The research reported in this manuscript was supported by Merck, Inc., though a contract with The University of Kansas Center of Research to co-author Elizabeth Topp. Co-authors Josef Vlasak and Roxana Ionescu are employees of Merck Research Laboratories, Inc. With the exception of these ongoing employer/employee relationships, the authors do not anticipate any direct financial gain as the result of the publication of this manuscript.
The effects of secondary structure on asparagine (N) deamidation in a 22 amino acid sequence (369-GFYPSDIAVEWESNGQPENNYK-390) of the crystallizable (Fc) fragment of a human monoclonal antibody (Fc IgG1) were investigated using high-resolution ultra performance liquid chromatography with tandem mass spectrometry (UPLC/MS). Samples containing either the intact Fc IgG (∼50 kD) (“intact protein”), or corresponding synthetic peptides (“peptide”) were stored in Tris buffer at 37°C and pH 7.5 for up to forty days, then subjected to UPLC/MS analysis with high energy MS1 fragmentation. The peptide deamidated only at N382 to form the isoaspartate (isoD382) and aspartate (D382) products in the ratio of ∼4:1, with a half-life of ∼3.4 days. The succinimide intermediate (Su382) was also detected; deamidation was not observed for the other two sites (N387 and N388) in peptide samples. The intact protein showed a 30-fold slower overall deamidation half-life of ∼108 days to produce the isoD382 and D387 products, together with minor amounts of D382. Surprisingly, the D382 and isoD387 products were not detected in intact protein samples and, as in the peptide samples, deamidation was not detected at N388. The results indicate that higher order structure influences both the rate of N-deamidation and the product distribution.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Therapeutic antibodies are one of the fastest growing segments of the pharmaceutical industry.1, 2 To maintain potency and minimize immunogenicity, antibodies and other protein drugs must be protected from physical and chemical degradation during manufacturing and storage. One of the most common routes of chemical inactivation of proteins is deamidation at asparagine (N) residues, a hydrolytic reaction that can influence the activity of endogenous proteins as well as protein drug products.3–5 In neutral to basic solution, deamidation proceeds via the nucleophilic attack of the N + 1 nitrogen of the protein backbone on the carbonyl group of the N side chain, forming a cyclic imide (succinimide, Su) intermediate. Hydrolysis then occurs at either of the succinimide carbonyl groups to form the aspartate- (D) and isoaspartate- (isoD)containing products, which differ from the parent sequence both in charge (acidic vs. neutral) and mass (+1 amu). Protein primary sequence strongly influences deamidation rate. For example, in unstructured model peptides (37°C, pH 7.4), deamidation half-lives for the NG sequence are typically on the order of 1 day, while half-lives at NV and NI sequences of more than 200 days have been reported.6 Secondary structure also influences deamidation rate,4, 7–10 but its effects have not been as clearly elucidated.
Various analytical methods have been used to detect deamidation in proteins.11 Ion-exchange chromatography (IEC) exploits the difference in charge of the acidic degradation products and the parent molecule,12, 13 and is routinely used in the pharmaceutical and biotechnology industry. Tryptic digestion with reversed-phase LC/MS analysis has been used to identify sites of deamidation and to quantify deamidation rates.14, 15 More recently, electron capture dissociation (ECD) in Fourier transform mass spectrometry (FTMS) has been used to distinguish D- and isoD-containing forms in tryptic digests based on their signature fragmentation patterns.16, 17 The enzyme protein isoaspartyl methyl transferase (PIMT) repairs isoD residues in vivo, and has been used to quantify isoD levels in vitro; PIMT detection kits are available commercially (e.g., IsoQuant®, Promega, Madison, WI).
Cumulative evidence from a number of monoclonal antibodies suggests that these proteins typically contain deamidation-prone sites: (i) in the complementarity determining regions (CDR), depending on sequence, and (ii) in several “hot spots” in the constant regions (Fc). Recent studies of deamidation in the Fc-domain of a humanized monoclonal antibody employed tryptic digestion and liquid chromatography/electrospray ionization mass spectrometry (HPLC/+ESI/MS).18 The results showed that the reaction is sensitive to antibody structure, with tryptic fragment G369-K390, (GFYPSDIAVEWESNGQPENNYK) showing rapid and structure-dependent deamidation at N382 and N387. Interestingly, the authors did not observe several expected deamidation products at these sites, including the isoD product at N387 (i.e., isoD387), the Asp product at N382 (i.e., D382), and the succinimide intermediate at N387. The authors infer that the mechanisms of deamidation differ at the two sites, perhaps due to differences in secondary structure. This is an important observation and a potentially significant contribution to our understanding of the effects of secondary structure on chemical reactivity in proteins. However, it is also possible that the “missing” products were simply poorly resolved chromatographically and so were not detected. This is particularly problematic if the isoD and D variants coeluted, because the two products have identical mass and cannot be resolved by the use of base ion chromatograms, as used by the authors. As a result, the effect of secondary structure on deamidation in this commercially important protein remains an open question.
The work reported here is a detailed study of deamidation at N382 and N387 in the Fc portion of a humanized IgG1 antibody. We have used ultraperformance liquid chromatography with electrospray ionization mass spectrometry (UPLC/+ESI/MS) in an attempt to achieve chromatographic resolution of deamidation products while maintaining relatively short analysis times. In addition, synthetic peptides corresponding to the tryptic fragment G369-K390 and its deamidation products have been used as analytical standards and as controls for the effects of structure on deamidation. The results demonstrate that both the product profiles and reaction kinetics differ for the synthetic peptide and the intact protein, supporting the hypothesis that higher order structure plays a significant role in deamidation in this system.
UPLC/+ESI-MS analysis of synthetic peptide standards
Four synthetic peptides (NNN, ND387N, IsoD382NN, NisoD387N) were analyzed by UPLC/+ESI-MS to establish chromatographic elution times for use in identifying deamidation products. Approximate elution times were: IsoD382NN = 36.0 min; NNN = 37.0 min; NisoD387N = 37.0 min; ND387N = 38.5 min. Baseline resolution was achieved for the IsoD382NN and ND387N peaks, and for the NNN and ND387N peaks; resolution of the IsoD382NN and NNN peaks was partial. The NNN and NisoD387N peptides showed identical retention times and were not distinguishable chromatographically. This suggests that the NisoD387N deamidation product may be masked by the NNN parent peak, and that careful analysis of NNN mass envelopes is warranted to establish the presence or absence of this product.
Deamidation products in the NNN synthetic peptide
The EIC for a representative sample of the NNN synthetic peptide following 23 h of storage (37°C, pH 7.4) shows three well-resolved peaks eluting at ∼36.2 min (Peak 1), 37 min (Peak 2), and 39.3 min (Peak 3) [Fig. 1(A)]. Figure 1(B) shows the molecular ion mass spectra of the first three of these peaks, presented in the order of elution. The second peak shows an m/z of 1273 [Fig. 1(B, 2)] and is thereby identified as the intact parent peptide (i.e., G369-K390). The first peak [Fig. 1(B, 1)] and the third peak [Fig. 1(B, 3)] both shown +1 amu shifts in their molecular isotope envelopes, consistent with singly deamidated products. Because the retention time of the first peak is comparable to that of the IsoD382NN synthetic peptide, this peak is tentatively assigned to the isoD382 deamidation product, suggesting that the third peak is the corresponding D382 product. A fourth peak has an m/z of 1264 (∼41 min, −17 amu; not shown) and so is tentatively identified as the succinimide intermediate at position 382 (i.e., Su382NN).
The site of deamidation was confirmed using the daughter ions (i.e., b- and y″-ions) formed during high energy MS1 analysis of each of these peaks (see Fig. 2). Figure 2(A,C) show y, y, y, and y ions with a mass increase of +1 amu, consistent with deamidation in these fragments and which could have occurred at the N382, N387, or N388 sites. However, Figure 2(A,C) show no mass changes in the y, y, and y ions, which excludes deamidation at the N387 and N388 sites. Thus, using these daughter ions, deamidation at N382 was confirmed for both the first and third peaks of Figure 1(A), indicating that they are the isoD382 and D382 products, respectively. Both the relative peak areas and the elution order provide further support for these assignments, because in rpHPLC, isoD-containing peptides typically elute earlier than their D-containing counterparts.19–22 The isoD product is generally favored in unstructured peptides, with a typical isoD:D ratio of 3:1 to 5:1.7, 23 On this basis, the product peaks in Figure 1(A) are definitively assigned as isoD382NN [Fig. 1(A), 36.5 min] and D382NN[Fig. 1(A), 39.3 min]. Based on the identification of the two deamidation products at the N382 site, the succinimide intermediate [Fig. 1(A), 41 min] is assumed to be associated with this site as well. It should be noted that racemization can also occur via the succinimide to produce the D-forms of asparagine, aspartate, and isoaspartate containing species. D- and L-forms were not resolved by the UPLC/+ESI-MS assay used here, so the species identified may be mixtures of racemates. At long storage times, loss of parent and/or product peptide species occurred; the appearance of lower mass species suggested that that this loss is due to peptide bond hydrolysis (“clipping”). Kinetic studies were truncated if these losses exceeded ∼10%.
Deamidation products in the intact protein
Figure 3(A) shows the EIC for the G369-K390 fragment of intact Fc stressed for 28 h before tryptic digestion and UPLC/+ESI-MS analysis. Five species are detected, eluting at 36 min (Peak 1), 37 min (Peak 2), 37.5 min (Peak 3), 38.5 min (Peak 4), and 39.2 min (Peak 5). The molecular isotope envelopes [Fig. 3(B)] show that: (i) the first, fourth and fifth peaks correspond to singly deamidated products [Fig. 3(B, 1, 4 and B, 5)], (ii) the second peak corresponds to the parent peptide [Fig. 3(B, 2)], and the third peak corresponds to a doubly-deamidated product with +2 amu mass increase relative to the parent [Fig. 3(B, 3)]. By comparison with results for the synthetic model peptide standards and with the products of NNN deamidation, and by high energy MS1 analysis, the first, fourth, and fifth peaks are identified as the isoD382NN, ND387N, and D382NN variants, respectively [see Fig. 4(B)]. By high energy MS1 analysis [Fig. 4(A)] and the molecular isotope envelopes [Fig. 3(B, 3)], the third peak is identified as a doubly deamidated product. The presence of y, y, and y daughter ions with a mass change of +1 amu and of a y daughter ion with unchanged mass indicate that deamidation occurs at N387. A +2 amu mass change for the y and y daughter ions indicate a second deamidation site at N382. Thus, the third peak is deamidated at both the N382 and N387 sites; the identity of the deamidated sites with respect to isoD or D cannot be determined with the present data. The succinimide intermediates (i.e., Su382NN, NSu387N) were not detected at either site, suggesting that they were not formed or were degraded during tryptic digestion.
Detection of the NisoD387N deamidation product
The NisoD387N synthetic peptide was analyzed in an attempt to confirm that this product is not produced by deamidation at the N387 site in samples of the intact protein. The NisoD387N form is expected to be the major deamidation product at this site but was not detected in stressed samples of Fc-IgG, although the corresponding ND387N form is clearly present [see Fig. 3(A)]. A similar result was reported by Chelius et al.18 Interestingly, UPLC/MS analysis demonstrated that the NisoD387N synthetic peptide coelutes with the parent NNN peptide in EIC chromatograms, suggesting that this product could go undetected if it is formed. Additional analysis of isotopic distributions was conducted to estimate the amount of NisoD387N in the samples. Because the isoD product has greater mass (+1 amu) than the parent NNN peptide, the presence of NisoD387N as an impurity in the NNN peak would be expected to skew the relative intensities of the peaks in the molecular ion isotope envelope (see [Fig. 1(B)]. Specifically, in a sample containing both species, the first isotopic peak would be unique to NNN (i.e., the NNN “M” peak), but the second peak would contain contributions from both NNN and NisoD387N (i.e., the NNN “M + 1” peak and the NisoD387N “M” peak). The amount of NisoD387N in the sample was estimated by subtracting the theoretical intensity of the M + 1 peak of NNN from the observed intensity of the second peak. The “theoretical intensity” of the M + 1 peak of NNN was assumed to be identical to that measured for isoD382NN, a species that was cleanly resolved in EIC. The percentage contribution of NisoD387N to the second peak was estimated for each of three samples analyzed for each of the twelve time points of the study. By this estimation, the average percentage contribution of NisoD387N to the second peak for each time point was less than 2% of the NNN intensity, supporting the assertion that the isoD product is not formed at the N387 site.
The deamidation products for the NNN synthetic peptide are consistent with the accepted mechanism for deamidation at a single Asn site. In the discussion below, deamidation kinetics in these samples will be modeled as an irreversible first-order reaction to produce the succinimide (Su382NN), followed by parallel irreversible first-order reactions to produce the isoD and D products (isoD382NN, D382NN; [Fig. 5(A)] as in previous reports.7, 10, 23 In contrast, deamidation in the intact Fc IgG produced singly deamidated products at two sites (i.e., isoD382NN and ND387N) and a doubly deamidated product. The corresponding succinimide intermediates (i.e., Su382NN and NSu387N) were not detected. Reaction kinetics for the intact protein are consistent with parallel irreversible first-order deamidation to produce the singly deamidated products, followed by their further degradation to produce the doubly deamidated product [Fig. 5(B)]. In the discussion below, reaction kinetics for the NNN model peptide are presented first followed by results for the intact protein.
Deamidation kinetics in the NNN synthetic peptide
Kinetic profiles for deamidation in the NNN synthetic peptide are shown in Figure 6. The NNN peptide degrades in a pseudo-first order manner with ∼40% remaining after 100 h. Loss of the parent peptide is accompanied by transient formation of the succinimide intermediate and monotonic increase in the isoD- and D-containing products at N382 (i.e., isoD382NN and D382NN, respectively), which are formed in the ratio of ∼4:1 (see Fig. 6). The percentages were calculated assuming that sum of the N-, succinimide, isoD- and D-containing species is 100%; possible racemization and side reactions (e.g., clipping) are not included in the analysis.
The kinetic data for the parent tryptic fragment and its deamidation products were fitted simultaneously to the reaction scheme in [Fig. 5(A)] to provide values for the rate constants k1p, k2p, and k3p (Table I). The regression lines are in close agreement with the data (see Fig. 6) and the coefficients of variation are relatively small (%CV, Table I), indicating a good fit. The calculated half-life for loss of the parent NNN synthetic peptide is ∼81 h (i.e., 0.693 (k1p)−1 = 3.4 days), in reasonable agreement with the value of 110 h previously reported by Chelius et al. for deamidation in this fragment of an Fc IgG.18 Because the two studies were conducted in solutions of identical pH, storage temperature, buffer, and buffer concentration, the reasons for this difference are unclear, but may involve differences in peptide concentration. Deamidation at N387 and N388 was not detected during the time course of the studies reported here. Because Tyr is C-terminal to N388, the absence of deamidation products at this site is not surprising as steric hindrance by the bulky Tyr residue typically slows deamidation.6 The absence of deamidation products at N387 suggests that the presence of a second N, C-terminal to the N of interest, slows deamidation rate. Long-term incubation of the digests at 37°C led to clipping so that the slower deamidation at N387 and N388 could not be monitored.
Table I. Kinetic Parameters for Deamidation in Model Peptides and in the Intact Fc Fragment
Kinetic profiles for deamidation in the intact Fc IgG are shown in Figure 7. The fully N-containing form of the protein (i.e., NNN) undergoes monotonic loss with corresponding increases in the singly deamidated (i.e., isoD382NN, ND387N) and doubly deamidated (i.e., isoD382D387N) products. At time zero, the percentage of the fully N-containing form is less than 100 and the percentage of the isoD382NN product is greater than zero, suggesting that partial deamidation of N382 occurs during sample preparation or is initially present in the protein. The D382 product (i.e., D382NN) was also detected throughout the study at a constant level of ∼2–3% (not shown), consistent with its formation during sample preparation.
Kinetic analysis of deamidation in intact Fc IgG was performed according to the reaction scheme shown in Figure 5(B). The D382NN product was treated as an artifact of sample preparation and measured amounts of D382NN (1.5–2%) were reassigned to the parent form (NNN). Kinetic data for the parent protein and its degradation products then were fitted simultaneously to provide the microscopic rate constants k1 and k2, assuming that the rate of deamidation at each site is unaffected by deamidation at the second site (i.e., k1 = k4 and k2 = k3; [Fig. 5(B)]. Simultaneously, fitting the time-dependent profiles for the four species with a model containing two global fitting parameters produced good agreement between experimental and theoretical values (see Fig. 7).
The calculated half-life for the overall loss of the parent protein is ∼108 days (i.e., 0.693 (k1 + k2)−1; Table I), ∼30-fold slower than the 3.4 day half-life observed for the NNN synthetic peptide. The rate-limiting step in formation of isoD382NN in the synthetic peptide is formation of the succinimide rather than its hydrolysis (i.e., k1p < k3p, Table I), as expected for deamidation in unstructured peptides. The rate of formation of isoD382NN in the synthetic NNN peptide is then roughly equal to the rate of succinimide formation (k1p), a rate that is ∼50-fold greater than the rate of isoD382NN formation in the intact protein (k1 vs. k1p, Table I). This suggests that slowing of deamidation at N382 is a dominant contribution to the overall reduction in degradation rate. The rates of formation of the two singly deamidated products, isoD382NN and ND387N, are comparable in the intact protein (k1 vs. k2, Table I).
Molecular dynamics simulation
Using the PDB crystal structure for human Fc-IgG1 (1h3u), SYBYL MDS showed that N382 is located in a loop between two β-sheet regions, whereas the N387 and N388 residues are located in a relatively unstructured loop (see Supporting Information, Appendix II, for visualization). Solvent accessibility values and interatomic distances were calculated for all three N residues (Table II). Previous reports suggest that deamidation is favored by an interatomic distance between the attacking backbone nitrogen and the N-side chain carbonyl (i.e., CO to (N + 1)NH, Table II) of 1.89 Å or less, while distances greater than 4.9 Å are unfavorable. Here, the calculated interatomic distances are greater than 1.89 Å and less than 4.9 Å for all three N residues (Table II), suggesting some susceptibility to deamidation based on structure. The smallest simulated interatomic distance corresponds to N382, experimentally observed to deamidate most readily (Table II). Calculated solvent accessibility values decreased in the order N382 > N387 > N388, again corresponding to the measured deamidation rates (Table II).
Table II. Solvent Accessibility and Interatomic Distances by SYBYL MDSa
Solvent accessibility (%)
CO to (N + 1)NH distance, (Å)
Values determined by SYBYL molecular dynamics simulation (MDS) of PDB entry 1h3u.
The MDS results also help to explain the somewhat unusual distribution of deamidation products in the intact protein, in which: (a) only the isoD product is observed at N382 (i.e., isoD382NN but not D382NN), (b) only the D product is observed at N387 (i.e., ND387N but not NisoD387N), and (c) no deamidation is observed at N388. The formation of isoD382NN to the exclusion of D382NN suggests that hydrolysis of the succinimide occurs preferentially at the carbonyl contributed by the backbone amide and not at the carbonyl contributed by the N side chain. Such preferential hydrolysis could be related to the location of N382 between two β-sheet regions, which may hinder the attack of water from the side-chain side. In contrast, hindered hydrolysis of the succinimide is an unlikely explanation for the preferential formation of D at N387, because the site is relatively solvent exposed. The absence of deamidation products at N388 probably reflects the combined effects of primary sequence, with deamidation hindered by the N388 + 1 Tyr residue, and the limited solvent exposure of this site (Table II).
The rapid, high-resolution UPLC method was able to separate the N-containing 22-amino acid tryptic fragments and their deamidated variants. The parent NNN peptide was well resolved from its deamidated counterparts [Figs. 1(A) and 3(A)], thus emphasizing the utility of UPLC as a rapid, high resolution technique. The synthetic NNN peptide deamidated ∼30-times faster than the intact protein, suggesting that the rate of deamidation is slowed by protein structure. Protein structure also affected the product profiles. In the NNN synthetic peptide, the N382 site deamidated to form the expected isoD382 and D382 products in the typical ratios of 4:1; in the intact protein, only the isoD382 is observed without detectable D382. Molecular dynamics simulations supported the formation of isoD382 because the N382 residue was shown to be located in a structurally constrained orientation. In the NNN synthetic peptide, deamidation was not observed at the N387 site, because peptide bond hydrolysis (“clipping”) near the N- and C-termini prevented observation of the slower N387 deamidation reaction. In the intact protein, only the D387 product was observed at this site with no appearance of the corresponding isoD387 form. Together, these results indicate a significant role of the local secondary structure in both deamidation rate and product profiles in this region of Fc IgG.
The product profile is similar to that reported previously by Chelius et al. for Fc IgG.18 That group also observed deamidation at the N382 and N387 sites with exclusive formation of the isoD382 and D387 products. However, the doubly deamidated product was not detected by Chelius et al. for the intact protein, while the succinimide intermediate at N382 was observed. Their detection of the succinimide at N382 supports the hypothesis that the exclusive formation of the isoD product at this site involves the preferential hydrolysis of backbone side of the succinimide. As noted above, our failure to detect the succinimide in the intact Fc IgG samples may be the result of its hydrolysis during tryptic digestion. Recently, Ren et al. reported the degradation products in an Fc fusion protein, and detected the N(isoD)387N deamidation product, albeit at levels considerably lower than the corresponding ND387N product.24 The group also detected succinimide intermediates at both the N382 and N387 sites.24 These results further support the role of structure-dependent hydrolysis of the succinimide in altering the D:isoD product ratio.
In the work by Chelius et al., the half-life for loss of a G369-K390 tryptic fragment (110 h) is comparable to that for our NNN synthetic peptide (81 h; 0.693 (k1p)−1, Table I). In contrast, that group reported a half-life for loss of the intact Fc IgG protein of 38 days, while we observed a half-life of 108 days (0.693 (k1 + k2)−1, Table I), a difference of nearly threefold. The 38-day half-life was estimated based on concentrations of the parent G369-K390 tryptic fragment measured at three time points; the 108-day half-life reported here is the result of simultaneous nonlinear regression of multiple products at more than a dozen time points, and is therefore supported by considerably more kinetic data. Minor differences in experimental conditions for the two groups may also have influenced the results.
Similar results have also been reported for isomerization of a D-residue located in the CDR1 region of an antibody,25, 26 a reaction that also proceeds via a succinimide intermediate. In that work, succinimide hydrolysis in the denatured protein yielded isoD to D in a ratio of 3.5:1, but in the native antibody only D was formed under neutral and basic conditions. Unpublished work on deamidation in the CDR1 region by Vlasak et al. has shown that the ratio of isoD to D changes over time, presumably due to isomerization of isoD to D as the process tends toward equilibrium. Wearne et al.27 studied deamidation in RNase A and observed that the structured protein deamidates 30-fold slower at N67 than in the reduced and denatured form, a difference attributed to the local β-turn. Capasso et al. later observed that the isoD:D ratio for this protein was initially as high as in unstructured peptides (∼3:1) but decreased to 1:2 as the reaction approached equilibrium.9 A study by Harris et al. of Herceptin®, a commercial monoclonal antibody drug product, showed that N30 in the light chain deamidates to form only the D product.21 Atypical isoD:D ratios have also been observed for crystalline.10, 28 Capasso et al.29–31 proposed that the D product will dominate over its isoD counterpart wherever the D configuration has a lower energy, which in turn depends on the protein three-dimensional structure, and noted that equilibration between isoD and D via the succinimide occurs more slowly than deamidation. A similar rationale may hold for deamidation at the N387 of the Fc IgG studied here.
An alternative deamidation mechanism has been proposed by Clarke et al.,32 in which the backbone carbonyl oxygen rather than the (N + 1) nitrogen attacks the side chain carbonyl to form an isonimide ring with release of water. Subsequent hydrolysis of the symmetric isonimide ring from either side yields only the D-containing product, unlike hydrolysis of the succinimide which produces both isoD- and D-containing products. This mechanism has been promoted in several studies in which the formation of the D-containing product was observed to the exclusion of the isoD product, with the absence of the succinimide intermediate.33, 34 In this study, the formation of the D387 to the exclusion of isoD387 is consistent with an isonimide mechanism at that site. However, because the isonimide intermediate was not detected here, this explanation must be regarded as speculative.
Abbreviations: D382NN, peptide GFYPSDIAVEWESDGQPENNYK; EIC, extracted ion chromatogram; +ESI, electrospray ionization in the positive ion mode; Fc IgG, crystalizable fragment (Fc) of a monoclonal antibody (IgG); isoD, isoaspartate; (isoD)382NN, peptide GFYPSDIAVEWES(isoD)GQPENNYK; ND387N, the peptide GFYPSDIAVEWESNGQPEDNYK; N(isoD)387N, the peptide GFYPSDIAVEWESNGQPE(isoD)NYK; NNN, the peptide GFYPSDIAVEWESNGQPENNYK, corresponding to tryptic fragment G369-K390 of Fc IgG; Su succinimide, cyclic imide; an intermediate in deamidation; Su382NN, the peptide GFYPSDIAVEWES(Su)GQPENNYK; UPLC/MS, ultra performance liquid chromatography with tandem mass spectrometry.
Fc-IgG was prepared at Merck Research Laboratories from a humanized monoclonal antibody expressed in NS0 cells, a murine myeloma cell line. The human IgG1 Fc fragment was prepared by digestion with immobilized papain (Pierce, Rockford IL), followed by protein A purification. Impurities were further removed by cation-exchange high performance liquid chromatography (HPLC) using a ProPac WCX-10 9 × 250 mm column (Dionex, Sunnyvale, CA) and gradient elution (mobile phase A: 10 mM sodium phosphate pH 6.5; mobile phase B: 10 mM sodium phosphate, 0.5M NaCl pH 6.5; gradient: 3–26% B in 20 min at a flow rate of 2 mL/min).
Synthetic model peptides corresponding to the tryptic fragment G369-K390 (i.e., GFYPSDIAVEWESNGQPENNYK) and its deamidated variants were purchased from several sources. The parent peptide GFYPSDIAVEWESNGQPENNYK (abbreviated “NNN”) and the D387-containing deamidated variant, GFYPSDIAVEWESNGQPEDNYK (abbreviated “ND387N”), were synthesized by the Biochemical Research Services Laboratory of the University of Kansas. The isoD-containing variants, GFYPSDIAVEWES(isoD)GQPENNYK (abbreviated “isoD382NN”) and GFYPSDIAVEWESNGQPE(isoD)NYK (abbreviated “NisoD387N”), were purchased from American Peptide Company (Sunnyvale, CA). All other chemicals and reagents were of the highest commercial grade and used without further purification. Milli-Q water was used to prepare all solutions.
Sample preparation and accelerated stability studies
Accelerated stability studies were conducted for two types of samples at 37°C and pH 7.5: (i) the synthetic model peptide, NNN, and (ii) the fully-folded, intact Fc IgG (∼50 kD) (“intact protein”). For the NNN peptide (i), samples contained 50 μL of 0.5 mg/mL of peptide in 100 mM Tris-HCl buffer (pH 7.5). Samples were stored in capped microcentrifuge tubes at 37°C for up to 5 days. Triplicate samples were withdrawn at designated time intervals and stored at −20°C before LC-MS analysis.
Stability studies for the intact protein (ii) were conducted similarly, except that samples were subjected to tryptic digestion prior to analysis. Samples contained 50 μL of 0.5 mg/mL Fc IgG in 100 mM Tris-HCl buffer (pH 7.5), and were stored in capped microcentrifuge tubes at 37°C for up to 70 days. At each time point, triplicate samples were withdrawn, stored at −20°C, and then digested with trypsin before LC-MS analysis. Tryptic digestion followed the protocol reported by Chelius et al.18 Briefly, an aliquot of intact Fc was first exchanged to 6M guanidine-HCl, 0.2M Tris-HCl, 1 mM EDTA (pH 7.5) such that 100 μL of 0.5 mg/mL Fc was obtained. Reduction was performed by addition of 2 μL of 0.5M dithiothreitol (DTT) at 37°C for 40 min followed by alkylation by 4 μL of 0.5M iodoacteamide at room temperature for 40 min. The reduced, alkylated sample was then buffer exchanged with a 10,000 MW cut off membrane (Microcon centrifugal filter, Millipore, Billerica, MA) to 100 mM Tris buffer (pH 7.5) such that the Fc concentration was maintained at 0.5 mg/mL. Trypsin (Worthington, MA) was then added in a trypsin:protein ratio of 1:50 (w/w) and incubated for 2 h at 37°C. A second equal amount of trypsin was then added with incubation for 2 more hours, bringing the final trypsin:protein ratio to 1:25 (w/w). Samples were withdrawn and the reaction quenched by addition of 20% formic acid to a final concentration of 2%.
Mass spectrometric analysis (UPLC/+ESI-MS) of Fc-IgG and fragments
The analytical strategy for monitoring deamidation in Fc-IgG involved tryptic digestion of intact protein, ultraperformance liquid chromatography with tandem electrospray ionization mass spectrometry (UPLC/+ESI-MS) analysis of the fragments, and quantitation of peptides and their deamidated variants using extracted ion chromatograms (EIC). UPLC is a recent innovation in high performance liquid chromatography (HPLC), in which unique small particles and very low column volumes allow for greater throughput and rapid analysis. In the G369-K390 sequence, deamidation can occur at each of the N residues (N382, N387, N388) to produce the succinimide (Su) intermediate, as well as the isoD and D variants. Because each of the three sites can therefore exist in four forms (i.e., N, isoD, D, Su), a total of 64 (i.e., 43) variants is possible. Identification and quantitation of the product mixture thus requires analytical methods capable of resolving similar peptides in complex mixtures. UPLC is well-suited to the analysis of such mixtures, and was selected as the separation method to minimize possible coelution of isobaric products.
UPLC was performed using a Waters Acquity UltraPerformance liquid chromatography system (Waters, Milford, MA) with a BEH C18 column (1 mm × 150 mm, 1.7 μm particle size) at a flow rate of 100 μL mL/min and a column temperature of 35°C. Approximately 10 μg of protein or peptide was injected for each analysis (10 μL/injection). Gradient elution was used, (Solvent A: 99% H2O, 1% methanol, 0.08% formic acid; Solvent B: 99% acetonitrile, 1% H2O, 0.06% formic acid), with the profile: 0–1 min, solvent B at 1% (v/v); 1–2 min, solvent B increased to 10% (v/v); 2–19 min, solvent B increased to 18% (v/v); 19–45 min, solvent B increased to 22% (v/v); followed by washing and re-equilibration. Mobile phase solvent compositions were: Solvent A = 99% H2O, 1% methanol, 0.08% formic acid; Solvent B = 99% acetonitrile, 1% H2O, 0.06% formic acid.
The UPLC eluate was coupled to a Micromass® Q-Tof II mass spectrometer (Waters, Milford, MA) operated in the positive ion mode (+ESI) with a cone voltage of 35 V and a collision voltage of 30 eV. The relatively high collision voltage produces partial fragmentation of the parent ions and was adopted as an alternative to MS/MS analysis because it produced a sufficient number of daughter ions to allow unambiguous identification of deamidation sites. The G369-K390 fragment and the corresponding synthetic peptide carry a charge of +2, a theoretical mass of 2544.67 Da and thus an expected m/z value of 1272.57. Data were analyzed using MassLynx® software (Waters). Quantitation of native peptides and their deamidated variants was based on the isotopic peak intensities of each form. EIC scans were conducted for the m/z range 1264–1277 to allow identification and quantitation of the isoD- and D-deamidation products (+1 amu, +0.5 m/z) and the Su intermediate (−17 amu, −8.5 m/z). Analyte elution times typically ranged from 30 to 45 min; to facilitate comparison of EIC scans, retention times were normalized by setting the retention time of the NNN peptide peak to 37 min [Figs. 1(A) and 3(A)]. The theoretical isotope distribution for the parent peptide was used to separate the contribution of a doubly deamidated product, which eluted as a shoulder with the parent in studies of the undigested protein. The ionization efficiency of the parent peptide was assumed to be equivalent to that of its deamidated forms.
The time-varying concentrations of the N-containing parent species and their deamidated variants were subjected to kinetic analysis to determine the rate constants for deamidation. The native species and the deamidation products form an ensemble that evolves in time based on the microscopic rates of interconversion between species. Assuming that the deamidation process can be well represented as a unimolecular, first-order reaction, the time-dependencies of the population of “n” species can be described by a system of linear differential equations of the form:
where Ai, Aj, represent molar fractions of any two species in the ensemble, and kji is the microscopic rate constant describing the conversion of species Aj into species Ai:
The system of linear differential equations was solved using classical procedures35 to determine the eigenvalues of the system λi (i = 1,n). The eigenvalues λi represent experimental rate constants and can be expressed as linear combinations of the microscopic rate constants, kji. For a system comprising “n” distinct species there are at most “n − 1” values for λi, because one solution is always zero, corresponding to the equilibrium value. The solutions of the system, [Ai](t), were expressed as a sum of exponentials of the form:
where αij represent coefficients that are specific to each species, “i”. The coefficients αij are determined from the system of linear equations and from initial (t = 0) and equilibrium (t = ∞) conditions. Because the system is relatively uncomplicated, the eigenvalues and eigenvectors were calculated “by hand”, that is, no particular application was used to solve the system of equations describing the time-dependence of each species. The explicit form of the system of differential equations, with their corresponding eigenvalues and solutions for the molar fractions of the species, is available as Supporting Information (online).
For a particular data set (i.e., describing either the deamidation of a model peptide or the intact Fc), the microscopic rate constants were determined by performing a global fit, in which the fitting of the experimental points was performed simultaneously for all kinetic traces using the program SigmaPlot (ver. 10.0, Systat Software, San Jose, CA). Statistical information regarding the fitting results, including standard and relative errors of regression coefficients, was provided by the SigmaPlot routines.
Molecular dynamics simulations (MDS)
The Fc IgG1 structure was visualized using SYBYL (Version 8.0, Tripos, St. Louis, MO) which enabled direct measurement of key interatomic distances as represented in the 1h3u RCSB Protein Data Bank (PDB) crystal structure.36 The relative solvent accessibility of specific residues was determined in SYBYL by generating Connolly surfaces for those amino acids and evaluating the total area of each resulting surface.
The authors gratefully acknowledge assistance provided by members of the Mass Spectrometry Service Laboratory of The University of Kansas. The Q-TOF-2 was purchased with support from KSTAR, Kansas-administered NSF EPSCoR (T.D. Williams) and the University of Kansas.