1. Introduction

In the quest for the next generation of imaging bio-markers, successful probes have to prove to be non toxic, bright, stable against long term excitation, and able to generate a sharp contrast against background fluorescence. In all these respects, Second Harmonic Radiation IMaging Probes (SHRIMPs) are receiving an increasing attention as they also open a series of alternative detection possibilities simply not accessible with the present generation of fluorescent dyes and quantum dots.[1, 2, 3, 4] Second Harmonic Generation (SHG) is intimately connected to the symmetry of the crystal structure of a material and its chemical composition. Often on the basis of known properties of bulk nonlinear materials, a series of SHG-active inorganic nanocrystals have been synthesized by very diverse processes, and their optical response characterized by various research groups: KNbO₃[5], KTiOPO₄ (KTP)[6], ZnO[7], BaTiO₃[₈]. In particular Hsieh et al. have compared the doubling efficiency of an individual SHRIMP particle (which scales as the 6th power of its diameter) to the action cross-section of the two-photon excited fluorescence of a quantum dot, assessing values compatible to imaging.[8] We have recently proposed the use as bio-marker for nonlinear microscopy of iron iodate (Fe(IO₃)₃)[9] a novel nano-material, which does not exist in the bulk form.[10] One of the advantages of iron iodate with respect to other probes resides in its non-toxic chemical composition, which very likely will not prevent its future application for in vivo studies. The optical characterization of new SHRIMPs has been often accompanied by the proposition of novel detection schemes largely based on the coherent nature of SH emission[3, 8, 4] or on their polarization properties [9, 6, 8], but besides these sophisticated approaches, the very nature of SHG from nanostructures presents straightforward advantages for imaging through turbid tissues. In fact, the sub-wavelength size of the SHRIMPs implies the absence of phase-matching constraints and consequently their capability to frequency-double any incoming spectrum within their transparency range.[11, 12] From this descends the possibility of tailoring the excitation source to the sample absorption and scattering properties. Such extreme spectral flexibility can turn out to be crucial for increasing the imaging penetration depth, and for limiting energy deposition and autofluorescence by avoiding absorption bands of tissues.

In this work we show how Fe(IO₃)₃ SHRIMPS can be efficiently excited at two different wavelengths (800 nm and 1.55 µm) by two distinct ultrafast sources. We also demonstrate how in an epi-detected SHG measurement through a turbid sample, signal collection can be severely affected by the choice of excitation wavelength. To this end, we first use artificial tissue phantoms made of submicrometric polystyrene spheres, and subsequently a section of murine liver.

 

Fig. 1. Left: NIR and IR light (red arrow) are focused by the objective onto the SHRIMP particle, placed on the upper side of a system of two microscope substrates of thickness t delimiting a diffusive layer (DL) (polystyrene beads suspension or liver section) of variable thickness d. The backwards traveling fraction of SH emission (blow arrow) is epi-collected by the same objective. Images are reconstructed by scanning sample position along x and y axes. Right: Example of the approach used in the Monte Carlo simulation. The incoming photon (red dashed line) is considered effective for SH excitation because, after two scattering events, it enters the SHRIMP region (red circle) with a small angle θ_IN with respect to normal incidence. On the other hand, the outgoing SH photon (blue dashed line) is not accounted for the signal calculation, as it exits the sample with a large angle exceeding θ_OUT, i.e. the collection capability of the objective.

Download Full Size | PDF

2. Experimental

The laser sources employed in the study are a Ti:Sapphire oscillator at 800 nm (Femtosource, Femtolasers) and an Erbium fiber laser at 1550 nm (EFO-150, Avesta) that in the following will be referred to as NIR and IR, respectively. The polarization of the NIR was matched to that of the IR by an half-wave plate, and their average power adjusted to coincide at 1.3 mW at the input of the microscope objective. As sketched in the left of Fig. 1, the beams were focused by a 100×N.A=1.3 or alternatively by a 40×N.A=0.6 objective (77% transmission for NIR and for 50% IR). The latter was used for all the layer transmission measurements because of its longer working distance (2 mm). Using the NIR excitation and the 100× (resp. 40×) objective, the spatial resolution of the set-up was 0.8 µm (resp. 1.8 µm). The harmonic signal was epi-collected by the same focusing objective, separated from the fundamental by a dichroic mirror, and further selected by an interference filter before being detected by a photo-multiplier (Hamamatsu H5701-51 for NIR excitation and H6780-20 for IR) and fed into a lockin amplifier. Alternatively, the spectrum was resolved by placing a scanning monochromator (1 nm resolution) in front of the photomultiplier. Scattering tissue phantoms were prepared by placing spacers of calibrated thickness (d=90, 180, 300 µm) between two microscope slides (170 µm or 1 mm thickness) and filling the gap with a water suspension of 0.1 µm polystyrene nanospheres at a concentration of 45.5 particles/µm³ (Polysciences). Mouse liver was first fixed with a solution of Phosphate buffered saline (PBS)-4% paraformaldehyde at 4°C overnight, then washed in PBS and incubated in a solution of PBS - 30% sucrose. Tissue was then embedded in OCT medium for freezing and stored at minus 20°C before cryosectionning at a nominal thickness of 20 µm.

3. Monte Carlo simulation

For simulating the epi-detection of the signal from a SHRIMP embedded in a turbid tissue, we employed a Monte Carlo code of light transport in multilayered samples, which already proved very successful for a variety of studies.[13] The simulations assume an infinitely narrow photon beam, perpendicularly incident on a tissue layer supposed much wider than the spatial extent of photon distribution. The model is restricted to a cylindrical symmetry by assuming an optically isotropic medium. At every computation step, a photon, which is treated as a classical particle, neglecting polarization effects, has a certain probability of being absorbed or scattered. These probabilities are determined directly from the macroscopic values of scattering efficiency (µ_s), absorption efficiency (µ_a), and anisotropy (g) calculated applying Mie theory for a suspension of nanospheres in water [14], using the real and imaginary reflective indices for polystyrene provided by ref.[15]. Alternatively, for the liver sample, we used the values of µ_s, µ_a, and g experimentally determined by Parsa et al..[16] The input parameters used for the simulation are summarized in table 1.

Table 1. Monte Carlo Input Parameters

View Table

As indicated on the right of Fig. 1, the effective excitation for a SHRIMP placed after a scattering layer of thickness d for an excitation wavelength ω_i (i=800, 1550 nm), was obtained by integrating the fraction of photons α(ω_i,d,R_IN,θ_IN) exiting from the substrates+layer system through a limited circular area of R_IN=1 µm diameter (comparable with the experimental focal spot) with propagation axis deviating θ_IN < 10° from the incident direction. This way we ensured to limit the excitation process to ballistic photons, which conserve the temporal structure (i.e. photon density) of the incoming pulse.[17] This quantity was then squared to account for the nonlinear power dependence of SHG. An independent simulation was successively run to calculate the fraction of SH photons β (2 ·ω_i,d,R_OUT,θ_OUT) that can reach the microscope objective after traveling backwards through the substrate/scattering system. Contrary to excitation, in this case, the transmitted photons were integrated over a large exit area R_OUT and over an angular range θ_OUT calculated by taking into account the objective N.A. and the refractions of the SH photons at the substrate/sample and substrate/air interfaces. The epi-detected signal I_SHRIMP as a function of the incident laser intensity I was finally determined as I_SHRIMP=(I · α)² · β.

4. Results and discussion

SHRIMP particles have so far been investigated only by means of Ti:Sapphire lasers, so we first characterized their response to IR excitation by reporting in Fig. 2 the up-converted spectrum generated by a single nanocrystal. The fundamental Erbium laser spectrum is shown in the inset for reference. The second harmonic peak between 730 and 850 nm is the predominant structure. Note that the tiny spectral dip at 780 nm correlates well with the minimum appearing in the calculated frequency-doubled spectrum (dotted line) and that the original bandwidth is completely up-converted. The peak at 520 nm corresponds to the third harmonic (TH) emission, which, although much weaker, is clearly visible in the semi-logarithmic plot. TH was always co-localized with SH, evidencing that it is an effect genuinely related to the presence of a nanocrystal.

 

Fig. 2. Semi-logarithmic spectrum of the second and third harmonic emission (continuous line) from a single Fe(IO₃)₃ nanocrystal excited by the IR laser. The dotted lines correspond to the calculated frequency doubled and tripled fundamental IR laser spectrum (shown in the inset).

Download Full Size | PDF

In Fig. 3, panels (a) and (b) present a comparison of the same region of a sample of randomly dispersed nanocrystals obtained by drying a drop of Fe(IO₃)₃ filtered solution on the upper microscope substrate as illustrated in Fig. 1. From independent atomic force microscope measurements of samples prepared in the same way, we know that filtered Fe(IO₃)₃ particles (~30 nm) have a tendency to aggregation on microscope substrate and their actual size on the sample is 80±40 nm.[9] The images were realized by detecting the SH signal from the particles excited first by the IR and successively by the NIR laser. A quick inspection indicates that all the particles are retrieved using both excitations and they show comparable relative intensities. The darker halo surrounding the particles in the IR-excited scan (Fig. 3(b)) can be originated from diminished imaging performances of the objective, which is optimized for the visible region, but it can also be ascribed to interference effects among the SHG radiation generated at different location within the focus, as previously observed for other coherent microscopy signals.[18]

The different light transport properties of NIR and IR excitation and their corresponding SH (400 and 775 nm) through a strongly scattering medium were then investigated using calibrated tissue phantoms. In the series of SH scans reported in Fig. 3 (c)–(f) and (h), the different rows correspond to different layer thickness, each containing two images of the same sample region excited by the two laser sources: NIR (left) and IR (right). Note that the images on different rows do not correspond to the same sample region. For a sample thickness of 90 µm [(c), (d)], two particles are present on the scan. The left one is characterized by a much weaker signal and its size is comparable with the lateral resolution of the microscope.

The poorer resolution of this particle in the IR scan is partially attributed to the larger diffraction limit at 1.55 µm, as well as the aforementioned reduced performances of the objective at this wavelength. The right SHRIMP is probably a larger particles aggregate, with a size exceeding the set-up resolution. At 180 µm [(e), (f)], the comparison between the IR scan evidence the presence of a SHRIMP which does not appear in the corresponding NIR scan, although several NIR-excited scans were performed systematically varying focusing and detection parameters. A similar observation resulted from the measurement performed with the diffusive layer of 300 µm thickness (Fig. 3(h)), even in this case we were not able to detect any particle in the NIR-excited image.

 

Fig. 3. Comparaison of nanocrystals samples illuminated by NIR (left column) and IR (right column) laser with no diffusive layer [(a), (b)], and with diffusive layer of polystyrene nanospheres suspension in water of 90 µm thickness [(c), (d)], 180 µm thickness [(e), (f)], and 300 µm [(h)]. Panel (g): Result of a Monte Carlo simulation showing the intensity of the epi-detected signal (I_SHRIMP) as a function of the penetration depth for NIR (▫) and IR excitation (∘). Experimental values measured on microscopic Fe(IO₃)₃ structures by NIR (▪) and IR (∙) are also reported on the plot as additional reference.

Download Full Size | PDF

 

Fig. 4. Comparison of excitation of Fe(IO₃)₃ nanocrystals through a 20 µm murine liver sample: (a) NIR excitation and (b) IR excitation. (c) Monte Carlo simulations of the epidetected signal fom a SHRIMP excited and epi-detected through a layer of murine liver as a function of thickness calculated for various excitation wavelengths: NIR (▫), IR (∘), and 1320 nm (▪). Inset: Semi-logarithmic representation of the same dataset.

Download Full Size | PDF

The corresponding simulation are presented in Fig.3 (g). One can observe that for d <80 µm, the NIR-generated signal is expected to be larger than the IR one. This finding can be correlated to the slightly higher contrast shown by the NIR scan in the comparison between panels (c) and (d). For thicker samples, the SH signal collected under IR excitation becomes dominant, and indeed already for d=180 µm, no NIR-excited signal could be detected. The dashed horizontal line corresponds to an upper estimate for the detection limit, determined by the absence of signal for I_SHRIMP(ω800,180 µm). We observe that its position is consistent with the detection of the SHRIMP at d=300 µm (Fig.3 (h)), whose simulated intensity lies above this threshold. To further assess the agreement between calculations and experiment, we measured the epidetected SH signal from a micrometric Fe(IO₃)₃ structure excited by NIR and IR. The relative intensities of the corresponding datapoints are well superimposed to the curves in Fig. 3(g) and corroborate the correctness of the numerical approach within the thickness range investigated.

Given that the performance of SHRIMP detection cannot be simply ascribed to the deeper penetration of longer wavelengths, but are set by the interplay between excitation and backward detection of the SH signal, we extended the investigation by substituting the sample with a 20 µm thick murine liver tissue. In this case, in fact, rather strong scattering is accompanied by the specific spectral response of the tissue, dominated by heme proteins absorption around 400 nm and water absorption around 1.4 µm.[16] Panels (a) and (b) of Fig. 4, report the results of NIR- and the IR-excited images, respectively. The intense spot on the lower left corner of both images is a reference SH signal generated by a micrometric Fe(IO₃)₃ structure, placed on the substrate close to the SHRIMP nanocrystal to verify the correct experimental settings for both measurements. As in the preceding comparison, the SHRIMP signal could be easily epidetected exclusively in the IR case. In Fig. 4(c), the simulated signal intensity again indicates that IR-excitation (∘) remains the most favorable choice for this representative biological sample, not only with respect to NIR (▫), but, contrary to our expectations, also with respect to 1320 nm excitation (▪), which was also simulated as it presents no spectral overlap with the tissue absorption for both excitation and SH. From calculations it appears that, even though scattering properties in the excitation process assume a major role in affecting the signal intensity by setting the fraction of ballistic photons reaching the SHRIMP with enough power density to nonlinearly excite it, the stronger diffusion of bluemost wavelenghts results the dominant factor at the origin of the intensity differences as penetration depth increases. Clearly, to identify the most efficient excitation option, signal intensity is not the unique criterion, as sample heating by water absorption should also be taken into consideration for long-term measurements.

In conclusion, we have demonstrated the possibility to excite Fe(IO₃)₃ SHRIMPs with two distinct laser sources and we have experimentally evidenced and rationalized by numerical simulations how the wavelength flexibility inherent to non-phase-matched SHG can be a crucial factor to increase sample penetration for nonlinear microscopic measurements based on these markers.