Tips and Tricks in Crystallography - Crystallization

Xinhua Ji: It was recently reported that microseeding was used to produce untwinned crystals of an O-demethylase using twinned crystals as seeds (Acta Crystal. F72:897-902, 2016). Microseeding separates nucleation events from crystal growth events so that it can yield different crystal forms. Hence, it is promising that this technique can be used to produce untwinned crystals from twinned seeds.

Alternative Reservoirs in Vapor Diffusion Experiments

At the 45th Mid-Atlantic Protein Crystallography Meeting, Dr. Ivan Shabalin from Professor Wladek Minor's laboratory at the University of Virginia shared various tricks in crystallization with the audience, including alterntive reservoirs in vapor diffusion experiments. Wondering what it is? Follow this link for an introduction.

The Cross-Linking Effect of PEG

Michael Garavito: One of the unfortunate by-products of keeping PEG stock solutions in water is that they will form peroxides and aldehydes. They will slowly cross-link the surface of some crystals. However, it is dependent on the nature of your protein's composition of surface residues, so not every protein cyrstal does this.

I had one case where PEG4000 grown crystals would be resistant to dissolving and would easily bend; the thinner rods would spring back staight. After placing the crystals into buffer known to dissolve them, I poked the crystals hard and the insides squeezed out like toothpaste, leaving an empty sack behind. the bottom-line is that fresh crystals diffracted better than old crystals because of this cross-linking.

Suggestions: (1) Make your PEG stocks up fresh or store them in the freezer as aliquots; (2) Remove oxidized PEGs from your stocks (See Ray et al. Biochemistry, 30, 6866-6875, 1991 and Jurnak, J. Cryst. Growth, 76, 577-582, 1986); (3) Check to see if freshly grown crystals behave better.

Seeded Screening with a Robot

Artem Evdokimov: By popular request here's my favorite version of the in-screen seeding. We use a Mosquito but it doesn't have to be a specific robot as long as it can dispense relatively tiny volumes of seed stock.


Caveats: (1) If I am desperate enough to do this, then the situation is pretty bad indeed and I don't mind wasting some protein; (2) my success rate is not hugely favorable but this does work on occasion when other things have failed.

(1) Identify a few likely conditions. Ideally they have microcrystals but desperation has made me try 'lovable precipitates' in the past, with a modest degree of success.

(2) Harvest entire drop using a few ul of mother liquor as diluent

(3) Break the existing crystals using your favorite method (sead beed, etc.) mine involves swirling a pipette tip in the mixture, running it along the walls, with rapid pipetting up and down. Dilute seed stock to useful volume (enough for screening).

(4) I do not normally centrifuge the resulting seed stock, but some people do

(5) Dispense your screen as always with the usual protein/reservoir ratio. Let's say you like drops of the 0.2ul+0.2ul variety - add 25 nanoliters of the seed stock *last*. Optional mixing of the condition is a fun thing to try but it seems not to matter very much. Note that I typically use the same tips to dispense seed stock, fully aware that this causes cross-contamination of conditions. I don't mind :)

  (5a) Variation - Add seed stock to protein, then dispense ASAP. Surprisingly not a bad option, practically speaking.

  (5b) Variation - Crosslink seed stock very gently in solution (with trace of glutaraldehyde) before use. Buffers/additives with primary or secondary amine groups do interfere, of course.

  (5c) Variation - Mix seeds from SEVERAL initial hit conditions, then use as one seed stock. Be ready for fireworks as they may not be compatible!

(6) Endure nail-biting wait for results :)

As noted earlier, it's not a sure-fire way to get new hit conditions but it does seem to work and it's a fun way to put to use a remainder of otherwise useless protein (when you've tried all other tricks you like to try).

Comments and suggestions are always welcome!

Ligand Concentration

Herman Schreuder: If you do the calculations, you will find that you need a FREE ligand concentration of >10 * Kd to get >90% occupancy of the binding site.

If you have e.g. a ligand with a Kd of 100 nM, you would need a free ligand concentration of 1 µM. However, a solution of 10 mg/ml of a protein of 30 kDa, has a protein concentration of 333 µM, so in theory you should have a total ligand concentration (free + bound) of 334 µM. In pratice, some of the ligand may have been degraded during (prolonged) storaged, or the compound may not be as pure as the chemist would have wished it to be, so it is wise to use a safety margin of at least two. We normally use 1-2 mM compound to be on the safe side. Having too much ligand usually does not hurt, except that you use more compound, but with too little ligand you end up with an empty binding site and you will have to repeat the experiment with more ligand.

Sean Seaver: Crystallization Resources - P212121

P. Hubbard: Crystal Screen Optimization Server
   It is a GUI written by P Hubbard to calculate 2D gradients for crystal screen optimization (much like a web 2.0 version of Hampton’s “Make Tray” utility). It gives a handy print out of recipes and records for people who make their optimization trays by hand.

CCP4 wiki: Crystallization of Protein-DNA Complexes

Annie Hassell: Hapton Research Crystal Screens 1 & 2 as Addititives
    Annie Hassell and co-workers at Glaxo Smithkline have had lots of success using the Hampton Research Crystal Screens (HRCS) 1 & 2 as additives. Sometimes these improve the crystal quality whereas the 96 Additive Screen does not. What they have not had a great deal of success with is determining which component of the added reagent is responsible for improving the crystal quality. Does not appear that it is just one component in many cases.

They generally use 5% of the sparse matrix screen added to the well. When they get a hit, they further optimize the concentration of the reagent to add to the well.

Drs. Jichun Ma and Di Xia (NCI): The Use of Blue-Native PAGE in the Evaluation of Membrane Protein Aggregation States for Crystallization
Crystallization has long been one of the bottlenecks in obtaining structural information at atomic resolution for membrane proteins. This is largely due to difficulties in obtaining high-quality protein samples. One frequently used indicator of protein quality for successful crystallization is the monodispersity of proteins in solution, which is conventionally obtained by size exclusion chromatography (SEC) or by dynamic light scattering (DLS). Although useful in evaluating the quality of soluble proteins, these methods are not always applicable to membrane proteins either because of the interference from detergent micelles or because of the requirement for large sample quantities. Here, we report the use of Blue Native-PAGE (BN-PAGE) to assess aggregation states of membrane protein samples. We demonstrate a strong correlation between the monodispersity measured by BN-PAGE and the propensity for crystallization of a number of soluble and membrane protein complexes. Moreover, we show that there is a direct correspondence between the oligomeric states of proteins as measured by BN-PAGE and those obtained from their crystalline forms. When applied to a membrane protein with unknown structure, BN-PAGE was found useful and efficient for selecting well-behaved proteins from various constructs and in screening detergents. Comparisons of BN-PAGE with DLS and SEC are provided. (Full Article)

Dr. Susan Buchanan (NIDDK): Crystallization of Integral Membrane Proteins
    X-ray crystallography has become a very powerful tool for determining the structures of integral membrane proteins, with almost 200 unique membrane protein structures solved as of December 2007 (for a complete list, see Stephen White’s summary at However, membrane protein structures still represent less than 1% of all structures in the PDB. The major bottlenecks in the field are the expression of sufficient quantities of functional membrane proteins and the growth of well ordered crystals for X-ray analysis. This short review covers only aspects pertaining to crystallization. Our approach to membrane protein crystallization was recently written up for Current Protocols in Protein Science (Unit 17.9, available from the NIH library website). An interesting review has been written by Patrick Loll. (Full Article)

Editor: Tantalum Cluster [(Ta6Br12)2+] Derivatization Kit - Available
Check out for information.

Wei Yang (NIDDK): Crystallization of Protein-DNA Complexes (updated 2007)
    Macromolecular interaction is essential, necessary and unavoidable in a living organism. Specific interactions among macromolecules are required for molecular machinery assembly and for progression and regulation of metabolic reactions.  To fully understand a biological process, it is essential to determine the atomic structures of and interactions among components of a macromolecular complex and to decipher how these structures and interactions change during a reaction or signaling cycle.  Some macromolecular complexes are naturally stable, for example tetrameric hemoglobin, nucleosome, and ribosome. But most macromolecular complexes are formed only transiently, e.g. an enzyme and substrate complex, a growth factor and its receptor interaction, or transcription factors assembled on a promoter.  To determine structures of macromolecular complexes, whether stable or transient, has become a common practice of structural biologists in the 21st century. (Full Article)

Editorial: The Silver Bullets: At the ACA 2006, Bob Cudney (Hampton Research) and Alexander McPherson (University of California Irvine) presented an alternative stretage for crystallizing macromolecules, as they put it, by searching the silver bullets. Examples of the silver bullets incude hexammine cobalt (III) chloride, 1,3-propanediol, sebacic acid, 4-aminobezonic acid, terephthalic acid, arginine, pentaglycine, glycerol 2-phosphate, trans-aconitic acid, trimesic acid, and putrescine. As you may realize, they are in fact additives. They tested 120 additives in the crystallization experiment of 81 proteins using two fundamental conditions: (1) 30% w/v PEG 3350, 0.1 M HEPES pH 7.0; and (2) 50% TacsimateTM pH 7.0. The succesful rate was very impressive: 65 out of 81 (85%) proteins crystallized. Most significant was that 35 of the 65 (54%) crystallized only in the presence of one or more reagent mixes, but not in control samples lacking any additives!

Peter D. Sun (NIAID): Crystallization of Protein-Protein Complexes
As crystallographers zest for larger and larger molecular machinery (spoiled public is part to blame), what used to be a novelty: crystallization of protein-protein complexes, has indeed become a way of life for most of us. Like everyone else, our lab often struggles to obtain that elusive crystal of so-and-so complex. Over the years, it became clear to us that protein complexes often favored certain conditions of crystallizations. Although they were not as well defined as DNA-protein complex crystallizations, these conditions appeared more narrowly distributed than those for general soluble proteins. This lead us to conduct a survey on protein-protein complex crystallizations a few years ago, which resulted in a 48-condition sparse matrix screening kit. More recently, we revisited the survey using a much larger database of published structures and expanded the initial 48-condition to a 96-condition sparse matrix kit (Radaev, Li, and Sun, Acta Cryst. D62:605-612).

Recommended Readings: Rational Protein Crystallization by Mutational Surface Engineering; Anomalous-scatterer-mediated crystal-packing interactions; Strategies in making cross-linked enzyme crystals.

Dr. David Waugh (NCI): When an otherwise well-behaved protein fails to crystallize, what do you (suggest to) do? It is not uncommon for proteins to have disordered termini, which may impede the formation of crystals. Therefore, when an otherwise well-behaved protein fails to crystallize, the first thing we do is subject it to limited proteolysis with thermolysin. We prefer thermolysin because its major specificity determinant is a hydrophobic residue in the P1’ position. Hydrophobic residues occur much less frequently than arginine and lysine in solvent-exposed loops of proteins. Consequently, thermolysin is less likely than trypsin or chymotrypsin to yield misleading results. We have also found that in general, digestion patterns generated by thermolysin are cleaner, which simplifies the identification of the digestion products. Metastable digestion products can be identified by mass spectrometry and N-terminal amino acid sequencing, and then new vectors can be constructed to overproduce the truncated polypeptides. Secondary structure prediction and sequence alignments can sometimes used to make educated guesses about the locations of domain boundaries when the proteolysis approach fails to yield definitive results. We also recommend trying reductive alkylation with formaldehyde and dimethylamine-borane complex (Rayment, I. Methods Enzymol. 276, 171-179 [1997]). The net result of this reaction is the dimethylation of all accessible lysine side chains (and the N-terminal amino group). Although this does not change the intrinsic charge of a protein, it may alter its isoelectric point slightly. One rational behind this strategy is that dimethylation of lysine side chains will reduce their interaction with solvent, thereby causing them to adopt more “ordered” conformations that may facilitate crystallization. Reductive methylation also frequently reduces the solubility of proteins, and so it may be a good approach to try when mostly clear drops are obtained from crystallization screens even with very concentrated solutions of protein. The nice thing about this approach is that it can be performed on the existing sample of protein (i.e., no new constructs need to be made). Surface entropy reduction mutagenesis, a strategy pioneered by Zygmunt Derewenda and coworkers, is another option. In this method, linear clusters of amino acid side chains with high conformational entropy (e.g., Lys and Glu), which are presumed to lie on the surface of the protein, are replaced by methyl groups (Ala) in an effort to create new epitopes that will facilitate crystallization. A growing number of proteins have been crystallized in this manner, suggesting that the method may be of general utility. Yet, because is impossible to predict which cluster mutant(s) will crystallize, the probability of a successful outcome is proportional to the number of mutants that are screened. Consequently, surface entropy reduction mutagenesis can be a very labor intensive undertaking. Finally, if the opportunity exists, working with multiple orthologs of a target protein generally improves the odds of obtaining crystals.

Tutorial: The pictorial library of crystallization drop phenomena

Dr. Wei Yang (NIDDK): Crystallization of protein-DNA complexes  

Macromolecular interactions within a living organism achieve many remarkable feats from formation of complex multi-molecular machines to coordination of cascades of signaling or metabolic events.  To fully understand biology one must understand the structure and interplay between components of macromolecular complexes and reaction pathways.  Some macromolecular complexes are stable, such as tetrameric hemoglobin, chromatin, and the ribosome.   Most macromolecular complexes are formed only transiently however, notable examples being an enzyme and its substrate, a growth factor and its receptor, and a transcription factor and its DNA recognition site.  To determine structures of macromolecular complexes, whether stable or transient, has become the goal of structural biologists in the 21st century.

Protein-DNA complexes were among the first macromolecular complexes characterized by X-ray crystallography.  Since the isolation of the lac repressor with the lac operator by Gilbert and Müller-Hill, understanding the nature of specific protein-DNA interactions has captivated and occupied many scientists.  How are protein-DNA interactions turned on and off?  How do such interactions alter the involved macromolecules, protein or DNA, so that they initiate the next reaction step?  The crystallographic studies of repressor-DNA and CAP-DNA complexes in the 1980s and early 1990s (Anderson et al., 1987; Jordan and Pabo, 1988; Otwinowski et al., 1988; Schultz et al., 1991) paved the road to pursuit of more complex and intricate protein-DNA assemblies in the last 10 years. In the following sections, I will describe strategies and current approaches employed to obtain cocrystals of protein-DNA complexes. [061]

Mr. Jerry Alexandratos (NCI): Conditions 49 and 50 of HRCS 1. Many people prefer to use only 48 out of 50 conditions from Hampton Research Crystal Screen 1, thus only using two 24-place plates. However, they decide arbitrarily not to use conditions 49 and 50. Take note that these two are rather good conditions for crystallization, even better than about 12 other conditions of the same Screen. See the "success profile" chart on page 11 of the 2003 HR catalog. If you still want to discard two, then select them scientifically. I recommend removing #25 (zero hits!) and either #13 or #27 (10 or fewer hits each).

Convenience is fine, but you may arbitrarily make bad choices. I made this same argument to a company representative who wanted to sell us a modified 48 solution crystallization screen for use with a crystallization robot. That company made the same choices, keeping #1-48. Companies (made up of people) are not infallible either. Keep this in mind, no matter what experiments you're running or what you're asked to buy.

Dr. Peter Sun (NIAID): Crystallization of protein-protein complexes. Protein-protein complexes are, in general, harder to crystallize than their non-complexed components due to their lesser solubility and instability. For example, it maybe difficult to achieve a 10 mg/ml concentration for protein-protein complexes while the solubility of their components could exceed well beyond 10mg/ml concentration. A blind application of the available crystallization screening kits often results in precipitation in majority of the conditions. Even if one can reach a desired concentration for a protein complex sample, the stability under which a complex remains often restricts its crystallization configuration space. This is mostly due to a lesser energy required to disrupt a protein complex formation than it is required to disrupt a protein structure or folding. As such, the crystallization configuration space is less restricted for tight binding protein complexes, such as antibody-antigen and cytokine-receptor complexes, and is more restricted for weakly interacting complexes, such as many transient signal transduction complexes that are at best of micromolar binding affinity. Thus, the principle concern of crystallizing a protein-protein complex is to maximize the complex solubility and limit to the conditions under which the complex remains associated. A brief survey of the crystallization conditions for 200 published protein complexes showed certain preference in their crystallization conditions compared to single proteins (1). Here is a summary of our findings: (A) The majority conditions are PEG related with ammonium sulfate represents less than 20% of the cases. As for the precipitant concentration, the average PEG used is 10-15%, which is somewhat lower than the 20-30% PEG represented in many screening kits. (B) Most of them are crystallized at or near neutral pH values with less than 10% cases having pH lower than 5 or higher than 8.5. (C) Salt concentrations are general less than 300 mM, except when ammonium sulfate is used. (D) A crystallization screen kit was assembled to represent the most favorable conditions for obtaining protein complex crystals.        

Reference: 1. Radaev S., and Sun P. Crystallization of protein-protein complexes. J. Appl. Cryst. (2002), 35:674-676.

Dr. Traci Hall (NIEHS): Crystallization of Protein-RNA complexes. Xinhua asked me to write something about how to grow crystals of protein:RNA complexes, so here are some tips from our lab.  (1) RNA source.  We order our synthetic RNA oligos from Dharmacon Research (  I’m told that TriLink is also a good source ( (2) Determining the minimal piece of RNA that is sufficient for binding.  A first analysis can be done with a standard band shift or filter-binding assay, but the shortest piece of RNA that binds in these assays can sometimes be longer than what is necessary for crystallization.  So, it is useful to analyze the ability to form complexes at crystallographic concentrations by gel filtration chromatography.  We use a Superdex 75 or 200 HR 10/30 column from Amersham Pharmacia for this, and it is helpful to monitor at 260 and 280 nm, if possible.  (3) Forming complexes for crystallization.  We use gel filtration chromatography to analyze complex formation at crystallographic concentrations and determine the ratio of protein:RNA to produce stoichiometric complexes.  Alternatively, we calculate a ratio and set up crystallization trials at slightly varying ratios.  It is also possible to purify protein:RNA complexes by gel filtration for crystallization.  (4) Crystallization screens.  Our favorite screens are Natrix and MembFac from Hampton Research.

This site is maintained by Dr. Xinhua Ji ( on the NCI-CCR-MCL server (