In order to breed a wild canid to friendliness to humans, you'd have to have a substantial number of them to work with, as they had with those foxes/sables that were being kept captive for the fur business. To do it with only a few, you'd have to be very lucky in the few that you started with - they'd have to already have the right variants in their genomes.
It's still early days for epigenetics, so you can hypothesize whatever you want. :) The differences between dogs and wolves and quite stable, though, and I would guess that means sequence differences. Here's another reference, also open access.
MHC variability supports dog domestication from a large number of wolves: high diversity in Asia.
http://www.ncbi.nlm.nih.gov/pubmed/23073392
It's commonly stated that dogs can breed successfully with wolves, and that has certainly happened (the melanistic, dark wolf phenotype in American wolves actually was transferred from domestic dogs.) However, there are hundreds of dog breeds, and no one has tested them all to see if any of them have lost fertility in dog/wolf crosses. I don't suppose anyone will give grant money to test that, but it seems possible to me that some dog breeds might in fact be incipient new species, by the "biological" species criterion.
