Tumor Intrinsic Efficacy by SHP2 and RTK Inhibitors in KRAS-Mutant Cancers

Protein tyrosine phosphorylation plays a central role in normal cell biology and pathogenesis of diseases including cancer (1). A large number of tyrosine kinases and several phosphatases are mutated in various types of cancer and function as oncogenes (2). The Src homology-2 (SH2) domain-containing phosphatase 2 (SHP2), encoded by PTPN11, was identified as the first protein tyrosine phosphatase (PTP) oncogene (3). SHP2 contains two SH2 domains (N-SH2 and C-SH2), a PTP catalytic domain and a C-terminal proline-rich motif containing tyrosyl phosphorylation sites (Y542 and Y580; ref. 4). Under basal conditions, intramolecular interactions between the N-SH2 and PTP domains impedes the accessibility of substrates, thus suppressing the catalytic activity of SHP2 (5). Upon RTK activation by upstream growth factors or cytokines, the autoinhibitory interactions of SHP2 are disrupted leading to its activation (6). In RTK-activated cells (RTK, e.g., EGFR, FGFR, VEGFR) or KRAS-amplified cells, SHP2 transduces the upstream signal with its phosphatase activity through guanine nucleotide exchange factors, including SOS1 (4, 7). This in turn promotes activation of KRAS by GDP to GTP exchange, and subsequent activation of the RAS–MAPK pathway resulting in pro-proliferative and prosurvival signals (4, 6, 8, 9).

SHP2 has been shown to be important for many biological functions that are dependent on upstream growth factor signaling including angiogenesis and differentiation and proliferation of tissues (10–14). Inhibition of SHP2, using a selective allosteric inhibitor, SHP099, has revealed therapeutic promise in cancers that are dependent on RTKs (15–17). Upstream signaling from RTKs, such as EGFR, can be dispensable in cells with constitutively active KRAS mutations, as evidenced by the lack of clinical response to the anti-EGFR antibody, cetuximab, in KRAS-mutant colorectal cancer patients (18). However, not all KRAS mutations impair intrinsic GTPase activity equally, and some KRAS-mutant cancers may still require robust upstream RTK activity to maintain the GTP-loaded active state. Recent reports have shown that the intrinsic hydrolysis of GTP to GDP in KRAS^G12C is higher than other KRAS mutations, almost approaching that of wild-type KRAS (19–21) and that inhibition of SHP2 can impact mutant KRAS signaling in genetically engineered mouse models and in cancer cell lines (17, 22–24). In addition to SHP2′s function of activating RAS downstream of RTKs, there is growing, yet confounding evidence for an immunomodulatory function of SHP2. SHP2 has been implicated in suppressing T-cell activation by dephosphorylating PD1-recruited CD28 (25), although it has also been shown to be dispensable for T-cell exhaustion and for anti-tumor activity by PD1 blockade in CD4⁺ T-cell specific knockout mouse models (26). Taken together, inhibition of SHP2 may exert antitumor effects in both tumor cell autonomous and nontumor cell autonomous manners.

In this report, we provide further evidence that inhibition of SHP2 or RTKs is effective at treating a variety of KRAS mutant cancers that are dependent on upstream growth factor signaling, including some KRAS G13D and Q61H mutants.

The following inhibitors were synthesized and structurally verified by NMR/LC-MS at Novartis Institute of Biomedical Research (Supplementary Materials): SHP2 inhibitor/SHP099 (15, 16), MEK1/2 inhibitor/trametinib (27), pan-VEGF/PDGFR/FGFR inhibitor dovitinib (28), VEGFR2 inhibitor/BFH772 (29), MET inhibitor/capmatinib (30), FGFR1-3 inhibitor/BGJ398 (31), EGFR-inhibitor/erlotinib, HER2 and EGFR-inhibitor/lapatinib. Cetuximab antibody was produced by Eli Lilly (catalog # NDC-66733-958-23).

Cell lines evaluated were obtained from Novartis' CCLE collection (32) and were tested to be free of Mycoplasma and murine viruses in the IMPACT1 PCR assay panel (RADIL, University of Missouri, Columbia, MO). These cells were maintained in ATCC specified medium at 37°C in a humidified atmosphere containing 5% carbon dioxide. In general, split ratios for subculturing consisted of 1:4 and 1:6 in T-225 flasks (Corning) and 1:4 and 1:6 in 3 bilayer cell factories (Nunc Cell Factory) and then 1:4 in 10-layer cell factories (Nunc Cell Factory). Cells were harvested at 85% to 95% confluence by treating with 0.25% trypsin-EDTA (Gibco, catalog #14175-095), and centrifuged at 1,200 RPM for 5 minutes at 4°C.

MIA PaCa-2_SHP2-KO clone was generated by CRISPR technology as previously reported (33). MIA PaCa-2_SHP2-KO_SHP2^T253M/Q257L cells were generated by infection of above MIA PaCa2_SHP2-KO clone with lenti-virus packaged using the pLKO-Trex-SBP-SHP2-T253M-Q257L plasmid (15) and selection in TET-free media containing 1 mg/mL G418. MIA PaCa-2_SHP2-shRNA cells were generated by infection of MIA PaCa-2 cells with retro-virus packaged using the pRSIUP-U6Tet-SHP2-shRNA-UbiC-TetRep-2A plasmid (34) and selection in TET-free media containing 3 μg/mL puromycin.

Mice were handled in accordance to the ILAR Guide for the Care and Use of Laboratory Animals in an AAALAC-accredited facility. Cell lines were confirmed to be free of mycoplasma and mouse viruses before use, cultured in medium according to ATCC guidelines then resuspended in ice-cold Hank's Balanced Salt Solution (HBSS; Gibco, catalog # 25200-056) containing 50% Matrigel (Corning, catalog #354234). Athymic female nude mice were inoculated subcutaneously into the right flank. Tumor-bearing mice were randomized into treatment groups once tumor volumes reached approximately 250 mm³. For the dox-inducible SHP2-shRNA xenograft model, mice were fed ad-libitum with chow containing doxycycline (Scott Pharma Solutions, catalog #1813539).

HUVEC cells were seeded into 6-well plates (0.2 × 10⁶) overnight in endothelial cell growth media (PromoCell, #C-22010) with 10% FBS. Following incubation, cells were treated with the respective compounds. After incubation, cells were harvested and diluted to 0.2 × 10⁵ cells/mL in endothelial cell basal media (PromoCell C-22010) plus 10% FBS and 72 μL of the suspension plus 8 μL of endothelial cell growth supplement (PromoCell C-39215) was plated within 96-well plates (Ibidi, #89646) coated with Matrigel (Millipore #ECM625). Brightfield images were captured (EVOS microscope, Thermo Fisher Scientific) 4 hours after incubation at 37°C, and tubes were quantified and analyzed using Image J (http://rsbweb.nih.gov/ij/).

For 2D proliferation assays, 500 to 1,500 cells per well were seeded into 96-well plates (Corning #3904) in 80 μL media per well. For 3D proliferation assays, 3,000 to 4,500 cells per well were seeded into low attachment 96-well plates (Greiner Bio-one, #655976) in 80 μL media alone or with 2% or 20% Matrigel (Corning, #356237). Twenty-four hours after seeding, cells were treated with 20 μL compounds at different final concentrations as indicated. After 6 days of incubation, 100 μL CellTiter-Glo reagent (Promega #G7573 for 2D; #G9683 for 3D) was added to each well and cell viability was determined according to the manufacturer's instruction. To determine the IC₅₀ values, data were fitted using the dose response algorithm in GraphPad Prism as Y = Bottom + (Top-Bottom)/(1+10⁁((X-LogEC50))), in which Top and Bottom are plateaus in the units of the y-axis and EC₅₀ is the concentration of inhibitor that gives a response half way between Bottom and Top. For details on the soft agar colony formation methods and 2D and 3D screen with SHP099 and the RTK inhibitor cocktail, please see Supplementary Materials.

MIA PaCa-2 spheroids were generated by plating at 5,000 cells/well of p-HEMA (Sigma, cat. # 192066) coated 384 microwell Elplasia plate (Kuraray cat. # SQ 200 100 384) according to the manufacturer's protocol. For 2D samples, subconfluent cells were seeded at 750 cells/well of a 384 well micro clear plate (Greiner cat. # 781091). Next day, cells were treated with SHP099 at 20 μmol/L, and DMSO control. At 3 and 19 hours posttreatment, 2D and 3D cultured cells were lysed with 500 or 50 μL, respectively, per well of lysis buffer with DTT. RNA was extracted according to the manufacturer's protocols within the RNeasy Kit (Qiagen cat. # 74134). Quantification and quality of RNA assessed by the Agilent 2100 Bioanalyzer (Agilent cat. # G2938C): 2100 expert Eukaryote Total RNA Nano Chip (Agilent cat. # 5067-1511). For transcriptome sequence and analysis details, please see Supplementary Materials. Gene expression datasets are available via NCBI BioProject accession number PRJNA558508.

The following primary antibodies were used: phospho-ERK (Cell Signaling Technology #4370), ERK (Cell Signaling Technology #4695), phospho-RSK3 (Cell Signaling Technology #9348), phospho-AKT (Cell Signaling Technology #4060), AKT (Cell Signaling Technology #9272), p-SHP2 (Cell Signaling Technology #3751), SHP2 (Cell Signaling Technology #3397), phospho-MET (Cell Signaling Technology #3077), phospho-STAT3 (Cell Signaling Technology #9145), and tubulin (Cell Signaling Technology #3873). Cells (2–7.5 × 10⁵) in 2 mL growth media were seeded in 6-well plates (Corning, #3506 for 2D, #3471 for 3D with ultralow binding). The 3D growth media contains 2% Matrigel (Corning, #356237). After one day (for 2D) or three days (for 3D, for spheroid formation), cells were treated with inhibitors at the indicated concentrations and duration. Cells were lysed on ice in RIPA buffer (Boston Bioproduct #BP-115) supplemented with 1 mmol/L EDTA and Halt Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific #1861281). Lysates were centrifuged at 14,000 rpm at 4°C for 15 minutes and protein concentrations were determined using BCA protein assay (Thermo Fisher Scientific). Equal amount of protein was separated by electrophoresis in NuPAGE 4%–12% Bis-Tris gel (Thermo Fisher Scientific #WG1402BX10), and transferred to nitrocellulose membranes (Bio-Rad, #1704159) for immunoblot with indicated primary antibodies. The bound primary antibodies were visualized using goat anti-rabbit IgG (H + L) secondary antibody conjugated with Alexa Fluor 700 and goat anti-mouse IgG secondary antibody conjugated with IRDye 800 CW and scanning with an Odyssey Infrared Imager System (Li-Cor).

It was previously reported that cancer cell lines sensitive to KRAS or NRAS depletion were refractory to SHP2 depletion in a pooled shRNA screen and RAS/RAF-mutant cell lines were not sensitive to SHP099 (15). The lack of sensitivity remains the case when KRAS mutation status is overlaid with sensitivity to either SHP2 (PTPN11) depletion in a pooled shRNA screen (Fig. 1A) or with sensitivity to SHP099 assessed across cell lines and cancer lineages (Fig. 1B).

Interestingly, a small number of KRAS-mutant models, displayed some sensitivity to SHP2 depletion by shRNA or inhibition by SHP099 (Supplementary Tables S1–S3). These data lightly suggest that activating mutants of KRAS may still be dependent on SHP2 or that other RAS isoforms (NRAS/HRAS or wild-type KRAS for heterozygotes) may contribute to MAPK pathway activation in those cells. Cancer cells have displayed differential sensitivity to targeted anti-cancer agents when grown in three-dimensions (3D) compared with the two-dimensional (2D) setting; among these are AKT/mTOR inhibitors (35, 36) and KRAS^G12C inhibitors (20). We therefore selected three KRAS G12 mutant cell lines that were not sensitive to SHP099 in 2D culture (MIA PaCa-2/KRAS^G12C, KP4/KRAS^G12D, Capan-2/KRAS^G12V) and tested them in a 3D culture setting with or without the addition of varying levels of Matrigel in the same growth media as in 2D (Fig. 1C). Strikingly, MIA PaCa-2 cells displayed greater sensitivity to SHP099 when cultured as spheroids with an IC₅₀ = 0.39 μmol/L (0% Matrigel), contrary to their resistance in the 2D setting (IC₅₀ > 30 μmol/L). Interestingly, the addition of either 2% or 20% Matrigel reduced the SHP099 efficacy against MIA PaCa-2 3D spheroids. In KP4 cells, there was a greater than 10-fold reduction in IC₅₀ for all 3D growth conditions (2D IC₅₀ >30 μmol/L; 3D 0% Matrigel IC₅₀ = 1.42 μmol/L; 3D 2% Matrigel IC₅₀ = 2.97 μmol/L; 3D 20% Matrigel IC₅₀ = 3.49 μmol/L). In Capan-2 cells, which only formed spheroids with the addition of Matrigel, there was a greater than 10-fold reduction in IC₅₀ in 3D (2D IC₅₀ >30 μmol/L; 3D 2% Matrigel IC₅₀ = 1.90 μmol/L; 3D 20% Matrigel IC₅₀ = 2.26 μmol/L). The 2D versus 3D sensitivity to the MEK inhibitor, trametinib, also displayed an IC₅₀ shift in the same manner as SHP099, with the exception of expected sensitivity to trametinib in 2D (IC₅₀ = 6.23–20.80 nmol/L; Fig. 1C). In contrast, T3M-4 cells (Fig. 1D) bearing the KRAS^Q61H mutation remained insensitive to SHP099 (IC₅₀ >30 μmol/L) even when cultured in 3D. To demonstrate that the SHP099 3D efficacy against KRAS G12–mutant cell lines is not a result of systematic sensitization to SHP099 from 2D to 3D growth, we also tested two RAS/RAF wild-type EGFR dependent cancer cell lines that are sensitive to SHP099 in 2D, Detroit-562, and KYSE520 (Supplementary Fig. S1A). There was either no IC₅₀ shift (Detroit-562) or a modest IC₅₀ reduction (∼3-fold, KYSE520) comparable with the EGFR inhibitor erlotinib.

To determine whether the SHP099 sensitivity of KRAS-mutant cell lines in 3D is a general phenomenon and to ascertain correlations to specific KRAS-mutant alleles, we examined a larger KRAS-mutant cell line panel to evaluate SHP099 under both 2D and 3D growth conditions (n = 14, Fig. 1E; Supplementary Table S4). For the majority of KRAS-G12–mutant cell lines (7/11) sensitivity to SHP099 in 3D was comparable with that of the KYSE520 in 2D or 3D. In summary, KRAS-mutant cancer cells, primarily those with G12 mutations, cultured as multicellular spheroids, were sensitive to SHP2 inhibition.

We next examined whether SHP099 had differential effects on the MAPK and/or PI3K pathways in KRAS-mutant cells cultured in 2D versus 3D. The 3D growth condition used was 2% Matrigel for all cell lines to enable 3D growth of those cell lines requiring Matrigel, as 20% Matrigel is too rigid for cell harvesting. p-ERK and its downstream marker p-RSK3 was overall similarly inhibited by 3 and 10 μmol/L SHP099 at 2 hours of treatment, in all three cell lines (MIA PaCa-2, KP4, and Capan-2) in both conditions, although not as complete as the levels achieved by 10 nmol/L trametinib (Fig. 2A). In addition, rebound of p-ERK and p-RSK3 at 24 hours following SHP099 treatment was similarly observed under both conditions, despite the different antiproliferation effects. Interestingly, the baseline levels of p-AKT, and to a lesser degree for p-ERK, were lower when all three KRAS-mutant cells were grown under 3D conditions. However, in T3M-4, baseline levels of p-AKT in 3D conditions were also lower than in 2D, while p-ERK levels were comparable in 2D and 3D conditions (Supplementary Fig. S1B). It is also worth mentioning that fluctuation of p-ERK levels was observed between experiments and are likely a result of the size and how long the spheroids have formed. Nonetheless, these baseline differences on p-ERK and p-AKT levels indicate differential wiring of cell signaling in 2D versus 3D culture, which is further supported by the literature (35, 37, 38). The combination of lower p-ERK levels, for SHP099 to inhibit, and a downregulated PI3K pathway, often implicated in cancer cell survival, may contribute to the sensitivity to SHP2 inhibition in a subset of KRAS-mutant cells cultured in 3D.

Given that SHP099 suppressed MAPK signaling similarly in 2D and 3D but had different impacts on cell proliferation and the acute MAPK suppression is not complete, we utilized gene expression analyses to determine whether alternative signaling pathways may contribute to the differential sensitivity to SHP099. MIA PaCa-2 cells, cultured in 2D or 3D and treated with SHP099 or trametinib, were subjected to RNA sequencing (RNA-seq) analysis. Using a recently described MAPK pathway signature with relevance to clinical responses to MEK inhibition (MPAS: SPRY2, SPRY4, ETV5, DUSP4, DUSP6, CCND1, EPHA2, EPHA4; ref. 39) along with several other MAPK pathway transcriptional markers (ETV1, DUSP5, DUSP7, EPHA7, BMF; ref. 40), we compared the effects of SHP099 with those of trametinib. The MAPK pathway transcriptional effects of SHP099, while less than those of trametinib, did demonstrate MAPK pathway inhibition (Fig. 2B), consistent with the effects on p-ERK and p-RSK3 in the MIA PaCa-2 cells (Fig. 2A). Interestingly, we also found that levels of SHP2 transcripts at baseline in 3D were approximately 50% lower than in 2D and were more comparable with those observed in tumor xenografts (Fig. 2C). The decreased SHP2 expression in MIA PaCa-2 was also confirmed at the protein level, concomitant with decreased SHP2 phosphorylation at Y542, a marker of SHP2 activation by RTK activity (Fig. 2D; ref. 41). We then examined SHP2 and p-SHP2 levels in the remainder of the cell lines and observed a similar decrease of both SHP2 and p-SHP2 in Capan-2 cells and a decrease only in p-SHP2 in KP4 cells and in T3M-4, when they were cultured in 3D (Fig. 2D; Supplementary Fig. S1C). It is conceivable that the decreased levels of SHP2 or that SHP2 activation in 3D may partially contribute to the increased sensitivity in 3D culture, as there is less SHP2 or less activated SHP2 to engage with the allosteric SHP2 inhibitor.

Deeper analysis of the transcriptional changes in the MIA PaCa-2 revealed substantial differences in the global gene expression changes in response to treatment with SHP099 in 3D culture as compared with 2D culture (Supplementary Fig. S1D) 19 hours after treatment. Gene ontology analysis (42, 43) revealed greater modulation of transcriptional programs after SHP099-treatment related to apoptosis, IFNγ signaling, cell cycle, and membrane localization of proteins and is likely a reflection on the observed enhanced efficacy in 3D. Interestingly, in 2D culture there are very few uniquely altered genes following SHP099 treatment, notably downregulation of genes associated with cell adhesion, p38 MAPK, and TGFβ/SMAD signaling (Supplementary Fig. S1D); these same genes display no differential expression with SHP099 treatment in the 3D culture setting. In summary, KRAS-mutant cells are dependent on SHP2 for full MAPK pathway activation in both 2D and 3D growth conditions. The efficacy of SHP2 inhibition may be more pronounced in KRAS-mutant cells in 3D growth conditions likely due to a variety of reasons including changes in MAPK pathway, the AKT pathway, and SHP2 expression and activation.

We next evaluated whether SHP099 would be efficacious in KRAS-mutant human tumor xenograft models. In line with the in vitro 3D findings, SHP099, dosed at 100 mg/kg daily, was efficacious in the MIA PaCa-2 model, with equivalent response to trametinib (Fig. 3A). In contrast to the lack of in vitro response to SHP099 (Fig. 1D), the KRAS^Q61H T3M-4 tumors, displayed sensitivity to SHP099 in vivo with a modest improvement in response when compared to trametinib (Fig. 3B). We next examined MAPK pathway suppression in vivo as assessed by p-ERK and p-RSK3 in MIA PaCa-2 tumors. MAPK pathway suppression was observed with SHP099 (Fig. 3C), consistent with the newly appreciated biology that the KRAS^G12C-mutant cancers retain intrinsic GTPase activity and are regulated by RTKs and SHP2 (17, 21–23). However, in T3M-4 tumors (KRAS^Q61H), suppression of p-ERK by SHP099 was variable and modest among tumors, while trametinib robustly and consistently suppressed p-ERK without impacting p-RSK3, suggesting a decoupling of the p-RSK effect from ERK activity in this cell line (Fig. 3D).

We extended the in vivo profiling of SHP099 across a panel of KRAS-mutant xenograft models, including the KYSE520 (KRAS^WT/EGFR-dependent) and RKO (BRAF^V600E; Fig. 3E; Supplementary Fig. S2A–S2I). Relative to the KYSE520 (Supplementary Fig. S2A), SHP099 achieved comparable antitumor activity in two KRAS^G12C models, NCI-H358 and MIA PaCa-2, as well as a KRAS^G12D model, KP4 (Fig. 3A and E; Supplementary Fig. S2B and S2C). In addition, tumor growth inhibition was observed in all other KRAS-mutant models tested (Supplementary Fig. S2D–S2G), except in NCI-H460 harboring homozygous KRAS^Q61K mutations, where there was no effect of SHP099 treatment (Supplementary Fig. S2H), and more comparable with SHP099 treatment in the RKO BRAF^V600E xenograft model (Fig. 3E; Supplementary Fig. S2I).

We next examined the extent of MAPK pathway inhibition in vivo following SHP099 treatment using DUSP6 mRNA expression, a downstream transcriptional readout for MAPK pathway activity, 3 hours after the last treatment dose (Fig. 3F). Overall, the magnitude of DUSP6 suppression correlated with the magnitude of antitumor activity of SHP099. For the KRAS^G13D model, HCT15, and the KRAS^Q61H model, T3M-4, we observed only modest changes in DUSP6 suppression, which is likely insufficient to cause the observed magnitude of antitumor activity. These data further suggest that mechanisms outside of the MAPK pathway may also contribute to the in vivo efficacy of SHP099 in KRAS mutant cancers.

Several reports have demonstrated that VEGFR signaling depends on SHP2 in endothelial cells and is required for angiogenesis (14, 44, 45). To ascertain whether the efficacy of SHP099 in KRAS-mutant models in vivo was in part due to an antiangiogenic effect of VEGFR2 signaling inhibition, we evaluated SHP2 inhibition in vitro in tube formation assays using human endothelial cells, HUVECs. SHP099 was unable to inhibit HUVEC tube formation in vitro at 5 μmol/L, a concentration above what can be achieved in vivo at the MTD of 100 mg/kg, daily (Fig. 4A). We next evaluated whether efficacy of SHP099 was due to antiangiogenic effects in vivo. In contrast to dovitinib, a multi-targeted RTK inhibitor with antiangiogenic effects (46) that clearly caused tumor devascularization in a KRAS-mutant colorectal cancer human PTX model, HCOX4087, in vivo treatment with SHP099, had no gross impact on tumor vascularization and was qualitatively comparable to the vehicle and trametinib-treated tumors (Fig. 4B; Supplementary Fig. S3A). We further examined in vivo efficacy of a selective VEGFR2 inhibitor, BFH722 (29), in MIA PaCa-2 and T3M-4 xenografts and found that it had minimal efficacy (Supplementary Fig. S3B and S3C), in contrast to SHP099, while also displaying inhibition of mouse CD31 mRNA expression (Fig. 4C), suggesting that these models are not significantly dependent on tumor vascularization for in vivo growth. SHP099 also maintained efficacy when MIA PaCa-2 cells were grown orthotopically by surgical implantation directly into the pancreas, demonstrating that the effects of SHP099 are unlikely to be an artifact of subcutaneous xenografts exhibiting enhanced permeability and retention (EPR; Supplementary Fig. S3D; ref. 47). Having established that SHP099 had little impact on tumor vascularization, we evaluated whether genetic depletion of SHP2 within KRAS-mutant tumor cells would replicate the effects of SHP099. Using CRISPR-Cas9 technology, SHP2 knockout (crKO) MIA PaCa-2 cells were generated and confirmed (Fig. 4D). Knock out of SHP2 in the MIA PaCa-2 cell line had only modest growth effects when grown in a monolayer culture, whereas significant growth inhibition was observed when the cells were grown in 3D (Fig. 4D) or when implanted subcutaneously into nude mice (Fig. 4E).

To assess the role of SHP2 in tumor maintenance versus initiation and to also rule out potential clonal effects from the SHP2-crKO model, MIA PaCa-2 cells transduced with a doxycycline (dox)-inducible shRNA targeting PTPN11 was generated (Supplementary Fig. S3E). Comparable with SHP099 treatment, SHP2 knockdown resulted in significant antiproliferative effects in a 3D colony formation assay (Supplementary Fig. S3F). We next performed a head-to-head comparison of efficacy between SHP099 and SHP2-shRNA using the engineered dox-inducible SHP2-shRNA MIA PaCa-2 cells. Following tumor randomization, efficacy of SHP099 and dox-induced SHP2 knockdown was evaluated for 15 days (Fig. 4F). Efficacy between SHP2 genetic depletion and pharmacologic inhibition was comparable, with a modest but statistically insignificant improvement with the inhibitor treatment group, attributed to greater MAPK pathway suppression as measured by DUSP6, and incomplete depletion of SHP2 in the dox-treated xenografts (Supplementary Fig. S3G). To demonstrate that the effects observed by SHP099 are due to on-target SHP2 inhibition, MIA PaCa-2 SHP2-crKO cells were transduced with a dox-inducible SHP099-binding deficient mutant, SHP2^T253M/Q257L, and in vitro 3D efficacy was evaluated in the presence of dox and SHP099 (Fig. 4G). The phosphatase activity of SHP2^T253M/Q257L mutant was demonstrated to be similar to that of wild-type SHP2 (15) and the expression level of SHP2^T253M/Q257L mutant in the SHP2 KO background was very similar to the endogenous level of SHP2 in the parental cells. SHP099 was inactive in cells expressing the compound-binding deficient mutant of SHP2, demonstrating that the antiproliferative activity of SHP099 is due to on target inhibition of SHP2 (Fig. 4G).

Because SHP2 is a node downstream of RTKs, we hypothesized that the effect of SHP2 inhibition in KRAS-mutant cancers could be recapitulated with RTK inhibitors. We first evaluated the efficacy of an RTK inhibitor cocktail, targeting ERBBs, FGFRs, and MET, across a panel of KRAS-mutant cell lines in both 2D and 3D culture conditions (Fig. 5A; Supplementary Table S4). Similar to treatment with SHP099, we found that the modest efficacy of the RTK inhibitor cocktail in 2D culture was also enhanced in the 3D growth conditions (Figs. 5A and 1E). We next confirmed these results in vitro and in vivo in both the MIA Paca-2 and KP4 cell lines (Fig. 5B–E). As observed with SHP099 (Fig. 1C), efficacy of the RTK-inhibitor cocktail (Fig. 5B) in the MIA Paca-2 was improved in the 3D setting as evidenced by a shift in the antiproliferation IC₅₀. In vivo efficacy of an EGFR mAb, cetuximab, or with an FGFR1-3 inhibitor, BGJ398, in the MIA PaCa-2 tumors resulted in modest tumor growth inhibition, 64% T/C and 100% T/C, respectively. In contrast, combination of these two inhibitors resulted in a significant improvement in efficacy (26% T/C); however, SHP099 remained the most active treatment in this xenograft model (7% T/C), presumably due to the ability of SHP099 to inhibit signaling from additional upstream RTKs, and potentially impacting other signaling pathways (Fig. 5C).

Consistent with the antitumor efficacy, MAPK pathway suppression, as assessed by p-RSK3 levels, was reduced by SHP099 or the combination of cetuximab and BGJ398, but not cetuximab or BGJ398 alone (Supplementary Fig. S4A). In contrast, p-ERK levels were comparable in all treatment groups (Supplementary Fig. S4A) and only DUSP6 suppression was observed by SHP099 (Supplementary Fig. S4A), likely due to its well-known tight regulation and the use of end of efficacy samples. Interestingly, there is a drastic p-SHP2 downregulation in MIA PaCa-2 xenografts and 3D organoids when compared with the cells cultured in 2D, consistent with the p-SHP2 reduction observed in 3D culture while p-AKT decrease in tumors is less pronounced compared wih the decrease in 3D spheroids (Supplementary Fig. S4B). Similarly, the p-EGFR and p-FRS2 signal in MIA PaCa-2 tumors were very low, preventing us to understand the EGFR and FGFR activation status and responses to cetuximab and BGJ398 (Supplementary Fig. S4A). We also evaluated the KP4 model, a model that upon further analysis was determined to be dependent on MET due to overexpression of the MET-ligand, HGF (Supplementary Fig. S4C). In KP4 we tested sensitivity in 3D and in vivo toward capmatinib (Fig. 5D and E). Similar to the MIA PaCa-2 model, SHP099 treatment in the KP4 model trended toward better efficacy than that of the MET inhibitor, although the results were not statistically significant. Consistent with the antitumor efficacy, p-RSK3 levels were modestly reduced by SHP099 or capmatinib (Supplementary Fig. S4D). Comparable with the MIA PaCa-2, p-SHP2, and p-MET levels, but not p-AKT levels, were also downregulated in KP4 tumors and 3D organoids compared to KP4 cells cultured in 2D (Supplementary Fig. S4E). To further test our hypothesis that in vivo sensitivity to SHP099 can potentially be related to sensitivity to RTK inhibitors, we evaluated efficacy of an RTK inhibitor combination in the KRAS^Q61H-mutant T3M-4 model. Combination of an FGFR1-3 inhibitor (BGJ398), MET inhibitor (capmatinib), and EGFR inhibitor (cetuximab) was well-tolerated and resulted in significant antitumor activity when compared with the control and was comparable with that of SHP099 (Fig. 5F). Given that comparable antitumor activity was observed between SHP099 and RTK inhibitor treatment, combination strategies with SHP2 inhibitors are warranted to deepen the in vivo response. Taken together, these data demonstrate that RTK activity is responsible for the activation of SHP2 in KRAS-mutant tumors and support further investigation toward the use of SHP2 inhibitors and combination partners to treat RTK-dependent KRAS-mutant cancer.

In this report, stimulated by the observation that a small subset of KRAS-mutant cells showed sensitivity to the allosteric SHP2 inhibitor SHP099, we profiled a panel of KRAS-mutant cell lines grown as multicellular spheroids and revealed an unexpected and widespread sensitivity to SHP2 inhibition when KRAS-mutant cancer cells are grown in 3D. The efficacy of SHP099 in KRAS-mutant models was further confirmed in vivo and recapitulated by SHP2 depletion or knockdown in cancer cells, suggesting a tumor cell intrinsic mechanism (e.g., in the MIA PaCa-2 model bearing KRAS^G12C homozygous mutations). Significant MAPK pathway suppression by SHP099, albeit to a lesser extent compared with the MEK inhibitor, trametinib, was observed in vivo, 3D, and 2D growth conditions, in all models bearing KRAS G12 alterations tested and to some extent in KRAS G13 and Q61 mutant models. However, MAPK suppression may not fully explain the activity of SHP099 in all KRAS-mutant models and other SHP2-dependent but MAPK-independent mechanisms may contribute to the activity of SHP099 in vivo, as illustrated in Supplementary Fig. S4F.

The enhanced activity of SHP099 in KRAS-mutant cells in 3D compared with 2D conditions, despite a similar degree of MAPK pathway suppression and rebound (Figs. 1C and 2A), is intriguing. The altered sensitivity of cancer cells to anticancer therapies in 3D growth conditions compared with 2D conditions has previously been reported in KRAS-mutant cancer (22), HER2⁺ breast cancer (37), and colorectal cancer (35). However, the mechanisms proposed to be responsible for these differential responses in 2D and 3D are not consistent and are possibly due to lineage and treatment specific differences. We observed a decreased baseline level of p-ERK and p-AKT and more importantly discovered a downregulation of SHP2 in three KRAS-mutant cell lines cultured in 3D conditions, manifested as either decreased SHP2 protein levels (transcriptional mechanism for MIA PaCa-2) or decreased phosphorylation at the Y542 site. A similar p-SHP2 downregulation was observed in MIA PaCa-2 and KP4 tumors compared with 2D growth conditions, as well as p-MET, but not p-AKT (Supplementary Fig. S4B and S4E). Those signaling differences in 3D and in vivo may partially explain why the sensitivity to SHP2 inhibitors of KRAS mutant cells is particularly revealed under 3D and in vivo conditions. Although one may conclude that the 3D response is more predictive of in vivo efficacy, the T3M-4 model that showed a response to SHP099 in vivo did not exhibit sensitivity in 3D. It is possible that the role of SHP2 in KRAS-mutant signaling is dependent on a complex tumor microenvironment beyond what a tumor spheroid in vitro model can reveal. It is also possible that different mechanisms could account for the 3D sensitivity and in vivo sensitivity to SHP099 in different KRAS-mutant models, as the p-AKT levels in 2D more resemble that in xenograft tumors of MIA PaCa-2 and KP4 (Fig. 2A; Supplementary Fig. S4B and S4E).

KRAS and MAPK pathway activation are regulated by upstream signaling proteins, such as EGFR, FGFR, MET (48–51), as well as by the SHP2 protein tyrosine phosphatase and the RAS-GEF, SOS1 (7, 17). Indeed, tyrosine kinase inhibitors have been shown to provide overall survival benefit in patients (50, 52, 53); however, retrospective analysis has also demonstrated that the activity of these selective inhibitors is mostly restricted to tumors harboring wild-type RAS or RAF (54–57). Furthermore, in vitro 2D screens of RTK and SHP2-inhibitors in cell lines confirmed that their activity was dependent on the wild-type KRAS/BRAF genetic status (Fig. 1A and B; ref. 15). These data had long supported the notion that gain-of-function KRAS mutations, commonly of the G12, G13, Q61 form, were constitutively active and that targeting upstream signaling would be ineffective. However, this paradigm has shifted in recent years, with several publications demonstrating biochemically that the KRAS^G12C mutation is not locked in the constitutively active form, but maintains intrinsic GTP-hydrolytic activity (19–21). In fact, these data helped the development of a selective and active KRAS^G12C inhibitor, ARS853 (20). Together with the newly appreciated KRAS biology and the effects of SHP099 in KRAS-mutant cancer, we demonstrate that KRAS-mutant cancers are still dependent on upstream receptor tyrosine kinase signaling, although the specific RTK's that regulate the SHP2-KRAS signaling axis appear to be cancer cell-line specific (Fig. 5). Furthermore, the activity observed with SHP099 appears to be comparable to that of RTK-inhibition in these cell lines, where only tumor stasis is observed. It is possible that treatment of cancers with a SHP2 inhibitor alone could prevent pathway feedback reactivation of the MAPK pathway or block RTK-bypass mechanisms where upregulation of other RTK's under treatment may contribute to tumor resistance (33). However, to induce tumors into regression the identification of optimal clinical combination agents with a SHP2 inhibitor warrants further investigation.

With the generation of the first selective SHP2 inhibitor, SHP099 (15, 16), we demonstrated that SHP099 was potently active in RTK-driven cancers harboring wild-type KRAS and BRAF (15). Consistent with the notion that KRAS-mutant cancers would be insensitive to SHP2 inhibition, we showed limited activity of SHP099 in KRAS-mutant cancer cells grown in 2D culture. However, broad evaluation of SHP099 in KRAS-mutant cell lines cultured as either tumor spheroids or as in vivo xenograft models revealed significant antitumor activity (Figs. 1 and 3). These findings mirror that which has been recently and independently reported, that KRAS-mutant cancers, specifically KRAS^G12C and KRAS^G12D variants still depend on SHP2 activity (17, 22, 23). Here, we show that the in vivo activity of SHP099 in the MIA PaCa-2 KRAS^G12C model in an immunocompromised mouse setting is due to tumor cell-intrinsic mechanisms. However, this does not exclude the potential for additional antitumor activity due to nontumor cell-intrinsic mechanisms that a SHP2-inhibitor may impart in humans, where SHP2-mediated effects may include both tumor cell autonomous effects and effects on stromal cells and immune cells (25, 58). In T-cells, SHP2 was shown to be recruited by programmed cell death-1 (PD1) to dephosphorylate and inactivate costimulatory receptor CD28, which suppresses T-cell activation (25). SHP099, intraperitoneally dosed, has significant anti-tumor efficacy in the CT-26 syngeneic model grown in immunocompetent mice, but little efficacy when the model is grown in immunodeficient nude mice or in vitro, although the dose and route of administration of SHP099 that were used are questionable (59). Using myelomonocytic cell–specific Shp2 knockout mice, Shp2 ablation in myeloid cells was shown to inhibit melanoma growth by potentiating macrophage production of T-cell chemoattractant CXCL9 in response to IFN-γ and tumor cell-derived cytokines, thereby facilitating the tumor infiltration of IFNγ-producing T cells (60). In addition, SHP2 inhibition synergized with PD-1 blockade in the MC-38 murine syngeneic colon cancer model (59).

The dependence of multiple RTKs/SHP2 for activation of the mutant KRAS signaling and the potential immunomodulatory effects of SHP2 inhibition provides a strong clinical rationale for the utility of a SHP2 inhibitor for treating KRAS-mutant cancers, where treatment options are currently limited.

S. Kovats is the scientific technical leader at Novartis Institute for Biomedical Research. M. Shirley is an investigator at Novartis Institute for Biomedical Research. M.J. Meyer is a former employee of Novartis Institute of Biomedical Research; is the Head, Discovery Pharmacology and In Vivo Biology, at BMS; and has ownership interest (including patents) in Novartis and BMS. M.J. LaMarche is a senior investigator at Novartis Institutes of Biomedical Research and has ownership interest (including patents) in NVS. S. Moody is the clinical program leader at Novartis Institutes for Biomedical Research. S.J. Silver is a former employee and stockholder at Novartis Institutes for Biomedical Research. D.D. Stuart is an executive director at Novartis Institutes of Biomedical Research and has ownership interest (including patents) in Novartis. T.J. Abrams is a senior investigator at Novartis Institutes of Biomedical Research and has ownership interest (including patents) in Novartis. J. Williams is an executive director at Novartis, has a commercial research grant for Novartis, and has ownership interest (including patents) in Novartis. J.A. Engelman is the global oncology head at Novartis Institutes of Biomedical Research and has ownership interest (including patents) in Novartis. M. Mohseni is an investigator at Novartis Institutes of Biomedical Research and has ownership interest (including patents) in Novartis. No potential conflicts of interest were disclosed by the authors.

Conception and design: H.-X. Hao, H. Wang, C. Liu, S. Kovats, M.J. Meyer, M.J. LaMarche, G. Caponigro, T.J. Abrams, J.A. Engelman, M. Mohseni

Development of methodology: H. Wang, C. Liu, S. Kovats, B. Pant, J. Lim, M. Fleming, S.J. Silver, T.J. Abrams, S. Goldoni

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Liu, S. Kovats, R. Velazquez, H. Lu, B. Pant, M.J. Meyer, M. Pu, J. Lim, M. Fleming, L. Alexander, A. Farsidjani, M.J. LaMarche, J. Williams

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.-X. Hao, H. Wang, C. Liu, S. Kovats, H. Lu, B. Pant, M. Shirley, M.J. Meyer, M. Pu, J. Lim, M. Fleming, A. Farsidjani, M.J. LaMarche, S.J. Silver, D.D. Stuart, S. Goldoni, M. Mohseni

Writing, review, and/or revision of the manuscript: H.-X. Hao, C. Liu, R. Velazquez, B. Pant, M.J. Meyer, S. Moody, G. Caponigro, P.S. Hammerman, J. Williams, S. Goldoni, M. Mohseni

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Wang, B. Pant, M. Pu, A. Farsidjani, M. Mohseni

Study supervision: H.-X. Hao, M.J. LaMarche, S.J. Silver, G. Caponigro, D.D. Stuart, P.S. Hammerman, J.A. Engelman, M. Mohseni

We thank the following Novartis colleagues, Yingnan Chen, Brant Firestone, Ho Man Chan, Michael Acker, Hui Gao, Alice Loo, Suzanna Clark, and Lisa Duong for their helpful technical advice, discussion of data interpretation, and suggestions of experiment design on the manuscript.

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.