Infections remain a significant challenge in treatment of childhood leukemia. Inherited genetic variants may influence the immune-inflammatory response in patients with compromised immune functions and may involve multiple signalling molecules reflecting the interplay of multiple genes. Using an extended candidate gene approach including pathway-analysis and protein-protein interactions we expanded our earlier study on mannose-binding lectin (MBL) polymorphisms to include 30 000 target SNPs within 900 candidate genes with possible relevance for childhood ALL. The candidate genes covered the following domains: i) pharmacogenetics, ii) immunogenetics, iii) apoptosis, iv) neurobiology, v) toxicity, vi) thrombosis, vii) cell cycle control genes, and viii) DNA repair and mitosis. We explored differences in host genomic patterns between patients with and without an infectious event during induction therapy.
Included in our study were 72 patients aged 1.0 to 14.9 years diagnosed and treated for non-B ALL at the University Hospital Rigshospitalet, Copenhagen, Denmark, from January 1, 1992, to December 31, 2000. The 7 weeks induction period (NOPHO-ALL 92 protocol) included oral prednisone 60mg/m2/d day 1–36/45, weekly i.v. vincristine 2mg/m2 (6 doses), i.v. doxorubicin 40mg/m2 (3 doses; 4 doses for the high risk groups), l-asparaginase 30 000IE/m2 daily day 36–45, and age adjusted i.t. methotrexate (4 doses). Main endpoints in our analysis were “infectious event” and “infectious event with positive blood culture”. Additional covariates included age, sex, risk group, immunophenotype, white blood cell count, and microbiological data. DNA library preparation was performed according to the SureSelect Target Enrichment System. Illumina GAIIx Genome Analyzer was used for sequencing. Associations were analysed by Fisher's exact test implemented in PLINK. Only SNPs with observed minor allele frequency (MAF) above 5% and at least 50% of non-missing genotypes above 10 times the sequencing depth were used in the analyses. The obtained p-values were adjusted for multiple testing by means of adaptive permutation. The empirical p-values were generated by 1 million permutations of the test and were then corrected using false discovery rate (FDR) for multiple testing. A set of high confidence protein-protein complexes was found. For association of protein-protein complexes similar methods were applied requiring at least 2 SNPs with observed MAF above 1% in at least 2 different genes of the complex. CART analyses were performed using R package applying 3-fold cross-validation using the genotypes of the SNPs associated with risk of infection and positive culture together with the clinical data. Only the SNPs with permutation corrected P-values below 0.01 were used in the analysis.
Of the 69 patients with appropriate DNA quantity, 48 (70%) patients experienced an infectious event. Of these, 23 (33%) patients had at least 1 positive blood culture during an infectious event which, among others, included 4 Pseudomonas, 4 Escherichia coli and 3 Candida infections. After correcting for multiple testing by means of adaptive permutation, a total of 103 and 94 SNPs were found to be associated with infectious event and positive culture, respectively (p<0.05). The QQ plots for the single SNP association analysis followed the null distribution and showed a few SNPs above the expected distributions, suggesting true biological signals. Manhattan plots for both analyses indicate that several loci were associated to both risk of having an infection and positive culture. CART analysis demonstrated rs12666401, rs2275287, rs10841795 and rs1136410 to be highly predictive of outcome characterizing 42 of 48 (88%) infectious event patients with > 89% accuracy.
Our data indicate that host genomic profiles can predict the risk of infectious events during induction therapy in children with ALL. Such knowledge might be helpful in adapting more individualised supportive care and treatment protocols.
No relevant conflicts of interest to declare.
This icon denotes a clinically relevant abstract
Asterisk with author names denotes non-ASH members.