Design, Synthesis, and Biological Evaluation of a Small-Molecule PET Agent for Imaging PD-L1 Expression

Immunotherapy blocking programmed cell death protein 1/programmed death ligand 1 (PD-1/PD-L1) pathway has achieved great therapeutic effect in the clinic, but the overall response rate is not satisfactory. Early studies showed that response to treatment and overall survival could be positively related to PD-L1 expression in tumors. Therefore, accurate measurement of PD-L1 expression will help to screen cancer patients and improve the overall response rate. A small molecular positron emission tomography (PET) probe [18F]LP-F containing a biphenyl moiety was designed and synthesized for measurement of PD-L1 expression in tumors. The PET probe [18F]LP-F was obtained with a radiochemical yield of 12.72 ± 1.98%, a radiochemical purity of above 98% and molar activity of 18.8 GBq/μmol. [18F]LP-F had good stability in phosphate buffer saline (PBS) and mouse serum. In vitro assay indicated that [18F]LP-F showed moderate affinity to PD-L1. Micro-PET results showed that the tumor accumulation of [18F]LP-F in A375 tumor was inferior to that in A375-hPD-L1 tumor. All the results demonstrated that [18F]LP-F could specifically bind to PD-L1 and had a potential application in non-invasive evaluation of PD-L1 expression in tumors.


Introduction
Immunotherapy based on immune checkpoint blockade has developed greatly in the past decade for the treatment of cancers, opening a new era of oncotherapy. PD-L1 is PD-1's major ligand, which can be overexpressed in a variety of malignancies [1]. Once PD-1 binds to PD-L1, not only can the PD-1/PD-L1 pathway inhibit the activation and physiological function of T cells, the proliferation of NK cells and the production of B cell antibodies, but can also promote the stability of Treg cells to enhance their inhibitory function, thus leading to immune suppression and tumor immune evasion [1,2]. The overall response rate of cancer patients followed by PD-1/PD-L1 blocked treatment is only 5-30% [3] in spite of the remarkable clinical results achieved. Some studies indicated that PD-L1-positive patients may have higher clinical therapy effectiveness and a longer median survival time after immunotherapy [4][5][6][7][8][9]. Therefore, accurate measurement of PD-L1 expression levels in tumors has important clinical guiding significance.
To date, immunohistochemistry (IHC) is still the commonly used approach to evaluate PD-L1 expression levels in the clinic. However, it has the following drawbacks. First, it is necessary to obtain tumor tissue by invasive procedures such as biopsy. Second, immunohistochemistry can only evaluate PD-L1 expression levels in pathological sections rather than reflect the whole body or metastasis [10]. Third, the expression level at tumor sites will be modulated by treatment administration, such as chemotherapy and radiotherapy [11][12][13], but IHC could not monitor it in real time. Some studies, however, have reported that there was no correlation between the immunotherapy response rate of patients receiving anti-PD-L1 immunotherapy and the PD-L1 expression level in tumors [14][15][16]. This may be caused by the however, have reported that there was no correlation between the immunotherapy response rate of patients receiving anti-PD-L1 immunotherapy and the PD-L1 expression level in tumors [14][15][16]. This may be caused by the intra-/inter-tumoral heterogeneity and the differences in the definition of PD-L1 positivity in IHC [17,18].
PET can non-invasive monitor mutations in PD-L1 expression levels in lesions in real time, dynamically and systemically, avoiding the defects of IHC. It provides a new strategy for determination of PD-L1 expression. At present, researchers have developed many PET imaging agents targeting PD-L1 based on monoclonal antibodies. For example, 64 Cu-atezolizumab could accurately assess PD-L1 expression [19] and 89 Zr-atezolizumab could predict clinical response to anti-PD-L1 blocking therapy [20]. However, antibody-based imaging agents need a long interval time between the injection and imaging to measure target expression owing to the long biological half-life [13,21]. Consequently, radionuclides with a longer half-life are demanded for labeling, resulting in high radiation dosimetry to body.
Compared with antibody-based imaging agents, small-molecule imaging agents have the advantages of low cost, good tumor penetration, fast tissue uptake and rapid imaging. For example, [ 64 Cu/ 68 Ga]DPA [22], D-dodecapeptide-based radiotracers, could image PD-L1 expression tumor with high a signal to noise ratio within 60 min. 68 Ga-NOTA-WL12 based on peptide could also acquire tumor images rapidly with reasonable radiation dosimetry [23]. Therefore, small-molecule imaging agents have great potential to measure PD-L1 expression levels rapidly. Recently, many reports have demonstrated that small molecules containing biarylmethyl aryl ether scaffold showed high affinity to PD-L1 [24][25][26][27][28]. In our previous works, two PD-L1 PET probes based on the biarylmethyl aryl ether scaffold, [ 18 F]LN and [ 18 F]LG-1, have been designed and synthesized ( Figure S16A) [29,30]. [ 18 F]LN had poor water solubility and could not show the outline of PD-L1 + tumor clearly and defluorination restricted and hindered its further application ( Figure S16B). As for [ 18 F]LG-1, the solubility and tumor to muscle ratio (T/M) value has been improved greatly ( Figure S16B), but it was necessary to take two steps of labeling to obtain the product.
Given this, a new small molecular radiotracer [ 18 F]LP-F (Scheme 1) bearing the same scaffold as [ 18 F]LN and [ 18 F]LG-1 was developed by simple one-step 18 F-fluorination. A poly(ethylene glycol) (PEG) moiety was introduced to mediate the pharmacokinetics of the tracer including increased water solubility and reduced renal clearance, which had been widely used in development of new radiotracers [31][32][33]

Synthesis of Precursor LP2 and Non-Radioactive Probe LP-F
Compounds LP2 and LP-F were prepared by the following reactions (Scheme 2). The synthesis of compounds L1-L3 had been reported in our previous work [29]. Compound L3 reacted with pyrazine through a reductive amination reaction to obtain compound LP1. Then, compound LP1 reacted with TsO-PEG4-OTs catalyzed by potassium carbonate to afford the precursor compound LP2. Substitution of compound TsO-PEG4-OTs with tetrabutylamine fluoride (TBAF) afforded TsO-PEG4-F. Similarly, LP-F was synthesized through the nucleophilic substitution reaction between LP1 and TsO-PEG4-F. Chemical structures of all the compounds were characterized by HPLC, ESI-MS and 1 H/ 13 C/ 19 F NMR spectra (Figures S1-S13).

Synthesis of Precursor LP2 and Non-radioactive Probe LP-F
Compounds LP2 and LP-F were prepared by the following reactions (Scheme 2). The synthesis of compounds L1-L3 had been reported in our previous work [29]. Compound L3 reacted with pyrazine through a reductive amination reaction to obtain compound LP1. Then, compound LP1 reacted with TsO-PEG4-OTs catalyzed by potassium carbonate to afford the precursor compound LP2. Substitution of compound TsO-PEG4-OTs with tetrabutylamine fluoride (TBAF) afforded TsO-PEG4-F. Similarly, LP-F was synthesized through the nucleophilic substitution reaction between LP1 and TsO-PEG4-F. Chemical structures of all the compounds were characterized by HPLC, ESI-MS and 1 H/ 13 C/ 19 F NMR spectra (Figure S1-S13).   Figure  1B). The radio-conversion was over 70% and the final radiochemical yield was 12.72 ± 1.98% after purification by pre-HPLC with the radiochemical purity more than 98% ( Figure 1B). The molar activity of [ 18 F]LP-F was estimated to be 18.8 GBq/μmol ( Figure  S15). The partition coefficient (Log P) of [ 18 F]LP-F was calculated to be 2.18 ± 0.16 (Table  S2). [ 18 F]LP-F showed high stability in PBS and mouse serum in 4 h as shown in Figure  1C and Figure 1D.   Figure 1B). The radio-conversion was over 70% and the final radiochemical yield was 12.72 ± 1.98% after purification by pre-HPLC with the radiochemical purity more than 98% ( Figure 1B). The molar activity of [ 18 F]LP-F was estimated to be 18.8 GBq/µmol ( Figure S15). The partition coefficient (Log P) of [ 18 F]LP-F was calculated to be 2.18 ± 0.16 (Table S2)

Cell Uptake and Specific Binding
The MTT assay was performed in PD-L1 + (A375-hPD-L1) cells and PD-L1 -(A375) cells to investigate the cytotoxicity of LP-F. The cell viability of all treated tumor cells was over 90% ( Figure 2B. However, for PD-L1 -cells, the uptake of [ 18 F]LP-F was always at a low level. The maximum uptake value was determined to be 1.41 ± 0.09 %AD at 4 h. After blockade by LP-F, the uptake of [ 18 F]LP-F in PD-L1 + cells obviously decreased to 1.42 ± 0.03 %AD at 0.5 h and 1.72 ± 0.02 %AD at 4 h, which was compared to that of PD-L1 -cells. The dissociation constant (Kd) value of [ 18 F]LP-F to the PD-L1 + cell was measured to be 226.0 nM by the saturation binding assay ( Figure 2C).

Pharmacokinetics
The pharmacokinetics was an important parameter for PET tracers, and small molecule-based probes always got benefits from the optimum pharmacokinetic characteristics. Thus, the pharmacokinetics of [ 18 F]LP-F was studied as the metabolism of radioactive dose in blood. As shown in Figure 3, the activity in blood decreased rapidly within

Cell Uptake and Specific Binding
The MTT assay was performed in PD-L1 + (A375-hPD-L1) cells and PD-L1 -(A375) cells to investigate the cytotoxicity of LP-F. The cell viability of all treated tumor cells was over 90% (

Cell Uptake and Specific Binding
The MTT assay was performed in PD-L1 + (A375-hPD-L1) cells and PD-L1 -(A375) cells to investigate the cytotoxicity of LP-F. The cell viability of all treated tumor cells was over 90% (

Pharmacokinetics
The pharmacokinetics was an important parameter for PET tracers, and small molecule-based probes always got benefits from the optimum pharmacokinetic characteristics. Thus, the pharmacokinetics of [ 18 F]LP-F was studied as the metabolism of radioactive dose in blood. As shown in Figure 3, the activity in blood decreased rapidly within

Pharmacokinetics
The pharmacokinetics was an important parameter for PET tracers, and small moleculebased probes always got benefits from the optimum pharmacokinetic characteristics. Thus, the pharmacokinetics of [ 18 F]LP-F was studied as the metabolism of radioactive dose in blood. As shown in Figure 3, the activity in blood decreased rapidly within 10 min and then eliminated slowly. After data fitting in DAS 2.1 software using a two-compartment model,

Micro-PET Imaging
Micro-PET imaging was performed in A375 (right) and A375-hPD-L1 (left) bilateral tumor-bearing nude mice. The representative dynamic images were obtained as shown in Figure 4A. [ 18 F]LP-F could show the outline of A375-hPD-L1 tumor more clearly than that of A375 tumor. The accumulation of [ 18 F]LP-F in PD-L1 + tumor increased constantly and reached to the maximum 3.53 ± 0.46 %ID/mL at 50 min ( Figure 4B) while the activity in PD-L1 -tumor was at a low level and always weaker than that in PD-L1 + tumor at any time. The maximum was observed to be 1.23 ± 0.39 %ID/mL at 60 min ( Figure 4B). [ 18 F]LP-F showed a low uptake in muscle from 0.75 to 1.62 %ID/mL within the whole measuring process. The tumor-to-muscle ratio for PD-L1 + tumor was calculated to be 2.08 ± 0.38 post injection of [ 18 F]LP-F at 30 min, and increased to 2.20 ± 0.29 at 50 min ( Figure  4C). However, the value of T/M for PD-L1 -tumor was below 1 all the time and ranged from 0.75 ± 0.19 at 50 min to 0.82 ± 0.24 at 10 min ( Figure 4C). In Figure 4D, the uptake of [ 18 F]LP-F in PD-L1 + tumor was 2.95 ± 0.39 times higher than that in PD-L1 -tumor at 50 min. Notably, there was no obvious activity on the bone indicating the non-defluorination and good in vivo stability of [ 18 F]LP-F. All the results confirmed that [ 18 F]LP-F could make a distinction between tumors with various PD-L1 expression levels.

Micro-PET Imaging
Micro-PET imaging was performed in A375 (right) and A375-hPD-L1 (left) bilateral tumorbearing nude mice. The representative dynamic images were obtained as shown in Figure 4A.  Figure 4B) while the activity in PD-L1tumor was at a low level and always weaker than that in PD-L1 + tumor at any time. The maximum was observed to be 1.23 ± 0.39 %ID/mL at 60 min ( Figure 4B). [ 18 F]LP-F showed a low uptake in muscle from 0.75 to 1.62 %ID/mL within the whole measuring process. The tumor-to-muscle ratio for PD-L1 + tumor was calculated to be 2.08 ± 0.38 post injection of [ 18 F]LP-F at 30 min, and increased to 2.20 ± 0.29 at 50 min ( Figure 4C). However, the value of T/M for PD-L1tumor was below 1 all the time and ranged from 0.75 ± 0.19 at 50 min to 0.82 ± 0.24 at 10 min ( Figure 4C). In Figure 4D, the uptake of [ 18 F]LP-F in PD-L1 + tumor was 2.95 ± 0.39 times higher than that in PD-L1tumor at 50 min. Notably, there was no obvious activity on the bone indicating the non-defluorination and good in vivo stability of [ 18 F]LP-F. All the results confirmed that [ 18 F]LP-F could make a distinction between tumors with various PD-L1 expression levels.

Discussion
Immunotherapy of blockage of the PD-1/PD-L1 interaction has made great achievements in the clinic. However, it is particularly crucial to improve the response rate. PD-L1 in tumors was the most studied biomarker to predict clinical effectiveness [34]. PET provides a new strategy to detect PD-L1 expression. Our group has designed two small-molecule radiotracers based on the biarylmethyl aryl ether scaffold, [ 18

Discussion
Immunotherapy of blockage of the PD-1/PD-L1 interaction has made great achievements in the clinic. However, it is particularly crucial to improve the response rate. PD-L1 in tumors was the most studied biomarker to predict clinical effectiveness [34]. PET provides a new strategy to detect PD-L1 expression. Our group has designed two small-molecule radiotracers based on the biarylmethyl aryl ether scaffold, [ 18 F]LN and [ 18 F]LG-1. Both of them showed specific binding to PD-L1. However, the defluorination of [ 18 F]LN and complicated two-step radiosynthesis of [ 18 F]LG-1 promoted us to further improve the design of the tracers. In this work, [ 18 F]LP-F was designed by introducing a PEG group to the same scaffold to optimize the pharmacokinetics. Furthermore, [ 18 F]LP-F could be obtained simply by one-step radiosynthesis via nucleophilic substitution of the introduced OTs group with a high radio-converting ratio. It showed high in vitro stability as proved by the radio-HPLC after incubation in PBS and mouse serum, and high in vivo stability since no obvious defluorination was observed in PET imaging. LP-F possessed more satisfactory biocompatibility than LN and LG-1.
The significant difference (p < 0.05) of the cellular uptake between the PD-L1 + and PD-L1cells demonstrated that [ 18  Introduction of hydroxyl groups to the biarylmethyl aryl ether scaffold in the solvent interaction region is another feasible way of modification to increase the solubility and the PD-L1 affinity of small-molecule PET agent, which could increase more accumulation and more retention time in PD-L1 + tumors and accelerate clearance from non-target organs. Construction of satisfactory PD-L1 tracers needs to coordinate all the factors, including lipophilicity, polarity, stability and radiolabeling methods to mediate the high binding Pharmaceuticals 2023, 16, 213 7 of 12 affinity, which still remains a big challenge to fit within the routine clinical work. Such small-molecule PET agent could visualize PD-L1 + tumor rapidly within 2 h, which would lessen the burden on patients in terms of the total time of examination and the dose of radiation absorbed. The diagnostic reports could provide significant clinical guidance for patient stratification and recommend that patients whose tumors are PD-L1 + should receive immunotherapy.   18 F nuclear reaction through a medical cyclotron (HM7, Sumitomo Heavy Industries) and trapped on an anion-exchange Sep-Pak light QMA column, which had been activated by NaHCO 3 aqueous solution (0.5 M, 10 mL) and distilled water (10 mL). The QMA cartridge with [ 18 F]fluoride was eluted with a solution of Kryptofix 2.2.2 and K 2 CO 3 in 1.5 mL of acetonitrile/water. The solution was evaporated using a stream of nitrogen at 110 • C and coevaporated to dryness with ACN (2.0 mL) to remove water. The labeling precursor, compound LP2 (approximately 3.6 mg), dissolved in DMSO (0.6~0.7 mL), was added and reacted with 18 F -(approximately 11.1 GBq) at 110 • C for 30 min ( Figure 1A). The 18 F-labeled crude product was analyzed by radio-HPLC by comparing the retention time (t R ) of [ 18 F]LP-F with that of the non-radioactive probe LP-F. The [ 18 F]LP-F was subjected to semi-preparative HPLC, diluted in water, concentrated on C18 light Sep-Pak cartridge, washed by water and eluted by ethanol (0.8~1.0 mL).

Molar Activity Test
The UV spectrum of compound LP-F was drawn by UV spectrophotometer to aquire the λ max of LP-F. Concentrations of LP-F were subjected to UV-HPLC at λ max (286 nm) to establish calibration curve using peak areas of corresponding concentrations of LP-F and afforded linear regression equation through peak areas and concentrations. After injection of purified [ 18 F]LP-F, the UV peak area at 286 nm of purified [ 18 F]LP-F was used to calculate the injection concentration according to above linear regression equation. The molar activity was calculated based on Eq.
Molar activity = Injection radioactivity/Amount of substance injected

In Vitro Stability Assay
In order to investigate the in vitro stability of the radioactive probe, [ 18 F]LP-F (~14.8 MBq, 40 µL) was incubated in PBS (pH 7.4, 360 µL) at 37 • C for 4 h. [ 18 F]LP-F (~14.8 MBq, 40 µL) was also mixed with mouse serum (360 µL) and maintained at 37 • C for 4 h. Such PBS solution was subjected to stability analysis using radio-HPLC at 0, 1, 2 and 4 h. A 30 µL sample from the above mouse serum after protein precipitation with 30 µL acetonitrile was also subjected to stability analysis at 0, 1, 2 and 4 h.

Partition Coefficient Test
Amounts of 0.37 MBq [ 18 F]LP-F, 1000 µL water and 1000 µL 1-octanol were mixed in a tube. The mixture was thoroughly shaken for 5 min. After centrifugation, 500 µL water was withdrawn from water phase and its radioactivity was measured by a gamma counter. An equal volume of 1-octanol was withdrawn from the organic phase and measured. The partition coefficient (Log P) was calculated based on Eq. Then, 500 µL water and 500 µL 1-octanol were added to the above tube to replenish the water and the organic phase. The two-phase system was shaken again. Water and the organic phase were sampled and measured and Log P was calculated again. Repeat two more times.
Female BALB/c nude mice at age of 5~7 weeks were inoculated subcutaneously with 1 × 10 6 A375-hPD-L1 cells in saline at the left flank and 1 × 10 6 A375 cells at the right flank in the same mouse in order to eliminate individual differences. Animal study can be performed in accordance with the principles laid out by the ethical committee of Jiangsu Institute of Nuclear Medicine.

Cell Viability Assay
Each well was seeded with 1 × 10 4 cancer cells in 96-well plates. LP-F dissolved in DMEM with gradient concentration (0, 12.5, 25, 50 and 100 µM) was added to each well at triple parallels, respectively, and incubated for 24 h. Then, MTT (5 mg/mL, 20 µL/well) was added and incubated for 4 h. 150 µL DMSO was added each well and shaken vigorously for 10 min to dissolve MTT crystallization. The absorbance of each sample was detected at 490 nm by ELISA, and percent cell viability was calculated based on Eq.: Percentage = sample well/reference well × 100%

The Saturation Binding Assay
PD-L1 + cells (2 × 10 5 ) were added to each tube and incubated with the gradient of concentration of [ 18 F]LP-F dissolved in DMEM (0 to 128 nM) at 37 • C for 1 h. After incubation, the cells were centrifuged and washed by PBS twice and then were measured for radioactivity by a γ counter as total bound value. For the non-specific bound value, PD-L1 + cells (2 × 10 5 ) were added to each tube and incubated with LP-F (50 µM) dissolved in DMEM for half an hour in advance and then incubated with above concentrations of [ 18 F]LP-F. After centrifugation and wash, the radioactivity was detected. The specific binding value is obtained by subtracting non-specific bound value from total bound value. The equilibrium dissociation constant (K d ) was calculated by non-linear fit in a one-site model through specific binding values and concentrations of [ 18 F]LP-F. Dynamic scanning was performed for one hour. ASIPro software (Siemens) was applied to process the PET images. Dynamic images were reconstructed in OSEM3D/MAP algorithm and split into twelve frames. Quantification and analysis of region of interest (ROI) in tumors and organs were carried out.

Pharmacokinetics
Female BALB/c normal mice (n = 3) were injected (i.v.) with [ 18 F] LP-F (~7.4 MBq, 200 µL) via tail vein and the tails were cut off (p.i.) to obtain the blood samples at setting time points. Meanwhile, 25 µL [ 18 F] LP-F (~0.925 MBq) was set as reference. After weighing, radioactivity of each blood sample was tested immediately. The drug concentrationtime curve was draw. DAS 2.1 software was employed to calculate the pharmacokinetic parameters. A compartment model was used to analyze the distribution and clearance process of the radioactive probe [ 18 F] LP-F in vivo to demonstrate the results of PET and the biocompatibility of the radiotracer.

Statistical Analysis
Data were presented as the mean ± SD. The differences among groups were analyzed by two-tailed Student's t-test indicated as * p < 0.05, ** p < 0.01 and *** p < 0.001.

Conclusions
In summary, a small molecule, [ 18 F]LP-F, radiolabeled with 18 F was explored as a novel PET imaging radiotracer for assessment of PD-L1 expression in tumors. [ 18 F]LP-F showed high stability in vitro and in vivo. The moderate affinity to PD-L1 made it selectively accumulate in PD-L1 overexpression tumor cells. The PET imaging studies demonstrated its ability to measure PD-L1 expression in vivo. This study provided opportunities to explore the design of small-molecule PET tracers for assessing PD-L1 expression in tumors.
Author Contributions: Methodology, investigation, software, formal analysis, visualization, writingoriginal draft preparation, L.X., L.Z., B.L. and S.Z. Methodology, validation, writing-review and editing, funding acquisition, G.L. Resources, data curation, writing-review and editing, visualization, supervision, project administration, funding acquisition, L.Q. and J.L. All authors have read and agreed to the published version of the manuscript.