Cytoplasmic Localization of Prostate-Specific Membrane Antigen Inhibitors May Confer Advantages for Targeted Cancer Therapies

Prostate cancer is the most common type of cancer in men in western societies and is one of the leading causes of cancer-related mortality (1, 2). Among prostate cancer biomarkers for imaging and therapy, the prostate-specific membrane antigen (PSMA) has proven to be an excellent target structure due to (i) its overexpression in prostate cancer, (ii) its absence or low expression rates in healthy tissue (3, 4), and (iii) increasing expression rates with tumor aggressiveness, androgen-independence, metastatic disease, and disease recurrence (4–9). PSMA is a transmembrane glycoprotein (100–120 kDa) with an extensive extracellular domain (amino acids 44–750), which undergoes clathrin-mediated internalization upon ligand binding (3, 10). A novel MXXXL motif of N-terminal amino acids is formed by the cytoplasmic tail–mediating PSMA internalization (11). In colocalization studies with internalized transferrin, PSMA was detected in the recycling endosomal compartment upon tracking with the monoclonal antibody mAB J591, which targets the extracellular domain of PSMA (11, 12).

Specific prostate cancer targeting has been successfully achieved by the development of peptidomimetic PSMA inhibitors. Radiolabeling turns these molecules into powerful tools in the diagnosis and therapy of prostate cancer. In diagnostics of all stages, the ⁶⁸Ga-labeled PSMA inhibitor Glu-urea-Lys(Ahx)-HBED-CC ([⁶⁸Ga]Ga-PSMA-11) has become the most widely used PET/CT imaging agent (13–19).

Particularly the metastatic, castration-resistant prostate cancer (mCRPCa) represents a major therapeutic challenge as treatment options are still limited. In therapy of mCRPCa, alpha- or beta-emitter–radiolabeled PSMA inhibitors (e.g., PSMA-617, PSMA-I&T) have been introduced as a treatment alternative and first studies have revealed a high efficacy with a favorable safety profile (20, 21).

The recently developed dual-modality PSMA inhibitors feature two reporter entities (radioactive and/or fluorescent) enabling both preoperative imaging and subsequent intraoperative (radio- or fluorescence-) guidance (22, 23). This approach guarantees the precise detection and resection of malignant tissue to the best possible extent, directly affecting treatment outcome and patient survival. The new class of dual-labeled peptidomimetic PSMA inhibitors paves the way for promising new strategies in the diagnosis and therapy of prostate cancer and several of these molecules are currently in preparation for clinical translation (22, 23).

However, the detailed internalization mechanism and the subcellular distribution of PSMA inhibitors are still unknown. In particular, their intracellular fate is of great clinical interest and crucial for obtaining a detailed understanding of the mechanism of action during endoradiotherapy. Stimulated emission depletion (STED) nanoscopy (24, 25) provides the spatial resolution to follow PSMA inhibitor trafficking at the nanoscale. Here, we elucidate the internalization mechanism of PSMA inhibitors and determine their subcellular distribution on the molecular level using STED nanoscopy. For visualizing the intracellular distribution patterns of PSMA inhibitors in relation to PSMA, we developed a novel nonstandard live cell immunofluorescence (IF) staining protocol. To the best of our knowledge, this study is the first to investigate and reveal the subcellular fate of PSMA/PSMA inhibitor complexes.

The STED-compatible dual-labeled PSMA inhibitors Glu-urea-Lys-HBED-CC-PEG₂-STAR RED and Glu-urea-Lys-HBED-CC-PEG₂-STAR 635P (hereinafter referred to as Glu-urea-Lys-HBED-CC-<dye> for <dye> conjugates) were synthesized according to previous procedures (22). Radioisotope labeling with ⁶⁸Ga and determination of fluorescence properties were conducted according to previously established protocols. Details on synthesis, radiolabeling, and fluorescence spectroscopy are provided in Supplementary Methods.

For cultivation of PSMA-positive androgen-sensitive human prostate adenocarcinoma cells (LNCaP, ATCC CRL-1740, RRID:CVCL_1379, high PSMA expression; 22Rv1, ATCC CRL-2505, RRID: CVCL_1045, moderate/heterogeneous PSMA expression) and PSMA-negative androgen nonreliant human prostate adenocarcinoma cells (PC-3, ATCC CRL-1435, RRID:CVCL_0035), RPMI medium was enriched with 10% FCS and 2 mmol/L l-glutamine (all from PAA). Cells were grown at 37°C in humidified air with 5% CO₂ and were harvested using trypsin-ethylenediaminetetraacetic acid (trypsin-EDTA, 0.25% trypsin, 0.02% EDTA, Gibco, Cat#25200056; <20 passages between thawing and experimental use). Cell line authentication and Mycoplasma testing is regularly performed. Authentication of the LNCaP and PC-3 cell lines was confirmed on 03/06/2020. The 22Rv1 cell line was directly obtained from the ATCC (purchased on 05/29/2020; Lot: 2215512).

Cell binding and internalization experiments were performed as previously described (13). Potential cytotoxicity was assessed by analyzing the duration and frequency of cell division via holographic time-lapse imaging with a HoloMonitor M4 cytometer (PHI AB). Details to cell binding, internalization, and cytotoxicity experiments are provided in Supplementary Methods.

Biodistribution and PET imaging studies in LNCaP- and PC-3-tumor xenograft mice were performed according to established protocols (22). For cryosectioning, tissue was directly frozen 1 hour after tracer injection (details in Supplementary Methods). All experiments are complied with the current laws of the Federal Republic of Germany and were conducted according to German Animal Welfare guidelines and ARRIVE guidelines. The experiments were approved by the regional authorities Regierungspräsidium Karlsruhe and Freiburg (approval numbers G158/15, G18/04).

All confocal and STED data were acquired with a custom-built STED system close to the one published by Gorlitz and colleagues (details in Supplementary Methods; ref. 26).

For assessing the time and concentration dependence of PSMA inhibitor internalization, cells were incubated with 50/100/250/500 nmol/L Glu-urea-Lys-HBED-CC-STAR RED in RPMI for 5/15/30/45 minutes at room temperature. To verify specific uptake, cells were co-incubated with 2-PMPA (500 μmol/L). In addition, potential internalization of free dye was tested by incubating cells with STAR RED and STAR 635P carboxylic acid (Abberior, Cat#STRED-0001, Cat#ST635P-0001) for 1 hour at 37°C. After fixation with paraformaldehyde (2% PFA in PBS) for 10 minutes, the samples were mounted with ProLong Diamond Antifade Mountant containing DAPI (Thermo Fisher Scientific, Cat#P36966).

For live cell imaging and colocalization experiments, a novel non-standard live cell staining protocol exploited the blocking effect of low temperatures on PSMA internalization without impairing cell health. Briefly, cells were immunolabeled against PSMA and/or incubated with dual-labeled PSMA inhibitors on ice to specifically target the entire cell membrane-bound PSMA fraction. Internalization of the antibody/PSMA/PSMA inhibitor complex was triggered by temperature increase. For PSMA IF, a monoclonal anti-PSMA antibody (1:50, mouse IgG1, clone 107–1A4, Sigma-Aldrich, Cat#SAB4200257, RRID:AB_11129838) and a STAR 600-labeled goat anti-mouse-IgG antibody (1:50, Abberior, Cat#2–0002–010–5) were used. For colocalization experiments with clathrin, LNCaP cells were transiently transfected with the fusion construct SNAP-tag/clathrin light chain (SNAP-CLC) and stained with the live dye 610CP-BG (27, 28). Clathrin-mediated endocytosis was blocked via co-incubation with 30 μmol/L Pitstop2 (Sigma-Aldrich, Cat#SML1169). For endosomal colocalization experiments, LysoTracker Green DND-26 (Thermo Fisher Scientific, Cat#L7526) or SiR-lysosome (SPIROCHROME, Cat#SC012) were used. Details on microscopy sample preparation and colocalization experiments are provided in Supplementary Methods.

Background correction was done by subtracting at most 10% of the maximum fluorescence signal. Linear deconvolution (Wiener filter) was applied with a Lorentzian PSF (FWHM STED PSF 60 nm, FWHM confocal PSF 200 nm) and only to the extent that data are smoothed and noise is reduced but resolution is not increased. Detailed information on microscopy data analysis is provided in Supplementary Methods.

The binding affinity of Glu-urea-Lys-HBED-CC-STAR RED and -STAR 635P (0.0/0.5/1.0/2.5/5.0/10/25/50/100/500 nmol/L) to PSMA on LNCaP and 22Rv1 cells was determined as the mean effective concentration (EC₅₀) by flow cytometry (BD FACS Canto II Flow Cytometer, BD Biosciences). To demonstrate the specificity of the binding, all conjugates in concentrations of 0.5/5.0/50 nmol/L were additionally co-incubated with 500 μmol/L 2-PMPA. The pH dependence of the binding was analyzed for Glu-urea-Lys-HBED-CC-STAR RED, -STAR 635P and the antibody complex (primary antibody: monoclonal anti-PSMA, mouse IgG1, clone 107–1A4, Sigma-Aldrich, Cat#SAB4200257, RRID:AB_11129838; secondary antibodies: STAR 600- or STAR RED-labeled goat anti-mouse-IgG, Abberior, Cat#2–0002–010–5 or Cat#2–0002–011–2, RRID:AB_2810982). Details on the flow cytometry studies are provided in Supplementary Methods.

Experiments were performed at least in triplicate. Quantitative data in text are expressed as mean ± standard deviation (SD). Bar plots depict the mean ± SD of the measurements for replicate experiments. Box plots indicate the interquartile range (box), the outer-most data points falling within 1.5× interquartile range (whiskers), the median (center line) and the mean (triangle) of the measurements for replicate experiments. If applicable, means were compared using Student t test (GraphPad Prism Version 7, GraphPad Software, Inc.). The P values <0.05 were considered statistically significant.

STED-compatible dual-labeled PSMA inhibitors were derived from Glu-urea-Lys-HBED-CC-PEG₂-IRDye800CW, as the latter preclinically performs comparably with clinically established molecules (e.g., PSMA-11 or PSMA-617; refs. 13, 22, 29, 30). The analytical data of the final products Glu-urea-Lys-HBED-CC-PEG₂-STAR RED and Glu-urea-Lys-HBED-CC-PEG₂-STAR 635P are summarized in Supplementary Table S1, their structures are shown in Fig. 1A, and their fluorescence spectra are displayed in Supplementary Fig. S1A–S1C.

Incubation of LNCaP (high PSMA expression) and 22Rv1 (moderate/heterogeneous PSMA expression) cells with ⁶⁸Ga-labeled Glu-urea-Lys-HBED-CC-STAR RED and -STAR 635P resulted in significant and specific binding to PSMA with affinities in the low nanomolar range. In addition, both conjugates specifically internalized (Supplementary Table S2). The pronounced difference in PSMA expression levels (31) accounts for the reduced signal of 22Rv1 cells. This trend is likewise reflected in all following fluorescence imaging experiments.

To ascertain and elaborate on the radioactively determined in vitro data, cell binding and internalization was investigated in confocal microscopy experiments (Fig. 1B). All subsequent fluorescence microscopy data were acquired with Glu-urea-Lys-HBED-CC-STAR RED, as this conjugate displayed superior photo stability and contrast in the fluorescence imaging experiments. Confocal imaging detected an increasing accumulation of Glu-urea-Lys-HBED-CC-STAR RED in PSMA-positive cells independent of its concentration over time (Supplementary Fig. S2A and S2B). A blocking effect was not detected for the tested concentration range (50 to 500 nmol/L). Blocking studies in the presence of 500 μmol/L 2-PMPA, internalization experiments with the free Abberior dyes STAR RED and STAR 635P and with PSMA-negative PC-3 cells proved the specificity of PSMA binding and internalization (Supplementary Fig. S3A–S3D).

For confocal time-lapse experiments, LNCaP cells were incubated with Glu-urea-Lys-HBED-CC-STAR RED for 20 minutes on ice to prevent premature internalization, subsequently washed and immediately imaged. Over a period of 120 minutes, steady internalization was observed (Supplementary Fig. S3E and S3F, Supplementary Movie S1). No morphological changes or other signs of impaired cell viability were detected during the experiment. To rigorously exclude cytotoxicity by the PSMA inhibitor, the frequency and duration of cell division was assessed via holographic time-lapse imaging. Cell proliferation was followed in the presence of either Glu-urea-Lys-HBED-CC-STAR RED or -STAR 635P for 48 hours and was compared with the proliferation of untreated cells. The data show no evidence of cytotoxicity but a slightly accelerated LNCaP cell proliferation in the presence of PSMA inhibitors (Fig. 1C and D; Supplementary Figs. S4A and S4B and S5A and S5B, Supplementary Table S3, Supplementary Movies S2–S4).

The ability of Glu-urea-Lys-HBED-CC-STAR RED and -STAR 635P to specifically target PSMA in vivo was evaluated in biodistribution studies at 1 hour post injection (p.i.). Both conjugates performed similarly in a PSMA-expressing LNCaP tumor mouse xenograft model with a high tumor uptake of around 5%ID/g and a favorable organ distribution profile comparable with PSMA-11, the reference compound without dye (Fig. 2A; Supplementary Table S4; ref. 30). Confocal imaging of tumor and muscle tissue cryosections confirmed a PSMA-specific uptake and revealed a homogenous subcellular distribution pattern throughout the tumor cells already 1 hour p.i. (Fig. 2B). The specificity of the in vivo tumor uptake of Glu-urea-Lys-HBED-CC-STAR RED was further proven with PSMA-negative PC-3 tumor xenografts. The uptake in PSMA-negative tumor tissue was comparable with the uptake in muscle tissue (Supplementary Table S5).

In addition, small-animal PET imaging was performed with ⁶⁸Ga-labeled Glu-urea-Lys-HBED-CC-STAR RED. Selective tumor uptake in the LNCaP xenograft model accompanied by rapid clearance from off-target tissue resulted in a high imaging contrast at early time points. In PSMA-negative PC-3 xenografts, no measurable uptake of the conjugate was observed during the experiment, demonstrating high PSMA specificity (Fig. 2C). The corresponding time activity curves showed rapid clearance from muscle, liver and heart tissues, but continuous accumulation of the conjugate in kidneys and bladder confirming the renal elimination (Fig. 2D).

To resolve details of the PSMA distribution in the cell membrane, Glu-urea-Lys-HBED-CC-STAR RED bound PSMA was imaged with STED nanoscopy. A heterogeneous distribution of the fluorescence signal along the cell membrane was observed, suggesting areas of higher and lower PSMA density, which could not be visualized with confocal imaging. Within the patches of high PSMA density, round to oval-shaped PSMA clusters of various sizes were detected (Fig. 3A; Supplementary Fig. S6A).

After short internalization times, the PSMA inhibitor signal was localized to the membrane of cytoplasmic vesicles. Although confocal imaging visualized the vesicles as blurred diffraction-limited spots, STED imaging enabled the nanometer resolution of defined, hollow, spherical structures (Fig. 3B and C). Heterogeneous vesicle distribution was observed with regions of higher and lower vesicle density. Although the nucleus always remained free of any Glu-urea-Lys-HBED-CC-STAR RED signal, the overall concentration of vesicles in cellular filopodia increased particularly (Fig. 3D). Line profiles of vesicles derived from raw confocal and STED data confirm the improvement in resolution gained by STED. The full width half maxima (FWHM) of the Lorentzian fits suggest a spatial resolution significantly below the diffraction limit (Fig. 3C; Supplementary Fig. S6B–S6D; for post processing of STED data refer to Supplementary Fig. S7).

Live LNCaP cell confocal colocalization experiments between Glu-urea-Lys-HBED-CC-STAR RED and SNAP-tagged CLC labeled with 610CP (27, 28) showed a significant signal overlap for early time points that decreased with progressing internalization time (Fig. 3E; Supplementary Fig. S8A). The Pearson correlation coefficient (PCC; ref. 32) dropped significantly within the first 15 minutes of internalization, suggesting clathrin-mediated uptake with rapid clathrin uncoating after internalization (Fig. 3F; Supplementary Table S6), which enabled the fusion with early endosomes and thus vesicle growth (33, 34). Co-incubation with the clathrin inhibitor Pitstop2 confirmed clathrin-dependent PSMA inhibitor internalization (Supplementary Fig. S8B and S8C).

As clathrin-coated vesicles (CCV) feature a well-defined diameter, the diameter of endocytic vesicles is an indicator for the degree of progress along the endocytic pathway. It was analyzed by fitting a two-dimensional ring function to the fluorescence signal (Supplementary Fig. S6D). The average diameter does not differ significantly within the first hour of internalization. The large SDs indicate a broad spread of diameter in the overall vesicle population, especially for later times (Fig. 3G; Supplementary Table S7). These values are in the range of but, especially for LNCaP cells, always larger than the size of CCVs as published in the literature (35, 36).

For assessing the extent and duration of colocalization of the PSMA inhibitors with PSMA during and after the internalization process, incubation with Glu-urea-Lys-HBED-CC-STAR RED was combined with indirect immunolabeling of PSMA with the Abberior dye STAR 600 in a novel nonstandard live cell staining protocol. To avoid premature PSMA internalization, living cells were stained on ice and internalization times were well defined by subsequent incubation at 37°C (Fig. 4A; Supplementary Fig. S9). The PSMA inhibitor and the primary/secondary antibody complex were both internalized without any evidence for mutual blocking effects. The intracellular distribution of the PSMA inhibitor and PSMA was visualized for different internalization times by STED (LNCaP) and confocal (22Rv1) microscopy.

Because of biological heterogeneity, we rather state qualitative trends than quantitative numbers. Colocalization trends can be illustrated in pixel fluorograms (hereinafter referred to as fluorograms; for details, see Supplementary Methods; ref. 37). For visualizing the time-dependent changes in the colocalization of PSMA inhibitor and PSMA, we subdivided the fluorograms of all dual color STED/confocal images in three defined sections separating (i) the PSMA antibody signal only (top section), (ii) the colocalizing signal (middle section), and (iii) the PSMA inhibitor signal only (bottom section). The background (quarter circle in bottom left corner) was set to the average intensity of Glu-urea-Lys-HBED-CC-STAR RED after 24 hours of internalization (Fig. 4B).

For short internalization times, the PSMA inhibitor signal (red) and the PSMA antibody signal (cyan) strongly colocalized in one fraction (white) falling into the middle section of the fluorogram representing specific binding of the PSMA inhibitor to PSMA in the plasma and vesicle membrane (Fig. 4C; Supplementary Fig. S10). The corresponding PCC peaked at 30 minutes with r(LNCaP) = 0.65 ± 0.08 (N = 9) and r(22Rv1) = 0.70 ± 0.06 (N = 15). During the first hour of internalization, continuing colocalization of the signals at the membrane of cytoplasmic vesicles was detected with an average PCC of r(LNCaP) = 0.54±0.09 (N = 43) and r(22Rv1) = 0.65 ± 0.07 (N = 56).

However, around 45 minutes, the cyan fraction (PSMA antibody) started to increase significantly in the fluorogram's top section and the red fraction (PSMA inhibitor) started to slowly vanish into the background. Over time, the relative intensities of both signals in the foreground of the middle section of the fluorogram, and thus the colocalizing fraction, significantly declined. The relative intensity of the PSMA inhibitor in the entire foreground of the fluorogram (all three sections taken together) significantly decreased whereas it increased in the entire background. However, the relative intensity of PSMA changed neither in the entire foreground nor in the entire background (Fig. 4C; Supplementary Figs. S10 and S11A–S11E).

These trends illustrate the gradual release of the PSMA inhibitor from PSMA. The PSMA inhibitor signal initially distributed homogeneously in the vesicles (Fig. 4D) and finally dispersed in the cytoplasm after >3 hours of internalization (Fig. 4E and F; Supplementary Fig. S11F, Supplementary Table S8). At these times, the PSMA antibody signal was still detected at the vesicle membrane and additionally in dotted patches at the cell membrane, presumably representing recycled PSMA (see next subsection).

At later time points (6 and 24 hours), no significant further changes of the signal distribution were observed (Supplementary Fig. S11A–S11F) and the colocalization between PSMA inhibitor and PSMA was significantly reduced with an average PCC of r(LNCaP) = 0.25 ± 0.12 (N = 18) and r(22Rv1) = 0.29 ± 0.15 (N = 21). Photobleaching was found to be negligible as the ratio of the integrated fluorescence intensities did not drastically change over time with an average of 1.17 ± 0.35 (LNCaP N = 57, PSMA to PSMA inhibitor). Dye-dependent effects could be excluded by substantiating the results of Glu-urea-Lys-HBED-CC-STAR RED with Glu-urea-Lys-HBED-CC-STAR 635P and the results of PSMA STAR 600 IF with PSMA STAR RED IF (Supplementary Fig. S11G and S11H).

To obtain more information on the PSMA recycling pathway, endosomal colocalization experiments were carried out. Living cells were immunolabeled for cell membrane-bound PSMA on ice to prevent premature PSMA internalization. Subsequently, internalization was triggered by incubation at 37°C. After different internalization times, endosomal colocalization was assessed during live cell confocal imaging experiments by staining with LysoTracker Green DND-26, which is selective for both lysosomes and endosomes. With progressing internalization time, the PSMA signal increasingly overlapped with the endosomal signal and the respective PCC significantly climbed (Fig. 5A and B; Supplementary Table S9), suggesting endosomal recycling of PSMA. Non-colocalizing LysoTracker signal was additionally detected due to unlabeled PSMA and non–PSMA-carrying lysosomes (Supplementary Movie S5). Co-incubation with 2-PMPA, Glu-urea-Lys-HBED-CC-STAR RED or -STAR 635P did not significantly affect the degree of colocalization (Supplementary Table S9). PSMA recycling was further supported by the occurrence of clustered PSMA antibody signals at the cell membrane not colocalizing with PSMA inhibitor in confocal and STED imaging experiments after 45 minutes of internalization (Fig. 5C).

The results were confirmed in live confocal and live STED colocalization experiments by co-staining with SiR-lysosome, which also targets both lysosomes and endosomes (Supplementary Fig. S12A, Supplementary Movie S5). To exclude phototoxic influence, low light intensity confocal times-series of lysosomal dynamics were recorded after STED imaging. No qualitative difference in lysosomal movement was observed compared with control cells not exposed to high STED laser intensities (Supplementary Movies S6 and S7).

The homogeneous cytoplasmic distribution of Glu-urea-Lys-HBED-CC-STAR RED for internalization times >3 hours suggested a separation of the PSMA/PSMA inhibitor complex in the endosomes by decreasing pH. To test this hypothesis, the pH dependence of the binding properties of the PSMA antibody and the PSMA inhibitor to PSMA on LNCaP cells were assessed. In flow cytometry studies, the fractions bound at different physiologically relevant pH values in percentage of the fraction bound at pH 7.0 were measured via the detected fluorescence signal of the PSMA inhibitors or the secondary antibodies. The binding of the antibody complexes (primary monoclonal anti-PSMA antibody decorated with secondary antibody conjugated to Abberior STAR 600 or Abberior STAR RED) to PSMA was not significantly affected up to an acidic pH of 4.6. In contrast, the binding of Glu-urea-Lys-HBED-CC-STAR RED and -STAR 635P to PSMA was drastically reduced with increasing acidity (Fig. 5D, Supplementary Tables S10 and S11). Temperature and/or pH-dependent fluorophore degradation of the PSMA inhibitors could be excluded (Supplementary Fig. S12B and S12C).

PSMA-targeted imaging and therapy have refashioned the diagnosis and the treatment of patients with prostate cancer over the last years (14–16, 19). In particular, endoradiotherapy with alpha- or beta-emitter–radiolabeled PSMA inhibitors offers a promising treatment approach for patients with late-stage mCRPCa beyond established treatment options (20, 21).

With the development of dual-labeled PSMA inhibitors, an additional imaging modality has been successfully introduced to the clinical routine (22, 23, 38). Besides intraoperative fluorescence-guided surgery, the additional fluorescence property enables intracellular tracking of the dual-labeled inhibitor, making it a perfect model compound for analyzing the subcellular fate of peptidomimetic PSMA inhibitors at the nanoscale. To the best of our knowledge, the detailed subcellular fate of PSMA/PSMA inhibitor complexes for peptidomimetic PSMA inhibitors has not been investigated with fluorescence nanoscopy so far.

All existing knowledge on the mechanism of internalization and the intracellular distribution of PSMA is solely based on diffraction-limited confocal microscopy studies (11, 12). The herein achieved spatial resolution is indeed sufficient to follow the coarse localization of PSMA and its inhibitors inside cells over time. However, to precisely investigate localization subtleties (e.g., cluster substructures) and to elucidate in detail the subcellular fate of PSMA and PSMA inhibitor (e.g., molecular colocalization), STED nanoscopy is the method of choice, as it combines the required spatial resolution with inherently coaligned fluorescence imaging channels (excitation multiplexed, one STED donut; ref. 39).

Because this nanoscopy technique places special demands on the fluorophores used (40), we synthesized dual-labeled PSMA inhibitors based on PSMA-11 equipped with the STED-compatible Abberior dyes STAR RED and STAR 635P. Both conjugates exhibited high PSMA-specific binding affinities, specifically internalized fractions, specific tumor uptake in PSMA-positive lesions, and high tumor-to-background ratios at early time points comparable with the parental structure (22) and reference compounds without dye (13, 29). Hence, the additional fluorescence moiety does not affect the main characteristics of our conjugates (e.g., pharmacokinetic properties, tumor accumulation), making them suitable for mimicking clinically used PSMA inhibitors to study their intracellular fate with STED nanoscopy.

Confocal and STED imaging as a function of time permitted the visualization of the intracellular distribution of PSMA and of our PSMA inhibitors in PSMA-expressing LNCaP and 22Rv1 cells. The data revealed an increase of the internalized PSMA inhibitor fraction over time, independently of the concentrations tested. These results support the radioactively determined in vitro data acquired with the ⁶⁸Ga-labeled PSMA inhibitors in cell binding and internalization studies.

Long-term PSMA inhibitor internalization experiments using confocal and STED imaging indicate enrichment in cells. These findings are in line with the in vivo time course of tumor accumulation, which agrees with our observations of a long-term accumulation of our PSMA inhibitors in prostate cancer tumor cells with a high signal-to-background ratio. Nonspecific signal of unbound inhibitors decreased over time, whereas the signal of the specifically internalized fraction remained. Targeting of nontumor tissue was very limited, but even if off-target internalization were to occur, our holographic cytometry data does not indicate cytotoxic effects by the non-radiolabeled PSMA inhibitors.

STED nanoscopy allowed the visualization of structural subtleties of the cell membrane-bound PSMA fraction (assessed with both PSMA IF and PSMA inhibitor). The PSMA signal was not homogeneously distributed along the cell membrane, but rather spread into patches of oval- to round-shaped signals of different sizes, both before internalization and after PSMA recycling. We assume that these individual signals consist of varying numbers of PSMA molecules grouped in clusters at the cell membrane. PSMA clustering at the cell surface was previously suggested on the basis of data of biochemical assays, in which it was artificially induced with antibodies. PSMA clustering is crucial not only for PSMA activation and its subsequent internalization, but also for many downstream signaling pathways directly or indirectly affecting cell proliferation (41–43) potentially explaining the slightly accelerated LNCaP division frequency and duration we observed.

With STED nanoscopy, we visualized PSMA clustering at the cell surface for the first time, induced either by antibodies or by our PSMA inhibitors. During the LysoTracker colocalization experiments, we did not observe significant differences in endosomal colocalization in the presence or absence of our PSMA inhibitors, as PSMA clustering, activation, and internalization was in any case induced by our PSMA IF-labeling approach.

During the early phase of the internalization process, cytoplasmic vesicle formation was observed. Live cell colocalization experiments suggest that the PSMA-specific internalization of peptidomimetic PSMA inhibitors is mediated via clathrin-dependent endocytosis. As resolved by STED imaging, the average vesicle diameter is more than twice the value for CCVs found in literature (35, 36) but the large SD indicates a very heterogeneous vesicle population in our samples. The FWHMs of selected vesicle line profiles suggest a resolution good enough to resolve vesicles with approximately 100 nm in diameter. Thus, we assume that we mostly detected early endosomes, as CCVs quickly uncoat and fuse to become early endosomes with diameters larger than CCVs. We could indeed confirm that PSMA itself passes the endosomal pathway to be either degraded in lysosomes or recycled back to the cell surface as previously described (10, 11).

Furthermore, we developed a novel nonstandard live cell IF staining protocol to visualize the intracellular distribution of our PSMA inhibitors in relation to PSMA. In general, antibody staining is limited to fixed cells as antibodies are not cell membrane permeable. Here, we exploited the fact that PSMA internalization can be blocked by low temperature without impairing cell health. Thus, the entire cell membrane-bound PSMA fraction could be labeled by incubating the cells on ice. A subsequent temperature increase triggered the internalization of the antibody/PSMA/PSMA inhibitor complex (Supplementary Fig. S9). While indirect IF provides strong signal amplification, our PSMA inhibitors add only one fluorophore per PSMA molecule. Despite this difference in fluorophore stoichiometry, the brightness and signal-to-noise ratio of the PSMA inhibitor signal are only slightly reduced to the immunolabeled PSMA signal.

Interestingly, peptidomimetic PSMA inhibitors appear to distribute fundamentally differently on the subcellular level than PSMA antibodies used for prostate cancer immunotherapy. For short internalization times, strong colocalization was observed between PSMA and our PSMA inhibitors, but it decreased over time and eventually vanished at later time points. The PSMA/PSMA inhibitor complex dissociates during the PSMA recycling process in the endosomal compartment allowing the PSMA inhibitors initially to distribute in the endosomes but to disperse gradually homogeneously in the cytoplasm without binding to other structures. Immunolabeled PSMA remained in the vesicle membrane, eventually to be recycled back to the cell membrane. Rajasekaran and colleagues (11) previously reported on internalization and vesicle formation of antibody-targeted PSMA after 2 hours of incubation. Our findings match these results but additionally visualize PSMA localization at later time points up to 24 hours.

A dye-dependent effect could be excluded by interchanging the fluorescent labels for immunolabeled PSMA and PSMA inhibitor. It could additionally be shown that the spectral properties of our PSMA inhibitors were not sensitive to acidic pH or elevated temperature and, as unnatural inhibitors of PSMA, our conjugates are presumably intrinsically inert to lysosomal digestion.

Remarkably, significant differences in the binding affinity to PSMA at acidic pH were detected between the antibody and PSMA inhibitor. The interaction between PSMA and PSMA inhibitor decreased drastically at acidic pH, whereas binding of the anti-PSMA antibody/secondary antibody complex remained unaffected. The difference in pH sensitivity of PSMA binding is one plausible explanation for the fundamentally different intracellular distributions observed for the antibody and PSMA inhibitor. With decreasing endosomal pH, the PSMA-binding affinity of the PSMA inhibitors is reduced, shifting the equilibrium between bound and unbound PSMA inhibitor toward the unbound state. We now speculate that, at endosomal pH, the acidic moieties of PSMA inhibitors are mostly protonated, allowing for endosomal escape and cytoplasmic dispersion. Following this hypothesis, unbound PSMA inhibitor would be continuously removed from the PSMA bound/unbound collective inside the endosomes. By pushing the position of the equilibrium further toward the unbound state, eventually, the entire PSMA inhibitor population would be located inside the cytoplasm, a process simply driven by entropy. Finally, the cytoplasm's neutral pH is assumed to trap the PSMA inhibitors inside the cell by deprotonating their acidic moieties, thereby restoring the molecules' membrane impermeability. However, further studies on externalization and membrane permeability need to confirm this hypothesis.

The performance of prostate cancer–targeting inhibitors in diagnostics and therapy highly depends on their cell binding and internalization properties as well as on their subcellular localization. This study describes the intracellular distribution of PSMA and peptidomimetic PSMA inhibitors at the nanometer scale for the first time. STED fluorescence nanoscopy allows insights into the internalization and precise localization of both PSMA inhibitors and of PSMA itself at clinically relevant time points. As observed in this study, strong and specific internalization in combination with a long retention allows the inhibitors' enrichment in the target cells over time affording high imaging contrasts while sparing healthy tissue and preventing off-target tissue effects. The striking difference between the PSMA antibodies' and the PSMA inhibitors' subcellular fate will potentially have significant impact on the therapeutic efficiency of the peptidomimetic inhibitors. The homogeneous dispersion of PSMA inhibitors in prostate cancer cells after internalization, which we have demonstrated here for the first time, is of particular interest in endoradiotherapy, for which this dispersed cytoplasmic distribution potentially leads to beneficial effects. Intracellular accumulation and localization in nuclear proximity may allow a more target-oriented application of effectively higher local radiation doses—especially with high linear energy transfer alpha particles—resulting in potent DNA damage and subsequent apoptosis.

We expect our results to fuel further ligand development, not only in the field of prostate cancer. Our findings will hopefully boost the development of refined and/or new targeting strategies for diagnostic and therapeutic approaches in prostate cancer and other human cancer treatment.

U. Haberkorn reports a patent for 21-15PCT-improved 18F-tagged inhibitors of prostate-specific membrane antigen (PSMA) and their use as imaging agents for prostate cancer pending, 46-14PCT 18F–tagged inhibitors of prostate-specific membrane antigen (PSMA), their use as imaging agents and pharmaceutical agents for the treatment of prostate cancer issued, 46-16PCT prostate-specific membrane antigen (PSMA) ligands comprising an amylase cleavable linker pending, and 40-16PCT treatment of PMSA-expressing cancers issued. M. Eder reports a patent for WO2015055318 issued, US20150110715 issued, WO2017054907A1 issued, EP 17201264.3 pending, EP 17202711.2 pending, EP19157214.8 pending, and WO2020165409A1 issued. K. Kopka reports a patent for PSMA-617 licensed and with royalties paid from Endocyte, a Novartis company, as well as a patent for PSMA-1007 licensed and with royalties paid from ABX GmbH, Radeberg, Germany, and a patent for PSMA-914 pending, and member of scientific advisory board (SAB) of Telix Pharmaceuticals. A.C. Eder reports a patent for EP 17201264.3 pending, EP 17202711.2 pending, and EP 19157214.8 pending. No disclosures were reported by the other authors.

J. Matthias: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. J. Engelhardt: Conceptualization, formal analysis, methodology, writing–review and editing. M. Schäfer: Investigation. U. Bauder-Wüst: Investigation. P.T. Meyer: Resources, writing–review and editing. U. Haberkorn: Resources, writing–review and editing. M. Eder: Conceptualization, resources, funding acquisition, project administration, writing–review and editing. K. Kopka: Resources, funding acquisition, writing–review and editing. S.W. Hell: Resources, methodology, writing–review and editing. A.-C. Eder: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft.

We acknowledge funding from Helmholtz International Graduate School for Cancer Research PhD stipends for A.C. Eder and J. Matthias and support by the VIP+ grant VP00130, Federal Ministry of Education and Research (BMBF), Germany. We thank Karin Leotta, Ursula Schierbaum, and Mareike Roscher for support in biodistribution and small-animal imaging studies, Martin Seefeld for providing access to UV/Vis and fluorescence spectrophotometers, Alexey Butkevich and Vladimir Belov for providing the fluorescent live dye 610CP-BG, Francesca Bottanelli for providing the SNAP-CLC construct, Steffen J. Sahl and Jade Cottam Jones for helpful discussions, and Rifka Vlijm for support in data analysis.

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.