Research ArticleCancer

HCMV-Encoded Chemokine Receptor US28 Mediates Proliferative Signaling Through the IL-6–STAT3 Axis

See allHide authors and affiliations

Science Signaling  03 Aug 2010:
Vol. 3, Issue 133, pp. ra58
DOI: 10.1126/scisignal.2001180

Abstract

US28 is a viral G protein (heterotrimeric guanosine triphosphate–binding protein)–coupled receptor encoded by the human cytomegalovirus (HCMV). In addition to binding and internalizing chemokines, US28 constitutively activates signaling pathways linked to cell proliferation. Here, we show increased concentrations of vascular endothelial growth factor and interleukin-6 (IL-6) in supernatants of US28-expressing NIH 3T3 cells. Increased IL-6 was associated with increased activation of the signal transducer and activator of transcription 3 (STAT3) through upstream activation of the Janus-activated kinase JAK1. We used conditioned growth medium, IL-6–neutralizing antibodies, an inhibitor of the IL-6 receptor, and short hairpin RNA targeting IL-6 to show that US28 activates the IL-6–JAK1–STAT3 signaling axis through activation of the transcription factor nuclear factor κB and the consequent production of IL-6. Treatment of cells with a specific inhibitor of STAT3 inhibited US28-dependent [3H]thymidine incorporation and foci formation, suggesting a key role for STAT3 in the US28-mediated proliferative phenotype. US28 also elicited STAT3 activation and IL-6 secretion in HCMV-infected cells. Analyses of tumor specimens from glioblastoma patients demonstrated colocalization of US28 and phosphorylated STAT3 in the vascular niche of these tumors. Moreover, increased phospho-STAT3 abundance correlated with poor patient outcome. We propose that US28 induces proliferation in HCMV-infected tumors by establishing a positive feedback loop through activation of the IL-6–STAT3 signaling axis.

Introduction

Human cytomegalovirus (HCMV), a member of the family of β-herpesviruses, is widespread, persisting as a latent infection in up to 90% of the U.S. population (1). Although HCMV is asymptomatic in immunocompetent individuals, it may cause pathologies such as pneumonitis, hepatitis, and retinitis in immunocompromised hosts (1). Furthermore, HCMV has been proposed to promote the development of colon cancer (2) and malignant glioblastoma (3). HCMV has also been detected as an active infection in several forms of cancer, including prostate cancer, colon cancer, and malignant glioblastoma. However, the exact role of HCMV as a promoting factor in these tumors remains elusive. HCMV-encoded proteins are thought to promote tumor formation either directly, through activation or inhibition of cellular signaling pathways, or indirectly, through induction of autocrine and paracrine signaling. One of the HCMV-encoded proteins, known to induce a proliferative and angiogenic phenotype in vitro and in vivo, is the viral chemokine receptor US28 (4).

The presence of four G protein (heterotrimeric guanosine triphosphate–binding protein)–coupled receptors (GPCRs) that are structurally similar to human chemokine receptors in the HCMV genome—US28, US27, UL33, and UL78—is intriguing (5, 6). Chemokine receptors are involved in the regulation of the immune system (7) and have been implicated in various aspects of oncogenesis (8). The HCMV-encoded GPCR US28, which is structurally similar to the human chemokine receptor CCR1 (9), has been studied most extensively (10). US28 binds various chemokines, including CCL2, CCL5, and CX3CL1 (11), and may thereby suppress the host immune response (12). In addition, US28 also signals constitutively and shows G protein promiscuity, traits that enable it to hijack the host cell’s signaling machinery (13).

Like chemokine receptors, viral GPCRs appear to play a role in oncogenesis and tumor growth (14, 15). Transgenic mice expressing ORF74, the chemokine receptor encoded by Kaposi sarcoma–associated herpesvirus (KSHV), develop lesions resembling Kaposi sarcoma (16). Similarly, US28 induces various oncogenic responses, including increased cyclin D1 and cyclooxygenase-2 (COX2) production, as well as that of vascular endothelial growth factor (VEGF), when expressed in NIH 3T3 fibroblasts (4, 17). Moreover, US28 contributes to HCMV-induced VEGF promoter activity and COX2 expression in HCMV-infected cells (4, 17) and promotes tumor formation in a mouse xenograft model (4, 17).

To investigate the molecular mechanism by which US28 contributes to oncogenesis, we analyzed US28-induced release of angiogenic factors. We used an antibody array that recognized different chemokines, growth factors, and cytokines to identify factors secreted by US28-expressing NIH 3T3 cells. US28 increased the production of interleukin-6 (IL-6), which is induced by HCMV infection (18, 19) and has been implicated in oncogenesis (20, 21). In addition, we identified a key role for the IL-6–STAT3 (signal transducer and activator of transcription 3) axis in US28-mediated proliferative signaling. Our data thus reveal a positive feedback loop initiated by US28 that is crucial for the proliferative phenotype shown by US28-expressing cells.

Results

US28 increases the secretion of IL-6 and VEGF

NIH 3T3 cells stably expressing US28 display oncogenic properties (4, 17). For instance, injection of US28-transformed NIH 3T3 cells into nude mice results in tumor formation (4). Because angiogenic factors are required for the formation of large tumors (22), we assessed their secretion by US28-expressing NIH 3T3 cells. Using a mouse antibody array for angiogenic factors, we analyzed conditioned medium from US28-expressing NIH 3T3 cells and compared it to that from mock-transfected NIH 3T3 cells. The array analysis showed a marked increase in both IL-6 (296 ± 3%) and VEGF (271 ± 45%) in medium from the US28-expressing cells (Fig. 1); the latter finding was consistent with earlier observations (4). The medium concentration of CCL2, a chemokine that binds US28 and is subsequently internalized, was decreased (13 ± 2%) (11), whereas the medium concentration of CCL11 and CXCL4, which do not bind US28, was unaffected. Because IL-6 has been implicated in oncogenesis (22, 23), we investigated the role of IL-6 in US28-induced proliferative signaling.

Fig. 1

Increased IL-6 and VEGF secretion in US28-transfected cells. Medium from NIH 3T3 cells stably transfected with US28 was compared with medium obtained from mock-transfected cells. Secreted cytokines and growth factors were analyzed with an antibody array. The mean result of three separate experiments is shown as a percentage of protein content in US28-expressing cells compared to mock-transfected cells. CCL2 secretion was significantly decreased in US28-expressing cells (***P < 0.001), whereas IL-6 and VEGF secretion were both significantly increased (**P < 0.01).

US28 induces STAT3 phosphorylation and STAT3-driven transcriptional activation

IL-6 binds to the IL-6 receptor to activate STAT3 signaling (24); therefore, we assessed the phosphorylation status of STAT3. Consistent with the observed increase in IL-6 concentration in the supernatant of stably US28-transfected cells, STAT3 phosphorylation was markedly increased in US28-expressing cells compared to mock-transfected cells (Fig. 2A). In contrast, no STAT3 phosphorylation was observed in cells stably transfected with the US28-R129A mutant, which fails to couple to G proteins (25). Furthermore, a luciferase-based reporter assay showed that STAT3-driven transcriptional activation was increased in human embryonic kidney (HEK) 293T transiently transfected with a plasmid encoding US28, but not in mock- or US28-R129A–transfected cells (Fig. 2B). The presence of US28 in transfected HEK293T (Fig. 2C) and NIH 3T3 (Fig. 2D) cells was verified by [125I]CCL5 binding.

Fig. 2

Increased STAT3 activity in NIH 3T3 and HEK293T stably transfected with US28. (A) STAT3 phosphorylation is increased by ~520% compared to mock in NIH 3T3 stably transfected with US28, whereas cells transfected with the G protein–uncoupled mutant US28-R129A show no increased STAT3 phosphorylation. (B) In HEK293T cells, STAT3-driven transcriptional activation is only observed when the US28 wild-type (WT) receptor is present. **P < 0.01, compared to mock. (C) [125I]CCL5 binding to US28 and its displacement by 10−7 M CX3CL1 in HEK293T cells. **P < 0.01, compared to mock. (D) [125I]CCL5 binding to US28 and US28-R129A and its displacement by 10−7 M CX3CL1 in stably transfected NIH 3T3. ***P < 0.001, for both WT and US28-R129A compared to mock.

JAK1 and nuclear factor κB mediate US28-induced STAT3 signaling

Next, we used specific inhibitors to identify components of the signaling pathway leading from US28 to STAT3 activation (26, 27). We performed these analyses both in transiently transfected HEK293T cells, in which we assessed STAT3 activity with STAT3 reporter gene assays, and in NIH 3T3 cells stably transfected with US28, in which we determined STAT3 phosphorylation by Western blot analysis. Overnight treatment with 10 μM pyridone 6 (P6) [a pan-JAK (Janus kinase) inhibitor] inhibited STAT3-dependent reporter gene activity by 51.6 ± 1.5% in HEK293T cells transfected with plasmids encoding US28 (Fig. 3A). In contrast, overnight exposure to several other kinase inhibitors [10 μM AG-490 (JAK2 inhibitor), PP-2 (Src inhibitor), or Tyrene CR4 (Abl inhibitor)] failed to affect STAT3 signaling. Signaling through the Gαo family of G proteins activates STAT3 (28). Therefore, we assessed the effects of treating cells with the Gαi/o inhibitor pertussis toxin (PTX; 100 ng/ml) on US28-induced STAT3 signaling and observed no significant effect on STAT3 reporter gene activity (Fig. 3A). Because IL-6 production is enhanced by the transcription factor nuclear factor κB (NF-κB) (29, 30), which is constitutively activated through both Gαq and Gβγ in cells expressing US28 (12), we treated cells with the NF-κB inhibitor BAY11-7082 (31, 32). BAY11-7082 reduced transcriptional activation of reporter genes by STAT3 78.7 ± 4.2% (Fig. 3A).

Fig. 3

US28-induced STAT3 phosphorylation and transcriptional activity is mediated by JAK1. (A) STAT3-driven transcriptional activation was inhibited by treatment with 10 μM P6 (pan-JAK kinase inhibitor) (**P < 0.01, compared to vehicle-treated), whereas similar concentrations of AG-490 (JAK2 inhibitor), PP-2 (Src inhibitor), and Tyrene CR4 (Abl inhibitor) had little or no effect. PTX (100 ng/ml) had no effect on STAT3 activation. The NF-κB inhibitor BAY11-7082 reduced STAT3-dependent reporter gene activation by ~80%. **P < 0.01, compared to vehicle. (B) NIH 3T3 cells stably transfected with US28 were treated for 30 min with the different kinase inhibitors. Only 10 μM P6 inhibited STAT3 phosphorylation (~6% compared to vehicle-treated cells).

We also assessed the effects of the kinase inhibitors P6, AG-490, and Tyrene CR4 on STAT3 phosphorylation in US28-expressing NIH 3T3 cells treated for 30 min with each inhibitor (10 μM). We observed a strong reduction of STAT3 phosphorylation only after treatment with P6 (Fig. 3B), implicating JAK1 in mediating US28-dependent STAT3 phosphorylation.

US28 activates STAT3 signaling in a paracrine and autocrine fashion

Next, we investigated the importance of US28-mediated IL-6 release in STAT3 activation. We observed a marked increase in STAT3 phosphorylation as assessed by Western blot analysis in mock and US28-expressing NIH 3T3 cells treated with IL-6 (10 ng/ml; Fig. 4A). Moreover, conditioned medium from US28-expressing NIH 3T3 cells induced STAT3 phosphorylation in both mock- and US28-R129A–transfected cells (Fig. 4B), whereas incubation of US28-expressing NIH 3T3 cells with a neutralizing antibody to IL-6 attenuated STAT3 phosphorylation (Fig. 4C). Treatment of US28-expressing NIH 3T3 cells with 10 μM madindoline A, which inhibits the formation of gp130 homodimers and thereby signaling of the IL-6Rα–gp130 complex (33), completely inhibited US28-induced STAT3 phosphorylation (Fig. 4D).

Fig. 4

US28 induces a positive feedback loop involving IL-6 signaling. (A) Incubation with IL-6 (10 ng/ml) for 10 or 30 min elicits STAT3 phosphorylation in both mock-transfected (~2500 and ~1800% compared to 0 min, respectively) and US28-transfected (~450 and ~500% compared to 0 min, respectively) NIH 3T3 cells. (B) Conditioned medium from NIH 3T3 cells transfected with US28 induced STAT3 phosphorylation after a 90-min incubation in both mock-transfected (~1300% compared to vehicle) and US28-R129A–transfected (~760% compared to vehicle) NIH 3T3 cells. (C) Incubation with IL-6–neutralizing antibody (1 μg/ml) for 90 min decreased STAT3 phosphorylation in US28-expressing cells (~20% compared to vehicle). (D) Overnight incubation with 10 μM gp130 inhibitor madindoline A also decreased STAT3 phosphorylation in US28-expressing cells (~4% compared to vehicle). (E) HEK293T cells expressing US28 were cotransfected with different amounts of shIL-6. At 100 ng of shIL-6 per 106 cells, STAT3-dependent transcriptional activation was reduced by almost 50% (*P < 0.05, compared to transfection with shEmpty), and 400 ng of shIL-6 per 106 cells abolished STAT3-driven transcriptional activation (***P < 0.001, compared to transfection with shEmpty). (F) Cotransfection of shIL-6 failed to significantly alter NFAT reporter gene activation.

To investigate the involvement of IL-6 in the US28-induced activation of STAT3 in HEK293T cells, we expressed shIL-6, the short hairpin RNA (shRNA) targeting and down-regulating IL-6 (23), in conjunction with the STAT3 reporter gene. Cotransfection of shIL-6 completely abolished US28-induced STAT3 transcriptional activity, whereas expression of scrambled shRNA (shEmpty) did not affect STAT3 transcriptional activity (Fig. 4E). In contrast, cotransfecting shIL-6 with an NFAT (nuclear factor of activated T cells) reporter gene did not significantly alter US28-induced NFAT signaling (Fig. 4F). These results indicate that in both HEK293T and NIH 3T3 cells, US28 induced IL-6 release, resulting in STAT3 activation by way of the IL-6Rα–gp130 complex.

STAT3 is involved in the US28-induced proliferative phenotype

Next, we used inhibitors of STAT3 and JAK1 to determine whether IL-6–driven STAT3 signaling was involved in the US28-mediated proliferative phenotype. The specific STAT3 inhibitor JSI-124 (34) inhibited US28-mediated STAT3 signaling in HEK293T cells (Fig. 5A) with an IC50 (median inhibitory concentration) of ~500 nM. Next, we used a [3H]thymidine incorporation assay to determine the effect of various inhibitors of the IL-6–STAT3 pathway on the proliferation of mock and US28-expressing NIH 3T3 cells. We found that treatment with either the STAT3 inhibitor JSI-124 or the JAK inhibitor P6 strongly reduced the proliferation of US28-transfected cells (Fig. 5B).

Fig. 5

STAT3 plays a critical role in US28-induced proliferation. (A) The STAT3 inhibitor JSI-124 inhibited STAT3 response element–dependent activation in HEK293T cells after a 24-hour incubation. (B) Overnight treatment with either JSI-124 or P6 inhibited DNA synthesis in NIH 3T3 cells expressing US28. **P < 0.01, compared to vehicle. (C) Knockdown of IL-6 with 400 ng of shIL-6 per 106 cells in HEK293T cells expressing US28 inhibits transcriptional activation of the VEGF promoter. **P < 0.01. (D) Treatment of HEK293T cells cotransfected with US28 and a reporter gene containing the VEGF promoter with either 500 nM JSI-124 or 25 μM celecoxib partially inhibits transcriptional activation of the VEGF promoter, and treatment with both compounds simultaneously has a synergistic effect. ***P < 0.001, **P < 0.01, compared to single treatment with JSI-124 or celecoxib, respectively. RLU, relative light unit. (E) Inhibition of foci formation by US28-expressing NIH 3T3 cells by either 250 or 500 nM JSI-124. **P < 0.01, compared to vehicle.

Because US28 increases expression of the gene encoding VEGF (4), which in part is regulated by STAT3 (35), we examined whether IL-6 was involved in US28-mediated transcriptional activation of the VEGF promoter. Coexpression of shIL-6 with US28 resulted in reduced US28-induced activation of the VEGF promoter (Fig. 5C). Furthermore, inhibition of STAT3 with JSI-124 partially inhibited US28-induced VEGF promoter activity (Fig. 5D). COX-2 also mediates US28-induced VEGF production (17); therefore, we investigated whether the combined inhibition of STAT3 and COX-2 with JSI-124 and celecoxib had a synergistic effect. Indeed, treatment with both inhibitors resulted in a 67.8 ± 4.2% reduction of VEGF promoter activation compared to 35.1 ± 7.7 and 48.1 ± 5.6% reduction when treated with either JSI-124 or celecoxib, respectively (Fig. 5D). In addition, the transforming potential of US28 was inhibited by treatment with JSI-124, resulting in a 48 ± 4 and 77 ± 8% reduction in number of foci when treated with 250 and 500 nM JSI-124, respectively, in a focus formation assay with NIH 3T3 cells (Fig. 5E).

HCMV infection results in STAT3 activation

To investigate the contribution of US28 to STAT3 activation during HCMV infection, we infected U373 MG (malignant glioblastoma) cells with HCMV Titan wild type or HCMV Titan ΔUS28 mutant at a multiplicity of infection (MOI; the amount of viral particles per cell) of 2. The presence of US28 or lack thereof was confirmed with [125I]CCL5 binding (Fig. 6A). The presence of US28 was further confirmed 24 hours after infection by immunofluorescence with an antibody targeting US28. Infected cells (green) were visualized by means of green fluorescent protein (GFP) incorporated in HCMV Titan, and US28 is shown in red (Fig. 6C, panels I and II). As previously shown, US28 is predominantly found in the perinuclear region of infected cells (36). In contrast, we did not detect any US28 in cells infected with HCMV Titan ΔUS28 (Fig. 6C, panels III and IV). We confirmed US28 functionality by measuring the intracellular accumulation of inositol phosphates, as previously described (4, 13). We observed increased inositol phosphate accumulation in HCMV Titan–infected cells, but not in cells infected with the ΔUS28 strain (Fig. 6B).

Fig. 6

STAT3 activation during HCMV infection is partly mediated by US28. (A) [125I]CCL5 binding to US28 on U373 MG cells 48 hours after infection with HCMV Titan. Cells infected with the WT virus showed CCL5 binding, whereas those infected with ΔUS28 virus did not. *P < 0.05, compared to binding on HCMV Titan ΔUS28. (B) Accumulation of inositol phosphate (IPx) in U373 MG cells 48 hours after infection with HCMV Titan. Cells infected with the WT virus showed increased inositol phosphate accumulation compared to both mock- and ΔUS28 mutant virus–infected cells. ***P < 0.001, compared to HCMV Titan ΔUS28. (C) U373 MG cells infected with HCMV Titan stained with antibody against US28 show specific staining of US28 (red) in panels I and II, whereas in panel II, infected cells are shown in green with a GFP tag incorporated in the viral genome. In panels III and IV, the same staining is performed on cells infected with HCMV Titan ΔUS28 as a negative control; in panel IV, infected cells are also shown in green with the same GFP tag. (D) STAT3-dependent transcriptional activation by HCMV Titan in U373 cells 48 hours after infection with HCMV Titan, which is less prominent in the ΔUS28 mutant virus. ***P < 0.001, compared to the WT virus. (E) STAT3 phosphorylation in U373 MG cells 24 hours after infection with HCMV Titan. STAT3 phosphorylation is induced more strongly upon infection with WT virus (~250% compared to mock) compared to HCMV Titan ΔUS28 (~130% compared to mock). Staining with antibody against IEA confirms viral infection in both samples. (F) HCMV infection induces IL-6 secretion in U373 MG in a US28-dependent fashion, IL-6 concentration was measured 24 hours after infection in serum-starved cells (*P < 0.05).

Experiments using the STAT3 reporter gene showed increased STAT3 activity 48 hours after infection in cells infected with the HCMV Titan strain but significantly less activity in cells infected with the HCMV Titan ΔUS28 mutant strain (P < 0.001) (Fig. 6D). Consistent with the reporter gene data, Western blot analysis of cells infected with the HCMV Titan strain showed increased STAT3 phosphorylation, which was significantly greater than that in cells infected with the HCMV Titan ΔUS28 mutant strain (Fig. 6E). IL-6 concentration in the supernatant of HCMV-infected cells 24 hours after infection was increased by 182 ± 13.8% compared to mock-transfected cells, whereas no increase in IL-6 concentration was apparent in cells infected with the HCMV Titan ΔUS28 strain (Fig. 6F).

Primary glioblastoma tumors contain US28 and activated STAT3

Finally, we examined whether US28, STAT3 phosphorylation, and IL-6 could be detected in primary tumor specimens from patients with malignant glioblastoma. We examined 21 different malignant glioblastoma specimens obtained from patients at debulking surgery, of which 20 were HCMV-positive and 1 was HCMV-negative. The presence of HCMV was confirmed by staining with antibodies directed against US28 and HCMV immediate-early antigen (IEA) (Fig. 7, A and C). Cells containing US28 and showing STAT3 phosphorylation were mostly confined to the vascular wall, with a few cells scattered over the tumor. Double-staining with antibodies directed against phospho-STAT3 and US28 revealed a similar pattern (Fig. 7, B and E; cells with phospho-STAT3 in brown and cells with US28 cells in red). HCMV IEA was present in tumor cells, smooth muscle cells (SMCs; identified with SMC α-actin) (Fig. 7D), and endothelial cells (Fig. 7G). IL-6 was abundant in tumor cells close to the vessels in all tissue samples (Fig. 7F; US28 in brown and IL-6 in red). As shown in Fig. 7G, CD31, which is present on endothelial cells and is a known marker for angiogenesis, was consistently detected in HCMV-positive glioblastoma specimens. We did not observe either HCMV IEA– or US28-stained cells in the HCMV-negative tumor (Fig. 7I). In this HCMV-negative tumor, phospho-STAT3 abundance was low and only detected in <10% of the cells (Fig. 7J), in contrast with its abundance in HCMV-infected tumor samples.

Fig. 7

HCMV IEA, US28, and STAT3 phosphorylation in primary glioblastoma specimens. Representative immunohistochemical stainings are shown. (A) Presence of US28 (brown). (B) Presence of phospho-STAT3 (brown). (C) Presence of HCMV IEA (brown). (D) Presence of SMC α-actin (brown). (E) Double-staining of a primary brain tumor tissue sample with antibody against US28 (red) and phospho-STAT3 (brown) in cells lining a blood vessel. (F) Double-staining for US28 (brown) and IL-6 (red). (G) Presence of CD31 in blood vessels. (H) Rabbit IgG served as an isotype control. (I and J) No US28 (I) or phospho-STAT3 (J) is apparent in tissue from the HCMV-negative patient. Scale bars, 50 μm. (K and L) Kaplan-Meier analysis showing decreased OS probability (K) for individuals with a high grade of STAT3 phosphorylation in the primary tumor (P = 0.039), as well as a shorter TTP (P = 0.0052) (L).

The abundance of US28 and the extent of STAT3 phosphorylation differed among different individuals, allowing us to grade tumor samples accordingly and relate these values to patient outcome. Median overall survival (OS) in patients (n = 9) with <30% US28-positive cells in the tissue was 19.5 versus 14.5 months in those (n = 12) with >30% US28-positive cells (P = 0.7), and median time to tumor progression (TTP) was 12 versus 6.5 months (P = 0.28) (Table 1). OS time and TTP were higher in one patient with <30% HCMV IEA–positive cells (OS, 34 versus a median of 14.5 months; TTP, 17 versus a median of 7 months) than in those (n = 17) with >30% infected cells (Table 1). Patients (n = 14) with <30% phospho-STAT3–positive cells survived significantly longer (median OS, 21 versus 11.5 months, P = 0.039; median TTP, 12 versus 4.5 months, P = 0.0052) than those (n = 7) with >30% phospho-STAT3–positive cells in their tumor tissue (Table 1 and Fig. 7, K and L). The presence of US28 was related to STAT3 phosphorylation (P = 0.006), and STAT3 phosphorylation was related to the presence of IL-6 as determined by a Wald test (P = 0.041). Thus, although we analyzed only a limited number of samples, our data suggest that HCMV infection is related to STAT3 phosphorylation, IL-6 production, and outcome for individuals with malignant glioblastoma.

Table 1

Median OS and TTP in tumors with low-grade (0 to 2) and high-grade (3 to 5) US28, phospho-STAT3, and HCMV IEA (n = 21 patients per group). P value was determined with the log-rank (Mantel-Cox) test.

View this table:

Discussion

Convincing evidence has linked viral infection to several forms of cancer. For example, KSHV and human papillomavirus (HPV) are considered the etiological agents of Kaposi sarcoma and cervical cancer, respectively (37, 38). HCMV proteins and DNA have been detected in several tumors (3942), but a causal relationship has yet to be demonstrated. Although HCMV is not considered an oncogenic virus like KSHV and HPV (43), accumulating evidence suggests that it may act as an oncomodulator (44). Furthermore, HCMV interferes with several key cellular signaling pathways, leading to enhanced tumor survival and angiogenesis, as well as alterations in cell motility and adhesion (45).

Here, we identified a signaling pathway involving NF-κB and IL-6–STAT3 through which US28 induces cell proliferation. Analysis of a set of secreted growth factors, chemokines, and cytokines enabled us to confirm increased VEGF secretion by US28-expressing cells (4). Medium concentration of CCL2 was decreased, reflecting its sequestration by US28 (11). Notably, the concentration of IL-6 was increased in the medium of US28-expressing cells. IL-6 is a pro-inflammatory cytokine that induces STAT3 phosphorylation by binding to its cognate receptor IL-6Rα and thereby activating gp130, the tyrosine kinase subunit of the IL-6 receptor (24). After IL-6 binding to IL-6Rα, the two gp130 subunits activate JAKs, leading to activation of STAT3 and its target genes. IL-6 is a regulatory factor in melanoma, inhibiting proliferation, and may play a pivotal role in the switch from cellular senescence to oncogenesis (23). Moreover, STAT3 is a transcriptional regulator that shows increased activity in solid tumors (46) as well as in lymphomas (47). Recent studies have shown that constitutively active gp130 mutants are responsible for increased STAT3 phosphorylation in hepatocellular tumors (48). As such, IL-6 and STAT3 are considered promising anticancer drug targets (49, 50), stimulating the clinical use of IL-6–neutralizing antibodies (51) and the discovery of several STAT3 inhibitors (34, 52).

The importance of IL-6 and STAT3 signaling in oncogenesis (20, 21) led us to investigate the role of the IL-6–STAT3 axis in US28-mediated proliferative signaling. The rise in IL-6 production and secretion in US28-expressing cells was associated with increased activation of STAT3 through the upstream activation of JAK1. This increase was apparent only in cells transfected with US28 and not in either mock-transfected cells or cells carrying the US28 mutant US28-R129A. Using conditioned growth medium, IL-6–neutralizing antibodies, an inhibitor of the IL-6 receptor, and shRNA targeting IL-6, we showed that US28 activates the IL-6–JAK1–STAT3 signaling axis. Because IL-6 itself is transcriptionally activated by STAT3 (53), this creates a positive feedback loop. Indeed, both IL-6–STAT3 autocrine and paracrine loops have been described in cancer cells (54, 55). A common factor in both US28 and IL-6 signaling is NF-κB, which is activated in response to US28 signaling (13) and drives the transcription of IL-6 (29, 30, 56). We found that NF-κB signaling was crucial to US28-mediated STAT3 activation. Using the G protein–uncoupled US28-R129A mutant, we demonstrated that like constitutive US28-driven NF-κB signaling (13), US28-induced STAT3 activation is G protein–dependent. Thus, a positive loop activating IL-6–JAK1–STAT3 is initiated in a US28-dependent manner through activation of NF-κB. Treatment of cells with the STAT3 inhibitor JSI-124 inhibited US28-induced [3H]thymidine incorporation, VEGF promoter activity, and foci formation, further demonstrating the relevance of the STAT3 pathway to the US28-induced proliferative phenotype. We previously identified a role for COX-2 in US28-induced VEGF production (17); our experiments here indicate that US28-induced VEGF promoter activation is regulated by both STAT3 and COX-2, suggesting that US28 acts as an oncomodulator through multiple mechanisms.

We also observed increased STAT3 activity and IL-6 abundance in HCMV-infected U373 MG glioblastoma cells, both of which were attenuated in cells infected with a strain lacking US28. However, STAT3 activity was not fully abolished in cells infected with the ΔUS28 virus, suggesting that STAT3 activation is also promoted by other viral factors. Furthermore, HCMV-infected cells secrete large amounts of IL-5 (57), which activates STAT3 (58). The increased IL-6 production in HCMV-infected U373 MG cells appears to depend on US28. Increased IL-6 abundance and subsequent activation of the STAT3 axis have been reported in HCMV-infected HUVECs (human umbilical cord endothelial cells) and U373 MG cells (57, 59). Moreover, gliomas and some cancer stem cells require IL-6 and STAT3 for tumor growth and survival (60).

The presence of US28 and STAT3 phosphorylation in cells lining the blood vessels in primary glioblastoma tumors suggests that US28 may be involved in the formation and maintenance of these tumors within the vascular niche. The IL-6 receptor is present in glioblastomas, and IL-6 triggers proliferation and migration in cerebral endothelial cells (61, 62). The specific localization of these cells suggests that they may play a role in vascularization of the tumor, particularly in light of our data showing IL-6–dependent VEGF secretion. Furthermore, IL-6 induces the expression of VEGF and other angiogenic factors in pulmonary hypertension (63). Moreover, increased IL-6 concentrations are associated with SMC proliferation and VEGF release in human cerebrovascular SMCs (62). Cells expressing US28 may also influence neighboring cells in a paracrine manner, effectively reprogramming these cells to display a more malignant phenotype. This notion is further supported by our data on disease progression and patient survival. An increased number of cells showing HCMV IEA and STAT3 phosphorylation in the glioblastoma specimens were associated with a poor prognosis, with median OS and TTP significantly reduced. A similar trend was observed for the presence of US28. That US28 abundance correlated with the degree of STAT3 phosphorylation in the tumor specimen indicates a potentially important role for US28 in glioblastoma: US28 and other viral factors that induce NF-κB signaling in infected cells may initiate a positive feedback loop activating the IL-6–STAT3 axis, thereby contributing to the severity of disease.

We suggest the following model for US28-induced STAT3 activation and subsequent proliferative signaling (Fig. 8). Cells carrying US28 produce IL-6 by way of a G protein–dependent pathway involving NF-κB. IL-6 binds to the IL-6Rα subunit, subsequently activating the gp130 subunits of the IL-6 receptor, eliciting tyrosine phosphorylation of STAT3 by JAK1. Phospho-STAT3 dimerizes and translocates to the nucleus to regulate target genes, including proliferative and angiogenic factors such as cyclin D1 (64) and VEGF (35), respectively. IL-6 itself is a STAT3 target gene (53), leading to the creation of a positive feedback loop. Our observations imply that in cells infected with HCMV, US28 initiates a pathway leading to secretion of IL-6 and resulting in enhanced proliferative signaling in an autocrine feedback loop. In addition, IL-6 stimulates STAT3 activation in neighboring uninfected cells through paracrine signaling. By locally altering cytokine concentrations, US28 may facilitate tumor progression and vascularization, thereby contributing to the oncomodulatory properties of HCMV. Together, our data indicate that the HCMV-encoded chemokine receptor US28 mediates proliferative signaling by establishing a positive feedback loop involving activation of the IL-6–STAT3 axis. Targeting the IL-6–STAT3 axis with inhibitors effectively inhibited US28-induced proliferative signaling and VEGF secretion; the IL-6–STAT3 axis may thus represent a target for antiangiogenic therapy and specifically for HCMV-related tumor formation.

Fig. 8

Model outlining the US28 positive feedback loop. (A) US28 activates STAT3 through autocrine stimulation initiated by inducing IL-6 production via the NF-κB pathway. IL-6 subsequently activates the IL-6 receptor, which results in STAT3 phosphorylation and activation of its target genes (among others, VEGF). (B) IL-6 secreted by US28-expressing cells can also activate STAT3 in a paracrine fashion. In both cases, IL-6 is a target gene of STAT3 and activation of the IL-6–STAT3 axis may result in a positive feedback loop. Arrows indicate direct interactions; dashed arrows indicate activation with intervening steps.

Materials and Methods

Materials and reagents

Antibodies directed against phospho-STAT3 (Tyr705) and STAT3 were purchased from Cell Signaling Technology. Neutralizing antibody against murine IL-6 was from BD Biosciences. Secondary antibodies used in the immunofluorescence experiments were obtained from Invitrogen. P6 (pan-JAK inhibitor), Tyrene CR4 (Abl inhibitor), AG-490 (JAK2 inhibitor), JSI-124 (STAT3 inhibitor), BAY11-7082 (NF-κB inhibitor), and PP-2 (Src inhibitor) were purchased from Calbiochem. Stocks were made in dimethyl sulfoxide, except for JSI-124, which was dissolved in ethanol. The inhibitors were subsequently diluted in the culture medium. Madindoline A was purchased from Alexis Biochemicals. Tris base, PTX, and linear polyethylenimine (25 kD) were obtained from Sigma-Aldrich; other chemicals were obtained from Applichem. Recombinant human CCL5, human CX3CL1, and murine IL-6 were obtained from PeproTech. Dulbecco’s modified Eagle’s medium (DMEM) was purchased from PAA Laboratories. Fetal bovine serum was purchased from Integro, and bovine serum was purchased from Invitrogen.

Cell culture

Human HEK293T, human U373 MG, and murine NIH 3T3 cells were cultured in DMEM supplemented with penicillin (50 IU/ml), streptomycin (50 μg/ml), and 10% fetal bovine, heat-inactivated fetal bovine, and bovine sera, respectively. Transient transfections of HEK293T cells were performed with the polyethylenimine method (65). Transient transfections of U373 cells were performed with the Lipofectamine method (66). The HCMV Titan strain described in (4) was used to infect U373 cells at an MOI of 2. Stable clones of NIH 3T3 expressing US28 or US28-R129A mutant (4) were kept under a selective pressure of neomycin (400 μg/ml) in the culture medium to ensure homogenous receptor expression in the cells. Expression of US28 in HEK293T, U373 MG, and NIH 3T3 cells was confirmed with [125I]CCL5 binding (specific binding was determined with 10−7 M CX3CL1) as previously described (13).

Angiogenesis array

A mouse angiogenesis array (RayBiotech) was used according to the manufacturer’s instructions to determine relative amounts of cytokines and chemokines involved in angiogenesis.

Measurement of IL-6 concentration in culture medium

IL-6 concentrations in culture medium supernatant from serum-starved U373 MG cells were measured 24 hours after infection with either HCMV Titan wild type or HCMV Titan ΔUS28. For the measurement, the Human IL-6 Quantikine ELISA kit from R&D Systems was used according to the manufacturer’s instructions.

Western blot analysis

A Bio-Rad minigel system was used to perform SDS–polyacrylamide gel electrophoresis, and a Bio-Rad electroblot system was used to transfer protein samples to a 0.45-μm nitrocellulose membrane. Cells were lysed in radioimmunoprecipitation assay buffer supplemented with α-Complete Protease Inhibitor Cocktail from Hoffmann–La Roche and with 1 mM phenylmethylsulfonyl fluoride, 1 mM NaVO4, and 1 mM NaF added. Samples were normalized with the BCA total protein determination kit obtained from Thermo Fisher Scientific. Blots were quantified with ImageJ (W. S. Rasband, National Institutes of Health). The data are representative of three independent experiments.

Reporter gene analysis

The Ly6E STAT3–response element luciferase construct (28) was used for STAT3 activity measurements. To determine STAT3 activity, we transfected 106 HEK293T cells with 625 ng of Ly6E STAT3–response element luciferase construct, with or without pcDEF3 containing a gene encoding HA-US28. U373 cells were transfected 24 hours before infection with 1300 ng of STAT3 reporter gene. When inhibitors were used, they were added immediately after transfection. Total DNA amounts were kept constant by addition of empty vector. Luciferase activity was measured 24 hours after transfection or 48 hours after infection with a Victor2 multilabel plate reader from Perkin-Elmer. The NFAT reporter gene pNFAT-luc was purchased from Stratagene. To measure VEGF promoter activation, we used the pGL2-VEGF-luciferase construct provided by G. Pages (Institute of Signalling Development Biology and Cancer, Nice, France).

IL-6 knockdown

IL-6 knockdown experiments were performed with the pRS-puro–shIL-6 (GAACTTATGTTGTTCTCTA) construct provided by D. Peeper (Netherlands Cancer Institute, Amsterdam, the Netherlands) (23). For the knockdown, HEK293T cells were transiently transfected with shIL-6 and a STAT3, VEGF, or NFAT reporter gene. Luminescence measurements were made 48 hours after transfection. As a negative control, pRS-puro expressing a nonsense sequence (CCAATGCTTTGATGCCAAA) was used.

[3H]Thymidine incorporation

Cell proliferation in NIH 3T3 cells was measured by [3H]thymidine incorporation. [6-3H]Thymidine was obtained from GE Healthcare Life Sciences. Cells were serum-starved overnight before labeling in DMEM supplemented with 0.5% bovine serum containing [6-3H]thymidine (1 μCi/ml) together with inhibitors.

Focus formation assay

Focus formation potential of NIH 3T3 cells transfected with either US28 or empty vector was determined by incorporating transfected cells into a monolayer of native NIH 3T3 cells as previously described (4). Treatment with JSI-124 was initiated 48 hours after seeding. Medium containing JSI-124 or vehicle was replaced every 48 hours. After 2 weeks, the cells were washed three times with phosphate-buffered saline and subsequently fixed in cold methanol for 10 min. Subsequently, the cells were stained with 0.4% methylene blue in H2O and the foci were counted.

Inositol phosphate accumulation

U373 MG cells were labeled 48 hours after infection in inositol-free DMEM supplemented with myo-[2-3H]inositol (2 μCi/ml) and incubated overnight. The cells were subsequently incubated for 2 hours in DMEM with 10 mM LiCl added. The incubation was stopped with 10 mM cold formic acid, and inositol phosphates were isolated by anion-exchange chromatography (Dowex AG1-X8 columns, Bio-Rad).

Generation of a rabbit antiserum against US28

To generate polyclonal antisera, we immunized rabbits with a synthetic peptide corresponding to the C-terminal 17 amino acids of the US28 protein (H2N-SSDTLSDEVCRVSQIIP-CO2H) (67) coupled to keyhole limpet hemocyanin through the N-terminal amino group. The specificity of the antisera was assessed by immunoprecipitation followed by Western blot and immunofluorescence analyses (fig. S1). Immunofluorescence was performed with Alexa Fluor 568–conjugated antibody against rabbit to detect antibody against US28, and subsequent imaging was done on an Olympus FSX-100.

Patient samples and immunohistochemistry

Paraffin-embedded primary brain tumor specimens were obtained from one HCMV-negative glioblastoma tumor and randomly from 20 glioblastoma patients in our sample collection in the Biobank at the Karolinska University Hospital Sweden. Twelve of 21 patients received temozolomide and radiation, 2 patients received gamma knife treatment, 5 patients received CCNU (lomustine) and radiation, and 2 patients received only CCNU after debulking surgery. Tissue sections (6-μm thickness) were stained for HCMV IEA, US28, phospho-STAT3, CD31, and IL-6 with immunohistochemistry staining protocols as previously described (68). Briefly, the sections were deparaffinized in xylene, rehydrated through an alcohol series, postfixed with 4% neutral buffered formalin, treated with pepsin (Biogenex), and then incubated in citrate buffer (Biogenex). The sections were treated with 3% H2O2 (Sigma-Aldrich) to inactivate endogenous peroxidase; Avidin/Biotin Blocking kit (DakoCytomation) was used to block endogenous biotin-avidin, and Fc receptor (FcR) blocker (Innovex Biosciences) to block FcR. Finally, the tissue sections were treated with Background Buster (Innovex Biosciences). Incubation with primary antibodies against US28 (created by A. Fraile-Ramos), phospho-STAT3 (Cell Signaling Technology), CD31 (DakoCytomation), IL-6 (Abcam), HCMV IEA (Chemicon), and HCMV late (Chemicon) was done overnight at 4°C. Antibodies against rabbit immunoglobulin G (IgG; R&D Systems) and SMC α-actin (IgG2a, Dako) were used as isotype controls. After incubation with primary antibodies, the sections were incubated with biotinylated antibodies against rabbit (Dako) or mouse (Biogenex). Finally, the antibodies were visualized with streptavidin-conjugated horseradish peroxidase and diaminobenzidine (Innovex Biosciences). For double-staining, the first staining was performed as described, and the second was performed with streptavidin-conjugated alkaline phosphatase (DakoCytomation) and FastRed (DakoCytomation). Hematoxylin (Sigma-Aldrich) was used for counterstaining, and slides were mounted in permanent mounting medium (DakoCytomation).

The percentage number of cells expressing different factors in the tissue specimens was graded as follows: grade 1, 0 to 1+ to <10%; grade 2, >10 to 30%; grade 3, >30 to 50%; grade 4, >50 to 70%; grade 5, >70%. This study was approved by the ethics committees at Karolinska Institutet.

Statistical analysis

All experiments were performed at least three times in triplicate. When comparisons between treated and vehicle-treated cells or mock and infected cells were made, Student’s t test was performed with the GraphPad Prism software. Bars and error bars represent the mean and SEM, respectively. Median TTP and survival were analyzed with Kaplan-Meier curves and log-rank (Mantel-Cox) test. The relationship between different stainings was determined with Wald statistics.

Acknowledgments

Acknowledgments: We thank M. Detlef (Abteilung Virologie, Universitätsklinikum Ulm, Germany) for providing both the HCMV Titan wild type and ΔUS28 strain. We also thank D. Peeper (Netherlands Cancer Institute, Amsterdam, the Netherlands) for providing the IL-6 shRNA, C. M. Horvath (Northwestern University, Chicago, IL) for providing the Ly6E STAT3–response luciferase construct, and M. Marsh (Medical Research Council Laboratory for Molecular Cell Biology, London, UK) for his assistance in the generation of the US28 antibody. Funding: Netherlands Organisation for Scientific Research grant ECHO 700.55.010 (to M.J.S.) and Swedish Medical Research Council grants 10350 and K2007-56X-12615-10-3, Swedish Children’s Cancer Foundation Projects 05/100 and 07/065, and Cancer Foundation grant 5044-B05-01XAB (to C.S.-N.). Author contributions: E.S., D.M., A.S., A.R., and S.A.L. performed the experiments and analyzed the data; E.S., M.S., and M.J.S. designed the experiments and wrote the paper. A.F.-R. generated and characterized the US28 antibody. C.S.-N. and S.A.L. proofread and corrected the paper. Competing interests: C.S.-N. has served on an advisory board and has given paid lectures for Roche on the topic of this paper.

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/3/133/ra58/DC1

Fig. S1. Western blot analysis of US28-expressing NIH 3T3 cells and U373 MG cells infected with either HCMV wild-type virus or the ΔUS28 deletion strain.

References and Notes

View Abstract

Stay Connected to Science Signaling

Navigate This Article