Association of Inherited Genetic Factors With Drug-Induced Hepatic Damage Among Children With Acute Lymphoblastic Leukemia

Key Points Question What are the genetic drivers of interpatient variability in the occurrence of hepatotoxic effects during treatment for acute lymphoblastic leukemia (ALL)? Findings In this genetic association study of 3557 children, adolescents, and young adults receiving ALL therapy, variants in UGT1A1 and PNPLA3 were associated with hyperbilirubinemia and elevated alanine aminotransferase and aspartate aminotransferase levels, respectively. A polygenic risk score–based analysis demonstrated that the UGT1A1 variant was the primary driver of elevated bilirubin levels, while other genetic variants contributed to aminotransferase levels even after adjusting for PNPLA3. Meaning These findings suggest that there is an association between genetic factors and interpatient variability in the occurrence of hepatotoxic effects during treatment for ALL.


Introduction
Childhood acute lymphoblastic leukemia (ALL) is the most common pediatric cancer.Cure rates now exceed 90%, 1 and decreasing treatment-related toxic effects and improving patients' quality of life have become increasingly important to further improve overall patient outcomes.ALL therapy is prolonged for up to 3 years and is characterized by several phases of intensive chemotherapy. 2tients can experience a plethora of adverse effects, including hepatotoxic effects, as evidenced by hyperbilirubinemia and elevated levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST).Hepatotoxic effects can be caused by chemotherapies such as methotrexate, asparaginase, and mercaptopurine, but can also be caused by infections, blood transfusions, and drugs used in supportive care.Even though these toxic effects usually resolve after withholding therapy, their long-term implications remain unclear.Uncovering genetic and nongenetic factors associated with elevated bilirubin, ALT, and AST levels can help identify patients who are likely to develop severe hepatotoxic effects and facilitate treatment individualization.
Elucidating the underlying biological mechanisms could provide insights into the development of rational interventions.
Previously, using a candidate gene approach, Kishi et al 3 identified the UGT1A1 promoter TA repeat variant (rs8175347) as associated with hyperbilirubinemia during all treatment phases in a singleinstitution study at St Jude Children's Research Hospital (Total XIIIB protocol).In a 2017 genome-wide analysis using modern Total XV and Total XVI study cohorts, 4 the PNPLA3 I148M variant (rs738409) was shown to be associated with elevated ALT levels after induction therapy.6][7] These findings suggest that genetic factors that influence healthy liver function can also have effects on drug-induced changes in liver function in patients with ALL.
Pharmacogenomics studies often are limited by the number of cases of the drug-induced phenotype of interest and therefore underpowered.One possible remedial approach is, when a phenotype is shared between a drug effect and a disease, prioritizing known disease-associated variants may help overcome this limitation.Herein we demonstrate the feasibility of this approach.
The purpose of this study was to replicate previously identified genetic factors and discover novel genetic factors associated with hepatotoxic effects during pediatric ALL treatment.We first identified candidate variants by mining an available public large-scale genome-wide association study (GWAS) repository 8 and tested their associations with hepatotoxic effects in 2 large multiinstitutional Children's Oncology Group (COG) trials, AALL0232 (n = 2283) 9 and AALL0434 (n = 1274). 10Subsequently, we undertook a complementary gene-agnostic approach by performing a GWAS in our cohorts to discover novel genetic variant associations.

Patients
Patients with newly diagnosed ALL were enrolled in the COG AALL0232 9  with toxic effects of grade 3 or above were considered as cases.Hyperbilirubinemia of grade 3 or above was defined as the bilirubin level above 3.0 times the upper limit of normal.Levels of ALT or AST of grade 3 or above were defined as greater than 5.0 times the upper limit of normal.
Toxic effects were reported within treatment phases (eTable 1 in Supplement 1), including induction, consolidation, interim maintenance, delayed intensification, and maintenance.For those patients who did not have toxic effects reported during the phase, if the patient's last follow-up time was later than the day that at least 80% toxicity events were reported during the treatment phase, then they were assumed to not have had any toxicity events for the whole phase and were considered a control; otherwise, the patient was treated as missing a value for the phase and subsequent treatment phases.Only the first delayed intensification or interim maintenance phase was considered.Total duration of multiagent protocol chemotherapy ranged from 2.5 to 3.5 years.

Genotyping
At the end of the induction phase, DNA were extracted and genotyped using a mapping array (Affymetrix SNP6; Thermo Fisher Scientific) for the AALL0232 cohort and an exome array (Infinium 2.5 Omni-exome array; Illumina) for the AALL0434 cohort. 13,14Variants were filtered by Hardy-Weinberg equilibrium (P > .0001)and a call rate of greater than 95%; variants with minor allele frequency of at least 3% were used for association analysis.Imputation was performed using a National Institutes of Health TOPMed Imputation server.Genetic ancestry was included as a covariate in the genetic association analysis to reduce false discovery attributable to population stratification.Genetic ancestry assignments are provided in the eMethods in Supplement 1.

Statistical Analysis
Both hyperbilirubinemia and elevated ALT and AST levels were treated as repeated binary phenotypes.Incidences of toxic effects were estimated using a Kaplan-Meier curve with the first

JAMA Network Open | Oncology
Association of Genetic Factors With Drug-Induced Hepatic Damage in Pediatric ALL event as time of onset.Patients with relapse or other outcome events were treated as censored at time of events.To account for subgroup effects between the 2 treatment protocol cohorts, models were fitted in the 2 cohorts separately for all genetic association analyses, and the statistics were combined in a meta-analysis using the Stouffer method. 17Details of statistical testing techniques and controls for multiple testing are described in the eMethods in Supplement 1. P < .05indicated statistical significance; all P values are 2 sided unless indicated otherwise.

Incidence of Hepatotoxic Effects and Clinical Risk Factors
The cumulative incidence of hyperbilirubinemia within the first 2 years of therapy was 11.4% (range, 10.4%-12.4%) in the AALL0232 and 10.1% (range, 8.6%-11.4%) in AALL0434 cohorts (eFigure 1 in Supplement 1).The percentage of patients who experienced hyperbilirubinemia was the highest during maintenance therapy in both cohorts (157 of 1757 [9.0%] in AALL0232 and 98 of 749 [13.1%]   in AALL0434) (Figure 1 and eFigure 2 in Supplement 1).Because transfusions and severe infections are rarer during maintenance therapy, this finding suggests that these elevations are more likely chemotherapy related.
The cumulative incidence of elevated ALT and AST levels within the first 2 years of therapy was estimated to be 48.4% (range, 46.8%-49.9%) in the AALL0232 cohort and 45.7% (range, 43.3%-47.9%) in the AALL0434 cohort (eFigure 1 in Supplement 1).The percentage of patients experiencing elevated ALT and AST levels was the highest during maintenance therapy in both cohorts (689 of 1785 [38.6%] in the AALL0232 cohort and 289 of 770 [37.5%] in the AALL0434 cohort) (Figure 1).Consolidation (374 of 2187 [17.1%] in the AAL0232 protocol and 186 of 973 [19.1%]   in the AALL0434 protocol) and interim-maintenance (374 of 2044 [18.3%] in the AAL0232 protocol and 158 of 889 [17.8%] in the AALL0434 protocol) phases were the next 2 treatment phases with the highest ALT and AST levels.

Association Between Hepatotoxic Effects and Known Genetic Variants
Between reported variants associated with hepatic function from the GWAS catalog and CPIC level A/B genes, we tested 698 variants.Hyperbilirubinemia was associated with UGT1A1 rs887829 after JAMA Network Open | Oncology adjusting for multiple testing (OR, 2.18 [95% CI, 1.89-2.53];meta-analysis P = 6.7 × 10 −27 ; adjusted P < .002)(Table 2) based on a mixed-effects logistic model adjusting for genetic ancestry and age.Variant rs887829 is in high linkage disequilibrium (LD) with the UGT1A1 promoter (TA)n variant (*28; rs8175347) across populations.The rs887829-T allele, which is in LD with (TA)7/8, was associated with higher risk of hyperbilirubinemia in the AALL0232 cohort (OR, 2.18 [95% CI, 1.83-2.59];P = 3. 1 × 10 −18 ) and the AALL0434 cohort (OR, 2.20 [95% CI, 1.67-2.89];P = 1.43 × 10 −8 ).No significant interaction between rs887829 and treatment phase was observed (adjusted P = .25).Patients with the rs887829-T allele were more likely to have hyperbilirubinemia across all phases in both cohorts (Figure 2A).No other candidate variants tested were statistically significant after multiple testing adjustment for either the main effect or genotype × treatment phase interaction.PNPLA3 rs738409 (I148M) was associated with elevated ALT and AST levels after adjusting for multiple testing (OR, 1.27 [95% CI, 1.15-1.40];meta-analysis P = 3.72 × 10 −7 ; adjusted P < .002).The G allele of rs738409 was associated with higher ALT and AST levels in both the AALL0232 (OR, 1.26; [95% CI, 1.12-1.41];P = .0001)and AALL0434 (OR, 1.29 [95% CI, 1.07-1.55];P = .007)cohorts.No interaction between PNPLA3 genotype and treatment phase was observed (adjusted P = .78)(Figure 2B).A significant interaction between TPMT rs1142345 (Y240C) and treatment phase was observed (meta-analysis P = 2.38 × 10 −5 ; adjusted P = .02).The C allele represents TPMT *3C and was only associated with lower incidence of elevated ALT and AST levels during maintenance phase   b Indicates admixed ancestry category for which the genetic ancestry did not fit into the specified categories.The method to determine ancestry and cutoffs is described in the eMethods in Supplement 1.Because the UK Biobank data set is composed primarily of participants of European ancestry, we also performed a PRS analysis in only patients of European ancestry, and the above associations remained, with similar effect size estimates (eTables 3 and 4 in Supplement 1).

Genome-Wide Association Analysis of Hepatotoxic Effects
To discover novel genetic associations with hepatotoxic effects, we performed genome-wide analyses within each cohort adjusting for age, treatment group, and genetic ancestry, and then performed meta-analysis using the Stouffer method. 17UGT1A1 was the only locus reaching genomewide significance for hyperbilirubinemia (eFigure 4 in Supplement 1), with the top variant being rs887829.We also performed conditional analysis adjusting for UGT1A1 rs887829, and the locus zoom plot showed it was the only significant signal at this locus (eFigure 5 in Supplement 1).Because age was associated with hyperbilirubinemia, we performed genome-wide analysis within subcohorts of patients younger than 10 years or 10 years and older (Figure 3A).UGT1A1 rs887829 continued to be significant in the age-stratified analysis for patients younger than 10 years (OR, 1.10 [95% CI, 1.06-1.14];meta-analysis P = 3.64 × 10 −7 ) vs patients 10 years or older (OR, 1.27 [95% CI, 1.20-1.34];meta-analysis P = 4.4 × 10 −19 ).No additional variants reached genome-wide significance in the age-stratified analysis.

Discussion
In this genome-wide association study, hepatic toxic effects were common in both pediatric B-cell lineage (AALL0232) and T-cell lineage (AALL0434) ALL, and over 40% of patients experienced elevated ALT and AST levels over the first 2 years of treatment.Age was associated with hyperbilirubinemia in both cohorts and with elevated ALT and AST levels in the AALL0232 cohort.
The incidence of both hepatotoxic effects also differed among ancestry groups, suggesting potential underlying genetic factors.By leveraging the power from existing common disease GWAS, we were able to apply a less stringent significance threshold and improve our power for discovering biologically relevant variants that were associated with phenotypes, despite our limited sample size.
For example, the PNPLA3 rs738409 variant was not significant genome-wide in our cohort (eFigure 7 in Supplement 1), but it could be discovered even after multiple testing adjustments by focusing on candidate SNVs from the GWAS catalog of relevant phenotypes.This replicated the association between rs738409 and hepatotoxic effects reported in a new independent cohort. 4Our findings indicate the benefit of leveraging common disease findings in future genetic association discovery studies of drug toxicities in ALL.
We observed a genome-wide significant SNV that was associated with elevated ALT and AST levels in younger patients.The SNV rs12283870 is located 89 kilobase (kb) downstream of IGHMBP2 and 148.6 kb upstream of the CPT1A gene.From GTEx, version 8, rs12283870 is an expression quantitative trait locus for IGHMBP2 and a splice quantitative trait locus for CPT1A in multiple tissues (eFigure 8 in Supplement 1).CPT1A encodes carnitine palmitoyltransferase 1, a key enzyme in transferring the acyl group of long-chain fatty acid-coenzyme A conjugates onto carnitine.9][20] However, in a mouse model, it did not reduce asparaginase-associated hepatotoxic effects. 21IGHMBP2 was initially discovered to bind to the immunoglobulin mu chain switch (S mu) region, 22 but its function related to IgM has not been well established.There is conflicting evidence whether IGHMBP2 is associated with IgA nephropathy. 23,24Polyethylene glycol conjugated (PEGylated) preparations of asparaginase (pegasparagase) were used in both AALL0232 and AALL0434, 25 and presence of anti-PEG IgM has been shown to accelerate blood clearance of the PEGylated products. 26,27It is possible that rs12283870 influences hepatic toxicity by modifying exposure to asparaginase.
By interrogating the interaction between genotypes and treatment phase, the likely drugs implicated with these genetic variants may be identified.We showed that the TPMT genotype was only associated with elevated ALT and AST levels during the maintenance phase, a period when mercaptopurine exposure is continuous for more than 1 year.TPMT intermediate and poor metabolizers were reported to have lower ALT levels, which could be due to lower concentrations of hepatotoxic methylated thiopurine metabolites. 28,29tients of African ancestry had a lower incidence of hyperbilirubinemia, even though they had a higher risk allele frequency for UGT1A1 rs887829 and a lower frequency of the PNPLA3 risk allele; thus, we were unable to identify a genetic basis for their lower hyperbilirubinemia.Hispanic patients had a higher frequency of the PNPLA3 rs738409 risk allele, which likely contributes to their higher ALT and AST levels.Furthermore, in the multivariable analysis with both ancestry and genotype, Hispanic patients continued to a have higher risk of elevated ALT and AST levels in the AALL0232 cohort (eTable 5 in Supplement 1).These findings suggest a genetic basis for racial and ethnic differences in hepatotoxic effects and further contribute to a growing body of evidence for racial and ethnic disparities in childhood ALL. 30

Strengths and Limitations
By using both a candidate gene and genome-wide approach in an ancestrally diverse study population, we validated previous associations found in other cohorts 3,4 and identified novel loci associated with hepatotoxic effects during pediatric ALL treatment.Our novel discovery of an association of rs12283870 on chromosome 11 with elevated ALT and AST levels in younger patients needs to be replicated in additional cohorts.We also discovered a new interaction between the maintenance phase and TPMT *3C protective effect on ALT and AST levels that has biological plausibility.
This study also has some limitations.Because patients were enrolled in prospective interventional clinical trials that use contemporary multimodal ALL therapy, definitive analysis of individual drug effects was not possible, and therefore treatment phase was used as a surrogate for differential drug effects.As ALL therapy evolves, future studies may elucidate drug-specific effects.
Although we found that genetic risk factors for elevated bilirubin, ALT, and AST levels were also risk factors for treatment-associated elevations in these biomarkers, conflicting findings were reported for nonalcoholic fatty liver disease-associated SNVs and drug-induced liver injury. 31The PRS used in our study was derived from older adults and therefore may have performance limitations; however, it was still associated with elevated bilirubin, ALT, and AST levels in this pediatric cohort.Finally, we did not have access to nonalcoholic fatty liver disease data that may partially explain the association between ALT and AST levels and PNPLA3 rs738409.

Conclusions
The findings of this genetic association study suggest that UGT1A1 and PNPLA3 were associated with increased risk of hyperbilirubinemia and elevated ALT and AST levels in independent multinational and ancestrally diverse cohorts treated for ALL.We also identified novel associations and interactions among age, treatment phase, and genotype and found that disease-risk alleles can inform candidate SNVs for treatment-related phenotypes.Our results suggest a strong genetic basis for interpatient variability in bilirubin, ALT, and AST levels during ALL therapy.
and AALL043410 protocols, which accrued data fromDecember 29, 2003, to January 21, 2011 (AALL0232), and from JAMA Network Open | OncologyJanuary 22, 2007, to July 24, 2014 (AALL0434), and both of which used augmented Berlin-Frankfurt-Münster treatment regiments.11,12Written informed consent from the parents or guardians and assent from the patients, as appropriate, were obtained with oversight by the local institutional review boards, all of which approved the study.This study followed the Strengthening the Reporting of Genetic Association Studies (STREGA) reporting guideline, an extension of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.Data were analyzed between January 1, 2019, and April 15, 2022.Toxic effects data were retrieved from protocol databases on May 29, 2019.Toxic effects were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events, version 4. Only patients

Figure 1 .
Figure 1.Proportion of Patients With Hepatotoxic Effects by Protocol and Treatment Phases

a
Odds ratio and P values were based on mixed-effects logistic regression.Numbers of patients used in the mixed-effects model are shown.

Figure 3 .
Figure 3. Genome-Wide Association Study Analyses of Hyperbilirubinemia and Elevated Alanine Aminotransferase (ALT) and Aspartate Aminotransferase (AST) Levels Stratified by Age
a b Allele frequencies are based on 1000 Genomes Project.Ancestry includes admixed American and Latino, African, East Asian, European, South Asian.cOdds ratio is associated with an additional copy of effect allele.dMeta-analysis P values were generated using the Stouffer method. 17 Computed based on permutation test (n = 500), considering 617 SNVs were tested.fIncludes patients younger than 10 years only.JAMA Network Open | Oncology

SUPPLEMENT 2. Data Sharing Statement
1. eMethods.Data Analysis eTable 1. Treatment Phase and Protocol eTable 2. Association Between GWAS Catalog or CPIC Level A/B Variants With Hyperbilirubinemia or Elevated ALT and AST eTable 3. Top PRS From PRS Catalog eTable 4. Association Between Liver Function-Related PRS With Hyperbilirubinemia or Elevated ALT and AST in Combined Cohort or European Ancestry-Only Cohort eTable 5. Multivariable Analysis Including Both Ancestral Group and Genotypes eFigure 1. Cumulative Incidence of Hepatotoxic Effects in AALL0232 and AALL0434 eFigure 2. Proportion With Toxic Effects by Age eFigure 3. Forest Plot of the Association Between Top PRS With Hyperbilirubinemia (A) and Elevated ALT and AST (B) eFigure 4. GWAS of Hyperbilirubinemia and Elevated ALT and AST eFigure 5. Locus Zoom for UGT1A1 and Hyperbilirubinemia Based on Results From Meta-analysis Combining Both AALL0232 and AALL0434 eFigure 6. Association of rs12283870 Genotypes and Elevated ALT and AST by Age and Treatment Phase eFigure 7. Locus Zoom for PNPLA3 and Elevated ALT and AST Based on Results From Meta-analysis Combining Both AALL0232 and AALL0434 eFigure 8. Locus Zoom for rs12283870 and Elevated ALT and AST (Age < 10 y) Based on Results From Metaanalysis Combining Both AALL0232 and AALL0434 eReferences