Wait, it's reversible?? Not irreversible? Whoa I didn't know that. You got any sources? Not that I don't believe you, I just wanna check it out lol.
when i was at university in sweden (where i had free access to full text PDFs for lots of the journals) i read an article that described the acute and long-term effects of phenelzine on neurotransmitter levels in rats. for GABA, GABA levels (which were about 200% elevated at whatever dose they used) returned to normal "baseline" about 48 hours after the last dose. i believe its in this article http://www.ncbi.nlm.nih.gov/pubmed/1610412
but since i don't have full text access right now i can't go through the PDF and check. i might have a couple PDFs saved locally though. this one seems good too:
Neurochemical effects of the monoamine oxidase inhibitor phenelzine on
brain GABA and alanine: A comparison with vigabatrin
Kathryn G. Todd and Glen B. Baker
Neurochemical Research Unit, Department of Psychiatry, University of Alberta, Edmonton, Canada
btw, the group at Department of Psychiatry, University of Alberta, Edmonton, Canada that wrote these articles has a lot of focus on phenelzine and the MAOIs. the fact that pretreatment with another MAOI (parnate) can block most of the GABA elevating effect suggests that the metabolite PEH that is produced from a secondary MAO reaction is largely responsible for the GABA-T inhibition.
also, GABA-T is B6-dependent like a lot of the other enzymes nardil inhibits (B6 is a cofactor bound inside of the protein), so i'm not sure if phenelzine (or its metabolites) directly destroys B6 or destroys B6 inside of the enzyme core...
as to the MAOI properties, there is some clue to the mechanism of how phenelzine might work based on a crystal structure of MAO-B in complex with reacted phenylethylhydrazine. it appears that the hydrazine binds directly to the cofactor FAD+ (flavin-adenine dinucleotide, a sort of single-stranded two base DNA analogue), leaving behind just phenylethane:
as to how tranylcypromine acheives MAO inhibition, check out this one (it's quite similar mechanism, covalent modification of FAD+):
this is pretty much zero-day sh*t, i had checked the PDB a few weeks ago and a lot of this stuff wasn't available yet
the reason why they use MAO-B and not A in these structures (there's actually quite a lot of them if you just search for "monoamine oxidase"), is probably just that it was easier to crystallize MAO-B (MAO enzymes are usually bound to mitochondrial membranes, so to make an MAO that crystallizes in solution can be tricky experimentally). also if you search for the individual compounds (tranylcypromine, phenethylhydrazine) there are complexes of other amine oxidases bound to these substrates, as well as another FAD-containing enzyme LSD1 (lysine-specific demethylase)...