Supplemental Methods

Mouse strains

K14-RtTA/TRE-miR-125b transgenic mice were in strain FVB. Details of their derivation and Dox-induction were described previously (Zhang et al. 2011).Nude mice (B6.Cg-Foxn1nu/J) were purchased from Jackson labs.

Animal care and use

All animal experiments were performed in the AAALAC-accredited Comparative Bioscience Center at The Rockefeller University. Experiments were in accordance with NIH guidelines for Animal Care & Use, approved and overseen by The Rockefeller University’s Institutional Animal Care and Use Committee.

Antibodies and immunoblot analysis

Primary antibodies used and their dilutions are listed in Supplemental Table 2. All immunoblots were scanned and quantified using the Odyssey Infrared Imaging System (LI-COR) and standard reagents/protocols provided by the vendor. All of the immunoblots were developed using a 2-color detection model. Specifically, αTubulin was used as the internal control and detected by IRDye® 680 (Red) conjugated secondary antibodies. The proteins of interest were detected by IRDye® 800 (Green) conjugated secondary antibodies. The two channels were simultaneous blotted, scanned and quantified with the Odyssey System. For FACS, β1-PE-Cy7 (ebioscience, Hamster, 1:200) and α6-Alex647 (AbDSerotec, 1:50) antibodies were used.

Immunofluorescence and in situ hybridizations

For immunofluorescence, fresh tissue was embedded in OCT, frozen, sectioned and fixed in 4% paraformaldehyde (Kaufman et al. 2003). In situ analysis of miR-125b was conducted following our previously described protocol (Zhang et al. 2011).

Cell culture

The human SCC-13 and SCC-15 cell lines were gifts from the Rheinwald lab (Rheinwald and Beckett 1980). HaCAT keratinocytes and A431, and Fadu SCC cell lines were from ATCC. Normal human keratinocytes (HEK) were purchased from Lonza. The mouse SCC CSC cell line 63A was generated previously in the Fuchs’ lab (Schober and Fuchs 2011). WT and DTG mouse keratinocytes (MKs) were isolated from backskin of P0 CD-1 and DTG mice, respectively. HEKswere cultured in growth factor supplemented KGM Medium. All other human cellswere cultured in E Medium(Rheinwald and Beckett 1980). SCC-63A, DTG and WT MKs were cultured in E low calcium Medium (E medium with 0.03 mM calcium). For suspension cultures, cells were trypsinized and plated into the Costar Clear Flat Bottom Ultra Low Attachment Multiple Well Plates in E low calcium Medium. MTT Cell viability assays were conducted using CellTiter 96® AQueous One Solution Cell Proliferation Assay (Promega) following the manufacturer's protocol. Apoptosis was assayed by Pacific Blue™ Annexin V/SYTOX® AADvanced™ Apoptosis Kit (Life tech). EDU incorporation was assayed by Click-iT® EdU Alexa Fluor® 647 or 594Flow Cytometry Assay Kit (Life tech). The P38/JNK MAP kinase inhibitor SB203580 was purchased from Cell Signaling and used at 10 µM final concentration.

Overexpression and knockdown of miR-125b in tumor cell lines.

For knockdown of miR-125b function, SCC63A-CSCs (Schober and Fuchs 2011)were first infected with lentiviruses harboring either GFP-control or GFP-miR-125b-Sponge. 48hrs later, transduced GFP+ cells were FACS-isolated and allowed to expand in culture without passage. These cells were then transfected overnight with 50nM Scramble or miR-125b antagomir(Zhang et al. 2011) in Hiperfect transfection reagent (Qiagen) according to the vendor’s protocol. GFPHI cells were FACS-isolated and immediately used for transplantation. A431 cells were transfected overnight with 50nM Scramble or miR-125bantagomir and intradermally transplanted with 50μM corresponding antagomir in the injection mixture. All miR-125b or scramble control overexpression experiments, other than those of the DTG MKs and luciferase assays, are conducted by infecting subject cells with PLKO-miR-125b or scramble overexpression lentivirus followed by puromycin selection of infected cells (Zhang et al. 2011). Successful overexpression of miR-125b in each of the selected cell lines was confirmed by qPCR (Data not shown).

Luciferase reporters and Luciferase assay

Dual Luciferase reporter assays in cultured mouse keratinocytes were conducted as described (Zhang et al. 2011). The previously-used pmiRGLO (Promega) dual luciferase reporter vector was modified by replacing the original SV40-hRluc-Neocassette with a CMV-hRluc cassette to improve its performance in keratinocytes. The PCR primers for cloning 3’UTR fragments of miR-125b target genes are listed in Supplemental Table 2. Mutations in the seed region of predicted miR-125b binding sites on these 3’UTR reporters were created using the Phusion Site-Directed Mutagenesis Kit (Finnzymes). The primers used for creating those mutations are also listed in Supplemental Table 2. For transfection in a 96 well format, miR-125b overexpression was achieved by co-transfection of 5ng Reporter plasmid + 100ng PLKO1-miR-125b plasmid(Zhang et al. 2011) + 5pmol Pre-miR-125bmiRNA Precursor (Applied Biosystems). Scrambled control overexpression was achieved by co-transfection of 5ng Reporter plasmid + 100ng PLKO1-Scramble plasmid(Zhang et al. 2011) + 5pmol Control Pre-miRNA Precursor (Applied Biosystems). For assaying miR-125b sponge activity, 5ng Reporter were co-transfected with 100ng Sponge construct. The Scarb1 3’UTR reporter and miR-125b sensor have been described (Zhang et al. 2011). All transfection are done with the Effectene reagent (Qiagen).

Plasmid Constructions

Sponge inhibitor for miR-125b: The sponge sequences are listed in Supplemental Table 2. A GFP-miR-125b Sponge or GFP-control Sponge cassette was cloned into the NheI-SalI sites of the pCDH-EF1-MCS-IRES-Puro cDNA Cloning and Expression lentiviral Vector (System Biosciences) respectively. In these constructs, the GFP-Sponge cassettes were placed after the EF1 promoter and in exchange of the original IRES-Puro cassette. ShRNA constructs for knocking down mouse and human VPS4B:the primers for making mouse and human VPS4B shRNA constructs are listed in Supplemental Table 2. As previously described (Beronja et al. 2010), the primers were annealed and cloned into the AgeI-EcoRI sites of the pLKO1 vector. The shRNA constructs were packaged into lentivirus and used to infect target cells. Transduced cells were selected by puromycin. Overexpression construct for VPS4B: The U6 promoter on the pLKO1 vector (with PGK-puro selection marker)was first replaced by the Tre2 promoter from the Tre2-miR-125b construct (Zhang et al. 2011) to make a pLKO1-Tre2 vector. Mouse VPS4B cDNA was purchased from Open Biosystems (Clone ID: 3488471). The cDNA sequence was then PCR amplified using primers listed in Supplemental Table 2 and cloned into the PLKO-Tre2 vector after the Tre2 promoter to make the pLKO1-Tre2-VPS4b construct, which was then packaged into lentivirus and used to infect DTG MK cells. The transduced cells were selected by puromycin. NotI-EcoRI sites were used to swap the Tre2-VPS4b cassette from pLKO1 vector to pLKO-H2BRFP vector (Beronja et al. 2010).

Illumina RNA-seq

For RNA seq, RNA samples were submitted to the Genomics Resources Core Facility of the Weill Cornell Medical College for library construction using Illumina TruSeq Stranded mRNA Sample Prep Kit and then sequencing using Illumina HiSeq2000. Results were analyzed via the Galaxy web platform (Giardine et al. 2005; Blankenberg et al. 2010; Goecks et al. 2010), using TopHat for initiate mapping and Cufflinks for transcripts assembling and expression level estimation (Computing FPKM: fragments per kilobase of exon per million fragments mapped). The MM9 genome assembly (UCSC Genome Browser (Karolchik et al. 2012; Rosenbloom et al. 2012)) was used as reference genome for all analyses. Low expression genes (Max FPKM <10 in all the samples) were excluded from all subsequent analyses. All miRNA qPCR analyses were conducted using the Taqman miRNA RT-PCR system (Life Tech) according to the vender’s protocol.

Kaplan Meier Curves

GraphPad Prism software was used to generate the Kaplan Meier curves and to calculate P value by two tailed log-rank test.

Data Analysis

To calculate the miR-125b correlation score (CS) for each protein-coding gene, Pearson’s correlation coefficient r was first calculated between its mRNA expression level (FPKM) and corresponding miR-125b level (determined by miRNA RT-PCR) in the same sample across all samples. The P value of each r was then calculated by t-test. Based on these, the CS for each gene was assigned as following: if r>=0 then CS= |log10(P)|, otherwise if r<0 then CS= -|log10(P)|. Therefore, high positive CS values indicated strong positive correlation while high negative CS value indicated strong negative correlation. Unsupervised Hierarchical clustering analysis was conducted using the Cluster software (Eisen et al. 1998). Gene set enrichment analysis (GSEA) was conducted using Pre-ranked model of GSEA software (Mootha et al. 2003; Subramanian et al. 2005). Pre-defined gene sets are either extracted from our previous studies (Schober and Fuchs 2011; Zhang et al. 2011) or the Molecular Signatures Database (MSigDB) (Subramanian et al. 2005). Official gene symbols are used for cross species comparison between mouse and human gene sets. TCGA data were extracted and analyzed by the cBioPortal tools (Cerami et al. 2012). P values for GSEA analysis, cBioPortal tool analysis of TCGA data andKalplan Meier Curves analysis with Graphpad Prism software were generated using the default setting of the built-in functions of the corresponding softwares used. All other P values were calculated by Student's t-test.

SupplementalFIGURE LEGENDs:

Supplemental Figure S1: (A) In situ hybridization of DMBA/TPA induced mouse SCC. Scale bar 50μm. (B) In situ hybridization of normal human epidermis, hair follicle and actinic keratosis. Scale bar 100μm. Arrow: potential bulge region. (C) Copy number variation of miR-125b-1/2 and their adjacent miRNA genes in genome of human HNSCCs (TCGA).

Supplemental Figure S2: (A) In vitro colony formation assay. Left panel: low passage (<4) WT MK cells transduced with miR-125b or scrambled control overexpression lentivirus. Right panel: high passage (>10) DTG MK cells, uninduced or Dox induced. n=3 independent wells, 2 tail unpaired t-test. (B) MTT Cell proliferation/survival assay of cultured WT MKs, normal bulge SCs, and SCC CSCs in suspension culture (n=8 independent wells). *: P<0.05, **: P<0.01, unpaired 2 tail t test, same in following. (C, D) MTT Cell proliferation/survival assay of DTG MKs or HaCAT cells in attached culture (n=8 independent wells each). (E, F) The % of apoptotic/anoikis cells [AnnexinV(+) and Dead cell stain(-)] and proliferating cells (EDU+ after 1hr labeling prior to harvesting) in attached culture or following 12hrs in suspension culture of SCC15, A431 or Fadu Cells (n=3 independent wells).

Supplemental Figure S3: (A) Photographs of spontaneously formed skin tumors in F1 line and F2 line ofmiR-125b overexpressing DTG transgenic mice. (B) Photographs of skin tumors that formed spontaneously after miR-125b overexpression and then were monitored after Dox withdrawal for 0 and 12 days as indicated. (C) Photographs of DTG and Tre-miR-125b mice following 10 weeks of DMBA/TPA treatment. Note that only DTG mice with Dox-induction of miR-125b overexpression developed large numbers of tumors. Scale bars: 1cm. (D) Photographs of DTG and K14-rtTA mice following 14 weeks DMBA/TPA, TPA alone or single time DMBA alone treatment. All mice on Dox. Scale bars: 1cm. (E)Colony formation assay of FACS-isolated α6hiβ1hi and α6lowβ1low cells from DMBA/TPA-induced DTG tumors after 2 weeks culture.3 independent wells were analyzed for each (n=3). *: P<0.05 based on unpaired two-tail t test, error bars represent standard error.(F)Photographs of Nude mice intradermally transplanted with FACS-isolatedα6hiβ1hior α6lowβ1low DTG tumor cells, as well as α6hiβ1hi cells from the skin of K14-rtTA mice on Dox food (negative control). (I) miRNA Q-PCR of DTG CSCs (FACS-isolatedα6hiβ1hiGFP+ cells from GFP+ DTG tumors), in comparison with SCC CSCs and normal adult skin SCs. Bulge: HF Bulge SCs, Epi: Epidermal Basal progenitors. WT and cKO SCC CSC: CSCs (α6Hiβ1Hi) of DMBA induced SCCs from WT skin or skin conditionally ablated for TGFβ receptor II (cKO). (J) Oil Red O staining of tumors derived from DTG skin.

Supplemental Figure S4: (A) Photographs of engrafted DTG tumors. The tumors were first allowed to grow up on Dox, then switched to Off-Dox condition for long-term (up to 120 day), then switched back to Dox (Re-Dox) for 30 days. (B) Cleaved Caspase 3 and TUNEL staining and quantification of engrafted GFP+DTG tumors during of Off-Dox regressing process. n.s.: no significant difference (P<0.05, 2 tail t-test) found between any two data points. (C) FACS profile and quantification of regressing GFP+DTG tumors. Arrow point to the α6lowβ1lowpopulation that quickly expanded after Dox withdrawal. **: P<0.01, 2 tail t-test. (D) MiRNA qPCR analysis of FACS isolated DTG CSCs (GFP+α6hiβ1hi) following DOX withdrawal for 0 day (n=2, independent tumors), 4days (n=3, independent tumors), and 7 days (n=3, independent tumors). Shown also are miR-125b levels in WT adult hair follicle (bulge) SCs (n=3), error bars: standard error. (E) Dual Luciferase reporter assay in cultured WT MK cells to examine the effectiveness of miR-125b sponge construct. Scarb1 is a previously identified miR-125b target in skin cells(Zhang et al. 2011). Vertical axis indicates relative Luciferase activity. *: P<0.05, **: P<0.01, T-test, N=4 independent wells for each sample, error bars: standard error. (F) FACS profile of SCC-63A cells harboring GFP-miR-125b or control sponge after antagomir treatment, analyzed by the FACS DIVA software. The GFP-miR-125b sponge construct also act as a miR-125b sensor. The SCC-63A cells harboring GFP-miR-125b sponge displayed much lower GFP signal then those with GFP-control sponge (middle panels), indicating that significant residue miR-125b activity were still present in these cells. Additional miR-125b antagomir treatment abolished most of the residue miR-125b activity, as indicated by significant up-regulation of GFP signal in SCC-63A cells harboring GFP-miR-125b sponge but not control sponge (right panels). (G) miRNA qPCR analysis of the above SCC-63A cells after modification by miR-125b knockdown (KD) (miR-125b sponge + miR-125bantagomir) or control KD(control Sponge + control antagomir). (H) In vitro MTT cell proliferation assay of SCC-63A cells after miR-125b KD or control KD. Error bars: standard error. n=8 independent wells of each data point. (I) Tumors generated fromSCC-63A cells miR-125bKD or control KD. (J) miRNA Q-PCR analysis of cultured normal human keratinocytes (HEK), HaCAT, A431 and Fadu cells. (K) miRNA Q-PCR analysis cultured A431 cells after miR-125b or scramble control antagomir transfection.

Supplemental Figure S5: (A) mRNA levels of P53 and P53 target genes in the RNA-seq data of regressing DTG tumors. (B) Vdr mRNA levels in the RNA-seq data of regressing DTG tumors. Vertical axis: FPKM (fragments per kilobase of exon per million fragments mapped), representing mRNA abundance. Horizonal axis: Days after taken Off-Dox, n as indicated. Error bars: standard error. (Right panel) GSEA analysis of genes carrying VDR transcriptional factor binding motif (MSigDB, M14782) in the RNA-seq data of regressing DTG tumors. Horizon axis: miR-125b correlation score (CS). Vertical axis: GSEA enrichment score. NES: normalized enrichment score. FDR: FDR (false discovery rate) adjusted P value of enrichment generated by the GSEA software. *: FDR<0.05. (C) Representative immunofluorescence picture of sections from engrafted H2BGFP-labeled DTG tumors after 3 days VDR ligand EB1089 or solvent control treatment. (D) DNA sequence of the Ha-Ras Codon 60-62 region of DTG CSCs and WT skin. (E) Two-color quantitative immunoblot analysis (each lane represents an independent well) of lysates from serum starved HACAT cells ±miR-125b overexpression and stimulated with 100ng/ml EGF ligand for the indicated duration. Numbers represent relative quantifications of pEGFR level. (F) Quantification of pEGFR (Y1068) and pERK1 (Y202) levels in Fadu cells ±miR-125b overexpression in attached (Atc) or suspension (Sup) culture, based on two-color quantitative immunoblot analysis. P values are based on unpaired two tail t test. n=3 each. (G) GSEA (Gene Set Enrichment Analysis) analysis of genes carrying miR-125b and let-7 seed sequences (MSigDB) in the RNA-seq data of regressing DTG tumors. Comparisons are against all the expressed genes in the RNA-seq data ranked by their miR-125b correlation score. Horizon axis: miR-125b correlation score (CS, see Materials and Methods). Vertical axis: GSEA enrichment score. NES: normalized enrichment score. FDR: FDR (false discovery rate) adjusted P value of enrichment generated by the GSEA software.**: FDR<0.01. ns: not significant. (H, I) Two-color quantitative immunoblot analysis of (G) cultured human SCC-13 cells or lysates from engrafted DTG tumors that were allowed to grow on Dox and then taken Off-Dox for days indicated. Numbers represent relative quantification of protein level in green channel. (J) Quantification of VPS4B protein level in cultured SCC-15, A431, Fadu cells ±miR-125b overexpression based on two-color quantitative immunoblot analysis. P Values are based on unpaired two tail t-test. N=3 each. (K) Targetscan predicted conserved miR-125b binding site on the 3’UTR of Vps4B mRNA.