Human cytomegalovirus infects human populations at a high frequency worldwide. During the long coevolution of virus and host, a fine balance has developed between viral immune evasion strategies and defense mechanisms of the immune system. Human cytomegalovirus encodes multiple proteins involved in the evasion of immune recognition, among them UL18, a MHC class I homologue. Despite almost 20 years of research and the discovery of a broadly expressed inhibitory receptor for this protein, its function in immune modulation is not clear yet. Recent data suggest that besides inhibitory effects on various immune cells, UL18 may also act as an activating component during CMV infection. In this review, we provide an overview of the biology of UL18 and discuss several attempts to shed light on its function.
Human cytomegalovirus (HCMV)2 is a highly prevalent β-herpesvirus that causes life-long latent infections. The virus possesses an impressive array of genes dedicated to subverting the human immune system (1, 2, 3, 4). During the long coevolution of virus and its host, a sophisticated adaptation to the human immune system has been established. HCMV becomes a serious threat to survival when this fine balance between infection and immunity is disrupted, such as in AIDS or transplant patients (3, 5). Although our knowledge about the individual HCMV gene products involved in immune evasion is constantly increasing, there are still many molecules with as yet unknown functions. One of the most studied HCMV-encoded proteins in the latter category is UL18, a MHC class-I homologue discovered in 1988 by Beck and Barrell while sequencing the laboratory strain AD169 (6). To date, all of the clinical HCMV isolates analyzed have retained the gene for UL18 (7), reflecting its importance for the virus. UL18 has been proposed to be involved in immune escape because it is not needed for viral replication in vitro (8). However, despite recent studies providing new insights into the effects of UL18 on specific immune cells such as NK and T cells (9, 10), the function of this protein in HCMV infection still remains an enigma. In this article, we provide an overview of the biology of UL18 and discuss attempts to shed light on its function.
The UL18 protein, a MHC class I homologue
UL18 is a 348-residue type-I membrane glycoprotein (11) that associates with β2-microglobulin (β2m) (12) and can bind endogenous peptides (13). Several approaches by different research groups have suggested a remarkable structural resemblance of UL18 to MHC-I complexes despite the fairly low sequence identity. The proposed extracellular domains of UL18 share only ∼21% sequence identity with the corresponding α1, α2 and α3 regions in classical MHC-I (6). In contrast, the secondary structure of UL18 is similar to that of MHC-I proteins (14). A three-dimensional molecular model of UL18 (Protein Model DataBase access code PM0074971; http://mi.caspur.it/PMDB/) suggested an overall fold very similar to that of classical MHC-I. However, it also contained several localized structural variations that could reflect functional differences (Fig. 1) (15, 16). Another distinction from MHC-I proteins is the high degree of glycosylation. The UL18 sequence encodes 13 potential N-linked glycosylation sites (6) vs one in human MHC-I (17). Of note, HCMV infection induces several glycosyltransferases (18), suggesting that the virus may be able to alter the glycosylation system of the host. Furthermore, the extent of the glycosylation of UL18 varies upon the use of different viral vectors for UL18 expression. For example, infection with HCMV results in greater posttranslational modifications than adenoviral delivery (19). The large shell of carbohydrates is not unique to UL18 in HCMV; the recently identified MHC-I homologue UL142 is also heavily glycosylated, with 17 potential N-linked glycosylation sites (7). The mouse CMV MHC-I homologue m144 contains four N-linked carbohydrate addition sites compared with two in most murine MHC-I molecules (20, 21). Glycosylation contributes to the stability of proteins (22). One may speculate that the relatively higher glycosylation of MHC-I homologues protects viral proteins from degradation or could sterically hinder interaction with MHC-I receptors (other than LIR-1) or coreceptors (22). In addition, the carbohydrates in UL18 have been suggested to prevent recognition by viral unique short (US) region proteins that down-regulate MHC-I but not UL18 (23).
Peptide binding to UL18
Due to the structural similarities between MHC-I and UL18, peptide binding to UL18 is of particular interest. Whether this common feature also reflects functional similarities remains unknown, i.e., if peptides are involved in UL18-receptor interactions or act as mere structural stabilizers of the molecule. The peptides eluted from UL18 are of similar length to those bound by MHC-I and are derived from proteins degraded in the cytoplasm (13). Sequence analysis revealed that UL18-binding peptides comprise a binding motif similar to some of the most common human MHC-I motifs, such as that of HLA-A2, consisting of a leucine or methionine at position 2 and a hydrophobic amino acid at position 9 (13). In the absence of peptides, the stability of UL18 is significantly reduced (14, 24). Four conserved tyrosine residues that make hydrogen bond interactions with the N terminus of the peptide in the A pocket of the peptide-binding groove in classical MHC-I are also found in UL18 (14). Residues that comprise the other end of the groove and interact with the C terminus of the bound peptides are suggested to differ between UL18 and MHC-I, providing a potential explanation for the capacity of UL18 to accommodate peptides longer than nine residues (13, 14).
TAP transfers cytosolic peptides into the endoplasmic reticulum where the peptides are loaded onto MHC-I molecules (25). UL18 has been proposed to bind peptides both in a TAP-dependent (23) and TAP-independent manner (19). Griffin et al. expressed UL18 by using a recombinant adenovirus vector in a TAP-deficient fibroblast cell line (19). UL18 was detected on the cell surface, yet in lower amounts than in fibroblasts with intact TAP. Park et al. used a recombinant vaccinia virus for UL18 expression, which was significantly reduced upon blocking TAP (23). The data obtained in these two studies are not contradictory, although the interpretations vary. Both studies detected lower levels of surface UL18 when TAP-dependent peptides were not available, indicating that UL18 is able to use different peptide-loading mechanisms even though the TAP-dependent pathway results in higher expression levels. Obviously, HCMV has evolved strategies to ensure proper peptide loading on UL18. During infection, there is only a partial temporal overlap of UL18 and US gene expression, which may contribute to the insensitivity of UL18 to the effect of US molecules (23). Interestingly, UL18 expression can be reduced by the HSV inhibitor of TAP, ICP47, but not by US6, the HCMV-encoded TAP inhibitor (19, 23). One possible explanation is that US6 may be less efficient than ICP47 in inhibiting TAP. Thus, TAP-transported peptides that escape US6-mediated inhibition in combination with TAP-independent peptides could allow for sufficient expression of UL18 on the surface of infected cells.
UL18 expression during HCMV infection
The complete protein composition of the HCMV virion, the “HCMV proteome,” has been determined (26). However UL18 was not found among the 71 identified HCMV-encoded virion proteins, suggesting that UL18 exerts its functions in the infected cell. Two differently glycosylated forms of UL18 are detected in HCMV-infected cells or when expressed by different vectors (19). The 67-kDa form is short lived and susceptible to endoglycosidase H but can nevertheless be found on the cell surface (19, 23). The 160-kDa form appears from 72 h onward in the infectious cycle and is endoglycosidase H resistant (19). Removal of all glycosylation reduces the size of this protein to 35 kDa (19), illustrating the contribution of carbohydrates attached to this protein.
UL18 mRNA is low in abundance and appears late during infection (23, 27, 28). Due to difficulties in detecting the low levels of UL18 protein on the surface of infected cells (9, 10, 19, 29), this molecule has been studied mostly in transfected cells (29, 30) or when expressed using recombinant (retro) viruses (19, 23). Detection of surface-expressed UL18 on HCMV-infected cells is further complicated by confounding Ab interactions with virus-encoded Fc receptors (10, 19). Another contributing factor to difficulties in detection may be the relatively low immunogenicity of UL18 due to the shading effect of glycans. The different expression systems that have been used as an alternative to HCMV infection are not flawless either; it seems to be impossible to create long-term stable cell lines for UL18 (29, 30, 31), and recombinant viruses have confounding effects on cell metabolism as well as on the glycosylation machinery. In addition, any indirect effects of UL18 mediated through interaction with other HCMV proteins cannot be detected in these systems.
Expression and function of LIR-1, a receptor for UL18
While searching for a cell surface expressed receptor for UL18, Cosman et al. discovered a novel molecule termed LIR-1 (32). The systematic denomination assigned to LIR-1 is LILRB1 (leukocyte Ig-like receptor, subfamily B, 1) (33); other names used in the literature are ILT2 (Ig-like transcript 2) (34) and CD85j (35). The inhibitory receptor LIR-1 is expressed predominantly on monocytes, macrophages, dendritic cells (DCs) and B cells, but is also found on the surfaces of subsets of NK and T cells (34). Cross-linking of LIR-1 on the surfaces of DCs suppresses Ca2+ mobilization, cytokine production, and induction of Ag-specific T cell proliferation (36). In T cells, LIR-1 signaling inhibits Ag-specific proliferation, cytokine production, and cytotoxicity (37, 38, 39). LIR-1 can compete with CD8 for MHC-I binding and may serve as an inhibiting coreceptor during TCR-MHC interactions (40, 41). Coligation of LIR-1 with B cell receptors prevents Ca2+ mobilization, cytokine production, and isotype switching (34, 42). For primary NK cells, the LIR-1/MHC-I interaction is too weak to inhibit lysis in the absence of other inhibitory receptors, with the exception of HLA-G (Ref. 43 and references within and Ref. 44). Having in mind the strong affinity of UL18 for LIR-1, one may very well envisage an inhibitory effect of UL18 also for NK cells.
LIR-1 recognizes a broad range of classical (HLA-A, -B, and -C) (24, 32, 34) as well as nonclassical (HLA-G and -F) (32, 40, 45, 46) MHC-I molecules. The reason for this remarkable ligand diversity lies in the nature of the LIR-1 binding template; LIR-1 recognizes the relatively nonpolymorphic α3 domain of the MHC-I H chain as well as UL18 (24) and, importantly, the invariant β2m contributes 70% of the total interaction surface as determined in the crystal structure of HLA-A2/LIR-1 (41). UL18 binds to LIR-1 with >1000-fold higher affinity compared with host MHC-I (24). It has been speculated that such a higher affinity may be due to local differences in LIR-1 contact residues between the α3 domains of MHC-I and UL18. Different spatial orientations of the α3 domain and of the β2m subunit in UL18/LIR-1 compared with MHC-I/LIR-1 complexes may also play an important role (16, 41). We have recently demonstrated the existence of two disulfide bonds in the α3 domain of UL18 (instead of only one found in MHC-I) and the requirement for the formation of both of these disulfide bridges for complex formation with β2m (Fig. 2) (16). The additional disulfide bridge may result in further stabilization of the α3 domain of UL18 and influence the orientation of β2m, which would contribute to a more favorable interaction with LIR-1. It has been previously suggested that the glycosylation of UL18 is probably not involved in binding to LIR-1, because insect cell-produced UL18 with modified shortened carbohydrates binds with similar affinity to LIR-1 as does fully glycosylated UL18 produced in mammalian cells (24). Furthermore, when the glycosylation sites of UL18 are mapped on the crystal structure of the HLA-A2/LIR-1 complex, no glycans are localized in the vicinity of the LIR-1 contact site in the α3 domain (41). Peptide binding to UL18 does not influence binding to LIR-1 according to Chapman and colleagues (24).
Do variations in UL18 molecules influence interactions with LIR-1?
UL18 proteins from several clinical isolates differ in their binding affinity to LIR-1 compared with AD169-UL18 (15, 47). The underlying variability of UL18 genes, which deviate in composition by as much as 20 amino acids, is considerably greater than that of other viral immune evasion genes (Ref. 47 and references within). Sequence variation is not only conferred to the α3 domain of UL18, the postulated main LIR-1 binding site (24, 41, 48), but is also found in the α1 and α2 domains (47). We and others correlated certain variations within and outside the proposed LIR-1 binding site with alterations in binding affinity to LIR-1, confirming that the α3 domain is an important binding region for LIR-1 (16, 24). However, we also proposed that α1 residues may modulate UL18 affinity to LIR-1 (15, 16). Provided that the interaction of UL18 with LIR-1 during immune evasion is as important as it seems, one may speculate that the varying affinities for LIR-1 by UL18 proteins of clinical HCMV isolates reflect a differential modulation of host immune responses. Some authors (47) have noted that UL18 proteins of AD169 and clinical HCMV isolates bind solely to LIR-1 but not to other members of the LIR-family. However, according to surface plasmon resonance binding assays, UL18 binds also to LIR-2, although with weak affinity (48, 49).
It is also possible that UL18 binds to another still undefined receptor. If so, genetic mutations could provide an advantage to the virus through the modulation of binding affinity to this unknown molecule. Should such a molecule “X” be expressed as an activating receptor on NK or T cells, several findings that established UL18 as stimulus for immune responses would be explained (9, 10, 29, 50). Alternatively, mutations in the α2 domain may alter the binding of peptides, modifying the stability of the UL18 complex or its interaction with undefined receptors.
The roles of UL18 in immune responses
Modulation of NK cells.
Initially, UL18 was proposed to be responsible for MHC-I down-regulation through the sequestration of β2m, but subsequent identification of several US proteins proved their accountability for this task (12). Cells infected with a deletion virus lacking the UL18 gene (dUL18) still displayed reduced MHC-I expression levels (8). In contrast, the deletion of US2 and US11 restored normal MHC-I expression in the laboratory strain AD169, which argues against a contributing role of UL18 in interference with MHC-I expression (51). For several years, UL18 was then assumed to act as a decoy molecule for NK cells. In mice, the biological significance of a viral MHC-I homologue was demonstrated in vivo. Deletion of the MCMV gene m144 resulted in an NK cell-dependent decrease in viral replication (20). Yet, the results for UL18-mediated inhibition of NK cells were subject to controversy. Leong et al. proposed that UL18 actually enhanced NK cell killing (29). When endothelial cells and macrophages were infected with AD169 or dUL18, no inhibitory effect of UL18 on NK cells could be detected either (52). Conversely, another study using UL18-transfected target cells in a xeno situation supported an inhibitory role (31). However, these investigations did not take into account the expression of LIR-1 on NK cells. Recently, the effect of UL18 on NK cell lysis and degranulation in the context of LIR-1 expression has finally been assessed. Comparing dUL18 to AD169 infection as well as using an assay system based on UL18 expression via a replication-deficient adenovirus vector, investigators demonstrated that UL18 inhibits LIR-1+ NK cells but presumably activates LIR-1− NK cells through an undefined mechanism (10). The outcome of the UL18/NK interaction should thus depend on the prevalence of LIR-1. In a polyclonal setting, UL18 encounter may lead to either inhibition or activation of NK cells or, alternatively, have no effect at all whenever inhibitory signals are equalled by activating stimuli.
The studies on UL18 and NK cells, despite seemingly controversial results, could be reconciled as follows: if UL18 is expressed to a sufficient level at the cell surface, accomplished by transfection or the use of a suitable viral vector, UL18 is able to inhibit NK cell activity (10, 30, 31). The presence of LIR-1 is a prerequisite for the UL18-mediated inhibitory effect, as observed upon using LIR-1+ NK clones or LIR-1+ NK cell lines as effector cells. However, the inhibition may not be measurable when polyclonal NK cells with varying subsets of LIR-1-expressing cells are used instead (10). The situation becomes even more complex in HCMV infection, because UL18 is expressed at low levels and is often not detectable on the cell surface (9, 19). Furthermore, UL18 may activate certain NK cells, although neither the potentially involved receptors nor the details about the localization of UL18 are known (10, 29).
The effects of UL18 on T cells and DCs.
Although most research regarding the function of UL18 has focused on NK cells, the wide distribution of LIR-1 suggests that other cells may be a target for UL18 (3, 53, 54). We have therefore recently studied the effects of isolated UL18 proteins on monocyte-derived DCs. Our results disclosed substantial alterations in both phenotype and function of these DCs (55). Saverino et al. postulated that UL18 can activate CD8+ T cells in a LIR-1-dependent fashion, yet no mechanism explaining LIR-1-mediated activation is known (50). Our own results also favor an activating role for UL18, however, independent of a potential interaction with LIR-1 (9). We established that if the UL18 protein is accessible for ligation with LIR-1 on T cells, an efficient inhibitory signaling cascade unfolds. TCR-mediated activation is abrogated through the UL18-LIR-1 interaction, which in turn drastically reduces effector functions such as cytokine production (9). However, comparison of responses to AD169- and dUL18-infected cells points toward an activating function of UL18 through an undefined mechanism (9). Similar to the results emerging from studies about the effects of UL18 on NK cells, LIR-1 expression on a subset of T cells may not be sufficient to mediate physiologically relevant inhibition, especially when considering the low levels of surface-expressed UL18 during infection. The situation may be different for cell types such as macrophages, DCs, or B cells that ubiquitously express LIR-1 at high levels.
About the different functions of UL18.
UL18 seems to be able to exert both activating and inhibitory functions, depending on the receptor repertoire of the target cell population and possibly involving different localizations (surface expressed vs intracellularly expressed) (Fig. 3). The inhibitory function of UL18 mediated via LIR-1 is rather clearly defined, whereas the mechanisms of UL18-induced activation remain obscure. The situation certainly becomes even more complex in vivo, where UL18 may be found at lower levels on the surface of infected cells compared to experiments performed in vitro. Support for UL18 expression during active HCMV replication is provided by a study that found UL18 mRNA to be translated in patients with a high level of viremia (27) and by the fact that UL18-specific T cells can be detected (56, 57). In combination with the observation that all clinical HCMV isolates have retained the gene for UL18, these findings argue convincingly for a role of UL18 during natural infection. However, whether the protein mediates its function at an intracellular or extracellular location has not yet been assessed. The presence of UL18-specific T cells does not necessarily argue for surface-expressed UL18. Theoretically, T cells could be primed for UL18 by cross-presentation of DCs that have taken up infected cells expressing UL18 solely intracellularly. Yet, several arguments favor an extracellular role for UL18. Its extremely high affinity for LIR-1 combined with the fact this inhibitory receptor is so widely expressed on the surface of many types of immune cells makes UL18 an ideal candidate to exert immunosuppressive functions, particularly since HCMV infection is linked to increased expression of LIR-1 on lymphocytes (58, 59, 60). It seems rather unlikely that the virus has evolved such a powerful tool as the inhibition of immune responses via the UL18/LIR-1 interaction without actually making use of this option. Even though multiple variations were found in UL18 proteins from clinical isolates, mutations never prevented LIR-1 binding (30), supporting an in vivo role for the UL18-LIR-1 interaction. Furthermore, from a biochemical point of view the notion that UL18 contains more disulfide bonds than classical MHC-I could reflect the extracellular environment as the site of function. Proteins stabilized by disulfide bonds are often found extracellularly, where a high degree of structural stability helps to resist proteases and other challenges in a rather disruptive milieu compared with that of intracellular locations (61).
Most studies on the function of UL18 in immune responses are based on the assumption that UL18 acts as a surface-expressed protein on infected cells. More thorough investigations on the intracellular processing and possible interaction partners of this viral molecule may help to define alternative roles. Several viral immune-evasion proteins can exert multiple functions. Examples include the MCMV-encoded m152/gp40, which prevents MHC-I traffic beyond the cis-Golgi compartment and also impairs NK cell recognition by down-regulating a ligand for the activating receptor NKG2D (62, 63). The HCMV tegument protein pp65 prevents the generation of antigenic peptides and reduces NK-mediated cell lysis through inhibition of the activating receptor NKp30 (64, 65). The virion-associated protein pp71 enhances viral replication and is also suggested to hinder MHC-I transport (66). Similarly, UL18 may also exert more than one function. Some of the effects induced by UL18, such as the inhibition of T cells, NK cell, or DCs would favor viral immune escape, whereas activation may actually support host defense.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Abbreviations used in this paper: HCMV, human cytomegalovirus; β2m, β2-microglobulin; DC, dendritic cell; dUL18, deletion virus lacking UL18 gene; US, unique short.