Skip to content

Downstream Processing

Cell disruption, centrifugation, precipitation, membranes, chromatography, extraction and drying: CSIR NET downstream processing notes.

  • 19 min read
  • Updated
On this page · 17 sections
  1. Prerequisites
  2. The Hook: 90% Pure, 0% Useful
  3. The Shape of a Purification Train
  4. Cell Disruption
  5. Solid–Liquid Separation
  6. Precipitation
  7. Membrane Separations
  8. Chromatography
  9. Extraction, Crystallisation and Drying
  10. Yield, Purity and the Purification Table
  11. Memory Tricks
  12. Don’t Confuse
  13. Common Mistakes
  14. Practice
  15. Quick Revision Table
  16. Key Takeaways
  17. What’s Next

Prerequisites

The Hook: 90% Pure, 0% Useful

How CSIR NET asks it. Part B: which operation comes where in the train, which method suits an intracellular versus a secreted product, what a given chromatography mode separates on, what lyophilisation does. Part C: compute overall yield and purification factor across a table, work out a diafiltration or a partition-coefficient recovery, read a chromatogram for resolution, choose the right sequence for a stated product.


The Shape of a Purification Train

graph TD
    A["Broth"] --> B["1. Solid-liquid separation
centrifugation, filtration,
flocculation"] B --> C{"Product location?"} C -- "Intracellular" --> D["2. Cell disruption
homogenizer, bead mill,
enzymatic, osmotic"] C -- "Extracellular" --> E["Clarified broth"] D --> F["Debris removal
centrifuge / microfiltration"] F --> G["3. Concentration and primary isolation
precipitation, ultrafiltration,
aqueous two-phase, adsorption"] E --> G G --> H["4. Purification
chromatography:
affinity, ion exchange, HIC, SEC"] H --> I["5. Polishing and formulation
crystallization, final UF/DF,
sterile filtration, drying"] I --> J["Product"]

RIPP is the standard mnemonic for the four stages: Recovery (removal of insolubles), Isolation (concentration and gross purification), Purification (high-resolution), Polishing (final form and stability).

Three rules govern the order:

  1. Reduce volume early. Every later operation costs money per litre.
  2. Remove the biggest contaminant class first; leave the highest-resolution, most expensive step for last, when the feed is small and clean.
  3. Do not repeat a separation principle twice in a row. Two ion-exchange steps back to back buy less than ion exchange followed by hydrophobic interaction.

Cell Disruption

Needed only for intracellular products — one strong argument for secreting the product in the first place.

MethodPrincipleScaleNotes
High-pressure homogeniserBroth forced through a valve; shear and cavitationLargeThe industry standard for bacteria and yeast; multiple passes; heat must be removed
Bead millAgitated glass or ceramic beads grind cellsLargeGood for tough yeast and fungal walls; bead size and loading are the variables
UltrasonicationCavitationLaboratoryHeats badly; poor at scale
French pressPressure dropLaboratory
Enzymatic lysisLysozyme (bacteria), lyticase/zymolyase (yeast)AnyGentle and selective; costly; adds a protein contaminant
ChemicalDetergents, chaotropes, solvents, alkaliAnyMay denature the product
Osmotic shock / freeze–thawPressure difference, ice crystalsSmall–mediumMild; good for periplasmic products

For a homogeniser, release follows a first-order-in-passes form:

$$\ln\frac{R_m}{R_m - R} = k\,N\,P^{a}$$

with \( R \) the protein released, \( R_m \) the maximum releasable, \( N \) the number of passes and \( P \) the operating pressure (\( a \) is an exponent around 2–3 in reported correlations). The practical lesson: release rises with passes, but so does micronisation of debris — over-homogenising creates fine fragments that are very hard to centrifuge or filter out, and it releases more host DNA, which raises viscosity. Periplasmic products are therefore released selectively by osmotic shock rather than by smashing the cell.


Solid–Liquid Separation

Centrifugation. Settling under an applied field follows Stokes’ law:

$$v_g = \frac{d^2(\rho_p-\rho_L)g}{18\mu}$$

Replace \( g \) with \( \omega^2r \) for a centrifuge. Separation gets harder as particles get smaller (the \( d^2 \) term) and as the density difference falls — which is exactly the problem with bacterial cells (about 1 µm, close to the density of broth) and worse with disrupted debris. Flocculants and pH adjustment aggregate fines so that \( d \) effectively rises.

The capacity of a centrifuge is written as \( Q = v_g\Sigma \), where \( \Sigma \) is an equipment-geometry factor with units of area; it allows scale-up between machines of the same type at constant \( v_g \) (that is, same separation performance). Formats: tubular bowl (highest \( g \), batch solids), disc-stack (continuous, intermittent solids discharge, the workhorse), decanter (high solids), and basket.

Filtration. Conventional dead-end filtration builds a cake; flux falls as the cake thickens, and compressible biological cakes get worse with pressure. Filter aids (diatomaceous earth, perlite) and rotary vacuum drum filters with a precoat and a knife are classic for mycelial broths. Crossflow (tangential-flow) filtration sweeps the membrane surface so the cake stays thin; it is the standard for bacterial and mammalian cell harvest.


Precipitation

Cheap, scalable, and the usual first concentration step.

AgentPrincipleComment
Ammonium sulfate (salting-out)High ionic strength removes water of hydration; hydrophobic patches associateThe Cohn equation \( \log S = \beta - K_sI \); proteins come out in order of surface hydrophobicity; mild and often reversible
Organic solvents (ethanol, acetone)Lowers dielectric constant; must be done coldThe basis of Cohn plasma fractionation
Polymers (PEG)Volume exclusionGentle
Isoelectric precipitationAt pH = pI the net charge is zero, so solubility is minimalNeeds a known pI; casein from milk is the classic
Heat / pH denaturationSelective denaturation of contaminantsOnly if the product is robust
Affinity precipitationLigand-linked polymer brings the target outSpecialised

Note the “salting-in” effect at low ionic strength: a little salt increases solubility by screening charges, before salting-out takes over at high strength. Precipitation is a concentration step, not a purification step with high resolution: expect two- to fivefold purification at best.


Membrane Separations

ProcessNominal cut-offRetainsDriving force
Microfiltration (MF)about 0.1–10 µmCells, debrisPressure
Ultrafiltration (UF)about 1–1,000 kDaProteins, virusesPressure
Nanofiltration (NF)about 200–1,000 DaSugars, divalent saltsPressure
Reverse osmosis (RO)IonsEverything but waterPressure above osmotic
DialysisSmall solutesMacromoleculesConcentration gradient
ElectrodialysisIons—Electric field
PervaporationVolatiles—Partial-pressure difference; used for ethanol recovery

Flux is \( J = \mathrm{TMP}/(\mu R_{total}) \), and \( R_{total} \) grows with concentration polarisation (a solute layer at the membrane face) and fouling (adsorbed and deposited material). Above a critical pressure the flux stops rising — the polarised layer simply gets denser. Remedies: crossflow velocity, turbulence promoters, backflushing, and cleaning cycles.

Diafiltration is buffer exchange by UF: add fresh buffer at the rate permeate leaves. For a freely permeating solute at constant volume,

$$\frac{C}{C_0} = e^{-N}$$

where \( N \) is the number of diavolumes exchanged. Five diavolumes remove 99.3% of the small solute; seven remove 99.9%. This is how a protein is moved from a salting-out buffer into a formulation buffer without a column.


Chromatography

The high-resolution core of any biopharmaceutical process.

ModeSeparates onBinding / elutionTypical place in the train
Affinity (Protein A, IMAC, lectin, dye, immunoaffinity)Specific biological recognitionElute by pH, competing ligand or imidazoleCapture; enormous purification in one step
Ion exchange (DEAE/Q anion; CM/S cation)Net surface charge at the working pHSalt or pH gradientCapture or intermediate; very high capacity
Hydrophobic interaction (HIC)Surface hydrophobicityBind at high salt, elute at low saltFits straight after an ammonium sulfate step
Reversed phaseHydrophobicity with organic solventAcetonitrile gradientPeptides, small molecules; denaturing for most proteins
Size exclusion (gel filtration)Hydrodynamic sizeIsocratic; no bindingPolishing, aggregate removal, buffer exchange at small scale
Mixed-mode / hydroxyapatiteCharge plus other interactionsPhosphate gradientPolishing

Rules of thumb. Size exclusion has no concentrating effect and a sample volume limit of a few percent of the bed, so it is always last and never a capture step. Ion exchange selection follows the pI: below its pI a protein is positively charged and binds a cation exchanger; above its pI it binds an anion exchanger. HIC pairs naturally with salting-out because both want high salt.

Performance measures.

$$N = 16\left(\frac{t_R}{w}\right)^2,\qquad \mathrm{HETP} = \frac{L}{N},\qquad R_s = \frac{2(t_{R2}-t_{R1})}{w_1+w_2}$$

\( R_s \ge 1.5 \) is baseline resolution. The van Deemter equation \( H = A + B/u + Cu \) explains the optimum flow velocity: eddy diffusion (A), longitudinal diffusion (B, matters at low flow), and mass-transfer resistance (C, dominant at high flow). Smaller, more uniform beads lower A and C — at the cost of pressure drop, exactly as with immobilised-enzyme beads.

Scale realities. Preparative columns are run on dynamic binding capacity (g product per litre of resin at a given residence time), not on analytical resolution; resin cost and lifetime (number of cycles) dominate economics; and expanded-bed adsorption lets a capture step run on unclarified feed, combining clarification and capture.


Extraction, Crystallisation and Drying

Liquid–liquid extraction suits small molecules (antibiotics, organic acids, steroids): penicillin is extracted into an organic solvent at low pH and back-extracted into aqueous buffer at higher pH, a pH-swing that needs fast, cold operation because penicillin degrades at acidic pH. Performance is set by the partition coefficient \( K = C_{top}/C_{bottom} \) and the phase volume ratio \( V_r = V_{top}/V_{bottom} \); for a single stage, the fraction recovered in the top phase is

$$Y = \frac{K V_r}{1 + K V_r}$$

Several stages in countercurrent recover far more than one stage with the same total solvent.

Aqueous two-phase systems (ATPS) — PEG/dextran or PEG/salt — give two water-rich phases, so proteins and even whole cells partition without denaturation. They are gentle, cheap, continuous and scalable, and they can remove cell debris and product in one operation. Partitioning depends on polymer molecular weight, tie-line length, salt type and pH relative to the pI.

Crystallisation is the purest and cheapest finishing step where it works: insulin, many antibiotics and citric acid are crystallised. It needs supersaturation (cooling, evaporation, antisolvent, pH change), control of nucleation against growth, and attention to polymorph and crystal habit because they decide filtration and dissolution behaviour.

Drying stabilises the product for storage:

MethodPrincipleSuits
Freeze-drying (lyophilisation)Freeze, then sublime ice under vacuum (primary drying), then remove bound water (secondary)Proteins, vaccines, labile biologics; needs cryo- and lyoprotectants (sucrose, trehalose, mannitol) and an eutectic/collapse temperature that the cycle must stay below
Spray dryingAtomise into hot gas; very short residence timeEnzymes, probiotics, bulk powders; cheap and continuous
Vacuum / tray dryingLow temperature evaporationHeat-sensitive solids
Fluidised-bed dryingHot gas suspends particlesGranules

Yield, Purity and the Purification Table

Every purification is reported as a table — and Part C loves to hand you one half-filled.

ColumnDefinition
Total activityactivity per mL × volume
Total proteinprotein per mL × volume
Specific activitytotal activity / total protein
Purification factor (fold)specific activity of step / specific activity of crude
Yield (%)total activity of step / total activity of crude × 100

Overall yield is the product of step yields, and so is overall purification. A step that doubles purity but halves the yield is only worth taking if the alternative route cannot reach the specification — which is why a single high-resolution affinity capture is prized over three mediocre steps.


Memory Tricks

Don’t Confuse

ThisvsThat
Dead-end filtration (cake builds)vsCrossflow (tangential sweep keeps it thin)
Microfiltration (cells, µm)vsUltrafiltration (proteins, kDa)
Concentration polarisation (reversible layer)vsFouling (adsorbed/deposited, needs cleaning)
Diafiltration (buffer exchange)vsUltrafiltration (concentration)
Salting-in (low salt raises solubility)vsSalting-out (high salt lowers it)
Ion exchange (bind low salt, elute high salt)vsHIC (bind high salt, elute low salt)
Size exclusion (no binding, no concentration, last)vsAffinity (binds, concentrates, first)
Purification factor (purity ratio)vsYield (recovery of activity)
Freeze-drying (sublimation under vacuum)vsSpray drying (hot gas, seconds)
\( \Sigma \) factor (centrifuge geometry)vs\( v_g \) (particle settling velocity)

Common Mistakes


Practice

All questions below are original practice questions written in CSIR NET style. They are not previous-year questions.

Level 1: Foundation

  1. In what order do the four stages of a downstream train run?
  2. A protein has pI 5.0 and the buffer is at pH 8.0. Which exchanger does it bind?
  3. How many diavolumes are needed to remove 99% of a freely permeating salt?
Answers
  1. Recovery, isolation, purification, polishing (RIPP).
  2. pH is above pI → the protein is negatively charged → an anion exchanger (DEAE or Q).
  3. \( C/C_0 = e^{-N} = 0.01 \Rightarrow N = \ln 100 = \mathbf{4.6} \) diavolumes (so 5 in practice).

Level 2: Part B Style

Q1 (practice). A monoclonal antibody is captured from clarified mammalian-cell supernatant in a single step to well over 95% purity. The most likely method is:

  • (a) size-exclusion chromatography
  • (b) Protein A affinity chromatography
  • (c) ammonium sulfate precipitation
  • (d) reversed-phase chromatography
Answer and explanation

(b). Protein A binds the Fc region specifically, so one capture step removes almost all host-cell protein. Gel filtration cannot be a capture step; precipitation gives only a few fold; reversed phase with organic solvent would denature the antibody.

Q2 (practice). Which sequence is the most sensible for an intracellular bacterial enzyme?

  • (a) Homogenise → centrifuge → gel filtration → ion exchange → ultrafiltration
  • (b) Centrifuge (harvest cells) → homogenise → remove debris → ammonium sulfate → HIC → ion exchange → gel filtration
  • (c) Ion exchange → homogenise → centrifuge → precipitation
  • (d) Ultrafiltration of whole broth → gel filtration → homogenise
Answer and explanation

(b). Harvest and wash cells first (removing spent medium), disrupt, clear the debris, concentrate by salting-out, go straight to HIC because the feed is already at high salt, then ion exchange after desalting, and finish with gel filtration for aggregates and buffer. (a) puts a low-capacity, dilution-prone gel filtration before the high-capacity steps; (c) and (d) attempt chromatography or filtration on whole cells.

Q3 (practice). During ultrafiltration the permeate flux stops increasing when the transmembrane pressure is raised further. The explanation is:

  • (a) the pump has reached its limit
  • (b) concentration polarisation — a denser gel layer forms at the membrane and becomes the controlling resistance
  • (c) the membrane pores have expanded
  • (d) the protein has denatured
Answer and explanation

(b). In the pressure-independent regime the retained solute accumulates in a polarised layer whose resistance grows with the applied pressure, so the net flux plateaus. The cure is hydrodynamic (crossflow velocity, turbulence promoters), not more pressure.

Q4 (practice). In an aqueous two-phase PEG/salt system, a protein has a partition coefficient \( K \) = 5 (top/bottom) and the phase volume ratio is 1.0. The fraction of the protein recovered in the top phase is:

  • (a) 50%
  • (b) 71%
  • (c) 83%
  • (d) 95%
Answer and explanation

(c). \( Y = KV_r/(1+KV_r) = 5/(1+5) = 0.833 \), i.e. 83%. Raising \( V_r \) raises recovery but dilutes the product — the usual trade-off.

Level 3: Part C Style

Q1 (practice). A purification gave the following data.

StepVolume (mL)Total protein (mg)Total activity (U)
Crude extract1,0005,00025,000
Ammonium sulfate (30–60%)2001,50020,000
Anion exchange6025015,000
HIC254011,000
Gel filtration10118,800

(i) Complete specific activity, fold purification and yield for each step. (ii) Which single step contributed most to purification? (iii) The gel-filtration step removes aggregates but costs 20% of the activity. If a customer accepts the HIC pool, what is the gain in delivered activity, and what is lost?

Answer and worked solution

(i)

StepSpecific activity (U/mg)FoldYield (%)
Crude25,000/5,000 = 5.01.0100
Ammonium sulfate20,000/1,500 = 13.32.780.0
Anion exchange15,000/250 = 60.012.060.0
HIC11,000/40 = 27555.044.0
Gel filtration8,800/11 = 80016035.2

(ii) Step-wise purification factors are 2.7, then 60.0/13.3 = 4.5 (anion exchange), then 275/60 = 4.6 (HIC), then 800/275 = 2.9. The HIC step is marginally the largest single contributor, with anion exchange effectively equal; ammonium sulfate (2.7-fold) and gel filtration (2.9-fold) give the smallest purification gains.

(iii) Stopping after HIC delivers 11,000 U instead of 8,800 U: 25% more activity (2,200 U), at 275 U/mg instead of 800 U/mg — roughly one-third of the final specific activity, with aggregates and any remaining size-distinct contaminants still present. For a technical-grade enzyme that is the right call; for an injectable product aggregate removal is non-negotiable because aggregates are an immunogenicity risk. The general lesson: each late step must be justified by what it removes, because by then there is the least contaminant left to remove.

Q2 (practice). A 50 L protein solution at 2 g L⁻¹ in 1.5 M ammonium sulfate must be concentrated tenfold and transferred into 20 mM phosphate buffer using a 10 kDa ultrafiltration membrane, with the salt freely permeating and the protein fully retained.

(i) What volume and concentration result from the concentration step, assuming no product loss? (ii) How many diavolumes of phosphate buffer at constant volume are needed to bring the ammonium sulfate below 1.5 mM? (iii) What volume of diafiltration buffer does that require, and why is it cheaper to concentrate before diafiltering?

Answer and worked solution

(i) Tenfold concentration: 50 L → 5 L, protein 2 → 20 g L⁻¹ (total 100 g retained).

(ii) Required reduction: 1.5 M → 1.5 mM is a factor of 1,000.

$$e^{-N} = 10^{-3} \Rightarrow N = \ln(1{,}000) = 6.91 \approx \mathbf{7\ diavolumes}$$

(iii) At constant volume, one diavolume = the retentate volume = 5 L, so 7 diavolumes = 35 L of buffer. Had the diafiltration been done at the original 50 L, the same 7 diavolumes would need 350 L — ten times the buffer, ten times the processing time and ten times the buffer-preparation and disposal cost. Hence the rule: concentrate first, then diafilter. The limit is protein solubility and viscosity at the concentrated point, and the risk of aggregation at 20 g L⁻¹.

Q3 (practice). Two peaks elute from a 20 cm column at 10.0 min and 12.0 min with baseline widths 1.0 min and 1.2 min. Which statements are correct?

(A) The resolution is about 1.8, i.e. better than baseline separation. (B) The plate number calculated from the first peak is about 6,400. (C) The corresponding plate height is about 31 µm. (D) Doubling the flow rate will necessarily improve resolution, because the peaks will be sharper.

  • (a) A, B and C
  • (b) A and D only
  • (c) B, C and D
  • (d) A, B, C and D
Answer and worked solution

(a) A, B and C.

  • (A): \( R_s = 2(12.0-10.0)/(1.0+1.2) = 4.0/2.2 = \mathbf{1.82} \); \( \ge 1.5 \) is baseline resolution. Correct.
  • (B): \( N = 16(t_R/w)^2 = 16(10.0/1.0)^2 = \mathbf{6{,}400} \) plates. Correct.
  • (C): \( \mathrm{HETP} = L/N = 200\ \mathrm{mm}/6{,}400 = 0.03125\ \mathrm{mm} = \mathbf{31\ \mu m} \). Correct.
  • (D): wrong. On the van Deemter curve \( H = A + B/u + Cu \), there is an optimum velocity. Above it the \( Cu \) (mass-transfer) term dominates, plates are lost, peaks broaden and resolution falls. Higher flow buys time, not resolution — and in preparative work it also lowers dynamic binding capacity.

Quick Revision Table

ItemFact
StagesRIPP: Recovery, Isolation, Purification, Polishing
EconomicsDownstream often 50–80% of cost; overall yield = product of step yields (0.85⁵ = 0.44)
DisruptionHomogeniser and bead mill at scale; enzymatic and osmotic shock are gentle; over-disruption micronises debris
Stokes\( v_g = d^2(\rho_p-\rho_L)g/18\mu \); centrifuge capacity \( Q = v_g\Sigma \)
CentrifugesTubular bowl, disc-stack (standard), decanter, basket
FiltrationDead-end (cake) vs crossflow (tangential); filter aids for mycelia
PrecipitationAmmonium sulfate (Cohn \( \log S = \beta - K_sI \)), solvents, PEG, isoelectric
MembranesMF µm, UF kDa, NF small solutes, RO ions
Flux\( J = \mathrm{TMP}/(\mu R) \); polarisation makes flux pressure-independent
Diafiltration\( C/C_0 = e^{-N} \); 5 DV = 99.3%, 7 DV = 99.9%
ChromatographyAffinity (capture), IEX (capacity), HIC (high salt binds), RP, SEC (last)
IEX ruleBelow pI → cation exchanger; above pI → anion exchanger
Performance\( N = 16(t_R/w)^2 \); \( \mathrm{HETP}=L/N \); \( R_s = 2\Delta t_R/(w_1+w_2) \), baseline at 1.5
van Deemter\( H = A + B/u + Cu \); an optimum velocity exists
Extraction\( Y = KV_r/(1+KV_r) \); penicillin by pH swing; ATPS is gentle and scalable
DryingLyophilisation (sublimation, cryoprotectants, collapse temperature) vs spray drying (fast, cheap)
TableSpecific activity = activity/protein; fold = ratio to crude; yield = activity ratio

Key Takeaways

What’s Next