Macrophages induce AKT/β-catenin-dependent Lgr5+ stem cell activation and hair follicle regeneration through TNF

Skin stem cells can regenerate epidermal appendages; however, hair follicles (HF) lost as a result of injury are barely regenerated. Here we show that macrophages in wounds activate HF stem cells, leading to telogen–anagen transition (TAT) around the wound and de novo HF regeneration, mostly through TNF signalling. Both TNF knockout and overexpression attenuate HF neogenesis in wounds, suggesting dose-dependent induction of HF neogenesis by TNF, which is consistent with TNF-induced AKT signalling in epidermal stem cells in vitro. TNF-induced β-catenin accumulation is dependent on AKT but not Wnt signalling. Inhibition of PI3K/AKT blocks depilation-induced HF TAT. Notably, Pten loss in Lgr5+ HF stem cells results in HF TAT independent of injury and promotes HF neogenesis after wounding. Thus, our results suggest that macrophage-TNF-induced AKT/β-catenin signalling in Lgr5+ HF stem cells has a crucial role in promoting HF cycling and neogenesis after wounding.

(a, b) The number of anagen follicles was proportional to the size of excisional wounds. Representative images showing areas of anagen follicles around the wound at the 15 th day post-wounding (PWD-15) (a). The numbers of anagen follicles in different groups were quantified (b). n = 7 for each wound size. Data are expressed as the mean ± s.e.m. *P<0.05, **P<0.01, unpaired t-test, two-tailed.
(c) Immunofluorescence (IF) analysis showed densely populated CD45 + leukocytes in the wounded skin. Double staining for CD45 and F4/80 showed that macrophages accounted for the majority of the inflammatory cells in the wound-adjacent tissue at PWD-3. For all IF analyses, representative images from 8-16 tissue sections of wounds in 4-6 mice are shown.
(f-h) Dex or JSH-23 treatment completely abolished the wounding-induced telogen-anagen HF transition as assessed at PWD-15. n = 8 mice for each group.
(a) IF analysis of normal skin of CX3CR1 CreER/+ :R26 iDTR/+ mice showed that in the unwounded skin, the CX3CR1 + residential macrophages are widely distributed in the skin tissue, while the Ly6C + myeloid-macrophages are rarely detected (upper channel). CX3CR1-YFP cells are positive for F4/80 and largely co-localized with MHC II + cells (lower channel).
(b) Real-time PCR from wound adjacent tissues (2 mm in width) revealed the kinetics of CCL2 expression at different times post-wounding (PWD: 0, 1, 2, 3, 5, 7, 10, 14). (Gene expression was normalized to GAPDH with 40 cycles, data are represented as the mean ± SD, and n = 3.) (c) IF analysis of the wound adjacent tissue at different times for the expression of CCL2. In unwounded skin, CCL2 was detected in the hair germ (HG) (yellow arrow); at PWD-1, CCL2 was detected in the HG, hair follicle infundibulum (IN) and some epidermal cells and dermal cells. At PWD-3.5, the levels of CCL2 increased in dermal cells but decreased in the HG and IN. Scale bars, 50 m.

Supplementary Figure 4
(a-c) Real-time PCR analysis of cells derived from wounds (PWD-3) showed much higher TNFA expression in macrophages than in epidermal cells, neutrophils and the rest of the cells (Other cells). (b, c) Double staining of wound tissue (PWD-3) indicated that TNF was closely co-localized to F4/80 + macrophages (b) and Ly6C + macrophages, but barely to CX3CR1 + cells (c). Scale bars, 50 μm.
(d) LPS-activated macrophages induced HF TAT, while LPS-activated macrophages with TNFA knockout failed to induce HF TAT. n=6 mice for each group.
(e) IF analysis showed that TNFR1 was relatively more highly expressed in the hair follicle and the basal layer of the epidermis, and TNFR2 showed higher expression in the upper layer of the epidermis.
(f) Real-time PCR results showed that in the Tg-TNF mice, the mRNA of TNF in wound-adjacent tissue (PWD-3) was approximately 3-fold higher than that in the non-wounded skin. Data are expressed as the mean ± s.e.m.* P<0.05, ** P<0.01, unpaired t-test, two-tailed.
(a, b) IF staining indicated that p-AKT was barely detected in the epidermis and HF of unwounded skin. Six hours (h) post-wounding, p-AKT was detected in cells in the basal layer of the epidermis adjacent to the wound, which were also strongly positive for integrin-6, and increased levels of p-AKT were found at 12 and 24 hours. Five mice for each time point, 2 excisional wounds per mouse, and 5-6 tissue sections per wound were analyzed.
(c-e) Treatment of 8-week-old C57/B6 mice with perifosine (c) or LY294002 (d) suppressed hair depilation-induced p-AKT in epidermal and follicle cells and HF TAT. Control mice, n = 9; perifosine-treated mice, n = 8; LY294002-treated mice, n = 8. Local injection of bpV (phen) into the skin without depilation induced p-AKT in epidermal and HF cells and HF TAT at the injection site (e). n = 8 for both control mice and mice treated with bpV. Scale bars, 50 μm.
(a) TNF-α markedly increased the expression of Wnt3a, Wnt7b and Wnt10b, but not that of Wnt5a, Wnt7a and Wnt10a, and blockade of PI3K with LY294002 or blockade of AKT with perifosine greatly attenuated TNF-α-induced Wnt ligand expression.
(b) TNF-α exerts a weak effect on the expression of Wnt ligands in fibroblasts.
(c) The efficiency of Lgr5-mTNFR-ShRNA (ShRNA) lentiviruses in the down-regulation of TNFR1 was examined in Lgr5 + follicle stem cells isolated from Lgr5-EGFP mice by cell sorting for EGFP and CD49f double positive cells, and a mock ShRNA (Sh-C) and culture medium alone (NC) were used as controls; the levels of TNFR1 expression and p-AKT after TNF- treatment were determined by Western blot analysis, which was repeated 4 times.
(d) A schematic illustration of the experimental scheme of the in vivo knockdown of TNFR1 using ShRNA lentiviruses and its influence on wounding-induced HF TAT.
(e, f) IF analysis of tissue sections of the skin at the injection site. ShRNA-TNFR1 detected the expression of ShRNA (green) in Lgr5 + cells (e). The expression of TNFR1 in Lgr5 + cells in the HF was substantially down-regulated in mice that received ShRNA-TNFR1 compared to mice receiving the mock sequence (f). Scale bars, 50 μm. (Fig.1c). Firstly, we tested the quality and specificity of the antibodyies (anti-CD11b and F4/80) with a macrophage cell line, set as a positive control. In the analysis, plot was firstly gated for the single cell population (FSC-H vs. SSCA), to remove the debris, air bubbles and laser noise (all which should be FSC-low) (a). The single cell population was further analyzed for their expression of CD11b (FITC) and F4/80 (PE) (c,d); cells stained with FITC/PE Rat IgG2b isotype antibodies were stained as a negative control (b). In the analysis of skin single cell suspensions, plot was also firstly gated for the single cell population (FSC-A vs. SSCA) (e), then the cells were analyzed for the expression of CD11b (FITC) and F4/80 (PE). Similarly, FITC/PE Rat IgG2b isotype antibodies were stained as a negative control, and thus the CD11band F4/80cells were gated accordingly (f). With the above parameter of negative and positive controls, cells positive for CD11b and F4/80 in Figure 1c were gated in the plot. (Figure 1h and i in main manuscript). In FACS analysis of blood macrophages, plot was firstly gated for single cells (wild type mice) (FSC-A vs. SSCA) (a), then the single cells were analyzed for the expression of CX3CR1 (FITC) (c) and Ly6C (Alexa Fluor 647) (d); Alexa Fluor 647 Rat IgG2b/FITC Mouse IgG2a, κ isotype antibodies stained cells served as a negative control (b). The blood lysis of CX3CR1CreER-YFP mice was also analyzed, to detect the sub-population of CX3CR1 + cells. With the parameters of negative control and single stains based compensation, cells positive for CX3CR1 or Ly6C were gated in Figure 1h and i. Figure 3d). In FACS analysis of blood macrophages, plot was firstly gated for single cells (wild type mice) (FSC-A vs. SSCA) (a), and the cells were analyzed for the expression of CX3CR1 (FITC) and Ly6C (Alexa Flour 647); Alexa Fluor 647 Rat rat IgG2b/FITC Mouse IgG2a, κ isotype antibodies stained cells served as a negative control (b). With the gating parameters of negative control, and single staining based compensation, gates for CX3CR1+or Ly6C+ cells were gated (c). The gating strategy was used in Supplementary Figure 3d. In (d) and Supplementary figure 3d, the graph type was present in "Density" style.