A protein phosphatase network controls temporal and spatial dynamics of differentiation commitment in human epidermis

Key results

1. ShownThe 4 h commitment window is a distinct cell state. Protein levels track transcripts poorly and dephosphorylation dominates, which points to post-transcriptional control of commitment.

   Clonogenicity fell sharply between 4 and 8 h. t-SNE and clustering separated 4 h from 8 and 12 h. Protein-transcript Pearson r was 0.19 at 4 h vs 0.51 (8 h) and 0.68 (12 h). Over 3,500 phosphosites (Andromeda score >= 30) were quantified; about two thirds of 4 h changes were dephosphorylation not explained by protein abundance. · Results and Discussion; Figure legends

2. ShownAn siRNA screen of phosphatases induced at commitment identified DUSP6, PPTC7, PTPN1, PTPN13 and PPP3CA as pro-commitment factors. Depleting them increases self-renewal, and depleting DUSP6, PTPN1, PPP3CA or PTPN13 blurs basal versus suprabasal identity in reconstituted epidermis.

   45 phosphatases changed from 0 to 4 h (20 up). Silencing seven of the 20 raised clonal growth, the five named most strongly (p-value <0.001; n=3 screens, 3 cultures each). Knockdowns delayed TP63 loss and TGM1 gain. In organotypic cultures (n=2 transfections) TP63+ fractions rose and TP63/Ki67+ cells appeared suprabasally. · Results and Discussion; Figure legends

3. ShownPro-commitment phosphatases lower ERK1/2 activity and promote induction of MAF-family AP1 factors, since their knockdown blunts it. DUSP10, also induced at 4 h, opposes commitment, and its knockdown does not raise ERK phosphorylation.

   ERK2 (MAPK1) and ERK1 (MAPK3) sites ranked in the top 15 most dephosphorylated at 4 h. Knockdowns raised phospho-ERK1/2 vs scrambled control (n = 2 blots) and blunted MAF, MAFB, MAFG induction. DUSP10 knockdown cut clonogenicity (p-value<0.001) without raising phospho-ERK. Overexpressed DUSP10 raised and DUSP6 lowered colony formation; DUSP6 C293S had no effect. · Results and Discussion; Figure legends

4. SuggestedPhosphatase cross-regulation rewires over time: mostly negative at 0, 8 and 12 h, but almost entirely positive around a self-amplifying DUSP6 at 4 h. Boolean reasoning indicates the network must switch, framing commitment as an unstable switch.

   Knockdown-by-time RT-qPCR networks (2-way ANOVA, Holm-Sidak) show all but one interaction positive at 4 h. RE:SIN needed added possible edges to satisfy constraints. RE:IN found that no single Abstract Boolean Network reproduces the dynamics. PKCi 12 h expression fit the 0 h network; neither PKCi nor TSA reached the 12 h network. · Results and Discussion; Network analysis; Figure legends

5. SuggestedCommitment phosphatases are spatially patterned in the basal layer of human epidermis, and the pattern is reproduced on topographic substrates in vitro. This implies that dermal cells are not required for the patterning.

   Wholemount confocal imaging: DUSP6, PTPN1 and PPP3CA were highest in beta1-integrin-high stem cell clusters, PPTC7 was enriched in integrin-low regions, and PTPN13 and DUSP10 were uniform. On collagen-coated PDMS substrates mimicking the epidermal-dermal junction, DUSP6-high clusters formed on feature tops while DUSP10 stayed uniform. · Results and Discussion; Figure legends