I have worked with the optimization of several viral proteins, from different viruses/families, with different results.

When Menzella, 2011 was published I tried that method. I found quite the contrary result, in that randomization did far poorer than straight maximization of codon usage. I chucked this up to working within the context of viral infection (though it was also true in transient expression from plasmid transfection under CMV promotion).

I was quite excited by Pechmann and Frydman's Nature article which highlighted the importance of the location of the codon within the protein when regarding optimal vs suboptimal codon selection. Oddly enough we were able to deploy this information most successfully with the purposeful deoptimization of a protein.

Pairing a locational awareness, screening for secondary structure, and then maximizing codon optimization has had the best results for more than a year. By maximization, I mean choosing the most optimal codon in every instance that doesn't violate:

  1. Needed restriction sites
  2. Predicated Secondary Structure
  3. Location gradient to a lesser extent (ibid.)

This is all well and good, but we've been having problems with my institutions preferred vendor for gene synthesis. A post-doc that just joined my lab recommended a company I had never heard of before, and they posed an optimization strategy I had never considered before. Instead of maximizing codon optimization, they propose a "balanced" codon optimization that matches the normal distribution of codons in the target organism.

For example, the codon bias for alanine in humans breaks down (roughly) as follows:

  1. GCC 65%
  2. GCT 20%
  3. GCA 11%
  4. GCG 4%

As opposed to a maximization strategy which would seek to use "GCC" when ever possible and only when there was an issue switch to "GCT," the balanced stratgy would seek to replicate the above distribution as closely as possible.

Has anyone tried a balanced codon approach, especially when compared to a maximization technique? My overall goal is to maximize expression of naturally poorly expressing protein. I know that straight maximization can occasionally lead to poor expression for a variety of reasons (often fatal secondary structure), but this seems like it would hardly increase expression at all.

I would be further interested if anyone else had experience within a viral context.

I know that there have been a few related questions, mostly in the context of E. Coli. I don't think this is a duplicate, but would be willing to remove if it is seen as such. If I run into some extra funds I may test this empirically, but often when I'm trying such a test it's only in relation to 2-3 proteins, which might not be representative.

  • $\begingroup$ Do you want to control the protein synthesis rate? Something like you do not want it to be produced too fast because of folding issues? This article summarizes different mathematical models for translation. You can obtain optimal codon usage by fixing the synthesis rate. $\endgroup$
    Commented Dec 16, 2014 at 15:35
  • $\begingroup$ This paper talks about codon usage and virus-host interactions. Is it what you are interested in ? $\endgroup$
    Commented Dec 16, 2014 at 15:37
  • $\begingroup$ @WYSIWYG I want to maximize protein expression in a human cell line. While I appreciate David's work, that's not what really is at play here. The empirical observation from the theories leaves more complex modeling to be done. I am mostly interested in the bold question, perhaps I included too much background. $\endgroup$
    – Atl LED
    Commented Dec 16, 2014 at 15:42
  • $\begingroup$ I suppose I mention Menzella solely as an example of what makes me weary to try something different. If I can reasonable suspect that it will end in failure, I'd just as soon not waste time or funds on it. $\endgroup$
    – Atl LED
    Commented Dec 16, 2014 at 15:46
  • $\begingroup$ I suspect that this maximization and balanced codon usage will depend on your protein; there would be some optimal rate of synthesis at which it will fold properly. IMHO there cannot be a generalized model unless folding dynamics can be predicted solely based on sequence (which is very difficult). $\endgroup$
    Commented Dec 16, 2014 at 15:55

2 Answers 2


Codon usage optimization is still pretty much heurestic. The most common factor we think are important in E. Coli are:

Most probably it's a mixture of all of those.

Disclaimer: We are working on a codon optimization strategy for e.coli/yeast/hela and it works in e. coli but is yet untested in human cells. If you want to take a long shot just shoot me a message and we could try it. Apart from this, I would also offer you to have a look at your sequence and see if there is anything 'strange' from the codon usage point of view and what I think what be the best optimization strategy.

  • $\begingroup$ It would be great if you provided some references to support your response! $\endgroup$ Commented Dec 17, 2014 at 16:21
  • $\begingroup$ Thanks for the answer. Small editing note, you included the "]" in several of the links (I could change it for you, but as new user you might want the practice). Looking at references, none seem to be advocating matching the levels found in the host (even if it's E. Coli). As far as the sequences, we're making about 14 or so different strains of a poly-protein, so that's probably not practical. $\endgroup$
    – Atl LED
    Commented Dec 18, 2014 at 14:26
  • $\begingroup$ @Atl LED, thanks! Just wanted to get the answers out before christmas since I thought it might be time critical for you. I will change the formatting and will look for the mentioned reference as soon as possbile. $\endgroup$
    – Jan
    Commented Dec 19, 2014 at 14:51

In a side-by-side comparison of 3 viral proteins (2 from paramyxoviridae, 1 from caliciviridae) each with both a "maximizing" and "matched" codon bias, we found the "maximized" version produced more protein the "matched" version in each instance. Protein levels were quantified via Western and flow, cell lines tested were 293T, HEp2 (HeLa contaminate), and Vero.

This is certainly not wide enough testing for any kind of general statement, it's just what we found. I think this will take a cell bio or pure micro-bio lab to fully study. We're in infectious diseases, and this question would be off grant and off topic for the paper. If we can find a way to squeeze it into a publication we will.

I'm going to accept Jan's answer because it contains more information on optimization, but with this answer in place to:

  1. Address what we observed.
  2. Note that Jan's doesn't offer any information on "matching" the codon bias of the target host (other than perhaps noting optimization is heuristic).

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .