Image-Guided Surgery: Are We Getting the Most Out of Small-Molecule Prostate-Specific-Membrane-Antigen-Targeted Tracers?

Abstract

Expressed on virtually all prostate cancers and their metastases, the transmembrane protein prostate-specific membrane antigen (PSMA) provides a valuable target for the imaging of prostate cancer. Not only does PSMA provide a target for noninvasive diagnostic imaging, e.g., PSMA-positron emission tomography (PSMA–PET), it can also be used to guide surgical resections of PSMA-positive lesions. The latter characteristic has led to the development of a plethora of PSMA-targeted tracers, i.e., radiolabeled, fluorescent, or hybrid. With image-guided surgery applications in mind, this review discusses these compounds based on clinical need. Here, the focus is on the chemical aspects (e.g., imaging label, spacer moiety, and targeting vector) and their impact on in vitro and in vivo tracer characteristics (e.g., affinity, tumor uptake, and clearance pattern).

Introduction

There has been a surge in interest in the diagnosis, staging, and treatment of prostate cancer (PCa), the second most frequent cancer and the fifth leading cause of cancer deaths in men.^1,2 PCa is highly heterogeneous: its manifestation ranges from indolent, slow-growing, low-grade tumors, to high-grade aggressive neoplasms with extensive metastasis that eventually are the cause of mortality.³ Overexpression of PSMA in >90% of PCa patients^2,4 has driven the development of PSMA-targeted tracers for imaging purposes. PSMA-targeted tracers come in all shapes and sizes (including proteins and nanoparticles), yet the clinical standard is set by glutamate–ureido-based “small-molecule” imaging agents designed to facilitate positron emission tomography (PET) imaging (e.g., ⁶⁸Ga-PSMA-11, ⁶⁸Ga-PSMA-617, ¹⁸F-DCFPyL, and ¹⁸F-PSMA-1007).⁵ PSMA-targeted tracers provide sensitive detection of metastatic lesions both at low PSA values⁶⁻¹⁰ and at tumor diameters <5 mm,¹¹ and have been successfully used to deliver PSMA-targeted endoradiotherapy¹² and theranostics¹³ in heavily metastasized castration-resistant PCa patients.

Surgery remains the mainstay of primary PCa treatment;¹⁴ surgery is used for (1) resection of the primary cancer (prostatectomy) and (2) management of locoregional progressive disease in PCa management.¹⁵ For surgery, pre- and perioperative imaging of lymph nodal tumor infiltration and extracapsular tumor expansion (ECE) remain a challenge to date.^16,17 At the moment, image-guidance targets in prostate cancer surgery are confined to sentinel- and/or PSMA-overexpressing lymph nodes.¹⁸ To realize precision PSMA-targeted PCa surgery, preoperative molecular imaging “roadmaps” supplied by PSMA–PET need to be transferred to the surgical setting, indicating a demand for interventional PSMA-targeted imaging techniques.^19,20 Molecular image-guided surgery can be divided into three classes, i.e., tracers for (1) radioguided surgery (positron emission tomography, PET; single-photon emission computed tomography, SPECT; and Cerenkov), (2) fluorescence-guided surgery, and (3) bimodal, i.e., hybrid approaches (SPECT/fluorescence or PET/fluorescence combinations). When these concepts are applied to small-molecule PSMA-targeted tracers, each class requires a different chemical design. Nonetheless, in all three cases tracer designs for image-guided surgery purposes should complement existing diagnostic imaging strategies that make use of the clinically accepted glutamate–ureido-based radiotracers for PSMA–PET.

A plethora of PSMA-targeted tracer designs has been presented with image-guided surgery applications in mind, including tracers that already are available for image-guided surgery in the clinic. In this review, we aim at complementing earlier reviews on PSMA-guided surgery²⁰⁻²³ by discussing: (1) the compatibility of glutamate–ureido-based “small-molecule” PSMA-targeted tracer designs with the binding pocket of PSMA, (2) the influence of tracer design on pharmacokinetics, and (3) the ability of a given design with all its characteristics to address the needs of a surgeon. In doing so, (radio)chemists, nuclear medicine physicians, and surgeons are familiarized with the status quo of small-molecule PSMA-targeted tracers for image-guided surgery.

The Biology of PSMA

PSMA goes by different names, i.e., N-acetyl-l-aspartyl-l-glutamate peptidase I (NAALADase I) in the nervous system and the more generally applicable name glutamate carboxypeptidase II (GCP II). The protein is encoded by folate hydrolase 1 (FOLH1) in humans²⁴ and has an extracellular domain performing enzymatic functions.^25,26 When PSMA is internalized, it activates the Protein Kinase B (AKT) and mitogen-activated protein kinase pathways, thus promoting proliferation and survival.^27,28

PSMA is expressed in all human prostate cells, including normal and hyperplastic epithelium and prostatic–intraepithelial neoplasia.² In prostate adenocarcinoma, a 100- to 1000-fold overexpression of PSMA is seen compared to benign prostate tissue.²⁹ The level of PSMA expression correlates with tumor aggressiveness²⁴ and Gleason score.³⁰ Next to its expression on prostate cancer, PSMA is also expressed on the endothelial cells of tumor-associated neovasculature of cancers such as lung carcinoma and neuroendocrine carcinoma of the pancreas.³¹ This characteristic implies that PSMA-targeting strategies could serve oncological challenges beyond prostate cancer in the future.

PSMA expression levels may be influenced by hormonal therapies commonly applied in prostate cancer. Since androgens downregulate FOLH1 gene expression,³² short-term androgen deprivation therapy (ADT) enhances PSMA expression.⁶ On the other hand, long-term ADT will decrease tumor volume and thereby limit the potential of PSMA-based imaging.³³

Requirements for Targeting PSMA

The apical domain of the extracellular portion of PSMA provides the target for small-molecule inhibitors as it contains the binuclear-zinc-ion-containing active site of the protein, which can be reached via a 20-Å-deep substrate-binding cavity (Figure 1A).^34,35 PSMA has two substrate-binding pockets within this cavity designated the S1 and S1′ pocket. The S1 pocket is fairly specific for glutamate and aspartate side chains and this pocket is optimized for binding of glutamate and glutamate-like residues.³⁶ The S1′ pocket is amphipathic, i.e., it exploits both polar (hydrogen-bonding, ionic) and nonpolar (hydrophobic, van der Waals) interactions to induce binding and stabilization of the substrate or inhibitor moieties.^37,38 Low nanomolar PSMA inhibition of this binding site can be achieved through small-molecule moieties of different compositions: phosphonates such as 2-(phosphonomethyl)pentanedioic acid,³⁹ thiols such as 2-(3-mercaptopropyl) pentanedioic acid,⁴⁰ and ureas such as N-[[[(1S)-1-carboxy-3-methylbutyl]amino]carbonyl]-l-glutamic acid.⁴¹ Tracer designs have only been based on organophosphorus-based (mimicking phosphinic acid inhibitors of PSMA) and glutamate–ureido-based (mimicking N-acetylaspartylglutamate) moieties, of which the latter are most abundantly exploited and thus described in this review.

Figure 1.

(A) Schematic cross-section view of PSMA illustrating intrinsic interactions between the three critical tracer components (targeting vector, spacer, and imaging flag) and the protein’s primary and secondary binding site. (B) Depiction of anatomical relevance in PCa surgery: bladder (in blue) relative to liver (red), kidneys (blue), lymph nodes (green), and prostate (pink). (C) Typical intraoperative sight during radical prostatectomy using nerve-sparing surgery.

Based on crystallographic data, Barinka et al.⁴² have described the existence of a “remote hydrophobic binding register” or rather “accessory hydrophobic pocket” that is formed by portions of β-sheets β13 (Arg463–Asp465) and β14 (Arg534–Arg536). This secondary binding pocket allows for binding of distally placed benzene-like moieties. Exploration of this pocket has led to various structure–activity relationship studies to determine, e.g., optimal spacer length between targeting vector and the benzene-like structure.^43,44 In radiochemistry, exploring the beneficial features of this secondary binding site has proven to be of paramount importance in tracer refinement.²⁶

Imaging labels (e.g., radioisotopes, chelates, or fluorophores) are generally placed outside of the amphipathic entrance funnel (Figure 1A). These moieties can impact the affinity; from radiochemical efforts, it seems that mostly steric hindrance and/or charge play a large role in these influences.^{13,26,36,43,45−48} For example, Banerjee et al.⁴⁹ placed various tridentate moieties (chelating ^99mTc) outside the amphipathic entrance funnel, resulting in a more than 60-fold increase in tracer affinity compared to analogs without a spacer moiety.

In an ideal setting, the above-mentioned characteristics imply that small-molecule PSMA-targeted tracers should be designed to meet three requirements: (1) enforcing primary binding with the S1/S1′ pocket(s) and Zn²⁺ ions located in the active site; (2) promoting secondary binding in the accessory hydrophobic pocket; (3) minimizing the impact of the externally placed imaging label (Figure 1A) on the binding affinity with the use of lengthy spacer moieties as previously researched for radiotracers.^{25,36,43,47,49}

Surgical Requirements

PSMA-targeted image-guided surgery can, in theory, address a number of challenges during PCa surgery. A first surgical application is found in the identification of locoregional metastatic lymph nodes (LN; Figure 1B).^18,50−52 In primary surgery, identifying disease-related LNs by lymphatic mapping has been shown to complement extended pelvic LN dissections.^18,53 The same may be true for PSMA-targeted image guidance approaches. Alternatively, in recurrent PCa, salvage surgery can address the resection of PSMA–PET-positive LNs.⁵⁴ However, as is well-known, PET-based detection of PSMA-positive nodes is size-dependent; analyses have shown PSMA–PET detection rates of 0% and >60% when the short axis diameter of the metastatic lesion was ≤2.0 and 2–4.9 mm, respectively.^55,56 This characteristic, combined with the fact that the amount of tracer uptake in micrometastases will lead to an insufficient signal-to-background ratio, implies that PSMA-targeted image-guided surgery is limited to the surgical treatment of macrometastases.¹⁸ Still, patient selection based on the PSMA-based PET scan as well as other clinical variables (PSA value, number of lesions, location of lesions, time from first treatment, Gleason score, and presence of androgen deprivation therapy) remains mandatory.⁵⁷

A second image-guided surgery application is found in defining surgical margins during prostatectomy (Figure 1C). Intraoperative margin assessment is of particular relevance when both ECEs might be present and a nerve-sparing radical prostatectomy is intended (to preserve erectile function). The diagnostic accuracy of current imaging methods in predicting ECE is poor. Intraoperatively, the ex vivo neurovascular structure-adjacent frozen-section examination (NeuroSAFE) technique can be used to study excised tissue and to pathologically assess the presence of tumor cells in the resection margins.^58,59 Intraoperative assessment of margins would improve surgical logistics and would also improve prediction of the presence of residual lesions in the patient.

Despite the specificity for PCa, basal PSMA expression has also been reported for other tissues, including benign prostate epithelium.^2,60 Basal PSMA expression may thus provide a background signal that needs to be considered when margin assessment is aimed for. While glutamate–ureido-based radiotracers have readily proven their ability to target the primary tumor,⁶¹ their predominant renal clearance could also compromise margin assessment; during prostatectomy, the bladder is disconnected from the prostate and urine might flow into the surgical field (Figure 1B). If urine contains PSMA-targeted tracer (or its metabolites), it can contaminate the margins and impair the ability to surgically assess the tumor spread relative to the background.⁵⁶ However, renal excretion rates are unknown for most tracers. While the renal cortex is said to express PSMA,^2,60,62−66 studies with the PSMA-targeted radiolabeled J591 monoclonal antibody did not confirm its presence.^67,68 As the proximal tubules are the prime location for a plethora of transporters and multiligand receptors such as megalin and cubilin, alternative mechanisms could potentially also influence renal accumulation.⁶⁹⁻⁷¹ The observation of ¹⁷⁷Lu-PSMA I&T depicting an 8-fold reduction in kidney uptake/retention at late (24–48 h) compared to early (1–4 h) time points strengthens the assumption that metabolism of renally accumulated PSMA-targeted tracers could influence urine contamination.⁷² It is thus assumable that renal tracer retention can to a certain degree be predictive for the presence of imaging agents or labels in urine. To avoid such contamination during surgery, a tracer either should not undergo renal clearance or the time between tracer administration and surgery should be extended to minimize contamination.

The physics of intraoperative guidance is dictated to a large degree by two related features, i.e., the tissue penetration of the imaging signature and the intensity of the signal. Identification of affected LNs is likely to demand imaging signatures that can readily penetrate tissue, as metastases are embedded in, e.g., adipose tissue, and can be rather small, hence favoring radioguidance modalities.⁷³ The assessment of ECE in the dissection planes requires a superficial feedback as signal from deeper-lying tissues limits the spatial resolution. This requirement may be optimally supported by the use of optical technologies. However, the availability of surgical modalities that enable detection of the tracer used for guidance is imperative. While perhaps not immediately obvious, the surgical technique itself may also influence the choice for a particular tracer. During open surgery, a small (approximately 15 × 10 cm²) but deep cavity is created that can easily accommodate a γ-probe but limits the use of a 2D fluorescence camera. In a (robot-assisted) laparoscopic setting, on the other hand, dedicated (DROP-IN) γ-probes are required for radiotracing, while the availability of a fluorescence (and 3D) laparoscopic system is quickly becoming standard practice.⁷⁴ For example, since the Da Vinci S model has become available, all these surgical robots are equipped with a fluorescence camera capable of imaging different fluorophores.^74,75

The preceding paragraph implies that PSMA-targeted image-guided surgery is limited to the surgical treatment of macrometastases and defining the surgical margins. Both applications require a high signal-to-background ratio and benefit from minimal tracer presence in urine.

PSMA-Targeted Tracers

Tracers for PSMA-Targeted Radioguided Surgery

The field of radioguided surgery revolves around the intraoperative use of oftentimes diagnostic radiotracers commonly used in the field of nuclear medicine. In this concept, noninvasive preoperative images can be used to provide both a surgical roadmap and intraoperative guidance. A range of radioactive isotopes, corresponding chelates, and combinations thereof have been coupled to PSMA-targeted vectors, thereby developing a variety of radiotracers. Unique for PSMA-targeted radiotracers is that many of them have already been translated to the clinic. Interestingly, relatively few preclinical efforts with image-guided surgery in mind have been reported in this area. (Scheme 1, Table 1, Figure 2).

Scheme 1. Chemical Structures of PSMA-Targeted Tracers for PET and SPECT Imaging.

Targeting moiety (green) and imaging flag (red) have been colored accordingly.

Table 1. Radioisotope-Based PSMA-Targeted Tracers for Image-Guided Surgery.

compound	receptor affinity (competitor)	tumor-to-background (time point, cell line, background tissue)	kidney-to-tumor (time point, cell line)	injected dose	biodistribution; time point	clearance pathway (time point, uptake clearance organs)	half-life	lipophilicity	plasma protein binding	optimized for entrance funnel?	imaging label	imaging signature (s)
68Ga-PSMA-1178	IC50 = 7.5 ± 2.2 (2-PMPA)	8b (1 h, LNCaP, muscle)	18.1 (1 h, LNCaP)	0.1–0.2 (10–20 MBq/nmol)	Yes; 1 h	Renal (1 h, kidneys 139.4 ± 21.4%ID/g, liver 0.87 ± 0.05% ID/g)	t1/2, rad = 1.1 h	LogP = −3.15a	Data not provided	Yes	HBED-CC, 68Ga	β (Emaxβ+ = 1899 keV), γ (Eγ = 511 keV), Cerenkov (>250 nm)
111In-PSMA I&T48	IC50 = 7.5 ± 1.5 nM ((125I-BA)KuE)	43 ± 6 (1 h, LNCaP, muscle)	23.8 (1 h, LNCaP)	0.2 nmol (7 MBq/nmol)	Yes; 1 h	Renal (1 h, kidneys, 191 ± 24%ID/g, liver 0.26 ± 0.04%ID/g)	t1/2, rad = 67 h	LogP = −4.5	83%93	Yes	DOTAGA, 111In	γ (Eγ = 171.3, 245.4 keV)
99mTc-PSMA I&S93	IC50 = 39.7 ± 1.2 nM((125I-BA)KuE)	21 (1 h, LNCaP, muscle)	22.5 (1 h, LNCaP)	0.1 nmol (30–40 MBq/nmol)	Yes; 1 h, 3 h, 5 h, 12 h, 16 h, 21 h	Renal (1 h, kidneys 186 ± 23%ID/g, liver 1.58 ± 0.24%ID/g)	t1/2, rad = 6.5 h	LogP = −3.0	94%	Yes	mas3,99mTc	γ (Eγ = 140.5 keV)

^aValues were calculated by using ChemAxon’s Chemicalize plugin.

^bEstimated from graph.

Figure 2.

Typical clinical workflow of radioguided surgery: (A) preoperative ⁶⁸Ga-PSMA-11 PET/CT followed by (B) preoperative ^99mTc-PSMA I&S SPECT/CT; (C) intraoperative guidance combined with minimally invasive robot-assisted laparoscopic surgery using a DROP-IN γ-probe with ^99mTc-PSMA I&S. (Reprinted from van Leeuwen et al.⁷⁶ with permission from Wolters Kluwer Health, Inc.).

Use of Established PSMA–PET Tracers for Image-Guidance Purposes

During patient selection for PSMA-targeted image-guided surgery, the availability of, e.g., diagnostic PSMA–PET images, has proven to be a critical prerequisite (Figure 2A).⁷⁷ Intraoperatively, radiotracers such as ⁶⁸Ga-PSMA-11⁷⁸ can potentially be traced via their β⁺-emission (E_maxβ⁺ = 1899 keV), high-energy γ-emission (Eγ = 511 keV), or Cerenkov light (>250 nm). Despite literature examples regarding intraoperative high-energy γ-emission tracing of PET tracers,⁷⁹ this application has not yet been reported for PSMA. Tracing the β⁺-emission of ⁶⁸Ga-PSMA-11 (IC₅₀ = 7.5 ± 2.2)⁷⁸ is currently being investigated for ex vivo margin assessment in primary PCa and to identify PSMA-positive LNs,⁸⁰ even though this emission type is limited by its tissue penetration of a few millimeters.⁸¹ Alternatively, Costa et al.⁸² have used ex vivo Cerenkov imaging to identify ⁶⁸Ga-PSMA-11 uptake in prostate sections containing primary PCa lesions. Despite the potential displayed by these technologies, the half-life of ⁶⁸Ga (t_1/2, rad = 1.1 h) limits the surgical time frame, the radiation dose to the surgical staff is potentially high,^83,84 and renal clearance of ⁶⁸Ga-PSMA-11 (kidney retention 139.4 ± 21.4%ID/g at 1 h) means that prostate samples need to be rinsed prior to imaging to remove urine contamination. Furthermore, the signal penetration of β⁺-emissions (mm range)⁸¹ and the resulting Cerenkov luminescence (<2 mm)⁸⁵ limits the implementation to superficial lesions. On a positive note, these technologies are also applicable for alternative PET tracers such as, e.g., ¹⁸F-PSMA-1007 (minimized renal clearance and extended half-life).⁸⁶ With regard to half-life, the use of, e.g., ⁶⁴Cu-based derivatives (t_1/2 = 12.7 h, E_maxβ⁺ = 653 keV)⁸⁷ could potentially help to further extend the time interval between tracer administration and surgery.

SPECT Tracers for PSMA-Targeted Image-Guided Surgery

In the field of radioguided surgery, γ-ray-emitting radiotracers are most commonly implemented.⁸³ Herein, radionuclides such as ¹²⁵I (Eγ = 31.0, 35.5, 27.2, 27.4 keV), ¹⁷⁷Lu (113, 210 keV), ^99mTc (Eγ = 140.5 keV), and ¹¹¹In (Eγ = 171.3, 245.4 keV) have proven to be compatible with conventional γ-probes.⁸¹ Given the clinical use of the chemically optimized ⁶⁸Ga-PSMA I&T, wherein a peptidic 3-iodo-d-Tyr-d-Phe spacer promotes secondary binding,¹³ and the ability of the 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTAGA) chelate to coordinate ¹¹¹In, ¹¹¹In-PSMA I&T was explored in radioguided surgery applications.^88,89 In direct comparison, the ^natIn-PSMA I&T (7.5 ± 1.5 nM)⁴⁸ showed a similar affinity for PSMA compared to its ^natGa and ^natLu-labeled analogs (9.4 ± 2.9 and 7.9 ± 2.4 nM, respectively). Clinical evaluation of ¹¹¹In-PSMA I&T in six patients revealed that although the sensitivity and spatial resolution of ¹¹¹In SPECT/CT are limited, uptake of the tracer in PSMA-positive LNs could be used to provide efficient surgical guidance.^77,88 Theoretically, tracer analogs like ¹¹¹In-PSMA-617(90) could also be explored in a similar radioguided surgery setting. To circumvent the suboptimal nuclear properties of ¹¹¹In (high cost, limited availability of ¹¹¹InCl₃, relatively poor SPECT quality, and high burden of radiation), a demand for a tracer analog labeled with ^99mTc arose, ^99mTc is one of the—if not the—most important radionuclide(s) used in nuclear medicine due to its availability from a generator, convenient half-life, negligible radiation dose to the patient, and low cost.⁸³ Given the versatile but very specific coordination chemistry of ^99mTc,⁹¹ various ^99mTc-chelating moieties used in PSMA-targeted radiotracers have been described.^{47,49,62,92−9599m}Tc PSMA I&S (IC₅₀ = 39.7 ± 1.2 nM)⁹³ employs the mas₃ chelate, and to optimize interaction with the accessory hydrophobic pocket in the PSMA protein the tracer contains a d-Tyr-d-2-Nal-d-Lys spacer moiety. The lipophilicity of PSMA I&S (LogP = −3.0) is higher compared to its parent analog PSMA-I&T (LogP = −4.5), resulting in delayed blood clearance (^99mTc PSMA I&S 1.73 ± 0.50%ID/g; ¹¹¹In PSMA I&T 0.24 ± 0.05%ID/g). ^99mTc PSMA I&S provides high-contrast preoperative PSMA-SPECT (at time points >4 h post injection, preferably 18 h) and supports tracing with a traditional γ-probe in open surgery,^93,96 or during robot-assisted surgery with a DROP-IN γ-probe (Figure 2B,C).⁷⁶ To date, this tracer has predominantly been used in patients that have received salvage surgery of metastasized LNs.^57,96 Following PSMA-guided surgery, a complete biochemical response (PSA < 0.2 ng/mL) could be achieved in 66% of 121 patients.⁵⁷ Notably, the clinical studies did also indicate that PSMA-guided surgery was particularly valuable in patients with low preoperative PSA values that displayed a single lesion on PSMA–PET.

The radioguided surgery studies have relied heavily on previous research and expertise obtained in the field of radiochemistry. Efforts are focused on both primary PCa as well as nodal metastases and set the current standard for PSMA-targeted image-guided surgery applications.

PSMA-Targeted Fluorescent Tracers

In contrast to tracers for radioguided surgery, fluorescent PSMA-targeted tracers have the potential to support real-time optical visualization of PSMA-positive lesions. Unfortunately, the ability to use this technique for lesion identification is limited to superficial applications (<1 cm from the surface), meaning it cannot be used to create surgical roadmaps.⁹⁷ Preclinically, various fluorescent PSMA-targeted tracers have been described; designs greatly vary in spacer moieties and fluorophores (visible to far-red fluorophores and near-infrared fluorophores). (Tables 2, 3, Schemes 2, 3, and Figure 3.)

Table 2. Visible to Far-Red Fluorescent Imaging Agents for PSMA-Targeted Fluorescence-Guided Surgery.

compound	receptor affinity (competitor)	tumor uptake (time point, cell line)	kidney uptake (time point, cell line)	injected dose	biodistribution; time point	clearance pathway (time point, uptake clearance organs)	half-life	lipophilicity	plasma protein binding	optimized for entrance funnel?	imaging label	imaging signature(s)
CYUE-Rhodamine B98	kD = 88 nM (2-PMPA)	No ex vivo quantification, no scale bars	No ex vivo quantification, no scale bars	No in/ex vivo experiments performed	No in/ex vivo experiments performed	No in/ex vivo experiments performed	No in/ex vivo experiments performed	LogP = −0.37a	Data not provided	Yes	Rhodamine-B (427 g/mol)	λabs = 510 nm, λem = 565 nm
DUPA-Dylight 68046	kD = 33 nM (2-PMPA)	No ex vivo quantification, no scale bars	No ex vivo quantification, no scale bars	10 nmol	No ex vivo quantification, no scale bars	No ex vivo quantification, no scale bars	Data not provided	LogP = 1.38a	Data not provided	Yes	DyLight 680 (790 g/mol)	λabs = 682 nm, λem = 715 nm
DUPA-Alexa Fluor 64746	kD = 4.5 nM (2-PMPA)	No ex vivo quantification, no scale bars	No ex vivo quantification, no scale bars	10 nmol	No ex vivo quantification, no scale bars	No ex vivo quantification, no scale bars	Data not provided	Fluorophore structure not available	Data not provided	Yes	AlexaFluor 647 (853 g/mol)	λabs = 650 nm, λem = 670 nm
PSMA-1-Cy5.599	IC50 = 2.1 ± 0.1 nM (Glutamate-ureido-cysteine)	5.08%ID (24 h, PC3-PIP) 2, 3, 7, 17, 18, 16, 51, 51, 50, 43, 41, 41 AUb (at 0.8, 0.17, 0.5, 1, 2, 6, 8, 24, 48, 72, 96, and 120 h PC3-PIP (in vivo))	0.025 AUc (120 h, PC3-PIP)	1 nmol	Qualitative; 0.8, 0.17, 0.5, 1, 2, 4, 6, 8, 24, 48, 72, 96, and 120 h (in vivo); Qualitative; 120 h (ex vivo)	Renal (120 h, kidneys 0.025 AUc, liver 0.01 AUc)	Data not provided	LogP = −1.02 ± 0.23	Data not provided	Yes	Cy5.5 (567 g/mol)	λabs = 675 nm, λem = 694 nm
Cy5.5–1101	ki = 90 ± 40 pM (NAAG)	30000 AUc (24 h, PC3-PIP)	60000 AUc (24 h, PC3-PIP)	1 nmol	Qualitative; 1, 4, 24 h	Renal (24 h, kidneys 60000 AUc, liver 10000 AUc)	Data not provided	LogP = 0.12a	Data not provided	Yes	Tetrakis sulfonate Cy5.5 (897 g/mol)	λabs = 675 nm, λem = 694 nm
Cy5.5–2101	ki = 50 ± 20 pM (NAAG)	50000 AUc (24 h, PC3-PIP)	50000 AUc (24 h, PC3-PIP)	1 nmol	Qualitative; 1, 4, 24 h	Renal (24 h, kidneys 50000 AUc, liver 10000 AUc)	Data not provided	LogP = −1.08a	Data not provided	Yes	Tetrakis sulfonate Cy5.5 (897 g/mol)	λabs = 675 nm, λem = 694 nm
Cy5.5–3101	ki = 50 ± 2 pM (NAAG)	70000 AUc (24 h, PC3-PIP)	50000 AUc (24 h, PC3-PIP)	1 nmol	Qualitative; 1, 4, 24 h	Renal (24 h, kidneys 50000 AUc, liver 10000 AUc)	Data not provided	LogP = 3.00a	Data not provided	Yes	Tetrakis sulfonate Cy5.5 (897 g/mol)	λabs = 675 nm, λem = 694 nm
Cy5.5(SO3–)4-EuK102	ki = 0.6 ± 0.1 nM (NAAG)	3700 AUb (24 h, 22Rv1)	24000 AUc (24 h, 22Rv1)	10 nmol	Qualitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 24000 AUc, liver 3540 AUc)	Data not provided	LogP = −1.36a	Data not provided	No	Tetrakis sulfonate Cy5.5 (897 g/mol)	λabs = 675 nm, λem = 694 nm
Cy5.5(SO3–)4-(EuK)2102	ki = 0.1 ± 0.0 nM (NAAG)	5000 AUb (24 h, 22Rv1)	20000 AUc (24 h, 22Rv1)	10 nmol	Qualitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 20000 AUc, liver 0 AUc)	Data not provided.	LogP = −0.33a	Data not provided	No	Tetrakis sulfonate Cy5.5 (897 g/mol)	λabs = 675 nm, λem = 694 nm
Cy5.5(SO3–)4-4-(aminomethyl)benzoic acid-(EuK)2102	ki = 0.4 ± 0.1 nM (NAAG)	5300 AUb (24 h, 22Rv1)	20000 AUc (24 h, 22Rv1)	10 nmol	Qualitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 20000 AUc, liver 3540 AUc)	Data not provided	LogP = 1.32ak	Data not provided	No	Tetrakis sulfonate Cy5.5 (897 g/mol)	λabs = 675 nm, λem = 694 nm
Cy5.5(SO3–)4-2-nitroimidazole-EuK103	ki = 0.3 ± 0.0 nM (NAAG)	18 AUb (24 h, 22Rv1)	29 AUb (24 h, 22Rv1)	10 nmol	Qualitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 29 AUb, liver 15 AUb)	Data not provided	LogP = −0.31a	Data not provided	No	Tetrakis sulfonate Cy5.5 (897 g/mol)	λabs = 675 nm, λem = 694 nm
Cy5.5(SO3–)4-(2-nitroimidazole)2-EuK103	ki = 3.0 ± 0.3 nM (NAAG)	20 AUb (24 h, 22Rv1)	46 AUb (24 h, 22Rv1)	10 nmol	Qualitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 46 AUb, liver 13 AUb)	Data not provided	LogP = −2.24a	Data not provided	No	Tetrakis sulfonate Cy5.5 (897 g/mol)	λabs = 675 nm, λem = 694 nm
YC-9100	ki = 0.2 ± 0.05 nM (NAAG)	15 AUc (24 h, PC3-PIP)	25 AUc (24 h, PC3-PIP)	10 nmol	Qualitative; 6 h	Renal (24 h, kidneys 25 AUc, liver 3 AUc)	Data not provided	LogP = −12.72a	Data not provided	Yes	IRDye700DX (1635 g/mol)	λabs = 689 nm, λem = 700 nm

^aValues were calculated by using ChemAxon’s Chemicalize plugin.

^bEstimated from graph.

^cEstimated from scale bar. AU = arbitrary units.

Table 3. Near-Infrared Imaging Agents for PSMA-Targeted Fluorescence-Guided Surgery.

compound	receptor affinity (competitor)	tumor uptake (time point, cell line)	kidney uptake (time point, cell line)	injected dose	biodistribution; time point	clearance pathway (time point, uptake clearance organs)	half-life	lipophilicity	plasma protein binding	optimized for entrance funnel?	imaging label	imaging signature(s)
Cy7–1101	ki = 1.0 ± 0.5 pM (NAAG)	30 AUdb (24 h, PC3-PIP)	20 AUdb (24 h, PC3-PIP)	1 nmol	Quantitative, 24 h	Renal (24 h, kidneys 20 AUdb, liver 10 AUdb)	Data not provided	LogP = 3.39a	Data not provided	Yes	Cy7 (665 g/mol)	λabs = 753 nm, λem = 775 nm
Cy7–2101	ki = 7.0 ± 0.4 pM (NAAG)	200 AUdb (24 h, PC3-PIP)	28 AUdb (24 h, PC3-PIP)	1 nmol	Quantitative; 24 h	Renal (24 h, kidneys 28 AUdb, liver 10 AUdb)	Data not provided	LogP = 2.19a	Data not provided	Yes	Cy7 (665 g/mol)	λabs = 753 nm, λem = 775 nm
Cy7–3101	ki = 5.0 ± 0.2 pM (NAAG)	200 AUdb (24 h, PC3-PIP)	30 AUdb (24 h, PC3-PIP)	1 nmol	Quantitative; 24 h	Renal (24 h, kidneys 30 AUdb, liver 10 AUdb)	Data not provided	LogP = 3.18a	Data not provided	Yes	Cy7 (665 g/mol)	λabs = 753 nm, λem = 775 nm
IRDye800CW-1101	ki = 70 ± 5 pM (NAAG)	500 AUdb (24 h, PC3-PIP)	3000 AUdb (24 h, PC3-PIP)	1 nmol	Quantitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 3000 AUdb, liver 28 AUdb)	Data not provided	LogP = 2.45a	Data not provided	Yes	IRDye800CW (983 g/mol)	λabs = 774 nm, λem = 789 nm
IRDye800CW-2101	ki = 40 ± 10 pM (NAAG)	2200 AUdb (24 h, PC3-PIP)	4000 AUdb (24 h, PC3-PIP)	1 nmol	Quantitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 4000 AUdb, liver 13 AUdb)	Data not provided	LogP = 1.25a	Data not provided	Yes	IRDye800CW (983 g/mol)	λabs = 774 nm, λem = 789 nm
IRDye800CW-3101	ki = 20 ± 5 pM (NAAG)	3200 AUdb (24 h, PC3-PIP)	1100 AUdb (24 h, PC3-PIP)	1 nmol	Quantitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 1100 AUdb, liver 20 AUdb)	Data not provided	LogP = 3.21a	Data not provided	Yes	IRDye800CW (983 g/mol)	λabs = 774 nm, λem = 789 nm
IRDye800RS-1101	ki = 100 ± 10 pM (NAAG)	0.05 AUc (24 h, PC3-PIP)	0.10 AUc (24 h, PC3-PIP)	1 nmol	Qualitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 0.10 AUc, liver 0.03 AUc)	Data not provided	LogP = 5.01a	Data not provided	Yes	IRDye800RS (825 g/mol)	λabs = 774 nm, λem = 789 nm
IRDye800RS-2101	ki = 200 ± 50 pM (NAAG)	0.20 AUc (24 h, PC3-PIP)	0.25 AUc (24 h, PC3-PIP)	1 nmol	Qualitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 0.25 AUc, liver 0.03 AUc)	Data not provided	LogP = 3.81a	Data not provided	Yes	IRDye800RS (825 g/mol)	λabs = 774 nm, λem = 789 nm
IRDye800RS-3101	ki = 4 ± 0.5 pM (NAAG)	0.29 AUc (24 h, PC3-PIP)	0.29 AUc (24 h, PC3-PIP)	1 nmol	Qualitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 0.29 AUc, liver 0.01 AUc)	Data not provided	LogP = 5.76a	Data not provided	Yes	IRDye800RS (825 g/mol)	λabs = 774 nm, λem = 789 nm
ICG-1101	ki = 700 ± 10 pM (NAAG)	0.05 AUc (24 h, PC3-PIP)	0.05 AUc (24 h, PC3-PIP)	1 nmol	Qualitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 0.05 AUc, liver 0.03 AUc)	Data not provided	LogP = 5.65a	Data not provided	Yes	ICG (714 g/mol)	λabs = 780 nm, λem = 800 nm
ICG-2101	ki = 400 ± 60 pM (NAAG)	0.08 AUc (24 h, PC3-PIP)	0.10 AUc (24 h, PC3-PIP)	1 nmol	Qualitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 0.10 AUc, liver 0.05 AUc)	Data not provided	LogP = 4.45a	Data not provided	Yes	ICG (714 g/mol)	λabs = 780 nm, λem = 800 nm
ICG-3101	ki = 200 ± 5 pM (NAAG)	0.29 AUc (24 h, PC3-PIP)	0.22 AUc (24 h, PC3-PIP)	1 nmol	Qualitative; 1, 2, 4, 24 h	Renal (24 h, kidneys 0.22 AUc, liver 0.05 AUc)	Data not provided	LogP = 6.40a	Data not provided	Yes	ICG (714 g/mol)	λabs = 780 nm, λem = 800 nm
DUPA-IRDye800CW46	kD = 12 nM (2-PMPA)	No ex vivo quantification, no scale bars	No ex vivo quantification, no scale bars	10 nmol	No ex vivo quantification, no scale bars	No ex vivo quantification, no scale bars	Data not provided	LogP = 0.62a	Data not provided	Yes	IRDye800CW (983 g/mol)	λabs = 774 nm, λem = 789 nm
ZJ-MCC-dEdEdEGK (IRDye800CW)G106	Data not provided	No ex vivo quantification, no scale bars	No ex vivo quantification, no scale bars	2–10 nmol	Qualitative; 1, 2, 3, 4 h	No ex vivo quantification, no scale bars	Data not provided	LogP = −2.67a	Data not provided	Yes	IRDye800CW (983 g/mol)	λabs = 774 nm, λem = 789 nm
PSMA-1-IRDye80099	IC50 = 1.5 ± 0.1 nM (EuC)	7.12%ID (120 h, PC3-PIP) 28, 35, 52, 58, 71, 65, 60,50, 38, 31, 35, 22 (at 0.8, 0.17, 0.5, 1, 2, 6, 8, 24, 48, 72, 96, and 120 h (in vivo)	0.004 AUc (120 h, PC3-PIP)	1 nmol	Qualitative; 0.8, 0.17, 0.5, 1, 2, 6, 8, 24, 48, 72, 96, and 120 h (in vivo), Qualitative; 4 h (ex vivo)	Renal (120 h, kidneys 0.004 AUc, liver 0.002 AUc)	Data not provided	LogP = −2.14 ± 0.17	Data not provided	Yes	IRDye800CW (983 g/mol)	λabs = 774 nm, λem = 789 nm
EUKL-cRGDfK-IRDye800107	Data not provided	1.53 AUc (24 h, PC3-PIP)	1.53 AUc (24 h, PC3-PIP)	1 nmol	Qualitative, 24 h	Renal (24 h, kidneys 1.53 AUc, liver 0.5 AUc)	Data not provided	LogP = −2.21a	Data not provided	No	IRDye800CW (983 g/mol)	λabs = 774 nm, λem = 789 nm
IRDye800CW-SCE104	Data not provided	160 AUb, 140 AUb (12 h, 24 h, LNCaP)	No ex vivo quantification, no scale bars	10 nmol	Qualitative; 12 h, 24 h	No ex vivo quantification, no scale bars	Data not provided	LogP = 0.94a	Data not provided	No	IRDye800CW (983 g/mol)	λabs = 774 nm, λem = 789 nm
KUE-PEG4-ZW800 + 3C105	Data not provided	4 AUb (4 h, LNCaP)	13 AUb (4 h, LNCaP)	10 nmol	Quantitative; 4 h	Renal (4 h, kidneys 13 AUb, liver 2.4 AUb)	t1/2, biol = 4.8 h	LogP = 1.39a	Data not provided	Yes	ZW800–1 (997 g/mol)	λabs = 772 nm, λem = 788 nm
OTL78108	kD = 4.7 nM (EuK-FITC)	7.5 AUc, 15 AUc, 17.5 AUc, 11 AUc, 15 AUc, 15 AUc (2 h, 22Rv1 for 1, 3, 10, 30, 60, 90 nmol)	1.3 AUc, 1.8 AUc, 4.1 AUc, 5.9 AUc, 4.2 AUc, 1.0 AUc (2 h, 22Rv1 for 1, 3, 10, 30, 60, 90 nmol)	1, 3, 10, 30, 60, 90 nmol	Qualitative; 2 h	Renal (2 h, kidneys 4.1 AUc, liver 0 AUc, for 10 nmol)	Data not provided	LogP = 1.19a	Data not provided	Yes	S0456 (983 g/mol)	λabs = 774 nm, λem = 789 nm

^aValues were calculated by using ChemAxon’s Chemicalize plugin.

^bEstimated from graph.

^cEstimated from scale bar.

^dNormalized to muscle tissue. AU = arbitrary units.

Scheme 2. Chemical Structures of Visible to Far-Red Fluorescent Imaging Agents for PSMA-Targeted Fluorescence-Guided Surgery.

Targeting vector (green) and imaging label (red) were colored accordingly.

Scheme 3. Chemical Structures of Near-Infrared Imaging Agents for PSMA-Targeted Fluorescence-Guided Surgery.

Targeting vector (green) and imaging label (red) were colored accordingly.

Figure 3.

(A) In vitro uptake of IRDye800CW-SCE in PSMA-positive LNCaP cells (top) with blocking by 2-PMPA (bottom) (adapted from Matusoka et al.¹⁰⁴ with permission from Elsevier). Example of in vivo imaging and typical biodistribution for (B) IRDye800CW-1 (adapted from Chen et al.)¹⁰¹ and (C) Cy5.5(SO₃^–)₄-EuK (adapted from Kwon et al.)¹⁰³

Visible to Far-Red Fluorescent PSMA-Targeted Tracers

Since evaluation conditions varied in, e.g., dose and time points, insights can only be inferred from direct comparisons that have been made in individual studies. For the same reason, it is difficult to discuss the findings of studies that reported individual compounds, i.e., CYUE-Rhodamine B,⁹⁸PSMA-1-Cy5.5,⁹⁹ and YC-9.¹⁰⁰

By standardizing the EuK targeting vector and Phe-Phe-Dap-Asp-Cys-maleimido-2-aminoethyl spacer, Kelderhouse et al.⁴⁶ were able to report the impact that the fluorophores Dylight 680 and AlexaFluor 647 have on PSMA affinity (DUPA-Dylight 680k_D = 33 nM, DUPA-Alexa Fluor 647k_D = 4.5 nM). These findings indicated that the fluorophore exerts an influence, but due to a lack of additional data, it is difficult to assess which exact feature of the chemical structure causes this influence.

The influence of spacers was studied for analogs that make use of Cy5.5-tetrakissulfonate by direct comparison.¹⁰¹ This study indicated that introducing a spacer has a positive effect on the affinity (from k_i = 90 ± 40 pM for Cy5.5–1 to k_i = 50 ± 20 pM for Cy5.5–2 and Cy5.5–3), but that variations in spacer type predominantly affected the lipophilicity (Cy5.5–2 LogP = −1.08 to Cy5.5–3, LogP = 3.00). Fluorescence imaging results suggested that tumor uptake is increased when the spacers are elongated. Dimerization studies by Kwon et al. underlined the influence of the spacer on the lipophilicity, but also suggested that increasing the number of targeting moieties has minimal impact on the affinity when comparing Cy5.5(SO₃^–)₄-EuK (LogP = −1.36, k_i = 0.6 ± 0.1 nM) with Cy5.5(SO₃^–)₄-(EuK)₂ (LogP = −0.33, k_i = 0.1 ± 0.0 nM) and (Cy5.5(SO₃^–)₄-4-(aminomethyl)benzoic acid-(EuK)₂; LogP = 1.32, k_i = 0.4 ± 0.1 nM)¹⁰² A follow-up study showed similar effects on the lipophilicity, but the bi-2-nitroimidazole-containing spacer seemed to limit the affinity.¹⁰³ Surprisingly, this chemical alteration did not have a clear impact on the fluorescence imaging findings.

Near-Infrared Fluorescent PSMA-Targeted Tracers

For fluorescence-guided surgery applications, near-infrared Cy7 analogs have been most commonly employed. For the compounds KUE-PEG₄-ZW800 + 3C,¹⁰⁵IRDye800CW-SCE,¹⁰⁴ZJ-MCC-dEdEdEGK(IRDye800CW)G,¹⁰⁶ and EUKL-cRGDfK-IRDye800(107) the affinity has not been reported, despite the availability of in vivo data suggesting that the compounds could be used to image PSMA-positive lesions. DUPA-IRDye800CW has an affinity of k_D = 12 nM, but no in vivo quantification was provided.⁴⁶ Individual studies with OTL78 (LogP = 1.19, k_D = 4.7 nM)¹⁰⁸ and PSMA-1-IRDye800 (LogP = −2.14, IC₅₀ = 1.5 nM)⁹⁹ provide promising in vivo findings, but further comparison is hampered by the fact that a reference compound has not been assessed and the chemical composition varies (Table 3, Scheme 3).

By varying spacers and fluorophores, Chen et al.¹⁰¹ examined how these molecular traits impacted tracer performance. For the fluorophores IRDye800CW, IRDye800RS, and indocyanine green (ICG), the PSMA affinity was highest with the longest Ahx-Lys spacer (k_i = 20, 4, and 5 pM, respectively), indicating that steric hindrance is reduced when a longer spacer is used. Cy7 displayed the highest affinity without the use of a spacer (k_i = 1 pM), suggesting this fluorophore yields little steric hindrance. No general trends could be inferred from in vivo fluorescence imaging findings or lipophilicity. During in vivo imaging, the fluorescence signal of the IRDye800CW analogs was substantially higher than that of the other analogs. This finding is surprising, as the brightness of Cy7 analogs generally does not differ to such an extent.¹⁰⁹

Studies that compared different tracer designs under the same conditions allowed for the isolation of general performance trends; not only does the length of the spacer play a role, but its composition also affects affinity. Furthermore, the structure of the fluorophore clearly impacts the tracer performance, e.g., steric hindrance in the secondary binding pocket.

Hybrid Tracers for PSMA

The beneficial traits of both radio- and fluorescent guidance can be integrated by creating hybrid tracers that incorporate a radiolabel and a fluorophore in a single molecule.¹¹⁰ To date, small-molecule hybrid PSMA-targeted tracers have only been reported in the preclinical setting. Next to structural differences (fluorophore, spacer), different radioisotopes have been included. (Table 4, Scheme 4, Figure 4).

Table 4. PSMA-Targeted Hybrid Tracers for Image-Guided Surgery.

compound	receptor affinity (competitor)	tumor-to-background (time point, cell line, background tissue)	kidney-to-tumor (time point, cell line)	injected dose	biodistribution; time point	clearance pathway (time point, uptake clearance organs)	half-life	lipophilicity	plasma protein binding	optimized for entrance funnel?	imaging label (molecular weight)	imaging signature(s)
IRDye800CW-Lys-DOTA-Gly-urea-Lys111	Ki = 1.2 ± 0.1 (Not described)	11.58, 19.00, 18.22, 36.50 (1, 2, 5, 24 h, PC3-PIP, muscle)	7.48, 7.29, 7.38, 3.29 (1, 2, 5, 24 h, PC3-PIP)	0.1 nmol (370 MBq/nmol)	Yes; 1, 2, 5, 24 h	Renal (1 h, kidneys 104 ± 9.4%ID/g; liver 2.4 ± 0.1%ID/g, 2 h, kidneys 97.1 ± 13.0%ID/g, liver 2.2 ± 0.0%ID/g, 5 h, kidneys 121.7 ± 21.6%ID/g, liver 3.2 ± 1.2%ID/g, 24 h, kidneys 48.1 ± 17.7%ID/g, liver 3.1 ± 0.4%ID/g)	t1/2, rad = 67 h	LogP = 0.12a	Data not provided	Yes	DOTAGA, 111In (387 g/mol); IRDye800CW (983 g/mol)	γ (Eγ = 171.3, 245.4 keV); λabs = 774 nm, λex = 789 nm
68Ga-PSMA I&F75	IC50 = 10.5 ± 2.7 nM ((I-BA)-KuE)	9.00 (1 h, LNCaP, muscle)	23.56 (1 h, LNCaP)	0.2 nmol (66 MBq/nmol) (2.0 nmol fluorescence imaging)	Yes; 1 h	Renal (1 h, kidneys 105.8 ± 22.7%ID/g, liver 0.9 ± 0.1%ID/g)	t1/2, rad = 1.1 h	LogP = −3.40	94%	Yes	DOTAGA, 68Ga (459 g/mol); Cy5 (625 g/mol)	γ (Eγ = 511 keV); λabs = 648 nm, λex = 664 nm
68Ga-Glu-urea-Lys-HBED-CC-FITC113	IC50 = 11.14 ± 1.16 nM (68Ga-EuK[(Ahx)]2-HBED-CC)	1.06 (1 h, PC3-PIP, muscle)	104 (1 h, PC3-PIP)	60 pmol (17–33 MBq/nmol) 0.5 nmol (fluorescence imaging)	Yes; 1 h	Renal (1 h, kidneys 138 ± 39%ID/g, liver 1.7 ± 0.5%ID/g)	t1/2, rad = 1.1 h	logD = −2.98 ± 0.09	Data not provided	Yesd	HBED-CC, 68Ga (499 g/mol); Fluorescein (346 g/mol)	γ (Eγ = 511 keV) λabs = 494 nm, λex = 512 nm
68Ga-Glu-urea-Lys-HBED-CC-AlexaFluor488113	IC50 = 13.32 ± 0.83 nM (68Ga-EuK[(Ahx)]2-HBED-CC)	1.35 (1 h, PC3-PIP, muscle)	69 (1 h, PC3-PIP)	60 pmol (17–33 MBq/nmol) 0.5 nmol (fluorescence imaging)	Yes; 1 h	Renal (1 h, kidneys 125 ± 32%ID/g, liver 1.0 ± 0.3%ID/g)	t1/2, rad = 1.1 h	logD = −2.01 ± 0.39	Data not provided	Yesd	HBED-CC, 68Ga (499 g/mol); AlexaFluor488 (505 g/mol)	γ (Eγ = 511 keV) λabs = 490 nm, λex = 525 nm
68Ga-Glu-urea-Lys-HBED-CC-IRDye800CW113	IC50 = 24.54 ± 5.70 nM (68Ga-EuK[(Ahx)]2-HBED-CC)	1.15, 4.05, 12.88 (1, 2, 6 h, PC3-PIP, muscle)	63.78, 15.11, 24.35 (1, 2, 6 h, PC3-PIP)	60 pmol (17–33 MBq/nmol) 0.5 nmol (fluorescence imaging)	Yes; 1, 2, 6 h	Renal (1 h, kidneys 205 ± 56%ID/g, liver 2.8 ± 0.5%ID/g, 2 h, kidneys 120 ± 5.1%ID/g, liver 1.2 ± 0.1, 6 h, kidneys 160 ± 28%ID/g, liver 1.0 ± 0.2)	t1/2, rad = 1.1 h	logD = −2.21 ± 0.36	Data not provided	Yesd	HBED-CC, 68Ga (499 g/mol); IRDye800CW (983 g/mol)	γ (Eγ = 511 keV) λabs = 774 nm, λex = 789 nm
68Ga-Glu-urea-Lys-HBED-CC-DyLight800113	IC50 = 21.41 ± 1.90 nM (68Ga-EuK[(Ahx)]2-HBED-CC)	1.46 (1 h, PC3-PIP, muscle)	52.11 (1 h, PC3-PIP)	60 pmol (17–33 MBq/nmol) 0.5 nmol (fluorescence imaging)	Yes; 1 h	Renal (1 h, kidneys 221 ± 24%ID/g, liver 6.2 ± 1.1%ID/g)	t1/2, rad = 1.1 h	logD = −2.95 ± 0.03	Data not provided	Yesd	HBED-CC, 68Ga (499 g/mol); DyLight800 (unknown)	γ (Eγ = 511 keV) λabs = 770 nm, λex = 794 nm
99mTc-EuK-Cy5-mas3114	IC50 = 175.3 ± 61.6 nM ((I-BA)-KuE)	18.95 ± 14.66 (2 h, LNCaP, muscle)	2.99 (2 h, LNCaP)	1 nmol (40 MBq/nmol)	Yes; 2 h	Renal (2 h, kidneys 9.0 ± 5.6%ID/g, liver 1.8 ± 0.8%ID/g)	t1/2, rad = 6.5 h	LogP = −2.13 ± 0.10	88%	Yes	mas3, 99mTc (408 g/mol); Cy5 (523 g/mol)	γ (Eγ = 140.5 keV); λabs = 648 nm, λex = 664 nm
99mTc-EuK-(SO3)Cy5-mas3114	IC50 = 19.2 ± 5.8 nM ((I-BA)-KuE)	5157 ± 949 (2 h, LNCaP, muscle)	1.20 (2 h, LNCaP)	1 nmol (40 MBq/nmol)	Yes; 2 h	Renal (2 h kidneys 18.4 ± 8.5%ID/g, liver 0.6 ± 0.5%ID/g)	t1/2, rad = 6.5 h	LogP = −2.86 ± 0.05	85%	Yes	mas3, 99mTc (408 g/mol); Cy5 (602 g/mol)	γ (Eγ = 140.5 keV); λabs = 648 nm, λex = 664 nm
99mTc-EuK-Cy5(SO3)-mas3114	IC50 = 118.8 ± 117.4 nM ((I-BA)-KuE)	0.46 ± 0.28 (2 h, LNCaP, muscle)	0.00 (2 h, LNCaP)	1 nmol (40 MBq/nmol)	Yes; 2 h	Renal (2h kidneys 10.5 ± 12.8%ID/g, liver 0.5 ± 0.1%ID/g)	t1/2, rad = 6.5 h	LogP = −2.70 ± 0.01	87%	Yes	mas3, 99mTc (408 g/mol); Cy5 (602 g/mol)	γ (Eγ = 140.5 keV); λabs = 648 nm, λex = 664 nm
99mTc-EuK-(Ar)Cy5-mas3114	IC50 = 113.1 ± 35.9 nM ((I-BA)-KuE)	172 ± 84.95 (2 h, LNCaP, muscle)	4.51 (2 h, LNCaP)	1 nmol (40 MBq/nmol)	Yes; 2 h	Renal (2 h, kidneys 15.6 ± 20.0%ID/g, liver 3.3 ± 2.2%ID/g)	t1/2, rad = 6.5 h	LogP = −1.75 ± 0.10	91%	Yes	mas3, 99mTc (408 g/mol); Cy5 (612 g/mol)	γ (Eγ = 140.5 keV); λabs = 648 nm, λex = 664 nm
99mTc-EuK-Cy5(Ar)-mas3114	IC50 = 164.4 ± 76.9 nM ((I-BA)-KuE)	5.79 ± 0.86 (2 h, LNCaP, muscle)	2.87 (2 h, LNCaP)	1 nmol (40 MBq/nmol)	Yes; 2 h	Renal (2 h, kidneys 11.9 ± 12.3%ID/g, liver 9.6 ± 6.8%ID/g)	t1/2, rad = 6.5 h	LogP = −1.84 ± 0.06	89%	Yes	mas3, 99mTc (408 g/mol); Cy5 (612 g/mol)	γ (Eγ = 140.5 keV); λabs = 648 nm, λex = 664 nm
18F-(SO3)Cy3(SO3)-EuK115	EC50 = 10.3 ± 0.7 nM (Unlabeled(SO3)Cy3(SO3)-EuK)	128 (2 h, PC3-PIP, blood)	0.06 (2 h, PC3-PIP)	2.5 nmol (7.0 ± 1.9 MBq/nmol)	Yes; 2 h	Renal (2 h, kidneys 1.0 ± 0.5%ID/g, liver 0.1 ± 0.0%ID/g)	t1/2, rad = 1.8 h	LogP = 1.02a	Data not provided	No	18F (278 g/mol); Cy3 (682 g/mol)	β (Emaxβ+ = 634 keV) λabs = 550 nm, λex = 570 nm

^aValues were calculated by using ChemAxon’s Chemicalize plugin.

^bEstimated from graph.

^cEstimated from scale bar. AU = arbitrary units.

^dNo spacer as opposed to parental compound PSMA-11.⁷⁸

Scheme 4. Chemical Structures of Hybrid Tracers for PSMA-Targeted Image-Guided Surgery.

Targeting vector (green) and imaging label (red) were colored accordingly.

Figure 4.

(A) Maximum-intensity projections of small-animal PET imaging of athymic nude mice bearing LNCaP tumors using 0.5 nmol ⁶⁸Ga-HBED-CC-IRDye800CW. (B) Use of HBED-CC-IRDye800CW in vivo for fluorescence imaging with a Da Vinci Firefly laparoscope (1 mg, 30 μg/kg; both adapted from Baranski et al.;¹¹³ in JNM). (C) Maximum-intensity projections of small-animal SPECT imaging of BALB/c nude mice bearing LNCaP tumors using 1.0 nmol ^99mTc-EuK-(SO₃)Cy5-COOH (adapted from Hensbergen et al. in JNM).¹¹⁴ (D) Use of ^99mTc-EuK-(SO₃)Cy5-COOH (100 μg, 2.85 μg/kg) for in vivo fluorescence imaging with a prototype Karl Storz camera that allows Cy5 imaging.

The first hybrid PSMA-targeted tracer design, IRDye800CW-Lys-DOTA-Gly-urea-Lys, has been introduced by Banerjee et al.¹¹¹ Here, both IRDye800CW and DOTA (to chelate ¹¹¹In) were conjugated to a distal EuK moiety using a Lys-Lys-Suberic acid backbone, a design that is in line with the multifunctional single-attachment point (MSAP) principle coined by Josephson et al.¹¹² Although this particular spacer design does not seem to actively promote binding in the secondary binding pocket, but rather aims to minimize steric hindrance by the fluorophore, a high affinity was reported for IRDye800CW-Lys-DOTA-Gly-urea-Lys (k_i = 1.2 ± 0.1 nM). SPECT imaging showed clear tumor visualization and in vivo biodistribution data (%ID/g assessment at 0.1 nmol) indicated tumor-to-background ratios in the range of 11–37 at time points between 1–24 h. Fluorescence imaging at the same dose (but different imaging time points) allowed for clear tumor visualization.¹¹¹ The same MSAP design concept was later used to introduce a bisulfonated cyanine pentamethine (−(SO₃)Cy5(SO₃−) fluorophore and DOTAGA chelate to a D-(3-iodo)Tyr-D-(3-iodo)Tyr spacer optimized for secondary binding,⁴⁸ yielding ⁶⁸Ga-PSMA I&F.⁷⁵ The high affinity of this tracer (IC₅₀ = 10.5 ± 2.7 nM) translated to a tumor-to-background ration of 9 %ID/g (0.2 nmol) and tumor visualization using SPECT imaging (0.2 nmol) and fluorescence imaging (at 2 nmol) using a customized Firefly laparoscope.

Baranski et al.¹¹³ have applied a design concept wherein the HBED-CC chelate was used to link the EuK binding moiety with a PEG spacer containing a variety of fluorophores. This ⁶⁸Ga-based hybrid design was derived from ⁶⁸Ga-PSMA-11 (Table 1, Scheme 1),⁷⁸ but the 6-aminohexanoic acid spacer between HBED-CC and EuK was removed. Varying the fluorophore (FITC, AlexaFluor488, IRDye800CW, or DyLight800) did not significantly impact the logD values [(2.01 ± 0.39)–(2.98 ± 0.09)], but did impact the PSMA affinity (IC₅₀ = 11.14 ± 1.16, 13.32 ± 0.83, 24.54 ± 5.70, and 21.41 ± 1.90 nM, respectively). Quantitative biodistribution using the ⁶⁸Ga isotope (%ID/g assessment at 60 pmol) at 1 h p.i. showed that the structural variations also extended to renal retention (138 ± 39, 125 ± 32, 205 ± 56, and 221 ± 24%ID/g, respectively), which could impact the performance of tumor imaging as discussed previously. This particular comparison allows for the isolation of the influence that a fluorophore exerts on tracer performance. In mice, ⁶⁸Ga-Glu-urea-Lys-HBED-CC-IRDye800CW could be used to identify PSMA-positive tumors with small-animal PET (Figure 4A; 0.5 nmol) or fluorescence imaging (0.5 nmol). Interestingly, basal prostate uptake of Glu-urea-Lys-HBED-CC-IRDye800CW in healthy pigs (1 mg, 30 μg/kg) was evaluated using a Da Vinci Firefly laparoscope (Figure 4B).¹¹³

Alternatively, the use of cyanine fluorophores as spacers has been explored. By using the fluorophore as spacer but keeping the primary targeting vector and mas₃ chelate consistent, it became possible to systematically study if and how a custom bifunctional Cy5 fluorophore (i.e., −Cy5–, −(SO₃)Cy5–, −Cy5(SO₃)–, −(Ar)Cy5–, and −Cy5(Ar)−) could influence secondary binding to the amphipathic entrance funnel.¹¹⁴ Affinities were greatly impacted by the fluorophore design: of the five fluorophores studied only −(SO₃)Cy5– showed IC₅₀ < 100 nM (EuK-(SO₃^–)Cy5-mas₃, IC₅₀ = 19.2 ± 5.8 nM). These results imply that the presence and location of an anionic sulfonate moiety on the fluorophore influence the affinity for PSMA. At 2 h p.i. using a dose of 1 nmol, ^99mTc-EuK-(SO₃^–)Cy5-mas₃ displayed a relatively low renal retention (18.4 ± 8.5%ID/g), as well as high tumor-to-muscle and good kidney-to-tumor ratios (Figure 4C, Table 4). This compound is currently under evaluation in a healthy pig model where it allows accurate prostate delineation using a microdosing regimen (100 μg, 2.85 μg/kg; Figure 4D). Kommidi et al.¹¹⁵ have used the commercially available HOOC-(SO₃)Cy3(SO₃)-COOH to create ¹⁸F-(SO₃)Cy3(SO₃)-EuK. In line with the above findings, this compound has a high affinity (EC₅₀ = 10.3 ± 0.7 nM) and tumors could be clearly visualized using small-animal PET imaging. Quantitative biodistribution studies indicated a high tumor uptake (tumor-to-blood ratio = 128) and low renal retention (kidney-to-tumor ratio 0.06 at 2 h) using a dose of 2.5 nmol. Here, it should be noted the fact that ¹⁸F-(SO₃)Cy3(SO₃)-EuK is radiohalogenated, which implies it has a reduced amount of charges compared to radiometalated ligands, and as such can influence renal accumulation and/or clearance.¹¹⁶

In summary, three types of hybrid tracer designs have been reported: (1) MSAP, (2) chelate-spacer-based, and (3) fluorophore-spacer-based. In general, these studies indicate that the affinity of the hybrid tracers is impacted by the addition of a fluorophore. This effect, however, is most substantial when the fluorophore resides within the secondary binding pocket. In some cases, the dose used for the biodistribution in%ID/g was different from the dose used for fluorescence imaging, making it difficult to assess in vivo performance for the latter application.

Discussion

By combining the binding characteristics of PSMA with the clinical need for accurate and sensitive detection of PSMA-positive lesions during PCa surgery, a general perspective could be created for the evaluation of the three families of PSMA-targeted tracers for image-guided surgery (i.e., radioactive, fluorescent, and hybrid). This overview suggests the “low-hanging fruit” in PSMA tracer design has been picked by incorporating off-the-shelf imaging labels and spacers on a well-known targeting moiety. Although the majority of PSMA-targeted tracers designed for image-guided surgery applications are fluorescent (Tables 2, 3, Schemes 2, 3) or hybrid (Table 4, Scheme 4) of nature, only radiotracers (Table 1, Scheme 1) have been reported in clinical trials so far. Hence, it does not seem that we are currently getting the most out of PSMA-targeted tracer developments for image-guided surgery purposes.

Published clinical data with ^99mTc-PSMA I&S focuses on the positive influence image-guided surgery has on the resection of tumor-positive LNs in the salvage setting (after staging with PSMA–PET).^57,96 Due to the lack of reports that describe the use of these radiotracers in primary cancer resection, it is difficult to judge the potential of PSMA-targeted image guidance for this application. Most likely, resection of PSMA-positive LNs in the primary setting will be equally as effective as in a salvage setting, but given the overall renal clearance and high renal accumulation of the PSMA-targeted tracers reported, the ability to detect primary tumor margins is likely more reliant on the chemical design of the tracer.

We have attempted to comprehend the requirements for the chemical design of a high-end tracer for PSMA-targeted image-guided surgery, but comparative interstudy performance assessment was impaired by a number of features: (1) lack of a standardized methodology for affinity assessment (k_i, k_D, IC₅₀; in some cases picomolar range affinities have been reported, while comparative studies yield a lower affinity (nanomolar range) for similar designs); (2) although it is known that the applied dose influences the biodistribution of PSMA-targeted tracers,^117,118 large variations in dose have been reported between, e.g., fluorescence- and radioactivity-based studies (0.1–1 nmol for radio- and hybrid tracers, 1–10 nmol for fluorescent tracers; Table 1–3); (3) differences in time points of imaging (1–120 h; Table 1–3), combined with the biological clearance profiles of individual tracers, are common and are likely to impact the intensity of background signals; (4) missing quantitative biodistribution data (%ID/g) for fluorescence tracers (Table 4) makes it particularly challenging to place the reported preclinical findings in perspective to the PSMA–PET tracers that are currently used for staging of patients; (5) use of different animal/tumor models complicates a direct comparison based on tumor uptake values—it has been shown that the use of different tumor models can increase the tumor uptake by almost 3-fold,¹¹⁶ it might cause one to wonder how accurately proof-of-concept studies in mice represent the challenges encountered during PCa surgery; (6) utility of radioactive and fluorescent signatures for imaging is intrinsically different, further complicating a comparison of the findings (Table 5).

Table 5. Differences between Fluorescence- and Radioactivity-Based Image-Guided Surgery.

	fluorescence	radioactivity
Spatial resolution	μm–mm’s	>2 mm
Signal display	Video rate	Only numeric/acoustic (γ-probe) or static image (γ camera)
Acquisition time	<50 ms	>500 ms (γ-probe), multiple seconds (γ camera)
Compatibility with microscopic tissue evaluation	Fluorescence microscopy (<μm resolution)	Autoradiography (>mm)
Biodistribution	Qualitative (image intensity only; impacted by autofluorescence and tissue-based signal attenuation)	Quantitative (%ID/g)
Signal penetration depth through tissue	<10 mm	>10 mm
Noninvasive whole-body imaging (surgical roadmap)	n.a.	2D scintigraphy/3D (e.g., SPECT/CT, PET/CT)

Despite the hampering of interstudy comparisons, a number of manuscripts cited in this review (one regarding radiotracers, four regarding fluorescent tracers and two regarding hybrid tracers) allowed for a structural intrastudy comparison of different tracer analogs. These studies therefore complement the existing radiochemistry literature on PSMA-targeted tracer refinement.⁵ Key performance characteristics could be inferred from these combined data, e.g., that the spacer can be used to influence the binding to the secondary pocket and/or to limit steric hindrance from “bulky” imaging labels.^102,103 Hereby the molecular features of different fluorophores exert a different impact on affinity and tracer kinetics, similar to what has been reported for chelates/radiolabels.^62,63 The influence of the molecular composition of fluorophores was largest when short spacers were used (meaning the fluorophores were placed in the amphipathic entrance funnel),¹⁰¹ or when the fluorophores were purposely exploited as spacers.^113,114 In the secondary binding pocket, cyanine fluorophore structures with sulfonates instead of additional aromatic rings yielded the highest affinity. Furthermore, it is interesting to note that a study by Wang et al. underlines that Cy5 fluorophores are not always outperformed by Cy7 analogs.⁹⁹

Besides strides that can be taken in optimization of the chemical design, clinical translation of fluorescent tracers is also affected by the sensitivity of detection. The translation of radiotracers such as ^99mTc-PSMA I&S was supported by the ability to apply these tracers in a microdosing regime, a concept that limits toxicity and eases the clinical translation of radiotracers.¹¹⁹ Despite clinical data indicating that fluorescence-guided surgery might also be feasible adhering to the microdosing principle,^73,120,121 as has also been suggested for PSMA using a porcine model (Figure 4D), the application of microdosing remains controversial in the field of fluorescence-guided surgery.¹²² It is certain, however, that intraoperative identification of (fluorescent) lesions is less efficient when they contain low tracer concentrations.⁷³ Other than PSMA–PET studies reporting on a >2 mm size limit for lesion identification,⁵⁵ little is known about the PSMA concentration on cancer cells in primary or metastatic lesions. Currently, analysis of the PSMA–PET roadmap is one of the clinical selection criteria for the inclusion of patients in PSMA-guided salvage surgery.²⁰ One might argue that inclusion of quantitative analysis (e.g., SUV) of the lesion is these scans, coupled to quantitative surgical detection and pathological analysis, could in the future help define which PSMA concentration levels are needed to support radio- and/or fluorescence-based lesion identification in a surgical setting.

Given the clinical impact of PSMA-targeted radioguided surgery, the limited insights provided for the in vivo performance of fluorescent PSMA-targeted tracers and literature suggesting that tissue-induced signal attenuation, e.g., as result of a high body mass index, limits the accuracy of fluorescence-guided surgery.^123,124 It is likely that PSMA-targeted tracers incorporating a radiolabel are most promising. Here, the “best-of-both-worlds” scenario offered by hybrid tracers (combination of features in Table 5) seems to provide a logical means to support preoperative imaging and radioguided surgery, while at the same time satisfying the surgeon’s desire for fluorescence guidance.¹¹⁰

Next to the small-molecule PSMA-tracers, antibody-based efforts are also explored for this molecular target.^125−129 In the future, these different tracer designs can jointly address PSMA-based resections in PCa and given the biology of PSMA, they can possibly also guide resections in other cancers. When PCa is PSMA-negative or PSMA-expression is low, other molecular targets may be used for surgical guidance, e.g., the fibroblast activation protein¹³⁰ or gastrin-releasing peptide receptor.¹³¹

Conclusion

Different PSMA-targeted tracers have been reported for the field of PSMA-targeted image-guided surgery. When comparing the reported radio-, fluorescent-, and hybrid PSMA-targeted tracers it becomes evident that (1) radiotracers have already created clinical impact, (2) fluorescent tracers are most abundant but are used at a relatively high dose and the preclinical studies lack pharmacokinetic assessment, and (3) hybrid tracers provide the potential to complement the beneficial values of radiotracers with fluorescence guidance.

The authors declare the following competing financial interest(s): F. W. B. van Leeuwen is Chief Innovation Officer at ORSI Academy, and H.-J. Wester is founder and shareholder of Scintomics GmbH.