Removing aggregates & host-cell-protein impurities from a Protein A-purified antibody using CEX / AEX chromatography as a polishing step
Every ion-exchange polishing step begins with a mode decision that shapes the entire unit operation: should the antibody bind the resin and be eluted (CEX), or should it flow through unretained while impurities bind (AEX)? The choice depends on the antibody's isoelectric point (pI), the impurity profile carried over from Protein A capture, and the purity target for the drug substance.
Ion-exchange chromatography separates proteins by exploiting differences in net surface charge, which arises from the ionization state of surface-exposed acidic (Asp, Glu) and basic (His, Lys, Arg) residues at a given buffer pH. A monoclonal antibody typically carries a pI between 7.5 and 9.5, meaning that at pH values below its pI the molecule is net positively charged and will bind a cation-exchange (CEX) resin bearing sulfopropyl (SP) or carboxymethyl (CM) ligands.
Crucially, no antibody preparation is charge-homogeneous. Post-translational and chemical modifications generate a spectrum of "charge variants":
• C-terminal lysine clipping/variable amidation — shifts net charge by discrete increments • Asparagine/glutamine deamidation — converts a neutral amide to a charged carboxylate, increasing acidity • Sialylation of Fc/Fab glycans — adds negatively charged sialic acid residues • Aspartate isomerization and succinimide formation — subtle pI shifts of 0.05–0.3 pH units • Aggregation — dimers and higher-order multimers present multivalent, spatially clustered charge patches that bind ion-exchange ligands with different apparent avidity than the compact monomer
These differences are small — often a fraction of a net charge unit — but ion-exchange resins, with their high ligand density and cooperative multi-point electrostatic interactions, can resolve them with remarkable selectivity. This is precisely why IEX, rather than size-exclusion alone, is the workhorse technology for aggregate and charge-variant polishing at manufacturing scale.
Two fundamentally different operating modes are available, and platform processes often use one or the other (or both in series):
CEX bind-and-elute: the load is conditioned to a pH below the antibody pI and a low conductivity, so the antibody binds the resin along with some charge-similar impurities. A rising salt (or pH) gradient then selectively displaces bound species in order of increasing apparent affinity, resolving the antibody monomer from aggregates and other charge variants as discrete elution peaks. This mode offers the highest resolving power and is preferred when aggregate levels are elevated (>2–5%) after Protein A capture.
AEX flow-through: the load is conditioned to a pH above the antibody pI, so the antibody carries near-neutral to net-negative surface character only at the most acidic patches and largely does not engage the positively-charged Q or DEAE resin, while acidic impurities — host-cell proteins (HCP, generally more acidic, pI 4–7), residual DNA (highly polyanionic), leached Protein A ligand, and a fraction of aggregates carrying negative charge patches — bind tightly and are retained. The antibody is collected in the flow-through/wash, in a single pass, at high throughput and with minimal buffer consumption.
Many commercial platforms run CEX first (to remove the bulk of aggregates and resolve charge variants) followed by an AEX flow-through polishing step (to scavenge residual HCP, DNA, and leached Protein A) — combining the resolving power of bind-elute chromatography with the throughput and orthogonality of flow-through capture.
The mode decision is rarely made in isolation: it is set jointly with the upstream Protein A elution pH and downstream viral filtration train, since each step must hand off a feed stream within the conductivity, pH, and turbidity specification the next unit operation requires.
Before the polishing column ever sees the feed, the Protein A eluate pool — typically held at low pH for viral inactivation — must be conditioned to the pH and conductivity window that favors selective binding. Loading is then executed at a flow rate that balances residence time against process throughput.
The Protein A eluate emerges at pH 3.0–3.6, both to elute the antibody from the Protein A ligand and to provide a dedicated low-pH viral inactivation hold (typically 30–60 minutes) that inactivates enveloped viruses. Before this stream can be loaded onto an ion-exchange column it must be:
• Neutralized — titrated to the target load pH (pH 5.0–5.5 for CEX; pH 7.0–8.0 for AEX) using a base such as Tris or sodium hydroxide • Diluted or diafiltered — to bring conductivity below the threshold at which competing counter-ions (Na+, Cl-) would block antibody-resin electrostatic interaction • Clarified — depth- or 0.2 µm-filtered to remove any precipitate formed during neutralization, protecting the resin bed from fouling
This conditioning step is itself an orthogonal impurity-clearance opportunity: neutralization-induced precipitation of denatured HCP and lipids, formed transiently at intermediate pH, is removed by the clarifying filter before it ever reaches the column.
Column loading is characterized by the dynamic binding capacity (DBC) — the mass of antibody the resin can bind per liter before product begins to break through into the flow-through at a defined threshold (commonly 5 or 10% of feed concentration). DBC depends on residence time (longer residence time allows deeper pore diffusion into the resin bead, raising apparent capacity), ligand density, and the antibody's own diffusivity and charge.
In CEX bind-elute mode, the column is deliberately loaded below its maximum DBC (typically 60–80% of DBC) to leave headroom for resolution during elution — overloading broadens elution peaks and degrades separation between monomer and aggregate.
In AEX flow-through mode, by contrast, the column is loaded far beyond a conventional "binding capacity" concept, since the product itself is not meant to bind — loading is instead limited by the impurity-adsorption capacity of the resin for HCP and DNA, which is typically much higher (100s of g/L) than a conventional bind-elute DBC, enabling very high product throughput per liter of resin per cycle.
Once loaded, the column becomes a molecular sorting device: monomer, aggregate, and residual process impurities distribute across the resin bed according to their apparent binding affinity, driven by the same small charge differences that define the antibody's charge-variant landscape.
Antibody aggregates — dimers, trimers, and higher-order soluble oligomers formed by self-association during expression, low-pH Protein A elution, or freeze-thaw stress — are not merely a potency-diluting impurity. Multimeric protein assemblies present repetitive, densely-spaced epitopes that are far more effective at cross-linking B-cell receptors than monomeric protein, a structural feature strongly associated with breaking immune tolerance and triggering anti-drug antibody (ADA) responses in patients.
Regulatory guidance (ICH Q6B and associated biologics guidances) therefore expects aggregate content to be tightly controlled and monitored by orthogonal methods (SEC-HPLC, analytical ultracentrifugation, field-flow fractionation) throughout the manufacturing process, with typical release specifications requiring high-molecular-weight (HMW) species below 1–2% of total protein. Because aggregates differ from monomer by only a small increment in apparent surface charge — arising from clustered, multivalent charge patches rather than a change in amino acid sequence — ion-exchange chromatography, with its high resolving power for subtle charge differences, is one of the few technologies capable of reducing aggregate content by an order of magnitude in a single polishing step.
A well-optimized CEX polishing step routinely reduces HMW aggregate content from several percent post-Protein A to well below 0.5% in the pooled product — directly derisking immunogenicity for the clinical or commercial product.
Beyond aggregate removal, the ion-exchange step is also where the antibody's intrinsic charge-variant heterogeneity becomes visible and, to some extent, controllable. Analytical charge-variant profiling — by imaged capillary isoelectric focusing (icIEF) or analytical CEX-HPLC — typically resolves a main peak flanked by acidic variants (deamidation, sialylation, glycation) eluting earlier and basic variants (C-terminal lysine, succinimide, aspartate isomerization) eluting later in a CEX gradient.
Because charge variants can differ subtly in potency, Fc receptor binding, or pharmacokinetic clearance, the relative proportion of acidic and basic species is itself a critical quality attribute. Selecting the pool boundary on the preparative IEX chromatogram is therefore not solely an aggregate-clearance decision — it also fixes the charge-variant composition of the final drug substance, tying the polishing step directly to product identity and comparability across manufacturing campaigns.
The elution phase is where separation becomes physically visible: in CEX mode a rising ionic-strength (or pH) gradient progressively displaces bound species from the resin in order of increasing apparent affinity, generating a UV280 chromatogram with resolved peaks; in AEX mode the antibody simply exits the column unretained while impurities remain immobilized on the resin.
As the mobile-phase ionic strength rises, competing counter-ions progressively displace bound protein from the resin's charged ligands, with each species desorbing once the ambient salt concentration exceeds its own binding threshold. Weakly-bound impurities (residual HCP and DNA that co-eluted at low affinity) desorb first, the antibody monomer main peak elutes in the middle of the gradient, and aggregates — which bind more avidly through multivalent charge-patch interactions — elute last, though the precise order can invert depending on resin chemistry and aggregate conformation.
Gradient slope controls the classic resolution/throughput trade-off: a shallow gradient (long column-volume span, small salt increment per CV) maximizes the time each species spends re-equilibrating between the mobile and stationary phase, producing narrow, well-separated peaks at the cost of extended run time and diluted pools. A steep gradient (or a step elution, jumping directly to a high-salt buffer) elutes everything in a compressed volume, dramatically shortening cycle time but co-eluting monomer with closely-related charge variants and aggregates — sacrificing resolution for speed.
In AEX flow-through mode there is no gradient at all: the column is run isocratically at the loading pH and conductivity, and the entire separation happens instantaneously at the point of contact between the feed stream and the resin bed. The antibody, carrying little to no net negative charge at the operating pH, passes through the packed bed largely unretained and is collected directly as the column effluent — typically in a single pooled fraction spanning the load and a short chase-buffer wash.
Meanwhis, acidic impurities and a fraction of the aggregate population bind the positively-charged resin and remain immobilized until the column is stripped and regenerated at the end of the cycle (a step that discards, rather than analyzes, the bound impurity fraction). Because there is no gradient to run and no peak to wait for, AEX flow-through polishing is markedly faster and more resin-efficient than CEX bind-elute — which is why it is so often deployed as the final polishing pass after CEX has already done the heavy lifting on aggregate removal.
The gradient-slope trade-off is not academic: doubling the gradient length from 15 to 30 column volumes can improve monomer/aggregate resolution enough to lift pooled purity by a full percentage point, but it also roughly doubles processing time and buffer consumption per batch — a trade-off weighed explicitly during process characterization.
The final act of the polishing step is a decision, not a physical operation: which fractions of the eluted (or flow-through) chromatogram get combined into the pooled drug substance intermediate. Pool boundaries are set against a matrix of in-process analytics, trading yield against purity to hit the release specification.
Pool selection is driven by a small panel of fast, at-line or rapid-turnaround assays performed on individual fractions (or on UV280/conductivity trace features in real time):
• SEC-HPLC — quantifies HMW aggregate and low-molecular-weight (LMW) fragment content per fraction; the pool is typically bounded to exclude fractions where HMW exceeds a pre-set threshold (often 0.5–1%) • HCP ELISA — measures residual host-cell protein, generally trending lowest in the fractions richest in monomer • qPCR for residual DNA — confirms clearance to well below the regulatory expectation (commonly <10 pg/mg product, or ~100 pg/dose) • Charge-variant profiling (icIEF/CEX-HPLC) — confirms the acidic/main/basic variant ratio of the pooled fractions matches the target product profile
Because narrowing the pool always trades yield for purity — excluding shoulder fractions removes trailing aggregate or leading impurity peaks but also discards some monomer — pool boundaries are typically fixed during process characterization studies and then applied consistently, run to run, as a validated in-process control rather than decided ad hoc on each batch.
Ion-exchange polishing does not operate in isolation: it is the second of three purification stages in the now-standard mAb "platform" purification train:
1. Protein A affinity capture — captures the antibody with high selectivity directly from clarified harvest, removing the bulk of host-cell protein, DNA, and media components in a single bind-elute step; the low-pH elution simultaneously serves as a dedicated viral inactivation hold for enveloped viruses
2. Ion-exchange polishing (this step) — resolves monomer from aggregates and residual charge-variant impurities that Protein A cannot distinguish (since Protein A binds the Fc region regardless of aggregation state or minor charge differences), while further reducing HCP, DNA, and any leached Protein A ligand
3. Virus-retentive nanofiltration — a size-exclusion-based orthogonal viral clearance step (typically 15–20 nm pore-size membranes) that physically retains any residual virus particles by size rather than charge or affinity, and which performs far more reliably on a low-aggregate, low-turbidity feed stream — meaning the IEX polishing step directly protects viral filter capacity and flux by removing the aggregate and particulate burden that would otherwise foul the membrane
Each step contributes an independent, mechanistically distinct impurity-clearance mechanism — affinity, charge, and size — so that combined log-reduction values for viral clearance and impurity clearance multiply rather than merely add, giving the overall three-step platform train the redundancy and robustness regulators expect from a commercial biologics process.
A typical platform train takes a clarified harvest at roughly 1–10 mg/mL HCP-laden feed and delivers a drug substance intermediate exceeding 99.9% monomeric purity with HCP and DNA reduced by 4–6 logs overall — the ion-exchange polishing step alone routinely contributing 2–3 logs of that total HCP clearance and the majority of the aggregate reduction.