Cytomegalovirus seropositivity is associated with herpes zoster

Abstract

Herpes zoster (HZ) is caused by VZV reactivation that is facilitated by a declined immunity against varicella-zoster virus (VZV), but also occurs in immunocompetent individuals. Cytomegalovirus (CMV) infection is associated with immunosenescence meaning that VZV-specific T-cells could be less responsive. This study aimed to determine whether CMV infection could be a risk factor for the development of HZ. CMV IgG serostatus was determined in stored serum samples from previously prospectively recruited ambulatory adult HZ patients in the UK (N = 223) in order to compare the results with those from UK population samples (N = 1545) by means of a logistic regression (controlling for age and gender). Furthermore, we compared the UK population CMV seroprevalence with those from population samples from other countries (from Belgium (N1 = 1741, N2 = 576), USA (N = 5572) and Australia (N = 2080)). Furthermore, CMV IgG titers could be compared between UK HZ patients and Belgium N2 population samples because the same experimental set-up for analysis was used. We found UK ambulatory HZ patients to have a higher CMV seroprevalence than UK population samples (OR 1.56 [1.11 2.19]). CMV IgG seropositivity was a significant risk factor for HZ in the UK (OR 3.06 [1.32 7.04]. Furthermore, high CMV IgG titers (exceeding the upper threshold) were less abundant in CMV-seropositive Belgian N2 population samples than in CMV-seropositive UK HZ patients (OR 0.51 [0.31 0.82]. We found CMV-seroprevalence to increase faster with age in the UK than in other countries (P < 0.05). We conclude that CMV IgG seropositivity is associated with HZ. This finding could add to the growing list of risk factors for HZ.

Keywords:

• CMV
• immunosenescence
• reactivation
• shingles
• susceptibility
• zoster

Introduction

Herpes zoster (HZ) is caused by the clinical reactivation of latent varicella-zoster virus (VZV). This reactivation is linked to a declined immunity against VZV as noted in immunocompromised individuals, but also in older individuals.^1,2 Interestingly, re-exposure to chickenpox patients has been hypothesized to boost T-cell immunity against VZV and subsequently postpone the risk on HZ.^3,4 Several clinical risk factors for HZ have been documented including auto-immune diseases, immunosuppressive medication, depression and asthma.^5,6 However, HZ is also known to occur in otherwise healthy ambulatory and young patients.^5,6 In addition, certain HLA types have been shown to be enriched in individuals with prolonged HZ (post-herpetic neuralgia).⁷

Cytomegalovirus (CMV), another human herpesvirus (HHV) with latency-reactivation cycles, is well known for its modulating effects on immunity. For e.g., VZV-specific T-cells, but also T-cells against other HHV, from CMV-seropositive individuals are more mature, further differentiated and thus possibly less responsive compared to the VZV-specific T-cells from CMV-seronegative individuals.⁸ From a clinical perspective it has been shown, although some debate is still ongoing, that CMV-seropositive older individuals have an increased frailty and mortality.⁹ Recently, analyses using VZV-specific antibody titers¹⁰ and herpes simplex virus type 1 (HSV-1) antibody titers¹¹ have indicated that CMV-seropositivity was significantly associated with reactivation of VZV and HSV-1.

The present study set out to assess the level of CMV-seropositivity in UK HZ patients compared to random population samples from the UK to determine whether there is an association with VZV reactivation. If so, this could be a major addition to the list of risk factors of HZ in otherwise healthy individuals.

Furthermore, it is known that CMV-seropositivity is associated with ethnic and socio-economic factors,¹² as such we compared CMV-seroprevalence data between general population reference samples from the USA, Australia and Belgium. Differences in CMV-seroprevalence might explain the inter-country differences noted in HZ incidence.¹³

Results

UK HZ patients vs. UK-reference sample

The best model choice (after step-wise logistic regression based on AIC):$\begin{array}{l}\text{logit}(CMV=1_i)={\beta }_0+{\beta }_1*ag{e}_i+{\beta }_2*Gende{r}_i\\ +{\beta }_3*Datase{t}_i+{ϵ}_i\end{array}$

with i denoting individual i, logit (CMV = 1ⁱ) the probability individual i has a positive CMV IgG serostatus, β_j=0,…,3 parameters, age as a continuous variable, Gender (Female = 0, Male = 1) and Dataset (HZ = 0, UK = 1) as dummy variables and ε_i error terms. Parameters were β₀ = − 1.20 (SE 0.23, P < 0.001), β₁ = 0.045 (SE 0.0033, P < 0.001), β₂ = − 0.16 (SE 0.11, P = 0.13) and β₃ = − 0.44 (SE 0.18, P = 0.011). We note that a sensitivity analysis focused only on HZ samples collected during acute HZ onset gave highly similar results. Figure 1 shows per classification (based on age < 50 or => 50, gender and data set) the observed versus predicted CMV-seroprevalence. We note that dichotomous age stratification (at 30 years, 40 years, 50 years, 60 years) did not present statistically significant differences in CMV IgG serostatus between the younger and older age groups most likely due to sample size restrictions. The parameter estimation illustrates that the likelihood for CMV-seropositivity increases with age and compared to the adult HZ patients our reference UK-population has a lower OR for CMV-seropositivity (0.64 [0.45–0.90]). Figure 1 suggests that the latter effect is mainly caused by individuals younger than 50 years.

Figure 1. Observed and predicted CMV-seroprevalence percentage for UK HZ patient and UK general population samples. Caption: Groups are classified using age (50 years as cut-off), gender, data set (UK HZ (HZ) or UK-reference (UK)). The number of participants per classification group is shown (N). For each classification group the left column shows the observed CMV-seroprevalence whereas the right column shows the CMV-seroprevalence as predicted by the logistic regression analysis.

CMV-seropositivity was associated with an increased risk of HZ (OR 3.06 [1.32–7.04], and the effect had a minor non-significant tendency to decrease per additional year of age (OR 0.99 [0.96–1.01]).

CMV-seropositivity as a function of country, age and gender

The comparisons between the UK-reference dataset and BE-2002, BE-2006, USA and AUS data sets illustrated that CMV-seroprevalence is not only a function of age and gender as was shown above, but also of country (Fig. 2 and Table S1). Our results show that CMV-seroprevalence increases slower with age in BE-2002, BE-2006, USA and AUS datasets compared to the UK-reference data set.

Figure 2. Observed CMV-seroprevalence for UK reference sample (‘UK’), BE-2002 population sample (‘BE-2002’), USA population sample (‘USA’) and Australian population sample (‘AUS’). Caption: raw observed data are shown for women in upper panel and for men in lower panel.

CMV IgG titers in UK HZ patients compared to Belgian population samples

Although CMV-seroprevalence data differed between countries, we nevertheless assumed that CMV-titers could be compared between the UK HZ and BE-2006 data sets. A logistic regression analysis (controlling for age and gender) showed that CMV IgG titers above the threshold of 500 U/ml occurred less frequently (OR 0.51 [0.31 0.82], P Value = 0.0050) in BE-2006 CMV-seropositive individuals than in UK HZ CMV-seropositive patients. In addition, we noted that 10/48 young (<50 years of age) UK HZ CMV-seropositive patients were CMV IgM positive, compared to 4/81 in the older (>= 50 years) age group (P = 0.0077).

Discussion

Both cytomegalovirus seropositivity and the CMV-specific IgG titer are associated with increased frailty and mortality.^14,15 Furthermore, CMV-seropositivity is associated with modified T-cell immunity⁸ and increased serum antibody titers against VZV¹⁰ and HSV-1.¹¹ These increased serum antibody titers in CMV-seropositive individuals could be caused by an increased reactivation rate in these individuals.¹⁶ Our present study shows that CMV-seroprevalence is higher in samples from UK HZ patients (1998–2003) compared to samples of the general population in the UK (1991, but generally similar to 2002 as illustrated by Vyse et al.¹⁷). Defining the UK HZ patients as cases and the UK population as controls, we found CMV-seropositivity to be associated with an increased risk of HZ with an OR of 3.06 [1.32 7.04].

Of interest, we also noted higher CMV IgG titers in CMV-seropositive HZ patients compared to CMV-seropositives in a Belgian general population sample. This finding could indicate that CMV reactivation, known to cause an increase in CMV-specific antibody titer via the secondary immune response,¹⁶ (or perhaps even de novo infection) has occurred more recently or more frequently in HZ patients. This hypothesis suggests a co-susceptibility for human herpesvirus reactivation. We also assessed whether CMV-seroprevalence population data from other countries could be used as a control. However, we found CMV-seroprevalence to differ between countries even after controlling for age and gender. One could speculate whether socio-economic, cultural or other factors may contribute to these differences.¹² We cannot establish with certainty whether such factors also influenced the comparison between the UK HZ patients and UK general population. However, we used samples that maximized representativeness for both these populations within the limits of our control, and we found no indications that socio-economic or cultural factors differed between these datasets.

Further studies are needed to confirm these results. In particular, prospective case-control studies should reassess the magnitude of the risk caused by CMV-seropositivity on clinical VZV reactivation. Furthermore, a prospective study should also assess CMV presence in HZ patients via the documentation of the viral load at various sites (blood, saliva, urine, etc.).¹⁸ Our study suggests that this risk could be higher in young individuals. Indeed, we also found a higher CMV IgM seroprevalence in young (<50 years) CMV-seropositive HZ patients (21%) compared to older HZ patients (5%), although we could not find an additional statistically significant effect of age on the difference in CMV IgG seroprevalence between HZ and control individuals (probably due to the limited sample size). Nevertheless, our finding regarding CMV IgM seroprevalence could support our speculation that not only CMV reactivation, but perhaps also de novo CMV infection could be implicated in VZV reactivation. These findings could help explain why a disease that is generally thought to be due to a decline in VZV-specific immunity also affects healthy young individuals. In the advent of a CMV vaccine that can prevent CMV infection, its benefits might therefore extend to a reduction in HZ-related morbidity. Indeed, our study identifies CMV-seropositive individuals, particularly under age 50 years, as a possible target group for HZ vaccination.^19,20 Furthermore, this study also provides support to add CMV-seropositivity to the growing list of confounders in clinical vaccine trials.²⁰ The currently reported 51% reduction of HZ incidence after HZ vaccination might actually be higher in CMV-seropositive individuals.

Although the analyses in the current paper focused on HZ, similar associations have also been shown for other human herpesviruses such as HSV-1. CMV-seropositive individuals not only have increased antibody titers against both VZV and HSV-1, but a recent epidemiological analysis showed strong associations between clinical HZ and HSV in the general population.⁶ A possible explanation could be a shared increased susceptibility caused by CMV for example caused by T-cell exhaustion.⁸

A limitation of our analysis is that the general population samples with which the HZ patients were compared contain an unknown proportion of individuals who have had a history of HZ (at most 10–30%^21,22). Clearly, this limitation makes the observation of a significant difference between the 100% HZ and the “general population” samples more convincing (and it may help explain why there was no significant difference with the US and Australian general population samples). Additionally, as recently shown, CMV-seronegative individuals can have CMV-responsive T-cells thereby suggesting a former encounter with CMV, and this might also have affected our analyses.^23,24 Finally, our study is the first to show an association between CMV infection and HZ, further causal relations should be studied in other studies.

We conclude that CMV-seropositivity is associated with HZ implying that CMV infection could be a possible clinical risk factor for HZ.

Methods

Sample collection

This analysis was restricted to adult (>= 18 year) samples.

223 adult HZ patients (age range = [18 89] with median 53 years; F/M = 55%) were obtained from 2 prospective longitudinal studies in England (samples from 1998–2003) as detailed elsewhere.^25,26 Serum samples were collected and frozen at various time points up to 12 months after acute HZ onset. CMV IgG measurements were performed on acute HZ onset serum samples (thus from the first time point = baseline). However, for 41/223 (=18%) patients CMV IgG measurements had to be performed on a sample post acute HZ onset (within 12 months), because there wasn't a baseline serum sample available. In order to minimize the risk of new CMV infections, CMV IgM was additionally determined on the 41 post acute HZ onset samples. Two of these samples had “gray zone” CMV IgM values and were excluded. Another sample from the total group of 223 was excluded because it had gray zone CMV IgG values. Thus we retained 220 adult HZ samples for further analysis. IgG directed against CMV pp150, pp28, p38, and p52 and IgM against CMV pp150, p52 were determined on the thawed samples in the Antwerp University Hospital laboratory (Roche Cobas, Basel, Switzerland).

Reference adult population CMV-seroprevalence data were obtained for 1545 individuals (age range = [18 100] with median 34 years; 55% women) in the UK (1991 - nearly identical to the 2002 UK data set - CMV-specific IgG ELISA (Enzygnost, Dade Behring),¹⁷ 1741 individuals (age range = [18 71] with median 32 years; F/M = 51%) in Belgium (2002 - CMV IgG (Enzygnost, Dade Behring) ),¹⁰ 576 individuals (age range = [18 65] with median 29 years; F/M = 50%) in Belgium (2006) (see²⁷ for sampling method; CMV testing is the same as for UK HZ samples), 5635 individuals (age range = [18 49] with median 31 years; F/M = 54%) from the USA National Health and Nutrition Examination Survey (NHANES)²⁸ (1999–2002 - multistep IgG detection involving SeraQuest enzyme immunoassay (Quest International) and VIDAS test (bioMerieux Vitek) for more information see Bate et al²⁸) and 2080 individuals (age range = [18 59] with median 33 years; F/M = 50%) in Australia (2006 - CMV IgG ELISA (Medac)).²⁹ See Table 1 for more information.

Table 1. CMV-seroprevalence shown per dataset

Data set	Sample size	# CMV-seropositive (%)
HZ UK	220	158 (72%)
UK-reference	1545	780 (51%)
BE-2002	1741	556 (32%)
BE-2006	576	218 (38%)
USA	5572	3610 (63%)
Australia	2080	1267 (61%)

Data analysis

First, multiple logistic regression was performed with CMV IgG serostatus as dependent variable and age (as a continuous variable), gender, data set (HZ or UK-reference ‘UK’) as predictor variables. The inclusion of age and gender allows us to control for possible confounding factors. Furthermore, interaction terms between age, gender and dataset were tested (age*gender, age*data set, gender*dataset and age*gender*data set). Interaction terms define how the effect of the predictor variable X1 on the dependent variable changes conditional on the value of another predictor variable X2. For example, the term age*gender defines how the effect of age on CMV-seroprevalence differs between men and women. Logistic regression was performed as (1) a step-wise logistic regression including 2-way interactions between all variables and (2) a logistic regression without these interactions. Akaike's Information Criterion (AIC)-based stepwise logistic regression analyses were performed using the Generalized Linear Model software in Matlab 2013b. Confidence intervals shown are 95% CI.

Next, we reversed our analysis and defined the UK HZ group as cases and the UK-reference group as controls, thus assuming that only a minority (maximum of 10–30%)^21,22 of the UK-reference sample donors had a HZ history. In this case-control analysis we assessed the importance of CMV-seropositivity as a risk factor for HZ (also controlling for age and gender).

Subsequently, we performed a between-countries comparison regarding CMV seroprevalence by means of multiple logistic regression including the following datasets: UK-reference (baseline), BE-reference 2002 ‘BE-2002’, BE-reference 2006 ‘BE-2006’, USA-reference ‘USA’ and Australia-reference ‘AUS.

Finally, we also assessed whether high CMV IgG titers (defined as exceeding the upper limit of detection, i.e. >= 500 IU/ml) occurred more frequently in HZ CMV-seropositive patients (UK HZ) than in CMV-seropositive general population samples (BE-2006). The latter comparison is possible because HZ UK and BE-2006 CMV IgG titers were obtained in the same lab using the same technology, although differences in CMV epidemiology between the UK and Belgium might bias our analysis. Higher CMV IgG titers in HZ patients could indicate a recent exposure to CMV (either de novo or reactivation¹⁶). We also assessed CMV IgM serostatus, occurring during de novo CMV infection or CMV reactivation,³⁰ seroprevalence in these CMV IgG seropositive individuals.

Disclosure of Potential Conflicts of Interest

J.B. has received funding from SPMSD for the VZV Identification Program, which undertakes genotyping of VZV from vaccine adverse events. J.B. also heads the VZV reference laboratory, which has genotyped VZV for GlaxoSmithKline. B.O., N.H., R.P., H.J., H.S., M.Q., H.G., H.T. and P.B. report no conflicts of interest.

Supplemental Material

Supplemental data for this article can be accessed on the publisher's website.

Ethical Approval

East London and the city research ethics committee LREC R&WF2002/38 (UK HZ patient samples).

Acknowledgments

We thank Helena Tutil and Veronique De Vroey for their assistance in HZ patient sample management. We also thank the laboratories and staff that processed and tested the samples of the previously published general population sera in Belgium, The UK and Australia. We specifically acknowledge Pierre Van Damme, Dominic Dwyer and Gwendolyn Gilbert.

Funding

This work was supported by the Research Foundation Flanders (FWO) [personal grants to B.O. and H.T. and project grant G040912N]; the University of Antwerp [Pfizer-sponsored scientific chair to N.H.]; The Wellcome Trust [081703/B/06/Z] and the National Institute for Health Research (NIHR) UCL/UCLH Comprehensive Biomedical Research Center [to J.B.].