PAVING THE WAY FOR ASSESSING IN VIVO DYNAMICS OF MULTIPLE QUALITY ATTRIBUTES FOR PROTEIN THERAPEUTICS

Haihong Zhou Principal Scientist, Biologics, Vaccines & Bioanalytics PPDM Merck & Co., Inc., Kenilworth, NJ, USA

CASSS Mass Spec September 21, 2017

1

Outline

• Workflow for measuring multiple quality attributes in vivo – Key drivers – Challenges – Method development considerations

• Proof of concept studies • Differential Analysis

2

In Vivo Dynamics of Quality Attributes: Bioanalytical Multi-Attribute Method

3

Drug (antibody, fusion protein, peptide, nanobody, etc)

Attribute

In Vivo Dosing Blood Collection

Background protein

Affinity Purification Elution/Digestion High-resolution LC/MS Characterization

Impact of PQAs on product safety and efficacy

Differential Analysis

Peptide Identification

Attribute Identification and Quantification

0 10 20 30 40 50 60 Time (min)

0 10 20 30 40 50 60 70 80 90

100

Rel

ativ

e A

bund

ance

mAb PK Profile Proportion with Attribute

Attribute Exposure Profile X =

Time In Vivo

mAb

Con

cent

ratio

n

Prop

ortio

n of

mAb

with

At

tribu

te

Attri

bute

Con

cent

ratio

n

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

0

Adapted from Goetze AM, et al. MAbs. 2010;Sep-Oct;2(5):500-507; Li Y, et al. MAbs. 2016;Aug-Sep;8(6):1079-1087.

X =

Key Drivers

• Help evaluate criticality of the quality attributes and facilitate manufacturing process development

• Assess the effect of product quality attributes on safety and efficacy • Establish better understanding of PK/PD relationship

4

LC/MS – QQQ • Simple processing steps

(Monitor one or two peptides) - Absolute quantitation - High throughput - Known

In Vivo • Limited volume • Low concentration • Complex matrix •May need purification

Determine drug concentration (PK)

Bioanalytical

Challenges in Quantifying Multiple Quality Attributes In Vivo

5

LC/MS – High Resolution • Complex processing steps

(Monitor all peptides) - Relative quantitation - Low throughput - Known & unknown

In Vitro • Volume not limited • High concentration • Buffer • No need for purification

Detect changes in multi-attributes in vitro (CMC)

MAM

LC/MS – High Resolution • Complex processing steps

(Monitor all peptides) - Relative quantitation - Low throughput - Known & unknown

In Vivo • Limited volume • Low concentration • Complex matrix • Require purification

Detect changes in multi-attributes in vivo (metabolism)

Bioanalytical MAM

Proof of Concept Using an IgG1 mAb

6

Drug in Serum ex vivo Immuno-enrichment

Drug Spiked in Serum Immuno-enrichment

Drug in Serum in vivo Immuno-enrichment

Drug in Buffer Direct analysis

Drug in Buffer Immuno-enrichment

Starting point: CMC MAM

Control for sample processing induced modifications

Method Development Considerations • Sensitivity requirement

– Concentrations in PK profile – Available sample volume

• Affinity purification – Choice of the capture antibody – Affinity purification media (binding capacity, nonspecific binding, ease of automation, etc.) – Optimization on yield and purity

• Digestion conditions – Choice of protease

• LC/MS instrumentation – Low flow vs high flow chromatography – Low resolution vs high resolution mass spectrometer

• Overall method reproducibility

7

Method Performance

• Reproducibility • Sequence coverage • Control for sample processing induced modifications

8

1 Heavy Chain 3 Leader Sequence 2 Light Chain 4 Unidentified

0

20

40

60

80

100

Rel

ativ

e In

tens

ity

69.13

68.94 70.61

68.45 72.27

74.87

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 RT (min)

Controls

Consistent recovery of all the quality attributes with and without affinity purification and with different digestion methods indicates that

1. There was no bias in the affinity purification step toward any specific quality attribute

2. Overall sample processing procedure did not alter quality attributes

9

Ref. Material #1 Ref. Material #2 12M25C #1

12M25C #2 0

10

20

30

40

HC D222 IsoD

% A

ttrib

ute

Direct Digestion (100 µg)

Direct Digestion (15 µg)

IP from Serum (15 µg)

% A

ttrib

ute

0

2

4

6

8

HC M429 Oxidation

0

1

2

3

4

HC N385 & N390 Deamidation

0.0

0.2

0.4

0.6

0.8

LC pE

Direct Digestion (100 µg)

Direct Digestion (15 µg)

IP from Serum (15 µg)

Direct Digestion (100 µg)

Direct Digestion (15 µg)

IP from Serum (15 µg)

% A

ttrib

ute

% A

ttrib

ute

Direct Digestion (100 µg)

Direct Digestion (15 µg)

IP from Serum (15 µg)

Bioanalytical MAM Study Design

10

D0 (0.5hr) D2 D4 D7

Predose

D0 D2 D4 D7

In Vivo IV infusion at 200 mg/kg

Ex Vivo 0.5 mg/mL spiked in cynomolgus monkey serum; incubate at 37°C

N-terminal Modification: HC Glutamine to Pyro-glutamate

• A complete conversion of N-terminal glutamine residue into N-terminal pyroglutamate was observed in vivo

11

0 2 4 6 8 80.0

85.0

90.0

95.0

100.0

In Vivo

Post Dose (Day)

Rel

ativ

e P

erce

ntag

e (%

)

0 2 4 6 8 80.0

85.0

90.0

95.0

100.0

Post Dose (Day)

Cyno-1 Cyno-2 Cyno-3

Replicate 1 Replicate 2 Replicate 3

Ex Vivo

Rel

ativ

e P

erce

ntag

e (%

)

Relative Percentage (%) = 𝑃𝑃𝑃𝑃 𝐴𝐴𝑃𝑃 (𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝑚𝑚𝑃𝑚)𝑃𝑃𝑃𝑃 𝐴𝐴𝑃𝑃 𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝑚𝑚𝑃𝑚 +𝑃𝑃𝑃𝑃 𝐴𝐴𝑃𝑃 𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝑚𝑚𝑃𝑚 +𝑃𝑃𝑃𝑃 𝐴𝐴𝑃𝑃 (𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝑢𝑢𝑚𝑚𝑃)

Deamidation: HC N385/N390

• A conserved site in the Fc domain commonly found in humanized monoclonal antibodies exhibited a rapid increase in deamidation

Deamidation Rate: In vivo: 0.9%/day (1.1%1) Ex vivo: 1.5%/day (1.7%1)

2.0

4.0

6.0

8.0

10.0

2.0

4.0

6.0

8.0

10.0 Cyno-1 Cyno-2 Cyno-3

Rel

ativ

e P

erce

ntag

e (%

)

Rel

ativ

e P

erce

ntag

e (%

)

Post Dose (Day) Post Dose (Day)

1Yin S, et al. Pharm Res. 2013;30:167-178.

In Vivo Ex Vivo

0 2 4 6 8 0 2 4 6 8 0.0

12.0

0.0

12.0

Replicate 1 Replicate 2 Replicate 3

12

C-terminal Modification: HC Lysine Processing

• C-terminal lysine processing was likely due to carboxypeptidase activity

13

0.0 0.0

2.0

4.0

6.0

8.0

2.0

4.0

6.0

8.0

Rel

ativ

e P

erce

ntag

e (%

)

Rel

ativ

e P

erce

ntag

e (%

)

Post Dose (Day) Post Dose (Day)

In Vivo Ex Vivo

0 2 4 6 8 0 2 4 6 8

Cyno-1 Cyno-2 Cyno-3

Replicate 1 Replicate 2 Replicate 3

High-mannose Glycans • High-mannose glycans on the Fc region of therapeutic IgG antibodies increase

serum clearance

14

0 2 4 6 8 0.0

0.2

0.4

0.6

0.8

1.0

Post Dose (Day)

Rel

ativ

e P

erce

ntag

e (%

) Man5

0 2 4 6 8 0.00

0.05

0.10

0.15

0.20

Post Dose (Day)

Man6 Man7 Man8

0 2 4 6 8 0

20

40

60

80

100

Post Dose (Day)

A2G0F A2G1F

Rel

ativ

e P

erce

ntag

e (%

)

Rel

ativ

e P

erce

ntag

e (%

)

Example of Attribute Exposure Profile

15

HC N385 & N390 Deamidation

0 2 4 6 8 0

2000

4000

6000

8000

PK Profile

Post Dose (Day)

Ser

um C

once

ntra

tion

(µg/

mL)

0 2 4 6 8 0

2

4

6

8

% Attribute

Post Dose (Day)

Rel

ativ

e P

erce

ntag

e (%

)

0 2 4 6 8 60

80

100

120

140

160

Attribute Exposure Profile

Post Dose (Day)

Cyno-1 Cyno-2 Cyno-3

Ser

um C

once

ntra

tion

(µg/

mL)

Looking for Unknown Changes in an Unbiased Manner

16

Ex Vivo Day 0 Ex Vivo Day 7

• SIEVE (ThermoFisher) • BioPharma Finder (ThermoFisher) • Progenesis QI (Waters-Nonlinear Dynamics)

Monkey Serum

Affinity Purification

LC/MS/MS

Differential Analysis

Identification

Quantification

SIEVE • Ex Vivo Day 0 vs Ex Vivo Day 7

– SIEVE was able to pick up most of the known changes in the sample

17

Deamidation HC N385/N390

0 2 4 6 8 0.0

2.0

4.0

6.0

8.0

10.0

12.0

Ex Vivo

Time (Day)

Rel

ativ

e Pe

rcen

tage

Replicate 1 Replicate 2 Replicate 3

SIEVE (continued)

• Low abundant deamidation was missed by SIEVE with threshold of 1e6

• Lower threshold (eg, 1e4) allowed SIEVE to detect HC76-86 deamidation. However, more false positives were observed when using lower threshold

18

HC76-86 Deamidation

38.0 38.5 39.0 39.5 40.0 40.5 Time (min)

0

10

20

30

40

50

60

70

80

90

100 0

10

20

30

40

50

60

70

80

90

100

Rel

ativ

e A

bund

ance

NL: 5.74E4 Base Peak m/z= 677.3388-677.3524 F: FTMS + p ESI Full ms [300.00-1800.00] MS 20170404_cyno_D0_1

NL: 2.10E5 Base Peak m/z= 677.3388-677.3524 F: FTMS + p ESI Full ms [300.00-1800.00] MS 20170404_cyno_d7_1

BioPharma Finder • Low abundant attribute changes were detected, and the percent attribute was calculated automatically

19

Progenesis QI • Low abundant HC76-86

deamidation was also detected using Progenesis

20

Deam

idat

ed

Inte

nsity

Inte

nsity

D0 D7

Unm

odifi

ed

Differential Analysis for Bioanalytical MAM

• All three algorithms can pick up known attribute changes – BioPharma Finder has better identification capability and calculates attribute percentage

automatically when peptide identity is known • Challenges for all three algorithms include:

– Setting up the appropriate threshold to minimize the number of false positives – Statistical analysis tools to help reduce false positives – Filtering strategies and visualization tools to help distinguish false positives from

true positives – The ability to normalize overall intensity, align retention time, and perform peak picking

when molecule concentration differs several folds across multiple time points

21

Summary

• We demonstrate that it is feasible to track changes in multiple quality attributes for a IgG1 monoclonal antibody in cynomologus monkey serum

• We present a general strategy on how to identify and quantify changes in multiple quality attributes of protein therapeutics in vivo

• Further developmental efforts are needed to improve – Overall sample processing sensitivity and reproducibility – Algorithms for differential analysis can be modified to streamline the data analysis workflow

(eg, comparison with multiple time points and better visualization tools)

22

Acknowledgments

Merck & Co., Inc. Yi Wang, Douglas Richardson, Bhumit Patel, Daniela Tomazela, Vibha Jawa, Richard Wong, Dong Hun Lee, Sejal Patel, Maribel Beaumont, Yan-hui Liu, Ayesha Sitlani, David Pollard, Shuangping Shi, Hetal Sarvaiya, Amy Beebe, Yaoli Song, Mohammad Tabrizifard, David McLaren, Stephen Previs, Maria Webb, Xiang Yu Just Biotherapeutics Richard Rogers ThermoFisher Michael Blank, Jennifer Sutton Rockefeller University Yinyin Li 23