An online four-dimensional HICxSEC-IMxMS methodology for proof-of

Subscriber access provided by UNIV OF NEW ENGLAND ARMIDALE

Article

An online four-dimensional HICxSEC-IMxMS methodology for proof-of-concept characterization of antibody drug conjugates. Anthony Ehkirch, Valentina D'Atri, Florent Rouvière, Oscar Hernandez-Alba, Alexandre Goyon, Olivier Colas, Morgan Sarrut, Alain Beck, Davy Guillarme, Sabine Heinisch, and Sarah Cianférani Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.7b02110 • Publication Date (Web): 20 Dec 2017 Downloaded from http://pubs.acs.org on December 21, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

An online four-dimensional HICxSEC-IMxMS methodology for proof-of-concept characterization of antibody drug conjugates. Anthony EHKIRCH1‡, Valentina D’ATRI2‡, Florent ROUVIERE3‡, Oscar HERNANDEZ-ALBA1, Alexandre GOYON2, Olivier COLAS4, Morgan SARRUT3, Alain BECK4, Davy GUILLARME2, Sabine HEINISCH3*, Sarah CIANFERANI1* 1

Laboratoire de Spectrométrie de Masse BioOrganique, Université de Strasbourg, CNRS, IPHC UMR 7178, 67000 Strasbourg, France

2

School of Pharmaceutical Sciences, University of Geneva, University of Lausanne, CMU - Rue Michel-Servet, 1, 1206 Geneva – Switzerland

3

Université de Lyon, Institut des Sciences Analytiques, CNRS UMR5280, Université de Lyon, Ens, 69100 Villeurbanne, France

4

IRPF - Centre d’Immunologie Pierre-Fabre (CIPF), 74160 Saint-Julien-en-Genevois, France

ABSTRACT: There are currently two main techniques allowing the analytical characterization of inter-chain cysteinelinked antibody drug conjugates (ADCs) under native conditions, namely hydrophobic interaction chromatography (HIC) and native mass spectrometry (MS). HIC is a chromatographic technique allowing the evaluation of drug load profile and calculation of average drug to antibody ratio (DAR) in quality control laboratories. Native MS offers structural insights into multiple ADC critical quality attributes, thanks to accurate mass measurement. However, both techniques used as standalone can lead to misinterpretations or incomplete characterization. Online coupling of both techniques can thus potentially be of great interest, but the presence of large amounts of non-volatile salts in HIC mobile phases makes it not easily directly compatible with native MS. Here, we present an innovative multidimensional analytical approach combining comprehensive online two-dimensional (2D)-chromatography that consists of HIC and size exclusion chromatography (SEC), to ion mobility and mass spectrometry (IM-MS) for performing analytical characterization of ADCs under non-denaturing conditions. This setup enabled comprehensive and streamlined characterization of both native and forced degraded ADC samples. The proposed 4D methodology might be more generally adapted for online all-in-one HICxSEC-IMxMS analysis of single proteins or protein complexes analysis in non-denaturing conditions.

INTRODUCTION Monoclonal antibodies (mAbs) and their related compounds make up the largest class in human therapeutics to treat cancers, autoimmunity and inflammatory diseases.1 The success of mAbs stems from their high specificity and affinity, long circulating half-lives, ability to induce immune cell effector response, and structural versatility. However, canonical mAbs often show a limited efficacy or face resistance, so armed antibodies (such as antibody drug conjugates, ADCs or immunocytokins) have been developed to overcome these limitations.2 ADCs are tripartite molecules consisting of a mAb onto which highly cytotoxic small molecules are conjugated by cleavable or non-cleavable linkers. They show better efficacy than the parent naked mAbs, due to the synergic effect of high selectivity for its antigen target and the potency of the highly cytotoxic drug.3 Drug conjugation is most frequently achieved via reactions on side chains of two different amino acids: lysine or cysteine (after reduction of the inter-chain disulfide bonds for the latter case).2 ADCs

are thus more complex than unconjugated mAbs, because of the increased inherent heterogeneity imparted by the addition of a controlled but variable number of druglinkers. Therefore, their structural characterization requires a combination of orthogonal analytical techniques that mostly rely on state-of-the-art chromatographic, electrophoretic and mass spectrometric methods.4 The analytical characterization of ADCs requires the investigation of several critical quality attributes (CQAs),5 such as the drug load distribution (DLD), the level of unconjugated antibody (D0), the average drug to antibody ratio (DAR),6 the site conjugation position of the cytotoxic drugs,7 the amounts of residual small molecule drugs (SMDs) and related products.8 The average DAR represents the number of linker-payloads that can be delivered to the tumor cell and thus is directly linked to the potency and the safety of the ADCs. The distribution of drug loads (i.e., the percentage of antibodies or subunits containing zero, one, two, three, n drugs) is also an important characteristic, since the different isoforms may have dif-

ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ferent aggregation, toxicological and pharmacological properties. In quality control (QC) laboratories, hydrophobic interaction chromatography (HIC) allows the separation of the individual DARs based on the number of attached drugs and their apparent hydrophobicity.6,9–12 HIC is considered as the reference technique for the analytical characterization of hinge cysteine-linked ADC (Cys-ADCs) and sitespecific ADCs, providing fast and easy evaluation of most CQAs. In many cases, the interpretation of the HIC-UV profile can be misleading due to the presence of overlapping or poorly resolved peaks, which are identified only based on HIC retention times (see Supporting Information Fig. S1 for examples of complex HIC profiles or Wiggins et al.13 for IgG2-based ADCs). Additional/overlapped/unresolved HIC peaks can account for either positional isomers of DAR species not distinguishable by native MS alone,6 DAR species with an odd number of cytotoxic drugs,14 or other type of variants and degradation products. On the other hand, MS analysis performed under nondenaturing conditions (also called native MS) has been reported as a valuable tool for the structural characterization of intact mAbs15 and ADCs.16–19 Due to its high mass accuracy, native MS offers structural insights into multiple ADC CQAs (DLD profiles, D0 and average DAR) within only a few minutes. Native MS data interpretation is also more straightforward than the combination of denaturing reversed-phase liquid chromatography (RPLC) and MS, as the harsh mobile phase conditions (acidic pH, high proportion of organic solvent, elevated mobile phase temperature) do not maintain weak non-covalent species in hinge Cys-ADCs,20 leading to indirect assessments of both DLD profiles and average DAR. However, native MS is not yet routinely used in QC environments, mostly because it is technically more complex, difficult to validate and requires strong expertise13 and, the information provided by native MS alone cannot allow the distinction among different positional isomers having the same masses. When dealing with complex ADC samples (Supporting Information Fig. S1), HIC fraction collection followed by manual desalting and subsequent native MS or IM-MS identification must be performed.16 However, such offline MS analysis is labor intensive, time consuming (multiple injections) and cannot be automatized, which is a considerable drawback in terms of throughput. Therefore, in the past few years, different groups have focused their efforts on the online coupling of HIC to MS. HIC is not inherently MS-compatible, due to the presence of non-volatile salts in the mobile phase, except when ammonium acetate at a concentration of 1M or less is used. Online HIC-MS hyphenation has been proposed by Chen et al. with such volatile mobile phase, but 50% acetonitrile was used in the mobile phase. Thus, this is not stricto senso non-denaturing conditions21, and such conditions do not allow keeping Cys-ADC structural integrity. Multidimensional 2D-LC strategies have also been

Page 2 of 15

reported for direct hyphenation of HIC to denaturing MS: HIC was used in the first dimension and RPLC in the second one, in either fully comprehensive14,22 or heartcutting8,23 modes. RPLC allows efficient removal of the non-volatile salts from the first dimension along with positional isomer relative abundance assessment in some cases.14 However, RPLC also involves classical organic and acidic solvents that lead to inter-chain cysteine-linked ADC sample denaturation,16,19 resulting in a more difficult characterization of the complex ADC samples and a denaturing coupling to MS. For all the above mentioned reasons, the online coupling of HIC (allowing to separate individual species and preserve the structure of the ADC) to non-denaturing MS would be of great interest for use in R&D laboratories, to assist and prepare complex HIC profile interpretations in QC labs. In this context, we present here a proof-of concept for the direct hyphenation of HIC to native MS or its hyphenation to ion mobility (IM-MS), by using a 2D-LC setup involving HIC in the first dimension and size exclusion chromatography (SEC) in the second one (i.e., HICxSEC), for ADCs characterization. EXPERIMENTAL SECTION Reagents and materials All chemicals were purchased from Sigma-Aldrich: ammonium acetate (A1542); cesium iodide (21004); phosphoric acid (345245); 2-propanol (I9516); sodium chloride (S7653); sodium phosphate monobasic (S8282); sodium phosphate dibasic (S7907). All the aqueous solutions were prepared using an ultra-pure water system (Sartorius, Göttingen, Germany). Brentuximab vedotin (Adcetris) was from Takeda. IgGZERO (A0-IZ1-010) enzyme was obtained from Genovis. Preparation of deglycosylated and forced degraded ADC samples ADC deglycosylation was performed by incubating for 30 min at 37°C one unit of IgGZERO per microgram of ADC. For forced degraded studies, a vial of 50 mg freeze-dried brentuximab vedotin was reconstituted under aseptic conditions with 10 ml water to yield a 5 mg/ml solution. Next, 1 ml of the solution was transferred in a sterile 1.5 ml tube under a laminar flow hood24 and incubated at 40°C for 4 weeks. Instrumentation The LCxLC-IMxMS system consists of a combination of H-Class and I-Class liquid chromatography systems hyphenated to a Synapt G2 HDMS Q-TOF mass spectrometer, both from Waters (Manchester, UK). In the first dimension, the H-Class system includes a high-pressure quaternary solvent delivery pump, an autosampler with a flow-through needle of 15 µL equipped with an extension loop of 50 µL. In the second dimension, the I-class system includes a high-pressure binary solvent delivery pumps, a column manager composed of two independent column ovens and two 6-port high pressure two-position valves acting as interface between the two chromatographic


Page 3 of 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


separation dimensions. A single wavelength ultraviolet (UV) detector and a diode array detector both equipped with 500 nL flow-cell were used for the first and second dimension, respectively. The dwell volumes were about 425 µL and 300 µL for the first and the second dimensions, respectively. It should be noted that the dwell volume of the second dimension includes the volume of sample loops used at the interface (i.e., 200 µL). Measured extra-column volumes were 12 µL and 17 µL for the first and second dimensions, respectively. An external two position switching valve (Vici Valco Instruments, Houston, USA) was also placed prior to the mass spectrometer. Non-denaturing MS and ion mobility experiments were performed on a TWIMS-MS Synapt G2 HDMS instrument (Waters, Manchester, UK). Data acquisition and instrument control were performed with MassLynx V4.1 software (Waters). 2D-data (UV and TIC) were exported to MATLAB V7.12.0635 to construct 2D-contour plots via house-made calculation routines. MS-Data were processed with MassLynx 4.1. Chromatographic conditions The HIC column was a MAbPac HIC-10 (100 mm x 4.6 mm, 5 µm, 1000 Å) from Thermo Scientific, Cheshire, UK. The SEC column employed in the second dimension was an AdvanceBio SEC (50 mm x 4.6 mm, 2.7 µm, 300 Å) from Agilent Technologies, Wilmington, DE, USA. For the HIC first dimension, the mobile phase A was composed of 2.5 M of ammonium acetate and 0.1 M phosphate buffer (Na2HPO4), pH 7.0 (adjusted with phosphoric acid), while the mobile phase B was composed of 0.1 M phosphate buffer (Na2HPO4 and NaH2PO4) with pH 7.0 (adjusted with sodium hydroxide solution). The following gradient was employed in HIC: 0 to 90% B in 36 min, 90 to 100% B in 21 min, followed by an isocratic step at 100% B for 8 min before re-equilibration for 25 min, for a total analysis time of 90 min. The HIC experiment was conducted at a flow-rate of 100 µL/min. Column temperature, wavelength and data acquisition rate were set at 30°C, 280 nm and 10 Hz, respectively. The injection volume was 40 µL. For the second chromatographic dimension in SEC, the separation was carried out in isocratic mode with an aqueous mobile phase composed of 100 mM ammonium acetate at a flow-rate of 700 µL/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, namely 1.5 min. A fraction of 0.45 min (from 0.42 to 0.87 min) was sent to MS thanks to a switching valve, to limit salt contamination of the electrospray ionization (ESI) source. A flow splitter divided the flow-rate by a factor 7 prior to entering MS (i.e., inlet flow of 100 µL/min). HICxSEC-MS in non-denaturing conditions The Synapt G2 HDMS was operated in sensitive mode and positive polarity with a capillary voltage of 3.0 kV. To avoid disruption of weak non-covalent interactions, the sample cone and pressure in the interface region were set

to 120 V and 6 mbar, respectively. Source and desolvation temperature were set to 100 and 450°C, respectively. Desolvation and cone gas flows were set at 750 and 60 L/hr, respectively. Acquisitions were performed in the m/z range of 1000-10000 with a 1.5 s scan time. External calibration was performed using singly charged ions produced by a 2 g/L solution of cesium iodide in 2propanol/water (50/50 v/v). To get more accurate mass measurements, masses reported in Table 1 were obtained at 160 V, which is not compatible with accurate TWCCSN2 measurements. HICXSEC-IMxMS in non-denaturing conditions For IM-MS measurements, the parameters were optimized with respect to the ion transmission and resolution in the drift cell, with as little gas-phase unfolding as possible. Especially, the sample cone voltage was set to 100 V and the backing pressure of the source was 6 mbar. The Ar flow rate was 5 mL/min and the trap collision energy was set at 4 V in the traveling-wave-based ion trap. Trap DC entrance, trap DC Bias, and trap DC exit were operating at 0 V, 60 V, and 1 V respectively, avoiding extensive ion activation. Ions were focused with a constant He flow rate of 130 mL/min before IM separation. The height and the velocity of the periodic waveform in the pressurized ion mobility cell were 40 V and 923 m/s, respectively. N2 was used as drift gas (45 mL/min) providing a constant pressure of 2.75 mbar. Transfer collision energy was fixed to 2 V to extract the ions from the IM cell to the TOF analyzer. A SEC-IMxMS calibration based on three different proteins (concanavalin A, pyruvate kinase and alcohol dehydrogenase) in non-denaturing conditions was used to perform collision cross-section (CCS) calculation as previously described.25,26 The three calibrants were individually injected through the SEC-IMxMS set-up (same set-up as for HICxSEC-IMxMS, but the HIC column was disconnected) to shorten the analysis time (run duration 8 min instead of 90 min for the complete HICxSEC-IMxMS run) at a concentration of 30 µM. SEC was operated in isocratic mode with an aqueous mobile phase composed of 100 mM ammonium acetate at a flow rate of 100 µL/min during 8 min (AdvanceBio SEC 50 mm x 4.6 mm, 2.7 µm, 300 Å from Agilent Technologies, Wilmington, DE, USA). The cone voltage was set to 80 V to avoid ion activation. IM parameters were the same as those used for HICxSEC-IMMS described above, which provides a reliable IM calibration method for CCS measurements in stressed and unstressed BV samples. Average DAR calculation The average DAR value represents the sum of relative peak area for each DAR multiplied by its corresponding number of drugs. It was calculated from HIC chromatogram, using Equation 1. ∑ ∙ (1) = ∑ Where k is the number of drugs and the HIC peak area of DARk.



Average DAR values from native MS were calculated from the relative peak intensities measured either from the raw mass spectrum (25+ to 29+ charge states). In Equation 1, Ak is then the relative peak intensity of DARk. For DAR calculations from native IM-MS, charge state intensities for each species (25+ to 29+) were extracted and fitted to a Gaussian distribution. The areas underneath each of these curves were quantified and used to calculate average DAR with Equation 1 (Ak is then the Gaussian peak area intensity of DARk). RESULTS Optimization of HICxSEC conditions The first point to be addressed for coupling HIC to native MS was the need to eliminate the salts employed in HIC. First, rather than employing ammonium sulfate as the major mobile phase component, ammonium acetate was selected here, as it is volatile and MS compatible. In addition, it was recently demonstrated that a highly similar HIC separation can be obtained with both ammonium sulfate and ammonium acetate, assuming that the amount of salt was adjusted.10 Besides HIC mobile phase modification, an additional SEC dimension was added as a fast online desalting27 before MS inlet (see Fig S2 in supporting information). SEC was selected in this set-up because it allows a sizebased separation of the species, that allow differentiation of the HIC salts (low molecular weight species of 150,000 Da). In this configuration, SEC was not intended for separating ADC high molecular weight aggregates, but exclusively used as a “fast desalting” step.

Figure 1. Effect of injection volume on the size exclusion chromatography (SEC) separation of brentuximab vedotin and ammonium acetate: (a) 1µL, 10µL and 50µL of a sample containing 2.5M ammonium acetate and brentuximab vedotin 0.4mg/mL, injected in 1D-SEC; (b) hydrophobic interaction chromatography (HIC)-fraction of 150 µL of brentuximab vedotin injected in 2D-SEC. Vertical lines delimit the SEC-fraction sent to mass spectrometry (MS). Other conditions are given in the experimental section. To achieve HICxSEC hyphenation, the optimization of 2D-LC conditions was required. Very fast SEC separation was achieved with a short 50 x 4.6 mm SEC column at a 0.7 mL/min flow-rate, corresponding to the maximum recommended inlet pressure. The stationary phase was found to minimize non-specific interactions with the

Page 4 of 15

hydrophobic ADC species28 while providing sufficient resolving power to discriminate HIC salts and ADC samples in the case of injection volumes as low as 1µL(Figure 1a). However, in online 2D-LC, much larger injection volumes are required since the two dimensions are linked by the sampling time. As a result, this latter must be long enough to achieve the separation in the second dimension (2D) and short enough to send a sufficient number of fractions in 2D. As shown in Figure 1a, strong peak distortion (broadening and tailing) was observed for both species injected at 10 and 50 µL, resulting in resolution decrease and separation time increase. In our HIC conditions, the minimum resolution required is 1.5. To maintain a sufficient separation between all pairs of peaks, the loss in resolution should not exceed 30% (i.e., Rs>1). According to the relationship established by Davis et al. to correct the resolution from under-sampling,29 a loss of 30% corresponds to a sampling rate of 2.7. Considering 1D peak widths (i.e., about 400 µl) obtained with the optimized HIC gradient, fractions of 150 µL or less must be sent in the second dimension. In case of 150 µL, a 2D analysis time of 1.5 min was required to achieve the complete elution of the salts, leading to a 1D flow-rate of 100 µL/min (i.e., 150/1.5). Therefore, a sampling time of 1.5 min and a 1 D-flow-rate of 100 µL/min were selected, with 150 µL injected in the second dimension. The resulting separation of a HIC fraction of the reference cysteine-linked ADC, brentuximab vedotin (BV) on the SEC column is shown in Figure 1b. As reported, the baseline resolution between the two peaks is not obtained in these conditions, due to the strong injection effects. Despite this, the fraction with no salts (delimited by the two vertical lines in Figure 1b) could be sent to MS by using a two-positions switching valve located between the SEC column outlet and the MS inlet. The rest of the SEC separation was sent to the waste (supporting information Fig S2). Rather than a multiple heart-cutting methodology, a fully comprehensive online HICxSEC approach was considered in this study, thereby allowing the whole HIC separation to be tracked with MS. The optimized HICxSEC method was then hyphenated with high resolution native MS (threedimensional (3D) approach), to achieve the identification of all the peaks observed in HIC, without any denaturation occurring. Finally, ion mobility (IM) was activated simultaneously to MS to have a conformational characterization of each HIC peak (four-dimensional (4D) approach). An overall schematic workflow of the fully comprehensive online HICxSEC-IMxMS strategy is provided in Figure 2. This configuration maintained the integrity of HIC 1D-separation, while providing a global MS picture within a single run. Application of HICxSEC-IMxMS to the reference Cys-ADC As a proof-of-concept, we first analyzed brentuximab vedotin (BV, Adcetris, reference Cys-ADC) by using the 4D HICxSEC-IMxMS setup (Figure 2). The optimized 1DHIC chromatogram (Figure 3a) was obtained without any sample preparation (no deglycosylation, no buffer exchange). Each 1.5 min, a HIC fraction was sent to 2D-SEC


Page 5 of 15

enabling online, continuous, fast buffer exchange of each individual HIC fraction. All SEC fractions were then continuously infused into MS equipped with an IM separator via an ESI source operating under non-denaturing conditions.

Table 1. Summary of experimental values for the different drug to antibody ratios (DARs) of non-stressed and stressed antibody drug conjugates (ADC). Molecular weights were obtained from mass spectra, experimental collision crosssection (TWCCSN2), hydrophobic interaction chromatography (HIC) peak area and average DAR calculations. To get more accurate mass measurements in HICxSEC-MS mode, a Vc voltage of 160 V was used while less energetic conditions were used when IM-MS was turned on to measure TWCCSN2 values (Vc = 100V). Abbreviations are defined as ND for not detected. Experimental masses of DARs were compared to the theoretical ones. Experimental collision cross-sections obtained from centroid ion mobility (IM) drift times were compared with predicted CCSN2 calculated through the equation CCS=2.435*M2/3 for spherical proteins.30

Non-stressed

DAR (G1F/G0F) Theoretical mass (Da) Experimental mass (Da) Mass accuracy (ppm) TW

CCSN2 (nm2)

Pred CCS (nm2) HIC peak area (%) Native MS (%) IM (%) Experimental mass (Da) Mass accuracy (ppm) TW

Stressed

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


CCSN2 (nm2)

Pred CCS (nm2) HIC peak area (%) Native MS (%) IM (%)

D0

D1

D2

D3

D4

D5

D6

D7

D8

148242

149559

150877

152195

153512

154830

156148

157465

158783

148259 ±6

149756 * Average mass

150901 ±3

152479 * Average mass

153538 ± 7

ND

156169 ± 8

ND

158800 ± 6

116

-

157

-

166

-

137

-

106

78.4 ± 0.1

-

78.9 ± 0.1

-

79.2 ± 0.1

-

79.5 ± 0.1

-

79.9 ± 0.1

68.2

-

69.0

-

69.8

-

70.6

-

71.4

5.8

1.8

24.1

4.3

32.8

ND

21.5

ND

9.7

5.9

1.9

23.9

4.4

33.5

ND

23.3

ND

7.1

6.8

ND

24.2

ND

38.8

ND

22.8

ND

7.4

148257 ±7

149588 ±5

150900 ±5

152221 ± 25

153535 ± 40

ND

156167 ± 21

ND

158818 ± 17

105

193

151

171

150

-

124

-

222

78.3 ± 0.1

78.3 ± 0.1

78.9 ± 0.1

79.2 ± 0.1

79.3 ± 0.1

-

79.6 ± 0.1

--

80.0 ± 0.1

68.2

68.6

69.0

69.4

69.8

-

70.6

-

71.4

13.5

10.6

25.9

7.3

33.4

-

8.2

-

1.1

10.5

9.2

26.2

7.6

35.0

ND

9.5

ND

2.0

9.6

11.1

27.1

6.5

31.2

ND

12.1

ND

2.4


Average DAR 4.0 Average DAR 3.9 Average DAR 4.0

Average DAR 2.8 Average DAR 3.0 Average DAR 3.0


Page 6 of 15

Figure 2. Flowchart of the analysis for brentuximab vedotin

Figure 3. Online HICxSEC-IMxMS of brentuximab vedotin. (a) Hydrophobic interaction chromatography (HIC) profile and (b) ion mobility mass spectrometry (IM-MS) characterization. For each individual drug to antibody ratio (DAR), (b, left panel) a zoom of the 13−18 ms td region of the driftscope plots, (b, middle panel) individual arrival time distributions (ATDs) corresponding to the 27+ charge state, and (b, right panel) deconvoluted native mass spectra were represented. Data for low abundant D1 and D3 species are reported in the Supporting Information Fig.S4.


Page 7 of 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


As expected, five main peaks were detected after the 1DHIC separation (Figure 3a), which were all unambiguously identified by non-denaturing MS intact mass measurement (Figure 3b right panels and Table 1). Considering the short acquisition time of only 0.45 min, all detected species were identified as drug load species (D0 to D8) with a remarkable mass accuracy of 100-160 ppm, which is classical for intact ADC analysis in non-denaturing conditions on quadrupole time-of-flight (Q-TOF) instruments.16,17 Even minor HIC species like D0 or D8 were clearly identified. For BV, an average DAR of 4.0 was unambiguously determined (Table 1) in agreement with expected values.16,31 To add an online additional dimension for conformational characterization, the ion mobility of the mass spectrometer was then turned on, resulting in a comprehensive 2D-IM-MS analysis (Figure 3b, left panels). Arrival time distributions (ATDs) of each individual species were isolated in the IM cell, allowing conformational characterization through collisional cross section (TWCCSN2) calculations of each HICxSEC separated drug load species (Figure 3b, middle panels). IM analyses led to the determination of the conformational homogeneity of each individual HIC fraction, as already reported for offline HIC-IMxMS analysis in non-denaturing conditions.16 Figure 3b (left panels) presents the IMxMS contour plots of each individual HIC-detected species under IM conditions in which ion activation was limited (see Experimental Section). A good agreement was obtained between TW CCSN2 values obtained by HICxSEC-IMxMS and those obtained by offline injection of manually desalted BV (Table 1), demonstrating that the HICxSEC configuration does not affect the BV global conformation (see Supporting Information Fig.S3). Altogether, our results demonstrate for the first time the ability to have a comprehensive analytical characterization of a cysteine-linked ADC within a single run, affording: i) simultaneous drug load profile and quantitative average DAR assessment (HIC); ii) the unambiguous identification of the number of drug conjugations through accurate intact mass measurement in non-denaturing conditions (native MS); and iii) conformational homogeneity assessment of each drug load species (native IM). Benefits of HICxSEC-IMxMS standalone HIC or native MS

compared

to

The benefits of our HICxSEC-IMxMS workflow compared to HIC or native MS used as standalone techniques are highlighted, through the examples of deglycosylated BV (Figure 4) and an in-house ADC under development (supporting information Fig.S5). In strictly identical chromatographic conditions, deglycosylated BV (Figure 4a) presents a far more complex HIC chromatogram than native BV, with three main peaks that can be attributed respectively to D0, D2 and D4 (comparison with BV HIC retention times), and four minor peaks (labeled A, B, C and D on Figure 4a)) that cannot be identified based on the solely HIC retention times, and thus preventing the accurate average DAR calculation. Similarly, Figure 4b presents the native MS profile of deglycosylated BV, revealing a very similar native MS profile as the one ob-

tained for native BV (see supporting information Fig.S6) enabling an average DAR calculation of 4.0, in agreement with already published data.16 However, HIC and native MS profiles poorly correlate in this example, with lesser species being observed on the native mass spectrum as compared to the HIC chromatogram. Online HICxSEC-IMxMS was then applied, allowing mass-based unambiguous identification of the HIC minor peaks that were identified as D4 variants (peaks A and B), D6 (peak C) and D8 (peak D), respectively (Figure 4c). Finally, the online HICxSEC-IMxMS allowed the accurate average DAR calculation of 4.1. It should be noted that although HIC separation allows the differentiation of D4 positional isomers, very similar CCS values, within the instrumental error, were calculated in native IM-MS, emphasizing the fact that current IM cells still lack the required resolution for a correct positional isomers identification at the intact mAb level.16,17 On a side note, D6 species were surprisingly observed to elute before D4 ones, as already observed by Gilroy and Eakin by comparing HIC profiles of a glycosylated versus deglycosylated cysteine ADC (Gilroy and Eakin 2017). It is also interesting to notice a significant signal-to-noise enhancement on the mass spectra obtained with HICxSEC-IMxMS (Fig.4c) compared to native MS alone (Fig.4b), especially for low abundant D0 and D8 species. That might originate from the reduction of both matrix effects and competing ionization efficiencies with HIC separation prior to MS, which further underlines the value of our multidimensional approach. The benefits of our analytical setup are also shown in Supporting Information Fig.S5, by using an investigational ADC with a target average DAR of 4.0. Again, more species were detected on the HIC chromatogram than on the native mass spectrum, which could only be unambiguously identified by online HICxSEC-IMxMS. Application of HICxSEC-IMxMS to a forced degraded sample To highlight the possibilities and relevance of this new 4D approach involving chromatographic and mass spectrometric methods, we next compared intact BV with temperature stressed BV samples (see material and methods). Of note, our HICxSEC-IMSxMS configuration is adapted for the investigation of the forced degradation effect on drug load distribution, but not for aggregate formation. HICxSEC contour plots of stressed and non-stressed BV can be compared in Figures 5a and 5b, respectively. For stressed BV, the contour plot clearly highlights an increase in the intensities of species eluted between D0 and D2, and between D2 and D4. In addition, both D6 and D8 intensities significantly decrease, while the one of D0 increases (Table 1 and comparison of Figures 5a and 5b). The direct coupling of online HICxSEC to non-denaturing MS allowed accurately measuring the molecular weights of all the compounds, and then to unambiguously identify D1 and D3 as species formed under thermal stress (Figures 5c and 5d). In unstressed sample, only an averaged mass of all glycoforms was measured for D1 and D3, due to the very low MS signal intensity hampering accurate mass



measurements. As D1 and D3 became more abundant in the stressed sample, accurate mass measurement of the most intense glycoforms was possible (Table 1). From MS identification of all detected species, the average DAR value, calculated from HIC data, was reduced to 2.8 in stressed conditions compared with 4.0 (Table 1), which is in agreement with previously published data.14 Simultaneously, online non-denaturing IM analysis allowed direct and unambiguous conformational characterization of odd DAR species (D1 and D3) through their TWCCSN2 calculations (Table 1). In addition, comparison of normalized extracted ATDs of unstressed versus stressed BV clearly highlights the intensity increase of odd DAR species in forced degradation conditions (Figures 5c and 5d). Altogether, these results provide a first proof-of-concept for using a comprehensive 4D HICxSEC-IMxMS methodology to gain structural insights of hinge cysteine-linked ADCs.

Figure 4. Benefits of online HICxSEC-IMxMS compared to HIC and native MS as standalone techniques for deglycosylated brentuximab vedotin. (a) Hydrophobic interaction chromatography (HIC) profile, (b) deconvoluted native mass spectrum and (c) deconvoluted native mass spectra obtained by online HICxSEC-IMxMS of each minor drug to antibody ratio (Dn) species. HICxSEC contour plots of stressed and non-stressed BV can be compared in Figures 5a and 5b, respectively. For stressed BV, the contour plot clearly highlights an increase in the intensities of species eluted between D0 and D2, and between D2 and D4. In addition, both D6 and D8 intensities significantly decrease, while the one of D0

Page 8 of 15

increases (Table 1 and comparison of Figures 5a and 5b). The direct coupling of online HICxSEC to non-denaturing MS allowed accurately measuring the molecular weights of all the compounds, and then to unambiguously identify D1 and D3 as species formed under thermal stress (Figures 5c and 5d). In unstressed sample, only an averaged mass of all glycoforms was measured for D1 and D3, due to the very low MS signal intensity hampering accurate mass measurements. As D1 and D3 became more abundant in the stressed sample, accurate mass measurement of the most intense glycoforms was possible (Table 1). From MS identification of all detected species, the average DAR value, calculated from HIC data, was reduced to 2.8 in stressed conditions compared with 4.0 (Table 1), which is in agreement with previously published data.14 Simultaneously, online non-denaturing IM analysis allowed direct and unambiguous conformational characterization of odd DAR species (D1 and D3) through their TWCCSN2 calculations (Table 1). In addition, comparison of normalized extracted ATDs of unstressed versus stressed BV clearly highlights the intensity increase of odd DAR species in forced degradation conditions (Figures 5c and 5d). Altogether, these results provide a first proof-of-concept for using a comprehensive 4D HICxSEC-IMxMS methodology to gain structural insights of hinge cysteine-linked ADCs.

Figure 5. 2D-HICxSEC contour plots of (a) brentuximab vedotin and (b) forced degraded brentuximab vedotin. Ion mobility mass spectrometry (IM-MS) structural characterization of (c-d) odd drug to antibody ratio (DAR) species of stressed (black lines) and not stressed (grey lines) brentuximab vedotin. Individual arrival time distributions (ATDs) corresponding to the 27+ charge state (cd, left panels), and native mass spectra (c-d, right panels) of (c) DAR1 and (d) DAR3 were represented. CONCLUSIONS We have presented here an innovative multidimensional analytical approach combining comprehensive online two-dimensional chromatography (HICxSEC) to ion mobility and mass spectrometry (IM-MS) for performing structural characterization of Cys-ADCs under nondenaturing conditions. This setup offers in-depth and detailed information on ADC samples that cannot be reached from HIC-UV or non-denaturing MS alone. Indeed, when the HIC profile is complex (this is often the


Page 9 of 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


case), due to the presence of positional isomers, odd DARs or degradation products, it becomes very difficult to have an unambiguous interpretation of all the HIC peaks, only based on the chromatographic method. Native MS could be a successful strategy for such complex ADCs. However, native MS used as a standalone approach cannot afford the characterization of positional isomers of identical masses and also offer limited sensitivity for the less abundant species. For all the above mentioned reasons, the direct coupling of HIC to native MS offers some clear and unambiguous benefits in biopharma laboratories. The proposed 4D HICxSEC-IMxMS approach addresses several analytical challenges, including the LC-MS characterization of engineered antibody constructs, and the online coupling of HIC with native MS. As illustrated in this work, several ADC CQAs required for process and formulation development, routine lot-characterization, and stability testing could be monitored within one unique run. The proposed 4D setup affords: i) simultaneous DLD profiles and quantitative average DAR assessment (HIC); ii) unambiguous identification of the number of drug conjugations, through accurate intact mass measurement in non-denaturing conditions (native MS); and iii) conformational homogeneity assessment of each drug load species (native IM). Having online native IM opens in theory the way to tackle positional isomer (separated by HIC) issues at the intact ADC level. However, resolving powers of current IM cells are too limited to address this point. Altogether, the described 4D approach (HICxSECIMxMS) could be proposed as the first multi-attribute method (MAM) 32,33 in non-denaturing conditions for intact cysteine-linked ADC characterization, combining the power of chromatographic separation to the specificity of MS identification. The all-in-one analytical strategy described here permits a deep, straightforward and rapid characterization of complex ADC samples. In comparison with the heartcutting (or multiple heart-cutting) procedure, much more exhaustive information can be reached on the sample. For example, both DAR1 and DAR3 species could be tracked with our comprehensive 2D-approach, while this would probably not be the case with a heart-cutting approach. Also, positional isomers can be directly mass identified. Furthermore, heart-cutting approaches usually require multiple injections (one injection per peak identification), whereas all the information is available in one single run in the presented comprehensive 2D-strategy.

Author contributions ‡These authors contributed equally.

Funding sources This work was supported by the CNRS, the Université de Strasbourg, the Université de Lyon, the Agence Nationale de la Recherche (ANR) and the French Proteomic Infrastructure (ProFI; ANR-10-INBS-08-03), the Swiss National Science Foundation (fellowship 31003A_159494).

ACKNOWLEDGMENTS This work was also supported by the “Agence Nationale de la Recherche” (ANR) and the French Proteomic Infrastructure (ProFI; ANR-10-INBS-08-03). The authors thank GIS IBiSA and Région Alsace for financial support in purchasing a Synapt G2 HDMS instrument. A.E. acknowledges the «Association Nationale de la Recherche et de la Technologie » (ANRT) and Syndivia for funding his PhD fellowship. The authors would like to thank David Lascoux (Waters) for his kind support and valuable assistance. The authors acknowledge Liz Bevan and Tony Edge from Agilent Technologies, for providing the AdvanceBioSEC column. Finally, the authors wish to thank Waters for the loan of 2D-LC instrumentation used in this work.

REFERENCES (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

(12) (13) (14)

ASSOCIATED CONTENT

(15)

Supporting information

(16)

The Supporting Information S1-S6 is available free of charge on the ACS Publications website.

(17)

AUTHOR INFORMATION Corresponding authors

(18)

* [email protected] * [email protected]

(19)

Carter, P. J.; Lazar, G. A. Nat. Rev. Drug Discov. 2017. DOI: 10.1038/nrd.2017.227. Beck, A.; Goetsch, L.; Dumontet, C.; Corvaïa, N. Nat. Rev. Drug Discov. 2017, 16, 315–337. Ornes, S. Proc. Natl. Acad. Sci. 2013, 110, 13695–13695. Fekete, S.; Guillarme, D.; Sandra, P.; Sandra, K. Anal. Chem. 2016, 88, 480–507. Wakankar, A.; Chen, Y.; Gokarn, Y.; Jacobson, F. S. MAbs 2011, 3, 161–172. Le, L. N.; Moore, J. M. R.; Ouyang, J.; Chen, X.; Nguyen, M. D. H.; Galush, W. J. Anal. Chem. 2012, 84, 7479–7486. Janin-Bussat, M.-C.; Dillenbourg, M.; Corvaia, N.; Beck, A.; Klinguer-Hamour, C. J. Chromatogr. B 2015, 981–982, 9–13. Birdsall, R. E.; McCarthy, S. M.; Janin-Bussat, M. C.; Perez, M.; Haeuw, J.-F.; Chen, W.; Beck, A. MAbs 2016, 8, 306–317. Rodriguez-Aller, M.; Guillarme, D.; Beck, A.; Fekete, S. J. Pharm. Biomed. Anal. 2016, 118, 393–403. Cusumano, A.; Guillarme, D.; Beck, A.; Fekete, S. J. Pharm. Biomed. Anal. 2016, 121, 161–173. Ouyang, J. In Antibody-Drug Conjugates; Ducry, L., Ed.; Methods in Molecular Biology; Humana Press: Totowa, NJ, 2013; Vol. 1045, pp 275–283. Haverick, M.; Mengisen, S.; Shameem, M.; Ambrogelly, A. MAbs 2014, 6, 852–858. Rogstad, S.; Faustino, A.; Ruth, A.; Keire, D.; Boyne, M.; Park, J. J. Am. Soc. Mass Spectrom. 2017, 28, 786–794. Sarrut, M.; Fekete, S.; Janin-Bussat, M.-C.; Colas, O.; Guillarme, D.; Beck, A.; Heinisch, S. J. Chromatogr. B 2016, 1032, 91–102. Beck, A.; Sanglier-Cianférani, S.; Van Dorsselaer, A. Anal. Chem. 2012, 84, 4637–4646. Debaene, F.; Bœuf, A.; Wagner-Rousset, E.; Colas, O.; Ayoub, D.; Corvaïa, N.; Van Dorsselaer, A.; Beck, A.; Cianférani, S. Anal. Chem. 2014, 86, 10674–10683. Marcoux, J.; Champion, T.; Colas, O.; Wagner-Rousset, E.; Corvaïa, N.; Van Dorsselaer, A.; Beck, A.; Cianférani, S. Protein Sci. 2015, 24, 1210–1223. Botzanowski, T.; Erb, S.; Hernandez-Alba, O.; Ehkirch, A.; Colas, O.; Wagner-Rousset, E.; Rabuka, D.; Beck, A.; Drake, P. M.; Cianférani, S.; Cianferani, S. MAbs 2017, 9, 801–811. Valliere-Douglass, J. F.; McFee, W. A.; Salas-Solano, O.



(20) (21) (22)

(23) (24) (25)

(26)

Anal. Chem. 2012, 84, 2843–2849. Fekete, S.; Veuthey, J.-L.; Beck, A.; Guillarme, D. J. Pharm. Biomed. Anal. 2016, 130, 3–18. Chen, B.; Peng, Y.; Valeja, S. G.; Xiu, L.; Alpert, A. J.; Ge, Y. Anal. Chem. 2016, 88, 1885–1891. Sarrut, M.; Corgier, A.; Fekete, S.; Guillarme, D.; Lascoux, D.; Janin-Bussat, M.-C.; Beck, A.; Heinisch, S. J. Chromatogr. B 2016, 1032, 103–111. Birdsall, R. E.; Shion, H.; Kotch, F. W.; Xu, A.; Porter, T. J.; Chen, W. MAbs 2015, 7, 1036–1044. Sanders, E. R. J. Vis. Exp. 2012, 63, 1–12. Salbo, R.; Bush, M. F.; Naver, H.; Campuzano, I.; Robinson, C. V.; Pettersson, I.; Jørgensen, T. J. D.; Haselmann, K. F. Rapid Commun. Mass Spectrom. 2012, 26, 1181–1193. Bush, M. F.; Hall, Z.; Giles, K.; Hoyes, J.; Robinson, C. V.; Ruotolo, B. T. Anal. Chem. 2010, 82, 9557–9565.

(27) (28) (29) (30) (31) (32) (33)

Page 10 of 15 Waitt, G. M.; Xu, R.; Wisely, G. B.; Williams, J. D. J. Am. Soc. Mass Spectrom. 2008, 19, 239–245. Goyon, A.; Beck, A.; Colas, O.; Sandra, K.; Guillarme, D.; Fekete, S. J. Chromatogr. A 2017, 1498, 80–89. Davis, J. M.; Stoll, D. R.; Carr, P. W. Anal. Chem. 2008, 80, 461–473. Ruotolo, B. T.; Benesch, J. L. P.; Sandercock, A. M.; Hyung, S.-J.; Robinson, C. V. Nat. Protoc. 2008, 3, 1139–1152. Sievers, E. L.; Senter, P. D. Annu. Rev. Med. 2013, 64, 15–29. Rogers, R. S.; Nightlinger, N. S.; Livingston, B.; Campbell, P.; Bailey, R.; Balland, A. MAbs 2015, 7, 881–890. Rogers, R. S.; Abernathy, M.; Richardson, D. D.; Rouse, J. C.; Sperry, J. B.; Swann, P.; Wypych, J.; Yu, C.; Zang, L.; Deshpande, R. AAPS J. 2018. DOI: 10.1208/s12248-017-01683.


Page 11 of 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Effect of injection volume on the size exclusion chromatography (SEC) separation of brentuximab vedotin and ammonium acetate: (a) 1µL, 10µL and 50µL of a sample containing 2.5M ammonium acetate and brentuximab vedotin 0.4mg/mL, injected in 1D-SEC; (b) hydrophobic interaction chromatography (HIC)fraction of 150 µL of brentuximab vedotin injected in 2D-SEC. Vertical lines delimit the SEC-fraction sent to mass spectrometry (MS). Other conditions are given in the experimental section. 85x50mm (300 x 300 DPI)



Flowchart of the analysis for brentuximab vedotin 178x140mm (300 x 300 DPI)


Page 12 of 15

Page 13 of 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Online HICxSEC-IMxMS of brentuximab vedotin. (a) Hydrophobic interaction chromatography (HIC) profile and (b) ion mobility mass spectrometry (IM-MS) characterization. For each individual drug to antibody ratio (DAR), (b, left panel) a zoom of the 13−18 ms td region of the driftscope plots, (b, middle panel) individual arrival time distributions (ATDs) corresponding to the 27+ charge state, and (b, right panel) deconvoluted native mass spectra were represented. Data for low abundant D1 and D3 species are reported in the Supporting Information Fig.S4. 371x372mm (300 x 300 DPI)



Benefits of online HICxSEC-IMxMS compared to HIC and native MS as standalone techniques for deglycosylated brentuximab vedotin. (a) Hydrophobic interaction chromatography (HIC) profile, (b) deconvoluted native mass spectrum and (c) deconvoluted native mass spectra obtained by online HICxSEC-IMxMS of each minor drug to antibody ratio (Dn) species. 192x375mm (300 x 300 DPI)


Page 14 of 15

Page 15 of 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


2D-HICxSEC contour plots of (a) brentuximab vedotin and (b) forced degraded brentuximab vedotin. Ion mobility mass spectrometry (IM-MS) structural char-acterization of (c-d) odd drug to antibody ratio (DAR) species of stressed (black lines) and not stressed (grey lines) brentuximab vedotin. Individual arrival time distributions (ATDs) corresponding to the 27+ charge state (c-d, left panels), and native mass spectra (c-d, right panels) of (c) DAR1 and (d) DAR3 were represented. 178x119mm (300 x 300 DPI)


An online four-dimensional HICxSEC-IMxMS methodology for proof-of

Recommend Documents