The Sartorius ambr250 is the most widely used high-throughput scale-down model for CHO fed-batch process development. It runs up to 24 vessels in parallel, each at 100-250 mL working volume, making it the workhorse for clone selection, media screening, and process characterization. But teams scaling down from 2,000 L or 15,000 L single-use bioreactors consistently report the same problem: ambr250 viability tracks manufacturing scale through the growth phase, then diverges by day 8-10, finishing 5-15 percentage points lower at harvest.
This viability gap is not a defect in the ambr250. It is a consequence of physics: a 200 mL vessel has fundamentally different gas transfer, evaporation, and sampling dynamics than a 2,000 L vessel. Understanding and mitigating these differences is the gap between a scale-down model that generates misleading data and one that reliably predicts manufacturing performance. This guide maps every root cause, quantifies each one, and gives you the engineering mitigations to close the gap.
Why Does Viability Drop Faster in the Ambr250?
The viability gap between the ambr250 and manufacturing scale is driven by five categories of engineering differences, not by cell biology. The same CHO clone in the same media at the same temperature, pH, and dissolved oxygen setpoint will reach different viability endpoints because the physical environment diverges as vessel size shrinks. Each root cause acts independently, but their effects compound over a 14-day run.
The critical insight is that these factors are not independent. Evaporation concentrates the medium, which raises osmolality, which stresses cells, which lowers viability, which changes metabolite profiles, which shifts pH control behaviour. A 200 mL vessel amplifies each link in this chain relative to a 2,000 L vessel. The compounding effect explains why the viability trajectories match early (days 0-7, when cumulative effects are small) and diverge late (days 8-14, when they are not).
Gas Transfer and CO2 Accumulation Differences
The ambr250 strips dissolved CO2 far more efficiently than manufacturing-scale vessels, typically maintaining pCO2 at 30-60 mmHg compared to 80-150 mmHg in a 2,000 L bioreactor. This difference alone can shift cell metabolism and pH control behaviour enough to reduce late-culture viability by 3-5 percentage points.
The physics is straightforward: CO2 leaves the liquid through both the sparged gas bubbles and the headspace surface. In a 200 mL vessel, the headspace-surface-to-volume ratio is roughly 10 times higher than in a 2,000 L vessel. Even with identical volumetric gas flow rates (vvm), the small vessel vents CO2 faster. The consequence is twofold:
- Lower steady-state pCO2 changes the intracellular pH and bicarbonate equilibrium, altering lactate metabolism and potentially shifting the lactate-to-consumption switch timing
- Less base addition because CO2-driven acidification is weaker, which means lower osmolality from base addition. This is the opposite of what happens at manufacturing scale, where base consumption drives osmolality above 350 mOsm/kg in late culture
| Parameter | Ambr250 (200 mL) | 5 L bench | 2,000 L SUB |
|---|---|---|---|
| Typical pCO2 (mmHg) | 30-60 | 50-100 | 80-150 |
| Headspace S/V ratio (m-1) | ~50 | ~15 | ~5 |
| CO2 stripping rate | High | Moderate | Low |
| Base consumption (relative) | 0.3-0.5x | 0.6-0.8x | 1x (reference) |
| Osmolality from base (mOsm/kg) | +10-20 | +20-40 | +40-80 |
Mitigation: To replicate manufacturing-scale pCO2 in the ambr250, reduce the overlay gas flow rate or blend CO2 into the overlay gas at 2-5%. Some teams cap the sparge rate at 0.01 vvm (versus the default 0.02-0.05 vvm) during production phase to reduce stripping. Monitor pCO2 offline at each sampling point to confirm the adjustment brings the profile within 20 mmHg of the target.
Hydrodynamic Stress and Shear
Matching P/V (power input per unit volume) between the ambr250 and manufacturing scale does not guarantee matched shear stress. The ambr250 impeller diameter is roughly 40% of the vessel diameter (D/T ratio of about 0.4), compared to 0.33-0.40 in a typical 2,000 L STR. At constant P/V, the smaller impeller spins faster, generating higher tip speed and higher maximum local energy dissipation rates near the blade tips.
Srom et al. (2022) used shear-sensitive PMMA micro-aggregates to characterize the maximum hydrodynamic stress in the ambr250 and found that above a critical stress threshold, CHO-K1 GS cells showed reduced growth rate and viability. Below that threshold, operating conditions had no significant impact on cell behaviour. This work established that the maximum stress, not the average P/V, controls the cell response across bioreactor scales.
Kenmoku et al. (2025) confirmed this in a broader study using a dry-wet approach: they measured that average shear stress correlated well with titer decrease in CHO-K1 fed-batch cultures, and that shear sensitivity varies significantly by cell line. A clone that tolerates 0.2 Pa average shear may fail at 0.3 Pa, while another clone is unaffected up to 0.5 Pa.
| Parameter | Ambr250 | 5 L bench | 2,000 L SUB |
|---|---|---|---|
| Impeller diameter (mm) | ~25 | ~50 | ~400 |
| Agitation speed (rpm) | 300-600 | 150-300 | 40-80 |
| Tip speed (m/s) | 0.4-0.8 | 0.4-0.8 | 0.8-1.7 |
| Mixing time (s) | 3-8 | 8-15 | 20-60 |
| Max energy dissipation rate (W/kg) | 0.5-2.0 | 0.2-1.0 | 0.1-0.5 |
Mitigation: Rather than matching P/V, match the maximum energy dissipation rate (epsilon_max) or tip speed. This typically means running the ambr250 at a lower agitation speed than a pure P/V match would dictate. For shear-sensitive cell lines, reduce agitation 10-20% below the P/V-matched setpoint and confirm oxygen transfer remains adequate by monitoring DO.
Evaporation and Osmolality Drift
Evaporation is the most underestimated root cause of ambr250 viability loss. Over a 15-day fed-batch without volume compensation, a 200 mL vessel can lose 30-40% of its liquid volume to evaporation driven by the continuous gas flow through the headspace. This concentrates every solute in the medium: salts, metabolites, and feed components. The result is an osmolality drift that can push the culture above 400 mOsm/kg, a level that is directly cytotoxic to most CHO cell lines.
Wiegmann et al. (2018) demonstrated the magnitude of this problem in miniature bioreactors: uncompensated 15-day fed-batch runs showed 36.7 ± 6.7% liquid loss, rising osmolality, and declining viability. By tracking sodium ion concentration as an indirect evaporation marker and adding sterile water when Na+ exceeded the day-0 baseline, they reduced liquid loss to 6.9 ± 6.5%. The sodium-based method had an R2 of 0.95 correlation with gravimetric evaporation measurements.
Worked Example: Estimating Evaporation Impact
Setup: 200 mL ambr250 fed-batch, 37 °C, overlay gas at 0.5 L/min, 14-day run
Measured evaporation rate: ~2.5 mL/day (1.25% of initial volume per day)
Day 14 volume without compensation: 200 - (2.5 × 14) = 165 mL = 17.5% loss
But: feeds add ~30 mL total, and sampling removes ~14 mL (1 mL × 14 samples). Net volume = 200 - 35 + 30 - 14 = 181 mL. Apparent loss is 9.5%, but solute concentration has increased by evaporation/volume_remaining = 35/(181+35) = 16%.
Osmolality impact: Starting at 300 mOsm/kg, a 16% concentration gives 300 × 1.16 = 348 mOsm/kg from evaporation alone, before accounting for lactate/ammonia accumulation and base addition.
With Na+-guided water additions (daily): Osmolality drift from evaporation reduced to <3%, keeping contribution below +10 mOsm/kg.
Mitigation: Implement one of two strategies: (1) humidify the overlay gas by passing it through a water reservoir at 37 °C, reducing evaporation at source by 60-80%, or (2) add sterile water periodically based on Na+ or weight measurements. Strategy 1 is simpler but requires hardware modification. Strategy 2 is more precise and works on any system.
Measurement and Sampling Artifacts
Each 1 mL sample from a 200 mL ambr250 vessel removes 0.5% of the culture volume. Over a 14-day run with daily sampling, that is 7% of the initial volume. By contrast, a 5 mL sample from a 2,000 L bioreactor removes 0.00025% of the volume. The ambr250 sampling impact is 2,000 times larger, proportionally.
This matters for three reasons:
- Volume depletion compounds with evaporation. The combined loss (sampling + evaporation) reaches 25-40% of initial volume by day 14 without compensation, concentrating all solutes.
- Feed ratio distortion. If feeds are dosed as a fixed percentage of the initial volume (e.g., 3% v/v per day of a 200 mL start), but the actual volume has dropped to 150 mL by day 10, the effective feed concentration is 33% higher than intended. This overfeeding concentrates amino acids and glucose, shifting metabolite profiles.
- Selective removal. Each sample removes cells, product, and medium together. In a high-density culture (20 × 106 cells/mL), a 1 mL sample removes 20 million cells. Over 14 samples, that is 280 million cells removed, representing a meaningful fraction of the total cell mass in a 200 mL vessel.
Mitigation: Minimize sample volume (0.3-0.5 mL is sufficient for most offline analytics). Use in-line sensors where available (pH, DO, biomass by capacitance) to eliminate some sampling entirely. Adjust feed volumes based on actual measured volume, not initial volume. Some teams use a volume-tracking algorithm that accounts for feeds added and samples removed at each time point.
Feed Volumes and Media Concentration Effects
Feed handling introduces two sources of variability that are negligible at manufacturing scale but significant in the ambr250. First, automated liquid-handling systems (pipettes) have a precision of ±1-3% for volumes below 1 mL. When a daily feed addition is 4 mL (2% of 200 mL working volume), a ±3% pipetting error means ±0.12 mL. Over 10 feed additions, the cumulative deviation can reach ±1.2 mL, which is 0.6% of the total volume. At manufacturing scale, peristaltic pump accuracy of ±1% on a 40 L feed addition introduces the same relative error but with far less impact on concentration.
Second, feed concentrates stored at room temperature in the ambr250 workstation degrade faster than feeds held in a refrigerated bag on a manufacturing-floor balance. Glutamine cyclisation (half-life ~14 days at 37 °C, ~7 days at 25 °C in some formulations), cysteine oxidation, and light-sensitive vitamin degradation (riboflavin, folic acid) can reduce effective nutrient delivery over a 14-day campaign. Bordoloi and Talebnia Rowshan (2025) found that CHO viability dropped to 72% in the ambr250 vessel receiving the lowest feed rate, compared to 88% in the highest-feed vessel, confirming that nutrient limitation at small scale is a direct driver of viability loss.
Mitigation: Prepare feeds in single-use aliquots and store at 2-8 °C. Bring each aliquot to room temperature immediately before use. For glutamine-containing feeds, prepare fresh every 3-4 days. Calibrate pipettes weekly and use gravimetric verification (weigh the vessel before and after feed addition) for critical experiments.
How to Screen Root Causes with DOE
A screening DOE run entirely in the ambr250 is the fastest way to identify which root cause dominates for your specific cell line and process. Because the ambr250 runs up to 24 vessels in parallel, you can execute a 6-8 run definitive screening design (DSD) or fractional factorial in a single 14-day campaign, with replicates.
The recommended screening factors and their ranges are:
| Factor | Low level | High level | Root cause tested |
|---|---|---|---|
| Overlay gas flow (L/min) | 0.1 | 0.5 | Evaporation, CO2 stripping |
| Sparge rate (vvm) | 0.005 | 0.05 | Gas transfer, bubble shear |
| CO2 in overlay (%) | 0 | 5 | pCO2 matching |
| Agitation speed (rpm) | 200 | 500 | Shear stress |
| Water addition frequency | None | Daily (Na+-guided) | Evaporation compensation |
| Sample volume (mL) | 0.3 | 1.0 | Volume depletion |
The response variables are day-14 viability (%), peak viable cell density, final titer, and osmolality trajectory. A Pareto chart of standardised effects from this design will rank the factors by impact, immediately showing whether your gap is driven primarily by evaporation (overlay gas flow and water addition will dominate), gas transfer (sparge rate and CO2 overlay), or shear (agitation speed).
Build Your Screening Design
Use the DOE Generator to create a definitive screening design for ambr250 root-cause diagnosis. The Pro tier adds D-optimal designs for constrained parameter spaces.
Engineering Mitigations That Close the Gap
Once you know which root cause dominates, apply targeted mitigations. The chart below ranks each root cause by how frequently it drives the gap (based on published case studies and industry reports) and the typical viability improvement when mitigated.
| Mitigation | Root cause addressed | Typical viability gain (pp) | Complexity |
|---|---|---|---|
| Na+-guided daily water additions | Evaporation / osmolality | 3-8 | Low |
| CO2 overlay (2-5%) | Gas transfer / pCO2 | 2-5 | Low |
| Humidified overlay gas | Evaporation | 2-6 | Medium |
| Reduced sparge rate (0.005-0.01 vvm) | CO2 stripping, bubble shear | 1-3 | Low |
| Reduced agitation (10-20% below P/V match) | Shear stress | 1-4 | Low |
| Minimised sampling (0.3 mL, fewer time points) | Volume depletion | 1-2 | Low |
| Volume-adjusted feed dosing | Feed ratio distortion | 1-3 | Medium |
| Refrigerated feed aliquots | Feed degradation | 0.5-2 | Low |
The combined effect of the top three mitigations (evaporation compensation, CO2 overlay, reduced sparge rate) typically recovers 6-12 percentage points of the viability gap, bringing the ambr250 within 3-5 percentage points of manufacturing scale. For most process characterisation studies, this is sufficient for qualification.
Scale-Up Calculator
Match engineering parameters (P/V, tip speed, kLa) between the ambr250 and your target manufacturing scale.
What Viability Gap Is Acceptable for Qualification?
There is no fixed regulatory threshold for the viability difference between a scale-down model and manufacturing scale. ICH Q11 and the FDA Process Validation Guidance require the scale-down model to be "representative," but leave the acceptance criteria to the applicant. In practice, industry consensus has converged on a tiered framework based on the magnitude of the gap at harvest:
- <5 percentage points: Generally acceptable without additional justification. The model is considered representative for process characterisation.
- 5-10 percentage points: Requires documented root-cause analysis and a demonstration that the gap does not affect the critical quality attributes (CQAs) under study. Regulators may ask for side-by-side data.
- >10 percentage points: The model needs engineering adjustments before it can support BLA-enabling process characterisation. The gap is too large to argue representativeness.
The key question is not the viability number itself, but whether the viability gap causes a product quality difference. If the ambr250 runs at 75% viability at harvest while manufacturing runs at 88%, but the glycosylation profile, charge variant distribution, and aggregate levels are statistically equivalent, the model may still qualify. Conversely, a 5-point viability gap that shifts galactosylation by 8% absolute is a disqualifying finding.
For formal qualification methodology, statistical equivalency testing, and the regulatory expectations for BLA-enabling studies, see the companion guide: How to Qualify a Scale-Down Model for Bioprocess Development.
Frequently Asked Questions
Why does viability drop faster in the ambr250 than at manufacturing scale?
The ambr250's small working volume (100-250 mL) creates a higher surface-to-volume ratio, more efficient CO2 stripping, greater relative evaporation, and disproportionate sample removal compared to manufacturing-scale vessels. These engineering differences compound over a 14-day fed-batch, typically producing 5-15% lower viability by day 10-12. The root cause is physics, not biology.
What is an acceptable viability gap between ambr250 and manufacturing scale?
Industry practice considers less than 5 percentage points acceptable for a qualified scale-down model. Gaps of 5-10% require documented root-cause analysis and CQA equivalency data. Gaps exceeding 10% generally indicate the model needs engineering adjustments before it can support process characterisation studies.
How do I compensate for evaporation in the ambr250?
Track sodium ion concentration as an indirect evaporation marker. When Na+ rises above its day-0 baseline, add sterile water to restore the original concentration. Wiegmann et al. (2018) demonstrated that this method reduced 15-day fed-batch liquid loss from 37% to 7% in miniature bioreactors. Alternatively, use humidified overlay gas (passed through a water reservoir at 37 °C) to reduce evaporation at source.
Can a DOE identify which root cause is driving viability loss in my ambr250?
Yes. A fractional factorial or definitive screening design with 6-8 runs in the ambr250 itself can rank the contribution of overlay gas flow rate, sparge rate, agitation speed, water-addition frequency, and sample volume to the viability gap. A single 24-vessel campaign produces enough data to identify the dominant factors.
Does the ambr250 Generation 2 fix the viability problem?
The Generation 2 system introduces continuous gassing control, improved temperature regulation, and better data integrity features, which address some root causes. However, evaporation, sampling artifacts, and hydrodynamic differences remain inherent to the 250 mL working volume. The Gen 2 narrows the gap but does not eliminate it.
CHO Troubleshooter
Diagnose other CHO culture problems: low viability, high lactate, slow growth, and more.
Related Tools
- Scale-Up Calculator — Match P/V, tip speed, kLa, and mixing time between the ambr250 and your target manufacturing scale.
- DOE Generator — Build screening and optimisation designs to identify the dominant root causes of your viability gap.
- OTR/kLa Estimator — Calculate volumetric oxygen transfer coefficients for scale-down parameter matching.
References
- Srom O, Travnikova V, Wutz J, Kuschel M, Unsoeld A, Wucherpfennig T, Soos M. Characterization of hydrodynamic stress in ambr250 bioreactor system and its impact on mammalian cell culture. Biochem Eng J. 2022;178:108240. doi:10.1016/j.bej.2021.108240
- Wiegmann V, Bernal Martinez C, Baganz F. A simple method to determine evaporation and compensate for liquid losses in small-scale cell culture systems. Biotechnol Lett. 2018;40(7):1029-1036. doi:10.1007/s10529-018-2556-x
- Kenmoku H, Kaneko A, Saito T, Nemoto T, Kato Y, Ohira S. Scale-up strategy focused on hydrodynamic stress for mammalian cell culture established by a dry-wet approach. Eng Life Sci. 2025;25(11):e70054. doi:10.1002/elsc.70054
- Bordoloi A, Talebnia Rowshan F. Process mapping and optimization study of CHO cell cultures for mAb production using Ambr 250 high-throughput parallel bioreactors. Bioprocess Biosyst Eng. 2025;48(12):2077-2088. doi:10.1007/s00449-025-03229-y
- Xu S, Hoshan L, Jiang R, Gupta B, Brodean E, O'Neill K, Seamans TC, Bowers J, Chen H. A practical approach in bioreactor scale-up and process transfer using a combination of constant P/V and vvm as the criterion. Biotechnol Prog. 2017;33(4):1146-1159. doi:10.1002/btpr.2489