The Myeloid Transcription Factor GATA-2 Regulates the Viral UL144 Gene during Human Cytomegalovirus Latency in an Isolate-Specific Manner

Abstract

It is generally accepted that, following primary infection, human cytomegalovirus (HCMV) establishes lifelong latency in CD34⁺ progenitor cells and other derivative cells of the myeloid lineage. In this study, we show that the viral UL144 gene is expressed during latent infection in two cell types of the myeloid lineage, CD34⁺ and CD14⁺ monocytes, and that the UL144 protein is functional in latently infected monocytes. However, this latency-associated expression of UL144 occurs only in certain isolates of HCMV and depends on the presence of functional GATA-2 transcription factor binding sites in the UL144 promoter, in contrast to the viral latency-associated gene LUNA, which we also show is regulated by GATA-2 but expressed uniformly during latent infection independent of the virus isolate. Taken together, these data suggest that the HCMV latency-associated transcriptome may be virus isolate specific and dependent on the repertoire of transcription factor binding sites in the promoters of latency-associated genes.

INTRODUCTION

Human cytomegalovirus (HCMV) is a species-specific betaherpesvirus and one of the largest viruses to infect humans, with a coding capacity of 150 to 180 open reading frames (ORFs). The prevalence of the virus varies geographically, with between 50 and 90% of the population carrying the virus (1). Like all herpesviruses, HCMV has two phases of life cycle: lytic or productive infection is characterized by the temporal transcription of immediate early (IE), early, and late viral genes with ensuing virus production, whereas latent infection, often established in CD34⁺ progenitor cells and cells of the myeloid lineage (2–5), is associated with the expression of only a few viral genes and the absence of production of virions. Unless latent infection is to be a dead-end for the virus, these sites of latency must be able to support virus reactivation, and much evidence now points to terminal differentiation of myeloid cells to macrophages and dendritic cells as a key signal for virus reactivation (4, 6–13).

Although our understanding of the latent HCMV transcriptome is far from complete, it is clear that during experimental latent infection viral gene expression is quite limited (2, 8, 14–16). Importantly, several HCMV transcripts shown to be expressed during experimental latent infection, including an alternate form of viral interleukin-10 (IL-10), i.e., UL111a (17), UL138 (18), and LUNA (19), have also been identified in naturally latently infected cells. To date, with the exception of the UL111a latent isoform (17), there is no evidence that any HCMV immune-modulatory proteins expressed during lytic infection are commensurately expressed during latency (20–23). While the highly restricted transcriptional profile during latency will almost certainly reduce the potential for host recognition of infected cells, it is highly likely that latency-expressed genes will need to function to both maintain the viral genome and restrict host immunity (24). Consequently, elucidating the functions of latency-associated viral genes is crucial to further our mechanistic understanding of HCMV persistence.

One region of the HCMV genome, termed UL/b′, encodes approximately 19 ORFs (UL133 to UL151, here referred to as UL133-151) (25) and is prone to deletion in viral strains extensively passaged in fibroblasts (e.g., strains AD169 and Towne). Several proteins encoded in this region have been established to perform immune-modulatory functions (26–32). Recently, this genomic locus has also been implicated in the regulation of experimental latency (18, 33). Using an experimental model in primary CD34⁺ cells, it has been reported that HCMV strains lacking the entire UL/b′ region are unable to establish latency, with a commensurate increase in lytic infection also being observed (18). Notably, although the UL133-138 region ORFs were shown to be required for latency establishment, the UL144-148 region was also implicated for optimal latency (18), suggesting that multiple UL/b′ region genes may regulate this process.

Tumor necrosis factor (TNF) family cytokines are critical regulators of HCMV immune defenses (34, 35). The latency-expressed UL138 ORF has recently been shown to modulate cell surface expression of TNF receptor 1 (26, 27), suggesting that TNF signaling may impact HCMV latency, as has been shown for mouse CMV (36). A second connection between the TNF family and UL/b′ region proteins is highlighted by the UL144 ORF, which shows high homology to the herpesvirus entry mediator (HVEM/TNFRSF14) (37, 38). UL144 is expressed early in lytic infection, encodes a transmembrane glycoprotein with a short intracellular cytoplasmic tail, and is highly variable in sequence between clinical isolates of HCMV (37, 39). During lytic infection, UL144 upregulates CCL22 via activation of NF-κB (28–30). CCL22 is a chemoattractant for Th2 and T regulatory cells (40), and consequently UL144 may dampen Th1-mediated immunity. Additionally, UL144 binds the B and T lymphocyte attenuator (BTLA) and potently inhibits T cell proliferation (41). Intriguingly, UL144 has lost the ability to bind to the HVEM ligand LIGHT, highlighting the selective evolution of this viral protein to target specific components of the HVEM-LIGHT-BTLA signaling network (42). On this point, nothing has been known regarding the potential function(s) of UL144 in regulating HCMV latency. However, its pleiotropic functions as an immune-modulating protein, several of which are reminiscent of the latent membrane protein 1 (LMP1) of Epstein-Barr virus (EBV) (40, 43, 44), suggest it as a potential candidate.

In this study, we have examined whether UL144 is expressed during natural latency in CD14⁺ monocytes, as well as in two experimental models of HCMV latent infection. We show that UL144 is expressed in monocytes of healthy seropositive donors as well as in experimentally latent monocytes and CD34⁺ myeloid progenitors but that this is in an HCMV isolate-specific manner. Analysis of the UL144 promoter regions from various HCMV strains showed that expression during latency is regulated by GATA-2 and is completely correlated with the presence or absence of GATA-2 binding sites. This contrasts with the HCMV latency-associated gene LUNA, which we also show is regulated by GATA-2 but expressed ubiquitously during latent infection, independent of the virus isolate. As we have previously shown that the reduction in the cellular microRNA (miRNA) hsa-miR-92a during latent infection of CD34⁺ progenitor cells results in enhanced levels of GATA-2 (45), we now conclude that a functional consequence of this is to regulate expression of both UL144 and LUNA. Thus, our data argue that during latent infection and in the absence of immediate early gene expression, the myeloid transcription factor GATA-2 plays a key role in the differentiation-dependent regulation of latency-associated transcription.

MATERIALS AND METHODS

Ethics statement.

Ethical permission for this project was granted by the Cambridgeshire 2 Research Ethics Committee (REC reference 97/092). Written consent was obtained from all of the volunteers included in this study prior to providing blood samples for the preparation of primary CD14⁺ monocytes.

Primary CD34⁺ hematopoietic progenitors were from normal hematopoietic stem cell transplant donors after stem cell mobilization (Lonza AG) in accordance with the Declaration of Helsinki.

Viruses and cells.

Primary monocytes were isolated from venous blood using CD14⁺-positive selection (Miltenyi Biotech). In brief, fresh venous blood was diluted 1:2 with phosphate-buffered saline (PBS) and peripheral blood mononuclear cells (PBMCs) isolated by Ficoll-Hypaque density gradient centrifugation (Axis-Shield, Norway) as described previously (46). PBMCs were removed from the gradient and washed twice in PBS before the addition of CD14⁺ direct beads (Miltenyi Biotech). CD14⁺ cells conjugated to antibodies were isolated from total PBMCs by positive selection using the MACS magnetic bead system and LS columns according to the manufacturer's instructions (Miltenyi, United Kingdom). Alternatively, primary CD34⁺ cells were commercially sourced (Stemcell). The THP1 cells (human acute monocyte leukemia cell line) were obtained from ATCC and were maintained in RPMI 1640 with 10% fetal bovine serum (FBS), 2 mM l-glutamine, and 100 units/ml penicillin and streptomycin.

Low-passage-number strains Merlin and TB40E have been described previously (47). Low-passage-number FIX virus was a kind gift of Finn Gray (Edinburgh, United Kingdom), and clinical AF1 virus was a kind gift from Andrew Davison (Glasgow, United Kingdom).

Natural and experimental HCMV latency in monocytes and CD34⁺ cells.

For experimental-latency studies, monocytes or CD34⁺ cells were infected for 6 days to achieve latency and then differentiated and matured into dendritic cells (DCs) for reactivation, where indicated, exactly as previously described (11, 45). For natural-latency studies, CD14 monocytes were isolated from fresh venous blood of seropositive donors by MACS separation (as described above), and RNA and DNA were isolated directly for analysis as described below (see Results).

Cytokine analysis.

CCL22 was quantified from supernatants using the RayBiotech C1000 kit according to the manufacturer's instructions (RayBiotech).

Immunofluorescence.

Cells were prepared and analyzed for immunofluorescence staining as described previously (30, 48). UL144 was detected with an IgG2a rat monoclonal antibody (clone 2F11) generated by fusing splenocytes from rats immunized with UL144:Fc (37) with mouse SP2/0 myeloma cells and subsequent selection for hybridomas producing antibody that (i) bound to UL144:Fc by enzyme-linked immunosorbent assay (ELISA) and (ii) detected UL144 expressed on the surface of transfected 293T cells (41) (see Fig. 6). This was followed by anti-rat-594 Alexa Fluor secondary antibody staining. IE was detected using E13 mouse monoclonal antibody conjugated to fluorescein isothiocyanate (FITC).

Fig 6.

The 2F11 anti-UL144 rat monoclonal antibody (2F11 mAb) recognizes UL144 protein from virus isolates of groups 1 and 3, but not group 2. Representative UL144 group 1 to group 3 sequences were cloned in the pND expression vector (kind gift of Peter Barry). The 2F11 antibody was generated by immunizing rats with UL144:Fc protein (ectodomain cloned from the Fiala strain) produced in both baculovirus and 293T cells (50 μg each protein). Forty thousand transfected 293T cells were treated with 5 μg/ml protein and analyzed by fluorescence-activated cell sorting (FACS) as described previously (41).

ChIP assays.

Chromatin immunoprecipitation (ChIP) assays were carried out as previously described (49) with a 1:200 dilution of the appropriate antibodies, i.e., control isotype matched control or anti-H3K4-Di-Me (Upstate) or GATA-2 (Cell Signaling) antibodies. Regions of the UL144 promoter were amplified using the following primers: first GATA-2 binding site forward, TCCATGGGAATCAACGGATC, and reverse, TCCGAACTTTTATACACGCC; deletion region GATA-2 binding site forward, GGCGTGTATAAAAGTTCGGA, and reverse, CAAAGTCCACCTACGACGCT.

Plasmids and PCR.

Plasmids containing UL144 promoter regions driving luciferase reporters were based on pGL3 (Promega). The UL144 promoters of Toledo, Merlin, and TB40E isolates of HCMV were amplified by PCR from viral genomic DNA with the primers 5′-AAGCTTCCTACCGGAAGAA-3′ and 5′-TCTCGAGTATATGCCATACC-3′, which also added HindIII and XhoI cloning sites to the ends of the amplified product. PCR was carried out with the following protocol: 95°C for 40 s, 55°C for 40 s, and 72°C for 1 min for 35 cycles. Following digestion of amplified products and vector and ligation, recombinant clones were verified by restriction digests and confirmed by sequencing. Site-directed mutagenesis was carried out using the XL Quickchange site-directed mutagenesis kit (Stratagene) to generate Merlin UL144 promoter constructs with deletion of either one or two GATA-2 sites. Mutagenesis primer pairs were designed for the GATA-2 deletion as follows: forward primer, 5′-GATTACCTATGCTCCTACGGCCTAAGAGGTAGAC-3′, and reverse primer, 5′-GTCTACCTCTTAGGCCGTAGGAGGCATAGGTAATC-3′. For the GATA-2 T-C point mutation, the following were used according to the manufacturer's instructions: forward primer, 5′-CAACGGATCAATTAACGTCCATCAGCTATGTGATTGTGC-3′, and reverse primer, GCACAATCACATAGCTGATGGACGTTAATTGATCCGTTG-3′. The same sets of primers were sequentially used to generate the double mutant. Plasmids were sequenced with the primers 5′-CTAGCAAAATAGGCTGTCCC-3′ and 5′-CTTTATGTTTTTGGCGTCTTCCA-3′. Plasmid pEF-BOS-GATA-2 was a kind gift of B. Gottgens (Cambridge, United Kingdom). The plasmid containing the LUNA promoter driving the overexpression of luciferase has been previously published (11).

Transfection and luciferase assays.

THP1 cells were transfected with LTX Lipofectamine/Plus Reagent. Samples were transfected at ratios of 1:2:4 of DNA:Plus reagent:LTX Lipofectamine. Experiments were performed in 24-well plates using 2 × 10⁵ cells per sample and 2 μg plasmid DNA. Cells were cultured in X-vivo 15 (Lonza) medium during transfection, and Opti-MEM was used to prepare plasmid DNA/LTX Lipofectamine reagent. Samples were harvested 24 h posttransfection, and luciferase was analyzed as previously described (28–30) with the modification that samples were measured in a 96-well format using the Promega GloMax 96 Microplate luminometer with dual infectors. Light intensity from a 10-s period was measured in accordance with the manufacturer's instructions (Promega).

RT-PCR and RT-qPCR.

Reverse transcription (RT)-PCR parameters and primers have been described elsewhere (30, 45). For RT-PCR, UL144 primers pairs were as follows: forward, 5′-GCCATACCCTATGGGCGCTAC-3′, and reverse, 5′-TCCGAACTTTTATACACGCC-3′, as well as forward, 5′-CGCACATGTAACCGTCAAAC, and reverse 5′-GAAAATTTTGCCGATTGAGC. The real-time quantitative PCR (RT-qPCR) was performed using the Qiagen SYBR green kit (Qiagen) with standard parameters and commercial GAPDH primers (Qiagen), and the RT-qPCR primers for UL144 were as follows: forward primer, 5′-AGGCGTCCAACATCACAAGC, and reverse primer, 5′-CGACCACTTTTCCCTTGTTTG. For the nested PCRs of these amplimers, the primers were forward primer, 5′-TCGTATTACAAACCGCGGAGAGGAT, and reverse primer, 5′-ACTCAGACACGGTTCCGTAA. For US28, forward primer, 5′-AAGGATCCTGGTGCTATCAT, and reverse primer, 5′-TTGAATTCTTACGGTATAAT, were used. For the nested PCRs of these amplimers, the primers were forward primer, 5′-CGTCGGATTCAATGCTCCGGCGATGTTTAC, and reverse primer, 5′ GAATGGCGATGATCACGGCAAAGATCCACC. Samples were run compared to standard curves from lytically infected human foreskin fibroblasts (HFFF) RNA on an ABI 7500 Fast Real Time PCR system machine, and relative values were plotted.

RESULTS

The hypervariable UL144 gene is expressed in a strain-specific manner during HCMV latency and is associated with markers of active chromatin.

We first tested whether UL144 transcripts could be detected during natural latent infection in healthy HCMV carriers. As monocytes are an established site of natural latency in vivo (50, 51), we analyzed monocytes from healthy seronegative and seropositive donors for UL144 expression by nested RT-qPCR.

We tested a total of 6 donors, two seronegatives and four seropositives. Figure 1A represents RT-qPCR from one seronegative and two seropositive donors. As expected, there were no UL144 or US28 viral transcripts detected in the seronegative individual. However, for all the HCMV-seropositive donors, US28 transcript was detectable, and this is consistent with the ability to detect US28 RNA during experimental HCMV latency (7). Interestingly, UL144 was only detected from one of these US28 transcript-positive, HCMV-seropositive donors. Furthermore, of the two other seropositives tested who also expressed US28 as expected, one was found to express UL144 whereas the other was negative for UL144 RNA (data not shown). This suggested that not all donors express detectable UL144 during natural latency, and we posited a number of potential explanations for this observation. Perhaps the most likely was the knowledge that UL144 is extremely hypervariable and the PCR primers used in this assay may not detect UL144 transcripts from all 3 identified groups (52). Alternatively, as latently infected donor monocytes were analyzed ex vivo, it is possible that the cytokine milieu required for latent UL144 expression is present only in some, but not all, donors in vivo. Finally, it may be that the apparent isolate-specific expression of UL144 reflects differences in the UL144 promoter between isolates resulting in differential transcription factor binding and thus disparate UL144 promoter activity.

Fig 1.

The HCMV UL144 gene is transcribed during natural and experimental latency in a strain-specific manner. RNA was isolated from CD14⁺ monocytes isolated from healthy seronegative or seropositive donors, and nested RT-qPCR was carried out for GAPDH, US28 (control latent gene product), or UL144. (A) The graph shows RNA from 3 donors; RT-qPCR was carried out alongside a positive control consisting of HFFF-infected RNA, and relative values were obtained and normalized to GAPDH. (B and C) Monocytes or CD34 cells that had been latently infected with Merlin and TB40E strains of HCMV for 6 days were also analyzed by RT-PCR for the presence of viral transcripts UL144, UL138, and IE and the cellular gene GAPDH. (D) Fibroblasts were infected for 48 h with Merlin and TB40E strains of HCMV, and RT-PCR was carried out for the IE, UL138, and UL144 genes. (E) Additionally, monocytes that had been latently infected with Merlin or TB40E strains of HCMV were chromatin immunoprecipitated with antibodies to the chromatin marker H3K4-Di-Me or the isotype control, and the UL144 promoter was amplified by PCR and analyzed by agarose electrophoresis.

Initially, to test whether UL144 was expressed during latency only from UL144 group 3-specific isolates (which the optimized qRT-PCR UL144 primers target), we analyzed UL144 expression in two different experimental models of HCMV latency (CD34⁺ progenitor cells and monocytes) after latent infection with TB40E (UL144 group 3) and Merlin (UL144 group 1) with these same primers.

Figure 1B shows that in CD34⁺ cells the latency-associated gene product UL138 is expressed with both isolates of HCMV in the absence of IE gene expression. In contrast, UL144 expression was detected during latent infection with only the Merlin but not the TB40E isolate of HCMV when using primers for RT-PCR specific for both isolates of virus.

We also analyzed UL144 expression from these isolates after establishing experimental latency in CD14⁺ monocytes (Fig. 1C). As with CD34⁺ cells, UL144 RNA (in the absence of IE expression) was observed only after infection with Merlin; infection with TB40E showed no UL144 transcription. In contrast, both virus isolates expressed IE transcripts as well as UL144 and UL138 RNAs during lytic infection of fibroblasts, as expected (Fig. 1D).

We have previously shown that, consistent with cellular gene expression, specific chromatin marks of transcriptional activation or repression correlate well with expression of viral genes during latent or lytic HCMV infection (10, 49). Specifically, in latently infected myeloid cells, the IE gene is associated with repressive chromatin markers, whereas, in contrast, the LUNA gene is associated with active chromatin markers (10, 11, 49). Consequently, we analyzed the chromatin around the UL144 promoter of TB40E and Merlin during latent infection of CD14⁺ cells.

Consistent with the isolate-specific expression of UL144 during latent infection shown in Fig. 1B, the Merlin UL144 promoter was associated with dimethylated histone H3 at lysine 4 (H3K4me2) during latent infection of monocytes—a well-established marker of actively transcribed genes (53). In contrast, and again consistent with the lack of detection of UL144 transcription in latently infected monocytes (Fig. 1E), the TB40E UL144 promoter showed no such chromatin marks of transcriptional activity (Fig. 1E).

These data also show that detection of UL144 expression during latency is not specific to group 3 viruses, as UL144 expression was also observed during latent infection with Merlin, a group 1 isolate. Furthermore, as the analyses of UL144 expression by both Merlin (which expresses UL144 during latency) and TB40E (which does not express UL144 during latency) were carried out in the same minimal media ex vivo, it is unlikely that the isolate-specific expression of UL144 during experimental latency was due to differential host cytokine signaling.

We reasoned, then, that a likely explanation for the differential expression of UL144 during latency was due to differences between isolates in the UL144 promoter sequence itself and that these differences might explain the expression of UL144 in myeloid cells in an apparently strain-specific manner.

The latency-associated gene LUNA contains functional GATA-2 binding sites in its promoter, as does the UL144 promoter of some HCMV isolates.

To try to address which transcription factor(s) may be responsible for activity of the UL144 promoter during latent infection, we initially examined the transcription factor binding sites in a promoter of a previously identified latency-associated gene, LUNA. It has previously been shown that experimental as well as natural latent infections of CD34⁺ progenitor by HCMV result in expression of the viral LUNA gene (11) and this is regulated by histone modifications around the LUNA gene promoter: the LUNA promoter is associated with acetylated histones during HCMV latency, in contrast to the viral major immediate promoter (MIEP), which shows repressive chromatin markers (11, 49, 54). Consequently, the LUNA promoter represents a viral gene promoter whose activity is clearly latency associated. On this basis, we examined the LUNA promoter of two clinical isolates (TB40E and Merlin) to try to identify potential transcription factor binding sites which would be consistent with its expression in undifferentiated myeloid cells. We noted the presence of potential binding sites for the cellular transcription factor GATA-2 (Fig. 2A). Interestingly, GATA-2 is an important transcription factor for myelopoiesis, and its expression is maintained in undifferentiated cells of the myeloid lineage (55–57). Perhaps even more suggestive of an involvement of GATA-2 in viral latency, we have previously shown that latent HCMV infection in CD34⁺ progenitor cells results in GATA-2 upregulation (45) and that LUNA is expressed during experimental latency after infection with either Merlin or TB40E isolates (24).

Fig 2.

The latency-associated gene LUNA binds GATA-2 at the promoter and is GATA-2 responsive. (A) Sequence alignment of the LUNA promoter was carried out using Clustal Omega. Predicted GATA-2 binding sites are highlighted in red, and the TATA box is in green. (B) Monocytes latently infected with the Merlin or the TB40E strain of HCMV were chromatin immunoprecipitated with antibodies to GATA-2 (or the isotype control), and the LUNA promoter was amplified by PCR and analyzed by agarose electrophoresis. (C) Alternatively, THP1 cells were transfected with luciferase constructs driving luciferase expression from the LUNA promoter or the vector-only control and the transfection control plasmid. Luciferase activity is presented as fold induction over vector only, and the graph represents data from 3 replicates with standard deviations of the means.

We first analyzed whether GATA-2 binds to the LUNA promoter during latent infection in monocytes by chromatin immunoprecipitation assays. Figure 2B shows that GATA-2 does, indeed, bind the LUNA promoter during latent infection with either TB40E or Merlin.

We next asked whether the LUNA promoter was responsive to GATA-2 by using a LUNA-luciferase reporter and expression vectors for GATA-2 in THP1 cells, a myelomonocytic cell line that recapitulates the known differentiation-dependent permissiveness of HCMV gene expression observed with primary monocytes (Fig. 2C) (58). In these transient cotransfection assays, GATA-2 reproducibly activated the LUNA promoter from the TB40E isolate of HCMV (Fig. 2C).

Taken together, these data argue that the LUNA promoter contains GATA-2 binding sites that can bind GATA-2 and enhance LUNA promoter activity, and it is likely that this factor helps mediate the active chromatin structure around the LUNA promoter observed during latency (11).

Together with LUNA, viral UL138 and viral IL-10 (cmv-IL-10) have also been shown to be expressed during latent infection in both experimental and natural models of HCMV latency (17, 18, 59). Interestingly, analyses of all three of these latency-associated genes in Merlin for defined GATA-2-binding consensus sequences (60) show putative GATA-2-binding sites in their promoters (data not shown). This suggests that GATA-2 responsiveness may be a characteristic of latent promoters in general—this is under investigation.

GATA-2 occupancy on the UL144 promoter during latency defines the isolates that express UL144 during latent infection.

In order to interrogate whether the isolate-specific expression of UL144 during latency was, indeed, due to differences in GATA-2 binding sites in the UL144 promoter, we analyzed the UL144 promoter sequences of 5 different strains of HCMV (Fig. 3A). Using sequence criteria previously described (60), two GATA-2 binding sites were predicted in the promoter region of the Merlin and AF1 isolates of HCMV, but these consensuses were not present in TB40E, FIX, or Toledo (Fig. 3A). To determine whether GATA-2 was differentially associated with these promoters during latent infection in myeloid cells, chromatin immunoprecipitation assays were carried out on monocytes that had been latently infected with either HCMV TB40E (no GATA-2 consensuses) or Merlin (carrying two GATA-2 consensuses). PCR primers spanning each of the two predicted GATA-2 binding sites were used in ChIP assays to determine whether GATA-2 bound to these promoter regions during latent infection. In Merlin, both sites were found to have GATA-2 occupancy in latently infected monocytes (Fig. 3B). In contrast, latent infection with TB40E showed no evidence of GATA-2 binding to either equivalent region of the promoter (Fig. 3C). Therefore, in the case of Merlin and TB40E, the transcription of UL144 appears to be correlated to the presence of two GATA-2 binding sites in the UL144 promoter.

Fig 3.

Sequence analysis shows strain variation in the prediction of GATA-2 binding sites in the UL144 promoter, and those sites have GATA-2 occupancy. (A) The UL144 promoter sequences from the HCMV strains Merlin, AF1, TB40E, FIX, and Toledo were aligned, and the predicted translation ATG sites (green), TATA box (yellow), and GATA-2 binding sites (red) are highlighted. (B and C) Monocytes were latently infected with Merlin and TB40E strains of HCMV, the chromatin immunoprecipitations were carried out using a GATA-2 antibody (or the isotype control), and the two GATA-2 binding sites in the UL144 promoter were amplified by PCR and analyzed by agarose electrophoresis.

The presence of two GATA-2 binding sites in the UL144 promoter predicts expression of UL144 during latent infection.

Since the presence (seen with Merlin) or absence (seen with TB40E) of expression of UL144 during latent infection appeared to be correlated with GATA-2 binding to the UL144 promoter, we asked whether expression of UL144 by two other isolates of HCMV during latency could also be predicted by GATA-2 binding to the UL144 promoter.

HCMV AF1 has two predicted GATA-2 binding sites in its UL144 promoter (Fig. 3A). In contrast, like TB40E, HCMV strain FIX has none. Consistent with this, ChIP analysis of GATA-2 binding to the UL144 promoters of these viruses during latent infection of monocytes showed that the AF1 UL144 promoter binds GATA-2 whereas the FIX UL144 promoter does not (Fig. 4A). Consequently, we reasoned that this should also predict that AF1 will express UL144 during latency but FIX should not. Figure 4B shows, again consistent with this prediction, that UL144 expression is detected only during AF1 latency; no UL144 expression is observed in monocytes latently infected with FIX. In contrast, and as expected, UL138 is expressed during latent infection with both viruses, and this is in the absence of viral IE expression. Interestingly, when the UL144 promoters were sequenced from the two natural-latency donors (Fig. 1), only the virus that expressed UL144 during natural latency contained sequences consistent with the presence of GATA-2 binding sites. It appears, then, that the presence of GATA-2 binding to the promoter of UL144 determines whether UL144 is transcribed during HCMV latency in monocytes.

Fig 4.

(A) The HCMV strain AF1 UL144 promoter, but not the FIX UL144 promoter, binds GATA-2 and is expressed during latency. (B) Monocytes were latently infected with either the AF1 or the FIX strain of HCMV for 6 days and then chromatin immunoprecipitated with antibodies to GATA-2 (or the isotype control), and the UL144 promoter was amplified by PCR and analyzed by agarose electrophoresis. Alternatively, cells were harvested for RT-PCR for the viral transcripts UL144, UL138, and IE and the cellular gene GAPDH.

Expression of UL144 during HCMV latency is dependent upon the presence of GATA-2 binding sites.

Having established that during latency, UL144 expression is dependent on whether GATA-2 binding to the UL144 promoter occurs, we next analyzed the role of GATA-2 in activation of the UL144 promoter in myeloid cells in more detail.

A number of constructs were prepared based on the UL144 promoter from Merlin, in which specific regions of the UL144 promoter from Merlin (functional during latency) were replaced with promoter sequences from the TB40E UL144 promoter (not functional during latency). An initial test with the wild-type (WT) Merlin and TB40E UL144 promoters in myelomonocytic THP1 cells showed that the Merlin UL144 promoter is responsive to additional GATA-2 whereas the TB40E UL144 promoter is not; this recapitulates the isolate-specific expression of UL144 in the context of virus latency (Fig. 5A). Additionally, we tested one other UL144 promoter, from the Toledo isolate, which is TB40E-like, and this also showed no response to GATA-2 in transfection assays (Fig. 5A).

Fig 5.

The Merlin UL144 promoter, but not that of TB40E, is responsive to GATA-2, and removal of GATA-2 binding sites prevents activation by additional GATA-2. THP1 cells were transfected with plasmids driving expression of luciferase from the WT UL144 promoter from Merlin or TB40E (A) or with plasmids driving expression of luciferase from the Merlin UL144 promoter that has either the first, the second, or both of the GATA-2 binding elements mutated (B) with a transfection control plasmid. Graphs represent between 3 and 10 experiments of triplicate samples with standard deviation error bars.

As we would predict, when the Merlin UL144 promoter was mutated to remove either one or both of the GATA-2 binding sites and replaced with equivalent sequences from TB40E, it was no longer responsive to GATA-2 in transfection assays (Fig. 5B). Further to this, when the UL144 promoter regions from the two seropositive donors (Fig. 1A) were sequenced, donor 2 had a sequence consistent with mutated GATA-2 regions. In contrast, the UL144 promoter sequence from donor 3 was consistent with the presence of GATA-2 binding domains. Thus, during both experimental and natural HCMV latency, the presence of functional GATA-2 binding sites in the UL144 promoter is predictive for UL144 expression.

UL144 is expressed at the protein level during HCMV latency and is functionally active.

Having established that UL144 is expressed at the RNA level after latent infection with some isolates of HCMV, we next determined whether this gene is expressed at the level of protein and, if so, whether the protein is functional during latency.

We first determined whether UL144 protein could be detected during lytic infection at the protein level. To do this, a rat monoclonal antibody that recognizes UL144 proteins from different virus groups (Fig. 6) was generated. Figure 7A confirms that UL144 antibody is able to detect UL144 in fibroblasts lytically infected with either the TB40E or the Merlin isolate of HCMV by indirect immunofluorescence. There was no UL144 detected in fibroblasts infected with AD169, which lacks UL/b′ and, hence, lacks UL144. In contrast, after latent infection of monocytes, UL144 protein was detected only when latency was established with Merlin; latent infection of monocytes with TB40E showed no detectable UL144 protein (Fig. 7B).

Fig 7.

UL144 protein is expressed in a strain-specific manner during HCMV latency, and this leads to an upregulation of the chemokine CCL22. (A) Human foreskin fibroblasts (HFFs) were infected with HCMV strains AD169, TB40E, and Merlin for 72 h and then fixed and stained with antibodies to IE and UL144 as well as the nuclear stain Hoechst. (B) Alternatively, monocytes were latently infected with Merlin or TB40E strains of HCMV for 6 days and then fixed and stained with antibodies for IE and UL144 as well as a nuclear stain. (C) Alternatively, cells were reactivated, and the same staining protocol was repeated. (D) Cells shown in panel C were also harvested in Trizol for RT-PCR analysis of the viral transcripts IE, UL138, and UL144 as well as cellular GAPDH. (E) CCL22 levels from supernatants of CD34⁺ cells that had been latently infected with either the Merlin or the TB40E strain of HCMV were analyzed from the secretome at 10 days postinfection; data represent duplicates.

We and others have previously shown that differentiation of latently infected monocytes to monocyte-derived dendritic cells results in reactivation of latent virus in both natural and experimental models of HCMV latency (7, 10, 50). To ensure that both TB40E and Merlin had efficiently established a latent infection capable of reactivation in monocytes, we induced reactivation of latent virus by differentiation and analyzed cells for UL144 expression, which would be predicted to occur during reactivation/induction of lytic cycle regardless of the ability of the viruses to express UL144 during latent infection. Figure 7C shows that UL144 protein expression by both TB40E and Merlin strains of virus was detectable upon reactivation of lytic cycle, and this was confirmed by RT-PCR analysis (Fig. 7D).

Finally, to examine whether the UL144 protein expressed during latency had any functional significance, we assayed the levels of CCL22 during latent infection, as we and others have previously shown that expression of UL144 (and its functional homologue in EBV, LMP-1) constitutively leads to the upregulation of the chemokine CCL22 (28–30, 44, 61). Figure 7E shows that monocytes latently infected with HCMV Merlin, but not TB40E, have increased CCL22 in their latency-associated secretome, consistent with functional UL144 expression during latent infection with Merlin but not TB40E isolates of HCMV.

DISCUSSION

It is becoming clear that, in the same way that HCMV modulates multiple cell functions during lytic infection, a complex interplay between the virus and the host cell also occurs during viral latency. Changes in the cellular transcriptome clearly occur during latent infection. This includes latency-associated changes in cellular coding RNAs (62), but modulation of cellular gene expression orchestrated by latency-associated changes in cellular miRNAs also occurs (45). Notably, altered expression of cellular miRNAs during latency can result in increased levels of the myeloid transcription factor GATA-2 (45), which we now show specifically promotes latency-associated gene expression of at least two HCMV ORFs, UL144 and LUNA.

First, our analyses show that expression of UL144 mRNA is detectable in naturally latently infected monocytes from healthy seropositive donors and that UL144 mRNA and protein are detected in two experimental models of HCMV latency. However, this latency-associated expression of UL144 appears to be isolate specific. We have previously shown that in both naturally and experimentally latently infected myeloid cells latent and lytic promoter activity is regulated by changes in posttranslational modifications of chromatin-associated histones (10, 11). Consistent with this, the chromatin around the Merlin UL144 promoter (active in latently infected monocytes) but not the TB40E UL144 promoter (inactive in latently infected monocytes) showed histone posttranslational modifications marks associated with actively transcribed genes.

Comparison of the UL144 promoter from virus isolates, which did or did not express UL144 during latency, suggested that this isolate-specific UL144 expression was, at least in part, due to differences in GATA-2 binding sites in the UL144 promoter. Consistent with this, in myeloid cells, the UL144 promoter of Merlin, which contains consensus GATA-2 binding sites, bound GATA-2 during latent infection and was active during latency. In contrast, the TB40E isolate, which does not contain consensus GATA-2 binding sites in its UL144 promoter, showed no evidence of GATA-2 binding to UL144 promoter sequences and showed no latency-associated UL144 expression.

Why the immune-modulatory protein UL144 appears to be expressed by only some isolates of HCMV during experimental latency, but not others, is unclear at present. Since the establishment and reactivation from experimental latency does not appear to require UL144 (Fig. 7) (24), we postulate that the functional consequence of UL144 expression may be more relevant/important in the context of natural latency. For example, the enhanced expression of CCL22 in latently infected cells expressing UL144 (Fig. 7E) might shape the “immune environment” that surrounds latently infected cells, subverting Th1-mediated defenses that would otherwise clear these cells from the host (30). In addition, UL144 binding to BTLA may inhibit similarly detrimental immune responses that would negatively impact viral persistence. In this regard, UL144 has the potential to suppress host defenses both in cis and in trans (63), as myeloid lineage cells express BTLA and may be subject to cell-intrinsic regulation of BTLA signaling by UL144 in the latently infected cells themselves. However, these explanations do not reconcile what the selective advantage may be for some clinical isolates of HCMV to express UL144 during natural latency and for others to not express it, as we have observed to be the case for two of the four healthy seropositive individuals analyzed in this study. Furthermore, of the 41 sequenced clinical isolate strains, 17 were found to contain GATA-2 binding sites (A. Davison, personal communication), suggesting that UL144 regulation by GATA-2 is not absolutely necessary for carriage of virus in vivo. Epidemiological studies indicate that specific strains of HCMV preferentially circulate in distinct subpopulations (64), suggesting that UL144 may differentially impact latency and/or host immunity depending upon the specific genetic background of the individual. Notably, conflicting conclusions have been reached when assessing the potential role of UL144 in the outcome of congenital HCMV infection (65–69), and these studies have been performed with people of differing ethnicities, further suggesting that the impact of UL144 on HCMV pathogenesis may be genetically linked and, as yet, remains unclear.

To our knowledge, UL144 is the first example of an HCMV latency-associated gene product that is differentially expressed by virus isolates during latency, apparently due to differences in myeloid-specific transcription factor binding sites in the gene promoter. Consequently, analyses of HCMV gene promoters could help to identify other latency-associated genes (e.g., those containing GATA-2 binding elements), further our understanding of how the latency-associated transcriptome is regulated, and help to delineate whether differential expression of latency-associated genes may be linked to HCMV-associated disease.