Paired Agent Fluorescence Imaging of Cancer in a Living Mouse Using Pre-assembled Squaraine Molecular Probes with Emission Wavelengths of 690 and 830 nm.

Abstract

New methods are described for construction of targeted fluorescence probes for imaging cancer and assessment of tumor targeting performance in a living mouse model. A novel non-covalent assembly process was used to fabricate a set of structurally related targeted fluorescent probes with modular differences in three critical assembly components: the emission wavelength of the squaraine fluorochrome, the number of cRGDfK peptide units that target the cancer cells, and length of the polyethylene glycol chains as pharmacokinetic controllers. Selective targeting of cancer cells was proved by a series of cell microscopy experiments followed by in vivo imaging of subcutaneous tumors in living mice. The mouse imaging studies included a mock surgery that completely removed a fluorescently labeled tumor. Enhanced tumor accumulation due to probe targeting was first evaluated by conducting Single Agent Imaging (SAI) experiments that compared tumor imaging performance of a targeted probe and untargeted probe in separate mouse cohorts. Although there was imaging evidence for enhanced tumor accumulation of the targeted probe, there was moderate scatter in the data due to tumor-to-tumor variability of the vasculature structure and interstitial pressure. A subsequent Paired Agent Imaging (PAI) study co-injected a binary mixture of targeted probe (with emission at 690 nm) and untargeted probe (with emission at 830 nm) into the same tumor-burdened animal. The conclusion of the PAI experiment also indicated enhanced tumor accumulation of the targeted probe but the statistical significance was much higher, even though the experiment required a much smaller cohort of mice. The imaging data from the PAI experiment was analyzed to determine the targeted probe’s Binding Potential (BP) for available integrin receptors within the tumor tissue. In addition, pixelated maps of BP within each tumor indicated a heterogenous spatial distribution of BP values. The results of this study show that the combination of fluorescent probe pre-assembly and PAI is a promising new way to rapidly develop targeted fluorescent probes for tumors with high BP and eventual use in clinical applications such as targeted therapy, image guided surgery, and personalized medicine.

Graphical Abstract

INTRODUCTION

There is major ongoing research efforts to create targeted fluorescent probes for tumors in living subjects, with the goal of developing imaging methods that facilitate pre-clinical cancer research or clinical procedures such as fluorescence-guided surgery.^1,2,3,4 It is well-recognized that the probe emission wavelength should be in the near-infrared window that permits optimal penetration of light through skin and tissue.⁵ Various probe architectures have been examined, ranging from small molecule conjugates to relatively large nanoparticles. To date, most in vivo fluorescence imaging studies have focused on experiments that dose each subject with a single fluorescent probe. In principle, more imaging information can be gained if the subject is dosed with two or more fluorescent probes that emit at different wavelengths and can be distinguished by using an imaging station equipped with multiple fluorescence filter channels. Some instrument vendors have focused on surgical applications using video-rate imaging stations that can distinguish two near-infrared wavelength channels, such as ~700 nm and ~800 nm,⁶ but recent development of new fluorescent probes which emit in the short wavelength infrared range of 1000–1400 nm suggests that the detection window for multicolor imaging can be expanded.⁵ Multicolor fluorescence imaging has been used to visualize the anatomy and flow of the lymphatic system,⁷ or alternatively to measure tumor accumulation by probes that are equipped with targeting unit(s) that have avidity for biomarkers on the surface of the cancer cells.^8,9,10 In the latter case, tumor imaging is performed after waiting for sufficient clearance of the unbound background probe from the blood stream and surrounding tissues. In principle, there are several different pharmacokinetic processes that can lead to enhanced retention of fluorescent probes within tumors. One of the best known is the Enhanced Permeation and Retention (EPR) effect which is attributed to the enhanced leakiness of tumor vasculature and reduced drainage to lymph nodes, leading to entrapment of the probe in the tumor parenchyma.^11,12 Probe accumulation in tumors by the EPR effect is sometimes called passive targeting, and there has been considerable effort to exploit it for tumor delivery. However, the ubiquity of the EPR effect makes it challenging to unambiguously determine if the targeting units on a targeted probe have any significant influence on the degree of tumor accumulation. Moreover, tumor-to-tumor variability of the vasculature structure and interstitial pressure means the influence of the EPR effect might be heterogeneous within a single subject.¹² The effectiveness of a fluorescent probe’s targeting units is often measured by conducting a set of comparative Single Agent Imaging (SAI) experiments that quantify the tumor uptake of a targeted probe and a homologous untargeted probe (lacking target units) in separate animal cohorts. An alternative experimental paradigm is Paired Agent Imaging (PAI), which simultaneously doses a subject with a binary mixture of targeted and untargeted fluorescent probes (each probe is endowed with a distinguishable imaging wavelength) and quantifies tumor accumulation of each probe in the same tumor.^8,9,10,13,14

Experimentally, the logical workflow for developing an effective fluorescent probe is an iterative cycle of structure optimization experiments that test imaging performance in cell culture and subsequently in animal models.^15,16,17 A typical fluorescent molecular probe is a covalently linked conjugate composed of three critical components: the fluorochrome, targeting unit(s), and pharmacokinetic controller.¹⁸ The conjugate is prepared by a sequence of synthetic reactions, often involving purification steps that consume time and resources. Because of this high cost, iterative cycles of probe optimization studies are rarely attempted. We are working to obviate this synthetic inefficiency by developing a much faster non-covalent probe fabrication process that we call Synthavidin Technology.^19,20,21 As shown in Scheme 1, a stable pre-assembled fluorescent probe is created by combining two molecular components. One component is a tetralactam macrocycle that has appended cell targeting units, and the other component is a near infrared fluorescent squaraine dye (that determines fluorescence emission wavelength) with flanking polyethylene glycol (PEG) chains (that determine pharmacokinetic properties). When the two components are mixed together in equimolar amounts, the macrocycle is spontaneously and quantitatively threaded by the dye to create a stock solution of stable pre-assembled fluorescent probe that can be used for molecular imaging experiments. The iterative structure optimization cycle starts with cell culture experiments and then progresses to animal models, with the imaging results providing feedback that informs structural revision of the probe components leading to incremental performance adjustments of the pre-assembled probe. In principle, the modular and combinatorial nature of the non-covalent probe pre-assembly process (Synthavidin Technology) can greatly accelerate the pace of probe development.

Scheme 1.

Schematic depiction of the iterative structure optimization cycle that uses non-covalent pre-assembly (Synthavidin Technology) to fabricate fluorescent probes for imaging experiments, first in cell culture and then in animal models. The insight provided by the imaging results leads to informed structural revision of a probe component and an incremental change in probe imaging performance.

This article presents several new advances in our fluorescent probe pre-assembly method. We describe the imaging properties of seven pre-assembled probes and present the experimental results in four related sections. The first section describes the probe pre-assembly process using a new class of tetralactam macrocycles, and the photophysical properties of the pre-assembled probes. The second section highlights the synthetic efficiency of the probe pre-assembly method by describing a series of cell microscopy experiments that assess the extent and selectivity of probe targeting to cancer cells. In these studies, the same macrocycles are recycled and threaded with squaraine dyes that have flanking PEG chains of different lengths. The third section describes a set of SAI studies that compare tumor imaging performance of a targeted probe and untargeted probe in separate mouse cohorts. Finally, the fourth section describes a PAI study that doses each mouse in a cohort with a pair of fluorescent pre-assembled probes (untargeted and targeted). The two probes have different emission colors that enable simultaneous imaging of probe accumulation in the same tumor, a procedure that greatly increases the statistical significance of the imaging results and reduces the number of mice that have to be used. In addition, PAI is used to determine a targeted probe’s Binding Potential (BP) which is a quantitative measure of its in vivo targeting effectiveness.^8,14

RESULTS AND DISCUSSION

Probe Pre-assembly.

Shown in Scheme 2A are the six probe assembly components (two tetralactam macrocycles and four squaraine dyes) that were used to pre-assemble the seven fluorescent probes used in this study. Our previous imaging work had used tetralactam macrocycles whose structures contained anthracene sidewalls, but here we employ a new family of tetralactam macrocycles with 2,3,5,6-tetramethyl-1,4-phenylene sidewalls which we recently showed can be threaded with squaraine dyes to give very stable non-covalent complexes with excellent water solubility and fluorescence properties.²² Using standard synthetic chemistry methods, we prepared two homologous tetralactam macrocycles (untargeted and targeted). Appended to each end of the targeted macrocycle T are two copies of a cyclic peptide with the sequence cRGDfK which is known to have high affinity for the integrin receptors that are overexpressed on many cancer cells including the A549 (lung cancer) and OVCAR-4 (ovarian cancer) cells employed in this study.^21,23 Thus, pre-assembled probes with macrocycle T were expected to target cancer cells in culture and xenograft tumors in mice. The untargeted counterpart of T is the control macrocycle C whose structure has two appended tricarboxylate units which provide high water solubility and has no affinity for any cell type. Pre-assembled probes with macrocycle C were expected to have no cancer targeting potential and only accumulate in tumors by passive targeting. Previous studies have also shown that the pre-assembled probes remain complexed during cell and animal imaging experiments.^19,21 It is important to note that if there is a small amount of unthreading during animal imaging, the blue-shifted and weakly fluorescent SQ690-P12, SQ690-P45, or SQ830-P45 dyes that are produced by probe unthreading cannot be detected (Figures S18, S22 and S23), and thus do not affect the imaging results.

Scheme 2.

Structures and identifiers for: (A) Six individual components used to pre-assemble seven fluorescent probes for SAI and PAI. Squaraine dyes are colored blue; the tetralactam macrocycles are red with conjugated peptide targeting groups colored green. (B) Pre-assembled fluorescent probes which are stabilized by a combination of hydrogen bonding between the squaraine oxygens and the tetralactam NH residues, and hydrophobic stacking of the squaraine aromatic rings with the 2,3,5,6-tetramethyl-1,4-phenylene sidewalls of the surrounding tetralactam macrocycle.

Three of the four squaraine dyes have the same emission wavelength of 690 nm and they differ by the length of the appended PEG chain at each end of the dye (SQ690-P3, SQ690-P12, SQ690-P45) which alters the pharmacokinetics of the pre-assembled probe. We expected that a longer PEG chain would more effectively shield the probe from biological surfaces.²⁴ The fourth squaraine (SQ830-P45) emits at 830 nm and was designed for use with a 690 nm probe for PAI.²⁵ Shown in Scheme 2B are simplified structures of the seven pre-assembled probes that were pre-assembled in quantitative yield by simply mixing a macrocycle and squaraine dye in equimolar amounts. Quantitative formation of a pre-assembled probe was proved by observing a diagnostic red-shift of the squaraine absorption maxima band and a large enhancement of squaraine emission intensity (Figures S6–S8).^22,25,26 Analysis by gel electrophoresis provided independent confirmation that the stock solutions of pre-assembled probes were pure, and there was no evidence of probe unthreading (Figures S9–S11). The spectral properties of each probe are provided in Table 1 and Table S1. As a general trend the squaraine fluorescence quantum yields for the pre-assembled probes were substantially higher than the free squaraine dyes which agrees with previous observations. Another trend in the fluorescence was higher emission intensity for the probes with macrocycle C compared to analogous probes with macrocycle T. This is likely due to the difference in molecular charge, with probes containing the hexa-anionic macrocycle C disfavoring collisional quenching processes compared to probes containing the near-neutral macrocycle T.²⁵

Table 1.

Photophysical properties of pre-assembled complexes in water at 3 μM.

Pre-assembled Probe	λabs (nm)	λem (nm)	log ε	Φf
C ⊃ SQ690-P3	663	690	5.393	0.22
T ⊃ SQ690-P3	663	691	5.382	0.15
C ⊃ SQ690-P12	665	691	5.361	0.21
T ⊃ SQ690-P12	665	694	5.310	0.16
C ⊃ SQ690-P45	668	689	5.250	0.13
T ⊃ SQ690-P45	670	693	5.069	0.12
C ⊃ SQ830-P45	804	830	5.242	0.10

Cell Microscopy Studies.

To evaluate cytotoxicity, the probes C ⊃ SQ690-P3 and C ⊃ SQ690-P45 were incubated with CHO-K1 (Chinese Hamster Ovary) cells for 24 h at various concentrations. Negligible cytotoxicity was observed with probe concentration up to 5 μM (the highest level tested) which was five times higher than the probe concentrations used for cell microscopy (Figure S12). The targeting abilities of T ⊃ SQ690-P3, T ⊃ SQ690-P12, and T ⊃ SQ690-P45 were then evaluated with integrin positive cell lines (A549 and OVCAR-4) and an integrin negative cell line (EMT-6). For the A549 cell line (human lung adenocarcinoma), high protein expression of α_vβ₅ integrin receptors was previously reported with an average of [10.2 ± 3.7] × 10⁴ β₅ protein molecules on the surface of each cell.²¹ Furthermore, α_vβ₅ integrin receptors strongly bind cRGDfK peptides, and the A549 cell line is a well-established system for investigating integrin targeting for cancer chemotherapy^27,28 and cancer imaging.²⁹ The OVCAR-4 cell line (high-grade serous ovarian carcinoma) was also evaluated for several reasons; it has a high expression of α_vβ₃ integrin receptors with a reported value of [9.8 ± 2.5] × 10⁴ per cell,³⁰ strong affinity for cRGDfK peptides, and it was recently classified as a highly relevant model of clinical high-grade serous ovarian carcinoma.³¹ Lastly, the EMT-6 cell line (mouse breast carcinoma) was utilized as a negative control that lacks receptors for the RGD peptide motif.^32,33 Flow cytometry experiments using antibodies also verified that the EMT-6 cell line does not express β₅ proteins (Figure S13).

The six different SQ690 probes listed in Scheme 2 were incubated with integrin positive A549 cells at 1 μM for 30 min. Integrin receptor blocking was achieved by adding excess free cRGDfK peptide prior to cell incubation with the fluorescent probes. All three targeted SQ690 probes showed significant cell uptake in comparison to their respective untargeted control probes (Figure 1A). The near-infrared probe fluorescence within a set of micrographs was quantified as mean pixel intensity (MPI) to give the following order of probe uptake by the cells: T ⊃ SQ690-P3 > T ⊃ SQ690-P12 > T ⊃ SQ690-P45 (Figure 1B). The same trend was also observed with the integrin positive OVCAR-4 cells (Figure S14). While the T ⊃ SQ690-P3 probe produced the highest cell uptake, it could not be blocked by excess cRGDfK, and there was also significant cell uptake of the probe into integrin negative EMT-6 cells (Figure S15). Thus, the probe was deemed to undergo non-specific cell uptake due to its short PEG chain length and relatively hydrophobic structure. In contrast, the pre-assembled probes with longer PEG chains, namely, T ⊃ SQ690-P12 and T ⊃ SQ690-P45, displayed selective cell targeting properties. That is, uptake into the integrin positive cells was blocked by the presence of excess cRGDfK peptide and there was minimal cell uptake of the two probes into integrin negative EMT-6 cells. Since the probe with the medium PEG chain length, T ⊃ SQ690-P12, displayed higher integrin mediated cell endocytosis, it was chosen for in vivo SAI using a subcutaneous A549 tumor mouse model.

Figure 1.

Targeting integrin positive A549 cells. (A) Representative epifluorescence micrographs showing probe uptake by A549 cells after a 30-min incubation: 1 μM C ⊃ SQ690 (untargeted), 1 μM T ⊃ SQ690 (targeted), or 1 μM T ⊃ SQ690 + 200 μM cRGDfK (targeted + block). Blue fluorescence shows Hoechst nuclear stain. For each row, the calibration bar for near-infrared fluorescence intensity is in the upper right corner where units are arbitrary. Scale bar = 30 μm. (B) Quantification of intracellular mean fluorescence intensities as a measure of probe internalization. The threshold p-values are: * p<0.05, ** p<0.01, *** p<0.001.

Single Agent Imaging (SAI) of Mouse Tumor Model.

The in vivo cancer imaging capabilities of two pre-assembled probes, targeted T ⊃ SQ690-P12 and untargeted C ⊃ SQ690-P12, were compared using a subcutaneous tumor mouse model. There were two separate experimental goals. One was to determine if the targeting units in T ⊃ SQ690-P12 were enhancing tumor accumulation compared to the untargeted C ⊃ SQ690-P12. The second goal was to determine the suitability of these probes for fluorescence-guided surgery where there are three highly desired probe performance benchmarks; (a) high probe accumulation in the tumor with low background signal in surrounding tissue, (b) complete and rapid clearance of background unbound probe from the blood stream (<2 hours after dosage), and (c) extensive probe clearance from the blood through the kidneys with very low non-specific retention in other major organs.

A549 cells mixed with Matrigel (1:1) were subcutaneously injected into the right rear flank of Foxn1 nude mice. Approximately 4 weeks later, the tumor-burdened mice were randomly divided into two cohorts where each mouse received an intravenous injection of either T ⊃ SQ690-P12 or C ⊃ SQ690-P12 (10 nmol). The workflow of the SAI experiment is summarized in Figure 2A. Shown in Figure 2B is a representative montage of live mouse fluorescence images that were acquired using an in vivo imaging station and a Cy5.5 filter set. Both sets of fluorescence images show probe accumulation in the tumor and also in the kidneys. After the 3 h time point the mice were sacrificed, and a mock surgery procedure was performed on the cohort that had been dosed with T ⊃ SQ690-P12. The skin tissue covering the tumor was removed, and the surgical field was imaged to reveal the fluorescent tumor which was completely excised (Figure 2C). Subsequently, all tumors (Figure S16) and major organs were removed and imaged. The images were analyzed to determine relative Mean Pixel Intensity (MPI) due to the presence of fluorescent probe. Shown in Figure 2D are plots of MPI for each excised tumor normalized to the muscle. For the cohort dosed with untargeted C ⊃ SQ690-P12 and sacrificed after 3 h (N= 6) the average tumor-to-muscle ratio was 2.7 ± 0.4. For the cohort dosed with targeted T ⊃ SQ690-P12 and sacrificed after 3 h (N= 10), the average tumor-to-muscle ratio was 4.2 ± 0.7 which is above the benchmark of 3 that is needed for effective fluorescence-guided surgery.³⁴ Both probes cleared primarily through the kidneys (Figure S17) with the targeted T ⊃ SQ690-P12 probe showing more accumulation in the liver and spleen, which is often observed with peptide-based probes.³⁵

Figure 2.

(A) Schematic of Single Agent Imaging (SAI) procedure. (B) Representative fluorescent images of tumor-burdened mice at −5 min, 1.5, and 3 h after intravenous injection of either C ⊃ SQ690-P12 (untargeted probe) or T ⊃ SQ690-P12 (targeted probe) (10 nmol/mouse). (C) Mock surgery images of tumor-burdened mouse 3 h post-injection of T ⊃ SQ690-P12 (targeted probe). Mouse images were taken initially (before surgery), after the skin surrounding the tumor was removed (during surgery), and after the tumor was fully excised (after surgery). (D) Plot of MPI for excised tumors normalized to thigh muscle from the same mouse treated with a single probe and sacrificed at 3 h or 6 h post probe injection. Averages for each cohort is indicated by a black line, with error bars indicating ±SEM.

The plot in Figure 2D shows scattered tumor-to-muscle ratios for the targeted probe, T ⊃ SQ690-P12, in mice sacrificed after 3 h, and it was not possible to conclude with statistical significance if the targeting units in T ⊃ SQ690-P12 were enhancing tumor accumulation compared to the untargeted C ⊃ SQ690-P12. Thus, an additional mouse tumor imaging experiment (N= 4) was conducted using T ⊃ SQ690-P12, but with a longer time point of 6 h before sacrifice to allow further clearance of the unbound probe and a lower background signal. As shown in Figure 2D, this new data set with T ⊃ SQ690-P12 (6 h) produced an average tumor-to-muscle ratio of 4.9 ± 0.8 and a p-value of 0.027 when it was compared to the data set for C ⊃ SQ690-P12, indicating the difference to be statistically significant.

Paired Agent Imaging (PAI) of Mouse Tumor Model.

The tumor-to-muscle ratio plots from the SAI experiments (Figure 2D) show moderate scatter. This is not surprising, as it is known that the rate and extent of probe accumulation within a tumor depends on physiological variables such as tumor microvessel density,³⁶ permeability of the tumor vasculature,³⁷ intratumoral pressure,³⁸ and tumor growth rate.¹² Thus, the animal cohort size for SAI experiments must be relatively large to ensure statistical averaging of the inherent differences in tumor EPR effect between the cohorts. One way to account for this experimental variability is to conduct a PAI experiment; that is, co-injection of a binary mixture of targeted and untargeted probes into the same mouse. A requirement of the fluorescent PAI experiment is that the two probes have distinguishable emission wavelengths that permit quantitative two-color imaging. In this regard, squaraine dyes are well suited for multicolor fluorescence imaging because they exhibit intense and narrow absorption/emission bands. Moreover, we recently reported SQ830-P45 as a new dye component to make pre-assembled fluorescent probes that are complementary imaging partners to pre-assembled probes that incorporate SQ690-P45.²² We reasoned that the untargeted pre-assembled probe C ⊃ SQ830-P45 and the targeted pre-assembled probe T ⊃ SQ690-P45 were good choices for a PAI experiment (see Table 1 for spectral properties). Appended to each dye were two long PEG-45 chains to ensure that both pre-assembled probes exhibited very similar brightness and pharmacokinetics. Thus, non-specific uptake of both probes from the blood stream into the tumors should be the same, and any enhanced accumulation of the targeted probe can be attributed to binding of integrin receptors within the tumor tissue by the appended cRGDfK peptides on the targeted probe. The two probes were pre-assembled using Synthavidin Technology (Scheme 1) and phantom imaging experiments on the imaging station (Figure S19) showed that separate samples of the two probes could be selectively imaged using the following filter channels (ex: 745 nm, em: 850 nm) and (ex. 640 nm, em: 710 nm).

A schematic summary of the PAI experiment is shown in Figure 3A. In short, a cohort of five mice bearing subcutaneous A549 tumors were each injected intravenously with a binary 1:1 mixture of T ⊃ SQ690-P45 and C ⊃ SQ830-P45 (20 nmol each). Each living mouse was imaged at 1.5, 3 and 5 h after dosing, with the SQ690 filter monitoring biodistribution of targeted probe and the SQ830 filter monitoring biodistribution of untargeted probe. The live animal images in Figure S20 show that both probes accumulated in the tumor and were cleared through the kidneys. At 5 h after dosing, the mice were sacrificed; the tumors and organs were excised and imaged with both filter channels to give false-colored image maps (Figure 3B) that enabled determination of tumor-to-muscle ratios. Shown in Figure 3C are plots of the tumor-to-muscle ratios for each probe in the tumors from each of the five mice. It is important to emphasize that the plot of tumor-to-muscle ratios for targeted T ⊃ SQ690-P45 is equally as scattered as the plot in Figure 2D (as expected), but now the analysis has much greater statistical certainty because each value of tumor-to-muscle ratio for T ⊃ SQ690-P45 is directly paired with the corresponding valued determined for untargeted C ⊃ SQ830-P45 in the same tumor. The pairing provides a correction factor that accounts for the tumor-to-tumor variability in non-specific probe uptake. In other words, both the SAI and PAI experiments support the conclusion that there is increased tumor accumulation of targeted T ⊃ SQ690-P45 compared to untargeted C ⊃ SQ830-P45, but the PAI experiment has much higher statistical significance and used a much smaller cohort of mice.

Figure 3.

(A) Schematic of Paired Agent Imaging (PAI) procedure (B) Near-infrared fluorescent images of excised tumors that were harvested at 5 h after intravenous co-injection of C ⊃ SQ830-P45 (untargeted probe) and T ⊃ SQ690-P45 (targeted probe) (20 nmol/mouse). Each tumor was imaged using two different filter sets. Top: C ⊃ SQ830-P45; Bottom: T ⊃ SQ690-P45. (C) Plot of MPI for excised tumors normalized to thigh muscle from the same mouse, black line connects the same tumor.

The general concept of PAI (sometimes called dual-reporter imaging) has been known for many decades and has been adapted for several imaging modalities beyond fluorescence.^10,39,40,41 Furthermore, the leading research groups in PAI have developed image analysis methods that determine a targeted probe’s binding potential (BP) for a specific biomarker in a Region of Interest (ROI).^8,10,14 The BP is a dimensionless value whose magnitude is determined by at least two parameters, the probe’s affinity for the biomarker and the abundance of the biomarker within the ROI. In this present study, we employed the single time point analysis method of Tichauer and coworkers to determine the average BP of T ⊃ SQ690-P45 for integrin receptors within each sample of excised tumor and muscle tissue.³⁹ This determination treats each tissue sample as a single ROI (see Experimental section for details). As shown in Figure 4A, the average BP for the five tumors is close to one (0.9 ± 0.1) indicating overexpression of integrin receptor, whereas the average BP value for corresponding muscle tissue is close to zero (−0.03 ± 0.06) which is expected since there is no overexpression of integrin receptor in muscle. A higher resolution version of this image analysis method is provided in Figure 4B, which shows a pixelated BP map for each of the five excised tumors. These maps clearly indicate a heterogenous distribution of BP values. So not only is there moderate tumor-to-tumor variation in the average BP (Figure 4A), but there is also substantial spatial variation of BP within each specific tumor (Figure 4B). Assuming that the affinity of T ⊃ SQ690-P45 for integrin receptors, and the extent of probe internalization after integrin binding, does not change with spatial location within the tumor, we can conclude that the cell surface expression of integrin receptors within an individual tumor is spatially heterogeneous and varies by about a factor of ten. This finding agrees with the general picture of heterogeneous cell populations within tumors that have differing receptor expression levels,^{4,10,42,43,44} and the more specific picture of a heterogeneous distribution of α_vβ₅ and α_vβ₃ integrin receptor levels throughout A549 tumors.⁴⁵ There are multiple factors driving integrin receptor heterogeneity in tumor tissue which has angiogenic regions with high integrin expression and necrotic regions with low integrin expression.^3,46,47,48

Figure 4.

(A) Plot of T ⊃ SQ690-P45 Binding Potential (BP) for each sample of excised tumor and muscle tissue harvested at 5 h after intravenous co-injection of C ⊃ SQ830-P45 (untargeted probe) and T ⊃ SQ690-P45 (targeted probe) (20 nmol/mouse). (B) Pixel map showing the heterogenous distribution of T ⊃ SQ690-P45 BP within each of the excised tumors.

Conclusions.

Tumor targeted fluorescent molecular probes are attractive for in vivo imaging because they can be employed at low doses, which minimizes toxicity, while producing high levels of tumor accumulation, which maximizes image contrast. While the concept of a targeted molecular probe is easy to comprehend, the current experimental processes for evaluating and optimizing imaging performance are tedious, and there is a need for improved probe development strategies using non-covalent click chemistry.⁵⁰ In this current study, a novel non-covalent assembly process called Synthavidin Technology was employed to fabricate seven fluorescent probes with modular differences in three critical assembly components: the fluorochrome, targeting unit(s), and pharmacokinetic controller. Targeted probes were prepared using the macrocycle T which was equipped with two copies of cRGDfK peptide which has high affinity for cancer cells that overexpress integrin receptors. Untargeted probes were prepared using the macrocycle C whose structure has two appended tricarboxylate units with no cancer targeting potential. Probe imaging performance was optimized by an iterative cycle of cell and mouse imaging studies that provided feedback for structural revision of each probe component. A set of Single Agent Imaging studies compared tumor imaging performance of a targeted probe and untargeted probe in separate mouse cohorts. Although there was imaging evidence for enhanced tumor accumulation of the targeted probe compared to the untargeted probe, the plots of tumor-to-muscle ratio showed moderate scatter likely due to tumor-to-tumor variability of the vasculature structure and interstitial pressure. A total of twenty mice were sacrificed before statistical significance was obtained. In contrast, a Paired Agent Imaging (PAI) study, that used a co-injected binary mixture of untargeted probe (C ⊃ SQ830-P45) and targeted probe (T ⊃ SQ690-P45), reached the same conclusion but with higher statistical significance and only sacrificed five mice. The potential of the targeted probe for fluorescence-guided surgery was demonstrated by a mock surgery that completely removed the fluorescently labeled tumor. The results of this study show that the combination of fluorescent probe pre-assembly and PAI is promising new way to rapidly develop targeted fluorescent probes for tumors with high Binding Potential. Furthermore, image analysis quantified the heterogenous spatial distribution of integrin receptors within an individual tumor. Looking to the immediate future, the combination of probe pre-assembly using Synthavidin technology and PAI can be used to further enhance binding potential by producing and evaluating pre-assembled fluorescent targeted probes with higher orders of multivalency.¹⁹ From a broader perspective, PAI is emerging as an effective method to visualize and quantify cell receptor heterogeneity in living and excised tissues, and thus can be used to facilitate pre-clinical research and clinical implementation of targeted therapy, image guided surgery, and personalized medicine.^10,14

EXPERIMENTAL SECTION

Synthesis and Preparation of Stock Solutions.

Synthetic methods and structural characterization (¹H NMR and mass spectrometry) for squaraines SQ690-P3, SQ690-P12, SQ690-P45, and SQ830-P45 and macrocycles C and T are provided in the Supporting Information. Briefly, each bis-alkyne squaraine dye underwent copper-catalyzed azide-alkyne cycloaddition (CuAAC) with azide-modified poly(ethylene glycol) chains to produce the water-soluble squaraines. Similarly, the bis-alkyne tetralactam precursor was also modified by CuAAC to append either a water-solubilizing tricarboxylate unit or a cRGDfK peptide. The 1 mM stock solutions of squaraines and macrocycle C were prepared in deionized water, and a 10 mM stock solution of T was prepared in dimethyl sulfoxide. Each squaraine was diluted to 3.0 μM in deionized water in a 1-mL quartz cuvette, and its absorption and fluorescence spectra was recorded. One molar equivalent of C or T was then added, and the absorption and fluorescence spectra collected again. Complexation was confirmed by a red shift in absorption and emission maxima as well as a six-fold increase in fluorescence intensity. Analysis by gel electrophoresis provided independent confirmation that each stock solution of pre-assembled probe was pure with no evidence of probe unthreading. Previous studies have shown that the pre-assembled probes remained complexed during cell and animal imaging experiments.¹⁹ Even if there is a small amount of unthreading during animal imaging, the blue-shifted and weakly fluorescent SQ690-P12, SQ690-P45, or SQ830-P45 dyes that are produced by probe unthreading cannot be detected (Figures S18, S22 and S23), and thus do not affect the imaging results. Furthermore, stability studies with C ⊃ SQ690-P45 in 50% fetal bovine serum showed no unthreading over 24 h and cell microscopy studies showed no evidence for loss of probe signal in the lysosome due to unthreading.

Cell Microscopy.

The A549 cells were grown in F-12K medium (ATCC) supplemented with 10% fetal bovine serum (Atlanta Biologicals) and 1% penicillin/streptomycin (Sigma Aldrich). The A549 cells were seeded and grown to 70% confluency on an 8-well chambered coverglass (Lab-Tek, Nunc, USA). The cells were incubated with 1 μM probe in media for 30 min at 37℃. For blocking, 200 μM excess cRGD was added to the cells for 5 min prior to probe treatment and remained present during probe incubation. The cells were washed three times with 1xPBS, fixed with 4% cold paraformaldehyde for 20 min at room temperature, washed again with 1xPBS, and co-stained with 3 μM Hoechst 33342 for 10 min. Afterwards, the cells were washed two times with 1xPBS and imaged on a Zeiss Axiovert 100 TV epifluorescence microscope equipped with UV filter (Ex. 387/11, Em. 447/60) and Cy5.5 filter (Ex. 655/40, Em. 716/40). For each micrograph, a background subtraction with a rolling ball radius of 200 pixels was conducted using ImageJ2 software. The average mean fluorescence intensity for each micrograph was then calculated from using 20 randomly generated 25 × 25 pixel extra-nuclear ROIs. The averages and SEM were calculated and plotted in GraphPad Prism.

In vivo Imaging.

Single Agent Imaging.

All animal experiments were conducted under protocols that were approved by the Notre Dame Institutional Animal Care and Use Committee. Female Foxn1 nude mice (N = 16) were inoculated with A549 cells (1 × 10⁶) in 1:1 Matrigel (corning):media on the right rear flank. Approximately 4 weeks later, the mice were divided into two cohorts and received a retro-orbital injection of either T ⊃ SQ690-P12 (N= 10) or C ⊃ SQ690-P12 (N= 6) (100 μL, 10 nmol/mouse). Five min prior to injection, each mouse was anesthetized with 2–3% isoflurane with an oxygen flow rate of 2 L min⁻¹ and imaged using an in vivo imaging station (Xenogen IVIS) (Filter: Cy5.5, Acquisition time: 3 s, Binning: small, F-stop: 2, Field-of-view 10 × 10 cm). The mice were then injected with probe and imaged at 0, 1.5, and 3 h. At 3 h, the mice were anesthetized and sacrificed via cervical dislocation, and blood was collected from the heart. To simulate a surgery, the skin surrounding the tumor was cut and the mice were imaged. The tumor was then fully removed, and the mice were imaged again. Afterwards, all major organs were removed, including the liver, lungs, heart, spleen, kidneys, and muscle. The organs were imaged and analyzed using ImageJ2. Image processing of full-body and tumor images was conducted using a background subtraction with a rolling ball radius of 1000 pixels. The maximum fluorescence was set to 500 and the images were pseudocolored “fire.” Biodistribution analysis was performed by importing images of excised organs from each cohort into ImageJ2. A background subtraction was applied with a rolling ball radius of 1000 pixels. The amount of probe in each organ was obtained by adjusting the color threshold to create a region of interest (ROI) around each organ. The mean pixel intensity (MPI) of each selected ROI was measured and graphed to show organ fluorescence relative to the muscle. The above protocol was repeated again with additional mice inoculated with A549 tumors. The mice received a retro-orbital injection of T ⊃ SQ690-P12 (N= 4) (100 μL, 10 nmol/mouse), and the experiment was carried out to 6 h.

Paired Agent Imaging.

All animal experiments were conducted under protocols that were approved by the Notre Dame Institutional Animal Care and Use Committee. Female Foxn1 nude mice (N = 16) were inoculated on the right rear flank with a 1:1 mixture of A549 cells (1 × 10⁶) and Matrigel. Approximately 4 weeks later, the mice received a retro-orbital co-injection of a binary mixture of T ⊃ SQ690-P45 and C ⊃ SQ830-P45 (N= 5,100 μL, 20 nmol/mouse). Five min prior to injection, each mouse was anesthetized with 2–3% isoflurane with an oxygen flow rate of 2 L min⁻¹ and imaged using an in vivo imaging station (Ami HT Spectral Instruments Imaging) with two channels (ex: 640 nm, em: 710 nm, exposure: 3 s, percent power: 50%, F-stop: 2, binning: small) and (ex: 745 nm, em: 850 nm, exposure: 3 s, percent power: 30%, F-stop: 2, binning: small)]. The living mice were then injected with the probe mixture and imaged at 0, 1.5, 3, and 5 h. At 5 h, the mice were anesthetized and sacrificed via cervical dislocation, and blood was collected from the heart. Afterwards, all major organs were removed, including the liver, lungs, heart, spleen, kidneys, muscle, and tumor. The organs were placed on a transparent imaging tray and imaged with two channels (ex: 640 nm, em: 710 nm, exposure: 3 s, percent power: 50%, F-stop: 2, binning: small) and (ex: 745 nm, em: 850 nm, exposure: 3 s, percent power: 30%, F-stop: 2, binning: small)] with minimal signal crosstalk (Figure S19). It is assumed that the fluorescence intensity in each channel is proportional to the respective probe concentration in each tissue due to the similar optical properties at 690 and 830 nm in most tissues.⁴⁹ Most of the image analysis compares fluorescence intensities from separate samples of the same tissue, which eliminates tissue dependent differences in optical properties.⁴³ Image processing using ImageJ2 software was conducted using a background subtraction with a rolling ball radius of 1000 pixels. The maximum fluorescence was set to 1000 and the images were pseudocolored “fire.” Biodistribution analysis was performed by importing images of excised organs from each cohort into ImageJ2. The amount of probe in each organ was obtained by creating a region of interest (ROI) around each tumor or excised organ. The mean pixel intensity (MPI) of each tumor was measured and divided by the MPI for thigh muscle from the same animal to give the tumor-to-muscle ratio. The average Binding Potential (BP) value for each tumor and muscle sample was determined by the single time point PAI method^14,51,43 which uses the following equation:

$$\text{BP}=\frac{{\text{ROI}}_{T\mathbf{\supset }SQ690\text{-}P45}-{\text{ROI}}_{C\mathbf{\supset }SQ830\text{-}P45}}{{\text{ROI}}_{C\mathbf{\supset }SQ830\text{-}P45}}$$

In the case of Figure 4A, the ROI is the entire tumor or muscle sample. In the case of Figure 4B, the ROI is a spatially defined pixel within a tumor image, and the overall result is a pixelated map of the BP distribution throughout the tumor. All image analyses were completed using the image calculator in ImageJ2 software.

Statistical Analysis.

For cell microscopy and in vivo SAI, an unpaired t test was employed using GraphPad Prism software. For in vivo PAI, a paired t test was employed using the same software. Statistical significance between groups/cohorts was defined as a p-value less than 0.05, and threshold p-values assigned in the following manner: * p<0.05, ** p<0.01, *** p<0.001.

ABBREVIATIONS

BP

Binding Potential

EPR

Enhanced Permeation Retention

MPI

Mean Pixel Intensity

PAI

Paired Agent Imaging

SAI

Single Agent Imaging