Troubleshooting and Optimization

Low expression is the most common bottleneck in biologics development, and the root cause is rarely where people first look. Before rewriting your codon table or switching expression hosts, check the fundamentals: Is your signal peptide compatible with your expression system? Is there a cryptic splice site or internal ribosome entry site in your construct? Is the protein toxic to the host cell? These are quick checks that eliminate the simplest failure modes before you invest in optimization.

If the construct is correct and the protein still doesn't express, the problem is usually folding. Disulfide-rich proteins often misfold in reducing cytoplasmic environments—moving to a periplasmic or secreted expression system can help. Proteins with extensive hydrophobic surfaces may aggregate during overexpression; lowering induction temperature (16–20 °C), reducing inducer concentration, or co-expressing chaperones (DnaK/DnaJ, GroEL/GroES) can improve yield. For antibodies and antibody fragments, the heavy chain:light chain ratio matters—imbalanced expression drives aggregation and reduces functional yield.

Aggregation kills biologics programs more often than any other developability issue. It reduces yield, creates heterogeneous preparations that confound assays, and raises immunogenicity concerns. The first step is diagnosis: is the aggregation occurring during expression (misfolding), during purification (pH or concentration-driven), or during storage (colloidal instability)?

SEC-MALS, DLS, and nanoDSF are your diagnostic workhorses. SEC-MALS tells you the oligomeric state and aggregate fraction. DLS detects sub-visible particles. NanoDSF gives you melting temperature (T_m) and onset of aggregation temperature (T_agg), which are the two most predictive metrics for long-term stability. If T_m is below 55 °C or T_agg is within 5 °C of your storage temperature, you have a problem that formulation alone is unlikely to solve—you need to engineer the molecule.

Common engineering fixes include removing unpaired cysteines, introducing proline substitutions in aggregation-prone regions, replacing surface-exposed hydrophobic patches with polar residues, and adding stabilizing disulfides. Computational tools like CamSol and Aggrescan3D can identify problematic regions and suggest mutations, but these predictions should always be validated experimentally—in silico aggregation scores correlate with but do not determine experimental behavior.

The hardest decision in protein engineering is knowing when to stop optimizing a problematic molecule and start over with a different scaffold, format, or lead. There is no universal rule, but there are useful heuristics. If three rounds of rational design or directed evolution haven't moved your key metric (affinity, expression, stability) by at least 3-fold, the problem is likely structural rather than parametric—you need a different starting point, not more mutations on the same one. If your molecule requires heroic formulation (high surfactant, narrow pH range, arginine stabilization) to stay soluble at therapeutic concentrations, it will be difficult to develop regardless of how good its binding data looks.

The best programs build pivot criteria into their project plans from the start. Define what "good enough" looks like for expression, stability, and potency at each stage, and commit to revisiting the lead selection if those thresholds aren't met. Sunk cost is the enemy of good decision-making in drug discovery.