IKZF1 Gene in Childhood B-cell Precursor Acute Lymphoblastic Leukemia: Interplay between Genetic Susceptibility and Somatic Abnormalities

Upon the advent of high-resolution genome-wide analysis, focal deletions, amplifications, point mutations, and rearrangements in genes encoding normal B-cell development regulators were identified in approximately 40% of B-cell precursor acute lymphoblastic leukemia (BCP-ALL) patients (1, 2). Aberrant copy number abnormalities (CNA) located in the IKZF1 gene have been identified in approximately 15% of BCP-ALL cases; IKZF1 deletions were consistently associated with two high-risk subtypes, BCR-ABL1 (3) and BCR-ABL1-like cases (4). These aberrations are somatic mutations demonstrated by their absence in remission samples.

On the other hand, common germline allelic variants in IKZF1 (7p12.2), ARID5B (10q21.2), and CDKN2A (9p21) have been repeatedly and significantly associated with childhood ALL risk (5–7). Interestingly IKZF1 rs11978267 and rs4132601 are among the most replicable SNPs to be associated with an increased risk of childhood BCP-ALL (5, 6). Although the IKZF1 somatic alterations result in dominant negative activity and/or loss-of-function mutations accelerating the onset of BCP-ALL in murine models (8), the functional significance of the IKZF1 SNP is not fully elucidated. The mRNA expression was evaluated in transformed Epstein–Barr virus lymphocytes, and significantly lower dose-dependent expression was found with each copy of the variant IKZF1 allele (5).

The IKZF1 deletions frequently cooccur with other CNAs in BCP-ALL (e.g., CDKN2A/B, PAX5, pseudoautosomal region 1 (PAR1) region, and BTG1 deletions; refs. 9, 10), and these frequencies may vary depending on the types of IKZF1 deletion (e.g., intragenic or complete gene deletion; ref. 11). The latter, BTG1 (B-cell translocation gene 1), is a transcriptional coregulator that promotes B-cell differentiation and has been shown to contribute to leukemia development (10, 12). In this context, we aimed to investigate a possible association between inherited genetic susceptibility (IKZF1 rs11978267 and rs4132601), IKZF1 somatic deletions, and other cooperative somatic deletions in BCP-ALL.

Samples from 743 children (276 BCP-ALL and 467 controls) were ascertained from 2004 to 2011 and included in this study. The eligibility criteria for BCP-ALL diagnosis and case selection have been previously described (10). Patients were 18 years old or younger at the time of diagnosis. The diagnosis and classification were performed through several methods, including morphologic, multiparametric flow cytometry, and molecular-cytogenetics analyses according to standard criteria recommended by the World Health Organization (13). First, the case–control study included 246 BCP-ALL cases without any further selection. On the basis of the data obtained in the initial analysis regarding IKZF1 and BTG1 alterations, we also included an additional series of 31 cases that harbored BTG1 deletions, aiming to analyze the risk association between genetic variants and acquisition of this specific type of abnormality. These supplementary BTG1-deleted cases were ascertained from 2012 to 2014 following the same eligibility criteria of the initial BCP-ALL series of cases.

The controls, which consisted of healthy children enrolled in the Brazilian Collaborative Study Group of Infant Acute Leukemia (BCSGIAL), were genotyped for rs11978267 in a previous study (14). BCSGIAL is a study group that investigates the acute leukemia risk factors linked to maternal environmental exposures, immunologic factors, and genetic susceptibility (15–17). Cases and controls were age-matched and from the same Brazilian regions.

Data collection and laboratory procedures were evaluated and approved by the Ethics Committee of all participating hospitals. Data analysis was approved by the Ethics and Scientific Committee of Instituto Nacional de Câncer (Rio de Janeiro, Brazil; #33243214.7.0000.5274; #33709814.7.0000.5274).

For cases, DNA isolation was performed either using peripheral blood cells from remission, minimal residual disease, or diagnostic samples, whereas for controls, DNA was isolated from buccal cells. DNA extraction was performed using the QIAamp DNA Blood Mini Kit (Qiagen) or with Oragene DNA technology (Genotek), respectively. Genotyping of IKZF1 rs11978267 and rs4132601 for cases and controls was conducted by TaqMan allelic discrimination assays: C_199413_10 and C_26019772_10, respectively (Applied Biosystems). Genotypes were defined upon allelic discrimination charts in which the germline, heterozygous or mutant patterns were identified by comparison with reference controls for each allele. To ensure quality and reproducibility of the method, 10% of random samples were analyzed in duplicates and concordance was absolute.

The genomic regions of deletion breakpoints within IKZF1 were described previously (18). On the basis of those breakpoint sequences, we retrieved a list of SNPs surrounding 100 bp of each deletion breakpoint using the UCSC Genome Browser. We analyzed data from LDlink 2.0 to compare the linkage disequilibrium between IKZF1 rs11978267 or rs4132601 and the remaining SNPs (19).

Genomic DNA was isolated from bone marrow or peripheral blood cells (blast percentage > 30%) of samples at patient diagnosis using the QIAamp DNA Blood Mini Kit (Qiagen). The presence of CNAs in IKZF1 and other genes, such as CDKN2A/B, PAX5, EBF1, BTG1, and pseudoautosomal region 1 (PAR1), was analyzed by multiplex ligation-dependent probe amplification (SALSA MLPA P335, MRC Holland). Data were analyzed using Gene Marker software v2.2.0. The types of IKZF1 deletions were confirmed using either long distance multiplex PCR, SALSA MLPA P202, or in-house designed MLPA, as described previously (10, 11, 18). IKZF1 deletions affecting exons 4–7 and exons 4–5 were classified as “dominant-negative”, whereas deletions of exons 1 or 2 were classified as “haploinsufficient”. The CNA groups were defined as “IKZF1 plus” (IKZF1 deletions plus CDKN2A/B, PAX5, and/or PAR1 deletions) and “9p deletions” (concurrent deletions of CDKN2A/B and PAX5).

Total RNA was extracted from bone marrow samples at the time of diagnosis with TRIzol reagent (Life Technologies). After DNase treatment, 2 μg of purified RNA was used to synthesize the cDNA by SuperScript II Reverse Transcriptase (Life Technologies). The oligonucleotides sequences for IKZF1 are described elsewhere (20). PCR reactions were performed in technical replicates using the BRYT GoTaq qPCR Master Mix (Promega BioSciences, LLC) and the Rotor Gene Q 2plex HRM Platform (QIAGEN). To determine the relative quantification of gene expression, the average quantification cycle (C_q) value of the endogenous control gene (GAPDH) was subtracted from the average experimental gene C_q values (ΔC_q). Next, gene expression data were presented as 2^−ΔCq, according to previous recommendation (21). Expression level was considered altered when augmented or diminished ≥2-fold compared with the reference group (22).

The expected SNP frequency was calculated using the Hardy–Weinberg law based on the allele frequency in the control group. To compare the distribution of genotypes between cases and controls, the χ² test (two-sided) was used (or Fisher exact test when expected values were less than five). To avoid calculating multiple comparisons, we selected the best genetic model for this case–control study based on a genetic model selection strategy (23). Both IKZF1 rs11978267 and rs4132601 best fitted the recessive genetic model for analyzing their relevance in BCP-ALL susceptibility. P < 0.05 was considered statistically significant. The disease risk associated with SNPs occurrence across overall or subgroups of patients was determined by calculating ORs with 95% confidence interval (CI). All statistical analyses were performed using the Statistical Product and Services Solutions statistical package, version 18.0 (SPSS Inc.).

The gene expression analysis was performed with GraphPad Prism 5 software. First, the Kolmogorov–Smirnov normality test was done for definition of appropriate statistical analyses. Then, expression data were compared according to genotype and demographic variables with the two-tailed Mann–Whitney test.

The distribution of demographic and genetic variant data of controls (n = 467) and cases (n = 246) is shown in Supplementary Table S1. Control genotypes for both SNP loci were in Hardy–Weinberg equilibrium (P > 0.05). Among cases, there was a predominance of boys (54.1%), 1 to 9 years old at diagnosis (73.6%), and the most frequent subtype was CD10⁺ ALL (86%). The IKZF1 somatic status was established for cases, being 196 wild-type and 50 with deletions (Supplementary Table S1).

First, the disease-associated risk has been evaluated (Supplementary Table S2). There was no risk association of developing BCP-ALL in patients with the variant allele of IKZF1 rs11978267. However, there was a significant increased risk for patients with the homozygous variant allele of IKZF1 rs4132601 for the codominant and recessive models (OR, 1.98; 95% CI, 1.13–3.48 and OR, 1.94; 95% CI, 1.13–3.34, respectively). As explained in the Materials and Methods section, both IKZF1 rs11978267 and rs4132601 best fitted the recessive genetic model for analyzing their relevance in BCP-ALL susceptibility and, therefore, all subsequent analyses were performed on the basis of this genetic model.

The risk associations between genetic variants and BCP-ALL were further stratified by age range, as displayed in Table 1. Although the age at diagnosis was not associated with any disease risk for the variant allele of IKZF1 rs11978267, the variant allele of IKZF1 rs4132601 conferred a two times risk of developing BCP-ALL between 1 and 9 years old (OR, 2.09; 95% CI, 1.16–3.74).

To evaluate whether such risk associations could be explained by IKZF1 expression, we quantified IKZF1 transcripts and compared them with genotypes. Although the relationship between genotype and phenotype indicated that IKZF1 expression might be influenced by its SNPs in a recessive pattern, our analyses did not show statistical differences between IKZF1 SNPs and IKZF1 transcript levels (Supplementary Fig. S1A–S1D).

IKZF1 expression was also tested by age strata and CD10 status. Infants presented 14- and 24-fold higher expression compared with children aged between 1 and 9 years (P = 0.009) and those older than 10 years (P = 0.007), respectively. Among patients with wild-type IKZF1, the same pattern of higher expression in younger children was observed (P < 0.05; Supplementary Fig. S1E). The results demonstrated that IKZF1 transcript level was lower among samples with CD10⁺ in the analysis that included all patients (P = 0.012; Supplementary Fig. S1F).

The risk associations between controls and BCP-ALL cases further stratified by IKZF1 somatic status are shown in Fig. 1. The homozygous variant of IKZF1 rs11978267 significantly increased the risk for somatic IKZF1 deletion occurrence (OR, 2.80; 95% CI, 1.25–6.23). We further tested whether this risk allele was associated to a specific phenotypic consequence of IKZF1 deletion (i.e., expression of dominant-negative isoforms or haploinsufficiency). The results show that individuals with IKZF1 rs11978267 variant genotype had a higher risk of developing BCP-ALL with IKZF1 haploinsufficiency in the blast cell (OR, 2.79; 95% CI, 1.14–6.78). A similar result was observed for individuals with IKZF1 rs4132601 variant allele, who had a significantly increased risk for IKZF1 deletion occurrence (OR, 2.88; 95% CI, 1.24–6.69). Higher odds of IKZF1 haploinsufficiency were observed for these individuals with IKZF1 rs4132601 variant allele (OR, 2.75; 95% CI, 1.06–7.08; Fig. 1). Regardless of IKZF1 somatic status, the gene expression was similar among all groups (Supplementary Fig. S1G). Then, we tested whether IKZF1 rs11978267 or rs4132601 were in linkage disequilibrium with SNPs that change recombination signal sequences (RSS) surrounding deletion breakpoints (Supplementary Table S3). Although 27 SNPs were located within breakpoint hotspots of IKZF1, 12 occurred at CAC/GTG sites. These SNPs were in linkage equilibrium with rs11978267 and rs4132601, and their minor allele frequency was lower than 0.01%.

We further tested the effect of cumulative variant alleles of both IKZF1 SNPs in the risk susceptibility to BCP-ALL. Children with the variant genotype for both IKZF1 rs11978267 and rs4132601 had significant higher odds of harboring an IKZF1 deletion (OR, 3.73; 95% CI, 1.49–9.33; Supplementary Table S4). Interestingly, when analyzing the cumulative risk associations using a case–case approach (comparing IKZF1 wild-type vs. deleted cases), individuals with both variant genotypes had increased risk to develop disease with IKZF1 deletion (OR, 2.71; 95% CI, 1.01–7.33; data not shown).

Because the somatic IKZF1 deletions frequently occur concomitantly with abnormalities in other genes in BCP-ALL, we also aimed to evaluate whether these risk alleles were associated to a specific combination of genetic alterations (Table 2). The disease risk was increased for individuals categorized as “IKZF1 plus” with the IKZF1 rs11978267 homozygous variant genotype (OR, 2.94; 95% CI, 1.13–7.63). The chances of harboring a combination of IKZF1 and 9p deletions were higher for individuals with the variant allele of IKZF1 rs11978267 (OR, 3.64; 95% CI, 1.14–11.66) and for those with the variant allele of IKZF1 rs4132601 (OR, 4.32; 95% CI, 1.34–13.95).

Finally, we observed that patients with the variant allele of IKZF1 rs4132601 had more than 7 times the risk of developing disease with both IKZF1 and BTG1 deletions; however, the number of cases with BTG1 deletions was very small precluding the opportunity to reach a reliable statistical result (OR, 7.55; 95% CI, 1.33–42.96, recessive model, Table 2). To further explore the magnitude of this risk association, we analyzed an extended cohort consisting of additional 31 cases that harbored BTG1 deletions (Supplementary Table S1). The results presented in Table 2 show a remarkable risk association for individuals harboring both IKZF1 and BTG1 deletions with either the variant allele of IKZF1 rs11978267 (OR, 4.90; 95% CI, 1.65–14.55) or rs4132601 (OR, 5.80; 95% CI, 1.94–17.41).

Childhood ALL is a malignancy characterized by the detection of primary genetic abnormalities in more than 75% of cases (24). The understanding of how such changes emerge is still poor. Therefore, despite the fact that ALL cases have long been very well characterized from a genetic point of view, it is still a challenge to associate the risk of childhood ALL with inherited susceptibility. Nevertheless, since the advent of genome-wide technologies, some studies have reported that two independent loci located within IKZF1 are exclusively associated with the risk of developing ALL (5, 6). After replication of this discovery by other studies, a meta-analysis provided large-scale evidence that both IKZF1 SNPs contribute to the occurrence of BCP-ALL (25), and our study could clearly confirm the association of IKZF1 rs4132601 with childhood BCP-ALL.

A striking feature in the risk association involving IKZF1 is that somatic mutations within the same gene are recurrently detected in cases of BCP-ALL (1, 2), which suggests that germinative variant alleles may favor an intrinsic vulnerability of precursor cells to acquire additional transforming events, such as IKZF1 somatic deletions. In accordance, the interplay between genetic susceptibility and somatic abnormalities in the etiology of childhood ALL has been previously demonstrated by others. For example, SNPs located at 10q21.2 (ARID5B) appear to be highly associated with the risk of developing hyperdiploid ALL (5, 6). Furthermore, an ARID5B SNP conferred increased risk to acquire specific MLL rearrangements in a cohort of early childhood leukemia (14). In addition, germline variation of TP63 and PTPRJ genes can contribute specifically to the risk of the subset of BCP-ALL with ETV6-RUNX1 (26). Our findings demonstrate that IKZF1 SNPs can distinctively contribute to the risk of BCP-ALL with IKZF1 deletions, especially those leading to haploinsufficiency.

Because breakpoints of intragenic deletions of IKZF1 are located nearby RSSs (3, 11), which are recognized by the RAG complex for V(D)J recombination, one hypothesis that could explain the relationship between IKZF1 SNPs and risk of acquiring IKZF1 deletions is that SNPs could modify RSSs nearby deletion breakpoints, thus influencing RAG complex recognition. To evaluate this hypothesis, we mapped recurrent SNPs located proximal to IKZF1 breakpoint hotspots and checked whether they altered the minimal recombination motif of RAG. Although our findings showed that some SNPs could change RSSs nearby recurrent deletion breakpoints, the allele frequency was very low to explain this risk association.

The relationship between IKZF1 SNP and disease risk may be explained by the genotype influence on gene expression. Papaemmanuil and colleagues investigated the relationship between rs4132601 and expression of IKZF1 transcript in Epstein–Barr virus–transformed lymphocytes. They found an association between mRNA expression and genotype in a dose-dependent manner; the homozygous variant genotype had lower expression levels and therefore might disrupt B-cell differentiation (5). As rs4132601 is not located on the promoter or enhancer region of IKZF1, the process linking its SNP and disease risk in not fully understood. One hypothesis is that it could indirectly alter gene expression when in linkage disequilibrium with another SNP. Notably, rs4132601 is in linkage disequilibrium with rs11978267 and rs1110701. The latter lies on enhancer regions of IKZF1 annotated in GM12878 cell line (19, 27, 28). Thus, our data suggest that haplotypes carrying variant genotypes of rs4132601 and rs11978267 may be related to differential expression of IKZF1 and contribute to leukemic transformation in a similar way of deregulation that is promoted by IKZF1 deletions (as summarized in Supplementary Fig. S2).

The IKZF1 transcripts are highly expressed in hematopoietic progenitor cells and during lymphopoiesis. Investigation of Ikaros function in early B-cell development suggested that it regulates its targets in a stage-specific manner (29). Our study demonstrated that IKZF1 expression is more pronounced in earlier stages of B-cell differentiation (CD10-negative B cells), where it could dictate lymphoid gene expression.

The analysis whether the risk alleles were associated to any specific combination of genes abnormalities showed that the presence of variant alleles of both IKZF1 SNPs increases the risk of BCP-ALL with: IKZF1+9p, “IKZF1 plus” and IKZF1+BTG1 deletions. One of the most interesting findings from this study was the risk association for individuals harboring both IKZF1 and BTG1 deletions; after analyzing an increased number of patients, we could confirm this result. According to Scheijen and colleagues, BTG1 deletions strongly enhance the risk of relapse in patients with IKZF1 deletions and augment the glucocorticoid resistance phenotype mediated by loss of IKZF1 function, while combined loss of IKZF1 and other deletions was also significantly associated but had no impact on prognosis (9, 12). The biological mechanism underlying such association is uncertain. Nevertheless, recent studies demonstrated that the acquisition of IKZF1 and BTG1 deletions is mediated by RAG (3, 30). As IKZF1 encodes a transcription factor that regulates many downstream targets in the lymphoid differentiation pathway (e.g., RAG and TdT), we hypothesize that reduced levels of IKZF1 transcripts may disrupt RAG expression and, thus, promote the concurrent occurrence of somatic deletions.

Some limitations in this analysis should be mentioned. First, the small number of cases after some subsets stratification raises concern with regard to statistical power. Second, the presence of bias in the selection of controls cannot be ruled out, once the samples were obtained from hospitalized children and not from the overall population. Third, we had missing biological sample in some cases that precluded us to have all samples screened for mRNA expression. Finally, the findings obtained in this study have not been validated by independent replication. Therefore, data presented here must be carefully interpreted, and future studies are needed to confirm these results.

We can also mention some study strengths. This study included children from varied geographical regions in Brazil, being the risk associations identified very consistent and concordant with previously published data indicating, therefore, good reliability and sensitivity of our data. Another important contribution of the current work is the replication of genome-wide association studies in a population different from the American and European ones, where most of the studies were so far concentrated. Moreover, studies involving gene–gene interactions are highly desirable in the context of the etiology of childhood ALL, and innovative results have been obtained in this field.

In summary, our findings demonstrate that germline genetic variation can contribute not only to the risk of childhood BCP-ALL in general, but also specifically contribute to the risk of BCP-ALL subtypes, in this case patients with IKZF1 deletions only or with additional somatic alterations.

No potential conflicts of interest were disclosed.

Conception and design: M. Emerenciano

Development of methodology: B.A. Lopes, T.C. Barbosa, B.K.S. Souza, M. Emerenciano

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.A. Lopes, T.C. Barbosa, B.K.S. Souza, C.P. Poubel, M.S. Pombo-de-Oliveira, M. Emerenciano

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.A. Lopes, T.C. Barbosa, C.P. Poubel, M. Emerenciano

Writing, review, and/or revision of the manuscript: B.A. Lopes, T.C. Barbosa, M. Emerenciano

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.S. Pombo-de-Oliveira

We are grateful to the children and their parents for participating in the study. We thank Priscilla M.S. Ferreira for assisting the genotyping of rs11978267. We thank Alessandra J. Faro, Camilla F.G. Andrade, MSc. Caroline Zampier, Dr. Elda P. Noronha, Dr. Eugenia T.G. Pina, and Dr. Gisele Vasconcelos, who contributed with laboratory diagnosis of acute leukemia.

The Brazilian Study Group for Childhood Leukemia

Bruno A. A. Gonçalves¹, Cynthia C. Neves², Jozina M. A. Agareno², Lilian M. B. Carvalho³, Flávia N. S. Araújo², Nilma P. Brito³, Isis Q. Magalhães⁴, José Carlos Cordoba⁴, Flávia Pimenta⁵, Andreia Gadelha⁵, Eloísa Cartaxo⁵, Rosania M. Basegio⁷, Atalla Mnayarji⁷, Marcelo S. Souza⁷, Gustavo R. Neves ⁸, Renato Melaragno⁸, Virgínia M. Cóser⁹, Thereza C. Lafayete⁹, Patricia C. Brito¹⁰, Adriana S. Deyl¹¹, Alejandro M. Arancibia¹², Teresinha J. M. Salles¹³.

Affiliations: ¹Research Center of Instituto Nacional de Cancer, Rio de Janeiro, RJ. ²Sociedade de Oncologia da Bahia, Salvador, BA; ³Hospital Martagão Gesteira, Salvador, BA; ⁴Hospital de Apoio Brasília, Unidade de Onco-Hematologia Pediátrica, Brasília, DF; ⁵Hospital Napoleão Laureano, João Pessoa, PB; ⁶Departamento de Pediatria, Faculdade de Medicina, Universidade Federal de Minas Gerais, Belo Horizonte, MG; ⁷Hospital Rosa Pedrossian, Campo Grande, MS; ⁸Hospital Santa Marcelina, São Paulo, SP; ⁹Departamento de Pediatria, Faculdade de Medicina, Universidade Federal de Santa Maria, Santa Maria, RS; ¹⁰Hospital Araújo Jorge; ¹¹Hospital das Clínicas de Porto Alegre, Porto Alegre, RS; ¹²Instituto do Câncer Goiânia, GO Infantil, Porto Alegre, RS; ¹³Hospital Universitário Oswaldo Cruz, Recife, PE.

This investigation was supported by the Brazilian National Counsel of Technological and Scientific Development (CNPq#447385/2014-3) and by the Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ#E-26/110.533/2014). M. Emerenciano has been supported by CNPq (PQ-2014#304142/2014-0) and FAPERJ-JCNE (E_26/201.539/2014) research scholarships.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.