I tried to use GAPDH, but the results were inconsistent, so I'm thinking about beta-actin, but I'm not sure if it would vary as well as the proteins of interest, being thus unsuitable as a control.
Good question. The choice of loading control may depend on a few variables, including tissue type (neuronal?) and sample preparation (whole cell lysate, nuclear, soluble?).
I quickly read some plasticity articles from Nature Neuroscience to see what high stature articles are using these days, and I found that GAPDH, B-actin, and A-tubulin were all being used (also, I've published before using B-actin as a loading control for hippocampal homogenate). All three controls seem to be valid and accepted.
This makes me worry a little bit about your samples. Maybe the "inconsistent" data are real, and what you are seeing are problems in sample preparation leading to protein degradation, or perhaps your protein assay isn't working correctly.
My recommendation is to next time perform both GAPDH and B-actin for controls and compare (you can multiplex a single blot, cut your blot into strips, or run two blots side by side identically loaded, whatever is easier). If they are identical, this points to your sample preparation as opposed to the loading control.
Probably the samples are OK, because we already run the same membranes for 3 other proteins, and all of them were where they should be, the problem seems to be only with GAPDH, because we ran for it several times and some of the bands, in different bands, did not appear.
And yes, it is neural tissue, from rat hippocampus.
I am doing the process with beta-actin right now (and using the computer while it is incubating hehe), because, as you said, in the literature it seems to be a fairly used control; I just wanted to be more sure by asking here.
If the blot from today does not work, I'll probably follow your idea of doing both controls and comparing.