A peptide's arrival time in a capillary is the sum of two transport terms with different owners. Electrophoretic mobility belongs to the analyte — net charge over hydrodynamic drag — and is the only term that separates anything. Electroosmotic flow belongs to the wall: a bulk displacement of the whole liquid column that carries charged and neutral zones alike. The detector reports the sum, which is why a change in surface chemistry looks exactly like a change in analyte charge.
Charge-to-size scaling is fitted, not derived
The Stokes picture — charge over 6πηr — is the right intuition and the wrong arithmetic, because you measure mass or residue count, not hydrodynamic radius. Grossman, Colburn and Lauer determined the size and charge contributions to mobility independently, by experiment, in a single solvent system, then combined them into one semiempirical relationship, fixing its constants on 40 peptides spanning 3 to 39 amino acids and charges from 0.33 to 14.0. Uncharged residues did contribute, but only as a minor perturbation on the effects of size and charge [1].
Janini and co-workers tested competing model forms against an independent set — 58 peptides, 2 to 39 amino acids, charge 0.65 to 7.82, measured at 22 °C in a 10% linear polyacrylamide-coated column with 50 mM phosphate at pH 2.5 — and report that the Offord model, correlating mobility with the charge-to-size parameter q/M2/3, offers the best fit to their reliable experimental data [2]. Inside that model, doubling net charge doubles mobility while doubling mass returns roughly 63% of it: charge is the lever, size is the tax.
Both treatments assume something close to a random coil. Sitaram, Keah and Hearn examined 33 synthetic peptides and noted significant divergences from linearity in plots of mobility against charge and against the charge-to-size parameter for several, possibly, they suggest, because particular amino acid arrangements have a proclivity to assume preferred α-helical or β-sheet features rather than random coil structures under their conditions [3]. A peak away from its predicted position is therefore not evidence of a different species: check whether the sequence would be expected to fold before calling it an impurity.
Electroosmotic flow, and why pH controls it
Silanols on fused silica deprotonate as pH rises, leaving a negative surface, a compensating cationic diffuse layer and, under field, bulk flow toward the cathode. Zhou and Foley applied an extended site-dissociation–site-binding model and showed that the logarithm of the number of apparent silanol sites on a fused-silica surface has a linear relationship with buffer pH. Their equation reproduced two behaviours a monotonic picture does not predict: a maximum in electroosmotic mobility as pH is varied at constant ionic strength, and both an inversion and a maximum when divalent cation concentration is varied at constant pH. The same equation indicates that electroosmotic mobility increases with temperature [4].
That temperature term deserves calibration. Evenhuis and co-workers extrapolated curves of electroosmotic mobility against power per unit length at several temperatures to isolate the dependence free of Joule heating, and found electroosmotic mobility increasing at 2.22% per °C [5]. Migration-time drift is therefore a usable diagnostic: if the neutral marker walks at constant applied field, either the wall chemistry is changing or the capillary is running hotter than the cassette setpoint.
Coatings that suppress or reverse the flow
Because the flow is generated at the surface, the surface is the control point. Hardenborg and co-workers physically adsorbed a cationic polyamine with short aliphatic blocks of combined 2- and 3-carbon length onto the negatively charged silica by flushing, followed by a self-stabilisation step. Those coated capillaries generated an anodal electroosmotic flow that was independent of pH over the range investigated, pH 4 to 8, and gave basic-protein peak efficiencies of 265,000 to 584,000 plates [6]. Reversal is not merely suppression: it changes which end of the capillary the detector sits at.
The same control is reachable with buffer additives. Badal and co-workers thermally cured a poly(dimethylsiloxane) layer onto fused silica and measured, in 50 mM ionic strength phosphate with 0.01% Brij 35, an electroosmotic flow about 25% of that in a bare capillary, showing no pH dependence between pH 6.3 and 8.2. Adding 2.7 to 6.3 µM cetyltrimethylammonium bromide to that buffer caused flow reversal, and CTAB between 0.0014 and 1.37 mM set electroosmotic mobility anywhere from −1 × 10−4 to −5.0 × 10−4 cm2 V−1 s−1 [7]. Those concentrations belong to the PDMS-coated surface with Brij 35 present and do not transfer to bare silica. Bilayer coatings work on the same electrostatic principle — a Polybrene–dextran sulfate bilayer held the surface for peptide deamidation work at pH 6.0 [8] — while a neutral polymer layer such as the 10% linear polyacrylamide column Janini used takes a different route again, building the coating into the capillary rather than replenishing it from the buffer [2].
Adsorption, and what a wash protocol does not undo
Peptides carrying exposed basic residues bind deprotonated silanols, with a characteristic failure mode: tailing, low recovery, and migration that shifts across a sequence as sites fill. Verzola, Gelfi and Righetti quantified it by flushing fluorescently labelled myoglobin into a capillary equilibrated in Tris-acetate at pH 5.0 until the potential adsorbing sites saturate, then driving SDS micelles in electrophoretically to desorb the bound material and reading the eluted peak fluorometrically. Screening amines as wall competitors, they found mono- and diamines rather poor quenchers, with 50% inhibition values of the order of 50–100 mM and 90% values reaching as high as 560 mM, while oligoamines — spermine and tetraethylenepentamine in particular — were most effective, with 50% molarities in the sub-millimolar range and 90% values of the order of 1 mM [9].
Routine practice fares worse. Washing with 1 M NaOH, with 1 M HCl, or with both, left behind traces of protein still bound to the wall, whereas the SDS micelle electrophoretic desorption seemed to be 100% effective [9]. If your conditioning sequence is base, acid, buffer and the first injection of each day is reliably the odd one, that is a plausible reason. Note the scope: those are myoglobin measurements in Tris-acetate at pH 5.0, and the quencher molarities belong to that system, not to a peptide run at pH 2.5.
The alternative to competing with the wall is to switch the wall off. Righetti, Bossi and Gelfi ran peptides and proteins in acidic isoelectric buffers in uncoated capillaries: at the low pH adopted, and with dynamic coating by cellulose derivatives, silanol ionisation is essentially suppressed. Because quasi-stationary isoelectric buffers have low conductivity, gradients up to 800 V/cm could be applied, giving fast peptide analysis with high resolving power and minimal diffusional peak spreading [10]. What you give up is real: EOF is no longer a transport term, and pH is no longer a selectivity variable, because everything separates on charge at one fixed pH.
Joule heating sets the bore
Applied volts dissipate as heat in the electrolyte, and that heat leaves only through the wall. The internal consequence is a radial temperature profile, hence a radial mobility profile, hence dispersion no longitudinal optimisation removes. What is easy to get wrong is where the thermal resistance sits. Evenhuis and co-workers derived the full profile and measured it: at a power per unit length of 1.00 W m−1, the radial temperature difference across the electrolyte was 0.14 °C, across the fused-silica wall 0.17 °C, across the polyimide coating 0.13 °C, and across the stationary air layer outside it 2.33 °C. Under typical conditions with active air cooling and a 74.0 µm internal diameter capillary, the mean electrolyte temperature was found to be up to 20 °C higher than the external temperature [5].
Two things follow. The bulk offset between setpoint and actual electrolyte temperature is dominated by heat transport outside the capillary, so cooling geometry matters as much as bore; and the internal gradient — the part that broadens bands — is small at 1 W m−1 and grows with power. The operating question is what power per unit length you can afford. Cianciulli and Wätzig measured external capillary temperature by infrared thermometry across a range of field strengths, buffers and capillary internal diameters, obtaining a slope of 6.3 °C m W−1 at the optimal 8 mm thermometer-to-capillary distance, against 7 to 10 °C m W−1 from indirect methods based on conductance, electroosmotic velocity or current. Their estimate: Joule heating can often be tolerated up to 4 W m−1, but sensitive analytes can already be affected by more than 1 W m−1 [11].
Diameter is the lever because, at fixed field strength and buffer conductivity, power per unit length scales with the cross-sectional area of the lumen. Going from 75 to 50 µm internal diameter cuts the heat load to about 44% at the same field, which is why fast, high-field CE runs narrow. The cost is optical: absorbance detection across the lumen loses path length in proportion, so narrow bore buys field strength with concentration sensitivity.
Zone electrophoresis, micellar electrokinetic chromatography, isoelectric focusing
Capillary zone electrophoresis is free solution: one homogeneous background electrolyte, separation strictly on effective mobility, neutral species all travelling at the EOF velocity and unresolved from one another. Micellar electrokinetic chromatography breaks that degeneracy by adding surfactant above its critical micelle concentration, so a charged micellar pseudostationary phase migrates at its own velocity and analytes partition between it and the aqueous phase, adding hydrophobicity as a second selectivity axis.
Whether it helps depends on the peptide. Popa, Mant and Hodges built a set of 18-residue amphipathic α-helical monomeric peptides in which a single site at the centre of the hydrophobic face was substituted by each of 19 L- or D-amino acids, so that each pair shares mass-to-charge ratio, sequence and intrinsic hydrophobicity and differs only in the stereochemistry of one residue. Comparing CZE in an uncoated capillary, MEKC with CHAPS, and open-tubular capillary electrochromatography on a C8-coated capillary with 25% trifluoroethanol or 25% ethanol, they found the OT-CEC methods most effective overall at separating the most peptide pairs, particularly those with hydrophilic side chains, while the MEKC approach proved most effective for pairs containing hydrophobic or aromatic side chains [12].
Capillary isoelectric focusing abandons steady migration entirely: a carrier ampholyte gradient sets a pH ramp along the capillary, each species focuses where local pH equals its pI, and the focused zones must then be mobilised past the detector. Dai and co-workers checked that premise in an online cIEF–MS configuration using pressure-assisted chemical mobilisation, verifying it by the linear correlation between pI values and migration time for a mixture of pI markers, and reporting beneficial effects of glycerol, which reduces non-cIEF electrophoretic mobility and limits band broadening [13].
These modes differ in what they can see. Schairer and co-workers compared cation-exchange chromatography, cIEF and CZE, each with MS detection, across ten monoclonal antibodies with isoelectric points from 7.3 to 8.7, without adjusting method parameters for each analyte. The lysine variant was resolved with high resolving power by all the separation methods tested; the monoglycosylation, by contrast, was found in the peak of the main variant with CEX and with cIEF, while CZE separated it from the main peak [14]. A mobility-based method and a pI-based method are not interchangeable even when both are nominally charge-variant assays.
Getting the effluent into a mass spectrometer
The interface is where CE's advantage is most easily discarded. A coaxial sheath-liquid sprayer closes the electrical circuit and stabilises the spray, but it dilutes a nanolitre-scale zone into a much larger liquid stream. Höcker, Montealegre and Neusüß compared three interfaces on one platform: a standard triple-tube sheath liquid, a nanoflow sheath liquid with the separation capillary inserted into a glass emitter, and a sheathless porous etched tip. Relative to the triple-tube interface, average peak areas with the nanoflow sheath liquid interface were higher by a factor of 13 for peptides in positive mode, 38 for six organic acids in negative mode, and 114 and 36 for the light and heavy chain of a monoclonal antibody respectively. Both nanoflow interfaces showed similar performance, and the nanoflow sheath liquid setup gave intra-day RSDs of 6.6 to 12.0% [15].
The decision is flow regime versus flexibility, not sheath versus sheathless: sheath liquid interfaces offer full flexibility in method development [15], while a sheathless porous tip removes the dilution term and leaves the sprayed composition to whatever the separation demands. Either way, MS detection restricts the background electrolyte to volatile constituents, which is why the deamidation work below runs in 150 mM ammonium formate at pH 6.0 rather than phosphate [8].
What CE resolves that reversed-phase does not
Reversed-phase separates on hydrophobicity, so modifications that change charge without much changing hydrophobicity are the class where CE is structurally better placed. Faserl and co-workers coupled a low-flow CE–electrospray interface to Orbitrap instruments and achieved excellent resolution of asparagine-, aspartic acid- and isoaspartic acid-containing peptides in a synthetic peptide mixture; the migration order then enabled clear assignment of in vitro deamidation and isomerisation sites in a 48-protein standard mixture of intermediate complexity. The same approach addressed citrullination, a modification that changes the positively charged guanidinium group of arginine to the uncharged ureido group of citrulline, and separated isobaric mono-phosphorylated peptides from a kinase activity study, quantifying the individual positional isomers [16].
The positional-isomer problem scales to hard cases. Dominguez-Vega and co-workers took a 4.5-kDa peptide containing five closely positioned potential deamidation sites, exposed it to acidic conditions and/or elevated temperatures for 1 to 14 h, and analysed the stressed material by CE–MS/MS with a Polybrene–dextran sulfate bilayer coating. Separation of the deamidated and deacetylated species, including several positional isomers, was greatly enhanced by adding up to 40 vol% acetonitrile–isopropanol to the background electrolyte, allowing reliable determination of the number of deamidations and deacetylations per degradation product. Where isomeric species still comigrated, comparing theoretical against measured isotope distributions for specific y ions yielded their identity and relative abundance [8] — recovered by isotope deconvolution, not by the separation.
Stereochemistry is the harder case, because a D/L pair is not a charge variant at all. In this comparison free-solution CZE had the weakest handle on such a pair, which is why the micellar and open-tubular modes carried those separations [12].
For complex digests the honest claim is complementarity, not replacement. Ibrahim and co-workers characterised 100 ng of yeast mitochondrial extract by sheathless CE with a high-sensitivity porous tip coupled to tandem MS, and concomitantly by nanoflow LC on a similar MS system, with matched 60-minute separations. Single injections identified 271 proteins from 976 unique peptides by nanoLC–MS/MS and 300 proteins from 1,765 unique peptides by CESI–MS/MS, which the authors read as significant specificity and complementarity depending on the physicochemical separation employed; CESI–MS/MS identified larger peptides and eased detection of those with extreme pI without impairing spectral quality [17].
The loading ceiling, and how it is worked around
A 50 µm internal diameter capillary 60 cm long holds 1.1 µL in total [19]. Injecting a few percent of that is a nanolitre-scale load; low loading capacity and a narrow separation window are the stated reason CZE has not displaced LC for large-scale peptide work. Chen, Shen and Sun attack it: their automated CZE–MS system approached at least half a microlitre of loading capacity, routinely used over 12% of the sample available in the vial, and generated a 140-minute separation window with a peak capacity of about 380 [18].
The other route is on-line preconcentration. Zhang and co-workers synthesised a sulfonate-silica hybrid strong cation-exchange monolith at the proximal end of the capillary, loaded sample in acidic buffer, eluted with basic buffer, and let the resulting dynamic pH junction refocus an otherwise ruinously wide elution plug. Loading 21 µL of 1 × 10−7 M angiotensin II gave an enrichment factor of 3000 over standard electrokinetic injection while retaining 44,000 theoretical plates; the monolith's loading capacity was about 15 pmol by frontal analysis. Applied to 5.5 µL of a 10−3 mg/mL E. coli digest, the platform identified 109 proteins and 271 peptides in a 20-minute separation at a median 25,000 theoretical plates [19]. The enrichment is real; so is the plate count you pay for it.