I wrote this in response to a direct email, in case what I have written is of use to anyone else having trouble with MuLE.

"The MuLE saga returns! I have actually completely stopped using the system after determining that the inconsistency, propensity for recombination, and mysterious mosaic expression was just taking up way too much time and effort to figure out, and seemed to be inherent to Multisite Gateway lentiviruses.

I did find that using NEB Stables at 30C was a much better bet than DH5a-type bacteria, but still had huge variability. Though sometimes I was able to get a clone using 3 to 5 fragment Gateway recombination on the first try, many times it would take between 2 and 24 screened clones, and sometimes I was never able to get the final clone. Many would transiently express each of the inserts but not make functional virus, or would make extremely poor virus, though some more or less worked great.

However, what I wanted for MuLE was to be able to express several fluorescently-tagged genes in neurons at equal stoichometries. Before beginning, I read that it is not the correct strategy to express the same promoter multiple times in one plasmid because of promoter interference, so I used Ef1a, ESYN, CaMK, and other neuron-friendly promoters. But in both my neuron constructs and those using the original MuLE entry vectors driven by SV40 or CMV, I found that there was mosaic expression. I would have a monoclonal prep of plasmid, I would make virus (but same problem with transient too actually), and would find that though all cells seemed to express the PGK-iRFP destination vector piece fine, they would shut down one of the two inserts, and only rarely express both. The expression of any one of the inserts was also so much lower than what I could achieve using regular lenti systems, that it was essentially unusable for difficult cell lines like neurons. I’ve never determined what it was about the inserts created this mosaic expression situation, and your suggestion about cell states could be right. I later learned that plasmids with repetitive sequences can have very complicated secondary structures that make them not only difficult to copy and push towards recombination in bacteria, but also to express. My last ditch effort was to use this fancy bacteria from Scarab Genomics that is incapable of recombination. I found I was also not able to efficiently derive final clones using this bacteria, so I shit-canned the whole MuLE enterprise, the 6 months of work and the near 100 entry vectors I made for my poor neuron project, and am now happily making regular, non-Gateway lentiviruses!"

Did you try to use dam-/dcm- e. coli? DNA is hypermthylated in bacteria and maybe this is how some of your promoters get inactivated?