Molecular Inhibitor of QSOX1 Suppresses Tumor Growth In Vivo

Cancer is a leading cause of death worldwide and distant metastases are the major cause of patient mortality. Initially the primary tumor grows in its microenvironment which consists of tumor cells and nonmalignant stroma, each secreting extracellular matrix (ECM; ref. 1). The constituents of tumor ECM are a critical factor for cancer invasion and metastasis (1) and many changes occur in the tumor microenvironment (TME) prior to physical migration of metastatic cells away from the primary tumor. These changes include upregulation of matrix metalloproteinases (MMP; ref. 2), aberrant integrin signaling (3), and a loss of adherens junctions (3, 4). Each step of the metastatic process is mediated by various ECM constituents (1). In the largest cohort (N = 126), to our knowledge, of matched patient primary and renal cell carcinoma (RCC) metastases, we identified the upregulation of ECM-related genes in metastases relative to primary RCC tumors. Because these ECM genes are upregulated in metastases, they may play an important role in the metastatic cascade across multiple solid tumors (5).

Because every protein in the ECM contains disulfide bonds, we hypothesized that QSOX1, an enzyme that generates disulfide bonds in substrate proteins, is important for tumor cell growth, adherence, and invasion. Overexpression of QSOX1 in cancer was discovered after a C-terminal peptide was detected by mass spectrometry of plasma from patients with pancreatic cancer (6). Subsequently, QSOX1 overexpression has been reported in many other cancers (7–12). Inhibition of QSOX1 activity with a mAb revealed that QSOX1 is active extracellularly in stromal fibroblasts and is required for proper incorporation of laminin and fibronectin into the ECM (13, 14). Further, a small molecule inhibitor of QSOX1, ebselen, reduced proliferation, and invasion of pancreatic and renal cancer cell lines in vitro, and reduced tumor growth in vivo (15).

Herein we demonstrate a novel small molecule derived from a high throughput screen of ∼50,000 compounds, 3-methoxy-N-[4-(1-pyrrolidinyl)phenyl]benzamide (“SBI-183”), inhibits the enzymatic activity of QSOX1, thereby suppressing the proliferative and invasive phenotype of 2 renal cancer cell lines (786-O and RCJ-41T2), a triple negative breast cancer (TNBC) cell line (MDA-MB-231), a lung adenocarcinoma cell line (A549), and a pancreatic ductal adenocarcinoma (MIA PaCa2). Furthermore, we did not observe any compound-related toxicity of normal adherent fibroblasts or nonadherent peripheral blood mononuclear cells (PBMC) supporting a role for QSOX1 in tumor-derived ECM.

SBI-183 (molecular weight 296.3723 g/mol) was purchased from ChemBridge Corp. Compounds were dissolved in tissue culture-grade DMSO (Sigma-Aldrich) and kept at −80°C as 100 mmol/L stock solutions. See Supplementary Fig. S1 for the chemical structure of SBI-183.

RCC line 786-O was purchased from the ATCC and maintained in RPMI1640 (Corning) containing 10% FBS (Atlanta Biologicals), 1% Penicillin-Streptomycin (Pen-Strep; Corning), and 1% Glutamax (Gibco). A recently derived sarcomatoid RCC line from Mayo Clinic, RCJ-41T2 (16), was maintained in DMEM in 10% FBS, 1% Pen-Strep, and 1% Glutamax. The TNBC adenocarcinoma cell line MDA-MB-231 (ATCC), lung adenocarcinoma cell line A549 (ATCC), and the pancreatic ductal adenocarcinoma cell line MIA PaCa2 (ATCC) were also maintained 10% DMEM. MDA-MB-231-Luc (Cell Biolabs) was maintained in 10% RPMI1640 without Pen-Strep. De-identified fibroblasts derived from a 28-year-old Caucasian male with no overt disease were a kind gift from Dr. Clifford Folmes. PBMCs were obtained under an IRB-approved protocol (#06010000548) from Arizona State University. The identity of all cell lines was confirmed by STR analysis. Each cell line also tested negative for mycoplasma and mouse pathogens throughout the study, and were maintained at 37°C in 5% CO₂. All cell lines were used immediately upon thawing throughout the study.

Short hairpin (sh) lentiviral particles were purchased from GeneCopoeia containing either sh742 RNA as described (7) (catalog no. LPP-CS-HSH273J-LVRU6GP-100) or a shScramble (shScr) control (catalog no. LPP- CSHCTR001-LVRU6GP-025). 786-O cells were seeded at 2.5 × 10⁴ cells/well in a 6-well plate in complete RPMI1640. Adherent cells were transduced in triplicate with lentiviral particles following the manufacturer's instructions. After 72 hours, cells were selected in puromycin and subcloned by limiting dilution. A monoclonal population denoted as 786-O sh742.E11 was expanded. KD of QSOX1 was determined to be 90% by qRT-PCR as compared with the 786-O shScr cells (Supplementary Fig. S2).

PcDNA3.1 containing the short form of human QSOX1 (rQSOX1) was used to transfect Freestyle 293F cells (Thermo Fisher Scientific). rQSOX1 was expressed by 293F cells, harvested from supernatants, and purified on a nickel column via the C-terminal histidine tag. Enzymatic activity of QSOX1 and inhibitory activity of SBI-183 was confirmed using a fluorogenic assay as previously reported (17). Briefly, a mixture of 150 μmol/L dithiothreitol (DTT) substrate (Sigma-Aldrich) was added to 150 nmol/L rQSOX1, 1.4 μmol/L horse radish peroxidase (HRP; Thermo Fisher Scientific), and 1 mmol/L homovanillic acid (HVA; Sigma-Aldrich) in PBS at ambient temperature, pH 7.5. Assays were performed in a black plate in a total volume of 150 μL in triplicate. Fluorescence was measured at 20 second intervals over 15 minutes after the addition of DTT at λ_ex 320 nm/λ_em 420 nm using a FlexStation spectrophotometer (Molecular Devices). SBI-183 was preincubated with rQSOX1 for at least 10 minutes at concentrations ranging from 6.25 to 50 μmol/L.

rQSOX1 was labeled with DyLight 650 Amine-Reactive Dye (Thermo Fisher Scientific). Briefly, Dylight-650 was dissolved at 10 mmol/L in dimethylformamide and added at 2:1 molar ratios to 86 μmol/L QSOX1 in 50 mmol/L NaPO₄³⁻, 150 mmol/L NaCl, pH 8.0. The mixture was incubated in the dark for 1e hour at room temperature on a rocker, and dialyzed to 50 mmol/L Tris, 150 mmol/L NaCl pH 8.0 overnight at 4^oC. The labeling ratio was estimated using ϵ = 250,000 M⁻¹cm⁻¹ at 655 nm for DyLight 650 and ϵ = 93110 M⁻¹cm⁻¹ at 280 nm for QSOX1, and found to be 1.1.

Microscale thermophoresis (MST) experiments were performed in a Monolith NT.115 (Nanotemper). Sixteen serial dilutions of SBI-183 (from 250 to 0.0076 mmol/L) with 50 nmol/L Dylight 650-labeled QSOX1 in 1x PBS, pH 7.4, 5% DMSO, and 0.05% Tween 20 were loaded into standard MST capillaries and scanned at MST power of 20% at 23°C. To obtain K_d, MST data were fitted using MO Affinity Analysis software (Nanotemper).

Docking for SBI-183 was performed using Glide (v.5.6) within the Schrödinger software suite (Schrödinger, LLC; ref. 18). Our modeling techniques have been described (19–25). Briefly, we started with conformation searches of the ligand via the method of Polak-Ribière conjugate gradient (PRCG) energy minimization with the optimized potentials for liquid simulations (OPLS) 2005 force field (26) for 5,000 steps (or until the energy difference between subsequent structure was less than 0.001 kJ/mol-Å; ref. 18). Our docking methodology has been described (19, 25, 27), and the scoring function utilized described elsewhere (28). Briefly, molecular refracting molecules were removed from the human QSOX1 crystal structure (PDB Codes: 3Q6O; ref. 29). Schrödinger's SiteFinder module focused the grid on the active site region for QSOX1 (Fig. 1C). Using this grid, initial placement for SBI-183 was docked using the Glide algorithm within the Schrödinger suite as a virtual screening workflow (VSW). The docking proceeded from lower precision through SP docking and Glide extra precision (XP; Glide, v.5.6; Schrödinger, LLC; refs. 20, 30). The top poses were ranked for best score and unfavorable scoring poses were discarded. Multiple orientations were allowed in the site. Site hydroxyls were allowed to move with rotational freedom. Full docking scores are given in Supplementary Data File S1. This method provides the ideal conformation of ligand binding as utilized within Schrödinger suite, and the top docked pose represents the conformation of the ligand required to inhibit QSOX1. Hydrophobic patches were utilized within the VSW as an enhancement. XP descriptors were used to obtain atomic energy terms that result during the docking run (20, 30). Molecular modeling for importing and refining the X-ray structure and generation of SBI-183, as well as rendering of figure images were completed with Maestro (Schrödinger, LLC).

Cells were plated at optimized densities (1,000 cells/well for 786-0 and RCJ-41T2; 750 cells/well for MDA-MB-231) in their respective media, and plated in Corning 3570 384-well white titerplates using a MultiFlo bulk dispenser (BioTek). The cells were allowed to adhere for 24 hours. Compound and assay controls diluted in 100% DMSO were added to the cells using an ATS Gen4 acoustic transfer system (EDC Biosystems). A 1:1,000 compound:cell volume ratio was enforced to avoid DMSO toxicity. After 72-hour compound incubation, 25 μL CellTiter Glo reagent (Promega; G7573) diluted 1:4 in MilliQ water was added to the plates using the Multiflo dispenser and luminescent signal was read per standard assay protocol using a Molecular Devices Paradigm multimode reader (TUNE cartridge, luminescent mode).

Prior to screening, assay optimization experiments were performed for each cell line in the assay conditions described above. Cell densities were titrated in control plates containing negative and positive controls (DMSO and 10 μmol/L staurosporine, 0.1% DMSO in assay wells) to identify optimal seeding densities within linear ranges of luminescent signal, minimizing CVs (<10%) and maximizing Z' factors (>0.5), per standard NIH assay guideline optimization criteria and methods. For primary screens, 20-point 2-fold serial dilutions of SBI-183 in 100% DMSO were prepared from a stock concentration of 40 mmol/L in acoustic-compatible Aurora microplates (Ref. ABA200100A); internal plate controls for live cells (100% viability) and dead cells (0% viability) were included in source plates. After 24-hour seeding time, compound was added acoustically as described above. Cellular viabilities for each test well were derived from raw luminescent signal by normalization to internal plate controls. Viability experiments were performed in triplicate, and normalized data points averaged per dose. Dose–response curves were calculated by logistic regression in TIBCO Spotfire (version 7.0.0).

786-O, RCJ-41T2, MDA-MB-231, A549, and MIA PaCa2 were seeded in triplicate at 2.5 × 10³ cells/well (786-O, RCJ-41T2, A549, and MIA PaCa2) or at 5.0 × 10³ cells/well (MDA-MB-231) in phenol-red free 10% RPMI1640 (786-O) or 10% DMEM (RCJ-41T2, MDA-MB-231, A549, and MIA PaCa2) in 96-well plates. Adhered cells were incubated with 2-fold dilutions of SBI-183 starting at 20 μmol/L, or vehicle (0.4% DMSO) for 5 days. Cell growth was determined at days 1, 3, and 5 in an MTT assay (Molecular Probes) following the manufacturer's directions.

A total of 1.0 × 10⁵ 786-O, RCJ-41T2, MDA-MB-231, and A549 or 5.0 × 10⁴ MIA PaCa2 cells were seeded in triplicate onto Matrigel-coated 24-well invasion 8-μm pore-size inserts (Corning) in serum-free media. Cells were allowed to adhere for 30 minutes prior to the addition of DMSO or SBI-183 giving a final concentration in the well of 0.2% DMSO vehicle or 20 μmol/L SBI-183. Inserts were incubated for 4-5 hours (786-O, RCJ-41T2, MDA-MB-231, and A549) or overnight (MIA PaCa₂) at 37°C. Noninvading cells were removed, membranes were fixed in ice cold 100% methanol, and mounted on slides with DAPI (Vector Laboratories). Three unique fields were captured using the 4× objective and then automatically counted on a Cytation 5 microscope (BioTek). Images were edited using ImageJ.

The following protocol was performed as described in Vinci and colleagues (2015) with slight modifications as stated (31). 786-O and RCJ-41T2 were seeded in triplicate at 1.25 × 10³ cells/well in 200 μL 10% RPMI1640 or 10% DMEM respectively in ultra low attachment (ULA) 96-well plates (Corning). MDA-MB-231, A549, and MIA PaCa2 were seeded at 2.5 × 10³ cells/well in 10% DMEM. Plates were centrifuged at 1,000 × g for 3 minutes then incubated and allowed to form spheroids for 3 days. Plates were chilled to 4°C for 20 minutes and all but 50 μL of media was removed. On ice, 50 μL of Matrigel Matrix (Corning) was added. Plates were centrifuged at 300 × g for 3 minutes at 4°C, then incubated for 1 hour at 37°C. Each well contained the following final concentrations of SBI-183 in complete media: 20, 10, 5, 2.5, or 0.4% DMSO vehicle. Cells were imaged on days 0, 2, 4, 6 (RCJ-41T2 and MDA-MB-231), and 8 (786-O). Invasion was quantified with ImageJ as total area of invaded cells.

Rescue invasions were performed as stated above with the following modifications. To the 50 μL of media remaining in the wells, 50 μL of media containing either PBS or rQSOX1, and SBI-183 or DMSO was added. Final concentrations in the well were 2.5 μmol/L SBI-183 and 5 μmol/L rQSOX1 or 0.025% DMSO (786-O), 5 μmol/L SBI-183, and 5 μmol/L rQSOX1 or 0.05% DMSO (RCJ-41T2), and 2.5 μmol/L SBI-183 and 2.5 μmol/L rQSOX1 or 0.025% DMSO (MDA-MB-231). Matrigel was added as above at a 1:1 dilution.

786-O: Fox1^nu/nu mice were inoculated with 1.0 × 10⁶ 786-O cells in the right hind flank. Seven days post-implant (study day 0), mice were dosed daily by oral gavage with 400 μg/mouse/day SBI-183 dissolved in 100% DMSO. Control mice received 100% DMSO. Tumor length and width measurements were obtained using Vernier calipers.

RCJ-41T2: A part of RCJ-41T2 tumor was minced with a sterilized blade to slurry and mixed with equal volume of Matrigel. One hundred microliters of the resulting mixture was injected subcutaneously into 8- to 10-week-old male NSG mice using 1 mL syringes equipped with a 16-gauge needle. When the tumor grew to approximately ∼1,500 mm³, mice were sacrificed, the tumors were harvested and reimplanted into 18, male NSG mice as described above. When the average size of the tumors was ∼100 mm³, mice were randomized into 2 groups: (i) Vehicle: 20% DMSO + 80% PEG-400, gavage daily, (ii) SBI-183, 100 mg/kg dissolved in the vehicle, oral gavage daily. Treatment was continued for 3 weeks. Then, mice were euthanized and tumors and organs were harvested for further analysis.

MDA-MB-231-Luc: Twenty-four female CB.17 SCID mice ages 8 weeks were obtained from Charles River. Mice were inoculated with 0.1 mL of 50% Matrigel/50% Media containing 5 × 10⁶ MDA-MB-231-Luc cells (Cell Biolabs) into the mammary fat pad. Seven days post-implant (study day 1), daily oral administration of 100 mg/kg (n = 12) SBI-183 or vehicle control (n = 12) began. Primary endpoint was assessment of treatment effects on spontaneous distal lung metastases determined by ex vivo bioluminescence imaging. SBI-183 was dissolved in DMA (10% total volume)/PEG400 (90% total volume). Stock solution was made fresh weekly and stored at −20°C.

All animal experiments were conducted in accordance with and approved by an Institutional Animal Care and Use Committee (IACUC).

RCJ-41T2 xenograft tumors from mice were mounted in paraffin on slides. Slides were deparaffinized, rehydrated, and then the antigen was retrieved with citrate buffer (pH 6.0, 125°C for 1 minute). Slides were incubated in 3% H₂O₂ for 10 minutes at room temperature. Rabbit anti-laminin-α4 antibody (Novus Biologicals) was added at 1:300 and incubated at 4°C overnight. HRP conjugated anti-rabbit secondary antibody (Vector Laboratories) was added and incubated for 30 minutes at room temperature. 3,3′-Diaminobenzidine (DAB) was used as the chromogen with hematoxylin counter staining.

Three unique images from each of 2 tumors per group were obtained using the 10× objective on an Olympus BX51 microscope. DAB intensity was measured using Fiji (32).

Unless otherwise noted, all statistical analyses were performed using GraphPad Prism version 7.04 for Windows, GraphPad Software, www.graphpad.com.

SBI-183 emerged as a QSOX1 inhibitor from cell-free high throughput screening assays described previously (15). As shown in Fig. 1A, SBI-183 inhibits rQSOX1 in a dose-dependent manner in a fluorescence assay developed by Raje and colleagues (17). H₂O₂ produced by QSOX1 activity activates HRP to dimerize HVA resulting in fluorescence at 420 nm. The first bar in Fig. 1A demonstrates that SBI-183 does not inhibit HRP, nor does it appear to scavenge H₂O₂, providing confidence that the target of SBI-183 is QSOX1.

Because SBI-183 appeared to inhibit the enzymatic activity of QSOX1, we wanted to determine if it bound to QSOX1. MST was performed in duplicate showing binding of SBI-183 to QSOX1 at a K_d of 20 μmol/L (Fig. 1B).

Because crystal structures of QSOX1 have been generated (33) and SBI-183 appeared to bind QSOX1, computer modeling was performed predicting the binding location of SBI-183. From a SiteFinder search, 2 sites (Site 1 and Site 2) were identified as possible binding locations on QSOX1, however Site 1 was optimal for SBI-183 binding (Fig. 1C). Supplementary Table S1 displays the results from our docking protocols.

At Site 1, SBI-183 fits deep into a wedge-like crevice inside QSOX1 that includes the following residues within 6Å of SBI-183: C237, Y238, L239, V251, L252, M253, F258, Y259, Y262, and L263. Interaction pairs are formed between SBI-183 and QSOX1 with frequent ring-ring pi-clouds, H-bonds, and charge–charge interactions participating in electrostatic interactions with the backbone carbonyls and hydroxyl residues, and transient pi-cloud interactions occurring with the phenyl-substituted tyrosine rings (Fig. 1D).

To determine if SBI-183 targets QSOX1 in tumor cells, we selected cell lines that were previously identified to express QSOX1. The ability of SBI-183 to inhibit viability of tumor cells in a dose-dependent fashion was determined using the CellTiter Glo assay (Supplementary Fig. S3). Cells were treated with 2-fold dilutions of SBI-183 between 40 μmol/L and 0.076 nmol/L (in triplicate), incubated for 72 hours and analyzed. As shown in Supplementary Fig. S3, inhibition of viability was observed for 786-O, RCJ-41T2, and MDA-MB-231 with IC₅₀s of 4.6, 3.9, and 2.4 μmol/L, respectively.

Because previous studies demonstrated reduced proliferation of tumor cells when QSOX1 was knocked down (KD) using shRNA (7, 12, 15), we hypothesized that a compound which inhibits QSOX1 would similarly decrease tumor growth in vitro. To test this, tumor cells were cultured for 5 days in the presence of SBI-183 or 0.4% DMSO vehicle control. An SBI-183 concentration-dependent reduction in cell growth was observed for each tumor cell line (Fig. 2A,–E). To determine if SBI-183 demonstrated selectivity for tumor cells, non-malignant adherent human fibroblasts and non-adherent PHA-stimulated PBMC from healthy human donors were incubated with SBI-183 for 5 days under the same conditions. No significant inhibition of cell growth was observed compared with vehicle control (Fig. 2G and H). Supplementary Table S2 lists percent growth for Fig. 2A–D.

We previously reported that silencing QSOX1 expression with shRNA and inhibiting QSOX1 with the small molecule ebselen reduced the invasiveness of cancer cell lines in vitro (7, 12, 15). Similarly, we hypothesized that another small molecule inhibitor of QSOX1 would also suppress invasion. It is well known that 3D culture systems more closely recapitulate in vivo tumor phenotypes than 2D cultures. As seen in vivo, compounds may have difficulty diffusing to the center of a spheroid, or may be inhibited by hypoxia, leading to decreased efficacy, increased cellular survival, and reduced compound sensitivity (34–36). Therefore, in order to more closely mimic how naturally occurring tumors would be affected by SBI-183, we utilized a 3D invasion model. 786-O, RCJ-41T2, MDA-MB-231, A549, and MIA PaCa2 were grown as spheroids. After the addition of Matrigel, spheroids were imaged on the indicated days (Fig. 3i, ii, iii, iv, and v). 786-O, RCJ-41T2, and A549 initially formed dense spheroids (Fig. 3C, F, and L), which expanded over the course of the experiment (Fig. 3A, D, and J). RCJ-41T2 formed wandering tendrils as it invaded the surrounding matrix (Fig. 3D). MDA-MB-231 and MIA PaCa2 initially formed loose, grape cluster-like spheroids (Fig. 3I and O). At the end of the experiment these clusters were greatly enlarged with projections from the main body, and single cells migrating from the spheroid (Fig. 3G and M). In each cell line tested, incubation with SBI-183 reduced invasion through Matrigel (Fig. 3i, ii, iii, iv, and vB, E, H, K, N), similar to the reduction observed in QSOX1 stable KD cell line 786-O sh742.E11 (Fig. 3viS).

To ensure the observed decrease in invasion was not simply due to a decrease in proliferation due to the length of the 3D experiment, modified Boyden chamber invasion assays were also performed (Fig. 3vii–xi). These invasion assays confirmed the 3D spheroid results.

QSOX1 is overexpressed by tumor cells, localizes to the ER and Golgi, and is secreted. Because SBI-183 inhibits the activity of QSOX1 resulting in a decrease in invasion, we added exogenous rQSOX1 to rescue the invasive phenotype (7, 15). Addition of a 2-fold molar excess of rQSOX1:SBI183 to 786-O cells partially rescued invasion (Fig. 4A). Addition of equimolar concentration of rQSOX1 partially rescued the invasive phenotype of both RCJ-41T2 cells (Fig. 4B) and MDA-MB-231 cells (Fig. 4C).

Because SBI-183 inhibits invasion in vitro, we tested the activity of SBI-183 in 2 independent RCC mouse xenografts. Tumor measurements were obtained at the intervals indicated in Fig. 5. One mouse from the test group was terminated according to IACUC protocol on day 21. At the end of the experiment (day 41), SBI-183-treated 786-O xenografts had average tumor volumes that were 86% smaller than vehicle-treated mice. These results indicate that SBI-183 inhibits the growth of a RCC tumor cell line in vivo.

Sarcomatoid RCC is associated with an aggressive, mesenchymal phenotype, and is intrinsically resistant to antiangiogenic therapy. To extend our findings to a sarcomatoid RCC line recently derived from a patient, 18 NSG mice were inoculated with minced RCJ-41T2 tumors obtained from patient-derived xenografts in 50% Matrigel and tumors were established for 10 days prior to dosage with 100 mg/kg SBI-183 or vehicle control. Data are from 9 control mice and 6 experimental mice (3 mice were lost in the experimental group due to an oral gavage problem, not due to the compound). Tumor volume was measured every 7 days with calipers and volume was calculated using the following formula: Tumor volume = 0.5 × a × b², where a and b are the longest and shortest diameters, respectively. Over the course of the experiment, treatment with SBI-183 resulted in an average 51% tumor volume reduction compared with control (Fig. 5). No differences were observed in the overall body weight. These data suggest that SBI-183 inhibits the growth of a highly aggressive sarcomatoid RCC in vivo.

Because QSOX1 has previously been shown to be involved in the deposition of laminin-α4 in the ECM (13), we hypothesized that laminin-α4 deposition would be reduced in xenograft tumors from mice treated with SBI-183. DAB staining intensity due to laminin-α4 deposition was shown to be significantly reduced in SBI-183–treated mice compared with vehicle control (Fig. 6).

Despite systemic therapy, distant metastases are the major cause of cancer mortality. QSOX1 secreted from tumor and stromal cells is involved in ECM formation including laminin and fibronectin deposition (13, 14), and posttranslational activation of MMPs (12). Taken together, QSOX1 plays an important role in ECM-mediated invasive processes. Because tumor-stroma-derived ECM is crucial for metastasis, targeting a potential master regulator of the ECM such as QSOX1 may affect multiple ECM proteins involved in invasion and metastasis. There are several lines of evidence supporting QSOX1 as a potential therapeutic target. First, QSOX1 is overexpressed in several malignancies (7–12, 37) and is an indicator of poor relapse free and overall survival in luminal B breast cancer (7, 8, 38). Second, enzymatic inhibition of QSOX1 using either small molecules or mAbs interferes with ECM deposition and reduces tumor invasion (13–15). Because shRNA silencing of QSOX1 previously demonstrated suppression of tumor growth and invasive phenotype (7, 15), we embarked on a screening strategy to identify chemical probes to examine the effects of QSOX1 inhibition in vitro and in vivo. We demonstrate that the small molecule, SBI-183, (i) inhibits QSOX1 enzymatic activity in vitro, (ii) binds to QSOX1, (iii) inhibits tumor cell growth and invasion in vitro, and (iv) reduces tumor size in 2 independent mouse models.

We used an enzymatic assay developed by Colin Thorpe's group (17) to screen for QSOX1 inhibitors in a library of ∼50,000 compounds. SBI-183 was identified as a lead compound for the inhibition of QSOX1 enzymatic activity (Fig. 1A). We previously reported that ebselen bound covalently to QSOX1 by LC/MS-MS analysis (15), but SBI-183 does not appear to bind covalently to QSOX1. Another measure of binding is MST which measures the motion of proteins along microscopic temperature gradients and is affected by changes in protein hydration, charge, and size originating from ligand binding. Incubation of serial dilutions of SBI-183 with QSOX1 demonstrated a temperature shift indicative of binding (Fig. 1B). This physical interaction between QSOX1 and SBI-183 supports computer models showing SBI-183 fitting into a crevice in QSOX1 at the C-terminal end of the second thioredoxin domain. The strong docking score via the SBI-183 benzyl-moiety and the tyrosine ring at Y262, along with the SBI-183 carbonyl oxygen electrostatic interactions at the nearby tyrosines (Y259, Y262) and various van der Waals interactions on the hydrophobic residues (V251, M253, L252, F258) with the alkane atoms of SBI-183 creates a solid anchored position for SBI-183 on QSOX1. Additionally, the area of interaction includes C237 within 6Å of SBI-183 binding (Fig. 1D). C237 is one of 2 cysteines that covalently bound ebselen in our previous study (15). It is thought that C237 is not involved in QSOX1 enzymatic activity (39), however, it is interesting that 2 compounds which inhibit QSOX1 interact with it in this location. These data suggest that this region may be important for QSOX1 activity.

In addition to metastatic processes, the ECM is involved in signaling. SBI-183 suppressed growth in each tumor cell line tested in a concentration-dependent manner (Fig. 2A–E), but no significant reduction in growth of fibroblasts or PHA-stimulated PBMC was observed (Fig. 2G–H). This finding agrees with previously published IHC results in a tumor tissue biopsy that show no QSOX1 protein expression in nonmalignant tissue or infiltrating lymphocytes (6, 40). Furthermore, Supplementary Table S3 demonstrates that the maximal tolerated dose of SBI-183 in healthy nude mice is over 200 mg/kg and was limited by solubility of the compound, not toxicity. We do note, however, that SBI-183 has a high IC₅₀ in the tested cell lines (Supplementary Fig. S3). As such, it is possible that there are other targets of SBI-183 in vivo that suppress tumor growth.

To further examine the cancer phenotype, we utilized a well-accepted model of 3D invasion of spheroids into Matrigel (31, 41–43). Incubation with SBI-183 significantly reduced invasion in all 5 cell lines (Fig. 3i-v). Our 3D results are consistent with reduced invasion observed in the trans-well invasion assay (Fig. 3vii-xi). All invasion results are consistent with decreased invasion observed in cells stably expressing shRNA specific for QSOX1 (7, 15), and could be at least partially due to a malformed ECM lacking laminin, fibronectin (13, 14), and reduced MMP-2 and MMP-9 activity (12). Furthermore, addition of exogenous rQSOX1 partially rescued SBI-183–induced invasion suppression observed in 786-O, RCJ-41T2, and MDA-MB-231 (Fig. 4; A549 and MIA PaCa₂ were not tested). Rescue experiments are difficult to perform with small molecules because the small molecule can enter the cell while the target protein remains extracellular. A review of polypharmacology discusses that most drugs interact with 5 or more targets (44). In line with this, some VEGF tyrosine kinase inhibitors such as sunitinib, are known to also interact with other kinases (45). Similarly, although SBI-183 is active against QSOX1 (Fig. 1A) it likely has various other targets in tumors, explaining why exogenous addition of rQSOX1 does not completely rescue the invasive phenotype. However, if QSOX1 is a master regulator of multiple disulfide-bonded proteins, even partial inhibition of QSOX1 may disrupt folding or proper association of proteins in the ECM.

To examine the in vivo effects of SBI-183 on tumor growth and metastasis, we inoculated mice with 786-O, RCJ-41T2, and MDA-MB-231 (Supplementary Fig. S4). Mice bearing 786-O or RCJ-41T2 tumors that were treated with SBI-183 exhibited a statistically significant reduction in tumor volume compared with controls (Fig. 5A ad B). Mice bearing the TNBC cell line MDA-MB-231-luc interestingly did not exhibit a reduction in primary tumor volume (Supplementary Table S4) but rather exhibited a suppression of metastasis as evidenced by a reduction in mean lung radiance of 76% when compared with controls (Supplementary Fig. S4). This reduction, while striking, did not reach statistical significance, likely due to the death of 2 control mice. Our MDA-MB-231 in vivo data differs from our in vitro data in that SBI-183 slows tumor growth in vitro, but did not slow primary tumor growth in mice. However, MDA-MB-231 cells were inhibited from invading in the 3D spheroid and modified Boyden chamber models which are in vitro surrogates for metastasis. It should be noted that the effect of SBI-183 observed depends on the cell line tested, suggesting that cells depend differently on QSOX1 activity.

Our data show that both in vitro and in vivo, SBI-183 suppresses QSOX1 enzymatic activity which results in inhibition of tumor growth, invasion, and possibly metastasis in vivo. Further, our data suggest that SBI-183 may be a useful tool to increase our understanding of the role of QSOX1 activity in the ECM of cancer and stromal cells during invasion and metastasis. Because metastasis is the main cause of death from cancer, even partial inhibition of this process may prolong patient survival. Finally, our study provides further evidence of QSOX1 as an anti-neoplastic target.

No potential conflicts of interest were disclosed.

Conception and design: A.L. Fifield, P.D. Hanavan, D.O. Faigel, J.L. Petit, T.H. Ho, D.F. Lake

Development of methodology: A.L. Fifield, P.D. Hanavan, E. Sergienko, A. Bobkov, T.R. Caulfield, J.A. Copland, D. Mukhopadhyay, T.H. Ho, D.F. Lake

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.L. Fifield, P.D. Hanavan, E. Sergienko, A. Bobkov, N. Meurice, J.L. Petit, A. Polito, E.P. Castle, D. Mukhopadhyay, K. Pal, S.K. Dutta, H. Luo, T.H. Ho, D.F. Lake

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.L. Fifield, P.D. Hanavan, D.O. Faigel, A. Bobkov, N. Meurice, J.L. Petit, A. Polito, T.R. Caulfield, D. Mukhopadhyay, K. Pal, T.H. Ho, D.F. Lake

Writing, review, and/or revision of the manuscript: A.L. Fifield, D.O. Faigel, A. Bobkov, N. Meurice, J.L. Petit, T.R. Caulfield, E.P. Castle, J.A. Copland, D. Mukhopadhyay, K. Pal, T.H. Ho, D.F. Lake

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.L. Fifield, D.O. Faigel, J.L. Petit, T.H. Ho, D.F. Lake

Study supervision: T.H. Ho, D.F. Lake

Other (Assay development and HTS that identified the studied molecule): E. Sergienko

This work was supported in part by the Gloria A. and Thomas J. Dutson Jr. Kidney Research Endowment. T.H. Ho is supported by NCI (R01 CA224917) and the Department of Defense (W81XWH-17-1-0546). Opinions, interpretations, conclusions, and recommendations are those of the author and are not necessarily endorsed by the Department of Defense. The funding agencies had no role in the study design. This work was also partially supported by grants to D.O. Faigel, E. Sergienko, and D.F. Lake (R01 CA201226) and to D. Mukhopadhyay (R01 CA78383-20).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.