Animal experiments were approved by the Animal Care and Use Committee at the Rush University Medical Center (IACUC No 20-079) or the state government of Schleswig-Holstein (Protocol 119-10/17), depending on where the experiments were conducted. All experiments were performed by certified personnel.

Tissue from five patients with confirmed inflammatory EBA, defined as a neutrophilic inflammatory infiltrate, positive direct immunofluorescence, and positive dermal side indirect immunofluorescence with corresponding anti-COL7 IgG, was retrospectively collected. All patients involved in this study provided written informed consent. All experiments with human samples were approved by the ethical committee of the Rush University Medical Center (IRB No. 20121406) and were performed in accordance with the Declaration of Helsinki.

C57BL6/J were purchased from Jackson Laboratories (Bar Harbor, ME, USA) (JAX stock 000664). Sperm of C57BL/6-Clec4d^{tm1.1Cfg/Mmucd} (Clec4d^−/− ; MMRRC stock: 031935-UCD) and C57BL/6-Clec4e^{tm1.1Cfg/Mmucd} (Clec4e^−/− ) mice (MMRRC stock 031936-UCD), designated with the MMRCR stock numbers 031935-UCD and 031936-UCD, respectively, were purchased from the Mutant Mouse Research and Resource Center (MMRRC; Davis, CA, USA). The sperm was used for in vitro fertilization. The heterozygous progeny was used to breed Clec4d^−/− and Clec4e^−/− mice and their respective wild-type littermates. MRP8-Cre-ires/GFP (MRP8-Cre) mice were obtained from the Jackson Laboratory (Bar Harbor, ME, USA). A conditional knockout of Clec4n was generated using Clec4n^{tm2a(KOMP)Wtsi} embryonic stem (ES) cells acquired from the KOMP Repository (Davis, CA, USA). Using the fully verified and karyotyped C57BL/6N ES cells, chimeric founder mice were generated by microinjection into BALB/c blastocysts. The resulting mice with the “knockout first allele” were crossed with a C57BL/6N-Tg(CAG-Flpe) deleter mouse to remove the selection cassette and create the conditional Clec4n^tm1c allele, hereafter referred to as Clec4n^fl allele. Polymorphonuclear neutrophil (PMN)-specific Clec4n^−/− (Clec4n^ΔPMN ) mice were generated by crossing homozygous Clec4n^fl/fl with MRP8-Cre mice inducing an MRP8 promotor-driven Cre recombination excision event during neutrophil differentiation of precursors. To confirm the gene knockout in neutrophils, bone marrow neutrophils were isolated using the Neutrophil Isolation Kit (Miltenyi, Teterow, Germany), and their RNA was isolated using the RNA Mini Kit (Analytik Jena AG, Jena, Germany). cDNA was generated using the Thermo Scientific Revert Aid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, Bremen, Germany). cDNA was used for quantitative PCR using the SYBR™ Select Mastermix (Thermo Fisher Scientific, Bremen, Germany). Data were acquired using the RealPlex (Eppendorf, Hamburg, Germany) cycler (data not shown).

Recombinant murine von Willebrand factor A-like domain 2 (vWFA2) of the NC1 domain of collagen 7 (aa1048-1238) was produced, as previously described (16, 17).

New Zealand white rabbits were immunized with COL7^vWFA2 as previously described (18). IgG was purified from rabbit serum using Protein G Sepharose Fast Flow affinity column chromatography (Amersham Biosciences, Freiburg, Germany) as previously described (18). Reactivity of IgG fractions was analyzed by the immunofluorescence microscopy on murine skin. Concentrations of purified rabbit IgG were measured at 280 nm by a spectrophotometer. Total rabbit IgG underwent an additional purification step of antigen affinity purification using COL7^vWFA2 coupled Affi-Gel 10 (Bio-Rad, Munich, Germany) as previously described (19).

Mice were housed under specific pathogen-free conditions at Rush University (Chicago, IL, USA) or at the University of Lübeck (Lübeck, Schleswig-Holstein, Germany) and were provided standard mouse chow and acidified drinking water ad libitum. Mice were conditioned at least 2 weeks prior to experiments and were under 12 h:12 h light/dark cycles. Sex-matched mice, age 6–12 weeks were used for the experiments. For transcriptomic analyses, antibody transfer EBA was performed, as previously described (17). Affinity-purified anti-COL7^vWFA2 IgG (200 µg) was injected intraperitoneally thrice weekly for 2 weeks. In experiments examining the contribution of Dectin-2, Dectin-3, and Mincle to experimental EBA, disease was induced and evaluated, as previously described (4, 20). Briefly, New Zealand rabbits were immunized against three epitopes of type VII collagen. IgG directed to the epitope C (anti-COL7c) was affinity purified, and 50 μg of anti-COL7c was injected subcutaneously on days 0, 2, and 4 of the experiment. The percentage of the total body surface area affected by skin lesions (erythema, blisters, erosions, crusts, or alopecia) was determined by an investigator blinded to the experimental conditions. All clinical examinations and bleedings were performed under anesthesia.

For mouse tissue, additional perilesional tissue from previous passive transfer experiments and optimizations were used to minimize animal numbers. Hematoxylin and eosin stains were performed using standard protocols. For both human and mouse, FFPE blocks were sectioned routinely and stained with the following antibodies: CLE4D (MBS9607710, rabbit polyclonal, Mybiosource, San Diego, CA, USA), CLEC4E (BS08541R, rabbit polyclonal, Bioss, Woburn, Massachusetts, USA), and CLEC4N (clone IMG3D1, ab107572, mouse IgG3 monoclonal, Abcam, Boston, Massachusetts, USA). Human neutrophils were stained for MPO (AF3667, goat polyclonal, R&D Systems, Minneapolis, Minnesota, USA), while mouse samples were stained for Ly6G (clone RB6-8C5, SC-53515, rat monoclonal IgG2b, Santa Cruz, Dallas, Texas). Secondary antibodies and DAPI counterstain were purchased from ThermoFisher (Rockford, IL, USA). Mouse direct immunofluorescence was performed as previously described, using anti-rabbit secondary antibody to confirm successful passive transfer (17). All slides were immediately photographed following staining using an Evos FL microscope (ThermoFisher, Rockford, IL, USA). The frequency of dual-positive cells was quantified in ImageJ v1.52 (Bethesda, Maryland, USA)

Surface antigen staining was performed according to standard flow cytometry procedures. The following antibodies and reagents were used: Zombie aqua viability stain, anti-Ly6G (clone 1A8, 127628, Rat IgG2a), anti-CD45 (clone 30-F11, 103151, Rat IgG2b), and anti-CD11b (clone M1/70, 101243, Rat IgG2b), all of which were purchased from BioLegend (San Diego, CA, USA). Anti-CLEC4D (clone MA5-24152, mouse IgG2b) was purchased from ThermoFisher (Rockford, IL, USA) and conjugated with APC using the Lightning-Link APC Conjugation Kit (Abcam, Boston, Massachusetts, USA). Recombinant anti-CLEC4N targeting the extracellular domain (50267-R001, Rabbit IgG, Sino Biological, Wayne, PA, USA), was conjugated with PE using the Lightning-Link PE Conjugation Kit (Abcam, Boston, Massachusetts, USA) and anti-CLEC4E (BS-8541R, Rabbit polyclonal, ThermoFisher, Rockford, IL, USA), was conjugated with Alexafluor-488 using the Lightning-Link Alexafluor488 Conjugation Kit (Abcam, Boston, Massachusetts, USA). Flow cytometry was performed on an LSRFortessa (BD Biosciences, San Jose, CA, USA), using compensation beads (BioLegend, San Diego, CA, USA) and FACSDiva software (BD Biosciences, San Jose, CA, USA). Gating was performed by removing doublets and dead cells, followed by use of fluorescent minus one control. Flow cytometry data were analyzed using FCS Express 7 Plus software (De Novo Software, Pasadena, CA, USA).

Blood was collected by retrobulbar bleeding from anesthetized mice, and immediately followed by RBC lysis (BioLegend, San Diego, CA, USA). Lesional (ear) skin from anesthetized mice was harvested by punch biopsies and enzymatically digested using the whole skin digestion kit (Miltenyi, Auburn, CA, USA) per manufacturer instructions with minor modifications. Following addition of enzymes, samples were incubated in a water bath at 37°C for 2 h and shaken every 15 min. Mechanical dissociation by pipetting was only performed at the end of the incubation. Cells were then filtered using a 70-μm filter, centrifuged at 300g for 10 min at 4°C, washed and resuspended in 0.04% BSA/HBSS (ThermoFisher, Rockford, IL, USA). Cell viability was immediately assessed by trypan blue (ThermoFisher, Rockford, IL, USA). Single-cell suspensions of 1,000 cells/μL were run into the chromium Controller (10x Genomics, Pleasanton, CA, USA) to encapsulate 5,000 and 10,000 cells for blood and skin samples, respectively. Ambion RNase inhibitor (Invitrogen, Waltham, Massachusetts, USA) was added to the master mix. Samples were processed using the Chromium Single Cell 3′ GEM, Library & Gel Bead Kit v3 (10x genomics, Pleasanton, CA, USA) per manufacturer’s instructions. Library QC was performed using the Agilent Tape station. Sequencing was performed using two NovaSeq 6000 SP lanes (Illumina, San Diego, CA, USA). Cell capture, RT, library preparation, and sequencing were performed at the University of Illinois at Chicago.

Raw FASTQ reads were mapped to mouse mm10 reference genome (GRCm38.93.dna/GRCm38.93.gtf) using Cell Ranger Version 3.0.0. Web summary alignment metrics are provided in Supplementary Table 3 .

Raw, digitized count matrices were pre-processed and doublets/multiplets were simulated using Single-Cell Remover of Doublets (Scrublet) (21) (version 0.2.1) with standard parameters enabled. Putative singlets were filtered in Seurat to remove low-quality cells and kept for downstream analysis if and only if they met the following user-defined, collective quality control metrics criteria: (a) 350 < genes/cell < 5,000, (b) cells contained no more than 10% mitochondrial gene expression, and (c) cells were not defined as outliers (22).

We performed anchoring and integration of pathogenic IgG or non-pathogenic IgG-treated mouse blood or pathogenic IgG or non-pathogenic IgG-treated mouse skin datasets using the Seurat package (Version 3.2.2, R Studio version 3.6.1) (23) as suggested by the developer. Briefly, Seurat objects were created using Scrublet-pre-processed individual, raw digitized count matrices and merged. Individual gene expression digital matrices were normalized, and the top 2,000 variable genes/features were identified. Integration anchors were identified using the dimensions set to 30, and datasets were integrated with dimensions set to 30. The integrated object was scaled, and significant principal components used for clustering and finding neighbors were identified using a combination of statistical and heuristic methods. Neighbors and clusters were then identified with dimensions specified by user and visualized using Uniform Manifold Approximation and Projection (UMAP). Putative cell identities were identified based on differential gene expression profiles (cluster markers) (log fold change of 25% with a minimum of 25% of cells expressing such gene in either one cluster). Genes were log-normalized and visualized as two-dimensional feature plots or bubble plots.

Differentially expressed genes between cell types across treatments were calculated with Seurat. Differentially expressed genes for a particular comparison were filtered using a p-adjusted threshold of 0.05 and a Log₂ fold-change of 0.58 or greater in either direction.

Differentially expressed genes were used to calculate Gene Ontologies (GOs). Manual curation of GOs was performed using Enrichr (24) and visualized as bubble plots. A list of significant GOs (p-adjusted < 0.05) is included in Supplementary Table 4 .

Aggregate, composite gene scores were assigned to neutrophils using the AddModuleScore function in Seurat. This “Neutrophil aggregate score” was defined by a core set of known neutrophil-associated marker genes, including S100a8, S100a9, and Ly6g. Aggregate, composite gene scores were log-normalized and visualized as two-dimensional feature plots.

Pseudo-ordering of individual cells was performed using Monocle2 (Version 2.10.1) (25). Briefly, neutrophils were bioinformatically gated in Seurat and a cellDataSet object was created in Monocle2. Subclustered cells were ordered based on variable features and dimensionality reduction was performed with the reduction method = ’DDRTree’ argument enabled and then ordered. To identify pseudotime-dependent gene expression changes in neutrophils in the putative blood–skin transition, we applied the single-cell Energy path (scEpath; Version 1; MATLAB Version 9.5) (26) algorithm on a subset of Monocle2-ordered neutrophils. To identify statistically significant pseudotime-dependent gene changes, we compared the standard deviation of the observed smoothed expressions with a set of similarly permuted expressions by randomly permuting the cell order (i.e., 100 permutations) as suggested by the developer. We considered all genes with a standard deviation greater than 0.01 and a Bonferroni-corrected p-value below a significance level α = 0.05 to be pseudotime-dependent. To analyze pseudotime-dependent mouse transcription factors, we used genes annotated in the Animal Transcription Factor Database (AnimalTFDB 3.0) (27) in scEpath. Pseudotime-dependent genes were represented and visualized using a rolling wave plot with user-defined optimal K clusters.

Percentages of CLEC4-positive neutrophils between blood in EBA versus control or EBA blood versus skin were determined by log-transformation of the percentage of dually positive cells, followed by Student’s t-test. Comparisons of the body surface area affected by skin lesions were performed by two-way ANOVA using GraphPad Prism 9 (GraphPad, San Diego, CA, USA), and comparisons of the area under the curve of the affected body surface area over time were performed using Student’s t-test. All tests were two sided with a p < 0.05. Given the relatively few numbers of comparisons and lack of statistically significant results, correction for multiplicity was not performed. All data are reported as mean ± SEM.

Mice were injected thrice weekly for 2 weeks with pathogenic rabbit anti-mouse antigen affinity-purified COL7^vWFA2 IgG or with non-pathogenic rabbit IgG, as previously described (17). As expected, the group treated with pathogenic IgG consequently developed a characteristic disease phenotype with positive direct immunofluorescence against the basement membrane zone. We then performed scRNA-seq on unsorted cells from RBC-lysed blood ( Figure 1A ) or ear skin ( Figure 2A ) in experimental EBA mice or controls at day 14. Following doublet removal and quality control filtering ( Supplementary Figure 1 ), 7,143 cells (k_EBA = 2,878 vs. k_IgG = 4,265) were anchored and integrated and included in our blood analysis ( Figures 1B, C ), and 8,173 cells (k_EBA = 3,445 vs. k_IgG = 4,728) in our skin analysis ( Figures 2B, C ).

Unsupervised clustering of blood cells was performed with Seurat (28), revealing eight distinct clusters. Using differentially expressed gene signatures, we assigned clusters based on their putative identities and hierarchical similarities. As expected, these clusters were primarily composed of immune cells ( Figures 1C, D ), of which lymphocytes were most abundant followed by neutrophils ( Supplementary Figure 2 ). We next compared the ratio of immune cells per cluster post-normalization between the pathogenic IgG and the non-pathogenic IgG-treated groups to identify shifts in immune populations because of disease. Neutrophils and myeloid cells appeared in larger amounts in samples from pathogenic IgG-treated mice relative to non-pathogenic IgG-treated mice ( Supplementary Figure 2 ). Putative neutrophils were identified based on an aggregate, composite gene score including S100a8, S100a9, and Ly6g. Notably, two-dimensional UMAP demonstrates an apparent neutrophil activation in EBA samples that is absent in IgG control samples, seen in an expanding population emanating from the basal neutrophil populations ( Figure 1E ). Bioinformatic gating was subsequently performed to isolate neutrophils based on these putative markers, with resolution decreased to consider neutrophils as a single, granular population for each treatment condition. Wilcoxon Rank Sum Test revealed 148 upregulated and 375 downregulated genes between neutrophils from the pathogenic IgG and the non-pathogenic IgG-treated groups, defined as >0.25 fold-change difference, and a p-adjusted < 0.05 ( Supplementary Table 1 ).

We next characterized the cutaneous transcriptome in active lesions, with an emphasis on capturing viable lesional-skin neutrophils. Given neutrophil fragility following skin dissociation and flow sorting, dead cell depletion, or alternative dissociation protocols (unpublished), we capitalized on our scRNA-seq workflow to minimize such manipulation. Thus, following dissociation, whole skin dissociates were immediately processed for single-cell capture. Viability threshold of >80% was set without dead cell depletion to minimize dead cell background signal but allow capture of viable neutrophils. Using differentially expressed gene signatures, we assigned clusters based on their putative identities and hierarchical similarities as before ( Figure 2D , Supplementary Table 2 ). Use of our putative neutrophil aggregate marker score highlighted a cutaneous neutrophil population in diseased but not IgG control skin ( Figure 2E ).

To better understand the trajectory of neutrophil activation, whole blood and skin neutrophils were combined into a single object using bioinformatic gating. Cells were then aligned in pseudotime using Monocle2 (25), with pseudotime-dependent genes identified in scEpath (26). Trajectory analysis revealed pseudo-infiltration from activated blood to intralesional neutrophils ( Figures 2H, I ). Control blood, EBA blood, and EBA skin each exhibited unique transcriptomes, of which the top 10 differentially expressed genes between each comparison are shown ( Figure 2F ). Gene ontology analyses with Enrichr (24) of the activated and cutaneous neutrophils versus basal neutrophils revealed increased neutrophil degranulation in blood and enhanced cytokine signaling in the skin ( Figure 2G ). While several genes demonstrated pseudotemporal activation from blood to skin, genes for CTRLs collectively displayed a clear pattern of regulation with several stimulatory CTLRs upregulated and inhibitory CTLRs downregulated ( Figure 2J ; Supplementary Figure 3 ). This intriguing finding prompted us to investigate these markers and their functional significance in more detail.

Next, we characterized the protein expression levels of Dectin-3, Mincle, and Dectin-2, the proteins encoded by Clec4d, Clec4e, and Clec4n, respectively, on neutrophils. To this end, skin sections from mice with EBA were stained for Ly6G and either Dectin-3, Mincle, or Dectin-2 revealing co-expression in the skin ( Figure 3B ). To distinguish whether this activation occurred in the blood or in the skin, we performed flow cytometry on blood from mice treated with COL7 antibodies or corresponding isotype controls. As previously described (17), the frequency of CD45⁺Ly6G⁺ neutrophils increased significantly in the EBA compared to the control group ( Figure 3A ). Dectin-2 was highly expressed in both basal and activated blood neutrophils (>95%), while much lower levels of Dectin-3 and Mincle were detected. There was no significant difference between control and EBA neutrophils ( Figure 3C ) The percentage of neutrophils expressing Mincle and Dectin-3 was significantly increased in the skin relative to blood neutrophils (p < 0.001). We next sought to confirm the expression of these three CLTRs in human EBA patients by assessing formalin-fixed paraffin-embedded (FFPE) sections retrospectively. These sections were co-stained for myeloperoxidase (MPO) and either Dectin-2, Dectin-3, or Mincle. Skin neutrophils from patients with inflammatory EBA demonstrated consistent co-expression of Dectin-2, Dectin-3, and Mincle ( Figure 3D ). MPO-positive cells were uniformly positive for each CTLR. CTLRs were also detectable in few MPO-negative cells with a mononuclear morphology.

We analyzed the functional significance of Dectin-3, Mincle, and Dectin-2 in EBA. To this end, we examined the course of EBA in globally genetically deficient Clec4d^−/− and Clec4e^−/− mice as well as in neutrophil-specific Clec4n^−/− mice. The latter were generated by breeding Clec4n^fl/fl with Mrp8^+/Cre mice. The clinical course of skin inflammation as well as the histopathology of skin lesions were analyzed, as described in the Materials and methods section. Genetic deficiency of each CTLR did not change the course of disease at the clinical or the histopathological level ( Figure 4 ).