 0 TABLE 1 | Definition of symbols. to coutss , solving Equation (1) gives: Symbol Definition Vcell (m3 ) Cell volume cin (t) = coutss (1 − exp(−Kt)), (2) Dout (m2 /s) Extracellular diff. coeff. of PI Din (m2 /s) Intracellular diff. coeff. of PI where K = VAk . Here, we assume that Apore , k, and Vcell are Dpore (m2 /s) Diff. coeff. of PI inside the pore cell constant during the integrating time, i.e., the period between Apore (m2 ) Pore area adding PI molecules to the sample and the time when the Dg (m2 /s) Diff. coeff. of PI in agarose gel FI is saturated (saturation occurred when the intracellular Dw (m2 /s) Diff. coeff. of PI in aqueous solution concentration of PI molecules reached the concentration level cout (mol/m3 ) Extracellular concentration of PI outside, i.e., cin = coutss ). cin (mol/m3 ) Intracellular concentration of PI According to Equation (2) and by measuring cin h (m) Membrane thickness experimentally at time t, the quantity K can be calculated: R (m) Pore radius Ra (m) Radius of PI molecule Rg (m) Radius of gel pores 1 1 K= ln ( ), (3) kB (J/K) Boltzmann constant t 1 − I(t) Imax T (K) Temperature η (N.s/m2 ) Viscosity of plasma membrane where assuming that the relationship between the concentration of fluorescent molecules and the FI is linear (Kim et al., 2020), we have ccinout(t) = II(t) max (I(t) and Imax are the intensity at time t and TABLE 2 | Constant values. ss maximum intensity, respectively). Constant Value Reference Third, as discussed previously by Zarnitsyn et al. (2008), we Vcell 0.7 ∗ 10−18 (m3 ) For E. coli (Yu et al., 2014) considered a three-part process for the diffusion of PI molecules Dw 4 ∗ 10−10 (m2 /s) Stokes-Einstein relationa into a cell: diffusion from the extracellular point to the pore; h 4 ∗ 10 −9 (m) For E. coli (Briegel et al., 2009) diffusion across the cell membrane within the pore; and diffusion Ra 6 ∗ 10−10 (m) (Bowman et al., 2010) away from the pore in the cytosol. The flow J (mol/s) is given by: Rg 8 ∗ 10−7 (m) (Pluen et al., 1999) kB 1.38 ∗ 10−23 (J/K) - cout − cin T 296 (K) - J= h 1 1 , (4) η 5 ∗ 10−9 (N.s/m2 ) (Daniels and Turner, 2007) Dpore Apore + 4Dout R + 4Din R aD = kB T/(6π ηw Ra ), where ηw is the viscosity of water. w where h is the membrane thickness; R represents the pore radius; and Dpore , Dout , and Din are the diffusion coefficients of PI molecules inside the pore, outside the cell, and inside the cell, We used the expression that Verkman (2002) introduced to respectively. Symbols are defined in Table 1. estimate the diffusion coefficient of small molecules (with size By substituting Apore = πR2 (where R is the pore radius), K and mass approximately equal to those of the PI molecule) can be expressed as a function of D and R: in cytoplasm: Din ≈ 0.25Dw . This approximation agrees with the intracellular diffusion coefficient that Zarnitsyn et al. (2008) used to estimate the wound radius in human cell membranes 1 K= . (5) after sonication. 1 1 Vcell ( D hπ R2 + 4Dout R + 4Din R ) As shown in Figure 1E, the diffusion of fluorescent molecules pore within the pore can be modeled as diffusion of molecules in As mentioned earlier, we immobilized the cells in agarose gel; a tube of length h and radius R. We used the expression that therefore, the extracellular diffusion coefficient (Dout ) was the Daniels and Turner (2007) proposed for calculating the diffusion rate of the diffusion for PI molecules in agarose gel (Dg ) and equal coefficient of proteins along long thin membrane tube to to σ Dw , where Dw is the diffusion coefficient in aqueous solution express Dpore : (Table 2). σ was taken from the literature (Pluen et al., 1999): kB T R Ra 3 5 Dpore = [ln( ) + O(1) + O( ) + ...], (7) σ = 1 − 2.1444(Ra /Rg ) + 2.08877(Ra /Rg ) − 0.94813(Ra /Rg ) 4πη Ra R −1.372(Ra /Rg )6 + 3.87(Ra /Rg )8 − 4.19(Ra /Rg )9 , where kB is the Boltzmann constant, T is the temperature in (6) Kelvin, η is the viscosity of the membrane, and Ra is the particle (PI) radius (all values in Table 2). O(.) elements are neglected where Ra and Rg are the radii of the PI molecule and gel pore terms from the Taylor series representation. We obtained the respectively, with the values given in Table 2. measured value for K by inserting I(t) and Imax (Equation 3) and Frontiers in Microbiology | www.frontiersin.org 5 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure tried to fit our simulated value for K (Equation 5) to the measured value for K (Equation 3). By doing so, and by substituting other parameters from the literature and physics of the cells (Table 2), we estimated the parameter R in Equation (5) which gave us the pore radius R. The time dependent value K made the link between our model and experimental data and since we calculated it from the measured data, it specifically affected the final estimation of the pore size, thereby calibrating the model for a specific experimental condition. Fourth, as we followed an identical protocol for microscopy, it was assumed that the impact of photobleaching (if it occurred) did not interfere with the results. 3. RESULTS 3.1. Monitoring the Diffusion of PI Molecules Into Damaged Bacterial Cells Through Membrane Pores To measure the diffusion rate of PI molecules through a damaged membrane, we measured red FI for each bacterial cell (L. monocytogenes) after adding PI molecules for 30 min at a rate of 1 fps (Figure 2). All obtained images and videos can be found in the Dryad Digital Repository (Nikparvar et al., 2020). Owing to variability in cell resistance, cells showed different degrees of pressure-induced damage. We used the k-means algorithm to cluster the intensity curves using a certain number of clusters (determined by the elbow method) for days 0, 1, 2, 3, and 4 after pressure exposure based on the rate of PI diffusion through membrane pores (see section 2.2). The mean intensity curve for each cluster on each evaluation day is shown in Figures 3A–E. In the negative control sample (untreated cells), the cells remained uncolored during the observation period, indicating that PI molecules did not diffuse into the cells (Figure 3F). By contrast, in the positive control sample (heat-treated cells), the intensity reached its maximum value after 30 min, indicating substantial membrane damage. The numbers of untreated, heat-treated, and pressure-treated cells in each cluster for days 0–4 are reported in Table 3. FIGURE 2 | Fluorescence microscopy images for pressure-treated (400 MPa, First, we found that for all evaluation days, the FI started to 8 min, 8◦ C) and control samples. To measure the diffusion rate of PI molecules through a damaged membrane, we measured red FI for each bacterial cell (L. increase after a delay of about 8 min, probably owing to the monocytogenes) after adding PI molecules for 30 min at a rate of 1 fps. All time taken for the PI molecules to diffuse from the extracellular images and videos are available from the Dryad Digital Repository (Nikparvar medium to the cells through the agarose gel. et al., 2020). Second, Equation (2) implies that the higher the slope of the intensity curve, the faster PI molecules diffuse inside the cell. According to Equations (2) and (5), the diffusion rate of PI molecules (slope of the curves in Figure 3) was positively consistent for the cluster of cells with the lowest rate of diffusion, correlated with the membrane pore size. This suggests that in which the final FI value (shown with a dashed line in Figure 3) the curves belonging to the cluster with the lowest slope of changed from 0.14 on day 0 to a final value of 0.02 on day 4, i.e., a the intensity curve correspond to the cells with the smallest seven-fold decrease. This is a strong indication that the pore size pore sizes. The intensity curve for the group with the lowest decreased over the 4 days after the pressure treatment. slope displayed an upward trend toward saturation as the time approached 30 min (Figure 3). Conversely, the clusters with the 3.2. Estimation of the Pore Area highest slope of the intensity curve correspond to the cells with As discussed earlier, we assumed that the cluster corresponding the largest pore sizes. to the curves with the lowest slope of the intensity curve was Most importantly, we detected a general decrease in the rate associated with the least-damaged bacteria. As we could detect of PI diffusion into the cells on consecutive days. This was a decay in the rate of PI uptake for this group of cells from day 0 Frontiers in Microbiology | www.frontiersin.org 6 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure FIGURE 3 | Intensity plots. The normalized mean and standard deviation of the red fluorescence intensity, FI, emitted from: (A–E) each pressure-treated cell on days 0–4 following HPP, and (F) two control samples. The FI values were extracted from images captured by the fluorescence microscopy technique for 30 min after PI addition. We used the k-means algorithm to cluster the intensity curves through a certain number of clusters (determined by the elbow method) for days 0, 1, 2, 3, and 4 after pressure exposure based on the rate of PI diffusion through membrane pores (see 2.2). The y-axis is in log scale. The intensity value after 30 min for the group of cells with the lowest slope (shown in blue) decreased with the days from 0.14 to 0.02, represented by the dashed lines. Intensity values for each cell (raw data) are available in Supplementary File 1. TABLE 3 | Number of studied untreated (Untr), heat-treated (H-T), and SD = 0.0014) was significantly (p < 10−6 ) higher than on the pressure-treated (PI-positive) cells in each cluster on days 0–4 (D0-D4). other days (Figure 4A). The mean values of the pore radius for Cluster D0 D1 D2 D3 D4 Untr H-T day 1 (mean = 1.191, SD = 0.0039) and day 2 (mean = 1.221, SD = 0.0041) were significantly different (p < 10−6 ) from those Lowest slope 98 12 11 10 7 – on days 3 (mean = 0.964, SD = 0.0043) and 4 (mean = 0.809, Mild slopea 5 28 3 30 12 – SD = 0.0051). Although we observed a general reduction in the Highest slope 15 9 7 4 8 – PI uptake during the 4 days (Figure 4B), we did not observe a Total 118 49 21 44 27 20 39 reduction from day 1 to day 2 probably owing to high variability in the results or due to other reasons such as a lag time before a The number of cells in the mild-slope cluster equals to the number of cells belonging to the clusters other than the lowest- and the highest-slope clusters. starting the recovery process. Furthermore, because of the small sample size on day 4 (due to a low number of PI-positive cells detected on day 4 which most likely resulted from the partial recovery of the membrane), we could not show a significant to day 4, we used this group to calibrate the model that estimates reduction from day 3 to day 4. Estimated pore size values for the the pore size. cluster with the lowest slope are presented in Table 4. The decay By substituting the FI value corresponding to this cluster for in the pore size suggests the presence of an embedded membrane each evaluation day into Equations (3–7), the mean value for repair mechanism in bacteria, which was activated in response pore size was estimated. First, the results of the model calibration to HPP (see section 4). The Bartlett test reported insufficient showed that the mean radius of the pore decreased from 1.3 to evidence to say the variances are different with a P-value higher 0.8 nm over the 4 days (Figure 4A). than 0.05. Second, we ran one-way ANOVA and multi-comparison tests Once the model was calibrated, we fitted a weighted least using the Bonferroni method (see section 2.3) and found that the squares model to the estimated pore size values on each day mean value of the estimated pore radius on day 0 (mean = 1.338, for the cluster with the lowest slope (Figure 5A). Based on the Frontiers in Microbiology | www.frontiersin.org 7 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure FIGURE 4 | Pore size estimation. (A) Estimated pore radius of the cells belonging to the cluster with the lowest slope. The mean value of pore radius for day 0 was significantly (p < 10−6 ) different than the rest of the days. The mean value of the pore radius for days 1–2 was significantly different than days 3–4 (stars represent statistically different groups). Error bars show the confidence intervals, calculated using the Bonferroni method for multi-comparison (see 2.3). (B) The last frame (t = 30 min monitoring) of the fluorescence microscopy image stacks obtained in each evaluation day. Here only one cell as a representative for each day is shown. TABLE 4 | Estimated pore radius for the cluster with the lowest slope (Figure 4). solutes, such as ions, amino acids, ATP, etc. (size <700 Da) could pass through 1-nm pores. Larger molecules such as ribosomes Day post-pressure R (nm) Confidence intervala (nm) and DNA were found to pass through pores larger than 2 nm D0 1.338 (1.296–1.380) (Figure 6). Although estimation of cell life expectancy requires D1 1.191 (1.113–1.269) extensive statistical experiments, it can be safely assumed that D2 1.221 (1.139–1.303) large pore sizes can cause leakage of larger essential compounds, D3 0.964 (0.878–1.051) e.g., DNA, from the cell through pores (Figure 6). In other D4 0.809 (0.702–0.915) words, the cell is more likely to die when the pore sizes a The Bonferroni method. are large. To investigate how sensitive the model prediction R was to different parameters (e.g., Vcell , σ , η, h, and Din ), we perturbed each parameter by 1p = ±10%, while the input cin was 50% of assumption that cells with different pore sizes in the range <5 its maximum value, 1 (local sensitivity analysis). We found that nm are repaired at the same rate (Zarnitsyn et al., 2008), we used the model was sensitive to Vcell , η, and h. The relative sensitivity the same weighted least squares model to fit the remaining cell R −R p groups (Figure 5B). Given an initial FI value on day 0, this model coefficient (S = | (p) 1p(p+1p) · R |) is listed in Table 5, where p is could predict the pore radius as a function of time. It was then the parameter and 1p is its perturbation. possible to estimate the time that the bacterial cell needed to To evaluate how sensitive our linear regression model was to regain its membrane integrity. In our case, this occurred when variations in the cell volume (i.e., the most sensitive parameter the pore radius became smaller than the radius of the PI molecule in our model; Table 5), we perturbed the parameter Vcell (1p = (dashed horizontal line, Figure 5B). Although extrapolation ±10% and ±50%) and checked the deviation in the predicted outside the region of experimental measurements may risk value R each day. Importantly, for both 1p = ±10% and entering different dynamics/regions, where the assumptions do 1p = ±50%, the local sensitivity of the regression model to not apply anymore, we believe that such error will not matter in the perturbation decreased over time (Figure 7), such that the our work because this is when the pores are too small to have estimated time for regaining the membrane integrity did not consequences on the cell (death). alter significantly. We also examined the other two sensitive This information could be useful for food industry to design parameters, i.e., membrane thickness h and viscosity of the more efficient pressure treatments by adjusting HPP strategy (the plasma membrane η, and found similar results. These results pressure strength and holding time). together with the fact that the size of L. monocytogenes is in Figure 6 gives a better understanding of pore size in terms the range defined by 10 and 50% perturbation from the size of allowing for the leakage of solutes with different molecule of E. coli (Jamshidi and Zeinali, 2019) indicate that the model sizes through the pore. A literature search revealed that small predictions of temporal recovery were robust to the uncertainty Frontiers in Microbiology | www.frontiersin.org 8 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure FIGURE 5 | Regression model for the pore size. (A) A weighted regression model showed that the estimated values of the pore size for the cluster with the lowest slope in the intensity curve decayed linearly over the four days. (B) The linear model was fitted to the data of the remaining clusters. Error bars in both (A,B) show the standard deviation for the estimated pore radius. The solid blue line shows the fitted linear model to the data for the least damaged cells (the cluster with the lowest slope in the intensity curve) and the solid gray lines show the predicted linear approach for resealing pores over evaluation days for the rest of the clusters. With this linear model, we could predict the time interval needed for each group to recover the membrane. FIGURE 6 | Leak of solutes through small pores. Literature research in filter membranes shows that solutes such as glucose, some ions, and ATP are small enough to pass through pores up to 1 nm size. Bigger pores may allow pass of bigger molecules such as DNA and dextrans1 . of parameters such as Vcell (which was substituted from values Exposure of L. monocytogenes to 400 MPa, 8 min, 8◦ C specified for E. coli, Table 2). led to a 7.79 ± 0.82 log CFU/mL decrease in viable cell counts, corresponding to a cell concentration below the 3.3. Population Growth Behavior of LOQ of the method (Figure 8). Viable cell counts in the Pressure-Treated L. monocytogenes pressure-treated sample remained constant and below the LOQ Absence of growth at the population level during the recovery during the subsequent 4 days, whereas the growth of the period was confirmed by measurement of viable plate counts. total population was only observed 7 days after HPP. This trend is compatible with the existence of a lag phase of 1 London Health Science Centre (2020). https://www.lhsc.on.ca/critical-care- at least 4 days, followed by the onset of the exponential trauma-centre/principles-of-crrt/ (accessed July 24, 2020). growth phase. Frontiers in Microbiology | www.frontiersin.org 9 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure 4. DISCUSSION distribution of red FIs for cells taking up PI, probably arising from different structural strengths of the cellular envelope. Entry We exposed L. monocytogenes cells to HPP of 400 MPa, 8 of a bacterial population into the stationary phase, as analyzed min, 8◦ C. Our results showed that some of the cells in the in this study, is related to an increased resilient cell envelope sample that were exposed to HPP became permeable to PI and the synthesis of stress response proteins (Casadei et al., molecules, which otherwise could not pass through the intact 2002; Huang et al., 2014). However, this phenotype is subject membrane, suggesting the formation of pores in the cellular to various possible drivers of intrapopulation heterogeneity, envelope under high pressure. Although the exact mechanism including stochasticity in gene expression, the effects of which of pressure-induced membrane permeabilization is not known, it has been linked to the denaturation of membrane proteins, as well as to the phase transition of membrane phospholipids from the physiological liquid-crystalline phase to the gel phase, which results from lateral compression and increased packing of the acyl chains (Pagán and Mackey, 2000; Casadei et al., 2002; Winter and Jeworrek, 2009; Patterson, 2014). To evaluate the degree of membrane damage, we measured the rate of FI change inside each cell after adding PI molecules to the extracellular medium. First, we detected large variations in the degree of membrane damage among single cells, which were clustered into groups (Figure 3) using the k-means algorithm (see section 2.2). This was consistent with a study by Ritz et al. (2001), in which HPP followed by PI staining and flow cytometry analysis of L. monocytogenes revealed a broad TABLE 5 | Relative sensitivity coefficient for 10% perturbation in each parameter. Parameter Relative sensitivity |S| Vcell 0.3799 FIGURE 8 | L. monocytogenes viable cell counts (log CFU/mL). Assessment η 0.3798 was done by plating on non-selective medium before (untreated) and after h 0.3798 HPP (day 0) and during storage at 8◦ C (day 1–7). LOD, limit of detection; LOQ, limit of quantification. Error bars show standard deviation of three biologically σ 9.758 ∗ 10−6 independent replicates. Raw data is available in Supplementary File 2. Din 3.188 ∗ 10−5 FIGURE 7 | Parametric uncertainty of the linear prediction for the cell volume (Table 5) diminished with pore size. (A) 10% perturbation and (B) 50% perturbation in p = Vcell = 0.7 ∗ 10−18 (m3 ). Shaded bounds show the deviation of the predicted R values from its nominal values (black solid line) due to parameter (Vcell ) uncertainty. For both perturbations of 10 and 50% of the nominal parameter value, the uncertainty region diminished as pores were resealed. Frontiers in Microbiology | www.frontiersin.org 10 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure are amplified for low-abundance molecules such as mRNAs and area, additional pressure buildup will most likely concentrate regulatory proteins; cell cycle and aging stage; and epigenetic around the same area which is structurally weaker than the regulation (Avery, 2006; Veening et al., 2008). Moreover, remaining parts of the membrane. Finally, finer resolution phenotypic heterogeneity is an inherent feature of bacterial of pore sizes can be estimated by using several fluorescent populations and, most importantly, is considered to be a molecules as in a previous work (Zarnitsyn et al., 2008) the component of their adaptation and survival strategy (Booth, authors compensated for this assumption (single pore area) by 2002). using a series of fluorescent molecules with different sizes. The Second, the synergy of our mass transfer mechanistic model linear relationship between the concentration of PI molecules (Equations 1–7) with specially tailored experiments to test and the FI is another fundamental assumption in our model the membrane recovery hypothesis allowed us to estimate the development. This assumption may limit the accuracy of the total size of the pressure-created pores at any time point after result if saturation occurs i.e., when the increase in the PI treatment at 400 MPa, 8 min, 8◦ C. Our observations indicated concentration does not affect the intensity anymore. However, that the rate of diffusion of PI, particularly in the cluster with because we used heat-treated bacteria as positive control cells the lowest slope (corresponding to the least-damaged cells), (see 2.1.3), in which the membrane damaged was highest, we was reduced during the 4 days after exposure (Figure 4). The guaranteed that for each day the intensity in pressure-treated existence of a recovery process in the bacterial membrane bacteria was not saturated. investigated in this work is consistent with the results of our The intensity curves for the cluster with the lowest slope previous works (Duru et al., 2021; Nikparvar et al., 2021), where (Figure 3) displayed an increasing trend until reaching a near- by using a time-series RNA sequencing data and conducting a saturation level. However, as the intensity approaches saturation, network component analysis we proved the presence of a repair the sensitivity of the signal decreases. An inaccurate saturation process in the membrane after HPP. value may cause to underestimate the pore size through Most importantly, we found that the estimated pore size Equation (5), particularly for days 0–2. For days 3 and 4, the decayed (most likely) linearly as a function of time (Figure 5). intensity curve for the lowest-slope cluster stayed at a steady If the trend is not linear, then the rate of repair (decay of size) state after 30 min and therefore the estimation of the pore size must be increased (e.g., log decay) or decreased (e.g., exponential for these 2 days is more accurate. However, the underestimation decay). We see no reason for an accelerated repair because cells of the pore size on days 0–2 does not affect our conclusion of will likely mobilize most of their resources to repair damage when the existence of a recovery process, because even if the actual it is at its maximum level. There is no reason to assume that the pore sizes for days 0–2 were larger than the estimated values, the cells will increase their repair resources (and thus the rate) as damage was still recovered over days 3 and 4. the pore size decreases. An exponential decay (decelerating rate Leakage of intracellular components due to membrane of repair) is otherwise more likely, but as pores are becoming damage following various treatments with antibiotics, smaller, the repair process will likely proceed faster. However, bacteriocins, or high pressure was demonstrated for both at that stage the cells likely divert resources to other essential Gram-positive and Gram-negative bacteria. In an early study, damaged sites than the membrane, and therefore the rate remains leakage of low-molecular-weight solutes upon treatment of nearly linear. The linear repair rate is consistent with a previous E. coli with different concentrations of 2-phenoxyethanol study of membrane recovery in human cells (Zarnitsyn et al., was proposed as an indicator of the disorganization of the 2008), in which the authors measured membrane wound closure cytoplasmic membrane (Gilbert et al., 1977). In line with this, using several fluorescent molecules. They found an exponential membrane damage induced by poly-hexamethylene biguanides decay for wound sizes on the scale of hundreds of nanometers resulted in leakage of potassium ions and inorganic phosphates over time, followed by an approximately linear decay for wound in E. coli (Broxton et al., 1983). Lambert and Hammond (1973) sizes less than 20 nm. showed leakage of potassium to be a primary indicator of The model suggested in this work for quantifying the membrane damage. Depending on the extent of membrane membrane damage is based on several assumptions (see 2.4). The damage, larger solutes such as ATP or DNA (500–700 Da) are assumption that the main pressure-induced membrane damage also released from the cell (Zhen et al., 2013; Singh et al., 2016). occurred in a single effective pore area (total area of membrane Although bacteria possess membrane repair mechanisms, the pores) may affect the prediction of the size of individual pores. duration and the extent of membrane damage (i.e., pore size) To date, pores smaller than fluorescent molecules (such as PI) may lead to bacterial cell death (Wortman and Bissonnette, 1988; are difficult to measure by direct methods. Thus, the model Vigouroux et al., 2020). As the model developed in this work predictions may appear biased when the number of small pores predicts the timely repair of pores, it could be integrated with (smaller than PI size) is large. Importantly, the total diffusion of future work to estimate the cumulative probability of survival molecules through the cell membrane is only dependent on the and life expectancy. total surface of the pores (Zarnitsyn et al., 2008), so cell fate, We did not observe a marked increase in the number of cells which is strongly affected by the total diffusion via the loss of able to form colonies in the culture medium during the 4 days of cytosol material, will not be affected by the wrong estimation the experiments, i.e., cell counts remained stable and below the of the number of pores. Additionally, there is no evidence LOQ. However, as we could observe a decay in the membrane supporting that high pressure causes multiple small pores: once pore radius by time, we inferred that the growth process (mass high pressure damages the membrane structure in a certain accumulation and division) of these individual cells was arrested, Frontiers in Microbiology | www.frontiersin.org 11 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure yet they were not dead. Although a population growth curve does PI molecules to the pressure-treated bacteria at time point not provide information about the physiological state of cells 0 (immediately after treatment) and on days 1, 2, 3, and giving rise to the exponential growth phase, single-cell studies 4 after HPP, and measured the FI emitted by DNA-bound have revealed that the presence of a non-growing fraction of cells PI molecules using a fluorescence microscopy technique. We is the main cause of the extended apparent population lag when developed a dynamic model to quantify the degree of damage stress conditions at the growth limit are applied (Koutsoumanis, in pressure-treated bacteria. The synergy between our diffusion 2008; Aguirre and Koutsoumanis, 2016), as in the present study. model and microscopy experiments revealed that some L. Moreover, several studies have shown an increased length of monocytogenes cells exposed to HPP repaired their damaged the lag phase, both at the population and single-cell levels, after membrane approximately linearly on a time scale of days. This exposure of L. monocytogenes to stress conditions, including high is the first time that membrane pores created by HPP have been pressure (Robinson et al., 1998; Francois et al., 2006; Muñoz- quantitatively described and shown to diminish. Cuevas et al., 2013). The increased time required to start division can be attributed to the metabolic processes needed for repair DATA AVAILABILITY STATEMENT of damaged cell components and is therefore indicative of the presence of sub-lethally injured cells (Guillier et al., 2005; Métris The original contributions presented in the study et al., 2008). In our experiments, cell counts exceeded the LOQ are publicly available. This data can be found here: only after 7 days. Thus, it is likely that the cells that resealed https://datadryad.org/stash/share/bP2iD9QM03-5tCl8njy_0BNr their membrane pores started to proliferate and, along with a cQAHgX5dZvCwznZwmvw. small fraction of non-injured cells (Aguirre and Koutsoumanis, 2016), contributed to the observed growth at the population AUTHOR CONTRIBUTIONS level. Our observation of growth arrest at the first days after pressure treatment led us to hypothesize that for each day the BN: conceptualization, methodology, software, formal analysis, cells in the group with the lowest slope came either from the cells investigation, and writing—original draft, visualization. AS: belonging to the lowest-slope group in previous days or injured methodology, investigation, and writing—original draft. MC: cells from other groups such as mild-slope group that transferred resources, writing—review and editing, and supervision. MH-H: to the lowest-slope group as they managed to recover partially. resources and supervision. PC: visualization and writing—review Therefore, we knew that the less number of PI-positive cells and editing. CR: writing—review and editing and supervision. obtained over the 4 days could not be due to proliferation of new NB: conceptualization, validation, formal analysis, investigation, cells but likely due to an increasing number of recovered cells. resources, writing—review and editing, supervision, and project We note that it is impossible to infer the rate of membrane administration. All authors contributed to the article and damage repair from the microscopy observations alone, because approved the submitted version. of the large biological variations and the number of cell clusters (Figure 3). It was the synergy of our dynamic model, trained by FUNDING our experimental data and tested on a subset of experiments, that enabled us to calculate the repair rate of any pore size on This study was supported within the ERA-IB2 consortium this scale (<20 nm) until the cellular envelope was repaired. SafeFood (ID: ERA712IB-16- 247 014) by grants of the the This could help to predict the duration time of growth arrest Research Council of Norway (to NB, grant number 263499), the after exposure to HPP until the cells restart the growth process. Spanish Government (to MC, grant number PCIN-2016-072), It should be noted that the results obtained in this work were and the German Ministry for Education and Research (to CR, specifically related to the pressure treatment of L. monocytogenes grant number 031B0268). The funders had no role in the design in 400 MPa, 8 min, 8◦ C, thereby generalizing the result to of the study and collection, analysis, and interpretation of data other HPP conditions and other microorganisms may not be and in writing the manuscript. valid. However, what we presented in this study may pave the road for future works such that the method suggested here can ACKNOWLEDGMENTS be applied for other HPP conditions or other microorganisms and fluorescent molecules. Repeating similar experiments with The authors thank Judit Pampalona Sala for technical assistance different pressure values and holding times can increase the with fluorescence microscopy experiments performed at Servei fidelity of the model. de Microscopia at Autonomous University of Barcelona and Pedro A. Lira-Parada and Dr. Gerd M. Seibold for discussions. 5. CONCLUSION SUPPLEMENTARY MATERIAL The recovery process in bacteria after exposure to high pressure has not been investigated well. Here, we focused on foodborne The Supplementary Material for this article can be found pathogenic bacteria L. monocytogenes and the effects of high online at: https://www.frontiersin.org/articles/10.3389/fmicb. pressure (400 MPa, 8 min, 8◦ C) on its membrane. We added 2021.598739/full#supplementary-material Frontiers in Microbiology | www.frontiersin.org 12 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure REFERENCES Guillier, L., Pardon, P., and Augustin, J. (2005). Influence of stress on individual lag time distributions of Listeria monocytogenes. Appl. Environ. Microbiol. 71, Aguirre, J., and Koutsoumanis, K. (2016). Towards lag phase of microbial 2940–2948. doi: 10.1128/AEM.71.6.2940-2948.2005 populations at growth-limiting conditions: the role of the variability in Hereu, A., Dalgaard, P., Garriga, M., Aymericha, T., and Bover-Cida, S. (2014). the growth limits of individual cells. Int. J. Food Microbiol. 224, 1–6. Analysing and modelling the growth behaviour of Listeria monocytogeneson doi: 10.1016/j.ijfoodmicro.2016.01.021 RTE cooked meat products after a high pressure treatment at 400 MPa. Int. Alpas, H., Kalchayanand, N., Bozoglu, F., Sikes, A., Dunne, C., and Ray, J. Food Microbiol. 186, 84–94. doi: 10.1016/j.ijfoodmicro.2014.06.020 B. (1999). Variation in resistance to hydrostatic pressure among strains Huang, H.-W., Lung, H.-M., Yang, B. B., and Wang, C.-Y. (2014). Responses of food-borne pathogens. Appl. Environ. Microbiol. 9, 4248–4251. of microorganisms to high hydrostatic pressure processing. Food Control 40, doi: 10.1128/AEM.65.9.4248-4251.1999 250–259. doi: 10.1016/j.foodcont.2013.12.007 Alvarez-Ordóñez, A., Broussolle, V., Colin, P., Nguyen-The, C., and Prieto, Jamshidi, A., and Zeinali, T. (2019). Significance and characteristics of M. (2015). The adaptive response of bacterial food-borne pathogens in the Listeria monocytogenes in poultry products. Int. J. Food Sci. 2019, 1–7. environment, host and food: implications for food safety. Int. J. Food Microbiol. doi: 10.1155/2019/7835253 213, 99–109. doi: 10.1016/j.ijfoodmicro.2015.06.004 Jofré, A., Aymerich, T., Bover-Cid, S., and Garriga, M. (2010). Inactivation and Avery, S. (2006). Microbial cell individuality and the underlying sources of recovery of Listeria monocytogenes, Salmonella enterica and Staphylococcus heterogeneity. Nat. Rev. Microbiol. 4, 577–587. doi: 10.1038/nrmicro1460 aureus after high hydrostatic pressure treatments up to 900 MPa. Int. Microbiol. Booth, I. (2002). Stress and the single cell: intrapopulation diversity is a mechanism 13, 105–112. doi: 10.2436/20.1501.01.115 to ensure survival upon exposure to stress. Int. J. Food Microbiol. 78, 19–30. Kassambara, A. (2017). Practical Guide to Cluster Analysis in R, 1 Edn. CreateSpace doi: 10.1016/S0168-1605(02)00239-8 Independent Publishing Platform, Scotts Valley, California. Bover-Cid, S., Belletti, N., Garriga, M., and Aymerich, T. (2010). Model for Listeria Kim, Y., Shin, E., Jung, W., Kim, M. K., and Chong, Y. (2020). A near-infrared monocytogenes inactivation on dry-cured ham by high hydrostatic pressure turn-on fluorescent sensor for sensitive and specific detection of albumin from processing. Food Microbiol. 28, 804–809. doi: 10.1016/j.fm.2010.05.005 urine samples. Sensors 20:1232. doi: 10.3390/s20041232 Bowman, A., Nesin, O., Pakhomova, O., and Pakhomov, A. (2010). Analysis of Klotz, B., Mañas, P., and Mackey, B. (2010). The relationship between plasma membrane integrity by fluorescent detection of Tl+ uptake. Membr. membrane damage, release of protein and loss of viability in Escherichia coli Biol. 236, 15–26. doi: 10.1007/s00232-010-9269-y exposed to high hydrostatic pressure. Int. J. Food Microbiol. 137, 214–220. Bozoglu, F., Alpas, H., and Kaletunç, G. (2004). Injury recovery of foodborne doi: 10.1016/j.ijfoodmicro.2009.11.020 pathogens in high hydrostatic pressure treated milk during storage. FEMS Koseki, S., Mizuno, Y., and Yamamoto, K. (2007). Predictive modelling Immunol. Med. Microbiol. 40, 243–247. doi: 10.1016/S0928-8244(04)00002-1 of the recovery of Listeria monocytogenes on sliced cooked ham Briegel, A., Ortega, D., Tocheva, E., Wuichet, K., Li, Z., Chen, S., et al. (2009). after high pressure processing. Int. J. Food Microbiol. 119, 300–307. Universal architecture of bacterial chemoreceptor arrays. Proc. Natl. Acad. Sci. doi: 10.1016/j.ijfoodmicro.2007.08.025 U.S.A. 106, 17181–17186. doi: 10.1073/pnas.0905181106 Koutsoumanis, K. (2008). A study on the variability in the growth limits of Briers, Y., Klumpp, J., Schuppler, M., and Loessner, M. (2011). Genome sequence individual cells and its effect on the behavior of microbial populations. Int. J. of Listeria monocytogenes Scott A, a clinical isolate from a food-borne listeriosis Food Microbiol. 128, 116–121. doi: 10.1016/j.ijfoodmicro.2008.07.013 outbreak. J. Bacteriol. 193, 4284–4285. doi: 10.1128/JB.05328-11 Lambert, P. A., and Hammond, S. M. (1973). Potassium fluxes, first indications of Broxton, P., Woodcock, P. M., and Gilbert, P. (1983). A study of the antibacterial membrane damage in micro-organisms. Biochem. Biophys. Res. Commun. 54, activity of some polyhexamethylene biguanides towards Escherichia coli ATCC 796–799. 8739. J. Appl. Bacteriol. 54, 345–353. doi: 10.1111/j.1365-2672.1983.tb02627.x Mañas, P., and Mackey, B. (2004). Morphological and physiological changes Casadei, M., Mañas, P., Niven, G., Needs, E., and Mackey, B. (2002). Role of induced by high hydrostatic pressure in exponential- and stationary-phase cells membrane fluidity in pressure resistance of Escherichia coli nctc 8164. Appl. of Escherichia coli: Relationship with cell death. Appl. Environ. Microbiol. 70, Environ. Microbiol. 68, 5965–5972. doi: 10.1128/AEM.68.12.5965-5972.2002 1545–1554. doi: 10.1128/AEM.70.3.1545-1554.2004 Daniels, D., and Turner, M. (2007). Diffusion on membrane tubes: a highly Métris, A., George, S., Mackey, B., and Baranyi, J. (2008). Modeling the discriminatory test of the saffman-delbruck theory. Langmuir 23, 6667–6670. variability of single-cell lag times for Listeria innocua populations after doi: 10.1021/la0635000 sublethal and lethal heat treatments. Appl. Environ. Microbiol. 74, 6949–6955. Davey, H., and Hexley, P. (2011). Red but not dead? Membranes of stressed doi: 10.1128/AEM.01237-08 Saccharomyces cerevisiae are permeable to propidium iodide. Environ. Muñoz-Cuevas, M., Guevara, L., Aznar, M., Martínez, A., Periago, P., and Microbiol. 13, 163–171. doi: 10.1111/j.1462-2920.2010.02317.x Fernández, P. (2013). Characterisation of the resistance and the growth Duru, I. C., Andreevskaya, M., Laine, P., Rode, T. M., Ylinen, A., Løvdal, variability of Listeria monocytogenes after high hydrostatic pressure treatments. T., et al. (2020). Genomic characterization of the most barotolerant Food Control 29, 409–415. doi: 10.1016/j.foodcont.2012.05.047 Listeria monocytogenes RO15 strain compared to reference strains Muntean, M.-V., Marian, O., Barbieru, V., Cătunescu, G. M., Ranta, O., used to evaluate food high pressure processing. BMC Genomics 21:455. Drocas, I., et al. (2016). High pressure processing in food industry doi: 10.1186/s12864-020-06819-0 – characteristics and applications. Agric. Agric. Sci. Proc. 10, 377–383. Duru, I. C., Bucur, F. I., Andreevskaya, M., Nikparvar, B., Ylinen, A., Grigore- doi: 10.1016/j.aaspro.2016.09.077 Gurgu, L., et al. (2021). High-pressure processing-induced transcriptome Nikparvar, B., Andreevskay, M., Duru, I. C., Bucur, F. I., Grigore-Gurgu, L., Borda, response during recovery of Listeria monocytogenes. BMC Genomics 22:117. D., et al. (2021). Analysis of temporal gene regulation of Listeria monocytogenes doi: 10.1186/s12864-021-07407-6 revealed distinct regulatory response modes after exposure to high pressure Francois, K., Devlieghere, F., Standaert, A., Geeraerd, A., Impe, J. V., and Debevere, processing. BMC Genomics 22:266. doi: 10.1186/s12864-021-07461-0 J. (2006). Effect of environmental parameters (temperature, ph and aw) on Nikparvar, B., Subires, A., Capellas, M., Hernandez, M., and Bar, N. (2019). “A the individual cell lag phase and generation time of Listeria monocytogenes. dynamic model of membrane recovery mechanisms in bacteria following high Int. J. Food Microbiol. 108, 326–335. doi: 10.1016/j.ijfoodmicro.2005. pressure processing,” in Proceedings of the 12th IFAC Symposium on Dynamics 11.017 and Control of Process Systems, including Biosystems DYCOPS (Florianópolis), Gänzle, M., and Liu, Y. (2015). Mechanisms of pressure-mediated cell death 243–250. and injury in Escherichia coli: from fundamentals to food applications. Front. Nikparvar, B., Subires, A., Capellas, M., Hernandez-Herrero, M., Crauwels, Microbiol. 6:599. doi: 10.3389/fmicb.2015.00599 P., Riedel, C.U., et al. (2020). Images obtained by fluorescence microscopy Gänzle, M., and Vogel, R. (2001). On-line fluorescence determination of pressure technique for monitoring diffusion of PI molecules into pressure-treated mediated outer membrane damage in Escherichia coli. Syst. Appl. Microbiol. 24, Listeria monocytogenes cells, Dryad Digital Repository (dataset), Available 477–485. doi: 10.1078/0723-2020-00069 online at: https://datadryad.org/stash/share/bP2iD9QM03-5tCl8njy_ Gilbert, P., Beveridge, E. G., and Crone, P. B. (1977). The lethal action of 2- 0BNrcQAHgX5dZvCwznZwmvw phenoxyethanol and its analogues upon Escherichia coli NCTC 5933. Microbios Pagán, R., and Mackey, B. (2000). Relationship between membrane damage 19, 125–141. and cell death in pressure-treated Escherichia coli cells: differences between Frontiers in Microbiology | www.frontiersin.org 13 May 2021 | Volume 12 | Article 598739 Nikparvar et al. Response of Listeria monocytogenes to High Pressure exponential- and stationary-phase cells and variation among strains. Appl. Verkman, A. (2002). Solute and macromolecule diffusion in Environ. Microbiol. 66, 2829–2834. doi: 10.1128/AEM.66.7.2829-2834.2000 cellular aqueous compartments. Trends Biochem Sci. 27, 27–33. Patterson, M. (2014). Food technologies: high pressure processing. Food Sci. 3, doi: 10.1016/S0968-0004(01)02003-5 196–201. doi: 10.1016/B978-0-12-378612-8.00262-6 Vigouroux, A., Cordier, B., Aristov, A., Alvarez, L., Özbaykal, G., Chaze, T., Pluen, A., Netti, P., Jain, R., and Berk, D. (1999). Diffusion of macromolecules in et al. (2020). Class-A penicillin binding proteins do not contribute to agarose gels: comparison of linear and globular configurations. Biophys. J. 77, cell shape but repair cell-wall defects. eLife 9:e51998. doi: 10.7554/eLife. 542–552. 51998 Ritz, M., Tholozan, J., Federighi, M., and Pilet, M. (2001). Morphological Winter, R., and Jeworrek, C. (2009). Effect of pressure on membranes. R. Soc Ch 5, and physiological characterization of Listeria monocytogenes subjected 3157–3173. doi: 10.1039/b901690b to high hydrostatic pressure. Appl. Environ. Microbiol. 67, 2240–2247. Wortman, A. T., and Bissonnette, G. K. (1988). Metabolic processes involved in doi: 10.1128/AEM.67.5.2240-2247.2001 repair of Escherichia coli cells damaged by exposure to acid mine water. Appl. Robinson, T., Ocio, M., Kaloti, A., and Mackey, B. (1998). The effect of the growth Environ. Microbiol. 54, 1901–1906. doi: 10.1128/AEM.54.8.1901-1906.1988 environment on the lag phase of Listeria monocytogenes. Int. J. Food Microbiol. Yang, Y., Xiang, Y., and Xu, M. (2015). From red to green: the propidium iodide- 44, 83–92. permeable membrane of Shewanella decolorationis S12 is repairable. Sci. Rep. Rubio, B., Possas, A., Rincón, F., García-Gímeno, R. M., and Martínez, B. 5:18583. doi: 10.1038/srep18583 (2018). Model for Listeria monocytogenes inactivation by high hydrostatic Yu, A., Loo, J., Yu, S., and S Kong, T.-F. C. (2014). Monitoring bacterial pressure processing in spanish chorizo sausage. Food Microbiol. 69, 18–24. growth using tunable resistive pulse sensing with a pore-based technique. doi: 10.1016/j.fm.2017.07.012 Appl. Microbiol. Biotechnol. 98, 855–862. doi: 10.1007/s00253-013-5 Shi, L., Günther, S., Hübschmann, T., Wick, L., Harms, H., and Müller, S. (2007). 377-9 Limits of propidium iodide as a cell viability indicator for environmental Zarnitsyn, V., Rostad, C., and Prausnitz, M. (2008). Modeling transmembrane bacteria. Cytom. Part A 71A, 592–598. doi: 10.1002/cyto.a.20402 transport through cell membrane wounds created by acoustic cavitation. Singh, M., Mallick, A. K., Banerjee, M., and Kumar, R. (2016). Loss Biophys. J. 95, 4124–4138. doi: 10.1529/biophysj.108.131664 of outer membrane integrity in Gram-negative bacteria by silver Zhen, H., Han, T., Fennell, D. E., and Mainelis, G. (2013). Release of free nanoparticles loaded with Camellia sinensis leaf phytochemicals: plausible DNA by membrane-impaired bacterial aerosols due to aerosolization and mechanism of bacterial cell disintegration. Bull. Mater. Sci. 39, 1871–1878. air sampling. Appl. Environ. Microbiol. 79, 7780–7789. doi: 10.1128/AEM.02 doi: 10.1007/s12034-016-1317-5 859-13 Smelt, J., Rijke, A., and Hayhurst, A. (1994). Possible mechanism of high pressure inactivation of microorganisms. High Pressure Res. 12, 199–203. Conflict of Interest: The authors declare that the research was conducted in the Subires, A., Yuste, J., and Capellas, M. (2013). Flow cytometry immunodetection absence of any commercial or financial relationships that could be construed as a and membrane integrity assessment of Escherichia coli o157:h7 in ready-to- potential conflict of interest. eat pasta salad during refrigerated storage. Int. J. Food Microbiol. 169, 47–56. doi: 10.1016/j.ijfoodmicro.2013.10.013 Copyright © 2021 Nikparvar, Subires, Capellas, Hernandez-Herrero, Crauwels, Valdramidis, V., Patterson, M., and Linton, M. (2015). Modelling the recovery of Riedel and Bar. This is an open-access article distributed under the terms of Listeria monocytogenes in high pressure processed simulated cured meat. Food the Creative Commons Attribution License (CC BY). The use, distribution or Control 47, 353–358. doi: 10.1016/j.foodcont.2014.07.022 reproduction in other forums is permitted, provided the original author(s) and the Veening, J.-W., Smits, W., and Kuipers, O. (2008). Bistability, epigenetics, copyright owner(s) are credited and that the original publication in this journal and bet-hedging in bacteria. Annu. Rev. Microbiol. 62, 193–210. is cited, in accordance with accepted academic practice. No use, distribution or doi: 10.1146/annurev.micro.62.081307.163002 reproduction is permitted which does not comply with these terms. Frontiers in Microbiology | www.frontiersin.org 14 May 2021 | Volume 12 | Article 598739