Chris Bailey-Kellogg, Andres H. Gutiérrez, Leonard Moise, Frances Terry, William Martin,

Anne S. De Groot

Immunoinformatic analysis of Chinese Hamster Ovary (CHO)

protein contaminants in therapeutic protein formulations

European meeting on HCPsMay, 2014

How we got to CHO?

2002Invitation to

Immunogenicity Conference at FDA

2011CHO

Genome Published

2006Immunogenicity scale

Tregitopes

CHO genome immunogenicity

analysis

ECI conferenceHCP / CHO Cells

Host Cell ProteinsParallels with Graves’ model

2004Clustered

T cell epitopesEpiBars

CHOPPIOn line . . .

Why examine CHO HCP Immunogenicity?

• What contaminants could be there?• Are they likely to be immunogenic?

IL-6 (pdb id 2l3y) Cathepsin Z (pdb id 1deu)

Immunoinformatics analysis

Databases available

Mousesecreted

165 proteins

Transcriptome32,801 contigs

Validated HCP contaminants

26 proteins

CHO genome24,383

predicted genes

Proteome6,164 proteins

Putatively Secreted

(signal peptide)

EpiMatrix

• EpiVax uses EpiMatrix to predict epitopes– Matrix based prediction algorithm

• Can predict either class I or class II MHC binding– MHC binding is a prerequisite for immunogenicity

• T cell epitopes are linear and directly derived from antigen sequence• Binding is determined by amino acid side chains (R groups) and ‘encoded’

in single letter code

Peptide epitope

`M

HC II Mature

APC

MHC II

Easy easy to deliver as peptidesClusters of MHC binding drive T cells

DRB1*0101

DRB1*0301

DRB1*0401

DRB1*0701

DRB1*0801

DRB1*1101

DRB1*1301

DRB1*1501

• T cell epitopes are not randomly distributed but instead tend to cluster in specific regions. – These clusters can be very powerful, enabling significant immune responses to low scoring proteins.

• ClustiMer recognizes T-cell epitope clusters as polypeptides predicted to bind to an unusually large number of HLA alleles.

What Makes Proteins Really immunogenic?Sequences that Contain EpiBars

Roberts CGP, Meister GE, Jesdale BM, Lieberman J, Berzofsky JA, A.S. De Groot, Prediction of HIV peptide epitopes by a novel algorithm, AIDS Research and Human Retroviruses, 1996, Vol. 12, No. 7, pp. 593-610.

ClustiMer - Locates highly immunogenic regions

EpiBar : A common feature of highly

immunogenic clusters

EpiBar

EpiVax Immunogenicity Scale- 80 -

- 70 -

- 60 -

- 50 -

- 40 -

- 30 -

- 20 -

- 10 -

- 00 -

- -10 -

- -20 -

- -30 -

- -40 -

- -50 -

- -60 -

- -70 -

- -80 -

Thrombopoietin

Human EPO

Immunogenic Antibodies*

Tetanus Toxin

Influenza-HA

Albumin

IgG FC Region

EBV-BKRF3

Fibrinogen-AlphaNon-immunogenic Antibodies†

Follitropin-Beta

PROTEIN_001 (35.13)

Protein Immunogenicity Scale

Proteins Scoring above +20 areconsidered to be potentiallyimmunogenic.

On the left of the scale weinclude some well-knownproteins for comparison

- 80 -

- 70 -

- 60 -

- 50 -

- 40 -

- 30 -

- 20 -

- 10 -

- 00 -

- -10 -

- -20 -

- -30 -

- -40 -

- -50 -

- -60 -

- -70 -

- -80 -

Thrombopoietin

Human EPO

Immunogenic Antibodies*

Tetanus Toxin

Influenza-HA

Albumin

IgG FC Region

EBV-BKRF3

Non-immunogenic Antibodies†

Follitropin-Beta

≥ 20: potentially immunogenic

Immunogenicity Scores distribution

≥ 20: potentially immunogenic

But what about conservation with Human?

What is the impact of the cross-conservation between CHO and Human?

Human

The God of Two Faces: JanusMatrix

MHC/HLA

TCR

MHC

T cell epitop

e

T cell receptor

Identifies cross-reactive peptides:• Identical T cell-facing residues• Same MHC allele, but …• OK if different MHC-facing

residues

Moise L et al. Hum Vaccin Immunother. 2013 Jul;9(7):1577-86

Source epitope

Human protein with cross-reactive epitopes

Cross-reactive human epitope

Source (pathogen) protein

pH1N1 - HA Hepatitis C virus

Effector T cells Regulatory T cells

Visualizing cross-reactivity patterns

Moise L et al. Hum Vaccin Immunother. 2013 Jul;9(7):1577-86. He L et al. BMC Bioinformatics 2014, 15(Suppl 4):S1

CHO-unique epitope content

CHOPPICHO Protein Predicted Immunogenicity

Biotechnology and Bioengineering (In press)

CHOPPI results

≥ 20: potentially immunogenic

`

C-X-C motif chemokine 3

Lysosomal alpha-mannosidase

CHOP cross-reactivity

Lysosomal protective protein

• Exploring immunogenicity adjustment for conservation with self – JanusMatrix, cross-reactive networks

• Define ratio of non cross-reactive vs. cross-reactive associated with immunogenicity

• Further in vitro studies required to determine the impact and implications of findings

• Add additional genomes.. Such as NS0 mouse? E. Coli --- the website is set up so we can add genomes as they become available.

Next Steps

How to do a Risk Assessment on a single HCP?

• Go into CHOPPI• Paste in the protein you are looking for• Run CHOPPI• Look at the overall score: If over 20, possibly a problem.• Look at the conservation with human: if high, then less

of a problem. • We don’t yet know what “ratio” of CHO unique vs.

conserved is clinically meaningful. • We need HCP immunogenicity data – if you have that

information, we will build it into the tool.

Thank you!

Dartmouth College

EpiVax, Inc.

Contact information:Dr. Anne De Grootdr.annie.degroot@gmail.com