Although high-resolution electron cryo-microscopy (cryo-EM) of purified biological macromolecules (Glaeser et al., 2021) has become a successful and mature method, preparing the required thin, vitrified samples often remains a major challenge. Efforts have been under way for some time to improve the way in which samples are thinned before they are vitrified, as has recently been reviewed by Weissenberger et al. (2021). Nevertheless, the classic goal of embedding particles within free-standing films of buffer, as depicted in the cartoon shown as Figure 2 of (Glaeser, 2021), appears to be rarely achieved. Instead, the desired encapsulation of biological macromolecules within vitrified buffer is generally thwarted by rapid diffusion of proteins to the air-water interface (AWI), often followed by formation of a sacrificial, denatured-protein monolayer (Han and Glaeser, 2021). While subsequent adsorption of additional particles to such sacrificial monolayers may still result in an acceptable outcome, it often does not, and better alternatives are generally needed.

Interaction of proteins with the AWI was considered to be a potential hazard in the early days of cryo-EM sample preparation—see section 6.6 of (Dubochet et al., 1988). Indeed, attempts were made at that time to develop some type of electron-transparent slide and coverslip that might be suitable for use in cryo-EM. Exploratory directions of work included sandwiching samples between thin, hydrophilic support films (Taylor and Glaeser, 1973; Taylor and Glaeser, 1976) and the use of fatty-acid monolayers as a conformal coverslip (Hayward et al., 1978; Chang et al., 1985a; Chang et al., 1985b).

Nevertheless, the simplicity and the success of blotting a holey support film with filter paper initially caused such concerns to be put aside.

The issue was reopened, however, in a retrospective of cryo-EM sample preparation—see Figure 5 in (Taylor and Glaeser, 2008). Awareness then continued to grow that the challenges encountered when preparing samples on grids were due to interaction with the AWI, rather than to how the sample was isolated. Perhaps the foremost indication was that the number of particles seen in images was often either more than, or fewer than, what would be expected from the known sample concentration and the thickness of the sample (Vinothkumar and Henderson, 2016). Finally, after cryo-tomography demonstrated that nearly all types of biological macromolecules were adsorbed to the AWI (Noble et al., 2018; D'Imprima et al., 2019; Fan et al., 2019), the problem again began to be discussed more widely in the literature (Drulyte et al., 2018; Carragher et al., 2019; Klebl et al., 2020). It now is widely acknowledged that avoiding interaction with the AWI remains one of the most important challenges still to be solved for single-particle cryo-EM.

This Perspective focuses on the strategy of immobilizing particles onto the surface of a thin support film, preferably followed by washing off unbound particles with buffer, and then removing all but a suitably thin layer of the wash-buffer. Other potential strategies for avoiding interaction with the AWI, which are not reviewed here for lack of space, include 1) thinning and then vitrifying samples so quickly that adsorption to the AWI does not have time to occur; 2) sandwiching samples between structure-friendly, electron-transparent windows, in effect confining particles between some type of electron-transparent “slide and coverslip” (Frederik et al., 1989); and 3) milling or sectioning thin samples from bulk-frozen material.

More specifically, this Perspective addresses two issues that remain relevant after macromolecules have been immobilized onto the surface of a grid. The first of these is the fact that immobilization does nothing that might reduce the unwanted variation in ice thickness that is produced by traditional blotting with filter paper. Second, although immobilization prevents diffusion of particles to the AWI, it does nothing to prevent the AWI from still touching the particles, should the thickness of buffer become comparable to, or less than, the size of the bound particles themselves.

Some macromolecules will hit and perhaps adsorb to the AWI, even while the drop of sample is initially forming at the tip of the pipette, because particles that are within 100 nm of a newly formed aqueous surface will diffuse to the AWI within a ms or less (Taylor and Glaeser, 2008; Naydenova and Russo, 2017). Something similar is expected to happen when protein solutions wet the surface of a dip pen, and subsequently the grid surface, a system that is roughly equivalent to the glass rod historically used to quantitatively deliver denatured protein to the surface of a Langmuir trough (Trurnit, 1960). In addition, diffusion will deliver particles to the fresh AWI that then spans the exposed side of ∼micrometer-sized, open holes of the holey carbon grid. In other words, the AWI on the underside (i.e., back) of the grid is just as much of a hazard to proteins as is the larger, continuous surface over the top of the deposited sample.

As mentioned in the Introduction, one alternative is to bind samples onto the surfaces of support films. Furthermore, excess sample might be washed from such grids, depending about the binding affinity, the intent being to remove proteins adsorbed to the AWI at the top of the applied sample. There are, in fact, many ways in which biological macromolecules can be immobilized at solid-liquid interfaces in a structure-friendly way, as is known from the field of biochemical chromatography. Grids that are intended to immobilize biological macromolecules in a structure-friendly way are therefore referred to here as being “affinity grids”, a terminology that is meant to be taken figuratively rather than literally.

At the same time, there are also many types of solid substrates that are not always structure friendly, as has been reviewed briefly in the Discussion section of (Joppe et al., 2020). Thus, as is summarized in Table 1, adsorption onto solid surfaces such as glow-discharge treated carbon film or graphene oxide may produce favorable results for some proteins, while adsorption of other proteins to the same substrates results in preferred orientation or even severe particle damage. It therefore seems likely that the same, specimen-dependent outcome will prove to be true for adsorption of proteins onto the silicon nitride windows of microfluidic EM grids (Huber et al., 2022).

A large number of alternatives have already been investigated for making affinity grids, as is summarized in Table 1. Some of those strategies have already given high-resolution results, even when using samples that previously resisted preparation with standard approaches (D'Imprima et al., 2019; Gillies et al., 2021; Kasinath et al., 2021; Lahiri et al., 2019; Sauer et al., 2021; Wang et al., 2020). While all such affinity grids have attractive features, the most appealing may be ones with high binding affinities, so that—as mentioned above—grids can be washed without eluting the immobilized particles. Furthermore, it is essential that there be little risk that the chosen immobilization strategy results in preferred orientation of particles. The point being made here is that many effective strategies are already available for making affinity grids, and it is certainly welcome if more of them can be developed.

As previously indicated, one can blot affinity grids with filter paper in the traditional way, but the usual, unsatisfactory features produced by blotting will still remain. These include the fact that some sample often finds its way to the back side of grids during blotting (Armstrong et al., 2020); there are large variations in the amount of area over which the sample is effectively opaque to 300 keV electrons; there often are many areas in which the ice may seem to be relatively transparent to the electron beam, but it is still not yet thin enough to get the best result; and there are other areas where the ice is either too thin or where the buffer has even dewetted the support film, causing those areas of the grid to dry before vitrification.

Newer alternatives to blotting with filter paper, which are intended to overcome at least some if not all of these shortcomings, have recently been reviewed in (Weissenberger et al., 2021). These alternatives include different ways in which samples are sprayed onto grids, which can be either self-wicking grids (Wei et al., 2018) or conventional holey-film grids, as well as ways in which samples are spread with either a capillary (Arnold et al., 2017) or a dip pen (Ravelli et al., 2020). However, since those methods are unlikely to be compatible with a washing step, it is hard to imagine ways to remove unbound material when using affinity grids. As a result, there still is reason to seek alternative ways to produce thinned films on affinity grids.

Among the alternatives that seem to have not yet been investigated, one might think of applying some type of body force, such as the inertial force employed in spin coating (Larson and Rehg, 1997), or using a strong air flow to “blow off” unwanted buffer. Other possibilities might include creation of a gradient of surface tension from one edge of a grid to the other in order to generate Marangoni flow (Velarde and Zeytourian, 2014), or mechanically squeezing excess buffer from the sample with an electron-transparent “coverslip”, as was attempted in some of the early work described above.

In addition, conspicuous by its near-absence from the cryo-EM literature, is the idea of simply “wicking” excess buffer by touching filter paper to the edge of a grid, as is often done during negative staining. That approach is unsatisfactory when wicking is done from one edge, of course, because it leaves behind a spherical cap of liquid that is several micrometers thick (Glaeser et al., 2016). Axially symmetric draining (wicking), on the other hand, has the potential to produce a uniformly thin film across much or all of the grid. This latter approach was referred to as “blotting from the perimeter” (Armstrong et al., 2020), or in the oxymoronic description used here, “blotting with a hole”.

The principles involved in making samples that are suitable for cryo-EM are the same as those involved 1) in making stable foams (juxtaposed air bubbles), emulsions, and colloidal suspensions, as well as 2) in some technological applications that employ modern microfluidics. As a result, an extensive literature exists that addresses the formation and stability of thin liquid films. Recent reviews (Andrieux et al., 2021; Chatzigiannakis et al., 2021), for example, contain much that can inform our approach to making cryo-EM samples on affinity grids, and an older, more extensive review (Stubenrauch and Klitzing, 2003) is also worth pointing to.

Equipment used previously to observe the removal of excess buffer from 3 mm diameter coverslips (Armstrong et al., 2020) has since been modified, as is described in Supplementary Figure S1, to obtain high-speed movies that show what happens when the same coverslips are blotted with an ∼2 mm hole in filter paper. Blotting pads used for the new experiments were fabricated with a laser-beam etching tool, and these are also described in the Supplementary Figure S1.

Both the relative humidity and the flow rate are important parameters to control in an air stream that is used to assist axisymmetric draining. Too little flow will still leave a too-large “cap” of liquid at the center; too large a flow might cause the liquid film to rupture and to dewet the substrate; too low humidity will cause excessive evaporation to occur; and too high humidity may produce condensation to form.

Demonstration of the potential usefulness of blotting with a hole thus requires that similar results first be achieved when using EM grids rather than a surrogate, 3 mm coverslip. In addition, such grids would have to be vitrified and examined in an electron microscope, preferably one equipped with an energy filter, so that their thickness values can be measured, for example, by the method described in (Rice et al., 2018). The final proof of usefulness will then depend upon whether the resolution achieved in single-particle cryo-EM maps proves to be as good as, or even better than, that achieved with other grid-preparation methods, assuming that all else remains constant.

While axisymmetric draining seems to be an appropriate approach to use with affinity grids, it may be that it will also be an effective alternative to consider when using standard, holey grids any type of affinity grid could be used, of course, to avoid diffusion to the AWI. Of the options already listed in Table 1, streptavidin affinity grids have the advantage that the resolution retained in the specimen can be easily and unequivocally determined from the highest spatial frequency at which Bragg peaks remain visible in the Fourier transforms of images (Han et al., 2017).