Intravesical VAX014 Synergizes with PD-L1 Blockade to Enhance Local and Systemic Control of Bladder Cancer

Bladder cancer ranks fourth among all newly diagnosed cancers worldwide and is the second most frequently diagnosed urologic cancer (1). An estimated 79% of newly diagnosed cases are classified as early-stage nonmuscle-invasive bladder cancer (NMIBC). These patients are treated with transurethral resection of bladder tumor(s) followed by intravesical (IVE) adjuvant chemotherapy or immunotherapy with Bacillus Calmette-Guerin (BCG; ref. 2). The majority of patients will have one or more recurrences that are usually nonmuscle invasive, and a subset of patients with high-risk disease will progress to more advanced stages of disease despite IVE treatment. Those with treatment-refractory T1 (lamina propria invasion) disease, carcinoma in situ, and T2-T3 (muscle-invasive) disease have limited treatment options, aside from radical cystectomy, a procedure that often comes with persistent comorbidities and reduction in quality of life (3, 4). Thus, bladder-sparing therapeutic interventions for patients with BCG-unresponsive NMIBC, muscle-invasive T2–T3 disease, as well as an effective treatment for metastatic disease, all represent areas of high unmet need.

The discovery of T-cell immune-checkpoint pathways and the introduction of systemic immune-checkpoint blockade (ICB) therapy into clinical practice have resulted in remarkable durable treatment outcomes in a variety of solid tumor indications (5). Recent accelerated marketing approvals of ICB therapies targeting the programmed death ligand-1 (PD-1/PD-L1) axis for the second-line treatment of metastatic, locally advanced muscle-invasive bladder cancer and BCG-unresponsive NMIBC highlight the importance and promise of these agents in this spectrum of urothelial carcinoma indications (6). Durable response in a minority of these bladder cancer patient subpopulations is unparalleled, yet unfortunately, limited by modest overall response rates.

An emerging approach that has improved overall response rates in melanoma and a growing number of other solid tumor indications is the combination of systemically administered ICB with locally administered intralesional oncolytic viral (OV) therapy (7). Local intratumoral OV treatment is thought to induce a tumor-localized inflammatory immune response, mediated by viral entry and replication, followed by the later lytic phase of the viral lifecycle leading to local release and increased availability of tumor antigens for cross-presentation (8). This effect is thought to facilitate both a broader T-cell repertoire through epitope spreading and better CD8⁺ T-cell activation and migration back to the tumor sites to ultimately turn immunologically “cold” tumors “hot” (9). The tumor-clearing effector phase of the tumor-infiltrating lymphocyte (TIL) response mediated by OVs is then supported by systemically administered ICB, which acts to curtail the tumor-orchestrated immunosuppression of effector lymphocytes and/or T-cell exhaustion (9).

VAX014 is a novel integrin-targeted oncolytic agent based on recombinant bacterial minicells (rBMC) and is currently undergoing clinical investigation for the IVE treatment of NMIBC, where it will likely be used in the BCG-unresponsive disease setting (10, 11). In general terms, rBMCs are small spherical, replication-incompetent bacterial particles that contain all molecular components of the bacterial cells from which they are produced, except the bacterial chromosome (11–13). VAX014 rBMCs are derived from an inducible minicell-producing strain of Escherichia coli containing an immune-attenuated form of lipopolysaccharide (LPS with penta-acylated Lipid A) and further engineered to contain a preformed oncolytic protein, perfringolysin O (PFO) from Clostridium perfringens, along with surface-localized expression of invasin from Yersinia pseudotuberculosis, which has a nanomolar affinity for both alpha3beta1 (α3β1) and alpha5beta1 (α5β1) integrin heterodimers (11, 14, 15). Mechanistically, VAX014 first selectively targets tumor cells expressing either of these integrins through invasin, which upon receptor engagement stimulates the endocytosis of VAX014. Following uptake, the rBMC vector is quickly degraded in the endosomal/lysosomal pathway, facilitating the intracellular release of PFO and rapid oncolysis (11). Both α3β1 and α5β1 integrins have been reported as being upregulated in human bladder carcinoma specimens (16–18). The expression of α3β1 integrin in tumors of high-risk NMIBC patients has been identified as a negative prognostic indicator of recurrence and progression, and α5β1 is a known receptor of BCG (16, 19).

In this preclinical work, we investigated the immunotherapeutic effects of VAX014 following weekly IVE administration using both the MB49 and MBT-2 syngeneic orthotopic (o.t.) mouse models of bladder cancer as a prelude to clinical investigation in BCG-unresponsive NMIBC and other advanced bladder cancer settings. We showed antitumor activity was dependent on both CD8⁺ and CD4⁺ T cells and that durable protective antitumor immunologic memory resulted in response to VAX014. PD-L1 was identified as an adaptive immune-resistance mechanism to VAX014 in the MB49 model, and the combination of VAX014 with PD-L1 blockade (αPD-L1) in this model resulted in enhanced systemic tumor-specific immunity to improve the durable complete response (CR) rate of bladder tumors, clearance of distally engrafted secondary intradermal (i.d.) tumors, and effective control of pulmonary metastases. Actively responding distal i.d. tumors demonstrated an increase in Th1-polarized CD4⁺ T cells and activated CD8⁺ T cells, although this occurred without any increase in total tumor leukocyte or lymphocyte numbers, suggesting an immunophenotypic switch occurs in situ in tumors in response to combination therapy. Finally, to help establish or refute the biological relevance and posit clinical translational potential of each model with respect to the antitumor activity of VAX014, an analysis of expression patterns of β1, α3, and α5 integrin subunits was conducted in o.t. bladder tumors from both mouse models. This was followed by an analysis of β1, α3, and α5 integrin expression in locally advanced-stage T1–T3 human bladder tumor specimens, where positive expression of at least one of these two integrin subtypes was detected in 88.3% of tumor specimens tested. Together, these results provide a strong rationale for the clinical investigation of IVE administration of VAX014 and systemic ICB combinations in patients with advanced bladder cancer, with potential use in other solid tumor indications where local administration is feasible.

The MB49 murine urothelial carcinoma cell line was obtained from Dr. W.T. Godbey (Tulane University). The MBT-2 murine cell line, B16F10 murine melanoma cell line, and the HEK-Blue hTLR4 reporter cell line were purchased from the Japanese Collection of Research Bioresources Cell Bank, American Type Culture Collection, and InvivoGen, respectively. MB49 cells were cultivated in “complete” DMEM (Corning) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (HI-FBS; Corning) and 100 U/mL penicillin–streptomycin (P/S; Thermo Fisher). MBT-2 and B16F10 cells were grown in “complete” RPMI-1640 medium (Corning) with 10% (v/v) HI-FBS and 100 U/mL P/S. HEK-Blue hTLR4 reporter cells were maintained in DMEM with 100 U/mL P/S, 100 μg/mL normocin (InvivoGen), and 4 μL/mL HEK-Blue Selection antibiotics (InvivoGen). All cell lines were incubated at 37°C in 5% (v/v) CO₂ using a humidified tissue culture incubator (Thermo Fisher). Working stocks of each cell line were generated from the initial thaw, aliquoted, and stored in the vapor phase of liquid nitrogen until expanded for experimental use. Cell lines expanded for the experimental use were grown to 70% to 80% confluency in T-25 or T-75 flasks (Thermo Fisher). Cells were used for experiments after 2 to 5 passages. Single-cell suspensions were obtained by trypsinization in 40 μL/cm² TrypLE Express (Thermo Fisher) as instructed by the manufacturer unless indicated otherwise, and cells were counted using a Scepter 2.0 automated cell counter (Millipore Sigma). Cell line authentication and mycoplasma testing were performed by each respective vendor and not reevaluated throughout the course of the studies.

Six- to eight-week old female C57BL/6J mice (Strain 000064) and Toll-like receptor 4 LPS-deficient C3H/HeJ mice (Tlr4^Lps-d; Strain 000659) were purchased from the Jackson Laboratories. Mice were housed in microisolator cages and treated in accordance with NIH and American Association of Laboratory Animal Care regulations. All mouse experiments and procedures were approved by the San Diego State University Institutional Animal Care and Use Committee under Protocols 16-04-005Mc and 19-05-004Mc. For preclinical in vivo studies, the number of mice per group varied in each experiment but ranged from 6 to 16 per treatment group in studies with initial tumor challenge to ensure statistical power, and the number of mice in each study is described in the figure legends. Mice were randomized into treatment groups after tumor implantation to avoid tumor size bias except for the depletion studies, where depletion antibodies were administered prior to tumor injection as described below. In all experiments, mice were monitored every day following tumor installation(s) for health observations. Experiments were repeated at least twice as indicated in figure legends, with the exception of immune cell depletion studies, and the combined data from all experiments are presented. At least 5 mice per treatment group were used in tumor immunophenotyping studies and cytotoxicity assays. Statistical tests used in each study are described in the main text and figure legends.

VAX014 rBMCs were manufactured and characterized by Vaxiion Therapeutics and met all applicable and current product specifications utilized for clinical release. The control rBMCs, VAX-I (containing invasin but lacking PFO), wild-type LPS vector rBMCs (lacking both invasin and PFO with wild-type LPS), or immune-attenuated LPS vector rBMCs (lacking both invasin and PFO with penta-acylated Lipid A containing LPS) were produced and characterized by Vaxiion Therapeutics as described previously (11). The reduced ability of rBMCs containing immune-attenuated LPS to activate human Toll-like receptor 4 (hTLR4) was confirmed in a reporter cell line (Supplementary Fig. S1).

The integrin expression profile, potency, and oncolytic effect of VAX014 against both the MB49 and MBT-2 cell lines were performed in vitro prior to conducting efficacy studies (Supplementary Fig. S2). Methods for each evaluation are included below.

Integrin expression in MB49 and MBT-2 cells was determined via flow cytometry. MB49 and MBT-2 cells were grown in T-25 flasks and trypsinized as described above to make single-cell suspensions prior to staining. Cells were washed twice in staining/blocking buffer [1× PBS with 0.2% (w/v) BSA (Millipore Sigma), 0.02% (w/v) sodium azide (Millipore Sigma), and 10% (v/v) heat-inactivated goat serum (Abcam)]. A total of 2 × 10⁵ cells in 200 μL were incubated at 4°C for 30 minutes with 10 μg/mL anti-integrin β1, 10 μg/mL anti-integrin α5, or 10 μg/mL isotype control hamster antibody (Supplementary Table S1). Cells were then washed twice with staining/blocking buffer and incubated with 200 μL of 10 μg/mL Alexa Fluor 488–conjugated goat anti-hamster antibody for 30 minutes at 4°C. Cells were then washed twice with staining/blocking buffer prior to evaluating integrin expression using an S1000 flow cytometer (Stratedigm). Histograms were generated for visual comparison between each integrin expression with the isotype control using CellCapTure 3.1 software (Stratedigm). The integrin expression experiment was conducted once.

For the evaluation of the in vitro potency of VAX014, MB49 and MBT-2 cells were seeded at 5,000 cells/well in 96-well tissue culture plates in a total of 100 μL/well complete culture media and allowed to adhere for 4 hours. Titrations of VAX014 rBMCs or VAX-I rBMCs in saline [0.9% (w/v) NaCl; Teknova] were added to test wells (10 μL/well) over descending ratios of rBMCs to cultured cells (referred to as MOI), ranging from 2,000:1 to 4:1. Volume-matched saline vehicle was added to the wells with cells only for 100% viability control and wells with culture media only for 0% viability control. Each assay condition was performed in triplicate. After the addition of rBMCs, plates were centrifuged at 671 × g for 10 minutes and incubated at 37°C for 20 hours. Cell viability was then assayed by the addition of 10 μL/well PrestoBlue Cell Viability Reagent (Thermo Fisher) according to the manufacturer's instruction. Fluorescence (excitation/emission, 560 nm/590 nm) was measured using a SpectraMax M3 plate reader (Molecular Devices) with SoftMax Pro 5.4 software (Molecular Devices). The mean fluorescence of the 0% viability control wells was subtracted from each sample as background correction using Excel software (Microsoft). Percent cell viability in each sample was calculated by dividing the background-corrected fluorescence by the mean background-corrected fluorescence of the 100% viability control wells multiplied by 100% and percent cell viability as a function of MOI was plotted. The half-maximal effective concentration (EC₅₀) expressed as an MOI value was calculated using GraphPad Prism 9.0 software (GraphPad Software). Potency assays and EC₅₀ determinations were conducted a total of three times for each cell line.

For the evaluation of the oncolytic effect of VAX014 in vitro, MB49 and MBT-2 cells were seeded at 10,000 cells/well in 96-well tissue culture plates in a total of 100 μL/well complete culture media and allowed to adhere for 4 hours. Titrations of VAX014 or VAX-I rBMCs in saline were added to test wells (10 μL/well) over descending ratios of rBMCs to cultured cells (referred to as MOI), ranging from 2,000:1 to 4:1. Volume-matched saline vehicle was added to wells containing cell and designated for later use in the assay as 100% oncolysis background lysis controls. Each assay condition was performed in triplicate. After the addition of rBMCs, plates were centrifuged at 671 × g for 10 minutes and incubated at 37°C for 20 hours. The presence of lactic dehydrogenase (LDH) in culture supernatants, a direct measure of oncolysis, was detected using an In Vitro Toxicology Assay Kit (Millipore Sigma). Prior to assaying supernatants, a 10 μL/well of LDH Assay Lysis Solution was added to the 100% oncolysis control wells and incubated for 45 minutes. The plates were centrifuged at 671 × g for 5 minutes, and the supernatant (25 μL/well) was transferred to clear 96-well plates. Lactate Dehydrogenase Assay Mixture was prepared using the kit-provided reagents as instructed by the manufacturer and added to the supernatant (50 μL/well). After a 1-hour incubation in the dark at room temperature, the LDH assay reaction was stopped by the addition of 10 μL/well 1N HCl (Thermo Fisher). LDH signal was detected by measuring the absorbances at 690 nm and 490 nm using a SpectraMax M3 plate reader with SoftMax Pro 5.4. The 490 nm absorbance values were subtracted from the corresponding 690 nm absorbance values, and the mean background signal obtained from the 0% oncolysis control condition was further subtracted from each. The percent LDH release in each sample was then calculated by dividing the background-corrected LDH signal by the mean background-corrected LDH signal of the 100% oncolysis control wells multiplied by 100%. The percent LDH release as a function of MOI was plotted using GraphPad Prism 9.0. The oncolysis experiment was performed in triplicate.

The expression of PD-L1 in MB49 or MBT-2 cells in response to treatment with VAX014 was determined by flow cytometry. Each cell line was seeded in 6-well plates (5 × 10⁵ cells/well) in a total volume of 3 mL complete media and allowed to adhere for 4 hours. Cells were stimulated by the addition of VAX014 rBMCs at their respective EC₅₀ (MOI of 300 for MB49 and 75 for MBT-2) in 200 μL/well. After the addition of rBMCs, plates were centrifuged for 10 minutes at 671 × g. An untreated control condition was included to obtain a baseline reading of PD-L1 expression in each cell line. After 20 hours, cells were physically detached by scraping and washed once with wash buffer [PBS with 10% (v/v) HI-FBS]. Cell concentrations were adjusted to 10⁶ cells/mL in wash buffer, and aliquots of 10⁵ cells/100 μL were made. Each aliquot was treated with 2 μg/100 μL Fc block anti-mouse CD16/CD32 (clone 2.4G2; Supplementary Table S1) and incubated at 4°C for 20 minutes. Cell-surface PD-L1 was stained by the addition of 0.5 μg/100 μL PE-conjugated anti–PD-L1 mAb (clone 10F.9G2; Supplementary Table S1) and incubated for 30 minutes at 4°C. Labeled cells were then washed three times in wash buffer and resuspended in 1 mL wash buffer. PD-L1 expression was measured using an S1000 flow cytometer, and the median fluorescent intensity (MFI) of untreated control cells and VAX014-treated cells was obtained with CellCapTure 3.1. The MFI ratio between untreated and VAX014-treated cells was calculated using Microsoft Excel. The experiment was repeated once, and the mean MFI ratio from two experiments was plotted using GraphPad Prism 9.0. PD-L1 expression in MB49 cells after IFNγ stimulation was determined as described above, except that 50 ng/mL mouse recombinant IFNγ (Millipore Sigma) was added to untreated cells.

A titration of vector rBMCs (no invasin and no PFO) containing either wild-type LPS or immune-attenuated LPS (20 μL/well to make final rBMC concentrations ranging from 10 to 10⁵ rBMCs/mL in PBS) was added to an empty 96-well tissue culture plate in triplicates. Next, HEK-Blue hToll-like receptor (TLR)4 reporter cells in HEK-Blue Detection medium (InvivoGen) were added to the plate at 25,000 cells/well at 180 μL/well. PBS (20 μL) and HEK-Blue Detection medium (180 μL) were added to control wells. After a 16-hour coincubation period, absorbance at 655 nm was measured as a signal for the transcriptional activation of the NF-κB reporter construct (SEAP activity) using a SpectraMax M3 plate reader. The cell-based TLR activation assay was repeated a total of three times, and mean absorbance as a function of rBMC concentration was plotted using GraphPad Prism 9.0.

Orthotopic MB49 or MBT-2 tumors were implanted into the urinary bladders of female C57BL/6 mice or C3H/HeJ mice, respectively. Mice were first anesthetized by intraperitoneal (i.p.) injection of 100 mg/kg ketamine (Akorn Ophthalmics) and 16 mg/kg xylazine (Akorn Ophthalmics), followed by the transurethral catheterization of urinary bladders using a sterile 24-gauge flexible angiocatheter (Exel International; refs. 11, 20). Once catheterized, residual urine was removed by gently pressing the abdomen area with a finger through the catheter prior to tumor installation.

For o.t. MB49 bladder tumor installation, two separate tumor attachment sites were made in the bladder wall by electrocauterization using a 2-second pulse of 5W monopolar output from a Bovie electrocautery unit (Symmetry Surgical), which was placed in contact with a platinum guidewire inserted through the catheter lumen and touching the bladder wall. This was performed at two sites by repositioning the guidewire between pulses. Following the cauterization, 10⁵ MB49 cells (in 50 μL of complete DMEM) were slowly injected into the bladder through the catheter attached to an insulin syringe (Exel International), and syringes were kept in place in the catheter to prevent voiding and ensure tumor take (20, 21). After a 1-hour tumor dwell time, catheters were gently removed, and mice were allowed to recover prior to randomization into treatment groups as indicated in the figures and/or legends. Tumor take was confirmed by gentle digital palpation of voided bladders and/or the development of gross hematuria.

For o.t. MBT-2 bladder tumor installation, mice were anesthetized as indicated above, and voided bladders pretreated with 50 μL of 0.1 mg/mL poly L-lysine hydrobromide (Millipore Sigma) for 15 minutes via injection through a syringe attached to the urinary catheter (21). After the removal of poly-lysine via the syringe, 2 × 10⁶ MBT-2 cells were instilled into the bladder (in 50 μL PBS), and the attached insulin syringe was kept in place to prevent voiding and ensure tumor take. After a 1-hour tumor dwell time, catheters were removed, and mice were allowed to recover prior to randomization into the indicated treatment groups. Orthotopic tumor take was confirmed by a gentle digital palpation of voided bladders and/or the development of gross hematuria.

IVE treatments with VAX014 rBMCs, VAX-I rBMCs, or saline vehicle control were instilled in a 50-μL dose volume into the emptied bladders of mice. Following anesthetization with ketamine/xylazine, mice were catheterized as described above and a slow pulsed injection of the entire dose volume was administered into the bladder. Syringes were kept in place in the catheter for a total treatment dwell time of 1 hour to help minimize leakage and backflow and to prevent voiding. Upon completion of treatment, catheters were gently removed, and mice were allowed to void naturally and recover under observation. In all experiments where VAX014 or VAX-I was given by the IVE route, the previously optimized fixed dose of 10⁸ VAX014 rBMCs/dose was used (11). IVE treatments in the MB49 model were administered 1 day after tumor installation, whereas IVE treatments in the MBT-2 model were initiated 5 days after tumor installation. IVE treatments were provided weekly thereafter up to a total of 6 doses or until mice expired. The treatment group and schedule in each study are indicated in the figure or figure legends. In both models, survival was the primary experimental endpoint unless bladder weights were used as a more proximal surrogate endpoint for tumor burden. When bladder weights were assessed, mice that had received weekly IVE treatments with VAX014, VAX-I or saline were euthanized on day 25 prior to bladder removal.

Mice surviving beyond day 60 were considered to have achieved CR of the o.t. bladder tumor and were rechallenged with a second o.t. MB49 or MBT-2 bladder tumor (each installed as described above). For each model, a group of tumor-naïve mice were implanted with o.t. tumors to control for tumor take. Rechallenged mice were observed for clinical symptoms and development of gross hematuria. Survival was the primary endpoint for rechallenge studies and mice surviving beyond day 50 were considered to have rejected tumors.

Immune cell depletion efficacy experiments were conducted using the same single o.t. MB49 and MBT-2 models, and a weekly IVE treatment schedule was used as described immediately above, except that weekly IVE treatments with VAX014 rBMCs or saline control were started on day 3 in the MB49 model and day 6 in the MBT-2 model. Immune cell depletion was achieved by the i.p. administration of CD8α, CD4, or NK1.1 depletion antibodies or PBS as a nondepleted control condition (Supplementary Table S1), each at 100 μg/100 μL/dose on the days indicated in the figure legends. Survival was the primary endpoint for immune depletion studies, and mice surviving beyond day 60 were considered to have achieved CR of the o.t. bladder tumor.

Weekly IVE treatment with VAX014 in combination with systemic ICB against PD-L1 was evaluated using both the single o.t. MB49 and MBT-2 bladder tumor models. Each model was established as described above. For both models, weekly IVE treatments were initiated 6 days following o.t. tumor installation. Systemic PD-L1 treatment (clone 10F.9.G2) given as a 100 μL i.p. bolus (100 μg/dose) was initiated 1 day following the initial IVE treatment and administered twice weekly for up to 5 total doses on days indicated in the figures. Where indicated, volume-matched vehicle controls (saline for IVE treatments and PBS for systemic treatments) were administered on the same schedule by the corresponding route of administration. The primary endpoint for combination treatment studies was survival.

Mice from the MB49 combination treatment group surviving beyond day 70 were considered to have achieved CR of the o.t. bladder tumor and were rechallenged with a second o.t. MB49 bladder tumor (installed as described above). A group of tumor-naïve mice were implanted with o.t. MB49 tumors to control for tumor take. Rechallenged mice were observed for clinical symptoms and development of gross hematuria. Survival was the primary endpoint for rechallenge studies, and mice surviving beyond day 50 were considered to have rejected tumors.

To evaluate systemic antitumor effects against distal secondary tumors after IVE treatment with VAX014 ± PD-L1 blockade, an experimental variation of the MB49 model was used where mice were implanted with both an o.t. tumor along with a second i.d. MB49 tumor placed in the right flank on the same day. To prepare for i.d. MB49 tumor implantation, the fur on the right flank was shaved 1 to 3 days before tumor implantation using a small animal grooming trimmer (PepPet) while mice were anesthetized by inhalation of 3 to 5% (v/v) isoflurane (Phoenix Pharmaceuticals) mixed with pure oxygen (Airgas) at a flow rate of 0.8 to 1 L/minute. On the day of dual o.t./i.d. tumor implantation, each mouse was anesthetized by ketamine/xylazine mixture as indicated above, and the i.d. tumor installed first using 1.25×10⁵ MB49 cells (in 30 μL PBS) at the shaved site via an insulin syringe with a 29-G needle (Exel International). Immediately after the i.d. tumor was implanted, the o.t. MB49 bladder tumor was installed as described above. Weekly IVE treatments were initiated 3 days following o.t. tumor installation. Systemic PD-L1 treatments (clone 10F.9.G2) were given as a 100 μL i.p. bolus (100 μg/dose) twice weekly on days indicated in the figures. Volume-matched vehicle controls were administered on the same schedule by the corresponding route of administration. As an additional control to demonstrate the required presence of an o.t. MB49 tumor for generating systemic antitumor activity against the distal untreated MB49 i.d. tumor in response to IVE treatment with VAX014 (i.e., in situ immunization effect), a group of mice were implanted with only an i.d. MB49 tumor and then treated weekly via the IVE route with VAX014 or saline as described above. The primary endpoints for these combination treatment studies were distal i.d. MB49 tumor growth rates and survival. The growth of i.d. MB49 tumors was measured in millimeters every 1 to 3 days with a digital caliper (Thermo Scientific), and tumor volumes were calculated using the equation, π/6 × length × width × width (mm³). Mice displaying signs of morbidity or having i.d. tumor volume >1,500 mm³ were euthanized.

Mice surviving beyond day 80 were considered to have achieved CR of both the o.t. MB49 bladder tumor and the distal i.d. MB49 tumor. These long-term survivors were rechallenged with a second i.d. MB49 tumor (implanted as described above) and then observed for clinical symptoms and development of gross hematuria. Survival was the primary endpoint for rechallenge studies, and mice surviving beyond day 50 were considered to have rejected tumors.

Mice that rejected a rechallenge with an i.d. MB49 tumor and survived until day 188 from initial tumor installation were rechallenged a second time with an i.d. MB49 tumor (n = 8 total). A group of tumor-naïve mice (n = 4) were implanted with i.d. MB49 tumors to control for tumor take. To assess the role of immunologic memory in the CD4⁺ lymphocyte compartment, the eight rechallenged mice were randomized into two groups (n = 4/group), and one group was depleted of CD4⁺ T cells prior to and during rechallenge. The CD4 depletion antibody was given by i.p. bolus as described in depletion studies above. Growth rates of i.d. MB49 tumors were then measured 2 to 3 times a week until CR was achieved, or tumor size–based termination criteria were met. Assessments for the development of neutralizing idiotypic antibodies against the CD4 depletion antibody were performed as described below.

To evaluate systemic antitumor effects against distal pulmonary MB49 tumors after IVE treatment with VAX014 ± PD-L1 blockade, an experimental variation of the MB49 model was used where mice were implanted with both an o.t. tumor and mice seeded with pulmonary metastases on the same day. To establish pulmonary metastases, each mouse was immobilized in a Tailveiner Restrainer (Braintree Scientific), and tail veins were illuminated using a Veinlite vein finder (TransLite). A total of 5×10⁵ MB49 cells (100 μL in PBS) was injected into the tail vein using an insulin syringe with a 29-G needle. Mice were then anesthetized by ketamine/xylazine i.p. injection, and o.t. MB49 tumors were implanted as described above. Mice were allowed to recover prior to randomization into saline, VAX014, PD-L1 blockade, and combination (VAX014 and PD-L1 blockade) treatment groups. Administration of IVE and systemic treatment with PD-L1 blockade were conducted as described above, and treatment schedules are indicated in the figures and figure legends. Where indicated, volume-matched vehicle controls on the same schedule by the corresponding route of administration. Experimental endpoints include survival and lung nodule counts. Mice surviving beyond day 60 were considered to have achieved CR of both the o.t. MB49 bladder tumor and pulmonary MB49 tumors. After expiration or euthanization, bladders were surgically removed for bladder weight measurements, and lungs were excised for visual tumor nodule count. Digital images of the lungs were captured using an iPhone 7 (Apple) with the built-in camera software. Lung nodules were counted manually, and any lungs containing nodules that were too numerous to count (TNTC) were assigned a value of 100 nodules.

PD-L1 expression in CD45⁺ immune cells and CD45^– nonimmune cells of distal untreated i.d. MB49 tumors following weekly IVE treatment of o.t. MB49 bladder tumors with VAX014 was assessed by flow cytometry. Dual o.t./i.d. MB49 tumors were established as described immediately above. On day 11, following 2 weekly IVE treatments with VAX014 or saline, mice (n = 5 per treatment group) were euthanized, and distal i.d. MB49 tumors were surgically removed and processed for flow cytometry. Bladders were removed for bladder weight assessment as a surrogate efficacy readout for o.t. bladder tumor burden. Tumors were dissociated by incubating minced tumor tissues in Tumor Dissociation Buffer [RPMI-1640 with 10% (v/v) HI-FBS, 3 mg/mL collagenase IV (Thermo Fisher), and 0.2 mg/mL DNaseI (Roche); 1 mL/100 mg tumor] at 37°C for 20 minutes followed by neutralization by the addition of an equal volume of RPMI-1640 containing 10% (v/v) HI-FBS and 10 mmol/L EDTA (Thermo Scientific). Large debris and clumps were removed by passage through a 70-μm cell strainer (Thermo Fisher). Cell suspensions were then washed once with PBS with 10% (v/v) HI-FBS (wash buffer), and cell concentrations were adjusted to 10⁷ cells/mL to make aliquots of 10⁶ cells in 100 μL. Fc receptors were then blocked with 1 μg/100 μL of anti-CD16/32 (clone 93) for 30 minutes at 4°C, and cells were labeled with 0.6 μg/100 μL of PE-conjugated anti–PD-L1 (clone MIH5) and 0.5 μg/100 μL of FITC-conjugated anti-CD45 for 30 minutes at 4°C. Stained cells were then washed twice in wash buffer and resuspended in 1 mL for data collection and analysis using the S1000 Flow Cytometer and CellCapTure 3.1. Unlabeled cells were used as the control. Target cell populations were gated using the forward (FSC) and side scatter (SSC) plot followed by doublet exclusion using FSC LinH and FSC LinA scatter plot, and unlabeled cells were used as PD-L1 and CD45-negative controls to set gates. Scatter plots of CD45 expression in a function of PD-L1 expression were generated using CellCapTure 3.1. The population of PD-L1-expressing cells per gram of tumor was calculated using Excel and plotted using GraphPad Prism 9.0. MFI of PD-L1-expressing cells based on CD45 expression status was calculated using CellCapTure 3.1 and plotted using GraphPad Prism 9.0.

IFNγ expression in distal untreated i.d. MB49 tumors following the weekly IVE treatment of o.t. MB49 bladder tumors with VAX014 was assessed by sandwich ELISA against IFNγ using a Mouse IFNγ Uncoated ELISA Kit (Thermo Fisher, #88-7314-22). Dual o.t./i.d. MB49 tumors were established as described immediately above. On day 11, following 2 weekly IVE treatments with VAX014 or saline, mice (n = 6 per treatment group) were euthanized and distal i.d. MB49 tumors were surgically removed, weighed, and processed for total protein by immersion (50 mg tissue/mL) in T-PER Tissue Protein Extraction Reagent (Thermo Fisher) supplemented with 1% (v/v) Halt Protease Inhibitor Cocktail (Thermo Fisher). Samples were homogenized using a rotor/stator tissue homogenizer (BioSpec Products) at 25,000 rpm. Insoluble debris was pelleted by centrifugation at 10,000 × g for 5 minutes, and supernatants were transferred to clean microfuge tubes as total protein extracts. Protein concentrations in the extracts were determined by Bradford assay. Protein extracts were diluted to 1:40 by adding 10 μL of protein extracts to 390 μL of protein extract solution. A standard curve was generated using a titration of BSA in protein extract solution (4.9 to 5,000 μg/mL). Each sample (10 μL) was mixed with 200 μL Bradford Reagent (Millipore Sigma) and assayed in triplicate in a 96-well flat-bottom clear plate (Greinerbio-one). After 10 minutes of incubation at room temperature, absorbance at 595 nm was read using a SpectraMax M3 plate reader, and protein concentration in each sample was calculated using the built-in Bradford assay calculator of SoftMax Pro 5.4 using the best-fit line generated from absorbance values of the BSA standard curve. Tumor protein extracts were stored at −70°C until being assayed for IFNγ. Per the instructions of the ELISA kit, a standard curve of recombinant IFNγ (100 μL/well) ranging from 15.6 to 2,000 pg/mL was built. Protein samples were diluted into 1× ELISA/ELISPOT Diluent (provided in the kit) and added to wells (100 μL/well) at a concentration of 1 mg/mL. Only the 1× ELISA/ELISPOT Diluent was added to the blank control wells. All standards and samples were assayed in triplicate per the instruction of the kit. Absorbance at 450 nm was measured using the SpectraMax M3 plate reader. IFNγ concentration (pg/mg protein) in each sample was calculated using the built-in ELISA calculator in the SoftMax Pro 5.4.

All antibodies used for the flow cytometry analysis of TILs in i.d. tumors are listed under flow cytometry antibodies in Supplementary Table S1. A dedicated experiment in the dual o.t./i.d. MB49 was set up as described above to analyze lymphocyte populations in distal i.d. MB49 tumors by flow cytometry. Treatment groups included saline-treated controls, weekly IVE treatment with VAX014, biweekly treatment with systemic PD-L1 blockade or the combination of weekly IVE treatment with VAX014 with systemic PD-L1 blockade (n = 7 saline treatment group, n = 5 per other treatment group). Treatments were provided as described above. Mice were euthanized on day 11, and excised i.d. tumor samples were placed in vials containing the complete RPMI media. Bladders were also removed for bladder weight assessment as a surrogate efficacy readout for o.t. bladder tumor burden. Single-cell suspensions from tumor samples were generated using the gentleMACS Tumor Dissociation Kit (Miltenyi Biotec). Tumors were washed once each with DPBS (Thermo Fisher) and minced to ∼1 mm³ after necrotic tissue was removed. Minced tumors were then placed in a gentleMACS C Tube (Miltenyi Biotec) containing Digestion Enzyme mix (2.35 mL of RPMI-1640, 12.5 μL of enzyme A, 100 μL of enzyme D, and 50 μL of enzyme R; Miltenyi Biotec). Tumor segments were digested by using the gentleMACS Octo Dissociator with Heaters (Miltenyi Biotec) and running the dissociation program (37_C_m_TDK 1). After the completion of the program, cells were pelleted, and the supernatant was removed. Each pellet was resuspended in 10-mL FACS buffer [DPBS with 10% (v/v) HI-FBS and 40 mmol/L EDTA], and the cell suspension was filtered through a 70-μm cell strainer. Cells were washed twice in FACS buffer, and cell concentrations were adjusted to 2×10⁷ cells/mL to make aliquots of 2×10⁶ cells in 100 μL. Fc receptors were then blocked with 2 μg/100 μL of anti-CD16/32 (clone 93) at room temperature for 10 minutes, followed by labeling with 1 μL/100 μL of eFluor 780 Fixable Viability Dye (Thermo Fisher) and 0.2 μg/100 μL each of surface markers (anti-CD3, anti-CD4, anti-CD8, Brilliant Violet 785 conjugated anti-CD45, and anti–PD-1) at 4°C for 30 minutes. Labeled cells were washed twice in FACS buffer, followed by fixation and permeabilization for staining intracellular markers. Each aliquot of cells was incubated in 200 μL 1× Foxp3/Transcription Factor Fixation/Permeabilization solution (Thermo Scientific) at room temperature for 30 minutes before cells were washed twice in 1× Permeabilization Buffer (Thermo Scientific). Each sample was resuspended in 100 μL 1× Permeabilization Buffer containing 0.2 μg/100 μL each of antibodies against internal markers (anti-FoxP3, anti-EOMES, and anti–T-bet) and incubated at room temperature for 30 minutes. Cells were washed twice in FACS buffer and resuspended in 200 μL. Flow cytometry data were acquired using BD Fortessa (Beckton Dickinson) and analyzed with Kaluza software (Beckman Coulter). The live singlet CD45⁺ population was used as the primary gate for analyzing the immune compartment, and the live CD45⁺CD3⁺ population was used as the primary gate on the T-lymphocyte population, and gates were established using fluorescence minus one samples. Scatter plots depicting the complete gating strategy used for analysis of TILs from a tumor from each treatment group are provided in Supplemental Data File S1. Flow cytometry data analysis presented here is % CD45⁺ cells: CD3⁺ (CD45⁺CD3⁺), CD4⁺ T cells (CD45⁺CD3⁺CD4⁺), and CD8⁺ T cells (CD45⁺CD3⁺CD8⁺) in the live-cell population. The population ratio between conventional T cells (Tconv; CD45⁺CD3⁺CD4⁺FoxP3^–) and regulatory T cells (Treg; CD45⁺CD3⁺CD4⁺FoxP3⁺), ratio of T helper type 1 cells (Th1; CD45⁺CD3⁺CD4⁺FoxP3^–Tbet⁺) to Treg, and % activated CD8⁺ T cells (CD45⁺CD3⁺CD8⁺Eomes^–Tbet⁺PD-1⁺) in CD8⁺ T-cell population were determined. The ratio and percentage of population of analyzed immune cells were calculated using Excel, and results were plotted using GraphPad Prism 9.0.

The same general methodology was used to prepare all formalin-fixed paraffin-embedded (FFPE) samples for IHC. All antibody reagents, including isotype controls, are listed in Supplementary Table S1, and all staining controls for all IHC studies are provided in Supplementary Fig. S3–S5.

First, to better expose antigenic epitopes, slides were first subjected to a deparaffinization step, followed by rehydration and heat antigen retrieval steps. Deparaffinization was achieved by three successive 5-minute immersions in fresh HistoClear Clearing Agent (Avantor). Rehydration was performed using four sequential 5-minute immersions in 100%, 95%, 70%, and 50% (v/v) ethanol (Thermo Fisher), followed by two 5-minute immersions in nano-purified water (Thermo Fisher). Slides were then immersed in heat antigen retrieval buffer [10 mmol/L Tris-base (Avantor), 1 mmol/L EDTA (Avantor), 0.05% (v/v) Tween 20 (Millipore Sigma), pH 9.0] and heated for 20 minutes in a steamer (Oster) placed on an electrical cooktop (Thermo Fisher). After heating, slides were cooled by a slow exchange with running deionized water and washed by sequential 5-minute immersions in nano-purified water, wash buffer [PBS containing 0.02% (v/v) Tween 20 (Millipore Sigma)], and PBS. Slides were then blocked for 1 hour in PBS with 20% (v/v) normal goat or rabbit serum (Abcam; matched with the host species of the secondary detection antibody) at room temperature and rinsed for 5 minutes in PBS, after which primary detection antibody was added at the concentration indicated for each IHC experiment indicated below.

Following incubation with primary detection antibody for at least 1 hour at room temperature, slides were then drained, washed twice by 5-minute immersion in wash buffer, and endogenous peroxidase was quenched by a 10-minute immersion in 3% (v/v) hydrogen peroxide (Thermo Fisher). Slides were washed twice in wash buffer, followed by 1-hour incubation with goat anti-rabbit IgG HRP conjugated (2 μg/mL) or rabbit anti-goat IgG HRP conjugated (2 μg/mL; Supplementary Table S1). Slides were drained of the secondary detection reagent, washed twice in wash buffer, and washed once with PBS before the addition of the metal-enhanced 3,3′-diaminobenzidine (DAB) detection substrate (Thermo Fisher). After DAB signal developed, the DAB substrate was drained, and slides were washed twice in wash buffer and once with PBS prior to counterstaining with hematoxylin (Avantor) for 5 minutes. Slides were washed twice in tap water before dehydration by four sequential 5-minute immersions in 50%, 70%, 95%, and 100% (v/v) ethanol. Dehydrated slides were then subjected to three successive 5-minute immersions in fresh HistoClear Clearing Agent and mounted with HistoChoice mounting media (Avantor) prior to placing glass coverslips. Slides were then viewed under the Axio Scope A1 light microscope (Zeiss) at 5× or 20× magnification after the mounting media solidified. Digital images were captured using a microscope-mounted OMAX camera system with ToupView software (OMAX Microscope).

The staining intensity of DAB in digital images was quantified using ImageJ 1.53c software as previously described (22). Because ImageJ assigns higher numerical values to pixels with a lighter color, grayscale of each image was inverted before tumor-restricted mean DAB stain intensity was quantified (23). DAB stain intensity of each sample was then scored based on the ratio of the mean DAB intensity between each sample and its corresponding isotype control. A positive signal cutoff ratio of 1.32 was utilized (two standard deviations above the mean of all isotype control stains). Staining scores of negative, +, ++, or +++ were given when the DAB intensity ratio was ≤1.32, >1.32, ≥2.0, or ≥3.0, respectively.

Bladder samples for IHC staining were obtained from mice randomly selected for IHC analysis from the survival experiments of the single o.t. MB49 or MBT-2 model described above (n = 3/model; date of organ collection ranging froms day 5 to 26 across both models). After expiration due to tumor burden, tumor-bearing bladders were surgically removed and placed into vials containing 1× Formal-Fixx formalin solution (Thermo Fisher). Tumor specimens were processed into FFPE tumor tissue blocks, sectioned via microtome, and 5 μm-thick serial sections were affixed to glass slides for IHC staining (Pacific Pathology). IHC was carried out as described above. Each primary detection antibody against integrin subunits β1, α5, or α3 was a rabbit antibody, and each was used at a working concentration of 1.2 μg/mL. Detection was performed as described.

For the IHC staining of VAX014 rBMCs in o.t. MB49 bladder tumors, mice (n = 3/group) were implanted with a single o.t. tumor and 5 days later provided with a single IVE treatment with VAX014 or saline as a control. One hour following treatment, mice were euthanized, and bladders were removed and placed in 1× Formal-Fixx formalin solution to be processed for IHC. Slides were generated, and IHC was performed as described above. The primary detection reagent used was a Protein A purified rabbit polyclonal antibody raised against the rBMC vector at a working concentration of 5 μg/mL (11). Detection was performed as described.

To evaluate PD-L1 expression in situ in o.t. MB49 or MBT-2 bladder tumors from mice that had failed weekly IVE treatment with VAX014 or saline, tumor-containing bladders were surgically excised from randomly selected mice that had expired or were euthanized because of tumor burden (n = 3/treatment group/model, date of organ collection ranging froms day 20 to 23 across both models). Tumor-bearing bladders from saline-treated control mice from each model were used as a control. Bladders were removed and placed in 1× Formal-Fixx formalin solution to be processed for IHC. Slides were generated and IHC was performed as described above. The primary detection reagent used was a goat polyclonal antibody specific for mouse PD-L1 at a working concentration of 5 μg/mL. Detection was performed as described but utilized an HRP-conjugated rabbit anti-goat IgG secondary polyclonal antibody.

Human bladder carcinoma microarray serial section slides were purchased from US Biomax (Array No. BC12011b). Each slide contains 60 individual bladder carcinoma sections, 2 adjacent normal bladder samples, and 1 normal bladder specimen as a control. A detailed description of the specimens in the array is available at the manufacturer's website (www.biomax.us). Each microarray slide was stained against integrin subunits β1, α3, or α5 or with isotype antibody control. Because of the high degree of homology at the amino acid level, IHC against individual integrin subunits β1, α3, and α5 in mouse and human tumor samples was conducted using the same primary detection reagents, concentrations, and conditions. Preparation of arrays for IHC was performed as described. Each primary detection antibody was a rabbit antibody, and each was used at a working concentration of 1.2 μg/mL. Detection was performed as described.

CTL assays were performed using splenocytes stimulated with nonviable MB49 stimulator cells. To generate stimulator cells, MB49 cells were cultured in T-75 flasks as described until reaching 70% to 80% confluency and then treated with a cytostatic concentration of 50 μg/mL mitomycin C (MMC) (Millipore Sigma) for 1 hour. MMC-treated MB49 stimulator cells were washed twice with 20 mL PBS and then physically detached from the flask with a cell scraper. Stimulator cells were then pelleted, resuspended in 10 mL Expansion Media [complete RPMI-1640 media supplemented with 40 U/mL recombinant murine IL2 (Thermo Scientific), 50 μmol/L β-mercaptoethanol (Thermo Scientific), and 1% (v/v) Minimum Essential Medium Non-Essential Amino Acids Solution (Thermo Scientific)], and counted. Stimulator cell concentration was adjusted to 1.5×10⁵ cells/mL in Expansion Media prior to admixing with splenocytes.

The spleens of mice used for CTL analysis were retrieved from mice of all treatment groups (n = 5 per treatment group) from experiments evaluating lymphocytes in i.d. MB49 tumors from the o.t./i.d. dual tumor model. Spleens were harvested from the euthanized mice and placed in vials containing complete RPMI media. Single-splenocyte suspensions were produced by grinding spleens on a 70-μm cell strainer placed above a 50-mL conical tube using a 10-mL syringe plunger (Exel International). The strainer was rinsed with 20-mL complete RPMI media to collect splenocytes into the conical tube. Splenocytes were pelleted, and the supernatant was replaced with 1 mL RBC lysis buffer [135 mmol/L ammonium chloride (Millipore Sigma) and 15 mmol/L Tris-HCl (Avantor)] and incubated for 10 minutes at room temperature. The lysis reaction was stopped by the addition of 111 μL of 10× PBS. Splenocytes were then washed once with complete RPMI media, and the RBC lysis step was repeated. Splenocyte concentration was determined and adjusted to 2.3 ×10⁶ cells/mL in complete RPMI-1640 media. Splenocytes were then expanded for 5 days before use as effector cells in the presence of stimulator MB49 cells at a ratio of 15:1 by mixing 1 mL stimulator cells with 100 μL splenocytes prepared above in 24-well tissue culture plates (Corning).

To set up the CTL assay, target MB49 or nonspecific target B16F10 cells were grown to ∼70% confluency in T-75 flasks as indicated above. In some experiments, 50 ng/mL, IFNγ was added to the flasks with MB49 cells one day before use to upregulate PD-L1 for experimental purposes. Target cells were removed from the flasks by scraping after being washed twice with PBS and resuspended in CTL media [RPMI-1640 media with 5% (v/v) HI-FBS and 100 U/mL P/S] at 2 × 10⁵ cell/mL. Target cells were then added to 96-well tissue culture plates (10,000 cells/well in 50 μL). Expanded splenocytes were washed once in CTL media, counted, and added to test wells containing MB49 or B16F10 target cells at effector-to-target ratios (E:T) of 50:1, 25:1, and 10:1 in duplicate with a total volume of 100 μL in each well. An equivalent number of splenocytes were also added in duplicate to the wells containing only the CTL media to serve as effector spontaneous controls. Volume-matched complete RPMI-1640 media were also added in duplicate to the wells containing the target cells to serve as target spontaneous controls. The plate was incubated for 4 hours. After 3 hours and 15 minutes of incubation, 10 μL of lysis buffer provided in the kit was added to target maximum and media and lysis buffer control wells. After incubation, cells were pelleted by centrifugation at 200 × g for 3 minutes, and 50 μL/well of undiluted supernatant was transferred to corresponding wells of a new 96-well plate. To assay for lactate dehydrogenase (LDH) in supernatants as a measure of target cell lysis, 50 μL/well of CytoTox 96 Reagent was added to each test well and allowed to incubate for 30 minutes before reactions were stopped by the addition of 50 μL/well of Stop Solution. LDH activity was measured by absorbance at 492 nm using SpectraMax M3 plate reader, and CTL activity as a percentage of target cell lysis was calculated according to the kit manufacturer's instructions. CTL activity as a function of E:T ratio was plotted using GraphPad Prism 9.0.

To assess the potential development of neutralizing idiotypic antibodies against the rat monoclonal mouse CD4 depletion antibody utilized for CD4⁺ lymphocyte depletion in MB49 tumor rechallenge experiments, blood was collected via cheek bleed into BD Microtainer Capillary Blood Collector tubes with dipotassium EDTA (Becton Dickinson) from naïve mice (n = 2), nondepleted mice (n = 4), and CD4 depleted mice (n = 4) 18 days after initial CD4 depletion antibody (rat monoclonal antibody) administration. Plasma was separated from blood cells by centrifugation at 3,500 × g for 90 seconds, transferred to clean microfuge tubes, and stored at −70°C until the ELISA was conducted to determine antibody titer against the CD4 depletion antibody. A high binding 96-well ELISA plate (Greinerbio-one) was coated with the CD4 depletion antibody (2 μg/mL, 100 μL/well) in coating buffer [3 mg/mL sodium carbonate (Millipore Sigma) and 6 mg/mL sodium bicarbonate (Millipore Sigma), pH 9.5] at 4°C overnight. The plate was washed twice in wash buffer [PBS with 0.05% (v/v) Tween 20 (Millipore Sigma)] and blocked with 200 μL/well blocking buffer (PBS with 10 mg/mL BSA) at room temperature for 2 hours. Plasma samples (10 μL) were diluted in blocking buffer (990 μL) and incubated for 1 hour at room temperature, followed by serial dilution in blocking buffer. The blocking buffer was removed, and the serially diluted plasma sample from each mouse (dilution factors ranging from 100 to 1,562,500) was added to the plate in duplicate and incubated at room temperature for 1 hour. After the incubation, the plate was washed three times with wash buffer. Antibodies against CD4 depletion antibodies were then detected by incubation with an HRP-conjugated goat anti-mouse IgG polyclonal antibody (0.2 μg/mL in blocking buffer, 100 μL/well) at room temperature for 1 hour. After the incubation, the plate was washed four times and 100 μL/well TMB Single Solution (Thermo Fisher) was added to each test well and incubated at room temperature for 10 minutes. Reactions were stopped by the addition of 100 μL/well 1N sulfuric acid (Avantor). Absorbance at 450 nm was read with SpectraMax M3 plate reader. The mean absorbance of each dilution of each sample was calculated using Excel. Antibody titer against the CD4 depletion antibody of each sample was determined as the highest dilution factor of the mean absorbance value that is more than 1.5-fold above the mean of the no plasma control wells. ELISA was repeated once. Combined data of two ELISA assays were plotted using GraphPad Prism 9.0.

GraphPad Prism 9.0 was used to plot and analyze all data. A two-tailed Student t test or Mann–Whitney U test was used for statistical analysis between two groups. Kaplan–Meier survival curves were analyzed for the statistical significance of median survival using the log-rank test. All statistical comparisons were made against saline or untreated controls unless indicated otherwise. Statistical significance was defined as a P ≤ 0.05. Where applicable, a power analysis was conducted with α = 0.05 and 85% power for individual in vivo studies using the Sample Size Calculator 1.041 (University of Vienna) or GraphPad StatMate 2.0 software (GraphPad Software).

All data associated with this study are available in the main text or the supplementary materials.

The research prototype to VAX014, termed VAX-IP, was previously shown to have modest but significant antitumor activity in the o.t. MB49 bladder tumor model when given twice weekly via the IVE route (11). Here, we reassessed the activity of the clinical formulation of VAX014 using the weekly clinical dosing schedule in both the MB49 and MBT-2 syngeneic o.t. mouse models of bladder cancer (Fig. 1A and B, treatment regimen in the legend). Weekly IVE administration of VAX014 led to a significant durable survival advantage in both models. No antitumor activity was observed in the o.t. MB49 model following weekly IVE administration of VAX-I rBMCs (control rBMCs containing invasin but lacking the oncolytic PFO toxin; Supplementary Fig. S6).

VAX014 has dual specificity for α3β1 and α5β1 integrins, which are conserved in amino acid sequence, structure, and function among mammals. To our knowledge, characterization of the expression patterns of α3β1 and α5β1 integrin subtypes has not been reported for either MB49 or MBT-2 o.t. bladder tumors. Therefore, we evaluated in situ expression patterns of the individual β1, α3, and α5 integrin subunits in MB49 and MBT-2 bladder tumor specimens and found unpolarized, uniform expression of at least one integrin subtype in o.t. bladder tumors from each model (Fig. 1C). Expression in healthy bladder tissue was confined to the basal urothelium, as has been previously reported (24). We next investigated if VAX014 could be detected within bladder tumors following IVE treatment. To address this, mice bearing o.t. MB49 bladder tumors were treated with a single IVE dose of VAX014 prior to evaluation of tumors for the presence of VAX014 by IHC (Fig. 1D). VAX014 was readily detectable in situ in the periphery of exophytic o.t. MB49 bladder tumors but not healthy urothelium.

VAX014 is an oncolytic agent, and based on the known immunotherapeutic effects of other locally administered OV therapies, we hypothesized VAX014 would elicit antitumor cellular immune responses in vivo (8, 9). As an initial test for protective antitumor responses following IVE treatment with VAX014 in both the MB49 and MBT-2 models, we rechallenged long-term survivors with a second round of o.t. bladder tumors (Fig. 2A and B). Rechallenge experiments demonstrated high o.t. tumor rejection rates as indicated by long-term survival in comparison with the survival of tumor-naïve mice implanted with either o.t. MB49 or MBT-2 bladder tumors as a control.

Based on these results, we reasoned that VAX014 likely mediated a lymphocytic response against the treated tumor, a generally accepted pharmacologic hallmark of most OV treatment modalities (8). To explore the role of different lymphocyte subsets in the antitumor activity of VAX014, efficacy experiments were conducted in mice depleted of either CD8⁺ T cells or CD4⁺ T cells in both the MB49 and the MBT-2 o.t. bladder tumor models (Fig. 2C and D). To ensure enough time to deplete lymphocyte subsets prior to treatment while ensuring immune cell depletion did not affect o.t. tumor take, treatment initiation was delayed until day 3 in the MB49 model (as opposed to day 1, as performed in Fig. 1C). Consistent with our previous report, delaying treatment from day 1 to day 3 resulted in decreased survival in this aggressive MB49 model (11). Depletion of either CD8⁺ T cells or CD4⁺ T cells resulted in a complete loss of antitumor activity in both models. Depletion of CD4⁺ or CD8⁺ T cells alone led to an observable trend in aggressiveness of the MB49 model, suggesting an antitumor lymphocytic response is primed by the tumor alone but is insufficient to meaningfully control the disease. This result mirrors that seen in previous reports, where lymphocyte depletion in the MB49 model in the context of treatment with BCG leads to more aggressive disease (25). NK cell depletion in the MB49 model also resulted in a noticeable loss of activity, although the effect was not significant (Supplementary Fig. S5).

Previous reports demonstrate the upregulation of PD-L1 in response to the local OV treatment of i.d. MB49 tumors and other syngeneic mouse models (26, 27). Therefore, we conducted a series of studies aimed at evaluating PD-L1 expression in response to VAX014 in vitro (Fig. 3A–C). This examination revealed MB49 cells upregulated PD-L1 nearly 15-fold in response to treatment with VAX014, whereas MBT-2 exhibited constitutive high expression of PD-L1, with no appreciable upregulation following treatment. Building on this, we then assessed the expression of PD-L1 in situ in o.t. MB49 or MBT-2 bladder tumors isolated from mice that had failed IVE treatment with VAX014 monotherapy compared with saline-treated tumors (Fig. 3D–F). Consistent with in vitro findings, saline-treated MBT-2 bladder tumors constitutively expressed high PD-L1, with no apparent upregulation after treatment with VAX014 (Fig. 3D), whereas PD-L1 expression was relatively low in saline-treated MB49 bladder tumors but was increased in tumors from mice failing VAX014 treatment (Fig. 3F).

Disparate upregulation of PD-L1 in response to VAX014 in the MB49 and MBT-2 models provided an opportunity to evaluate and compare therapeutic synergy between VAX014 and systemic anti–PD-L1 in a tumor model where PD-L1 has constitutively high expression in tumors (MBT-2) versus a tumor model where PD-L1 expression is minimal (MB49; Fig. 3E–G). As a monotherapy, anti–PD-L1 was so effective in the MBT-2 model that therapeutic synergy with VAX014 could not be reasonably determined (Fig. 3E). On the other hand, in the MB49 model, systemic anti–PD-L1 monotherapy was less effective, but when combined with IVE treatment with VAX014, it resulted in profound therapeutic cooperation that increased the frequency of complete tumor clearance and overall survival rates in comparison with either agent alone (Fig. 3G). When rechallenged with o.t. MB49 bladder tumors, all long-term survivors from the combination treatment arm were capable of complete tumor rejection despite receiving no further treatment (Supplementary Fig. S8).

We next explored whether the IVE treatment of an o.t. bladder tumor with VAX014 could lead to systemic antitumor immune responses capable of acting upon distal extravesical tumors and, if so, could that effect be enhanced by the addition of systemic anti–PD-L1. We first evaluated this using an experimental dual tumor variation of the MB49 model where an o.t. tumor is implanted in the bladder along with placement of a second i.d. MB49 tumor in the flank of the same mouse (dual o.t./i.d. model; Fig. 4). Using the dual o.t./i.d. tumor model, we discovered PD-L1 was upregulated in CD45^– nonimmune cells of distal i.d. tumors following IVE treatment with VAX014 (Fig. 4A–C). These distal untreated tumors were also found to have elevated IFNγ (Fig. 4D), a known mediator of PD-L1 upregulation whose receptor has recently been identified as critical in response to IVE treatment with BCG in the MB49 model (25). We then used the o.t./i.d. dual tumor model to evaluate the treatment effect of combination therapy with IVE VAX014 and systemic anti–PD-L1 in comparison with either agent alone (Fig. 4E–H). As monotherapies, each agent resulted in a modest but significant systemic treatment effect as measured by individual and mean growth rates of the distal i.d. MB49 tumors (Fig. 4F and G). The IVE administration of VAX014 in combination with systemic anti–PD-L1 led to complete clearance of distal i.d. and o.t. MB49 tumors in 100.0% and 90.0% of mice, respectively, and provided a clear survival advantage in comparison with either agent alone (Fig. 4F and H). The observed response was durable to at least 88 days (longest timepoint evaluated), after which 88.9% of long-term survivors were capable of rejecting rechallenge with a single i.d. MB49 tumor (Fig. 4I). Response of the distal i.d. MB49 tumor was dependent on the presence and treatment of o.t. MB49 bladder tumor with VAX014, supporting that VAX014 facilitated in situ immunization against the treated tumor (Supplementary Fig. S9). CTL activity against MB49 target cells was significantly higher with combination treatment (Fig. 4J), was specific for the treated tumor type, and could be dampened by IFNγ-mediated upregulation of PD-L1 by target cells, providing further evidence that this immune-checkpoint molecule contributes to immune escape in the MB49 model (Supplementary Fig. S10).

Next, we used multicolor flow cytometry to evaluate changes in leukocyte populations and lymphocyte phenotype(s) in distal i.d. MB49 tumors actively responding to treatment in the dual o.t./i.d. MB49 model (Fig. 5). Evaluation of o.t. bladder tumors was not feasible due to a paucity of o.t. tumor tissue at the time of analysis (Supplementary Fig. S11). Distal i.d. MB49 tumors actively responding to combination treatment demonstrated no change in the total percentage of CD45⁺ leukocytes or CD3⁺ lymphocytes across treatment groups, along with no change in the relative proportions of either CD4⁺ or CD8⁺ T cells (Fig. 5A–D). However, there were significant changes among the phenotypes of both CD4⁺ and CD8⁺ T cells. Tumors from mice treated with a combination of VAX014 and anti–PD-L1 exhibited a significant increase in the ratios of CD4⁺FoxP3^– conventional T cells (Tconv) to CD4⁺FoxP3⁺ Tregs (Fig. 5E) and in the ratio of CD4⁺Foxp3^–Tbet⁺ Type 1 helper T cells (Th1) to Tregs (Fig. 5F). A significant increase in the proportion of activated CD8⁺ T cells (CD8⁺Eomes^–Tbet⁺PD-1⁺) among the total CD8⁺ T-cell population was also observed in the combination group (Fig. 5G).

A recent report indicates that CD4⁺ T cells are critical for protective memory responses in the MB49 model following IVE treatment with BCG (25). To preliminarily evaluate the role of CD4⁺ T cells in the memory response to rechallenge in the MB49 model after combination treatment with VAX014 and anti–PD-L1, mice that had been previously rechallenged with i.d. MB49 after achieving complete clearance of dual MB49 tumors following combination therapy were rechallenged for a second time with a single i.d. MB49 tumor but were depleted of CD4⁺ T cells before and during rechallenge (Fig. 5H). All nondepleted mice completely rejected tumor rechallenge with no observed tumor outgrowth. In the CD4⁺ depleted group, there was an initial palpable i.d. tumor growth phase in all tumors that was not observed in the nondepleted group, yet not all tumors continued to grow. Ultimately, mice depleted of CD4⁺ T cells demonstrated a complete rejection rate of 50.0%, implying the development of protective antitumor immunologic memory in CD4⁺ T cells for some mice in this model. The eventual rejection and slower tumor growth rates of tumors in CD4⁺ T cell–depleted mice may be due to compensatory CD8⁺ T-cell memory but was not due to the development of neutralizing antibody titers to the rat IgG anti-mouse CD4 depletion antibody (Supplementary Fig. S12).

Urothelial carcinomas of the urinary bladder have a propensity to metastasize to distal organs, including the lung (28). Therefore, we examined the ability of combination therapy to control distal pulmonary tumors in an experimental variation of the MB49 model, where an o.t. bladder tumor was implanted in mice along with an i.v. bolus of MB49 cells, the latter resulting in extensive pulmonary metastases in all mice within 10 days (Fig. 6). Despite the more aggressive nature of the model in comparison to the dual o.t./i.d. model used above (Supplementary Fig. S13), weekly IVE treatment with VAX014 alone led to improved survival (Fig. 6B) and significant control of pulmonary metastases as determined by lung nodule count (Fig. 6C and D), and these outcomes were further enhanced in the VAX014/anti–PD-L1 combination treatment group. Systemic monotherapy with anti–PD-L1 led to a subtle yet insignificant improvement in survival, with no improvement in lung nodule count compared with saline-treated controls. Individual lung nodule data from all mice appear in Supplementary Fig. S14.

To further strengthen the rationale for the clinical combination of VAX014 with systemic PD-1/PD-L1 blockade in patients with advanced bladder cancer, we reasoned it would be important to demonstrate tumor restricted expression of VAX014 target integrins α3β1 and α5β1 in stage T1–T3 bladder tumor specimens. Therefore, we evaluated the expression of each individual integrin subunit by IHC using a commercially available bladder tumor microarray (Table 1, with individual scored IHC images appearing in Supplementary Data File S2). This analysis revealed that one or both integrin subtypes were selectively expressed in over 88.3% of all stage T1–T3 tumor specimens evaluated.

Recent marketing approvals for nivolumab, pembrolizumab, and avelumab in various advanced bladder cancer indications highlight the promise of PD-1/PD-L1 blockade therapy in this diverse patient population (6). Clinical data supporting early approval of each of these agents in advanced bladder cancer indications demonstrate remarkable response durability, yet only in a minority of patients (29). The combination of local intralesional administration of immunotherapy with systemic ICB is an emerging clinical approach that appears thus far to result in a promising increase in durable response rates, with no additive toxicity over the respective systemic single-agent ICB (7). To our knowledge, limited clinical investigation of local immunotherapy combined with systemic ICB has occurred or been reported in advanced bladder cancer patients. It stands to reason that based on the success of this approach in other solid tumor types, improved outcomes in advanced bladder cancer patients might be expected.

Here, we used two different immune-competent o.t. bladder cancer models to explore the immunotherapeutic effects and translational potential of VAX014, a novel bacterial minicell-based targeted oncolytic agent. In both models, weekly IVE administration of VAX014 led to CD4⁺ and CD8⁺ T cell–dependent clearance of o.t. tumors and provided protective antitumor immunologic memory. The antitumor activity of VAX014 was found to be PFO-dependent, which helps in ruling out the possibility that the immune-mediated antitumor effects observed after IVE treatment of o.t. bladder tumors with VAX014 were mediated simply by the rBMC vehicle itself. This was important to understand because although VAX014 rBMCs are engineered to contain immune-attenuated penta-acylated LPS, which limits its ability to activate human TLR4, there are several lines of evidence to suggest this modified form of LPS retains full TLR4 activation in the mouse (30–32). Data presented here demonstrated VAX014 still had appreciable activity in the o.t. MBT-2 model, despite the loss of LPS-mediated TLR4 activation in the C3H/HeJ (Tlr4^Lps-d) mouse, whereas control VAX-I rBMCs lacking the oncolytic PFO protein showed no antitumor activity against o.t. MB49 tumors in the wild-type TLR4 C57BL/6 mouse. Although not conclusive, these data together provide orthogonal evidence that the immunotherapeutic activity of VAX014 is dependent on PFO-mediated oncolysis and not on the activation of TLR4 by the LPS component of the rBMC vector.

Although single-agent VAX014 led to immune-mediated clearance of o.t. bladder tumors after weekly IVE administration, not all achieved a CR, which was suggestive of established or acquired immunosuppression in o.t. tumors. Upon exploration, we found disparate expression of PD-L1 between models, and a lack of expression in the MB49 model correlated with limited response to anti–PD-L1, which is consistent with the growing clinical theme where a lack of PD-L1 expression is associated with poor response to PD-1/PD-L1 blockade (33–35). PD-L1 was identified as an acquired immune-resistance mechanism in response to IVE treatment with VAX014 in the MB49 model, and IVE treatment of bladder tumors with VAX014 in combination with systemically administered anti–PD-L1 demonstrated therapeutic synergy and enhanced clearance of bladder tumors as well as distal untreated extravesical MB49 tumors at two different anatomic sites, including the lungs. Systemic responses to IVE treatment were completely dependent upon the presence of an o.t. MB49 tumor in the bladder, providing compelling evidence that VAX014 facilitated in situ immunization. The immunologic findings of this study share several similarities with those of IVE treatment with BCG in the o.t. MB49 model, including an absolute requirement for CD4⁺ T cells, generation of protective CD4⁺ T-cell immunologic memory, and presence of elevated IFNγ in responding tumors (25). A previous report investigating the combination of avelumab (anti–PD-L1) and BCG in the o.t. variation of this model demonstrated minimal therapeutic synergy above avelumab monotherapy (36). However, systemic effects of this combination against distal tumors were not evaluated and, although outside of the scope of this work, warrant additional investigation.

Immune analysis of accessible distal i.d. MB49 tumors actively responding to treatment led to novel results in the dual o.t./i.d. MB49 model used to evaluate systemic effects following IVE treatment with VAX014. First, PD-L1 was upregulated in nonimmune cells of distal i.d. tumors in response to IVE treatment of o.t. bladder tumors with VAX014. The mediator(s) of PD-L1 upregulation following IVE treatment with VAX014 are presently unknown, but the concomitant increase of IFNγ observed in distal tumors may mediate this effect following IVE treatment of o.t. bladder tumors with VAX014. Although there were no changes in total leukocytes or TILs in any treatment group, there was a pronounced shift in the proportion of Th1-polarized CD4⁺ T cells in relation to Tregs, accompanied by a significantly elevated proportion of activated CD8⁺ T cells within tumors actively responding to combination therapy. This observation is aligned to a degree with clinical findings, where a gene signature reflective of CD4⁺ T-cell effector response in proportion to that of a Treg signature and independent of the presence of CD8⁺ T cells was associated with response to anti–PD-L1 therapy in metastatic bladder cancer tumors (37). An increased proportion of Th1-polarized CD4⁺ T cells in relation to Tregs was also observed in tumors from mice treated with VAX014 monotherapy, but unlike the combination group, there was no increase in the proportion of activated CD8⁺ T cells. Although local activation of antitumor lymphocytic response, followed by lymphocyte migration to distant tumor sites, cannot be ruled out, the findings of this study are more suggestive of a phenotypic switch to Th1 polarity in the CD4⁺ T-cell population within tumors in response to IVE administration of VAX014 and systemic PD-L1 blockade. The elevated proportion of Th1-polarized CD4⁺ T cells was associated with reduced Tregs, suggesting that tumor-localized Tregs had differentiated into Th1 lineage cells in situ in response to signals received following treatment with VAX014. One possibility behind this is that IVE treatment of bladder tumors with VAX014 stimulates production of Type-I IFN, a known mediator of Treg to Th1 differentiation that can also promote PD-L1 upregulation (38, 39). The role of Type-I IFN and other potential adaptive and innate immune mechansims will be explored in future work. One would expect a shift to Th1 polarity in the CD4⁺ T-cell population to influence local activation and better support CD8⁺ T-cell function, particularly when systemic ICB is administered to counteract tumor-orchestrated checkpoint-mediated immunosuppression. Consistent with this supposition, we found VAX014 monotherapy induced systemic tumor-specific CTL activity that was enhanced in the combination treatment setting, providing evidence that combination therapy leads to better effector responses. Minimal tumor-specific CTL activity was observed in saline or PD-L1 blockade monotherapy controls, suggesting T-cell activation is in response to VAX014.

The expression patterns of VAX014 target integrin subtypes, α3β1 and α5β1, in both the MB49 and MBT-2 model, as well as in tumor specimens from patients with advanced bladder cancer, were also evaluated as part of this work as a potential indicator and proxy for clinical translation of the preclinical pharmacology findings summarized above. At least one of these two integrin subtypes was selectively expressed in situ in o.t. bladder tumors from each preclinical model, and VAX014 was demonstrated to associate with o.t. bladder tumors but not normal urothelium in vivo. Finally, similar integrin expression patterns were observed in locally advanced bladder tumor specimens (stages T1–T3). No fewer than 80.0% of specimens expressed at least one integrin subtype. The commercially available microarrays used in this study are limited in sample size and lack supporting clinical histories, yet the data are consistent with previous reports and do suggest a significant number of T1–T3 bladder tumors express one or more VAX014 target integrins (16–18). Exploration of integrin expression in distinct subpopulations with known clinical histories should be the focus of future work. Because integrins are conserved in amino acid sequence between species, we were able to utilize the same conditions and primary detection reagents for IHC for both mouse and human specimens. The high degree of conservation in integrin amino acid sequence, structure, function, and apparent parallel role in bladder tumor pathobiology between species supports higher confidence in the interpretation of the observed results with respect to the potential for clinical translation.

In summary, the results of this study provide evidence supporting VAX014 as an effective complementary treatment to systemic PD-L1 blockade for advanced bladder cancer patients with tumors that do not express high PD-L1 in primary biopsies. At the same time, it is prudent to note that immune-competent mouse models do not always translate clinically (40). Nonetheless, the encouraging evidence presented here suggests this combination may be useful not only as a bladder-sparing treatment of locally advanced treatment-refractory disease but may also potentially provide systemic protection against disseminated metastatic disease, both of which are currently major therapeutic and clinical hurdles for PD-1/PD-L1–directed checkpoint blockade in the treatment of patients with advanced bladder cancer. Given that VAX014 is already undergoing a phase I clinical trial, there are several attractive and potentially bladder-sparing therapeutic intervention points for VAX014 in combination with ICB that can be evaluated in the near term. Taken together, these data provide a rationale for clinical investigation of this combination in bladder cancer patients and perhaps in other solid tumor indications that may benefit from this emerging therapeutic strategy.

S. Tsuji reports grants from NCI during the conduct of the study; personal fees from Vaxiion Therapeutics outside the submitted work. K. Reil reports grants from NCI during the conduct of the study; personal fees from Vaxiion Therapeutics outside the submitted work. K. Nelson reports grants from NCI during the conduct of the study; personal fees from Vaxiion Therapeutics outside the submitted work. V.H. Proclivo reports grants from NCI during the conduct of the study; personal fees from Vaxiion Therapeutics outside the submitted work. K.L. McGuire reports grants from NCI during the conduct of the study; other support from Vaxiion Therapeutics outside the submitted work. M.J. Giacalone reports grants from NCI during the conduct of the study; personal fees from Vaxiion Therapeutics outside the submitted work; in addition, M.J. Giacalone has a patent for US 9,267,108 issued, a patent for US 9,566,321 issued, a patent for US 10,039,817 issued, a patent for US 10,124,024 issued, a patent for US 10,561,719 issued, a patent for US 11,167,008 issued, and a patent for US 11,219,679 issued.

S. Tsuji: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. K. Reil: Data curation, investigation, methodology, writing–original draft, writing–review and editing. K. Nelson: Data curation, investigation, methodology. V.H. Proclivo: Data curation, visualization, methodology, writing–review and editing. K.L. McGuire: Conceptualization, resources, formal analysis, supervision, methodology, writing–original draft, project administration, writing–review and editing. M.J. Giacalone: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, methodology, writing–original draft, project administration, writing–review and editing.

Thank you to Seth P. Lerner MD, FACS at Baylor College of Medicine, Michael O'Donnell MD, FACS, at the University of Iowa Department of Urology, Neal D. Shore MD, FACS, at the Carolina Urologic Research Center, Steven N. Fiering Ph.D. at Dartmouth University Geisel School of Medicine, Trevor Hallam Ph.D. at Sutro Biopharma, and Stanley Maloy, Ph.D. at San Diego State University for their thoughtful review of the manuscript. This work was funded in part by NCI Phase I SBIR Grant No. 1R43CA180403-01.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.