Here’s why. CHO is not a monolith. It became the industry default not because it’s optimal for every molecule — it isn’t — but because it’s good enough for most, and good enough combined with ubiquitous infrastructure and regulatory familiarity is an almost unbeatable combination. When you’re under timeline pressure and program risk is high, you don’t run platform selection experiments. You use what you know. That’s rational. It’s also how you end up using the same platform for molecules where it genuinely isn’t the best fit.

The right question is: for what molecule, in what manufacturing context, does a novel host give you a meaningful and defensible advantage in cost, speed, or a built-in quality attribute that CHO cannot match without extensive engineering?

Three dimensions where the answer can be yes.

1. Cost structure. If your product doesn’t require commercial-scale multi-ton production, and a simpler or lower-capital platform can meet your GMP requirements at your volume, the economics may favor a novel host.
2. Speed. Three weeks from gene sequence to first protein is not a marginal improvement on six months. For pandemic response or first-to-clinic competitive scenarios, that difference changes the entire strategic calculus.
3. Intrinsic product quality. If your molecule needs afucosylation for efficacy, and one platform delivers that natively while CHO requires glycoengineering, the novel host isn’t exotic — it’s just the better fit.

In Part 1, we covered five platforms with five honest assessments: moss with Phase 1b clinical data and an EMA viral testing waiver, microalgae with a compelling sustainability case and a 1,000-fold yield gap, plant pharming with a fully approved vaccine and the fastest gene-to-protein timeline in the field, silkworms with PMDA Phase I alignment and an unresolved batch consistency challenge, and cyanobacteria as the most distant from clinical reality of the five. Now the question is: what does all of that actually mean for how you make host selection decisions?

The Three Platforms with the Clearest Near-Term Path

Based on the clinical and regulatory evidence available right now, three platforms have demonstrated a credible near-term path to human biologics.

Plant Pharming — Nicotiana benthamiana

One approved product. The fastest sequence-to-dose timeline in the field. Techno-economic data validating cost of goods competitive with CHO for VLPs. This is not a speculative platform. It has been validated in a real regulatory process with a real approval.

The clearest fit: VLP-based vaccines, pandemic response, and decentralized manufacturing in settings where bioreactor capital and cold-chain logistics are real constraints, not theoretical ones. If you need a vaccine candidate within weeks of identifying a new pathogen sequence, there is no other platform that matches this.

⚠️The honest constraint: there’s no master cell bank equivalent in transient plant expression. Consistency depends on plant growth conditions and the efficiency of Agrobacterium infiltration. Covifenz showed this is manageable. But it is a regulatory dimension you don’t encounter with a clonal CHO line, and you need to plan for it from the start.

Insect Cells — Sf9 and Hi-5 with baculovirus

I think this is the most underappreciated platform in the alternative host conversation, partly because it’s been around long enough that people take it for granted. Flublok and Nuvaxovid are FDA-approved and commercially manufactured at scale. Cervarix was FDA-approved but withdrawn from the US market in 2016 — it remains approved and in use in other markets. Several major CDMOs have the infrastructure. The platform is de-risked in a way that most novel hosts aren’t.

For VLPs and large multimeric proteins that don’t express well in mammalian systems, insect cells are the best available option. The glycan profile — paucimannosidic structures — differs from mammalian, which is limiting for therapeutic mAbs but actually advantageous for vaccine applications where you want immune activation.

⚠️The real constraint: baculovirus kills the cells. The infection cycle is lytic. Cells die at harvest, which makes continuous processing difficult and fundamentally constrains this as a batch platform.

Cell-Free Protein Synthesis — Sutro Biopharma’s XpressCF platform

This is the one I think gets the least attention, and it may have the most under-appreciated near-term relevance for a specific and growing application. Sutro has ADC candidates in clinical trials produced using a cell-free system. No cells. No traditional bioreactor.

The differentiating capability: this is the only production system that allows you to site-specifically incorporate non-natural amino acids at defined positions in the protein sequence. For next-generation ADCs — where the drug-to-antibody ratio and the exact conjugation site both matter for efficacy and safety — that’s not a nice-to-have. It’s the entire technical basis of the differentiation.

⚠️The honest constraint: cost of goods is roughly 80 percent higher than CHO for bulk protein production. For a standard mAb where site-specific chemistry isn’t required, CHO is the right economic choice. Cell-free wins when you’re making something CHO fundamentally can’t produce — or when the speed and decentralization advantages outweigh the cost premium.

Two Platforms Worth Watching on a 5–10 Year Horizon

Two platforms that aren’t ready for broad clinical use today, but have the data trajectory to become serious contenders.

Moss — Eleva

Phase 1b clinical data for Fabry disease. EMA viral testing waivers. Built-in afucosylation without glycoengineering. That is a substantive package — not just a compelling technology story, but an active clinical program with real regulatory engagement.

⚠️The watch signal: if a moss-derived afucosylated oncology monoclonal antibody advances to Phase III, the platform’s credibility for therapeutic antibodies shifts materially. Right now, the regulatory precedent is enzyme replacement therapy. An oncology mAb in late-stage trials is a completely different regulatory category, and that signal would change the risk calculus for a lot of programs.

Filamentous Fungi — Dyadic’s C1 platform

Phase I cleared for DYAI-100, a COVID-19 subunit vaccine. GRAS status. Thermophilic growth at 50 degrees Celsius, which is a real practical advantage for contamination control. High secretion capacity that rivals CHO for some protein classes.

In March 2025, CEPI awarded a grant to Fondazione Biotecnopolo di Siena (FBS) for pandemic preparedness research using Dyadic’s C1 technology. The Rubic Africa manufacturing partnership is active, targeting vaccine production for low- and middle-income countries where cost and infrastructure constraints are the dominant variables.

⚠️The watch signal: Phase II or Phase III data for DYAI-100 or a successor would move this platform from “interesting single cleared IND” to “proven in a real disease program.” That is the threshold where risk perception changes for regulatory reviewers.

What Should Actually Drive Host Selection

Let me give you a concrete framework — something you can actually use when you’re sitting in a program initiation meeting and someone asks which platform you’re using.

➡️ By molecule type:

Glycosylated mAb with standard Fc function — CHO. The infrastructure, the regulatory familiarity, the CDMO availability. Nothing displaces it for this application.

Non-glycosylated peptide or protein — E. coli or yeast, depending on complexity and folding requirements.

VLP or recombinant vaccine — insect cells for batch scale, plant pharming if you need speed or decentralization.

ADC requiring site-specific non-natural amino acid incorporation — cell-free protein synthesis. There is no other viable route for this product class at present.

Afucosylated oncology mAb where ADCC is the therapeutic mechanism — seriously evaluate moss. Not because it’s novel, but because the biology delivers what you need from the start.

➡️ By manufacturing context: pandemic speed or emergency response, plant pharming is the only real answer on timeline. Decentralized or LMIC manufacturing with cold-chain constraints, plant pharming or lyophilized cell-free formats. Maximum ADCC potency without the engineering burden, moss.

➡️ By regulatory position: insect cells and plant pharming have the most clinical and commercial precedent today. Moss is building that precedent in Phase II. Cell-free protein synthesis is building it specifically for ADCs.

The decision is not about picking the most exciting platform. It’s about matching platform biology to molecule requirements, manufacturing constraints, and where regulatory precedent actually exists.

Closing

Here’s my honest summary, and I’ll say it plainly: CHO is not being replaced. But I also think that framing has become a distraction.

The real development over the past decade is that bioprocess scientists now have a credible, validated toolkit of alternatives for specific molecules in specific contexts. That toolkit did not exist 20 years ago. Regulatory agencies weren’t equipped to review it. Clinical data didn’t support it. Today, both of those things have changed — and that change is significant.

The practical implication: know this landscape well enough to ask the right host selection question at program initiation, before you’ve built months of process development around a platform you chose by default.

And before I close — one more thing I want to leave you with.

Of all five platforms we reviewed, the silkworm is the only one where the upstream production system is a living organism that fits in the palm of your hand. No bioreactor. No temperature control systems. No complex media formulation. KAICO has PMDA alignment for a Phase I VLP trial. The infrastructure cost is a fraction of stainless steel.

If the batch consistency challenge is solved, this will raise a disruptive question: what if the next generation of biologics is not manufactured in bioreactors at all? What if they’re produced in pupae, in small distributed facilities, close to the patients who need them?

That is not science fiction. That is the direction the data is pointing.

And here’s the pattern I keep coming back to: the platforms that look exotic at one moment in time tend to look obvious in retrospect. High-throughput screening seemed extreme in the early 1990s. Continuous manufacturing seemed like a regulatory bridge too far just ten years back, industry 4.0 factories seemed impossible.

The smart biotech scientist doesn’t wait to be surprised by what becomes standard.

Show notes include links to all five guest episodes — moss, microalgae, plant pharming, silkworms, and cyanobacteria.