Secondary Logo

Journal Logo


Listeriosis: A Model for the Fine Balance Between Immunity and Morbidity

Lavi, Orita; Louzoun, Yorama; Klement, Eyalb

Author Information
doi: 10.1097/EDE.0b013e3181761f6f
  • Free


Listeriosis is a dangerous disease; its case fatality can reach 60%.1 It is caused by the intracellular bacteria, Listeria monocytogenes. This pathogen is a psychrotrophic bacterium (ie, it can survive and replicate at low temperatures, but grows optimally between 30°C and 37°C).2 Thus, exposure to L. monocytogenes is enhanced by poor hygiene and food storage conditions, especially when the cold chain is not adequately maintained. L. monocytogenes is primarily spread by food, with transmission from human to human playing a negligible role.3 Adaptive immunity to L. monocytogenes is mediated mainly by cellular response and is dominated by CD8+ T-cells,4 though CD4+ T-cells are required for the development of adequate memory response.5,6 As listeriosis can be fatal, there are no experimental data regarding the longevity of the human immune response to L. monocytogenes. However, infection of mice by L. monocytogenes has been shown to induce long-term and even life-long protective immunity.4,7 This acquired immunity is not related to the induction of clinical signs by infection with L. monocytogenes.8,9 T-cells specific for L. monocytogenes are frequently present in healthy individuals,10 which suggests that exposure to L. monocytogenes via food can stimulate specific immune response against the bacteria.

Gillespie et al11 recently described a sharp and steady increase in the incidence of listeriosis among the elderly population of England and Wales. A similar increase has also been observed in France, Switzerland, Finland, Germany, Denmark, and Norway during the last few years,12,13,13a although in the latter increase was also apparent in people aged 40–59 years. At least for England, France, and Germany the authors found no basis for attributing the upsurges to increased reporting or referral of listeriosis cases. Serotyping data, which were available only in the report from England, did not show an increase in any particular subtype of L. monocytogenes. Thus, it is unlikely that this increase is due to the penetration and dissemination of highly pathogenic L. monocytogenes. Other possibilities, such as improved diagnostic assays, demographic changes in the population, or an increase in patients with recorded immunodeficiency, could not explain this sharp increase (a 2-fold increase in 6 years). This upsurge remains unexplained.

In view of this rise in the incidence of listeriosis, and taking into account the above-mentioned characteristics of L. monocytogenes, we hypothesize that the upsurge of listeriosis incidence originates from a decrease in the overall population immunity to L. monocytogenes infection. In cases of contagious diseases such as measles and rubella, for which seriousness of illness increases with age of infection, a weakening of the force of infection (eg, due to immunization) can alter the age distribution of attack among the remaining susceptible population, thereby causing an increase in the incidence of serious cases.14–17 We expand this hypothesis to listeriosis, which is a noncontagious infection. As an extension of the classic SIR model18 (“susceptible,” “infectious,” “recovered”) to noncontagious diseases, we developed a mathematical model to study the balance between the immunized population fraction and the force of infection in noncontagious diseases in general (eg, food-borne diseases or vector-borne diseases for which humans are considered dead-end hosts), and in L. monocytogenes in particular.


General Model

We first consider a population divided into 3 compartments: naive (x), sick (y) and immunized (z) (Fig. 1). Each susceptible person can be infected with an age-dependent probability λ(T) = λσ(T), defined as a constant (λ) multiplied by an age-dependent function (σ(T)), where T is the age. The function σ(T) represents the susceptibility to infection and should be practically zero for young persons. It should increase to a maximum for the elderly. Each naive person can also be affected by a low dose of L. monocytogenes, leading to immunity with a probability of λv. Thus, the naive compartment has a transition rate of λv to the immunized compartment and of λσ(T) to the sick compartment. Recovery takes place in a rate of ψ. Immunity may wane and thus immunized individuals can become naive again with a transition rate of β. If immunity is life-long, we simply set β = 0. Even if immunity is not for life, only cases in which β << 1 are considered, since it is assumed that immunity for L. monocytogenes lasts many years.4,7 A constant low probability λ′σ(T) for the immunized to become sick can be introduced, and we set a probability of μ that sick persons would die as a result of the disease. As for natural mortality, a type I survivorship function is assumed in the generic model; that is, 100% of the population survives until the age of 80 and then everyone promptly dies. This model (Fig. 1) results in the following set of equations:

Generic model. See text for explanation. Terms neglected from the full generic model are in square brackets.

Simplified Model

The sick population fraction (y) is considerably smaller than either the naive (x) or immunized (z) populations (o(10−5)). Thus in the equation for z, the term ψy has a negligible effect. Furthermore, the illness probability in the immune state (z) is much smaller than in the naive state (x). We thus also neglect λ′σ(T)z in the equations for y and z. If indeed λ′ << λ (by at least a 2-order factor) then the removal of this term has no effect. If, on the other hand, λ′ ≈ λ, then the only way to reduce listeriosis is by reducing the exposure to L. monocytogenes; in that case, the results presented in the current manuscript would not be valid. Finally one can ignore the −λσ(T)x in the equation for x (but not in the equation for y), since λ << λv, again since the listeriosis rate is of the order of o(10−5). One is thus left with a simplified model (all the terms with no brackets in Fig. 1):

Given the above-mentioned assumptions, the solution of the simplified and original model are similar, as is shown in the results section.

Model Solution

The equations are solved for the lifetime of a patient from birth to death (0 → Tend). Typical listeriosis frequencies are of o (10−5 to 10−6). We thus set λ << 1. The total probability of a person to be sick during his lifetime, independent of the outcome of the disease, isL(λ) = ʃTλσ(T)xT)dT. Since x and z are not a function of y, one can solve their dynamics without taking y into account and obtain:

For this simple case, we can compute L(λ) analytically and numerically.

Detailed Model

Given the lifetime axis of an individual T:0 → Tdeath, and the year of birth tbirth, the life of an individual can be adapted to an absolute time—t through T = ttbirth.

A model more realistic than the simple model presented in the previous section (and therefore more detailed) can be computed using the following functions: λ(t), σ(T), N(T, t). λ(t) represents the exposure probability to L. monocytogenes in a given year t. σ(T) is the probability of being sick following an exposure at a given age T, and N(T, t) is the population at age T in the year t. Using these functions, one can compute the listeriosis frequency in a given year L(t) (which is different from the total listeriosis probability of an individual L(λ)) as:

We here assume that N(T, t) = N(T). In other words we assume that the age distribution of the population has not changed drastically in the last few years. We then normalize N(T) → N(T)/NTotalTN(T)dt = 1 to obtain: L(t = ʃTN(T)σ(T)x(T, t)λ(t)dT. This value is compared with the listeriosis frequency in the population in a year t.

Realistic Functions

To build the argument, we start the analysis from very simple functions and then, step by step, elaborate these functions.

At the first stage we assume the simplest model where σ(T) is a step function after 60 years,

, N(T, t)=const, λ(t)=const. This model can be fully solved analytically.

At the second stage, we maintain the assumptions on σ(T) and N(T, t), but vary λ(t) either as a step function or as a linear decrease, ie,

The possibility of a sharp and significant reduction in λ is realistic. There was a substantial shift in hygienic and storage conditions during the 1950s and the 1960s with the purchase of refrigerators in England that resulted in a significant change in parameter space.19 Hence, λ(t) depends on the absolute chronological time (t).

At the third stage, we include a complex population based on recent population estimates for England and Wales.20 We determine N(T) as a decay linear function: N(T) = 400 − 20(T − 60)(thousands) fitted to the observed data, and the observed values of σ(T).11 We ignore at the current stage the population over age 80 since it contributes a negligible number of listeriosis cases.

Finally, a realistic model is designed that enlarges the generic features observed in the simplified model. It considers a detailed adaptation of the simplified model to the observed data.

Four-Components Model

If one assumes that boosting may be required to maintain the immune state, a slightly more complex model can be built in which another state (w) is introduced. This model can be written as:

This additional state represents people originally immunized (by exposure to L. monocytogenes), who will become immune again following a boosting of the immune system, with a rate of λB. The probability of losing immunity is again β but, in this model, the transition from the immune population (z) is to the postimmune population (w) and not to the naive population (x). This model can again be solved using the simplifying assumptions mentioned above:

We show below that, as long as λ′′ ≤ λ, this model reproduces the maximum listeriosis frequency for intermediate values of λ, and it can actually produce the same results even for values of β higher than in the 3-compartments model.


Constant Exposure

To test the possibility that a reduction in the exposure rate to L. monocytogenes can indeed lead to an increase in the rate of listeriosis, we first solved a simple model [Eq. (2)]. Here we assumed that both the probability of immunization (λv) and the probability of being sick (λ) are constant and are actually a function of the exposure rate to L. monocytogenes. One can thus write λv = ελ and compute the listeriosis rate (L(λ)) as a function of λ. If one assumes that immunization is lifelong (β = 0), the model shows a maximal incidence of listeriosis for intermediate values of λ (Fig. 2). Increasing or decreasing λ toward this maximal value increases listeriosis incidence. For very low values of λ, the population is not immunized, but the L. monocytogenes prevalence is low enough to induce low disease levels. If, on the other hand, the L. monocytogenes prevalence is very high, by the age of 60 years, practically all the population is immunized and old-age listeriosis rarely occurs despite high exposure to L. monocytogenes. There is an in-between range where not all the population is immunized by the age of 60 years and exposure to L. monocytogenes is high enough to induce disease in parts of the population. The reduction of L. monocytogenes prevalence in many European countries may have led to this situation. Note that one can relax the assumption that immunity is life-long. However, the maximal listeriosis rate at intermediate exposure to L. monocytogenes is valid if immunity is long-lived (at least 20 years; β = 0.05 [Fig. 2 inset]). The ratio ε = λv/λ determines the worst value of λ (Fig. 2). This ratio is proportional to the ratio between the L. monocytogenes dose required for immunization and the dose eliciting disease in the susceptible population. The model at this stage contains only 2 parameters: β, ε; all other parameters have no effect on the model results. A maximum incidence is obtained for all values of ε.

Effect of λv and λ on the total number of sick elderly. Incidence of listeriosis L(λ) as a function of λ and ε = λv/λ. It can be seen that there is a value of λ for which the incidence is maximal. The worst value is a function of the ratio λv/λ. The figure is depicted for β = 0. The black line represents the current observed values of listeriosis. Crossing this black line toward the center would increase the listeriosis rate. The inset shows the effect of changing β for a constant value of ε = 50. If β is lower than 0.02 (dark lines) there is an intermediate value of λ maximizing the number of listeriosis cases. However, for β > 0.02 (red dashed lines) increasing λ can only increase listeriosis incidence.

Chronological Increase in Listeriosis

The generic mathematical model shows that, assuming a constant prevalence of L. monocytogenes, there is a value maximizing listeriosis. Raising or lowering the prevalence toward this value merely leads to a higher incidence. To further examine our specific hypothesis, we then considered the evolution from a high L. monocytogenes prevalence to a lower value as a result of a sharp drop in exposure to the pathogen following the utilization of new refrigeration technologies in the 1960s.17 Immediately following the decrease in exposure, the susceptible population is still immunized but is substantially less exposed to L. monocytogenes. The incidence of listeriosis thus drops sharply (thick solid line and triangles in Fig. 3). Then, for a long period, a part of the population slowly loses its immunity, and eventually this part of the population gets old (30–40 years later, ie, now) and becomes susceptible to clinical listeriosis. Following the argument presented in Figure 2, listeriosis frequency in the population can then increase even beyond its prior level, until stabilized at a new high value (thick solid line and triangles in Fig. 3).

Time evolution of L(λ) following a sharp drop in the L. monocytogenes prevalence during the 1960s. Symbols represent solution of the full model [Eq. (1)] and lines represent solution of the simplified model [Eq. (2)]. Following the drop in λ, L(λ)is substantially reduced (thick line). Then the population starts losing its immunity until most of the old population ages in a low λ environment. At that time, L(λ) is stabilized on a new high value. The triangles represent the same model with the full model solution. The dashed line and circles represent the same model with a slow drop in the listeriosis rate (40 years). The dashed dotted lines and the dotted thin lines with the diamonds and squares are for a realistic age distribution function and a realistic illness probability function, respectively. The values used for this analysis are: λ1 = 0.001, λ2 = 0.0001, ε = 70, β = 0.

The shape of the curve depicted in Figure 3 is generic and is not sensitive to the specific parameters used. The observed behavior is the result of 2 contrasting mechanisms; the immediate sharp drop in listeriosis following the decrease in L. monocytogenes, and the increase in the steady state of listeriosis incidence following the same drop in the exposure rate, and thus in the overall level of population immunity. As long as these 2 mechanisms occur, the same behavior will be observed. It can be further noted that if one starts to measure listeriosis after the initial drop, one would observe only the rapid increase as is actually the situation now. Similar results are obtained if one uses a gradual decrease over forty years in the exposure to L. monocytogenes [Eq. (5), dashed lines and circles in Fig. 3], and if one uses a realistic age distribution (dotted-dashed line and squares in Fig. 3) and a realistic function of the illness probability—σ(T)11 (thin dotted line and diamonds in Fig. 3).

We have simplified the model and neglected many elements. To validate that the simplifying assumptions have no important effect on the results, we have reproduced the curves in Figure 3 using the full model. As can be seen, there is practically no difference between the end result of the simplified model (lines) and the full model (symbols).

Realistic Model

A more realistic model reproduces the features observed in the simplified model. Our realistic model contains a linear decrease in the L. monocytogenes exposure rate—λ(t), a linear decrease in the population after the age of 60 as observed in the population20N(T, t), and the experimentally measured values of listeriosis probability—σ(T).11 Using realistic values for all of these functions reproduces the observed sharp rise in listeriosis cases (Fig. 4) for many arbitrary parameter sets. This model predicts, that even if no measures are taken, listeriosis incidence will reach a new plateau within a few years, with values higher than the ones obtained before the reduction in the exposure rates to L. monocytogenes. The long time lag between the decrease in L. monocytogenes results from the late onset of listeriosis in the elderly.

Evolution of L(λ) for a gradual decrease in λ. (Large figure: 1980–2020. Inset: the entire period 1940–2050.) Adaptation of model for observed data. The circles are the observed listeriosis rates in England over the last 20 years. The dashed thick line is the model for the elderly and the full line is the same model for pregnant women. We assumed a gradual decrease in the exposure rate from 0.001 to 0.00025 over 20 years from the 1960s to 1980. We further assumed that ε = 70 and β = 0, the observed age distribution and illness probabilities and a linear decrease in λ.

Pregnant Women

A similar model can be developed for pregnant women. The only difference between the model for pregnant women and for the elderly is the age distribution of disease susceptibility σ(T). While old persons start being sensitive around age 60 and keep increasing their sensitivity, pregnant women are sensitive between the ages of 20 and 40. We have therefore developed a model in which the disease susceptibility σ(T) is constant between the ages of 20 and 40, and zero otherwise. This model, as in the case for the elderly, has a peak listeriosis frequency in the middle range of exposure rates (data not shown). The main difference between the elderly and pregnant-women populations is that changes in the exposure rate to L. monocytogenes affect pregnant women faster than elderly. Therefore, pregnant women reach a steady state much faster, and are expected to show an increase in listeriosis long before the elderly. We may now be in a period after a new steady state has been achieved, and may not be able to see a reduction in the listeriosis rate among pregnant women (Fig. 4, dashed line).


The model presented up to now has only 3 possible states. One could in principle assume that, to maintain immunity, the immune system has to be boosted by a de novo exposure to L. monocytogenes. We have developed a model to include the possibility of boosting. A new compartment of postimmunized population was introduced with a lower probability of being sick in the postimmunized state. This extended model has 4 compartments: naive, immunized, postimmunized and sick. We compute again the total number of listeriosis cases as a function of the exposure rate to L. monocytogenesL(λ) or the total number of listeriosis cases in a given year—L(t). The introduction of a fourth state allows us to relax the requirement that the immune memory is eternal by adding a requirement that the illness probability following exposure is substantially higher in the first exposure than in the exposure following an immune period. A maximal listeriosis frequency at middle values of the exposure rate λ can be obtained even if immunity lasts for only 10 years (Fig. 5). The inclusion of this additional compartment replaces the requirements that immunity lasts for many years by the possibility of a graded-illness probability.

Two-immunity-states model (boosting). Range of parameters where the 2 immunity states model has a maximal listeriosis rate for intermediate ranges of L. monocytogenes exposure. The x and y axes represent the values of ε and β. Gray regions represent parameter ranges where the model has a maximal listeriosis rate for intermediate ranges of L. monocytogenes exposure. The presented results are for λB = λv,λ′′ = 0.05λ.


The interplay between pathogen exposure and the resultant development of protective immunity plays a strong role in determining the incidence of infectious diseases. In diseases associated with old ages there is a balance between the immunization obtained at younger ages and the pathogen prevalence. In the current study we discussed the influence of this balance on the old-age morbidity of listeriosis which is a noncontagious infection. We considered here the effect of hygienic improvement on the incidence of listeriosis, a food-borne infection affecting pregnant women, immune-compromised patients, and the elderly. However, the model we have constructed may be implemented for noncontagious infectious diseases in general (eg, food-borne diseases or vector-borne diseases for which human are considered dead-end hosts).

We have provided support for the hypothesis that the increase in listeriosis incidence observed in many countries in Europe in the last decade is explained by a reduction of exposure to L. monocytogenes a few decades ago due to changes in hygiene and food storage conditions.

Understanding the nature of immune response to L. monocytogenes is of high importance for assessing the validity of our model. Two important aspects are the existence of long-term immunity after infection by L. monocytogenes, and cross-protection by various L. monocytogenes strains. Studies in mice have shown that a long-term protective immunity was elicited after infection by L. monocytogenes,4,7 and have also shown cross-protection among the various strains.21 However, data are available only from animal studies. Furthermore, in most of the existing studies the infection route is parenteral, while natural infection occurs through the oral route. The nature of immunity acquired in humans by oral exposure to L. monocytogenes should be further studied. We have showed that our results are less sensitive to the length of the immune period if a boosting effect is assumed. However, there are no available data to show that such an effect exists.

So far as we are aware, none of the risk assessments of listerosis consider the possibility of acquired immunity elicited by exposure to L. monocytogenes. About 1 in 15,000 servings of ready-to-eat food contains very high doses of L. monocytogenes (approximately 105 colony-forming units of L. monocytogenes in 1 g of food).22 These levels have been associated with several outbreaks of listeriosis. Taking into consideration that in the United States the number of ready-to-eat food servings consumed by high-risk populations is estimated to be around 1010 per annum,23 the incidence of listeriosis should have been orders of magnitude higher than the current incidence of about 2500 cases a year.24 A possible explanation for this relatively low incidence of listeriosis may be in the immunity acquired through previous exposure to the organism. Thus, taking into account this fact and the data on immunity to L. monocytogenes, we believe that our explanation for the current upsurge of listeriosis observed among the elderly is plausible and should be considered when analyzing the current trends in incidence of this highly important disease.

Nevertheless, there are no longitudinal data on exposure to L. monocytogenes, and obviously no direct measures have been performed to reveal what is the exact level of exposure that leads to the development of immunity and what level leads to listeriosis. While our results show the plausibility of the suggested scenario in a reasonable parameter range, they cannot prove this assumption. Studies to examine the existence of cellular immunity to L. monocytogenes in various age groups, as well as prospective studies to correlate surrogates of immunity with actual protection from disease, will contribute significantly to assessments of exposure and immunity to L. monocytogenes, and to a better understanding of the dynamics of listeriosis in the population.


We thank Ran Nir-Paz (Department of Clinical Microbiology and Infectious Diseases, Hadassah Hebrew University Medical Center, Jerusalem, Israel) for taking part in the brain-storming during model construction.


1.Vazquez-Boland JA, Kuhn M, Berche P, et al. Listeria pathogenesis and molecular virulence determinants. Clin Microbiol Rev. 2001;14:584–640.
2.Farber JM, Peterkin PI. Listeria monocytogenes, a food-borne pathogen. Microbiol Rev. 1991;55:476–511.
3.Sauders BD, Pettit D, Currie B, et al. Low prevalence of Listeria monocytogenes in human stool. J Food Prot. 2005;68:178–181.
4.Pamer EG. Immune responses to Listeria monocytogenes. Nat Rev Immunol. 2004;4:812–823.
5.Shedlock DJ, Shen H. Requirement for CD4 T cell help in generating functional CD8 T cell memory. Science. 2003;300:337–339.
6.Sun JC, Bevan MJ. Defective CD8 T cell memory following acute infection without CD4 T cell help. Science. 2003;300:339–342.
7.Lara-Tejero M, Pamer EG. T cell responses to Listeria monocytogenes. Curr Opin Microbiol. 2004;7:45–50.
8.Darji A, Mohamed W, Domann E, et al. Induction of immune responses by attenuated isogenic mutant strains of Listeria monocytogenes. Vaccine. 2003;21(suppl 2):S102–S109.
9.Goossens PL, Milon G. Induction of protective CD8+ T lymphocytes by an attenuated Listeria monocytogenes actA mutant. Int Immunol. 1992;4:1413–1418.
10.Munk ME, Kaufmann SH. Listeria monocytogenes reactive T lymphocytes in healthy individuals. Microb Pathog. 1988;5:49–54.
11.Gillespie IA, McLauchlin J, Grant KA, et al. Changing pattern of human listeriosis, England and Wales, 2001–2004. Emerg Infect Dis. 2006;12:1361–1366.
12.Antal EA, Hogasen HR, Sandvik L, et al. Listeriosis in Norway 1977–2003. Scand J Infect Dis. 2007;39:398–404.
13.Koch J, Stark K. Significant increase of listeriosis in Germany—epidemiological patterns 2001–2005. Euro Surveill. 2006;11:85–88.
13a.Goulet V, Hedberg C, Le Monnier A, et al. Increasing incidence of listeriosis in France and other European countries. Emerg Infect Dis. 2008;14:734–740.
14.Anderson RM, Grenfell BT. Quantitative investigations of different vaccination policies for the control of congenital rubella syndrome (CRS) in the United Kingdom. J Hyg (Lond). 1986;96:305–333.
15.Anderson RM, May RM. Vaccination against rubella and measles: quantitative investigations of different policies. J Hyg (Lond). 1983;90:259–325.
16.Griffiths D. A catalytic model of infection for measles. Appl Stat. 1974;23:330–339.
17.Knox EG. Strategy for rubella vaccination. Int J Epidemiol. 1980;9:13–23.
18.Kermack WO, McKendrick AG. A contribution to the mathematical theory of epidemics. Proc R Soc Lond. 1927;115:700–721.
19.Thévenot R. A History of Refrigeration Throughout the World. Paris: International Institute of Refrigeration (IIR); 1979.
20.Population Estimates for UK, England and Wales, Scotland and Northern Ireland. Available at:
21.Kaufmann SH. Acquired resistance to facultative intracellular bacteria: relationship between persistence, cross-reactivity at the T-cell level, and capacity to stimulate cellular immunity of different Listeria strains. Infect Immun. 1984;45:234–241.
22.Gombas DE, Chen Y, Clavero RS, et al. Survey of Listeria monocytogenes in ready-to-eat foods. J Food Prot. 2003;66:559–569.
23.U.S. Department of Health and Human Services and United States Department of Agriculture. Draft Assessment of the Relative Risk to Public Health from Foodborne Listeria monocytogenes Among Selected Categories of Ready-to-Eat Foods. Washington, DC: U.S. Department of Health and Human Services and United States Department of Agriculture; 2001.
24.Mead PS, Slutsker L, Dietz V, et al. Food-related illness and death in the United States. Emerg Infect Dis. 1999;5:607–625.
© 2008 Lippincott Williams & Wilkins, Inc.