Trastuzumab and T-DXd samples were provided by the Institut de Cancérologie Strasbourg Europe (Strasbourg, France).

For middle-up level experiments, IdeS digestion was performed by incubating one unit of FabRICATOR enzyme (Genovis, Lund, Sweden) per microgram of mAb or ADC for 60 min at 37 °C.

Detailed protocols for peptide mapping and rpLC-MS are available in SI (Supplementary Text 1 and Supplementary Text 2, respectively).

An Acquity UPLC H-class system (Waters, Wilmslow, UK) composed of a quaternary solvent manager, a sample manager set at 10 °C, a column oven and a TUV detector operating at 280 nm and 214 nm was coupled to an Orbitrap Exactive Plus EMR mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) equipped with an Ion Max source with a heated electrospray ionization HESI probe. 20 and 10 $\mu$g were injected for intact and middle-up level analyses, respectively. The SEC column used was an ACQUITY Premier Protein SEC 250 Å, 1.7 $\mu$m, 4.6 $\times$ 150 mm (Waters). The separation was carried out in isocratic mode with a 50 mM AcONH${}_4$ mobile phase at pH 6.9. The flowrate was set to 250 $\mu$L/min and 150 $\mu$L/min for intact and middle-up analyses, respectively. The ESI parameters were set as follows: sheath gas flowrate to 30 a.u., auxiliary gas flowrate to 10 a.u., ion transfer capillary temperature to 250 °C, vaporizer temperature to 150 °C, and capillary voltage 4.0 kV. The S-lens RF level was set to 200%. In-source collision induced dissociation (CID) and higher-energy collisional dissociation (HCD) energies were set to different values along the elution for intact-level experiments, in order to achieve an optimal desolvation for each species: 90/100 eV (CID/HCD) for dimers, 100/10 for monomers, and 80/10 for fragments 100 kDa. For middle-up level analyses, CID/HCD energies were lowered to 50/10 eV. The trapping gas pressure was 7 a.u. (corresponding to an ultra-high vacuum of 1e-9 mbar). The voltages on the injection, inter, and bent flatapoles were tuned to 8, 7, and 6 V, respectively. Acquisitions were performed in the m/z range 1000–12,000 with a 3 s scan time and a resolution of 35,000 (17,500 for dimers) at 200 m/z. The automatic gain control (AGC) target was set to 1e6, and the maximum injection time was 300 ms. Data were interpreted using MassLynx v4.1 (Waters, Milford, MA, USA) and BioPharma Finder v3.2 (Thermo Fisher Scientific, San Jose, CA, USA). The deconvolution mass tolerance was set to 10 ppm.

The online 2D HICxSEC-nMS setup has been extensively described elsewhere [28]. The HIC column was a MAbPac HIC-10, 1000 Å, 5 $\mu$m, 4.6 $\times$ 100 mm (Thermo Fisher Scientific). The SEC column used in the second dimension was an AdvanceBio SEC 300 Å, 2.7 $\mu$m, 4.6 $\times$ 50 mm (Agilent Technologies, Santa Clara, CA, USA). 200 and 150 $\mu$g were injected for intact and middle-up level analyses, respectively. For the first dimension (HIC), the mobile phase A was composed of 2.5 M of AcONH${}_4$ and 0.1 M phosphate buffer (Na${}_2$HPO${}_4$) at pH 7.0 (adjusted with phosphoric acid), whereas the mobile phase B was composed of 0.1 M phosphate buffer (Na${}_2$HPO${}_4$ and NaH${}_2$PO${}_4$) with pH 7.0 (adjusted with a NaOH solution). The following gradient was employed in HIC: 0 to 90% B in 36 min, 90 to 100% B in 21 min, then maintained at 100% B for 8 min before re-equilibration for 25 min, leading to a total analysis time of 90 min. The HIC experiment was conducted at a flowrate of 100 $\mu$L/min. Column temperature, wavelength, and data acquisition rate were set to 30 °C, 280 nm, and 10 Hz, respectively. For the second dimension (SEC), the separation was carried out in isocratic mode with 100 mM AcONH${}_4$at pH 6.9, using a flowrate of 700 $\mu$L/min. Column temperature, wavelength, and data acquisition rate were set at 25 °C, 210/280 nm, and 40 Hz, respectively. The analysis time of the second-dimension run corresponds to the sampling time of the first-dimension separation (1.5 min). In order to limit salt contamination of the ESI source, a fraction of 0.45 min is sent to MS thanks to a switching valve. A flow splitter was employed to reduce the flowrate to 100 $\mu$L/min prior to MS analysis.

The Synapt G2 HDMS mass spectrometer (Waters, Manchester, UK) was operated in sensitive mode and positive polarity with a capillary voltage of 3.0 kV. The cone voltage and backing pressure were set to 120 V and 6 mbar, respectively, to preserve noncovalent interactions. Source and desolvation temperature were held to 100 and 450 °C, respectively. Desolvation and cone gas flows were 750 and 60 L/h, respectively. Acquisitions were performed in the m/z range of 1000–10,000 with a 1.5 s scan time. Data were analyzed with MassLynx v4.1.

CEX-nMS analyses use the same LC-MS system as SEC-nMS experiments (Acquity H-Class coupled to the Orbitrap Exactive Plus EMR). The CEX column was a YMC-BioPro 5 $\mu$m, 4.6 $\times$ 100 mm (YMC, Dinslaken, Germany). 63 and 31 $\mu$g were injected for intact and digested samples, respectively. The mobile phase A was composed of 20 mM AcONH${}_4$ at pH 5.6, while the mobile phase B contained 140 mM AcONH${}_4$ and 10 mM NH${}_4$HCO${}_3$ at pH 7.4. For intact samples, the following gradient at a flowrate of 250 $\mu$L/min was used: 40 to 50% B in 2 min, 50 to 70% B in 11.5 min, increase to 100% B in 0.1 min and maintained for 5.4 min, followed by a re-equilibration step of 3 min. For middle-up level analyses, the %B was increased from 10 to 100% in 16 min, and kept at 100% for 4 min, at a flowrate 0.25 $\mu$L/min. MS parameters were identical to SEC-nMS experiments, except for CID/HCD values which were set to 120/10 eV for intact-level analyses. Data were interpreted using BioPharmaFinder v3.2. The deconvolution mass tolerance was set to 10 ppm.

Intact mass analysis in denaturing conditions—Intact-level analysis consists of measuring the mass of the full mAb/ADC, generally by classical denaturing reversed phase LC (rpLC) using C8 or C4 columns, coupled to denaturing MS. As rpLC-MS is implemented in most labs, it can be routinely performed to provide accurate intact mass measurements. rpLC-MS is particularly well adapted for mAbs and highly homogeneous ADC formats, but it might be trickier for heterogeneous ADC [31], or even ADCs partially or completely constituted of noncovalent assemblies, such as BV [32] or T-DXd. In absence of interchain disulfide bonds, the light chain bearing one drug (24,472.8 $\pm$ 0.1 Da) and the heavy chain containing three drugs (53,696.4 $\pm$ 0.2 Da) are observed (Fig. 1A). For each subunit, a minor peak associated to a fragmentation within the payload (–475 Da) was also detected. In addition, a minor species with a mass of 24,473.6 $\pm$ 0.4 Da was observed for the light chain, which may correspond to a deamidation (Fig. 1A). For the heavy chain, an oxidized form was also detected (53,714.4 $\pm$ 0.5 Da) as a shoulder. However, no species corresponding to free light or heavy chains was detected with rpLC-MS, suggesting an avDAR of 8.0. In addition, rpLC-MS can also give a direct snapshot of the glycosylation profile, and revealed four glycoforms for T-DXd, namely G0F-N/G0F, G0F/G0F (major one), G0F/G1F, and G1F/G1F.

Middle-up level subunit analysis in denaturing conditions—To precise the localization of drug conjugation and because mass measurements performed by rpLC-MS are more accurate for lower masses, downsizing of $>$150 kDa intact ADC can be achieved using a variety of enzymes [33, 34, 35, 36] that specifically cleave the mAb format into smaller subunits. After IdeS enzymatic digestion, which cleaves mAbs below the hinge region, three major species corresponding to the light chain with one drug (24,472.9 $\pm$ 0.1 Da), Fd subunit linked to three drugs (28,481.2 $\pm$ 0.2 Da), and unconjugated Fc/2 region (25,232.7 $\pm$ 0.1 Da for the G0F/G0F glycoform) were observed (Fig. 1B). No signals associated to the light chain or Fd without any bound DXd could be detected from extracted ion chromatograms (avDAR = 8.0), in agreement with intact rpLC-MS results. Middle-up analysis thus allowed confirming the localization of all drugs on the upper region of the mAb, being consistent with the position of the interchain disulfide bridges of the naked mAb. One minor species identified as the Fc/2 subunit bearing 1 oxidation was also observed.

Bottom-up level peptide mapping analysis—Finally, bottom-up strategies are conventionally used in biopharmaceutical environment to obtain sequence information and localize modification sites [37]. Peptide mapping performed on T-DXd after digestion with two different enzymes (trypsin and pepsin) resulted first in a sequence coverage of 96% for the heavy chain and 100% for the light chain, confirming the primary amino acid sequence. Several post-translational modifications (PTMs) were also detected in both chains (Fig. 2, Table 1). More precisely, the light chain contains two modifications: one methionine oxidation (M4, noted ox), and one deamidation (on residues Q27, N30, Q37, and/or Q38—in particular, N30 is known to be responsible for charge heterogeneity in the parent mAb trastuzumab [38, 39]). The heavy chain comprises five modifications: the formation of pyroglutamate from the N-terminal Glu (E1, called pyroGlu), two methionine oxidations (M83 and M255), an aspartate isomerization (D283, referred to as isoAsp), and the presence of a C-terminal lysine (K450). Of note, no peptide corresponding to the succinimide ring opening (+18 Da on the payload) [40] was detected on the reference compound. Finally, conjugation was confirmed at positions C223, C229, and C232 on the heavy chain, and C214 on the light chain, in agreement with a Cys-conjugation strategy following reduction of interchain disulfide bonds (Fig. 2). Peptides comprising the conjugated Cys residues could be identified with both the intact (1034 Da) and fragmented (–475 Da) payloads, in line with rpLC-MS data.

SEC-nMS for size variants analysis—Aggregates (high molecular weight species, HMWS) and fragments (low molecular weight species, LMWS) are critical quality attributes that have to be characterized for ADCs, conventionally by SEC-UV [41]. Data interpretation is thus only based on peak retention times. Recent development of SEC-nMS has paved the way for more accurate, less time-consuming all-in-one HMWS/LMWS assignment along with drug substance characterization [22, 42].

Intact glycosylated T-DXd was thus first analyzed using SEC-nMS. The main peak corresponds to monomeric T-DXd (peak 2, $>$95% based on the SEC-UV chromatogram) with its glycoforms (Fig. 3A). A mass of 156,339 $\pm$ 1 Da was measured for the main glycoform G0F/G0F of the monomer (Fig. 3B, Table 2), in good agreement with the mAb bearing 8 conjugated drugs (expected mass of 156,339 Da for G0F/G0F). Overall, the straightforward spectrum of T-DXd resulting from homogeneous Cys-conjugation affords easy DAR determination, as opposed to heterogeneous second-generation ADCs such as T-DM1 or BV which present highly complex spectra (Supplementary Fig. 1). In addition, the use of a novel bioinert SEC column (see Materials and Methods) that minimizes secondary interactions between the analyte and the stationary phase proved to be particularly interesting here, as it has been shown to yield more reliable HMWS/LMWS quantification [43]. Low amounts of HMWS were detected at lower retention times (~4 min) and attributed to dimers bearing a total of 16 drugs from the mass measurement (peak 1) (Fig. 3A, Table 2), while no significant amounts of LMWS (1%) were detected at greater elution volumes.

As mass accuracies are superior for lower size analytes, SEC-nMS experiments were also performed after enzymatic IdeS digestion. Additionally, this approach enables a more precise overview of the localization of the drugs on the different parts of the mAb (Fab and Fc). Hence, two major species attributed to Fab and Fc subunits are partially separated on the SEC column (Fig. 3C). The Fc subdomain (peak 3) presents a baseline-resolved glycoforms profile without any drug attached (50,463.9 $\pm$ 0.4 Da for the major glycoform G0F/G0F, Fig. 3D), in agreement with all previous LC-MS data. As expected, each Fab fragment contains 4 payloads (peak 2, 52,954.4 $\pm$ 0.5 Da, Fig. 3E). Note that the F(ab’)${}_2$ subdomain, eluted in peak 1, only accounts for 4% of the UV signal. As all interchain disulfide bonds have been reduced during the conjugation process, the F(ab’)${}_2$ subdomain is supposed to be held only by noncovalent interactions and appears to partially dissociate at the hinge region, leading to the release of the Fab part.

HICxSEC-nMS for drug conjugation assessment—As HIC is the gold standard for Cys-ADC analysis to assess DLD, we next moved to HIC-nMS analysis of T-DXd using our previously developed bidimensional HICxSEC-nMS methodology [28].

For the naked parent trastuzumab, a unique population corresponding to the unmodified monomer (148,385 $\pm$ 5 Da) is eluted at ~22.4 min (Supplementary Fig. 2A). T-DXd possesses 8 drugs and is eluted significantly later, at approximately 36 min. Two main populations are partially separated at the intact level for T-DXd (Fig. 4A). The peak at ~36.2 min corresponds to the monomer bearing 8 drugs (156,500 $\pm$ 1 Da), in agreement with the major DAR8 species detected using SEC-nMS. The more hydrophilic variant H1 at ~35.4 min could not be clearly mass-identified due to coelution with the main form. It is likely that other variants are comprised within the main broad peak (FWHM = 2 min versus 0.91 min for T-DXd and trastuzumab, respectively), and so performing HIC at the middle-up level may help to decipher the HIC profile of intact T-DXd.

After IdeS digestion, the HIC profile of T-DXd is more populated compared to the intact level, as several variants are uncovered for both subunits. The Fc region elutes first at ~17.9 min (same time as the naked mAb, Supplementary Fig. 2B), as all hydrophobic drugs retained by the HIC stationary phase are localized on Fab subunits of T-DXd. Two species are separated within the Fc region: the major unmodified species (50,466 $\pm$ 2 Da), and a hydrophilic variant (H1) with one oxidation (50,481 $\pm$ 3 Da, Fig. 4B), consistent with rpLC-MS results already reported. The Fab subunit profile consists of a main unmodified species bearing 4 intact drugs (52,955 $\pm$ 1 Da), and one hydrophilic variant H’1 containing an oxidation (52,974 $\pm$ 2 Da). Neither species bearing lower numbers of drugs nor free unconjugated Fab forms were detected.

These results clearly show that the online coupling of HIC separation to nMS remains challenging. For Cys-ADC, it is highly desirable that the separation profile achieved from a HIC-nMS method is comparable to that from more conventional HIC-UV methods, so that valuable information from MS analysis could be directly applied to assist peak assignment and identity elucidation. While our 2D HICxSEC-nMS strategy successfully enabled online coupling of HIC to nMS in an indirect manner but with optimal chromatographic performances, the MS sampling resolution from the HIC separation remains limited by the run time of the second dimension (1.5 min), which happens to be problematic to differentiate species with very close retention times. Additionally, because samples are diluted along elution in the HIC and SEC columns, the intensity of the MS signal may be too low for an accurate determination of variants. Nonetheless, we were able to clearly assign intense hydrophilic oxidized forms, which are particularly amenable to HIC separation as seen with unidimensional HIC-nMS [26].

CEX-nMS for charge variants—We next aimed at analyzing charge variants of T-DXd to provide a more comprehensive analysis of the ADC, with an accurate identification of PTMs. The mAbs generally exhibit a high level of heterogeneity, due to their PTMs and their susceptibility to chemical and physical degradations. Among the chemical degradations, oxidation, deamidation, isomerization, glycation, succinimide ring opening etc., have to be monitored and quantified in the drug substance since those modifications are often related to immunogenic effects, thus impairing patient safety [44].

The first experiments performed on the parental mAb trastuzumab demonstrate the presence of three distinct species (Supplementary Fig. 3A): (i) a major unmodified species, (ii) one acidic variant corresponding to a deamidation, and (iii) a basic variant identified as an isoAsp, in agreement with several already published studies [39, 45, 46, 47]. As T-DXd is derived from a neutral hydrophobic drug (DXd), CEX charge profiles of trastuzumab and T-DXd should be similar under identical experimental conditions. Interestingly, T-DXd presents a different profile characterized by a broad major peak at ~17.9 min (FWHM = 0.45 versus 0.16 min for T-DXd and trastuzumab, respectively) and a poor resolution of separation with substantial fronting (Fig. 5A). This result stems from secondary interactions between the ADC and the stationary phase mainly driven by the hydrophobic interactions with the hydrophobic drugs of T-DXd, thus being detrimental to the quality of the CEX separation [29, 30, 48]. Despite this lack of chromatographic resolution, nMS sensitivity and accuracy enabled the assignment of different species: a major form (M) whose glycoform and drug load profiles are consistent with the one observed in SEC-nMS for T-DXd (see inset in Fig. 5A), two acidic variants comprising one (A1) or two (A2) oxidations, and a basic variant (B1) which may be linked to an isoAsp (Table 3A), all bearing 8 drugs. It is likely that other PTMs are present, but could not be identified due to poorly resolved chromatographic separation. Again, no species with lower DAR values were detected in intact CEX-nMS analysis.

Overall, both the inherent charge heterogeneity and drug hydrophobicity of the ADC make it challenging to interpret the CEX profile of intact T-DXd, highlighting some limitations of CEX analysis for hydrophobic ADCs. In addition, unambiguous isoAsp (+0 Da) or deamidation (+1 Da) identification is often not achieved at the intact level as mass shifts fall within the margin of measurement error. As a result, PTM assignment strongly relies on associated retention shift (isoAsp = basic variant, deamidation = acidic variant). Thus, middle-up level analyses are generally more adapted for a more precise characterization of charge variants [49, 50].

Analyzing IdeS-digested subunits with CEX-nMS affords clear benefits over intact-level experiments, providing localization of charge size variants along with more accurate mass measurements to ease PTM assignment. For trastuzumab, the middle-up level CEX profile accurately mirrors PTMs observed at the intact level with a chromatographic resolution comparable to that of intact trastuzumab (Supplementary Fig. 3B). The deamidations and isoAsp previously identified at the intact level occur on the F(ab’)${}_2$ region, as reported in another study [51]. Advantages of performing CEX-nMS at middle-up level are obvious for T-DXd as distinct chromatographic peaks are now observed. The peak width of the Fc subunit peak becomes thinner (FWHM = 0.12 min, equal to trastuzumab) compared to the intact profile, as the contribution of drug hydrophobicity is now avoided (no drug on the Fc subunit). On the other hand, the Fab region still displays a broad peak with significant fronting (Fig. 5B), in part explained by the increasing ionic strength over elution time. Indeed, low ionic strength mobile phases focus chromatographic peaks, whereas higher ionic strength leads to broader peaks [52]. Five different basic variants are observed for the Fc subunit: B1 may be related to an isoAsp, B2 and B5 contain one and two oxidations, respectively, while B3 and B4 correspond to the presence of one or two C-terminal Lys (+128 Da) (Fig. 5B, Table 3B). The Fab subunit has two acidic variants, associated to an oxidation (A’1) and two deamidations (A’2). One minor basic species resulting from a N-terminal pyroglutamate (B’1) is also observed. All Fab variants bear 4 drugs in addition to the PTMs. These experiments particularly highlight that the ADC (here T-DXd) is more prone to oxidation than its naked counterpart (here trastuzumab). More importantly, all PTMs identified with peptide mapping were also detected with middle-up CEX-nMS analysis, confirming the suitability of CEX to identify PTMs in a rapid manner without extensive prior sample preparation. Hence, the CEX-nMS method represents a significant advantage over bottom-up strategies which require considerable sample treatment that may generate artifactual PTMs.

Forced degradation studies are essential for the development of therapeutic proteins to evaluate their stability in different conditions (storage, time, temperature, etc.) [53]. They can also be of interest in analytical method development and validation. In forced degradation conditions, minor species can be either artificially generated or enriched, constituting an interestingly more complex analytical matrix. For therapeutic mAbs, forced degradation studies highlight hot spots or points of fragility of proteins (deamidations, oxidations, aggregation, fragments, etc.) while simultaneously ensuring that the developed analytical methods are able to detect them. We thus thermally stressed T-DXd (15 days at 50 °C) and evaluated our different LC-nMS workflows for the detection of new low amount proteoforms.

Peptide mapping—We first performed classical bottom-up peptide mapping analysis as reference method to precisely identify and quantify PTMs generated in forced degradation conditions. Upon thermal stress, an additional deamidation site is observed within T-DXd, at position N328 on the heavy chain (Fig. 2, Table 1), along with a peptide from the hinge region with succinimide ring opening (222-251 SCDKTHTCPPCPAPELLGGPSVFLFPPKPK). In addition, higher amounts of deamidation on the light chain (+6%), isoAsp (D283 on the heavy chain, +17%) and pyroGlu (E1 on the heavy chain, +3%) were detected (Table 1).

SEC-nMS experiments—SEC-nMS was used first for the identification and quantification of degradation products. For forced degradation studies, SEC-nMS experiments are only performed on intact T-DXd, as information on thermally-generated HMWS and LMWS would be lost upon digestion. Interestingly, a slight shift in retention time is observed for the monomer (peak 2, 62%), which might suggest that thermal stress induces a small conformational change of the ADC. The mass measured for the main monomeric species is strictly identical to the reference sample (156,339 $\pm$ 2 Da), with the same glycoforms profile. However, the presence of a minor form with two oxidations (156,376 $\pm$ 1 Da) is now detected along with the unmodified monomer (Table 2). More importantly, no signals corresponding to either unconjugated mAb or DAR 8 are observed, indicating that no deconjugation occurs upon forced degradation. Upon thermal stress, the amount of HMWS increases to 10%, and additional LMWS are detected on the right of the main peak (Fig. 3A). The hinge region is a known mAb degradation hotspot, and peak 3 (21%) corresponds to a series of Fc-Fab fragments comprising 6 drugs, with a continuum of cleavage sites in the upper hinge region D224/K/T/H/T/C229, as previously reported for another thermally-stressed IgG1-based ADC [54] (Table 2). The peak 4 (7%) contains dimeric and free light chain. By comparing T-DXd to its non-functionalized mAb trastuzumab in similar forced degradation conditions (thermal stress), we determined that T-DXd is more prone to degradation into LMWS than the naked mAb (~28% of LMWS compared to 5% for trastuzumab) and that T-DXd is more prone to aggregation (~5% HMWS) than trastuzumab (HMWS 1%). This result is in good agreement with several publications confirming that ADCs are more likely to generate degradation products than their parent non-functionalized mAb [54, 55, 56].

HICxSEC-nMS experiments—Significant retention time shifts in HIC chromatograms after thermal stress are used as indications to tackle drug deconjugation issues. Forced degradation of T-DXd does not impact the retention time at the intact level, and mass measurements are still in line with the mAb bearing 8 drugs (156,504 $\pm$ 13 Da), confirming that thermal stress does not induce drug deconjugation (Fig. 4A). However, the major chromatographic peak broadens (+0.7 min at FWHM), and the populations corresponding to the main monomer and H1 are partially overlapped, which suggest that multiple PTMs contribute to peak widening. For more in-depth investigation, we performed HIC-nMS at the middle-up level: forced degradation leads to an early retention shift for both the major Fc peak and the hydrophilic variant H1, without any mass modification (50,466 $\pm$ 2 Da for the main peak), which could correspond to an isoAsp and/or deamidation, as both PTMs can lower HIC retention times [41] (Fig. 4B). The main Fab peak is now eluted later (from ~46.0 to 47.4 min), but its mass remains strictly identical to the one measured for the non-stressed sample. As peptide mapping did not identify any isoAsp in the Fab region, and in light of SEC data, we hypothesize that the retention time delay might rather be explained by a conformational change occurring upon thermal stress, which could favor access to additional hydrophobic regions.

CEX-nMS experiments—As abovementioned, CEX-nMS experiments are expected to be more informative than HIC for the separation and identification of PTMs. At the intact level, the major peak broadens with a late shift in retention time (peak now centered on ~18.1 min, FWHM = 0.90 min, Fig. 5A). No mass shifts are observed compared to the non-stressed sample, which suggests that the amount of basic variants increases, and/or that a conformational rearrangement happens upon heat exposure, giving access to additional surface charges. The amount of oxidized species does not increase, in line with peptide mapping results. As intact-level CEX analyses of T-DXd are once again not highly conclusive, we moved to middle-level analysis. After IdeS digestion, only few peaks in the CEX chromatographic profile of T-DXd are impacted by high temperatures (Fig. 5B). Among PTMs that are commonly encountered and susceptible to increase upon thermal stress, isoAsp leads to late retention shift, while deamidation induces an early shift in retention time [41]. The Fc region exhibits two newly-generated acidic variants, corresponding to one (A1) and two deamidations (A2, one on each Fc/2) (Table 3B). Based on peptide mapping performed on the stressed sample, the deamidation site is localized at N328. Of note, additional early-eluting species (not shown on the chromatogram) corresponding to free and dimeric light chain were also observed after thermal stress, in agreement with fragments identified with SEC-nMS.