Scale-Up, Plant Design, Metabolic Engineering and Synthetic Biology
Scale-up criteria and power correlations, bioprocess plant design, flux balance analysis, metabolic control analysis and synthetic biology.
- 20 min
On this page · 13 sections
Prerequisites
- Microbial Growth Kinetics, Bioreactors and Fermentation — \( k_La \), reactor types and fed-batch strategy.
- Engineering Principles and Bioprocess Thermodynamics — balances, the stoichiometric matrix and dimensionless groups.
- Regulation of Gene Expression and Epigenetics — promoters, repressors and inducible systems, the raw material of synthetic circuits.
The Hook: The Process That Worked Perfectly at 5 Litres
How CSIR NET asks it. Part B: which scale-up criterion is standard, what the power number means, what a flux control coefficient is, what the summation theorem states, which tool does what in synthetic biology. Part C: compute impeller speed and power on scale-up under a stated criterion, interpret flux distributions or knockout data, reason through a control-analysis or FBA argument with multiple correct options.
Scale-Up
Mixing and power
For a stirred tank the power drawn by the impeller is
$$P = N_p\,\rho\,N^3 D_i^5$$where \( N_p \) is the power number (a constant for a given impeller geometry in the turbulent regime — of the order of 5 for a standard Rushton turbine with baffles), \( N \) is in revolutions per second and \( D_i \) is the impeller diameter. In the laminar regime \( N_p \propto 1/Re_i \) and the power scales differently; fermenters are almost always turbulent.
Aeration reduces the power draw (gassed power \( P_g < P \)) because gas cavities behind the blades lower the effective density — a correction that matters when sizing a motor.
Mixing time \( t_m \) is the time to reach a stated uniformity after a tracer pulse. In a geometrically similar series it rises with scale: a small vessel mixes in a second or two, a large one in tens of seconds. Because substrate uptake and oxygen consumption happen on a timescale of seconds, large vessels inevitably have gradients in substrate, dissolved oxygen, pH and CO₂.
The choice: you cannot hold everything constant
| Criterion held constant | Consequence for the large vessel | When used |
|---|---|---|
| \( P/V \) | Tip speed rises, mixing time rises, \( Re_i \) rises | The commonest default; keeps \( k_La \) roughly comparable |
| \( k_La \) | Power may have to rise; sometimes achieved by aeration or enrichment instead | The right criterion for oxygen-limited aerobic processes |
| Impeller tip speed \( \pi N D_i \) | \( P/V \) falls sharply, so \( k_La \) falls | Shear-sensitive cultures (mycelia, animal cells) |
| Mixing time | Power requirement becomes absurd at large scale | Essentially impossible to hold |
| \( Re_i \) | \( P/V \) collapses | Rarely useful for fermenters |
The algebra of constant \( P/V \). With \( P/V \propto N^3D_i^5/D_i^3 = N^3D_i^2 \), holding \( P/V \) constant requires
$$N_2 = N_1\left(\frac{D_{i1}}{D_{i2}}\right)^{2/3}$$so impeller speed falls on scale-up, but only as the two-thirds power, while diameter rises linearly — which is why tip speed rises and shear at the blade tip increases. That is the hidden cost of the default criterion.
Scale-down: the modern answer
Because no single criterion reproduces the large vessel, process development now uses scale-down models: a small reactor deliberately operated to mimic large-scale heterogeneity — for example a two-compartment system in which cells are circulated between a well-mixed stirred tank and a plug-flow loop that is substrate-rich or oxygen-poor. Strains and feeding strategies are then screened for robustness to the oscillations they will actually meet. Computational fluid dynamics coupled to kinetic models does the same job in silico.
Bioprocess Plant Design
A fermentation plant is a sequence of operations wrapped in utilities, containment and documentation.
graph TD
M["Media prep and sterilization
batch or continuous HTST"] --> SF["Seed train
flask to seed fermenters,
5-10% inoculum per stage"]
SF --> PF["Production fermenter
control loops, feeds, off-gas"]
AIR["Air compression,
0.2 um filtration"] --> PF
PF --> H["Harvest"]
H --> DSP["Downstream train"]
DSP --> FF["Formulation, fill, finish"]
U["Utilities: clean steam, WFI,
chilled water, process air, CIP/SIP"] --> PF
U --> DSP
QA["QA/QC: validation, batch records,
environmental monitoring"] --> FF| Design area | Key decisions |
|---|---|
| Seed train | Usually 5–10% inoculum per stage, so each stage is roughly a ten- to twentyfold step; number of stages set by the final volume and the organism’s lag |
| Materials | 316L stainless steel, electropolished, with sanitary (tri-clamp) fittings and no dead legs where liquid can stagnate |
| Sterility | SIP of the vessel and lines; 0.2 µm filtration of inlet and outlet air; positive pressure; single-use assemblies at clinical scale |
| Utilities | Clean steam, water for injection, chilled water or glycol for the jacket, oil-free compressed air, nitrogen |
| Containment | Biosafety level for the organism; inactivation of effluent (kill tank) for GMOs; HEPA-filtered areas and airlocks |
| Instrumentation and automation | DCS/SCADA with cascade control of DO (through agitation, flow and enrichment), pH, temperature, feed; electronic batch records |
| Process analytical technology (PAT) | Online probes and spectroscopy to measure quality while the batch runs rather than testing it afterwards |
| Quality | GMP, validation (installation, operational and performance qualification), cleaning validation, change control, traceability |
| Single-use vs stainless | Single-use: fast changeover, no CIP/SIP, lower capital, limited scale, plastic waste and extractables. Stainless: large scale, lower running cost, heavy validation burden |
Regulatory and biosafety context (India). Research and manufacture with genetically engineered organisms is governed by the Rules notified under the Environment (Protection) Act, 1986, with institutional biosafety committees and national committees above them; the Department of Biotechnology and the Central Drugs Standard Control Organisation are the relevant central bodies for research oversight and for biologics approval respectively. Learn the structure — a named committee acronym occasionally appears as a Part B option — rather than the procedural detail.
Economic levers at plant scale, in rough order of impact: titre, batch time (turnaround included), downstream step count and yield, raw-material cost, and labour and QC. A small rise in titre pays twice, because the downstream train shrinks with it.
Metabolic Engineering
Metabolic engineering is the directed improvement of cellular properties by modifying specific biochemical reactions, with analysis of the resulting flux distribution. Its defining feature is the cycle: analyse, design, build, measure the fluxes again.
graph LR
A["Analyse
flux maps, bottlenecks,
genome-scale model"] --> D["Design
which gene to delete,
overexpress or import"]
D --> B["Build
CRISPR, recombineering,
synthetic operons"]
B --> T["Test
titre, rate, yield,
13C flux, omics"]
T --> AFlux analysis
| Method | What it needs | What it gives |
|---|---|---|
| Metabolic flux analysis (MFA) | A stoichiometric model plus measured uptake and secretion rates | The internal flux distribution, if the system is determined |
| ¹³C-MFA | Labelled substrate plus mass-spectrometric labelling patterns of metabolites or amino acids | Resolves parallel and cyclic pathways that stoichiometry alone cannot (for example the split between glycolysis and the pentose phosphate pathway) |
| Flux balance analysis (FBA) | A genome-scale stoichiometric model, \( \mathbf{S}\mathbf{v}=0 \), flux bounds, and an objective (commonly maximum growth) | An optimal flux distribution by linear programming; no kinetic parameters needed |
| FVA, MOMA, knockout algorithms | The same model | Ranges of feasible fluxes; predicted behaviour of mutants; candidate deletions that couple product to growth |
FBA’s assumptions are its limits: steady state, an optimality objective the cell need not share, and no information about metabolite concentrations or regulation. It predicts what is stoichiometrically possible, not what the cell will do.
Metabolic control analysis
The key insight that replaced the idea of a single “rate-limiting step”. Define the flux control coefficient
$$C^J_i = \frac{\partial \ln J}{\partial \ln E_i}$$the fractional change in pathway flux per fractional change in the activity of enzyme \( i \).
- Summation theorem: \( \sum_i C^J_i = 1 \). Control is shared; if one enzyme had \( C^J = 1 \), every other would be zero, which is almost never observed.
- Connectivity theorem links control coefficients to elasticities \( \varepsilon^{v_i}_S \) (the local sensitivity of one enzyme’s rate to a metabolite).
- Practical consequence: overexpressing the enzyme you assume is limiting typically gives a disappointing flux increase, because its control coefficient is small. Real gains need several steps changed together, or the removal of an allosteric or transcriptional constraint.
The engineer’s toolkit
| Strategy | Example of the idea |
|---|---|
| Overexpress the committed step, or replace it with a feedback-insensitive variant | Classic amino-acid production relies on deregulated allosteric enzymes |
| Delete competing pathways | Remove by-product routes to push carbon towards the target |
| Balance cofactors | Match NADPH supply to demand by rerouting through the pentose phosphate pathway or swapping cofactor specificity |
| Import heterologous pathways | Move a plant or fungal pathway into a microbial host |
| Modulate expression levels | Promoter and RBS libraries to tune each step rather than maximise it |
| Transport engineering | Improve uptake of substrate or export of product to relieve inhibition |
| Compartmentalise | Target enzymes to mitochondria or peroxisomes, or build scaffolds, to raise local concentrations |
| Dynamic control | Sense an intermediate and switch flux on or off during the run, so growth and production can be separated |
Flagship achievements to recognise: the microbial route to the artemisinin precursor artemisinic acid in engineered yeast; 1,3-propanediol from sugar by an engineered organism for polymer manufacture; microbial production of amino acids (glutamate, lysine) at enormous scale; and engineered strains for biofuels, bioplastic monomers and natural-product scaffolds.
Synthetic Biology
Where metabolic engineering modifies a pathway, synthetic biology aims to design biological systems from characterised, reusable parts, with abstraction and standardisation borrowed from engineering.
| Layer | Content |
|---|---|
| Parts | Promoters, ribosome binding sites, coding sequences, terminators, insulators; registries of standardised parts; assembly standards and methods (Gibson assembly, Golden Gate, yeast assembly) |
| Devices | Sensors, logic gates, oscillators, switches |
| Systems | Whole pathways and circuits in a chassis |
| Chassis | E. coli, B. subtilis, S. cerevisiae, Pichia, cyanobacteria, mammalian cells, cell-free extracts |
Founding circuits. The genetic toggle switch (two mutually repressing repressors, giving bistability and memory) and the repressilator (three repressors in a cycle, giving oscillation) were built in E. coli in 2000 and established that circuit behaviour can be designed and modelled. Later came quorum-sensing-based population circuits, band-detect “edge detectors”, and logic gates built from orthogonal regulators.
Enabling tools.
| Tool | Use |
|---|---|
| DNA synthesis and assembly | Build whole operons, genomes and codon-optimised genes |
| CRISPR–Cas editing | Precise, multiplexed edits in almost any chassis |
| CRISPRi / CRISPRa (dCas9) | Tunable repression or activation without changing the genome — ideal for screening which step limits flux |
| Riboswitches and biosensors | Metabolite-responsive regulation; coupling a product to fluorescence for high-throughput screening |
| Orthogonal systems | Expanded genetic codes, unnatural amino acids, orthogonal ribosomes |
| Genome reduction and minimal cells | Strip non-essential genes to make a predictable chassis |
| Cell-free systems | Transcription–translation in a tube: fast prototyping, no toxicity or growth burden |
Where the field is honest about its limits: context dependence (a part behaves differently in a new circuit), metabolic burden (expression competes with growth, so engineered cells are outgrown by escape mutants), genetic instability, and the gap between a model and a living chassis. Biosafety and biocontainment — kill switches, auxotrophy, synthetic amino-acid dependence — are an explicit design requirement, not an afterthought.
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Power number \( N_p \) (geometry constant) | vs | Power per volume \( P/V \) (operating choice) |
| Constant \( P/V \) (tip speed rises) | vs | Constant tip speed (\( P/V \) falls) |
| Mixing time (homogenisation) | vs | Residence time (\( 1/D \), throughput) |
| MFA (measured rates + stoichiometry) | vs | FBA (optimisation of a genome-scale model) |
| Flux control coefficient (systemic, \( \sum = 1 \)) | vs | Elasticity (local, one enzyme’s response to a metabolite) |
| Metabolic engineering (modify an existing pathway) | vs | Synthetic biology (design from standardised parts) |
| CRISPR–Cas9 editing (permanent change) | vs | CRISPRi with dCas9 (reversible, tunable repression) |
| Toggle switch (bistable) | vs | Repressilator (oscillator) |
| Validation (documented evidence the process performs) | vs | Calibration (an instrument reads correctly) |
Common Mistakes
Practice
All questions below are original practice questions written in CSIR NET style. They are not previous-year questions.
Level 1: Foundation
- Write the turbulent power correlation for a stirred tank.
- State the summation theorem of metabolic control analysis.
- What does dCas9 provide that Cas9 does not?
Answers
- \( P = N_p\rho N^3D_i^5 \).
- \( \sum_i C^J_i = 1 \): the flux control coefficients of all enzymes in a pathway sum to one.
- Catalytically dead Cas9 binds DNA without cutting, giving reversible, tunable repression (CRISPRi) or, with an activation domain, activation (CRISPRa) — no permanent genomic change.
Level 2: Part B Style
Q1 (practice). A process is scaled from 1 m³ to 10 m³ with geometric similarity, holding \( P/V \) constant. The impeller speed in the large vessel, relative to the small one, is a factor of:
- (a) 2.15
- (b) 1.00
- (c) 0.60
- (d) 0.46
Answer and explanation
(c). Linear scale factor \( = 10^{1/3} = 2.154 \). Constant \( P/V \) requires \( N_2 = N_1(D_{i1}/D_{i2})^{2/3} = (1/2.154)^{2/3} = \mathbf{0.60} \). Speed falls, but only modestly — which is why tip speed rises.
Q2 (practice). Which scale-up criterion is most appropriate for a shear-sensitive mycelial culture?
- (a) Constant impeller Reynolds number
- (b) Constant tip speed
- (c) Constant rpm
- (d) Constant mixing time
Answer and explanation
(b). Tip speed \( \pi ND_i \) is the standard proxy for the maximum shear a cell meets at the blade. Holding it constant protects the organism, at the price of a lower \( P/V \) and therefore lower \( k_La \) — usually compensated with more aeration or oxygen enrichment.
Q3 (practice). An enzyme in a pathway has a flux control coefficient of 0.05. Doubling its activity is expected to raise the pathway flux by approximately:
- (a) 100%
- (b) 50%
- (c) 5%
- (d) 0%
Answer and explanation
(c). \( C^J = \Delta\ln J/\Delta\ln E \approx (\Delta J/J)/(\Delta E/E) \). A 100% increase in activity with \( C^J = 0.05 \) gives roughly a 5% flux increase. This single number explains most disappointing overexpression experiments.
Q4 (practice). Flux balance analysis requires all of the following except:
- (a) a stoichiometric matrix
- (b) upper and lower bounds on fluxes
- (c) an objective function
- (d) measured \( K_m \) values for every enzyme
Answer and explanation
(d). FBA is deliberately kinetics-free: \( \mathbf{S}\mathbf{v}=0 \) plus constraints plus an objective, solved by linear programming. Needing no kinetic parameters is exactly what lets it work at genome scale — and is also why it cannot predict concentrations or dynamics.
Level 3: Part C Style
Q1 (practice). A 1 m³ fermenter uses a Rushton turbine with \( N_p \) = 5.0, \( D_i \) = 0.50 m, at \( N \) = 2.0 s⁻¹, in broth of density 1,000 kg m⁻³. The process is to be scaled to 10 m³ with geometric similarity.
(i) Calculate the ungassed power and \( P/V \) at small scale. (ii) Under constant \( P/V \), calculate the large-scale impeller diameter, speed, power and tip speed, and compare tip speeds. (iii) If constant tip speed were chosen instead, what would \( P/V \) become?
Answer and worked solution
(i) \( P = N_p\rho N^3D_i^5 = 5.0 \times 1{,}000 \times (2.0)^3 \times (0.50)^5 = 5{,}000 \times 8 \times 0.03125 = \mathbf{1{,}250\ W} \), so \( P/V = \mathbf{1.25\ kW\ m^{-3}} \) — a typical aerobic value.
(ii) Linear factor \( 10^{1/3} = 2.154 \), so \( D_{i2} = 0.50 \times 2.154 = \mathbf{1.077\ m} \).
$$N_2 = N_1(D_{i1}/D_{i2})^{2/3} = 2.0 \times (1/2.154)^{2/3} = 2.0 \times 0.5995 = \mathbf{1.20\ s^{-1}}$$Check the power: \( P_2 = 5.0\times1{,}000\times(1.199)^3\times(1.077)^5 = \mathbf{12{,}500\ W} \), i.e. \( P_2/V_2 = 12.5\ \mathrm{kW}/10\ \mathrm{m^3} = 1.25\ \mathrm{kW\ m^{-3}} \). Confirmed. Tip speeds: small \( = \pi(2.0)(0.50) = 3.14\ \mathrm{m\ s^{-1}} \); large \( = \pi(1.199)(1.077) = 4.06\ \mathrm{m\ s^{-1}} \) — 29% higher, which is \( 2.154^{1/3} \). Holding \( P/V \) constant therefore increases the shear the cells experience at the blade tip.
(iii) Constant tip speed means \( N_3 = N_1D_{i1}/D_{i2} = 2.0/2.154 = 0.928\ \mathrm{s^{-1}} \). Then
$$P_3 = 5.0\times1{,}000\times(0.928)^3\times(1.077)^5 = 5{,}802\ \mathrm{W} \Rightarrow P_3/V = \mathbf{0.58\ kW\ m^{-3}}$$less than half the small-scale value, with a correspondingly lower \( k_La \). This is the central trade-off of scale-up in one calculation: protect the cells from shear and you starve them of oxygen; feed them oxygen and you shear them.
Q2 (practice). An engineered strain converts glucose to a three-carbon product. Three constructs are compared in identical fed-batch runs.
| Strain | Product titre (g L⁻¹) | Carbon yield (C-mol product / C-mol glucose) | By-product acetate (g L⁻¹) | Final biomass (g L⁻¹) |
|---|---|---|---|---|
| Parent + pathway | 12 | 0.28 | 6.0 | 14 |
| + deletion of the acetate route | 18 | 0.44 | 0.8 | 11 |
| + deletion and overexpression of the first pathway enzyme | 19 | 0.46 | 0.9 | 10 |
(i) Interpret the effect of each modification. (ii) The theoretical maximum carbon yield for a three-carbon product from glucose with no carbon loss is 1.00 — why is the measured value so much lower, and what two routes would you try next? (iii) What does the third strain’s result suggest about the flux control coefficient of the first pathway enzyme?
Answer and worked solution
(i) Deleting the acetate route removes an overflow sink: acetate falls from 6.0 to 0.8 g L⁻¹, titre rises 50% and carbon yield rises from 0.28 to 0.44. Biomass falls from 14 to 11 g L⁻¹, as expected when carbon is diverted away from growth (and possibly because acetate overflow was relieving a redox or energy imbalance). The extra overexpression adds only 1 g L⁻¹ of titre and 0.02 of yield — within the noise of a fermentation comparison.
(ii) Carbon is lost to biomass, to CO₂ in the decarboxylations of central metabolism, and to maintenance energy; the stoichiometric maximum assumes none of these. Two next steps: (a) decouple growth from production — a two-phase fed-batch or a dynamically controlled promoter so that carbon goes to biomass first and to product later; (b) reduce carbon loss in the pathway itself — a route with fewer decarboxylation steps, or cofactor rebalancing so that the required NAD(P)H is generated without extra oxidation of substrate to CO₂. Deleting further by-product routes and improving product export are reasonable additional answers.
(iii) The small gain suggests the first pathway enzyme has a low flux control coefficient under these conditions — control has moved elsewhere (a later pathway step, cofactor supply, or precursor availability). Consistent with the summation theorem, the next experiment should test several steps, for example by modulating each with CRISPRi or by a promoter library, rather than pushing harder on the same enzyme.
Q3 (practice). Which of the following statements about synthetic-biology circuits and their hosts are correct?
(A) A repressilator uses three repressors in a cyclic arrangement and produces oscillating gene expression. (B) A genetic toggle switch built from two mutually repressing repressors is bistable and can store a state. (C) dCas9 fused to a repression domain allows reversible, tunable knock-down without changing the genome sequence. (D) Expressing a heterologous pathway imposes no measurable burden on the host, so engineered strains are genetically stable over long fermentations.
- (a) A, B and C
- (b) A and B only
- (c) B, C and D
- (d) A, B, C and D
Answer and worked solution
(a) A, B and C.
- (A) correct — an odd-numbered ring of repressors is a negative-feedback loop with delay and, with suitable parameters, oscillates.
- (B) correct — mutual repression between two repressors gives two stable states, hence memory; the state is set by a transient inducer pulse.
- (C) correct — catalytically dead Cas9 blocks transcription (or recruits effectors) at a guide-specified locus, and the effect is titratable and reversible, which is why it is the standard tool for finding which step controls a flux.
- (D) wrong on both counts: heterologous expression diverts ribosomes, amino acids and energy, measurably slowing growth (metabolic burden), and precisely because burdened cells grow slower, non-producing escape mutants are selected during long fermentations. Plasmid loss, transposon insertion and promoter mutations are the usual routes, which is why production strains use genome integration, selection that is coupled to production, or tightly induced expression late in the run.
Quick Revision Table
| Item | Fact |
|---|---|
| Power | \( P = N_p\rho N^3D_i^5 \); \( N_p \approx 5 \) for a baffled Rushton turbine (turbulent) |
| Gassed power | \( P_g < P \): aeration lowers the draw |
| \( P/V \) | \( \propto N^3D_i^2 \); typical aerobic value 1–5 kW m⁻³ |
| Constant \( P/V \) | \( N_2 = N_1(D_{i1}/D_{i2})^{2/3} \); tip speed rises as \( D_i^{1/3} \) |
| Constant tip speed | \( N \propto 1/D_i \); \( P/V \) falls sharply; for shear-sensitive cultures |
| Geometric similarity | \( D_i/D_T \approx 1/3 \); tenfold volume = 2.154× every length |
| Mixing time | Rises with scale; large vessels have substrate, DO, pH and CO₂ gradients |
| Scale-down | Two-compartment reactors and CFD to mimic heterogeneity |
| Seed train | 5–10% inoculum per stage |
| Plant | 316L steel, sanitary fittings, SIP/CIP, 0.2 µm air filters, WFI, DCS, PAT, GMP validation |
| Single-use | Fast changeover, no CIP/SIP, scale-limited, extractables |
| MFA vs FBA | Measured rates + stoichiometry vs genome-scale optimisation (\( \mathbf{Sv}=0 \), objective, LP) |
| ¹³C-MFA | Resolves parallel and cyclic pathways |
| MCA | \( C^J_i = \partial\ln J/\partial\ln E_i \); \( \sum C^J_i = 1 \); elasticities are local |
| Engineering strategies | Deregulate committed step, delete by-product routes, balance cofactors, tune expression, dynamic control, compartmentalisation |
| Synthetic biology | Parts–devices–systems–chassis; toggle switch (bistable), repressilator (oscillator) |
| Tools | DNA synthesis and assembly, CRISPR–Cas, CRISPRi/a (dCas9), riboswitch biosensors, cell-free systems |
| Limits | Context dependence, metabolic burden, genetic instability, biocontainment |
Key Takeaways
What’s Next
- Quick Revision — the whole unit on one page.
- Downstream Processing — the other half of the plant.
- Regulation of Gene Expression and Epigenetics — the natural regulation that synthetic circuits reuse.
- Life Sciences and Biotechnology — the rest of the subject.