Not So Innocent: Impact of Fluorophore Chemistry on the In Vivo Properties of Bioconjugates

Abstract

The combination of targeting ligands and fluorescent dyes is a powerful strategy to observe cell types and tissues of interest. Conjugates of peptides, proteins, and, in particular, monoclonal antibodies (mAbs) exhibit excellent tumor targeting in various contexts. This approach has been translated to a clinical setting to provide real time molecular insights during the surgical resection of solid tumors. A critical element of this approach is generation of highly fluorescent bioconjugates that maintain the properties of the parent targeting ligand. A number of studies have found that fluorophores can dramatically impact the pharmacokinetic and tumor targeting properties of the bioconjugates they are meant to only innocently observe. In this review, we summarize several examples of these effects and highlight strategies that have been used to mitigate them. These include the application of site-specific labeling chemistries, modulating label density, and altering the structure of the fluorescent probe itself. In particular, we highlight the significant potential of fluorophores with hydrophilic, but net-neutral, structures. Overall, this review highlights recent progress in refining the in vivo properties of fluorescent bioconjugates and, we hope, will inform future efforts in this area.

Introduction

Surgical intervention remains a primary treatment for patients with solid tumors. Surgeons typically rely only on palpation and visual cues to differentiate between healthy and tumor tissue. A number of imaging modalities have been utilized to assist surgeons in demarcating tumor boundaries. Among these, optical methods stand out for their non-invasive nature, simple technical requirements and good spatiotemporal resolution. Recent years have seen dramatic developments in the clinical translation of fluorescence-guided surgery (FGS) procedures [1,2]. These approaches provide molecular-level insights in real-time during surgical interventions. Early efforts used fluorophores that emit light in the visible range, which are hampered by high background signal. It has become clear that using near-infrared (NIR) wavelengths is essential, and now most commercial instrumentation operates within this range. However, using these wavelengths requires the accompanying fluorophores, which present a specific set of chemical challenges.

Two untargeted NIR fluorophores, methylene blue (MB) and indocyanine green (ICG) are approved by the FDA and have been investigated extensively for FGS applications (Figure 1a). ICG, in particular, has found use in a variety of oncology applications, with promising results in hepatocellular carcinoma and glioblastoma multiforme [3,4]. However, these applications rely entirely on the ability of the fluorophore to be taken up by tumors, which limits the breadth of feasible applications [5]. Consequently, there have been significant efforts to identify approaches that employ an active targeting element. This approach was first validated through the use of fluorescein-folate conjugates, whose target, the folate receptor, is overexpressed on numerous solid tumors [6–11]. Over the past two decades, the use of labeled monoclonal antibodies (mAbs) that target tumor-associated antigens have also been validated for FGS, with several NIR fluorophore-mAb conjugates progressing through clinical trials [12].

Figure 1:

Overview of common fluorophores and imaging variables. (a) Structures of clinically tested NIR fluorophores with absorbance and emission maxima [21–23]. (Blue highlighting is used throughout to indicate cationic functional groups and red for anionic functional groups). (b) Variables involved in mAb targeted optical imaging strategies.

The construction of bioconjugates with bright NIR emission and high target occupancy in vivo is a significant challenge with many variables to consider. These include aspects of receptor biology (ligand binding, internalization kinetics, receptor recycling rate), various properties of the targeting group, and finally the effects of fluorophore chemistry and bioconjugation method (Figure 1b). This review will focus on the latter, with special attention paid to strategies that improve the properties of bioconjugates for in vivo imaging applications. Bioconjugation has long been known to dramatically alter the properties of many fluorophores, the extent of which can be easily be determined through changes in their spectral properties [13–16]. This challenge was first addressed through the development of persulfonated fluorophores, a successful strategy that led to the Cy-Dye and Alexa-Fluor series [17,18]. However, as detailed below, it has become clear that highly anionic, persulfonated dyes are not optimal for many in vivo applications. This review will describe the role of fluorophores and conjugation strategies in modulating the in vivo properties of the corresponding bioconjugates, and the strategies that have been developed to overcome these effects [19,20].

Peptide-Fluorophore Conjugates

Due to their small size, amenability to high-throughput ligand optimization schemes, and straightforward chemical modification, peptides have been important ligands in the development of molecular imaging [24]. Particularly useful examples include those that target the somatostatin receptor, such as octreotate, and peptides based on a cyclic forms of arginine-glycine-aspartic acid (RGD), which target integrins [25]. Several optical and dual modality imaging agents based on these peptides have been reported and are promising candidates for FGS applications [26]. Another example is the development of synthetic derivatives of chlorotoxin, a tumor-targeting scorpion venom, which has progressed into clinical trials for FGS in recurrent gliomas [27,28].

Given the relatively small size of peptides, it is not surprising that fluorophore modification often has a dramatic effect on the targeting and pharmacokinetic (PK) properties of the resulting bioconjugates [29–31]. While these issues are well established for in vitro cellular imaging, the last decade has provided several examples that highlight the equally large role of fluorophore chemistry for in vivo imaging. Foundational studies by the Choi and Henary labs examined the impact of modifying heptamethine cyanines with butyl trimethylquaternary amines, which they term zwitterionic cyanines (Figure 2a) [29,32,33]. In addition to modifying the clearance pathways of the free dyes, these dyes also alter the in vivo distribution of cRGD conjugates. In a xenograft mouse model bearing α_vβ₃-positive and α_vβ₃-negative tumors, all three peptide conjugates exhibited similar target-positive tumor uptake 4 h post injection (p.i.). However, nonspecific uptake of cRGD-IR-800CW or cRGD-Cy5.5 in the α_vβ₃-negative tumor is significantly higher than with zwitterionic dye conjugates (Figure 2b) [29]. As a result, peptide conjugates of ZW800–1 exhibited significantly higher tumor-to background ratio compared to the highly anionic compounds cRGD-IR-800CW and cRGD-Cy5.5 (Figure 2b). More recent examples by Bunschoten and Bao have also examined the role of fluorophore chemistry for cRGD and anti-PSMA peptides [30] [31]. These efforts also found a large impact of fluorophore charge on the in vivo properties of the resulting peptide bioconjugates.

Figure 2.

Impact of fluorophore structure on peptide conjugates. (a) Structure of cRGD conjugate of Zwitterionic dye (ZW800–1). (b) Imaging 4h post-injection of cRGD-ZW800–1, or cRGD-IR-800CW, or cRGD-Cy5.5 in a xenograft mouse model with transplanted human melanoma cells (either the α_vβ₃-positive M21 cell line or the α_vβ₃-negative M21-L cell line). T (+), integrin α_vβ₃-positive tumor; T (-), integrin α_vβ₃-negative tumor; arrows, kidneys; red circle, region of interest used for background measurement and quantification of fluorescence signal with respect to tumor-to-background ratio (TBR), signal-to-background ratio (SBR) and Positive-to-negative tumor ratio (PNR)

mAb Conjugates in Molecular Imaging

Since the first approval in 1986, mAbs have become invaluable therapeutics with over 60 approved by the FDA [34]. The first efforts Joutoapplythesefor molecular imaging were carried out in the context of radionuclide imaging, and several Zr-89 conjugates now find routine use as PET imaging agents [35,36]. These methods have proven particularly useful for clinical staging and determining antigen presence, thereby informing therapeutic decision making. By contrast, optical methods are particularly useful for FGS applications due to simpler technical requirements and being amenable to real-time readouts. Among the various fluorophores tested, conjugates of heptamethine cyanine NIR fluorophores have proven to be uniquely useful [37–40]. IR-800CW, a commercial fluorophore originally developed for DNA sequencing applications, emerged from these early efforts as the first bioconjugatable variant to progress into clinical testing [21]. Studies by Rosenthal and coworkers have employed IR-800CW conjugates of cetuximab in the context of head and neck squamous cell cancer (HNSCC) [41]. Since then, conjugates of trastuzumab, bevacizumab and girentuximab have all progressed into clinical FGS studies for HNSCC [42,43], breast cancer [44] and renal cancer respectively [45]. The significant clinical potential of these conjugates has motivated a detailed examination of the role of the labeling and fluorophore chemistry.

Impact of Labeling Method

A number of approaches have been developed for preparing small molecule-antibody conjugates. While these have mostly been directed towards the preparation of antibody drug conjugates (ADCs), significant efforts have also been directed towards molecular imaging [46–50]. For optical imaging, the most common method for preparing labeled bioconjugates is the reaction of fluorophore activated esters with lysine residues [51]. This approach is operationally much simpler, however it produces a heterogenous mixture of labeled proteins, which can have detrimental effects on certain properties [52–54]. These issues are due to a subset of the antibody population, which is significantly over labeled. As a consequence, it is quite reasonable to hypothesize that site-specific protein labeling methodologies would offer significant advantages [55,56]. Notably, for full antibodies the impact of site-specific labeling is more dramatic when examining issues of Fc domain interactions [57,58], rather than the binding affinities alone [59,60].

While these issues have mostly been explored for radioisotope imaging and ADCs, the impact of labeling chemistry on optical imaging was recently examined in a study by Hernot and coworkers [61]. These studies compared the effect of two labeling methods with respect to PK and biodistribution in the context of human epidermal growth factor receptor 2 (HER2) targeted nanobodies labeled with IR-800CW. Lysine labeled nanobodies exhibited lower binding affinity to EGFR compared to the site-specific labeled variants. As shown in Figure 3, the site-specific labeled IR-800CW nanobody conjugate showed rapid tumor accumulation with useful TBR as early as 3 h. By contrast, randomly labeled conjugates showed higher liver uptake and dramatically delayed selective tumor accumulation.

Figure 3.

Impact of site-specific labeling on nanobody conjugates. (a) In vivo images of mice bearing HER2+ SKOV3 subcutaneous (s.c.) xenografts after injecting random conjugate (lysine label) or site-specific conjugate (cysteine label) of nanobody 2Rs15d to the NIR dye IR-800CW. (b) In vitro receptor binding constant and in vivo TBR ratio up to 24h post-injection for these conjugates.

Effect of Labeling Density

Increasing the degree of labeling (DOL) of antibody conjugates should in principle increase the brightness of the fluorophore-antibody conjugates. However, in practice, higher DOL is not necessarily accompanied by higher emission due to ground-state quenching effects. Additionally, higher label density can negatively impact the PK properties of the resulting conjugates. These effects have been extensively characterized in the context IR-800CW conjugates, where it was determined that a DOL of less than 1.5 is optimal [62–64]. The impact of DOL on plasma clearance of trastuzumab and bevacizumab conjugates was characterized by Cilliers et al. [65]. This work found that only very low DOL (~0.3) antibody conjugates have comparable plasma clearance to that of total antibody over the course of 4 days and diverge significantly after this time frame. By contrast, moderately higher DOL (~1.3) antibody conjugates exhibited higher plasma clearance as early as 8 h post injection. Notably, these effects are not limited to fluorescence agents, and related observations looking at effect of chelating agents and dual modal agents have also been reported [62,63,66–68].

Role of Fluorophore Structure

Most of efforts looking at the in vivo properties of mAb conjugates have been carried out with IR-800CW and a handful of related commercial fluorophores. While existing polysulfonated fluorophores are superior to their unmodified laser dye precursors, obtaining high-density labeling and tumor uptake comparable to naked or radiolabeled mAb conjugates has been a significant challenge. To address this, our group has optimized the properties of the fluorophore component of mAb conjugates through an iterative series of efforts over the past several years [69]'[70,71]. These studies were initiated by the discovery of a new variant of the classical Smiles rearrangement reaction, providing the first access to C4’-O-alkyl cyanines. Specifically, we found that cyanines substituted at the C4’ position with an N-methylethanolamine undergo efficient N- to O- rearrangement with concurrent incorporation of an electrophile. The reaction proceeds readily with cyclic anhydrides to form a bioconjugatable carboxylate product, which provided our first generation mAb label, the tetrasulfonated Frederick Near IR-774 or FNIR-774. Driven by the long-standing characterization of heptamethine cyanines as chemically unstable, we characterized the chemical reactivity of these C4’-O-alkyl derivatives relative to commonly used C4’-phenol derivatives [1]. We found that phenol- and thiol-modified cyanines react readily with primary thiol nucleophiles under mild conditions (1 mM GSH, pH 7.0 PBS, Figure 4b) to provide the C4’-thio substituted products. By contrast, alkyl ethers, such as FNIR-774, are not reactive under these conditions. To test if analogous reactivity occurs with cellular proteins, these molecules were incubated with whole-cell lysate (Figure 4b). We found extensive cyanine labeling of a variety of proteins with IR-800CW, but not with FNIR-774 or other C4’-O-alkyl cyanines [72]. While the relevance of this covalent reactivity to applications of IR-800CW and related molecules remains to be fully determined, C4’-O-alkyl derivatives represent a simple linkage chemistry that avoids this issue altogether.

Figure 4.

(a) Evolution of C4’-O-alkyl heptamethine cyanines developed by our group. (b) Stability of IR-800CW and FNIR-774 in the presence of biological thiols. Dye (10 μM) was reacted with glutathione (1 mM) in pH 7.4 PBS and reaction was monitored by HPLC over time (inset graph). IR-800CW reacts with thiols t_1/2= 95 min) significantly faster than FNIR-774 (t_1/2 > 5000 min). (c) In vivo serial fluorescence images of MDAMB-468 xenograft tumor-bearing mice (right dorsum) injected with 100 μg each of panitumumab (Pan) conjugates of IR-800CW and FNIR-Tag at a DOL of 4. At a 10 μg dose, Pan-FNIR-Tag showed higher tumor signal and TBR compared with the IR-800CW conjugate and reduced non-specific uptake defined by the liver to background ratio (LBR).

Taking note of several prior efforts, we have sought to optimize the photophysical and in vivo properties of these molecules as mAb conjugates. Needless to say, we were inspired by the studies highlighted above demonstrating the effect of fluorophore properties on peptide conjugates. Additionally, prior efforts looking at mAb radioisotope conjugates found a dramatic role of charge on targeting. These studies suggested that alterations in net molecular charge of a conjugate have a dramatic effect on isoelectronic point, affinity, and disposition of antibody, and resulted in marked variation in tissue distribution and PK [73–75]. Lastly, we also took note of significant efforts from the materials science community, where zwitterionic particles and surfaces have been shown to exhibit dramatically improved properties [76].

Our first studies with FNIR-774 revealed that while low DOL conjugates (~2.0) exhibit improved targeting relative to IR-800CW, higher labeled conjugates (DOL 4–5) were overall similar to IR-800CW in both brightness and tumor targeting. To address this, we first used a strategy similar to that described above and examined cyanines that were substituted with propyl-trimethyl ammonium substituents on the indolenine nitrogen atoms. While these Zwit-FNIR dyes had moderately decreased liver uptake, they still exhibited significant aggregation at higher DOL with relatively similar properties to IR-800CW (Figure 4). We then examined the placement of a quaternary amine at the central carbon and triethylene glycol substituents on the indolenine positions [77]. To our delight, this compound, FNIR-Tag (Figure 4A), exhibited unique properties as a bioconjugatable tag, with properties unlike any other dye we have tested. We observed a nearly linear increase in emission as a function of labeling density (up to DOL 5–6), with no evidence of H-aggregation typically observed with persulfonated dyes. Additionally, in vivo imaging studies found excellent tumor targeting, with minimal liver uptake. Using a commercial IVIS system, we were able to visualize a 1μg dose in mouse studies and found a 10μg dose to be optimal for in vivo use (compared to 100 μg dose for IR-800CW conjugates). Recent studies demonstrated that FNIR-tag can provide improved properties relative to IR-800CW in other contexts, including labeling nanoparticles and for small molecule conjugates [77,78].Recent studies by Smith and coworkers also employed further modification to the C4’ position and observed dramatic effects on both aggregation and photobleaching kinetics [79].

Conclusions

The combination of fluorescent labels and biological ligands is a well-established approach with applications across the fundamental and applied biomedical sciences. The challenges of this approach for live cell imaging of intracellular targets are well established and it is now evident that extensive optimization of the fluorophore component is often required [80]. The demands for carrying out related experiments in living organisms are quite different, and it shouldn’t be surprising that a different set of fluorophores are required. In vivo imaging has been most successful for extracellular targets, though this isn’t always the case [81]. As a consequence, using charged fluorophores is not a concern, and can serve to help confine small molecule and peptide bioconjugates to the extracellular environment. What makes in vivo imaging challenging is the role of competitive clearance pathways. In particular, the reticular endothelial system (RES), and especially Kuepfer cells, within the liver are responsible of clearance of many exogenous macromolecules from circulation [82]. These mechanisms have recently been characterized in great detail with nanoparticles and ADCs and are likely, at least in part, responsible for the altered clearance of fluorophore-labeled bioconjugates [83,84]. Thus, the general observation that highly charged but net neutral conjugates can evade these uptake processes is significant, though more detailed mechanistic studies are needed to rationalize this observation.

Going forward, the identification of not only target presence, but enzymatic activity is of significant interest. Most of the fluorophores discussedin this review are in their ‘on state’, i.e., they always emit fluorescence signal. Fluorogenic probes whose fluorescence properties respond to biological stimuli can, in principle, provide a high contrast without waiting for biological washout. Approaches to develop probes that target various enzymes and reactive oxygen species have been reported [85][86–89]. Underexplored, and an area of significant potential going forward, is the union of turn-on probes and active targeting. Critical in such efforts is the development of probes with properties that tolerate biomolecule labeling. Overall, there is significant potential for molecular imaging to interrogate biological processes in live animals and clinical settings, and the optimization of probe chemistry will be critical to the realization of these goals.