A small molecule inhibitor of ARF GTPase protein 1 limits liver and colon cancer cell growth and metastasis
Hui Peng, Jyoti Chhimwal, Wei Fan, Jiaohong Wang, Lucía Barbier-Torres, Sonal Sinha, Avradip Chatterjee, Yi Zhang, Maria Lauda Tomasi, José M. Mato, Ramachandran Murali, Shelly C. Lu
TL;DR
A new small molecule disrupts a key protein interaction in liver and colon cancer cells, reducing their growth and spread in mice.
Contribution
A novel small molecule inhibitor targeting GIT1-MAT2B interaction is developed for liver and colon cancer treatment.
Findings
Compound 3 selectively interacts with GIT1 and shows anti-cancer effects in liver and colon cancer cells.
C3 inhibits tumor growth and metastasis in mice models of liver and colon cancer.
C3 disrupts GIT1-MAT2B interaction and affects downstream signaling pathways like RAS-RAF-MEK-ERK.
Abstract
ARF GTPase protein 1 (GIT1) is a scaffold protein that is overexpressed in hepatocellular carcinoma (HCC) and colorectal cancer (CRC). GIT1 forms a complex with methionine adenosyltransferase 2B (MAT2B) that activates RAS-RAF-MEK-ERK signaling in HCC and CRC to enhance tumorigenicity. Here, we investigated in a proof-of-concept study whether a small molecule that disrupts GIT1-MAT2B interaction can be effective in HCC and CRC treatment. Since the GIT1 crystal structure is unavailable, we developed a molecular model and used computer-based drug discovery approach to screen for small molecules targeting the GIT1 ankyrin repeat domain, the region closest to where MAT2B interacts that is accessible. Of nine compounds tested, compound 3 (C3) selectively interacts with GIT1 and shows an anti-cancer effect in a GIT1-dependent manner. C3 is antiproliferative, induced apoptosis and G2/M cell…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6- —https://doi.org/10.13039/100000054U.S. Department of Health & Human Services | NIH | National Cancer Institute (NCI)
- —https://doi.org/10.13039/100006955U.S. Department of Health & Human Services | NIH | Office of Extramural Research, National Institutes of Health (OER)
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsProtein Kinase Regulation and GTPase Signaling · Endoplasmic Reticulum Stress and Disease · Cancer Research and Treatments
Introduction
Liver cancer is the fourth most common cause of cancer-related death and the second most lethal tumor with a 5-year survival of 18% [1]. Majority of hepatocellular carcinoma (HCC) patients present late and only 30% are candidates for surgical resection but still face a recurrence rate of about 50% at 3 years [2]. Colorectal cancer (CRC) is the second and third most common cancer diagnosed in women and men, respectively [3]. Approximately 35% of CRC patients present with metastatic colorectal cancer (mCRC) at time of diagnosis and up to 50% of patients with initially localized CRC develop mCRC later [4]. Much progress has been made in treating mCRC and advanced HCC, but the overall prognosis remains poor. Although traditional chemotherapies can prolong survival, they often have significant adverse side effects [2]. Therefore, targeted treatment strategies using small molecule inhibitors are still highly sought after.
In our search to characterize cellular pathways that are important for liver and colon cancer growth, we identified ARF GTPase protein 1 (GIT1), a scaffold protein, and methionine adenosyltransferase 2B (MAT2B) to be overexpressed in HCC and CRC [5, 6]. The human GIT1’s structure is composed of an amino-terminal zinc finger-like motif, an ADP ribosylation factor (ARF) GTPase-activating protein (GAP) domain, three ankyrin (ANK) repeats, and a conserved carboxyl-terminal region that interacts with paxillin. GIT1 affects the function of both G protein-coupled receptors and tyrosine kinase-coupled receptors. Acting as a scaffold, GIT1 links the binding of these receptors by agonists to ERK1/2 activation by associating with MEK1 [7]. MAT2B was previously known to encode for a regulatory subunit (β) that regulates the activity of the methionine adenosyltransferase 2 A (MAT2A) encoded isoenzyme, but we found its encoded protein is also a binding partner of GIT1 [5]. Overexpression of MAT2B enhanced growth of HCC [8, 9]. The GIT1-MAT2B complex interacts with all of the components of the RAS-RAF-MEK-ERK signaling pathway and activates each step [5, 6]. This proved to be important in tumorigenesis because overexpression of MAT2B with GIT1 exerted an additive effect on growth and metastasis in an orthotopic HCC model [5]. Conversely, knockdown of endogenous MAT2B or GIT1 lowered MEK1 and ERK1/2 activity and reduced growth [5].
In the present study we identified small molecule inhibitors that can disrupt GIT1-MAT2B interaction and examined their effects on HCC and CRC tumorigenesis. Our work revealed a small molecule (compound 3 or C3) can effectively disrupt GIT1 from interacting with MAT2B and RAF/MEK/ERK to halt cancer cell growth in vitro and in vivo. Furthermore, we uncovered a previously unknown function of GIT1. We found GIT1 interacts with cyclin B1 and anaphase promoting complex subunit 3 (APC3) and APC’s co-activator CDC20 but treatment with C3 enhanced GIT1-cyclin B1 interaction while lowering GIT1-APC3/CDC20 and cyclin B1-APC3/CDC20 interactions, resulting in sustained cyclin B1 expression and mitosis arrest. These studies suggest that a small molecule that can distrupt GIT1 proteome will not only provide insight into the fine aspects of GIT1 function, but also will form the basis to develop therapeutics in the future.
Results
Design and screening of small molecule inhibitors of GIT1
GIT1’s crystal structure is unavailable; hence we used computer-based drug discovery approach. In order to design compounds that could interfere with MAT2B binding to GIT1, we first investigated the binding domain of GIT1 for MAT2B. GIT1 has several domains that can bind with different signaling molecules (Fig. 1A), among which the Spa2-homology domain (SHD) was identified as the MEK1 binding site [10]. We reported that MAT2B binds to GIT1 and MEK1, thus we surmised that MAT2B-GIT1 binding site also involves the SHD domain. To test this, we overexpressed Flag-GIT1 (1–635aa), Flag-GIT1 (250–770aa), Flag-GIT1 (420–770aa) fragments in HepG2 cells and performed immunoprecipitation by anti-flag antibody. As shown in Fig. 1B, Flag-GIT1 (1–635aa) and Flag-GIT1 (250–770aa) pulled down endogenous MAT2B successfully, while Flag-GIT1 (420–770aa) failed, suggesting that 250–420aa of GIT1 is critical for MAT2B binding. Since the SHD domain lies within the 250–420aa region of GIT1, we believe MAT2B also binds to the SHD domain.Fig. 1. Interaction of GIT1 with MAT2B and design of compound 3 (C3).A Design of C3 targeting the accessible ANK domain of GIT1. B FLAG-tagged GIT1 constructs were overexpressed in HepG2 for 48 hours, various lengths flag-GIT1 were immunoprecipitated by anti-FLAG antibody, endogenous MAT2B binding to GIT1 fragments were detected by western blot. C C3 binds to GIT1-GFP but not GIT2-GFP on MST. D EdU assay after treating non-malignant (HEK293, AML12) and malignant (HepG2, MzChA-1, HT-29, RKO) cells with varying concentrations of C3 for 24 hours. Data are presented as mean values ± SEM. *p < 0.05 from at least 3 independent experiments. E MTT assay after treating primary human and mouse hepatocytes and liver and colon cancer cells with varying concentrations of C3 for 24 hours. Results are mean ± SE from at least 3 independent experiments, *p < 0.05 vs. DMSO.
GIT1 SHD domain is inaccessible for small molecules. However, our structural model suggests that MAT2B binding involves at the interface of ANK and SHD domain. Based on the result we chose the ANK domain for the design of small inhibitory molecules (Supplemental Fig. 1). We identified two compounds, C3 and C9, that showed a dose-dependent inhibition of ERK activity in HepG2 cells (Supplemental Fig. 2). We next compared C3 to C9 on viability and growth in cancer cells (HepG2, Hep3B, MzChA-1, MC38, HT-29, RKO) and found C3 much more potent than C9 (Supplemental Fig. 3A, B) while not affecting growth in AML12 cells (Supplemental Fig. 3B) hence we used C3 for all subsequent analyses. To ensure C3 specifically binds to GIT1, we used MST, which allows measuring ligand binding to fluroscoscently labeled protein in a cell-lysate MST [11], and compared between GIT1 and GIT2, which share 65% sequence identity and 85% similarity [10], and was used as a negative control. We found C3 binds selectively to GIT1 with K_D_ of 6.2 µM but not to GIT2 (Fig. 1C). The selectively can be explained by C3’s interaction with GIT1 at nonconserved amino acid residues (Supplemental Fig. 4). In vitro, C3 inhibited growth of multiple cancer cell lines (HepG2, Hep3B, MzChA-1, RKO, HT-29) with inhibition occurring at 10 µM but not in non-malignant AML12 or HEK293 cells (Fig. 1D). C3 did not reduce discernable viability in primary mouse or human hepatocytes at concentrations up to 100 µM (Fig. 1E). In the cancer cell lines C3 exerted a dose-dependent inhibition on cell viability with IC_50_ of around 20 µM (Fig. 1E, Supplemental Fig. 3A).
C3 arrested cell cycle progression at G2/M phases and induced cell death only in liver and colon cancer cells
Next, the effect of C3 on cell cycle progression was tested in HepG2, RKO, AML12, and HEK293 cells. Cells were synchronized by serum starvation for 48 hours, released into serum containing media, and treated with either 5 µM of C3 or DMSO control for 24 hours. In both HepG2 and RKO cells treatment with C3 lowered % of cells in G1 phase, while G2/M phase was significantly increased, suggesting G2/M phase arrest in cancer cells. This did not occur with AML12 or HEK293 cells (Fig. 2A). G2/M phase arrest was induced by concentrations as low as 1-2 µM and was dose dependent in HepG2 and RKO cells (Fig. 2A).Fig. 2C3 induces cell cycle arrest at G2/M, inhibits clonogenicity and migration in cancer cells.A HepG2, RKO, AML12 and HEK293 cells were synchronized by serum starvation for 48 hours, released into serum containing media and treated with 1–5 µM of C3 or DMSO for 24 hours prior to FACS. Results are from representative experiments, done independently at least three times. *p < 0.05 vs. DMSO. B RKO, HepG2 and AML12 cells were treated with 1–10 µM C3 for 24 hours and maintained in drug-free medium for 10 days for clonogenicity assay. Results are from n = 3. *p < 0.05 vs. DMSO. C RKO and MzChA-1 cells were treated with 1–2 µM C3 for 24 hours for migration assay. Results are from n = 3. *p < 0.05 vs. DMSO.
To investigate whether C3 causes cell death, an annexin V-FITC/PI binding assay was performed. HepG2, RKO, AML12, and HEK293 cells were treated with DMSO or C3 at 5 μM for 24 to 48 hours, stained with Annexin V-FITC and PI, and analyzed by flow cytometry. We found 5 µM C3 induced significant apoptosis in RKO cells that increased with time (Supplemental Fig. 5). For HepG2 cells, apoptosis and necrosis significantly increased after 48 hours but the absolute values remained low, whereas C3 did not increase apoptosis or necrosis in non-malignant cells (AML12, HEK293) or primary mouse and human hepatocytes after 48 hours (Supplemental Fig. 5).
C3 inhibits both clonogenic potential and migration of cancer cells
The clonogenic cell survival assay determines the ability of a cell to proliferate indefinitely [12]. To assess if C3 affects cellular clonogenic potential of cancer cells, RKO and HepG2 cells were treated with C3 at one to 10 µM for 24 hours and maintained in drug-free medium for 10 days. Compared to DMSO control, C3 inhibited the clonogenic potential of both cancer cells even at 1 µM, and the effect was dose dependent, but not in AML12 cells (Fig. 2B). Cancer cells are characterized by migration from their origin to distant sites causing invasion and metastasis [13]. To assess if C3 can influence cancer cell migration, we performed the scratch assay using RKO and MzChA-1 cells and found a dose-dependent inhibition (Fig. 2C).
C3 disrupts MAT2B/GIT1 scaffold and inhibits MEK and ERK activity
Since C3 was designed to bind to the GIT1 ANK domain (Fig. 1A), adjacent to the SHD domain where MEK and MAT2B bind, we tested if C3 can influence MAT2B-GIT1 interaction. We found C3 reduced MAT2B-GIT1 interaction in HepG2 (Fig. 3A, B) and RKO cells (Supplemental Fig. 6A). To test if RAF/MEK/ERK/GIT1 complex was affected by C3, co-IP was performed by anti-MEK1/2 antibody in HepG2 and RKO cells after 24 hours C3 treatment. There was less recruitment of cRAF, BRAF, ERK1/2 and GIT1 to MEK1/2 (Fig. 3C, D). The results demonstrated C3 disrupted GIT1-MAT2B scaffold and RAF/MEK/ERK recruitment.Fig. 3. Effects of C3 on MAT2B-GIT1 and MEK1/2-GIT1-RAF-ERK interactions, MEK and ERK activities.A, B HepG2 cells were treated with 10 µM C3 for 24 hours and co-IP with anti-MAT2B or anti-GIT1 antibodies followed by immunoblotting (IB). Graphs show densitometric changes from at least 3 independent experiments in mean ± SE. *p < 0.05 vs. DMSO. C, D HepG2 and RKO cells were treated with 10 µM C3 for 24 hours followed by co-IP with anti-MEK1/2 antibody and IB for cRAF, BRAF, ERK1/2 and GIT1. Graphs show densitometric changes from at least 3 independent experiments in mean ± SE. *p < 0.05 vs. DMSO. E, F AML12, HepG2, RKO, and MC38 cells were treated with 10 µM C3 for 24 hoursfollowed by western blotting using GIT1, pMEK1/2 (Ser218/222), MEK1/2, pERK1/2 (T202, Y204), ERK1/2, cyclin D1 and Actin antibodies. Graphs on the right show densitometric changes from at least 3 independent experiments in mean ± SE. *p < 0.05 vs. DMSO.
Since the GIT1-MAT2B scaffold is important for activation of the RAS/RAF/MEK/ERK signaling pathway [5, 6], we next examined the activity of MEK and ERK in HepG2, RKO, MC38 cancer cells, and AML12 cells. MEK activity fell in HepG2 and RKO cells, but not in AML12 cells (Fig. 3E), ERK1/2 activity and its downstream target cyclin D1 also fell in HepG2 and MC38 cells but not in AML12 cells (Fig. 3F) or human hepatocytes, the latter up to 50 µM (Supplemental Fig. 6B).
C3 induces cyclin B1 stabilization and activates cyclin B1/CDK1 but arrests cells in M phase as degradation of cyclin B1 is inhibited
Cyclin B1, the regulatory subunit of CDK1, is essential for the transition from G2 phase to mitosis. Just prior to mitosis, a large amount of cyclin B1 is present in the cell, but it is inactive due to phosphorylation of CDK1 by the Wee1 kinase [14]. The complex is activated by dephosphorylation by the phosphatase Cdc25 on two conserved residues, Thr 14 and Tyr 15 of CDK1, and trigger the cells to enter into M phase [15]. Since C3 caused G2/M arrest, we first examined factors that are known to cause G2/M arrest. DNA damage is known to elevate nuclear cyclin B1 content and cause G2 arrest [16]. We first checked whether C3 caused DNA damage by measuring phosphorylated (Ser139) histone 2AX (γH2AX) protein level, which is known to increase following DNA damage [16]. C3 did not cause any DNA damage in HepG2 or RKO cells (Fig. 4A). Treatment with C3 caused a dose-dependent increase in cyclin B1 level, which is due in part to a slight increase in cyclin B1 mRNA level (<30%) and mostly to stabilization of cyclin B1 protein (Fig. 4B–D). Higher cyclin B1 led to higher complex formation with CDK1 (Fig. 4E, F), activating CDK1 activity (indicated by loss of p-Thr14/p-Tyr15) (Fig. 4G) only in cancer cells, and higher cyclin B1 nuclear level (indicated by higher phospho-cyclin B1 S116 in the cytoplasmic retention sequence that is known to promote nuclear import/inhibit nuclear export of cyclin B1/CDK1 [17]) (Fig. 4H).Fig. 4. Effect of C3 on DNA damage, cyclin B1 expression, CDK1 activity and interaction of GIT1 with cyclin B1.A HepG2 and RKO cells were treated with DMSO or C3 for 24 h and blotted for phospho-γH2AX (S139) and cyclin B1. B, C HepG2 cells were treated with DMSO or C3 (2–5 µM) for 24 hours. Cyclin B1 expression by RT-PCR (B) and western blotting (C). Numbers below are mean ± SE densitometric values. D HepG2 cells were pretreated with cycloheximide (CHX) and then with DMSO or C3 (5 µM) and blotted for cyclin B1 over time. Graph to the right shows cyclin B1 t_1/2_ is prolonged with C3. E, F HepG2 and RKO cells were treated with 5 µM C3 for 24 hours, followed by co-IP with anti-CDK1 antibody and western blots. G AML12, MC38 and HepG2 cells were treated with 2–10 µM C3 for 24 hours and western blotted for pCDK1 (Thr14/Tyr15), total CDK1 and Actin. H HepG2 cells were treated with C3 for 24 hours and western blotted for p-cyclin B1 (S116). I HepG2 cells were treated with C3 (5 µM) for 24 hours followed by co-IP with anti-GIT1 or anti-cyclin B1 and blotted for the same proteins. The densitometric changes in (G–I), normalized by input. J In vitro translated GIT1, recombinant MAT2B and cyclin B1 (both 0.5 µg) were used to evaluate their direct binding. The results are from at least 3 independent experiments in mean ± SE. *p < 0.05 vs. DMSO.
To exit M phase cyclin B1 must be degraded by the E3 ubiquitin ligase APC/C [17, 18]. APC/C requires co-activators, CDC20 or CDH1, to be active [19]. APC/C is composed of 14 proteins [19, 20]; within this complex APC3 is the center for regulation by recruiting CDC20 and CDH1 [21]. We next examined whether C3 influenced interaction of cyclin B1 with GIT1 and APC3, a key component of APC/C. Interestingly, GIT1 interacts with cyclin B1 at baseline and C3 treatment enhanced their interaction (Fig. 4I). Using recombinant proteins, we found GIT1 directly interacts with cyclin B1 (Fig. 4J). GIT1 also interacts with APC3 and CDC20 under basal conditions and whereas C3 enhanced GIT1-cyclin B1 interaction (Fig. 4I), it lowered the interaction between GIT1 with CDC20 and APC3 in HepG2 and RKO cells (Fig. 5A, B), as well as cyclin B1 with CDC20 and APC3 (Fig. 5C). Interestingly, knocking down GIT1 resulted in more interaction between cyclin B1 and CDC20 and APC3 in both HepG2 and RKO cells (Supplemental Fig. 6B), which suggests GIT1 normally limits access of CDC20 and APC3 to cyclin B1. Importantly, C3’s ability to raise cyclin B1 level requires GIT1 (Fig. 5D), and its suppressive effect on growth was lost if GIT1 expression was first lowered (Fig. 5E). Knocking down GIT1 first also greatly attenuated C3’s ability to cause G2/M arrest (Fig. 5F). Taken together, C3 treatment enhanced GIT1 to interact with cyclin B1 and lowered interaction between APC/C-CDC20 and cyclin B1-GIT1, resulting in accumulation of cyclin B1 and arrest in M phase.Fig. 5C3 lowers interaction of CDC20/APC3 with GIT1 and cyclin B1 and its ability to raise cyclin B1, reduce growth and cause G2/M arrest requires normal expression of GIT1.A, B HepG2 and RKO cells were treated with C3 at 5 µM for 24 hours and subjected to co-IP with anti-CDC20 or anti-APC3 antibodies and immunoblotted for the same proteins and GIT1. Densitometric changes as compared to DMSO are shown below the blots, normalized by input from 3 independent experiments. C HepG2 cells were treated with C3 at 5 µM for 24 hours and subjected to co-IP with anti-cyclin B1 antibody and immunoblotted for APC3 and CDC20. The ratios of APC3 to cyclin B1 and CDC20 to cyclin B1 are compared to DMSO control after normalizing to input from at least 3 independent experiments. D–F HepG2 cells were transfected with siRNA against GIT1 for 24 hours and treated with C3 at 5 µM for another 24 hours and subjected to western blotting for cyclin B1 (D), EdU assay (E), and cell cycle analysis (F, representative flow cytometry images are shown on the left, summarized in bar graph on the right) from at least 3 independent experiments. Densitometric values are normalized to Actin for western blots. *p < 0.05 vs. SC + DMSO.
C3 inhibited colon cancer growth in vivo
To evaluate in vivo anti-cancer effect of C3, we started with a subcutaneous MC38 CRC tumor in syngeneic mouse model. C3 was administered intratumorally at 100 µM (5xIC_50_) with DMSO as control. Treatment with C3 was started after tumors had reached 0.5 cm in diameter and continued every other day for 6 days, which led to a dramatic inhibition in tumor growth (Fig. 6A). Body weights were not significantly different and H&E staining of major organs showed no tissue damage or toxicity in both DMSO and C3 treatment groups (data not shown). On H&E, the spindle-shaped MC38 tumor cells turned into small round cells upon C3 treatment (Fig. 6B), TUNEL assay showed more apoptotic cells (Fig. 6C) and PCNA staining showed C3 had a dramatic effect in reducing its staining in the tumors (Fig. 6D). Western blots showed decreased MEK1/2 activity in C3 treated tumors (Fig. 6E), indicating effectiveness of C3 in inhibiting MEK signaling.Fig. 6C3 inhibits CRC growth in multiple in vivo models.A–E Subcutaneous model in syngeneic mice. A MC38 cells (1.8×10^6^ in 100 µl) were injected subcutaneously in both flanks. C3 and DMSO were injected intratumorally every other day at 100 µM (5 x IC_50_) starting at day 7 when the tumors averaged 0.5 cm in diameter. Each mouse received C3 injection in tumors on one flank and DMSO in the contralateral tumor. Gross tumor appearance at the end and growth curves are shown, results are from 6 mice. B–E representative H&E staining of tumors (B), TUNEL staining and score (C), PCNA staining and score from n = 3 (D), western blotting of representative tumors and densitometric changes of C3 treated as compared to DMSO (E). F Human CRC implanted in livers of nude mice. RKO cells (2 × 10^6^) stably expressing luciferase were injected directly into the liver of nude mice. Twenty days later DMSO and C3 were injected intraperitoneally (ip) at a dose of 25 mg/kg daily for 5 days and imaged on day 0 and 7 after start of treatment to track tumor progression. Gross liver appearance with arrows pointing at the tumor and graph summarizing luciferase intensity are shown to the right from n = 5 per group, *p < 0.05 vs. DMSO. G–I CRLM model in syngeneic mice. MC38 cells (2.5 × 10^5^) stably expressing luciferase were injected intrasplenically followed by splenectomy. 7 days later C3 (20 mg/kg on day 1, then 15 mg/kg daily x 3 days, rest for 2 days, then 15 mg/kg × 2 days) or DMSO were administered ip. Small animal imaging after treatment started, liver with tumors and luciferase intensity from n = 5 per group (G), representative H&E from n = 3 per group (H), ALT and AST levels (I) from n = 5 per group, *p < 0.05 vs. DMSO.
To make sure that C3 is effective against human CRC cells when delivered systemically without causing toxicity, we injected RKO cells stably expressing luciferase directly in the liver of nude mice and 20 days later administered C3 intraperitoneally (ip) at 10 mg/kg/d (pilot experiment for 2 weeks, Supplemental Fig. 7) and 25 mg/kg/d for 5 days (higher dose was used since the experiment was shorter). C3 was effective in reducing tumor growth in this model without causing overt toxicity (Fig. 6F). To ensure systemic delivery of C3 can inhibit CRC metastasis in an immune competent host, we injected MC38 cells stably expressing luciferase intrasplenically into C57BL/6 mice and 7 days later began treatment with C3. We again observed inhibition in tumor growth without toxicity (shown as ALT/AST levels, and H&E of different organs, data not shown) (Fig. 6G–I). Treatment with C3 up to 21 days at 15 mg/kg/day ip also did not cause any toxicity (Supplemental Figs. 8 and 9). Lastly, we performed pharmacokinetic (PK) study after intraperitoneal injection to ensure systemic exposure to C3 was achieved. Following IP administration, C3 exhibited a maximum plasma concentration (Cmax) of 10.9 ± 1 µg/mL, which was achieved at approximately 0.67 ± 0.3 hours (Tmax). The elimination half-life (t½) of C3 was estimated to be greater than 4 hours, indicating relatively sustained systemic exposure. The area under the plasma concentration–time curve (AUC) exceeded 16 h*µg/mL. The mean residence time (MRT) of C3 was approximately 2 hours, and the compound remained detectable in plasma up to 24 hours post-administration (Supplemental Fig. 10).
Discussion
Although traditional chemotherapy and radiation therapy have been effective in cancer treatment, they can result in severe side effects. Specific molecular targeting is attractive for avoiding these side effects. Two approaches, small molecules and monoclonal antibodies (mAbs) have been studied. mAbs are large molecular weight proteins of around 150 kDa that can only act on molecules that are expressed on the cell surface [22]. Small molecule cancer drugs are much smaller in size (≤500 Da), which allows them to translocate through the plasma membrane and interact with any portion of a molecular target regardless of the target’s cellular location [23].
RAS-RAF-MEK-ERK pathway plays a fundamental role in cell proliferation and mutations in this pathway occur in many cancers, prompting approaches to block the aberrant signaling by inhibiting RAF using small molecule inhibitors such as sorafenib and vemurafenib [24]. However, clinical success has been disappointing, and inconsistent activation in RAF kinase can occur [25]. Furthermore, cancer cells often develop resistance by switching to alternate signaling pathways, thus identification of additional novel signaling molecules to treat cancers is urgently needed.
Our previous finding that MAT2B and GIT1, which are overexpressed in human liver and colon cancer, form a scaffold complex that positively regulates RAS signaling at multiple steps suggests that targeting this complex may be a novel and effective approach in cancer therapy. Although multiple scaffold proteins, such as KSR1/2, IQGAP1, MP1, and β-Arrestin1/2, have been described to regulate RAS-RAF-MEK-ERK signaling, none of them interact with all the components of this signaling pathway [26]. Furthermore, to our knowledge, it has not shown that targeting a large protein complex by a small molecule as a strategy can limit tumor growth.
We screened nine compounds and focused on C3 due its potency in cancer cells and lack of toxicity in human and mouse hepatocytes. We intentionally included cancer cell lines with different mutations and C3 was effective regardless of the mutations. RKO are CpG island methylator phenotype positive, harbor microsatellite instability [27] and HT-29 are microsatellite stable but have chromosomal instability [28]. RKO and HT-29 also differ in p53 status (RKO has wild type, HT-29 has R273H mutation) but both have activating BRAF mutation [29]. Similarly, we used HepG2 (hepatoblastoma cell line, mutated in p53 and chromatin modulators), Hep3B (HCC line, wild type p53), and MzChA-1 (gallbladder cancer cell line, wild type p53, ATM mutation) [30]. Even though C3 targets the ANK domain of GIT1, which is adjacent to where MAT2B and MEK interact with GIT1, it was able to effectively reduce the interaction between GIT1 and MAT2B as well as with MEK. Since GIT1-MAT2B complex serves as a scaffold for RAF-MEK-ERK, consistently, C3 treatment suppressed the interaction of cRAF, BRAF, ERK with MEK, lowered MEK and ERK activity and cyclin D1 expression. This is one mechanism of C3’s growth suppressive effect but not the only one.
Another mechanism for C3’s anti-cancer effect is via mitosis arrest. In liver and colon cancer cells but not in non-malignant cells, C3 treatment arrested cells in G2/M phase. This was not due to increased DNA damage, but to sustained cyclin B1 expression as a result of decreased degradation. Dynamic changes in cyclin B1/CDK1 occur during the transition from G2 phase to mitosis. The cyclin B1/CDK1 complex is activated to stimulate its own nuclear import to trigger entry into the mitosis phase [31]. To exit mitosis, cyclin B1 must be degraded by APC/C [18]. However, C3 treatment resulted in stabilization of cyclin B1 and a sustained increase in cyclin B1 expression. This most likely is the explanation for why cells were arrested in the mitosis phase after C3 treatment.
We next asked how C3 was able to block cyclin B1 degradation. One intriguing observation is the direct interaction of GIT1 with cyclin B1, which was enhanced by C3 treatment. C3 treatment also influenced access of APC/C to cyclin B1. APC/C requires co-activators, CDC20 or CDH1, to be active [19]. APC3 is one of the subunits of APC/C and the center for regulation by recruiting CDC20 and CDH1 [21]. Interestingly, GIT1 also interacts with APC3 and CDC20 under basal conditions and C3 treatment lowered their interactions, as well as the interaction between cyclin B1 and CDC20/APC3. One hypothetical scenario is that GIT1 acts as a scaffold that brings cyclin B1 and APC/C together, along with the co-activator CDC20. However, the finding that silencing GIT1 increased the interaction between cyclin B1 and APC3/CDC20 excluded this possibility. In fact, it suggests that GIT1 normally limits the interaction between cyclin B1 and CDC20/APC3 and C3 treatment further accentuates this by enhancing GIT1 to bind to cyclin B1. We speculate that C3 binding to GIT1 changes the latter’s conformation and interactome, with secondary changes in APC/C access to cyclin B1. We showed C3 binds GIT1 (not GIT2) specifically and that C3’s anti-cancer effects are dependent on its influence on GIT1, since C3’s effect was greatly attenuated if GIT1 was silenced. This rules out off-target effects. Future work in solving the crystal structure of C3-GIT1 complex will be required to address the exact molecular mechanisms of how C3 binding to GIT1 affects its interactome.
Another anti-cancer effect of C3 is induction of cell death in cancer cells. The most likely explanation is due to prolonged mitosis arrest, which is known to trigger apoptosis [32] and may even lead to mitotic catastrophe that can result in apoptosis, necrosis, and autophagic cell death [33].
To test the efficacy of C3 in vivo, we employed three different models. First was a syngeneic subcutaneous CRC model. This was done as proof of concept that C3 administration can inhibit tumor growth and lowered MEK activity. Next, we ensured C3 administered systemically would be well tolerated and effective against human CRC cells implanted directly in the liver. Our pilot experiment tested C3 at 10 mg/kg/day ip, 5 days/week for 2 weeks and found C3 highly effective so we increased to a higher dose (25 mg/kg/day for 5 days) and found C3 was still effective and well tolerated without overt toxicity. Lastly, we tested C3 administration systemically in a syngeneic colorectal liver metastasis (CRLM) model with an intact immune system and found it to work well there as well. PK analysis validated systemic C3 exposure was achieved with i.p. administration and paves the way for further optimization of C3.
GIT1 was reported to be concentrated at centrosomes at all cell cycle stages [34, 35]. Centrosome functions as the major microtubule-organizing center in animal cells, and regulates cell shape, motility, spindle formation and chromosome segregation [36]. Centrosome duplication and separation is required to complete the normal cell cycle, and this requires activation of the centrosomal kinase Aurora-A. GIT1 targeting to the centrosome was believed to scaffold PIX and recruit PAK, which then activates Aurora-A [35]. However, our results showing that GIT1 interacts with cyclin B1, APC3 and CDC20 even under basal condition, which were changed after C3 treatment suggest GIT1 may be involved in regulating the cell cycle progression more than just activating Aurora-A kinase. Much more work will be required to better understand the role of GIT1 in cell cycle progression.
Taken together, we have identified a small molecule inhibitor that targets GIT1 and disrupts its normal interactome, resulting in less RAF-MEK-ERK signaling and mitosis arrest in liver and colon cancer cells in vitro and in vivo. Moreover, we found GIT1 interacts with many regulators of the cell cycle machinery and may be involved in regulating cell cycle progression. Lastly, our study demonstrated a new approach in developing new therapeutics for liver and colon cancers.
Materials and methods
Materials and reagents
See Supplemental Table 1 for a list of the reagents and sources.
Cell lines, mouse and human hepatocytes
AML12 (murine hepatocytes), HepG2 (human hepatoma), Hep3B (human HCC), HEK293 (human embryonic kidney cell), RKO (human CRC), HT-29 (human CRC), and MC38 (mouse CRC) were obtained from the ATCC (Manassas, Virginia), while MzChA-1 (human biliary adenocarcinoma) was obtained from Dr. Gianfranco Alpini (Indiana University, Indianapolis, Indiana). Each cell line was passaged fewer than 6 months, authenticated and tested for mycoplasma contamination prior to use. See Supplemental Methods for culturing conditions.
Primary mouse hepatocytes were isolated from 3-month-old male C57BL/6 mice liver as we described [37]. Cryopreserved human hepatocytes were provided by ThermoFisher (Waltham, MA). See Supplemental Methods for culturing conditions.
Reagents and plasmids
Plasmids Flag-GIT1(1–635aa), Flag-GIT1(250-770aa), Flag-GIT1(420–770aa) were kindly provided by Dr. Bradford Berk [10].
Computer model of GIT1 and drug discovery
Three-dimensional structure of full-length human GIT1 (Uniprot: Q9Y2X7) is not available and hence a structural model was built using a homology modeling. The full-length GIT proteins have an N-terminal ARF GTPase activating protein (ARF-GAP) domain, three ANK repeats, a Spa2-homology domain (SHD), a coiled-coil domain and a paxillin-binding site [10]. Using BLAST search no satisfactory homology domain was identified to develop a complete model. Hence, an initial structural model for the ANK domain of the human GIT1 was subsequently generated using I-TASSAR [38]. The most reliable model obtained from I-TASSAR was subsequently refined using the Prime module in Schrodinger (Schrodinger, Inc. San Diego, CA). The final structural model was obtained after a cycle of minimization followed by a short 1.2 ns molecular simulation. The putative binding site of small molecules was identified with site score 0.684 (Dscore 0.630). Subsequently, the model was subjected to an extended molecular simulation (2.4 ns). Models were averaged over 0.5 ns and the sitemap score was reevaluated for each ensemble. Two models with site score 0.92 were selected for virtual screening. For virtual screening chemicals from Hit2Lead consisting of 500,0000 small molecules from Chembridge (San Diego, CA) were prepared using ligprep to select drug-like compounds. Finally, Gide using XP mode (Schrodinger, Inc., San Diego, CA) was used to identify hits. Selected compounds were synthesized by Chembridge (San Diego, CA).
GFP-GIT1/GIT2 lysate binding to C3 inhibitor using microscale thermophoresis
Microscale thermophoresis (MST)-based binding experiment was carried out using a Monolith NT.115 (NanoTemper Technologies) instrument. The assay was performed in phosphate buffered saline (PBS) + 5% Glycerol + 3.3% DMSO. The lysate containing either GFP-GIT1 or GFP-GIT2 was titrated against a serial dilution of the C3 ranging from 6 nM to 200 µM. Following an incubation for 15 minutes at room temperature, the MST measurements were taken. Data was analyzed using MO. Affinity Analysis software (NanoTemper technologies).
Cell viability assay
We used the MTT assay as described in Supplemental Methods.
5-Ethynyl-2’-deoxyuridine (EdU) proliferation assay
See Supplemental Methods.
Cell cycle analysis
See Supplemental Methods.
Apoptosis assay
See Supplemental Methods.
Western blots and co-immunoprecipitation (co-IP) assay
Western blotting was performed as previously described [37]. Immunoprecipitation studies were carried out as described previously [5, 6]. See Supplemental Methods for details.
Clonogenicity assay
Clonogenic cell survival assay was performed as described [12]. See Supplemental Methods for details.
Migration scratch assay
Cell migration assay was performed as described in Supplemental Methods.
Effect of C3 on tumor growth in syngeneic models and nude mice
All procedure protocols and the care of the animals were reviewed and approved by the Institutional Animal Care and Use Committee of Cedars-Sinai Medical Center (IACUC009378). Both sexes were used in experiments using C57BL/6 wild type mice (syngeneic models) and since results were similar they were combined in the analyses. In one experiment C3 was tested in the MC38 (CRC cell line from a C57BL/6 mouse) syngeneic mouse model. MC38 cells (1.8 × 10^6^ in 100 µl) were injected subcutaneously in both flanks. Administration of C3 and DMSO intratumorally was performed every other day at a dose of 100 µM concentration of C3 (5 x IC_50_), starting at day 7 when the tumors averaged 0.5 cm in diameter. Tumor size was measured by caliper, with the tumor volume calculated according to the formula: π/6 (length x width x height). Each mouse received C3 injection on tumors on one flank and DMSO injection on the contralateral tumor. Animals were sacrificed on day 13. Tumor tissues were used for protein analysis and some were fixed in 4% formalin for histology, immunohistochemistry (IHC) study and TUNEL assay.
In another experiment, syngeneic CRC liver metastasis (CRLM) model was created by an intrasplenic injection of 2.5 × 10^5^ MC38 cells stably expressing luciferase (Luc) in 50 μl of 0.9% saline, followed by splenectomy after injection. The size of the liver tumor and metastasis were monitored by PerkinElmer IVIS® SPECTRUM system. One week after surgery, mice were treated with C3 or DMSO vehicle control by intraperitoneal (ip) injection, and were euthanized nine days later. Plasma and tumor tissues were saved for further analysis.
To test if C3 is effective in human CRC in vivo, RKO cells (2 × 10^6^) stably expressing Luc were injected directly into the left lobe of the liver in 4-month old male nude mice and tumor growth was monitored by small animal imaging as above after treatment with C3 ip. We used only male mice for this experiment since the above experiments showed C3’s effectiveness was sex-independent.
For all in vivo tumor experiments the tumor volume did not exceed 2 cm^3^, which is the maximum tumor size allowed by our institution in mice. Mice were monitored daily for body weight and behavior. Source data for figures that describe tumour growth complied with the journal’s data availability policy and are included in the Supplemental Material.
IHC and tunel assay
See Supplemental Methods for details. Percent cells staining positive and intensity of staining were separately scored and totaled for each specimen as previously described [39] with minor modifications described in Supplemental Methods.
Alanine transaminase (ALT) and aspartate transaminase (AST) levels
ALT and AST levels from plasma of mice were measured with the ALT and AST activity assay kits (Sigma-Aldrich, MO) following manufacturer’s instructions.
In vivo toxicity study
In vivo toxicity studies were performed as described in Supplemental Methods.
Pharmacokinetic (PK) analysis
PK analysis was done by WuXi AppTech, Shanghai, China and detailed in Supplemental Methods.
Statistical analysis
Data are expressed as mean ± standard error. Data were analyzed using two-tailed unpaired Student’s t-test for comparing two groups and analysis of variance followed by Fisher’s test for multiple comparisons. The ratios of genes or proteins expression levels to housekeeping genes or proteins were calculated. Student’s t-test was used for Pearson correlation. Significant difference was defined by p < 0.05. No statistical methods were used to pre-determine sample sizes; however, they are consistent with standard practices in the field, and a minimum of n = 5 animals was systematically analyzed and mentioned in figure legend.
Supplementary information
Supplemental Figures and Legends Supplemental Methods Supplemental Table