Supplementary Information forTwo-color multiphoton in vivo imaging with a

femtosecond diamond Raman laserEvan P. Perillo1, Jeremy W. Jarrett1, Yen-Liang Liu1, Ahmed Hassan1, Daniel C.

Fernée1, John R. Goldak2, Andrei Bonteanu1, David J. Spence3, Hsin-Chih Yeh1 & Andrew K. Dunn1*

1Department of Biomedical Engineering, The University of Texas at Austin, TX, USA.2Department of Physics, The University of Texas at Austin, TX, USA.

3MQ Photonics, Department of Physics and Astronomy, Macquarie University, NSW 2109, Australia.

S1

Materials and Methods

Ytterbium fiber amplifier design. The amplifier consists of a 7 m long segment of double-clad ytterbium-doped large-mode-area optical fiber with a core size of 20 μm and cladding size of 400 μm (LMA-YDF-20/400-9M, Nufern). To avoid damage from high pulse energy, the fiber was custom endcapped with a 1.2 mm long coreless fiber and angle polished in FC:APC connectors (Coastal Connections, Ventura, CA). By coiling the fiber to a radius of less than 10 cm, the higher order modes of the fiber are attenuated and single-mode output is enforced (Supplementary Figure S1). The fiber is pumped with 7.5 W by a wavelength-stabilized laser diode emitting at 976±0.1 nm (maximum available pump power of 12 W). The amplifier is seeded by a commercial oscillator emitting 80 fs pulses at 1050 nm (Origami-10, OneFive GmbH). The oscillator repetition rate (80 MHz) drives the repetition rate of the amplifier and diamond laser, in addition to defining the cavity length of the diamond laser. An optical isolator (IO-5-1050-HP, Thorlabs) placed between the amplifier and the seed laser blocks amplified spontaneous emission feedback into the seed source. After exiting the amplification fiber, the pulses are compressed with a two-grating two-pass geometry (G1, G2, in Figure 1) and the output is split using a Glan-type calcite polarizer, with the majority (2.5 W) going to pump the diamond laser, and the remainder (500 mW) going to the microscope to be used for imaging. A detailed build protocol (Supplementary Figures S10 & S11) and bill of materials (Supplementary Tables 1 & 2) are also provided for both lasers.

Microscope setup. The imaging experiments were conducted on a custom built upright microscope consisting of, the custom dual output laser system, XY galvanometer scanning mirrors (6125H, Cambridge Tech.), and NIR-coated water-immersion objective (25×, 1.0 NA, XLPN25XSVMP, Olympus). The fiber amplifier and diamond Raman laser were combined onto a single optical axis with a dichroic mirror (DMSP1180, Thorlabs). The fiber amplifier was synchronized with the diamond Raman laser through a mirror delay line with motorized adjustable position (PT1-Z8, Thorlabs). The laser powers were independently controlled using variable neutral density filters (0.0-2.0 ND), and their collimation was adjusted with telescope lens pairs. Both beam polarizations were linear. The excitation wavelengths are λ1 = 1055 nm and λ2 = 1240 nm. The virtual 2C2P excitation wavelength is λ3 = 2( λ1

-1 + λ2-1)-1 = 1140 nm. A Plössl design scan lens (40 mm effective focal

length), and tube lens (200 mm) relay the image formed from the scan mirrors to the objective back aperture. Fluorescence light was epi-collected by GaAsP photomultiplier tubes (H10770PB-40, Hamamatsu Photonics) in a non-descanned configuration. The collected light was then split from excitation with a long pass dichroic mirror (FF775-Di01-52x58, Semrock). The emission light was further split into two channels with a long pass dichroic mirror with edge at 570 nm (570dcxr, Chroma). For Texas Red, a 610 nm bandpass filter was used (HQ/610/75M, Chroma). For Hoechst 33342, a 457 nm band pass filter was used (FF01-457/50-25, Semrock). For all other fluorophores, a 675 nm bandpass filter was used (FF01-675/67-25, Semrock). Acquisition was performed with a custom written software in LabVIEW, with a multifunction DAQ (PCI-6353, National Instruments).

Two-Color Alignment Procedure. For optimal two-color excitation, the two laser sources must be well co-aligned both spatially and temporally. Temporal alignment can be performed during imaging, by optimizing the 2P signal while changing the distance of the fiber amplifier delay line (Figure 4a).

S2

The spatial alignment is performed in a similar way to a standard two-photon microscope 1, by illuminating the back aperture with the laser source. However, there is added complexity involved with the relative alignment of both excitation sources. The alignment begins with fiber amplifier being aligned to the objective back aperture based on standard protocol1. Next, two reference pinholes are placed, one just before the scan mirrors and another just after the lasers are combined with the dichroic (DM1 in Figure 4a). The diamond laser is then aligned through the two pinholes, and should be nominally close in co-alignment relative to the first laser. Once the reference pinholes are positioned, they can be used as a guide for daily alignment. Once, coarse alignment in achieved and both lasers are illuminating the objective back aperture, we place a GaAsP photodiode (G1116, Hamamatsu) at the objective focus to measure two-photon induced photo-voltaic current. The induced current is then amplified through a transimpedance amplifier (SR570, Stanford Research Systems) and detected on an oscilloscope. The photodiode signal measures the two-photon excitation contributions from both lasers at the objective focus. By maximizing the signal from the photodiode, we can ensure that the optimal 2C2P overlap occurs at the objective focus. Typically we optimize several parameters, position and angle of diamond laser (by walking the beam with mirror pair), diamond laser collimation (with a telescope assembly), and the temporal overlap (fiber amplifier delay line). The photodiode alignment procedure is done before every day of experiments and takes approximately 5 minutes to perform. Although spatial overlap of the two lasers can be performed through image analysis (i.e. observing the spatial shift in diffraction limited bead images), we find that the image-based analysis is less accurate, more time consuming, and prone to photo-bleaching.

Spheroid preparation. The MCF-10A cell cultures were kept in humidified atmosphere with 5% CO2 at 37ºC. The transient transfection of MCF-10A cell line was conducted with the plasmid of tdKatushka2-VASP-5 (Addgene plasmid # 56049)2 using Lipofectamine® 3000 transfection reagent (L30000008, Thermo Fisher Sceintific). After being transfected with tdKatushka2-VASP-5, the MCF-10A cells were initiated to form spheroids by following the protocol developed by Kunz-Schughart’s group3. In brief, single suspensions were prepared by mild enzymatic dissociation using a 0.25% trypsin/EDTA solution (25200-056, Thermo Fisher Scientific), and then cells (10,000 cells per well) were seeded into agarose-coated 96-well plates (130188, Thermo Fisher Scientific). The plates were incubated for 96 hours in humidified atmosphere with 5% CO 2 at 37ºC. Agarose coating was conducted by filling each well with 50 µL DMEM sterilized agarose solution (1.5% by weight, A9539-100G, Sigma-Aldrich) (Supplementary Figure S4).

Animal preparation. All animal procedures were approved by the University of Texas at Austin Institutional Animal Care and Use Committee (protocol number AUP-2015-00011) and were under regulations consistent with the Guide for the Care and Use of Laboratory Animals, The Public Health Service Policy on Humane Care and Use of Laboratory Animals, and the Animal Welfare Act and Animal Welfare Regulations. Surgical instruments and artificial cerebral spinal fluid were sterilized via autoclave immediately before surgery. Mice (C57 males, 22 to 24 g, Charles River) were stabilized by a stereotaxic frame (Narishige, Narishige Scientific Instrument Lab, Tokyo, Japan), anesthetized with 3% isoflurane delivered through a nose-cone, and thermoregulated at 37ºC using a feedback temperature control system (FHC Bowdoin, ME, USA). Lidocaine was applied in small aliquots directly to the scalp as an anesthetic agent prior to removing the skin above the skull and resecting muscle tissue. Mice were then administered carprofen (5 mg/kg, subcutaneous), dexamethasone (2 mg/kg, intramuscular injection) to inhibit inflammation during drilling. Working a dental drill in controlled circular motions (Ideal Microdrill, 0.8 mm burr, Fine Science Tools, Foster City, CA, USA) the brain was exposed, leaving the dura intact. 5 mm round cranial windows (#1.5 cover slips, World Precision

S3

Instruments, Sarasota, FL, USA) were implanted and secured with dental cement for chronic in vivo imaging. A thin layer of cyanoacrylate was used to seal the porous dental cement. Animals were allowed to recover from anesthesia and monitored for a week before imaging. Immediately before imaging experiments, Texas Red dissolved in saline (100 µl, 5% w/v, ThermoFisher Scientific, Waltham, MA, USA) was delivered by retro-orbital injection using a fine syringe (29 gauge, Thomas Scientific, Swedesboro, NJ).

Spheroid experiments. All images were acquired with 512×512 pixel sampling at a frame rate of 1.34 Hz. For spheroid imaging, the excitation power used was approximately 10 mW for the diamond Raman laser, and 8 mW for the fiber amplifier. Five frame averaging was used, equating to a total acquisition time of 3.5 seconds per frame. Steps of 2 μm were taken along the z dimension over the 200 μm deep stack. The pixel size was 1.1×1.1 μm2 (Figure 5a), 0.28×0.28 μm2 (Figure 5b), and 0.57×0.57 μm2 (Supplementary Video 1).

Animal experiments. Image stacks were collected near the center of the cranial windows (bregma: -2.5 mm, interaural: 1.3 mm, lateral: 2.5 mm), avoiding large blood vessels at the surface. All images were acquired with 512×512 pixel sampling at a frame rate of 1.34 Hz. For surface level imaging, the excitation power was evenly distributed between fiber amplifier and diamond Raman laser (10 mW for each) equating to a total power of 20 mW at the objective back aperture. The power was adjusted at depth up to 150 mW total at the objective back aperture; the depth when max power was reached was approximately 400 μm. Averaging at 4 seconds per frame was used for surface level images, and 7.5 seconds per frame after 600 μm. The voxel size was 0.57×0.57×5 μm3 for all stacks recorded (Figure 6, Supplementary Figure S9, and Supplementary Video 2).

Data analysis and visualization. Data was saved into a binary file format directly from the LabVIEW acquisition software, and read into MATLAB (MathWorks) for processing. Images were filtered with a 2×2 pixel median filter and then normalized. A histogram contrast stretch function was employed (imadjust and stretchlim, Matlab) to force the top 1% of signal pixels to be saturated, and the bottom 1% to be zero. The stretch was performed on a per image basis. So images with lower SNR appear much noisier such as the delayed versus overlapped 2C2P in vivo images (Figure 6b). After stretching, the images are written as .tif file format. To visualize the 3D stacks, a rendering software (Avizo standard, VSG) was used.

Description of supplementary videos.

Video S1. Animation of two-color three-photon excitation signal versus delay time (τ). The sample is a spheroid of Hoechst 33342 stained MCF-10A cells. The excitation power was split between diamond laser (10 mW) and fiber amplifier (8 mW) for a total excitation power of 18 mW at the back aperture of the objective. Delay time, τ, is adjusted using a motorized mirror delay line in the fiber amplifier optical path. Steps of +30 fs are taken, which equates to a physical path length change of 9 μm. The scale bar is 40 μm.

Video S2. Visualization of a 3D image stack in a live mouse brain recorded using two-color two-photon excitation (τ = 0 fs). The excitation power was equally split between the fiber amplifier (λ1 = 1055 nm) and diamond laser (λ2 = 1240 nm). The total power at the surface was 20 mW, and the power was increased to a maximum of 150 mW at a depth of 400 μm, beyond which the maximum power was maintained. The box dimensions are 300×300×960 μm3.

S4

Supplementary Figure S1. Fiber amplifier beam quality. Fiber amplifier (λ = 1055 nm) beam waist diameter versus focus position, using a 300 mm focal length spherical lens. The beam quality was measured to be an average of M2 = 1.19, along x and y dimensions. Inset: spatial profile of the beam, the scale bar is 50 μm.

S5

Supplementary Figure S2. Laser characterization. a. Fiber Amplifier 1055 nm output power dependence on absorbed 976 nm pump power for a seed input power of 50 mW at 1050 nm. The amplification threshold is at 2 W of pump power. b. Diamond Raman laser output power at 1240 nm versus input pump power at 1055 nm (left axis). The residual 1055 nm power is also displayed (right axis). The lasing threshold is at 750 mW pump power. c. Average output power of the 1055 nm Fiber Amplifier versus time. d. Average output power of the 1240 nm diamond Raman laser when pumped at 2.5 W.

S6

Supplementary Figure S3. 2C2P excitation power ratio. Fluorescence signal of tdKatushka2 versus pulse synchronization delay time under 2C2P excitation for differing levels of Ф. Where Ф is the ratio of average powers of 1240 nm and 1055 nm excitation. Ф > 1, refers to higher 1240 nm average power. In this case of fluorophore and 2C excitation wavelength pair it is possible to adjust the signal modulation depth from 35% to 70% for increasing values of Ф. With 1240 nm excitation, tdKatushka2 has a near zero cross section, whereas at 1055 nm the cross section is almost 50% of the peak, so there will always signal due to direct 1C2P excitation. Although the excitation efficiency is not increased with high values of Ф, the increased modulation depth could be beneficial for multicolor experiments requiring rapid signal modulation of 2C2P while still maintaining direct 1C2P excitation.

S7

Supplementary Figure S4. Spheroid formation and initiation protocol for MCF-10A cells. a. Schematic of the preparation and imaging of tdKatushka2-expressing spheroids. First, The MCF-10A cells were transfected by the plasmid of tdKatushka2-VASP-5 (Addgene plasmid # 56049) and then transferred to an agarose-coated plate for spheroid formation (10,000 cells per well). After 4 days culture, the spheroids were fixed with 4% paraformaldehyde and embedded in a chambered coverglasses with 1.3% agarose for cell imaging. b. A bright field imaging of formed spheroid in liquid, overlaid with fluorescence imaging. Scale bar is 100 μm.

S8

Supplementary Figure S5. 2C2P excitation of far-red fluorescent proteins. Two-color two-photon excitation images of HeLa cells transiently expressing a. mKate2-tubulin (FP185, Evrogen) b. tdKatushka2-CENPB-N-222 (plasmid # 56033, Addgene), c. mCardinal-vimentin2 (plasmid # 56178, Addgene), d. dKatushka-cytERM (plasmid # 56027, Addgene), and e. mNeptune2-tubulin (plasmid # 56150, Addgene). The scale bar is 7.5 μm. f. Comparison of detected fluorescence signal from the cell images in a.-e. under both 1C2P and 2C2P excitation. The total power at the back aperture was kept constant with each excitation scheme. The average power ranged from to 4 mW to 16 mW depending on the brightness of the fluorophore and expression level in the cell. In all of the tested cells, 2C2P offers the highest signal level, on average 90% brighter than with 1C2P at λ=1055 nm. This brightness enhancement can be attributed to the higher 2P fluorescence cross section at the effective 2C2P wavelength of 1140 nm, compared with either of the 1C2P excitation wavelengths. The error bars are the standard deviation of the image signal.

S9

Supplementary Figure S6. Excitation dependency of tdKatushka2 and Hoechst 33342. Log-log plot for tdKatushka2 and Hoechst33342 under 1055 nm excitation. The slope of tdKatushka2 is 1.97 indicating a two-photon process, while the slope for Hoechst is 2.97 indicating a three-photon process.

S10

Supplementary Figure S7. Excitation dependency of Texas Red. Log-log plot of fluorescence signal versus laser power for excitation wavelengths of 1055 nm (orange) and 1240 nm (red). In both cases, the slope is 2.0, indicating a two-photon process.

S11

Supplementary Figure S8. Additional in vivo images. a. in vivo stack taken with 2C2P (reproduced from Figure 6 for comparison), The dotted blue line is the approximate depth limit of 1C2P (λ = 1055 nm). b. in vivo stack of the same region taken with 2C2P when the pulses are delayed by -600 fs relative to each other. c. in vivo stack of the same region taken with 1C2P excitation at 1055 nm and the equivalent average power level as a. and b.

S12

Supplementary Figure S9. Neuron in vivo images. a. in vivo stack in a live mouse using the fiber amplifier for 1C2P excitation (λ=1055 nm). Neurons are labelled with tdTomato through an adeno-associated virus injection. The injection causes local expression to all cell bodies within a region of the brain. b. Maximum intensity projection images of 10 μm thick slices at depths, 225 μm, 500 μm, 775 μm, and 950 μm, respectively.

S13

Supplementary Figure S10. Summary of build protocol for fiber amplifier.

1. Refer to Table 1 for a summary of the components used in the build, including ID descriptions. 2. Coupling the pump light [1] into the gain fiber cladding [2]. Using a 10 mm focal length lens [4]

collimate the light from the pump fiber [1] and couple into the cladding of the gain fiber [2] using a 18 mm focal length lens [5]. The magnification factor is 1.8 and pump light should illuminate a ~200 μm circle on the gain fiber end-face. The distance from pump lens to gain fiber lens should be ~150 mm or less. A fold mirror may be necessary to achieve optimal alignment.

3. Refine pump coupling [Optional]. Using a visible alignment laser (632 nm) coupled into the opposite end of the gain fiber [3], align the output of the visible light from the gain fiber [2] to the pump light [1] using an IR card. If the visible light and pump light are overlapped in two locations, then the pump coupling will be well aligned into the gain fiber.

4. Launch the seed light. Align the seed laser [12] through the isolator [10] and into the gain fiber input end [3] with aspheric lens [6]. Ensure the seed light enters the core of the gain fiber by co-aligning the input seed light with the output residual pump light from the cladding.

5. Refine Seed coupling. The seed output should be seen through the output end of the gain fiber [2]. The seed should be primarily in the single mode core of the fiber (a ~2mm spot should be visible on an IR card). Place a power meter far away from the output end of the collimator (>500 mm), slowly optimize the alignment of the seed into the input end of the gain fiber [3] by walking the beam using two fold mirrors and the fiber launch stage [9], periodically optimize the seed focus into the fiber during the process.

6. Set up pulse compressor [11]. Follow the guidelines to align a grating compressor from previous literature.4

7. Characterize. The pump power dependence, spectrum, and pulse width should match closely with the results obtained here, otherwise further optimization of pump and seed coupling is required.

S14

ID Descriptiona Supplier Manufacturer's Part # Price Qty.

[1] 18 W 967 nm Laser Diode Pump BWT Laser

K976AA2RN-18.00W$486.00 1

[2],[3] Yb-Doped Double Clad LMA Optical Fiber Nufern LMA-YDF-20/400-

9M$252.00 7

- FC/APC Fiber Endcap 1.3 mm

Coastal Connections CUSTOM

$414.00 1

- 30W TEC High Power LaserMount

Arroyo Instruments 264-BB-9W4

$895.00 1

- LaserSource, 20A/7V Driver

Arroyo Instruments 4320-20-07

$2,395.00 1

[4] f = 10 mm, Pump Lens ThorlabsAL1210-B

$183.00 1

[5],[6] f = 18 mm, Amplifer Lens Thorlabs

AL2018-B$222.00 2

[7],[8] 1000 nm Shortpass Dichroic Thorlabs

DMSP1000$261.00 2

[9] Fiber Launch Stage ThorlabsMAX355D

$1,729.00 1

[10] 1050 nm Isolator ThorlabsIO-5-1050-HP

$2,490.00 1

[11] 1000 lines/mm Trans. Gratings LightSmyth T-1000-1040-3212-

94$400.00 2

[12] Origami – 10 fs Fiber Laser

OneFive GmbH

OR/SM/80/10530/100/100/FS

$28,080.00 1

aThe majority of opto-mechanical components have been omitted and can be selected based on preference.

Supplementary Table 1. Select list of components used in fiber amplifier build.

S15

Supplementary Figure S11. Summary of build protocol for diamond Raman Laser.

1. Refer to Table 2 for a summary of the components used in the build, including ID descriptions. 2. Bring pump laser to a 1/e2 radius of 20 μm with three mode-matching lenses [1], [2], and [3]. The

approximate distances between the lenses are 124 mm between [1] and [2], 40 mm between [2] and [3], and 165 mm between [3] and diamond [6].

3. Insert first curved mirror [4] approximately 102 mm from focal position4. Insert second curved mirror [5] 208 mm from first mirror [4].5. Insert diamond [6] equidistant between both mirrors. Diamond should be mounted so that the 4 mm

edge is 35º from the horizontal plane of polarized pump light. Diamond should be placed slightly off normal incidence to avoid back reflection. A 3-axis translation stage is recommended for the diamond mount.

6. Insert the remaining fold mirrors [7], [8], [9], and the output coupler [10]. The total cavity length should correspond to the period of the pump source (~3.75 m for 80 MHz pulsed source). Cavity length should account for the refractive index of the 8 mm diamond slab. If space is not a concern, then two fold mirrors can be removed and the distance between the bottom path can be lengthened. The distance between the top and bottom arms of the cavity should be made as small as possible to reduce aberration from the curved mirrors.

7. Align the pump light through the cavity. The pump light exiting through the first curved mirror [4] after one round trip should overlap the pump light reflected by the backside of the mirror [4]. Overlap both pump beams and move the cavity distance with mirror [9]. When the cavity length is matched to the pump repetition (to within 10 μm), fringes can be seen in the overlapped pump beams. The cavity length should be measured carefully by ruler before attempting to find fringes. Align the pump beams with an IR card or IR viewer so that the fringes are perfectly circular.

8. Optimize cavity alignment until lasing is detected from the output coupler [10] on a power meter. Parameters to optimize include tip-tilt on [9] and [10], distance of mirror [9], and diamond [6] position in xyz. Lasing should begin occur near 750 mW of pump power.

S16

ID Description R at 1240 nm

T at 1050 nm Supplier Part # Price Qty.

[1] f= -50 mm Achromat Lens - 99.75% Thorlabs ACN254-

050-B $85.25 1

[2] f = 150 mm Achromat Lens - 99.75% Thorlabs AC254-150-

C $94.75 1

[3] f = 200 mm Achromat Lens - 99.75% Thorlabs AC254-200-

C $94.75 1

[4],[5] Curved Mirror ROC=200 mm 99% 93% ATF Films -

$800.00 2

[6] CVD Diamond 8×4×1 mm3 <0.5% - Element 6 155-104-

1456 $5,000.00 1

[7],[8],[9] Fold mirror 99.50% - Layertec 102023$469.00 3

[10] Output Coupler (Rmax @1060 nm) 92% <1% CVI -

$700.00 1

Supplementary Table 2. Select list of components used in diamond Raman laser build.

S17

Supplementary References

1. Rosenegger, D. G., Tran, C. H. T., Ledue, J., Zhou, N. & Gordon, G. R. A High Performance , Cost-Effective , Open-Source Microscope for Scanning Two-Photon Microscopy that Is Modular and Readily Adaptable. PLoS One 9, (2014).

2. Shcherbo, D. et al. Far-red fluorescent tags for protein imaging in living tissues. Biochem. J. 418, 567–574 (2009).

3. Friedrich, J., Seidel, C., Ebner, R. & Kunz-Schughart, L. A. Spheroid-based drug screen: considerations and practical approach. Nat. Protoc. 4, 309–24 (2009).

4. Miesak, E., Negres, R. & Florida, C. Alignment Procedure for a Dual Grating Pulse Compressor. 1, 8146–8147

S18