Use of Solvent Iodide Ions as an Effective In-house Tool for Crystallographic Phasing Michael R. Sawaya & Duilio Cascio January 17, 2002

Phasing by solvent halide ions promises to be a revolutionary new way to phase protein structures  Quick soak in halide (30 seconds)  Easy to perform  Applicable to any protein  Non-toxic, no heavy atom waste  High quality phasing Acta Cryst (2000). D56,

Iodide is the choice of halide for in-house phasing Atomf” I-6.9 Pt7.0 Br-1.3 C0.01 N0.02 O0.03 S0.56 Anomalous scattering correction factors at CuK  wavelength  At CuK , f” is comparable to Pt, Hg, Au  54 electrons  F-RD generator high flux, fine focus

How successful is phasing by iodide at UCLA? Original study performed with only 4 proteins using a synchrotron source. Will this method prove to be generally applicable in a real academic research lab with real proteins with real problems? What are the limitations? (e.g. resolution limits, lack of isomorphism, data quality, soaking conditions, number of iodide sites required for good phases) Will all proteins bind a sufficient number of iodides to generate good phases? Dauter et al., suggest that the number of iodides bound is simply proportional to the surface area of the protein. Is this always true?

Students, Post docs, and Staff from MBI generously donated their crystals to test the effectiveness of iodide soaks in phasing protein structure proteincontributor Rv1926cCelia Goulding Rv3697cCelia Goulding Rv2878cCelia Goulding DsbD-N-termCelia Goulding P51Chongwoo Kim SmAPCameron Mura NarLc complexAnn Maris RNase ds-trimerYanshun Liu Protein isoaspartyl methyl transferase Scott Griffith Daniel Boutz MyoglobinMaria Grzeskowiak Proteinase KHelty Adisetiyo ThaumatinHelty Adisetiyo XylanaseHelty Adisetiyo lysozymeStudents of M230B (2001) Thank You for the crystals!

Experimental methods flow chart SOAKING PROCEDURES Weigh 0.008g KI (one medium sized grain) Dissolve KI in 100 uL of reservoir solution Add appropriate % of glycerol (for cryoprotection) Soak crystal 30 seconds Mount on cold stream DATA COLLECTION PROCEDURES Collect data on F-RD when possible Collect 360 degrees of data Process with Denzo/Scalepack DETERMINE IODIDE SUBSTRUCTURE Import data with xprepx (Bruker) Calculate difference Patterson coefficients using SAS, SIRAS, SIR data Locate Iodide sites with ShelxD Verify quality of sites by overlapping predicted Patterson peaks on Patterson map PHASING CCP4 suite: scalepack2mtz, truncate, cad, scaleit, mlphare, dm MODEL BUILDING Arp/wArp

12/14 crystals soaked showed clear evidence of iodide binding

8/14 Iodide soaks led to complete structure determination 1 structure had not been previously determined

Example of electron density generated by SIRAS phasing based on 11 iodide sites DsbD N-term

Table 1. Data Statistics on Iodide Soaked Crystals Summary: Iodide is my first choice for derivatization. No other heavy atom as successful with so many proteins under so many different conditions Eight structures solved using phases based on iodide Clear iodide sites but poor phasing Non- Isomorph ous 3 13% 1.8A

Tips for successful phasing with iodide  Poor peak heights in difference Patterson map? High redundancy of intensity measurements is crucial to locating heavy atom sites and phasing. Collect 360 degrees of data. Not just iodide, but any derivative would benefit.  Iodide soak is non isomorphous with native? Non- isomorphorism can be reduced by a quick back-soak in cryo- conditions lacking iodide. (eg. Rv2878c)  Iodide sites not convincing? ShelxD often succeeds at finding iodide sites based on anomalous differences alone. But, If the solution is not clear, try using isomorphous differences (SIR) or a combination of isomorphous and anomalous differences (SIRAS) output by xprepx.

Data collected using FR-D generator can produce better quality maps than RU200 generator I/  (2.0 A) R sym (2.0 A) 37.3% 7.1%

Poor phasing is a direct consequence of too few iodides/surface area Need 1 iodide bound per residues

Why do some proteins bind disproportionately fewer iodides/surface area? Two possibilities 1)Soaking conditions (e.g. pH, salt, buffer) disfavor or compete with iodide binding. If true then we could search for conditions that favor iodide binding. or 2) Residue composition of the protein surface disfavors iodide binding. Make predictions about iodide binding based on amino acid composition.

Iodide binding appears insensitive to the composition of the cryo-solvent Thaumatin 1.3M Na,K tartrate 35% glycerol Bis-Tris pH M KI 1 iodide/14 residues Rv1926c 0.1M (NH 4 ) 2 SO 4 30% PEG 4000 Tris pH M KI 1 iodide/47 residues Soaking a Rv1926c crystal in thaumatin’s cryo-conditions did not increase the number of iodides bound. But, why expect conditions that are optimal for iodide binding to one protein to also be optimal for another protein? Thaumatin is a more basic protein (pI=8.5) than Rv1926c (pI=6.1). Perhaps if I tried a more substantial change in pH to change the electrostatic potential of the surface… Experiment to test effects of cryo-solvent on iodide substitution

Higher pH appears to weaken iodide binding Proteinase K 0.1M (NH 4 ) 2 SO 4 30% glycerol Cacodylate pH M KI Proteinase K 20% PEG % glycerol CHES pH M KI Top 3 negative peaks in F obs(pH9.5) –F obs(pH6.5) difference Fourier map correspond to iodide sites. Experiment to test effects of pH on iodide substitution

Tally of side chains in contact with 102 iodide sites Note: Arginine and lysine are the two residues most frequently found in iodide binding sites.

The amino acid composition favored by iodide is significantly depleted in negatively charged side chains compared to the average amino acid composition on the surface of most proteins Red data points taken from The Atomic Structure of Protein-Protein Recognition Sites by Lo Conte, Chothia & Janin, J. Mol. Biol., 285, CC=0.77

Table 1. Data Statistics on Iodide Soaked Crystals Most successful iodide experiments were conducted at a pH below the pI with the exception of Rv2878c 3 13% 1.8A >>>>><>><<<<>>>>><>><<<<

A protein may still bind iodide even if pH > pI since iodide binding sites are often non-polar No consensus coordination geometry Polar & nonpolar Hydrogens at a radius Angstroms Peptide planes Could involve any of the 20 amino acids

Conclusions  SIRAS phasing from iodide soaks in-house is effective, quick, easy, and non-toxic. 8/14 structures could be determined at UCLA  Even in cases where there are too few iodide sites to produce a good map, iodide sites could be used in combination with other derivatives (e.g. CsCl).  High redundancy, high resolution, and a bright, focused x-ray source (F-RD) are important factors for success.  Soaking at pH < pI improves chances of success Future: Lower the pH of cryo-conditions of Rv1926c or xylanase to increase iodide binding and solve another structure.

Acknowledgements Duilio Cascio- partner in experiments, advice, inspiration CRYSTALS Celia Goulding Chongwoo Kim Cam Mura Ann Maris Yanshun Liu Scott Griffith Daniel Boutz Maria Grzeskowiak Helty Adisetiyo SUPPORT David Eisenberg Todd Yeates Richard Dickerson James Bowie Zbigniew Dauter-advice on back-soaking, shelxD, xprepx. Peter Muller-xprep connections Kim Ma –X-ray maintenance STATISTICS Gary Kleiger Todd Norcross