A diagnostic company's key antibody was producing 1.1 g/L at R&D scale — insufficient for commercial supply and with no validated scale-up process. AntibodyLLM rebuilt the upstream process from the ground up using a fed-batch platform approach, achieving 2.7 g/L yield with identical product quality from 500 mL shake flask all the way to 100L bioreactor.
The client, a diagnostics company supplying clinical testing laboratories, had developed a proprietary antibody that served as the capture reagent in their flagship immunoassay. At R&D scale — 500 mL shake flasks — the process produced approximately 1.1 g/L, which was adequate for early development but completely insufficient for commercial supply volumes.
Attempts to transfer the process to a 5L stirred-tank bioreactor had resulted in a significant yield drop to 0.7 g/L, along with increased high-molecular-weight aggregates that compromised product quality. The team could not determine whether the loss was due to process parameters, cell stress, or medium limitations — and had no roadmap for the 100L production scale their commercial forecast required.
They needed a robust, reproducible upstream process that could be transferred to contract manufacturing, with documented scale-up parameters and quality equivalence data across all three scales.
A shake flask process cannot be simply "transferred" to a bioreactor. The physical environment changes dramatically at scale, and cells that thrived in one format can underperform significantly in another without deliberate process re-engineering.
Increasing volume changes volumetric oxygen transfer coefficients (kLa) and shear stress profiles. CHO cells are shear-sensitive; over-agitation causes cell damage, under-agitation causes hypoxia and metabolic stress — both reduce productivity.
Batch culture at scale rapidly depletes glucose, glutamine, and key amino acids. Without a fed-batch feeding strategy tuned to the specific cell line's metabolic demands, cells shift into lactate-producing metabolism — toxic to the culture and wasteful of nutrients.
CO₂ dissolved in large-volume cultures shifts pH and inhibits cell growth. Headspace sparging strategies and bicarbonate buffering that work in small formats become inadequate at scale, requiring active pH control and CO₂ stripping strategies.
Three parallel workstreams — process development, scale engineering, and quality assurance — ran concurrently to compress the timeline and ensure that each scale transition was supported by data, not assumption.
The first step was diagnosing the 1.1 g/L ceiling in the original process. Metabolic profiling of the shake flask culture revealed that glucose was depleted by day 5 of a 14-day run, causing cells to shift to glutamine as the primary carbon source — generating excess ammonia, which is toxic to CHO cells. The culture entered a metabolic crisis before peak viable cell density was reached.
AntibodyLLM redesigned the feeding strategy using a Design of Experiments (DoE) framework: 24 parallel conditions varying glucose feed rate, glutamine substitution, and amino acid supplementation were run in miniature bioreactor ambr® systems. The optimized fed-batch strategy maintained glucose between 3–6 g/L throughout the run, eliminated ammonia accumulation, and extended the high-viability culture phase from 10 to 16 days.
The result at shake-flask equivalent scale: 2.4 g/L — more than a 2× improvement before any bioreactor-specific optimization.
The transition from shake flask to stirred-tank bioreactor required deliberate engineering of the physical culture environment. AntibodyLLM applied a constant kLa scale-up strategy: impeller speed and sparging rate at each scale were set to maintain equivalent oxygen transfer and mixing time, preventing the dissolved oxygen crashes that had caused aggregate formation in the client's earlier attempts.
At 5L scale, the optimized agitation and sparging parameters, combined with the new fed-batch feeding regime, produced 2.6 g/L — essentially matching the shake-flask result. This confirmed that the process was genuinely scale-independent rather than accidentally performing well in one format.
Scale-up to 50L and then 100L followed the same engineering principles. Active pH control (base addition + CO₂ stripping via overlay air) maintained culture pH at 7.0 ± 0.05 throughout — eliminating the CO₂ accumulation problem that typically increases with vessel volume. Both 50L and 100L runs achieved 2.7 g/L, within measurement uncertainty of each other.
Upstream improvements generate value only if the downstream purification process can handle the increased titer and maintain product quality. AntibodyLLM implemented a three-step purification platform: Protein A affinity capture, cation exchange chromatography (polishing step 1), and anion exchange flow-through (polishing step 2).
The Protein A loading density and elution gradient were re-optimized for the higher-titer feed to prevent column overloading. The polishing steps were specifically tuned to reduce the high-molecular-weight aggregates that had appeared in earlier bioreactor runs — targeting the root cause (process-induced cell stress) upstream while using the polishing train as a safety net.
Final product at 100L scale: >98% monomer purity by SEC-HPLC, with glycosylation profile, charge variant distribution, and biological activity all confirmed equivalent to the R&D-scale reference material. The process was documented with full batch records and scale-up rationale for CMO technology transfer.
| Scale | Process Version | Titer (g/L) | HMW Aggregates | Outcome |
|---|---|---|---|---|
| 500 mL Shake Flask | Original (batch) | 1.1 | <3% | Baseline — insufficient for scale |
| 5L Bioreactor | Original (batch) | 0.7 | ~8% | Failed — yield drop + aggregation |
| 500 mL Shake Flask | New (fed-batch, DoE) | 2.4 | <2% | Process breakthrough |
| 5L Bioreactor | New (fed-batch + kLa control) | 2.6 | <2% | Scale-independent confirmed |
| 50L Bioreactor | New (+ CO₂ management) | 2.7 | <2% | Pilot scale qualified |
| 100L Bioreactor | New (commercial) | 2.7 | <2% | Commercial scale — supply secured |
Process documented and ready for CMO technology transfer — full batch records, scale-up rationale, and quality equivalence data included.
The yield ceiling wasn't a cell line problem — it was a metabolic crisis caused by glucose depletion. Metabolic profiling identified the root cause in 48 hours. The right intervention (fed-batch feeding) fixed it. Changing medium or the cell line would have been the wrong fix.
Scale-up failures are usually physics failures: wrong mixing, wrong oxygen transfer, wrong CO₂ management. Matching kLa rather than impeller speed or volumetric flow rate is the correct scaling principle — and it's what kept yield constant across 200× volume increase.
Higher titer is worthless if product quality degrades. Tracking glycosylation, charge variants, and aggregate levels at every scale — not just the final one — is what makes a process "validated" rather than "lucky."
Whether you're hitting a yield ceiling at R&D scale or need a process ready for commercial manufacturing, our bioprocess engineering team has the platform and the expertise to get you there — with data at every scale to back it up.
Discuss Your Scale-Up