Deep InquiryπŸ”¬ Analytical chemistry & quality

LC-MS is not a molecular balance: earning quantitative rigor across electrospray

β€œHow is the quantitative rigor of HPLC maintained when interfaced with the vacuum phase and severe matrix effects of Mass Spectrometry (LC-MS)?”

The short answer

LC-MS maintains quantitative rigor only when its entire measurement chain is calibrated and validated; adding a selective mass detector does not preserve concentration information automatically. Electrospray ionization (ESI) converts a changing liquid composition into gas-phase ions with efficiencies affected by coeluting material, mobile-phase additives, charge-state distribution, and source conditions.[1,2,3] Stable-isotope-labeled standards, matrix-effect experiments, appropriate calibration, and controlled acquisition can make that response quantitative, but none removes the need to demonstrate selectivity and recovery.[4,8] β€œMS identity confirmed” establishes a narrower claim than chemical purity, correct stereochemistry, or amount.[11,12,15] For abundant, adequately separated peptides, a calibrated UV assay can be the more robust quantifier while MS supplies complementary identity and impurity evidence.[13,16]

Where the concentration information becomes vulnerable

The PHYS202 notes correctly distinguish mass-to-charge ratio (m/z) from neutral mass. Their electric-field and trajectory picture explains how ions can be sorted or their motion measured, not how many molecules in the original solution become detected ions. The CHEM202L summary's statement that electrospray MS β€œconfirms identity” needs a similar qualification: it can strongly support an expected mass assignment, but one matching mass is not a complete structural proof. USP 〈736βŒͺ describes MS as an m/z-based technique capable of molecular-mass, composition, and structural information, with the evidence depending on the experiment.[15]

An LC peak reaches the electrospray source as a liquid stream at or near atmospheric pressure, not as a liquid suddenly placed intact inside the analyzer's high vacuum. The source and pressure-reduction interface remove solvent and transport a small selected ion population into the mass analyzer.[1] A vacuum requirement is therefore not the same as a quantitative transfer of every solute molecule. Losses, adduct formation, transmission, and detection all stand between sample amount and recorded intensity.

I find it useful to write an explicitly conceptual model: signal = amount Γ— recovery Γ— ionization efficiency Γ— transmission Γ— detection response. These factors need not be independent or constant, so this is an accounting aid rather than a universal physical law. A matrix-dependent change in ionization can bias amount estimates while mass calibration remains excellent. The detector may identify the ions it receives to a few parts per million and still undercount the original analyte badly.

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”     β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ SAMPLE MASS β”‚ ──→ β”‚ PREPARATION /  β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜     β”‚ LC SEPARATION  β”‚
                    β””β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                            β”œβ”€β”€β†’ UV CELL β†’ ABSORBANCE VS TIME
                            β”‚
                            β–Ό
                  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
                  β”‚ CHARGED DROPLETS    β”‚ ← SALTS / ACIDS / MATRIX
                  β”‚ EVAPORATION / FISSIONβ”‚
                  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                            β–Ό
                    GAS-PHASE IONS
                            β”‚ INTERFACE / TRANSMISSION
                            β–Ό
                  M/Z + ISOTOPE ENVELOPES
                            β”‚ CALIBRATION / DECONVOLUTION
                            β–Ό
                    IDENTITY OR AMOUNT?
                    DIFFERENT VALIDATION

Electrospray is a competitive physical process

The mechanism is not simply β€œvoltage gives each molecule a proton.” Reviews by Kebarle and Verkerk and by Konermann and colleagues distinguish droplet formation and evolution from final gas-phase-ion release, with multiple mechanistic models relevant to different analytes.[1,2] A useful sequence is:

  1. The electric field creates a spray of charged droplets from the liquid stream.
  2. Solvent evaporation increases charge density; unstable droplets can undergo Coulomb-driven fission.
  3. Small ions may be released through ion evaporation, while charged-residue and chain-ejection descriptions apply to other molecular regimes.
  4. Desolvated analyte ions, adducts, and background ions encounter interface and ion-optical selection before reaching the analyzer.

I would not assign every long therapeutic peptide to one exclusive ion-release model on the basis of its residue count. Conformation, solvent, charge, and surface behavior matter. The quantitative consequence is less controversial: the number and types of detected ions depend on the environment as well as analyte abundance.[1,2] A clean solvent standard is not automatically response-equivalent to a counterion-rich sample.

Ion suppression occurs when coeluting components reduce analyte response; enhancement is also possible. Salts, ion-pairing agents, proteins, metabolites, and other less volatile material can change droplet formation, evaporation, and the available gas-phase-ion population.[3] Competition for charge or droplet-surface access is a useful part of the picture, but β€œthe other molecules steal all the protons” is too narrow a universal explanation. Suppression can be compound-specific, as Bonfiglio and colleagues demonstrated with different model compounds and sample-preparation procedures.[5]

That means suppression cannot be inferred from the total-ion chromatogram alone. A strong signal at one retention time does not prove that another component was efficiently ionized, and an absent impurity signal may reflect poor response rather than absence. Conversely, a minor compound with favorable ionization may look disproportionately large. My practical rule is that MS peak percentages need analyte-specific response evidence before they become composition percentages.

TFA, formic acid, and the compromise called DFA

Trifluoroacetic acid (TFA) is useful in peptide/protein reversed-phase chromatography because its strong ion-pairing behavior can improve peak shape and suppress secondary interactions. It can also suppress electrospray response. Apffel and colleagues attributed suppression in their peptide-mapping system to combined ion-pairing and surface-tension effects, and demonstrated recovery of signal through a post-column modification.[6] Their result shows that source response and separation can be partly decoupled, not that one additive recipe universally fixes TFA.

Formic acid (FA) is generally more compatible with sensitive MS detection, but switching from TFA to FA can change retention and selectivity, sometimes degrading resolution. The best-looking mass spectrum may therefore come from a chromatographically worse separation. A UV purity result acquired with TFA and an MS identity result acquired with FA are not automatically measurements of the same resolved components. I would document that they are separate methods and establish which impurity each can actually see.

Difluoroacetic acid (DFA) offers another possibility rather than a guaranteed optimum. Nguyen and colleagues reported approximately threefold higher MS sensitivity than TFA and useful chromatographic performance in their antibody-drug-conjugate work.[7] The system involved antibody-derived species and carefully prepared MS-quality DFA; it is evidence to motivate a peptide-method screen, not a predicted threefold benefit for PP's products. I would compare critical-pair resolution, recovery, response, background, and robustness across additives rather than optimize target signal alone.

A dual-method strategy may be more defensible than forcing every objective into one run. One method can maximize related-substances resolution with UV detection, while an orthogonal MS-compatible method characterizes masses and trace impurities. Where serial UV-MS detection is used, delay volume, splitting, and acquisition alignment need to be understood. Agreement between channels is meaningful only after accounting for what each measures and when it measures it.

Charge states, isotope envelopes, and what deconvolution assumes

For an ion carrying z protons, m/z = (M + z Γ— 1.007276)/z, where M is neutral monoisotopic mass in daltons. One peptide can therefore produce several charge states, each with its own isotope distribution. Summing one arbitrary m/z trace may lose response if conditions redistribute the envelope. Adducts and overlapping ions complicate the apparently straightforward conversion back to mass.[1,9]

Here is my own hypothetical worked example, not a spectrum from a supplier. A neutral peptide of monoisotopic mass 4800.0000 Da would give protonated ions at approximately 1601.0073 (+3), 1201.0073 (+4), 961.0073 (+5), and 801.0073 (+6). Adjacent isotopic peaks dominated by substitution of one ΒΉΒ³C are separated by approximately 1.00335/z m/z: about 0.2508 at +4 and 0.2007 at +5. Correctly resolving that spacing can support charge assignment; seeing a peak around 1201 without it may leave more ambiguity.

Monoisotopic mass uses the lightest relevant isotopes, whereas average mass is the abundance-weighted mean over isotope compositions. The most intense isotopic peak is neither definition in general. Large peptides contain enough atoms that the all-light-isotope species may not be the strongest or even readily observable. Senko and colleagues showed how isotope-distribution analysis can recover monoisotopic assignments and warned that an isotope-index error can shift the inferred mass by roughly a dalton despite very precise measurement.[9]

I therefore want the theoretical formula, mass convention, isotope assignment, and tolerance stated on an identity report. A comparison of a measured monoisotopic value with a vendor's rounded average molecular weight is not a valid ppm check. Modifications, terminal chemistry, disulfides, conjugates, and salt conventions also need explicit treatment. TFA and water in a weighed powder are not necessarily retained as part of the neutral peptide mass inferred from its protonated gas-phase ions.

Deconvolution combines charge-state information to estimate a neutral-mass distribution. Marty and colleagues' UniDec work uses a Bayesian framework to separate mass and charge dimensions and illustrates the power of this approach.[10] The algorithm still depends on data quality, allowed charge/mass ranges, peak-shape assumptions, background handling, and assignment choices. A tidy deconvoluted peak is processed evidence, not proof that the raw spectrum contained only one chemical species.

I would retain both raw and processed spectra, inspect agreement across multiple charge states, and challenge software settings with known mixtures. A sodium adduct, overlapping isotope series, or excluded low-intensity charge state should not disappear silently from the interpretation. For quantitative work, the selected ions and charge states must be fixed or controlled in the validated method. β€œThe software found the expected mass” is not enough information to reproduce an identity decision.

How to measure matrix effects rather than discuss them abstractly

Post-column infusion is a diagnostic map. A constant analyte infusion is introduced after the LC column while a blank or representative matrix extract is chromatographed; depressions in the otherwise steady analyte signal reveal suppression zones.[3,5] It is particularly useful for discovering that a target elutes into a problematic region even when the matrix has no obvious target-like peak. It does not by itself quantify extraction recovery or prove acceptable performance for every concentration and sample type.

The Matuszewski approach separates effects using appropriately matched preparations; the original study introduces the framework, and Cortese and colleagues reproduce its equations in an accessible tutorial.[4,17] Define responses, with equal final analyte amounts and suitable volume correction, as:

  1. A: analyte in neat solution.
  2. B: blank matrix processed first, then fortified with analyte after extraction/preparation.
  3. C: blank matrix fortified before extraction/preparation and carried through the procedure.

Then matrix factor = B/A, extraction recovery = C/B, and process efficiency = C/A. Multiply by 100 to express percentages. Those ratios should use consistent integration, volumes, concentrations, and response definitions. Calling C/A β€œrecovery” without separating the matrix effect mixes loss in preparation with loss or enhancement in ionization.[17]

For my numerical example, let A = 100,000 area units, B = 60,000, and C = 48,000. The matrix factor is 0.60, showing 40% suppression; recovery is 80%; and process efficiency is 48%. A neat-standard comparison alone could misleadingly suggest only 48% extraction recovery. The arithmetic identifies different corrective actions: improve preparation recovery, remove or separate suppressors, or use a demonstrably compensating standard.

PP's lyophilized-peptide matrix is not plasma, and a genuine blank matrix may be hard to obtain. The same experimental logic can be adapted using representative counterions, solvents, excipients where applicable, and lot-specific samples, but a surrogate must be challenged for commutability. Standard addition, dilutional parallelism, spike recovery, and comparison with an orthogonal assay can expose residual bias when blank-matrix designs are imperfect. This is my proposed adaptation of the literature, not a claim that a synthetic salt mixture reproduces every supplier sample.[3,4]

One matrix lot and one concentration are not enough to establish generality. Counterion burden, contaminant composition, and analyte concentration can all vary, and coeluting related peptides may suppress each other differently. I would examine representative lots, concentration levels, and preparation conditions, including stability and adsorption losses. A low replicate coefficient of variation can coexist with a stable, substantial bias.

What an isotope-labeled internal standard fixesβ€”and what it cannot

A stable-isotope-labeled internal standard (SIL-IS) closely matched to the target can track preparation and ionization when it coelutes and behaves equivalently. Quantitation then uses an analyte/IS response ratio against calibrators, rather than assuming a constant absolute analyte signal. Stokvis and colleagues consider SIL standards a first choice but warn that deuterium-labeled compounds can show different retention or recovery and that apparently successful correction can conceal assay problems.[8] Carbon-13/nitrogen-15 labels are often attractive, but label position, purity, mass separation, and behavior still need qualification.

The addition point matters. An IS added after sample preparation cannot correct target lost to an earlier vessel, incomplete dissolution, or extraction step. A labeled fragment is not automatically a surrogate for an intact modified peptide's recovery, and a heavy peptide added after digestion cannot report digestion completeness. I would choose the standard to match the exact measurand and the earliest stage it is intended to control, then test rather than assume equivalence.

Consider my matrix example with equal amounts of analyte and matched IS. If both suffer the same 40% suppression, their ratio can remain unchanged even though both absolute signals fall substantially. If the analyte retains 60% response but the IS retains 75%, the ratio becomes 0.60/0.75 = 0.80, leaving a 20% bias. This calculation shows why an IS-normalized matrix factor and its variability matter more than the mere presence of a labeled peak.

An IS cannot create selectivity where an isomer produces the same monitored response, nor restore precision after signals approach the detection limit. Isotopic impurity, channel cross-talk, insufficient mass offset, detector saturation, and concentration-dependent response need checks. Calibration standards also need assigned content: a SIL assay calibrated against unqualified β€œ99% HPLC” powder can inherit that powder's missing-mass error. Internal standardization corrects specific shared behavior, not all analytical uncertainty.[8,16]

What β€œMS identity confirmed” does not establish

An intact-mass match is strong evidence for a compatible molecular composition when charge, isotope pattern, calibration, and modifications agree. It does not uniquely establish residue order, linkage topology, stereochemistry, or purity. Isobaric species may share nominal mass yet separate at high resolution; isomeric species can have exactly the same elemental formula and exact mass. Higher mass accuracy does not overcome identical mass by itself.

Leu/Ile substitutions and D/L epimers illustrate the structural problem. Ordinary intact mass cannot distinguish them; routine fragmentation may also leave ambiguity, although specialized methods can supply additional evidence. Jia and colleagues demonstrated site-specific D-amino-acid localization using chromatographic separation, fragmentation, and ion-mobility differences of epimeric fragments.[12] That is evidence that sophisticated MS workflows can address chirality indirectly, not that routine identity scans do so automatically.

Asp/isoAsp is another important blind spot. Backbone rearrangement from Asp to isoAsp preserves mass, while deamidation of Asn produces a mass change of about +0.984 Da but does not alone reveal which Asp/isoAsp or stereochemical products formed. Ni and colleagues used selective Asp-N proteolysis with electron-transfer dissociation to distinguish and locate isoAsp.[11] The lesson is to choose a chemical or fragmentation handle specific to the structural distinction, rather than demand more decimal places from intact mass.

Coeluting different-mass impurities are more accessible, but their quantitation still needs calibration. Kodidasu and colleagues demonstrated validated HRMS quantitation of coeluting glucagon deletion impurities, rather than simply normalizing their uncorrected intensities.[14] A useful CoA should therefore state whether MS provided intact-mass compatibility, sequence-supporting fragmentation, targeted impurity measurement, or a validated quantitative assay. Those are distinct levels of evidence.

When UV is the better quantifier

For an abundant target that is adequately resolved from its relevant impurities, LC-UV with a qualified reference standard can offer a simpler and more stable amount measurement. Absorbance avoids the gas-phase ionization and charge-distribution dependence that complicates ESI. It does not avoid sample loss, detector nonlinearity, coelution, or reference-standard errors. USP 〈1503βŒͺ describes reference-standard-based chromatographic assay as a common peptide approach while distinguishing it from peptide-content and related-substances measurements.[16]

At 214 or 220 nm, peptide bonds and side chains contribute differently, so UV area normalization across different peptides is not automatically mass quantitation. Kuipers and Gruppen's constituent-response measurements at 214 nm support this caution.[13] A target assay calibrated with the same target is a different proposition from assuming equal response for every impurity. UV may be best for the main component while MS, with appropriate impurity standards, is best for a low-level coeluting deletion.

My method-selection question is therefore β€œWhich detector provides the most defensible result for this measurand at this concentration?” rather than β€œWhich instrument is more advanced?” A well-separated UV assay can outperform a poorly controlled LC-MS assay for content. A selective, isotope-dilution LC-MS method can outperform UV for a low-level analyte in a difficult matrix. Neither conclusion generalizes without the validation data.

What this means for Peptide Partners

I would separate identity testing, related-substances testing, and content assay in our analytical plan and CoAs. For identity, the report should state calculated and observed mass, mass convention, charge states, isotope evidence, mass error/tolerance, relevant modifications, and the extent of MS/MS confirmation. It should not imply exclusion of D/L, Asp/isoAsp, or every isobaric impurity unless a method specifically addresses them. Raw spectra and deconvolution settings should remain available for review.

For quantitative LC-MS, I would require a documented calibration model and range, assigned reference content, IS identity/addition stage, selectivity, recovery, matrix-factor assessment, carryover, stability, dilution integrity, and precision/accuracy appropriate to the intended decision. The acquisition cycle must sample the chromatographic peaks adequately: adding more MS/MS transitions or high-resolution scans can reduce points across a narrow peak. These are proposed PP qualification elements, not an assertion that every listed test is already validated in-house. Larissa should be able to explain how each element constrains an actual failure mode.

The highest-value early comparison would use representative supplier lots measured by a qualified UV assay, an independent peptide-content approach, and targeted LC-MS impurity characterization. Disagreement should initiate investigation of denominators, standards, coelution, sample recovery, and suppressionβ€”not an automatic preference for MS. PP's RUO context makes accurate research concentrations and defensible identity claims important without making a clinical safety or dosing claim. The objective is reproducible chemistry, not a more impressive instrument name on the certificate.

Questions to bring to the lab

  1. Which current MS results are only intact-mass matches, and which genuinely quantify a target or specified impurity?
  2. Can we map suppression at the target retention time and measure lot-to-lot matrix-factor variation for our actual counterion-rich samples?
  3. Does changing TFA to FA improve sensitivity while hiding a critical impurity through worse chromatographic selectivity?
  4. Would a DFA screen improve the combined separation/response problem, and what evidence would justify adopting it?
  5. At what stage is the SIL standard added, and which losses or structural differences does it fail to track?
  6. Can we reproduce charge-state assignment and deconvolution from raw data, including minor envelopes and adducts?
  7. Which products need dedicated epimer or Asp/isoAsp testing because accurate intact mass leaves the critical identity question unanswered?

What I still don't know

I do not know the magnitude of suppression in PP's actual supplier matrices, the source conditions used by our third-party laboratory, or the content assignments of its calibrators. The literature provides mechanisms and experimental designs, not those lot-specific answers. My numerical examples are calculations illustrating measurement bias, not validation results.

I also cannot promise a universally best acid modifier or a SIL standard that compensates every modified peptide equally. Some identity problems will require custom standards, orthogonal separations, specialized fragmentation, or external expertise. Until those boundaries are mapped, I would describe each MS result at the level the evidence actually supports and reserve quantitative claims for a demonstrated quantitative method.

References

  1. Kebarle P, Verkerk UH. Electrospray: From ions in solution to ions in the gas phase, what we know now. Mass Spectrometry Reviews. 2009;28(6):898-917. doi:10.1002/mas.20247
  2. Konermann L, Ahadi E, Rodriguez AD et al. Unraveling the Mechanism of Electrospray Ionization. Analytical Chemistry. 2013;85(1):2-9. doi:10.1021/ac302789c
  3. Annesley TM. Ion Suppression in Mass Spectrometry. Clinical Chemistry. 2003;49(7):1041-1044. doi:10.1373/49.7.1041
  4. Matuszewski BK, Constanzer ML, Chavez-Eng CM. Strategies for the Assessment of Matrix Effect in Quantitative Bioanalytical Methods Based on HPLCβˆ’MS/MS. Analytical Chemistry. 2003;75(13):3019-3030. doi:10.1021/ac020361s
  5. Bonfiglio R, King RC, Olah TV et al. The effects of sample preparation methods on the variability of the electrospray ionization response for model drug compounds. Rapid Communications in Mass Spectrometry. 1999;13(12):1175-1185. doi:10.1002/(SICI)1097-0231(19990630)13:12<1175::AID-RCM639>3.0.CO;2-0
  6. Apffel A, Fischer S, Goldberg G et al. Enhanced sensitivity for peptide mapping with electrospray liquid chromatography-mass spectrometry in the presence of signal suppression due to trifluoroacetic acid-containing mobile phases. Journal of Chromatography A. 1995;712(1):177-190. doi:10.1016/0021-9673(95)00175-M
  7. Nguyen JM, Smith J, Rzewuski S et al. High sensitivity LC-MS profiling of antibody-drug conjugates with difluoroacetic acid ion pairing. mAbs. 2019;11(8):1358-1366. doi:10.1080/19420862.2019.1658492
  8. Stokvis E, Rosing H, Beijnen JH. Stable isotopically labeled internal standards in quantitative bioanalysis using liquid chromatography/mass spectrometry: necessity or not? Rapid Communications in Mass Spectrometry. 2005;19(3):401-407. doi:10.1002/rcm.1790
  9. Senko MW, Beu SC, et al. Determination of monoisotopic masses and ion populations for large biomolecules from resolved isotopic distributions. Journal of the American Society for Mass Spectrometry. 1995;6(4):229-233. doi:10.1016/1044-0305(95)00017-8
  10. Marty MT, Baldwin AJ, Marklund EG et al. Bayesian Deconvolution of Mass and Ion Mobility Spectra: From Binary Interactions to Polydisperse Ensembles. Analytical Chemistry. 2015;87(8):4370-4376. doi:10.1021/acs.analchem.5b00140
  11. Ni W, Dai S, Karger BL et al. Analysis of Isoaspartic Acid by Selective Proteolysis with Asp-N and Electron Transfer Dissociation Mass Spectrometry. Analytical Chemistry. 2010;82(17):7485-7491. doi:10.1021/ac101806e
  12. Jia C, Lietz CB, Yu Q et al. Site-Specific Characterization of d-Amino Acid Containing Peptide Epimers by Ion Mobility Spectrometry. Analytical Chemistry. 2014;86(6):2972-2981. doi:10.1021/ac4033824
  13. Kuipers BJH, Gruppen H. Prediction of Molar Extinction Coefficients of Proteins and Peptides Using UV Absorption of the Constituent Amino Acids at 214 nm To Enable Quantitative Reverse Phase High-Performance Liquid Chromatographyβˆ’Mass Spectrometry Analysis. Journal of Agricultural and Food Chemistry. 2007;55(14):5445-5451. doi:10.1021/jf070337l
  14. Kodidasu A, Acharyya K, Ramu VG. Absolute Quantitation of Coeluting Impurities in Peptide Drugs Using High Resolution Mass Spectrometry: Glucagon a Case Study in Pharmaceutical Development. Journal of the American Society for Mass Spectrometry. 2025;36(10):2072-2078. doi:10.1021/jasms.5c00105
  15. United States Pharmacopeia. 〈736βŒͺ Mass Spectrometry. USP–NF. 2017. Rockville, MD: United States Pharmacopeia. doi:10.31003/USPNF_M99525_02_01
  16. United States Pharmacopeia. 〈1503βŒͺ Quality Attributes of Synthetic Peptide Drug Substances. USP–NF. 2021. Rockville, MD: United States Pharmacopeia. doi:10.31003/USPNF_M12935_02_01
  17. Cortese M, Gigliobianco MR, Magnoni F et al. Compensate for or Minimize Matrix Effects? Strategies for Overcoming Matrix Effects in Liquid Chromatography-Mass Spectrometry Technique: A Tutorial Review. Molecules. 2020;25(13):3047. doi:10.3390/molecules25133047