Nonsymmetrically Substituted 1,1′-Biphenyl-Based Small Molecule Inhibitors of the PD-1/PD-L1 Interaction

Therapeutic antibodies directed against either programmed cell death-1 protein (PD-1) or its ligand PD-L1 have demonstrated efficacy in the treatment of various cancers. In contrast with antibodies, small molecules have the potential for increased tissue penetration; better pharmacology; and therefore, improved antitumor activity. A series of nonsymmetric C2 inhibitors were synthesized and evaluated for PD-1/PD-L1 interaction inhibition. These compounds induced PD-L1 dimerization and effectively blocked PD-L1/PD-1 interaction in a homogeneous time-resolved fluorescence (HTRF) assay with most inhibitors exhibiting IC50 values in the single-digit nM range and below. Their high inhibitory potency was also demonstrated in a cell-based coculture PD-1 signaling assay where 2 exhibited an EC50 inhibitory activity of 21.8 nM, which approached that of the PD-L1 antibody durvalumab (EC50 = 0.3–1.8 nM). Structural insight into how these inhibitors interact with PD-L1 was gained by using NMR and X-ray cocrystal structure studies. These data support further preclinical evaluation of these compounds as antibody alternatives.

−9 ICIs based on the antibodies against the PD-1/PD-L1 pathway are currently the cornerstone of this cancer immunotherapy. 1,9−15 In addition, higher rates of toxicity are expected when these drugs are combined with chemotherapy and other immunotherapeutic agents. 2,4−23 Nevertheless, small molecule checkpoint inhibitors have been actively pursued, 15,24−26 and the ongoing clinical trials of orally bioavailable checkpoint inhibitors are the culmination of these efforts. 15,26,27erein, we show a group of "nonsymmetric" compounds that bind to PD-L1 and are based on an amide linker attached to a biphenyl structure (Table 1).These compounds effectively block the PD-1/PD-L1 interaction in a number of in vitro and ex vivo assays.
A review of the literature suggests two distinct families of low-molecular-weight molecules that bind to PD-L1: (a) those derived from the biphenyl moiety (Figure S1), 23 originally identified by researchers at Bristol-Myers Squibb (BMS) scientists, 28,29 and (b) amino acid-based small molecules that mimic the PD-1/PD-L1 interface interaction. 15,24,30The detailed properties of the compounds based on biphenyl have been described in several publications. 22,23,31,32By analyzing these data together, we were able to divide the biphenyl core compounds into two pharmacophores.The first Table 1.PD-1/PD-L1 Inhibitory Activity of the Compounds, wherein "Nonsymmetric" Means That R1 ≠ R2 a Compound A in Park et al. 25 was used here as a reference compound.The EC 50 of Compound A reported is 17 nM. 25b BMS-1166 of BMS. 29,30harmacophore, termed "short," consists of a core group, a linker, an aryl moiety, and a tail group (Figure S1A). 32,33The core is a biphenyl moiety that is located in a hydrophobic pocket composed of the amino acids Tyr56, Met115, and Ala121of PD-L1. 22,31The aryl group is typically a five-or sixmembered aromatic ring or fused ring connected to the core group by a linker.The terminal tail is oriented toward the solvent region and could form interactions with nearby residues via hydrogen bonding.The second pharmacophore includes the compounds with the so-called C2 symmetry or pseudosymmetry, which contain polar groups at the termini (Figure S2). 34These compounds, referred to here as "long" or "C2," have the same composition as the first pharmacophore but they have an additional linker, aryl, and tail group on the other side of the core moiety (Figure S2). 34MR and X-ray cocrystal structure studies of small molecule/PD-L1 complexes showed that the C2 molecules, such as LH1307 (Figure S2B), 34 form a more symmetric PD-L1 dimer than previously reported for the short inhibitors.The 2-(acetamido) ethylamine polar groups of BMS-202 incorporated into the C2 symmetric molecules of LH1307 are reported to extend from the hydrophobic cleft and occupy the solvent-exposed region sandwiched between the β strands of the PD-L1 dimer, thereby further enhancing binding to the two PD-L1 monomers. 34Another C2 compound induces a side chain flip of the Tyr56 protein residue to form a new cavity, which results in higher binding affinity to PD-L1 and higher PD-1/PD-L1 inhibitory activity under physiological conditions, as reported for Compound 4. 35 Representative structures of compounds that have been disclosed by BMS, Incyte, Gilead, and others are shown in Figure S2B.
The C2-"long" compounds, themselves, can be symmetric or nonsymmetric depending on the linkers, aryl, and tail groups used.25,34 The linkers of the compounds in Figure S1 and Figure S2 include the ether, 36 alkene, and the amide moieties.Among these, the amide linker is a privileged linker given its druglike properties.37,38 The amide linker was first disclosed by Incyte and later on adopted by several other groups and companies (Figure S1 and Figure S2).25 In summary, our compounds are composed of a biphenyl in a core, an amide linker, and an arylpyridine group, and various tail solubilizing groups attached to the core (Table 1).
Three types of paths were used to obtain the final structures.These are described in Scheme 1, Scheme 2, and Scheme 3.
Schemes 1−3 present synthesis descriptions for a representative compound for each pathway.The term "xxx" refers to a specific final compound.The intermediates are additionally marked with the step number of each pathway.
In Scheme 1, in the first steps, we obtained the biphenyl core, which was synthesized in the Suzuki coupling of tertbutyl(3-bromo-2-methylphenyl)carbamate and 2-methyl-3-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)aniline.Next, in the amide formation reaction with 5-(methoxycarbonyl)picolinic acid, we obtained an ester intermediate, which in further transformations gave a halide derivative.Then, in the S N 2 reaction, we introduced the first solubilizer, R1, and then, after protection with tert-butyldimethylsilyl chloride and the second formation of the amide, we introduced the aldehyde, which in the reductive amidation reaction allows the entry of the second solubilizer�R2.The route described in Scheme 2 differs from the previous one in that both amide groups are introduced sequentially, distinguishing the sides of the molecule by protecting one hydroxyl group.The solubilizers were then introduced via S N 2 reactions.
For the pathway shown in Scheme 3, another intermediate was used in the second amide formation reaction: 5bromopicolinic acid.The bromine derivative was converted to vinyl by Suzuki−Miyaura cross-coupling reactions of potassium vinyl trifluoroborate.The vinyl was then converted to an aldehyde for reductive amidation.During the synthesis, modifications were made to the synthetic pathways to improve efficiency.
The synthesized inhibitors were evaluated by HTRF (homogeneous time-resolved fluorescence assay; Figure S3A), which allows for the determination of the range of inhibition of the compounds on the interaction of PD-1 with PD-L1 (Table 1).The compounds were further checked for their reactivation of PD-L1-blocked effector T cells using the cell-based immune checkpoint blockade (ICB) assay [EC 50 (half-maximal effective concentration) values in Table 1, Figure S3B].The results show that the type of solubilizer used influences the affinity of the target protein.Alkyl alcoholsolubilizing groups, e.g., derivatives of ethanolamine, generally perform better in the ICB cell-based assay than their cyclic counterparts, such as hydroxycyclopentyl.
BMS-1166 was used as a reference compound in the HTRF measurements.As can be seen in Table 1, only two compounds showed lower activity in HTRF than reference BMS-1166.The rest of the synthesized compounds showed a lower amount of undissociated protein complex at 5 nM inhibitor concentration.A total of 83% of dissociation of the PD-1/PD L1 complex was achieved using 2-(ethylamino)ethan-1-ol and N- [3-(ethylamino)propyl]methanesulfonamide (17a) as a solubilizer.Interestingly, this compound an HCl salt of 17 where we observed only 17% of the undissociated complex in the HTRF assay, which proves that the presence of the hydrochloric salt increases the solubility of the compound (Table S2).A similar trend was observed for 2 and its HCl salt 2a (Table S2).
Good results can also be seen in the Promega cell assay.Compound A, 25,39 a resynthesized compound from a patent with symmetric analogues, 39 was used as a reference and showed activity in our cell assay with an EC 50 of 14.7 nM.However, for compound 4 with additional (1S,2S)-2-aminocyclopentan-1-ol, the EC 50 was above 2000 nM.The compound with a less polar solubilizing group�pyrrolidine and methyl ethyl alanine (7)�has no activity in the assay.EC 50 values above 180 nM were observed for compounds with N-methyl-4-piperidinol and 1-ethyl-4-(methylsulfonyl)piperazine as solubilizing tags (9, 10, 11, and 15).
Representative molecules were evaluated for their interaction with PD-L1 by means of an 1 H NMR titration experiment.NMR showed that the compounds induced oligomerization of the protein (Figure 1).Upon addition of compounds 14, 17, and 17a to hPD-L1 (human PD-L1), the proton NMR lines of hPD-L1 broaden, thereby indicating oligomerization of hPD-L1, which we have previously observed with similar compounds. 31,40,41n addition, we performed a PD-1/PD-L1 NMR-based antagonist-induced dissociation assay (AIDA). 42,43The AIDA assay was performed by adding a slight molar excess of PD-L1 to PD-1, which causes 1 H NMR signals to broaden corresponding to complex formation (Figure 2).The addition of compounds that displace PD-1 from PD-L1 results in rescue of the PD-1 signal.There is a 100% rescue percentage for 17a, which indicates that 17a was able to completely displace PD-1.The spectra in Figure 2 (purple) and Figure 1 (purple) match those in the broader signals of PD-L1.However, in Figure 2 (purple), these are obscured by the sharper NMR signals of monomeric PD-1 released from the PD-L1/PD-1 complex.Hence, the similarity of the purple spectrum (bottom) is the upper blue spectrum in Figure 2.
Structural Studies of the 17a Interactions with PD-L1.We managed to cocrystallize PD-L1 with compound 17a and solve the structure at the resolution of 2.5 Å (Figure 3 and Table S1).Attempts to crystallize 2 and 2a failed to produce suitable diffracting crystals.The asymmetric unit of the PD-L1 costructure with 17a with the P12 1 1 space group contains six chains arranged in three dimers bridged by the inhibitor found at the dimer interface.The 17a electron density was welldefined at the PD-L1 dimer interface with the exception of sulfonamide solubilizer at the R2 position (Figure 3B).The inhibitor pose on PD-L1 dimer interface resembles those reported for C 2 -symmetric inhibitors with highly symmetrical hydrophobic interactions, including π−π stacking between pyridine rings and A,B Tyr56, as well as π−alkyl interactions with A,B Ile54, A,B Met115, and A,B Tyr123 (comparison between 17a pose and C2-symmetric anti-PD-L1 compound PDB ID: 6RPG in Supplementary Figure S4). 34However, the positioning of distal phenyl rings are different, likely because of different linkers between biphenyl core and more branched substitution in the case of PDB ID: 6PRG.Moreover, numerous symmetrical hydrogen bonds are also reported, such as between amide carboxylate linkers and A,B Met115, or nonsymmetric with A Lys124 or B Asp122.Sulfonamide polar group, although smilingly stabilized by hydrogen interactions with B Thr20, is not resolved well in the electron density and is mainly water-exposed outside of the "hydrophobic tunnel" between PD-L1 monomers, likely contributing to the solubility of the 17a.
Our work provides the characterization of low-molecularweight C2 nonsymmetric binders to PD-L1 obtained by providing the central biphenyl moiety on either side with the amide linkers attached to an aryl ring and a tail group that are different on these two sides (Table 1).Similar C2-long compounds have been reported in the scientific and patent literature.However, the first group is based on an ether linker (i.e., different from the amide linker), and these compounds are rather weak inhibitors. 34,35For the C2 compounds with the amide linker, the solubilizing groups were identical at both ends of the molecules, 25,37 and the majority of the reports included only the in vitro HTRF binding data and did not provide the activity in an ex vivo cellular environment. 37,39TRF assay measures compound activity in PD-1/PD-L1 complex dissociation.Many potent compounds in the HTRF assay have little or even no activity in the cellular context; see, for example, Konieczny et al. 44 where the "best" compound  The complex of hPD-1/hPD-L1 with 17a (purple) in a 1:1 molar ratio of protein to compound 1:1.Rescue of the signal intensity by addition of 17a to the complex (purple) suggests that the compound displaces PD-1 from PD-L1.
there for PD-L1 had an excellent HTRF IC 50 of 14.9 nM but a poor EC 50 activity in the PD-1/PD-L1 cells of only 6632 nM.
In conclusion, the compounds described in this study are potent antagonists of PD-1/PD-L1 complexation, as seen in conventional in vitro assays, such as NMR and HTRF and, importantly, the NFAT assay.Activation of Jurat cells by our "best" compounds was at EC 50 levels approaching those of the control antibodies despite the fact that they were significantly smaller in size.Like other PD-L1 binders described in the literature, 31 our compounds are dimerizers of human PD-L1.Their specificity is to hPD-L1.In the NFAT assay, the compounds had an EC 50 (half-maximal effective concentration) below 1000 nM, with 79% of the compounds having an EC 50 cutoff below 300 nM (Table 1).−37 The PD-1/PD-L1 inhibitors with an EC 50 above this cutoff are not considered potent enough to stimulate T cells to clinically relevant anticancer activity.Our most potent compound, 2, features exceptionally high target affinity and demonstrated potency in cell-based assays with an EC 50 of 21.8 nM.This identifies 2 as an excellent candidate for further preclinical and clinical studies in anti-PD-L1 cancer therapy.The X-ray structure presented helps to explain the enhanced inhibitory activity of these inhibitors.
■ EXPERIMENTAL SECTION General Chemistry.Syntheses were carried out following the procedures summarized in Scheme 1, Scheme 2, and Scheme 3. Additional information can be found in the Supporting Information.
UHPLC-MS/MS analyses were performed using a Waters ACQUITY UPLC instrument (Waters Corporation, Milford, MA, USA) coupled to a Waters TQD mass spectrometer.Chromatographic separation was performed using a 2.1 × 100 mm Acquity UPLC BEH C18 column with a particle size of 1.7 μm and equipped with a VanGuard Acquity UPLC BEH C18 column with a size of 2.1 × 5 mm and a particle size of 1.7 μm.The column was maintained at 40 °C and eluted with a gradient from 95% to 0% eluent A over 10 min and then isocratically using 100% eluent B over 2 min at a flow rate of 0.3 mL min −1 .Eluent A was water/formic acid (0.1%, v/v); eluent B was acetonitrile/formic acid (0.1%, v/v).Diode array detector (DAD) chromatograms were recorded by using a Waters eλ PDA detector.The spectra were analyzed in the range of 200−700 nm with a resolution of 1.2 nm and a sampling rate of 20 points s −1 .The MS analysis parameters were as follows: source temperature 150 °C, desolvation temperature 350 °C, desolvation gas flow 600 L h −1 , shield gas flow 100 L h −1 , capillary potential 3.00 kV, and cone potential 30 V. Gas nitrogen was used for nebulization and desolvation.Data was collected in the range from 50 to 1000 m/z in intervals of 0.5 s.
Homogeneous Time-Resolved Fluorescence.The certified CisBio kit was used to conduct the HTRF assay, adhering to the manufacturer's guidelines.The assay setup comprised hPD-1, hPD-L1, anti-Tag1 tagged with Europium cryptate (HTRF donor), and anti-Tag2 tagged with XL665 (HTRF acceptor).The experiments utilized final concentrations of 5 nM hPD-L1 and 50 nM hPD-1 in a 20 μL total volume performed in triplicate.Per the CisBio instructions, all components were combined, the plate was incubated at ambient temperature for 2 h, and the TR-FRET (time-resolved fluorescence energy transfer) was carried out using the Tecan Spark 20M.The negatives in the acquired data were subtracted for background, the positives were standardized, and then the data were averaged.
PD-1/PD-L1 Blockade Bioassay.The PD-1/PD-L1 immune checkpoint bioassay (PD-1/PD-L1 Bioassay, Promega) was conducted according to the manufacturer's protocol.PD-L1+ aAPC/ CHO-K1 cells were seeded at a density of 40 × 10 4 cells in 100 μL of medium [Ham's F12, 10% fetal bovine serum (FBS)] in 96-well white flat-bottomed plates and incubated overnight at 37 °C in the presence of 5% CO 2 .The next day, the medium was removed, and serial dilutions of small molecule inhibitors were added at 40 μL per well in assay buffer (RPMI1640 + 1% FBS + 1% DMSO).PD-1 effector Jurkat cells resuspended in assay buffer (RPMI 1640 + 1% FBS) at 1.25 × 10 6 cells/mL were then added at 40 μL per well (total of 50 × 10 4 cells).The cells were cocultured for 6 h at 37 °C with 5% CO 2 followed by a 5 min equilibration at room temperature.Bio-Glo Reagent (Promega) was prepared and added at 80 μL per well.After a 15 min incubation at room temperature, luminescence was measured using a Tecan Spark microplate reader.Data are presented as fold induction of luminescence relative to DMSO-treated cells.EC 50 values were calculated using the log(inhibitor) vs response variable slope (four-parameter) analysis model with GraphPad Prism software.
Comparison of the biphenyl-based inhibitor structures.Experimental procedures for substrate preparation; results of the PD-1/PD-L1 blockade bioassay assay; Xray data collection and refinement statistics (molecular replacement) for the PD-L1 cocrystal structure with 17a; results of the ICB assay; compound purity and NMR data (PDF) Strings of the molecular formula (CSV)

Accession Codes
The structure factors and final models of PD-L1 complexes with inhibitor 17a have been deposited in the Protein Data Bank under accession number 9EO0.

Figure 2 .
Figure 2. NMR-based AIDA assay showing that a compound displaces PD-1 from PD-L1.The aliphatic part of the 1 H NMR spectra of hPD-1 (blue), hPD-L1 (red), the complex of hPD-1/hPD-L1 (green), and the PD-1 signals at −0.4 ppm (green) are broadened by the addition of PD-L1 (red) compared with PD-1 alone (blue).The complex of hPD-1/hPD-L1 with 17a (purple) in a 1:1 molar ratio of protein to compound 1:1.Rescue of the signal intensity by addition of 17a to the complex (purple) suggests that the compound displaces PD-1 from PD-L1.

Figure 3 .
Figure 3. X-ray structure of 17a in complex with hPD-L1 (PDB: 9EO0).(A) Arrangement of one of the three dimers in the asymmetric unit.Two PD-L1 molecules with chain A in gray and B in orange form an interface with 17a represented as blue sticks.(B) Detailed interactions of 17a with PD-L1 dimer.The inhibitor interacts with PD-L1 to induce its dimerization.Color coding is as in (A).The electron density of 17a (2Fo-Fc map at 1 σ) represented as a gray isomesh.