Aaron O. Bailey1, Bobby Gamez1, Terry Zhang1, Stephane Houel1, Shanhua Lin2, Guanghui Han3, Katie Southwick1, Wendy Sandoval3, Jonathan L. Josephs1. 1Thermo Fisher Scientific, 355 River Oaks Parkway, San Jose, CA, USA, 95134; 2Thermo Fisher Scientific, 1228 Titan Way, Sunnyvale, CA, USA, 94085; 3Department of Microchemistry, Proteomics and Lipidomics, Genentech Inc., South San Francisco, CA, USA, 94080

MODERATE TOLERANCE SLIDING WINDOW DECONVOLUTION (15 ppm) ALLOWS SENSITIVE AND CONFIDENT DETECTION OF LOW LEVEL CQA ISOFORMSConfident detection of low-level charge variant CQAs on intact mAbNative IEC-MS analysis of intact trastuzumab resulted in a chromatographic profile consistent with previously published data. Time-resolved deconvolution analysis of the innovator drug resulted in identification of several specific isoforms comprised of differential N-glycosylation in combination with a variety of low level charge-imparting PTMs, such as deamidation, C-terminal lysine removal, or sialylation of N-glycans. We confirmed that the main peak consisted of several isoforms corresponding to variable N-glycosylations of the trastuzumab amino acid sequence. When comparing the innovator and biosimilar products, we found marked differences in glycoform profile measured at the main peak. Furthermore we found measureable differences in several charge variant isoforms when comparing innovator and biosimilar. These differences observed were consistent with a comparison of these same two samples via MAM HRAM peptide mapping analysis (data not shown).

ABSTRACTPurpose: To demonstrate feasibility of workflow for monitoring multiple critical quality attributes of a therapeutic protein drug as analyzed in the intact form.

Methods: We analyzed proteins using pH-elution ion exchange coupled directly to native ESI-MS. Isoform identification was accomplished using Thermo Scientific™ BioPharma Finder™ software and XIC-based isoform relative abundance measurements were performed using Thermo Scientific™ Chromeleon™ software.

Results: We demonstrate that ion exchange native MS provides high precision intact mass measurement of Herceptin® and a trastuzumab biosimilar and that Chromeleon software may be used to measure the XIC-based relative abundances of individual CQA isoforms.

INTRODUCTIONManufacturing of innovator biologics can be successfully mimicked to produce generic “biosimilar” drug products. In order to satisfy safety and efficacy requirements, a biosimilar drug must be reasonably comparable to an innovator. Comparability is directly assessed by measuring a panel of critical quality attributes (CQAs). Recently, a multi-attribute method has been demonstrated to measure several CQAs simultaneously using LC-MS peptide mapping data. Additional CQAs, such as charge variants or size variants (i.e., aggregation/fragmentation), are conventionally monitored at the intact level using ion exchange (IEC) or size exclusion chromatography (SEC), respectively. Native LC-MS approaches, such as SEC-MS and IEC-MS, combine the traditional gold standard chromatographic assays with accurate mass measurement to allow isoform-specific monitoring of multiple CQAs of intact therapeutic proteins.

MATERIALS AND METHODSIntact trastuzumab was separated by IEC (Thermo Scientific™ ProPac™ WCX-10 column) directly coupled to MS. Mobile phases consisted of aqueous 50 mM ammonium acetate, using a pH gradient from 6.8 to 10.1. LC-MS was accomplished using a Thermo Scientific™ Vanquish™ H-Class UHPLC system with a variable wavelength detector directly connected to a Thermo Scientific™ Q Exactive™ HF-X Orbitrap™ mass spectrometer. Native LC-MS raw data were analyzed using a time-resolved deconvolution approach utilizing Sliding Window and ReSpect algorithms in BioPharma Finder software. Chromeleon software was used to integrate XIC peak areas and relatively compare CQAs of innovator and biosimilar drug samples.

CONCLUSIONS• Method for monitoring multiple CQAs of an intact biotherapeutic compound to allow direct

comparison of innovator and biosimilar products.

Orbitrap mass analyzer is capable of mass measurements with high (sub-Dalton) precision

BioPharma Finder software can be used to manually export mass lists from native ion exchange MS experiments, and m/z values may be targeted and quantified in Chromeleon software.

REFERENCES1. Harris RJ, Kabakoff B, Macchi FD, Shen FJ, Kwong M, Andya JD, Shire SJ, Bjork N, Totpal K,

Chen AB. Identification of multiple sources of charge heterogeneity in a recombinant antibody. J Chromatogr B Biomed Sci Appl. 2001 Mar10;752(2):233-45

2. Rogers RS, Nightlinger NS, Livingston B, Campbell P, Bailey R, Balland A. Development of a quantitative mass spectrometry multi-attribute method for characterization, quality control testing and disposition of biologics. MAbs. 2015;7(5):881-90. doi: 10.1080/19420862.2015.1069454.

TRADEMARKS/LICENSING© 2018 Thermo Fisher Scientific Inc. All rights reserved. Herceptin is a registered trademark of Genentech, Inc. All other trademarks are the property of Thermo Fisher Scientific and its subsidiaries. This information is not intended to encourage use of these products in any manner that might infringe the intellectual property rights of others.

Assessing Biosimilarity by Monitoring Multiple Critical Quality Attributes of an Intact Monoclonal Antibody Drug Using Orbitrap Native LC-MS

Figure 4. Glycoform profiles are easily identified using ReSpect and Sliding Window algorithms in Biopharma Finder software. Our “moderate” mass tolerance of 15 ppm is wide enough to detect both deamidated and main peak forms of trastuzumab. We readily observed distinct glycoform profiles for Herceptin and biosimilar trastuzumab samples.

DEAMIDATION / ISOMERIZATION PATHWAY DOMINATES TRASTUZUMAB CHARGE VARIANT MS PROFILEDeamidation and isomerization isoforms pose challenge of near-isobar heterogeneity

Trastuzumab includes several known residues in the light chain and heavy chain sequences, which may undergo deamidation (LC-N30) or isomerization (HC-D102) or both at once (HC-N55). These modified isoforms are considered to be ‘near-isobars’, which exhibit mass differences in the range of 0-2 Da, and pose an issue for accurate mass identification by MS. These deamidation and isomerization isoforms are observed in the charge variant profile and can be observed directly using pH-elution ion exchange coupled directly to Orbitrap native MS.

MONITORING CQA ISOFORMS IN CHROMELEON SOFTWARELarge scale relative abundance measurements of isoforms identified by BioPharma Finder software

Quantitative monitoring of protein drugs can be easily accomplished using peptide mapping LC-MS using a so-called Multi-Attribute Method (MAM) (Reference 2). This type of approach allows CQAs to be become characterized and then monitored in a routine fashion. The combination of BioPharma Finder software and Chromeleon software serves as a powerful framework for MAM applications to enable deep protein characterization and further method streamlining for downstream process development.

We sought to develop a workflow that performs targeted quantitation of isoform charge states identified by Sliding Window deconvolution of native IEC-MS datasets. We manually exported the centroid m/z values of the charge states of isoforms identified by ReSpect / Sliding Window analysis in BioPharma Finder software. Centroid m/z values were averaged from the 4 replicates and these average values were targeted in Chromeleon.

Modification Example of Interfering Isoforms ∆M (Da / ppm)

ReSpect Deconvolution and Sliding WindowMass / Merge Tolerances Needed for Method

Deamidation G0F/G0F vs. G0F/G0F + Deamidation 1 / 6.6 3 / 3

Succinimide (from Aspartic Acid) G0F/G0F vs. G0F/G0F + AspSucc 18 / 118 15 / 15

Lysine G0F/G1F vs. G0F/G0F + Lysine 34 / 224 15 / 15

Sialic Acid G2F/G2F vs. G0F/S1G1F 34 / 224 15 / 15

Hexose G0F/G0F vs. G0F/G1F 162 / 1069 15 / 15

Isoform Modification ∆M (Da)

†† 2 x Deamidation LC-N30 + 2

‡ IsoAspartate HC-N55 + 1

† Deamidation LC-N30 + 1

†! Deamidation LC-N30IsoAspartate HC-D102 + 1

* Relatively unmodified 0

! IsoAspartate HC-D102 0

‼ 2 x IsoAspartate HC-D102 0

Aspartic acid conversion to succinimide (“basic” RT) identified

on innovator, Herceptin

Sialic acid glycoforms (“acidic” RT shift) identified

on innovator, Herceptin

C-terminal lysine-retaining isoforms (“basic” RT shift) identified on

trastuzumab biosimiliar

Figure 1. Platform for on-line native IEC-MS method. Essential instrumentation for our platform includes the ProPac WCX-10 weak cation exchange column, the Vanquish Horizon UHPLC, the Q Exactive HF-X Orbitrap mass spectrometer, and BioPharma Finder data analysis software.

LOW TOLERANCE SLIDING WINDOW DECONVOLUTION (3 ppm) ALLOWS SUB-Da PRECISION FOR INTACT MASS MEASUREMENT High precision of Orbitrap mass analyzer is demonstrated by reproducible results.

We injected 1 µL (20 µg) volumes of trastuzumab samples in formulation buffer and acquired native LC-MS data on a Q Exactive HF-X system. The results were analyzed using ReSpect and Sliding Window algorithms in BioPharma Finder software. We found a reproducible difference (4 replicates per sample) of ~1 Da between the abundant glycoforms corresponding to the acidic (deamidation) and main (relatively unmodified) LC peaks. Interestingly, we also observed a further ~1 Da difference between innovator Herceptin and a trastuzumab biosimilar, which we could not immediately explain.

0

1000

100

20.586 NL: 6.29E4

0

10020.452 NL: 3.57E4

0

10020.318 NL: 7.62E4

0

10020.384

10 20 30 40RT (min)

NL: 4.31E4

25.874 NL: 4.32E5

25.608 NL: 4.71E5

25.672 NL: 4.40E5

25.671

10 20 30 40RT (min)

NL: 4.94E5

Herceptin

23

4

G0F/G1FDeamidation

G0F/G1FMain and Isomerization

1

BPC BPC

0

1000

100

20.718 NL: 5.52E4

0

10020.781 NL: 5.85E4

0

10020.651 NL: 6.32E4

0

10020.517

10 20 30 40RT (min)

NL: 5.69E4

26.072 NL: 5.42E5

26.134 NL: 4.87E5

26.070 NL: 5.26E5

26.001

10 20 30 40RT (min)

NL: 5.53E5

Trastuzumab biosimilar

23

4

1

BPC BPC

G0F/G0FDeamidation

G0F/G0FMain and Isomerization

Sample Modification (isoform)Mean Average

Mass (Da) *4 replicates

Avg Mass Standard Dev

(Da /ppm)

Mean Mass Accuracy vs. Theoretical (Da /ppm)

Mean Relative Abundance (%)

Mean Apex RT

(min)

Herceptin 2xA2G0F 148057.95 0.28 / 1.85 3.06 / 20.23 63.53 25.94

Herceptin 1xDeamidation (N),2xA2G0F 148059.36 0.83 / 5.46 3.50 / 23.09 4.76 21.34

Herceptin 1xA2G0F,1xA2G1F 148219.81 0.18 / 1.19 2.86 / 18.89 100.00 25.71

Herceptin 1xDeamidation (N),1xA2G0F,1xA2G1F 148221.33 0.43 / 2.85 3.41 / 22.48 10.63 20.44

Herceptin 2xA2G1F 148381.81 0.39 / 2.59 2.79 / 18.44 70.36 25.49

Herceptin 1xDeamidation (N),2xA2G1F 148383.05 0.99 / 6.55 3.05 / 20.15 5.08 20.58

Trastuzumab Biosimilar 1 2xA2G0F 148058.74 0.15 /1.01 3.87 / 25.55 100.00 26.07

Trastuzumab Biosimilar 1 1xDeamidation (N),2xA2G0F 148060.26 0.30 / 1.95 4.42 / 29.16 7.87 20.67

Trastuzumab Biosimilar 1 1xA2G0F,1xA2G1F 148220.65 0.19 / 1.25 3.71 / 24.50 40.16 25.89

Trastuzumab Biosimilar 1 1xDeamidation (N),1xA2G0F,1xA2G1F 148222.84 0.65 / 4.26 4.95 / 32.67 4.44 20.57

Trastuzumab Biosimilar 1 2xA2G1F 148382.58 0.37 / 2.44 3.58 / 23.62 15.76 25.74

Trastuzumab Biosimilar 1 1xDeamidation (N),2xA2G1F 148383.60 0.91 / 6.00 3.62 / 23.87 0.90 20.32

Q Exactive HF-X Orbitrap MSwith BioPharma Option

Vanquish UHPLC

BioPharma Finder 3.0 software

ProPac WCX-10 column4 x 250 mm

Chromeleon 7.2.7 software

Flow rate = 0.3 mL / min

Buffer A: 50 mM NH4Ac, pH = 6.8Buffer B: 50 mM NH4Ac, pH = 10.1

100

6 7 8 9 10 11 12 13 140

*††† ! ‼‡ †!

%

time

Mainpeak∆M

+ 2 + 1 + 1 + 1 0 0 0

Herceptin

0

20

40

60

80

100

5460 5480 5500 5520m/z

25.940 NL: 2.52E5

25.808 NL: 3.79E5

0

1000

100

0

100

0

10025.741

15 20 25 30 35RT (min)

NL: 2.98E5

BPC

Trastuzumab1xA2G0F, 1xA2G1F148219.80

147500 148000 148500 149000Mass

Trastuzumab2xA2G1F148382.52

Trastuzumab2xA2G0F148058.35

0

20

40

60

80

100 Trastuzumab biosimilar

5480 5500 5520m/z

0

20

40

60

80

100

26.072 NL: 5.45E5

25.940

NL: 2.70E525.675

15 20 25 30 35RT (min)

NL: 1.10E5

BPC

0

1000

100

0

100

0

100

Tras2xA2G0F148059.09

148000 148500 149000Mass

Tras1xA2G0F, 1xA2G0F148220.56

Tras2xA2G1F148382.57

0

20

40

60

80

100

Modification Average Mass

Theoretical Mass (Da)

Matched Mass Error

(ppm)Apex RT

1xA2G1F,1xA2S1G1F 148834.37 148832.22 14.5 23.75871xA2G0F,1xA2S1G1F 148674.54 148670.17 28.4 23.75871xA2G2F,1xA2S1G1F 148997.35 148994.27 20.7 23.7587

Modification Average Mass

Theoretical Mass (Da)

Matched Mass Error

(ppm)Apex RT

1xAspSucc,1xA2G0F,1xA2G1F 148202.70 148199.00 24.9 33.93671xAspSucc,2xA2G1F 148363.70 148361.06 17.8 33.67231xAspSucc,2xA2G0F 148041.88 148036.94 33.4 33.9367

Modification Average Mass

Theoretical Mass (Da)

Matched Mass Error

(ppm)Apex RT

1xLys,2xA2G0F 148188.56 148183.13 36.7 33.93671xLys,1xA2G0F,1xA2G1F 148349.36 148345.19 28.1 33.6723

1xLys,2xA2G1F 148513.21 148507.25 40.1 33.9367

Figure 5. Screenshots from BioPharma Finder software show that lower-level CQA isoforms such as succinimide conversion from aspartic acid, sialic acid glycoforms, and C-terminal lysine-retaining isoforms are distinctly identified by mass and retention time shift relative to the main peak.

Figure 2. Asparagine deamidation and aspartic acid isomerization are observed as distinctly separated chromatographic peaks in the charge variant profile. These isoforms are near-isobars, ranging in theoretical mass differences of 0-2 Da, making independent intact mass analysis a challenging endeavor. Isoform identities were previously attained by Harris et al., using fractionation and peptide mapping of each LC peak (Reference 1).

Figure 3. High precision measurements show that an approximately 1 Da mass shift is reproducibly measured for abundant glycoforms at the main vs. acidic peaks, which is highly consistent with deamidation.

Figure 8. Intact glycoforms and charge variant CQAs can be simultaneously monitored. Charge variants CQAs are analyzed as groups of glycoforms, which may be useful for comparing innovator and biosimilar drugs. As shown below, we observed the biosimilar trastuzumab to contain decreased levels of succinimide isoforms compared to Herceptin, but to have markedly higher levels of C-terminal lysine retention.

Figure 6. BioPharma Finder intact protein analysis results can be curated to determine which species should be monitored as CQAs. Interesting isoforms can be selected (box checked) to mark for export in spreadsheet format. To export the constituent charge states’ centroid m/z values for the isoforms identified by BioPharma Finder software, right-click > “Export Checked” > “All Levels”.

PO65315-EN0518S

Main Peak(2xG0F, G0F/G1F,

2xG1F)

C-term Lysine(2xG0F, G0F/G1F,

2xG1F)

Sialic Acid(G0F/S1G1F,G1F/S1G1F)

Succinimide (2xG0F,G0F/G1F, 2xG1F)

Deamidation (2xG0F,G0F/G1F, 2xG1F)

Herceptin 15 ppm 100.00% 0.00% 3.28% 6.44% 12.88%Biosimilar 15 ppm 100.00% 18.13% 0.00% 1.13% 15.59%

0%10%20%30%40%50%60%70%80%90%

100%

Charge Variant Glycoforms Relative to Main Peak Glycoforms

1xA2G0F 1xA2G1F 1xA2G0,1xA1G0F 2xA2G0 1xA2G0,1x

A2G0F 2xA2G0F 1xA2G0F,1xA2G1F 2xA2G1F 1xA2G1F,1

xA2G2F1xA2G1F,1xA2S1G0F 2xA2G2F 1xA2G1F,1

xA2S1G1FHerceptin 30 ppm 1.85% 2.52% 2.37% 3.37% 12.94% 67.28% 100.00% 72.16% 22.84% 3.37% 6.39% 4.51%Biosimilar 30 ppm 1.47% 4.01% 100.00% 65.86% 24.00% 8.64%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Relative Abundance of Main Peak Glycoforms

Figure 7. Isoform charge state m/z values and RT information can be used to create a processing method with flexible options for XIC integration. We quantified 3–4 charge states for each isoform, which is shown here in an overlay plot shaded for customized integration specific for the main peak and deamidated isoforms of G0F/G0F, extracted at 3, 15, 30, and 45 ppm tolerances.

Raw data acquisitionQ Exactive BioPharma Native IEC-MS (“CVMS”)R = 45,000

Discovery data processingSliding Window deconvolution and sequence matching

Low tolerance (3 ppm)• High precision• High abundance species

CQA Targeting

XIC relative abundance

High tolerance (15 – 45 ppm)

• High sensitivity• Low abundance species