A PAI-1 (SERPINE1) polymorphism predicts osteonecrosis in children with acute lymphoblastic leukemia: a report from the Children's Oncology Group

As cure rates for childhood acute lymphoblastic leukemia (ALL) increased, adverse effects of chemotherapy became an important area of study. Osteonecrosis is one of these, attributed primarily to corticosteroids (dexamethasone and prednisone) that play a vital role in treatment of ALL, as they have improved cure rates.¹  Osteonecrosis may lead to severe joint pain, limitations on physical activity, with some cases culminating in surgical intervention to restore function.²  Osteonecrosis affects up to one-third of patients treated for ALL, and host-related factors such as age more than 10 years, female sex, and white race increase risk of development.^3-5  Better definition of risk factors for osteonecrosis might permit tailoring therapy to minimize this complication.

We have shown that therapy-related adverse events (gastrointestinal, infectious, hepatic, and neurologic toxicities) can be attributed to germline polymorphisms in genes linked to pharmacodynamics of chemotherapy.⁶  In a relatively small series of patients at high risk of osteonecrosis, we found 2 inherited genetic polymorphisms, vitamin D receptor FokI start site polymorphism and thymidylate synthetase enhancer repeat, were associated with risk of osteonecrosis.⁵  To further explore genetic predictors of osteonecrosis, we herein studied 12 candidate polymorphisms potentially involved in osteonecrosis in a mature, completed study of the Children's Cancer Group (CCG1882), whose protocol resulted in a relatively high incidence of osteonecrosis in children 10 years and older.⁴ 

Approval was obtained from St Jude Children's Research Hospital institutional review board for these studies (institutional review board approval number XMP03-045 for protocol AALL03B2, Study of pharmacogenetic risk factors for avascular necrosis approved by the Children's Oncology Group). Informed consent was obtained in accordance with the Declaration of Helsinki.

A total of 980 patients were 10 years of age or older when enrolled on CCG1882; 108 developed symptomatic osteonecrosis (11.0%) diagnosed by radiographic imaging; patients were not prospectively screened with radiologic imaging.⁴  A total of 361 patients had sufficient archived DNA for inclusion in genotyping; 51 of these patients developed symptomatic osteonecrosis, an incidence not significantly different from the overall CCG1882 population (P = .188; Table 1). Patients were scheduled to receive the following total doses of steroids: during induction, prednisone (1815 mg/m²); delayed intensification, dexamethasone (standard arm, 235 mg/m²; augmented arm, 470 mg/m²); maintenance, prednisone (standard arm: males = 7000 mg/m², females = 4400 mg/m²; augmented arm: males = 6200 mg/m², females = 3600 mg/m²).⁴ 

DNA was isolated from archived diagnostic tissue samples preserved on glass slides, including bone marrow aspirates and peripheral blood smears.⁷ 

Genotyping was performed for 12 candidate polymorphisms (Table 2): TYMS enhancer repeat, VDR start site Fok1 (rs2228570⁸),⁵ MTHFR C677T (rs1801133), PAI-1 (rs6092),⁹ ESR PvuII (rs2234693), LRP5 (rs2306862), BGLAP (rs1800247), ACP5 (rs2305799 and rs2229531, respectively), ABCB1 (rs1045642), PTH (rs6254), and PTHR (rs1138518). The TYMS enhancer repeat was tested as described.⁵  With samples that failed initial testing, a second set of primers was designed to produce smaller 116 and 144 bp products (corresponding to an enhancer 2-repeat or 3-repeat variation).¹⁰  For all polymorphisms except TYMS, genotyping was performed using the Beckman Coulter GenomeLab SNPstream (Fullerton, CA) by DNAprint. Additional genotyping for the VDR start site Fok1 polymorphism for samples failing initial genotyping was performed as described.⁷  Not all patients were successfully genotyped for each polymorphism (Table 2; www.pharmgkb.org PS207425).

Univariate analysis was performed using logistic regression for each individual genotype as a predictor for osteonecrosis. Multivariate logistic regression analyses were performed for genotype, adjusting for age (10-15 years and 16+ years), sex, and treatment (rapid early response patients: with vs without prophylactic cranial radiation; slow early response patients: standard vs augmented; Table 2). Adjustment for ethnicity was not performed because of a low number of patients of nonwhite ethnicity (n = 6) affected by osteonecrosis (Table 1). Logistic regression analyses were performed to test for possible correlations among polymorphisms. Among the largest racial/ethnic group (whites), we compared whether genotype distributions varied from those expected under Hardy-Weinberg assumptions using a χ² test; only the TYMS and LRP5 polymorphisms varied from Hardy-Weinberg predictions (P < .004).

A total of 43 of 291 patients who were 10 to 15 years of age and 8 of 70 patients 16 years of age and older developed osteonecrosis (Table 1). Osteonecrosis was more common in females and whites (P = .034 and P = .035, respectively; Fisher exact test). Univariate analysis (Table 2) indicated that only the PAI-1 polymorphism was associated with osteonecrosis (P = .002), with an odds ratio of 2.79 (95% confidence interval, 1.45-5.34). A total of 26.9% of the combined GA and AA genotypes developed osteonecrosis compared with 11.7% of the GG genotypic group.

Multivariate logistic regression analysis adjusted for age, sex, and treatment arm (Table 2) did not significantly alter the results: PAI-1 was associated with osteonecrosis (P = .002 and adjusted odds ratio = 2.89). Logistic regression analysis revealed no significant correlations among genotypes.

PAI-1 rs6092 allele frequencies differed significantly between racial/ethnic groups. The combined GA and AA genotypes were present in 30.9% of whites but only 12.5% of blacks. The incidence of GA or AA genotype in Hispanics was 20%, not significantly different from the white population (P = .572).

Other than older age,^4,5,11-14  host-related risk factors for osteonecrosis remain incompletely defined. Herein, after adjusting for age, sex, and treatment arm, the only genetic risk factor for osteonecrosis was a PAI-1 polymorphism.

PAI-1 inhibits fibrinolysis. Increased serum levels have been associated with increased incidence of thrombophilia¹⁵  and osteonecrosis,^16-19  although reports are not consistent. High levels of PAI-1, induced by corticosteroid treatment,²⁰  or through polymorphisms in PAI-1, lead to suppression of fibrinolysis through inhibition of tissue plasminogen activator and promotion of thrombosis.²¹  Resulting increased intraosseous venous pressure blocking blood flow to the femoral head may culminate in hypoxic bone death or osteonecrosis.^17,19 

The PAI-1 SNP (rs6092) we found associated with osteonecrosis resides in a haplotype containing a 4G/5G repeat promoter insertion/deletion polymorphism.⁹  The 4G allele has been linked with variation in PAI-1 serum levels,^9,22,23  metabolic syndrome, coronary atherosclerosis, myocardial infarction, increased serum triglycerides,²⁴  and also with increased risk of osteonecrosis among renal transplant recipients who received glucocorticoids.²⁵  Our findings confirm an association of PAI-1 germ line variation with osteonecrosis risk in an entirely different clinical setting: children who received glucocorticoids as part of ALL chemotherapy.

We did not confirm our prior finding⁵  in a St Jude ALL cohort that polymorphisms in VDR and TYMS were associated with osteonecrosis. This discrepancy may be related to technical challenges posed by DNA quality or to differences in chemotherapy between the St Jude and CCG cohorts, in that patients at St Jude received more antimetabolites (particularly methotrexate) than patients on CCG1882. Methotrexate is associated with high plasma homocysteine and folate depletion, both of which have been linked to thrombosis and osteonecrosis.^26,27  Thus, it is plausible that lower expression of thymidylate synthetase (a target of methotrexate) associated with the TYMS polymorphism may be more relevant for patients receiving methotrexate-intensive chemotherapy regimens but perhaps not for patients receiving other ALL regimens. As in all of pharmacogenetics, the important target genes will depend on therapy.

Interestingly, the risk of osteonecrosis was higher among whites than other ethnic/racial groups in this study, as well as in the larger CCG1882 cohort and in another group ALL cohort, with incidence in the larger CCG1882 cohort being approximately 2.5-fold more common in whites than nonwhites.^4,5  Although the number of nonwhites in our cohort available for genotyping was relatively small and therefore limits our power, the magnitude of increased osteonecrosis risk associated with the PAI-1 A allele (∼2.8-fold) is similar to the increased frequency of harboring at least one A allele in whites compared with nonwhites (∼2.4-fold), suggesting that racial differences in germline genomic variations might account for differences in frequency of osteonecrosis among racial/ethnic groups.

Our findings support the hypothesis that inherited variation in PAI-1 contributes to variation in risk of constitutive and drug-induced phenotypes and to the notion that some individuals are more prone to develop serious adverse effects of ALL therapy than others.

The authors thank Nancy Anderson, Pam McGill, and Diana Chan for technical and analytical assistance, our clinical and research faculty and staff, and the patients and their families for participating.

This work was supported by National Cancer Institute (NCI) CA51001, CA78224, CA21765, the National Institutes of Health/ National Institute of General Medical Sciences (NIGMS) Pharmacogenetics Research Network and Database (U01 GM61393, U01GM61374; www.pharmgkb.org PS207424) and the Children's Oncology Group Chairman's grants CA98543 and CA98413 from the National Institutes of Health; and American Lebanese Syrian Associated Charities.

National Institutes of Health

Contribution: L.A.M., J.B.N., and M.V.R. conceived and designed the project; H.N.S. and M.D. carried out and interpreted the statistical analyses; D.F. and L.H.H. interpreted the data and were primary authors; all authors contributed to the writing of the paper.

L.A.M., H.N.S., M.D., J.B.N., and M.V.R. are study participants in the Children's Oncology Group.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Mary V. Relling, Department of Pharmaceutical Sciences, St Jude Children's Research Hospital, 332 N Lauderdale St, Memphis, TN 38105-2794; e-mail: mary.relling@stjude.org.