Cytomegalovirus: Shape-Shifting the Immune System

In recent years, systems approaches have been used to begin elucidating the extent that the immune system differs between individuals. These have focused largely on measuring the composition, phenotype, and gene-expression patterns of circulating immune cell populations (1–3). The hope is that this will begin to clarify how variation in these parameters relates to disease, how they might impact vaccination responses, and, ultimately, how they help develop a blueprint for better overall health. Concurrent technological advancements have allowed the measurement of multiple, individual immune cell frequencies and immune-regulating cytokines from patient blood samples (e.g., mass cytometry and multiplex ELISA), and in combination with unbiased genomics approaches will ultimately help to spawn personalized medicine (4, 5). Notably, nonheritable factors are major drivers of immune system variation, with one prime example being the microbiome (6). However, quite surprisingly, another major factor is human CMV (HCMV) infection (HHV-5, a β-herpesvirus) (7). Identical twins discordant for HCMV infection vary in >50% of ∼200 measured immune parameters (7), a remarkable impact for a single, common viral infection. In this review, we discuss unique aspects of the broad immune response to HCMV infection that arise during its distinct phases of infection, specific strategies used by CMV to target them, and how this may ultimately impact health and disease.

HCMV is the prototypic member of the Betaherpesvirinae and is a dsDNA virus with an ∼236-kb genome expressing up to ∼750 protein-encoding open reading frames emanating from complex control of transcription and splicing (8, 9). CMVs, similar to all Herpesviridae, have coevolved with their individual hosts for millions of years to establish a persistent/latent infection that is never cleared (10). Consequently, CMV replication is species specific, in large part because the virus has evolved many fine-tuned strategies to inhibit various immune defenses unique to those hosts (11, 12). In the United States, infection rates of prepubescent children range from 30 to 60%, varying by sex, ethnicity, and socioeconomic status, increasing steadily to 50–90% by the age of 50 y. In general, infection is lowest in non-Hispanic white males, persons of higher education, and those living in less crowded conditions (13). The overall incidence of CMV infection is significantly higher in South America, Asia, and Africa (>90%) as compared with the United States and Western Europe (14). Consequently, the fact that many people living today in the Western developed world remain uninfected with CMV into adulthood has facilitated assessing how it impacts immunity.

CMV infection has three distinct phases: 1) a systemic replication phase in many peripheral tissues that strongly activates the innate immune system and specific NK cell populations and primes a diverse Ab and T effector memory (Tem) cell response; 2) a tissue-localized persistent phase that continues for months to years and continues to shape innate and adaptive immunity; and 3) multisite latency with restricted viral gene expression that promotes immune inflation during a subsequent lifetime (15–18) (Fig. 1, Table I). Despite this prolonged and multifaceted interaction with its host, CMV normally only causes acute disease when immunity is naive or compromised, exemplifying how coevolution over millennia has resulted in a largely nonpathogenic détente. However, the extremely broad impact of CMV on immune system homeostasis in healthy adults likely impacts aspects of health and disease over time.

Primary CMV infection in healthy people is essentially asymptomatic, and therefore studies assessing the activation and contribution of innate defenses during the phase I infection in people are sparse. Additionally, the species-specific replication of CMV necessitates using animal models to help define key innate control mechanisms. In this regard, mouse CMV (MCMV) has been the most instructive (19), but rhesus (20, 21) and guinea pig (22) CMV models have also been very instructional. Most mouse and guinea pig infections are done systemically (i.p. or i.v.), with s.c. infection often performed in monkeys. How the initial activation/priming of innate and adaptive immune responses with these infection routes and doses might differ from that of natural mucosal infection is an important question and is discussed later.

NK cells and type I IFN (IFN-I) are key regulators of early CMV control (23, 24). MCMV induces two major systemic phases of IFN-I. The initial phase is produced by splenic stromal cells and is proportional to the infecting dose, peaks in sera at 10–12 h, and is dependent on B cell–derived lymphotoxin αβ signaling (25). The second phase occurs at 36–48 h in response to the first burst of MCMV production and is produced by both plasmacytoid dendritic cells (DC) and conventional DC. In addition to IFN-I, these DC subsets produce high levels of tissue-localized and systemic IL-12 and IL-18 (26, 27), which further NK cell activation (28–30) and help to prime adaptive responses. Notably, second-phase IFN-I does not directly restrict MCMV replication levels (31), whereas the first phase does (32).

Work with MCMV during the last three decades has revealed many mechanisms by which NK cells contribute to antiviral defense in various inbred mouse strains (33), and more recent advances in genome sequencing has furthered this (34). It is clear that altered NK cell licensing and expression of disparate activating and inhibitory receptors underlie many of these differences (35, 36). In C57BL/6 mice ∼50% of splenic NK cells express the Ly49H activating receptor, which binds the MCMV m157 protein expressed on the surface of infected cells, promoting their resistance to viral replication in the spleen (37–42). m157 induces the activation and proliferation of Ly49H⁺ NK cells during the first week of infection (43, 44), which then restrict MCMV replication mainly by cytolysis (45). Notably, in the absence of Ly49H–m157 interactions, the role of NK cells in restricting MCMV replication can be minimal (e.g., in BALB/c mice) (33, 46), in large part due to CMV neutralizing several NK cell effector pathways (35).

Several other innate effector cell types also make important contributions in both the direct and indirect control of early MCMV infection. Lymphoid tissue sinus-localized macrophages can capture MCMV during initial infection and restrict it from productively infecting reticular stromal cells (47). Mast cells are activated within hours of infection and produce chemokines that recruit CD8 T cells to infection sites within the first week (48, 49). Neutrophils are rapidly recruited by NK-produced IL-22 and contribute to viral control via TRAIL (50). Invariant NKT cells are activated by the first and second waves of innate cytokines, independently of CD1d, and help to control early MCMV replication (51, 52). The importance of invariant NKT cells in responding to HCMV infection is also suggested by viral restriction of CD1d expression (53, 54). Various myeloid cell types also help to shape the earliest response to CMV infection, but they also play a key role in disseminating CMV to sites of persistence and harboring latent viral genome in mice and humans, highlighting their complex role in CMV defenses (55–60).

A key role for NK cells in controlling HCMV infection is supported from a patient case study who lacked them (24). Historically, NK cells were thought to respond generally to viral infection, functioning en masse as front-line innate defenders. However, more recent data, many in mice, indicate that populations of memory-like NK cells that display properties of adaptive lymphocytes can be primed during phase I (61) and may be preferentially maintained during later phases of infection. One of these is the m157-driven expansion of Ly49H⁺ NK cells in C57BL/6 mice, which remain at higher levels for months postinfection and provide better MCMV protection when adoptively transferred into naive mice (62). Interestingly, healthy CMV-infected people >60 y of age show a significant increase in the numbers of CD56^dim NK cells in blood (63). These cells also display a unique CD57⁺NKG2C^hi cell surface phenotype (64), a distinct transcription factor expression profile, and epigenetic remodeling of the Ifnγ gene locus (65). They also preferentially expand in CMV-infected transplant patients, correlating with control of systemic viremia (64). Some studies indicate that NKG2C/HLA-E interactions might be impacting this NK cell subset (66, 67), and a currently unidentified HCMV-encoded NKG2C ligand has also been postulated (similar to Ly49H-m157 in mice), but the precise mechanisms underlying why NKG2C^hi NK cells expand in CMV⁺ patients are yet to be fully defined (68). However, as NKG2C^hi NK cells can comprise as much as 50% of the entire repertoire in some infected individuals (64), which results in a substantial reduction in overall NK pool diversity, this may impact the ability to respond to heterologous infections or tumor immune surveillance. Interestingly, a population of CD56⁺CD8⁺ NKT-like cells also expands in peripheral blood of middle-aged persons infected with CMV, but may be more a result of aging than being virus imposed per se (69).

The CMV-specific T cell response shows several unique characteristics (18) and is differentially impacted by the three infection phases. MCMV replication is ultimately controlled during the first week in most peripheral organs, correlating with the initial expansion of effector T cells, but replicates for many more weeks at high levels in selected mucosal sites such as the salivary gland. A robust and diverse epitope-specific CD8 and CD4 T cell response is primed in mice during this first week, most of which contract and go on to establish a stable memory pool in the next 1–2 wk (70, 71). In contrast, selected populations of T cells do not significantly contract, expand only during this second phase, and/or continue to increase during viral persistence, and these have been termed inflationary memory cells (15, 71–73). These inflationary T cells display a largely Tem phenotype, are maintained by a relatively small population of CD27^hi CD8 T cells that are more central memory–like (74), depend less on the immunoproteasome than do stable memory cells, are dependent on the initial dose of viral inoculum (75), and require the continuous transcriptional activity of FOXO1 (76). Notably, despite the phenotype of these inflationary T cells being consistent with their continued exposure to Ag, they do not show typical signs of exhausted T cells induced by several other chronic viral infections (77, 78).

As in mice, CMV induces a broad CD8 and CD4 T cell response that typically comprises 10–20% of all circulating T cells in CMV⁺ persons (79). HCMV also induces the expansion of selected, epitope-specific Tem cell subsets over time in healthy individuals, reaching staggeringly high numbers in some people (>20% for a single, epitope-specific response) (80, 81), and again typically show no signs of exhaustion. The expansion of these Tem cells scores as one of the most prominent CMV-imposed impacts in discordant monozygotic twins (7). Interestingly, these inflationary HCMV-specific T cells can use unique cosignaling pathways for their activation (82). Many lack expression of CD27 and/or CD28 (83), and the emergence of high numbers of HCMV-specific CD4 T cells displaying a cytolytic phenotype is also seen (84). Differential cosignaling requirements for stable and inflationary memory pools (85–89) and induction of CD4 CTLs (90) is also observed in CMV-infected mice.

CMV-specific Ab responses are strongly induced. Polyclonal human Ig containing high HCMV Ab levels has been reported to provide some protection against congenital CMV infection and/or disease in at-risk pregnant women (91), but not all trials could reproduce this (92). This may be due to differing HCMV IgG avidity (93), or its specificity for particular envelope protein complexes that mediate viral entry into fibroblasts (gB/gH/gL, trimer) and other cell types (gH/gL/UL128–131, pentamer) (94, 95). In mice, transfer of MCMV Ig can reduce viral spread within tissues and restrict dissemination from sites of initial infection (96). The levels of MCMV IgG increase steadily throughout persistence/latency, although avidity appears to plateau several weeks before phase II is ultimately controlled, whereas IgM levels drop significantly after phase I (17). Inflation of HCMV IgG has also been observed in healthy infected persons (97). Mice deficient in B7-CD28 cosignaling are severely compromised in MCMV IgG responses (17), and in combination with their crippled T cell response never control phase II infection (R. Arens and C. Benedict, unpublished observations). Adoptive transfer of memory B cells into immune-deficient RAG^−/− mice also reduces MCMV replication in both prophylactic and therapeutic settings, and it provides longer protection than serum Ig transfer (98). As seen for MCMV T cell responses (75), the magnitude of the Ig and memory B cell response is dose-dependent, although Ig inflation and affinity maturation does still occur at low doses (17). As natural infection by nursing or saliva transfer in both mice and people is likely to be relatively inefficient (99, 100), the relative initial infection efficiency may partly explain why CMV⁺ individuals often show significant variation in immune inflation decades later. Whether selected memory B cell subsets inflate in CMV-infected mice or humans is currently unknown, but B cells in CMV⁺ twins do show altered IL-10 signaling responses (7). Interestingly, some HCMV Ags appear to induce poor memory B cell responses during primary infection (101), suggesting that well-designed vaccine approaches may be able to outperform natural immunity. Despite all evidence that CMV induces a robust B cell/Ig response that is protective in several therapeutic settings, it is noteworthy that immune control of phase I and II MCMV infection is unchanged in mice lacking B cells (102), although control of viral reactivation in phase III is compromised to some extent (103), indicating that significant redundancy exists for immune control of primary infection.

After high-level persistent CMV replication during phase II is ultimately brought under control, viral latency is established and a host–virus détente ensues for life in phase III. Latency as defined originally for HSV is generally appreciated to be 1) stable, nuclear, and extrachromosomal maintenance of the viral genome in selected cells and tissue sites for the life of the host, 2) a state of restricted gene expression with no viral production, and 3) associated with the ability to reactivate from this state to full-replicative potential under the right conditions (104). CMV encodes the largest genome of the Herpesviridae (105), displays a broad cellular tropism, and is capable of establishing latency in several cell types (106). This differs from HSV, which establishes latency only in neurons. For HCMV, latency has been studied almost exclusively in hematopoietic precursor and myeloid lineage cells, in part because these cells are relatively easy to access from patients, but primarily because these cells clearly harbor latent viral genome in CMV⁺ people and good experimental models have been established (59). In mice, MCMV can establish latency in both endothelial and lymphoid-tissue reticular cell types (107, 108), and can be experimentally reactivated from various tissues (109).

Although less is known about the molecular and cell-intrinsic regulators of latency for MCMV than HCMV, mouse studies have revealed key aspects of how phase III infection impacts immune inflation (16). Inflationary T cells are maintained at higher levels when adoptively transferred into latently infected mice than into naive uninfected mice, indicating that sporadic/abortive CMV gene expression during phase III latency facilitates continued exposure of immune cells to viral Ag and accentuates immune inflation (73). The relevant cellular source driving this inflation is radioresistant stromal cells (110, 111), suggesting that CMV latency in nonmyeloid cells is a major driver. Interestingly, a spread-deficient MCMV mutant can also induce memory inflation, albeit at lower absolute levels (112). This suggests that a latency program can be triggered shortly after initial CMV entry in some cells, perhaps due to their specific differentiation/activation state, with other cellular states favoring the initiation of lytic replication and phase I infection. Evidence exists for a key role of lymphotoxin and IFN-I signaling in regulating this choice (113, 114). Importantly, these two paths do not have to be mutually exclusive, and they are likely to be conserved among the Herpesviridae (115). However, recent studies show that significantly more viral gene expression likely occurs during HCMV latency than dogma might predict (116–118), many of which perform immune-modulatory functions (119, 120), and this may help to explain why T cells of particular Ag specificity undergo immune inflation and others do not.

Intriguingly, healthy CMV⁺ people with strong pre-existing immunity can be reinfected (121), and in mice reinfection enhances T cell inflation (122). The significant variability in the magnitude of immune inflation between individuals may depend in part on reinfection frequency. As many genetically distinct strains of HCMV exist (123), reinfection may just simply represent an insufficient ability of prior immunity to protect against a heterologous virus. However, a genetically identical rhesus CMV isolate can reinfect monkeys multiple times and induce strong primary Tem responses, a strategy that is being developed as a vaccination approach (124). Interestingly, the ability of rhesus CMV to superinfect in the presence of strong pre-existing immunity is dependent on viral immune evasion strategies that dampen CD8 T cell responses (125).

Finally, γδ T cells also contribute to controlling MCMV replication and limiting virus-induced organ disease (126), and specific clonotypes expand in mice and show inflation within selected organs, similar to that observed for αβ T cells. Support for a role of γδ T cells in controlling, or at a minimum strongly responding to, HCMV infection also exists in both healthy persons and transplant patients (7, 127, 128). However, as it remains unclear what self and/or CMV Ags γδ T cells recognize, it is difficult to know what phase of infection may be most critical in promoting their inflation.

Despite inducing a robust and diverse innate and adaptive immune response, CMV successfully progresses through its three-phase infection and establishes lifelong détente with its host. To aid in this, CMV has developed many sophisticated mechanisms targeting host immunity. It is likely that only about a third of the >750 CMV open reading frames are required for the nuts-and-bolts of entry and lytic replication (129), suggesting that the rest have evolved to combat the host immune response. Our laboratory has studied several of these immune-modulatory genes during the last two decades, primarily focusing on those that target and/or intersect with CD28 family and TNF family immune signaling pathways. These and many other CMV immune evasion mechanisms have been reviewed in recent years (12, 35, 55, 130–135), so we only briefly discuss those that we deem to be the most likely to impact immune inflation.

As already discussed, CMV induces the expansion of specific NK cell subsets. In addition to encoding proteins that specifically bind and activate NK cells (e.g., m157-Ly49H), CMV commensurately uses proteins to restrict NK cell–activating ligands and their receptors on the surface of infected cells. HCMV blocks expression of NKG2D ligands such as MICA/B and ULBP family proteins (35). HCMV UL141 restricts the surface expression of CD155 and CD122, which activate NK via DNAM-1 and CD94 (136). Additionally, UL141 also binds and inhibits expression of the TRAIL death receptors to block NK-mediated killing by this TNF family cytokine (137), highlighting the polyfunctionality that a single CMV immunomodulatory protein can have. MCMV utilizes a similar strategy via m166 to inhibit TRAIL apoptosis (46). HCMV also encodes proteins that bind NK inhibitory receptors, such as the MHC class I homolog UL18 that binds to LIR-1 (138), and UL40 that stabilizes HLA-E expression to promote NKG2A engagement (66). In fact, m157 can bind the inhibitory receptor Ly49i encoded in Ly49H⁻ 129/J mice, which likely attenuates NK cell activation (39). Collectively, this multilayered approach will almost certainly contribute to the CMV-induced expansion, function, and/or memory of specific NK subsets

CMV also utilizes multiple strategies to regulate detection by both CD8 and CD4 T cells. First, multiple HCMV and MCMV proteins operate to reduce cell surface MHC expression during lytic replication, thereby limiting T cell activation during phase I and II of infection (133, 139). In infected cells, the HCMV US2, 3, 6, 10, and 11 and MCMV m152/gp140, m04/gp34, and m05/gp48 proteins downregulate MHC membrane expression and limit peptide presentation to T cells. MCMV restricts expression of costimulatory molecules (e.g., B7 and CD40) and preferentially enhances expression of inhibitory ligands (e.g., PD-L1) as another T cell blockade strategy (140–143), and viral mutants unable to block B7-CD28 signaling induce higher CD4 T cell response and show reduced replication in phase II (144). HCMV employs similar strategies to inhibit B7-CD28 signaling and promote death of T cells that encounter infected DC (145), although the specific viral proteins responsible have not been identified as for MCMV. HCMV UL144 is an ortholog of HVEM that selectively binds the coinhibitory molecule BTLA and potently blocks T cell proliferation (146–148). UL144 can also activate NF-κB and is expressed during latency (120, 149), suggesting that it may impact T cell responses in phase III of infection. Additionally, all primate CMVs encode orthologs of IL-10 (150), and HCMV viral IL-10 is expressed during both lytic and latent infection, perhaps contributing to the creation of a latency niche where effector T cells are suppressed (117). Notably, HCMV can also promote production of host IL-10 by virus-specific CD4 T cells that encounter MHC class II–presented latency Ags (119, 151). This strategy of hijacking host IL-10 immunosuppression to promote viral persistence is similarly employed by MCMV (152, 153). Finally, CMV encodes both chemokines and chemokine receptors, which impact T cells and operate to fine-tune the inflammatory environment during all phases of infection (154).

Does the broad impact of CMV on the immune system ultimately contribute to human disease? CMV can be a serious human pathogen when immunity is compromised or naive, as in the case of transplant patients and fetal infection, and this is why vaccine development is a top priority (19, 155). However, there is a paucity of direct evidence for whether the lifelong détente in healthy people is deleterious, and the field remains largely agnostic on whether CMV is a driver or a passenger in associated diseases. CMV⁺ persons have been reported to exhibit a higher incidence of all-cause mortality, mainly from increased cardiovascular disease (CVD) (156–158), and recent meta-analysis indicates CMV⁺ individuals have a 22% higher risk of developing CVD (159). As CMV infects vascular endothelial cells, a direct role for the virus is possible. Intriguingly, CMV CD4 T cells can express CX3CR1, which binds fractalkine expressed by activated/damaged endothelium, suggesting a mechanism by which CMV may kick-start or amplify CVD during multiple phases of infection (81, 160, 161). CVD is an autoinflammatory disease, and consequently whether CMV infection may be linked with other common autoimmune disorders such as rheumatoid arthritis, systemic sclerosis, systemic lupus erythematosus, diabetes, Sjögren’s syndrome, and several intestinal disorders has also been examined (162). However, because most data are from small patient cohorts and/or case studies, causation remains unclear.

Evidence for CMV negatively impacting immune function over time is suggested by its association with inflammaging and immune senescence in the elderly (163), with recent results indicating that it likely accelerates these processes (164). Very old Swedish cohorts exhibit an immune risk phenotype that includes CD4T/CD8T ratios <1 in blood (165, 166). However, the apparent ability of CMV to dial up the homeostatic rheostat of the immune system may also provide some benefit, as younger CMV⁺ people have been reported to mount better vaccination responses to influenza (167).

Although HCMV is not generally considered an oncogenic virus, viral nucleic acid has been reported to be present in various tumor types at high frequency (168), with the most notable example in recent years being that of glioblastoma (169). However, the absence of replicating virus, the low percentage of glioma cells harboring viral genomes, and the inability of some groups to detect HCMV at all has again made the contribution of CMV controversial. Nevertheless, some recent clinical results suggest that including CMV-targeted approaches in the treatment of glioblastoma may be helpful (170, 171). Future studies assessing whether the high proportions of inflationary CMV Tem might be preferentially recruited to tumor sites will be of interest, as their unique phenotype may impact the tumor microenvironment regardless of whether viral Ag-specific recognition occurs.

It is becoming more clear that immune system diversity among healthy individuals does not segregate into distinct groups, but instead varies across a continuum to generate multiple immunotypes (164). These immunotypes vary less in younger people, and even less in twins, and they highlight the importance of nonheritable factors such as CMV infection in shaping an individuals immune system (7, 164). Interestingly, if someone varies highly in the levels of one particular immune cell type, this does not correlate with them being an outlier for overall immune cell composition (164). CMV activates many arms of the immune system, and together with its modulatory strategies results in a major impact on its homeostasis. Given the multitude of acute and persistent infections that we encounter during a lifetime, it is quite extraordinary that CMV can shape immunity to this degree, and we would argue that this stems from unique aspects of its three-phase infection. It seems reasonable to hypothesize that this may ultimately impact health and/or disease at various levels, and the rapid advancements being made in systems-based immunity are likely to bring this into clearer focus in the coming years.

The authors have no financial conflicts of interest.