Coexistence of inhibitory and activating killer-cell immunoglobulin-like receptors to the same cognate HLA-C2 and Bw4 ligands confer breast cancer risk

Human leukocyte antigen (HLA) class I-specific killer-cell immunoglobulin-like receptors (KIR) regulate natural killer (NK) cell function in eliminating malignancy. Breast cancer (BC) patients exhibit reduced NK-cytotoxicity in peripheral blood. To test the hypothesis that certain KIR-HLA combinations impairing NK-cytotoxicity predispose to BC risk, we analyzed KIR and HLA polymorphisms in 162 women with BC and 278 controls. KIR-Bx genotypes increased significantly in BC than controls (83.3% vs. 71.9%, OR 1.95), and the increase was more pronounced in advanced-cancer (OR 5.3). No difference was observed with inhibitory KIR (iKIR) and HLA-ligand combinations. The activating KIR (aKIR) and HLA-ligand combinations, 2DS1 + C2 (OR 2.98) and 3DS1 + Bw4 (OR 2.6), were significantly increased in advanced-BC. All patients with advanced-cancer carrying 2DS1 + C2 or 3DS1 + Bw4 also have their iKIR counterparts 2DL1 and 3DL1, respectively. Contrarily, the 2DL1 + C2 and 3DL1 + Bw4 pairs without their aKIR counterparts are significantly higher in controls. These data suggest that NK cells expressing iKIR to the cognate HLA-ligands in the absence of putative aKIR counterpart are instrumental in antitumor response. These data provide a new framework for improving the utility of genetic risk scores for individualized surveillance.


Discussion
Aggregation of breast cancer in families indicates a predisposing genetic component for breast cancer risk 30 .
Family studies using linkage analysis have identified several rare mutations with strong effects (i.e., highly penetrant), notably at BRCA1, BRCA2, PALB2, ATM, and CHEK2 loci, conferring lifetime risk of breast cancer 31 .
The large-scale genome-wide association studies (GWAS) have identified more than 200 susceptibility loci, each of which confers a small risk for breast cancer development 32 . However, the mechanism steering these genetic associations remains largely unknown because most variants are located in non-coding regions and are not in strong linkage disequilibrium with known protein-coding variants 33 . Moreover, GWAS includes relatively few informative SNPs in the KIR region, and therefore analysis of the KIR region has been impractical because its extraordinary structural diversity leaves few locations suitable for designing binary SNP markers 34 . Therefore, much of the KIR genetic contribution to breast cancer risk remains unknown. NK cell surveillance is an essential activity in defending tumor initiation and metastasis 35 . According to the "missing-self " hypothesis, NK cells complement T cell immunity by killing cancer cells that downregulate MHC class I molecules to escape class I-restricted T cell response 36 . Presumably, defects in NK cell number and activity play a role in breast cancer initiation and progression. Consistent with this notion, a substantial reduction of blood NK cell cytotoxicity in women with breast cancer, particularly in women with advanced-stage breast cancer, was noted compared to healthy individuals 9,10 . Individuals with high incidences of familial breast cancer exhibit significantly reduced NK cell cytotoxicity in peripheral blood 37 . The advanced breast cancer patients have an increased proportion of more immature and less cytotoxic CD56 bright CD16 +/− NK cell subset in their peripheral blood, which might account for at least part of the reduced levels of cytotoxic functions observed in these patients 38 . The molecular mechanism underlying the impaired NK cell cytotoxicity and antitumor effect in breast cancer is not identified.
The interaction of inhibitory KIR with specific cognate HLA class I ligand makes NK cells matured to acquire full effector function, developmental programming termed "licensing" 16,17 . In the absence of inhibitory KIR-HLA interactions, NK cells became hyporesponsive or anergic. The distribution of four inhibitory KIR-HLA class I ligand combinations is comparable between breast cancer patients and controls, indicating that the development of functionally active NK cells in patients might be similar to those of controls. However, NK licensing is not entirely permanent, and the functional activity of mature NK cells can be reset by new HLA environment in tumor tissue with reduced HLA class I expression 39 , a mechanism that tumor develops to evade from adoptive immune response 40 .
The activating KIR-based HLA class I-dependent licensing may also influence NK cell unresponsiveness to transformed cells. Although the mechanisms have not yet been identified, education by activating KIRs shares features with the hyporesponsiveness induced by chronic stimulation of other activating receptors expressed by NK cells. For example, chronic exposure to NKG2D ligands in mice renders NK cells hyporesponsive to target cells 41 . A recent study found that the expression of NKG2D on blood NK cells was higher in breast cancer patients than the levels documented in healthy females 10 . Similarly, when the ligand (m157) for the activating Ly49H is constitutively expressed, mouse Ly49H + NK cells become hyporesponsive 42 . NK cells expressing activating KIR2DS1 are hyporesponsive in the presence of self-HLA-C2 ligands 43 and thus unable to mount an efficient response against breast cancer. This disarming model of licensing emphasizes the crucial role of activating KIRs 2DS1 and 3DS1 and interactions with their HLA class I ligands in developing anergic NK cells (disarmed), which are hyporesponsive and are not able to defend against the tumor. All patients with advanced cancer carrying 2DS1 + C2 or 3DS1 + Bw4 also had their inhibitory KIR counterparts 2DL1 and 3DL1, respectively. In contrast, the 2DL1 + C2 and 3DL1 + Bw4 pairs without their activating KIR counterparts were significantly higher in controls than patients. These data suggest that NK cells expressing inhibitory KIR to the cognate HLA class I ligands in the absence of putative activating KIR counterpart are instrumental in antitumor response.
It also remains possible that activating KIR receptors could recognize altered HLA class I complexes, e.g., specific HLA/peptide complexes 44 . It is possible that the activating KIRs could directly bind neoantigens explicitly expressed on breast cancer cells, which may suppress cytolytic function but trigger cytokine release. Supporting this possibility, KIR2DS4 has been suggested binding to an unidentified protein expressed on melanoma-derived tumor cells, independently of HLA class I 45 . Particularly, KIR2DS1 has been shown to displays a certain degree of peptide selectivity in its binding to HLA class I 46 , indicating that the functional outcome of activating KIRs can be modulated by the nature of the presented peptide.
The human KIR genotypes can be simply divided into two groups, AA and Bx, with quantitively and qualitatively contrasting KIR gene content 15 . We found a striking association of Bx genotypes with breast cancer. Three of four prior studies suggested an association between breast cancer and B-haplotype-specific KIR genes [47][48][49][50] . Consistent with our findings, Jobim et al. found a strong association between KIR2DL2 and Brazilian women with breast cancer 48 . However, this study could not find any association with other KIR genes, such as KIR2DS2, 2DS3, and 2DL5, which are located at close proximity to 2DL2 with strong linkage disequilibrium. Jobim et al. also reported an association between KIR2DL2 + C1 in breast cancer, which was not observed in our study. Oztruk et al. reported a strong association between KIR2DS1 and patients with breast cancer in Turkey 49 , which is in agreement with our findings. However, they could not find an association with KIR2DS1-linked genes, such as 3DS1 and 2DS5. www.nature.com/scientificreports/ Using a new cohort of breast cancer patients and controls from the Fars province, our collaborator Prof. Abbas Ghaderi and team recently reported an association between Iranian breast cancer patients and Bx KIR genotypes, centromeric Bx genotypes, and B-haplotypes carrying C4T4 motif (positive for seven KIRs: 2DL2, 2DS2, 2DS3, 2DL5, 3DL1, 2DS1, and 2DS3) 50 , which is in agreement with our findings. However, the study confirmed the association of individual B-haplotype-associated KIR genes only with breast cancers expressing estrogen receptors. Breast cancer positive for progesterone receptor or human epidermal growth factor 2 (HER2) were not associated with B-haplotype-specific KIRs. Moreover, the HLA class I ligands were not analyzed in this study. In total contradiction to our findings, Alomar et al. reported a significant decrease in the frequencies of KIR2DS2, 2DS3, and Bx genotypes in 50 Saudi women with breast cancer compared to 65 controls 47 . HLA ligands were analyzed by only Turkish and Saudi studies. Inconsistent results observed between the studies are presumably contributed by multiple factors, including ethnic and population disparity in KIR and HLA genome, the differential composition of histologic breast cancer phenotypes, and small sample sizes.
The B-haplotype KIRs were correlated with an increased risk of other solid and hematological malignancies, including leukemia 51 , cervical neoplasia 52 , Hodgkin lymphoma 53 , gastric cancer 54 , head and neck squamous cell carcinoma 55 , urothelial bladder cancer 56 , colorectal adenocarcinoma 57 , systemic sclerosis 58 , and meningioma 59 . The B haplotype-specific KIRs, particularly those located at the telomeric half (3DS1, 2DS1, and 2DS5), were observed to be prominently increased in patients with an advanced stage of breast cancer. These results contrast with the classical view that activating NK cell receptors mediate spontaneous lysis of transformed cells and protect against the tumor 60 .
Given our study's retrospective design, further investigations are warranted in a prospectively accrued patient population to substantiate our findings. The number of patients included in our study was insufficient to evaluate the impact of KIR-HLA combinations in tumors with different phenotypes, such as estrogen receptor-positive (ER +), progesterone receptor-positive (PgR +), and HER2 + . Therefore, further systematic studies should be focused on determining the impact of combined KIR + HLA combinations using multivariate analysis. The limitation of our KIR-binding HLA epitope typing is its inability to discriminate HLA allotypes (e.g., Cw*05:01, Cw*02:02), which can differ in binding affinity 61 . In summary, our results provide a genetic basis for impaired NK cell antitumor activity in breast cancer. The KIR-HLA associations observed in this study provide further insight into genetic susceptibility to breast cancer, improving the utility of genetic risk scores for individualized screening and follow-up recommendations for earlier implementation of breast cancer risk-reduction strategies. Moreover, our results suggest that autologous activated NK cell clones with select KIR-HLA composition favoring antitumor activity could be a promising immunotherapeutic strategy against breast cancer.

Materials and methods
Study subjects and samples. A cohort of 162 women with breast cancer and 278 healthy controls from the southern part of Iran (Fars province) were included in this study. The patients were recruited at Faghih hospital, Shiraz University of Medical Sciences. The age-matched controls were collected from the same geographical area. The clinical and pathological characteristics were collected from patient medical records. Table 3 shows the distribution of clinicopathological characteristics of breast cancer. The breast cancer patients were categorized according to TNM staging 62 , and grouped as either early stage (0, I, and II) or advanced stage (III and IV) of disease. The study was reviewed and approved by the Medical Research Ethics Committee of Shiraz University of Medical Sciences and UCLA Institutional Review Board of human research protection. Genomic DNA was extracted from peripheral blood samples using either the standard salting-out method or by QIAamp blood kit (Qiagen, Hilden, Germany). The quality and quantity of DNA were determined by UV spectrophotometry, and the concentration was adjusted to 100 ng/μL. All DNA samples received at UCLA were de-identified and only marked as having been obtained from patients with breast cancer or controls. Informed consent was obtained from all subjects. Data obtained were Health Insurance Portability and Accountability Act (HIPAA) compliant, and the study adhered to the tenets of the Declaration of Helsinki. All methods were carried out in accordance with relevant guidelines and regulations.
KIR genotyping and genotype/haplogroup classification. The presence and absence of 16 KIR genes were determined using our previously developed duplex SSP-PCR typing method 63 . Ambiguous and unusual KIR genotypes were resolved by using the alternative SSP-PCR typing method 27 . Based on the presence and absence of KIR genes, we divided the study subjects into two groups: the AA and Bx genotype carriers. The AA genotype subjects carried only KIR3DL3-2DL3-2DL1-2DP1-3DP1-2DL4-3DL1-2DS4-3DL2 genes that are characteristic of A-haplotype. The rest were regarded as Bx genotype carriers (AB heterozygous and BB homozygous carriers). Based on our previous linkage disequilibrium analyses, we determined the frequency of B-haplotype-specific KIR gene clusters 64,65 . One of them comprises KIR2DS2-2DL2-2DS3-2DL5B genes and is located at the centromeric half of the KIR gene complex (termed C4 linkage group). In contrast, another cluster contains KIR3DS1-2DL5A-2DS5-2DS1 genes and is located at the telomeric half of the complex (termed T4 linkage group).
HLA class I ligand typing by novel direct sequencing. We developed a novel direct DNA sequencing method to determine KIR-binding HLA-A, -B, and -C ligands. The procedure starts with gene-specific amplification of exon 2 and 3 of HLA-A, -B, and -C loci followed by direct sequencing of PCR amplicons (Suppl. Fig. 1, Suppl. Table 2). The primers amplify all common and well-documented HLA-A, -B, and -C alleles 66 . However, the HLA-B amplification excludes HLA-B*73:01. The reverse primers used in the HLA-B specific amplification (3BIn3-37R) binds to the intron-3 region from nucleotide 1028 to 1050. Since the HLA-B*73:01 allele has mutations at reverse primer annealing site at nucleotide 1032 from G to A and 1038 from G to C, the HLA-B*73:01 is not be amplified by HLA-B PCR. However, HLA-B*73:01 is amplified by HLA-C specific amplification. Gene- The PCR amplicons were purified from unincorporated primers and dNTPs by digesting with ExoSAP-IT exonuclease-I (USB Corporation, Cleveland, OH) according to the manufacturer's protocol and were used as a template in the sequencing reactions. Then the segments of exon 2 that encode the KIR-ligands were sequenced at both directions using the BigDye terminator V1.1 cycle sequencing kit (Applied Biosystems, Foster City, CA). Sequencing reactions (10 μL) comprise 2 μL of sequencing reagent premix, 1 μL dilution buffer, 0.3 μL of sequencing primer (10 pM/μL), and 2 μL of purified PCR amplicon. The following PCR thermal cycling profile was used: 25 cycles of 96 °C for 20 s, 53 °C for 20 s, 60 °C for 1 min, and soak at 4 °C. Once the cycling was completed, the sequencing reactions were precipitated using sodium acetate/EDTA buffer and ethanol to concentrate the reactions and to eliminate unincorporated fluorescent-labeled nucleotides. The precipitates were resuspended in 15 µL of Hi-Di deionized formamide (Applied Biosystems), denatured by heating at 95 °C for 2 min, and loaded into the ABI PRISIM TM 310 capillary sequencer (Applied Biosystems). Finally, sequence analysis was performed using Assign SBT v3.5.1 software (Conexio Genomics, Western Australia), which can combine both forward and reverse sequences files to inspect and edit the electropherograms. The Assign program assigned the alleles by comparing the test sequences with a library of known HLA-A, -B, and -C sequences downloaded from the international ImMunoGeneTics (IMGT-HLA Database (http:// www. ebi. ac. uk/ imgt/ hla). The KIR-binding HLA class I ligands were deduced from the assigned alleles. We have validated this method by using a panel of 31 UCLA DNA standards that includes most core HLA class I types (Suppl. Table 3).
Data analysis and statistical methods. The percentage of each KIR gene in control and patient groups was determined by direct counting (individuals positive for the gene divided by individuals tested per popula- www.nature.com/scientificreports/ tion × 100). Differences between the study groups in the distribution of each KIR genotypes, KIR genes, HLA ligands, and KIR-HLA combinations were estimated by the two-tailed Fisher Exact probability (P) test, and p < 0.05 was considered to be statistically significant. Odds ratio (OR) and 95% Confidence Intervals (CI) were calculated to determine the magnitude and statistical significance of associations 67 .

Data availability
The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.