Development and In Vivo Evaluation of Small-Molecule Ligands for Positron Emission Tomography of Immune Checkpoint Modulation Targeting Programmed Cell Death 1 Ligand 1

A substantial portion of patients do not benefit from programmed cell death protein 1/programmed cell death 1 ligand 1 (PD-1/PD-L1) checkpoint inhibition therapies, necessitating a deeper understanding of predictive biomarkers. Immunohistochemistry (IHC) has played a pivotal role in assessing PD-L1 expression, but small-molecule positron emission tomography (PET) tracers could offer a promising avenue to address IHC-associated limitations, i.e., invasiveness and PD-L1 expression heterogeneity. PET tracers would allow for improved quantification of PD-L1 through noninvasive whole-body imaging, thereby enhancing patient stratification. Here, a large series of PD-L1 targeting small molecules were synthesized, leveraging advantageous substructures to achieve exceptionally low nanomolar affinities. Compound 5c emerged as a promising candidate (IC50 = 10.2 nM) and underwent successful carbon-11 radiolabeling. However, a lack of in vivo tracer uptake in xenografts and notable accumulation in excretory organs was observed, underscoring the challenges encountered in small-molecule PD-L1 PET tracer development. The findings, including structure–activity relationships and in vivo biodistribution data, stand to illuminate the path forward for refining small-molecule PD-L1 PET tracers.


■ INTRODUCTION
Cancer immunotherapy has transformed the landscape of cancer treatment over the past decade.Among the remarkable advances in this field, immune checkpoint therapy, particularly the blocking of programmed cell death 1 ligand 1 (PD-L1) and its receptor, programmed cell death protein 1 (PD-1), has emerged as a pivotal strategy.This approach harnesses the power of the immune system to target and eliminate cancer cells, leading to unprecedented clinical responses in a subset of patients. 1owever, the clinical success of PD-L1 checkpoint therapy has unveiled complex challenges, including response heterogeneity, resistance, and the need for accurate patient stratification.In this context, positron emission tomography (PET) radiotracers, particularly radiolabeled antibodies, have emerged as promising tools to address these challenges by enabling longitudinal, noninvasive, real-time assessment of PD-L1 expression and immune response dynamics. 2,3One of the most promising applications of PD-L1 PET radiotracers is patient stratification.By identifying patients with high PD-L1 expression and an active antitumor immune response, PET imaging can guide the selection of individuals who are most likely to respond to PD-L1 checkpoint inhibitor therapy. 2 This personalized approach holds the potential to minimize treatment-related adverse events (fatigue, pruritus, diarrhea, endocrine dysfunction, pneumonitis) and optimize therapeutic outcomes.
Human studies of radiolabeled anti-PD-1/PD-L1 antibodies, e.g., [ 89 Zr]Zr-atezolizumab, [ 89 Zr]Zr-durvalumab and [ 89 Zr]Zrpembrolizumab, demonstrated that radiotracer tumor uptake was higher in patients with a response to immune checkpoint therapy. 2,4,5Additionally, tumor uptake correlated better with clinical response than immunohistochemistry or RNA-sequencing, 2,4 and substantial intra-and intertumoral uptake heterogeneity was observed, reflecting the heterogeneity of PD-L1 expression.Recently, the peptide-based radiotracer [ 68 Ga]Ga-NOTA-WL12 was investigated in a first-in-human study indicating its potential benefits for clinical immunotherapy. 6−16 Small-molecule PET tracers targeting PD-L1 represent a promising avenue for addressing critical challenges associated with antibody-and peptide-based radiotracers.These small molecules offer potential advantages, including expedited pharmacokinetics, cost-effectiveness, increased stability, and enhanced tissue and tumor penetration, facilitating comprehensive evaluation of PD-L1 expression within the heterogeneous tumor microenvironment.Significant efforts have been invested in advancing small molecules for therapeutic applications despite the intricate nature of the target.PD-L1 lacks a dedicated binding pocket for small molecules and its binding mode with the endogenous receptor PD-1 is characterized by a large and flat protein−protein interaction interface. 17This characteristic makes it challenging to effectively target PD-L1 with small molecules.Among these compounds, the biphenyl substructure emerged as a prominent and recurrent moiety found in potent inhibitors patented by companies and institutes in the pharmaceutical field, e.g., Bristol Myers Squibb (BMS), Polaris Pharmaceuticals, Incyte Corporation and Institute of Materia Medica. 18These compounds exhibited selectivity for human PD-L1 (hPD-L1) over murine PD-L1 (mPD-L1) 19,20 and induced dimerization of PD-L1 through binding modes that overlap with anti-PD-L1 antibodies, e.g., atezolizumab and durvalumab. 21,22Nevertheless, the development of nonpeptidic small-molecule PD-L1 PET tracers is still in its early stages with limited published research and constrained achievements to date (Table 1).The observed uptakes in PD-L1 expressing (PD-L1 + ) tumor xenograft over controls were modest, with increases of 2.2-fold, 2.9-fold, 3.0-fold, ∼1.4-fold, and 1.9-fold, resulting in uptake values of 1.2% ID/g, 4.0% ID/g, 3.5% ID/mL, ≤ 5% ID/g, and 4.2% ID/g for radiotracers [ 18 F]LN, 23 [ 18 F]LG-1, 24 [ 18 F]LP-F, 25 [ 64 Cu]Cu-43b, 26 and [ 68 Ga]BMSH, 27 respectively.[ 18 F]FDHPA 28 and [ 18 F]LGSu-1 29 demonstrated ≤1% ID/g and 3.3% ID/mL uptake, respectively; however, control xenografts were not available for comparison.
Recent research explored the potential of the 4-fluorophenylthiophene-3-carbonitrile moiety as an alternative to the biphenyl core substructure.Ex vivo tissue section autoradiography experiments showed that this radiotracer (2-((4-(aminomethyl)benzyl)oxy)-4-(4-[ 18 F]fluorophenyl)thiophene-3-carbonitrile) exhibited a 1.4-fold higher uptake in PD-L1 + compared to PD-L1 − H358 tumors (lung adenocarcinoma).This difference was not observed in PD-L1 ± ES2 tumors (ovarian carcinoma), in contrast to a radiolabeled biphenylbased BMS-1166 derivative. 30We previously demonstrated that commercially accessible biphenyl-based lead structures and derivatives designed via a ligand-based drug design approach exhibit suboptimal binding affinity.Nonetheless, this inves-tigation provided valuable insights into the effects of structural modifications. 31ur primary objective was to design and develop small molecules for noninvasive PET imaging targeting PD-L1, with the overarching goal of enhancing patient stratification within the framework of personalized medicine.This endeavor was rooted in the identification of promising substructures through rigorous in silico investigations and an extensive review of existing literature 18,32−34 (Figure 1).De novo synthesized compounds underwent extensive in vitro evaluations, with a particular focus on assessing their binding affinity toward PD-L1.Viable candidates were subjected to carbon-11 radiolabeling processes, culminating in the selection of the most promising candidate for further investigations, both in vitro and in vivo.

■ RESULTS AND DISCUSSION
Pharmacophore-Based Virtual Screening.A consensus feature-based ("shared feature") pharmacophore model was derived from six distinct crystallographic data sets (PDB: 5J89, 5J8O, 5N2D, 5N2F, 6R3K, and 6NM8) using small-molecule ligands that interact with PD-L1 (Figure S1).This model encompassed three hydrophobic features and a positive ionizable area (Figure 2B).Hydrophobic features represent the 2-methylbiphenyl core substructure situated at the base of the hydrophobic pocket formed within the interplay of two PD-L1 monomers. 21,31o identify novel potential structures substituting the 2methylbiphenyl moiety, the generated pharmacophore model underwent screening against a data set of 34,207 low molecular weight compounds (≤200 g/mol) including bioactive molecules with drug-like properties, marking the positive ionizable area as an optional feature.A total of 2695 in silico hits (7.9%) were acquired, exhibiting Pharmacophore-Fit Scores spanning from 34.73 to 38.89.Upon transposition to the PDB entry 5J89, Binding Affinity Scores were computed, encompassing a range from −34.55 to 26.44.These hits were then ranked based on their scores and were allocated up to 10 points per score.Top 10 hits are represented in Table 2.All hits passed the Pan Assay Interference Compounds (PAINS) test.
The phenyl moiety and its bioisosteric counterparts emerged as recurring substructures, with the 2-methylbiphenyl structure (entry 3) and modifications being frequently represented among the top hits.It is worth mentioning that the tertiary amine in entry 1 (calculated pK a : 7.34) would undergo protonation within the acidic tumor microenvironment (pH 6.4−7 35 ), which might have adverse effects on the binding mode.In the case of entry 2, the presence of an additional methyl group at the distal phenyl ring compared to entry 3 implies the applicability of specific modifications.However, it has been shown that methoxy, ethoxy, and methylenedioxy substituents exhibit detrimental effects, while compounds containing an ethyl- enedioxy group displayed comparable or enhanced binding affinities. 32Interestingly, pyrrole (entry 4) was identified as a superior bioisosteric replacement for the distal phenyl ring compared to pyridine (entry 9), underscoring the significance of the heteroatom's position and basicity.In summary, our pharmacophore-based virtual screening investigations did not uncover any novel structures capable of enhancing pharmacophore fitting and binding affinity beyond the 2-methylbiphenyl structure (entry 3).Anyway, it is worth mentioning that pyrrole may serve as a potential bioisosteric replacement with reduced hydrophilicity.Following this observation, we synthesized compounds with pyrrole substitutions, replacing the 2-methylbiphenyl moiety at R 1 (Scheme 1) in the subsequent step.
Multistep De Novo Synthesis of Ligands.Novel ligands were synthesized by incorporating potentially beneficial substructures identified through pharmacophore-based virtual screening (vide supra) and extensive literature research, 32−34 along with previously unexplored molecular entities and bioisosteric replacements.The multistep synthetic pathway is presented in Scheme 1. Overview of exemplary structures from previously reported biphenyl-based ligands, 32−34 featuring potentially advantageous substructures that serve as the foundation for the development of our compounds.A comprehensive patent review has been published before. 18ompounds 1 represent the main pharmacophore deemed essential for PD-L1 binding as described before. 21These compounds are sourced either from commercial suppliers (1a) or synthesized through Suzuki coupling (1b) of a boronic acid and aryl halide, or reduction (1c) of the respective carboxylic acid, resulting in good yields of 64 and 74% (Table 3).Intermediates 1 were subsequently joined with polysubstituted (hetero)aromatic molecules, that bear functional groups suitable for subsequent modifications, via Mitsunobu reactions delivering yields within the range of 24−52% (2a−d).(Un)substituted (hetero)cyclic aromatic molecules were added through nucleophilic substitutions under basic condition in good yields of 68−83% (3a−e).Intermediates 3 served as precursors for intermediates 4b−o or final compounds 5a−d, 5f, and 5i in reductive amination reactions using NaBH(OAc) 3 or NaBH 3 CN as reducing agents achieving yields of 12−64%.Compound 4a, which lacks R 2 , was synthesized from 2a through reductive amination with 41% yield.Intermediates 4 were used as starting material for nucleophilic substitution reactions (i.e., fluoroethylation and carbamylation) giving final compounds 5e,g,h,j in 40−52% yields, and as precursors for radiolabeling (i.e., 11 C-methylation).All intermediates and products passed the PAINS test.
In summary, the synthetic pathways involving Mitsunobu reactions, nucleophilic substitutions, and reductive aminations yielded the desired final products effectively.This approach resulted in the generation of 37 compounds that in the further course facilitated the exploration of structure−activity relation-ships.Furthermore, it provided a collection of six methylated or carbonylated and four fluorinated products, which were used for subsequent in vitro evaluations and served as essential reference compounds for radiolabeling endeavors.
Structure−Activity Relationships.The lipophilicity of the compounds was evaluated using an established HPLC method 36,37 as the logarithm of the partition coefficient at pH 7.4 (μHPLC logD pH7.4 ) (Table 3).Obtained lipophilicity data was compared with calculated parameters such as the clogP, clogD pH7.4 , and the topological polar surface area (tPSA) (Table S2).The measured μHPLC logD values for compounds 4a−o and 5a−j fell within the range of 2.33−5.6 except for compounds 5g and 5h with logD values >5.75, indicating their overall lipophilic nature.
Our measurements clearly showcased how structural modifications affected lipophilicity (μHPLC log D).The introduction of pyrrole at R 1 reduced lipophilicity compared to the distal 1,4-benzodioxanyl moiety, and the introduction of one or more heteroatoms at R 2 in the form of picolinonitrile or oxazole reduced lipophilicity compared to benzonitrile.R 3 and R 3 * significantly influenced the lipophilic character of our compounds.Indeed, fluoroethylation and methylation increased lipophilicity, although O-methylation resulted in a more significant increase in lipophilicity compared to N-methylation.
Our smallest tested compounds (intermediates 2b,c) did not exhibit binding to PD-L1 in the competitive HTRF assay.First

Journal of Medicinal Chemistry
observations of PD-L1 binding proficiency were made with intermediates 3 upon the introduction of R 2 .This is in contrast to the findings by Skalniak et al., 38 where 1 H− 15 N HMQC NMR measurements have elucidated that the minimal functional fragment capable of engaging with PD-L1 corresponds to the biphenyl structure, mirroring our 1b intermediate.Significantly enhanced binding affinities were achieved with the incorporation of polar residues R 3 , surpassing, in certain instances, the antibody atezolizumab (IC 50  = 4.1 nM, Table S2).Compounds 4b−o encompassing all three residues (R 1 , R 2 , and R 3 ) demonstrated exceptional IC 50 values, spanning from 3.7 to 50 nM.Furthermore, final products 5, featuring chemically modified R 3 residues (R 3 *), also displayed remarkable PD-L1 binding affinities in the low nanomolar range, but not superior when compared to compounds 4.
The correlation analysis between various calculated and measured parameters and IC 50 values revealed the following observations: a moderate correlation (ρ = −0.51,p = 0.003, n = 32) was found between molecular weight and affinity (IC 50 ) supporting the indication that an adequate molecular size was required for sufficient binding (Figure 4A).There was a moderate correlation between clogP (ρ = 0.51, p = 0.003, n = 32) and clogD (ρ = 0.55, p = 0.001, n = 32) and measured IC 50 values (Figure 4B,C).Calculated tPSA demonstrated a strong inverse correlation (ρ = −0.70,p = 0.000007, n = 32), implying an affinity−hydrophilicity relationship, although only a weak and statistically not significant correlation was observed for the measured μHPLC logD values (ρ = 0.30, p = 0.14, n = 25) (Figure 4D,E).These results highlight the complexity of reliably predicting PD-L1 binding affinities and suggest that multiple factors beyond molecular size, lipophilicity, and polar surface area may influence binding.
In further attempts to predict affinity in silico, 37 literatureknown ligands encompassing a broad range of affinities 32,33 and 25 novel compounds 4a−o and 5a−j underwent ligand docking (Table S3).Subsequently, the calculated binding affinity parameters were extracted and subjected to correlation analysis with literature HTRF IC 50

Journal of Medicinal Chemistry
results of our extensive ligand docking study revealed only very weak and weak correlations between docking parameters (i.e., Binding Affinity Score and Affinity) and the affinity of both literature-known ligands (ρ = −0.10,p = 0.57 and ρ = 0.26, p = 0.11, respectively) and novel compounds (ρ = 0.01, p = 0.95 and ρ = 0.33, p = 0.11, respectively) (Figure S3).These findings emphasize the challenges and significance of comprehending the precise determinants of binding affinity.
In summary, de novo synthesized small-molecule compounds reached excellent PD-L1 binding affinities in the low nanomolar range comparable to the antibody atezolizumab.This work represents a significant advancement in binding capabilities when compared to commercially available small molecules, such as PD-1/PD-L1 Inhibitor 1 (BMS-1; IC 50 = 202 nM), PD-1/ PD-L1 Inhibitor 2 (BMS-202; IC 50 = 101 nM) and PD-1/PD-L1 Inhibitor 3 (a macrocyclic peptide; IC 50 = 113 nM) (Table Table 3  S2), as well as small molecules described in our previously reported ligand-based drug design approach. 31ell Viability and Cell-Based Competitive Binding Assay.Compounds 5a and 5c, exhibiting high affinities (IC 50 ) of 6.2 and 10 nM, respectively, were chosen as promising candidates and underwent further in vitro evaluation using cellbased assays.The MTT assay, employed to assess cell viability, revealed cytotoxicity profiles similar to those of small-molecule compounds, including PD-1/PD-L1 Inhibitor 1 and 2 (BMS-1 & BMS-202) (Figures 5A and S4), as well as various other BMS compounds.38 Cell viability provided the basis for establishing a concentration range for subsequent cell-based in vitro investigations to avoid interferences with the results based on cellular death.
The cell-based PD-L1 binding affinity of 5a and 5c was evaluated using a competitive radioligand binding assay with the zirconium-89 labeled anti-PD-L1 antibody atezolizumab ([ 89 Zr]Zr-atezolizumab) (Figure 5B).Indeed, [ 89 Zr]Zr-atezolizumab exhibited no binding to PD-L1 negative CHO cells, but bound to PD-L1 positive CHO-hPD-L1 cells, and this binding was effectively blocked by preincubation with excess (>100-fold) unlabeled antibody (p = 0.0081), demonstrating its specificity.When 5a and 5c were administered at their highest noncytotoxic concentrations, a 50% blockade of antibody binding was observed (p = 0.034, p = 0.025, respectively).This can be translated into K i values using the Cheng−Prusoff equation, 39 resulting in a range of ∼700 to 4500 nM.−43 These results indicate that, on one hand, small-molecule compounds and atezolizumab share binding motifs on the PD-L1 protein as anticipated. 21,22On the other hand, it implies that their binding affinities might not be as robust as initially indicated by the cell-free HTRF assay, when using a more complex, biological system.
Radiolabeling of High-Affinity Ligands.Lead structures 5a and 5c were subjected to carbon-11 labeling by conventional 11 C-methylation.Small-scale reactions were conducted to optimize the reaction conditions for enhanced radiochemical conversion (RCC) and selectivity for the desired N-methylated products  involved varying precursor concentration, reaction temperature, and the addition of a base, using the demethylated precursors 4c or 4e and the [ 11 C]CH 3 I synthon (Scheme 2).
[ 11 C]5a was obtained with an RCC of up to 32.9% and concurrent formation of 37.7% byproduct [ 11 C]5b at 100 °C (Figures 6A and S5).Optimization of reaction conditions improved selectivity (Figure 6B).A reaction temperature of 60 °C, without the addition of a base, appeared to strike a favorable balance between achieving high radiochemical conversion and maintaining selectivity.Furthermore, a precursor concentrationdependent RCC was found, although the precursor concentration did not impact the isomer selectivity (Figure 6C).
Similarly, [ 11 C]5c was selectively produced with exceptional RCCs of up to 67% at 100 °C (Figures 6D and S6).The absence of the O-methylated byproduct [ 11 C]5d even in the presence of a base, suggests that DIPEA (calculated pK a = 10.7) may not effectively deprotonate the alcohol of 4e (calculated pK a = 15.6).Alternatively, it could indicate that the formation of [ 11 C]5c is kinetically favored over [ 11 C]5d.Again, a precursor concentration-dependent correlation with RCC was identified (Figure 6E).As precursor concentration decreased and base was added, a more lipophilic, unidentified product emerged (Figure 6F), concurrent with the decrease of [ 11 C]5c (Figure 6D), suggesting the formation of a potential dimethylated byproduct.
In summary, small-scale reactions successfully attained satisfactory RCC for both [ 11 C]5a and [ 11 C]5c.The superior RCC, coupled with feasible chromatographic separation of the product from precursor and byproducts, rendered [ 11 C]5c the more favorable choice for subsequent in vitro and in vivo assessments.
Plasma Stability, Plasma Protein Binding, and Metabolic Stability.Radiotracer [ 11 C]5c underwent additional evaluation to assess its plasma stability and metabolic stability

Journal of Medicinal Chemistry
for subsequent in vivo investigations.It exhibited remarkable stability, with over 99% remaining intact for 60 min in both mouse and human plasma (Figure 7).A high plasma protein binding of 98.9% aligns with expectations, given the tracer's lipophilic properties.Furthermore, when subjected to incubation with human liver microsomes for 60 min, 49.3% of the radiotracer remained intact.These findings connote that [ 11 C]5c exhibits sufficient stability for in vivo investigations.
In Vivo μPET Imaging and Biodistribution.PD-L1 expression in both cell lines (CHO and CHO-hPD-L1) was confirmed in vitro through flow cytometry (Figure 8A) and ex vivo via immunohistochemistry (Figure 8B).Tumor vascularization was verified by CD31 staining (Figure 8B).Dynamic μPET/CT imaging was conducted using [ 11 C]5c to evaluate its in vivo potential for quantifying PD-L1 expression.NSG mice with both PD-L1 negative (CHO) and PD-L1 overexpressing (CHO-hPD-L1) xenografts were used for this assessment.Following the imaging, the mice were sacrificed for subsequent ex vivo biodistribution analysis.
Neither the radiotracer nor its potential radiometabolites accumulated in white adipose tissue (WAT) or the brain.Correspondingly, the ex vivo biodistribution results demonstrated substantial uptake in the liver, lung, and kidneys (Figure 9B).The elevated kidney uptake may be attributed to a combination of perfusion, excretion, and reabsorption processes.As anticipated for this lipophilic tracer, the hepatobiliary system emerged as the principal pathway for tracer metabolism and excretion, as evidenced by the time-dependent increase observed in the intestines.Conversely, minimal radioactivity was discerned in the bladder, indicative of limited urinary excretion.The elevated lung uptake observed in our ex vivo study, along with the observed uptake in the liver and kidneys ex vivo and in vivo, is likely a consequence of perfusion effects rather than specific target binding.This phenomenon is expected due to the considerable blood volume and rapid perfusion rates in mice within these organs. 44n summary, [ 11 C]5c exhibited an absence of specific binding to both PD-1 negative and PD-1 expressing xenografts.This observation could potentially be attributed to rapid radiotracer metabolism, constrained tissue penetration due to pronounced plasma protein binding, or heightened levels of nonspecific or off-target interactions.To gain deeper insights into the underlying factors contributing to the absence of specific binding observed in [ 11 C]5c, further comprehensive inves-tigations and experiments may be warranted.These endeavors could encompass conducting additional in vivo studies to elucidate metabolic pathways, refining the radiotracer formulation to enhance tissue penetration, and assessing potential off-target binding via competitive binding studies or other pertinent methodologies.These findings would provide valuable insights for the development and refinement of [ 11 C]5c or similar radiotracers for PD-L1 imaging applications.

■ CONCLUSIONS
A plethora of intermediates and a collection of 10 final compounds were effectively synthesized through a series of judiciously selected reactions, including Suzuki coupling, Mitsunobu reaction, nucleophilic substitution, and reductive amination.These synthesized compounds exhibited remarkable nanomolar affinities as ascertained through a well-established HTRF assay, enabling the elucidation of profound structure− activity relationships.Notably, our synthesis efforts culminated in the creation of a novel database housing potential PD-L1 ligands.Additionally, a highly promising candidate was efficiently radiolabeled via 11 C-methylation, achieving a radio-   ■ EXPERIMENTAL SECTION General Information.Solvents and chemicals were obtained from commercial suppliers and used without further purification unless otherwise stated.All synthesized compounds were ≥95% pure as assessed by high-performance liquid chromatography (HPLC).
High-Performance Liquid Chromatography.Setup 1. Reaction progress, compound purity, and in vitro stability were measured with an Agilent 1200 series LC system (Agilent Technologies, Inc., Santa Clara, USA) paired with an Agilent 1100 series autosampler and an XBridge C18 HPLC column, 5 μm, 4.6 mm × 150 mm (Waters Corporation, Eschborn, Germany).GINA Star Software (Raytest Isotopenmessgeraẗe GmbH, Straubenhardt, Germany) was used for data acquisition.Solvent "A" consisted of a 10 mmol/L sodium phosphate (Merck KGaA, Darmstadt, Germany) buffer adjusted to pH 7.4 with 1 mol/L NaOH (Merck KGaA, Darmstadt, Germany) and solvent "B" of 90% v/ v acetonitrile (MeCN) (Merck KGaA, Darmstadt, Germany) plus 10% v/v Milli-Q H 2 O (Merck KGaA, Darmstadt, Germany).The flow rate was set to 1.5 mL/min.A mobile phase gradient of 70% A: 30% B to 20% A: 80% B within 20 min and a hold until the end of the run was used.
Setup 2. For semipreparative purification, an Agilent 1200 series LC system was paired with a SUPELCOSIL ABZ+ HPLC Column, 5 μm, 25 cm × 10 mm (Merck KGaA, Darmstadt, Germany).Solvent "A" consisted of 90% v/v MeCN plus 10% v/v Milli-Q H 2 O and solvent "B" of 10 mmol/L sodium phosphate buffer adjusted to pH 7.4 with 1 mol/ L NaOH.The flow rate was set to 5 mL/min.A mobile phase gradient of 30% A: 70% B to 60% A: 40% B within 20 min and a hold until the end of the run was used.
Setup 3.For log D measurements, an Agilent 1200 series was paired with an Agilent 1100 autosampler and Agilent 1100 UV detector, an apHera column, 5 μm, 10 mm × 6 mm (Merck KGaA, Darmstadt, Germany), GINA Star Software for data acquisition and a mobile phase gradient of 10% A and 90% B to 100% A within 9.4 min and back to starting conditions until minute 12.An equilibration time of 2 min before measurements has been set.Solvent "A" consisted of methanol (MeOH) (Merck KGaA, Darmstadt, Germany) and solvent "B" of 10 mmol/L sodium phosphate buffer pH 7.4.The flow rate was set to 1.5 mL/min.Setup 4. For semipreparative purification after radiosynthesis the GE TRACERlab FX2 C synthesis module (General Electric Medical Systems, Sweden) was paired with a Sykam S1122 pump (Sykam, Eresing, Germany), a BlueShadow UV detector (KNAUER Wissenschaftliche Geraẗe GmbH, Berlin, Germany), and a SUPELCOSIL ABZ + HPLC Column, 5 μm, 25 cm × 10 mm.The solvent consisted of 90% MeCN and 10% Milli-Q H 2 O.The flow rate was set to 5 mL/min.
For high-performance liquid chromatography measurements after radiosynthesis, an Agilent Technologies 1620 Infinity system was utilized with an XBridge BEH RP18 XP column, 2.5 μm, 3 cm × 5 cm (Waters Corporation, Eschborn, Germany), as stationary phase and GINA Star Software for data acquisition."A" consisted of 90% v/v MeCN in Milli-Q H2O and "B" of 50 mmol/L ammonium dihydrogen phosphate (Honeywell International, Inc., Charlotte, USA) adjusted to pH 9.3 with 5 mol/L NaOH.For biocide purposes, a spatula tip's worth of NaN3 was added to "B".
Setup 5. A mobile phase gradient of 40% A: 60% B to 55.5% A: 44.5% B within 5 min and a hold until the end of the run was used.Flow rate was set to 1.0 mL/min.Setup 6.An isocratic mobile phase of 50% A: 50% B and a flow rate of 1.5 mL/min was used.
Full-scan high-resolution mass spectra (m/z 50−1600) of the compounds dissolved in MeCN/MeOH and 1% H 2 O were obtained by direct infusion measurements on a maXis ESI-Qq-TOF mass spectrometer (Bruker, Mannheim, Germany).The sum formulas of the detected ions were determined using Compass DataAnalysis 4.0 (Bruker, Mannheim, Germany) based on the mass accuracy (Δm/z ≤ 5 ppm) and isotopic pattern matching (SmartFormula algorithm).
Compound characterization data is provided in the Supporting Information (Figures S10−S19).
Interference Compounds Test.Virtually and synthetically obtained structures were filtered for Pan Assay Interference Compounds (PAINS) using the ZINC online filter. 45igand Docking Experiments.Pharmacophore Screening Study.A ChEMBL database 46 containing bioactive compounds with molecular weights ranging from 4 to 200 g/mol was subjected to screening against a consensus feature-based pharmacophore derived within LigandScout 4.4 software (Inte:Ligand GmbH, Vienna, Austria).The pharmacophore was constructed based on X-ray crystallography data extracted from specific Protein Data Bank (PDB) entries, 47 including codes 5J89, 5J8O, 5N2D, 5N2F, 6R3K, and 6NM8.Pharmacophore-Fit Scores were computed for each compound, resulting hits were transposed to the PDB entry 5J89, and Binding Affinity Scores were calculated.Hits were ranked according to their Pharmacophore-Fit Scores and Binding Affinity Scores, respectively, and assigned to a maximum of 10 points each score (max.20 points total).
Ligand Docking Study.PDB 6R3K ligand was re-docked for validation of the docking procedure. 48A root-mean-square deviation (RMSD) of 0 Å was achieved between the docked and original pose, highlighting its reliability.
Newly synthesized and literature-known compound structures were protonated to pH 7.4 using MarvinSketch 22.13 software.Ligand docking was then performed with LigandScout 4.4 software using the AutoDock Vina 1.1 program and PDB code 6R3K (PD-L1 monomer C and D).The PD-L1 protein structure was maintained as a rigid entity, enabling flexible ligand docking.Water and ethylene glycol molecules were removed prior to docking.The grid box dimensions, approximately 30 × 30 × 30 Å, were automatically determined by LigandScout.Docking, performed in triplicates for enhanced consistency, adhered to default settings (Exhaustiveness: 8; Max.number of modes: 9; Max.energy difference: 3).Postdocking refinement of docking poses was not undertaken to maintain the integrity of the results.
Syntheses.General Procedures.Mitsunobu Reaction�General Procedure 1.The alcohol (1 equiv), acid (1−2 equiv), and triphenylphosphine (1.5 equiv) were dissolved in an organic solvent on ice under N 2 atmosphere, followed by slow addition of azodicarboxylate (1.5 equiv), and finally stirred at room temperature for several days.Products were purified by semipreparative silica gel chromatography.
Mitsunobu Reaction�General Procedure 2. General procedure 2 was used when general procedure 1 showed low conversion: Triphenylphosphine (2 equiv) and azodicarboxylate (2 equiv) were first mixed on ice�preforming the betaine�under N 2 atmosphere, followed by the addition of the alcohol (1 equiv), the acid (1 equiv), and eventually stirred at room temperature for several days.Products were purified by semipreparative silica gel chromatography.
Nucleophilic Substitution�General Procedure 3. The electrophile (2 equiv) was premixed with catalytic amounts of iodide salt, dissolved in organic solvent, and added to the nucleophile (1 equiv) and base (2− 3 equiv) under inert atmosphere, and stirred at room temperature for 1 day.Products were purified by semipreparative silica gel chromatography.
Radiosyntheses with Carbon-11.Radiosyntheses were performed using a GE TRACERlab FX2 C module (General Electric Medical Systems, Uppsala, Sweden).Radionuclide production and production of [ 11 C]methylating agents was performed as described before. 49In short, [ 11 C]CO 2 was produced in a GE PETtrace cyclotron (General Electric Medical Systems, Uppsala, Sweden) by irradiation of a gas target containing N 2 and 0.5% O 2 using the 14 N(p,α) 11  For automated radiosynthesis of [ 11 C]5c, [ 11 C]CH 3 I (∼51 GBq) was trapped in the reactor of the synthesis module containing 4 mg/mL precursor 4e in 250 μL of DMSO.The reaction mixture was heated for 5 min at 100 °C.After cooling, the product was purified by semiprep.HPLC setup 4. The product-containing fraction was diluted with 90 mL of H 2 O ad inj.(B.Braun, Maria Enzersdorf, Austria) and pushed through a preconditioned Sep-Pak C18 Plus Light cartridge (Waters Corporation, Eschborn, Germany).The cartridge was washed with 5 mL of H 2 O ad inj.The product was eluted with 1.4 mL of ethanol (Merck KGaA, Darmstadt, Germany) and concentrated for in vivo application by means of a SpeedVac vacuum concentrator (Thermo Fisher Scientific, Inc., Waltham, USA) at 60 °C for 30 min.The residue was reconstituted in a 0.9% NaCl solution (B.Braun, Maria Enzersdorf, Austria) and subjected to quality control assessments.Radiochemical and chemical purity evaluations were conducted utilizing HPLC setup 6.The osmolality and pH of a product sample were measured using an osmometer (Sanova Medical Systems, Vienna, Austria) and a pH meter (Metrohm, Herisau, Swiss), respectively.
In Vitro Stability Tests.Plasma stability was tested against pooled mouse plasma (Merck KGaA, Darmstadt, Germany) and pooled human plasma (Merck KGaA, Darmstadt, Germany).25 μL of formulated radiotracer were incubated with 1250 μL of plasma at 37 °C.The amount of intact tracer (%) was determined after 0, 15, 30, and 60 min.100 μL aliquots were quenched with the same amount of MeCN, centrifuged for 4 min at 4 °C with 21,380g, and analyzed by HPLC setup 1.
Plasma protein binding was assessed using 10 kDa centrifugal filters (Merck KGaA, Darmstadt, Germany).After centrifugation for 30 min at 21,380g, filtrates, and filters were measured separately in a γ counter.In addition, water was used instead of plasma to assess the nonspecific binding to filters.
Aliquots were drawn 0, 15, 30, and 60 min after the addition of 5 μL radiotracer, subsequently quenched with the same amount of MeCN, centrifuged for 4 min at 21,380g, and analyzed by HPLC setup 1 for the amount of intact tracer (%).
Animals.8-to 10-week-old male NOD.Cg-Prkdc scid Il2rg tm1Wjl /SzJ ("NSG") mice (Center for Biomedical Research and Translational Surgery, Vienna, Austria) were kept under conventional housing conditions, with food and water supply ad libitum and a 12 h day/night cycle.All animals were treated according to the European Union rules on animal care.The corresponding animal experiments were approved by the Austrian Ministry of Sciences (2021-0.422.476).
Immunohistochemistry. Immunohistochemistry was performed using primary antibodies against PD-L1 (Cell Signaling Technology, Danvers, USA, Cat.#13684) and CD31 (Cell Signaling Technology, Danvers, USA, Cat.#77699).In an autostainer (Lab Vision AS 360, Thermo Fisher Scientific, Waltham, USA), a polymer detection system with a secondary antibody conjugated to an enzyme-labeled polymer was applied.For details regarding antibodies, dilution, pretreatment, and chromogen used, see Table S1.
Tumor Grafting.The optimization of engraftment in NSG mice (n = 8) involved consideration of the quantity of injected cells and the timing of inoculation to attain an appropriate tumor size of approximately 250 mm 3 and ensure uniform tumor growth rates among experimental groups.Optimal outcomes were obtained by administering 1.5 × 10 6 cells in a PBS/matrigel (1:1) matrix (Corning, Corning, USA) over a 9-to 12-day inoculation period.
For imaging studies, NSG mice (n = 4) were injected subcutaneously with 1.5 × 10 6 CHO-K1 cells into one flank and 1.5 × 10 6 CHO-hPD-L1 cells in the opposite flank.Body weight and tumor development were monitored every second day by caliper measurement.The respective tumor volume was calculated according to the following equation: tumor volume (mm 3 ) = d 2 × D/2 (where d is the shortest diameter and D the longest diameter).The animals were subjected to μPET imaging, when tumors reached a volume of at least 200 mm 3 .Tumor volume never exceeded 1 cm 3 .There were no losses.
The acquired PET data was reconstructed reconstructed with Inveon Acquisition Workplace (Siemens Preclinical Solutions, Knoxville, TN, USA) using the OSEM3D/MAP algorithm and 18 MAP iterations on a 256 × 256 × 159 grid with a voxel size of 0.388 × 0.388 × 0.796 mm.Volumes of interest (VOIs) were created semiautomatically based on fused μPET and μCT images using PMOD software (Version 3.807; Bruker, Mannheim, Germany).The tracer uptake in the VOIs is normalized to injected dose and volume and expressed as percentage injected dose per cubic centimeter (% ID/cc).
Ex Vivo Biodistribution.Ex vivo biodistribution was assessed 30, 40, and 70 min after tracer application in NSG mice.Radioactivity was determined using a Wizard 2 γ counter.Samples were measured for 30 s, CPM-corrected for background counts, and half-life corrected to time of tracer injection.Organs were wet-weighted, and the percentage of injected dose per gram of organ was calculated (% ID/g).
Statistical Analysis.Values are depicted as mean ± standard deviation (SD), and experiments were performed in triplicates and repeated at least three times.Peak areas in the radioactivity channel were corrected for decay during HPLC measurements and radiochemical conversion (RCC) was calculated according to Formula 1. (1) where A the peak area; Rt is the retention time (min); x is the substance of interest; and i denotes other entities.

Data Availability Statement
Data is contained within the article or Supporting Information.

* sı Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/

Figure 1 .
Figure1.Overview of exemplary structures from previously reported biphenyl-based ligands, 32−34 featuring potentially advantageous substructures that serve as the foundation for the development of our compounds.A comprehensive patent review has been published before.18

Figure 3 .
Figure 3. Comprehensive overview of the identified structure−activity relationships concerning residues R 1−3 *.The moieties are systematically arranged based on ascending HTRF IC 50 values.

Figure 7 .
Figure 7. Stability of [ 11 C]5c over time in mouse plasma, human plasma, or human liver microsomes at 37 °C.

Table 1 .
Overview of Reported Small-Molecule PET Radiotracers Targeting PD-L1 and the Corresponding Results of In Vivo Investigations, Specifically Tumor Uptake

Table 2 .
Top 10Hit Structures of the Pharmacophore Screening Accounting to Their Performance According to the Pharmacophore-Fit Score and Binding Affinity Score Expressed as Overall Points (max.20) a Higher Pharmacophore-Fit Score and lower Binding Affinity Score indicate better pharmacophore fitting and affinity, respectively. a or measured HTRF IC 50 values.The Scheme 1. Synthesis Scheme of Intermediates 1, 2, and 3, Intermediates/Precursors 4, as well as Methylated, Carbonylated, or Fluorinated Final Products 5 a

Table 3 . continued
. continued a CA = commercially available.ND = not determined.* No full dose−response curves were observed, and values are represented as relative IC 50 .