Association of killer cell immunoglobulin-like receptors with endemic Burkitt lymphoma in Kenyan children

Endemic Burkitt lymphoma (eBL) is an aggressive pediatric B cell lymphoma, common in Equatorial Africa. Co-infections with Epstein-Barr virus (EBV) and Plasmodium falciparum, coupled with c-myc translocation are involved in eBL etiology. Infection-induced immune evasion mechanisms to avoid T cell cytotoxicity may increase the role of Natural killer (NK) cells in anti-tumor immunosurveillance. Killer immunoglobulin-like receptor (KIR) genes on NK cells exhibit genotypic and allelic variations and are associated with susceptibility to diseases and malignancies. However, their role in eBL pathogenesis remains undefined. This retrospective study genotyped sixteen KIR genes and compared their frequencies in eBL patients (n = 104) and healthy geographically-matched children (n = 104) using sequence-specific primers polymerase chain reaction (SSP-PCR) technique. The relationship between KIR polymorphisms with EBV loads and eBL pathogenesis was investigated. Possession of ≥ 4 activating KIRs predisposed individuals to eBL (OR = 3.340; 95% CI 1.530–7.825; p = 0.004). High EBV levels were observed in Bx haplogroup (p = 0.016) and AB genotypes (p = 0.042) relative to AA haplogroup and AA genotype respectively, in eBL patients but not in healthy controls. Our results suggest that KIR-mediated NK cell stimulation could mute EBV control, contributing to eBL pathogenesis.

www.nature.com/scientificreports/ do not kill healthy cells when only the inhibitory receptors are ligated to HLA-I ligands on the target cells, since there is no activating signal generated 15 . Down-regulation of HLA-I in target cells by viral infections or neoplastic transformation results in a lack of ligation of inhibitory NK cell receptors to their ligands, hence the absence of NK cell inhibition. Instead, only the activating NK cell receptors are ligated to activating ligands, resulting in NK cell stimulation to kill target cells 16 . The outcome of the interaction of NK cells with tumor cells containing ligands for both inhibitory and activating receptors depends on the balance of the strength of signals generated 15 . The KIR gene family comprises rapidly evolving genes present in all primates 17 . The genes contain two (2D) or three (3D) domains in the extracellular region, with a short (S) or a long (L) cytoplasmic tail 18 . The KIR genes 2DL1, 2DL2/2DL3, 2DL5, 3DL1, 3DL2, and 3DL3 have a long tail with an inhibitory motif. Short cytoplasmic tailed KIRs have activating motifs and include 2DS1, 2DS2, 2DS4, 2DS3/2DS5, and 3DS1 18 . KIR2DL4 is the only long-tailed receptor with both inhibitory and activating motifs 19 . There are two pseudogenes, 2DP1 and 3DP1 20 (Fig. 1). These genes are arranged in a head-to-tail order in the long arm of chromosome 19 (19q13.4), within the Leukocyte Receptor Complex (LRC) 21 . Each KIR gene is 10-16 kb in length, with a 2 kb sequence separating each gene pair, except a 14 kb sequence that occurs upstream of KIR2DL4 21 . The expression of KIR genes varies between NK cell subsets and is controlled by four types of promoters 22 . CD56 dim NK cells express all KIR genes except 3DL3 22 . 2DL4 occurs on both CD56 bright and -dim NK cells in a non-variegated manner 22 . Some KIR genes demonstrate variations in their sequences, for example, KIR2DS4 has a 22 base pair (bp) deletion in the second extracellular domain which results in a non-functional gene 21 . A deletion of one base-pair in exon 4 of KIR2DP1 introduces a stop codon, resulting in a pseudogene 23 . Another pseudogene, KIR3DP1 has a deletion of 1.5 kb that removes exon 2. There are no transcripts for the 2 pseudogenes 23 . KIR2DL5 has two variants A and B, encoded by different loci 24 . 2DL5B is in the centromeric region, while 2DL5A occurs in the telomeric region 25 .
Polymorphisms within the KIR locus result from gene content, allelic, and copy number variations 26 . Based on gene content and copy number, KIRs are grouped into inhibitory haplotype A and activating haplotype B 27 . The haplotypes are further subdivided into AA and Bx genotypes, where x can be either A or B 28 . There are more than 500 different Bx groups in the database (http:// www. allel efreq uenci es. net) 28 21 . This haplotype has a Bx group containing one (AB heterozygous) or two (BB homozygous) genotypes. Genotype BB does not have one or more of the group A KIR genes. All the remaining genotypes in haplotype B are defined as AB 29 . 3DL3, 3DL2, 2DL4, and 3DP1 are framework genes, hence they appear in all haplotypes 23 . KIR haplotypes are split into centromeric A or B (cA, cB) and telomeric A or B (tA, tB) halves 30,31 . Both cA, cB and tA, tB regions exhibit an even balance in East Africa population 32 . Classification of KIR based on presence/absence of a gene generates eight telomeric regions (tA01, tB01, tB02, tB03, tB04, tB05, tB06 and tB07) and nine centromeric regions (cA01, cA02, cA03, cB01, cB02, cB03, cB04, cB05 and cB06) 33 . KIR2DL5, 2DS5, and 2DS3 are duplicated and can occur in centromeric and/or telomeric locations 34 . Genes occurring in different regions of the KIR complex may undergo homologous recombination, resulting in expanded and contracted haplotypes 34,35 The B content score is the sum of cenB and/or telB motifs in each genotype 36 . The Bx group can be classified further into four subsets, by considering two gene clusters; T4, containing KIR2DL5-3DS1-2DS1-2DS5 genes, and C4, which has KIR2DL2-2DS2-2DS3-2DL5 genes. The C4T4 contains both C4 and T4 genes, while the C4Tx subset has C4 but lacks T4 genes. CxT4 lacks C4 genes, thus it contains T4 genes. The absence of both C4 and T4 genes results in CxTx subset 29,37 .
Studies have suggested that specific KIRs influence the generation of either inhibitory versus activating signals. A balance between these signals determines whether NK cells bypass or kill viral-infected or tumor cells 38 . www.nature.com/scientificreports/ Consequently, these signals can influence an individual's susceptibility to diseases and malignancies 16,30,39,40 .
The presence of certain KIRs has been associated with cancer pathogenesis. For instance, an increased number of activating KIRs predispose individuals to EBV-related nasopharyngeal carcinoma (NPC) 16 , whereas the presence of genotype B, which mainly contains activating KIRs is associated with gastric cancer lesions 30 . In contrast, the Bx haplogroup protects against colorectal adenocarcinoma 41 . However, there is little understanding of the impact of KIR polymorphisms on eBL pathogenesis. Therefore, to improve our understanding of how KIR genes may contribute to eBL pathology, we performed KIR genotyping using commercially available kits and analyzed the haplotype, genotypes, centromere-telomere regions, Bx subsets, and B score contents in eBL patients and healthy controls. Given the strong link between EBV and eBL 42 , we further evaluated the association of haplogroups AA/ Bx and genotypes AA, AB, and BB with EBV loads, to determine viral control.

KIR genes.
To characterize the frequencies of KIR genes in the study population, we genotyped the genes responsible for inhibitory signals (2DL1, 2DL2/2DL3, 2DL5, 3DL1), activating signals (2DS1, 2DS2, 2DS4, 2DS3/2DS5, and 3DS1), the framework and pseudogenes (2DL4, 3DL2, 3DL3, 2DP1 and 3DP1) from genomic DNA using sequence-specific primers polymerase chain reaction (SSP-PCR) technique. KIR genotypes were classified based on the presence or lack of each gene locus and were analyzed to determine differences in their frequencies between eBL patients and healthy controls (HC). The genes KIR3DP1, KIR2DP1, KIR2DL1, KIR2DL4, KIR3DL2 and KIR3DL3 occurred at a frequency ≥ 99% and were excluded from the association analysis. The KIR genes were not statistically different between the study groups ( KIR centromeric and telomeric distribution. The KIR gene contents vary in the centromeric and telomeric regions. To investigate these differences in our study population, genotypes AA and Bx were grouped into centromeric and telomeric contents 31 . A total of 6 centromeric and 2 telomeric genetic regions were reported. There were no significant differences in these regions when comparing eBL patients with the control group (Supplementary Table S1).  Table S2). However, the observed differences were not statistically significant.
Bx Subsets and the number of activating KIRs. We observed that all the four Bx subsets (C4T4, CxT4, C4Tx, CxTx) were present in the study population (Table 2). C4T4 had the least frequency, with no representation among the healthy individuals. CxTx had the highest frequency in both eBL patients and healthy controls. Further, in order to investigate possible differences in the number of KIRs, we compared the iKIRs and aKIRs in cases and controls. There were an increased proportion of eBL patients with ≥ 4 aKIRs relative to HC.
KIR haplotypes, genotypes, and EBV Viral loads. We compared EBV viral loads and observed significantly higher viremia in children with eBL (median 6496.571 EBV copies/μg of DNA) compared to healthy children from the malaria holoendemic region (median 202.697 EBV copies/μg of DNA) (p-value <0.0001). Next, we investigated whether the EBV loads differed between the haplotypes and genotypes. Considering the KIR haplotypes and genotypes, we observed significant differences in EBV load in eBL patients but not in healthy controls (Fig. 3).

Discussion
KIR gene polymorphisms predispose individuals to various malignancies associated with viruses 16 . However, few studies have evaluated the role of such polymorphisms in eBL etiology. To address this issue, we evaluated the association of KIR genes with eBL development. The most common KIR genotype in our study population was homozygous A, with the genotype id AA1. This genotype has previously been shown to be the most frequent in all worldwide populations, including Africans [44][45][46] . Interestingly, while its frequency in healthy controls was consistent with the expected frequency in African populations (35.6%) 45 , the representation was lower in eBL patients (27.9%). However, such variations have been reported in a few African populations, from 12.0% in the Xhosa population of South Africa, 28.1% in the Ugandan population, and 42.0% in Senegal 44,47,48 . In our study, the Bx genotype was highly variable, with a frequency range of 0.0-10.6%. This genotype consists of two haplotypes; AB and BB. Most of the study participants were heterozygous AB, and there was a very low frequency of homozygous BB. The considerable diversity for Bx but not AA genotypes may be a result of copy number variation 33 due to selection pressure from environmental, climatic, chronic, and infectious diseases that have prevailed in our study population for many years 40 . Recently, malaria has been shown to drive selection for this haplotype in a Ugandan population 49 . In our study, there were more eBL patients carrying ≥ 4 aKIRs compared to healthy controls; suggesting that individuals with ≥ 4 aKIRs may have a high risk of developing eBL. Consistent www.nature.com/scientificreports/ with our findings, an increased number of aKIRs predispose individuals to colorectal adenocarcinoma, human papillomavirus-associated cervical cancer and EBV-associated nasopharyngeal carcinoma 16,39,50 . The role of the number of aKIRs in the etiology of cancers is explained by two hypotheses 39 . First, an increased number may protect individuals against cancers, due to enhanced cytolysis of tumor cells, resulting from increased NK cell activation 30 . In contrast, increased immune activation of NK cells by aKIRs may cause non-specific inflammatory responses, such as oxidative DNA damage 16,30 . Such responses may increase the risk of cancer development 30 . Therefore, considering the second hypothesis, our findings raise a possibility that an increasing number of aKIRs coupled with repeated infections with Plasmodium falciparum in our study population 51 could be associated with increased NK cell activation resulting in inflammation-associated oncogenesis.
The study participants had a higher frequency of centromeric B region and a lower frequency of telomeric B region. Similar observations were reported in a Ugandan population 44 . Generally, cenB region is common in the African population relative to telB region 44 . The number of B motifs in the centromere and telomere regions influences NK cell activation 30 . Subsequently, the B motif is associated with disease outcome 30 . In this study, we evaluated how the number of KIR B gene motifs of centromeric or telomeric origin influences eBL development. There were no significant differences in the B score when comparing eBL patients with the control group.
Previous studies have reported that children living in malaria holoendemic areas experience primary EBV infection at an early age compared to children residing in areas with lower incidences of malaria 52 . In addition, repeated exposure to malaria is associated with poor EBV control, hence higher viremia 53 . Consistent with these findings, we observed that eBL patients had higher median EBV loads relative to healthy controls (6496.571 versus 202.697 EBV copies/μg of DNA), respectively, p-value < 0.0001). The EBV levels were significantly different Figure 2. The occurrence of KIR genotypes in the study population. Thirty-five different KIR genotypes were observed in the 208 persons. The genotypes differed from each other by the presence of (black box) or absence (open box) of KIR genes. The variants KIR2DL5A and 2DL5B were considered as KIR2DL5 while KIR2DS4 mutant and 2DS4 full length were considered as KIR2DS4, we analyzed a total of 16 KIR genes. The data are expressed as percentage frequency, obtained by dividing the number of individuals having the genotype by the number of individuals in the studied group. Genotype ID reference numbers were acquired from the KIR genotype database (http:// www. allel efreq uenci es. net). www.nature.com/scientificreports/ when considering the KIR haplogroup and genotypes in children with eBL but not in healthy controls, with higher EBV loads observed in Bx relative to AA haplogroup. Considering the genotypes, EBV load differed between AA and AB. NK cells are essential in the control of infections associated with viruses 54 . Their subsets expand upon infection with herpes viruses 55 , and the proliferation positively correlates with EBV viral loads 55 . Individuals deficient in NK cells are predisposed to herpes viruses-associated infections 56 . The persistence of viruses in an individual may cause chronic recruitment and activation of NK cells, up to when in some individuals; the NK cell activation is deregulated. Therefore, increased viral load in Bx relative to AA haplogroup in eBL could be related to continuous stimulation and subsequent loss of NK cell control of EBV. A previous study reported the accumulation of dysfunctional CD56 negative CD16 positive subset of NK cells in eBL patients 9 . www.nature.com/scientificreports/ Consistent with these findings, our results raise the possibility that activation of NK cells that are mediated by KIRs may impair NK cell functions in our study population 9 . Further studies are required to confirm this observation and the role of KIR-expressing T cells [11][12][13] . A limitation of our study was the small sample size and convenience sampling of healthy controls which led to them being younger than eBL cases. However, since KIR genotypes do not change with age, we don't believe this biased our findings. In addition, our conclusion is restricted by a lack of HLA ligands data, as KIR/HLA combinations influence NK cell activity. As KIRs can vary at the gene or allele level, we only investigated the presence or absence of each KIR gene, and we could not evaluate allelic and copy number variations that can impact NK cell functions. A previous study reported decreased expression of KIR2DL1/S1 inhibitory/activation marker and increased expression of KIR3DL1 in children exposed to malaria and eBL patients relative to healthy children 9 , hence future studies will need to assess the mechanistic implications of KIR proteins and gene expression profiles in eBL pathogenesis. We acknowledge that whereas our findings suggest a possible association of the Bx haplogroup and increased number of aKIRs with EBV load and eBL pathogenesis respectively, there is a need for a larger validation cohort and functional studies to confirm their biological relevance.

Materials and methods
Study site and subjects. The study enrolled 208 children; 104 eBL patients, and 104 healthy individuals.
Patients with eBL were children aged 0-14 years old, who were enrolled at Jaramogi Oginga Odinga Teaching and Referral Hospital (JOOTRH), located in Kisumu County, western Kenya. JOOTRH is one of the two regional referral centers for childhood cancer cases in western Kenya. Morphologic diagnosis of eBL was performed by staining fine-needle aspirates (FNA) with Giemsa/May-Grünwald and observed under a microscope 57 . The control group consisted of convenience samples selected from healthy non-eBL children, aged 0-12 years old, living in the same malaria-holoendemic regions of western Kenya as the eBL patients. Convenience sampling did not affect our results since KIR genotypes do not differ by age.

DNA extraction and KIR genotyping. Genomic DNA was isolated from 200 μl of blood using Qiagen
QIAamp DNA Mini Kit (Valencia, CA, USA), following the manufacturer's instructions, and frozen at − 20 °C until genotyping. The samples were analyzed for KIR gene content 58 . A commercially available KIR genotyping sequence-specific Primers (SSP) kit (Miltenyi, Biotec, Inc, Germany), was used to test for the presence or absence of KIRs generating inhibitory signals (KIR2DL1, 2DL2, 2DL3, 2DL5A, 2DL5B, 2DL5 (A and B) 3DL1), activating KIRs (2DS1, 2DS2, 2DS3, 2DS4del, 2DS4ins, 2DS5, and 3DS1) and the framework and pseudogenes (2DL4, 3DL2, 3DL3, 2DP1 and 3DP1) following the manufacturer's recommendations. The amplified sequences were examined by electrophoresis in 2% agarose gel stained with SYBR Safe (Invitrogen, Burlington, ON, Canada) and visualized on a UV transilluminator using a gel documentation system (ChemiDoc, BioRad) for the presence or lack of amplicons specific to each gene, according to the manufacturer's instructions (See full-length gel in Supplementary Fig. 1 online).
Definitions for KIR gene content polymorphisms. KIR polymorphisms were analyzed by determining the presence or absence of 16 KIRs genes; 2 pseudogenes, 8 inhibitory, and 6 activating KIR genes. The KIR haplotypes were classified into AA (inhibitory) and Bx (activator), where x can be A or B. The homozygote AA genotype was defined by the absence of KIR2DL2, 2DL5 (2DL5A and B, 2DL5A, 2DL5B) 2DS1, 2DS2, 2DS3, 2DS5, and 3DS1 genes. Individuals in the Bx genotype contained at least one of the genes above 28 . The genes 2DL4, 3DP1, 3DL2, and 3DL3 are framework genes 23 . The genotypes AA and Bx were evaluated according to the distribution of the centromeric and telomeric genes and the B content score as previously reported 31 .
EBV Load. Epstein-Barr virus load was determined by quantitative Polymerase Chain Reaction (qPCR) 59 .
Briefly, amplification of genomic DNA was done in a Bio-Rad CFX96 Real-Time System with C1000 Thermo Cycler base for the primers and probes (BioRad Laboratories, Hercules, CA), following the manufacturer's instructions. The PCR amplification conditions were: 180 s at 95 °C, 10 s at 95 °C, 30 s at 63.5 °C-plate read, 10 s at 95 °C (39 times). Statistical analysis. The frequencies of KIR genes, haplotypes, genotypes, B score, and centromere-telomere gene content in eBL patients were compared with the healthy controls. Differences between KIR genes and EBV loads were assessed by Fisher's exact test using Graphpad Prism version 8.0.2 software (GraphPad Software, La Jolla, CA). Comparisons of log-transformed EBV load between the genotypes were performed using the Mann-Whitney test and one way Kruskal-Wallis statistic. When the p-value in the Kruskal-Wallis test was statistically significant, pair-wise comparisons were assessed by Dunn's test. Multivariable logistic regression analyses were performed in R, controlling for age and sex as variables influencing the risk of eBL etiology. The statistical significance of associations was assessed using odds ratios (OR) with 95% confidence intervals (CI). A p ≤ 0.05 was considered significant.
Ethical approval. This research was approved by the Scientific and Ethical Review Unit (SERU) at the Kenya Medical Research Institute (KEMRI), and the Institutional Review Board at the University of Massachusetts Medical School (UMMS), Worcester, USA. All experiments were performed in accordance with relevant guidelines and regulations. Study participants were informed about the study and since they were all below 18 years, the parent and/or legal guardian provided written informed consent, before enrollment. In addition, children aged 13 years and above provided assent as per the requirements of the local IRB.