The impact of temperature on the inactivation of enteric viruses in food and water: a review

Authors


Christophe Gantzer, Laboratoire de Chimie Physique et Microbiologie pour l’Environnement (LCPME), Faculté de Pharmacie, Université de Lorraine, CNRS, 5 rue Albert Lebrun, BP 80403, F-54001 Nancy Cedex, France. E-mail: christophe.gantzer@univ.lorraine.fr

Summary

Temperature is considered as the major factor determining virus inactivation in the environment. Food industries, therefore, widely apply temperature as virus inactivating parameter. This review encompasses an overview of viral inactivation and virus genome degradation data from published literature as well as a statistical analysis and the development of empirical formulae to predict virus inactivation. A total of 658 data (time to obtain a first log10 reduction) were collected from 76 published studies with 563 data on virus infectivity and 95 data on genome degradation. Linear model fitting was applied to analyse the effects of temperature, virus species, detection method (cell culture or molecular methods), matrix (simple or complex) and temperature category (<50 and ≥50°C). As expected, virus inactivation was found to be faster at temperatures ≥50°C than at temperatures <50°C, but there was also a significant temperature–matrix effect. Virus inactivation appeared to occur faster in complex than in simple matrices. In general, bacteriophages PRD1 and PhiX174 appeared to be highly persistent whatever the matrix or the temperature, which makes them useful indicators for virus inactivation studies. The virus genome was shown to be more resistant than infectious virus. Simple empirical formulas were developed that can be used to predict virus inactivation and genome degradation for untested temperatures, time points or even virus strains.

Introduction

There is growing concern over human exposure to enteric viruses through contaminated water or food products (Koopmans and Duizer 2004). Data on viral waterborne and foodborne diseases are still fragmented (WHO/FAO 2008), focusing either on particular countries or on particular pathogens. Thus, ‘the foodborne viruses in Europe’ (FBVE) network was initiated in 1999 to promote the exchange of data on outbreaks of gastroenteritis because of noroviruses (NoV) (Kroneman et al. 2007). Epidemiological evidence indicates that NoV is the major cause of foodborne outbreaks of gastroenteritis worldwide. To a lesser extent, rotavirus is also implicated in foodborne illness and constitutes an important cause of death in young children in low-income countries (WHO/FAO 2008; EFSA 2011). Hepatitis A virus (HAV) is less common in countries with a high standard of hygiene where it may cause infection later in life, with the risk of leading to a more severe disease outcome (WHO/FAO 2008). While consumption of ready-to-eat foods contaminated by infected food handlers remains an important risk factor for enteric virus outbreaks (Barrabeig et al. 2010), shellfish, particularly oysters, are commonly considered as the most frequently associated food vehicles in NoV outbreaks (EFSA 2011). Other foods, especially raw materials like soft fruits and vegetables, are also being recognized as relevant food vehicles of enteric viruses (Maunula et al. 2009; Ethelberg et al. 2010; EFSA 2011). RT-PCR shows relatively high positivity rates of NoV and/or HAV genome in bivalve molluscs, with positivity prevalence ranging from 4·8 to 53·2% (Croci et al. 2007; David et al. 2007; Phan et al. 2007; Le Guyader et al. 2008; Nakagawa-Okamoto et al. 2009; Mesquita et al. 2011). Contamination rates of NoV genome ranging from 6·6 to 34·5% on soft red fruits and from 28·2 to 50% on leafy greens have been observed in Belgium, Canada and France (Baert et al. 2011).

Water can also be a source of disease outbreaks either directly via recreational and drinking water (Maunula et al. 2005; Sinclair et al. 2009) or indirectly via irrigation or contamination of shellfish growing areas (Koopmans and Duizer 2004; Westrell et al. 2010). Viral outbreaks linked to recreational or drinking water are often underestimated, because of low virus concentrations in water together with poor detection efficiency (Maunula et al. 2005; Sinclair et al. 2009). Human pathogenic viruses, such as enterovirus, adenovirus, NoV, reovirus, rotavirus and HAV, have been detected in groundwater with molecular and/or cell culture techniques with prevalence rates varying from 8 to 23% (Borchardt et al. 2003, 2007; Fout et al. 2003). Consumption of drinking water sourced from groundwater contaminated with human pathogenic viruses may lead to epidemics that cause severe illness and even death (Parshionikar et al. 2003; Kim et al. 2005; Gallay et al. 2006; Jean et al. 2006). When groundwater was involved, it often concerned vulnerable geologic settings as fractured rock aquifers, cross-connecting well bores, or leaking well-cases in sandstone and shallow aquifers (Powell et al. 2003; Borchardt et al. 2007) in combination with the presence of significant sources of contamination, such as wastewater treatment facilities, septic tanks and animal manure (Parshionikar et al. 2003; Gallay et al. 2006; Jean et al. 2006; Fong et al. 2007).

When a viral inactivation process is available, even for some food products (Deboosere et al. 2004; Butot et al. 2009), the assessment of treatment efficiency remains difficult. In the absence of cell culture systems for some enteric viruses, different virus types have been evaluated as surrogates (Bae and Schwab 2008; Baert et al. 2008b). Even when a culture system is available, laboratory strains might not reflect the resistance of naturally occurring strains. Viral inactivation is also highly variable between virus types, type of treatment or type of matrix. In addition, viral inactivation is usually studied for a limited number of viruses, matrices and inactivating parameters. Therefore, it is not easy to determine the most resistant virus for a particular treatment in a particular matrix, and so there is no single treatment regimen which is applicable for every virus in every matrix. Other concerns include the methodology of viral detection used to estimate treatment efficiency and the data analysis (Teunis et al. 2009). The detection of infectious virus is based on cell culture but no susceptible cell lines has yet been identified for noroviruses (Vashist et al. 2009), and cell culture options are limited for strains of HAV or rotaviruses. Molecular tools may detect all viral types but it is now generally accepted that they may largely underestimate treatment efficiency (Gassilloud et al. 2003; Hewitt and Greening 2006). Thus, a recommended practice is to use a cultivable virus indicator to represent uncultivable pathogenic viruses (Gassilloud et al. 2003; Baert et al. 2008b; Butot et al. 2009).

The temperature is largely known as the major factor determining virus inactivation in the environment and also is widely applied in food industries. The objective of this study was to perform a statistical analysis of literature data on virus inactivation to develop empirical formulae predicting inactivation of specific viruses in specific matrices as a function of temperature. The data were collected by the Working Group 4 ‘Viral Inactivation’ of COST Action 929 ‘A European Network for Food and Environmental Virology’ in a spreadsheet (available at the following website http://www.cost929.pcrlab.net/login.php).

Data Collection and Statistical Analyses

Selection of the reviewed literature

Peer-reviewed papers on viral inactivation and genome degradation mainly in food and water at different temperatures were collected. Data obtained with biological samples such as blood were not included in the present work. The major information of interest coming from these published studies was the time needed (in days) to observe the first log10 reduction [time to first log (TFL) value] either for infectious virus by using cell culture (CC) or for viral genome by using molecular methods (PCR). This work is not an exhaustive review; this is notably because of the criteria of inclusion in the spreadsheet described below. In this review, a decrease (linear or nonlinear) in viral infectivity or genome of more than 0·5 log10 below the initial viral titre during the time of experiment was considered the threshold for inclusion in the analyses. If a linear decrease was shown over the whole period of analysis by regression analysis (e.g. squared correlation given) or suggested by the authors, the TFL value could be estimated in the following ways: (i) For a decrease higher than 1 log10, the slope of the regression analysis (decay rate) or a graphical estimation was used. (ii) For a decrease between 0·5 and 1 log10, an extrapolation was allowed if the decrease was observed over at least 10 days by using at least five experimental points. (iii) The TFL value could also be estimated from the T90 or T99 value, if available.

In case of a nonlinear decrease over the whole period of analysis, the TFL value could be estimated graphically or by using the slope value (decay rate) if either (i) a linear line could be drawn for the first log reduction over at least three experimental points or (ii) two experimental points were given with the second point at least one log higher than the detection limit.

Data classification

A spreadsheet was created for the collection and the analysis of the selected studies. For each study, the spreadsheet (available at: http://www.cost929.pcrlab.net/login.php) was filled with 13 different parameters organized in columns. Among them, the taxonomy of the viruses (family, genus and species), the detection method (either cell culture [CC] for inactivation of infectious virus or molecular methods [PCR] for virus genome degradation), the type of matrix and the experimental temperature were used for the classification of the data. Ten types of matrices were identified from the published studies and were considered as ‘simple’ or ‘complex’ matrices. Three types of matrices were considered as simple: (i) synthetic media (synthetic and sterile media without suspended matter, e.g., phosphate-buffered saline, cell culture medium, artificial seawater, artificial groundwater), (ii) drinking water (dechlorinated tap water, bottled water, filtered water, sterilized water, distilled or deionized water), and (iii) groundwater (filtered or non-filtered groundwater or well water). The seven other types of matrices did not correspond to the above criteria and were considered as complex: (i) freshwater (water from river, stream, lake, bog, pond and well water if the authors reported high level of suspended matter), (ii) natural seawater (seawater and estuarine water), (iii) sewage (primary and secondary sewage), (iv) soil, (v) dairy products (e.g. milk, skim milk, cream, reconstituted dry skim milk…), (vi) food (e.g. vegetables, fruits, mussels, meat…) and (vii) urine (reuse in agricultural purpose). A distinction was made between matrices as an influence of the matrix composition has been previously suggested, especially for food (Bidawid et al. 2000; Deboosere et al. 2004). Other parameters in the spreadsheet were temperature, TFL and log10TFL values. The other data reported in the spreadsheet were R2, the use of graphical estimation, log decrease, number of days corresponding to the log decrease, study period, initial viral titre, literature reference and remarks if needed (such as ‘linear regression analysis’, ‘TFL value estimated from the data’…). The studies included in the spreadsheet and in the figures of the present work are labelled in the References with a (*) and/or a (‡) according to their use for the statistical analysis of viral inactivation and/or genome degradation, respectively.

Statistical analyses

A linear relationship between log10TFL and temperature was apparent from plotting the data, as shown in Figs 1 and 2; therefore, the following relationship was analysed:

image(1)
Figure 1.

 Values of log10 time to first log (n = 563) as a function of temperature, categorized according to detection by cell culture (CC), in simple (S) or complex (C) matrices and temperatures <50°C or ≥50°C. The values shown in this figure were obtained from studies marked by the * after the year of publication in the References.

Figure 2.

 Values of log10 time to first log (n = 95) as a function of temperature, categorized according to detection by molecular methods (PCR), in simple (S) or complex (C) matrices and temperatures <50°C or ≥50°C. The values shown in this figure were obtained from studies marked by the ‡ after the year of publication in the References.

with Y-intercept α0, slope α1 [°C−1], and temperature T [°C]. If α1 = 0, then α0 = log 10TFL, which means that, in this case, virus inactivation is not affected by temperature. A higher value of α0 implies a longer time to the first log10 reduction in virus concentration, hence a more stable virus. It should be noted that α1 ≤ 0. A stronger negative value of the slope (α1) implies a greater sensitivity to temperature leading to a faster viral inactivation at higher temperature.

The governing equation for first-order rate virus inactivation is as follows:

image(2)

where Ct is the virus concentration [number of viral units per volume or mass] at time t, C0 is the initial virus concentration and μ is the virus inactivation rate coefficient [per time unit].

Equation (2) can be rearranged as follows:

image(3)

From Eqn (3), it is possible to determine the virus inactivation rate coefficient:

image(4)

Equation (3), Eqns (4) and (1) can be combined in the following way:

image(5)
Figure 3.

 Distribution of the data on viral inactivation (n = 563) in function of virus species.

Figure 4.

 Distribution of the data on viral inactivation (n = 563) in function of the type of matrix.

Figures 1 and 2 show that the values of log10TFL, as a function of temperature, can be categorized according to detection method (CC and PCR), matrix (S and C) and temperature < or ≥50°C (low and high temperature). Each category of log 10TFL values may be characterized by their distinct values of α0 and α1. To that aim, the effects of virus species, detection, matrix and temperature categories were analysed by means of linear model fitting using the statistical package R (ver. 2.12.2; Bell Laboratories, Lucent Technologies, http://www.r-project.org). Effects of virus species, detection, matrix, temperature categories and their interactions may be significant or not. By means of stepwise model selection by Akaike’s Information Criterion (AIC) with k, the multiple of the number of degrees of freedom used for the penalty in AIC set to 3·84, the best model describing the data was selected. A k value of 3·84 corresponds to the Chi-square value with 95% confidence and one degree of freedom. The best model selected in R was transferred to Mathematica 8 (Wolfram Inc, Champaign, Illinois), where it can be easily implemented into a function for abstracting the linear equations for combinations of values of the categorical variables.

In addition, Eqn (1) was extended to calculate so-called mean prediction bands that encompass 95% of the mean predicted log10TFL values. The governing equation for calculating the 95% confidence interval of the mean log 10TFL is:

image(6)

The same way, single prediction bands can be derived that encompass 95% of single log 10TFL values. The latter equation should be used for predicting the 95% confidence interval for a single log 10TFL value. The governing equation for calculating the 95% confidence intervals of a single log 10TFL value is:

image(7)

In Eqns (6) and (7), βm, βs, β1 and β2 are linear combinations of the standard errors of the estimated model parameters, including covariances.

It should be noted that for a particular virus–matrix combination, Eqns (6) and (7) should only be applied in the relevant low (0–50°C) or high (50–100°C) temperature range. In other words, the equation for a particular virus–matrix combination that was found for the low temperature range cannot be used to calculate virus inactivation in the high temperature range and vice versa.

Results

A total of 563 data were collected on viral inactivation (CC) from 73 published studies. The data represent twenty viral species belonging to nine virus families (Fig. 3). Among these families, Picornaviridae (Hepatitis A virus, Poliovirus, Coxsackievirus, Echovirus) and Leviviridae (F-specific RNA phages) were the most represented and covered 34·3 and 22·2% of the data, respectively. In contrast, only 0·7% of the data consisted of Astroviridae (astrovirus). The decreasing classification of the six other viral families is: Tectiviridae (14·6%), Siphoviridae (10·3%), Caliciviridae (7·3%), Adenoviridae (4·3%), Microviridae (3·9%) and Reoviridae (2·5%). Figure 4 shows that among the viral inactivation data, 315 data (56%) resulted from experiments performed with complex matrices.

Ninety-five PCR data on viral genome degradation were collected from 13 publications. These data came from 11 viral species belonging to the Picornaviridae, the Caliciviridae and the Leviviridae families. The higher percentage of data (67·5%) was represented by the Caliciviridae, whereas this family represented only 7% of the data on inactivation. This is a result of the numerous studies performed on the norovirus genome. The Picornaviridae and Leviviridae represented 24 and 8·5% of the data, respectively. As for CC, 56% of the PCR data came from studies using complex matrices.

As shown in Figs 1 and 2, two sets of data were observed in function of the temperature applied to the samples. Data of the first set were obtained from experiments performed at temperatures ranging from 5 and 42°C and the second one corresponded to data obtained at temperatures ranging between 51 and 95°C. Virus inactivation data were analysed for low and high temperatures (< and ≥50°C), respectively. The range of temperatures <50°C concerns a wide range of environmental samples included in our work, especially water (e.g. bottled mineral water, surface water, groundwater, wastewater), whereas higher temperatures (≥50°C) were deliberately applied for virus inactivation in industrial processes (e.g. in food or dairy products). Thus, all the experiments with food or dairy products were conducted at temperatures ≥50°C. Irrespective of the type of matrix and the type of method, 69% (452 data) of the experiments on virus inactivation were performed at temperatures <50°C.

According to the best model selection in R, the variables and their interactions that best describe log10TFL values as a function of temperature are listed in Table 1. Note that the variable virus species did not significantly interact with any of the other variables except for low/high temperature. This implies that virus species are inactivated at different rates in the two temperature categories.

Table 1.   Best model description: significant variable and interactions
Dependent variableLog10 TFL
  1. TFL, time to first log.

  2. *<50°C/≥50°C, respectively.

Numerical variableTemperature
Categorical variablesSpecies
Detection
Matrix
Low/high temperature*
Significant variables and interactionsSpecies
Detection
Matrix
Temperature
Low/high temperature
Detection × Matrix
Detection × Temperature
Matrix × Temperature
Species × Low/high temperature
Detection × Low/high temperature
Matrix × Low/high temperature
Temperature × Low/high temperature
Detection × Matrix × Temperature
Detection × Matrix × Low/high temperature
Matrix × Temperature × Low/high temperature

By consecutively substituting values of the categorical variables in the general model, linear equations in the form of Eqn (1) were derived for each of the virus species. Tables 2–9 summarize all the values of α0 and α1 corresponding to these equations, and also the values of βs, βm, β1 and β2 for calculation of 95% confidence intervals. These tables also include predicted single log10TFL values, including 95% confidence intervals for 0 and 50°C for the low temperature category and for 50 and 100°C for the high temperature category. As examples, Fig. 5(a)–(d) show the data and the model predictions with mean and single 95% prediction bands for poliovirus, HAV, PRD1 phage and F-specific RNA phage genogroup I, respectively.

Table 2.   Parameters for calculation of virus inactivation and log10 TFL values including 95% prediction intervals (95% CI) calculated by using Eqn (7) for 0 and 50°C. Detection by cell culture (CC), inactivation at temperature <50°C (lo), in simple matrices (S)
CC-lo-S
α1 = −0·036, β2 = 0·000019log10 TFL (0°C)log10 TFL (50°C)Virus species
α0βsβmβ1NMean95% CIMean95% CI
  1. CaCV, canine calicivirus; FCV, feline calicivirus, FRNAPH, F-specific RNA phages; HAV, hepatitis A virus; MNV, murine norovirus; TFL, time to first log reduction.

0·870·380·093−0·0009920·87−0·342·1−0·91−2·10·3CaCV
1·40·310·021−0·00075121·40·312·5−0·37−1·50·73FCV
1·40·310·019−0·00060141·40·352·5−0·34−1·50·78Echovirus
1·40·600·310−0·0014011·4−0·0813·0−0·34−1·81·2Human rotavirus
1·60·320·034−0·0007241·60·482·7−0·18−1·30·95Simian rotavirus
1·70·340·051−0·0008341·70·542·8−0·092−1·21·1FRNAPH genogroup II
1·70·340·051−0·0008341·70·562·8−0·076−1·21·1FRNAPH genogroup III
1·70·330·046−0·0008251·70·582·8−0·065−1·21·1FRNAPH genogroup IV
1·80·310·020−0·00062101·80·702·90·0096−1·11·1Coxsackievirus
1·80·300·010−0·00068451·80·732·90·027−1·11·1FRNAPH GGI
1·90·360·077−0·0006431·90·73·10·11−1·11·3Human astrovirus
1·90·300·011−0·00073571·90·823·00·12−0·971·2Poliovirus
2·10·310·020−0·0006462·10·993·20·3−0·811·4Human adenovirus
2·30·370·083−0·0008422·31·13·50·52−0·681·7MNV
2·50·310·024−0·00071112·51·43·60·71−0·41·8HAV
2·50·300·013−0·0006762·51·43·60·72−0·381·8PRD1 phage
2·60·320·035−0·0006612·61·53·70·82−0·322·0PhiX174 phage
Table 3.   Parameters for calculation of virus inactivation and log10TFL values including 95% prediction intervals (95% CI) calculated by using Eqn (7) for 0 and 50°C. Detection by cell culture (CC), inactivation at temperature <50°C (lo), in complex matrices (C)
CC-lo-C
α1 = −0·030, β2 = 0·000014log10 TFL (0°C)log10 TFL (50°C)Virus species
α0βsβmβ1NMean95% CIMean95% CI
  1. FCV, feline calicivirus; FRNAPH, F-specific RNA phages; HAV, hepatitis A virus; MNV, murine norovirus; TFL, time to first log reduction.

0·780·580·290−0·0005710·78−0·712·3−0·73−2·20·77Bacteroides fragilis phage
0·910·310·020−0·0005550·91−0·182·0−0·61−1·70·49FCV
0·940·310·022−0·0005760·94−0·152·0−0·57−1·70·53Echovirus
1·10·320·032−0·0005771·1−0·00892·2−0·42−1·50·71Simian rotavirus
1·20·340·050−0·0006121·20·052·3−0·33−1·50·82FRNAPH genogroup II
1·20·340·050−0·0006121·20·0652·3−0·31−1·50·84FRNAPH genogroup III
1·20·330·045−0·0006021·20·0862·3−0·3−1·40·84FRNAPH genogroup IV
1·30·310·021−0·0005881·30·22·4−0·23−1·30·88Coxsackievirus
1·30·300·009−0·00054441·30·242·4−0·21−1·30·88FRNAPH genogroup I
1·30·320·033−0·00050101·30·232·5−0·18−1·30·95FRNAPH all genogroups
1·40·370·079−0·0005611·40·22·6−0·13−1·31·10Human astrovirus
1·40·300·010−0·00055311·40·332·5−0·12−1·20·97Poliovirus
1·60·300·015−0·00041181·60·52·70·065−1·01·20Human adenovirus
1·80·370·083−0·0006411·80·613·00·29−0·91·50MNV
2·00·310·025−0·0006162·00·93·10·48−0·631·60HAV
2·00·300·011−0·00066762·00·933·10·48−0·591·60PRD1 phage
2·10·320·030−0·00046102·11·03·20·59−0·541·70PhiX174 phage
Table 4.   Parameters for calculation of virus inactivation and log10TFL values including 95% prediction intervals (95% CI) calculated by using Eqn (7) for 50 and 100°C. Detection by cell culture (CC), inactivation at temperature ≥50°C (hi), in simple matrices (S)
CC-hi-S
α1 = −0·058, β2 = 0·000096log10 TFL (50°C)log10 TFL (100°C)Virus species
α0βsβmβ1NMean95% CIMean95% CI
  1. CaCV, canine calicivirus; FCV, feline calicivirus; FRNAPH, F-specific RNA phages; HAV, hepatitis A virus; MNV, murine norovirus; TFL, time to first log reduction.

0·160·770·48−0·013 2−2·7−3·9−1·5−5·6−6·9−4·3Poliovirus
0·230·810·53−0·013 2−2·6−3·9−1·4−5·5−6·9−4·2CaCV
0·410·710·42−0·012 6−2·5−3·6−1·3−5·3−6·6−4·1FCV
0·580·700·42−0·012 4−2·3−3·4−1·2−5·2−6·5−3·9MNV
0·630·780·49−0·013 3−2·2−3·5−1·0−5·1−6·5−3·8Lactobacillus paracasei phage
0·710·700·42−0·012 3−2·2−3·3−1·1−5·0−6·3−3·8HAV
0·710·780·49−0·013 3−2·2−3·4−0·95−5·0−6·4−3·7Lactobacillus casei phage
1·100·790·50−0·012 1−1·8−3·1−0·48−4·7−6·2−3·2Simian rotavirus
1·200·730·44−0·01318−1·7−2·8−0·56−4·6−5·8−3·3Lactococcus lactis phage
1·300·730·44−0·013 8−1·6−2·7−0·42−4·4−5·7−3·2Lactobacillus helveticus phage
1·400·820·54−0·013 1−1·5−2·8−0·23−4·4−5·7−3·0FRNAPH genogroup I
1·500·720·44−0·01310−1·4−2·5−0·2−4·3−5·6−3·0PhiX174 phage
Table 5.   Parameters for calculation of virus inactivation and log10TFL values including 95% prediction intervals (95% CI) calculated by using eqn (7) for 50 and 100°C. Detection by CC, inactivation at temperature ≥50°C (hi), in cell culture (CC)/complex matrices (C)
CC-hi-C
α1 = −0·012, β2 = 0·000047log10 TFL (50°C)log10 TFL (100°C)Virus species
α0βsβmβ1NMean95% CIMean95% CI
  1. FCV, feline calicivirus; FRNAPH, F-specific RNA phages; HAV, hepatitis A virus; MNV, murine norovirus; TFL, time to first log reduction.

−2·70·570·28−0·00663−3·3−4·5−2·1−3·9−5·1−2·7Poliovirus
−2·50·580·29−0·00724−3·1−4·2−1·9−3·7−4·8−2·5FCV
−2·30·560·27−0·00683−2·9−4·0−1·8−3·5−4·7−2·3MNV
−2·20·590·30−0·00651−2·8−4·1−1·6−3·5−4·7−2·2Lactobacillus paracasei phage
−2·20·560·27−0·007138−2·8−3·9−1·6−3·4−4·5−2·3HAV
−2·20·590·30−0·00651−2·8−4·0−1·5−3·4−4·6−2·1Lactobacillus casei phage
−1·80·590·30−0·00664−2·5−3·7−1·2−3·1−4·3−1·8Bacteroides fragilis phage
−1·80·670·39−0·00671−2·4−3·7−1·1−3·0−4·3−1·6Simian rotavirus
−1·80·630·34−0·00622−2·4−3·7−1·1−3·0−4·4−1·6Echovirus
−1·70·530·25−0·006611−2·3−3·4−1·2−2·9−4·0−1·7Lactococcus lactis phage
−1·70·600·31−0·00572−2·3−3·6−0·98−2·9−4·3−1·5Coxsackievirus
−1·50·540·25−0·00655−2·1−3·3−1·0−2·8−3·9−1·6Lactobacillus helveticus phage
−1·50·620·33−0·00662−2·1−3·3−0·84−2·7−4·0−1·4FRNAPH genogroup I
−1·40·550·27−0·00661−2·0−3·1−0·85−2·6−3·8−1·4PhiX174 phage
−0·950·490·20−0·00494−1·6−2·7−0·39−2·2−3·5−0·83FRNAPH all genogroups
−0·870·550·27−0·00573−1·5−2·7−0·26−2·1−3·4−0·76Somatic coliphage
Table 6.   Parameters for calculation of virus inactivation and log10TFL values including 95% prediction intervals (95% CI) calculated by using Eqn (7) for 0 and 50°C. Detection by PCR methods (PCR), inactivation at temperature <50°C (lo), in simple matrices (S)
PCR-lo-S
α1 = −0·001, β2 = 0·000092log10 TFL (0°C)log10 TFL (50°C)Virus species
α0βsβmβ1NMean95% CIMean95% CI
  1. CaCV, canine calicivirus; HuNV, human norovirus; FCV, feline calicivirus; HAV, hepatitis A virus; TFL, time to first log reduction.

0·940·450·170−0·005720·94−0·382·30·88−0·352·1CaCV
1·20·430·140−0·005731·2−0·0622·51·2−0·0372·4HuNV
1·50·370·083−0·004751·50·282·71·40·232·6FCV
1·90·370·083−0·004531·90·673·11·80·63·0Coxsackievirus
1·90·360·076−0·004521·90·693·11·80·633·0HAV
2·00·360·076−0·004552·00·783·11·90·723·1Poliovirus
Table 7.   Parameters for calculation of virus inactivation log10TFL values including 95% prediction intervals (95% CI) calculated by using Eqn (7) for 0 and 50°C. Detection by PCR methods (PCR), inactivation at temperature <50°C (lo), in complex matrices (C)
PCR-lo-C
α1 = −0·041, β2 = 0·000085log10 TFL (0°C)log10 TFL (50°C)Virus species
α0βsβmβ1NMean95% CIMean95% CI
  1. FCV, feline calicivirus; FRNAPH, F-specific RNA phages; MNV, murine norovirus; TFL, time to first log reduction; HuNV, human norovirus.

0·960·350·064−0·002240·96−0·22·1−1·1−2·40·22HuNV
1·20·340·057−0·002921·2−0·062·4−0·84−2·10·42FCV
1·50·380·088−0·003111·50·292·7−0·56−1·90·73FRNAPH genogroup II
1·50·380·088−0·003111·50·312·7−0·55−1·80·74FRNAPH genogroup III
1·50·370·084−0·003111·50·332·7−0·54−1·80·75FRNAPH genogroup IV
1·60·340·052−0·003031·60·472·8−0·44−1·70·81FRNAPH genogroup I
1·70·340·053−0·003021·70·562·8−0·35−1·60·9Poliovirus
2·10·400·120−0·003312·10·863·40·05−1·31·4MNV
Table 8.   Parameters for calculation of virus inactivation and log10TFL values including 95% prediction intervals (95% CI) calculated by using Eqn (7) for 50 and 100°C. Detection by PCR methods (PCR), inactivation at temperature ≥50°C (hi), in simple matrices (S)
PCR-hi-S
α1 = −0·023, β2 = 0·000074log10 TFL (50°C)log10 TFL (100°C)Virus species
α0βsβmβ1NMean95% CIMean95% CI
  1. CaCV, canine calicivirus; FCV, feline calicivirus; HAV, hepatitis A virus; HuNV, human norovirus; MNV, murine norovirus; TFL, time to first log reduction.

−0·700·820·54−0·012 1−1·9−3·1−0·59−3·0−4·2−1·8Poliovirus
−0·630·960·68−0·013 2−1·8−3·2−0·43−3·0−4·2−1·8CaCV
−0·450·800·52−0·012 2−1·6−2·8−0·39−2·8−3·9−1·6FCV
−0·280·760·47−0·01210−1·4−2·6−0·25−2·6−3·7−1·5MNV
−0·150·770·49−0·012 2−1·3−2·5−0·11−2·5−3·6−1·3HAV
−0·0360·760·47−0·012 5−1·2−2·40·0018−2·4−3·5−1·2HuNV
Table 9.   Parameters for calculation of virus inactivation and log10TFL values including 95% prediction intervals (95% CI) calculated by using Eqn (7) for 50 and 100°C. Detection by PCR methods (PCR), inactivation at temperature ≥50°C (hi), in complex matrices (C)
PCR-hi-C
α1 = −0·023, β2 = 0·000053log10 TFL (50°C)log10 TFL (100°C)Virus species
α0βsβmβ1NMean95% CIMean95% CI
  1. FCV, feline calicivirus; HAV, hepatitis A virus; HuNV, human norovirus; MNV, murine norovirus; TFL, time to first log reduction.

−1·00·660·37−0·00868−2·3−3·5−1·1−3·5−4·6−2·4FCV
−1·00·630·35−0·00834−2·2−3·3−0·98−3·3−4·4−2·2MNV
−0·870·630·34−0·00839−2·0−3·2−0·86−3·2−4·3−2·0HAV
−0·760·630·35−0·008416−1·9−3·1−0·75−3·1−4·2−1·9HuNV
Figure 5.

 Data on viral inactivation and model prediction with mean and single 95%-prediction bands for poliovirus (a), HAV (b), PRD1 phage (c) and F-specific RNA phage genogroup I (d) in simple (circle) and complex (square) matrices.

Within each of the eight possible combinations of the variables, detection method, temperature category and matrix, the values of α1 and β2 were the same for all virus species, implying that within each one of these eight combinations, the sensitivity of the viruses to temperature change seemed to be similar for all virus species. All other parameters (α0, βs, βm and β1) are not only dependent on detection method, temperature category and matrix, but also on virus species. In Tables 2–9, within each of the eight combinations of detection method, temperature category and matrix, virus species are ordered from low to high values of α0. Because inline image, virus species are ordered from high inactivation rate to low inactivation rate. Note that α0 values obtained at high temperatures were all positive in simple matrices (Table 4), whereas those in complex matrices were all negative (Table 5). This would imply that at 0°C, one-log10 inactivation would take <1 day in case of complex matrices and more than 1 day in the case of simple matrices. This is not realistic, but, as mentioned above, inactivation predictions should not be performed outside of the range of the temperature category. Thus, the linear relation observed for the high temperature category should not be used for temperature <50°C. On the same way, the linear relation observed for the low temperature category should not be applied for the high temperature category.

Data on virus inactivation (CC) studied for 15 virus species in the low temperature range resulted in the same order of virus species according to α0 values yielded in simple and complex matrices (Tables 2 and 3, respectively). The PhiX174 and PRD1 phages, and HAV showed the highest α0 values, whereas feline calicivirus (FCV), echovirus and rotavirus resulted in the lower α0. A factor 10 was observed between the farthest TFL values. For this low temperature range, α0 values could be determined for 15 virus species or genogroups in both simple and complex matrices. In the high temperature range, ten virus species were studied both in simple and complex matrices (Tables 4 and 5, respectively) that showed the same order in function of their α0 value. For temperatures above 50°C, the higher α0 values were observed for somatic and F-specific RNA bacteriophages whereas poliovirus and FCV show the lowest α0 values. Virus inactivation proceeds faster in the high temperature range than in the low temperature range, which is not only a temperature effect, but a temperature–matrix effect because temperature category and matrix significantly interact (Table 1). For each virus species, comparison of the log10TFL values between matrices showed consistently faster inactivation in complex than in simple matrices at a given temperature. The α1 values (slope values) showing the sensitivity of virus species to temperature change could also be used to analyse the results. For the low temperature range, virus inactivation appeared to have similar sensitivity to temperature change whatever the type of matrix. The α1 values reached −0·03 and −0·036 in simple and complex matrices, respectively. Note that in the high temperature range, α1 for simple matrices is −0·058 (Table 4) and for complex matrices it is −0·012 (Table 5), implying that virus inactivation is more sensitive to temperature change in simple matrices than in complex matrices.

Data on genome degradation (PCR) monitored by real-time PCR methods were collected for eight viral species (Tables 6–9). In the low temperature range, degradation of the viral genome appeared to be sensitive to temperature changes in complex matrices only, indeed the α1 values were 0·001 and −0·041 in simple and complex matrices, respectively (Tables 6 and 7). In contrast, genome degradation showed the same sensitivity to the increase of temperature between 50 and 100°C in simple and complex matrices (α1 = −0·023, Tables 8 and 9). In the high temperature range, genome degradation occurs faster in complex matrices as shown by the log10TFL values. The comparison of the α1 values obtained for genome and infectivity resulted in a similar observation whatever the temperature range: a lower sensitivity to temperature increase for genome degradation in comparison with inactivation in simple matrices (Tables 2, 4, 6 and 8) and the opposite in complex matrices (Tables 3, 5, 7 and 9). Nevertheless, for temperature below 50°C, PCR showed similar or higher log10TFL values than cell culture in simple and complex matrices at 0 and 50°C. Moreover, at high temperature range, log10TFL values are systematically higher for PCR than cell culture at 50 and 100°C. Generally, virus genomes may be considered to appear more persistent as compared with infectious viruses in both simple and complex matrices.

Discussion – Conclusion

Selected literature data concerning virus inactivation by temperature were collected on a spreadsheet. The selected data were organized in function of the temperature range, the type of matrix and the detection method (RT-PCR or cell culture). The spreadsheet was filled with 563 data on infectivity decrease and 95 data on genome degradation.

The statistical analyses resulted in regression equations (α0 values and log10TFL values at 0, 50 and 100°C) which allow better estimates of inactivation rates associated with a notion of variability of these values. Note that this information should be used by taking account of the temperature range and the type of matrix (simple or complex). Moreover, the data collected in the spreadsheet correspond to the time needed to obtain a first log10reduction (TFL) and they were obtained either from a linear decrease over a whole period of analysis or from the linear part of a biphasic decrease. Thus, the estimate of the conditions needed to obtain several log10-units of either inactivation or genome degradation should be made under the assumption of a linear decrease. Note that the statistical analyses have been performed with data provided in the formats required for publication (e.g. graphs, T90 values).

For a given range of temperature (< or ≥50°C) and a given type of matrix (simple or complex), the slope value (α1) was the same for all the virus species, which showed a similar sensitivity to temperature change and allowed comparison of virus survival using α0 values.

Viruses were first compared in terms of infectivity estimated by cell culture. At lower temperatures, the highest α0 values were observed for bacteriophages PhiX174 and PRD1, and human pathogenic virus HAV in both simple and complex matrices (Tables 2 and 3). Thus, these viruses showed the higher survival under these conditions. At higher temperatures, somatic and F-specific bacteriophages were classified as the most persistent and HAV seemed to be less persistent than these bacteriophages (Tables 4 and 5). Several studies comparing bacteriophages and other viruses are in agreement with a longer persistence of bacteriophages (Girones et al. 1989; Mocé-Llivina et al. 2003; Allwood et al. 2004, 2005; Charles et al. 2009). Moreover, viruses or phages with a DNA genome seemed to have a slower loss of infectivity as shown by the results obtained for Siphoviridae (somatic coliphages, Bacteroides fragilis, Lactobacillus helveticus, Lactococcus lactis bacteriophages), Tectiviridae (PRD1 bacteriophage) and Adenoviridae (human adenovirus) (Girones et al. 1989; Straub et al. 1992; Governal and Gerba 1997; Gantzer et al. 2001; Mocé-Llivina et al. 2003; Charles et al. 2009). By taking account of the range of temperature and the type of matrix, a ‘worst-case’ virus classified among the most persistent viruses in the present work could be selected for the evaluation of a thermal inactivation treatment.

The survival of several viruses was compared in terms of the degradation of their genome. The genomes of poliovirus and murine norovirus (MNV) were classified as the most persistent at lower temperatures (Tables 6 and 7). The human norovirus genome seemed to be the most persistent at higher temperatures (Tables 8 and 9), whereas it was completely the opposite at low temperature. In the present work, the log 10TFL values obtained by quantitative PCR were similar or higher to those observed by cell culture. This observation is in agreement with almost all the studies performing a simultaneous and quantitative comparison between infectivity loss and genome degradation, as they usually show a higher persistence of viral genome for HAV, poliovirus, coxsackievirus, MNV, FCV, F-specific RNA phages, rotavirus and astrovirus (Gassilloud et al. 2003; Duizer et al. 2004; Hewitt and Greening 2006; Kirs and Smith 2007; Bae and Schwab 2008; Baert et al. 2008a,b; Espinosa et al. 2008; Hewitt et al. 2009; de Roda Husman et al. 2009). Whatever the range of temperature, the slope values showed a higher sensitivity to temperature change for the viral genome in comparison with infectivity in complex matrices. Nevertheless, genome degradation represented only 15% of the data collected, and no data on DNA viral genomes were obtained. More data are needed for DNA viruses before confident predictions of their survival under different conditions can be made.

Virus inactivation proceeds faster in the high temperature range than in the low temperature range, which is not only a temperature effect, but a temperature–matrix effect. Temperature category and matrix were found to be highly correlated. In the higher temperature range (≥50°C), the slope value (α1) was higher in simple than complex matrices, indicating lower virus sensitivity to temperature change in complex matrices. In the lower temperature range (<50°C), the sensitivity to temperature change was similar in both simple and complex matrices. HAV was the most frequently studied virus in complex matrices at temperature ranging between 60 and 90°C. In some complex matrices, especially in food, a protective effect of a high content of protein, fat or sucrose on HAV infectivity has been suggested by several studies performed at high temperature range (>50°C) (Parry and Mortimer 1984; Murphy et al. 1993; Croci et al. 1999; Bidawid et al. 2000; Strazynski et al. 2002; Deboosere et al. 2004).

This study will also offer to researchers a comparison tool between their experimental values and a prediction of viral inactivation based on a statistical analysis of published data. The values of α1 and β2 are given in Tables 2–9 for particular combinations of method, matrix and temperature category, and the values of α0 and βs are given in terms of virus species. By using these data, the log TFL and the 95% interval can be estimated by using Eqns (1) and (7), respectively. At present, this comparison can be performed for recently published studies that were not included inside the spreadsheet when the statistical analysis was performed. Thus, Chandran et al. (2009) observed an inactivation of at least 8 log10 for phage MS2 in fresh and diluted urine at 30°C within 3 and 4 weeks, respectively. With the inactivation parameters obtained in the present work, the TFL value of MS2 in these experimental conditions (cell culture, low temperature range, complex matrix) could be estimated at 2·5 days [95% interval: 0·2–29 days], and a mean inactivation reaching 8 log10 could be obtained in a mean period estimated near 3 weeks under a linear decrease hypothesis. The results obtained by Ogorzaly et al. (2010) for human adenovirus in natural groundwater at 4 and 20°C (130 and 35 days respectively) are quite similar to those described in the present work as the values reach 90 days (95% interval: 7 days–3 years) and 24 days (95% interval: 2–289 days), respectively. The results obtained by Shieh et al. (2009) for HAV in spinach leaves (TFL value = 28·6 days) and in phosphate-buffered saline (PBS) (no decrease observed in 49 days) at 5·4 ± 1·2°C are also in agreement with our mean values reaching 69 days (95% interval: 6–840 days) and 202 days (95% interval: 16–2460 days), respectively. These comparisons between the estimations based on the present work and recent studies demonstrate that our approach allows a realistic prediction of the inactivation rates of different viral species by taking into account the range of temperature and the type of matrix.

Recently published data on virus genome degradation can also be compared with prediction values. However, the quantity of data on genome degradation remains low. Sow et al. (2011) observed TFL values reaching 39 s for MNV and 77 s for HAV in shellfish at 85°C. For the same viruses, the TFL values could be estimated at 96 s [95% interval: 45 s–3 min] for MNV and 129 s [95% interval: 10 s–26 min] for HAV. Thus, the mean TFL values estimated from our statistical analysis are twice those observed by Sow and colleagues; however, the published data are inside the 95% interval obtained from the statistical analysis. Moreover, all these data are in agreement with a very fast genome degradation showed by a maximal TFL value of 2 min. An overestimation of genome persistence could in some cases constitute a ‘worst-case scenario’.

Several outcomes of this review should be underlined:

  •  A significant interaction between temperature and matrix in a temperature–matrix effect was observed.
  •  Bacteriophages PRD1 and PhiX174 appeared to be highly resistant to temperature in comparison with pathogenic viruses. These bacteriophages may therefore be proposed as model viruses in a ‘worst-case study of treatment based on temperature’.
  •  The statistical analysis offers a prediction tool of virus inactivation for untested conditions (temperature, matrix, virus strain) in the form of empirical formulae.

This work also underlines the need for further studies in several areas:

  •  Among all the viruses and bacteriophages included in this work, none of them was homogeneously and significantly represented in the different categories of temperature, matrix and detection method. Studies of the same virus over a wide range of temperature are needed. For instance, experiments could be performed with HAV at temperatures below 50°C and with Poliovirus at temperatures above 50°C. Inactivation data on bacteriophage PhiX174 at temperatures below 50°C in simple matrices and above 50°C in complex matrices could also be very useful to confirm the high resistance of this phage.
  •  To validate the case for phages PRD1 and PhiX174 being ‘worst case’ for evaluation of thermal treatments, further studies comparing bacteriophages and pathogenic viruses under the same conditions are needed.
  •  Few data are published for temperatures around 50°C. More data in the temperature range of 40–60°C would give a more homogeneous distribution of the data in function of the temperature and improve the evaluation of viral sensitivity to temperature change.
  •  At low temperature, especially at temperature below 10°C, viral inactivation occurs relatively slowly and longer term studies, as those performed by Charles et al. (2009) and de Roda Husman et al. (2009) of more than 1 year, are needed to obtain strong data on viral inactivation.
  •  More data are needed on the thermal inactivation (infectivity loss or genome degradation) of DNA viruses.

Acknowledgements

This work was supported by COST Action 929 ‘A European Network for Food and Environmental Virology’. G. Sánchez is the recipient of a JAE doctor grant from the ‘Consejo Superior de Investigaciones Científicas’ (CSIC).

Ancillary