Fluorochrome Selection for Imaging Intraoperative Ovarian Cancer Probes

The identification and removal of all gross and microscopic tumor to render the patient disease free represents a huge challenge in ovarian cancer treatment. The presence of residual disease is an independent negative prognostic factor. Herein, we describe the synthesis and the “in vitro” evaluation of compounds as cyclooxygenase (COX)-1 inhibitors, the COX-1 isoform being an ovarian cancer biomarker, each bearing fluorochromes with different fluorescence features. Two of these compounds N-[4-(9-dimethylimino-9H-benzo[a]phenoxazin-5-ylamino) butyl]-2-(3,4-bis(4-methoxyphenyl)isoxazol-5-yl)acetamide chloride (RR11) and 3-(6-(4-(2-(3,4-bis(4-methoxyphenyl)isoxazole-5-yl)acetamido)butyl)amino-6-oxohexyl)-2-[7-(1,3-dihydro-1,1-dimethyl-3-ethyl 2H-benz[e]indolin-2-yl-idene)-1,3,5-heptatrienyl]-1,1-dimethyl-3-(6-carboxilato-hexyl)-1H-benz[e]indolium chloride, 23 (MSA14) were found to be potent and selective inhibitors of cyclooxygenase (COX)-1 “in vitro”, and thus were further investigated “in vivo”. The IC50 values were 0.032 and 0.087 µM for RR11 and 23 (MSA 14), respectively, whereas the COX-2 IC50 for RR11 is 2.4 µM while 23 (MSA14) did not inhibit COX-2 even at a 50 µM concentration. Together, this represented selectivity index = 75 and 874, respectively. Structure-based virtual screening (SBVS) performed with the Fingerprints for Ligands and Proteins (FLAP) software allowed both to differentiate highly active compounds from less active and inactive structures and to define their interactions inside the substrate-binding cavity of hCOX1. Fluorescent probes RR11 and 23 (MSA14), were used for preliminary near-infrared (NIR) fluorescent imaging (FLI) in human ovarian cancer (OVCAR-3 and SKOV-3) xenograft models. Surprisingly, a tumor-specific signal was observed for both tested fluorescent probes, even though this signal is not linked to the presence of COX-1.


Introduction
Ovarian cancer (OC) is the most aggressive gynecological cancer, and the 5-year survival rate is still <50%. This is due to delayed diagnoses, development of chemoresistance, and tumor evasion of the host's immune responses.
OC of epithelial origin (EOC) constitutes 90% of all OC histological subtypes diagnosed. The remaining 10% are of germinal and stromal origin. Most patients with newly diagnosed EOC receive the same standard treatment, which consists of cytoreductive surgery together with adjuvant therapy consisting of first-line platinum-based

Rationale behind the Design of the Novel Target Compounds
Novel compounds have been designed to target the COX-1 [12], especially in tumor cells and tissues where it is overexpressed [13][14][15][16][17]. From a chemical viewpoint, the molecules must carry at least three moieties: a mofezolac unit recognized through its carboxylic group by COX amino acid residues (Arg120, Tyr355, and Glu524) located at the entrance of the enzyme long hydrophobic channel that has the catalytic site on its top [18], a linker and a fluorochrome. Several fluorochromes were used to uncover the one endowed with the highest Stokes shift. Mofezolac was selected as the most potent and selective COX-1 inhibitor used in humans [19]. Flexible and rigid linkers were used to join mofezolac and the selected fluorochrome.

Chemistry
Most of the reaction conditions used to prepare the herein described target compounds for "in vitro", "ex vivo", and "in vivo" COX-1 detection use coupling reagents to form amide bonds. In other words, the carboxylic acids are "in situ" activated by the addition of HOBt to the reaction mixture to form an ester, before being mixed with the proper amine in the presence of the coupling reagent such as the EDC.

Fluorescent Properties
The λabs and λem, of the target compounds were measured in aqueous solution (phosphate-buffered saline solution, PBS) at the concentration of 10 −5 M (Table 1).

Fluorescent Properties
The λ abs and λ em , of the target compounds were measured in aqueous solution (phosphate-buffered saline solution, PBS) at the concentration of 10 −5 M (Table 1). The dansyl-bearing compounds 3a-c, 6a-b and 9 displayed λ ex~2 30-360 nm and λ em in the range 492-514 nm, with a fairly large Stokes shift.
NBD-bearing compounds 13a-c showed λ abs between 482 and 495 nm and λ em ranging from 550 to 564 nm, with the Stokes shift decreased, probably due to its well-known solvatochromic properties.
Compound 17a, bearing the naphthalimide fluorophore, displayed λ abs = 450 nm and λ em = 546 nm. The excitation and emission wavelength values of 17b were measured in EtOH because of its low solubility in PBS. Despite of the solvent, the Stokes shift was very low with this fluorophore.
Overall, the measured λ abs and λ em for all the novel fluorescent molecules were in accordance with the properties conferred by the insertion of the specific fluorescent moiety. The fluorescent dyes used herein are known to be "environment sensitive", since they display sensitivity to the local environment polarity. These pharmacodynamic properties combined with the optimal fluorescent spectroscopic properties, such as good Stokes shift and high fluorescence emission in aqueous and EtOH, make compounds 23 (MSA 14) and RR11 [11] valuable tools to investigate COXs as targets "in vitro" and "in vivo" as well.

Cyclooxygenase Catalytic Activity Inhibition Evaluation of the Novel Target Compounds
The novel fluorescent compounds were tested to evaluate their inhibitory activity and selectivity towards cyclooxygenases ( Table 2) [26]. Regarding 3a-c and 6a-b, the presence of dansylsulfonamide moiety, as fluorophore, does not determine a change in the selectivity towards the cyclooxygenase isoforms. They remain selective COX-1 inhibitors. Additionally, 3a is almost inactive as an inhibitor towards both COXs. The nature of the linker does not affect the inhibitory potency of 3b-c (COX-1 IC 50 = 0.05 and 0.06 µM, respectively); it is important in 6a-b (COX-1 IC 50 = 12.0 and 0.1 µM, respectively). In fact, 3a, 6a-b showed an increase in potency as the length of the alkyl chain became longer, passing from an almost inactive 3a, which determines only 45% COX-1 inhibition, to 6a, which displays a 74% inhibition with a linker of six methylenes and a COX-1 IC 50 value equal to 12 µM. The potency increases by an order of magnitude as the straight chain reaches the length of 12 carbon atoms (6b) (IC 50 = 0.1 µM). In 6b, the linker (twelve methylenes) confers a suitable flexibility to determine a greater inhibition potency, although the 55% of inhibition at compound concentration of 50µM is still low. The corresponding molecules with rigid linkers, phenyl (3b), and a benzidine (3c) showed a potency gain of an order of magnitude. The best result in terms of inhibition was obtained with 6b (COX-1 IC 50 = 0.06 µM and 71% inhibition). Bearing in mind that mofezolac (4), the most potent and selective COX-1 inhibitor, has COX-1 IC 50 = 0.0079 µM [27][28][29], it seems that the presence of the dansyl moiety does not affect the selectivity of all new compounds because they remain as a COX-1 selective inhibitor, but with a reduced inhibitory potency. Among dansyl-bearing derivatives, 3c remains the most potent inhibitor. When the fluorophore is bound to 3b through a flexible linker with ten carbon atoms (9), the percentage of inhibition remains high (83%) but with a lower potency (IC 50 = 0.7 µM). Unlikely dansyl-bearing compounds, the length of the linker does not determine any variation of the inhibitory activity of the NBD-bearing derivatives. On the contrary, the presence of this fluorophore completely cancels also the inhibitory activity induced by mofezolac. Specifically, all compounds 13a-c result as inactive and non-selective. For the naphtylimide derivatives 17a-b, a recovery of the selectivity in favor of the COX-1 isoform (IC 50 = 0.08 µM and 0.15 µM for 17a and 17b, respectively) was observed, and was almost inactive towards COX-2 at 50µM final concentration.
In the dansyl derivatives 3a-c and 6a-b and naphthylamide derivatives 17a-b, the dansylsulfonamide and naphthylimide carbonyls confer selectivity and potency towards COX-1, probably due to a certain electronic availability that allows them to interact with COX-1 active site, as revealed by the inspection of the binding poses in Fingerprints for Ligands and Proteins (FLAP) modeling studies. The cyanine-bearing compound 23 is still a potent and selective COX-1 inhibitor, with an IC 50 = 0.087µM and 57% of inhibition, and is almost inactive towards COX-2 at 50µM final concentration.

Computational Studies Based on the Fingerprints for Ligands and Proteins (FLAP) Algorithm
To ascertain the specific interactions established by the new fluorescent inhibitors with amino acids residues forming the hCOX-1 binding site, structure-based virtual screening (SBVS) with the Fingerprints for Ligands and Proteins (FLAP) software [30] employing the recently published crystal structure of human cyclooxygenase (hCOX)-1 (PDB ID: 6Y3C) as template, [31] was performed.
In the first step of the computational study, the protein cavities (pockets) were calculated. Accordingly, five internal cavities in the structure of hCOX-1 have been determined. As reported in Figure 1, it was possible to identify the substrate-binding site (P1) and the heme catalytic site (P3), in agreement with the analysis of hCOX-1 crystal structures [31] [further details are in the Supporting materials ( Figure S1)].
The SBVS approach allowed us to generate binding poses of the newly synthesized compounds in the substrate-binding cavity (P1) based on the similarity between their GRID fields; the FLAP binding pose for the reference compound RR11 was also calculated in the new protein template and used as a reference ( Figures S2 and S3). The FLAP calculated scores are reported in Table 3; the global similarity score (Glob-Sum) was used to rank the inhibitors. The SBVS approach allowed us to generate binding poses of the newly synthesized compounds in the substrate-binding cavity (P1) based on the similarity between their GRID fields; the FLAP binding pose for the reference compound RR11 was also calculated in the new protein template and used as a reference ( Figures S2 and S3). The FLAP calculated scores are reported in Table 3; the global similarity score (Glob-Sum) was used to rank the inhibitors. Overall, the results reported in Table 3 shows that the SBVS procedure allows us to differentiate highly active compounds (IC50 < 0.1 µM, Glob-Sum > 3.0) from less active and inactive structures, the only exception being compound 13a. It is noteworthy that structurally related compounds showing different inhibitor activity, i.e., 3a and 3b, gave  Overall, the results reported in Table 3 shows that the SBVS procedure allows us to differentiate highly active compounds (IC 50 < 0.1 µM, Glob-Sum > 3.0) from less active and inactive structures, the only exception being compound 13a. It is noteworthy that structurally related compounds showing different inhibitor activity, i.e., 3a and 3b, gave reasons for different Glob-sum values, confirming the good correlation between this global parameter and the trend observed for the inhibitor activity.
A close inspection at the binding poses, reported in Figures 2a-d and S4-S7, for the most active compounds-namely 3b-c and 17a-b-revealed some common features: (i) The fluorophore moieties are always located at the bottom of the P1 cavity and gave reasons for the strongest CH-π, S-π and π-π interactions with Tyr385, Trp387, Met522 (ii) The mofezolac units are placed in the outer region of the binding site realizing strong CH-π interaction with the external parallel α-helixes and strong π-π and H-bonds with Tyr 355 and Arg120, respectively, and lower cavity walls stabilize the inhibitors in the central region of the binding cavity.
global parameter and the trend observed for the inhibitor activity. A close inspection at the binding poses, reported in Figures 2a-d and S4-S7, for the most active compounds-namely 3b-c and 17a-b-revealed some common features: (i). The fluorophore moieties are always located at the bottom of the P1 cavity and gave reasons for the strongest CH-π, S-π and π-π interactions with Tyr385, Trp387, Met522 (ii). The mofezolac units are placed in the outer region of the binding site realizing strong CH-π interaction with the external parallel α-helixes and strong π-π and H-bonds with Tyr 355 and Arg120, respectively, and lower cavity walls stabilize the inhibitors in the central region of the binding cavity. Amongst the active compounds, the cyanine derivative 23 (Figure 2d) used the two fluorophore units for the interaction with both the entry and the bottom of the cavity; the mofezolac moiety showed negligible interaction, as it spanned behind the binding site but does not interfere with the heme catalytic site.
The binding pose obtained for 3a and the NBD derivatives exhibited a similar arrangement of the different substrate moieties. However, no significant interaction occurred between the fluorophore units and the hydrophobic region of the binding cavity.

Biodistribution of RR11 and 23 (MSA14) in OVCAR-3 Xenograft Models
Due to deeper tissue penetration of near-infrared (NIR) light, intraoperative imaging agents bearing NIR fluorophores are more valuable candidates for clinical translation than Amongst the active compounds, the cyanine derivative 23 (Figure 2d) used the two fluorophore units for the interaction with both the entry and the bottom of the cavity; the mofezolac moiety showed negligible interaction, as it spanned behind the binding site but does not interfere with the heme catalytic site.
The binding pose obtained for 3a and the NBD derivatives exhibited a similar arrangement of the different substrate moieties. However, no significant interaction occurred between the fluorophore units and the hydrophobic region of the binding cavity.

Biodistribution of RR11 and 23 (MSA14) in OVCAR-3 Xenograft Models
Due to deeper tissue penetration of near-infrared (NIR) light, intraoperative imaging agents bearing NIR fluorophores are more valuable candidates for clinical translation than the shorter-wavelength fluorescent ligands. Consequently, two fluorescent probes, RR11 and 23 (MSA14), were used for preliminary NIR FLI in OVCAR-3 bearing NSG mice.
To assess the capacity of both conjugates to visualize ovarian xenograft tumors "in vivo" and evaluate accumulation and clearance behavior of the conjugates, kinetic imaging was performed at six different time-points. Subcutaneously engrafted OVCAR-3 NSG mice were injected with 100 µL of 500 µM RR11 or 23 (MSA14) intravenously into the tail vein ( Figure 3). RR11 imaged mice (n = 2) showed a high non-specific background signal at the first imaging time-points, with high signal intensities in kidneys and gastroin-testinal tract ( §, ¥, respectively, Figure 3A). Tumor-specific fluorescence was obtained 4, 6, and 24 h after RR11 injection, with a mean tumor-to-background ratio (TBR) of 1.74 after 24 h. In OVCAR-3 bearing NSG mice (n = 3), injected with 23 (MSA14), high liver signals (ventral view, #) were obtained at all time points with high compound accumulation in the tumors at 24 h (*, Figure 3B). The best TBR of 1.44 was achieved after 24 h. Residual background fluorescence was predominantly observed in the spine and limbs as well as in the excretion organs. Following the biodistribution over time "in vivo", we assessed the accumulation of the compounds "ex vivo" at 6 h. In Figure 4A, RR11-induced tumor-specific signal, along with high fluorescence in kidneys, gallbladder, stomach, and lungs, can be observed. In contrast, "ex vivo" 23 (MSA14) FLI of tumor tissues and organs reveals distribution to kidneys, liver, spleen, and lungs with lower intensities in the stomach ( Figure 4A). The higher gastrointestinal fluorescence signal in RR11-imaged mice can derive from auto-fluorescent alfalfa containing diet, which is acquired in the lower wavelength of RR11. The main differences between the accumulation and clearance behavior are marked with an arrow, indicating two different clearance patterns of the compounds through liver (23, MSA14) and kidney (RR11). Finally, the fluorescence of 23 (MSA14) was evaluated in a clinically compatible intraoperative setting using the FLARE ® system to assess the suitability of the COX-1 targeted NIR compound for FIGS and its potential clinical translation ( Figure 4B). The "in vivo" and "ex vivo" intraoperative images confirm the results from Figures 3B and 4A. Tumor-specific accumulation was observed in OVCAR-3 s.c. models. While the tumor fluorescence intensities were higher than those of kidneys, the fluorescence intensities in the gallbladder and lungs were the highest.  Figure 5B). However, "ex vivo" analysis revealed high signal in lungs (at high dose) and spleen as observed in OVCAR-3 models. Intraoperative imaging showed superior tumor signal at high dose and 1000 ms exposure time after 24 h ( Figure 5C). However, the tumor fluorescence intensity was not sufficient to discriminate between healthy organs and malignant tissue at low exposure times of 100 ms "ex vivo". The fluorescence signal in COX-1 low-expressing SKOV-3 tumors might be an indication of non-specific compound accumulation. and background in the flank muscle (yellow square), is presented in one representative mouse. All animals were euthanized when indicated and "ex vivo" NIR FLI was performed. (C) 23 (MSA14) injected SKOV-3 mice were additionally imaged with the intraoperative imaging system "in vivo" and "ex vivo". NIR exposure times are indicated in brackets.

Materials and Methods
1 H NMR and 13 C NMR spectra were recorded on a Bruker 600 MHz or AGILENT 500 MHz spectrometer and chemical shifts are reported in parts per million (δ). The following abbreviations were used to explain the multiplicities: s-singlet; d-doublet; t-triplet; q-quartet; m-multiplet; quin-quintuplet; sext-sextet; sep-septet; b-broad. FT-IR spectra were recorded on a PerkinElmer 681 spectrophotometer. GC analyses were performed on a HP 6890 model, Series II, using a HP1 column (methyl siloxane; 30 m × 0.32 mm × 0.25 µm film thickness). Analytical thin-layer chromatography (TLC) was carried out on precoated 0.25 mm thick plates of Kieselgel 60 F254; visualization was accomplished by UV light (254 nm). Column chromatography was accomplished by using silica gel 60 with a particle size distribution 40-63 µm and 230-400 ASTM. GC-MS analyses were performed on HP 5995C model. High-resolution mass spectrometry (HRMS) analyses were performed using a Bruker microTOF QII mass spectrometer equipped with an electrospray ion source (ESI). Reagents and solvents were purchased from Sigma-Aldrich AcOEt/hexane = 9/1 as mobile phase for 3a and CHCl 3 /CH 3 OH = 9:1 as mobile phase for 3b-c) and washed with aqueous 10% NaHCO 3 (3 × 10 mL). Then, the solvent was removed under reduced pressure. The resulting brown semisolid underwent chromatography on silica gel and AcOEt/hexane = 9/1 as mobile phase for 3a and CHCl 3 /CH 3 OH = 9.5:0.5 as mobile phase for 3b-c.

Cytotoxicity Study
Determination of SKOV-3 and OVCAR-3 cell growth was performed using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay [34]. On day 1, 20,000 cells/well were seeded into 96-well plates at a volume of 50 µL. On day 2, 50 µL of the various drug concentrations was added. In all the experiments, the drug solvent (DMSO) was added to each control to evaluate solvent cytotoxicity. After 48 h incubation time with drugs, MTT (10 µL, 0.5 mg/mL)) was added to each well, and after 3-4 h incubation at 37 • C, the supernatant was removed. The formazan crystals were solubilized using 100 µL of DMSO/EtOH (1:1), and the absorbance values at λ = 570 nm were determined on the Tecan Infinite 200 Microplate Reader.

Cyclooxygenase Activity Inhibition
The COX-1/2 inhibition screening experiments were carried out according to the manufacturer's instructions. The plate was incubated for 60 min at r.t. under shaking and read with the Tecan Infinite 200 Microplate Reader at λ = 405 nm.
The graph between percent inhibition of the enzyme and corresponding concentration of the compound provided IC 50 values for the probe compounds.

Subcutaneous Xenograft Models of Ovarian Cancer
Female NOD-scid IL2rγ null (NSG) mice, aged 6 to 25 weeks were maintained under defined floral conditions and no more than five mice were housed in individually ventilated (HEPA-filtered air) cages, kept on a 12 h dark/night schedule at a constant temperature of 21 • C and at 50% relative humidity at the University of Bergen's animal facility. Observations of body weight and general condition including activity levels, appearance, and food intake were monitored twice weekly, and humane endpoints were defined with the use of score sheets. When indicated, animals were euthanized according to institutional guidelines.
3.9.5. Fluorescence Imaging (FLI) "In vivo" and "ex vivo" FLI was performed with both COX-1 targeting compounds, the reference compound RR11 and the novel 23 (MSA14), at different concentrations and time points. Prior to FLI, mice were shaved to avoid autofluorescence background signal. The desired concentration (100 µM, 500 µM) of the compound stock solutions in DMSO were freshly diluted with PBS, before 100 µL was injected intravenously into the tail vein. FLI was performed with the IVIS Spectrum In Vivo Imaging System with a λ abs 640 ± 15 nm bandpass filter (BP), λ em 680 ± 10 nm BP for RR11, and λ abs 745 ± 15 nm BP, λ em 820 ± 10 nm BP for 23 (MSA14), 0.5-24 h after small molecule administration. All scans were acquired with epi-illumination and auto exposure. Regions of interest (ROI) were manually gated around the tumor, organs, or the whole ventral and lateral positioned mouse and calculated using the Living image software (Perkin Elmer). The ROI of the muscle at the flank region was used to calculate the tumor-to-background (TBR) ratios.
A second intraoperative compatible imaging system (FLARE ® , Curadel LLC, Nattick, MA, USA) was used to assess the clinical utility of 23 (MSA14). Images were acquired using the NIR channel #2 (λ abs 760 ± 3 BP; λ em 781 long pass filter) with an exposure time of 500 ms to 2 s and a gain of one. The color video channel has a 400-660 nm illumination source; exposure time was set to 8.5 ms. Both images were simultaneously acquired and merged as a pseudo-colored image. The imaging head was positioned at 24.5 cm working distance, resulting in a field of view (FOV) of 7,7 cm 2 (no zoom).

Computational Methods
FLAP (fingerprints for ligands and proteins) [36,37] developed and licensed by Molecular Discovery Ltd. (Molecular Discovery, Borehamwood, United Kingdom) was employed to perform the virtual screening experiments, as reported in recent literature reports. [33,38].
The molecular interaction fields (MIFs) calculated by GRID [39] are employed in FLAP to describe both receptor and small molecules in terms of four-point pharmacophoric fingerprints. In the SBVS procedure, the GRID MIFs provide a complete description of both the ligands and the binding site (pocket), essentially in terms of shape (H), hydrophobic (DRY), H-bond acceptor (N1), and H-bond donor (O) properties of the pocket. The screening process involves the matching of the pockets and ligand quadruplets, which is quantitatively scored by considering the MIFs similarity. The Glob-Prod and Glob-Sum are two global scores that are produced by multiplying and summing all the scores of the individual probes.

Conclusions
Five different fluorochromes (dansyl, nitrobenzofuran-NBD, isoquinoline, cyanine, and nile-blue) were chosen as moieties of new compounds projected to target "in vivo" COX-1 overexpressed in the human ovarian cancer models. In particular, many such compounds were found to be potent and mostly highly selective COX-1 inhibitors, with their selective index ranging between 4 and 1000. COXs substrates and inhibitors are recognized by Arg120, Tyr355, and Glu524 located at the entry that gives access to the enzyme catalytic site. In this case, mofezolac moiety present in all twelve new compounds was expected to be recognized by the three amino acids (Arg120, Tyr355, and Glu524) and then to cross the long hydrophobic channel to reach Tyr385, responsible as a tyrosyl radical, to initiate the reaction that converts arachidonic acid into PGH 2 (precursor of prostaglandins, thromboxane and prostacyclin). Instead, this portion remains outside the catalytic zone while the fluorochrome is inside, as shown in the FLAP measurements. RR11 and 23 (MSA14) "in vivo" results show a high tumor-unspecific signal, which causes overall low TBRs. COX-1 is constitutively expressed in many organs, e.g., liver, spleen, intestines, and bone marrow, so the signal could be a result of mouse and human COX-1 cross reactivity. Therefore, COX-1 expression in ovarian cancer cell lines might not be specific enough, which renders both compounds unsuitable for FIGS. Further investigations are underway to better address the problem of fluoresce-guided surgery in ovarian cancer and to identify the real target of the newly prepared compounds.
Author Contributions: P.V., S.F., A.C. and R.S. carried out the synthesis of compounds and their chemical structural assignment. M.G.P. and M.M. performed COXs assays and fluorescence measurements. FLAP investigation by C.B., C.M. and C.G.F. "In vivo" and "ex vivo" experiments were accomplished by K.K., L.B. and A.S. supervised the experimental work and supported data interpretation. The manuscript was written through the contribution of all authors. All authors have read and agreed to the published version of the manuscript.