Modelling and predicting the simultaneous growth of Listeria monocytogenes and psychrotolerant lactic acid bacteria in processed seafood and mayonnaise-based seafood salads

lable at ScienceDirect

Food Microbiology 46 (2015) 1e14

Contents lists avai

Food Microbiology

journal homepage: www.elsevier .com/locate/ fm

Modelling and predicting the simultaneous growth of Listeriamonocytogenes and psychrotolerant lactic acid bacteria in processedseafood and mayonnaise-based seafood salads

Ole Mejlholm*, Paw DalgaardNational Food Institute (DTU Food), Technical University of Denmark, Kgs. Lyngby, Denmark

a r t i c l e i n f o

Article history:Received 16 December 2013Received in revised form16 June 2014Accepted 4 July 2014Available online 15 July 2014

Keywords:Predictive modelMicrobial interactionJameson effectLactobacillusRisk assessmentRisk managementOrganic acids

* Corresponding author. Division of Industrial FoInstitute (DTU Food), Technical University of DenmarkDK-2800 Kgs. Lyngby, Denmark. Tel.: þ45 45252567;

E-mail address: [email protected] (O. Mejlholm).

http://dx.doi.org/10.1016/j.fm.2014.07.0050740-0020/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

A new combined model for Listeria monocytogenes and psychrotolerant Lactobacillus spp. was con-structed and evaluated for processed seafood and mayonnaise-based seafood salads. The new model wasconstructed by combining existing cardinal parameter models for L. monocytogenes and Lactobacillus spp.using the classical Jameson effect to model microbial interaction. Maximum population density (MPD)values of L. monocytogenes were accurately predicted in processed seafood with a known initial cellconcentration of Lactobacillus spp. For these experiments, average MPD values of 4.5 and 4.3 log (cfu/g)were observed and predicted, respectively for L. monocytogenes. In seafood salads, growth ofL. monocytogenes continued at a reduced growth rate after Lactobacillus sakei had reached their MPD. Thisgrowth pattern was successfully described by an expanded version of the classical Jameson effect model,but only for products with pH of 6.0 or higher. For seafood salads with pH below 6.0 the performance ofthe new model was unacceptable, primarily due to prediction of no-growth by L. monocytogenes whengrowth was actually observed. Within its range of applicability the new model can be valuable for riskassessment and risk management of processed seafood as well as for evaluating the compliance ofproducts with the EU regulation for ready-to-eat foods.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Over the last couple of decades, numerousmathematical modelshave been developed to predict growth responses of pathogenicand spoilage microorganisms in different types of food. Thesemodels are important as they facilitate the assessment and man-agement of microbial food safety and quality. Some predictivemodels have been included in user-friendly software and arewidely used by food processors and regulatory authorities(McMeekin et al., 2006). As one example, the Seafood Spoilage andSafety Predictor (SSSP) software has been recommended by theDanish Veterinary and Food Administration (DVFA) as a means topredict growth of Listeria monocytogenes and to document thecompliance of ready-to-eat foods with the EU-regulation (EC, 2005;DVFA, 2013). SSSP includes a growth and growth boundary modelfor L. monocytogenes that takes into account the effect of 12

od Research, National Food, Søltofts Plads, Building 221,fax: þ45 45884774.

environmental parameters (Mejlholm and Dalgaard, 2009). Thismodel has been validated and is increasingly used for differenttypes of food with various preserving parameters (Mejlholm et al.,2010). However, further studies are needed to determine its rangeof applicability with respect to other food types and preservingparameters. Previously, modelling the effect of microbial interac-tion between L. monocytogenes and lactic acid bacteria (LAB) wasdetermined to be important in order to accurately predict themaximum population density (MPD) of the pathogen e.g. for cold-smoked and marinated fish products (Gim�enez and Dalgaard,2004; Mejlholm and Dalgaard, 2007a; Vermeulen et al., 2011).These predictions relied on (i) the assumption of the Jameson effecti.e. that L. monocytogenes stopped growing when LAB reached theirMPD and on (ii) accurate predictions of growth responses of bothL. monocytogenes and LAB. Existing predictive models for growth ofL. monocytogenes and psychrotolerant LAB have been successfullyvalidated for wide ranges of 12 preserving parameters andincluding products with benzoic, citric and/or sorbic acids(Mejlholm and Dalgaard, 2009, 2013). However, the ability of thesemodels to accurately predict microbial interaction in chilled prod-ucts including these organic acids remains to be documented and

Delta:1_given name

Delta:1_surname

mailto:[email protected]

http://crossmark.crossref.org/dialog/?doi=10.1016/j.fm.2014.07.005&domain=pdf

www.sciencedirect.com/science/journal/07400020

http://www.elsevier.com/locate/fm

http://dx.doi.org/10.1016/j.fm.2014.07.005



O. Mejlholm, P. Dalgaard / Food Microbiology 46 (2015) 1e142

this is important e.g. for brined shell-fish, roe-products and seafoodsalads, including mayonnaise-based products, where benzoic, citricand/or sorbic acids often are used as preserving parameters(Mejlholm et al., 2012; Skalina and Nikolajeva, 2010; Vermeulenet al., 2007b).

Several studies have shown the presence and growth potentialof L. monocytogenes in mayonnaise-based salads, with the highestprevalence of the pathogen being observed for products containingprocessed seafood (Coillie et al., 2004; Di Pinto et al., 2010; Gombaset al., 2003; Hartemink and Georgsson, 1991; Levine et al., 2001;Uyttendaele et al., 2009). With respect to microbial spoilage ofthese products, LAB has often been identified as the dominatingmicrobiota at the time of sensory rejection, together with yeast andmoulds (Manios et al., 2009; Smittle and Flowers, 1982). Somemathematical models for growth responses of L. monocytogenes orLAB in mayonnaise-based salads and related products are available(Gysemans et al., 2007; Hwang and Tamplin, 2005; IFST, 1993;Manios et al., 2009, 2014; Vermeulen et al., 2007a,b). Thesemodels, however, do not include the effect of benzoic, citric andsorbic acids or the effect of microbial interaction betweenL. monocytogenes and LAB. In addition, several of the models aredeveloped exclusively for ambient temperature storage and furtherstudies of chilled mayonnaise-based seafood salads are needed.

The objective of the present study was to model and predict thesimultaneous growth of L. monocytogenes and psychrotolerant LABin processed and chilled seafood and mayonnaise-based seafoodsalads with different combinations of preservatives includingbenzoic, citric and/or sorbic acids. Existing growth and growthboundary models for L. monocytogenes and psychrotolerant Lacto-bacillus spp. including the individual as well as the interactive effectof 12 environmental parameters (Mejlholm and Dalgaard, 2009,2013) were combined to take into account the effect of microbialinteraction between the two types of microorganisms. The perfor-mance of the new combinedmodel was evaluated by comparison ofobserved and predicted growth of L. monocytogenes and Lactoba-cillus sakei in 30 challenge tests with brined shrimp andmayonnaise-based shrimp salad including different combinationsof organic acids as preservatives, and stored at both dynamic andconstant chill storage temperatures. Literature data (n ¼ 82) forkinetics of L. monocytogenes and LAB in different types of seafoodwas included to strengthen the validation of the new combinedmodel.

2. Material and methods

2.1. Development of the new combined model for L. monocytogenesand psychrotolerant Lactobacillus spp.

Cardinal parameter secondary models for the growth rate andthe growth boundary of L. monocytogenes and psychrotolerantLactobacillus spp. (Mejlholm and Dalgaard, 2009, 2013) werecombined by using the expanded and differential form of the lo-gistic model (Equation (1)) to model the effect of microbial in-teractions as previously described (Gim�enez and Dalgaard, 2004;Mejlholm and Dalgaard, 2007a). Both secondary models includethe effect of temperature, NaCl/water activity, pH, smoke compo-nents (phenol), CO2, nitrite, acetic acid, benzoic acid, citric acid,diacetate, lactic acid and sorbic acid as well as their interactiveeffects. The present study concerns chilled products and unlessotherwise specifically stated then predictions for LAB and Lacto-bacillus spp. refer to psychrotolerant species.

An expanded version of the differential form of the logisticmodel was used to model microbial interaction betweenL. monocytogenes and LAB (Equation (1)) (Gim�enez and Dalgaard,2004; Møller et al., 2013). This model is based on the assumption

that L. monocytogenes and LAB inhibit each other to the same extentas they inhibit their own growth. However, due to higher concen-trations and faster growth of LAB, the inhibiting effect ofL. monocytogenes on LAB was not evaluated in the present study.

8>>>>>>>>>>>>><>>>>>>>>>>>>>:

t<tlag�Lm;dLm=dtLmt

¼0

t� tlag�Lm;dLm=dtLmt

¼mmaxLm��1� Lmt

Lmmax

��1�g�LABt

LABmax

�

t<tlag�LAB;dLAB=dtLABt

¼0

t� tlag�LAB;dLAB=dtLABt

¼mmaxLAB��1� LABt

LABmax

�

(1)

where Lm and LAB, both >0 cfu/g, signify concentrations ofL. monocytogenes and LAB, respectively, and g is a competitionfactor that allows the predicted cell concentration ofL. monocytogenes to increase (g < 1) or decrease (g > 1) after the cellconcentration of LAB has reached their maximum populationdensity (MPD) value (LABmax) (Møller et al., 2013). Growth rates ofL. monocytogenes (mmax Lm) and LAB (mmax LAB) were obtained fromthe models of Mejlholm and Dalgaard (2009, 2013).

2.2. Bacterial strains and preculture conditions

Four isolates of L. sakei (LA-5, A1, MAP23-1 and F46-4) (LABmix)and four isolates of L. monocytogenes (94-203D, 95-54A, 95-442A,94-167B) (Lmmix) from processed seafood (e.g. brined and drainedMAP shrimp and seafood salad) (Gim�enez and Dalgaard, 2004;Jørgensen and Huss, 1998; Mejlholm et al., 2008; Mejlholm andDalgaard, 2007a) were used for inoculation of challenge tests.Initially, each isolate was transferred from a �80 �C culturecollection at DTU Food and grown at 25 �C for 24 h using All Pur-pose Tween (APT) broth (Difco 265510, Becton, Dickinson andCompany, Le Point de Claix, France) for L. sakei and Brain HeartInfusion (BHI) broth (Oxoid CM1135, Basingstoke, Hampshire, En-gland) for L. monocytogenes. Then isolates were precultured at 8 �Cfor 2e3 days in fresh medium of the same type with pH 6.0 and 2%NaCl to simulate the conditions of many processed seafood. Pre-cultures were harvested in late exponential growth phase, definedas a relative change in absorbance of 0.05e0.2 at 540 nm (NovaspecII, Pharmacia Biotech, Allerød, Denmark). As needed, precultureswere diluted in 0.85% NaCl prior to inoculation of the products usedin the various experiments.

2.3. Evaluation of the new combined model for L. monocytogenesand psychrotolerant Lactobacillus spp.

112 sets of environmental conditions and corresponding growthresponses of L. monocytogenes and LAB in processed seafood andmayonnaise-based seafood salads were used to evaluate the per-formance of the new combined LmeLAB model (Tables 1 and 2). Atotal of 30 data sets from brined and drained shrimp (n ¼ 11) andmayonnaise-based shrimp salad (n ¼ 19) were generated in thepresent study (Table 1). In addition, data from 82 experiments withprocessed seafood and mayonnaise-based seafood salads wereobtained from the literature (Table 2). 57 experiments were inoc-ulated with L. monocytogenes and LAB, whereas for the remaining55 experiments the products were inoculated exclusively with thepathogen (Tables 1 and 2).

For literature data, growth responses of L. monocytogenes andLAB were described by (i) the maximum population density (MPD,

O. Mejlholm, P. Dalgaard / Food Microbiology 46 (2015) 1e14 3

log cfu/g), (ii) the maximum specific growth rate (mmax, 1/days) and(iii) growth and no-growth responses. Growth rates (mmax values)were obtained (i) directly from the reported data or (ii) by fitting ofgrowth curves with the integrated and log-transformed form of thelogistic model with delay (Dalgaard, 2009; Rosso et al., 1996). Todifferentiate between growth and no-growth of L. monocytogenesthe latter was defined as an increase in cell concentration of lessthan 0.5 log (cfu/g) within the duration of the experiment (FAO/WHO, 2009).

The performance of the new combined LmeLAB model wasevaluated by comparing observed and predicted MPD values ofL. monocytogenes. Observed and predicted mmax values ofL. monocytogenes or LAB were compared for experiments at con-stant storage temperatures by calculation of bias and accuracyfactor (Bf and Af) values (Ross, 1996). The Bf value indicates a sys-tematic over- or underestimation of mmax values and the Af value is ameasure of the average difference between observed and predictedmmax values. The Bf value was calculated so that numbers higherthan 1 always indicated that predicted growth was faster thanobserved growth. As an example, a Bf value of 1.2 indicates thatpredicted mmax values on average are 20% faster than observed mmax

values. For experiments at dynamic storage temperatures, thepercentage of observations within the acceptable simulation zone(ASZ) was determined. With the ASZ defined as ±0.5 log-units fromthe simulated L. monocytogenes or LAB concentrations. To beacceptable at least 70% of the observed values should be within thiszone (Møller et al., 2013; Oscar, 2005; Velugoti et al., 2011).Observed and predicted growth and no-growth responses ofL. monocytogenes were compared by calculating the percentage ofexperiments that were correctly predicted. Incorrect predictionswere categorized as fail-dangerous (i.e. no-growth predicted whengrowth was actually observed) or fail-safe (i.e. growth predictedwhen no-growth was actually observed).

2.3.1. Brined and drained MAP shrimpA total of 11 challenge tests with brined and drained shrimp

including benzoic, citric and/or sorbic acids were carried out toevaluate the performance of the new combined LmeLAB model atdynamic and constant storage temperatures (Table 1, ExperimentsB1eB11).

Three types of brine including sodium chloride (Akzo Nobel SaltInc., Mariager, Denmark), sodium benzoate (SFK Food Inc., Viborg,Denmark), citric acid (Jungbunzlauer AG, Basel, Switzerland) and/orpotassium sorbate (SFK Food) were studied (Table 1).

Individually quick frozen shrimp (Pandalus borealis) from theNorth Atlantic Ocean were used for the experiments. Shrimp werecaught, cooked, peeled, frozen and glazed in Greenland, and sup-plied in their frozen state to DTU Food where they were kept at�20 �C until the start of the experiments. Frozen shrimp werethawed overnight at 5 �C. Thawed shrimpweremarinated in plasticcontainers for 2 days at 1e2 �C, after which the brine was drainedoff. The ratio (w/w) between shrimp and brine was 1.0:0.7.

Brined and drained shrimp (experiments B1eB11) were inocu-lated with a mixture of L. sakei (LAB mix) and L. monocytogenes (Lmmix) to initial concentrations of approx. 2e3 log (cfu/g). Followinginoculation, 50 ± 1 g of brined and drained shrimp for experimentsB1eB3 were packaged in modified atmosphere initially containing40% CO2 and 60% N2 (AGA Ltd., Copenhagen, Denmark). A MultivacA 300/16 packaging machine (Multivac Ltd., Vejle, Denmark) and apackaging film (NEN 40 HOB/LLPDE 75, Amcore Flexibles, Horsens,Denmark) with low gas permeability (0.45 ± 0.15 cm3 m�2 atm�1

for O2 and 1.8 ± 0.6 cm3 m�2 atm�1 for CO2) was used. The gas toshrimp ratio (v/w) was approx. 3:1. Similarly, 50 ± 1 g of brined anddrained shrimp were vacuum packed (VP) for experiments B4eB11.Brined and drained MAP shrimp (Table 1, Exps. B1eB3) were stored

for (i) 12 days at 5 �C and then 9 days at 12 �C, (ii) 5 days at 12 �C andthen 16 days at 5 �C or (iii) 5 days at 10 �C, 4 days at 5 �C and then 10days at 12 �C. Brined and drained VP shrimp (Table 1, Exps. B4eB11)were stored at 5 �C. The storage temperature was recordedcontinuously throughout the experiments by data loggers (Tiny-tagPlus, Gemini Data Loggers Ltd., Chichester, UK).

Product characteristics of brined and drained shrimp weredetermined by analysis of three packages for each of the challengetests at the start of the experiments. The concentration of drymatter and the pH were measured as previously described(Dalgaard et al., 1993). Salt content was determined using anautomated potentiometric titration method. Organic acids weredetermined by using a previously described HPLC method(Dalgaard and Jørgensen, 2000). External standards were used foridentification and quantification of individual organic acids. Con-centrations of undissociated organic acids in the products werecalculated from Equation (2), using pKa values of 4.76, 4.19, 3.13,3.86 and 4.76 for acetic, benzoic, citric, lactic and sorbic acids,respectively, together with pH and measured concentrations (mM)of organic acids in the water phase of brined and drained shrimp.

Undissociated organic acid ðmMÞ ¼ Organic acid ðmMÞ1þ 10ðpH�pKaÞ (2)

The composition of headspace gas for the brined and drainedMAP shrimp in experiments B1eB3 was measured at regular in-tervals during storage using a Combi Check 9800-1 gas analyzer(PBI, Dansensor, Ringsted, Denmark).

At regular intervals during storage, three packages from each ofthe challenge tests were analysed by microbiological methods.Twenty grams of shrimp was diluted 10-fold in chilled (5 �C)physiological saline (0.85%, w/v NaCl) with 0.1% (w/v) BactoPeptone (PS) and homogenised for 60 s in a Stomacher 400 (SewardMedical, London, UK). Further appropriate 10-fold dilutions of thehomogenates were made in chilled PS. Aerobic Plate Counts (APC)were enumerated by spread plating on Long and Hammer agar with1% NaCl and incubation for 7 days at 15 �C (NMKL no.184, 2006).L. monocytogeneswas determined by spread plating on Palcam agar(Oxoid CM0877) with Palcam Selective Supplement (Oxoid SR0150)and incubation for 2 days at 37 �C. L. sakei was enumerated byspread plating on Rogosa agar (Oxoid CM0627) with pH 5.4 ± 0.2and incubation for 5 days at 25 �C using anaerobic jars with anatmosphere consisting of 90e95% N2 and 5e10% CO2.

2.3.2. Mayonnaise-based shrimp saladA total of 19 challenge tests with mayonnaise-based shrimp

salad (from now on shrimp salad) were carried out to evaluate theperformance of the new combined LmeLAB model for emulsifiedproducts with a structural composition different from e.g. brinedshrimp (Table 1, Exps. M1eM19).

Two types of brine including (i) sodium chloride, sodium ben-zoate, citric acid and potassium sorbate; or (ii) sodium chloride,sodium-acetate (Macco Organiques, Vallyfield, Quebec, Canada)and sodium-lactate (60% w/w syrup, Purac, Gorinchem, TheNetherlands) were studied. For selected experiments, pH of thebrine was adjusted with HCl/NaOH to obtain the desired pH of thefinal shrimp salad product.

When used for shrimp salad, brined and drained shrimp wereproduced as described in Section 2.3.1. The ratio (w/w) betweenshrimp and brine with acetic and lactic acids was 1.0:1.0. Mayon-naise was produced from rapeseed oil, water, egg yolks, vinegar (6%acetic acid), sodium-chloride and stabilizer (Grindsted FF5105,Danisco, Denmark) using a Stephan vacuum mixer (Stephan Ma-chinery GmbH, Hamelin, Germany). The pH of the final mayonnaisewas approx. 4.0. Shrimp salad was produced by mixing equal

Table 1Effect of product characteristics and storage conditions on growth of Lactobacillus sakei and Listeria monocytogenes in challenge tests (n ¼ 30) with brined shrimp and mayonnaise-based shrimp salad.a

Products andexperiments

Added organicacids

nb Packaging Temp. (�C) Measured product characteristics and storage conditions Growthof Lactobacillussakei

Growth ofListeriamonocytogenes

gd

Water phasesalt (%)

pH % (w/v) of organic acids in the water phase of the products % CO2c % Oil Temp. (�C)

Acetic Benzoic Citric Lactic Sorbic Lagphase(days)

mmax

(days�1)Lagphase(days)

mmax

(days�1)

Brined shrimpe

B1 Benzoic, citric,sorbic

1 MAPf Dynamicg 2.4 ± 0.0 5.8 ± 0.0 NDh 0.082 ± 0.015 0.44 ± 0.06 0.063 ± 0.001 0.049 ± 0.008 31.1 ± 1.2 e Dynamicg NSi e NSi e e


1 MAPf Dynamicj 2.4 ± 0.1 5.7 ± 0.0 NDh 0.077 ± 0.010 0.42 ± 0.04 0.046 ± 0.013 0.046 ± 0.007 33.1 ± 1.0 e Dynamicj NSi e NSi e e


1 MAPf Dynamick 2.4 ± 0.0 5.7 ± 0.0 NDh 0.068 ± 0.002 0.39 ± 0.02 0.044 ± 0.001 0.041 ± 0.001 31.7 ± 0.4 e Dynamick NSi e NSi e e

B4 Benzoic, citric 1 VPl 5.0 0.67 ± 0.07 6.1 ± 0.0 NDh 0.074 ± 0.013 0.49 ± 0.10 0.034 ± 0.014 NDh e e 5.1 ± 0.1 NSi 1.58 NSi 0.5 0.85B5 Benzoic, citric 1 VPl 5.0 0.69 ± 0.02 6.0 ± 0.1 NDh 0.14 ± 0.04 0.45 ± 0.10 0.035 ± 0.003 NDh e e 5.1 ± 0.1 NSi 1.30 NSi 0.41 0.54B6 Benzoic, citric 1 VPl 5.0 0.68 ± 0.08 6.0 ± 0.1 NDh 0.18 ± 0.07 0.36 ± 0.13 0.041 ± 0.009 NDh e e 5.1 ± 0.1 NSi 1.13 NSi 0.14 0.30B7 Benzoic, citric 1 VPl 5.0 0.65 ± 0.07 6.1 ± 0.0 NDh 0.39 ± 0.12 0.49 ± 0.21 0.054 ± 0.016 NDh e e 5.1 ± 0.1 NSi 0.98 NSi 0.07 0.00B8 Citric, sorbic 1 VPl 5.0 0.61 ± 0.07 6.1 ± 0.0 NDh NDh 0.36 ± 0.10 0.035 ± 0.008 0.019 ± 0.002 e e 5.1 ± 0.1 NSi 1.70 NSi 0.74 0.99B9 Citric, sorbic 1 VPl 5.0 0.59 ± 0.01 6.2 ± 0.1 NDh NDh 0.37 ± 0.11 0.038 ± 0.011 0.043 ± 0.022 e e 5.1 ± 0.1 NSi 1.68 NSi 0.58 0.55B10 Citric, sorbic 1 VPl 5.0 0.69 ± 0.01 6.0 ± 0.0 NDh NDh 0.48 ± 0.10 0.063 ± 0.006 0.074 ± 0.002 e e 5.1 ± 0.1 NSi 1.61 NSi 0.41 0.58B11 Citric, sorbic 1 VPl 5.0 0.63 ± 0.02 5.9 ± 0.1 NDh NDh 0.43 ± 0.09 0.056 ± 0.012 0.11 ± 0.00 e e 5.1 ± 0.1 NSi 1.51 NSi 0.34 0.42

Mayonnaise-based shrimp saladm

M1 Acetic, benzoic,citric, sorbic

1 Aerobic 5.0 2.6 ± 0.0 6.1 ± 0.0 0.14 ± 0.02 0.057 ± 0.010 0.35 ± 0.09 NDh 0.011 ± 0.009 e 37.5 5.4 ± 0.3 NSi 1.26 NSi 0.19 0.34


1 MAPf 7.0 2.6 ± 0.2 5.2 ± 0.1 0.18 ± 0.05 0.036 ± 0.004 0.34 ± 0.02 NDh 0.022 ± 0.002 34.9 ± 0.2 27.1 7.4 ± 0.2 NSi 0.98 NSi 0.055 0.40


1 MAPf 7.0 2.7 ± 0.0 5.8 ± 0.1 0.24 ± 0.03 0.030 ± 0.001 0.33 ± 0.00 NDh 0.023 ± 0.002 34.9 ± 0.5 36.6 7.4 ± 0.2 NSi 1.37 NSi 0.48 0.25


1 MAPf 7.0 2.4 ± 0.1 6.1 ± 0.0 0.20 ± 0.02 0.029 ± 0.003 0.25 ± 0.01 NDh 0.022 ± 0.003 36.1 ± 0.5 26.8 7.4 ± 0.2 NSi 1.56 NSi 0.54 1.04


1 MAPf 7.0 2.8 ± 0.3 6.0 ± 0.3 0.30 ± 0.10 0.028 ± 0.002 0.33 ± 0.01 NDh 0.022 ± 0.003 34.8 ± 0.0 34.7 7.4 ± 0.2 NSi 2.00 NSi 0.630 0.53


1 MAPf 7.0 2.3 ± 0.1 5.6 ± 0.0 0.22 ± 0.03 0.055 ± 0.003 0.41 ± 0.03 NDh 0.033 ± 0.003 35.5 ± 0.6 35.1 6.4 ± 0.2 NSi 1.48 NSi 0.1 0.78


1 MAPf 10.0 2.4 ± 0.4 5.5 ± 0.2 0.24 ± 0.01 0.057 ± 0.007 0.44 ± 0.05 NDh 0.039 ± 0.005 37.4 ± 0.3 36.6 9.1 ± 0.3 NSi 1.69 NSi 0.2 0.59


1 MAPf 10.0 2.2 ± 0.1 5.6 ± 0.0 0.25 ± 0.1 0.051 ± 0.003 0.40 ± 0.03 NDh 0.032 ± 0.002 36.6 ± 0.7 37.5 9.1 ± 0.3 NSi 2.05 NSi 0.34 0.31


1 MAPf 12.0 2.6 ± 0.2 5.6 ± 0.0 0.19 ± 0.02 0.054 ± 0.003 0.55 ± 0.05 NDh 0.032 ± 0.002 34.4 ± 0.6 27.6 11.2 ± 0.2 NSi 2.02 NSi 0.386 0.20


1 MAPf 12.0 2.4 ± 0.1 6.1 ± 0.1 0.13 ± 0.05 0.029 ± 0.001 0.25 ± 0.01 NDh 0.021 ± 0.001 37.0 ± 0.7 29.1 11.2 ± 0.2 NSi 3.71 NSi 0.86 0.72

M11 Acetic, lactic 1 Aerobic 5.0 2.9 ± 0.1 6.3 ± 0.0 0.14 ± 0.01 NDh NDh 0.088 ± 0.007 NDh e 37.1 5.4 ± 0.3 NSi 1.40 NSi 0.61 0.93M12 Acetic, lactic 1 Aerobic 7.0 2.3 ± 0.1 6.4 ± 0.0 0.52 ± 0.04 NDh NDh 0.062 ± 0.008 NDh e 33.3 7.2 ± 0.2 NSi 1.99 NSi 0.45 1.00M13 Acetic, lactic 1 MAPf 7.0 2.9 ± 0.1 5.3 ± 0.0 0.20 ± 0.01 NDh NDh 0.054 ± 0.004 NDh 35.9 ± 0.5 28.5 7.4 ± 0.2 NSi 1.59 NSi 0.17 0.15M14 Acetic, lactic 1 MAPf 7.0 3.0 ± 0.1 5.3 ± 0.0 0.19 ± 0.03 NDh NDh 0.051 ± 0.008 NDh 36.2 ± 0.2 31.2 7.4 ± 0.2 NSi 1.61 NSi 0.26 0.59M15 Acetic, lactic 1 MAPf 7.0 3.1 ± 0.6 6.2 ± 0.1 0.23 ± 0.07 NDh NDh 0.064 ± 0.017 NDh 35.0 ± 0.0 32.6 7.4 ± 0.2 NSi 2.30 NSi 0.37 0.71M16 Acetic, lactic 1 MAPf 7.0 2.7 ± 0.2 6.2 ± 0.0 0.18 ± 0.02 NDh NDh 0.049 ± 0.009 NDh 35.3 ± 0.2 32.2 7.4 ± 0.2 NSi 2.33 NSi 0.32 0.74M17 Acetic, lactic 1 MAPf 10.0 2.7 ± 0.2 5.6 ± 0.2 0.43 ± 0.06 NDh NDh 0.20 ± 0.02 NDh 36.7 ± 1.0 36.5 9.1 ± 0.3 NSi 2.17 NSi 0.3 0.03M18 Acetic, lactic 1 MAPf 10.0 2.6 ± 0.2 5.9 ± 0.1 0.39 ± 0.00 NDh NDh 0.22 ± 0.00 NDh 37.4 ± 0.0 35.1 9.1 ± 0.3 NSi 2.32 NSi 0.380 0.25M19 Acetic, lactic 1 MAPf 12.0 2.9 ± 0.0 5.7 ± 0.0 0.19 ± 0.02 NDh NDh 0.055 ± 0.006 NDh 36.7 ± 0.2 35.2 11.2 ± 0.2 NSi 3.42 NSi 0.64 0.65

a Inoculum for all experiments consisted of Lactobacillus sakei (4 strains) and Listeria monocytogenes (4 strains).b n, number of experiments.c Measured equilibrium concentrations in headspace gas.d Fitted competition factor (g) that allows the predicted concentration of Listeria monocytogenes to increase (g < 1) or decrease (g > 1) after Lactobacillus sakei has reached their maximum population density (MPD) value

(LABmax) in Equation (1).

O.M

ejlholm,P.D

algaard/Food

Microbiology

46(2015)

1e14

4

eCoo

kedan

dpee

ledco

ldwater

shrimp(Pan

dalusbo

realis).

fMod

ified

atmospherepacka

ging(M

AP)

withan

initialg

asco

mpositionof

40%CO2an

d60

%N2.

g11

.8day

sat

5� C

andthen

9day

sat

12� C

(mea

sured:11

.8day

sat

4.8±0.2

� Can

dthen

9day

sat

11.8

� C).

hND,m

easu

redbu

tnot

detected.

iNS,

not

sign

ificant.

j4.8day

sat

12� C

andthen

16day

sat

5� C

(mea

sured:4.8day

sat

11.8

±0.1

� Can

dthen

16day

sat

4.6±0.2

� C).

k4.8day

sat

10� C

,4day

sat

5� C

andthen

10day

sat

12� C

(mea

sured:4.8day

sat

9.6±0.1

� C,4

day

sat

4.5±0.1

� Can

dthen

10day

sat

11.7

±0.1

� C).

lVP,

vacu

um

packe

d.

mMay

onnaise-based

saladincludingbrined

anddrained

shrimp(Pan

dalusbo

realis).


amounts of mayonnaise and brined and drained shrimp. To allowthe equilibrium pH of shrimp salad to be reached, products werestored for 2 days at 1e2 �C prior to inoculation and packaging.

Shrimp salad was inoculated with a mixture of L. sakei (LABmix)and L. monocytogenes (Lm mix) to an initial concentration ofapprox. 1e3 log (cfu/g) and 3e5 log (cfu/g), respectively. Shrimpsalad for experiments M2eM10 and M13eM19 was packaged inmodified atmosphere as described in Section 2.3.1. For theremaining experiments (M1, M11 to M12), 100 ± 1 g of shrimpsalad was placed in plastic cups that were closed with a lid. Shrimpsalad was stored at constant temperatures between 5 and 12 �C(Table 1), which were recorded by data loggers as described abovein Section 2.3.1.

Product characteristics of shrimp salad were determined byanalysis of three samples from each experiment at the start of thechallenge tests. Dry matter, pH, salt, organic acids and thecomposition of headspace gas in MAP was determined asdescribed above in Section 2.3.1. The method of Bligh and Dyer(1959) was used to determine the oil content of shrimp salad.In addition to the classical HendersoneHasselbalch equation(Equation (2)), a modified version (Equation (3)) taking parti-tioning of organic acids between the oil and the water phase intoaccount was also used to calculate the concentration of undis-sociated organic acids in the water phase of the products (Wilsonet al., 2000).

Undissociated organic acid ðmMÞ ¼ Organic acid ðmMÞ1þ Kp

�f

1�f

�þ 10ðpH�pKaÞ

(3)

where Kp is the partition coefficient defined as the ratio betweenthe concentration of undissociated organic acid in the oil and thewater phase, respectively, and f is the volume fraction of oil.Partition coefficients (Kp values) of 0.02, 7.22, 0.033 and 4.19 wereused for acetic, benzoic, lactic and sorbic acids, respectively, asestimated for sunflower, rapeseed or fish oil:water systems at 25 �C(Brocklehurst and Wilson, 2000; Cheng et al., 2010). For citric acidno usable Kp values were found in the literature and, thus, a Kp

value of 0.01 was assumed and used due to the low solubility of thisorganic acid in oil.

2.4. Curve fitting, statistical analyses and model simulation

To determine lag phase (days) and maximum specific growthrate (mmax,1/days) of L. monocytogenes and L. sakei in challenge testswith brined and drained shrimp and shrimp salad the integratedand log-transformed form of the logistic model with delay(Dalgaard, 2009; Rosso et al., 1996) was fitted to log(cfu/g) dataobtained on the different types of agar. MS Excel with Solver add-inwas used for curve fitting (Microsoft Corp., Redmond, WA, USA). AnF test to compare fits of the logistic models with and without delaywas used to evaluate whether lag phases of the microbial growthcurves were significant. Simultaneous growth of L. monocytogenesand LAB was predicted by Equation (1) and using numerical inte-gration with a time step of 0.43e2.35 h depending on the storagetemperature (MS Excel).

3. Results and discussion

3.1. Challenge tests with processed seafood

L. monocytogenes and L. sakei grew without a significant lagphase in all experiments (n ¼ 30) with processed seafood and

Table 2Product characteristics and storage conditions in experiments (n ¼ 82) used for evaluation of the new combined model for Listeria monocytogenes and psychrotolerant Lactobacillus spp.

Experiments Products References na Inoculum Packaging Temp. (�C) Water phasesalt (%)

pH Phenol(ppm)

% CO2b % (w/v) of organic acids in the water phase of the products

Acetic Benzoic Citric Diacetate Lactic Sorbic

S1 Brined shrimpc Mejlholm andDalgaard(2009)

2 Mixture(8 strains)d

Aerobic 5.4 2.7e2.9 6.23e6.29 e e 0e0.17 0e0.037 0e0.29 NDe 0.050e0.12 0e0.016

S2 Brined shrimpc Mejlholm et al.(2012)


MAPf 7.7 1.8e2.3 5.72e6.28 e 35.0e37.9 0e0.85 0e0.096 0e0.54 NDe 0.050e0.97 0e0.039

S3 Brined shrimpc Mejlholm andDalgaard(2013)


MAPf 4.2e9.6 2.5 5.81 e 35.8e36.4 NDe 0.14 0.57 NDe 0.05 0.05

S4 Brined crayfishg Mejlholm andDalgaard(2013)


MAPf 4.2e9.6 2.2 5.95 e 35.4e35.9 NDe NDe 0.73 NDe NDe 0.13

S5 Brined surimi Mejlholm andDalgaard(2013)


MAPf 4.2e9.6 2.9 5.72 e 36.3e37.5 NDe 0.1 0.32 NDe NDe NDe

S6 Lumpfish roe Mejlholm andDalgaard(2013)


MAPf 4.2e9.6 5.4e6.4 5.45e5.93 e 36.1e37.4 NDe 0.16 0.26e0.46 NDe NDe NDe

S7 Cold-smokedsalmon

Mejlholm andDalgaard(2007a)

4h Listeriamonocytogenes(4 strains)

VPi or MAPf 7.8e14.3 3.2e5.3 5.94e6.10 8.9e14.1 0e30.0 NDe NDe NDe 0e0.26 0.93e1.49 NDe

S8 Cold-smokedsalmon

Mejlholm andDalgaard(2007b)

5 Mixture(9 strains)j

MAPf 7.9 5.2 6.02e6.22 3.8 25.1e61.0 NDe NDe NDe 0e0.30 0.66e2.93 NDe

S9 Gravad salmon Mejlholm andDalgaard(2007a)


MAPf 14.3 3.0e3.1 6.11e6.26 4.8e5.0 25.0 NDe NDe NDe 0e0.17 0.93e0.98 NDe

S10 Marinated salmon Mejlholm andDalgaard(2007a)


MAPf 7.8 3.6 6.08e6.24 e 21.5 NDe NDe NDe 0e0.16 0.50e0.65 NDe

S11 Smoked salmon Vermeulenet al. (2011)


VPi 8.0 4.2e5.7 6.05e6.11 2.2e6.2 e e e e e 1.01e1.09 e

S12 Cold-smokedGreenland halibut



VPi or MAPf 7.6e7.7 3.7e4.6 6.17e6.50 16.2e21.2 0e26.0 NDe NDe NDe 0e0.23 0.14e3.16 NDe

S13 MarinatedGreenland halibut



MAPf 7.8 3.4 6.68e6.84 e 16.2 NDe NDe NDe 0e0.15 0.25e0.89 NDe

S14 Seafood saladk Hwang andTamplin (2005)

30 Listeriamonocytogenes(6 strains)

Aerobicor VPi

4.0e12.0 2.5l 6.3e6.6 e e 0.063l e e e e e

a n, number of experiments.b Measured equilibrium concentrations in headspace gas.c Cooked and peeled cold water shrimp (Pandalus borealis).d Inoculum consisting of Lactobacillus sakei (4 strains) and Listeria monocytogenes (4 strains).e ND, measured but not detected.f Modified atmosphere packaging (MAP) with an initial gas composition of 40% CO2 and 60% N2.g Cooked and peeled Louisiana crayfish (Procambarus clarkii).h Naturally contaminated with LAB.i VP, vacuum packed.j Inoculum consisting of Listeria monocytogenes (4 strains), Lactobacillus sakei (1 strain), Lactobacillus curvatus (1 strain), Carnobacterium maltaromaticum (1 strain), Enterococccus malodoratus (1 strain) and Leuconostoc spp. (1

strain).k Seafood salad including mayonnaise and a mixture of cooked shrimp and imitation crabmeat. The ratio (w/w) between mayonnaise and seafood was 1:3.l Bold typed: Assumed values.

O.M

ejlholm,P.D

algaard/Food

Microbiology

46(2015)

1e14

6

0 5 10 15 20 25

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

Log

(cfu

/g)

5 °C 12°C

(a)

0 5 10 15 20 25

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

Log

(cfu

/g)

12°C 5°C

(b)

0 5 10 15 20 25

Storage time (days)

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

Log

(cfu

/g)

5°C10°C 12°C

(c)

Fig. 1. Observed and predicted growth of Listeria monocytogenes (▫) and Lactobacillussakei (B) during storage of brined and drained shrimp in modified atmospherepackaging (MAP). Brined and drained MAP shrimp including benzoic, citric and sorbicacids were stored for 11.8 days at 4.8 �C and then 9 days at 12.1 �C (a) (Exp. B1), 4.8days at 11.8 �C and then 16 days at 4.6 �C (b) (Exp. B2), and 4.8 days at 9.6 �C, 4 days at4.5 �C and then 10 days at 11.7 �C (c) (Exp. B3). Growth of L. monocytogenes (solid line)and L. sakei (dashed line) was predicted by the new combined LmeLAB model of thepresent study (Equation (1)).


shrimp salad irrespective of product characteristics and storageconditions (Table 1).

3.2. Evaluation of the new combined model for L. monocytogenesand psychrotolerant Lactobacillus spp.

Compared to the most extensive and available microbial inter-action model (Mejlholm and Dalgaard, 2007a), the new combinedmodel for L. monocytogenes and LAB was expanded with the effectof nitrite as well as acetic, benzoic, citric and sorbic acids to includea total of 12 environmental parameters. For the 112 experimentsused to evaluate the performance of the new combined LmeLABmodel, between 4 and 8 environmental parameters were used at atime as input (Tables 1 and 2). The effect of all the included envi-ronmental parameters was evaluated with the exception of nitritenot being allowed as additive for seafood products.

3.2.1. Brined and drained MAP shrimp at dynamic storagetemperatures

For the three dynamic storage temperature scenarios (Table 1),growth of L. monocytogenes was accurately predicted with respectto both the effect of fluctuating storage temperatures andmicrobialinteraction (i.e. the classical Jameson effect) (Fig. 1). All cell con-centrations of L. monocytogenes were within the acceptable simu-lation zone (ASZ) of ±0.5 log-units and average maximumpopulation density (MPD) values of 5.3 and 5.2 log (cfu/g) wereobserved and predicted, respectively (i.e. predicted as 1.3 times toolow). Without the effect of microbial interaction, an average MPDvalue of 6.5 log (cfu/g) was predicted for L. monocytogenes (i.e. 16times too high) (Result not shown). For L. sakei, the percentage ofcell counts within the ASZ varied between 63 and 88% for the threeexperiments. Previously, Oscar (2005) suggested that the percent-age of observations within the ASZ should be at least 70% toconsider the performance of a predictive model acceptable. Onaverage, 75% of the cell counts for L. sakeiwere within the ASZ withthe greatest difference between observed and predicted valuesbeing 0.8 log (cfu/g) (Results not shown).

3.2.2. Processed seafoodThe performance of the new combined LmeLAB model was

evaluated on data from 60 experiments with processed seafood(Tables 1 and 2). For products including benzoic, citric and/or sorbicacids (n ¼ 21), average MPD values of 4.6 and 4.5 log (cfu/g) wereobserved and predicted, respectively for L. monocytogenes (i.e.predicted as 1.3 times too low) (Fig. 2, Table 3). Without the effect ofmicrobial interaction, an average MPD value of 5.4 log (cfu/g) waspredicted for L. monocytogenes (i.e. 6 times too high). Average MPDvalues of 5.0 and 4.4 log (cfu/g) was observed and predicted,respectively, for L. monocytogenes (i.e. predicted as 4 times too low)in products including acetic acid/diacetate and lactic acid (n ¼ 39)(Fig. 2, Table 3). Without the effect of microbial interaction, anaverage MPD value of 6.3 log (cfu/g) was predicted forL. monocytogenes (i.e. predicted as 20 times too high). When eval-uated exclusively on products where growth of L. monocytogeneswas observed the importance of including microbial interactionwas even more evident. For products with benzoic, citric and/orsorbic acids (n ¼ 11) and acetic acid/diacetate and lactic acid(n ¼ 29), average MPD values of L. monocytogenes was predicted tobe 12 and 60 times too high, respectively, when the effect of mi-crobial interaction was ignored. With the effect of microbial inter-action, the corresponding MPD values were predicted to be 2 and 7times to low. Importantly, it was found that prediction of too lowMPD values for L. monocytogenes mainly was observed for experi-ments where the initial cell concentration of Lactobacillus spp. wasnot reported and as an alternative the initial total cell concentration

of LAB as determined on NAP or MRS agar was used as input for thenew combined LmeLAB model (Fig. 2). For these experiments(n ¼ 25) (Table 2, Exps. S7, S9eS13), average MPD values of 5.4 and4.6 log (cfu/g) were observed and predicted, respectively, for


L. monocytogenes (i.e. predicted as 6 times too low). However, forproducts inoculated with a well-defined concentration of L. sakei(n ¼ 35) (Tables 1 and 2, Exps. B4eB11, S1eS6 and S8), a betteragreement between observations and predictions was found asindicated by average MPD values of 4.5 and 4.3 log (cfu/g),respectively, for L. monocytogenes (i.e. predicted as 1.6 times toolow). Previously it has been found that Lactobacillus spp. typicallygrows faster than other LAB (e.g. Carnobacterium spp. and Leuco-nostoc spp.) of relevance to chilled seafood and meat products(Borch et al., 1996; Devlieghere and Debevere, 2000), and this wasalso shown during the evaluation of the Lactobacillus spp. modelincluded in the new combined LmeLAB model (Mejlholm andDalgaard, 2013). In a constructed example with cold-smokedsalmon stored at 5 �C and having typical product characteristics(pH 6.1, 3% WPS, 10 ppm phenol and 0.80% water phase lactic acid),three different scenarios were used to illustrate the importance ofthe initial cell concentration of LAB and its composition on pre-dicted MPD values of L. monocytogenes (Fig. 3). Assuming an initialcell concentration of 0 and 2 log (cfu/g) for Lactobacillus spp. theonset of microbial interaction, defined as the time when LABt iswithin 0.05 log (cfu/g) of LABmax (in this example set as 8.5 log (cfu/g)) in Equation (1), was predicted as 24 and 19 days, respectively,with corresponding MPD values of 3.2 and 2.5 log (cfu/g) forL. monocytogenes (Fig. 3, dashed and solid lines). For Carnobacte-rium spp. and Leuconostoc spp. with an initial cell concentration of 2log (cfu/g) the onset of microbial interaction was predicted as 29days with a corresponding MPD value of 3.8 log (cfu/g) forL. monocytogenes (Fig. 3, dotted line). Growth rates (mmax values) ofCarnobacterium spp. and/or Leuconostoc spp. were assumed to be35% slower than for Lactobacillus spp. based on recent findings(Mejlholm and Dalgaard, 2013). Thus, it was predicted that even ifLactobacillus spp. constitutes only 1% of the initial cell concentra-tion of LAB it would outgrow e.g. Carnobacterium spp. and Leuco-nostoc spp. and determine the onset of microbial interaction. Ourresults shows that if Lactobacillus spp. is known to constitute thedominating microbiota of a certain product it is important todetermine their initial cell concentration in order to accuratelypredict the MPD value of L. monocytogenes. A measure of less thanthe detection limit for Lactobacillus spp. as determined on e.g.Rogosa agar is clearly preferable compared to measures of “total”LAB determined on e.g. NAP or MRS agar. The use of too high initial

2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0

MPD (log cfu/g) - observed

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

MPD

(log

cfu/

g) -

pred

icted

Line of Perfect Match (observed = predicted)

Fig. 2. Comparison of observed and predicted maximum population density (MPD)values (log cfu/g) of Listeria monocytogenes in processed seafood including benzoic,citric and sorbic acids (▫) and acetic acid/diacetate and lactic acid (B,▵). Processedseafood was inoculated with a mixture of L. monocytogenes and Lactobacillus sakeistrains (▫,B) or a mixture of L. monocytogenes strains (▵). MPD values ofL. monocytogenes were predicted using the new combined LmeLAB model of thepresent study.

cell concentrations for LAB would result in prediction of microbialinteraction earlier than observed and, consequently, the estimatedMPD values of L. monocytogeneswould be too low. If LAB other thanLactobacillus spp. is known to constitute the dominating microbiotaalternative mathematical models for LAB should be used to avoidprediction of too low MPD values for L. monocytogenes. Theimportance of Carnobacterium spp. and Leuconostoc spp. as spoilagebacteria has previously been documented for seafood and meatproducts (Bj€orkroth, 2005; Leisner et al., 2007; Nieminen et al.,2012; Pothakos et al., 2014), however, mathematical models witha complexity similar to our model for Lactobacillus spp. are not yetavailable for these types of LAB and further developments areneeded in this area. Obviously, it would be very time-consumingand expensive for the seafood industry to determine the initialcell concentration of LAB for each batch of the production. Alter-natively, samples should be collected and analysed at regular in-tervals (e.g. twice a month) to build knowledge about the initial cellconcentration of LAB for relevant types of products. In a conser-vative approach, the lowest estimated initial cell concentration ofLAB should be used as input for the new combined LmeLAB modelas this would result in the highest predicted MPD value ofL. monocytogenes (i.e. longer time to the onset of microbialinteraction).

The initial cell concentration of L. monocytogenes in inoculatedsamples of processed seafood used to evaluate the performance ofthe new combined LmeLAB model was in the range of 1.5e5.2 log(cfu/g), being much higher than typical initial cell concentrations ofthe pathogen in naturally contaminated products. In a recent studywith naturally contaminated samples of cold-smoked Greenlandhalibut and cold-smoked salmon, it was shown that MPD values ofL. monocytogenes were accurately predicted by the LmeLAB modelof the present study, even when initial cell concentrations of thepathogen were as low as �0.49 log (cfu/g) (Mejlholm et al., 2014).Comparable results were found for a previous version of ourcombined LmeLAB model when evaluated on data from naturallycontaminated samples of cold-smoked salmon (Mejlholm andDalgaard, 2007a). These results indicate that the new combinedLmeLAB model is valid for initial cell concentrations ofL. monocytogenes being much lower than the ones examined in thepresent study and, thus, can be used to predict the time needed toreach the critical limit of 2.0 log (cfu/g) (EC, 2005).

Fifty-four of 60 growth and no-growth responses ofL. monocytogenes in processed seafood were correctly predicted,corresponding to 90%. The six incorrect predictions were distrib-uted as two fail-dangerous (3%) and four fail-safe predictions (7%)(Table 3). These new results are in agreement with previous vali-dation studies of our model for L. monocytogenes on the ability topredict growth and no-growth responses (Mejlholm et al., 2010;Mejlholm and Dalgaard, 2009). In the present study, no clear dif-ferences were noticed between products with different types/compositions of organic acids (i.e. benzoic, citric and sorbic acids oracetic acid/diacetate and lactic acid) with respect to prediction ofgrowth and no-growth responses of L. monocytogenes (Table 3).Predicted mmax values of L. monocytogenes and LAB as evaluated byaverage Bf and Af values were within the acceptable range set forthe pathogen and spoilage bacteria, respectively (Table 3)(Mejlholm and Dalgaard, 2013; Ross, 1999). The individual modelsfor L. monocytogenes and Lactobacillus spp. have previously beensuccessfully validated for their ability to predict growth rates of thetwo types of microorganisms (Mejlholm and Dalgaard, 2009, 2013),and this is a prerequisite in order to accurately predict microbialinteraction. Notably, for experiments where the initial cell con-centration and growth of LAB was determined on NAP or MRS agar(n ¼ 25), average Bf and Af values of 1.20 and 1.28 were obtained.This confirms that growth of LAB in these experiments on average

Table 3Comparison of observed and predicted growth responses of lactic acid bacteria and Listeria monocytogenes in processed seafood (n ¼ 60) and mayonnaise-based seafood salads (n ¼ 49).a

Products andexperiments

Environmentalparameters

nc Observed growth/no-growth ofListeria monocytogenes

Bias/accuracy factor values based on comparisonof observed and predicted mmax values

Prediction of growth/no-growthresponses by L. monocytogenesb

Maximum population density (log cfu/g) ofL. monocytogenes

Growth No-growth Model for psychrotolerantLactobacillusspp. (Mejlholm andDalgaard, 2013)d,e

Model forL. monocytogenes(Mejlholm andDalgaard, 2009)f,g

Correct Fail-safeh Fail-dangerousi

Observed Predictedwithoutmicrobialinteractionj

Predicted withmicrobialinteractionk

Predictedwithmicrobialinteractionl

Processed seafoodAdded organic acids

B4eB11, S1eS6 Benzoic, citric,sorbic acids

21 11 10 0.91/1.24 (n ¼ 21)m 0.88/1.42 (n ¼ 10)m 17 (81%) 3 (14%) 1 (5%) 4.6 ± 1.3 5.4 ± 2.1 4.5 ± 1.3 e

S1eS2, S7eS13 Acetic acid/diacetate,lactic acid

39 29 10 1.07/1.27 (n ¼ 39)m 1.04/1.36 (n ¼ 28)m 37 (95%) 1 (3%) 1 (3%) 5.0 ± 1.8 6.3 ± 2.5 4.4 ± 1.2 e

Average 1.01/1.26 (n ¼ 60)m 0.99/1.38 (n ¼ 38)m 54 (90%) 4 (7%) 2 (3%) 4.9 ± 1.6 6.0 ± 2.4 4.4 ± 1.2 e

Mayonnaise-based saladsS14 e 30 30 0 e 1.07/1.30 (n ¼ 30)k 30 (100%) 0 (0%) 0 (0%) e e e e

Added organic acidsM1eM10 Acetic, benzoic, citric,

sorbic acids10 10 0 0.75/1.34 (n ¼ 10)m 0.84/1.61 (n ¼ 4)m 4 (40%) 0 (0%) 6 (60%) 6.1 ± 1.1 5.4 ± 2.1 4.5 ± 1.1 4.9 ± 1.7

M11eM19 Acetic, lactic acids 9 9 0 0.73/1.37 (n ¼ 9)m 1.13/1.32 (n ¼ 5)m 5 (56%) 0 (0%) 4 (44%) 6.4 ± 0.9 6.3 ± 2.5 4.9 ± 1.4 5.6 ± 1.8

pHM1, M4eM5, M10eM12,

M15eM16>6.0 8 8 0 0.77/1.29 (n ¼ 8)m 1.01/1.48 (n ¼ 8)m 8 (100%) 0 (0%) 0 (0%) 6.7 ± 0.7 8.0 ± 0.8 5.9 ± 0.7 6.8 ± 0.9

M3, M6, M8, M18eM19 5.6e6.0 5 5 0 0.71/1.41 (n ¼ 5)m 0.84/1.19 (n ¼ 1)m 1 (20%) 0 (0%) 4 (80%) 6.6 ± 1.0 4.9 ± 2.1 4.1 ± 0.6 4.6 ± 1.5M2, M7, M9,

M13eM14, M17<5.6 6 6 0 0.72/1.40 (n ¼ 6)m e 0 (0%) 0 (0%) 6 (100%) 5.4 ± 1.0 3.6 ± 0.2 3.6 ± 0.2 3.6 ± 0.2

Average 0.74/1.36 (n ¼ 19)m 0.99/1.44 (n ¼ 9)m 9 (43%) 0 (0%) 10 (57%) 6.2 ± 1.0 5.8 ± 2.3 4.7 ± 1.2 5.2 ± 1.7

Mayonnaise-based salads (with predictions based on partitioning of organic acids between the oil and the water phase of the products)n

Added organic acidsM1eM10 Acetic, benzoic, citric,

sorbic acids10 10 0 0.88/1.15 (n ¼ 10)m 0.62/2.10 (n ¼ 7)m 7 (70%) 0 (0%) 3 (30%) 6.1 ± 1.1 5.9 ± 2.0 4.8 ± 1.1

M11eM19 Acetic, lactic acids 9 9 0 0.73/1.37 (n ¼ 9)m 1.14/1.32 (n ¼ 5)m 5 (56%) 0 (0%) 4 (44%) 6.4 ± 0.9 6.3 ± 2.5 4.9 ± 1.4

pHM1, M4eM5, M10eM12,

M15eM16pH > 6.0 8 8 0 0.79/1.26 (n ¼ 8)m 1.11/1.46 (n ¼ 8)m 8 (100%) 0 (0%) 0 (0%) 6.7 ± 0.7 8.0 ± 0.8 5.9 ± 0.7

M3, M6, M8, M18eM19 pH 5.6e6.0 5 5 0 0.78/1.28 (n ¼ 5)m 0.39/2.54 (n ¼ 3)m 3 (60%) 0 (0%) 2 (40%) 6.6 ± 1.0 4.9 ± 2.1 4.1 ± 0.6M2, M7, M9,

M13eM14, M17pH < 5.6 6 6 0 0.83/1.21 (n ¼ 6)m 0.49/2.05 (n ¼ 1)m 1 (17%) 0 (0%) 5 (83%) 5.4 ± 1.0 3.6 ± 0.2 3.6 ± 0.2

Average 0.80/1.25 (n ¼ 19)m 0.80/1.73 (n ¼ 12)m 12 (63%) 0 (0%) 7 (37%) 6.2 ± 1.0 6.1 ± 2.2 4.9 ± 1.2

a See Tables 1 and 2 for information on product characteristics and storage conditions in the different experiments.b Predicted by the Listeria monocytogenes model of Mejlholm and Dalgaard (2009).c n, number of experiments.d Experiments B4eB11 have not previously been used to validate the performance of the model for psychrotolerant Lactobacillus spp., whereas experiments S1eS13 previously have been used (Mejlholm and Dalgaard, 2013).e Interpretation of the bias factor value for graduating the performance of the model for lactic acid bacteria: (i) 0.85e1.25 good/acceptable and (ii) <0.85 or >1.25 unacceptable (Mejlholm and Dalgaard, 2013).f Experiments B4eB11, S2eS6 and S11 have not previously been used to validate the performance of themodel for Listeria monocytogenes, whereas experiments S1, S7eS10 and S12eS13 previously have been used (Mejlholm

and Dalgaard, 2009; Mejlholm et al., 2010).g Interpretation of the bias factor value for graduating the performance of the model for Listeria monocytogenes: (i) 0.95e1.11 good; (ii) 0.87e0.95 or 1.11e1.43 acceptable and (iii) <0.87 or >1.43 unacceptable (Ross, 1999).h Growth observed and no-growth predicted.i No-growth observed and growth predicted.j Predicted by the new combined model for Listeria monocytogenes and psychrotolerant Lactobacillus spp. (using Equation (1) with a g value of 0.0).k Predicted by the new combined model for Listeria monocytogenes and psychrotolerant Lactobacillus spp. (using Equation (1) with a g value of 1.0).l Predicted by the new combined model for Listeria monocytogenes and psychrotolerant Lactobacillus spp. (using Equation (1) with fitted g values).

m Values in parentheses indicate the number of experiments used to calculate average bias and accuracy factor values.n See explanation in Section 2.3.2.

O.M

ejlholm,P.D

algaard/Food

Microbiology

46(2015)

1e14

9


was slower than predicted by the Lactobacillus spp. model ofMejlholm and Dalgaard (2013) which to some degree explains theprediction of slightly too low MPD values for L. monocytogenes aspreviously discussed in this section (Fig. 2).

3.2.3. Mayonnaise-based shrimp saladContrary to brined shell-fish, and cold-smoked and marinated

fish products, the performance of our predictive models forL. monocytogenes and Lactobacillus spp. (Mejlholm and Dalgaard,2009, 2013) had not previously been evaluated for mayonnaise-based salads. Nine of 19 growth and no-growth responses ofL. monocytogenes in shrimp salad were correctly predicted, corre-sponding to 43%. All incorrect predictions were fail-dangerous(57%) (Table 3). With respect to prediction of growth and no-growth responses of L. monocytogenes, no clear differences wereobserved between shrimp salad with different types/compositionsof organic acids. However, a clear correlation was observed be-tween the percentage of fail-dangerous predictions and the pH ofthe shrimp salad. At pH values higher than 6.0 (n ¼ 8), growth ofL. monocytogenes was correctly predicted for all challenge tests,whereas for the pH intervals 5.6e6.0 (n ¼ 5) and lower than 5.6(n ¼ 6) the percentage of fail-dangerous predictions was 80% and100%, respectively (Table 3). The high percentage of fail-dangerouspredictions clearly indicated that growth of L. monocytogenes wasseverely underestimated in shrimp salad with pH lower than 6.0.Most critically, L. monocytogeneswas observed to increase by 3.4 log(cfu/g) in one of the challenge tests where no-growth of thepathogenwas predicted (Table 1, Exp. M3). The lowest pH of shrimpsalad where growth of L. monocytogenes was both observed andcorrectly predicted was 5.7 (Table 1, Exp. M19). For this experimentand the 8 challenge tests with pH values higher than 6.0, average Bfand Af values of 0.99 and 1.44 were calculated. Thus, an acceptableagreement between observed and predicted mmax values ofL. monocytogenes was found for these experiments. Hwang andTamplin (2005) examined the growth of L. monocytogenes in sea-food salad (mixture of mayonnaise and shrimp/imitated crabmeat)with pH values in the range of 6.3e6.6 (Table 2). For these 30challenge tests carried out at 4, 8 and 12 �C, average Bf and Af valuesof 1.07 and 1.30 were obtained when the L. monocytogenesmodel ofMejlholm and Dalgaard (2009) was used. For L. sakei, average Bf andAf values of 0.74 and 1.36 were obtained for the 19 challenge testswith shrimp salad (Table 3). Unlike for L. monocytogenes, the abilityto predict growth of L. sakeiwas not influenced by pH of the shrimpsalad (Table 3). The more consistent performance of the model forLactobacillus spp. compared to the model for L. monocytogenes waspartly explained by the fact that the examined combinations ofproduct characteristics and storage conditions were located closerto the growth boundary of the pathogen (McMeekin et al., 2002;Mejlholm and Dalgaard, 2009; Vermeulen et al., 2007b). In thissituation, changes in the product characteristics (e.g. pH) of shrimpsalad during the duration of the challenge tests would have a largereffect on growth of L. monocytogenes than on LAB. In the presentstudy, following mixing of mayonnaise and brined and drainedshrimp, the product was allowed to stand for 2 days at 1e2 �C priorto inoculation and packaging, and during this period the equilib-rium pH of the shrimp salad was almost reached (Fig. 4, Exp. M6 asan example). For the 10 fail-dangerous predictions (Table 3) it wasestimated that pH values of the shrimp salad should have beenbetween 0.3 and 0.6 units higher in order to correctly predictgrowth/no-growth of L. monocytogenes and to obtain a Bf value of1.0 for the observed and predicted mmax values. Such increases inthe product pH of shrimp salad are unlikely and other reasonsshould be identified to explain the discrepancy between observedand predicted growth responses of L. monocytogenes at pH valueslower than 6.0. The potential occurrence of micro environments

within the shrimp salad having e.g. higher pH and/or lower con-centrations of organic acids than the average product could to somedegree explain the underestimation of growth by L. monocytogenesas this might result in high locally distributed cell concentrations ofthe pathogen. In future studies, the use of micro electrodes formeasurement of e.g. pH and organic acids could be relevant forevaluating such locally distributed differences in the productcharacteristics of mayonnaise-based salads.

In an attempt to explain and improve the performance of themodels for L. monocytogenes and Lactobacillus spp. (Mejlholm andDalgaard, 2009, 2013), partitioning of organic acids between theoil and the water phase of the products was taken into account(Equation (3)). In food, preservatives are partitioned in the waterand the lipid phase depending on their solubility in each of the twophases. Microorganisms are hydrophilic and almost exclusivelypresent in the water phase of food. Therefore, the proportion of apreservative dissolved in the lipid phase of a food will notcontribute to the inhibition of microbial growth. Benzoic and sorbicacids are more soluble in lipids than acetic and lactic acids, whichare almost exclusively dissolved in the water phase (Brocklehurstand Wilson, 2000). Previously, evaluation of model performancefor products including benzoic and sorbic acids has primarily beencarried out for brined shrimp with a lipid content of 1e2%(Mejlholm and Dalgaard, 2009, 2013). For these products, only avery small proportion of the total concentration of benzoic andsorbic acids would be dissolved in the lipid phase and the impor-tance of partitioning could be ignored without affecting the pre-dicted growth of e.g. L. monocytogenes. For shrimp salad the contentof lipids was substantially higher in the range of 27e38%, with themajority originating from the mayonnaise (i.e. rapeseed oil)(Table 1). Using Equation (3), 73e81% and 61e72% of benzoic andsorbic acids, respectively, was estimated to be dissolved in the lipidphase of the shrimp salad with the remaining part being distrib-uted in the water phase (Table 1, Exps. M1eM10). In comparison,less than 1% of acetic, citric and lactic acids was estimated to bedissolved in the lipid phase of the shrimp salad (Table 1, Exps.M1eM19). The performance of the models for L. monocytogenesand Lactobacillus spp. were improved for shrimp salad includingbenzoic and sorbic acids due to the lower estimated concentrationsof these organic acids in the water phase of the products. ForL. monocytogenes, the number of fail-dangerous predictions wasreduced from six to three in shrimp salad with benzoic and sorbicacids. However, by including the three new correctly predictedgrowth responses, the average Bf value changed from 0.84 to 0.62,indicating that mmax values of L. monocytogenes were severelyunderestimated in these challenge tests (Table 3). The overall per-formance of the model for Lactobacillus spp. was improved withrespect to prediction of mmax values when partitioning of organicacids between the oil and thewater phase of shrimp salad includingbenzoic and sorbic acids were taken into account (Table 3). As theconcentrations of acetic and lactic acids in the water phase ofshrimp salad were only slightly reduced, the performance of theevaluated models was almost unchanged for these products(Table 3).

Average MPD values of 6.2 and 4.7 log (cfu/g) were observedand predicted, respectively, for L. monocytogenes (i.e. 32 times toolow) in shrimp salad (n ¼ 19). Comparable results were obtainedfor shrimp salad with different types/compositions of organicacids (Table 3). Two main reasons were identified to explain theprediction of too low MPD values for L. monocytogenes. Firstly, forshrimp salad with pH values lower than 6.0, no-growth ofL. monocytogenes was predicted for 10 of 11 challenge tests eventhough growth of the pathogen was observed for all these ex-periments with increments between 0.7 and 3.8log(cfu/g) (i.e.5e6300 times higher than predicted). Secondly, for 11 of the 19

0 5 10 15 20 25 30 35 40

Storage time (days) at 5°C

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

Log

( cfu

/g)

Fig. 3. The influence of the initial cell concentration of lactic acid bacteria (LAB) on thepredicted maximum population density (MPD) value of Listeria monocytogenes in aconstructed example with cold-smoked salmon having typical product characteristics(pH 6.1, 3% water phase salt, 10 ppm phenol and 0.80% water phase lactic acid). Cold-smoked salmon with an initial cell concentration of (i) 0 log (cfu/g) for L. mono-cytogenes (solid bold line) and L. sakei (solid fine line), (ii) 0 log (cfu/g) for L. mono-cytogenes (dashed bold line) and 2 log (cfu/g) for L. sakei (dashed fine line), and (iii)0 log (cfu/g) for L. monocytogenes (dotted bold line) and 2 log (cfu/g) for Carnobacte-rium/Leuconostoc spp. (dotted fine line). Growth of L. monocytogenes and LAB waspredicted by the new combined LmeLAB model of the present study. The growth rate(mmax, days�1) of Carnobacterium/Leuconostoc spp. was assumed to be 35% slower thanfor L. sakei based on recent findings (Mejlholm and Dalgaard, 2013).


challenge tests, cell concentrations of L. monocytogenes wereobserved to increase by 0.5e1.7 log (cfu/g) after L. sakei hadreached their MPD value (Fig. 5). Thus, the approach used tomodel microbial interaction (i.e. the classical Jameson effect) be-tween L. monocytogenes and L. sakei seemed not to be generallyapplicable for shrimp salad. Prediction of too low MPD values forL. monocytogenes and Salmonella due to continued growth afterthe expected onset of microbial interaction has previously beenreported for e.g. brined shrimp and fresh pork meat (Besse et al.,2006; Mejlholm et al., 2012; Møller et al., 2013). In contrast,Cornu et al. (2011) showed that predicted MPD values ofL. monocytogenes in cooked and smoked diced bacon were toohigh when the classical Jameson effect was used to model mi-crobial interaction. In the present study, predicted MPD values of

0 10 20 30 40 50

Time (hours)

3.5

4.0

4.5

5.0

5.5

6.0

6.5

pH

0.0

2.0

4.0

6.0

8.0

10.0

T emp (°C

)

Fig. 4. Changes in pH of mayonnaise-based shrimp salad following mixing ofmayonnaise and brined and drained shrimp in a ratio (w/w) of 1:1 (▫) (Exp. M6,Table 1). The temperature profile is shown by a solid line.

L. monocytogenes in brined and drained MAP shrimp (Table 1, Exps.B4eB11), comparable to the ones used for shrimp salad, were onaverage 0.3 log (cfu/g) lower than the observed values. However,this can only partly explain the prediction of too low MPD valuesfor L. monocytogenes in shrimp salad. Recently, Møller et al. (2013)described the continued but reduced growth of Salmonella inground pork after the natural microbiota had reached their MPDby using an expanded Jameson effect model. This model is moreflexible than the classical Jameson effect model due to the intro-duction of a temperature dependent competition factor (g) thatallows the predicted concentration of e.g. Salmonella to increase(g < 1) or decrease (g > 1) after the natural microbiota has reachedits MPD value. With a g value of 1.0 the expanded and the classicalJameson effect model are identical. In the present study, g valueswere determined from the 19 challenge tests with shrimp salad(Table 1, Exps. M1eM19) using a two-step fitting procedure forL. monocytogenes. Firstly, mmax values of L. monocytogenes weredetermined on growth data obtained from the beginning of theexperiments and until the time where LAB reached their MPDvalue (LABmax in Equation (1)). For LAB, mmax values were deter-mined on growth data obtained during the full length of the ex-periments. Secondly, g in Equation (1) was estimated to describethe development in L. monocytogenes from the time of LABmax anduntil the end of the experiments. For the 19 challenge tests withshrimp salad, g values between 0.03 and 1.04 were determined,thus, ranging from no microbial interaction betweenL. monocytogenes and LAB to microbial interaction as described bythe classical Jameson effect (g ¼ 1.0). Contrary to Møller et al.(2013), no significant correlation (p ¼ 0.27) between g valuesand storage temperatures was determined in the present study.Using the estimated g values and the expanded Jameson effectmodel (Equation (1)), average MPD values of 6.7 and 6.9 log (cfu/g)were observed and predicted, respectively, for L. monocytogenes(i.e. 1.6 times too high) in shrimp salad where growth of thepathogen was both observed and predicted (n ¼ 9), primarilybeing products with pH values higher than 6.0 (n ¼ 8) (Fig. 5, Exps.M10 and M15 as an example). Average MPD values of 5.9 and 8.1log (cfu/g) were predicted for L. monocytogenes when a g value of1.0 (i.e. the classical Jameson effect model) and no microbialinteraction was applied (g ¼ 0), respectively (i.e. 6 times too lowand 25 times too high, respectively). For the 10 fail-dangerouspredictions, the introduction of a more flexible means of pre-dicting microbial interaction obviously had no impact on thepredicted MPD values of L. monocytogenes, still being identical tothe initial cell concentrations of the pathogen. For the 8 experi-ments with brined shrimp at 5.1 �C, g values between 0 and 0.99were estimated (Table 1, Exps. B4eB11), again ranging from nomicrobial interaction to microbial interaction as described by theclassical Jameson effect (g ¼ 1.0). Using the estimated g values anda g value of 1.0, respectively, average MPD values of 5.6 and 5.0 log(cfu/g) were predicted for L. monocytogenes, both being acceptablewhen compared to the observed value of 5.3 log (cfu/g) (Result notshown). Mejlholm et al. (2012) has previously reported that theclassical Jameson effect might be too restrictive for modellingmicrobial interaction between L. monocytogenes and LAB in brinedshrimp. On data from that study, MPD values of L. monocytogenesin brined shrimp were over- and underestimated by 0.6 and 0.5log (cfu/g) when estimated g values and the classical Jamesoneffect was used, respectively. Thus, based on our results the clas-sical Jameson effect seem to be too restrictive for brined shrimp,whereas the use of estimated g values and the expanded Jamesoneffect model is too conservative. In the future, more studies areneeded to understand the quantitative effect of microbial inter-action including the significance of different food matrices andtheir microbial substrates.

0 5 10 15 20 25 30 35

Storage (days) at 7.5°C

0

1

2

3

4

5

6

7

8

9

Log

(cfu

/g)

(b)

0 5 10 15 20 25

Storage (days) at 11.2°C

0

1

2

3

4

5

6

7

8

9

Log

(cfu

/g)

(a)

Fig. 5. Observed and predicted growth of Listeria monocytogenes (▫) and Lactobacillussakei (B) during storage of mayonnaise-based shrimp salad including acetic, benzoic,citric and sorbic acids (Exp. M10, Table 1) (a) or acetic and lactic acids (Exp. M15,Table 1) (b). Growth of L. monocytogenes (solid line) and L. sakei (dashed line) waspredicted by the new combined LmeLAB model of the present study. Microbialinteraction between L. monocytogenes and L. sakei was also predicted using theexpanded Jameson effect model with estimated g values of 0.72 (a) and 0.71 (b),respectively (dotted line).


Other studies have examined the growth potential ofL. monocytogenes in different types of mayonnaise-based salads(Burnett et al., 2005; Hwang and Tamplin, 2005; Skalina andNikolajeva, 2010; Uyttendaele et al., 2009; Vermeulen et al.,2007b), however, to our knowledge the simultaneous growth ofthe pathogen and LAB has not previously been described for thesetypes of products. Andrighetto et al. (2009) studied the simulta-neous growth of L. monocytogenes and LAB in Italian marinatedseafood salad (i.e. marinated seafood mixed with sunflower seedoil). However, the importance of microbial interaction was notestablished for this particular product as growth ofL. monocytogenes was prevented by the relatively low pH of 5.0,together with the storage temperature of 4 �C. For mayonnaise-based salads at chill temperatures, no growth of L. monocytogenesis typically reported for products with pH values below 5.0 (Alaliet al., 2012; Burnett et al., 2005; Uyttendaele et al., 2009). Skalinaand Nikolajeva (2010) showed that L. monocytogenes was able togrow in smoked ham salad with pH in the interval from 5.07 to

5.09, even at 3 �C. In the present study, growth of L. monocytogeneswas observed in shrimp salad with pH as low as 5.2 when stored at7.4 �C (Table 1, Exp. M2). Vermeulen et al. (2007b) described thetypical product characteristics of mayonnaise-based salads aspH < 5.6, aw < 0.988 and >0.1% acetic acid, in addition to thefrequent use of benzoic and sorbic acids as preservatives. It wasestimated that 0.030% and 0.060% of benzoic and sorbic acids in aratio of 1:1 would be sufficient to prevent growth ofL. monocytogenes in mayonnaise-based salads (pH 5.6, aw ¼ 0.985)at 4 and 7 �C, respectively (Vermeulen et al., 2007b). However, inthe present study, substantial growth of L. monocytogenes wasobserved in shrimp salad with even higher concentrations ofbenzoic and sorbic acids, and otherwise comparable productcharacteristics and storage conditions (Table 1, Exp. M6). Thus, boththe model of Vermeulen et al. (2007b) and the new combinedLmeLAB model of the present study seem to be too restrictivewhen predicting growth of L. monocytogenes in mayonnaise-basedsalads.

4. Conclusions

The new combined LmeLAB model accurately predicted themaximum population density (MPD) value for L. monocytogenes indifferent types of processed seafood including brined shell-fish aswell as cold-smoked and marinated fish products. Predictions wereaccurate for products with different combinations of organic acidsincluding (i) acetic acid/diacetate and lactic acid or (ii) benzoic,citric and/or sorbic acids. This extensive model seems valuable for(i) risk assessments where concentrations of L. monocytogenes inchilled seafood products during distribution can be markedlyinfluenced by growth of LAB, and (ii) risk management includingevaluation of compliance of processed seafood with the EU regu-lation for ready-to-eat foods. For non-complying products, the newcombined LmeLAB model can be used to identify combinations ofenvironmental parameters or the maximum shelf-life that wouldprevent L. monocytogenes from exceeding the critical limit of 2.0 log(cfu/g). Without the effect of microbial interaction, the predictedrisk associated with L. monocytogenes would in many cases bemarkedly overestimated, and the changes needed to preventgrowth of L. monocytogenes tomore than 2.0 log (cfu/g) would oftenbe too restrictive in form of excessive use of preservatives and/or anunnecessary shortening of the shelf-life. Importantly, it wasconcluded that to accurately predict MPD values ofL. monocytogenes the initial cell concentration of Lactobacillus spp.should be known. Using the “total” initial cell concentration of LABas input for the new combined LmeLAB model, the onset of mi-crobial interaction was typically predicted earlier than observedresulting in estimation of too lowMPD values for L. monocytogenes.

In mayonnaise-based shrimp salad with pH of 6.0 or higher,growth of L. monocytogenes continued at a reduced growth rateafter L. sakei reached their MPD. This growth pattern was success-fully described by an expanded version of the classical Jamesoneffect model. However, the performance of the new combinedLmeLAB model was unacceptable for shrimp salad with pH valueslower than 6.0. This was mainly due to prediction of no-growthwhen growth was actually observed (i.e. fail-dangerous pre-dictions). The new combined LmeLABmodel should not be used forshrimp salad with pH values below 6.0 as it might result inmisleading predictions for L. monocytogenes, potentially beingdangerous. In the future, it seems relevant to obtain a better un-derstanding of growth by microorganisms in mayonnaise-basedsalads for improved management of microbial quality and safety.Clearly, this can be facilitated by development of accurate predic-tive models for these products.


Acknowledgements

The present study was supported by the Technical University ofDenmark through the FoodDTU programme and by the DanishFood Industry Agency (3405-11-0305) as part of a project carriedout in collaboration between DTU Food and the seafood processingindustry in Denmark.

References

Alali, W.Q., Mann, D.A., Beuchat, L.R., 2012. Viability of Salmonella and Listeriamonocytogenes in delicatessen salads and hummus as affected by sodiumcontent and storage temperature. J. Food Prot. 75 (6), 1043e1056.

Andrighetto, C., Lombardi, A., Ferrati, M., Guidi, A., Corrain, C., Arcangeli, G., 2009.Lactic acid bacteria biodiversity in Italian marinated seafood salad and theirinteractions on the growth of Listeria monocytogenes. Food Control 20 (5),462e468.

Besse, N.G., Audinet, N., Barre, L., Cauquil, A., Cornu, M., Colin, P., 2006. Effect of theinoculum size on Listeria monocytogenes growth in structured media. Int. J. FoodMicrobiol. 110 (1), 43e51.

Bj€orkroth, J., 2005. Microbiological ecology of marinated meat products. Meat Sci.70 (3), 477e480.

Bligh, E.G., Dyer, W.J., 1959. A rapid method of total lipid extraction and purification.Can. J. Biochem. Physiol. 37 (8), 911e917.

Borch, E., Kant-Muermans, M.-L., Blixt, Y., 1996. Bacterial spoilage of meat and curedmeat products. Int. J. Food Microbiol. 33 (1), 103e120.

Brocklehurst, T.F., Wilson, P.D.G., 2000. The role of lipids in controlling microbialgrowth. Grasas Aceites 51 (1e2), 66e73.

Burnett, S.L., Mertz, E.L., Bennie, B., Ford, T., Starobin, A., 2005. Growth or survival ofListeria monocytogenes in ready-to-eat meat products and combination delisalads during refrigerated storage. J. Food Sci. 70 (6), 301e304.

Cheng, H., Friis, A., Leth, T., 2010. Partition of selected food preservatives in fishoilewater systems. Food Chem. 122 (1), 60e64.

Coillie, E. v., Werbrouck, H., Heyndrickx, M., Herman, L., Rijpens, N., 2004. Preva-lence and typing of Listeria monocytogenes in ready-to-eat food products on theBelgian market. J. Food Prot. 67 (11), 2480e2487.

Cornu, M., Billoir, E., Bergis, H., Beaufort, A., Zuliani, V., 2011. Modeling microbialcompetition in food: application to the behavior of Listeria monocytogenes andlactic acid flora in pork meat products. Food Microbiol. 28 (4), 639e647.

Dalgaard, P., Gram, L., Huss, H.H., 1993. Spoilage and shelf life of cod fillets packed invacuum or modified atmosphere. Int. J. Food Microbiol. 19 (4), 283e294.

Dalgaard, P., Jørgensen, L.V., 2000. Cooked and brined shrimps packed in a modifiedatmosphere have a shelf-life of >7 months at 0 �C, but spoil in 4e6 days at25 �C. Int. J. Food Sci. Technol. 35 (4), 431e442.

Dalgaard, P., 2009. Modelling of microbial growth. Bull. Int. Dairy Fed. 433,45e60.

Devlieghere, F., Debevere, J., 2000. Influence of dissolved carbon dioxide on thegrowth of spoilage bacteria. Lebensm. Wissensch. Technol. 33 (8), 531e537.

Di Pinto, A., Novello, L., Montemurro, F., Bonerba, E., Tantillo, G., 2010. Occurrence ofListeria monocytogenes in ready-to-eat foods from supermarkets in SouthernItaly. Microbiol. New. Microbiol. 33 (3), 249e252.

DVFA e Danish Veterinary and Food Administration, 2013. This is How You Evaluateif L. monocytogenes Can Grow in Ready-to-eat Lightly Preserved Seafood byUsing Computer Software. Available at: http://www.foedevarestyrelsen.dk/SiteCollectionDocuments/25_PDF_word_filer%20til%20download/04kontor/Mikro%20zoonose/Listeria/Listeria%20fisk%20end.pdf (in Danish, accessed22.07.13).

EC, 2005. Commission Regulation (EC) No 2073/2005 of 15 November 2005 onmicrobiological criteria for foodstuffs. Off. J. Eur. Commun. 338, 1e26.

FAO/WHO, 2009. Codex/ALINORM 09/32/13 Proposed Draft Microbiological Criteriafor Listeria monocytogenes in Ready-to-eat Foods. FAO/WHO, Rome.

Gim�enez, B., Dalgaard, P., 2004. Modelling and predicting the simultaneous growthof Listeria monocytogenes and spoilage micro-organisms in cold-smokedsalmon. J. Appl. Microbiol. 96 (1), 96e109.

Gombas, D.E., Chen, Y.H., Clavero, R.S., Scott, V.N., 2003. Survey of Listeria mono-cytogenes in ready-to-eat foods. J. Food Prot. 66 (4), 559e569.

Gysemans, K.P.M., Bernaerts, K., Vermeulen, A., Geeraerd, A.H., Debevere, J.,Devlieghere, F., Van Impe, J.F., 2007. Exploring the performance of logisticregression model types on growth/no growth data of Listeria monocytogenes.Int. J. Food Microbiol. 114 (3), 316e331.

Hartemink, R., Georgsson, F., 1991. Incidence of Listeria species in seafood andseafood salads. Int. J. Food Microbiol. 12 (2e3), 189e196.

Hwang, C.A., Tamplin, M.L., 2005. The influence of mayonnaise pH and storagetemperature on the growth of Listeria monocytogenes in seafood salad. Int. J.Food Microbiol. 102 (3), 277e285.

IFST, 1993. Code for the production of microbiologically safe and stable emulsifiedand non-emulsified sauces containing acetic acid (The CIMSCEE Code). In:Shelf-life of Foods e Guidelines for its Determination and Prediction. Instituteof Food Science & Technology, London, UK, pp. 49e51.

Jørgensen, L.V., Huss, H.H., 1998. Prevalence and growth of Listeria monocytogenes innaturally contaminated seafood. Int. J. Food Microbiol. 42, 127e131.

Leisner, J.J., Laursen, B.G., Prevost, H., Drider, D., Dalgaard, P., 2007. Carnobacterium:positive and negative effects in the environment and in foods. FEMS Microbiol.Rev. 31 (5), 592e613.

Levine, P., Rose, B., Green, S., Ransom, G., Hill, W., 2001. Pathogen testing of ready-to-eat meat and poultry products collected at federally inspected establish-ments in the United States, 1990 to 1999. J. Food Prot. 64 (8), 1188e1193.

Manios, S.G., Lambert, R.J.W., Skandamis, P.N., 2014. A generic model for spoilage ofacidic emulsified foods: combining physico-chemical data, diversity and levelsof specific spoilage organisms. Int. J. Food Microbiol. 170 (1), 1e11.

Manios, S.G., Skiadaresis, A.G., Karavasilis, K., Drosinos, E.H., Skandamis, P.N., 2009.Field validation of predictive models for the growth of lactic acid bacteria inacidic cheese-based Greek appetizers. J. Food Prot. 72 (1), 101e110.

McMeekin, T.A., Baranyi, J., Bowman, J., Dalgaard, P., Kirk, M., Ross, T., Schmid, S.,Zwietering, M.H., 2006. Information systems in food safety management. Int. J.Food Microbiol. 112 (3), 181e194.

McMeekin, T.A., Olley, J., Ratkowsky, D.A., Ross, T., 2002. Predictive microbiology:towards the interface and beyond. Int. J. Food Microbiol. 73 (2e3), 395e407.

Mejlholm, O., Dalgaard, P., 2009. Development and validation of an extensivegrowth and growth boundary model for Listeria monocytogenes in lightly pre-served and ready-to-eat shrimp. J. Food Prot. 72 (10), 2132e2143.

Mejlholm, O., Dalgaard, P., 2007b. Modeling and predicting the growth boundary ofListeria monocytogenes in lightly preserved seafood. J. Food Prot. 70 (1), 70e84.

Mejlholm, O., Dalgaard, P., 2007a. Modeling and predicting the growth of lactic acidbacteria in lightly preserved seafood and their inhibiting effect on Listeriamonocytogenes. J. Food Prot. 70 (11), 2485e2497.

Mejlholm, O., Gunvig, A., Borggaard, C., Blom-Hanssen, J., Mellefont, L., Ross, T.,Leroi, F., Else, T., Visser, D., Dalgaard, P., 2010. Predicting growth rates andgrowth boundary of Listeria monocytogenes e an international validation studywith focus on processed and ready-to-eat meat and seafood. Int. J. FoodMicrobiol. 141 (3), 137e150.

Mejlholm, O., Kjeldgaard, J., Modberg, A., Vest, M.B., Bøknæs, N., Koort, J.,Bj€orkroth, J., Dalgaard, P., 2008. Microbial changes and growth of Listeriamonocytogenes during chilled storage of brined shrimp (Pandalus borealis). Int. J.Food Microbiol. 128, 250e259.

Mejlholm, O., Bøknæs, N., Dalgaard, P., 2014. Development and evaluation of astochastic model for potential growth of Listeria monocytogenes in naturallycontaminated lightly preserved seafood. Food Microbiol. (accepted).

Mejlholm, O., Devitt, T.D., Dalgaard, P., 2012. Effect of brine marination on survivaland growth of spoilage and pathogenic bacteria during processing and subse-quent storage of ready-to-eat shrimp (Pandalus borealis). Int. J. Food Microbiol.157 (1), 16e27.

Mejlholm, O., Dalgaard, P., 2013. Development and validation of an extensivegrowth and growth boundary model for psychrotolerant Lactobacillus spp. inseafood and meat products. Int. J. Food Microbiol. 167 (2), 244e260.

Møller, C.O.A., Ilg, Y., Aabo, S., Christensen, B.B., Dalgaard, P., Hansen, T.B., 2013.Effect of natural microbiota on growth of Salmonella spp. in fresh pork e apredictive microbiology approach. Food Microbiol. 34 (2), 284e295.

Nieminen, T.T., V€alitalo, H., S€ade, E., Paloranta, A., Koskinen, K., Bj€orkroth, J., 2012.The effect of marination on lactic acid bacteria communities in raw broiler filletstrips. Front. Microbiol. 3, 1e8.

NMKL no.184, 2006. Aerobic Count and Specific Spoilage Microorganisms in Fishand Fish Products. Nordic Committee on Food Analysis, pp. 1e6.

Oscar, T.P., 2005. Validation of lag time and growth rate models for SalmonellaTyphimurium: acceptable prediction zone method. J. Food Sci. 70 (2),M129eM137.

Pothakos, V., Snauwaert, C., De Vos, P., Huys, G., Devlieghere, F., 2014. Psychro-trophic members of Leuconostoc gasicomitatum, Leuconostoc gelidum and Lac-tococcus piscium dominate at the end of shelf-life in packaged and chilled-stored food products in Belgium. Food Microbiol. 39, 61e67.

Ross, T., 1996. Indices for performance evaluation of predictive models in foodmicrobiology. J. Appl. Bacteriol. 81 (5), 501e508.

Ross, T., 1999. Predictive Food Microbiology Models in the Meat Industry. Meat andLivestock Australia, North Sydney, Australia, p. 196.

Rosso, L., Bajard, S., Flandrois, J.P., Lahellec, C., Fournaud, J., Veit, P., 1996. Differentialgrowth of Listeria monocytogenes at 4 and 8 �C: consequences for the shelf lifeof chilled products. J. Food Prot. 59 (9), 944e949.

Skalina, L., Nikolajeva, V., 2010. Growth potential of Listeria monocytogenes strainsin mixed ready-to-eat salads. Int. J. Food Microbiol. 144 (2), 317e321.

Smittle, R.B., Flowers, R.S., 1982. Acid tolerant microorganisms involved in thespoilage of salad dressings. J. Food Prot. 45 (10), 977e983.

Uyttendaele, M., Busschaert, P., Valero, A., Geeraerd, A.H., Vermeulen, A.,Jacxsens, L., Goh, K.K., De Loy, A., Van Impe, J.F., Deviieghere, F., 2009. Prevalenceand challenge tests of Listeria monocytogenes in Belgian produced and retailedmayonnaise-based deli-salads, cooked meat products and smoked fish between2005 and 2007. Int. J. Food Microbiol. 133 (1e2), 94e104.

Velugoti, P.R., Bohra, L.K., Juneja, V.K., Huang, L., Wesseling, A.L., Subbiah, J.,Thippareddi, H., 2011. Dynamic model for predicting growth of Salmonella spp.in ground sterile pork. Food Microbiol. 28 (4), 796e803.

Vermeulen, A., Devlieghere, F., Bernaerts, K., van Impe, J., Debevere, J., 2007a.Growth/no growth models describing the influence of pH, lactic and acetic acidon lactic acid bacteria developed to determine the stability of acidified sauces.Int. J. Food Microbiol. 119 (3), 258e269.

Vermeulen, A., Devlieghere, F., De Loy-Hendrickx, A., Uyttendaele, M., 2011. Criticalevaluation of the EU-technical guidance on shelf-life studies for

http://refhub.elsevier.com/S0740-0020(14)00170-1/sref1
































































http://www.foedevarestyrelsen.dk/SiteCollectionDocuments/25_PDF_word_filer%20til%20download/04kontor/Mikro%20zoonose/Listeria/Listeria%20fisk%20end.pdf


































































































































































L. monocytogenes on RTE-foods: a case study for smoked salmon. Int. J. FoodMicrobiol. 145 (1), 176e185.

Vermeulen, A., Smigic, N., Rajkovic, A., Gysemans, K., Bernaerts, K., Geeraerd, A., vanImpe, J., Debevere, J., Devlieghere, F., 2007b. Performance of a growth-nogrowth model for Listeria monocytogenes developed for mayonnaise-based

salads: influence of strain variability, food matrix, inoculation level, and pres-ence of sorbic and benzoic acid. J. Food Prot. 70 (9), 2118e2126.

Wilson, P.D.G., Wilson, D.R., Waspe, C.R., 2000. Weak acids: dissociation in com-plex buffering systems and partitioning into oils. J. Sci. Food Agric. 80 (4),471e476.












