An Optimized Radiosynthesis of [¹⁸F]DK222, a PET Radiotracer for Imaging PD-L1

Abstract

A radiochemical synthesis of [¹⁸F]DK222, a peptide binder of programmed death ligand 1 protein, suitable for human PET studies is described and results from validation productions are presented. The high specific activity radiotracer product is prepared as a sterile, apyrogenic solution that conforms to current Good Manufacturing Practice (cGMP) requirements. In addition, the production is extended to use a commercial synthesizer platform (General Electric FASTlab 2).

Introduction

Elevated expression of the immune checkpoint protein programmed death-ligand 1 (PD-L1) is associated with many cancers including non-small cell lung carcinoma, melanoma, Hodgkin’s lymphoma, kidney, bladder, and breast cancer.¹ Recently [¹⁸F]DK222, a 14-amino acid human PD-L1 specific cyclic peptide modified with a bifunctional chelator ((2,2′-(7-(4-isothiocyanatobenzyl)-1,4,7-triazonane-1,4-diyl)diacetic acid) was described as a novel radiotracer for detecting PD-L1 levels non-invasively using positron emission tomography (PET).² In that publication, a preliminary synthesis of the radiotracer was described that was useful for animal studies, but not scalable for anticipated future human PET studies.

This report describes an optimized and scalable radiosynthesis of [¹⁸F]DK222 with full cGMP-compliant quality control specifications and results. This procedure is suitable for future human PET studies.

Standard Reagent Statement

Reagents and solvents were obtained from Millipore, Sigma-Aldrich, or Thermo Fisher Scientific in either ACS or HPLC grade, unless otherwise noted. The precursor (cyclo-(Ac-Tyr-NMeAla-Asn-Pro-His-Glu-Hyp-Trp-Ser-Trp(Carboxymethyl)-NMeNle-NMeNle-Lys(NODA_NCS)-Cys-)-Gly-NH2; known hereafter as simply DK-NODA; Figure 1) was custom synthesized from CPC Scientific (San Jose, CA). The authentic standard (cyclo-(Ac-Tyr-NMeAla-Asn-Pro-His-Glu-Hyp-Trp-Ser-Trp(Carboxymethyl)-NMeNle-NMeNle-Lys(NODA_NCS-AlF)-Cys-)-Gly-NH2; known as DK222, Figure 1) was synthesized in-house.²

Figure 1:

Synthesis of [¹⁸F]DK222

The reaction vials used for this synthesis were washed in nitric acid (6 M), then rinsed sequentially with Milli-Q water and absolute ethanol. The vials were dried and stored at >80° C until their use, at which time they were allowed to cool to room temperature.

The analytical chromatography system included an Agilent 1260 Infinity System equipped with a quaternary pump, HiP ALS autosampler, and DAD UV detector with a Max-Light flow cell set to 254 nm as well as a Bioscan Flow-Count interface with a NaI radioactivity detector. The analytical HPLC column was a Phenomenex Luna C18(2) 5μ, 4.6 x 150 mm column eluted with a gradient starting at 30% acetonitrile : 70% water : 0.3% trifluoroacetic acid and ending with 47.2% acetonitrile : 52.8% water : 0.3% trifluoroacetic acid at a flow rate of 1.5 mL/min. Chromatographic data were acquired and analyzed with Agilent OpenLAB CDS EZChrom edition chromatography data system (Rev. A.04.09).

Radioactivity measurements were made using a Comecer model IBC-LITE dose calibrator (Castel Bolognese, Italy) with Veenstra VIK-202/203 ionization chambers (Joure, The Netherlands).

Residual solvent levels were analyzed using an Agilent 7890A gas chromatograph, Agilent OpenLAB chromatography data system for data acquisition and analysis, and a WAX (Polyethylene glycol phase: USP G16, G20) 30 meters, 0.25 mm ID, 0.25 mm film column.

Endotoxin testing was performed on a Charles River Laboratories Endosafe^® nexgen-PTS (Wilmington, MA).

Sterility testing was conducted using aerobic and anaerobic media according to USP <71>.³

Radionuclidic purity was examined on an AccuSync SA1000-1S 1024-channel spectrum analyzer (Milford, CT).

Synthesis Procedure

Preparation of Reagents

The following reagent solutions were prepared and used on the day of the synthesis.

L-ascorbic acid (22 g) was dissolved in Milli-Q water (250 mL) to produce a 0.5 M L-ascorbic acid solution. Potassium acetate (10 mg) was dissolved in TraceSelect^® water (1 mL, Fisher Scientific).

Aluminum chloride hexahydrate (1.5 mg) was dissolved in an aliquot of the 0.5 M L-ascorbic acid solution to produce a solution with a calculated concentration of 0.43 mg/mL.

An 80/20 mixture of acetonitrile (3.2 mL) with the 0.5 M L-ascorbic acid (0.8 mL) solution was prepared. The precursor (DK-NODA, 1 mg) was dissolved in an aliquot of the 80/20 mixed acetonitrile/ascorbic acid solution to produce a solution with a calculated concentration of 1 mg/mL.

(+)-Sodium L-ascorbate (338 mg) was dissolved in a 50-mL vial of Sodium Chloride for Injection, USP (Hospira).

A tC18 Sep Pak plus cartridge (Waters, MA) was flushed with 10 mL of 0.5 M ascorbic acid. A Chromafix 30-PS-HCO₃ SPE resin cartridge (ABX GmbH, Germany) was conditioned with 1 mL of water (TraceSelect).

To a previously cleaned, heat-dried and cooled vial, the aluminum chloride solution (45 μL) is added, followed by acetonitrile (150 μL), 0.5 M ascorbic acid (50 μL), and the DK-NODA precursor solution (200 μL).

Production and Trapping of [¹⁸F]Fluoride

[¹⁸O]Water (98%, Huayi Isotopes (Shanghai, China), approx. 1.7 mL) was loaded into a niobium-body, high-yield [¹⁸F]fluoride target of a General Electric Medical Systems PETtrace cyclotron (GEMS; Waukesha, WI). The target was bombarded with a 16 MeV proton beam of 60 μA for up to 30 min to produce aqueous [¹⁸F]fluoride via the ¹⁸O(p,n)¹⁸F nuclear reaction. Target water was transferred into a clean and dry vial where its radioactivity content was assayed using a dose calibrator. Using remote manipulators, an aliquot (between 50 to 300 mCi; 1.8 to 11 GBq) was loaded onto the anion cartridge for trapping [¹⁸F]fluoride.

Synthesis of [¹⁸F]DK222

The resin cartridge was connected directly to the vial containing DK-NODA and the cartridge was eluted with the potassium acetate solution (150 μL) after which the cartridge was removed, and the vial was assayed for radioactivity content. The reaction vial was heated to 110° C for 10 minutes, then cooled to less than 50° C. The vial was then connected to a custom-built solid phase extraction (SPE) module for purification and formulation (see supplemental data).

Purification and Formulation of [¹⁸F]DK222

Purification of the crude [¹⁸F]DK222 reaction mixture was accomplished by the SPE module by the addition of 0.5 M ascorbic acid (4.5 mL) to the reaction vial, after which the diluted solution was aspirated into a glass syringe. The contents were passed through the conditioned tC18 Sep Pak plus cartridge at 300 μL/second. The module washed the cartridge with 0.5 M ascorbic acid (5 mL at 400 μL/second), then eluted the tC18 cartridge with absolute ethanol (1 mL at 185 μL/second) followed by 14 mL (300 μL/second) of Sodium Chloride for Injection, USP containing (+)-Sodium L-ascorbate through a Millipore 0.2 μm FG sterilizing filter into an empty sterile and pyrogen-free product vial (Huayi, NUCMEDCOR, San Francisco, CA).

GE FastLab2^™ synthesis of [¹⁸F]DK222

The [¹⁸F]DK222 radiosynthesis procedure was adapted to run on a GE Fastlab2 synthesis module using the GE FastLab developer software version 3.2.0.2. [¹⁸F]DK222 synthesis cassettes were built using GE developer kit components (see Figure 4 in supplemental data). The Chromafix 30-PS-HCO₃ SPE resin cartridge was manually conditioned with 1 mL of water then attached to the cassette immediately before the synthesis. The reagent preparation was the same as described above. The 1mg/mL potassium acetate stock solution was transferred to a GE glass vial (11 mm, 3 mL) fitted with rubber stopper and 11 mm cap and inserted into position V2 (see Figure 5 in supplemental data). Immediately before delivery of [¹⁸F]fluoride, the GE FastLab2 reactor vent line (right) and addition line (center) were removed so the precursor solution could be manually added (Note: Due to the small volumes being added, manual addition of the reactants to the reactor was necessary). Using variable volume pipettes (Fisher Scientific), aluminum chloride (45 μL), acetonitrile (150 μL), 0.5M ascorbic acid (50 μL), and DK222-NODA solution (200 μL) were added to the reactor via the center addition port and the lines were reconnected. Cyclotron-produced [¹⁸F]fluoride ion (between 226 - 995 mCi; 8.4-36.8 GBq) was delivered to the GE Fastlab2 and trapped on the Chromafix 30-PS-HCO₃ SPE resin cartridge. The FASTlab2 sequence directed syringe 1 to elute the fluoride ion from the resin with 150 μL of the potassium acetate solution into the reactor. The resin cartridge flow path was flushed with nitrogen for 30 seconds, then the reactor was heated at 110° C for 10 minutes. The reactor was cooled to 50° C followed by purification and reformulation sequence which was completed by the FastLab2 using the same procedure as described above.

Quality Control Procedures

Quality control (QC) procedures for [¹⁸F]DK222 performed based upon current requirements for radiotracers set forth in the U.S. Pharmacopeia⁴ are summarized below. Data for 3 validation batches of [¹⁸F]DK222 produced using the methods disclosed herein are summarized in Table 1. Each of the 3 batches met all established QC criteria.

Table 1.

Release and Stability Test Data for Three Qualification Batches of [¹⁸F]DK222 Injection Manufactured at the JHU PET Center

Test	Specification	[18F]DK222 Validation 1	[18F]DK222 Validation 2	[18F]DK222 Validation 3
Initial Appearance	Clear, colorless solution, no visible particulate matter	Conforms	Conforms	Conforms
Appearance 240 minutes after EOS	Clear, colorless solution, no visible particulate matter	Conforms	Conforms	Conforms
Initial Radiochemical Purity, % (t = 0 minutes)	≥ 95%	100%	100%	100%
Expiry Radiochemical Purity, % (t =240 minutes)	≥ 95%	100%	98.3%	100%
pH, Initial	3.5 – 5.5	4	4	4
pH, Expiry	3.5 – 5.5	4	4	4
Chemical Purity	No acceptance specification for this item.	DK222: 3.83 μg/mL Precursor: 5.54 μg/mL Others: 0.03 μg/mL	DK222: 2.78 μg/mL Precursor: 5.54 μg/mL Others: 0.02 μg/mL	DK222: 2.01 μg/mL Precursor: 1.97 μg/mL Others: 0.11 μg/mL
Yield	> 20 mCi [18F]DK222 (referenced to assay recorded at end-of-filtration)	64.7 mCi	71.4 mCi	74.9 mCi
Specific Activity	> 250 mCi/μmole of [18F]DK222 (referenced to end of filtration)	911 mCi/μmole	1227 mCi/μmole	1429 mCi/μmole
Identity (HPLC)	HPLC Retention Time is Within ±10% of Reference Standard	1.61%	1.71%	1.81%
Radionuclidic Purity	T1/2 Calc = 105 - 115 min	111.95 min	107.01 min	108.14 min
Bubble-Point	> 13 psi	16 psi	17 psi	17 psi

Quality control test procedures including visual inspection, radiochemical identity, radiochemical purity, specific activity (also known as molar activity) calculation, residual solvent analysis, pH measurements, filter integrity testing, radionuclide identity by half-life measurement, endotoxin analysis and sterility testing have all been described previously⁵ and are not repeated here.

Figure 2 shows typical analytical chromatograms observed during the determination of the radiochemical identity, radiochemical purity, and specific activity. This includes a standard of the authentic non-radioactive product DK222 to establish analytical HPLC system suitability, the final radiotracer product, and a co-injection of DK222 with the radiotracer product.

Figure 2:

Quality control chromatograms of [¹⁸F]DK222.

Results

From the validation runs reported here using the in-house synthesis module, [¹⁸F]DK222 had an isolated average product yield of 70.3 ± 5.2 mCi (2.60 ± 0.19 GBq) (n = 3) starting from 136 ± 19 mCi (5.0 ± 0.7 GBq) of [¹⁸F]fluoride in a synthesis time of 17 minutes, resulting in a non-decay corrected radiochemical yield of 52.2% ± 3.6%. The average effective specific activity (also known as molar activity) was 1189 ± 261 mCi/μmole (44.0 ± 9.7 GBq/μmole).

When the automated GE FASTlab2 synthesis was used, [¹⁸F]DK222 was prepared with an average product yield of 127 ± 74.7 mCi (4.7 ± 2.8 GBq) starting from 435 ± 289 mCi (16.1 ± 10.7 GBq) of [¹⁸F]fluoride in a synthesis time of 23 minutes. The non-decay corrected yield of [¹⁸F]DK222 was 30.0% ± 4.6% with an average effective specific activity of 2,044 ± 1,297 mCi/μmole (75.6 ± 48.0 GBq/μmole) (n=6).

Discussion

Over the past decade, evaluation of PD-L1 levels by immunohistochemistry has emerged as a companion diagnostic for guiding therapeutics targeting immune checkpoint proteins PD-L1 and its receptor programmed death 1 (PD-1). However, inadequacy of those tests in capturing the heterogeneity in PD-L1 levels within and across patients has necessitated the development of non-invasive methods to quantify PD-L1 dynamics.⁶ To meet this need, radiotracers incorporating a wide variety of radioisotopes (⁶⁴Cu, ¹⁸F, ⁶⁸Ga, ⁸⁹Zr, or ^99mTc) have been developed for both PET and SPECT imaging of the PD-L1 levels in vivo. The use of preclinical and clinical PET in this effort has been summarized,⁷ but few of those radiotracers have demonstrated the potential of PD-L1 PET for capturing total PD-L1 levels and their heterogeneity. Radiotracers for this site have been synthesized via various routes: nucleophilic substitution (¹⁸F, often in multistep reactions,^{8 18}F for ¹⁹F exchange⁹), metal chelation (e.g., ⁶⁴Cu, ⁶⁸Ga),^10,11 and most recently using the AlF complex as a “pseudo-metal” for “chelation” with NODA and NOTA.^2,12

The first use of aluminum [¹⁸F]fluoride (Al[¹⁸F]F) was reported by McBride et al.^13,14 Since that date, there have been dozens of publications describing the use of Al[¹⁸F]F-based radioconjugates and much of this previous work was recently summarized by Archibald and Allott¹⁵ in their review of the field. Notable amongst the publications in this field are [¹⁸F]-Alfatide as the first Al[¹⁸F]F-labeled radiotracer that has been used for imaging lung and breast cancer in humans¹⁶, and recent reports with Al[¹⁸F]F-NOTA-octreotide and Al[¹⁸F]F-FAPI-74 for studying an even wider variety of cancers in humans.^17,18 The radioactivity decay of fluorine-18 offers advantages over various radiometals (e.g., gallium-68) and hence used as a better substitute, in theranostic pairing of diagnostic and therapeutic radioisotopes when Al[¹⁸F]F is incorporated.

Reviewing the Al[¹⁸F]F literature, the optimal pH for radiolabeling appeared to be 4.0±0.2. However, Wan et al. reported in their supplemental data that the optimal pH for labeling was between 2 – 4.¹⁶ The reported synthesis methods appear split between one- and two-step syntheses with the one-step method^19,20 in which [¹⁸F]fluoride, AlCl₃ solution, and ligand are mixed and heated together vs. the two-step syntheses^13,21 in which it has been reported that [¹⁸F]fluoride is first added to an AlCl₃ solution and heated, then the ligand is added.

During attempts to scale up previously described [¹⁸F]DK222 synthesis,² decreased product yields were observed with what appeared to be significant radiolysis as the amount of [¹⁸F]fluoride increased (Figure 3). The literature often reports reactions with small amounts of starting radioactivity for an AlF synthesis that may not translate to larger scale cGMP production methods.¹⁶

Figure 3:

Product yield vs. starting [¹⁸F]fluoride examined by analytical HPLC of [¹⁸F]DK222 crude reaction mixture

Examining the literature for ways to inhibit radiolysis, many of the references focus on radiotracer stability in the final radiotracer formulation.¹⁵ Far fewer references look to mitigate radiolysis during the synthesis and purification.^23-25. These references point to inhibition using scavengers such as ascorbic acid, potassium permanganate, sodium ascorbate, and ethanol during the synthesis.

In an attempt to increase the initial starting amount of [¹⁸F]fluoride, reduce the radiolysis and provide an avenue to automate the radiosynthesis of [¹⁸F]DK222, changes to the initial published synthesis using acetate/acetic acid buffer at pH 4.0±0.2 were investigated. Similar radiochemical yields were observed using ascorbic acid/sodium ascorbate 4.0±0.2, 0.5 M acetic acid, and 0.5 M ascorbic acid. However, when using 0.5 M ascorbic acid (pH ~2), noticeably less radiolysis was observed even when larger amounts of starting radioactivity were used. No noticeable differences were seen between a one- or two-step radiosynthesis using the above buffers.

Radiolysis was further reduced by an efficient and rapid purification and reformulation. The SPE module, and later the GE FastLab2, steps were optimized to reduce the amount of time that [¹⁸F]DK222 remained on the SPE cartridge. Figure 4 shows an example of radiolysis from [¹⁸F]DK222 left on the SPE for an extended period (5 minutes) after flushing the SPE with air.

Figure 4:

Analytical HPLC of decomposing [¹⁸F]DK222

Table 1 presents the data from 3 consecutive validation runs for this radiotracer. It is noted that the specific activities reported are “effective specific activities” since it is not possible to isolate the final radiotracer separate from the DK-NODA precursor. These data have been deemed acceptable to Federal and institutional regulatory authorities to support human investigational use.

Conclusion

In summary, a cGMP protocol for the radiosynthesis of [¹⁸F]DK222, a radioligand for PD-L1, was developed. The synthesis was reproducible from both an in-house module and a commercial radiosynthesizer (GE FASTlab 2). The specific activity, chemical and radiochemical purity of the final radiotracer product are all considered acceptable for use in future human PET studies.