Multicolor two-photon light-sheet microscopy Pierre Mahou, Julien Vermot, Emmanuel Beaurepaire & Willy Supatto

Mahou et al Supplementary Information Page 1 / 16 Multicolor two-photon light-sheet microscopy Pierre Mahou, Julien Vermot, Emmanuel Beaurepaire & Willy Supatto SUPPLEMENTAR INFORMATION SUPPLEMENTAR ITEM TITLE Supplementary Figure 1 Supplementary Figure 2 Supplementary Figure 3 Multicolor 2P-SPIM optical setup Mixed wavelength 2P-SPIM is less affected by chromatic aberrations than mixed wavelength 2P-LSM Pulse synchronization for wavelength mixing in bidirectional 2P-SPIM Supplementary Figure 4 Supplementary Figure 5 Supplementary Table 1 Reduced photobleaching in multicolor 2P-SPIM compared to 2P-LSM Fast multicolor 2P-SPIM imaging of the beating zebrafish heart: absence of significant heating and photobleaching Acquisition parameters for photobleaching comparison of 2P-SPIM with 2P-LSM Supplementary Table 2 Acquisition parameters for 4D multicolor imaging of the zebrafish beating heart Supplementary Methods Supplementary References

Mahou et al Supplementary Information Page 2 / 16 Supplementary Figure 1 Multicolor 2P-SPIM optical setup a Synchronized pulse trains Delay τ HWP Beam divergence τ Time TiS (λ 1 ) HWP HWP OPO (λ 2 ) Power λ 1 Power λ 2 DM Shutter HWP Periscope + -Scanner b 2P-SPIM bidirectionnal illumination microscope Tube Lens Illumination arm B τ B τ A Time Time Illumination arm A Sample Chamber illumination objective B illumination objective A detection objective Widefield multispectral detection Tube Lens Imaging plane Image Splitter Tube Lens Scan Lens EM- CCD HWP Scan Lens Z PBS HWP c Widefield multispectral detection d Image Splitter Imaging plane 2PEF at λ 2 Detection Objective Imaging plane EM- CCD 2C-2PEF at λ 3 Tube Lens DM1 DM2 2PEF at λ 1

Mahou et al Supplementary Information Page 3 / 16 Supplementary Figure 1 Multicolor 2P-SPIM optical setup (a) Synchronized pulse trains for mixed wavelength multicolor excitation. The pulse trains are produced by a Ti:sapphire laser (TiS, λ 1 ) and by an optical parametric oscillator (OPO, λ 2 ). Beam powers are adjusted using half wave plates (HWP) and polarization beam splitters. The time delay between the pulse trains is adjusted with a motorized delay line. Beams sizes and divergence are controlled with separate telescopes, before combination with a dichroic mirror (DM). (b) 2P-SPIM bidirectional illumination microscope. Two non-symmetrical illumination arms are generated by splitting the laser beams with a polarization beam splitter (PBS). Each illumination arm is composed of a low numerical aperture (NA) water immersion objective (10 0.30 NA, illumination objective), a scan lens and a tube lens chosen to minimize dispersion imbalance between the two paths. A galvanometer mirror (-scanner) scans the excitation beams vertically ( direction) in order to produce the illumination light sheet. (c) Widefield multispectral detection. Fluorescence from the illuminated plane is collected by a high NA detection objective (16 0 NA), and three spectral channels are spatially split and projected onto on an electron-multiplying charged coupled device camera (EMCCD) using an image splitter. Note that since the field of view is usually larger in the axis than in the axis in a SPIM setup, it is advantageous to split the image along the axis and use the full collecting area of the camera. (d) Trichromatic 2PEF images recorded on a live fly embryo at gastrulation showing endogenous 2PEF from the yolk and the vitelline membrane excited at λ 1 = 820 nm in blue, 2PEF from RFP-labeled nuclei excited at λ 2 = 1,175 nm in red, and 2C-2PEF from GFP-labeled membranes excited with wavelength mixing corresponding to an equivalent twophoton excitation wavelength λ 3 = 965 nm in green (exposure time 1.5 s; scale bar 50 µm).

Mahou et al Supplementary Information Page 4 / 16 Supplementary Figure 2 Mixed wavelength 2P-SPIM is less affected by chromatic aberrations than mixed wavelength 2P-LSM a -scanning direction b 2-photon point scanning microscopy 2P-LSM Illumination -scanning direction High NA illumination Detection δ Excitation (λ 2 ) Beam overlapping parameters Z δ Excitation (λ 1 ) δz 2-photon scanned light-sheet microscopy 2P-SPIM Light-sheet (λ 2 ) Scanning direction Light-sheet (λ 1 ) Illumination beam direction Effect of axial chromatic aberration on 2C-2PEF signal δz Low NA illumination δ Detection Z δz δ δ c -scanning direction (µm) Important 2C-2PEF signal reduction with axial mismatch δ -scanning direction (µm) Moderate 2C-2PEF signal reduction with axial mismatch Effect of lateral chromatic aberration during scanning on 2C-2PEF signal δ δ δ Large scan angles, tight focus Z Small scan angles, moderate focus 2C-2PEF (a.u.) -scanning direction (µm) -scanning -scanning 0 0 200 400 600 -scanning direction (µm) Lateral chromatic aberration limits field of view 2C-2PEF (a.u.) -scanning direction (µm) 0 0 200 400 600 -scanning direction (µm) Field aberration has limited effect

Mahou et al Supplementary Information Page 5 / 16 Supplementary Figure 2 Mixed wavelength 2P-SPIM is less affected by chromatic aberrations than mixed wavelength 2P-LSM (a) Parameters describing the spatial overlap between two illumination foci used for mixed wavelength excitation (at λ 1 and λ 2 ) in a two-photon point scanning geometry (2P-LSM, left) and in a two-photon scanned light-sheet geometry (2P-SPIM, right). In 2P-LSM, the illumination is done along the Z axis and the tightly focused excitation beams are usually raster-scanned along the and directions to record an image. In the SPIM geometry, the illumination () and detection (Z) axes are orthogonal, and the weakly focused excitation beams are scanned along the third dimension () to generate an illumination sheet. Spatial mismatches are described by δ, δ, δz in each spatial direction. In both geometries, a good spatial overlap between foci can be obtained by maximizing a reference two-color two-photon (2C-2PEF) signal. During imaging, this overlap is mainly affected by chromatic aberrations of the illumination objective. (b) Axial chromatic aberrations are generally independent on the scanning angle in the range used for imaging. They can be compensated for by adjusting the relative amount of defocus between the two beams (e.g. using telescopes). However residual chromaticity affects beam overlap more strongly in the tightly focused 2P-LSM geometry than in the low-excitation NA 2P-SPIM one. Indeed, for the same axial resolution Z (i.e. NA SPIM 2 NA LSM 3.3n), the 2-photon excitation point spread function is elongated along the beam propagation axis by a factor FOV Z (3.3n NA LSM ) 2 in 2P-SPIM compared to 2P-LSM. In our experiments, this extension factor is FOV Z = 30 with NA LSM = 0 and Z = 2.5 3.0 μm at 820/1,100 nm. As a consequence, the large extension of the excitation along the laser propagation axis makes beam overlap easier to achieve and less sensitive to chromatic aberrations in this direction than using 2P- LSM. (c) Lateral chromatic aberrations depend on the scanning angle and are more difficult to correct. Such aberrations limit the extension of the field of view over which mixed wavelength excitation is achieved. When imaging similar areas of homogeneously labelled samples (bottom graphs), this limitation is less severe in 2P-SPIM than in 2P-LSM due to tighter foci and larger scan angles: (i) for the same axial resolution the beam extensions are larger due to lower excitation NA ( SPIM LSM = 3.3n NA LSM 5.5 for NA LSM = 0) and (ii) for a similar field of view along the scanning direction, scanning angles are smaller due to lower magnification of the excitation objective. For instance, using a 10 illumination objective with 2P-SPIM and a 25 objective with 2P- LSM lead to scanning angles θ that are 2.5 times smaller for the same scanning area in both geometries (θ M Obj y 2f Tube where M Obj is the objective magnification, f Tube the focal lens of the nominal tube lens and y the scanning).

Mahou et al Supplementary Information Page 6 / 16 Supplementary Figure 3 Pulse synchronization for wavelength mixing in bidirectional 2P-SPIM a Z Detection b Arm B Arm A Endo GFP RFP Illumination arm B τ B Time Sample Time Illumination arm A τ A 2PEF Intensity (a.u) τ FWHM 400 ± 7 fs τ B τ A 51 ± 2 fs 0-400 -200 0 200 400 Delay τ (fs) c Detection d Detection e Arm B Arm A Arm B Detection Arm A τ B Embryo Embryo Embryo τ A τ B τ A Time Time Time Endo (2PEF at λ 1 ) GFP (2C-2PEF at λ 3 ) RFP (2PEF at λ 2 ) 50 µm

Mahou et al Supplementary Information Page 7 / 16 Supplementary Figure 3 Pulse synchronization for wavelength mixing in bidirectional 2P-SPIM (a) Principle of temporal synchronization between two pulse trains (e.g. λ 1 = 820 nm and λ 2 = 1,175 nm, leading to λ 3 = 965 nm) in a 2P-SPIM setup with bidirectional illumination. In the case of a single illumination path, the time delay τ A between the pulse trains at the focus of the excitation objective can be nulled by adjusting a delay line. In the case of bidirectional illumination, beam synchronization is more complex, since the time delay between the pulse trains for the two illumination arms, noted τ A and τ B, must be simultaneously cancelled. In our case, bidirectional illumination is implemented with two non-symmetrical paths, and we equalize τ A and τ B by matching the optics differential dispersion between both arms, without using a second delay line. (b-e) Simultaneous blue (endogenous, Endo), and red (RFP) two-photon excited fluorescence (2PEF) and green (GFP) two color two-photon-excited fluorescence (2C-2PEF) signals recorded in a live fly embryo. (b) Signals produced by each illumination arm as a function of the delay τ between the pulse trains adjusted with the delay line. Blue and red fluorescence signals are mainly produced by singlecolor 2PEF and do not depend on time delay between the pulse trains, as expected. However, green signal mainly results from 2C-2PEF and depends on pulse timing. The time delays are simultaneously cancelled in both excitation arms within τ B τ A = 51 ± 3 fs. This value is well below the measured width of the temporal intensity cross-correlation between the two pulse trains τ FWHM = (τ 1 2 + τ 2 2 ) 1/2 = 400 ± 7 fs. This demonstrates that mixed wavelength excitation can be efficiently implemented in a bidirectional 2P-SPIM setup by choosing appropriate relay optics. (c-d) Trichromatic 2PEF images recorded in using each illumination path separately. (e) Similar images using simultaneous bidirectional illumination, illustrating that multicolor 2PEF excitation is uniformly achieved over the field of view (Integration time 1.5 s, Scale bar 50 µm).

Mahou et al Supplementary Information Page 8 / 16 Supplementary Figure 4 Reduced photobleaching in multicolor 2P-SPIM compared to 2P-LSM a Excitation τ Detection c Excitation Detection Excitation Time Sample τ B Time Sample Time τ A b RFP 2PEF (a.u.) GFP 2C-2PEF (a.u.) 0 50 100 150 200 Frame Number ( 1.5 s) 0 50 100 150 200 Frame Number ( 1.5 s) d RFP 2PEF (a.u.) GFP 2C-2PEF (a.u.) RFP 2PEF 0 50 100 150 200 Frame Number ( 1.5 s) GFP 2C-2PEF 0 50 100 150 200 Frame Number ( 1.5 s) λ 1 λ 2 <N Ph blue> <N Ph red> 30 ± 5 mw 25 ± 5 mw 18 ± 3 49 ± 9 20 ± 5 mw 25 ± 5 mw 7 ± 2 47 ± 9 12 ± 5 mw 25 ± 5 mw 5 ± 2 53 ± 9 30 ± 5 mw 0 mw 21 ± 3 8 ± 3 0 mw 25 ± 5 mw 66 ± 9 λ 1 λ 2 <N Ph blue> <N Ph red> 63 ± 5 mw 54 ± 5 mw 88 ± -5 222 ± 4 44 ± 5 mw 54 ± 5 mw 39 ± 4 226 ± 8 22 ± 5 mw 54 ± 5 mw 11 ± 5 185 ± 10 63 ± 5 mw 0 mw 112 ± 5 70 ± 10 0 mw 54 ± 5 mw 200 ± 5 e f 50 µm 50 µm GFP Frame 1 RFP Frame 1 GFP Frame 1 RFP Frame 1 GFP Frame 200 RFP Frame 200 GFP Frame 200 RFP Frame 200

Mahou et al Supplementary Information Page 9 / 16 Supplementary Figure 4 Reduced photobleaching in multicolor 2P-SPIM compared to 2P-LSM We compared photobleaching rates of GFP and RFP during multicolor two-photon imaging of live fly embryos using either point-scanning (2P-LSM) (a, b, e) or bidirectional light-sheet (2P-SPIM) (c, d, f) geometries. In both cases, excitation was achieved using mixed wavelength excitation (λ 1 = 820 nm and λ 2 = 1,175 nm). Fluorescence signals were collected using a three channel non-descanned detection in 2P-LSM 2, and with an image splitter and an EMCCD in 2P-SPIM. Photobleaching rates were measured from time lapse series of 200 images (e-f) using similar detection filters, image acquisition time (1.5 s per image), initial signal levels, and 3 µm axial resolution for the two modalities. Tables in b, d indicate the mean powers used in the experiments and the estimated average initial number of photons collected in the blue (N Ph blue) and red (N Ph red) channels, from endogenous and RFP fluorescence, respectively. We used slightly higher illumination powers in 2P-SPIM than in 2P-LSM to ensure that the photobleaching in 2P-SPIM experiments was not underestimated. Although red FPs can be excited using one-color excitation at short wavelengths (λ 1 = 820 nm) 6, such a strategy results in significant photobleaching in 2P-LSM (pink curve in b top graph) 3. This issue is common to all multicolor imaging schemes employing such wavelengths. In contrast, the photobleaching of RFP is attenuated when using one-color excitation at longer wavelengths (λ 2 = 1,175 nm, orange curve in b top graph). The excitation of multiple fluorophores requires combined excitation at short and long wavelengths resulting in significant photobleaching of RFP in 2P-LSM (green, blue and red curves in b top graph). Importantly, for all investigated conditions of single- and dual-wavelength excitation, we measured that the photobleaching rate of RFP was reduced by more than 25 times in 2P-SPIM compared to 2P-LSM (d top graph). In addition, while we could observe photobleaching of GFP by recording 2C-2PEF signal obtained with wavelength mixing in 2P-LSM (b bottom graph), it was not detectable with multicolor 2P-SPIM imaging using similar conditions (d bottom graph). Together, these data demonstrate that light-sheet illumination enables sustained multicolor two-photon imaging of live tissue with significantly reduced photobleaching compared to standard point scanning approaches. (e,f) First and last GFP/RFP images of 200-frame series recorded using point-scanning (e) and light-sheet (f) illumination.

Mahou et al Supplementary Information Page 10 / 16 Supplementary Figure 5 Fast multicolor 2P-SPIM imaging of the beating zebrafish heart: absence of significant heating and photobleaching a Excitation Objective Detection Objective b No photobleaching of fluorescent proteins during continuous multicolor 2P-SPIM illumination Single beam excitation CFP DsRed 2PEF 2PEF λ 1 λ 2 840 nm 1,090 nm t Two-color excitation GFP 2C-2PEF 2 λ = 3 1 + 1 950 nm λ 1 λ 2 2PEF Intensity (a.u) CFP GFP DsRed λ 1 at 840 nm using 63 ± 5 mw λ 2 at 1,090 nm using 58 ± 4 mw 0 0 50 100 150 200 Frame number ( s) c T = 146 ms Zebrafish heart (60-84 hpf) 2C-2PEF intensity (a.u.) Spectral density (a.u.) Heart beat fluorescence signal 0 0 1 2 3 4 Time in s Heart beat Fourier analysis Heart beat frequency at 2.3 ± Hz 0 0 2 4 6 8 10 Frequency in Hz d Heart beat frequency in Hz 4.0 3.5 3.0 2.5 2.0 Multicolor 2P-SPIM (N = 14) λ 1 = 63 ± 5 mw λ 2 = 58 ± 4 mw Control from Kopp et al, 2005 1.5 15 20 25 30 35 Temperature in C

Mahou et al Supplementary Information Page 11 / 16 Supplementary Figure 5 Fast multicolor 2P-SPIM imaging of the beating zebrafish heart: absence of significant heating and photobleaching (a) Multicolor transgenic embryos expressing red, green and blue fluorophores in the heart (DsRed, GFP and CFP in red blood cells, myocardical cells, and pericardial cells, respectively) were imaged using a single illumination path with two synchronized pulse trains at λ 1 = 840 nm (from the TiS laser, orange) and λ 2 = 1,090 nm (from the OPO, dark red), for exciting CFP and DsRed respectively. Spatio-temporal overlap of the two light sheets at λ 1 and λ 2 creates a mixed wavelength excitation at λ 3 = 950nm (simultaneous absorption of a photon at λ 1 and a photon λ 2 ) to excite GFP with two-color two-photon excited fluorescence (2C-2PEF). (b) To investigate DsRed, GFP and CFP photobleaching when imaging zebrafish embryonic heart with multicolor 2p-SPIM, we continuously illuminated the heart using monodirectional multicolor 2P-SPIM with illumination powers of 63 ± 5 mw at λ 1 and 58 ± 4 mw at λ 2 and recorded one image per second. Plotting the average signal level from DsRed, GFP and CFP depending on time shows no significant photobleaching after 200 seconds of sustained illumination. (c) Quantification of heart beating rate. Green fluorescence from GFP (2C-2PEF signal) was recorded within a specific area of the heart (white dashed square in inset) during monodirectional multicolor 2P-SPIM imaging using illumination powers of 63 ± 5 mw for λ 1 and 58 ± 4 mw for λ 2. Average signal depending on time (top graph) shows oscillations due to beating cycles and its Fourier analysis (bottom graph) provides a quantification of the heart beat frequency. (d) The measured heart beat frequency (red) remains within the expected range at imaging temperature 21 C (gray curve obtained from Kopp et al. 7 ) demonstrating no detectable heating during multicolor 2P-SPIM imaging (N = 14 embryos).

Mahou et al Supplementary Information Page 12 / 16 Supplementary Table 1 Acquisition parameters for photobleaching comparison of 2P-SPIM with 2P-LSM Modality 2P-SPIM 2P-LSM TiS excitation wavelength OPO excitation wavelength Two-color two-photon equivalent excitation wavelength 820 nm 1,175 nm 965 nm Lateral pixel size 7 µm 0 µm Axial resolution (TiS) 2.7 µm 2.0 µm Monodirectional light sheet size (TiS) Bidirectional light sheet size (TiS) Excitation NA (TiS) Axial resolution (OPO) Monodirectional light sheet size (OPO) Bidirectional light sheet size (OPO) 112 µm in and 500 µm in 224 µm in and 500 µm in 0.11 3.6 µm 140 µm in and 500 µm in 280 µm in and 500 µm in n/a n/a 0 3.0 µm n/a n/a Excitation NA (OPO) 0.12 0 Image acquisition time 1.5 s 1.5 s Dichroic mirrors FF484-FDi01 FF560-FDi01 FF500-Di01 FF560-FDi01 Rejection filter FF01-680/SP FF01-680/SP Detection bandpass filters FF01-447/55 FF01-525/50 FF01-607/70 + FF01-582/75 FF01-447/55 FF01-525/50 FF01-607/70 + FF01-582/75 Excitation power (TiS) 22-63 mw 12-30 mw Excitation power (OPO) 54 mw 25 mw

Mahou et al Supplementary Information Page 13 / 16 Supplementary Table 2 Acquisition parameters for 4D multicolor imaging acquisition of the beating heart Modality TiS excitation wavelength OPO excitation wavelength Two-color two-photon equivalent excitation wavelength 2P-SPIM 840 nm 1,090 nm 949 nm Lateral pixel size 7 µm Z-step between acquisitions 2 µm Axial resolution (TiS) Light sheet size (TiS) Excitation NA (TiS) Axial resolution (OPO) Light sheet size (OPO) Excitation NA (OPO) Image exposure time Frame rate Dichroic mirrors Rejection filter Detection bandpass filters Excitation power (TiS) Excitation power (OPO) 2.7 µm 112 µm in and 250 µm in 0.11 3.6 µm 140 µm in and 250 µm in 0.12 10 or 20 ms 85 fps or 48 fps FF484-FDi01 FF560-FDi01 FF01-680/SP FF01-450/80 FF01-525/50 FF01-607/70 63 mw 58 mw

Mahou et al Supplementary Information Page 14 / 16 Supplementary Methods 2P-SPIM optical setup The 2P-SPIM optical setup is described in Supplementary Fig. 1. Synchronized 80MHz pulse trains with 820/1,175 nm or 840/1,090 nm central wavelengths and 200-400 fs pulse durations are obtained from a Ti:sapphire laser (λ 1 = 820 or 840 nm, Chameleon Ultra II, Coherent) and an optical parametric oscillator (λ 2 = 1,175 or 1,090 nm, OPO, APE). Beam powers are adjusted using half wave plates and polarization beam splitters. Time delay between the pulse trains is adjusted with a motorized delay line and permits pulse synchronization. Beams sizes and divergence are controlled with separate telescopes, before combination with a dichroic mirror (DM, 1000-DCR, Chroma). Two nonsymmetrical illumination arms each consist of a low NA water immersion objective (10 0.30 NA, Nikon), a scan lens and a tube lens chosen to minimize dispersion imbalance between the two paths. The illumination light sheet is produced using a galvanometer mirror (-Scanner, GSI Lumonics) scanning the beams in the -direction at 500 Hz. Fluorescence from the illuminated plane is collected by a high NA water immersion objective (16 0 NA, Nikon), and three spectral channels are spatially split on an electron-multiplying charged coupled device camera (EMCCD, ion3 885, Andor) with a spectral image splitter (OptoSplit III, Cairn Research) equipped with two dichroic mirrors (FF484-FDi01 and FF560-FDi01, Semrock). The three objectives are mounted on translational stages for alignment. Sample is maintained in chamber filled with water solution and positioned from the top of the chamber with a combination of a motorized stage (MP285, Sutter Instrument) for translation in, and Z directions, and a rotation stage for rotation about the axis. Z-stack acquisition is obtained by moving the sample in the Z direction across the light sheet. All peripheral devices, including motorized half-wave plates, motorized delay line, shutter, galvanometer mirror, EMCCD camera, and motorized sample stage, are controlled using custom-written LabVIEW (National Instruments) software. Note on 2p-SPIM principle, spatial resolution and imaging depth The principle of light-sheet generation, the spatial resolution, and the imaging depth in 2p-SPIM compared to 1p-SPIM and 2P-LSM are discussed in previous works 1,4,8. In light-sheet microscopy, the lateral and axial resolutions of imaging are uncoupled (see Truong et al. 1 for a detailed discussion in the case of 2P-SPIM): the lateral resolution is determined by the detection objective NA and the axial resolution is usually dominated by the thickness of the light sheet. This thickness is controlled by the illumination objective NA. In our 2P-SPIM optical setup, the lateral resolution is however limited by the pixel size of the camera. The effective pixel size of 7 µm sets the lateral resolution to 1.34 µm (compared to < 0 µm lateral resolution expected with a 0 NA detection objective). In the illumination arms, we underfilled the back aperture of the objective to obtain 0.11-0.12 effective illumination NA and reach 2.7 and 3.6 µm of axial resolution using the TiS and the OPO, respectively. Spatial overlap between λ 1 and λ 2 light sheets The efficiency of wavelength mixing to generate two color two-photon excited fluorescence (2C- 2PEF) is governed by the spatio-temporal overlap between the two pulse trains at λ 1 and λ 2. The parameters describing beam spatial overlap, the effects of chromatic aberrations and the comparison between 2P-LSM and 2P-SPIM are presented in Supplementary Fig. 2. After precise beam alignment, the spatial overlap between the foci of two pulse trains at different wavelengths is mainly affected by the chromatic aberrations of the illumination objective. The axial mismatch between foci is due to the axial chromatic aberrations. These aberrations do not depend on the scanning angle and can be pre-compensated with static optical elements to adjust the beam divergence (telescopes in our case). However, any residual axial chromatic aberration has a stronger effect on wavelength mixing in 2P- LSM than in 2P-SPIM due to larger excitation NA (Supplementary Fig. 2b). The lateral foci mismatch is due to lateral chromatic aberrations. These aberrations are generally difficult to compensate in a scanning geometry since their amount depends on the scanning angle. However, wavelength mixing is less sensitive to lateral chromatic aberration in 2P-SPIM compared to 2P-LSM for the following reasons: (i) for the same axial resolution the excitation beams are larger in 2P-SPIM

Mahou et al Supplementary Information Page 15 / 16 due to lower excitation NA, (ii) for similar field of view extension in the scanning direction the scanning angles are smaller in 2P-SPIM (Supplementary Fig. 2c). Pulse synchronization in bidirectional 2P-SPIM Pulse synchronization was achieved simultaneously in both illumination arms of the 2P-SPIM setup by matching the optics dispersion between both arms, and by using a single delay line placed before the microscope. The two arms had a different length and each one consists of two achromatic doublets (scan lens and tube lens) and a 10 0.30 NA illumination objective (Supplementary Fig. 1b). Pulse synchronization was then achieved simultaneously in both arms by equalling the dispersion introduced by the scan and tube lenses, since they are the only different elements (Supplementary Fig. 3). For each lens, the differential time delay τ 1/2 accumulated by pulse trains at λ 1 and λ 2 is given by: N τ 1/2 = L k n gk (λ 2 ) n gk (λ 1 ) k=1 Where N is the number of optical elements making up the lens, L k the thickness of the k th element and n gk λ 1/2 the group refractive index of the k th element at wavelength λ 1/2. We used the following achromatic doublets (Thorlabs): illumination arm A, scan lens AC-508-150-B, tube lens AC-508-150- B, τ 820/1,100 nm = 788 + 788 fs; illumination arm B, scan lens AC-508-100-B, tube lens AC-508-100-C, τ 820/1,100 nm = 649 + 988 fs (the chromatic dispersion of materials was obtained from http://refractiveindex.info/). In this configuration, the differential time delay between pulse trains is expected to differ for the two arms by τ B τ A = 61 fs. This value is in good agreement with experiment τ B τ A = 51 ± 3 fs (Supplementary Fig. 3b) and is small enough compared to the measured width of the intensity cross-correlation between the two pulse trains τ FWHM = (τ 2 1 + τ 2 2 ) 1/2 = 400 ± 7 fs to efficiently obtain 2C-2PEF with both illumination arms simultaneously (Supplementary Fig. 3e). Fly embryo imaging We used a transgenic line of Drosophila melanogaster exhibiting a ubiquitous expression of red (RFP stands for mrfp1) and green (GFP stands for egfp with F64L, S65T mutations) fluorescent proteins (gift from A. McMahon and A. Stathopoulos at the California Institute of Technology). Fusion protein between histone 2A and the red fluorescent protein (such as in Bloomington stock center strain #23650, http://flystocks.bio.indiana.edu/reports/23650.html) provides red labeling of nuclei (H2A- RFP). In addition, a fusion protein between mouse lymphocyte marker mcd8 and the green fluorescent protein (mcd8-gfp) labels cell membranes 9. Fly embryos were collected, staged, and dechorionated following standard procedure 10. They were glued either on a glass cover slip for 2P- LSM imaging (as in 10 ) or on a glass capillary tube for 2P-SPIM imaging (as in 1 ). Embryos were imaged during gastrulation in water at 21 C using λ 1 = 820 nm and λ 2 = 1,175 nm for exciting endogeneous fluorescence (yolk and vitelline membrane) and RFP respectively. Spatio-temporal overlap of pulsed trains at λ 1 and λ 2 creates a mixed wavelength excitation at λ 3 = 965 nm to excite GFP. Comparison of photobleaching rate using multicolor 2P-SPIM and 2P-LSM Photobleaching of GFP and RFP is investigated by imaging mcd8-gfp and H2A-RFP expressing fly embryos with 2P-LSM or 2P-SPIM (Fig. 1d and Supplementary Fig. 4) using acquisition parameters listed in Supplementary Table 1. 2P-LSM imaging was performed using the multicolor 2P-LSM setup described in 2. Zebrafish embryonic heart imaging Zebrafish (Danio rerio) embryos with triple fluorescent labeling was obtained by crossing the transgenic lines Tg(myl7:eGFP) 11 and Tg(gata1:DsRed) 12 (providing green and red labeling of myocardial and red blood cells, respectively) and by injecting on cell stage embryos with H2B-CFP mrna (CFP stands for Cerulean, mainly providing blue labeling of pericardial cells in our experiments). H2B-CFP mrna (a kind gift from S.G. Megason) was prepared using mrna express mrna synthesis kit (System Bioscience) and injected with a nanoject II (Drummond Scientific). The

Mahou et al Supplementary Information Page 16 / 16 embryos were raised at 28 C in the dark and treated with 1-phenyl-2-thiourea (PTU) (Sigma Aldrich) at 10 hpf to inhibit pigment formation. For 2p-SPIM imaging, embryos were anesthetized with 0.01% (100 mg/l) Tricaine (Sigma Aldrich) solution and embedded in 1% (10 g/l) Low Melting Point agarose (Sigma Aldrich) as previously described 1. Heart imaging conditions used for Figure 1e-f, Supplementary Fig. 5, and Supplementary Videos 1-2 are listed in Table 2. Embryonic beating hearts were imaged using monodirectional illumination 2P-SPIM at λ 1 = 840 nm (from the TiS laser) and λ 2 = 1,090 nm (from the OPO), for exciting CFP and DsRed respectively. Spatio-temporal overlap of the two light-sheets at λ 1 and λ 2 creates a mixed wavelength excitation at λ 3 = 950nm to excite GFP. We used effective numerical apertures of NA = 0.11 and NA = 0.12 for the TiS and OPO beams, respectively, corresponding to a filling of one third of the excitation objective. Axial resolutions of 2.7 µm and 3.6 µm for the TiS and OPO light sheets, respectively correspond to a field of view along -direction at full width half maximum equal to 110 µm and 140 µm for the TiS and OPO beam, respectively. Time-lapse series of 170 images were recorded at 48 or 85 frames per second at each Z- depth and every 2.0 µm in the Z-direction. To reconstruct heart movement in 3D, fast time-lapses were synchronized using Cardiac Image-Sequence Synchronization Software (University of California at Santa Barbara) from Michael Liebling 5 (Figure 1e). 3-color images were then spatially overlapped and signal was unmixed using linear unximing from Fiji software 13. 3D image reconstruction, contrast adjustment, 3D-cell segmentation and tracking (Figure 1f-g, and Supplementary Videos 1-2) were performed using Imaris software (Bitplane). All experiments performed with zebrafish complied with the European directive 2010/63/UE and IGBMC guidelines validated by the regional committee of ethics. Z directions Throughout this article, the spatial Cartesian coordinate system Z refers to the geometry of the 2P- SPIM optical setup as shown in Fig. 1b., and Z axes correspond to illumination, scanning and detection axes, respectively. The illuminating beams propagate along the axis and are scanned along the axis to generate an illuminating light-sheet in the - plane. Fluorescence detection is performed along the Z axis. The camera records images in the - plane and 3D-stacks are obtained by moving the sample in the Z direction. When comparing 2P-SPIM with 2P-LSM (Supplementary Fig. 2), Z remains the detection axis and - the image plane in 2P-LSM. The terms lateral and axial resolution refer to spatial resolution in - and Z directions, respectively. Finally, lateral and axial chromatic aberrations investigated in Supplementary Fig. 2 are viewed from the illumination axis, meaning that lateral and axial aberrations are along - and Z directions in 2P-LSM and along the and directions in 2P-SPIM. Supplementary References 6. Drobizhev, M., Makarov, N.S., Tillo, S.E., Hughes, T.E. & Rebane, A., Nature Methods 8 (5), 393 (2011). 7. Kopp, R., Schwerte, T. & Pelster, B., Journal of Experimental Biology 208 (11), 2123 (2005). 8. Supatto, W., Truong, T.V., Debarre, D. & Beaurepaire, E., Current Opinion in Genetics & Development 21 (5), 538 (2011). 9. Lee, T. & Luo, L.Q., Neuron 22 (3), 451 (1999). 10. Supatto, W., McMahon, A., Fraser, S.E. & Stathopoulos, A., Nature Protocols 4 (10), 1397 (2009). 11. Huang, C.J., Tu, C.T., Hsiao, C.D., Hsieh, F.J. & Tsai, H.J., Developmental Dynamics 228 (1), 30 (2003). 12. Traver, D. et al., Nature Immunology 4 (12), 1238 (2003). 13. Schindelin, J. et al., Nature Methods 9 (7), 676 (2012).