Heterologous Protein Expression Hosts by Product Class

Why Product Class Should Drive Your Host Decision

The choice of heterologous protein expression host is the single most consequential decision in recombinant protein production, yet most guides organize it backwards. They start with the host (E. coli, then yeast, then mammalian) and list what each can do. The problem: process developers rarely begin with a blank slate. They know the product. A peptide hormone, a monoclonal antibody, an industrial protease, a virus-like particle. The host follows from the product's biology, not the other way around.

This article takes the product-class-first approach. You arrive knowing your protein and leave knowing the 1–2 hosts most likely to work. For the complementary technology-first comparison (E. coli vs Pichia vs CHO vs insect cells with detailed titer data, timelines, cost curves and a decision tree), see our expression system comparison guide.

Five product classes account for the vast majority of heterologous protein expression worldwide:

1. Peptide hormones and small therapeutics — non-glycosylated proteins under ~30 kDa (insulin, hGH, interferons, G-CSF)
2. Monoclonal antibodies and glycosylated biologics — IgGs, Fc-fusion proteins, bispecifics
3. Industrial enzymes — proteases, amylases, lipases, cellulases produced at ton scale
4. Complex structural proteins and VLPs — multi-subunit assemblies, virus-like particles for vaccines
5. Pathway products — small molecules made via engineered metabolic pathways (artemisinin precursors, terpenes, flavonoids)

Each class has a dominant expression host, driven not by tradition but by the decisive biological constraint that rules out the alternatives. The sections below walk through each class, name that constraint, and give the titer ranges and regulatory context a process developer needs.

Peptide Hormones and Small Therapeutics

E. coli is the default heterologous protein expression host for non-glycosylated therapeutic proteins smaller than approximately 30 kDa. The decisive constraint is simple: these products do not need glycosylation, so the speed and cost advantages of E. coli (20–30 minute doubling time, media under $1/L, gene-to-protein in 2–5 days) far outweigh the refolding burden.

Recombinant human insulin, the first approved therapeutic protein produced by heterologous expression in E. coli, established the template in 1982. The proinsulin precursor accumulates as inclusion bodies at 20–30% of total cell protein, then undergoes solubilization in 6–8 M urea or guanidine HCl, oxidative refolding to form the correct disulfide bonds, and enzymatic processing (trypsin + carboxypeptidase B) to excise the C-peptide. Despite this multi-step downstream process, E. coli remains the dominant production host for insulin and its analogs (lispro, aspart, glargine).

Pichia pastoris offers a secreted alternative. A 2010 study demonstrated approximately 1.5 g/L of purified recombinant human insulin from optimized Pichia fed-batch culture, where the insulin precursor was secreted into the medium and then enzymatically converted. Secretion eliminates the inclusion body solubilization step, but Pichia's longer development timeline (1–2 weeks for stable integrants) and the risk of hyperglycosylation on O-linked sites mean it remains the secondary choice for this product class.

Other products in this class follow the same logic. Human growth hormone (hGH, 22 kDa), granulocyte colony-stimulating factor (G-CSF, 19 kDa), and interferons (IFN-alpha, 19 kDa) are all produced commercially in E. coli. Each lacks glycosylation requirements, each tolerates inclusion body refolding, and each benefits from the sub-$50/g production cost that E. coli enables at scale.

For optimizing the expression conditions (IPTG concentration, induction temperature, feed rate) in any E. coli campaign, codon optimization and fusion tag selection are the two design levers that most influence soluble yield before you reach for DOE.

Monoclonal Antibodies and Glycosylated Biologics

CHO cells produce over 80% of all approved monoclonal antibodies, and this dominance is driven by a single biological constraint: N-linked glycosylation. The Fc region of an IgG antibody carries a conserved N-glycan at Asn297 that is essential for effector functions including antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC). Only mammalian expression hosts produce the human-like complex biantennary glycans (G0F, G1F, G2F) that regulators and clinicians expect.

Modern CHO fed-batch processes routinely achieve 5–10 g/L monoclonal antibody titer, with some proprietary platforms reporting above 10 g/L. Glycosylation control during the culture (temperature, dissolved oxygen, manganese supplementation, pH, and media composition) determines the final glycan profile and is a critical quality attribute (CQA) subject to regulatory specification.

HEK293 cells serve a complementary role. They are the preferred heterologous protein expression host for viral vectors (AAV, lentivirus) because they support efficient viral replication and packaging. For therapeutic proteins, HEK293 is chosen when the molecule folds poorly in CHO or when rapid material supply is needed: transient transfection in HEK293 can deliver milligram quantities of a glycosylated protein in 1–2 weeks, versus the 3–6 months required for stable CHO cell line development.

Fc-fusion proteins (etanercept, abatacept, aflibercept) and bispecific antibodies also require CHO or HEK293 for the same glycosylation reason. The regulatory precedent is overwhelming: as of 2026, no monoclonal antibody or Fc-fusion protein produced in a non-mammalian host has been approved by the FDA or EMA.

The mAb Market: What the Numbers Actually Say

The global monoclonal antibody market is large, but how large depends on who you ask. Research firms define the market boundary differently (therapeutics only vs. diagnostics included, biosimilars in or out, companion diagnostics), and the result is a spread of estimates that tells you more about methodology than about the underlying market.

Bar chart showing five market size estimates for the 2026 monoclonal antibody market: Business Research Company at 288.8 billion USD, Grand View Research at 294.1 billion, Global Market Insights at 319.5 billion, GII Research at 331.8 billion, and Fortune Business Insights at 344.0 billion.

The honest reading: the global mAb therapeutics market in 2026 is somewhere in the range of $289–$344 billion, growing at 10–15% CAGR. The precise number matters less than the scale. CHO cells and their supporting infrastructure (media suppliers, CDMO capacity, analytical methods) are the backbone of this market, and any new glycosylated biologic will almost certainly be produced in CHO.

Industrial Enzymes: Bacillus subtilis and the Secretion Advantage

Bacillus species manufacture approximately 60% of all commercial industrial enzymes. The decisive constraint for this product class is not folding complexity or glycosylation but economics: industrial enzymes are commodity products sold at $5–50/kg, so the expression host must secrete the active enzyme directly into the fermentation broth at gram-per-liter levels to keep downstream processing costs viable.

Bacillus subtilis is the workhorse of this product class for three reasons that E. coli cannot match:

• Direct secretion. B. subtilis has no outer membrane and naturally secretes proteins into the extracellular medium via the Sec and Tat pathways. Secreted titers of 1–25 g/L are routine for well-optimized enzymes. E. coli retains most recombinant protein in the cytoplasm, requiring cell lysis and additional purification.
• GRAS status. B. subtilis carries Generally Recognized As Safe (GRAS) status from the FDA and Qualified Presumption of Safety (QPS) certification from EFSA. This is critical for food-grade and feed-grade enzymes (detergent proteases, baking amylases, textile cellulases) where the regulatory burden of a non-GRAS host would be prohibitive.
• No endotoxin. As a Gram-positive organism, B. subtilis produces no lipopolysaccharide (LPS). For pharmaceutical-adjacent applications (enzyme reagents, diagnostics) this eliminates the endotoxin removal step that adds cost and yield loss to E. coli-derived products.

B. licheniformis is the other major Bacillus production host, often chosen for thermostable amylases used in starch processing. Together, B. subtilis and B. licheniformis cover the majority of commercial enzyme production. For more on Bacillus fermentation at scale (sporulation control, dissolved oxygen management, autolysis prevention), see our B. subtilis fermentation guide.

Filamentous fungi (Aspergillus niger, Trichoderma reesei) dominate cellulase and certain food enzyme production where secretion titers can exceed 100 g/L of total protein. However, their slower growth, morphology-dependent productivity, and more complex genetic engineering make them less versatile than Bacillus for new enzyme development.

Optimizing expression host performance for a new industrial enzyme is a classic screening problem. After selecting B. subtilis or B. licheniformis as the chassis, the variables that most influence secreted titer (signal peptide, promoter strength, growth temperature, media composition, induction timing) are best explored with a structured design of experiments approach rather than one-factor-at-a-time testing.

Stacked horizontal bar chart with four categories. Therapeutic proteins: CHO/HEK approximately 70%, E. coli 20%, other 10%. Industrial enzymes: Bacillus approximately 60%, filamentous fungi 25%, other 15%. Food/feed enzymes: Bacillus approximately 45%, filamentous fungi 40%, yeast 15%. Research-grade proteins: E. coli approximately 65%, yeast 15%, insect 10%, mammalian 10%.

Complex Proteins and VLPs: Insect Cell-Baculovirus Expression

The baculovirus expression vector system (BEVS) in Sf9 or Hi5 insect cells is the standard heterologous protein expression platform for virus-like particles and complex multi-subunit protein assemblies. The decisive constraint here is the need for eukaryotic folding and post-translational processing (disulfide bond formation, cleavage of signal peptides, limited glycosylation) combined with the ability to co-express multiple genes simultaneously from a single baculovirus vector.

Two licensed vaccines demonstrate the platform's strengths at commercial scale:

• Cervarix (GSK) — HPV-16/18 VLP vaccine produced in Hi5 insect cells using BEVS. The L1 capsid protein self-assembles into VLPs during expression.
• FluBlok (Sanofi Pasteur) — Recombinant influenza vaccine expressing hemagglutinin (HA) in expresSF+ insect cells. The cell-free, egg-free manufacturing platform enables faster strain updates for seasonal flu.

Typical BEVS titers range from 50 to 500 mg/L for soluble proteins, though VLP yields depend heavily on assembly efficiency rather than raw protein expression. The main limitations are the lytic nature of baculovirus infection (cells lyse at 48–72 hours post-infection, releasing proteases and DNA into the harvest) and the insect-cell glycosylation pattern, which produces paucimannose (Man3GlcNAc2) rather than complex sialylated glycans.

For applications requiring human glycosylation on a complex protein, stable CHO expression is preferred. For VLPs, structural proteins, and multi-subunit complexes where the glycan structure is not a CQA, BEVS offers faster development (weeks vs months) and excellent folding fidelity. For more on optimizing MOI and harvest timing, see our guide on baculovirus MOI and harvest optimization.

Pathway Products and Plant Natural Compounds: Engineered Yeast

Saccharomyces cerevisiae is the go-to heterologous expression host when the product is not a protein but a small molecule synthesized by an engineered metabolic pathway. The decisive constraint is pathway length and compartmentalization: yeast has the endomembrane system (ER, mitochondria, peroxisomes) needed to house plant cytochrome P450 enzymes that bacteria cannot functionally express, and decades of metabolic engineering tools (CRISPR, homologous recombination at near-100% efficiency, well-characterized promoter libraries) make it the fastest chassis for building multi-gene pathways.

The most commercially significant example is artemisinic acid, the precursor to the anti-malarial drug artemisinin. In 2006, Ro et al. demonstrated production of artemisinic acid at up to 100 mg/L in engineered S. cerevisiae by expressing the amorphadiene synthase gene and a plant cytochrome P450 (CYP71AV1) from Artemisia annua. Amyris subsequently optimized the strain and transferred it to Sanofi, which launched commercial production of semi-synthetic artemisinin in 2013, supplying approximately one-third of global demand.

Other pathway products produced in engineered yeast at commercial or pilot scale include:

• Farnesene (Amyris) — a sesquiterpene converted to squalane for cosmetics and fuels. Production exceeds 1,000 tonnes/year.
• Resveratrol (Evolva/DSM) — the stilbene antioxidant, produced by expressing a stilbene synthase and feeding coumaric acid.
• Naringenin and other flavonoids — phenylpropanoid pathway reconstruction for food and nutraceutical applications.
• Recombinant spider silk proteins — while E. coli achieves higher protein titers (up to 14.5 g/L for small spidroins), yeast and Bacillus hosts are explored for secreted production that avoids inclusion body refolding.

Pichia pastoris also plays a role in pathway products, particularly where the product benefits from methanol-inducible strong expression (AOX1 promoter) or where the heterologous pathway includes membrane-bound enzymes. Cell-free systems offer a complementary approach for pathway prototyping at microgram scale before committing to a cellular host.

How Large Is the Recombinant Protein Market?

The global recombinant protein market was valued at approximately $600–$800 billion in 2025 when therapeutic proteins, industrial enzymes, diagnostic reagents, and research-grade proteins are included. The monoclonal antibody segment alone accounts for roughly $289–$344 billion depending on which research firm you ask (Figure 1). The industrial enzyme market adds another $7–$9 billion, and research reagents contribute approximately $3–$5 billion.

The spread across market estimates is instructive. For the mAb segment alone, five research firms publishing 2026 estimates disagree by 19%:

• Business Research Company: $288.8 billion
• Grand View Research: $294.1 billion
• Global Market Insights: $319.5 billion
• GII Research: $331.8 billion
• Fortune Business Insights: $344.0 billion

This spread reflects methodology, not uncertainty. Some firms include biosimilars; others exclude them. Some count companion diagnostics; others do not. The takeaway for a process developer is not which number is "right" but that the market is measured in hundreds of billions of dollars, growing at double-digit rates, and overwhelmingly served by CHO cell expression.

The industrial enzyme market (dominated by Bacillus expression) is two orders of magnitude smaller in revenue but vastly larger in tonnage. Detergent proteases alone account for 900 tonnes of pure enzyme per year in Europe. This tonnage asymmetry explains why Bacillus expression economics are built around secretion yield and media cost per kilogram, while CHO economics focus on product quality and regulatory compliance per milligram.

Five-Question Framework for Host Selection

Narrowing from five candidate expression hosts to one or two takes five questions, asked in order. Each question eliminates hosts that cannot meet the product's biological requirements. The order matters: glycosylation is the strongest filter (it eliminates all prokaryotes), followed by disulfide complexity, subunit count, volume economics, and regulatory pathway.

Decision map showing five product classes mapped to expression hosts: small non-glycosylated therapeutics to E. coli or Pichia (constraint: no glycosylation needed), glycosylated biologics to CHO or HEK293 (constraint: human-like N-glycosylation), industrial enzymes to Bacillus or fungi (constraint: GRAS plus secretion), complex proteins and VLPs to insect cells via BEVS (constraint: multi-gene eukaryotic folding), and pathway products to engineered S. cerevisiae (constraint: plant P450 compartments).

The five questions, in order:

1. Does the product need N-linked glycosylation? If yes, you need a mammalian host (CHO, HEK293). This single question eliminates E. coli, Bacillus, and yeast for any product where the glycan profile is a CQA.
2. How many disulfide bonds, and how complex is the fold? Proteins with 0–2 simple disulfides can be refolded from E. coli inclusion bodies or expressed in the periplasm. Proteins with 4+ disulfides or complex domain architectures favor eukaryotic hosts with ER-resident chaperones.
3. Is the product a multi-subunit assembly? VLPs, heterodimeric proteins, and complexes requiring co-expression of 3+ genes point toward BEVS (insect cells) or mammalian transient systems.
4. What is the volume economics target? Commodity enzymes at $5–50/kg demand high-secretion prokaryotic hosts (Bacillus). Therapeutic proteins at $100–1,000/mg can tolerate the higher per-gram costs of mammalian cell culture.
5. What is the regulatory pathway? An IND filing for a therapeutic biologic favors CHO (deepest regulatory precedent). A food enzyme filing favors Bacillus (GRAS). A research reagent has no regulatory constraint on host choice.

Frequently Asked Questions

Related Calculators

• Fermentation Economics Calculator — Model COGS per gram for E. coli, Pichia, and Bacillus expression at different scales.
• Media and Feed Estimator — Estimate media volumes and costs for fed-batch and perfusion campaigns.
• Scale-Up Calculator — Translate bench-scale expression results to pilot and production bioreactors.

References

1. Rosano GL, Ceccarelli EA (2014). Recombinant protein expression in Escherichia coli: advances and challenges. Frontiers in Microbiology, 5:172. doi:10.3389/fmicb.2014.00172
2. van Dijl JM, Hecker M (2013). Bacillus subtilis: from soil bacterium to super-secreting cell factory. Microbial Cell Factories, 12:3. doi:10.1186/1475-2859-12-3
3. Wurm FM (2004). Production of recombinant protein therapeutics in cultivated mammalian cells. Nature Biotechnology, 22:1393–1398. doi:10.1038/nbt1026
4. Ro DK et al. (2006). Production of the antimalarial drug precursor artemisinic acid in engineered yeast. Nature, 440:940–943. doi:10.1038/nature04640
5. Kost TA, Condreay JP, Jarvis DL (2005). Baculovirus as versatile vectors for protein expression in insect and mammalian cells. Nature Biotechnology, 23:567–575. doi:10.1038/nbt1095