24
Ellen Adams Selective Plane Illumination Microscopy (SPIM) for Hydra Development Draft Project Report Advised by Eva-Maria Collins and Maggie Delano 1. Motivation: A challenge in biology is imaging live biological samples in 3D to capture larger-scale dynamic processes. These include whole-organism processes like development and regeneration, which involve dramatic rearrangements of cells and changes in cell chape. To image these morphological changes in biological samples, transgenic animals can be generated that have fluorescent proteins to label subcellular structures, cells, or tissues (Takeuchi et al., 1999; Wittlieb et al., 2006). This fluorescence allows for the visualization of the labelled sample and can be used to track cells and cell contents, as well as shape changes (Wittlieb et al., 2006; Hadjantonakis et al., 2004). One of the difficulties of imaging these fluorescent live samples is achieving good resolution in live samples due to the size of the specimen and the scattering properties of intact tissue (Huisken and Stainier, 2009). Due to these challenges, often live samples will instead be fixed and sectioned at various timepoints to achieve better visualization, however this approach does not lend itself to studying the continuous timeline of development. Selective Plane Illumination Microscopy (SPIM) allows for optical, instead of physical, sectioning for deeper penetration into the sample, to image live, intact embryos in 3D over time (Huisken and Stainier, 2009). SPIM has been used to study the development of embryos of the polyclad flatworm M. crozieri and fruit fly Drosophilia melanogaster (Girstmair et al., 2016; Schmied and Tomancak, 2016), as well as zebrafish embryo vasculature development over the course of up to five days (Daetwyler et al., 2019). In these studies, SPIM proved to be a useful tool for longer term continuous imaging of development in live embryos. We want to use this microscopy technique to visualize the development of Hydra, which are small freshwater cnidarians famous for their regenerative abilities. They regenerate from a small tissue piece or cell aggregate, and we aim to track cell movements throughout this process to follow the patterning of tissues that occurs. By applying the SPIM technique to studying regeneration and development in Hydra, we can gain insights into the dynamics of development and regeneration to gain a greater understanding of them. 2. Background: The challenge of imaging live tissues in 3D has been solved by current conventional microscopes. The most basic method is widefield microscopy, in which the entire sample is illuminated, as shown in Figure 1. The fluorescence produced by the sample is then all detected at once, making this a very fast method. However, since widefield microscopy involves illumination of the entire sample, fluorescence from areas around the focal plane are also

Selective Plane Illumination Microscopy (SPIM) for Hydra

  • Upload
    others

  • View
    4

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Selective Plane Illumination Microscopy (SPIM) for Hydra

Ellen Adams

Selective Plane Illumination Microscopy (SPIM) for Hydra Development

Draft Project Report

Advised by Eva-Maria Collins and Maggie Delano

1. Motivation:

A challenge in biology is imaging live biological samples in 3D to capture larger-scale dynamicprocesses. These include whole-organism processes like development and regeneration, whichinvolve dramatic rearrangements of cells and changes in cell chape. To image thesemorphological changes in biological samples, transgenic animals can be generated that havefluorescent proteins to label subcellular structures, cells, or tissues (Takeuchi et al., 1999;Wittlieb et al., 2006). This fluorescence allows for the visualization of the labelled sample andcan be used to track cells and cell contents, as well as shape changes (Wittlieb et al., 2006;Hadjantonakis et al., 2004). One of the difficulties of imaging these fluorescent live samples isachieving good resolution in live samples due to the size of the specimen and the scatteringproperties of intact tissue (Huisken and Stainier, 2009). Due to these challenges, often livesamples will instead be fixed and sectioned at various timepoints to achieve better visualization,however this approach does not lend itself to studying the continuous timeline of development.Selective Plane Illumination Microscopy (SPIM) allows for optical, instead of physical,sectioning for deeper penetration into the sample, to image live, intact embryos in 3D over time(Huisken and Stainier, 2009). SPIM has been used to study the development of embryos of thepolyclad flatworm M. crozieri and fruit fly Drosophilia melanogaster (Girstmair et al., 2016;Schmied and Tomancak, 2016), as well as zebrafish embryo vasculature development over thecourse of up to five days (Daetwyler et al., 2019). In these studies, SPIM proved to be a usefultool for longer term continuous imaging of development in live embryos. We want to use thismicroscopy technique to visualize the development of Hydra, which are small freshwatercnidarians famous for their regenerative abilities. They regenerate from a small tissue piece orcell aggregate, and we aim to track cell movements throughout this process to follow thepatterning of tissues that occurs. By applying the SPIM technique to studying regeneration anddevelopment in Hydra, we can gain insights into the dynamics of development and regenerationto gain a greater understanding of them.

2. Background:

The challenge of imaging live tissues in 3D has been solved by current conventionalmicroscopes. The most basic method is widefield microscopy, in which the entire sample isilluminated, as shown in Figure 1. The fluorescence produced by the sample is then all detectedat once, making this a very fast method. However, since widefield microscopy involvesillumination of the entire sample, fluorescence from areas around the focal plane are also

Page 2: Selective Plane Illumination Microscopy (SPIM) for Hydra

detected (What is Widefield Imaging? 2020). This produces out-of-focus light, which gives apoor signal-to-noise ratio (SNR).

Figure 1. Comparison of different microscopy illumination modalities (Krieger et al., 2015).

Another microscopy method was developed to improve the SNR of images: confocalmicroscopy. Confocal microscopy involves illumination of a defined spot at a specific depthwithin the sample, multiple of these points are shown in Figure 1. Detection of the samplefluorescence is only at the point of illumination, to reduce out-of-focus light. This illuminationpoint is scanned through the entire sample point by point, which is very time consuming,however the resulting images do have a better SNR (Selective Plane Illumination MicroscopySPIM or LSFM Overview 2020). This method therefore gives a good spatial resolution, at theprice of more time-consuming imaging.

SPIM is the ideal combination of these two methods, as it optimizes this tradeoff between fastimaging and good spatial resolution. Illumination is in the form of a thin light sheet sent at thesample, which illuminates only a plane of the sample at a time, allowing for good opticalsectioning (Selective Plane Illumination Microscopy SPIM or LSFM Overview 2020). Thefluorescence emitted by the illuminated plane of the sample is detected orthogonal to the plane ofillumination in a widefield manner, allowing for faster image collection than that of confocalmicroscopy (Girstmair et al., 2016). Additionally, compared to conventional widefieldillumination, which always illuminates the whole sample, SPIM reduces the amount of light sentat the sample, resulting in minimized photodamage and stress of the live sample (Huisken et al.,2004; Stelzer, 2015). The comparison between these three methods of 3D microscopy issummarized in Table 1.

Pros Cons

Widefield Fastest Poor SNRPoor Spatial Resolution

Widefield -~ Confocal

t t

Lightsheet detection

Page 3: Selective Plane Illumination Microscopy (SPIM) for Hydra

Confocal Good SNRBest Spatial Resolution

SlowPotential Photodamage

SPIM

FastGood SNR

Good Spatial ResolutionMinimal Photodamage

Advanced Image ProcessingSpecialized Application

Table 1. Comparison of Widefield, Confocal, and SPIM methods of microscopy.

SPIM results in a more specialized microscope that is customized for one application, so it is lessgeneralizable, and it is harder to switch between sample types, which is one downfall.Additionally, SPIM images require advance image processing to output a fused 3D image of thesample. However, using SPIM results in minimal photodamage, good optical sectioning, and fastacquisition rates, making it ideal for imaging 3D development in one type of sample.

SPIM works by sending a sheet of light to excite the sample, shown in blue below (Figure 2) thatis created using optics. The axis of detection is orthogonal to the illumination (green arrow) andconsists of a microscope objective lens and other widefield optics to image the sample onto acamera (Huisken and Stainier, 2009). This light sheet is set up to only illuminate the sample atthe focal plane of the detection objective (Girstmair et al., 2016).

Page 4: Selective Plane Illumination Microscopy (SPIM) for Hydra

Figure 2. Model of SPIM illumination and detection axes (Huisken and Stainier, 2009).

Utilizing OpenSPIM (https://openspim.org/), which is an open access platform for building andadapting SPIM technology, a version of a SPIM microscope can be constructed. Shown below isthe final setup achieved for the SPIM (Figure 3). The laser sits on a heatsink, and the path of thelaser is shown in blue, where it passes through mirrors and filters to illuminate the samplechamber as a sheet. The sample is then illuminated, and the detected fluorescence is imaged ontothe camera through an objective lens and tube lens. The 4D stage is used to position the samplein the sample chamber and to rotate the sample during imaging.

Figure 3. My constructed SPIM setup based on OpenSPIM guidelines. Important elements arelabelled including the laser, which is directed into the sample chamber, where the sample ismoved using the 4D-stage. The fluorescence from the sample is then detected though thetubelens and imaged by the camera.

To move the sample using the 4D stage, it is suspended in a 0.1-1.5% agarose gel that isstiff enough to keep the sample in place during imaging and allow for the gel to be moved withthe sample. Using the 4D stage, the sample in gel is both translated, to illuminate and collectimages from successive parts of the sample horizontally with the light sheet, and rotated tocollect images from all sides of the sample. This gives a large set of images that must be‘stitched’ together to give a 3D image of the sample for each timepoint. This can be compared tothe process of image reconstruction for computerized tomography (CT) scans. In a CT scan,multiple X-ray images are taken from different angles around a patient, and then are combined

Page 5: Selective Plane Illumination Microscopy (SPIM) for Hydra

using reconstruction algorithms (Willemink and Noël, 2019). In a similar manner, Fiji can beused for processing the data collected from a SPIM setup. This involves registration of the viewsusing beads and fusion of the data from different views into a single output image usingmulti-view deconvolution. The fused data can then be analyzed, and 3D renderings can be doneto better visualize the data.

Page 6: Selective Plane Illumination Microscopy (SPIM) for Hydra

3. Theory:

The two key aspects of SPIM are the physical microscope and the reconstruction of the 2Dimages captured by the camera into a 3D image of the sample. Therefore, I will first describe thetheory of the optics involved in the setup, and then discuss image reconstruction.

3.1 Optics:

The optics of the microscope are designed to take the outputted laser beam and convert it into alight sheet that is focused and thinnest at the sample, to give the best spatial resolution. This isdone using multiple components such as mirrors, lenses, and objectives (Figure 4).

Figure 4. Layout of my microscope components (a) showing the path the laser light takes in blue.Simplified diagram (b) with optical components and light paths. The illumination axis (blue)consists of light from the laser (L) that is reflected by mirrors (M), expanded by a telescope (T1),sent through a vertical slit (VS), cylindrical lens (C), and passed through another telescope (T2)to narrow the beam. This sheet of light is then sent through the illumination objective (IO) to thesample (S). Fluorescence from the sample (green line) is detected by the detection objective(DO), sent though filters (F), and passed through a tube lens (TL) before being imaged by thecamera (CMOS).

First, light from the laser is reflected by two mirrors to change the angle of the beam (M, Figure4b). Then light is passed through two lenses forming a telescope (T1, Figure 4b). Lenses work byrefracting a beam of parallel light onto a single point, the focus. The distance between this focusand the lens is the focal length, shown as f in Figure 5. Parallel light that enters at angle φ willshift the focus perpendicularly to the optical axis by a distance Δx. When two lenses withdifferent focus lengths, f and f’ are combined, they can be used as a telescope to change thediameter of the incoming beam of light, as seen in Figure 5. The first lens in the telescoperefracts the incoming parallel light onto the focus, and then this light can be passed through

a b

Page 7: Selective Plane Illumination Microscopy (SPIM) for Hydra

another lens to do the opposite: take the light put into the focus and collimate it to produceparallel light again. The telescope is used to increase or decrease the diameter of the beam oflight from D to D’. In this case, the light is expanded by a factor of M, which can be calculatedusing:

𝑀 = 𝐷'

𝐷 = 𝑓'𝑓

In the case of the first telescope in the SPIM setup (T1, Figure 4b), the first lens has a focallength of 25mm, and the second a focal length of 50mm, which expands the beam by a factor oftwo.

Figure 5. Diagram of telescope constructed from two lenses (OpenSPIM).

After the telescope, the beam is sent into the vertical slit (VS, Figure 4B). This is used to controlthe numerical aperture and thickness of the light sheet. As seen below in Figure 6, the verticalslit (VS) cuts off some amount of the incoming expanded beam. After the vertical slit, light ispassed through the cylindrical lens (C), which focuses the light into a horizontal sheet, which isthen imaged into the back focal plane of the illumination objective (IO) by another telescope(T2) that narrows the beam by a factor of ½ using the same two lenses as T1, but in reverseorder. This created light sheet is sent through the illumination objective (IO), which focuses thesheet in the z-direction so that it is thinnest when passing through the sample (S) to illuminate it.

Figure 6. Changes in light beam as sent though optical elements as observed from the side of themicroscope.

Page 8: Selective Plane Illumination Microscopy (SPIM) for Hydra

Sample fluorescence is then detected by the detection objective (DO), which converts light fromfluorescing points in the sample into a parallel beam of light. Points are located at positions x inthe sample along the illuminating sheet of light, which exit the detection objective underdifferent angles φ(x) (Figure 7). The tube lens then takes this parallel light and converts it backinto an image in its focal plane, where the camera is located. The magnification from this processcan be calculated as:

𝑀 = 𝑥'

𝑥 = 𝑓

𝑇𝐿

𝑓𝑂

Light passing between the objective and the tube lens is in ‘infinity space’, since the light isparallel, and optical elements like filters that are planar can be inserted at this point withoutdisturbing any optical properties of the microscope. Filters are used to filter out any light otherthan the fluorescence emitted by the sample (bandpass filter with center wavelength of 500nm),since the illumination laser light is at 488nm and emission is at 510nm. Additionally, a filter isused to block light from the illumination from being picked up by the camera (longpass filterwith cut-on wavelength of 500 nm).

Figure 7. Diagram of detection axis (OpenSPIM).

3.2 Reconstruction:

Once the camera has taken stacks of images from a defined number of angles, these images mustbe reconstructed into a 3D image of the entire sample. This involves taking the images that aretaken from different angles and determining how they are related to one another so that they canbe fused together. The transformation between different views can be found by using stationarymarkers within the images that are visible from different angles. In the case of SPIM data, wesuspend 0.5-1 µm fluorescent beads in the gel with the sample to function as these stationarymarkers. Beads are detected in an image using a smoothed 3D LaPlace filter , which∇2

accurately detects beads while limiting high frequency noise (Preibisch et al., 2010). The of∇2

an image is approximated as the difference of two Gaussian convolutions of the image (DoGfilter) with standard deviations, , of 1.4px and 1.8px (Preibisch et al., 2010). Beads are locatedσby finding local minima in a 3×3×3 neighborhood in the of the image, which represent∇2

intensity maxima in the original image. The locations of these intensity maxima are estimated byfitting a 3D quadratic function to the neighborhood (Preibisch et al., 2010). This allows for the

Page 9: Selective Plane Illumination Microscopy (SPIM) for Hydra

identification of all beads in an image, however it also over-segments the image, allowing forstructures such as corners and edges within the sample to be detected. These detections do nothinder the registration, since they are filtered out by local descriptor matching later, and the onlydetections that are repeatable between views are the actual beads.

Once the beads are identified, then the different views with the identified beads have to beregistered. This is done by finding the corresponding bead pairs between two views. Atranslation and rotation invariant local geometric descriptor was developed that identifies eachbead based on the unique set of nearby beads (Preibisch et al., 2010). The local geometricdescriptor is defined by the location of a bead’s 3 nearest neighbors in space (Figure 8c), orderedby distance to that bead. This descriptor made up of four beads is used to define a localcoordinate system, and since all remaining bead coordinates not used for the local coordinatesystem become rotation invariant, descriptors can be compared efficiently using kd-trees(Preibisch et al., 2010). After this, a set of identified bead pairs is created, which can be used todefine an affine transformation that maps one view to the other. The transformation is found byleast square bead correspondence displacement.

Figure 8. Bead-based registration process. (a) Stacks of images are taken from multiple views ofthe sample, which must be registered. (c) Example descriptors made up of four beads (centralbead and the three nearest neighbors) used to find corresponding beads between different views.(d) Visualization of the global optimization process with displacement of corresponding beaddescriptors color-coded from red (maximum displacement) to green (minimal displacement).Three iteration numbers (0, 10 and 283) are shown with average displacement for each one(Preibisch et al., 2010).

After the transformation is found between two views, registration of additional views requiresgroupwise optimization, which is solved using iterative optimization, as seen in Figure 8d. Foreach iteration, the optimal transformation for each view, relative to the configuration of all other

d 1

10 -"'

Iteration O lteralion 10 Iteration 283 <1

(r = 386.4 pixels} (t = 19.4 pixels) (r = 1.1 pixels)

Page 10: Selective Plane Illumination Microscopy (SPIM) for Hydra

views, is estimated and applied to all beads in that view. The optimization is complete when theoverall bead correspondence displacement converges, as in the final green image in Figure 8d.

Once all the views are registered using beads, the views can be combined to create a single 3Dimage. A fusion algorithm using Gaussian Filters is utilized to approximate the imageinformation at each pixel in the contributing views (Preibisch et al., 2010).

4. Realization:

4.1 Physical Microscope:

The microscope was designed and constructed utilizing OpenSPIM (https://openspim.org/) as aguide. Because OpenSPIM provides a detailed description of all the parts and the assembly, Iwill only provide a brief overview here, highlighting custom approaches. A list of all parts can befound in Appendix II, including purchased and self-made parts. All self-made parts can be seenbelow in Table 2. These include 3D-printed parts, which were designed and modified inRhinoceros 3D. The majority were 3D printed either from PLA using an FDM machine, or wereprinted form a resin using an SLA machine. All parts that needed to be clear, including thesample chamber were printed from a clear resin. Additionally, a few parts were machined out ofaluminum including the heat sink.

PART: MATERIAL: NUMBER:

Heat sink Aluminum 1

RC1 vertical slit stilt PLA 1

RC1 1/2" lens stilt PLA 6

RC1 1" mirror stilt PLA 2

Modified Rail carrier Modification 1

1" mirror stand PLA 1

Milled acrylic sample chamber Resin 1

Holder for the acrylic sample chamber Resin 1

Objective holder ring Resin 2

Detection axis holder Resin 1

Infinity space tube PLA 1

1" microscopy fluorescence emission filter holder PLA 2

Page 11: Selective Plane Illumination Microscopy (SPIM) for Hydra

Table 2. List of all the self-made parts including 3D printed parts and parts made by the machineshop.

Once all self-made and purchased parts were acquired, the microscope was assembled and thelaser was aligned as described in OpenSPIM. Once the physical construction was done, all theelements that are controlled by the computer during imaging were connected. These include thelaser, camera, and Picard stage, which are all run by µManager, embedded in Fiji. There isOpenSPIM software embedded in µManager that directs image acquisition, allowing for thestage to be moved, laser to be turned on and off, and the camera to take images at the right times.To take images, the acquisition protocol would be run, and all the images taken would be savedin a format ready for image processing.

4.2 Sample Image Processing:

The image processing procedure was developed using sample data from an existing SPIM setupcreated by the Tomancak group at the Max-Planck Institute (Preibisch et al, 2010; Preibisch et al,2014). Their SPIM setup was used to image Drosophila melanogaster embryos expressingHistone-YFP in all cells, which labels all the cell nuclei. Images were taken every 6 minutesfrom five different angles (Figure 9) during embryogenesis (OpenSPIM). The entire datasetcovers a 24hr time period, so data was only taken every 6 minutes, as it was unnecessary to takedata more often. The sample dataset used for image processing here only consists of the first 11timepoints, and each view contains a stack of 51 images that were taken 6 apart through theµ𝑚embryo (OpenSPIM). This step size was determined to minimize the necessary number ofimages for processing, while still collecting enough data.

Page 12: Selective Plane Illumination Microscopy (SPIM) for Hydra

Figure 9. Maximum intensity z-projections of sample at timepoint 5. All 5 angles are displayed,and the sample can be observed from various sides, along with the beads suspended with thesample. Z-projections were created using Fiji.

The first step is pre-processing, which involves converting the initial dataset, which are ome.tiffsinto .tif files. This is necessary for later processes in Fiji. I did this conversion using a script inFiji (see Appendix I). Once pre-processing is complete, registration of the views must be done. Iused bead-based registration for this sample data, and this method will also be used for datataken using my SPIM setup. Registration is done using the Multiview Reconstruction plugin inFiji (Preibisch et al, 2010; Preibisch et al, 2014). Beads are detected using an interactivethreshold that can be manually adjusted based on a visualization of bead detections on a userselected timepoint, which was timepoint 5 for this data, seen in Figure 10. I manually determinedthe threshold visually by maximizing the number of correct bead detections with no repeateddetections of the same bead. For most SPIM data, this threshold is around 0.01, which can likelybe applied to future collected data. As seen below, detections in green are not solely of beads, butalso include points within the sample, but unless these are conserved between views, they will bediscarded at a later point in registration.

Page 13: Selective Plane Illumination Microscopy (SPIM) for Hydra

Figure 10. Detected beads at timepoint 5 of sample data. Registraion is initially optimized usingtimepoint 5, and beads are detected using an optimized threshold of 0.0101, shown to the right asgreen dots.

The manually selected threshold was then applied to all timepoints and angles to detect beads inall stacks. Once beads were detected, all the views were then registered by finding the affinetransformation between views. Views were paired up to find the number of corresponding beadpairs and the transformation between views. This is summarized in Figure 11, the registrationquality. The quality for different timepoints can be seen, with the timepoint with lowest error andhighest number of inliers displayed as the reference timepoint. This is an important readout andcan be used to determine if the registration should be performed again, when parameters like thethreshold used for detecting beads can be modified.

Page 14: Selective Plane Illumination Microscopy (SPIM) for Hydra

Figure 11. Plot of quality of bead-based registration. The average, minimal, and maximaldisplacement of corresponding bead descriptors are shown, as well as the RANSAC ratio ofinliers for all time-points. The RANSAC ratio is calculated as the number of truecorrespondences divided by all correspondences.

After registration was performed, fusion and deconvolution were done using the same plugin inFiji. I visualized the fused views using the BigDataViewer plugin in Fiji, as seen in Figure 12.The five separate views were combined for all timepoints, and timepoint 5 is visible from anintermediate angle below. Since the different imaged angles correspond to different parts of thesample, the image edges surrounding the sample are determined by that angle’s bounding edges.This fusion can be cropped to minimize unnecessary data.

Figure 12. Reconstructed and fused views of sample data using BigDataViewer. All views werecombined for a single timepoint, timepoint 5, to achieve a 3D rendering of the sample that can bevisualized from any angle.

Registration Quality 100

-1.6 -- ------ 90

(')

1.4 80 0 "'I "'I

1.2

x ,:: 1.0

---------------- ---------------________ _,__ ---------------_,___ ......__ ......

(I)

70 "C 0

60 ::::; 0.. (I) ...

0 0 ,8 ... - 50 ::::; n (I) ...

U,J 0 ,6

40 ;,::, a, ,..

30 o· 0 .4

0 ,2

20

10

0 .0 R11f1u·wnc11 lim•paint 3

0

0 1 2 3 4 5 6 7 8 9 10 Time point

- minError - avgError - maxError - minRatio - avgRatio - maxRatio

Page 15: Selective Plane Illumination Microscopy (SPIM) for Hydra

This fusion can be viewed in 3D, as the image can be rotated to see all side of the sample, as wellas throughout time. Additionally, the fused image can be viewed in slices to view the internalstructure of the sample during the development process.

5. Outlook:

This implementation of this SPIM setup is fully constructed and ready to take data of Hydradevelopment. Additionally, an image processing procedure was developed using sample data thatcan be minimally modified to be applied to new data taken. Due time constraints, I was unable totake my own data, however, because all SPIM data is taken in the same manner, and using asimilar setup, the image processing pipeline is generalizable. There are various interactivethresholds throughout the process that can be adjusted for the specific data being used. Based onthe usage of SPIM for Drosophila embryo development, as was seen in the sample data, thismicroscope will be useful for studying similar processes in Hydra. Since Hydra are moretransparent than Drosophila embryos, the image quality is likely to be improved, and the entirethickness of the regenerating Hydra will be visualized using SPIM. Given more time, this SPIMsetup will hopefully produce informative images and movies that can be used to betterunderstand the process of regeneration in Hydra.

Page 16: Selective Plane Illumination Microscopy (SPIM) for Hydra

References

Daetwyler, S., Günther, U., Modes, C. D., Harrington, K., & Huisken, J. (2019). Multi-sampleSPIM image acquisition, processing and analysis of vascular growth in zebrafish. Development,146(6).

Girstmair, J., Zakrzewski, A., Lapraz, F., Handberg-Thorsager, M., Tomancak, P., Pitrone, P. G.,Simpson, F., & Telford, M. J. (2016). Light-sheet microscopy for everyone? Experience ofbuilding an OpenSPIM to study flatworm development. BMC developmental biology, 16(1), 22.

Hadjantonakis, A. K., & Papaioannou, V. E. (2004). Dynamic in vivo imaging and cell trackingusing a histone fluorescent protein fusion in mice. BMC biotechnology, 4(1), 1-14.

Huisken, J., & Stainier, D. Y. (2009). Selective plane illumination microscopy techniques indevelopmental biology. Development, 136(12), 1963–1975.

Huisken, J., Swoger, J., Del Bene, F., Wittbrodt, J., & Stelzer, E. H. K. (2004). Optical sectioningdeep inside live embryos by selective plane illumination microscopy. Science, 305(5686),1007–1009.

Krieger, J. W., Singh, A. P., Bag, N., Garbe, C. S., Saunders, T. E., Langowski, J., & Wohland, T.(2015). Imaging fluorescence (cross-) correlation spectroscopy in live cells and organisms.Nature protocols, 10(12), 1948-1974.

Preibisch, S., Saalfeld, S., Schindelin, J., & Tomancak, P. (2010). Software for bead-basedregistration of selective plane illumination microscopy data, Nature Methods, 7(6), 418–419.

Preibisch, S., Amat, F., Stamataki, E., Sarov, M., Singer, R.H., Myers, E., & Tomancak, P.(2014). Efficient Bayesian-based Multiview Deconvolution, Nature Methods, 11(6):645-648.

Schmied, C., & Tomancak, P. (2016). Sample preparation and mounting of Drosophila embryosfor multiview light sheet microscopy. Drosophila, 189-202.

Selective Plane Illumination Microscopy SPIM or LSFM Overview. Applied ScientificInstrumentation. (2020, April 28).http://www.asiimaging.com/products/light-sheet-microscopy/selective-plane-illumination-microscopy-spim-overview/.

Stelzer, E. H. K. (2015). Light-sheet fluorescence microscopy for quantitative biology. NatureMethods, 12(1), 23–26.

Takeuchi, Y., Yoshizaki, G., & Takeuchi, T. (1999). Green fluorescent protein as a cell-labelingtool and a reporter of gene expression in transgenic rainbow trout. Marine Biotechnology, 1(5),448-457.

What is Widefield Imaging? Teledyne Photometrics. (2020, October 22).https://www.photometrics.com/learn/microscopy-basics/what-is-widefield-imaging.

Page 17: Selective Plane Illumination Microscopy (SPIM) for Hydra

Willemink, M. J., & Noël, P. B. (2019). The evolution of image reconstruction for CT-fromfiltered back projection to artificial intelligence. European radiology, 29(5), 2185–2195.

Wittlieb, J., Khalturin, K., Lohmann, J. U., Anton-Erxleben, F., & Bosch, T. C. (2006).Transgenic Hydra allow in vivo tracking of individual stem cells during morphogenesis.Proceedings of the National Academy of Sciences, 103(16), 6208-6211.

Wulstein, D. M., Regan, K. E., Robertson-Anderson, R. M., & McGorty, R. (2016). Light-sheetmicroscopy with digital Fourier analysis measures transport properties over large field-of-view.Optics Express, 24(18), 20881–20894.

Page 18: Selective Plane Illumination Microscopy (SPIM) for Hydra

Appendix I: Detailed Image Processing Pipeline

1. Expand the sample data file into a working directory using

tar -xvzf OpenSPIM_Drosophila_11tp.ome.tiff.tar.gz

Each view has a stack of 51 images (imaged planes through the specimen 6 microns apart) savedas spim_TL<tt>_Angle<a>.ome.tiff

- tt is zero padded time-points (01 - 235)- a is the five acquisition angles (0-4)

2. Pre-processing- To use the OpenSPIM ome.tiff data in Fiji they need to be saved as regular tiffs- Use this in Fiji script editor (only 11 of the 235 timepoints):

/* setBatchMode( true );* won't show each image as it goes*/for ( t = 0; t < 11; t++ )

{for ( a = 0; a <= 4; a++ )

{open("5. Senior Spring/E 90/Test Operation/test_data/spim_TL"

+ IJ.pad(t, 2) +"_Angle" + a +".ome.tiff");saveAs("Tiff", "C:/Users/eadam/Documents/5. Senior Spring/E

90/Test Operation/test_data/tiffs/spim_TL" + IJ.pad(t, 2) +"_Angle" +a +".tif");

close();}

}

- File->New->Script- Paste above script (change angles/timepoints for real data)- Select ImageJ Macro under Languages

- Run

[i,;l •Fiji.ijm (Running)

[fie fdit !,anguage I emplates Run TQOls T.!!.bS Qptio ns

I I 0- : •New .ijm r *Fiji.ijm (Running) I o- Genetics l 1 • set8otch1'1ode ( true );

:: :/ won't show each image os it goes

4 for ( t = 0; t < 11 ; t ++ ) o- Digital ViewE s [I = ~~~~b :: fo r ~ a = 0;a <= 4 ;a ++ )

X

10

open ("S. Senior Spring / E 90/ Test Operation / test_data / spim_TL" + I J . pad(t, saveAs ( ~Tiff" , "C :/ Users / eadam/ Docurr.ents/ 5. Senior Spring / E 90/ Test Operati close () ;

o- Summer 202 o- Downloads

o- LabWebsite

o- local Settings

o- My Documents

11 } 12

Page 19: Selective Plane Illumination Microscopy (SPIM) for Hydra

- (Make sure in right directory on the left)3. Bead-based registration

- In Fiji: Plugins->SPIM Registration->Bead based registration

- Choose Single channel registration- Difference-of-Gaussian (DoG) is more precise but slower compared to Difference of Mean

which takes advantage of integral images (usually DoM good enough)

- Click Browse to locate the directory with the .tif files- Enter the pattern of the files, where tt is a zero padded place holder for time points and a a

zero padded placeholder for angles/views. If we have a three digit series (000 - 100) we use{ttt}; in case of our sample data that range from 0 to 10 we need tt as time point placeholder

- Adjust the extension to .tif- We will initially optimize the registration using time point 5. To register all time-points in the

time series we would enter 0-10. Can also use commas for discontinuous ranges- Specify the Angles to process, in our case 0-4 (again, can use commas)

- In the second section of the dialog we need to initially pay attention only to the parametersaffecting the initial segmentation of the bead. Bead Brightness is a pull down menu offering6 options:

o Very weak - the threshold for bead detection set to 0.0025 (typically when the beadsemit in a different wavelength compared to sample, i.e. red green beads for redfluorescent sample)

o Weak - the threshold for bead detection set to 0.02

[1;J Bead based registration

Select type of registration j Single-channel

Selecttype of detection I• · -· Please note that the SPIM Registration is based on a publication. If you use it successfully for your research please be so kind to cite our work: Preibisch et al., Nature Methods (201 0), 7(6): 418-41 9

G;J Single Channel Bead-based Registration

SPIM data directory lc :\Users\eadam\Documents Browse .. I Pattern of SPIM fil es lsp im_ TL{tt}_Angle{a}.tif

Time points to process 15

Angles to process

r Load seame nted beads

r Load segme nted beads

Bead brightness I interactive .

Subpixel loca lization 13-d ime nsional quad ratic fit (all detections) ..:J P-Spec ify ca libration ma nually (Note: oth erw is e read from fil e - s low )

xy reso luti on (um/px)

z resoluti on (um/px)

X

X

Page 20: Selective Plane Illumination Microscopy (SPIM) for Hydra

o Comparable to sample - the threshold for bead detection set to 0.075 (typically whenusing beads emitting in the same wavelength as the sample, i.e. green beads and GFPin the sample)

o Strong - the threshold for bead detection set to 0.25o Advanced - the threshold for bead detection can be set manually in the following

dialogo Interactive - the threshold can be manually adjusted based on visualization of bead

detection on a user selected time point- In our initial run we will select the option Interactive to play around with the threshold that is

best suited for our data- The Subpixel localization is a pull down menu offering three option for more-or-less

precisely localizing the beadso 3-dimensional quadratic fit (all detections) - fastest, but least precise option,

sufficient for OpenSPIM datao Gauss-fit (true correspondences) - more precise but slower, limited only to true

correspondences for performance reasons.o Gauss-fit (all detections) - the slowest option where Gauss fitting is applied to all

bead detections which can be many and it can take a while.- We will select 3-dimensional quadratic fit for our sample OpenSPIM data.- For OpenSPIM data it is very important to select the Specify calibration manually checkbox

and enter the xy and z resolution. For the sample OpenSPIM data discussed here thefollowing parameters work:

o xy resolution (um/px) = 0.645 - this is dictated by the optics of the detection path ofOpenSPIM and the size of the pixels on the CCD camera.

o z resolution (um/px) = 6 - the step size of the motors on OpenSPIM are 1.5 um andwe used 4 z steps between planes when acquiring the sample data, i.e. 4 x 1.5 = 6 um.

o Note that the important number here is the ration between the z and xy resolutions. Inthe case of OpenSPIM sample data it is 6/0.645 = 9.3023. We could also express theratio as xy = 1 and z = 9.3023. The 9.0323 or z-scaling is an important number toremember and write down. It will also be stored as the value in the *.registration filesas discussed below. We will need it during all subsequent steps.

- Finally we will examine the pull down menu Transformation model that offer three optionso Translation - uses transformation model that takes into account only translation

between views (nto particularly useful for multi-view OpenSPIM data).o Rigid - included additionally rotation

Transformation model IMolll ..:l r Re-use per time point registration

r Timelapse registration Select reference timepoint ~I M_a_n_u_a-11-y-(i-nt-e-ra_ct_i_ve_)_..:J_T~

lmgli b container I Array container (images sma ller - 2048x2048x450 px) ..:J

This Plugin is developed by Stephan Preibisch http://fly.mpi-cbg.de/preibisch

Page 21: Selective Plane Illumination Microscopy (SPIM) for Hydra

o Affine - includes scaling and shearing (necessary for OpenSPIM data due toaberrations introduced by diffraction index mismatch between water and agarose).

- We will leave Affine as a useful and, in fact, necessary default for OpenSPIM data.- Ignore the remaining options of the dialog and proceed

- Select a time-point to perform the segmentation optimization on (Click Browse and locate thefile spim_TL05_Angle0.tif)

- Typically only need to play with the Thresholdo The lower the threshold the more 'beads' are segmented. The goal here is to find a

threshold where most beads are detected - ideally only once. We can examine theperformance of a particular threshold by browsing through the stack and zooming inand out using the Fiji toolbar tools. We want to avoid the situation where severaldetections are shown for what appears to be a single bead and conversely whenclearly visible beads are not detected at all.

o The threshold that works well for OpenSPIM sample data is 0.01o Done, run the registration

- Then fusion and deconvolution- Evaluate fusion, return to image processing (registration) if dissatisfied

R.i.dius l • 2

Radius 2 • 3

Thresho ld - 0.01 0126 623

P.i' file Edit font

_J LookforMinima( red)

LookforMa xima(gree n)

Loading /home/tomancak/Desktop/OpenSPIM_f ~ B&C

D 207.2S 618.89

(Fiji Is Just) lmageJ

Page 22: Selective Plane Illumination Microscopy (SPIM) for Hydra

Appendix II: All Parts

Page 23: Selective Plane Illumination Microscopy (SPIM) for Hydra

Pur chased Part s

Laser 1/2" SM OS-M ount ed Frosted Glass Alignme nt Disk w/ 1 mm Hole Optical Breadboard Doveta il Optica l Rail, 300 m m Dovet ai l Opt ical Rail, 150 m m Dovet ai l Opt ical Rail, 75 mm Rail carr ier , 1"x1" Adjustable mechanical sli t 1/2" Optic mou nt Rotat ion Mo unt for 1" opt ics POLARIS-K1-H - 1" M irror Mo unt POLARIS-Kl - 1" M irror Mo unt 1" Broad band dielect ric mirr or 1/2" Achro matic Doublet , f = 50m m 1/2 " Achro matic Doublet, f = 25mm 1" Cyli ndrical Achrom at, f = 50mm Wa t er di pping objective lens for il lum inat ion (lOx/0.3) NBR 70 20x3 mm 0 -Ring Water di pping objective lens fo r detect ion {20x/0 .5) 1" Fluo rescent Microscopy Filter (one bandpass, one long pass) Tube lens for detect ion objecti ve lx Magn ifyi ng camera mount 0 .5x Magnify ing came ra mou nt Com pact 4 axis mo tor assembly, USB-40-Stage Rubber Band

Ball bearing system Breadboard clamps (CL2/M or CL3/M ?) Digital camera USB to serial adaptor for the laser USB cables for the 4 D position ing system USB Hub (at least 5 USB posit ions) USB led lamp fo r brightfie ld il lumi natio n M4 X 25mm cap screw (1 needs to be cut down to M4x22mm) M4 x 20 mm capscrew M4 nut M4 x 45 mm capscrew M6 x 12mm screws M3 x 8 mm screws MS x 50m m screws M6 x 25m m screws M2.5 x 6mm setscrews M6 x 25 mm capscrews M4x 10 M6 x 16 mm capscrews M2.5 x 10 M6x8 M3x6 M6 x 20 mm screws Needle nosed pliers 1.5, 5 mm Hex key/dr iver Heat tra nsfer paste

I I I I Isource ICobo lt/Co here nt/Om icron/Vortran ;Thor Labs ;Thor Labs ;Thor Labs IThor Labs ;Tho r Labs :Thor Labs ;rhor Labs 1Thor Labs ;Thor Labs ;Thor Labs ;rhor Labs 1Tho r Labs ;Thor Labs ;Thor Labs ;rhor Labs 1Qlym pus ; Lelebeck.de ;01ym pus ;rho r Labs 10lym pus ;Olym pus ;01ym pus 1 Picard Indust ries I I

RS Component s Thor Labs Ando r/ Hama matsu/pco Keyspan Any Any Any Any Any Any Any Any Any Any Any Any Any

_JAny Any Any

I Any Any Any Any Any Any

Quant i_!y

TOTA L:

12 6

10 3

Cost (each) Link 1 2 I 3 1.52 https://www .thor labs .de/thorProduct.cfm ?partN umb er=DGOS-1500 -Hl -M D

193.64 https://www.thorlabs.de/thorProduct.cfm?partNumber=MB3045/M 78 .54 https://www .thor labs.de/thor Product.cfrn ?partNu rnber -RLA300/ M 44 .39 https://www.thorlabs.de/thorProduct.cfm?partNumber-RLA150/M 29 .94 https://www.thor labs.de/tho rProduct.cfm?partNu m ber- RLA075/M 26 .16 https://www.thorlabs.de/thorProduct.cfrn?partNumber=RCl 265.8 https://www.thor labs .de/thorProduct.cfm?partNumber=VA100/M 15 .7 1 https://www.thor labs .de/thorProduct.cfm?partNumber=LMROS/M

137 .63 https://www .tho rlabs.de/thorProduct.cfm ?partNumb er=RSP1X15/M#tad- imag e-O 142 .88 https://www.thor labs.co m/t horProduct.cf m ?partNum ber=POLARIS-Kl -H#tad-image -O 14 7 .4 7 https://www .thor labs .com/thorProduct.cf m ?partNumber=POLARIS-Kl#tad-image-0

75. 1 https://www.thorlabs .de/thorProduct.cfm?partNumber=BB1-E02 82 . 7 4 https://www .thorlabs.de/thorProduct.cfm ?partNu mb er=AC127 -050-A -M L 82 .74 https://www.thorlabs.de/thorProduct.cfm?partNumber=AC127-025-A-ML

395 .03 https://www .thorlabs .de/thorProduct.cfm ?partNumber=ACY254-050-A 1https://www.olympus-lifescience.com/en/objectives/ lumplfln-w/ http://lelebeck.de/1006.htm/ https://www.olympus-lifescience.com/en/objectives/lumplfln-w/#tlcms[tabJ-%2Fobiectives%2Flumplfln-w%2F20xw

100 .http://www.thorlabs.de/newgrouppage9.cfm ?ob ject group id - 1001/

4260 http://www.picard-industries .co m/ prod ucts/ usb-4d-sta ge.html

5,05 €: https: //de. rs-online.com /we b/p/kuge llager /4090057 / ; http s://www .thorlabs.de/thorProduct .cfm ?partNu mb er=CL2/M I I

40 1 http s://tr ipplit e.com/keys pan-hi gh-speed-usb-to -serial-adapterUSA19HS I I I I I I I I I I I I I I I I I I I I I I I I

-l I I I I I I I I I I I I I I

7045.43 1 I

Page 24: Selective Plane Illumination Microscopy (SPIM) for Hydra

Opt ional Parts

SD1 - 1/4" (may need extra?) M4 Cap Screw and Hardware Kit M6 Cap Screw and Hardware Kit SM1 Spanner Wrench

Self-Made Parts

:Thor Labs 1Thor Labs :Thor Labs _:_Thor Labs I I I I I

!Mate rial Heat sink 1alum inium ~c1 vert ical slit stilt PLA

C11/2" lens stilt PLA .. C11 " mirror sti lt PLA Modified Rail carrier

[1" mir ror stand ~PLA Milled acryl ic sample chamber 1Acry lic Holder for the acrylic samp le chamber l meta l Objecti ve holder ring :metal Detect ion axis ho lder 1met al Infinity space tub e PLA 11" microscopy fluo rescence emission filter holder PLA

8.15: https://www .thorlabs.de/thorProduct.cfm ?partNu mber=SD1 59 .361 http://www.tho rlabs.de/tho rProduct .cfm ?partNum ber=HW-KIT1/M

117 .67: http://www.tho rlabs.de/tho rProduct.cfm ?partNumbe r=HW-KIT2/M 27 .55: https://www .thorla bs.de/thorProduct.cf m ?partNu mber=SPW602

3D Print? I Quant i_!y no yes yes yes no yes no no no

yes

1, 1 ' I 21

I 1 ,

I _j_

I I I