Introduction

Atopic dermatitis (AD) is a common, pruritic, inflammatory skin disease characterized by life-long periods of acute disease flares as well as remissions that has been associated with genetic variation1,2,3,4,5,6. AD is associated with T-cell dysregulation7,8,9,10. Several recent studies, have associated AD with altered numbers of natural killer (NK) cells and diminished NK cell function11,12,13,14,15,16. NK cell function is directly associated with Human Leukocyte Antigen (HLA) Class I and NK cell surface receptors like the killer cell Ig-like receptor family (KIR) as well as other NK cell surface receptors17,18,19,20,21,22. NK cells are lymphocytes that have classically been described as part of the innate immune system but NK cells also have adaptive immune properties18,23,24,25. NK cell receptors recognize highly conserved nonpolymorphic HLA class I leader sequence sites as well as polymorphic HLA Class I ligands24,25,26,27. NK cell functions include the identification and removal of virally infected cells and malignant clones12,25,27.

The NK cell membrane bound KIR ligands are the primary regulator of NK cell function18,20,21. There are up to fifteen KIR genes found on chromosome 19 that code for the KIR ligands that are simply thought to have either activating (usually denoted by “S” for short tail, e.g., KIR2DS1) or inhibitory functions (usually denoted by “L” for long tail, e.g., KIR2DL5) for NK cells18,26,28,29. Most mature NK cells express KIR, although the total number of KIR genes observed and expressed as well as the type of KIR ligands produced by the genes (i.e., inhibitory and activating ligands) varies by individual18,20,21. We have recently shown that the KIR genes KIR2DL5, KIR2DS5, and KIR2DS1 are more frequently found in individuals with AD22. This is also true for the HLA-C2 epitope and HLA-B*-21T leader sequence22.

The goal of this study was to examine allelic variation in KIR genes KIR2DL5, KIR2DS5, and KIR2DS1 that were previously associated with AD22,30,31. This investigation is important in that the pathophysiology of AD has not been fully elucidated and dysregulation of immune function is being actively investigated in the quest for new therapeutics.

Results

Study cohort

We focused on the allelic variation of the three KIR genes previously shown to be associated with AD; KIR2DL5, KIR2DS5, and KIR2DS122. Sufficient DNA for KIR typing was available from 506 Genetic of Atopic Dermatology (GAD) cohort participants. The KIR typing was successful for 489 (96.6%) subjects. Frequencies for all KIR gene alleles are provided in the supplement. Among those with full allelic typing, there were 313 (64.1%) AD cases and 176 (35.9%) controls. The AD cases included 167 (53.0%) Whites and 125 (40.0%) Blacks. The controls included 114 (65.1%) Whites and 59 (33.7%) Blacks. Of those with AD, 175 (56.6%) had asthma and 194 (62.8%) had seasonal allergies (4 individuals were missing this information). The median age of onset of AD was 0.75 years (IQR: 0.25–7.0). On the day of enrollment, about 25% of the subjects in the GAD case group had moderate to severe AD based on the patient-oriented eczema measure (POEM), with a group mean POEM score of 7.31 (6.65, 7.90)32.

KIR gene prevalence

The overall prevalence of KIR2DL5 gene was 52.5% (95% CI 48.0,57.0), KIR2DS5 gene was 33.0% (28.8,37.3), and KIR2DS1 gene was 33.6% (29.4,38.0) in the GAD cohort. As previously reported, subjects with these genes were more likely to have AD; KIR2DL5 (OR:1.57 95% CI (1.07,2.31)), KIR2DS5 (1.80 (1.18,2.76)), and KIR2DS1 (1.57 (1.03,2.39)). The copy number variation (CNV) of KIR2DL5 and KIR2DS5 varied from 0 to 4 copies and increasing CNV was associated with AD (1.52(1.22,1.89), p < 0.0001 and 1.66(1.16,2.37), p = 0.006, respectively). The KIR2DL5 was not highly correlated (R2 approximately 0.45) to the two other KIR genes. KIR2DS1 and KIR2DS5 were moderately in linkage disequilibrium (LD) (R2 approximately 0.61). The total number of alleles discovered by genotyping for KIR2DL5 was 10, KIR2DS5 was 9, and KIR2DS1 was 8 (see Supplement for all KIR gene frequencies). As noted in the methods section, a priori we decided to evaluate only those alleles with an allelic frequency of ≥ 0.05 (Table 1). The number of alleles meeting this criterion for KIR2DL5 was 2, KIR2DS5 was 1, and KIR2DS1 was 3 (Table 1).

Table 1 Allelic frequencies ≥ 0.05 for 2DL5, 2DS1, and 2DS5.
Full size table

KIR allele frequency

In the GAD cohort, the presence of the KIR2DL5*001:01 allele increased the odds of AD by about 86% (1.86(1.23,2.82) p = 0.003). More specifically, those homozygous for KIR2DL5*001:01 had an OR of 2.16 (1.31,3.53; p = 0.002) (Tables 2 and 3). The risk of having AD for KIR2DL5*001:01 was similar in Whites or Blacks (Table 3 and Supplementary Table 2). The alleles from the other KIR genes of interest were not associated with AD. KIR2DL5*001:01 is part of the KIR2DL5A gene complex which is located in the telomeric half of the KIR gene locus on chromosome 19 and can include the KIR2DS1 and KIR2DS533,34. Those with KIR2DL5A alleles were at an increased risk for having AD (1.86 (1.23, 2.82) p = 0.003) (Table 2).

Table 2 Odds ratios and 95% CI for the association of atopic dermatitis by KIR alleles of interest (allelic frequency of > 0.05).
Full size table
Table 3 Odds ratios and 95% CI for the association of atopic dermatitis for KIR alleles of interest evaluated by the presence or absence of the allele.
Full size table

KIR allele and HLA interactions

Interactions between HLA ligands and KIR alleles are presented for the presence or absence of the KIR allele and HLA ligand (Table 4). Similar results were obtained for the KIR alleles (categorized as heterozygote or homozygote). There is no known HLA ligand for KIR2DL5. A previous report showed an increased effect of KIR2DL5 in the presence of HLA-B *-21TT22. The effect of KIR2DL5*001:01 increased in the presence of HLA-B *-21TT leader sequence (2.46 (1.37,4.41) p = 0.0025), C2 (2.07 (1.37,4.41)), and the weakly binding Bw4 epitope variant B80T (2.91(1.50,5.63) p = 0.0016). The association between AD and KIR2DL5*001:01 was not found to be confounded by race, HLA-A*01:01, HLA-A*02:01, HLA-B*07:02, HLA-C*07:02, C2, Bw4 (an HLA ligand), B80T (an HLA variant associated with HLA-B position 80 coding for threonine), seasonal allergies or asthma.

Table 4 ORs for the KIR allele with 95% CI by the presence or absence of known HLA KIR ligands.
Full size table

Discussion

This is the first case–control study of AD to explore KIR allelic variation. As previously reported, the presence of three KIR genes is associated with AD; KIR2DL5, KIR2DS1, and KIR2DS522. However, only the KIR2DL5*001:01 allele is independently associated with an increased risk of AD. This association is greatest for the presence of KIR2DL5*001:01. The other KIR2DL5 variants did not increase the risk of AD. This is consistent with recent studies that have demonstrated that the KIR2DL5*002:01, which is the second most frequent and is the dominant centromeric KIR2DL5 variant, is silenced and not expressed on the NK cell surface35,36. The increased risk associated with KIR2DL5*001:01 is augmented by an interaction with HLA-B*-21TT leader sequence, which is associated with less “educated” or less active NK cells22,24. The magnitude of the effect of KIR2DL5*001:01 may also be augmented by HLA-C epitope C2.

Our findings are consistent with other diseases. Previous studies of HIV have shown that within a KIR gene, KIR allelic variation can have important influences on the functioning of NK cells as well as how the KIR receptor interacts with HLA ligand37. In addition our findings refine our previous investigation, which was based on simply genotyping the presence or absence of the KIR gene rather than focusing on specific alleles22. The risk of AD due to KIR2DL5*001 likely does not vary significantly by race; although the effect of KIR2DL5*001:01 may be greater in Blacks with AD than Whites.

Two recent studies described the potential physiology of an association between AD and NK cell function and number11,16. In a cohort of individuals with moderate to severe AD by Mack et al. , individuals with AD had fewer circulating NK cells that are more homogenous than expected11. The number of circulating NK cells could be returned to control subject levels after treatment with anti-IL-4 blockade and IL-15 super-agonism11. Furthermore, Mack et al. showed a low frequency of KIR2DL5 cell markers on the circulating NK cells but did not compare the frequency of these cell markers between groups with and without AD. Some of the subjects evaluated by Mack et al. were from the GAD cohort. Mobus et al. evaluated NK cells using skin transcriptomics in a likely White European group with moderate to severe AD16. These investigations noted an increased number of NK cells in AD lesional skin as compared to AD non-lesional skin or healthy control skin16. Mobus et al. further showed that surface markers on the NK cells from AD lesional skin consistently had an increase in inhibitory receptors and were poorly educated16. There findings are consistent with KIR2DL5*001:01 function. Their findings could be reversed after treatment with anti-IL-4 blockade and calcineurin inhibition16. However, the inhibitory receptors evaluated by these investigators were not KIR receptors but natural cytotoxicity receptors (NCR)16,38,39. While Mobus et al. implied that their results contradicted those of Mack et al., the disagreement is mostly based on the interpretation of mouse studies reported in Mobus et al.11,16. Experimentally induced mouse AD many not fully represent human disease11,16.

Focusing on the human subject results from the three studies, NK cell numbers differ in circulation and lesional skin in those with AD as compared to those without AD. Those with AD are more likely to have poorly educated NK cells (less active) in circulation that are more likely to express inhibitory receptors from NCR or KIR families11,12,16. Mack et al. and Mobus et al. both found larger numbers NK cells in human lesional skin11,16. Although the full functional consequences are unknown, it is possible that allelic polymorphisms such as KIR2DL5*001:01 impact KIR protein expression and function in a group of mature NK cells found in AD lesional skin. A previous study showed that specific KIR2DL5A alleles such as KIR2DL5*001:01 are expressed on the NK cell surface, whereas by a variety of mechanisms such as silenced transcription (e.g., KIR2DL5B*006:01) or intracellular retention (e.g., KIR2DL5A*005), some KIR2DL5 alleles are not expressed35.

This is an epidemiologic study and inherently has limitations. We focused on common allelic variants, not rare variants. This was primarily due to sample size and power considerations. It is likely that less common alleles in the genes KIR2DS1 and KIR2DS5 are associated with AD. However, from a population perspective, we focused on alleles that are likely relevant to population at large. In addition, while rare variants might help explain disease mechanisms, common variants are critical for conducting translational studies on humans to further our understanding of the immunologic mechanisms associated with NK cell function and AD. Larger studies should be conducted to investigate less common variants. We evaluated genetic variation but did not evaluate the expression of NK surface receptors and how these receptors interact with keratinocytes, antigen presenting cells or other cells. Based on our studies, these evaluations should be done specifically for KIR2DL5*001:01 in future studies. We were also not able to evaluate NK cell numbers in circulation or in tissue. These studies are essential to further our understanding of how KIR alleles effect NK cell function with respect to AD. Finally, it is possible that our study does not generalize to other populations with AD. However, strengths of our study include that our participants were recruited from multiple centers and, unlike previous studies, we were able to assess associations in a relatively large number of African Americans.

In summary we conducted a case–control study of White and Black individuals from the United States with physician confirmed AD. We demonstrated that the common allele KIR2DL5*001:01 is associated with AD and that this effect is augmented in the presence of HLA-B*-21TT. This finding is consistent with previous tissue based reports11. Our study adds to the growing literature on the genetic basis of the immune dysregulation of AD and, specifically, that KIR allelic variation is associated with an increased risk of AD20,27,29,40,41. Further a paradigm shift has occurred in understanding of the immunology of AD suggesting that NK deficiency may be real, may interplay with barrier dysfunction and that treatment studies specifically examining NK induction and modulation are now warranted11. Specifically, studies of NK cell immunologic function in those with AD may help explain disease variability, different disease immunologic endotypes, differences in genetic predilection for AD by race, as well as treatment failure10,42,43,44,45.

Materials and methods

Study population

Study subjects were enrolled between 2015 and 2020 as participants in the Genetics of Atopic Dermatitis (GAD) cohort; details of this study were published earlier30,46. 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 (controls). We evaluated self-reported race. This decision was based on previous evaluations of US studies on AD of race and genetic admixture that have shown that self-reported race is highly correlated with genetically determined ancestry (also see Supplementary Table 2)47,48,49. All subjects in the GAD cohort provided informed consents and/or assents that was approved by the appropriate Institutional Review Boards including the institutional review boards from the University of Pennsylvania, Washington University of St. Louis, Pennsylvania State University, and Children’s Hospital of Philadelphia. DNA samples analyzed for this study were anonymized. All studies were conducted consistent with ethical standards and guidelines of Scientific Reports.

DNA analysis

HLA genotyping

DNA was collected using Oragene DNA collection kits (DNA Genotek, Ottawa Canada) as previously reported49,50,51,52. HLA Class I genes (-A, -B, and -C) were sequenced using targeted amplicon-based NGS as previously described50,51,52.

We focused on HLA ligands known to interact with KIR genes studied. HLA-B leader sequence is defined at location − 21 (amino acids T or M) of HLA-B. The HLA-C epitope that interacts with these KIR genes is at position 80 and is defined by amino acids N and K (i.e., C1/C2 KIR binding site)24,26,27,53. Residue locations were based on IMGT data54,55. These data were previously published22,30,51.

KIR allelic sequencing

KIR allele typing, requiring 15 µl of DNA in the concentration range of 20–50 ng/µl, was conducted by the Institute for Immunology and Infectious Diseases (IIID) of Murdoch University in Perth, Western Australia. Uniquely indexed primers were designed to target KIR exons 3, 4, and 5 (D0, D1 and D2 Domain). The primers chosen were adapted and refined from published literature56. Primer mixes were prepared containing 2–4 primers per amplicon to cover all KIR genes in a multiplexed assay with an amplicon size ~ 400 bp. Five separate PCR reactions were setup for each sample to cover all targeted exons in 12.5 μl volume in 96-well plates with GoTaq DNA Polymerase (Promega) and the associated buffer system. As each sample was uniquely indexed during the PCR reactions, all amplicons from the PCR reactions were pooled using volumes appropriate to obtain balanced read coverage for each KIR exon. The minimum coverage of 10 reads and > 10% of the reads was accepted to be called an allele, however in most cases we have a read coverage of over 50 reads and majority of the heterozygous calls were larger. The products were sequenced on the Miseq using 600V3 chemistry. Post sequencing, quality filtered paired reads were demultiplexed based on unique molecular sequence for each sample and each set of overlapping paired reads were merged based on the Q30 scores leading to a single read with a median read length of up to 500bps. spanning the full amplicon. Since these amplicons span across the full exon, the SNPs within the exon were phased. These reads were then aligned to a reference sequence containing all KIR genes using CLCbio genomics workbench. Results were based on the current KIR dataset available IMGT/KIR Sequence Database—release: v2.10.0, date: 16 December 2020 (https://www.ebi.ac.uk/ipd/kir/docs/version.html). Multi mapped reads were extracted and reassigned based on set defined rules based on KIR haplotypes and heterozygosity and final assignments are reported as igroups57. Please refer to the link below for the list of condensed alleles in the i-groups. http://www.iiid.com.au/s/iiid_KIR_iGroups_v2100.xlsx. Each pool was then ligated with unique Illumina indexes and sequencing adapters ready for Illumina sequencing. The alignment files (bam files) were then used for assigning the KIR allele calls using VGAS which generates the cluster consensus sequences. Following alignment, using an inhouse developed application All Class the reads are moved to the correct gene by comparing each mapped read to the reference gene dataset generated from IMGT KIR database reads that do not map to any reference sequence data and reads mapping below the defined minimum cut off are discarded KIR alleles are presented to two fields. Validation of this pipeline has been shown by using this technique to demonstrate that an allele distribution analysis of the Perth HIV cohort dataset using this pipeline compares quite well to the frequency distribution in the United States58.

Analysis

The prevalence of KIR genes was estimated at the subject level and are presented with 95% confidence intervals (CI). The frequencies of KIR alleles were estimated at the chromosome level and are also presented with 95% confidence intervals (CI). These parameters were estimated for those with and without AD and by self-reported race (White or Black) which we found to be highly correlated with genetically determined ancestry in a previous study49. We limited our analyses to common alleles (i.e., frequency of ≥ 0.05). The odds ratio (OR) of having AD was estimated using logistic regression. Analyses of KIR genes and alleles are complicated and, in prior studies, have been inconsistently presented18,19,59,60. Our analyses focused on three primary comparisons of 2DS5, 2DL5, and 2DS1 KIR genes: (1) presence or absence of the KIR gene, (2) the association of a KIR allele (i.e., having the gene but not the allele, heterozygote for the allele, and homozygote for the allele) to the absence of the KIR gene, and (3) presence or absence of a specific KIR allele in the whole cohort. The association of known interactions between KIR and HLA ligand pairs was evaluated based on the presence or absence of the HLA ligand24,60,61.

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 pathway62. Our hypotheses were determined a priori and, based on previously published data of KIR genes known to be associated with AD, and hence p-values were corrected per gene for multiple testing22.

All statistical analyses were conducted using Stata Version 17.0 (College Station, TX).

Ethics statement

Human subject involvement was reviewed and approved by the Institutional Review Board of the University of Pennsylvania. All studies were conducted consistent with the ethical and regulatory requirements of Scientific Reports.