Indirect effects of cytomegalovirus infection: Implications for vaccine development

Abstract Development of a cytomegalovirus (CMV) vaccine is a high priority due to its significant global impact—contributing to mortality in immunosuppressed individuals, neurodevelopmental delay in infected neonates and non‐genetic sensorineural hearing loss. The impact of CMV on the general population has been less well studied; however, a wide range of evidence indicates that CMV may increase the risk of atherosclerosis, cancer, immunosenescence, and progression of tuberculosis (TB) and human immunodeficiency virus. Due to the high seroprevalence of CMV worldwide, any modulation of risk by CMV is likely to have a significant impact on the epidemiology of these diseases. This review will evaluate how CMV may cause morbidity and mortality outside of the neonatal and immunosuppressed populations and consider the potential impact of a CMV vaccine on these outcomes.

damage to tissues. However, current evidence has only demonstrated an association rather than causal role in these disease processes.
The primary objectives of developing a vaccine against CMV have been to eliminate congenital infection and to reduce morbidity and mortality in highly immunosuppressed individuals. However, this review presents evidence that CMV may also contribute to the development of vascular, infectious, and oncological conditions in the general population. Globally, CMV seroprevalence is estimated to be between 60% and 90% 21 and therefore if there is any modulation in risk by CMV, vaccination is likely to have a significant impact on the incidence of these diseases. We used PubMed to perform a literature review of the epidemiology and potential pathogenic mechanisms of CMV using search terms 'CMV' or 'cytomegalovirus' or 'human cytomegalovirus' in combination with 'atherosclerosis', 'cancer', 'HIV', 'TB' or 'vaccination'. We included these themes as they represent the most studied areas of CMV in the general population, and areas where there would be the greatest implications if a causal interaction did exist. We performed a separate ClinicalTrials.gov search for CMV vaccination trials. In this review, we focus on the potential indirect effects of CMV in the general population, and methods to evaluate the effect of vaccination on associated disease development during and after vaccine trials.

| CMV in immune development
CMV seropositivity becomes increasingly prevalent with age. In lowmiddle income country (LMIC) settings, the age of acquisition is typically younger-studies in the Gambia and Uganda demonstrated 86%-95% seroprevalence at 12 months of age compared to 15% in UK 1-4-year-olds. 22,23 Even in high income settings, the seroprevalence eventually reaches 60%-80% by 65 years of age. [22][23][24] The age of CMV acquisition and the role of co-infection with other viruses are factors that may influence how CMV interacts with the host immune system. The human virome is an emerging concept in virology, and similar to the microbiome, the interactions between host and virus are likely to shape immune development and response. 25 CMV is a major determinant of variation in the immune system between individuals 26 and therefore likely exerts an important effect on its development. CMV-specific CD8+ effector memory T-cells dominate the immune repertoire of elderly patients who are seropositive for CMV. 17 This has been proposed due to chronic reactivation and low-level presentation of antigens to CMV-specific T-cells leading to memory T-cell inflation. 27 In healthy monozygotic twins who are discordant for CMV seropositivity, the number of effector CD8 T-cells and gamma-delta T cells are poorly correlated. 26 Similarly, there are significantly different plasma levels of IL-10 and IL-7 level between CMV seropositivity discordant twins, however this study did not adjust for any confounding factors that may explain this discrepancy. 26 Overall, evidence indicates that CMV is strongly associated with phenotypical differences between individual immune systems. With increasing age, there is a diminishing impact of genetic factors in determining the proportion of cell populations, cytokines and signalling molecules-this may reflect an increasing role of microbial influences, of which CMV is a likely a major contributor.
CMV is linked to profound changes in various cellular subsets, including reduction in naïve cell populations, and presence of cellular phenotypes and functions. 28 Overall, it appears that with age the T-cell compartment becomes increasing dominated by CMV-specific CD8+ T-cells with a corresponding reduction in the naïve T-cell population. In CMV seropositive experimental mice models naïve CD8+ T-cells are reduced and a similarly reduced naïve cell population appears to contribute to an overall mortality risk profile. 29,30 CMV-specific CD8+ cells typically re-express CD45RA alongside a reduction in co-stimulation markers such as CD28 and CD27. 20 The CD45RA isoform becomes increasingly expressed during chronic CMV infection and is associated with expansion of low-avidity cells.
CD45RA, the longer isoform of CD45, is typically expressed on naïve T-cells and impacts on the T-cell receptor (TCR) signalling pathway through altering the interaction of the TCR with CMV antigens.
Interestingly, these CMV-specific T-cells express few markers of T-cell exhaustion such as programed cell death protein 1 (PD-1), in contrast to T-cells seen in chronic infection secondary to HIV and hepatitis C virus (HCV). 28 This may be related to the low-level of chronic exposure of antigens in CMV latent infection compared to the high level of damage and replication seen in HIV and HCV. CD8+ CD28− CD57+ T-cells is a specific subset which has been found to be elevated in cohorts of CMV seropositive patients compared to controls. 31 These cells represent a highly differentiated phenotype with limited proliferative ability. CD57 is a surface sulphated glycan carbohydrate that has been found to be a marker of chronic immune activation in humans often in association with suppression of CD28. Studies suggest that this cell population can remain highly polyfunctional and maintain ability to kill pathogens ex vivo. 31 Nevertheless, several studies have associated expansion of this T-cell subset with increased mortality and morbidity. 32,33 The apparent contradiction of these findings may suggest that these cells are not directly pathogenic and may represent a proxy of chronic immune dysfunction and inflammation induced by CMV. Alternatively, these polyfunctional cells may be directly pathogenic through generation of excessive inflammation and bystander host damage and thus contributing to 'inflamm-ageing'.
Studying the relationships between age of acquisition and outcomes at different ages has been challenging due to the inherent difficulties in conducting long-term observational studies spanning decades and the inability to date when primary infection occurred.

| ATHEROSCLEROSIS
Many epidemiological studies have suggested an association between CMV seropositivity and cardiovascular mortality risk 5,8,[34][35][36] ; however, distinguishing whether this is a confounding effect of co-variants such as socioeconomic status, country of origin and smoking status, makes establishing causality challenging. A large longitudinal study of 14,000 subjects in the USA recruited from the National Health and 2 of 14 - Nutrition Examination Survey (NHANES) III (1988)(1989)(1990)(1991)(1992)(1993)(1994) showed that CMV seropositivity at initial sampling (age ranging from 25 to 90 years old) was associated with increased cardiovascular and all-cause mortality after 2 decades of follow up 35 (Table 1). The effect on all-cause mortality (hazard ratio [HR] 1.19, 95% CI: 1.01-1.41) remained significant after adjusting for confounders including diabetes, age and gender; however, the effect on cardiovascular disease (CVD) mortality disappeared after adjustment. 35 A nested USA case-cohort of 726 participants reported that patients with the highest 20% of CMV immunoglobulin G (IgG) titres had an increased rate of CVD, with a HR of 1.76 (95% CI: 1-3.11), as compared to patients with the lowest 20% of CMV IgG titres. 8 Conversely, in a 13,090 participant prospective study based in the UK which included 2514 deaths, there was no significant association between CVD mortality and CMV IgG seropositivity over an average 14.5 years follow up. 34 Evidence for the contribution of CMV to the development of atherosclerosis is derived from several sources. Firstly, antiviral therapy reduces the rate of transplant graft atherosclerosis. 46,47 This may be the mechanism by which antiviral therapy reduces both overall mortality and CMV related disease in solid organ transplants at 3-18 months post-transplant. 48 Secondly, CMV antigens have been directly isolated from smooth muscle cells of the carotid artery and aorta in surgical specimens. [49][50][51] There is mixed evidence for isolation of CMV antigens or CMV transcripts within the athero-  Figure 1a). CMV pre-disposes to a Type 1 T-helper (Th1) cytokine phenotype 6,52 and an M1 macrophage phenotype 19,53,54 in healthy patients during latency. Type 1 T-helper cell dominance has been mechanistically linked to the initiation and maintenance of atherosclerosis in mouse models. 55 Observational studies in humans show that Th1 phenotypes are associated with increased atherosclerosis as measured by carotid artery calcification. 56 35 in combination with raised CMV IgG levels therefore suggesting persistent reactivation as a mechanism. 34,59 The excess cardiovascular mortality risk may be due to CMV modulating the risk of plaque rupture and subsequent complications ( Figure 1a). CD4/CD8 inversion is associated with plaque rupture 60 and as mentioned, CMV is associated with an inversed CD4/CD8 ratio. Additional co-infection with other chronic viruses may exacerbate the phenotype; for example, HIV and CMV co-infection leads to inversed CD4/CD8 ratio and has a particularly high cardiovascular risk. 61,62 High serum levels of CMV DNA were detected in patients at time of admission with acute coronary syndrome. 63

| CANCER
The role of CMV in cancer is controversial and the evidence is highly contradictory. The supportive evidence for CMV's role as an oncogenic or onco-modulatory agent 64 is twofold; (1) isolation of CMV proteins and DNA from tumours with surrounding non-infected tissue and (2) well-studied in vitro effects of CMV on key cellular mechanisms of tumourigenesis-these are briefly summarised below and discussed in more detail elsewhere ( Figure 1b). 65,66 The key mechanisms of CMV tumourigenesis include inhibition of MHC II molecules alongside upregulation of PD-1 signalling. Several CMV proteins promote cell-cycling such as UL82, which both promotes entry into S phase from quiescence and increases the progress of cells through G 1, potentially through an interaction with cell-cycle regular retinoblastoma protein. 71 Whilst CMV proteins primarily function to promote viral persistence in the human host, in the appropriate context these may contribute to tumourigenesis.
CMV DNA and protein has been reported within a wide range of tumours-including breast, prostate, glioblastomas-but not the surrounding tissues. [72][73][74] CMV may come to infect these tissues through local infiltration of infected immune cells with subsequent local infection; however, why tumour cells appear to be susceptible to CMV infection is unclear. Many studies have been unable to isolate CMV DNA or protein in tumours. 40,56 Glioblastoma is the most studied example of CMV interaction with cancer-particularly in relation to epidemiology, prognosis, and pathogenesis 64,75 -however overall, the evidence is limited. CMV transcripts and proteins have been found within glioblastomas, although this is not a universal finding and there is a large amount of MOSELEY ET AL.   challenged with a lethal load of listeria were protected-compared to those with no CMV exposure-and this effect persisted for several months. 81 Subsequently, the persistence of this effect was questioned as it appeared that the protection waned past 6.5 months in mice. 82 Protection against unrelated pathogens is often felt to be due to the non-specific priming of the innate immune system. Alternatively, cross-reaction between peptides induced in the CMV immune response is another possible mechanism.

| CMV AND SUSCEPTIBILITY TO INFECTION
TB is a highly prevalent pathogen which has remarkable overlapping epidemiology and cellular tropism to CMV. 16 The mechanism of interaction between TB and CMV is unclear. 15 scores which persisted compared to CMV-seronegative HIV-exposed infants, even when controlling for socioeconomic factors, breastfeeding duration, and education. 14 CMV-seropositivity in combination with HIV-exposure is associated with an increased CRP level in six-week-old infants compared to CMV-seronegative HIV-exposed infants, which may suggest that CMV and HIV-exposure are synergistic in promoting an early pro-inflammatory state in infants. 13,92 Despite this, CMV and HIV co-infection does not appear to increase mortality when infants received early anti-retroviral therapy 14,88 and the long-term impact after 2 years of age is unclear.

| CMV AND VACCINATION RESPONSE
CMV may impair response to vaccination-particularly in elderly individuals where both T-cell and B-cell mediated effector functions are affected (Figure 1c). In studies evaluating influenza vaccine responses, CMV seropositivity has been associated with poor humoural vaccine response in adult populations. 11,12 One study demonstrated reduced influenza vaccine IgG production in CMV seropositive adults compared to seronegative adults across all ages, 12 whereas another found that CMV seropositivity was associated with reduced vaccine response only in adults over the age of 60. 11 However several studies have shown no association between CMV seropositivity and vaccine responsive to both influenza and pneumococcal vaccination. 10,93,94 These studies vary in the strain of influenza evaluated and the degree to which they adjust for confounding effects (most studies only adjusted for age and sex).
Various measures of immunogenic response to influenza vaccination-such as neutralising antibodies titres on haemagglutinin inhibition assays and antibody-dependent-cellular-cytotoxicity assays-decrease with age. 95,96 Poor immunogenic response to influenza vaccination is associated with an increase in the CD8+ CD28− T-cell subset 96,97 which is typically expanded in CMV infected elderly individuals as part of a late-differentiated CD8+ cell phenotype. 20,31,32 Similarly, late-differentiated CD4+ phenotypes, present in CMV+ patients, associated with poor humoural immunogenic response in elderly individuals. 11 In a study evaluating the immune responses to the Ebola vaccines, UK CMV seropositive young adults were found to have reduced antibody production. 98 Poor response to Ebola vaccination is correlated with increased levels of KLRG1, a marker of the terminal differentiated T-cell found in CMV seropositive individuals. 98 The expansion of these differentiated cells is potentially a consequence of low-level reactivation; it is not known whether these cells contribute directly to poor immune responses or are a proxy marker of an alternative pathological process.
In infants, most of the evidence regarding immunogenic responses to vaccination shows no association between CMV seropositivity and response as measured by vaccine antibodies titres and T-cell responses. Exposure to CMV occurs early in life in the majority of LMIC's, where seropositivity can reach 90% within 6 weeks of birth. 92 In a Zambian cohort of 369 CMV-seropositive infants, there was no difference in polio antibody titres at 18 months of age compared to CMV-seronegative infants. 44 In 9-month-old Gambian infants, CMV seropositivity was associated with mildly impaired IFNy response to measles antigen 1 week after vaccination-however this effect disappeared 3 months later. 45,99

| CMV RELATED MORTALITY
One interesting aspect of many epidemiological studies is the mortality gap between CMV seropositive and seronegative patients which is not explained by cancer or CVD, even once various MOSELEY ET AL. socioeconomical factors have been controlled for. 34,35,38 This nonspecific increase in death is due to a highly heterogenous group of causes.
In a Swedish longitudinal cohort, the accumulation of CD8+ CD28− CD57+ T-cells, a subset driven by CMV seropositivity as discussed above, was associated with increased mortality (Table 1). 37 CMV is thought to have a generalised pro-inflammatory effect potentially through persistent low-level replication. Raised CRP has been found to be an independent predictor of mortality in the elderly and in another study CRP is synergistic with CMV to increase the risk of mortality. 35 Overall mortality and CVD-related mortality was correlated with CMV-IgG levels and most of this effect was reduced when controlling for IL-6 and tumour necrosis factor-alpha levels. 38 It is likely that CMV interacts with other diseases through modulation of the immune system and localised reactivation in inflammatory niches. Whilst CMV may not initiate disease processes, through biasing immune responses towards pro-inflammatory phenotypes it may contribute to the maintenance of inflammation and impaired resolution.

| CMV VACCINATION AND THE INDIRECT EFFECTS OF CMV INFECTION
Several CMV vaccine platforms are in phase II and phase III trials