Abstract
Mutant cystathionine gamma-lyase was targeted to phosphatidylserine exposed on tumor vasculature through fusion with Annexin A1 or Annexin A5. Cystathionine gamma-lyase E58N, R118L, and E338N mutations impart nonnative methionine gamma-lyase activity, resulting in tumor-localized generation of highly toxic methylselenol upon systemic administration of nontoxic selenomethionine. The described therapeutic system circumvents systemic toxicity issues using a novel drug delivery/generation approach and avoids the administration of nonnative proteins and/or DNA required with other enzyme prodrug systems. The enzyme fusion exhibits strong and stable in vitro binding with dissociation constants in the nanomolar range for both human and mouse breast cancer cells and in a cell model of tumor vascular endothelium. Daily administration of the therapy suppressed growth of highly aggressive triple-negative murine 4T1 mammary tumors in immunocompetent BALB/cJ mice and MDA-MB-231 tumors in SCID mice. Treatment did not result in the occurrence of negative side effects or the elicitation of neutralizing antibodies. On the basis of the vasculature-targeted nature of the therapy, combinations with rapamycin and cyclophosphamide were evaluated. Rapamycin, an mTOR inhibitor, reduces the prosurvival signaling of cells in a hypoxic environment potentially exacerbated by a vasculature-targeted therapy. IHC revealed, unsurprisingly, a significant hypoxic response (increase in hypoxia-inducible factor 1 α subunit, HIF1A) in the enzyme prodrug–treated tumors and a dramatic reduction of HIF1A upon rapamycin treatment. Cyclophosphamide, an immunomodulator at low doses, was combined with the enzyme prodrug therapy and rapamycin; this combination synergistically reduced tumor volumes, inhibited metastatic progression, and enhanced survival. Mol Cancer Ther; 16(9); 1855–65. ©2017 AACR.
This article is featured in Highlights of This Issue, p. 1727
Introduction
As research continues to unravel the molecular circuitry behind cancers, targeted therapies will likely enhance the duration and quality of life of cancer patients through creative approaches, including toxic payload delivery, signaling attenuation, immunostimulation, and combination strategies (1–3). One targeted adaptation of a toxic payload delivery involves the local activation of chemotherapeutic prodrugs within the tumor using enzyme prodrug therapies (4, 5). Enzyme prodrug strategies localize an enzyme to the tumor microenvironment either through gene delivery (6, 7), antibody targeting (8, 9), local administrations, or cellular delivery systems (10). Once the enzyme is localized to the tumor environment, a nontoxic prodrug is administered that is then enzymatically converted to a toxic drug in the tumor, minimizing systemic exposure to the cytotoxic agent.
Enzyme prodrug therapies attempt to achieve a high therapeutic index and selectivity through the generation of toxic levels of drug from an inert prodrug at the tumor. The methionine gamma-lyase (MGL) enzyme prodrug system converts the prodrug selenomethionine to methylselenol, enhancing cytotoxicity of the compound by several orders of magnitude (11, 12). The methylselenol created in the tumor leads to a rapid antitumor effect due to oxidative stress as a result of the generation of toxic reactive oxygen species, which activate the caspase cascade and apoptosis (11, 13). Alone, selenomethionine exhibits minimal toxicity because mammalian cells do not express the enzymes necessary to convert it to a toxic selenol compound (14, 15), allowing for controlled activation of the prodrug through the targeting of the MGL system.
A primary criterion for enzyme prodrug system success, site-specific prodrug activation, requires the use of foreign enzymes to avoid off-target cytotoxic activity, and in the case of MGL, the Pseudomonas putida enzyme is often utilized. Administration of large foreign proteins presents complications even for immunocompetent preclinical models. Cystathionine gamma-lyase (CTH) acts in a mammalian cystathionine metabolism pathway in a manner similar to bacterial methionine salvage with MGL. Wild-type mammalian CTH displays no activity toward methionine or selenomethionine; however, three amino acid substitutions impart MGL activity to CTH (16), providing a viable alternative to the administration of a foreign protein in the MGL enzyme prodrug system.
Current enzyme prodrug systems face limitations regarding enzyme localization, including gene-directed enzyme prodrug therapy and its derivative virus-directed enzyme prodrug therapy. Most significantly, gene delivery strategies suffer from poor gene expression in vivo; however, selective delivery, insertional mutagenesis, and immunogenicity are persistent concerns as well (17–19). Delivery of active enzyme, as with antibody-directed enzyme prodrug therapies, circumvents expression issues; however, antibody–enzyme conjugate accessibility to the tumor-specific targets can limit this approach (17, 20).
A potential target for delivery of therapies is phosphatidylserine, which is expressed externally on cancer cells and the tumor vasculature but remains on the cytoplasmic side of the membrane of healthy cells and normal vasculature (21–25). Targeting of an enzyme to phosphatidylserine can be achieved through the fusion of the enzyme to a member of the Annexin protein family, which binds to anionic phospholipids, such as phosphatidylserine, with high affinity (26–30). Phosphatidylserine acts as a unique target for the delivery of an enzyme prodrug system because adhesion to the endothelial wall of tumor vessels allows for tumor killing through the destruction of vasculature and permeation of the small-molecule cytotoxic drug into the tumor, creating a bystander killing effect (31, 32). The Annexin-assisted delivery of mutant mammalian CTH (mCTH) to the tumor vasculature and the subsequent occlusion of vessels following conversion of selenomethionine to methylselenol present a unique opportunity for anticancer combination therapies to address the tumor response to the resultant intensification of the hypoxic tumor environment.
Hypoxia-inducible factors (HIF) activate mechanisms to improve oxygenation as well as reprogram metabolic processes to enhance cell survival under oxygen deprivation, such as in the case of a vasculature-directed cytotoxic delivery system. HIF1 and HIF2 regulate over 1,000 genes involved in cancer biology, including the proangiogenic VEGF and enzymes responsible for matrix remodeling and cellular migration (33–35). The hypoxic tumor microenvironment activates this cellular survival mechanism, which cascades and ultimately plays a role in the progression of a number of the hallmark traits of cancer, including angiogenesis, metastasis and invasion, altered metabolism, and apoptotic resistance. Rapamycin, an inhibitor of mTOR, acts as an upstream antagonist of HIF1 α subunit (HIF1A). mTOR regulates a signaling cascade through control of phosphorylation of proteins S6K1 and 4E-BP-1, important for translation of mRNAs involved in cell growth and proliferation processes, including HIF (10, 36). Effective reduction of HIF1A levels with rapamycin could significantly reduce the prosurvival signaling of the hypoxic response and inhibit tumor progression. Combination therapy with rapamycin and vasculature-targeted enzyme prodrug therapies may be especially complementary as targeted enzyme prodrug therapy has been shown to result in necrotic tumor cores and reduced blood flow (27).
Induction of tumor cell apoptosis with methylselenol and reduction of prosurvival signaling with rapamycin can be further exploited through augmentation of antitumor immunity. Tumor cell apoptosis increases antigen presentation and stimulation of tumor-specific effector T cells (37), an effect that could be amplified through the dose-dependent immunomodulatory properties of cyclophosphamide. In particular, low-dose cyclophosphamide has been shown to selectively deplete regulatory T cells (Treg), potentially tipping the balance from immune tolerance and evasion to an active anticancer immunity. Cyclophosphamide acts by alkylating DNA, creating interstrand and intrastrand cross-links that ultimately result in cell death (38, 39). Increased apoptosis and decreased immunosuppressive activity of Tregs upon low-dose cyclophosphamide treatment result from a hypersensitivity possibly related to a decreased capacity for DNA damage repair (39). Combined with other antitumor therapies, low-dose cyclophosphamide can enhance the antitumor immune response, leading to tumor rejection and improved survival (40–42).
The work described aims to address the efficacy of a three-pronged anticancer strategy: targeted delivery of a cytotoxic compound with an enzyme prodrug system, modulation of the prosurvival hypoxic response within the tumor microenvironment, and promotion of antitumor immunity. Together, this strategy aims to address many of the hallmark traits of cancer while minimizing the impact on healthy tissue. We validate the efficacy of the targeted enzyme prodrug system using human MDA-MB-231 and murine 4T1 triple-negative breast cancer models. Our findings using the highly metastatic 4T1 mammary tumor in immunocompetent mice demonstrate a synergistic effect on tumor volumes when combining rapamycin with the enzyme prodrug therapy and a significant antimetastatic effect when including cyclophosphamide.
Materials and Methods
Fusion protein construction, expression, and purification
The methionine gamma-lyase and Annexin A5 (ANXA5) fusion protein (MGL-ANXA5) was produced and purified as described previously (26). Wild-type human CTH displays no activity toward l-methionine or l-selenomethionine; however, three mutations impart MGL activity to human CTH (16). Equivalent mutations were applied to mouse CTH to impart MGL activity (E58N, R118L, E338N). Mutant mouse CTH (mCTH), ANXA5, and Annexin A1 (ANXA1) gene fragments were codon optimized for Escherichia coli (E. coli) protein production using DNAWorks (Helix Systems; NIH, Bethesda, MD) and synthesized (Life Technologies) with each fusion, mCTH-ANXA5 and mCTH-ANXA1, consisting of three fragments. Fragment sizes ranged from 371 to 1,000 amino acids with 40-bp overlapping regions on each end of the fragment. Gene fragments were directly assembled into pET-30Ek/LIC (EMD Chemicals) using the Gibson Assembly method (43). Gibson assembly Master Mix (New England Biolabs) was held at 50°C for 60 minutes in a thermocycler with the gene fragments and vector. The assembled product was transformed into NEB 5-alpha competent E. coli, expanded, and cultured on kanamycin plates. Colonies were sequenced (Oklahoma Medical Research Foundation) and transformed into BL21(DE3) competent cells for expression or into T7 Express lysY competent cells (New England Biolabs) for enhanced protein yields.
BL21(DE3) E. coli were grown in Luria broth medium at 37°C and 200 rpm to an OD600 nm of 0.6. T7 Express lysY E. coli were grown in TB medium at 37°C and 200 rpm to an OD600 nm of 1.2. Growth occurred in two steps: initially at 10 mL and then expanded to 1 L. Isopropyl β-D-thiogalactopyranoside (IPTG) was added to a concentration of 0.4 mmol/L for 6 hours at 30°C at 180 rpm for mCTH-ANXA1 or 1.0 mmol/L for 19 hours at 25°C and 200 rpm for mCTH-ANXA5.
Recombinant fusion proteins were purified using immobilized metal affinity chromatography with immobilized Ni2+ to isolate the fusion protein as described previously (26, 28, 29). An endotoxin removal step using a 1% Triton X-114 wash was included, and the 6× His Tag was removed with HRV-3C protease (Thermo Scientific Pierce).
Endotoxin levels were assessed through a chromogenic Limulus Amebocyte Lysate assay (Lonza) according to the manufacturer's instructions. Purity was assessed through a densitometric analysis with ImageJ software (FIJI build; NIH) from SDS-PAGE gels with Coomassie brilliant blue staining of protein samples. Enzyme activity with l-methionine and l-selenomethionine was assessed according to the methioninase assay described previously (26).
Cell lines and culture conditions
Human MDA-MB-231 and murine 4T1 breast cancer cell lines were purchased from ATCC in 2010 and 2011, respectively. Human endothelial HAAE-1 cells were purchased from Coriell Cell Repositories in 2012. All cells were cultured according to conditions recommended by the suppliers and described previously (29, 44). ATCC verifies cell identity with isoenzymology methods and short tandem repeat analysis. Coriell Cell Repositories perform cell identification with nucleoside phosphorylase, glucose-6-phosphate dehydrogenase, and lactate dehydrogenase isoenzyme electrophoresis and chromosome analysis. No further authentication was performed; however, no cells were maintained in culture for more than 6 months. Experiments were performed within six passages, except in the generation of TdTomato-expressing 4T1 cells and GFP-expressing MDA-MB-231 cells.
In vitro binding and cytotoxicity analysis
The specific, calcium-dependent binding of the Annexin fusion proteins was quantified as described previously (26, 28, 29). Live cell imaging with GFP-expressing MDA-MB-231 cells and Dylight 680–conjugated fusion protein confirmed binding to the cell membrane (see Supplementary Fig. S1). The cytotoxicity assays were performed over 3 days in 24-well plates under standard culture conditions with calcium-supplemented medium (2 mmol/L Ca2+). On day 0, cells were incubated with 100 nmol/L fusion protein for 2 hours at 37°C. The plates were washed, and medium containing varying concentrations of selenomethionine was added. An Alamar Blue assay was performed daily with a 4-hour incubation of 10% Alamar Blue (Life Technologies) to assess viability using a fluorescence measurement at an excitation wavelength of 530 nm and emission at 590 nm. Medium containing the prodrug selenomethionine was replaced daily.
Mouse tumor models and treatments
Six-week-old female BALB/cJ and SCID mice were purchased from The Jackson Laboratory and housed in a pathogen-free facility at the University of Oklahoma Health Sciences Center (Oklahoma City, OK) in accordance with protocols approved by the Institutional Animal Care and Use Committee. Mice were injected in the mammary fat pad number 4 with 105 4T1-TdTomato or MDA-MB-231 mouse breast cancer cells suspended in 50 μL PBS and an equal volume of Matrigel. Mouse body weight and tumor volume were monitored every 3 to 4 days. Tumor volume was calculated with the modified ellipsoid formula volume = (1/2) × (length × width2) using caliper measurements of the longest dimension and perpendicular width.
Mice bearing orthotopic 4T1 or MDA-MB-231 tumors were randomized into groups (7–10/group) prior to the start of treatment on day 10 postinoculation of 4T1 cells or day 48 for MDA-MB-231 xenografts. Fusion protein was administered daily at 10 mg/kg by intraperitoneal injection. Selenomethionine (5 mg/kg i.p.), rapamycin (5 mg/kg i.p.), and cyclophosphamide (10 mg/kg i.p.) were administered daily 10 hours after fusion protein injection.
Antibody titers
Protein-specific antibody titers were determined on the basis of modified protocols (45, 46). Blood samples were collected from mice at 0, 1, 2, and 3 weeks of treatment. A sandwich ELISA assay was performed to determine protein-specific antibody titers. A 0.1 mol/L carbonate coating buffer and 20 μg/mL fusion protein were incubated overnight at 4°C on high binding capacity ELISA 96-well plates. Plates were then washed and blocked with FBS, and plasma dilutions were incubated overnight at 4°C. Following additional washes, goat anti-mouse IgG and IgM conjugated to HRP (Jackson ImmunoResearch Laboratories, Inc.) were used to develop o-phenylenediamine for quantification on a BioTek Synergy plate reader.
Tissue analysis and IHC
At the conclusion of the treatment period for mice with 4T1 tumors, 3 mice per group were sacrificed for tissue analysis. Tumors were removed and fixed in 10% neutral-buffered formalin. IHC slides were produced using 3,3′-diaminobenzidine (DAB) staining for activated caspase-3 (Abcam 2302), Ki-67 (Abcam 15580), CD-31 (Abcam 2302), and HIF1A (Abcam 82832) with hematoxylin counter staining. Heat-mediated antigen retrieval was performed at pH 6. Tumor sections for activated caspase-3 and Ki-67 staining were imaged at 20× using a Nikon Eclipse E800 microscope collecting 15 images per section. An automated ImageJ macro was developed for quantification of DAB staining and visually verified (Supplementary Figs. S2 and S3). HIF1A images were collected with a Leica stereomicroscope to capture the entire tumor section.
The lungs were removed and immediately analyzed for the presence of TdTomato-positive metastatic nodules using a Leica stereomicroscope. The size and number of metastatic nodules were quantified with ImageJ software. In addition, the spleen was removed, and the cells were mechanically dissociated from the organ in FACS buffer and passed through a 70-μm cell strainer. A Mouse Regulatory T Cell Staining Kit (eBioscience) was used to stain splenocytes for flow cytometry according to the manufacturer's instructions. A BD Biosciences Accuri C6 flow cytometer was used for data acquisition and analysis. CD4+ CD25+ Foxp3+ cells were considered to be Tregs and quantified as a percentage of total spleen lymphocytes. Gating is shown in Supplementary Fig. S4.
Statistical analysis and synergism assessment
Statistically significant differences of cell viability, tumor volume, and section staining were assessed using a one-way ANOVA and Tukey–Kramer multiple comparisons test with GraphPad Prism software. Assessment for synergism of combination therapies was performed using the Bliss independence model on primary tumor growth inhibition (47, 48). The assessment of synergism was determined by calculating a synergism assessment factor, which is defined as the percent inhibition observed for the combination therapy minus the additive probability of the % inhibitions observed for the separate therapies (see the Supplementary Data for additional information). Statistical significance for survival was assessed on the basis of Kaplan–Meier survival curves using the log-rank (Mantel–Haenszel) test with a 0.05 significance level corrected for family-wise significance based on the number of comparisons according to the Bonferroni-corrected threshold (significance level/number of comparisons).
Results
The mCTH enzyme was fused to ANXA1 and ANXA5 and confirmed to have methioninase activity that allows for the conversion of selenomethionine to methylselenol. As this enzyme prodrug therapy utilizes the conversion of l-selenomethionine to methylselenol and not l-methionine depletion alone, we evaluated the activity of the engineered mouse CTH toward l-selenomethionine. mCTH-ANXA1 and mCTH-ANXA5 exhibited an activity of 0.75 ± 0.2 U/mg and 0.95 ± 0.1 U/mg, respectively. Enzyme activities with the substrate l-methionine for mCTH-ANXA1 and mCTH-ANXA5 were 1.0 ± 0.1 and 1.3 ± 0.2 U/mg, respectively, compared with 1.0 U/mg for MGL-ANXA5 (26). Yields of purified fusion protein were achieved in excess of 120 mg/L of culture medium, with >95% purity and endotoxin levels of <10 EU/mg.
Strong in vitro binding and cytotoxic efficacy on human breast cancer cells and endothelial cells representative of the phosphatidylserine exposure of tumor vasculature were apparent and comparable among the three systems (mCTH-ANXA1, mCTH-ANXA5, and MGL-ANXA5). Figure 1A summarizes the dissociation constants, all in the nanomolar range, for the three systems, and binding to the cell membrane is shown (Fig. 1B); this binding image is consistent with the protein's being designed for Annexin V to bind to phosphatidylserine expressed on the cell membrane. A cytotoxicity analysis confirms the therapeutic potential of the enzyme prodrug system on MDA-MB-231 and 4T1 breast cancer cells (Fig. 1C and D). The Supplementary Data contain additional in vitro binding data, including the binding curves and hyperbolic fit for dissociation constant determination (Supplementary Fig. S5) and additional live cell confocal multiphoton microscopy data (Supplementary Fig. S1). Relatively minor in vitro differences between the ANXA1 and ANXA5-targeted mCTH failed to distinguish the superiority of one system over the other, warranting a comparative in vivo study.
In vitro analysis of fusion protein binding and enzyme prodrug system cytotoxicity. A, The dissociation constants for the fusion proteins were determined from the specific binding curves (available in Supplementary Data) on breast cancer and nonconfluent endothelial cells representative of tumor vasculature endothelium. B, Live cell imaging indicates membrane binding of Dylight 680 conjugated fusion protein the cell membrane of GFP expressing MDA-MB-231 cells. C and D, The cytotoxic effects of the mCTH-ANXA1, mCTH-ANXA5, and MGL-ANXA5 enzyme prodrug systems were compared for MDA-MB-231 (C) and 4T1 (D) breast cancer cells. Groups that received fusion protein were treated on day 0. Selenomethionine was administered daily. Viability was determined by the Alamar Blue assay on day 3, and each sample was represented as a percentage of the vehicle-treated control. Statistical analysis was performed with a one-way ANOVA test with data presented as mean ± SEM (n = 3). Statistical significance versus vehicle-treated control is denoted by * (P < 0.001).
Pharmacokinetic analysis of the mammalian fusion protein indicates rapid entrance into the bloodstream and complete clearance from circulation within 10 hours, shown in Fig. 2A. This result is consistent with similar fusion proteins and allows for daily dosing of both the fusion protein and the prodrug. IHC staining confirms accumulation of the fusion protein on the tumor vasculature and within the tumor tissue (Supplementary Fig. S6).