Modular multimodal platform for classical and high throughput light sheet microscopy

Light-sheet fluorescence microscopy (LSFM) has become an important tool for biological and biomedical research. Although several illumination and detection strategies have been developed, the sample mounting still represents a cumbersome procedure as this is highly dependent on the type of sample and often this might be time consuming. This prevents the use of LSFM in other promising applications in which a fast and straightforward sample-mounting procedure and imaging are essential. These include the high-throughput research fields, e.g. in drug screenings and toxicology studies. Here we present a new imaging paradigm for LSFM, which exploits modularity to offer multimodal imaging and straightforward sample mounting strategy, enhancing the flexibility and throughput of the system. We describe its implementation in which the sample can be imaged either as in any classical configuration, as it flows through the light-sheet using a fluidic approach, or a combination of both. We also evaluate its ability to image a variety of samples, from zebrafish embryos and larvae to 3D complex cell cultures.

switch between two directions of the light sheet, parallel or perpendicular to the optical table, which is essential for the multimodality of the system. This can be also achieved in SPIM configuration, although in a manual manner, since cylindrical lenses are mounted on a manual rotational mount (CRM1). Moreover, both illumination and detection can be performed either simultanously either sequentially, multiplying the number of possible configurations.
It is important to notice that, due to the geometry of the system, depending whether if we generate the DSLM light-sheet parallel or perpendicular to the table, the dual side illumination will be synchronous or in opposite directions. This has a capital importance if we aim to implement the line confocal detection that will allow a reduced level of background signal in our images. In confocal line scanning, the laser beam is conjugated to the rolling shutter of the sCMOS camera, which acts as scanning slit. Hamamatsu cameras offer two rolling shutter modes. For fast scanning, reaching 100 fps full frame, the chip is divided in two and read out from center pixels towards periphery. Another mode, the light sheet mode, read outs the full chip in a single direction, thus reducing the overall speed to 33 fps. With our system, the first method could be implanted on the Classic/Flow LSFM mode, while the second could be applied to Hybrid LSFM mode. This will allow take full advantage of our systems capabilities.
The values reported in Supplementary Fig. 2c-d were obtained as follow. For each aperture diameter (R1 to R5), an image of the focused beam was acquired. The Gaussian beam appears propagating along the horizontal direction of the image. The FWHM beam diameter was calculated by Gaussian fitting the intensity profile of the beam (perpendicularly to the propagation direction) at its waist. The FWHM diameter, representing the LS thickness, was calculated from the standard deviation of the Gaussian fit as d waist = FWHM waist = 2.355 σ. From this value, the diameter of the beam at the Rayleigh range was calculated as d rayleigh = √2 • d waist .
Through a custom-made FIJI macro, the intensity profile of each column of the image was extracted and sent to the FIJI built-in Gaussian fit, retrieving for each column the FWHM of the Gaussian peak as described above. This array of values represents the FWHM beam diameter as function of the propagating distance. The number of columns presenting values smaller than d rayleigh were counted and converted from pixel to micrometers, delivering the illumination FoV.  For the upper part we have designed two variants of the detection scheme ( Supplementary Fig. 3(bc)). The chosen one will depend on the objective's brand used, since Nikon and Olympus present different parfocal distances and different tube lenses, of 200mm and 180 mm focal length respectively. When low magnification objectives are used, the sCMOS chip can be optically split in two parts using a knife-edge prism (Thorlabs, MRAK25-P01) inserted in the central cage cube (Supplementary Fig. 3(f)), so that each half visualizes simultaneously one side of the sample. For higher magnifications, alternate visualization of each side is performed with an Arduino controlled motorized mirror (Thorlabs PFR10-P01 on a Nanotec, L4018S1204-M6 stepper motor) ( Supplementary Fig. 3(g)).
Alternatively, detection can be performed in an up-right configuration ( Supplementary Fig.3 (e)) using water dipping objectives. An achromatic doublet with focal distance of 200 mm (AC254-200-A-ML) is used to form the image onto the camera chip. This module can be easily attached to the central cage cube of the upper module as shown in Supplementary Fig 3(a), while removing the prism or motorized mirror.
Depending on the chosen modality, sample scanning is performed through different motors. When samples are mounted in the classical configuration, i.e. in agarose filled capillaries, the capillar is attached to the sample scanning system, composed of a linear stepper motor stage (Thorlabs, LNR25ZFS) for sample translation across the light-sheet plane and an Arduino controlled stepper motor (Nanotec, L4018S1204-M6) for sample rotation. When the Flow LSFM mode is used in order to increase throughput of the system, samples are loaded into FEP tubes and transported towards the detection objective field of view using a syringe pump (Tecan, Cavro Centris). In both cases the lightsheet is created perpendicular to the optical table. For the Hybrid LSFM mode the scanning of the sample is performed by translation of the chamber with a motor (PI M-501.1DG) through a fixed horizontal light sheet plane, or through a synchronized movement of the light-sheet (with a galvo) and the detection objective lens mounted on a piezo motor (PiezoConcept PiFoc), keeping the sample static.

Supplementary Figure 4: High-Throughput Imaging Chamber with Rotation
One of the main advantages given by the Flexi-SPIM setup is the straightforward sample mounting in the Hybrid configuration. As described, in this operational mode the sample is loaded into a FEP tube, which in turn is connected to a programmable syringe pump. This pump aspires the sample in a controlled manner until it reaches the field of view of the vertical detection objective. Here the imaging process takes place, with the possibility to scan the sample in the vertical direction. An imaging chamber is needed in order to: (i) maintain steady the FEP tube during the scanning; (ii) contain the water needed for the water-dipping detection objective; (iii) permit the rotation along the longitudinal axis of the sample; (iv) provide the system with quasi-brightfield imaging capability.
To accomplish these tasks, we designed and 3D printed a custom imaging chamber, and additional components that permits sample rotation and water-sealing. The 3D printed PLA model is the core part. It gives the structure to the entire HT chamber; it contains the imaging medium (water-based), and prevents its leaking. Centred on its longitudinal axis, a channel is present, through which the rotation FEP is placed. The PLA model features an almost-symmetric geometry: the central compartment contains the medium, and the detection objective is placed on its top. Its lateral walls are windows, glued with 22x22mm coverslips, to permit the illuminating lightsheet enter the chamber. Under this section, the model offers a void space for the insertion of a blue LED, which provides the illumination for the brightfield imaging. This LED is assured to the chamber's centre thanks to a 3D printed part, complementary to the void cavity. Light is delivered to the water-compartment thanks to a circular coverslip, glued with the 2-components UHU Endfest glue. Right outside the water-compartment and following the longitudinal channel, two small void cubicles contain some grease (High Vacuum Grease, DOW CORNING) which prevents the water leakage and helps the motion of the rotation FEP tube in the channel. At the opposite sides of the model, two vertical panels offer the possibility to attach and screw the two stepper motors (L4118S1404-M6X1 -NEMA 17, Nanotec Electronic GmbH & Co. KG). These motors offer a threaded (M6x1) hollow shaft and are driven by an Arduino-based controller, which gives the user the possibility to digitally control them. A custom-designed and internally machined part, called RotationScrew, is screwed to the rotator shaft of each stepper. The rotation FEP is fixed to it. Another mechanical part, called MotorConnector, is also internally produced and connected through 4x M3 screws to the motor's body. The MotorConnector offers a centred cavity where two o-rings (sized 3x3 and 4x3) and one bearing ball (623ZZ, ID=3mm, OD=10mm) are glued with the 2-components UHU Schnellfest glue. In the 3mm diameter remaining void cylinder, the tip of the RotationScrew is inserted. Both the MotorConnector and the RotationScrew feature a channel (ID 1.6mm) along the longitudinal direction. Finally, the FEP tube, carefully cut in three different parts (loading, rotation, and aspiring) is inserted in the central channel of the system. The first tip of the loading tube is reserved for the insertion of the sample. The other end is inserted in the left sided MotorConnector, for a length of 5mm, and fixed with the two retaining screws. This position corresponds to the interface between loading and rotation tube. Symmetrically, the first end of the aspiring tube is fixed with retaining screws in the right sided MotorConnector, and the other end is connected to the syringe pump. Between them, the rotation tube is present. It is fixed on both sides with retaining screws into the two RotationScrew components. From each RotationScrew, the rotation tube exceeds for a length of 10mm. The two tips of the rotation tube are in direct contact, inside the MotorConnector, with the loading and aspiring tubes. The Flexi-SPIM setup features different hardware components that, thanks to a single LabVIEW-based homemade software, work in synergy. In order to interact with these components, we use an Arduino UNO board, which is connected via USB to our workstation, and is integrated in the LabVIEW software through the LIFA (LabVIEW Interface for Arduino) interface, provided by National Instrument. A principle schematic of the connections is shown in Supplementary Fig. 8(a). The front panel layout of the controller is divided in 4 different areas, respect to the specific function and devices to be controlled, as shown in Supplementary Fig. 8(b). These devices are: -1x stepper motor for sample rotation for Classic LSFM mode or tunable filter in Raman LSFM mode.
-1x stepper motor for the alternate detection in Classic LSFM mode -2x servomotors used as shutters for the alternate-illumination configuration.
-2x stepper motors and 1x LED of HT Chamber for Hybrid LSFM mode.
-2x stepper motors, 1x servomotor and 2x switches that act as contact sensors for the multi-well plate reader platform.
The Arduino UNO board is the central part of the controller. Through its I/O pins, it provides the digital control of the devices. For a general stepper motor however, an additional driver is required, as the Arduino Board cannot provide the needed power supply. We choose the Big Easy Driver v1.2 ( Supplementary Fig. 8(c)) from by Brian Schmalz (Schmalz Haus LLC).
The two motors relative to the HT Chamber share the same Arduino pins for ENABLE, STEP, and DIR. This assures that the two motors receive the same signal at the same time, moving synchronously. However, the connections between the two coils of the motors and the A and B pairs of pins of the drivers are inverted between each other. In this way, the two motors move in opposite directions, as needed for the HT Chamber. The stepper motor drivers relative to the XYZ platform only share the ENABLE pin, as they supply two independent motors. During our experiment we always used the 1/16 micro-step resolution, which permits a smoother movement of the motor.
While the servomotors signal pins are directly connected to the output pins of the Arduino board, the LED is connected to a potentiometer and a MOSFET. Through the potentiometer (10 KΩ) we can regulate the correct intensity of the LED, while the transistor closes the circuit, and then powers the LED, when the GATE pin is triggered, through a digital output pin of the Arduino. Two Arduino digital pins are also used as input pins, to detect the status change of the switches implemented in the XYZ Platform, which indicate the Home position of the Platform. A resuming table of the Arduino pins needed is shown in Supplementary Fig. 8(e). Additionally, some pins of the Arduino are left free, which could be further used for future development and integration (e.g. temperature sensor).
Thanks to the LIFA (LabVIEW Interface For Arduino) code, developed by National Instruments, we can communicate with the Arduino board through LabVIEW and dedicated VIs (Virtual Instruments). In our application we need only digital signals, thus we need to use some analogue pins of the Arduino board as digital pins. Although this is possible when programming the Arduino through its own textbased code, in LabVIEW we need to modify a default setting of the "Check For Pin Out Of Range" VI of the LIFA interface. In the "Uno, Default" case, "Digital" sub-case, changing the Num Pins constant to a value of 20 will let the usual analogue pins work as digital, as shown in Supplementary Fig. 8(f We implemented a multiwell plate reader (MPR) system, in order to automatize the loading procedure in the Flow LSFM and Hybrid LSFM modes. These imaging modalities, instead of using low melting point agarose, permits to bring the sample under the imaged field of view along a FEP tube, connected to a programmable syringe pump. The main aim of the MPR is to automatize the loading process, thus increasing the number of samples imaged, for high-throughput application. It consists of 3D printed components connected through metallic rails. Two stepper motors provide the possibility to move the translational part in the horizontal (xy) plane, while a servomotor displaces in the vertical direction (z) the tip of the FEP tube. On the base of the platform, a 96-multiwell plate, commonly used in biological laboratories, is placed and fixed. The MPR is connected to the Arduino-based controller, which provides the power and signals to the motors, reads the two switches that sense the "home" position of the platform, and permits the customization and automation of the protocols. A custom-made LabVIEW interface allows easy control of all the elements of the MPR, as well the automation of complex protocols. In this way, the MPR is able to move the loading head above the multiwell plate in order to reach the well of interest, place the tip of the FEP tube into the selected well, and let the pump aspire the sample. Additionally, a webcam is placed under the transparent platform base, in order to detect if the loading of the sample has been successful. If not, force the MPR to repeat the sample loading protocol.
The MPR can also being used to implement more complex or diversified protocols, like loading and, after imaging, sorting of the sample in specific well, or even direct and automated drug delivery to specific wells. Note that in the real application, when using the FEP tube mounting, no agarose embedding is needed. However, in order to maintain the beads steady during imaging, they were embedded in 0.5% lmpa. As shown, the introduction of the embedding medium between the objective and the sample (either lmpa or FEP tube) slightly degrade the resolutions with respect to the theoretical ones, but do not interfere with the capability of imaging at cellular levels. (d) created with PowerPoint Office 2019 (www.microsoft.com).

Supplementary Note 1: Control software and image processing tools
The Flexi-SPIM microscope offers a high automation degree, letting also non-microscope-specialists being able to use all its functions. This has been possible thanks to the home-made LabVIEW software, which controls almost all the aspects in a user friendly graphic interface.
The software is mainly divided in two layers: the first is the "LIVE" layer. Through the graphical front panel, the user can directly act on the different components. This layer include all the controls, and it is used to load the sample, position it, acquire snapshots, receive feedbacks on the status of the devices, define the light sheet parameters through the galvo mirrors, and regulate the best proprieties for the acquisition (e.g. exposure time, camera's ROI, laser power, etc.). Different functions are separated in different tabs, providing an intuitive organization. On the block diagram, the software provides the initialization of all the connected devices (camera, motors, Arduino controller, galvo mirrors, syringe pump, filter wheel, lasers) and then will execute two parallel loops. The first one is entirely dedicated to the camera settings, and to the displaying of the current detected image. The second one manages the actions to be taken in the case the user changes any of the parameters on the other devices, or the protocols. This permits to visualize the "live" image and permits to define the best parameters for the automated acquisition.
The second layer of the software consist of a set of so-called subVIs. For each of the operational modes of the microscopes at the hardware level in fact, a number of different protocols can be implemented, e.g. simultaneous or alternate illumination, simultaneous or alternate detection, translation or rotation of the sample during imaging, etc. Thus, each subVI corresponds to a different acquisition mode. Once the settings are ready for the imaging the first layer of the software, based on the parameters inserted by the user, passes the control of the hardware to one of these subVI, which performs the actual imaging procedure and save the acquired data and related metadata file.
The different acquisition modes could be resumed as follow: For Classic LSFM: We used the open source Fiji 1 software to process all the acquired data. For image fusion and stitching we used "Pairwise stitching" 2 plugin, for image registration "StackReg" 3 plugin, "Gaussian_Stack_Focuser" 4 plugin for bright-field stack focusing, and "TransformJ" 5 for image shearing.