Western blotting: from brain tissue to quantified bands

Questions to investigate

1. Run the knockout design with the specific antibody. Is the target band absent from every knockout lane, and what normalised signal remains?
2. Switch to the antibody with a cross-reactive band and move that band close to the target molecular weight. When does the knockout appear to retain the protein, and which gel percentage separates the two bands?
3. Lengthen the exposure until bands saturate. Which bands saturate first, why is quantification then invalid, and how does a dilution series reveal the linear range?
4. Group the lanes by condition with the transfer gradient switched on, then randomise them. How does lane order bias the comparison, and how much does normalisation correct?
5. Let the treatment change the housekeeping protein. How does normalising to it distort the estimated effect compared with total-protein normalisation?
6. Omit the protease inhibitors, or assay the lysates undiluted. What happens to the bands, to the standard-curve estimates and to the loading volumes?

In a real laboratory

Animals are allocated to groups at random, and tissue is collected under an approved protocol from the smallest number of animals that answers the question (the 3Rs: replacement, reduction and refinement); tissue from one set of animals can often serve several assays. The region is homogenised on ice in lysis buffer with protease (and, where needed, phosphatase) inhibitors, cleared by centrifugation and assayed in duplicate against fresh albumin standards, diluted so that every sample falls within the standard range. Equal amounts of protein are loaded in randomised lane order, with a molecular-weight marker, on a gel whose percentage suits the size of the protein; after transfer, a total-protein stain records loading and transfer quality. Antibodies are validated, ideally on knockout tissue, and the linear range of detection is checked with a dilution series. Images are acquired at exposures that do not saturate the bands, and densitometry is performed blind, with background subtraction and normalisation; total-protein normalisation is often preferred because housekeeping proteins can change with treatment. Lanes from the same animal are averaged, because the animal is the experimental unit. Full uncropped images of every blot are kept and reported.

How the simulation works

The level of the target protein in each animal is drawn from a log-normal distribution around its group mean (the knockout has none), and tissue mass and protein yield set the lysate concentration. BCA absorbances follow a saturating curve with noise; the chosen standard curve (linear or quadratic) is fitted by least squares and used to estimate each concentration and the loading volume, so assay error carries into the loading. Proteins migrate approximately linearly in the logarithm of molecular weight over a range set by the gel percentage. Transfer falls with molecular weight at short transfer times, very small proteins can pass through the membrane in long transfers, and an optional gradient weakens transfer towards the right of the membrane. The antibody signal saturates with antibody concentration and is proportional to the amount of antigen, so loss of linearity comes from the camera: a 16-bit image with shot and read noise that saturates at 65,535 counts. Background rises with short blocking and concentrated antibodies. Densitometry integrates pixel values in adjustable boxes, with local background taken as the median of strips above and below each box. Degradation without protease inhibitors, a cross-reactive band, faint nonspecific bands and smiling are included. All values are approximate. Post-translational modifications, stripping and reprobing, fluorescent multiplex detection, substrate kinetics and the animal procedures are not simulated.