Attack of the Clones?

Studies on monozygotic twins with concordant leukemia and retrospective scrutiny of neonatal blood spots of patients with leukemia indicate that chromosomal translocations characteristic of pediatric leukemia often arise prenatally, probably as initiating events.With the exception of leukemias that occur in the first year or so of life, twin studies and animal studies all suggest that additional postnatal exposure and acquisition of subsequent genetic abnormalities are required for clinically overt leukemia development. This leads to the possibility that chromosomal translocations, functional fusion genes, and preleukemic clones might be present in the blood of healthy newborns at a rate that is significantly higher than the cumulative risk of the corresponding leukemia.To address this hypothesis, Mori H, et al [Proc Natl Acad Sci 99:8242-7, 2002] have studied approximately 500 normal cord bloods. The data indicate that the chromosomal translocations, TEL-AML1 and AML1-ETO, can be identified in approximately 1% of normal infants.This study used meticulous care to prevent contamination or amplification of non-functional translocations from confounding the analysis. The 1% frequency of fusion gene positive cord blood is approximately 100 times the cumulative frequency of overt clinically diagnosed leukemia with the same fusion genes.The authors suggest that the fusion gene positive cells represent expanded clones of preleukemic cells that will remain pathologically and clinically silent, or covert, in the absence of additional postnatal genetic events. Of note, the fusion gene products detected were all in-frame and potentially producing functional protein. Cells with the TEL-AML1 fusion in normal cord blood were exclusively of B-lineage, and the AML1-ETO fusion was found in cells of myeloid lineage in agreement with the cells of origin of the leukemia associated with these fusions.

COMMENT: This study was carefully performed in an area that can be fraught with technical difficulties as low frequency clones are being identified with a powerful technique (PCR) that can be susceptible to contamination, or amplification of related but non-leukemogeneic clones. The data are of significance for all investigations of the etiology of leukemia as it identifies the time point (prenatal) when leukemia is initiated. This study suggests that investigations of susceptibility or exposure that might cause such an event need to be directed to the nine months spent in utero.  Studies to investigate events associated with acquisition of secondary abnormalities should be directed toward the postnatal period. In addition, it raises the more distant possibility of neonatal screening for the presence of clones, if a preventive intervention could be devised to protect children destined to evolve to full blown leukemia. 

Hyperdipoidy is frequent in childhood ALL (30%) and is typically associated with low risk features and a good prognosis.Among leukemias with more than 50 chromosomes, there is evidence that the hyperdiploid karyotype arises from a single abnormal cell division.Panzer-Grumayer R, et al [Blood, 2002, 100: 347-349] used an analysis of immunoglobulin heavy chain gene rearrangements to define the order of cellular events leading to nondisjunction. The study investigated a single leukemia that has 3 copies of chromosome 14 and 3 different non-functional IgH rearrangements (one on each chromosome 14). The data show that nondisjunction occurred during very early B-cell differentiation. Moreover, the authors show the presence of the 3 IgH rearrangements in neonatal blood spots from the child, indicating that nondisjunction took place in utero. Clinical leukemia was diagnosed at 2.6 years of age.

COMMENT: These data again show the importance of in utero events in the investigation of initiation of leukemia, and the potential for long latencies. This offers a clearly defined time frame for investigating etiology. 

There have been several studies suggesting a link between autoimmune disease in families and risk of childhood malignancy. In particular, maternal (non-gestational) diabetes has been associated with a slight increased risk of childhood cancer (primarily leukemia and lymphoma). In this report, Westbom L, et al [Br J Cancer 2002; 86:1078-80] performed a record linkage between the Swedish Cancer Registry and the Medical Birth Registry to explore potential associations between maternal autoimmune diseases and childhood cancer risk. Autoimmune diseases included insulin-dependent diabetes, systemic lupus erythe-matosus, rheumatic arthritis, scleroderma, Crohn’s disease or ulcerative colitis, multiple sclerosis, and thyroiditis. Of 4380 children born in 1987-1997 to women with diabetes and who lived beyond one year of age, ten were diagnosed with cancer before the end of 1998. After adjusting for year of birth, maternal age, parity, multiple birth, and birth weight, the odds ratio was 2.3 (95% CI=1.2-4.2), suggesting an increased risk of childhood malignancy associated with maternal (non-gestational) diabetes. Malignancies included retinoblastoma, kidney cancer (n=2), leukemia (n=5), bladder cancer and sarcoma. No other notable associations were found among the 5842 children born of mothers who had other autoimmune diseases. The authors suggest that their findings indicate a mechanism other than autoimmunity, given that only diabetes was positive. However, they stress that even if diabetes was a risk factor, the risk is small, and is unlikely to contribute substantially to population cancer rates.

COMMENT: As we have noted in the past, there is always concern regarding recall bias in a case-control study, where mothers of cases may over-report insignificant events. However, this is essentially a population-based cohort study; the Swedish records are nearly 100% complete, and the exposure measure (in this case, maternal autoimmune disease) was recorded prior to the child developing cancer. Thus, this study lacks the problem of recall bias. The strengths associated with a record linkage study provide some impetus to explore this association in additional studies. In particular, the positive associations with kidney cancer (2 cases) and leukemia (5 cases) may be more noteworthy than the other malignancies. It is unclear, however, why maternal insulin-dependent diabetes would be associated with an increased risk of childhood cancer, and it would be helpful if there was some biological foundation behind it. 

In a recent report from California, Reynolds P, et al [Am J Epidemiol 155:603-13, 2002] explored birth characteristics and leukemia in children under the age of 5. A total of 1957 leukemia cases (including 1407 ALL cases and 240 ANLL cases) were ascertained during the period 1988-1997 using the statewide cancer registry in California and linked with birth certificates; 88% of cases matched to a California birth certificate. Two randomly selected control birth certificates, matched on date of birth and sex were obtained for each case. Using conditional logistic regression, the investigators reported results separately for ALL and ANLL. They confirmed a positive association with Down syndrome for both subtypes of leukemia; 22 cases and no controls had Down syndrome recorded on the birth certificate. They further noted that African American children had a substantially reduced risk of ALL (OR=0.29; 95% CI=0.20-0.42) and Asian/Pacific Islanders an increased risk of ANLL (OR=2.00; 95% CI=1.19-3.36). Older maternal age (?35 years) was associated with a slight increased risk of ALL.

What was most striking, however, among the investigators’ findings was the lack of relationship with high birth weight. For ALL, there was no relationship with birth weight.  For cases diagnosed less that 2 years of age, the OR for birth weight ? 4000 grams compared to birth weight between 2500-3999 grams was 0.93 (95% CI=0.63-1.39). Similar associations were found when the investigators examined results for cases diagnosed less than 12 months of age, and for cases diagnosed at older ages. For ANLL, there was a similar lack of a relationship. Strengths of this study include the use of record linkage and the large sample size, allowing the investigators to explore associations among different ethnic groups and leukemia subtypes.

COMMENT: As we have reported previously [see C3 vol 9, No 2; Vol 10, no 3], there has been growing evidence that high birth weight (usually defined as greater than 4000 grams) is associated with an increased risk of childhood cancer, particularly leukemia. These positive findings have been noted in both record linkage studies and case-control studies. Birth weight is one of the few ‘exposure’ variables for which there is little concern regarding recall bias; most mothers can accurately recall their child’s birth weight years after the child’s birth [Olson JE et al, Am J Epidemiol 145:58-67,1997]. One criterion for establishing a cause/effect relationship in epidemiological studies is consistency among studies. The vast majority of studies have demonstrated an association between high birth weight and risk of childhood leukemia, especially for children diagnosed before the age of 2 [Ross JA et al, Cancer Causes & Control 1996; 7:272-9], leaving most epidemiologists to contend that high birth weight is a risk factor for younger age childhood leukemia. Well-conducted large studies such as this by Reynolds et al, however, challenge this belief. Only 12% of cases had an unsuccessful record linkage with birth certificates and it is unlikely that lack of these non-participants in the analysis would notably alter the results. It will be important to perform similar record linkage analyses in other states. Moreover, we will await the results of the current Children’s Oncology Group case-control study of infant leukemia. Julie A. Ross

Editors’ Note

In a previous C3 [Vol 12, No 6] we commented on a study by Thompson et al [Lancet 358:1935-40, 2001] that explored folate deficiency and childhood ALL, and we evaluated this report in the context of a recent molecular study by Prof Greaves’ group regarding the frequency of polymorphisms in folate-metabolizing genes [Vol 12, No 2] among leukemia subgroups. Prof Greaves has kindly pointed out that the report by Thompson does not necessarily conflict with the findings of his group, which reported that homozygosity for the C677T MTHFR polymorphism was linked with a decreased risk of infant leukemia, and there was a suggestion (although not statistically significant) that this relationship also held for TEL-AML1 leukemia. Larger sample sizes will be needed to confirm this association. Further, Prof Greaves noted that folate supplementation during pregnancy, by maintaining integrity of DNA, could help to diminish the initiation of breaks in utero, and thus, decrease the risk of leukemia even in later childhood.These are both excellent comments. As always, we appreciate any input from our readers.

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