What makes fusion proteins difficult

Three properties make fusion protein purification harder than purifying either partner alone: heterogeneity, clipping and aggregation, each originating in the construct design rather than the process.

Heterogeneity. The fused partner brings its own post-translational modifications. A receptor domain fused to Fc adds N-glycosylation sites with different occupancy and sialylation than the Fc glycan, and flexible linkers can carry O-glycans. Each variant differs in charge and hydrophobicity, so the product elutes as a broad population, and the specification must define which part is acceptable.

Clipping. Linkers such as (G4S)n and the Fc hinge are exposed, unstructured sequences that host proteases cut during culture and harvest. The clipped fragments retain one domain, often the Fc, and co-purify through Protein A capture. Clipping in the bioreactor can only be removed downstream, at a yield cost.

Aggregation. The two domains have different pH and conductivity stability windows. Conditions routine for an antibody, such as a pH 3.5 elution or a low-pH viral inactivation hold, can unfold a receptor domain and seed aggregates. Hydrophobic patches at the domain interface also drive self-association during UF/DF.

Fc-fusion versus albumin-fusion versus tag-based constructs

The fusion partner decides the capture chemistry and the dominant impurity, so fusion protein purification is planned by construct class before any resin is screened.

Fc-fusions dominate the therapeutic class because Protein A gives them a platform-like capture step. Albumin-fusions and tag-based constructs need a custom first step, which is why fusion protein purification services are scoped construct by construct rather than at a platform price.

Capture strategies by construct type

Capture is where fusion protein purification departs most from the antibody platform: the resin may be the same, but elution conditions, loading limits and the impurities that follow the product are not.

Fc-fusion on Protein A. Binding is through the Fc CH2–CH3 interface and works as for an antibody; elution is the problem. Where the partner domain unfolds below pH 4, options are elution at pH 4.0–4.5 with arginine or other additives, a stepped pH gradient, or an alternative ligand with milder elution. Dynamic binding capacity is often lower than for a mAb because the partner adds steric bulk, so loading is measured, not assumed.

Albumin-fusion. Albumin-affinity resins give a selective capture; otherwise, anion exchange in bind-elute mode exploits albumin's low isoelectric point, binding the fusion at neutral pH while basic host proteins flow through. A hydrophobic interaction step then separates intact fusion from free albumin.

His-tagged constructs on IMAC. Nickel or cobalt chelating resins bind the His-tag; elution is by imidazole or pH. Metal leaching, histidine-rich host proteins and imidazole removal are the recurring issues. IMAC is a development-stage tool for most therapeutic programs; the tag is cleaved and the tag-free protein is purified on ion exchange.

Untagged, non-Fc fusions. Capture relies on the protein's own charge or hydrophobicity. Resin screening across cation exchange, anion exchange, mixed-mode and HIC at several pH values is the only way to find the selective step, and it is done at 1 mL scale first.

Tag removal with site-specific proteases

Tag removal is a unit operation in fusion protein purification with its own parameters, yield and clearance requirements, not a footnote to capture. The protease, the cleavage site engineered into the construct and the post-cleavage separation are chosen together.

Cleavage runs at a defined enzyme-to-substrate ratio, temperature and time, with completeness tracked by CE-SDS and confirmed by intact mass. A chromatography step then removes the protease, the free tag and any uncleaved fusion; a reverse IMAC step does all three in one pass for His-tagged protease and tag. For clinical material, the protease is a process-related impurity with a clearance claim, and the enzyme grade determines the supporting documentation.

Detecting clipped and misfolded variants

Clipped and misfolded species are the variants most likely to escape a standard antibody panel, so fusion protein purification needs analytics built around the linker and the domain interface.

• CE-SDS, reduced and non-reduced. Reduced CE-SDS resolves clipped chains by size; non-reduced shows whether a clipped chain is still disulfide-linked to its partner. A band that appears between harvest and capture pool localizes the clipping to a step.
• Intact and subunit mass by LC-MS. Intact mass confirms the full-length fusion; subunit analysis after reduction and deglycosylation assigns each clipped species to a cleavage position.
• Peptide mapping with N- and C-terminal analysis. Identifies the exact clip site in the linker or hinge and the N-terminal variants left by tag removal, and quantifies oxidation and deamidation at the domain interface.
• SEC-HPLC and DLS. Aggregates and high-molecular-weight species; DLS catches early self-association in formulation screening.
• icIEF and HIC-HPLC. Charge heterogeneity from sialylation, and misfolded isoforms that share the correct mass but expose different hydrophobic surfaces.
• Potency assays for each domain. A binding assay for the partner domain and, for Fc-fusions, an FcRn or Fc-receptor binding assay confirm both halves are functional.

The practical rule in fusion protein purification development is to run the full panel on the clarified harvest, the capture pool and the final bulk, so that each variant is traced to the step that creates it and the step that removes it.

Scale-up considerations for fusion proteins

Fusion protein purification scales on the same rules as any protein, constant bed height and residence time with growing column diameter, but four points need explicit checks before the first pilot batch.

1. Viral inactivation. If the construct cannot tolerate the low-pH hold used for antibodies, the alternative, usually solvent/detergent treatment, adds reagent removal and a different clearance claim.
2. Hold times and temperature. Clipping continues in process intermediates. Pool hold times harmless for a mAb may need shortening or 2–8 °C storage, demonstrated in small-scale hold-time studies.
3. UF/DF concentration. Aggregation at the domain interface rises with concentration. Run the final concentration step with formulation excipients already present, and verify membrane loading and crossflow for the specific protein.
4. Resin lifetime. Milder elution can leave product on Protein A, so cleaning and lifetime studies must use the actual fusion protein, not platform antibody data.

A fusion protein that has been through this sequence at pilot scale carries far less risk into a cGMP campaign, which is why fusion protein purification services are best engaged at process development, not at technology transfer.