Going bananas over brain tumors!

Only limited data exist regarding associations between diet and cancer in children. This is partly due to a) methodological limitations in measuring dietary exposures; b) difficulty in determining whose diet (dad, mom, child) might have importance and when; and c) lack of supporting biological mechanisms. One previous area of focus has been the potential association between maternal consumption of N-nitroso compounds (e.g., hot dogs and other cured meats) during pregnancy and the risk of brain tumors in offspring [See C3 Vol 4 No 5; Vol 5 No1]. A recent study from Israel by Lubin et al [Int J Cancer, 86:139-143, 2000] explored the role of dietary habits during pregnancy and postnatal life and the occurrence of pediatric brain tumors. Mothers of 300 children with incident brain tumors diagnosed at less than 18 years of age, and 574 controls were interviewed. A detailed food frequency questionnaire was administered which included 100 items, enabling the authors to explore nitrate and nitrite intake as well as other dietary components. For all cases, no increased risk was observed with increased consumption of foods containing nitrates or nitrites nor was there any protective effect observed with increasing intake of fruits and vegetables.  Increased consumption of vegetable fats by the children was significantly associated with an increased risk (OR 1.9; 95%CI 1.1-3.2; highest vs lowest intake; p trend 0.01). Vitamin E consumption, which is highly correlated with vegetable fat intake, was also associated with an increased risk. Finally, increased maternal consumption of potassium during pregnancy was associated with a significantly increased risk (OR 2.1; 95% CI 1.1-3.7).  The analysis was stratified by two main diagnostic groups (group 1 (95 cases): PNET/medulloblastoma, neuroblastoma; group 2 (151 cases): astrocytoma, glioma, oligodendroglioma, glioblastoma). Results were consistent with the total group combined.  The authors acknowledge that their study failed to find an association with n-nitroso compounds. However, they suggest further exploration of the associations between vegetable fat and potassium and the risk of childhood brain tumors. 

COMMENT: This study provides further evidence that the association between maternal consumption of n-nitroso compounds and childhood brain tumors is weak, if present at all. Secondly, this study provides an 
opportunity to discuss the limitations of our current methodology. It is somewhat surprising that the authors failed to mention the number of dietary associations that were explored; it is very possible that their significant positive findings were purely chance. Further, it is imperative that we acknowledge the limitations of interpreting p values less than 0.05, as they do not necessarily represent “the truth”. In particular, the positive association with potassium was discussed by the authors as possibly being supported by a study that showed a ‘change in membrane K+ conductase that induces swelling of cultured glioma cells in rats’. These laboratory conditions have little supportive significance in explaining the relationship between maternal consumption of potassium-rich foods and the occurrence of brain tumors in offspring. While it is recognized that hypothesis-generating studies might be useful in offering directions for further study, it is imperative that we remain cautious when looking for supporting evidence in the literature to explain the positive associations found in these types of analyses. Julie A. Ross

Polyoma viruses and brain tumors: A game of hide and seek?

As noted previously [C3, Vol 10, No 6; Vol 8, No 3 ], there is considerable interest in exploring the role of polyoma viruses in carcinogenesis. In particular, JC, BK, and SV40 have been postulated to be important in the development of brain tumors. Most of the definitive associations have been reported in animal models such as hamsters, owls, and macaque monkeys (e.g., intracerebral injection of polyoma viruses into these animals can cause ependymomas, astrocytomas, and glioblastomas). Although polyoma viral genomes have been found in some human cancers (including ependymomas), it is unclear whether or how these viruses may be important in tumor etiology. However,  it is known that the transforming activity of the polyomavirus large T-antigen (one of the proteins encoded by the virus) primarily relies on its ability to inactivate p53 and proteins of the Rb family. In this study from Germany, Weggen et al [Brain Pathology 10:85-92, 2000] examined brain tumors that typically do not contain p53 alterations including 116 medulloblastomas, 131 meningiomas, 25 ependymomas, and 2 subependymomas. They also included 60 hepatoblastomas (which can arise in transgenic mice by liver directed large T- antigen expression ) and 31 brain samples from patients with schizophrenia (controls). Polyoma viruses were infrequently found in these specimens; only 2 medulloblastomas, 1 meningioma, 1 ependymoma and 1 subependymoma contained an SV40- like sequence. One menigioma contained a JC virus sequence. BK virus sequences were not detected in any of the samples. Moreover, the hepatoblastomas and brain samples from patients with schizophrenia were negative for all viruses. 

COMMENT: The authors conclude that these polyoma viruses are unlikely to play a major role in the etiology of medulloblastomas, meningiomas, and ependymomas. However, they do bring up an interesting potential explanation for the low frequency of observed viral  sequences in these specimens. There are data in transgenic mice to suggest a “hit and run” mechanism, whereby the tumor relies less on the viral sequence for maintenance of transformation than for initiation. For example, one study observed increased malignant potential in a rat brain tumor cell line after loss of the large T-antigen coding region and concurrent mutation in p53 [Cancer Res  59; 1980-6, 1999].  The potential role of viruses in brain tumors is intriguing and is likely very complex. We look forward to further work in this area.  Julie A. Ross

Another imprinting disease?

We have noted previously the role of alterations of imprinting (expression of only the paternally inherited allele) of insulin-like growth factor 2 (IGF2) at 11p15.5 in Beckwith Wiedeman syndrome (BWS) and its associated tumors [C3 vol 4, No 3; vol 2, No 5].  In a fascinating report, Sperandeo et al [Am J Hum Genet. 66:841, 2000] describe two cousins (sons of two sisters), one with BWS, and one with the overgrowth syndrome Klippel-Trenaunay-Weber (KTW). KTW is a disease of unknown etiology, characterized by cutaneous hemangiomata and regional hypertrophy of bones and soft tissue, but no cancer predisposition.  Little is known about the molecular etiology of KTW. In a single case, a translocation at 11p15.1 was reported [Whelan et al, Am J Med Genet, 59:492, 1995]. Rare familial aggregation occurs, suggesting a predisposing single gene defect. In this study, the child with BWS and the child with KTW showed relaxation of imprinting (expression of both alleles) at IGF2 in cultured skin fibroblasts. Interestingly, both children have hemihypertrophy, and imprinting was relaxed in fibroblasts from the large and the normal-sized sides of the body in both cases. The boys inherited different haplotypes at 11p15.5, excluding a common genetic defect linked to this chromosomal locus.  In contrast to the child with KTW, the child with BWS showed altered methylation at the locus KvDMR1, 350kb centromeric to IGF2. 

COMMENT: The authors propose that these data support the following hypotheses: 1)These children may carry a defect in a gene that regulates imprinting, causing different phenotypes depending on which genes are demethylated or lose their imprint; 2) Demethylation of KvDMR1 may be a BWS specific alteration; 3) Over-expression of IGF2 may play a role in KTW;  4) Loss of imprinting of IGF2 alone is not sufficient to cause hemihypertrophy; and 5) BWS and KTW may represent the result of a clinical spectrum of somatic overgrowth caused by partially overlapping epigenetic alterations. Further studies of KTW may provide some answers. Stella M. Davies

Looking for clues in all the right places 

Familial hemophagocytic lymphohistiocytosis (FHL) is a rare, rapidly fatal immune disorder.  Clinically, the disease presents in young children, typically with fevers, lymphadenopathy, hepatosplenomegaly, pancytopenia, coagulopathy, neurological abnormalities, and high serum levels of interferon-^a and tumor necrosis factor-^a. Pathologically, there is accumulation of activated macrophages and lymphocytes and hemophagocytosis in the bone marrow, liver, spleen, lymph nodes, and central nervous system. Stepp et al [Science 286:1957-59, 1999] demonstrate that at least some cases of this disorder are due to defects in perforin. The authors started with the hypothesis that the primary inherited defect in FHL could be a failure of cytolytic lymphocyte function and that this, together with childhood infections induces the fatal immune deregulation typical of FHL. The gene encoding perforin is an important mediator of lymphocyte cytotoxicity and has been mapped to 10q22 which is near one of the previously identified FHL loci. Importantly, perforin knock-out mice appear to be generally healthy when maintained in a pathogen-free controlled environment. However, when infected with lymphocytic choriomeningitis virus (LCMV), similar immunopathology and mortality to human FHL are seen. The authors first confirmed that the perforin gene (PRF1) is in the candidate deleted region of chromosome 10. They then studied 8 unrelated 10q21-22 linked FHL patients by sequencing exons 2 and 3 of the perforin gene. Of note, 5 of the patients were from consanguin-eous families. Nine independent mutations in exons 2 and 3 of the perforin gene were detected. Cultured lymphocytes taken from the patients had defective cytotoxic activity and immunostaining revealed almost no perforin in the granules.

COMMENT: FHL and its non-familial version, HLH, are troublesome to diagnose and to treat. This study starts to unravel the complex etiology and inheritance of this disease. These new data will surely stimulate additional studies to determine the role of perforins in cases that do  not appear to be linked to 10q21, and in cases that do not appear to be familial. Better understanding of the etiology of this disorder may lead to improved treatment options which are currently generally disappointing.  Stella M. Davies

Briefly noted
Guo et al [Oncogene 18: 4948, 1999] note that loss of heterozygosity (LOH) at 11q23 is common in MYCN single copy neuroblastoma, suggesting the presence of a tumor suppressor gene  within 11q23. In contrast to allelic deletion at 1p, 11q LOH is inversely correlated with MYCN amplification, so may define a genetically distinct set of neuroblastomas. In univariate analysis, 11q LOH was associated with decreased survival, which was not shown with multivariate analysis. Stella M. Davies
 

C3 Quarterly Newsletter
Children's Cancer Research Fund
Epidemiology Research Unit
Division of Pediatric Epidemiology
Clinical Research
University of Minnesota
420 Delaware St. SE, Box 422
Minneapolis, MN 55455
pedsepi@umn.edu

Editors: 
Stella M. Davies, MD, PhD, and Julie A. Ross, PhD