PRMT1 inhibition induces differentiation of colon cancer cells

Differentiation therapy has been recently revisited as a prospective approach in cancer therapy by targeting the aberrant growth, and repairing the differentiation and cell death programs of cancer cells. However, differentiation therapy of solid tumors is a challenging issue and progress in this field is limited. We performed High Throughput Screening (HTS) using a novel dual multiplex assay to discover compounds, which induce differentiation of human colon cancer cells. Here we show that the protein arginine methyl transferase (PRMT) type 1 inhibitor, MS023, is a potent inducer of colon cancer cell differentiation with a large therapeutic window. Differentiation changes in the highly aggressive human colon cancer cell line (HT-29) were proved by proteomic and genomic approaches. Growth of HT-29 xenograft in nude mice was significantly delayed upon MS023 treatment and immunohistochemistry of tumor indicated differentiation changes. These findings may lead to development of clinically effective anti-cancer drugs based on the mechanism of cancer cell differentiation.


HTS for compounds inducing differentiation of colon cancer cells. Some cell lines were reported
to be irresponsive for ALP induction by the differentiation agent sodium butyrate (SB) 25 . Therefore choosing a suitable cellular model is an essential for a HTS campaign. Two colon cancer cell lines HT-29 and HCT-116, and a normal colon epithelial cell line CCD-841 were tested for basal ALP activity. As expected, both cancer cell lines had very low ALP activity, while activity of the enzyme in CCD-841 cells was significantly higher (Fig. 1a). SB was reported as an inducer of differentiation in HT-29 cells 25 . Therefore we examined the effect of this molecule on ALP activity and found a dose dependent increase of activity together with a reduction in cell number after 5 days of treatment (Fig. 1b). Short duration (5 h) of SB treatment did not affect ALP signal or cell viability ( Supplementary Fig. S1), suggesting a modulation of the phenotype rather than a direct impact on ALP enzyme activity and cell survival. Based on this data, HT-29 was considered as a suitable candidate for the HTS screening, and SB can be used as a positive control for phenotypic modulation. 5790 compounds from different chemical libraries, including 30 compounds from The Structural Genomics Consortium (SGC) Epigenetic Chemical Probe Collection, were screened at 10 µM concentration for their ability to induce ALP activity and delay growth of HT-29 cells (Z′ > 0.4). Entinostat, a well-known histone deacetylase (HDAC) inhibitor, potently induced ALP activity, however its toxicity in the primary screen was ~ 91%, leading us to de-prioritize this compound. However, due to the superior potency of entinostat over SB, it was used in further experiments as a control. Entinostat was toxic for normal cells, while its effect on colon cancer cells was stronger (Growth Inhibition of 50% (GI50) was 8.3 µM and 1.2 µM, respectively). The most pronounced ALP-inducing hit was PRMT type 1 inhibitor MS023. Figure 2a shows that MS023 moderately increased ALP activity in comparison to entinostat (e.g. 5-fold  www.nature.com/scientificreports/ increase versus about 30-fold increase respectively at 2.5 µM). Effective concentration of 50% (EC50) of MS023 was 5.8 µM, and it significantly delays HT-29 cell proliferation (GI50 was 2.3 µM). Importantly, MS023 had no effect on either ALP activity or proliferation of CCD-841 cells. Interestingly, MS023 reduced cell growth in four colon cancer cell lines, previously reported as irresponsive to SB treatment 25 : HCT-116, Colo-205, SW-620, and HCT-15 (Table 1). This suggests that MS023 effect was not cell line specific 36 . Three dimensional cultures (3D, tumor spheroids and organoids) are believed to be an improved in vitro model since they represent a more physiologically relevant pattern of cell growth. In comparison to monolayer, 3D culture may demonstrate profound differences in proliferation, differentiation, morphology and other cellular functions 37 . HT-29 rapidly forms spheroids in an alginate-based matrix. Following formation of established spheroids (5 days), MS023, entinostat or vehicle (DMSO) were added and growth was monitored during 6 additional days. In order to detect dead cells, we performed propidium iodide (PI) staining of the spheroids shortly before imaging (Fig. 2b,c). While entinostat rapidly induced cytotoxic effect, MS023 showed reduced growth, but low toxicity. Thus, PRMT type 1 inhibition induces effects associated with cell differentiation in both monolayer and 3D culture.
Target validation: inhibition of PRMT1 mediates a differentiated phenotype. Due to the welldefined target, selectivity of a chemical probe, inactive but structurally similar compounds are extremely useful. MS094 is a close analog of MS023, and was reported to be inactive against PRMT type 1 in biochemical assays 38 . Indeed, MS094 failed to induce ALP activity in HT-29 cells and did not affect cell proliferation (Fig. 3a), confirming that engagement of PRMT type 1 is likely driving the observed phenotype. MS023 was reported to be a potent inhibitor of PRMT1, 3, 4, 6 and 8 38 . Chemical probes of PRMT3, 4 and 6 (SGC707, TP-064 and MS049) are present in the Epigenetic Chemical Probe Collection (SGC) library, and they were unable to induce a differentiated phenotype in colon cancer cells, while PRMT8 is a brain specific protein 39 . Considering the selectivity profile of the epigenetic probe library, it is likely that PRMT1 is the sole target driving the probe induced differentiation phenotype. In order to prove this hypothesis, we employed genetic ablation of PRMT1 using siRNA. As shown in Fig. 3b,c we observed a significant elevation of ALP activity and growth delay in siPRMT1 treated HT-29 cells compared to controls (Full-length blots from Fig. 3c are shown in Supplementary Fig. S2).
Target validation: cellular thermal shift assay (CETSA) validate PRMT1 as a target of MS023.. CETSA is a well-established assay for target validation and evaluation of physiological binding conditions inside of cells 40 . First, stabilization of PRMT1 by MS023 was analyzed in living HT-29 cells using temperature gradient. We found a significant PRMT1 stability shift compared with DMSO treated control (about 12 °C) (Fig. 4a). While PRMT1 band completely disappeared at the temperature of 60 °C in the group of DMSO treated cells, similar effect in MS023 treated group was observed in the temperature of 72 °C. In order to more quantitatively understand the MS023 binding affinity, we performed isothermal dose response fingerprint analysis based CETSA. Figure 4b shows MS023 dose dependent stabilization of PRMT1. Full-length blots from Fig. 4a,b are shown in Supplementary Figs. S3 and S4 respectively).
Phenotype validation: PRMT1 inhibition drives colon cancer cell differentiation. Broad changes in gene expression patterns are expected after treatments with differentiation inducing chemical probes. Since both entinostat and MS023 treatments drove an increase in ALP activity, the associated gene-expression profile of each phenotype was examined. Figure 5a shows that in spite of some overlap, the majority of genes changed differently. Clustering of genes shown in Fig. 5b demonstrates significant distinction between all treated groups. Focusing on genes with documented involvement in positive regulation of differentiation status of normal intestinal cells 41 showed that MS023 treatment resulted in elevation of more than 65% of them, while ~ 23% were differentially expressed, and none of the candidate genes was significantly decreased (Fig. 5c).
Next, we performed proteomic analysis of non-treated and MS023 treated samples. Our focus was on changes in the expression of proteins, which are considerably involved in either tumor progression or cell differentiation. As shown in Fig. 6, the majority of the proteins involved in tumor progression were significantly down regulated. Importantly, such changes were observed not only in tumor proliferation pathways, but also in cancer cell motility and invasiveness pathways, suggesting that PRMT1 inhibition might prevent metastatic progression. In contrast, expression of proteins that characterize less aggressive tumor or even normal tissue was significantly upregulated. This trend was even more evident in the proteins responsible for cell-cell contact and cellular adhesion. Yet, expression of ALP was increased ~ 20 fold in MS023 treated samples, proving the efficacy of our assay www.nature.com/scientificreports/ in the measurement of ALP activity, and strongly supporting the usage of ALP as a differentiation marker of colon cancer cells. Full list of the differentially changed proteins and list of the proteins associated with particular pathways are presented in the Supplementary Tables S1 and S2.
Intestinal epithelium and CRC cells differ in the expression and intracellular localization of Junctional Adhesion Molecules (JAMs) 20,21 . We analyzed both, gene and protein expression of JAMs (namely E-Cadherin and ZO-1). Immunostaining of cells allowed us to monitor also intracellular localization of JAMs. MS023 and entinostat treatments slightly increased JAM gene and protein expression and causing corrected intracellular localization in HT-29 cells (Fig. 7a-c). Intriguingly, HCT-116 colon cancer cell line, which was reported as not responder to SB treatment, as suggested by the lack of induction of ALP activity 25 , still demonstrated a slight increase in E-Cadherin, while ZO-1 pattern was similar to HT-29 cells (Supplementary Fig. S5). This implies that the differentiation phenomenon could be a more general effect in CRC.
Differentiation therapy is effective against human colon cancer in mouse xenograft model. MS023 efficacy in-vivo was examined using HT-29 xenograft model in nude mice. In order to calculate desirable frequency of MS023 injection, a preliminary experiment was performed, where cell growth medium containing MS023 was replaced 5 days later with a fresh one containing or not a second dose of MS023. The absence of MS023 led to restoration of the proliferative potential, while addition of a second dose kept the cells in low-proliferating status (data not shown). Therefore, the animals received 2 doses of the compound/week. The treatment started when tumors reached 6-7 mm diameter. MS023 treatment significantly delayed tumor growth (Fig. 8a) without any signs of toxicity. Animal organs (spleens, lungs, livers) were taken and inspected postmortem, and were found similar to those of normal animals, suggesting absence of systemic toxicity.
Reduced proliferation is a characteristic of cancer cells undergoing differentiation. A well-known marker of cell proliferative status is Ki67, proliferation-associated nuclear protein, which is detected exclusively in dividing cells in any cycle of cell division. Its expression strongly correlates with clinical stage of tumor and metastatic burden 42 . In accordance to a scale proposed by Jonat et al. 43 , tumor proliferation status assessed by Ki67 index is divided into 3 different categories, where tumor containing of more than 30% of Ki67 expressing cells is considered as highly proliferating, 16-30% as intermediate proliferating, and 15% or less than as low proliferating.

Discussion
While the genome-driven targeted drug discovery model was the accepted path for most industrial campaigns of the last 20 years, the number of approved drugs discovered through phenotypic screens was found to be statistically greater than those discovered through the molecular target-based approach 44 . Therefore, in the last decade, phenotypic screens have been reestablished as a main methodology in drug discovery. Although it is a challenging path, a compound with a validated target from a relevant phenotypic assay seems to be better fit for probable progression towards a drug. Phenotypic assays are frequently costly and difficult to automate, thus impacting the number of compounds that can be screened, making the choice of chemical matter of primary importance to the success of the screen. The use of bioactive compounds and chemical probes can bridge this gap, as they will sample druggable targets, and provide a focus for expansion of the chemotypes around that target. Cancer cell phenotype is significantly different from that of normal cells. There are some "hallmarks of cancer", well described by Hanahan and Weinberg 45 . The key cancer hallmark is uncontrolled cell proliferation, therefore inhibition of proliferation, induction of apoptosis and killing of cancer cells were and remain the main goals in anticancer drug discovery 46 . Unfortunately, the majority of drug candidates discovered by such methods exhibit little if any therapeutic benefit in clinical trials 47 , mostly due to unexpected toxicity 48 . In contrast, differentiation therapy is potentially less aggressive if used as a single agent or in combination with lower-dose of cytotoxic drugs. Given that majority of cancer mortality is caused due to metastases, a decrease in metastatic potential by differentiation might have a great benefit as a therapy. Yet, the majority of differentiation focused therapies were performed using leukemia models, while solid tumors have been less investigated 6 .
Epigenetic alterations are common in cancer and can promote various features such as malignant cell transformation, development of the disease, metastatic procession and resistance to chemotherapy 49 . Epigenetic modulation, via HDAC inhibitors, has been reported as a means for inducing therapeutic differentiation 21 . It was confirmed in the screen by the differentiation activity of sodium butyrate and entinostat. Whereas considered to be a highly selective HDAC type 1 inhibitor, entinostat demonstrated high toxicity against both cancer and normal colon cells. One can speculate that such low therapeutic window can be one of the reasons for failure of entinostat in clinical studies 20 . In contrast, PRMT type 1 inhibitor, MS023, significantly delayed colon cancer cell/spheroids growth without any effect on growth of normal cells (Fig. 2), and thus was selected as a probe with the potential to modulate the CRC phenotype. Since MS023 can inhibit a few different PRMTs type 1 in-vitro 38 , it was important to discover a particular PRMT subtype, which is responsible for the induction of differentiation phenomenon in cellular models. Using both siRNA and CETSA approaches, it was shown that: (1) PRMT1, a member of PRMT type 1, is a target for MS023, which was able to bind and stabilize it in living LGR5
PRMT type 1 has attracted a close attention of pharmaceutical industry as a promising anticancer target. GlaxoSmithKline (GSK) performed intensive studies across 249 cancer cell lines, representing 12 tumor types 50 . Some of the cancer lineages especially blood cancers showed a strong cytotoxicity profile, while such effect in solid tumors was less pronounced. It is remarkable that in their studies the IG50 in the majority of colon cancer cell lines was reported to be between 1 and 10 µM, which is similar to IG50 observed in our experiments (2.3 µM, Fig. 2a).
Model in vitro cell culture systems with limited predictivity can affect the ability to succeed in a clinical trial. For example, 2D monolayer cultures significantly differ from 3D spheroids in many aspects such as physical pressure; cell/cell contact; cell shapes, orientation and polarity; nutrition and oxygenation; paracrine status, etc. For example, Karlsonn et al. reported that colon cancer (HCT-116) spheroids were significantly less sensitive than cells growing in 2D monolayer 51 to four frequently used CRC chemotherapeutic drugs. Therefore, the effect of MS023 and entinostat in 3D culture of HT-29 cells was examined. This assay revealed similar effect of these compounds on cell growth in both 2D and 3D models. Importantly, MS023 and entinostat were added when www.nature.com/scientificreports/ spheroids were already formed, implying that the growth delay was not preventing spheroid formation and would more accurately represent a clinical context of treatment. In both models, entinostat had strong cytotoxic effect, while no remarkable toxicity was observed in MS023 treated cells/spheroids (Fig. 2b,c). In order to further characterize the pro-differentiation phenotype, gene expression in non-treated HT-29 cells and cells treated with either entinostat or MS023 was compared. It was found that the gene expression profile in treated groups was significantly different (Fig. 5a,b). Since downstream targets of HDAC and PRMT type 1 are different, the mechanisms of differentiation mediated by inhibition of PRMT type 1 can be different from those mediated by inhibition of HDAC, suggesting a potential for combination therapy. MS023 treatment resulted in upregulation of many genes involved in intestinal cell differentiation (about 23% of all such genes were differentially expressed) (Fig. 5c). For example, the expression of Neurogenin 3, a gene that stimulates a program of terminal enteroendocrine cell development 52 , and is critical for epithelial cell functionality 53 was elevated about 25-fold. The expression of CDX2 gene, a master regulator of intestinal phenotype that plays a tumor-suppressive role in colon cancer expression 54 was increased ~ tenfold. Another important gene upregulated under MS023 treatment is a regulator of enterocyte/colonocyte differentiation named ATOH1 (Math1/Hath1). In colorectal cancer, ATOH1 is a tumor suppressor and is silenced 55 .
The fact that MS023 treatment resulted in less aggressive phenotype of colon cancer cells was further proved by deep analysis of protein expression profile. In this study, we focused on changes in proteins, which can characterize tumor phenotype. HT-29 is aggressive, metastatic colon cancer cell line 56 . MS023 treatment significantly changed its phenotype toward less aggressive status. The proteins involved in nine critical pathways related to tumor aggressiveness were significantly down regulated. Importantly, we identified the down regulation of the proteins responsible for cancer cell movement and invasive potential. Moreover, MS023 treated cells showed increasing expression of cell-cell contact and adhesion proteins, which is a hallmark of less aggressive and more differentiated cells. www.nature.com/scientificreports/ Junctional Adhesion Molecules (JAMs) are important for proper functions of cell-to-cell tight junctions. They regulate epithelial and endothelial cell adhesion and polarity. Abnormality in JAMs are involved in the Epithelial-Mesenchymal Transition (EMT), a process that plays a crucial role in the invasiveness and metastasis of various cancers 57 . E-cadherin, encoded by the CDH1 gene, is a transmembrane protein critical for the normal function of epithelial cell-cell junction, and neoplastic processes in the colon are characterized by loss and/or aberrant localization of E-cadherin. This results in two main consequences: (1) disruption of cell-cell contacts that promotes metastatic process; (2) dissociation of E-cadherin-β-catenin complex. Ultimately, this results in loss of differentiation and activation of β-catenin mediated cell proliferation 22 . ZO-1, is a member of the Zonula occludens protein family which consists of scaffold-forming intracellular proteins located between the transmembrane proteins and the actin cytoskeleton. They bind actin, occludin, and claudins and regulate the assembly of cellular junctions. In differentiated normal epithelial cells, ZO-1 mostly localizes at the cell-cell adhesion membrane complexes. Similar to β-catenin, ZO-1 detaches from the membrane complexes in cancer cells, accumulates in the cytoplasm, and eventually translocates to the nucleus to promote proliferation and invasiveness 58 . Both gene and protein expression of JAMs as well as their intracellular localization were analyzed. MS023 and entinostat treatments slightly increased JAM gene and protein expression and corrected intracellular localization (Fig. 7). Intriguingly, approximately half of colon cancer cell lines, including HCT-116, did not respond to the treatment with known differentiation agent (SB) by increasing of ALP activity 25 . However, HCT-116 demonstrated JAMs pattern similar to HT-29 cells (Fig. 6), implying that the differentiation phenomenon could be a more general effect in CRC.
MS023 is a well characterized molecular probe 38 , however its effects were not examined in animal models. Effectiveness of a molecule in whole organisms does not always correlated with that in vitro. Absorption, distribution, metabolism, and excretion (ADME) properties of a potential drug as well as its toxicity can be nonappropriated that might result in failure of such compounds in animal studies. We observed significant delay of HT-29 xenograft growth, while there was no statistical difference in tumor volume between two groups treated with MS023 (15 and 30 mg/kg). We hypothesize that we reached a dose plateau, and increasing the frequency of injections rather dose augmentation, will strengthen anti-tumor effect of MS023. Given that numerous cytotoxic drugs lead to tumor growth delay, it was important to examine differentiation status of tumor xenograft. Proliferation index is an established marker of tumor aggressiveness 39,41 . Its measurement allowed us to conclude that MS023 treatment was able to significantly reduce the proliferative potential of HT-29 xenograft, suggesting differentiation changes in the treated tumors. Our data shows that MS023 was able to reduce tumor growth and proliferation in vivo through differentiation of malignant cells, demonstrating a tractable model of pharmacological manipulation of colon cancer differentiation, with clinical implications. Screening procedure and CDP/CTG multiplex luminescence assay. 5760 compounds from Selleck Chemicals Bioactive, MEGxp Pure natural compounds (Analyticon), Drug-Like Set (DLS, Enamine) and whole SGC Epigenetic Chemical Probe Collection chemical libraries (30 compounds) were screened using HT-29 cells by previously described multiplex method for detection and normalization of ALP level in cells 29 . Briefly, 250 cells/well were plated in 50 µl of growth medium into white/white 384-well TC plate (Greiner, #781080) and treated with either compounds or controls (sodium butyrate, DMSO). After 5 days cells were washed 4 times, lysed, and two consequent luminescent signals (CDP for alkaline phosphatase activity and CTG for cell viability) were measured. Similar procedure was used in all follow up experiments, where ALP activity and cell viability were measured. Data analysis, normalization and statistics. Screening and hit validation data were plotted and analyzed using GeneData 12 and 15 (Basel, Switzerland), and Collaborative Drug Discover (CDD) softwares (Cambridge, United Kingdom). Other statistical significances were evaluated by two tailed t test, *p < 0.05, **p < 0.01. ALP activity and cell viability data were normalized to DMSO treated controls. Genomic and proteomic methods for data analysis described in the related sections below. www.nature.com/scientificreports/ Cell staining and image processing. 250 cells/well were plated in 50 µl of growth medium into black/ transparent 384-well TC plate (Greiner, #781091) and treated with appropriated compounds/controls. After 5 days cells were fixed with 4% PFA (Santa Cruz, #30525-89-4), permeabilazed with 0.2% Triton in 2% BSA (both from Sigma Aldrich) and stained with appropriated antibodies (anti-E-Cadherin-Alexa Fluor 647 conjugated, Abcam, #ab194982; anti-ZO1, BD, #610966; Alexa Fluor 488 goat anti-mouse IgG, Invitrogen, #A11001) in accordance to the manufacture protocols. Images were taken using ImageXpressMicro confocal microscope (Molecular Devices, Sunnyvale, CA, USA) using × 20 lens with appropriate filters. Image analysis was done using MetaXpress in house developed algorithm (Custom Module Editor).

Methods
Spheroids growth in 3D culture. Green Fluorescent Protein (GFP) expressed HT-29 cells (500 cells/ well) were plated in 50 µl of growth medium supplemented with SpheraMax synthetic polymer (provided by Nissan Industries, Japan) into repellent surface plate (Greiner, #781970). Compounds were added to the formed spheroids 5 days after cell plating. In compound treated spheroids, DMSO was adjusted to the maximum concentration of 0.1%. Spheroid growth was monitored during additional 6 days and then cells were stained with propidium iodide (Sigma Aldrich, #P4864). Images were taken and analyzed as described above.  Mass spectrometry. Mass spectrometry analysis was performed as previously described 60 . Briefly, the nanoUPLC was coupled online through a nanoESI emitter (10 μm tip; New Objective; Woburn, MA, USA) to a quadrupole orbitrap mass spectrometer (Q Exactive HFX, Thermo Scientific) using a FlexIon nanospray apparatus (Proxeon). Data was acquired in data dependent acquisition (DDA) mode, using a Top10 method. MS1 resolution was set to 120,000 (at 400 m/z), mass range of 375-1650 m/z, AGC of 3e6 and maximum injection time was set to 60 ms. MS2 resolution was set to 15,000, quadrupole isolation 1.7 m/z, AGC of 1e5, dynamic exclusion of 30 s and maximum injection time of 60 ms. The nanoUPLC was coupled online through a nanoESI emitter (10 μm tip; New Objective; Woburn, MA, USA) to a quadrupole orbitrap mass spectrometer (Q Exactive HFX, Thermo Scientific) using a FlexIon nanospray apparatus (Proxeon). Data was acquired in data dependent acquisition (DDA) mode, using a Top10 method. MS1 resolution was set to 120,000 (at 400 m/z), mass range of 375-1650 m/z, AGC of 3e6 and maximum injection time was set to 60 ms. MS2 resolution was set to 15,000, quadrupole isolation 1.7 m/z, AGC of 1e5, dynamic exclusion of 30 s and maximum injection time of 60 ms (as described in 60 ).
Genomic and proteomic data processing and analysis RNASeq. Poly-A/T stretches and Illumina adapters were trimmed from the reads using cutadapt 61 resulting reads shorter than 30 bp were discarded. Reads were mapped to the H. sapiens reference genome GRCh38 using STAR 62 using Ensembel annotations (v.92) with EndToEnd option and outFilterMismatchNoverLmax was set to 0.04. Counts for each gene were quantified with htseq-count 63 , using the same gtf. Differential expression analysis was done using DESeq2 64 with the betaPrior, cooksCutoff and independentFiltering parameters set to False. Raw P values were adjusted for multiple testing using the procedure of Benjamini and Hochberg. Genes that had absolute log2 fold change > 1 and p adjusted value < 0.05 were considered as differentially expressed. Bioinformatics Pipeline was run using snakemake 65 . All sequencing data that support the findings of this study have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GEO) and are accessible through the GEO Series accession number GSE142314. Proteomics raw data was processed in Maxquant version 1.6.6.0. Data was searched against the SwissProt human database (November 2018 version) appended with common laboratory contaminant proteins 66 . Fixed modification was set to carbamidomethylation of cysteine and variable modifications were set to protein N-term acetylation and oxidation of methionine. Search results were filtered to achieve maximum false discovery rate of 1% at the protein level. Protein LFQ intensities were calculated based on unique peptides. The LFQ values were further processed in Perseus version 1.6.2.3. A Student's t-test, after logarithmic transformation, was used to identify significant differences in LFQ intensities across the biological replica. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE 67 partner repository with the dataset identifier PXD016799. Pathway enrichment: Canonical Pathways enrichment and functional analysis were performed using Ingenuity-Pathway-Analysis (IPA) software (QIAGEN Inc., CA, US). Genes were considered differentially expressed according to the threshold of p value < 0.05 and fold change > 2 or < − 2. www.nature.com/scientificreports/ subcutaneously in the right flank of animal. Tumor volumes were calculated based on the formula: tumor volume = (Length × Width × Height × 3.14/6). Following randomization into study groups (n = 5 per group), when the mean tumor size reached ~ 150 mm 3 (6 days after tumor inoculation), animals were injected i.p. with either vehicle (saline) or indicated doses of MS023 twice/week. Animals were weighted twice/week and weight was monitored during the experiment.

HT
Immunohistochemistry. At 19 th day after tumor inoculation, mice were sacrificed and internal organs were visually examined. Tumor xenograft tissues were formalin-fixed and paraffin-embedded using standard procedures. 4 μm rehydrated samples were boiled in 10 mM citric acid solution for 10 min, blocked with 20% normal horse serum (NHS, Vector Laboratories, #S-2000) in PBS with 0.3% triton for 90 min, and avidin/biotin blocking kit (Vector Laboratories, #SP-2001) following by incubation with anti-Ki67 abs (Cell Marque, #275R-15) 1:50 overnight in RT. Then sample were treated with ABC kit (Vector Laboratories, PK-6100) for 30 min and stained with 3,3′-diaminobenzidine (DAB, Sigma, #D4293) for 3 min. Finally, cell nucleus were stained with hematoxylin by standard staining procedure. DAB images (6/sample) were randomly acquired using Nikon E600 microscope equipped with Nikon ds-Fi2 camera with both 10 × and 20 × lenses. Acquired RGB images were then splitted. The blue channel was inverted and saved in 16 bit TIF format. Image analysis was done using Metamorph "cell scoring" application, where cells expressing Ki67 were segmented with intensity threshold eightfold higher than total cells (Ki67 positive cells). Number of analyzed cell in any field was > 500.