Clofarabine inhibits Ewing sarcoma growth through a novel molecular mechanism involving direct binding to CD99

Abstract

Ewing sarcoma (ES) is an aggressive bone and soft tissue malignancy that predominantly affects children and adolescents. CD99 is a cell surface protein that is highly expressed on ES cells and is required to maintain their malignancy. We screened small molecule libraries for binding to extra cellular domain of recombinant CD99 and subsequent inhibition of ES cell growth. We identified two structurally similar FDA-approved compounds, clofarabine and cladribine that selectively inhibited the growth of ES cells in a panel of 14 ES vs. 28 non-ES cell lines. Both drugs inhibited CD99 dimerization and its interaction with downstream signaling components. A membrane-impermeable analog of clofarabine showed similar cytotoxicity in culture, suggesting that it can function through inhibiting CD99 independent of DNA metabolism. Both drugs drastically inhibited anchorage-independent growth of ES cells, but clofarabine was more effective in inhibiting growth of three different ES xenografts. Our findings provide a novel molecular mechanism for clofarabine that involves direct binding to a cell surface receptor CD99 and inhibiting its biological activities.

Introduction

Ewing sarcoma (ES) is the second most common bone and soft tissue malignancy occurring in children and young adults ^1–3. It is characterized by highly aggressive, poorly differentiated, small round blue cells ⁴. The recent advances in multimodal therapy of patients with localized disease have significantly raised survival up to 60–70% ⁵, but the survivors are prone to short- and long-term treatment-related side effects including secondary malignancies, cardiac toxicity and infertility ⁶. Unfortunately, the prognosis of patients with metastatic or recurrent ES remains poor, with a 5-year survival rate of less than 30% ⁵.

High expression of a cell surface protein CD99 on ES cells holds diagnostic significance and therapeutic potential ^{7, 8}. Immunohistochemical analysis of CD99 expression is routinely used as a diagnostic marker to distinguish ES from other tumors with similar histological appearance ^9–15. CD99 is a heavily O-glycosylated type 1 transmembrane protein that has an apparent molecular weight of 32 kDa on sodium dodecyl (SDS) polyacrylamide gel electrophoresis (PAGE) ^{16, 17}. Human CD99 has a unique characteristic in that it shows no resemblance to any known protein family except for CD99L2 and PBDX, which display only moderate amino acid identity ^{18, 19}. The mouse homolog of CD99 represents only 45% amino acid identity to its human counterpart, suggesting that the human and mouse proteins may not be functionally equivalent ²⁰. CD99 is broadly distributed on normal cells with a high expression in specific cell populations of the hematopoietic system including T- and B-lymphocytes ^{17, 21}. The functional role of CD99 in oncogenesis remains largely unknown. A true functional ligand(s) for human CD99 has not been identified yet other than CD99 itself, as leukocyte diapedesis through endothelial junctions involves homophilic interaction between CD99 molecules ²².

CD99 is highly expressed in ES and contributes to its pathogenesis ^{7, 8}. The tumor specific chromosomal translocation product EWS-FLI1 fusion protein up-regulates the expression of CD99 either by directly binding to CD99 promoter elements ^23–25 or through modulating miRNA expression ²⁶. Knockdown of CD99 expression in ES cells induces terminal neuronal differentiation and results in reduced malignant phenotype including slower growth rate on plastic, smaller colonies in soft agar and decreased xenograft growth ^{25, 27}. CD99 engagement by blocking murine monoclonal antibodies (mAb) induces massive apoptosis in cultured ES cells and inhibits tumor growth and metastasis formation in ES bearing xenograft mice ^28–31. Moreover, anti-CD99 treatment with blocking mAbs sensitizes ES cells to doxorubicin under both in vitro and in vivo conditions ³¹. Thus, a substantial body of evidence makes CD99 an attractive target for therapeutic intervention against ES.

We performed a broad chemical screening assay to identify small molecules that can directly bind to CD99 with a selective growth inhibitory activity against ES cells. We identified clofarabine and cladribine, two FDA-approved drugs currently used in the treatment of leukemia, as novel small molecule inhibitors targeting CD99 function. We validated their functional activity and demonstrated their mode of action in a comprehensive panel of biochemical and cellular assays. Finally, we demonstrated the antitumor activity of clofarabine in xenograft mouse models of ES. These results provide a novel molecular mechanism underlying the action of clofarabine on the plasma membrane that involves direct binding to the cell surface protein CD99 and inhibiting its biological activities.

Results

A Chemical Library Screen For Small Molecules that Bind to CD99 with Selective Growth Inhibitory Activity in ES Cells

We screened a set of 2,607 compounds representing diverse molecular structural classes from the National Cancer Institute’s Developmental Therapeutics Program (The NCI/DTP Open Chemical Repository at https://dtp.cancer.gov) to identify small molecules that directly bind to purified CD99 protein with selective growth-inhibitory activity against ES cell lines. A summary of the screening approach is given in Supplementary Figure S1.

We prepared recombinant extracellular domain of CD99 (CD99-ECD) from Escherichia coli in a highly purified form as a fusion protein with N-terminal Bioease and C-terminal 6xHis tags (Supplementary Figure S2A, left panel). The purified protein was analyzed to check its biotinylation by immunoblotting using streptavidin-HRP (Supplementary Figure S2A, right panel). Surface plasmon resonance (SPR) analysis confirmed that CD99-ECD protein made in bacteria was recognized by monoclonal antibodies, indicating that antigenic determinant sites were preserved (Supplementary Figure S2B). The primary screen was based on a direct binding assay using SPR technology with purified CD99-ECD protein that was immobilized on a neutravidin-coated sensor chip. A negative control protein on the neighboring flow cell of the same sensor chip was used in order to eliminate non-specific binders. Compounds with a CD99 binding level higher than 20% of R_max (the analyte binding capacity or theoretical maximum response) were included for hit selection. Primary hits were then defined as compounds that showed at least 20-fold difference in binding to CD99 over negative control protein (Figure 1A). 160 compounds that passed hit selection criteria were further tested in a functional screen to prioritize them for having higher specificity towards ES. Compounds were evaluated based on their ability to inhibit cell proliferation at 10 μmol/L concentration following 48 h incubation in ES cell lines RDES and TC-71, which express high levels of CD99, and show reduced growth response to CD99 blocking antibody treatment ^28–31. We used a negative control cell line U-2 OS that expresses low levels of CD99 and is resistant to CD99 blocking antibodies ^32,33. While a majority of the compounds did not show any cytotoxicity, some compounds killed both ES and OS cells (Figure 1B). From this screen, 2 compounds were identified as secondary hits (NSC606869 and NSC662825), which inhibited the growth of both TC-71 and RDES cell lines by >70%, and U-2 OS cell line by <10% (Supplementary Figure S1 and Figure 1B). Dose-response studies were performed with two compounds at 1, 3, 10 and 30 μmol/L concentration in ES cell lines. NSC606869 (clofarabine, Clolar^®), a FDA-approved adenosine nucleoside analog, exhibited a higher antiproliferative activity against ES cells as shown by an IC₅₀ value lower than 1 μmol/L compared to 9 μmol/L of NSC662825 (Supplementary Figure S1). Therefore, we selected clofarabine over NSC662825 as the main lead compound. Furthermore, NSC105014 (cladribine, Leustatin^®), another FDA-approved adenosine nucleoside analog, with high structural similarity to clofarabine that differ only by the absence of one fluorine atom, was present among 160 primary hits (Supplementary Figure S1). Cladribine exhibited some degree of selective cytotoxicity in our secondary screening against ES cells, and a subsequent dose-dependent cytotoxicity assay revealed an IC₅₀ lower than 1 μmol/L for TC-71 and RDES cell lines too. For these reasons, we also included cladribine in many of the follow up studies.

Figure 1. Screening of small molecule libraries for inhibitors of CD99-ECD.

(A) Recombinant CD99-ECD was captured on a neutravidin-coated sensor chip and small molecules (50 μmol/L) were individually injected over the surface to measure direct binding in a Biacore T200 instrument. Each dot represents a small molecule. Relative binding given in Y-axis was normalized for binding to an empty flow cell, a buffer alone injection and binding to a negative control protein. Any compound showing at least 20-fold difference in relative binding to CD99 over negative control protein in terms of normalized response was selected as primary hit. (B) 160 primary hits identified from the initial binding screening were further screened for their effect on cell viability in two ES cell lines (TC-71 and RDES) vs. a negative control cell line U-2 OS to distinguish nonspecific CD99 binders from functionally relevant compounds. After 48 h treatment with each compound at 10 μmol/L concentration, cell viability was determined by WST-1 assay. Single points in shaded area represent the secondary hits that reduced cell viability of ES cells more than 70% and OS cells less than 10%. Values are presented as the mean of three replicates. A representative steady-state affinity curve and sensorgram (inset) showing binding kinetics of clofarabine (C) and cladribine (D) to the purified recombinant human full-length CD99. Recombinant human full-length CD99 protein produced in HEK293 cells was immobilized by amine coupling on a CM4 Biacore sensor chip in a Biacore T200 instrument. Compounds were injected over the chip surface at 1.25, 2.5, 5 and 10 μmol/L concentrations for clofarabine and 0.625, 1.25, 2.5, 10 μmol/L concentrations for cladribine in duplicate. The final K_D values are 5.3 ± 3.2 μmol/L (n = 3 separate experiments) and 3.8 ± 3.3 μmol/L (n = 4 separate experiments) for clofarabine and cladribine, respectively.

Clofarabine and Cladribine Directly Bind to CD99 and Selectively Inhibit the Growth of ES Cell Lines

We subsequently determined binding affinity constants of clofarabine and cladribine for CD99 by detailed SPR analysis using the recombinant human full-length CD99 expressed and purified from mammalian cells in a Biacore T200 instrument (Figure 1C and D). Titration of compounds onto immobilized CD99 yielded a K_D value of 5.3 ± 3.2 μmol/L (n = 3 separate experiments) and 3.8 ± 3.3 μmol/L (n = 4 separate experiments) for clofarabine and cladribine, respectively. The use of properly folded and post-translationally modified CD99 protein produced in eukaryotic cells in binding assays is particularly more relevant for analysis of the protein-inhibitor interactions, given that CD99 is a heavily glycosylated protein. Analysis of binding kinetic data using CD99-ECD purified from bacteria also yielded comparable binding constants with K_D values of 4.6 ± 3.2 μmol/L (n = 5 separate experiments) and 3.8 ± 2.4 μmol/L (n = 7 separate experiments), for clofarabine and cladribine, respectively, thereby validating our screening approach.

To further verify the selective cytotoxic activity of clofarabine and cladribine against ES, we performed a detailed IC₅₀ analysis on a panel of 14 ES cell lines and 23 non-ES cell lines including 22 human cancer lines derived from 7 different tumor histotypes and HEK293 cell line (Figure 2A and B, and Supplementary Figure S3 and S4). Clofarabine and cladribine are approved by FDA for their use in patients with relapsed leukemia. Therefore, we included two leukemia cell lines, MOLT-4 and Jurkat, as positive controls in this experiment. Consistent with their clinical indications, both drugs inhibited the growth of leukemia cell lines with IC₅₀ of <1 μmol/L. Similar to leukemia cell lines, all ES cell lines, except for A673, which has a BRAF V600E mutation ^{34, 35}, were highly sensitive to both clofarabine and cladribine with IC₅₀ concentrations in the submicromolar range, and exhibited 31-fold and 18-fold increased sensitivity compared to non-ES cell lines, respectively.

Figure 2. ES and leukemia cell lines are more sensitive to clofarabine and cladribine.

Cell viability was evaluated by WST-1 or MTT assay following 48 h treatment with clofarabine (A) and cladribine (B). The IC₅₀ values were determined by nonlinear regression analysis using GraphPad Prism version 6.0h software. The average IC₅₀ concentrations of clofarabine were 0.44 ± 0.44 for ES, 0.18 ± 0.01 for leukemia and 13.73 ± 11.54 for non-ES cell lines. The average IC₅₀ concentrations of cladribine were 1.09 ± 1.85 for ES, 0.34 ± 0.03 for leukemia and 20.05 ± 13.16 for non-ES cell lines. The average IC₅₀ concentration for non-ES cell lines was calculated using the highest concentrations tested (either 10 or 30 μmol/L) for resistant cell lines. Values are presented as the mean ± SD of three technical replicates.

CD99 Protein Expression Correlates with Clofarabine and Cladribine Sensitivity in Cancer Cells

To investigate the relationship between the level of CD99 expression and drug sensitivity, we determined the endogenous expression of CD99 in a panel of 9 ES and 13 non-ES cell lines. The recombinant purified CD99-ECD protein was used to generate standard curves for immunoquantification of CD99 expression by immunobotting (Supplementary Figure S5A). A significant negative correlation was found between CD99 expression and IC₅₀ values for clofarabine (Spearman rho=−0.53, p=0.0120) and cladribine (Spearman rho=−0.56, p=0.0071) (Supplementary Figure S5B and C). Cells with high CD99 expression had smaller IC₅₀ values and more sensitive to clofarabine and cladribine. Cells with low CD99 expression had bigger IC₅₀ values and less sensitive to clofarabine and cladribine. CD99 expression analysis of sarcoma cell lines by flow cytometry in a panel of 7 ES, 8 OS and 4 rhabdomyasarcoma cell lines and the subsequent correlation analysis also produced significant negative correlations for clofarabine and cladribine (Supplementary Table S1).

We next determined the cytotoxicity of cladribine and clofarabine in TC-71 cells silenced for CD99 expression. We generated two stable transfectants of TC-71 cell line, TC-CD99-shRNA#1 and TC-CD99-shRNA#2, by using plasmids expressing either a shRNA targeting the 3’ untranslated region (UTR) of CD99 or a scrambled shRNA control ^{25, 27} (Supplementary Figure S6). CD99 knockdown resulted in increase of IC₅₀ values by 2.85- and 2.35-fold of cladribine and 3.70- and 1.63-fold of clofarabine in TC-CD99-shRNA#1 and TC-CD99-shRNA#2, respectively, compared with TC-71 parental cells (Supplementary Table S2). These findings validate the functional involvement of CD99 as target of clofarabine and cladribine in ES cells.

Clofarabine and Cladribine Inhibit CD99 Dimerization and Downstream Signaling

As CD99 could form homodimers through extracellular domain-mediated interactions ^{36, 37} and the homophilic CD99 interactions regulate transendothelial migration of immune cells ²², we examined whether clofarabine and cladribine could block CD99 dimerization in STA-ET-7.2 and RDES cells, which express the highest level of CD99 protein among the ES cell lines tested (Supplementary Figure S5A). To address this, in a chemical cross-linking assay, STA-ET-7.2 ES and RDES cells were preincubated with the inhibitors for 1 h at a final concentration of 5 μmol/L followed by the addition of 1 mmol/L BS³, a membrane-impermeable chemical cross-linking agent. Immunoblot analysis demonstrated that cladribine significantly reduced the formation of CD99 homodimers. CD99 dimer formation was also blocked by treatment of the cells with clofarabine, albeit at a lower level than cladribine (Figure 3A and Supplementary Figure S7).

Figure 3. Clofarabine and cladribine inhibit homodimer formation of CD99 and cyclophilin & PKA binding, and the membrane-impermeable analog of clofarabine shows potent cytotoxicity in ES cells.

(A) STA-ET-7.2 ES cells were incubated with either drug at 5 μmol/L concentration or DMSO for 1 h. The cells were treated with cross-linking agent BS³ as described in Materials and Methods. The lysates were resolved in 12% SDS-PAGE followed by immunoblotting with anti-actin and anti-CD99 clone 013 antibodies. M and D represent monomeric and dimeric forms of CD99, respectively. Values given below the lanes on the immunoblot represent the relative density of the bands, and were determined using ImageJ 1.48v software. (B) STA-ET-7.2, RDES and A4573 ES cells were treated with 3 μmol/L of either drug for 6 h. Endogenous CD99 was immunoprecipitated from total cell lysates using anti-CD99 clone 013 antibody. Immunoblot (IB) analysis was performed for cyclophilin A, CD99 and PKA-RIIα using total cell lysates (TCL) and immunoprecipitated (IP) samples. Values given below the lanes on each immunoblot represent the relative density of the bands, and were determined using ImageJ 1.48v software. (C) A steady-state affinity curve and sensorgram (inset) showing binding kinetics of clofarabine-5’-triphosphate to the purified recombinant human full-length CD99. Recombinant human full-length CD99 protein produced in HEK293 cells was immobilized by amine coupling on a CM4 Biacore sensor chip in a Biacore T200 instrument. Compound was injected over the chip surface at 2.5, 5, 10 and 20 μmol/L concentrations in duplicate. (D) Cell viability was evaluated by WST-1 assay following 48 h treatment of TC-71 and A4573 ES cells with clofarabine and clofarabine-5’-triphosphate. The IC₅₀ values were determined by nonlinear regression analysis using GraphPad Prism version 6.0h software.

The interaction of CD99 with cyclophilin A ³⁸ and protein kinase A regulatory subunit IIα (PKA-RIIα) ³⁹ is implicated in downstream signaling pathways. We next tested if clofarabine and cladribine could block the interaction of CD99 with cyclophilin A and PKA-RIIα. We immunoprecipitated endogenous CD99 from STA-ET-7.2, RDES and A4573 ES cells and immunoblotted for CD99, cyclophilin A and PKA-RIIα. Consistent with the results of cross-linking experiments, the amount of co-precipitated cyclophilin A and PKARIIα proteins decreased significantly by clofarabine and cladribine in all the ES cell lines tested, except for no change in PKARIIα protein immunoprecipitated from cladribine-treated A4573 cells (Figure 3B).

Membrane-impermeable Analog of Clofarabine Shows Potent Cytotoxicity in ES Cells

Nucleoside analogs exert their cytotoxic effects through disrupting DNA synthesis. The phenotypes we observed in ES cells in response to clofarabine and cladribine treatment were most likely the result of both inhibition of CD99 activity and DNA synthesis. We hypothesized that a membrane-impermeable analog of clofarabine and cladribine could separate anti-CD99 activity from inhibition of DNA synthesis in ES cells. Because high polarity of phosphate moieties results in extremely poor, if any, membrane permeability ^{40, 41}, we treated ES cells with a triphosphate analog of clofarabine. We first determined the binding affinity of clofarabine-5’-triphosphate for full-length CD99 as 3.7 μmol/L (Figure 3C). For the CD99-ECD purified from bacteria, the K_D was calculated as 7.6 μmol/L. The dose-response curves showed that sensitivity of the ES cells to clofarabine and clofarabine-5’-triphosphate were highly similar with IC₅₀ concentrations of 0.15 and 0.17 μmol/L in TC-71 cells and 0.31 and 0.46 μmol/L in A4573 cells, respectively (Figure 3D). Furthermore, both clofarabine and clofarabine-5’-triphosphate selectively induced apoptosis in ES cells but not in OS cells as shown by caspase 3 and apoptotic PARP-1 cleavage (Supplementary Figure S8). Therefore, membrane-impermeable analogues of clofarabine and cladribine can function extracellularly by inhibiting CD99 to induce cell death in ES without affecting DNA synthesis.

Clofarabine and Cladribine Inhibit ROCK2 Expression and the Motility of ES Cells

ROCK2 is a key regulator of actin cytoskeletal remodeling and a crucial mediator of CD99-regulated cell adhesion and migration ³³. We next sought to determine whether pharmacological inhibition of CD99 could lead to a reduction in ROCK2 expression. We treated TC-71, RDES, 6647 and IOR/CAR ES cells with clofarabine and cladribine for 24 h. Immunoblot analysis showed that there was a notable decrease in ROCK2 expression in cells treated with either drug compared with DMSO-treated control cells (Supplementary Figure S9A). To ascertain the role of CD99 in drug-induced loss of ROCK2 expression, we took a loss-of-function approach. Knockdown of CD99 in TC-CD99-shRNA#2 cells abolished the inhibitory effect of drugs on ROCK2 expression, suggesting that the loss of ROCK2 expression by cladribine or clofarabine is mediated through CD99 signaling (Supplementary Figure S9B). We next determined the inhibitory effect of clofarabine and cladribine on the migration of 6647 and TC-71 ES cells by transwell migration assay. The cells were pre-treated with anti-CD99 compounds for 24 h followed by seeding of viable cells on the upper chamber. The number of migrated cells in the bottom chamber was significantly decreased when the cells exposed to either drug compared with control cells (Supplementary Figure S9C).

Clofarabine and Cladribine Kill ES vs. OS by Different Mechanisms

We performed cell cycle analysis to characterize the mechanism by which clofarabine and cladribine exerts their growth-inhibitory effects in ES vs. OS cells. Treatment of TC-71, 6647 and RDES ES cells and U-2 OS cells with either clofarabine or cladribine for 48 h led to a significant increase of hypodiploid sub-G1 peak only in ES cells compared with U-2 OS cell line (Figure 4A). Furthermore, clofarabine and cladribine led to a cell cycle arrest in S phase with a concomitant decrease in G1 phase in all the cells tested, however, this effect was more pronounced in OS cells compared with ES cells (Supplementary Figure S10). These observations suggest that cell cycle arrest in S phase by clofarabine and cladribine may be the result of inhibition of DNA synthesis due to their nucleoside analog activity. However, these findings also reflect a cytotoxic action of drugs in ES cells through an alternative mechanism involving CD99 rather than a cytostatic effect alone as observed in OS cells, and suggest that CD99 seems to be required for shifting the treatment response from cytostatic to cytotoxic.

Figure 4. Clofarabine and cladribine induce cell death and inhibit anchorage-independent cell growth in ES cells.

(A) TC-71, 6647 and RDES cells and U-2 OS cells were treated with both drugs at indicated concentrations for 48 h. Histograms display the percentage of hypodiploid cells based on DNA contents analyzed by cell sorting. Values are presented as the mean ± SD of three independent experiments. (B) TC-71, 6647 and RDES cells were plated in semi-solid medium with either clofarabine or cladribine and all ES cells exhibited a dose-dependent inhibition in colony formation as determined by soft agar assay. The data are represented as the mean ± SD of triplicate determinations. (C) Shown are representative images of brightfield colonies from the soft agar assay (scale bar 200 μm).

Asterisks indicate statistically significant differences between treatments (*p<50.05; vs. control using a Student’s unpaired t test, two-tailed).

Clofarabine and Cladribine Suppress Anchorage-independent Cell Growth

We employed a soft-agar colony formation assay to examine the effects of CD99 inhibitors on the clonogenic survival/proliferation of cells under anchorage-independent conditions. Consistent with cell proliferation inhibition, both clofarabine and cladribine significantly inhibited colony formation of RDES, TC-71 and 6647 ES cells in a dose-dependent manner (Figure 4B–C).

In order to provide additional evidence that clofarabine inhibits anchorage-independent growth of ES cells through blocking CD99, we evaluated TC-CD99-shRNA#1 and TC-CD99-shRNA#2 cells that showed increase of IC₅₀ values compared with TC-71 parental cells in regular culture conditions (Supplementary Table S2). Both TC-CD99-shRNA#1 and TC-CD99-shRNA#2 formed fewer number of colonies compared to parental TC-71 cells (Supplementary Figure S11A). However, treatment with clofarabine did not result in further decrease in colony formation of TC-CD99-shRNA#1 cells in soft-agar. This finding suggested that the lack of CD99 protein in TC-71 cells rendered them resistant to clofarabine.

We also evaluated the effect of clofarabine-5’-triphosphate, which showed comparable cytotoxicity to clofarabine in normal culture conditions (Figure 3D). Both clofarabine and clofarabine-5’-triphosphate inhibited the anchorage-independent growth of RDES, 6647 and TC-71 ES cells in a dose-dependent manner (Supplementary Figure S11A and S11B). Similar to clofarabine, responsiveness of TC-CD99-shRNA#1 cells to clofarabine-5’-triphosphate in soft agar was diminished when CD99 expression was inhibited by an shRNA (Supplementary Figure S11A).

Clofarabine and Cladribine Inhibit Tumor Growth in Vivo

To evaluate the anti-tumor effect of clofarabine and cladribine, we established orthotopic ES xenografts of TC-71, SKES and A4573 cells in SCID/beige mice. We first sought to determine the antitumor activity of drugs through intraperitoneal (i.p.) administration given once daily. When tumor sizes reached to 150–200 mm³, the mice were randomly allocated into 3 treatment groups; DMSO, clofarabine and cladribine. Mice tolerated the drug treatment well. We did not observe any meaningful reduction in total body weight that was monitored daily (Supplementary Figure S12). The in vivo potencies of drug treatments were evaluated by event-free survival curves, where tumors that reached 1.0 cm³ in size considered an event. Clofarabine treatment significantly improved the event-free survival of mice in all three xenograft models, whereas the event-free survival of cladribine-treated mice was significantly different from that of the vehicle-treated group only in TC-71 xenograft model (Supplementary Figure S13). Treatment of mice with clofarabine induced a significant inhibition of tumor growth compared to control mice in all xenograft models (Supplementary Figure S14). By contrast, cladribine was less effective, however the activity of this compound was encouraging in TC-71 xenograft model as the tumors in cladribine-treated mice grew slower than those in vehicle-treated mice (Supplementary Figure S14A). After 14 days of treatment surviving animals were monitored for tumor growth in TC-71 xenograft-bearing mice. A tumor regrowth was detected immediately upon drug withdrawal in mice treated with either drug (Supplementary Figure S14A).

Histopathological analysis of SKES and A4573 xenografts showed typical small round blue cell morphology in H&E stained samples and strong CD99 positivity in immunohistochemistry studies (Supplementary Figures S15 and S16). We did not observe any meaningful changes in CD99 expression following clofarabine treatment. A4573 xenografts showed relatively stronger CD99 staining compared to SKES. In the DMSO control group there were more than 15 mitotic figures per field in 200-fold magnification, which were reduced to 3 or less in treatment groups. There was significant increase in apoptotic cells in the treatment group that was observed in H&E slides. Increase in apoptosis was also confirmed by caspase-3 immunohistochemistry.

We next investigated the antitumor activity of clofarabine administered orally against the TC-71 xenograft model. Clofarabine (30 mg/kg) administered orally once daily for 14 days resulted in markedly enhanced antitumor activity compared to i.p. administration as shown by more significant inhibition of tumor growth and prolonged event-free survival (Figure 5A–B). Clofarabine treatment significantly increased the median event-free survival from 7 to 33 days (p<0.0001) (Figure 5A). TC-71 xenografts also showed typical small round blue cell morphology in H&E staining and CD99 staining even stronger than A4573 xenografts (Supplementary Figure S17). In this study, reduction in number of mitotic figures and increase in caspase-3 staining was less prominent. We hypothesize that the reason for apparently reduced response to drug treatment in this group is the time of tissue harvest. In the oral study animals responded to 14-day clofarabine treatment and they survived more than 30 days. Therefore, when the animals were euthanized for tumor regrowth they were not receiving any clofarabine. Their last dose of clofarabine was more than two week before the tissue harvest.

Figure 5. Clofarabine administered orally shows potent antitumor activity in an orthotopic TC-71 xenograft mouse model of ES.

SCID/beige mice bearing TC-71 xenografts were treated once daily for 14 days with vehicle or clofarabine (30 mg/kg) administered orally by gavage. (A) Kaplan-Meier event-free survival curves generated for mice treated with clofarabine were compared with that of vehicle-treated mice. n indicates the number of mice per group. Statistical significances between treatments were calculated using long-rank (Mantel-Cox) test. (B) Tumors were measured each day. The number of mice per group for each day throughout the study is given on top of the graph. The data are represented as the mean ± SD. The bar on the bottom of the figure shows the duration of the drug treatment. Asterisks indicate statistically significant differences between treatments (CTRL, control; CLF, clofarabine; ns, non-significant; *p<0.05; vs. control using non-parametric Mann-Whitney U test, two-tailed).

In order to observe the true effect of clofarabine on TC-71 xenografts we repeated the experiment with smaller cohort and euthanized the animals after 2 days of drug treatment (30 mg/kg administered orally). When the TC-71 xenografts were harvested immediately after 2 days of clofarabine treatment we observed complete loss of mitotic figures and significant increase in apoptotic cells in H&E stained slides, which was validated by very strong caspase-3 staining (Supplementary Figure S18). In this two-day clofarabine treatment study we saved half of the tumors for biochemical analysis. Earlier work suggested that CD99 activity may modulate phosphorylation of Src protein in different systems. We first validated this hypothesis by treating STA-ET-7.2 cells with clofarabine in culture. Inhibition of CD99 by clofarabine resulted in time-dependent reduction in phopsho-Src levels without altering expression of Src protein (Supplementary Figure S19A). We then compared phospho-Src levels in harvested TC-71 xenografts by western blotting. Four mice that were treated with clofarabine for two days showed reduced Src phosphorylation compared to 3 mice in the vehicle control group (Supplementary Figure S19B). There was no difference in total Src levels.

We evaluated the lung metastasis in animals with TC-71 xenografts. None of the animals showed any gross metastatic foci during necropsy. At least 4 H&E stained lung sections from each animal were examined by a board certified pathologist (B.T.H.) in a blinded format for the presence of metastatic ES cells. In the control group 3 out of 13 mice (23%) had lung metastasis. In the clofarabine treatment group 6 mice out of 14 (43%) had lung metastasis. All metastatic cells were in or around blood vessels (Supplementary Figure S20). We hypothesized that increased lung metastasis in the clofarabine group was due to extended survival of the animals in this cohort (7 days vs. 33 days). A true effect of clofarabine on metastasis remains to be tested in an animal study where the primary tumors are removed by amputation and then treated with clofarabine or placebo to achieve comparable observation time between two groups.

Discussion

CD99 is a promising target in ES, because the engagement of CD99 by monoclonal antibodies promotes significant cancer cell death both in cultured cells and in xenografts ^{29, 31, 42–44}. Effective antibody-based therapies targeting CD99 have not yet been developed. In this study, we identified and characterized first-in-class small molecules targeting CD99 function. We took an innovative and a stringent screening approach that combined a chemical library screening for testing direct binding ability of small molecules to the CD99 protein followed by a secondary cell-based cytotoxicity screening. This allowed us to identify two FDA-approved chemotherapy drugs clofarabine and cladribine as novel inhibitors of CD99. We further characterized and validated the drugs in a variety of biochemical/molecular, cellular and in vivo assays. Clofarabine and cladribine selectively inhibited the growth of cultured ES cells with 18–30 fold lower IC₅₀ values in comparison to non-ES cell lines. Furthermore, clofarabine significantly suppressed the growth of tumor in three different xenograft models of ES.

From a mechanistic point of view, the anti-leukemic activity of clofarabine and cladribine is thought to be due to the multiple mechanisms including abrogation of DNA synthesis, inhibition of DNA repair and induction of DNA strand-breaks and apoptosis. Following entry into the cells, they have to be metabolized by host cell kinases to undergo stepwise addition of phosphate groups to yield the mono-, di- and triphosphate forms, with the triphosphate being the active lymphotoxic metabolite ^{45, 46}. The relatively higher expression of deoxycytidine kinase (a phosphorylating enzyme) and lower expression of 5’-nucleotidase (a dephosphorylating enzyme) in lymphocytes compared with other cell types has been suggested as the key determinant for their higher responsiveness to the cytotoxic action of cladribine ⁴⁷. It has been recently demonstrated that CD99 is highly expressed in acute myeloid leukemia (AML) leukemic stem cells and myelodysplastic syndromes hematopoietic stem cells compared to their normal hematopoietic counterparts, hematopoietic stem and progenitor cells ⁴⁸. Furthermore, anti-CD99 monoclonal antibodies exhibited antileukemic activity in AML xenografts ⁴⁸. Strikingly, our results suggest that higher expression of CD99 in leukemia cells ^{21, 49–55} might also be associated with the particular vulnerability of those cells to clofarabine and cladribine.

CD99 engagement by specific antibodies block lymphocyte diapedesis and reduce the infiltration of inflammatory leukocytes into the central nervous system in relapsing-remitting experimental autoimmune encephalomyelitis, an animal model of multiple sclerosis ⁵⁶. These findings indicate that CD99 is a valid promising target for the treatment of autoimmune diseases as well. It is interesting that cladribine had significant efficacy against multiple sclerosis in a large randomized study ⁵⁷, supporting our findings that blocking of CD99 function could be an important mechanism for the observed therapeutic effects of cladribine in patients with multiple sclerosis.

From a clinical perspective, targeting CD99 represents a promising strategy, although with some concerns on tumor selectivity regarding that CD99 is broadly expressed on many cell types, albeit at low levels. While clofarabine is indicated for the treatment of pediatric patients with relapsed or refractory acute lymphoblastic leukemia, consecutive to at least two prior regimens ⁴⁵, cladribine is used for the treatment of hairy cell leukemia, chronic lymphocytic leukemia (progressive and resistant to first-line chemotherapy drug forms), non-Hodgkin’s lymphoma of low and intermediate grade as well as for the therapy of relapsing multiple sclerosis ^{46, 57}. The primary toxicity of both clofarabine and cladribine is myelosuppression leading to increased risk of infectious complications ^{45, 46}. In our study, we did not observe any significant body weight loss in none of the drug-treated xenograft groups (Supplementary Figure S12).

Mouse models of ES provide valuable proof of principle data but may fall short in guiding physicians to design new clinical trials. There are both biological and technical differences in growing ES cell lines in immunocompromised mice and detecting the disease dissemination at the microscopic level. It is possible to improve the primary tumor response by testing higher doses or longer treatment schedules including multiple cycles of one week on one week off clofarabine regimen. Since mice had to be euthanized when the primary tumor reaches the limit size, animals do not have enough time to develop established metastasis. Therefore, potential effect of clofarabine cannot be tested in this model. A more appropriate experiment would have been amputating the tumor bearing leg to allow animals to survive long enough to develop lung metastasis. Mice then can be randomized to control and treatment groups following the amputation.

Because the antimetabolic effects of clofarabine and cladribine depend on their intracellular phosphorylation ^{45, 46}, we hypothesized that membrane-impermeable, carboxylic acid or phosphate ester derivatives of these drugs could be more specific in targeting CD99 with much lower toxicity on normal proliferating cells (Supplementary Figure S21). Accordingly, our preliminary findings are encouraging in that clofarabine 5’-triphosphate binds to CD99 with a K_D value of 3.7 μmol/L, and inhibits growth of ES cells with a very similar IC₅₀ value to that of clofarabine (Figure 3C–D).

In conclusion, we have identified novel small molecule inhibitors of CD99 from screening of small molecule libraries that bind directly to CD99 on ES cells. These findings establish that anti-CD99 activity on the cell surface is a novel molecular mechanism underlying the action of clofarabine.

Materials and Methods

Study Approval

All animal studies were conducted under a protocol approved (approval number 2016-1174) by the Georgetown University’s Institutional Animal Care and Use Committee (IACUC) in accordance with NIH guidelines for the ethical treatment of animals.

Chemical Libraries and Drugs

The chemical library collection consisting of Diversity Set, Mechanistic Set and Natural Products Set used in this work was kindly supplied by NCI/DTP Open Chemical Repository. Clofarabine was purchased from Sigma Aldrich (St. Louis, MO; #C7495) or Selleck Chemicals (Houston, TX; #S1218). Cladribine was obtained from Sigma Aldrich (#C4438 or # 1134200). Clofarabine-5’-triphosphate was purchased from Jena Bioscience (Jena, Germany; #NU-874).

Cell Lines and Culturing

ES cell lines RDES, TC-71, STA-ET-7.2 and MHH-ES were grown in RPMI supplemented with 10% FBS. TC32 and A4573 cells were maintained in RPMI with 10% FBS and 10 mmol/L HEPES. SKES cells were grown in McCoy’s medium with 15% FBS. COG-E-352, CHLA-9 and CHLA-10 cells were grown in IMDM with 15% FBS and 1% (v/v) ITS (Sigma). IOR/CAR, SK-N-MC, A673 and 6647 cells were grown in IMDM, supplemented with 100 U/mL penicillin, 100 μg/mL streptomycin and 10% FBS. The leukemia cell lines MOLT-4 and Jurkat were maintained in RPMI with 10% FBS. The glioblastoma cell line A172 was grown in DMEM with 10% FBS. U-87 MG cells were grown in DMEM with 10% FBS and 1% (v/v) nonessential amino acids (Gibco). The human OS cell lines HOS-MNNG, SAOS-2/LM7, SAOS-2, U-2 OS and MG63.3 were grown in DMEM with 10% FBS. IOR/MOS, MG-63, IOR/OS20, IOR/OS14, IOR/OS9, IOR/OS10 cells and the rhabdomyosarcoma cell lines RD/18, RH4, RH30 and RH1 were routinely cultured in IMDM supplemented with 10% FBS, 100 U/mL penicillin and 100 μg/mL streptomycin. The breast cancer cell line MCF-7 and human lung adenocarcinoma cell line A549 were maintained in RPMI with 10% FBS. The prostate cancer cell line PC3 was grown in RPMI with 10% FBS with 10 mmol/L HEPES. The human embryonic kidney cell line HEK293 was grown in DMEM with 10% FBS. All cell lines were maintained in a fully humidified atmosphere of 5% CO₂ at 37°C, and tested mycoplasma-negative using MycoAlert kit (Lonza).

Cell Viability Assays

Cell viability was assayed by using the MTT (Trevigen) or WST-1 (Roche Diagnostics) cell proliferation assays according to the manufacturer’s instructions.

Cloning and Preparation of Recombinant CD99-ECD Protein

The Champion^™ pET104 BioEase^™ Gateway^® Expression System (Invitrogen) was used to produce CD99-ECD fused with a N-terminal biotin tag, as directed by the manufacturer’s instructions. Briefly, the DNA sequence encoding the extracellular domain (Asp23-Asp122) of human CD99 was amplified by PCR using pcDNA3.1 expression vector (Life Technologies) carrying the full-length human CD99 cDNA transcript variant 1 as template. The primers were designed to include a C-terminal 6xHis tag in the protein. The PCR product was gel purified with a QIAquick gel extraction kit (Qiagen) and subcloned into the pDONR 221entry vector. The entry clone was then subjected to a recombination reaction with pET104.1-DEST destination vector, an expression vector with a N-terminal BioEase tag.

CD99-ECD was expressed in Escherichia coli BL21 (DE3) cells. A saturated overnight culture of cells containing 100 μg/mL ampicillin was diluted by 20-fold into fresh LB medium. To induce the expression of CD99-ECD, cells were grown to OD_600nm 0.5–0.7 at 37°C and isopropyl-β-D-thiogalactopyranoside was added to a final concentration of 1 mmol/L. The cultures were grown for an additional 3 h at 37°C. Cells were harvested by centrifugation at 8,000 × g for 15 min, resuspended in 50 mmol/L sodium phosphate buffer, pH 7.4, containing 500 mmol/L NaCl, 10 mmol/L imidazole and 1 Complete EDTA-free protease inhibitor tablet/50 mL (Roche Diagnostics), and lysed by sonication. The cleared lysates were then subjected to an affinity chromatography on a nickel-charged Hi-Trap chelating high-performance column (GE Healthcare Bio-Sciences) in an AKTA Explorer chromatography system (GE Healthcare Bio-Sciences). Protein fractions were eluted with a linear gradient of imidazole (10 to 1000 mmol/L). The purity of the protein was assessed by SDS-PAGE followed by Coomassie staining. The eluted fractions were stored at −80°C for further use.

Gene Silencing with shRNA

Stable silencing of CD99 was obtained in TC-71 as previously described ²⁵. Briefly, an shRNA plasmid (pSilencer 2.1-U6 Neo vector; Ambion) expressing CD99 siRNA-1 (5’-GATCCGGCTGGCCATTATTAAGTCTTCAAGAGAGACTTAATAATGGCCAGCCTTTTTGGAAA-3’) was created and ES cells were transfected using the calcium phosphate transfection method. TC-CD99-shRNA clones (TC-CD99-shRNA#1 and TC-CD99-shRNA#2) were established after selection in neomycin (500 μg/mL) (Sigma-Aldrich).

Cell Cycle and Cell Death Analysis

ES and OS cell lines were treated with cladribine or clofarabine at different concentrations for 48 h. For evaluation of the cell cycle, cell cultures were incubated with 10 μmol/L bromodeoxyuridine (BrdU) (Sigma-Aldrich) for 1 h in a CO₂ atmosphere at 37°C. Harvested cells were fixed in ice-cold 70% ethanol for 30 min. After DNA denaturation with 2 N HCl for 30 min at room temperature, cells were washed with 0.1 mol/L Na₂B₄O₇, pH 8.5 and processed for indirect immunofluorescence staining, using α-BrdU (BD Biosciences) diluted 1:4 as a primary MAb and α-mouse FITC (1:100 – Thermo Scientific) as a secondary antibody. After treatment with 0.5 mg/mL RNase and staining with 20 μg/mL propidium iodide, cells were and analyzed by flow cytometry (FACSCalibur; Becton Dickinson) for cell cycle evaluation and for assessing cell death by DNA content analysis.

Chemotaxis/Cell Migration Assay

Motility assay was done using Trans-well chambers (Costar) according to manufacturer’s instructions. 6647 and TC-71 cells were pre-treated for 24 h with the drugs (cladribine 1 μmol/L or 3 μmol/L and clofarabine 0.3 μmol/L or 0.5 μmol/L, respectively), counted, and 100,000 viable cells were seeded in the upper chamber for migration analysis to rule out possible effects of drug treatment on cell vitality that might affect cell migration.

Soft Agar Colony Formation Assay

Anchorage-independent growth was determined in 0.33% agarose (Sea-Plaque Agarose, Lonza) with a 0.5% agarose underlay. Cell suspensions (3,300 cells/60-mm dish) were plated in semisolid medium with or without clofarabine, clofarabine-5’-triphosphate or cladribine and incubated at 37°C in a humidified 5% CO₂ atmosphere. Colonies were counted after 7–12 days.

SPR Experiments

The initial binding screening of the NCI/DTP chemical libraries and kinetic studies were performed on a Biacore T200 instrument (GE Healthcare, Piscataway, NJ, USA). A negative control protein, Ly6k was immobilized on another flow cell of the same chip, and the first flow cell was left empty for reference subtraction. Analytes were then injected individually over all flow cells as the binding interactions were recorded.

For kinetic studies, purified recombinant human full-length CD99 produced in HEK293 cells was purchased from Origene (Rockville, MD) (#TP304056). K_D (equilibrium dissociation constant) values were obtained using BiaEvaluation software (version 1.0).

Immunoprecipitation and Immunoblotting

Immunoprecipitation and immunoblotting experiments were performed as previously described ⁵⁸ using the following primary antibodies: anti-CD99 clone 12E7 (Dako, Inc.; #M3601); anti-CD99 clone 013 (Invitrogen, #180235 or MA5-12287); anti-CD99 antibody [EPR3097Y] (Abcam, #ab75858); anti-cyclophilin A (Abcam, #ab58144); anti-PKA-RIIα (Santa Cruz Biotechnology, #sc-908); anti-Src (Cell Signaling, #2123); anti-phospho-Src family (Tyr416) (Cell Signaling, #6943); streptavidin-HRP (Cell Signaling, #3999); anti-actin-HRP, (Santa Cruz Biotechnology, #sc-1615); anti-ROCK2 clone C20 (Santa Cruz Biotechnology, #sc-1851); anti-GAPDH (Santa Cruz Biotechnology, #sc-25778); and apoptosis western blot cocktail (Abcam, # ab136812). Quantitative analysis of protein bands was performed using ImageJ 1.48v software ⁵⁹.

Chemical Cross-linking with BS3

A monolayer of STA-ET-7.2 and RDES cells were grown to 80– 90% confluency in 10-cm tissue culture dishes, and incubated with clofarabine and cladribine at 5 μmol/L concentration or vehicle control (DMSO) for 1 h. The cells were rinsed 3 times with ice-cold PBS, and treated with 1 mmol/L BS³ (Thermo Scientific) in PBS for 10 min at 37°C according to manufacturer’s instructions. The reaction was quenched with the addition of 20 mmol/L Tris-HCl, pH 7.5 for 15 min at RT.

Orthotopic Mouse Xenograft Studies

The 5-weeks-old female SCID/bg mice (Taconic Farm, Inc., Germantown, NY) were injected intratibially with TC-71 (one million/50 μL PBS), SKES (two million/100 μL PBS) and A4573 (one million/100 μL PBS) cells. Sample sizes for all experiments were estimated using StatMate 2.0a software (GraphPad) assuming a mean tumor volume of 1.0 and 0.75 cm³ for control and treatment groups as the endpoints, respectively, with control group standard deviation of 150 mm³ at 0.95% power and a significance level (alpha) of 0.05 (two-tailed). After primary tumors reached 150–200 mm³ in size, mice were randomly allocated to vehicle control (DMSO), clofarabine and cladribine treatment groups using the random number generator function in Microsoft Excel. Clofarabine and cladribine were solubilized in DMSO as 120 or 80 mg/mL stock solutions for i.p. administration, and the dosing solutions were prepared by 10x dilutions in sterile PBS. For oral administration, clofarabine was dissolved in PEG 400, then diluted to a final concentration of 25% PEG 400 in sterile 0.9% sodium chloride solution. Mice carrying TC-71 xenografts were treated by i.p. injection with clofarabine (30 mg/kg), cladribine (20 mg/kg) or vehicle (DMSO, 10% (v/v)) in a volume of 50 μL, or with clofarabine (30 mg/kg) administered orally by gavage in a volume of 100 μL, once daily, for 14 days. SKES and A4573 xenograft-bearing mice were treated by i.p. injection with clofarabine (30 mg/kg), cladribine (20 mg/kg) or vehicle (DMSO, 10% (v/v)), once daily for the indicated days. The tumor volumes were determined by the formula (π/6) × length² × width and measured every day using a slide caliper. Animals found dead overnight with tumors smaller than 1.0 cm³ in size or mice euthanized for tumor ulceration were censored from the survival analyses.

Histology and Immunohistochemistry

All tumor tissues were fixed in 10% neutral buffered formalin for 24 h, dehydrated through a graded series of alcohols and cleared in xylenes prior to embedding in paraffin. Embedded tissues were cut into 5 μm thick sections and stained with hematoxylin and eosin (H&E).

Immunohistochemical staining was performed for CD99 and caspase 3 (cleaved). Five micron sections from formalin fixed paraffin embedded tissues were de-paraffinized with xylenes and rehydrated through a graded alcohol series. Heat induced epitope retrieval was performed by immersing the tissue sections at 98°C for 60 min in 10 mmol/L citrate buffer (pH 6.0). Immunohistochemical staining was performed using a horseradish peroxidase labeled polymer from Agilent (#K4001, #K4003) according to manufacturer’s instructions. Briefly, slides were treated with 3% hydrogen peroxide and 10% normal goat serum for 10 min each, and exposed to primary antibodies for CD99 (Abcam, #ab8855) at a 1:100 dilution and for caspase 3 (cleaved) (Biocare, #CP229A) at a 1:90 dilution in Da Vinci Green for one hour and stained on an DAKO Autostainer. Slides were exposed to the appropriate HRP labeled polymer for 30 min and DAB chromagen (Dako) for 5 min. Slides were counterstained with Hematoxylin (Fisher, Harris Modified Hematoxylin), blued in 1% ammonium hydroxide, dehydrated, and mounted with Acrymount. Consecutive sections with the primary antibody omitted were used as negative controls. The sections were evaluated in a blinded fashion by a board certified pathologist (B. T. H.).

Statistics

All statistical analyses were performed using Prism version 6.0c (GraphPad Software, La Jolla, CA). Statistical significance was defined as P<0.05. Statistical analysis of differences in tumor volumes between the control and drug-treated animal groups was performed by Mann-Whitney U test. A long-rank (Mantel-Cox) test was used for determining the statistical significances in event-free survival differences of animals between treatments. Statistical analysis for correlation was performed using Spearman’s correlation analysis with two-tailed P values and 95% confidence interval. A Student’s unpaired t test, two-tailed, was used for statistical analysis of in vitro data. As part of the standard unpaired t test analysis, GraphPad Prism tests the assumption that the variance between the groups is identical using a F test.