A technique for improved focused ion beam milling of cryo-prepared life science specimens

Authors


M. F. Hayles. Tel: +31 (0) 610900030; e-mail: Mike.Hayles@fei.com

Summary

The combination of focused ion beam and scanning electron microscopy with a cryo-preparation/transfer system allows specimens to be milled at low temperatures. However, for biological specimens in particular, the quality of results is strongly dependent on correct preparation of the specimen surface. We demonstrate a method for deposition of a protective, planarizing surface layer onto a cryo-sample, enabling high-quality cross-sectioning using the ion beam and investigation of structures at the nanoscale.

Introduction

Cryo-transfer systems have been available for electron optical instruments for a few decades, and cryo-scanning electron microscopy (cryo-SEM) has established itself as a major tool in its own right, especially for specimens with high water content. In general, the advantages of cryo-immobilization in conjunction with SEM are widely accepted for applications in the life sciences (Echlin, 1992) and for soft materials such as polymers, gels and emulsions (Goff et al., 1999; Kishi et al., 2003; Stokes et al., 2004; van Duynhoven et al., 2005; Gonzalez-Meijome et al., 2006; Schaper et al., 2006).

Frozen-hydrated samples examined by cryo-SEM demonstrate superior preservation compared with chemically fixed and dried specimens because they retain all or most of their water. Structures such as cells appear fully turgid, showing little distortion or shrinkage, and extracellular secretions are well preserved (it is well-known that chemical fixation and dehydration of living cells can introduce many preparation artefacts). Indeed, there are abundant reports in the literature of cryo-SEM of botanical specimens, used as a routine technique since the 1980′s (see, e.g. Vartanian et al. 1982; Burton et al. 2000 and reviews by Sargent, 1988 and Read and Jeffree, 1991). Cryo-immobilization allows more rapid immobilization of the sample than chemical fixation and can therefore capture dynamic or transient events, for example spore discharge (McLaughlin et al., 1985). Cryo-prepared samples are not exposed to solvent extraction and therefore any fine structures and the spatial relationships of delicate, labile materials are well preserved, such as the conidia-bearing structures of fungi (Sadanandom et al., 2004) and powdery mildew growing on the surface of leaves (Xiao et al., 1997). Such samples cannot be as successfully prepared by chemical fixation and critical point drying, and so cryo-preparation has become the method of choice.

Samples are generally cryo-immobilized by a process known as plunge-freezing, in sub-cooled (‘slushy’) liquid N2 at −210°C (Umrath, 1974), liquid ethane or liquid propane. This method has been well-investigated, both experimentally and theoretically (Elder, 1989). Samples can also be prepared by high-pressure freezing, depending on the size and nature of the samples to be studied and the quality and depth of freezing required (Studer et al., 2001; Walther, 2003). Frozen samples are then transferred in vacuo to a cryo-preparation chamber where they can be fractured and sputter-coated with a conductive layer. After transfer to the SEM chamber, samples can be examined in high vacuum on the cryo-SEM stage.

In addition to the high-quality surface information that can be gained from biological samples using this method, cryo-fracture techniques in the SEM combined with high resolution, field emission microscopes, have enabled investigation of internal ultrastructural details, such as intra-membranous particles in yeast (Walther et al., 1990). Fracturing, however, does not readily show the contact relationship between, for example, membranes, and the extent and complexity of structures continuing within the cell volume. Fracture planes generally propagate through the weakest points of a sample, and are therefore considered to be relatively uncontrolled and random. One might argue that, instead of freeze-fracturing, one should turn to embedding and sectioning to prepare the sample for the transmission electron microscope. However, some samples are not only difficult to chemically fix and embed, but they can also be difficult to physically section using an ultramicrotome, due to density or hardness variations within a sample. In plants, there are often areas, such as the interface between resin, waxy cuticle and thick epidermal cell wall in some leaf samples, which make successful sectioning difficult. A further technique, known as cryo-planing, has been used to prepare the surfaces of high-pressure frozen blocks for high resolution cryo-SEM (Nijsse and van Aelst, 1999; Walther and Muller, 1999; Refshauge et al., 2006). Ultimately, researchers may be interested in specific points within a three-dimensional tissue, which are difficult to see or access unless first viewed at electron optical magnification in order to precisely choose a specific cell or feature before cutting. Clearly a method for in situ site-specific cutting would be advantageous.

An alternative means of sectioning is to use a focused ion beam (FIB) to selectively remove material and thereby create a cross-sectional face. Using FIB, one can decide precisely where one wishes to cut through a sample in a site-specific manner. The ion beam can successfully be used to mill any material, regardless of its hardness, to explore, for example, the interfaces between muscle cells and bone or gum tissue and teeth. FIB milling capabilities can also be combined with SEM imaging in one instrument (FIB SEM), and some recent examples of the use of FIB SEM for biological materials include cyanobacteria on calcite crystals (Obst et al., 2005), gland cells (Drobne et al., 2005), tissue-biomaterial interfaces (Burkhardt and Nisch, 2005) and bone cells on hydroxyapatite (Stokes et al., 2006).

FIB milling of temperature-sensitive samples such as polymers, foodstuff, cosmetics and many life science samples can result in damage and/or deformation because of lack of rigidity at room temperature and acute sensitivity to either the ion or electron beam. These samples could therefore benefit from cryo-stabilization to permit successful milling and imaging, although the development of cryo-techniques in conjunction with FIB or FIB SEM is at quite an early stage (Heymann et al., 2006; Marko et al., 2006).

Before FIB milling of specimens, it is advantageous to carry out localized in situ chemical vapour deposition (CVD) of material, commonly a metal such as platinum or tungsten, via a gas injection system (GIS). This helps to protect top surfaces from unwanted ion beam erosion, preserves topographical features against re-deposition of material and minimizes ‘curtaining’ artefacts (uneven vertical striations) on cross-sectioned faces. We will term the latter effect ‘planarization’. It is thought that low energy excited or secondary species, generated by either the primary ion or electron beam, degrade a precursor gas into metallic and (volatile) organic components, leaving a metallic layer on the sample surface (see, e.g. Orloff et al., 2003). Cryo-prepared specimens should also be protected by CVD prior to FIB milling, but unfortunately the normal practice does not work satisfactorily at low temperatures due to the relatively high vapour pressure of the precursor organometallic gas (such as methylcyclopentadienyl platinum trimethyl, (CH3)3(CH3C5H4)Pt). Precipitation onto the cold specimen surface occurs before the gas can be degraded. This leaves a thick layer of non-conductive organic compound, as shown in Fig. 1, with an inhomogeneous morphology that is unsuitable for the purposes of protection and planarization.

Figure 1.

A heavy deposition of mainly non-conductive organic precursor deposited on a cold surface using standard GIS practice. The crystalline structures and possible internal porosity will subsequently increase the risk of irregularities in the cut surface.

We have successfully developed a new approach that has enabled us to perform controlled deposition onto biological samples, polymers, foodstuff and cosmetics, etc, at low temperatures. Specifically, we demonstrate the application of this technique using Nicotiana tabaccum flower petals as an example of a challenging botanical sample, with a high degree of topography, and where embedding and sectioning for transmission electron microscopy (TEM) or freeze-fracturing techniques for SEM were previously inadequate to reveal the desired area of interest. We also discuss the issues and limitations concerning sample preparation for obtaining the best ultrastructural details from within a frozen sample.

Materials & methods

Experiments were performed using an FEI Nova NanoLab DualBeam™, with a field emission electron source and in-lens electron detectors, and an FEI Quanta 3D DualBeam™ with a tungsten hairpin filament electron source. Both employ similar ion columns with gallium ion sources. Low-temperature experiments were carried out using a Quorum PP2000T cryo-transfer system (Quorum Technologies, Newhaven, UK) (fitted to the Nova NanoLab DualBeam™) and a Gatan Alto 2500 cryo-transfer system (Gatan, Oxford, UK) (fitted to the Quanta 3D DualBeam™).

Small samples (cut to roughly 5 mm × 5 mm) of tobacco (Nicotiana tabaccum) petals were mounted with Tissue-Tek adhesive (O.C.T. compound, BDH laboratory supplies, Poole, England, U.K.) onto an aluminium stub on the specimen holder. The stub was then plunge-frozen in liquid nitrogen slush at −210°C and transferred under vacuum to the cryo-preparation chamber attached to the FIB SEM. The sample temperature was raised to −95°C for approximately 3 min to sublime any condensed ice from the surface which was gained during transfer. The temperature of the sample was then reduced to −125°C. Essentially, to avoid charging problems while searching for a suitable site, the sample was sputter-coated with platinum for 160 s, giving a thickness of approximately 15 nm. The sample was then passed through the transfer lock to the FIB SEM cryo-stage, which was held at −125°C. Imaging was performed using an accelerating voltage of 3 kV (on the field emission instrument) to 10 kV (on the tungsten instrument).

The GIS is specifically designed to deposit or etch materials at or near room temperature. The GIS has a crucible containing an organic precursor material with metal ligands (platinum, tungsten, etc). Under normal operation, the compound is heated in the crucible to produce a desired vapour pressure. The gas is then released by a valve mechanism and delivered to the sample by means of a narrow hypodermic-style needle that is positioned over the sample target area. In this case, the GIS temperature was set to 25°C. Since this is a lower temperature than generally used, together with the fact that the sample is frozen, we term this ‘cold-deposition’.

For standard ambient conditions, a needle-to-sample distance as little as 100 μm is acceptable. But for low-temperature conditions, and in particular for highly topographical samples, a much greater distance of between 300 μm and 2 mm was found to be necessary, so that the deposit was dispersed more evenly and also to avoid collision of the needle with any protruding surface structures.

The GIS device can be set so that the needle-to-sample distance at the eucentric position is within these limits and, therefore, retains coincidence of the ion and electron beams at a specific point on the sample. If the sample topography is of an acute nature then the stage can also be lowered, but eucentricity must be re-established after deposition has taken place. For this work, once the required features of the sample had been located by electron beam imaging, the sample stage was lowered slightly to increase the distance from the inserted needle of the GIS. The sample stage was tilted perpendicular to the GIS needle, for a more even surface coating.

Thicknesses of 0.5 to 2 μm of deposition are preferable, particularly for topographic observation of underlying surface features. For 3 s deposition time at a GIS temperature of 25°C, approximately 1–2-μm thicknesses of Pt were achieved. Electron beam imaging was used to observe the deposition process at low magnification (3 kV, 270 pA). The ion beam was not in operation during this procedure. Unlike beam-assisted CVD, it is the thermal gradient between the deposition gas and the cold specimen surface that drives the deposition process.

After cold-deposition, the GIS needle was retracted and the sample stage was returned to its imaging position. The required milling parameters (e.g. size and type of cutting area) were programmed and milling with the ion beam began, typically with ion beam current = 20 nA, accelerating voltage = 30 kV. The milled face of the sample was given a final ‘polish’ with the ion beam at a much lower current, typically 500 pA. The sample temperature was raised to −95°C, and the ice allowed to etch briefly, whilst imaging with the electron beam. For a final, high-quality image, the sample was withdrawn to the cryo-preparation chamber for sputter-coating with a further thin layer of platinum (at least 2 nm), before being returned to the cryo-stage in the main chamber of the FIB SEM for electron beam imaging.

Results and discussion

Sputter coating is a common method for making surfaces electrically conductive for imaging. However, repeated passes of a FIB quickly causes the coating to break down into its constituent islands, even for a thickly sputtered coating (tens of nanometres), leaving gaps through which the ion beam can penetrate. This was previously demonstrated for Au/Pd coating on silicon by Kempshall et al. (2002), who also showed that the more continuous nature of sputtered chromium is better in this regard. As a result of spurious ion beam penetration between any gaps, the sample surface becomes damaged, and irregular structures at the cutting edge cause differential milling (i.e. curtaining, as described in the Introduction). An example of these effects is shown in Fig. 2. The curtaining effect caused by the disintegrated sputter coating tends to be more pronounced for soft materials; hence the greater need to have a better protective layer on the surface.

Figure 2.

A sample of cryo-immobilized yeast cells that have been sputter coated with approximately 30 nm of Pt. This image demonstrates the effect of the ion beam on sputter-coated surfaces. Exposure of the top surface (a) to the ion beam will gradually disintegrate the metal coating due to its fragile structure. Disintegration of the sputter coating along the cutting edge (b) promotes the curtaining effects (c) seen here.

As mentioned in the previous section, using a GIS to deposit a platinum-containing layer under low temperature conditions requires a different approach to that for samples at ambient temperatures. In order to improve the method, we made some changes to the set-up of the GIS and the sample. The GIS heater temperature can be set between room temperature and 50°C, with 45°C being the standard setting. Since the vapour pressure of the precursor gas decreases with temperature, we found that the lower temperature of 25°C helped to reduce the flow of gas being released from the GIS to the extent that the deposition process could be more easily controlled, and therefore the layer thickness reduced to an appropriate level.

In a further refinement, we were able to modify the deposited Pt layer to form a very smooth surface and therefore provide an even better cutting platform. This was achieved by exposing an area to a 30 kV ion beam at a current of 1 nA for 30 s. Figure 3 shows a rectangular area that has been smoothed in this way. When the deposited layer is subsequently milled, it shows a solid homogeneous structure, similar to that expected of material conventionally deposited by beam exposure at room temperature. At no time was any damage to the underlying true sample surface observed using this technique, provided that the deposition thickness was maintained at around 2 μm or more.

Figure 3.

Modifying a deposited surface with the ion beam for a set time of 30 s with a current of 1 nA at 30 kV leaves a smooth surface (rectangular region, centre of image) that improves the cutting surface quality.

The tilt angle of the sample to the GIS is another important factor. Adjusting the tilt angle of the sample to a more favourable position, in this case with the features of interest perpendicular to the GIS needle, played a significant part in overcoming shadowing effects likely to be caused by other parts of the sample structure. Figure 4 shows the result of deposition onto a cryo-sample using a non-optimal tilt angle. The structures have received a highly directional coating and shadowing from adjacent structures, leaving some parts of the surface exposed to the ion beam. Unwanted exposure of the unprotected surface to the ion beam leads to irregularities in the surface texture that causes curtaining as the face of the sample is cross-sectioned. Conversely, the deposition shown in Fig. 5 has been carried out with the GIS needle perpendicular to the sample surface, resulting in a more even distribution of the protective layer and improved planarization. In another example (Fig. 6), we see that features such as the cell wall and associated membranes are clearly visible. In both cases, features are no longer obscured by curtaining artefacts. Note that the deposited material in Fig. 6 was smoothed as a result of a brief period of ion beam imaging across the whole field of view as described on page 11. The specimen seen in Fig. 5 was not exposed to ion beam imaging.

Figure 4.

Petal epidermis sample (Nicotiana tabaccum) following GIS deposition, under low temperature conditions, but at less than the optimum tilt angle, showing partial deposition. On the unprotected left-hand side (a), the upper surface has been degraded by exposure to the ion beam and curtaining effects (b) on the cross-sectional face are obvious. On the protected right-hand side (c), the beneficial effects (d) of the planarizing layer can be seen. Imaged using Quanta 3D DualBeam.

Figure 5.

As Fig. 4, but with an optimized tilt angle, resulting in a much more even distributed coating (Pt), (∼1-μm thick). The cross-section is vastly improved. Imaged using Quanta 3D DualBeam ESEM.

Figure 6.

As Fig. 5; The protection of the GIS metal deposition allows nanoscale details to seen on the cut surface. Here cell wall details and membranes are visible. Pt, even distributed GIS coating; Cw, cell wall; Pm, plasma membrane; Cy, cytoplasm; Eu, eutectic ice ridges; V, vacuole. Imaged using Nova NanoLab DualBeam.

After the milled face of the sample is given a final ‘polish’ with the ion beam, the surface is very smooth and shows little contrast. This is common to cellular material that has been cryo-immobilized but not chemically treated with an electron-dense stain. A brief temperature-controlled ice-etch is therefore very effective in providing surface relief for topographic imaging (Figs 5 and 6).

Meanwhile, poor conductivity is always an issue when using samples of high water content which have been cryo-immobilized from the native state, without the addition of electron dense stains. It should be noted that the platinum-based organic compound deposited in the way described does not aid electrical conductivity, but is used simply to enhance surface protection and provide the necessary planarization for smooth cross-sectioning. Hence samples were transferred back to the cryo-preparation chamber for sputtering with a thin layer of Pt to make the freshly etched surface conductive before being returned to the sample chamber of the FIB SEM for imaging.

Although cryo-preservation at the specimen surface will be good, plunge frozen-hydrated specimens will generally be poorly preserved in the centre, due to ice crystal growth during freezing. Only if the sample is extremely small (typically less than 20 μm for plunge-frozen specimens and less than 300 μm if high-pressure frozen), can cryo-immobilization produce true vitreous ice inside the sample, that is, to solidify the water without going through a phase transition and therefore without forming damaging ice crystals (Dubochet et al., 1988). Therefore, cryo-fracturing can only yield really high-quality information about internal microstructures if the sample is very small and the method of freezing, and the cryogen chosen, provides a fast enough cooling rate to result in a well-frozen sample. Note that, if using high-pressure freezing (to obtain better preservation of the sample for higher quality internal detail), it would not be possible to sublime water without the risk of phase changes that allow re-crystallization of ice.

The simplicity, speed and low cost of simple plunging-freezing methods have led to their widespread and accepted use in many circumstances for cryo-SEM sample preparation. Using no chemical treatments, i.e. chemical fixation or cryoprotection (replacement of water with compounds such as glycerol that freeze amorphously), can be more attractive and often preferable to dealing with the consequences of interpreting chemically induced artefacts. However, this may not prevent all cryo-artefacts, such as eutectic ice ridges.

Botanical specimens present us with special difficulties; thick cell walls hinder rapid freezing and it is almost impossible to freeze the large, water-filled vacuoles present in many plant cells without the formation of some ice crystals. In practice, freezing methods aim to remove the heat of a specimen as quickly as possible to keep ice crystals to a sufficiently small size (less than a few nanometres) so that they do not obscure cell ultrastructure. This subject has been reviewed fully by Sitte et al. (1987). For a large range of cryo-fractured specimens, plunge freezing in liquid nitrogen slush is perfectly adequate, since this reveals sufficient details at the cellular level to enable, for example, measurements of cell size and cell wall thickness (Ryden et al., 2003).

The emphasis of the present study is the successful FIB milling of cryo-prepared biological samples with minimal artefacts. The eutectic ice ridges seen in the samples shown in Figs 5 & 6 are within the vacuoles of the cells, which are not the areas under investigation, and are virtually impossible to freeze without any ice crystals forming, whatever method of freezing is chosen. The cell wall, plasma membrane and cytoplasm were adequately frozen to be able to measure the thickness of the cell wall and demonstrate the relationship between the shape of the cell and how the cell wall thickness varies around the contours of the outer epidermal wall. The epidermal cells on these petals vary in shape from flat, slightly domed to cone shaped, in various positions on the petal. For this sample we aimed to see, in particular, how the cell wall thickness varies at the top of one of the cone-shaped cells when compared with the base of the same cell. We had been unable to explore this by freeze-fracture methods alone, since the fracture plane almost always goes around the base of the cell and not through the tip.

Using the combination of cryo-SEM and FIB, we were able to site-specifically cross-section a single, chosen cell. Any devitrification of ice that may have occurred at the surface does not appear to have impaired the information being sought. It is also possible to take a series of successive images during FIB-sectioning and reconstruct the volume data to make a 3-D model of, say, a whole cell with internal details. We have begun using this technique to study the difference between wild type and mutant samples where the cells on the petal have abnormal shapes, thought to be due to cell wall deposition.

Conclusions & further work

We have described a technique that combines cryo-preparation of a sample and gas chemistry to allow milling of frozen, temperature-sensitive samples which, as a result, display little or no damage or degrading effects. This has been achieved by modification of the procedure used for the GIS, its physical settings and those of the cryo sample.

For high-resolution ultrastructural studies at greater depths or to study other types of sample, we intend to progress to high-pressure freezing. In these cases, we shall asses the advantages and disadvantages of aiding contrast by osmication of the tissues prior to freezing and view the contrast using a backscatter detector method (Walther, 2003), which may mean that the need for any sublimation is avoided. Despite using this method, the bulk of any plant tissue may still contain large ice crystals.

The transfer of high-pressure frozen material from the planchettes used to freeze such samples to the SEM cryo-specimen holder is also challenging and there are often problems with specimen instability once such samples have been transferred to the cryo-SEM specimen holder. However, good progress is being made in this field and it is becoming more common to see publications where high-pressure freezing and cryo-SEM are combined (Osumi et al., 2006). These issues need to be overcome before successful milling with a FIB can be attempted, as there must be no movement or drift in the sample during this process.

There are many samples from which adequate resolution can be obtained after simple plunge-freezing that would benefit from the cryo-SEM FIB combination. We have tested this process using a variety of samples which would otherwise be difficult, time consuming or impossible to prepare by other methods, such as longitudinally FIB-sectioning along the full length of a single, chosen Arabidopsis root hair or through the embryo of a Brassica seed (data not shown).

Work on techniques to improve the status of the FIB milled surface of cryo-immobilized samples will also continue, along with more detailed studies of the chemical and physical mechanisms involved.

Acknowledgements

We thank John Nichols and Cathie Martin, of the John Innes Centre, for supplying the Nicotiana tabaccum samples. We are grateful to Dr Steve Reyntjens for assistance with the operation of the FIB instrumentation in Eindhoven and for helpful discussion. KF was supported by a grant in aid from the Biotechnology and Biological Sciences Research Council (BBSRC) to the John Innes Centre.

Ancillary