Association of KIR Genes and MHC Class I Ligands with Atopic Dermatitis.

The primary source of subjects for this study was the Genetics of Atopic Dermatitis (GAD) cohort and details of this study were recently published.( 28 , 29 ) All subjects were examined and diagnosed by dermatologists with expertise in the diagnosis of AD. All subjects had a history and an exam consistent with AD (cases) or no history of AD by history and examination (controls). Subjects were enrolled between 2015 and 2020. The informed consents and/or assents were approved by their appropriate Institutional Review Board. DNA was collected using Oragene DNA collection kits (DNA Genotek, Ottawa Canada) as previously reported.( 30 – 33 ) The three classic HLA Class I genes (-A, -B, and -C) were sequenced using targeted amplicon-based NGS with Omixon Holotype HLA™ V2 kits (Budapest, Hungary) using Illumina MiSeq (San Diego, CA) using paired-end 2×150 V2 chemistry. Omixon Twin™ (7,000 pairs/locus, v, 2.5.1) and GenDx NGSengine® (300,000 pairs/sample) analyzed each set of Fastq files.( 31 – 33 ) We focused on HLA ligands known to interact with KIR. The HLA-B ligand contact residues for the Bw4 epitope is at locations 77, 80, 81, 82, and 83 and are best defined as 80 (amino acids I or T) and 83(R).( 26 , 34 ) HLA-B leader sequence is defined at location −21 (amino acids T or M) of HLA-B. The HLA-C epitope that interacts with KIR is at position 80 and is defined by amino acids N and K (i.e., C1/C2 KIR binding site).( 4 , 6 , 7 , 34 ) Residue locations were based on IMGT data.( 23 , 25 ) The presence/absence of each KIR gene was determined using Illumina NGS platform with a total of 64 amplicons designed specifically for 15 KIR genes by 6 reactions of gene-specific or group-specific multiplex PCR, consisting of three to six amplicons for each gene. The typing was conducted by HistoGenetics (Ossining, NY). Illumina NGS data was then analyzed by a Multiplex KIR typing algorithm for the classic KIR genes developed by HistoGenetics using IMGT KIR database v2.6.0. The results per gene were confirmed in at least two amplicons. The prevalence of KIR specific HLA Class I epitopes and KIR genes was estimated by subject and presented with 95% confidence intervals (CI). These parameters were estimated for those with and without AD and by self-reported ancestry (European or African). We have previously reported in a group of United States children with AD that self-reported ancestry is highly correlated with genetically determined ancestry.( 30 ) The odds ratio (OR) of having AD was estimated using logistic regression assuming an additive model. Analyses of KIR genes are complicated and, in prior studies, have been inconsistently presented.( 2 , 22 , 35 , 36 ) Overall, we analyzed the association of the KIR gene, the association of HLA Class I amino acid variants known to interact with suspected KIR, association of the HLA B −21 leader sequence with respect to AD, and the association of known interactions between KIR and HLA ligand pairs with respect to AD.( 4 , 36 , 37 ) A previous study noted that KIR2DS5 might interact with C2 and we accepted this for our study.( 37 ) Models were not adjusted for other atopic illnesses like asthma, seasonal allergies, or food allergies because, as previously noted by studies of the “atopic march”, these illnesses are likely on the same causal pathway.( 38 ) P -values were not corrected in view of the fact that our hypotheses were determined a priori and were not agnostic. Our primary study question was whether KIR genes were associated with AD based on the knowledge that KIR genes affect the function of NK cells. We also evaluated gene-gene interaction (epistasis), which can complicate alpha control of type I error. We present point estimates (i.e., odds ratios) and 95% confidence intervals as evidence for the association of KIR genes and AD. All statistical analyses were conducted using Stata Version 16.1 (College Station, TX). We previously reported on the Pediatric Elective Eczema Registry (PEER). Study subjects from PEER (PEER; www.thepeerprogram.com ) were enrolled as a United States nationwide cohort of more than 8,000 subjects with pediatric-onset AD.( 30 ) Here, we analyzed a sub-cohort of PEER children, who provided a genetic sample (PEER DNA cohort).( 30 , 33 , 39 ) PEER DNA enrollment occurred between November of 2004 and January 2015. At the time of enrollment, children were two to seventeen years old, had a physician-confirmed diagnosis of AD, and, prior to enrollment, had used pimecrolimus cream for at least six months, but may and did use other therapies.( 30 ) With respect to outcome measures (e.g., the persistence of AD), subjects are followed for up to 10 years and during that time were not required to (and most did not) continue therapy with pimecrolimus.( 40 ) The primary PEER outcome was determined every 6-months to determine if a child had persistent AD based on the self-reported outcome of whether or not a child’s skin was AD symptom-free during the previous six-months without need for medication.( 30 ) This outcome response has been validated and shown to correlate with other tools used to evaluate symptom control and is likely a marker of long-term disease severity.( 41 – 43 ) As compared to a case-control study, which is an ideal design to evaluate the association of a variation with a disease, a longitudinal cohort study is ideal to study disease course. We used this cohort for a secondary study, in which the presence of KIR genes was typed in 274 PEER subjects (35% of the PEER cohort). The sub-cohort included 81 of self-reported African Ancestry and 193 of self-reported European Ancestry subjects. We hypothesized that variants or composites shown to be associated with AD in the GAD cohort would be associated with more persistent AD in the PEER cohort and, therefore, a priori decided to only evaluate the GAD significant findings in the PEER based secondary study.

HLA genotyping was conducted in 849 individuals (385 controls and 464 cases) from the GAD cohort. The individuals with AD (cases) included 232 (50.0%) individuals of European ancestry, 203 (43.8%) individuals of African ancestry, 31(6.7%) individuals of other ancestries, and 303 (65.4%) were women. The controls included 271 (70.4%) individuals of European ancestry, 112 (29.1%) individuals of African ancestry, 2 (0.5%), individuals of other ancestries, and 174 (45.2%) were women. Seasonal allergies were noted in 16.8% of the controls and 64.4% of the cases. Asthma was reported in 12.8% of the controls and 56.4% of the cases. Based on the patient-oriented eczema measure (POEM), 71.5% of the individuals with AD had moderate to very severe AD (mean group score was 12.0 (CI: 11.3,12.7)).( 44 ) On the day of enrollment, 95.1% classified themselves as having active AD and 88.2% were using topical prescription medication to treat their AD (66.5% noted their use was frequent or continuous treatment). Per case and control status, the demographic and non-medical subject-level characteristics were essentially identical in the study cohort to the full GAD (data not presented). Sufficient DNA for KIR genotyping was available for analysis from 753 individuals in the GAD cohort (88.7% of the total) including 409 individuals with AD (cases) and 344 controls ( Table 1 ). The prevalence of HLA Class I KIR associated epitopes as well as the KIR genes are presented in Table 1 . Based on HLA-B frequencies, the Bw6 epitope (B80N) was more common than the Bw4 (B80T or B80I) epitope as was the C1 epitope (C80N) than the C2 epitope (C80K).( 2 , 4 , 5 ) We found that those of African ancestry were less likely to have B80T residue of Bw4 epitope than those of European ancestry (OR= 1.89 (95% CI=1.53,2.42)) and more likely to have B80I residue (0.61 (0.48,0.78)). For the HLA-C C1 (C80N) epitope, those of African ancestry were less likely to have the C1 epitope than those of European ancestry (0.65 (0.54,0.78)) and vice versa for the C2 epitope. Most KIR gene frequencies were very similar between ancestries ( Table 1 ). KIR3DS1 and KIR2DS1 were less common among those of African ancestry than those of European ancestry (0.36(0.26,0.52)) and 0.53 (0.381,0.74)) and the presence of these genes are in LD=0.70. KIR3DL3, KIR3DP1, KIR2DL4 , and KIR3DL2 as well as the framework KIR genes were noted in all subjects. Overall, HLA Class I KIR associated ligands seemed to have little influence on the likelihood of having AD, with one exception ( Table 2 ). The Bw6 epitope, not known to interact with KIR, was associated with a decreased risk of having AD in those of African ancestry ( Table 2 ). Several KIR genes were associated with AD and the effect size was mostly consistent between ancestries ( Table 2 ). An increased risk of AD was noted for KIR2DL5 (1.51(1.13,2.01)), KIR2DS5 (1.72(1.26,2.34)), and KIR 2DS1 (1.41 (1.04,1.91)). These KIR genes are associated with the KIR haplotype B. The genes KIR2DS5 and KIR2DS1 are in LD (r 2 =0.60). The effect estimates for the presence of these three KIR genes varied by race ( Table 2 ). The known interactions between HLA Class I epitopes and KIR genes with respect to the risk of AD are presented in Table 3 . Individuals with KIR2DS5 and the C2 epitope are at an increased risk of AD. This finding is significant in the full cohort as well as separately in those of European and African ancestry. Those of African ancestry with KIR2DS1 and the HLA C2 ligand are also at an increased risk of AD ( Table 3 ). The B-21T (TT) leader sequence as a composite with KIR genes also increased the risk of AD with respect to KIR2DL3 , KIR2DL5 , KIR2DS1 , and KIR2DS5 in the full cohort ( Table 4 ). An increased risk of AD was noted among those of European Ancestry for KIR2DL3 , KIR2DL5 , KIR2DS1 , KIR2DS5 , and KIR3DS1. An increased risk of AD was noted for KIR2DL5 in association with B-21T among those of African ancestry. Composites of KIR genes and the HLA-B*−21M (MM or MT) were not associated with risk of AD ( Table 4 ). In addition, the Bw6 protective effect was more pronounced amongst those who also had the B-21M leader sequence (0.68(0.52,0.90)). KIR gene typing was conducted in about one-third of the PEER cohort. The purpose of this evaluation was to see if the presence of KIR genes or KIR gene HLA Class I composites as noted above are associated with individuals with more persistent AD (odds ratio of <1.0). Overall, the KIR genes and KIR/HLA composites associated with AD in the GAD cohort were also associated with higher rates of persistence of AD and four of these variants were statistically associated with more persistent AD ( Table 5 ).

This case-control study of individuals with and without AD focused on NK cell KIR and HLA Class I associated epitopes. HLA Class I associated epitopes were determined using NGS of HLA Class I region and, unlike many previous studies, the HLA genotyping was not imputed from GWAS or based on immunophenotyping.( 45 – 48 ) The overall frequency of the HLA Bw4 epitope as well as the frequency of the HLA-C C1/C2 epitopes was consistent with previous reports.( 22 , 49 ) These HLA Class I epitope frequencies varied little by ancestry. The African American frequency of Bw4 and C1 was similar to Mbuti people of Gabon, a region consistent with African American ancestry.( 49 ) By ancestry, KIR gene frequency only varied with respect to KIR3DS1 . Individuals with Bw6, which is not a KIR associated HLA-B ligand, who were of African ancestry were less likely to have AD. Individuals with KIR2DS1 , KIR2DL5, and KIR2DS5 , all part of KIR B haplotype, were more likely to have AD. Two of these genes are thought to be activating (S) and one inhibitory (L). The magnitude of this effect was more pronounced in those of African ancestry and increased in the presence of the KIR gene appropriate HLA Class I epitope (C2). Finally, individuals with less well-educated NK cells (HLA-B −21TT leader sequence) were more likely to have AD. To the best of our knowledge, these findings have not been previously reported. The HLA-B leader sequence influences the education of NK cells through interactions with HLA-E.( 4 ) The presence of the epitope B-21M results in more diverse and functionally potent NK cells.( 4 ) With respect to AD risk, B-21T as compared to B-21M, has a greater effect on these genes. When occurring with B-21T, KIR2DL3 , KIR2DL5 , KIR2DS1 , and KIR2DS5 were associated with an increased risk of AD in the full cohort. The interaction between the HLA-B*−21T leader sequence and these KIR genes was also noted in those of European ancestry. The increased risk of AD in the presence of HLA-B*−21T is consistent with and corroborates a recent study demonstrating that the circulating fraction of immature NK cells was increased in those with moderate to severe AD as compared to a control population.( 9 ) Our findings are also consistent with a recent study that demonstrated an imbalance of NK cell phenotypes that favored NK cells with less educated and more inhibitory phenotypes in lesional versus non-lesional or resolving AD lesions.( 15 ) In our study, KIR2DL5, KIR2DS5, and KIR2DS1 were associated with an increased risk of AD. A previous study reported a decreased association between KIR2DS1 and AD in those of European ancestry.( 10 ) Like our study, the previous study focused on the presence of KIR genes, but differed in that we used high resolution HLA and KIR typing technique.( 10 ) The control frequency for the KIR genes was very similar between studies with the exception of KIR2DS1 .( 10 ) Interestingly, the frequency of KIR2DS1 in our European American controls was smaller than in the previous study.( 10 ) In the previous study there was no change as would have been expected in the protective effect of KIR2DS1 in the presence of its known associated HLA ligand, the C2 epitope.( 10 ) The KIR2DS5 gene, the only KIR associated with an increased risk of AD in our study in both European and African Americans, is also associated with decreased risk of preeclampsia and endometriosis.( 37 , 50 ) KIR2DS5 has not been specifically associated with the C1 or C2 epitope.( 2 , 3 ) However, in the studies of preeclampsia and endometriosis, it was shown to have a greater effect in the presence of C2 epitope.( 37 , 51 ) Recent studies have confirmed that the likely ancestral allelic variant KIR2DS5*005 does bind to C2.( 20 ) However, KIR2DS5*002 , a very common European variant, does not bind the C2 epitope but KIR2DS5*006 , a common variant in Africans, binds C2 and is protective of preeclampsia.( 20 ) KIR2DS5 in the presence of the C2 epitope was also noted in our study to increase the risk of AD.( 37 , 50 ) Another study has shown that different haplotype B KIR genes are likely protective of preeclampsia in Europeans.( 51 ) Of interest, based on secondary analyses and limited to childhood eczema, AD may be associated with pre-eclampsia and endometriosis.( 52 , 53 ) A direct mechanism regarding how the KIR2DS5 gene increases the risk of AD is not known. It is possible that AD risk is increased due to the presence of these KIR genes in the setting of less-educated NK cells, perhaps leading to fewer NK cells.( 9 ) Alterations in NK cell function have also been associated with increased risk for bacterial (e.g., staphylococcus and streptococcus) and viral illnesses (e.g., herpes simplex) that are also associated with AD.( 19 , 36 ) NK cell function has been associated with autoimmune illness and evidence also exists that AD may have a component that is consistent with autoimmune mechanisms.( 54 – 58 ) Thus, the precise manner of how NK cell regulation by KIRs influences the immune response in AD remains to be shown. Furthermore, reported interactions between KIR genes and HLA ligands are evolving.( 21 , 36 ) For example, one of the most extensively studied KIR genes is KIR3DL1 , which has been shown to be associated with milder course of HIV.( 6 , 59 ) KIR3DL1 was originally noted to interact with Bw4 but later noted to interact most strongly with the Bw4 B80T subgroup.( 3 ) More recent studies using KIR allele genotyping have shown that the reactivity of KIR3DL1 alleles (e.g. *001/*005/*015) to Bw4 epitopes (e.g. B80T and B80I) varies greatly resulting in variable inhibition of NK cell function.( 60 ) Specifically, of the KIR3DL1 alleles, the KIR3DL1*005 allele is the least sensitive allele to the Bw4 ligand.( 60 ) Recent studies have also shown that KIR3DL1 preferentially recognizes HLA-B*57:01 over HLA-B*57:03 , yet these two alleles share more than 99% homology and are both part of the Bw4 epitope.( 35 , 60 ) Clinically, individuals with HIV infection and KIR3DL1 ~ HLA-B*57:01 have the best prognosis.( 35 , 59 , 61 ) A limitation of our study is that we were only able to type for the presence or absence of KIR genes. Future studies will need to directly evaluate NK cell surface expression levels of HLA Class I ligands or KIR gene alleles. As well summarized by Aiello et al , the analytic methods used to show genetic associations in humans between KIR and human illnesses have varied and have included reports of unadjusted associations, associations after adjustment, conditioning on HLA ligand variation, and/or associations based on epistasis between KIR genes and the assumed HLA ligand.( 36 ) Studies seldom use more than one statistical analytic method.( 36 ) As can be seen in our study, due to the heterogenous nature of an individual’s genome and the lack of a full understanding of the interactions between KIR variation, HLA Class I variation, and other genetic and epigenetic regions, person-level associations between KIR and HLA Class I are somewhat dependent on the method used. A complex attempt was recently published to rectify this issue.( 22 ) These authors evaluated families and full sequencing of KIR genes in order to properly determine KIR haplotypes for proper analyses of these genes.( 22 ) This approach is outside of the scope of the GAD cohort and is a limitation of our study. In summary, we found that individuals with KIR2DS1 , KIR2DL5 and KIR2DS5 were more likely to have AD. In general, the presence of KIR genes from the B haplotype tended to favor a diagnosis of AD, especially in those of African ancestry. In individuals with these KIR genes and the HLA-B*−21T leader sequence, there was also an increased risk of AD. Unlike filaggrin , the most commonly reported gene to be associated with AD which primarily is found in Europeans, it appears that the risks due to KIR variation is primarily noted in African Americans.( 30 , 62 , 63 ) In this study, we did not measure allelic variation within KIR genes, so it is possible that KIR gene allelic variation may behave differently than the presence of the KIR gene.( 26 , 35 ) However, overall our findings add to the growing literature that indicate an important association between NK cell function, circulating numbers of NK cells, and AD.( 9 , 11 – 14 ) Future AD treatments and even improved understanding of the mechanism of action of current AD treatments could benefit from a better understanding of the impact that NK cells have on AD. Cancer immunotherapy has focused on the therapeutic targeting of NK cells.( 64 ) The targeted pathways include interleukins 2 and 15 pathway, which downstream is associated with JAK kinases.( 64 , 65 ) JAK kinase inhibitors are currently under investigation for the treatment of AD.( 66 , 67 ) NKG2D is also an important activating receptor for NK cells.( 64 ) Studies in mice have recently shown that inhibition of this target using bortezomib alleviated atopic dermatitis.( 68 ) In order to more fully understand the association of NK cell function and AD, future studies will need to focus on KIR gene allelic variation as well as consider cell-based measurements of KIR and polymorphisms with the HLA Class I epitopes.( 61 ) These studies were beyond the scope of this report and are important for future research.

Atopic dermatitis (AD) is a chronic illness that is caused by skin barrier dysfunction and immune dysregulation. Abnormalities of both the innate and adaptive immune systems have been associated with AD.( 1 ) Natural Killer (NK) cells are classically considered innate lymphocytes.( 2 – 5 ) Unlike cytotoxic T-cells, NK cells do not express T-cell receptors that recognize antigen bound to MHC Class I receptor. NK cells directly interact with polymorphic MHC Class I ligand variants via NK cell immunoglobulin-like receptors (KIRs).( 2 , 4 ) NK cells also interact through their conserved receptor CD94/NKG2A with HLA-E in complex with HLA-class I leader sequence peptides. ( 4 – 7 ) NK cell function is most frequently associated with the identification and removal of virally infected cells and malignant clones via cell-cell interactions.( 1 , 5 , 7 ) NK cell function, usually represented by decreased circulating numbers of NK cells or decreased activity of circulating NK cells, has been associated with moderate to severe AD.( 1 , 8 – 13 ) In mouse models, increased numbers of NK cells that poorly function in the skin were associated with increased risk of disseminated vaccinia infection, an AD associated comorbidity which is also seen in humans.( 14 ) A recent study by Mack et al further confirmed the relationship between NK cells and AD by revealing that the number of circulating NK cells in adults with moderate to severe AD was significantly less than in a healthy control population.( 9 ) An imbalance in activated and resting NK cells has also been documented in lesional skin of AD subjects.( 15 ) Additionally, a study in Poland demonstrated that the activating KIR gene KIR2DS1 was associated with a decreased risk of AD.( 10 , 16 ) Finally, we have recently shown that, Bw4, an HLA class I epitope that interacts with KIR genes, Bw4, is associated with AD prognosis.( 17 ) Fifteen KIR genes are found on chromosome 19 and have activating (usually noted by “S” for short tail, e.g., KIR2D S 1 ) or inhibitory function (usually as noted by an “L” for long tail, e.g., KIR3D L 1 ) for NK cells.( 2 , 6 , 18 , 19 ) KIR genes are often classified in many ways including: part of a haplotype (A or B), by their location on chromosome 19, by the specific KIR gene, and by specific KIR alleles.( 2 , 20 , 21 ) Four genes are framework genes: 3DL3 (centromeric end), 3DP1 (middle), 2DL4 (middle), and 3DL2 (telomeric end).( 22 ) Two KIR loci, KIR2DP1 and KIR3DP1 , are classified as pseudogenes .( 2 ) KIR haplotype A is defined by the presence or absences of one or more genes that encode inhibiting KIR and can include KIR2DL1, KIR2DL3 , KIR3DL1, and KIR2DS4 .( 23 ) In this haplotype, KIR2DS4 could be an activating KIR gene but most of the alleles of KIR2DS4 have an exonic deletion, rendering the protein non-functional.( 24 ) KIR haplotype B is defined by the presence or absences of one or more genes that encode activating KIR and can include KIR2DL5, KIR2DS, KIR2DS2, KIR2DS3, KIR2DS5, and KIR3DS1 .( 23 ) KIR3DL1 and KIR3DS1 are likely not distinct genes but allelic variants of the same KIR gene.( 2 ) The location (telomere or centromere region) for KIR2DL5 , KIR2DS5 , and KIR2DS3 is dependent on allelic variation.( 20 , 21 ) The two KIR haplotypes evolved by balancing selection.( 24 ) Most mature NK cells express KIR and often express both inhibitory and activating ligands.( 2 ) KIRs recognize specific HLA Class I amino acid variants.( 7 , 25 ) The HLA-B ligand variants can interact with KIRs as part of the Bw4 epitope, which is best described by residue variation in HLA-B amino acid positions 77 to 83, but essentially variation in position 80 amino acids (B80T (HLA-B, position 80, amino acid Threonine) or B80I) and position 83 amino acids (B83R).( 26 ) B80I most strongly binds with KIR.( 2 ) HLA-C ligand and KIR interactions are defined by residue variation in position C80.( 4 , 26 , 27 ) The residue variation in this position is defined by the C1 (C80N) and C2 (C80K) epitopes. All HLA-C Class I carry either a C1 or C2 KIR binding site, while not all HLA-B have KIR associated ligands (i.e., HLA-Bw6 epitope, B80N, and B83G does not bind KIR). Finally, the strength of the interaction between KIR genes and the relevant HLA Class I ligand is dependent on allelic variation within a KIR gene.( 21 ) HLA Class I variation as represented by HLA-B dimorphism in the −21 position of the leader peptide appears to educate or modulate the function of NK cells and this activity is related to the form of the dimorphism (i.e., homozygote or heterozygote).( 4 ) The HLA-B*−21M residue appears to be more active than the ancestral T residue and as a result this locus is often considered by any presence of M residue (i.e. M/M and M/T versus T/T).( 4 ) HLA-C1 and Bw6 epitope each form the common ancestral haplotype with HLA-B*−21M.( 4 , 5 ) Individuals with HLA-B*−21M rarely encode KIR genes that interact with the Bw4 and C2 ligands.( 19 ) Given these underlying biological mechanisms, the purpose of this study was to evaluate the potential influence of NK cells on AD by evaluating variation in HLA Class I KIR-specific ligands as well as the presence or absence of specific KIR genes in patients with AD.