Article Text

Download PDFPDF

Cytomegalovirus and glioma: putting the cart before the horse
  1. Mahua Dey,
  2. Atique U Ahmed,
  3. Maciej S Lesniak
  1. The Brain Tumor Center, The University of Chicago, Chicago, Illinois, USA
  1. Correspondence to Maciej S Lesniak, The Brain Tumor Center, The University of Chicago Pritzker School of Medicine, 5841 South Maryland Ave, MC 3026, Chicago, IL 60637, USA; mlesniak{at}


In 1908, Oluf Bang and Vilhelm Ellerman laid the foundation for theory of oncoviruses by demonstrating that the avian erythroblastosis (a form of chicken leukaemia) could be transmitted by cell-free extracts. Since then, it has been shown very convincingly that viruses can directly cause several human cancers by various mechanisms. Epidemiological data imply that viruses are the second most important risk factor for cancer development in humans, exceeded only by tobacco consumption. Although the ability of certain viruses (hepatitis B and C, human papillomavirus, etc) to cause cancer has been time tested and proven scientifically, there are several other potential viral candidates whose role in oncogenesis is more controversial. One such controversial scenario involves the role of cytomegalovirus (CMV) in malignant gliomas, the most common form of primary brain tumour. CMV first attracted attention about a decade ago when CMV gene products were found in glioma tissue but not in normal brain. Since this initial observation, several different groups have shown an oncomodulatory effect of CMV; however, direct association between CMV infection and incidence of glioma is lacking. In this review, we will evaluate the evidence, both preclinical and clinical, regarding the possible role of CMV in gliomagenesis and maintenance. We will also critically evaluate the rationale for using antiviral drugs in the treatment of patients with glioma.

View Full Text

Statistics from


In 1964, Epstein et al identified Epstein–Barr virus (EBV), a herpes virus, as the first human cancer causing virus from Burkitt's lymphoma cells by electron microscopy.1 This started an era of intense research in the field of oncoviruses, leading to establishment of seven human viruses as the cause of 10–15% of all human cancers.2 Remarkable advances in the prevention of virus-associated cancers have been made through vaccination against human papillomavirus and hepatitis B virus.3 Malignant gliomas, comprising anaplastic astrocytoma and glioblastoma multiforme (GBM), are the most common primary brain tumours in adults. Current standard of care for newly diagnosed GBM involves gross total resection when possible, followed by adjuvant chemotherapy with temozolomide (TMZ) and radiation.4 Over the last decade, there has been much speculation about the role of cytomegalovirus (CMV) in gliomagenesis. Based on these speculations, a recent clinical trial, Valcyte Treatment of Glioblastoma Patients in Sweden (VIGAS), treated glioma patients with antiviral drug valganciclovir with disappointing outcome.5 Thus, glioma therapeutics represent a highly active area of basic science research and clinical trials.

In this review, we will outline the evidence, both preclinical and clinical, regarding CMV and its role in gliomagenesis. We will also discuss the scope of CMV as a causal agent of glioma and the rationale for/against treating glioma patients with antiviral drugs.

CMV biology

Infectious agents like viruses and bacteria are long known to cause various pathological cellular alterations such as DNA mutations, cell cycle modulation and dysregulation of DNA repair mechanisms, all leading to carcinogenesis. Also, recent literature convincingly demonstrates the role of chronic inflammation resulting from various infectious agents as a clear aetiology of various cancers.6 For example, Helicobacter pylori is associated with gastric cancer, hepatitis B and C viruses with liver cancer, and human papilloma viruses with cervical cancer. Thus, treating the offending infection is an effective way of interfering with carcinogenesis or even preventing cancer.7 Both preventing infection by vaccination (such as hepatitis B) and treating active infection (such as HIV) decreases the risk of associated cancer (hepatocellular carcinoma and Kaposi sarcoma).8 ,9 CMV, a double-stranded DNA virus from the Herpesviridae family, genome is approximately 230 kb in size and carries approximately 200 genes encoding proteins, some non-coding RNAs and approximately 14 microRNAs.10 CMV is extremely prevalent worldwide with an overall 50% seroprevalence among US adults but as high as 90% in Mexican-Americans and 96% in southern Brazilian population.10 At an early age, it is associated with lower socioeconomic classes;11 however, these criteria dissolve over the age of 65. Like any herpes virus, CMV becomes clinically inactive after the usually asymptomatic primary infection phase is brought under control by the host immune system. The virus persists at the quiescent state in the infected host and has the potential to become active in the late stage. In latency, infected cells are not producing any infectious virus but retain the complete genome and have the potential to start producing virus at a later time. The immature cells of the myeloid lineage within the bone marrow are the major reservoir for CMV latency; however, other sites of latency cannot be completely ruled out.12

Glioma aetiology

Like most other cancers, GBM is a sequential multistep evolution of cumulative genetic alterations from intrinsic factors such as rare hereditary syndromes and environmental factors such as therapeutic irradiation.13 ,14 The risk of GBM increases with older age, male gender, higher socioeconomic class and non-Latino white race.15 Besides the hereditary syndromes and therapeutic irradiation, epidemiological studies have failed to identify any clear aetiology of brain cancer. Over the years, many different aetiologies have been implicated as the causal agent of glioma, such as exposure to cell phones, viruses, head injury and foods containing N-nitroso compounds. Even though none of these stood the test of rigorous scientific evaluation,16 ,17 some of these, especially cell phone usage and viruses, have captured public attention.

CMV is not the first virus that has been implicated as a causal agent for GBM. Between 1955 and 1963, Simian virus 40 (SV40) was found to have contaminated the polio vaccines and adeno vaccines, resulting in inadvertent injection of millions of people with this virus. SV40 viral sequences have been detected in 35% of brain tumours and thus have been implicated as a causative agent in the development of neurological malignancies.1820 However, a clear relationship between virus-contaminated vaccinations and an increased risk for brain cancer was not supported by any epidemiological data.21 ,22 In the case of CMV, since Cobbs et al first reported the presence of CMV gene product in 100% of the tested GBM samples compared with none in the normal tissue,23 the relationship between CMV and glioma has been highly controversial. On the one hand, several groups have reported similar findings, but on the other hand, many groups failed to identify any CMV gene products in the glioma samples (table 1). A recent, large-scale transcriptome sequencing study reading 22.8 billion sequences from 167 tumours found only one sequence corresponding to CMV RNA.24 To date, different studies have reported the presence of seven different CMV proteins and two different CMV genes within human glioma specimens.25 However no group has been able to isolate the actual viral particle from the glioma tissue. This brings us to the critical question that begs our attention: Is CMV a pathogen or bystander when it comes to gliomagenesis? To consolidate these confounding results, we need to first define how to establish a causal relationship between a causative microbe, CMV and the disease, glioma.

Table 1

Studies analysing CMV gene products in human tissue

The earliest and the simplest way of proving weather a microorganism is the cause of a disease was described by a German physician, Robert Koch, in the late 1800s. Koch formulated guidelines known as Koch's postulates in an attempt to create guidelines for proving disease causation by microbes, formalising how we think about the connection between cause and effect in medicine.26 His postulates were mostly derived from his work on bacterial infectious diseases such as anthrax and tuberculosis. The fundamental limitation of his work was that it could not be applied faithfully to obligatory intracellular parasites such as viruses that propagate by usurping cellular machinery and thus cannot be propagated in pure culture.26 Last century saw a tremendous leap in our medical knowledge and a revolution in biotechnology affecting microbiology, molecular biology and cell biology. These technical advances brought forward various improvisations of Koch's postulates for establishing causality based on new technology (sequence-based detection of microorganism), epidemiological data (Hill's criteria) and immunological data (Elements of Immunological Proof of Causation), etc.26 Even though the debate continues in the field of microbiology regarding the best set of guidelines to establish microbial pathogenesis and disease causation, Koch's criteria still provide a valid standard for judging disease causation as long as we keep in mind the associated limitations. So the question then becomes: Does CMV meets Koch's criteria to qualify as the causal agent of glioma?

Koch's postulates

The microorganism must be found in abundance in all organisms suffering from the disease, but should not be found in healthy organisms

In the USA, National Health and Nutrition Examination Survey reported overall age-adjusted CMV seroprevalence (CMV-specific immunoglobulin G antibody) of 50.4%.27 CMV seroprevalence was higher among non-Hispanic black and Mexican-American children compared with non-Hispanic white children and increased more quickly in subsequent age groups. CMV seropositivity was independently associated with older age, female sex, foreign birthplace, low household income, high household crowding and low household education.27 In contrast, the prevalence rate of malignant glioma in USA is only 29.5 cases per 100 000.28 So far there has been no epidemiological study correlating timing of CMV infection and subsequent risk of glioma development. Even the studies that did identify CMV gene products in glioma samples reported very low viral DNA in the samples as compared with total DNA.29 ,30 In patients with glioma containing CMV DNA, only about 50% had detectable viral DNA in the peripheral blood.31 Since bone marrow resident monocytes are known reservoirs of latent CMV, among the patients with CMV gene product positive glioma, with no evidence of CMV in the normal brain, the presence of CMV in bone marrow monocytes cannot be ruled out. Just by the virtue of ubiquitous nature of CMV infection and comparatively very low incidence of glioma, CMV fails to satisfy this criterion.25 However, we should keep in mind that the now well-established Burkitt's lymphoma caused by EBV also failed to satisfy Koch's criteria as a cancer-causing agent. This can be explained by recognition that chronic viral infection may function together with multifactorial non-viral risks to contribute to cancer.

The microorganism must be isolated from a diseased organism and grown in pure culture.

As we discussed previously in the limitations of Koch's postulates, intracellular pathogens such as CMV are difficult to grow in pure culture. However, the infectious CMV virion has yet to be isolated from the GBM samples. Inability to isolate CMV virus thus far excludes the possibility of an active CMV infection in the GBM tissue. Another possibility is whether CMV exists within the glioma as a latent disease instead of an active infection? During its latency phase, CMV expresses only a small number of genes including CMV IL-10 and US28, which are also present in glioma.12 ,32 However, the presence of early gene products such as IE1 is a strong indication that the CMV infection in glioma is present not just in a latent phase. Even though some of the studies specified the presence of CMV in primary glioma,33 ,34 the evidence is not clear and consistently displayed. One possible scenario may be that antiglioma chemotherapy in the form of TMZ can reactivate the latent CMV infection in glioma patients. Although TMZ is generally well tolerated, it can cause lymphopenia and may lead to opportunistic infections.35 The frequency of latent CMV reactivation post or during TMZ treatment in patients with GBM is not well documented; however, there have been anecdotal cases of CMV reactivation reported in the literature.35 ,36 Once the latent CMV gets reactivated in patients treated with TMZ, the glioma tissue can be a prime target for CMV because the cellular entry receptor for CMV is the platelet-derived growth factor receptor (PDGFR) and specific PDGFR haplotypes are associated with greater incidents of glioblastoma.37 ,38 Though promising, such hypothesis will require intense research to prove.

The cultured microorganism should cause disease when introduced into a healthy organism

Active CMV infection has been shown to cause several illnesses such as retinitis, mononucleosis and encephalitis; however, glioma is not one of them.10 Even in animal models, infection with CMV by itself failed to induce gliomagenesis.39 In the mouse model of congenital CMV infection, studies have shown that early infection with CMV causes significant loss of neural stem cells (NSCs), decreased proliferation of neuronal precursor cells and marked loss of young neurones.40 ,41 These studies suggest that CMV infection may be associated with several neurological pathologies such as mental retardation and microcephaly; however, none of these studies report increased risk of glioma. When undifferentiated neural precursor cells infected with CMV differentiate into glial cells, they retain their susceptibility to CMV concomitantly but CMV susceptibility is actively repressed following their differentiation into neurons.42 Hence unlike EBV that can immortalise primary B cells,43 CMV neither causes glioma by itself nor does the infection of NSCs with CMV render these cell immortal.

The microorganism must be reisolated from the inoculated, diseased experimental host and identified as being identical to the original specific causative agent

Failure of CMV to satisfy any of the above three criteria automatically rules out this possibility. Thus, even accounting for all the limitations of Koch's postulates, CMV clearly does not qualify as a causal agent of glioma. Hence treating CMV infection with antiviral drugs in this setting cannot be justified on the evidence present so far.

Although CMV does not clear the bar for being a causal agent of glioma, several studies so far have clearly demonstrated the presence of CMV gene products in glioma samples. It is not clear whether these gene products are a result of active, latent or past infection or just contamination from glioma infiltrating immune cells; however, they do possess oncomodulatory activity of interacting with downstream signalling molecules. The term ‘oncomodulation’ was coined to infer that the virus may modulate the malignant properties of the tumour cells through mechanisms that affect the cell cycle, survival, invasive potential, chromosomal stability, immunodetection and angiogenic properties.44 The next important question that we will focus on in the following section is: What are the molecular and clinical consequences for the expression of the CMV gene products within the glioma tissues?

Molecular mechanisms of CMV-mediated oncomodulation

The possible association between CMV oncogenic events has been speculated for decades. In 1970, the first report published demonstrated the ability of CMV to transform normal human embryonal cells in vitro.45 Even though CMV-infected cells were able to form tumour in the immunocompromised animals, the expression of CMV-specific antigens in the transformed cells as well as in the tumour–derived cell lines diminished over time without altering their oncogenic potential.46 Similar observations were also reported with regards to the ability of CMV to transform rodent cells and the viral DNA was mostly absent in the transformed cells.4749 These observations led to the hypothesis that CMV may initiate oncogenesis by ‘hit-and-run’ mechanisms.48 ,49 However, such theory is invalid in the case of glioma because majority of the glioma specimens are positive for both CMV genes and their products. Moreover, up to now, there is no report conclusively demonstrating the transformative ability of CMV after infecting any normal cells from the central nervous system, thus the mechanism by which CMV might contribute to gliomagenesis remains unclear. The majority of the published studies so far point towards the notion that human glioma tissue may be susceptible to CMV infection and postinfection CMV may enhance the oncogenic properties of human glioma. This concept, also known as oncomodulation, is a process where CMV may have the propensity to infect glioma tissue and as a consequence may contribute to gliomagenesis without being directly involved in the initiation process of glioma. The candidate CMV genes products involved in the oncomodulatory activity have been extensively studied and are summarised in table 2. In the following section, we will discuss the three most commonly detected CMV gene products within human glioma tissue and elaborate their possible oncomodulatory activity during gliomagenesis.

Table 2

Role of CMV in gliomagenesis

Immediate early gene 1 (IE1)

During the early phase of CMV infection, the major immediate-early gene IE1 is expressed as a 491 aa nuclear protein from UL 123 open reading frame (ORF). IE1 is the most commonly detected protein in GBM. Several studies have reported that 93–100% of GBM samples are positive for IE1 by immunohistochemical (IHC) analysis, while much lower percentage are positive in the low-grade gliomas.23 ,34 ,5052 These findings identify IE1 as one of the viable CMV gene products that may contribute to the oncomodulatory activity in human glioma.

Like most of the small DNA viruses, CMV infects non-cycling cells in vivo, and such conditions are not likely to be conducive to viral DNA replication. To bypass this situation, IE1 expression stimulates the S phase entry of the infected cells via interaction with the pocket protein p107 and/or phosphorylation of critical host cell cycle regulators such as E2F through its reported kinase activity and interaction with p53 and Rb tumour suppressor proteins.5355 However, the biological outcome of such interaction is cell type dependent: in some cell types, this interaction results in growth arrest and in others it alters the cell cycle distribution towards S and G2/M phase. Stable expression of IE1 in human GBM also induces cell line-dependent responses—promoting proliferation in U87 and U118 cells and unchanged or decreased proliferation in U251 and LN229.56 ,57 Thus, these reports point towards the notion that the effects of CMV on glioma cell proliferation and cell cycle may depend on the context of intracellular microenvironment. Taking into consideration that all human gliomas are extremely heterogeneous in nature, it remains obscure how the cell cycle regulatory capacity of CMV IE1 can influence the oncogenic capacity of the glioma cells, where the cell cycle phases are inherently deregulated. Most importantly, these reports also demonstrated that IE1 expression in the immortalised fibroblast and epithelial cells did not alter their proliferation rate, DNA synthesis or their cell cycle distribution. Since other established oncogenes have been shown to transform these immortalised lines into tumorigenic lesions,58 it is reasonable to conclude that IE1 may not be associated with the direct oncogenic initiation during gliomagenesis.

However, IE-1 gene can contribute to maintenance and aggressive behaviour of GBM. Resistance to apoptosis is a common feature of glioma cells and represents a possible mechanism of chemoresistance. The antiapoptotic effects of the ectopic IE1 expression have been well documented in the literature (reviewed in59). IE1-mediated AKT/PI3K survival signalling is central to such regulation.56 The direct modulation of p53-related proteins by IE1 has been also reported.56 But this response was also cell line dependent, thus making it difficult to understand how such interaction might regulate the apoptotic pathway in glioma cells. In the future, it is imperative to investigate the functional aspects of apoptotic pathways by investigating how the IE1 expression alters the efficacy of antiglioma therapeutics, which are known to activate cellular apoptotic pathways.


CMV pp65 is the most abundant tegument protein that is responsible for modulating/evading the host cell immune response during active infection.60 It has been shown that pp65-mediated phosphorylation of CMV immediate-early proteins blocks their presentation to the major histocompatibility complex (MHC) class I molecules.61 Furthermore, several studies have presented evidence that pp65 is involved in modulating a decrease in MHC molecules62 which has been shown to be downregulated in glioma specimens as a potential immune escape mechanism.63 ,64 IHC analysis, deep sequencing and immunoblot analysis showed presence of pp65 in human glioma samples.30 ,34 Recently, Heimberger and colleagues reported propensity of CMV to infect glioma initiating cancer stem cells (GSC) and microglia populations.25 This resulted in the induction of immunosuppressive phenotype in human monocytes with downregulation of the MHC complex and costimulatory molecules. However, the CMV gene(s) that is responsible for immunomodulation within glioma is yet to be determined. Taking into consideration that pp65 is critical for modulating the host immune responses during CMV infection and the well-documented ability of malignant gliomas to evade immune detection, it is logical to postulate a possible role of pp65 in establishing the immunosuppressive microenvironment within glioma. Whether or not such modulation has any clinical significance with respect to disease progression or development of an effective antiglioma immunotherapy will still require vigorous collaborative investigation among the fields of tumour immunology, virology and molecular oncology.


US28 is a chemokine receptor encoded by CMV that shares the highest sequence homology with human chemokines receptors and binds to a broad spectrum of chemokines, including CCL2/MCP-1, CCL3, CCL4, CCL5/RANTES and CX3CL1.65 Consequently, US28 activates many different signalling pathways and constitutively activates NF-κB, the cAMP response element-binding protein (CREB) and nuclear factor of activated T cells (NFAT).66 In GBMs, about 60% of the tested patient samples were positive for US28 by IHC analysis.67

US28 overexpression has been shown to be enough to induce activation of several oncogenic pathways including signal transducer and activator of transcription 3 (STAT3) and e-NOS, secretion of vascular endothelial growth factor (VEGF) and promotion of invasion and angiogenesis in the neural progenitor cells and the established glioma cell lines.67 It has been convincingly shown that STAT3 plays an important oncogenic role in glioma pathogenesis.68 ,69 Overexpression of US28 in the immortalised NIH 3T3 fibroblast-induced transforming phenotype and targeted expression of US28 in the intestinal epithelial cell (IECs) of the VS28 transgenic mice showed hyperplastic intestinal epithelium resulting in mice eventually developing adenomas and adenocarcinomas by week 40.70

In the Mut3 transgenic model, murine CMV (MCMV) infection significantly shortened the survival of mice with glioma as compared with uninfected controlled mice or mice infected with a different neurotropic virus.39 MCMV infection also increased the activation of STAT3 in the neural progenitor cells isolated from the subventricular zone of the Mut3. Patient-derived glioma cells infected with CMV demonstrated similar hyperactivation of the STAT3 as well as increased neurosphere forming ability; moreover, a STAT3 inhibitor reversed the transforming phenotype post CMV infection in vitro and in vivo. These reports collectively support the oncogenic potential of the US28 in gliomagenesis. However, so far there has not been any study directly linking the presence of US28 gene to active CMV infection.

Taken together, none of the published results are indicative of CMV infection as one of the initiating events for gliomagenesis. To prove this, one must demonstrate that CMV infection alone or any molecular component(s) of the CMV are able to successfully transform immortalised normal cells from the target organ or induce tumour formation in the transgenic model as shown in the VS28 model.70 On the other hand, the oncomodulatory effects of the CMV gene products are quite convening though the clinical significance of such modulations remains obscure. In order to understand the clinical relevance of experimental definition of oncomodulation, we must move away from using the homogeneous glioma cell line system and use the transgenic animal model with heterogeneous spontaneous glioma. We also need to expand our investigative effort to study CMV-induced oncomodulation not only in infected tumour cells but also in tumour stroma in order to unravel the complete picture of CMV's role in gliomagenesis.

Clinical evidence of CMV in glioma

Atopic diseases such as asthma, eczema and hay fever have been shown to have inverse association with glioma, implicating a strong role of immune system in the context of glioma.71 Cumulatively all the preclinical studies point towards the possibility that CMV may exist as a persistent infection capable of immune evasion, in which certain viral genes capable of altering pro-inflammatory and anti-inflammatory cytokine production levels in the host are expressed without contributing to a significant amount of host cell lysis. Thus, the impact of CMV on glioma risk and prognosis could potentially be more related to viral manipulation of immunological factors and the host's immune response towards the virus, rather than virus-related host cell damage or a direct carcinogenic effect of a viral gene product. Clinically this immune modulating aspect of HCVM raises some very interesting questions and opens up a promising therapeutic window.

Does CMV seropositivity influence glioma risk?

Few epidemiological studies have tried to address this question over time without any conclusive results. Two separate studies over 5 years by the same US group found no difference in anti-CMV IgG between glioma cohort and control cohort.72 ,73 To take into account the possibility that the relationship between CMV and glioma could be influenced by treatments such as steroids and chemotherapy, Sjostrom et al analysed prediagnostic samples from three large Scandinavian cohorts and found no association between glioma risk and prediagnostic immunoglobulin levels for CMV.74 Some studies found confounding relationships between CMV seroprevalence and glioma based on gender, age and various ethnicities. For example, CMV seroprevalence is significantly lower in Caucasians than in African-Americans or Hispanics, while GBM incidence is higher; CMV seroprevalence rates are significantly higher in women than men, although GBM is more common in men.75 ,76 Contrary to these studies, Amirian et al assessed the potential associations between anti-CMV antibodies and glioma risk and found that among anti-CMV IgG-positive participants, increasing anti-CMV IgG levels were associated with decreasing glioma risk and those with the lowest level of anti-CMV IgG had the highest glioma risk, controlling for age, sex and race/ethnicity. Antibody levels were not associated with survival among glioma cases.77 Even in the VIGAS study only 29% of the treated patients were seropositive for CMV IgG.5 These handful of epidemiological studies fail to conclusively demonstrate any correlation between CMV seropositivity and glioma risk. This follows the common logical explanation that a ubiquitous agent is unlikely to play a key role in the initiation of a disease with very low prevalence. The more interesting question here is whether gliomagenesis fosters a local immunosuppressive environment in the brain creating an inviting environment for local CMV reactivation from its natural reservoir, aka haematopoietic cells. Well-designed long-term studies are needed to answer this question satisfactorily.

Does the presence of CMV gene product in glioma also correlate with anti-CMV serum antibody?

Starting from the original study by Cobbs et al in 2002, almost every study that reported the presence of CMV gene product (listed in table 1) in glioma is a retrospective analysis of patient tissue samples embedded in paraffin block. Apart from IHC, most of the other methods of CMV gene detection (PCR, qPCR, immunohistochemical staining, DNA sequencing, etc) fail to take into account the heterogeneous nature of the tumour composition involving immune cells, tumour-derived vascular endothelial cells along with the glioma cells. There is some evidence that suggests that undifferentiated glial cells are less permissive to CMV infection compared with fully differentiated astrocytes.78 One explanation for detecting CMV gene by PCR in glioma could be that GBMs with profoundly leaky blood–brain barrier have high infiltration of immune cells (some of them possibly containing latent CMV virus) compared with normal brain and thus PCR of the tumour sample will have some CMV gene product compared with the normal brain. This is supported since most studies reported low ratio of viral DNA compared with cellular DNA in the glioma samples.30 ,79 Moreover, none of the reported studies confirming the presence of CMV gene product looked at the seropositivity of the patients for anti-CMV immunoglobulin. One study found only half of the patients with glioma in their patient pool to be seropositive for CMV.80 Even though CMV is a ubiquitous virus, as discussed previously, its prevalence varies within various ethnic backgrounds. Thus, it is very important to establish a relationship between presence of CMV gene product in glioma samples with seropositivity for anti-CMV immunoglobulin in future studies.

Can the CMV gene products be effectively targeted as glioma-specific antigen?

One of the most fascinating observations is that CMV antigens such as IE-1 and pp65 are present in GBM cells but not in normal brain tissue. The most important implication of this is that the virally encoded proteins can be targeted for immune-based therapies. While conducting a phase 1 trial of autologous dendritic-cell vaccination pulsed with tumour lysate as adjunctive therapy in glioma, Prins et al reported one case (out of 14 recruited at that point) of robust CD8+ T-cell response to the pp65 immunodominant epitope of human CMV immediately after one injection of vaccine. The patient's tumour also stained positive for CMV protein pp65.81 Since it is much easier to elicit an immune response against viral antigens compared with ‘self’ tumour antigens, this observation opened up the possibility of CMV in glial cells to serve as an immunotherapeutic target in GBM. Subsequently, Ghazi et al successfully reactivated and expanded CMV-specific T cells from patients with GBM using antigen-presenting cells transduced with an adenoviral vector encoding pp65 and IE1. CMV-specific T-cell lines were able to recognise pp65 and IE1 targets and killed CMV-infected autologous GBM cells.82 Taking this step forward, Crough et al adoptively transferred in vitro-expanded CMV-specific T cells in combination with TMZ therapy into a patient with recurrent GBM and showed a long-term disease-free survival. Interestingly, several phase I/II clinical trials assessing the safety of administrating CMV-specific cytotoxic T cells in patients with GBM have been terminated due to poor accrual (NCT01205334, NCT00990496). Another clinical trial from Duke (ERaDICATe) designed to evaluate whether vaccinating adult patients with newly diagnosed GBMs using CMV-DCs during recovery from therapeutic TMZ-induced lymphopenia with autologous lymphocyte transfer in patients that are seropositive for CMV enhances the T-cell response is not currently recruiting. Although the concept of targeting CMV antigen expressing GBM with CMV-specific T cell is very intriguing, it needs to be rigorously tested clinically and preclinically before it is ready for primetime, especially with the possibility of inducing immune response against CMV reservoir haematopoietic cells.

Finally, a recent letter published in the New England Journal of Medicine pertaining to the use of valganciclovir in the setting of gliomas is gathering increasing public attention. However, the study itself is problematic. First, the authors claim they examined 250 patients with CMV proteins, but only 75 were evaluated for survival. Of these, patients with ‘low-grade’ CMV infection did better than those with ‘high-grade’ infection. It would be helpful to define and quantify the difference between low versus high level of infection as the authors provide no information or data regarding these definitions. Moreover, while they admit that in their VIGAS study valganciclovir showed no benefit with regards to tumour volumes and median overall survival, an ‘exploratory analyses’ of 22 patients receiving 6 months of the drug versus contemporary controls showed an increased survival. Patients could stay on valganciclovir for compassionate use even though there was no difference in defined endpoints for as long as they lived.

Their retrospective analysis of those 22 patients and 28 additional patients treated off trial (for compassionate use again) was then used to claim benefit. What they found is that patients on standard therapy who continued to take valganciclovir despite it not working as far as defined clinical trial endpoints ‘lived longer’. Essentially, if you lived, you continued the drug for ‘compassionate reasons’, and that is the basis for claiming it works. It is therefore interesting that negative results of a clinical trial are now used to suggest efficacy. It is even more perplexing that an agent that is used to inhibit viral replication is posited to be used in patients with glioma without any evidence of active viral CMV infection in those individuals. Caution should be used in interpreting studies that are not designed to answer the question at hand, and clinical judgement should not be replaced by a publication simply because it appeared in the New England Journal of Medicine.


Taking all the preclinical and clinical studies together, the relationship between CMV and glioma can at best be described as an observational association. Although there is a lot of preclinical evidence of CMV playing a role in gliomagenesis, clinical studies are lagging far behind and fail to substantiate the claims put forward by the preclinical studies. A consensus statement on the role of CMV on glioma also concluded that current literature supports an oncomodulatory role for CMV in malignant gliomas, but future studies are needed to focus on determining the role of CMV as a glioma-initiating event.25 One thing that can be conclusively deduced is that CMV infection does not cause glioma. Hence, there is absolutely no rationale for treating patients with antiviral medications when there is no objective evidence of active viral replication. However, CMV does appear to play a role as an oncomodulator, and elucidating the mechanisms, which contribute to this process, is the first step in the right direction.


This work was supported by NIH R01CA122930, RR01CA138587, R01NS077388, U01NS077388 (MSL), R00C160775, R25NS065744 and ACS Resident Research Fellowship Grant.


View Abstract


  • Contributors MD wrote and organised the manuscript. AUA wrote and reviewed the manuscript. MSL participated in the writing, review and correction of the final version of the manuscript.

  • Competing interests None.

  • Provenance and peer review Commissioned; externally peer reviewed.

  • Data sharing statement This is not an original research article. It is a review.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles