2.1.

Viral Injection and Brain Slice Preparations

A C1V1-mCherry transducing viral vector [AAV-DJ-CaMKIIa-C1V1-(E122T/E162T)-TS-P2A-mCherry, Stanford Neuroscience Gene Vector and Virus Core] was injected into the dentate gyrus of hippocampi in transgenic Thy1-GCaMP6s mice (Stock # 025776, Jackson Labs) of both sexes, 1 to 3 months old (P30 to P90). Animals were euthanized up to 3 months postinjection for experiments. Animals were anesthetized with 2% to 3% oxygen-vaporized isoflurane (Clipper Distributing Company) and placed in a stereotaxic apparatus (Kopf Instruments). Carprofen ($5\mathrm{mg}/\mathrm{kg}$, Zoetis) was administered subcutaneously prior to surgery. The viral solution was loaded into a $10\text{-}\mu \mathrm{L}$ Nanofil syringe with a 33G needle driven by an injection pump controller (Micro4, World Precision Instruments). A $1\text{-}\mu \mathrm{L}$ volume of the viral solution at a titer of $\sim {10}^{12}\text{genomes}/\mathrm{mL}$ was injected into the hippocampal dentate gyrus region (1.8 mm posterior to bregma, $\pm 1.3\mathrm{mm}$ lateral from bregma, and 1.4 ventral to the cranial surface) at $0.12\mu \mathrm{L}/\min$. The syringe was left stationary 3 to 5 min postinjection to promote diffusion of the viral vector throughout the injection site. The wound was closed with silk sutures (Perma-hand, Ethicon). Finally, 2.5% lidocaine + 2.5% prilocaine cream (Hi-Tech Pharmacal) and Neosporin (Johnson & Johnson) were applied topically to the injection site to promote healing and reduce inflammation.

For slice preparation, animals were anesthetized by 2% to 3% oxygen-vaporized isoflurane, followed by decapitation. The mouse brain was then isolated and quickly transferred to chilled cutting solution (254 mM sucrose, 11 mM glucose, 2.5 mM KCl, 1.25 mM ${\mathrm{NaH}}_2{\mathrm{PO}}_4$, 10 mM ${\mathrm{MgSO}}_4$, 0.5 mM ${\mathrm{CaCl}}_2$, and 26 mM ${\mathrm{NaHCO}}_3$). A vibratome (VT1000S, Leica Biosystems) was then used to cut $300\text{-}\mu \mathrm{m}$-thick slices. Regions near the injection site were saved for experimentation, whereas all other brain slices were discarded. Sliced tissue was incubated in artificial cerebrospinal fluid (126 mM NaCl, 2.5 mM KCl, 10 mM glucose, 1.25 mM ${\mathrm{NaH}}_2{\mathrm{PO}}_4$, 1 mM ${\mathrm{MgSO}}_4$, 2 mM ${\mathrm{CaCl}}_2$, and 26 mM ${\mathrm{NaHCO}}_3$ at $\sim 298\mathrm{mOsm}$) at 32°C for 60 min, then transferred to room temperature for at least 15 min prior to imaging. Slices were continuously perfused with 95% oxygen and 5% ${\mathrm{CO}}_2$ at all times.

2.2.

Calcium Imaging and Single-Cell Photostimulation Experiments

The experimental setup is illustrated in Fig. 1. A large mode area PCF (PM-LMA-15, NKT Photonics) was pumped by a 1030-nm femtosecond fiber laser (Satsuma, Amplitude Systemes) to generate the highly polarized coherent fiber supercontinuum.²¹ Although the pump laser can output an average power of 10 W, we only applied 3.5 W for pumping, which was more than sufficient for our experiments. The repetition rate was also tuned to 20 MHz in these studies. The broadband coherent supercontinuum was directed to a 640 pixel, 4-f pulse shaper (MIIPS Box 640, Biophotonics Solutions Inc.) for further spectral–temporal modulation. The broadband spectrum was separated into two bands (890 to 950 nm and 1010 to 1160 nm) inside the pulse shaper by a knife-edge-based amplitude modulation, which minimized the excitation crosstalk between GCaMP6s and C1V1-mCherry, as shown in the inset of Fig. 1. With the programmable capability of the spatial light modulator inside the pulse shaper, optimized band-specific phases were simultaneously applied to these two spectral bands to achieve individually near transform-limited pulse compressions (890 to 950 nm, $\sim 51\mathrm{fs}$ and 1010 to 1140 nm, $\sim 35\mathrm{fs}$). This approach of temporal modulation is more convenient and versatile compared to the common pulse compressor methods, e.g., prism/grating pairs, which require two separate sets with careful alignment, along with the need to be realigned for different spectral band applications. The relatively broad bandwidth of each spectral band, along with the low pulse repetition rate of 20 MHz, was advantageous for generating higher peak power compared to standard 80 MHz pulses, as reported in our previous work.²¹ This becomes more explicit with the formal relation between peak power $P_{\text{peak}}$, average laser power $P_{\mathrm{avg}}$, pulse duration $t_p$, and laser repetition rate $f_p$ defined as $P_{\text{peak}}=\frac{P_{\mathrm{avg}}}{t_p{f}_p}$. Because higher peak power can reduce the average laser power requirement for 2P excitation and minimize phototoxicity and photobleaching, the low repetition rate strategy has also recently begun to be utilized within the neuroscience community.^6,7,17

Fig. 1

Optical system used for experiments. A 1030-nm, 350-fs, and 20-MHz repetition rate laser is used to pump a PCF for supercontinuum generation. The light is directed to a pulse shaper which compresses the pulses at two bands to maximize 2P efficiency and improve image quality. After the pulse shaper, the light is split with a dichroic mirror to separate paths. The GCaMP6s path is driven by a resonant scanner paired to galvanometers for high-speed imaging. The C1V1-mCherry path is directed to a pair of galvanometers to scan for imaging and for spiral scanning of target neurons. The two paths are recombined at the sample for experiments. Separately, a 561-nm CW laser was directed by a multimode optical fiber to the sample for widefield optogenetic activation.

The beam after the pulse shaper was then split into two paths by a 980-nm long-pass filter (FF980-Di01, Semrock) (DM1) for excitation of GCaMP and C1V1-mCherry, which were recombined (DM2) after the scanners and directed (DM5) to a modified Olympus BX51 microscope. Calcium imaging was performed at video rate using an 8-kHz resonant scanner (SC-30, Electro-Optical Products) for one transverse axis with a standard galvanometric-scanned mirror (6215H, Cambridge Tech) for the second transverse axis. For C1V1-mCherry imaging and excitation, the beam was scanned using a pair of galvanometric-mounted mirrors. Fluorescence from calcium imaging was detected during point-scanning using an analog PMT (H7422A-40, Hamamatsu) then amplified using a PMT Transimpedance amplifier (TIA60, ThorLabs). For C1V1-mCherry imaging, the fluorescence was detected using a photon counting PMT (H7421-40, Hamamatsu). All images covered a $300\mu \mathrm{m}\times 300\mu \mathrm{m}$ field-of-view. A movable dichroic mirror (DM3) was placed to separate the near-infrared excitation beam from the fluorescence for PMT detection (735 nm high pass, Semrock), with an additional dichroic placed to separate the GCaMP and C1V1-mCherry fluorescence (593 nm high pass, Semrock) in front of the PMT. The DM3 mirror can be slid out for wide field fluorescence and DIC imaging with a CCD camera (QImaging Retiga Electro). Filters were implemented for detection of GCaMP6s and C1V1-mCherry fluorescence at $510\pm 42\mathrm{nm}$ and $641\pm 37.5\mathrm{nm}$, respectively (FF01-510/84-25 and F-F02-641/75-25, Semrock). The sample was illuminated with $\sim 18\mathrm{mW}$ of power for excitation and imaging of C1V1-mCherry. About 18 mW of power was used for imaging GCaMP6s in brain tissue. Experiments were also conducted with a CW laser source operating at a 561-nm wavelength (Sapphire, Coherent). A partially under-filled objective (XLPlan N $25\times$ 1.05 NA, Olympus) was used for all experiments. Previous studies have demonstrated that a lower effective numerical aperture, rather counter-intuitively, improves 2P optogenetic activation in brain tissue due to the increased number of excited opsins.^52,53

For optogenetic experiments, the brain tissue was first examined by widefield GCaMP imaging, followed by C1V1-mCherry imaging, to verify the presence of GCaMP6s and determine regions with dense C1V1-mCherry expression. Once a region was established, the optogenetic stimulation protocol (both CW and 2P) was used to evoke responses from cells near transduced areas. A spiral scanning approach was implemented for 2P stimulation, with controlled exposure times, as reported in Table 1. For CW stimulation, widefield illumination across the region of interest was employed, targeting all cells in the field-of-view. Stimulation protocols were initiated after a wait time of at least 1 min after identifying C1V1-mCherry positive regions to promote recovery of the proteins to their resting, ground state. Similarly, cells illuminated with either CW or 2P illumination were given adequate time ($>1\min$) to recover prior to subsequent stimulation. The response was then visualized in real time and the data were recorded for postprocessing. All data were processed offline using custom-written MATLAB software. All hardware for imaging and stimulation protocols were controlled using a custom-written LabVIEW program.

Table 1

Stimulation protocols for Fig. 4.

2.3.

Image Denoising

To reduce the presence of noise inherent in fluorescent microscopy systems, image denoising was implemented prior to image analysis. In particular, the Noise2Void denoising technique⁵⁴ was implemented to remove sources of noise from the captured calcium imaging sequences. Calcium transients were visible without the algorithm, but the image quality was dramatically improved, which allowed for ready identification of sources of these dynamic changes. The algorithm was run in PyCharm. Representative calcium imaging datasets demonstrating the efficacy and reliability of Noise2Void are shown in Fig. S1 in the Supplementary Materials. A representative video demonstrating the effectiveness of denoising is also included in Video S1 [URL: https://doi.org/10.1117/1.NPh.7.4.045007.1]. Additionally, the signal-to-noise ratio is quantified before and after denoising in Fig. S2 in the Supplementary Materials, quantified using the equation, $\mathrm{SNR}=10{\log }_{10}(\frac{{\mu }_{\mathrm{sig}}}{{\mu }_{\mathrm{bckg}}})$, where ${\mu }_{\mathrm{sig}}$ and ${\mu }_{\mathrm{bckg}}$ represent the mean intensity from an image representative of signal and background, respectively.

2.4.

Data Analysis

All analyses were performed offline using custom MATLAB software, using a previously devised algorithm.⁵⁵ Calcium transients were isolated from individual cells by selecting a point on the soma and averaging around a five-pixel crosshair. Raw images were saved in binary format. Videos were processed with a mean projection along time using ImageJ to generate the final images, unless otherwise noted. Stimulation times were validated by measuring the image power throughout the time series.

3.1.

Imaging and Transduction Validation

Initial experiments were designed to ensure that the imaging parameters were optimal for both beam paths from the supercontinuum source. Brain slices from GCaMP6s mice injected with C1V1-mCherry were imaged with and without pulse compression to demonstrate image quality optimization. Results highlighted in Fig. 2 show images of these cells and consequently the validation of the expected emission for the desired spectral bands. In particular, Fig. 2(a) shows fluorescence from hippocampal tissue both before and after pulse compression, where the green channel is GCaMP6s and the red channel is mCherry. The results verify both that adequate detection of sufficient image quality can be obtained with the system and that pulse compression does indeed result in improving the signal strength via enhanced 2P absorption efficiency. The line plots in Fig. 2(b), which correspond to the region highlighted by the blue dashed line in Fig. 2(a), show a cross section of the image that further exemplifies the increased signal obtained by pulse compression. Histograms of the pixel intensities of these images also illustrate the dramatic increase in signal after pulse compression. Figure 2(c) clearly illustrates a larger distribution of higher intensity pixels after pulse compression for both GCaMP6s and C1V1-mCherry. The presence of visible structures in the sample in Fig. 2(a), along with the line plots in Fig. 2(b) and histograms in Fig. 2(c), together highlight the efficacy of the system for the desired wavelengths, and the improved image quality following pulse compression optimization.

Fig. 2

Image pulse compression characterization. (a) Images of hippocampal brain slices imaged with the optical setup show a dramatic increase in image quality after pulse compression. (b) A cross section of the image (green and red) is also plotted with and without pulse compression for both the intended GCaMP6s and C1V1-mCherry channels, respectively. The cross sections were normalized to the peak intensity for the compressed images. (c) The histograms also quantify the considerable increase for both the GCaMP (top) and mCherry (bottom) images, before and after pulse compression. Notably, the distribution of intensities from the samples are much broader and illustrate a clear increase in signal intensity overall. Scale bars represent $50\mu \mathrm{m}$; $n=1$ slice, 1 mouse.

Similarly, brain slices harvested from GCaMP6s-positive mice were imaged under the microscope to verify image quality and successful transduction of the targeted sites. Figures 3(a) and 3(c) show widefield fluorescence images of C1V1-mCherry from the harvested brain slices using a $4\times$ objective (Olympus). Here we see positive expression of C1V1-mCherry in neurons, which verifies the virus reached the intended delivery site. Further multiphoton imaging with the optical setup in Figs. 3(b) and 3(d) shows expression of GCaMP-positive neurons, in addition to structures of the hippocampus from C1V1-mCherry. An additional slice preparation in Fig. 3(c), coupled with the 2P imaging results from Fig. 3(d), demonstrates the reliability of the system for producing high-quality images for the fluorophores of interest. Collectively, these results demonstrate the capabilities of the system to visualize the intended fluorescent structures at targeted injection sites.

Fig. 3

Brain tissue expressing GCaMP and C1V1-mCherry. (a), (c) Widefield fluorescence imaging of two separate slice preparations shows very bright fluorescence in distinct regions of the hippocampus. (b), (d) Magnified 2P images corresponding to the areas in the red dashed boxes in the widefield images on the left illustrate positively transduced cells throughout the mouse brain at different brain sites. A wire harp was used to keep the tissue stationary, and one wire strand is shown in (b) as a large diagonal bar across the hippocampus. (a), (c) $4\times$ scale bars represent $300\mu \mathrm{m}$. (b), (d) $20\times$ scale bars represent a distance of $50\mu \mathrm{m}$; $n=2$ slices, 2 mice.

3.2.

Simultaneous Imaging and Stimulation

Finally, verification that this system could simultaneously perform 2P calcium imaging and 2P excitation of C1V1-mCherry was performed. Expression for mCherry is shown in Figs. 3(a) and 3(c), and both GCaMP6s and C1V1-mCherry are shown in the multiphoton images in the remainder of Fig. 3, as well as in Figs. 4(a) and 4(c). Sites visibly present with GCaMP6s, C1V1-mCherry, and co-expression of the two can thus be easily identified. Expression of GCaMP6s can clearly be seen throughout the tissue and targeted delivery of C1V1-mCherry expresses well throughout the targeted region of the hippocampus. These validate, again, that targeted delivery of the virus and the genetic expression was successful. Next, actuation of the opsin was validated by excitation with a CW laser source. As previous literature has shown that the efficiency for absorption in opsins is significantly higher for CW excitation than 2P,⁵² this was performed as an initial validation that the opsin was functional. Stimulation protocols for both CW and 2P illumination are summarized in Table 1 for ease of reference. Figure 4(b) shows the calcium response of wide-field excitation of the transduced sample with a 561-nm CW laser ($\sim 2\mathrm{mW}/{\mathrm{mm}}^2$), corresponding to the locations in Fig. 4(a). Immediately, a dramatic increase in signal from GCaMP6s was evoked and captured from dozens of cells. Plots from three representative cells are shown in Fig. 4(b), with the average response (red) across multiple stimulation epochs ($n_{\text{epochs}}=5$) superimposed on individual trials (pink). The experiment was repeated three times [Fig. 4(b), 1P: P1 to P3] to demonstrate the repeatability of these responses. These data verify the presence of each probe of interest, and the capability of both eliciting and detecting neural responses. Furthermore, they illustrate the capability of monitoring dynamic calcium transients, optogenetically evoked or otherwise.

Fig. 4

Photoactivation of GCaMP using (a), (b) single-photon excitation of C1V1-mCherry and (c), (d) from multiphoton excitation. The red tick marks in (b) and (d) represent the time of the optical stimulus. Three separate stimulation protocols for CW (CW: P1 to P3) and 2P stimulation (2P: P1 to P3) were run for each sample. For CW excitation, there was widefield illumination (b), whereas for 2P protocols (d), a single area was targeted. In (c), the blue arrows highlight the cells excited for protocol 2P: P1, the orange arrow indicates the targeted cell for protocol 2P: P2, and the green arrow indicates the cell targeted in 2P: P3. Cells were labeled between (i) and (vi), and their calcium signals are plotted for each dataset (b), (d). Red traces represent the mean signal from the calcium traces plotted in pink. Data from protocol P1 under 2P stimulation (d) had a single, burst stimulation epoch, and hence, no individual trials. CW light was incident for 500 ms. Multiphoton sources were incident on the sample for 30 ms for protocol P1, 50 ms for protocol P2, and 200 ms for protocol P3. Image scale bars represent $50\mu \mathrm{m}$; $n=5$ cells under CW stimulation; and $n=6$ cells under multiphoton stimulation. Both CW and multiphoton protocols were performed on the same brain slice. Calcium plot scale bars represent 60% (2P) or 2000% (single photon) $\frac{\mathrm{\Delta }F}{F}$ on the $y$ axis, and 2 s on the $x$ axis.

Finally, single-cell excitation and imaging using the supercontinuum source was tested. Initial reference images of GCaMP6s and C1V1-mCherry were acquired and used to guide localization of the stimulation beam. Thereafter, a cell of interest was selected, and the stimulation protocol was initiated. Two-photon-excited imaging of GCaMP6s was performed at a 13.12-Hz framerate, whereas 2P stimulation pulses with 30, 50, and 200 ms exposure were applied to cells i/ii, v, and ii, respectively. The results are shown in Figs. 4(c) and 4(d), where protocol 2P: P1 corresponds to the 30-ms stimulus applied to cells i and ii, 2P: P2 the 50-ms stimulus at cell v, and 2P: P3 to the 200-ms stimulus at cell ii. In our studies, protocols were defined by the cells targeted, the scanning geometry, and the stimulation time, as detailed in Sec. 2 and Table 1. Protocol 2P: P1 had only a single-stimulation epoch, compared to five for all other protocols. Previous literature has demonstrated that repeated 2P optical stimuli in a given timeframe result in more reliable calcium responses,⁷ and this tested for this particular protocol. For ease of reference, all cells stimulated under different protocols at the same site are identified in Fig. 4(c), with the blue arrows specifying cells stimulated under protocol 2P: P1, the orange arrow specifying the identified cell for protocol 2P: P2, and the green arrow the cell targeted under protocol 2P: P3. Spiral scanning was implemented for excitation of the entire cell to elicit sufficient 2P responses. Notably, the calcium transients for each protocol resulted from a single or pair of activated cell(s) near the stimulation site, rather than from several as with CW excitation. This suggests the beam was localized to a designated spot as intended for each protocol, and consequently, resulted in cells within a confined region being excited. This is especially clear for protocols 2P: P2 and 2P: P3, where cells v and ii, respectively, were targeted and showed a clear response, coupled with consistent increases in intensity for each stimulation epoch as well. In contrast, cells iv and i for protocol 2P: P2 and cells i and vi in protocol 2P: P3 did not respond. When a targeted region encompassing two neighboring cells was activated, as in Fig. 4(d) (protocol 2P: P1), there was an increase in fluorescence from both cells i and ii at the time of stimulation, with an absence thereof for cell iii. This provides further support that the targeted regions alone resulted in a calcium response. Unlike the widefield CW illumination, 2P protocols targeted different single cells. These targeted regions, again, show cells at the site of interest that responded to the optical stimulus. More specifically, cells i and ii in protocol 2P: P1, cell v in protocol 2P: P2, and cell ii in protocol 2P: P3 all showed a clear and consistent response to 2P activation. In each case, the targeted cell showed, on average, an increase in fluorescence.

Notably, when cell ii was stimulated individually, in contrast to simultaneously stimulating both cells i and ii in protocol 2P: P1, only cell ii showed a clear response. Figure 4(d) along protocol 2P: P1 targeted both cells for stimulation, whereas protocol 2P: P3 targeted only cell ii. Also the calcium signals evoked with CW excitation are considerably higher than those for 2P activation, which is consistent with reduced amplitudes from patch-clamp recordings reported from other groups.^52,56 These results provide evidence that in addition to imaging, supercontinuum generation is useful for neural applications and optogenetics, reducing the need for complex multilaser systems, associated space occupation, and flexibility with regards to multispectral excitation.