Fractal Aggregates in Protein Crystal Nucleation Bin Chen,* Ricky B. Nellas, and Samuel J. Keasler Department of Chemistry, Louisiana State UniVersity, Baton Rouge, Louisiana 70803-1804 ReceiVed: December 6, 2007 Monte Carlo simulations of homogeneous nucleation for a protein model with an exceedingly short-ranged attractive potential yielded a nonconventional crystal nucleation mechanism, which proceeds by the formation of fractal, low-dimensional aggregates followed by a concurrent collapse and increase of the crystallinity of these aggregates to become compact ordered nuclei. This result corroborates a recently proposed two-step mechanism for protein crystal nucleation from solution. 1. Introduction Understanding the nucleation phenomenon, in which a new phase is born out of a metastable supersaturated phase, has important implications in many scientific and technological areas. As applied to proteins, the importance of this problem has been recently re-emphasized by the rapid growth of structural proteomics. 1 In particular, the preparation of high- quality crystals from protein solutions has become a bottleneck for the structural determination of new classes of biomolecules. 2 Therefore, over the past decade, finding the optimal conditions and/or physical mechanisms for enhanced protein crystal nucleation has been the major theme for a great deal of protein crystallization research, which, in turn, has illustrated a rather intriguing side of this phenomenon. A decade ago, George and Wilson 3 demonstrated that the second osmotic virial coefficient could be used as an effective predictor for achieving successful crystallization of proteins. For a variety of proteins, the solvent conditions that favor crystal- lization confine B 2 to a narrow range of small negative values. While large negative values of B 2 lead to protein aggregation instead of crystallization, large positive values of B 2 disable the crystallization completely. This interesting correlation im- mediately inspired some theoretical work. Rosenbaum, Zamora, and Zukoski 4,5 showed that this surprising result arises partly from the unusual phase behavior of protein solutions, i.e., the existence of a metastable fluid-fluid coexistence curve below the fluid-solid coexistence curve, due to the short-range attractive interactions between proteins (see Figure 1). 6-8 In addition, they mapped the solvent conditions onto the phase diagram of protein solutions and observed that proteins crystal- lize in a narrow temperature range, equivalent to the narrow range of small negative B 2 . 8,9 To gain microscopic insight into the physical mechanism of this process, ten Wolde and Frenkel 8 performed the first protein crystal nucleation simulation. They found that the free energy barrier for crystal nucleation is strongly reduced at the meta- stable fluid-fluid critical point (CP). The underlying physical mechanism was attributed to the large density fluctuations around the CP, which lead to the initial formation of liquid- like droplets (density fluctuation) followed by crystal nucleation (structure fluctuation) inside these high-density droplets. A similar conclusion was also drawn by Talanquer and Oxtoby with density functional theory. 10 However, the prediction of nucleation enhancement at the CP was soon questioned by other theoretical and experimental work. 11-13 The main problem is that protein solutions gel at the CP as a result of the high concentration and the associated high viscosity. In addition, experiments using lower concentrations of lysozyme also observed crystal nucleation enhancement, but at temperatures close to the fluid-fluid coexistence temperatures. 2 The fact that this enhancement follows along the fluid-fluid coexistence curve at different concentrations led Vekilov 14 to suggest that density fluctuations (i.e., the formation of liquid-like droplets) are not just a part of the crystal nucleation mechanism around the CP. Therefore, a general two-step mechanism was proposed in which density fluctuations occur first, followed by the structural fluctuations, over a much broader range of phase space than previously thought. Although this mechanistic proposal provides an explanation of the protein crystal nucleation enhancement observed there under different temperature condi- tions, it requires further theoretical support, as, until now, the superposition of these two fluctuations was only revealed at the CP. 8 More importantly, many of the microscopic details regarding how these two fluctuations are superimposed onto each other remain unclear. Even simpler questions about the nature of these liquid-like intermediates have barely been addressed. For example, what do the quasi-droplets look like? Could they be doubly metastable, i.e., metastable with respect to both the solution and the crystals as Vekilov and co- workers suggested? 14,15 Also, what are the relative barriers that the two fluctuations have to overcome in the crystal nucleation landscape? To address these questions, we extended the aggregation- volume-bias Monte Carlo approach, which has led to our recent * Corresponding author. E-mail: binchen@lsu.edu. Figure 1. (a) A typical protein phase diagram showing a metastable fluid-fluid coexistence region (diamonds) below the solubility line (circles). 6,7 (b) A model potential with short-range attractive interactions reproducing this behavior. 8 (c) Comparison of this potential (dashed- dotted line) to the common LJ potential (dashed line). 4725 J. Phys. Chem. B 2008, 112, 4725-4730 10.1021/jp8002728 CCC: $40.75 © 2008 American Chemical Society Published on Web 03/22/2008