1. Introduction

The cells in the brain rely on circulating blood for continuous oxygen and nutrient supply. Blood flow speed is a critical hemodynamic parameter which can reflect neuronal activity and certain brain disease. Among the various methods for measuring blood flow speed in the brain, multiphoton microscopy (MPM) is especially advantageous in that it combines subcellular spatial resolution, fast imaging speed and simultaneous tracking of both hemodynamic and neuronal activity [1,2].

MPM includes many imaging modalities such as multiphoton fluorescence microscopy and harmonic generation microscopy. So far the most widely used MPM modality for measuring blood flow speed is 2-photon fluorescence microscopy. The standard protocol is [1,2]: first fluorescent dye is injected into animals, delineating blood vessels. Then repetitive line scan is performed along the blood vessel over time, yielding alternating bright and dark stripes. Finally the blood flow speed can be derived by calculating the slope of the stripes. In order not to incur the exogenous fluorescent label which might be toxic to the animals, it is desirable to develop a label-free technique for measuring blood flow speed through MPM.

Recently, two groups demonstrated that label-free third-harmonic generation (THG) microscopy can be used for measuring blood flow speed in animals in vivo [3,4]. Their results show that the blood vessel walls and red blood cells flowing in the blood vessels yield THG contrast, facilitating both structural and hemodynamic imaging. Most pertinent to neuroscience [4], is the first demonstration of THG imaging for measuring blood flow speed in the animal brain in vivo, reaching an imaging depth of at least 900 µm below the surface of the brain judging from the 2D image. This experiment utilizes excitation at the 1300-nm window, exploiting both the deep penetration capability and resonant enhancement property of red blood cells excited at this window.

Despite its successful applications, there remain several unanswered questions and challenges for THG microscopy for measuring the blood flow speed: (1) blood vessel walls were claimed to be imaged through THG imaging, however, without justification. (2) Although cortical blood vessels could be imaged and measured, it is not clear if blood vessels in subcortical regions, such as the white matter, could be imaged and measured as well. (3) For their demonstration in the mouse brain in vivo, a complicated laser system at 1300-nm (optical parametric amplifier) was used [4]. It is not clear if excitation at the more promising deep-imaging window, the 1700-nm window (covering roughly from 1600 nm to 1840 nm) [5,6], with a simpler laser system could be used, even though there is no resonant enhancement of THG at this wavelength. (4) What is the difference and relationship between the multiphoton fluorescence and THG measurement techniques?

Here we demonstrate that THG microscopy exited at the 1700-nm window can be used to visualize and measure blood flow speed in the mouse brain in vivo down to the white matter layer. Combining selective labeling technique, we show that blood vessel walls indeed generate THG signals. Comparison between 3-photon fluorescence (3PF) and THG imaging further illustrates that the two signals are complementary to each other in line scan images for blood flow speed measurement.

2. Methods

The experimental setup is shown in Fig. 1(a). In order to generate energetic femtosecond pulses at the 1700-nm window, 1-MHz 500-fs pump pulses at 1550-nm from a fiber laser (FLCPA-02CSZU, Calmar) were coupled into a 44-cm PC rod (SC-1500/100-Si-ROD, NKT Photonics) with a 100-µm core diameter. Soliton self-frequency shift (SSFS) [7,8] shifted the soliton to 1665 nm (peak wavelength). A 1635-nm long-pass filter (1635lpf, Omega Optical) was used to remove the residual. Slight angle tuning was employed to shift the cut-on wavelength of the long-pass filter to ∼1610 nm, in order not to block the soliton spectrum. The soliton spectrum was measured with an optical spectrum analyzer (OSA203B, Thorlabs). The measured soliton spectrum is shown in Fig. 1(b). Using a home-built second-order interferometric autocorrelator, the soliton pulse width was measured to be 86 fs after deconvolution [Fig. 1(c)]. Briefly, the interferometric autocorrelator was a Michelson interferometer, consisting of a 50/50 beam splitter cube, and two silver-coated mirrors, one of which was mounted on a motorized stage to scan the optical delay. The 2-photon silicon photodetector was placed after a focusing lens for 2-photon current measurement and retrieval of the interferometric autocorrelation trace, with an estimated error of ±3.3 fs. The measured soliton energy was 120 nJ, suitable for deep-brain 3PM (3-photon microscopy). In comparison with an OPA (optical parametric amplifier) system operating at the 1300-nm window, our laser system is much simpler and less costly, while delivering excitation pulses at the deeper-penetrating 1700-nm window.

 

Fig. 1. (a) Experimental setup. M: silver-coated mirror, L1∼L4: lens, LPF: long-pass filter, DM: dichroic mirror, BPF1: band-pass filter (593lp and 630/92), BPF2: band-pass filter (558/20), PMT: photomultiplier tube. (b) Measured soliton spectrum. (c) Measured soliton interferometric autocorrelation trace. (d) Measured 3-photon excitation spectrum (characterized by wavelength-dependent 3-photon action cross section ησ₃) covering the 1700-nm window for Alexa Fluor 633.

Download Full Size | PDF

The soliton pulses were then sent into a multiphoton microscope (MOM, Sutter) for 3PM, similar to that in [9]. The objective lens was a customized one with high transmittance at the 1700-nm window for 3PM (XLPLN25XWMP2-SP1700, Olympus). The lateral and axial 3-photon point spread functions (PSFs) were 0.76 µm and 2.5 µm, respectively, measured by 3 dimensional imaging of 200-nm diameter fluorescent beads. Simultaneous dual-channel detection of THG and 3-photon fluorescence signals from fluorescent dyes were performed. A GaAs photomultiplier tube, i.e., PMT (H7422p-50, Hamamatsu) with a 558/20 band-pass filter (FF01-558/20-25, Semrock, OD>4) was used to detect THG signals, while a GaAsP PMT (H7422p-40, Hamamatsu) with a 593-nm long-pass filter (FF01-593/LP-25, Semrock, OD>4) and a 630/92 band-pass filter (FF01-630/92-25, Semrock, OD>4) was used to detect 3PF signals from exogenous dyes.

Animal procedures were reviewed and approved by Shenzhen University. We used adult Balb/C mice (9∼10 weeks old, Guangdong Medical Laboratory Animal Center) for imaging. Mice were anesthetized using a gaseous anesthesia system (Matrx VIP 3000, Midmark). Their body temperature was hold at 36.5 °C using a heating blanket. Craniotomies were performed centered at 2 mm posterior and 2 mm lateral to the bregma point and sealed with 0.16-mm-thick cover glass (BT10162, Best).

3. Results and discussion

First we verify that blood vessel walls yield THG signals. In both our [ Fig. 2(a)] and previous THG imaging results [3], in 2 dimensional (2D) THG images, bright lines that run parallel to the blood vessels can be clearly visualized and were naturally assumed to be blood vessel walls. In order to verify this, we used fluorescent labeling technique developed in [10]: Alexa Fluor 633 (A30634, Thermofisher) specifically labels arteriole (diameter≥15 µm) walls in animal brains. Experimentally we also verified that 3-photon fluorescence from Alexa Fluor 633 could be excited at the 1700-nm window and measured the 3-photon excitation spectrum [Fig. 1(d)], using our previously developed technique detailed in [11]. Although 3-photon fluorescence can be most efficiently at 1620 nm within the 1700-nm window [corresponding to the peak in Fig. 1(d)], we chose excitation at 1665 nm because: (1) our main aim is to perform THG imaging rather than 3-photon fluorescence imaging; (2) soliton energy is higher at 1665 nm compared with 1620 nm. Excitation at 1665 nm clearly reveals the blood vessel walls labeled by Alexa Fluor 633 in 3PF image [Fig. 2(b)]. The merged image [Fig. 2(c)] of both THG and 3PF images show spatial overlap of the imaged blood vessel walls. We also plotted line profiles of the merged image perpendicular to the blood vessel walls [Figs. 2(d), 2(e)], from which we can clearly see colocalization of THG and 3PF signals in blood vessel walls. Combining label-free THG imaging and structure-specific 3PF imaging, we thus verify that blood vessel walls are indeed visualized by THG imaging. We also note that the cell-free layer (dark lines between the blood vessel walls and circulating blood) can also be visualized in THG imaging excited at the 1700-nm window, similar to those observed with other excitation wavelengths [3].

 

Fig. 2. 3PM in the mouse brain in vivo. (a) THG image of blood vessels and circulating red blood cells. (b) 3PF image of blood vessel walls labeled by Alexa Fluor 633. (c) Merged image of (a) and (b). (d, e) Normalized line profiles plotted along the lines perpendicular to the blood vessel walls indicated in (c), showing colocalization of THG signals (green) and 3PF signals (red). The frame rate is 4 s/frame. Scale bars: 50 µm.

Download Full Size | PDF

Next we reveal the difference and relationship between THG and 3PF imaging in measuring blood flow speed. Experimentally we used low-cost dye SR101 (sulforhodamine 101, Sigma-Aldrich) [11] to label the plasma (while not the red blood cells) for 3PF imaging. Figure 3(a) shows representative merged image of a segment of labeled blood vessel 500 µm below the brain surface with both THG (green) and 3PF (red) imaging. We performed simultaneous THG and 3PF line scan along the center of the blood vessel (indicated by the dashed line). The corresponding merged line scan image is shown in Fig. 3(b), yielding the typical bright and dark stripes. Inside blood vessels, THG signal originates from red blood cells (not labeled by SR101), while 3PF signal originates from plasma (labeled by SR101) and not red blood cells. It can be expected that the line scan results will be complementary to each other in THG and 3PF. This is exactly what we have observed: by plotting line profiles along the time axis, the THG and 3PF signals are complementary to each other [Figs. 3(c)–3(e)]. Physically, this means that at each time point, a spatial space position cannot be occupied by red blood cells and plasma simultaneously, so the two are mutually exclusive in space. We used a method similar to [12] to calculate the blood flow speed, given by mean ± standard error of the mean. From the stripes, we calculated a blood flow speed of 0.80 ± 0.07 mm/s.

 

Fig. 3. (a) Simultaneously acquired THG (green) and 3PF (red) image of a segment of blood vessel 500 µm below the surface of the mouse brain in vivo. (b) Line scan image along the center of the blood vessel indicated by the dashed line in (a). (c, d, e) Normalized line profiles along the lines indicated in (b), showing time-lapse THG (green) and 3PF (red) signals at 3 spatial positions. The frame rate is 1 s/frame in (a), and the line rate is 2 ms/line in (b).

Download Full Size | PDF

The deep penetration capability of 3PM at the 1700-nm window is illustrated by the reconstructed 3D stack of THG imaging of the entire mouse neocortex and the subcortical white matter layer [ Fig. 4(a)]. In our experiment, the white matter layer spans from 880 µm to 1000 µm below the surface of the brain. In the brain, besides blood vessels and circulating blood cells, THG signals also arise from myelinated axons [13] and the neuropil surrounding cells [14]. The white matter layer generates strong THG signals [Figs. 4(a), 4(b)], since it is a densely packed layer of myelinated axons. In spite of this, the blood vessels in this layer can still be visualized by THG imaging alone [area delineated by dotted lines in Fig. 4(c)], by its streamlined appearance due to the flowing red blood cells. Simultaneous 3PF imaging [Fig. 4(d)] of SR101 in the circulating plasma shows that both 3PF and THG signals originate from the same segment of blood vessel. Using THG imaging excited at the 1700-nm window, we thus imaged blood vessels down to the white matter layer.

 

Fig. 4. (a) 3D THG image stack of the mouse brain in vivo. The white matter layer (from 880 µm to 1000 µm below the surface of the brain) generates strong THG signals. (b) 2D THG image 908 µm below the surface of the brain reveals myelinated axons, and a segment of blood vessel in the zoomed-in image (area delineated by dotted lines) in (c). (d) Simultaneous 3PF imaging from SR101 (red) and THG imaging (green) verifies the observed blood vessel in (c). Scale bars: 25 µm in (a-d). (e) THG line scan image along the center in (c) for blood flow speed measurement. The frame rate is 2 s/frame in (b), (c) and (d), and the line rate is 2 ms/line in (e).

Download Full Size | PDF

We performed line scan along the center of the blood vessel in the white matter in Fig. 4(c) for blood flow speed measurement. The line scan image [Fig. 4(e)] shows clear bright and dark stripes, facilitating speed calculation. Using this technique, the blood flow speed for the imaged vessel 908 µm below the surface of the brain and in the white matter is 1.50 ± 0.11 mm/s. Although here we only show THG imaging of blood vessels in one mouse, experimentally we have imaged blood vessels and measured the blood flow speed in more mice using THG imaging, summarized in Table 1. In all these four mice we could image blood vessels using THG imaging down to the white matter layer. However, limited by the requirement that the blood vessels have to be parallel lying, within the white matter layer we could only measure blood flow speed in three mice as shown in Table 1. Our measured results below 900 µm are similar to the measured value (1.4 mm/s) at a depth of 900 µm reported in [15].

Table 1. Blood flow speed measured using THG imaging in four mice.

View Table

4. Conclusion

3PM excited at the 1700-nm window has shown its great potential for structural deep-brain imaging. Here we demonstrate that THG imaging excited at this window is a promising technique for extracting the hemodynamic information, specifically the blood flow speed, through the cortical grey matter layer and down to the subcortical white matter layer. In combination with selective labeling technique and 3PF imaging, we further verify: (1) blood vessel walls can generate THG signals; (2) THG signals and 3PF signals are complementary in the line scan images in the blood vessel. Currently, our THG line scan image is limited to parallel-lying blood vessels. This limitation is expected to be circumvented by user-defined scanning technique demonstrated in [1].