A prostate-specific membrane antigen activated molecular rotor for real-time fluorescence imaging

Surgery is an efficient way to treat localized prostate cancer (PCa), however, it is challenging to demarcate rapidly and accurately the tumor boundary intraoperatively, as existing tumor detection methods are seldom performed in real-time. To overcome those limitations, we develop a fluorescent molecular rotor that specifically targets the prostate-specific membrane antigen (PSMA), an established marker for PCa. The probes have picomolar affinity (IC50 = 63-118 pM) for PSMA and generate virtually instantaneous onset of robust fluorescent signal proportional to the concentration of the PSMA-probe complex. In vitro and ex vivo experiments using PCa cell lines and clinical samples, respectively, indicate the utility of the probe for biomedical applications, including real-time monitoring of endocytosis and tumor staging. Experiments performed in a PCa xenograft model reveal suitability of the probe for imaging applications in vivo.

Prostate cancer (PCa) is the second leading cause of cancer-related death in men. Patients with localized tumors can benefit greatly from radical prostatectomy1,2. Long-term control can be achieved if the tumor-positive margin is completely resected. However, that can be challenging unless there is a reliable way to visualize very small amounts of tumor that may still reside in the margin after resection. Currently, histopathology is the most reliable way to identify a positive margin, but it is time-consuming, and is not often uncovered until the post-operative period. An imaging agent that could reliably identify a positive margin intraoperatively may change the surgical plan in real-time and contribute to a better outcome.

Prostate-specific membrane antigen (PSMA) is a type II transmembrane glycoprotein specifically overexpressed in most PCa with limited expression in normal tissue3,4. It is regarded as a high-value biomarker for PCa and has received worldwide attention for PCa diagnosis and treatment. PSMA-targeting ligands have succeeded in guiding a variety of functional groups for such purposes, including radionuclides, fluorophores, photosensitizers, Gd (III) and so on5,6,7,8,9,10,11,12. With the development of high-quality near-infrared dyes, PSMA-targeted fluorescent ligands are being actively pursued5,13. Upon administration, a PSMA-targeted fluorescent dye can specifically retain within PCa, and after clearance of background signal from the blood pool, which normally takes several hours, PSMA-specific tumor imaging can be achieved, and may enable surgical guidance. To date optical probes targeting prostate cancer have largely been based on fluorescent dyes (Cy7, IRdye800CW, indocyanine green etc.) that are “always on”, requiring substantial time to achieve suitable image contrast14. PSMA-activatable fluorescent probes, which have an off-on response, will be desirable as they may have less background than fluorescent agents that are fluorescent upon excitation, but only a few examples have been reported15,16,17. Kobayashi et al. carried out pioneering studies based on the quenching effect when a fluorescent dye (such as indocyanine green) was conjugated with an antibody or minibody, demonstrating a 30-fold increase in fluorescence upon activation and internalization of the agent by PSMA15,16. Based on that activation effect, they could detect PSMA-positive tumors specifically at 6 h post-injection with minibodies conjugated with indocyanine green. However, the requirement of internalization for activation, and the long clearance times of monoclonal antibodies from the blood pool, prevent imaging shortly after administration. Urano et al. recently reported a novel fluorogenic method using the carboxypeptidase activity of PSMA. The method could be utilized for ex vivo fluorescence imaging of PCa in surgically resected clinical specimens17. The method enables high fluorescence enhancement, while the generation of signal depends on the enzymatic activity of PSMA, and requires minimum 30 min for sample staining. Overall, existing methods need at least tens of minutes to hours to reach suitable image contrast, encouraging us to develop probes that may enable imaging at shorter time post-administration.

Fluorescent molecular rotors (FMRs) are a family of fluorophores sensitive to local microenvironment (e.g., polarity and viscosity)18,19,20. Upon photoexcitation, the molecule can form a low-energy state, referred to as a twisted intramolecular charge transfer (TICT) state, so that the excitation energy can be dissipated accompanied by red-shifted emission or non-radiative relaxation. The strategy for applying FMRs is to restrict the formation of the TICT state, in which case specific fluorescence enhancement (usually quantum yield and fluorescent lifetime) can be obtained21. The fluorescence response of FMRs is more sensitive and faster compared with other on-off probes mediated by specific chemical reactions22, enabling real-time and in situ detection. In this regard, some FMRs have been developed for sensing viscosity in the microenvironment, such as derivatives of julolidine (DCVJ, CCVJ)23,24, meso-phenyl-substituted derivatives of BODIPY25,26,27, porphyrin derivatives28, and merocyanine dyes29. Increasingly, FMRs have been exploited for imaging biomolecules (protein, DNA, RNA, peptidoglycan) and biomolecular interactions30,31,32,33,34,35,36,37,38. We hypothesize that by conjugating to a low-molecular-weight PSMA inhibitor, the FMR can interact with residues lining the entrance funnel of the enzyme, whereupon activation will occur12,39 (Fig. 1a). Since that process is triggered simply by binding to PSMA and does not involve other steps, such as endocytosis and catalysis, it will be expected to respond rapidly and specifically. Here we report the discovery of FMR-based probes specifically activated by PSMA. We reveal the underlying mechanism of the fluorescent activation upon probe-PSMA binding. Furthermore, these probes are successfully applied to real-time monitoring of PSMA-mediated endocytosis, rapid prostate cancer tissue staining, and in vivo imaging of PCa.

a Schematic of probe binding and activation. b Structures, absorbance, and fluorescence spectra of PSMA activated fluorescent probes: Glu-490, ODAP-490, and ODAP-436. RFU relative fluorescence units. c Changes in fluorescence intensity of probes in solutions with increasing glycerol concentrations. Probe concentration was 0.1 mM and data were normalized to PBS control. Data are presented as mean values ± s.d. (n = 3 biologically independent experiments). d Inhibition of PSMA enzymatic activity using the radioenzymatic assay. Data are presented as mean values ± s.d. (n = 2 biologically independent experiments). e Saturation binding of rhPSMA/probe complexes. Data are presented as mean values ± s.d. (n = 3 biologically independent experiments). f Fluorescence intensity of the rhPSMA/probe complex in response to concentration changes. Data are presented as mean values ± s.d. (n = 3 biologically independent experiments). g Time frame for rhPSMA/probe complex formation. Data are presented as mean values ± s.d. (n = 3 biologically independent experiments). F.I. fluorescence intensity. Source data are provided in Source Data file.

Full size image

Design and synthesis of PSMA-targeted activatable probes

FMRs based on benzonitrile and julolidine moieties are well known24. Recently, these have been successfully modified by Yen-Pang Hsu et al. and FMRs with wavelengths closer to the near-infrared region (λex/λem = 490/660 nm) have been generated. Those modifications enabled better tissue penetration of a fluorescence signal for imaging applications in vitro or in vivo37. In this study, we aimed at the design of activatable probes by linking a benzonitrile fluorescent rotor to the Lys-Urea-Glu scaffold, a well-established PSMA-targeting moiety used clinically for PCa imaging and radiotherapy39. To maximize critical interactions between the benzonitrile fluorescent rotor and the protein upon ligand binding, the fluorophore was directly conjugated to the Lys-Urea-Glu scaffold at the ε-amine of lysine to yield Glu-490 (Supplementary Fig. 1). Additionally, we synthesized ODAP-490 and ODAP-436, Lys-Urea-oxalyldiaminopropionic acid analogs of the traditional glutamate-containing scaffold, to increase its water solubility and reduce nonspecific interaction. (Fig. 1b)12.

Biochemical and biophysical characterization of probes

To evaluate fluorescence enhancement of our FMR probes in relation to their molecular environment, initially, we used a glycerol solution to mimic rotationally constrained conditions likely found upon PSMA binding and collected fluorescence and UV-Vis spectra of the compounds37. ODAP-436 exhibited UV absorption and fluorescence emission maxima at 436 and 490 nm, respectively. Both Glu-490 and ODAP-490 exhibited UV absorption and fluorescence emission maxima at 490 and 660 nm, respectively (Fig. 1b). All probes showed a marked glycerol-dependent fluorescence increase, with an over 27-fold signal enhancement at 80% glycerol concentration compared to pure buffer (Fig. 1c).

Using recombinant human PSMA (rhPSMA), we next determined critical biophysical and biochemical parameters of the probes upon complexation with the enzyme in solution. First, IC50 values of the probes were determined using an established NAAG-hydrolyzing assay40. Inhibition constants for Glu-490, ODAP-490, and ODAP-436 were 63.1, 99.6, and 118.4 pM (Fig. 1d), respectively, identifying these compounds as potent PSMA-specific inhibitors. Next, we evaluated the stoichiometry and fluorescence enhancement of probes upon rhPSMA binding. To that end, probes at a concentration of 500 nM were titrated with a dilution series of rhPSMA revealing the expected 1:1 stoichiometry of binding (Fig. 1e). Furthermore, we observed a 29.9- to 38.3-fold increase in fluorescence intensity (Fig. 1f), pointing towards efficient rotational constraints imposed on the fluorophore upon PSMA binding. Finally, we evaluated the time required for PSMA activation of the probe. Here, a probe at 500 nM concentration was mixed with a 10-fold molar excess of PSMA, and the fluorescence signal measured over a 30 min time interval (Fig. 1g). The fluorescence signal reached its maximum in less than 30 sec, the time necessary for mixing the components and start of the measurement, and remained stable during the entire experiment.

Overall, compounds Glu-490, ODAP-490, and ODAP-436 provide nearly instantaneous and robust, stable fluorescence upon rhPSMA binding. The combination of these characteristics makes these probes ideal for biological and biomedical applications where time resolution is critical.

Mechanism of fluorescent activation upon probe-PSMA binding

To provide mechanistic rationale for the fluorescent enhancement of Glu-490 upon rhPSMA binding, we solved the crystal structure of the rhPSMA/Glu-490 complex and refined it to the 1.73 Å resolution limit (Table 1). The active-site-bound Glu-490 was fitted into well-defined interpretable Fo-Fc electron density peaks in the final stages of the refinement (Fig. 2b). Positioning of the Lys-Urea-Glu motif is virtually indistinguishable from rhPSMA/urea ligand complexes reported previously41. The P1’ glutamate moiety binds the S1’ pocket in the “canonical” mode, the urea linker engages several residues in the vicinity of the active-site zinc ion, and at the nonprime side, the most noticeable contacts involve the arginine patch of PSMA and the P1 α-carboxylate function of Glu-490 lysine function42,43.

a Molecular formula of the Glu-490. b A stereo view of the Gluo-490 inhibitor. The Fo-Fc omit map (green) is contoured at 3.0 σ and the inhibitor is shown in stick representation with atoms colored red (oxygen), blue (nitrogen), yellow (sulfur), and cyan (carbon). c Details of interactions between residues of the glutarate sensor (green carbons) and Glu-490 (cyan carbons). CH–π interactions are depicted as dashed lines with distances to the ring centers in Angstroms. The active-site zinc ions are shown as orange spheres. d Surface representation of PSMA with residues of the glutarate sensor interaction with the FMR moiety colored blue, PDB code (7BFZ).

Full size image

The FMR moiety consists of a distal N-substituted tetrahydroquinoline donor group, a thiophene ring spacer and a nitrile acceptor group (Fig. 2a). In our structure, the three subgroups adopt a near-planar configuration (Fig. 2b, c) that is sustained via intermolecular interactions with PSMA residues. That spatial arrangement prevents the intramolecular twisting motion of the subgroups relative to each other and such rigidity is critical for efficient fluorescence emission upon FMR photoexcitation44. The nitrile group is inserted into a deep pocket delineated the main chains of G206–F209 and the Y700 side chain. Furthermore, the FMR ring system is propped against the PSMA surface delineated by residues N698–G702 of the “glutarate sensor” (Fig. 2d)45. Here, the most conspicuous are CH–π interactions between the thiophene ring and the Y700 methylene group (4.0 Å) and the tetrahydroquinoline moiety and the A701 side chain methyl group (3.8 Å and 4.1 Å; Fig. 2c).

It can be surmised that the FRM moiety also comes intermittently into a contact with residues F546 and S547 of the entrance lid, a flexible segment at the entrance into the internal PSMA cavity (amino acids Y541–G548)42. However, the weaker and discontinuous Fo-Fc electron density map prevented us from generating this part of the model with high confidence (Supplementary Fig. 6).

Real-time monitoring of live cell receptor binding and endocytosis

As Glu-490 and ODAP-490 were virtually indistinguishable in our biophysical experiments, and ODAP-490 exhibited less nonspecific staining than Glu-490 (Supplementary Fig. 8), we selected ODAP-490 for the ensuing biological applications. Using LNCaP, 22RV1, and PC3 cell lines expressing high, medium-to-low, and negligible levels of PSMA46,47, respectively, we first evaluated the cytotoxicity of ODAP-490 using an established MTT-based viability assay48, the probe proved non-toxic at concentrations up to 25 μM (Fig. 3a, Supplementary Fig. 7). For imaging of PSMA-expressing cells, cells were incubated with 10 μM ODAP-490 for 2 h and subsequently visualized by fluorescence microscopy (Fig. 3b). Quantification of the fluorescent signal revealed correlation between PSMA expression levels and fluorescence intensity with LNCaP and 22RV1 showing a 4.4- and 2.3-fold higher signal compared to PC3 (Fig. 3c).

a Cytotoxicity of ODAP-490 determined by the MTT viability assay. Data are presented as mean ± s.d. (n = 3 independent experiments). b Wash-free imaging of LNCaP (PSMA+++), 22RV1 (PSMA+), and PC3 (PSMA−) cells. 10 μM ODAP-490 was applied at 37 °C for 2 h. Cell nuclei were counter-stained with hochest 33342. (red, ODAP-490; blue, hochest 33342). Scale bar: 20 μm. cQuantification of the fluorescence intensity of samples in panel b. Upper and lower bounds of boxes represent 25th and 75th percentiles, horizontal lines indicate the median values, whiskers represent the minimum and maximum ranges. (n = 20 biologically independent cell samples). ****P < 0.0001, P values = 6.35e−22(LNCaP); 2.90e−11(22RV1); two-tailed unpaired Student t-test. A.U. arbitrary units. d Inhibition of PSMA-mediated endocytosis at 4 °C. The bottom figures show the fluorescence intensities corresponding to cross-sections (yellow lines) in the upper figures. (red, ODAP-490; blue, hochest 33342). Scale bar: 5 μm. Experiment was repeated three times independently with similar results. e The real-time monitoring of PSMA ligand binding and endocytosis. LNCaP cells were treated with 10 μM ODAP-490 and images were captured with a confocal microscope at 37 °C for 2 h at 3 min intervals. (red, ODAP-490; blue, hochest 33342). Scale bar: 20 μm. f Inhibition of the PSMA binding and PSMA-mediated endocytosis. The 2 h time point images are shown. (red, ODAP-490; blue, hochest 33342). CPZ chlorpromazine. Scale bar: 20 μm. g Quantification of fluorescence intensity changes of endocytosis with/without inhibitors in panels e and f. Data are presented as mean + s.d. (n = 20 biologically independent cell samples). Source data are provided in Source Data file.

Full size image

To monitor PSMA endocytosis in real-time, LNCaP cells were treated with 10 μM ODAP-490 and cells were imaged using a confocal microscope at 37 °C for 2 h in 3 min intervals (Fig. 3e, Supplementary Movie 1). Fluorescence intensity rose rapidly within the first 6 min and then the signal increased gradually for the remaining time, plateauing at around 90 min. The time profile likely reflects rapid binding of the probe to the plasma membrane-resident PSMA molecules within the first 6 min followed by continuous PSMA internalization and recycling back to the plasma membrane (Fig. 3g). The specificity of fluorescence monitoring was verified using ZJ-43, an established, potent PSMA inhibitor49. By pretreating LNCaP cells with 100 µM ZJ-43, the fluorescence signal was significantly reduced (15 fold), consistent with competition between the probe and ZJ-43 for PSMA binding (Fig. 3f, g, Supplementary Movie 2).

Several known modulators of endocytosis were then used to evaluate the mechanism of PSMA trafficking. As expected, the endocytosis was blocked nearly completely by incubating cells at 4 °C, yet the strong fluorescence signal was still observed at the cell surface (Fig. 3d). Chlorpromazine (CPZ), nystatin, and 5-(N-ethyl-N-isopropyl) amiloride (EIPA) were used to target clathrin-dependent50, caveolin-dependent51, and macropinocytotic52 pathways, respectively. Cells were pretreated with 10 μM concentration of each inhibitor for 30 min prior to addition of ODAP-490, and the fluorescence signal was monitored. CPZ exhibited the most significant, ~86%, inhibition of the PSMA-mediated ODAP-490 uptake, conforming a major role for the clathrin-dependent pathway PSMA endocytosis (Supplementary Fig. 9, Supplementary Movie 3).

Fast and convenient staining of PCa tissues ex vivo

Human histological samples were used ex vivo to evaluate the feasibility of monitoring PSMA expression levels by ODAP-490. We first co-stained frozen sections of PCa surgical specimens using an anti-PSMA antibody and ODAP-490 (Fig. 4a), and observed correlation between the green fluorescence of antibody and the red fluorescence of ODAP-490 with a Pearson’s coefficient 0.88 (Fig. 4b).

a Fluorescence image of surgical specimens co-stained by PSMA antibody and ODAP-490. Nucleus were stained with DAPI and shown in blue. Antibody and ODAP-490 staining were shown in green and red, respectively. Scale bar, 10 μm. Experiment was repeated three times independently with similar results. b Correlation between antibody and ODAP-490 stains. Fluorescence intensities corresponding to cross-section (yellow lines) in panel a. c Fluorescence image of surgical specimens incubating with ODAP-490. Representative images of corresponding PSMA immunohistochemistry (IHC) of in different regions of interest (ROIs) are shown below. Scale bar: 100 μm. IHC results were repeated three times independently with similar results. d Correlation of the fluorescence intensity, Gleason score, cancer existence, and PSMA expression level of each ROIs in panel c. e Summary of fluorescence increase at ROIs in six clinical specimens incubated with ODAP-490. PBCC point-biserial correlation coefficient. Source data are provided in Source Data file.

Full size image

Six prostatectomy specimens were harvested and stained by ODAP-490. Differences in fluorescence intensity were observed in focal regions of each sample, likely reflecting heterogeneity of PSMA expression (Fig. 4c). For each sample, three regions of interest were selected, excised, and examined by a pathologist for PSMA expression and Gleason score, which correlated with a Pearson’s coefficient of 0.69 (Fig. 4c, d, e). Without additional interventions, such as fixing or washing, the method enabled a convenient way to identify PCa rapidly in patient specimens.

In vivo imaging of PCa

Coupled with the fluorescence switching characteristics of ODAP-490, we speculated that this probe can detect PCa in vivo at early timepoints. We tested the ability of ODAP-490 to detect PCa after intravenous injection into mice bearing PSMA-expressing prostate tumors. 22RV1 tumors with medium-to-low level of PSMA expression were selected to challenge the detection limit, and PC3 tumors were used as a negative control. After administration of 25 nmol ODAP-490, we found the tumor could be clearly visualized at 2 h after injection, with the combination effect of specific signal activation inside the tumor and the clearance of the background signal from blood pool (Fig. 5a, b). At the same condition, there was no obvious fluorescence signal accumulation in PC3 xenografts. This result indicates that ODAP-490 is suitable for specifically in vivo imaging of PSMA-expressing tumor.

We next compared the in vivo imaging capability of PCa of ODAP-490 with an established “Always-On” PSMA-targeting probe ODAP-800CW12. Mice were administrated with ODAP-490 or ODAP-800CW, and the images at 4 h post-injection were collected. With the tumor specific activatable property, ODAP-490 could reach a high tumor-to-background ratio of 29.1 folds, in comparison with ODAP-800CW barely reaching 2.4 folds (Fig. 5c, d). These preliminary in vivo results demonstrated the potential advantages of PSMA-activatable probe for faster tumor detection, although the wavelength of the activatable dye still need further effort to improve.

PSMA, specifically overexpressed in nearly all PCa with limited expression in normal tissues, is an ideal biomarker for imaging PCa with high sensitivity and specificity. To overcome the time-consuming limit of fluorescent PSMA detection, we designed a series of PSMA-activatable fluorescence probes by facile conjugation of FMRs to a PSMA-targeting moiety. The probes bind to PSMA with very high potency (63–118 pM). We showed that upon PSMA binding, the FMR tightly fitted into the entrance funnel of the enzyme restricting conformation of the fluorophore (PDB code: 7BFZ), and therefore eliciting the fluorescent signal (29.9 to 38.3 fold). Since that process is triggered directly and only by PSMA/probe complex formation, the instant signal turning-on can be achieved within 30 seconds, which represents the fastest PSMA activated probes discovered so far. The high PSMA binding affinity and fast signal response allowed us to test a series of biomedical applications.

The PSMA expression on live cells can be visualized using a convenient wash-free protocol by simply incubating with the probe. With this protocol, we were able to monitor and quantify the entire PSMA/probe complex formation and endocytosis process in living cells in real-time. By quantifying the effects of different types of endocytosis inhibitors on PSMA-mediated endocytosis, we confirmed that the clathrin-dependent endocytosis pathway plays the most important role in the endocytosis of PSMA53,54, while the contribution of the caveolin-dependent and micropinocytosis pathway is quite limited. That is consistent with the established results that most transmembrane receptors are internalized via a clathrin-dependent mechanism, as recently shown by Matthias et al., who did so using stimulated emission depletion (STED) nanoscopy55.

Meanwhile, a similar wash-free protocol could be adapted for staining PCa surgical specimens and we were able to achieve a quick and accurate PSMA quantification within 10 min, correlating with cancer existence and Gleason score. Compared with conventional time-consuming histopathology, the wash-free method could potentially be adapted intraoperatively. Furthermore, the specific in vivo imaging of ODAP-490 could be achieved in mouse xenografts within a few hours. Compared with the “Always-On” PSMA-targeted dye ODAP-800CW12, ODAP-490 achieved a significant improved tumor-to-background contrast at the early timepoints, making it a special candidate for intraoperative surgical guidance when imaging time is critical.

In summary, we have developed a series of PSMA-activatable fluorescence probes, which could exhibit a fast fluorescence intensity increase upon binding to PSMA. The probe was successfully applied to real-time monitor the whole process of PSMA ligand binding and mediated endocytosis in live cells, and quickly wash-free stain PCa surgical specimen. In vivo imaging of PCa was also successfully demonstrated with the advantage for detecting early timepoints. Effort to further improve the wavelength of PSMA-activatable probes is currently under investigation.