Combinations with Allosteric SHP2 Inhibitor TNO155 to Block Receptor Tyrosine Kinase Signaling

The MAPK pathway plays a crucial role in regulation of cell growth, proliferation, and differentiation, and is frequently activated in cancers (1). A large arsenal of inhibitors targeting different nodes of this pathway including RTKs, KRAS^G12C, BRAF, CRAF, MEK1/2, and ERK1/2 have been developed to treat RTK- and MAPK-dependent cancers (2). However, the effectiveness of these RTK and MAPK inhibitors (RTKi and MAPKi) is often limited by pathway feedback activation at the RTK level in response to pathway suppression, particularly ERK inhibition. Even in cases of durable responses to RTK inhibitors such as EGFR tyrosine kinase inhibitors (TKI), acquired resistance inevitably occurs, which often involves reactivation of EGFR via secondary EGFR “gatekeeper” mutations or bypass activation of alternative RTKs such as MET (3–5). Combination therapies with RTK inhibitors to block reactivation or bypass activation of RTKs are needed for sustained MAPK pathway suppression and durable anti-tumor efficacy.

SHP2, encoded by PTPN11, is a nonreceptor protein tyrosine phosphatase that transduces signaling from various RTKs to promote the activation of RAS and subsequently the downstream MAPK pathway. The recently discovered allosteric SHP2 inhibitors (6–8) enable simultaneous inhibition of multiple RTKs. Several SHP2 inhibitors including TNO155 (9) and RMC-4630 are currently in clinical development (ClinicalTrials.gov identifier: NCT03114319 and NCT03634982) for various types of cancers and exhibited single-agent activity on ERK inhibition, preliminary efficacy, and tolerability (10). Given the intratumor and intertumor heterogeneity of RTK utilization and the technical challenges of identifying the specific RTK(s) activated in the absence of genetic alterations, SHP2 inhibitors offer an appealing one-size-fits-all approach to overcome RTK-mediated feedback activation of RAS, thus enhancing the efficacy of RTKi and MAPKi.

SHP2 is ubiquitously expressed in the human body and also plays a crucial role in immune cells such as T cells and macrophages (11). In T cells, SHP2 mediates the programmed cell death-1 (PD-1) signaling through several mechanisms including dephosphorylation of ZAP70, a kinase downstream of the T-cell receptor (12), and CD28, a receptor that provides costimulatory signals for T-cell activation (13). In immune-suppressive tumor-associated macrophages (TAM), SHP2 transduces colony-stimulating factor 1 receptor (CSF1R) signaling, which is essential for T-cell suppression by TAM (14). Therefore, SHP2 inhibition may reverse immune suppression in the tumor microenvironment through downstream inhibition of both PD-1 signaling in T cells and CSF1R signaling in TAMs. With both tumor intrinsic and immune-mediated mechanisms of antitumor activity, SHP2 inhibitors are also a rational combination partner for immunotherapies such as checkpoint inhibitors.

In addition to the RAS–MAPK pathway, RTKs also activate other downstream effector pathways such as PI3K–AKT, JAK–STAT, and PLCγ–PKC signaling (15). Although SHP2 is implicated in both PI3K–AKT (16) and JAK–STAT signaling in certain cellular contexts (17), allosteric SHP2 inhibitors are often most effective in inhibiting the RAS–MAPK pathway in RTK-driven cell lines (6). In addition, we recently reported that allosteric SHP2 inhibitors are less effective in a subset of FGFR-driven cell lines with rapid feedback activation (18). It remains unclear whether SHP2 inhibitors are effective at combating all types of RTK activation in cancers with different genetic backgrounds. Therefore, we studied the combination efficacy and synergistic mechanisms of TNO155 with inhibitors of EGFR, BRAF, and KRAS^G12C. We also observed combination benefit of the CDK4/6 inhibitor ribociclib with TNO155 in a large panel of lung and colorectal cancer patient-derived xenograft (PDX) models including those with KRAS mutation, with comparable efficacy to the combination of the MEK inhibitor trametinib with TNO155 but improved tolerability. Finally, we explored the immunomodulatory effects of SHP2 inhibition in CSF1R-driven models of TAMs and also found combination benefit of TNO155 and anti-PD-1 antibody.

Human cancer cell lines were obtained from the Novartis Cancer Cell Line Encyclopedia stock (19) and authenticated by SNP analysis and tested for Mycoplasma infection using a PCR-based method (IDEXX BioAnalytics). Cell lines were cultured in ATCC-specified medium: RPMI (Thermo Fisher Scientific) for all cell lines except HT-29 (McCoy's 5A), RKO and A-427 (MEMα), MDST8, and MIA PaCa-2 (DMEM), supplemented with 10% FBS (VWR). HCC827-GR cells (3) and PC-9 EGFR^T790M/C797S cells (20) were provided by the laboratory of Jeffrey Engelman at MGH (Boston, MA). Cell lines were used within 15 passages of thawing and continuously cultured for less than 6 months.

Human peripheral blood mononuclear cell (PBMC) was isolated by spinning CPT tubes (BD Biosciences, catalog no. 362761) containing exsanguinated whole blood at 1,800 rpm for 20 minutes. CD14⁺ cells and CD3⁺ cells were isolated from PBMC by following the instruction of Human Pan Monocyte Isolation Kit (Miltenyi Biotec, catalog no. 130-096-537) and Human Pan T Cell Isolation Kit (Miltenyi Biotec, catalog no. 130-096-535), respectively.

All small-molecule inhibitors used were synthesized and structurally verified by nuclear magnetic resonance and LC/MS according to cited references at Novartis. Cetuximab antibody was produced by Eli Lilly (catalog no. NDC-66733-958-23) and murine anti-PD-1 clone 332.1D2 is a kind gift of Gordon Freeman at DFCI (Boston, MA).

For CellTiter-Glo assay, 2,000–3,000 cells per well (3-day assay) or 500–1,000 cells per well (6-day assay) were seeded in 96-well plates [for CD14⁺ cells, the medium is supplemented with 50 ng/mL macrophage colony-stimulating factor (M-CSF) except no M-CSF control wells]. Cells were treated 24 hours after seeding and cell viability was measured by the CellTiter-Glo Assay (Promega, catalog no. G7573).

The combination dose matrix assay and the colony formation assay were performed as described previously (21). For colony formation assay quantification, the crystal violet staining in each well was dissolved in 10% acetic acid and the absorbance at 590 nm was measured using a spectrometer.

For IncuCyte assay, 500 cells per well were seeded in 96-well plates and treated 24 hours after seeding. The plates were then placed into IncuCyte incubation chamber, and four pictures per well were taken every 24 hours. Cell numbers were determined by the contrast analysis by the IncuCyte system.

Immunoblotting was performed as described previously (21). The following primary antibodies were used: phospho-ERK (p-ERK; Cell Signaling Technology, catalog no. 4370), cleaved PARP (c-PARP; Cell Signaling Technology, catalog no. 5625), BIM (Cell Signaling Technology, catalog no. 2933), phospho-MEK (p-MEK; Cell Signaling Technology, catalog no. 9154), phospho-RSK3 (p-RSK3; Cell Signaling Technology, catalog no. 9348), NRAS (Proteintech, catalog no. 10724-1-AP), HRAS (Proteintech, catalog no. 18925–1-AP), KRAS (Proteintech, catalog no. 12063-1-AP), phospho-RB (p-RB; Cell Signaling Technology, catalog no. 8516), cyclin D1 (Cell Signaling Technology, catalog no. 2978), phospho-CRAF (p-CRAF; Cell Signaling Technology, catalog no. 9427), phospho-CSF1R (Cell Signaling Technology, catalog no. 3155), phospho-EGFR (Cell Signaling Technology, catalog no. 2234), phospho-GAB1 (Cell Signaling, Technology, catalog no. 3233), phospho-AKT (Cell Signaling Technology, catalog no. 4060), phospho-PLCγ (Cell Signaling Technology, catalog no. 2821), phospho-STAT3 (Cell Signaling Technology, catalog no. 9145), Tubulin (Cell Signaling Technology, catalog no. 3873), and Actin (Cell Signaling Technology, catalog no. 3700).

The siRNA SMARTpool targeting HRAS (Dharmacon, L-004142-00) and individual siRNA-targeting NRAS (Dharmacon, J-003919-08) were transfected into NCI-H1373 cells using DharmaFECT 1 (Dharmacon, T-2001-02) following the manufacturer's instructions. Three days after transfection, NCI-H1373 cells were treated with compounds at the indicated concentrations and duration before harvesting for immunoblot analysis.

After isolation, CD3⁺ cells were frozen down in Recovery Cell Culture Freezing Medium (Thermo Fisher Scientific, catalog no. 12648010) and stored at −80°C. CD14⁺ cells were resuspended in RPMI media and seeded in 96-well plates (10,000 cells/well in 80 μL media, and M-CSF were supplemented into each well in 10 μL media at 450 ng/mL as indicated). A total of 10 mmol/L compounds were then added to each well by Tecan D300e digital dispenser to achieve indicated concentrations. Three days after seeding, M-CSF were supplemented again into each well in another 10 μL medium at 500 ng/mL. Four days after seeding, CD3⁺ cells were thawed and resuspended in RPMI media, and 25,000 CD3⁺ cells in 80 μL media were added into each well of the CD14⁺ cells. Twenty-four hours later, Dynabeads Human T-Activator CD3/CD28 beads (Thermo Fisher Scientific, catalog no. 11132D) in 20 μL media were added to each well. After 6 hours, the supernatant were harvested and IFNγ and IL2 levels were analyzed using the V-PLEX Human IFNγ Kit (MSD, catalog no. K151QOD-2) and V-PLEX Human IL2 Kit (MSD, catalog no. K151QQD-2), respectively.

All animal studies were performed in accordance with the NIH Guide for the Care and Use of Laboratory Animals and the Novartis Institutes for Biomedical Research Animal Care and Use Committee guidelines. Female athymic nude mice (8–12 weeks of age) were inoculated with HT-29 cells and LU-99 cells [5 million cells in 200 μL suspension containing 50% Matrigel (BD Biosciences) in Hank's Balanced Salt Solution] subcutaneously into the right superciliary region. Mice were monitored twice a week and tumors were measured and calculated by length × width²/2. Once tumors reached roughly 250 mm³, mice were randomly assigned into different groups to receive treatments at the indicated regimen. Tumor size and body weight were continuously measured at indicated frequencies. The high-throughput in vivo compound profiling with PDX models was performed as described previously (22).

Seven days after dosing as described, tumor tissues were put into gentleMACS C tubes (MACS Miltenyi Biotec) containing 2 mL RPMI1640 and minced into fine pieces (∼1 mm³) using scissors. Another 7 mL warmed RPMI1640 and 1 mL Liberase TM/DNase I stock solution [0.17 mg/mL Liberase TM (Roche, catalog no. 5401127001), 14 U/mL DNase I (Roche, catalog no. 4716728001)] were added into C tubes. Tightly closed C tubes were attached upside down onto the sleeve of the gentleMACS Dissociator and ran through Program h_impTumor_01. Next, C tubes were incubated at 37°C for 5 minutes and quenched with 1:10 dilution of FBS. Then program h_impTumor_02 was ran twice on C tubes. After termination of the program, tumor tissue cell suspensions were applied to 70 μm cell strainers placed on 50 mL conical tubes. The collection tubes containing the suspensions were centrifuged (1,500 rpm, 10 minutes, 4°C) and the cells were counted. Pellets were resuspended in flow cytometry buffer and subjected to immunostaining. After gating CD45⁺ live immune singlets, Ly6G⁻, Ly6C⁺, CD11b⁺, F4/80⁺, and MHCII⁻ gating was used for identifying M2 macrophages. The following antibodies were used: CD45 (e-Bioscience, catalog no. 56-0451-82), Ly6G (e-Bioscience, catalog no. 11-5931-82), Ly6C (e-Bioscience, catalog no. 12-5932-82), CD11b (e-Bioscience, catalog no. 25-0112-82), F4/80 (e-Bioscience, catalog no. 45-4801-82), MHC-II (e-Bioscience, catalog no. 47-5321-82), and Live/Dead Yellow (Invitrogen, catalog no. L34968).

Statistical significance was determined using GraphPad Prism 8 software by paired Student t test or Mann–Whitney test. Significance is reported at three levels: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Despite the remarkable clinical efficacy of EGFR TKI in treating patients with EGFR-mutant non–small cell lung cancer (NSCLC; ref. 23), acquired resistance inevitably occurs in the majority of patients. One common mechanism of resistance is the acquisition of gatekeeper mutations in EGFR, such as T790M and C797S (4). Because SHP2 mediates RAS activation downstream of EGFR, the efficacy of SHP2 inhibitors is not expected to be affected by the EGFR T790M and C797S mutations, even if they cooccur on the same DNA strand (in cis) as seen in some patients who relapsed on treatment with the third-generation EGFR TKI osimertinib (4). We first tested whether TNO155 is broadly efficacious in EGFR-mutant NSCLC cell lines. Among the eight cell lines we tested, six were sensitive to the mutant-selective third-generation EGFR TKI, nazartinib (EGF816; ref. 24) and TNO155 exhibited activity in three cell lines with IC₅₀ values lower than 1.5 μmol/L (NCI-H3255, HCC827, and PC9; Supplementary Table S1). The differential responses to TNO155 and nazartinib suggest that the dependency on SHP2 signaling of EGFR-mutant cells varies, possibly related to the cell context. Next, we took advantage of the previously described PC-9 cells (EGFR^ex19del) overexpressing EGFR^T790M/C797S (20) to test whether TNO155 remains effective against activity of EGFR-harboring TKI-resistant gatekeeper mutations. As expected, expression of EGFR^T790M/C797S led to resistance to both the first-generation EGFR TKI erlotinib (IC₅₀: PC-9 = 4.98 nmol/L, PC-9 EGFR^T790M/C797S > 1 μmol/L) and nazartinib (IC₅₀: PC-9 = 0.85 nmol/L, PC-9 EGFR^T790M/C797S > 1 μmol/L) but had no significant effect on the sensitivity to TNO155 (IC₅₀: PC-9 = 1.56 μmol/L, PC-9 EGFR^T790M/C797S = 1.38 μmol/L; Fig. 1A).

Another common acquired resistance mechanism to EGFR TKI is alternative RTK activation such as amplification of ERBB2 or MET, which were found in approximately 15% of the patients who relapsed on the first- or the third-generation EGFR TKI (25, 26). Because SHP2 also mediates RAS activation by HER2 and MET, TNO155 is expected to overcome this type of resistance mechanism. We tested this hypothesis using the previously published gefitinib-resistant HCC827 cells with MET amplification (HCC827-GR; ref. 3). As shown in Fig. 1B, HCC827-GR cells were cross-resistant to nazartinib as expected (IC₅₀: HCC827 = 7.55 nmol/L, HCC827-GR > 10 μmol/L) but remained sensitive to TNO155 (IC₅₀: HCC827 = 0.77 μmol/L, HCC827-GR = 1.38 μmol/L), despite the possible additional activation of SHP2 by MET amplification. Moreover, TNO155 only inhibited the RAS–MAPK pathway downstream of RTK signaling in HCC827 cells without any acute effect on the PI3K–AKT, PLCγ—PKC, and JAK–STAT pathways, which were also suppressed by EGFR TKI gefitinib (Fig. 1C). Taken together, TNO155 is efficacious in a subset of EGFR-mutant NSCLC cell lines and may overcome two common resistance mechanisms to EGFR TKIs, secondary EGFR mutations (T790M and/or C797S) and bypass activation of alternative RTKs.

Given the variable sensitivity of EGFR-mutant NSCLC cells to TNO155, we tested the possibility of combining TNO155 with nazartinib to prevent the emergence of acquired resistance. Intriguingly, TNO155 and nazartinib exhibited strong synergy (synergy score > 2) in five out of the six nazartinib-sensitive cell lines tested, including two cell lines that are insensitive to TNO155 (PC-14 and NCI-H1975; Supplementary Table S1). The synergy of nazartinib and TNO155 in PC-14 and NCI-H1975 was observed across a wide range of nazartinib concentrations and at low concentrations of TNO155 (e.g., 0.124 μmol/L), which lack single-agent activity in both lines (Fig. 1D, Loewe excess grids), suggesting that the contribution of TNO155 may result from the inhibition of alternative RTK signaling. Consistent with this hypothesis, in PC-14 cells, rebound of phospho-ERK1/2 (p-ERK) levels was observed after 24 hours treatment with 0.1 μmol/L nazartinib, which could not be blocked by a higher dose of nazartinib (0.3 μmol/L; Fig. 1E). Similarly, TNO155 effectively reduced p-ERK levels at 4 hours but suffered a rebound at 24 hours, while the combination of TNO155 and nazartinib achieved sustained inhibition of ERK. The combination also induced a stronger apoptotic response than either of the single agents at 24 hours, as evidenced by increased levels of c-PARP and BIM (Fig. 1E).

Next, we evaluated the combination of TNO155 and osimertinib in six EGFR-mutant lung cancer patient-derived tumor models in a mouse clinical trial format as described previously (22). The single-agent dosing regimen of TNO155 in mice was optimized as 20 mg per kilogram body weight (mpk) twice a day (9). In mouse clinical trial combinations, TNO155 was dosed at 10 mpk twice daily to optimize tolerability with many combinations. As shown in Fig. 1F, TNO155 10 mpk twice daily overall had limited activity as seen in some EGFR-mutant cell lines in vitro (Supplementary Table S1) while osimertinib 10 mpk daily achieved tumor regression or stasis except in HXXTM 29666 and 29667. Combination benefit of osimertinib and TNO155 was seen in four out of the six models tested (Fig. 1F; HXXTM 2996, 29997, 29665, and 29670). In HXXTM29666, osimertinib transiently slowed tumor growth while TNO155 had a more durable response. The combination achieved near complete tumor regression (Fig. 1G). In HXXTM29667 and HXXTM29665, modest to strong antitumor activity by osimertinib was enhanced by TNO155 (Fig. 1G). These data suggest that TNO155 can overcome acquired resistance to EGFR TKIs and also enhance their efficacy, providing a strong rationale to explore this combination in the clinic.

Despite its success in treating BRAF^V600E/K patients with melanoma, the combination of BRAF (BRAFi) and MEK inhibitors (MEKi) had limited activity in BRAF^V600E/K colorectal cancer with a response rate of only 12% in a 43-patient study compared with approximately 65% in melanoma (27). A common mechanism driving this intrinsic resistance was thought to be the EGFR-mediated feedback reactivation of the MAPK pathway as reported in preclinical colorectal cancer models (28). However, the combination of BRAF (e.g., dabrafenib and encorafenib) and MEK inhibitors (e.g., trametinib and binimetinib) with EGFR blocking antibodies (e.g., cetuximab and panitumumab) only modestly improved the response rate (21%–26%; refs. 29, 30). One likely explanation for the limited improvement provided by EGFR inhibition is that other RTK(s) may also be feedback activated in some patients with colorectal cancer and contribute to the intrinsic resistance. To identify such RTKs in colorectal cancer models, we used the MEK inhibitor selumetinib to induce feedback activation and tested the ability of individual RTK inhibitors such as MET inhibitor capmatinib (INC280; ref. 31), FGFR inhibitor infigratinib (BGJ398; ref. 32) or SHP2 inhibitors to abolish such feedback activation as described in KRAS-mutant models (21). Because of the robust activity of BRAF^V600E, p-CRAF (S338) was used instead of phospho-MEK1/2 (p-MEK), or dabrafenib was added with selumetinib to clearly reveal the RTK-mediated p-MEK induction. This p-CRAF/p-MEK assay led to the identification of MET and FGFR as the feedback activated RTK in RKO and MDST8 cells, respectively (Supplementary Fig. S1A).

Next, we evaluated whether cotreatment with inhibitors of the feedback activated RTK can enhance the efficacy of the combination of dabrafenib and trametinib in HT-29 (EGFR-mediated; ref. 33), RKO (MET-mediated), and MDST8 (FGFR-mediated) cells in a colony formation assay and whether TNO155 can replace those RTK inhibitors. As shown in Fig. 2A, the combination of dabrafenib and trametinib at clinically achievable concentrations (10 nmol/L dabrafenib and 1 nmol/L trametinib) failed to fully inhibit the proliferation of all three cell lines. Erlotinib, capmatinib, or infigratinib at 1 μmol/L had no obvious effect as expected because of the downstream BRAF activation but dramatically enhanced the efficacy of the dabrafenib plus trametinib combination in HT-29, RKO, and MDST8 cells, respectively. TNO155 at 1 μmol/L was as effective as RTK inhibitors at enhancing the efficacy of dabrafenib plus trametinib in all three cell lines. Moreover, TNO155 and dabrafenib plus trametinib exhibited strong synergy (synergy score > 2) in all three BRAF-mutant colorectal cancer cell lines (Fig. 2B), comparable with the synergy scores of dabrafenib plus trametinib with capmatinib in RKO or with infigratinib in MDST8 (Supplementary Fig. S1B). Consistently, erlotinib had no synergy with dabrafenib plus trametinib in either RKO or MDST8 (Supplementary Fig. S1B).

We next examined whether the combination benefit was due to sustained inhibition of the MAPK pathway. As shown in Fig. 2C, p-ERK levels in all three cell lines were effectively reduced after 2-hour treatment with dabrafenib plus trametinib but rebounded by 24 hours. TNO155 or RTK inhibitor treatment had limited effects on the p-ERK levels but successfully prevented the pathway feedback activation following treatment with dabrafenib plus trametinib. These data suggest that TNO155 can be seen as almost a pan-RTK inhibitor to enhance the efficacy of dabrafenib plus trametinib in BRAF V600–mutant colorectal cancer, in which EGFR and other RTKs may be feedback activated and mediate intrinsic resistance to MAPKi.

We next evaluated the in vivo efficacy of the combination of MAPK and SHP2 inhibitors using nude mice bearing HT-29 xenografts. Dabrafenib plus trametinib at a clinically achievable dose (dabrafenib, 30 mpk daily; trametinib, 0.3 mpk daily), as well as TNO155 (20 mpk twice daily), only resulted in moderate tumor growth inhibition (Fig. 2D). The combination of dabrafenib plus trametinib and TNO155 (10 mpk, twice daily) was well tolerated as judged by stable mouse body weight and maintained tumor stasis for more than 40 days, comparable with the efficacy of the combination of dabrafenib plus trametinib and cetuximab (Fig. 2D). These data suggest the dabrafenib plus trametinib combination with TNO155 can be as effective as the combination with an EGFR inhibitor (EGFRi) in vivo.

KRAS^G12C-specific inhibitors have shown promising activity in patients with NSCLC (34, 35), while their activity in colorectal cancer is limited (36, 37). Combination therapies are needed for KRAS^G12C inhibitors (G12Ci) to improve patient responses in both colorectal cancer and NSCLC. Because G12Ci only binds to KRAS^G12C in its inactive GDP-bound state, TNO155 is an attractive combination partner to shift KRAS^G12C to the GDP-bound state by reducing KRAS GTP loading. In addition, TNO155 can also prevent pathway feedback activation mediated through wild-type (WT) KRAS, HRAS, and NRAS, which cannot be targeted by G12Ci.

To evaluate these two distinct but related hypotheses of combination benefit, we first examined the combination of TNO155 and the G12Ci Compound 12a (Cpd 12a; ref. 38) shortly after treatment when the pathway feedback activation is not expected to be robust. One-hour treatment with 0.5 μmol/L Cpd 12a in NCI-H2122 cells caused an electrophoretic mobility shift of KRAS^G12C protein as described previously (35), which was not observed following treatment with a 5-fold lower concentration of Cpd 12a or TNO155 (Fig. 3A). The combination with TNO155 further enhanced the KRAS^G12C mobility shift induced by 1 hour treatment of both 0.1 and 0.5 μmol/L Cpd 12a and translated to further inhibition of MEK and ERK at as early as 15 minutes of treatment (Fig. 3A). Next, we titrated Cpd 12a and TNO155 concentrations to enable the examination of TNO155 effects on preventing the pathway feedback activation without acute combination benefit. In NCI-H1373 NSCLC cells (Fig. 3B; KRAS^G12C/G12C), the initial reduction of p-MEK, p-ERK, and p-RSK3 levels after 2 hours treatment with either Cpd 12a or TNO155 fully rebounded after 24 hours. The rebound was prevented by the combination of Cpd 12a and TNO155, which was maintained for at least 48 hours (Fig. 3B). Similar observations were made in LIM2099 colorectal cancer cells (KRAS^G12C/G12C) at 24 hours of treatment and more importantly treatment with Cpd 12a at a 4-fold higher concentration did not reduce the rebound of p-ERK levels (Fig. 3B), consistent with the notion that the G12Ci-induced pathway feedback activation may be mediated through HRAS, NRAS, and WT KRAS (for KRAS^G12C/+ cells). To test this hypothesis, we knocked down HRAS and NRAS in NCI-H1373 cells using siRNA, which did not inhibit the MAPK pathway as expected, and slightly increased p-ERK levels possibly due to further activation of KRAS^G12C in the absence of competition by HRAS and NRAS (Fig. 3C). Knocking down HRAS and NRAS effectively prevented the rebound of p-ERK levels as seen in Fig. 3B following 24-hour treatment with Cpd 12a (lane 8 vs. lane 5), similar to the effect of combing Cpd 12a with TNO155 (lane 13; Fig. 3C). In addition, with TNO155 treatment, no additional reduction of p-ERK and p-RSK levels by knockdown of HRAS and/or NRAS was observed, consistent with the hypothesis that the effect of TNO155 is through HRAS and NRAS inhibition. Combination benefit of TNO155 and Cpd 12a on enhanced and sustained MAPK pathway inhibition was also observed in NCI-H2122 and SW837 cells with heterozygous KRAS^G12C that is more common in primary tumors (Supplementary Fig. S2A).

To determine whether the sustained pathway inhibition by the Cpd 12a and TNO155 combination leads to improved antiproliferative effects, we used the IncuCyte system to monitor cell confluency over 14 days of compound treatment in several KRAS^G12C cell lines (Fig. 3D; Supplementary Fig. S2B). In NCI-H1373 and LIM2099 cells, Cpd 12a at 0.2 μmol/L inhibited cell proliferation to various degrees while 0.3 μmol/L TNO155 treatment had little effect as expected for KRAS-mutant cells cultured as a monolayer (39). The combination led to enhanced inhibition of cell proliferation (Fig. 3D). Similar antiproliferation combination benefit was observed in three additional KRAS^G12C cell lines; in a fourth cell line tested, SW837, single-agent Cpd 12a exhibited antiproliferative effects too strong to allow discernment of a combination benefit (Supplementary Fig. S2B). We next examined whether the combination benefit of Cpd 12a and TNO155 is additive or synergistic and performed a combination dose matrix cell proliferation assay with seven KRAS^G12C cell lines (Fig. 3E; Supplementary Fig. S2C) including the five lines tested via the IncuCyte assay (Fig. 3D; Supplementary Fig. S2B). Cpd 12a and TNO155 exhibited strong synergy (synergy score > 2) in all seven lines across a wide range of concentrations of both Cpd 12a and TNO155 (Fig. 3E; Supplementary Fig. S2C).

In the three combinations described above, TNO155 was used to enhance the efficacy of inhibitors targeting oncogenic drivers (mutant EGFR, BRAF, and KRAS). SHP2 inhibitors such as SHP099 demonstrated remarkable activity against selected models of RTK-driven (6, 7) and KRAS-mutant cancers in vivo (39). However, SHP2 inhibitors alone often only achieved tumor stasis at best and their activity was variable across models. Therefore, we looked for agents that may enhance the efficacy of TNO155. Activation of the MAPK pathway increases the expression of cyclin D which enables CDK4/6 to promote the cell-cycle progression to S phase (40). CDK4/6 inhibitors such as ribociclib have shown combination benefit with both RTK and MAPK inhibitors (41). Therefore, we hypothesized that ribociclib may improve the efficacy of TNO155.

First, we explored the combination of TNO155 and ribociclib in EGFR-mutant NSCLC HCC827 cells via the colony formation assay (Fig. 4A). Both TNO155 and ribociclib had dose-dependent single-agent efficacy in HCC827 and the combination achieved greater efficacy in all concentrations tested. We then studied the effect of this combination on the cell cycle and the MAPK pathway (Fig. 4B). TNO155 prevented the cyclin D1 accumulation as a result of cell-cycle arrest following 48-hour ribociclib treatment and modestly further reduced p-RB levels (lane 4 vs. lane 6), consistent with the report on the combination benefit of nazartinib and ribociclib (41). TNO155 modestly inhibited the MAPK pathway following 48 hours treatment as evidenced by the mild reduction of p-RSK3 levels, possibly due to pathway reactivation. As expected, ribociclib did not enhance the MAPK pathway inhibition by TNO155. We then expanded our in vitro combination efficacy study to additional NSCLC and colorectal cancer cell lines (Fig. 4C). Single-agent activity and combination benefit of TNO155 and ribociclib was also observed in HCC366 (KRAS^WT), A-427 (KRAS^G12D/+), and NCI-H747 (KRAS^G13D/+) but not in LoVo (KRAS^G13D/+) or SW948 (KRAS^Q61L/+).

We next evaluated the TNO155 plus ribociclib combination (TNO155: 10 mpk, twice daily; ribociclib: 75 mpk, daily; combination dose selected by tolerability study) using a large panel of NSCLC and colorectal cancer patient-derived tumor models in a mouse clinical trial format (Fig. 4D), as described in Fig. 1E and previous studies (22). Ribociclib single-agent activity was evaluated previously (24 NSCLC and 34 colorectal cancer models) with largely overlapping models (19 overlapping models for NSCLC and 18 for colorectal cancer) as the combination study (28 NSCLC and 27 colorectal cancer models) and at a much higher daily dose (250 mpk vs. 75 mpk). For both NSCLC and BRAF^WT colorectal cancer models, tumors in the combination study (Vehicle 2) seemed to grow faster than they did in the prior study (Vehicle 1). We hypothesize that ribociclib would have been less efficacious if the study was done along with the TNO155 and ribociclib combination at the same ribociclib dose. Therefore, the combination group would likely have greater efficacy in both NSCLC and BRAF WT colorectal cancer models than either of the single agents, although this conclusion is only supported by the statistical significance of the difference between the combination and the TNO155 treatment group. The efficacy of TNO155 plus ribociclib was also comparable with that of TNO155 plus trametinib, a synergistic combination in KRAS-mutant (21, 33) and RTK-driven cancers (42) but much less tolerated. The variable responses of TNO155 plus ribociclib we observed in NSCLC and BRAF^WT colorectal cancer prompted us to further examine whether the combination is more effective in KRAS-mutant or WT cancers. Among the NSCLC models, the combination of TNO155 and ribociclib achieved greater tumor responses in the KRAS^WT group than the mutant group (average: −21% vs. −8%). Interestingly, the improvement of the combination over TNO155 alone was only significant in KRAS mutant but not in KRAS^WT BRAF^WT colorectal cancer models. Importantly, in models with either genetic background or lineage, TNO155 plus ribociclib was not inferior to TNO155 plus trametinib at the tolerated regimen.

We further compared the combination of TNO155 plus ribociclib versus TNO155 plus trametinib in the KRAS^G12C NSCLC cell line xenograft Lu-99 (Fig. 4E). In this study, the TNO155 dose was maintained at 20 mpk twice daily and the trametinib dose was further reduced to 0.075 mpk. TNO155 plus ribociclib achieved tumor stasis and was tolerated by mice as judged by body weight. In contrast, TNO155 plus trametinib was less efficacious and caused significant body weight loss. Taken together, these data suggest that the combination of TNO155 and ribociclib may be better tolerated and equally efficacious compared with TNO155 plus trametinib in both KRAS-mutant and WT cancers.

In addition to driving tumor cell growth, RTK signaling also plays an important role in regulating immune cells. For example, M-CSF or CSF1 secreted by tumors can reprogram TAM to an immune-suppressive state (M2) by activating the CSF1 receptor (CSF1R) signaling (14). SHP2 is also implicated in the RAS activation downstream of CSF1R signaling. Indeed, treatment with TNO155 abolished ERK phosphorylation in an acute myelomonocytic leukemia cell line, GDM-1, bearing a CSF1R^Y571D-activating mutation (43), similar to the effect of trametinib or the CSF1R kinase inhibitor BLZ945 (ref. 44; Supplementary Fig. S3A). However, it remains unclear whether the RAS–MAPK pathway is the most critical effector pathway for the differentiation or the immunosuppressive properties of M2 TAMs. Tumor macrophages mainly derive from circulating monocytes in the blood (45). The M2-like macrophages can be generated in vitro by stimulating CD14⁺ monocytes isolated from human peripheral blood with M-CSF, which exhibit T-cell suppressive phenotypes in coculture systems (45, 46). As shown in Supplementary Fig. S3B, both the survival and the proliferation of CD14⁺ monocytes are dependent on M-CSF stimulation. TNO155 effectively blocked the M-CSF–stimulated proliferation of CD14⁺ monocytes, comparable with the effect of BLZ945 (IC₅₀, TNO155 = ∼0.05 μmol/L, BLZ945 = ∼0.3 μmol/L; Fig. 5A). As expected, M-CSF treatment activated the MAPK pathway as evidenced by elevated levels of p-ERK, which was blocked by treatment with either TNO155 or trametinib (Fig. 5B).

We next examined whether TNO155 can abolish the T-cell suppression by those M-CSF–differentiated macrophages in a coculture assay (Supplementary Fig. S3C). We used anti-CD3 plus anti-CD28 antibodies to activate T cells and measured the levels of IFNγ and IL2 in the culture media as a surrogate readout for T-cell activity. As shown in Fig. 5C, M-CSF–differentiated monocyte-derived macrophages reduced both IFNγ and IL2 levels in the coculture media, suggesting T-cell suppression, while treatment with TNO155 or BLZ945, but not trametinib, restored IFNγ and IL2 secretion. This functional rescue by TNO155 or BLZ945 is likely a result of blocking the M-CSF–stimulated proliferation of monocytes as revealed by a viability assay of the monocyte only groups performed in parallel with the coculture assay (Supplementary Fig. S3D). Trametinib at 50 nmol/L also suppressed the M-CSF–stimulated proliferation of monocytes but failed to reverse the T-cell suppression (Supplementary Fig. S3D; Fig. 5C), likely due to its inhibition of T-cell activity as reported previously (47).

Combination benefit of another SHP2 inhibitor, RMC-4550, and anti-PD-L1 antibody was reported in the CT-26 syngeneic colon tumor model (48). We also observed such combination benefit of TNO155 with an anti-PD-1 antibody in the MC38 syngeneic colon tumor model that is resistant to TNO155 when grown in immune-deficient mice (manuscript in revision), suggesting SHP2 inhibition in immune cells is largely responsible for the combinatorial antitumor activity. On the basis of our in vitro data (Fig. 5C), we hypothesize that the combination benefit in the MC38 model may be due to the inhibition of immunosuppressive M2 macrophages by TNO155 in vivo. Indeed, immunophenotyping analysis showed a significant decrease of M2 TAMs following seven days of TNO155 treatment (Fig. 5D; Supplementary Fig. S3E). PD-1 blockade had no inhibitory effect on M2 TAMs as expected. Intriguingly, there was a further reduction of TAMs in the TNO155 plus anti-PD-1 antibody combination group, which was also observed in the CT-26 syngeneic colon tumor model in the RMC-4550 and anti-PD-L1 antibody combination (48). The synergistic mechanism on reduction of M2 macrophages by coinhibition of SHP2 and PD-1 signaling remains to be elucidated. Taken together, in addition to tumor intrinsic effects, TNO155 also has immunomodulatory effects, particularly inhibiting immune-suppressive macrophages in the tumor, and can enhance the efficacy of PD-1 blockade.

In this article, we showed in vitro and in vivo combination benefit and elucidated the likely underlying synergistic mechanisms of five different combinations involving TNO155, demonstrating the broad utility of TNO155 as a combination partner across multiple cancer types with different genetic drivers. This broad utility is largely attributed to the ability of TNO155 to block both signaling from multiple RTKs and the RTK-mediated feedback pathway activation and to reverse immune suppression in tumor microenvironment (6, 13, 14, 21, 33).

TNO155 can overcome two common mechanisms of resistance expected for osimertinib and nazartinib. TNO155 can also enhance the efficacy of nazartinib by preventing MAPK pathway reactivation from adaptive activation of other RTKs and eliminating a subset of EGFRi-tolerant cells where RTK drives MAPK pathway reactivation (49), potentially deepening response and preventing subsequent emergence of resistance. SHP2 inhibitors are often considered as a pan-RTK inhibitor but in most cases we tested, TNO155 only inhibits the RAS–MAPK pathway downstream of RTK signaling without any acute effect on the PI3K–AKT, PLCγ–PKC, and JAK–STAT pathways (Fig. 1C). These pathways can also play important roles in cancer cell survival and proliferation, which probably explains why single-agent SHP2 inhibitors are not always as effective as RTK inhibitors in cells with activating mutations of RTK. In addition to mutations, RTKs can also be activated by ligands (autocrine or paracrine) and by downregulation of negative feedback regulators such as SPRY family proteins (50). We speculate that the feedback activation of RTKs is more susceptible to SHP2 inhibition as the RAS–MAPK pathway seems to be preferentially feedback activated over other effector pathways (21). Therefore, preventing RTK-mediated feedback activation of the RAS–MAPK pathway is likely a more effective use of TNO155 than targeting RTK-mutated cancers. Indeed, in BRAF-mutant colorectal cancer cells, dabrafenib plus trametinib showed similar combination benefits with TNO155 compared with RTK inhibitors (Supplementary Fig. S1B; Fig. 2D). SHP2 inhibitors also have advantages over RTK inhibitors due to the technical challenge to identify the various and/or even multiple feedback activated RTKs in patient tumors.

SHP2 inhibitors may also benefit from vertical combinations targeting the RAS–MAPK pathway due to the extensive feedback regulation of this pathway. MEK inhibitors were previously identified as synergistic combination partners for SHP2 inhibitors in KRAS-mutant (21, 33) and RTK-mutated cancers (42). Such a combination is being explored in the clinic (ClinicalTrials.gov identifier: NCT03989115); however, the tolerability of this combination is a concern. In this study, we showed that the CDK4/6 inhibitor ribociclib can serve as an alternative to trametinib to enhance the efficacy of TNO155 in both KRAS-mutant and WT cancers (Fig. 4D), suggesting targeting downstream effectors of ERK signaling such as cell-cycle regulators can be as effective as targeting MEK when combined with TNO155. More importantly, TNO155 combined with ribociclib showed comparable efficacy and improved tolerability than combined with trametinib (Fig. 4E).

The synergy of TNO155 and KRAS G12Ci is the strongest among the TNO155 combinations we examined in this study with synergy scores greater than 5 in more than half of the cell lines tested. This could be attributed to the additional synergy from TNO155 enhancing G12Ci target engagement besides preventing pathway feedback activation (common in all those combinations). TNO155 treatment may shift the KRAS to the GDP-bound state, to which G12Ci bind. Alternative approaches to enrich the KRAS-GDP pool include using a RTK inhibitor, which can only block one RTK, or a SOS1 inhibitor, which cannot block other RAS guanine nucleotide exchange factors. Given the large unmet medical need of patients with KRAS^G12C cancer and the encouraging clinical efficacy of both KRAS^G12C (36, 37) and SHP2 inhibitors (10), TNO155 plus KRAS G12Ci represents one of the most promising combinations for KRAS^G12C cancers to be pursued in the clinic. The mutant selectivity of G12Ci may also allow the addition of a third combination partner such as ribociclib. In addition, it was reported that G12Ci may induce a proinflammatory tumor microenvironment and synergizes with checkpoint inhibitors (34). Therefore, a triplet combination of TNO155, G12Ci, and anti-PD-1 antibody may provide a multifaceted and highly synergistic regimen against KRAS^G12C cancers.

A major challenge in combination therapies is tolerability. Many efficacious combinations in preclinical models cannot achieve similar exposures in patients due to overlapping toxicities of the involved agents. The combinations we investigated here have a good chance to be tolerated in patients for distinct reasons. First, the third-generation EGFR TKIs and the KRAS^G12C inhibitors are mutant selective, sparing the WT EGFR or KRAS, respectively, in normal cells. They are well tolerated as single agents and suitable for combining with agents with acceptable tolerability profiles such as a SHP2 inhibitor (8, 10). Second, type 1 and 1.5 BRAF inhibitors including dabrafenib cause paradoxical pathway activation in normal cells with WT BRAF and improves the tolerability of MAPK inhibitors such as trametinib. In addition, the triplet combination of BRAFi, MEKi, and EGFRi was shown to be efficacious and tolerated in human (30) and we anticipate the combination of dabrafenib, trametinib, and TNO155 to be similarly tolerated at efficacious doses. Third, anti-PD-1 antibodies exclusively target the exhausted T cells and have exhibited tolerability with a variety of targeted therapies. Finally, although ribociclib is not mutant nor tissue selective as those above four agents, we have demonstrated its combination tolerability with TNO155 in mice where severe body weight loss was only observed when TNO155 was combined with trametinib (Fig. 4E).

With the synergistic combination activity observed in preclinical models and promising tolerability of these combinations described in this article, several clinical trials have been launched: TNO155 plus spartalizumab in NSCLC (ClinicalTrials.gov identifier: NCT04000529), TNO155 plus MRTX849 in KRAS^G12C NSCLC and colorectal cancer (ClinicalTrials.gov identifier: NCT04330664), TNO155 plus nazartinib in EGFR-mutant NSCLC (ClinicalTrials.gov identifier: NCT03114319), and TNO155 plus ribociclib in KRAS WT NSCLC and KRAS-mutant colorectal cancer (ClinicalTrials.gov identifier: NCT04000529). The preclinical data, particularly the mouse clinical trial data, guided our patient selection strategy to prioritize patients with KRAS WT NSCLC and patients with KRAS-mutant colorectal cancer for the TNO155 plus ribociclib combination, despite the current lack of mechanistic understanding of this interesting response differences. In summary, the preclinical data described here have largely influenced the clinical development strategy for TNO155, a first-in-class and versatile molecule that can be used in various combinations for the management of both RTK-activated and KRAS/BRAF-mutant cancers.

H.-X. Hao reports a patent for TNO155 combination with ribociclib pending and a patent for TNO155 combination with spartalizumab pending. C. Liu reports a patent for TNO155 and ribociclib combination pending and a patent for TNO155 and spartalizumab combination pending. S. Goldoni reports a patent for TNO155 combination with spartalizumab pending. W.D. Hastings reports a patent for TNO155 combination with spartalizumab pending and is a full-time employee of Novartis. S.E. Moody is an employee of Novartis Institutes for BioMedical Research and has a patent for TNO155 combination with EGFR inhibitors pending. M.J. LaMarche reports personal fees from Novartis during the conduct of the study and personal fees from Novartis outside the submitted work, and has multiple patents and applications pending and issued. J.A. Engelman is an employee of Novartis and has equity in Novartis. T.J. Abrams is an employee and shareholder of Novartis. G. Caponigro is an employee and shareholder of the Novartis Institutes for BioMedical Research. No disclosures were reported by the other authors.

H.-X. Hao: Conceptualization, formal analysis, supervision, investigation, visualization, writing-original draft, project administration, writing-review and editing. C. Liu: Conceptualization, data curation, formal analysis, investigation, visualization, writing-original draft, writing-review and editing. H. Lu: Conceptualization, data curation, formal analysis, investigation, visualization, writing-original draft, writing-review and editing. H. Wang: Data curation, visualization. A. Loo: Resources, data curation. X. Zhang: Data curation. G. Yang: Data curation. C. Kowal: Data curation. S. Delach: Data curation. Y. Wang: Data curation. S. Goldoni: Supervision. W. Hastings: Supervision. K. Wong: Resources, methodology. H. Gao: Resources. M.J. Meyer: Supervision. S.E. Moody: Supervision, investigation, writing-original draft, writing-review and editing. M.J. LaMarche: Resources. J.A. Engelman: Resources, supervision. J.A. Williams: Resources, supervision. P.S. Hammerman: Supervision, writing-original draft. T.J. Abrams: Supervision, writing-review and editing. M. Mohseni: Formal analysis, supervision, visualization. G. Caponigro: Conceptualization, resources, supervision.

This research was funded by Novartis Institutes for BioMedical Research.

The authors thank the following Novartis colleagues for their technical advice, discussion of data, and suggestions and comments on the study and this manuscript: Ying-Nan Chen, Matt Shirely, Matthew J. Niederst, David Kodack, Saskia M. Brachmann, Ralph Tiedt, Nadia Hassounah, Jennifer Mataraza, Tina Yuan, Vesselina Cooke, Darrin Stuart, Roberto Velazquez, Michael Fleming, Joanne Lim, Pushpa Jayarman, Eugene Tan, and Scott Loftus-Reid. They also thank the Novartis team who performed the high-throughput compound profiling of PDX in mouse and data analysis.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.