Quantifying Charge Heterogeneity by Capillary Isoelectric Focusing: The Limits of pI Calibration, Focusing Completeness and Peak-Area Reporting

1 Introduction

Charge heterogeneity is usually covered by two methods: cation-exchange chromatography (CEX) and capillary isoelectric focusing (cIEF / icIEF). Regulatory dossiers often state that "both methods were used to determine charge variants and the results agree". Methodologically this statement is problematic—the two methods do not measure the same thing.

CEX separates by the strength of electrostatic interaction between the protein and a negatively charged stationary phase, and its output is a retention time. That retention time depends on column lot, salt-gradient slope, pH and column temperature; a change in any of them shifts the peaks. The CEX "acidic peak percentage" is therefore a relative quantity that holds only within a specific separation system, and comparison between laboratories requires system suitability first.

icIEF separates by the point at which the protein's own net charge is zero, and its output is the pI. The pI is determined mainly by the primary structure and post-translational modifications of the molecule, so it supports lot-to-lot comparison when method conditions are controlled. But the reported value is an apparent pI calibrated against markers, and systematic differences can still arise between platforms, ampholyte and marker combinations; comparisons across instruments or laboratories presuppose defined method conditions and system suitability. That is the value of icIEF and the reason it carries high weight in release and comparability studies.

Yet there are four links in this chain that raise no alarm:

The pI is calculated, not read. The instrument records the pixel position of a peak in the detection window; the pI is converted through a marker regression. If the markers are poorly chosen, cover too narrow a range or regress poorly, the pI cannot be trusted—and the software will not warn you.

There is no objective signal that focusing is complete. When voltage is applied the current falls, but "how far it must fall" is a judgement. Under-focusing and over-focusing both produce readable peaks, and both can give a wrong pI or broadened peaks.

The normalization denominator goes unexamined. Peak-area normalization requires the denominator to be "the sum of all protein peak areas". If system peaks, marker peaks or baseline artefacts are included, the denominator grows and every percentage is proportionally understated.

The detection wavelength hides an assumption. icIEF detects at 280 nm, and the molar absorptivities of the charge variants are not strictly equal; normalization implicitly assumes "equal response factors".

This article takes these four links apart and gives the basis for judgement and the reporting requirements for each.

2 Principles

2.1 Focusing: building the pH gradient and self-positioning of the protein

The capillary is filled with carrier ampholytes, and its two ends are immersed in an acidic anolyte and a basic catholyte. When voltage is applied, the ampholytes order themselves by pI along the capillary, forming a pH gradient that rises continuously from anode to cathode.

Once in this gradient, a protein sitting at a pH more acidic than its pI is positively charged and migrates towards the cathode; at a more basic pH it is negatively charged and migrates towards the anode. Both movements push it towards the point of zero net charge, where it stops migrating. Diffusion and focusing reach equilibrium and the protein forms a narrow band.

This self-positioning mechanism has three direct consequences:

• The pI is positional information, not intensity information. Peak height is far less reproducible than peak position, because height depends on load, adsorption and detector linearity, while position depends only on the gradient.
• Focusing is a dynamic equilibrium, not a static deposit. Band width results from competition between focusing (narrowing) and diffusion (broadening), so there is a suitable range of focusing time; beyond it the band may widen again, for example through thermal effects or gradient drift.
• The gradient is built by ampholytes, so its linearity is limited. Commercial ampholytes give steeper gradients at the pH extremes (<3 and >10), where resolution falls. This is the physical reason why proteins with extreme pI show poorer recovery and resolution.

2.2 The pI is a regression result: markers, pixels and linearity

ChP General Chapter 3129 is explicit about how the pI is calculated: the pI of each isoelectric-point marker is regressed linearly against its pixel position, and the pixel position of each charge variant is entered into the regression equation to obtain its pI.

This sentence contains three constraints, and each can fail:

The markers must bracket the target. If the target pI lies between the highest and lowest markers, it is interpolated; outside them, it is extrapolated. The uncertainty of extrapolation grows rapidly with distance and is not reflected in the regression R²—R² only shows that the markers themselves fit well, not that the extrapolated range is reliable.

Marker pI values are nominal. The markers carry their own uncertainty (typically around ±0.05 pH), which propagates as a systematic bias into every result. When lots are compared using markers from different suppliers, the systematic pI bias can reach 0.05–0.1 pH—enough to swallow a real small shift.

Pixel-to-pH linearity holds only in the linear part of the gradient. The gradient may deviate from linearity at the ends, and a linear regression averages such curvature away. Marker residuals that are positive at both ends and negative in the middle suggest systematic curvature; whether it really comes from steepening at the ends needs more markers or independent data to confirm. Until then, narrow the working range instead of continuing to use the whole regression line.

2.3 Resolving power: 0.02 pH is a typical value under specific conditions

ChP General Chapter 0542 states that gels prepared with carrier ampholytes can usually separate proteins whose pI differs by about 0.02 pH units, and that immobilized pH gradients can improve resolution further. This is the typical capability under specific gel IEF conditions.

The figure is a useful reference point for judgement, not a fixed boundary for every capillary method and sample. Actual resolving power should be taken from data for the specific sample and method. It suggests that:

• the closer the ΔpI of neighbouring peaks is to the method's actual resolving power, the less raising the voltage or extending focusing will separate them—often both peaks simply broaden together;
• when a report claims "five acidic variants detected", the first question is not the integration parameters but whether each neighbouring ΔpI lies above the resolving power demonstrated for that method;
• when several modifications combine and neighbouring components differ by a very small ΔpI, the peak may appear broadened rather than split—a shape that is easily misread. Whether such species can be resolved should be judged from resolution data in method development, not inferred from the type of modification.

2.4 The response-factor assumption behind 280 nm detection

icIEF detects proteins at 280 nm. The chapter reports relative peak areas under the specified detection conditions; reading relative peak areas as molar proportions assumes that every peak has the same molar absorptivity.

For charge variants of intact IgG the assumption approximately holds: deamidation, residual C-terminal lysine and terminal sialylation of N-glycans change only a few residues or sugars, and their effect on absorbance at 280 nm is usually a few per cent, smaller than the uncertainty of the integration itself.

But the assumption clearly fails in some situations:

• Samples containing fragments. Fab and Fc differ substantially in aromatic-residue composition and therefore in molar absorptivity; if intact molecules and fragments coexist, area ratios deviate systematically from molar ratios.
• Samples with markedly different glycoforms. Glycans do not absorb at 280 nm. The molar-mass differences caused by glycoforms mainly affect the conversion from molar to mass proportions; a difference in molar mass alone does not imply a change in molar response at 280 nm.
• Aggregates. The 280 nm response of aggregates is affected by light scattering, so their apparent absorbance is high.

The assumption that "relative peak area equals molar proportion" should therefore be stated in the method's limitations rather than taken for granted.

3 Experimental method

3.1 Conditions of Method 1 in ChP General Chapter 3129

ChP 2025 Volume IV General Chapter 3129 specifies that charge variants of monoclonal antibodies are determined by whole-column imaged capillary isoelectric focusing (icIEF, Method 1) or capillary isoelectric focusing (cIEF, Method 2). Charge variants are separated by their isoelectric points according to the capillary electrophoresis chapter (General Chapter 0542), the pI is determined and percentages are calculated. The elements of Method 1 are:

Two details deserve separate comment.

Methylcellulose is not an excipient; it suppresses flow. It raises viscosity and suppresses electro-osmotic flow and convection, so the narrow bands formed by focusing are not smeared by fluid movement. Poor preparation (incomplete dissolution, wrong concentration) shows up as broader peaks without necessarily shifting the pI—similar to under-focusing and easily misjudged.

The sample compartment (4–10 °C) and the capillary (18–25 °C) are two separate temperature controls. The first prevents degradation or aggregation while samples wait for injection; the second sets conductivity and viscosity during focusing. Lowering the sample-compartment temperature to "protect the sample" is reasonable, but if the capillary temperature is lowered too, the focusing kinetics change and the focusing time must be re-confirmed.

3.2 Desalting: a necessary step, not an option

Chapter 3129 is explicit: samples with high salt concentration must be desalted before analysis.

The reason is conductivity. During focusing, the current in the capillary is carried by both the carrier ampholytes and the ions in the sample. If the sample brings in a high salt load:

1. with a fixed applied voltage and capillary length, the average field strength is set by voltage and length; but high salt raises conductivity and current and changes the local field distribution, heat load and gradient formation, so focusing may not complete within the set time;
2. the higher current increases Joule heating; viscosity in the warm zone drops, diffusion intensifies and bands widen;
3. salt ions redistribute within the gradient, the local pH is disturbed and the pI readings shift.

Together these appear as "broader peaks, slightly shifted pI, poor reproducibility". Such problems are often blamed on a "dirty capillary" or "lot differences", when they are in fact matrix effects.

The goal of desalting is not to remove all salt but to bring conductivity down to a level the method can tolerate. That level should be determined during method development with a salt-spiking study: add increasing concentrations of NaCl to the same sample, observe when the pI and peak width begin to deviate, and write that concentration into the SOP as the method's salt tolerance limit.

3.3 Sequence and system suitability

Chapter 3129 specifies the injection order: system suitability solution, blank, reference standard, then test samples. The system suitability solution is injected at least twice at the start of the sequence and at least once at the end; the blank and reference standard once each.

The design intent is threefold:

• At least two injections at the start separate the first-injection effect from real results. The first injection often shows abnormal peak shape because the capillary wall is not yet equilibrated; the difference between the two injections is itself useful information.
• At least one injection at the end captures drift. If the end result differs markedly from the start, the data of the whole sequence must be re-evaluated, not just the last samples.
• The blank comes after system suitability and before the reference standard to confirm first that the integration region is clean. The chapter requires both isoelectric-point markers to be detected in the blank, and the sample integration region between the marker peaks to be free of negative peaks, spikes and other non-protein peaks that would interfere with integration. This bears directly on the denominator discussed in §4.2.

3.4 Choosing between Method 1 and Method 2

Method 1 (icIEF) images the whole column and captures the pI distribution of the entire capillary in a single exposure, without a migration step, so there is no "migration-time drift" error source and repeatability is usually better than Method 2. Method 2 (cIEF) scans point by point through a single detection window: after focusing, chemical or pressure mobilization pushes the gradient past the window, and mobilization introduces additional peak broadening.

In laboratories without an icIEF instrument, Method 2 remains viable, but the method must additionally confirm the effect of mobilization conditions (mobilization pressure or chemical mobilizer concentration) on the pI reading and include that effect in the method uncertainty.

4 Data analysis and integration

4.1 Four records for the pI regression

The reliability of the pI is determined by the regression, so the quality of the regression must be recorded—not just a single result. The method should always record four items:

The residual pattern is the most often overlooked and the most diagnostic of the four. R² is insensitive to systematic curvature—five points bent into an arc can still give R² of 0.999. The sign pattern of the residuals can reveal curvature; whether it stems from gradient non-linearity or from the markers themselves has to be judged with more markers or independent data.

4.2 The normalization denominator: which peaks count as protein peaks

Chapter 3129 calculates percentages by peak-area normalization: the area of each charge variant as a percentage of the sum of all protein peak areas.

Everything depends on how a "protein peak" is defined. Four kinds of material produce peaks in the electropherogram, but they are different in nature:

• Charge variants of the sample: included in the denominator.
• Isoelectric-point markers: not included. Markers are added internal standards; their positions are used for the regression and their areas have nothing to do with sample content.
• System peaks: generated by the ampholytes, methylcellulose or capillary coating; not included. The requirement that the blank be free of interfering peaks in the integration region exists precisely to exclude them.
• Baseline artefacts: caused by bubbles, particles or detector noise, appearing as very narrow spikes or inverted peaks; not included.

The consequence of a contaminated denominator is one-directional: count one extra peak and every percentage is proportionally understated, systematically diluting main-peak purity. The bias may escape indicators that monitor only the main-peak pI and percentage; the chapter's requirements on interfering peaks in the blank and on electropherogram similarity exist to guard against it procedurally.

4.3 Where acidic, main and basic peaks divide

The boundaries of the three regions must be defined in advance by the method, not decided automatically by the integration software from peak shape. The usual practice uses the pI of the main peak as the divide: variants with a lower pI are acidic, those with a higher pI are basic.

This definition has an implicit premise: that the main peak itself is the "unmodified" species. If most of the sample is already modified (for example a highly deamidated sample), the main peak may no longer be the species at the theoretical pI. The meaning of "acidic/basic" should then be reconsidered, and the report should give the pI and percentage of each peak rather than a single "acidic percentage".

The dividing points should also be fixed to two decimal places and written into the method. Moving a dividing point by 0.05 pH can change the acidic percentage by several percentage points, a change that has nothing to do with the sample.

4.4 Judging focusing completeness: several indicators together

The most important point in this section: focusing time cannot be copied; it must be confirmed for each sample. A range as wide as 4.5–15 min in the chapter shows that the end point has to be set for the sample and method.

Indicators to observe together during method development:

Current levelling off. As the gradient forms and the ampholytes focus, conductivity in the capillary falls and so does the current. When the current approaches a stable value (change between consecutive readings below a preset threshold), focusing is close to complete; stopping while the current is still clearly falling may mean under-focusing.

Main-peak width, apparent pI and resolution of key peaks. Record how the main-peak FWHM, the main-peak apparent pI and the resolution of key neighbouring peaks change with focusing time. The end point should be chosen where all three are stable. If the width rises again, investigate pI drift, current and sample matrix rather than attributing it straight away to one mechanism.

No single indicator is enough: current alone misses cases where width or pI are still changing; peak width alone misses cases where current is still high and the matrix has not equilibrated.

4.5 What the report must contain

An icIEF report suitable for a regulatory filing should contain, besides the result values:

1. the marker list with the nominal pI and lot of each marker;
2. the regression equation and the position of the target pI relative to the marker range (interpolated / extrapolated);
3. the boundaries of the integration region (defined by marker peak positions);
4. the pI dividing acidic/main/basic peaks and the basis for setting it;
5. main-peak pI, main-peak percentage and their SD / RSD for the three system suitability injections;
6. the blank electropherogram (showing a clean denominator);
7. the response-factor assumption and a statement of its applicability.

Of the seven, items 2, 6 and 7 are most often missing from current reports and have the greatest effect on the credibility of the conclusions.

5 Results and discussion

5.1 icIEF and CEX: one sample, two kinds of "acidic"

Consider a hypothetical example (constructed values, not measured): suppose a recombinant IgG1 sample gives 43.4% acidic peaks by icIEF and, by CEX, 27.0% acidic, 61.7% main and 11.3% basic.

The two acidic values differ by 16.4 percentage points. This does not necessarily mean either method is wrong; one possible explanation is that "acidic" means something different in each:

• In icIEF, "acidic" is a pI criterion: any species with a pI below the main peak is acidic, including deamidation, sialylation, loss of C-terminal lysine and every other modification that lowers the pI.
• In CEX, "acidic" is a retention criterion: anything eluting before the main peak is acidic, depending on whether the modification produces enough retention difference under the specific gradient.

Some modifications may change the pI noticeably but barely change CEX retention (for example, those with little change in net charge that mainly redistribute surface charge); others may separate well on CEX but differ by less than the icIEF method's actual resolving power. The two sets do not contain one another; they only partly overlap.

Two rules of interpretation follow:

Never cross-calibrate the values; use them only to corroborate trends. If a lot shows higher icIEF acidic peaks but unchanged CEX acidic peaks, do not conclude straight away that "the methods disagree"; first consider whether the modification is of the "changes pI but not retention" type. A rise in both methods strengthens the judgement about a trend, but confirming the identity or total amount of the variants still needs other evidence (for example fraction collection and mass spectrometry).

Report the two separately in a filing; do not merge them. Writing "the two methods agree" hides information; the accurate statement gives both results and explains that the methods define "acidic" differently. If measured data are cited, give the sample source, the conditions of both methods, the raw electropherograms and chromatograms, and the integration rules.

5.2 Method transfer: combine numeric criteria, similarity to the reference electropherogram and predefined transfer criteria

Chapter 3129 states explicitly that when icIEF instruments differ in brand, or capillaries differ in brand or specification, the electropherogram of the system suitability standard may differ slightly in peak shape from the reference electropherogram.

This settles a dispute that recurs in method transfer: if the receiving laboratory's electropherogram does not look exactly like the sending laboratory's, can the method be released?

The answer: slight differences in peak shape are not in themselves grounds for rejection, but nor can the electropherogram be ignored in favour of numbers alone. The chapter's system suitability requirements cover both numeric criteria and similarity to the reference electropherogram, and allow slight peak-shape differences caused by different instruments or capillaries. A transfer should be judged on all of the following together:

• system suitability values: whether the main-peak pI, main-peak percentage and their SD / RSD fall within the specified ranges;
• similarity to the reference electropherogram: whether the number, relative positions and distribution of peaks match, with differences limited to slight changes in peak shape;
• predefined transfer criteria: whether the comparison indicators and acceptance limits set in advance in the transfer protocol are met.

This avoids rejecting a method because of unavoidable peak-shape differences between brands, without reducing method transfer to a pure comparison of numbers or of pictures.

5.3 Three traps

Trap 1: pI extrapolation. When the target pI lies outside the marker range, the software still reports a pI, without warning. How to spot it: check the target pI against the marker list. What to do: switch to a marker set with suitable coverage, or label the value as extrapolated in the report and state the extrapolation distance.

Trap 2: attributing broader peaks to a single cause. Under-focusing, thermal effects and matrix (salt) interference can all appear as "broader peaks, poor reproducibility", yet call for different actions. How to spot it: look together at the current–time curve, whether the main-peak apparent pI is still drifting, and the resolution of key peaks. A current that is still clearly falling suggests under-focusing; if the current is stable while width and pI keep changing, investigate heat load, matrix and gradient stability. Base the remedy on evidence—a misdiagnosis makes things worse with every adjustment.

Trap 3: a contaminated denominator. If system peaks in the blank fall within the integration region and are not excluded, the denominator is inflated and every percentage is understated. How to spot it: review the blank in every sequence, not just once during method development. What to do: adjust the integration boundaries or change the ampholyte lot.

5.4 Limitations

• Resolution is limited by gradient linearity. At extreme pI (<3 or >10) the gradient is steep and resolution low; variants in that region can only be characterized qualitatively.
• Hydrophobic and membrane proteins recover poorly. Adsorption to the capillary wall during focusing lowers recovery, seen as a low total peak area while the relative peak proportions may remain correct.
• It cannot identify the type of modification. icIEF groups species only by pI, and several modifications may overlap at the same pI. Identifying them requires peptide mapping, glycan analysis or CEX fraction collection.
• The response-factor assumption (see §2.4) may fail for samples containing fragments.

6 Pharmacopoeia comparison

* Check chapter numbers and current status of the USP and EP chapters against the latest official editions before citing.

7 Conclusions

1. The pI is a regression result, not a direct reading. Its reliability is set by the marker coverage and the residual pattern of the regression, not by R² alone. A target pI within the marker range is interpolated; outside it is extrapolated, and the two should not carry equal weight in the same report.
2. Focusing completeness is a state, not a time. Set the end point during method development by observing current levelling off, main-peak width, stability of the apparent pI and resolution of key peaks together. The 4.5–15 min in the chapter is a range, not a recommended value.
3. Percentages depend on what is counted. The make-up of the denominator is set by the integration boundaries and the cleanliness of the blank; a contaminated denominator causes a one-directional bias that may escape indicators monitoring only the main peak, and has to be prevented procedurally.
4. "Acidic" in icIEF and in CEX are not the same set. The two methods cannot cross-calibrate values, only corroborate trends; filings should report them separately and explain the difference in definition.
5. Judge method transfer on combined evidence. Combine system suitability values, similarity to the reference electropherogram and predefined transfer criteria; the chapter allows slight peak-shape differences between instruments or capillaries, but that does not mean the electropherogram can be ignored.
6. Relative peak areas assume equal molar response at 280 nm—approximately true for charge variants of intact IgG; samples containing fragments need separate confirmation, stated as a limitation. Glycoform differences mainly affect the conversion between molar and mass proportions.

References