Validation of SV2A-Targeted PET Imaging for Noninvasive Assessment of Neuroendocrine Differentiation in Prostate Cancer

Neuroendocrine prostate cancer (NEPC) is an aggressive and lethal variant of prostate cancer (PCa), and it remains a diagnostic challenge. Herein we report our findings of using synaptic vesicle glycoprotein 2 isoform A (SV2A) as a promising marker for positron emission tomography (PET) imaging of neuroendocrine differentiation (NED). The bioinformatic analyses revealed an amplified SV2A gene expression in clinical samples of NEPC versus castration-resistant PCa with adenocarcinoma characteristics (CRPC-Adeno). Importantly, significantly upregulated SV2A protein levels were found in both NEPC cell lines and tumor tissues. PET imaging studies were carried out in NEPC xenograft models with 18F-SynVesT-1. Although 18F-SynVesT-1 is not a cancer imaging agent, it showed a significant uptake level in the SV2A+ tumor (NCI-H660: 0.70 ± 0.14 %ID/g at 50–60 min p.i.). The SV2A blockade resulted in a significant reduction of tumor uptake (0.25 ± 0.03 %ID/g, p = 0.025), indicating the desired SV2A imaging specificity. Moreover, the comparative PET imaging study showed that the DU145 tumors could be clearly visualized by 18F-SynVesT-1 but not 68Ga-PSMA-11 nor 68Ga-DOTATATE, further validating the role of SV2A-targeted imaging for noninvasive assessment of NED in PCa. In conclusion, we demonstrated that SV2A, highly expressed in NEPC, can serve as a promising target for noninvasive imaging evaluation of NED.


Introduction
Prostate cancer (PCa) remains the second leading cause of deaths across all cancer types in American men [1]. Although the 5-year cancer-specific survival rate for patients with clinically localized and regional PCa is nearly 100%, the 5-year cancer-specific survival sharply drops to 31% for metastatic castration-resistant prostate cancer (mCRPC) [2]. Classified as neuroendocrine PC (NEPC) [3][4][5], a subset of mCRPC with universal neuroendocrine differentiation (NED) features manifests an even worse prognosis with an ical imaging. By extracting SV2A genomic data in clinical CRPC and NEPC specimens from the available gene expression repositories, we found that SV2A gene expression was significantly elevated. Given that several SV2A specific radiotracers (e.g., 11 C-UCB-A [46], 11 C-UCB-J [47], 18 F-UCB-H [48], and 18 F-SDM-8/SynVesT-1/2 [49,50]) have been reported for synaptic density PET imaging of neurodegenerative diseases [51,52], we performed a PET imaging evaluation in an NEPC xenograft model using 18 F-SynVesT-1, after having confirmed SV2A protein expression in NEPC cell lines and mouse xenografts through a comparative study.

SV2A Gene Expression in Clinical Tumor Tissues: NEPC vs. CRPC-Adeno
Our bioinformatic analyses were performed on combined data from three sources: a recent report by Beltran et al. [3], WCMC11/14, and SU2C 2015. We found that the normalized RNA expression of SV2A is significantly higher in NEPC (n = 20) than that in CRPC-Adeno (n = 147) ( Figure 1A, p = 0.0034). Similar results were observed in another dataset (GSE104786, Supplementary Figure S1A). In addition, we observed an intriguing trend of SV2A expression in a patient-derived xenograft (PDX) model (GSE59986, Figure 1B) as AdPC was progressing to NEPC after castration; the SV2A expression maintained at a relatively stable and low level within 12 weeks post-castration until the emerging of the terminally differentiated NEPC, resulting in a drastic increase. Furthermore, we found that the frequency of SV2A gene amplification in mCRPC and NEPC samples (54 out of 96 or 56%, dataset by Beltran et al. [3]) was markedly higher than that in AdPC samples (8 out of 499 or 1.6%, TCGA Firehose Legacy dataset) ( Figure 1C). As anticipated, NEPC samples also present higher expression levels of other neuronal markers in addition to SV2A than CRPC-Adeno samples (Supplementary Figure S1B,C). Taken together, SV2A expression is significantly elevated in NEPC and we reason it may serve as a potential target to develop efficacious theranostics for NEPC and other innervated cancers. desired feature for oncological imaging. By extracting SV2A genomic data in clinical CRPC and NEPC specimens from the available gene expression repositories, we found that SV2A gene expression was significantly elevated. Given that several SV2A specific radiotracers (e.g., 11 C-UCB-A [46], 11 C-UCB-J [47], 18 F-UCB-H [48], and 18 F-SDM-8/Syn-VesT-1/2 [49,50]) have been reported for synaptic density PET imaging of neurodegenerative diseases [51,52], we performed a PET imaging evaluation in an NEPC xenograft model using 18 F-SynVesT-1, after having confirmed SV2A protein expression in NEPC cell lines and mouse xenografts through a comparative study.

SV2A Gene Expression in Clinical Tumor Tissues: NEPC vs. CRPC-Adeno
Our bioinformatic analyses were performed on combined data from three sources: a recent report by Beltran et al. [3], WCMC11/14, and SU2C 2015. We found that the normalized RNA expression of SV2A is significantly higher in NEPC (n = 20) than that in CRPC-Adeno (n = 147) ( Figure 1A, p = 0.0034). Similar results were observed in another dataset (GSE104786, Supplementary Figure S1A). In addition, we observed an intriguing trend of SV2A expression in a patient-derived xenograft (PDX) model (GSE59986, Figure  1B) as AdPC was progressing to NEPC after castration; the SV2A expression maintained at a relatively stable and low level within 12 weeks post-castration until the emerging of the terminally differentiated NEPC, resulting in a drastic increase. Furthermore, we found that the frequency of SV2A gene amplification in mCRPC and NEPC samples (54 out of 96 or 56%, dataset by Beltran et al. [3]) was markedly higher than that in AdPC samples (8 out of 499 or 1.6%, TCGA Firehose Legacy dataset) ( Figure 1C). As anticipated, NEPC samples also present higher expression levels of other neuronal markers in addition to SV2A than CRPC-Adeno samples (Supplementary Figure S1B,C). Taken together, SV2A expression is significantly elevated in NEPC and we reason it may serve as a potential target to develop efficacious theranostics for NEPC and other innervated cancers.

SV2A Protein Expression in PCa Cell Lines and Patient Tumor Specimens
Prompted by the encouraging results from the bioinformatic analyses, we did a comparative assay of SV2A protein expression in multiple PCa cell lines, including non-NEPC cell lines (LNCaP and 22Rv1) and NEPC lines (DU145, PC-3, NCI-H660). As expected, all NEPC cell lines exhibited markedly higher SV2A expressions than the non-NEPC cell lines (Figure 2A,B). Specifically, NCI-H660 had about a 23-fold higher SV2A expression than 22Rv1. Of note, SV2A is detectable in AdPC cell lines, likely indicating the early existence of NED and the proneness to further NED progression upon treatment. Indeed, when 22Rv1 cells were maintained under hypoxia for a few days, neurite-like structures became to emerge along with the elevated NED marker of neuron-specific enolase (NSE) [53]. In addition, 22Rv1 xenografts were found expressing NED markers within highly hypoxic tumorigenic regions [54].

SV2A Protein Expression in PCa Cell Lines and Patient Tumor Specimens
Prompted by the encouraging results from the bioinformatic analyses, we did a comparative assay of SV2A protein expression in multiple PCa cell lines, including non-NEPC cell lines (LNCaP and 22Rv1) and NEPC lines (DU145, PC-3, NCI-H660). As expected, all NEPC cell lines exhibited markedly higher SV2A expressions than the non-NEPC cell lines (Figure 2A,B). Specifically, NCI-H660 had about a 23-fold higher SV2A expression than 22Rv1. Of note, SV2A is detectable in AdPC cell lines, likely indicating the early existence of NED and the proneness to further NED progression upon treatment. Indeed, when 22Rv1 cells were maintained under hypoxia for a few days, neurite-like structures became to emerge along with the elevated NED marker of neuron-specific enolase (NSE) [53]. In addition, 22Rv1 xenografts were found expressing NED markers within highly hypoxic tumorigenic regions [54]. We were able to obtain and analyze a limited number of NEPC (n = 3) and CRPC (n = 7) patient tissue samples ( Figure 2C,D). Similar to SYP and CgA, SV2A showed strong immunohistochemistry (IHC) staining in NEPC samples, but with variations in CRPC samples. In contrast, it is noteworthy that the expression of SSTR2, the molecular target of 68 Ga-DOTATATE, showed expressions in CPRC but became absent in NEPC.

SV2A Localization in PCa Cells
To determine the location of SV2A in PCa cells, we extracted cell membrane and cytoplasmic fractions from NCI-H660 cells. Immunostaining of subcellular fractionation demonstrated that SV2A predominately presents in the membrane rather than in the cytoplasm ( Figure 3A). Of note, we observed that the membrane-bound SV2A isoform possesses a higher molecular weight than its cytoplasm counterpart, perhaps indicative of its higher extent of post-translational modifications. Given that the oncological version of SV2A-targeted agents likely lacks the desired cell membrane permeability for the brain imaging agent, 18 F-SynVesT-1, clearly, these results further support that SV2A protein can be used for the development of targeted theranostics of NEPC. We were able to obtain and analyze a limited number of NEPC (n = 3) and CRPC (n = 7) patient tissue samples ( Figure 2C,D). Similar to SYP and CgA, SV2A showed strong immunohistochemistry (IHC) staining in NEPC samples, but with variations in CRPC samples. In contrast, it is noteworthy that the expression of SSTR2, the molecular target of 68 Ga-DOTATATE, showed expressions in CPRC but became absent in NEPC.

SV2A Localization in PCa Cells
To determine the location of SV2A in PCa cells, we extracted cell membrane and cytoplasmic fractions from NCI-H660 cells. Immunostaining of subcellular fractionation demonstrated that SV2A predominately presents in the membrane rather than in the cytoplasm ( Figure 3A). Of note, we observed that the membrane-bound SV2A isoform possesses a higher molecular weight than its cytoplasm counterpart, perhaps indicative of its higher extent of post-translational modifications. Given that the oncological version of SV2A-targeted agents likely lacks the desired cell membrane permeability for the brain imaging agent, 18 F-SynVesT-1, clearly, these results further support that SV2A protein can be used for the development of targeted theranostics of NEPC.

Automated Radiosynthesis of 18 F-SynVesT-1 on GE TRACERlab FX-N Synthesizer
Our previously reported radiosynthesis of 18 F-SynVesT-1 was carried out in a semiautomated manner with ~19% of isolated yields (decay-uncorrected), in which the 18 F-fluoride was loaded onto and eluted off a PS-HCO3 cartridge using a reversed flow elution [49]. Our initial attempts to fully automate the radiosynthesis on a GE TRACERlab FX-N synthesizer resulted in low yields (~5%) due to the low recovery (~35% vs. >90% semiautomated) of 18 F-fluoride from the PS-HCO3 cartridge. In addition, it was hard to implement the reversed flow elution on the synthesizer without extra hardware modifications. Therefore, we applied a low-vacuum reaction vessel (~40 kPa pressure) to slowly drain the elution mixture through the cartridge. This modification allowed the low amount of K2CO3 solution to thoroughly interact with the anion-exchange resin, thus reproducibly improving the 18 F-fluoride recovery over 95% (>3 runs) for the following radiolabeling procedures. The automated radiosynthesis of 18 F-SynVesT-1 afforded ~ 15% of the isolated yield (decay-uncorrected; n > 3) with > 99% of radiochemical purity, similar to what we previously reported [49]. The total automated radiosynthesis time was 95-100 min.

PET Imaging of SV2A with 18 F-SynVesT-1 in NEPC Xenograft Model
Given the significantly elevated SV2A expression in NEPC cells and xenografts, we performed a pilot small animal PET/CT imaging evaluation with 18 F-SynVesT-1 in NOD-SCID mice bearing NCI-H660 xenografts. We acknowledge that although 18 F-SynVesT-1 has entered clinical trials [41,51], it is a neuroimaging agent that targets SV2A in the brain for synaptic density assessment but not in NEPC for oncological imaging. Despite the undesired high hepatic accumulation due to its lipophilic nature, 18 F-SynVesT-1 was able to clearly visualize the SV2A + NCI-H660 tumor within 1 h p.i. (Figures 4A and Supplementary Figure S2A). The quantitative data analysis of the dynamic PET imaging displayed that the uptake of 18 F-SynVesT-1 in the tumor was consistently higher than that in the muscle, starting from 10 min p.i. It maintained a stable and increasing trend within 60 min p.i., peaked at 0.70 ± 0.14 %ID/g in between 50-60 min p.i., and then gradually decreased to 0.29 ± 0.04 %ID/g at 4 h p.i. (Figure 4B and Supplementary Figures S3A,B). Within the 4 h of imaging, the uptake of 18 F-SynVesT-1 stayed virtually unchanged in the muscle. As such, the highest tumor-to-muscle ratio was 1.47, observed in between 50-60 min p.i. (Supplementary Figure S3C). To confirm the SV2A-specific uptake of 18 F-SynVesT-1 in the NCI-H660 tumor, we performed the blocking study with a co-injection of 19 F-SynVesT-1 (0.1 mg, 0.033 µmol). The uptake of 18 F-SynVesT-1 in NCI-H660 tumor was significantly reduced from 0.70 ± 0.14 %ID/g to 0.25 ± 0.03 %ID/g at 50-60 min p.i. (p = 0.025), the same level as in the muscle. At the same time, the brain uptake also showed a drastic decrease (p < 0.01) ( Figure 4D), indicating the successful blockade of SV2A-specific uptake and the SV2A-specific binding of 18 F-SynVesT-1 in the NCI-H660 tumor. As anticipated, 18 F-Syn-VesT-1 showed high accumulation in both liver and brain (Figures 4C and Supplementary

Automated Radiosynthesis of 18 F-SynVesT-1 on GE TRACERlab FX-N Synthesizer
Our previously reported radiosynthesis of 18 F-SynVesT-1 was carried out in a semiautomated manner with~19% of isolated yields (decay-uncorrected), in which the 18 Ffluoride was loaded onto and eluted off a PS-HCO 3 cartridge using a reversed flow elution [49]. Our initial attempts to fully automate the radiosynthesis on a GE TRAC-ERlab FX-N synthesizer resulted in low yields (~5%) due to the low recovery (~35% vs. >90% semi-automated) of 18 F-fluoride from the PS-HCO 3 cartridge. In addition, it was hard to implement the reversed flow elution on the synthesizer without extra hardware modifications. Therefore, we applied a low-vacuum reaction vessel (~40 kPa pressure) to slowly drain the elution mixture through the cartridge. This modification allowed the low amount of K 2 CO 3 solution to thoroughly interact with the anion-exchange resin, thus reproducibly improving the 18 F-fluoride recovery over 95% (>3 runs) for the following radiolabeling procedures. The automated radiosynthesis of 18 F-SynVesT-1 afforded~15% of the isolated yield (decay-uncorrected; n > 3) with > 99% of radiochemical purity, similar to what we previously reported [49]. The total automated radiosynthesis time was 95-100 min.

PET Imaging of SV2A with 18 F-SynVesT-1 in NEPC Xenograft Model
Given the significantly elevated SV2A expression in NEPC cells and xenografts, we performed a pilot small animal PET/CT imaging evaluation with 18 F-SynVesT-1 in NOD-SCID mice bearing NCI-H660 xenografts. We acknowledge that although 18 F-SynVesT-1 has entered clinical trials [41,51], it is a neuroimaging agent that targets SV2A in the brain for synaptic density assessment but not in NEPC for oncological imaging. Despite the undesired high hepatic accumulation due to its lipophilic nature, 18 F-SynVesT-1 was able to clearly visualize the SV2A + NCI-H660 tumor within 1 h p.i. (Figure 4A and Supplementary Figure S2A). The quantitative data analysis of the dynamic PET imaging displayed that the uptake of 18 F-SynVesT-1 in the tumor was consistently higher than that in the muscle, starting from 10 min p.i. It maintained a stable and increasing trend within 60 min p.i., peaked at 0.70 ± 0.14 %ID/g in between 50-60 min p.i., and then gradually decreased to 0.29 ± 0.04 %ID/g at 4 h p.i. (Figure 4B and Supplementary Figure S3A,B). Within the 4 h of imaging, the uptake of 18 F-SynVesT-1 stayed virtually unchanged in the muscle. As such, the highest tumor-to-muscle ratio was 1.47, observed in between 50-60 min p.i. (Supplementary Figure S3C). To confirm the SV2A-specific uptake of 18 F-SynVesT-1 in the NCI-H660 tumor, we performed the blocking study with a co-injection of 19 F-SynVesT-1 (0.1 mg, 0.033 µmol). The uptake of 18 F-SynVesT-1 in NCI-H660 tumor was significantly reduced from 0.70 ± 0.14 %ID/g to 0.25 ± 0.03 %ID/g at 50-60 min p.i. (p = 0.025), the same level as in the muscle. At the same time, the brain uptake also showed a drastic decrease (p < 0.01) ( Figure 4D), indicating the successful blockade of SV2A-specific uptake and the SV2A-specific binding of 18 F-SynVesT-1 in the NCI-H660 tumor. As anticipated, 18 F-SynVesT-1 showed high accumulation in both liver and brain ( Figure 4C and Supplementary Figure S2A). The partial renal clearance of 18 F-SynVesT-1 was reflected by the increasing uptake in the bladder (Supplementary Figure S2A). Figure S2A). The partial renal clearance of 18 F-SynVesT-1 was reflected by the increasing uptake in the bladder (Supplementary Figure S2A). To demonstrate the diagnostic value of PET imaging of SV2A for noninvasive assessment of NED in NEPC, we carried out another set of small animal PET/CT imaging evaluations with 18 F-SynVesT-1, in direct comparison with 68 Ga-PSMA-11 (targeting PSMA) and 68 Ga-DOTATATE (strong binding to SSTR2 [55]) in NOD-SCID mice bearing the NEPC DU145 xenografts. Similar to the observations in the NCI-H660 xenograft model, DU145 tumors were clearly visualized by 18 F-SynVesT-1 ( Figure 5A), with the uptake at 1.17 ± 0.36 %ID/g (10-30 min p.i.) and 1.48 ± 0.46 %ID/g (40-60 min p.i.) ( Figure 5B). In contrast, neither 68 Ga-PSMA-11 nor 68 Ga-DOTATATE displayed meaningful tumor-tobackground uptake ratios ( Figure 5A). Although their tumor uptake values (0.80 ± 0.23 %ID/g and 1.08 ± 0.16 %ID/g, respectively) were at a similar level to that of 18 F-SynVesT-1 within 10-30 min p.i., a significant reduction was observed for both agents (p ≤ 0.01) within 40-60 min p.i. It indicated the retention of 18 F-SynVesT-1 was SV2A-specific in the tumors, while the tumor accumulation of 68 Ga-PSMA-11 and 68 Ga-DOTATATE was not related to NED ( Figure 5B). Immediately after the imaging, the DU145 tumors were harvested for IHC staining of SV2A, PSMA, and SSTR2. Shown in Figure 5C, the tumors displayed clearly positive staining of SV2A but not of PSMA or SSTR2, which further validates the PET imaging observations ( Figure 5A,B). To demonstrate the diagnostic value of PET imaging of SV2A for noninvasive assessment of NED in NEPC, we carried out another set of small animal PET/CT imaging evaluations with 18 F-SynVesT-1, in direct comparison with 68 Ga-PSMA-11 (targeting PSMA) and 68 Ga-DOTATATE (strong binding to SSTR2 [55]) in NOD-SCID mice bearing the NEPC DU145 xenografts. Similar to the observations in the NCI-H660 xenograft model, DU145 tumors were clearly visualized by 18 F-SynVesT-1 ( Figure 5A), with the uptake at 1.17 ± 0.36 %ID/g (10-30 min p.i.) and 1.48 ± 0.46 %ID/g (40-60 min p.i.) ( Figure 5B). In contrast, neither 68 Ga-PSMA-11 nor 68 Ga-DOTATATE displayed meaningful tumor-to-background uptake ratios ( Figure 5A). Although their tumor uptake values (0.80 ± 0.23 %ID/g and 1.08 ± 0.16 %ID/g, respectively) were at a similar level to that of 18 F-SynVesT-1 within 10-30 min p.i., a significant reduction was observed for both agents (p ≤ 0.01) within 40-60 min p.i. It indicated the retention of 18 F-SynVesT-1 was SV2A-specific in the tumors, while the tumor accumulation of 68 Ga-PSMA-11 and 68 Ga-DOTATATE was not related to NED ( Figure 5B). Immediately after the imaging, the DU145 tumors were harvested for IHC staining of SV2A, PSMA, and SSTR2. Shown in Figure 5C, the tumors displayed clearly positive staining of SV2A but not of PSMA or SSTR2, which further validates the PET imaging observations ( Figure 5A,B).

Discussion
The extent of NED in PCa has been recognized, with an essential role in the prognosis of patients with PCa [56]. The lack of an imaging method for noninvasive assessment of NED in PCa and other cancer types represents not only a major unmet clinical need in the current cancer patient care but also an opportunity to leverage this biomarker for developing novel cancer theranostics.
Given the clinical rarity of NEPC specimens and our experience in SV2A imaging, we started our work with bioinformatic analyses, using the reported PCa databases where SV2A had been profiled but not analyzed in detail (Figure 1). It is noteworthy that there were only 20 NEPC specimens available in the analysis and their gene expression levels of SV2A varied in a wide range. However, we found that SV2A gene expression is indeed upregulated in CRPC and further amplified in NEPC. As such, we performed a screening for SV2A in a variety of PCa cell lines, non-NEPC and NEPC (Figure 2). Not surprisingly, we observed that higher SV2A expressions are in association with NEPC cell lines.
Earlier studies on NETs and neuroblastoma [40,43] reported that the higher molecular weight isoform of SV2A was presented in the cytoplasm. However, in the PCa cell lines assayed in this work, we found that it is predominately cell membrane-bound instead (Figure 3), which is of advantage when considering SV2A as a target for the design and development of cancer theranostics.
To provide a proof of concept, we performed an imaging evaluation of SV2A using 18 F-SynVesT-1 in NEPC xenograft mouse models established by NCI-H660 and DU145. Even with the neuroimaging agent that targets SV2A in the brain for synaptic density

Discussion
The extent of NED in PCa has been recognized, with an essential role in the prognosis of patients with PCa [56]. The lack of an imaging method for noninvasive assessment of NED in PCa and other cancer types represents not only a major unmet clinical need in the current cancer patient care but also an opportunity to leverage this biomarker for developing novel cancer theranostics.
Given the clinical rarity of NEPC specimens and our experience in SV2A imaging, we started our work with bioinformatic analyses, using the reported PCa databases where SV2A had been profiled but not analyzed in detail (Figure 1). It is noteworthy that there were only 20 NEPC specimens available in the analysis and their gene expression levels of SV2A varied in a wide range. However, we found that SV2A gene expression is indeed upregulated in CRPC and further amplified in NEPC. As such, we performed a screening for SV2A in a variety of PCa cell lines, non-NEPC and NEPC (Figure 2). Not surprisingly, we observed that higher SV2A expressions are in association with NEPC cell lines.
Earlier studies on NETs and neuroblastoma [40,43] reported that the higher molecular weight isoform of SV2A was presented in the cytoplasm. However, in the PCa cell lines assayed in this work, we found that it is predominately cell membrane-bound instead (Figure 3), which is of advantage when considering SV2A as a target for the design and development of cancer theranostics.
To provide a proof of concept, we performed an imaging evaluation of SV2A using 18 F-SynVesT-1 in NEPC xenograft mouse models established by NCI-H660 and DU145. Even with the neuroimaging agent that targets SV2A in the brain for synaptic density assessment but not SV2A in NEPC for oncological imaging [41,51], we were able to clearly visualize the SV2A + tumors and confirm the desired SV2A-imaging specificity (Figure 4) for noninvasive assessment of NED in NEPC or other innervated cancer types. Clinically, NEPC features reduced expressions of PSMA, while the expressions of SSTRs (1)(2)(3)(4)(5) in NEPC remain inconclusive. Indeed, the results of our work presented herein demonstrated the limited diagnostic value of PET imaging with 68 Ga-PSMA-11 or 68 Ga-DOTATATE for the detection of NED in NEPC.
Because of the lipophilic nature of 18 F-SynVesT-1, which is required for a neuroimaging agent to cross the blood-brain barrier (BBB), we observed its high uptake and retention in the brain, liver, and intestines. The high accumulation of the agent in the abdominal region is of less concern for brain imaging but it presents a likely insurmountable obstacle for PCa imaging. Of note, the high brain uptake may also present a challenge for neurooncological imaging of SV2A due to the inevitable high presence of SV2A in the brain. We also observed an appreciable level of renal excretion of 18 F-activity, likely resulting from the less lipophilic metabolites of 18 F-SynVesT-1. These metabolites might not be able to cross the BBB to adversely affect the brain imaging, but they may complicate the image interpretation and data quantification when using the agent for cancer imaging, particularly for PCa. Therefore, although our proof-of-concept data is promising, we must re-design the SV2A-targeted brain imaging probes for oncological applications.
While maintaining the high affinity and specificity to SV2A, the oncological SV2A agents must have a substantially different in vivo kinetic distribution profile from their neuroimaging counterparts. For neuroimaging, in general, a lipophilicity measured by logP from 2 to 4 is required for optimal passive brain entry [57][58][59]. The measured logP value of 18 F-SynVesT-1 (2.32) is right in the range [49]. Due to the lipophilicity, the plasma free fraction (f p) of 18 F-SynVesT-1 is~43 ± 2% [49], which means~57% of the radiotracer stays protein-bound in the blood. For oncological applications, the SV2A targeting agents will have to be modified to be hydrophilic so that their f p values can be high for effective cell membrane SV2A binding and rapid renal clearance.
Although this work focuses on providing a proof-of-concept for SV2A-targeted PET imaging of NED in PCa, it can certainly be extended to other innervated cancer types beyond PCa. For instance, somatostatin receptor-based PET imaging is routinely practiced in NETs [60], but it is most useful in well-differentiated gastroenteropancreatic NETs (GEP-NETs) [61][62][63][64]. In contrast, SV2A has been reported with high expressions in a broad spectrum of cancer types with NED [41,43,65]. Therefore, SV2A can be potentially used to develop early detection methods for NED, which will provide new insights into the diagnosis and treatment of cancers with NED features, and theranostic agents for innervated cancers.

General
All reagents and solvents were purchased from commercially available sources and used as received unless otherwise stated. All aqueous solutions were prepared with Milli-Q water (18 MΩ-cm), which was obtained from a Millipore Gradient Milli-Q water system.

Bioinformatic Analyses Data Source
The SV2A gene expression data in NEPC patient tumors (WCMC11/14, SU2C 2-15, and Beltran et al. [3]) were downloaded from the cBioPortal database (http://www. cbioportal.org, accessed on 19 February 2019). The data from "The Expression data from Neuroendocrine Prostate Cancer and Primary Small Cell Prostatic Carcinoma" (GSE104786) and NEPC patient-derived xenograft (PDX) study (GSE59986) were downloaded from the Gene Expression Omnibus repository (https://www.ncbi.nlm.nih.gov/gds, accessed on 31 January 2020). The frequency of increased SV2A gene copy number was analyzed by the built-in analysis tool in the cBioPortal website for the comparative assay of two sets of data: the NEPC data from the cBioPortal database vs. the PCa data from the Cancer Genome Atlas (TCGA) database, linked through the cBioPortal website. Prism 8.0 (GraphPad Inc., San Diego, CA, USA) was used to run statistical analyses and plot the graphs.

Cell Culture and Animal Models
All animal studies were performed in accordance with relevant guidelines and regulations through an animal protocol (APN: 2020-102858; effective from 26 May 2020 to 26 May 2023) approved by the Institutional Animal Care and Use Committee (IACUC) at UT Southwestern, which adheres to the ARRIVE guidelines. Human LNCaP, 22Rv1, PC-3, DU145, and NCI-H660 PCa cell lines were obtained from the American Type Culture Collection (ATCC). The NCI-H660 cells were cultured in a complete growth medium according to the culture method supplied by ATCC. 22Rv1 cells were cultured in DMEM media with 10% fetal bovine serum (FBS) and 1% Penicillin/Streptomycin (PS). Other cells were cultured in RPMI media with 10% FBS and 1% PS. All cells were cultured in a humidified chamber at 37 • C with 5% CO 2 .

Determination of the SV2A Location in PCa Cells
Membrane and cytoplasmic fractions were extracted using a subcellular protein fractionation kit for cultured cells (78840, Thermo Scientific, Waltham, MA, USA) per manufacturer's protocol. The fractions were then subjected to gel electrophoresis and blotted. Anti-SV2A antibody (ab32942, Abcam, Waltham, MA, USA) was used at a dilution of 1:1000, and Anti-rabbit IgG, HRP-linked Antibody, Cell Signaling Technology #7074, was used as the secondary antibody; GAPDH antibody (sc-166574, Santa Cruz Biotechnology, Inc., Dallas, TX, USA) was used at a dilution of 1:2000, and Anti-mouse IgG, HRP-linked Antibody (#7076, Cell Signaling Technology, Danvers, MA, USA) was used as the secondary antibody; and NaK ATPase subunit 1 antibody (3010S, Cell Signaling Technology, Danvers, MA, USA) was used at a concentration of 1:500. Specificity for membrane fractionation was evaluated by blotting for NaK ATPase subunit 1, while specificity for cytoplasmic fraction-ation was evaluated by blotting for GAPDH. Unstained gel and membrane were imaged with the ChemiDoc MP Imaging system (BioRad Laboratories, Hercules, CA, USA). Protein quantification was done via analyzing relative intensities of each band using ImageJ. Both the top membrane-specific SV2A and the total SV2A protein were normalized to protein levels in the unstained gel. The quantification was assessed via densitometry analysis.

Automated Radiosynthesis of 18 F-SynVesT-1
The preparation of 18 F-SynVesT-1 was conducted in a GE TRACERlab FX-N synthesis module (GE Healthcare, Waukesha, WI, USA) with a modification from the reported procedure [49]. Briefly, the cyclotron-produced 18 F-fluoride was trapped in a pre-conditioned Chromafix PS-HCO 3 cartridge, through which a mixture of KOTf in Milli-Q water (10 mg/mL, 0.45 mL), K 2 CO 3 in Milli-Q water (1 mg/mL, 50 µL), and MeCN (0.5 mL), was pulled into a low vacuum reaction vessel (~40 kPa pressure to start the drainage). The following radiochemical procedures were performed as reported [49].

Small Animal PET/CT Imaging
The imaging study was started when the tumor size reached 200-500 mm 3 on a Inveon PET/CT Multimodality System (Siemens Medical Solutions USA, Inc.,Knoxville, TN, USA). Followed by CT data acquisition, which was conducted at 80 kV and 500 µA with a focal spot of 58 µm, a dynamic scan (0-60 min) was performed immediately after intravenous injection of~3.5 MBq of 18 F-SynVesT-1 in 100 µL saline containing < 5% ethanol into each tumor-bearing mouse under anesthesia, with 2% isoflurane in oxygen. After the dynamic scan, the mouse was allowed to recover in a cage and then re-anesthetized for a 20 min scan at 2.5 and 4 h post-injection (p.i.) (at each time point, n = 3-4). Both CT and PET images were reconstructed with the manufacturer's software. Reconstructed CT and PET images were fused for quantitative data analysis. Regions of interest (ROIs) were drawn as guided by CT and quantitatively expressed as percent injected dose per gram of tissue (%ID/g).

Statistical Analysis
Statistical analyses were performed using GraphPad Prism 8.0. A p-value less than 0.05 (unpaired t-test) was considered statistically significant. All quantitative data are presented as mean ± standard deviation.

Conclusions
In this work, high SV2A expressions were found in association with NED in CRPC and NEPC by a set of bioinformatics analyses performed on available databases. Further screening and assays of PCa cell lines confirmed the upregulated SV2A in the NEPC cell lines and that the high molecular weight SV2A isoform is predominately cell membranebound. The specific SV2A imaging with 18 F-SynVesT-1 in an NEPC xenograft model provided the proof of concept and laid the foundation for oncological applications of SV2A-targeted theranostic agents.

Conflicts of Interest:
The authors declare no conflict of interest.