Detection of human cytomegalovirus in medulloblastomas reveals a potential therapeutic target

Medulloblastomas are the most common malignant brain tumors in children. They express high levels of COX-2 and produce PGE₂, which stimulates tumor cell proliferation. Human cytomegalovirus (HCMV) is prevalent in the human population and encodes proteins that provide immune evasion strategies and promote oncogenic transformation and oncomodulation. In particular, HCMV induces COX-2 expression; STAT3 phosphorylation; production of PGE₂, vascular endothelial growth factor, and IL-6; and tumor formation in vivo. Here, we show that a large proportion of primary medulloblastomas and medulloblastoma cell lines are infected with HCMV and that COX-2 expression, along with PGE₂ levels, in tumors is directly modulated by the virus. Our analysis indicated that both HCMV immediate-early proteins and late proteins are expressed in the majority of primary medulloblastomas. Remarkably, all of the human medulloblastoma cell lines that we analyzed contained HCMV DNA and RNA and expressed HCMV proteins at various levels in vitro. When engrafted into immunocompromised mice, human medulloblastoma cells induced expression of HCMV proteins. HCMV and COX-2 expression correlated in primary tumors, cell lines, and medulloblastoma xenografts. The antiviral drug valganciclovir and the specific COX-2 inhibitor celecoxib prevented HCMV replication in vitro and inhibited PGE₂ production and reduced medulloblastoma tumor cell growth both in vitro and in vivo. Ganciclovir did not affect the growth of HCMV-negative tumor cell lines. These findings imply an important role for HCMV in medulloblastoma and suggest HCMV as a novel therapeutic target for this tumor.

Medulloblastomas are molecularly and clinically diverse tumor types that together encompass the most frequent malignant brain tumors in children (1). Despite intensive multimodal therapy, survival in high-risk patients is still poor, and long-term survivors frequently experience physical and neuropsychological sequelae (2, 3), underscoring the need for new therapeutic options. Although medulloblastomas contain significantly fewer mutations than adult cancers (4), specific subsets have been identified, and aberrations of significant developmental pathways, including the Hedgehog and Wnt pathways, as well as frequent mutations of the MLL family genes have been described. However, since many medulloblastomas have no apparent mutations of any known cancer gene, it has been suggested that epigenetic changes also may be responsible for tumor initiation (4). Hence, additional etiological factors clearly connected to medulloblastoma development remain to be identified.

HCMV DNA and proteins are found in tumors of various origins, including 90%–100% of glioblastomas and prostate, breast, and colon cancers (5–7). This virus is not considered to be oncogenic; rather, its potential role in cancer seems to be oncomodulatory (8). Its 230-kbp genome has 252 open reading frames and encodes approximately 200 proteins, of which only about 50 are known to be essential for viral replication. Hence, the vast majority of HCMV proteins fulfill other functions in the viral life cycle and affect cellular and immunological functions relevant to tumor biology and immune surveillance. HCMV proteins control telomerase activity, cellular differentiation, proliferation (through specific effects on the cell cycle), apoptosis, epigenetic functions, migration, and angiogenesis (8–10). HCMV induces inflammation, but in parallel, it escapes immune recognition (8, 10). These tumor-promoting functions of HCMV proteins, together with the detection of HCMV DNA and proteins in tumors of different origins, suggest that HCMV has a pivotal, hitherto unknown role in the tumorigenesis of certain cancers (8).

The inflammatory mediator PGE₂, a COX-2 metabolite implicated as a growth factor for tumor cells, drives cancer progression through its ability to stimulate cell proliferation, promote angiogenesis and invasion, inhibit apoptosis, and suppress immune responses (11–13). For unknown reasons, many tumors, including medulloblastomas, express COX-2, and inhibition of PGE₂ synthesis with nonsteroidal antiinflammatory drugs induces apoptosis in medulloblastoma cell lines and significantly inhibits the growth of medulloblastomas in vivo (14). Aspirin reduces the incidence and growth of various cancer forms in animal models (15, 16), an effect that may be mediated at least in part by inhibition of COX-2 enzymes, resulting in decreased production of prostaglandins and other inflammatory mediators. In clinical trials, aspirin given daily to prevent vascular events for at least 4 years reduced death from different forms of cancer both during and after the trials (17). In clinical studies, nonspecific and specific COX-2 inhibitors have shown promise as anticancer agents (16, 18–20).

Interestingly, the HCMV protein US28, a constitutively active chemokine receptor homolog, induces COX-2 expression and results in STAT3 phosphorylation, which increases production of vascular endothelial growth factor and IL-6 as well as tumor formation in vivo (21). Both aspirin and selective COX-2 inhibitors efficiently prevent viral production, which is partly dependent on PGE₂ (22, 23). Transgenic mice that express US28 in intestinal epithelial cells exhibit a hyperplastic intestinal epithelium and tumor development. US28 expression in intestinal epithelial cells inhibits glycogen synthase–3β (GSK-3β) function, promotes accumulation of β-catenin, and increases expression of Wnt target genes involved in the control of cell proliferation (24); these findings provide a direct molecular link between this HCMV protein and oncogenesis. Moreover, we recently demonstrated that the HCMV protein IE72 induces telomerase activity, which is likely another important event in tumor transformation (25).

In this study, we examined the prevalence of HCMV in medulloblastomas. We also assessed the connection between HCMV and COX-2 expression in tumors and sought to determine whether inhibition of viral replication with antiviral treatment and COX-2 inhibition affects tumor growth in vitro and in vivo.

High prevalence of HCMV in primary medulloblastomas, medulloblastoma cell lines, and medulloblastoma xenografts. Of the 37 primary medulloblastomas we examined for HCMV, 34 (92%) expressed immediate-early (IE) proteins and 27 (73%) expressed late proteins (Table 1). The IE and late proteins were widely expressed in cells within the tumor (Figure 1), as determined by immunohistochemistry (n = 37, Figure 1, A, B, D, E, M, and N) or indirect immunofluorescence (n = 6, Figure 1, J, K, and L) in frozen tissue sections. Detection of IE proteins with two different IE-specific antibodies resulted in nuclear (Figure 1) or nuclear/cytoplasmic (see below) staining patterns. In situ hybridization confirmed the presence of HCMV in primary medulloblastomas (Figure 1, G–I). Flow cytometric analysis of these tumors revealed expression of HCMV IE and UL83 (pp65 in Figure 1O) proteins in 90%–95% of the cells (n = 2). Six available frozen tissue samples were confirmed to contain HCMV IE, pp150, and gB DNA and IE and gB RNA by TaqMan PCR (data not shown). Sequencing of the PCR products revealed that the HCMV genome in medulloblastoma cell lines was distinctly different from that in 4 commonly used laboratory strains (Merlin, AD169, TB40E, and VR1814; data not shown), which excluded the possibility of contamination.

Figure 1

Detection of HCMV DNA and proteins in a majority of primary medulloblastoma tissue samples. (A–F) Immunohistochemical assessment of HCMV IE (A and D) and late (B and E) proteins in primary medulloblastomas; α-actin served as an isotype control antibody (C and F). Original magnification, ×20 (A–C) and ×40 (D–F). (G–I) Confirmation of immunohistochemical findings by in situ hybridization. (G) HCMV DNA. (H) Negative control. (I) Positive control (Alu). Original magnification, ×40, inset, ×100 (G); ×40 (H); ×20 (I). (J–L) Indirect immunofluorescence staining for HCMV IE (J) and late (K) antigens in frozen sections of a primary medulloblastoma, and isotype control for immunofluorescence (L). Original magnification, ×20 (J–L). (M) Quantification of HCMV expression in medulloblastomas positive for HCMV IE and late proteins. (N) HCMV expression graded 1–4 according to the percentage of HCMV-positive cells in 37 primary medulloblastoma tissue samples (see Methods for details). L, late. (O) Flow cytometric analysis of HCMV UL83 expression in two primary medulloblastomas (M2 and M3). Isotype antibodies served as a negative control.

Table 1

IE and late protein expression in medulloblastoma primary tumors

Next, we analyzed 8 human medulloblastoma cell lines (D324 MED, D283 MED, UW228-3, MEB-MED-8A, D458 MED, PFSK-1, D384 MED, and D425 MED) for HCMV DNA and RNA by PCR. Surprisingly, HCMV IE and pp150 DNA and IE and gB RNA were detected at low levels in all 8 cell lines (Supplemental Table 1; supplemental material available online with this article; doi: 10.1172/JCI57147DS1). FISH analyses with a probe utilizing the whole HCMV genome in bacterial artificial chromosome clones revealed HCMV DNA in medulloblastoma cells and in HCMV-infected (AD169) MRC-5 human fibroblasts, but not in K562 cells (control human chronic myelogenous leukemia cells; Figure 2A). To examine whether the viral genome was integrated in human chromosomes, we performed FISH analyses on colcemid-treated D324 MED cells. No sign of integration of HCMV DNA in chromosomes was observed in metaphase cells (Figure 2A).

Figure 2

Detection of HCMV DNA and proteins in medulloblastoma cell lines and human medulloblastoma xenografts. (A) FISH analysis showing the presence of HCMV DNA (red) in colcemid-treated D324 MED cells and HCMV-infected (AD169) MRC-5 cells in metaphase, but not in uninfected K562 cells. (B) Flow cytometric analysis of D324 MED cells stained for the HCMV IE antigen or an isotype control antibody (IgG1). (C) Flow cytometric analysis of D283 MED cells double-stained for HCMV IE and CD133. (D) Detection of HCMV IE and late proteins in established human medulloblastoma xenografts in NMRI nu/nu mice. α-Actin served as an isotype control.

Further analysis by flow cytometry revealed HCMV IE and pp65 protein expression in 9%–39% of the cells; D324 MED cells stained for IE are shown in Figure 2B, and UW228-3 cells stained for IE and pp65 are shown in Figure 3B. On different sampling occasions over a 12-month period, HCMV proteins were expressed in 0.2%–75% of cells in individual medulloblastoma cell lines. Given this consistent presence of viral DNA, we hypothesized that HCMV resides in cancer stem cells. To test this hypothesis, we double stained the medulloblastoma cell line D283 MED for CD133 (a stem cell marker) and HCMV IE proteins. 57% of the CD133-positive cells coexpressed HCMV IE proteins (Figure 2C). When D283 MED cells were xenografted subcutaneously in NMRI nu/nu mice, HCMV protein expression distinctly increased, and a majority of the tumor cells were positive for HCMV IE proteins. While IE protein expression was diffuse in the xenografts, HCMV late protein expression was distinctly located adjacent to vessels (Figure 2D).

Figure 3

COX-2 expression in primary medulloblastomas and cell lines is restricted to positivity for HCMV proteins. The HCMV protein UL83 and COX-2 are coexpressed in primary medulloblastomas, as shown by (A) immunohistochemical double staining for COX-2 (green) and either HCMV IE1/IE2 protein (red) or HCMV protein UL83 (red). Nuclei were stained with DAPI (blue). Isotype antibodies served as negative control. Original magnification, ×40. (B) Flow cytometric analysis of UW228-3 cells. Top row: Cells were stained for isotype control (IgG1), IE antigen (IEA), or UL83. Bottom row: Cells were stained with the secondary antibodies goat anti-mouse (GAM) and swine anti-rabbit (SAR) (negative control) or for IE antigen and COX-2.

HCMV-positive medulloblastoma cells express COX-2. Since medulloblastomas express COX-2 and HCMV induces expression of COX-2 (14, 21), we investigated whether HCMV protein expression correlated with COX-2 expression in tumor cells. In frozen tissue sections of primary medulloblastomas, HCMV protein expression correlated with COX-2 expression (Figure 3A). IE proteins were expressed in both the nucleus and cytoplasm of tumor cells. In medulloblastoma cell lines, cells positive for the HCMV IE pp65 also expressed COX-2 (Figure 3B). HCMV and COX-2 expression also correlated in human medulloblastoma xenografts in mice (data not shown). These observations imply that HCMV may control the expression of COX-2 in medulloblastomas.

HCMV infection induces COX-2 expression and PGE₂ synthesis in medulloblastoma cells. HCMV induces COX-2 expression in fibro­blasts and epithelial and smooth muscle cells (22, 23). We hypothesized that HCMV also regulates COX-2 expression in medulloblastomas. In vitro infection (superinfection) of medulloblastoma cells (D324 MED) with HCMV resulted in increased IE RNA expression at 3 hours after infection (relative mRNA expression, 0.04296, n = 3 in duplicate) that greatly increased at 6–72 hours after infection (Figure 4A). The late HCMV pp150 RNA was detectable at 3 hours after infection (relative mRNA expression, 0.00316, n = 3 in duplicates), and was greatly increased at 24 hours after infection (Figure 4A). Both IE and pp150 were detectable at low levels in the uninfected cells (mean relative mRNA expression, 0.0000322 and 0.00000132, respectively; n = 3; 7 time points in duplicate). Interestingly, expression of COX-2 mRNA was already increased at 1 and 3 hours after infection in cells infected with both HCMV- and UV-treated virus inoculum (Figure 4A), before induced IE expression was detected. Also, an increased level of the COX-2 protein was detectable at 3 hours after infection (Figure 4B). As US28 has been demonstrated to induce COX-2 expression, we examined whether US28 was expressed at early time points after infection. We found that US28 mRNA could be detected in cells already at 15 minutes after infection: low US28 mRNA levels were also detected in UV-HCMV–infected cells. This was consistent with the increased expression of COX-2 mRNA observed in both HCMV- and UV-HCMV–infected cells at 1 and 3 hours after infection. At 3 days after infection, HCMV had significantly increased PGE₂ synthesis (Figure 4C), which was significantly reduced when D324 MED cells were treated with ganciclovir, celecoxib, or both drugs (Figure 4D); this was consistent with the detection of COX-2 mRNA in HCMV-infected but not UV-HCMV–infected cells at 72 hours after infection (Figure 4A).

Figure 4

HCMV induces COX-2 expression and PGE₂ synthesis in medulloblastoma cells. (A) HCMV IE mRNA, US28 mRNA, and low levels of the late gene pp150 mRNA were expressed in D324 MED cells after in vitro infection with HCMV VR1814 as shown by quantitative PCR at 15 minutes to 72 hours after infection. COX-2 mRNA was detected when US28 was expressed, but earlier than IE mRNA. (B–D) Superinfection of HCMV VR1814 increases COX-2 expression and PGE₂ levels in D324 MED cells as shown by Western blotting (B). ctr, control. (C) PGE₂ levels measured by ELISA 48 hours after infection with HCMV VR1814. (D) Ganciclovir (gan) and celecoxib (cel) significantly inhibit PGE₂ synthesis in D324 MED cells as measured by PGE₂ ELISA. *P < 0.05, ***P < 0.0001 (1-way ANOVA). Values are mean ± SD.

Antiviral treatment and COX-2 inhibition significantly inhibit medulloblastoma growth in vitro and in vivo. As HCMV appears to be present in a majority of medulloblastoma tumors, and viral mechanisms control COX-2 expression, we sought to test whether inhibition of viral replication by antiviral drugs or COX-2 inhibitors affects tumor growth in vitro and in vivo. Both specific and nonspecific COX-2 inhibitors can prevent virus-mediated accumulation of PGE₂ and efficiently inhibit viral production, as viral replication appears to be dependent on PGE₂ (22, 23). Therefore, we examined whether ganciclovir, which targets the DNA polymerase during HCMV replication, or the COX-2 inhibitor celecoxib, which prevents HCMV replication by decreasing PGE₂ levels, affects medullo­blastoma growth in vitro.

In clonogenic assays with the native medulloblastoma cells, treatment with increasing concentrations of ganciclovir reduced the clonogenic capacity of D324 MED (P < 0.0001), D283 MED (P < 0.0001), and UW228-3 (P < 0.001, Figure 5A). Since none of the screened medulloblastoma cell lines were HCMV negative, we tested the effect of ganciclovir on the growth of two tumor cell lines that were negative for HCMV by PCR and flow cytometry (data not shown): PC-3 (established from a bone metastasis of a grade IV prostatic adenocarcinoma) and BxPC-3 (established from a pancreatic adenocarcinoma). Treatment with increasing concentrations of ganciclovir had no significant effect on the clonogenic capacity of either cell line in the presence of ganciclovir (37.5, 75, or 150 μM) (Figure 5A).

Figure 5

Ganciclovir and celecoxib significantly impair the growth of medulloblastoma cell lines in vitro. (A) Ganciclovir (gan) significantly inhibits the clonogenic capacity of D324 MED, D283 MED, and UW228-3 cells in a dose-dependent manner but has no growth-inhibiting effect on HCMV-negative PC-3 or BxPC-3 cells. ***P < 0.0001 (1-way ANOVA). (B) Ganciclovir significantly augments the growth-inhibitory effect of celecoxib in D324 MED, UW228-3, and D283 MED cells. At the same concentrations that significantly inhibited the growth of medulloblastoma cells, neither drug alone nor the combinations of the two drugs affected the survival of MRC-5 cells in survival assays (MTT) in human MRC-5 fibroblasts. Values are mean ± SD.

Celecoxib also reduced the clonogenic capacity of D324 MED (P < 0.0001) and D283 MED (P < 0.0001) but not of UW228-3 (Figure 5B). In combination, valganciclovir and celecoxib significantly reduced the clonogenic capacity of medulloblastoma cell lines compared with single treatments alone (D324 MED, P < 0.0001; D283 MED, P < 0.0001; UW228-3, P < 0.001; Figure 5B). To further test whether ganciclovir had a general cytotoxic effect, we performed survival assays (MTT) in human fibroblasts. At the same concentrations that significantly inhibited the growth of medulloblastoma cells, neither drug alone nor the combinations of the two drugs affected the survival of MRC-5 cells (Figure 5B). These data imply that the antitumor response in HCMV-positive medulloblastoma cell lines reflects a specific antitumor effect involving HCMV rather than cytotoxicity.

To further investigate whether the growth-suppressing effect of ganciclovir and celecoxib is dependent on the presence HCMV in medulloblastoma cells, we treated D283 MED cells with different concentrations of ganciclovir (37.5, 75, and 150 μM). After 2 weeks of continuous treatment, viral protein expression in cells treated with 75 μM ganciclovir had decreased substantially (data not shown and Figure 6A), but neither HCMV protein expression nor HCMV DNA was completely eliminated (Figure 6A). Although D283 MED cells pretreated with ganciclovir (D283 MED^gan) were able to grow in soft agar, they formed significantly fewer colonies than untreated D283 MED (D283 MED^WT) cells (P < 0.05; Figure 6A). The clonogenic capacity of D283 MED^gan cells was not affected by increasing concentrations of ganciclovir or celecoxib but was significantly reduced by the combination of the two drugs (Figure 6A).

Figure 6

Antiviral treatment reduces CMV protein expression and clonogenic capacity of medulloblastoma cells in vitro. (A) D283 MED cells pretreated with 75 μM ganciclovir (D283 MED^gan) exhibited markedly reduced HCMV IE and pp65 protein expression (left panel), had significantly reduced clonogenic capacity (*P < 0.05, t test; upper-right panel, and were resistant to ganciclovir (0–150 μM) or celecoxib treatment but demonstrated reduced clonogenic capacity in treatment with a combination of ganciclovir and celecoxib (right lower panel). (B) Ganciclovir and celecoxib also inhibited the clonogenic capacity of D324 MED, D283 MED, and UW228-3 cells superinfected with HCMV VR1814. *P < 0.05, ***P < 0.0001 (1-way ANOVA). Values are mean ± SD.

In clonogenic assays with HCMV-superinfected medulloblastoma cells, treatment with increasing concentrations of ganciclovir or celecoxib also significantly reduced the tumorigenic capacity of HCMV-D324 MED, HCMV-D283 MED, and HCMV-UW228-3 cells in a dose-dependent manner (Figure 6B). Combining cele­coxib with ganciclovir augmented the effect; growth was reduced by 81%, 84%, and 97% (P < 0.0001, Figure 6B), respectively, in superinfected cells. Consistent with these observations, valganciclovir and celecoxib alone or in combination also reduced PGE₂ production in D324 MED cells (Figure 4D).

To further investigate the therapeutic effects of valganciclovir and celecoxib in vivo, we established a human medulloblastoma xenograft model using D283 MED cells in NMRI nu/nu mice. We first determined the bioavailability of valganciclovir in these mice. After administration of valganciclovir at 100 mg/kg, the plasma ganciclovir concentration was sufficient for a potential therapeutic effect (Figure 7A). To observe a combinatorial effect with celecoxib, we reduced the valganciclovir dose. Mice were randomized to daily treatment for 12 days with oral valganciclovir (14 mg/kg twice daily) or oral celecoxib (90 mg/kg once daily), starting at a tumor volume of 0.1 ml. The mice were euthanized, and the tumors were collected for analysis. HCMV protein expression in the xenografts was strongly increased compared with that in the cells in culture (Figure 2D). Each drug inhibited the growth of established medulloblastomas by approximately 40% (valganciclovir, P < 0.0001; celecoxib, P < 0.0001; Figure 7, B and C). In combination, celecoxib and valganciclovir reduced tumor growth by 72% (tumor volume index [TVI], 3.3 versus 10.3 in untreated controls, P < 0.0001; Figure 7, B and C). The combined drug regimen reduced the number of clusters positive for HCMV late protein in xenograft tumors by 80% (P < 0.0001; Figure 7, D and E).

Figure 7

Ganciclovir and celecoxib significantly impair the growth of established human medulloblastoma xenografts in NMRI nu/nu mice. (A) Pharmacokinetics of valganciclovir (val) (plasma concentration versus time) after 1 oral dose (100 mg/kg). Solid line represents concentration-time data estimated by pharmacokinetic modeling. NMRI nu/nu mice were engrafted with 7 × 10⁶ D283 MED cells subcutaneously and randomized to receive 14 mg/kg valganciclovir twice daily, 90 mg/kg celecoxib daily, both drugs, or no treatment. All treatments were given orally through a gastric feeding tube for 12 days, starting when tumors were 0.1 ml (mean, 0.13 ml). (B and C) Valganciclovir augments the inhibitory effect of celecoxib on medulloblastoma growth in vivo, as shown by the TVI (B) and tumor weight (C) at autopsy. TVI values are mean ± SD; tumor weight is shown as median and quartiles (TVI, 2-tailed ANOVA; tumor weight, 1-way ANOVA). (D and E) Clusters of cells positive for HCMV late protein in xenograft tissues were reduced by 80% by valganciclovir and celecoxib treatment. Untreated (n = 5), 75 ± 30; valganciclovir + celecoxib (n = 7), 15 ± 5. Values are mean ± SD. ***P < 0.0001 (t test). NMRI nu/nu mice were also engrafted subcutaneously with the HCMV-negative cell lines PC-3 (30 × 10⁶ cells) and BxPC-3 (5 × 10⁶ cells) and randomized to receive 14 mg/kg valganciclovir by gastric feeding twice daily or no treatment for 23 days (PC-3) or 32 days (BxPC-3). (F and G) PC-3 or BxPC-3 xenograft growth was not inhibited in mice treated with ganciclovir, as shown by the TVI (F) and tumor weight (G) at autopsy. TVI values are mean ± SD; tumor weight is shown as median.

We also established xenografts of the two HCMV-negative tumor cell lines (PC-3 and BxPC-3) to test the effect of ganciclovir on tumor cell growth. In NMRI nu/nu mice carrying either PC-3 or BxPC-3 xenografts, tumor growth was unaffected by treatment with oral valganciclovir (14 mg/kg twice daily) (tumor weight: PC-3, P = 0.4810; BxPC-3, P = 0.7237; Figure 7, F and G).

This study shows that HCMV proteins are expressed in the majority of primary medulloblastoma tissue samples. Surprisingly, HCMV DNA, RNA, and proteins were detected in 8 examined medulloblastoma cell lines, and viral protein expression was highly induced by xenografting of human medulloblastoma cells in vivo. Sequencing of PCR products excluded the possibility that the cells lines were contaminated by a laboratory strain. Targeting of HCMV by antiviral or COX-2–inhibitory drugs with different modes of action on viral replication significantly impaired tumor growth in vivo. These observations imply a persistent source of HCMV in neuroprecursor cells and suggest that the virus has a pathogenetic role in medulloblastoma rather than representing an epiphenomenon. Importantly, our findings also suggest HCMV as a novel therapeutic target for this tumor.

Medulloblastomas can originate from neural precursor cells in the external granule layer in the cerebellum or from cells in the dorsal brainstem (26). Medulloblastoma cells exhibit many characteristics similar to those of progenitor cells of the embryonic brain, further suggesting the presence of a tumor-initiating or cancer stem cell population in these tumors (26). This statement is supported by the clinical presentation and treatment response and by the fact that cells positive for stem cell markers (e.g., CD133 or CD15 in primary medulloblastomas) give rise to new tumors in xenograft models (27, 28). Since HCMV appears to persist for years in laboratory cell lines, the virus is most likely maintained in a stem-like cell phenotype within this population. Indeed, we observed HCMV protein expression in 57% of CD133-positive D283 MED cells. The viral genome did not appear to be integrated into human chromosomes, and we concluded that HCMV integration evidently is not a major event in the medulloblastomas, although it cannot be excluded in single cells.

HCMV can infect neuronal progenitor cells and is the major infectious cause of birth defects in newborns, including sensorineural hearing loss or neuronal migration disturbances during brain development and, in the most severe cases, microcephaly or anencephaly. Previously, we showed that infection of neuronal progenitor cells blocks their ability to differentiate into neurons (29) or astrocytes (30). Interestingly, prevention of the normal differentiation of neuronal stem cells has been implicated as an important step in the early development of brain tumors (31).

We speculate that HCMV infection of a defined tumor precursor cell is a decisive early event in the disordered cerebellar development that leads to medulloblastoma (32) through HCMV-specific effects on tumorigenesis, inflammation, and immune evasion strategies. HCMV infection and expression of US28 result in STAT3 phosphorylation, Wnt activation, enhanced COX-2 expression, production of vascular endothelial growth factor, increased production of PGE₂ and IL-6, and inflammation. Transgenic mice (VS28 mice) with US28 expression targeted to intestinal epithelial cells exhibit a hyperplastic intestinal epithelium and develop adenomas and adenocarcinomas at 40 weeks of age (24). When exposed to inflammation, VS28 mice have a significantly higher tumor burden (24). These observations imply that US28 by itself promotes dysplasia and cancer development, especially during enhanced inflammation. HCMV induces inflammation both by its presence as a foreign antigen in tissues and by inducing expression of COX-2 and 5-lipoxygenase, which results in production of prostaglandins and leukotrienes (33). Here, we observed COX-2 expression in HCMV protein–positive cells in primary medulloblastoma samples and in medulloblastoma cell lines, and the induction of COX-2 expression and PGE₂ production in experimentally infected medulloblastoma cells. US28 mRNA was detected at 15 minutes and at 1 and 3 hours after infection and before the detection of HCMV IE RNA. It is likely that US28 RNA was delivered with the viral particle, a phenomenon that is known to occur during HCMV infection (34). US28 may also be carried in the virus envelope, and may directly affect COX-2 expression immediately after infection (35). Indeed, we found that induced COX-2 expression was observed in HCMV- and UV-HCMV–infected cells at 1 and 3 hours after infection. We speculate that US28 (protein and RNA) delivered by the virus particle results in the early expression of COX-2 protein, while newly transcribed US28 RNA molecules produced during viral replication were responsible for the induced COX-2 expression at 48 and 72 hours after infection. This statement is substantiated by the observation that COX-2 expression was not increased in cells infected with the UV-treated virus at 48 and 72 hours after infection.

Besides US28, many other factors may facilitate the development of cancer in HCMV-infected individuals. Several HCMV-encoded gene products control cellular pathways that could be involved in cancer development, among them cellular differentiation, cell cycle regulation, DNA damage and repair, epigenetic functions, apoptosis, cellular migration, angiogenesis, and immune evasion (9). For example, HCMV provokes chromosomal instability (36), and HCMV IE proteins induce telomerase activity (25) and exert fundamental effects on the cell cycle. HCMV IE86 binds to p53 and controls proliferation, whereas IE72 induces proliferation of cells by dysregulating PI3K/Akt activity, retinoblastoma (Rb) phosphorylation, and expression of p53 family proteins (p53, p63, and p73) (37).

HCMV UL97 phosphorylates the Rb protein and prevents apoptosis (38). Thus, interactions with the p53 and Rb families of tumor suppressor proteins will promote cell cycle progression and increase DNA synthesis while simultaneously elongating telomeres and blocking apoptosis (39–42). In primary glioblastomas, we found that mainly HCMV IE–positive cells expressed hTERT, and the HCMV IE72 protein directly interacted with the hTERT promoter at SP1 binding sites to induce telomerase activity and telomere elongation (25). This scenario will most likely be important in tumor transformation.

The two HCMV IE–specific antibodies we used to detect IE proteins in primary medulloblastoma tissue specimens resulted in different staining patterns. The CH160 clone from Abcam recognizes both IE72 and IE86, while the Chemicon antibody recognizes a 68- to 72-kDa protein, which likely explains the different staining pattern of IE protein expression revealed by direct immunofluorescence and immunohistochemistry. We and others have noted that IE protein is expressed in both the cytoplasm and the nucleus in tumor cells, a pattern that may depend on different IE splicing variants encoded by the IE gene (43).

At the same time as HCMV induces inflammation and may affect tumorigenesis, many HCMV proteins use sophisticated mechanisms that make the infected cells invisible to the immune system. The result is enhanced viral replication and tumor growth. In support of this statement, the viral load in glioblastomas appears to be associated with patient outcome. We recently observed significantly longer median survival in glioblastoma patients with low-grade HCMV infection in the tumor than those with high-grade infections. The 2-year survival rate was also greater in patients with low-grade infections. These findings imply that the HCMV infection level in glioblastomas has a high prognostic value and that HCMV may contribute to the pathogenesis of glioblastoma rather than representing an epiphenomenon of the disease (our unpublished observations).

Targeting HCMV may be a new therapeutic option for medulloblastoma patients. In medulloblastoma cell lines, treatment with ganciclovir markedly reduced the number of HCMV-positive cells but did not eliminate the virus. Furthermore, treatment of medulloblastoma cells with two drugs with different modes of action efficiently inhibited tumor growth in vitro and in vivo. In combination, the drugs reduced tumor growth by 81%–97% in clonogenic assays and by 72% in NMRI nu/nu mice carrying established human medulloblastoma xenografts. In sharp contrast, ganciclovir had no effect on the clonogenic capacity or tumor growth of two HCMV-negative cell lines (PC-3 and BxPC-3) in vitro or in vivo. Importantly, ganciclovir had no significant toxic or antiproliferative effect on non-tumor cells that were HCMV negative.

These observations imply that the inhibitory effect of ganciclovir on the growth of medulloblastoma cell lines and tumors is HCMV specific and is not mediated by potential nonspecific drug effects on cellular proliferation. Since COX-2 inhibitors efficiently prevent HCMV replication (22, 44) and reduce the growth of US28-expressing tumor cells (21, 45), the ability of COX-2 inhibitors to protect against cancer in clinical trials may be mediated by control of HCMV replication. Since HCMV has strong oncomodulatory effects, treatment to control HCMV infection in tumor tissues may suppress tumor progression, improve the patient’s immune response, and increase the tumor’s sensitivity to conventional therapy.

Antiviral drugs for HCMV and COX-2 inhibitors are well tolerated and may be a new therapeutic option to be used in combination with conventional therapies in patients carrying HCMV-infected tumors. Indeed, we are evaluating antiviral drugs against HCMV as an adjuvant therapy for glioblastoma in a double-blind, placebo-controlled pilot study of 42 patients.

In summary, our observations suggest that HCMV is a novel etiological factor linked to medulloblastoma development and that nontoxic antiinflammatory and antiviral drugs targeting HCMV may be used in addition to conventional chemotherapy to treat patients with medulloblastoma.

View Supplemental data

This work was supported by grants from Torsten and Ragnar Söderberg’s Foundations, Swedish Children’s Cancer Foundation, Swedish Cancer Society, Swedish Research Council, Märta and Gunnar V. Philipson Foundation, Mary Bevé Foundation, Hans and Märit Rausing Charitable Fund, Dämman Foundation, Swedish Society for Medical Research (SLS), Goljes Memory Foundation, Magnus Bergvalls Foundation, Swedish Society for Medical Research (SSMF), and Tore Nilsons Foundation. We thank Stephen Ordway for editing the manuscript.

Address correspondence to: Cecilia Söderberg-Nauclér, Center for Molecular Medicine, L8:03, Cell and Molecular Immunology, Karolinska Institutet, Karolinska University Hospital, SE-171 76 Stockholm, Sweden. Phone: 46851779896; Fax: 468313147; E-mail: cecilia.naucler@ki.se. Or to: John Inge Johnsen, Karolinska Institutet, Childhood Cancer Research Unit, Q6:05, Department of Women’s and Children’s Health, Karolinska Institutet, Stockholm 17176, Sweden. Phone: 46851772989; Fax: 48851773475; E-mail: John.Inge.Johnsen@ki.se.

Conflict of interest: The authors have declared that no conflict of interest exists.

Reference information: J Clin Invest. 2011;121(10):4043–4055. doi:10.1172/JCI57147.

See the related article at Viruses and human brain tumors: cytomegalovirus enters the fray.

1. Thompson MC, et al. Genomics identifies medulloblastoma subgroups that are enriched for specific genetic alterations. J Clin Oncol. 2006;24(12):1924–1931.
   View this article via: PubMed Google Scholar
2. Mulhern RK, Merchant TE, Gajjar A, Reddick WE, Kun LE. Late neurocognitive sequelae in survivors of brain tumours in childhood. Lancet Oncol. 2004;5(7):399–408.
   View this article via: PubMed CrossRef Google Scholar
3. Rood BR, Macdonald TJ, Packer RJ. Current treatment of medulloblastoma: recent advances and future challenges. Semin Oncol. 2004;31(5):666–675.
   View this article via: PubMed CrossRef Google Scholar
4. Parsons DW, et al. The genetic landscape of the childhood cancer medulloblastoma. Science. 2011;331(6016):435–439.
   View this article via: PubMed Google Scholar
5. Cobbs CS, et al. Human cytomegalovirus infection and expression in human malignant glioma. Cancer Res. 2002;62(12):3347–3350.
   View this article via: PubMed Google Scholar
6. Harkins L, et al. Specific localisation of human cytomegalovirus nucleic acids and proteins in human colorectal cancer. Lancet. 2002;360(9345):1557–1563.
   View this article via: PubMed CrossRef Google Scholar
7. Samanta M, Harkins L, Klemm K, Britt WJ, Cobbs CS. High prevalence of human cytomegalovirus in prostatic intraepithelial neoplasia and prostatic carcinoma. J Urol. 2003;170(3):998–1002.
   View this article via: PubMed CrossRef Google Scholar
8. Michaelis M, Doerr HW, Cinatl J Jr. Oncomodulation by human cytomegalovirus: evidence becomes stronger. Med Microbiol Immunol. 2009;198(2):79–81.
   View this article via: PubMed CrossRef Google Scholar
9. Soroceanu L, Cobbs CS. Is HCMV a tumor promoter? Virus Res. 2011;157(2):193–203.
   View this article via: PubMed CrossRef Google Scholar
10. Powers C, DeFilippis V, Malouli D, Fruh K. Cytomegalovirus immune evasion. Curr Top Microbiol Immunol. 2008;325:333–359.
    View this article via: PubMed Google Scholar
11. Wang D, Dubois RN. Eicosanoids and cancer. Nat Rev Cancer. 2010;10(3):181–193.
    View this article via: PubMed CrossRef Google Scholar
12. Clin Cancer Res. 2005;11(20):7362–7368.
    View this article via: PubMed CrossRef Google Scholar
13. Soumaoro LT, Uetake H, Higuchi T, Takagi Y, Enomoto M, Sugihara K. Cyclooxygenase-2 expression: a significant prognostic indicator for patients with colorectal cancer. Clin Cancer Res. 2004;10(24):8465–8471.
    View this article via: PubMed Google Scholar
14. Baryawno N, et al. Tumor-growth-promoting cyclo­oxygenase-2 prostaglandin E2 pathway provides medulloblastoma therapeutic targets. Neuro Oncol. 2008;10(5):661–674.
    View this article via: PubMed CrossRef Google Scholar
15. Thun MJ, Henley SJ, Patrono C. Nonsteroidal anti-inflammatory drugs as anticancer agents: mechanistic, pharmacologic, and clinical issues. J Natl Cancer Inst. 2002;94(4):252–266.
    View this article via: PubMed Google Scholar
16. Elwood PC, Gallagher AM, Duthie GG, Mur LA, Morgan G. Aspirin, salicylates, and cancer. Lancet. 2009;373(9671):1301–1309.
    View this article via: PubMed CrossRef Google Scholar
17. Rothwell PM, Fowkes FG, Belch JF, Ogawa H, Warlow CP, Meade TW. Effect of daily aspirin on long-term risk of death due to cancer: analysis of individual patient data from randomised trials. Lancet. 2011;377(9759):31–41.
    View this article via: PubMed CrossRef Google Scholar
18. Wang L, Chen W, Xie X, He Y, Bai X. Celecoxib inhibits tumor growth and angiogenesis in an orthotopic implantation tumor model of human colon cancer. Exp Oncol. 2008;30(1):42–51.
    View this article via: PubMed Google Scholar
19. Cuzick J, et al. Aspirin and non-steroidal anti-inflammatory drugs for cancer prevention: an international consensus statement. Lancet Oncol. 2009;10(5):501–507.
    View this article via: PubMed Google Scholar
20. Bosetti C, Gallus S, La Vecchia C. Aspirin and cancer risk: a summary review to 2007. Recent Results Cancer Res. 2009;181:231–251.
    View this article via: PubMed Google Scholar
21. Maussang D, et al. The human cytomegalovirus-encoded chemokine receptor US28 promotes angiogenesis and tumor formation via cyclooxygenase-2. Cancer Res. 2009;69(7):2861–2869.
    View this article via: PubMed CrossRef Google Scholar
22. Speir E, Yu ZX, Ferrans VJ, Huang ES, Epstein SE. Aspirin attenuates cytomegalovirus infectivity and gene expression mediated by cyclooxygenase-2 in coronary artery smooth muscle cells. Circ Res. 1998;83(2):210–216.
    View this article via: PubMed Google Scholar
23. Zhu H, Cong JP, Yu D, Bresnahan WA, Shenk TE. Inhibition of cyclooxygenase 2 blocks human cytomegalovirus replication. Proc Natl Acad Sci U S A. 2002;99(6):3932–3937.
    View this article via: PubMed CrossRef Google Scholar
24. Bongers G, et al. The cytomegalovirus-encoded chemokine receptor US28 promotes intestinal neoplasia in transgenic mice. J Clin Invest. 2010;120(11):3969–3978.
    View this article via: JCI PubMed Google Scholar
25. Straat K, et al. Activation of telomerase by human cytomegalovirus. J Natl Cancer Inst. 2009;101(7):488–497.
    View this article via: PubMed Google Scholar
26. Baryawno N, Sveinbjornsson B, Kogner P, Johnsen JI. Medulloblastoma: a disease with disorganized developmental signaling cascades. Cell Cycle. 2010;9(13):2548–2554.
    View this article via: PubMed Google Scholar
27. Read TA, et al. Identification of CD15 as a marker for tumor-propagating cells in a mouse model of medulloblastoma. Cancer Cell. 2009;15(2):135–147.
    View this article via: PubMed CrossRef Google Scholar
28. Singh SK, et al. Identification of human brain tumour initiating cells. Nature. 2004;432(7015):396–401.
    View this article via: PubMed CrossRef Google Scholar
29. Odeberg J, Wolmer N, Falci S, Westgren M, Seiger A, Soderberg-Naucler C. Human cytomegalovirus inhibits neuronal differentiation and induces apoptosis in human neural precursor cells. J Virol. 2006;80(18):8929–8939.
    View this article via: PubMed Google Scholar
30. Odeberg J, et al. Late human cytomegalovirus (HCMV) proteins inhibit differentiation of human neural precursor cells into astrocytes. J Neurosci Res. 2007;85(3):583–593.
    View this article via: PubMed CrossRef Google Scholar
31. Sanai N, Alvarez-Buylla A, Berger MS. Neural stem cells and the origin of gliomas. N Engl J Med. 2005;353(8):811–822.
    View this article via: PubMed Google Scholar
32. Scotting PJ, Walker DA, Perilongo G. Childhood solid tumours: a developmental disorder. Nat Rev Cancer. 2005;5(6):481–488.
    View this article via: PubMed CrossRef Google Scholar
33. Qiu H, Straat K, Rahbar A, Wan M, Soderberg-Naucler C, Haeggstrom JZ. Human CMV infection induces 5-lipoxygenase expression and leukotriene B4 production in vascular smooth muscle cells. J Exp Med. 2008;205(1):19–24.
    View this article via: PubMed CrossRef Google Scholar
34. Bresnahan WA, Shenk T. A subset of viral transcripts packaged within human cytomegalovirus particles. Science. 2000;288(5475):2373–2376.
    View this article via: PubMed CrossRef Google Scholar
35. Zipeto D, Bodaghi B, Laurent L, Virelizier JL, Michelson S. Kinetics of transcription of human cytomegalovirus chemokine receptor US28 in different cell types. J Gen Virol. 1999;80(pt 3):543–547.
    View this article via: PubMed Google Scholar
36. Fortunato EA, Dell’Aquila ML, Spector DH. Specific chromosome 1 breaks induced by human cytomegalovirus. Proc Natl Acad Sci U S A. 2000;97(2):853–858.
    View this article via: PubMed CrossRef Google Scholar
37. Cobbs CS, Soroceanu L, Denham S, Zhang W, Kraus MH. Modulation of oncogenic phenotype in human glioma cells by cytomegalovirus IE1-mediated mitogenicity. Cancer Res. 2008;68(3):724–730.
    View this article via: PubMed CrossRef Google Scholar
38. Hume AJ, Finkel JS, Kamil JP, Coen DM, Culbertson MR, Kalejta RF. Phosphorylation of retinoblastoma protein by viral protein with cyclin-dependent kinase function. Science. 2008;320(5877):797–799.
    View this article via: PubMed CrossRef Google Scholar
39. Zhu H, Shen Y, Shenk T. Human cytomegalovirus IE1 and IE2 proteins block apoptosis. J Virol. 1995;69(12):7960–7970.
    View this article via: PubMed Google Scholar
40. Yu Y, Alwine JC. Human cytomegalovirus major immediate-early proteins and simian virus 40 large T antigen can inhibit apoptosis through activation of the phosphatidylinositide 3′-OH kinase pathway and the cellular kinase Akt. J Virol. 2002;76(8):3731–3738.
    View this article via: PubMed CrossRef Google Scholar
41. Poma EE, Kowalik TF, Zhu L, Sinclair JH, Huang ES. The human cytomegalovirus IE1-72 protein interacts with the cellular p107 protein and relieves p107-mediated transcriptional repression of an E2F-responsive promoter. J Virol. 1996;70(11):7867–7877.
    View this article via: PubMed Google Scholar
42. Castillo JP, Kowalik TF. Human cytomegalovirus immediate early proteins and cell growth control. Gene. 2002;290(1–2):19–34.
    View this article via: PubMed CrossRef Google Scholar
43. Awasthi S, Isler JA, Alwine JC. Analysis of splice variants of the immediate-early 1 region of human cytomegalovirus. J Virol. 2004;78(15):8191–8200.
    View this article via: PubMed CrossRef Google Scholar
44. Zhu H, Cong JP, Yu D, Bresnahan WA, Shenk TE. Inhibition of cyclooxygenase 2 blocks human cytomegalovirus replication. Proc Natl Acad Sci U S A. 2002;99(6):3932–3937.
    View this article via: PubMed CrossRef Google Scholar
45. Maussang D, et al. Human cytomegalovirus-encoded chemokine receptor US28 promotes tumorigenesis. Proc Natl Acad Sci U S A. 2006;103(35):13068–13073.
    View this article via: PubMed CrossRef Google Scholar
46. Wentworth BB, French L. Plaque assay of cytomegalovirus strains of human origin. Proc Soc Exp Biol Med. 1970;135(2):253–258.
    View this article via: PubMed Google Scholar
47. Kleihues P, et al. The WHO classification of tumors of the nervous system. J Neuropathol Exp Neurol. 2002;61(3):215–225.
    View this article via: PubMed Google Scholar