RUNX2/CBFB modulates the response to MEK inhibitors through activation of receptor tyrosine kinases in KRAS-mutant colorectal cancer.

Intrinsic and acquired resistances are major hurdles preventing the effective use of MEK inhibitors for treatment of colorectal cancer (CRC). Some 35-45% of colorectal cancers are KRAS-mutant and their treatment remains challenging as these cancers are refractory to MEK inhibitor treatment, because of feedback activation of receptor tyrosine kinases (RTKs). We reported previously that loss of ERN1 sensitizes a subset of KRAS-mutant colon cancer cells to MEK inhibition. Here we show that the loss of RUNX2 or its cofactor CBFB can confer MEK inhibitor resistance in CRC cells. Mechanistically, we find that cells with genetically ablated RUNX2 or CBFB activate multiple RTKs, which coincides with high SHP2 phosphatase activity, a phosphatase that relays signals from the cell membrane to downstream pathways governing growth and proliferation. Moreover, we show that high activity of SHP2 is causal to loss of RUNX2-induced MEK inhibitor resistance, as a small molecule SHP2 inhibitor reinstates sensitivity to MEK inhibitor in RUNX2 knockout cells. Our results reveal an unexpected role for loss of RUNX2/CBFB in regulating RTK activity in colon cancer, resulting in reduced sensitivity to MEK inhibitors.

Introduction

KRAS-mutant colorectal cancers (CRCs) are highly refractory to targeted treatments, including inhibition of direct downstream targets of KRAS, such as MEK [[1], [2], [3]]. We reported previously that genetic ablation of the endoplasmatic reticulum–embedded kinase ERN1 sensitizes KRAS-mutant colorectal cancer cells to MEK inhibition [4]. To reveal the mechanistic connection between ERN1 and the MAPK signaling pathway, we performed a genome-wide CRISPR/Cas9 genetic screen in ERN1 knockout (KO) cells to find modulators of the MEK inhibitor response. Using this approach, we identified a number of genes whose inactivation can support the growth and proliferation of ERN1 KO cells in the presence of MEK inhibitor [4]. Besides previously validated genes, DUSP4, DET1, and COP1, we also identified in the same genetic screen RUNX2 (formerly known as core-binding factor subunit alpha-1 or CBFA1) and its transcriptional coactivator; CBFB (core-binding factor subunit beta) as regulators of the MEK inhibitor response [4].

The RUNX (runt-related) family of transcription factors are heterodimeric proteins composed of a DNA-binding alpha subunit and a non-DNA–binding beta subunit. All three mammalian RUNX proteins (RUNX1, RUNX2, and RUNX3) bind to a common DNA motif and heterodimerize with CBFβ, which facilitates DNA binding of RUNX proteins without making a direct contact with DNA itself [5]. The functional and mechanistic interaction between RUNX2/CBFA1 and CBFβ has been well documented using various in vitro [6] and in vivo model systems [[7], [8], [9], [10]]. While all three RUNX proteins are involved in skeletal development and differentiation, RUNX2 is best characterized in terms of its role in control of bone cell proliferation and differentiation [11]. RUNX2 is often referred to as the principal osteogenic master switch, as it is essential for the formation of mature osteocytes and for controlling the expression of genes required for mineralization of the bone extracellular matrix. Heterozygous loss of RUNX2 causes the skeletal disease cleidocranial dysplasia [12], whereas mice with a homozygous mutation in the Runx2 locus die at birth without breathing, because of complete lack of ossification (bone formation) [13,14]. In normal development, RUNX2 is robustly expressed during early embryogenesis, before formation of bone tissue, and its functional relevance at later stages of development remains less defined. It is, however, known that RUNX2 can regulate cell migration [15] and vascular invasion in bone tissue [16]. These were the first findings supporting the role of RUNX2 in cell fate determination in cells that are not of osteogenic lineage.

RUNX2 is regulated by a number of posttranscriptional control mechanisms including selective proteolysis and phosphorylation. Specific ERK/MAPK phosphorylation sites on RUNX2 have been identified and functionally characterized, which suggest that RUNX2 is activated by the MAPK pathway [17,18]. Here, we set out to investigate how the loss of either RUNX2 or CBFB can cause MEK inhibitor resistance in colorectal cancer. Our studies reveal an unexpected role for RUNX2 in controlling signaling through the MAP kinase pathway through regulation of multiple RTKs.

Materials and Methods

Cell Culture, Transfection, and Lentiviral Infection

We used HEK293 cells, cultured in DMEM, for lentiviral production. All other cell lines were maintained in RPMI1640 medium containing 10% FBS and 1% penicillin/streptomycin at 37 °C and 5% CO2. All cell lines were purchased from the American Type Culture Collection (ATCC), STR profiled (by Eurofins Medigenomix Forensik GmbH, Ebersberg, Germany) and routinely tested negative for mycoplasma. Transfection of HEK293 cells with linear polyethylenimine (PEI) 25K from polysciences (cat# 23966-2) and subsequent infection of target cells was performed as described previously [19]. For knockout of individual genes, the following single guide (sg) RNAs were cloned in the lenti CRISPR version 2.1 (LC2.1) vector by Gibson cloning: sgERN1, 5′-ACATCCCGAGACACGGTGGT-3’; sgRUNX2, 5′-GCTGTCGGTGCGGACGAGTT-3’; sgCBFB, 5′-GCCGACTTACGATTTCCGAG-3’. Nontargeting (NT) sgRNA 5′-ACGGAGGCTAAGCGTCGCAA-3′ was used as a control.

Cell Proliferation Assays and Growth Curves

For long-term cell proliferation assays cells were seeded in 6-well plates at densities of around 1 × 104 cells per well, in 12-well plates at around 5 × 103 cells per well, or in 48-well plates at around 2 × 103 cells per well, and cultured with or without inhibitors, as indicated. When control cells reached confluency, all cells were fixed in 4% formaldehyde and stained with 0.1% crystal violet (in water).

For short-term growth inhibition assays, cells were seeded in 96-well plates at around 800 cells per well. Twenty-four hours after seeding, serial dilutions of AZD6244 were added to cells to final drug concentrations ranging from 0.04 to 10 μM. Cells were then incubated for 72 hours and cell viability was measured using the CellTiter-Blue Viability Assay (Roche). Relative survival in the presence of AZD6244 was normalized to the untreated controls after subtracting the background.

Live cell growth was measured by automated determination of confluency every 4 hours using IncuCyte Zoom (Essen Bioscience). Between 600 and 800 cells were plated per well of a 96-well plate and all experiments were carried out in triplicates. MEK inhibitors selumetinib (AZD6244) and trametinib (GSK1120212) were purchased from Selleck chemicals and kept as 10 mM stock solution aliquots in DMSO. SHP2 inhibitor SHP099 was synthesized as described previously [20]. Afatinib and pan-RAF inhibitor LY3009120 were purchased from MedKoo Inc. and neratinib was purchased from Bio-Connect BV. Anti-EGFR monoclonal antibody cetuximab was obtained from the Hospital Pharmacies at The Netherlands Cancer Institute.

Protein Lysate Preparation and Western Blot Analysis

Cells were lysed and western blots were performed as described previously [19]. Primary antibodies against HSP90 (H-114; sc-13119), ERK1 (C16; sc-93), ERK2 (C14; sc-154), p-ERK1/2 (E-4; sc-7383), and SHP2 (SH-PTP2 C-18; sc-280) were purchased from Santa Cruz Biotechnology. Antibodies against EGFR (ab40815), p-EGFR (Y1068; ab5644), and p-SHP2 (Y542; ab62322) were obtained from Abcam. Antibodies against ERN1 (3294), p-ERK1/2 (9101), RUNX2 (12556), CBFB (12902), JUN (9165), RSK1 (8408), p-p90 RSK (Ser380; 9335), and p-RET (3221) were purchased from Cell Signaling. Antibody against p-RSK1 (Thr359/Ser363; 04-419) was obtained from Millipore. Secondary antibodies were obtained from Bio-Rad Laboratories. Human Phospho-Receptor Tyrosine Kinase Array Kit was purchased from R&D and processed according to manufacturer's instructions. All experiments shown, except RTK array analysis, were performed independently at least three times.

Computational Analysis of Drug Response Data

Drug response and (RMA normalized) RUNX2 expression and mutation data in colorectal cancer cell lines from the GDSC1000 panel were downloaded from www.cancerrxgene.org [21]. IC50 values of the 5 MEK inhibitors in the panel were median centered to make them comparable. T-tests were performed to assess statistical significance of difference in response between groups.

Analysis of RNA-seq Data

Transcriptomic analysis was performed using the R-package limma [22]. Non- or lowly expressed genes (<1 count per million in at least 2 samples) were removed before analysis. Read counts were transformed using the voom function. Multidimensional scaling (MDS) analysis was performed with the plotMDS function. Because of the transcriptional and phenotypic similarity, the ERN1/CBFB and ERN1/RUNX2 DKO lines were treated as biological replicates. Differential gene expression analysis was performed using a linear model with drug-treatment, CBFB/RUNX2 KO status, and the interaction between drug-treatment and KO status as variables, using the standard limma functions lmFit and eBayes. Raw and processed data from the next-generation RNA sequencing of samples have been deposited to the NCBI Gene Expression Omnibus (GEO) under accession number GSE139169.

Results

We recently performed a genetic screen for MEK inhibitor resistance in ERN1 null LoVo CRC cells, which identified both RUNX2 and CBFB as potential modifiers of the response to MEK inhibitors [4]. To validate these findings, we asked whether loss of RUNX2 or CBFB indeed confers resistance to MEK inhibitors in these cells. We introduced gRNAs targeting the RUNX2 or CBFB genes in LoVo ERN1 KO cells. After selection, we tested the growth of ERN1/RUNX2 double knockout (DKO) and ERN1/CBFB DKO cells in the presence and absence of MEK inhibitors selumetinib and trametinib (Figure 1A). As predicted by the genetic screen, LoVo ERN1 KO cells, which are sensitive to low nanomolar amounts of MEK inhibitor, showed strong resistance to MEK inhibitor treatment when gRNAs targeting RUNX2 or CBFB were introduced. We also tested the effect of RUNX2 or CBFB loss in HCT-116 ERN1 KO cells as an additional model for ERN1-loss–induced MEK inhibitor sensitivity. However, we observed no difference in MEK inhibitor response between HCT-116 ERN1 KO cells transduced with nontargeting (NT) control sgRNA and sgRNAs targeting RUNX2 or CBFB (Figure 1A). Biochemical analyses of protein cell lysates showed pronounced levels of RUNX2 protein in LoVo cells, but much lower levels in HCT-116 (Figure 1B), which might explain why we observed no change in MEK inhibitor response upon its loss. CBFβ shows similar levels of expression in LoVo and HCT-116 (Figure 1C), but its loss also failed to rescue the MEK inhibitor sensitivity of HCT-116 ERN1 KO cells (Figure 1A). This could be explained by the notion that CBFβ acts through RUNX2, which is poorly expressed in HCT-116 cells.

Figure 1

MEK inhibitor sensitivity of LoVo and HCT-116 . (A) Colony formation assays comparing the growth of CBFB and RUNX2 KO cells with cells expressing nontargeting (NT) gRNA in the presence of indicated concentrations of two different MEK inhibitors; selumetinib (left) and trametinib (right). After 10 days of culture, cells were fixed and stained. Image is representative of three independent experiments. (B) Western blot analysis of LoVo and HCT-116 ERN1/RUNX2 double KO cells compared with parental cells and cells expressing NT gRNA control. All (+) samples were treated with 1 μM of MEK inhibitor selumetinib for 24 hours before collection, and compared with vehicle treated (−) samples. Protein extracts were probed with specific antibodies against ERN1, phosphorylated ERK, ERK2, and RUNX2 (to estimate the efficiency of CRISPR editing in a polyclonal population). Specific antibody against HSP90 was used as a loading control. (C) Western blot analysis of LoVo and HCT-116 ERN1/CBFB double KO cells compared with parental cells and cells expressing NT gRNA control. All (+) samples were treated with 1 μM of MEK inhibitor selumetinib for 24 hours before collection, and compared with vehicle treated (−) samples. Protein extracts were probed with specific antibodies against CBFβ (to estimate the efficiency of CRISPR editing), and HSP90 (as a loading control).

We computationally analyzed RUNX2 expression in 45 colorectal cancer cell lines from the GDSC1000 panel according to Iorio et al. [21]. Consistent with our Western blot data (Figure 1B), we found that HCT-116 cells, unlike LoVo, are on the lower end of the RUNX2 gene expression scale (Figure 2A). Together, these data indicate that expression of RUNX2 in CRC is rather heterogeneous. Next, we compared the IC50 values of MEK inhibitors of the upper 25% of RUNX2 expressing cell lines with lower 25% (Supplemental Table 1). The data, represented in a box plot, indicate that RUNX2 expression levels are not per se linked to MEK inhibitor response (Figure 2B). However, when we compared cell lines with RUNX2 mutations to nonmutant cells, we found that RUNX2 mutant cells have significantly higher MEK inhibitor IC50 values (Figure 2C). Considering that RUNX2 gene does not have a hotspot mutation site, it is likely that the mutations are inactivating. These data suggest that loss of RUNX2 leads to resistance to MEK inhibitors in CRC. To test directly whether loss of RUNX2 or CBFB could decrease the response of CRC cells to MEK inhibitors in the absence of ERN1 loss, we made polyclonal populations of RUNX2 and CBFB knockout cells in parental LoVo cells. As can be seen in Figure 2D, these polyclonal populations had reduced expression, but not complete loss, of RUNX2 and CBFβ. Nonetheless, this was sufficient to cause a marked increase in resistance to the MEK inhibitor selumetinib (Figure 2E). This indicates that RUNX2 or CBFB can modulate MEK inhibitor responses in a broader context than just in cells having ERN1 loss of function (Figures 2D and E).

Figure 2

Analyses of RUNX2 expression and mutation status in a CRC panel. MEK inhibitor response of LoVo parental cells with loss of . (A) RUNX2 expression in colorectal cancer cell lines from the GDSC1000 panel. Blue dots represent cell lines in upper-most quantile and yellow dots lowest quantile of RUNX2 expression. (B) IC50 values for five different MEK inhibitors in lowest 25% RUNX2 expressing cells (C2BBe1, CCK-81, COLO-205, COLO-741, CW-2, GP5d, HCC-2998, HCT-116, LS-180, NCI-H630, SNU-81 and SW-1463; shown on the left) and highest 25% RUNX2 expressing cells (CL-11, COLO-320-HSR, HT-55, LS-123, LoVo, RCM-1, SNU-407, SW-1116, SW-1417, SW-48 and SW-837; shown on the right). (C) IC50 values for five different MEK inhibitors in four RUNX2 mutant CRC cell lines (GP5d, HT-115, SW-620, SW-948; right) compared with nonmutant lines (41 RUNX2 wild-type CRC cell lines; left). (D) Western blot of LoVo RUNX2 KO and CBFB KO cells compared with NT gRNA controls. (E) Colony formation assay of LoVo RUNX2 KO and CBFB KO cells in the presence and absence of the MEK inhibitor selumetinib, relative to LoVo parental cells expressing nontargeting (NT) gRNA control. Shown is a representative example of three biological replicates.

To address the mechanism of loss of RUNX2-induced MEK inhibitor resistance, we established a collection of CBFB and RUNX2 KO clones in LoVo in an ERN1 KO background (Figure 3A). For subsequent analyses, we used CBFB KO clone B and RUNX2 KO clone A. We tested the growth properties of these clones in the presence of MEK inhibitor selumetinib using both a colony formation assay (Figure 3B) and an IC50 assay (Figure 3C). These assays show that the MEK inhibitor resistance induced by the loss of RUNX2 or CBFB surpasses intrinsic resistance of the parental line, as IC50 values of the parental line fall in between values for ERN1 KO and ERN1/RUNX2 DKO or ERN1/CBFB DKO cells.

Figure 3

Characterization of LoVoanddouble KO clones. (A) Western blot analysis of LoVo ERN1/RUNX2 and ERN1/CBFB double KO clones. (B) Colony formation assay of ERN1/RUNX2 and ERN1/CBFB double KO cells in the presence and absence of the MEK inhibitor selumetinib, relative to control ERN1 KO cells and LoVo parental cells. Shown is a representative example of three biological replicates. (C) IC50 growth curves of LoVo ERN1/RUNX2 and ERN1/CBFB double KO cells as compared with LoVo ERN1 KO and LoVo parental cells in the presence of indicated concentrations of AZD6244 (selumetinib). Error bars represent standard deviation of three independent experiments. (D) Multidimensional scaling (MDS) plot of RNA sequencing data from ERN1 KO, ERN1/CBFB KO and ERN1/RUNX2 KO cells treated with 1 μM MEK inhibitor selumetinib or vehicle control. (E–H) Live cell proliferation assays of (E) LoVo parental, (F) LoVo ERN1 KO, (G) LoVo ERN1/RUNX2 double KO, and (H) LoVo ERN1/CBFB double KO cells in the presence and absence of indicated concentrations of MEK inhibitor AZD6244 (selumetinib). Error bars show standard deviation of three experiments.

To investigate possible transcriptomic changes giving rise to this drug resistance phenotype, we performed RNA sequencing of CBFB and RUNX2 knockout clones (both in the presence and absence of MEK inhibitor). Analysis of RNA sequencing data revealed a strong transcriptional similarity between CBFB and RUNX2 knockout clones indicating a similar resistance mechanism in both clones (Figure 3D). Interestingly, DKK-1 (Dickkopf-1), a secreted inhibitor of the WNT/β-catenin pathway, is the most significant differentially expressed gene when comparing CBFB and RUNX2 knockouts with their parental counterparts (false discovery rate <10−6), as shown by a volcano plot in Figure S1A. In both treated and untreated conditions, expression levels of DKK-1 were significantly higher in knockout clones (Figure S1B), indicating that RUNX2 might act as a repressor of DKK-1.

To further quantify the degree of MEK inhibitor resistance in CBFB and RUNX2 KO cells, we performed IncuCyte® real-time cell proliferation assays using a dose range of the MEK inhibitors selumetinib and trametinib. We compared the response of parental cells to ERN1 KO cells and confirmed their increased sensitivity to a wide concentration range of the MEK inhibitors selumetinib (Figures 3E and F) and trametinib (Figures. S2A and B). On the other hand, ERN1/RUNX2 DKO cells and ERN1/CBFB DKO cells showed uninhibited proliferation in the presence of selumetinib (Figures 3G and H) or trametinib (Figures S2C and D) treatment.

In order to delineate the mechanism of RUNX2-loss–induced MEK inhibitor resistance, we first analyzed phosphorylation status of a panel of receptor tyrosine kinases using a Human Phospho-Receptor Tyrosine Kinase Array Kit (R&D). The results showed that ERN1/RUNX2 DKO cells exhibit higher levels of phospho-HER3 (ERBB3) and phospho-RET compared with their parental ERN1 KO cells, and these differences were even more pronounced under selumetinib treatment (Figure 4A). We validated these results by immunoblotting with specific antibodies in both long-term (Figure 4B) and short-term assays (Figure 4C). Interestingly, prolonged treatment of LoVo cells with MEK inhibitor (14 days) resulted in markedly reduced levels of RUNX2 protein, suggesting that low RUNX2 expressing cells were positively selected under MEK inhibitor pressure (Figure 4B). In addition, we note that ERN1/RUNX2 DKO cells exhibit sustained levels of phospho-ERK and phospho-RSK despite MEK inhibitor treatment, explaining their poor response to these drugs. Interestingly, this is correlated with high phospho-EGFR as compared with parental LoVo and ERN1 KO cells. To further confirm the intensity of RTK signaling in RUNX2 and CBFB KO cells, we blotted for phospho-SHP2 (encoded by the PTPN11 gene), which is essential in signal transduction from the receptor tyrosine kinases in the cell membrane to the RAS-MEK-ERK pathway [23]. Our results show persistent SHP2 activity in RUNX2 and CBFB KO cells compared with MEK inhibitor sensitive ERN1 KO counterparts. On the other hand, parental LoVo cells exhibit intermediate levels of phospho-SHP2 (Figure 4C).

Figure 4

Loss ofinduces multiple receptor tyrosine kinases (RTKs). (A) LoVo ERN1 KO and ERN1/RUNX2 KO cells were cultured in the presence or absence (untreated = UT) of 1 μM MEK inhibitor selumetinib (1 μM Sel.). After 24 hours, cells were collected and protein lysates were processed with Human Phospho-Receptor Tyrosine Kinase (RTK) Array Kit (R&D) according to manufacturer's instructions. (B) Western blot showing the effects of long-term MEK inhibitor treatment on LoVo, ERN1 KO and ERN1/RUNX2 double KO cells. Cells were cultured in the presence and absence of 1 μM selumetinib for the indicated time (in days). (C) Western blot showing the effects of short-term treatment with MEK inhibitor on LoVo parental and LoVo ERN1 KO cells compared with LoVo ERN1/RUNX2 and ERN1/CBFB double KO cells. Cells were cultured in the presence and absence of 1 μM selumetinib for the indicated time (in hours). All experiments shown, except RTK array analysis, were performed independently at least twice.

In order to find out whether SHP2 activity is causally related to MEK inhibitor resistance observed in RUNX2 negative cells, we used SHP2 inhibitor SHP099 and combined it with MEK inhibitor selumetinib in LoVo ERN1 KO (Figure 5A) and ERN1/RUNX2 DKO cells (Figure 5B). As expected, LoVo ERN1 KO cells were not growing even in relatively modest amounts of selumetinib. However, ERN1/RUNX2 DKO cells, whose growth was uninhibited in MEK inhibitor alone, exhibited complete growth arrest when SHP2 inhibitor was combined with the MEK inhibitor (Figure 5B). Considering that we observed high activity of ERBB3/HER3 in ERN1/RUNX2 DKO cells (Figure 4A), and that ERBB3 is a kinase impaired protein that signals through obligatory heterodimers with other members of the ERBB receptor family, we used EGFR and HER2 inhibitors to see if those could recapitulate the effects of SHP2 inhibition. Interestingly, however, we could not identify a single RTK inhibitor that could mimic the phenotype observed with SHP2 inhibition (Figures S3 A–C). Only combination of MEK inhibitor with pan-RAF inhibitor could replicate synthetic lethality observed with SHP2 and MEK inhibitor combination (Figure 5C). To quantify these effects in time-dependent fashion, we used IncuCyte® real time cell proliferation imaging and quantified cell confluence for LoVo ERN1/RUNX2 DKO cells in the presence and absence of MEK inhibitor selumetinib, SHP2 inhibitor, or their combination. These data again indicated that the combination of a SHP2 and a MEK inhibitor is highly effective in the context of RUNX2 null colorectal cancer cells (Figure 5D), highlighting the central role of RTK signaling in the resistance to MEK inhibitors of RUNX2 KO cells.

Figure 5

SHP2 is an essential driver of loss ofinduced MEK inhibitor sensitivity. (A) Colony formation assay of LoVo ERN1 KO cells in the presence of indicated concentrations of the SHP2 inhibitor SHP099 and MEK inhibitor selumetinib. (B–C) Colony formation assays of LoVo ERN1/RUNX2 double knockout (DKO) cells in the presence and absence of the MEK inhibitor selumetinib (shown vertically) and in the indicated concentrations of (B) SHP2 inhibitor SHP099, and (C) pan-RAF inhibitor LY3009120. Shown are representative examples of three biological replicates. (D) Live cell proliferation assay comparing the growth of LoVo ERN1/RUNX2 double KO cells in the presence and absence of indicated concentrations of MEK inhibitor selumetinib (AZD6244), SHP2 inhibitor SHP099 and their combination (UT = untreated). Error bars show standard deviation of three experiments. (D) Model showing the interplay between RTK activation, SHP2, RUNX2 and RAS/RAF/MEK/ERK signaling pathway.

Discussion

This study reveals that RUNX2, a critical transcription factor for bone development, and its cofactor CBFβ can regulate MEK inhibitor response in colorectal cancer cells. Previous work has shown that RUNX2 is aberrantly expressed in cancer cells as compared with normal mammary epithelial cells [24]. In cancer cells, RUNX2 has thus far been shown to regulate beta-casein [25], osteopontin [26], sialoprotein [27], calcitonin, and RANKL [28]. More recently, RUNX2 was shown to be required for the growth of multiple myeloma, a malignancy driven by the accumulation and proliferation of abnormal plasma cells in the bone marrow. The suppression of RUNX2 inhibited the progression of the disease and the expression of metastasis-promoting RUNX2 target genes RANKL and DKK-1 [29]. However, the role of RUNX2 in regulating the expression or activation of receptor tyrosine kinases or MEK inhibitor sensitivity, in the context of colorectal cancer, has thus far not been identified. In melanoma, however, RUNX2 knockdown by short hairpin RNAs resulted in RTK downregulation [30]. The same study demonstrated that melanoma cells resistant to the BRAF V600E inhibitor PLX4720 had a significant increase in RUNX2 expression which was associated with an increase in both RTK expression and activation. This is the opposite of the RUNX2 effect on RTK activity, we find in KRAS-mutant colorectal cancer cells. A differential response between melanoma cells and colon cancer cells in terms of their response to inhibition of the MAPK pathway has been shown before by our lab and others [[31], [32], [33]]. In colorectal cancer, unlike melanoma, BRAF inhibition causes rapid feedback activation through the epidermal growth factor receptor (EGFR). For this reason, BRAF inhibitor treatment is ineffective as a monotherapy in colorectal cancer and combination treatment approach is needed. A combination of BRAF inhibitor encorafenib, MEK inhibitor binimetinib, and an EGFR inhibitor cetuximab recently resulted in a successfully completed phase 3 clinical trial [34], underscoring the importance of the regulation of RTK activity in colorectal cancer.

A study by Kim et al. [35] has shown that deletion of Mek1 and Mek2 kinases resulted in severe osteopenia and cleidocranial dysplasia, similar to that seen in humans and mice with impaired RUNX2 function. In the present study, we show that RUNX2 directly controls MAPK pathway activity. We first show that loss of RUNX2 reverts the MEK inhibitor sensitivity phenotype of LoVo ERN1 KO cells. Moreover, lack of RUNX2 expression increased resistance of parental LoVo cells as well, and this was also confirmed for its binding factor CBFβ. Interestingly, all CBFB knockout clones display reduced levels of RUNX2 protein confirming previous findings that CBFβ plays an important role in the stabilization of RUNX proteins by inhibiting ubiquitination-mediated degradation [8,9]. This could explain why whole exome sequencing efforts identified CBFB as one of six most significantly mutated genes in breast cancer, right after TP53, PIK3CA, and AKT1 [36]. More studies are required to investigate whether CBFB loss of function mutations in breast cancer are also associated with increased RTK activity.

Differential gene expression analysis of RNA sequencing data point to genes whose expression pattern changes under MEK inhibitor treatment. Interestingly, transcriptional analysis of CBFB and RUNX2 knockout clones (in the presence and absence of MEK inhibitor) revealed DKK-1 (Dickkopf-1), a secreted inhibitor of the WNT/β-catenin pathway and a negative regulator of bone formation, as the most significant differentially expressed gene when comparing CBFB and RUNX2 knockouts with their parental counterparts. In both treated and untreated conditions, expression levels of DKK-1 are significantly higher in knockout clones, indicating that RUNX2 might act as a repressor of DKK-1. By contrast, Gowda et al. [29] found that reduced expression of RUNX2 correlates with a reduction in DKK-1 in multiple myeloma. Our results suggest context specificity of DKK-1 regulation by RUNX2. This is potentially significant as it impacts our understanding of WNT signaling regulation in colorectal cancer. For example, Tentler et al. [37] have shown that DKK-1 is among the core genes in the WNT pathway with increased expression in KRAS mutant CRC cells resistant to AZD6244 (selumetinib).

In this work, we demonstrate that RUNX2 functions as a repressor of RTK activity in LoVo colorectal cancer cells. We observed sustained levels of MEK/ERK signaling in CBFB or RUNX2 null cells, even in the presence of MEK inhibitor. These sustained levels of phospho-ERK (resulting in continued proliferation) coincide with high phosphorylation status of EGFR, ERBB3, and RET in RUNX2 null cells. These findings complement our previous work showing that high RTK activity confers resistance to MEK inhibitors in colon cancer cells [38]. Moreover, here we report high levels of phosho-SHP2 in MEK inhibitor–treated RUNX2 null cells indicating persistent signaling from the cell membrane which helps to explain the observed resistance phenotype.

To validate the functional significance of SHP2 activity in RUNX2 negative cells, we used SHP2 inhibitor in combination with MEK inhibitor. The observed synthetic lethality demonstrated that RTK activation in RUNX2 negative cells is causal to their lack of sensitivity to MEK inhibition. The MEK and SHP2 inhibitor combination has already been demonstrated as a powerful treatment strategy to overcome RTK activity-driven MEK inhibitor resistance; in the context of pancreatic ductal adenocarcinoma [39], KRAS-mutant non-small-cell lung cancer [40] and in wild-type KRAS-amplified gastroesophageal cancer [41]. In these cancers, combination of SHP2 and MEK inhibitors was shown to synergistically inhibit signaling through the MAPK pathway.

Results presented here suggest that resistance to MEK inhibition via loss or downregulation of RUNX2 can be circumvented by concomitant treatment with SHP2 inhibitor. This is particularly relevant when we consider that deep deletions of CBFB are seen in around 6% of malignant peripheral nerve sheath tumors and in 5% of metastatic prostate cancers (cbioportal.org). Moreover, mutations in RUNX2 are seen in ∼10% of small-cell lung cancer and in ∼10% of desmoplastic melanomas [42]. Our data reveal an unexpected relationship between loss of function mutations in RUNX2 and CBFB and activity of RTKs in colon cancer. These data may help explain why loss-of-function mutations in these two genes are seen in a variety of cancers.

Based on our findings and those in the literature, we propose a model for MAPK pathway regulation by RUNX2 (Figure 5E). Previous work has shown that RUNX2 expression and activity are positively regulated by the PI3K and MAPK pathways [17,18,43]. Moreover, RUNX2 has been shown to repress RTK signaling [44]. Our present data add to this by showing that the RUNX2/CBFB complex acts as a repressor of RTK activity. Thus, inhibition of MEK has a dual effect on RTK activity. First, ERK inhibition leads to activation of RTK signaling, as previously demonstrated [35,42]. Second, ERK inhibition, by inhibition RUNX2/CBFB activity, also leads to RTK activation (Figure 4). These two effects both counteract the effect of the MEK inhibitor, contributing to drug resistance. In summary, our data are compatible with a model in which RUNX2 mutant tumors are resistant to MEK inhibitors, but respond to the combination of MEK and SHP2 inhibitors.

Conflict of Interest

The authors declare no conflict of interest.

Credit Author Statement

Tonći Šuštić; : Conceptualization, Investigation, Methodology, Validation, Writing – original draft preparation. Evert Bosdriesz: Formal Analysis, Software, Validation, Data curation, Visualization. Sake van Wageningen: Methodology, Validation. Lodewyk F.A. Wessels: Supervision, René Bernards: Conceptualization, Writing – Reviewing and Editing, Supervision, Funding Acquisition.