1.

Introduction

Since the seminal report of the in-human safety and tumor-specificity of fluorescent antibodies in work described by Rosenthal et al., leveraging FDA-approved therapeutic antibodies as fluorescent clinical molecular probes for image-guided surgery is becoming an ever-growing reality.^1–4 Rosenthal et al. fabricated a fluorescent conjugate of IRDye 800CW with the human-murine chimeric monoclonal antibody cetuximab for the in vivo clinical imaging of epidermal growth factor receptor (EGFR) in head and neck cancer patients.¹ This development in the clinical imaging of cancer using targeted, optically active biologics has been warranted by the long-standing unmet clinical need for assistance in surgical navigation.⁴ Prior attempts to address this critical need were initially approached in 1948 by administering nonspecific fluorescein dye to image perfused brain tumors in patients.⁵ More recently, far-red and near-infrared dyes, namely methylene blue (MB) and indocyanine green (ICG), exhibiting superior suitability for imaging deep tissue have proven to be valuable fluorescence contrast agents for the mapping of sentinel lymph nodes, monitoring blood perfusion, and imaging vascular and lymphatic pathologies to assist surgical procedures.^6,7 In the context of oncology, limited reports have demonstrated the detection of hepatic and breast malignancies using intravenously administered ICG, leveraging delayed dye interstitial clearance from tumors to provide selectivity.^8,9 However, the weak selectivity and absence of discrete tumor specificity of fluorophores, such as ICG and MB, render them unreliable for accurate tumor detection and image-guided surgery. The necessity for increased tumor specificity has since lead to strategies that leverage the tumor tissue’s inherent capacity to synthesize an accumulated amount of fluorescent protoporphyrin IX following the exogenous administration of 5-aminolevulinic acid.¹⁰ In June 2017, 5-aminolevulinic acid received FDA approval as Gleolan^® for image-guided surgery of glioma and a number of clinical trials also leveraging the approach to guide surgical resection are ongoging.⁴ The described strategies, although powerful in their own right, provide weak specificity when imaging cancer and can provide heterogeneous intratumoral signals, requiring secondary approaches to improve their specificity and homogeneity of labeling.^11,12 The earliest demonstration of leveraging fluorescent antibodies for imaging human disease tissue was reported by Folli et al. who intravenously administered fluorescent anticarcinoembryonic antigen antibodies to patients with primary colorectal carcinoma and imaged the tissue ex vivo following surgical resection.¹³ The first demonstration of in-human molecular imaging in patients was reported in 2011 by van Dam et al. who visualized ovarian cancer metastases, of which 90% to 95% overexpress the folate receptor-α, using a folate–fluorescein conjugate to guide radical cytoreductive surgery.¹⁴ Following the first in-human demonstration of image-guided surgery using cetuximab–IRDye 800CW in 2015,¹ a phase I trial was initiated for the use of the conjugate for image-guided surgery of head and neck cancer, (NCT01987375). A second trial imaging EGFR using cetuximab–IRDye 800CW is also pending. Four phase I, clinical trials using an IRDye 800CW conjugate of the antivascular endothelial growth factor receptor antibody have also been initiated to image familial adenomatous polyposis (NCT01691391), breast cancer (NCT01508572), rectal cancer (NCT01972373), and premalignant esophageal lesions (NCT02129933).⁴ An additional pending clinical trial will also leverage an antiprostate-specific membrane antigen antibody to image prostate cancer (NCT02048150). Of particular relevance to this study, a current clinical trial is performing intraoperative imaging of pancreatic cancer using a cetuximab–IRDye 800CW conjugate (NCT02736578), further motivating our demonstration here of a dual-activateable probe approach in an orthotopic model of pancreatic ductal adenocarcinoma.

Building on the wealth of antibody-based molecular probes for image-guided surgery, the concept of proteolytic probe activation is an elegant means to enhance the specificity of protein-based molecular probes. We have previously reported an activatable photosensitizer-cetuximab conjugate whereby proteolytic intracellular degradation of the antibody conjugate resulted in tumor-specific activation of imaging and photodynamic therapy.^15–18 These antibody probes have been demonstrated to be effective therapeutic agents with the potential to serve a dual diagnostic and therapeutic function.^15,19–21 Furthermore, they can be leveraged to mop up the surgical bed from residual, undetectable microscopic disease.^15,20 Static quenching of fluorophores when confined in proximity to one another on an antibody can exist through self-quenching or by the introduction of a Förster resonance energy transfer- and contact-based dark quencher.^22–26 In this approach, activation is dependent on the spatial separation of statically quenched fluorophores upon lysosomal degradation. In this study, we present a flexible platform optimized for the activation of far-red and near-infrared antibody fluorophore conjugates using IRDye QC-1 as a broad-spectrum quencher to enable the synthesis of molecular-targeted, activatable fluorescent probes over a spectrum of fluorophore colors, given that sufficient spectral overlap exists between the fluorophore’s emission and the quencher’s absorption profile.

As described in the methods section, varying ratios of Alexa Fluor 700 (AF700) and IRDye QC-1 (QC-1) were conjugated to cetuximab to obtain a final dye/antibody ratio of 8 [Fig. 1(a)] to provide conjugates with increasing degrees of the dark quencher QC-1 [Fig. 1(b)]. This dye ratio was chosen to avoid over labeling that compromises the biological activity and specificity of the antibody. The conjugates were then subject to a simulation of intracellular lysosomal proteolysis using trypsin in vitro to assess the degree of activation the probes exhibited. With increasing QC-1 composition, the degree of cetuximab activation, reaching a maximum of $\sim 10$-fold activation with the optimal $1:2:6$ ratio of cet:AF:QC-1 [Fig. 1(c)]. Further incorporation of QC-1 at a $1:1:7$ ratio of cet:AF:QC-1 resulted in an inferior degree of activation compared to $1:2:6$. The enhanced specificity provided by the QC-1 in the cet-AF700 conjugates also results in a 29.4% compromise in signal, as shown by the raw fluorescence emission spectra [Fig. 1(d)], which is not anticipated to be problematic in the clinic, as probe specificity rather than fluorophore brightness remains the major unmet clinical need.¹ Enhanced activation is also demonstrated by the incorporation of the QC-1 in cetuximab AF660 conjugates [Fig. 1(e)], exemplifying the flexibility of this approach to synthesize activatable conjugates using other fluorophores.

Fig. 1

(a) Schematic representation of the panel of cetuximab (cet) conjugated to varying ratios of Alexa Fluor 700 (AF) and IRDye QC-1 (QC-1) amounting to a total of eight chromophores per antibody. (b) Normalized UV–visible absorption spectra of the varied ratio cetuximab conjugates with increasing degrees of QC-1 incorporation. (c) Proteolytic activation of the panel of varying fluorophore-quencher ratio cetuximab conjugates by 24 h of trypsin digestion revealed that $1:2:6$ ratio of cetuximab:Alexa Fluor 700:IRDye QC-1 exhibited the highest degree of activation with a 9.8-fold increase in fluorescence following probe digestion. (d) Raw fluorescence emission spectra of digested, activated cetuximab:Alexa Fluor 700 with and without IRDye QC-1 demonstrate that the increased specificity provided by the quencher results in only 29.4% compromise in brightness. (e) Activation was also demonstrated in Alexa Fluor 660 cetuximab conjugates showing an improvement in fold activation with QC-1 incorporation at a ratio of $1:2:6$. Data points are $\mathrm{mean}\pm \mathrm{S.D.}$, statistical significance was calculated using a one-way ANOVA analysis with Tukey post-test comparison in (c) and using a two-tailed $t$-test in (e). Cetuximab is represented by the refined structure of an intact IgG2a monoclonal antibody (Protein Data Bank; PDB ID: 1IGT)²⁷ and the three-dimensional (3-D) structure was projected using Jmol: an open-source Java viewer for chemical structures in 3-D.²⁸

Fundamental principles of the static quenching and intracellular activation approach are shown schematically [Fig. 2(a)] along with basic spectroscopic data showing the overlap of the QC-1 absorption spectrum with the fluorescence emission of the exemplary far-red AF dye pre- and postactivation for the optimal $1:2:6$ ratio [Fig. 2(b)].

Fig. 2

(a) Schematic representation of the targeted intracellular proteolysis of optimal cetuximab–Alexa Fluor 700–IRdye QC-1 ($1:2:6$) leading to the spatial separation of the fluorophore from the quencher and subsequent fluorescence activation (the structure of the proprietary Alexa Fluor 700 is depicted by the structure of the related dye Alexa Fluor 647). (b) Fluorescence emission spectra of quenched, intact cet:AF:QC-1 ($1:2:6$) and the activated, proteolyzed cet:AF:QC-1 ($1:2:6$) with respect to the absorption spectrum of the IRDye QC-1 quencher.

To further test that the optical ratio identified for cetuximab-based probes holds for other antibodies, we compared the efficiency of self-quenching ($1:8:0$; no QC-1 dark quencher) with the $1:2:6$ ratio for trastuzumab [Fig. 3(a)] and an IgG isotype control [Fig. 3(b)]. The enhanced activation potential using the QC-1 dark quencher at the optimal ratio held for all three of the antibody conjugates [Fig. 3(c)]. This finding suggests that this approach can be applied to a wide range of fluorescent dyes to potentially construct a panel of multicolor probes (i.e., any dye that overlaps with the broad QC-1 absorption spectrum, which extends from $\sim 500$ to 900 nm)²⁴ for efficient quenching of visible, far-red, and near-infrared dyes).

Fig. 3

Schematic representations and degree of proteolytic activation of (a) trastuzumab and (b) IgG isotype control conjugated to the optimal Ab:AF:QC-1 ratio of $1:2:6$. (c) Percentage increases in activation over the entire emission spectrum of Alexa Fluor 700 by introducing IRDyeQC-1 at the optimal ratio. Data points are $\mathrm{mean}\pm \mathrm{S.D.}$ and statistical significance was calculated using a one-way ANOVA analysis with Tukey post-test comparison. Cetuximab, trastuzumab, and human IgG isotype control are represented by the refined structure of an intact IgG2a monoclonal antibody (PDB ID 1IGT)²⁷ and the 3-D structure was projected using Jmol: an open-source Java viewer for chemical structures in 3-D.²⁸

We then performed a preliminary xenograft study in orthotopic pancreatic cancer (AsPC-1) to demonstrate tumor-specific activation of dual-activatable tracers (Fig. 4). Fluorescence image-guided surgery was simulated by acquiring hyperspectral near-infrared fluorescence images of dual-activatable cetuximab-AF660-QC-1 ($1:2:6$) and IgG-AF700-QC-1 ($1:2:6$) sham probes coinjected intravenously into mice bearing the orthotopic AsPC-1 pancreatic cancer xenografts. The tumor margin is clearly visualized by the cetuximab-based activatable probe whereas the nonspecific IgG-activatable probe enabled visualization of nonspecific accumulation with significant background in the pancreas and surrounding organs [Figs. 4(a)–4(j)]. Figure 4(j) is a heat map (using the “fire” look up table, LUT, in ImageJ) image for 0 to 5 that quantitatively shows specific binding where cet is above the IgG nonspecific level, thus delineating the tumor margins more accurately, as compared to the activateable cet alone in Fig. 4(i). The qualitative margin delineation presented here is preliminary and would require histological analysis for validation. The heat map was generated by subtraction of the image intensity histogram mode of each raw image (cet and IgG) to set the mean background to zero. Each image is then normalized (divided) by its overall mean signal intensity. For each pixel, the ratio cet/IgG – 1 was calculated, such that 0 (1 − 1) indicates no specific binding and values greater than 0 highlight specific binding, and this is presented as the heat map image in Fig. 4(j). The ratio used for image analysis of the dual-activateable probe technique was adapted from a previously reported study.²⁹

Fig. 4

Hyperspectral near-infrared fluorescence imaging a murine model of orthotopic AsPC-1 pancreatic cancer xenografts simulating fluorescence image-guided surgery using the dual-activatable cetuximab-AF660-QC-1 ($1:2:6$, red) and IgG-AF700-QC-1 ($1:2:6$, blue) sham probes. (a) White light and (b) hyperspectral composite images of the abdominal cavity and the pancreas bearing an orthotopic ASPC-1 tumor 15 days after implantation. (c) Following resection of the tumor-bearing pancreas (white light), (d) no evidence of residual disease remains as shown in the hyperspectral composite image. (e) Following resection of the AsPC-1 tumor-bearing pancreas, tumor margins are undetected by white light, (f) whereas unmixed cetuximab-AF660-QC-1 ($1:2:6$, red) enables tumor visualization, (g) unmixed IgG-AF700-QC-1 ($1:2:6$, blue) enables visualization of nonspecific antibody binding profiles, and (h) the composite hyperspectral image of the pancreas identifies areas where nonspecific binding contributes to the signal of the targeted probe. (i) Tumor margins delineated by activatable cetuximab-AF660-QC-1 ($1:2:6$, red) . A heat map image (using the “fire” LUT in ImageJ) subtracting the nonspecific component of cet binding provides superior specificity of imaging and an improved tumor margin delineation are shown by yellow arrows in (j) (see text for image computation details).

Future work will exploit this new activatable dual tracer approach to perform quantitative imaging of cellular uptake and molecular target levels in the tumor, extending a new powerful approach to perform dual tracer molecular fluorescence imaging that enables quantitative imaging of receptor binding and cellular uptake based on pharmacokinetic modeling.^30,31

Finally, an exciting area of future development will able to integrate this multicolor, activatable imaging probe platform with theranostic agents that in addition to fluorescence also produce cytotoxic-reactive species (e.g., singlet oxygen) to perform therapy^15,20 (i.e., tumor-targeted, activatable photoimmunotherapy, and taPIT). This development will have potential to enable surgeons to selectively “mop up” residual microscopic disease left behind in the surgical bed.

2.

Materials and Methods