In vivo near-infrared imaging using ternary selenide semiconductor nanoparticles with uncommon crystal structure In Vivo Near-Infrared Imaging Using Ternary Selenide Semiconductor Nanoparticles with an Uncommon Crystal Structure.

The implementation of in vivo fluorescence imaging as a reliable diagnostic imaging modality at the clinical level is still far from reality. A lot of work remains ahead to provide medical practitioners with solid proof of the potential advantages of this imaging technique over well-established ones such as magnetic resonance imaging or nuclear imaging. To do so, one of the key objectives is to improve the optical performance of dedicated contrast agents, thus improving the resolution and penetration depth achievable with in vivo imaging. In the present study, we move along this direction and report on the use of AgInSe 2 nanoparticle-based contrast agents (nanocapsules) for fluorescence imaging. The use of a Ag 2 Se seeds-mediated synthesis method allows stabilizing an uncommon orthorhombic crystal structure. This structure endows the material with emission in the second biological window (1000-1400 nm), where deeper penetration in tissues is achieved. The nanocapsules, obtained via phospholipid-assisted encapsulation of the AgInSe 2 nanoparticles, comply with the mandatory requisites for an imaging contrast agent–colloidal stability and negligible toxicity in vitro –and show superior brightness compared with well-established and widely used Ag 2 S nanoparticles. Imaging experiments point to the great potential of the novel AgInSe 2 -based nanocapsules for high-resolution, whole-body in vivo imaging. In particular, their extended permancence time within blood vessels make them especially suitable for prolonged imaging of the cardiovascular system.


ABSTRACT
The implementation of in vivo fluorescence imaging as a reliable diagnostic imaging modality at the clinical level is still far from reality. A lot of work remains ahead to provide medical practitioners with solid proof of the potential advantages of this imaging technique over wellestablished ones such as magnetic resonance imaging or nuclear imaging. To do so, one of the key objectives is to improve the optical performance of dedicated contrast agents, thus improving the resolution and penetration depth achievable with in vivo imaging. In the present study, we move along this direction and report on the use of AgInSe2 nanoparticle-based contrast agents (nanocapsules) for fluorescence imaging. The use of a Ag2Se seeds-mediated synthesis method allows stabilizing an uncommon orthorhombic crystal structure. This structure endows the material with emission in the second biological window (1000-1400 nm), where deeper penetration in tissues is achieved. The nanocapsules, obtained via phospholipidassisted encapsulation of the AgInSe2 nanoparticles, comply with the mandatory requisites for an imaging contrast agent-colloidal stability and negligible toxicity in vitro-and show superior brightness compared with well-established and widely used Ag2S nanoparticles. Imaging experiments point to the great potential of the novel AgInSe2-based nanocapsules for highresolution, whole-body in vivo imaging. In particular, their extended permancence time within blood vessels make them especially suitable for prolonged imaging of the cardiovascular system.

Introduction.
Among other conditions, the prosperity of our society depends on a sustainable health system.
To that end, access to diagnostic approaches that are fast, inexpensive, and that minimize the discomfort of the patient is pivotal. Currently available techniques, such as magnetic resonance imaging, X-ray computed tomography, and nuclear imaging, offer ever improving performance, but in turn they require sophisticated equipment and elevated maintenance costs. In recent years, fluorescence imaging (FI) has been emerging as a credible complementary technique to those above mentioned. [1] The advantages of FI reside in the use of non-ionizing radiation, simple and less expensive setups, minimal invasiveness, and real-time image-acquisition capabilities.
However, its main limitation is the difficulty in obtaining deep-tissue and high-resolution in vivo images owing to the large attenuation of light into tissues. [2] The future application of FI at the clinical level as a viable alternative to conventional imaging techniques for selected applications passes through the development of more efficient detection systems as well as novel contrast agents. In this vein, luminescent nanoparticles (NPs) hold great promise.
In search for the most suitable NP-based contrast agents, it soon became clear that in vivo images of the highest quality could be obtained with NPs emitting in the so-called biological transparency windows. [3] These are near-infrared (NIR) spectral ranges where biological tissues are less impervious to the electromagnetic radiation, with NIR-I and NIR-II spanning over the 750-950 nm and 1000-1400 nm ranges, respectively. [4] The latest generation of luminescent NPs employed in FI emit in the NIR-II, wherein the low scattering coefficient and reduced absorption from tissue components result in significantly improved penetration depth and resolution of the fluorescence images. [3b] The use of NPs emitting in the NIR-II has made possible, for example, the acquisition of high-resolution in vivo images of the cardiovascular system, anatomical images of animal models, as well as molecular imaging of infarcted heart. [5] Carbon nanotubes, lanthanide-doped NPs, and semiconductor NPs are the most representative FI contrast agents. [6] In fact, many of the most significant advances in NIR-II FI have been obtained with semiconductor NPs made of Ag2S. [5f, 7] Ag2S NPs possess desirable properties like good biocompatibility and physicochemical stability. [8] Nonetheless, their performance is severely limited by their low brightness, which stems from low photoluminescence quantum yield (PLQY; < 0.1% for commercially available Ag2S), and limited molar absorption coefficient (4·10 -5 M -1 cm -1 at 800 nm, an optimal excitation wavelength for in vivo FI). [9] Moreover, methods to increase their optical performance are scarce, could require special equipment, and often afford limited improvement. [9a, 10] Therefore, the development of NIR-II-emitting semiconductor NPs that are brighter is an urgent matter. AgInSe2 (AISe) NPs, for instance, offer the possibility of achieving large brightness compared with Ag2S NPs, owing to their higher (> 20%) PLQY and their good photostability. [11] Their ternary nature also allows to finely manipulate their composition via selective cation exchange. [12] However, to the best of our knowledge, AISe NPs have been investigated only once as in vivo FI contrast agents by D. Deng et al., who performed tumor imaging using RGD-peptide-decorated AISe/ZnS core/shell NPs. [13] In that study, AISe NPs have their emission centered at 800 nm, imposing the use of an excitation wavelength lying outside of the biological windows (660 nm). Thus, the authors obtained in vivo fluorescence images of limited resolution. Being able to push the operating range of AISe NPs into NIR-I and NIR-II, in terms of excitation and emission respectively, would allow achieving images of much higher quality.
To address this issue, in this work we take advantage of a synthesis method that affords AISe NPs with an uncommon crystal structure, which in turn leads to a NIR-II-centered emission.
The hydrophobic NPs are transferred to aqueous milieus by encapsulating them in a phospholipid layer. These colloidally stable nanocapsules (NCs) are tested for their toxicity in vitro and later utilized as contrast agents for in vivo FI. The comparison of the results obtained with AISe NCs and commercial Ag2S NPs underscores the excellent performance of the newly developed contrast agents particularly for the visualization of the vasculature, opening the door to further studies on the use of this material for FI.

Results and discussion.
The synthesized AISe NPs showed a triangular shape in TEM images, which is suggestive of a tetrahedral habit, and an average size of (8 ± 2) nm ( Figure 1A and S1). They could be easily dispersed in non-polar organic solvents such as TCE, giving an optically clear dispersion ( Figure 1B). The diffraction pattern of AISe NPs featured broad reflections characteristic of nanometric crystalline domains ( Figure 1C). Both the morphology and the diffraction pattern were consistent with the data reported by Langevin et al. [14] It is worth noting that these AISe NPs possess an unusual orthorhombic crystalline phase that was first reported by Vittal and coworkers. [15] rather than the most common tetragonal polymorph. [16] The reason for this has to be searched in the NP formation mechanism, which entails orthorhombic Ag2Se seeds as an intermediate. The subsequent cation exchange with In 3+ yields AgInSe2 that are isostructural to AgInS2 crystallized in the orthorhombic Pna21 space group. [14] Therefore, the crystal structure of AISe NPs has memory of the structure of parent Ag2Se seeds. This uncommon crystal structure is likely the responsible for the unique optical properties of these NPs (vide infra).
Upon transfer to aqueous media with the aid of phospholipids (Figure 1D), the AgInSe2 NPs arranged in quasi-spherical aggregates ( Figure 1A, right). These nanocapsules form due to two competitive hydrophobic interactions: i) between DDT molecules attached to the surface of the NPs and the hydrophobic tail of the phospholipids, ii) between DDT molecules on the surface of different NPs. The stability of the AISe NCs in aqueous media was confirmed by DLS measurements, which showed a sizeable increase of the effective diameter when passing from tetrachloroethylene (TCE) to phosphate buffer saline (PBS) 1x (Figure 1E, left). We further tested the colloidal stability of the AISe NCs in PBS over the 15-55 ºC temperature range ( Figure 1E, middle), observing negligible changes in the hydrodynamic diameter (approximately 150 nm; Z-average) and in the polydispersion index (PDI; approximately 0.2).
As expected, the measured Z-potential was negative (-33 mV), due to the partial hydrolyzation of the maleimide groups on the surface of AISe NCs (Figure 1E, right). [17] This hydrolyzation ultimately resulted in the appearance of carboxylic groups, whose deprotonation in aqueous media leads to an overall negative surface charge. The use of phospholipids with maleimide was preferred due to the documented increased interaction of nanoparticles featuring this functional group towards cellular thiols (vide infra). [18] To further confirm the presence of phospholipids on the surface of the NCs, FTIR spectra of the as-synthesized AISe NPs and AISe NCs were collected and compared with the spectra of pristine organic molecules ( Figure   1F). Upon encapsulation of the AISe NPs in phospholipids, the signals originating from DSPE-PEG2000-MAL became dominant over those coming from DDT, thus confirming the presence of amphiphilic molecules in the NCs. We then tested the cytotoxicity of AISe NCs, incubating HeLa cells at different concentrations (5-100 μg/mL) for 2 and 24 h ( Figure 1G). The results did not indicate appreciable toxicity, with a maximum cell viability decrease of 13% after 2 h of incubation at an AISe NCs extracellular concentration of 100 μg/mL. The absence of cell toxicity of our NCs is in agreement with previous works reporting good biocompatibility of AgInSe2 NPs against cancerous and normal cell lines. [19] Altogether, the observed excellent colloidal stability in PBS along with the lack of substantial toxicity indicated the amenability of the developed NCs to applications in the biomedical context. Subsequently, the optical properties of the material were assessed. As can be observed in Figure 2A, while the extinction spectrum is not affected by the dispersing solvent, AISe NCs in water exhibited an emission that is slightly blue-shifted and narrower compared to the one of parents AISe NPs dispersed in TCE. This change in the emission spectral profile could be mainly attributed to the prominent inner filter effect exerted by water, which has strong absorption bands in the wavelength range of interest. [20] As expected, the transfer to water induces a sizeable intensity decrease close to 80% (Figure S2), which is attributed to the activation of non-radiative de-excitation events involving water vibrations. [21] To benchmark the brightness of AISe NCs, we compared their optical properties with those of commercial Ag2S NPs (Figure 2B-D). NIR images acquired under 800-nm excitation with the in vivo optical setup (see Experimental Section) show that AISe NCs have a much brighter luminescence compared with Ag2S NPs (Figure 2B).
Upon integrating the emission spectra acquired on these dispersions ( Figure 2B), we estimated that our AISe NCs are approximately 7.5 times brighter than Ag2S NPs under our experimental conditions. The molar extinction coefficient (ε) of AISe NCs was calculated to be roughly 3.5 times smaller than the one Ag2S NPs ( Figure 2D; see also Supporting Information for detail around these calculations). Therefore, since the brightness is defined as the product between ε and PLQY, according to the above observations the PLQY of AISe NCs should be approximately 25 times larger than the one of Ag2S NPs. Considering a value of 0.1% for Ag2S NPs, [22] this translates to a PLQY of 2.5% for AISe NCs. This value well agrees with an estimate of 3.5%, based on the 21% PLQY reported by Langevin et al. for AISe NPs in TCE [14] and an 80% intensity loss observed when passing from TCE to water ( Figure S2). Admittedly, optimization of the reaction conditions [10] or use of ultrafast photochemistry methods [23] allows producing Ag2S NPs whose PLQY is on par or higher than the values herein reported for AISe NCs. However, we should point out that our synthesis protocol did not undergo any optimization. Therefore, it is expected that further adjustment of the synthesis conditions and, possibly, exploitation of selective cation exchange [12] could push the brightness of AISe NCs to even higher values.
To complete the characterization of our AISe NCs, we also investigated the temperature dependence of their photoluminescence. As described in detail in the Supporting Information ( Figure S3), the thermal sensitivities retrieved for intensity-and lifetime-based thermometric approaches were 4.5 and 3% ºC -1 respectively at 37 ºC. These numbers are similar to those reported previously for Ag2S NP-based luminescent thermometers used for in vivo thermal sensing. [7b, 7d, 24] Moreover, as demonstrated above, AISe NCs are brighter and should therefore provide higher spatiotemporal resolutions. Thus, although the use of AISe NCs as thermal sensors lies outside the scope of this study, these nanoprobe hold great potential as biocompatible luminescent thermometers.
Encouraged by the low toxicity and high brightness of our NPs, we moved to in vivo measurements (Figure 3). We administered 150 μL of a 1.0-mg/mL dispersion of either AISe NCs or commercial Ag2S NPs in PBS 1x to two female CD1 mice via retro-orbital injection. For Ag2S NPs, the NIR intensity decreased at both locations to 40-45% of the maximum value over the course of 3 h. On the other hand, the signal coming from AISe NCs at the liver increased over time, and the intensity of the emission at the femoral vessels only decreased by 15% 3 h after injection. This prolonged imaging of the vasculature via FI is not common and it was ascribed to the unique combination of bright emission and surface modification of the nanostructures. Firstly, the encapsulation in PEGylated phospholipids is expected to minimize the formation of a protein corona around the AISe NCs. [25] In addition, the surface modification with maleimide was shown to induce a strong interaction with cell membranes via surface thiols. [26] Therefore, specific interaction with cellular components of the blood vessels (mainly epithelial cells) can contribute to the extended permanence time within the vascular system. [27] Motivated by the apparent advantage of using AISe NCs in vasculature imaging, we applied the principal components analysis (PCA) to the set of in vivo images obtained in the time frame of the experiment. [28] Our goal was to check if a sharper distinction between the veins/arteries and the surrounding tissues could be achieved. The rationale for using PCA is that this analysis converts a set of observations based on certain variables into another set of linearly uncorrelated new variables ordered according to their importance. [29] A detailed description of the procedures applied to our set of data is described in Supporting Information.
Hence the pixels containing static or slowly variable signals (highly correlated across the time frames) could be separated from the ones with faster dynamics. [30] Therefore, in some components (PCX, with X ≥ 1), certain organs could be more highlighted than others. Since the accumulation of the NCs was preferential at the liver, it comes as no surprise that this organ would be in the spotlight in the first components ( Figure S5). [5d] For our purposes, however, PC3 is particularly interesting. We selected a representative vessel in the right ventral region (white circle in Figure 3C, left) and analyzed the projection of the pixels corresponding to the vessel onto PC3 (Figure 3C, right). The thickness of the vessel can be retrieved from a Gaussian fit of the signal and considering the full width at half maximum of the fitting curve (260 μm in this case). What is even more interesting is that, aside from a difference in the magnitude (observed in PC1 and PC2 too), there is also a change in the sign of the signal arising from the vessel and the surrounding tissue. This means that for at least one of the new linearly independent variables, the behavior presented by the vessels was in direct opposition to the one of the surrounding tissues. For this reason, the vessels become markedly visible in this component of the PCA (Figure 3C). Moreover, although PC1 and PC2 did not highlight the vasculature as well as PC3, the contrast offered by them was still remarkable (Figure S5).
Overall, the analysis suggests a good capability of the AISe NCs to work as a contrast agent in enhanced angiography. The extended permanence time of AISe NCs in the vasculature make them suited for applications in angiography and detection of local deficiencies in blood perfusion caused by, for instance, stroke. [5g, 31] Further confirmation of the capability of the developed AISe NCs to act as effective vasculature markers, was obtained from ex vivo NIR images (Figure 4). Strong NIR signal was observed in correspondence of blood vessels in the organs. As it can be observed, the vasculature is highlighted in a wide number of organs, such as dorsal skin and muscle, the kidneys, and even the brain. Regarding the two latter organs, the internal vasculature of the kidneys is well resolved in their sagittal and transversals sections, while the characteristic Polygon of Willis is observed prior to any dissection in the ventral part of the brain. To further confirm this preferential accumulation in the vasculature, we dissected the aorta, observing an increased signal at the vessel compared to the superficial perivascular fat, similarly to what observed ex vivo for the femoral vessels in the right hindlimb. Depending on the analysed part of the animal, the contrast achieved with AISe NCs varies, with some vessels (e.g., those in the muscle of the hindlimb and in the skin) being better discerned than other (e.g., those in the kidneys), with an observed maximum ratio between the signal coming from the vessel and the surrounding tissue varying between 1.3 and 6.

Conclusion.
We have herein presented the first example of AgInSe2-based contrast agent for fluorescence imaging in the second biological window. The emission of these nanoparticles is shifted towards longer wavelengths compared to the usual emission featured by AgInSe2 due to an uncommon crystal structure, imparted by the Ag2S-seed mediated synthesis that we adapted from the literature. Upon transfer to water with the aid of maleimide-bearing phospholipids, nanocapsules composed of several AgInSe2 nanoparticles are formed, which feature high colloidal stability and lack of in vitro cytotoxicity. Their brighter emission and longer permancence time within the vascular system compared to commercial Ag2S nanoparticles make these nanocapsule an invaluable addition to the currently relatively limited library of contrast agents for near infrared fluorescence imaging. We also showed that the use of principal components analysis (PCA) allows observing more clearly the vasculature upon dynamical acquisition of fluorescence images after injection of AgInSe2 nanocapsules.
Since the improvement of the performance of fluorescence imaging passes through the development of brighter contrast agents, this study represents an important step in this direction.
While the developed AgInSe2-based nanocapsules already perform better than commercially available Ag2S nanoparticles, there is plenty of room for improvement via adjustment of the reaction conditions and exploitation of selective cation exchange procedures. Future research effort will be therefore directed along these lines. The viability of HeLa cells exposed to AISe NCs was analyzed by the MTT assay. [32] Twentyfour hours after appropriate treatments with AISe NCs, 3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide (MTT) solution was added to each well at a concentration of 0.5 ng/mL, and plates were incubated at 37 ºC for 2 h. The resulting formazan crystals were dissolved by the addition of DMSO and absorbance was measured at 540 nm. Cell viability was estimated as a percentage relative to the mean absorption obtained from control cells (not incubated with the AISe NCs; 100% viability).
Imaging system. NIR-II in vivo and ex vivo images were obtained with an InGaAs NIR camera (ZephIRTM 1.7). A fibre-coupled diode laser operating at 808 nm was used as excitation source (LIMO30-F200-DL808). The illumination power density was controlled by adjusting the diode current and set to 50 mW/cm 2 (a low power density that prevented excessive tissue heating and subsequent drying) and the laser height was adjusted to illuminate the whole animal.
A short-pass filter (Thorlabs FES 850) was placed immediately in front of the laser fiber to minimize specular and diffuse reflection effects. Additionally, 3 long-pass filters (Thorlabs FEL850) were used to minimize tissue autofluorescence signal.

Animal handling.
In vivo experiments were approved by the regional authority for animal   NPs at 800 nm (grey dot) obtained from ref [9a] .

Supporting Information
In vivo near-infrared imaging using ternary selenide semiconductor nanoparticles with uncommon crystal structure. Size distribution of AgInSe2 nanoparticles before transfer to water. 24 Comparison of emission intensity before and after transfer to water. 24 Concentration and molar extinction coefficient calculation for AgInSe2 nanocapsules. 25 Thermometric performance of AgInSe2 nanocapsules. 26 Temporal evolution of contrast and resolution. 27 Principal Components of dynamically acquired fluorescence images. 28

References. 28
Size distribution of AgInSe2 nanoparticles before transfer to water. Figure S1. Size distribution for AgInSe2 nanoparticles (NPs) obtained from TEM observations.
Comparison of emission intensity before and after transfer to water. Figure S2. Comparison of the emission spectra obtained under 790-nm excitation for AISe NPs in TCE (grey) and AISe NCs in water (purple).

Concentration and molar extinction coefficient calculation for AgInSe2 nanocapsules.
To determine the molar extinction coefficient of the AgInSe2 nanocapsules (AISe NCs), an extinction spectrum was recorded in water on a dispersion at a mass concentration ([mass]) of 0.25 mg/mL. To estimate the molar concentration of this dispersion, the following considerations were made.
The single nanoparticles composing the NCs have a tetrahedral shape with an edge (a) of 8 nm.
The volume of the single nanoparticle was calculated to be 60 nm 3 according to the following formula: To calculate the density of the material, we considered that orthorhombic AgInS2 (which is isostructural to our AgInSe2) has four molecular units in the unit cell (Z = 4). Considering a molecular weight (MW) of AgInSe2 of 380.6 and a unit cell volume (Vcell) of 0.4243 nm 3 (as obtained by the group of Vittel), 1 the density (d) of orthorhombic AgInSe2 was estimated to be 5.96 g/cm according to the following formula: = · · 1.66 · 10 −24 (S2) Where 1.66·10 -24 g is the value of the atomic mass unit (Dalton).
Using VNP and d, the mass (mNP) of a single nanoparticle was calculated to be 3.59·10 -19 g. Therefore, the molar concentration of the AISe NCs dispersion is given by: Where NA is the Avogadro's number (6.022·10 23 ). For a concentration of 0.25 mg/mL, the molar concentration found was of 1.15·10 -6 mol/L. The data obtained from the absorption spectrometer was divided by this value to obtain the graph in Figure 2D of the main manuscript.
Thermometric performance of AgInSe2 nanocapsules. Figure S3. A) Emission spectra and B) photoluminescence decay curves of a AISe NCs dispersions obtained as a function of temperature under 800-nm excitation. C) Integrated emission intensity and D) average lifetime as a function of the temperature of the dispersion. The solid lines are respectively the parabolic and linear fit of the data. E) Relative thermal sensitivity obtained from the fit of the data in C (purple line) and D (grey line).
The relative sensitivity was obtained according to the following equation: 2 Where Δ is the either the integrated intensity of the NIR emission or the average lifetime obtained from the integration of the normalized decay curves.
Temporal evolution of contrast and resolution. Figure S4. A) NIR image obtained 15 min after administration of AISe NCs and zoom-in of the area where the profile of the blood vessel was monitored (white dashed box). B) Profile of the intensity ratio between the signal from the vessel and from the surrounding tissue at three different time points (100 s, 30 min, 60 min). Temporal evolution of C) the width at half maximum (σ) obtained of the Gaussian fit of the profiles and D) the vessel-to-tissue intensity ratio. Note that a broadening of the peak can be correlated qualitatively to a loss of spatial resolution.

Principal Components of dynamically acquired fluorescence images.
For every pixel in the 150 fluorescence images (512×640 of resolution) acquired after the intravenous injection of NPs, the evolution of its intensity could be represented by a vector of dimension 150. Thus, each of the 150 coordinates would correspond to the intensity value at an instant t = (m-1)×t, where m is a natural number between 1 and 150 and t is the integration time used in the measurement. Under this perspective, the experimental dataset is no longer viewed as a list of 150 matrices of dimensions 512×640 but instead as a single matrix of 327680 lines and 150 columns. In other words, if the m-th luminescence image is: To pre-process the data, we compute the means, Ai, and the standard deviations, i, of each of the 150 columns. With these values we then calculate the following matrix: is the r-th coordinate value of the k-th eigenvector and ̅ is an image whose pixels values are all equal to Ar. We call Pk the projection of the experimental data on the k-th principal component. As representative examples, we include the first three projections in Figure S5.