Purifying Antibody Fragments and Bispecific Antibodies

1. Translate Molecular Architecture into Purification Behavior

Map the molecule before selecting a resin: constant domains, light-chain or single-chain elements, engineered interfaces, affinity tags if present, linker sequences, disulfide pattern, theoretical size, charge, hydrophobic patches, and expected assembly. Identify which domains remain accessible in the folded product.

List product-related impurities from expression and assembly. These can include unpaired chains, half molecules, fragments, homodimers, mispaired bispecifics, aggregates, and correctly sized but functionally incorrect species. Process impurities from host cells remain important, but the hardest separation may be between structurally similar product variants.

Choose assays that can see the problem

Total protein and a single size measurement are rarely sufficient. Combine identity, intact or subunit mass, size distribution, charge or hydrophobic profile, and binding to each intended target as appropriate. A bispecific molecule can have the expected mass while lacking one functional arm.

2. Choose Capture by Accessible Domain and Selectivity

Determine whether an affinity interaction recognizes the format and whether that interaction is present on all unwanted species. A constant-domain capture may bind the desired molecule and certain homodimers equally. A tag can enable selective capture during research, but its removal creates another purification and verification step.

Test antibody affinity media with representative feed. Measure dynamic capacity, breakthrough, recovery, impurity profile, and product quality. Do not infer capacity from a conventional full-length antibody when the fragment has different size, valency, or domain accessibility.

Use wash conditions to remove host impurities and weakly associated variants without disrupting the molecule. If the product binds weakly, a milder wash or shorter residence may improve recovery. Confirm that apparent loss is not caused by proteolysis or aggregation during feed preparation.

3. Protect Formats with Narrow Stability Windows

Screen pH, conductivity, temperature, concentration, and time before locking the chromatography sequence. Some fragments lack stabilizing domains and can be more sensitive to low pH, interfaces, or concentration. Bispecific formats can expose engineered surfaces that alter solubility and hydrophobic interaction.

Evaluate both immediate and delayed quality after stress. A sample can look monomeric immediately after elution and aggregate during a later hold. Use realistic concentration and container surfaces. Include target-binding assays because structural changes may not be visible by size alone.

Design neutralization and buffer transitions to avoid local extremes. Keep collection vessels mixed without foam, and limit unnecessary transfers. If the format requires a stabilizing additive, verify its compatibility with resin binding, analytical methods, and downstream operations.

4. Separate Assembly Variants with Orthogonal Mechanisms

Mispaired or homodimeric species may differ only slightly in charge, hydrophobicity, exposed domains, or avidity. Use analytical profiling to identify exploitable differences before screening preparative modes. Charge-based separation can resolve engineered charge asymmetry; hydrophobic or multimodal interactions can reveal surface differences; affinity combinations can select for the presence of two distinct domains.

Design the sequence so each step removes a defined variant family. If the initial capture binds all Fc-containing species, a later step must distinguish correct bispecific assembly. If a dual-affinity strategy is used, prove that sequential binding enriches molecules possessing both functional elements rather than simply selecting high avidity.

Retain small fractions across variant peaks and test both binding functions. Pooling by absorbance or a single purity assay can accept a functionally incomplete population. Map purity versus recovery before choosing boundaries.

5. Use Charge-Based Chromatography Deliberately

Estimate isoelectric behavior only as a starting point; engineered domains, glycosylation, clipping, and conformational differences can shift actual retention. Screen pH and conductivity in a stability-compatible region. Decide whether bind-and-elute or flow-through creates the clearest separation from the critical variant.

Use a shallow gradient to map closely spaced species, then evaluate whether a practical step or shorter gradient can reproduce the result. Correct for system delay and extra-column dispersion. Small analytical peaks may merge at preparative load.

Challenge loading because displacement and overload can change variant order or peak overlap. Analyze boundary fractions with format-specific assays. The best condition is one with a robust pooling window, not necessarily the longest chromatographic resolution under low load.

6. Apply SEC with Realistic Expectations

SEC can separate aggregates, monomers, fragments, and unassembled chains when their hydrodynamic sizes differ sufficiently. It is nonbinding and usually operated isocratically, but resolution depends strongly on sample volume, concentration, bed efficiency, flow, and pore range.

Select pores that allow the desired format to enter without total inclusion or exclusion. Test the real molecule because shape can affect hydrodynamic behavior. A compact fragment and an elongated construct of the same mass need not elute identically.

Limit injection volume to the resolution required and account for pool dilution. At high protein concentration, nonideal interactions or self-association can distort peaks. Evaluate recovery on all relevant surfaces and determine whether SEC belongs in a high-value polishing role or mainly in analytical monitoring.

7. Control Clipping and Proteolysis during Purification

Differentiate intended fragments from proteolytic products. Minimize uncontrolled hold time, maintain an appropriate temperature, and consider compatible protease control where scientifically justified. Clarification delays can create new variants before chromatography begins.

Track N- and C-terminal integrity or domain-specific identity across steps. A clipped product may retain one binding assay and lose another. Use nonreducing and reducing analyses, mass-based methods, and functional testing as appropriate.

If clipping increases during low-pH elution or concentration, adjust exposure, buffer, or sequence rather than relying on a later column to remove all products. Preventing formation generally preserves more yield than separating a growing impurity family.

8. Close Mass Balance across Closely Related Species

Quantify desired product separately from total antibody-related material. A step can show high total recovery while converting or co-pooling unwanted variants. Use species-resolved assays for feed, flow-through, wash, product, side fractions, and strip.

Report accepted recovery using the fraction population that meets all critical identity and function criteria. This prevents a superficially high yield from hiding loss of correctly assembled product.

9. Scale Sensitive Separations by Preserving the Critical Window

Preserve resin identity, bed geometry or justified residence time, load per resin volume, inlet conditions, phase volumes, and gradient slope per column volume. Recheck system mixing and collection delay because narrow variant peaks are sensitive to dispersion.

At larger scale, confirm feed homogeneity and concentration. Slowly associating species or settling particulates can change the material presented during a long load. Keep time from feed preparation to column entry within the stability evidence.

Challenge likely variability: expression ratio, clipping, aggregate burden, pH, conductivity, load, and resin age. A method developed on an unusually clean feed may not have enough selectivity for routine production. Define an operating region with a measurable pool decision.

10. Resolve a Bispecific Mispairing Problem through Fraction Mapping

Imagine a bispecific construct produced with the desired heterodimer plus two homodimers, trace half molecules, and aggregates. All Fc-containing species bind the initial affinity medium, so capture removes process impurities but does not establish correct assembly. The capture pool is analyzed by intact mass, subunit analysis, SEC, charge profile, and binding to both targets.

A charge gradient reveals three overlapping populations. Fractions on one side of the main peak retain binding to target A but show reduced binding to target B; mass analysis identifies enrichment of one homodimer. Fractions on the opposite side contain more aggregate and a second mispaired form. This fraction map provides a mechanistic basis for a preparative charge step.

Test whether the separation survives preparative load

The team increases load gradually while holding residence time and inlet conditions constant. At high load, the first homodimer displaces into the desired-product region, narrowing the pool window. The operating load is set below that point, and gradient delivery is verified at the column inlet because small composition differences shift the closely spaced variants.

A subsequent SEC step removes remaining aggregates and half molecules, but sample volume is limited to preserve resolution. The team compares a narrow SEC pool with a broader pool by dual-target binding and intact-complex recovery. The final pool rule is based on both functions and monomer state, not absorbance symmetry.

During development, apparent recovery drops after low-pH affinity elution. A no-column hold control shows that the format aggregates during collection rather than on the resin. Faster neutralization and a stabilizing buffer improve correctly assembled recovery without changing capture chemistry. This prevents a polishing column from being used to correct avoidable upstream damage.

Tag removal is evaluated in a separate route. Cleavage improves the intended final construct but creates uncleaved material, free tag, and protease. A subtractive affinity step removes those species, and the entire route is compared on accepted bispecific mass. A higher total-protein yield from incomplete cleavage is rejected because it fails the defined product identity.

At scale, the process includes feed-age limits, neutralization mixing, gradient verification, pool-delay correction, and enhanced early-lot analytics. The final control strategy identifies capture as process-impurity removal, charge chromatography as assembly-variant control, and SEC as aggregate and size-state polishing.

This case illustrates why a bispecific process cannot be judged with one conventional antibody assay. The selected product must contain the intended subunits, correct pairing, both binding functions, acceptable aggregate, and stable behavior after concentration. Each analytical result is connected to a specific chromatographic decision.

11. Build a Format-Specific Purification Train

A fragment workflow may use tag or domain-selective capture followed by charge-based polishing and SEC if structural homogeneity is critical. A bispecific workflow may require capture, an orthogonal step aimed at homodimers or mispairs, and a final aggregate or charge-variant polish. The appropriate sequence depends on which species co-bind at each step.

Evaluate cumulative recovery, both binding functions, monomer or assembly state, pool concentration, and downstream compatibility. Eliminate a step only when its impurity control is replaced by upstream improvement or another demonstrated mechanism.

For format-specific support, provide domain architecture, expression system, analytical profiles, target-binding assays, feed volume, and quality requirements through protein purification services. The workflow should be built around the molecule that exists, not a conventional-antibody template.

Establish reference materials for desired product and major assembly variants when feasible. They support method resolution, fraction identity, and training, but must be stored and qualified so their own degradation does not distort comparisons. When a variant standard is unavailable, use orthogonal evidence and retain characterized fractions.

Review the purification design with expression-development colleagues. A modest change in chain ratio, secretion, linker, interface engineering, or protease control can reduce a variant more effectively than another chromatography step. Quantify the upstream change against the same accepted-product yield and function endpoints.

At transfer, define how both target-binding functions are tested and how discordant results are investigated. If one assay drifts, the process should not be adjusted until analytical performance is separated from a true assembly change. Preserve raw fraction data and construct identity throughout the handoff.

Before release of a preparative method, test a deliberately variant-rich feed or a justified worst-reasonable assembly profile. Confirm that pool analytics detect the enriched impurity and that the selected loading and boundaries still protect both functions. This challenge demonstrates the method's discriminatory ability more directly than repeated runs with only ideal feed.

References

1. Carta G, Jungbauer A. Protein Chromatography: Process Development and Scale-Up. 2nd edition. Wiley-VCH; 2020. doi:10.1002/9783527824045
2. ICH. Q6B: Specifications—Test Procedures and Acceptance Criteria for Biotechnological/Biological Products. 1999.
3. Cassedy A, O’Kennedy R. Antibody purification using affinity chromatography. In: Ayyar BV, Arora S, eds. Affinity Chromatography. Vol 2466. Springer US; 2022:3-22. doi:10.1007/978-1-0716-2176-9_1