Long and Short Isoforms of the Human Cytomegalovirus UL138 Protein Silence IE Transcription and Promote Latency

INTRODUCTION

Human cytomegalovirus (HCMV) is a betaherpesvirus that causes birth defects and disease in immunocompromised and immunosuppressed patients and has been associated with cancers, cardiovascular disease, and immune dysfunction (1–3). HCMV productively (lytically) infects differentiated cells, such as fibroblasts, endothelial cells, and epithelial cells, and latently infects incompletely differentiated cells of the myeloid lineage, such as monocytes and CD34⁺ hematopoietic progenitor cells (HPC) (4, 5). Productive infection in multiple cell types within the human host facilitates viral dissemination, while latency ensures lifelong persistence despite an effective immune response against lytic-phase antigens. Periodic reactivations of latent infections into the productive phase perpetuate both lifelong infection and virus spread to new hosts. No drugs that target latent reservoirs exist. Antiviral drugs targeting active replication are effective but have deleterious side effects and select for drug-resistant viruses (6).

The development of facile viral genetics (7) permitted the evaluation of viral sequences that are required for, augment, or are dispensable for productive infection of fibroblasts (8). While many of the mechanisms important to productive infection are well understood in fibroblasts (3), the mechanisms critical to latency remain poorly defined. However, in the last decade, an in vitro system to model experimental HCMV latency emerged, permitting genetic (9–15) and molecular (16–18) analyses of latently infected cells. During latency, the expression of productive-phase genes is suppressed. For this suppression to occur, the accumulation of the viral immediate early (IE) proteins that stimulate lytic infection must be controlled. IE protein accumulation is limited during latency by both transcriptional (19–21) and posttranscriptional (22) suppression. Transcriptional suppression of viral IE genes during latency is instituted, in part, by the viral UL138 protein (16).

HCMV UL138 resides within a clustered set of ULb′ genes (UL133, UL135, and UL136 genes) encoding proteins that all localize within secretory membranes, including the Golgi apparatus (12). Genetic analyses indicate that UL135 and some UL136 isoforms are required for efficient productive infection in endothelial cells (23–25) and that UL138, when expressed in the presence of UL133 and UL136 but in the absence of UL135, impairs efficient productive infection of fibroblasts (13). Notably, viruses deficient in UL133, UL138, or the soluble 23-/19-kDa isoforms of UL136 generate many infectious virus progeny in CD34⁺ hematopoietic cells, whereas wild-type viruses generate few such progeny (11, 12, 15). Molecular studies have revealed that UL135 and UL136 promote cytoplasmic maturation of viral particles in endothelial cells (23, 24), UL135 impairs immune synapse formation (26), and UL138 promotes cell surface expression of the tumor necrosis factor alpha (TNF-α) receptor (TNFR) (27, 28) as well as impairs cell surface expression of the multidrug resistance-associated protein-1 (MRP1) (18).

During latency, UL138 also promotes the suppression of IE1 gene transcription that is driven by the major immediate early promoter (MIEP). MIEP-directed transcription initiates productive replication in fibroblasts when the viral pp71 protein, a component of the virion tegument delivered upon infection, migrates to the nucleus and degrades Daxx, thereby counteracting histone deacetylase (HDAC)-mediated repression (29). When HCMV infects incompletely differentiated myeloid cells, in which it enters latency, tegument-delivered pp71 remains in the cytoplasm and IE1 transcription is suppressed (20, 21, 30, 31). For high-passage-number laboratory strains, the intrinsic defense instituted by Daxx may be the only significant restriction to IE1 transcription during latency because IE1 transcription from AD169 can be rescued in THP-1 and CD34⁺ cells (20, 21) by either Daxx knockdown or the HDAC inhibitors trichostatin A (TSA) or valproic acid (VPA). Unlike AD169, low-passage-number strains impose additional restrictions (21), one of which is enacted by UL138. UL138 impairs the recruitment of cellular lysine demethylases (KDMs) to the MIEP, thereby stabilizing repressive epigenetic histone modifications and silencing IE1 transcription (16).

Estimates of the protein-coding capacity of HCMV have substantially increased with recent studies (32, 33) and suggest a complexity in viral protein expression and likely function not previously appreciated. Genomic approaches have also cataloged previously undetected transcribed and translated sequences during both lytic and latent infections (34–37). The UL136 gene described above generates a series of coterminal protein isoforms with both overlapping and distinct functions (9, 15), illustrating the functional impact of previously unrecognized viral coding capacity. Here we show that UL138 gene also produces two coterminal proteins originating from separate in-frame methionine codons that support HCMV latency with different temporal phenotypes.

MATERIALS AND METHODS

RESULTS

Upon transfection of an expression plasmid carrying a UL138 allele with an added C-terminal hemagglutinin (HA) epitope tag, NHDF lysates contained a single HA-reactive band on Western blots (Fig. 1A) consistent with expression of full-length UL138 (here called pUL138-L for “long”). However, THP-1 lysates (Fig. 1A) contained approximately equimolar levels of this seemingly full-length protein and one of smaller size (here called pUL138-S for “short”). Both pUL138-L and pUL138-S were detected in NHDFs (Fig. 1B) infected with AD-138HA, a laboratory strain AD169-based recombinant (16) expressing C-terminally HA tagged UL138. The steady-state levels of pUL138-S were much lower than those of pUL138-L. We also detected both pUL138-L and pUL138-S in AD-138HA-infected THP-1 cells (Fig. 1C). Once again, pUL138-S accumulated to much lower levels than did pUL138-L. Furthermore, pUL138-S was also detected during infection of MRC5s or THP-1s with a TB40/E clinical strain in which the 3× Flag epitope tag was added to the C terminus of the endogenous UL138 gene (see Fig. 8). Our ability to visualize pUL13-S expression is frustratingly inconsistent (Fig. 1; see also below) due to the low-level expression of the isoform (expected for a latency protein), which hovers near our limits of detection. Various and sundry attempts to increase sensitivity (higher multiplicities of infection [MOIs], new virus preps, different time points, different lysis buffers and protocols, immunoprecipitation [IP]-Western, VPA treatment, stabilization with proteasome inhibitors) have failed to increase the frequency with which pUL138-S is detected (please see Discussion for more details on pUL138-S low-level expression). Nevertheless, because both the long and short forms of pUL138 can be visualized with antibodies to C-terminal epitope tags, and because disruption of the first translation initiation site results in increased accumulation of pUL138-S, we conclude that the UL138 gene is capable of producing two protein isoforms with identical carboxy termini.

The UL136 gene generates multiple protein isoforms through alternative transcription initiation and, likely, translation initiation from sequential in-frame methionine codons (9, 44). To evaluate the potential transcriptional mechanisms responsible for the expression of two pUL138 isoforms, we used 5′ rapid amplification of cDNA ends (RACE) over a time course of infection to detect any transcripts with transcriptional start sites (TSS) within the UL138 coding sequence (cds). Our analyses detected transcripts just upstream of the first methionine codon (M1) of UL138 gene but did not detect transcripts initiating within the UL138 cds (data not shown), consistent with previous data (11, 45) and suggesting that the two pUL138 isoforms may be encoded on every pUL138-specific transcript. The UL138 cds contains four in-frame methionine (M) codons (M1, M16, M40, and M134), the last three of which represent potential start sites for a smaller, coterminal protein (Fig. 2A). As proteins initiating at the final methionine (M134) would seem to be significantly smaller than the observed size of pUL138-S (Fig. 1), we focused a mutagenesis approach (Fig. 2B) on M16 and M40.

Converting codon 10, which normally encodes leucine (L), into a stop codon (L10stop), deleting the first 15 codons (Δ1–15), or deleting codon 1, normally encoding methionine (ΔM1), had no effect on the expression of pUL138-S in transfected THP-1 cells (Fig. 2C). However, converting codon 43, which normally encodes a tyrosine (Y), into a stop codon (Y43stop) abrogated the ability to detect pUL138-S (Fig. 2C), indicating that either M16 or M40 was the potential start site for pUL138-S translation. Because deleting the first 39 codons (Δ1–39) but retaining the M40 codon also abrogated the ability to detect pUL138-S (Fig. 2C), we suspected that translational initiation at M16 was responsible for generating pUL138-S. To directly test this prediction, we generated an allele (M16I) in which codon 16, normally encoding methionine, was changed to one encoding isoleucine (I). The M16I allele generated pUL138-L but not pUL138-S (Fig. 2D). We conclude that pUL138-L initiates from translation at codon M1 and that pUL138-S initiates from translation at codon M16. It remains possible that translational initiation at M40 or even M134 also generates protein. In fact, we have on occasion observed smaller HA-reactive bands on Western blots of lysates from UL138-HA-transfected cells, but those species have not been consistently observed or mapped.

The long form of the UL138 protein contains an amino-terminal signal sequence and a putative transmembrane domain (Fig. 2A) and is a membrane-localized resident of the Golgi apparatus (11). Methionine 16, where translation of the short form begins, is distal to the signal sequence and within the putative transmembrane domain. Therefore, pUL138-S is unlikely to be a transmembrane protein. During transient expression, both pUL138-L and pUL138-S localize to the Golgi apparatus in NHDFs (Fig. 3A) and in THP-1 cells (Fig. 3B), identified by costaining for the Golgi body resident protein (46) golgin A2/Golgi matrix protein (GM) 130 (additional images are provided in Fig. S1 and S2 in the supplemental material). In a portion of cells, pUL138-S also showed a more diffuse cytoplasmic localization. We conclude that both pUL138-L and pUL138-S localize to the Golgi apparatus during transient transfection.

Despite its localization in the Golgi body, UL138 has profound impacts on viral gene expression in the nuclei of latently infected myeloid cells (16), where it silences IE1 transcription by preventing the recruitment of cellular KDMs to the viral MIEP. As an initial test of the individual abilities of pUL138-L and pUL138-S to modulate MIEP function, we used a transient-transfection reporter assay in THP-1 cells. Transfection of different amounts of expression constructs encoding either wild-type UL138 (that makes both pUL138-L and pUL138-S), one encoding only pUL138-S (Δ1–15), or one encoding only pUL138-L (M16I) generated the predicted protein products to differing levels and allowed us to compare how well equivalent levels of protein isoforms could suppress a cotransfected MIEP reporter (Fig. 4). When roughly equivalent levels of proteins were compared, we found no statistical differences in the ability of wild-type UL138, pUL138-S, or pUL138-L to silence the MIEP. We conclude that both pUL138-L and pUL138-S repress an MIEP reporter in THP-1 cells.

A more physiologic test for viral transcriptional regulation requires targeted promoters to be in their native context (the viral genome) and modulating genes to be expressed at endogenous levels. We previously accomplished this by making a recombinant virus (AD-UL138HA) expressing the wild-type UL138 gene (with an additional C-terminal HA tag) from its native promoter within the context of the laboratory-adapted strain AD169 genome (16). Here we made two additional AD169-based recombinant viruses, one (AD-UL138HA-M16I) encoding only pUL138-L and another (AD-UL138HA-Δ1–15) encoding only pUL138-S (Fig. 5A).

AD-UL138HA-M16I and AD-UL138HA-Δ1–15 replicate with kinetics similar to those of wild-type virus in NHDFs (Fig. 5B). As shown earlier (Fig. 1B), wild-type virus expresses both long and short forms of UL138 upon infection of NHDFs (Fig. 5C) or MRC5s (Fig. 5D). AD-UL138HA-M16I expresses only pUL138-L, while AD-UL138HA-Δ1–15 expresses only pUL138-S in NHDFs (Fig. 5C) or MRC5s (Fig. 5D). As shown earlier, pUL138-L accumulates to greater levels during infection than does pUL138-S. In THP-1 cells, we detected pUL138-L made by both wild-type virus and AD-UL138HA-M16I but were unable to detect pUL138-S made by either wild-type or AD-UL138HA-Δ1–15 (Fig. 5E). That we could detect it under similar conditions in other experiments (Fig. 1C; see also Fig. 8D) underscores that the steady-state level achieved by pUL138-S is at or below our limit of detection. Nevertheless, we conclude that AD-UL138HA-M16I generates only pUL138-L and AD-UL138HA-Δ1–15 generates only pUL138-S. pUL138-L expressed by these AD169-based viruses localized to the Golgi apparatus in THP-1 cells (Fig. 6A) and primary CD34⁺ cells (Fig. 6B). We were unable to detect pUL138-S in THP-1 or CD34⁺ cells by indirect immunofluorescence after AD-UL138HA-Δ1–15 infection (Fig. 6A and B). However, when we infected VPA-treated THP-1 cells with AD-UL138HA-Δ1–15, we were able to visualize pUL138-S in a small number of cells, in which it localized to the Golgi apparatus (Fig. 6C).

Consistent with previous work, IE transcription in THP-1 (Fig. 7A) and CD34⁺ (Fig. 7B) cells was rescued by VPA addition upon infection with AD169 but not the AD-UL138HA that expresses both protein isoforms of UL138. VPA was unable to rescue IE transcription from AD-UL138HA-M16I genomes in either THP-1 (Fig. 7A) or CD34⁺ (Fig. 7B) cells, indicating that pUL138-L suppresses IE transcription in these assays. AD-UL138HA-Δ1–15 showed lower IE mRNA levels in the presence of VPA than did wild-type AD169 in THP-1 cells, indicating that pUL138-S also suppressed IE transcription (Fig. 7A). pUL138-S was not as strong of a repressor as pUL138-L or wild-type UL138 when expressed from the AD169 genome in THP-1 cells. Interestingly, the transcriptional repression associated with the presence of the UL138 Δ1–15 allele (Fig. 7A) indicates that pUL138-S has significant biological consequences at levels below the limit of detection on our Western blots.

In CD34⁺ cells, IE transcription from AD-UL138HA-Δ1–15 genomes was fully activated by VPA (Fig. 7B). At present, it is unclear if the difference between the ability of AD-UL138HA-Δ1–15 to repress VPA-induced IE transcription in THP-1 and CD34⁺ cells represents a cell type specificity for function or expression of pUL138-S (see Discussion). We conclude that pUL138-L silences IE transcription during the early stages of experimental latent infections (e.g., the first 18 to 24 h). While pUL138-S shares this ability, the magnitude of the effect is reduced and may show a dependence on cell type or whether latency was achieved naturally (in vivo) or experimentally (in vitro) (see Discussion).

UL138-null viruses made by either deletion of the entire open reading frame (14) or insertion of an in-frame stop codon (11) generate more infectious progeny than wild-type virus does after infection of CD34⁺ cells. Because the stop codon replaced methionine 16, this mutant (as well as the deletion virus) expresses neither pUL138-L nor pUL138-S. We therefore asked whether the long or short form of UL138 (or both) suppresses infectious virion formation during latency. Such experiments require the use of low-passage-number virus strains. Thus, we created TB40/E-UL138-3xFlag-based recombinant viruses encoding either an allele that expresses only pUL138-L (TB-UL138F-M16A) or an allele that expresses only pUL138-S (TB-UL138F-M1A) (Fig. 8A). TB-UL138F-M16A and TB-UL138F-M1A replicate with kinetics similar to those of wild-type virus in MRC5s (Fig. 8B) and synthesize UL138 proteins of the expected sizes in MRC5 (Fig. 8C) and THP-1 (Fig. 8D) cells.

As previous reports have demonstrated differences in localization between proteins expressed exogenously and those expressed in the natural context of low-passage-number virus strains (9), we addressed the localization of the two pUL138 protein isoforms during infection of MRC5s and THP-1 cells with TB-UL138F mutants. pUL138 colocalized with the Golgi body in addition to being distributed in puncta throughout the cytoplasm in both MRC5s (Fig. 9A) and THP-1 cells (Fig. 9B) infected with the parental TB-UL138F virus expressing both pUL138 isoforms. Colocalization of pUL138-S with the Golgi apparatus was not observed during infection with TB-UL138F-M1A in either cell type. Rather, pUL138-S exhibited a cytosolic distribution (Fig. 9A and B). In contrast, pUL138-L exhibited Golgi body and punctate cytosolic staining during infection with TB-UL138F-M16A similar to what was seen with the wild type in both cell types (Fig. 9A and B). As expected, no pUL138 staining was observed in either TB-UL138stop or mock infection (Fig. 9A and B). Together, these data suggest that during TB40/E infection of MRC5s, pUL138-L is Golgi body localized, while both isoforms are also distributed in the cytosol. As localization of pUL138-S during TB40/E infection differed from its localization in the context of transient expression (Fig. 3) or during recombinant AD-UL138HA infection (Fig. 6), it is possible that viral factors specific to clinical strains modulate the localization of pUL138-S. Given their demonstrated interactions with pUL138 (47), other proteins from the UL133-UL138 locus are candidates for this putative function.

Because TB40/E maintains an additional, unidentified restriction to viral IE1 transcription in undifferentiated myeloid cells treated with VPA (16), we cannot use IE mRNA levels as a measure of function for the short- or long-form proteins in this context. However, viruses lacking the UL138 gene make significantly greater numbers of infectious particles during experimental latent infections than do UL138-expressing viruses (11, 12, 14). Therefore, we asked if pUL138-S, pUL138-L, or both proteins could suppress productive replication during latency. We performed assays in ESCs because they provide a robust readout (16, 31) and in CD34⁺ cells because of their physiologic relevance (11, 12, 14). In ESCs infected with wild-type virus for 10 days, very few infectious progeny virions were produced, whereas infections with a UL138-null virus (M16stop) produced ∼15-fold-greater numbers of infectious virus (Fig. 10A). TB-UL138F-M1A showed an ability to suppress progeny virion formation during 10 days of experimental latency that was statistically indistinguishable from wild-type virus. In contrast, TB-UL138F-M16A produced more infectious virus than the wild-type but less than the UL138-null virus. From these data, we conclude that pUL138-S prevents infectious virion formation during experimental latency in ESCs as efficiently as wild-type UL138 and that pUL138-L also suppresses infectious virion formation during experimental latency in ESCs, but not as effectively as wild-type UL138.

In CD34⁺ cells (Fig. 10B), wild-type virus produced very few infectious centers while the UL138-null virus produced ∼8-fold more. TB-UL138F-M1A and TB-UL138F-M16A both produced quantities of infectious virus more similar to the quantities produced by a UL138-null virus than to those of a wild-type virus. Thus, while pUL138-L appears to be more potent than pUL138-S at silencing IE transcription at the beginning of latency (Fig. 7B), pUL138-S is more efficient than pUL138-L at suppressing replication, at least in the context of ESC infection (Fig. 10A). In total, we conclude that both the long and short forms of UL138 promote HCMV latency.

DISCUSSION

The work presented here maps and analyzes a novel isoform of the HCMV latency determinant UL138. No transcriptional start sites within the UL138 cds have been mapped (11, 33, 37), and inhibiting translation initiation at codon 1 by mutation does not preclude the production of pUL138-S (Fig. 2C). These data indicate that pUL138-S is likely to be expressed from the same transcript encoding pUL138-L. We mapped the translational start site of pUL138-S to the codon for M16 (Fig. 2), indicating that this isoform does not represent a cleavage or degradation product of the full-length (long) protein. Taken together, these data suggest that pUL138-S most likely arises via alternative translation mechanisms. Recently, lactimidomycin-stalled (initiating) ribosomes were detected within the UL138 gene at both codon 1 (∼93%) and codon 16 (∼7%), providing independent evidence for translation of a short form of pUL138 (33). An internal ribosome entry site (IRES)-like element has been previously shown to support translation of pUL138-L from transcripts with open reading frames upstream of the UL138 cds (45). Although this study did not examine the expression of pUL138-S, it is possible that the IRES-like element could also support expression of pUL138-S. When the methionine at codon 1 of UL138 is mutated (ΔM1 or M1A), the accumulation of pUL138-S increases (Fig. 2C and 8C), which is a result consistent with a leaky scanning mechanism of translation (48).

The differences in pUL138-S localization in AD169 and TB40/E infections could be due to the differences in the context of expression. In AD169 infection, expression of pUL138-S was achieved by deletion of the first 15 amino acids, whereas in TB40/E infection the first methionine codon was disrupted. These differences in mutagenesis approaches may contribute to changes in expression levels and may contribute to differences in the results obtained with the different strains. However, it should also be noted that pUL138 is expressed in the context of the UL133-UL138 locus in TB40/E infection. It is possible that expression of other UL133-UL138 proteins influences the function of pUL138 isoforms, as pUL133, pUL135, and pUL136 are known to interact with pUL138 (47). We also cannot exclude the possibility that other less well defined differences between AD169 and TB40/E influence pUL138 isoform localization and function.

Golgi body localization or retention signals within UL138 have not been demonstrated. The N-terminal signal sequence and putative transmembrane domain should be absent and disrupted, respectively, in pUL138-S, yet this protein can be found at the Golgi apparatus as well as the cytoplasm. There is some heterogeneity of pUL138-S localization when expressed in different contexts. Why this occurs is unclear but could result from, among other things, differences in expression levels, cis-acting mRNA sequences (49), or other viral proteins present during infection. pUL138-L interacts with a number of the other proteins encoded within the UL133-UL138 locus (47). While it is small enough to diffuse freely through nuclear pores, pUL138-S is not detected in the nucleus. This may be due in part to the low levels of the protein achieved in cells during infection. However, transient transfection results in the relatively abundant accumulation of pUL138-S but not its nuclear localization. It remains to be determined how UL138 exerts transcriptional effects in the nucleus from its position at the Golgi apparatus and in the cytoplasm.

pUL138-S is expressed at low levels during infection and accumulates to much lower levels than pUL138-L, making it difficult to detect. In our reporter assays (Fig. 4), we needed to transfect three times the amount of Δ1–15 encoding plasmid to achieve protein levels similar to those generated by wild-type or M16A alleles. Viral proteins in persistent infections, for example, the HCMV UL136 protein and the human papillomavirus E7 protein, are often found at low levels, presumably to limit immune detection (9, 50, 51). Low-level protein accumulation may result from weak transcription, inefficient translation, codon usage, poor protein stability, or a combination of these factors. For example, the UL138 gene in TB40/E has a codon adaption index (CAI) (52) of 0.7 (GenScript), which is considered suboptimal for expression. Intriguingly, it may also result from microRNA (miRNA)-mediated translational repression, as a viral microRNA, miR-UL36, has been reported to downregulate UL138 (53) and is expressed during both experimental and natural latent infections (54). Regulation by cellular miRNAs is also possible, as has been detected for viral IE1 transcripts in latently infected cells (22, 55). The very low levels of UL138 proteins detected are consistent with viral attempts at immune evasion during latency, as UL138 possesses known cytotoxic T-cell-specific epitopes (56). Interestingly, the identified CD8⁺ T cell epitope is from the amino terminus of the protein found in pUL138-L but not pUL138-S (56).

Reporter assays (Fig. 4) and quantitation of IE transcripts in the presence of VPA (Fig. 7) indicate that the long form of UL138 may play a more prominent role than the short form at suppressing IE gene expression during latency. In contrast, assays in ESCs (Fig. 10A) indicate that the short isoform better suppresses infectious virion production during latency than does the long form. These subtle differences in function may reflect a time dependence in activity. One possible model is that early and prominent accumulation of pUL138-L silences IE transcription at the beginning of latency prior to the protein being downregulated. At later times, weaker expression of pUL138-S might suppress progeny virion production without resulting in the presentation of UL138 epitopes unique to the long form and recognized by CD8⁺ T cells (56). Interestingly, C-terminal peptides found in pUL138-S elicit a CD4⁺ T cell response that includes secretion of the immunosuppressive cytokine cellular interleukin-10 (cIL-10) (57). Thus, pUL138-S might promote latency by a combination of transcriptional repression of lytic-phase genes and by protecting latently infected cells from immune detection and clearance.

We were unable to detect the suppressive functions of pUL138-S observed in THP-1s or ESCs in the context of CD34⁺ HPCs. The cell-type-dependent differences may reflect different kinetics of isoform synthesis or accumulation, cell type specificity for function, or a combination of these features. Interestingly, an unbiased examination of HCMV latent transcripts (36) demonstrated that the fraction of all mapped HCMV latent messages represented by UL138 was ∼5-fold higher during experimental infection of monocytes than during natural infection of these cells but ∼15-fold lower in experimental latent infection of CD34⁺ cells than in natural infections. Thus, it appears that UL138 is expressed better in monocytes in vitro and better in CD34⁺ cells in vivo, at least at the transcript level. The low levels of UL138 expression in in vitro systems may contribute to the difficulty in discerning functional outcomes of pUL138-S in CD34⁺ cells.

We describe pUL138 as silencing transcription because it clearly represses an HCMV promoter, the MIEP. pUL138 decreases luciferase enzymatic activity when the gene encoding the protein is driven by the MIEP but not when it is expressed from the simian virus 40 (SV40) promoter (16), making posttranscriptional regulation of this reporter mRNA unlikely. Furthermore, pUL138 expression has been shown to modify the chromatin structure of the MIEP in a manner correlated with transcriptional repression (16). Thus, despite its cytoplasmic localization, pUL138 modulates viral transcription. However, because of its cytoplasmic localization, supplementary effects on mRNA stability or translation are not excluded by any of our data. More work is required to determine if pUL138 regulates latent viral gene expression posttranscriptionally in addition to its transcriptional effects.

While UL138 has been associated with multiple effects during HCMV infection (13, 16, 18), previous studies did not take into account the individual roles of the pUL138 protein isoforms. Our current study demonstrates that the two protein isoforms of pUL138 may have both overlapping and distinct functions during HCMV infection. In recent years, many studies have contributed to a body of work that begins to effectively demonstrate the power and function of increased coding capacity in HCMV through transcriptional, translational, and posttranslational mechanisms (9, 15, 32, 33, 37, 58, 59). Further, many examples of viral protein isoforms that have similar or unique functions exist (60–67), but these examples likely only begin to scratch the surface of the important roles that protein isoforms play in viral infection and virus-host interactions. More work is needed to fully understand the functions of the two pUL138 isoforms and how their functions, in cooperation with the UL133-138 locus, contribute to the persistence of HCMV within the human host.

Supplementary Material