1. Andrew Weiskopf, Ph.D. 2010 WCBP CMC Strategy Forum 19 July 2010
CQA Assessment for Glycosylation and Oxidation of a Monoclonal Antibody: An FDA OBP Pilot Program Case Study
Andrew Weiskopf, Ph.D.
2010 WCBP CMC Strategy Forum
19 July 2010

2. Biogen Idec & The FDA Pilot Program
Biogen Idec a participant in the FDA’s OBP Pilot Program for QbD
Meetings with OBP senior staff and reviewers to discuss approach to product and process risk assessments, design space development, and control strategy
Case Study for Pilot Program: Antibody X (AbX)
Monoclonal IgG1, standard Fc N-glycosylation
Oncology drug entering late-stage clinical development
Mechanism-of-Action
CDR of AbX binds to cellular receptor of target cells
Fc of AbX is involved in promoting apoptosis, but Fc-binding entity in vivo is unknown
In vitro assessment indicates ADCC capability is modest
CDC is not involved
FcgRIIIa binding used as metric of Fc integrity
Risk Assessment Examples
Glycosylation: Galactosylation & Fucosylation
Methionine Oxidation
Isotype control
Antibody X 2

3. Product Risk Assessment
Risk-ranking & Filtering
Impact & Uncertainty
Impact
Review each attribute and determine severity associated with failure to control
Consider effects on potency, PK, immunogenicity, safety
SAR studies
Nonclinical studies
Clinical exposure history
Toxicology reports
Data from related IgGs
Uncertainty
Evaluate quantity & relevancy of body of data
Reliance on in vitro / in vivo data
Relevance of related molecules
Range of clinical exposure
Process additives: body of toxicology data and history of use
Final Risk Priority Number (RPN)
RPN > 24: CQA
RPN ≤ 24: Non-CQA
Considerations of process operation excluded from product risk assessment
Keeping process capability out of the assessment makes CQA assessment more “modular” 3

4. Product Risk Assessment: Galactosylation
Potency
SAR studies to assess impact on CDR and FcgRIIIa binding
No effect observed on CDR binding from 36 – 90% G0
Modest but clear effect on FcgRIIIa binding over the same range
Consistent with literature report: fully galactosylated glycoform of an IgG exhibits 2-fold higher FcgRIIIa binding affinity than agalactosyl form 4

5. Product Risk Assessment: Galactosylation
Pharmacokinetics
Literature: Fc glycans of IgG molecules not involved in FcRn binding, thus not expected to impact clearance of AbX
Nonclinical & Clinical Experience
Wide range of terminal galactosylation exposure established with AbX in both nonclinical and clinical development history (36 – 75% G0)
AbX Fc glycosylation variants are typical of human IgG’s, and are not novel structures
Conclusion: CQA
Impact = Moderate (12)
Uncertainty = Low-to-Moderate (3)
Risk Priority Number: 12 x 3 = 36 5

6. Product Risk Assessment: Fucosylation
Potency
SAR study with afucosylated AbX prepared by fucosyltransferase knockout cell line
No effect on CDR binding, but substantial impact on FcgRIIIa binding
Consistent with numerous literature reports
Pharmacokinetics
Fc glycans not expected to impact half-life
Nonclinical & Clinical Experience
Limited: no detectable levels of afucosyl-AbX observed in DS
Conclusion: CQA
Very high Impact (20) x low-moderate Uncertainty (3) = 60
CDR Binding
FcgRIIIa Binding
AbX DS (no afucosyl)
100%
2% afucosyl AbX co-mix
103%
191%
100% afucosyl AbX
98%
>250% 6

7. Product Risk Assessment: Methionine Oxidation
Determination of susceptible residues
Site(s) of methionines prone to facile oxidation govern approach and rationale for risk assessment
Forced oxidation of Antibody X by treatment with H2O2
Five potential sites for oxidation: one on light chain, four on heavy chain
Peptide mapping by LC-MS
Results
Two Lys-C peptides (KH15 and KH30) found to contain hotspots for methionine oxidation
Both hotspots located in Fc region: Met253 & Met429
Oxidation susceptibility is limited to Fc methionines, conserved among IgGs
Can leverage knowledge gained from other IgGs for risk assessment 7

8. Product Risk Assessment: Methionine Oxidation
CDR Binding (FRET)
FcgRIIIa Binding (AlphaScreen)
AbX Control
105%
83%
“25% Oxidized”
23% M253, 11% M429
101%
95%
“50% Oxidized”
47% M253, 23% M429
99%
89%
“100% Oxidized”
100% M253, 100% M429
97%
100%
Potency
SAR study conducted with oxidized preparations of AbX
No effect on activity observed with either CDR or FcgRIIIa binding assays
Literature: Fc methionines reside outside of FcgR binding region, no impact observed with other IgGs
Pharmacokinetics
In-house studies with another IgG
Highly-oxidized Fc methionines (67% at Met253, 41% at Met429) yield several-fold decrease in FcRn binding affinity
Literature: highly oxidized Fc Mets exhibit similar reductions in FcRn binding affinity
Fc mutants demonstrating several-fold reduction in FcRn binding affinity showed no correlation with IgG half-life in rats
Literature: changes in FcRn affinity do not necessarily correlate to impact on half-life 8

9. Product Risk Assessment: Methionine Oxidation
Nonclinical & Clinical Experience
3-8% oxidized Met253 in clinical lots of AbX
No adverse effects traced to differences
Wider range of experience with other IgG products
Other Considerations
Oxidation-induced increase in aggregation rates not studied in isolation for AbX, though reported in literature
Conclusion: Non-CQA
Low-Moderate Impact (8) x Low Uncertainty (2) = 16 9

10. Product Risk Assessment: Attribute Criticality & Control Strategy
Galactosylation
Classified as a CQA, due to modest impact on FcgRIIIa binding
Only critical control parameters linked to galactosylation are readily controlled
Cell culture duration, pH, and temperature
Primary controls achievable through process, justifying in-process testing
Incorporate into comparability assesssments
Fucosylation
Classified as a CQA, due to high impact on FcgRIII binding activity
No critical control parameters identified within the design space
Monitor by proxy through release test for FcgRIIIa binding activity
Variations will be evident due to sensitivity of assay to afucosyl content
Afucosyl levels controlled indirectly by FcgRIIIa binding activity specification
FcgRIIIa binding activity evaluated in comparability assessment
Methionine Oxidation
Classified as a non-CQA, due to low impact on potency & PK
Comparability assessment, selected process monitoring 10

11. Take-Home Messages From the Pilot Program
Keep process considerations separate from product risk assessment
Process changes shouldn’t have to trigger review of attribute criticality
Maintains CQA assessment as an evaluation of the biology, efficacy, and safety of the product
Certainty scoring for CQAs is helpful, but take care with implementation
Scoring system should not overweigh certainty at expense of impact
Don’t mathematically de-risk a critical attribute on the basis of having data
RPNs should accurately reflect degree of criticality and convince one of the risk
Attributes reside on a “continuum of criticality”
“CQA” and “Non-CQA” are useful terms, but developing an appropriate control strategy looks beyond the name
How critical is the attribute?
What process controls can be implemented to control the attribute?
How readily controlled are those parameters?
Control strategy should be proportional to overall degree of risk
Not all CQAs need to be controlled through release testing 11

12. Acknowledgments Analytical Development Research Bioprocess Development
Testing Support
Joseph Siemiatkoski
Michael Molony
Martha Berardino
Namrata Chansarkar
Tongjun Lu
Christine Poliks
Molly Roudabush
Bioassay Development
Svetlana Bergelson
Yucai Peng
Aeona Wasserman
Method Development / Characterization
Yelena Lyubarskaya
Zoran Sosic
Richard Strong
Li Zang
Andy Blum
Tyler Carlage
Hans Fajardo
Damian Houde
Amy Diggle Moran
Samnang Tep
AD CMC Management Group
Damian Cunningham
Mia Kiistala
Maria Lacap
Research
Graham Farrington
Ann MacLaren
Nuzhat Pathan
Bioprocess Development
Lynn Conley
Eric Cunningham
Victoria Delerue
Tia Estey
John Pieracci
Valerie Tsang
Michael Villacorta
Matt Westoby
Helena Yusuf-Makagiansar
QbD Project Team
Vince Narbut
Stephen Notarnicola
Suzanne Stella
Kazumi Kobayashi
Helena Madden
Rohin Mhatre 12