Preclinical Evaluation of a Radiotheranostic Single-Domain Antibody Against Fibroblast Activation Protein a

,

Tar geted radionuclide therapy (TRT) can be a relevant strategy for systemic malignancies.In TRT, radiolabeled compounds are systemically administered with an aim to deliver cytotoxic radiation to diseased sites (1).In doing so, targeting moieties are coupled to a-emitting, b-emitting, or Auger electron-emitting radionuclides.Different types of moieties have been used, of which monoclonal antibodies have been explored abundantly (2,3), but their limited tissue penetration and off-target effects have hindered progress.With the introduction of monoclonal antibody fragments, peptides, and small molecules, the field of TRT has undergone a revitalization, highlighted by recent success stories with [ 177 Lu]Lu-DOTATATE and [ 177 Lu]Lu-PSMA-617 (4,5).An interesting class of monoclonal antibody fragments comprises single-domain antibodies (sdAbs).High binding affinity and specificity, fast blood clearance, and the ability to homogeneously distribute in target tissues make sdAbs promising targeting moieties for diagnostic imaging and TRT (6,7).
Fibroblast activation protein a (FAP) is overexpressed on cancer-associated fibroblasts in various carcinomas, contributing to tumor growth, metastasis, and immunosuppression (8).Breast, colon, and pancreatic tumors often show strong desmoplastic reactions, which result in a dominant presence of stromal cells.In particular cases, cancer cells express FAP as well, such as in sarcomas, melanomas, or glioblastomas.FAP targeting is gaining attention for molecular imaging and TRT.Quinoline-based FAP inhibitors (FAPIs) have shown potential in cancer imaging, but their prognostic value remains uncertain.Some FAPI molecules have been labeled with therapeutic radionuclides and evaluated in preclinical and clinical studies, although clinical evidence is limited and shows mixed responses to anti-FAP TRT.Challenges include small patient cohorts, heterogeneity, and optimizing FAPI tumor retention (9).
Here we describe the generation, selection, and characterization of a lead sdAb with excellent binding characteristics for FAP.We assessed its PET and SPECT imaging potential, as well as its use for TRT.

MATERIALS AND METHODS
The methodologies for sdAb generation (10,11), lead compound selection and biochemical characterization, cell culture conditions, radiochemical procedures, radioligand binding studies, and dosimetry calculations can be found in the supplemental materials (supplemental materials are available at http://jnm.snmjournals.org).

Animal Models
Female CRL:n-FoxN1nu mice (Charles River) received subcutaneous neck injections of 2 3 10 6 U87-MG cells.Tumors were allowed to grow to 100-250 mm 3 for biodistribution and therapy studies.We monitored animal welfare by measuring weights and tumor volumes, applying a humane endpoint of body weight loss of more than 20% or tumor volume of more than 1,500 mm 3 .Tumor dimensions were calculated using a digital caliper.Doseescalation studies were performed on female C57BL/6 mice (Charles River).Ethical approval was granted by Vrije Universiteit Brussel's ethical committee.

Biodistribution of Radiolabeled sdAbs
CRL:n-FoxN1nu mice with human FAP (hFAP)-positive U87-MG tumors sized 100-250 mm 3 received intravenous injections of 99m Tc-and 68 Ga-labeled sdAbs.Imaging was conducted 1 h after tracer administration using small-animal SPECT/CT Vector1/CT (MILabs) or small-animal PET/CT b/X-CUBE (Molecubes) systems.[ 131 I]I-GMIB-R3B23 and [ 131 I]I-GMIB-4AH29 were imaged at various time points using the Vector1/CT system.After imaging, mice were euthanized, and organs and tissues were weighed.Time-dependent biodistributions of [ 131 I]I-GMIB-4AH29 and [ 225 Ac]Ac-DOTA-4AH29 were determined through dissections up to 96 h after injection, and dosimetry calculations were performed as described previously (13,14).[ 131 I]I-GMIB-4AH29 and [ 225 Ac]Ac-DOTA-4AH29 were administered to healthy mice in a dose-escalation study, with a 6-mo follow-up period.C57BL/6 mice (n 5 5) received intravenous injections 6 times over 3 wk, with doses ranging from 37 to 4.63 MBq for [ 131 I]I-GMIB-4AH29 and from 40 to 2.5 kBq for [ 225 Ac]Ac-DOTA-4AH29 (each , 30 mg).Control groups received only vehicle solution.Mice were euthanized on reaching specific endpoints: more than 20% weight loss, body condition scoring, or significant changes in behavior.Organs and tissues were collected, fixed, and examined for signs of toxicity via hematoxylin-and-eosin staining and light microscopy.

Statistical Analysis
Kaplan-Meier curves that show the time to reach humane endpoints were statistically analyzed using the log-rank (Mantel-Cox) test.Student t testing or 1-way ANOVA were performed for all other results.

Generation and In Vitro and In Vivo Screening of Anti-FAP sdAbs
As described in the supplemental materials, we developed 4 anti-FAP sdAbs, of which 4AH29 is the lead for further clinical development.In an enzyme-linked immunosorbent assay, 4AH29 binds both mouse FAP (mFAP) and hFAP, 4HB139 is cross-reactive but shows weaker binding to mFAP, 2MR158 is an hFAP binder, and 2MS56 is an mFAP binder.A homology plot of the FAP sequence of different species is shown in Supplemental Figure 1.All hFAP binders equally recognize cynomolgus FAP protein.In these assays, no sdAb binds human dipeptidyl peptidase-4, the closest homolog of FAP in humans, showing the specificity of the binding (Supplemental Fig. 1A).Biolayer interferometry measurements confirmed these binding profiles, with the lead compound, 4AH29, having an affinity of 38 pM for hFAP protein (Table 1; Supplemental Fig. 3).Next, we investigated whether the sdAbs influenced enzymatic activity of the FAP protein.In contrast to FAPI-4 and talabostat, which bind to the catalytic pocket of the enzyme and inhibit it, none of the 4 sdAbs influenced the rate of conversion of a fluorogenic substrate (Table 2; Supplemental Fig. 2D).The 4 sdAbs bound to GM05389 human fibroblasts, mFAP-or hFAP-transfected HEK293 cells, and U87-MG cells, all cells that express either mFAP or hFAP, as measured by flow cytometry (Supplemental Figs.2C, 4, and 5).4AH29, 4HB139, 2MR158, and 2MS56 were successfully radiolabeled with 99m Tc (Supplemental Table 1).The half-maximal effective concentration (EC 50 ) values of [ 99m Tc]Tc-4AH29, [ 99m Tc]Tc-4HB139, [ 99m Tc]Tc-2MR158, and [ 99m Tc]Tc-2MS56 were obtained on the same cell lines as described earlier (Supplemental Figs.2B and 6) and were in line with results obtained through flow cytometry.In vivo, the 99m Tc sdAbs revealed a typical pattern of fast-clearing radiolabeled small proteins, with low levels of radioactivity in all organs and tissues, including blood, and high radioactivity levels in the kidneys (Figs.1A and 1C; Supplemental Table 2).As expected, the 99m Tc-labeled hFAP-binding sdAbs 4AH29, 4HB139, and 2MR158 showed significantly higher uptake in hFAP-expressing U87-MG tumors than did [ 99m Tc]Tc-2MS56, which binds only mFAP (P , 0.0001).Levels in blood and vascularized tissues were slightly elevated for [ 99m Tc]Tc-4AH29, [ 99m Tc]Tc-2MS56, and [ 99m Tc]Tc-4HB139, which bind the mFAP protein.This elevated uptake could potentially be attributed to the presence of circulating FAP protein in mice (but not in humans), as suggested by Keane et al. (15).[ 99m Tc]Tc-2MS56 exhibited higher liver uptake than that of 4HB139, 4AH29, and 2MR158.This might result from its lower melting temperature of 61.5 C, in contrast to the melting temperature of more than 65 C of the others.This could result in more aggregate formation during radiolabeling at 50 C, potentially causing formation of aggregates that are trapped in the liver (12).

Therapeutic Efficacy
The therapeutic potential of [ 131 I]I-GMIB-4AH29 and [ 225 Ac]Ac-DOTA-4AH29 was evaluated in U87-MG-bearing mice (Fig. 4A).Tumor growth was delayed in all mice receiving aor b-TRT (Fig. 4B; Supplemental Figs.10A and 11).No relevant weight loss was observed in the treatment groups (Supplemental Fig. 10B).Consequently, treated mice lived longer than those in the control group (P , 0.05; Fig. 4C).No difference in survival was observed between cohorts receiving a high and those receiving a low cumulative radioactive dose of [ 131 I]I-GMIB-4AH29, with mean survival of 52 and 56 d, respectively.One animal of the high-dose group was tumor-free.Median survival was 63 d for a high-dose treatment with [ 225 Ac]Ac-DOTA-4AH29 and 42 d for a low-dose treatment (P 5 0.0001).

DISCUSSION
FAP is highly expressed on cancerassociated fibroblasts of carcinomas.Breast, colon, and pancreatic tumors often show strong desmoplastic reactions, which result in a dominant presence of stromal cells.FAP has low expression or is undetectable in most normal adult tissues (8).Various FAPIs have been used as tools to image FAP expression.Some have also been labeled with therapeutic radionuclides, and although significant efforts were made to optimize the tumor half-life of FAPI compounds, it seems that limited tumor retention still poses a challenge (9).In this study, we identified the lead FAP-targeting sdAb, 4AH29, which has subnanomolar affinity for an mFAP or hFAP conserved epitope and recognizes both purified and membraneanchored FAP protein.Other relevant FAP-targeting moieties, such as FAPI-04 ( 16), FAPI-46 (17), and FAP-2286 ( 18), have been extensively described and underline the relevance of FAP as a promising diagnostic imaging modality.The herein described sdAb 4AH29 is different from the reported FAPIs, because it targets FAP away from the catalytic site.This characteristic is important because FAP enzymatic activity is believed to be involved in normal physiologic processes (19).By targeting FAP without affecting its enzymatic activity, this sdAb-based radiopharmaceutical offers a promising avenue for therapeutic interventions without interfering with inherent biologic processes.Moreover, we report here a binding affinity of 38 pM for FAP compared with a nanomolar affinity for most FAPIs, as reported in the literature.We have shown successful specific detection of FAP-positive tumors in mice using [ 68 Ga]Ga-DOTA-4AH29 and tracer-dose [ 131 I]I-GMIB-4AH29.Higher and specific accumulation in FAP-positive tumors was measured, and retention in other tissues was lower than in radiolabeled control sdAb.In the case of tracer-dose [ 131 I]I-GMIB-4AH29, repeated small-animal SPECT/CT underlined its optimal binding characteristics, with radioactivity cleared quickly from circulation and kidneys, whereas retention in the FAP-positive tumor was maintained.This result is in line with what has been reported previously for other 131 I-labeled sdAbs, such as the HER2targeting sdAb 2Rs15d (14).Likewise, small-animal PET/CT imaging with [ 68 Ga]Ga-DOTA-4AH29 showed fast (after 1 h) and high-contrast imaging of FAP-positive tumors.The obtained results are in line with what has been described previously for 68 Galabeled sdAbs such as HER2-targeting 2Rs15d (20) and a CD206-targeting sdAb (21), both of which are in clinical evaluation (NCT03331601 and NCT03924466, respectively).
For therapeutic purposes, we selected b-emitting 131 I and a-emitting 225 Ac.The radiohalogen 131 I was chosen for the excellent biodistribution that is obtained for sdAbs when radiolabeled with 131 I using the prosthetic group SGMIB (14).The radiometal 225 Ac was chosen for different factors, such as better availability and more straightforward chemistry than with the use of 211 At.In theory, the latter is more suited to comparison with 131 I. Similar to 131 I, the half-life of 225 Ac allows centralized manufacturing.However, when used in tandem with fast-clearing moieties, extensive kidney accumulation is an important risk (13).Systemic administration of [ 131 I]I-GMIB-4AH29 revealed high and sustained tumor accumulation in hFAP-positive U87-MG-xenografted mice.In this study, we observed varying absolute tumor targeting among different experiments and compounds.For instance, absolute tumor uptake after 1 h measured about 15 %IA/g in the long-term biodistribution of [ 131 I]I-GMIB-4AH29 (Supplemental Table 5), whereas the ex vivo dissections after small-animal SPECT/CT imaging of the same compound revealed tumor uptake of approximately 5 %IA/g (Supplemental Table 3).We and others have observed that the level of FAP expression is dynamic in this naturally FAPexpressing U87-MG cell line because of the variable influence of tumor microenvironmental factors (22,23).Hence, comparison of uptake values between different experiments should be done with care.In this study, we assessed the time-dependent biodistribution of [ 131 I]I-GMIB-4AH29 and [ 225 Ac]Ac-DOTA-4AH29 head to head, because only this allows a true comparison.Both drug products accumulated rapidly and specifically in tumors.Unbound [ 131 I]I-GMIB-4AH29 was rapidly cleared from kidneys, with less than 1 %IA/g after 24 h, whereas it was slower for [ 225 Ac]Ac-DOTA-4AH29, with roughly 10 %IA/g still in the kidneys after 24 h.This difference in kidney clearance was expected and has been described previously (13,14).
Subsequent assessment of long-term toxicity after repeated administration of therapeutic  over 3 wk was associated with dosedependent histopathologic changes in the kidneys after 6 mo.Renal changes were considered adverse and started at cumulative radioactivity doses of 30 kBq.Mice receiving a cumulative dose of 30 kBq did reach the end of the study, after 6 mo, without relevant weight loss.In the case of [ 131 I]I-GMIB-4AH29, mice in all treatment groups lived until the end of the study without the occurrence of weight loss.Histopathologic analysis revealed a limited number of cellular changes in the kidneys for the highest cumulative doses of [ 131 I]I-GMIB-4AH29, whereas the damage was more striking for [ 225 Ac]Ac-DOTA-4AH29.This observation underlines the impact of the radiolabel and linking chemistry on the overall behavior of the radiopharmaceutical.
A few strategies to reduce the retention of cytotoxic radiation in the kidneys after peptide-or small-protein-based TRT have been explored.SdAbs are excreted via the kidneys' glomerulus with moderate retention, making kidneys the dose-limiting organ of sdAb-mediated TRT.A pretargeting approach could circumvent the limitations imposed by kidneys for sdAb-mediated TRT, as was demonstrated successfully for a 177 Lu-labeled Affibody molecule (Affibody AB) (24).Other possible avenues are modulation of the sdAb protein sequence, cleavable linkers, and coupling methods to avoid a high radiation burden on the kidneys (25).
Repeated administration of [ 225 Ac]Ac-DOTA-4AH29 and [ 131 I]I-GMIB-4AH29 in mice bearing hFAP-positive U87-MG xenografts translated into dose-dependent therapeutic effects, with inhibition of tumor development and extended survival for treated mice versus controls.For both radiopharmaceuticals, we aimed to include 1 treatment arm that gave rise to an absorbed dose in the kidneys of roughly 23 Gy, which is the generally accepted threshold for kidney toxicity.However, any reference to this limit should be made with care, because it originates from external-beam irradiation and assumes homogeneous distribution, which is not the case for TRT or for a-particles emitted from 225 Ac (13).In addition, when considering a-particle radiation, we should consider the relative biological effectiveness to enable comparability of doses from different radiation types.A relative biological effectiveness value of 5 is often used for studies with 225 Ac (26,27), which always results in higher relative biological effectiveness-weighted absorbed doses in tissues.The second radioactive dose level was selected because of the results from the repeated-dose toxicology in  mice.For [ 131 I]I-GMIB-4AH29, cumulative radioactive doses of 222 and 111 MBq were selected, which correspond to absorbed doses in the kidneys of 50 and 25 Gy, respectively.For [ 225 Ac]Ac-DOTA-4AH29, cumulative radioactive doses of 120 and 30 kBq were selected, leading to kidney-absorbed doses of 14 and 3.5 Gy, respectively.Here, 30 kBq was selected because of the 100% overall survival in the repeated-dose toxicology.These dosimetry estimations assume that the cumulative radioactive dose is delivered in a single administration, which is not the case in this study.A cumulative radioactive dose of 120 kBq was selected to understand the extent of an antitumor effect while considering the likelihood of inducing late-stage toxicity in the kidneys.Because of the short range of a-particles, it remains difficult to determine correctly the maximum tolerable dose and toxicities related to a-particles in small animals, which makes it even more difficult to translate the maximum tolerable dose into patients.Hence, future dose-escalation studies in larger species will be conducted, providing better understanding because the sizeable organs form a more realistic setting to assess the maximum tolerable dose (28).High-dose [ 225 Ac]Ac-DOTA-4AH29 performed best, as reflected in the most pronounced tumor inhibition and extended survival, whereas the low-dose treatment gave rise to limited therapeutic signals.This is an important observation, because the low-dose regimen (and beyond) gave rise to adverse renal changes in the repeated-dose toxicology study.Both high-and low-dose [ 131 I]I-GMIB-4AH29 performed well, with relevant impact on both tumor growth inhibition and improved survival of tumor-bearing mice and limited cellular changes in the kidneys, as determined by histopathology.

CONCLUSION
This pioneering study explores the radiotheranostic potential of anti-FAP sdAbs, highlighting sdAb 4AH29's promise in TRT.
[ 68 Ga]Ga-DOTA-4AH29 and [ 131 I]I-GMIB-4AH29 enable precise FAP-positive tumor detection.Therapeutic [ 225 Ac]Ac-DOTA-4AH29 and [ 131 I]I-GMIB-4AH29 exhibit potent and sustained tumor targeting, leading to dose-dependent therapeutic effects in FAP-positive tumor-bearing mice.However, further research is necessary to understand adverse events at high therapeutic doses and to align radiation type with treatment indications.

TABLE 1
Kinetic Binding Analysis of sdAbs to Biotinylated hFAP or mFAP Protein association rate constant; k off 5 dissociation rate constant; K D 5 equilibrium dissociation constant; ND 5 not determined.Data are representative of 3 independent experiments.

TABLE 2
Influence of Equimolar Concentration of SdAbs and Small-Molecule Inhibitors Talabostat and FAPI-4 on hFAP Enzymatic Activity