A key element in a nonlinear microscope is the use of an ultrafast laser. These are natural sources that are able to produce the required high intensities needed for exciting nonlinear processes. Historically, Ti:sapphire sources have been used in NLM due to its available large peak powers along with its large tunability range. However, its complexity, high price and maintenance requirements, have limited the widespread adoption of these powerful imaging techniques into daily routine biomedical applications.

Efforts in the past have explored developing compact, lower-cost, and easy to use ultrafast semiconductor saturable absorber mirrors (SESAM) for modelocking diode-pumped solid-state lasers [3,4]. Compact designs can be realized by “folding” long cavities using careful mechanical design, or by increasing the repetition rate of the lasers which naturally allows short cavity lengths. However these sources have been limited by the available peak powers. More recently, compact ultrafast Cr-doped laser systems such as Cr:LiCAF, Cr:LiSAF, Cr:LiSGAF, lasers [4–6] (see Ref [4], Table. 2.1.2 for an overview on these sources) have been produced. Some of these have been employed for nonlinear TPEF imaging [7–11] and although average powers of up to 500mW have been demonstrated [6], they are often limited in their ability to sufficiently scale their average power.

Other alternative sources based on fiber lasers [12–14] and semiconductor laser diodes with amplification schemes [15–17] have also been successfully demonstrated as compact lasers for NLM applications. Fiber lasers can generate very short pulses via passive mode-locking, however, they are limited to wavelengths around 1030 nm and 1550 nm. In semiconductor lasers with amplification schemes, i.e., gain-switched laser source based on vertical cavity surface emitting lasers (VCSELs) [16], gain-switched InGaAsP Distributed-Feedback-Bragg (DFB), laser diode [17] and an external cavity mode-locked laser diode consisting of multiple quantum wells (AlGaAs) [15], the simplicity and compactness of these systems is compromised, as they need several stages to compress and/or amplify the pulses.

One key aspect for optimizing a compact laser for TPEF or SHG microcopy is the critical trade-off between repetition rate of the laser and the multiphoton signal strength generated [18, 19]. The signal strength in TPEF or SHG imaging scales as the product of the peak power times the average power (assuming image spot size, absorption, sample, detection path, etc. remain constant). This two-photon figure of merit (FOM_2p) allows one to make relative comparisons of laser sources as a function of their average power, pulse duration, and repetition rate (since peak power equals average power divided by repetition rate divided by pulse dlocking howeveruration). Typical Ti:sapphire lasers on a TPEF sample give a peak-power-average-power product (FOM_2p) of approximately 1 W² (e.g. 200 fs pulse duration, 80 MHz repetition rate, and 4 mW average power at the sample). Alternative laser sources should have a FOM_2p value of a similar order of magnitude. Higher repetition rate lasers allow for more compact designs, but require higher average power to achieve the same FOM_2p value. Repetition rates in the multi-Gigahertz range have been demonstrated in TPEF imaging, but ultimately require average powers that begin to impact sample viability.

Here we experimentally show, for the first time, that an ultrashort pulse semiconductor disk laser (SDL) or vertical extended cavity surface emitting laser (VECSEL) that is modelocked by a “quantum dot-engineered” SESAM [20] can be in fact used for nonlinear microscopy. In fact, SDLs provide a compelling source for TPEF and SHG, as they combine key features such as excellent beam quality, output power, short pulse durations, amplitude stability, and can be made to operate at a large set of wavelengths, well-matched to key TPEF dyes, while maintaining simplicity and ease of operation. In addition, the repetition rate of these lasers can be adjusted into the 100’s of MHz to 1 GHz, resulting in a compact laser cavity, but still working in a range where the FOM_2p can be reasonable (compared to the standard Ti:sapphire laser) without requiring excessively large average powers. SDL’s also have the potential for low-cost manufacturing, as their gain element is based on wafer technology and because they have a relatively simple laser cavity design requiring only relatively low-cost and low-brightness pump diodes. Thus, in this work, we demonstrate the use of a modelocked VECSEL [20], with a footprint of only 140x240x70 mm, for in vivo multiphoton microscopy. The VECSEL gain chip is mode-locked with a quantum-dot SESAM [21–23] to produce 1.5 ps pulses at 500 MHz with an output average power of 287 mW at 965 nm. Importantly the laser output wavelength brings the advantage that TPEF of the Green Fluorescent Protein (GFP), one of the most widely used fluorescent markers for biological applications, has its two-photon action cross section maxima around this operating wavelength [2, 24–27]. Exciting the GFP at this wavelength substantially relaxes the required FOM_2p and corresponding average and peak power values needed for TPEF-based imaging. This is demonstrated in both, fixed samples and in vivo by imaging prepared slides containing different dyes and Caenorhabditis elegans (C. elegans) nematodes expressing GFP in a specific set of motoneurons respectively. In addition, the extended versatility of the laser is shown by presenting SHG images of pharynx, uterus and body wall muscles, followed by a demonstration of the ability of this source to excite various commercially available dyes.

The beads shown in Fig. 2a containing green fluorescing firefly dye can be excited at 468nm and therefore these are very efficiently imaged with our laser. From image 2b it is possible to see the nuclei of goblet cells labeled with SYTOX green, and the border surrounding these cells arrangement labeled with Alexa Fluor 568 phalloidin. In this sample, SYTOX green possess an absorption peak at 940 nm. This is very close to the operating wavelength of our laser and therefore, cells are visualized very efficiently. For the cell structure labeled with Alexa fluor 568 (filamentous actin in the brush border) the image region is very faintly observed as the dye is more efficiently excited in the 720-840 nm wavelength range. Finally, for the cells labeled with Alexa Fluor 350 (mucus of goblet cells) there is no TPEF signal observed as the excitation wavelength range is restricted to the 720 - 800 nm.

We then proceed to test the laser for in vivo imaging. For this purpose we employed genetically modified C. elegans nematodes expressing GFP in VD-Type motoneurons and pharyngeal nerve ring neurons. Furthermore, to emphasize the imaging capabilities of the SDL all the TPEF images where simultaneously recorded with any SHG signal that could be originated from the nematode. Sample imaging was performed employing an average power of 34 mW (measured at the sample plane) which corresponds to a peak power of ~0.04 KW and a FOM_2p = 1.5 W². Employing this configuration, no damage at the sample was observed. In our setup, the peak power threshold for getting a TPEF image from the GFP expressed in the nematode was ~0.01KW, corresponding to a FOM_2p of approximately 0.1 W².

The first region to be imaged was the head of an adult C. elegans nematode (see Fig. 3 ). Figure 3a (TPEF signal depicted in blue), shows the 3D projection of the GFP tagged neurons in the C. elegans pharyngeal region. Image 3b (SHG signal depicted in orange) shows the anterior bulb of the pharyngeal region. In this figure, some parts of the neurons forming the nerve ring are also observed even though a narrow band SHG filter was used (see materials and methods). This could be possibly due to fluorescence leaking through the SHG filter or SHG emission from the neuron [30]. Mapping the TPEF and SHG signals (Fig. 3c) enables the observation of how the neurons forming the nerve ring innervate with the isthmus of the pharynx.

To further demonstrate the capabilities of this laser we then proceeded to image an adult C. elegans nematode near the vulva (see Fig. 4 ). Figure 4a shows a set of VD-type motoneurons expressing GFP. On the upper part of image 4a, the ventral nerve cord and motoneurons cell bodies (near the vulva) can be clearly observed (bright spots at the top end of the image). Additionally, in the same image on the lower and upper parts, a weak epi-detected SHG signal from body wall and vulval muscles is observed (see green arrows Fig. 4a).

In panel 4b, the vulval muscles (C. elegans diagonally positioned) can be observed together with a motoneuron cell body (bright spot). This spot, similarly to Fig. 3b, could be due to autofluorescence leaking through the filter or SHG emission by the same motoneuron observed in Fig. 4a. The superposition of both signals enables the observation of the vm-type 1 and 2 vulval muscles and a portion of the body wall muscles in close relation with the motoneurons that are responsible for locomotion (blue arrow).

Based on both imaged regions, we observe a very efficient TPEF signal despite the fact that we are using a very low peak power. The main reason for this is that the employed wavelength is located around GFP maximum of its two-photon action cross-section [2, 27–27]. Furthermore, the relaxation of the required intensities to excite GFP, would, in principle, maximize sample viability given that the ratio of the energy absorbed to the input energy flux determines the possible induced sample damage.

Finally, the use of this laser for TPEF imaging was assessed with other markers. To do so, different dyes (available in our lab) in solutions were excited with a constant laser output power of ~10 mW (measured at the sample plane). Then the relative two-photon signal intensities were recorded using the PMT voltage value. For each image the background signal was first set to zero and then the emitted signal was adjusted so that only a few pixels were saturated (with an intensity of 255). In each case, low TPEF signal resulted in high PMT voltages and vice versa [31]. Figure 5 shows images of the edge of drops of different dyes that were placed on top of a cover glass. In each image, the voltages resulting from the excitation with the SDL laser system were recorded. It is important to mention that, the precise detector setting may vary not only between instruments but also between samples. However the excitation/emission general trend should not change, meaning that the PMT values shown here are not absolute but they are relative guides to help in understanding the imaging capabilities of our laser.

From above we can observe that the laser can be used for exciting different biological markers. To further illustrate its applicability for these purposes, we have theoretically estimated how efficiently the laser would excite other dyes. Our calculations were based on the two-photon action cross-section of other dyes relative to GFP at 965 nm (~180 GM) for different markers (data taken from ref. 27). By performing such calculation it is possible to determine that the expected TPEF signal compared to GFP are 0.05 for Bodipy, 0.20 for Alexa 488, 0.47 for dsRed. 1.00 for mVenus, 1.15 for YFP, and 1.54 for mCitrine

For instance, the compact design of this laser, having a footprint of only 140x240x70 mm combines a simple and maintenance free operation, as no alignment, cleaning or specialized knowledge is required to operate this laser. Its novel mode-locking mechanism, based on a quantum-dot SESAM, enables to have a self-starting pulsed output immediately after the laser is switched on.

The laser’s operating wavelength centered at 965 nm, offers several benefits. Besides inheriting all the advantages of NLM, this laser enables a very efficient TPEF signal excitation as the GFP two-photon action cross section maxima falls around this operating wavelength. This results in extremely efficient generation of images with a very small light dose (~0.04KW peak power and 34 mW average power at the sample) which should be important for preserving the sample viability. This is demonstrated by presenting in-vivo TPEF images of C. elegans nematodes expressing GFP in a set of motoneurons. As this laser is used for nonlinear imaging, the enhanced signal to noise collection should enable larger penetration depths, reduced production of auto fluorescence from endogenous molecules and slower photo-bleaching rates compared to CLSM.