Background
Acute myeloid leukemia (AML) is a heterogeneous hematological malignance derived from the hemopoietic progenitors with highly diverse clinical traits, molecular pathogenesis and clinical outcomes [
1]. Apart from known prognostic factors including age, white blood cell (WBC) counts, complex karyotype, antecedent hematologic disease and secondary leukemia [
2], accumulated evidence has shown that the presence of somatic mutations in genes such as fms-like tyrosine kinase 3 (
FLT3), nucleophosmin 1 (
NPM1), CCAAT enhancer binding protein alpha (
CEBPA), KIT proto-oncogene receptor tyrosine kinase (
KIT), tumor protein p53 (
TP53) [
3] and DNA methyltransferase 3 alpha (
DNMT3a) [
4,
5] have clinical prognostic significance. Chemotherapy with cytarabine (Ara-C) based regimen remains the major therapy for AML except for the M3 subtype, and the complete remission (CR) rate varies between 60 and 70% for the adult patients during induction therapy [
6]. Half of the patients achieved CR in primarily induction chemotherapy relapsed due to the existence of minimal residual disease [
7]. Overall long-term survival rate of AML ranges from 21.9 to 44% [
8]. Primary or acquired chemoresistance is the major problem faced in AML treatment [
9]. Therefore, identification of factors related to Ara-C-resistance and better elucidation of potential mechanisms involved in Ara-C resistance will help optimize regimens for treatment of AML and improve the clinical outcomes as well.
Ara-C is a prodrug and undergoes biotransformation into the active metabolite cytarabine triphosphate (Ara-CTP) to exert its pharmacological activity; the latter can incorporate into replicating DNA and interfere with DNA synthesis, resulting in apoptosis of cells. Three kinases are required to accomplish intracellular phosphorylation of Ara-C to the formation of Ara-CTP: the rate-limiting enzyme deoxycytidine kinase (DCK), deoxycytidine monophosphate kinase and nucleoside diphosphate kinase [
10]. On the other hand, 5′-nucleotidases [
11], cytidine deaminase (CDA) [
12] and SAM domain and HD domain-containing protein 1 (SAMHD1) [
13,
14] are the major deactivating enzymes of Ara-C that act through prevention of the formation or directly increase the degradation of the active triphosphate metabolite. Alteration in the activity of enzymes involved in Ara-C metabolism may result in a change in the proportion of its active form in the cells, and thus affects both sensitivity and toxicity of Ara-C in AML patients receiving Ara-C based chemotherapy [
15]. Our previous studies have reported that polymorphisms in
DCK, nucleoside diphosphate kinase 2 (
NME2), ribonucleotide reductase catalytic subunit M2 (
RRM2), and
SAMHD1 are associated with chemosensitivity to Ara-C based therapy and disease prognosis in Chinese AML patients [
10,
16].
The Hedgehog (HH)/Glioma-associated Oncogene Homolog (GLI) signaling pathway that plays a role in chemotherapy resistance and cellular self-renewal is supposed to be a novel therapeutic target in AML [
17]. In a phase 2 clinical trial with AML and high-risk myelodysplastic syndromes (MDS) patients, combined therapy with low dose Ara-C and a selective HH/GLI pathway inhibitor glasdegib is observed to improve overall survival (OS) as compared with Ara-C treatment alone [
18]. Our recent study also showed that GLI1 expression was upregulated in bone marrow mononuclear cells from patients with refractory or relapsed AML, and GLI1 inhibition is sufficient to increase Ara-C sensitivity [
19]. Furthermore, Zahreddine and colleagues demonstrated that UDP-glucuronosyltransferase 1A subfamily (UGT1A) enzymes can inactivate Ara-C through glucuronidation in leukemia cells, and GLI1 is involved in Ara-c resistance through enhancing the stability of UGT1A enzymes [
20]. However, there is no report on associations of genetic polymorphisms in
GLI1 and the
UGT1A subfamily with Ara-C presently.
The human UGT1A subfamily enzymes are encoded by the
UGT1A gene locus on Chromosome 2q37.1 by alternative splicing. Nine functional proteins (UGT1A1, 1A3, 1A4, 1A5, 1A6, 1A7, 1A8, 1A9, and 1A10) were translated by the locus. All the isoforms share 4 common exons from exon 2 to exon 5, but have unique exon 1 and individual promoter pairs in turn of 1A8, 1A10, 1A9, 1A7, 1A6, 1A5, 1A4, 1A3, 1A1 in the
UGT1A pre-mRNA [
21]. The
UGT1A1 exon 1/promoter is the nearest one to the common exon 2. UGT1A1 plays important roles in the clearance and metabolism of many endogenous or exogenous compounds such as irinotecan [
22]. Two functional single nucleotide polymorphisms (SNPs) in
UGT1A1, i.e.
UGT1A1*28 (rs8175347) and
UGT1A1*6 (rs4148323) are reported. The
UGT1A1*28 polymorphism that results in TA repeating number alteration in the TATA box in the promoter can decrease the
UGT1A1 expression and accounts for increased risk for irinotecan-induced neutropenia [
23,
24] and myelosuppression [
25,
26]. In 2005, the USA Food and Drug Administration (FDA) warned that patients with
UGT1A1*28/*28 genotype are at increased risk for neutropenia when irinotecan is used, and a lower starting dose of irinotecan was recommended for
UGT1A1*28/*28 homozygotes [
27].
UGT1A1*6 is a missense variants (Gly71Arg) in
UGT1A1 exon 1 that leads to decreased UGT1A1 enzyme function [
28] and is also a risk factor for irinotecan toxicity.
UGT1A1*6 variant is mainly observed in the Asians with allele frequency ranges in 15–30%. In 2008, the Ministry of Health, Labour, and Welfare of Japan also warned increased risk of severe irinotecan-related neutropenia in Japanese patients carrying the
UGT1A1*6 or
*28 allele, and approved diagnostic test for
UGT1A1 genotypes [
29]. As UGT1A is involved in Ara-C detoxification, we hypothesized that the functional polymorphisms in
UGT1A1 may affect chemosensitivity to Ara-C based therapies in AML patients through influence Ara-C metabolism, which could eventually improve responses to Ara-C and AML prognosis.
In this study, we investigated the impact of UGT1A1*28 and *6 variants on CR rate after induction chemotherapy, treatment-related mortality (TRM), OS and event-free survival (EFS) in 726 Chinese AML patients treated with Ara-C.
Discussion
UGT1A is a newly identified enzyme subfamily that is involved in Ara-C detoxification. In this study, we performed an association study on UGT1A1 polymorphisms with responses to Ara-C and disease prognosis in AML patients for the first time. We observed that carriers of the UGT1A1*6 variant or carriers of any of the UGT1A1*6 and UGT1A1*28 alleles showed significantly decreased risk of non-CR after one and two courses of Ara-C based induction chemotherapy for AML patients. We also observed that carriers of the UGT1A1*6 or UGT1A1*28 alleles showed significantly better OS in AML patients.
Glucuronidation is an important way of metabolism for many small endogenous and exogenous lipophilic compounds and is mediated by UDP-glucuronosyltransferases (UGTs) located in the endoplasmic reticulum. UGT1A1 is the most abundant member of the UGT1A family in human liver and is also the major isoform responsible for the glucuronidation of bilirubin (UGT1A1 specific), SN-38 (the active metabolite of irinotecan), β-estradiol, etc. [
33,
34].
UGT1A1*6 and
*28 are two functional polymorphisms that lead to decreased glucuronidation activity.
UGT1A1*28 is characterized by an extra TA repeat (TA-7) in
UGT1A1 promoter that decreases the gene transcription, and SN-38 glucuronidating activity was decreased by approximately 25 and 50%, respectively, in liver microsomes from
UGT1A1*28 heterozygotes and homozygotes [
35]. Similarly,
UGT1A1*6 is a missense variant that results in decreased UGT1A1 activity by about 50% as indicated by bilirubin and SN-38 glucuronidation [
36,
37], and the variant is primarily observed in the Asians. Clinical studies have shown that the two polymorphisms were associated with increased risk of Gilbert’s syndrome or irinotecan-reduced toxicity in Caucasians and Asians [
38].
A previous study by Zahreddine reported for the first time that Ara-C is a substrate of UGT1A enzymes and could be inactivated through glucuronidation [
20]. Increased UGT1A expression in Ara-C resistant M5 AML THP-1 cells and relapsed AML after standard Ara-C therapies were observed, and elevation in GLI1 expression is sufficient to drive UGT1A dependent glucuronidation of Ara-C and drug resistance [
20]. However, by analyzing the mRNA expression profile of the Cancer Genome Atlas (TCGA) dataset, we observed that the expression of
UGT1A1 in blast cells from AML patients was nearly negligible in contrast to the main Ara-C inactivating enzyme
CDA (Additional file
5: Fig. S3) [
39]. These findings suggest that the observed influence of
UGT1A1*28 and
*6 polymorphisms on Ara-C response in AML patients is less likely to be explained by decreased Ara-C detoxification in AML blast cells. As the UGT1A subfamily is mainly expressed in human liver, we speculate that the
UGT1A1*28 and
*6 polymorphisms may improve Ara-C response and OS of the AML patients through decreasing hepatic glucuronidation and increasing the systemic exposure of Ara-C. It’s a pity that we failed to detect the plasma concentrations of Ara-C during Ara-C infusion. Of note, AML is usually treated by combined therapy with Ara-C and anthracyclines, and whether UGT1A1 is involved in the metabolism of anthracyclines remains unknown. Therefore, we could not rule out the possibility that difference in disease outcomes among
UGT1A1 genotypes is due to difference in anthracycline metabolism. Influence of both
UGT1A1 polymorphisms on pharmacokinetics and systemic exposure of Ara-C and glucuronidated Ara-C (AraC-Glu) deserved further study in future.
We noticed that the predictive value of
UGT1A1*28 and
*6 polymorphisms on CR after Ara-C based induction therapy is modest in AML patients (CR rate 76.9% in carriers of the
UGT1A1*28 and
*6 alleles and 63.8% in
UGT1A1*1/*1 for both loci), and the association of
UGT1A1*28 alone with non-CR risk was nonsignificant after Bonferroni correction for two SNPs (significance set at
P < 0.05/2 for 2 SNPs). This may be explained by two reasons: firstly, the lower allele frequency of the
UGT1A1*28 polymorphism in our patient cohort. Secondly, the multigenic characteristic of drug response including Ara-C. In our previous studies, we observed that polymorphisms in other genes encoding enzymes in the Ara-C metabolic pathway such as
DCK,
NME2 (DNPK-B),
RRM2, and
SAMHD1 are also associated drug response to Ara-C based therapies in AML [
10,
16]. We suggest that the construction of a decision tree based on multiple genetic variations concomitantly may increase the predictive values of pharmacogenetics biomarkers in AML induction therapy. Of course, the exact usefulness requires to be explored with a large sample size.
Regarding disease prognosis including OS and EFS, we observed better OS in carriers of UGT1A1*28, or carriers of at least one of the UGT1A1*6 and *28 alleles. Neither polymorphisms nor the combined genotypes were associated with EFS in our study. As EFS was considered only in patients achieved CR after induction therapies, lack of association between UGT1A1 polymorphisms and EFS observed in our study suggests other mechanisms other than chemosensitivity might also play a role in AML prognosis.
Authors’ contributions
PC and K-WZ collected the clinical samples, genotyped, performed the follow-up, analyzed the data and prepared the manuscript. D-YZ, HY, HL, and Y-LL collected and characterized the data. SC, GZ, HZ, S-PC, X-LZ and JY contributed to patients’ recruitment, X-PC designed the study and revised manuscript. All authors read and approved the final manuscript.