This was an open-labeled randomized, dual-arm, phase Ib study assessing the safety and preliminary efficacy of sequential administration of atezolizumab and sipuleucel-T in subjects with asymptomatic or minimally symptomatic mCRPC. Subjects were randomly assigned in a 1:1 ratio to receive either atezolizumab followed by sipuleucel-T (Arm 1) or sipuleucel-T followed by atezolizumab (Arm 2). The primary endpoint was safety and tolerability of the combination treatment. Preliminary efficacy was a secondary objective. Correlative studies investigating changes in immune activation in response to treatment including characterization of molecular changes, immunogenicity profile, immune responses, and cytokine profile were also objectives in this trial.

Men with confirmed metastatic adenocarcinoma of the prostate and castrate levels of testosterone with sequentially rising prostate specific antigen (PSA) levels were eligible for this study provided they met the definition of asymptomatic or minimally symptomatic (ie, no use of opiate analgesia for cancer-related pain as identified on a patient reported brief pain inventory). Eligible subjects had to meet the following criteria: serum creatinine <1.5× the upper limit of normal (ULN), total bilirubin <1.5× ULN and serum aspartate aminotransferase <2.5× ULN, white blood cells ≥2500/μL, an absolute neutrophil of ≥1500, and platelets >100,000). Subjects were excluded if they had received prior immune therapy or were simultaneously undergoing treatment with any chemotherapy, radiotherapy, immunotherapy, or hormonal therapy besides medical or surgical castration. Subjects with any history of autoimmune diseases, clinically significant cardiac or pulmonary disease, uncontrolled infection, diseases of the central nervous system, active secondary malignancy, HIV, or hepatitis were excluded. Subjects currently receiving chronic corticosteroids, whose dose cannot be decreased to 10 mg or less of prednisone equivalent, were excluded from this study. All subjects signed an Institutional Review Board (IRB) approved informed consent.

Treatment in both arms consisted of an induction phase lasting 12 weeks followed by a maintenance phase beginning at week 13 and continuing until subjects experienced disease progression, unacceptable toxicity, or loss of clinical benefit. Subjects in Arm 1 received treatment in the induction phase as follows: atezolizumab 1200 mg by intravenous infusion on weeks 1 and 4 followed by sipuleucel-T administered intravenously on weeks 6, 8, and 10. Subjects in Arm 2 received induction treatment as follows: sipuleucel-T administered intravenously on weeks 1, 3, and 5 followed by atezolizumab 1200 mg intravenously on weeks 7 and 10. Subjects with an objective response, defined as either a complete response (CR) or partial response (PR), or stable disease (SD) at the end of week 12 could receive atezolizumab 1200 mg intravenously every 3 weeks until disease progression or a loss of clinical benefit occurred (maintenance phase). Subjects were categorized into one of three Halabi risk groups (Low, Group 1; Intermediate, Group 2; High, Group 3) based on the following variables: Eastern Cooperative Oncology Group (ECOG) performance status (0, 1, or 2), baseline PSA, lactate dehydrogenase, alkaline phosphatase, albumin, hemoglobin, presence of bone metastases, and presence of lymph node metastasis.31 Once categorized into one of the three risk groups, subjects were stratified and randomized using a block design. Subjects who discontinued or withdrew from the induction phase of the study were replaced.

Samples for hematology, serum chemistries, coagulation, and urinalysis were obtained prior to each infusion. Furthermore, these blood-based analyses along with serum PSA were evaluated at week 12 and subsequently every 12 weeks thereafter until disease progression.

Sipuleucel-T infusions were prepared from peripheral blood mononuclear cells (PBMCs) as previously reported.8 The dose level of atezolizumab in this study was 1200 mg (equivalent to an average body weight−based dose of 15 mg/kg) administered by intravenous infusion over 30 (±10) min every 3 weeks (±2 days). The initial dose of atezolizumab was delivered over 60 (±15) min.

Tumor assessments were conducted at baseline, every 12 weeks thereafter, and at the end of treatment by CT and bone scanning. Efficacy outcome measures included radiographic progression-free survival (rPFS), objective response rate (ORR), and disease control rate (DCR) by Prostate Cancer Working Group 3-modified Response Evaluation Criteria in Solid Tumors (RECIST) V.1.1,32 and OS. PSA levels were not used to determine disease progression or to trigger radiographic evaluations. At the conclusion of the study, a blinded, independent radiological review was used to confirm the time to objective disease progression.

All treatment-emergent adverse events were reported until the time of objective disease progression. Thereafter, only events that were determined by the investigators to be at least possibly related to sipuleucel-T and/or atezolizumab were reported. Monitoring for treatment-emergent adverse events (AEs) and survival occurred at 2 and 6 months after disease progression and at intervals of 6 months or less thereafter. All subjects who had undergone at least one leukapheresis procedure or atezolizumab infusion were included in the safety population. Adverse events and laboratory values were graded using the National Cancer Institute Common Terminology Criteria for Adverse Events, V.4.0. Immune-related AEs (irAEs) were defined as per previous report.33 Multiple occurrences of specific events were counted once per patient; the event with the greatest severity was summarized. Additional anticancer interventions and causes of death were collected for all subjects. The cut-off date for data presented here was October 15, 2020.

A diagnostic antihuman PD-L1 monoclonal antibody (Clone 22C3, Agilent Technologies/Dako, Carpinteria, California, USA) was used to detect PD-L1 expression on formalin-fixed, paraffin-embedded tumor tissue by immunohistochemistry (IHC), as previously described.34 PD-L1 positivity was defined as a combined positive score (CPS) of >10. The CPS was calculated by: (1) For an area of 200 tumor cells, count the number of PD-L1 positive cells (tumor cells, lymphocytes, and macrophages) and divide by 200 and then multiply by 100 to generate a score for this 200 tumor cell area. (2) Repeat this for a total of four 200 tumor cell areas. (3) Calculate the average of the scores from these four areas to calculate the final CPS. A board certified pathologist (OTMC) reviewed all slides.

Correlative testing was performed to assess immune response associated with study drug administration. Whole blood samples were collected at baseline and weeks 12, 16, 20, 32, 45, and 58 (Arm 1) or weeks 7, 11, 15, 27, 40, and 53 (Arm 2). Immune assays, as detailed below, were performed at Dendreon (Seal Beach, California, USA).

This study was designed to obtain preliminary safety and clinical activity information in subjects with asymptomatic or mildly symptomatic mCRPC. Analyses were based on all subjects who received any amount of study treatment (safety evaluable population). The ORR and DCR with corresponding 95% CIs were calculated using the Clopper-Pearson method. The rPFS and OS were assessed using the Kaplan-Meier method, with 95% CIs for median rPFS and OS estimated using the Brookmeyer-Crowley method. While this study was not powered to detect a survival difference between the two treatment arms, there was a protocol-specified requirement to follow each patient for survival (and treatment-related AEs) for up to 12 months after randomization.

From January 15, 2017 through October 30, 2019, 37 subjects were enrolled in this study (n=19 Arm 1, n=18 Arm 2). All 37 subjects were included in the safety population and thus analyzed for both safety and efficacy (online supplemental figure 1). Notably, three subjects did not complete induction treatment. These three subjects were replaced (two in Arm 1 and one in Arm 2); otherwise, all subjects received three doses of sipuleucel-T and two doses of atezolizumab. The most common reason for treatment discontinuation was disease progression. Median duration of treatment was 4.9 months (range, 0–9.9 months) for Arm 1 and 5.1 months (range, 1.4–13.7 months) for Arm 2. Baseline characteristics were generally as expected for a sipuleucel-T eligible mCRPC population (table 1). Median age for the total population was 75.0 years (range, 53–86 years), 73.0% of subjects had an ECOG performance status 0, 18.9% had received prior docetaxel-based chemotherapy, and 62.2% had received two or more previous antiandrogen therapies. Median PSA level for the total population was 21.9 ng/mL at study entry

At least one treatment-related AE was reported in 31 subjects (83.8%), including 7 (18.9%) with at least one grade 3 treatment-related AE. The most common treatment-related AEs were diarrhea, nausea, fatigue, and hypertension in Arm 1 and fatigue, pain, joint pain, platelet decrease, constipation, and anemia in Arm 2 (table 2A). irAEs, which were based on a list of common terms, occurred in 5 (13.5%) subjects with similar incidences between the two arms and all of which were grade 1 or 2 (table 2B). There were no grade 4 or five toxicities attributed to either study drug and no irAE required systemic steroid therapy.

In total, 23 subjects were evaluable for response (table 3). No patient in either arm had a CR and only one patient in Arm 2 had a PR. Three subjects (30%) in Arm 1 and one patient (7.7%) in Arm 2 had SD. Thus, for the overall population the ORR was 4.3% and the DCR was 21.7%. For the individual arms, the DCR was 30% for Arm 1 and 15.4% for Arm 2. These subjects are subsequently noted as responders. Among the 12 subjects who had at least one measurable target lesion by CT at baseline, 4 (33.3%) had a decrease from baseline in the sum of target lesions, including 1 (8.3%) with a PR and 2 (16.7%) subjects with SD (figure 1). Among the five subjects with radiographic response, one remained on treatment at data cut-off, while the remaining three had experienced subsequent disease progression (figure 2A). Overall, the median rPFS was 3.0 months (95% CI 2.8 to 5.6 months). Median rPFS was 3.3 months (95% CI 2.7 to 7.8 months) for Arm 1 and 2.9 months (95% CI 2.6 to 5.5 months) for Arm 2 (figure 2B). Median OS was not reached (NR; 95% CI 11.1 to NR) in Arm 1 and 21.4 months (95% CI 16.6 to NR) in Arm 2 (figure 2C). The estimated 12-month survival rates were 70.6% in Arm 1 and 83.3% in Arm 2. For both arms, median OS was 23.6 months (95% CI 16.6 to NR) and the estimated 12-month survival rate was 77.2%. A total of 15 subjects (40.5%) out of 37 had some decrease in PSA level, including 1 patient (5.3%) in Arm 1 and 3 subjects (16.7%) in Arm 2 who had a >50% decrease (table 3 and online supplemental figure 2). PD-L1 expression in patient samples is shown in online supplemental figure 3). Only one subject, who was classified as a non-responder in Arm 2, had CPS >10. Furthermore, non-responders in both arms had numerically higher levels of PD-L1 expression compared with responders.

First, we performed TCR repertoire analysis using whole blood samples from 16 subjects (n=9 Arm 1 and n=7 Arm 2) to assess the functional phenotype associated with each treatment. TCR clonotype analysis revealed that the main population consists of (T Cell Receptor Alpha Constant) TRAC and (T Cell Receptor Beta Constant) TRBC subtypes (figure 3A, online supplemental figure 4). Interestingly, the TRAC population tended to decrease and TRBC population tended to increase after the treatment in Arm 1, but not in Arm 2. In addition, we found that TCR diversity was relatively higher in the one responder in Arm 2 compared with non-responders in both arms (figure 3B).

Next, using a panel of T-cell related cytokines, we compared cytokine expression in serum pretreatment and post-treatment using specimens from 22 subjects (11 in each arm). Overall, the heatmap shows that cytokine levels in the post-treatment responder is higher than those in baseline (figure 3C and online supplemental figure 5) compared with non-responder changes from baseline, suggesting that T-cell activation by sipuleucel-T may be associated with response.

Further analysis showed mean PA2024-specific ELISpot counts (averaged over time) were similar in both arms (52.3 spots in Arm 1 vs 45.4 spots in Arm 2; p=0.62). In both arms, PA2024-specific ELISpot counts were significantly increased compared with baseline at each postbaseline visit (Arm 1: baseline vs week 12; p=0.022 and Arm 2: baseline vs week 11; p<0.0001) (figure 3D and online supplemental table 1). When averaged across all time points, PA2024-specific T-cell proliferation responses were numerically but not statistically higher in Arm 1 compared with Arm 2 (p=0.34; figure 3D and online supplemental table 1). At all time points through week 32, PA2024-specific T-cell proliferation responses were significantly higher compared with baseline in both arms (Arm 1; p=0.034; figure 3D, baseline vs week 12; p=0.0097) (Arm 2; p=0.019; figure 3D, baseline vs week 11; p=0.011, baseline vs week 11; p=0.049, and baseline vs week 15; p=0.014). All subjects developed PA2024-specific T-cell responses after sipuleucel-T treatment. PA2024 antibody titers after sipuleucel-T treatment in Arm 1 and Arm 2 were 16.8 times (p=0.00066) and 12.8 times (p=0.00025) higher, respectively, on average compared with baseline, and similar between arms, remaining significantly elevated through week 32 (figure 3 and online supplemental table 1).

Exploratory analysis of PA2024 T-cell stimulation index was performed in subjects for whom cells could be processed within 24 hours of collection, thus precluding the need to freeze the cells before analysis. The median ratio of the T-cell stimulation index at 12 weeks in Arm 1 and 11 weeks in Arm 2 was 32.9 times (p=0.00033) and 24.0 times (p=0.0014) higher, respectively, on average compared with baseline (preinfusion). Increased IgG levels to secondary antigens such as PSA, PAPi, and PAPm were observed in both arms at all time points through week 32 (online supplemental table 1); however, there was no significant difference between responders and non-responders in either arm.

In addition, to assess the T-cell population, PBMCs were collected and analyzed by flow cytometry for CD45+ (general leucocyte marker) and CD4+ and CD8+ T cells expression. The percentage of total T cells (Arm 1 29.4% vs Arm 2 30.3%, p=0.86), CD4+ T cells (Arm 1 13.2% vs Arm 2 12.5%, p=0.85), and CD8+ T cells (Arm 1 14.4% vs Arm 2 14.8%, p=0.92) were similar in both arms (representative results are shown in figure 3E). Interestingly, the percentage of total T cells was significantly higher in responders versus non-responders in Arm 1 (p=0.044) but not in Arm 2 (p=0.51). However, CD4+ or CD8+ T cells were numerically but not statistically higher in responders compared with non-responders in both arms (CD4+ T cell; p=0.72 for Arm 1 and p=0.92 for Arm 2, CD8+ T cell; p=0.15 for Arm 1 and p=0.46 for Arm 2).

The SERA assay was performed on samples from 37 subjects to assess autoantigen signal in subjects pretreatment and post-treatment in both arms. SERA uses a large random bacterial peptide display library with the PIWAS method to assess outlier antigens in samples. At baseline, a sum-of-IWAS calculation identifies a non-significant difference in the baseline number of outlier antigens in subjects with prostate cancer relative to a large cohort of healthy controls that becomes significant over time (online supplemental figure A,C). No differences were observed either at baseline or after therapy between responders and non-responders (online supplemental figure B,D). PIWAS identified 92 outlier antigens that were significantly different in responders versus. non-responders at baseline (online supplemental table 2).39–45 Serial PIWAS analysis also identified 40 antigens that increased between baseline and week 7 from Arm 2 (online supplemental table 3) and 18 antigens that increased in responders versus non-responders in either Arm 1 or Arm 2 (online supplemental table 4). Arms were combined for this analysis given the limited numbers of responders. The trajectories of response to three putative antigens, SIK3, KDM1A/LSD1, and PIK3R6, that were shared in at least two subjects in response to sipuleucel-T at (week 7, Arm 2) and also in response to combined therapy (week 12, either Arm 1 and Arm 2) are shown in figure 4.