Association between TGF-β signaling and HMGA2, a potential biomarker for Poster No. 1573 bintrafusp alfa in triple-negative breast cancer

T.-L. Yeung, G. Locke, A. Lazorchak, G. Qin, H. Yu, J. Qi, B. Marelli, M. H. Jenkins, A. Rolfe, L. S. Ojalvo, I. Dussault, Y. Lan EMD Serono Research & Development Institute, Inc., Billerica, MA, USA; a business of Merck KGaA, Darmstadt, Germany

• HMGA2 expression signi cantly correlated with the INTRODUCTION Figure 3. Best change in target lesions from RESULTS CONCLUSIONS baseline assessed by investigator expression of TGF-β receptors, ligands, collagen,

• Bintrafusp alfa is a rst-in-class bifunctional fusion 150 and EMT-related genes in bintrafusp alfa–treated BOR • HMGA2 expression signi cantly correlated with TGF-β signaling–related genes⁴ • Patients with TNBC who experienced disease 130 PR SD mice (Figure 7) protein composed of the extracellular domain of the 110 PD in tumors treated with isotype control (29%) or bintrafusp alfa (55%) (Figure 5), control with bintrafusp alfa had high expression of TGF- RII receptor, to function as a TGF- “trap”, fused to 90 * Figure 7. Correlation between HMGA2 and TGF-β genes β β 70 indicating a strong association between HMGA2 and TGF-β signaling HMGA2 PD-L1 50 Tgfbr1 Core genes Tgfbr2 a human IgG1 antibody blocking PD-L1 (Figure 1) expression 6 5 30 0 Figure 5. Association between HMGA2, TGF-β signaling, and bintrafusp alfa • In a murine model, the correlation of HMGA2 and * ≥1 10 Figure 1. Proposed mechanism of action of bintrafusp alfa * * * * * ≥5 pharmacodynamic effects in a TNBC model 4 -10 ≥10 TGF-β signaling was stronger in bintrafusp alfa–treated 4 Tumor cells 20 -30 Isotype control Bintrafusp alfa 3 NK cell Mesenchymal-like tumors compared with isotype control–treated tumors

tumor cell Change in sum of diameters, %

-50 expression expression 2 TAM -70 29% correlated 55% correlated 2 Immune Hmga2 1.0 Hmga2 1.0 HMGA2 Smad3 HMGA2 Smad3 Tgfb1 Tgfb1 N=15 N=15 Dendritic Tgfb2 Tgfb2 • These data demonstrate a link between HMGA2 phenotype Tgfb3 Tgfb3 1 cell EMT (leading to metastasis and resistance Tgfbr1 Tgfbr1 Tgfbr1 r=0.846 Tgfbr2 r=0.911 TC PD-L1 expression Tgfbr2 Tgfbr2 Tumor angiogenesis Tgfbr3 Tgfbr3 to therapy [including checkpoint inhibition]) Fibroblast I Bmp2 Bmp2 Bmp3 Bmp3 p<0.001 p<0.001 Bmp4 Bmp4 TME PD-L1 expression E Excluded Bmp5 Bmp5 Bmp6 Bmp6 0 0 expression and TGF- ; bintrafusp alfa and Trap control CAF Bmp7 Bmp7 β D Desert Bmp10 Bmp10 Bmp15 Bmp15 Cytotoxic Immune phenotype E E E E E E D E E E ‡ D D E E E E D E E E D Bmpr1a Bmpr1a 2.53.0 3.54.0 4.55.0 5.5 2.53.0 3.54.0 4.55.0 5.5 † Bmpr1b Bmpr1b T cell Bmpr2 Bmpr2 Acvr1 Acvr1 Hmga2 expression Hmga2 expression Acvr1b Acvr1b Acvr1c Acvr1c reduced expression of HMGA2 and individual TGF- Acvr2a Acvr2a β Patients with a BOR of NE (n=5) are not included in this gure. Five additional patients were not Acvr2b 0.5 Acvr2b 0.5 Acvrl1 Acvrl1 Suppression of immune response TGF-β trap moiety sequesters Fibrosis and impaired Nodal Nodal Gdf1 Gdf1 Tgfb1 Tgfb2 TGF- to inhibit downstream signaling drug access Gdf11 Gdf11 β included because they lacked a valid postbaseline target lesion measurement. Inha Inha Inhba Inhba TGF- * Inhbb Inhbb signaling–related genes relative to anti–PD-L1 β Inhbc Inhbc 6 0.0 Inhbe Inhbe *Patients with HMGA2-high tumors who had an evaluable response. Smad2 Smad2 Smad1 Smad1 Smad5 Smad5 † Smad4 Smad4 T cell Smad6 Smad6 Sample had an indeterminate immune phenotype. Smad7 Smad7 -0.2 Bintrafusp Tumor cells Smad9 Smad9 Sptbn1 Sptbn1 ‡ Tgfbrap1 Tgfbrap1 • Collectively, these observations warrant further alfa Zfyve9 Zfyve9 PD-L1 expression and immune phenotype were not determinable. Aldh1a1 Aldh1a1 PD-1 PD-L1 Bcl2l11 Bcl2l11 4 Cdh1 Cdh1 Cdh2 0 Cdh2 0 -0.4 Cdkn1a Cdkn1a Cdkn2b Cdkn2b Fos Fos Jun Jun analysis of the potential link between bintrafusp alfa Myc Myc Anti–PD-L1 mAb moiety Col1a1 Col1a1 expression expression HMGA2 expression was 32-fold higher in patients who Col1a2 • Col1a2 blocks PD-L1 interactions with PD-1 Col3a1 Col3a1 -0.6 Dab2 Dab2 Dapk1 Dapk1 2 Fn1 Fn1 Foxk1 Foxk1 Foxk2 Foxk2 antitumor activity and HMGA2 Foxo1 N=15 Foxo3 Foxo1 N=15 experienced disease control with bintrafusp alfa Foxo4 Foxo3 Tgfb1 Tgfb2 -0.8 Gadd45b Foxo4 Id1 Gadd45b r= 0.590 r=0.696 Igf2 Id1 Il6 Igf2 Itgb6 Il6 Itgb8 Itgb6 p=0.023 p=0.005 Mapk14 -0.5 Itgb8 Mmp2 Mapk14 -0.5 0 -1.0 Mmp2 • Based on phase 1 trial results and biomarker compared with those who had PD (Figure 4) Mmp9 Pard6a Mmp9 Pja1 Pard6a Pja2 Pja1 2.53.0 3.54.0 4.55.0 5.5 2.53.0 3.54.0 4.55.0 5.5 CAF, cancer-associated broblast; EMT, epithelial-mesenchymal transition; NK, natural killer; Pja2 Runx1 Runx2 Runx1 Runx3 Runx2 Hmga2 expression Hmga2 expression Serpine1 Runx3 Smad7 Serpine1 TAM, tumor-associated macrophage. Smad7 analysis, a phase 2 study is planned to evaluate the Snai1 Figure 4. HMGA2 expression by investigator-assessed Snai2 Snai1 Stat3 Snai2 Tert Stat3 Twist1 Tert *Tumor cells are also a major source of TGF- in the tumor microenvironment (TME). Twist2 Twist1 β Twist2 Vegfa Vim Vegfa Tgfb3 Smad3 10 Zeb1 Vim BOR Zeb2 Zeb1 clinical activity of bintrafusp alfa in patients with † Zeb2

l 4.5 5 3 5 2 1 1 0 1 2 1 5 4 6 7 9 c 1 2 3 7 1 2 a 2 3 1 1 1 1 2 3 n f2 m Il6 a1 a2 e1 l b2 p2 p9 p2 p3 p4 p5 p6 p7 ai ai 3 l1 1 7 1 2 a 1 2 3 c 2 1 0 5 1 2 1 5 4 6 7 9 2 3 at 1 1 1 1 2 3 Id n gf • In a global, phase 1, open-label trial of bintrafusp alfa pr p1 pk p1 ad ad ad ad ad ad ad ad ga ad f2 m Fos Ju Fn1 Ig gb6 gb8 Vi Tert My l1a1 l1a2 l3a1 Il6 a1 a2 e1 p2 p9 b2 p2 p3 p4 p5 p6 p7 Pj Pj Inha pr1a rd6a tbn1

9 pr1b l2 ai ai l1 kn1a Gdf1 at Id Zeb1 Zeb2 kn2b gf It It pr Da St Cdh1 Cdh2 pk ad p1 p1 ad ad ad ad ad ad ad ga ad Fn1 Ig Fos Ju gb6 gb8 Inhba Inhbc Inhbe Sn Sn Vi Inhbb Tgfb1 Tgfb2 Tgfb3 Ve Tert Acvr Noda Bm Bm Bm Bm Bm Bm My l1a1 l1a2 l3a1 Gdf1 in Foxk1 Foxk2 Mm Mm dh1a1 Pj Pj Inha Foxo1 Foxo3 Foxo4 rd6a pr1a tbn1 Da rp in pr1b Acvrl1 Runx Runx Runx l2 Tgfbr Tgfbr Tgfbr Twist1 Twist2 kn1a Gdf1 Bm Zfyve9 Zeb1 Zeb2 kn2b It It Sm Sm Sm Sm Sm Sm Sm Co Co Co Sm Sm St Da Acvr1c Acvr2a Pa Bm Bm Hm Cdh1 Cdh2 Acvr1b Acvr2b Sp Bc Sn Sn Inhba Inhbc Inhbe Inhbb Tgfb1 Tgfb2 Tgfb3 Ve Cd Acvr Bm Bm Bm Bm Bm Bm Noda Cd Mapk14 Gdf1 Bm Foxk1 Foxk2 Mm Mm dh1a1 Bm Foxo1 Foxo3 Foxo4 Da rp Al Runx Runx Runx Acvrl1 Twist1 Twist2 Tgfbr Tgfbr Tgfbr Bm Zfyve9 Sm Co Co Co Sm Sm Sm Sm Sm Sm Sm Sm Pa Acvr1c Acvr2a Bm Bm Hm Se Acvr1b Acvr2b Sp Tgfbrap Gadd45b Bc Cd Mapk14 Cd Bm Bm Al Se Gadd45b Tgfbrap 4.0 4 HMGA2-high TNBC (NCT02517398), 33 patients with triple-negative breast 8 TGF-β core genes TGF-β target genes TGF-β core genes TGF-β target genes 7 3.5 3

cancer (TNBC), who received bintrafusp alfa 1200 mg 6 expression • Bintrafusp alfa or Trap control (TGF-β trap linked to an inactive anti–PD-L1) reduced expression (TPM+0.5) 2 3.0 2 5 every 2 weeks until con rmed progression, were log REFERENCES 4 expression of these HMGA2-correlated TGF-β signaling–related genes relative to Tgfb3 2.5 N=15 Smad3 1 N=15 r=0.850 r=0.943 analyzed for best overall response (BOR) 3 p<0.001 p<0.001 anti–PD-L1 or isotype control treatment (Figure 6), suggesting that HMGA2 is correlated 2.0 0 1. Lan Y, et al. Sci Transl Med. 2018;10:eaan5488. 2 2.53.0 3.54.0 4.55.0 5.5 2.53.0 3.54.0 4.55.0 5.5 NE PD* SD PR Hmga2 expression Hmga2 expression Response with TGF- –related signaling more strongly in the presence of TGF- sequestration – The investigator-assessed median duration of TPM, transcripts per million. β β 2. Spira A, et al. Cancer Res. 2020;80(Suppl 4):Abstract P3-09-06. response was 5.4 months (range, 4.7-5.5) in patients *Three samples from patients with a BOR of PD are not included in this gure due to lack of Figure 6. Effects of bintrafusp alfa on HMGA2 expression and TGF- signaling activity Target genes 3. Locke G, et al. Cancer Res. 2020;80(Suppl 4):Abstract P3-09-13. sample or sequencing failure. β Col1a2 Col1a1 4. Korkut A, et al. Cell Syst. 2018;7:422-37. with con rmed responses (Figure 2) †One sample from a patient with a BOR of PR was excluded from the formal biomarker Core genes Target genes 8 7.0 assessment due to failing quality control. Tgfbr1 Tgfbr2 Col1a1 Col1a2 6.5 5 p=0.434 1 8 8 p=0.002 6 – 4 patients had a BOR of partial response (PR) per p<0.001 6.0 • HMGA2 expression is upregulated by TGF-β signaling p=0.001 7 investigator assessment (3 con rmed, 1 uncon rmed; 7 4 5.5 ACKNOWLEDGMENTS and is an important factor in mediating TGF-β–induced 4 6 expression expression 5.0 objective response rate,12.1%) (Figures 2 and 3) 0 5 6 2 N=15 N=15

EMT Col1a1 r=0.714 Col1a2 4.5 r=0.746 The authors thank the patients and their families, investigators, 3 4 p=0.004 p=0.002 – An additional patient had stable disease (SD), giving a 5 0 4.0 co-investigators, and the study teams at each of the participating • Using the same cutoff for HMGA2-high expression, 3 2.53.0 3.54.0 4.55.0 5.5 2.53.0 3.54.0 4.55.0 5.5 disease control rate of 15.2% Gene expression Gene expression Gene expression Gene expression Hmga2 expression Hmga2 expression centers and at Merck KGaA, Darmstadt, Germany, and EMD approximately 13.0% of tumor samples annotated as 2 -1 2 4 Vim Vegfa Serono Research & Development Institute, Inc., Billerica, MA, USA; • Tumor cell (TC) PD-L1 expression was not associated Hmga2 Tgfb1 Tgfb2 Vim Vegfa 10.5 6 TNBC in the Breast Invasive Carcinoma TCGA (The 6 6 p<0.001 1 11 6 p=0.475 p=0.004 p<0.001 p=0.036 a business of Merck KGaA with response to treatment (Figure 3) 10.0 Cancer Genome Atlas) data set had high expression of 5 4 4 9.5 This study was funded by Merck KGaA and is part of an alliance Figure 2. Change in target lesions assessed by investigator 10 4

HMGA2 after adjustment with the ComBat algorithm expression 4 0 9.0 between Merck KGaA and GlaxoSmithKline PR expression 2

% 170 SD 2 9 2 N=15 N=15

Vim 8.5 Vegfa , 150 PD 3 r=0.732 r=0.750

Gene expression Gene expression Gene expression Gene expression p=0.003 p=0.002

METHODS Gene expression 130 Patient off treatment 8.0 0 All authors are employees of EMD Serono Research & 110 First occurrence of new lesion 2 0 -1 8 0 2.53.0 3.54.0 4.55.0 5.5 2.53.0 3.54.0 4.55.0 5.5 • This study evaluated the association between HMGA2 Hmga2 expression Hmga2 expression Development Institute, Inc.; a business of Merck KGaA, ameter s 90 i Tgfb3 Smad3 Fn1 Zeb1

d PD-L1

70 f p<0.001 Darmstadt, Germany Anti– 5 5 10 2.0 o 50 expression and TGF- signaling in TNBC and the Trap control p=0.02 p<0.001 Fn1 Zeb1

β Isotype control Bintrafusp alfa p=0.250 8.5 1.0 30 1.5 effect of bintrafusp alfa on HMGA2/TGF- signaling 4 9 su m 10 β 8.0 4 1.0 0.5 n

i -10

3 8 e -30 • The syngeneic murine 4T1 tumor model was used in 0.5 7.5 g 0.0 Copies of this poster obtained through Quick Response (QR) -50 2 7

3 0.0 7.0 expression -70 this study because it closely mimics human stage IV expression Code are for personal use only and may not be reproduced

Cha n 1 6 -0.5 N=15 -0.5 N=15 Gene expression Gene expression Gene expression Gene expression 6.5 Fn1 r=0.721 Zeb1 r=0.525 without permission from AACR® and the author of this poster 0 1 2 3 4 5 6 7 8 9 10 11 TNBC and has high baseline levels of HMGA2 expression, 2 0 5 -1.0 p=0.003 p=0.047 Months 6.0 -1.0 2.53.0 3.54.0 4.55.0 5.5 2.53.0 3.54.0 4.55.0 5.5 Correspondence: Tsz-Lun Yeung, [email protected] NE, not evaluable; PD, progressive disease. and bintrafusp alfa directly modulates canonical TGF- PD-L1 PD-L1 PD-L1 β PD-L1 Hmga2 expression Hmga2 expression GET POSTER PDF Anti– Patients with a BOR of NE (n=5) are not included in this gure. Five additional patients were not Anti– Trap control Anti– Trap control Trap control Anti– Trap control Isotype control Bintrafusp alfa Isotype control Bintrafusp alfa Isotype control Bintrafusp alfa All expression values are log (TPM+0.5) included because they lacked a valid postbaseline target lesion measurement. signaling and induces antitumor activity in this model Isotype control Bintrafusp alfa 

Abstract No. 1053. Presented at the 2020 AACR Virtual Annual Meeting II, June 22–24, 2020