Investigation of Adenovirus Occurrence in Pediatric Tumor Entities

Abstract

Adenoviruses (AdVs) contain genes coding for proteins with transforming potential, and certain AdV serotypes have been shown to induce tumors in rodents. However, data on the possible oncogenicity of AdVs in humans are scarce. We have therefore employed a real-time quantitative PCR (RQ-PCR) assay permitting highly sensitive detection of all 51 currently known human AdV serotypes to screen more than 500 tumor specimens derived from 17 different childhood cancer entities including leukemias, lymphomas, and solid tumors. Most tumor entities analyzed showed no evidence for the presence of AdV sequences, but AdV DNA was detected by RQ-PCR in different brain tumors including 25/30 glioblastomas, 22/30 oligodendrogliomas, and 20/30 ependymomas. Nonmalignant counterparts of AdV-positive brain tumors, including specimens of ependymal cells, plexus choroideus, and periventricular white matter, were screened for control purposes and revealed the presence of AdV DNA in most specimens tested. Identification of the AdV types present in positive malignant and nonmalignant brain tissue specimens revealed predominantly representatives of species B and D and, less commonly, C. To exclude contamination as a possible cause of false-positive results, specimens with AdV sequences detectable by PCR were subsequently analyzed by in situ hybridization, which confirmed the PCR findings in all instances. The central nervous system appears to represent a common site of AdV infection with virus persistence, thus providing the first evidence for the possible contribution of AdVs to the multistep process of tumor pathogenesis in brain tissue.

The role of infectious agents in the pathogenesis of human cancer has been an important focus of research over the past decades. In particular, virus infections appear to be associated with a number of malignant disorders (4, 12, 41, 46) in which tumorigenicity may be attributable to virus-induced immune suppression or transforming activity mediated by viral oncogenes (19, 25). Well-known examples include the human T-cell lymphotropic virus type 1 in acute leukemia (41), the human papillomavirus in cervical carcinoma (25), the hepatitis B virus in hepatocellular carcinoma (46), and the Epstein-Barr virus in Burkitt's lymphoma (12, 25). The first evidence of the induction of malignant tumors by a human virus dates back to 1962, when adenovirus (AdV) type 12 was shown to induce multiple tumors in newborn hamsters including sarcomas, neuroectodermal tumors, adenocarcinomas, retinoblastomas, and medulloblastomas (26, 47). Similar observations were later made with different AdV serotypes, particularly of the species A, B, and D. For example, AdV serotype 9 has been associated with mammary gland tumors in rats (29). Moreover, serotype 5 was identified in human small-cell lung carcinoma (33). However, several studies failed to provide conclusive evidence for the presence of human AdV sequences in adult human cancer (13, 21, 37). Despite epidemiological evidence for virus involvement in childhood cancer (36), no studies focusing on AdVs in pediatric neoplasia have been performed to date.

Human AdVs constitute a large family, currently including 51 different serotypes divided into six species (A to F) (2, 45). Primary, clinically often inapparent infection with AdVs usually occurs in early childhood. A latent form of AdV infection persists within tonsillar lymphocytes in nearly 80% of children investigated (17) or in peripheral blood lymphocytes (27). AdVs enter susceptible hosts by either the mouth, the nasopharynx, or the ocular conjunctiva (15). The uptake of virus particles by host cells occurs via the CAR (coxsackievirus and AdV receptor) or CD46 receptor and integrins (3, 16, 50). Subsequently, inside the cytoplasm, the virus capsid is removed and the viral DNA is transported into the nucleus (20). Early viral proteins (E1A and E1B) act as transcription factors for viral and cellular genes and interact with important tumor suppressors, including the retinoblastoma and p53 proteins (10, 35). This indicates a rather complex pattern of interaction between early AdV proteins and key players of cell cycle control. Moreover, AdVs carry the E3 and E4 genes, which code for proteins capable of blocking the onset of apoptosis (18, 38, 42). The E1A and E1B proteins of the AdV serotypes 5 and 12 were demonstrated to transform rodent cells (44, 49), and the E4orf1 protein of AdV serotype 9 was shown to possess transforming capacity (29). The transforming properties of human AdVs are well documented in animal models but could be reproduced only in selected types of human cells (10). Data indicating a possible role of AdVs in human oncogenesis are restricted to the investigation of serotype 5, and little is known about the oncogenic potential of other members of the AdV family. Moreover, there seem to be major functional differences between identical genes in different AdV types, suggesting, for example, that the transforming capacity demonstrated for the E1A gene of serotype 12 may not necessarily apply to other AdV serotypes (30). The CAR mediates the entry of most, but not all, AdV serotypes into human cells (28, 40). The CD46 molecule has recently been identified as the receptor for infection by AdV species B (16), and the CAR and the CD46 receptor are differentially expressed in various tissues. There appears to be a tissue-dependent mechanism for the transformation of cells (24), thus rendering predictions of the oncogenicity of individual AdV serotypes in humans difficult.

We have recently determined the complete nucleotide sequence of the hexon gene in all AdV serotypes (7). This gene codes for the most abundant capsid protein, which carries most recognition epitopes. The comprehensive DNA sequence data enabled us to establish a highly sensitive real-time quantitative PCR (RQ-PCR) assay for the detection of all currently known human AdV serotypes (9), which was an important prerequisite for the current study. We have employed the new pan-AdV detection assay to screen a large spectrum of pediatric malignancies including leukemias, lymphomas, and solid tumors for the presence of AdV sequences to assess a possible implication of human AdVs in childhood neoplasia.

MATERIALS AND METHODS

Patients and tumor specimens.

The tumors studied were derived from archival material collected at the time of diagnosis in pediatric patients from St. Anna Children's Hospital, Vienna, Austria. Written informed consent had been obtained from each patient and/or the parents, and the study was carried out in agreement with the Helsinki Declaration. The specimens studied were frozen diagnostic samples including peripheral blood or bone marrow from patients with B-cell precursor acute lymphocytic leukemia (B-ALL), T-cell acute lymphocytic leukemia (T-ALL), acute myeloid leukemia, and chronic myeloid leukemia; affected lymph nodes from patients with Hodgkin's disease and B-cell and T-cell non-Hodgkin's lymphoma (B-NHL and T-NHL, respectively); and diagnostic tumor tissue specimens from patients with Wilms' tumor, Ewing's sarcoma, and neuroblastoma. An additional source of material was paraffin-embedded, diagnostic tumor tissue of embryonal and alveolar rhabdomyosarcoma, osteosarcoma, and germ cell tumors. Histological sections were performed using fresh disposable blades for each specimen to prevent cross-contamination. Moreover, targeted core needle biopsy samples were excised from paraffin-embedded brain tumor tissue including ependymoma, glioblastoma, oligodendroglioma, astrocytoma, and medulloblastoma. For brain tumor entities for which insufficient or no specimens from pediatric patients were available, samples from adult patients were analyzed (ependymoma, n = 5; glioblastoma, n = 30; oligodendroglioma, n = 30; astrocytoma, n = 7; and medulloblastoma, n = 0). Additionally, targeted core needle biopsy samples from paraffin-embedded specimens of choroid plexus papilloma and nonmalignant brain tissue (derived from surgery in patients with focal epilepsy), including samples from the ependymal cell layer, plexus choroideus, and periventricular tissue, were investigated.

DNA extraction.

DNA isolation from peripheral blood and bone marrow specimens from leukemia patients and from lymph node samples of lymphoma patients was performed with the QIAmp Blood DNA Mini Kit (QIAGEN, Germany); DNA extraction from paraffin-embedded tissue was carried out with the QIAmp Tissue DNA Mini Kit (QIAGEN, Germany) according to the manufacturer's recommendations. DNA extraction from fresh solid tumor specimens was performed by conventional tissue lysis at 56°C overnight with proteinase K and subsequent ethanol precipitation. The DNA content was quantified by photometric analysis using the Spectrophotometer U-2000 (Hitachi, Japan).

Real-time PCR analysis.

An RQ-PCR assay detecting all 51 human AdV serotypes in two PCRs was used for screening of the pediatric tumor specimens studied (9). The PCR tests were carried out using the 7700 ABI Genetic Analyzer (Applied Biosystems, CA). About 50 to 100 ng DNA was used as template in individual PCRs. To control for adequate quantity and quality of the DNA, a single-copy human housekeeping gene (the beta-2-microglobulin gene) was tested in parallel by RQ-PCR (48), and samples were regarded as adequate for virus analysis if a minimum of 5 × 10³ copies of the control gene were present. Satisfactory results were obtained in specimens of all tumor entities, medulloblastoma being the only exception. In this tumor entity, DNA quality adequate for AdV screening was available only in 12 of 30 specimens investigated.

For species identification in samples that tested positive by the pan-AdV assay, RQ-PCR assays detecting each of the species A to F in individual reactions were employed, as described elsewhere (34). Analysis of the adenoviral oncogenes E1A, E1B, and E4orf1 has been performed by RQ-PCR using the primers and probes listed in Table 1 under amplification conditions described earlier (48). Several negative controls were included in each assay.

TABLE 1.

Primers and probes for RQ-PCR detection of the adenoviral oncogenes E1A, E1B, and E4orf1 of serotypes from species B^a

Primer	Oligonucleotide sequence (5′-3′)	Concn (μM)
Ad B E1A forward	CAKACYGCAGYGAATGAGGGAG	0.4
Ad B E1A probe	CCTGGACATGGCTGTAAGTCTTGTGAATTTCAC	0.2
Ad B E1A reverse	CAYAAYAGTTCYTTYATTCCAGTGTTWTTC	0.4
Ad B E1B forward	GAGTTGGCTTTAAGTTTAATGAG	0.4
Ad B E1B probe	TTTGGTGGCATGAGGTTCAGA	0.2
Ad B E1B reverse	AGTGAATATTTCTCCTGC	0.4
Ad B E4orf1 forward	CTGATGAGGCTTTGTATGTGTATTTAGA	0.4
Ad B E4orf1 probe	ACGTTGCCTGAACAGCAGCAGAGAAATAATT	0.2
Ad B E4orf1 reverse	CAAAGTAAAAGGCACAGGAGAATAAAA	0.4

^aAll probe sequences indicated are TaqMan probes labeled with 6-carboxyfluorescein on the 5′ end and with 6-carboxytetramethylrhodamine on the 3′ end.

Sequencing and serotype identification by PCR fragment analysis.

PCR products for subsequent serotyping by fragment length analysis were generated using primers specific for the AdV species A, B, C, E, and F, as described previously. Serotype identification within species D was performed by nucleotide sequence analysis of a short PCR product (8). AdV reference strains from the ATCC were used as positive controls.

In situ hybridization (ISH).

Sections displaying a thickness of about 5 μm were prepared from paraffin-embedded tumor tissue and placed onto silane-coated slides. Pretreatment of the sections, proteolytic treatment, hybridization, and detection were performed using the Universal DISH&AP detection kit (PanPath, The Netherlands), according to the manufacturer's recommendations. For specific detection of human AdV, 10 μl hybridization solution containing biotinylated AdV type 5 probe (Enzo Diagnostics, NY) was used per section. Biotinylated oligonucleotides complementary to ALU repeats were used as positive-control probes, and the biotinylated vector pSP served as a negative control.

Statistics.

A minimum of 30 specimens per tumor entity were studied. In the case of negative virus DNA tests in all specimens, an association between the presence of AdV and the respective tumor type can be excluded with the upper limit of the one-sided 95% confidence interval being 90%. The distribution of AdV species in brain tumors was compared by using Fisher's exact test (SAS).

RESULTS

AdV screening in pediatric leukemia and lymphoma.

The great majority of 120 leukemia samples investigated did not reveal any evidence of AdV sequences. Only 1 of 30 B-ALL and 1 of 30 T-ALL specimens investigated showed results positive for AdVs (lower limit of the 95% confidence interval, <15%; Table 2 ). Among 30 Hodgkin's disease specimens, four tested AdV positive by the two-reaction RQ-PCR assay and the species-specific RQ-PCR assay (9, 34), revealing the presence of AdV species B in three patients and species D in one patient. Within 29 B-NHL specimens investigated, one tested positive for species B and one for species C (Table 2).

TABLE 2.

Investigation of AdV in pediatric tumors^e

Tumor type	No. of samples		Species	Serotype(s)	Viral oncogenes detected
	Total	AdV positive
Leukemias/lymphomas
B-ALL	>30	1	C	NA	NA
T-ALL	>30	1	B	NA	NA
AML	30
CML	30
NHL	>30a	2	B, C	NA	NA
Hodgkin's disease	30	4	B, D	D08, NA	NA
Solid tumors
Wilms' tumor	30
Neuroblastoma	30
Ewing tumor	30
Rhabdomyosarcoma	>30b
Osteosarcoma	30
Ependymoma	30c	20	D, B, C	C01, B50, NA	E1A, E1B
Glioblastoma	30d	25	B, D	B50, B16, D10	E1A, E1B, E4orf1
Oligodendroglioma	30d	22	B	B50	E1A, E1B, E4orf1
Astrocytoma	30
Medulloblastoma	>30
Germinal cell tumor	30
Total	538	75	B, C, D	B50, B16, C01, D08, D10	E1A, E1B, E4orf1

^aIncluding 29 B-NHLs and 9 T-NHLs.

^bIncluding 26 embryonal and nine alveolar rhabdomyosarcomas.

^cIncluding five adult and 25 pediatric ependymomas.

^dGlioblastoma and oligodendroglioma specimens were from adult patients.

^eAbbreviations: AML, acute myeloid leukemia; CML, chronic myeloid leukemia; NA, not available.

AdV screening in pediatric solid tumors.

Investigation of 30 specimens derived from Wilms' tumor, Ewing tumor, neuroblastoma, rhabdomyosarcoma, osteosarcoma, and germinal cell tumors showed negative results in all instances, suggesting the absence of association between AdV infection and these tumor entities (lower limit of the 95% confidence interval, 10%). Among brain tumors, which account for about 20% of pediatric malignancies (23), AdV positivity has been detected in three different entities including glioblastoma (25/30), oligodendroglioma (22/30), and ependymoma (20/30) (Table 2). By contrast, all 30 astrocytoma specimens and all evaluable medulloblastoma specimens investigated tested AdV negative (Table 2). The RQ-PCR assays described elsewhere (9, 34) revealed a predominance of species B in glioblastoma and oligodendroglioma, while species D was present in only two glioblastoma specimens (Fig. 1a). In ependymoma, the majority of AdV-positive specimens revealed species D and less commonly species B or C (Fig. 1a). Quantitative analysis performed in relation to a single-copy control gene (the beta-2-microglobulin gene) (48) mostly indicated low virus copy numbers, ranging from one viral genome per 100 to one per 1,000 human cells (Fig. 1b). An adequate DNA amount and quality for serotype identification were available only in a proportion of AdV-positive brain tumor specimens. The AdV serotypes detected are displayed in Table 2.

FIG. 1.

Real-time PCR analysis of AdV-positive brain tumors. (a) In glioblastoma and oligodendroglioma, AdV species B was most commonly detected, whereas in ependymoma species D was predominant. The predominant occurrence of AdV species D in ependymoma was statistically significant (P < 0.001), while the preponderance of AdV species B in the other brain tumor entities did not achieve statistical significance. n. a., not available for AdV species identification due to inadequate DNA quality; CNS, central nervous system. (b) Amplification curves of the hexon gene in AdV-positive brain tumors (glioblastomas) generally showed threshold cycles in a range between 35 and 38, as depicted in the displayed examples of six brain tumor specimens positive for AdV species B. The single-copy control gene, the beta-2-microglobulin gene, revealed threshold cycles at or around 27, indicating a difference in copy numbers from AdV target genes of up to 3 logs. Other viral genes, such as E1A, E1B, and E4orf1, were detected in all AdV-positive brain tumor samples (Table 2), at the same level as the hexon gene.

Five specimens from each brain tumor entity that tested AdV positive by RQ-PCR analysis were randomly selected and investigated by an independent technical approach, ISH, using a biotinylated pan-AdV DNA probe and color-based detection (Fig. 2a to d). In all instances, a number of AdV-positive cells could be detected within distinct areas of histological brain tumor sections. Cells displaying positive hybridization signals were identified as tumor cells by microscopic examination performed by experienced neuropathologists. Adequate performance of the ISH investigation was documented by positive and negative controls (Fig. 2e and f). The observation of loosely scattered clusters of AdV-positive cells by ISH (Fig. 2d) is in accordance with the low virus load detected by RQ-PCR. Brain tumor samples derived from glioblastoma, oligodendroglioma, and ependymoma that tested AdV negative by RQ-PCR were also investigated by ISH. In most instances, AdV-positive cells were found in the tumor sections, indicating that RQ-PCR investigation may have underestimated the number of positive brain tumors, most likely due to the small size of core needle biopsy samples used for PCR analysis.

FIG. 2.

Detection of AdV DNA in brain tumor tissue by ISH. (a to c) AdV-positive cells in ependymoma (a), glioblastoma (b), and oligodendroglioma (c) are visualized by a color reaction (panels a, b, and c, 1,000-fold magnification). Cells were counterstained with nuclear fast red. (d) Loosely scattered AdV-positive cells in an ependymoma specimen. (e) Positive control: biotinylated oligoprobes against ALU repeats. (f) Negative control: biotinylated pSP vector DNA (panels d, e, and f, 500-fold magnification).

AdV screening in nonmalignant brain tissue.

To assess whether the detectability of AdV DNA in brain specimens is restricted to malignant cells, we have screened corresponding nontumorous brain tissue sections and specimens of choroid plexus papilloma, a well-defined low-grade tumor. RQ-PCR screening revealed the presence of AdV DNA of species B and D in ependymal cells (5/10), in plexus choroideus specimens (four/five), in the periventricular white matter (two/five), and in choroid plexus papilloma samples (two/five) (Table 3).

TABLE 3.

RQ-PCR analysis of AdV in nonmalignant brain tissue

Brain tissue specimen type	No. of samples		Species
	Total	AdV positive
Ependymal cells	10	5	B, D
Plexus choroideus	5	4	B, D
Periventricular white matter	5	2	B
Choroid plexus papilloma	5	2	B

DISCUSSION

The aim of the study presented was to investigate the putative occurrence of AdVs in childhood tumors. Among more than 500 specimens derived from 17 different tumor entities, the majority tested negative for human AdV sequences. The absence of AdV DNA in all 30 specimens of a respective tumor investigated suggests the lack of association between this entity and AdV infection associated with the presence of the complete virus genome (lower limit of the 95% confidence interval, 10%). In view of the fact that pediatric malignancies had previously not been systematically studied for the presence of adenoviral sequences, the negative findings for most tumor entities provide a definitive answer to the long-lasting enigma regarding the potential involvement of these viruses in most types of childhood cancer.

The observation of one AdV-positive specimen within 30 samples of T-ALL and B-ALL, respectively, might be attributable to occasional persistence of the virus in peripheral blood lymphocytes. Of lymph node specimens derived from pediatric lymphoma patients, about 10% were AdV positive. These observations are in line with an earlier study reporting the detection of AdV serotype 5 in a subset of adult lymphoma patients (13). In animal model systems, the oncogenic potential of AdVs has been described as serotype specific. Primary infections with AdVs in humans usually occur in childhood, and lymphatic tissue is one of the sites of virus persistence (17, 45). It would be of interest to analyze specimens from adult patients with lymphomas and lymphocytic leukemias for the entire spectrum of AdV serotypes to determine whether the long-term presence of the virus in lymphocytes is more frequently associated with malignant disorders derived from this cell type.

To the best of our knowledge, the findings presented in this report provide the first piece of evidence for the frequent occurrence of human AdV DNA in different brain tumors. It is important to point out, however, that our observations also revealed the presence of AdV DNA in nonmalignant brain tissue. The proportion of cells containing viral sequences, including the early transcribed oncogenes and the major capsid gene, was small in all positive specimens. It is not possible, therefore, to exclude the possibility that our observations merely reflect a tropism of AdVs to brain tissue, but it is also conceivable that the viruses may persist in brain cells after primary infection and that their oncogenic properties could contribute to the pathogenesis of different brain tumors. The low percentage of brain tumor cells carrying viral sequences is compatible with a hit-and-run mechanism, which has recently been proposed for AdVs on the basis of in vitro studies (39). Similarly, low viral genome copy numbers in tumor cells have also been described for simian virus 40, which is well known for its transforming potential and is a putative tumor-inducing virus in humans (32).

The detection of adenoviral DNA in brain tumor cells is in accordance with the described capability of the virus, particularly of species A and B, to induce neural tumors in rodents (10). Moreover, one of the rare human cell lines transformed with AdV (HEK293) was reported to reveal an expression pattern of intermediate filament proteins similar to those of early-differentiating neurons. Another human cell line successfully transformed with adenoviral genes (HER 911) is also of neuronal origin, thus providing additional support for the notion that AdVs are capable of transforming human neuronal cells (11, 43).

Detection of AdV DNA by RQ-PCR in a large number of clinical specimens raises questions about contamination as a source of potentially false-positive results. The brain tissue specimens analyzed were provided by a single institution, and the specimens were derived from different years of diagnosis scattered across more than 2 decades, thus minimizing the possibility of contamination with a particular viral pathogen during initial processing and storage of the specimens. Moreover, different AdV types have been reproducibly detected in the brain tumors studied, whereas astrocytoma and medulloblastoma tested completely negative for AdV DNA sequences. These observations and the detection of AdV sequences in brain specimens by ISH, a technique not prone to false-positive results due to contamination, indicated that our findings were not attributable to inadvertent introduction of extraneous AdV DNA at any stage of sample processing.

The central nervous system is a site of infection by different neurotropic viruses, such as human immunodeficiency virus or mumps virus, which have the capacity to cross the blood-brain barrier, consisting of a single layer of endothelial cells. Various theories on how virus particles may gain access to the brain are currently discussed: (i) access via infected lymphocytes and migration through the brain microvascular endothelium, (ii) passage of free virus particles either by migration between or transcytosis through the endothelial cells, and (iii) release of virus into the brain by the infected endothelium (1). Virus passing from blood into the stroma of the choroid plexus potentially can infect epithelial cells and seed virus directly into the cerebrospinal fluid or potentially can be transported via pinocytotic vesicles through the elongated epithelial cells (31). It is conceivable that AdVs also follow one of the pathways described. Moreover, the absence of AdV DNA in medulloblastoma could be attributable to the different origin of the tumor and its localization in the cerebellum, whereas the negative results for astrocytomas might be related to the poor vascularization of these tumors compared with other malignancies in the brain. The presence of the virus in tumor cells might be linked to the degree of vascularization, which could facilitate hematogenic spreading of the virus to or within the central nervous system. Most AdV species enter human cells via the CAR, while members of species B use the cellular CD46 receptor for recognition and initiation of virus uptake (16). The ependyma has been described as the only cell type in the brain that expresses the CAR. The occasional observation of AdV species C and the predominance of species D in malignant and nonmalignant ependymal cells are in line with the reported receptor expression. By contrast, in glioblastoma and oligodendroglioma serotypes of species B were the predominant AdV type observed.

To our knowledge, there have been no previous reports on the systematic investigation of the entire spectrum of AdVs in the brain covering both malignant and nonmalignant tissue. Although brain tissue tropism has been previously described for murine AdV type 1 in a mouse model (22), only a small number of case reports describing the occurrence of AdV in the human brain have been published to date (5, 14). The screening data presented in this study provide new insights into AdV infection of cerebral tissue and the brain as a potential sanctuary for virus persistence and raise questions about the possible involvement of these viruses in the pathogenesis of brain tumors. However, over the past decades, no clear association of AdV infection with any human tumor could be established. AdVs have therefore been commonly regarded as nonharmful to humans with regard to oncogenicity, and they are currently being exploited as vectors for targeted gene therapy (6). Nevertheless, data have been lacking to firmly exclude any long-term risk associated with the therapeutic use of AdV vectors in humans. It is conceivable that the adenoviral oncoproteins interact with the cell cycle regulation of infected cells, thereby contributing to the process of malignant transformation. For a number of well-established tumor viruses, such as the human papillomavirus, human T-cell lymphotropic virus type 1, or the Epstein-Barr virus, the first indication of their implication in tumorigenesis was the detection of their presence in specific cancer entities (12, 25, 41). The identification of AdV DNA in brain tumors may therefore be seen in the light of earlier observations for different viruses which were later identified as tumorigenic in humans. A possible causal relationship of AdV infection with persistence in brain tissue and the pathogenesis of malignant brain tumors remains to be established.