Microparticle Encapsulation of a Prostate-targeted Biologic for the Treatment of Liver Metastases in a Preclinical Model of Castration-resistant Prostate Cancer

The liver is a common site of metastasis for many primary cancers, including colorectal, breast, lung, kidney, skin, pancreatic, and prostate (1). In fact, the liver is the third most frequent site of hematogenously disseminated disease in men with metastatic castration-resistant prostate cancer (mCRPC) after bone and lung (2). Liver metastasis is associated with a worse median overall survival in these men compared with those with disseminated disease in other tissue sites, including lung, lymph nodes, and bone (3, 4). These facts document that secondary liver metastases are a significant clinical health burden in need of novel targeted therapeutic approaches.

Proaerolysin is an attractive candidate agent that can be engineered for such a purpose. Proaerolysin is a secreted bacterial toxin from the species Aeromonas hydrophila that induces cell lysis via oligomeric pores inserted into the target cell membrane (5, 6). To regulate this lytic function, proaerolysin requires specific proteolysis of its C-terminus by furin, a ubiquitously expressed mammalian protease that removes a small inhibitory peptide to induce oligomerization and subsequent pore formation (7). Thus, to target this toxin to prostate cancer cells, we have previously mutated the native furin activation site to a peptide sequence selectively recognized by PSA (8). PSA is a chymotrypsin-like serine protease that is uniquely expressed by normal and malignant prostate epithelial cells and is only enzymatically active in the extracellular fluid surrounding these cells (8–10). This tumor-restricted activity is the result of PSA binding to highly abundant serum protease inhibitors when present in the circulation of men with prostate cancer (10).

This mutant PSA-activated protoxin, known as PRX302 (Supplementary Fig. S1), is selectively lethal to cell lines at picomolar doses only in the presence of enzymatically active PSA (8). Furthermore, intratumoral injections of PRX302 yield robust tumor regressions in PSA-expressing xenografts (8). As a part of its mechanism of action, PRX302 binds with high affinity to cell membranes via ubiquitously expressed glycosylphosphatidylinositol (GPI)-anchored proteins present on the surface of cells throughout the body (11), which leads to poor biodistribution and toxicity when delivered systemically. This is supported by the increased therapeutic index observed with a mutant form of the protoxin that is unable to bind to GPI anchor proteins (12, 13). However, because intraprostatic injection of PRX302 induces significant localized cell death, it is being evaluated clinically as a local therapy for men with primary prostate cancer or symptomatic benign prostate hyperplasia (BPH) after successfully meeting primary endpoints in phase II and III trials, respectively (14–16). Consequently, to harness the potent and selective toxicity of PRX302 as a systemic anticancer therapy for metastatic disease, it must be selectively delivered to the tumor microenvironment in a form that prevents it from immediately interacting with GPI anchors on cell membranes in nontarget tissue.

This challenge can potentially be overcome by using microparticles composed of poly-lactic-co-glycolic acid (PLGA), a biodegradable polymer capable of encapsulating small-molecule drugs and biologics, which slowly degrade over time for controlled release of the encapsulated agent (17, 18). We hypothesized that encapsulation of PRX302 within microparticles would physically prevent it from immediately interacting with GPI anchor proteins to enable systemic distribution of the protoxin. In addition, the pharmacokinetics and biodistribution of systemically infused particles are greatly influenced by their chemical and physical properties, including size (19, 20). Indeed, the percent of the injected dose deposited in the liver following intravenous injection for PLGA particles positively correlates with particle size (i.e., larger particles, greater liver deposition; ref. 20). This distribution pattern suggests that PLGA particles with large diameters may be ideal vectors for targeting liver disease, such as prostate cancer metastases. Importantly, PLGA is used clinically in multiple FDA-approved formulations for a variety of therapeutic indications, including successful applications in the context of prostate cancer (e.g., Lupron Depot formulation; ref. 21).

To ultimately realize this systemic delivery strategy for PRX302, we first need to develop robust reagents to accurately characterize protoxin functionality and release kinetics from PLGA microparticles. Herein, we developed and optimized a robust and highly sensitive sandwich ELISA to quantify proaerolysin/PRX302 in multiple biologic matrices. Release kinetics of two independent PRX302-loaded microparticle formulations with distinct release profiles were assessed, in addition to confirming PSA-dependent hemolytic function of the released protoxin and cytotoxicity to prostate cancer cells in vitro. Importantly, microparticle encapsulation blocks the interaction between PRX302 and GPI anchor proteins present on cell membranes. This results in an increased therapeutic index leading to significant antitumor efficacy in vivo in a preclinical model of prostate cancer liver metastasis following a single dose of intravenously administered PRX302-loaded microparticles. Consequently, the studies reported herein document two critically important proof-of-principle concepts necessary for optimization of the proposed microparticle-based PRX302 systemic drug delivery platform. Specifically, the development of sensitive and robust biochemical assays to detect the release and function of encapsulated PRX302 from PLGA microparticles, in addition to proof-of-concept demonstration of the in vivo efficacy of systemically delivered PRX302 using a well-established clinically compatible drug delivery platform.

See Supplementary Materials and Methods for additional details.

See Supplementary Materials and Methods for additional details.

PRX302 was expressed in Aeromonas salmonicida as described previously (8). See Supplementary Materials and Methods for additional details.

See Supplementary Materials and Methods for additional details.

See Supplementary Materials and Methods for additional details.

A modified double emulsion protocol was used to prepare PRX302-loaded PLGA microparticles similar to methods described previously (22, 23). Briefly, PLGA [10 kDa (50:50)], the encapsulating polymer, was dissolved in dichloromethane (DCM). The PLGA-DCM solution was then homogenized (15,000 rpm, 1 minute) on ice with dropwise addition of the PRX302-containing solution. The emulsion was then homogenized two additional times, once with 1% polyvinyl alcohol (PVA) followed by a 0.3% PVA solution. The resultant emulsion was allowed to evaporate for 4 hours to remove the organic solvent component. After evaporation, the microparticle solution was centrifuged (5,000 rpm, 10 minutes) to pellet the microparticles and the supernatant containing PVA was aspirated. The pellet was then suspended in a small volume of ultrapure water, frozen at −80°C, and lyophilized.

Subsequently, one batch of PRX302-loaded microparticles was coated with two lipid layers. For coating the first lipid layer, 1, 2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) and 1, 2-dioleoyl-3-trimethylammonium-propane (methyl sulfate salt; DOTAP) were dissolved in DCM followed by the slow evaporation of solvent using a rotary vacuum evaporator, resulting in the formation of a thin lipid film. The lipid film was then hydrated using water at 45°C for 1 hour to form liposomes. PRX302 microparticles were added to the liposome suspension and hydrated at 40°C for 1 hour. DPPC-DOTAP–coated PRX302 microparticles were centrifuged at 10,000 × g for 10 minutes and resuspended in water. For coating the second lipid layer, a pure DPPC thin film was prepared by slow evaporation of a DPPC solution in DCM. DPPC-DOTAP–coated PRX302 microparticles were hydrated with the DPPC thin film and water at 40°C for 1 hour. Lipid-coated PRX302 microparticles were centrifuged at 10,000 × g for 10 minutes and resuspended in water prior to lyophilization as described above. Lipid-coated blank (i.e., unloaded) PLGA microparticles were made according to similar methods to be used as controls.

Microparticle morphology was characterized by high-resolution scanning electron microscopy (SEM; Zeiss Merlin High-resolution SEM; acceleration voltage, 2 kV). For electron microscopic analysis, particles were lyophilized and placed on carbon tape, and coated with a thin layer of gold using a sputtering machine.

See Supplementary Materials and Methods for additional details.

Hemolysis assays were performed as described previously (8). Briefly, human red blood cells (RBC) were isolated from whole blood via centrifugation at 800 × g for 15 minutes, and the buffy coat was discarded. An aliquot of RBCs was diluted to 4% in normal saline and centrifuged at 1,000 × g for 5 minutes to remove cellular debris followed by a second wash in saline. Dilutions of recombinant PRX302 or unknown samples were made in PBS with 1% BSA after being incubated with or without recombinant human PSA (5 μg/mL) for 2 hours at 37°C. These solutions were then mixed with the washed 4% RBC solution and incubated for 2 hours at 37°C while rotating. Samples were centrifuged at 1,000 × g for 5 minutes to remove intact RBCs. The supernatant was transferred to a 96-well plate and absorbance was read at 540 nm. The samples were then plotted as a function of percent hemolysis when compared with RBCs incubated with 2% Triton X-100 as the 100% lysis-positive control.

To accurately determine PRX302 release rates from different microparticle formulations, we used the previously described sandwich ELISA protocol. Lyophilized PRX302 microparticles were washed twice in saline containing 1% BSA for 5 minutes each to remove any free PRX302 and then resuspended in this solution at a concentration of 1 mg/mL. The solution was sonicated on the lowest power setting for 3 × 10 seconds pulses on ice and then incubated at 37°C while shaking under atmospheric conditions (i.e., 21% O₂ and 0.04% CO₂). At the indicated timepoints, microparticle suspensions were centrifuged for 15 minutes at 800 × g with the resulting supernatants and pellets stored at −80°C until analysis. Supernatants were diluted (1:2,000) in ELISA assay buffer to obtain readings in the dynamic linear range of the assay.

PRX302-loaded microparticles were allowed to condition in serum-free media (i.e., RPMI + 1% l-glutamine) for 96 hours prior to collection of the supernatant via centrifugation at 800 × g for 15 minutes to remove the remaining microparticles. For incubations with PRX302-loaded microparticle-conditioned supernatant, PC3 and DU145 cells were plated at a density of either 500 or 1,000 cells per well in a 96-well plate, respectively, in serum-containing media for 48 hours prior to changing to serum-free media. The PRX302 microparticle supernatant was then diluted serially directly into each well and incubated for 48 hours with or without PSA (5 μg/mL). Cell morphology was monitored regularly using a light microscope, and cell viability was determined by MTT Assay (Promega). The number of cells per well was calculated on the basis of a standard curve of known cell numbers.

Peripheral blood mononuclear cells (PBMC) were isolated via Ficoll gradient centrifugation from discarded samples obtained from the clinical chemistry laboratory at Johns Hopkins University (Baltimore, MD) via institutional review board–approved protocols. PBMCs (300,000/sample) were incubated with 1 nmol/L FLAER [i.e., Alexa 488–labeled proaerolysin variant (Cedarlane)] in the presence or absence of increasing concentrations of PRX302 or PRX302-loaded microparticles (1–100 nmol/L) for 20 minutes in PBS supplemented with 2 mmol/L EDTA and 0.5% BSA in the dark at room temperature. After washing, FLAER binding was analyzed by flow cytometry (ex 488/em 530 ± 30) using a BD FACSCalibur (BD Biosciences) with competition for binding documented by a decrease in mean fluorescence intensity (MFI). Assay was performed in triplicate.

All animal experiments were performed in accordance with protocols reviewed and approved by the Johns Hopkins University Animal Care and Use Committee. The hemi-spleen liver metastasis model was performed as described previously with minor modifications using LN95 cells (24–26).

When circulating PSA was detectable in all animals (i.e., ≥0.01 ng/mL, ∼30 days postinoculation), they were stratified into four treatment groups on the basis of total PSA levels (n = 5/group). Animals were injected intravenously with a single dose of PRX302-loaded or blank (i.e., empty/unloaded) microparticles suspended at 0.5 mg/mL (i.e., 100 μg total × 5.9% PRX302 by weight = 5.9 μg total PRX302) suspended in 200 μL sterile saline, or PRX302 alone (sans microparticles) at the median lethal dose (LD₅₀ = 1 μg) in 200 μL sterile saline. Total PSA plasma levels and body weights were monitored weekly as surrogates of tumor burden and toxicity, respectively. At the end of study, animals were euthanized via CO₂ asphyxiation and total liver weights were measured. Total tumor burden was calculated by subtracting the average liver weight determined from 15 nontumor-bearing age-matched animals from the total liver weight in each tumor-bearing animal, and this difference was considered the liver tumor burden in that animal. Livers were subsequently formalin-fixed, paraffin-embedded, and processed for hematoxylin and eosin (H&E) and androgen receptor [AR (N-20), #sc-816, Santa Cruz Biotechnology] staining by the Sidney Kimmel Comprehensive Cancer Center Tissue Services Core (Baltimore, MD) according to standard protocols.

See Supplementary Materials and Methods for additional details.

To accurately determine the quantity of PRX302 released from protoxin-loaded PLGA microparticles, a sandwich ELISA was developed. The final optimized 8-point standard curve has a linear range over approximately 1.5 orders of magnitude between approximately 0.78 and 50 ng/mL (∼15–1,000 pmol/L) and fits a standard 4-parameter sigmoidal model (Supplementary Fig. S1). Additional validation of the ELISA can be found in the Supplementary Data, including documentation of intra- and interassay variation, specificity, parallelism, linearity of dilution, spike and recovery assays, and validation in different biological matrices (Supplementary Fig. S2; Supplementary Tables S1–S5). Consequently, PRX302 concentrations can be robustly and accurately quantified using this validated assay.

To determine the amount of functional PRX302 released from microparticles, a hemolytic titer assay was performed (8, 27). Briefly, this assay measures the release of hemoglobin from lysed RBCs in solution relative to a detergent-lysed positive control (Fig. 1A). Utilizing this assay, a hemolytic “standard curve” of known concentrations of functional recombinant PRX302 versus lytic activity was generated in which a 4-parameter sigmoidal relationship was observed (Fig. 1B). The linear range of this assay is between 300 and 1,000 ng/mL with an LOD of approximately 150 ng/mL (i.e., ∼3 nmol/L). PRX302 concentrations in a dilution series determined independently using the ELISA and hemolysis assays document a strong positive linear correlation (R² = 0.975) between the two assays (Fig. 1C). Therefore, the amount of hemolysis observed in a sample can be used to determine the percentage of functional protoxin released from microparticles relative to the amount of total PRX302 present as determined by the ELISA.

Next, these validated assays were used to accurately determine the amount of functional PRX302 released from different PLGA-based microparticle formulations that were generated using a modified double emulsion protocol (22). One batch of PRX302-loaded PLGA microparticles was left unmodified (Fig. 2A). Unmodified microparticles are often associated with an immediate release of surface-associated drug (28). This is undesirable for the protoxin delivery platform because it leads to elevated levels of free protoxin in the circulation where it can bind to GPI anchor proteins prior to reaching the tumor microenvironment for microparticle-based protoxin release as intended. Therefore, in a second batch, the surface of the microparticles was coated with lipids to reduce the immediate release of this surface-associated PRX302 (Fig. 2B). DPPC is a neutral lipid (zero net surface charge), while DOTAP is a cationic lipid (net surfaced charge is +1). It was hypothesized that a combination of the two lipids for coating would render the net surface of the microparticles slightly positive, which would enhance the uptake by negatively charged target cells. In addition, the double layer was introduced to ensure that the net surface charge was not very high (i.e., >+30 mV), because highly cationic microparticles would significantly disrupt plasma membrane integrity, in addition to causing mitochondrial and lysosomal damage (29). Because, the second layer only contains DPPC, the net surface charge was controlled at approximately 24 mV compared with −17.5 mV for the unmodified microparticles (Fig. 2C). The microparticles have an average diameter of approximately 1.3–1.5 μm with a polydispersity index of 0.2 and 0.3 for the coated and uncoated microparticles, respectively; each with an encapsulation efficiency of approximately 80%. PRX302 (i.e., drug) loading was 3% for the unmodified and 1.47% for the lipid-coated microparticles (Fig. 2C).

PRX302 release from each of these microparticle formulations over time was quantified using the validated ELISA. Two 5-minute washes were performed to remove surface-associated protoxin from the unmodified microparticles to accurately assess release kinetics of the encapsulated protoxin from each preparation. As determined using the ELISA, approximately 30% of the total encapsulated protoxin was released from the unmodified microparticles over 10 days in solution (i.e., saline + 1% BSA) at 37°C (Fig. 2D); reaching concentrations of approximately 300 nmol/L in the microparticle supernatant. In addition to the initial release phase, there was potentially a second enhanced release phase at approximately 180 hours that may be due to initiation of bulk microparticle degradation; thus, releasing PRX302 at a higher rate compared with the slow release observed at previous timepoints when polymer degradation was slow. However, the high variability in drug release at later timepoints was not statistically different and a biphasic response cannot be concluded without additional data points. It is also possible that this was due to a lack of homogeneous microparticle distribution in the medium and/or aggregation of microparticles following centrifugation.

In contrast, and as predicted because of the surface modification, release kinetics were significantly slower from the lipid-coated microparticles, resulting in <2% of the total encapsulated protoxin being released over the same period of time under the same conditions (Fig. 2E). These results document that the validated assays described above can accurately assess PRX302 delivery from microparticle formulations with distinct release profiles, which is essential for continued optimization of the drug delivery platform. Because insignificant amounts of protoxin were released from the lipid-coated microparticles, only the unmodified microparticles were used in subsequent assays.

To demonstrate the PSA-dependent pore formation of PRX302 released from unmodified microparticles, the hemolysis assay described above was utilized. Supernatant from PRX302-loaded microparticles incubated in saline + 1% BSA for 24 hours was mixed with RBCs in the presence or absence of a physiologically relevant concentration of exogenous PSA (5 μg/mL). This assay demonstrated that the conditioned supernatant contains functional PSA-dependent hemolytic activity that is not present in unconditioned supernatant (Fig. 3A).

Next, cytotoxicity of the released protoxin was confirmed by incubating PSA-negative PC3 and DU145 prostate cancer cells with PRX302-loaded microparticle-conditioned supernatant in the presence or absence of exogenous PSA (Fig. 3B and C). No toxicity was observed using conditioned supernatant from unloaded (i.e., “blank”) PLGA microparticles independent of exogenous PSA. In contrast, substantial cell death (>85%) was observed in the presence of conditioned supernatant from protoxin-loaded microparticles, but only in the presence of exogenous PSA. Thereby, documenting that PRX302 is released from the microparticles as an intact functional prodrug that retains its PSA-dependent cytotoxic activity against prostate cancer cells.

To determine whether microparticle encapsulation could prevent PRX302 from interacting with GPI anchor proteins on the surface of cells as needed for effective systemic delivery of the protoxin to tumors, a competition assay was performed with a fluorescently labeled inactive variant of the parental toxin (i.e., proaerolysin) known as FLAER (30). This reagent is used clinically to diagnose a rare hematopoietic disorder [i.e., paroxysmal nocturnal hemoglobinuria) and fluorescently labels all cells expressing GPI anchor proteins (including PBMCs). Consequently, it can be used to quantify binding to GPI anchor proteins on the surface of cells. As expected, unencapsulated PRX302 competes with FLAER for binding to GPI anchor proteins in a dose-dependent manner as documented by a decrease in fluorescence with increasing concentrations of PRX302 [IC₅₀ = ∼10 nmol/L; Fig. 4A and C). In contrast, PRX302 encapsulated inside of microparticles had a >20-fold lower relative binding affinity (IC₅₀ >200 nmol/L) for GPI anchor proteins (Fig. 4B and C). This abrogation of GPI anchor protein binding in the context of microparticle encapsulation is likely a function of sequestration of the protoxin within the polymer, which physically prevents it from immediately interacting with outside partners until the protoxin is released from the microparticle.

To document the in vivo efficacy of this approach, hemi-spleen injection of LN95 cells was performed to seed mCRPC liver metastases (24–26). LN95 was derived from LNCaP prostate cancer cells by long-term passaging under continuous androgen deprivation conditions, which produced a line that grows robustly in castrated hosts and maintains PSA expression in the absence of androgen (i.e., castration resistant; refs. 31, 32). Therefore, LN95 accurately recapitulates a common mCRPC phenotype observed clinically, and circulating PSA can be used as a surrogate of tumor burden as we have documented previously (26).

Castrated animals seeded with liver metastases were treated with a single dose of blank or PRX302-loaded microparticles [0.2 mL × 0.5 mg/mL suspension containing 5.9% PRX302 by weight (i.e., ∼6 μg PRX302) in this specific batch] administered via intravenous injection once circulating PSA became detectable (∼30 days postinoculation). Circulating PSA levels in animals receiving the blank (i.e., “unloaded/empty”) microparticles increased at comparable rates with that observed in untreated control animals, while significantly lower levels of PSA could be measured in the plasma of animals receiving PRX302-loaded microparticles (Fig. 5A). Treatment was well-tolerated with minimal toxicity (Fig. 5B).

To determine whether similar effects could be achieved using PRX302 in the absence of microparticle encapsulation, PRX302 was administered intravenously at its previously determined median lethal dose (i.e., LD₅₀) of 1 μg; a dose that is approximately one-sixth less (i.e., smaller therapeutic index) than that which can be safely administered in the context of microparticle encapsulation. In addition to being poorly tolerated as anticipated (6/10 animals had to be euthanized), this dose did not suppress circulating PSA as effectively as PRX302-loaded microparticles (Fig. 5D). At the end of study (day 35 posttreatment), livers were harvested from all animals to determine total tumor burden (Fig. 5E) and histology was performed to identify prostate cancer foci in the livers of each animal (Fig. 5F). These results confirm those obtained via measurement of circulating PSA and further validated its use as a noninvasive surrogate biomarker of tumor burden in this model. Therefore, microparticle encapsulation increases the therapeutic index of PRX302 in vivo and its antitumor efficacy in a model of mCRPC liver metastasis.

The major goal of this study was to develop a systemically delivered targeting vector for PRX302, a highly potent prostate-targeted pore-forming biologic protoxin (Fig. 6). Liver metastases are a significant healthcare burden in the clinical management of patients with metastatic prostate cancer (2–4) and many other solid malignancies (1). Thus, innovative targeting strategies to selectively deliver therapeutic agents to sites of liver metastasis are urgently needed. One such strategy is the use of PLGA-based microparticles to deliver anticancer agents due to the known propensity of particles with large diameters to be deposited in the liver, which can receive approximately 50% or more of the injected dose following intravenous injection (19, 20).

To characterize release and functionality of the protoxin from these microparticles, two independent bioassays to accurately quantify the concentrations of PRX302 were developed and optimized. The first of these was a highly sensitive (LOD = 0.2 ng/mL) sandwich ELISA with a robust signal to noise ratio (average intra- and interassay variability <10%) and a dynamic linear range spanning approximately 1.5 orders of magnitude. The methodology was validated by demonstrating parallelism, linearity of dilution, and specificity, in addition to spike and recovery assays to determine necessary sample dilutions for accurate PRX302 quantification in various biologic matrices as needed for future in vitro studies and pharmacokinetics analyses. The second assay, a hemolytic assay, provided independent validation of the ELISA in addition to accurately documenting PSA-dependent functional activity of the protoxin upon release from microparticles.

Once validated, these independent assays were used to quantify PRX302 release from two independent microparticle formulations (unmodified and lipid coated) designed to produce different protoxin release profiles to document assay utility for assessing the kinetics of protoxin release for further optimization of the drug delivery platform. As predicted, the encapsulated protoxin was released more rapidly from the unmodified PLGA microparticles compared with the lipid-coated ones, achieving concentrations as high as 300 nmol/L in the microparticle supernatant. These release kinetics are comparable with those observed with other proteins encapsulated in PLGA microparticles (33, 34). PRX302 released from these microparticles retains PSA-dependent hemolytic activity and selective toxicity to prostate cancer cells in the presence of enzymatically active PSA in vitro and in vivo. This is demonstrated by the efficacy observed in a model of mCRPC liver metastasis following a single dose of PRX302-loaded microparticles administered intravenously with no obvious toxicity. This is in contrast to free PRX302, which results in substantial toxicity when administered systemically at doses that can produce an antitumor effect.

It is noteworthy that this microparticle strategy can achieve encapsulation and sustained release of PRX302, a biologic, while maintaining sufficient activity to achieve functional outcomes in vivo. This encapsulation strategy can, therefore, be used in the future to potentially revive many biologics and other anticancer agents that may have failed or stalled in development due to poor biodistribution or other unfavorable pharmacokinetics parameters. As an example, PRX302 (8), the potent PSA-activated protoxin utilized herein, is in clinical development as a local therapy for BPH and primary prostate cancer with encouraging results (14–16). Unfortunately, PRX302 cannot be developed as a systemic therapy for metastatic prostate cancer due to the fact that it binds to GPI-anchored proteins (11), which are ubiquitously expressed on the surface of cells throughout the body including the vasculature. This leads to poor biodistribution and prevents the protoxin from accumulating in tumor tissue to levels sufficient to achieve a therapeutic dose without causing systemic toxicity. Encapsulation of PRX302 in PLGA microparticles, as described herein, prevents this interaction and selectively delivers the protoxin to the liver, which improves the therapeutic index and enhances antitumor efficacy in a preclinical model of mCRPC liver metastasis that recapitulates key features of clinical disease.

These results also provide proof-of-principle data to suggest that other strategies to block this PRX302–GPI anchor interaction outside of the tumor microenvironment, such as coupling to albumin via a maleimide-based linker (i.e., maleimide-coupled albumin drug delivery or MAD; refs. 35, 36), have therapeutic potential for systemic targeting. One limitation of this study is that the current microparticle formulation was specifically designed to target prostate cancer liver metastases. However, similar approaches can potentially be coupled with strategies to further enhance selective targeting of disseminated tumor lesions throughout the body, including bone, using cell-based delivery methods or targeting moieties such as bisphosphonate analogs or a prostate-specific membrane antigen (PSMA) ligand or inhibitor (22, 37–40). Importantly, PSMA is not only overexpressed in prostate cancer, but is highly overexpressed on the tumor neovasculature of nearly all solid tumor types as well (41–43), suggesting this may be a pan-tumor targeting strategy that can be coupled with other anticancer agents. Another potential limitation is that the liver is a frequent site of metastasis for PSA-negative neuroendocrine tumors (44–46). While purely AR⁻ tumors would be unresponsive to a PSA-targeted approach such as this one, it should be noted that tumors with neuroendocrine features are often admixed with canonical AR⁺ adenocarcinoma as evidenced by the presence of lower, but detectable, levels of circulating PSA in these patients in many cases (2, 46–48). Consequently, these PSA-negative cells will also be killed via a bystander effect due to the fact that the released protoxin is activated in the extracellular fluid of the tumor microenvironment (10), which contains PSA from neighboring AR⁺ prostate cancer cells in these admixed tumors. To target purely AR⁻ tumors, a protoxin engineered for activation by an alternative protease expressed in an AR-independent manner could be employed, such as fibroblast activation protein or PSMA (41, 49).

In summary, this work demonstrates two important proof-of-principle concepts necessary for optimization of the proposed clinically compatible systemic PRX302 drug delivery platform. The first is the development of two independent, sensitive, and robust biochemical assays to accurately detect the release and function of a prostate-targeted biologic (i.e., PRX302) from PLGA microparticles. These tools will allow us to further optimize a controlled release protoxin delivery platform, analyze in vitro and in vivo release kinetics, and biodistribution in animal models, in addition to serving as informative assays for other therapeutic strategies utilizing PRX302. Second, sustained release of significant concentrations of functional PRX302 from PLGA microparticles over an extended period of time was demonstrated. Critically, the released protoxin retained its PSA-dependent cytotoxicity against prostate cancer cells in vitro and demonstrated antitumor efficacy in vivo in a preclinical model of castration-resistant prostate cancer liver metastases. Future work will aim to optimize dosing regimens and the release of functional PRX302 from alternative PLGA microparticle formulations to facilitate targeted release of the protoxin into the metastatic prostate cancer microenvironment following systemic delivery.

S.H. Ranganath reports being a director of Tvastra Innotech Solutions Private Limited, a company that has an option to license IP generated by S.H. Ranganath and may benefit financially if the IP is licensed and further validated, and S.H. Ranganath's interests were reviewed and are subject to a management plan overseen by Siddaganga Institute of Technology, Tumkur, Karnataka, India in accordance with their conflict of interest policies. J.M. Karp reports paid consultancy for companies including Stempeutics, Sanofi, Celltex, LifeVault Bio, Gecko Biomedical, Alivio Therapeutics, Skintifique, Molecular Infusions, Landsdowne Labs, Takeda, Quthero, and Mesoblast (see: https://www.karplab.net/team/jeff-karp), is an inventor on a patent that was licensed to Mesoblast, and holds equity in Frequency Therapeutics, a company that has licensed IP generated by J.M. Karp that may benefit financially if the IP is further validated. S.R. Denmeade reports grants from NIH (R01) during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

O.C. Rogers: Conceptualization, investigation, writing-original draft, writing-review and editing. L. Antony: Investigation, writing-review and editing. O. Levy: Conceptualization, investigation, writing-review and editing. N. Joshi: Conceptualization, investigation, writing-review and editing. B.W. Simons: Investigation, writing-review and editing. S.L. Dalrymple: Investigation, writing-review and editing. D.M. Rosen: Investigation, writing-review and editing. A. Pickering: Investigation, writing-review and editing. H. Lan: Investigation, writing-review and editing. H. Kuang: Investigation, writing-review and editing. S.H. Ranganath: Conceptualization, investigation, writing-review and editing. L. Zheng: Methodology, writing-review and editing. J.M. Karp: Conceptualization, funding acquisition, writing-review and editing. S.P. Howard: Conceptualization, resources, funding acquisition, investigation, writing-review and editing. S.R. Denmeade: Conceptualization, funding acquisition, writing-review and editing. J.T. Isaacs: Conceptualization, funding acquisition, writing-review and editing. W.N. Brennen: Conceptualization, funding acquisition, investigation, writing-original draft, writing-review and editing.

The authors would like to acknowledge the expert assistance of the Sidney Kimmel Comprehensive Cancer Center (SKCCC) Cell Imaging Facility, Flow Cytometry Core, Tissue Services, and Immunohistochemistry Cores supported by the SKCCC Cancer Center Support Grant [CCSG, (P30 CA006973)]. This work was supported by a Prostate Cancer Foundation (PCF) Young Investigator Award (to W.N. Brennen), SKCCC CCSG developmental funds (P30 CA006973 to W.N. Brennen), PCF/Movember Challenge Award (to J.T. Isaacs, S.R. Denmeade, and J.M. Karp), NIH-Prostate SPORE Grant P50 CA058236 (to S.R. Denmeade and J.T. Isaacs), the Department of Defense (W81XWH-13-1-0304 to J.T. Isaacs, S.R. Denmeade, and J.M. Karp), the NIH grant R01HL095722 (to J.M. Karp), and a Natural Sciences and Engineering Research Council of Canada Discovery grant (to S.P. Howard).

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.