We present a three-color multiplex coherent anti-Stokes Raman scattering (CARS) setup that facilitates a prompt recording of broadband CARS spectra along with a fast CARS imaging. With separate narrowband Stokes and probe beams being introduced in the near IR, we are able to incorporate a stable, wideband Ti:sapphire femtosecond laser as a pump beam that covers the full range of Raman shift for CHn stretching vibrational modes. Experimentally, high-resolution multiplex CARS signals are allowed to investigate molecular vibrations over the range of 2650 cm-1–3050 cm-1, which are spectrally integrated to construct lipid-sensitive images. It is demonstrated that the proposed implementation promises a particular benefit on CARS imaging of lipid-rich tissue structures by providing detailed information on CHn Raman-active vibrations at points of interest on the CARS images that can be obtained at high frame rates.
© 2009 Optical Society of America
Coherent anti-Stokes Raman scattering (CARS) has received considerable attention in the field of biomedical imaging, for its high sensitivity to molecular vibrations offering inherent chemical contrast . Without exogenous labeling agents introduced to biological samples, laser-beam-scanning CARS microscopy has demonstrated a real-time performance and a submicrometer-scale spatial resolution in visualizing the three-dimensional (3D) structure of cells and tissues with vibrational selectivity . The current implementations of high-speed CARS microscopy  are mostly depending on a synchronized dual picosecond laser system that can resonate with a single Raman band allowing a high vibrational contrast. So far, one of the most successful applications of CARS microscopy has been the lipid-selective imaging in unlabeled living organisms, making good use of the characteristic abundance of carbon-hydrogen (CH) bonds in lipids . Many lipid studies, however, often require further detailed vibrational information to be obtained for high molecular specificity. To this purpose, CARS microspectroscopic techniques have been pursued to acquire a vibrational signature that spans over a number of Raman resonances. It has been cumbersome and time-consuming to record CARS spectra by changing the picosecond laser frequency to tune the Raman shift point by point [3, 4], which is deemed impractical especially for a dynamic biological system . Therefore incorporating high-speed 3D vibrational histology of a lipid-rich sample with prompt multiplex CARS spectral analysis at desired sites on a same imaging platform, would significantly benefit comprehensive studies on lipid storage mechanisms  as well as tissue diagnosis for various human metabolic diseases .
Multiplex CARS microspectroscopy using broadband Stokes pulses in combination with narrowband pump/probe pulses [7, 8], can be a straightforward way to overcome the shortcoming of the monochromatic CARS imaging. Recent progress in the supercontinuum (SC) generation from photonic crystal fibers (PCF)  has dramatically enlarged the multiplex CARS bandwidth to >2000 cm-1 , reducing the technical complexity and cost burden involved with multiplex CARS light sources as well. Compared to monochromatic CARS using picosecond pulses, however, the signal level of broadband multiplex CARS is significantly weaker because the supercontinuum pulse energy is spread over broad spectral window. As a result, the sensitivity of broadband multiplex CARS degrades for chemical imaging and longer image pixel dwell times (>10 ms) are required [11, 12], much slower compared to that (<10 µs) of current monochromatic CARS imaging techniques. While further improvements such as low-noise SC generation  and soliton self-frequency shift of broadband light pulses , have significantly increased the stability and tunability of the PCF-based light source, high-speed chemical imaging remains quite challenging due to their low Stokes pulse spectral density.
In this paper, we present a CARS spectroscopic imaging platform based on a 3-color multiplex CARS scheme that is designed to perform fast lipid-sensitive imaging and concomitant broadband analysis in the entire CHn stretching vibrational region. It is the fingerprint region (1000 cm-1–1800 cm-1) that contains a wealth of vibrational information for detailed spectral analysis. However, the high-wavenumber Raman region (2600 cm-1–3100 cm-1) presents its own merit within the lipid research; Raman signals in this region do provide complementary but unique chemical contents such as lipid concentration , molecular ordering in lipid packing , degree of carbon chain unsaturation , etc. More importantly, strong CHn stretching vibrational bands that lipids possess in the high-wavenumber region  would be ideal for realizing high-speed chemical imaging along with multiplex spectral analysis. In order to optimize the broadband CARS excitation for both the spectral bandwidth and the chemical imaging sensitivity, we use an adjustable-bandwidth Ti:sapphire femtosecond laser at 800 nm wavelength to serve as a wideband pump light which provides sufficient power and superior stability. The 3-color CARS scheme is employed to configure all CARS excitation laser sources in the near-IR region. Owing to the separate use of a narrow-linewidth probe light with a shorter wavelength away from the pump laser bandwidth, multiplex CARS spectra covering the Raman shift from 2600 cm-1 to 3100 cm-1 can be obtained with high spectral resolution. The approach of integrated detection of the entire multiplex CARS signal is found to be effective for a rapid construction of vibrational contrast images, without any need for a significant change in the configuration of CARS excitation light sources. Finally, we demonstrate the use of our 3-color multiplex CARS setup in 3D en-face chemical imaging of mouse cardiovascular tissues with atherosclerotic lesion, illustrating its particular benefit for a high-frame-rate imaging of lipid-rich tissue structures with a concomitant multiplex analysis of CHn stretching vibrations for the chemical details.
2. Three-color broadband multiplex CARS spectroscopy
As a four-wave-mixing process, CARS involves three incident laser pulses at the pump (ωp), Stokes (ωs), and probe (ω′p) frequencies, interacting simultaneously at a tight focus within the sample. The pump (ωp) and Stokes (ωs) pulses put molecules into coherent oscillation at the beating frequency (ωp-ωs), and the probe (ω′p) pulse is scattered off of this molecular vibration to generate an anti-Stokes signal at frequency ωas=ωp-ωs+ω′p. We refer to this process as “3-color CARS” [17, 18] throughout this paper, in contrast with “2-color CARS” process that has been widely adopted in many practical CARS experiments where the probe pulse is derived from the same pulse as the pump (ω′p=ωp) to yield an anti-Stokes signal at frequency ωas=2ωp-ωs.
A CARS spectrum can be measured by tuning the difference frequency (ωp-ωs) of narrow-linewidth laser pulses, which could be carried out by changing either the pump frequency (ωp) or the Stokes one (ωs). During the scan, strong anti-Stokes signals are observed when the difference frequency (ωp-ωs) matches the frequency of a particular Raman resonance (Ω). A more direct and convenient means to record a CARS spectrum is the multiplex CARS approach that uses a broadband light pulse to simultaneously excite multiple vibrational modes within the frequency band of interest (ΔΩ). Based on a 2-color CARS scheme as shown in Fig. 1(a), multiplex CARS usually requires a broadband light pulse (Δωs) for stimulating the Stokes process, in combination with another narrow-linewidth pulse with a higher frequency (ωp) acting as both the pump and the probe. Resulting anti-Stokes components that are multiplexed at frequency ωas=Ω+ωp, allow CARS spectra with a frequency resolution determined essentially by the linewidth of the pump pulse.
In contrast to the 2-color CARS spectrum produced with a broadband Stokes pulse (Fig. 1(a)), a broadband pump pulse inevitably gives rise to a blur and distortion of the 2-color CARS spectrum (Fig. 1(b)); this is because the pump pulse simultaneously acts as a probe pulse to convolute its spectral profile (of a bandwidth Δωp) with the anti-Stokes Raman features occurring at ωas=Ω+ωp. For a broadband light source to provide a multiplex pump pulse, it is straightforward to use a 3-color CARS scheme as shown in Fig. 1(c) where a separate a narrow-linewidth probe pulse (ω′p) is added to produce a spectrally resolvable multiplex CARS spectrum. Providing that the probe frequency (ω′p) is set well above that of the pump pulse (ωp) by ~3Δωp/2, one can record a high-resolution 3-color multiplex CARS spectrum at ωas=Ω+ω′p, which is isolated away from the unwanted 2-color contribution in the spectral domain. Since the 3-color laser pulses allow for several pulse combinations to generate 2- and 3-color CARS signals, it is necessary to configure a spectroscopic setup with the optimal probe frequency in order to collect the desired multiplex CARS signals.
2.2. Multiplex CARS spectroscopy and imaging of lipid-rich samples
Despite its apparent complexity involved in the experimental setup, the 3-color multiplex CARS approach provides a straightforward solution to incorporate a broadband laser source as a pump, which can be particularly suitable for high-sensitivity imaging and concomitant chemical analysis of lipid-rich biological samples that exhibit strong and distinct vibrational signatures within the high-wavenumber region ranging from 2700 cm-1 to 3100 cm-1.
Previously, fast lipid mapping has been carried out by detecting strong CARS contrast arising from the symmetric CH2 vibration with Raman shift around 2845 cm-1. The vibrational spectra of lipids have been obtained by either tuning the Raman shift point by point [3, 4] or performing multiplex CARS measurements with a conventional 100-fs Ti:sapphire laser oscillator as a Stokes source [7, 8]. However, the former method is inappropriate for dynamics studies and the latter one has an insufficient bandwidth ~150 cm-1 to cover the entire highwavenumber region. The 3-color multiplex CARS scheme allows us to eliminate this inconvenience by employing a broadband CARS excitation source with optimum spectral bandwidth for a vibrational analysis with full coverage of CHn stretching vibration bands and the efficient integration of entire multiplex CARS signals to form high lipid contrast. A 900-mW wideband femtosecond mode-locked Ti:sapphire laser (Micra 10, Coherent Inc.) emitting a pump beam at 817 nm is adjusted to have bandwidth of 30 nm (~450 cm-1), which is combined with a 7-ps narrowband (~3.5 cm-1) Stokes laser beam at 1064 nm to simultaneously excite molecular vibrations within the frequency band of interest (from 2650 cm-1 to 3050 cm-1). A 6-ps narrowband (~5 cm-1) laser beam at 776 nm, well separated from the pump bandwidth, is additionally combined to produce 3-color multiplex CARS spectra with spectral resolution.
Of note, the proposed laser configuration streamlines the procedure of switching between the multiplex CARS measurement and the fast lipid-selective imaging. An integrated detection of the multiplex vibrational signal at a photomultiplier tube having a passband pertinent to the multitude of Raman resonances of lipid-related CHn vibrations, readily allows for an efficient lipid-selective CARS mapping without changing the excitation light source used for the multiplex CARS spectroscopy. This obviates the need for a careful beam alignment to ensure precise matching of the focal volume positions when switching between spectroscopic and microscopic measurements.
In the CARS signal integration over the range of 2650 cm-1–3050 cm-1, we detect not only the signal from the symmetric CH2 vibration band at 2845 cm-1 but all CHn stretching vibration bands in the multiplex CARS window. Based on the fact that lipids are characteristically abundant in CH2 bonds compared to other biomolecules, this specific Raman band has been almost exclusively employed for gaining the lipid contrast in many previous studies [2, 3, 4]. Various kinds of lipids, however, do have many other strong Raman bands over the high-wavenumber region, which might be a great benefit for fast vibrational imaging of lipids in a particular class of samples. Since a majority of cellular components and tissue matrix constituents have high-wavenumber Raman bands that are either relatively weak or negligible compared with lipids, the broadband-integrated CARS detection can be effective in lipid-rich structures, especially when the broadband CARS excitation is wide just enough to concentrate on CHn stretching vibrations. Previously, there has been a similar attempt to demonstrate lipid-selective CARS imaging with a Raman excitation bandwidth of ~180 cm-1, which has resulted in a high signal-to-background ratio of 20:1 . In this study, we enlarged such spectral integration bandwidth to ~400 cm-1 for high-throughput imaging of lipid-rich tissue samples.
Compared with multiplex CARS using PCF-based light sources, we can take advantage of the optimized bandwidth of a powerful broadband laser to allow high sensitivity for imaging lipids as well as sufficient multiplex bandwidth. The main reason for taking the 3-color CARS scheme is to utilize this broadband laser as a CARS pump to configure all excitation laser in the near-IR. In case of using the 2-color multiplex CARS scheme with the 817 nm broad-band source as a Stokes laser, the pump laser should be prepared at around 660 nm for Raman resonance at 2850 cm-1, resulting in anti-Stokes signals appearing at 555 nm. Such short wavelengths should be avoided in CARS imaging of biological tissue samples because penetration depth of the CARS excitation beam decreases and the scattering loss of CARS signals increases . However, there is no fundamental difference in performance, compared to a system with a broadband Stokes and a narrowband pump. Another comparison can be made with a multiplex CARS with a stable ultrashort femtosecond oscillator with large bandwidth of up to 100 nm. Such a light source, however, has been adopted for single-pulse CARS approaches where all CARS excitation pulses are derived from the same laser pulse through a spectral-domain shaping , resulting in a limited coverage of vibrational frequencies only below 1500 cm-1. By tailoring the spectral phase of broadband pump and/or probe pulses, CARS signals could be obtained in high resolution for high Raman shift [21, 22]. Compared to spontaneous Raman microspectroscopy, strong multiplex CARS signals can be further enhanced by mixing with the nonresonant background while spontaneous Raman signals are inherently multiplex and background free. Combining monochromatic CARS (for fast selective imaging) with spontaneous Raman spectroscopy (for broadband spectroscopy) could be a simple solution but may require a careful spatial calibration for the position and size of the two different probing volumes. In this study, a 3-color CARS measurement scheme is implemented to enable broadband multiplex CARS microspectral analysis covering the Raman shift from 2650 cm-1 to 3050 cm-1, readily compatible with lipid-sensitive 3D mapping based on the integrated broadband CARS signals through a spectral window for Raman shifts around 2850 cm-1.
3. Experimental setup
The 3-color CARS imaging platform consisted of a near-infrared pulsed laser system generating synchronized CARS excitation beams in 3 colors and a modified commercial laser-scanning confocal microscope (IX81/FV300, Olympus) combined with a grating spectrometer for multiplex CARS microspectral analysis and a forward CARS bandpass detector for fast lipid imaging.
A 1064-nm mode-locked Nd:YVO4 laser (PicoTrain, High-Q Laser) delivering 10-W average power of 7-ps pulse train in the repetition rate of 76 MHz was used to generate the CARS Stokes beam by splitting off 10% power of its output and guiding into the microscope through a pulse delay line. The main portion (9 W) of its output was utilized for synchronously pumping an intracavity-doubled optical parametric oscillator (OPO; Levante, APE) to generate the 1.3-W CARS probe beam in a 6-ps, 76-MHz pulse train at 776-nm wavelength. The multiplexed CARS pump beam centered at the wavelength of 817 nm was produced from a broadband femtosecond mode-locked Ti:sapphire laser (Micra 10, Coherent) providing 900-mW average power and its spectral bandwidth adjusted to about 30 nm, whose output pulse train was actively synchronized with that of the 1064-nm CARS Stokes beam to maintain the same pulse timing with repetition rate at 76 MHz by the use of a cavity stabilization feedback servo (SynchroLock-AP, Coherent). The synchronization system was found to have a pulse timing jitter of less than 250 fs in rms for the bandwidth of 0.02 Hz–1 kHz and a good long-term stability to perform 3-color CARS experiments without significant CARS signal fluctuations (<5%) for more than 10 hours.
The beam diameter and divergence of each laser beam were adjusted by a telescope beam expander placed in each beam path to match one another. The three CARS excitation beams were collinearly overlapped in space by using two beam-combining optics in series; the pump and probe beams were combined at a 50:50 broadband beam splitter and the Stokes beam was then added to them by a dichroic mirror (from Chroma Technologies, USA) with high reflectivity (>99%) for near-IR wavelength in the range of 730 nm–960 nm and high transmittance (>90%) for the Stokes beam at 1064 nm. The combined laser beams are delivered to a 1.2 NA 60× water-immersion microscope objective (UPlanSApo/UIS2, Olympus) through the 2-axis beam scanning unit (FV300, Olympus) consisting of a pair of galvanometer mounted gold mirrors with reflectivity of about 95% for wavelengths longer than 600 nm.
Tight focusing of the attenuated 3-color excitation beams on a biological sample  were expected to produce 3-color multiplex anti-Stokes signals in the wavelength range of 623 nm–641 nm, separated from those of the 2-color contribution residing in region of 644 nm–683 nm. Guided by a free-space relay optics, the forward-collected multiplex CARS signals were sent into a grating monochromator (Triax320, HORIBA Jobin-Yvon) equipped with a Peltier-cooled EMCCD camera (DU970N-BV, Andor) with a 1600×200 image sensor with pixel size of 16×16 µm2. The spectrometer was configured to use a 600-groove/mm diffraction grating and an entrance slit of 150-µm width, allowing the optical spectra to be recorded with a wavelength span of 130 nm and instrumental resolution of 0.1 nm (2.5 cm-1 in wavenumbers, equivalently). Depending on the intensity of CARS signals arising from the focal volume, spectral data recording was carried out with the camera exposure time of 20 ms–150 ms, which was externally synchronized to the gating of CARS excitation laser beams that was achieved by using an electro-mechanical beam shutter (LS6, Uniblitz) in order to avoid excessive local heating of the sample. Quantitative multiplex CARS spectra of a sample under investigation could be obtained by the normalization procedure using the nonresonant CARS spectrum from a cover glass, which took into account both the nonuniform spectral profile of the broadband pump laser source and the overall spectral transmission characteristics of its optical path through the sample to the detector.
In the same broadband multiplex CARS setup, lipid-sensitive CARS images could be obtained by raster scanning the focus of the collinear pump and Stokes laser beams over the specimen and collecting the entire multiplex CARS signals in the range of 2650 cm-1–3050 cm-1. To achieve high throughput, we utilize not only the symmetric CH2 vibration mode at 2845 cm-1 but all the multitude of lipid-associated vibrations within the multiplex CARS spectral window, including aliphatic CH2, CH3, vinyl CH stretches, and so on. Various kinds of lipids have many strong Raman bands over the high-wavenumber region while a majority of cellular components and tissue matrix constituents have high-wavenumber Raman bands that are either relatively weak or negligible compared with lipids. The broadband-integrated CARS detection can be effective in lipid-rich structures, especially when the CARS excitation is wide just enough to concentrate on CH stretching vibrations. There has been a similar previous approach to perform lipid-selective imaging with a detection bandwidth of ~180 cm-1, which has demonstrated signal-to-background ratio of 20:1 and image pixel dwell time of ~80 µs . Here, we enlarged such spectral integration bandwidth to ~400 cm-1 in order to fully cover the CHn stretching vibrational region.
Unlike the high-resolution multiplex CARS spectral measurement, the lipid-contrast imaging was attempted without the separate probe beam from the OPO source, relying on the spectrally integrated 2-color broadband CARS signals. With the 776-nm OPO output blocked by a mechanical shutter, the resulting lipid-window CARS signal was collected in the forward direction through a 0.55NA condenser with 27-mm working distance (IX2-LWUCDA2, Olympus) and reflected off a long-pass dichroic (750dcxr, Chroma Technology) to dump out most of unwanted CARS excitation beams. The isolated CARS signals in the wavelength range of 645 nm–685 nm, after further purification through a stack of three 40-nm-width bandpass filters centered at 660 nm (42-7120, Coherent-Ealing), were detected by a red-sensitive photomultiplier tube (R3896, Hamamatsu).
The CARS microscopy setup could acquire depth-sectioned images in 512×512 pixels with the maximum field of view of 250×250 µm2 and a motorized focusing stage built in the inverted microscope was actuated in micrometer steps along the axial direction to collect z-stacks of 2D image slices. CARS microscopic inspection and image acquisition for areas of interest were facilitated by using a 2-axis motorized translation stage (H117, Prior Scientific). After points of interest in the sample were selected for CARS spectral analysis, the CARS imaging setup could be readily reconfigured into a 3-color broadband multiplex CARS microspectrometer.
4. Result and discussion
4.1. Characteristics of a 3-color broadband multiplex CARS spectral measurement
To test the 3-color multiplex CARS microspectroscopy setup, we measured CARS spectra of several chemical compounds in neat solution over the Raman shift ranging from 2600 cm-1 to 3100 cm-1. Balsam oil, acetone, ethanol, isopropyl alcohol, and methanol were dropped on microscope cover glasses and used as test samples, which exhibited their own characteristic multiplex CARS spectra as shown in Fig. 3. All multiplex CARS spectra were normalized by the nonresonant CARS spectrum from the cover glass, in order to correct the effect of the nonuniform spectral nature of the broadband pump excitation as well as the overall transmission and detection of anti-Stokes signals.
The power levels of the pump, Stokes, and probe beams incident on the samples were 14 mW, 7 mW, and 12 mW, respectively, and the camera exposure time for spectral recording was fixed to 90 ms, except for the glass which required a longer integration time as large as 1.0 s for a sufficient signal-to-noise ratio to be obtained. The time delays between the pump, Stokes, and probe pulses were adjusted to maximize the CARS spectral intensity and no additional time delay was applied to the probe pulse for reducing the nonresonant CARS contributions. The multiplex CARS spectrometer was calibrated for Raman vibrational frequency by using the CARS peaks observed in the spectra to match the previously reported data, which included the symmetric CH2 stretch at 2850 cm-1 abundant in oily substances ,=C-H stretch at 3015 cm-1 in unsaturated fatty acids , carbon-hydrogen vibrations of acetone, ethanol, and methanol at 2923 cm-1, 2877 cm-1, and 2846 cm-1, respectively . Considering the spectrometer’s instrumental resolution (2.5 cm-1) and the 776-nm probe beam linewidth (~5 cm-1), the present CARS microspectroscopy setup were expected to yield multiplex CARS spectra broadened by about 6 cm-1.
To confirm that the measured multiplex CARS spectra originated from the 3-color contribution of multiplex CARS, we investigated the dependence of multiplex CARS spectra on the the power of laser beams used for CARS excitation. As displayed in Fig. 4(a), changing the power of CARS pump beam did not alter the characteristic spectral profile of a balsam oil sample, but only led to a variation in the spectral intensity of the multiplex CARS spectra. The linear dependence of the integrated spectral intensities on the individual excitation laser power was demonstrated as Fig. 4(b), directly indicating that the CARS spectra was produced via the 3-color CARS process as desired.
4.2. Lipid-sensitive microanatomic imaging and chemical profiling of atherosclerotic lesions
We applied our 3-color broadband multiplex CARS microscope platform to lipid-selective microanatomic imaging and concomitant chemical profiling of intact atherosclerotic lesions. As the most prevalent form of cardiovascular diseases being on the global rise in incidence, atherosclerosis has been attracting a great deal of attention for its pathology and diagnosis [6, 24, 25]. In the pathogenesis of atherosclerosis, lipids have been known to play a crucial role in the progression of the affected lesion into a certain type of plaques that are vulnerable to rupture, precipitating thrombosis to cause acute myocardial infarction (AMI) [26, 27]. Autopsy studies have characterized ruptured atherosclerotic plaques likely to contain a thin fibrous cap (consisting of collagen, macrophages, lymphocytes, etc.), proteoglycans, and calcification as well as enriched lipid contents (including extracellular lipids such as free and modified cholesterol, phospholipid, triacylglycerol, fatty acid, etc.) in either a soft gruel-like phase or various crystalline forms .
Focusing on demonstrating the capability of our proposed setup to characterize the atherosclerotic lipids, we carried out a label-free 3D imaging and prompt spectral analysis of atheroma tissues that were taken from apolipoprotein E knock-out (ApoE-/-) mice fed a 0.15% high-fat high-cholesterol diet for 16 weeks. After harvesting the heart and aorta, the connective tissue of the aorta was carefully removed and the aortic segments with lesser arch curvature and carotid arteries were dissected and perfused in phosphate buffered saline (PBS) solution. Without any chemical mounting solution or fixatives, the dissected samples were then longitudinally opened and mounted lumen side down on a microscope coverslip, for en face CARS imaging of the arterial wall from the luminal view.
Depth-sectioned en face CARS image slices were acquired for an atherosclerotic plaque as in Fig. 5, exquisitely delineating lipid-rich microanatomical components embedded in the arterial lesion from the lumen surface to the deep intima. Individual foam cells engulfing intracellular lipid droplets were clearly imaged in the superficial layers (5–10 µm deep from the lumen, Fig. 5 (a–c)), whereas extracellular lipid debris along with cholesterol crystals were observed in the deep intima region (>20 µm in depth, Fig. 5 (e–i)), as validated further by the comparative study done for the same section with histological Oil-Red-O staining for lipids and whole mount immunohistochemistry using the CD68 marker specific to macrophages.
Higher concentration of lipid contents appeared as brighter pixels on the pseudo-color CARS image, in which the lipid contrast was attained by detecting the integrated intensity of the multiplex CARS arising from entire CH vibrational modes in the Raman band of 2650 cm-1–3050 cm-1. With the probe beam blocked in our 3-color CARS setup, this 2-color broadband CARS contribution, while its spectral details were smeared out due to the convolution of the broad pump spectrum, still yielded a reasonable signal-to-background ratio as large as 15:1 for lipids, comparable to those of previous approaches to imaging lipid-rich tissue structures . The CARS microscope setup could be operated to acquire en-face images at the rate of 2.5 s/frame with the field-of-view (FOV) up to 250×250 µm2. The spatial resolution was characterized to be ~0.4 µm in the lateral (x–y) plane and ~1.5 µm along the axial (z) direction when the image pixel dimension does not exceed the CARS excitation volume. The 3D imaging of mouse atherosclerotic plaques was restricted in depth around 50–70 µm depending on the type of lesions. Average power of the combined pump and Stokes laser beams illuminating the sample was limited to below 20 mW in total by attenuating the incident laser beam with neutral density filters, in order to avoid the risk of laser-induced damage to the tissue sample .
The demonstrated capability of the CARS imaging platform in fast 3D mapping of intact lipid-rich tissue structures would be useful not only for the quantification of volumetric coverage of lipid contents in the lesion but also for the morphometric assessment of atherosclerotic lipids that could help distinguish different types of plaques  as well as facilitate the microspectral analysis of the plaque components. From CARS imaging inspection over a good amount of atherosclerotic lesions in the affected artery samples, various microanatomical features of atherosclerotic lipids could be observed and classified into, but not limited to, four representative groups according to their appearance: lipid-laden foam cells, extracellular lipid deposits, stratified laminae lipid crystals, and needle-shaped lipid crystals, as illustrated in Fig. 6. Each image was taken at a different site and depth of independent lesions in the sample, where a representative feature of the lipid accumulation morphology was revealed.
A lipid-laden foam cell was viewed in the CARS imaging as lipid droplets closely packed within a cellular envelope having a dark circular void at the center, presumably a nucleus (Fig. 6(a)). The foam cells are known as one of the markers characterizing the stage of atherogenesis, which can be usually found at the subendothelial layer or the intima in the proximity of lumen even from the early stage. Depending on the type and stage to which atherosclerotic lesions belonged, foam cells were observed to differ in the density and extent of accumulation as well as their cellular shape (streak, elongated, rounded). In the intermediate lesions or the advanced ones, a large mass of extracellular lipid deposits could be found in the intima (>10 µm deep from the lumen) as shown in Fig. 6(b). The lipid accumulation in the extracellular space had several indeterminate forms such as interstitial lipid droplets, scattered lipid granules, and gruel-like lipid lumps. The cholesterol deposit, a major component contained in the dense lipid cores of vulnerable plaques, also showed up to have crystalline forms with the morphology of stratified laminae or needles as clearly imaged in Fig. 6(c) and (d), respectively. Since the formation of cholesterol crystals has been recognized as a relevant factor for plaque rupture [28, 29], the capability of CARS imaging to visualize cholesterol crystals in intact arterial tissue ex vivo would be useful in the investigation of the plaque vulnerability and its clinical assessment. Previously, cholesterol crystals have been studied in vitro  or examined under a polarized light microscope after dissection from the atheroma .
More in-depth chemical information was obtained of the lipid-rich structures observed within the intact atherosclerotic lesions, by switching the CARS setup from a lipid-sensitive micro-scope to a spectrometer with full coverage of high-wavenumber CH stretching vibrations. Based on our 3-color broadband multiplex CARS scheme, the procedure could be promptly done just by forwarding the probe beam reserved for high-resolution CARS spectroscopy into the same microscope and diverting the multiplexed anti-Stokes signal from the sample to a spectral detection setup. In the point-scan mode operation of the microscope system, a coinciding focus of the three laser beams was made to target a point of interest on the sample, yielding the multiplex CARS spectra as recorded on the right panel of Fig. 6 which reveals distinctive chemical profiles according to the type of atherosclerotic lipids as presented. The multiplex CARS spectra were measured for each representative lipid morphology at the site indicated by arrows in the left panel of Fig. 6, using the CARS excitation laser power of less than 40 mW in total with an exposure time of 50 ms. Being irrespective of the location where a lipid accumulation took place, the CARS spectra were found to be highly characteristic only to the lipid morphology and reproducible in the measurement as well.
From the CARS spectra measured, intracellular lipid droplets bounded in foam cells, were found to exhibit a prominent CARS band at 2850 cm-1 responsible for the symmetric CH2 stretching vibration with several adjacent vibrational bands in close contact diminishing in their intensity toward higher Raman shifts (in Fig. 6(a’)). Little difference was observed for extra-cellular lipid droplets and lumps in the CARS spectral profile (in Fig. 6(b’)) when compared with that of intracellular lipid droplets. In contrast, the lipids in the form of stratified laminae crystals manifested a remarkable difference in the CARS bands appearing at 2871 cm-1, 2910 cm-1, and 2950 cm-1 (in Fig. 6(c’)) which were tentatively assigned as CH2 asymmetric, CH3 symmetric, and CH3 asymmetric vibrations, respectively. The CARS spectra of needle-shaped crystalline lipids were also distinct from those of lipid droplets, but had vague peaks at 2910 cm-1 and 2950 cm-1 as compared to stratified laminae lipid crystals (in Fig. 6(d’)). A satellite CARS peak was observed in both types of lipid crystals around 2978 cm-1, possibly arising from to CH3 vibrations of alkane residues pertinent to lipid crystallization. Such differences in the CARS profiles for CH vibrational bands are believed to suggest post-translational modifications of lipids undergone in the progression of atherosclerosis as well as their changes in the phase such as liquid, gel, and crystal . In keeping with the recognition of Raman spectroscopy that the greater ratio of the asymmetric to symmetric CH2 vibrations indicates the higher ordering of polymethylene chains [4, 8], the lipids with crystalline morphology of laminae and needles in this study showed greater ratios between the spectral peak at 2871 cm-1 and 2850 cm-1, as compared with lipid droplets in liquid phase with less-ordered methyl chains. Furthermore, the significant increase of CH3 vibrational bands observed from laminae-shaped lipids (in Fig. 6(c’)), is likely attributed to high protein content coming into play in the crystalline formation where CH2 groups are less dominant in amino acid side chains than in lipids .
Previously, spectral CARS microscopy measurements on the lipid components in tissue specimen have been done through laborious recording of CARS signals with a step-by-step tuning of the Raman shift . In another work based on multimodal nonlinear optical microscopy, atherosclerotic tissue lesions have been examined by lipid-selective CARS imaging in which chemical modification of lipids, however, has been traced by the two-photon excitation fluorescence method . In the aspect of convenience and measurement throughput, our broadband multiplex CARS microspectroscopy setup is thought to have demonstrated a practical benefit in the label-free biomedical imaging combined with spectral analysis capability.
We have developed a broadband multiplex CARS microspectroscopic imaging platform capable of recording the high-wavenumber CH vibrational spectra in full coverage, with a concomitant lipid-sensitive tissue imaging available on the same setup. Unlike the conventional approaches of multiplex CARS measurements, we have employed a broadband near-IR femtosecond laser source to serve as a CARS pump beam, based on a 3-color CARS scheme with all CARS excitation laser sources configured in the near-IR region favorable to biological tissue analysis in terms of penetration depth and photodamage. By the use of a narrowband probe beam in the 3-color CARS scheme, high-resolution (~7 cm-1) multiplex CARS spectra isolated from the 2-color CARS contribution, have been produced in the Raman shift range of 2650 cm-1–3050 cm-1. Interestingly, the spectral integration of multiplex signals from the 2-color CARS contribution has been shown to provide a high sensitivity to lipids suitable for a fast 3D imaging of tissue specimen, being affordable with the superior power and stability of the broadband CARS pump source used in this study. It has been demonstrated that switching the presented CARS setup between a lipid-sensitive microscope and a high-wavenumber CH vibrational spectrometer, can be done quick and straightforward without any inconvenience such as laser wavelength tuning, optical beam realignment, focus overlap adjustments, etc.
We have demonstrated the use of our broadband multiplex CARS measurement setup in lipid-selective 3D microanatomic imaging of intact mouse cardiovascular tissues and multiplex CARS chemical analysis of the atherosclerotic lipids of pathological concern. The feasibility of high-throughput CARS microspectral imaging to acquire chemical details in direct correlation with label-free morphological features has been successfully established, thereby illustrating its particular benefit for better understanding of lipid-associated diseases on both cellular and tissue levels. Using the presented CARS microspectral imaging, more comprehensive investigations on atherosclerosis are under progress and will appear elsewhere in the literature.
This study was supported by the Next-Generation New-Technology Development Program of MKE Korea, and in part by the Bio-signal Analysis Technology Innovation Program of the MEST Korea. The authors gratefully acknowledge Eun-Soo Lee for the preparation of arterial tissue samples from atherosclerosis mouse models.
References and links
1. A. Zumbusch, G. R. Holtom, and X. S. Xie, “Three-dimensional vibration imaging by coherent anti-Stokes Raman scattering,” Phys. Rev. Lett. 82, 4142–4145 ( 1999). [CrossRef]
2. J.-X. Cheng, Y. K. Jia, G. F. Zheng, and X. S. Xie, “Laser-scanning Coherent anti-Stokes Raman scattering microscopy and applications to cell biology,” Biophys. J. 83, 502–509 ( 2002). [CrossRef] [PubMed]
3. C. L. Evans, E. O. Potma, M. Puoris’haag, D. Cote, C. P. Lin, and X. S. Xie, “Chemical imaging of tissue in vivo with video-rate coherent anti-Stokes Raman scattering (CARS) microscopy,” Proc. Natl. Acad. Sci. USA 102, 16807–16812 ( 2005). [CrossRef] [PubMed]
4. T. Hellerer, C. Axäng, C. Brackmann, P. Hillertz, M. Pilon, and A. Enejder, “Monitoring of lipid storage in Caenorhabditis elegans using coherent anti-Stokes Raman scattering (CARS) microscopy,” Proc. Natl. Acad. Sci. USA104, 14658–14663 ( 2007). [CrossRef] [PubMed]
5. H. A. Rinia, M. Bonn, E. M. Vartiainen, C. B. Schaffer, and M. Müller, “Spectroscopic analysis of the oxygenation state of hemoglobin using coherent anti-Stokes Raman scattering,” J. Biomed. Opt. 11, 050502 ( 2006). [CrossRef] [PubMed]
6. A. H. Chau, J. T. Motz, J. A. Gardecki, S. Waxman, B. E. Bouma, and G. J. Terney “Fingerprint and high-wavenumber Raman spectroscopy in a human-swine coronary xenograft in vivo,” J. Biomed. Opt. 13, 040501 ( 2008). [CrossRef] [PubMed]
7. M. Müller and J. M. Schins, “Imaging the thermodynamic state of lipid membranes with multiplex CARS microscopy,” J. Phys. Chem. B 106, 3715–3723 ( 2002). [CrossRef]
8. J.-X. Cheng, A. Volkmer, L. D. Book, and X. S. Xie, “Multiplex coherent anti-Stokes Raman scattering microspectroscopy and study of lipid vesicles,” J. Phys. Chem. B 106, 8493–8498 ( 2002). [CrossRef]
9. J. M. Dudley, G. Genty, and S. Coen, “Supercontinuum generation in photonic crystal fiber ,” Rev. Mod. Phys. 78, 1135–1184 ( 2006).
10. M. Okuno, H. Kano, P. Leproux, V. Couderc, and H. Hamaguchi, “Ultrabroadband multiplex CARS microspectroscopy and imaging using a subnanosecond supercontinuum light source in the deep near infrared,” Opt. Lett. 33, 923–925 ( 2008). [CrossRef] [PubMed]
12. H. Kano and H. Hamaguchi, “Vibrationally resonant imaging of a single living cell by supercontinuum-based multiplex coherent anti-Stokes Raman scattering microspectroscopy,” Opt. Express 13, 1322–1327 ( 2005). [CrossRef] [PubMed]
13. S. Murugkar, C. Brideau, A. Ridsdale, M. Naji, P. K. Stys, and H. Anis, “Coherent anti-Stokes Raman scattering microscopy using photonic crystal fiber with two closely lying zero dispersion wavelengths,” Opt. Express 15, 4848–4856 ( 2007). [CrossRef]
14. E. R. Andresen, V. Birkedal, J. Thøgersen, and S. R. Keiding, “Tunable light source for coherent anti-Stokes Raman scattering microspectroscopy based on the soliton self-frequency shift,” Opt. Lett. 31, 1328–1330 ( 2006). [CrossRef] [PubMed]
15. C. Heinrich, A. Hofer, A. Ritsch, C. Ciardi, S. Bernet, and M. Ritsch-Marte, “Selective imaging of saturated and unsaturated lipids by wide-field CARS-microscopy,” Opt. Express 16, 2699–2708 ( 2008). [CrossRef] [PubMed]
16. G. Socrates, Infrared and Raman characteristic group frequencies, 3rd/ed., (John Wiley & Sons, New York, 2001), Chap. 23.
17. D. Pestov, R. K. Murawski, G. O. Ariunbold, X. Wang, M. Zhi, A. V. Sokolov, V. A. Sautenkov, Y. V. Rostovtsev, A. Dogariu, Y. Huang, and M. O. Scully, “Optimizing the laser-pulse configuration for coherent Raman spectroscopy,” Science 316, 265–268 ( 2007). [CrossRef] [PubMed]
19. F. Ganikhanov, S. Carrasco, X. S. Xie, M. Katz, W. Seitz, and D. Kopf, “Broadly tunable dual-wavelength light source for coherent anti-Stokes Raman scattering microscopy,” Opt. Lett. 31, 1292–1294 ( 2006). [CrossRef] [PubMed]
22. S. Postma, A. C. W. van Rhijn, J. P. Korterik, P. Gross, J. L. Herek, and H. L. Offerhaus, “Application of spectral phase shaping to high resolution CARS spectroscopy,” Opt. Express 16, 7985–7996 ( 2008). [CrossRef] [PubMed]
23. Y. Fu, H. Wang, R. Shi, and J.-X. Cheng, “Characterization of photodamage in coherent anti-Stokes Raman scattering microscopy,” Opt. Exp. 14, 3942–3951 ( 2006). [CrossRef]
24. T. T. Le, I. M. Langohr, M. J. Locker, M. Sturek, and J.-X. Cheng, “Label-free molecular imaging of atherosclerotic lesions using multimodal nonlinear optical microscopy,” J. Biomed. Opt. 12, 054007 ( 2007). [CrossRef] [PubMed]
25. H.-W. Wang, I. M. Langohr, M. Sturek, and J.-X. Cheng, “Imaging and quantitative analysis of atherosclerotic lesions by CARS-based multimodal nonlinear optical microscopy,” Arterioscler. Thromb. Vasc. Biol. 29, 1342–1348 ( 2009). [CrossRef] [PubMed]
26. D. M. Small, “George Lyman Duff memorial lecture - Progression and regression of atherosclerotic lesions: Insights from lipid physical biochemistry,” Arterioscler. Thromb. Vasc. Biol. 8, 103–129 ( 1988). [CrossRef]
28. R. Virmani, a. P. Burke, A. Farb, and F. D. Kolodgie, “Pathology of the vulnerable plaque,” J. Am. Coll. Cardiol. 47(C), C13–18 ( 2006). [CrossRef]
29. G. S. Abela and K. Aziz, “Cholesterol crystals cause mechanical damage to biological membranes: a proposed mechanism of plaque rupture and erosion leading to arterial thrombosis,” Clin. Cardiol. 28, 413–420 ( 2005). [CrossRef] [PubMed]
30. R. K. Tangirala, W. G. Jerome, N. L. Jones, D. M. Small, W. J. Johnson, J. M. Glick, F. H. Mahlberg, and G. H. Rothblat, “Formation of cholesterol monohydrate crystals in macrophage-derived foam cells,” J. Lipid Res. 35, 93–104 ( 1994). [PubMed]