Targeting MET kinase with the small-molecule inhibitor amuvatinib induces cytotoxicity in primary myeloma cells and cell lines

Previously studies have correlated plasma HGF levels with MM clinical parameters such as diagnosis[20‐23] disease stage, aggressiveness[22, 24, 25], prognosis[22, 23, 26], and response[26‐29]. While expression of both HGF and MET transcripts has been shown to be present in myeloma cells[18, 19] and HGF mRNA has also been demonstrated to be expressed in bone marrow stromal cells[39] the levels of HGF and MET mRNA in patient plasma cells have not been well evaluated nor correlated with disease status.

To determine the levels of MET and HGF gene expression in malignant and normal plasma cells, we analyzed data from the Mayo Clinic Patient Dataset available in the public domain[40, 41]. The 162 samples evaluated represented 15 healthy individuals (normal), 22 patients with monoclonal gammopathy of undetermined significance (MGUS), 24 with smoldering MM (SMM), 74 with newly diagnosed MM (MM-N), and 27 with relapsed/refractory MM (MM-R). Among these five groups, there was no significant difference (P = 0.708) in the expression of MET in the CD138+ cells (Figure 1A). In contrast, there was a significant trend (P = 2.5 × 10^-06) for increases in HGF mRNA levels in CD138+ plasma cells, with progressive severity of disease from healthy donors to patients with relapsed or refractory MM (Figure 1B). Within each group, there was heterogeneity in HGF expression as evinced by the 75th percentile mark. It is interesting to note that even samples with lower HGF mRNA levels in the plasma cells, typically had higher levels than the samples from healthy individuals; the 25th percentile for the myeloma patient HGF levels was > the 75th percentile for healthy individuals.

Gene array analysis along with numerous other studies by us[32, 33] and others[22, 23, 26], identified the HGF/MET axis as a therapeutic target in myeloma. To test this, we assessed the sensitivity of primary myeloma cells to the MET-kinase inhibitor, amuvatinib. CD138+ (myeloma plasma cells) and CD138– (non-malignant cells) cells were isolated from bone marrow samples obtained from eight myeloma patients (Table 1) and treated with 25 μM amuvatinib for 24 h. This dose was chosen based on findings that >95% of the compound is bound and sequestered by serum proteins (unpublished data) as well as a small preliminary screen in myeloma cell lines (data not shown). CD138+ cells from six out of eight patient samples showed a cell death induction in the amuvatinib treated cells compared to time-matched dimethyl sulfoxide (DMSO)-treated control cells as measured by annexin V/propidium iodide (PI) staining (Figure 2A). Cell death in these six samples increased by 20% to 67% compared to time-matched controls; patients 3 and 5 showed <10% increase in cell death.

An assessment of clinical characteristics of the myeloma patients did not reveal any correlations with amuvatinib-induced cell death, including prior treatment, patient age, or cellularity of the bone marrow (Table 1). We were able to measure the HGF levels in plasma samples from patients 2, 4, and 5 which were 11.9, 1.7, and 1.4 ng/mL, respectively. Although the number of total samples is small, there seems to be a relationship between levels of HGF and amuvatinib-induced apoptosis. Additional studies are needed to determine if there is any correlation between HGF level and sensitivity of CD138+ cells to MET inhibition.

In contrast to the sensitivity of CD138+ cells to amuvatinib, CD138– cells did not show any sizeable inductions (all, <10%) of death compared to time-matched controls (Figure 2B), suggesting that non-malignant bone marrow cells are not affected by amuvatinib. Overall, these results indicate that treatment of CD138+ MM cells with a MET inhibitor is detrimental to their survival.

To determine whether death of CD138+ cells was associated with an effect on the target, total and phosphorylated-MET (p-MET) expression levels were determined by flow cytometry. There were sufficient numbers of cells from patients 6 and 8 to perform this assessment. In both samples, MET phosphorylation was reduced on Tyr 1234/1235 in the CD138+ cells by 40% and 50%, respectively, as compared to the time-matched controls (Figure 3A and B). In contrast, there was no detectable level of p-MET in CD138– cells from patient 8 as compared to isotype control (Figure 3C). The lack of p-MET in CD138– cells likely explains why this population of cells was not affected by amuvatinib treatment in any of the eight patient samples.

Since the number of the primary CD138+ cells in eight patient samples were insufficient for performing a detailed investigation of HGF/MET signaling, we decided to further investigate the effects of amuvatinib in a myeloma cell line. Because our results in the patient samples suggested that higher levels of HGF may be associated with an increased sensitivity to MET inhibition, we used the U266 cell line which expresses high levels of HGF[19]. Compared to DMSO-treated control cells, amuvatinib-treated U266 cells showed a dose- and time-dependent decrease in growth (Figure 4A). The growth inhibition was ~40% at a dose of 5 μM after 48 and 72 h of incubation and 50% at a dose of ~7 μM at 72 h.

We also tested whether the observed amuvatinib-induced growth inhibition was associated with an alteration in the cell cycle. At low micromolar doses (3 and 5 μM), U266 cells were arrested at G₁ after 48 and 72 h (Figure 4B and C). In the DMSO-treated controls, approximately 65% of cells were in G₁ phase at 72 h, while cells treated 3 μM amuvatinib significantly increased to 75% in G₁ phase at the same time point (p < 0.05). Incubation with a higher level of amuvatinib (25 μM) resulted in a lower percentage of cells in G₁ phase (Figure 4C), with a concomitant increase in the subG₁ fraction (data not shown).

To determine whether the observed cell cycle changes were associated with an effect on DNA synthesis, we measured incorporation of thymidine in total DNA. Compared to DMSO-treated (control cells), amuvatinib-treated cells had decreased thymidine incorporation at doses of both 5 and 25 μM (Figure 4D), which was significantly higher for cells treated with 5 μM amuvatinib for 24 and 72 h, (P = 0.008 and P = 0.048, respectively). At 24 h, the inhibition of thymidine incorporation was greater than 50% with 5 μM amuvatinib. Additionally, as expected, the decrease in the cells’ S-phase DNA replicative capacity was discernible 24 h before there was a measurable change in the cell cycle profile as the doubling time for U266 cells is ~36 hours.

Similar to what was seen in the primary CD138+ cells, U266 cells treated with 25 μM amuvatinib exhibited significantly greater cell death (28%, 40%, and 54%) than DMSO-treated controls (6%, 7%, and 7%) for 24, 48, and 72 h, respectively (Figure 5A and B) (P = 0.045, 0.015, and 0.018 respectively). This apoptotic induction was blocked by a pan-caspase inhibitor, ZVAD, suggesting a role for caspases in amuvatinib-mediated cell death (data not shown). Consistent with the annexin V/PI staining results was our finding that amuvatinib induced poly ADP ribose polymerase (PARP) cleavage in these cells in a dose-dependent manner. Under full serum conditions (10% fetal bovine serum (FBS)), an induction of PARP cleavage was seen after 24 h with doses as low as 3 μM amuvatinib with the highest level of cleaved PARP product was seen at 25 μM (Figure 5C).

Because 95% of amuvatinib binds serum proteins and is unavailable to the cells, we also examined the efficacy of amuvatinib under low serum (0.1% FBS) conditions. When cells were treated with 5 μM amuvatinib for 16 hours in the presence of low serum, but in the presence of endogenous autocrine HGF, we found maximum induction of PARP cleavage (Figure 5D). To further confirm the cytotoxic effects of amuvatinib in myeloma cells is associated with inhibiting the HGF/MET signaling axis, we compared the efficacy of amuvatinib to induce apoptosis in U266 cells versus RPMI-8226/S, a myeloma cell line which express 75% lower levels of HGF and 95% lower levels of MET (Additional file1: Figure S1). As expected, there was a significant amuvatinib dose-dependent apoptosis-induction in the U266 cells after 48 h treatment (Figure 5E) (25 μM versus 0, 2.5, 5, and 10; P = 0.022, 0.018, 0.013, and 0.029, respectively). In contrast there was only a minor apoptosis induction in the RPMI-8226/S cells which was not statistically significant (P = 0.053, 0.300, 0.427, and 0.503, respectively). These results suggest that the apoptosis induction is due to targeting MET kinase which U266 cells are addicted to while RPMI-8226/S cells are not.

Bone marrow stroma provides a protective environment for MM cells[42]; thus, it is important to assess the efficacy of therapeutic agents in the context of a stromal environment. To assess this, we treated U266 cells co-cultured with and without stromal cells with amuvatinib for 48 h and measured viability by using flow cytometry analysis of annexin V/PI staining. Under these conditions, the U266 do not attach to the stromal cells, but are protected by them through both cell to cell contact and by various soluble factors produced by the stromal cells[43]. Amuvatinib induced ~50% cell killing during this time period and co-culture with the stromal cells provided no protection from this effect (Figure 5F). In contrast, these stromal cells were able to protect U266 cells from bortezomib treatment as they reduced the amount of bortezomib-induced apoptosis from ~75% to ~40% (Additional file1: Figure S2A). To determine whether amuvatinib had an effect on the survival of stromal cells, stromal cells cultured alone were treated with amuvatinib, harvested by trypsinization, and similarly assessed for viability. Interestingly, amuvatinib had a very minimal effect on the survival of this population of cells (Figure 5F), though they express MET (Additional file1: Figure S2B). These results indicate that the tumoricidal action of amuvatinib was largely restricted to the U266 myeloma cells, whereas the stromal cells, which are not addicted to MET, are not affected by this inhibitor. Furthermore, the stromal cells were not able to protect the U266 cells from amuvatinib’s tumoricidal activity.

Since amuvatinib also inhibits PDGFR and KIT, we validated MET kinase inhibition as the primary cause of cell death by using imatinib as a negative control. In addition to ABL, imatinib is known to also inhibit PDGFR and KIT but not MET[44]. In contrast to amuvatinib, 25 μM imatinib did not induce significant cell death (Figure 5G; P = 0.07) indicating that amuvatinib-mediated cell death is not due to its effects on PDGFR and KIT.

To further investigate the effect of amuvatinib on MET signaling, we first measured MET receptor tyrosine kinase activity in U266 cells by flow cytometry. Similar to the results seen with the CD138+ cells from the patient samples, a reduction in MET phosphorylation was detected in cells treated for 24 hr with 25 μM amuvatinib when cells are grown in normal growth conditions (10% FBS) (Figure 6A and B). This decrease of p-MET was associated with cell death as cell death was induced with 25 μM amuvatinib when cells are grown in normal growth conditions (10% FBS) (Figure 5B).

To assess the effects of amuvatinib on HGF-specific signaling, protein lysates from U266 cells serum starved in 0.1% FBS for 16 h with and without various concentrations of amuvatinib followed with 15 min HGF stimulation were examined by immunoblot analysis. The results showed that under serum starved conditions, treatment with 5 μM amuvatinib, decreased phosphorylation of the processed ~140 kDa MET β-chain at Tyr1349 by ~60% (Figure 6C and D). Because of the autocrine stimulation of MET by the endogenous HGF produced in these cells, MET was phosphorylated under serum-starved conditions even without the addition of exogenous HGF. Furthermore, an amuvatinib-dependent decrease of total MET levels of ~30% was also observed. A ~170 kDa phosphorylated MET band was detected at ~2 fold higher levels than the 140 kDa band in untreated U266 cells. A comparison with total MET shows both bands were present but the levels of the total 140 kDa band was ~4 times greater than the levels of the 170 kDa band. Although unprocessed pro-MET, containing both the α and β subunits, has been detected by SDS PAGE as a 170 kDa band, it has not been associated with kinase activity. Conversely, a splice variant of MET containing an additional 54 nt of exon 10 has been reported to be expressed at low levels[45]. This splice form produces a MET isoform that has kinase activity, though it cannot be processed into α and β subunits. In U266 cells, amuvatinib inhibited phosphorylation of a 170 kDa MET by ~70% (Figure 6C and D, B). Again, the decrease of HGF-specific phosphorylation of both isoforms of MET under low serum conditions is associated with cell death under low serum conditions (Figure 5D). The lower concentration of amuvatinib needed to decrease MET phosphorylation under 0.1% serum versus 10% serum conditions is in agreement with binding of the drug by serum proteins. Additionally, the concentration of amuvatinib required to decrease MET phosphorylation correlates with the concentration required to induce cell killing under either growth conditions. These results suggest the amuvatinib-induced cell death was associated with reduced MET activity.

Previous studies have shown that inhibition of MET causes a reduction in the phosphorylation of both AKT and extracellular signal-regulated kinases (ERK)1/2 in the MAPK signaling pathway[11]. The regulation of AKT activity by MET plays a prominent role in promoting cell survival. Moreover, MET regulation of the ERK pathway is important for proliferation and both the ERK1/2 and AKT pathways are involved in MET-induced cell spreading and motility. To examine AKT activity, p-AKT (S473) levels were measured in U266 cells by flow cytometry. Similar to the results seen with p-MET, a reduction in AKT phosphorylation was detected in cells treated for 24 hr with 25 μM amuvatinib when cells were grown in normal growth conditions (10% FBS) (Figure 7A and B). An assessment of amuvatinib’s effects on HGF-specific signaling was also performed in the U266 cells cultured in 0.1% FBS for 16 h with and without various concentrations of amuvatinib followed with 15 min HGF. Immunoblot analysis again showed that lower concentrations amuvatinib is needed to decreased AKT phosphorylation at Ser473 (Figure 7C and D), even though in these cells the levels were low and difficult to detect. Interestingly, total AKT decreased by 60% with amuvatinib treatment. To better assess the effect of amuvatinib on the AKT pathway, we examined the phosphorylation of an AKT target, glycogen synthase kinase 3 β (GSK3β) on Ser9. Amuvatinib-treated cells showed, in addition to reduction of AKT, a 65% decrease in phosphorylation of GSK3β, with a 24% decrease in total GSK3β (Figure 7C and E).

A similar assessment of phospho-ERK1/2 levels under HGF specific signaling demonstrated that amuvatinib inhibited phosphorylation of both the 44-kDa and the 42-kDa ERK isoforms by 55% and 50%, respectively, while total ERK1/2 levels did not significantly change (Figure 7C and F). These results demonstrate that amuvatinib treatment inhibits both ERK1/2 and AKT signaling through the MET pathway.

Amuvatinib (MP470) was obtained from Astex Pharmaceuticals, Inc. (Dublin, CA) and was dissolved in DMSO (Sigma Aldrich, St. Louis, MO). Because of stability constraints, amuvatinib solution was prepared fresh for each experiment. Imatinib was purchased from Novartis (St. Louis, MO) and was dissolved in DMSO and stored in aliquots at -20°C. [³H] thymidine (60 Ci/mmol) was obtained from Moravek Biochemical Inc. (Brea, CA).

The myeloma cell lines U266[61] and RPMI-8226/S[62] were obtained from Dr. William Dalton at H. Lee Moffitt Cancer Center (Tampa, FL). NK-tert human bone marrow stromal cells were obtained from Dr. Jan Burger at UT MD Anderson Cancer Center[63]. The cell lines were maintained as described[32, 63] and routinely tested for Mycoplasma infection and authenticated by short tandem repeat analysis by UT MD Anderson Cancer Center’s Characterized Cell Line Core facility.

Myeloma cell-stromal co-cultures were performed using U266 cells and NK-tert cells at a ratio of 20 to 1. Stromal cells were plated at a concentration of ~2 × 10² cells/mm² surface area 5 hours before adding U266 cells at a 20 fold higher concentration. The cells were co-cultured for 2 h prior to treatment with or without amuvatinib or bortezomib for 48 h. At the end of incubation, the U266 cells, which are free floating in these cultures, were carefully removed for analysis, leaving the adherent stromal layer undisturbed. Additionally, the stromal cells were also harvested by trypsinization and similarly assessed.

The effect of amuvatinib treatment on cell growth inhibition was measured in exponentially growing U266 cells. Cells were counted using a Coulter counter (Beckman Coulter, Fullerton, CA). DNA synthesis was measured using [³H]thymidine incorporation as described[64].

Expression data from 162 CD138+ bone marrow plasma cell samples from healthy individuals as well as patients with MGUS, SMM, MM-N, and MM-R, which were measured by using Affymetrix U133A microarrays, were downloaded from GEO (GSE6477)[40, 41]. Robust Multichip Average (RMA) algorithm was used for normalization/quantification of the data. The maximal values for the respective probe-sets of MET and HGF were used for gene expression profiling. The Kruskal-Wallis test was applied to assess whether expression of MET and HGF was associated with defined clinical groups, and results are presented as box-plots.

Primary samples were obtained from both male and female myeloma patients being treated at MD Anderson Cancer Center (Table 1). Patient samples were obtained using an MD Anderson Cancer Center Institutional Review Board approved protocol. All patients signed an informed consent form to provide peripheral blood and bone marrow samples. After collection of bone marrow samples, CD138+ cells were isolated as described[65], suspended in RPMI 1640 with 10% human AB serum (Cambrex Biosciences, East Rutherford, NJ) and used immediately for experiments. Peripheral blood samples were collected from patients 2, 4, and 5 for assessment of plasma HGF levels.

The effects of amuvatinib on HGF dependent signaling were assessed in U266 cells that had been serum starved for 24 h in RPMI 1640 containing 0.1% FBS; for the last 16 h of starvation the cells were treated with various concentrations of amuvatinib or DMSO. They were then treated with 50 ng/ml HGF for 15 min to stimulate MET. Amuvatinib-mediated induction of PARP cleavage was performed on U266 cells cultured in full serum (10% FBS) as well as under low serum conditions (0.1% FBS). Protein lysates and immunoblots were prepared as previously described[66]. Experiments were performed in triplicates, and bands were quantified by using an Odyssey Infrared Imaging System (LI-COR Biosciences, Lincoln, NE). Primary antibodies were: mouse monoclonal antibodies to MET clone 3D4 (Invitrogen, Carlsbad, CA); GSK-3β clone 7/GSK-3b, PARP clone C2-10, cleaved PARP Asp 214 clone F21-852, AKT clone 9Q7 (BD Biosciences Pharmingen, San Diego, CA); GAPDH clone 6C6 (Abcam, Inc, Cambridge, MA); phospho-ERK1/2 (Thr202/Tyr204) clone E10 (Cell Signaling Technology, Danvers, MA); β-actin clone AC-15 (Sigma Aldrich); rabbit monoclonal antibodies to phospho-GSK-3β (Ser9) clone 5B3 (Cell Signaling Technology); rabbit polyclonal antibodies to phospho-MET (Tyr1349) (Invitrogen); ERK1/2, and phospho-AKT (Ser473) (Cell Signaling Technology).

Intracellular protein expression in U266 cells was measured using BD Cytofix/Cytoperm Fixation/Permeabilization Kit (BD Biosciences). Primary antibodies used were anti-phospho-HGF R/c-MET (Tyr1234/1235) (R&D Systems, Minneapolis, MN), MET (C-12): sc-10 (Santa Cruz Biotechnology, Santa Cruz, CA), phospho-AKT (Ser473), AKT antibody (Cell Signaling Technology); and caspase-9 (Ser196) (Santa Cruz Biotechnology). Secondary antibody was a fluorescein isothiocyanate–conjugated (FITC) Affinipure goat anti-rabbit (Jackson ImmunoResearch, West Grove, PA). Cell cycle analysis and annexin V/propidium iodide (PI) staining were performed, respectively, as described[32, 33]. All flow cytometry analysis was performed using a Becton Dickinson FACSCalibur flow cytometer (San Jose, CA, USA). Statistical significance of changes was assessed by paired t-test analysis using Prism software (Graphpad, San Diego, CA).

HGF levels in primary patient plasma were determined using the Human HGF Immunoassay Kit as per the manufacturer’s protocol (Invitrogen). The absorbance of this horseradish-peroxidase based assay was measured at 450 nm. Each sample was assayed in triplicate.