1. Introduction

Since its initial demonstration in 1990 [1], multi-photon microscopy (MPM) has become one of the most powerful imaging modalities in diverse areas of biology, including neuroscience, immunology, embryology and cancer research [2, 3]. Because of its highly spatially confined ‘nonlinear’ excitation, MPM does not suffer from emission generated out of focal volume and can achieve sub-micrometer, three-dimensional (3D) optical resolution without a detector pinhole. Since virtually all of the emission light contributes to useful signal, the detection systems for MPM usually have high efficiency that enhances imaging at larger depths in optically turbid media. In addition, the long excitation wavelengths used in MPM allow deeper penetration into biological tissue and produce less damage to specimens [4]. In addition to two-photon fluorescence excitation (2PE) [2–9], which is the most commonly used in MPM, other nonlinear processes have been applied for imaging including: second harmonic generation (SHG) [10–16], coherent anti-Stokes Raman scattering [17], three-photon fluorescence excitation (3PE) [18–21] and third harmonic generation (THG) [22–26].

One of the main limiting factors preventing widespread use of MPM is the high cost and complexity of short pulse laser sources [4, 27]. The price of the laser system can constitute up to 50% of the total cost of commercial multi-photon microscopes, and the percentage could be significantly higher for the custom-built microscopes [4]. In the following paragraphs we will briefly describe the requirements for the laser sources in MPM and the reasons for their high cost. More details on this subject can be found in [2, 4, 27–30].

Several parameters of the laser, such as average power (P_ave), pulsewidth (τ), pulse repetition rate (f), and central wavelength (λ_exc), are critical for the quality of the MPM image. Due to the nonlinear nature of excitation in MPM, the average emission signal intensity is proportional to Pⁿ _ave/(τf)^n-1, where n is the order of the nonlinear process. One can ideally increase the average power of the laser in order to increase the average emission signal intensity; however, the average power that can be applied to a biological sample is limited by destructive thermal effects, photo-toxicity and photobleaching.

Although it has been demonstrated that using very short pulses (~10 fs) may be beneficial for MPM [31], pulse durations of 50–100 fs offer several practical advantages [2, 32]. It is challenging to compensate for the effects of group delay dispersion (GDD) of the microscope optics for very short pulses, since higher order dispersion effects become dominant for broad bandwidths. Without careful dispersion precompensation a 10-fs pulse may even produce a weaker fluorescent signal than a 100-fs pulse [31]. Also, laser sources generating 50–100 fs are more common than 10-fs lasers. Finally, the spectrum of a very short pulse may be wider than the absorption spectrum of the fluorophore. This reduces the excitation efficiency and can also make it difficult to separate emission from excitation light.

Higher pulse repetition rates of the laser source may increase the fluorescence emission rate and reduce the required scan time, if the average power is not a limiting factor. However, most fluorophores used in MPM have an upper state lifetime of ≈5 ns, and for very high repetition rates (f>200 MHz) absorption saturation may limit the obtainable signal strength [30] and decrease the resolution due to accumulation of fluorophores in the excited state. A pulse repetition rate of ~100 MHz is close to an optimum choice for many MPM applications [8, 28, 30]; for historical reasons, most lasers used in MPM have a repetition rate of 80 MHz.

The excitation wavelengths for MPM are typically in the near-infrared spectral region. In particular, two-photon absorption (2PA) cross sections of the most common fluorophores are centered between 700 nm and 1000 nm and are usually broader and blue-shifted with respect to corresponding one-photon spectra, with full-width half-maximum (FWHM) bandwidths typically exceeding 100 nm [8, 21, 33–41]. Therefore, it is usually not critical to tune the excitation wavelength to the exact maximum of the 2PA cross section when imaging a single fluorophore. However, wavelength tunability may be very important for imaging using several fluorophores in the same sample [38]. The maximum two-photon ‘action’ cross section for common fluorophores is in the range of 1–300 GM, and depending on this cross section, only a few tens of milliwatts or less of the excitation power at the focus of high numerical aperture (NA) objective is sufficient to saturate the excitation, assuming 100 MHz pulse repetition rate and 100 fs pulsewidth [1, 8, 21]. Propagation of the optical beam through the microscope optics and focusing deep into the scattering biological tissue produce attenuation from absorption and scattering as well as optical aberrations, which reduce the excitation efficiency within the focal volume. Investigators typically compensate for this reduced efficiency by increasing the laser power. For a laser producing 50–100 fs pulses at a repetition rate of ~100 MHz, an average power of 200 mW is sufficient for most MPM applications [2, 4, 8, 21, 42].

Table 1 provides a representative list of laser sources that have been used for MPM. In its early years, MPM imaging was performed using colliding pulse dye lasers [1, 2, 30]: the only source at the time that could generate sub-picosecond pulses with sufficient peak powers, at repetition rates of ~100 MHz [30]. However, due to their complexity as well as to maintenance issues [2, 27], dye lasers are difficult to use for MPM. The real breakthrough in

Table 1. Representative list of laser sources that have been used for multi-photon microscopy. OPO: Optical parametric oscillators, PCF: Photonic crystal fiber, SHG: Second harmonic generation. * Denotes the Cr:LiCAF laser that will be described in this work.

View Table

MPM came with the development of Ti:Sapphire lasers, which are much more powerful and user friendly than dye lasers [2, 58]. Today’s commercial Ti:Sapphire lasers can generate sub-100-fs pulses with hundreds of kW peak power and are broadly tunable (680 to 1080 nm) [43–45]. The tunability can be extended further to the visible and mid-infrared regions using optical nonlinear schemes such as optical parametric oscillators (OPOs), photonic crystal fibers (PCFs) and second harmonic generation (SHG) [46, 47]. The majority of today’s MPM systems use fully automated Ti:Sapphire lasers, where the central wavelength and average power of the laser are adjusted under computer control [43, 44]. The main disadvantage of Ti:Sapphire femtosecond laser technology is its high cost. Ti:Sapphire gain medium requires pumping in the green spectral region. Unfortunately, direct diode pumping is not available at these wavelengths, and pump sources use the second harmonic generated by the diode-pumped Nd-doped laser systems. These Nd-based systems are bulky and a standard pump laser costs in the range of $50-100k, making the overall cost of the Ti:Sapphire lasers and the MPM systems quite high, thus limiting widespread adoption of MPM [2, 4, 27].

Beyond the popular Ti:Sapphire laser, several other lasers have been demonstrated for MPM. One alternative is Neodymium or Ytterbium based systems that can generate ~200-fs pulses with ~100 kW of peak power at ~1050 nm wavelengths [30, 48, 49]. The main disadvantage of these rare earth element lasers is their narrow emission bandwidth, which prevents tunability. Most fluorophores used in MPM have a TPE spectrum centered below 1000 nm [8, 21, 34–37] and these laser sources cannot reach this part of the spectrum directly. However, as with Ti:Sapphire, nonlinear effects can be used to generate tunable outputs from these sources, at the expense of increased complexity and cost [50–52].

Diode lasers, which are compact, easy to use and low-cost, represent another possible light source for MPM. Gain-switched diode lasers can provide picosecond pulses and by using external amplifiers, peak powers above ~1 kW level can be obtained [53–56]. Repetition rate is easily adjustable by simply changing the current driving frequency, which might be advantageous in some MPM experiments. However, low peak powers and issues related to amplified-spontaneous-emission and tunability currently limit their effective use in MPM.

In this paper, we describe the Cr:LiCAF laser as a promising light source for MPM. Cr:LiCAF belongs to the family of Cr³⁺-doped colquiriite solid-sate laser gain media, where the most well-known members are: Cr:LiSAF [59–64], Cr:LiSGaF [65], and Cr:LiCAF [66–69]. Similar to Ti:Sapphire, Cr:Colquirites have broad emission bandwidths around 800 nm, enabling the generation of pulses as short as ~10 fs [69]. The possibility of developing a low cost and efficient femtosecond technology based on diode pumped Cr:Colquirites was suggested more than five years ago with the demonstration of battery powered femtosecond operated Cr:LiSAF laser [61–64]. The main advantage of Cr:Colquirites over Ti:Sapphire is that they can be directly diode-pumped by inexpensive diodes in the red region of the spectrum; hence, the total material costs of an entire laser system could be below ~$10k [27]. We believe that, with the recent advances, Cr:Colquiriites lasers have the potential to become efficient and low-cost excitation light sources for MPM and could enable wide-spread use of MPM technology, by reducing cost.

Cr:Colquiriite lasers have previously been demonstrated for MPM imaging (Table 1) [42, 57]. Using an 500-mW master oscillator power amplifier (SDL, Inc.) with a diffraction limited beam as the pump source, Svoboda et al. developed a Cr:LiSAF laser producing 90-fs pulses with ~0.3 nJ pulse energy at ~860 nm and a repetition rate of 150 MHz for MPM [42]. However, this pump source is very expensive and the available peak power from the laser was below 5 kW, limiting deep tissue imagining in scattering media. A multimode diode pumped Cr:LiSAF laser with similar specifications of 100–200 fs pulse duration, 0.375 nJ pulse energy at 80 MHz repetition rate was also used to demonstrate MPM imaging; however, peak powers were limited [57].

The Cr:LiCAF laser used in this study employs inexpensive single mode diodes, costing only ~$150 each, as the pump source and uses double-chirped mirrors for dispersion compensation. A semiconductor saturable absorber mirror (SESAM) [70], also referred to as a saturable Bragg reflector (SBR) [71], was used for mode locking, which enables self-starting and robust laser operation. The laser produces ~70 fs long pulses (FWHM), with 1.8-nJ of energy, centered around 800 nm wavelengths, with a repetition rate of 100 MHz. The average output power is 180 mW, and the peak power is as high as 25 kW. This is approximately an order of magnitude higher peak power than earlier Cr:Colquiriite lasers that were used for MPM, and the laser design is also simpler. These improvements over the previous lasers are the result of advances in high-power visible diodes, mirror coatings, SESAMs/SBRs and the quality of Cr:Colquiriite crystals.

This paper is organized as follows: Section 2 describes the Cr:LiCAF laser setup used in the MPM studies. Section 3 describes methods for representative examples of MPM imaging with the Cr:LiCAF laser in neuroscience, immunology, endocrinology and cancer research. Section 4 presents results and Section 5 discussion and conclusions.

2. Low-cost and highly efficient femtosecond Cr:LiCAF laser

Figure 1 shows a schematic of the single-mode diode-pumped Cr³⁺:LiCAF laser used in these multi-photon microscopy experiments. A similar version of this laser was recently described in [68], so only a brief description is presented here. Four microlensed single-mode diodes at ~660 nm (D1-D4) with a diffraction-limited circular beam profile, each costing only ~$150, were used as the pump source (VPSL-0660-130-X-5-G, Blue Sky Research). At a drive current of 220 mA (2.6 V), each diode provided up to ~160–170 mW of output power, without requiring water cooling. Polarizing multiplexing of the linearly polarized diode outputs using polarizing beam splitter (PBS) cubes resulted in up to ~660 mW of incident pump power on the Cr:LiCAF (x-tal). Two 65-mm focal length lenses were used to focus the pump beams inside the crystal. The 2.5-mm long, 11 mol. % chromium-doped Cr3+:LiCAF crystal absorbed ~600 mW of the incident pump power. Water cooling of the crystal was not required at these low pump power levels. The continuous-wave (cw) laser resonator was a standard x-folded cavity, with two curved pump mirrors, each with a radius of curvature of 75 mm (M1 and M2), a flat end high reflector (M3), and a flat 2% output coupler (OC). The laser output beam was circular and had a TEM₀₀ transverse mode structure. In cw operation, the laser produced more than 270 mW of output power at ~785 nm, with a slope efficiency of about 50% with respect to absorbed pump power. Using a birefringent filter, it was possible to tune the cw laser wavelength from 754 nm to 871 nm.

 

Fig. 1. Schematic of the single-mode diode-pumped Cr³⁺:LiCAF laser used in multi-photon microscopy experiments. D1-D4: Single-mode pump diodes at 660 nm, HR: 45° high reflectors at 660 nm, f: visible achromatic doublets with a focal length of 65 mm, PBS: polarizing beam splitter cube, M1-M2: pump mirrors with R=75 mm, M3: flat high reflector, M4: curved high reflector with R=150 mm, OC: 2% output coupler, DCM1-DCM3: flat double-chirped mirrors with ~-50 fs² dispersion per bounce, SESAM/SBR: semiconductor saturable absorber mirror/saturable Bragg reflector. Dashed lines indicate the mode-locked laser cavity.

Download Full Size | PDF

 

Fig. 2. Measured spectrum and second harmonic autocorrelation taken with the single-mode diode-pumped mode locked Cr³⁺:LiCAF laser using the 2% output coupler at an absorbed pump power of ~600 mW. The FWHM of the autocorrelation is 108 fs, corresponding to a ~70-fs pulse duration (assuming sech² pulse shape). The average output power is 180 mW, corresponding to a pulse energy of 1.8 nJ for the 100-MHz repetition rate cavity. The spectrum has a bandwidth of 10.2 nm (FWHM) centered around ~800 nm. The corresponding time bandwidth product is ~0.335.

Download Full Size | PDF

For mode locked operation, the flat high reflector (M3) was replaced with a curved high reflector (M4, R=150 mm) and a flat SESAM/SBR. The DCM mirrors had ~-50 fs² of GVD per bounce and produced a total round trip cavity GVD of ~-100 fs². The SESAM/SBR initiated and sustained mode locking, but the pulse shaping occurred mainly by a soliton pulse-shaping mechanism. The SESAM/SBR mode locked laser was self-starting, insensitive to environmental fluctuations and did not require careful cavity alignment, enabling turn-key operation. When mode locked, the laser produced 70-fs pulses (assuming sech² pulses) with 180 mW average power and 10.2 nm spectral bandwidth near 800 nm at 100 MHz (~1.8-nJ pulse energy) (Fig. 2). The corresponding peak powers were as high as 25 kW. The time bandwidth product was ~0.335, close to the transform limit of 0.315 for sech² pulses. The demonstrated electrical-to-optical conversion efficiency of the laser was close to ~8%.

The Cr:LiCAF laser has a simple layout, uses commercial optomechanics (except the SESAM/SBR), and the total material cost of the entire laser system could be reduced below ~$10k. The laser has also a low maintenance cost compared with traditional frequency doubled pump lasers [30] and very low electrical consumption. Moreover, the laser does not require water cooling and the diodes can be powered by batteries [63, 64]; hence, the laser can be made compact and portable. Due to SBR/SESAM-initiated mode locking the laser is quite stable, thus enabling long-term use in a laboratory environment.

The main disadvantage of the Cr:LiCAF laser compared to Ti:Sapphire is its limited tunability. Ti:Sapphire is tunable from 680 to 1080 nm in mode locked operation [43, 44], whereas for Cr:LiCAF we expect tunability from 750 to 850 nm. This limitation will be discussed in more detail in Section 5. The emission cross section of Cr:LiCAF is ~30 times smaller than that of Ti:Sapphire and for SESAM/SBR mode locked lasers, this causes higher susceptibility to q-switching instability during mode locked operation [72, 73]. However, the SESAM/SBR parameters and laser cavity design can be adjusted to obtain stable cw mode locking. Cr:LiCAF has lower gain than Ti:Sapphire and requires the use of low loss optics. This is not a problem with today’s optical coating technology and slope efficiencies approaching the intrinsic slope efficiency of 69% were obtained using commercial mirrors. Lastly, the output power level for the Cr:LiCAF laser is limited by the available pump power. In this work, we demonstrate that pumping with relatively low-power and low-cost diodes provides sufficient output powers for most multi-photon imaging applications.

3. Methods

3.1 Multi-photon microscopy setup

Imaging was performed using a commercial two-photon microscope (Ultima, Prairie Technologies, Inc.). Excitation was provided in the epi-illumination configuration and scanning a single optical beam in the xy plane, perpendicular to the optical axis z, was performed by a pair of conventional galvanometer-based scanners. The fine positioning of the microscope objective along the optical axis was controlled by a motorized stage. The emission from the sample was reflected by a high-pass dichroic mirror positioned close to the back aperture of the objective and detected by a four-channel detector. The emission light inside the four-channel detector was split by the dichroic mirror into two arms, each containing a filter cube and a pair of photomultiplier tubes (PMTs). The PMTs output current was amplified and the signal digitized by a 12-bit analog-to-digital converter (ADC) at 2.5x10⁶ samples/sec. The intensity of a single pixel in the image frame was calculated as a sum of all of the ADC samples obtained during the pixel dwell time. A 3×3×3 median filter and histogram equalization was used for image processing.

The total optical intensity transmission efficiency of the 800-nm laser light through the microscope together with the Olympus LUMPlanFI/IR 40X WI objective (NA=0.8; working distance=3.3 mm; water immersion) used to image the most demanding samples in this study was 20%. This relatively low transmission was mostly due to a pair of additional dichroic mirrors in our Ultima microscope setup, and additional losses because of sub-optimal beam diameter. A continuously variable optical neutral density filter was placed in front of the laser and manually adjusted to control the optical power presented on the sample. Four SF10 prisms were used for dispersion pre-compensation to provide approximately -8000 fs² group velocity dispersion (GVD) and the measured pulsewidths after the objective lens, without the sample, were ~100 fs.

3.2 Animal preparation for the brain imaging

Sprague Dawley rats (250–320 g) were anesthetized with isofluorane and cannulas were inserted in the femoral artery and vein to monitor heart rate, blood pressure, body temperature and blood gases. A tracheotomy was performed and rats were ventilated with a mixture of air and oxygen. A cranial window 3×3 mm² in size was opened in the parietal bone and the dura was removed. The cranial window was filled with 1.5% agarose and sealed with a 150-µm-thick microscope coverslip. During the measurement isofluorane was discontinued and anesthesia was maintained with 50 mg/kg intravenous bolus of alpha-chloralose followed by continuous intravenous infusion at 40 mg/(kg h). All experimental procedures were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care.

Blood plasma was labeled by injection of a green fluorescent dye, Dextran-conjugated fluorescein (FD2000S, Sigma-Aldrich; 500 nM concentration in blood), through the femoral vein. To label the cortical astrocytes we first topically applied the red fluorescent dye Sulforhodamine 101 (Sigma-Aldrich, 100 µM concentration in the extracellular saline) to the exposed cortical surface. The cortical surface was subsequently rinsed with the extracellular saline, filled with the agarose and sealed with the coverslip [74].

3.3 Animal preparation for the imaging of the lymph nodes

Adult C57BL/6 mice received an initial intraperitoneal injection of 100 mg/kg Ketamine and 10 mg/kg body weight of Xylazine for induction, and subsequent intramuscular injections for maintenance of surgical anesthesia. The right popliteal lymph node was prepared for intravital microscopy as described in [75]. Briefly, the mice were placed in a prone position on a custom-made microscope stage and the popliteal lymph node was microsurgically exposed through a 1-cm skin incision in the popliteal fossa of the right hind leg and blunt dissection of the overlying connective and fatty tissue, taking care not to damage blood or lymph vessels. The lymph node was immersed in normal saline and covered with a 150-µm-thick coverglass. A circular metal tube was sealed to the top of the coverslip, and the resulting pool was filled with water for immersion of the objective lens. The metal tube was perfused with warm water to maintain the temperature of the lymph node at ~37°C; the temperature was monitored through a small thermocouple placed in the direct vicinity of the lymph node.

To visualize blood vessels, we intravenously injected Dextran-conjugated fluorescein (FD2000S, Sigma-Aldrich) through the retroorbital plexus to achieve a calculated 500 nM plasma concentration. To stain lymph node cells we intravenously injected 20 mg/kg body weight of the red-fluorescent intravital dye Rhodamine 6G, which rapidly diffuses into cells and tissues and accumulates in mitochondria [76]. The lymph node collagen network was visualized using SHG imaging [75]. To visualize the conduits accompanying the collagen reticulum in the draining lymph node, 1 µg of the small molecular weight protein Lysozyme was conjugated to Alexa Fluor 633 in 10 µl Hank’s buffered salt solution (HBSS) and injected subcutaneously into the right footpad [77]. Time-lapse movies were generated by recording stacks of 5 optical sections, spaced 2.5 µm apart, at 15-second intervals. SHG signals were collected using a 400/40 bandpass filter (Chroma Technology Corp.).

3.4 Preparation of the microscope slides

To assess the utility of our new system for histology we prepared the following specimens: 1. Wild type GFP-expressing 9L gliosarcoma tumors [78] were grown in nude mice (nu/nu, Massachusetts General Hospital Radiation Oncology breeding facilities). After two weeks, mice were sacrificed and tumors were excised, snap frozen in liquid nitrogen, cut into 7-µm sections and mounted on histological slides using VECTASHIELD Mounting Medium with the fluorescent dye DAPI (Vector Laboratories, Burlingame, CA). 2. Samples of the pancreas were taken from BALB/c mice, snap frozen and cut into 7-µm sections. Staining for insulin was performed using rabbit anti-human insulin polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA) followed by FITC-labeled secondary anti-rabbit IgG (H+L) (Vector). Slides were mounted using VECTASHIELD Mounting Medium with DAPI.

4. Results

4.1 Imaging of the cortical vasculature and blood flow

Imaging of highly scattering brain tissue is an important application area of MPM [3, 79–81]. The signal intensity generated within the focal volume decays exponentially with imaging depth z and is approximately proportional to exp(-2z/l_s), where l_s is the optical scattering mean free path [6, 81–83]. The maximum imaging depth in optically turbid media is a complex function of the optical properties of the sample, the excitation light, and the detection system, and is ultimately limited by fluorescence generated near the sample surface [83]. By reducing the pulse repetition rate while keeping the average power constant, it is possible to penetrate up to 1 mm into the neocortex [84, 85]. However, increases in background fluorescence are usually significant at depths beyond 600 µm and resolution may be severely degraded due to spherical aberrations, unless adaptive optics compensation methods are used [86–89]. Due to technical difficulties related to the availability of optical excitation sources, increased background fluorescence and loss of resolution, most brain imaging applications of MPM today are limited to a depth of ~600 µm [3, 29, 90]. To achieve this imaging depth, typically 180–250 mW of the optical power was presented on the sample surface [81, 82, 91, 92].

Figure 3(a) shows a maximum intensity projection (MIP) along the z axis of a 250-µm-thick 3D stack of the cortical vasculature labeled with Dextran-conjugated fluorescein (FITC). The imaging was performed using an Olympus UPlanFL 10X objective (NA=0.3). The 3D stack contains 26 frames acquired with a Δz=10 µm step over a large, 932×932-µm² field of view (FOV). Figure 3(b) represents the MIP along the z axis of a 400-µm-thick stack of the FITC-labeled cortical vasculature. The position of the stack was marked with the window in Fig. 3(a) and contains 400 frames acquired with a Δz=1 µm step. Imaging was performed using an Olympus LUMPlanFI/IR 40X WI objective. Each frame was averaged four times to improve signal-to-noise ratio (SNR) and to smooth the appearance of gaps due to non-labeled red blood cells. In order to compensate for the loss of SNR with the imaging depth, the excitation light power incident on the sample and the dwell time per pixel were manually adjusted every Δz=50 µm. For imaging deeper than 200 µm, we used the maximum available power of 35 mW. A 2 µs dwell time per pixel was used for imaging first 100 µm and up to 4 µs for imaging depths below 300 µm. For depths of 300–350 µm and 350–400 µm, the dwell time was increased to 10-µs and 20-µs per pixel, respectively. The relatively small power levels required for imaging at the larger depths, compared to previously reported values [81, 82, 91, 92], could be attributed to the 2–3 times shorter pulse widths used here as well as increased averaging, and possible differences in the design of the detection system. In addition, imaging depth might be reduced by increased absorption from the close proximity of the large vessels, the increase in the scattering coefficient of the brain tissue in older animals, or blood and dye leakage from the vasculature.

The volumetric image of the whole 3D stack, spanning 233×233×400 µm³, was created using a software package, Imaris (Bitplane AG), and is shown in Fig. 3(c). Figures 3(d) and 3(e) show the MIPs of 50-µm-thick regions at depths of 200 µm and 350 µm, respectively.

One important application of 2PM in brain imaging is the measurement of red blood cell (RBC) velocity [81]. The arrow in Fig. 4(a) shows the location of the velocity measurement 120 µm below the sample surface. We performed 1000 line scans along the vessel axis [Fig. 4(b)] with a line scan speed of 16.08 µm/ms, a line scan period of 1.49 ms, dwell time per pixel of 4 µs, and average power of 20 mW. Imaging was performed using the Olympus LUMPlanFI/IR 40X WI objective. The dark bands in Fig. 4(b) were produced by the moving red blood cells. Blood cells appear as the dark spots in the two-photon microscopy images because they are not labeled with the dye. From the slope of the dark bands in Fig. 4(b), the RBC velocity is estimated as 0.6 mm/s.

Another major application area of MPM is imaging of neuronal activity and organization [3]. Figure 5 shows examples of structural images of astrocytes and cortical vasculature. Astrocytes are the most abundant macroglial cells in the brain with numerous projections that anchor neurons to their blood supply. Their active role in the brain function is currently the subject of intensive research [93]. Figure 5 shows the MIP along the z axis of a 20-µm-thick volumetric image of the neocortex. The blood plasma was labeled with FITC and the astrocytes were labeled with Sulforhodamine 101 (SR101). Imaging was performed with the Olympus LUMPlanFI/IR 40X WI objective. The distance between consecutive frames was Δz=1 µm and the imaging depth was 200 µm. The average excitation optical power on the sample was 20 mW and the pixel dwell time was 4 µs. Each frame was averaged eight times. Numerous processes of astrocytes and their connections with the microvasculature endothelium are clearly visible.

 

Fig. 3. Images of cortical vasculature stained with FITC. (a) MIP along the z direction of a 250-µm-thick 3D stack with a 932×932 µm² FOV. The data set consists of 26 frames acquired at 10-µm increments. (b) MIP along the z direction of a 400-µm-thick 3D stack with a 233×233 µm² FOV. The lateral (x, y) position of the image is marked with the window in (a). The data set consists of 400 frames acquired at 1-µm increments. The same data set is presented as a volumetric image in (c) and two MIPs along the z direction of 50-µm-thick regions are presented at depths of 200 µm and 350 µm in (d) and (e), respectively (Media 1). Scale bar is 100 µm.

Download Full Size | PDF

 

Fig. 4. Measurement of the RBC flow. (a) Two-dimensional image of cortical vasculature labeled with FITC, with the white arrow marking the line scan position. The scale bar is 50 µm. (b) The intensity plot of 1000 line scans taken along the axis of the blood vessel during 1.49 s. The length of the individual line scan is 18.69 µm. The estimated RBC velocity is 0.6 mm/s.

Download Full Size | PDF

 

Fig. 5. MIP of a 20-µm-thick 3D stack image of the neocortex with astrocytes labeled with SR101 (red) and cortical vasculature stained with FITC (green). The data set has a FOV of 233×233 µm² and it consists of 20 frames acquired at 1-µm increments (Media 2). The imaging depth is 200 µm below the cortical surface. Scale bar is 100 µm.

Download Full Size | PDF

4.2 Imaging of lymph nodes

In situ and intravital imaging has become a burgeoning field of immunology and the dynamic behavior of immune cells in lymph nodes has been a primary focus of initial studies [94]. The utility of the Cr:LiCAF laser for MPM intravital microscopy was demonstrated using a representative example. Popliteal lymph nodes were exposed microsurgically and static and time-lapse high-resolution recordings were obtained using an Olympus LUMPFL60X/W IR objective with NA=1.1 (Fig. 6). An incident power of 20 mW was sufficient to obtain an excellent signal-to-noise ratio for cells and structural tissue elements at depths of more than 100 µm in the tissue. The dwell time per pixel was 3.2 µs.

4.3 Imaging of histological specimens

Figure 7 presents representative MPM images of histological specimens described in Section 3. Figure 7(a) shows GFP-expressing 9L fliosarcoma tumor with readily detectable green cytoplasmic staining depicting green fluorescent protein (GFP). Insulin-producing beta cells in pancreatic islets are clearly recognizable by characteristic punctuate FITC staining of insulin granules [green color, Fig. 7(b)]. In both images, blue DAPI staining depicts cell nuclei. Figs. 7(a) and 7(b) were obtained using Olympus LUMPFL60X/W IR and LUMPlanFI/IR 40X WI objectives, respectively, with 4-µs pixel dwell times.

 

Fig. 6. Intravital micrographs of a mouse popliteal lymph node. The image (a) was obtained after intravital staining of mitochondria in lymph node cells though i.v.-injection of Rhodamine 6G (red) and labeling of the blood plasma with FITC (green). SHG signals from reticular collagen fibers are shown in blue (Media 3). Media 4 represents MIPs of five optical sections spaced 2.5 µm apart. This time-lapse recording highlights the cellular dynamics within the lymph node in relation to the structural components, such as collagen fibers and blood vessels. The image (b) shows the lymph node conduit system filled through subcutaneous injection of the low-molecular-weight protein lysozyme tagged with Alexa Fluor 633 (red) into the footpad drained by this lymph node (Media 5). The conduits accompany the collagen network (blue). Scale bars are 50 µm. The imaging depths are 100×150 µm (a) and 30×75 µm (b).

Download Full Size | PDF

 

Fig. 7. Imaging of the histological specimen. (a) GFP-expressing 9L fliosarcoma tumor. Green color depicts cell cytoplasm stained with GFP. (b) Pancreatic islets with insulin-producing beta cells. Green color represents FITC staining of insulin granules. In both images cell nuclei were stained with DAPI (blue color). Scale bars are 50 µm.

Download Full Size | PDF

5. Summary and discussion

These results demonstrate that a femtosecond Cr:LiCAF laser can be successfully used as a light source for multi-photon microscopy. Examples of two-photon and second harmonic generation imaging from several research areas such as neurology, immunology, cancer and diabetes research are presented. The images obtained are of equal quality to previously published MPM data [2–4, 93]. The average laser power was 180 mW, which should be appropriate for the majority of MPM applications. In our setup, as much as 35 mW average power was available on the sample. This is generally a safe power level, even relatively close to the sample surface when microscope objectives with NA <1 and ~1-µs dwell times per pixel are used [21]. No signs of thermal damage, photobleaching or phototoxicity were evident during experiments. The available power is sufficient for MPM imaging up to several hundreds of micrometers deep into highly scattering brain tissue. However, with additional effort, transmission through the microscope optics could be improved more than twofold. For applications requiring even higher average power levels, it is possible to use Cr:Colquiriite lasers pumped by multimode diodes. A femtosecond Cr:LiCAF laser pumped by broad-stripe single-emitter diodes [67, 95], has the potential to generate average mode locked powers approaching 1 W level. The drawback of a multimode diode pumped system would be increased complexity and price compared to single-mode diode pumping.

In our imaging experiments, the central wavelength of the Cr:LiCAF laser was fixed at around ~800 nm, which is near the gain peak of Cr:LiCAF (~780 nm). However, the emission bandwidth of the Cr:LiCAF medium is broad enough to provide wavelength tuning from ~750 nm to ~850 nm. In the current setup, tunability would be limited because a standard SESAM/SBR with only ~50–60 nm bandwidth was used for mode locking [67, 68]. The tuning range could be extended by using multiple SESAM/SBRs, at the expense of interchanging optics. Broadband oxidized saturable absorber mirrors [96, 97], with ~300 nm bandwidth have been developed which would enable the Cr:LiCAF laser to operate across its entire tuning range. Oxidized SESAM/SBR technology can be also applied to generate ~10–20 fs pulse durations from Cr:LiCAF laser, which, with proper dispersion compensation, would be beneficial in some MPM experiments. However, oxidized SESAM/SBRs currently have a relatively high level of passive losses (~2%), which must be decreased below the 0.5% level in order to apply this technology for low gain Cr:LiCAF media. Another alternative approach to obtain broader tunability (or shorter pulses) is to use Kerr lens mode locking (KLM). However, it is challenging to build a turn key KLM mode locked Cr:LiCAF laser system, since the third-order nonlinearity of Cr:LiCAF gain media is quite low (~8 times lower than Ti:Sapphire).

Although the more expensive Ti:Sapphire femtosecond oscillator, or frequency doubling from optical parametric oscillators (OPO), offers a broader range of wavelengths, the tuning range of the Cr:LiCAF could be sufficient for excitation of most available dyes. Based on the examples provided in this work, and from the available 2PE cross sections in the literature [8, 21, 33–41], the major classes of fluorescent dyes should be efficiently excitable with the Cr:LiCAF laser, including the following: fluorescein-, rhodamine-, and coumarine-based fluorophores; calcium indicators; quantum dots; cyanines (Cy2, Cy5); some fluorescent proteins (wild type GFP, eCFP, eGFP); intrinsic tissue fluorophores (folic acid, riboflavin, retinol); and others. Depending on the dye concentration and the tissue environment, it is also possible to have efficient excitation of the fluorophore even at wavelengths that are further from its 2PE cross-section maximum (see, for example, DAPI fluorescence in Fig. 7). To reach wavelengths above 850 nm, it is possible to use Cr:LiSAF, another Cr:Colquirite crystal, which has a tuning range between ~800 nm and ~950 nm (Cr:LiSAF has a gain peak around 850 nm). Since these lasers are low cost and they can be made compact, it may be possible to build two such lasers in parallel that would cover the entire spectral range from ~750 nm to ~950 nm. Lastly, we note that even though Cr:LiSAF has broader tunability and higher gain than Cr:LiCAF, the tuning range of Cr:LiCAF enables the generation of shorter wavelengths which may be important to access the TPE spectra of many commonly used fluorophores.

In summary, this study demonstrates that the performance of diode pumped femtosecond Cr:LiCAF lasers are sufficient to enable a wide range of MPM imaging applications which do not require broad excitation wavelength tunability. The use of low cost, single mode pump diodes for Cr:LiCAF lasers enables a significant reduction in cost and complexity compared to femtosecond Ti:Sapphire technology. Multi-photon microscopy is one of the most powerful optical imaging modalities in biomedical research. The development of lower cost femtosecond laser technology promises to be an important step in enabling wider spread use of multi-photon microscopy, accelerating fundamental research in biology and medicine.