Targeting lymphoid-derived IL-17 signaling to delay skin aging

Skin aging is characterized by structural and functional changes that contribute to age-associated frailty. This probably depends on synergy between alterations in the local niche and stem cell-intrinsic changes, underscored by proinflammatory microenvironments that drive pleotropic changes. The nature of these age-associated inflammatory cues, or how they affect tissue aging, is unknown. Based on single-cell RNA sequencing of the dermal compartment of mouse skin, we show a skew towards an IL-17-expressing phenotype of T helper cells, γδ T cells and innate lymphoid cells in aged skin. Importantly, in vivo blockade of IL-17 signaling during aging reduces the proinflammatory state of the skin, delaying the appearance of age-related traits. Mechanistically, aberrant IL-17 signals through NF-κB in epidermal cells to impair homeostatic functions while promoting an inflammatory state. Our results indicate that aged skin shows signs of chronic inflammation and that increased IL-17 signaling could be targeted to prevent age-associated skin ailments.

Skin aging is characterized by structural and functional changes that contribute to age-associated frailty. This probably depends on synergy between alterations in the local niche and stem cell-intrinsic changes, underscored by proinflammatory microenvironments that drive pleotropic changes. The nature of these age-associated inflammatory cues, or how they affect tissue aging, is unknown. Based on single-cell RNA sequencing of the dermal compartment of mouse skin, we show a skew towards an IL-17expressing phenotype of T helper cells, γδ T cells and innate lymphoid cells in aged skin. Importantly, in vivo blockade of IL-17 signaling during aging reduces the proinflammatory state of the skin, delaying the appearance of age-related traits. Mechanistically, aberrant IL-17 signals through NF-κB in epidermal cells to impair homeostatic functions while promoting an inflammatory state. Our results indicate that aged skin shows signs of chronic inflammation and that increased IL-17 signaling could be targeted to prevent age-associated skin ailments.
The skin contains a multilayered epidermis interspersed with hair follicles and sebaceous glands. The outermost layer of the skin, the epidermis, is a stratified epithelium that forms an impermeable protection for the organism. Daily renewal of the epidermis is fueled by interfollicular epidermal stem cells whereas hair follicle stem cells maintain hair follicles by periodical generation of new hair shafts 7 . In the dermis, an organized mesh of extracellular matrix (ECM) embedded with distinct cell types forms the niche for all epidermal stem cells 7 . Epidermal aging is characterized by alterations in its regenerative potential and barrier functions [8][9][10][11][12] . Moreover, age-related changes in the dermal cellular composition and ECM properties affect its functionality, both as a structural scaffold and as a niche for epidermal stem cells [13][14][15][16][17][18] . Collectively these changes lead to a slower epidermal turnover, breaching of the barrier and lower quality of wound healing, all of which contribute to increasing the incidence of infections and chronic traits in the elderly [19][20][21] .
Previous single-cell RNA sequencing (scRNA-seq) studies focused on defining changes in skin cell types during homeostasis, aging and disease suggest that some immune cell types change their abundance or behavior during aging in skin 13,[22][23][24][25][26][27][28][29] . However, much remains unknown regarding the relationship between immune cells of the dermis and other dermal and epidermal cells during aging.
Here we unbiasedly profiled and characterized the single-cell transcriptome of dermal cells in aged skin in mice. CD4 + T helper (T H ) cells, γδ T cells and innate lymphoid cells (ILCs) showed the most prominent changes during aging and appeared to orchestrate many of the alterations observed in the aged skin. Specifically, these cells became polarized towards an IL-17-producing phenotype, strongly contributing Article https://doi.org/10.1038/s43587-023-00431-z 10X scRNA-seq in these populations were subtle and did not allow for in-depth analysis of their specific age-related alterations.

Immune cells.
Visual representation of immune cell clusters rendered two major cell groups, myeloid and lymphoid (Fig. 2a). We observed statistically significant changes in proportions in some immune populations during aging (Extended Data Fig. 3a,b).
To measure the effect of aging on the transcriptome of immune cells in a global and unbiased manner, we developed a deep learning model that quantifies the deviance of each immune cell type in aging. We assumed that the most age-affected cells would present the most altered transcriptomic changes compared with their adult counterparts and, therefore, would be more distant in multidimensional gene expression space. We used our 10X scRNA-seq data from adult mice as a reference to train our model for each different cell type signature and then evaluated the similarity between adult and aged cells within each cell type. Cells with the highest deviance in their identity were those accumulating more differences between ages. Notably, we found that CD4 + T H cells, γδ T cells, ILCs and dendritic cells were the types with the highest age deviance, showing the highest proportion of cells affected by aging (Fig. 2b). This suggests that lymphoid cell subsets present the most predominant age-related changes. Focusing on lymphoid cells, UMAP representation of exclusively lymphoid cells showed that CD4 + T H cells, γδ T cells and ILCs possess a large proportion of cells with a high age-deviance score (Fig. 2c).
Therefore, because cell proportion changes accompany transcriptomic changes in some dermal immune cells during aging, we focused our further analysis on these immune cell types.

Aged dermal myeloid cells upregulate proinflammatory genes
Dermal myeloid cells balance pro-and anti-inflammatory functions in skin in a context-dependent manner 38 . Specifically, they scan the tissue to detect antigens, orchestrate an early response to pathogens and trigger the first steps in the wound-healing response 39 . We detected eight clusters of myeloid cells: macrophages 1-3, monocytes 1-3, dendritic cells and proliferating myeloid cells (Fig. 2a).
Monocyte 1 and macrophage 1 clusters were defined by proinflammatory features (for example, the former expressed Ccr2, Cxcl2, Fcgr4 and Cxcr4 and for macrophage 1 we obtained Il1b, Tnfaip3 and MHC class II complex genes including H2ab1, H2-Eb1, H2-Aa and Cd74). Conversely, cluster macrophage 2 expressed regulatory macrophage markers including Il4l1, Cd200r1 and Lgals1 (Supplementary Table 1). Of note, although the proportion of both macrophage 1 and monocyte 3 clusters changed in aged tissue (Extended Data Fig. 3a) they showed no clear transcriptome variations (Supplementary Table 2). By contrast, monocyte 1 and 2 clusters showed more pronounced gene expression changes with aging, with a trend towards expression of higher levels of proinflammatory cytokine secretion and responses to proinflammatory stimuli (Extended Data Fig. 3c and Supplementary Tables 2 and 3). This included upregulation of the gene encoding IL-1β in monocyte 1 (Extended Data Fig. 3d), which is necessary for these cells to initiate a full immune response 39 . The expression of Il1b was also upregulated in clusters macrophage 2 and dendritic cells during aging (Extended Data Fig. 3d and Supplementary Table 2), indicating an increase in proinflammatory gene expression during aging.
Together, this indicates that myeloid cell activation exacerbates a chronic inflammatory state in aged skin.

Increased IL-17 in the lymphoid compartment of aged skin
Lymphoid cells in the mouse dermis include T cells (CD4 + , CD8 + , γδ T cells and regulatory T (T reg ) cells), ILCs and natural killer (NK) cells (Fig. 3a, left). These cells scan for tissue damage and perform immune surveillance in steady state, and also trigger an inflammatory response during wound healing and tumorigenesis 38 . The proportion of γδ T cells and CD4 + T H cells increased in aged dermis ( Fig. 2a and Extended Data to the inflammatory environment found in aged skin. Importantly, in vivo blockade of IL-17 signaling during the aging process delayed the development of several hallmarks of skin aging, resembling the state of adult skin.

Aged dermal cells reveal cell type-and age-specific changes
We analyzed the nonepithelial (negative for epithelial cell adhesion molecule (EpCAM -)) dermal population of dorsal skin from aged (80-to 90-week-old) and adult (17-to 25-week-old) mice by scRNA-seq. We removed epidermal cells enzymatically and isolated dermal cells by fluorescence-activated cell sorting (FACS) (Fig. 1a and Extended Data Fig. 1a). To maximize sampling of less abundant immune cells, we enriched by FACS for CD45 + cells and sequenced these separately from other dermal cells (EpCAM -CD45 -) (Extended Data Fig. 1a). Using the 10X Genomics platform (v.3) we characterized 11,940 cells for CD45 + cells and 5,213 for their CD45 -EpCAMcounterparts. After batch correction and data integration, we verified good sample mixing 30 and representation of all replicates across all populations (Extended Data Fig. 1b,c). We clustered all cells by generation of a shared nearest-neighbor graph using the Louvain algorithm 31 and visualized clustering with uniform manifold approximation and projection (UMAP). We then performed differential gene expression between cell populations to obtain cluster markers (Fig. 1b and Supplementary Table 1) and plotted discriminatory population markers to ensure correct clustering (Fig. 1c). In UMAP visualization we observed three major groups of clusters pertaining to nonimmune (CD45 -EpCAM -), immune myeloid and immune lymphoid lineages (both CD45 + ) (Fig. 1b,c).

Nonimmune cells.
Although we detected clusters for endothelial cells, along with pericytes and Schwann cells (Fig. 1b-d), fibroblasts were the most abundant CD45 -EpCAMcell type and were separated into five clusters containing subtypes with distinct features (Fig. 1d). Some of these clusters showed high similarity to well-defined fibroblast subtypes regarding marker expression-cluster 1 was reticular like, cluster 2 papillary like and cluster 3 proinflammatory 23,24,32-34 . Cluster fibroblast 1 included markers of dermal papilla cells 35 , a specialized cluster of fibroblasts located immediately beneath hair follicles that is essential for modulation of the hair follicle cycle (Fig. 1c,d).
With fibroblasts being the most predominant cell type in the dermis, their age-associated alterations probably affect neighboring cells. To further understand their age-related changes we analyzed the expression of senescence-related markers. The accumulation of senescent cells is suggested to be a key mechanism underlying aging 3 . These nonproliferating senescent cells generate a proinflammatory environment by secretion of an array of cytokines and ECM remodeling enzymes among others. Although detection of senescent cells in vivo is challenging, transcriptional upregulation of Cdkn2a (which encodes p16) and/or Cdkn1a (p21) is considered a sign of senescence 3 . We did not observe increased expression of these genes in our dermal nonimmune clusters, which include fibroblasts (Extended Data Fig. 2a and Supplementary Table 2). Besides, there was no upregulation of Ifna and Ifnb1, additional bona fide markers of senescence 36 . Moreover, there was no global difference in expression of the skin aging-associated secreted protein signature 37 in any of the fibroblast populations with aging (Extended Data Fig. 2b). This suggests that there is no increased expression of senescence in our dataset, which could have been caused by either a lack of sufficient depth to detect these marker changes in our experimental setup or to senescence not being a major driver of fibroblast aging in the age group analyzed.
The nonimmune compartment showed stable cell proportions during aging (Extended Data Fig. 2c). This was accompanied by transcriptomic changes in some populations, including fibroblasts 1 and 3 (Supplementary Table 2    Article https://doi.org/10.1038/s43587-023-00431-z Fig. 3b). We also confirmed that there was a higher frequency of CD4 + cells by immunostaining the dermis of aged mice (Extended Data Fig. 3e). Accordingly, γδ T cells, CD4 + T H cells and ILCs presented the most pronounced transcriptomic changes with age from all cell types analyzed (Fig. 2b,c). Focusing on these lymphoid populations, aged CD4 + T H cells, γδ T cells and ILCs expressed much higher levels of inflammatory genes than their adult counterparts (Fig. 2d-f and Supplementary Tables 2 and 3). The proinflammatory cytokine interleukin 17 (IL-17) family members Il17a and Il17f were among the most strongly upregulated genes in these cell types (Fig. 2g). There are six IL-17 members, IL-17A-IL-17F, all of which signal via binding to the IL-17R family of receptors. The highly homologous cytokines IL-17A and IL-17F bind as homo-or heterodimers to IL-17RA and IL-17RC receptors 40 . The IL-17 family of cytokines intervene in tissue repair and host defense and also have pathogenic roles in autoimmune and chronic inflammatory diseases 40 . Importantly, therapeutic inhibition of aberrant IL-17A and IL-17F activities is used as treatment for skin diseases such as psoriasis and other autoimmune conditions [41][42][43][44] . Notably, during aging there is a polarization of circulating CD4 + T H cells towards an IL-17-expressing phenotype and, in mice, γδ T cells also show this skew in peripheral lymph nodes [45][46][47] . However, whether these changes affect peripheral tissue aging is unknown.
The increase in expression of Il17a and Il17f was restricted to these specific aged lymphoid cell types (Extended Data Fig. 4a). By contrast, expression of the receptors that bind these cytokines specifically (Il17ra and Il17rc) remained mostly unaltered across all cell types during aging (Extended Data Fig. 4a). The upregulation of Il17a and Il17f in CD4 + T H cells, γδ T cells and ILCs during aging was due to increased expression in individual cells ( Fig. 3a and Supplementary Table 2), suggesting that there was a polarization towards an IL-17-expressing phenotype. We also detected, by immunostaining, a higher proportion of IL-17A-positive cells in aged mouse back skin dermis (Fig. 3b). Furthermore, we analyzed IL17A and IL17F expression in adult and aged human dermis by fluorescent in situ hybridization (FISH) to determine whether this upregulation prevailed in other species. We observed a clear trend towards a higher (although not statistically significant) percentage of IL17A-and IL17F-positive cells in aged dermis compared with adult dermis. This confirms that our results in mouse skin are, to some extent, also observed in human skin (Extended Data Fig. 4b).
In our 10X scRNA-seq data, zooming in on the CD4 + T H cell cluster further revealed three subclusters with specific marker gene expression (Fig. 3c,d and Supplementary Table 1). While the CD4 + T H (a) cell cluster could not be assigned to any specific T H cell subtype, the CD4 + T H (c) cell cluster showed markers usually expressed by T H 1 cells such as Ifng. The CD4 + T H (b) cell cluster showed markers compatible with their being bona fide T H 17 cells, including Il17f, Rora, Tmem176a/b, Ccr6 and Junb among others [48][49][50] (Fig. 3d). This cell subtype, CD4 + T H (b) cells (T H 17 cells), showed an increase in abundance in aged dermis compared with control adult skin (Fig. 3e). Accordingly, the presence of CD4 + T H (c) cells (T H 1 cells) was decreased in aged dermis (Fig. 3e), pointing again to a skewing of dermal CD4 + cells towards a T H 17 cell phenotype during aging.
Similar to T H cells, γδ T cells could also be further clustered into two subpopulations (Fig. 3c,f and Supplementary Table 1). The γδ T(b) cell cluster expressed markers compatible with γδ T17 cells, as indicated by the expression of Il17a, Rora, Junb and Jak1 among others 50 , whereas the γδ T(a) cell cluster could not be assigned to any specific γδ T cell subtype (Fig. 3f). Both γδ T cell clusters were more abundant in aged skin although only the γδ T(a) cell subcluster was significantly increased, showing an increased presence of IL-17-expressing γδ T cells in aged mouse skin (Fig. 3g).
ILCs are crucial for the development of psoriatic pathogenesis through sustained increased secretion of IL-17 (ref. 51). Our analysis also showed a marked increase in the expression of genes associated with the switch to ILC subtype 3 (refs. 51,52), such as Il17a, Il17f, Tmem176a and Tmem176b in aged skin (Fig. 3a,h and Supplementary Table 2).

Inhibition of IL-17A/F prevents age-related inflammation
We next sought to determine whether increase in the expression of Il17a and Il17f contributes to the general proinflammatory state of aged skin. Treatment with antibodies that neutralize IL-17A and IL-17F (anti-IL-17A/F) is currently used as a therapy for patients with inflammatory and autoimmune diseases such a psoriasis 44 . We therefore blocked IL-17 signaling by systemic administration of anti-IL-17A/F (or an IgG isotype control) in aging mice (73 weeks old) for 12 weeks (Fig. 4a). We then isolated dermal cells by FACS and performed 10X scRNA-seq, following a previous workflow (Fig. 1a). In total, 16,975 CD45 + cells and 33,262 CD45 -EpCAMcells were analyzed with the same bioinformatics pipeline as the aged sample analysis. We further verified batch effect, good sample mixing and representation of all replicates although no batch effect correction was deemed necessary for these samples (Extended Data Fig. 5a,b). We clustered the cells and confirmed that the same cell types were detected (Fig. 4b).
Nonimmune dermal cells. Nonimmune dermal clusters did not show major changes in gene expression following IL-17A/F inhibition. However, neutralization of IL-17A/F activity decreased the abundance of fibroblast 3 cells, previously described as the proinflammatory type of fibroblast (Extended Data Fig. 6a, bottom). This suggests that, even though the transcriptome of these cells is not markedly altered, they still responded to the blocking treatment.

Immune cells.
Myeloid cells respond to IL-17 signaling in inflammatory situations, activating the expression of more proinflammatory cytokines 53 . The upregulation of proinflammatory cytokines (such as Il1b in clusters monocyte 1, macrophage 2 and dendritic cells) that we observed in aged cells was attenuated following neutralization of IL-17A/F signaling (Extended Data Fig. 6b and Supplementary Tables 4  and 5). We also observed a general decrease of proinflammatory genes in the monocyte 1 cluster (Extended Data Fig. 6c), which we previously identified as proinflammatory monocytes, although no significant changes in proportion were detected (Extended Data Fig. 6a).
Anti-IL-17A/F treatment led to no differences either at the transcriptomic level or in cell proportion analysis on lymphoid cells (Supplementary Table 4 and Extended Data Fig. 6a).
Consequently, our results indicate that blockade of IL-17A/F signaling affects certain dermal cell populations in a cell type-specific manner.

Delay in epidermal aging traits following IL-17A/F blockade
We next investigated whether anti-IL-17A/F treatment would also ameliorate age-associated traits in keratinocytes. Sustained inflammation in aged epidermis is thought to affect epidermal stem cell fitness, potentially by reducing their regenerative function 2,10,13,17 . Bulk RNA-seq of aged and adult epidermal cells confirmed that the expression of proinflammatory cytokines and chemokine signaling increases in aged skin (Extended Data Fig. 7a and Supplementary Table 6). Importantly, IL-17A/F blockade reduced expression of cytokine-related signaling and chemotactic genes, including Ccl5, Cxcr3 and Xcl1, as well as downregulated genes involved in proinflammatory processes including Tnf  Table 6). Concomitant to this, there was an increase in the expression of genes related to healthier wound healing such as Wnt7a 54 (Fig. 4c, Extended Data Fig. 7b and Supplementary Table 6).

Cornified layer thickness.
A well-defined age-related phenotype in the epidermis is the increased thickness of the cornified layer 2 . Indeed, the thickness of this layer doubled in aged as compared with adult mice Cornified layer thickness ( m) A g e d / c o n t r o l I g G A g e d / a n t i -I L -  Article https://doi.org/10.1038/s43587-023-00431-z ( Fig. 4d,e). In vivo systemic inhibition of IL-17 (using anti-IL-17A/F) in aging mice resulted in a thinner cornified layer compared with aged/ IgG-treated control mice (Fig. 4d,e).
Hair follicle growth. Aged hair follicles show a lower capacity to enter the growing phase of the hair cycle (anagen) 17 . Anagen can be subdivided into phases that indicate how far the growth of the hair follicles has advanced 55 . As expected, hair follicles in our cohort of aged mice showed a delay in the anagen stage as compared with their adult counterparts at 8 days post epilation (Fig. 5a,b and Extended Data Fig. 7c). Although anti-IL-17A/F treatment in aged mice did not induce spontaneous anagen entry (Extended Data Fig. 7d), hair follicles of aged/anti-IL-17A/F-treated mice showed a faster pace in their anagen entry and progression almost identical to that in adult mice. This points towards a strong amelioration of the capacity of hair follicle stem cells to activate compared with aged/IgG-treated control mice (Fig. 5a,b and Extended Data Fig. 7c). Dermal papilla fibroblasts express secreted factors essential for hair follicle anagen entry, such as RSPO3 (ref. 34). We observed higher Rspo3 mRNA expression by FISH in aged/anti-IL-17A/F-treated mice compared with aged/IgG-treated control mice, although this difference was not statistically significant (Extended Data Fig. 7e). This suggests that dermal papilla fibroblasts could be more fit to regrow hair following IL-17A/F blockade.
Transepidermal water loss. The cornified layer confers impermeability to the skin due to the presence of crosslinked proteins and lipids, and is altered during the aging process 12 . When this layer is removed by mechanical means (for example, by tape stripping), the epidermal barrier becomes permeable (higher transepidermal water loss (TEWL)). Recovery of basal TEWL values after tape stripping is achieved in a few hours under homeostatic conditions. Nonetheless, the efficiency of recovery is affected by aging 8,12,56 . Accordingly, TEWL values recorded during the first 24 h after tape stripping showed a significant delay in aged/IgG-treated mice compared with adults ( Fig. 5c and Extended Data Fig. 7f), but delayed barrier recovery was not as strong in aged/ anti-IL-17A/F-treated mice when compared with aged/IgG-treated mice ( Fig. 5c and Extended Data Fig. 7f). This shows an improvement in the barrier recovery of aged/anti-IL-17A/F-treated mice after sublesional injury to the cornified layer compared with their aged/IgG-treated counterparts.
Wound healing. Efficient wound healing requires a rapid tissue response for efficient re-establishment of the skin barrier. The dialog between dermal and epidermal cells is impaired during aging, leading to a delay in wound healing 18,27 . To test the impact of in vivo IL-17A/F blockade on wound healing we created a 5-mm wound on each flank of adult, aged/IgG control and aged/anti-IL-17A/F-treated mice. Wound closure did not show differences between age groups by macroscopic measurement of wound diameter in vivo throughout the ten post-wounding days (Extended Data Fig. 7g), as suggested by previous reports 18 .
After wounding, functional barrier regeneration is achieved by complete maturation of keratinocytes, including the enucleation of the upper layers of the epidermis (orthokeratosis) 57,58 . Nevertheless, incomplete maturation of keratinocytes can occur in aberrantly healing and chronic wounds where immature nucleated cells can populate the uppermost layers of the epidermis (parakeratosis) 59 . Importantly, when we assessed the epidermal maturation status of the re-epithelized area 10 days after injury we observed that, in adult mice, most of the new epidermis was composed mainly of orthokeratotic epithelium (that is, enucleated cells in the cornified layer). Conversely, in the aged/ IgG-treated control mice the presence of parakeratosis in the majority of wounds points to a delay in epidermal maturation and incorrect barrier recovery (Fig. 5d,e) previously reported for aged wounds 59 . In aged/anti-IL-17A/F-treated mice the epithelization maturation state 10 days after wounding reached a degree intermediate between adult and aged/IgG-treated control wounds (Fig. 5d,e). Moreover, in the dermis a paramount event for correct wound healing is the maturation process of granulation tissue-newly formed blood vessels, immune cells and fibroblasts that help close the wound 60 . As wound healing progresses the granulation tissue matures towards a fibrotic tissue that will eventually form a scar. We observed immature granulation tissue in the majority of aged/IgG-treated control wounds 10 days after wounding. This trend was partially reversed in adults and intermediate maturation was observed in most of these wounds (Fig. 5d,f). The blockade of IL-17A/F signaling in aged mice allowed the progression of granulation tissue maturation to an adult-like state (Fig. 5d,f).
The amelioration of these parameters during wound healing suggests that short-term in vivo blockade of IL-17A/F signaling in aging mice aids the wound-healing process at two different levels: it helps in efficient maturation of both the re-epithelized area and subjacent granulation tissue. To summarize the foregoing, aberrant IL-17A/F signaling is at the hinge between dermal and epidermal cells and its neutralization results in a synergy that delays age-associated skin traits.

IL-17A/F-NF-κB signaling is key in epidermal aging
Epidermal cells are responders to IL-17A and IL-17F in the skin 42 . Once IL-17 receptors IL-17RA and IL-17RC become activated, they signal through downstream transcription factors such as the canonical nuclear factor κ-light-chain-enhancer of activated B cells (NF-κB) 40 . Because tight regulation of NF-κB activity is needed for correct epidermal homeostasis and defense response 38 , we hypothesized that upregulation of IL-17A/F-dependent inflammation-related genes in aged epidermis might rely on NF-κB transcriptional activity. To test this, we performed chromatin immunoprecipitation followed by sequencing (ChIP-seq) for p65 (encoded by Rela) in epidermal cells from aged/IgG-treated control, aged/anti-IL-17A/F-treated and adult mice. We annotated the peaks to their closest gene, filtered out those not associated with protein-coding genes and analyzed their biological function by Gene Ontology (GO) term analysis (Extended Data Fig. 8a and Supplementary Table 7).
Genes bound by p65 in the aged/control IgG-treated samples showed an increase in inflammation-related functions when compared with those in adults, as reflected by the higher statistical significance of inflammation-related categories in GO analysis (Fig. 6a). This was accompanied by the presence of p65 in regulatory areas of specific proinflammatory genes in aged/IgG-treated control samples, such as Il9 and Crlf2, which have been described as relevant in immune responses 61,62 (Fig. 6b). After in vivo blockade of IL-17A/F activity in aging mice, this age-associated inflammatory trend was reversed (Fig. 6a,b). Nonetheless, this effect was only partial because some of these inflammatory genes remained bound by p65 following anti-IL-17A/F treatment (Extended Data Fig. 8b,c). This suggests that the effect of in vivo anti-IL-17A/F treatment in aging mice, although sufficient to delay many aging-related skin traits, may be acting only partially in a p65-binding-dependent manner.
In addition, we observed that binding of p65 to the regulatory areas of certain genes relevant to epidermal homeostasis was lost during aging (Fig. 6c). Such is the case for Evpl and Krt14 (which encode envoplakin and keratin 14, respectively), essential in skin barrier function and stem cell maintenance, respectively 63,64 (Fig. 6d). Importantly, binding of p65 to those genes was recovered following IL-17A/F in vivo blockade, suggesting that decreased IL-17A/F activity might redirect p65 to genes essential for epidermal function that are lost during aging (Fig. 6d). This effect was also partial, because other genes important for epidermal development and maturation, such as Dll1 (ref. 65), remained unbound in aged/anti-IL-17A/F-treated epidermis (Extended Data Fig. 8d).
Thus, we have found that p65/NF-κB is an IL-17A/F-responsive transcription factor in epidermal cells that accounts for part of the

Discussion
During aging, tissue-specific alterations in the niche synergize with stem cell-intrinsic changes to contribute to the development of ageassociated traits 17,66-68 . Aging has been proposed to drive a tissuedependent proinflammatory microenvironment that perturbs adult stem cell behavior [69][70][71][72] . Infiltration of immune cells into the stem cell niche 69 , or a transcriptional switch of stem cells, contributes to this proinflammatory environment that negatively feeds back to their own fitness 70 . Here we have characterized the effects of the proinflammatory cytokine IL-17 on skin aging. Our results show that elevated IL-17 signaling, secreted by aged dermal CD4 + T H cells, γδ T cells and ILCs, orchestrated many of the age-associated tissue dysfunctions by exertion of pleiotropic effects. IL-17-mediated signaling is heavily linked to the development of chronic inflammatory and autoimmune diseases 42,46,73 . In the skin, these diseases include psoriasis, pemphigus and alopecia areata 43 . Even if none of the clinical signs of these diseases are common with physiological aging, they share an increased aberrant IL-17-based signaling that impedes correct skin function. Our results strongly suggest that, intriguingly, the local environment of the aged skin resembles a low-level but persistent state of chronic inflammation, with deficient permeability and impaired wound healing that is reminiscent of what occurs in serious skin diseases such as psoriasis. Consequently, anti-IL-17 therapies, already approved for treatment of psoriasis, might be repositioned to other age-associated ailments such as excessive skin dryness or difficulty in repairing damaged skin in the elderly.

Mouse handling and husbandry
Mice were housed under a regimen of 120h/12-h light/dark cycles and specific-pathogen-free conditions. The temperature of the animal facility was maintained between 20 and 24 °C and humidity ranged from 45 to 65%. Animals were handled following the ethical regulations and guidelines of Scientific Park of Barcelona and the Government of Catalunya. All procedures were evaluated and approved by the Ethical Committee for Animal Experimentation of the Government of Catalunya (approval reference no. 10712). All mice used were from the C57BL/6 J strain. Aged mice were either bred in-house or were retired C57BL/6 J breeder females purchased from Charles River, and were kept until the desired age in the animal facility at Barcelona Science Park. Control adult mice were either bred in-house or purchased from Charles River to generate matching cohorts. Mostly female mice were used due to this fact, and also to differences in skin-digesting efficiency between sexes, because longer male skin digestion times are required and this reduced the survival of sorted dermal cells, skewing the results towards the most resilient cell types. For experiments that did not require enzymatical digestion of the dermis, both males and females were used. Mice were always killed during darkness, to coincide with their active phase. Aged mice were between 80 and 90 weeks of age  -log 10 (adj. P value) Aged/anti-IL-17A/F treated Article https://doi.org/10.1038/s43587-023-00431-z and adult mice were between 17 and 25 weeks of age. Mice with signs of skin inflammation (dermatitis, wounds or redness) were excluded from the experiments to avoid interference with the results.

Epidermal and dermal cell isolation for genomics analysis
Mice were killed and whole-torso skin was removed as rapidly as possible.

RNA-seq
For bulk RNA-seq of adult and aged, and aged/control IgG-treated and anti-IL-17A/F-treated, mice, four mice were used per condition. Cells in single-cell suspensions were frozen in 1 ml of Trizol (Invitrogen) for posterior RNA isolation. RNA was extracted from epidermal cell pellets frozen in Trizol using the RNeasy Mini Kit (Qiagen) and further processed for mRNA-seq with Illumina sequencing technology.

In vivo anti-IL-17A/F-neutralizing treatment
Cohorts of aging (73-week-old) mice were randomly distributed into two groups and were treated with either (1) a mixture of 105 μg of anti-IL-17A (clone 17F3, no. BE0173, BioXCell) and 105 μg of anti-IL-17F (clone MM17F8F5.1A9, no. BE0303, BioXCell) or (2) 210 μg of control IgG1 (clone MOPC-2, no. BE0083, BioXCell). Injections of 100 μl of antibody solution in PBS were administered intraperitoneally and performed three times per week during the dark cycle at the same time of day (2-3 h into the dark phase). After 12 weeks of treatment, mice were either (1) killed to obtain samples of either dermal cells for 10X scRNA-seq or epidermal bulk RNA-seq of the epidermis and histology analysis or (2) used to measure skin aging traits.

Epilation
Adult (n = 6, 18-20-week-old), aged/IgG control and aged/anti-IL-17A/F (n = 6, 85-week-old) female mice treated for 12 weeks with either IgG control (n = 6, 85-week-old) or anti-IL-17A/F were used (n = 6, 85 weeks old). Mice were anesthetized using a mixture of ketamine (75 mg kg -1 body weight) and medetomidine (1 mg kg -1 ) by intraperitoneal injection. Buprenorphine (0.05 mg kg -1 ) was injected subcutaneously as analgesic and anti-inflammatory treatment. An area of about 2-3 cm 2 of back skin was epilated using wax papers until hair was completely removed (usually two or three rounds of waxing). Atipamezole (1 mg kg -1 ) was injected to reverse anesthesia, and mice were left on heating pads until completely recovered. Mice were housed individually to avoid scratching and contact of the epilated areas. Animal health status was monitored daily. After 8 days post hair removal, mice were killed and images of shaved back skin captured. Samples of back skin were taken, fixed in neutral buffered formalin (10%) for 3 h at room temperature, dehydrated and embedded in paraffin blocks for later histological assessment.

Wounding and wound-healing experiments
Adult mice (n = 15, 18-20 weeks old) and aged mice (n = 14, 85 weeks old) treated for 12 weeks with IgG control as previously described, and aged male and female mice (n = 15, 85 weeks old) treated for 12 weeks with anti-IL-17A/F-, were used. For the IL-17-blocked group, IL-17A/F-blocking injections were stopped 1 week before starting the wound-healing assay to normalize IL-17 endogenous activity and separate the role of this cytokine in wound healing 74 from its role during aging. Mice were anesthetized using 3-4% isoflurane, and buprenorphine (0.05 mg kg -1 ) was injected subcutaneously for analgesia. Body temperature of the animals was maintained by placing them on a heating mat covered by a sterile surgical drape throughout the process. Hair was shaved in the back region and the area sterilized with 10% povidoneiodine in distilled H 2 O solution. Two skin biopsies were carried out per animal, one on each flank using a 5-mm-diameter sterile, disposable circular biopsy punch (5 mm diameter, Kai medical, no. BP-50F). Sharp scissors were used to specifically dissect the skin layer biopsy after using the punch, to avoid affecting other underlying tissue. Mice were placed on warm pads until fully recovered from anesthesia and then housed individually throughout the entire experiment. Wound areas were macroscopically assessed by daily measurement of length and width using a digital caliper (Traceable Digital Caliper). For wound area analysis, measurements of two wounds per animal were averaged. All wounds were collected, fixed in neutral buffered formalin (10%) for 3 h at room temperature, dehydrated and embedded in paraffin blocks for later histological assessment.

Barrier recovery measurement
Barrier integrity was assessed by measurement of TEWL, a parameter that measures water evaporation directly on the skin surface using a probe (Tewameter TM Nano, Courage+Khazaka). The animals used included adult (n = 6, 18-20-week-old), aged/IgG control (n = 7, 85-week-old) and aged/IL-17A/F-treated (n = 5, 85-week-old) female mice. Back skins were shaved 24 h before barrier disruption assay. Mice were anesthetized using 3-4% isoflurane and their temperature maintained constant at 37 °C by keeping them on a tightly regulated thermal plate during TEWL measurements. The mice were left to acclimate after induction of anesthesia and before starting tape stripping and measurements. All measurements were carried out at the same location by the same person, and room temperature and humidity were recorded. The order of data collection was randomized on the first occasion but was then replicated for each consequent measure.
The epidermal barrier was disrupted by tape stripping (Corneofix CF 20, Courage+Khazaka) and the number of tape strips adjusted according to mouse age to obtain TEWL > 20 g m -2 h -1 (ref. 75). A range of between six and ten strips was needed for adults, and 12-16 for aged/ IgG controls and aged/anti-IL-17-treated mice. For TEWL measurements the probe was used according to the manufacturer's instructions. The probe was placed in each location of the skin surface being measured for 40 s before stabilization, each measurement lasting 1 min. TEWL probe measurements were carried out immediately after tape stripping (t 0 ) and 6, 10, 24 and 48 h post tape stripping (t 6 , t 10 , t 24 , t 48 , respectively). Two measurements per animal were recorded, and we averaged their TEWL value to obtain one single value per animal and per time point. Also, for consistency, the probe was placed at these same two locations for each consecutive measurement. The locations used for measurements were carefully selected to avoid areas where tape stripping had caused the appearance of small wounds, to prevent alteration of the actual barrier recovery values. Article https://doi.org/10.1038/s43587-023-00431-z

Chromatin immunoprecipitation
Four independent mice per condition of adult, aged/IgG-treated control and aged/anti-IL-17A/F-treated were used for p65 chromatin immunoprecipitation sequencing (ChIP-seq). To isolate epidermal cells, the epidermal cell isolation protocol described below was followed (Epidermal and dermal cell isolation for genomics analysis). Instead of dispase II solution, 0.8% trypsin solution (trypsin 1:250, dissolved in PBS) was used to separate dermis from epidermis for 30-40 min at 37 °C. Around 30 million pelleted epidermal cells were used per reaction. The cells were crosslinked using Gold fixative (no. C01019027, Diagenode) following the manufacturer's directions. After two PBS washes, a second fixation was carried out in methanol-free 1% formaldehyde (no. 28908, Thermo Fisher Scientific) in MEM calcium-free medium with 10% calcium-chelated fetal bovine serum for 10 min with rotation. Glycine was then added to a final concentration of 125 mM to stop crosslinking, with incubation for 5 min under rotation. The pellet was washed twice with cold PBS and the cells then resuspended in 7.5 ml of swelling buffer (25 mM Hepes pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.1% NP-40 supplemented with 1× protease inhibitors without EDTA). This cell suspension was homogenized 50 times with a Dounce homogenizer and tight pestle (no. 885302-0015, Kimble) and the extracts were centrifuged (3,000g, 5 min). Pelleted nuclear extracts were resuspended (in onetenth of the volume used for the swelling buffer) in ChIP buffer (10 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 5 mM EDTA, 0.5 mM DTT, supplemented with 0.2% SDS and 1× protease inhibitors without EDTA). This suspension was incubated for 15 min and then transferred to sonication tubes (no. C01020031, Diagenode). Sonication was carried out in 30 cycles of 30/30 min on/off at 4 °C (Bioruptor Pico, Diagenode) and the sonicated supernatant was centrifuged (14,000g, 10 min). The supernatant containing the chromatin was quantified for ChIP-seq.
For immunoprecipitation, 30 μg of chromatin was diluted with ChIP buffer without SDS to dilute the SDS concentration to 0.1% in all samples. Then, 4 μg ml -1 anti-p65 (no. 8242, Cell Signaling Technology) was added and samples rotated overnight at 4 °C. Sepharose beads (no. 17-5280-01, Merck) were used according to the manufacturer's instructions, added to the tubes that included each sample and incubated with rotation for 4 h at 4 °C. Samples were then centrifuged (1,000g, 3 min) and washed with low (L)-and high (H)-salt buffers (L, H, L, H, L), centrifuging between washes. Low-salt buffer comprises 50 mM HEPES pH 7.5, 140 mM NaCl, 1% Triton and 1× protease inhibitors without EDTA, and high-salt buffer contains 50 mM HEPES pH 7.5, 500 mM NaCl, 1% Triton and 1× protease inhibitors without EDTA. A final wash was carried out with TE buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA). Immunoprecipitated chromatin was eluted in elution buffer (1% SDS and 100 mM NaHCO 3 , freshly prepared) by incubation of samples in the buffer for 30 min at 65 °C with agitation. Samples were then centrifuged (1,000g, 3 min), 5 M NaCl was added to a final concentration of 200 mM per sample followed by incubation for between 5 h and overnight at 65 °C, with agitation. Next, Tris-HCl pH 6.8 was added to a final concentration of 40 mM, EDTA to 10 mM, protease K to 50 μg ml -1 and incubation for 1 h at 45 °C. This product was purified with PCR columns.
ChIP-seq library construction and sequencing. Libraries were prepared using the NEBNext Ultra DNA Library Prep for Illumina kit (no. E7370) according to the manufacturer's protocol. Briefly, input and ChIP-enriched DNA were subjected to end repair and the addition of 'A' bases to 3′ ends, ligation of the NEB adapter and USER excision. All purification steps were performed using AgenCourt AMPure XP beads (no. A63882, Beckman Coulter). Library amplification was performed by PCR using NEBNext Multiplex Oligos for Illumina (96 Unique Dual Index Primer Pairs, nos. E6440, E6442, E6444, E6446). Final libraries were analyzed using either an Agilent Bioanalyzer or Fragment analyzer High Sensitivity assay (no. 5067-4626 or DNF-474) to estimate the quantity and check size distribution, and were then quantified by quantitative PCR using the KAPA Library Quantification Kit (no. KK4835, KapaBiosystems). Libraries were sequenced 1 × 50 + 8 + 8 base pairs (bp) on Illumina's NextSeq2000. Between 35 and 40 million reads were obtained per sample.
IL-17A: following heat-mediated antigen retrieval (20 min at 97 °C with citrate pH 6.0), sections were blocked with 10% goat serum in PBS for 1 h at room temperature. Primary antibody incubation (anti-IL-17A, no. ab79056, abcam 1:200 in EnVision FLEX antibody diluent, Dako) was performed overnight at 4 °C. After three washes in PBS, incubation of secondary antibody (anti-rabbit Alexa Fluor 488, no. A-21206, Molecular Probes, 1:400 in EnVision FLEX antibody diluent, Dako) was carried out at room temperature for 1 h. Nuclei were counterstained with DAPI (5 μg ml -1 for 10 min at room temperature). Six adult and five aged mice per condition were analyzed.
CD4: antigen retrieval was performed using BOND Epitope Retrieval 2 (no. AR9640, Leica). For primary antibody incubation, rat IgG2b kappa anti-CD4 (clone 4SM95, no. 14-976-682, Thermo Fisher) was used at 1:100 and incubated for 120 min using Leica BOND RX. Nonspecific unions were blocked using 5% goat normal serum mixed with 2.5% bovine serum albumin diluted in wash buffer for 60 min at room temperature. For secondary antibody incubation, goat anti-rat IgG (H + L) and Alexa Fluor 647 (no. A-21247, Thermo Fisher) were used at 1:500 for 60 min. Nuclei were stained with DAPI and slides mounted with fluorescence Mounting Medium (no. S3023, Dako -Agilent). Five adult and five aged mice per condition were analyzed.

FISH
For IL17A and IL17F, human adult (22-29 years old, n = 4, two women and two men) and aged (60-72 years old, n = 4, two women and two men) skin FFPE 5-μm tissue sections were purchased from Genoskin.
Scaled images were analyzed with Qupath (v.0.3.0 and v.0.3.2). For cornified layer thickness we performed ten measurements per mouse, which were then averaged. Eight mice per condition and age were used. We always ensured that the area analyzed was in the resting hair follicle stage (telogen), to avoid interference of this parameter with the desired quantification. For wound-healing assay, qualitative analysis of dermal granulation tissue and epidermal maturation were performed in a double-blind study by histopathologists. For anagen stage assessment we assigned each hair follicle to an anagen stage based on previous reports 55 .
For immunostaining, multiple selections of the region of interest of the skin dermis, excluding the epidermis and hair follicles, were done manually in two portions of skin. The 'positive cell count' algorithm was used to detect positive cells. Results are presented as the percentage of positive cells. For Il-17A, NDP view2 (v.2.9.29, Hamamatsu) was used to generate snapshots from representative areas. Brightness and contrast were adjusted to help distinguish positive cells from background, using the same parameters for both conditions. For CD4 immunofluorescence, the images displayed in the manuscript were captured with an SP5 confocal microscope (Leica). Images of whole stacks were projected on a single image and brightness and contrast were adjusted to enhance the signal of positive cells, using the same parameters for both conditions. For all immunostaining, these modified images were used exclusively for visualization and never for quantification.
For IL17A and IL17F FISH, skin upper dermis-excluding the epidermis, hair follicles and skin appendages-was selected manually. In case of tissue artifacts, these were excluded from the region of interest (for example, broken areas, folds). The positive cell count algorithm was used to detect positive labeling in the TRITC channel. Results are presented as the percentage of cells with IL17A or IL17F signal (either spots-single mRNA copy-or clusters-accumulation of several mRNA copies). For visualization purposes only, representative snapshots were obtained with NDP view2 (v.2.9.29m, Hamamatsu) and brightness and contrast were adjusted to help distinguish FISH signal from background, using the same parameters for both conditions.
For Rspo3 FISH, dermal papillae were selected manually in well-oriented telogen hair follicles. The 'Cell detection' algorithm was used to detect nuclei within the selected areas, and the 'Subcellular spot detection' algorithm to detect Rspo3-positive labeling in the TRITC channel in those nuclei. Individual nuclei were not always distinguishable due to high packing of dermal papillae cells. Therefore, FISH signal (spots and clusters) was not calculated per cell, but per dermal papillae. Five mice per condition were used. Graphical representations were performed with GraphPad Prism 9. For visualization purposes only, representative snapshots were obtained with NDP view2 (v.2.9.29, Hamamatsu) and brightness and contrast were adjusted to help distinguish FISH signal from background, using the same parameters for both conditions. All FISH quantifications were performed on the raw images and never on adjusted ones. Modified images were used exclusively for clearer visualization in figures.

Bulk RNA-seq library construction and sequencing
For the adult versus aged comparison, libraries were prepared using the TruSeq Stranded Total RNA Library Prep Kit with the Ribo-Zero Human/Mouse/Rat Kit (no. RS-122-2201/2202, Illumina) according to the manufacturer's protocol, using 150-300 ng of total RNA; for ribosomal RNA depletion, RNA was then fragmented for 4.5 min at 94 °C. The remaining steps were followed according to the manufacturer's instructions. Final libraries were analyzed on an Agilent Technologies 2100 Bioanalyzer system using the Agilent DNA 1000 chip to estimate the quantity and validate size distribution; libraries were then quantified by quantitative R using KAPA Library Quantification Kit KK4835 (no. 07960204001, Roche) before amplification with Illumina's cBot. Finally, libraries were sequenced on the Illumina HiSeq 2500 sequencing system using paired-end 50-bp-long reads.
For IL-17-neutralized versus control samples, libraries were prepared using TruSeq stranded mRNA Library Prep (no. 20020595, Illumina) according to the manufacturer's protocol, to convert total RNA to a library of template molecules of known strand origin suitable for subsequent cluster generation and DNA sequencing. Briefly, 500-1,000 ng of total RNA was used for poly(A)-mRNA selection with poly-T oligonucleotides attached to magnetic beads, and two rounds of purification. During the second elution of poly-A RNA, RNA was fragmented under an elevated temperature and primed with random hexamers for complementary DNA synthesis. Cleaved RNA fragments were then copied into first-strand cDNA using reverse transcriptase (SuperScript II, no. 18064-014, Invitrogen) and random primers. Note that the addition of actinomycin D to the First Stand Synthesis Act D mix (FSA) improved strand specificity by prevention of spurious DNAdependent synthesis while allowing RNA-dependent synthesis. Secondstrand cDNA was then synthesized by removal of the RNA template and synthesis of a replacement strand, incorporating dUTP in place of dTTP to generate double-stranded cDNA using DNA polymerase I and RNase H. These cDNA fragments then had a single A base added to the 3′ ends of the blunt fragments, to prevent them from ligating to one another during adapter ligation. A corresponding single T-nucleotide on the 3′ end of the adapter provided a complementary overhang for ligation of the adapter to the fragments, ensuring a low rate of chimera (concatenated template) formation. Subsequent ligation of the multiple indexing adapter to the ends of dscDNA was carried out. Finally, PCR selectively enriched DNA fragments with adapter molecules on both ends, and the amount of DNA in the library was amplified. PCR was performed with a PCR Primer Cocktail that anneals to the ends of the adapters. Final libraries were analyzed using either Bioanalyzer DNA 1000 or Fragment Analyzer Standard Sensitivity (Agilent) to estimate the quantity and validate size distribution; libraries were then quantified by quantitative PCR using KAPA Library Quantification Kit KK4835 (Roche) before amplification with Illumina's cBot. Finally, libraries were sequenced on the Illumina HiSeq 2500 sequencing system using single-end 50-bp-long reads.

10X scRNA-seq
Stained dermal single-cell suspensions were prepared as described above. CD45 + and CD45 -EpCAMcells were FACS sorted separately in a BD Fusion cell sorter to enrich for the less abundant CD45 + cells following the sorting strategy shown in Supplementary Fig. 1a. Cells were collected in PBS + 0.5% bovine serum albumin at 4 °C in LoBind tubes (Eppendorf) and processed immediately with the microfluidics Chromium platform (10X Genomics). For the adult and aged experiment, three technical replicates were performed for CD45 + sorting. Replicate 1 included skin dermal cells of three adult and one aged mice, replicate 2 had two adult and one aged mice and replicate 3 comprised two adult and two aged mice. For CD45 -EpCAMcell sorting two replicates were carried out, both comprising one mouse per condition. For aged/IL-17-blocked and aged/IgG control 10X scRNA-seq of dermal skin cells, both CD45 + and CD45 -EpCAMconditions included four replicates in total. In both conditions, replicate 1 consisted of two aged/IL-17-blocked and two aged/IgG control mice. Replicates 2, 3 and 4 consisted in all cases of one aged/IL-17-blocked and one aged/ IgG control mouse.

Data preprocessing
Sequences were demultiplexed and aligned according to the Cell Ranger pipeline (v.6.0.0) with default parameters. Sequencing reads were mapped against the mouse GRCm38 reference genome to generate feature-barcode matrices separately for all CD45 + and CD45replicates.  80 . Replicates were merged, analyzed and annotated separately for CD45 + and CD45datasets before integration. Cells were filtered with >10% of mitochondrial gene content and genes not found in at least five cells. As part of quality control, cells situated between the minimum and first quartile (according to the distribution of number of genes per cell of each compartment dataset) were removed. To avoid contamination of epithelial cells in the immune compartment, EPCAM + cells found in CD45 + sorted cells were filtered out.
In addition, to remove technical bias arising after merging the different replicates we computed differentially expressed genes among replicates using the function FindAllMarkers, then we ensured that the top 500 differentially expressed genes between replicates were not present in highly variable genes (HVGs). Thus we removed the intersection between computed differentially expressed genes and HVG.
To evaluate integration of the different replicates we used the local inverse Simpson's index (LISI) (https://github.com/immunogenomics/ LISI) 30 . This score defines the effective number of samples in the local neighborhood of a cell. Under ideal mixing we would expect to obtain LISI scores equal to the number of different replicates in our datasets. This indicates that the neighborhoods are well represented by all samples and that the cell types/states previously identified exhibit good mixing across all replicates. In addition we assessed the performance of replicate integration by qualitative inspection of UMAPs. We checked that all cell types and states were well represented by each replicate and that our dataset did not contain any technical cluster entirely driven by replicate effect.

Clustering
Cell-to-cell variations were normalized by expression values using a scale factor of 100,000 and log transformation. Gene expression measurements were scaled and centered. Scaled z-score values were then used as normalized gene measurement input for both clustering and visualization of differences in expression between cell clusters. HVGs were selected by assessment of the relationship of log(variance) and log(mean) and choosing those with the highest variance:mean ratio. Principal component analysis was used to reduce the dimensionality of the dataset, and ElbowGraph to select the number of dimensions for the clustering of significant principal components. Cluster identification was performed using the functions FindNeighbors and FindClusters, which calculate k-nearest neighbors and generate the shared nearestneighbor graph to cluster cells. The algorithm applied was the Louvain method, which allows tuning of the number of clusters with a resolution parameter. To explore clusters in more detail, either the resolution parameter was increased in FindClusters function or FindSubCluster was used for specific clusters. UMAP was used as a nondimensional reduction method to visualize clustering.
In the case of the aged/control IgG-treated versus aged/anti-IL-17A/ F-treated fibroblast clusters, we conducted a subclustering analysis of all populations obtained to detect potential fibroblast populations responding to the blocking treatment. With this we separated the previously named fibroblast 1 cluster obtained in the aged versus adult comparison into three subclusters (fibroblasts 1.0-1.2). Then, fibroblast 2 cluster was divided into two new clusters (fibroblasts 2.0 and 2.1) and fibroblast 3 cluster was separated into four clusters (fibroblasts 3.0-3.3); fibroblast 4 and 5 clusters remained as separate units. This separation allowed detection of subtle differences in the IL-17A/F blockade response in fibroblast subgroups.

Cell type annotation
For annotation of cell types and states, enriched genes were first identified in each of the clusters with the function FindAllMarkers using the Wilcoxon rank-sum test to find cluster-specific markers. These cluster-specific genes were then explored to find previously reported cell population marker genes. Examples of marker genes used to annotate the cell populations include: CD4 + T cells:

Differential expression analysis for each cluster
To find differentially expressed genes between adult versus aged, and IgG-treated versus anti-IL-17A/F-treated, in different annotated cell type populations, differential expression analysis was performed between conditions for each cluster with the function FindMarkers. To control for multiple comparisons we used the FDR Benjamini-Hochberg method.

Age effect analysis
To assess age-related effects among immune cells, a deep learning approach was created based on autoencoders. During training this method can identify features relevant to the data structure and then use them to predict different cell types in a test dataset. To generate the model, adult data were divided into two smaller subsets that could be used as training and test sets. The model displayed high probability scores (P > 0.5) in the prediction of cell types from the test set among the different immune cell types and a low rate of unpredicted cells (unclassified rate <5%). If the rate of unclassified cell type is strictly biased for aged cells, this would reflect the effect of aging on the cell phenotype. Based on this assumption an age-deviance score was defined as 1 - q, where q is the probability of the unpredicted cell being in the true corresponding cell type class in aged cells. The proportion of unclassified cells within each cell type in aged cells, and the proportion of age-affected cells, were then measured and normalized by the model's error rate (for example, by subtracting the cell proportion in each adult cell type that could not be predicted by the model).

Cell composition analysis
To evaluate the significance of differences in cell type abundance between conditions we applied sccomp 81 , a tool designed for differential composition and variability analyses that relies on sum-constrained independent beta-binomial distributions. This model uses meanvariability association, which allows modeling of the compositional properties of the data while enabling the exclusion of outliers. We assessed significant differences (FDR < 0.025, using a Benjamini-Hochberg procedure to control for multiple testing) for composition and/ or group-specific variability in each cell type population, comparing adult versus aged conditions and aged/anti-IL-17A/F-treated versus aged/IgG-treated control.

GO analysis
GO analysis for scRNA-seq was performed on lists ranked by increasing Benjamini-Hochberg-adjusted P values with g:Profiler 86 (https://biit. cs.ut.ee/gprofiler/gost), using the biological processes database. A GO category was considered significant with adjusted P < 0.05. In the aged versus adult comparison, genes with fold change > 0.35 (log 2 ) were considered. In the aged/anti-IL-17A/F-treated versus aged/IgGtreated control, genes with fold change > 0.25 (log 2 ) were considered.

Statistics and reproducibility
No statistical methods were used to predetermine sample size. This was decided taking into the account variability between samples, particularly in aged mice, and was based on our expertise with the skin of aged mice 1,2,22 .
In general graphs show individual values and median (depicted as a line) and P values were obtained using the nonparametric two-tailed Mann-Whitney U-test with Prism v.9, unless otherwise stated in the figure legends.
Data collection and analysis were not performed blind to the conditions of the experiments, except for the histopathological wound healing analysis.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Cell−group proportion
Not-significant changes    Ig G co n tr o l A n ti−