Differential Enzyme-Linked Immunosorbent Assay ... - ACS Publications

Mei Han , Josh T. Pearson , Yunan Wang , Dwight Winters , Marcus Soto , Dan A. ... FGF21 Is Not a Major Mediator for Bone Homeostasis or Metabolic Act...
0 downloads 0 Views 411KB Size
Article pubs.acs.org/ac

Differential Enzyme-Linked Immunosorbent Assay and LigandBinding Mass Spectrometry for Analysis of Biotransformation of Protein Therapeutics: Application to Various FGF21 Modalities Todd Hager,†,∥ Chris Spahr,‡,∥ Jing Xu,§ Hossein Salimi-Moosavi,† and Michael Hall*,† †

PKDM, ‡Protein Sciences, and §Metabolic Disorders, Amgen Inc., Thousand Oaks, California 91320, United States S Supporting Information *

ABSTRACT: Novel protein therapeutics have become increasingly important modalities for treating diseases. Such therapeutics include recombinant fusions of pharmacoactive polypeptides to half-life extenders such as monoclonal antibodies, fragments of antibodies, and albumin. Half-life extension can also be achieved via chemical attachment to polymers such as polyethylene glycol. Any of these therapeutics may be susceptible to biotransformation, most notably in vivo proteolytic truncation, and it is vital to understand this phenomenon during early drug development to ensure correct pharmacokinetic profiling and optimize the in vivo stability through re-engineering. In this paper, we describe an integrated approach that combines differential enzyme-linked immunosorbent assay (ELISA) with ligand-binding-mass spectrometry (LBMS) to provide a thorough understanding of the biotransformation of novel protein therapeutics. Differential ELISA allows for a fast, high-throughput means to reveal gross in vivo proteolytic liabilities. Ensuing LB-MS analysis provides higher resolution details such as specific vulnerable loci to allow design refinement of the molecule. In this work, the power of the approach is elucidated by application to the optimization of a promising drug candidate, FGF21.

T

immunoreactivities of antibody reagents to unique epitopes on the target analyte are elucidated during reagent development, and this information is subsequently used to combine assay reagents with the desired specificities to enable a highly selective analytical system. The basic principle for this approach is to select reagents that would specifically recognize the sequences associated with biological activity thereby precluding the detection of the inactive truncated forms produced from degradation in this region. For application to fusion proteins, reagents that recognize the conventionally stable domains of the fusion partner (e.g., Fc) may be applied in a combinatorial fashion with reagents that are specific or nonspecific for the active moieties, thus enabling detection of the intact or total (i.e., intact plus truncated) forms. Such differentiating methods, when appropriately designed, would provide a highly sensitive and quantitative means of evaluating protein therapeutics in distinct forms as a function of in vivo degradation. However, despite offering these advantages over a traditional single assay approach, differential ELISA does have limitations inherent to immunoassays. Chiefly, reagent unavailability and/or lack of adequate reagent characterization may preclude a differential ELISA approach. Additionally, precise resolution at the amino acid sequence level would not be easily attainable with differential ELISA alone. To circumvent these limitations, differential ELISA can be combined with ligand-binding mass spectrometric (LB-MS)

he development of biologically stable, long-acting protein therapeutics remains challenging when many investigational molecules are peptides and small proteins that are susceptible to rapid clearance from circulation. This rapid clearance, or short half-life, is generally attributed to fast renal clearance due to low molecular weight, an effect that may substantially decrease clinical utility.1 Common half-life extension strategies involve increasing molecular size by conjugation to polymers such as polyethylene glycol (PEG) or genetic recombinant fusion with stabilizing molecules such as immunoglobulin fragment crystallizable region (Fc).1 While these methods may indeed improve the pharmacokinetics (PK) of the native peptide, the resulting increase in exposure to in vivo processes such as proteolytic degradation may then become a significant destabilizing factor for such bioengineered proteins, potentially leading to deactivation and loss of the desired pharmacodynamic (PD) effect. From an analytical perspective, the in vivo proteolysis may confound accurate quantification of the active therapeutic, which is commonly determined by an enzyme-linked immunosorbent assay (ELISA) that may not have the required specificity to distinguish among the intact molecule and its truncated forms. Furthermore, information derived from a single assay is most likely insufficient to direct further reengineering of the molecule to improve in vivo stability. Differential ELISA methods have been described as a means to distinguish between various forms of the analyte of interest (intact, total, etc.), in applications for evaluating the PK of large-molecule therapeutics,2,3 as well as for monitoring proteolysis of endogenous proteins.4 In differential ELISA, © 2013 American Chemical Society

Received: November 2, 2012 Accepted: February 2, 2013 Published: February 4, 2013 2731

dx.doi.org/10.1021/ac303203y | Anal. Chem. 2013, 85, 2731−2738

Analytical Chemistry

Article

terminal deletions were generated in E. coli as described elsewhere.11 Fc-FGF21 Fusion Proteins. Dimeric Fc fusion constructs were recombinantly generated, consisting of a human IgG1 Fc domain attached by a 15 amino acid linker to the N-terminus of human FGF21. The FGF21 sequence was either wild type (WT) or incorporated two mutations (RG) to suppress aggregation and reduce proteolytic liability in the wild type sequence, L98R and P171G, respectively.9 Structures of the various constructs are shown in Figure 1.

methods to provide a molecular understanding of the in vivo processing of biotherapeutics. LB-MS methods have been advantageously employed to generate sequence-specific information regarding biotransformation of candidate molecules as well as forming a basis for molecular re-engineering.5 LB-MS employs an initial ligand-binding (LB)/immunoaffinity solid phase capture to recover and enrich the therapeutic and its biotransformed species from dosed biological fluids. The eluted therapeutic and biotransformed species can then be directly queried by matrix-assisted laser desorption/ionization (MALDI)- or liquid chromatography (LC)−MS or digested with an appropriate endoproteinase such as trypsin and analyzed by LC−tandem mass spectrometry (MS/MS).5 Mass spectral analysis provides the resolution required to determine specific proteolytically labile loci that differential ELISA analysis cannot easily elucidate. On the other hand, LBMS approaches for molecular re-engineering are low throughput, quasi-quantitative at best (internal standards are not created for each truncated form), and initial screening by differential ELISA can immensely aid in guiding LB-MS analyses. Fibroblast Growth Factor 21 (FGF21) is a 19 kDa protein currently under preclinical investigation as a potential therapeutic for type 2 diabetes.6 The native molecule has a very short terminal half-life of ∼1−2 h in rodents and nonhuman primates, most likely due to rapid renal clearance.7,8 In order to enhance PK/PD properties of FGF21, recombinant C-terminal fusions with human Fc were generated yielding a homodimer composed of two Fc-linker-FGF21 moieties which increases the overall mass to 92 kDa.9 Fusion to human Fc reduces clearance and prolongs PD effects due to both the increase in molecular weight as well as recycling via FcRn.10 However, prolonged circulatory residence may expose these constructs to endogenous proteases resulting in blunted or abolished pharmacological activity. In fact, the pharmacological effects of FGF21 are highly dependent on the presence of Nand C-terminal residues; truncation at either terminus results in dramatic loss of biological activity.11 Accordingly, Fc fusions of native FGF21 demonstrated PK/PD characteristics in mice that were unimproved relative to the native protein.9 Further investigation by LB-MS identified major clipping at proline 414, corresponding to position 171 in the FGF21 sequence.9 Constructs were subsequently re-engineered with proline at this position replaced by glycine (P171G) to reduce proteolysis and incorporated a further alteration to address previously recognized aggregation issues (L98R).9 The re-engineered construct demonstrated comparable potency in vitro relative to the unmodified construct.9 Moving forward, it was important to examine the in vivo stability of the modified Fc-FGF21 fusion protein. In this paper, we demonstrate a powerful coordinated analytical strategy whereby differential ELISA and LB-MS are used as complementary methods for evaluating distinct forms in serum samples from preclinical species administered with FcFGF21 fusion proteins. The integrated approach provides a more effective and comprehensive biotransformation assessment to guide candidate selection than either method platform can provide alone.

Figure 1. FGF21 and Fc-FGF21 constructs. FGF21 point mutations are shown in red and numbered according to their position within FGF21 and not within the construct sequence: (A) native FGF21 [1181], (B) Fc-FGF21 WT [1-424], and (C) Fc-FGF21 RG [1-424].

Differential ELISA Reagent Characterization. Mouse monoclonal antibodies (mAb) and rabbit polyclonal antibodies (pAb) were generated using human FGF21. Two mouse mAb were identified (epitope mapping and specificity characterization sections and Figure S-1A,B in the Supporting Information): mAb 3.13 was specific for the N-terminus and mAb 3.7 specific for an epitope approximately 20 residues upstream from the C-terminus at amino acids 160-164 of the FGF21 sequence. C-terminal-specific pAb were obtained from high-titer rabbit serum by affinity purification using a peptide column consisting of FGF21 C-terminal residues. Specificity studies showed decreasing immunoreactivity of the C-terminal pAb with sequential loss of 1-4 C-terminal residues, which correlated with the loss of FGF21 bioactivity11 (Figure S-1C in the Supporting Information). Two mouse mAb were raised and developed against distinct epitopes of human Fc (mAb 100 and 1.35). Preclinical Biotransformation Study Design. All study procedures were approved by the Amgen Inc. Institutional Animal Care and Use Committee and conducted in accordance to the Guide for the Care and Use of Laboratory Animals, 8th edition. Mouse Study of Native FGF21. Native human FGF21 was intravenously (IV) administered to C57BL/6 mice at 10 mg/ kg. Serum samples were collected at each time point from three



EXPERIMENTAL SECTION FGF21 Constructs. Native FGF21 and Truncated Forms. Native full-length FGF21 and truncated forms with various 2732

dx.doi.org/10.1021/ac303203y | Anal. Chem. 2013, 85, 2731−2738

Analytical Chemistry

Article

capture antibody (assay-dependent) was diluted to 2 μg/mL with 1× Dulbecco’s phosphate buffered saline (Gibco) and passively adsorbed onto microtiter plates for 12−24 h at 4 °C. Standards and quality controls (QCs) were prepared by spiking the appropriate FGF21 analogue into 100% plasma or serum of the studied species (Bioreclamation, Inc.). Standards, QCs, matrix blank, and unknown samples were diluted with I-Block buffer (Applied Biosystems) and loaded into the wells. After incubation for 2 h at 25 °C and washing steps, the analytes of interest were detected by the addition of a biotinylated antibody specific for FGF21 or Fc (assay-dependent). After incubation for 1 h at 25 °C and subsequent washing, a streptavidin-peroxidase conjugate (R&D Systems, Inc.) was added to the wells and bound to the biotinylated antibody. After incubation for 30 min at 25 °C and final washing steps, colorimetric development was initiated by the addition of a 3,3′,5,5′-tetramethylbenzidine peroxide substrate (KPL, Inc.), and the bound analyte was quantified by measuring the optical density at 450 nm on a SpectraMax 340PC microplate reader (Molecular Devices). The conversion of optical density units to concentrations in the unknown samples was achieved through a software-mediated comparison to the analyte standard curve assayed on the same plate. Data regression was performed in Watson v7.0.0.01 (Thermo Fisher Scientific). LB-MS. LB-MALDI-TOF MS. Plasma samples (60 μL) were reduced with tris(2-carboxyethyl)phosphine (5 mM, 30 min, room temperature) followed by alkylation with iodoacetamide (15 mM, 1 h, room temperature, dark). These steps decrease the background components and permit MALDI analysis of a better resolved ∼46 kDa monomer versus a ∼92 kDa homodimer. The samples were diluted to 25% (v/v) plasma with 10 mM Hepes, 150 mM NaCl, 3 mM EDTA, 0.005% Tween-20 and processed with mAb 1.35 (anti-huFc) immunoaffinity tips.5 After washing, the tips were eluted with 5 μL of sinapinic acid solution [12 mg/mL in 33% (v/v) aq. methanol, 0.8% (v/v) aq. trifluoroacetic acid] and transferred directly to a stainless steel MALDI target. The plate was allowed to dry, and MALDI analysis was performed on a Bruker Autoflex II using an optimized positive-ion linear mode method. LB-Digestion-LC−MS/MS. Serum samples were incubated with anti-huFc mAb 1.35 coupled to Sterogene Bioseparations (Carlsbad, CA) Actigel (4.5 mg mAb/mL gel). The samples were washed extensively with PBS, then eluted in ∼200 μL of 10 mM HCl. An aliquot of 30 μL (∼2.4 μg for the early time points) was dried down, resuspended in 100 mM Tris, pH 8.0/ 2 M urea, and then digested with Asp-N (1:40) for 4 h at 37 °C. Asp-N digest aliquots (4 μL) were injected onto an Eksigent 1D-Plus HPLC (Dublin, CA) system equipped with a Waters (Milford, MA) NanoEase Symmetry300 C18 trapping column and a Grace (Columbia, MD) Vydac C18 75 μm × 250 mm analytical column. The mobile phase consisted of 0.1% formic acid (A) and 0.1% formic acid/90% acetonitrile (B) at a flow rate of 300 nL/min. The 75 min gradient consisted of the following conditions: isocratic for 2 min at 1% B, up to 10% B over 3 min, up to 40% B over 45 min, up to 80% B over 10 min, isocratic at 80% B for 5 min, back to 1% B over 5 min, and then isocratic at 1% B for 5 min. Effluent from the HPLC column was sprayed into an AB Sciex (Foster City, CA; Toronto, Canada) 4000 Q TRAP mass spectrometer equipped with a nanospray ionization source, an IonSpray II head, and a New Objective (Woburn, MA) fused silica emitter tip. Multiple reaction monitoring (MRM) transitions were created to

mice and stored frozen at −80 °C until analysis by ELISA or LB-MS. Cynomolgus Monkey Study of Fc-FGF21 Constructs. FcFGF21 WT and RG were injected IV into cynomolgus (cyno) monkeys at 23.5 mg/kg and serum and plasma time points were collected out to 840 h post dose. There were no significant differences between serum and plasma by ELISA (data not shown). Because of limitations of sensitivity, no time points beyond 168 h were analyzed for the LB-MS analyses. Evaluation of Drug Levels. ELISA. Two distinct assays for evaluation of either intact or C-terminal truncated FGF21 were applied to the mouse study (Figure 2A,B). Four distinct assays

Figure 2. Schematic of differential ELISA formats. Monomeric Fc constructs are drawn for simplicity to illustrate specificity. Capture Ab are depicted below the construct and detection Ab above. (A) Intact FGF21, recognizes only intact FGF21 (ELISA no. 1); (B) C-terminal truncated FGF21, recognizes intact and midstream C-terminal truncated FGF21 (ELISA no. 2); (C) total fusion, recognizes intact and truncated forms with linker (ELISA no. 3); (D) total Fc, recognizes all forms containing Fc (ELISA no. 4).

specific for different forms of the administered FGF21 fusion proteins were developed for the cyno monkey study (Figure 2A−D). The lower limits of quantification for all the assays were in the range of 0.5−0.9 nM. Standard calibrators were prepared with the intact forms of native FGF21, Fc-FGF21 WT, or Fc-FGF21 RG, for the concentration determination of the corresponding forms in unknown samples. The method accuracy and validity of comparative quantification were demonstrated by performing spike-recovery experiments with full-length and truncated forms of Fc-FGF21 (relative immunoreactivity evaluations and Figure S-2 in the Supporting Information). Although we cannot feasibly produce nor evaluate all potential truncations in a complex in vivo sample, we are confident in making the assumption that the major species present would retain comparable immunoreactivity and be quantified by the current methods. Assay procedures were based on a sequential-addition sandwich ELISA. Briefly, either an FGF21- or Fc-specific 2733

dx.doi.org/10.1021/ac303203y | Anal. Chem. 2013, 85, 2731−2738

Analytical Chemistry

Article

assay (ELISA no. 1). These data are of utmost importance since they reflect levels of the pharmacologically active forms and therefore direct the LB-MS analyses. The results from the other assays also contribute to the understanding of in vivo stability of the constructs and any potential re-engineering improvement. As Figure 4 depicts, the intact FGF21 form (ELISA no. 1) had the greatest disparity between the two constructs; the intact form was quantifiable in the plasma of the Fc-FGF21 RG treated monkey for 14 days but was below the lower limit of quantification (0.9 nM) after 24 h of Fc-FGF21 WT administration. The C-terminal truncated FGF21 form (ELISA no. 2) remained quantifiable for 28 days, far longer than the intact form in either construct. Finally, the total fusion and total Fc forms (ELISA no. 3 and no. 4, respectively) were the most persistent, remaining well above the quantification limit in the final samples collected on day 35. PK analyses of ELISA data were performed and parameter estimates are summarized in Table S-1 in the Supporting Information. The terminal half-life estimate determined in plasma via the intact FGF21 ELISA was 30.2 h for Fc-FGF21 RG, approximately 17-fold longer than 1.80 h for Fc-FGF21 WT. Likewise, the exposure expressed as AUC0‑inf was approximately 5-fold higher for the intact form in Fc-FGF21 RG than that in Fc-FGF21 WT. Systemic clearance estimation for the intact form in Fc-FGF21 RG was approximately 20% of that of Fc-FGF21 WT. These results indicated that the P171G point mutation in the Fc-FGF21 construct markedly reduced the proteolytic truncation of the construct. Although the intact ELISA data show that the RG construct was more stable than the WT in vivo, the information derived from the other ELISAs offered additional insight into degradation vulnerability and potential areas for improvement. The longer half-life of the truncated and total forms relative to that of the intact form implies that significant proteolytic cleavage still occurred within the FGF21 moiety. The intermediate profile derived from the C-terminal truncated assay implies that clipping occurred both upstream and downstream of the detection mAb’s recognition site at residues 160-164 within the FGF21 moiety. The exposure proportionality when determined by intact relative to C-terminal truncated FGF21 and C-terminal truncated relative to total forms was approximately equivalent (0.4−0.5). This equivalence indicates that clipping at sites both upstream and downstream of residues 160-164 occurred with similar prevalence. Within the same construct, the half-life derived from the total fusion assay was slightly lower than the total Fc measurement, indicating that the linker segment connecting the Fc and FGF21 moiety was relatively stable with only moderate hydrolysis. LB-MS. On the basis of tracking various epitopes of FGF21 constructs by differential ELISA, the results provided quantitative evidence of proteolytic processing of Fc-FGF21 WT and RG, showing the WT construct to be more unstable in vivo. However, in order to reveal specific amino acids where truncation was occurring, high resolution LB-MS studies had to be undertaken as this information was not obtainable by differential ELISA alone. The overall role of LB-MS analysis is to determine the proteolytically sensitive residues in a time dependent manner and therefore the data are inherently qualitative or quasi-quantitative at best. Absolute quantification of the parent or its truncated forms is not the goal of this analysis. LB-MALDI-TOF MS. LB-MALDI-TOF MS provided an overview of the in vivo stability profile of Fc-FGF21 WT and

specifically monitor for various truncated and nontruncated forms of the Asp-N derived C-terminal peptide of WT and RG as described in detail in the optimization of LB-digestion/MRM section of the Supporting Information. Pharmacokinetic Analysis. Noncompartmental analysis of plasma Fc-FGF21 WT and Fc-FGF21 RG concentrations derived from the ELISA assays were performed using WinNonlin Enterprise (v.5.1.1 Pharsight Corporation, Mountain View, CA). PK parameters are defined in the pharmacokinetic analysis section of the Supporting Information.



RESULTS AND DISCUSSION Differential ELISA. PK Profile of Native FGF21. Data from previous PK studies of native FGF21 in mice and cyno demonstrated that the native protein possessed a short terminal half-life.7,8 Prior to availability of FGF21 C-terminal specific pAb, we determined the N-terminal region was relatively stable. Samples from mice dosed with native FGF21 were evaluated using ELISAs comprised of mAb 3.7 for capture, with either mAb 3.13 or a total-FGF21 pAb for detection antibodies. The resulting concentration−time profiles were superimposed, which demonstrated N-terminal stability (data not shown). The question remained whether instability of the C-terminal region contributed to rapid clearance. In the current study, to investigate whether the rapid kinetics originated from proteolytic events or systemic clearance, native FGF21 was administered as a single IV dose of 10 mg/kg to C57BL/6 mice. Serum samples were evaluated for intact and C-terminal truncated forms of FGF21 (ELISA no. 1 and no. 2). Rapid clearance was observed with both assays, and serum FGF21 was below the detection limit (0.5 nM) at 12 h post injection. The concentration−time profiles were similar, implying the fast clearance of native FGF21 was a function of mass-based systemic elimination rather than proteolytic degradation in circulation (Figure 3).

Figure 3. Mean serum concentration−time profiles for native FGF21 in mice. Levels were determined by both intact (ELISA no. 1, purple trace) and C-terminal truncated (ELISA no. 2, blue trace) assays.

PK Profiles of Fc-FGF21 WT and Fc-FGF21 RG. Concentrations of Fc-FGF21 WT and Fc-FGF21 RG in cynomolgus monkey plasma were determined by differential ELISA assays to monitor various forms of the fusion proteins over time (Figure 4). The most significant difference between the two constructs was observed in the data from the intact FGF21 2734

dx.doi.org/10.1021/ac303203y | Anal. Chem. 2013, 85, 2731−2738

Analytical Chemistry

Article

Figure 4. PK profiles for (A) Fc-FGF21 WT and (B) Fc-FGF21 RG in cynomolgus monkeys. Concentrations were determined using ELISA no. 1 (purple), no. 2 (light blue), no. 3 (dark blue), and no. 4 (red). The left panels are the full PK profiles; the right panels are expansions of the early time points of the PK profiles (through 1 week postinjection).

at ∼41.4 kDa. The calculated mass loss of ∼4715 Da suggested truncation after P133 [1-376]. The identity of this truncation point was further corroborated by Asp-N digestion of the sample followed by LC−MS/MS (data not shown). This truncation at P133 was also visible after 168 h in the WT spectrum. Two other minor truncated components appeared between 42.5 and 45 kDa (also circled in purple) that corresponded to likely truncations after P393 and L409 based on mass differentials; however, these truncations were not confirmed by LC−MS/MS. LB-Digestion-LC−MS/MS. LB-digestion-LC−MS/MS of the Fc-FGF21 WT and RG samples provided higher resolution detail over a more focused mass range on the mixture of Cterminal truncations (i.e., loss of 1−3 residues) that was suggested by the LB-MALDI data (Figure 6). Predictably, the total signal of each form of the Asp-N derived C-terminal peptide decreased over time due to both systemic elimination as well as proteolytic degradation (signal intensity not shown). Forms containing [407-424] and [407-423] (derived from intact [1-424] or [1-423]) retained full bioactivity, while the bioactivity of [407-422] was reduced. The loss of a third amino acid ([407-421] and shorter) rendered Fc-FGF21 inactive. For this reason, the relative abundance/distribution of each Cterminal form was critically important. No internal standards of the various truncated peptide species were utilized as correction

RG time points over a large mass range. Since this is a “topdown” analysis, any truncated forms, where small numbers of residues are lost and there are no large changes in overall molecular size, are expected to have similar ionization efficiencies as the parent. The results for WT are shown in Figure 5A. Truncation at P171 [1-414] occurred rapidly (evident after 2 h), and by 6 h, most of the intact construct [1-424] was converted to this truncated form. The results for RG are shown in Figure 5B, with the dashed red line indicating intact construct. In contrast to WT, no truncation at P171 was observed in RG due to mutation to G171. The main biotransformation of RG was the gradual loss of up to three C-terminal residues which appear to be occurring from 48 to 168 h (i.e., the broadening peaks slightly to the left of the dashed red line). These peaks represent the incompletely resolved mixture of the loss of 1−3 C-terminal residues; at 168 h, the peak corresponding to loss of the three terminal residues was readily identifiable (circled in blue). The data suggested that the intact construct had mostly disappeared between 96 and 168 h. The results correlated well with the “intact” ELISA profile (no. 1) in Figure 4 which slightly diverged from the other ELISA profiles at 24 h, with a more apparent divergence from 48 h onward. The MALDI spectrum of the 168 h sample clearly showed larger truncations with significant mass losses (circled in purple). Of these, the most abundant one appeared 2735

dx.doi.org/10.1021/ac303203y | Anal. Chem. 2013, 85, 2731−2738

Analytical Chemistry

Article

Figure 6. LB-digestion-LC−MS/MS results: (A) Fc-FGF21 WT and (B) Fc-FGF21 RG. The colors of the traces correspond to the intact or truncated C-terminal peptides shown in the same color in the figure.

414] was complete by 24 h, indicating a rapid loss of 10 residues from the C-terminus of the WT molecule. In contrast, the RG construct slowly lost up to three Cterminal residues throughout the time course. After 6 h, ∼87% of the RG C-terminal peptide [407-424] existed. After 24 h, ∼63% of the RG C-terminal peptide [407-424] existed, with ∼7% appearing as [407-423], ∼12% as [407-422], and ∼16% as [407-421] (Figure 6B). After 48 h, the RG sample ∼44% of the RG sample remained as [407-424], ∼9% as [407-423], ∼15% as [407-422], and ∼28% as [407-421]. At 168 h, < 10% of the nontruncated C-terminal peptide [407-424] remained; however, there was still ∼6% [407-423], ∼8% [407-422], and ∼66% of [407-421]. These results correlated well to those of LB-MALDI. LBMALDI provided a quick read over a larger mass range whereas the LC−MS/MS approach had better resolution to track the losses of the last few residues from the C-terminus that were only suggested by the MALDI data. At 168 h, the LB-MALDI method, with the larger mass range output, identified larger truncated forms that were not monitored by the focused LBdigestion-LC−MS/MS method. Significantly, the trends in

Figure 5. LB-MALDI results: (A) Fc-FGF21 WT and (B) Fc-FGF21 RG.

factors. This was deemed unnecessary as all peptides were derived from the same base peptide sequence (DPLSMVGPSQGRSPSYAS [407-424]), and similar parameters were used for the MRM (i.e., the same b6 fragment ion [DPLSMV] derived from the N-terminal end of the peptide that is not undergoing biotransformation was used for all MRM transitions). Still, small biases and differences in ionization efficiency were likely. After 5 min, ∼81% of the WT intact C-terminal peptide remained [407-424], while ∼18% was estimated to be clipped and appeared as [407-414] (Figure 6A). After 6 h, roughly 19% of the intact C-terminal peptide remained [407-424], while 78% had been clipped to [407-414]. The processing to [4072736

dx.doi.org/10.1021/ac303203y | Anal. Chem. 2013, 85, 2731−2738

Analytical Chemistry

Article

MS-based results were only quasi-quantitative at best, due to the lack of standards of the truncated forms. Regardless, the complementary nature of all three assays provided a complete picture of the biotransformation of Fc-FGF21. LB-MS biotransformation studies during the molecular refinement stage of a therapeutic aided in the design of a better ELISA (one that measures only bioactive species and not inactive truncated forms). As long as any further molecular refinement of a therapeutic is occurring, the LB-MS tools can serve as a complementary aid to differential ELISA. This analytical strategy has potential application to a wide range of fusion proteins for assessing in vivo stability and to provide detailed information for biotherapeutic construct design.

biotransformation with respect to time were similar in both methods. In these studies, the FGF21 portion of the fusion molecule was cleaved with Asp-N. The enzymatically derived C-terminal peptides DPLSMVGPSQGRSPSYAS [407-424] (WT) and DPLSMVGGSQGRSPSYAS [407-424] (RG) were both large enough to monitor the rapid C-terminal cleavage of 10 residues from WT as well as minor carboxypeptidase-like cleavage in RG introduced following the P171 → G171 mutation. Alternative digestion strategies were examined and rejected since they were less than ideal in producing desired surrogate peptides. For other therapeutics, the optimal enzyme for digestion needs to be determined empirically based upon primary sequence. With the current assay, measuring a diagnostic fragment ion rather than the precursor yielded greater assay specificity; however, the total signal dropped much quicker, preventing the analysis of time points beyond 168 h. To increase the sensitivity of the MRM, a higher dose of therapeutics could be used, along with LB of a larger volume of serum. Alternatively, newer highresolution instruments (e.g., Fourier transform-mass spectrometry (FT-MS)) could replace the need for fragmentation and therefore MRM. Specifically, narrow extracted ion chromatograms (XIC) from a high-resolution survey scan MS data could provide similar biotransformation profiles.



ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*Address: One Amgen Center Dr., MS 30E-3-C, Amgen Inc., Thousand Oaks, CA 91320. Phone: 805/313-6484. Fax: 805/ 499-9027. E-mail: [email protected].



Author Contributions

CONCLUSIONS During early exploration of FGF21 as a potential therapeutic, we recognized the limitations of the native molecule with respect to rapid clearance, and subsequently, the construction of recombinant fusion proteins was initiated to improve PK. However, fusion to human Fc only resulted in modest improvement, with an undesirable t1/2 of the bioactive form caused by in vivo proteolysis. Using a differential ELISA approach coordinated with LB-MS analysis, we have implemented a highly sensitive and specific strategy to study the effect of biotransformation of Fc fusions of FGF21 on drug exposure and guide construct re-engineering. The complementary technologies provided qualitative and quantitative aspects of the intact peptide regions as well as truncation products arising from in vivo degradation. Significantly, the terminal half-life of the intact form of Fc-FGF21 RG was ∼17-fold greater than that of the intact form of Fc-FGF21 WT, indicating the mutated sequence conferred greater protection from degradation. From the well-defined assay specificity, we now have a better understanding of the proteolytic liabilities in these FGF21 fusion proteins and can use this information to assist in further re-engineering to enhance the PK of these molecules. Minor changes to the MRM transitions would allow other C-terminal point mutations to be monitored and provide data complementary to both ELISA and LB-MALDI. Each technology played a complementary role to overcome a limitation present in the other. ELISA, while lacking the sequence-level resolution, was quantitative, had the greatest sensitivity, and could measure Fc-FGF21 therapeutic longer than either LB-MALDI or LB-digestion-LC−MS/MS. The LBMALDI readily provided the best “global” overview of the construct, though it lacked sufficient resolution for tracking the carboxypeptidase-like C-terminal degradation of the RG mutant. The LB-digestion-LC−MS/MS method had the highest specificity but the lowest throughput and was blinded to any component without an MRM transition (e.g., endopeptidase truncation after P376 was ignored). Both of the



Both authors contributed equally to this work and are listed in alphabetical order. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors would like to acknowledge the late George Doellgast for contributions to the differential ELISA methods and Seth Fisher for additional sample purification support. The authors would also like to thank Jean Lee, Hsieng Lu, and Stone Shi for support and critical review of this manuscript.



REFERENCES

(1) Kontermann, R. E. Curr. Opin. Biotechnol. 2011, 22, 868−876. (2) Gan, J.; Kendra, K.; Ricci, M.; Hank, J. A.; Gillies, S. D.; Sondel, P. M. Clin. Diagn. Lab. Immunol. 1999, 6, 236−242. (3) Kendra, K.; Gan, J.; Ricci, M.; Surfus, J.; Shaker, A.; Super, M.; Frost, J. D.; Rakhmilevich, A.; Hank, J. A.; Gillies, S. D.; Sondel, P. M. Cancer Immunol. Immunother. 1999, 48, 219−229. (4) Diamandi, A.; Mistry, J.; Krishna, R. G.; Khosravi, J. J. Clin. Endocrinol. Metab. 2000, 85, 2327−33. (5) Hall, M. P.; Gegg, C.; Walker, K.; Spahr, C.; Ortiz, R.; Patel, V.; Yu, S.; Zhang, L.; Lu, H.; DeSilva, B.; Lee, J. W. AAPS J. 2010, 12, 576−585. (6) Kharitonenkov, A.; Shiyanova, T. L.; Koester, A.; Ford, A. M.; Micanovic, R.; Galbreath, E. J.; Sandusky, G. E.; Hammond, L. J.; Moyers, J. S.; Owens, R. A.; Gromada, J.; Brozinick, J. T.; Hawkins, E. D.; Wroblewski, V. J.; Li, D. S.; Mehbod, F.; Jaskunas, S. R.; Shanafelt, A. B. J. Clin. Invest. 2005, 115, 1627−1635. (7) Kharitonenkov, A.; Wroblewski, V. J.; Koester, A.; Chen, Y. F.; Clutinger, C. K.; Tigno, X. T.; Hansen, B. C.; Shanafelt, A. B.; Etgen, G. J. Endocrinology 2007, 148, 774−781. (8) Xu, J.; Stanislaus, S.; Chinookoswong, N.; Lau, Y. Y.; Hager, T.; Patel, J.; Ge, H.; Weiszmann, J.; Lu, S. C.; Graham, M.; Busby, J.; Hecht, R.; Li, Y. S.; Li, Y.; Lindberg, R.; Véniant, M. M. Am. J. Physiol. Endocrinol. Metab. 2009, 297, 1105−14. (9) Hecht, R.; Li, Y.-S.; Sun, J.; Belouski, E.; Hall, M.; Hager, T.; Yie, J.; Wang, W.; Winters, D.; Smith, S.; Spahr, C.; Tam, L.-T.; Shen, Z.; Stanislaus, S.; Chinookoswong, N.; Lau, Y.; Sickmier, A.; Michaels, M.;

2737

dx.doi.org/10.1021/ac303203y | Anal. Chem. 2013, 85, 2731−2738

Analytical Chemistry

Article

Boone, T.; Véniant, M. M.; Xu, J. PLoS ONE 2012, 7, No. e49345, DOI: 10.1371/journal.pone.0049345. (10) Suzuki, T.; Ishii-Watabe, A.; Tada, M.; Kobayashi, T.; KanayasuToyoda, T.; Kawanishi, T.; Yamaguchi, T. J. Immunol. 2010, 184, 1968−76. (11) Yie, J.; Hecht, R.; Patel, J.; Stevens, J.; Wang, W.; Hawkins, N.; Steavenson, S.; Smith, S.; Winters, D.; Fisher, S.; Cai, L.; Belouski, E.; Chen, C.; Michaels, M. L.; Li, Y.-S.; Lindberg, R.; Wang, M.; Véniant, M.; Xu, J. FEBS Lett. 2009, 583, 19−24.

2738

dx.doi.org/10.1021/ac303203y | Anal. Chem. 2013, 85, 2731−2738