Secretory Pathway Engineering: Improving Protein Export in Yeast and Mammalian Hosts

When a target protein accumulates in the cytoplasm instead of being secreted, the consequences compound quickly. Downstream purification requires cell lysis, which introduces host-cell protein contamination and increases buffer volumes. Product quality can suffer from incomplete folding in the cytosolic environment. And the cost-per-gram climbs. Secretory pathway engineering is one of the most effective interventions available to bioprocess teams — but it's also one where the wrong choice wastes more time than it saves.

This post covers the main secretory routes in yeast (Pichia pastoris, S. cerevisiae) and mammalian (CHO) hosts, the common bottlenecks in each, and the practical engineering strategies we've applied across our partner programs.

The Three Main Secretory Routes

Not all secretion is the same. Understanding which pathway your chassis uses — and which your target protein is compatible with — determines which interventions are worth attempting.

The Sec (general secretory) pathway

Used by bacteria and partially by yeast for short, unstructured proteins. In E. coli, the Sec pathway translocates unfolded polypeptides post-translationally through SecYEG translocon complexes into the periplasm. It works well for small proteins that can remain unfolded until translocation. For proteins that begin folding co-translationally in the cytoplasm, Sec translocation stalls. The classic symptom is periplasmic inclusion body formation — the protein gets partway through translocation and aggregates at the membrane.

The SRP (signal recognition particle) pathway

The SRP pathway is the dominant co-translational route in eukaryotes — Pichia, S. cerevisiae, CHO, and other mammalian hosts. As the ribosome synthesizes the signal peptide, SRP recognizes the hydrophobic core and arrests translation temporarily while the ribosome-mRNA-nascent chain complex docks to the ER membrane. Translocation proceeds co-translationally, reducing the window for cytosolic aggregation. Most secreted therapeutic proteins and most yeast-secreted enzymes use this route.

The Tat (twin-arginine translocation) pathway

The Tat pathway is the specialist. It translocates fully folded proteins across the inner membrane of bacteria (or thylakoid membrane in chloroplasts). The twin-arginine motif in the signal sequence is required. Tat is useful when your target protein coordinates a cofactor or forms a disulfide bond that must happen before translocation — conditions where Sec would fail. The tradeoff is that Tat pathway capacity is limited and typically lower-yield than Sec.

Common Bottlenecks in Yeast Secretion

In our work with Pichia pastoris and S. cerevisiae, three bottlenecks account for the majority of poor secretion cases.

Signal peptide mismatch

The signal peptide is the first variable to test when secretion is low. The canonical choice for Pichia is the S. cerevisiae alpha-mating factor prepro sequence — it's used in the majority of published protocols and works well for many targets. But we've seen cases where switching from the alpha-MF signal to a synthetic signal peptide designed for the specific target increased secreted titer by 3.5-fold without any other change. Signal peptide optimization is fast to run in parallel and high-return when secretion is the bottleneck.

ER folding capacity limitation

Overexpression of recombinant secretory proteins saturates the ER chaperone pool — BiP (Kar2p in yeast), PDI, and protein disulfide isomerase family members. The result is unfolded protein response (UPR) activation, which induces ER stress and eventually reduces viability. Co-overexpression of BiP or PDI is a documented intervention that improves secretion of disulfide-bonded proteins in both Pichia and CHO. We've seen 25–60% improvements in secreted titer from BiP co-expression alone for certain targets. It's not universally effective — for proteins where folding isn't the bottleneck, it has no impact.

Kex2 and Ste13 processing errors

When using the alpha-MF prepro signal, the mature protein is released by Kex2 protease cleavage at a KR or KRLA site. If the target protein's N-terminus has additional Glu-Ala repeats that aren't efficiently removed by Ste13 dipeptidyl aminopeptidase, you end up with N-terminal heterogeneity in the secreted product. For therapeutic applications, this is a quality attribute that can create regulatory discussion. The fix is usually to engineer the construct junction to remove the Glu-Ala repeats entirely and rely only on the Kex2 KR cleavage site.

Common Bottlenecks in CHO Secretion

Mammalian secretory pathway optimization follows different logic from yeast. CHO cells are generally better at folding complex glycoproteins, but have their own set of constraints.

Signal peptide competition

In CHO, endogenous signal peptides generally outperform the heterologous signals sometimes imported from yeast or bacterial work. When possible, use a validated mammalian signal peptide — the human immunoglobulin kappa light chain signal and the mouse IgG heavy chain signal are two frequently used sequences with broad compatibility. Testing 3–4 signal peptide candidates in a 96-well microfermentation format before committing to a development construct adds only two weeks but can prevent a suboptimal choice from propagating through the entire cell bank development process.

Golgi processing and glycan site occupancy

N-linked glycosylation site occupancy is not guaranteed. Sites that are theoretically glycosylated based on NXS/T sequon may be partially occupied depending on the local protein structure and competition for OST complex capacity. Incomplete glycan occupancy leads to product heterogeneity. If occupancy is a concern, testing alternate signal peptide and promoter strength combinations can shift the rate of synthesis relative to Golgi processing capacity and improve site occupancy.

Endoplasmic reticulum exit rate

Some proteins fold correctly in the ER but exit slowly — they accumulate at the ER-Golgi interface rather than progressing to the secretory vesicle pathway. This is sometimes caused by off-rate interactions with ER retention machinery (KDEL-receptor saturation, ERAD pathway competition). Identifying this bottleneck requires comparing intracellular localization at different time points post-induction. It's a less common bottleneck but worth ruling out if your protein is well-folded by biochemical assay but secretion remains low.

Practical Engineering Interventions — What We Test First

Based on our experience across yeast and mammalian programs, here is the order in which we typically apply interventions when secretion is below target:

1. Signal peptide panel. Test 3–5 candidates in parallel. Two weeks, high impact potential, no downstream consequences if the winner is different from the starting candidate.
2. Chaperone co-expression. BiP and PDI are the first co-expression constructs to try for disulfide-bonded targets. Low-risk intervention with documented upside.
3. Promoter strength reduction for very high expression targets. Counter-intuitive, but reducing expression rate can increase secreted yield by giving the folding machinery time to process each molecule. We've seen cases where halving induction strength increased secreted titer by 40%.
4. Host strain selection. For yeast programs, moving from a standard X-33 Pichia to a UPR-attenuated strain or a strain with modified ERAD pathway genes can improve secretion of difficult targets without any construct-level engineering.
5. Fermentation parameter optimization. Temperature reduction during induction (from 28°C to 20°C in Pichia, or 37°C to 33°C in CHO fed-batch) slows synthesis rate, reduces ER load, and frequently improves secreted fraction and product quality simultaneously.

When Secretion Engineering Isn't the Right Answer

Some targets are genuinely intractable for extracellular secretion in a given host. If a protein misfolds in the ER regardless of chaperone co-expression, has structural features incompatible with the secretory pathway, or requires cytosolic cofactors, the better answer is often to accept intracellular production and invest in a more efficient cell disruption and refolding workflow downstream. We've had two partner programs where six weeks of secretion optimization produced marginal gains, and the decision to pivot to periplasmic or cytosolic production with optimized recovery unlocked the program.

The question is always: where is the rate-limiting step? Secretion engineering helps when secretion is the bottleneck. When folding or host compatibility is the root constraint, pathway engineering adds cost and time without solving the underlying problem.