Automated deep-phenotyping of the vertebrate brain

  1. Amin Allalou
  2. Yuelong Wu
  3. Mostafa Ghannad-Rezaie
  4. Peter M Eimon  Is a corresponding author
  5. Mehmet Fatih Yanik  Is a corresponding author
  1. Massachusetts Institute of Technology, United States
  2. Uppsala University, Sweden
  3. ETH Zürich, Switzerland
6 figures, 3 videos and 2 additional files

Figures

Figure 1 with 4 supplements
Automated OPT platform for automated 3D in situ phenotyping.

(A) From left to right, the optical projection tomography platform consists of the following components: (1) a post-mountable broadband emission quartz tungsten-halogen light source, (2) a ground …

https://doi.org/10.7554/eLife.23379.002
Figure 1—figure supplement 1
Capillary fabrication and OPT image acquisition.

(A–C) Fabrication of a refractive index-matched tapered insertion within the glass capillary. (A) After being cleaned in Nano-Strip, the distal end of the capillary is dipped into index-matching …

https://doi.org/10.7554/eLife.23379.003
Figure 1—figure supplement 2
Registering 3D zebrafish images.

(A) Iterative shape averaging (ISA) is used to generate an average unstained reference fish (URF) for each developmental stage of interest from 5–7 wild-type embryos. Left: 2 dpf reference fish …

https://doi.org/10.7554/eLife.23379.004
Figure 1—figure supplement 3
Registration workflow for in situ pattern alignment.

(A) Iterative shape averaging is used to select probe reference fish (PRF). All wild-type fish from the same age stained with the same probe (8+) are aligned in the green channel. The initial …

https://doi.org/10.7554/eLife.23379.005
Figure 1—figure supplement 4
alternate registration workflow for in situ pattern alignment.

A wild-type probe reference fish (PRF) is selected using iterative shape averaging following the workflow in Figure 1—figure supplement 3A,B. All wild-type and mutant embryos from the same age …

https://doi.org/10.7554/eLife.23379.006
Automated registration and alignment of 3D WISH images.

(A) Alignment accuracy of 3D registration algorithms was verified using 26 wild-type embryos stained with tryptophan hydroxylase 2 (tph2). Stained embryos are randomly divided into three groups and …

https://doi.org/10.7554/eLife.23379.008
Figure 3 with 4 supplements
Automated detection and statistical quantification of fezf2 mutant deficits.

(A) Right: alterations in the expression of individual probes can be detected using a correlation-significance analysis approach. For each brain region, all wild-type embryos are compared with each …

https://doi.org/10.7554/eLife.23379.009
Figure 3—figure supplement 1
3D anatomical brain atlases.

Surface renderings of all regions in the 2 dpf (top) and 3 dpf (bottom) 3D anatomical zebrafish brain atlases. Both atlases comprise all 22 regions listed at right. Regions are subdivided between …

https://doi.org/10.7554/eLife.23379.010
Figure 3—figure supplement 2
Expression data mapped to brain regions.

Quantification of expression data for all probes mapped to all brain regions at 2 dpf (left) and 3 dpf (right). Expression data are visualized using a method developed for the Virtual Brain …

https://doi.org/10.7554/eLife.23379.011
Figure 3—figure supplement 3
Negative controls for automated 3D phenotyping.

(A) A null data set, consisting of 16+ negative control embryos (i.e. age-matched siblings that are either genetically wild-type or heterozygous for the fezf2 mutation) per probe, is used to define …

https://doi.org/10.7554/eLife.23379.012
Figure 3—figure supplement 4
Negative controls for maximum intensity projections.

(A) The null data set from Figure 3—figure supplement 3A is used to define the false discovery rate (FDR) for significance analysis of maximum intensity projections (MIPs) over a range of p-value …

https://doi.org/10.7554/eLife.23379.013
Figure 4 with 1 supplement
Automated phenotyping uncovers known and novel diencephalic deficits in fezf2 mutants.

(A–D) Overlay analysis of in situ expression patterns in wild-type and fezf2 mutant embryos at 2 dpf. Wild-type expression patterns are shown in green and fezf2 mutants are shown in magenta. …

https://doi.org/10.7554/eLife.23379.015
Figure 4—figure supplement 1
Additional diencephalon phenotypes in fezf2 mutant embryos.

(A–F) Overlay analysis of in situ expression patterns in wild-type and fezf2 mutant embryos at 2 and 3 dpf. Wild-type expression patterns are shown in green and fezf2 mutants are shown in magenta. …

https://doi.org/10.7554/eLife.23379.016
Figure 5 with 2 supplements
Fezf2 mutants exhibit telencephalic glutamatergic deficits during early development.

(A–F) Overlay analysis of in situ expression patterns in wild-type and fezf2 mutant embryos at 2 dpf. Wild-type expression patterns are shown in green and fezf2 mutants are shown in magenta. (A) …

https://doi.org/10.7554/eLife.23379.017
Figure 5—figure supplement 1
Additional telencephalon phenotypes in fezf2 mutant embryos.

(A–I) Overlay analysis of in situ expression patterns in wild-type and fezf2 mutant embryos at 2 and 3 dpf. Wild-type expression patterns are shown in green and fezf2 mutants are shown in magenta. …

https://doi.org/10.7554/eLife.23379.018
Figure 5—figure supplement 2
3D segmentations of telencephalic expression domains.

3D segmentations of gene expression domains within and immediately adjacent to telencephalon at 2 dpf are generated in a semi-automated manner using adaptive thresholding. The gray background …

https://doi.org/10.7554/eLife.23379.019
Figure 6 with 1 supplement
Segmentation and volume measurements.

(A) 3D segmentations and volumetric quantification of gene expression domains in the telencephalon of wild-type (green) and fezf2 mutant (magenta) embryos at 2 dpf. 3D segmentations are done …

https://doi.org/10.7554/eLife.23379.020
Figure 6—figure supplement 1
Two-photon analysis of telencephalon morphology.

(A) 3D segmentations of the telencephalon in 2 dpf Nissl-stained embryos. Six wild-type and six fezf2 mutant embryos from the same clutch were imaged using two-photon excitation microscopy and …

https://doi.org/10.7554/eLife.23379.022

Videos

Video 1
3D reconstruction of a wild-type 2 dpf embryo stained with tyrosine hydroxylase.
https://doi.org/10.7554/eLife.23379.007
Video 2
3D visualization of aggregate difference imaging at 2 days post fertilization.

Voxel intensity difference is calculated between wild-types and fezf2 mutants for each probe within each affected brain region. Summing the absolute value of the significant differences for all …

https://doi.org/10.7554/eLife.23379.014
Video 3
Visualization of vglut1 expression in wild-type and fezf2 mutant embryos.

The video shows overlay analysis of the vglut1 expression pattern in wild-types and mutants at 2 dpf. Wild-types are shown in green and mutants in magenta. The position and orientation of each 2D …

https://doi.org/10.7554/eLife.23379.023

Additional files

Supplementary file 1

Whole mount in situ probe library.

All probes used for automated 3D phenotyping are listed along with the forward and reverse primers used in PCR cloning.

https://doi.org/10.7554/eLife.23379.024
Supplementary file 2

Significance analysis overlaid on maximum Intensity projections (MIPs).

MIPs have been color coded to highlight all regions in which the expression of a given probe is significantly reduced (cyan) or increased (magenta) in mutant embryos (p<0.5 × 10−3). Significant intensity differences between mutants and wild-types were determined by performing a Mann-Whitney U-test to compare corresponding voxels for each probe within each brain region showing alterations in Automated Correlation Analysis. All MIPs show dorsal (left) and lateral (right) views.

https://doi.org/10.7554/eLife.23379.025

Download links