Reversed-phase chromatography sorts peptides by how much non-polar surface they can press against a bonded alkyl ligand. Ion exchange sorts them by how many charges they present to a fixed counter-charge at the pH you chose. Those are different physical quantities measured on the same molecule, which is why an ion-exchange method can split a pair that co-elutes in reversed phase, and why it can just as easily merge a pair reversed phase resolved cleanly. Everything else follows from taking that literally: the separation coordinate is charge, you set the analyte's charge with mobile-phase pH, and any retention you cannot account for from charge is the column doing something you did not ask for.
What the column is counting
The clearest statement of the mechanism comes from a sequence-specific retention model built on roughly 30,000 tryptic peptides from a Saccharomyces cerevisiae whole-cell lysate, run on a Polysulfoethyl A column in offline two-dimensional SCX–RPLC [1]. In reversed phase and HILIC, retention is driven by hydrophobic or hydrophilic contributions from all the individual residues; in strong cation exchange it depended mainly on peptide charge — the number of basic residues at acidic pH — and on size [1]. An additive model over all 20 residues combined with a peptide-length correction reached an R2 of 0.976, reported as significantly higher than the additive models for either HILIC or RPLC, and the finished SSRCalc SCX model reached about 0.99 [1]. A separation that a charge-and-length model predicts that well is one where you can reason about elution order from the sequence before you inject.
Position matters as well as count. In that dataset, basic residues (Arg, His, Lys) increased retention when they sat closer to the N-terminus, an effect the authors attribute to the spatial orientation of tryptic peptides against the negatively charged surface groups, with the positively charged N-terminus serving as the primary point of interaction [1]. Two peptides of identical composition and identical net charge can therefore separate on an SCX column because a lysine moved.
Charge and chain length move together, and the two were deliberately pulled apart on purpose-built polymers. Burke and colleagues synthesised three series spanning 5 to 50 residues — Ac-(Gly-Lys-Gly-Leu-Gly)n-amide, Ac-(Leu-Gly-Leu-Lys-Ala)n-amide and Ac-(Leu-Gly-Leu-Lys-Leu)n-amide, with n = 1, 2, 4, 6, 8 and 10 — and used their retention together with other peptides varying in charge density, net charge, chain length and hydrophobicity to establish a linearisation, and hence a prediction, of cation-exchange retention behaviour [2]. Adding residues adds charge and adds chain length at the same time, and the two contributions do not cancel.
Setting the analyte's charge: the pK arithmetic
You do not choose a peptide's charge; you choose a pH, and the ionisable groups decide. Getting the sign right means knowing where those groups titrate in a peptide. Thurlkill and co-workers measured them by potentiometric titration of alanine pentapeptides with appropriately blocked termini, in 0.1 M KCl at 25 °C: 3.67 ± 0.03 (α-carboxyl), 3.67 ± 0.04 (Asp), 4.25 ± 0.05 (Glu), 6.54 ± 0.04 (His), 8.00 ± 0.03 (α-amino), 8.55 ± 0.03 (Cys), 9.84 ± 0.11 (Tyr) and 10.40 ± 0.08 (Lys) [3]. They explicitly flag that several of these differ from the older values widely used in the literature — Asp, His, the α-amino group, Cys and Tyr all moved — so a pH picked off an inherited table can put a residue on the wrong side of its own pK [3]. Arginine is not in that measured set.
The exchanger chemistry is the mirror of the same arithmetic. A strong exchanger carries a group that stays ionised across essentially the whole usable pH range — sulfonate for cation exchange, quaternary ammonium for anion exchange — so the stationary-phase charge is a constant and pH acts only on the analyte. A weak exchanger carries a carboxylate or an amine that titrates within the working range, so a pH change moves both sides of the interaction at once. That is not automatically a disadvantage, but it does mean a retention shift you observe on changing pH cannot be attributed to the peptide without a control.
Whether a peptide binds is a question about the sign of its net charge at the running pH, and the isoelectric point is a weaker predictor of retention than its reputation suggests. In one-dimensional pH-gradient SCX of cell-lysate digests, the correlation coefficient between retention time and theoretical pI across identified peptides was higher than 0.4, which the authors took as showing that elution depended on charge state [9] — a demonstration that charge governs the axis, not a licence to predict a retention time from a pI. On a sulfonated ethylstyrene–divinylbenzene resin, five dipeptides with pI values from 4.0 to 9.7 showed a good relationship between retention factor and pI, but only above pI 5, and with Val-Glu (pI 4.0) excepted [4].
Salt gradients and pH gradients resolve different things
A salt gradient does not change the peptide. It raises the concentration of a competing counterion until the peptide loses the contest for the fixed charges, and that contest is stoichiometric. A model of cation-exchange/reversed-phase mixed-mode retention fitted a parameter γ defined as the number of salt molecules needed to desorb one bound biomolecule, and found experimentally that its value is close to the biomolecule's net charge; the model was validated against retention times for two peptides on a C8 column and four cation-exchange/RP columns of differing sulfonic-group content [5]. On the same materials the number of organic-modifier molecules needed to displace one adsorbed molecule was fitted as a separate parameter, which is the disciplined way of saying salt and organic solvent are not interchangeable knobs [5]. Seen from the other end, the retention factors of dipeptides on a resin of 0.81 wt% sulfo-group content fell as mobile-phase phosphate rose from 2.5 to 20 mmol/L [4].
A pH gradient does change the peptide. Titrating the mobile phase progressively neutralises the groups holding the analyte on the column, so species carrying the same charge at load pH but titrating differently can separate — something a salt gradient at fixed pH has no mechanism to do. The operational prize is that no rising salt gradient has to be washed back out before the mass spectrometer. An online pH-gradient SCX column plumbed straight into a nanoelectrospray source identified more than five thousand unique peptides per 90-minute run from 500 ng of digest, the column effluent going directly into the mass spectrometer [9].
Salt gradients keep what the pH approach gives up: a fixed mobile-phase pH means a fixed analyte charge state throughout the run, and the displacing counterion and its concentration, not the analyte's ionisation state, are what the gradient moves. Cation-exchange separation of six 26-residue amphipathic α-helical peptide standards of net charge +5 to +10 was run on sodium perchlorate gradients at pH 6.5 [8]. The two can also be run together: complete separation of the five dipeptides above was obtained with a 60-minute gradient that raised methanol from 10% to 50% while the 5 mmol/L phosphate buffer went from pH 4.8 to pH 8.9 [4].
Ion exchange is never purely ion exchange
Ion-exchange packings are hydrophobic and peptides notice. Across both silica- and polymer-based ion-exchange packings, non-specific hydrophobic interactions with those peptide polymer series were large enough that removing them, by adding acetonitrile to the mobile phase, was in some cases necessary just to elute the peptides from the matrix [2]. If a peptide will not come off a cation exchanger under a salt gradient that should be far past its charge, charge may not be the thing holding it. In the large SCX retention study, hydrophobic contributions that could otherwise produce a mixed-mode mechanism were described as largely suppressed at 20–30% acetonitrile in the eluent [1]; in the +5 to +10 standards work, acetonitrile from 0 to 40% was used to overcome hydrophobic interactions with the cation-exchange matrix, and resolution of the most highly charged members (+8, +9, +10) was particularly sensitive to how much was present [8].
Turned deliberately, the same effect becomes a separation mode. Cation exchange run at high organic content — optimum levels of 60–80% acetonitrile depending on the application — overlays a hydrophilicity-based separation on the charge-based one. On a 27-peptide mixture built as three groups of nine peptides at net charge +1, +2 and +3, in which adjacent members within a group generally differed by only a single carbon atom, and on peptide pairs of the same composition differing only in the position of one substitution, that mixed mode separated both mixtures better than RP-HPLC did, and the greater the charge on the peptides, the better the separation achieved [10]. Alpert pushed the electrostatics further: elute an ion-exchange column with a predominantly organic mobile phase and solutes are retained by hydrophilic interaction even when they carry the same charge as the stationary phase, so, in mixtures whose solutes differ greatly in charge, repulsion can be used to selectively antagonise the retention of the solutes that would normally be best retained. That permits isocratic resolution of mixtures that normally require gradients, affords separations of highly charged peptides that cannot readily be resolved by other means, and allows phosphopeptides to be isolated selectively from a tryptic digest [11].
The corollary for method development is blunt. Two peptides can share a pI and still separate on charge-based hardware: among four dipeptides at pI 6.4 (Gly-His, Val-His, Ile-His and Phe-His), retention on the sulfonated resin rose with log P [4]. Charge sets the order you expect; the packing's residual hydrophobicity sets some of the order you get.
Orthogonality, and what an orthogonality number licenses
Orthogonality has been quantified on peptides rather than assumed. Sixteen two-dimensional LC–ESI-MS pairings were compared using a retention dataset of roughly 30,000 tryptic peptides collected for each pairing. Separation orthogonality generally increased in the order RP < SCX < HILIC < SAX, with a PolySAX LP first dimension highest at 74% and, as the stated exception to that ordering, high-pH RP–low-pH RP second at 68% [12]. The finding that study insists on is the one easiest to skip: identification output is driven by both separation orthogonality and separation efficiency, which made high-pH reversed phase the best choice of first dimension — not the more orthogonal anion exchanger [12]. A high orthogonality number establishes that two dimensions disagree about elution order. It does not establish that the pair will find more.
Complementarity also shows up directly in identity lists. Overlap in unique peptides between one-dimensional pH-gradient SCX–MS/MS and reversed-phase MS/MS was 30% or less, and the SCX method identified a higher proportion of basic peptides, histidine-containing ones especially [9].
This is where ion exchange meets the reasoning behind confirming a purity number on an orthogonal separation. A reversed-phase area-percent figure is a statement about what one hydrophobicity axis could resolve. Re-running the sample on an ion exchanger is informative only for impurities whose charge differs from the main component at the ion-exchange pH. Something that alters neither net charge at that pH nor non-polar surface will co-elute in both, and two agreeing purity numbers will not make it visible.
Deamidation and the charge variants the technique was built for
Asparagine deamidation is the standard test case because it converts a neutral side-chain amide into a carboxyl group. The chemistry runs through a cyclic imide. In the hexapeptide L-Val-L-Tyr-L-Pro-L-Asn-Gly-L-Ala, deamidation to an aspartyl succinimide had a half-life of 1.4 days at 37 °C and pH 7.4; the succinimide then reacted further, by hydrolysis with a half-time of 2.3 h and by racemisation with a half-time of 19.5 h, so the net product was a mixture of L- and D-normal-aspartyl and isoaspartyl hexapeptides [13]. Sequence context dominates the rate: replacing the asparagine with an aspartic acid residue gave a 34-fold decrease in the rate of succinimide formation, and replacing the following glycine with a bulky leucine or proline gave a 33- to 50-fold decrease in the rate of degradation, with peptide cleavage products observed for those two analogues [13]. In Gly-Gln-Asn-X-His pentapeptides at 37 °C and I = 0.5, degradation followed apparent first-order kinetics and the rates were pH dependent, increasing with pH [14].
That last point is a method constraint before it is a result. Every buffer the sample sits in, and every pH a gradient runs to, is a condition under which the variant you are trying to quantify keeps forming.
What an ion exchanger sees depends on whether the new carboxyl is ionised at the running pH, and a careful retention study makes the distinction concrete. Comparing four separation modes for their power to tell citrullination and carbamylation apart from deamidation, strong cation exchange had the most differentiating power: elimination of basic residues on citrullination or carbamylation produced a retention decrease of about 58 mM KCl, while retention of the deamidated products decreased only slightly [15]. Keep the referents straight — the 58 mM figure belongs to the loss of a basic residue, not to deamidation, whose SCX shift in that work was small. The same study found similar retention shifts for all three modifications in high-pH RP (a decrease) and low-pH RP (an increase), which limits the applicability of that particular two-dimensional combination for distinguishing them [15]. An Asp carboxyl titrating near pK 3.67 [3] is substantially protonated at the acidic pH conventional SCX is run at; if the charge shift from deamidation is what you need to see, the separation has to be run somewhere that group is charged, or on an anion exchanger.
Note also what a charge-based separation cannot report. The normal-aspartyl and isoaspartyl products of one deamidation event carry the same ionisable groups. Resolving a deamidated peak from an intact one establishes that a deamidation-type charge change occurred; it does not establish which isomer formed, or at which residue. Site and isomer assignment need a different measurement.
Load, capacity and the displacement regime
Ion-exchange capacity is not a count of charged residues divided into a count of ligands. The steric mass-action formalism was introduced precisely to handle this: it explicitly accounts for the steric hindrance of salt counterions caused by adsorption, in multicomponent equilibria, and yields an analytical solution for the isotachic effluent profiles of displaced components and the salt gradients their binding induces, with the order of feed components in the displacement train established by a stability analysis and the predictions compared against cation-exchange displacement separations of proteins [7]. Peptides have been treated in the same framework: adsorption of tripeptides on a mixed-mode resin, measured by isocratic pulse-response experiments across combinations of salt concentration, temperature and pH, was described by an extension of the stoichiometric displacement model together with the steric mass-action model, yielding a distinct thermodynamic parameter set for each resin–peptide–temperature and resin–peptide–pH combination [6].
Two consequences for a working method. A capacity figure obtained with one analyte does not transfer to another of different charge, because both the binding stoichiometry [5] and the counterion shielding [7] differ between them. And past the linear region components compete for the same sites, which is the regime the displacement formalism describes [7]; peak shapes and fraction boundaries stop behaving as the analytical-scale run predicted, even though the underlying charge ranking has not changed. Charge-based reasoning about retention is a statement about the dilute limit.
Getting the salt back out
A salt-gradient fraction arrives in whatever buffer eluted it, at a conductivity chosen to break the peptide off a charged surface. That eluate is not directly compatible with electrospray, and it cannot be loaded onto a second ion exchanger without dilution or exchange. Desalting is part of the method, not an afterthought, and it is one more step at which material goes missing.
The standard small-scale answer is a micro-scale extraction device: ordinary pipette tips containing small disks of reversed-phase, cation-exchange or anion-exchange beads embedded in a Teflon mesh, which can be combined within one tip and supplemented with loose titania or zirconia beads for phosphopeptide enrichment; desalting and concentration take approximately 5 min, while fractionation or enrichment takes approximately 30 min [16]. The caveat is one of selectivity. A reversed-phase desalting bed retains by hydrophobicity, which is exactly the property the ion-exchange step was chosen to ignore, so the very hydrophilic, highly charged peptides an exchanger retains and resolves best are the ones most at risk of breaking through the trap unretained. Recovery through the desalting step is something to measure on the analyte in hand, not to inherit from a generic protocol. Eluting with a pH gradient avoids the step by never building a salt gradient that has to be removed, with the effluent going straight into the mass spectrometer [9].