Simultaneous two-photon calcium imaging at different depths with spatiotemporal multiplexing

Transcription

1 Nature Methods Simultaneous two-photon calcium imaging at different depths with spatiotemporal multiplexing Adrian Cheng, J Tiago Gonçalves, Peyman Golshani, Katsushi Arisaka & Carlos Portera-Cailliau Supplementary Figure 1 Supplementary Figure 2 Supplementary Figure 3 Supplementary Figure 4 Supplementary Figure 5 Effect of tissue depth on light scattering as measured by camera-type detection or photomultiplier tube. Multifocal in vivo calcium imaging and simultaneous electrophysiology. Optical configuration for multiple beam imaging. Read-out electronics. Optical transmission through various components of the prototype spatiotemporally multiplexed microscope. Supplementary Figure 6 Supplementary Figure 7 Supplementary Figure 8 Supplementary Note 1 Negligible beam cross-talk for Fluo-4 with 4-beam temporal multiplexing. Negligible beam cross-talk with OGB-1 AM with two-beam temporal multiplexing. Multiple beam scanning increases brightness and scanning speed for constant average laser power per beam. Longer fluorescence lifetimes, improving acquisition rates, and additional fluorophores.

2 Supplementary Figure 1 Effect of tissue depth on light scattering as measured by camera-type detection or photomultiplier tube. Cooled CCD PMT (HPD) Axial depth (µm) Radial distance (µm) µm -100 µm -200 µm -300 µm Cortical slice thickness 20 0 Intensity (db) Supplementary Figure 1 Effect of tissue depth on light scattering as measured by camera-type detection or photomultiplier tube. Resolution is measured as a function of brain depth. Fluorescence spheres (5 µm in diameter) were placed under fixed brain slices of varying thickness and imaged with 2PLSM (single beam), using either a cooled CCD camera (left) or a PMT (right) for detection. Note the degradation in resolution radially (laterally) seen with camera-type detection, due to visible wavelength scattering. Axial resolution is preserved with camera-type detection, demonstrating the degradation of the detection point spread function (PSF) rather than the excitation PSF. Dilution of signal with depth requires longer integration times for the camera (over minutes for 300 µm in depth, vs. seconds for PMT detection).

3 Supplementary Figure 2 Multifocal in vivo calcium imaging and simultaneous electrophysiology. bm. 4 bm. 3 bm. 2 bm. 1 a b µm c d 0.5 F/F 1 10 Cell Number s 20 s 10 pa Cell-attached recording 0.2 F/F e Simulated calcium imaging (from Ephys) Actual calcium imaging 2s Supplementary Figure 2 Multifocal in vivo calcium imaging and simultaneous electrophysiology. (a) Field of view scanned with 4 laser beams. The tip of the patch-clamp electrode is outlined in yellow and the recorded cell is indicated by an arrow. (b) Cell contours of 91 individual neurons. (c) Raw calcium traces from the first 10 cells. The simultaneous cell-attached electrophysiology trace is shown at the bottom (blue). The shaded area is shown with finer temporal resolution in e. (d) Raster plot of the calcium transient events for all 91 cells. (e) Detail of simultaneous electrophysiology and calcium imaging. From top to bottom: cellattached recording trace (black), simulated (convolved) calcium imaging trace based on electrophysiology (red), actual calcium imaging trace (green).

4 Supplementary Figure 3 Optical configuration for multiple beam imaging. a c SM1 SM1 PBS1 θ +0 ns +3 ns +6 ns +9 ns +0 ns p-pol L4 b SM1 M3 M2 PBS1 SE-M1 PBS2 M4 L1 L2 L3 +0 ns p-pol +3 ns s-pol +6 ns s-pol +9 ns p-pol L1 +6 ns s-pol L2 +3 ns s-pol L3 +9 ns p-pol d Telescope (T) Back aperture image Critical point (x) e ~ f o 2 T / d for Td / f o >> 1 x o o Back focal plane Back aperture d 1 f o d 2 o x o d 1 d 2 Back focal space Object space focus d critical point zero focal shift Back focal space focus d Object space Supplementary Figure 3 Optical configuration for multiple beam imaging. (a) To generate a vertical pattern of beam spots for single plane multibeam imaging, beams were converged onto scanning mirror (SM1) with an angular spacing θ = M x / f o where M is the demagnification by relay lens telescopes, x is the desired pitch at sample, and f o is the focal length of the objective lens, as approximated by f t / M o, where f t is the manufacturer tube lens focal length and M o is the objective magnification rating. (b) To generate a scan of four imaging planes, beams of varying divergence introduced by lenses (L1, L2, and L3) were combined by polarizing beam splitters (PBS1 and PBS2) and further combined by a sharp edge mirror (SE-M1). This led to a shift of two beams by ~ 20 µm. Note this could be done with a pinhole mirror of sufficient precision as well, to increase co-linearity. A pinhole mirror method would allow for a generalization of this configuration to N beams. (c) To generate a scan of two imaging planes with two beams each, a single polarizing beam splitter (PBS1) was used to combine four beams. Lenses introduced varying divergence or convergence (L1-L4). In all cases a-c, the back aperture was filled uniformly by maintaining uniform beam diameter at SM1. (d) An idealized optical train is shown for generating axial focal shift, consisting of the object space imaged by the objective lens, and a back focal space after a relay telescope with magnification T. To examine focal shift, let the objective focal length f o << f t, the tube length. Then, the back focal plane (BFP) will approximately Nature coincide Methods: with doi: /nmeth.1552 the objective exit pupil. Small displacements δ from the BFP will be imaged by the telescope at approximately Tδ from the BFP image (critical point x). (e) Ultimately the focal shift d 1 - f o = will be approximately f o2 T / d 1 where d 1 is the beam focus distance from x, making the approximation Tδ/f o >> 1.

5 Supplementary Figure 4 Read-out electronics. Photocathode 8kV APD Bias 300V Scanner synchronization Color select/shutter/piezo/etc γ e- HAPD Preamplifier Laser monitor Color select/ interleave V REF Demultiplexer Transimpedance/ charge-sensitive amplifier To display To disk array Frame grabber PCIe DMA transfer Gate generator Delay lines Host PCU Supplementary Figure 4 Read-out electronics. HPD signals are amplified and color channel is selected or interleaved by a high-speed analog switch (blue). The signals from the four separate beams are de-multiplexed by analog multiplication by four delayed gates with 3 ns width and 12 ns period (yellow). The delayed gates are generated by a laser monitor signal (green). The resulting current-voltage signal is then shaped for digitization by a commercial four-channel high-speed frame grabber (orange). PCI-express DMA transfer using the supplied frame grabber software library allows for real-time display, processing, and disk array recording (purple).

6 Supplementary Figure 5 Optical transmission through various components of the prototype spatiotemporally multiplexed microscope. 1 1 Power Transmission Spatiotemporal multiplexing prototype 2PLSM Zeiss 40x 1.0NA Olympus 40x 0.8NA Laser After delay lines After scanners After objectives 0.1 Supplementary Figure 5 Optical transmission through various components of the prototype spatiotemporally multiplexed microscope. Measurements were made before the system (at the laser output), after the delay lines section, after the scanning mirrors, and after the objective lens.

7 Supplementary Figure 6 Negligible beam cross-talk for Fluo-4 with 4-beam temporal multiplexing. a z y z x y Beam 2 only x Beam 1 only z y x Beam 2 (with beam 1) b 9 25 µm -0.2% 4-1.7% % 5 A(x,y) 10 B(x,y) +2.8% C(x,y) = A(x,y) + γb(x,y) Supplementary Figure 6 Negligible beam cross-talk for Fluo-4 with 4-beam temporal multiplexing. (a) Field of view scanned with beam 2 alone (left), beam 1 alone (middle), keeping other beams blocked, and field of view of beam 2 while simultaneously scanning with beam 1 (right). These images were acquired using the 3-D spatiotemporal multiplexing configuration and are average time intensity (xyt) projections of 45 sec-long in vivo calcium imaging movies. Regions of interest (ROI; yellow outlines) were selected around somas of presumed neurons (ROIs 1-9) and neuropil (ROI 10) in the beam 2 image. Note that bright astrocytes seen in layer 1 (beam 1 image) are not present in the beam 2 image. The rectangular region outlined in the left panel (red box) is shown at higher zoom in b. (b) Higher magnification detail of boxed region in panel a. Examples of cross talk values are shown for 3 representative cell body ROIs and 1 neuropil ROI. The average intensity change in beam 2 ROIs in the presence of beam 1 excitation was 3.0 ± 4.1 % (mean ± st. Nature Methods: doi: /nmeth.1552 dev.), suggesting that fluorescence leak-through between beams is not a dominant noise factor (see Online Methods).

8 Supplementary Figure 7 Negligible beam cross-talk with OGB-1 AM with 2-beam temporal multiplexing. Beam 1 only 10.5% 7.1% 4.1% 3.3% 5.6% 6-12 ns Cell 1 Cell 3 Cell 4 Cell 5 Cell 2 75 µm 0-6 ns Supplementary Figure 7 Negligible beam cross-talk with OGB-1 AM with 2-beam temporal multiplexing. A single beam was used to scan the field of view (top), with images at 0-6 ns (bottom) and 6-12 ns (top) after the excitation pulse shown images were acquired at 100 Hz and summed. Representative cells were selected (yellow, bottom) and the corresponding % cross-talk (i.e., contribution in the 6-12 ns image) for each cell is shown (top). The average static cross-talk for OGB was calculated to be 6.1 ± 2.9% (mean ± st. dev.). Because this percentage corresponds to somatic OGB-1 that on average already exhibits its longer lifetime of ~2.5 ns, action potential-induced calcium transients of ~ 10% F/F would be expected to result in cross-talk transients of 0.6%, which is on the order of Poisson fluctuations.

9 Supplementary Figure 8 Multiple beam scanning increases brightness and scanning speed for constant average laser power per beam. Beam 1 (0-12 ns) Beam 1 (0-6 ns) Beam 2 (6-12 ns) 75 µm 75 µm 7.13 ADC counts per pixel 1.43 x 10 6 ADC counts per frame (mean over 4000 frames) 800 x 250 pixels, 100 fps 1.5 W into system 7.98 ADC counts per pixel 3.19 x 10 6 ADC counts per frame (mean over 4000 frames) 800 x 500 pixels, 100 fps 1.5 W per beam into system Supplementary Figure 8 Multiple beam scanning increases brightness and scanning speed for constant average laser power per beam. To demonstrate specific advantages of multiple beam scanning in vivo, a field of view of mouse barrel cortex (layer 2/3) was stained with OGB-1 AM and imaged first with a single beam (left) and then with two beams 6 ns apart (right). The average power per beam was held constant and the laser repetition rate was 80 MHz. This implies a constant laser pulse energy, with 80 MHz pulse delivery at the sample for the case of a single beam, and 160 MHz pulse delivery for the case of two beams. Not only was scanning speed increased in the case of two beams (800 x 250 pixels at 100 frames per second versus 800 x 500 pixels at 100 frames per second), but fluorescence yield was approximately doubled (3.19 x 10 6 vs x 10 6 ADC counts). Thus, in a hypothetical situation where laser pulse energy is limited by non-linear photodamage, increasing laser pulse delivery rates is an option for increasing signal level. However, rather than merely increasing delivery rate to a single point, spatially distributing and resolving beams also provides an increase in scanning rate. Laser pulse energy may not necessarily be limited in all types of imaging, and non-linear damage limits (in terms of input pulse energy at the brain surface) certainly will have a functional dependence on depth. This functional dependence and the precise ideal pulse energies versus depth remain to be explored.

10 SUPPLEMENTARY NOTE Spatiotemporal multiplexing for fluorophores with longer lifetimes The number of spatiotemporally multiplexed beams could be greater than 4 provided one uses a laser with a lower repetition rate, which is determined by the length of the laser cavity or additional components such as regenerative amplifiers 1 or cavity dumpers 2. In addition, the lifetime of the calcium indicator used for temporally multiplexed excitation must be carefully considered to avoid cross-talk between beams. Much of in vivo 2PCI these days is done with OGB, so there may be great interest in the potential suitability of this dye for spatiotemporal multiplexing. OGB has both short and long lifetimes, (~350 ps and ~2.5 ns in live cells, respectively), the relative proportion of which appears to change in the presence of calcium as determined by its dissociation constant (K d ~ 200 nm) 3-4. This is due to a change in the radiative quantum yield in the presence of calcium. Fluo-4, whose excitation efficiency (but not its quantum yield) increases in the presence of calcium, has a relatively constant lifetime of ~1 ns. Thus, Fluo-4 was the preferred dye for our calcium imaging experiments using a 80 MHz laser and a 4-beam design. In contrast, at spiking levels ([Ca 2+ ] > 1 µm) 5, OGB was not ideally suited for our 4-beam multiplexing approach. However, using only 2 beams separated in time by 6 ns, we demonstrate in vivo calcium imaging with OGB with negligible cross-talk (Supplementary Fig. 7). In the future, any of the above mentioned approaches for lowering the laser repetition rate would permit the use of spatiotemporal multiplexing for 2PCI of dyes with longer lifetimes (~3 ns), including GFP-based genetically encoded calcium indicators. Additionally, in light of its calcium-dependent lifetime, OGB would also benefit from time-gated detection by excluding photons detected for the duration of its short [Ca 2+ ] free lifetime.

11 Improving acquisition rates by combining spatiotemporal multiplexing with AODs and targeted path scanning In principle, spatiotemporal multiplexing can also be combined with other approaches that address the closely related issues of signal to noise ratio (SNR), field of view (FOV) in two and three dimensions, and time resolution for in vivo 2PCI, including random access point scanning with acousto-optical deflectors (AODs) or targeted-path scanning with closed loop mirrors 6-8. Both approaches seek to increase time resolution and SNR by restricting laser delivery to a region of interest (e.g., neuronal cell bodies, but not the surrounding neuropil). AODs allow millisecond time resolution by scanning a limited number of points coinciding with preselected subsets of labeled neurons over a large FOV. Unfortunately, the number of AODs necessary to implement 3-D scanning introduces considerable spatial and temporal dispersion, as well as power loss, although this can be easily compensated for using prism correction or pulse pre-chirping 7. In addition, the axial extent of scanning with AODs is less than 50 µm 8. The arbitrary targeted path scanning technique takes advantage of closed-loop mirrors to scan optimized contours coinciding with cell bodies 6. For 3-D imaging, this approach can be combined with a piezo-electric objective scanner, resulting in acquisition rates of several Hz for small volumes 9. More importantly, an advantage of the raster scanning approach using multiple beams that we used is that it preserves full frame image scanning in 3-D, which allows a full range of image processing methods for cell identification and image registration. To combine these approaches with spatiotemporal multiplexing would require multiple sets of scanners and simultaneous projection of their apertures on to the objective back aperture. One solution would be to simply project the apertures of multiple sets of AODs or closed-loop scanning mirrors on to the objective back aperture with different access angles. This would allow each individual beam to scan

12 distinct FOVs. To achieve independent beam scanning in the same FOV, polarization optics generally provides a passive solution for combining the optical paths of two sets of scanners. The availability of more elegant solutions will depend on the specific scan method and pattern. Fluorescence lifetime-induced cross-talk We measured the potential beam cross-talk due to the finite fluorescence lifetime of the dye compared to the 3 ns beam delays during normal experimental imaging conditions. A region of layer 2/3 in barrel cortex was stained with Fluo-4 AM and in vivo calcium imaging was carried out as in Fig. 3, with different beams scanning different axial planes in 3-D (Supplementary Fig. 6). First, a 45 s time-lapse sequence of images was obtained with only beams 2 scanning the tissue (the other beams were blocked). Next, 1 minute later, another 45 s time-lapse sequence of images was obtained with only beam 1. A final 45 s time-lapse sequence was taken with beams 1 and 2 scanning the tissue at different planes (30 µm apart in the z direction). This last calcium movie is subject to crosstalk between beams. Regions of interest around Fluo-4 labeled cells in the beam 2 image plane were manually selected and their average pixel intensity throughout the movie was calculated. The average leakage (γ) of beam 1 into the beam 2 image was calculated using the following linear combination, C(x,y) = A(x, y) + γ(x,y) B(x,y) where A is the pixel intensity in the beam 2 alone image, B is the pixel intensity in the beam 1 alone image, and C the pixel intensity in the beam 2 image when simultaneously scanning with beam 1. Cross talk varied for different ROIs, but the average leak-through was 3.0 ± 4.1% (mean ± st. dev.). Cross-talk likely also varies slightly both spatially due to other fluorophore compositions (e.g., autofluorescence),

13 and also temporally (e.g., neuronal activity), which is why these measurements were done over 45 sec. For example, if cross-talk occurs during a 10% ΔF/F calcium transient, the dynamic contribution of this transient to the adjacent beam would be small (a ~0.3% transient). This effect will be the same order or less than Poisson fluctuations in our system and is ameliorated by the fact that the fluorescence lifetime of Fluo-4 decreases during a calcium transient 5, which brings somatic calcium concentration to the dye dissociation constant k 10 d. In the case of OGB, using two beams spaced by 6 ns in time, the average signal leak-through for cellular regions was 6.1 ± 2.9% (mean ± st. dev.; Supplementary Fig. 7). REFERENCES 1 Theer, P., Hasan, M. T. & Denk, W. Two-photon imaging to a depth of 1000 microns in living brains by use of a Ti:Al2O3 regenerative amplifier. Opt. Lett. 28, (2003). 2 Ramaswamy, M., Ulman, M., Paye, J. & Fujimoto, J. G. Cavity-dumped femtosecond Kerr-lens mode-locked Ti:A1(2)O(3)laser. Opt. Lett. 18, , doi:12002 [pii] (1993). 3 Wilms, C. D., Schmidt, H. & Eilers, J. Quantitative two-photon Ca2+ imaging via fluorescence lifetime analysis. Cell Calcium 40, 73-79, doi:s (06) [pii] /j.ceca (2006). 4 Wilms, C. D. & Eilers, J. Photo-physical properties of Ca2+-indicator dyes suitable for two-photon fluorescence-lifetime recordings. J. Microsc. 225, , doi:jmi1746 [pii] /j x (2007).

14 5 Helmchen, F., Imoto, K. & Sakmann, B. Ca2+ buffering and action potential-evoked Ca2+ signaling in dendrites of pyramidal neurons. Biophys. J. 70, , doi:s (96) [pii] /S (96) (1996). 6 Lillis, K. P., Eng, A., White, J. A. & Mertz, J. Two-photon imaging of spatially extended neuronal network dynamics with high temporal resolution. J. Neurosci. Methods 172, , doi:s (08) [pii] /j.jneumeth (2008). 7 Kremer, Y. et al. A spatio-temporally compensated acousto-optic scanner for two-photon microscopy providing large field of view. Opt. Express 16, , doi: [pii] (2008). 8 Duemani Reddy, G., Kelleher, K., Fink, R. & Saggau, P. Three-dimensional random access multiphoton microscopy for functional imaging of neuronal activity. Nat. Neurosci. 11, , doi:nn.2116 [pii] /nn.2116 (2008). 9 Gobel, W., Kampa, B. M. & Helmchen, F. Imaging cellular network dynamics in three dimensions using fast 3D laser scanning. Nat. Methods 4, (2007). 10 Yoshiki, K., Azuma, H., Yoshioka, K., Hashimoto, M. & Araki, T. Finding of optimal calcium ion probes for fluorescence lifetime measurement. Optical Review 12, (2005).