The Identification and Quantification of Residual Host Cell Proteins (Hcps)

The Identification and Quantification of Residual Host Cell Proteins (HCPs)

Steve Taylor

In purified product there is a high concentration of ‘prod uct’ proteins low concentration of ‘host’ proteins

Waters can Identify and Quantify these with UPLC-MSE

Background to HCPs and the Guidelines of the EU Regulatory Authority

©2009 Waters Corporation | COMPANY CONFIDENTIAL Host Cell Proteins (()HCPs) Recombinant Proteins produced in host cells Proteins from cells can co-purify with therapeutic protein of interest —e.g. Chinese Hamst er Ovary (CHO) cell proteins in recombinant monoclonal antibody therapeutics Purification steps should remove contaminants. Low levels can remain because of —Poor process control —Process changes: can affect HCP pattern and abundances

Safety drives the need for removal/minimization —Link between HCPs and immunogenicity European regulations in effect since 2007 —‘6.2 Validation of the purification procedure - …. The ability of the purification ppprocess to remove other specific contaminants such as host-cell proteins … should also be demonstrated’ —ICH Guidelines: 2009 review in progress (http://www.emea.europa.eu/pdfs/human/bwp/BWPworkprogramme.pdf)

2008: approval of follow-on Diagramatic representation of amino biologic/ biosimilar (EU; USA) acid sequence of human — Initially the application was turned growth hormone down due to a potential immunogenetic effect due to HCPs — Issue was resolved before drug release – positive opinion given —‘The cause of immunogenicity was linked to excess host cell protein contamination, which was resolved by the manufact urer with additional purification steps’. (http://www.pubmedcentral.nih.gov/articlerender.fcgi? artid= 2638545)

Image source: http://www.rxlist.com/omnitrope-drug.htm

HCPs and approval problems for some Biosimilars: — Appli catio n by b ios imila r co mpa ny fo r Inte rfe ro n Alfa 2a ( HepC) — Marketing permission rejected 2006: —“The reasons for the rejection by the EMEA included quality and clinical differences between [the biosimilar product] and the reference product, … inadequate validation of the process for the finished process and insufficient validation of immunogenicity testing.”

Liver Damage from Hepatitis

Thousands of possible protein contaminants HCPs can be present at extremely low levels o Typically ppt to ppm (relative to biotherapeutic) o Guidelines suggest monitoring to ppm (1-100ppm) Developing methods is expensive and time consuming Business Impacts of Failure to identify and remove contaminants: o Can reduce drug efficacy o May lead to adverse events o Drug development and introduction delays o Longer cycle to introduce process improvements o Perceived product quality issue = competitive disadvantage o Kill a promising candidate

Develop a method — To identify, — Quantify and — Monitor HCPs from recombinant means Results must have a means of validating the results (e.g. peptide sequence; concentration confirmation) Needs to be complementary/ compatible with existing methods Needs to provide improvements compared to existing methods (generality; efficiency; speed) Needs to provide cost-effective benefits for process improvements for Waters customers

Current Methods for Host Cell Protein Analysis

Narrow dynamic range (<100)

Biopharm International, Volume 13, Number 6, pp. 38-45, May 2008

©2009 Waters Corporation | COMPANY CONFIDENTIAL 11 EFFICIENCY SPEED Simultaneous i.d. and Quan Significantly shorter of proteins development time for assay Efficient assay and use of Ability to provide high resources throughput/ flexible monitoring assay

WhC/Shy UPLC/MSE ofCf HCPs ?

ACCURACY GENERALLY APPLICABLE Accurate quan of HCPs in HCPs do not need to be known complex mixture prior to analysis Quan over > 4 orders of Can be widely applied and easily magnitude modified

User Expertise Needed for Routine Use

LC/MS Tof LC/UV ELISA: High Low Very Low

Gel and Blot: Low

LC/MS Quad Reasonably Low

Ability to Quantify over wide dynamic range

ELISA: LC/UV LC/MS Poor 2 - 3 orders 4 orders Good Good

Gel and Blot : Poor

Ability to Detect at very low levels

LC/UV: ELISA: LC/MS Acceptable Excellent Good (ppm) (high ppm) (ppt)

Gel and Blot: Variable

User Interpretation required for Analysis

LC/UV LC/MS Objective by Highly RT Objective

Gel and ELISA: Blot: Subjective Subjective

Unambiguous Identification of HCPs

LC/UV: LC/MS High with Extremely SOP High

ELISA: Gel and Blot: Low Acceptable

Method Flexibility

LC/MS LC/UV: Extremely Flexible Flexible

ELISA: Gel and Blot: Very Flexible Inflexible

Summary of Waters Host Cell Protein Methodology

Similarity to proteomics applications - Similar tools can be used with minor changes - Complex Samples by tryptic digest - Same data mininggyg and rules for identifying - Databases used - MSE acquisition Differences: - Greater need for dynamic range (>4) - Needhhhdd to cope with high product concentration and a small amount of HCP (ppm) - Not normally sample limited - Databases can be tailored because Host is known

Informatics PLGS and IdentityE: validated protein identification rediducing flfalse-positive space. BiopharmaLynx 1.2 for automated sequence coverage and confirmation of primary structure of biomolecules (intact mass; peptide mapping) VerifyE for the determination of the most appropriate peptides for quantification (by MRM) Instrumentation NanoUPLC with 2D RP-RP – more reproducible chromatography (greater sensitivity) Synapt/ XevoQtof – accurate mass MSMS TQD/ Xevo TQ – high dynamic range quantitation Chemistry Rapigest: aids tryptic digestion PST/ BEH: Peptide Separation Columns HILIC (‚normal phase‘); Glycan columns

IdentityE to discover proteins - Peptide sequences matched (dB) - Confidence ranking of identification

BiopharmaLynx to - Compare samples vs control - Mit/Monitor/ quan modifidifitications - (e.g. glycoforms) E ExpressionE to quantify proteins - Confirm peptide sequence with MS - Measure of amount - Es ta blis he d protoco l - Generally applicable

VerifyE for ‘signature’ peptides - Relevant Peptides obtained for MRMs - Output of MRM method for Tandem Quad

The Application of 2D nanoAcquity Chromatography

FIRST DIMENSION: 1 mm x 0.5 cm X-Bridge packed with BEH130, 5 µm; 10 µL/min at pH 10 to elute all peptides. — HIGH RESISTANCE to extreme pH Trap column for this resea rch project 500 µm x 2 cm packed with Symmetry C18, 5 µm — High loading capacity

SECOND DIMENSION: 300 µm x 15 cm with BEH130, 1.7 µm; 4 µL/min Standard ESI probe with narrow bore capillary

MS Precursor

MSE Fragments

TIME

MSE

1189.5802 46.71 78430 1.98

765.3742 46.67 449 1

522.2606 46.67 554 1 ursor Ma

cc 800.4481 46.67 3754 1

963.5187 46.69 3658 1

515.3250 46.70 2325 1

Pre 687.3742 46.70 2351 1

1100.5773 46.71 1112 1

822.4154 46.71 436 1

781.4823 46.71 163 1

ct Mass 896.5183 46.72 675 1

685.3210 46.72 862 1

Produ 1009.6112 46.74 125 1 retention time 498.3296 46.75 709 1

432.2357 46.75 356 1

‘Capil ary’ scale Chromat ograph y Mass Spectrometry

nanoAcquity UPLC system SynaptTM HDMSTM with 2D Technology

Informatics

IdentityE Software ProteinLynx Global Server

Methdlhodology for Host Ce ll Prote in Monitoring by Tandem Quadrupole

Once HCPs have been established a high-throughput method can be developed for process monitoring.

23.80 418.7 100 E Discovery phase Verify data processing

22.45 458.7 33.13 E 27.37 422.2 484.7 LC-MS data (N) proteotypic pep

26.99 540.2 Acquisition as above (X) trans per peptide

24.93 29.12 29.93 459.7 626.3 30.73 407.7 582. 3 19.40 33.24 406.2 724.3

24.54 461.7 26.77 % 18.74 20.87 416.7 31.35 551.2 411.7 527.2 21.60 26.26 722.2 435.2 28.22 407.7 26.16 531.7 22.86 447.2 25.89 28.29 559.2 536.7 31.90 721.8 33.44 598.3

17.47 20.22 32.28 16.93 575.2 373.2 669.7 488.5

17.96 30.35 35.84 419.9 625.3 790.9 33. 76 36. 90 740.3 35.70 39.11 700.3 499.2 41.79 789.8 782.3

0 Time 16.00 18.00 20.00 22.00 24.00 26.00 28.00 30.00 32.00 34.00 36.00 38.00 40.00 42.00

Automatically generated Xevo TQ MRM exp file

Use MSE data to develop MRM assay Use in high-throughput monitoring and absolute quantification of HCPs Xevo TQMSTM Xevo TQ with selected MRMs

MS1 Collision Cell MS2

Workflow Overview: —Shotgun enzymatic digestion of sample into peptides —1D or 2D LC/MSE with IdentityE to DISCOVER contaminant proteins —2D for more loading capacity —Develop proprietary host cell protein database —(Hi3 for absolute quantitation, label-free) —Data mined for MRMs —Transfer to Tandem Quad for absolute quantitation (e.g. labeled peptides)

2D Chromatography requires optimisation because: —Column loading is non-linear (more loading does not equate to more dynamic range) —Product is present in much higher concentrations - 2D approa ch means low-lev el impur ities Gilar M. et. al, J. quantifiable despite disparity in Sep. Sci. 2005, concentrations 28, 1694-1703 Chemistries specifically selected for RP/RP 2D approach: —Retention time models applied (based on hydrophobicity scale of tryptic peptides) —All Chemistries readily available (but dimensions adapted)

1D Chromatography (75 um scale) — 0.05 ug o f p roduct d igest loaded fo r pept ide mapp ing — 90 min gradient (5-40% acetonitrile) Gilar M. et. al, J. Sep. Sci. 2005, 2D RP-RP (High/Low pH) Chromatography 28, 1694-1703 — 5 ug (80 pmol) of prod uct digest (over)l oad ed + 100 fmol ADH + 1 fmol BSA — 1st Dimension (pH 10): 5 or 10 step gradient (0 - 45% acetonitrile) — 2nd Dimension (pH 2.6): 90 min gradient (5 - 40% acetonitrile)

UPLC: nanoACQUITY® 2D UPLC® QTof: SYNAPT MS (MSE mode) DtData: PLGS 242.4 (Id(IdtitentityE)

©2009 Waters Corporation | COMPANY CONFIDENTIAL 35 Diaggpgram of 2D setup at higher scale 2D pH=2.4. 0.3 mm x 150 mm BEH Online dilution of 1D flow C18 1.7 µm, Flow at 4 µL/min. 90 to Trap column – change min gradient from 3 to 40% pH and dt the eeoeseectrefore selectivity acetitiltonitrile (01%(0.1% FA-fiformic acid).

Trap: 5-µm Symmetry 1D pH=10. flow 10 µL/min. XBridge high C18 trap - resistance to pH regime. Mobile phase peptides 20 mM ammonium formate in water washed on to (Solvent A) and ACN (Solvent B). Five 2D column. Fractions (to 50.0% B). ©2009 Waters Corporation | COMPANY CONFIDENTIAL 36 2D HPLC using High/Low pH RPLC

pH 10.0 20 mM ammonium formate 0-42% acetonitrile in 5 or 10 steps of 15 min

100 TIC 4.37e8

pH 10

18.95 95 41.26 25.68 % 35.85 15.79 29.41 37.65

8.53 14.03 10.38 13.21 16.41 29.00 41.92 5.70 6.77 11.73 22.39 39.58 9.64 25.92 35.48 4.10 5.21 18.58 21.04 19.89 24.20 18.21

0-56%1 B in 70 minutes 20 mM NH4OH pH 10 Time Bovine_Hemoglobin_Digest_Stored_091803_12.50 5.00 7.50 10.00 12.50 15.00 17.50 20.00 22.50 25.00 27.50 30.00 32.50 35.00 37.50 40.001: Scan 42.50 ES+ 45.00 28.55 TIC 100 neutral acidic 4.51e9 acidic basic pH 2.6 basic 18.75 75

23.86 27.00 pH 2.6 % 26.68 35.05 17.36 26.51 30.68 0.2% Formic acid 16.30 10.99 22.79 8.91 31.41 22.39 0-42% acetonitrile in 90 min 11.40 4.70 6.29 13.24 19.61 11.93 36.19 4.29 19.93 26.06 34.27 14.09

Gilar M. et. al, J. Sep. Sci. 2005, 28, 1694-1703

One peptide may appear in multiple Chromatographi c 45% ACN steps - ‘merge’ step to create coherent 23.6% ACN fractions 20. 8% ACN

18.9% ACN

17.4% ACN

15.9% ACN

14.5% ACN

13.0% ACN

11.7% ACN

8.2% ACN

Example Results from Biosimilar of Trastuzumab

Methodology for Greater Confidence: RdRandom peptide sequences addddded as Decoy strategy to ensure identified peptides real = Total 27,216 entries in database created — 13,600 entries from Swissprot for Golden Hamster and Mouse (homologs) — 6 protein sequences from spiked in proteins: LA, ADH, PHO, BSA, ENL, porcine trypsin, — 2 sequences from TrastuzumAb (heavy and light chains) — Equal number of random seqq()uences as known entries (13,608) False Positive Rate of Protein Return: 5% (user adjustable) Concentration range found here was 10 to 50 ppm relative to therapeutic Lower confidence hits (nearing random) not reported

Host cell: Chinese Hamster Ovary (CHO) Database with combination of all the mouse and hamster protein sequences listed in the Swiss- Prot database (http://www.expasy.ch/sprot/) —Chinese Hamster database held privately so homology database used (Golden hamster; M)Mouse)

Listing of Proteins — Accession details — Names and sequences generated — Confidence ranking Spectral overview (can be zoomed)

Sequence information for identified protein

Sequence Information available even on low level proteins

IDENTITYE (Discovery Data) Input .csv/.txt Proteotypic Peptide Filters Efficient Transition Filters

Retention Time UPLC/MRM (& TargetLynx) Method Optimization File Creation

Scouting Run Targeted (MRM) Targeted (DDA) .exp UPLC/MRM Quanpedia dB File

OPTIMISED TARGET PROTEIN ANALYSIS UPLC/MRM

RSD = 3% 1 picomole BSA digest on column Reproducible Chromatography RdiblReproducible RTs

RSD = 3% 1 picomole BSA digest on column RdiblReproducible ChtChromatograp hy Reproducible RTs

RSD = 8% 200 fmoles ENL digest on column Reproducible Chromatography Reproducible RTs

RSD = 13% 200 fmoles ENL digest on column Reproducible Chromatography RdiblReproducible RTs

Workflow Models: — UPLC-MSE is well-established — Proteomics tools already exist and are developing (e.g. HDMSE) Applications Benefits: — Confident Identification of individual HCPs – with ranking of confidence — Quantitation of each identified HCP o Label-free with discovery stage (Synapt/ XevoQT) o Using isotopically labelled peptides (XevoTQ/ TQD) — Much faster development time than immunoassay — Provide a multi-purpose platform for many other tasks — Sensitivity levels comparable to ELISA (low ppm) — Also applicable to subunit (recombinant) Vaccines

The 2D-LC/MSE setup is able to identify low abundance pppprotein contaminants present in biopharmaceuticals over more than 4 order of magnitude

The 2D-LC setup using the second chromatographic dimension provides the sensitivity and robustness required for HCP analysis

A high-throughput MRM assay on the Xevo TQ MS can qqyuantify these p rotein imp urities ( absolute q uantification can be done using isotopically labeled peptides)

The combination of 2D-LC/MSE and Xevo TQ MS provides a total system solution for HCP analysis

Waters Biopharmaceutical Development: Catalin Doneanu Hongwei Xie Keith Fadgen Martha Stapels Jim Kehoe Weibin Chen Scott Berger