Human MX2/MxB: a Potent Interferon-Induced Postentry Inhibitor of Herpesviruses and HIV-1

MX proteins.

MX proteins are found in all vertebrates, including birds and fish (11). They are IFN-induced dynamin-like GTPases that consist of a globular GTPase domain connected to an extended stalk via a flexible bundle signaling element (BSE) (Fig. 1). Mammals usually contain two or more MX genes that arose by ancient gene duplications, conversions, and deletions (11–13). The human MX1 gene codes for a cytoplasmic protein designated MxA that restricts the growth of a surprisingly large panel of positive- and negative-strand RNA and certain DNA viruses but, notably, not herpesviruses or HIV-1 (reviewed in reference 14). The mechanism of MxA antiviral action is incompletely understood. A fraction of MxA is associated with a subcompartment of the endoplasmic reticulum (ER) (15–17), and MxA appears to interfere with proper trafficking of viral components, in particular viral nucleoproteins, in infected cells. For example, MxA was shown to trap incoming Thogoto virus nucleocapsids, thereby interfering with transport of the viral genome into the nucleus (18). In the case of La Crosse bunyavirus, MxA associates with viral nucleocapsids and sequesters them into perinuclear complexes (19). In the case of hepatitis B virus (HBV), MxA interacts with the viral core antigen (HbcAg) and blocks proper assembly of capsids (20). Influenza A virus variants which escape MxA-mediated inhibition carry distinct signature mutations in their nucleoprotein (21). Nevertheless, attempts to demonstrate direct interactions between MxA and influenza A virus nucleoproteins remained elusive (22). Thus, it is unclear whether the observed or suspected interactions of MxA with viral components are direct or whether they require additional cellular factors.

The MX2 gene of humans codes for full-length MxB and an N-terminally truncated variant [MxB_(26–715)] (23) that lacks a nuclear localization signal (NLS) and is abundantly produced in IFN-stimulated cells (24) (Fig. 1A). The three-dimensional (3D) structure of MxB (Fig. 1B) resembles that of MxA (25, 26), except that MxA lacks the NLS-containing N-terminal domain present in MxB. Biologically active MxB probably is a noncovalent antiparallel dimer (Fig. 1C), but it can also form higher-order oligomeric structures (26, 27), like MxA (14, 28). Initial studies suggested that MxB, in contrast to MxA, has no antiviral activity but instead regulates the nucleocytoplasmic transport of cellular factors (23, 29, 30). Dominant-negative MxB mutants defective in GTP binding (K131A) or GTP hydrolysis (T151A), as well as an MxB deletion mutant lacking the NLS motif, disrupted nuclear import of certain reporter proteins (30). It was further shown that full-length MxB shuttles between the cytoplasm and the nucleus (24) and localizes to the cytoplasmic face of nuclear pore complexes (30).

Human MxB is a pan-herpesvirus restriction factor.

A first hint that MxB might interfere with the replication of herpesviruses came from a screen for IFN-induced genes conferring resistance to murine gammaherpesvirus 68 (MHV68) (31). Crameri et al. (8) reported that small interfering RNA (siRNA)-mediated knockdown of MxB renders IFN-α-treated human T98G glioblastoma cells partially susceptible to herpes simplex virus 1 (HSV-1). Correspondingly, overexpression of MxB in human A549 epithelial cells provided a high degree of resistance to various HSV-1 strains and to HSV-2 (belonging to the Alphaherpesvirinae subfamily), as well as to Kaposi’s sarcoma-associated herpesvirus (KSHV), a representative of the Gammaherpesvirinae. However, MxB-expressing A549 cells remained susceptible to unrelated viruses, such as influenza A virus (IAV) and human adenovirus 5, and susceptibility to vesicular stomatitis virus (VSV) was reduced only marginally (8). Additional experiments showed that the accumulation of HSV-1 early gene products was strongly reduced in MxB-overexpressing A549 cells compared with that in control cells, and that MxB inhibited the delivery of the HSV-1 genome into the nucleus. To identify the MxB-sensitive step of the HSV-1 replication cycle, Crameri et al. first asked whether MxB inhibited the early dissociation of tegument proteins from the viral capsid, but found that this was clearly not the case. They then used a combined electron microscopy, immunofluorescence, and click chemistry approach to determine whether MxB might interfere with the trafficking of incoming HSV-1 capsids to the nuclear envelope or the transfer of the viral genomes into the nucleus. Indeed, they observed fewer viral capsids arriving at the nuclear pore complexes of MxB-expressing A549 cells compared with those arriving at A549 control cells, and they noted strongly reduced nuclear import of incoming viral DNA (8). These results suggest that MxB interferes with the delivery of herpesviral genomes into the nucleus. Using transmission electron microscopy, they could show that DNA-containing viral capsids accumulated in the cytoplasm of MxB-expressing cells, whereas in control cells lacking MxB, the majority of viral capsids observed in the cytoplasm and at nuclear pore complexes were empty (8). Additional mechanistic studies demonstrated that MxB mutants lacking the N-terminal NLS or lacking GTPase function lost antiviral activity, indicating that both proper intracellular localization and GTP hydrolysis are important for the antiherpesvirus activity of MxB.

A second recent report, by Schilling et al. (9), supports and extends the concept that MxB is a panherpesvirus restriction factor. These authors generated human U87MG glioblastoma cell lines stably overexpressing either full-length MxB, N-terminally truncated MxB_(26–715), or MxB devoid of GTPase activity. They found that full-length MxB inhibited the growth of several herpesviruses, including HSV-1 and MHV68, as well as murine and human cytomegaloviruses (9). No inhibitory effects were observed in cell lines expressing MxB_(26–715) or GTPase-deficient MxB. Likewise, monomeric MxB carrying an M574D substitution that prevents self-assembly via the stalk interface was inactive, indicating that dimerization and perhaps oligomerization of MxB is required for antiviral activity. The growth of viruses unrelated to herpesviruses, including IAV, VSV, human adenovirus 5, and vaccinia virus, was not significantly affected (9), in agreement with Crameri et al. (8). To evaluate the contribution of MxB to herpesvirus resistance in IFN-α-treated cells, human T98G glioblastoma cell lines lacking functional MX2 genes were generated by clustered regularly interspaced short palindromic repeat (CRISPR)-Cas9 knockout technology. IFN-α failed to induce robust HSV-1 resistance in such cells (9), demonstrating that MxB is a major mediator of herpesvirus resistance in IFN-treated cells. Intriguingly, an engineered chimeric MxB/MxA protein carrying the first 85 N-terminal amino acids of MxB was sufficient to confer antiherpesvirus activity to human MxA, which, as mentioned, is not intrinsically active against herpesviruses. The chimeric protein also gained anti-HIV-1 activity, as previously reported (32). In contrast, a triple arginine motif near the N terminus of MxB that is required for anti-HIV-1 activity (33) was found to be dispensable for herpesvirus restriction (9). These results reveal apparent similarities and differences in the underlying mechanisms by which MxB inhibits herpesviruses and lentiviruses (see below).

A third report, by Jaguva Vasudevan et al. (10), demonstrates that MxB is a potent inhibitor of murine cytomegalovirus (MCMV), a member of the Betaherpesvirinae subfamily. Human 293T cells overexpressing either full-length or mutant MxB were analyzed in this study. After infection with MCMV, cells expressing full-length MxB showed decreased levels of nuclear viral genomes, resulting in strongly reduced expression of viral immediate early genes relative to control cells. MxB required the NLS and an intact GTPase domain for its anti-MCMV activity (10), whereas the latter domain was dispensable for anti-HIV-1 activity. Interestingly, no interaction of MxB with MCMV capsid proteins could be demonstrated in assays that readily revealed MxB binding to HIV-1 gag gene products, again pointing at mechanistic differences between how MxB acts against herpesviruses and HIV-1. Taken together, available results indicate that MxB interferes with the delivery of viral genomes into the host cell nucleus by targeting an early postentry step in the replication cycle, which is highly conserved among members of the Herpesviridae family (Fig. 2A).

Human MxB inhibits HIV-1.

In 2013, three research groups reported independently that full-length MxB is an inhibitor of HIV-1 (34–36). Overexpression of MxB resulted in strong inhibition of authentic and moderate inhibition of VSV-G-pseudotyped HIV-1 (36, 37), demonstrating that inhibition was independent of receptor usage. siRNA-mediated silencing of MxB partially rescued HIV-1 growth in IFN-treated cells, revealing that MxB plays a role in the IFN-mediated restriction of HIV-1. Mechanistic studies showed that MxB had no inhibitory effect on viral minus-strand cDNA synthesis but reduced the levels of 2–long terminal repeat (2-LTR) circles (a marker of viral cDNA nuclear localization) in infected cells. A logical conclusion was that nuclear uptake of the viral preintegration complex (PIC) was inhibited by MxB (Fig. 2B).

Distinct amino acid substitutions in the HIV-1 capsid protein CA resulted in escape from MxB inhibition (35, 36, 38, 39). Escape mutations were mapped to different domains of CA, indicating that the MxB restriction mechanism is most likely multifactorial and relies on multiple interactions. MxB can bind to HIV-1 capsids, but binding does not strictly correlate with virus inhibition (24–26, 35, 39–41). Apparently, binding to the HIV-1 capsid is necessary but not sufficient for antiviral activity. Some MxB escape mutations map to a flexible loop of the capsid (34) that is recognized by the peptidyl-prolyl isomerase cyclophilin A (CypA), a host factor that enhances HIV-1 infection by guiding viral capsids to nucleoporins (Nups) and facilitates nuclear import of the viral genome (39, 42). When the interaction between CypA and HIV-1 was disrupted by cyclosporine A, MxB lost most of its anti-HIV activity (29, 34), indicating that MxB might somehow interfere with the function of CypA.

The situation is presumably even more complex; different Nups are involved in the nuclear import of the HIV-1 genome in different cell types or during different stages of the cell cycle, and the anti-HIV activity of MxB seems to depend on which Nup is involved (29). Interestingly, by manipulating the levels of different Nups, both enhancing and inhibiting effects of MxB on various HIV-1 capsid mutants were observed (29). Thus, depending on which nuclear entry pathway incoming viral capsids choose, MxB is either a potent or a weak restriction factor. Of note, MxB can similarly control the nuclear import of host cell proteins that use the same pathways as HIV-1 (29).

Natural HIV-1 isolates usually exhibit MxB sensitivity (34–36), although there are some exceptions (40), suggesting that capsid alterations allowing escape from MxB may cause fitness loss. Overall, a picture emerges indicating that MxB inhibits HIV-1 by interfering with the function of a subset of Nups and possibly that of additional components of the host cellular machinery, such as CypA, involved in the nuclear import of the viral genome (Fig. 2B).

It is unclear, however, whether nuclear pore complex localization of MxB is essential for antiviral activity or whether MxB simply accumulates at the nuclear pore because Nups accumulate at this site (29). In fact, critical interactions between viral capsids, Nups and MxB might take place earlier, somewhere on the way of the viral capsids through the cytoplasm to the nuclear pore.

Crystallized MxB forms antiparallel dimers that are critical for HIV-1 restriction, and modeling suggests that the stalk domains of MxB provide the appropriate spacing for efficient interaction with hexamer interfaces in the HIV-1 capsid lattice (25–27) (Fig. 1C). The N-terminal domain of MxB, which is missing in isoform MxB_(26–715), is required for restriction of HIV-1 (34–36). This domain contains an NLS, which contains a lysine and a tyrosine at amino acid positions 20 and 21, respectively, and a triple arginine (RRR) motif at positions 11 to 13 that is essential for HIV-1 capsid binding (24, 33) and engages in interactions with Nups and the nuclear transporter TNPO1 (29) (Michael H. Malim, personal communication). Transfer of the N-terminal 91 amino acids of MxB converted MxA into a potent HIV-1 restriction factor without affecting its antiviral activity against influenza virus (32). Similar results were obtained when the N-terminal domain of MxB was fused to unrelated proteins, such as Fv1 or SAMHD1 (24, 33), demonstrating that the N-terminal domain is the only element of MxB that is necessary for HIV-1 inhibition in the context of a heterologous fusion partner. The N-terminal domain of MxB conferred anti-HIV activity to fusion partners only if the partner was able to form dimers or trimers (33), confirming the conclusions that dimeric (or perhaps oligomeric) forms of MxB confer antiviral activity and that monomeric MxB lacks antiviral activity (43).

The MxB mutant T151A, which binds but cannot hydrolyze GTP (30), showed normal activity against HIV-1 (35, 36). This was surprising, as the known functions of other members of the dynamin-like GTPase family all depend on GTP hydrolysis. A notable exception is the MxA-mediated inhibition of hepatitis B virus (44), which depends on a direct interaction of the viral capsid with a poorly defined interaction domain located in the stalk of MxA and which appears to function autonomously. In contrast, GTP hydrolysis is essential for the antiviral activity of MxA against IAV (45, 46), Thogoto virus (47) and bunyaviruses (48), and, as discussed above, for the antiviral activity of MxB against herpesviruses (8–10). Thus, MxB, as well as MxA, appears to employ at least two fundamentally distinct mechanisms by which they can restrict virus growth, and these mechanisms differ with regard to GTP hydrolysis.

Common and distinct features of MxB antiviral activities.

Evolutionary analyses of mammalian MX proteins revealed that the N-terminal tail is the most divergent region of human MxB. It evolved under diversifying selection, suggesting that this structurally flexible part of the molecule was engaged in a long-standing host-virus genetic conflict during primate evolution (11). The data discussed above revealed that the N-terminal domain of human MxB is essential for antiherpesvirus and antilentivirus activity. This domain directs the protein to the nuclear envelope with the help of its NLS. Moreover, it carries specific antiviral determinants required for inhibiting herpesviruses (8–10) and HIV-1 (24, 32, 38, 49). Not surprisingly, the amino acids in the N-terminal tail of MxB that dictate the antiviral specificities for the two unrelated viruses are not the same and must have been selected independently. Yet neither the amino acids that govern antiherpesvirus specificity nor those that confer anti-HIV-1 specificity overlap sites found to be positively selected in evolution, suggesting that diversification of the N-terminal tail might have been driven by other, as yet unknown pathogens (8, 9, 11). Be that as it may, the N-terminal domain of MxB serves as an autonomous module that mediates antiviral activity against HIV-1 (24, 32) and herpesviruses (9) when fused to partner proteins that can form stable dimers. Thus, the N-terminal domain of MxB is functionally comparable to loop L4 in human MxA (50, 51), yet these two functional domains reside at opposite ends of the molecules (26, 28).

The mechanistic details of how MxB inhibits herpesviruses and HIV-1 are still poorly understood. A common feature seems to be recognition and trapping, missorting and/or affecting the stability of viral capsids on their way to the nuclear entry port (Fig. 2). However, how HIV-1 and herpesvirus capsids are targeted by MxB is not entirely clear. MxB binds to the HIV-1 capsid protein, and a few amino acid substitutions in the viral capsid allow partial escape from MxB restriction (38). Binding seems to be a prerequisite but it is not sufficient for restriction (29, 38). In contrast, comparable interactions with herpesvirus capsid proteins cannot so far be demonstrated (10), and no MxB escape mutations have been identified (26, 27). Various host factors, such as Nups and CypA, appear to be involved in the interaction of the HIV-1 capsid with MxB (29, 34). It is tempting to speculate that host factors also mediate the interaction of MxB with components of the herpesvirus capsids. Another common characteristic is a requirement for MxB oligomerization. Dimers or lower-order oligomers are necessary for HIV-1 inhibition (26, 27). Similar oligomerization requirements were found for inhibition of herpesviruses (9, 10).

Despite commonalities, a striking functional dissimilarity is also evident: GTP binding and hydrolysis are essential for the antiviral effect of MxB against herpesviruses but not for the antiviral effects against HIV-1. At present, the role of GTPase activity in viral target recognition and viral inhibition is not well understood. In the case of MxA, it has been proposed that oligomerization of the molecule on viral substrates activates the enzyme and that nucleotide hydrolysis induces a conformational change in the oligomer. These molecular movements would result in a power stroke, which could mediate the antiviral effect (46, 52–54). On the other hand, antiviral effects of MxA against hepatitis B virus do not require GTPase activity (20). Thus, clearly, MxB, as well as MxA, can exert antiviral functions with or without a GTP substrate, depending on the type of virus and perhaps the host cell.

Open questions.

In spite of exciting recent progress, the precise mechanism by which MxB interferes with herpesvirus replication is presently not understood. In analogy to what we currently know from work on HIV-1, it appears likely that MxB binds to a highly conserved structural component of the viral capsid, which presumably becomes accessible only after limited dismantling of viral particles in the cytosol of infected cells. The complex herpesvirus uncoating process involves stepwise shedding of a large variety of tegument proteins away from the DNA-containing viral capsid, with a few critical tegument proteins remaining associated until the capsid has docked to the nuclear pore (reviewed in reference 55). Hence, a large number of different viral proteins are potential targets of MxB. Some recent straightforward attempts to detect physical interactions between MxB and herpesvirus proteins did not yield any hits (10) (Jovan Pavlovic, personal communication). Thus, it remains possible that such interactions are too dynamic or too weak for detection by conventional methods. More sophisticated technologies may eventually identify the herpesvirus target(s) of MxB.

Alternatively, the antiherpesvirus activity of MxB may not involve any direct interactions with viral components but may rather result from an inhibitory effect of MxB on some host cell pathways that are important for herpesvirus replication. In fact, recent work revealed that MxB inhibits nuclear import of various host cell proteins if they carry nonconventional NLS sequences, indicating that particular nuclear import pathways are more vulnerable to blockage by MxB than others (29). Thus, it is conceivable that herpesviruses are sensitive to MxB simply because they prefer to use a particular pathway for nuclear import of viral DNA that is efficiently blocked by MxB.

Another open question is why GTPase activity of MxB is required for inhibition of herpesviruses but not that of HIV-1. It should be noted, however, that this surprising difference is not a general principle. The equine MX2 orthologue differs from human MxB in that it does require GTPase activity to inhibit a range of lentiviruses, including HIV-1, in assays using VSV-G-pseudotyped luciferase reporter vectors (56), indicating that the mode of MxB/MX2 action may differ between species.

Another interesting yet unresolved issue is the relationship between the two MxB isoforms that differ from each other by a 25-amino-acid extension at the N terminus. The N-terminally truncated MxB_(26–715) is abundantly expressed, yet it has no known antiviral activity and its biological role is presently unclear (8–10). Mx proteins have a propensity to oligomerize (26, 28), and it was previously shown that MxB_(26–715) readily forms heterodimers with full-length MxB (57). What is the role, if any, of such heterodimers? Does the short isoform somehow modulate the activity of full-length MxB? Could the short isoform perhaps alleviate the negative effect of full-length MxB on host cell protein nuclear import and thereby limit cellular damage or even counteract antiviral activity? We should be open to surprises. It is well known that herpesviruses have evolved sophisticated ways to modulate the host immune response and to prevent the induction and function of interferons, but it will be important to know whether herpesviruses possess specific means to directly antagonize the MxB antiviral mechanism. In fact, a novel isoform of MxA is translated from an alternatively spliced transcript in HSV-1-infected cells that appears to favor rather than inhibit herpesvirus replication (58). The identification of a viral antagonist might help to define the function of MxB more precisely and to eventually provide a target for therapeutic antiviral interventions.

Finally, what is the clinical relevance of these findings? The role of MxB remains to be defined in clinical settings of primary and recurrent virus infections. To what extent does MxB contribute to innate immune surveillance in herpesvirus latency or during reactivation? It will be important to determine whether some of the sporadic cases of enhanced susceptibility to herpesvirus infections in children (7) are due to genetic alterations that negatively affect MX2 gene expression or MxB action. Allelic variants in the coding region of the human MX1 gene are rare (59), and the same appears to hold for the MX2 gene (Laura Graf and Georg Kochs, personal communication). Nevertheless, some rare allelic variations might code for MxB variants, with dominant negative effects on full-length MxB. Unfortunately, we cannot expect that some of these important issues will be solved quickly using standard mouse models. The reason is that the MX2 orthologue was deleted during rodent evolution, and the Mx2 gene of present-day mice and rats is the result of an ancient MX1 gene duplication and codes for an MxA-like GTPase (11). Accordingly, mouse MX2 has no antiviral activity against MCMV (10). Transgenic mouse models similar to the one used for recent MxA research (60) might be useful. In spite of these technical challenges, we hope that the exciting new findings discussed in this review article will soon lead to a better understanding of herpesvirus control by the innate immune system.