X-Ray Laser Reconstructs a 1-MDa Membrane Protein in Vacuo

Single-particle imaging with X-ray free-electron lasers has long aimed at molecules too small for conventional diffraction experiments, but residual gas scattering makes that goal difficult for weakly scattering biological particles. Previous XFEL single-particle demonstrations have therefore concentrated on larger, higher-contrast targets such as viruses, organelles, and cells. The authors report a three-dimensional reconstruction of isolated photosystem I (PS I) trimers, a roughly 1-MDa membrane-protein complex, in vacuum.

Core Contribution

The central advance is an injection and background-control scheme that makes the diffraction signal from an isolated membrane-protein complex usable at an XFEL. PS I trimers are aerosolised by charge-reduction electrospray ionisation, then transferred into the X-ray interaction region. Replacing most of the nitrogen and carbon-dioxide sheath gas with helium in the low-pressure injector region reduces measured background scattering by 80%.

That reduction matters because it addresses the experimental bottleneck rather than a reconstruction algorithm alone: for particles of this size, diffuse gas scattering can obscure the weak coherent signal before classification or phasing begins. The paper consequently demonstrates an end-to-end measurement, from electrospray delivery through diffraction classification and three-dimensional phase retrieval, on a membrane protein in vacuo.

Technical Approach

Figure 1 lays out the apparatus. Charged droplets containing PS I trimers are charge-reduced with bipolar ions generated by a vacuum-ultraviolet source, focused through an aerodynamic lens, and intersected with the European XFEL pulse train. Helium enters between the skimmer box and lens, lowering the gas-scattering contribution at the interaction point. Diffraction was recorded 536 mm downstream on a 1-megapixel DSSC detector; the detector flange restricted the usable scattering angle to 15.6°, giving a maximum momentum transfer of $q = 0.26\,\mathrm{nm}^{-1}$ and a geometry-limited resolution of 3.8 nm.

The authors classified diffraction patterns into particle arrangements before reconstruction. Representative classes distinguish a single trimer viewed edge-on or face-on from stacked and side-by-side pairs. They assigned 32,788 patterns to single PS I trimers, assembled those data into a common three-dimensional intensity volume with the EMC algorithm implemented in Dragonfly, and imposed the trimer's three-fold rotational symmetry at the intensity-assembly stage. Phase recovery then combined difference-map and error-reduction iterations; no symmetry was imposed during phasing.

Figure 2 is useful evidence that the sorting step separates the intended single-particle population from multi-trimer events. The inferred single-trimer projections and their 20-nm scale bars also make clear that this is a low-resolution shape reconstruction, not a molecular model.

Results and Analysis

The reconstructed density is disc-shaped, about 22 nm across and 10 nm thick, consistent with a PS I trimer embedded in a detergent micelle. Face-on and edge-on projections, along with central density slices, show three higher-density regions around the disc centre. Their arrangement agrees with the three monomers expected for the trimer, although the authors explicitly caution that these features sit near the resolution limit.

Figure 3 also reports a phase retrieval transfer function calculated over 1,000 independent phasing runs, with the $1/e$ threshold used to assess resolution. The reported 3.8-nm resolution is therefore not an atomic or subunit-resolving result; it is bounded primarily by detector geometry. Still, it is enough to recover the overall particle envelope and broad trimeric organization from randomly oriented, isolated complexes.

To assess whether vacuum and detergent altered the sample substantially, the authors ran a 1,500-ns molecular-dynamics simulation of PS I in a DDM micelle in vacuum. The measured envelope is reported to agree with the simulated compaction and with native mass-spectrometry observations. This supports the narrower claim that the detected particles remain structurally interpretable under the injection conditions, rather than proving preservation of fine native structure.

The evidence supports a meaningful experimental demonstration: background suppression, tens of thousands of classified single-particle patterns, and an internally checked reconstruction all point in the same direction. Its practical significance is conditional. The work establishes a route to imaging membrane-protein complexes that had been below the usual scattering threshold, but the present setup cannot yet distinguish molecular detail beyond a few nanometres. Better angular coverage and higher-resolution data will be required before this approach can answer structural questions normally addressed by cryo-EM or crystallography.