The Adhesion G-Protein-Coupled Receptor GPR115/ADGRF4 Regulates Epidermal Differentiation and Associates with Cytoskeletal KRT1

Among the 33 human adhesion G-protein-coupled receptors (aGPCRs), a unique subfamily of GPCRs, only ADGRF4, encoding GPR115, shows an obvious skin-dominated transcriptomic profile, but its expression and function in skin is largely unknown. Here, we report that GPR115 is present in a small subset of basal and in most suprabasal, noncornified keratinocytes of the stratified epidermis, supporting epidermal transcriptomic data. In psoriatic skin, characterized by hyperproliferation and delayed differentiation, the expression of GPR115 and KRT1/10, the fundamental suprabasal keratin dimer, is delayed. The deletion of ADGRF4 in HaCaT keratinocytes grown in an organotypic mode abrogates KRT1 and reduces keratinocyte stratification, indicating a role of GPR115 in epidermal differentiation. Unexpectedly, endogenous GPR115, which is not glycosylated and is likely not proteolytically processed, localizes intracellularly along KRT1/10-positive keratin filaments in a regular pattern. Our data demonstrate a hitherto unknown function of GPR115 in the regulation of epidermal differentiation and KRT1.


Introduction
The epidermis, a stratified epithelium, continuously renews from basal keratinocyte stem cells that migrate outwards to give rise to differentiated keratinocytes that ultimately form the skin barrier. In healthy epidermis, proliferation is restricted to keratinocytes in the basal layer, where the various stem cell types reside [1]. Their progeny constantly undergo differentiation to form the upper suprabasal layers, the stratum spinosum and stratum granulosum, and ultimately the stratum corneum. In normal skin, keratinocyte proliferation and differentiation are highly regulated to maintain epidermal homeostasis. The disruption of the underlying regulatory circuits contributes to conditions such as psoriasis, which is characterized by hyperproliferation and delayed differentiation [2]. Epidermal cytoarchitecture and tissue cohesion largely depend on the expression of keratins, which form the cytoskeletons of all epithelia, including the epidermis. Keratins protect the epidermis against mechanical stress, secure intercellular adhesion, and act as signaling hubs via multiple protein interactions [3,4]. They form extensive filament arrays through

Ethics Statement
The Ethics Committee of the Medical Faculty of Leipzig University approved the study of human skin for aGPCR expression (no. 386/18). All patients gave informed consent.

Antibodies (Abs) and Plasmids
The Abs used are summarized in Table S1. ADGRF4, amplified from human keratinocytes by RT-PCR, was cloned into the pcDNA3.1 plasmid. mGPR115 pcDps, encoding mouse GPR115, was kindly provided by I. Liebscher (Leipzig University).
For detection purposes, an HA tag was inserted directly downstream of the signal peptide, and a Flag tag was inserted at the C-terminus of the receptor using the Q5 ® Site-Directed Mutagenesis Kit (New England Biolabs GmbH, Frankfurt, Germany). The constructs encoding GPR115 with mutated glycosylation sites (GPR115 mutGlyc and GPR115 noGlyc) were purchased from OriGene Technologies GmbH (Herford, Germany). All created mutations were confirmed by sequencing. The plasmids were transfected into suitable cells using Lipofectamine 2000 (Thermo Fisher Scientific, Darmstadt, Germany).

Organotypic Skin Constructs
NIH-3T3 fibroblasts were treated with 4 µg/mL mitomycin C (Merck) for 2 h. Then, 4 × 10 4 postmitotic fibroblasts were embedded in 400 µL of rat tail collagen type I (Corning GmbH, Kaiserslautern, Germany) in a cell culture insert used in a 12-well carrier plate. After 24 h, the gel was coated with 5 µg/mL fibronectin (Merck) for 1 h, overlayered with 4 × 10 5 HaCaT cells, and cultured for 10 days. Constructs were fixed in phosphate-buffered 4% formaldehyde for 45 min and embedded in Tissue Tek (Thermo Fisher).

Crystal Violet Assay
To quantify the cell numbers of HaCaT clones, 1 × 10 3 , 2 × 10 3 , and 4 × 10 3 cells/well (three wells/density) were seeded in a 96-well plate and cultured for 48 h. The adherent cells were fixed for 10 min in 96% ethanol then washed and stained with 0.5% crystal violet/20% methanol for 25 min. After washing, the cells were lysed overnight with 100 µL of 0.5% Triton X-100. The absorbance was measured at 590 nm.

Immunolabeling and Proximity Ligation Assay (PLA)
Attached cultured cells or 6 µm tissue cryosections were fixed with ice-cold acetone for 10 min and immunostained as described in [16]. The Abs used for labeling are listed in Table S1. To detect and localize the specific interactions of endogenous proteins, the DuoLinkTM in situ PLA (Olink Bioscience, Uppsala, Sweden) was performed [17]. All immunofluorescence stains were imaged by confocal laser scanning microscopy (LSM800 or LSM880 fast Airyscan; Carl Zeiss AG, Oberkochen, Germany).

Immunoelectron Microscopy
Postembedding immune-gold labeling was performed as described in [18]. Briefly, human skin samples were fixed in 0.1 M phosphate buffer with 0.1% glutaraldehyde and 4% paraformaldehyde. Samples were embedded in LR White resin (Merck) and polymerized at 4 • C via UV radiation. Ultrathin sections were stained with the GPR115 ECD Ab in combination with the secondary Ab conjugated with NanogoldTM (Nanoprobes, Yaphank, NY, USA). For subsequent silver enhancement [19], the particle size was adjusted by the duration of incubation. Sections counterstained with ethanolic uranyl acetate underwent an analysis using an FEI Tecnai12 BioTwin TEM equipped with an SIS Mega-View3 CCD camera (Olympus Soft Imaging Solutions GmbH, Muenster, Germany). The images were analyzed using Fiji.sc.

Flow Cytometry
Cell surface expression was quantified by flow cytometry. Cells were incubated with the primary Ab for 30 min, followed by a fluorophore-labeled secondary Ab for 20 min. Cells were fixed with 1% paraformaldehyde/PBS and analyzed in an FACSAria III.

Western Blot Analysis
To isolate the epidermis, skin samples were incubated in PBS for 2 min at 60 • C. The epidermis was pulled off and stored in liquid nitrogen for further applications. Cells and epidermal samples were lysed either in 6.5 M urea buffer (50 mM Tris pH 7.4, 1 mM EGTA, 6.5 M urea, 2 mM DTT), RIPA lysis buffer, or M-PER (Thermo Fisher), each containing Halt™ protease and a phosphatase inhibitor cocktail (Thermo Fisher). For the glycosylation analysis, lysates were incubated with PNGase F (NEB GmbH, Frankfurt, Germany). The Western blot analysis was performed as described in [20].

Single-Cell RNA Sequencing (scRNAseq) Data Reanalysis
Data (GSE147482) [1] were downloaded from the GEO database. Using 300 marker genes for basal cell types I-IV (BAS I-IV), spinous types I-II (SNP I-II), and granular (GRN) keratinocytes as well as melanocytes (MEL) identified by Wang et al. [1] (library clustering), we ran a self-organizing map (SOM) analysis of all cells to identify cells with specific expression profiles [21]. The SOM spot analysis allowed the identification of cells that significantly overexpressed BAS I-III, GRN, and MEL signatures. The BAS IV and SPN I-II signatures were rather broadly distributed. A cluster of cells overexpressing these three signatures together and a part of the GRN signature was labeled as 'MIX'. Violin plots of the expression of individual genes were generated using the R-package Seurat 4.1.1 [22].

Statistical Analysis
Means ± SEM are given. Differences between HaCaT WT and GPR115 clones were analyzed by the Mann-Whitney U test. p values < 0.05 were considered significant.

High ADGRF4 Levels in Rare Basal and in Most Suprabasal Keratinocytes
ADGRF4 shows a distinct skin-dominated transcriptomic profile when comparing different human tissues (Figure 1a) [11]. To define this more precisely, we reanalyzed scRNAseq data of human neonatal epidermis [1]. We identified three subsets of basal keratinocytes (Figure 1b). The ADGRF4 level was low in most type-III basal cells, characterized by high COL17A1, encoding a structural component of hemidesmosomes. Notably, these are stem cells with an unlimited capacity for self-renewal that are generally quiescent. Most type-I and -II basal cells proliferate; they are characterized by high levels of cell-cyclerelated CDC20. The type-II restricted expression of UHRF14, playing a role in the G1/S phase transition, discriminates this population from type-I basal cells, and 14.5% and 20.3% of type-I and -II basal cells, respectively, had high ADGRF4 levels. Thus, GPR115 is likely expressed in a rare subset of basal keratinocytes. High ADGRF4 was present in 71.0% of the cells in the MIX cluster, including spinous keratinocytes, and in almost all granular keratinocytes. Consistent with the predominant ADGRF4 in differentiated keratinocytes, DSC1, encoding the desmosomal protein desmocollin-1, and IVL, encoding involucrin starting to be synthesized in the stratum spinosum, showed profiles similar to that of AD-GRF4. The closest relative of ADGRF4, ADGRF2, was almost absent in human keratinocytes. Melanocytes, positive for the specific transcription factor MITF, were ADGRF4-negative.  [11], given as RPKM (reads per kilobase per million mapped reads). (b) Reanalysis of scRNAseq data of human epidermis [1]. Violin plots of relative gene expression from basal cell types I-III (BAS), a mixed cluster (MIX, characterizing the transition from type-IV basal cell to spinous and granular keratinocytes), granular keratinocytes (GRN), and melanocytes (MEL). (c) Costaining for GPR115 and keratins of normal skin cryosections (star: cornified layer). (d) In images of a horse radish peroxidase based immunostained epidermis, few GPR115 + basal cells were seen at higher magnification (arrows; broken line: basal membrane). (e) Quantitation of GPR115 + cells in the basal layer, which either "sit at" or "touch" the basal membrane with extensions (dorsal skin, n = 4 donors; n = 1000-2000 basal cells/donor, mean ± SEM).

GPR115 Shows Intracellular Distribution in Human Epidermis
To validate these transcriptomic data at the protein level, we first carefully characterized GPR115 antibodies (Abs). We applied HaCaT wild-type (WT) keratinocytes and a HaCaT GPR115 knockout (KO) clone in which most ADGRF4 transcripts were lost as well as Cos-7 cells expressing HA-tagged GPR115 ( Figure S1a,b). We focused on the GPR115 ECD Ab, which binds the receptor at its ECD and stained 5-8% of the HaCaT WT cells but almost none of the ADGRF4-deficient GPR115KO2/2 cells. In HA-GPR115 Cos-7 cells, a one-to-one overlapping of the Tag with GPR115 ECD Ab labeling was seen.
Applying this Ab on cryosections of human epidermis, we detected GPR115 in few basal and in almost all KRT10-positive suprabasal, noncornified keratinocytes (Figure 1c), which is in full agreement with the transcriptomic profile. We noted that basal GPR115positive cells touched the basal membrane only with little membrane extensions but not across their entire basal site (Figure 1d,e). Unexpectedly, GPR115 displayed an apparent intracellular distribution in keratinocytes in vivo and lacked the expected membrane-associated localization pattern (Figure 1d). To further control GPR115 ECD Ab specificity, epidermal staining was repeated after the preincubation of the Ab with Cos-7 cells expressing GPR115, which strongly reduced intracellular epidermal staining ( Figure S1c). a b adrenal appendix bone marrow brain colon duodenum endometrium esophagus fat gallbladder heart kidnes liver lung lymph node ovary pancreas placenta prostate salivary gland skin small intestine spleen stomach testis thyroid urinary bladder

GPR115 Shows Intracellular Distribution in Human Epidermis
To validate these transcriptomic data at the protein level, we first carefully characterized GPR115 antibodies (Abs). We applied HaCaT wild-type (WT) keratinocytes and a HaCaT GPR115 knockout (KO) clone in which most ADGRF4 transcripts were lost as well as Cos-7 cells expressing HA-tagged GPR115 ( Figure S1a,b). We focused on the GPR115 ECD Ab, which binds the receptor at its ECD and stained 5-8% of the HaCaT WT cells but almost none of the ADGRF4-deficient GPR115KO2/2 cells. In HA-GPR115 Cos-7 cells, a one-to-one overlapping of the Tag with GPR115 ECD Ab labeling was seen.
Applying this Ab on cryosections of human epidermis, we detected GPR115 in few basal and in almost all KRT10-positive suprabasal, noncornified keratinocytes (Figure 1c), which is in full agreement with the transcriptomic profile. We noted that basal GPR115positive cells touched the basal membrane only with little membrane extensions but not across their entire basal site (Figure 1d,e). Unexpectedly, GPR115 displayed an apparent intracellular distribution in keratinocytes in vivo and lacked the expected membraneassociated localization pattern (Figure 1d). To further control GPR115 ECD Ab specificity, epidermal staining was repeated after the preincubation of the Ab with Cos-7 cells expressing GPR115, which strongly reduced intracellular epidermal staining ( Figure S1c).
In addition to the interfollicular epidermis, GPR115 expression was examined in skin appendages such as sebaceous and sweat glands and hair follicles ( Figure S2), where it showed a very similar distribution as KRT1 and KRT10 (Table S2).

Delayed GPR115 Expression in Suprabasal Psoriatic Keratinocytes
Our in vivo data are compatible with an involvement of GPR115 in epidermal differentiation. To obtain further insight, we turned to psoriatic skin as a model for disturbed epidermal homeostasis. In lesional psoriatic skin, characterized by excessive keratinocyte proliferation and altered differentiation, the distribution of GPR115 and keratins was changed, especially in rete ridges (Figure 2a,b). The expression of GPR115 and KRT1/10 was delayed until keratinocytes reached the granular layer. Simultaneously, many suprabasal cells expressed the proliferation marker Ki-67 ( Figure 2b). Of these Ki-67 + keratinocytes, 25% were positive for GPR115. In granular keratinocytes outside the rete ridges, GPR115 and KRT1 are always present but consistently appear non-homogenously in various cells. In addition to the interfollicular epidermis, GPR115 expression was examined in skin appendages such as sebaceous and sweat glands and hair follicles ( Figure S2), where it showed a very similar distribution as KRT1 and KRT10 (Table S2).

Delayed GPR115 Expression in Suprabasal Psoriatic Keratinocytes
Our in vivo data are compatible with an involvement of GPR115 in epidermal differentiation. To obtain further insight, we turned to psoriatic skin as a model for disturbed epidermal homeostasis. In lesional psoriatic skin, characterized by excessive keratinocyte proliferation and altered differentiation, the distribution of GPR115 and keratins was changed, especially in rete ridges (Figure 2a,b). The expression of GPR115 and KRT1/10 was delayed until keratinocytes reached the granular layer. Simultaneously, many suprabasal cells expressed the proliferation marker Ki-67 ( Figure 2b). Of these Ki-67 + keratinocytes, ~25% were positive for GPR115. In granular keratinocytes outside the rete ridges, GPR115 and KRT1 are always present but consistently appear non-homogenously in various cells. To address whether GPR115 preceded KRT1 or vice versa, immunostaining for both proteins was performed, focusing on the first GPR115-positive or KRT1-positive cells upon leaving the basal layer (Figure 2c). The notion that GPR115 + KRT1 -but no GPR115 -KRT1 + keratinocytes exist supports the hypothesis that GPR115 precedes the expression of KRT1. To investigate this, we turned to HaCaT-based organotypic skin equivalent models and performed loss of function studies. To address whether GPR115 preceded KRT1 or vice versa, immunostaining for both proteins was performed, focusing on the first GPR115-positive or KRT1-positive cells upon leaving the basal layer (Figure 2c). The notion that GPR115 + KRT1 − but no GPR115 − KRT1 + keratinocytes exist supports the hypothesis that GPR115 precedes the expression of KRT1. To investigate this, we turned to HaCaT-based organotypic skin equivalent models and performed loss of function studies.

Loss of ADGRF4 Negatively Regulates KRT1 Expression and Epidermal Differentiation
In 2D cultures, HaCaT keratinocytes formed monolayers in which 5-8% of the cells expressed KRT1/10, consistent with previous findings [13], and GPR115 (Figure 3a,b). Notably, when the experiment was repeated with GPR115KO HaCaT clones, the percentage of KRT1/10-positive cells declined obviously (Figure 3a,b), caused by a strong, up to 220fold decrease in KRT1 transcripts compared with HaCaT WT cells (Figure 3c). Notably, the level of remaining KRT1 transcripts and the percentage of KRT1/10-positive cells in the various clones were correlated. The moderate up to 12-fold reduction in KRT10 in GPR115KO HaCaT clones is compatible with the individual, but not pairwise, regulation of type I and type II keratins [23].

Loss of ADGRF4 Negatively Regulates KRT1 Expression and Epidermal Differentiation
In 2D cultures, HaCaT keratinocytes formed monolayers in which 5-8% of the cells expressed KRT1/10, consistent with previous findings [13], and GPR115 (Figure 3a,b). Notably, when the experiment was repeated with GPR115KO HaCaT clones, the percentage of KRT1/10-positive cells declined obviously (Figure 3a,b), caused by a strong, up to 220fold decrease in KRT1 transcripts compared with HaCaT WT cells (Figure 3c). Notably, the level of remaining KRT1 transcripts and the percentage of KRT1/10-positive cells in the various clones were correlated. The moderate up to 12-fold reduction in KRT10 in GPR115KO HaCaT clones is compatible with the individual, but not pairwise, regulation of type I and type II keratins [23]. Upon the 3D organotypic growth of HaCaT WT cells, schematically shown in Figure  3d, GPR115 was strongly induced in nearly all suprabasal cells, whereas most basal cells were GPR115-negative, as in a normal epidermis (Figure 3e,f). Remarkably, in HaCaT   Upon the 3D organotypic growth of HaCaT WT cells, schematically shown in Figure 3d, GPR115 was strongly induced in nearly all suprabasal cells, whereas most basal cells were GPR115-negative, as in a normal epidermis (Figure 3e,f). Remarkably, in HaCaT GPR115KO clones, the suprabasal cells remained negative for KRT1/10 (Figure 3e), indicating their likely inability to differentiate, and the keratinocyte layer formation looked disturbed.
To investigate this further, the number of keratinocyte layers in organotypic-grown HaCaT WT or GPR115KO clones was quantified (Figure 3g). The loss of ADGRF4 significantly decreased the number of suprabasal, differentiated epidermal layers, and thus the number of epidermal cells in organotypic constructs built from these two clones showed the lowest ADGRF4 levels (Figure 3h). To exclude that these differences between HaCaT WT cells and GPR115KO clones might be due to a direct regulation of proliferation by GPR115, their cell numbers were quantified in monolayer cultures where no difference was found (Figure 3i). Thus, HaCaT WT cells proliferate more compared to the GPR115KO clones under organotypic conditions but not under monolayer conditions. In sum, our data strongly indicate that GPR115 promotes epidermal differentiation and regulates KRT1.

Endogenous GPR115 Colocalizes with KRT1/10-Positive Filaments
In human epidermis, GPR115 was localized inside keratinocytes ( Figure 1d). A highly similar distribution was found in immunostained monolayered HaCaT keratinocytes (Figure 4a). Flow cytometry confirmed the intracellular localization of endogenous GPR115; HaCaT keratinocytes did not show GPR115 at the cell surface (Figure 4b). In contrast, Cos-7 cells transfected with GPR115 pcDNA3.1 localized GPR115 along the cell surface. One major difference between Cos-7 and HaCaT cells relates to the set of keratins they express. While the former are characterized by the simple epithelial keratins KRT8/18, the latter express KRT5/14 under proliferative and KRT1/10 under differentiation conditions.
To exclude the possibility that only a fragment of the GPR115 ECD locates intracellularly or that a C-terminal fragment might escape detection by the GPR115 ECD Ab, an Ab binding to the ICD (GPR115 ICD ) was applied ( Figure S1a, Table S1). This Ab confirmed the former results ( Figure 4c). Consistently, both Abs yielded an intracellular, filament-like pattern resembling that of cytoskeletal keratins (Figure 4a,c). Costaining with keratinspecific Abs showed a strong overlap of GPR115 with KRT1/10. All GPR115-positive cells expressed KRT1, and most were also positive for KRT10 (Figure 4d). To independently probe for this unexpected association of an aGPCR and a cytoskeletal keratin, a PLA was performed for endogenous proteins in HaCaT keratinocytes (Figure 4e), confirming the double immunofluorescence results.
To further substantiate the close proximity of epidermal GPR115 and keratins, we performed silver-enhanced immunogold electron microscopy using the GPR115 ECD Ab. In normal skin, immunogold particles were present in suprabasal, noncornified keratinocytes, where KRT1/10 represents almost all keratins, in addition to low, residual KRT5/14 (Figure 4f). The label was highly enriched at and restricted to bundles of keratin filaments (Figure 4g). Of note, the distribution of immunogold particles at the filaments showed a remarkable periodicity of~48 nm (Figure 4h). Given the length of~45 nm for the coiled-coil backbone of keratins, this is compatible with an association of GPR115 with the KRT1 head or tail domains [24]. Further, groups of silver-enhanced gold particles located along the desmosomes with a distance of~58 nm to the plasma membrane (Figure 4i), in approximate agreement with the known distance of keratins inserting at the C-terminus of desmoplakin [25]. In sum, GPR115 and KRT1 are highly colocalized inside suprabasal keratinocytes in vivo and in HaCaT keratinocytes, despite GPR115 being a 7TM receptor. cytes (Figure 4a). Flow cytometry confirmed the intracellular localization of endo GPR115; HaCaT keratinocytes did not show GPR115 at the cell surface (Figure 4b) trast, Cos-7 cells transfected with GPR115 pcDNA3.1 localized GPR115 along the face. One major difference between Cos-7 and HaCaT cells relates to the set of they express. While the former are characterized by the simple epithelial KRT8/18, the latter express KRT5/14 under proliferative and KRT1/10 under dif tion conditions.  (i) Alongside desmosomes (arrowheads), immunogold particle groups were also located at filaments. The distance of such groups to the desmosomes was 57.6 ± 1.3 nm (n = 8 pictures, 2-6 particle groups/picture, means ± SEM).

Endogenous GPR115 Is Unglycosylated and Uncleaved in HaCaT Keratinocytes
To gain some insight into principles that might explain this unexpected intracellular localization of GPR115, we turned to biochemical analysis. First, we probed whether endogenous GPR115 extracted from the epidermis and HaCaT keratinocytes represents the naked protein core, as predicted from its keratin-based localization. In total protein extracts, a strong band at~68 kDa was detected in the epidermis and HaCaT cells, and a weaker one was detected at 75 kDa in HaCaT cells (Figure 5a First, the blot was consecutively incubated with the rabbit GPR115 ECD and mouse KRT1/10 primary Abs. Afterwards both fluorophore-labeled secondary Abs were applied; pictures taken from this blot are shown (the single Abs in grey and their overlay colored). Lower part: same blot after stripping and restaining; the stripped blot was horizontally cut, and the three parts were incubated with either the KRT1, KRT5, or β-actin Abs.
To further substantiate this, HA-GPR115 was transfected into Cos-7 cells, which resulted in protein species that migrate as multiple bands at ~90-110 kDa (Figure 5a). Their extractability with RIPA buffer is consistent with that of other aGPCRs [16,17]. The GPR115 expressed in Cos-7 cells is N-glycosylated because, in PNGase F-treated lysates or in lysates of these cells cultured before with tunicamycin, the GPR115 molecular weight was reduced to 65 kDa (Figure 5c).
To examine whether human GPR115 was cleaved at the GPS as with most aGPCRs [10], N-terminal HA-and C-terminal Flag-tagged GPR115 was expressed in Cos-7 cells. Most human GPR115 remained noncleaved (Figure 5d), confirming data obtained from mouse GPR115 [9]. Because the molecular weights of endogenous GPR115 and of the de-N-glycosylated GPR115 expressed in Cos-7 cells are similar, endogenous GPR115 is very likely also not cleaved, arguing that the naked protein core localizes to KRT1/10.
Biochemical analyses of HaCaT WT cells and the GPR115KO clones further confirmed our data (Figure 5e). The amount of GPR115 and KRT1/10 protein decreased simultaneously in all HaCaT GPR115KO clones. (d) Lysates of Cos-7 transfected with constructs encoding N-terminal HA-and C-terminal Flag-tagged mouse or human GPR115 and human CD97 were analyzed with the indicated Abs by Western blot; the coding part of the constructs is shown schematically above (arrowhead: putative GPS). In the blots, arrowheads indicate uncleaved full-length human GPR115, and the arrows indicate the Nand C-terminal fragments of cleaved CD97 (positive control). Mouse GPR115, a little smaller than human GPR115 also was not cleaved. (e) Western blot analysis of urea-based lysates of HaCaT WT cells and the various GPR115 clones. The applied Abs are indicated on the right. Upper part: First, the blot was consecutively incubated with the rabbit GPR115 ECD and mouse KRT1/10 primary Abs. Afterwards both fluorophore-labeled secondary Abs were applied; pictures taken from this blot are shown (the single Abs in grey and their overlay colored). Lower part: same blot after stripping and restaining; the stripped blot was horizontally cut, and the three parts were incubated with either the KRT1, KRT5, or β-actin Abs.
To further substantiate this, HA-GPR115 was transfected into Cos-7 cells, which resulted in protein species that migrate as multiple bands at~90-110 kDa (Figure 5a). Their extractability with RIPA buffer is consistent with that of other aGPCRs [16,17]. The GPR115 expressed in Cos-7 cells is N-glycosylated because, in PNGase F-treated lysates or in lysates of these cells cultured before with tunicamycin, the GPR115 molecular weight was reduced to 65 kDa (Figure 5c).
To examine whether human GPR115 was cleaved at the GPS as with most aGPCRs [10], N-terminal HA-and C-terminal Flag-tagged GPR115 was expressed in Cos-7 cells. Most human GPR115 remained noncleaved (Figure 5d), confirming data obtained from mouse GPR115 [9]. Because the molecular weights of endogenous GPR115 and of the de-Nglycosylated GPR115 expressed in Cos-7 cells are similar, endogenous GPR115 is very likely also not cleaved, arguing that the naked protein core localizes to KRT1/10. Biochemical analyses of HaCaT WT cells and the GPR115KO clones further confirmed our data (Figure 5e). The amount of GPR115 and KRT1/10 protein decreased simultaneously in all HaCaT GPR115KO clones.

Transfected GPR115 Fails to Colocalize with KRT1 in Cos-7 Cells, Even upon Deglycosylation
One major difference between endogenous and transfected GPR115 resides in the extensive N-glycosylation of the latter. Thus, we asked whether this difference determined either the KRT1 association or the membrane localization. Upon the transfection of Cos-7 cells with a GPR115 variant in which most N-glycosylation sites were mutated (GPR115 mutGlyc, Table S3), its molecular weight decreased to~75-80 kDA; when treated with PNGase F, it decreased to 65 kDa (Figure 6a). The mutation of all N-glycosylation sites (Table S3) decreased the size to 63-65 kDa (Figure 6b). These N-glycosylation mutants lost their membrane localization, and unglycosylated GPR115 showed a diffuse intracellular distribution (Figure 6c,d). This was not surprising, given that Cos-7 cells lack KRT1.

Transfected GPR115 Fails to Colocalize with KRT1 in Cos-7 Cells, Even upon Deglycosylation
One major difference between endogenous and transfected GPR115 resides in the extensive N-glycosylation of the latter. Thus, we asked whether this difference determined either the KRT1 association or the membrane localization. Upon the transfection of Cos-7 cells with a GPR115 variant in which most N-glycosylation sites were mutated (GPR115 mutGlyc, Table S3), its molecular weight decreased to ~75-80 kDA; when treated with PNGase F, it decreased to 65 kDa (Figure 6a). The mutation of all N-glycosylation sites (Table S3) decreased the size to 63-65 kDa (Figure 6b). These N-glycosylation mutants lost their membrane localization, and unglycosylated GPR115 showed a diffuse intracellular distribution (Figure 6c,d). This was not surprising, given that Cos-7 cells lack KRT1. Thus, we transfected HaCaT keratinocytes with HA-tagged WT GPR115, which appeared partly at the cell surface (Figure 6e). The fraction of transfected GPR115 that was localized inside cells did not costain with keratins when using either GPR115 or the HAtagged antibody (Figure 6f)  Thus, we transfected HaCaT keratinocytes with HA-tagged WT GPR115, which appeared partly at the cell surface (Figure 6e). The fraction of transfected GPR115 that was localized inside cells did not costain with keratins when using either GPR115 or the HAtagged antibody (Figure 6f). Further, the localization of transfected GPR115 was similar in HA-GPR115-transfected keratinocytes of WT mice and of mice deficient for type I keratins (KtyI -/-) (Figure 6g). Thus, transfected GPR115 localized independent of keratins.
To address whether endogenous GPR115 might be associated with intracellular membranes via keratins, we focused on the recent finding that the peripheral endoplasmic reticulum (ER) partly exhibits nanometer proximity to keratin filaments [22]. We stained HaCaT keratinocytes using the GPR115 ECD Ab and Abs directed to reticulon 4 (RTN4), a marker for peripheral ER tubules, and to cytoskeleton-associated protein 4 (CKAP4), a marker of ER sheets. These proteins and GPR115 localized in distinctly different patterns (Figure 6h). Thus, the colocalization of endogenous GPR115 and KRT1 most unlikely does not occur along the peripheral ER.

Discussion
By now, the occurrence of actin in the nucleus, in addition to the cytoplasm, and of β-catenin at adherens junctions and in the nucleus represents textbook knowledge and serves as examples for many other proteins that display noncanonical localization and function [26,27].
Here, we provide strong evidence for the localization of GPR115, an aGPCR, inside keratinocytes in either skin cryosections or cultured HaCaT cells. In agreement with the reanalyzed epidermal scRNAseq data [1], we detected the protein in rare basal and in almost all suprabasal, nonkeratinized keratinocytes of normal skin by immunotechniques applying a carefully validated specific Ab directed to the ECD of this aGPCR. In psoriatic skin, characterized by excessive keratinocyte proliferation and altered differentiation, GPR115 expression was diminished. GPR115 also occurred in a subset of Ki-67-positive suprabasal keratinocytes. These data suggest an involvement of GPR115 in epidermal differentiation.
To further evaluate the roles of GPR115 and keratinocyte differentiation, organotypic cocultures of HaCaT keratinocytes with fibroblasts were established, conditions that induce differentiation and the formation of a stratified epithelium with all features of a normal epidermis [28]. In these constructs, almost all suprabasal keratinocytes coexpressed GPR115 and KRT1/10. Most notably, a CRISPR/Cas9-mediated deletion of ADGRF4 decreased the number of keratinocyte layers and abrogated KRT1. In similar experimental settings, a disruption of PTTG1 reduced epidermal stratification to the extent that cells appeared as a simple epithelium [1,29]. Disruptions of either HELLS or UHRF1 also resulted in a thinner epidermis and the suppression of epidermal homeostasis [1]. In light of these observations, it is likely that GPR115 not only regulates KRT1 but controls the early stages of epidermal differentiation, which is consistent with its expression in few proliferating basal cells [1]. In support, Kty1 -/mice showed a near-normal epidermis in which all epidermal layers were maintained, although these mice died a few hours after birth [23].
Our hypothesis that GPR115 is involved in epidermal differentiation is not contradicted by the lack of an overt skin phenotype in Adgrf4 -/mice [9]. Given that ADGRF4 clusters tightly with ADGRF2 and that both are only present in land-living mammals [9], they likely exert redundant functions in mouse skin. To what extent GPR111/ADGRF2, nearly absent in human keratinocytes [1], can substitute for GPR115 remains to be analyzed. Interestingly, Adgrf4 -/mice revealed that this protein plays an important role in enamel mineralization via the regulation of carboanhydrase-6 expression in ameloblasts [30].
The most unexpected finding of our study is the close colocalization of endogenous GPR115 and KRT1/10, which is strongly supported by immunogold electron microscopy. At first glance, such a colocalization appears unusual for a GPCR that is commonly present at and signaling from extra-and intracellular membranes [31,32]. Is it conceivable that only a GPR115 fragment such as the ECD or a part thereof associates alongside KRT1/10 filaments? Although not rigorously disproving this, the fsct that both GPR115 Abs, binding near the N-and C-terminus, respectively, reveal a close keratin association supports the view that intact GPR115 associates with keratins. Moreover, both human and mouse [9] GPR115 are not proteolytically cleaved at the GPS, and they display a similar molecular weight predicted for the nonglycosylated protein species in Western blots. This infers that full-length endogenous GPR115 is associated with KRT1/10.
The noncanonical, intracellular localization of an aGPCR has not been reported until now. At the same time, the noncanonical occurrence of cytoskeletal keratins is widely accepted, as exemplified by nuclear KRT17 and by extracellular KRT19 [33,34]. Given that GPCRs contain hydrophobic 7TM helices, it is unlikely that they exist as freely floating entities inside the cytoplasm but reside in membrane compartments, which is fully consistent with our localization data for transfected but not endogenous GPR115.
Having said this, KRTs 8, 18, and 19 have all been reported to functionally interact with the cystic fibrosis transmembrane regulator (CFTR), a multispan transmembrane receptor. The interaction of KRT8 with the CFTR∆F508 mutant, the most common mutation in patients with cystic fibrosis, at nucleotide-binding domain 1 (NBD1) inhibits the translocation of CFTR to the cell surface [35,36]. A deletion of KRT8 rescued CFTR cell surface localization [37]. KRT18 and KRT19, on the other hand, were shown to stabilize CFTR at the cell surface by direct protein interactions [38,39].
Most recently, epidermal keratins were identified in close association with peripheral ER compartments at desmosomes, where they are crucial to maintaining desmosomeassociated ER compartments, as seen by the disruption of ER-desmosome complexes upon the expression of a disease-associated KRT14 mutation [40]. Based on parallel staining of GPR115 and the ER markers RTN4 and CKAP4 in HaCaT keratinocytes, it appears unlikely that the GPR115 species that localize close to KRT1/10 reside in an ER compartment. Further experiments are required to identify the nature of the intracellular, KRT1/10-associated compartment in which GPR115 resides. This should pave the way to understanding the biological significance of GPR115's association with KRT1/10.