Supporting Information

Ischenko et al. 10.1073/pnas.1219592110 SI Materials and Methods and soft agar colony formation assays were performed as de- We used previously described LSL KrasG12D mice (1) and LSL scribed previously (3). We used replication-defective retroviral p53R172H mice (2). MEFs were prepared from embryonic day vectors encoding constitutively active mutants KrasG12D, 13 to 14 embryos. Adult fibroblasts and pancreatic epithelial cells KrasG12V, HrasG12V, NrasG12D, EGFR XZ066, Mek2 were prepared from 4- to 6-wk-old mice. All fibroblast cells, KW71, Myr Akt1, Myr PDK1, Rac G12V, CDC42 Q61L, RalA unless otherwise specified, were grown in DMEM (Gibco) sup- Q75L, CDK2, CDK4, c-, L-Myc, Bmi1, FoxC2, Stat3, plemented with 5% FBS (Sigma) and 1× antibiotic–antimycotic. Twist1, Ezh2, and dominant-negative mutants c-MycΔMBII and Epithelial cells were grown on gelatinized plates in CnT-17 Myc W136E. Mission lentiviral vectors expressing shMyc were media (CellnTec). Cells (2 × 104) were injected s.c. into nude purchased from Sigma. mice at four sites on the back of the mouse. We defined tumor For expression microarrays, total cellular RNA was extracted by latencies as the period between injection of tumorigenic cells using the RNeasy Mini Kit (Qiagen). One-color hybridizations of into mice and the appearance of tumors of ≥1 mm in diameter. labeled cRNAs (two technical replicas) were performed against the The survival end point was a tumor diameter of 1 cm. For mouse MouseRef-8 Expression BeadChip 25K microarray (Illu- sphere-forming assays, single cell suspensions were cultured in mina) at Roswell Park Cancer Institute (Buffalo, NY). Microarray ultralow-attachment six-well plates (Corning) with CnT-17 me- signals were processed with GenomeStudio Expression dia at 2 × 104 to 105 cells per well. For flow cytometry, cells were Module (GSGX; Illumina), version 1.6.0. Differentially expressed lifted with Accutase (Sigma), stained with antibodies to Thy1.2, were defined based on greater than two- or fourfold change. Sca1, EpCam, CD104, CD49f, CD44, and CD24 (BD Pharmin- Data were background-corrected and quantile-normalized. A gen), and analyzed by using FACSCalibur (BD) with CellQuest small offset was applied to bring values above 0. The heat maps software. Standard protocols were followed for Western blot (Fig. S3) were generated by calculating ratios of expression in each analysis. We used antibodies specific for Kras (Ab1), Ras Asp12 sample vs. control. The log2 values were then supplied to the heat (Ab-1), Pan-Ras (F132-62; all from Millipore), Hras (6100001; map function of the R statistical package. Module expression BD), c-Myc (N-262; all from Santa Cruz), Max (4732), and analysis was conducted as described previously (4). Average gene MAPK (05-157; Upstate). The Ras activation assay was per- expression values (log2) of all genes were set as baseline 0. The formed by using an agarose-bound GST-fused Ras-binding do- gene expression values (log2) of each module relative to the main of Raf-1 (Cell Biolabs). Briefly, cells were plated at equal overall average were represented as mean ± SEM. Definition of density, grown in 2% (wt/vol) FBS (fibroblasts) or CnT-17 media each module is as follows: the Kras module comprises 220 targets (epithelial cells), washed with cold PBS solution, and lysed. The derived from preclinical and clinical datasets (5); the AP1 module activated Ras was pulled down with agarose-conjugated Raf-1 is composed of 124 targets of the Jun and Fos family (6– Ras-binding domain, followed by SDS/PAGE gel electrophoresis 8); the Myc module comprises 500 targets of Myc, Max, N-Myc, and immunoblotting with anti-Kras, Hras, and Pan-Ras anti- Dmap1, , , and Zfx (4); the PRC module is composed bodies. Real-time RT-PCR was performed by using whole-cell of 560 targets of PRC cluster proteins, Phc1, Rnf2, Eed, and RNAs prepared from primary cells or tumors. RNAs were pre- Suz12 (4). Statistical analyses were performed by using a Student pared by using the TRIzol reagent (Invitrogen). Focus formation t test. P ≤ 0.05 was considered statistically significant.

1. Tuveson DA, et al. (2004) Endogenous oncogenic K-ras(G12D) stimulates proliferation 5. Loboda A, et al. (2010) A gene expression signature of RAS pathway dependence and widespread neoplastic and developmental defects. Cancer Cell 5(4):375–387. predicts response to PI3K and RAS pathway inhibitors and expands the population of 2. Olive KP, et al. (2004) Mutant gain of function in two mouse models of Li- RAS pathway activated tumors. BMC Med Genomics 3:26. Fraumeni syndrome. Cell 119(6):847–860. 6. Ozanne BW, et al. (2000) Transcriptional regulation of cell invasion: AP-1 regulation of 3. Petrenko O, Fingerle-Rowson G, Peng T, Mitchell RA, Metz CN (2003) Macrophage a multigenic invasion programme. Eur J Cancer 36(13 Spec No):1640–1648. migration inhibitory factor deficiency is associated with altered cell growth and 7. Leaner VD, Kinoshita I, Birrer MJ (2003) AP-1 complexes containing cJun and JunB reduced susceptibility to Ras-mediated transformation. J Biol Chem 278(13): cause cellular transformation of Rat1a fibroblasts and share transcriptional targets. 11078–11085. Oncogene 22(36):5619–5629. 4. Kim J, et al. (2010) A Myc network accounts for similarities between embryonic stem 8. Florin L, et al. (2004) Identification of novel AP-1 target genes in fibroblasts regulated and cancer cell transcription programs. Cell 143(2):313–324. during cutaneous wound healing. Oncogene 23(42):7005–7017.

Ischenko et al. www.pnas.org/cgi/content/short/1219592110 1of7 A Cre WT LSL KrasD12 WT KrasD12 25 p53-/- LSL KrasD12 20 p53-/- KrasD12 15 10 Adi Zaltsman 5

Fold cell increase 0 0 5 10 15 20 25 Passages B

LSL KrasG12D p53KO MEFs p4 KrasG12D p53KO MEFs p4 C Cl11 Cl17 Cl15 Cl1 Cl6 Cl8 Control Cl7 Cl2 Cl3 Control Cl22 Cl21 Cl11 Cl12

KrasG12D

Total Kras

MAPK

Fig. S1. (A) Growth curves of WT LSL KrasG12D (inactive allele), KrasG12D (activated allele), LSL KrasG12D p53−/− (p53KO; inactive allele), and KrasG12D p53KO (active allele) MEFs. (B) Morphological changes in LSL KrasG12D p53KO MEFs (Left) caused by activation of the mutant Kras allele (Right). (C) Immunoblot analysis of KrasG12D and total Kras expression in control LSL KrasG12D p53KO and clonal KrasG12D p53KO fibroblasts.

Ischenko et al. www.pnas.org/cgi/content/short/1219592110 2of7 D E

F

G H

Fig. S2. (A–C) To assess whether KrasG12D-expressing cells need to undergo many rounds of DNA replication to acquire a malignancy-associated phenotype, we compared KrasG12D p53KO MEFs that were grown for 12 passages under various serum conditions (0.1–10% FBS). Cells maintained in low serum underwent fewer population doublings compared with cells cultured in high serum (A). However, this growth disadvantage had no effect on the capacity of cells to form transformed foci (B) or tumors in nude mice (C), thus indicating that KrasG12D-mediated transformation is relatively independent of cell division rate. (D–F) We next used the Luria–Delbrück fluctuation method, which estimates the mutation rate of dividing cells (D). The premise of the method is that, among a large set of identical cultures, the number of transformed colonies arising in each culture will follow the Poisson distribution. However, the number of colonies per culture will vary greatly if random mutations conferred a strong proliferative and/or survival advantage. We monitored 10 parallel cultures of KrasG12D MEFs over a 2-mo time period (E)toconfirm that new malignant cells develop from premalignant cells via a Poisson process (with variance equal to the mean) and not the Luria–Delbrück process (with variance much larger than the mean) (F). (G and H) To assess the role of added genomic instability, we cultured LSL KrasG12D p53KO MEFs for 2, 10, or 20 passages (6, 30, and 60 d, respectively) to render them increasingly aneuploid, before induction of Cre-mediated re- combination and activation of KrasG12D allele (G). However, we observed no obvious effects of long-term growth on the ability of KrasG12D p53KO MEFs to form transformed foci. Overexpression of exogenous KrasG12D in p53KO MEFs produced similar results (H), effectively excluding the possibility that appreciable DNA damage had been caused by transient expression of Cre in LSL KrasG12D p53KO MEFs.

Ischenko et al. www.pnas.org/cgi/content/short/1219592110 3of7 rsini uosdrvdfo ASsre h1S,dul oiie c1S,adD uppltoso rs1Dp53 KrasG12D of subpopulations DN and Sca1-SP, positive, double Thy1-SP, FACS-sorted from derived tumors in pression p53 S3. Fig. shnoe al. et Ischenko KO Es ( MEFs. ( A h itiuino h1snl oiie(P,dul oiie(P,Sa-P n obengtv D)sbouain nttl notdKrasG12 unsorted total, in subpopulations (DN) negative double and Sca1-SP, (DP), positive double (SP), positive Thy1-single of distribution The ) B www.pnas.org/cgi/content/short/1219592110 xeietldsg otakclsrsosbefrtmriiito mn rs1Dp53 KrasG12D among initiation tumor for responsible cells track to design Experimental ) B A Thy1 h1Sa-(DN) Thy1-Sca1- cell population 65% 24% Thy1-Sp 3.5% 7% 42% 14% DP 41% 3% Sca1 Thy1-Sca1+ (Sca1-SP) Less differentiated cells: (Thy1-SP)andThy1+Sca1+(DP) Thy1+Sca1- More differentiated cells: Sca1-SP 86% 1.2% 13% 0% KO 94% 4% fi DN rbat.FC nlsso h1adSa ex- Sca1 and Thy1 of analysis FACS broblasts. 1.2% 0% KO fi broblasts. 4of7 D 18 51 A B C Metabolism and transport 33 69 Signal transduction Cell adhesion and cytoskeleton 46 Transcription factors Tumorigenic cells 16 Extracellular matrix (320 genes) Growth factors Cell growth 8 Proteolysis 30 Other 4 Unknown 46 27 4 9 6 10 7 12 Morphologically 9 25 21 transformed cells Morphologically transformed Morphologically transformed (132 genes) non-tumorigenic cells tumorigenic cells D SOX5 KLF5 FOXG1 DLX1 DLX2 SIX2 ETV4 HOXD9 HOXB5 HOXC10 HOXB2 HOXC6 IRF9 DDIT4L STAT3 SMAD6 DMRTA2 ANKRD1 FOSB FOS JUN PAWR BARX1 HOXB6 STAT2 HOXB9 CREB3L1 HEY1 IRX1 HOXD4 HOXB7 TCFL5 HOXB4 HOXC9 HOXD8 TCF3 HOXD10 SSBP2 PKNOX2 LMCD1 NFATC4 EYA4 RUNX1T1 PITX2 MEOX1 Morph. transformed Tumor 1 Tumor 2 Tumor 3 E Low High Expression (log2) HOXB5 DLX2 HOXB2 HOXD4 HOXB9 HOXB6 DLX1 PITX2 HOXD10 MEOX2 HOXD8 HOXC9 HOXB4 HOXB7 HOXC10 HOXD9 BARX1 HOXA2 MEOX1 PKNOX2 HOXC6 IRX1 IRX2 HOXD3 HOXA9 SIX2 Control Morph. transformed Tumor 1 Tumor 2 F Tumorigenic cells (combinations of DN and parental phenotypes)

Myc, PRC, Malignant other factors stages “Primed” Thy1-Sca1- (DN) Myc phenotypically immature cells KrasG12D Premalignant stages

Normal cells Morphologically transformed non-tumorigenic cells Thy1-SP Thy1-SP Sca1-SP Sca1-SP Th1/Sca1-DP Th1/Sca1-DP

Fig. S4. (A) The number of differentially expressed genes in morphologically transformed and tumor-derived cells. Genes with fourfold or greater difference are indicated. (B and C) Functional classification of differentially expressed genes in morphologically transformed (B) and tumor-derived cells (C). The number of genes belonging to each category is shown. (D) Heat maps showing induction of transcription factors in morphologically transformed KrasG12D p53KO cells and the respective tumors relative to untransformed controls. Genes with fourfold or greater difference are shown. (E) Heat maps comparing expression of genes in morphologically transformed and tumor cells relative to untransformed controls. (F) Ras-mediated transformation comprises separable stages. The earliest stage is morphological transformation that creates stable nontumorigenic cells. Contingent upon levels of Myc, these cells become competent to undergo malignant conversion.

Ischenko et al. www.pnas.org/cgi/content/short/1219592110 5of7 A B C Contr Vector KrasD12 Contr Vector KrasD12 80 Vector Myc KrasD12 Total Kras 60 shMyc

Myc Myc 40 20 MAPK MAPK

0 1 2 Tumor latency (days) DN Thy1+

Fig. S5. (A and B) Western blot analysis of control LSL KrasG12D p53KO (inactive allele) and KrasG12D p53KO (active allele) MEFs transduced with vector alone or KrasG12D-expressing retroviruses. (C) Tumor formation in nude mice by DN and Thy1-positive p53−/− KrasG12D fibroblasts transduced with the indicated vectors (104 cells per injection site). The error bars correspond to SD.

A B 100 Sca1+ Sca1- 80 60 EpCAM 40 EpCAM % Cells 20 Sca1 Sca1 0 1 2 3 p53+/+ p53+/-X Data p53-/- C Sca1- Sca1+ KrasD12, p4 tumor 1 tumor 3 Sca1- Sca1+ Cl16 Cl29 Cl6 Cl10 Cl18 Control p8 KrasD12, p12 KrasD12, Cl13 tumor 2 tumor 4

KrasD12 KrasG12D

Total Kras Total Kras

P-MAPK P-MAPK

Total MAPK Total MAPK D E

0.6 4 Sca1- Vector Sca1+ 3 Myc 0.4 2 0.2 1

0.0 1234567 0 1234 X Data X Data Relative levels mRNA Relative mRNA levels Relative NES LEF1 ZEB2 ZEB1 HES1 HNF6 SOX9 SNAIL2 SNAIL1 F TWIST1 TWIST2

5 Vector 4 KrasG12D 3 Myc 2 1

0 1 2 3 4 X Data Relative mRNA levels LEF1 ZEB2 TWIST1 TWIST2

Fig. S6. (A) The distribution of Sca1+ and Sca1− subpopulations in purified KrasG12D p53+/+, KrasG12D p53+/−, or KrasG12D p53KO pancreatic epithelial cells. + − − + (B) When isolated by FACS, the original Sca1 cells retain their parental phenotype after 2 wk culture (Left), whereas Sca1 cells give rise to Sca1 and Sca1 populations (Right). (C) Western blot analysis of control LSL KrasG12D p53KO cells, KrasG12D p53KO cells (passage numbers are indicated), tumors derived from − + passage 4 KrasG12D p53KO cells (Left), or Sca1 and Sca1 subsets of primary and clonal populations of KrasG12D p53KO pancreatic epithelial cells (Right). (D) − + Quantitative real-time RT-PCR (qRT-PCR) analysis of Sca1 and Sca1 populations of KrasG12D p53KO cells transduced with the indicated genes. Data were normalized to the corresponding HPRT value to obtain relative changes in the indicated mRNAs. (E) qRT-PCR analysis of Sca1+ KrasG12D p53KO cells transduced with empty vector or KrasG12D- or Myc-expressing retroviruses. (F) qRT-PCR analysis of Sca1+ KrasG12D p53KO cells transduced with empty vector or Myc- expressing retroviruses.

Ischenko et al. www.pnas.org/cgi/content/short/1219592110 6of7 Table S1. Acquisition of DN phenotype by tumor cells Pretumor phenotype

Clone ID % Thy1-positive % DN Tumor phenotype, % DN

+ 14%Thy1; 96% DP 0% DN 24–95* + 2 88% DN; 11% Sca1 88% DN 96–98* 3 60% Thy1+; 40% DP 0% DN 36–76* 4 84% Thy1+; 15% DP 0% DN 50 + 7 27% Thy1 ; 73% DP 0% DN 45 + 9 47% Thy1 ; 40% DP; 9% DN 9% DN 72 10 9% Thy1+; 90% DP 0% DN 16 + 11 33% Thy1 ; 66% DP 0% DN 33–93* + 13 40% Thy1 ; 58% DP 0% DN 36 16 80% Thy1+; 20% DP 0% DN 49 17 53% DN; 9% Sca1+ 53% DN 56 + 21 17% Thy1 ; 81% DP 0% DN 21 + 22 70% Thy1 ; 17% DP; 11% DN 11% DN 27 1/3 subclone 96% DP 0% DN 30 1/5 subclone 94% DP 0% DN 49 1/8 subclone 96% DP 0% DN 26 1/9 subclone 94% DP 0% DN 33 2/3 subclone 96% DN 96% DN 96 2/5 subclone 94% DN 94% DN 99 + 3/1 subclone 60% Thy1 ; 40% DP 0% DN 51 + 3/6 subclone 64% Thy1 ; 34% DP 0% DN 30 3/8 subclone 62% Thy1+; 38% DP 0% DN 24 + 3/9 subclone 58% Thy1 ; 40% DP 0% DN 69

FACS analysis of monoclonal KrasG12D p53KO cell lines and tumors. DN, double negative; DP, double positive. *Based on analysis of multiple tumors.

Table S2. Changes in the percentage of DN cells DN cells, %

Transduced genes Clone 1 Clone 2 Clone 3 Clone 11

Vector 0 96 0 0 KrasG12D* 18 100 24 52 KrasG12V* 39 100 20 46 HrasG12V* 36 100 28 48 NrasG12D* 24 100 21 36 Akt1* 5 99 10 5 Mek2* 14 99 7 14 EGFR 5 99 4 4 Myc 58992545 Bmi1 22 99 9 34 0 10 1 1 Oct4 6 98 3 2 24 69 16 12 Twist 3 98 13 5 p53 3 97 2 0 Myc + Bmi1 44 100 41 40 Myc + Klf4 10 48 8 16 Myc + Oct4 11 97 12 19 Myc + Sox2 66 89 31 45 Myc + p53 10 99 4 1 KrasG12D + Myc59994458 KrasG12D + MycΔMBII22981023

FACS analysis of monoclonal KrasG12D p53KO cell lines transduced with the indicated genes. DN, double negative. *Constitutively active mutants.

Ischenko et al. www.pnas.org/cgi/content/short/1219592110 7of7