Coherent anti-Stokes Raman scattering (CARS) microscopy based on inherent vibration of molecule bonds permits label-free chemically selective live cell imaging. This work presents the implementation of a complementary microscopy technique on a standard multiphoton laser scanning microscope based on a femtosecond Ti:sapphire laser and an OPO laser.
Laser scanning microscopes combining a femtosecond Ti:sapphire laser and an optical parametric oscillator (OPO) to duplicate the laser line have become available for biologists. These systems are primarily designed for multi-channel two-photon fluorescence microscopy. However, without any modification, complementary non-linear optical microscopy such as second-harmonic generation (SHG) or third harmonic generation (THG) can also be performed with this set-up, allowing label-free imaging of structured molecules or aqueous medium-lipid interfaces. These techniques are well suited for in-vivo observation, but are limited in chemical specificity. Chemically selective imaging can be obtained from inherent vibration signals based on Raman scattering. Confocal Raman microscopy provides 3D spatial resolution, but it requires high average power and long acquisition time. To overcome these difficulties, recent advances in laser technology have permitted the development of nonlinear optical vibrational microscopy, in particular coherent anti-Stokes Raman scattering (CARS). CARS microscopy has therefore emerged as a powerful tool for biological and live cell imaging, by chemically mapping lipids (via C-H stretch vibration), water (via O-H stretch vibrations), proteins or DNA. In this work, we describe the implementation of the CARS technique on a standard OPO-coupled multiphoton laser scanning microscope. It is based on the in-time synchronization of the two laser lines by adjusting the length of one of the laser beam path. We present a step-by-step implementation of this technique on an existing multiphoton system. A basic background in experimental optics is helpful and the presented system does not require expensive supplementary equipment. We also illustrate CARS imaging obtained on myelin sheaths of sciatic nerve of rodent, and we show that this imaging can be performed simultaneously with other nonlinear optical imaging, such as standard two-photon fluorescence technique and second-harmonic generation.
Optical microscopy has become a major technique for nondestructive visualization of dynamic processes in living biological systems with a subcellular resolution. Fluorescence microscopy is currently the most popular imaging contrast used in live cells due to its high specificity and sensitivity1. A large palette of fluorescent probes has emerged (exogenous dyes, genetically encoded proteins, semiconductor nanoparticles). Various sample illumination fluorescent-based techniques have flourished (such as confocal or two-photon microscopy) to perform 3D imaging and to reduce a main drawback of this technique which is photobleaching2. Other limitations include the requirement of fluorophore labeling because most of molecular species are not intrinsically fluorescent and therefore these fluorophores have to be artificially introduced in the imaged sample. This artificial manipulation may be disruptive especially for small molecules or induces potential photo-toxicity. These reasons make fluorescence microscopy not well suited for in-vivo observations. Hence, the use of optical imaging techniques with high sensitivity and specific molecular contrasts without the use of fluorescent molecules is highly desirable in biomedical science.
Several nonlinear optical imaging techniques without labeling or staining have emerged, including second-harmonic generation (SHG)3,4 and third-harmonic generation (THG)5. SHG microscopy has been used to image structural arrangements at the supramolecular level such as microtubules or collagen6. THG is generated from optical heterogeneities such as interface between an aqueous medium and lipids7. THG was also demonstrated to image myelin8,9. Both techniques can be implemented on a two-photon fluorescence microscope and require only one laser beam. However they require high power laser intensity (typically 50 mW at 860 nm for SHG10, 25 – 50 mW at 1,180 nm for THG9), which is deleterious in living samples, and do not provide the chemical specificity that is required to unambiguously image specific biological structures.
Chemically selective imaging can be obtained from inherent molecular vibration signals based on Raman scattering. When a beam of light hits matter, photons can be absorbed and scattered by atoms or molecules. Most of the scattered photons will have the same energy, i.e., frequency, as the incident photons. This process is called Rayleigh scattering. However, a small number of photons will be scattered at an optical frequency different from the frequency of the incident photons, i.e., with an inelastic scattering process called Raman scattering. The difference in energy originates from excitation of vibrational modes depending on molecular structure and environment. Therefore, spontaneous Raman scattering provides chemically selective imaging as different molecules have specific vibrational frequencies. However it is limited because of its extremely weak signal. Confocal Raman microscopy has been developed and provides 3D spatial resolution, but it requires high average power and long acquisition time11. To overcome these difficulties, recent advances in laser technology have allowed the rise of nonlinear optical vibrational microscopy, in particular coherent anti-Stokes Raman scattering (CARS)11,12,13.
CARS is a third-order nonlinear optical process. Three laser beams, composed of a pump beam at frequency ωP, a Stokes beam at frequency ωS and a probe beam (most often being the pump) are focused in a sample and generate an anti-Stokes beam at frequency ωAS= (2ωP – ωS)14. The anti-Stokes signal can be significantly enhanced when the frequency difference between the pump and the Stokes beams is tuned to a Raman molecular vibration ΩR=(ωP – ωS). CARS signal is based on multiple photon interaction. It generates therefore a coherent signal orders of magnitude stronger than spontaneous Raman scattering.
CARS microscopy was first experimentally demonstrated by Duncan et al.15. Zumbusch et al. improved then the technique, by using two focused near-infrared femtosecond laser beams with an objective lens of high numerical aperture, allowing the phase matching condition of CARS and avoiding the two-photon non-resonant background16. CARS microscopy has therefore emerged as a powerful tool for live cell and tissues imaging, by chemically detecting molecules such as lipids (via C-H stretch vibration)17,18, water (via O-H stretch vibrations), proteins, DNA in live cells19,20 but also deuterated chemical compounds for pharmaceutical21 and cosmetic applications22.
The major limitation of nonlinear microscopy originates from the complexity and the cost of the optical sources. A CARS system requires two wavelength tunable lasers with short pulse durations and with temporally and spatially synchronized pulse trains. Early CARS microscopes were based on two synchronized picosecond Ti:sapphire lasers20. CARS imaging was also obtained from a single femtosecond Ti:sapphire laser generating a supercontinuum light source23. Recently, laser sources composed of a single femtosecond Ti:sapphire laser pumping a tunable optical parametric oscillators (OPO) have been used for CARS microscopy. This set-up allows intrinsically temporally synchronized beams with a difference of frequency between the pump and the Stokes beam covering the full molecular vibrational spectrum24. In addition, laser scanning microscopes based on a turn-key fs laser and an OPO, primarily used for two-photon fluorescence (TPF) are now available for non-physicists. The potential of such set-ups can be greatly enhanced without requiring supplementary investment by the incorporation of other nonlinear optical imaging, since each nonlinear (NLO) imaging modality is sensitive to specific structures or molecules. Multimodal NLO imaging therefore capitalizes the potential of NLO microscopy for complex biological samples25. The coupling of these techniques has allowed the investigation of many biological questions, in particular on lipid metabolism, skin or cancer development26, skeletal muscle development27, atherosclerotic lesions28. Moreover, the implementation of laser beam scanning with CARS gives the capability of high-rate imaging, i.e., an appealing tool to study dynamical processes in vivo.
The aim of this work is to show each step to implement the CARS technique on a standard multiphoton laser scanning microscope. The microscope is based on a fsec Ti:sapphire laser and an OPO (pumped by the Ti:sapphire laser) operated by a software for biologists. The integration was performed by adjusting the length of one of the laser beam path in order to synchronize in time the two beams. We describe the step-by-step implementation of this technique which requires only basic background in experimental optics. We also illustrate CARS imaging obtained on myelin sheaths of sciatic nerve of rodents, and we show this imaging can be performed simultaneously with other nonlinear optical imaging, such as standard two-photon fluorescence technique and second-harmonic generation.
Figure 1. Schematic view of the general set-up. It includes the Ti:sapphire (680 – 1,080 nm) and the OPO (1,050 – 1,300 nm) lasers, the delay line with the 4 mirrors (M1 to M4), the fast oscilloscope, the photodiode and two fixed iris diaphragms I1 and I2. Mirrors M2 and M3 are fixed on a linear translation stage enabling to change the delay line length with a micrometer resolution. A 660 – 685 nm band pass filter was positioned in front of the PhotoMultiplier Tube (PMT) used for CARS imaging. Please click here to view a larger version of this figure.
1. Startup of the Laser System
2. Microscope Settings
3. Temporal Synchronization
Note: The two laser beams originate from the same Ti:sapphire laser but the OPO beam is delayed when it is generated so the two beams are not synchronized in time when they reach the microscope. The goal here is to delay one of the two beams to re-synchronize them in time before they reach the microscope.
4. Spatial Overlap of the Beams
Note: To produce a CARS signal, the spatial overlapping of the two laser beams is required. The alternate illumination of both beams on the same beads stained throughout with two different fluorescent dyes can be used to indicate the spatial shift. Fine adjustments of the mirror positions can then minimize the shift.
5. Final Adjustments and Coherent Anti-Stokes Raman Scattering (CARS) Signal Observation from Olive Oil Droplets
6. Enclosure of the Light Path of the Delay Line
7. Wavelength Tuning for CARS
8. Observation of CARS Signal and Stained Myelin from Sciatic Nerve Cuts
Note: All animal experiments were conducted in accordance with institutional regulations.
9. Observation of CARS and SHG Signals from Sciatic Nerve Cuts
The pulse train frequency of standard Ti:sapphire laser is typically around 80 MHz. The OPO has the same frequency since it is pumped by the Ti:sapphire laser. A fast oscilloscope of at least 200 MHz is therefore required. A fast photodiode in the range 600 to 1,100 nm is also required. The maximal temporal shift occurs when the Ti:sapphire and the OPO signals are shifted of 1/(2×80×106) = 6.2 nanoseconds. It corresponds to a maximum beam path shift of 1.9 m. Figure 2 shows the pulse train recorded from the OPO laser beam (A) and from the Ti:sapphire laser beam without the delay line (B) and with the adjusted delay line (C). The pulse trains are thus roughly synchronized in time with the implemented delay line as shown in Figure 2A and 2C.
Figure 2. Laser pulse train recordings for the temporal synchronization. Recording of pulse trains at the microscope sample plane with a 10X objective of the OPO beam (A), the Ti:sapphire laser beam without (B) and with the delay line (C). The top curve of each graph plots the signal directly recorded from the BNC connector on the back of the Ti:sapphire laser (Sync. out connector). The oscilloscope trigger is adjusted on this signal. Please click here to view a larger version of this figure.
The spatial overlap of the two beams can be obtained by the visualization of fluorescent polystyrene beads fixed on top of a microscope slide (Figure 3A). Figure 3B shows the image of three nearby beads given by the OPO laser at 1,107 nm (red false color) and by the Ti:sapphire laser at 830 nm (green false color). The spatial shift between both laser beams has been compensated following the procedure of the spatial synchronization previously described (Figure 3C).
Figure 3. Spatial overlap of the beams. General view of the fluorescent microspheres (A). Zoom on 3 adjacent microspheres. The illumination of beads at 830 and 1,107 nm shows a spatial shift of the bead images (false colors) (B). After x, y, z adjustments, the bead images are merged (C). Please click here to view a larger version of this figure.
The final step to achieve the precise temporal adjustment to activate CARS can be easily realized by imaging olive oil drops (containing numerous C-H bonds32) and by moving slightly the mirrors of the delay line to complete the synchronization in time of both laser pulse trains. Figure 4 shows images of an olive oil droplet delivered from the Ti:sapphire laser illumination only (Figure 4A) and from the OPO illumination only (Figure 4B). Both beams simultaneously applied induce a clear signal enhancement (Figure 4C) at 670 nm, which corresponds to a CARS signal from intrinsic carbon-hydrogen (C-H) stretching vibration with a Raman vibration range around 2,800 – 3,100 cm-1. CARS signal (C) disappears when one the two beams is switched off. Since olive oil contains natural fluorophores33, in particular chlorophyll molecules emitting in the range 650 – 670 nm, the weak signal in (A) and (B) is a multiphoton fluorescence process. The intensity of these signals is several orders of magnitude lower than the CARS signal. Therefore, it does not perturb the measurement of the CARS signal.
Figure 4. Final fine temporal adjustments for CARS signal outbreak. Olive oil droplets can be used to tune in time both beams by adjusting the delay line length with the linear translation stage with micrometer drive. Droplets under (A) 830 nm laser light only and under (B) 1,107 nm laser light only show a weak signal. Under simultaneous illumination (C), clear signal enhancement is observed, corresponding to the CARS signal. Scale bars are 100 μm. Raman vibration peak at 3,015 cm-1. Please click here to view a larger version of this figure.
The myelin sheath is a biological structure produced by specific cells that wrap and condensate their plasma membrane around axons of neurons in central and peripheral nervous systems34. This sheath is highly enriched in different lipids. We used here the myelin sheaths of the sciatic nerve of a mouse. Transversal and longitudinal sciatic nerve cross-section segments have been stained for two-photon fluorescence imaging of myelin at 1,095 nm as shown in Figure 5B and Figure 5E, respectively. A fluoromyelin red dye, which has selectivity for myelin was used35. These samples have been simultaneously illuminated at 830 and at 1,095 nm and the CARS signal was observed (Figure 5A and Figure 5D). In Figure 5A-C, the circles correspond to myelin sheaths around axons (empty space inside the rings) in transversal cut. As shown in Figure 5C and 5F, the same structure is found while overlapping CARS and fluorescence images. We conclude that similar level of details can be obtained with CARS and therefore without the use of fluorescent labeling.
Figure 5. CARS and stained myelin imaging of sciatic nerve cuts. (A) Transversal and (D) longitudinal cuts of label-free CARS imaging of myelin sheaths. (B) Transversal and (E) longitudinal cuts of the same sample stained by the fluoromyelin red dye under two-photon fluorescence illumination at 1,095 nm. (C) Transversal and (F) longitudinal cut overlap of both images. Scale bars are 10 μm. Average optical powers were 5 mW at 830 nm and 16 mW at 1,095 nm for CARS signal. Average optical power was 8 mW at 1,095 nm for two-photon fluorescence imaging. Raman vibration peak at 2,915 cm-1. CARS signal disappears by stopping the scan of one the laser or by moving the translation stage out of the Raman resonance. Please click here to view a larger version of this figure.
Additionally to CARS imaging, the available microscopy system allows to simultaneously generate second harmonic from the collagen fibers at the outer surface of the sciatic nerves. This additional image only requires the use of a different detector to record a signal at 550 nm (half the wavelength of the OPO) since a narrow band-pass filter at 670 nm is used for the CARS signal recording. Figure 6 shows a contrast image of myelin sheaths depicted in false green color (CARS signal only in Figure 6A) surrounded by collagen fibers illustrated in false magenta color (SHG only in Figure 6B). CARS signal was achieved with average optical powers of 4 mW at 830 nm and 13 mW at 1,095 nm, while it required 50 mW at 1,095 nm to obtain the SHG signal. Therefore much lower optical energy is required to obtain CARS signal compared to SHG or THG (not shown) signals in our biological samples.
Figure 6. CARS and SHG imaging of mouse sciatic nerve. (A) CARS imaging, (B) SHG imaging and (C) simultaneous visualization of myelin by CARS imaging (in green false color) and of collagen by SHG imaging (in magenta) on the mouse sciatic nerve. Average optical powers were 4 mW at 830 nm and 13 mW at 1,095 nm for CARS. Average optical power was 50 mW at 1,095 nm for SHG. Scale bar is 10 μm. Raman vibration peak at 2,915 cm-1. Please click here to view a larger version of this figure.
The most challenging part of the work is the temporal synchronization of the laser beams. It requires a fast photodiode combined with a fast oscilloscope, but only a rough overlapping in time can be performed at first. Then a further adjustment of few cm is required. Finally, micrometer moves by a linear translation stage allows performing the final fine adjustment of the delay line length in order to trigger the CARS signal. This signal is maintained in a narrow range of around 20 micrometers, as observed by tuning the translation stage micrometer drive. It corresponds to a peak shift of the beams of around half of the pulse duration, i.e., 140 fsec. The spatial overlapping of the two beams which is required for the CARS signal is similar to a normal beam alignment for such laser scanning microscope system.
The implementation of the delay line can be achieved in few weeks by biologists having a basic background in experimental optics or in collaboration with physicists. Special care has to be taken for the choice of the mirrors to minimize reflection losses. If the majority of the materials to build the delay line is affordable, the oscilloscope which is required is not a standard oscilloscope. Thus, since few days are enough to complete the temporal synchronization procedure, borrowing the oscilloscope from a physic laboratory appears to be the best option. Spatial overlapping can be time consuming since it demands the tuning of several mirrors and lenses (which can be motorized depending on the laser scanning system). Earlier CARS laser system based on two electronically synchronized Ti:sapphire laser had an issue with long term (tens of minutes to hours) temporal pulse overlapping. This problem has been totally solved with OPO pumped with a unique laser source20,21. With such a laser system, and using firmly clamped optics, no temporal walk-off between pump and Stokes beams was observed over months of operation.
A major limitation of the CARS technique is that the CARS signal can occur without the presence of resonant molecules (from the solvent or the scatterers)11. This non-resonant background can limit the sensitivity. Ways to minimize this background include the configuration of detection (forward or backward detection), the laser characteristics (femtosecond or picosecond), the choice of the solution used to immerse the objective or image treatment. A backward detection allows to minimize the non-resonant background11 as it was used in this work (Figure 1). Femtosecond Ti:sapphire lasers provide high peak intensities highly desirable for non-linear optical processes. However, in the spectral domain, the width of most Raman lines is smaller than the spectral width of a femtosecond pulse whereas it fits the spectral width of a picosecond pulse. Therefore the use of picosecond laser increases the ratio of resonant signal to non-resonant background. For C-H bonds in lipid, because of the high density of the bonds, the non-resonant background is negligible for a femtosecond laser. To reduce the background effect, the subtraction of a reference image recorded under off-resonant can also be applied (not done for this work).
Despite the utilization of near-infrared laser beams allowing deep penetration in tissue (200 – 300 microns), the practical penetration depth obtained in nerve imaging was around 50 microns with the experiments conducted in this work. However this parameter is heavily dependent of the biological tissue imaged and, because all myelinated fibers are aligned and concentrated in a small space, peripheral nerves probably do not constitute the best samples for evaluating this light penetration. In addition, few seconds are required to record high resolution images (1,024×1,024 pixels), which can limit the visualization of in vivo samples since even anesthetized, the breathing can induce slight movements.
CARS microscopy allows chemically specific selective imaging with a resolution close to the diffraction limit. This technique does not require any labeling, and since the Ti:sapphire laser is tunable from 700 to 1,000 nm and the OPO laser from 1,100 to 1,300 nm, it covers the entire spectrum region of molecular vibrations in biological systems (from 500 to 7,000 cm-1). Thus, it is applicable and highly discriminative to DNA, proteins, lipids, or water, while the stronger signal arises from the lipid C-H stretching bond. Photodamage is also reduced due to the low level of laser power used to generate the CARS signal compared to the second-harmonic generation imaging. Since CARS imaging can be realized with low laser power and without invasive tags, it is therefore a tool of choice for in vivo studies.
Laser scanning microscopes with a femtosecond Ti:sapphire laser source combined with an OPO have become largely available for two-photon microscopy in many biological laboratories. Since the spatial synchronization is already implemented in these systems, the temporal overlapping procedure of the two beams described in this work brings an additional technique of microscopy without expensive additional equipment. Thus, the modified set-up allows for simultaneous imaging of cells or living tissue with Second or Third Harmonic Generation (SHG, THG), two-photon fluorescence and CARS. CARS technique has demonstrated great potential for in vivo imaging of live animals allowing the study of dynamical processes (without stains or genetic manipulations). Promising applications include in vivo label-free non-invasive lipid imaging, used for instance for real-time imaging of neuronal myelin in live spinal tissues36, for studying lipid-related diseases (obesity, demyelinating and cardiovascular diseases)37, human skin penetration mechanisms22 or skin disorders38. We believe that future improvements will concern the use of special objective lenses, such as GRIN micro-objectives or miniature objective to enhance the penetration depth of the CARS signal.
The authors have nothing to disclose.
The authors want to thank Dr. Philippe Combette (IES, UM, Montpellier, France) for the loan of the fast oscilloscope and acknowledge financial supports from Montpellier RIO Imaging (MRI). HR acknowledges ANR grants France Bio Imaging (ANR-10-INSB-04-01) and France Life Imaging (ANR-11-INSB-0006) infrastructure networks for coherent Raman imaging developments. This work was mainly supported by an European Research Council grant (FP7-IDEAS-ERC 311610) and an INSERM – AVENIR grant to NT.
Oscilloscope | Tektronix | TDS 520D | 500 MHz |
Photodetector | Thorlabs | DET08C/M, T4290 | 5 GHz InGaAs, 800-1700 nm |
Ti:Sapphire laser Chameleon Ultra Family II | Coherent | ||
Optical parametric oscillator OPO Compact Family | APE Berlin | ||
Axio Examiner microscope LSM 7 MP | Carl Zeiss | ||
Motorized periscope | Newport | ||
Objective W Plan-Apochromat 20x/1.0 | Carl Zeiss | ||
Beam combiner | Carl Zeiss | ||
Acousto-optic modulator | Carl Zeiss | ||
OPO power attenuator | Carl Zeiss | ||
Photomultiplier tube | Carl Zeiss | ||
ZEN software | Carl Zeiss | ||
Bandpass filters | Carl Zeiss | LSM BiG 1935-176 | 400-480 nm ; 500-550 nm ; 465-610 nm |
Dichroic mirror | Carl Zeiss | Cutoff wavelength 760 nm | |
Silver mirrors | Newport | 10D20ER.2 | λ/10, 480-20,000 nm , Quantity 4 |
Single-axis translation stage with standard micrometer | Thorlabs | PT1/M | Quantity 1 |
Aluminium breadboard | Thorlabs | MB1015/M | Quantity 1 |
Mirror mount | Thorlabs | KMSS/M | Quantity 4 |
Mirror holder for Ø1" Optics | Thorlabs | MH25 | Quantity 4 |
Iris diaphragms | Thorlabs | ID8/M | Quantity 3 |
Protective box | Thorlabs | TB4, XE25L900/M, T205-1.0, RM1S | Quantity 1 |
Optical posts | Thorlabs | TR40/M, PH50/M, PH75/M, BA2/M | Quantity 8 (lengths depending on the set-up) |
661-690 nm bandpass filter | Semrock | 676/29 nm BrightLine® single-band bandpass filter | Quantity 1 |
Fluorescent beads | ThermoFisher | TetraSpeck™ Fluorescent Microspheres Size Kit | |
Laser viewing card | Thorlabs | IR laser viewing card | |
Laser safety glass | Newport | LV-F22.P5L07 | |
FluoroMyelin™ Red Fluorescent Myelin Stain | ThermoFisher | F34652 |