Expression in E.coli with rare codons

mccallo at rand mccallo at rand
Mon Mar 14 12:58:04 EST 1994

In article <2lmn54$ch6 at>, ANDY PHILLIPS <andy.phillips at> (Tel +44 275 392181 EXT 257) writes:
> We're trying to express a number of Arabidopsis cDNAs in E.coli
> to get large amounts of protein for biochemistry, NMR etc. We've
> so far tried the pTrcHis vectors from Invitrogen, pUC19 and
> pET-3d, but with no great success. We can detect active enzyme, but
> get no impressive bands on SDS gels. 
> When we compare the codon usage of these genes, we find that they
> use some codons for Ile, Pro, Arg etc which are very rare in E.coli, 
> particularly in highly-expressed genes. We've tried adding a plasmid which
> carries rare Arg codons, with no effect. 
> How worried should we be about this codon usage? Is E.coli should worth
> pursuing, by playing about with growth and induction conditions, or
> new vectors? Or should we switch systems, and try a different one?
> Pischia? Baculovirus?
> Any comments would be very welcome.
> Andy Phillips (andy.phillips at
> Long Ashton Research Station, Bristol, UK

Andy, when you say you tried _adding a plasmid which carries rare Arg codons,
with no effect_ do you really mean that you added a plasmid carrying a rare Arg
tRNA gene?  I ask because this has solved similar problems for me in the past. 
Rare Arg codons (ones starting with _A_) can hurt you in two ways:  first, if
they ever appear in tandem, you have a false RBS.  Presumably, ribosomes bind
here, and since they don't have a properly spaced fMET codon downstream, they
just sit there and block properly initiated translation proceeding from
upstream.  Second, when you attempt to overexpress a gene containing these
(argU cognate) codons, the small pool of charged tRNAs is depleted and the
essential gene dnaG (primase) cannot be translated at sufficient level to allow
DNA replication to proceed.  This is why this gene was originally misnamed as
dnaY:  their ts mutant had the phenotype of an essential replication protein. 
When they got around to sequencing the thing they saw that it was a tRNA gene. 
If your gene consists of 3% or more of argU cognate codons (AGA/AGG), this
could well be your problem.  My solution was to make a synthetic argU gene on
the synthesizer (its short enough to do in one shot) and clone it into
pACYC184.  This plasmid is compatable with pMB1-based expression vectors like
yours.  This has completely fixed the problem in more than one case;  in one
case the expression went from undetectable on coomassie to 15% TCP after the
addition of the second plasmid.  If you have tandem argU codons, you must
change with site-directed mutagenesis to loose the false RBS.  Check out these
references:  Nucleic Acids Res. 12:6683, 1984
	     Gene 100:59, 1991
	     J. Bacteriol. 175:716, 1993
	     Nucleic Acids Res. 20:6707, 1992
	     Gene 85:109, 1989
	     NAR 18:5031, 1990
	     Cell 37:669, 1984
	     Cell 45:453, 1986
	     J. Bacteriol. 174:1956, 1992
Good luck.
Owen McCall
D47N, AP9A, Abbott Laboratories, Abbott Park, IL USA

More information about the Methods mailing list