Preclinical evaluation of FAP-2286 for fibroblast activation protein targeted radionuclide imaging and therapy

Formalin-fixed paraffin-embedded (FFPE) tissue microarrays (US Biomax) and whole sections (Indivumed, Tissue Solutions, and Discovery Life Sciences) were stained with FAP antibody (SP325, Abcam; 1:50) according to the manufacturer recommendation. Staining was performed on the Bond Rx Autostainer (Leica Biosystems) and detection by the Bond Polymer Refine Detection (Leica Biosystems) according to manufacturer’s protocol. Whole slide scanning was performed on the Aperio AT2 (Leica Biosystems), and analyses were conducted using Aperio ImageScope (Leica Biosystems), Visiopharm Integrator System (Visiopharm), or HALO (Indica Labs) image analysis platform.

In vitro FAP autoradiography was performed as previously described with minor modifications (32). The 20-µm-thick frozen tissue sections were incubated with ¹¹¹In-labeled FAP-2286 (15 MBq/nmol). Subsequently, an x-ray film (Biomax MR films; Carestream Health) was placed onto positioned slides and exposed for 2 days. Relative optical film density was determined using MCID™ software (InterFocus) and correlated with known amounts of radioactivity via a separately recorded calibration curve.

HEK-FAP cell line was generated using the vector pTAR to insert FAP complementary DNA by recombinase-mediated cassette exchange into the pre-tagged HEK293 cells (InSCREENeX). For in vivo HEK-FAP studies, female nude or SCID beige mice (Charles River Laboratories) were subcutaneously implanted with 5 × 10⁶ HEK-FAP cells in 150 µL serum-free media:Matrigel (1:1) (Corning). For the patient-derived xenograft (PDX) efficacy study, female NMRI nu/nu mice (Janvier Labs) were subcutaneously implanted with Sarc4809 fragments (Experimental Pharmacology & Oncology Berlin-Buch). All animal experiments were carried out under licenses approved by the respective authorities and in compliance with guidelines on animal welfare.

Thirty megabecquerels (30 MBq/nmol) ¹¹¹In-FAP-2286 (n = 9) or ¹¹¹In-FAPI-04 (n = 6) was injected intravenously (IV) into HEK-FAP tumor-bearing mice and the distribution of the radiotracers was assessed using the NanoSPECT/computed tomography (CT) system (Mediso). For PET imaging, 10 MBq of ⁶⁸Ga-FAP-2286 or ⁶⁸Ga-FAPI-46 (10 MBq/nmol) was given IV into HEK-FAP tumor-bearing mice (n = 3) followed by PET/CT analysis using the nanoScan PET/CT system (Mediso). For SPECT imaging of ¹⁷⁷Lu-labeled FAP-2286 or ¹⁷⁷Lu-labeled FAPI-46, 30 MBq was administered IV (30 MBq/nmol; n = 3) and the distribution of the radiotracers was assessed using the nanoScan SPECT/CT system (Mediso).

A single dose of the indicated treatments was administered IV on day 0, with vehicle (saline), nonradioactive ^natLu-FAP-2286, 30 or 60 MBq ¹⁷⁷Lu-FAP-2286 (30 MBq/nmol and 60 MBq/nmol, respectively), or 30 MBq ¹⁷⁷Lu-FAPI-46 (30 MBq/nmol). Tumor length and width were measured by caliper, and tumor volume was calculated as 0.5 × (length × width²). Tumor growth and body weight were monitored 3 times weekly. Sarc4809 tumor growth was normalized to the mean tumor volume (MTV) at the time of injection for each group due to variability in tumor volume between the groups. Statistical analysis was performed using GraphPad Prism 8.4.

FAP-2286 was labeled with Alexa Fluor 488 (AF488) at the C-terminus of the peptide. For FAPI-46, a free carboxy moiety of the DOTA was used for labeling using AF488 Cadaverine (Thermo Fisher Scientific).

Gene expression analysis across multiple cancer types using The Cancer Genome Atlas dataset (34) showed elevated FAP mRNA levels in different tumor types with the highest median observed in breast (1014 RNA-Seq by Expectation Maximization [RSEM], n = 1082) and pancreatic tumors (1213 RSEM, n = 177; Supplementary Fig. S1). FAP levels were validated by immunohistochemistry in human tumor samples on tissue microarrays (TMA) of 19 cancers (n = 1829 cores). Cell-surface staining patterns of varying intensity were detected in several indications with the highest median levels in pancreatic cancer (H-score = 63; n = 91) and the second highest in breast cancer (H-score = 33; n = 265), cholangiocarcinoma (H-score = 33; n = 83), and esophageal cancer (H-score = 33; n = 69; Fig. 1A).

For a more comprehensive examination, FAP immunohistochemistry was performed on whole FFPE tissue sections of 16 tumor types that included primary and metastatic samples of both low- and high-grade histology (n = 359, Supplementary Fig. S2A). High FAP expression (H-score cut-off of ≥ 30) was observed in multiple tumor types. More than 40% of pancreatic, cancer of unknown primary (CUP), salivary gland and mesothelioma tumors highly express FAP. High FAP expression was detected in both primary and metastatic samples (Supplementary Fig. S2B) and was independent of tumor grade (Supplementary Fig. S2C).

In addition, FAP levels in the tumor versus the stroma compartments were examined using automated imaging in a subset of cases with very high FAP staining across cancer types. This quantitative analysis confirmed that in most epithelial cancers, FAP expression was primarily confined to the CAFs in stroma (Fig. 1B). However, tumor cell staining was common in sarcoma and mesothelioma, and was also occasionally observed in epithelial tumors including HNSCC, CUP, esophageal and others (Fig. 1C).

FAP-2286 is a FAP-binding cyclic peptide consisting of seven amino acids, wherein two cysteine residues are cyclized by an aromatic moiety linked to a DOTA chelator (Fig. 2A). The chelator can be labeled with radionuclides for imaging and therapeutic applications. The affinity and selectivity of FAP-2286 along with its complexes of natural nonradioactive metals (^natGa-FAP-2286, ^natLu-FAP-2286, and ^natIn-FAP-2286) as surrogates of the three radiotracers were evaluated in vitro.

The binding of FAP-2286 and its natural metal complexes to recombinant human and mouse FAP was assessed by SPR (Fig. 2B). FAP-2286 demonstrated potent binding to human and mouse FAP, with mean equilibrium dissociation constant (K_D) values of 1.1 and 4.7 nM, respectively. The high affinity was maintained for the metal complexes with K_D values ranging from 0.2 to 1.4 nM for human FAP and from 1.9 to 7.7 nM for mouse FAP. These results are consistent with the 89% amino acid sequence identity between the 2 species (5).

The binding of FAP-2286 to cell surface FAP was evaluated in a competition assay against a fluorophore-labeled competitor peptide using the human WI-38 fibroblast-like fetal lung cell line that endogenously expresses FAP. FAP-2286 binding to FAP markedly reduced fluorophore-labeled competitor peptide bound to cells with a mean IC₅₀ of 2.7 nM, as measured by flow cytometry. The three metal complexes were comparably potent in inhibiting binding of the competitor peptide, with mean IC₅₀ values ranging from 1.7 to 3.4 nM (Fig. 2B).

To further characterize binding in vitro, the potency of FAP-2286 was measured against FAP endopeptidase enzymatic activity on a fluorophore-labeled substrate. FAP-2286 inhibited FAP enzymatic activity with mean IC₅₀ values of 3.2 and 22.1 nM against human and mouse FAP, respectively. Similarly, the mean IC₅₀ values for the metal complexes ranged from 1.3 to 2.2 nM against human FAP and from 8.4 to 16.3 nM against mouse FAP (Fig. 2B).

Selectivity of FAP-2286 binding against the related family members DPP4 and PREP (4) was evaluated in enzymatic inhibition assays employing recombinant human proteins. FAP-2286 and its metal complexes showed minimal activity against the DPP4 and PREP proteases with IC₅₀ values of  >1 and 10 μM for human recombinant PREP and DPP4, respectively, confirming lack of binding to related family homologs (Supplementary Table S1).

Confirmation of FAP-2286 peptide stability in plasma was also performed to assess potential degradation or metabolism of the radiotracers in the in vivo studies. Plasma stability assays revealed that FAP-2286 was stable for 24 h at 37 °C with > 95% and 85% retained in human and mouse plasma, respectively (Supplementary Fig. S3).

Prior to in vivo evaluation of FAP-2286, the HEK-FAP xenograft was assessed for FAP expression by immunohistochemistry and FAP-2286 binding by in vitro autoradiography. The HEK-FAP tumor displayed high, homogeneous FAP levels, with a strong ¹¹¹In-FAP-2286 binding signal by autoradiography (approximately 22,000 counts per minute [CPM]), which was corroborated by the immunohistochemistry H-score of 225. As a negative control, normal human kidney was utilized, and non-specific background levels were detected (Fig. 3A). In addition, a comparison between the 2 techniques was performed using a set of cholangiocarcinoma and sarcoma whole tissue sections that were selected for varying FAP H-scores (Fig. 3B). A direct correlation was observed in these human specimens with the two methods (Pearson r = 0.79, P = 0.002, n = 12).

The biodistribution of ¹¹¹In-FAP-2286 was evaluated in vivo by SPECT imaging following a single dose of 30 MBq in HEK-FAP tumor bearing mice. The radiotracer was rapidly enriched within the HEK-FAP xenografts with low off-target accumulation and predominantly renal elimination (Fig. 4A). High tumor-to-background signal ratio was observed from 1 h post injection (p.i.) onward. At 1 h p.i. of ¹¹¹In-FAP-2286, tumor uptake was 11.1 percent injected dose per gram (%ID/g). Accumulation of ¹¹¹In-FAP-2286 was stably maintained in the tumors with 9.1%ID/g at 48 h p.i. (Fig. 4B, Supplementary Table S2). Normal tissues including liver, salivary gland, and shoulder showed very limited uptake. The organ with the highest non-target uptake was the kidney; however, an increasing tumor-to-kidney (T/K) ratio was observed over time with the highest differential uptake of 7.5 T/K obtained at 48 h p.i. (Supplementary Table S2).

Based on the SPECT imaging observations, ¹⁷⁷Lu-FAP-2286 antitumor efficacy was evaluated as a single dose of 30 or 60 MBq in HEK-FAP tumor bearing mice. Tumors in the control groups injected with vehicle or the natural nonradioactive metal-labeled ^natLu-FAP-2286 reached a mean tumor volume (MTV) of 1338 and 1392 mm³, respectively, on day 14 p.i., when < 50% of the animals in the vehicle group were sacrificed due to tumor volumes reaching humane endpoint criteria. In contrast, the MTV for both ¹⁷⁷Lu-FAP-2286 treatment groups were reduced to ≤ 37 mm³ on day 14 p.i. (Fig. 5A). Statistically significant antitumor activity was observed with tumor growth inhibition (TGI) of 111% and 113% (P < 0.05) in mice treated with 30 or 60 MBq ¹⁷⁷Lu-FAP-2286, respectively, relative to the vehicle-treated group (Fig. 5C).

HEK-FAP xenografts were monitored for regrowth out to the end of the study (day 42), with 3 and 9 tumor-free (< 10 mm³) mice out of 10 in the 30 and 60 MBq ¹⁷⁷Lu-FAP-2286 treatment groups, respectively. Furthermore, there was a statistically significant difference in the MTV on day 42 in animals treated with the low dose (365 mm³) versus the high dose (43 mm³) of ¹⁷⁷Lu-FAP-2286, suggesting a dose response relationship (P < 0.01). The median survival time (MST) for the two ¹⁷⁷Lu-FAP-2286 groups was not reached while the MST for the control groups was 17.5 days. The maximum body weight loss ranged from 3 to 5% on day 2 for all groups, with no ¹⁷⁷Lu-FAP-2286–related weight loss observed (Fig. 5B).

The efficacy of ¹⁷⁷Lu-FAP-2286 was evaluated in the Sarc4809 PDX model of sarcoma. Analogous to the HEK-FAP model, immunohistochemistry and autoradiography demonstrated high FAP expression in the Sarc4809 xenografts with a H-score of 202 and bound radioactivity of 16,000 CPM, respectively (Fig. 3).

A single dose of vehicle, ^natLu-FAP-2286, or ¹⁷⁷Lu-FAP-2286 at 30 or 60 MBq was administered to Sarc4809 tumor bearing mice. Both dose levels of ¹⁷⁷Lu-FAP-2286 resulted in significant antitumor activity (Fig. 6A). Because of starting tumor volume variability on day 0, all tumor volumes were normalized to the starting volumes to obtain relative tumor volume (RTV). On day 19 p.i., a single dose of 30 or 60 MBq ¹⁷⁷Lu-FAP-2286 demonstrated significant tumor growth inhibition as the treatment-to-control ratio was 45% for both doses (P < 0.01). By day 42, there was a statistically significant difference in the RTV in animals treated with the low versus high dose of ¹⁷⁷Lu-FAP-2286 (P = 0.018). The MST of vehicle and ^natLu-FAP-2286 control groups was 28.5 and 26.0 days, respectively, but was not reached for ¹⁷⁷Lu-FAP-2286 treatment groups after 42 days of monitoring (Fig. 6C). No ¹⁷⁷Lu-FAP-2286–related body weight loss was observed during the study (Fig. 6B).

We compared the biodistribution of FAP-2286 against FAPI-46, the current lead therapeutic compound from the FAPI radiotracer series (28). ⁶⁸Ga-FAP-2286 and ⁶⁸Ga-FAPI-46 were rapidly enriched within the HEK-FAP xenografts with predominantly renal elimination (Fig. 7A, Supplementary Table S3). High tumor specificity was observed at all 3 timepoints assessed (0.5, 1, and 3 h p.i.). At 0.5 h after administration of ⁶⁸Ga-FAP-2286 and ⁶⁸Ga-FAPI-46, tumor uptake was 9.8 and 9.3 %ID/g, respectively. Accumulation of ⁶⁸Ga-FAP-2286 and ⁶⁸Ga-FAPI-46 was maintained at 3 h after injection with 10.8 and 9.2 %ID/g, respectively, and no significant difference in tumor distribution was observed between both compounds (Fig. 7C).

The ¹⁷⁷Lu-radiolabeled FAP-2286 and FAPI-46 also demonstrated rapid accumulation within the HEK-FAP xenografts with low background signal at 3 h after dosing (Fig. 7B, Supplementary Table S3). However, significantly higher tumor retention of ¹⁷⁷Lu-FAP-2286 compared to ¹⁷⁷Lu-FAPI-46 was observed at 24 and 72 h p.i. (P < 0.001). Tumor uptake of ¹⁷⁷Lu-FAP-2286 at the 3-h timepoint was 21.1 %ID/g and demonstrated durable retention with 16.4 %ID/g 72 h later. The time-integrated activity coefficient (TIAC) and absorbed dose were calculated to be 11.2 MBq^*h/MBq and 2.8 Gy/MBq, respectively. In contrast, ¹⁷⁷Lu-FAPI-46 accumulation decreased from 15.3 %ID/g at the 3-h timepoint to 1.6 %ID/g after 72 h resulting in a lower TIAC of 0.9 MBq*h/MBq and absorbed dose of 0.3 Gy/MBq (P < 0.001, Fig. 7D, Supplementary Table S3 and Fig. S4). Similar to the gallium-labeled compounds, normal tissues showed limited uptake and rapid clearance of the radiotracers with the highest level in the kidneys. A higher kidney accumulation was observed for ¹⁷⁷Lu-FAP-2286 compared to ¹⁷⁷Lu-FAPI-46 with the greatest difference at 3 h after injection with 2.2 %ID/g and 0.7 %ID/g, respectively, which decreased to 0.6 %ID/g and 0.2 %ID/g, respectively, at 72 h p.i. (Supplementary Table S3).

The efficacy of ¹⁷⁷Lu-FAP-2286 and ¹⁷⁷Lu-FAPI-46 was evaluated in the HEK-FAP model as a single 30 MBq dose administration. Tumors in the vehicle control group reached an MTV of 952 mm³ on day 9 p.i., the last timepoint when < 10% of the animals in the vehicle group had been sacrificed due to tumor volumes reaching humane endpoint criteria. In contrast, the MTV for ¹⁷⁷Lu-FAP-2286 and ¹⁷⁷Lu-FAPI-46 treatment groups were reduced to 107 and 245 mm³, respectively, on day 9 p.i. (Fig. 8A). Statistically significant antitumor activity was observed with TGI of 108% and 90% (P < 0.001) in mice treated with ¹⁷⁷Lu-FAP-2286 and ¹⁷⁷Lu-FAPI-46, respectively, relative to the vehicle control group.

HEK-FAP xenografts were monitored for regrowth out to the end of the study (day 41), when 4 mice out of 10 in the ¹⁷⁷Lu-FAP-2286 treatment group had no measurable tumors (< 10 mm³), while all mice treated with ¹⁷⁷Lu-FAPI-46 had tumors. In addition, there was a statistically significant difference in the MTV in animals treated with ¹⁷⁷Lu-FAP-2286 (12 mm³, n = 10) versus ¹⁷⁷Lu-FAPI-46 (1210 mm³, n = 9, P < 0.0001) on day 23 when < 10% of the animals in the treatment groups were sacrificed due to tumor volume (Fig. 8C). The MST for the ¹⁷⁷Lu-FAP-2286 group was not reached while the MST for the vehicle control and ¹⁷⁷Lu-FAPI-46 groups were 16.5 and 27.5 days, respectively (Fig. 8B).

In order to gain more insight into the different tumor retention time of both molecules, AF488-labeled FAP-2286 and FAPI-46 were synthesized for studying cellular internalization, localization, and retention by live cell fluorescence microscopy. AF488 was conjugated to a free carboxy moiety of the DOTA for FAPI-46, and attached to the C-terminus of FAP-2286 since the chelator and metal complex may contribute to FAP-2286 binding affinity (Supplementary Methods).

Both compounds bound in a dose-dependent, saturable manner to HEK-FAP cells with half-maximal effective concentration (EC₅₀) values of 4.9 nM for AF488-FAP-2286 and 1.7 nM for AF488-FAPI-46 (Fig. 9A). The binding of AF488-labeled compounds was completely blocked by increasing concentrations of unlabeled FAP-2286 or FAPI-46, suggesting FAP-specific binding and an overlapping binding site on FAP (Supplementary Fig. S5).

AF488-labeled FAP-2286 and FAPI-46 retention on FAP-expressing HEK293 cells was visualized by fluorescence microscopy. After 1-h incubation with 5 nM of AF488-FAP-2286 or AF488-FAPI-46 at 37 °C, compounds bound to the cell surface at comparable intensity indicated by a green staining of the cell membrane, which was blocked by competition with unlabeled ^natLu-FAP-2286 or FAPI-46, respectively (Fig. 9B). Stained HEK-FAP cells were further incubated for additional 1, 3, 8, 24, and 72 h at 37 °C in compound-free medium (Fig. 9C). Intracellular structures with endosomal appearance were visible after 1 h of incubation indicated by green fluorescent puncta. From 3 h onward, the number of orange and yellow fluorescent endosomal structures increased indicating internalization into late endosomes or lysosomes and colocalization with LysoTracker. After 8 h of incubation, FAPI-46 had greater reduction in cell surface fluorescence and less intracellular fluorescence than FAP-2286. By 72 h, FAPI-46 was not detectable while FAP-2286 retained intracellular signal with little to no cell surface fluorescence.