David Cooper University of Virginia

The Integrated Center for Structure and Function Innovation:A PSI-2 Specialized Technology Center

David Cooper

University of Virginia

ISFI Mission StatementThe Integrated Center for Structure and Function Innovation (ISFI) is an NIH Protein Structure Initiative Specialized Center focused on developing and applying a set of synergistic technologies organized to overcome recognized bottlenecks in structure determination at the key steps of production of soluble protein and protein crystallization.

Addressing high-throughput bottlenecks:Protein Solubility and Crystallization

• Only 50.5% of “Expressed” proteins are soluble

• Only 35.5% of “Purified” proteins produce crystals

• Only 18.8% of “Purified” proteins produce X-ray quality crystals

Target DB Statistics

# o

f P

rote

ins

Data from Target DB as of Nov. 20, 2006

Los Alamos National Lab

University of Chicago

University of Virginia

Lawrence Livermore National Laboratory

University of California

Los Angeles

The ISFI

Lawrence Berkeley National Laboratory

Los Alamos

UChicago

The University of Virginia

LBNL

LLNL

UCLA

The ISFI

The Joint Center for Structural Genomics

Midwest Center for Structural Genomics

ISFI Protein Pipeline

Los AlamosProtein Production

Facility

UCLAAnalysis of protein complexes

Co-expression of partners

UChicagoCrystallization

chaperone design

LANLDirected EvolutionProtein Production

UVAProtein surface

engineering

LBNLData Collection

Facility

LLNLCrystallization

Facility

Los Alamos: Geoff Waldo et alFinding Soluble Domains using a HT pipeline

LANL

Polyketide Synthase Example

UCLA: David Eisenberg et alIdentification and Crystallization of Protein Complexes

•ProLinks 3.0 http://mysql5.mbi.ucla.edu/A database of inferred functional linkages

Uses: Phylogenetic Profiles Rosetta Stone Gene neighbor Gene cluster

•Crystallization of Complexes Cloning partners identified by ProLinksCo-expression using modified Duet vectorsCrystallization of predicted functional complexes

UCLA

Prolinks identifies 44% (107/242) of PDB prokaryotic complexes

Benchmarking ProLinks Identification of Prokaryotic PDB Complexes

PDB polypeptide chains (17,844 unique chains)Source: 8/2004

Identify functionallylinked sequences

(BLAST against Prolinks)

>= 1 High confidencefunctional linkage

Source organism presentin both PDB and Prolinks

Group sequences by PDB structure(24,475 structures)

NoFuture studies

Identification ofFunctionally

Linked Proteins

Filtering

Yes

Structures w/ >= 2 different chains (complexes)

782 complexes

242 non-redundant complexes

107 non-redundant complexes

NoFuture studies

NoFuture studies

Prokaryotic filter:

Prolinks filter

Biological filter

Yes

http://mysql5.mbi.ucla.edu/

Co-crystallization of predicted complexes

University of Chicago: Tony Kossiakoff et al Chaperone-Assisted CrystallographyAntibody fragments are produced from phage display experiments. This

method has several advantages: Fast antibody production. Improvement in crystal formation. Capability of coupling biochemical studies to selection. Acquisition of necessary phase information for structure

determination of complex (via Molecular Replacement or SeMet antibodies)

Target Validation and Prep1 week

Automated Screening w/KingFisher (3 rounds)

Confirm Enrichment of Hits1.5 weeks

Subclone into Expression Vectors1 week

HTP Protein ProductionBIAcore Assay0.5 week

UChicago

High-Affinity S/Y antibodies

Target BC DE FG Kd

(nM) MBP SYSSSVS GSKS YSYYYYYYSS 18 SYSSVY GSKS YSYYYYYYSS 23 SSYYYYYVS GYSS YSYSSYYSYYYS 20 ySUMO YYSYYSYSVS YYYS YSSYSSSSYS 14 SSSSVS SYYS YYYSYYYYYSYS 7 hSUMO4 SYYYVS SYYS YYSSYYSSYYYSYS 22 SYYYVS SYYS YYYYSYYYYYYS 7

A reduced code of amino acids can be used to generate binding sites with high affinities.A binary code works great! (Ser / Tyr)

FABsfor14

MCSGTargets

FN3 affinities for 3 targets

Two strategies for protein crystallization:

UVA

UVA: Zygmunt Derewenda et al Surface Entropy Reduction

Systematically altering the protein surface to facilitate crystallization

Varying the protein parameter• Homologues• Different construct ends• Reductive Methylation • Alanine scanning• Directed Evolution• Rational Mutagenesis

Promotes crystallization by altering surface features that inhibit crystallization.

Large, flexible residues on the surface can inhibit crystallization.

Lysine and Glutamate are primarily responsible for the “entropy shield”

Candidate Proteins: Soluble and purify well Difficult to crystallize or diffract

poorly Contain a cluster of highly-

entropic residues

Surface Entropy ReductionSystematically altering the protein surface to facilitate

crystallization

Lysine GlutamateRotamers Rotamers

Our Model Protein -- RhoGDI

Observations from initial experiments. Mutated residues are often found at or in crystal contacts. Single mutations may change the kinetics of crystallization, but

double and triple mutations lead to new crystal forms.

Meets all SER criteria Rich in lysines (10.1%) and glutamates (7.9%)

(average incidence of 7.2% and 3.7%, respectively) It took years to get a poorly-diffracting wild-type crystal.

(Longenecker, et al Acta Cryst. D57:679-688. 2001)

(Mateja, et al Acta Cryst. D58:1983-91. 2002)

The RGSL domain of PDZRhoGEFLongenecker KL, et al. & Derewenda Z.S. Structure (2001) 9:559-69

The LcrV antigen of the plague-causing bacterium Yersinia pestisDerewenda, U. et al. & Waugh, D.S. Structure (2001) 9:559-69

Product of the YkoF B. subtilis gene Devedjiev, Y. et al. & Derewenda, Z.S. J Mol Biol (2004) 343:395-406

Product of the YdeN B. subtilis gene Janda, I. et al. & Derewenda, Z.S. Acta Cryst (2004) D60: 1101-1107

Product of the Hsp33 B. subtilis gene Janda, I. et al. & Derewenda, Z.S. Structure (2004) 12:1901-1907

The product of the YkuD B. subtilis gene Bielnicki, J. et al. & Derewenda, Z.S. Proteins (2006) 1:144-51

Human Doublecortin N-terminal domainCierpicki, T. et al, & Derewenda, Z.S. Proteins (2006) 1:874-82

The Ohr protein of B. subtilisCooper, D. et al. & Derewenda, Z.S. in preparation

Human NudC C-terminal domainZheng, M. et al. & Derewenda, Z.S. in preparation

APC1446 -- Crystals diffracting to 3.0 Å, but unsolved.

**MCSG Targets**

Novel proteins crystallized by SER:

RGSL domain of PDZ-RhoGEF Structure 9:559-69 (2001)

YkoF JMB 343:395-406 (2004)

LcrV Structure 12:357-8 (2004)

Hsp33 Structure 12:1901-7 (2004)

The recurrence of crystal contacts involving mutated sites validates the hypothesis that crystallization is facilitated by surface entropy reduction.

Ongoing work: Optimization of the screening protocols Evaluation of other amino acids at crystal forming interfaces:

Alanine, Histidine, Serine, Threonine, Tyrosine Use of bioinformatics for prediction of crystallizable mutants

Ongoing work: Optimization of the screening protocols Evaluation of other amino acids at crystal forming interfaces:

Alanine, Histidine, Serine, Threonine, Tyrosine Use of bioinformatics for prediction of crystallizable mutants

A B C D

E F G H I

The Most successful MutantK138Y, K141Y

•34 hits in the traditional screen•35 hits in the salt screen

•Wild Type•No hits in the traditional screen•1 hit in the salt screen

Conclusions:

Alanine, tyrosine and threonine can be effectively used as crystal-contact mediating residues.

The salt screens produced almost 33% more hits – 242 vs. 183.

Performing traditional and alternative reservoir screening greatly increases the chances of getting a hit and greatly increases the number of conditions that give hits.

At certain surface locations some amino acids seem to nucleate crystal contacts better than others. Thus, different amino acids may be tried at each selected site to increase chances of success.

SER Prediction Server(Luki Goldschmidt, UCLA)

http://nihserver.mbi.ucla.edu/SER/

The server is designed to predict mutations that may increase the likelihood of crystallization. It has many user editable parameters, but is designed to be ready out of the box.

http://nihserver.mbi.ucla.edu/SER/

The SERp Summary Page

The Server presents ranked mutation suggestions. It also links to homologous structures, potential interacting partners and conserved blocks. The secondary structure prediction and blast results are also presented.

http://nihserver.mbi.ucla.edu/SER/

Get Pretty Results – then get crystals.

As of Nov. 21, 260 users have submitted 1430 jobs.

HSP33Structure

Publications by other groups using SER Novel proteins (black) or preparations of higher quality crystal forms (green)

The CUE:ubiquitin complex Prag G et al., & Hurley JH, Cell (2003) 113:609-20

Unactivated insulin-like growth factor-1 receptor kinaseMunshi, S. et al. & Kuo, L.C. Acta Cryst (2003) D59:1725-1730

Human choline acetyltransferaseKim, A-R., et al. & Shilton, B. H. Acta Cryst (2005) D61, 1306-1310

Activated factor XI in complex with benzamidineJin, L., et al. & Strickler, J.E. Acta Cryst (2005) D61:1418-1425

Axon guidance protein MICALNadella, M., et al. & Amzel, M.L. PNAS (2005) 102:16830-16835

Functionally intact Hsc70 chaperoneJiang, J., et al. & Sousa, R. Molecular Cell (2005) 20:513-524

L-rhamnulose kinase from E. coliGrueninger D, & Schultz, G.E. J Mol Biol (2006) 359:787-797

T4 vertex gp24 protein Boeshans, K.M., et al. & Ahvazi, B. Protein Expr Purif (2006) 49:235-43

Borrelia burgdorferi outer surface protein AMakabe, K., et al. & Koide, S. Protein Science, (2006) 15:1907-1914

SH2 domain from the SH2-B murine adapter proteinHu, J., & Hubbard, S.R J Mol Biol, (2006) 361:69-79

Mycoplasma arthriditis-derived mitogenGuo, Y., et al., & Li, H. J., Acta Cryst (2006) F62:238-241

2HDX – SH2-B with JAK2pY813

2HDV – Unliganded SH2-B

Principle in ActionE583A,E584A,(W593H)

Extending the MethodMulti-domain proteins.

2CGJ

2CGK

Red and Green denote domains.

E69A, E70A, R73A

Another Multi-domain Example

Several wild-type crystal forms were “not suitable for x-ray diffraction studies”

Made double Lys->Ala mutant K141A and K142A

“Well diffracting crystals of the mutated protein were readily obtained.”

Red and Green denote domains.

Extending the MethodMutating Multiple Clusters

First MutationsK48A, K60A, K83A,K196ADidn’t work

Added E37S, E45S, K46S,K64S, E104S, K107S, K239S, E240S, and K254S

University of VirginiaZygmunt DerewendaDavid CooperTomek BoczekWonChan ChoiUrszula DerewendaKasia GrelewskaNatalya OlekhnovichGosia PinkowskaMichal ZawadzkiMeiying Zheng

Lawrence Livermore National LaboratoryBrent Segelke Dominique ToppaniMarianne KavanaghTimothy Lekin

Lawrence Berkeley National LaboratoryLi-Wei Hung Evan BurseyThiru

RadhakannanJim WellsMinmin Yu

University of ChicagoAnthony Kossiakoff Shohei Koide Magdalena BukowskaVince CancasciSanjib DuttaKaori EsakiJames HornAkiko KoideValya TerechkoSerdar UysalJingdong Ye

Los Alamos National LaboratoryTom Terwilliger Geoffrey WaldoChang Yub KimEmily AlipioCarolyn BellStephanie

CabantousNatalia FriedlandPawel ListwanJin Ho MoonJean-Denis PedelacqTheresa Woodruff

UCLADavid Eisenberg Daniel AndersonSum ChanLuki GoldschmidtCelia GouldingTom HoltonMarkus KaufmannArturo Medrano-

SotoMaxim PashkovTeng Poh KhengMichael StrongPoh Teng

Acknowledgements

A Success from Screening Alone

A MCSG abandoned target.Wild-type crystallized only in the salt screen!

Meets all SER criteria Rich in lysines 10.1% and glutamates 7.9% average incidence

of 7.2% and 3.7%, respectively It took years to get a poorly diffracting wild-type crystal

Previous Successes w/ RhoGDI