Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Jul 31.
Published in final edited form as: Science. 2026 Jun 25;392(6805):eaeb3900. doi: 10.1126/science.aeb3900

Ubiquitin-like proteins NEDD8 and SUMO2 control epithelial homeostasis, regeneration, and inflammation

Mårten CG Winge 1,†, Leandra V Jackrazi 1,2,†, Douglas F Porter 1, Suhas Srinivasan 1, Vanessa Lopez-Pajares 1, Dayan J Li 1,3, Benjamin Pham 3, Aubrey Houser 4, Spencer H Cha 1, Robin M Meyers 1,5, Lisa A Ko 1, Luca Ducoli 1, Weili Miao 1, Lindsey M Meservey 1, Brian J Zarnegar 1,6, Mark Smith 7, Andrew L Ji 4, Michael T Longaker 3, Paul A Khavari 1,2,8,*
PMCID: PMC13421985  NIHMSID: NIHMS2190488  PMID: 42348677

Abstract

Stratified epithelial differentiation involves transcriptional and proteomic remodeling. Here, multi-omic profiling implicated ubiquitin and related post-translational networks in differentiation dynamics. Systematic perturbation of ubiquitin-like machinery in primary human keratinocytes uncovered opposite functions of NEDD8 and SUMO2. Generation of conditional knockout mice established essential roles for NEDD8 in progenitor maintenance, skin regeneration, and inflammation, whereas SUMO2 was required for differentiation. Beyond ubiquitin-proteasome-concordant changes, NEDD8 directed proteomic regulation correlated with RNA abundance. Integration of immunoprecipitation mass spectrometry (IP-MS) with genome-wide suppressor screening revealed context-specific NEDDylation dependencies. Among effectors, heterogeneous nuclear ribonucleoprotein U (HNRNPU) emerged as a post-transcriptional regulator of epithelial cell state whose RNA binding repertoire was modulated by NEDDylation. Thus, NEDD8 and SUMO2 play opposite roles in epithelial homeostasis, regeneration, and inflammation, demonstrating multiple ways ubiquitin-like networks govern tissue homeostasis.

Epithelial barrier function, regeneration, and inflammation are essential for terrestrial life (1). Homeostasis in stratified epithelia balances progenitor self-renewal with differentiation mediated barrier formation. This balance is augmented by the ability to regenerate after injury and the capacity for inflammatory responses to infection (2,3). Multiple signaling pathways converge on the proteome in such contexts to mediate transcriptional and post-translational control (4). Keratinocytes embedded in a lipid-protein matrix express a proteome specialized to provide barrier protection and immunosurveillance (5–9). Disruptions to epidermal homeostasis are hallmarks of psoriasis, failed wound healing, and squamous cancers (10–12).

The transcriptomic and epigenomic landscape of epidermal differentiation is highly dynamic and involves over 40,000 regulatory elements linked to more than 3,500 genes (13). Proteomic diversity is equally pronounced, varying markedly between cell types and among keratinocytes within the epidermis (14). These observations suggest that proteomic remodeling is a fundamental feature of stratified epithelial differentiation. We examined regulatory mechanisms of the ubiquitin-proteasome pathway and related ubiquitin-like proteins (UBLs) using single-cell perturbation screening, spatial transcriptomics, quantitative proteomics, and genome-wide CRISPR screening. We identified essential and opposite roles for UBLs NEDD8 and SUMO2 in epidermal homeostasis, inflammation, and tissue healing after injury.

Concordant and divergent RNA–protein dynamics during epidermal stratification

We performed multi-omic profiling across serial timepoints in a well-characterized in vitro model of keratinocyte differentiation (13). This included RNA sequencing, label-free liquid chromatography-tandem mass spectrometry (LC-MS/MS) proteomics with data-independent acquisition (DIA), and di-glycine remnant (K-ε-GG) immunoprecipitation mass spectrometry (diGly IP-MS) (Fig. 1A). This approach captured 8,526 RNA transcripts, 9,141 proteins, and 31,298 diGly modified peptides (fig. S1, A to F; Data S1). K-means clustering of dynamically regulated proteins identified five major expression trajectories (Fig. 1B). Gene Ontology (GO) analysis revealed functional enrichment for terms consistent with known biological processes active during epithelial differentiation (Fig. 1C).

Fig. 1. UBLs as opposite regulators of epidermal differentiation.

Fig. 1.

Open in a new tab

(A) Schematic overview of experimental integration of RNA-seq (n=2, 10 timepoints), label-free DIA-MS proteomics (n=2, 10 timepoints), and di-glycine remnant (diGly, K-ε-GG) IP-MS (n=3, 3 timepoints) across primary human keratinocyte in vitro differentiation. (B) K-means clustering (k=5) of dynamic proteins. (C) Heatmap of dynamic protein clusters with associated GO: BP enrichments. (D) Percentages of clustered dynamic proteins with multiple dynamic diGly modifications. (E) Alluvial plot of RNA-protein expression dynamics. (F) Perturb-seq library design (boxed top left). UMAP embedded transcriptional profiles of gene level effects in Perturb-seq screen. Points represent an aggregated perturbation signature (bottom, n=4 guides/gene). (G) Volcano plot of Perturb-seq gene expression changes along pseudotime, relative to safe-target controls (top). Aggregated NEDDylation (green) and SUMOylation (pink) pathway scores (bottom). (H) UBL conjugation cascade schematic with pharmacological inhibition strategy. (I) Bulk RNA-seq after 48hr UBL inhibitor treatment (n=2). Differentiation transcripts (orange), progenitor transcripts (blue). (J) Quantification of differentiation (K10 intensity), proliferation (Ki-67+ cells), and apoptosis (TUNEL+ cells) (n indicated by dots for each graph) after UBL knockdown and rescue. (K) RNA-seq of differentiation gene expression after siNEDD8 (left) or siSUMO2 (right) knockdown, with (gray) and without (colored) rescue (n=2). Graphs show mean ± SEM. Statistical analyses: (D) Fisher’s exact test. (G) Mann-Whitney U-test. (E, I, K) Wald test in DESeq2 with BH FDR correction. (J) Unpaired two-tailed t-test with Welch correction. *=p<5e-2, **=p<1e-2, ***=p<1e-3; ns, not significant.

Abbreviations: DIA-MS, Data independent acquisition mass spectrometry; UBL, ubiquitin like protein; NEDD8i, pevonedistat (MLN-4924); SUMOi, subasumstat (TAK-981); diGly, diglycine; KO KC, knockout keratinocyte; sc, stratum corneum; e, epidermis; d, dermis.

DiGly IP-MS mapped proteome-wide ubiquitin/UBL conjugation and revealed a significant enrichment for multiple modifications amongst differentiation dynamic proteins (Fig. 1D). Rate modeling compared mRNA and protein dynamics using matched RNA and proteomic profiles. This identified both concordant and discordant RNA-protein differentiation trajectories. Some transcripts remained stable despite pronounced protein abundance differences (Fig. 1E; fig. S1, G to I).

Proteins with decreasing expression trajectories during differentiation were enriched for multiple diGly modifications (fig. S1, J to N). This implicated post-translational regulatory mechanisms in the modulation of protein abundance. Ubiquitin and ubiquitin-like proteins are known to regulate protein stability and function, yet the core ubiquitination machinery remained stable across differentiation (fig. S2, A to D) (15). This contrast nominated ubiquitin/UBL pathways as selective drivers of epithelial proteomic remodeling and motivated interrogation of specific UBL circuits.

Opposite roles of specific UBLs in epidermal differentiation

To test functional roles of ubiquitin-related networks in keratinocyte differentiation, we performed Perturb-seq (CRISPR-based gene perturbations combined with single-cell RNA sequencing (16)). The sgRNA library targeted 202 genes involved in the ubiquitin–proteasome system, NEDDylation and SUMOylation pathways, predicted co-essential genes, differentiation regulators, and non-targeting controls (Fig. 1F, fig. S3, A to C; Data S2). Perturb-seq analysis demonstrated depletion of essential guides and reduced sgRNA target mRNA expression, validating CRISPR activity in the screen (fig. S3, D and E). Pseudotime differentiation trajectories established opposite roles for NEDDylation and SUMOylation (fig. S3, F to H). Perturbation of NEDDylation components induced differentiation, whereas disruption of SUMOylation machinery prevented it (Fig. 1G; fig. S3I). RNA-sequencing of individual siRNA knockdowns supported these findings (fig. S3, J to L). Pharmacologic inhibition using pevonedistat (NAE1 inhibitor; NEDD8i) and subasumstat (SAE1 inhibitor; SUMOi) recapitulated these phenotypes in primary human keratinocytes (Fig. 1, H to I; Data S3) (17, 18). Inhibitor doses did not affect viability or modulate off-target pathways (fig. S4, A to D). Thus, NEDD8 was necessary to sustain progenitor gene expression whereas SUMO2 enabled differentiation.

To validate these opposite effects in a three-dimensional tissue, we engineered reconstituted human skin organoids with genetic depletion or pharmacologic UBL conjugation inhibition (19, 20). Although this system is limited by spatial distribution of progenitor markers such as keratin 14, both pharmacologic and genetic UBL modulation in skin corroborated screening results (fig. S4, E and F) (21). NEDD8i reduced proliferation in both primary human and mouse keratinocytes (fig. S4G). Re-expression of the corresponding UBL using siRNA-resistant lentivectors reversed phenotypic effects of UBL depletion (Fig. 1, J and K, fig. S4, H to J). To test the requirement for UBL conjugation, we generated SUMO2 and NEDD8 conjugation deficient mutants (CM) lacking the C-terminal GG motif (fig. S5, A and B). Wild-type (WT) but not CM SUMO2 rescued expression of differentiation transcripts in primary human keratinocytes (fig. S5C). Conversely, WT but not CM NEDD8 rescued progenitor RNA levels (fig. S5D). In colony forming assays, NEDD8 knockout keratinocytes formed fewer colonies, whereas SUMO2 knockouts had increased colony formation (fig. S5, E to H). To investigate possible convergence of UBL signaling dependencies, we depleted SUMO2 followed by NEDD8i treatment. NEDD8i modulated interaction effects included cell cycle programs (fig. S5I). NEDD8i treatment or NEDD8 knockdown also impaired in vitro migration velocity, effects rescued by WT but not CM NEDD8 (fig. S5, J and K). We intersected NEDD8i transcriptional signatures generated in additional primary human stratified epithelial cell types (cervix, cornea) to identify a conserved NEDD8i-induced differentiation program (fig. S5, L to O).

Nedd8 and Sumo2 knockout mice

To validate findings with human epithelial cells and skin organoids in living animals, we generated tamoxifen-inducible epithelial specific knockout mice (Krt14-CreERT) for Nedd8 (Nedd8fl/fl, Krt14-CreERT) and Sumo2 (Sumo2fl/fl, Krt14-CreERT) (Fig. 2A; fig. S6A). After recombination, both sets of mice developed visible skin phenotypes (Fig. 2B; fig. S6B) following target protein depletion throughout the epidermis (Fig. 2C). This indicated that both UBLs were necessary for normal gross skin morphology. Consistent with human data, gene set enrichment analysis (GSEA) of knockout mouse skin bulk RNA-sequencing exhibited opposite transcriptional effects on differentiation. Nedd8−/− skin upregulated cornified envelope (CE) genes whereas Sumo2−/− skin showed repression (Fig. 2D). Histological analysis of UBL knockout epidermis revealed epidermal hyperplasia, hyperkeratosis, and parakeratosis in Nedd8−/− skin. This contrasted with the impaired stratification and moderate hyperplasia seen after Sumo2−/− knockout. Again, Nedd8 ablation induced premature differentiation, while Sumo2 knockout prevented differentiation (Fig. 2, E and F). There was no corresponding apoptosis or senescence (fig. S6, C and D). Systemic UBL deletion in Krt14 positive tissues affected stratified epithelia in the tongue, palate and rectum (fig. S6, E and F). This demonstrated consistent effects of Nedd8 and Sumo2 loss across diverse epithelial tissues.

Fig. 2. UBLs control epithelial homeostasis and regeneration.

Fig. 2.

Open in a new tab

(A) Generation of Nedd8 and Sumo2 K14-CreERT driven conditional knockout mice. (B) Rep. micrographs of UBL knockout mice (n=10). (C) Confocal microscopy of immunostaining of Nedd8 (red, left) in and Sumo2 (red, right) (n=3). Scale bar 50 μm. (D) Bulk RNAseq GSEA of cornified envelope (CE) genes in Nedd8−/− (top) or Sumo2−/− (bottom) versus controls (n = 3). (E) H&E staining (left), K10 IHC (middle, brown), Ki-67 IHC (right, brown) in dorsal skin from rep. control and UBL knockout mice. Scale bar 50μm (H&E) / 25μm (IHC). (F) Quantification of differentiation (K10 intensity) and proliferation (Ki-67+ cells) in mouse skin (n =3). (G) Trans epidermal water loss (TEWL) measurements in UBL knockout and control mice (Sumo2 n=3, Nedd8 n=7, controls n=7). (H) Rep. wound images at POD 14 (n=4/group). White dashed initial edge, black dashed 14-day edge. (I) Re-epithelialization measurements for 14 days post full thickness skin wounding (n=4). (J) Spatial transcriptomic analysis of wound sections with annotated immune cell types (monocytes/macrophages, red; neutrophils, yellow; T cells, blue). Dots represent active wound remodeling (Krt16, Krt6a, Krt17, Lamc2, Areg, Itgb1, Mmp13, Mmp9; cyan) or maturation (Krt10, Flg, Dsg1b, Dsg1a, Grhl3, Klf4; red) transcripts. (K) Spatial transcriptomics quantification of activation transcripts (left), maturation transcripts (right) at specified distances from wound edges. Graphs show mean ± SEM. *=p<5e-2, **=p<1e-2, ***=p<1e-3. Statistical analyses: (D) One-tailed GSEA (weighted Kolmogorov–Smirnov–like statistic, NES from adaptive multilevel permutation). (F, G) One-way ANOVA with Sidak’s correction. (I) Two-way ANOVA with Sidak’s correction. P values as indicated. H&E, hematoxylin and eosin; POD, post-operative day; ns, not significant; IFE, interfollicular epidermis; CE, cornified envelope; GO, Gene Ontology; ST, spatial transcriptomics; DEG, differentially expressed genes; rep, representative.

Integration of transcriptomic and proteomic data from Nedd8−/− mice, resolved 495 genes modulated at both the RNA and protein levels. This included upregulation of CE formation, Nfe2l2 (Nrf2) signaling, and neutrophil degranulation pathways (fig. S6G). The opposite impacts of Nedd8 and Sumo2 on terminal differentiation programs prompted functional assessment of formation of the cutaneous barrier. Barrier integrity was measured by trans-epidermal water loss (TEWL). Both Nedd8−/− and Sumo2−/− mice displayed significantly increased TEWL (Fig. 2G). Thus, normal skin barrier function required both UBLs. Consistent with this, topical application of NEDD8i or SUMOi to mice led to dose-dependent impairment of epidermal barrier formation and differentiation (fig. S6, H and I). Regenerative capacity was also assessed following fullthickness wounding (22). Nedd8−/− skin failed to re-epithelialize at 14 days, while Sumo2−/− wounds retained normal healing capacity (Fig. 2, H and I). Spatial transcriptomics of Nedd8−/− wounds revealed an enrichment of immature barrier transcripts at the wound edge and a corresponding mature transcript depletion. NEDD8 loss increased neutrophil infiltration at the wound edge, indicating a persistent acute wound phase (Fig. 2, J and K). Finally, barrier function was examined under inflammatory stress. After topical TLR7 agonist imiquimod, Nedd8−/− skin had enhanced barrier dysfunction, while Sumo2−/− skin showed no additional impairment (fig. S6, J and K) (23). Nedd8−/− tissue displayed altered transcriptional responses to imiquimod compared to wild-type controls (fig. S6L). Nedd8 and Sumo2 were both necessary for normal epidermal barrier function, however re-epithelialization and inflammatory responses to TLR7 stimulation required only Nedd8.

UBL loss alters the epithelial microenvironment

Integration of single-cell RNA-seq and spatial transcriptomics of skin from Nedd8 and Sumo2 knockout mice revealed distinct microenvironmental shifts (Fig. 3A; fig. S7A; Data S4). Nedd8 deficient skin accumulated neutrophils. Sumo2 deletion induced T cell infiltration most pronounced in the dermis (Fig. 3A; fig. S7, B to F). Nedd8−/− skin showed depletion of basal keratinocytes and emergence of a distinct population termed “NEDD8 KO KCs” within the interfollicular epidermis (IFE) (Fig. 3, B and C). Cell communication analysis identified rewired keratinocyte-immune signaling in both models. In Nedd8−/− skin, altered ligand-receptor interactions occurred between keratinocytes and neutrophils via the known neutrophil inciter axis Anxa1-Fpr1/Fpr2 (fig. S7G) (24). Conversely, neutrophil-derived signals involved Tnf-Tnfrsf1a and Sema4d-Plxnb2, known to be associated with neutrophil-driven skin inflammation (25, 26). In Sumo2−/− skin, T cell infiltration accompanied increased epithelial adhesion signals, including Cdh1 on keratinocytes and Itgae/b7 on T cells. T cell derived Tgfb1 engaged keratinocyte Tgfbr1/2. These interactions mediate T cell adhesion, localization, and retain resident memory T-cell niches (27, 28). T cells communicated with Sumo2−/− keratinocytes via Tnf-Tnfrsf1a and Sema4d-Plxnb2 (25, 29) (fig. S7H). Infiltrating T cell subsets included NKT, Th1, Th2, Th17 and T-regs (fig. S7, I to K). Spatial transcriptomics resolved the localization of intraepithelial DETC and intradermal T cells (fig. S7L). Immune and stromal signaling network analysis in Nedd8−/− and Sumo2−/− skin showed communication between IFE keratinocytes with fibroblasts and monocytes/macrophages (fig. S8, A and B). Both programs highlighted basal keratinocytes as a major source of Notch and Wnt ligands (fig. S8, C and D). In Nedd8−/− skin, fibroblasts secreted Nrg, Thbs, Cypa, and Bmp family signals (fig. S8E). In Sumo2−/− tissue, fibroblasts expressed galectins, Fn1, laminins, and collagens (fig. S8F). Spatial transcriptomics mapped neutrophil and T cell infiltration in UBL knockout skin (Fig. 3C). Thus, epidermal UBL loss perturbed cell intrinsic homeostasis and led to differentiation dependent reprogramming of both stromal and immune crosstalk.

Fig. 3. Single cell and spatial profiling refine UBL-dependent epithelial state transitions.

Fig. 3.

Open in a new tab

(A) UMAP of Harmony-integrated scRNA-seq showing skin cellular composition with cell type annotations (pooled n=3 per group Sumo2−/− and Sumo2+/+; pooled n=6 per group for Nedd8−/− and Nedd8+/+). Insets: neutrophil and T cell counts per genotype. (B) IFE composition in UBL knockout and control mice. (C) Spatial transcriptomics (ST) of control and knockout mice with annotated IFE compartments (suprabasal KC, cyan; basal KC, purple; NEDD8 KO-specific keratinocyte, red) and immune cells (neutrophils, yellow; T cells, blue). Scalebar 100 μm. (D) scRNA-seq differentiation signature scores in basal keratinocytes from control and UBL knockout mice. (E) scRNA-seq Nrf2 signature in control and Nedd8−/− basal and NEDD8 KO specific keratinocytes. (F) Generation of Krt14-CreERT Nedd8−/− x Nrf2tm1Ywk mice. (G) H&E of rep. control and knockout mice five days post tamoxifen-induced recombination. (H) Quantification of barrier dysfunction (TEWL, n=8 Nedd8+/+ Nrf2+/+, n=6 Nrf2−/−, n=5 Nedd8−/−, n=5 Nedd8−/− Nrf2−/−). (I) scRNA-seq differentiation signature scores in basal keratinocytes from control and knockout mice. (J) GO: BP enrichments for DEGs from NEDD8 KO-specific keratinocytes in Nedd8−/− and Nedd8−/− Nrf2−/− mice. (K) ST of control and knockout mice with IFE cell type annotations (suprabasal KC, cyan; NEDD8 KO KC, red; basal KC, purple; hair follicle KCs, gray). Scalebar 200 μm. (L) Spatial density map of inflammatory transcripts from ST. Scalebar 100μm. Graphs show mean ± SEM. *=p<5e-2, **=p<1e-2, ***=p<1e-3. Statistical analyses: (D, E, I) Kruskal-Wallis test with Dunn post-hoc comparisons. (H) One-way ANOVA followed by Tukey post hoc test. (J) Fisher’s Exact test with BH FDR correction. P values as indicated. H&E, hematoxylin and eosin; ns, not significant; Ctrl, control; IFE, interfollicular epidermis; GO, Gene Ontology; ST, spatial transcriptomics. TEWL; trans-epidermal water loss.

DEG, differentially expressed genes; scRNA-seq, single cell RNA sequencing; Rep, representative.

Nedd8 loss activates epidermal Nrf2 and drives oxidative stress-associated inflammation

Nedd8−/− and Sumo2−/− basal keratinocytes had opposite differentiation states, with altered expression of Trp63, Irf6 and Keratin 14 (Fig. 3D; fig. S8, G to I). Given the enrichment of Nrf2 signaling at both the protein and RNA level (fig. S6G), we stratified Nedd8−/− basal and NEDD8 KO specific keratinocytes by Nrf2 activity (Fig. 3E; Data S5). Confocal microscopy corroborated nuclear NRF2 in Nedd8−/− epidermis and spatial transcriptomics revealed increased Nrf2 transcripts in NEDD8 KO specific keratinocytes (fig. S9, A to C). To assess NEDD8 conjugation, we used proximity ligation assays for NEDD8-UBE2M as a surrogate. This confirmed efficacy of topical NEDD8i treatments in wild type C57BL/6J mice (fig. S9, D and E). NEDD8i treatment increased expression of NRF2 protein and cornified envelope transcripts in vivo (fig. S9, F to H). NRF2 CUT&RUN mapped genomic occupancy in primary human keratinocytes treated with NEDD8i or vehicle. This identified 2,404 differential NRF2 binding peaks (fig. S9, I to K). NEDD8i altered NRF2 occupancy near inflammatory loci, including IL36G, IL1RL1, CXCL1/8, and IL1A. These factors, which promote neutrophil recruitment and cytokine-driven inflammation, were responsive to NEDD8i treatment (fig. S9, L and M) (30,31). The genetic interaction effect between NRF2 knockdown and NEDD8i tested the requirement of NRF2 for NEDD8i-induced transcriptional responses (fig. S9N). Approximately 10% of the transcriptomic responses to NEDD8i were NRF2-dependent (fig. S9, O and P). NRF2 therefore mediated a defined subset of the NEDD8i induced transcriptional impacts.

We crossed Nedd8 knockout mice (Nedd8fl/fl Krt14-CreERT) to constitutive Nrf2 knockout mice to dissect the contribution of NRF2 to the Nedd8−/− in vivo phenotype (Fig. 3F) (32). Nedd8−/− Nrf2−/− skin demonstrated visible scaling but lacked the erythema characteristic of Nedd8−/− loss alone (fig. S10A). Histologically analysis noted a hyperkeratotic epidermis with reduced proliferation but persistent premature differentiation (Fig. 3G; fig. S10, B to D). Double knockout mice maintained barrier impairment despite normalization of epidermal thickness (Fig. 3H; fig. S10B). scRNA-seq of Nedd8−/− Nrf2−/− and controls identified a persistence of the NEDD8 KO specific population (fig. S10, E to H). Double knockout skin had a depleted NRF2 signature but retained an elevated differentiation score (Fig. 3I; fig. S10, I and J). The NEDD8 KO-specific population in double knockout mice had decreased oxidative stress and inflammatory response enrichment. There was more pronounced Wnt signaling, differentiation and wound healing pathways than in Nedd8−/− cells (Fig. 3J; fig. S10K). Basal-state analysis identified increased basal keratinocyte differentiation in both Nedd8−/− and Nedd8−/− Nrf2−/− skin (Fig. 3I). We noted fewer basal cells along the basement membrane by spatial transcriptomics and reduced basal cell abundance by scRNA-seq (Fig. 3K; fig. S10, L and M). Nedd8−/− Nrf2−/− skin had altered immune infiltrates and reduced spatial density of NRF2-dependent pro-inflammatory transcripts (Fig. 3L; fig. S10, N and O). Expression of pro-differentiation transcription factor Irf6 remained perturbed (fig. S10, P and Q). Thus, the inflammatory component of the Nedd8−/− phenotype was altered by concurrent Nrf2 loss, while the epidermal differentiation defects persisted. This supported NRF2-independent roles of NEDDylation in skin homeostasis. Together, transcriptomic, proteomic and spatial mapping of UBL perturbations (fig. S11, A to H) underscored mechanistic differences among Nedd8−/−, Sumo2−/− and Nedd8−/− Nrf2−/− phenotypes. These mice also highlighted the role of Nrf2 in driving inflammation seen with Nedd8 loss.

Contrasting UBL-mediated and ubiquitin-proteasome-concordant programs

NEDDylation-dependent NRF2 stabilization is known to be a ubiquitin-proteasome system (UPS)-dependent effect (17). NEDDylation also has non-UPS impacts (33, 34). We undertook a proteome-wide pharmacologic screen to assess UPS-concordant and discordant effects of NEDDylation. Primary keratinocytes were treated for 4 or 16 hours with NEDD8i, SUMOi, ubiquitination inhibitor (UBQi), proteasome inhibitor (PROi), or vehicle (Fig. 4A; fig. S12, A to G). Proteomic profiling clustered NEDD8i-treated cells separately from other treatments, with both UPS-shared and NEDD8-specific signatures (Fig. 4B).

Fig. 4. NEDD8 coordinates proteomic and transcriptomic levels of regulation.

Fig. 4.

Open in a new tab

(A) Workflow for proteomic survey of NEDDylation, ubiquitination and proteasomal pharmacological inhibition for 16hrs (n=3). (B) PCA of DIA-MS samples. (C) Bayesian model determination of NEDD8i effects with stratification into unique and non-unique categories (left). Posterior probabilities for individual NEDD8i effect models with model sums (right). (D) Volcano plot of NEDD8i unique (green), non-unique (purple) and non-significant effects (gray). Representative proteins labeled. (E) Odds ratios of NEDD8-dependent proteins in differentiation dynamic protein clusters (from Fig. 1B) relative to stable proteins for combined NEDD8i effect models (left) and NEDD8i unique effect model (right). (F) PPI networks of functionally enriched proteins from NEDD8i non-unique (left) and NEDD8i unique (right) models. (G) Correlation of NEDD8i induced transcriptomic and proteomic changes grouped by NEDD8i effect model. Lines represent linear regression model; shading represent regression coefficients. (H) Enrichment for NEDD8-modulated proteins amongst RNAs with consistent RNA stability changes over an actD time course ± NEDD8i. (I) Distribution of transcript-level RNA stability effects for proteins increased (left) or decreased (right) by NEDD8i treatment. Boxes indicate median + IQR. (J) Proportion of NEDD8i regulated PPI networks with associated RNA transcript level stability changes. Graphs show mean ± SEM. *=p<5e-2, ***=p<1e-3. Statistical analyses: (C) hierarchical empirical Bayesian framework to estimate posterior probabilities. (E, H) Fisher’s exact test with BH FDR correction. (G) grouped linear regression of RNA expression and protein abundance. (I) Unpaired Wilcoxon rank-sum test. PCA, principal component analysis; NEDD8i, NEDDylation inhibitor pevonedistat (MLN-4924); UBQi, ubiquitylation inhibitor TAK-243; PROi, proteasome inhibitor MG-132; DIA-MS, Data independent acquisition mass spectrometry; GO, Gene Ontology; PPI, protein-protein interaction; OR, odds ratio.

Bayesian modeling of 16-hour proteomic impacts stratified NEDD8i responses into unique and non-unique effects. This identified 760 significantly modulated proteins split between the two models (Fig. 4C; fig. S13, A and B). A dominant subset had discordant effects relative to other UPS perturbations (NEDD8i-unique, n=480) (Fig. 4, C and D; Data S6). A 4-hour proteomic time-course enabled early-response comparisons. UBQi and PROi remained similarly correlated at both 4 and 16 hours (R= 0.65 and 0.67) whereas UBQi and NEDD8i effects diverged over time (R = 0.50 and 0.38) (fig. S12E). This suggested temporal uncoupling of NEDD8-specific responses from the UPS. Furthermore, NEDD8i modulated proteins accumulated across the differentiation-dynamic clusters defined in Fig. 1B (fig. S13C). The strongest enrichment was in differentiation induced cluster 1 (Fig. 4E). Consistent with this, the NEDD8i effect model preferentially classified differentiation dynamic proteins (fig. S13D).

SUMOi produced limited changes in protein expression at both timepoints (fig. S14, A to C). Endogenous SUMO2 IP-MS recovered a set of SUMO2 associated proteins (fig. S14, D to G; Data S7) (35, 36). Among enriched proteins was TEAD1, a progenitor-associated transcription factor with SUMOylation-sensitive regulatory domains (fig. S14H) (37). We verified SUMOi dependent proximity in keratinocytes by PLA (fig. S14, I and J). The impact of SUMOi was consistent with established roles of SUMO2 in transcriptional regulation and lesser impacts on immediate protein turnover (35, 36).

UPS overlapping NEDD8i impacts included stress response factors and adhesion complex assembly. These protein-protein interaction (PPI) networks highlighted NRF2 signaling and cell substrate junction organization (Fig. 4F). NEDD8i-unique differentiation networks included cornified envelope formation, intermediate filament organization, and NF-κB signaling (Fig. 4F). These results established distinct UPS-concordant and NEDD8 unique impacts on the cellular proteome.

NEDDylation regulates RNA metabolism and transcript stability

The extensive protein-level changes induced by NEDDylation loss raised the possibility of RNA-level alterations. NEDD8i-modulated protein changes correlated with transcriptional shifts across effect sub-models (Fig. 4G). We compared mRNA expression, stability and splicing in keratinocytes to investigate how NEDD8 altered RNA levels. NEDD8i had limited effects on RNA splicing, which did not correlate with transcriptomic or proteomic changes (fig. S15, A to C). An actinomycin D time course profiled RNA stability (fig. S15D). This resolved multiple RNA stability trajectories (fig. S15, E to G). Differentiation-associated transcripts were consistently stabilized, whereas cell cycle regulators were consistently destabilized (fig. S15H). RNA stability changes were overrepresented amongst NEDD8i-unique proteins and reflected protein level impacts across models (Fig. 4H; fig. S15I). NEDD8i altered thousands of transcripts, including many relevant for downstream inhibitor effects (fig. S15, J and K). Concordant changes in RNA abundance and stability occurred across both classes of NEDD8i modulated proteins (Fig. 4I; fig. S15L). NEDD8i-unique protein interaction networks linked to differentiation preferentially corresponded with transcripts displaying time-dependent stabilization upon NEDD8i treatment. In contrast, neutral or destabilized transcripts dominated non-unique stress/adhesion networks (Fig. 4J). These findings linked the NEDD8i proteomic program to widespread remodeling of mRNA expression and stability.

Functional and biochemical mapping of NEDDylated RNA regulators

We probed NEDD8-dependent effects in primary keratinocytes in parallel with a pharmacological CRISPR suppressor screen and endogenous NEDD8 IP-MS (Fig. 5A; Data S7 and S8). Genome-wide CRISPR screening explored growth dependencies unique to NEDD8i treatment (fig. S16, A to D). Keratinocytes were cultured with NEDD8i at doses permissive of growth while still depleting guides targeting essential genes (fig. S16, E to K). This allowed for NEDDylation related guide enrichment including the NEDDylation E2 enzyme UBE2M and NEDD8 itself (Fig. 5B; fig. S16, L and M; Data S8) (38). GO term analysis nominated protein NEDDylation, ubiquitin conjugation, chromatin remodeling, Wnt signaling, cell adhesion, and RNA processing as NEDDylation dependent (Fig. 5C; fig. S16M). Next, NEDD8 IP-MS enriched 748 proteins relative to IgG control (fig. S17, A to F; Data S7). Incorporation of NEDD8i dependency yielded 172 high-confidence targets (Fig. 5D; fig. S17G). Modified proteins included NEDDylation enzymes (UBE2M, UBE2F, NAE1, UBA3), and NEDD8 interactors such as EGFR (39). GO terms included NEDDylation, small protein conjugation, catabolism, and cell growth (fig. S17H). Intersection of IP-MS and NEDD8i dependent suppressor screen hits identified NEDDylation dependencies and nominated direct binding partners (Fig. 5E).

Fig. 5. NEDDylation modulates RNA stability via HNRNPU.

Fig. 5.

Open in a new tab

(A) Workflow for context specific NEDDylation dependency identification. (B) Scatterplot of β-scores for DMSO and NEDD8i CRISPR screens. NEDDylation pathway (green). (C) Network clustering of significant suppressor screen hits. (D) Euler diagram of high confidence NEDD8-associated proteins from endogenous IP-MS (n=6). (E) Ranked enrichment of high confidence IP-MS proteins; known NEDD8-interactors (dark green), CRISPR hits (light green), and others (gray). (F) Confocal microscopy of NEDD8: HNRNPU PLA (red). (G) quantification of NEDD8: HNRNPU after pharmacologic (top) or genetic (bottom) NEDDylation modulation (n=3). scalebars 10μm (z), 25μm. (H) Volcano plot of irCLIP HNRNPU RNA-binding following 16hr NEDD8i treatment; increased (red) or decreased (blue) HNRNPU binding events (n=2). (I) Re-CLIP enrichment ratio (1st IP: HNRNPU, 2nd IP: NEDD8/1st IP: HNRNPU, 2nd IP IgG) by binding category. (J) DTS workflow (K) GO: BP enrichments amongst HNRNPU differential stabilized RNAs, grouped by NEDD8i co-dependency. (L) NEDD8i dependent HNRNPU RNA binding changes by stability category (n=2). (M) ECDFs of NEDD8i induced protein expression changes associated with HNRNPU-dependent differentially stable transcripts, stratified by NEDD8i co-dependency. (N) Provisional model. Graphs show mean ± SEM. *=p<5e-2, **=p<1e-2, ***=p<1e-3. Statistical analyses: (B) MAGeCK-MLE (maximum likelihood estimation framework). (C, I, K) Fisher’s exact test with BH FDR correction. (G) two-sided unpaired t-test with Welch correction. (E, H) limma (linear model + empirical Bayes) with BH FDR correction. (M) Kolmogorov-Smirnov test. NEDD8i, NEDD8 inhibitor pevonedistat; IP-MS, immunoprecipitation–mass spectrometry; irCLIP-v2, infrared crosslinking immunoprecipitation v2; DTS, differential transcript stability; GO, Gene Ontology; PLA, proximity ligation assay; ECDF, empirical cumulative distribution functions.

Experimental candidates included the RNA-binding protein, HNRNPU, implicated in keratinocyte progenitor maintenance (fig. S18A) (40). PLA further established HNRNPU-NEDD8 proximity (Fig. 5, F and G; fig. S18B). We confirmed a progenitor function of HNRNPU in primary human keratinocyte cultures and skin organoids (fig. S18, C to H). Human xenografts seeded with HNRNPU CRISPR knockout cells phenocopied the barrier impairment observed in Nedd8−/− mice or after topical NEDD8i treatment (fig. S18, I to K). This corresponded with the barrier dysfunction noted in embryonic lethal Hnrnpu-knockout murine skin (40). NEDD8i and HNRNPU knockdown both impaired keratinocyte migration velocity (fig. S5J; fig. S18, L and M). DiGly IP-MS of NEDD8i treated keratinocytes found multiple putative sites of NEDD8 attachment across HNRNPU. These included previously nominated modification sites and novel residues (fig. S18, N to P) (33, 34). HNRNPU thus represented a NEDDylated RNA binding protein essential for keratinocyte progenitor maintenance.

NEDDylation modulates RNA binding and stabilization by HNRNPU

Next, we investigated the effects of inhibiting NEDDylation on the HNRNPU RNA binding repertoire by infrared crosslinking immunoprecipitation (irCLIPv2-seq) (41). This identified 3,884 RNA transcripts with increased and 2,560 transcripts with decreased HNRNPU binding (Fig. 5H). After primary HNRNPU irCLIPv2, Re-CLIP-seq of NEDD8 enriched for co-bound RNAs (Fig. 5I) (41). This corroborated that HNRNPU NEDDylation modulated a subset of the HNRNPU RNA binding changes observed with NEDD8i.

To interrogate the functionality of these binding changes, we assessed RNA modulation in keratinocytes after HNRNPU loss. Alternative splicing events did not overlap with NEDD8i effects (fig. S19A). HNRNPU differentially stable transcripts (DTS) were classified according to NEDD8 co-dependence (fig. S19B). There were 179 stabilized and 103 destabilized NEDD8-dependent HNRNPU-modulated transcripts (Fig. 5J; Data S9). This contrasted with the reduced impact of SUMOi on RNA stability (fig. S19, C to E). With NEDD8i, stabilized HNRNPU-bound RNAs were enriched for differentiation transcripts, such as KRT16. Destabilized targets included cell cycle and progenitor transcripts such as COL17A1 (Fig. 5K; fig. S19F). HNRNPU DTS events corresponding with NEDD8i induced HNRNPU binding changes (Fig. 5L). This was consistent with HNRNPU stabilization of bound RNAs. HNRNPU-NEDD8 Re-CLIP-seq enriched for transcripts linked to progenitor and cell cycle regulation (fig. S19G) (42–47). Distributional shifts in NEDD8i-dependent and independent HNRNPU stability categories were significant for proteins uniquely regulated by NEDD8i relative to other UPS inhibitors (Fig. 5M; fig. S19H).

To test a direct requirement for HNRNPU NEDDylation, we generated an eight-residue lysine to arginine mutant HNRNPU construct (8M). WT and 8M constructs had comparable expression and localization (fig. S20, A and B). In skin organoids, the WT but not 8M construct restored HNRNPU knockdown driven differentiation defects (fig. S20, A to D). WT also rescued the HNRNPU dependent hypo-proliferative phenotype, whereas 8M did not (fig. S20, E and F). Transcriptomic profiling of the 8M HNRNPU detected 102 differentially expressed genes, which included cell cycle machinery. This also represented collapsed regulation of differentiation and wound response transcripts (fig. S20G; Data S10). We compared diGly IP-MS with NEDD8i, UBQi or PROi to assess whether the mapped diGly sites reflected NEDDylation dependent ubiquitination sites rather than residues of direct NEDD8 attachment. NEDD8i treatment differentially enriched five residues (fig. S20H). Proximity ligation assays verified that NEDD8i preferentially reduced HNRNPU-NEDD8 proximity compared to HNRNPU-UBB (fig. S20, I to K). Depletion of the de-NEDDylase SENP8 accentuated NEDD8-HNRNPU proximity in the WT construct, while the NEDDylation deficient 8M proximity remained low (fig. S20, L and M). These findings supported a provisional model in which NEDDylated HNRNPU stabilized progenitor mRNAs, whereas de-NEDDylated HNRNPU shifted to binding and stabilizing pro-differentiation mRNAs (Fig. 5N).

Undifferentiated keratinocytes displayed high levels of HNRNPU-NEDD8, which was reduced in differentiated cells (fig. S21, A and B). This suggested that NEDDylated HNRNPU marked progenitor-like states. Consistent with this, the poorly differentiated squamous cell carcinoma (SCC) cell line A431 had higher HNRNPU-NEDD8 signal than moderately differentiated CAL27 SCC cells (fig. S21, A and B) (48, 49). We analyzed 41 spontaneous human epidermal cancers and found elevated HNRNPU-NEDD8 proximity compared to control (fig. S21, C and D). Transcripts with NEDD8i dependent HNRNPU stability effects were amongst the NEDD8i induced SCC transcriptional signatures (fig. S21, E and F). These bidirectional HNRNPU binding signatures were found in altered RNA profiles of 20 primary SCC samples (fig. S21G) (50). These correlative data raise the possibility that NEDD8-HNRNPU associations may play a role in progenitor-like programs in epithelial cancer.

Conclusions

Here, we identified NEDD8 and SUMO2 as reciprocal regulators of differentiation in stratified epithelia that control divergent epithelial-intrinsic, stromal, and immune interactions operative in homeostasis and disease. Both UBLs were indispensable for normal cutaneous barrier function, albeit via opposite effects. NEDD8 supported the progenitor state, enabled wound healing, and suppressed NRF2-dependent neutrophilic inflammation. NEDD8 loss perturbed proteomic and transcriptomic circuits that induced aberrant differentiation across multiple stratified epithelial tissues. In contrast, SUMO2 was required for the induction of terminal differentiation genes. SUMO2 co-localized with context-specific transcription factors, such as TEAD1. The impacts of NEDD8 inhibition diverged from the ubiquitin-proteasome system and instead corresponded with RNA abundance alterations. Moreover, NEDDylation controlled RNA stability in part via HNRNPU, an RNA binding protein whose bound RNAs were modulated by its NEDDylation status. NEDD8-modified HNRNPU preferentially bound and stabilized RNAs essential for progenitor self-renewal. De-NEDDylated HNRNPU shifted to binding and stabilizing differentiation mRNAs. Together, these dynamic UBL modifications link protein and RNA regulation in the control of tissue homeostasis.

Supplementary Material

Supplementary Materials
Data File 10
Date File 5
Data File 11
Data File 9
Data File 8
Data File 1
Data File 6
Data File 4
Data File 7
Data File 2
Data File 3

Materials and Methods are available in the Supplementary Materials.

Acknowledgements:

This work was supported by the Vincent Coates Foundation Mass Spectrometry Laboratory, Stanford University Mass Spectrometry (RRID:SCR_017801) utilizing the Bruker timsTOF Ultra & nanoElute 2 system (RRID: SCR_025639) with assistance from Casey Powers, Garvey McKenzie, and Juan Pablo Galindo Lazo. We would like to thank Stanford FACS facility for assistance with flow cytometry and Sharzad Talebian and Dhananjay Chu at the Stanford Genomics core helped with spatial transcriptomics processing. We would also like to acknowledge Patrick James Atkinson for assistance with confocal microscopy, Kerriann Casey and Kerri Rieger for histologic analysis, Pauline Chu and Histo-Tec Laboratory in Hayward, CA for histology processing. We acknowledge Rudy Wycallis at the Stanford shared FACS facility for assistance with cell sorting and Nicole Dow, A. Dazey and P. Bernstein for administrative assistance. We would like to thank David L. Reynolds for assistance with perturb-sequencing, Konnie Q. Guo for genome-wide CRISPR sequencing, Smarajit Mondal for assistance in construct generation, Shiying Tao for cell culture and Rieke-Marie Hackbarth for assistance with scRNA-seq annotation. We would like to thank K Fields and G. Rayant for their generous support.

Funding:

This work was supported by the USVA Office of Research and Development, by USVA Merit Review grant BX001409 to P.A.K., and by NIAMS/NIH grants AR045192 and AR049737 to P.A.K. This work was also supported in part by NIH P30 CA124435 utilizing the Stanford Cancer Institute Proteomics/Mass Spectrometry Shared Resource. LMM was supported by National Science Foundation Graduate Research Fellowship Program under grant DGE-2444107.

Footnotes

Competing interests:

M.S. is a founder and equity holder of Riboscience LLC and founder and board member of Salacia Therapeutics, Inc. M.C.G.W. is a co-founder of PSOMRI and PHAMRI. M.C.G.W., L.V.J., M.S. and P.A.K. are inventors on patent applications S25-326 and S25-409, held and submitted by Stanford University. They are related to targeting UBLs in the treatment of skin diseases and disorders.

Data, code and materials availability:

High throughput sequencing data of human cell lines or mouse samples are deposited in GEO under accession number GSE326753. Primary human keratinocyte sequencing data is available under controlled access from dbGaP to protect patient privacy at accession code phs004614.v1.p1. Access is available to all users by request from dbGaP for general research use. Raw mass spectrometry data is available through MassIVE repository MSV000100932 (51). This study does not include original software code. Further information can also be obtained from the corresponding author upon request. Reagents generated in this study are available from the corresponding author upon reasonable request and completion of a material transfer agreement.

References and notes:

  • 1.Madison K, Barrier Function of the Skin: “La Raison d’Être” of the Epidermis. Journal of Investigative Dermatology, 121, 231–241 (2003). doi: 10.1046/j.1523-1747.2003.12359.x [DOI] [PubMed] [Google Scholar]
  • 2.Werner S et al. , The function of KGF in morphogenesis of epithelium and reepithelialization of wounds. Science 266, 819–822 (1994). doi: 10.1126/science.7973639 [DOI] [PubMed] [Google Scholar]
  • 3.Nizet V et al. , Innate antimicrobial peptide protects the skin from invasive bacterial infection. Nature 414, 454–457 (2001). doi: 10.1038/35106587 [DOI] [PubMed] [Google Scholar]
  • 4.Choudhary C, Mann M, Decoding signaling networks by mass spectrometry-based proteomics. Nat Rev Mol Cell Biol 11, 427–439 (2010). doi: 10.1038/nrm2900 [DOI] [PubMed] [Google Scholar]
  • 5.Miroshnikova YA et al. , Adhesion forces and cortical tension couple cell proliferation and differentiation to drive epidermal stratification. Nat Cell Biol 20, 69–80 (2018). doi: 10.1038/s41556-017-0005-z [DOI] [PubMed] [Google Scholar]
  • 6.Jones PH, Watt FM, Separation of human epidermal stem cells from transit amplifying cells on the basis of differences in integrin function and expression. Cell 73, 713–724 (1993). doi: 10.1016/0092-8674(93)90251-K [DOI] [PubMed] [Google Scholar]
  • 7.Rompolas P et al. , Spatiotemporal coordination of stem cell commitment during epidermal homeostasis. Science 352, 1471–1474 (2016). doi: 10.1126/science.aaf7012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Kuo IH, Yoshida T, De Benedetto A, Beck LA, The cutaneous innate immune response in patients with atopic dermatitis. J Allergy Clin Immunol 131, 266–278 (2013). doi: 10.1016/j.jaci.2012.12.1563 [DOI] [PubMed] [Google Scholar]
  • 9.Sawada Y et al. , Cutaneous innate immune tolerance is mediated by epigenetic control of MAP2K3 by HDAC8/9. Sci Immunol 6, eabe1935 (2021). doi: 10.1126/sciimmunol.abe1935 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Zhou X, Chen Y, Cui L, Shi Y, Guo C, Advances in the pathogenesis of psoriasis: from keratinocyte perspective. Cell Death Dis 13, 81 (2022). doi: 10.1038/s41419-022-04523-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Martin P, Pardo-Pastor C, Jenkins RG, Rosenblatt J, Imperfect wound healing sets the stage for chronic diseases. Science 386, eadp2974 (2024). doi: 10.1126/science.adp2974 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Chang MS, Azin M, Demehri S, Cutaneous Squamous Cell Carcinoma: The Frontier of Cancer Immunoprevention. Annu Rev Pathol 17, 101–119 (2022). doi: 10.1146/annurev-pathol-042320-120056 [DOI] [PubMed] [Google Scholar]
  • 13.Kim DS et al. , The dynamic, combinatorial cis-regulatory lexicon of epidermal differentiation. Nat Genet 53, 1564–1576 (2021). doi: 10.1038/s41588-021-00947-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Dyring-Andersen B et al. , Spatially and cell-type resolved quantitative proteomic atlas of healthy human skin. Nat Commun 11, 5587 (2020). doi: 10.1038/s41467-020-19383-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Lee JM, Hammaren HM, Savitski MM, Baek SH, Control of protein stability by post-translational modifications. Nat Commun 14, 201 (2023). doi: 10.1038/s41467-023-35795-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Dixit A et al. , Perturb-Seq: Dissecting Molecular Circuits with Scalable Single-Cell RNA Profiling of Pooled Genetic Screens. Cell 167, 1853–1866 e1817 (2016). doi: 10.1016/j.cell.2016.11.038; [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Soucy TA et al. , An inhibitor of NEDD8-activating enzyme as a new approach to treat cancer. Nature 458, 732–736 (2009). doi: 10.1038/nature07884 [DOI] [PubMed] [Google Scholar]
  • 18.Langston SP et al. , Discovery of TAK-981, a First-in-Class Inhibitor of SUMO-Activating Enzyme for the Treatment of Cancer. J Med Chem 64, 2501–2520 (2021). doi: 10.1021/acs.jmedchem.0c01491 [DOI] [PubMed] [Google Scholar]
  • 19.Kretz M et al. , Control of somatic tissue differentiation by the long non-coding RNA TINCR. Nature 493, 231–235 (2013). doi: 10.1038/nature11661 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Miao W et al. , Glucose dissociates DDX21 dimers to regulate mRNA splicing and tissue differentiation. Cell 186, 80–97 e26 (2023). doi: 10.1016/j.cell.2022.12.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Sarmin AM, Connelly JT, Fabrication of Human Skin Equivalents Using Decellularized Extracellular Matrix. Curr Protoc 2, e393 (2022). doi: 10.1002/cpz1.393 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Mascharak S et al. , Preventing Engrailed-1 activation in fibroblasts yields wound regeneration without scarring. Science 372, eaba2374 (2021). doi: 10.1126/science.aba2374 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.van der Fits L et al. , Imiquimod-induced psoriasis-like skin inflammation in mice is mediated via the IL-23/IL-17 axis. J Immunol 182, 5836–5845 (2009). doi: 10.4049/jimmunol.0802999 [DOI] [PubMed] [Google Scholar]
  • 24.Ansari J et al. , Targeting the AnxA1/Fpr2/ALX pathway regulates neutrophil function, promoting thromboinflammation resolution in sickle cell disease. Blood 137, 1538–1549 (2021). doi: 10.1182/blood.2020009166 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Kumari S et al. , Tumor necrosis factor receptor signaling in keratinocytes triggers interleukin-24-dependent psoriasis-like skin inflammation in mice. Immunity 39, 899–911 (2013). doi: 10.1016/j.immuni.2013.10.009 [DOI] [PubMed] [Google Scholar]
  • 26.Zhang C et al. , CD100-Plexin-B2 Promotes the Inflammation in Psoriasis by Activating NF-kappaB and the Inflammasome in Keratinocytes. J Invest Dermatol 138, 375–383 (2018). doi: 10.1016/j.jid.2017.09.005 [DOI] [PubMed] [Google Scholar]
  • 27.Cepek KL et al. , Adhesion between epithelial cells and T lymphocytes mediated by E-cadherin and the alpha E beta 7 integrin. Nature 372, 190–193 (1994). doi: 10.1038/372190a0 [DOI] [PubMed] [Google Scholar]
  • 28.Mackay LK et al. , The developmental pathway for CD103(+)CD8+ tissue-resident memory T cells of skin. Nat Immunol 14, 1294–1301 (2013). doi: 10.1038/ni.2744 [DOI] [PubMed] [Google Scholar]
  • 29.Fisher TL et al. , Generation and preclinical characterization of an antibody specific for SEMA4D. MAbs 8, 150–162 (2016). doi: 10.1080/19420862.2015.1102813 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Lentini G et al. , Neutrophils Enhance Their Own Influx to Sites of Bacterial Infection via Endosomal TLR-Dependent Cxcl2 Production. J Immunol 204, 660–670 (2020). doi: 10.4049/jimmunol.1901039 [DOI] [PubMed] [Google Scholar]
  • 31.Clancy DM et al. , Extracellular Neutrophil Proteases Are Efficient Regulators of IL-1, IL-33, and IL-36 Cytokine Activity but Poor Effectors of Microbial Killing. Cell Rep 22, 2937–2950 (2018). doi: 10.1016/j.celrep.2018.02.062 [DOI] [PubMed] [Google Scholar]
  • 32.Chan K, Lu R, Chang JC, Kan YW, NRF2, a member of the NFE2 family of transcription factors, is not essential for murine erythropoiesis, growth, and development. Proc Natl Acad Sci U S A 93, 13943–13948 (1996). doi: 10.1073/pnas.93.24.13943 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Lobato-Gil S et al. , Proteome-wide identification of NEDD8 modification sites reveals distinct proteomes for canonical and atypical NEDDylation. Cell Rep 34, 108635 (2021). doi: 10.1016/j.celrep.2020.108635 [DOI] [PubMed] [Google Scholar]
  • 34.Vogl AM et al. , Global site-specific neddylation profiling reveals that NEDDylated cofilin regulates actin dynamics. Nat Struct Mol Biol 27, 210–220 (2020). doi: 10.1038/s41594-019-0370-3 [DOI] [PubMed] [Google Scholar]
  • 35.Hendriks IA, Vertegaal AC, A comprehensive compilation of SUMO proteomics. Nat Rev Mol Cell Biol 17, 581–595 (2016). doi: 10.1038/nrm.2016.81 [DOI] [PubMed] [Google Scholar]
  • 36.Hendriks IA et al. , Site-specific characterization of endogenous SUMOylation across species and organs. Nat Commun 9, 2456 (2018). doi: 10.1038/s41467-018-04957-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.DelRosso N et al. , Large-scale mapping and mutagenesis of human transcriptional effector domains. Nature 616, 365–372 (2023). doi: 10.1038/s41586-023-05906-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Mamun M et al. , Discovery of neddylation E2s inhibitors with therapeutic activity. Oncogenesis 12, 45 (2023). doi: 10.1038/s41389-023-00490-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Najor NA et al. , Epidermal Growth Factor Receptor neddylation is regulated by a desmosomal-COP9 (Constitutive Photomorphogenesis 9) signalosome complex. Elife 6, e22599 (2017). doi: 10.7554/eLife.22599 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Hong SP et al. , Hnrnpu Is Essential for Proper Murine Skin Development. J Invest Dermatol 145, 965–968 e964 (2025). doi: 10.1016/j.jid.2024.09.016 [DOI] [PubMed] [Google Scholar]
  • 41.Ducoli L et al. , irCLIP-RNP and Re-CLIP reveal patterns of dynamic protein assemblies on RNA. Nature 641, 769–778 (2025). doi: 10.1038/s41586-025-08787-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.He H et al. , Glycomic and transcriptomic response of GSC11 glioblastoma stem cells to STAT3 phosphorylation inhibition and serum-induced differentiation. J Proteome Res 9, 2098–2108 (2010). doi: 10.1021/pr900793a [DOI] [PubMed] [Google Scholar]
  • 43.Feng L et al. , Comprehensive Analysis of E3 Ubiquitin Ligases Reveals Ring Finger Protein 223 as a Novel Oncogene Activated by KLF4 in Pancreatic Cancer. Front Cell Dev Biol 9, 738709 (2021). doi: 10.3389/fcell.2021.738709 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Choi S et al. , Corticosterone inhibits GAS6 to govern hair follicle stem-cell quiescence. Nature 592, 428–432 (2021). doi: 10.1038/s41586-021-03417-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Bish R, Vogel C, RNA binding protein-mediated post-transcriptional gene regulation in medulloblastoma. Mol Cells 37, 357–364 (2014). doi: 10.14348/molcells.2014.0008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Chen GL et al. , Xanthine dehydrogenase downregulation promotes TGFbeta signaling and cancer stem cell-related gene expression in hepatocellular carcinoma. Oncogenesis 6, e382 (2017). doi: 10.1038/oncsis.2017.81 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Ho TLF et al. , Domain-specific p53 mutants activate EGFR by distinct mechanisms exposing tissue-independent therapeutic vulnerabilities. Nat Commun 14, 1726 (2023). doi: 10.1038/s41467-023-37223-3; pmid: [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Giard DJ et al. , In vitro cultivation of human tumors: establishment of cell lines derived from a series of solid tumors. J Natl Cancer Inst 51, 1417–1423 (1973). doi: 10.1093/jnci/51.5.1417 [DOI] [PubMed] [Google Scholar]
  • 49.Gioanni J et al. , Two new human tumor cell lines derived from squamous cell carcinomas of the tongue: establishment, characterization and response to cytotoxic treatment. Eur J Cancer Clin Oncol 24, 1445–1455 (1988). doi: 10.1016/0277-5379(88)90335-5 [DOI] [PubMed] [Google Scholar]
  • 50.Ji AL et al. , Multimodal Analysis of Composition and Spatial Architecture in Human Squamous Cell Carcinoma. Cell 182, 497–514 e422 (2020). doi: 10.1016/j.cell.2020.05.039 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Winge MCG*, Jackrazi LV*, Srinivasan S, Khavari PA, Raw mass spectrometry data for NEDD8 and SUMO2 ubiquitin-like proteins control epithelial homeostasis, regeneration, and inflammation. MassIVE MSV000100932 (2026); 10.25345/C5C24R21M. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Griffin MF et al. , Fibroblasts of disparate developmental origins harbor anatomically variant scarring potential. Cell 189, 783–799 e720 (2026). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Ridky TW, Chow JM, Wong DJ, Khavari PA, Invasive three-dimensional organotypic neoplasia from multiple normal human epithelia. Nat Med 16, 1450–1455 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Sanson KR et al. , Optimized libraries for CRISPR-Cas9 genetic screens with multiple modalities. Nat Commun 9, 5416 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Saelens W, Cannoodt R, Todorov H, Saeys Y, A comparison of single-cell trajectory inference methods. Nat Biotechnol 37, 547–554 (2019). [DOI] [PubMed] [Google Scholar]
  • 56.Lopez-Pajares V et al. , Glucose modulates IRF6 transcription factor dimerization to enable epidermal differentiation. Cell Stem Cell 32, 795–810 e710 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Skowronek P et al. , Rapid and In-Depth Coverage of the (Phospho-)Proteome With Deep Libraries and Optimal Window Design for dia-PASEF. Mol Cell Proteomics 21, 100279 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Feng Z, Fang P, Zheng H, Zhang X, DEP2: an upgraded comprehensive analysis toolkit for quantitative proteomics data. Bioinformatics 39, btad526 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Cui H, Srinivasan S, Korkin D, Enriching human interactome with functional mutations to detect high-impact network modules underlying complex diseases. Genes 10, 933 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Szklarczyk D et al. , The STRING database in 2023: protein–protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic acids research 51, D638–D646 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Enright AJ, Van Dongen S, Ouzounis CA, An efficient algorithm for large-scale detection of protein families. Nucleic acids research 30, 1575–1584 (2002). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Dobin A et al. , STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 63.Mudge JM et al. , GENCODE 2025: reference gene annotation for human and mouse. Nucleic Acids Res 53, D966–D975 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 64.Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C, Salmon provides fast and bias-aware quantification of transcript expression. Nat Methods 14, 417–419 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 65.Zhang Y et al. , Model-based analysis of ChIP-Seq (MACS). Genome Biol 9, R137 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 66.Landt SG et al. , ChIP-seq guidelines and practices of the ENCODE and modENCODE consortia. Genome research 22, 1813–1831 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 67.Thrane K et al. , Single-Cell and Spatial Transcriptomic Analysis of Human Skin Delineates Intercellular Communication and Pathogenic Cells. J Invest Dermatol 143, 2177–2192 e2113 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 68.Joost S et al. , The Molecular Anatomy of Mouse Skin during Hair Growth and Rest. Cell Stem Cell 26, 441–457 e447 (2020). [DOI] [PubMed] [Google Scholar]
  • 69.Bauer-Rowe KE et al. , Creeping fat-derived mechanosensitive fibroblasts drive intestinal fibrosis in Crohn’s disease strictures. Cell 188, 6536–6553 e6526 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 70.Agrawal A, Thomann S, Basu S, Grun D, NiCo identifies extrinsic drivers of cell state modulation by niche covariation analysis. Nat Commun 15, 10628 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Materials
NIHMS2190488-supplement-Supplementary_Materials.pdf (19.5MB, pdf)
Data File 10
NIHMS2190488-supplement-Data_File_10.xlsx (2.5MB, xlsx)
Date File 5
NIHMS2190488-supplement-Date_File_5.xls (11.3MB, xls)
Data File 11
NIHMS2190488-supplement-Data_File_11.xlsx (12.4KB, xlsx)
Data File 9
NIHMS2190488-supplement-Data_File_9.xlsx (11.1MB, xlsx)
Data File 8
NIHMS2190488-supplement-Data_File_8.xlsx (4.9MB, xlsx)
Data File 1
NIHMS2190488-supplement-Data_File_1.xlsx (15.9MB, xlsx)
Data File 6
NIHMS2190488-supplement-Data_File_6.xlsx (8MB, xlsx)
Data File 4
NIHMS2190488-supplement-Data_File_4.xlsx (1MB, xlsx)
Data File 7
NIHMS2190488-supplement-Data_File_7.xlsx (6.7MB, xlsx)
Data File 2
NIHMS2190488-supplement-Data_File_2.xlsx (78.8KB, xlsx)
Data File 3
NIHMS2190488-supplement-Data_File_3.xlsx (5.2MB, xlsx)

Data Availability Statement

High throughput sequencing data of human cell lines or mouse samples are deposited in GEO under accession number GSE326753. Primary human keratinocyte sequencing data is available under controlled access from dbGaP to protect patient privacy at accession code phs004614.v1.p1. Access is available to all users by request from dbGaP for general research use. Raw mass spectrometry data is available through MassIVE repository MSV000100932 (51). This study does not include original software code. Further information can also be obtained from the corresponding author upon request. Reagents generated in this study are available from the corresponding author upon reasonable request and completion of a material transfer agreement.

RESOURCES