Novel associations between activating killer-cell immunoglobulin-like receptor genes and childhood leukemia

Acute lymphoblastic leukemia (ALL) is the predominant form of leukemia occurring in children and accounts for approximately 75% of all leukemia cases. It is a heterogeneous group of white blood cell cancers that usually involve B- or T-cell precursors (blasts) and are accordingly called pre-B ALL (B-ALL) or pre-T ALL (T-ALL). In Western countries, B-ALL is the predominant form of ALL in children and comprises 70%-80% of the pediatric ALL cases, whereas T-ALL is a rare tumor in these populations and accounts for approximately 15% of the cases.^1,2  The exact causes of leukemia are not known. Several environmental factors, including infections at an early age, exposure to ionizing radiation and certain chemicals (eg, benzene), and parental use of alcohol and tobacco, have been recognized as risk factors for ALL in children. Epidemiologic studies suggest that the genetic makeup of an individual plays an important role in determining his or her innate resistance or susceptibility to the development of this cancer. In this respect, we and others^3-6  have shown that mutations in genes that compromise a person's ability to metabolize carcinogens, maintain DNA integrity, adequately respond to oxidative stress, and maintain cell-cycle check points increase the risk for developing ALL in children (reviewed in Stiller¹ ). Recent genome-wide association studies have identified single-nucleotide polymorphisms in several new genes that could enhance susceptibility for childhood ALL. The genes include the AT-rich interactive domain (ARID)-5B, CCAAT/enhancer binding protein-ϵ (CEBPE), hydroxy acid oxidase (HAO)-1, erythrocyte membrane protein band 4.1-like (EPB4.1L)-2, and IKAROS family zinc finger protein (IKZF)-1.^7-9 

Genome-wide association studies are very powerful and have the ability to identify key genetic determinants implicated in conferring susceptibility or resistance to human diseases, including B-ALL. An inherent limitation of these studies is that currently available chips cover only areas of the genome that have been well characterized for single-nucleotide polymorphisms and small insertion-deletions. Unfortunately, there are numerous regions of the genome that have not yet been well characterized, and variations therein have not been well represented in these chips. For such regions, a candidate gene approach would be necessary to identify potential susceptibility/resistance genes. One such region in the genome is the leukocyte receptor complex, which is located on human chromosome 19q13.4 and houses the killer-cell immunoglobulin-like receptor (KIR) genes (reviewed in Yawata et al,¹⁰  Parham,¹¹  and Ianello et al¹² ). Investigation of this region for potential genetic determinants for B-ALL is of great interest because some KIR genes have been shown to be associated with development of certain forms of leukemia in humans.¹³ 

The KIR gene family comprises 14 functional genes. The genes encode synonymous receptors that are expressed mainly on natural killer (NK) cells and regulate their functional activities.¹²  NK cells have been well documented for their ability to kill cancer cells in the body, including leukemic ones.^14,15  Of the KIR genes, 6 encode receptors that, on binding with their cognate ligands on target cells, activate NK cells and trigger NK cell–mediated killing and secretion of soluble mediators. Hence, these genes have been named as activating KIR genes and include KIR2DS1-3, KIR2DS4, and KIR3DS1.^11,12  Although all KIR genes are highly polymorphic, an important variation with respect to the activating KIR genes is that humans may carry 1 or more of these genes. This variation in the inheritance of activating KIR genes results from differences in KIR haplotypes, which have been divided into 2 groups, A and B. The group A haplotypes are less variable and carry only 1 activating KIR gene, KIR2DS4, which often has a 22-bp deletion in exon 5 and encodes a soluble nonfunctional receptor.¹⁶  Thus, humans with the AA genotype may not carry any functional activating KIR gene. The group B haplotypes are more variable and may carry a full complement or a subset of the activating KIR genes. Some of these haplotypes lack KIR2DS4 but contain 1 or more of the other 5 activating KIR genes. Several studies have shown that KIR genotypes and the number of activating KIR genes inherited by a person may affect resistance/susceptibility to infectious agents, development of autoimmune diseases, reproductive efficiency, GVHD, and cancer^17-22  (reviewed in Kulkarni et al²³  and Farag et al²⁴ ). As mentioned above, they have also been shown to protect humans from certain types of leukemia.¹³  However, little is known concerning the impact of activating KIR gene content of humans on their innate resistance/susceptibility to develop B-ALL. In the present study, we show that activating KIR genes are associated with decreased risk for development of B-ALL in children. Furthermore, we also show that the inheritance of a larger number of these genes in humans is associated with an enhanced reduction in the risk for developing this malignancy.

To investigate associations between activating KIR genes and B-ALL, we performed a case-control study at the Centre Hospitalier Universitaire Sainte-Justine Research Center in Montreal. Cases of B-ALL (n = 102) were recruited from the hematology/oncology clinic of the hospital during 1989-2004 and included children diagnosed before 18 years of age. Diagnosis of the leukemia was based on established criteria that included histologic examination of the blood and bone marrow smears, as well as immunophenotyping of the lymphoblasts. The relevant demographic, clinical, and histopathologic data were extracted from the medical records of these patients or acquired from the children's parents and are shown in Table 1.

Given the potential heterogeneity in the distribution of KIR gene frequencies within seemingly ethnically homogeneous populations, control subjects (n = 245) were selected from different sources to enhance population representation. These included children visiting the orthopedic department of the study hospital for minor fractures, their siblings, and randomly selected children and adults from the general population. The control subjects were restricted to people who were without any type of leukemia or any other cancer or autoimmune disorder. These control subjects have been used previously for replication or confirmation of associations between several susceptibility genes and complex diseases such as inflammatory bowel diseases.^25-27  It is noteworthy that the KIR gene frequencies within the control population are reflective of the distribution within other populations of similar ancestry (Belgian and French; http://www.allelefrequencies.net; Table 2), which further demonstrates that bias, if any, in the selection of control subjects was limited. Both case subjects and control subjects were restricted to those with self-reported French Canadian ancestry and who resided in Quebec at the time of recruitment.

To assess whether associations between the KIR genes and ALL differed according to the ALL phenotype and ethnicity, we further investigated these associations in a sample of 30 T-ALL pediatric patients of mixed white ancestry and in another sample of 40 pre-B ALL pediatric patients of non-French ancestry. The characteristics of these 2 cohorts are also given in Table 1.

Blood or saliva samples were collected from the participants as a source of DNA. Genomic DNA from these samples was extracted by a salting out technique or with commercial kits (Oragene, DNA Genotek). The quantity and quality of the extracted DNA samples were determined by UV spectrophotometry. The samples were coded and kept at the BioBank maintained at the Centre Hospitalier Universitaire Sainte-Justine Research Center.

The study was approved by the institutional ethics committee of the Centre Hospitalier (CHU) Ste-Justine, and written informed consent was acquired from all participants in accordance with the Declaration of Helsinki.

Genomic DNA samples from patients and control subjects were analyzed for the presence or absence of individual activating KIR genes by use of PCR and published sequence-specific primers as described previously.^16,19,28-33  Standard PCR protocols were used. The general procedure included initial denaturation at 95°C for 5 minutes followed by 30-35 cycles, each comprising 30 seconds of denaturation at 95°C, 30 seconds of annealing at 60-68°C, and 2 minutes of extension at 72°C. The final cycle included extension for 10 minutes. The number of PCR cycles and annealing temperatures were adjusted for each gene in preliminary experiments. The presence or absence of the genes was determined by running the reaction products on 2%-2.5% agarose gels. The gels were stained with ethidium bromide, scanned, and imaged. The appearance of the amplicon band of the expected molecular weight was considered to indicate the presence of the gene in the genomic DNA sample. DNA samples known to be positive or negative for the gene were used as positive and negative controls, respectively. The PCR reactions were also run without the DNA template as a safeguard against false-positive cases, as well as for a housekeeping gene, GAPDH, as a control for false-negative cases. Despite all the precautions, the PCR reactions for 2 patients did not succeed, and these patients were excluded from the analyses.

On the basis of the genotyping results, patients were classified as those in whom the gene was present or absent. Because we could not distinguish between heterozygotes and homozygotes for a particular gene, it was not possible to examine deviation from Hardy-Weinberg equilibrium for the studied genes. Initial analysis included examination of the distribution of the genes in case and control subjects with χ² tests. Subsequently, logistic regression analysis was performed. A single model for each gene as the independent variable was first fit. Because many of the KIR genes studied are in high linkage disequilibrium with each other, a multivariate model that included all 6 genes was then fit to assess the independent effect of each gene after controlling for the effects of other genes. In addition to the effect of individual genes, we also assessed the combined effects of harboring 1 or more activating genes. For this purpose, we stratified the samples according to the number of genes harbored (≤ 1, 2-3, and ≥ 4). Odds ratios and corresponding 95% confidence intervals were estimated. All analyses were performed with STATA software (Version 10; STATA Corp).

A total of 100 B-ALL cases of French Canadian ancestry and 245 control subjects were successfully genotyped for the 6 activating KIR genes. Table 2 shows results from univariate analysis for the association between the KIR genes and risk for ALL. The frequencies of all the activating KIR genes were lower in patients than in control subjects, and all of them significantly (P < .05) decreased the risk for developing B-ALL. Similar results were observed in the unrelated white cohort of B-ALL patients of non–French Canadian ancestry. As shown in Table 3, all of the genes showed significant associations in this sample except for the KIR2DS4 gene. A similar pattern of associations was evident for the unrelated sample of white T-ALL patients. As shown in Table 4, significant associations were observed with 3 genes (KIR2DS1, KIR2DS2, and KIR2DS3). In this smaller cohort, associations did not reach statistical significance (P > .05) in the case of the other 3 genes (KIR2DS4, KIR2DS5, and KIR3DS1), potentially because of low power. On the basis of the similarity of associations across the ALL phenotypes, we performed an analysis that combined all of the ALL patients. A multivariate logistic regression analysis of the combined cohort that accounted for the effects of other genes showed that each gene independently conferred protection for ALL (Table 5). The strongest associations were noted with the KIR2DS2 gene (odds ratio 0.27, 95% confidence interval 0.16-0.43, P = 1.14 × 10⁻⁷), with a similar pattern of association across each of the different ALL phenotypes.

Because of haplotypic variations, humans may vary with respect to the number of inherited activating KIR genes, and therefore, we investigated the impact of the number of inherited activating KIR genes on the risk for developing this leukemia (including all of the ALL phenotypes). We observed that carrying a higher number of KIR genes was strongly associated with increasing protection for ALL (Table 6). Compared with children who harbored 1 or no genes, children who harbored 4 or more KIR genes were least susceptible to ALL (odds ratio 0.06, 95% confidence interval 0.03-0.13, P = 1.1 × 10⁻¹⁵).

In this case-control study, we investigated whether activating KIR genes were associated with susceptibility or resistance to B-ALL in Canadian children. We observed that all 6 activating KIR genes tended to reduce risk and confer protection against the acquisition of this type of leukemia after adjustment for the effects of individual genes (P values as low as 10⁻⁷). However, the decrease in risk was not the same with different genes. For example, the maximum reduction in the risk for developing B-ALL was conferred by inheritance of the KIR2DS2 gene. Furthermore, the present data strongly suggest that harboring increasing numbers of KIR genes provides additional protection against the acquisition of ALL. Furthermore, observed associations with the KIR genes extended to other phenotypes of ALL (T-ALL) and to white individuals of differing ancestry (whites of non-French ancestry).

To the best of our knowledge, this is first study that has described associations of activating KIR genes with reduction in risk for developing childhood ALL. A few earlier studies focused on adult patients. One group^34,35  (reviewed in Verheyden et al¹⁵ ) has shown potential associations of certain inhibitory KIR-HLA gene interactions but did not demonstrate independent effects of activating KIR genes with chronic lymphoid and chronic myeloid leukemias in a Belgian population. No associations with ALL were noted in that study, probably because of low power (n = 8). Furthermore, the researchers did not discriminate between functional and mutant variants of the KIR2DS4 gene. As mentioned in the introduction, this gene is frequently mutated and shows a 22-bp deletion,¹⁶  which results in the expression of a nonfunctional receptor. In all the analyses performed in the present study, we considered those who showed only this 22-bp mutant form as being negative for the gene. Subjects were considered positive for the gene when a nonmutated amplicon band was seen on the gel with or without the presence of the mutated band. In another study conducted in Polish and German individuals, Giebel et al¹⁷  found a protective association between the KIR2DS4 gene and chronic myeloid leukemia, but no associations with ALL were evident. Similar to the previous study, the lack of association with ALL may have been the result of the small number of ALL patients (n = 21) investigated. A more recent study in a Chinese population¹³  has shown that activating KIR genes were more frequent in chronic myeloid leukemia patients than in healthy control subjects, although differences achieved statistical significance only for the KIR2DS4 gene. In contrast, that group found significant protective associations between the KIR2DS3 and ALL. The present findings related to the KIR2DS3 gene and ALL are consistent with that study. The lack of associations noted for the other KIR genes in the Chinese study may be the result of low power, age-of-onset effects, different types of leukemia, or population-specific differences in the frequencies of the KIR genes.

It is noteworthy that all the previous studies^13,17,34,35  examined associations between KIR genes and adult-onset leukemia. Because the present study was focused on childhood-onset leukemia, direct comparisons between it and the earlier studies may not be valid and conclusive. This is particularly the case given that early-onset leukemia may involve different causes and hence differential susceptibilities vis-à-vis adult-onset leukemia. Furthermore, even within childhood leukemia, causes of ALL may differ between different types (B-ALL, T-ALL, and ALL with unspecified cell types). In the pediatric context, the present findings suggest that associations between the KIR genes may be common to at least 2 ALL phenotypes, B-ALL and T-ALL. Because of the lower prevalence of T-ALL, we could only manage to test a limited number (n = 30) of these cases in the present study. Despite their limited number, frequencies of all activating KIR genes were lower in the case subjects and reached significance (P < .05) for 3 of the genes (KIR2DS1, KIR2DS2, and KIR2DS3; Table 4). These results suggest that all of the activating KIR genes could have shown significant associations with this type of childhood leukemia had there been a larger number of case subjects. Certainly, additional studies will be required to confirm whether the findings observed here are unique to early-onset ALL. In the context of case-control studies performed for examining genetic associations, confounding because of population stratification (ethnicity) is a potential concern; however, samples in the present study were homogeneous with regard to ethnicity. Residual stratification cannot be ruled out; however, such stratification, if any, is unlikely to account for the strong associations noted in the study.

Ideally, we would have investigated the expression of ligands for activating KIR genes in the case and control subjects in the present study. Unfortunately, the ligands for these receptors remain largely unknown. Unlike their inhibitory counterparts, activating KIRs bind to HLA class I antigens with very low affinities.³⁶  In fact, it is believed that these activating receptors may bind to the ligands that are expressed de novo in cancer or virus-infected cells, or they may recognize foreign peptides bound with HLA class I antigens. In the case of KIR2DS4, the receptor was described as binding an unidentified ligand expressed by primary melanoma cells.³⁷  Another activating KIR gene, KIR3DS1, which was shown to be protective against B-ALL in the present study, has been associated with protection against infection with HIV type 1, as well as with delayed progression of the infection toward AIDS.^18,22  In fact, KIR3DS1 is the collective name given to several allelic variants of KIR3DL1. These variants encode a short-tailed activating version of the receptor. KIR3DL1 is one of the most polymorphic genes in the KIR family. It exists in 59 allelic variants, of which 13 encode the short-tailed activating version of the receptor and hence are collectively designated as KIR3DS1, whereas the remaining 46 allelic variants encode a long-tailed inhibitory version of the receptor and are designated as KIR3DL1.^10-12  Researchers have discovered that KIR3DL1 receptors bind HLA-Bw4 allotypes that have an isoleucine (Ile) at position 80 in their α-1 domain. The activating versions of the receptor were shown to bind the HLA-Bw4 allotypes with threonine (Thr) at position 80. However, subsequent studies have not confirmed these observations.^10-12  Thus, like other short-cytoplasmic-tail activating KIR, the exact ligands for KIR3DS1 remain unknown. Furthermore, in the absence of antibodies (commercial or otherwise) that bind specifically with activating KIRs, functional studies on these receptors and their ligands must await advances in these areas.

It is noteworthy that in addition to NK cells, activating KIRs are also expressed on the surface of certain subsets of T cells. It is believed that activation of these receptors decreases the activation threshold of the immune cells and helps in controlling cancers and viral infections.^23,24  The present findings related to childhood ALL fit with this paradigm and provide important novel insights concerning the immunopathogenesis of this malignancy. Because of their expression on the cell surface, KIRs are amenable to therapeutic interventions with antibodies or peptide mimics, unlike other target proteins that may be expressed inside cells. Furthermore, appropriately selected NK cells may be used as effector cells in future immunotherapeutic approaches against childhood leukemia.

In conclusion, the present study demonstrates novel associations between activating KIR genes and protection against childhood ALL.

The authors thank the Canadian Cancer Society Research Institute for support.

Contribution: Z.A. and S.S. designed and performed the experiments and wrote the first version of the manuscript; A.I. and O.D. helped in data interpretation and in writing the manuscript; M.D., C.-I.R., and D.S. helped in patient selection; D.K.A. provided important insights on the disease, analyzed and interpreted the results, and contributed to the manuscript; A.A. conceived the idea, supervised the study, and contributed to the manuscript; and all authors read and contributed to the final version of the manuscript.

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

Correspondence: Ali Ahmad, Laboratory of Innate Immunity, CHU Sainte-Justine Research Center/Department of Microbiology & Immunology, 3175 Cote Ste Catherine, Montreal, QC, H3T 1C5, Canada; e-mail: ali.ahmad@recherche-ste-justine.qc.ca.