Modelling growth of two Listeria monocytogenes strains, persistent and non-persistent: Effect of temperature Authors: Ľubomír Valík, Jana Minarovičová, Eva Kaclíková, Adriana Véghová, Tomáš Kuchta Published: 2024 ABSTRACT Better growth is a phenotypic trait that can contribute to persistence of Listeria monocytogenes in food processing environments. To test the hypothesis objectively, persistent and non-persistent strains were selected and grown in different media to gain reliable quantitative growth characteristics. In this study, the effect of temperature in the range from 6 °C to 43 °C on the planktonic growth of genotypically and phenotypically different strains LM9611-19 (LM-P, persistent) and LM120/5 (LM-S, sporadic - potentially non-persistent) in Tryptone Soy Broth (TSB) and in semi-synthetic cheese medium (SCM) was investigated. Two steps of growth modelling were applied to primary growth data and growth parameters using Baranyi and cardinal temperature models (CM), respectively. No statistically significant differences were found between the growth rates of the strains within the temperature range of 6 °C–37 °C in both media. However, the average growth rates were significantly higher (p < 0.05) for LM-P than for LM-S at 40 °C and 43 °C in both media. Regardless of whether calculated on μmax or λ basis, in TSB or SCM, Tmin for LM-P strain ranged from −1.2 to 0.7 °C with an average of 0.0 ± 0.9 °C (mean ± SD). Other averages of cardinal values were in TSB (aw = 0.995; pH 7) Topt = 37.8 ± 2.0 °C, Tmax = 43.6 ± 0.5 °C and μopt = 1.27 ± 0.2 h−1. In SCM (aw = 0.970, pH 7), the averages of Topt, Tmax, and μopt were 38.0 ± 1.2 °C, 45.2 ± 2.9 °C and 0.92 ± 0.04 h−1, respectively. Generally, the parameters of the CM model for the growth rate of sporadic strain in cheese medium were lower than for the persistent strain. This includes also μopt, which reflects lower experimental growth rates in the range from Topt to Tmax. However, based on the results found in the suboptimal temperature range, it seems that the growth rate did not play an important role in the persistency characteristics. It should be noted that the study was accompanied not only by low errors in model parameters but also by acceptable external validation indices for μmax values in SCM. To consider the ComBase Predictor data (n = 8), the bias factors (Bf) of 1.08 and 1.05 and accuracy factor (Af) of 1.09 and 1.07 were calculated for the LM-P strain and LM-S strain, respectively. The approach used in this study revealed different growth responses in the range of temperatures higher than Topt. It can be extended also for mild inactivation temperature range as similar differences in behaviour between persistent and non-persistent strains might also be expected. Introduction Listeria monocytogenes is an important food-borne pathogen causing life-threatening disease in immunocompromised populations [1,2]. The main vehicle of infection is ready-to-eat food [3], which is often contaminated from the processing environment [[4], [5], [6]]. Except for a decrease in the number of cases to 1887 in 2020 (the COVID-19 pandemic period), the general trend for listeriosis did not show any significant increase or decrease in the last decade. However, the mortality rate in EU increased to 18.1 % in 2022, higher than in 2021 and 2020 (13.7 % and 13.0 %, respectively) [3]. L. monocytogenes is a Gram-positive, facultatively anaerobic, non-spore-forming, motile rod bacterium (in the range of 20–25 °C). Its cardinal growth parameters referring to the main intrinsic and extrinsic food environmental factors are well known (Tmin = −1.5 °C, Topt = 37 °C, pHmin = 4.39, pHopt = 7.0, pHmax = 9.6, aw min = 0.92 μopt = 2.0 h−1) and are successfully used in growth predictions [7]. However, variability in growth should be expected and can be quantified considering experimental, biological, and strain differences [8]. Recently, the term microbial ‘persistence’ was defined as the ability of a given organism to be established in niches (or harbourage sites) within the food and feed processing environment for a long term, despite the regular application of cleaning and disinfection [9]. Long-term presence means months or years accompanied by growth or survival in specific places of the processing environment. Despite extensive research, the persistence mechanism of L. monocytogenes in the food processing environment remains a topic of interest [10,11]. The main candidate phenotypic traits responsible for persistence include increased biofilm formation [12,13], formation of mixed biofilms [14,15], higher resistance to sanitisers [16,17] and better growth in suboptimal conditions [18,19]. While the contribution of biofilm formation and resistance to sanitisers has been extensively discussed for L. monocytogenes persistence, this study aims to add information on the contribution of growth rate to this phenomenon. Primary mathematical modelling enables the analysis of the behaviour of microorganisms involving spore germination, growth, inhibition, inactivation, or survival over time. It provides kinetic models with parameters that are, however, sensitive to changes in food environmental factors [20]. The mathematical relationships between the primary model parameters and the independent variables, such as intrinsic and extrinsic factors, such as aw, pH, temperature or modified atmosphere, are usually the subject of secondary modelling. Combining primary and secondary models, even more sophisticated tertiary models or expert systems can be created to predict the fate of microorganisms in specific foods with known environmental factors. This approach may help food microbiologists and manufacturers ensure the microbiological quality and safety of food throughout the food chain. For these reasons, mathematical modelling and predictive microbiology approach are integral parts of microbiological exposure assessment and, in combination with dose-response models, are also an inevitable part of risk assessment [21,22]. The present study focused on growth quantification and modelling using classical primary and secondary modelling to characterise the growth of a persistent and a sporadic L. monocytogenes strain. The main goal was to compare their growth responses in two media, Tryptic Soy Broth (TSB) and semi-synthetic cheese medium (SCM), and to assess the variability as well as the statistical significance of potential differences in their behaviour. Materials and methods Bacterial strains L. monocytogenes LM9611-19 (LM-P) and L. monocytogenes LM120/5 (LM-S) were used as representatives of persistent and sporadic potentially non-persistent strains, respectively. LM-P was isolated from ewes' milk stored in a cooling tank at a farm that produces raw milk cheese and identified by the State Veterinary and Food Institute, Dolný Kubín, Slovakia. It was classified as molecular serogroup IIa, clonal complex 14, sequence type 14 and was considered persistent based on <10 allelic differences in cgMLST (39 isolates during 18-month period) in our previous study [23,24]. Based on testing according to Stepanović et al. [25] and Di Ciccio et al. [26], the strain was classified as a moderate biofilm producer (data not shown). Upon testing substrate utilization with API Listeria (Biomérieux, Marcy l'Étoile, France), the strain gave typical results with a numerical profile 6510. LM-S was isolated from the milking equipment filter in the same facility as LM-P and characterised in our laboratories. It was classified as molecular serogroup IIa, clonal complex 14, sequence type 91 and considered sporadic (single isolate for 18 months) in our previous study [23,24]. Based on testing according to Refs. [25,26], the LM-S strain was classified as a weak biofilm producer (data not shown). Upon substrate utilization testing with API Listeria, the strain gave atypical results of positive fermentation of D-xylose and D-tagatose, resulting in a numerical profile 6711. Preparation of the basic bacterial suspension For growth experiments, freeze-dried L. monocytogenes strains were revitalised by incubation in TSB (Merck, Darmstadt, Germany) at 37 °C for 24 h with mild shaking (1.7 Hz). The culture was streaked on TSA (Merck) and incubated at 37 °C for 48 h. Plates with cultures were stored at 5 °C with re-inoculation every 14 days for a maximum of 5 times. Cultures preserved in this way were used to prepare the basic suspension by picking a well-isolated colony into 10 ml of TSB, mixed with vortex, and incubated at 37 °C with mild shaking (1.7 Hz). After 18 h, microbial counts were determined by spreading 200 μl of decimal dilutions in peptone saline (containing 8.5 g NaCl (Mikrochem, Pezinok, Slovakia) and 1 g peptone (Lab M, Heywood, United Kingdom) per 1000 ml) on the surface of TSA plates in duplicate. Counts were calculated and expressed according to ISO 4833–2 [27]. The numbers determined for the basic suspension were used to adjust the inoculum so that, after adding 1 ml–300 ml of growth medium, the concentration at the beginning of each experiment was within the range of 102–103 CFU/ml. Semi-synthetic cheese medium The semi-synthetic cheese medium (SCM) was prepared according to Kagkli et al. [28] and Schrama et al. [29] to imitate the nutritive potential of cheese. The following components were dissolved in 275 ml deionised water: 0.06 g CaCl2·6H2O (Sigma-Aldrich, St. Louis, USA), 0.30 g yeast extract (HiMedia, Mumbai, India), 0.31 g MgSO4·6H2O (Lachema, Brno, Czech Republic), 1.8 g L-methionine (Merck, Darmstadt, Germany), 2.04 g of KH2PO4 (Centralchem, Bratislava, Slovakia), 3 g NaCl (Centralchem), 4.5 g casein (bioreagent for cell culture grade, Sigma-Aldrich) and 11.4 ml sodium DL-lactate (Sigma- Aldrich). The lactose stock solution was prepared by steam sterilisation at a temperature of up to 100 °C for 20 min. After cooling, 25 ml of the solution was added aseptically to the cheese medium base which finally contained 8.4 g of lactose (Slavus, Bratislava, Slovak Republic) in 300 ml. Using a 10 % NaOH solution (Mikrochem) or lactic acid (Lachema), the pH value of the complete medium was adjusted to pH 7.2 before sterilisation to obtain pH 7.0 ± 0.1 after sterilisation. The aw value of the medium was 0.974 ± 0.003. Quantification of L. monocytogenes and cultivation experiments A volume of 1 ml of the appropriate dilution of the basic bacterial suspension was aseptically added to a flask with 300 ml of TSB or SCM and thoroughly mixed. Inoculated growth media were incubated at temperatures of 6 °C, 10 °C, 16 °C, 22 °C, 25 °C, 30 °C, 37 °C, 40 °C and 43 °C. For quantification, 1 ml of the culture from each flask was taken at predetermined time intervals over a period until the bacterial population reached a stationary phase or for up to 14 h. Subsequently, 200 μl of certain set dilutions were plated on the TSA surface in duplicate. The agar plates were incubated at 37 °C for 24–48 h. Duplicate growth curves were constructed within each experiment in TSB at each temperature, except for 6 °C, 30 °C and 43 °C, at which two independent experiments were carried out. In SCM, two independent experiments were carried out in duplicates at each temperature. Evaluation of variability A)Inspired by Aryani et al. [8,30], who defined the experimental, biological, and strain variabilities for growth and inactivation rates of 20 strains of L. monocytogenes in detail, indices such as MSE and RMSE were estimated in this study. Indices were calculated for the growth parameters of LM-P and LM-S strains at each temperature. According to the mentioned authors, experimental variability was defined as the difference between duplicate experiments conducted in parallel at the same time on the same experimental day. Biological variability was defined as the difference between independently reproduced experiments of the same strain performed on different experimental days from new pre-cultures and newly prepared media, and strain variability means the difference between strains of the same species.B)Alternatively, experimental RMSE values were also calculated from primary growth curve data (experiments with two replicates). Then, biological RMSE resulted from the combined data of 2 experiments, each with 2 replicates (4 replicates altogether), and analogously, the strain RMSE was calculated from all combined growth data for both strains in each medium used. These data sets involved data from 4 experiments (8 replicates; Table 1 and Table 2 for 6 °C and 30 °C, respectively). The mentioned temperatures were chosen because they generally represent low and optimal growth temperature ranges, and also because the highest number of replicates was available at these temperatures. This approach can provide an additional perspective to the evaluation of experimental, biological, and strain variabilities.Table 1Growth data of L. monocytogenes strains in TSB and SCM at 6 °C.Table 1Strainnrep/nexpMediumModel parametersStatistical indicesnλ (h)ratelog N0log NendRMSERMSEaveVariabilityR2(log CFU/h)(log CFU/mL)LM-P2/1TSB45.00.0223.589.240.0630.072experimental0.9991542.60.0223.559.430.0840.9981570.00.0223.228.970.0720.9981267.00.0223.199.000.0690.99912LM-S78.70.0203.489.710.1090.0800.9971259.40.0223.319.490.0950.9981387.70.0223.148.980.0420.9991289.60.0213.059.150.0750.99913LM-P2/1SCM98.90.0203.187.650.1100.121experimental0.99611107.90.0213.237.520.211a0.98711103.90.0213.367.630.0720.99813108.80.0233.277.580.0910.99713LM-S116.30.0252.537.490.361a0.2640.97011116.00.0212.527.630.333a0.97111127.50.0242.947.300.264a0.9791695.90.0202.997.550.1000.99717LM-P4/2TSB42.30.0213.339.210.3880.289biological0.96454LM-S78.90.0213.289.450.3710.96850LM-PSCM104.00.0213.287.600.1330.99448LM-S105.40.0202.897.550.2660.97855LM-P + LM-S8/4TSB51.90.0203.259.350.4870.394strain0.945104LM-P + LM-S8/4SCM105.00.0203.107.560.3010.971103aDeclination of the numbers observed during the lag phase. nrep - number of replicates within the experiment, nexp - number of experiments.Table 2Growth data of L. monocytogenes strains in TSB and SCM at 30 °C.Table 2Strainnrep/nexpMediumModel parametersStatistical indicesnλ (h)ratelog N0log NendRMSERMSEaveVariabilityR2(log CFU/mL/h)(log CFU/mL)LM-P2/1TSB2.60.4693.589.450.1060.069experimental0.998162.50.4643.559.430.0750.999162.00.4523.639.340.0410.999151.70.4463.579.380.0540.99916LM-S3.30.4923.079.560.1130.0840.998153.20.4933.069.520.1070.998152.50.4643.169.380.0480.999162.50.4753.179.450.0680.99916LM-P2/1SCM3.60.3833.807.930.1070.085experimental0.996144.10.4393.777.820.1030.997124.00.3633.758.030.0810.998183.40.3443.737.970.0490.99918LM-S4.40.3153.278.010.1300.0850.995183.50.3083.218.000.0890.998183.20.3512.687.940.0570.999212.80.3312.688.100.0660.99921LM-P4/2TSB2.30.4753.499.250.1810.157biological0.99463LM-S2.80.4763.109.480.1500.99662LM-PSCM3.70.3723.767.930.1310.99462LM-S3.50.3262.968.000.1670.99278LM-P + LM-S8/4TSB2.30.4753.499.250.1810.312strain0.985125LM-P + LM-S8/4SCM3.50.3413.367.790.4440.937140nrep - number of replicates within the experiment, nexp = number of experiments. A)Inspired by Aryani et al. [8,30], who defined the experimental, biological, and strain variabilities for growth and inactivation rates of 20 strains of L. monocytogenes in detail, indices such as MSE and RMSE were estimated in this study. Indices were calculated for the growth parameters of LM-P and LM-S strains at each temperature. According to the mentioned authors, experimental variability was defined as the difference between duplicate experiments conducted in parallel at the same time on the same experimental day. Biological variability was defined as the difference between independently reproduced experiments of the same strain performed on different experimental days from new pre-cultures and newly prepared media, and strain variability means the difference between strains of the same species. Inspired by Aryani et al. [8,30], who defined the experimental, biological, and strain variabilities for growth and inactivation rates of 20 strains of L. monocytogenes in detail, indices such as MSE and RMSE were estimated in this study. Indices were calculated for the growth parameters of LM-P and LM-S strains at each temperature. According to the mentioned authors, experimental variability was defined as the difference between duplicate experiments conducted in parallel at the same time on the same experimental day. Biological variability was defined as the difference between independently reproduced experiments of the same strain performed on different experimental days from new pre-cultures and newly prepared media, and strain variability means the difference between strains of the same species. B)Alternatively, experimental RMSE values were also calculated from primary growth curve data (experiments with two replicates). Then, biological RMSE resulted from the combined data of 2 experiments, each with 2 replicates (4 replicates altogether), and analogously, the strain RMSE was calculated from all combined growth data for both strains in each medium used. These data sets involved data from 4 experiments (8 replicates; Table 1 and Table 2 for 6 °C and 30 °C, respectively). The mentioned temperatures were chosen because they generally represent low and optimal growth temperature ranges, and also because the highest number of replicates was available at these temperatures. This approach can provide an additional perspective to the evaluation of experimental, biological, and strain variabilities.Table 1Growth data of L. monocytogenes strains in TSB and SCM at 6 °C.Table 1Strainnrep/nexpMediumModel parametersStatistical indicesnλ (h)ratelog N0log NendRMSERMSEaveVariabilityR2(log CFU/h)(log CFU/mL)LM-P2/1TSB45.00.0223.589.240.0630.072experimental0.9991542.60.0223.559.430.0840.9981570.00.0223.228.970.0720.9981267.00.0223.199.000.0690.99912LM-S78.70.0203.489.710.1090.0800.9971259.40.0223.319.490.0950.9981387.70.0223.148.980.0420.9991289.60.0213.059.150.0750.99913LM-P2/1SCM98.90.0203.187.650.1100.121experimental0.99611107.90.0213.237.520.211a0.98711103.90.0213.367.630.0720.99813108.80.0233.277.580.0910.99713LM-S116.30.0252.537.490.361a0.2640.97011116.00.0212.527.630.333a0.97111127.50.0242.947.300.264a0.9791695.90.0202.997.550.1000.99717LM-P4/2TSB42.30.0213.339.210.3880.289biological0.96454LM-S78.90.0213.289.450.3710.96850LM-PSCM104.00.0213.287.600.1330.99448LM-S105.40.0202.897.550.2660.97855LM-P + LM-S8/4TSB51.90.0203.259.350.4870.394strain0.945104LM-P + LM-S8/4SCM105.00.0203.107.560.3010.971103aDeclination of the numbers observed during the lag phase. nrep - number of replicates within the experiment, nexp - number of experiments.Table 2Growth data of L. monocytogenes strains in TSB and SCM at 30 °C.Table 2Strainnrep/nexpMediumModel parametersStatistical indicesnλ (h)ratelog N0log NendRMSERMSEaveVariabilityR2(log CFU/mL/h)(log CFU/mL)LM-P2/1TSB2.60.4693.589.450.1060.069experimental0.998162.50.4643.559.430.0750.999162.00.4523.639.340.0410.999151.70.4463.579.380.0540.99916LM-S3.30.4923.079.560.1130.0840.998153.20.4933.069.520.1070.998152.50.4643.169.380.0480.999162.50.4753.179.450.0680.99916LM-P2/1SCM3.60.3833.807.930.1070.085experimental0.996144.10.4393.777.820.1030.997124.00.3633.758.030.0810.998183.40.3443.737.970.0490.99918LM-S4.40.3153.278.010.1300.0850.995183.50.3083.218.000.0890.998183.20.3512.687.940.0570.999212.80.3312.688.100.0660.99921LM-P4/2TSB2.30.4753.499.250.1810.157biological0.99463LM-S2.80.4763.109.480.1500.99662LM-PSCM3.70.3723.767.930.1310.99462LM-S3.50.3262.968.000.1670.99278LM-P + LM-S8/4TSB2.30.4753.499.250.1810.312strain0.985125LM-P + LM-S8/4SCM3.50.3413.367.790.4440.937140nrep - number of replicates within the experiment, nexp = number of experiments. Alternatively, experimental RMSE values were also calculated from primary growth curve data (experiments with two replicates). Then, biological RMSE resulted from the combined data of 2 experiments, each with 2 replicates (4 replicates altogether), and analogously, the strain RMSE was calculated from all combined growth data for both strains in each medium used. These data sets involved data from 4 experiments (8 replicates; Table 1 and Table 2 for 6 °C and 30 °C, respectively). The mentioned temperatures were chosen because they generally represent low and optimal growth temperature ranges, and also because the highest number of replicates was available at these temperatures. This approach can provide an additional perspective to the evaluation of experimental, biological, and strain variabilities.Table 1Growth data of L. monocytogenes strains in TSB and SCM at 6 °C.Table 1Strainnrep/nexpMediumModel parametersStatistical indicesnλ (h)ratelog N0log NendRMSERMSEaveVariabilityR2(log CFU/h)(log CFU/mL)LM-P2/1TSB45.00.0223.589.240.0630.072experimental0.9991542.60.0223.559.430.0840.9981570.00.0223.228.970.0720.9981267.00.0223.199.000.0690.99912LM-S78.70.0203.489.710.1090.0800.9971259.40.0223.319.490.0950.9981387.70.0223.148.980.0420.9991289.60.0213.059.150.0750.99913LM-P2/1SCM98.90.0203.187.650.1100.121experimental0.99611107.90.0213.237.520.211a0.98711103.90.0213.367.630.0720.99813108.80.0233.277.580.0910.99713LM-S116.30.0252.537.490.361a0.2640.97011116.00.0212.527.630.333a0.97111127.50.0242.947.300.264a0.9791695.90.0202.997.550.1000.99717LM-P4/2TSB42.30.0213.339.210.3880.289biological0.96454LM-S78.90.0213.289.450.3710.96850LM-PSCM104.00.0213.287.600.1330.99448LM-S105.40.0202.897.550.2660.97855LM-P + LM-S8/4TSB51.90.0203.259.350.4870.394strain0.945104LM-P + LM-S8/4SCM105.00.0203.107.560.3010.971103aDeclination of the numbers observed during the lag phase. nrep - number of replicates within the experiment, nexp - number of experiments.Table 2Growth data of L. monocytogenes strains in TSB and SCM at 30 °C.Table 2Strainnrep/nexpMediumModel parametersStatistical indicesnλ (h)ratelog N0log NendRMSERMSEaveVariabilityR2(log CFU/mL/h)(log CFU/mL)LM-P2/1TSB2.60.4693.589.450.1060.069experimental0.998162.50.4643.559.430.0750.999162.00.4523.639.340.0410.999151.70.4463.579.380.0540.99916LM-S3.30.4923.079.560.1130.0840.998153.20.4933.069.520.1070.998152.50.4643.169.380.0480.999162.50.4753.179.450.0680.99916LM-P2/1SCM3.60.3833.807.930.1070.085experimental0.996144.10.4393.777.820.1030.997124.00.3633.758.030.0810.998183.40.3443.737.970.0490.99918LM-S4.40.3153.278.010.1300.0850.995183.50.3083.218.000.0890.998183.20.3512.687.940.0570.999212.80.3312.688.100.0660.99921LM-P4/2TSB2.30.4753.499.250.1810.157biological0.99463LM-S2.80.4763.109.480.1500.99662LM-PSCM3.70.3723.767.930.1310.99462LM-S3.50.3262.968.000.1670.99278LM-P + LM-S8/4TSB2.30.4753.499.250.1810.312strain0.985125LM-P + LM-S8/4SCM3.50.3413.367.790.4440.937140nrep - number of replicates within the experiment, nexp = number of experiments. Table 1Growth data of L. monocytogenes strains in TSB and SCM at 6 °C.Table 1Strainnrep/nexpMediumModel parametersStatistical indicesnλ (h)ratelog N0log NendRMSERMSEaveVariabilityR2(log CFU/h)(log CFU/mL)LM-P2/1TSB45.00.0223.589.240.0630.072experimental0.9991542.60.0223.559.430.0840.9981570.00.0223.228.970.0720.9981267.00.0223.199.000.0690.99912LM-S78.70.0203.489.710.1090.0800.9971259.40.0223.319.490.0950.9981387.70.0223.148.980.0420.9991289.60.0213.059.150.0750.99913LM-P2/1SCM98.90.0203.187.650.1100.121experimental0.99611107.90.0213.237.520.211a0.98711103.90.0213.367.630.0720.99813108.80.0233.277.580.0910.99713LM-S116.30.0252.537.490.361a0.2640.97011116.00.0212.527.630.333a0.97111127.50.0242.947.300.264a0.9791695.90.0202.997.550.1000.99717LM-P4/2TSB42.30.0213.339.210.3880.289biological0.96454LM-S78.90.0213.289.450.3710.96850LM-PSCM104.00.0213.287.600.1330.99448LM-S105.40.0202.897.550.2660.97855LM-P + LM-S8/4TSB51.90.0203.259.350.4870.394strain0.945104LM-P + LM-S8/4SCM105.00.0203.107.560.3010.971103aDeclination of the numbers observed during the lag phase. nrep - number of replicates within the experiment, nexp - number of experiments. Growth data of L. monocytogenes strains in TSB and SCM at 6 °C. Declination of the numbers observed during the lag phase. nrep - number of replicates within the experiment, nexp - number of experiments. Table 2Growth data of L. monocytogenes strains in TSB and SCM at 30 °C.Table 2Strainnrep/nexpMediumModel parametersStatistical indicesnλ (h)ratelog N0log NendRMSERMSEaveVariabilityR2(log CFU/mL/h)(log CFU/mL)LM-P2/1TSB2.60.4693.589.450.1060.069experimental0.998162.50.4643.559.430.0750.999162.00.4523.639.340.0410.999151.70.4463.579.380.0540.99916LM-S3.30.4923.079.560.1130.0840.998153.20.4933.069.520.1070.998152.50.4643.169.380.0480.999162.50.4753.179.450.0680.99916LM-P2/1SCM3.60.3833.807.930.1070.085experimental0.996144.10.4393.777.820.1030.997124.00.3633.758.030.0810.998183.40.3443.737.970.0490.99918LM-S4.40.3153.278.010.1300.0850.995183.50.3083.218.000.0890.998183.20.3512.687.940.0570.999212.80.3312.688.100.0660.99921LM-P4/2TSB2.30.4753.499.250.1810.157biological0.99463LM-S2.80.4763.109.480.1500.99662LM-PSCM3.70.3723.767.930.1310.99462LM-S3.50.3262.968.000.1670.99278LM-P + LM-S8/4TSB2.30.4753.499.250.1810.312strain0.985125LM-P + LM-S8/4SCM3.50.3413.367.790.4440.937140nrep - number of replicates within the experiment, nexp = number of experiments. Growth data of L. monocytogenes strains in TSB and SCM at 30 °C. nrep - number of replicates within the experiment, nexp = number of experiments. Primary and secondary modelling The density of the cultures of both L. monocytogenes strains was evaluated as a function of time in the two growth media according to Baranyi's D-model [31] using the Excel tool DMFit v. 3.5 (Institute of Food Research, Norwich, United Kingdom). To evaluate the effect of temperature on specific growth rate and lag phase (secondary modelling), non-linear regression using Excel Solver (Microsoft, Redmont, USA) was applied. Baranyi's model is described by the following equations:(1)y(t)=y0+μmax⋅A(t)−1m⋅ln(1+em⋅μmax⋅A(t)−1em⋅(ymax−y0))(2)A(t)=t+ln(e−μmax⋅t+e−h0−e−μmax⋅t−h0)μmaxwhere y(t) = ln N(t), y0 = ln N0, ymax = ln Nmax, m is the parameter of the curvature of the growth line when the growth of microorganisms slows down from the exponential phase to the stationary phase. A(t) is an integral form of the adjustment function α(t) whose value increases with increasing time, t is time and h0 is a dimensionless factor that quantifies the initial physiological state of cells and is related to lag phase (λ) by the equation h0 = μ·λ [31]. Using log10 N in the DMFit tool, the main outputs of primary modelling were the growth rate kmax (log CFU/mL/h) and the duration of lag phase λ (h). Before secondary modelling, kmax was converted to μmax (h−1) according to the equation μmax=kmax⋅ln(10). To describe the effects of temperature on the growth of L. monocytogenes strains in TSB and SCM at aw values of 0.995 and 0.974, respectively, the square root Cardinal model (sqrtCM) of Rosso et al. [32] was used in the following form:(3)μmax=μopt⋅CM(T)(4)1λ=1λopt⋅CM(T)where(5)CM(T)=(T−Tmax)⋅(T−Tmin)2(Topt−Tmin)⋅[(Topt−Tmin)⋅(T−Topt)−(Topt−Tmax)⋅(Topt+Tmin−2T)] In this work, a square root transformation of the whole cardinal model was used for both growth parameters μmax and reciprocal λ. The secondary model includes five parameters with biological meaning, namely, μopt, λ, Tmin (theoretical minimum temperature), Topt (optimum temperature) and Tmax (maximum temperature above which growth is not likely). Except for normalising residuals, the modification allowed the estimation of parameters μopt and 1/λopt in the original non-squared form, which is more practical particularly for μopt (Eqs. (3), (4))). Statistical analysis and model validation Specific growth rates of the two strains at 6–37 °C were used for internal cross-validation, in which a set of μmax calc values for LM-P was validated with LM-S μmax exp values and vice versa. Due to statistically significant differences in μmax values, the data at 40 °C and 43 °C were excluded from this validation. External validation of secondary rate models for both strains in cheese medium was performed using growth rates acquired from the ComBase Predictor (CBP) at aw 0.974, pH 7. For CMs in TSB, the experimental μmax values estimated by Bajard et al. [33] in Mueller-Hinton broth (MHB) were used for validation. Data are collected in Suppl. Table S1. To evaluate the accuracy of fitting, the coefficient of determination (R2) and the root mean square error (RMSE) were calculated:(6)R2=∑i=1n(yiobs−yical)2(yiobs−y‾ical)2(7)RMSE=(yiobs−yical)2n−pwhere yiobs, yical and y‾ical are the observed (experimental), calculated (predicted) and averaged μmax or λ data, respectively; n is the number of observations; and p is the number of model parameters [34]. The CMs for growth rate were externally validated using independent data excerpted from the Combase database at aw of 0.974 and pH 7 as well as the data from Ref. [33]. The following equations were used to calculate validation indices bias and accuracy factors [7], [32]:(8)Bf=10∑i=1nlogμmaxiobs-logμmaxipredn(9)Af=10∑i=1n|logμmaxipred−logμmaxiobs|/n According to te Giffel and Zwietering [7], the bias factor (Bf) provides information on the distribution of the observed values towards the identity line between the observed (experimental) and predicted (calculated) μmax values. In this case, a value of Bf lower than 1 indicates a ‘fail-safe’ prediction and a value of Bf greater than 1 means a ‘fail-dangerous’ model. The accuracy factor (Af) describes how close are the predicted μmax to μobs values. The statistical differences between μmax values of the two L. monocytogenes strains at each temperature were statistically evaluated using the Analyse-it Method Validation package ver. 6.15 (Analyse-it Software, Leeds, United Kingdom). Data were treated using Student's t-test followed by Tukey's test with the least significant difference of 95 %. Results and discussion Growth of L. monocytogenes strains In this work, 63 independent cultivation experiments were carried out with duplicate trials for the two strains, persistent LM-P and sporadic – potentially non-persistent LM-S, in TSB and SCM. TSB represented a theoretically optimal culture medium (aw of 0.995, pH 7.0), while the semi-synthetic cheese medium simulated conditions that more closely resembled the internal environment of cheese (aw of 0.947 ± 0.003, pH 7.0). The nine cultivation temperatures used (6–43 °C) covered almost the entire range of L. monocytogenes growth. In total, more than 120 growth curves were provided with growth parameters and statistical indices. Experimental, biological and strain variability was evaluated according to Aryani et al. [8] using specific growth rates and were also based on the RMSE values of the growth curves from individual experiments or combined data from 2 to 4 experiments (4–8 replicates) performed at different times. Evaluation of variability To quantify the internal experimental, biological and strain variability of μmax according to Aryani et al. (2015a), we used μmax data recalculated from growth rates of all individual curves carried out in SCM at each temperature. In this way, we could evaluate the MSE values and compare RMSE values for our persistent and sporadic strains with those of 20 strains by Aryani et al. [8,30]. The variances of μmax for our strains were independent of temperature. The following RMSEexp, RMSEbiol, and RMSEstr of 0.016 h−1, 0.022 h−1, and 0.017 h−1 were calculated, respectively, for both strains LM-P and LM-S in this study. Aryani et al. [8] observed almost identical variability and the same trend between the values as we did in the current study, as RMSEbiol was the highest, followed by RMSEstr and RMSEexp. However, in the second case the RMSE values of growth curves, which are log10N-based and express the goodness of fit for the primary growth model, can also provide an opportunity to evaluate experimental, biological and strain variability. The RMSE average values (RMSEave) may come not only from individual trial duplicates conducted at the same time (experimental variability) or from experiments with duplicates performed at different times (biological variability) but also, in our case, from data combined from experiments with LM-P and LM-S (strain variability). Together with other growth parameters, the RMSE values are presented in Table 1, Table 2 for temperatures of 6 °C and 30 °C, respectively. The experimental RMSE values fluctuated between 0.063 and 0.130 log CFU/mL for both strains, temperatures, and media with almost the same mean RMSEexp of 0.082 and 0.081 for 6 °C and 30 °C, respectively. Similarly, the RMSEave values representing biological variability (2 experiments performed at different times containing data from 4 individual replicates) were 0.289 and 0.157 log CFU/mL for 6 °C and 30 °C, respectively. Finally, the highest average RMSE values of 0.394 and 0.312 log CFU/mL described the strain variability in both media (TSB and SCM) at 6 °C and 30 °C, respectively. In addition, the growth curves fitted with the Baranyi D-model using experimental data for both strains (LM-Pexp + LM-Sexp; 8 replicates) and representing strain variability in each medium are shown in Fig. 1 (TSB in blue; SCM in yellow). We suppose that similar values could be found in the suboptimal temperature range. However, it could not be properly demonstrated as only one experiment with two replicates was performed in TSB at other temperatures within the study.Fig. 1Growth curves of L. monocytogenes in TSB (blue) and SCM (yellow) at 6 °C and 30 °C. Merged data for LM-P and LM-S strains (LM-Pexp + Sexp) come from 4 experiments and 8 replicates in total. Dashed lines represent fitted values ± standard error (SE) of the fit and the solid red line in the right button figure represents the recalculated growth curve using equalized N0 of 2.96 log CFU/mL.Fig. 1 Growth curves of L. monocytogenes in TSB (blue) and SCM (yellow) at 6 °C and 30 °C. Merged data for LM-P and LM-S strains (LM-Pexp + Sexp) come from 4 experiments and 8 replicates in total. Dashed lines represent fitted values ± standard error (SE) of the fit and the solid red line in the right button figure represents the recalculated growth curve using equalized N0 of 2.96 log CFU/mL. The comparison of variability based either on specific growth rates or fitting errors of growth curves needs to use relative indices. To do so, the ratios between the average RMSE values and μ-values were calculated for replicates, experiments, and combined strain data. Similarly, for relative indices based on growth curves, the ratio of RMSE and growth increases (logNend - logN0) averages was used. The experimental, biological and strain variabilities based on μ data were represented with relative errors of 5.2 %, 4.0 % and 8.3 %, respectively. On the other hand, the relative indices based on the logN data showed slightly lower values of 1.6 %, 2.9 % and 6.1 %, respectively. It seems that both methods of evaluation provided acceptable outputs and, in this case, also certain relations between them could be defined. Except for one case, the relative indices doubled from experimental to biological and from biological to strain variability. Effect of temperature on growth parameters of L. monocytogenes strains in TSB and SCM The cardinal temperature model with inflection (CM) is one of the most widely used secondary models that reliably describes the effect of temperature in the full range of growth on microbial growth parameters. The model is often preferred as it provides four parameters that have biological interpretation. In contrast, the extended square root model contains two regression coefficients and parameters Tmin and Tmax, which are only theoretical minimum and maximum temperatures for growth and may differ significantly from the actual cardinal temperatures [35]. Graphic presentations of the dependences of sqrt μmax and sqrt 1/λ on the temperature for LM-P and LM-S in TSB (aw 0.995) and SCM (aw 0.974) are shown in Fig. 2, Fig. 3, respectively. The dashed lines represent the calculated data using standard errors of model parameters, which are summarised in Table 3.Fig. 2Modelling the effect of temperature on sqrt μmax of LM-P and LM-S strains in Tryptone Soy Broth (TSB) and semi-synthetic cheese medium (SCM).Fig. 2Fig. 3Modelling the effect of temperature on sqrt 1/λ of LM-P and LM-S strains in Tryptone Soy Broth (TSB) and semi-synthetic cheese medium (SCM).Fig. 3Table 3Parameters of the square root CM model and statistical indices of fit for LM-P and LM-S strains in TSB and SCM.Table 3ParametersCM sqrt 1/λ LM (TSB)CM sqrt μmaxLM (TSB)CM sqrt 1/λ LM (SCM)CM sqrt μmaxLM (SCM)LM-PSELM-SSELM-PSELM-SSELM-PSELM-SSELM-PSELM-SSETmin (°C)−1.1510.0121.0760.0080.1970.0010.8210.0030.6750.0031.2520.0070.0860.001−1.1150.005Topt (°C)40.8200.13436.8950.13337.2030.03036.3760.06638.9230.10538.8110.10338.0260.09136.3670.063Tmax (°C)43.2230.04543.1800.01544.3030.02143.5800.01843.3560.01744.2330.09749.5180.46643.8900.0291/λopt (h−1)0.7040.0240.4950.016–-0.3920.0090.3870.010––μopt (h−1)––1.2820.0091.2590.020––0.9440.0180.8940.013RMSE0.0400.0380.0110.0300.0180.0210.0260.022R20.9710.9700.9990.9910.9890.9870.9910.992 Modelling the effect of temperature on sqrt μmax of LM-P and LM-S strains in Tryptone Soy Broth (TSB) and semi-synthetic cheese medium (SCM). Modelling the effect of temperature on sqrt 1/λ of LM-P and LM-S strains in Tryptone Soy Broth (TSB) and semi-synthetic cheese medium (SCM). Table 3Parameters of the square root CM model and statistical indices of fit for LM-P and LM-S strains in TSB and SCM.Table 3ParametersCM sqrt 1/λ LM (TSB)CM sqrt μmaxLM (TSB)CM sqrt 1/λ LM (SCM)CM sqrt μmaxLM (SCM)LM-PSELM-SSELM-PSELM-SSELM-PSELM-SSELM-PSELM-SSETmin (°C)−1.1510.0121.0760.0080.1970.0010.8210.0030.6750.0031.2520.0070.0860.001−1.1150.005Topt (°C)40.8200.13436.8950.13337.2030.03036.3760.06638.9230.10538.8110.10338.0260.09136.3670.063Tmax (°C)43.2230.04543.1800.01544.3030.02143.5800.01843.3560.01744.2330.09749.5180.46643.8900.0291/λopt (h−1)0.7040.0240.4950.016–-0.3920.0090.3870.010––μopt (h−1)––1.2820.0091.2590.020––0.9440.0180.8940.013RMSE0.0400.0380.0110.0300.0180.0210.0260.022R20.9710.9700.9990.9910.9890.9870.9910.992 Parameters of the square root CM model and statistical indices of fit for LM-P and LM-S strains in TSB and SCM. The solid lines representing the sqrtCM model with sqrt μmax for both strains in both media agreed with the observed μmax in the suboptimal temperature range. Differences were found in the region beyond Topt (which is approximately 37 °C; Fig. 2). The μmax exp values for LM-P and LM-S strains in TSB and SCM did not show significant differences in the suboptimal temperature range (6 °C–30 °C) and at 37 °C in both media. At the 5 % significance level, this was confirmed by ANOVA, Student's t-test, and Tukey's test at each temperature. However, in the region between optimal and maximum temperature, which contained two experimental temperatures (40 °C and 43 °C), the differences between μmax exp values from repeated trials were statistically confirmed for both strains and both media (p = 0.0229 and 0.0145 at 40 °C and p < 0.0001 and 0.0053 at 43 °C in TSB and SCM, respectively). Interestingly, the values of sqrt μmax were higher for LM-P than for LM-S at the three highest incubation temperatures, but they were statistically significant only at the last two, 40 °C and 43 °C. Furthermore, both cardinal parameters Topt and Tmax determined by the sqrtCM model for sqrt μmax were also higher for LM-P than for LM-S in both media (Table 3). In terms of growth rate, Magalhães et al. [19] found that persistent and non-persistent strains showed different responses to NaCl and acidic conditions (pH 5). Average rates were significantly higher (p < 0.05) for persistent than for non-persistent isolates when grown at 22 °C, 2.5 %, 4 %, 8 % NaCl, and at pH 5. In our case, the situation was different as significantly higher growth rates were observed for the persistent strain in both media only at temperatures higher than 37 °C (at 40 °C and 43 °C). A similar trend could be observed in Fig. 3, which is the output of the sqrtCM model application to the lag phase. In this case, higher values of sqrt 1/λ, which means a shorter lag phase, could be seen more clearly in the region of temperatures beyond Topt. Likewise, the Topt values resulting from secondary modelling of sqrt 1/λ against temperature were higher in the case of LM-P in both media (Table 3). Magalhães et al. [19] estimated similar lag times for persistent and non-persistent isolates when grown at 37 °C, 22 °C and 4 °C but significantly shorter (p < 0.05) for persistent strains grown at 2.5 %, 4 % and 8 % NaCl, and at pH 5. In our case, significantly shorter (p < 0.05) lag times were observed for persistent strain (in both media) only at temperatures 40 °C. and 43 °C. However, the estimated Tmax values from the lag phase modelling were within the range of 43.2–44.2 °C for both strains and the media. The sqrtCM model fitted well all the data obtained from our experiments, as documented not only by the standard error of the parameters but also by the values of other statistical parameters such as RMSE and coefficient of determination R2 (Table 3). Associated with the minimal errors of estimation, all cardinal parameters for L. monocytogenes strains in this study were consistent with those mentioned by Bajard et al. [33], te Giffel and Zwietering [7] and Jia et al. [36]. Regardless of whether calculated on μmax or λ basis, in TSB or SCM, for a persistent or non-persistent strain, the most discussed cardinal parameter Tmin, here with estimation errors from 0.0003 °C to 0.012 °C, ranged from −1.2 °C to 1.3 °C. For LM-P strain, the range of Tmin values determined with the CM model in the two media was −1.2 to 0.7 °C with an average of 0.0 ± 0.8 °C (mean ± SD; n = 4). As in Table 3, strain LM-S showed a slightly higher average Tmin of 0.5 ± 1.1 °C. A similar estimated minimum growth temperature of 0.6 ± 0.2 °C was reported for L. monocytogenes in pasteurised milk by Jia et al. [36]. Model validation Even with sound statistical indices of the sqrtCM and model parameters shown in Table 3, the results should be validated using independent data. In this work, growth data for two strains with different genetic and physiological characteristics offered an option of internal cross-validation, within which the sqrtCM model for LM-P could be validated with μmax exp values of LM-S and vice versa. As statistically significant differences were found between the μmax model and experimental data at 40 °C and 43 °C for the two strains, these data were excluded from internal cross-validation. Fig. 4 shows the comparison of two sets of predicted and observed μmax values (recalculated from sqrtCM) in SCM at aw of 0.974 for the suboptimal temperature range (6 °C–37 °C).Fig. 4Comparison of the observed and predicted values at internal cross-validation with LM-P and LM-S in SCM.Fig. 4 Comparison of the observed and predicted values at internal cross-validation with LM-P and LM-S in SCM. Table 4 provides accompanying linear regression data for the predicted and observed μmax values with low parameter errors and acceptable RMSE 0.027 and 0.029, respectively.Table 4Comparison of predicted and observed μmax values as a part of internal cross-validation.Table 4Comparison betweenLinear regression and statistical parametersInterceptSESlopeSERMSER2predicted LM-Pand observed LM-S μmax values−0.00470.00931.05440.01810.02700.9927predicted LM-Sand observed LM-P μmax values0.01210.00970.93220.01760.02870.9912 Table 4Comparison of predicted and observed μmax values as a part of internal cross-validation.Table 4Comparison betweenLinear regression and statistical parametersInterceptSESlopeSERMSER2predicted LM-Pand observed LM-S μmax values−0.00470.00931.05440.01810.02700.9927predicted LM-Sand observed LM-P μmax values0.01210.00970.93220.01760.02870.9912 Comparison of predicted and observed μmax values as a part of internal cross-validation. The values predicted using the CM model for the LM-P strain slightly overestimated the experimental data for the LM-S strain. On the other hand, the calculated data for LM-S underestimated the experimental data for LM-P. However, the Bf values of 1.12 and 0.98 for the LM-P and LM-S CM models, respectively, were close to 1.0. Furthermore, Fig. 4 showed that, on average, the experimental values for LM-S were higher than for LM-P in the range of 6 °C–37 °C. However, the opposite situation could be seen in the temperature range between Topt and Tmax (Fig. 4). Square root CMs for growth rate were externally validated using independent data from Combase at aw of 0.974, pH 7 and from Bajard et al. [33]. For Combase data, the bias factors for the LM-P and LM-S models in SCM were 0.93 and 0.96, respectively, while the accuracy factors were 1.09 and 1.07, respectively. This means that the sqrtCM model for both L. monocytogenes strains provided a ‘fail-safe’ prediction to Combase data with 9 % and 7 % accuracy. On the other hand, the validation of LM-P and LM-S models based on data from this study in TSB with experimental growth rates by Bajard et al. [33] performed in Mueller-Hinton broth revealed ‘fail-dangerous’ predictions with Bf 1.06 and 1.07, respectively. Both accuracy factors were equal to 1.15. Conclusions In this work, the question of whether persistent or sporadically L. monocytogenes strains can grow better was dealt with by a quantitative study using two strains in two media in response to various temperatures. In the suboptimal temperature range, the growth parameters of persistent and sporadic, potentially non-persistent strains were almost indistinguishable, and it seems that growth predictions of persistent strains may be performed using the data of non-persistent strains (and vice versa), e.g., in exposure assessment. However, the models revealed differences in the behaviour of the strains at 40 °C and 43 °C. The results suggest that secondary modelling is suitable for testing a panel of L. monocytogenes strains regarding the possible correlation of their growth characteristics with persistence in food processing environments. The approach used in this study proved to be effective in revealing different growth responses in the range of higher temperatures than Topt. It can be extended also for mild inactivation temperature range as some similar differences in behaviour between persistent and non-persistent strains might also be expected. CRediT authorship contribution statement Ľubomír Valík: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Methodology, Funding acquisition, Conceptualization. Jana Minarovičová: Supervision, Methodology, Investigation. Eva Kaclíková: Writing – review & editing, Writing – original draft, Supervision, Methodology. Adriana Véghová: Methodology, Investigation. Tomáš Kuchta: Writing – review & editing, Writing – original draft, Resources, Conceptualization. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.