Nucleated eukaryotic cells encode a number of defence mechanisms to protect them from pathogens. Central to this is the activity of cell autonomous/innate immune responses. The innate immune response to HCMV is of critical importance to the intracellular environment encountered by the virus and the routes of entry determine some of the ways the virus is detected. Furthermore, given the variety of entry routes utilised by HCMV to infect different cell types, the innate immune response may be shaped differently in different cell types.
3.1. Evasion of Host Cell Apoptosis
One of the major challenges facing HCMV in its attempt to establish latency is a requirement to evade host immune responses following infection in the absence of expression of the vast array of cell modulators encoded within its genome. One such host response is the activation of apoptosis as a mechanism to remove a replication niche for the pathogen. For instance, during lytic infection, the expression of proteins such as UL36, UL37, UL38 and TRS1 as well as lncRNAs such as β2.7 restrict cell death pathways (Figure 2
A), allowing the virus to complete its replication cycle [80
]. However, does this present an issue during the establishment and maintenance of latency? Many of these gene products are not expressed during the early stages of infection of non-permissive cells, so how does the virus keep infected cells alive?
A simple solution to the conundrum would be that viral infection of non-permissive cells does not activate cell death pathways at all. However, studies with multiple viruses imply that this is not the case. This is exemplified in the context of HCMV, CD14+ monocytes, CD34+ cells and the THP1 cell line (Figure 2
B). It has been demonstrated that infection upregulates host cell survival signals and that blockade of this, or depletion of the survival signal itself, promotes cell death when challenged with the virus [86
]. The key aspect was that the virus upregulated a cell-encoded anti-apoptotic protein that is a pivotal regulator of myeloid cell viability—myeloid cell leukaemia-1 (MCL-1) protein—to counter the inevitable host activation of cell death pathways [86
]. A subsequent study demonstrated that infection of CD34+ cells with HCMV results in a transient activation of the pro-apoptotic protein Bak [88
]. In cells where the survival signal is blocked, the activation of Bak is prolonged, likely contributing to cell death. The HCMV-induced survival effect was highly dependent on ERK-MAPK signalling, which was activated in a gB-dependent manner [86
]. However, the precise identity of the receptor activated by gB in CD34+ cells remains to be determined, although a putative role for the DLD domain of gB [86
] may argue for a key role for integrins [53
]. It is likely that the identification of the receptor(s) being used by HCMV will also shed light on the upstream pathways activating the MEK-ERK pathway required for survival (Figure 3
A similar model applies to the infection of CD14+ cells, whereby MCL-1 upregulation was also demonstrated to be required for the survival of CD14+ monocytes following HCMV infection. However, differences are evident. Firstly, the effect in monocytes was PI3K-dependent [87
]. Secondly, MCL-1-mediated protection from apoptosis is a relatively transient event in CD34+ cells, with latently infected CD34+ cells no longer protected from cisplatin A-induced apoptosis 12 h post-infection [86
]. However, unlike in CD34+ cells, MCL-1 levels remain elevated in CD14+ monocytes until at least 48 h post infection (hpi) [87
], a long time after the initial binding events responsible for triggering its upregulation. Reasons for these differences are unclear, but the extended upregulation of MCL-1 and prolonged activation of PI3K may be linked to the unusual route that HCMV utilises during the infection of monocytes [78
]. Additionally, activation of Akt by PI3K and the triggering of ERK-MAPK signalling lead to inhibition of Tuberous Sclerosis Complex 2 (TSC2) [89
]. TSC1/2 inhibits mammalian target of rapamycin complex 1 (mTORC1) function, thereby inhibiting translation of proteins and ribosome biogenesis [91
]. This signalling activity may compensate for the inability to express pUL38, which inhibits TSC1 function during lytic infection [93
It is highly likely that the ability of HCMV to act as a poly-ligand through engagement of multiple receptors is critical for the survival response. Essentially, survival is the outcome of the activation of multiple signalling cascades working in concert to promote viability and only when these pathways are activated at the same time is protection observed.
3.2. Innate Immune Mechanisms for Detection of CMV
Underpinning the cell-intrinsic innate immune signalling apparatus that detects different components of virus particles is the activation of the production of interferon (IFN) and antiviral cytokines. This system consists of a series of pattern recognition receptors (PRRs) that detect various pathogen-associated molecular patterns (PAMPs). The role of the innate immune response to a viral infection is dual—to initiate a primitive antiviral state in the infected and neighbouring cells, and to help recruit cells of the adaptive immune system to commence a more specific response.
It is likely that viral glycoproteins such as gB and gH are detected at the cell surface, whereas other components such as viral nucleic acid can be detected at both the cell surface and intracellularly (Figure 3
). Toll-like receptors (TLRs) are a series of membrane-associated PRRs that have been shown to detect a variety of PAMPs involved in HCMV infection, such as TLR9 detecting hypomethylated CpG DNA motifs, and TLR2 recognition of gB and gH at the cell surface [94
]. Activation of TLRs leads to a downstream signalling cascade involving adapter molecules MyD88 and TRIF, culminating in the activation of the transcription factors NF-κB and IRF3 respectively. Monocyte activation in response to HCMV has been shown to be MyD88-dependent, suggesting a role for TLR sensing in response to HCMV in monocytes [95
]. Whilst the induction of cytokines such as IFN has been shown to be partially dependent upon TLR signalling, it has been demonstrated that infection of mice with murine cytomegalovirus (MCMV) leads to a biphasic production of IFN [96
], with the initial phase of IFN being, in fact, TLR-independent [97
Beyond cell-surface receptors for viral components, the other major route for detection of viral infection involves the detection of viral DNA after the viral nucleic acid has become exposed to the cell. Mechanisms of detection of viral DNA have come under intense scrutiny following initial reports that demonstrated activation of IRF3 in response to viral, bacterial and synthetic DNA [98
Multiple receptors, including cGAS, IFI16, DNA-PK, DAI, LRRFIP-1 and RNA Polymerase III have been shown to recognise DNA, activating innate signalling that leads to the transcription of interferon and other antiviral cytokines [99
]. Following detection by DNA sensors, the downstream signalling pathway utilises long-known kinase molecules such as TBK-1, but there are upstream proteins specific to DNA. The number of DNA sensors described highlights a complexity to DNA sensing that is not yet fully appreciated. Various DNA viruses occupy separate niches that may require sensing by different molecules. For example, vaccinia virus is a cytoplasmic DNA virus that has been shown to be detected by DNA-PK [100
] and the virus encodes an inhibitor of this molecule [101
]. However, whilst DNA-PK has been shown to be present in many cell types, it is not expressed in macrophages, for example, which suggests that another DNA sensor is active in these cells. Whilst much attention has been focused on DNA sensing using synthetic DNA, the sheer variety of sensors is indicative either of redundancy in the system or, perhaps more likely, of the fact that different sensors could play cell or intracellular compartment specific roles in response to pathogens.
A key player on which the DNA sensing pathways converge is the adapter molecule STING. As such, STING has been shown to be critical for the activation of IRF3 in response to DNA via TBK1 during HCMV infection [102
]. Similarly, STING was identified to be essential for the initial burst of interferon in response to HCMV infection in primary human endothelial cells [103
], suggesting a role for intracellular DNA sensing in the initial response to HCMV. Further evidence of the importance of STING to the sensing of HCMV DNA comes from recent studies that have demonstrated that HCMV encodes multiple inhibitors of STING (US9 [104
] and pp71 [105
]). Tegument protein pp71, which is delivered with the virion upon entry, targets STING for degradation. It is likely that this is dependent on the LxCxD motif present in pp71 that is required for its interaction with retinoblastoma protein [106
], since work from the Stetson laboratory has shown that the related LxCxE motif in SV40 T antigen is required for directed degradation of STING [107
]. Indeed, the study of pp71 that identified the domain responsible for targeting STING contains the LxCxD motif [105
The mechanism of detection of HCMV DNA upstream of STING is not yet fully established and could possibly involve multiple detectors acting in cell-type-specific ways. It has been reported that DAI plays a role in the detection of HCMV DNA in fibroblasts [108
], despite the importance of DAI having been doubted given the absence of any deficit of IFN production in response to DNA in DAI-/- MEFs and mice [108
]. Furthermore, IFI-16 has been suggested to play a role in detection of HCMV detection as disruption of IFI-16 with shRNA resulted in reduced expression of cytokines and reduced activation of IRF-3 following infection with HCMV [93
]. The same study also found a role for the tegument protein pp65 in disrupting IFI-16-mediated signalling, suggesting an important role for IFI-16 in the detection of HCMV DNA. Thus two tegument proteins (pp65 and pp71) delivered with the virion are targeting key components of the DNA sensing pathway. Indeed, it is possible that the delivery of the antagonists with the virion becomes pertinent during the establishment of latency—a time where very limited gene expression is observed. Finally, a further study also showed a role for cyclic GMP-AMP synthase (cGAS) in innate immune control of HCMV since endothelial cells lacking cGAS produced lower levels of interferon compared with control cells. This finding was replicated with STING and TBK-1 knockdown, further emphasising the importance of IRF-3-mediated signalling in the initial response to HCMV [103
Beyond IRF-3 signalling, another PRR termed AIM2 has been shown to recognise DNA and initiate the formation of an inflammasome that leads to the elaboration of IL-1β and IL18-independent of IRF-3 signalling. It is not fully understood if inflammasomes play an important role in the control of HCMV infection. Inflammasomes are multimolecular bodies that convert pro-IL-1β and pro-IL-18 into their active forms, IL-1β and IL-18. AIM2 has been well characterised as a DNA sensor present in macrophages; however, the exact role it plays in HCMV has yet to be established. It has been noted that monocyte activation in response to HCMV is independent of the inflammasome adapter protein ASC [95
], perhaps suggesting that DNA sensing via AIM2 is either powerfully inhibited by the virus, or non-essential.
Whilst cytokines and chemokines undoubtedly primarily function to mount an immune response against HCMV, the cytokines involved have also been shown to have an effect on the latency of herpesviruses. HSV-1 latency in vitro has been shown to be enhanced by the production of type I IFNs [110
], and several other herpesvirus latency states are shown to be IFN-sensitive. For example, systemic administration of type I IFN to mice has been shown to prevent MCMV reactivation [111
]—presumably via the known inhibitory activity of IFN against viral IE gene expression [112
]. The authors postulated that IFN treatment enhanced the activity of PML—an interferon-stimulated gene that is a well-established inhibitor of lytic gene expression of multiple herpes viruses [113
]. However, no direct evidence for a role for PML was shown and it is interesting to note that in HCMV infection of non-permissive THP1 cells the presence or absence of PML has been reported to have no impact on the establishment of latent infections [114
What all these studies of HCMV may reveal is that the nature of the innate immune response to HCMV differs depending on which cell types are studied and the entry mechanism used. The influence of the differing responses of cell types and how this might impact on the progression of the infection of HCMV warrants further investigation. One potential conceptual challenge is the question of whether the host or the virus benefits the most from an IFN-based response against the virus. Whilst IFN may be important for keeping herpesviruses at bay, if it contributes to or even enhances the establishment of latency then, on an epidemiological level, this is potentially advantageous. An obvious benefit from lifelong cycles of latency and reactivation is the increased opportunity for transmission to new hosts.
The lack of HCMV virus production during latency suggests that many of the pathways used to detect viruses may not be activated during latency. However, whether HCMV DNA is a potential long-term PAMP during latency is unclear. This question is linked to the broader issue of how the viral DNA is distinguished from cellular DNA by the DNA sensing machinery. In the case of vaccinia, a cytoplasmic DNA virus, the cellular location of the DNA may be the defining feature. However, HCMV DNA exists in a circularised form in the nucleus during latency [41
], and, furthermore, the HCMV episomes are extensively chromatinised [33
]. It is possible that this renders it indistinguishable to DNA sensors from cellular DNA and therefore not detected at all during latency. This would assume that sequence information is not important for recognition by DNA sensors. For example, if viral genomes contain an abundance of repetitive sequences favoured by DNA sensors, it may be a mechanism for the sensors to selectively identify foreign DNA. Certainly, sequence-specific recognition of foreign DNA could be important. A recent study of HIV revealed that the presence of CG di-nucleotides is markedly reduced in the genome—an adaptation that potentially protects it from recognition by host zinc finger antiviral proteins with potent antiviral activity [115
The sensors playing a role in sensing of HCMV have yet to be fully tested. Whilst a role for HCMV was suggested by the existence of an inhibitor of IFI-16 in pp65, more recent studies have demonstrated that genetic knockout of AIM2-like molecules (including IFI-16) using CRISPR have shown that IFI-16 is dispensable for the induction of IFN in response to HCMV, conflicting with data generated using an siRNA approach in human fibroblasts [116
]. The presence of the HCMV-encoded inhibitor pp65 has been accounted for by using a virus deficient in this molecule in the study and IFI-16 remained non-essential. Clearly, there is a complexity to HCMV DNA sensing that has yet to be fully understood and may again indicate that cell-type-specific responses are key.