|Home | About | Journals | Submit | Contact Us | Français|
The balance between chromatographic performance and mass spectrometric response has been evaluated using an automated series of experiments where separations are produced by the real-time automated blending of water with organic and acidic modifiers. In this work, the concentration effects of two acidic modifiers (formic acid and trifluoroacetic acid) were studied on the separation selectivity, ultraviolet, and mass spectrometry detector response, using a complex peptide mixture. Peptide retention selectivity differences were apparent between the two modifiers, and under the conditions studied, trifluoroacetic acid produced slightly narrower (more concentrated) peaks, but significantly higher electrospray mass spectrometry suppression. Trifluoroacetic acid suppression of electrospray signal and influence on peptide retention and selectivity was dominant when mixtures of the two modifiers were analyzed. Our experimental results indicate that in analyses where the analyzed components are roughly equimolar (e.g., a peptide map of a recombinant protein), the selectivity of peptide separations can be optimized by choice and concentration of acidic modifier, without compromising the ability to obtain effective sequence coverage of a protein. In some cases, these selectivity differences were explored further, and a rational basis for differentiating acidic modifier effects from the underlying peptide sequences is described.
Reversed-phase liquid chromatography (RPLC) has become a powerful approach for the identification, characterization, and purification of peptides and peptide mixtures. This is particularly true for the biopharmaceutical industry, where the number of peptide and protein therapeutics is increasing rapidly1 and peptide mapping has become a standard method to assess the identity, stability, and batch-to-batch consistency of protein therapeutics.2–7
The combination of reversed-phase peptide separations with electrospray ionization mass spectrometry (MS) has provided an even more powerful approach for quickly developing effective peptide mapping8,9 and purification protocols, troubleshooting those methods, and rapidly detecting and identifying differences between samples. This combination of ultraviolet and mass detection creates the additional need to balance the requirements for effective separations with efficient mass spectrometry detection.
Developing effective peptide separations requires the integration of such diverse parameters as sample complexity, sorbent characteristics, gradient length and shape, and LC eluent composition. Optimization of the LC eluent composition plays a particularly important role at the early stages of developing an effective peptide separation, as each sorbent will have its own selectivity and performance based on the gradient and type of acidic modifier used for the separation.10,11 In LC coupled with MS detection, the positive chromatographic performance of an acidic modifier must also be balanced with the potential for suppressing electrospray ionization. A study12 using model intact proteins has shown this interplay can be quite dynamic, as protein retention, recovery, and MS signal can be affected by the choice of reversed-phase column and mobile phase modifiers.
The most common RP methods used today for peptide separations involve isocratic or gradient elution of peptides, where the mobile phase consists of water and acetonitrile in the presence of an acidic modifier. The acidic modifier serves several functions. The low pH produces tryptic peptides with net positive charge (particularly useful for coupling with positive ion electrospray MS), suppresses silanol effects by protonation of exposed hydroxyl groups, and potentially generates ion-pairing effects by simultaneously interacting with analyte and sorbent ligands. It has been reported that the type and concentration of acidic modifier used in RPLC has significant effects on the selectivity of peptide separations.13–15 Hancock et al. and Guo et al. first described the effects of three different acidic modifiers (phosphoric acid, trifluoroacetic acid, and heptafluorobutyric acid) on the retention and selectivity of synthetic peptide standards. Two models have been proposed to explain the mechanism for ion-pair separations. In one model, the acidic modifier first forms an ion pair with the analyte in solution, and then is retained by the column via analyte molecules.16,17 In an alternative model, the acidic modifier is first retained on the column, and the analyte binds with adsorbed modifier via counter-ion exchange.18,19 Regardless of the mechanism, the ion-pairing ability of an acidic modifier depends on its ability to interact with both the sorbent and a basic site on a peptide.20,21 If the desired peptide retention and elution order is not achieved with a particular modifier, another acidic modifier or a mixture of modifiers may be applied to achieve the desired separation.
Although acidic modifiers play a significant role in the selectivity of peptide separations,22 most scientists do not routinely investigate these effects when faced with a challenging separation. The increased time and effort involved in producing a range of eluents with varied levels and types of acidic modifier inhibit exploration of this variable during methods development. To more readily automate the process of optimizing a reversed-phase peptide separation, we have adopted the automated eluent blending (AutoBlend) methodology first described by Warren et al. for the automated development of ion exchange separation methods.23 In their approach, real-time quaternary blending of the basic and acidic forms of a buffer, a neutral salt (e.g., NaCl), and water was applied for the automated evaluation of various ion exchange conditions. In this work, we have shown how the AutoBlend methodology can be expanded to efficiently automate RP methods development for peptide separations, and how the balance between chromatographic performance and MS response can be addressed with this approach.
MassPREP enolase digestion standard was obtained from Waters Corporation (Milford, MA). Trifluoroacetic acid (TFA) was obtained from Pierce (Rockford, IL). Formic acid (FA) was obtained from EM Science (Gibbstown, NJ).
Enolase digestion standard was dissolved in 0.1% FA or 0.1% TFA to a final concentration of 5 pmol/μL. Sixty-five microliters (320 pmol) was loaded on column for each analysis.
An Alliance Bioseparations System (Waters) was configured with a 2796 Bioseparations module, column heater/chiller, 2487 dual-wavelength UV/VIS detector containing an biocompatible analytical flow cell, and a Micromass LCT ESI-Tof MS (Waters). A post-column split (1:7 MS to UV) allowed simultaneous MS and UV detection. A single MassLynx 4.0 workstation was used for instrument control, data acquisition, and processing. LC/MS data were processed by the PeptideAuto processing module, operating within ProteinLynx Global Server 2.0 (Waters) to determine peptide map coverage.
RP separations were produced using a 2.1 × 150 mm BioSuite PA-A C18 column (Waters) at a flow rate of 0.5 mL/min. A wavelength of 215 nm was used for peptide UV detection. The automated blending methodology involves the real-time blending of multiple chromatographic inlet channels to produce a chromatographic eluent. In this work, four inlet solutions (1% TFA, 1% FA, acetonitrile, and water) were mixed to permit the generation of RP gradients with adjustable levels of two acidic modifiers. To generate more stable UV baselines, gradients of organic solvent (2 to 48% acetonitrile over 86 min) were coupled with a linearly decreasing acidic modifier gradient (initial concentration to 80% or initial concentration) over the same period. This is described in more detail later in the text.
Positive-ion mass spectra were acquired on a Micromass LCT ESI-Tof mass spectrometer equipped with an orthogonal electrospray ionization source operated in the positive-ion mode. The electrospray ionization (ESI) capillary was maintained at 3-kV potential and heated to 110°C. The sample cone voltage was set to 25V, while the desolvation temperature was maintained at 200°C. The system was optimized with a nitrogen cone (50 L/h) and desolvation (450 L/h) gas flow, a maximum flight time of 65 μsec, and 2700 V on the multichannel plate. The mass analyzer was calibrated (m/z 300–3000) with 2 mg/mL sodium cesium iodide (NaI/CsI) in 50% isopropanol. Data were acquired in the mass-to-charge range of 300–3000.
Combining RP separations with electrospray MS detection has proven to be a powerful approach for the identification and characterization of peptides and peptide mixtures. Conditions that influence the quality of a peptide separation (or resulting map) can be optimized during methods development, but this optimization must also consider potential effects on downstream MS detection. In this paper, we have applied an automated blending (AutoBlend) chromatographic methodology for development of LC/UV and LC/MS peptide maps of yeast enolase. In these experiments, the real-time quaternary blending of acetonitrile, water, and up to two acidic modifiers permitted automated parallel generation of LC/UV and LC/MS peptide maps using a range of acidic modifier conditions and combinations.
LC/UV chromatograms (Figure 11)) for an enolase tryptic digest were produced using a typical concentration (0.1%) of formic acid (FA) and trifluoroacetic acid (TFA). The resulting chromatograms are consistent with literature observations that the use of TFA as an acidic modifier produces a general increase in peak retention time and decrease in peak width, relative to separations employing FA.24 As expected, the overall UV response is not affected by the choice of acidic modifier, and roughly equivalent peak areas were obtained for selected peptides analyzed from the two analyses (not shown). Of greater interest are the significant selectivity differences observed between the two separations. In the presence of TFA, the stronger ion-pairing agent, sample components are not only retained longer, but several have changed their relative elution order with the surrounding components (see below). Directly identifying these selectivity differences is not possible using UV detection, unless individual peptide standards are employed, but can be routinely observed using mass spectral analysis. Such changes in selectivity can be critical in designing methods for purifying a peptide from closely related impurities, or for resolving a complicated region of a peptide map.
The acidic modifier also has a dramatic effect on electrospray time-of-flight (TOF)MS response (Figure 22).). Whereas UV response was largely unaffected by the change in modifiers, the ESI-MS signal in the presence of TFA is suppressed approximately ninefold relative to FA results. This suppression effect has been well established for the analysis of both proteins and peptides,20 but the relative level of suppression varies with the nature of the analytes and separation conditions employed. This reduced sensitivity is critically important where sample is limited and/or components of high dynamic range are present. Peptide maps generally produce digest fragments at equivalent concentrations, while minor protein variants (e.g., low stoichometry of modification) produce specific modified peptides with much lower abundance. Thus, a peptide map used for identification purposes can be relatively immune to suppression effects, while maps used to assess stability of a protein drug may be compromised by reduced sensitivity.
The ability to manipulate the selectivity and MS response of a peptide separation/analysis by adjusting levels of acidic modifiers is seldom applied, due to the labor involved in producing the aqueous and organic solvent mixtures needed to evaluate different acids at various levels. For this reason, we configured a quaternary AutoBlend system (Figure 33)) for real-time automated blending of simple eluents (water, acetonitrile, 1% TFA, 1% FA) to evaluate the effects of varying acidic modifier levels of an individual acid, and the combinations of the two acidic modifiers. RP separations of an enolase digest were generated using each modifier at initial concentrations ranging from 0.025% to 0.2%. The ability of the AutoBlend system to change modifier levels during the run was exploited to maintain level UV baselines by linearly reducing the level of a modifier to 80% of the starting concentration by the end of the gradient. In a second set of experiments to examine the behavior of acidic modifier mixtures, a constant modifier concentration was maintained (0.1%), but the relative composition of FA and TFA was varied between runs.
Figure 44 shows the effects of increasing FA concentration on resulting LC/MS total ion current (TIC) (top) and LC/UV chromatograms (bottom) of an enolase digest. In general, average peak widths remain unchanged as a function of FA concentration, and general peptide retention is increased as the level of FA is increased. Concentration-dependent retention effects could be peptide specific, as seen in the region between 42 and 45 min, where the retention of a lower-intensity component is significantly altered relative to two other peptides, whose retention varies minimally over the different run conditions. Similar changes in peptide selectivity can be observed in chromatogram regions located between 10–15 and 57–60 min.
Examination of peak intensity information from extracted ion chromatograms (see supplementary data) for the dominant charge state of six enolase peptides revealed that increasing the FA concentration to the maximum tested level (0.2%) reduced MS signal by 25% (range 20% to 29%) of the maximum signal at the lowest concentration tested (0.025%). The consistent chromatographic peak widths and small retention shifts for these peaks favors the interpretation that this intensity decrease is a function of increased electrospray suppression rather than a decrease in chromatographic performance.
Extracted ion chromatogram (XIC) for these same five peptides analyzed under TFA conditions revealed an average intensity decrease of 60% (range 43–75%) from lowest to highest modifier concentration, and overall 90% (range 80–98%) average decrease in intensity from an equivalent FA concentration. Average XIC peak width (half height) also decreased slightly for the five peptides, from 0.19 min (0.025% TFA) to 0.17 min (0.2% TFA), indicating that actual electrospray suppression effects of TFA are slightly greater than indicated by intensity information alone. Consistent with the greater ion pairing capacity of TFA, peptide maps (Figure 55)) show a larger overall retention time shift with increased modifier concentrations than with FA. As with FA, examples of selectivity and elution order differences can be observed as a function of TFA modifier concentration (e.g., between 5–10 and 17–20 min).
The selectivity effects of varying the type and concentration of an acidic modifier can be demonstrated by following the relative retention of specific enolase peptides using XIC under the various modifier conditions. Under low (0.025%), medium (0.1%), and high (0.2%) levels of formic acid (Figure 66),), three enolase tryptic peptides (T10, T12, and T19) demonstrate subtle changes in elution pattern with increasing FA concentrations. All three peptides demonstrate increasing retention as FA concentration increases, but the elution of the T10 peptide increases more than the other two peptides. Indeed, this peptide is the earliest eluter under low-FA conditions, but becomes the latest eluter under the highest concentrations tested.
The XIC spectra for these same peptides when FA is replaced with TFA (Figure 77)) show the chromatographic resolution of the three components, and the increased retention of the T10 peptide at the lowest TFA concentration. Raising the TFA concentration showed a significant retention time increase for the T19 and T10 peptides relative to the T12 peptide.
A more detailed examination of the sequences of these peptides reveals a likely explanation for these observations. The T12 (ANIDVK) peptide contains two basic sites (N-terminus and C-terminal lysine) that can pair with acids in the mobile phase. The T10 (GLVLHAVK) and T19 (HLADLSK) peptides both possess an internal histidine than can provide an additional ion-pairing interaction under acidic conditions. In the T19 peptide, the position of this histidine plays a significant role in determining the extent to which these sites can interact with ion-pairing agents. The likelihood of charging both the N-terminal histidine and the N-terminus is reduced due to charge repulsion effects, and interactions of the pairing agent with both charged sites will be reduced further by steric effects. In combination, these effects can explain why the ion-pairing effects of TFA are more pronounced with the T10 peptide. In samples of known composition (e.g., peptide map of a protein drug), this rationale should be generally applicable to identify areas of the separation for which additional selectivity may be obtained by manipulation of acidic-modifier concentration. In unknown samples, the altered selectivity effect may provide useful information (e.g., number of basic sites) about the structural composition of a peptide.
The ultimate quality of a peptide separation may be quantitatively judged by the completeness of identification of all the peptides in the mixture. In this study, mass spectral data from the FA and TFA experiments was processed by the PeptideAuto and BioLynx applications in MassLynx 4.0 to generate protein coverage plots for the various enolase peptide maps. Under low, medium, and high FA and TFA concentrations, the same set of 35 tryptic peptides were observed for sequence coverage of 83%. There were a small number of additional minor missed cleavage products detected in FA that were not detected in the TFA analyses. Thus, there is no absolute detriment to using TFA for peptide mapping studies, given sufficient sample quantities, but a practical limitation on the detection of lower-abundance components may be encountered.
In a final set of experiments, the quaternary Auto-Blend system was used to produce enolase LC/UV/MS maps using mixtures of FA and TFA. Maintaining the total modifier concentrations at 0.1%, a series of six enolase separations were generated (Figure 88).). This set of experiments shows that the MS-suppression and chromatographic effects of TFA dominate even when it is present at only 20% of the total modifier concentration in the sample. This behavior was specifically confirmed with XIC data for the three peptides profiled above (data not shown). Blending of modifiers in the concentration range shown here does not appear to provide significantly differing selectivity to TFA alone, but the ability to change from an FA separation to a TFA or hybrid FA/TFA system during the course of a separation run may prove practical for some specific situations.
The ability to resolve a complex region of an analytical RP peptide map, or a peptide from an impurity in a preparative RP purification, requires a method that combines the necessary amount of separation capacity with the optimal chromatographic selectivity. The application of smaller particles, longer columns, higher pressures, and longer run times to obtain additional theoretical plates complements the changes in selectivity achievable from alterations to eluent composition. Our results indicate that both the type and concentration of an acidic modifier can produce useful selectivity differences in reversed-phase peptide separations.
It was interesting to note that the order of magnitude ESI-TOF MS sensitivity difference between FA and TFA separations did not affect our ability to derive roughly equivalent sequence coverage from the resulting peptide maps. Thus, given sufficient material for analysis, there was no benefit from using either acidic modifier to detect equi-abundant peptides in a map, and the choice of modifier can be made according to its ability to resolve specific peptides within a mixture. The search for minor variants, low-occupancy modification sites, or low-level contaminating proteins could be significantly affected by sensitivity/dynamic range limitations imposed by the use of modifiers that strongly suppress MS signal.
The AutoBlend methodology gave us an automated approach to evaluate the effects of varying both the type and concentration of two common acidic modifiers, without the need to prepare a large number of chromatographic eluents. The extension of these studies to optimize other gradient parameters (e.g., limits, shape, length, slope, flow rate) is inherent in the automated approach, and the addition of simple column-selection valving can give us the ability to include sorbent selectivity parameters for a more complete methods development matrix.
The authors wish to thank Drs. John Gebler and Thomas Wheat for their review and many useful discussions relating to this manuscript.
Supplementary data have been provided by the authors to give more detailed information concerning the effects of acidic modifier choice and concentration on electrospray MS response for selected enolase peptides. A link to this material is provided at http://jbt.abrf.org/.