Liquid Phase Biological Electron Microscopy: Many Published Results and Claimed Benefits Are Fantasy, Not Fact

Abstract

While the idea of imaging biological molecules by electron microscopy in the liquid phase might seem to be quite attractive, fundamental problems inherent in this approach preclude success at high resolution. One attractive goal, for example, is to image macromolecular machines in action, but radiation-inactivation of enzymatic function severely limits the electron exposures that can be used. Furthermore, although nanometer resolution or better has been claimed in some papers for macromolecular complexes said to be freely suspended in the liquid phase, Brownian motion must limit the achievable resolution to dimensions that are much larger than the macromolecules themselves. While Brownian motion can be avoided by adsorption of particles to a substrate or a thin window, there still is a risk that air drying occurs. We have analyzed publicly-available EM images of GroEL that were putatively obtained in the liquid phase, and show that these particles clearly had been dried. While the original authors argued that the contrast reversal seen in their images must have been due to liquid water scattering 120 keV electrons much more strongly than either vitreous ice or proteins, we show that the particles were likely negatively stained, presumably by remaining buffer salts.

Keywords: biological samples; electron microscopy; liquid phase; protein structure.

Figure 1.

A comparison of Rotavirus Double Layered Particles (DLPs) images by cryo-EM (left) and in “liquid” (right), as published in Kelly (2025). The contrast of the particles (black on a white background) is the same for both. Reproduced under the Creative Commons Attribution 4.0 International License.

Figure 2.

2D class averages 0f GroEL from (a) cryo-EM images (EMPIAR-12667) and (b) “liquid” phase images of Kong et al. No class averages from the Kong et al. images showed the highly characteristic side view of GroEL, seen in both negative stain and cryo-EM, with the four stripes. The class averages in (a) came from ~ 1.5k particles, while those in (b) came from ~ 102k particles.

Figure 3.

Comparison of GroEL reconstructed volumes. (a) The C7 symmetry volume from Kong et al. (b) the D7 symmetry volume from Kong et al. Both (a) and (b) have been filtered to 15 Å resolution. (c) The cryo-EM reconstruction of GroEL (EMD-45079) filtered to 15 Å and (d) the cryo-EM volume at 3.4 Å resolution. Shown are side views (top) and top views (bottom). The most striking difference between the reconstructions in (a) and (b) and the “ground truth” volumes in (c) and (d) is that (a) and (b) are squashed, turning a cylindrical GroEL into something closer to a sphere. Such a deformation is not unexpected from dried samples.