Identification of novel small molecule inhibitors of centrosome clustering in cancer cells

Abstract

Most normal cells have two centrosomes that form bipolar spindles during mitosis, while cancer cells often contain more than two, or "supernumerary" centrosomes. Such cancer cells achieve bipolar division by clustering their centrosomes into two functional poles, and inhibiting this process then leads to cancer-specific cell death. A major problem with clinically used anti-mitotic drugs, such as paclitaxel, is their toxicity in normal cells. To discover new compounds with greater specificity for cancer cells, we established a high-content screen for agents that block centrosome clustering in BT-549 cells, a breast cancer cell line that harbors supernumerary centrosomes. Using this screen, we identified 14 compounds that inhibit centrosome clustering and induce mitotic arrest. Some of these compounds were structurally similar, suggesting a common structural motif important for preventing centrosome clustering. We next compared the effects of these compounds on the growth of several breast and other cancer cell lines, an immortalized normal human mammary epithelial cell line, and progenitor-enriched primary normal human mammary epithelial cells. From these comparisons, we found some compounds that kill breast cancer cells, but not their normal epithelial counterparts, suggesting their potential for targeted therapy. One of these compounds, N2-(3-pyridylmethyl)-5-nitro-2-furamide (Centrosome Clustering Chemical Inhibitor-01, CCCI-01), that showed the greatest differential response in this screen was confirmed to have selective effects on cancer as compared to normal breast progenitors using more precise apoptosis induction and clonogenic growth endpoints. The concentration of CCCI-01 that killed cancer cells in the clonogenic assay spared normal human bone marrow hematopoietic progenitors in the colony-forming cell assay, indicating a potential therapeutic window for CCCI-01, whose selectivity might be further improved by optimizing the compound. Immunofluorescence analysis showed that treatment with CCCI-01 lead to multipolar spindles in BT-549, while maintaining bipolar spindles in the normal primary human mammary epithelial cells. Since centrosome clustering is a complex process involving multiple pathways, the 14 compounds identified in this study provide a potentially novel means to developing non-cross-resistant anti-cancer drugs that block centrosome clustering.

Figure 1. High-content screening strategy for the identification of candidate compounds that inhibit centrosome clustering

(A-C) Analysis of BT-549 cells for de-clustered centrosomes using a Cellomics Array Scan VTI imaging platform with Thermo Scientific Compartmental Analysis algorithm. (A and B) Identification of nuclei/cells and mitotic cells. Cells were labeled with Hoechst 33342 (A) and TG-3, a marker of mitotic cells (B). The nuclei of TG-3-positive (mitotic) cells are outlined in blue. TG-3-negative (non-mitotic) cells are outlined in orange. A border was introduced by expanding 5 pixels outward from the nuclear boundary of a TG-3 positive cell to define a region of interest (ROI, green circles) (B). (C) Identification of centrosomes in mitotic cells. Centrosomes were labeled with anti-pericentrin (light blue dots) and the number of pericentrin foci within the ROIs was enumerated by the software. Cells with de-clustered centrosomes were defined as those mitotic cells with greater than two pericentrin foci. Bar = 50 μm. (A'-C') Corresponding raw images before compartmental analysis. Bar = 50 μm. (D) Representative results of a primary screen evaluating 1,200 compounds. The distribution of mitotic cells with de-clustered centrosomes for the test group is similar to that for the untreated group except that it displays additional high scores. Those compounds that increased the score by at least 2.5x the standard deviation were considered “hits” and were subjected to a secondary screen. (E) Representative results of a secondary screen. Four different concentrations (27, 9, 2.7 and 0.9 μM) were tested for each primary hit. The frequency of cells with de-clustered centrosomes and mitotic indices are shown for primary hits, A-H, and negative DMSO control. Arrowheads indicate hits in the secondary screen. Compound F was a false positive due to fluorescent debris present in the well.

Figure 2. Chemical structures of three similar hits

Three active compounds that share structural similarities

Figure 3. Effects of identified small molecules on spindle organization in vitro

(A) Representative immunofluorescence images of BT-549 cells showing the centrosome de-clustering properties of the positive compounds identified in the high-content screen. Cells were treated with the various compounds for 5 hours at a concentration of 10 μM, and then immunolabelled for pericentrin (green) and tubulin (red). Nuclei were labeled with Hoechst 33342 (blue). Bar = 10 μm. (B) Quantification of the frequency of mitotic cells with de-clustered centrosomes after treatment as described in (A). Centrosome de-clustering was analyzed using pericentrin labeling, and tubulin labeling was also used to ensure multipolar spindle formation. At least 25 mitotic cells/treatment/experiment were measured for each compound tested. Average of at least two independent experiments is shown. *P < 0.03, student's t-test.

Figure 4. Comparison of cytotoxicity of CCCI-01 in cancer cells and normal cells

(A) Cell viability in cancer and normal cells was examined after 2 days of incubation with CCCI-01 by MTT assay. BT-549 was more sensitive to CCCI-01 compared to five primary normal human mammary epithelial cells (HMECs, passage 0) from five different reduction mammoplasty samples and MCF-10A, an immortalized normal mammary epithelial cell line. *P < 0.005 (compared to normal cells), student's t-test. (B) MTT assay was carried out with finer titration steps between 1 and 10 μM. In addition to BT-549, MDA-MB-231, A549 and HCT-116 were also examined to compare cytotoxicity in different types of cancer. IC₅₀ values for cancer cells were between 2 and 8 μM, while that for primary HMEC was above 10 μM. P < 0.001 for BT-549 at 1-10 μM; P = 0.045 and P < 0.005 for MDA-MB-231 at 1 μM and 3-10 μM, respectively; P < 0.02 for A-549 and HCT-116 at 3-10 μM, student's t-test, compared to primary HMEC. (C) Time course analysis of cell cycle and cell death by flow cytometry was carried out in BT-549 treated with 5 μM CCCI-01. Mitotic arrest represented by the increased G2/M population was observed with CCCI-01 treatment, while the G1 population decreased. Apoptosis, the sub-G1 population, was elevated in the presence of CCCI-01 with prolonged treatment. The distribution of S phase population remained similar, indicating there were no significant effects on DNA synthesis by CCCI-01. DMSO had essentially no effects on these peaks (Supplemental Figure 3). A representative result from three independent experiments is shown. (D) Apoptosis was examined by Cell Death Detection ELISA (Roche). BT-549 and MCF-10A were treated with CCCI-01 or DMSO for 19 hours. Values were normalized (= 0) to the highest DMSO concentration (0.1 %) applied for CCCI-01 treatment. The level of apoptosis was increased in BT-549 as low as at 3 μM, while apoptosis was not induced in MCF-10A even at 10 μM. This assay was carried out three times, and produced consistent results. A representative result is presented. *P < 0.01, student's t-test.

Figure 5. Effects of CCCI-01 in colony formation of cancer and normal cells

(A) Clonogenic assay was carried out in BT-549 and MCF-10A treated with CCCI-01. Colonies consisting of more than 50 cells were counted after 9 days of incubation with CCCI-01 or DMSO. All treatments received 0.06 % DMSO. Surviving fraction was obtained by normalizing to DMSO control (= 1). Values are average from at least two independent experiments. P, student's t-test. The blue line in the graph indicates IC₅₀. (B and C) The CFC assay was carried out in primary normal bone marrow cells. Cells were cultured in the presence of varying concentrations of CCCI-01. (B) Surviving fractions of CFC-G/M and BFU-E/CFU-E colonies were obtained by normalizing to the DMSO control (= 1). Values are average of two independent experiments with two normal bone marrow samples: one is nomonuclear cells and the other is selected for CD34+. Separate data are available in Supplemental Table 1. For CFC-G/M, no inhibitory effects were detected up to 1 μM, and colony size and number were reduced in both samples at 10 μM. BFU-E/CFU-E colonies did not seem to be affected up to 3 μM, and showed inhibition in colony formation at 10 μM in both samples. The blue line in the graph indicates IC₅₀. (C) Representative images of each type of colonies treated with different concentrations of CCCI-01.

Figure 6. Effects of CCCI-01 on spindle polarity in BT-549 cancer cells and normal primary HMECs

Centrosome arrangement and spindle multipolarity were examined by immunofluorescence after treating cells with CCCI-01 for 5 hours. As little as 5 μM increased centrosome de-clustering to 70 % in BT-549 breast cancer cells (average of two independent experiments, at least 30 mitotic cells/treatment/experiment). In contrast, even at 8 μM, de-clustering was not induced (P > 0.7, student's t-test) in normal primary HMEC from three different reduction mammoplasty samples (average of two independent experiments with three HMEC samples in total, at least 30 mitotic cells/treatment/sample). Scale bar = 10 μm. *P < 0.02, student's t-test.

Figure 7. Analysis of inactive compounds structurally similar to CCCI-01, 02 and 03

(A) Chemical structures of two inactive compounds, Inactive-A and E, and active CCCI-01. (B) Centrosome arrangement and spindle multipolarity were examined by immunofluorescence after treating cells with Inactive-A or E or CCCI-01. Inactive-A and E did not inhibit centrosome clustering even at 30 μM, while as low as 5 μM of CCCI-01 resulted in 70% of mitotic cells with de-clustered centrosomes. Only the highest concentration of DMSO (0.6 %) applied is presented as a negative control here. All corresponding DMSO showed no detectable effects on the spindle organization. Average from three independent experiments (at least 42 mitotic cells/treatment/experiment). (C) Cell viability in BT-549 was examined after 2 days of incubation with Inactive-A or E or CCCI-01 by the MTT assay. The cell viability in BT-549 was not affected by Inactive-A and E up to 10 μM, while at this concentration, CCCI-01 reduced the cell viability to less than 10 %. At 30 μM, Inactive-A and E showed some cytotoxicity, but this is not caused by centrosome de-clustering as centrosomes did not display scattered configuration at this concentration (see Figure 6B). Average of three independent experiments.