Radium-223 prolongs survival in a fraction of men with bone metastatic prostate cancer (PCa). However, there are no markers for monitoring response and resistance to Radium-223 treatment. Exosomes are mediators of intercellular communication and may reflect response of the bone microenvironment to Radium-223 treatment. We performed molecular profiling of exosomes and compared the molecular profile in patients with favorable and unfavorable overall survival.
We performed exosomal transcriptome analysis in plasma derived from our preclinical models (MDA-PCa 118b tumors, TRAMP-C2/BMP4 PCa) and from the plasma of 25 patients (paired baseline and end of treatment) treated with Radium-223. All samples were run in duplicate, and array data analyzed with fold changes +2 to −2 and P < 0.05.
We utilized the preclinical models to establish that genes derived from the tumor and the tumor-associated bone microenvironment (bTME) are differentially enriched in plasma exosomes upon Radium-223 treatment. The mouse transcriptome analysis revealed changes in bone-related and DNA damage repair–related pathways. Similar findings were observed in plasma-derived exosomes from patients treated with Radium-223 detected changes. In addition, exosomal transcripts detected immune-suppressors (e.g., PD-L1) that were associated with shorter survival to Radium-223. Treatment of the Myc-CaP mouse model with a combination of Radium-223 and immune checkpoint therapy (ICT) resulted in greater efficacy than monotherapy.
These clinical and coclinical analyses showed that RNA profiling of plasma exosomes may be used for monitoring the bTME in response to treatment and that ICT may be used to increase the efficacy of Radium-223.
The bone targeting Radium-223 prolongs survival in a fraction of men with bone metastatic prostate cancer. Monitoring bone microenvironment response and resistance to Radium-223 will allow for treatment improvement in selected patients. Molecular messages in plasma exosomes provide information on tumor microenvironment in bone and also reveal that patients with unfavorable response to Radium-223 have higher levels of immune checkpoint modulators, suggesting that combination of immune checkpoint therapy with Radium-223 treatment may increase therapy efficacy.
Bone metastases occur in approximately 70%–80% of men with advanced prostate cancer (1). They contribute to the morbidity of disease and are often the site of resistance to therapy (2, 3). The interaction between the cancer cells and host stromal cells within the bone tumor microenvironment (TME) has been shown to promote prostate cancer progression (4, 5–8). To date, most life-prolonging systemic therapies in advanced prostate cancer, such as androgen ablation and chemotherapy, mainly target the tumor cells. In contrast, bone-directed therapies, such as denosumab [receptor activator of NFκβ (RANK) ligand], bisphosphonates, samarium-153, and strontium-89, target the bone TME and reduce the morbidity associated with bone metastases without providing a survival benefit (9). Radium-223 (Ra-223) was the first bone-homing radiopharmaceutical to improve median overall survival (OS) when compared with placebo in the ALSYMPCA phase III clinical trial (10). This was the first evidence supporting the hypothesis that targeting the metastatic bone TME can improve symptoms and prolong survival in men with metastatic castration-resistant prostate cancer (CRPC).
Ra-223 is an alpha-emitting radiopharmaceutical that releases high linear energy with a range of 100 μm (11), promoting double-strand DNA breaks within the bone TME (12). This bone-targeting agent focuses on tumor-induced osteoblasts by mimicking calcium that is complexed with hydroxyapatite (13). Consistent with this, Ra-223 treatment was associated with a 30% or greater decline in serum levels of alkaline phosphatase (ALP; 47% of patients in Ra-223 arm vs. 3% in placebo arm), a marker of osteoblastic activity. Surprisingly only a small subset of patients experienced a 30% or greater reduction in serum levels of PSA (16% in Ra-223 arm vs. 6% in placebo arm), which is produced by prostate cancer cells. These results suggest that Ra-223 mainly targets the tumor-associated microenvironment within bone metastases.
Because of the difficulty in obtaining potentially informative serial bone biopsies, there are currently no specific markers of efficacy for Ra-223 in patients with metastatic CRPC. It is a clinical challenge to identify which patients respond to Ra-223 and which do not and the mechanisms of resistance. The development of noninvasive methods that can continuously monitor the responses to therapy is particularly appealing. Exosomes are secreted vesicles that are detected at increased levels in the plasma of patients with cancer (14, 15). Exosomes contain lipids, RNA, DNA, and proteins from the parental cells, and they can modulate the phenotype of recipient cells by delivering their cargo. Exosomes have been shown to play a key role in homotypic and heterotypic cell-to-cell interactions within the local and distant TME (16). In particular, exosomes are implicated in the formation of a TME that favors progression and metastasis in many cancers including prostate cancer (17).
In this study, we sought to address this clinically unmet need to identify which patients are resistant to Ra-223 by interrogating plasma-derived exosomes for predictive markers. We conducted parallel clinical and coclinical studies to gain insights into molecular changes associated with Ra-223 treatment in plasma exosomes. We found evidence that plasma-derived exosomes can be used for monitoring tumor-associated bone microenvironment in response to Ra-223 treatment and identify the possible resistance mechanisms.
Materials and Methods
We conducted a clinical trial (ClinicalTrials.gov ID: NCT02135484) to determine the effect of Ra-223 on the bone marrow microenvironment in patients with CRPC and bone metastases (Supplementary Table S1). Eligible patients had CRPC metastatic to bone, Eastern Cooperative Oncology Group ≤ 2, white blood cell count > 3,000/μL, absolute neutrophil count > 1,500/μL, hemoglobin ≥ 8.0 g/dL independent of transfusion, and platelet count ≥ 100,000/μL. Patients with visceral metastases and/or lymph nodes > 6 cm in the short axis were excluded. There were no exclusions based on number of prior androgen receptor–targeting therapies or chemotherapies, but prior treatment with a bone-homing radiopharmaceutical (e.g., radium, strontium, or samarium) was not permitted. Subjects were treated with the standard dosing of Ra-223 dichloride 50 kBq (0.0014 mCi)/kg intravenously every 4 weeks for six cycles (six doses total). Blood for biomarker discovery was collected at baseline and prior to each Ra-223 dose. Transiliac bone marrow biopsies were collected at baseline, week 12 (prior to Ra-223 dose 4), and end of treatment (EoT). Patients were scheduled to be treated with six doses of Ra-223 barring patient withdrawal, toxicity, or disease progression. The majority of the patients in both favorable OS (FavOS) and unfavorable OS (UnFavOS) received all six doses till the EoT, and we analyzed the plasma from that timepoint. For the patients that did not complete the Ra-223 course, we used the plasma from the last treatment. (Supplementary Table S2).
The murine prostate cancer cell line MyC-Cap was purchased from ATCC. MyC-Cap-Luc+-RFP cells were kindly provided by Guocan Wang (MDACC, Houston, TX). TRAMPC2-BMP4 cells were kindly provided by Sue-hwa Lin and used as described previously (18). The cell lines were maintained in DMEM at 37°C and 5% CO2 as the vendor recommended. The DMEM was supplemented with 10% FBS and 1% penicillin/streptomycin. Routine testing for Mycoplasma contamination was performed. Identification of cells was confirmed by fingerprinting analysis at IDEXX Laboratories Inc.
MDA-PCa 118b cells (1 million/site) were implanted subcutaneously in SCID mice. The mice were killed when tumors reached 8 to 10 mm in size and the plasma was collected for exosome isolation. For the Ra-223 study, mice with MDA-PCa 118b tumors were treated with vehicle (control group) or Ra-223 (300 kBq/kg). Plasma was collected 2 weeks after Ra-223 treatments.
FVB mice were purchased from Taconic BioSciences and the mice were housed at the animal facility of Experimental Radiation Oncology, Department of Veterinary Medicine & Surgery, MD Anderson Cancer Center (Houston, TX). All animal studies were conducted in accordance with the current regulations and standards of the U.S. Department of Agriculture, the U.S. Department of Health and Human Services, and the NIH and were approved by The University of Texas MD Anderson Cancer Center Institutional Animal Care and Use Committee.
For the combination treatment studies, we chose to utilize the syngeneic Myc-Cap as our model system instead of TRAMP-C2. When TRAMP C2 cells are transplanted into BL/6 mice, they develop neuroendocrine tumors that are clinically rare compared with the adenocarcinomas that are most often seen in patients with prostate cancer (19). Furthermore, it is an osteolytic model of prostate cancer. Myc-Cap cells form androgen-dependent adenocarcinomas in mice and reflect more closely to the biology of prostate cancer seen in the clinic (20, 21). The mice used in this experiment were 5 to 6 weeks old and all were surgically injected with 1 × 106 MyC-Cap cells in the right femur with 5 μL of PBS. The cells used for femoral injection were at 60% to 70% confluency. All animals were regularly monitored for changes in their health status and weight. Tumor growth in bone was measured by MRI every 2 weeks after initiation of treatment, and tumor volume was calculated using ImageJ (ImageJ, RRID:SCR_003070). Mice were euthanized when tumor growth was found outside of the femur or if they had weight loss exceeding 30%. Both femurs were collected and fixed in formalin buffer saturated with 10% PBS. Hematoxylin and eosin and IHC staining were conducted.
Exosome isolation from plasma
Plasma samples were thawed on ice and gently mixed by rotating for 2 minutes at 4°C. Then the samples were spun down at 500 × g for 20 seconds at 4°C. Continuously, 500 μL of plasma sample was mixed by inversion with 500 μL of cold PBS and centrifuged at 12,000 g for 45 minutes at 4°C to remove cellular debris. The supernatant was transferred to an ultracentrifuge tube, and 8 mL of cold PBS was added and ultracentrifuged for 2 hours at 120,000 × g for 2 hours. The supernatant was discarded, and 10 μL of the exosome-enriched pellet was diluted in 40 μL of PBS for nanosight analysis while the rest was lysed with RIPA buffer for Western blot analysis. For RNA isolation, the exosome pellet was resuspended directly in Qiazol Lysis Reagent and proceeded immediately to RNA extraction.
RNA extraction from plasma-derived exosomes
The miRNeasy Micro Kit (Qiagen, catalog no.: 217084) was used to extract RNA from plasma-derived exosomes, according to the manufacturer's instructions, with minor modifications. Briefly, exosomes were isolated as described above. In the exosome-enriched pellet, 700 μL of Qiazol Lysis Reagent was added and homogenized by vortexing. The homogenate was incubated for 15 minutes at room temperature, and 140 μL of chloroform was added. The samples were mixed vigorously for 15 seconds, incubated at room temperature for 3 minutes, and centrifuged for 15 minutes at 4°C. Continuously, the upper aqueous phase containing the RNA was transferred to another tube and 1.5 volumes of 100% ethanol was added, mixed, and transferred to an RNeasy MinElute spin column and centrifuged at 8,000 × g for 15 seconds, at room temperature. Then, 700 μL of RWT buffer was added to the RNeasy MinElute spin column and centrifuged at 8,000 × g for 15 seconds at room temperature, and the flow-through was discarded. In the RNeasy MinElute spin column, 500 μL of RPE buffer was added and centrifuged at 8,000 × g for 15 seconds at room temperature, and the flow-through was discarded. Next, 500 μL of 80% ethanol was added, the samples were centrifuged at 8,000 × g for 2 minutes at room temperature, and the flow-through was discarded. The spin column was then dried by centrifuge at 20,800 × g for 5 minutes at room temperature. Finally, 14 μL of RNase-free water was added to the spin column membrane and centrifuged at 20,800 × g for 1 minute at room temperature. The RNA quantification was performed with an Implen NanoPhotometer.
Exosomal RNA sequencing
Illumina-compatible libraries were prepared using the Ovation RNA-seq System V2 (Nugen) and the KAPA Hyper Library Preparation Kit (Kapa Biosystems, Inc). Briefly, exosomal RNA samples treated with DNase-1 were assessed for size distribution and quantity using the Fragment Analyzer High Sensitivity RNA Analysis Kit (Advanced Analytical) and the Qubit RNA HS Assay Kit (Thermo Fisher Scientific), respectively. One and a half nanograms of RNA were converted to double-stranded cDNA, then amplified using Nugen's proprietary single primer isothermal (Ribo-SPIA) protocol. Five hundred nanograms of the resulting cDNA were fragmented to an average size of 200 bp, and libraries were constructed using the KAPA Hyper Library Preparation Kit, followed by two cycles of PCR library enrichment. Following cleanup, the libraries were mixed (three libraries per pool), then quantified by qPCR using the KAPA Library Quantification Kit (KAPA Biosystems) and sequenced with one pool per lane on the Illumina HiSeq4000 Sequencer using a 75 bp paired-end format.
Gene expression array data analyses
The Affymetrix GeneChip Human Transcriptome Array 2.0 (HTA 2.0) was utilized by the core gene expression analysis unit as described by the manufacturers. Affymetrix CEL files were processed and quantile normalized with GC-RMA background correction using the Expression Console and Transcriptome Analysis Console (TAC) 2.0 (Affymetrix). Differential expression of genes was statistically evaluated using TAC 2.0. P < 0.05 was considered significant.
To assess the biological processes affected, the identified transcripts were submitted to the Ingenuity Pathway Analysis (IPA; RRID:SCR_008653) program for functional analysis or canonical pathway analysis. Functional analyses identify functions that are significantly enriched in the dataset. The microarray dataset, which includes log2 fold changes and differential regulation status for each identified transcript, was uploaded into IPA using the core analysis platform (Qiagen, Ingenuity Systems). The differentially expressed microarray data from all of our preclinical and clinical datasets were matched with those in the Ingenuity Knowledge Base (human/mouse) and also with the IPA mouse database, separately. Unmapped RNA was excluded from further analysis. Right-tailed Fisher exact test was used to calculate P values to determine the probability that each biological function assigned to the dataset was due to chance alone. Ingenuity-defined canonical pathways that were most significantly enriched in the dataset were reported. In this case, Fisher exact test was used to calculate the probability that the association between the genes in the dataset and the canonical pathways could be explained by chance alone. The results were based on log (P).
All results were expressed as mean ± SD of three independent experiments. The statistical significance of the difference in the means between groups was analyzed using Student t test with GraphPad Prism 6.0 (GraphPad Prism; RRID:SCR_002798). P < 0.05 was considered statistically significant.
Detection of bone-related markers in plasma exosomes from a bone-metastatic patient-derived xenograft
To test whether the molecular profile of circulating exosomes reflects tumor-induced changes within the bone microenvironment, we utilized the MDA-PCa 118b, a well-established xenograft [patient-derived xenograft (PDX)] model derived from a patient with CRPC with bone metastasis. This model has been shown to induce bone formation when implanted either into the mouse femur or subcutaneously (22, 23). Plasma from mice with or without MDA-PCa 118b tumors were collected and exosomes were isolated. The exosomal RNA was subjected to RNA sequencing (RNA-seq) analysis. There were no significant differences in the amount or quality of the RNA in the exosomal RNAs from mice with or without MDA-PCa 118b tumor (Supplementary Fig. S1). Species-specific comparative analysis was utilized to distinguish exosomal transcriptional changes in cancer epithelial cells (human) versus bone stromal cells (mouse) by Ra-223 (Fig. 1A). The bioinformatics analysis of the mouse and human exosomal RNA revealed that there was little species overlap in the genes identified in the exosomes (2.3% overlap for the upregulated genes and 1.3% for the downregulated genes) that were identified in the exosomes (Fig. 1B). Ingenuity analysis of the top pathways identified in the mouse transcriptome revealed pathways associated with junctions, cytoskeletal remodeling, and bone formation (Fig. 1C), which were host specific and were not identified in the human transcriptome. In contrast, the human transcriptome was enriched in pathways regulating tumor growth such as mTOR, Ephrin, ERK/MAPK, and CXCL4 signaling cascades (Fig. 1D), which were consistent with the human origin of the tumor.
Because the MDA-PCa 118b is an osteogenic prostate cancer model, we interrogated the exosomal transcriptome for molecular evidence of bone-related changes and transcripts. There was an enrichment in mouse-specific pathways associated with bone formation, such as TGFβ, Wnt/β-catenin, and calcium signaling pathways, and a concomitant decrease of genes associated with osteoclast activation (Fig. 1E). The most enriched, mouse-specific, bone-related transcript was fstl3, which is known to inhibit osteoclast differentiation and promote bone formation (Supplementary Fig. S2A). Bone morphogenic proteins (BMP) also play a critical role in bone formation and were identified in the mouse transcriptome. Bmpr1b, bmp3, and bmp6, which are known to be involved in new bone formation, were the most enriched, mouse-specific, BMP family members compared with nontumor-bearing mice, whereas there was a downregulation of BMP1, BMP5, and BMP7, which have been associated with tumor dormancy in the bone. BMP7, in particular, has been shown to promote dormancy in prostate cancer (ref. 24; Supplementary Fig. S2B). These data indicate that plasma-derived exosomes are enriched in mouse-specific bone-related transcripts, reflecting the TME of the osteogenic prostate cancer model.
Detection of pharmacodynamic changes induced by Ra-223 within the bone TME and plasma exosomes from a bone-metastatic PDX
Ra-223, an alpha-particle emitter, has selectivity for remodeling bone, and it has been shown to induce cell death in osteoblasts (25, 26). To determine whether the exosomal transcriptome can reflect Ra-223 treatment-related changes, tumor-bearing mice were treated with or without Ra-223 for 2 weeks. Plasma from these mice was collected at the end of treatment, exosomes were isolated, and exosomal RNA was subjected to RNA-seq analysis (Fig. 2A and B). There was no significant difference in the amount of exosomal RNA from mice with or without Ra-223 treatment. The only noticeable difference was an alteration in the size and concentration of exosomal RNA from mice treated with Ra-223 (Supplementary Fig. S1). However, the significance of these findings remains unknown. It may reflect changes in the mechanisms of secretion, of molecular content, or of the type of cells that the exosomes are originating from.
Species-specific comparative transcriptome analysis was again utilized to distinguish Ra-223–induced exosomal transcriptional changes in cancer epithelial cells (human) versus bone stromal cells (mouse; Fig. 2A and B). In the mouse-specific transcriptome analysis, there were changes in 2,803 mouse genes in response to Ra-223 treatment, whereas the human exotranscriptome had 737 genes with species overlap of 1% for the upregulated genes and 1.4% for the downregulated genes (Fig. 2B). To determine whether the pharmacodynamic changes observed in bone biopsies could also be detected in the plasma exosomes, we interrogated the exosomal transcriptome for changes in bone-related pathways. To confirm the targeted efficacy of Ra-223 on osteoblasts in mice subcutaneously implanted with MDA-PCa 118b tumors, changes in the expression of osteocalcin (a well-established osteoblast-secreted protein) were examined. As expected, Ra-223 reduced the expression of osteocalcin (Fig. 2C). The analysis of exosomal transcriptome revealed that the bone-related pathways that were enriched in the plasma exosomes from the MDA-PCa 118b model were downregulated upon treatment with Ra-223 (Supplementary Fig. S2C and D).
Ra-223 has been shown to induce DNA damage within the bone TME, leading to the activation of downstream DNA damage repair (DDR) signaling cascades (25). We examined the efficacy of Ra-223 to induce DNA damage by IHC staining in the MDA-PCa 118b tissue, for the activated form of ATM 9 (ATM phosphorylated on Ser 1891), an established marker for irradiation-induced DNA damage (27). Increased nuclear accumulation of the activated ATM was observed in the MDA-PCa 118b tissue (Fig. 2E). Importantly, Ra-223 induced the enrichment of exosomal transcripts related to DDR pathways, including the BRCA1, ATM, CHK, and NER pathways (Fig. 2F; Supplementary Fig. S2E). In summary, our osteogenic PDX model demonstrated that exosome profiling can detect Ra-223–induced molecular changes in the bone tumor and its associated microenvironment.
Plasma exosomes from patients detect pharmacodynamic changes in Ra-223
To confirm whether similar molecular features were found in plasma-derived exosomes from patients, we examined the exosomal transcriptomes of patients with metastatic CRPC to the bones treated with Ra-223 in a single-center, open-label clinical trial (NCT02135484; Fig. 3A). Between October 2014 and May 2015, 25 patients were enrolled and treated on the study. Patient characteristics are shown in Supplementary Table S1, and the median OS was 24.3 months. Serial plasma samples were collected prior to initiation of the first Ra-223 dose (baseline) and at the EoT. Bone marrow biopsies were performed at baseline.
Nanoparticle tracking analysis for the exosomes isolated from the plasma of patients at baseline and EoT demonstrated size and concentration typical of exosomes before and after Ra-223 treatment (Fig. 3B). The patient-derived exosome transcriptome was evaluated for changes in bone and DDR-related signaling pathways. Similar to the data from the PDX model, Ra-223 downregulated the bone forming TGFβ, calcium transport, and Wnt/β-catenin pathways and upregulated RANK signaling (Fig. 3C). Furthermore, DDR pathways were enriched in the patient-derived exosomes at the EoT (Fig. 3D).
Plasma exosomes detect biomarkers associated with prolonged OS in patients treated with Ra-223
To associate changes in exosomal transcriptomes with clinical outcomes, we stratified patients based on OS. Those patients who experienced OS greater than the median of 24.3 months were assigned to the FavOS cohort (n = 13), whereas those with lower than the median were assigned to the UnFavOS cohort (n = 12; Fig. 4A; Supplementary Table S3). Subsequently, the exosomal transcriptomes of the two groups were compared.
As part of the initial exosome characterization, the size and relative concentration of the vesicles isolated from the plasma of FavOS and UnFavOS patients were compared. The size of the plasma-derived exosomes at baseline was around the typical 100 nm, and the relative concentration of the vesicles between the two groups was similar (Supplementary Fig. S3). Interestingly, at EoT, the mean size and relative concentration of the exosomes differed between the FavOS and UnFavOS patients (Supplementary Fig. S3). The FavOS group of patients had a higher concentration and smaller mean size compared with the UnFavOS. The significance of these phenotypic changes is currently unknown.
RNA isolated from the exosomes was subjected to the Affymetrix Human Transcriptome Array 2.0. The analyses demonstrated differential enrichment of signaling pathways associated with tumor growth, DDR, and immune responses between FavOS and UnFavOS cohorts at EoT (Fig. 4B). With regard to tumor growth, several oncogenic pathways, including PLK, Tec, AhR, and NGF pathways, were enriched in exosomes derived from the UnFavOS group. With regard to bone signaling, IPA revealed that a number of bone formation–related pathways are altered in the transcriptome (Fig. 4C; Supplementary Fig. S4; ref. 28). Although there was a decrease in WNT/β, BMP, and RANK signaling in the plasma exosomes derived from FavOS patients (Fig. 4C), exosomes from the UnFavOS did not demonstrate any changes in these pathways. The data from patients are in concordance with our observations from the PDX model. To validate the transcriptome analyses at the protein level, we performed a Luminex multiplex array for bone-related markers in a small cohort of patients who had sufficient samples available. There were a number of bone-related chemokines such as MCSF, Tweak, MIG, MIP-1b, and DKK1 (Supplementary Fig. S5A–S5E), which have all been shown to promote osteoclast differentiation or activation were enriched at EoT of the UnFavOS patients. Interestingly, there were dynamic changes of these markers after cycle 1 and cycle 3 of Ra-223 and at EoT. These data collectively show that there is good concordance between RNA transcripts and protein levels and that we can monitor dynamic changes in a longitudinal fashion for individual patients.
Furthermore, we found that the exosomes were enriched with transcripts related to DDR pathways at the EoT compared with baseline (Fig. 4D). The enriched transcripts that we observed indicate that specific DDR responses to ionizing irradiation are induced by Ra-223. In summary, these data demonstrate that we can monitor Ra-223–induced changes in bone and DDR-related pathways in the exosomes derived from Ra-223–treated patients.
Immune pathways identified from plasma exosomes in patients treated with Ra-223
In the exosome transcriptome derived from patients that participated in the trial, 3 of the top 10 altered pathways were involved in immune regulation and immune checkpoint activation (Fig. 4B). Upon further dissection of the immune pathways, there was an upregulation of transcripts in UnFavOS exosomes related to PD-1 signaling and, in particular, Pd-l1, a central immune checkpoint (Fig. 4E). The presence of PD-L1 within the exosomes was confirmed at the protein level by Western blot analysis. The UnFavOS patient-derived exosomes had higher levels of PD-L1 at baseline compared with the FavOS cohort (Fig. 5A and B).These findings were further validated at the protein level using a Luminex multiplex array for a number of immune checkpoint modulators, which demonstrated that exosomes derived from the UnFavOS group had significantly higher levels of PD-L1, LAG3, and IDO at EoT than those from FavOS (Fig. 5C; Supplementary Fig. S6). There were no differences between the two patient groups for PD-1 and CD80 (Supplementary Fig. S6). Importantly, these changes were Ra-223 dependent because there were no differences in the same immune checkpoint modulators upon cabazitaxel treatment (Supplementary Fig. S7). To determine whether the enrichment of PD-L1 that we observe in the exosomes correlates with its presence in solid tumor biopsies, we performed IHC analysis for PD-L1 in one of the UnFavOS patients (Patient 20) at baseline. IHC demonstrated high levels of PD-L1 which is in agreement with the levels of this immune checkpoint activator detected in the plasma-derived exosomes at the RNA and protein level (Fig. 5D).
Targeting of the Ra-223 plus PD-(L)1 pathway in an immunocompetent prostate bone cancer model
To further examine the immune-related pathways in exosomes, we used a syngeneic prostate cancer model, which has a fully competent adaptive immune system. TRAMP-C2/BMP4 tumors, a bone-forming model of metastatic prostate cancer, were injected into the femurs of syngeneic C57BL/6 mice (18). Exosomes derived from the Ra-223–treated mice were enriched in transcripts related to cell death and apoptosis, bone, DDR, and immune response, similar to those in SCID mice (Supplementary Fig. S8A–S8D). Several immune-related pathways were enriched in the exosomes derived from mice treated with Ra-223, such as the immune checkpoint activators CTLA-4 signaling cascades and PD-L1 (Supplementary Fig. S8E and S8F).
On the basis of the data from exosome transcriptional profiling, which show the apparent induction of pd-l1 expression in response to Ra-223, we examined whether the combination of Ra-223 and immune checkpoint blockade (ICB) would potentiate the efficacy of Ra-223. To this end, we utilized the Myc-Cap prostate cancer syngeneic mouse model which is more clinically relevant than the TRAMP-C2 model. Prior to implantation, we performed a series of in vitro experiments to determine the sensitivity of Myc-Cap cells to Ra-223 (Supplementary Fig. S9). We found that after 24 hours of treatment, the low concentration of cells was partially sensitive to Ra-223, whereas the higher concentration was not (Supplementary Fig. S9A and S9B). These data indicate that tumor burden may be a determining factor in the efficacy of Ra-223. We measured the protein levels of PD-L1, active caspase-3 (apoptosis marker), and Bcl/xL (antiapoptotic marker) and found that active caspase-3 is induced, suggesting the activation of an apoptotic process. PD-L1 is induced, but Bcl-xL does not change (Supplementary Fig. S9C–S9F).
For in vivo study, FVB mice were injected in the femur with 1 × 106 Myc-CaP cells. The groups were normalized for tumor volume in the femur, and the tumor-bearing mice were randomized into four groups: Degarelix (Deg), Deg+Ra-223, Deg+ICB (anti-PD-1 and anti-CTLA-4), and Deg+Ra-223+ICB (Fig. 6A). Mice were then treated for 45 days. We monitored tumor growth in the femur by MRI and found that Ra-223+ICB promotes greater tumor regression than was observed with either monotherapy (Fig. 6B). We determined the efficacy of Ra-223 by measuring the number of osteoblasts/bone surface or per bone perimeter and total area and found that Ra-223 reduced the number of osteoblasts whereas degeralix and ICB alone did not have any effect (Fig. 6C; Supplementary Fig. S11). We isolated exosomes from the plasma of these mice, measured their size and concentration (Supplementary Fig. S10), and found that the exosomes had typical representative size of about 100 to 150 nm and that Ra-223+ICB increased the relative concentration of exosomes, whereas ICB decreased them (Supplementary Fig. S10A–S10C). We subjected the isolated exosomes to Western blotting to determine PD-L1 protein levels and found that mice treated with Ra-223 had higher levels of PD-L1 in the exosomes compared with degarelix alone (Fig. 6D). To determine the efficacy of Ra-223 in inducing PD-L1 expression and the efficacy of ICB in inducing T-cell activation (ICOS), we performed IHC analysis for PD-L1 and ICOS (Fig. 6E). We found that in the Deg+Ra-223 and Deg+Ra-223+ICB groups, there was an increase of PD-L1 in tumor cells, and an increase in ICOS was evident only in Deg+Ra-223+ICB. These results suggest that Ra-223 treatment increases PD-L1 and combination therapy with Ra-223+ICB can improve treatment outcomes.
There is an unmet clinical need to identify predictive markers that can inform the biology, the mechanisms of drug action, and development resistance to Ra-223 and upcoming bone-targeting agents. Such markers will allow for the development of rationally designed bone-targeting therapies. Bone metastatic prostate cancer is most frequently characterized by new bone formation, and the serum concentrations of bone ALP correlate well with this progression and serve as the golden standard in the clinic (29). Ra-223 targets the bone microenvironment and osteoblasts, in particular, by binding to sites of active mineralization, which are often observed in bone metastatic prostate cancer, resulting in declines in ALP. We attempted to stratify the patients based on ALP levels at baseline; however, the pathologic level of ALP is set to more than 200 UI/L, allowing for only 4 of 25 patients in our cohort to be analyzed. Furthermore, correlation of ALP decline (more or less than 30%) in this small cohort of patients did not correlate well with OS. Thus, we stratified patients based on OS with 24.3 months as the cutoff into two groups, FavOS and UnFavOS, and we analyzed the exosomal transcriptome in a dynamic manner at baseline and EoT (or in certain cases after cycle 1 and cycle 3).
Our analysis showed evidence of changes in the exosomes at the protein and RNA levels related to bone forming and bone lytic pathways, the balance of which defines the bone-forming or bone-erosive nature of the prostate cancer. In fact, our data show that there is decreased osteoblastic and increased osteoclastic activity at EoT with Ra-223 both in the preclinical and clinical settings. We observed increased RANK signaling in osteoclasts upon Ra-223 treatment, and we could identify in the tissue sections the presence of osteoclasts surrounding the cluster of Myc-Cap prostate cancer cells. Interestingly, we observed a similar enrichment of RNA and protein levels for a number of factors that are associated with osteoclast differentiation, maturation, and activity in the UnFavOS group compared with the FavOS group. It is tempting to speculate that patients who do not respond to Ra-223 are suffering from the osteolytic type of prostate cancer, and the addition of Ra-223 exacerbates this phenomenon, worsening their condition. Further studies are required to determine whether Ra-223 is more effective in bone-forming than bone-erosive metastatic CRPC.
Ra-223 is an alpha particle–emitting radiopharmaceutical that has been shown to induce DNA damage and, in particular, to induce double-strand breaks (DSB; ref. 30). The correlation between Ra-223 and DDR machinery was recently suggested in a retrospective study in a small cohort of patients; those who inherited or acquired DNA repair gene mutations derived greater benefit from Ra-223 compared with those who did not (31). DSBs are repaired via two main pathways, nonhomologous end joining and homologous recombination. A number of genes involved in DDR were identified in the transcriptome of the exosomes analyzed from the preclinical and clinical studies, such as ATM, ATR-interacting protein (Atrn), XRC family members, and ZEB. Interestingly, ZEB and ATM are key mediators for DDR, responsible for the induction of ATM radioresistance (32). Collectively, these data indicate that DDR-related changes can be monitored in the exosomal transcriptome, a feature that could be a pharmacodynamic measure of the efficacy of Ra-223 efficacy. Furthermore, these data suggest that a potential combination with a DDR inhibitor (PARP or ATM inhibitor) would potentiate the efficacy of Ra-223 alone.
It is well established that the bone and immune system are intricately related, not only because multiple members of the immune system homing at the bone marrow but also because progenitors of the immune system differentiate to osteoclasts. Thus, it is not surprising that following treatment with a bone-targeting agent like Ra-223, we observed a number of changes in immune-related markers. Our approach has demonstrated that Ra-223 treatment leads to an increase in immune checkpoint modulators including PD-L1 in vitro and in vivo. This induction was specific for Ra-223, because it was not detected in samples from patients treated with chemotherapy. The presence of PD-L1 in exosomes and its functional significance in mediating the activation of immune checkpoint and T-cell inactivation have recently been shown (33, 34). Thus, we tested the significance of our findings by combining Ra-223 with ICB and found that the combination had greater efficacy than Ra-223 alone, suggesting that this combination may be a good candidate for a clinical trial. Furthermore, following our observations of DDR induction in response to Ra-223, it is feasible that a triple combination of Ra-223, inhibitors for DDR (ATM/ATR inhibitor), and ICB may be even more beneficial.
It has been recently shown that PARP inhibitors can induce the upregulation of PD-L1 in cancer cells by various mechanisms and that is the rationale in combining PARP inhibitors with ICT and indeed various preclinical and clinical studies have been conducted in different cancer types (35). In patients with mCRPC, there is an ongoing clinical trial with the PARP inhibitor olaparib and the anti-PD-L1 antibody durvalumab where the combination therapy reduced PSA in 47% of the patients and the benefit was greater in those with mutation in DDR machinery (36).
Taken together, our clinical and coclinical studies suggest that plasma exosomes may be used to predict therapy benefit and to monitor therapeutic efficacy in men with prostate cancer bone metastases treated with Ra-223 or other bone-targeted therapies.
Our observations support the hypothesis that it is possible to longitudinally monitor the biology of prostate cancer progression in the bone by serial molecular profiling of exosomes. In the case of therapy with Ra-223, it is feasible to monitor changes in exosomal DDR as a pharmacodynamic measure. In addition, we could detect changes in the immune system as shown with activation of immune checkpoint modulators. Our coclinical and clinical findings are consistent with each other and with observed clinical outcomes, and they support the notion that exploitation of the molecular information encapsulated in exosomes may provide important information about the mechanisms of action and resistance and, thus, lay the foundation for the implementation of combination therapies to improve therapeutic efficacy for bone metastasis.
E. Efstathiou reports advisory board membership with Bayer. S.K. Subudhi reports personal fees from Apricity Health, Janssen, Dendreon, Polaris, Parker Institute for Cancer Immunotherapy, Amgen, AstraZeneca, Bayer, Bristol-Myers Squibb, Dava Oncology, and the Society for Immunotherapy of Cancer outside the submitted work. No disclosures were reported by the other authors.
I. Vardaki: Data curation, formal analysis, validation, investigation, writing–original draft, writing–review and editing. P. Corn: Conceptualization, formal analysis, writing–original draft, writing–review and editing. E. Gentile: Formal analysis, investigation. J.H. Song: Data curation, formal analysis, investigation. N. Madan: Formal analysis, investigation. A. Hoang: Formal analysis, investigation. N. Parikh: Data curation, formal analysis, investigation. L. Guerra: Formal analysis, writing–original draft, writing–review and editing. Y.-C. Lee: Data curation, formal analysis. S.-C. Lin: Data curation, investigation. G. Yu: Formal analysis, investigation. E. Santos: Formal analysis. M.P. Melancon: Formal analysis. P. Troncoso: Data curation, formal analysis. N. Navone: Resources, data curation. G.E. Gallick: Resources, funding acquisition, investigation. E. Efstathiou: Resources, data curation. S.K. Subudhi: Resources, funding acquisition, investigation, writing–original draft. S.-H. Lin: Conceptualization, resources, formal analysis, writing–original draft, writing–review and editing. C.J. Logothetis: Resources, funding acquisition, investigation, writing–original draft. T. Panaretakis: Conceptualization, data curation, formal analysis, supervision, investigation, writing–original draft, project administration, writing–review and editing.
We acknowledge Bayer for providing Radium-223 for the clinical and preclinical studies. This work was supported by grants from the NIH (CA174798, 5P50 CA140388, P30 CA16672), the Prostate Cancer Foundation, Cancer Prevention and Research Institute of Texas (CPRIT RP150179, RP190252), as well as funds from The University of Texas MD Anderson Moon Shot Program, the Swedish Cancer Foundation (Cancerfonden), and the Radiumhemmets Research Foundation (Radiumhemmets forskning fonder).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.