pH and Proton Sensor GPR65 Determine Susceptibility to Atopic Dermatitis

Extracellular acidic pH occurs in certain epithelia of the colon and skin and at sites of inflammation (1). It is sensed by several proton-sensing G protein–coupled receptors (GPCRs), including GPR65, which is widely expressed on immune cells (2). It signals through Gαs, leading to cAMP production, which phosphorylates the transcription factor cAMP response element binding protein (CREB) and blocks its binding to the NF-κB complex, thus limiting various proinflammatory responses (3). Normal signaling through GPR65 under low pH inhibits the release of inflammatory cytokines from immune cells (4–6). Furthermore, GPR65 is of great disease significance that GPR65-deficient animals exhibit exacerbated inflammation in preclinical models of inflammatory bowel disease (IBD) (7, 8) and acute lung injury (9). A GPR65 missense variant rs3742704 (I231L) and a GPR65-intronic single-nucleotide polymorphism (SNP) rs8005161 are among the top variants associated with the risk of human IBD (both ulcerative colitis and Crohn disease) in multiple studies (10–13). Together, pH and GPR65 signaling may be a major mechanism for the regulation of inflammation, particularly at the epithelia.

Atopic dermatitis (AD) is a chronic and relapsing inflammatory skin disease characterized by intense pruritic and eczematous lesions. AD affects 15–30% of children and 2–10% of adults in Western societies (14). Mechanisms thought to underlie AD pathogenesis include both excessive type 2 inflammatory responses and impairment of barrier function (15, 16). Although AD is highly heritable (17, 18), findings of the existing genome-wide association studies (GWAS) only account for ∼30% of AD heritability (19, 20). Regulation of skin pH may be an overlooked important factor. The surface pH of healthy skin is typically 4.5–6 (21). AD lesions have been reported to show higher pH values than healthy skin, and pH was positively associated with the severity of AD (22, 23). In a mouse model of AD, exposure to alkaline agents increased the severity of dermatitis lesions, whereas topical application of acids improved skin barrier function (24–26). However, the role of pH sensing by GPR65 in skin remain unknown.

In this study, we performed a phenome-wide association study (PheWAS) for GPR65-intronic SNP rs8005161 to understand the broader disease significance of GPR65 in human health. Apart from the known association with IBD, the minor T allele of rs8005161 was found to be strongly associated with human AD. In alignment with this, the absence of GPR65 in mice (GPR65^gfp/gfp) led to exacerbated skin inflammation in an experimental AD model, with significant elevation of skin pH. Furthermore, the exacerbated AD in GPR65^gfp/gfp mice was characterized by increased infiltration of various types of immune cells. Gpr65^gfp/gfp neutrophils showed changes in cell migration properties in vitro. In addition, mouse Gpr65^gfp/gfp T cells and human rs8005161-CT T cells showed increased production of the AD pathogenic cytokine TNF-α (27). Together, these data highlight a critical role of GPR65 in leukocyte migration, cytokine responses, and the pathogenesis of AD in mice and humans.

PheWAS was conducted in the 23andMe participant cohort from 647,776 participants of European ancestry. A total of 1,228 phenotypes collected within 23andMe’s data were included for exploration. Phenotype data are derived from responses to self-reported questionnaires. The association of rs8005161 with each phenotype was evaluated using a likelihood ratio test comparing the full model against null model with or without genotype adjusted for age, gender, genotyping array design, and PCs to correct for population stratification. Binary phenotypes were analyzed using logistic regression, quantitative phenotypes were analyzed using linear regression, and survival phenotypes were analyzed using Cox proportional hazard regression. We assume additive allelic effects. False discovery rate for each phenotype was calculated using the Benjamini–Hochberg procedure.

All experimental procedures involving mice were carried out according to protocols approved by the Animal Ethics Committees of Monash University and Victoria University of Wellington. Gpr65^gfp/gfp mice were purchased from The Jackson Laboratory (stock number 008577). Gpr65^gfp/gfp mice have 90% of exon 2 coding sequences replaced by promoterless IRES-EGFP sequences to disrupt GPR65 function (28). Wild-type (WT) controls were C57BL/6 mice obtained from the Monash Animal Research Platform, Monash University. All mice were maintained under specific pathogen–free and controlled environmental conditions.

Mice were anesthetized using isoflurane. MC903 (calcipotriol; Cayman Chemicals, Ann Arbor, MI) was dissolved in 100% ethanol and topically applied on mouse ears (1 nmol in 20 μl, 10 μl per side of ear) on days 0, 2, 5, 7, 9, 12, and 14. As vehicle control, the same volume of ethanol was applied. All experiments were performed on age- and sex-matched animals.

Transepidermal water loss (TEWL) was measured while mice were sedated, using the DermaLab TEWL probe (Cortex Technology, Hadsund, Denmark) at end point. TEWL was measured on both ears of the mice at room temperature, and results were recorded when TEWL regions stabilized. Two readings from each ear were taken and averaged for each mouse.

Skin pH was measured while mice were sedated, using the skin & scalp pH meter (Hanna Instruments) on day 14. Skin pH was measured on both ears of the mice at room temperature, and results were recorded when pH regions stabilized. Two readings from each ear were taken and averaged for each mouse.

Mice were recorded via time-lapse videography, and itch events were determined and quantified. This pathophysiological measurement was obtained from video observation of the relevant treatment groups 24 h before experimental end point.

Ear tissue was fixed in 10% neutral buffered formalin for 24 h, processed, paraffin embedded, and sectioned at 4 μm. H&E staining was conducted on ear sections. H&E stained was evaluated at 20× magnification and measurements of acanthosis (epidermal thickening) and dermal thickening were quantified by ImageJ analysis (National Institutes of Health, Bethesda, MD). Images of the tissue sections were taken on an Olympus IX71 inverted bright-field/fluorescence microscope with an Olympus DP70 digital camera (Olympus, Tokyo, Japan).

The human study was reviewed and approved by the human ethics committee of Wuxi Hospital of Integrated Traditional Chinese and Western Medicine. The study population included 189 healthy subjects with Chinese Han ancestry. All subjects provided written informed consent to be included in the study. Genotyping of SNP rs8005161 was performed by DNA sequencing. Briefly, genomic DNA was extracted from buccal swab by TIANamp Swab DNA kit (TIANGEN, Beijing, China). The target DNA was amplified by PCR by a PTC-200 PCR instrument (Bio-Rad Laboratories, Hercules, CA) using the following cycling conditions: 5 min at 96°C, 10 cycles of 96°C for 20 s, 62–52°C touchdown for 20 s, 72°C for 30 s, 35 cycles of 96°C for 20s, 52°C for 20s, 72°C for 30 s, and 5 min at 72°C. The amplified DNA were sequenced by an ABI3730XL DNA analyzer (Applied Biosystems, Foster City, CA) using the forward primer. The primers used for PCR and sequencing are forward: TAAAGGAGGCAAAATAAT and reverse: CTTCTTTGTAGTGACGCA. Eight subjects carrying rs8005161-CC and six subjects carrying rs8005161-CT provided blood samples for the cytokine detection from T cells. Demographic data were obtained at the time of blood collection.

For mouse skin cell suspensions, ear tissue was split into dorsal and ventral layers and was minced in 2 ml RPMI 1640 (Life Technologies, Grand Island, NY) with 10% FCS. The tissue was then digested with 2 mg/ml Collagenase Type IV (Life Technologies, Grand Island, NY) and 120 μg/ml DNase I (Roche Diagnostics, Mannheim, Germany) for 20 min at 37°C under agitation. After the digestion, the remaining tissue was passed through a 70-μm strainer and washed with cold RPMI 1640 containing 5 mM EDTA. For mouse ear draining lymph nodes cell suspensions, ear draining lymph nodes were mechanically disrupted and passed through a 70-μm strainer. For mouse splenic cell suspensions, spleens were mechanically disrupted and passed through a 70-μm strainer. Cells were then subjected to RBC lysis and washed with PBS. Human PBMCs were isolated by density gradient centrifugation using Ficoll-Paque (GE Healthcare).

Mouse splenic cells and human PBMCs were cultured in cell culture media, which was altered to pH 6 and 7 using HCl and NaOH. Cells were stimulated with 100 ng/ml PMA (Sigma-Aldrich) and 1 μg/ml ionomycin (Sigma-Aldrich) for 4 h.

Cells were resuspended with FACS buffer (PBS containing 2% FCS and 4 mM EDTA) and incubated in FcR Blocking Reagent, mouse (Miltenyi Biotec, Auburn, CA) to block nonspecific Ab binding. For T cells analysis, cell surface staining was conducted for 30 min using the following analysis panel: anti-mouse CD45 Alexa Fluor 700 (AF700; 30-F11; BD Pharmingen, San Jose, CA), anti-mouse CD3e PE-Cy7 (145-2C11, BD Pharmingen), anti-mouse TCR-β BV421 (H57-597; BD Horizon), anti-mouse TCR-γδ APC (GL3; BioLegend, San Diego, CA), anti-mouse CD4 BV510 (RM4-5; BD Horizon), and anti-mouse CD8α PE (53-6.7; BD Pharmingen). For myeloid cell analysis, cell surface staining was conducted for 30 min using the following analysis panel: anti-mouse CD45 AF700, anti-mouse CD11b APC (M1/70; BD Pharmingen), anti-mouse Siglec-F BV421 (E50-2440; BD Horizon), anti-mouse Ly-6G PE (1A8; BD Pharmingen), anti-mouse FcεR1α BV510 (MAR-1; BioLegend), anti-mouse c-kit BUV395 (2B8; BD Horizon), and anti-mouse CD49b PE-Cy7 (DX5; Invitrogen eBioscience). Dead cells were excluded using 7-aminoactinomycin D (BD Biosciences, San Jose, CA). Cell counts were determined using flow cytometry counting beads (CountBright Absolute; Life Technologies) following manufacturer instructions. For intracellular cytokine staining, stimulated mouse splenic cells were harvested and surface stained with anti-mouse CD4 Pacific Blue (RM4-5; BD Pharmingen) and anti-mouse CD8α AF700 (53-6.7; BD Pharmingen); stimulated human PBMCs were harvested and incubated in Zombie Green fixable viability dye (BioLegend); human cells were then incubated in FcR Blocking Reagent, human (Miltenyi Biotec) and surface stained with anti-human CD45 AF700 (2D1; BioLegend), anti-human TCR-α/β BV510 (BioLegend), anti-human CD4 PE-Cy7 (IP26; BioLegend), and anti-human CD8α APC (HIT8α; BioLegend). Cells were then fixed and permeabilized using an Foxp3/Transcription Factor Staining Buffer Set (eBioscience). Mouse cells were then stained with anti-mouse TNF-α PE-Cy7 (MP6-XT22; BioLegend); human cells were then stained with anti-human TNF-α BV421 (MAb11; BioLegend). Sample data were acquired using a five-laser BD LSRFortessa X-20 flow cytometer and BD FACSDiva software (BD Biosciences) and analyzed using FlowJo software (Tree Star).

Neutrophils were isolated from bone marrow as previously described (29). Briefly, bone marrow was subject to RBC lysis and filtered. Neutrophils were purified by gradient density centrifugation. Neutrophils were resuspended in complete media (pH 7.4 or 6.5). A total of 3 × 10⁵ neutrophils were added to the top wells of a Transwell plate (Corning). The bottom wells contained complete media with 0, 0.3, 1, or 3 nM rCXCL8 (574202; BioLegend). Cells were incubated at 37°C for 1.5 h. Neutrophils in the bottom wells were quantified by flow cytometry.

All graphical representation of data were done using Prism 8.2.1 (GraphPad Software, La Jolla, CA). Data were presented as mean ± SEM. Normality of distribution of continuous data were assessed by Shapiro–Wilk test. Comparison between multiple groups of different treatments and different mouse strains was analyzed using two-way ANOVA with Tukey multiple comparisons test. Comparison between two groups was analyzed using independent two-tailed Student t tests. All p values <0.05 were considered significant. Details of statistical analysis for each experiment can be found in the figure legend (including test details, n, definition of significance, average, and precision metrics).

The T allele of GPR65-intronic SNP rs8005161 is present at high frequency (∼25% in most populations worldwide based on the Ensembl database of 1000 Genomes Phase 3) (30). The PheWAS for rs8005161 was conducted on 1228 phenotypes collected from 647,776 genotyped subjects of European ancestry within 23andMe’s dataset. As reported by previous studies, rs8005161 showed significant association with IBD (p = 3.4 × 10⁻⁵) and asthma (p = 3.6 × 10⁻⁵) after false discovery rate correction (Fig. 1, Supplemental Table I). The other major human disease associated with rs8005161 was eczema (AD) (p = 9.2 × 10⁻⁵) (Fig. 1, Supplemental Table I). Although the functional validations are abundant for the roles of GPR65 in IBD (7, 8, 31) and asthma (32), the association between GPR65 and AD has never been reported before. However, rs8005161 did not stand out in any existing GWAS for AD to date (20, 33–35).

The SNP rs8005161 is in perfect linkage disequilibrium with the GPR65 missense variant rs3742704 (I231L), which results in ∼50% less signaling by GPR65, as measured by cAMP production (7). To investigate the effect of abrogated GPR65 signaling on disease pathogenesis, we studied Gpr65^gfp/gfp mice that have 90% of an exon in Gpr65 replaced with GFP to disrupt GPR65 function (36). Given the robust correlation between AD and rs8005161, we used a mouse model of AD using MC903 (also known as calcipotriol) (Fig. 2A) to interrogate the effect of GPR65 in the pathogenesis of this disease. Topical application of MC903 induces allergic skin inflammation characterized by pruritic eczematous skin lesions and elevated skin thickness in mice, which highly resembles human AD (37). Gpr65^gfp/gfp mice developed exacerbated eczematous skin lesions with far worse ear thickening after MC903 treatment compared with WT (C57BL/6) mice (Fig. 2B, 2C, Supplemental Fig. 1A, 1B). Barrier function is considered central in the pathogenesis of AD (38). We measured the TEWL, the loss of water from the skin through the epidermis to the outer environment via diffusion and evaporation processes, to indicate the severity of skin barrier disruption as previously described (39). Gpr65^gfp/gfp mice showed significantly higher TEWL than WT mice (Fig. 2D), indicating exacerbated barrier impairment after MC903 treatment. To assess pruritus (itch severity), we counted the times mice scratched their ears over 30 min on day 14, which was 24 h prior to the end point. In accordance with the severity of disease shown above, Gpr65^gfp/gfp mice showed higher pruritus (Fig. 2E). Elevated skin pH has been widely reported in AD patients (22–24). Consistent with this, MC903-treated skin exhibited a significantly higher pH than healthy skin, but there was no difference between WT and Gpr65^gfp/gfp mice (Fig. 2F). Histopathological assessment demonstrated that the skin of MC903-treated Gpr65^gfp/gfp mice showed exacerbated inflammation (Fig. 2G) and acanthosis (thickening of the epidermis) (Fig. 2G, 2H). No significant difference was observed in the dermal thickening between WT mice and Gpr65^gfp/gfp mice (Fig. 2H). Collectively, these data demonstrate that GPR65 deficiency results in exacerbation of AD in a mouse model.

Following the exacerbated phenotypes observed in Gpr65^gfp/gfp mice, we examined the impact of GPR65 deficiency on leukocyte infiltration in the inflamed site by flow cytometry. As expected, we observed a marked increase of CD45⁺ leukocytes infiltrating the skin and ear draining lymph nodes in Gpr65^gfp/gfp mice after MC903 treatment, compared with WT mice (Fig. 3A). Furthermore, we observed significant increases in neutrophils (Fig. 3B) and mast cells (Fig. 3C), and a nonsignificant increase of basophil infiltration (Supplemental Fig. 3A) in the inflamed skin of Gpr65^gfp/gfp mice. These subsets are known to contribute to the pathogenesis of AD (40–43). Eosinophils also play causal roles in AD (44). However, eosinophil infiltration was slightly reduced in Gpr65^gfp/gfp mice (Supplemental Fig. 3A). This finding is supported by a previous report on allergic airway disease in mice (32), suggesting that the exacerbation of AD in Gpr65^gfp/gfp mice is not through eosinophilia.

Effector functions by T cells contributes to the pathogenesis of human AD (15). Using a rigorous gating strategy for T cells (Supplemental Fig. 2), we noticed that GPR65 deficiency results in a drastic elevation of all T cell subsets in inflamed skin, including CD4⁺ T cells (Fig. 3D), CD8⁺ T cells (Fig. 3E), and conventional γδ T cells (Fig. 3F). The elevation of T cell numbers caused by GPR65 deficiency was also observed in ear draining lymph nodes (Supplemental Fig. 3B).

GPR65 in humans is expressed mostly by leukocytes, and much less so by epithelial cells (45). We used the GFP reporter of heterozygous Gpr65^+/gfp mice to indicate GPR65 expression in different leukocytes infiltrating the inflamed skin. As expected, the reporter GFP was widely detected in cutaneous leukocytes both under resting and inflamed states (Supplemental Fig. 4A). Although GPR65 was downregulated in other leukocytes during MC903-induced inflammation (Supplemental Fig. 4B), it was upregulated by neutrophils (Fig. 4E, 4F). Collectively, these results suggest that loss of GPR65 increases the presence of inflammatory cells, particularly neutrophils, T cells, and mast cells, in the skin after MC903 treatment.

Neutrophils are highly migratory and pathogenic in several autoimmune inflammatory diseases, including AD (40, 41). Skin-infiltrating neutrophils are critical for the priming of AD and inducing chronic itch (41). The application of monoclonal Abs against the main neutrophil chemokine receptor, CXCR2, can effectively attenuate MC903 AD in mice (46). Neutrophils express high levels of GPR65 (Fig. 4A, 4B). Protein kinase A (PKA) and cAMP, downstream of GPR65 signaling, play a central role in cytoskeletal reorganization and cell migration (47, 48). That leukocytes migrate differently with changes in pH was noted more than 20 y ago (49–51). Compromised GPR65 signaling, for instance by genetic polymorphisms or high pH, may increase leukocyte recruitment to inflamed tissues. We assessed the migratory properties of Gpr65^gfp/gfp neutrophils toward the neutrophil chemoattractant CXCL8 in an in vitro chemotaxis assay at physiological (pH 7.4) and slightly acidic (pH 6.5) pH. Interestingly, the migration of Gpr65^gfp/gfp neutrophils at either pH was increased compared with WT neutrophils (Fig. 4C), indicating that lack of GPR65 signaling somehow altered the machinery involved in cell migration. Of note, WT neutrophil migration (pH 6.5) was completely inhibited, whereas migration of Gpr65^gfp/gfp neutrophils was only partially reduced (Fig. 4C). These results demonstrated an inhibitory function of GPR65 in neutrophil migration and the essential anti-inflammatory role of GPR65 signaling at barrier surfaces.

From the perspective of adaptive immune responses, cytokine production by T cells is a key factor regulating inflammation. For instance, TNF-α is a major inflammatory cytokine and is easily quantified by intracellular staining. TNF-α is elevated the serum of AD patients (52) and plays a causal role in inducing AD-like features on engineered human skin equivalents (27). Furthermore, TNF-α promotes Th2 responses, which is central to the development of AD (53). GPR65 signaling and cAMP production is known to suppress LPS-induced TNF-α production in macrophages (54). Because we observed a significant increase of T cell infiltration in the skin of Gpr65^gfp/gfp mice (Fig. 3D–F), we decided to further understand the influence of pH and GPR65 signaling on T cell functions. We examined intracellular cytokine staining for TNF-α in both mouse splenic T cells and human T cells from blood. Acidic pH (pH 6) significantly inhibited TNF-α production by mouse CD4⁺ and CD8⁺ T cells stimulated with PMA (Fig. 4A, 4B). Interestingly, the production of TNF-α under acidic pH was partially restored in Gpr65^gfp/gfp T cells (Fig. 4D, 4E). GPR65 still suppressed TNF-α production by CD8⁺ T cells under neutral pH (pH 7), whereas no difference was observed between WT and Gpr65^gfp/gfp CD4⁺ T cells (Fig. 4D, 4E).

To investigate whether the T allele of the rs8005161 polymorphism affects TNF-α production by T cells in humans, we identified six healthy rs8005161-CT individuals (Chinese Han ancestry; Supplemental Tables II and III) and tested the production of TNF-α from their T cells from human PBMCs upon PMA/ionomycin stimulation. TNF-α production by T cells was also significantly inhibited at pH 6 (Fig. 4F, 4G). rs8005161-CT CD4⁺ T cells produced more TNF-α under acidic pH (pH 6) compared with rs8005161-CC CD4⁺ T cells (Fig. 4F, 4G). rs8005161-CT CD8⁺ T cells showed a nonsignificant increase of TNF-α (Fig. 4F, 4G). We did not identify any rs8005161-TT individuals, as these appeared to be rare in the Chinese population under study despite the large number of CT individuals. Collectively, these data demonstrate that acidity naturally inhibits TNF-α production by T cells partially through GPR65 and that the rs8005161 polymorphism is a genetic element that appears to account for variability in cytokine production in human.

GPR65 is a receptor that negatively impacts on immune cell responses (1, 2). This relates to coupling of GPR65 with Gαs and resulting downstream anti-inflammatory activities (2, 3, 55). The missense rs3742704 polymorphism (I231L) knocks down cAMP production under acidic pH through GPR65 signaling by ∼50%, at least in HeLa cells (7). We speculated that GPR65 may be extremely important for the regulation of immune responses that would serve to limit immune cell migration or cytokine production, for instance at sites of intense inflammation.

SNPs within GPR65, rs3742704 and rs8005161 are associated with IBD (10–13). Animal studies have shown the involvement of GPR65 in the intestinal and airway inflammations (7–9, 32). Aligned with these reports, our PheWAS revealed a significant association between GPR65 SNP rs8005161 and IBD and asthma. Additionally, we found that the rs8005161 polymorphism in the GPR65 gene is associated with AD, suggesting a broader influence of GPR65 on barrier responses. Although no existing GWAS for AD supported this association, GPR65 deficiency exacerbated AD phenotypes and led to drier and thicker skin and scalier lesions and compromised barrier integrity in the experimental AD animal model. These results clearly demonstrate a protective role for low pH and GPR65 in AD pathogenesis and, moreover, suggest that PheWAS is able to identify some novel genetic risk loci that did not appear significant in GWAS. AD now adds an important part to a picture: most of the major barriers appear to rely on GPR65 and proton-sensing of low pH for proper homeostasis.

GPR65 is a proton sensor and is activated under acidic conditions. The pH of normal skin is usually below pH 6 and can be as low as pH 4.5 in healthy people (56). Skin pH in AD patients is often increased into the neutral to basic range (22, 23, 57), a pH in which GPR65 signaling is nonoperational. This may be important for AD progression because an acidic pH is essential for maintaining skin barrier integrity (21) and normal microbiome composition (57). Similar to the case in human AD patients, MC903 treatment drastically elevated the skin pH in mice from below pH 6 to approximately neutral. It is still not fully clear how the skin maintains acidic conditions, but our studies show that low pH is important for proper immune regulation through GPR65 signaling.

Consistent with the exacerbated AD phenotypes in Gpr65^gfp/gfp mice, immune cell infiltration was also significantly elevated in Gpr65^gfp/gfp mice after MC903 treatment. These results suggest that GPR65 may regulate AD through immune cell recruitment. Although epidermal cells express low levels of GPR65, GPR65 is highly expressed by immune cells (45). The expression level of GPR65 in most leukocytes was reduced after MC903 treatment, and this may result from pH elevation. Increased cell infiltration in Gpr65^gfp/gfp mice suggested that GPR65 signaling restricts leukocyte recruitment. It has been largely ignored that pH markedly alters leukocyte migration (49–51) and GPR65 signaling might be the mechanism underlying this. Neutrophils are highly migratory and induce itchiness and scratching during allergic skin inflammation (40, 41). Notably, neutrophils upregulated GPR65 during the inflamed state. Moreover, GPR65 signaling inhibited neutrophil migration in vitro.

The association of cAMP with cytokine responses led us to investigate whether pH and GPR65 regulate cytokine production by T cells. Acidity inhibited TNF-α production by both CD4⁺ and CD8⁺ T cells, which was partially dependent on GPR65 in mice. Furthermore, T cells from rs8005161-CT individuals also produced more TNF-α compared with T cells from rs8005161-CC individuals, validating the significant anti-inflammatory function of GPR65 in humans. Although the cohort of participants lacks ethnical diversity, our findings align with the known anti-inflammatory function of GPR65 (2). Therefore, we propose that circumstances of low pH, such as lactic acid production, or high amounts of short chain fatty acids in the colon elicit immune-suppressive responses through GPR65. The normal pH of the skin, and the colon, is ∼4.5–6.5, and interestingly, these tissues are targets for inflammation in rs8005161-CT individuals, presumably because pH has become dysregulated.

Despite that IBD and AD rarely co-occur clinically, the shared genetic risk loci suggest that they might have a complex genetic relationship. A recent meta-analysis reported a positive association between AD and IBD (58), further supporting our findings. GPR65 might be a critical bridge linking the cross-talk between the homeostasis on both major epithelial tissues. AD is the first step of atopic march, which may eventually progress to asthma. AD and asthma both came up in our PheWAS results, suggesting rs8005161 and GPR65 may underly the development of atopic march.

Our study has some limitations. First, our PheWAS was based on self-reported questionnaires, which is not as sensitive and specific as a clinical diagnosis. Nevertheless, this was compensated by a large sample size of 647,776 samples. Second, our PheWAS and human experiments are limited by the single ancestry of the cohorts, so we are not able to provide the information on the ancestral differences. Third, we acknowledge our findings need to be independently validated in other models of AD, which was outside the scope of the current study.

In conclusion, our study identified an important role for the proton sensor GPR65 in AD as well as other inflammatory diseases. GPR65 affected processes closely aligned with inflammatory responses, such as cytokine production and cell migration. These findings suggest the possibility of treating AD, IBD, and other inflammatory diseases with pH-lowering agents or with GPR65 agonists (54, 59).

We thank the research participants from 23andMe for contribution to this study.

Michelle Agee, Adam Auton, Robert K. Bell, Katarzyna Bryc, Sarah L. Elson, Pierre Fontanillas, Karen E. Huber, Aaron Kleinman, Nadia K. Litterman, Jennifer C. McCreight, Matthew H. McIntyre, Joanna L. Mountain, Elizabeth S. Noblin, Carrie A. M. Northover, Steven J. Pitts, J. Fah Sathirapongsasuti, Olga V. Sazonova, Janie F. Shelton, Suyash Shringarpure, Chao Tian, Joyce Y. Tung, and Vladimir Vacic (23andMe, Inc., Sunnyvale, CA)

N.B. was employed by Pfizer at the time of this research. N.F. and D.H. were employed by 23andMe at the time of this research and hold stock or stock options in 23andMe. The other authors have no financial conflicts of interest.