For decades, structural biology has operated under a strict prerequisite: to map a molecule's three-dimensional shape at atomic resolution, you must first convince billions of copies of it to line up into a perfect, rigid crystal lattice. X-ray crystallography has yielded thousands of structures this way. However, this requirement introduces a severe selection bias. Many of the most biologically critical molecules—such as hydrophobic membrane proteins, highly flexible enzymes, and transient macromolecular complexes—stubbornly resist crystallization. They aggregate, unfold, or precipitate when extracted from their native environments.
To bypass this physical bottleneck, a powerful research direction has emerged. By combining alternative physical measurements, evolutionary data, and computational modeling, researchers are learning to determine high-resolution protein structures without ever needing to grow a crystal.
Why the Crystal Lattice Fails
To understand why alternative methods are necessary, we must look at why certain proteins are "recalcitrant" to traditional structural pipelines. Consider membrane proteins like G protein-coupled receptors (GPCRs), which are embedded in a hydrophobic lipid bilayer. As noted by Yang Wang et al., when these proteins are stripped of their native lipid membrane for purification, they frequently lose their structural integrity and aggregate.
Similarly, large modular enzymes or proteins involved in inhibition pathways often suffer from extreme flexibility or a tendency to precipitate during the high-concentration preparation phases required for crystallography or solution NMR (Jingmin Gu et al.; Hai‐Xi Hu et al.). Even complex organic metabolites can fail to form the high-quality crystals needed for standard diffraction, forcing researchers to turn to alternative spectroscopic and computational modeling techniques to resolve their 3D configurations (Philip W. Le Quesne et al.).
Diffraction Without Crystals
If growing a large crystal is the primary bottleneck, can we eliminate the crystal entirely? This is the core intuition behind Single-Particle Diffractive Imaging (SPI) using X-ray Free-Electron Lasers (XFELs).
In a traditional experiment, a crystal amplifies the weak diffraction signal of a single molecule by repeating it billions of times. SPI replaces this physical amplification with raw laser power and speed. By hitting an isolated, non-crystalline particle with an intense, ultrashort X-ray pulse, researchers can capture a diffraction snapshot before the molecule is vaporized by radiation damage—a principle known as "diffract-before-destroy."
As demonstrated by D. Starodub et al., spatial frequency correlation analysis can reconstruct 3D structures from these noisy snapshots without knowing the particle's orientation beforehand. To make this practical for delicate biological specimens, Anna Munke et al. leveraged the Linac Coherent Light Source (LCLS) to capture high-quality diffraction data from single virus particles, establishing a path toward near-atomic resolution imaging in native, non-crystalline states.
Molecular Rulers and Hybrid Modeling
When direct imaging is impossible, another powerful strategy is to gather sparse physical constraints from the protein in its native environment and use computational algorithms to fill in the blanks.
Consider chemical cross-linking mass spectrometry (XL-MS). Instead of forcing a protein into an artificial crystal, researchers introduce bifunctional chemical reagents into a solution (even complex fluids like blood serum). These reagents act as "molecular rulers," covalently bonding to pairs of amino acids that happen to be physically close in 3D space. By digesting the protein and analyzing the cross-linked fragments via mass spectrometry, researchers map out a network of maximum distance constraints.
[Residue A] <------- Chemical Cross-linker -------> [Residue B]
(Max Distance: ~20 Å)
Adam Belsom et al. demonstrated that using highly promiscuous cross-linking reagents provides enough dense spatial constraints to model protein domains de novo. Furthermore, Adam Belsom et al. argued that integrating these experimental distance constraints with computational docking and structure prediction algorithms creates a robust "hybrid" methodology capable of solving structures that are completely intractable by X-ray crystallography or NMR alone.
Reading Structure from Evolutionary History
The constraints do not even have to come from a wet lab; they can be inferred from the history of life itself. If two residues in a protein complex are physically touching, a mutation in one residue will likely disrupt the interaction unless a complementary mutation occurs in the other. This is sequence co-evolution.
By analyzing thousands of homologous genetic sequences, computational frameworks can identify pairs of amino acids that evolve in lockstep. Thomas A. Hopf et al. leveraged this intuition with EVcomplex, showing that evolutionary co-variation yields highly accurate 3D contact maps. This allows the direct calculation of protein complex architectures without relying on physical crystallization or resource-intensive screening.
Engineering Around the Bottleneck
Rather than abandoning traditional methods entirely, some researchers focus on re-engineering the proteins or their delivery mechanisms to make them cooperative.
For membrane proteins, the challenge is maintaining stability during delivery to the X-ray beam. Uwe Weierstall et al. developed a lipidic cubic phase (LCP) injector that continuously streams microcrystals embedded in a membrane-mimicking lipid matrix directly into an X-ray laser beam. This drastically reduces the sample volume required, mitigating the low-yield expression bottleneck characteristic of membrane targets.
Alternatively, the protein itself can be computationally optimized for stability. Yang Wang et al. paired homology modeling with molecular dynamics simulations to identify single-point mutations that significantly increase the thermal stability of fragile GPCR complexes, transforming them into viable candidates for successful crystallization.
Key Foundational Work
Structural and synthetic studies of pithomycolide (1994) — contributed 2 concept elements to this direction.