Ras drives malignancy through stem cell crosstalk with the microenvironment – Nature

Animals

TRE-HRAS^G12V mice have been described previously⁴⁸. The original TRE-HRAS^G12V C57Bl/6 mice have been backcrossed 10 generations to an FVB/N background. FVB/N TRE-HRAS^G12V mice were bred to FVB/N R26-LSL-YFP mice to create the TGFβ-reporter lineage-tracing model. For the Tgfbr2-cKO experiment, FR-LSL-Hras^G12V;Tgfbr2^fl/fl;R26-LSL-YFP mice were crossed in-house. For tumour transplantation experiments, 7–9-week-old female NU/NU Nude mice from Charles River were used. All other studies used a mix of male and female mice, which for the assays used here, behaved similarly. The animals were maintained and bred under specific-pathogen-free conditions at the Comparative Bioscience Center (CBC) at The Rockefeller University, an Association for Assessment and Accreditation of Laboratory Animal Care (AALAC)—an accredited facility. Adult animals were housed in a cage with a maximum of five mice unless specific requirements were needed. The light cycle was from 07:00 to 19:00. The temperature of the animal rooms was 20–26 °C, and the humidity of the animal rooms was 30–70%. All mouse protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at The Rockefeller University.

As tumours began to progress to the malignant stage (size > 10 mm), mice were housed individually and antibiotic cream was applied to the surface of the ulcerated tumour. When tumour sizes approached 15 mm, intraperitoneal injection of Bup was used every 8 h to minimize pain. Mice were euthanized once the tumour size exceeded 20 mm or if mice showed any signs of distress, for example, difficulty in breathing.

Cell lines

The mouse cutaneous SCC cell line PDVC57 was cultured in the E-low medium (E.F.’s laboratory)⁵. Mouse keratinocyte cell line FF (Tgfbr2^f/fPGK-Hras^G12V) and ΔΔ (Tgfbr2^nullPGK-Hras^G12V) were cultured with the E-low medium as previously discribed⁶. The HNSCC cell line A431 was cultured in DMEM medium (Gibco) with 10% FCS, 100 U ml⁻¹ streptomycin and 100 mg ml⁻¹ penicillin. The HEK 293TN cell line for lentiviral production was cultured in DMEM medium supplemented with 10% FCS (Gibco), 1 mM sodium pyruvate, 2 mM glutamine, 100 U ml⁻¹ streptomycin and 100 mg ml⁻¹ penicillin. The 3T3J2 fibroblast feeder cell line was expanded in DMEM/F12 medium (Thermo Fisher Scientific) with 10% CFS (Gibco), 100 U ml⁻¹ streptomycin and 100 mg ml⁻¹ penicillin. It was then treated with 10 µg ml⁻¹ mitomycin C (Sigma-Aldrich) for 2 h to achieve growth inhibition.

The human skin SCC line A431 was from ATCC; mouse skin SCC PDVC57 was a gift from the original laboratory that created it (Balmain lab); mouse keratinocyte cell lines FF (Tgfbr2^f/fPGK-Hras^G12V) and ΔΔ (Tgfbr2^nullPGK-Hras^G12V) were generated in E.F.’s laboratory; mouse fibroblast 3T3/J2 has been passaged in the laboratory as feeder cells and originated from the laboratory of H. Green;  HEK 293TN cells were purchased from SBI directly as low passage (P2) for lentiviral packaging. PDVC57 was validated by karyotyping and grafting tests. Mouse keratinocyte cell lines were validated previously in E.F.’s laboratory. 3T3/J2 has been functionally and morphologically validated as feeder cells. HEK 293TN cells were functionally tested as packaging cells producing lentivirus. A431 was not authenticated.

Human tumour samples

Human skin and SCC tumour samples were acquired as frozen tissue from B. Singh at Weill Cornell Medical College. All of the samples were de-identified according to National Institutes of Health and Federal/State regulations. Informed consent was obtained from all human research participants at Weill Cornell Medical College, and in accordance with approved Institutional Review Broad (IRB) protocols from The Rockefeller University, Weill Cornell Medical College and Memorial Sloan Kettering Cancer Center.

Lentiviral in utero transduction

Lentiviral constructs were previously described (SBE-NLSmCherry-P2A-CreERT2 PGK-rtTA3)⁶ or cloned in E.F.’s laboratory (SBE-NLSmCherry PGK-rtTA3, Lepr peak reporter-eGFP PGK-rtTA3, TRE-Lepr-IRES-eGFP, PGK-rtTA3, TRE-Vegfa EEF1A1-rtTA3, TRE-STOP EEF1A1-rtTA3). The lentiviral production and in utero injection were performed as previously described^6,23. In brief, pregnant female mice with a doxycycline-inducible HRAS^G12V transgene were anaesthetized with isoflurane (Hospira) when their embryos were at E9.5. Lentivirus (500 nl to 1 µl) was injected into the amniotic sacs of the embryos to selectively transduce a small number of individual epidermal progenitors within the surface monolayer that gives rise to the skin epithelium⁴⁹. Postnatal induction of tumorigenesis in clonal patches was achieved by doxycycline administration (2 mg per g) through the feed.

Tumour formation and grafting

To induce spontaneous tumour formation, transduced TRE-HRAS^G12V or TRE-HRAS^G12VR26-LSL-YFP mice were continuously fed doxycycline-containing chow (2 mg per g) from postnatal day 0 to 4 to activate the rtTA3 transcription factor and induce tumorigenesis. Papillomas appeared by around 4 weeks and progressed to SCCs by about 8 weeks. To activate the creER^T2 in lineage-tracing experiments, 100 μg tamoxifen (Sigma-Aldrich) was injected intraperitoneally into tumour-bearing mice daily for 3 consecutive days. For tumour allograft studies, 1 × 10⁵ mouse PDVC57 SCC cells were mixed with growth-factor-reduced Matrigel (Corning) and intradermally injected into NU/NU Nude immunocompromised mice. Visible tumours appeared after 3 weeks. For metastatic tumour xenografts, 1 × 10⁵ human SCC A431 cells were resuspended in sterile PBS and tail-vein injected into immunocompromised Nude mice. Mouse lung tissue with metastatic lesions was collected after 3 weeks. The volume of the tumour was calculated using the following formula: \(\frac{4}{3}\pi \left(\frac{x}{2}\times \frac{y}{2}\times \frac{z}{2}\right)\), where x, y and z are three-dimensional diameters measured using digital callipers (FST).

Immunofluorescence and histology

For both histology and immunofluorescence analysis, tumour tissues were fixed in 4% PFA at room temperature for 15 min, and then washed three times with PBS at 4 °C. For histology, samples were dehydrated in 70% ethanol overnight, and were sent to Histowiz for Oil Red O and H&E staining. For immunofluorescence, after PBS washes, the samples were dehydrated in 30% sucrose in PBS solution overnight at 4 °C. The dehydrated tissues were embedded in OCT medium (VWR). Cryosections (10 µm) were blocked in PBS blocking buffer with 0.3% Triton X-100, 2.5% normal donkey serum, 1% BSA, 1% gelatin. After blocking, the sections were stained with primary antibodies: ITGA6 (rat, 1:2,000, BD), RFP/mCherry (guinea pig, 1:5,000, E.F.’s laboratory), K14 (chicken, 1:1,000, BioLegend), CD31 (rat, 1:100, BD Biosciences), K5 (guinea pig, 1:2,000, E.F.’s laboratory), K8 (rabbit, 1:1,000, E.F.’s laboratory), mLEPR (goat, 1:200, R&D Systems), hLEPR (rabbit, 1:100, Sigma-Aldrich), RUNX1 (rabbit, 1:100, Abcam), FOS (rabbit, 1:100, Cell Signalling), GFP (chicken, 1:500, BioLegend), pSTAT3-Y705 (rabbit, 1:100, Cell Signalling), pSMAS2-S465/467 (rabbit, 1:1,000, Cell Signalling) or pS6-S240/244 (rabbit, 1:100, Cell Signalling). For pSTAT3 immunolabelling, sections were pretreated with ice-cold 100% methanol for 30 min before blocking. After primary antibody staining, all sections were washed three times with PBS wash buffer containing 0.1% Triton X-100 for 5 min at room temperature. For pSMAD2 immunolabelling, sections were pretreated with 3% H₂O₂ for 1 h before blocking, stained using the appropriate HRP-conjugated secondary antibody (Jackson ImmunoResearch) and amplified using the TSA plus Cy3 kit (Akoya Biosciences) in combination with other regular co-stains. The sections were then labelled with the appropriate Alexa 488-, 546- and 647-conjugated secondary antibodies (Thermo Fisher Scientific) and imaged using the Zeiss Axio Observer Z1 with Apotome 2 microscope. Images were collected and analysed using Zeiss Zen software.

For immunofluorescence microscopy of thick tumour sections, all collected tumours were fixed with 1% (v/v) paraformaldehyde/PBS overnight at 4 °C and washed three times with PBS. After an overnight incubation with 30% (w/v) sucrose/PBS at 4 °C and embedding in OCT, 100 μm cryosections were washed with PBS and transferred to a 24-well dish. After overnight permeabilization with 0.3% Triton X-100/PBS at room temperature with rotation, tissue was blocked for 4–6 h with 5% donkey serum and 1% bovine serum albumin in 0.3% Triton X-100/PBS (blocking buffer). Tissue was then incubated with the following primary antibodies for 2 days at room temperature: mCherry (Abcam, 1:1,000), CD31 (Sigma-Aldrich, 1:300), keratin 14 (E.F.’s laboratory, 1:400), Keratin 18 (rabbit, 1:300, E.F.’s laboratory), GFP (chicken, 1:300, E.F.’s laboratory) and ITGA6 (Rat, 1:300, BD Biosciences) before several washes with 0.3% Triton X-100/PBS. The tissue sections were incubated with secondary antibodies (Alexa Fluor-RRX, -488 or -647 hamster, rat, chicken and rabbit at 1:1,000) diluted in blocking buffer overnight (16–20 h) together with DAPI at room temperature and washed with 0.3% Triton X-100/PBS with several exchanges. Immunolabelled tissue sections were then dehydrated with a graded ethanol series by incubation in 30% ethanol, 50% ethanol and 70% ethanol, each set to pH 9.0 as described previously⁵⁰ for 1 h per solution, before a 2 h incubation with 100% ethanol, and cleared to optimize optical sectioning and imaging penetration by overnight incubation with ethyl cinnamate (Sigma-Aldrich). Cleared tumour samples were imaged in 35 mm glass-bottom dishes (Ibidi) with an inverted LSM Zeiss 780 laser-scanning confocal microscope and/or Andor dragonfly spinning disk. Images were then analysed using Imaris imaging software (Bitplane). The shortest distance and volume measurements were performed by the creation of individual objects of CD31⁺ blood vessels, K14⁺ tumour mass, K18⁺ tumour cells or Lepr reporter⁺ tumour cells.

Cell sorting and flow cytometry

To sort the target tumour cell populations by FACS, tumours were first dissected from the skin and finely minced in 0.25% collagenase (Sigma-Aldrich) in HBSS (Gibco) solution. The tissue pieces were incubated at 37 °C for 20 min with rotation. After a wash with ice-cold PBS, the samples were further digested into a single-cell suspension in 10 ml 0.25% trypsin/EDTA (Gibco) for 10 min at 37 °C. The trypsin was then quenched with 10 ml FACS buffer (5% FCS, 10 mM EDTA, 1 mM HEPES in PBS). The single-cell suspension was centrifuged at 700 rcf. The pellet was resuspended in 20 ml FACS buffer and strained through a 70 μm cell strainer (BD Biosciences). The filtered samples were centrifuged at 700 rcf to pellet cells, and the supernatant was discarded. The cell pellet was then resuspended in primary antibodies. A cocktail of antibodies against surface markers at the predetermined concentrations (CD31–APC, 1:100, BioLegend; CD45–APC, 1:200, BioLegend, CD117–APC, 1:100, BioLegend; CD140a–APC, 1:100, Thermo Fisher Scientific; CD29–APCe780, 1:250, Thermo Fisher Scientific; CD49f–PerCPCy5.5, 1:250, BioLegend; CD44–PECy7, 1:100, BD Biosciences) was prepared in the FACS buffer with 100 ng ml⁻¹ DAPI. Furthermore, CD44–BV421 (1:100, BD Biosciences), CD49f–PECy7 (1:250, BioLegend) and CD29–APCCy7 (1:250, BioLegend) were also used as interchangeable staining in the panels for the same purpose. The samples were incubated on ice for 30 min, washed with FACS buffer twice and resuspended in FACS buffer with 100 ng ml⁻¹ DAPI before FACS and analysis.

To sort the skin stem cell populations (IFE and HFSCs), whole back skins were first dissected from the mouse. After scraping off the fat tissues from the dermal side, the tissues were incubated in 0.25% trypsin/EDTA (Gibco) for 45–60 min at 37 °C. After quenching the trypsin with cold FACS buffer, the epidermal layer and hair follicles were scraped off the epidermal side of the skin. The tissues were mechanically separated/strained into a single-cell suspension for staining. A cocktail of antibodies for surface markers at the predetermined concentrations (CD31–APC, 1:100, BioLegend; CD45–APC, 1:200, BioLegend; CD117–APC, 1:100, BioLegend; CD140a–APC, 1:100, Thermo Fisher Scientific; CD29–APCe780, 1:250, Thermo Fisher Scientific; CD49f–PerCPCy5.5, 1:250, BioLegend; CD34–BV421, 1:100, BD Biosciences; CD200–PE, 1:100, BioLegend; SCA1–PECy7, 1:100, BioLegend) was prepared in the FACS buffer with 20 ng ml⁻¹ DAPI when using an ultraviolet laser. The sorting was performed on BD FACS Aria equipped with FACSDiva software.

For the in vivo Lepr reporter SCC cell experiment, reporter PDVC57 cells were treated with TGFβ1 (10 ng ml⁻¹) for 7 days. The treated reporter PDVC57 cells were stained with 100 ng ml⁻¹ DAPI in FACS buffer and analysed on the BD Biosciences LSR Fortessa system together with the control treatment (BSA only).

For the phosphorylated protein flow cytometry experiment, single-cell suspensions were obtained from papilloma or SCC tumour tissues as described above. After washes with cold PBS, cells were stained with Live/Dead Blue (1:200, Thermo Fisher Scientific) on ice for 30 min, and then washed with FACS buffer and blocked with FACS buffer with 5% normal mouse serum (Thermo Fisher Scientific), 5% normal rat serum (Thermo Fisher Scientific) and 1× Fc Block (BioLegend) for 15 min on ice. The live cells were then stained with FITC-conjugated surface marker (DUMP) antibodies (CD31, CD45, CD117, CD140a) (Thermo Fisher Scientific and BioLegend) for 30 min. After washing with FACS buffer, cells were fixed with 1× Phosflow Lyse/Fix Buffer (BD Biosciences) at 37 °C for 10 min. After centrifuging and another wash with FACS buffer, the fixed cells were permeabilized with −20 °C prechilled Phosflow Perm Buffer III (BD Biosciences) for 30 min on ice. The samples were then washed twice with 1× Phosflow Perm/Wahs Buffer I (BD Biosciences), and stained with CD49f–BV510 (BD Biosciences), CD29–APCe780 (Thermo Fisher Scientific), CD44–BB700 (BD Biosciences), pAKTs473–BV421 (BD Biosciences) and pJAK2y1007/1008–Alexa647 (Abcam) antibodies for 2 h on ice. The samples were then washed and analysed on the BD Biosciences LSR Fortessa system. The flow cytometry data were analysed using FlowJo (BD Biosciences).

RNA purification and ATAC-seq library preparation

For bulk RNA-seq, targeted cell populations from 2 (SCC) to 15 (papilloma) tumours per population were directly sorted into TRI Reagent (Thermo Fisher Scientific) and the total RNA was purified using the Direct-zol RNA MiniPrep Kit (Zymo Research) according to the manufacturer’s instructions. The integrity of purified RNA was determined using the Agilent 2100 Bioanalyzer. Library preparation, using the Illumina TrueSeq mRNA sample preparation kit (non-stranded, poly(A) selection), and sequencing were performed at the Genomic Core Facility at Weill Cornell Medical College on the Illumina HiSeq 4000 system with the 50 bp single-end setting or the NovaSeq with the 100 bp paired-end setting.

For accessible chromatin profiling, target cell populations from 2 (SCC) to 15 (papilloma) tumours per population were sorted into FACS buffer, and ATAC-seq sample preparation was performed as described previously⁵¹. In brief, a minimum of 2 × 10⁴ cells were lysed with ATAC lysis buffer on ice for 1 min. Lysed cells were then tagmented with Tn5 transposase (Illumina) at 37 °C for 30 min. Cleaned-up fragments were PCR-amplified (NEB) and size-selected with 1.8× SPRI beads (Beckman Coulter). Libraries were sequenced at the Genome Resource Center at The Rockefeller University on the Illumina NextSeq system with the 40 bp paired-end setting.

For scRNA-seq, target cell populations were sorted from 3–5 SCC tumours per mouse, for a total of 3 biological replicates (2 male and 1 female mice). Single-cell libraries were prepared according to a slightly modified Smart-seq2 protocol⁵². In brief, cells were sorted into 96-well plates containing hypotonic lysis buffer, snap-frozen with liquid nitrogen and stored at −80 °C until further processing. To semi-quantitatively assess technical variation between cells, ERCC spike-ins (1:2 × 10⁶ dilution, Thermo Fisher Scientific) were added with the lysis buffer. After thawing, cells were lysed at 72 °C for 3 min. Released RNA was reverse-transcribed using dT30 oligos, template switching oligos and Maxima H- reverse transcriptase. cDNA was amplified by 15 cycles of whole-transcriptome amplification using KAPA HiFi DNA polymerase (Roche) and then size-selected using 0.6× AmpPure XP beads (Beckman Coulter). To exclude wells containing multiple cells, as well as low-quality and empty wells, quantitative PCR with reverse transcription (RT–qPCR) for Gapdh was performed before proceeding. Illumina sequencing libraries were then prepared using the Nextera XT DNA library preparation kit (Illumina) and indexed with unique 5′ and 3′ barcode combinations. After barcoding, the samples were pooled and size-selected with 0.9× AmpPure XP beads. The integrity of the pooled library was assessed using the TapeStation (Agilent) before sequencing on two lanes of the Illumina NovaSeq S1 system using 100 bp paired-end read output (Illumina). For optimal sequencing depth, each sequencing library was sequenced twice, once in each lane of the Illumina NovaSeq system. Sequencing reads per cell from each lane were combined during alignment to the reference genome.

CRISPR-mediated Lepr knockout

Our Lepr^null PDVC57 cell line was generated using the Alt-R CRISPR–Cas9 system (IDT). In brief, a recombinant Cas9 protein, validated sgRNA (GAGUCAUCGGUUGUGUUCGG) targeting exon 3 of the mouse Lepr gene or a negative control sgRNA (IDT), and an ATTO-550-conjugated tracer RNA were used to form a ribonucleoprotein (RNP) were mixed with RNAiMax reagent (Thermo Fisher Scientific). PDVC57 cells were then transfected with the mixture overnight and FACS-purified into 96-well plates to produce clonal cell lines. The Lepr^null PDVC57 cell line was selected after validating by immunoblot analysis of LEPR as well as sequencing of the target region for indel efficiency using the MiSeq system. The Lepr^null PDVC57 cell line and Lepr^ctrl PDVC57 cell line were intradermally injected into the immunocompromised Nude mice, and the tumours were analysed for growth and progression.

Rescue of LEPR in Lepr
^null PDVC57 SCC cells

Lepr^null PDVC57 cells were transduced in vitro with 1:1 ratio of PGK-rtTA3 lentivirus and TRE-FL-Lepr-IRES-eGFP or TRE-Lepr^ΔSig-IRES-eGFP lentivirus. After culturing in 1 μg ml⁻¹ of doxycycline (Sigma-Aldrich) containing E-Low medium, eGFP^high cells expressing Lepr were isolated by FACS and expanded in vitro. These two different cell lines were later intradermally grafted onto immunocompromised Nude mice, and the tumours were analysed for growth and progression.

Limiting dilution assay

To compare the tumour-initiating ability between Lepr^null PDVC57 and Lepr^ctrl PDVC57 cell lines, a preset number of cells were intradermally grafted onto Nude mice, and the tumour growth was tracked for 5 weeks to calculate the tumorigenicity of cells. As previously described²³, SCC cells were diluted serially from 10⁴ to 10⁶ cells per ml and 100 µl cell mixtures in 1:1 PBS:Matrigel were injected. Four injections per mouse were performed under the animal facility regulations (for 10⁵ and 10⁴ per injection, n = 4; for 10³ per injection, n = 8). Photos of mice were recorded, and tumours were counted at the end point 5 weeks after injection.

Osmotic pump for systemic delivery

To achieve continuous systemic delivery of compounds, Alzet osmotic pumps were implanted as previously described⁵³ into the back skins of Nude mice. Three weeks after the initial intradermal tumour grafts, tumour-bearing Nude mice were anaesthetized and sterilized for surgical procedures. A small cut was created with scissors and the osmotic pump containing a predetermined concentration of compounds or vehicle was inserted underneath the back skin and the opening was clipped. For the leptin experiment, 4-week-long delivery pumps were used with 2 mg ml⁻¹ leptin (R&D Systems), 0.5 mg ml⁻¹ leptin, 0.5 mg ml⁻¹ SMLA (BioSources) and PBS vehicle. Where indicated, fluorescently labelled (680RD) leptin as previously discribed⁵⁴ was used to detect the ability of circulating leptin to reach the skin stroma. For the VEGFA experiment, 4-week-long delivery pumps were used with 50 μg ml⁻¹ VEGFA (R&D Systems) and PBS vehicle. For the rapamycin experiment, 2-week-long delivery pumps were used with 10 mM rapamycin (SelleckChem) in PBS solution with 10% DMSO or with the respective vehicle control. Tumour sizes were then monitored for tumour growth and progression.

To achieve local delivery of compound, intradermal injections were performed into the skin adjacent to or underneath the grafted tumours of Nude mice. 50 μg ml⁻¹ VEGFA (R&D Systems) and PBS vehicle were injected in a 50 μl volume every 3 days with a 1 ml syringe and a 26G needle (BD Biosciences).

RT–qPCR

To measure Lep mRNA levels in whole tissues, normal skin from wild-type FVB/N mice was first separated from the underlying white fat tissue, and then both were independently snap-frozen in liquid nitrogen. Papilloma and SCC tumours were trimmed from the adjacent aphenotypic skin. Frozen tissues were crushed and then dissolved in Tri-Reagent (Thermo Fisher Scientific). RNA was isolated using the Direct-zol RNA miniprep kit (Zymo Research). To measure Lep levels in specific cells from the tumour or normal microenvironment, CD45⁺ (immune cells), CD140a⁺ (fibroblasts and other mesenchymal cells), CD117⁺ (melanocytes) and CD31⁺ (endothelial cells) were FACS-isolated from single-cell suspensions of normal skin, papilloma and SCC in Tri-Reagent (Thermo Fisher Scientific). RNA was isolated using the Direct-zol RNA microprep kit (Zymo Research). Equivalent amounts of RNA were reverse-transcribed using the SuperScript VILO cDNA Synthesis Kit (Thermo Fisher Scientific). cDNAs were mixed with the primers listed below and the Power SYBR Green PCR Master Mix (Thermo Fisher Scientific), and then quantified using the Applied Biosystems QuantStudio 6 Real-Time PCR system. Lep levels were normalized to equal amounts using primers against B2m. Primer sequences were as follows: Lep forward, 5′-GAGACCCCTGTGTCGGTTC-3′; Lep reverse, 5′- CTGCGTGTGTGAAATGTCATTG-3′; B2m forward, 5′- TTCTGGTGCTTGTCTCACTGA-3′; B2m reverse, 5′-CAGTATGTTCGGCTTCCCATTC-3′.

PI3K inhibitor gavage in tumour-bearing mice

As previously described³³, the pan-PI3K inhibitor BKM120 (MedChemExpress) was dissolved in DMSO to 100 mg ml⁻¹. A 10% (v/v) solution was then sequentially diluted in 40% PEG300, 5% Tween-80 and 45% PBS. DMSO (10%) was used as vehicle control. The course of treatment was daily gavage for 14 days. Tumour-bearing mice were first anaesthetized lightly and 100 µl of the solution was delivered to the mouse stomach through a feeding needle (Thermo Fisher Scientific). The study was blinded by one experimentalist performing gavage and the other one measuring the tumour sizes every 2–3 days without knowing the treatment or control. The results were analysed at day 15 after the initial treatment.

Colony-forming assay

After LPER⁺ and LEPR⁻ tumour basal cells (CD29/CD49f^highCD44⁺) were FACS isolated and counted, 5 × 10⁴ cells of each replicate per condition were plated in a 10 cm dish with a growth-inhibited 3T3/J2 feeder layer with the SY medium (E.F.’s laboratory, see below) at 7.5% CO₂ and 37 °C. After 14 days, the cultures were fixed and stained with Alexa647-conjugated CD49f antibodies (BioLegend). The plates were then imaged using the LiCor Odyssey Imager and quantified on the basis of the numbers and sizes of colonies.

SY mouse skin stem cell culture medium

The base medium was made with calcium-free DMEM/F12 (3:1) (E.F.’s laboratory) with 1× Glutamax (Thermo Fisher Scientific) and 1× penicillin–streptomycin (Thermo Fisher Scientific). Additives included 15% chelated fetal bovine serum (Thermo Fisher Scientific), 418.5 ng ml⁻¹ of hydrocortisone (Sigma-Aldrich), 9.405 ng ml⁻¹ of cholera toxin (Sigma-Aldrich), 10 μM of Y-27632 (Selleck Chemicals), 0.0525 mg ml⁻¹ insulin (Sigma-Aldrich), 0.0525 mg ml⁻¹ Apo-transferrin (Thermo Fisher Scientific), 300 mM CaCl₂ (Sigma-Aldrich), 36.5 mM of NaHCO₃ (Sigma-Aldrich) and 2.1 × 10⁻⁸ M of 3,3′,5-triiodo-l-thyronine (Sigma-Aldrich).

Mouse leptin ELISA

For quantification of leptin level in the tissue or plasma, the Quantikine Mouse/Rat Leptin ELISA (R&D Systems) kits were used. Tumour tissues were snap-frozen without adjacent skin in liquid nitrogen and sonicated in lysis buffer (R&D Systems) before centrifuging at full speed at 4 °C for 10 min to obtain total lysates. Plasma was obtained by centrifuging clean blood for 15 min at 2,000g at 4 °C. The manufacture’s protocol was followed for these assays.

Fluorescence assay for detecting labelled proteins

For quantification of 680RD-labelled leptin level in tissue, the Biotek Cytation 5 System (BioTek) was used. Tumour and skin tissues were snap-frozen in liquid nitrogen and sonicated in Lysis Buffer (R&D Systems) before centrifuging at full speed at 4 °C for 10 min to obtain total lysates. Serial dilutions of labelled recombinant protein were used as standards to generate a curve to estimate the amount of protein in the tissue lysates. A total of 100 μl of lysates and standards in duplicates were loaded into 96-well black assay plate (Thermo Fisher Scientific) and then read at an excitation of 680 nm and emission of 695 nm. Estimated concentrations were then calculated.

Immunoblotting

To collect cells, cultured cells were washed on the plate in cold 1× PBS, lysed in RIPA Buffer (Millipore) supplemented with protease and phosphatase inhibitors (Roche), and collected by scraping. Cells were lysed for 30 min on ice and then centrifuged to collect the supernatant. The protein concentration was determined by the BCA assay (Pierce) against a bovine serum albumin standard curve. Protein (20 μg) of each sample was run on NuPAGE 4–12% Bis-Tris Gels (Invitrogen) for 1 h at 200 V in NuPAGE MES SDS Running Buffer (Invitrogen). Protein was transferred overnight onto the Immunoblon FL PVDF membrane (Millipore) in NuPAGE transfer buffer (Invitrogen) with methanol at 15 V and 4 °C. Membranes were blocked in Odyssey TBS blocking buffer for at least 1 h at room temperature before incubating with primary antibodies overnight at 4 °C in Odyssey buffer with Tween-20. Membranes were washed several times in 0.1% Tween-20 in PBS before incubating with fluorescent secondary antibody.

The following primary antibodies and dilutions were used: primary antibodies (anti-mLEPR 1:1,000, R&D Systems; anti-AKT, 1:1,000, Cell Signaling; anti-pAKT(S473), 1:1,000, Cell Signalling; anti-S6, 1:1,000, Cell Signaling; anti-pS6(S240/244), 1:1,000, Cell Signaling; anti-S6K, 1:1,000, R&D Systems; anti-pS6K(T389), 1:1,000, Cell Signaling; anti-GAPDH, 1:5,000, Thermo Fisher Scientific; anti-α-tubulin, 1:5,000, Sigma-Aldrich), secondary antibodies were used at 1:10,000 (donkey anti-rabbit HRP and donkey anti-mouse Alexa647, Jackson ImmunoResearch). Membranes were imaged with an GE Amsham AI600 Imager. Owing to multiple targeted proteins in each experiment, one set of identical samples with the same sample volumes and processing procedure was blotted for GAPDH or α-tubulin in one of the gels in the same experiment as a loading control.

Bulk RNA-seq analysis

Trimmed fastq files were obtained from the Genomic Core Facility of Weill Cornell Medical College. The analysis was performed by the cluster at the High-Performance Computing facility. For RNA-seq analysis of C57BL/6J TRE-HRAS^G12V driven papilloma and SCC samples (Fig. 1 and Extended Data Fig. 1), raw sequencing reads were aligned to the mouse reference genome (UCSC release mm10) using Bowtie2 (v.2.2.9)⁵⁵ using the default parameters. The expression values of each gene were quantified as transcripts per million (TPM), as well as raw counts, using RSEM (v.1.2.30)⁵⁶. Differential gene expression analysis was performed on raw counts using DESeq2 (v.1.24.0) with a negative binomial distribution and Wald test for significance⁵⁷. Genes with average counts of greater than 10, log₂[fold change] > |1| and adjusted P < 0.05 were considered to be differentially expressed. Differentially expressed genes were presented as a heat map with z-score-normalized expression values. To examine temporal changes in regulators of angiogenesis as cells transit from normal to benign to invasive states, the expressed genes related to the GO term ‘positive regulation of angiogenesis’ (GO:0045766, AmiGO2) were plotted as a z-score-normalized heat map.

For RNA-seq analysis of FVB TRE-HRAS^G12V-driven TGFβ-reporter papilloma and SCC samples, genome indices were generated with the genome sequence (GRCm38.p5) and the comprehensive gene annotation on the primary assembly (GENCODE M16). Raw reads were aligned to the genome indices and gene counts were generated using STAR (v.2.6)⁵⁸ with the default parameters. For differential gene expression analysis, low-expressed genes (minimum average read count < 10) were filtered out before DESeq2 analysis (v.1.16.1) in R Studio (v.3.4.2). Paired mCherry-positive and -negative samples were identified as batches and disease stages (papilloma versus SCC) as conditions for differential gene expression modelling using a negative binomial distribution and Wald test by DESeq2. Genes were considered to be differentially expressed when log₂[fold-change] > |1| and adjusted P < 0.05.

For PDV-WT and Lepr^KO grafted SCC samples, raw reads were mapped to the decoy-aware mouse genome (UCSC release mm10) using Salmon (v.1.4.0)⁵⁹. The expression level of each gene was quantified as TPM, as well as by raw counts, using Tximport (v.1.12.3)⁶⁰ in R (v.3.6.1). For differential gene expression analysis, low-expressed genes (minimum average read count < 10) were filtered out before DESeq2 analysis (v.1.16.1) in R Studio (v.3.4.2). Differential gene expression modelling used a negative binomial distribution and Wald test by DESeq2. Genes were considered to be differentially expressed for log₂[fold-change] > |1| and adjusted P < 0.05.

GO term and KEGG pathway analysis were performed using the DAVID online tool (NIH).

Bulk ATAC-seq analysis

Fastq files were obtained from the Genomic Resource Center of The Rockefeller University. The analysis was performed on the computational cluster at the High-Performance Computing facility. The raw sequencing reads were aligned using Bowtie2 (v.2.2.9)⁵⁵ to the mm10 reference genome (UCSC). The aligned reads were de-duplicated with Picard (v.2.3.0; Broad Institute, 2019) and shifted to correct for Tn5 insertion bias. Peaks were called using MACS2 (v.2.1.1) with the default settings⁶¹. Next, all peaks from IFE, HFSC, PAPneg, PAPpos, SCCneg and SCCpos were concatenated to a union peak set, and the read coverage of each sample at these peaks was calculated with Bedtools (v.2.25). The R (v.3.6.1) package pheatmap (v.1.0.12) was then used to generate the heat map. For motif analysis, HOMER findMotifGenome.pl was used with a customized motif database from JASPAR 2018. The motif input for HOMER (v.4.11)⁶² was generated from JASPAR 2018 vertebrates CORE central motifs using 80% of the maximum log-odds expectation for each motif as the detection threshold for HOMER.

Single-cell RNA-seq analysis

Sequence and transcript coordinates for the mouse release M23 (GRCm38.p6) genome and gene models were downloaded from GENCODE. Paired sequencing reads for scRNA-seq libraries were aligned to the mouse reference genome, combined with sequences for ERCC spike-ins as artificial chromosomes, using STAR (v.2.5.2a)⁵⁸ with the default parameters for paired-end reads. Transcript expression values were calculated using the Salmon quantification software (v.0.14.1)⁵⁹ and gene expression levels as TPMs and counts were obtained using Tximport (v.1.12.3)⁶⁰. TPMs were transformed to log₂[TPM + 1]. For downstream analyses, cells with <2,500 genes detected per cell and genes expressed in <5% of the cell population were removed. Cells expressing lower levels of the lineage marker Krt14 (log₂[TPM + 1] < 7) were excluded. After filtering, there were 1,504 cells (159 integrin^low suprabasal, 500 integrin^high, mCherry⁻ basal, and 845 integrin^high, mCherry⁺ basal cells) (n = 3 mice) in the dataset.

Analyses and visualization of data were conducted in a Python environment built on the Numpy, SciPy, matplotlib, scikit-learn package and pandas libraries. To distinguish true biological variability in gene expression from technical noise, we used a statistical model for identifying highly variable genes compared to ERCC spike-ins as described⁶³. In brief, we used a custom script based on the methodology described⁶³, running in R v.3.6.1, to identify those genes with a higher level of variation (at least 10% above the technical variation) and a false-discovery rate (FDR) value of less than 0.1. To identify cell clusters and visualize the data, we first centred and scaled the highly variable gene dataset and performed principal component analysis on the list of highly variable genes. The first 201 principal components, which captured 50% of the variance in the dataset, were used as an input for nonlinear dimensionality reduction, performed using UMAP implemented in scikit-learn. To identify clusters, we used a graph-based clustering approach based on building a k-nearest neighbours graph and clustering with the Louvain algorithm (with k set to one fifth of the dataset size, and a resolution parameter of 1 × 10⁻⁴). Euclidean distance in PCA space served as input for both UMAP generation and Louvain clustering.

Differential gene expression was used to identify genes specific to each cluster. In brief, we used raw count matrices for expressed genes and applied them to the DESeq2 package (v.1.24.0)⁵⁷ using R. We used a negative binomial fit to model differential gene expression, factored the dataset based on the Louvain cluster assignments, and used a threshold of 0.75 to construct Wald tests of significance. Genes were considered to be differentially expressed if log₂[fold change] > |1| and adjusted P < 0.05. Low-expressed differential genes (baseMean expression < 200) were discarded from visualization and further analysis. The expression levels of specific genes of interest were visualized as log₂[TPM + 1] values on the corresponding UMAP representation of the data. GO term and KEGG pathway analyses were performed using the DAVID online tool (NIH).

To generate comprehensive gene set scores based on GO term analyses, for example, for angiogenesis or AKT signalling pathways, the corresponding Mus musculus gene lists were obtained from AmiGO 2 through the Gene Ontology consortium. The AddModuleScore function of Seurat (v.3.1.1) was used to calculate the average expression levels of each gene set at the single-cell level, subtracted by the aggregated expression of control feature sets, as originally described⁶⁴. The resulting gene set scores for each cell were colour coded on corresponding UMAP visualizations of the data.

Statistics and reproducibility

For the mouse experiments, group sizes were determined by power analysis using preliminary experimental data. All of the experimental measurements were taken from independent distinct samples. Unless stated otherwise, statistical analysis was performed using unpaired two-tailed Student’s t-tests with a 95% confidence interval under the untested assumption of normality on the GraphPad Prism (v.9.0). All of the error bars in the box plots and growth curves are mean ± s.e.m. The degree of statistical significance is denoted by asterisks; NS, P ≥ 0.05; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Whenever representative plots or images are shown, the datasets with similar results were generally generated from more than one litter of mice and with n ≥ 3 independent biological replicates, due to the nature of the tumour staging being histopathology driven rather than age or size driven. All attempts at replication in this study were successful. In general, the experiments were not randomized or performed by the investigator in a blinded manner, except where stated.

Schematics

Schematics were prepared using Office 365 (Microsoft), BioRender (www.biorender.com) with publication permissions and Affinity Design (Serif Europe).

Reporting summary

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