2.1.

Preparation of Dye-Loaded Lipid Nanoparticles

Suppocire NC™ is a kind gift from Gattefossé (Saint-Priest, France). Myrj 53™ (polyethylene glycol 50 stearate) is from Croda Uniqema (Chocques, France), and DiD and DiR dyes from Invitrogen. Soybean oil, L- $\alpha$ -lecithin (soybean lecithin at $>30\%$ of phosphatidylcholine), and other chemical products are purchased from Sigma Aldrich (Saint-Quentin Fallavier, France).

LNPs are composed of a lipid core (mixture of soybean oil and Suppocire NC), stabilized by a surfactant shell (L- $\alpha$ -lecithin and Myrj 53), and dispersed in an aqueous phase, as depicted in Fig. 2 . For preparation of a $2\text{-}\mathrm{mL}$ dispersion batch, $200\mathrm{mg}$ of the lipid components ( $150\mathrm{mg}$ of Suppocire NC™/ $50\mathrm{mg}$ of soybean oil for formulations A1 and A2, $200\mathrm{mg}$ of Suppocire NC for formulation B2, $50\mathrm{mg}$ of Suppocire NC/ $150\mathrm{mg}$ of soybean oil for formulation C2) are mixed with L- $\alpha$ -lecithin ( $138\mathrm{mg}$ for formulation A1, $100\mathrm{mg}$ for formulations A2, B2, C2) and the appropriate amount of dye dissolved in dichloromethane. After homogeneization at $45°\mathrm{C}$ , the organic solvent is evaporated under reduced pressure and the lipid phase is crudely mixed with the continuous aqueous phase composed of Myrj 53 ( $228\mathrm{mg}$ for formulation A1, $331\mathrm{mg}$ for formulations A2, B2, C2), $50\mathrm{mg}$ of glycerol, and the appropriate buffer [1X PBS (10 mM phosphate, 150 mM NaCl, pH 7.2) if not stated otherwise, qsp $2\mathrm{mL}$ ]. Sonication (VCX750 Ultrasonic processor, power output $190\mathrm{W}$ , $3\text{-}\mathrm{mm}$ probe diameter, Sonics) is performed at $40°\mathrm{C}$ for a whole $5\text{-}\min$ period to yield $35\mathrm{nm}$ (formulation A1) or $30\mathrm{nm}$ (formulations A2, B2, C2) diameter particles. In order to eliminate free (nonencapsulated) DiD or DiR dyes and other components, LNP dispersions are dialyzed overnight at room temperature against 1000 times their volume in the appropriate aqueous buffer ( $12–\mathrm{14,000}\mathrm{Da}$ MW cut off membranes, ZelluTrans). Finally, the LNP dispersion is filtered through a $0.22\text{-}\mu \mathrm{m}$ Millipore membrane for sterilization before characterization and/or injection.

Fig. 2

Schematic representation of the dye-loaded LNP structure.

2.2.

Structural Characterizations

Cryo-transmission electron microscopy (TEM) analysis is performed using an FEI CM200 microscope with a $\mathrm{La}{\mathrm{B}}_6$ electron source operating at $200\mathrm{kV}$ . The sample is deposited onto a holey carbon grid, blotted to remove excess liquid, and frozen by plunging rapidly into liquid ethane at the temperature of liquid nitrogen. Grids are transferred to a Gatan 626 cryoholder and imaged by low-dose techniques, maintaining the low temperature $(-185°\mathrm{C})$ throughout. Images are recorded at a nominal magnification of $\mathrm{38,000}\mathrm{x}$ . Micrographs are recorded on Kodak SO-163 film and developed for $12\min$ in full-strength Kodak D19. The hydrodynamic diameter and zeta potential of the lipid nanoparticles are measured with a Malvern Zeta Sizer Nano instrument (NanoZS) in 0.1 X PBS buffer. Values indicate mean and standard deviation of at least three measures per batch, for three different batches per formulation.

2.3.

Steady-State Optical Characterizations

The absorbance and fluorescence measurements are performed respectively using a Cary 300 Scan (Varian) UV-Visible spectrophotometer and a Perkin Elmer LS50B fluorimeter. The absolute fluorescence quantum yields $\Phi$ are calculated by comparison to standards of known $\Phi$ [Nile blue perchlorate in ethanol $(\Phi =0.27)$ ⁴⁰ for $600\text{-}\mathrm{nm}$ excitation (Cy5 and DiD), HITC in methanol $(\Phi =0.28)$ ⁴¹ for $700\text{-}\mathrm{nm}$ excitation (Cy7 and DiR)]. A continuous increase of the absorbance with decreasing wavelengths (Absorbance $\propto {\lambda }^{-4}$ ) due to a slight Rayleigh scattering effect has been observed for unloaded LNPs. This effect is corrected for the calculation of the fluorescence quantum yields of the dyes whenever encapsulated in the nanocarriers. All steady-state characterizations are carried out using dye or particle dispersions in the relevant buffers, with optical densities of 0.1–1 for absorbance measurements, and $\approx 0.1$ for fluorescence measurements.

The efficiency of DiR and DiD encapsulation is measured by spectrophotometric titration, using the absorbance of the LNP dispersion before and after extensive dialysis. Encapsulation efficiencies of $>99\%$ are observed. An estimate of the average number of dyes per particle $⟨N⟩$ is calculated based on the number of encapsulated dyes $\left(n_{\mathrm{f}}\right)$ , and hypothesizing that the volume of the dispersed phase $V_{\mathrm{d}}$ (representing the total volume occupied by the $n_{\mathrm{p}}$ particles) include both the oily core and the surfactant shell. $⟨N⟩$ is then obtained using the following formula:

The photobleaching measurements are carried out with the same setup as mouse imaging, with an excitation power of $200\mu \mathrm{W}∕{\mathrm{cm}}^2$ at $633\mathrm{nm}$ . Then, $10\mu \mathrm{M}$ free or encapsulated dye solutions are prepared in the appropriate medium $\left(100\mu \mathrm{L}\right)$ and introduced into glass capillaries that are simultaneously lightened in the homogeneous lightening field (homogeneity $>95\%$ in $20{\mathrm{cm}}^2$ ). Half-lives of the dyes in the different conditions are calculated using a linear fit of the photobleaching curves.

2.4.

Lifetime Measurements

Fluorescence lifetime measurements are carried out in plastic cuvettes with $\approx 0.1$ absorbance solutions at the excitation wavelength, using a time-correlated single-photon counting (TCSPC) setup. For DiD and Cy5, the excitation source is a $630\text{-}\mathrm{nm}$ laser diode (BHL-600, Becker&Hickl) operating at $50\mathrm{MHz}$ ( $150\mu \mathrm{W}$ , $80\text{-}\mathrm{ps}$ pulses). For DiR and Cy7, the excitation source is a Ti:Sa femtosecond laser (Tsunami, Spectra-Physics) operating at $730\mathrm{nm}$ and $80\mathrm{MHz}$ ( $10\mathrm{mW}$ , $100\text{-}\mathrm{fs}$ pulses broaden to a few picoseconds after propagation in an optical fiber). The detection of the emission signal, filtered by a RG665 (Cy5 and DiD) or a RG780 (Cy7 and DiR) filter (Schott), is carried out at right angle using a photomultiplier R3809U-50 (Hamamatsu), an amplifier UFAC-26 (Becker&Hickl), and the TCSPC module SPC-730 (Becker&Hickl). The fluorescence decays are analyzed using the SPCM software (Becker&Hickl), after deconvolution of the instrument response function. Monoexponential decay is used to fit recorded data, eventually taking into account a scattering contribution.

2.5.

Numerical Simulations

In order to illustrate the benefit for in vivo imaging of developing fluorescent probes with long lifetimes, together with high quantum yields, a series of simulations are provided later (Fig. 8). To stay as close as possible to an actual situation, we consider a digitized 3-D volume of a living mouse acquired with a system previously described.⁴² Inside this volume, a fluorescence object has been introduced (ellipsoid, semiaxes length: $X$ : $1\mathrm{cm}$ , $Y$ : $0.25\mathrm{cm}$ , $Z$ : $0.1\mathrm{cm}$ ). Fluorescence light propagation through biological tissues is usually modeled within the diffusion approximation as a two-step light-propagation phenomenon.⁴³ Suppose a diffusing medium, in the present case a mouse, with a total volume $\Omega$ and bounded by the surface $\partial \Omega$ . Suppose a pointlike source located at the position ${\mathbf{r}}_{\mathrm{s}}$ inside the diffusing medium with intensity $Q_o\left({\mathbf{r}}_{\mathrm{s}}\right)$ . The fluorochromes present within the medium are excited by the incident diffuse photon density wave (DPDW) $u_x$ at the excitation wavelength ${\lambda }_x$ . Each excited fluorescent particle then acts as a secondary point source and gives birth to a fluorescent DPDW, propagating with a larger wavelength ${\lambda }_m$ . This is commonly modeled, within the diffusion approximation, by the following set of weakly coupled partial differential equations:

Fig. 8

(a) Description of the synthetic mouselike medium used for simulations: it is composed of three subdomains, representing the body of the mouse, the ellipsoidal fluorescence inclusion, and the position of the excitation source. (b) Temporal fluorescence profiles detected at position (0.2, 0, $0.25\mathrm{cm}$ ), for different relative fluorochromes concentrations (conc. background:inclusion), and increasing inclusion lifetime $\tau$ (background lifetime is kept unchanged and equal to $0.5\mathrm{ns}$ ).

For simplicity, the optical properties are supposed to be the same at both the excitation and emission wavelengths: $D_x=D_m=D=0.033\mathrm{cm}$ and ${\mu }_{\mathrm{ax}}={\mu }_{\mathrm{am}}={\mu }_a=0.3{\mathrm{cm}}^{-1}$ , and the different subdomains are supposed to be homogeneous. The source is modeled as a sphere with radius $0.05\mathrm{cm}$ , located at position (0,0,0). The resolution of the equations is performed using the finite element method with the commercial software Comsol Multiphysics®. To that purpose, the volume has been meshed resulting in $\mathrm{20,844}\mathrm{deg}$ of freedom. The chosen linear system solver is GMRES with a geometric multigrid preconditioner. The calculation is performed within the time interval [0; $12\mathrm{ns}$ ], corresponding to usual time-resolved measurements, and the time stepping is $0.1\mathrm{ns}$ . The background lifetime is set to ${\tau }_0=0.5\mathrm{ns}$ , and the fluorescence inclusion lifetime (0.5, 1.1, 1.8, 2.1, 5, $10\mathrm{ns}$ ), as well as its concentration relative to background (background:inclusion 1:5 and 1:50) are varied.

2.6.

Animal Models

Ts/Apc cells (murine breast carcinoma) are cultivated in RPMI 1640 medium (Gibco) supplemented with $4.5\mathrm{g}∕\mathrm{L}$ glucose, 10% fetal calf serum, 1% glutamine, $50\mathrm{U}∕\mathrm{mL}$ penicillin, $50\mathrm{mg}∕\mathrm{mL}$ streptomycin, and $2.5\times {10}^{-5}$ M 2-mercaptoethanol. Human ovarian adenocarcinoma OVCAR-3 cells are cultivated in RPMI 1640 medium supplemented with 20% fetal calf serum, 1% glutamine, $50\mathrm{U}∕\mathrm{mL}$ penicillin, and $50\mathrm{mg}∕\mathrm{mL}$ streptomycin. All cell lines are cultured in $75\text{-}{\mathrm{cm}}^3$ flasks with $15\mathrm{mL}$ of growth medium and kept at $37°\mathrm{C}$ in a humidified 95% air/5% $\mathrm{C}{\mathrm{O}}_2$ atmosphere.

All the animal experiments are performed in agreement with the EEC and U.S. guidelines.⁴⁴ About two weekes prior to imaging, ${10}^6–{10}^7$ Ts/Apc or OVCAR-3 cells are implanted subcutaneously in anesthetized female six-to-eight-week-old Nude mice (Janvier, le Genest Saint Isle, France). Imaging is carried out whenever the tumor reaches about $6–8\mathrm{mm}$ diam. For imaging, $200\mu \mathrm{L}$ of DiD-loaded LNP, representing $10\mathrm{nmol}$ of dye ( $3.2\mathrm{mg}$ of oil, $5\times {10}^{14}$ particles with a local DiD concentration of $1.55\mathrm{mM}$ ) are injected in the tail vein of the anesthetized animals (isoflurane/air 1.5%) placed on a warm plate $(37.5°\mathrm{C})$ . Animals are subsequently imaged at the desired time points using a fluorescence-reflectance-imaging setup. This homemade system is adapted from the commercially available Aequoria™ system from Hamamatsu. It is made up of a dark box equipped with a mouse-body temperature controller and a gas anesthesia device. The excitation device is composed of 10 LED emitting at $633\mathrm{nm}$ (adapted from the LuxiFlux™ device available from Hamamatsu) and equipped with interference bandpass filters (633BP10 nm from Schott) for a light illumination power of $15\mu \mathrm{W}∕{\mathrm{cm}}^2$ . The filtered fluorescence signal (filter RG665 from Schott) is measured by a cooled CCD camera (Orca II BT $512\mathrm{G}$ , Hamamatsu), placed at $160\mathrm{mm}$ from the imaging field, with an exposure time that is typically $20\mathrm{ms}$ . The Wasabi™ software (Hamamatsu) is used to drive the setup and for image processing. Fluorescence images can be superimposed on the black-and-white picture for better localization of the fluorescent signal. Semi-quantitative data can be obtained from the fluorescence images by drawing regions of interest (ROI) on the area to be quantified. The results are expressed as a number of relative light units per pixel.

3.1.

Encapsulation of Lipophilic Dialkylcarbocyanine Dyes in Lipid Nanoparticles

Figure 2 depicts the structure of the dye-loaded LNPs. The LNP core is composed of a mixture of vegetable oil (from 0 to 75% w/w) and wax (Suppocire, from 25 to 100% w/w), to which can be mixed the DiD or DiR dyes (structures represented in Fig. 1). The presence of Suppocire allows the solubilization of a high concentration of lecithin (6.9% w/w for formulation A1, 5% w/w for formulations A2, B2, C2), which, in combination with polyoxyethylene (PEG) surfactants (11.4% w/w for formulation A1, 16.6% w/w for formulations A2, B2, C2), stabilizes very efficiently the droplets in aqueous buffer. Practically, the oily phase (oil, wax, lecithin, and eventually the DiD or DiR dye) is mixed with the aqueous phase (PEG surfactant, glycerol, buffer), and the microemulsion is sonicated to yield $35\text{-}\mathrm{nm}$ -diam (formulation A1) or $30\text{-}\mathrm{nm}$ -diam (formulations A2, B2, C2) LNPs. In this paper, we mainly focus on nanoparticles issued from the formulation A1 composed of 2.5% soybean oil, 7.5% Suppocire, 6.9% lecithin, 11.4% Myrj53 (PEG surfactant), 2.5% glycerol, 69.2% PBS (unless stated otherwise). The size of the particles is mainly governed by the nature, concentration, and ratio of the surfactants used. When increasing the ratio of the PEG:lecithin surfactants from 33%:66% (formulation A1) up to 50%:50% (formulation A2), we observe a small decrease of the nanoparticle size (30 versus $35\mathrm{nm}$ ). However, no significant differences are observed for formulations A2, B2, and C2, for which the core lipid content (oil/wax ratio) differs, but are processed using the same surfactant (PEG, lecithin) content. The typical dye concentration used is $400\mu \mathrm{M}$ , which corresponds to a number of 21 dyes/particle ( $35\mathrm{nm}$ diam). The efficient dialysis of the LNP dispersions ensures elimination of any nonincorporated molecules (dyes, surfactants) at the end of the process.

The absorption and emission spectra of the DiD and DiR fluorophores in methanol or encapsulated within the LNPs (formulation A1) are displayed in Fig. 3 . No major modification is observed in the spectral properties of the dyes upon encapsulation (data, Table 1), except a slight increase of the $600\text{-}\mathrm{nm}$ absorption band, possibly indicative of the presence of a few dye dimers. The absorption and emission maximum wavelengths of DiR and DiD are also very slightly redshifted $(1–3\mathrm{nm})$ upon encapsulation and are similar to those of their hydrophilic counterparts, respectively, Cy5 and Cy7 (Fig. 1, Table 1). For nondoped LNPs, it has to be noted that there is a continuous increase of the absorbance with decreasing wavelengths (Absorbance $\propto {\lambda }^{-4}$ ), as a slight Rayleigh scattering effect (Fig. 3). This effect is taken into account for the calculation of the fluorescence quantum yields of the encapsulated dyes.

Fig. 3

Absorption and emission spectra of (a) DiD and (b) DiR in methanol or encapsulated within the LNPs, and absorption spectrum of the blank formulation ( $35\text{-}\mathrm{nm}$ -diam particles; formulation A1 dispersed in 1 X PBS).

The perfect match of the absorption and emission spectra of the formulation before and after dialysis demonstrates the 100% efficiency of incorporation of the lipophilic cyanines in the LNPs, for total dye concentration in the dispersion in the range of $75–1000\mu \mathrm{M}$ . This corresponds to the incorporation of 4–53 dyes per particle, or a local dye concentration of $0.30–3.9\mathrm{mM}$ in the droplets. Figure 4a displays the hydrodynamic diameter of the nanoparticles either loaded or not by the fluorophores, measured by dynamic light scattering (DLS) in PBS buffer. For formulation A1, only one-sized population centered at $35\pm 2\mathrm{nm}$ is evidenced by DLS, with a very reproducible size polydispersity of 0.17. Whatever the oily core composition, only one-sized population centered at $30\pm 2\mathrm{nm}$ is evidenced by DLS for formulations A2, B2, and C2. However, the cryo-TEM image of Fig. 4b evidences different sizes and forms, which are attributed to oblong particles viewed from different angles. Platelet or oblong particles have also been observed by cryo-TEM in the case of solid LNPs dedicated to drug delivery.⁴⁵ In this work, we will describe the particle size by their mean DLS hydrodynamic diameter; moreover, only one population is observed by this technique. As already noted, modification of the oil/wax ratio of the LNP core does not modify the particle size, which is mainly governed by the surfactant properties, concentrations, and ratios. Similarly, the dye encapsulation does not modify the droplet diameter, neither their zeta potential, which remains neutral $(-5\pm 2\mathrm{mV})$ . The neutrality of the surface of the particles is certainly ensured by the use of zwitterionic lecithin and neutral PEG surfactants. The fact that the LNP surface remains neutral despite the positively charged dye encapsulation confirms that the fluorophores are encapsulated within the lipid core rather than adsorbed on, or included in, the particle surfactant surface.

Fig. 4

(a) Size distribution of undoped and DiR- or DiD-loaded LNP measured by diffusion light scattering in 0.1 X PBS (formulation A1, 21 dyes/particle). The inset demonstrates the transparent aspect of the colloidal suspension because the particle diameter is $<100\mathrm{nm}$ (left: DiR-loaded LNPs, middle: undoped particles; right: DiR-loaded LNPs). (b) Cryo-TEM image of undoped LNPs.

3.2.

Stability of Dye-Loaded Lipid Nanoparticles

The inset in Fig. 4a demonstrates the transparent aspect of the colloidal suspension since the particle diameter is $<100\mathrm{nm}$ . The transparency of the solution is therefore a good indicator of its colloidal stability. Fluorescence and colloidal stabilities of 21 DiD/particle LNP dispersions (formulation A1) prepared at pH 5.5 in HBS buffer (HEPES 0.02 M; EDTA 0.01 M) and kept protected from light on the laboratory shelf are further assessed by DLS and fluorescence measurements [Fig. 5a ]. A very good stability of both the fluorescence properties of the encapsulated dyes and the size of the droplets is observed for more than one year in HBS buffer. This stability is observed in the pH 3–8 range (adjusted by NaOH or HCl), either in water, PBS, or HBS buffers, for at least $100\text{days}$ . However, the fluorescence properties of DiD-loaded LNP become altered rapidly in basic medium $(\mathrm{pH}>8)$ [Fig. 5b]. The colloidal stability of the solution is maintained, but its color shifts from blue to green and yellow. The loss of fluorescence signal can therefore be accounted for by a chemical degradation of the cyanine dye, a class of fluorophores known to be sensitive to pH conditions.¹⁸ This is confirmed by the disappearance of the $646\text{-}\mathrm{nm}$ absorption band of the dye and the appearance of a new absorbance band at $467\mathrm{nm}$ , associated with a weak emission centered at $497\mathrm{nm}$ . On the contrary, in acidic media, the loss of the fluorescence properties, which occurs on a longer time scale [Fig. 5b], goes along with the appearance of opaqueness of the dispersions. In this case, the absorption spectra of the LNPs always display the absorption and emission bands of the DiD dye with reduced intensities, but also an important increased scattering: there exist both a degradation of the fluorophore (but different from in basic media because no color change is observed) and a destabilization of the nanoparticles.

Fig. 5

(a) Colloidal and fluorescence stability of DiD-loaded LNP (formulation A1, 21 dyes/particle) dispersed in HBS buffer (Hepes 0.02 M, EDTA 0.01 M, pH 5.5). (b) Fluorescence stability of DiD-loaded LNPs (formulation A1, 21 dyes/particle) in deionized water added by concentrated HCl or NaOH to adjust the pH.

Figure 6 displays the evolution of the fluorescence of the DiD fluorophore in methanol or encapsulated in LNPs, in comparison to that of its hydrophilic counterpart (Cy5) whenever submitted to a continuous light irradiation at $633\mathrm{nm}$ for $24\mathrm{h}$ (excitation power: $200\mu \mathrm{W}∕{\mathrm{cm}}^2$ ). The photobleaching rate of DiD-loaded LNPs dispersed in aqueous buffer is higher than for the free dye in methanol. Moreover, this photobleaching rate is slightly sensitive to the continuous medium ( $154\mathrm{mM}$ NaCl or 1X PBS buffers) in which the particles are dispersed. However, this photobleaching rate is far reduced in comparison to that of Cy5 dissolved in the same buffers.

Fig. 6

Compared photobleaching of Cy5, DiD, and DiD-loaded LNPs (formulation A1, 21 dyes/particle, $154\mathrm{mM}$ NaCl or 1X PBS ( $10\mathrm{mM}$ phosphate, $137\mathrm{mM}$ NaCl, $2.7\mathrm{mM}$ KCl, pH 7.4)). The excitation wavelength is $633\mathrm{nm}$ , and the excitation power $200\mu \mathrm{W}∕{\mathrm{cm}}^2$ . Half-lives $\left(t_{1∕2}\right)$ of the dyes, calculated using a linear fit of the curves, are indicated.

Because dye degradation is observed in basic buffer, and the fluorophore photobleaching kinetics is sensitive to the continuous phase, it might be indicative that despite its LNP encapsulation, the fluorophore might exchange with the continuous medium. This could be due to the nature of the LNPs, which are permeable and deformable nanoparticles, the fluid surfactant interface through which oxygen and ions could diffuse. The increased photobleaching observed in the LNP core in comparison to the free dye in methanol could also be due to possible photoreactions occurring between the dye in its excited state and the lipid components composing the LNP core. In particular, soybean oil is mainly constituted from unsaturated glycerides, which might be prone to oxydation. However, all previous results (Fig. 5) demonstrate the very good encapsulation and stability of the lipophilic dialkylcarbocyanine dyes in the LNP in the pH 3–8 range, compatible with in vivo conditions. This is certainly due to the choice of the fluorophores, with long lipophilic chains (Fig. 1), which confer them a very good affinity for the lipid core of the droplets.

3.3.

Fluorescence Quantum Yields and Lifetimes of Dye-Loaded Lipid Nanoparticles

The fluorescence quantum yields and lifetimes of the DiD and DiR dyes, either in methanol or encapsulated in the LNP of different formulations, as well as those of their hydrophilic counterparts Cy5 and Cy7 in PBS buffer are listed in Table 1. As already noted, both the fluorescence quantum yields and lifetimes decrease when shifting from the red part (Cy5 and DiD) to the near-infrared part (Cy7 and DiR) of the spectrum. The role of the buffer on these optical parameters is also very important: despite very similar structures, Cy5 (respectively, Cy7) dispersed in an aqueous buffer displays lower fluorescence quantum yield than DiD (respectively, DiR) dispersed in methanol (DiD and DiR are not soluble in water or aqueous buffers). These lower performances of organic dyes in aqueous media are often accounted for by fluorescence quenching of buffer components, limited dye solubility or dye aggregation, or the presence of oxygen, which is known, in particular, to favor dye photobleaching.

However, it is necessary for biological applications and especially for in vivo imaging in animals to dispose of efficient fluorescent tracers with optimized optical properties in aqueous buffers, compatible with injection. Therefore, the encapsulation of dyes within “organiclike” medium nanoparticles, in which they could display improved optical properties, is attractive if these nanoparticles can be dispersed with a good colloidal stability in water. LNPs respond to this demand. Indeed, DiD and DiR encapsulated within LNPs display fluorescence quantum yields and lifetimes similar or even higher than the free dyes dispersed in organic solution (Table 1, Fig. 7 ). A high dye loading ratio can be achieved ( $>99\%$ encapsulation efficiency, up to at least 53 dyes per $35\mathrm{nm}$ diameter particle, corresponding to a dye local concentration of $3.9\mathrm{mM}$ ), without any significant fluorescence auto-quenching (Table 1). This is noticeable since cyanines are known to be prone to aggregation and auto-quenching in aqueous buffers.¹⁸ Therefore, DiD and DiR loading within LNP provides aqueous soluble fluorescent labels with far improved fluorescence quantum yields $\Phi$ (respectively 0.38 and 0.25) and longer fluorescence lifetimes $\tau$ (respectively $1.8\mathrm{ns}$ and $1.1\mathrm{ns}$ ) in comparison to their hydrophilic counterparts Cy5 ( $\Phi =0.28$ , $\tau =1.0\mathrm{ns}$ ) and Cy7 ( $\Phi =0.13$ , $\tau =0.57\mathrm{ns}$ ) (Table 1, Figure 7). Moreover, very high global dye concentrations (at least $1\mathrm{mM}$ in water) can be achieved without any solubility problem or deterioration of the optical properties due to aggregation, contrary to the direct solubilization of hydrophilic cyanines in water.

Fig. 7

(a) Fluorescence decays and (b) lifetimes of Cy5 (1X PBS), DiD (MeOH), and DiD-loaded LNP (1X PBS, formulation A1, 21 dyes/particle). The excitation wavelength is $630\mathrm{nm}$ .

We also notice that the fluorescence quantum yield and lifetime of DiD whenever encapsulated in the LNP are increased in comparison to those of the free dye in methanol (Table 1). In order to better understand the LNP encapsulation effects on the dye photophysics, radidative $[{\mathrm{k}}_{\mathrm{r}}=(\Phi ∕\tau )]$ and nonradiative $[{\mathrm{k}}_{\mathrm{nr}}=(1-\Phi )∕\tau ]$ rate constants derived from steady-state and time-resolved experiments are listed in Table 1. Fluorescence quantum yields and lifetimes are dependant on both the radiative and nonradiative processes. Larson reported a 2.2-fold increase of the radiative rates of TRITC (a rhodamine dye) upon silica encapsulation, relatively to water.³⁷ They account for this increase of the radiative rates by the change of refractive index between silica $(n=1.46)$ and water $(n=1.33)$ , according to a model described by Toptygin ⁴⁶ On the contrary, in this study, for a similar change of the refractive index between soybean oil $(n=1.47)$ and methanol $(n=1.33)$ , a decrease of the radiative rate of DiD and DiR is observed. The radiative rate modifications observed in this study should therefore be attributed to other interactions between the dyes and the lipid matrix.

Nonradiative decays can be accounted for by three process: (i) intersystem conversion from the excited singlet state to the triplet; (ii) photoisomerization, which is a well-acknowledged process for cyanine dyes,¹⁸ and (iii) internal conversion closely linked to the vibrational and rotational degrees of freedom of the dyes. Intersystem conversion has been reported as negligible for cyanine dyes in the absence of heavy atoms.⁴⁷ Photoisomerization and internal conversion are both influenced by the polarity and viscosity of the dye-surrounding medium. Two studies carried out on several tricarbocyanines in solvents of different polarities and viscosities attributed the decrease of radiative rates observed in nonpolar media in comparison to more polar solvents mainly to polarity effects, with low-viscosity influence.^{47, 48} On the contrary, other studies point out the importance in the nonradiative decay rates of the dye rotational degrees of freedom, which is linked to the viscosity.^{18, 37} In particular, Larson reported that the nonradiative rate constants of a rhodamine dye in a silica matrix is directly linked to the rotational mobility of the fluorophore.³⁷ The less movement the dye can have, the slower the nonradiative decay is. LNPs display both a high viscosity ( $79\mathrm{cP}$ for soybean oil, Suppocire is solid at $25°\mathrm{C}$ ) and a very low polarity (solvent orientation polarizability as defined in Ref. 48 is $\Delta f=0.06–0.09$ for soybean oil) in comparison to methanol ( $0.58\mathrm{cP}$ , $\Delta f=0.31$ ). Therefore, both effects could explain the decrease of the nonradiative rate observed for DiD from methanol to LNPs. The fact that no significant difference is observed between the nonradiative rates for the different A2, B2, and C2 formulations, which display different wax contents and therefore different viscosities, seems to indicate that the modification of the nonradiative rate observed between methanol and the LNP should be due mainly to a polarity effect. Surprisingly, no significant difference is observed for the DiR dye between methanol and the LNP core in comparison to what is observed for DiD.

Figure 8b displays the temporal fluorescence profiles simulated for a dye inclusion placed in the 3-D digitized volume of a living mouse [geometry described in Fig. 8a].⁴² The results represent the normalized intensities measured at position (0.2,0,0.25), for a simulated reflection detection mode. The same kind of curves have been obtained for simulated transmission acquisitions. The calculations are performed for different relative background to inclusion concentrations (background:inclusion 1:5 or 1:50), and for different values of the inclusion lifetime, corresponding to the same lifetime as the background $\left(0.5\mathrm{ns}\right)$ , DiD in MeOH $\left(1.1\mathrm{ns}\right)$ , DiD in A1 formulated LNPs $\left(1.8\mathrm{ns}\right)$ , and DiD in B2-C2 formulated LNPs $\left(2.1\mathrm{ns}\right)$ . Longer lifetimes (5 and $10\mathrm{ns}$ ) have also been considered. A first global remark is that the curves, plotted with a logarithmic scale, highlight two distinct behaviors: for short times (basically, $>0.1$ and $<0.3\mathrm{ns}$ ), the slope of the linear curve is equal to the inverse of the background lifetime, and at long times, to the inverse of the inclusion lifetime. Hence, by time gating the measured signal, one could selectively decouple the signal coming from the background and the signal coming actually from the ROI. Increasing the concentration of the fluorescence inclusion (conc. 1:5 to conc. 1:50) does not exhibit any difference in the time profiles, the modifications appear only on the level of signal detected. On the contrary, increasing the inclusion lifetime allows an easier detectability of the molecular probes. Nevertheless, if the lifetimes are too long $(>5\mathrm{ns})$ , it becomes difficult to exploit the slope of the curves (almost horizontal) to identify the marker through its lifetime.

To summarize, developing fluorescence molecular probes with lifetimes in the $1.0–3.0\mathrm{ns}$ range provides two major advantages: (i) possibility of time gating the fluorescence signal in order to selectively measure the emitted photons coming from the targeted zone; (ii) identifying the fluorochromes through the slope of the logarithmic curves at long times. In this perspective, DiD- and DiR-loaded LNPs (respectively, 1.8 and $1.1\mathrm{ns}$ ) appear more interesting labels in comparison to their classical water soluble counterparts Cy5 and Cy7 (respectively, 1.0 and $0.57\mathrm{ns}$ ). Moreover, DiD lifetimes appear to be sensitive to the LNP dye encapsulation. Therefore, in addition to its use as an imaging agent, DiD-loaded LNPs could be an interesting system to model and investigate the release of a lipophilic drug from the LNP carrier.

3.4.

Passive Tumor Targeting of Dye-Loaded LNP

Preliminary in vivo experiments are undertaken in order to demonstrate the potential of the new LNP nanocarriers for fluorescence imaging and other potential diagnostic and therapeutic applications. First, DiD-loaded LNPs [ $5\times {10}^{14}$ particles ( $3.2\mathrm{mg}$ of oil) at a local DiD concentration of $1.55\mathrm{mM}$ , dispersed in $200\mu \mathrm{L}$ of 1X PBS buffer, corresponding to $10\mathrm{nmol}$ of dye] are injected intravenously in healthy Nude mice $(n=4)$ . The fluorescence signal distributes very rapidly in the whole animal, and $24\mathrm{h}$ after injection, no organ seems to preferentially accumulate the dye (Fig. 9 ). This is outstanding because most nanocarriers of similar size can be quite rapidly uptaken by the macrophages of the reticuloendothelial system [(RES): mainly liver, spleen, bone marrow] after plasma protein binding.⁴⁹ The high density in the LNP shell of surfactants with a polyoxyethylene chain (PEG), which has been demonstrated to considerably reduce nanoparticle protein adsorption and therefore increase the plasma half-life of various nanocarriers, is certainly very important for conferring the LNP this suitable pharmacokinetics.^{50, 51, 52} Because of their very probable long lifetime circulation, efficient passive tumor accumulation of the LNPs in cancer models displaying leaky fenestrations should be achieved due to the EPR effect.³³ Ts/Apc $(n=4)$ and OVCAR-3 $(n=2)$ xenografted model mice have been used for imaging the passive tumor uptake of DiD-loaded LNPs. Ts/Apc (murine breast cancer model) and OVCAR-3 (human ovarian adenocarcinoma model) are tumor cells with a fast growth, leading to an imperfect vascular endothelium with leaky fenestrations. Figure 9 displays the fluorescent images obtained $24\mathrm{h}$ after intravenous injection in tail of DiD-loaded LNPs at the same dose than for a healthy animal ( $10\mathrm{nmol}$ of dye, $5\times {10}^{14}$ particles dispersed in $200\mu \mathrm{L}$ 1X PBS) for the two different tumor models. A significant accumulation (tumor/skin fluorescence $\text{ratio}=2.1–2.3$ ) is observed for both the Ts/Apc $(n=4)$ and OVCAR-3 $(n=2)$ models; $24\mathrm{h}$ corresponds to the time after injection for which the tumor fluorescence intensity as well as the tumor/skin fluorescence ratio are maximum in the two models. These preliminary experiments therefore demonstrate the possibility to use DiD-loaded LNPs for passive tumor uptake and imaging.

Fig. 9

In vivo imaging with DiD-loaded LNPs (formulation A1, 21 dyes/particle). Fluorescence images obtained $24\mathrm{h}$ after injection of DiD-loaded LNPs ( $3.2\mathrm{mg}$ of oil, $10\mathrm{nmol}$ of DiD in $200\mu \mathrm{L}$ PBS buffer) in nude mice, either healthy or bearing xenografted subcutaneous tumors. Fluorescence images (in color scale, lower part of the figure) are superimposed to visible light images (in white and black). For fluorescence images, the integration time is $20\mathrm{ms}$ and the contrast has been set between 2100 and 16,000. The tumor-over-skin ratio has been calculated by drawing a region of interest and quantifying the fluorescence signal with the Wasabi™ software. (Color online only.)