Rb Anti-vGlut2 Preview

Rabbit anti-vGliuT2 (1:16k) in the LHA (approx. Level 28). Image was acquired with Volocity and assembled with Photomerge/PS (CS5). The original image was gamut and profile corrected, and cropped and rotated. This version was also, sharpened & brightened, converted to sRGB & reduced to 1024x1024 (From 3737 x 3737)

Immuno Test Results-11/16/12

The immuno came out great—basically all primaries tested worked best at the highest dilution tested. That is,

• 1:40k for Paul's MCH
• 1:16k for the rabbit anti-vGlutT2
• 1:5k for Zamir's MCH
• 1:10k worked well for the NEI 

To document the results we should image at least one section of each at as close to the same atlas level as possible.

Sections to image for Paul's MCH 
1:10k = S1 on the 10, 20 40k slide
1:20k = S4 on the 10, 20 40k slide
1:40k = S1 on the 40k slide (see below)

For Zamir's MCH, an image of S3 on the 1, 2.5, 5k slide will be enough

The NEI-labeled sections are generally too rostral or caudal for a direct comparison, but there is labeling in the LHA and ACB. Time permitting, it would be nice to image S1 (ACB) and S3 (LHA, Lvl 27).

The vGlutT2 needs to be imaged as a 1 panel z-stack at 20X to clearly distinguish labeled elements. The background is very high at 1:4k so image 8 and 16k only. 


I've captured the 1st image (Paul's MCH, 1:40k) using Volocity on the 428c Zeiss, leaving only 5 or so to go...

1024 X 1002 image of Paul's MCH (1:40k), acquired as a 10X,  5 X 5 composite with Volocity, and assembled using PS and Photomerge. 

LHA Imaging Protocol

After examining the test images and crunching the numbers, here's the image acquisition protocol for the LHA project:

Image Acquisition:

Acquire images at 10X, 2048x2048, 8 bit, spd 7, Avg 2, 5x4 tiles (15% overlap), with an optical slice thickness of 8.5 µm, which gives 4z in 25 µm sections or 6z in 40 µm sections.

z-Stack:
The slice thickness is based on a pinhole of 40.9 µm, which is approx. 1 A in the Far Red (FR), 1.08 A in the Red, 1.25 A in the Green and 1.66 A in the Blue channels.

Acquisition parameters are optimized across all 4 channels, even though we will rarely use the 4th channel. The alternative is optimizing each combination of tracers for the longest wavelength, which would mean that the x, y and z resolution (lateral resolution and z-step) of a channel would change depending on the combination of channels.

All of these parameters can be found in the 2nd and 3rd sheets of the Image Acquisition Google Document.

Imaging Strategy

The file size, acquisition time, and the number of immuno series to be imaged makes it impractical to image all available LHA sections (except perhaps in the 40 µm 1:6 series). Instead sections to be imaged will be based on Atlas levels prioritized by region of interest. For us (as opposed to UTEP), we will initially focus on the sections corresponding to Atlas level 26, then levels 24 and 25, followed by levels 29 and 30 and finishing with levels 23 and 32. These levels represent the iconic view of the LHA (26), the ACB-recipient LHA (24 and 25), levels central to Arshad's region of interest (29 and 30) and the rostral and caudal  LHA  ends (23 and 32; not quite the pole, but easier to find/identify). This strategy covers 8 of 11 levels of the LHA. Exactly how this translates into images acquired is difficult to estimate because it is likely that more than one image will to deal with plane-of-section issues.

Immuno Strategy

Assuming that the same antibody raised in a different species gives identical results, our goal is to prepare series for 11 different antibodies (plus AVP), replicating one or more labels in each brain to facilitate analysis of the patterns of expression across brain. Ensuring that each brain is in register with the others will require 5 brains of 4 series each, for a total of 20 immuno series, which works out to be 2 series each for 9 of the 11 series, and 1 series for the last 2 (which will likely be AChE and vGluT). Because we want a single-labeled series for each peptide, this also means that 11 of 20 will be single labeling (plus DAPI), and 9 of 20 can be double or triple labeled (plus DAPI).

Files Sizes & Acq Time

 Single Images:
2 Channels: 
10x, 2048, spd 7, Avg 2, 5x4x4x2 (25 µm) = 800 MB and 52 min/image
10x, 2048, spd 7, Avg 2, 5x4x6x2 (40 µm) = 1120 MB and 73 min/image

3 Channels:
10x, 2048, spd 7, Avg 2, 5x4x4x3 (25 µm) = 960 MB and 62 min/image
10x, 2048, spd 7, Avg 2, 5x4x6x3 (40 µm) = 1440 MB and 93 min/image

Level 26:
Let's assume we'll acquire 2 images (not counting the Nissl) to capture level 26 X 20 series to cover all the immuno = a library of 40 images.

If each image is 800 to 1400 MB, and takes somewhere between 52 to 93 minutes to acquire, imaging level 26 take anywhere from 35 to 62 hours and generate between 31 to 56 GB of images. 

All Images/Levels:
Let's assume it will take 12 images to cover the 8 levels of interest. That's 240 images to to cover all series. Using the same numbers for size and duration, it will take roughly 200 to 372 hours (8.5 to 15.5 days) and generate between 187 and 338 GB of images—not including the Nissls. Overall, these numbers aren't as bad as I expected, but clearly the sooner we start, the better off we'll be

 Next Steps

Three things are needed before we start generating data: 

• Work out the DAPI (or NeuroTrace) protocol and get accustomed to using it
• Generate the Nissls and figure out the correspondence between section number and atlas level
• Finalize our list of brains to use, 1º combinations (both which peptides go with each brain and which will be singles & which labels will be combined)

And only two things are needed until acquiring the data is all that remains:

• Optimize the remaining antibodies (5-HT, vGluT2, DYN, and any species variants such as sh ∝ GAD)
• Run the immuno...

ms ∝ GAD67 & ms ∝ VGlut2

In brief, all GAD67's worked great, and the 1:5k grouped worked the best (and should be used for all future GAD67 series). It could probably be diluted further, but isn't necessary (& therefore not worth the time to try it).

The VGlut2 on the other hand wasn't so clear cut. There was clear and obvious labeling in the MH at all dilutions, but it was difficult to say that differences in intensity in other brain regions were specific vGlut2 labeling. This may be due to the nature of vGlut2 labeling, and for this reason, tests should be run with the rb ∝ and/or gp ∝ VGlut2 before writing off the ms ∝ VGlut2.

Imaging Nomenclature

When working with 4-dimensional images, terms like 'image', 'slice' and 'section' become ambiguous, so here's a tentative nomenclature scheme for referring to an image and its parts. 

• Aperture: The pinhole opening that allows light to pass. Aperture can be measured in µm, which is constant, or Airy (A or AU), which varies with wavelength. Aperture also determines slice thickness. Note that the resolving power of a given aperture varies by wavelength, but slice thickness does not.
• Brain: All series volumes
• Channel: The part of the image containing the signal, or labeling. Each image can have multiple channels.
• File: The saved/stored/named components of an image
• Frame: One field-of-view (f-o-v). The most basic component image.
• Image: The assembled frames (Tiles + Stacks)
• Library: A collection of images from multiple brains
• Mosaic: An image composed of stitched tiles.
• Series: Set of brain sections collected at the same interval or frequency. Can also 
• Slice: One of two or more frames in register in z
• Stack: All slices in register in z
• Stitched Stack: A mosaic with x, y, and z components
• Section: A brain section.
• Slab: The z distance of a section that is imaged. Ideally, slab thickness = section thickness, but is usually less
• Tile: A component of a stitched image. Can be a frame or a stack.
• Volume: All sections of a series
• z-Step, or slice interval: The distance moved, in z before imaging the next slice. Note that Nyquist oversamping is 150%, meaning the actual number of slices is 1.5 x (slab thickness/slice thickness.

1024 vs 2048: x,y,z preview

512 x 512 image of NT labeled w/Alexa 488. Source is 10X, 2048x2048, single channel, and 2x2x5 tiles (20 component images, 4 MB/tile and 80MB for the image). Acquisition time was 15.49 s/ tile and 5m 15s for the image

512 x 512 image of NT labeled w/Alexa 488. Source is 20X, 1024x1024, single channel, and 4x4x16 tiles (256 component images x 1 MB/Tile = 256 MB for the image). Acquisition time was 7.75 s/ tile and 35+ m the image.

Larger verision of 10X, 2048: 900 x 897 (but compression = 7)

Larger verision of 20X, 1024: 900 x 897 (but compression = 7)

Clearly more detail available in the 20X image, but differences diminish at the perspective needed to view the whole image (original images are 3634 x 3594, which is approx. 12 x 12" at print resolution, and 38 x 38" at screen resolution!)