Sperm count, too!

 

Genomic imprinting, or the preferential expression of   a gene depending on the parent of origin, results from the differential methylation of CpG islands. The imprint is erased during gametogenesis through demethylation and reset later according to the parent of origin of the transmitted chromosome. Demethylation of maternally transmitted alleles is a mechanism by which some genetic disorders and malignancy arise. While demethylation has been described in the maternal allele (see below), there has been little evidence that this phenomenon occurs in the paternal allele. Marques CJ, et al [Lancet 2004; 363:1700-1702] examined spermatozoan DNA from 123 individuals (27 normozoo-spermic and 96 oligozoospermic, including 46 moderately affected and 50 severely affected) undergoing routine analysis for infertility. Methylation profiles were studied in two imprinted genes: MEST (maternally imprinted) and H19 (paternally imprinted). For MEST, maternal imprinting was correctly erased in all 123 samples. In contrast, while the H19 paternal imprint was correctly erased in the normozoospermic samples, 8 (17%) of the moderately affected and 15 (30%) of the severely affected men showed incomplete methylation. The authors conclude that abnormal spermatogenesis is associated with a rise in methylation defects at H19. Since the maternal imprint was correctly erased for MEST, this may indicate that changes in DNA methyltransferase activity are responsible for hypospermato-genesis. Importantly, there is coordinated regulation of expression of insulin-like growth factor-2 (IGF2) (a maternally-imprinted gene) by repressor binding sites on the H19 gene [Reik W et al Nat Gen Rev 2001; 2:21 -32]. Thus, paternal H19 hypomethylation could lead to the inactivation of both IGF2 genes, which could have deleterious consequences during fetal development.

 

The Egg and Me(thylation)

 

Several recent studies have investigated the potential association between assisted reproductive technology (ART) and risk of childhood cancer (e.g., retinoblastoma) or genetic syndromes at high risk of malignancy (e.g., Beckwith-Wiedemann Syndrome (BWS)) [see C3, Vol 14, No 1; Vol 15, No 2]. Most recently, Halliday J et al [Am J. Hum Genet 2004; 75:526-528] conducted a case-control study in Victoria , Australia to measure the association between ART and BWS.   Of 1.3 million live births during the period 1983-2003, 37 cases of BWS were born (confirmed by clinical geneticists) for an overall population prevalence of about 1 case per 35,500 births. These cases were each matched (age, maternal age, and parity) to four live-born controls. Record linkage was performed with the providers of ART services in Victoria to determine which infants were conceived using ART methods. Of the 37 BWS cases, 4 were conceived by ART compared to 1 of the 148 matched controls (Odds Ratio 17.8, 95% CI: 1.8-432.9). In the context of all 14,894 live births conceived by ART during this time period, the absolute risk of BWS was approximately 1 case per 4,000 births or about 9 times higher than in the general population. As the authors note, previous reports of ART-associated BWS consistently demonstrate hypo-methylation of the maternal KvDMR1/LIT1 locus at 11p15.5 [Gicquel C, et al Am J Hum Genet 2003; 72:1338-1340; Maher ER, J. Med Genet 2003; 40:62-64]. In contrast, this mechanism is only observed in about 45% of the overall BWS population. It is possible that the process of ART preferentially results in maternal allele demethylation, which would have important implications not only for BWS but childhood cancer as well.

 

COMMENT: Due to the excellent follow-up and population linkage available, the study by Halliday et al is one of the first to quantify the risk of ART-associated BWS. Moreover, the study by Marques et al indicates that demethylation patterns are not necessarily confined to the maternal allele. A limited institution study has recently been awarded to Dr. John Heath, a young investigator in the Children's Oncology Group, to investigate the potential association between ART and childhood cancer. In addition to collecting epidemiological data, this study will also describe methylation patterns of H19, KvDMR1/LIT1, and other imprinted genes in selected childhood malignancies relative to ART.

Julie A. Ross

 

Tall tales about bone cancer

 

The descriptive epidemiology of osteosarcoma (OS) and, to a lesser extent Ewing 's sarcoma (ES), suggests an etiology related to bone development. Incidence rates for both OS and ES mirror the pediatric growth curve and peak earlier in girls than in boys.   However, the case that OS is associated with bone growth is stronger in that a large proportion of tumors develop in the bones of the leg, which lengthen the most during the adolescent growth spurt, whereas ES tumors occur on the skeleton more evenly.   Based on the supposition that taller children reach their height faster, findings have been inconsistent regarding stature at diagnosis of the bone sarcomas. Recently, Cotterill S et al. [Pediatr Blood Cancer 2004; 42: 59-63] revisited the topic of height in bone sarcomas. Information was collected on sex, tumor location, age, and height at diagnosis of 364 cases of OS and 356 cases of ES treated in clinical trials run by two United Kingdom cooperative groups between 1978 and 1997. Each case was assigned an age- and sex-appropriate standard deviation score (SDS) based on UK growth charts for 1990. An SDS of -1, 0, and 1 signify heights at the 16 th , 50 th (i.e. median), and 84 th percentiles, respectively; mean SDS is significantly different from the population mean if the 95% CI does not include 0. Height and age at diagnosis of various groups of cases was compared using t-tests and linear regression.   Firstly, they confirmed that females are diagnosed at significantly earlier ages than males for both OS (median 15.6 vs. 17.4 years; p = 0.009) and ES (median 12.8 vs. 14.3 years; p = 0.01).   Males did not have significantly higher SDS than females for OS (p = 0.2) or ES (p = 0.9).   Cases of OS were significantly taller than UK children on average (mean SDS = 0.20; 95% CI: 0.07-0.33).   The mean SDS score of OS cases with tumors of the femur (mean SDS = 0.45; 95% CI: 0.26-0.64) was significantly different (p = 0.0001) from that of OS cases with tumors on other bones (mean SDS = -0.06; 95% CI: -0.22-0.10). ES cases were not significantly taller than the general population overall (mean SDS = 0.09; 95% CI: -0.02-0.21).   In ES sub-group analyses, children diagnosed at ages <15 years were significantly taller (mean SDS = 0.20; 95% CI: 0.05-0.35), while those diagnosed at ages > 15 years (mean SDS = -0.07; 95% CI: -0.25-0.10) and those with femoral tumors (mean SDS = 0.19; 95% CI: -0.11-0.49) were not.

 

COMMENT: This study establishes that OS cases and younger ES cases are taller at diagnosis than average children in the UK .   As the investigators note, it is possible that the cancers themselves increased growth. Still the results are interesting, especially the novel finding that it is OS cases with femoral tumors, specifically, that are taller than average. The humerus, like the femur, lengthens considerably during puberty. However, OS cases with humeral tumors were not taller than average. One explanation for this is that it is the combination of rapid growth and the stress of weight bearing that increases risk of OS in the femur.   Certainly it will pay in future studies of OS to perform tumor-site-specific analyses.   Younger ES cases were taller than average while those with femoral tumors were not.   No data were available in this study on pubertal stage, which along with age and sex is a predictor of height.   The authors suggested that taller height in ES cases < 15 years old may be a marker for earlier puberty and, if so, then it is hormones rather than growth velocity that influence ES risk.   This study could easily be replicated elsewhere and should be, but research into bone sarcoma etiology ought not to stop there.   Rather, it would be most interesting to examine the molecular mechanisms that control a substantial portion of growth and pubertal development.  

Logan G. Spector

 

 

Take this job and study it

 

As noted previously [see C3 Vol 13, no 5; Vol 13, no 6; Vol 15, no 1], infections are of interest in the etiology of childhood leukemia either through potential direct exposure to an infectious agent or through an unusual response to infection. Because of difficulty in measurement, researchers have preferred to study proxies such as daycare attendance or birth order. These characteristics are easily recalled and increase the likelihood of infection exposure.   Another proxy is parental occupation, since parents whose jobs bring them into contact with many people may be more likely to acquire infections and pass them to their children.   A new study from the UK [Pearce MS et al. Epidemiology 2004; 15(3): 352-356] is the latest contribution to the growing literature on the topic. Cases of leukemia and non-Hodgkin lymphoma (NHL) diagnosed in the population-based Northern Region Young Persons' Malignant Disease Registry between 1968 and 1997 at ages 0 to 14 years were compared to two control groups. The first consisted of all patients in the registry with diagnoses other than leukemia and NHL who were the same sex and year of birth as a case (registry controls). The second consisted of 100 children for each case chosen from the Cumbrian Births Database, matched for sex and year of birth (CBD controls). Data on paternal employment, which is available on all UK birth certificates, were abstracted and classified into jobs the authors considered to have standard, high, or very high level of contact.   Odds Ratios (ORs) and 95% CIs relating leukemia, overall and at the childhood peak of 2-5 years of age, and NHL to paternal occupational contact level were obtained using logistic regression. No significant association of high paternal occupational contact level was seen compared to standard contact levels in any diagnostic or age groups. However, there were significant associations of very high paternal occupational contact level (ORs were 1.5 (95% CI: 1.2-1.9) for leukemia and NHL combined, 1.4 (95% CI: 1.1-1.8) for all leukemia, 1.8 (95% CI: 1.3-2.5) for ALL at ages 2-5 years, and 1.9 (95% CI: 1.0-3.6) for acute non-lymphoblastic leukemia comparing cases to CDB controls). Similar results were obtained for the comparison to registry controls.   These associations appeared stronger with residence in rural compared to urban areas.   For instance,   ORs comparing very high or high paternal occupational contact level to standard contact were 2.0 (95% CI: 0.8-5.0) and 1.3 (95% CI: 0.9-1.8) among those with rural or urban residence, respectively, for ALL at ages 2-5 years and CBD controls.   Analysis of single categories of very high contact jobs showed significantly increased risk of NHL and leukemia, overall and in most sub-groups, for children of teachers and policemen.

 

COMMENT: A solid population base and use of routinely collected data served to minimize bias in this study. This study had findings comparable to several previous studies on parental occupational contact level but also shared their limitations. Parental occupational contact during a child's lifetime was only partially accounted for, since occupation was derived from birth certificates for fathers alone, and the estimation of occupational contact is mostly educated guesswork.   The study of infection proxies would benefit from validation studies to verify that they affect exposure to infections as is commonly assumed.

Logan G. Spector

 

 

 

 

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