Durvalumab was first produced using hybridoma technology and immunization of IgG2 and IgG4 XenoMouse mice models (60). Thereafter, the constant domain of the Ab was substituted for a human IgG1 domain with a mutated Fc region that leads to reduced antibody dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) (60). The same group reported an IC₅₀ value of 0.1 nmol/L in a competition assay with PD-1. Tan et al. (61) demonstrated a high binding affinity of the single chain Fv fragment of durvalumab toward human PD-L1 by SPR and measured a K_D of 0.667 nM. At these earlier timepoints of development, little information was known about the binding mechanism and interaction surface of these emerging mAbs. Lee et al. (62) reported some of the first co-crystallization results of PD-L1 with PD-1 and anti-PD-L1 mAbs, including durvalumab. Sixteen amino acid residues of PD-L1, primarily located within the central CC′FG β-sheet, CC′loop and N-terminal region, are responsible for the high affinity interaction of durvalumab (62). It was further shown that the variable domain of the heavy chain (VH) and light chain (VL) of durvalumab contribute to PD-L1 binding (61). Since its approval by the FDA in 2017, durvalumab has been a key role player in the development of immunotherapy toward improving patient outcomes, and results from clinical trials have been encouraging (16). Durvalumab has led to improved major pathological response (MPR) and disease-free survival (DFS) when administered as neoadjuvant and maintenance treatment for resectable and unresectable stage III non-small cell lung cancer (NSCLC) patients, respectively (63, 64). However, the clinical trial for resectable NSCLC patients receiving neoadjuvant durvalumab highlighted a need for better patient selection since an unexpectedly high mortality related to cardiovascular and respiratory comorbidities resulted in premature termination of the trial (63). Recent clinical PET-imaging studies have been initiated to test the feasibility and safety of using [⁸⁹Zr]Zr-DFO-durvalumab for imaging PD-L1 expression. In a study by Smit et al. (65) PET/CT results were compared for patients receiving [⁸⁹Zr]Zr-DFO-durvalumab only and patients receiving a 750 mg unlabeled durvalumab co-injection to reduce the tracer sink effect. However, due to the occupation of PD-L1 by a much higher dose of therapeutic durvalumab in the latter condition, overall tracer uptake was decreased. Furthermore, while they could show that [⁸⁹Zr]Zr-DFO-durvalumab PET/CT was well-tolerated in patients with NSCLC, they were unable to demonstrate a significant correlation between patient response and tracer uptake in lesions (65). Similarly, [⁸⁹Zr]Zr-DFO-durvalumab PET imaging proved well-tolerated in patients with squamous cell carcinoma of the head and neck (SCCHN) (66). In line with the study by Smit et al. (65) treatment response was also not correlated with tumor uptake or PD-L1 expression (66).

Atezolizumab was identified using phage display technology by screening phage libraries expressing human VL and VH against human and murine PD-L1 (67). The clone selected and developed to become atezolizumab bound both human and mouse PD-L1 with high affinities as measured by SPR (K_D = 0.2 and 0.6 nM, respectively) (67). Affinities measured with binding assays on recombinant human and mouse PD-L1-expressing HEK293 cells were reported as K_D = 0.4 and 0.1 nM, respectively (67). In the study performed by Lee et al. (62), structural details about the interaction of atezolizumab and PD-L1 were elucidated as well. Twenty-three amino acid residues in the central CC′FG β-sheet, BC, CC′, C′C′′ and FG loops are responsible for the tight interaction with atezolizumab (62). By alanine scanning studies, Zhang et al. (68) identified Glu58 and Arg113 of PD-L1 as two major contributors to the high affinity of atezolizumab. Interestingly, while PD-L1 backbone conformation remains rigid upon binding to anti-PD-L1 mAbs, only atezolizumab binding induces a slight change in the PD-L1 BC loop to allow more interactions with the mAb (62). Atezolizumab was the first FDA-approved PD-L1 inhibitor for the treatment of cancers patients with urothelial carcinoma, metastatic NSCLC and triple-negative breast cancer (TNBC) (69–71). Given the favorable results from the IMpower010 clinical trial (NCT02486718, Impower010), atezolizumab has recently been approved as adjuvant therapy following surgery and chemotherapy in patients with stage II to IIIA NSCLC based on PD-L1 expression score using the Ventana PD-L1 assay (SP263, Ventana Medical Systems) as companion diagnostic (72). As with other PD-L1 blocking mAbs, the therapy response rate is difficult to predict. Bensch et al. (73) performed the first-in-human PET imaging study using [⁸⁹Zr]Zr-DFO-atezolizumab to test feasibility and response prediction in metastatic urothelial carcinoma, NSCLC, and TNBC. Contrary to [⁸⁹Zr]Zr-DFO-durvalumab, this study reported that better tumor response correlated well with increased tumor tracer uptake, while PD-L1 IHC expression gave no indication of such a correlation (73). Currently, a phase II diagnostic imaging trial is underway to evaluate whether [⁸⁹Zr]Zr-DFO-atezolizumab PET/CT imaging can be used as a predictive tool to select patients with metastatic TNBC to receive PD-L1 inhibitors in addition to chemotherapy (74).

Unlike durvalumab and atezolizumab, avelumab is a human IgG1 designed without a modified Fc region and can therefore mediate ADCC (75). While the binding affinities of durvalumab and atezolizumab are somewhat comparable, avelumab binds human PD-L1 in the picomolar range (K_D = 47 and 42 nM, measured by SPR independently) (61, 75). Crystal structures of the PD-L1-avelumab complex have revealed that even though both VL and VH are involved, the interaction with PD-L1 is dominated by the VH of avelumab (75). Additionally, the C strand, C′ strand, F strand, G strand, and CC′ loop present in PD-L1 are primarily involved in interaction (75). Avelumab is a promising treatment option for advanced and metastatic urothelial carcinoma (76, 77). Results from a recent phase III clinical trial (NCT02603432) have demonstrated that OS at 1 year was significantly longer in patients receiving avelumab as maintenance treatment (71.3%) compared to the control group who received best supportive care alone (58.4%) (76). In addition, the patient population with PD-L1-positive tumors vs. PD-L1-negative tumors determined by the Ventana PD-L1 assay (SP263, Ventana Medical Systems), which qualitatively detects PD-L1 expression in histology tissues, showed a significant increase in OS, further supporting the requirement for assessment of PD-L1 status (76). Studies to assess PD-L1 expression in humans using non-invasive PET imaging and radiolabeled avelumab have been limited. Zirconium-89-labeled avelumab has been investigated primarily in preclinical studies in which high specificity and affinity to PD-L1 were demonstrated on PD-L1-positive MDA-MB-231 human breast adenocarcinoma cells (78). In addition, [⁸⁹Zr]Zr-DFO-avelumab was investigated in mice models bearing MDA-MB-231 tumors and could show the feasibility of imaging PD-L1 with this tracer (39, 78). A phase I clinical trial is currently underway to evaluate the feasibility of using [⁸⁹Zr]Zr-DFO-avelumab to assess PD-L1 expression in patients with NSCLC and whether prediction of avelumab treatment response is possible (NCT03514719, PINNACLE).

Nivolumab, developed and characterized by Bristol-Myers Squibb (BMS), is an anti-PD-1 mAb that was produced in humanized mice using hybridoma technology (79). The lead antibody clone was grafted onto a human IgG4 bearing an S228P mutation for increased stability and reduced ADCC and CDC activity (79). In vitro characterization showed that nivolumab could bind PD-1-expressing CHO cells and activated T cells with EC₅₀ = 1.7 and 0.6 nM, respectively (79). The crystal structures of nivolumab in complex with PD-1 have been studied extensively to gain more insight into the PD-1 epitope and molecular mode of inhibition (80–82). Tan et al. (80) solved the crystal structures of this complex and showed that both VH and VL of nivolumab are involved in binding. Residues present in the FG and BC loops of the IgV domain and the N-loop of PD-1 are responsible for interaction with nivolumab (80). Interestingly, most of the hydrogen bonds (10 of 16) are formed within the N-loop, making it the most dominant interacting region of nivolumab (80). This study further confirmed that, unlike previously speculated, N-glycosylation is not required for nivolumab binding (80). Results from clinical trials in recent years have shown encouraging results for a multitude of cancers, such as NSCLC, melanoma, and esophageal squamous-cell carcinoma, to name a few (21, 83–85). Compared to a chemotherapy treatment, docetaxel, nivolumab led to significant improvement in OS, objective response rate (ORR) and overall tolerability in NSCLC (83). It has also been shown that nivolumab as adjuvant therapy after resection in urothelial carcinoma and melanoma patients resulted in overall prolonged DFS and recurrence-free survival (RFS), respectively (21, 86). While treatment outcomes are promising, these studies also highlighted the current lack of methods to predict ICI response accurately. In a preclinical PET imaging study to determine tracer clearance and biodistribution in healthy cynomolgus monkeys, nivolumab was conjugated to DFO and subsequently labeled with zirconium-89 (87). [⁸⁹Zr]Zr-DFO-nivolumab uptake was visualized clearly in the spleen, an organ with distinct populations of PD-1-expressing DCs, as well as in the liver, evident of mAb catabolism (87). England et al. (88) further demonstrated in another preclinical study the feasibility and efficiency of imaging PD-1 on T cells in a humanized lung cancer mouse model with [⁸⁹Zr]Zr-DFO-nivolumab. Niemeijer et al. (41) performed the first-in-human PET imaging study using [⁸⁹Zr]Zr-DFO-nivolumab in patients with NSCLC. Similar to the study in cynomolgus monkeys, higher [⁸⁹Zr]Zr-DFO-nivolumab uptake was visualized in the spleen and the liver compared to other organs (41). A positive correlation between nivolumab treatment response at 3 months and [⁸⁹Zr]Zr-DFO-nivolumab peak standard uptake value (SUV_peak) at baseline prior to treatment could be drawn, however, the sample population for this study was small (13 patients) (41).

Merck & Co., Inc. developed pembrolizumab, an anti-PD-1 mAb able to block the PD-1/PD-L1 interaction with high affinity (K_D ≈ 29 pM) (89). Similar to nivolumab, pembrolizumab is a humanized mAb that is also based on a human IgG4 antibody containing the S228P mutation (89). Na et al. (90) investigated the molecular mode of action by solving the crystal structures of the human PD-L1-pembrolizumab complex. While the complementarity-determining regions (CDRs) and framework region (FR) of pembrolizumab are involved in the interaction, two key regions within PD-1 are responsible for contact (90). The first region consists of residues present in the C′D loop, and the second region consists of residues present in the C, C′, and F strands of PD-1 (90). The structural analysis performed by Tan et al. (80) also confirmed that while the interaction surfaces are close, there are no overlapping regions between nivolumab and pembrolizumab on the surface of human PD-1. The last decade has shown impressive clinical results for pembrolizumab as first-line treatment, monotherapy and neoadjuvant therapy for patients with PD-L1-expressing NSCLC, melanoma, and advanced gastric cancer who benefit from prolonged OS, manageable side effects and durable responses (91–95). A couple of preclinical studies have evaluated radiolabeled pembrolizumab in rodents (96, 97). England et al. (97) showed that pembrolizumab modified with p-SCN-deferoxamine and radiolabeled with zirconium-89 could accumulate in the salivary glands (containing PD-1-expressing T-cells) of humanized NSG mice engrafted with human peripheral blood mononuclear cells (PBMCs). In a similar study, Natarajan et al. (96) imaged PD-1 expression on tumor infiltrating lymphocytes (TILs) in humanized NSG mice bearing A375 human skin melanoma tumors with ⁸⁹Zr- and ⁶⁴Cu-labeled pembrolizumab. For both tracers, uptake in the tumor and spleen could be visualized, indicative of PD-1 detection (96). Another preclinical study was performed with healthy cynomolgus monkeys to better evaluate the potential clinical translation of radiolabeled pembrolizumab for tracking PD-1 expression (98). In this study, uptake of the tracer (pembrolizumab conjugated to N-suc-desferal-TFP ester and radiolabeled with zirconium-89) was visible in the expected lymphoid organs such as the spleen, lymph nodes, and tonsils (98). Clinical PET imaging with [⁸⁹Zr]-DFO-pembrolizumab in patients with NSCLC was able to detect tumor lesions and, notably, response to anti-PD-1 treatment could be correlated with higher [⁸⁹Zr]Zr-DFO-pembrolizumab uptake (42, 99).

While the therapeutic mAbs described above have made great strides in the field of immunotherapy, new mAbs are continuously being developed. A couple of emerging anti-PD-1 mAbs that have gained FDA approval in more recent years include cemiplimab (2018), dostarlimab (2021), retifanlimab (2023), and toripalimab (2023) (52). Similar to nivolumab and pembrolizumab, all are humanized mAbs based on human IgG4 (100–103). Structural analysis studies have revealed that the binding interface of cemiplimab and dostarlimab with PD-1 are highly similar and include the BC, C′D and FG loops present in PD-1 (104, 105). Interaction with toripalimab was shown to be dominated by the FG loop of PD-1, and binding was independent of glycosylation (106), while glycosylation of N58 in PD-1 improved the affinity of cemiplimab substantially (107). Similar to the other two anti-PD-1 mAbs, cemiplimab, dostarlimab, retifanlimab, and toripalimab can bind human PD-1 with high affinity as measured by SPR (K_D = 6.1, 0.3, 0.6, and 0.2 nM, respectively) (100, 101, 106, 108). Various studies have shown the clinical benefit and durable responses following treatment with these mAbs as monotherapy or combined with chemotherapeutic agents, and many clinical trials are still ongoing (103, 109–112). Considering the development and clinical assessment of these recently approved mAbs are still in the early stages, the number of imaging studies performed to date has been limited. Thus far, toripalimab is the only candidate from this group that have been explored as a PET (and SPECT) imaging tracer (113–115). Huang et al. (114) radiolabeled toripalimab with iodine-124 and evaluated the tracer in humanized PD-1-C57BL/6 mice bearing mouse sarcoma S180 tumors. [¹²⁴I]I-toripalimab uptake could be visualized in tumors and the highest tumor-to-organ ratios were obtained at 72 h p.i. (114). The encouraging results from the preclinical study prompted the first-in-human pilot clinical translation study by the same group (115). Twelve patients, having either melanoma or urologic cancer, were administered a single dose of [¹²⁴I]I-toripalimab (115). The tracer was well tolerated and proven safe in all patients, and peak uptake in tumor lesions was visualized 24 h p.i. (115). In addition, PET/MR was the preferred imaging modality over PET/CT for the detection of lesions located in the liver (115).

Since their serendipitous discovery 30 years ago, Nbs have been at the forefront of both treatment and diagnosis of a multitude of human diseases due to their unique and favorable biophysical properties (125, 126). Nbs are uniquely present in the serum of mammals belonging to the Camelidae family and are often referred to as single-domain antibodies or VHH because they are composed of only the variable (V)-domain of the heavy (H)-chain of a conventional IgG (124). Their small size of 15 kDa, well below the cutoff for glomerular filtration by the kidneys (∼50 kDa), makes Nbs suitable for applications requiring rapid tissue penetration or blood clearance, such as targeted drug delivery and imaging (127, 128). Another striking feature of Nbs is their excellent in vivo stability. This could be explained by the replacement of highly conserved and hydrophobic amino acids in the VH that would usually interact with the VL (for conventional immunoglobulins) with amino acids that are either smaller in size or are more hydrophilic (124, 129). Another attractive characteristic of Nbs is low immunogenicity, which is an important requirement for the clinical implementation of any pharmaceutical (130). Similar to a VH of a conventional Ab, three CDRs are present in Nbs. However, differences such as a longer H3 loop (so-called because of its position on CDR3 of a H-chain variable region), higher sequence conservation and solubility-enhancing mutations present in the FR, contribute to the high specificity and affinity (nM to pM) achieved by Nbs (131). Taken together, these properties are ideal for developing Nbs as PET imaging tracers. A couple of promising Nbs targeting the PD-1/PD-L1 axis have already been identified and are currently being investigated (132).

Monobodies, also known as Adnectins, form part of a group of molecular scaffold proteins based on domain type III of the 10th human fibronectin (10Fn3) (156). They are structurally comparable to the heavy chain variable domain of immunoglobulins and consist of an anti-parallel β-sheet sandwich (156). Their target-binding properties are primarily attributed to the shared similarity between the diversified loops connecting the two β-sheets situated on opposite poles of 10Fn3 and the CDRs within immunoglobulin variable domains (57). Sequences present in these loops, similar to Abs and Nbs, can be subjected to diversification, enabling screening by display technologies (57). Due to the lack of Cys residues in their sequence, monobodies do not require linking disulfide bridges between β-sheets for proper folding and stability, providing them with more favorable thermodynamic properties and enhanced structural integrity (57). Favorable biophysical properties, especially high-affinity binding ability, make monobodies ideal candidates for therapeutic and imaging applications. Monobodies are small in size (approximately 94 amino acid residues) and can allow fast clearance of radioactively labeled tracers, quick imaging, and rapid penetration of solid tumors (157). Furthermore, a lysine residue is situated on the pole opposite to the binding region, allowing straightforward thiol- or amine-conjugation of chelators and radiolabeling (156). To date, only a handful of monobodies have been developed as PET imaging tracers to assess PD-L1/PD-1 expression and dynamics (158).

A receptor, designated staphylococcal protein A (SPA), that is commonly found in the cell wall of S. aureus naturally binds to the Fc region of IgG and consists of five homologous domains (EDABC) (165). Domain Z is a modified analog of the SPA B domain, characterized by a helix bundle structure containing three α-helices and differs from its parent domain by a Gly29Ala substitution of helix two to increase stability (165). This domain forms the structural basis for another engineered scaffold protein, namely, an affibody (166). Unlike other binding scaffolds that make use of CDRs to introduce variability, the target binding residues of affibodies are located within solvent-exposed surfaces along two of the three α-helices and yield a flat-surface binding architecture (167). Moreover, the randomization of residues in the binding region of affibodies to obtain highly diverse combinatorial libraries is slightly more challenging, since the non-binding residues on the opposite side of the α-helix need to remain constant to preserve structural integrity and stability of the protein (166). Nevertheless, highly diverse combinatorial affibody libraries have successfully been constructed by randomization of 13 residues located on the solvent-exposed surface of helix one and two of the bundle (167). Affibodies consist of only 58 amino acid residues (7 kDa), rendering them a promising tool for imaging cancer-associated targets in vivo (167). Affibodies are further characterized by high proteolytic, chemical and thermal stability, a short folding time of 3 μs independent of disulfide bridge formation and are water-soluble (168–170). Affibodies are also suitable for solid-phase peptide synthesis, allowing the introduction of a handle at the N-terminus to incorporate various labeling moieties (171). Due to the absence of a cysteine residue in the polypeptide chain of an affibody, a cysteine can be introduced in the protein to allow site-directed conjugation of chelators, linkers and prosthetic groups using thiol chemistry (172, 173). Combined, these properties allow affibodies to perform well as molecular imaging tracers and therapeutic agents, with many demonstrating impressive results in preclinical and clinical studies (174–177).

In 2014 BMS disclosed a library containing three classes of macrocyclic peptide inhibitors of the PD-L1/PD-1 interaction that sparked tremendous interest in developing these binders as both therapeutic and imaging agents (188). A promising PD-L1-targeting peptide selected from this disclosure that has gone through the most phases of development up to now is the highly specific 14-mer cyclic peptide WL-12. Chatterjee et al. have already explored the use of WL-12 as a potential PET tracer in the earliest stages of development by conjugating it with DOTAGA and radiolabeling with copper-64 (189). The IC₅₀ of the metalated version of WL-12-DOTAGA was measured as 2.9 nM by a FRET-based assay (189). PET imaging of mice bearing human PD-L1 expressing CHO tumors could show rapid and specific uptake of the [⁶⁴Cu]Cu-DOTAGA-WL-12 (14.9 ± 0.8 %ID/g 60 min p.i. in tumors) (189). To further obtain insight into the binding mode of WL-12 with PD-L1, docking studies were also performed. Overlays of PD-1 and WL-12 bound to PD-L1 revealed that WL-12 forms a beta-sheet comparable to a beta-sheet structure found within the interaction surface of PD-1 (189). The use of [⁶⁴Cu]Cu-DOTAGA-WL-12 as a tool to evaluate the dynamic expression of tumor PD-L1 and target engagement of therapeutic anti-PD-L1 mAbs was successfully demonstrated (190). The same group further explored using gallium-68 (t_1/2 = 67.7 min) with a half-life more closely matching the peptide pharmacokinetics (191). While tumor uptake was comparable for both versions of radiolabeled WL-12, [⁶⁸Ga]Ga-DOTAGA-WL-12 resulted in faster blood clearance and improved imaging contrast but higher kidney uptake and lower liver uptake, compared to [⁶⁴Cu]Cu-DOTAGA-WL-12 in the same tumor model (191). To ensure the feasibility of clinical translation and due to its easy access, the group explored an additional commonly used radioisotope, fluorine-18, for labeling of WL-12 (192). The ability of the native WL-12 and its non-radioactive analog (FPy-WL-12) to block the PD-L1/PD-1 interaction was measured by FRET-based assays, and EC₅₀ values of 26.4 and 37.1 nM were reported, respectively (192). Compared to the previously developed labeled versions of WL-12, [¹⁸F]FPy-WL-12 exhibited lower tumor uptake and higher uptake in normal tissues, especially in the liver, which the group attributed to the influence of hydrophilicity and low labeling molar activity (192). Another group pursued this peptide further by investigating the effect of using a NOTA chelator instead of DOTAGA (193). The affinity of the non-radioactive [^natCu]Cu-NOTA-WL-12 to human PD-L1 was determined by SPR (K_D = 3.0 nM) and comparable to affinities previously reported (193). The group demonstrated that [⁶⁴Cu]Cu-NOTA-WL-12 could be prepared with higher radiochemical yield and purity compared to the DOTA counterpart (189, 193). More recently, Quigley et al. (194) reported on preclinical PET imaging in low PD-L1-expressing MDA-MB-231 human breast carcinoma and H2009 human lung adenocarcinoma tumor xenograft murine models using WL-12 conjugated to another chelator – TRAP – and radiolabeled with gallium-68. Compared to [⁶⁸Ga]Ga-DOTA-WL-12, uptake in non-target organs, especially the liver, was lower and blood clearance more rapid for [⁶⁸Ga]Ga-TRAP-WL-12 (194). Recently, Zhou et al. (195) performed the first-in-human evaluation of [⁶⁸Ga]Ga-NOTA-WL-12 in patients with NSCLC and demonstrated its safety and feasibility to be used as a companion diagnostic to quantify PD-L1 expression. In a small number of lung cancer patients, different tumor uptake levels of [⁶⁸Ga]Ga-NOTA-WL-12 correlated well with two different therapy response outcomes even though for these two particular cases, the PD-L1 expression determined by IHC was the same (195). [⁶⁸Ga]Ga-NOTA-WL-12 uptake in the liver increased substantially and shifted toward the small intestines after adding WL-12 blocking (195). The group continues exploring ways to optimize the biological properties of this peptide-derived radiotracer further by testing other chelators, such as HBED-CC (196). Besides the initial study performed by Chatterjee et al. (189), no further studies to elucidate structural features of the peptide have been conducted yet. This is most likely due to the fact that sufficient affinity of this peptide, as well as that of its modified analogs, has already been achieved.

Recently Ferro-Flores et al. (197) used structure-based design and extensive docking studies to develop another cyclic peptide, iPD-L1, based on the structure of WL-12 for SPECT imaging of PD-L1. The methylation of four residues and thioester could be removed in addition to replacement of the ornithine with a lysine present in WL-12. These modifications in combination with the introduction of a HYNIC heterocycle could result in a more rigid peptide structure (197). HYNIC-iPD-L1 was radiolabeled with technitium-99m and the tracer evaluated both preclinically and clinically. Molecular docking software was used to estimate the affinity to PD-L1 of both iPD-L1 and HYNIC-iPD-L1 in terms of molecular binding energy scores (−6.7 and −7.2 kcal/mol, respectively) (197). Affinity estimations of iPD-L1 were more favorable compared to WL-12 and biodistribution studies in mice bearing human lung cancer tumors (HCC827) showed higher tumor uptake at 24 h p.i. of [^99mTc]Tc-iPD-L1 (5.65 %ID/g) compared to [^99mTc]Tc-WL-12 (3.21 %ID/g), however for [^99mTc]Tc-iPD-L1 activity in non-tumor tissues was higher as well (197). A patient with plantar malignant melanoma underwent [^99mTc]Tc-iPD-L1 SPECT/CT imaging and the tracer could in fact distinguish between lesions with and without PD-L1 expression (197). This study clearly demonstrated the potential of using this modified cyclic peptide for detecting PD-L1 expression by SPECT imaging.

The same group of researchers who initially developed WL-12 designed a derivative thereof, defined as DK221. In this work, the research focus was shifted toward developing imaging agents that could not only be used for imaging PD-L1, but that could be used more specifically as a tool to assess the pharmacodynamic behavior of therapeutic mAbs and determine the accessible PD-L1 target levels within the tumor (198). To this end, WL-12 was modified by introducing additional carboxylate groups in the peptide: L6E and Trp(Me)10Trp(carboxymethyl) (198). The ornithine group was replaced with lysine for modification with the bifunctional chelator NCS-MP-NODA and subsequently radiolabeled with fluorine-18 to form Al[¹⁸F]F-NODA-DK222 (198). The non-radioactive form was able to inhibit the PD-1/PD-L1 interaction with an EC₅₀ = 25 nM measured by FRET assays and in vitro cell binding assays using Al[¹⁸F]F-NODA-DK222 confirmed its specificity for binding to PD-L1 (198). PET imaging and biodistribution studies in NSG mice bearing high PD-L1-expressing triple-negative breast cancer human xenografts using Al[¹⁸F]F-NODA-DK222 resulted in high contrast images with good tumor uptake (13.4 ± 0.1 %ID/g 60 min p.i.), high accumulation in the kidneys (57.7 ± 0.5 %ID/g 60 min p.i.) and low accumulation in the liver (198). Al[¹⁸F]F-NODA-DK222 was then further used to assess the residence time, or target engagement, of different anti-PD-1 and -PD-L1 therapeutic mAbs at the tumor site during treatment. Interestingly, Al[¹⁸F]F-NODA-DK222 PET was able to successfully quantify changes in PD-L1 occupation levels in real-time and could further reveal differences in the pharmacological activity of different anti-PD-L1 mAbs (199). An optimized protocol for the radiosynthesis of Al[¹⁸F]F-NODA-DK222 was later established that met the requirements for current GMP that could swiftly facilitate in-human studies, clinical translation and commercial production of this tracer (200). With similar aims in mind but exploring different radiochemistry, the same group performed a study by conjugating DK221 with the chelator DOTA-NHS for radiolabeling with gallium-68 to form [⁶⁸Ga]Ga-DOTA-DK223 (201). Results were highly comparable to those obtained with Al[¹⁸F]F-NODA-DK222, showcasing the broad applicability and potential of this peptide-based imaging probe (201).

Liu et al. (202) were able to identify a linear 12-mer peptide, CLP002, by phage display biopanning with an affinity determined by SPR in the mid-nanomolar range (K_D = 366 ± 150 nM) able to block both the human and murine PD-1/PD-L1 interaction efficiently. Molecular docking studies revealed that the interaction surface of the peptide with PD-L1, similar to WL-12, overlaps with that of PD-1, however, the specific amino acid interactions between CLP002 and PD-L1 responsible for binding were not described (202). An interesting finding from this study was that one of the leads tested, CLP001, did not share a significant interaction surface area with PD-1, confirmed by blocking studies that showed little to no blocking efficiency of the human PD-1/PD-L1 interaction using recombinant PD-L1 and PD-L1 positive cancer cells (202). The group further showed that CLP002 had better tumor penetration and similar tumor growth inhibition capabilities compared to an anti-PD-L1 antibody (202). A known characteristic often observed mainly for linear peptides, is poor stability in vivo due to their susceptibility to protease degradation (203). The same group explored ways to improve CLP002 characteristics by performing side-chain macrocyclization scanning (204). This technique entails the substitution of two amino acids in the peptide sequence with a cysteine to produce all possible corresponding cyclic derivatives of the peptide. They discovered two cyclic CLP002 derivatives, CP7 and CP12, that had improved blocking efficiency of the PD-1/PD-L1 interaction, serum stability and anti-tumor activity compared to the parent peptide (204). Despite its inferior in vitro and in vivo results, Zhang et al. (205) decided to pursue the linear format of CLP002 further for its potential as a radiotracer to image PD-L1 in a colon cancer mouse model. The DOTA-conjugated CLP002, defined now as HKP2201, was radiolabeled with both copper-64 and gallium-68. A dimerized version, [⁶⁸Ga]Ga-DOTA-HKP2202, was tested alongside HKP2201 to determine whether peptide dimerization can improve tumor uptake (205). Both ⁶⁴Cu- and ⁶⁸Ga-labeled HKP2201 showed modest tumor uptake, however, a significant reduction in liver uptake was observed for [⁶⁸Ga]Ga-DOTA-HKP2201, making it the preferred labeling strategy for HKP2202. [⁶⁸Ga]Ga-DOTA-HKP2202 could improve tumor uptake significantly at 60 min p.i. compared to [⁶⁸Ga]Ga-DOTA-HKP2201 while maintaining the reduced uptake in the liver (205).

The first D-peptide identified by mirror-image phage display (MIPD) was identified and characterized in vitro (206). MIPD uses a D-protein of interest, in this case, D-PD-L1, in combination with a phage display library containing L-peptides for biopanning to identify L-peptide binders (207). The next step is synthesizing the corresponding D-enantiomeric form of the L-peptide binder that would bind to the original L-protein (207). The rationale behind developing such a peptide binder was to circumvent the inherent limitation of small proteins or peptides that have poor in vivo stability due to their sensitivity to proteolytic degradation in serum (203). Chang et al. (206) identified a 12-mer peptide inhibitor of the PD-L1/PD-1 interaction, defined as ^DPPA-1, that showed no degradation in 10% human serum within a period of 24 h and that could bind to human PD-L1 with a K_D of 0.51 μM as measured by SPR. In addition to excellent stability, ^DPPA-1 injected into CT26 tumor-bearing mice reduced tumor growth and increased survival time (206). A Cy5.5-conjugated ^DPPA-1 was subsequently used for near-infrared fluorescence imaging to assess its biodistribution in the same mouse tumor model. Interestingly, while ^DPPA-1 showed good uptake in the tumor, the peptide accumulated substantially in the liver and kidneys, in addition to some accumulation in the stomach and lungs (206). Since little was known about the metabolic profiles of D-peptides and their potential, the same group set out to investigate using this peptide as a PET imaging tracer (208). ^DPPA-1, now referred to as D-dodecapeptide antagonist or DPA, was modified by adding a PEG₃ linker and DOTA chelator for radiolabeling with copper-64 and gallium-68 (208). [⁶⁴Cu]Cu-DOTA-PEG₃-DPA was injected into WT (C57BL/6J) mice to determine the normal distribution of the radiotracer, and rapid clearance (within 20 min) from all major organs, including the heart, kidneys, and liver, was achieved (208). Urine samples were also collected at different time points up to 2 h p.i. and no degraded [⁶⁴Cu]Cu-DOTA-PEG₃-DPA could be found (208). Both [⁶⁴Cu]Cu-DOTA-PEG₃-DPA and [⁶⁸Ga]Ga-DOTA-PEG₃-DPA were injected into mice bearing B16F10 mouse melanoma or human glioblastoma tumors. In both cases, modest and rapid tumor uptake 60 min p.i. was observed in addition to slow tumor clearance, fast clearance from other organs and elimination via the kidney-bladder system (208). No significant decrease in red blood cells, platelets, or hematocrit was observed following injection of [⁶⁴Cu]Cu-DOTA-PEG₃-DPA injection measured up to 30 days p.i. suggesting potentially low toxicity (208). Another group recognized the potential of this peptide as a PET tracer and developed a fluorine-18 labeled version, defined it as Al[¹⁸F]F-NOTA-NF12, and performed the first human evaluation of this PD-L1 peptide binder (209). To take the characterization of this peptide tracer a step further, Zhou et al. (209) performed molecular docking to get insight into the binding mode of NF12 to PD-L1. The prediction revealed that NF12 binds to four amino acids on PD-L1, three of which have previously been identified as hot spot residues (48, 195). Healthy volunteers and patients with either NSCLC or esophageal cancer injected with Al[¹⁸F]F-NOTA-NF12 showed no clinical adverse or pharmacological events, and the tracer was eliminated primarily by the renal-urinary system (209). Compared to [⁶⁸Ga]Ga-NOTA-WL-12, the accumulation of Al[¹⁸F]F-NOTA-NF12 in other organs, such as the liver, was substantially decreased (195, 209). Tumor PD-L1 expression was efficiently detected by Al[¹⁸F]F-NOTA-NF12 within a clinically relevant timeframe. Moreover, fast clearance and low non-specific uptake resulted in high-contrast images (209).

Targeting PD-L1 peptide 1 (TPP-1) was identified by bacterial surface display and has shown to be highly specific for PD-L1 (210). TPP-1 could bind to PD-L1 with high affinity as determined by a single cycle kinetic SPR measurement (K_D = 95 nM) and exhibited good binding to tumor cell lines expressing PD-L1 (210). TPP-1 could not only block the PD-1/PD-L1 interaction, as determined by both ELISA and cell-based blocking assays, but could efficiently inhibit tumor growth mediated by T cell reactivation in a tumor xenograft mouse model (210). A prediction of the binding site of TPP-1 on PD-L1 was explored by the same group, and similar to WL-12, a beta-sheet-like secondary structure is formed that interacts in the same PD-L1 binding pocket as PD-1 (189, 210). Kuan et al. (211) further investigated the potential TPP-1 as a PET imaging tracer by directly comparing fluorine-18 and copper-64 labeling strategies. Additionally, a pegylated tetramer of TPP-1 was also radiolabeled and tested to ascertain whether the pharmacokinetics and biodistribution of the tracer could be improved (211). Pegylated tetramer tracers were more stable than their native counterparts, confirmed by serum stability tests and longer retention in circulation up to 2 h p.i. (211). They found that TPP-1 and its pegylated tetramer labeled with copper-64 could successfully detect PD-L1 at the tumor site of MDA-MB-231 tumor-bearing mice compared to the negligible signal detected using those labeled with fluorine-18 (211). While significant liver accumulation was observed for the ⁶⁴Cu-labeled tracers, higher spleen uptake could be achieved in normal (C57BL/6J) mice compared to ¹⁸F-labeled tracers (211). While there is still room for improvement, this study successfully described a strategy that could potentially improve radiotracer behavior in vivo.

A 25-mer peptide, RK-10, was identified by exploiting X-ray crystal structure data available for the PD-1/PD-L1 interaction and using computational techniques to determine which amino acid sequences could have high probabilities of forming interactions with PD-L1 in the binding region (212). Different levels of PD-L1 could easily be detected by biotinylated or dye-conjugated RK-10, as demonstrated by IHC, flow cytometry, and patient tissue immunofluorescent staining (212). Notably, the excellent in vitro results for RK-10 suggest a high-affinity peptide, however, the actual affinity has not been measured by any standard techniques. Sun et al. (213) recognized the potential of RK-10 as a PET imaging tracer to evaluate PD-L1 status in tumors. RK-10 was modified with NOTA, radiolabeled with fluorine-18 to form Al[¹⁸F]F-NOTA-IPB-PDL1P and injected into mice bearing human colon carcinoma (HCT116), human prostate carcinoma (PC3) or CHO-K1 tumors (213). In vivo stability of Al[¹⁸F]F-NOTA-IPB-PDL1P was tested in serum collected from the blood of injected mice and showed intact tracer up to 1 h after injection, which was reflected by the prolonged tumor retention seen in the PET scans up to 2 h p.i. (213). However, uptake of Al[¹⁸F]F-NOTA-IPB-PDL1P in non-specific organs was considerable, even 2 h p.i., in addition to evident liver metabolism of this tracer (213). While this peptide has potential as a PD-L1-targeting tracer, further characterization and optimization are required to improve in vivo properties.

A couple of cyclic peptides have recently emerged as promising PD-L1 binders, some with the potential to be used as imaging tracers (214–216). Similar to how WL-12 and DK221 were developed, these peptides are also derived from the earlier disclosure of BMS describing macrocyclic peptides inhibiting the PD-L1/PD-1 interaction (188). Not too long after, a group published their findings on the structural binding mode of two representative cyclic BMS peptides with PD-L1, BMS-71, and BMS-57, with IC₅₀ values of 7 and 9 nM measured by HTRF assays and reported by BMS, respectively (188, 217). Jouini et al. (214) used the BMS-71/PD-L1 crystal structure complex as a guide and identified the C-terminal Gly-NH₂ as a potential site for modification with DOTA and subsequent radiolabeling with gallium-68 to form their radiotracer, defined as [⁶⁸Ga]Ga-DOTA-NJMP1. However, due to complications with the synthesis of BMS-71, they opted for BMS-78 (reported IC₅₀ = 14 nM), a peptide identical to BMS-71 aside from an unmethylated glycine in BMS-78 (188, 214). Unfortunately, cellular binding studies demonstrated the inability of [⁶⁸Ga]Ga-DOTA-NJMP1 to bind to PD-L1 (214). Accordingly, a significant decrease in the affinity of radiolabeled NJMP1 was observed, and it was concluded that contrary to the original hypothesis, the site for modification had a major influence on the binding mode (214). In addition, this group found a discrepancy in the affinity for BMS-78 with a 10-fold decrease in the value reported (214). Soon after, another group reported an optimized version of BMS-71, PG1, with superior in vitro and in vivo behavior (215). It was hypothesized that the methyl group of Gly in BMS-71 is a preferred site for modification since molecular docking of both BMS-71 and PG1 on a tetrameric PD-L1 structure revealed it is directed away from the other three monomers (215). The high affinity of the precursor, DOTA-PG1, was confirmed by BLI assays (K_D = 9.9 nM) and rapid [⁶⁸Ga]Ga-DOTA-PG1 uptake in PD-L1-expressing A375 cells (215). PET/CT imaging and biodistribution in A375-PD-L1 tumor-bearing mice demonstrated rapid and high uptake of [⁶⁸Ga]Ga-DOTA-PG1 as soon as 30 min p.i. (5.01 ± 0.46 %ID/g) that is retained in the tumor up to 2 h p.i. (6.96 ± 1.02 %ID/g) (215). In contrast with ⁶⁸Ga-labeled WL-12 and DK223, high liver uptake was indicative of [⁶⁸Ga]Ga-DOTA-PG1 metabolism primarily by the hepatobiliary system (191, 201, 215). Another macrocyclic peptide, pAC65, to keep an eye on, has recently been reported by Rodriguez et al. (216) to have affinity and bioactivity equivalent to current FDA-approved anti-PD-L1 mAbs. Based on the BMS-57/PD-L1 crystal structure complex previously reported by the same group, six hydrophobic amino acids in the structure of BMS-57 were identified and replaced by more hydrophilic amino acids to form pAC65 (216). pAC65 showed no evidence of toxicity at concentrations tested, however Jurkat effector cells could effectively be activated by pAC65 in a dose-dependent manner in the presence of PD-L1-expressing cells (EC₅₀ = 0.6 nM) similar to atezolizumab (EC₅₀ = 0.1 nM) (216). Initial in vitro results are promising for pAC65, and its potential to be developed further as a peptide-based therapeutic or imaging agent has been established.

While the number is limited, a couple of PD-1-targeting peptide inhibitors have recently been described (218–220). The exact interaction surface of PD-L1 and PD-1 was uncovered by X-ray crystallography some time ago (48). Based on these structures, Boohaker et al. (218) selected a stretch of highly conserved amino acid residues present in both human and murine PD-L1 (Gly120 to Asn131) that has been shown to interact with PD-1 and investigated the 12-mer interface peptide as a PD-1 inhibitor. Another group explored the potential of this interface peptide to be used as a PD-1 PET imaging tracer (219). Human and mouse interface peptides were defined as hPep-1 and mPep-1, respectively, with mPep-1 containing a Leu instead of a Val at position 128 in the sequence of PD-L1 (219). Both hPep-1 and mPep-1 were conjugated with DOTA at the N-terminal and subsequently radiolabeled with copper-64 (219). Both tracers were evaluated in mice bearing murine melanoma B16F10 tumors (characteristic of PD-1 overexpression on TILs) and human hepatocellular carcinoma Huh-7 tumors. No tumor uptake was observed in the Huh-7 tumor model, while mPep-1-⁶⁴Cu was able to detect PD-1 as soon as 20 min p.i. in the B16F10 tumor model (219). A cyclic 7-mer peptide, C8, has been identified by phage display and was shown to bind to PD-1 with an affinity of 0.64μM as measured by microscale thermophoresis (MST) (220). It was further shown that C8 was able to block the PD-L1/PD-1 interaction, activate T cells to a similar extent as an anti-PD-1 mAb, and could inhibit the growth of CT26 and B16 tumors, the latter being a known anti-PD-1 mAb resistant tumor model (220). These promising in vitro results warrant further exploring C8 as an imaging agent to quantify PD-1 expression levels.