Crystallization of protein-ligand complexes, where the ligand has limited or no water solubility can be tricky. Here are some suggestions.

A review on the crystallization of protein ligand complexes can be found in Crystallization of protein–ligand complexes, Annie Hassel et al, Acta Cryst. (2007). D63, 72–79.

Vary the ratio of protein to ligand Try 1:3, 1:5 or 1:10 ratios of protein: ligand. This can make a huge difference. There are instances where the organic solvents (DMSO, Ethanol) used to dissolve the ligands interfere with the crystallization process. In such instances, try adding the dry compound to the crystallization drop with the protein. Or add the dry compound to the protein solution and put on a shaker overnight with gentle rotation. Centrifuge the next morning and set up your crystallization screens. It does not matter if it does not appear if the solid compound dissolved. Enough solid compound may go into solution to effect ligand binding.

Successful complexation depends on the concentration of protein, ligand, and the Kd of the protein-ligand complex. For Kd much greater than protein concentration, try the ligand concentration at greater than 10 x Kd. As Kd approaches the protein concentration, roughly superstoichiometric quantities will be sufficient for full occupancy. For Kd less than protein concentration, stoichiometric quantities of ligand will typically suffice. In general, the ligand concentration should be such that it is near saturation on the binding isotherm.

Solubilize the ligand in DMSO so it is maximally concentrated (100mM works fine). Add enough ligand to achieve two to three fold excess ligand to protein. Keep the DMSO concentration to no more than 3% to avoid damaging the protein. Set crystallization experiment using this sample. If you cannot achieve a high enough stock concentration of DMSO to be below the 3% threshold, dilute you protein in the storage buffer to approximately 1mg/ml. Add compound to two to three fold excess, incubate and co-concentrate to the desired concentration. This may help to avoid the DMSO shock. Alternatively one can incubate the concentrated protein with the compound solubilized in water for 24 to 48 hours and solubility of the ligand will be sufficient to complex with the protein. (Carsten Schubert ccp4bb December 2007)

One method that worked for me was to dissolve my ligand in 100% DMSO, as suggested in the previous response, then add a 3 molar excess of ligand to protein so that the final concentration of DMSO in the protein-ligand solution was no greater than 10% - of course the maximum concentration of DMSO that your protein can suffer will be protein specific but you could investigate this by incrementally adding DMSO to just a solution of your protein at your working crystallization concentration then measuring scattering at 600nm in a spectrophotometer to determine the critical DMSO concentration that causes your protein to precipitate (if you have sufficient protein to 'waste'!). If your ligand binds tightly to your protein at an equimolar concentration you can then remove excess ligand and DMSO by passing your sample through a G25 Sephadex column. (Rob Hussey ccp4bb December 2007)

If you grow crystals in polyethylene glycols or similar reagent you might try to solubilize the compound in a small amount of this reagent. This is helpful if you want just a 1:4 protein:ligand ratio. Sometimes solubility is low even in 5% DMSO (or diluted solutions of glycerol, alcohols and similar molecules). In these cases setting up drops in the presence a saturated solution and some precipitate of the compound may also lead to good co-crystals. As one molecule passes from the saturated solution to the “bound” state, a new molecule is solubilized from the precipitate, which gradually dissolves and passes from the solution to its binding site in the protein. Indeed, this sounds like a “soaking” experiment and it works well if the compound is colored, so that you can see if the crystals actually become of the same color. Just remember to wash them thoroughly before measurements, in order to remove traces of the ligand precipitate that would result in poor diffraction. (Marco Mazzorana ccp4bb December 2007)

We routinely obtain structures from protein solutions with a big pellet of ligand in the bottom of the tube. For co-crystallizations we add 1mM compound to a 0.3mM solution of the protein and incubate overnight. Many of the compounds are only soluble to 50 micromolar, so we get a lot of precipitate. The next day, we spin the tube at high speed, and use the supernatant for crystallization trials. We have started from 100 mM stocks in 100% DMSO or ethanol. This has worked for compounds ranging for picomolar to micromolar affinity, which surprised us, but it worked. (Kendall Nettles ccp4bb December 2007)

We normally prepare our ligand stocks in DMSO and add this to the protein in 3-fold molar excess. The majority of our ligands are quite insoluble and precipitate when the DMSO concentration decreases upon addition to the protein……. so I am not surprised that you are seeing this. If your compound does not bind your protein tightly, you might consider using a 5-fold molar excess of ligand.

Some proteins crash out if the protein concentration in high when you add the ligand. For those situations, we complex the ligand with dilute protein (1-2 mg/ml), and then concentrate this for crystallization trials. I have had proteins where we had to complex the dilute protein with ligand, and then let it sit overnight at 4 degrees Celsius before we concentrated the protein. We normally incubate the protein and ligand at 4 degrees Celsius for 1-3 hours for binding before we set up the crystallization experiments. Another scenario might be addition of ligand to the protein followed by incubation at room temp for about 1 hour. Then centrifuge at 4 degrees Celsius, keep protein at 4 degrees Celsius and set up your trays. (Annie Hassell ccp4bb August 2011)

Try solubilizing the ligand in low molecular polyethylene glycol (PEG 200, PEG 300 or PEG 400).

When adding ligands to a drop containing crystal, consider optimizing for smaller crystals. 10-20 micron crystals can reach 80% occupancy in less than 30 seconds. A short soak time can minimize or prevent crystal damage. Larger crystals can equal longer soak times can equal higher risk for crystal damage.

Recommended reading:
Guidelines for the successful generation of protein–ligand complex crystals. Ilka Muller, Acta Crystallogr D Struct Biol. 2017 Feb 1; 73(Pt 2): 79–92.