A series of fluorescent FAPI conjugates were synthesized to understand the impact of internalization and plasma kinetics on in vivo FAP molecular imaging in preclinical arthritis models (Fig. 1). FAPI-AF488 and FAPI-AF647 probes were synthesized to characterize in vitro receptor internalization kinetics via flow cytometry and microscopy, respectively. Notably, AF488 was selected due to ease of Trypan Blue quench in cytometry applications. For in vivo imaging, selection of the detectable label is not as straight forward. On one hand, lipophilic fluorophores such as non-sulfonated cyanine dyes can impart both albumin-binding properties to slow clearance and enable NIR tracking. However, direct conjugation of Cy7 significantly weakened FAP binding in vitro (data not shown). To maintain binding affinity while also increasing the systemic half-life of FAPI, alternative fluorophores and protein binding constructs were evaluated. The red dye (AF647) and near-infrared (NIR) dye (800CW) conjugates of FAPI designs were prepared48 (Fig. 1). The biphenyl acylsulfonamide moiety—with high binding affinity towards domain III of human serum albumin—was introduced to slow plasma half-life, and these synthesized FAPI compounds have an ALB suffix.
Compound lipophilicity was predicted through a cLogP calculation from the molecular structure, whereas plasma protein binding (PPB) was measured experimentally using equilibrium dialysis (Table 1). Whereas FAPI alone had relatively low plasma protein binding (44.7%), conjugation to biphenyl acylsulfonamide significantly increased albumin binding (> 98%). Addition of the hydrophilic sulfated fluorophore AF647 did not substantially change the plasma protein binding relative to the FAPI-ALB binder alone.
Cell surface affinity data were generated to assess the impact of fluorophore conjugations on binding affinity of the parental FAPI small molecule. While low degree of label modifications minimally impact binding for higher molecular weight protein ligands, the same may not be true for low molecular weight ligands as the label may be proximal to binding pockets. Nine-point affinity curves were generated for AF488 and AF647 FAPI derivatives via flow cytometry using HEK293-hFAP cells (Fig. 2).
Binding affinity (KD) for unconjugated and 800CW conjugates were calculated indirectly through competition assays conducted at equimolar concentrations of probes with known binding affinity and unknown binding affinity. All FAPI derivatives maintained low nanomolar KD whether measured directly or indirectly (Table 2A).
Fluorophore conjugation had minimal impact on the affinity; FAPI-AF488 and FAPI-AF647 had a KD of 4.8 nM and 5 nM, respectively (Table 2A). The KD of FAPI-800CW was measured to be 8 nM, comparable to FAPI-AF488 and FAPI-AF647. Similarly, the addition of the albumin-binding moiety minimally impacted affinity (KD of 4 nM to 10 nM for ALB compounds). Lastly, we estimated cell surface FAP expression using an overexpressing system (HEK293-hFAP cells) as well as primary RAFLS cells. RAFLS cells had ~ 25,000 copies per cell; HEK293-hFAP had ~ 1 million per cell (Table 2B). Both cell types had a negligible amount of non-specific sticking to FAPI ligands (Figure S1).
As there are reported differences in both the rate and extent of internalization for cell surface FAP in various cells35,49, we evaluated receptor internalization using live-cell time-lapse microscopy and quantitative flow cytometry (Fig. 3).
Qualitatively, live-cell time-lapse imaging of HEK293-hFAP cells showed rapid surface binding upon FAPI addition, followed by receptor mediated endocytosis. Cell surface binding equilibrated rapidly, which agrees well with the strong affinity of the ligand—namely fast kon— and the concentration of ligand added in the microscopy experiments (Fig. 3A). At longer times, the punctate, intracellular AF647 signal co-localizes well with lysosomal markers (Figure S2), suggesting that FAPI is eventually trafficked to lysosomes. Following visual confirmation of endocytosis, we sought to quantify the net internalization rate constant using a modified protocol from Schmidt et al50. The net internalization rate constant was measured on overexpressing HEK293-hFAP cells and endogenously expressing RAFLS cells. Cells were allowed to internalize continuously for predetermined amounts of time and the total intensity per cell was compared to the surface quenched intensity per cell. Importantly, this approach to quantifying the net uptake rate accounts for time-dependent cell-surface expression. After accounting for the quenching efficiency, the fluorescent intensities from total, surface, and internal cellular compartments were plotted against time for the overexpressing and endogenous cells (Fig. 3B). Surface-bound signal remained constant after 1 h, suggesting limited modulation of surface FAP expression during the time course of the experiment. Additionally, near constant surface expression was observed with an AF647 conjugated anti-FAP antibody (Figure S3). Non-specific uptake was assessed for this assay and showed minimal uptake (less than 0.2% of internal signal) for HEK293-hFAP cells, while RAFLS cells showed increased non-specific uptake (~ 20% of internal signal). The internalized signal adjusted for non-specific uptake was plotted against the integral of the surface signal over time and fit via linear regression with the slope being equivalent to the net internalization rate constant (Fig. 3C). The net internalization half-lives for RAFLS and HEK293-hFAP cells were estimated at 3.7 h and 7.2 h, respectively. As primary RAFLS cells are collected from various donors, we conducted additional net internalization assays to measure donor variability. Interestingly, the estimated net internalization half-lives ranged from 3.7 h to 18.2 h (Figure S4) with differences in non-specific uptake as well. While the absolute flow cytometry signals for donor cell experiments were low, our results highlight challenges for conducting these quantitative assays in low-expressing and/or primary cells.
The moderate internalization half-life measured for FAP and single digit nanomolar affinity of the FAPi probes highlights their potential for target engagement, however, an understanding of systemic exposure in vivo is important to characterize imaging probe capabilities. The impact of FAPI conjugations on pharmacokinetics (PK) was investigated in healthy mice (Fig. 4) and then PK parameters were generated with a non-compartmental analysis (Table 3).
Unconjugated FAPI demonstrated the fastest clearance rate with a moderate volume of distribution at steady-state, while conjugation of biphenyl acylsulfonamide, AF647, or 800CW resulted in decreased clearance rates and increased exposures. Addition of biphenyl acylsulfonamide further decreased clearance rates of both fluorescently labeled FAPI derivatives resulting in a terminal clearance half-life of ~ 5.5 h. In addition to an improved half-life, the fluorescent ALB conjugates demonstrated greater maximum plasma concentrations consistent with lower volumes of distribution, while the rapidly clearing conjugates had moderately lower maximum concentrations associated with larger volumes of distribution (Fig. 4, Table 3).
Given the significant differences in clearance between FAPI derivatives, yet comparable in vitro cell surface affinity, all FAPI derivatives were assessed via in vivo near-infrared imaging in a murine model of RA.
The CIA murine model of RA was used to evaluate in vivo molecular imaging of FAP preclinically. After the first symptoms of disease51,52, paw tissues were collected at 7 d, 14 d, and 21 d with corresponding IHC to confirm disease progression (Fig. 5).
Enhanced pannus formation was observed at 21 d vs healthy controls (Fig. 5A). Immunohistochemistry suggests increasing FAP expression from 7 d to 21 d and correlated with disease severity (Fig. 5B). After confirming the expression of FAP in this murine model, the FAPi probes were assessed via near-infrared imaging (Fig. 6). As it can be challenging to compare in vivo fluorescent intensities for different fluorophores, the compounds in this work were evaluated based on their ability to block and compete the near-infrared tagged FAPI ligands through co-administration of compounds, rather than imaging each standalone probe in their respective fluorescent channels. The detectable NIR probe was paired with a blocking probe with similar albumin binding capabilities; for example, probes (FAPI and FAPI-AF647) without the albumin-binding small molecule were used to block FAPI-800CW whereas the compounds with albumin-binding small molecule (FAPI-ALB and FAPI-ALB-AF647) were used to block FAPI-ALB-800CW.
FAPI-800CW and FAPI-ALB-800CW uptake in inflamed joints following intravenous (IV) dosing was observed in all groups. Unsurprisingly, the highest 800CW intensity was observed in animals that did not receive blocking doses and the lowest intensity was observed in healthy paws (Fig. 7, Figure S5). When FAPI-800CW was co-dosed with a molar excess of FAPI-AF647, FAPI-800CW uptake in inflamed paws was reduced to a similar MFI as healthy imaging controls at 5 h (Fig. 7A). Interestingly, a molar excess dose of the parental FAPI ligand did not provide a similar reduction of FAPI-800CW signal, suggesting improved in vivo targeting properties for FAPI-AF647. A comparison of the target specific signal blocked between the parental FAPI ligand and FAPI-AF647 demonstrated a 2-fold improvement in blocking for the faster clearing probes (Fig. 7C). The analogous comparison with the slower clearing FAPI-ALB probes showed similar trends (Fig. 7B and 7C). Here, FAPI-ALB-AF647 was a better blocker of FAPI-ALB-800CW than FAPI-ALB (24 h: 0.05 ± 0.01 vs 0.07 ± 0.01 for FAPI-ALB-AF647 block vs FAPI-ALB block, respectively). Expectedly, slower systemic clearance increased overall exposure and led to increased non-specific signal, as demonstrated by a 4-fold increase in mean fluorescence intensity in healthy paws at 5 h (0.026 ± 0.007 vs 0.10 ± 0.01 for FAPI-800CW in healthy vs FAPI-ALB-800CW in healthy, respectively, P < 0.001). The significant reduction in FAPI-800CW and FAPI-ALB-800CW signals upon co-administration of the AF647 conjugated versions of FAPI and FAPI-ALB, respectively, suggest improved targeting of the AF647 reagents imparted through a combination of their improved hydrophilicity (Table 1) and enhanced systemic circulation (Fig. 4 & Table 3).