Why Is My Codon-Optimized Gene Expressing Poorly? Common Mistakes to Check
7 min read · Updated July 10, 2026
A codon-optimized gene that still expresses poorly is one of the most frustrating results in the lab, because codon optimization is supposed to fix exactly this problem. The construct looks fine on paper, the codon usage checks out by every metric you ran, and the gel or western blot still comes back weak or empty. This guide walks through the likely causes in a useful order to check them, from problems the optimization step itself introduced to the possibility that the protein, not the DNA, is the real bottleneck.
Cause 1: naive optimization introduced new problems
The simplest optimization strategy substitutes the single most frequent codon for each amino acid, everywhere in the gene. That produces a high score on any codon-usage metric, but it is not how real highly expressed genes look, since natural coding sequences mix codons across the available synonyms rather than using the top choice at every position. Forcing one codon everywhere can create runs of identical or near-identical codons, unintended sequence repeats, and local or overall GC content that drifts toward an extreme the host handles poorly, in either direction. A codon-usage-only score like CAI will not flag any of this, because it measures how often the preferred codon was chosen, not the sequence context that choice creates.
Check GC content and scan for repeats specifically after optimizing, not just before. These are new properties of the optimized sequence, not something you would have caught reviewing the original protein-level design, and they are easy to miss if the only number you looked at was the overall CAI.
Cause 2: mRNA structure near the start codon
A gene can have excellent codon usage from the first residue to the last and still fail to express if the region around the ribosome binding site or start codon folds into a stable hairpin. Translation initiation happens at that specific stretch of mRNA, and a stable structure there is enough to block a ribosome from loading, no matter how well-optimized the rest of the gene is.
This is a structural property of the mRNA sequence, not a codon-frequency property, so a high CAI score gives no information about it one way or the other — it has to be checked separately. If codon usage looks fine and expression is still poor, checking secondary structure around the start codon and the 5' end, rather than re-running the same codon-usage check again, is one of the highest-yield things to look at next.
Cause 3: construct-level problems the back-translation step introduced
Back-translation invents a DNA sequence that did not exist in the original protein-level design, and that process can introduce construct-level problems a codon-usage check will not catch. Because most amino acids have several synonymous codons, the specific arrangement chosen at every position is new information, and a handful of coincidental choices is enough to create:
- A cryptic ribosome-binding-site-like motif appearing by chance somewhere in the coding sequence, giving the ribosome an internal site to misfire on.
- A premature in-frame stop codon in an alternate reading frame that matters for your expression system.
- An accidentally introduced restriction site that conflicts with the cloning plan.
- A locally extreme GC stretch, even when the sequence's overall average looks reasonable.
Why these need a separate QC pass
These are exactly the kind of problems that only show up after optimizing, since the optimization step itself created them. A QC pass on the finished coding sequence, not just the starting protein design, is what catches them, and it is worth doing even on a sequence that scored well on codon usage.
Cause 4: the protein itself is the bottleneck
If GC content, repeats, mRNA structure and construct-level motifs all check out and expression is still poor, the coding sequence is probably not the problem — the protein is. Aggregation-prone regions, long transmembrane segments, and sequences that are simply hard for the host's folding machinery to handle can all limit expression independent of how well the underlying gene was optimized.
A hydrophobicity plot flags long or unusually placed hydrophobic stretches worth factoring into the expression strategy before concluding that codon optimization itself failed. Depending on what turns up, that might mean planning for inclusion bodies and refolding, switching to a different host, or adding a solubility-enhancing fusion tag, rather than reworking the coding sequence again for a problem it does not actually have.
Cause 5: conditions outside the coding sequence
Promoter strength, plasmid copy number, induction conditions, and host strain choice can each independently limit expression regardless of how well-optimized the gene is. Before spending more time on the sequence itself, it is worth confirming these basics are in order, since a poor result does not automatically indict the codon optimization step at all.
What a CAI score does and doesn't tell you
A good CAI score confirms one thing: that codon usage in the sequence matches the chosen host's preferences. It says nothing about GC extremes, repeats, mRNA structure, cryptic motifs, or how the protein itself will fold once made. Treat codon optimization as one necessary check among several, not the whole diagnosis, and re-verify GC content, repeats and motifs specifically after optimizing rather than assuming a high CAI score settles the question on its own.
SeqBench's Codon Adaptation Index tool scores a coding sequence against a chosen expression host, so you can check where a gene actually stands before concluding optimization worked or failed. If the causes above point to the sequence needing rework, the Codon Optimizer tool re-optimizes a protein or CDS for E. coli, human, yeast, CHO, Pichia and other hosts. And if the protein itself looks like the limiting factor, the Hydrophobicity Plot tool maps hydrophobic stretches across the sequence to help judge transmembrane or surface-exposure risk before ruling out the coding sequence as the cause.
Frequently asked questions
Why is my codon-optimized gene not expressing at all?
Work through the likely causes in order: check GC content and repeats introduced by the optimization itself, check mRNA structure around the start codon, check for cryptic motifs or stop codons the back-translation introduced, then consider whether the protein itself is prone to aggregation or hard to fold.
Does a high CAI score guarantee good expression?
No. CAI only measures how closely codon usage matches the host's preferences; it says nothing about GC extremes, repeats, mRNA secondary structure, cryptic motifs, or the protein's own folding behavior.
Can codon optimization make expression worse instead of better?
Yes, if it forces the single most frequent codon at every position without checking the result. That can create repeats or GC extremes the original sequence did not have, even while the CAI score goes up.
How do I know if my expression problem is the protein and not the codon optimization?
If GC content, repeats, mRNA structure and cryptic motifs all check out on the coding sequence, the protein itself is the likely bottleneck, particularly if a hydrophobicity plot shows long or unusually placed hydrophobic stretches consistent with aggregation or transmembrane behavior.
Related references
Related tools
Score a coding sequence's codon usage against an expression host, before optimising.
Optimise a protein or CDS for expression in E. coli, human, yeast, CHO, Pichia and more.
Sliding-window hydropathy plot to spot transmembrane and surface regions.