Derivation of a small molecule inhibitor of TNBC growth.
In this study, we used phenotypic screening of TNBCs as an approach to identify the nature of signal transduction pathways activated in these cells and simultaneously identify compounds that can target these pathways. To accomplish this goal, we used four TNBC cell lines MDA-MB-231, MDA-MB-468, Hs-578T and BT-20 as well as CD44high/CD24low BCSCs purified cell population derived from these cell lines to screen a compound library synthesized in our laboratory to identify compounds that induce apoptosis of these cells. This strategy identified 108600 (Fig. 1A), which showed remarkable toxicity towards these tumor cells and the purified BCSC population with little toxicity to normal cells. This compound was assayed against a panel of 285 functional kinases by Reaction Biology Corporation (RBC), which revealed that 108600 is a multi-kinase inhibitor and inhibits CK2α1, α2, DYRK1A, 1B, and DYRK2 and TNIK at low nanomalar concentration (Fig. 1B). We independently tested the inhibitory activity in in vitro kinase assays using recombinant CK2α1 and CK2α2 as well as DYRK1A and DYRK1B kinases (data not shown). Our results showed that 108600 is a potent inhibitor of all four kinases, thus validating the analysis by RBC.
108600 inhibited endogenous CK2 and DYRK activity toward its substrates in a dose-dependent manner in triple negative breast cancer cells. (Fig. 1C). We evaluated the effect of 108600 treatment on established substrates of these kinases. CK2 has been shown to phosphorylate AKT1 at serine (S) 129, which promotes a hyper-activated state of AKT115, 16. Additionally, threonine (Thr) 286 of CYCLIN D1 is a known phosphorylation site for DYRK1A 17. Treatment with 108600 led to decreased phosphorylation at both of these sites, indicating that 108600 treatment suppresses endogenous CK2α and DYRK1A activities in TNBC cells.
108600 inhibits viability of TNBC CD44 high /CD24 low breast cancer stem cells.
We evaluated the effect of 108600 treatment on growth of an expanded panel of breast cancer cell lines, as well as five normal and non-transformed cells. The kinases targeted by 108600 are robustly expressed in most TNBC cell lines with relatively lower expression in normal cells (fibroblasts, immortalized human mesenchymal stem cells) (Supplementary Fig. 1). 108600 potently inhibited viability of all triple negative and other breast tumor cell lines with little or no toxicity against normal cells, including normal mammary epithelial cell lines derived from 2 different donors (Fig. 1D). The growth suppression of TNBC cells (MDA-MB231) observed with 108600 treatment was phenocopied by simultaneous transfection of MDA-MB-231 cells with three siRNA pools targeting CK2, DYRK1 and TNIK, all of which are targets of 108600 (Fig. 1E). Transfection of MDA-MB-231 cells with single siRNA pools targeting CK2, DYRK1, or TNIK individually had no or modest effect on growth, supporting the need for simultaneous multi-kinase targeting (Supplementary Fig. 2).
To specifically examine the effect of 108600 treatment on the BCSC population, we purified CD44high/CD24low cells from two TNBC cell lines (MDA-MB-231 and BT-20) by FACS and examined the effects of 108600 treatment on colony and mammosphere forming capabilities of this CD44high/CD24low fraction. 108600 potently inhibited colony growth (Fig. 1F). We also examined the effects of 108600 on patient derived TNBC cells in 2D and 3D cultures using commercially available breast cancer stem cells (XLC 479) (CD44high/CD24low/ALDH+) derived from a TNBC patient (Creative Bioarray, Shirley, NY). 108600 potently inhibited mammosphere formation by these cells in in vitro 2D and 3D cultures (Fig. 1F and Supplemental Fig. 3), which is in agreement with the data obtained from sorted populations isolated from established TNBC cell lines.
Structure of CK2a1 in complex with 108600
To gain an understanding of the mechanisms by which 108600 inhibits CK2 kinases, we expressed, purified, crystallized, and determined the structure of the human CK2a1 catalytic domain in complex with 108600 (Supplemental Fig. 4). The structure was solved by molecular replacement and refined to 1.80 Å resolution. The structure of CK2a1 has the familiar bilobal architecture of a protein kinase, composed of N-terminal lobe followed by a larger C-terminal lobe (Fig. 2B). 108600 binds in the cleft between the two domains, where ATP would normally be accommodated (Fig. 2A and 2B). The drug adopts a conformation where the NO2 group and the aromatic ring are coplanar. The density for 2,6-dichlorobenzyl ring of the drug is disordered due to the conformational flexibility of this aromatic moiety in the solvent-exposed region of the active site (Fig. 2A-2D).
Earlier work has shown that CK2 is able to use both, ATP and GTP as phosphoryl donors with similar efficiencies. The crystal structure of the human CK2a1 in complex with the ATP non-hydrolysable analogue AMP-PNP has been determined at a 1.3 Å resolution (Fig. 2E) 18. The CK2α1 (Zea mays) complex structure with GMP-PNP resolved at 2.2 Å showed that GMP-PNP and water molecules mimic AMP-PNP in the active site of protein kinase CK2 19. Our studies show that the O3 and N4 of 108600 take the positions of the O6 and N1 atoms of GMP-PNP, respectively, and maintain the same interactions with the backbone atoms of His115 and Val 116 (Fig. 2D and 2F). Moreover, the NO2 group of the drug is near the position that is occupied by the Pα group of AMP-PNP or GMP-PNP and establishes a hydrogen bond with the side chain of Lys68 (Fig. 2D, 2E and 2F). In addition, the central benzothiazinone and aromatic rings are involved in numerous hydrophobic contacts with residues such as Leu45, Val66, Ile95, Phe113, and Ile174 that line the ATP/GTP binding cavity. The carbonyl group of the thiazone ring interacts with the main chain NH group of His115 and the NH proton of the ring establishes a hydrogen bond with the main chain carbonyl group of Val116. Importantly, a water molecule (wat 1) in the drug complex forms hydrogen bonds with the OH group of the drug and the side chain of Glu81 (Fig. 2D). Wat1 is also observed in the complex with AMP-PNP (Fig. 2E). Moreover, the drug covers the area where wat 2, wat 3 and wat 4 are bound in the AMP-PNP/ CK2 α1 complex (Fig. 2E) and wat 7 in the GMP-PNP/ CK2α1 complex (Fig. 2F). To compare the binding modes of 108600 and AMP-PNP/GMP-PNP, the coordinates for all complexes were superimposed (Fig. 2G). Significantly, the drug mimics not only the shape and electrostatics of AMP-PNP/GMP-PNP, but also their hydration patterns in the CK2a1 active site pocket (Fig. 2D, 2E, 2F and 2G).
108600 induces a conformational change in CK2α1 which is not conducive to holoenzyme formation
The CK2 holoenzyme consists of two catalytic subunits (CK2α1 or CK2α2), which interact with a CK2β dimer forming a transient complex 20. Interestingly, the CK2 holoenzyme has been shown to bind substrates and activate them in a kinase-independent manner 21. CK2a1-AMP-PNP either alone or integrated into the holoenzyme has the loop connecting β4 and β5 sheets in the “open” conformation (Fig. 2H) conductive to integration with CK2β subunit (Fig. 2I and 2J). In the absence of AMP-PNP, the β4-β5 loop is in the “closed” conformation (Fig. 2H). Intriguingly, in the presence of 108600, the β4-β5 loop adopts the “closed” conformation that is obstructive to holoenzyme formation (Fig. 2H, 2I and 2J). Hence, 108600 binding to CK2a1 appears to induce a conformational change, which represents a state that is not conducive to holoenzyme formation (Fig. 2I and 2J). 108600 may thus inhibit CK2 as a both a competitive inhibitor of ATP/GTP and as an allosteric inhibitor that weakens the interaction between C2Ka and C2Kb in vivo.
108600 inhibits tubulin polymerization and causes abnormal mitosis
Small interfering RNAs directed against CK2 a and b subunits exhibit mitotic abnormalities ranging from defects in centrosome duplication, anaphase spindle elongation and chromosome mis-segregation22. CK2 holoenzyme is an established Microtubule Associated Protein (MAP) that aids in tubulin polymerization and regulates microtubule cytoskeletal reorganization 23, 24. Since 108600 induces a conformational change in the CK2a subunit potentially interfering with its ability to bind to the b-subunit, we examined the effects 108600 on CK2-holoenzyme mediated tubulin dynamics in vitro and assessed for effects on mitosis. We adapted and employed a published protocol 24, which measures CK2 holoenzyme mediated de novo tubulin polymerization. We incubated MAP-free tubulin preparations with either recombinant purified CK2a, CK2b-His-Sumo or pre-assembled (CK2a-CK2b) holoenzyme respectively and measured the extent of tubulin polymerization. We did not observe any tubulin polymerization in the presence of individual CK2 subunits (CK2a or CK2b), even after a prolonged incubation (Fig. 3A and B). However, addition of CK2 holoenzyme to tubulins resulted in enhanced polymerization, in accordance to the published observation that CK2 holoenzyme promotes tubulin polymerization. When we pre-treated the CK2 holoenzyme with 108600, we observed a decrease in CK2-holoenzyme-mediated tubulin polymerization (Fig. 3A and B). These effects suggest that 108600 inhibits CK2 holoenzyme formation and hence tubulin dynamics.
To study the mitotic phenotype induced by 108600, cells on collagen coated glass coverslips were treated with either DMSO or 108600 for varying periods of time. The cells were stained for α-tubulin and the centrosomal marker pericentrin. While DMSO-treated mitotic cells displayed normal bipolar spindle architecture, the cells treated with 108600 formed multipolar spindles and multiple centrosomes (Fig. 3C). These effects are similar to those observed with CK2 subunit inhibition reported by Bettencourt-Dias et al 22, further supporting the fact that 108600 inhibits CK2-dependent functions.
108600 treatment induces cell cycle arrest and apoptosis of TNBC cells
Given the potent inhibition of TNBC cell growth observed with 108600 treatment (Fig. 1) and effects on tubulin polymerization and mitosis (Fig. 3), we assessed cell cycle progression in triple negative MDA-MB-231, BT-474 and Hs578T cells treated with increasing concentrations of 108600 for varying periods of time. Treatment of all three cell lines with 108600 resulted in a potent, dose and time-dependent G2/M arrest (Fig. 4A and Supplemental Fig. 5) as assessed by flow cytometric analysis. The percentages of cells in the G2/M phase were clearly increased with 108600 treatment, while the fraction of cells in the G1 or S phases concomitantly decreased. Notably, 108600 treatment increased the percentage of cells in the sub-G1 fraction, indicating apoptotic cells, as well as increasing the percentage of polyploid or aneuploid cells.
Similar results were observed with organoid cultures derived from triple negative patient-derived xenograft (PDX) that express the targets of 108600 (CK2, DYRK, TNIK) (Fig. 4B, Supplemental Fig. 6). These observations strongly suggest that treatment of TNBC with 108600 leads to prolonged G2/M arrest and eventually induces cell death. To more specifically determine whether 108600 treatment induces apoptosis, we examined levels of cleaved PARP and cleaved Caspase 3 in cells treated with increasing doses of 108600 for 24-48 hours. Increased cleaved PARP was detected in both MDA-MB-231 cells and TNBC PDX organoids (Fig. 4C and D), supporting the induction of apoptosis by 108600 treatment. Interestingly, when we performed these studies with normal proliferating cells such as hMSC-hTERT (immortalized human-mesenchymal stem) or MCF-10A cells (human mammary stem cells) which express very low levels of CK2, we observed minimal or no mitotic arrest at concentrations of 1 μM or higher of 108600 (data not shown). The majority of these cells remained in the G1 phase, with a modest increase in the G2/M population. Taken together, these results suggest that 108600 does not induce mitotic arrest of normal cells, which might explain its selective toxicity to tumor cells.
Chemotherapy-Resistant TNBCs are Sensitive to 108600 Treatment
Patients with TNBC are often treated with anthracycline and taxane-based chemotherapy as standard of care. However, as response to these chemotherapies is heterogenous, we determined whether 108600 treatment would also be effective in suppressing viability and inducing apoptosis of chemotherapy-resistant TNBC cells in vitro. We generated paclitaxel-resistant variants of the MDA-MB-231 and BT-20 cell lines by exposing them to increasing concentrations of paclitaxel (PTX). Resistance was validated by cell viability, colony formation and apoptosis assays. After continuous exposure and passaging in PTX-containing medium, we obtained MDA-MB-231 and BT-20 cells (MDA-MB-231-TR and BT-20-TR), which exhibited greater than 250-fold resistance to PTX as compared to the parental cell lines (Fig. 5B).
We investigated whether cells with the identified paclitaxel-resistant phenotype were associated with increased expression of the CD44high CD24-/low stem-cell like population, similar to those of patients who are refractory to taxane therapy 25. Interestingly, protein expression analysis of parental and MDA-MB-231-TR cells showed an increase in the level of CD44 expression relative to the parental cell line, which is similar to that observed with the BCSC population (Fig. 5C). In addition, several target kinases of 108600 were also upregulated in the PTX-R variant cell line (Fig. 5C).
When parental and PTX-R MDA-MB-231 and BT-20 cells were treated with varying concentrations of 108600 for 72-96 hours, we observed that 108600 potently inhibited growth and colony formation of both parental and PTX-resistant cells with the same efficiency (Fig. 5D), suggesting that 108600 effectively overrides PTX-resistance of the MDA-MB-231-TR and BT-20-TR variant cell lines. When we performed mammosphere assays with PTX-resistant cells, we again observed that while PTX had little or no inhibitory effect on mammosphere formation, 108600 inhibited mammosphere formation in a concentration-dependent manner (data not shown).
Pilot toxicity, pharmacokinetics and in vivo efficacy studies.
We next performed studies to determine the acute maximum tolerated dose of 108600 in mice. Three sets of CD-1 female mice at 10-12 weeks of age (n=3 per group) were injected intraperitonially (ip) with a single dose of 100, 200 or 400 mg/kg of 108600. Viability, signs of toxicity and body weight were monitored over the next 7 days. The survival rate was 100%, and we did not observe any signs of toxicity or weight loss. We next administered 100mg/kg of 108600 (ip) each day for 7 days and again, did not observe any signs of toxicity. We also conducted a preliminary pharmacokinetic study where we determined 108600 plasma level concentrations where female CD-1 mice were treated with 108600 (Phosphate salt) (50mg/kg) dissolved in water via ip injection). Mice were euthanized at 10 minutes, 30 minutes, 1 hour, 2 hours, 4 hours and 8 hours post-administration and blood was collected. Plasma samples were analyzed by liquid chromatography/mass spectrometry (LC/MS) (Agilent 1200 HPLC /Agilent Triple Quad). Area Under the Curve (AUC) was calculated according to the trapezoidal rule: (1/2)(C1+C2)(t1-t2). n=2 per time point. LC/MS studies showed that the area under the curve (AUC) values were 5.96 µg/ml/hr in mice which reflects a concentration of 11.09 µM/ml/hr in the plasma. Significant concentrations of 108600 were found in the heart, kidney, liver and mammary tissues of mice up to eight hours of dosing (data not shown).
108600 inhibits the growth of TNBC in vivo
To determine the efficacy of 108600 against TNBC growth in vivo, MDA-MB-231 cells were injected subcutaneously into the mammary fat pads of female athymic (NCr-nu/nu) mice and the tumors allowed to grow for 10-14 days to a size of 100-150 mm3. The mice were then grouped (n=8) and treated every other day with 50 mg/kg or 100 mg/kg of 108600 or an equal volume of placebo. Treatment doses were chosen based on toxicity data and efficacy with respect to suppression of 108600 target kinases (Fig. 6A and Supplmental Fig. 7). Tumor measurements and body weights were determined every alternate day. 108600 is a potent inhibitor of TNBC tumor growth as judged by the inhibition of tumor growth (Fig. 6B). The study had to be terminated at the end of 22 days due to the large size of tumors in the control group, which began to undergo necrosis. Measurement of body weights and gross pathological examination of drug-treated animals showed lack of any detectable toxicity (data not shown).
Following the termination of the experiment, H&E staining and immunofluorescence was performed on tumor tissues (Fig. 6C). 108600-treated tumors showed massive apoptosis with very few tumor cells that stained for CD44. Immunofluorescence staining for cleaved caspase-3 showed significant activation of apoptosis in 108600-treated tumor tissues, confirming our in vitro findings (Fig. 4) and supporting apoptosis as a mechanism underlying the decreased growth of the tumor xenografts.
To expand the clinical relevance of 108600 as a therapeutic for TNBC, we also examined the effect of 108600 on the growth of a TNBC patient-derived xenograft (PDX), which, like the parental MDA-MB-231 cell line, is sensitive to taxane therapy. Tumor-bearing nod/scid/gamma (NSG) mice were randomly assigned to 2 groups and treated with 108600 (100mg/kg) or an equal volume of placebo as described above. The results presented in Figure 6 show that treatment with 108600 suppressed the growth of this tumor in vivo without adverse effects on animals’ body weights (Supplemental Fig. 7)
108600 suppresses aggressive TNBC PDX growth in combination with chemotherapy.
To further evaluate the efficacy of 108600 in suppressing growth of chemotherapy-resistant TNBC in vivo, we utilized clinically annotated PDX models of drug-resistant TNBC that we have been developing as part of an ongoing co-clinical study. We prioritized models based on relative chemotherapy resistance of the original patient cancer from which the PDX model was derived, as well as levels of expression of 108600 targets (CK2, DYRK and TNIK kinases, Supplemental Fig. 6). After establishing that 100mg/kg q2D dosing of 108600 is sufficient to suppress established targets of CK2, specifically pAKT S129 (Fig. 6A and B), we propagated two PDX models (CTG1883 and CTG2397) for efficacy studies based on robust expression of 108600 targets. CTG1883 is derived from a primary TNBC with intrinsic resistance to anthracycline and taxane-based chemotherapy, and CTG2397 is derived from a metastatic TNBC that recurred despite anthracycline and taxane-based chemotherapy. For these studies, mice bearing PDX tumor grafts were treated with either vehicle control, PTX (10mg/kg q4D), 108600 (100mg/kg q2D) or a combination of PTX/108600, and tumor growth was monitored over a period of 12-24 days. While paclitaxel had a moderate (CTG1883) or minimal inhibitory effect (CTG2397) on tumor growth, single agent 108600 significantly suppressed tumor growth in both models. However, the combination of 108600 and paclitaxel showed near complete suppression of tumor growth in both chemotherapy-resistant PDX models, as assessed by caliper measurement and endpoint tumor weight (Fig. 6E) without adverse effects on body weights of the treated animals (Supplemental Fig. 7).
We next evaluated the effect of 108600 on the stem-cell like population in these PDX tumors by assessing ALDH activity in organoids derived from these PDX models (Fig. 6F). 108600 treatment suppressed the ALDH-high fraction in CTG1883 organoid cultures in a dose-dependent manner. These results collectively support 108600’s efficacy against chemotherapy-resistant TNBC by suppressing the BCSC population and enhancing sensitivity to chemotherapy.
108600 suppresses growth of pre-established TNBC metastases
The results from the above in vivo studies encouraged us to investigate the efficacy of 108600 treatment against already established TNBC metastatic disease. This is particularly important as most novel therapeutics are first evaluated in the context of advanced, metastatic disease that has already been exposed to prior chemotherapy. For these studies, MDA-MB231 LM2-4/mCherry Luc cells (LM2-4/mCherry), a luciferase and mCherry-expressing derivative of MDA-MB-231 that readily disseminates from the primary site to form metastases in lung and liver, was used for modeling metastatic breast cancer 26. For these studies, LM2-4/mCherry Luc cells were implanted into the mammary fat pads of NSG mice and the primary tumors removed prior to evidence of metastasis (11 days post-implantation). Serial in vivo imaging (IVIS) was performed to monitor metastases development and once metastases were detected in all mice (primarily in the lung), the mice were randomized into four treatment groups: 1) DMSO vehicle control; 2) 108600 (100mg/kg q2D); 3) PTX (10mg/kg q4D); or 4) a combination of 108600 and PTX. Metastases were serially imaged on days 18, 23, 28 and 32 by IVIS (Figure 7A, B), using the same acquisition settings for all treatment groups. Acquisition timing after luciferin injection (15 minutes), exposure time (5 minutes), lung gates (3.3 mm horizontal x 3.3 mm vertical), were all kept constant for all treatment groups and time points to enable comparisons. At baseline, there was no significant difference in metastatic disease burden among the four groups (Fig. 7C). We also determined the ratio of luciferase signal from lung metastases at different treatment timepoints (2nd, 3rd 4th IVIS) relative to the signal obtained at time of randomization (1st IVIS) (Figure 7D). While treatment with paclitaxel had little or no effect on metastatic progression, treatment with 108600 alone had an early suppressive effect on metastatic burden but the effect was not sustained. However, combined treatment with 108600 and PTX significantly and stably suppressed growth of already established lung metastases. (Fig. 7D). These promising results show that 108600 synergizes with chemotherapy to suppress TNBC metastatic growth, and support clinical translation of combined 108600/chemotherapy treatment for stage 4 TNBC.