1. Introduction

Nonlinear microscopy is a manifold of imaging techniques that leverage laser technology to reveal the chemical composition and structure of a specimen. In biomedical imaging, nonlinear microscopes exploit femtosecond (fs) laser pulses to target intrinsic biomolecules of cells and tissues, chemical species that respond to the incident light by emitting radiation. Because this radiation is akin to a molecular fingerprint, nonlinear microscopy enables mapping the distribution and concentration of specific biochemical species in a label-free and nondestructive fashion, providing information on the tissue microenvironment that leads to better diagnostics and assessment of therapeutic interventions.

The continuous progress in laser technology has pushed forward the advance of nonlinear microscopy. As early as 1982, Duncan et al. used dye lasers to build the first coherent anti-Stokes Raman microscope [1]. Later, in 1990, fs pulses from a dye laser enabled Webb and colleagues to develop the earliest two-photon excited autofluorescence microscope (2PAF) [2]. The advent of mode-locked lasers based on solid-state gain materials, notably the Ti:Sapphire [3] and Nd:YLF [4], enabled tissue explorations using three-photon excited autofluorescence (3PAF) [5], second-harmonic generation (SHG) [6], and third-harmonic generation (THG) [7]. Although these laser technologies revealed the potential of nonlinear microscopy, they also constrained it to research laboratories – dye lasers are a cumbersome, unstable, and toxic technology, while the complex and large footprint of solid-state laser cavities prevented the practical translation and wide adoption of nonlinear imaging platforms.

The emergence of fiber-based lasers, with their cost-efficiency, compactness, and high-peak power at a low repetition rate [8–11], enables new routes for the translation of nonlinear microscopes to clinics and wet laboratories [12]. However, fiber-based lasers have narrow bandwidth and limited wavelength tunability, deficiencies that prevent the simultaneous exploration of biochemical species that expose the chemistry and structure of tissues, reducing the quantitative information extracted by nonlinear microscopes.

A solution to this problem is supercontinuum generation: a complex nonlinear optical process that stems from the distortion – by a fs pump laser – of the refractive index of a transparent material. This refractive index distortion triggers several nonlinear phenomena – self-phase modulation, cross-phase modulation, stimulated Raman scattering – that collectively alter the phase, amplitude, and frequency of the pump beam, resulting in a spatially and temporally coherent beam with an excellent spatial profile and the same noise level as the pump beam but exhibiting a much broader spectrum [13] – laser properties that are ideal for driving nonlinear microscopy.

Several research groups have leveraged supercontinuum generation to drive multiphoton nonlinear microscopes, mainly using light beaming from photonic crystal rods and fibers [14–18]. Notably, Xu’s group launched 1,550 nm light from a fiber laser to a commercially available 36-cm photonic-crystal rod to produce radiation at 1,700 nm, light that enabled in vivo 3PAF imaging of subcortical structures within an intact mouse brain [19].

By pumping a photonic crystal fiber with an Ytterbium-based fiber laser (oscillating at 20 MHz, centered at 1030 nm and a bandwidth of 30 nm), Boppart’s group routinely produced white light supercontinuum covering the spectral range within 900-1200 nm [20]. The spectral window within 1080-1140 nm enabled simultaneous label-free autofluorescence multi-harmonic (SLAM) microscopy, a powerful imaging technique that co-registers 2PAF, 3PAF, SHG, and THG signals from pristine tissues, even in vivo [21,22]. A SLAM microscope leverages the 2PAF from flavin adenine dinucleotide (FAD) and the 3PAF from reduced nicotinamide adenine dinucleotide (NADH) and reduced nicotinamide adenine dinucleotide phosphate (NADPH) to gain functional metabolic information of the tissue microenvironment, while also simultaneously exploiting the SHG from collagenous structures and the THG from refractive index inhomogeneities to reveal interfacial structural features of tissues.

An elegant approach for multimodal nonlinear microscopy was recently introduced by the Yoo group [23]. The authors extracted 2PAF, SHG, and coherent anti-Stokes Raman scattering (CARS) signals from biological specimens using a Ti:Sapphire laser and a single detector. To increase the efficiency of the nonlinear optical processes, they reduced the repetition rate of the laser with a pulse picker, thereby increasing the pulse-peak power. The modulated Ti:Sapphire beam was then split into two branches. The first branch – the source for SHG, 2PAF, and the pump for CARS – was dispersion-corrected and directly coupled to a microscope, whereas the second branch pumped a PFC to produce Stokes radiation for CARS. Upon signal generation, a set of dichroic mirrors and filters spectrally separated the SHG, 2PAF, and CARS signals. Each of these homodyne signals was sent to specific optical fibers with different lengths. The path mismatch between the optical fibers set a temporal delay between the nonlinear signals upon arrival to a single detector, a delay that allowed a fast digitizer to separate them in time, thus enabling multiplex chemical imaging.

Recently, Polli’s group used an Ytterbium fiber laser (oscillating at 2 MHz, centered at 1035 nm with a 10 nm bandwidth) to produce white light supercontinuum from a bulk crystal. This approach, widely used in ultrafast pump-probe spectroscopy [24], allowed the generation of broadband red-shifted Stokes pulses for coherent Raman microscopy. This supercontinuum, coupled with a narrow band pump beam, allowed for probing the molecular vibrations of cells and tissues, thereby revealing their chemical composition [25–27]. Thus, bulk nonlinear crystals [28–31], coupled with high-power fiber lasers, might lead to simple and robust light sources for multiphoton microscopy, providing not only an alternative to photonic crystal fiber technologies but also for a practical design for ready integration into clinical applications.

In this paper, we demonstrate the first application – to the best of our knowledge – of supercontinuum generation from a bulk crystal as the driving source for multiphoton microscopy. We show that the white light supercontinuum from a Yttrium Aluminum Garnet (YAG) crystal [32] offers a robust, stable, alignment-free, and cost-effective approach for generating the driving beam for a nonlinear microscope, emerging as an appealing alternative to photonic crystal fibers and rods. Additionally, we present the optical properties of the supercontinuum from the YAG crystal, particularly its spectrum, beam profile, and noise performance. We then present a multimodal nonlinear microscope driven by fs-pulses from the YAG supercontinuum – radiation that was compressed to 50 fs full width at a half maximum (FWHM) by a Fourier-transform pulse shaper. This system – called the YAG-SLAM – was spectroscopically characterized and adapted for label-free imaging of biological specimens, a platform that successfully mapped the tissue microenvironment of freshly excised rodent tissues in a label-free and non-destructive fashion. Because it shows long-term stability, a reduced footprint, and turn-key performance, YAG-SLAM validates the use of supercontinuum from bulk crystals for multiphoton biomedical imaging, paving the way for the translation of sophisticated yet simple nonlinear microscopes toward clinical applications and settings.

2. System description and performance

2.1 YAG-SLAM microscope

The YAG-SLAM system is depicted in Fig. 1 (a). The microscope has an inverted geometry, and a platform that uses a mechanical stage (U-761.25, Physik Instrumente) for coarse translation of the sample, enabling large (over 350 µm) displacements for recording different fields of view. For fine raster scanning, the YAG-SLAM microscope uses a pair of galvanometer mirrors (6215 H, Cambridge Technologies) conjugated with a 4f telescope (L₅ & L₆). Four dichroic mirrors (DM₁: 705 nm, DM₂: 409 nm, DM₃: 560 nm, DM₄: 495 nm) guide the nonlinear signals emitted at the sample plane to four photon counting photomultiplier tubes (PMTs, H742, Hamamatsu). To achieve spectral specificity, YAG-SLAM has specialized optical filters placed prior to each PMT– F₁: 390 nm short pass; band-pass filters F₂: 450/100 nm, F₃: 555/30 nm, and F₄: 610/60 nm, with the second number indicating their bandwidth. Spherical lenses L₇-L₁₀ focus the nonlinear signals on their corresponding PMTs. The objective lens used in YAG-SLAM is the LUMFLN60XW (water immersion, 60× magnification, numerical aperture = 1.1, working distance = 1.5 mm) from Olympus. Thus, YAG-SLAM is a multichannel imaging platform with channels 1-4 designed for simultaneously detecting THG, 3PAF, 2PAF, and SHG, respectively [21,22].

 

Fig. 1. The YAG-SLAM microscope platform. a) A schematic representation of the YAG-SLAM system. b) Scheme of the optical set-up used for optimum supercontinuum generation. c) Effective optical path of the Fourier-transform pulse shaper of YAG-SLAM. L_x: lens, GVS: galvanometer scanner, F_x: filter, LP: Low pass filter, YAG: Yttrium Aluminum Garnet, PMT: Photomultiplier tube, SLM: Spatial light modulator, SP: Sample, MO: Microscope objective.

Download Full Size | PDF

Our emission detection system uses four photon-counting PMTs, which are connected to counters of a DAQ card (NI PCIe 6351). The software reads the number of counts per dwell time and uses the accumulated number to generate an image [33]. Although YAG-SLAM currently operates with a photon counting scheme, we would like to mention that it is also possible to couple this method with other detectors such as analog and hybrid-PMTs, single photon avalanche diodes (SPAD), or silicon-photo multipliers (SiPM). To further increase the signal-to-noise ratio and imaging rate, YAG-SLAM could exploit a modulation transfer technique, particularly lock-in [34–36], a resonant circuit [37], or a box-car integrator [38].

2.2 Optical properties of the YAG supercontinuum

The light source of YAG-SLAM starts with a Ytterbium fiber-based laser (Satsuma HP, Amplitude), delivering 350 fs pulses at a repetition rate of 2.5 MHz centered at 1035 nm with a bandwidth of 30 nm (see the red curve in Fig. 2(a)). A 5 W fraction of this pulse train is passed through a metallic iris to produce a beam waist of 3 mm, which is then telescoped-down to 1.5 mm and focused to a 10 mm-thick YAG crystal by a 75 mm focal length lens. This configuration yields a beam waist of 60 µm at the center of the YAG crystal, resulting in a Rayleigh range of 10 mm [39], a length that perfectly matches the longitudinal dimension of the crystal, see Fig. 1(b). This reduction of the fundamental waist not only relaxes the focusing condition, thereby extending the lifetime of the crystal, but also increases the red-shifted spectral components of the supercontinuum, see blue curve in Fig. 2(a) [31,40,41]. This spectral window exhibits a remarkable beam shape, characterized by a Gaussian profile with outstanding spatial stability and high symmetry (σ_x/σ_y= 99.8%), see Fig. 2(b). Finally, the YAG supercontinuum does not show any polarization scrambling (ellipticity < 0.11), a result that not only is consistent with previous studies [32,42] but also emphasizes the versatility and potential of the source.

 

Fig. 2. Optical properties of the YAG supercontinuum. a) Spectrum of the fundamental beam (red) and that of the supercontinuum in the visible (green) and in the NIR (blue). b) Beam profile of the YAG supercontinuum in the NIR. This beam profile was acquired prior to the pulse shaper. c) Autocorrelation of the NIR of the YAG supercontinuum. d) The Relative Intensity Noise (RIN) of the pump beam (red) and the YAG supercontinuum in the NIR (blue dots). The inset shows the difference between the RIN of the pump and the NIR parts of the supercontinuum.

Download Full Size | PDF

In addition to NIR wavelengths, the YAG supercontinuum contains visible white light – a colorful cone of radiation that propagates collinearly with the NIR. This visible radiation has a lower power (< 1/500) relative to the NIR, but covers a spectral range within 550-900 nm, see green curve in Fig. 2 (a). To remove the visible light, we sent the YAG supercontinuum through a long-pass filter (LP02-1064RE, Semrock). Because the edge wavelength of this optical element is at 1064 nm, we slightly rotated it to transmit all the light within 1030-1150 nm, effectively eliminating all the visible components, resulting in about 1.5 W at the spectral window of interest, optical power that translates to 30% conversion efficiency at 1080-1140 nm.

Contrary to other nonlinear optical processes for up and down conversion, e.g., Type 0 phase matching crystals [43], the nonlinear interactions leading to supercontinuum generation within the YAG crystal do not produce additional noise components. Instead, the supercontinuum follows the power spectral density of the fundamental beam, producing a virtually identical relative intensity noise (RIN) trace – the RIN is a metric for how noisy a laser source is at a given modulation frequency and is defined as the power noise δ(f) normalized by its mean power [44,45]. A comparison of the RIN spectrum of the fundamental beam with that of the NIR supercontinuum from the YAG crystal is shown by the red and blue dots in Fig. 2(d) and their difference in the inset therein. This finding suggests that the noise in the supercontinua from bulk crystals is set by the driving source, implying that the supercontinuum from these materials pumped by lower-noise lasers enable applications that demand both broadband and low-noise excitation, e.g., microscopy based on interferometry [46,47], transient absorption [48,49], or coherent Raman scattering [50,51].

It is worth mentioning that under certain experimental conditions, the supercontinuum from nonlinear crystals can even be less noisy than the driving laser, while under other experimental conditions, the supercontinuum can be exceedingly unstable, flickering, and exhibiting a total lack of light generation [41]. This flickering occurs when the nonlinear crystal is misaligned, or when it is pumped with either too much or too low power. In the high-power regime, a multi-filament structure appears within the crystal, while in the low-power regime, the energy deposited in the crystal does not exceed a threshold to trigger the nonlinear effects that lead to supercontinuum. Therefore, care must be taken to prevent noisy supercontinuum and secure optimum performance.

The YAG crystal exhibits less generation efficiency relative to nonlinear fibers. This lower efficiency is accompanied by a narrower supercontinuum spectrum, an energy distribution with the highest power spectral densities around the pump wavelength. In this contribution, for example, the 1080 nm, 1125 nm, and 1140 nm bands of the red curve in Fig. 2(a) contain about 20%, 7%, and 4% of the total power of the light within 1080-1140 nm – the spectral window that YAG-SLAM uses to drive the nonlinear signals for imaging. In our experiments, the average power spectral density of the YAG supercontinuum within this range resulted in 20 µW/nm, a figure of merit that is 50 times smaller than that attained with a PCF pumped by an 80 MHz Ytterbium laser [20]. Nevertheless, the YAG crystal can successfully drive intrinsic nonlinear signals from biological tissues, thus enabling label-free chemical imaging.

2.3 On the implications of low-repetition rate lasers in nonlinear microscopy

While high (> 20 MHz) repetition rate lasers have traditionally driven nonlinear microscopes, they might not lead to optimal imaging performance and can even be detrimental to sample integrity. This unsuitability stems from the fact that the generation efficiency of nonlinear signals is proportional to the pulse peak power but inversely proportional to the repetition rate [38]. To effectively produce nonlinear signals, a microscope with a high repetition rate laser needs to deposit elevated average powers on the sample, power densities that might not only be phototoxic but also obliterate the specimen. This high average power limits the pulse peak power, thus reducing the generation efficiency of nonlinear signals and the signal-to-noise ratio.

By contrast, a low repetition rate light source might be more suitable for nonlinear microscopy due to three key aspects. 1) A low repetition rate laser – operating at the same average power of its high repetition rate counterpart – yields higher peak power per pulse, simultaneously applying less average power on the specimen. This average power reduction averts sample photodamage. Additionally, a low repetition rate, such as the 2.5 MHz used in this work, results in prolonged temporal delays (in our case 0.4 µs) between sequential pulses, a delay that gives the sample enough time for thermal energy dissipation, thus reducing photothermal damage. 2) By driving nonlinear signals more efficiently, a low repetition rate laser guarantees high-quality chemical imaging of biological specimens. 3) The higher energy pulses delivered by a low repetition rate laser enables the generation of white light supercontinuum in bulk crystals, bypassing the conventionally used PCF system along with its propensity to misalignment and photodamage. Therefore, low repetition rate lasers for nonlinear microscopy not only simplify the system and deliver high-quality images but also are gentler with the samples.

2.4 Dispersion compensation of the YAG supercontinuum

Because the generation efficiency of nonlinear optical signals is inversely proportional to the temporal distribution of the excitation pulses [2,6,52], YAG-SLAM requires pulses as short as possible at the sample plane. This condition maximizes the strength of the nonlinear contrasts, increasing the signal-to-noise ratio. However, the second-order dispersion introduced by the optical elements of the microscope – especially the microscope objective – stretches the optical pulses, reducing the generation efficiency of nonlinear signals [53]. Therefore, to achieve peak performance, YAG-SLAM pre-compensates the dispersion of the pump pulses. To this end, the YAG supercontinuum travels through a Fourier transform pulse shaper (MIIPS Box640, Biophotonics Solutions). Figure 1 (c) illustrates the effective optical path of the YAG supercontinuum within the pulse shaper.

This device, which slightly deviates from Weiner’s original design [54], consists of a diffraction grating (490.4 lines/mm), a spherical mirror (radius of 400 mm), and a programmable phase mask (a spatial light modular with 640 linearly-arrayed liquid crystal elements). The first propagation through the grating angularly disperses the spectral components of the YAG supercontinuum, while the spherical mirror focuses them on the spatial light modulator. An amplitude mask spatially filters out all the spectral components of the supercontinuum except those within 1030-1150 nm. The liquid crystal elements of the spatial light modulator tweak the phase of each of these spectral components, introducing an overall negative group delay dispersion (GDD) of circa –8000 fs². After the spatial light modulator, a second propagation through the spherical mirror and the grating recombines the spectral components of the shaped supercontinuum. The applied GDD on the YAG supercontinuum effectively compensates for the dispersion introduced by the optics of our imaging platform, producing optical pulses with a full width at half maximum of ∼ 50 fs at the sample plane – see the autocorrelation trace in Fig. 2(c). Note that other dispersion compensation devices – e.g., prism and grating pairs, chirped mirrors [55–57] – would successfully replace the Fourier transform pulse shaper presented in Fig. 1 (c), maintaining performance but reducing the overall cost of the system.

To quantify the effects of dispersion compensation, we imaged a reference sample – a commercially available slide (FluoCells prepared slide #2, Invitrogen) containing bovine pulmonary artery endothelial (BPAE) cells labeled with DAPI, BODIPY FL goat anti-mouse IgG, and Texas Red-phalloidin. These labels have an excitation/emission peak at 358 nm/461 nm, 505 nm/513 nm, and 591 nm/608 nm, respectively. Through multiphoton excitation, the YAG supercontinuum successfully drives these labels, allowing YAG-SLAM to measure their emission with channels 1-3. We located a bubble of distilled water near labeled cells to determine the effects of pulse-shaping on the generation efficiency of nonlinear signals below 390 nm. The interface between the distilled water droplet and the coverslip produced intense ultraviolet radiation through THG [58], light that is readily detected by the YAG-SLAM channel 4. The top row in Fig. 3(a) shows images of the reference sample without applying the phase mask, while the bottom row shows the same field of view but launching dispersion-compensated pump pulses. Note the increased photon counts on all the YAG-SLAM detection channels upon dispersion compensation, see Fig. 3(b). It is clear from these results that the applied GDD produces Fourier-transformed limited pulses that efficiently boost the generation of nonlinear signals, leading to enhanced contrasts and images with a higher signal-to-noise ratio.

 

Fig. 3. Reference sample for dispersion pre-compensation assessment – a labeled slide containing BPAE cells labeled with DAPI, BODIPY FL, and Texas Red. a) Images of the reference sample with YAG-SLAM individual channels. The top row shows data acquired with uncompensated pulses whereas the bottom row shows the same field of view but after dispersion compensation of the YAG supercontinuum. Imaging settings: 900 × 900 pixels; Pixel dwell time: 22 µs; Scale bar: 100 µm. Power at the sample plane: 3.5 mW. b) Graph summarizing the effects of dispersion on each individual channel.

Download Full Size | PDF

2.5 On the advantages of bulk nonlinear crystals over PCFs for driving nonlinear microscopes

Although the premise of this work is to demonstrate the potential of the supercontinuum from bulk crystals for biomedical imaging, we observed some aspects that make the YAG crystal more appealing to the widely used PCFs in our lab, key aspects that we summarize here.

Cost and time efficiency. PCFs require matching the pump waist to the fiber core. This mode matching demands tight focusing, a configuration that leads to high energy fluences and photodamage of the PCFs. This damage is negatively reflected on the performance of the fiber, leading to a progressive reduction of coupling efficiency, narrowing of the supercontinuum spectrum and degradation of its beam profile. For example, our records indicate that a 15 µm core PCF (LMA-PM-15, NKT Photonics) pumped with 160 nJ pulses (350 fs width and centered at 1030 nm) drops its coupling efficiency to 60% after only 77 hours of usage. Below this coupling efficiency, SLAM imaging is not possible and the fiber needs to be replaced – a cumbersome process that in our lab takes place over 10 times a year. Because each fresh segment of fiber costs around $\${100}$ and requires over 5 hours of specialized work – cutting, placing, and aligning the fiber, followed by system optimization – fiber photodamage has negative implications on the capital and work efficiency of the lab.

By contrast, the YAG crystal produces optimal supercontinuum via loose focusing of the pump, an arrangement that prevents the deposition of high energy densities on the crystal, thereby avoiding photodamage and securing spectra with consistent optical properties for over a year. Additionally, the diameters of commercial YAG crystals are as large as 2.54 cm, dimensions that are orders of magnitude larger than the spot size of the focused pump (consider, for example, the 60 µm pump waist discussed in Section 2.2). For instance, if a single spot on the YAG crystal is damaged due to long usage time (mainly because of dust or other irregularities on the YAG), the user simply needs to move the crystal to another unused region, easily recovering the supercontinuum.

Our empirical observation is supported by the laser-induced damage-thresholds reported in the literature: A polished YAG crystal tolerates up to 90 J/cm² before breaking [59,60], while (bulk) fuse silica undergoes damage at 6-26 J/cm² [61,62]. Although the historical literature emphasizes that numerous parameters of the pump – wavelength, repetition rate, pulse duration – influence the damage threshold of dielectric materials, there are plenty of reports that attest to increased laser-induced damage by reducing the spot size [61–66]. Thus, since the cost of a 10-mm long YAG crystal is about $\${300}$ and lasts longer than a PCF, this nonlinear crystal is not only more cost-effective than PCFs but also more suitable for long-term applications and commercial systems.

Robustness against misalignment. Optical systems invariably suffer from misalignment. Regardless of its origin [67], misalignment changes the design trajectory of a beam, causing tightly focused light to miss its target. The impact of misalignment grows in high-precision applications, such as supercontinuum generation in PCF – structures with circular input facets of diameters ranging from 1-100 µm. Since squeezing all the pump photons into the PCF core is crucial for optimal supercontinuum generation, any slight misalignment of the pump beam, or deviation of the PCF from the focal spot, reduces the coupling efficiency and can lead to fiber damage if the pump hits the cladding. Therefore, PCF-based nonlinear microscopes require time-consuming re-alignment before and during an experiment, which demands expensive actuators, such as automatic high-precision x-y-z fiber mounts or piezoelectric mirror mounts. Conversely, a nonlinear microscope using a nonlinear crystal with window diameters as large as 5-25.4 mm for supercontinuum generation is virtually immune to severe misalignment. This robustness against misalignment not only increases the productivity of the workflow but also drops the cost and complexity of the system.

Simplicity. Nonlinear optical fibers efficiently produce broad supercontinuum radiation, coherent light that can cover up to two octaves (370–1545 nm) [68]. However, nonlinear fibers, including PCFs, achieve this extreme broadening at the expense of tight focusing, which – in our opinion – is the main disadvantage of nonlinear fibers. A PCF user needs to align the fiber core with the pump, ensuring that all the pump photons fall within this micrometric structure. This mode matching requires thick lenses, lens- and fiber mounts with nanometric resolution and a burdensome alignment. Instead, a nonlinear crystal only requires telescopes with relatively thin lenses and loose focusing. Loose focusing not only simplifies the set-up and prevents adding dispersion to fs-pulses but also averts misalignment and protects nonlinear crystals against photodamage. Therefore, the YAG-based supercontinuum achieves sophistication through simplicity, making bulk crystals an appealing source for nonlinear microscopy.

3. YAG-SLAM for label-free imaging

3.1 Calibration of YAG-SLAM

YAG-SLAM aims to extract optical signals from intrinsic biochemical species of tissues, signals that are mediated through two and three photon interactions. The former type leads to 2PAF and SHG, while the latter to 3PAF and THG [69]. Although the origin of these nonlinear signals is different – e.g., multiphoton absorption involves electronic transitions and energy deposition into the specimen while harmonic generation generally does not – their intensity (I_ns) scales with the average ($\langle \rangle$) of the excitation power (${P_{ex}}$), obeying the number of photons (n) involved in the nonlinear interaction, i.e., ${I_{ns}} \propto {P_{ex}}^n$. [2,69–71].

We leveraged the power dependence of ${I_{ns}}$ to determine what nonlinear signal strikes each YAG-SLAM detector. To this end, we imaged a set of spectroscopic-grade substances that produce 2PAF, 3PAF, SHG, and THG, namely, FAD, NAD(P)H, β-Barium Borate (β-BBO), and glass-distilled water interfaces, respectively. These samples produce nonlinear signals that spectrally overlap with those of intrinsic multiphoton absorbers and harmonophores of tissues [72–78], light that enables assigning labels to channels 1-4 of the YAG-SLAM microscope. By pumping these samples at various excitation powers ${P_{ex}}$, we obtained multichannel images that allowed us to calculate the mean intensity of a given field of view per channel, i.e., ${\bar{I}_{ns}}({C{h_x}} )$. Insets in Fig. 4 show the raw data of these power-dependent experiments (black dots) and their respective fits (blue curves). Second-order polynomials fitted the signals that rely on two-photon interactions (2PAF and SHG), whereas third-order polynomials fitted those mediated by three-photon interactions (3PAF and THG). The goodness of these fits was determined by their R² values, which were all above 0.98.

 

Fig. 4. Spectroscopic validation of the YAG-SLAM microscope. Log-log plots of the power dependent emission of a) FAD (1.3 mM), b) NADH (1.3 mM), c) β-BBO, and d) a glass-distilled water interface. The insets show the raw data (black dots) and their respective polynomial fits (blue curves).

Download Full Size | PDF

By calculating the slope of ${\log _{10}}[{{{\bar{I}}_{ns}}({C{h_x}} )} ]$, we revealed the order of the multiphoton process. We found that the slopes of the SHG and THG experiments are exact, exhibiting a value of two and three, respectively – see Fig. 4(c-d). Conversely, we observed that the slopes of both the 2PAF and 3PAF curves slightly deviate from the anticipated results, Fig. 4(a-b). Although it challenges theoretical expectations, the drift on the slopes of the fluorescent channels was expected. As pointed out by other researchers [70–81]: other nonlinear optical processes compete with emissive decays, processes that effectively quench the fluorescence, thus deviating their expected power curves. The nonlinear processes that reduce fluorescent decays are stimulated emission, stimulated Raman scattering, and transient absorption. In stimulated emission, a pump photon forces the emissive species to return to the ground state; in stimulated Raman scattering, frequency detuning between a pair of photons – the pump and the Stokes – within the broad supercontinuum brings the emissive species to an excited vibrational state; while in transient absorption another pair of photons – the pump and the probe – brings the emissive species to an even higher electronic energy level through excited state absorption. Collectively, these processes prevent an emissive species in an electronically excited state from relaxing via fluorescence. Nevertheless, the results presented in Fig. 4 confirm that YAG-SLAM microscope channels 1-4 detect, respectively, 2PAF, 3PAF, SHG, and THG, thereby validating the spectroscopic specificity of our imaging platform.

3.2 YAG-SLAM tissue imaging

We applied YAG-SLAM for label-free multimodal imaging of freshly excised perirenal adipose tissue – the fatty portion of tissue that surrounds the kidney. This tissue is a crucial component in kidney homeostasis and exhibits a complex microenvironment characterized by a heterogeneous mixture of cells, sympathetic nerve endings, and vascular structures – all held together by a versatile extracellular matrix [82–85]. By probing a specimen of perirenal adipose tissue at different depths, YAG-SLAM revealed the composition and arrangement of this tissue substrate. As we approached the specimen from the coverslip, the edges of the most characteristic cell type of this visceral fat deposit, namely, the adipocyte, started to emerge, see white arrows in Fig. 5(a). The THG channel effectively uncovered tissue inhomogeneities, delineating the borders of the adipocytes, see magenta edges and interfaces around the fat cells in Fig. 5(b). At a depth of 10-30 µm, the sensitivity of the SHG channel toward non-centrosymmetric species, particularly the macromolecules elastin and collagen types I&III [6], enabled YAG-SLAM to reveal the extracellular matrix, especially near the adipocytes – note the green mesh in Fig. 5(a-c) which appears to be giving structural support to the cells. As we imaged deeper into the tissue – beyond 20 µm from the tissue surface – YAG-SLAM exposed the adipocyte bodies, see Fig. 5(c-d). These cells produced strong 3PAF, light that might have originated from three-photon excitation of biochemical species that emit light around the 450 nm spectral range, e.g., NAD(P)H, fatty acids, or other lipopigments [75]. The 2PAF channel identified tightly packed round-shaped structures with a diameter of circa 5 µm, a signal that might have originated from FAD of metabolic or red blood cells [75]. Finally, by probing the tissue beyond 30 µm deep, YAG-SLAM resolved the structures inside the perirenal adipose tissue, mostly the adipocytes, Fig. 5(c). This optical sectioning experiment not only reiterates the potential of SLAM microscopy for biomedical imaging but also validates the YAG supercontinuum as a suitable and effective source for multiphoton excited fluorescence and harmonic microscopy.

 

Fig. 5. YAG-SLAM tissue-sectioning of chinchilla perirenal adipose tissue at a) 5 µm, b) 10 µm, c) 20 µm, and d) 30 µm depths. Imaging settings: 900 × 900 pixels; Pixel dwell time: 22 µs; Scale bar: 100 µm. Power at the sample plane: 3.5 mW. Color code: Yellow: 2PAF, Cyan: 3PAF, Green: SHG, and Magenta: THG.

Download Full Size | PDF

4. Sample preparation

Solutions of NADH (10128023001, Roche Diagnostics) and FAD (F6625, Sigma Aldrich) were prepared by serial dilution of a stock solution to a final concentration of 1.3 mM, solutions that were sandwiched between a glass slide and a coverslip prior to imaging.

Kidney tissue was harvested from a healthy female chinchilla (Chinchilla lanigera) weighing 568 grams. The chinchilla was sacrificed as part of another research study by using CO₂ followed by cervical dislocation, and the tissue was harvested to maximally utilize this animal for research purposes. After the rodent was euthanized, a midline incision was made to access the abdominal cavity to perform a bilateral nephrectomy. Excised tissue was placed in a chilled saline solution prior to ex vivo imaging, which took place approximately 4 hours after removal. All procedures were performed under oversight and approval from the Institutional Animal Care and Use Committee (#21182) at the University of Illinois Urbana-Champaign.

5. Conclusion

In this study, we demonstrated the potential of the white light supercontinuum from a bulk crystal as the driving source for multimodal multiphoton microscopy, presenting data that shows that the light from a YAG crystal successfully excites native biomolecules of tissues to map the tissue microenvironment in a label-free and non-destructive manner. We observed that the YAG supercontinuum showed a robust performance, exhibiting a stable spectrum, intensity, and beam profile, optical properties that perdured for over 1,000 hours of use, carrying the same noise level as the pump beam and displaying an outstanding polarization-preservation nature. The white light supercontinuum from the YAG crystal not only covered the visible spectrum (550-900 nm) but also the near-infrared region (950-1200 nm), delivering an excitation window within 1030-1150 nm, a wavelength range of light that enables the simultaneous acquisition of intrinsic signals that inform about the composition, structure, and metabolic state of tissues, i.e., SHG, THG, 2PAF, and 3PAF. This excitation window required minimal quadratic phase compensation to achieve Fourier-transformed limited pulses at the sample plane, thus enabling high-quality images without requiring expensive pulse-shaping devices. Additionally, the near-infrared region of our YAG supercontinuum contains broad coherent light to attain frequency detuning for driving vibrational coherences of biomolecules, e.g., lipids, nucleic acids, or proteins. This frequency detuning could enable coherent Raman scattering microscopy by using the 980 nm band as a pump and the 1030-1190 nm spectral range as a Stokes to investigate the entire “fingerprint region” of molecular vibrations.

To conclude, this YAG-SLAM microscope validates the supercontinuum from a bulk crystal as a powerful source for multimodal multiphoton microscopy. This supercontinuum offers a robust, stable, alignment-free, and cost-effective broadband source with a simple and compact footprint. Bulk crystals such as YAG, as well as others, coupled with low repetition rate lasers, can potentially replace the photonic crystal fibers and rods that have been traditionally used in these systems, not only breaking the paradigm set by these technologies for label-free histology [86] but also opening new potential for the practical translation of multiphoton microscopy to clinical applications and clinical settings.