MTB/TICNX vs. TICNX Cell lines with A TICNX MTB/TICNX B N1 fusions 15 2.0 vs. wild-type ** * *** 1.5 10 1.0 237072 748 5 0.5 Relative expression 0 0.0 NICD1 Hey1 Hey2 C D 2.5 TICNX *** 2.0 P < 0.0001 MTB/TICNX P < 0.0001 2.0 trend 1.5 1.5 *** 1.0 1.0 0.5 0.5 Estimated Notch activity 0 Estimated Notch activity 0 48h + doxycycline 96h + doxycycline ControlNOTCH1 L1601P NICD1 Testing data Training data E F 2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

Estimated Notch activity P = 0.0087 Estimated Notch activity P < 0.0001 0 0 Wild-type Constitutively active DMSO GSI NOTCH1 status Supplemental Figure 1 Generation and validation of a Notch pathway signature. (A) qRT-PCR for human NICD1 transgene and mouse Hey1 and Hey2 in TICNX and MTB/TICNX mice induced with doxycycline for 4d; n = 4. (B) Venn diagram showing overlap between differentially expressed in MTB/TICNX versus TICNX mice (GSE51628) and in human breast cancer cell lines with and without NOTCH1 gene fusions (GSE36133, ref. 31). (C) Estimation of Notch pathway activity in TICNX and MTB/TICNX mammary glands induced with doxycycline for 48h or 96h (GSE51628); n = 4. (D) Estimation of Notch pathway activity in MCF-10A cells transduced with increasingly potent NOTCH1 constructs (GSE20285, ref. 33); n = 4. (E) Estimation of Notch pathway activity in human T-ALL cell lines grouped by Notch1 activation status, wild-type (n = 4) or constitutively active (n =10) (GSE36133, refs. 31, 36, 89). The outlier in the wild-type group corresponds to TALL-1, a cell line that is known to be GSI sensitive (89) and exhibits activated Notch3 signaling (90). (F) Estimation of Notch pathway activity human T-ALL cell lines with activating NOTCH1 mutations treated with vehicle or GSI for 24h (GSE5716, ref. 36); n = 10. Data in A, C and D are shown as the mean ± s.d. Error bars in E represent mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001 calculated by two-tailed Student’s t tests in A, C, E and F, one-way ANOVA in D. A B C 0.4 0.20 0.15 P = 2.6e−43 P = 3.8e−21 P = 2.8e−11

0.3 0.15 0.10 0.2 0.10 0.05 0.1 0.05 Relative Notch activity 0 0 0 Normal LumA LumBHER2 Basal G2G1 G3 ER+ ER− Intrinsic molecular subtype Tumor grade ER status

D E Estimated Notch pathway activity versus RFS Estimated Notch pathway activity versus RFS adjusted for intrinsic molecular subtype adjusted for tumor grade

P = 0.00042 P = 0.0051

F Overall p-value: 0.0051 Estimated Notch pathway activity versus RFS adjustedj for ER status Supplemental Figure 2 Notch pathway activity is independently associated with decreased relapse-free survival in women with breast cancer. (A-C) Association between estimated Notch pathway activity and established prognostic factors in breast cancer, and (D-F) forest plots of association between estimated Notch pathway activity and recurrence-free survival (RFS) in women with breast cancer after adjusting for correlated prognostic factors using multivariate Cox proportional hazards regression: (A and D) intrinsic molecular subtype, (B and E) increasing tumor grade (G1 – G3) and (C and F) estrogen (ER) status. In A-C, P values were calculated by the two-tailed Student’s t test or one-way P = 0.0052 ANOVA. In D-F, the overall hazard ratios and associated confidence intervals are shown for the fixed-effects model. ABEstimated Notch pathway activity versus RFS Estimated Notch pathway activity versus RFS within basal-like tumors within high grade (G3) tumors

P = 0.00084 P = 0.00025

C Estimated Notch pathway activity versus RFS within ER− tumors Supplemental Figure 3 Notch pathway activity is associated with decreased relapse-free survival within high-risk subgroups. (A-C) Forest plots of association between estimated Notch pathway activity and recurrence-free survival (RFS) within: (A) basal-like tumors, (B) high grade (G3) tumors, and (C) (ER) negative tumors. The overall hazard ratios and associated confidence intervals are shown for the fixed-effects model.

P = 0.012 A B 2.0 P = 0.015 NS 4 Vehicle GSI *** Ptrend = 0.006 LAP LAP + GSI 1.5 3 *** *** 1.0 2 *** *** NS NS 0.5 *** 1 Relative expression Estimated Notch activity 0 Vehicle 0.1 μM 1 μM 0 Hey1 Hey2 Dll1 Lapatinib

Supplemental Figure 4 Notch signaling is activated in HER2+ SKBR3 human breast cancer cells following HER2 targeted therapy. (A) Estimation of Notch pathway activity in SKBR3 cells treated with vehicle or lapatinib for 24h (GSE38376, ref. 38). (B) qRT-PCR analyses of Notch signaling components in SKBR3 cells 48h after treatment with lapatinib alone and in combination with GSI. Data are shown as the mean ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001 by one-way ANOVA followed by the post test for linear trend or Bonferroni’s multiple comparison test; n = 3. NS not statistically significant. A 4h treatment 48h treatment B 4h treatment 48h treatment doxycycline doxycycline doxycycline doxycycline AKTi AKTi AKTi AKTi MEKi+AKTi MEKi+AKTi DMSO MEKi DMSO MEKi DMSO MEKi MEKi+AKTi DMSO MEKi MEKi+AKTi − − − −

pErk1/2 HER2/neu

pAkt -Tubulin -Tubulin

C 48h treatment DMSO -dox MEKi ERKi Supplemental Figure 5 Effects AKT, MEK and ERK inhibition in HER2/neu-dependent primary tumor NICD cells derived from MTB/TAN mice. (A and B) Western blot analysis of (A) pErk1/2 (B) pAkt and HER2/neu, 4h or 48h after doxycycline withdrawal or treatment with MEKi and/or AKTi. (C) Western blot alysis of NICD 48h after doxycycline withdrawal or treatment with MEKi or ERKi. β-tubulin and GAPDH Gapdh are shown as loading controls. Western blot results are representative of 3 different experiments. A pk1 B pk1 100 Hes1 10000 Nrarp

1000 10 100

10 1 1 Relative Hes1 expression 0.1 Relative Nrarp expression 0.1 DMSO GSI DMSO GSI DMSO GSI DMSO GSI +dox −dox +dox −dox

C D 1000 MigR1 * 4 0 hours NICD1 24 hours Ptrend = 0.0007 48 hours 100 *** 3

10 2

Ptrend = NS 1 1 Relative expression normalized to MigR1

0.1 Relative Hey1 expression 0 Hey1 Hey2 MigR1 dnMAML

Supplemental Figure 6 Experimental manipulation of Notch signaling in primary MTB/TAN tumor cells. (A) qRT-PCR analysis of Hes1 expression in primary MTB/TAN tumor cells transduced with control (pk1Empty) or Hes1 expression constructs. (B) qRT-PCR analysis of Nrarp expression in primary MTB/TAN tumor cells transduced with control (pk1Empty) or Nrarp expression constructs. (C) qRT-PCR for Hey1 and Hey2 expression in primary MTB/TAN tumor cells transduced with control (MigR1) or NICD1 expression constructs. (D) qRT-PCR for Hey1 at 0, 24h and 48h after doxycycline withdrawal in primary MTB/TAN tumor cells transduced with control (MigR1) or dnMAML expression constructs. β-tubulin is shown as a loading control. Data are shown as the mean ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001 by one-way ANOVA followed by Bonferroni multiple comparison test or the post test for linear trend; n = 3. A B 100 15 AdGFP ) 3 AdCre 75 10 50 5 25

Tumor volume (mm Tumor P = NS P = NS 0 to regression (days) Time 0 0 5 10 15 20 25 AdGFP AdCre Days post injection C D 100 Rbpjflox/flox + AdGFP Rbpjflox/flox + AdCre AdGFP AdCre 75 RbpJ

50 -Tub

25

P = NS 0 Recurrence-free survival (%) 020 40 60 80 100 Days post doxycycline withdrawal

E F 1.5 flox/flox

) 100 Rbpj + AdGFP 3 Rbpjflox/flox + AdCre 1.0 75

50 0.5 25

Tumor volume (mm Tumor Relative RbpJ P = 0.008 p = ns P = NS 0 0 Rbpjflox/flox Rbpjflox/flox 07525 50 +AdGFP +AdCre Days post injection Supplemental Figure 7 Transduction with AdCre effectively excises floxed alleles without altering tumor progression. (A) Representative tumor growth rates for primary MTB/TAN tumor cells transduced with AdGFP (n = 16) or AdCre (n = 18). (B) Time to regression after doxycycline withdrawal for orthotopic tumors generated from primary MTB/TAN tumor cells transduced with AdGFP (n = 8) or AdCre (n = 15). (C) Kaplan-Meier survival curves showing indistinguishable latencies for recurrence-free survival in mice harboring fully regressed orthotopic tumors generated from primary MTB/TAN tumor cells transduced with AdGFP (n = 6) or AdCre (n = 14). (D) Western blot analysis of Rbpj and β-tubulin (control) in primary orthotopic tumors generated from MTB/TAN/Rbpjflox/flox tumor cells infected with AdGFP or AdCre. (E) Quantification of western blot in D showing Rbpj expression normalized to β-tubulin. (F) Tumor growth rates for primary orthotopic tumors generated from an MTB/TAN/Rbpjflox/flox tumor that was digested to yield a single cell suspension and infected with AdGFP (n = 30) or AdCre (n = 26). Error bars in A, B and D-F represent mean ± s.e.m. P values calculated by two-tailed Student’s t test in A, B, E and F, and by the Mantel-Haenszel method in C. NS not statistically significant. A Primary Recurrent AdGFP AdCre

B

Primary tumors Recurrent tumors AdGFPAdCre AdGFP AdCre Ladder Wild-type H2O Ladder

Rbpj Δ 400bp

Gapdh 590bp

Supplemental Figure 8 Tumor cells with Rbpj deletion are selected against during tumor recurrence. (A) PCR analysis of Gapdh loading control and Rbpjflox/flox excision (Rbpj Δ) in primary and recurrent tumors generated from orthotopic injections of MTB/TAN/Rbpjflox/flox tumor cells infected with AdGFP or AdCre. (B) Immunohistochemistry analysis of Rbpj expression in primary and recurrent tumors generated from orthotopic injections of MTB/TAN/Rbpjflox/flox tumor cells infected with AdGFP or AdCre (40x original magnification). Images are representative of 6 AdGFP primary tumors, 3 AdCre primary tumors, 5 AdGFP recurrent tumors and 3 AdCre recurrent tumors. A +doxycycline (+HER2/neu) −doxycycline (−HER2/neu) Tumor Progression: Latency Primary Tumorigenesis Regression Dormancy Recurrence

Treatment paradigm Preventative (Fig. S9A) GSI Single-agent (Fig. S9B) GSI Adjuvant (Fig. 6C) GSI Salvage (Fig. S10B) GSI

B C

100 2000 Vehicle Vehicle ) 150mg/kg GSI 3 150mg/kg GSI 75 300mg/kg GSI 1500 300mg/kg GSI

50 1000

25 500

Tumor volume (mm Tumor

Tumor-free survival (%) Tumor-free P = NS P = NS 0 0 050 100 150 082 4 6 Days of treatment Weeks of treatment

Supplemental Figure 9 Notch signaling is dispensable in HER2/neu-induced primary tumorigenesis. (A) Schematic of tumor progression in MTB/TAN mice and treatment paradigms tested in B and C. (B) Kaplan-Meier survival curves showing primary tumor-free survival in MTB/TAN mice treated with vehicle (n = 15), 150mg/kg GSI (n = 12) or 300mg/kg GSI (n = 15) initiated concurrently with doxycycline induction. P values calculated by the Mantel-Haenszel method. (C) Primary tumor growth curves in MTB/TAN mice treated with vehicle (n = 14), 150mg/kg GSI (n = 13) or 300mg/kg GSI (n = 13) initiated at the time of primary tumor detection shown as mean ± s.e.m. P values calculated by one way ANOVA. NS not statistically significant. A +doxycycline (+HER2/neu) −doxycycline (−HER2/neu) Tumor Progression: Latency Primary Tumorigenesis Regression Dormancy Recurrence

Treatment paradigm Preventative (Fig. S9A) GSI Single-agent (Fig. S9B) GSI Adjuvant (Fig. 6C) GSI Salvage (Fig. S10B) GSI

B

100 Vehicle 150mg/kg GSI Supplemental Figure 10 75 300mg/kg GSI Notch signaling is dispensable following recurrence in MTB/TAN mice. (A) Schematic of tumor progression in 50 MTB/TAN mice and treatment paradigm tested in B. (B) Kaplan-Meier survival curves showing progression-free survival of MTB/TAN mice in which treatment with vehicle 25 (n = 11), 150mg/kg GSI (n = 11), or 300mg/kg GSI (n = P = NS 10) was initiated after detection of recurrent tumors. P Ptrend = NS value calculated by the Mantel-Haenszel method. NS not 0

Progression-free survival (%) statistically significant. 0 50 100 150 Days of treatment Supplemental Table 1. 72-gene Notch pathway activity signature

ENTREZID SYMBOL GENE NAME 25890 ABI3BP ABI family, member 3 (NESH) binding protein 84752 B3GNT9 UDP-GlcNAc:betaGal beta-1,3-N-acetylglucosaminyltransferase 9 57673 BEND3 BEN domain containing 3 83990 BRIP1 BRCA1 interacting protein C-terminal helicase 1 753 C18orf1 18 open reading frame 1 81563 C1orf21 chromosome 1 open reading frame 21 767 CA8 carbonic anhydrase VIII 153733 CCDC112 coiled-coil domain containing 112 126731 CCSAP centriole, cilia and spindle-associated protein 7203 CCT3 chaperonin containing TCP1, subunit 3 (gamma) 947 CD34 CD34 molecule 51148 CERCAM cerebral endothelial cell adhesion molecule 57530 CGN cingulin 1105 CHD1 chromodomain helicase DNA binding protein 1 92421 CHMP4C charged multivesicular body protein 4C 26097 CHTOP chromatin target of PRMT1 79827 CLMP CXADR-like membrane protein 91522 COL23A1 collagen, type XXIII, alpha 1 84570 COL25A1 collagen, type XXV, alpha 1 55172 DNAAF2 dynein, axonemal, assembly factor 2 1824 DSC2 desmocollin 2 1832 DSP desmoplakin 1942 EFNA1 ephrin-A1 8664 EIF3D eukaryotic translation initiation factor 3, subunit D 51313 FAM198B family with sequence similarity 198, member B 2191 FAP fibroblast activation protein, alpha 10116 FEM1B fem-1 homolog b (C. elegans) 10690 FUT9 fucosyltransferase 9 (alpha (1,3) fucosyltransferase) 55127 HEATR1 HEAT repeat containing 1 3280 HES1 hairy and enhancer of split 1, (Drosophila) 9026 HIP1R huntingtin interacting protein 1 related 3339 HSPG2 heparan sulfate proteoglycan 2 3608 ILF2 interleukin enhancer binding factor 2, 45kDa 3615 IMPDH2 IMP (inosine 5'-monophosphate) dehydrogenase 2 3625 INHBB inhibin, beta B 81875 ISG20L2 interferon stimulated exonuclease gene 20kDa-like 2 25959 KANK2 KN motif and ankyrin repeat domains 2 286826 LIN9 lin-9 homolog (C. elegans) 54799 MBTD1 mbt domain containing 1 4170 MCL1 myeloid cell leukemia sequence 1 (BCL2-related) 92312 MEX3A mex-3 homolog A (C. elegans) 6183 MRPS12 mitochondrial ribosomal protein S12 4609 v-myc myelocytomatosis viral oncogene homolog (avian) 4645 MYO5B myosin VB 51602 NOP58 NOP58 ribonucleoprotein homolog (yeast) 4851 NOTCH1 notch 1 131870 NUDT16 nudix (nucleoside diphosphate linked moiety X)-type motif 16 57451 ODZ2 odz, odd Oz/ten-m homolog 2 (Drosophila) 23659 PLA2G15 phospholipase A2, group XV 84172 POLR1B polymerase (RNA) I polypeptide B, 128kDa 5437 POLR2H polymerase (RNA) II (DNA directed) polypeptide H 5447 POR P450 (cytochrome) oxidoreductase 8495 PPFIBP2 PTPRF interacting protein, binding protein 2 (liprin beta 2) 401494 PTPLAD2 protein tyrosine phosphatase-like A domain containing 2 5867 RAB4A RAB4A, member RAS oncogene family 65997 RASL11B RAS-like, family 11, member B 171177 RHOV ras homolog family member V 6194 RPS6 ribosomal protein S6 1901 S1PR1 sphingosine-1-phosphate receptor 1 6303 SAT1 spermidine/spermine N1-acetyltransferase 1 90701 SEC11C SEC11 homolog C (S. cerevisiae) 6560 SLC12A4 solute carrier family 12 (potassium/chloride transporters), member 4 388662 SLC6A17 solute carrier family 6, member 17 30812 SOX8 SRY (sex determining region Y)-box 8 6713 SQLE squalene epoxidase 90627 STARD13 StAR-related lipid transfer (START) domain containing 13 163183 SYNE4 spectrin repeat containing, nuclear envelope family member 4 6876 TAGLN transgelin 25907 TMEM158 transmembrane protein 158 (gene/pseudogene) 57659 ZBTB4 and BTB domain containing 4 7580 ZNF32 zinc finger protein 32 79755 ZNF750 zinc finger protein 750

Supplemental References

89. O’Neil J et al. FBW7 mutations in leukemic cells mediate NOTCH pathway activation and resistance to gamma-secretase inhibitors.. J Exp Med. 2007;204(8):1813–24.

90. Masiero M et al. Notch3-mediated regulation of MKP-1 levels promotes survival of T acute lymphoblastic leukemia cells.. Leuk Off J Leuk Soc Am Leuk Res Fund, UK. 2011;25(4):588– 98.