Application of environmental scanning electron microscopy to determine biological surface structure

Authors


S.E. KIRK. Tel: 01223 337007; fax: 01233 337000; e-mail: sek32@cam.ac.uk

Summary

The use of environmental scanning electron microscopy in biology is growing as more becomes understood about the advantages and limitations of the technique. These are discussed and we include new evidence about the effect of environmental scanning electron microscopy imaging on the viability of mammalian cells. We show that although specimen preparation for high-vacuum scanning electron microscopy introduces some artefacts, there are also challenges in the use of environmental scanning electron microscopy, particularly at higher resolutions. This suggests the two technologies are best used in combination. We have used human monocyte-derived macrophages as a test sample, imaging their complicated and delicate membrane ruffles and protrusions. We have also explored the possibility of using environmental scanning electron microscopy for dynamic experiments, finding that mammalian cells cannot be imaged and kept alive in the environmental scanning electron microscopy. The dehydration step in which the cell surface is exposed causes irreversible damage, probably via loss of membrane integrity during liquid removal in the specimen chamber. Therefore, mammalian cells should be imaged after fixation where possible to protect against damage as a result of chamber conditions.

Introduction

Scanning electron microscopy (SEM) has been used for many years to give high-resolution, surface images of biological specimens. Its improved resolution over light microscopy combined with the large depth of field makes it ideal for viewing a range of sample types, but the challenge for imaging biological specimens is preparing the samples for imaging without changing their morphology. Therefore, the ability to view hydrated, non-conductive samples without treatment makes variable-pressure SEM an appealing technique for viewing biological samples.

We aim to explore the application of environmental scanning electron microscopy (ESEM) for high-resolution imaging of biological samples and make comparisons to the use of SEM in this field, with particular reference to possible artefacts and sources of sample damage from both imaging techniques. Focusing on mammalian cells as a case study, which offers challenges to both techniques, exhibiting both extremely fine details in their surface structure and great sensitivity to environmental conditions, we hope to establish where ESEM can offer advantages in the high-resolution imaging of biological samples as well as outlining the limitations of the technique. To this end, we begin with a review of standard SEM preparation techniques for biological samples and a brief description of the operation of the ESEM instrument.

Sample preparation for SEM

Hydrated samples need to be dehydrated before viewing in the high vacuum SEM. Typically 99% of the water needs to be removed as outgassing degrades the chamber vacuum, scattering electrons and partially hydrated samples will dehydrate and collapse within the SEM chamber. Sample preparation consists of a series of processes that aim to remove water but minimize changes in sample volume and morphology.

Typically, the sample is first chemically fixed, which preserves the structure of the sample by crosslinking the proteins within it, and strengthens the sample for drying. Primary fixation is usually in glutaraldehyde alone or in combination with formaldehyde, in Piperazine-1,4-bis(2-ethanesulfonic acid) (PIPES), N-(2-hydroxyethyl)-piperazine-N′-2-ethanesulfonic acid (HEPES), phosphate or sodium cacodylte buffers. This is followed by secondary fixation in osmium tetroxide and tertiary fixation in uranyl acetate.

The sample must then be dehydrated. Air drying results in the sample collapsing by more than 80% and losing its fine structure (Bray et al., 1993), so alternative methods are employed. The most common technique is critical point drying (CPD). Here, the water in the sample is first replaced with an organic solvent, generally using an ascending series of ethanol or acetone in aqueous solution up to an absolute dehydrating agent (Ruffolo, 1974). Next, the sample is placed in a sealed vessel which is filled with liquid carbon dioxide which substitutes for the solvent (Ris, 1985). The carbon dioxide is then removed by conversion to its gaseous form by raising the temperature and pressure to the critical point. This process avoids artefact formation by never allowing liquid/gas interfaces to develop, so the material is not exposed to surface tension forces (Bergmans et al., 2005).

An alternative technique is freeze drying. Samples are frozen and freeze dried either directly from water or after chemical fixation from an organic solvent, the latter generally being faster and less likely to cause disruptive crystal growth as the solvent will act as a cryo-protectant. A range of different organic solvents have been used for freeze drying (Katoh, 1978, 1979; Osatake et al., 1980). Chemically fixed samples dehydrated in solvent can be frozen either using liquid nitrogen (Katoh, 1979; Osatake et al., 1980) or by evacuation (Boyde & Maconnachie, 1979; Katoh, 1979), but hydrated samples are frozen in propane or ethane cooled to below their boiling point in liquid nitrogen. This cools the sample more quickly, reducing ice crystal formation. Freeze drying can be carried out from liquid nitrogen temperatures in a modified carbon coater (Warley & Skepper, 2000). In some cases it is considered advantageous to pre-cool samples before evacuation in order to prevent destruction of fine detail, especially in cases where the solvent has a relatively high freezing point (Inoue & Osatake, 1988).

Though these techniques preserve most of the features of the specimens, they may still cause some shrinkage or distortion of the samples, particularly bulk samples. It has been shown (Boyde, 1978) that prior to CPD, samples swell in ethanol and then shrink during the CPD process, resulting in some distortion of the sample surface. Volume loss in the CPD can reach 60% in some samples such as cells and embryonic tissue. This is lower, around 15%, for freeze dried samples and the volume loss in this case is thought to be due to the loss of firmly bound water after the passage of the freeze drying front. Also, ice crystal formation can be a problem if the internal surfaces of bulk samples are to be analysed, and thermal stress can cause sample cracking. Therefore, rapid freezing is crucial and this limits the size of a bulk sample to be frozen without fixation and solvent dehydration/cryo-protection. In all cases, any volume change is likely to be non-uniform and can cause distortion. At best, a level of shrinkage of around 5% can be achieved by freeze drying, but only when using very small samples such as monolayers of cells, and it is difficult to process many samples using this method. Artefacts manifest as cracking of fine cell processes, which shrink a small amount during drying whereas the glass cover slip they are attached to does not shrink at all.

Non-conductive specimens require coating in order to avoid charging during the imaging process. This usually consists of sputter coating with gold or gold palladium though chromium and platinum can also be used to give a finer coating for high-resolution work (Echlin, 1981). Film thicknesses vary, but must be large enough to allow the coating to be conductive, meaning the coatings must be at least a few nanometres thick. Coating can be avoided using low voltage methods, in which samples are viewed with a lower primary beam voltage, which, if carefully controlled, can avoid charge building up on the sample (Pawley, 1984). Alternatively conductive staining can be used where the sample is loaded with osmium followed by a mordant such as thiocarbohydrazide or tannic acid, then more osmium. The mordant and osmication steps can be repeated to successively load the sample with more osmium to make it electrically conductive.

Cryo-SEM is an alternative way to view cells frozen-hydrated in the SEM. Cells are frozen and are then solid, eliminating the need for the usual drying procedures for viewing biological samples in the SEM.

Applications of cryo-SEM are too numerous to mention but in biology tend to fall into two categories; surface imaging of hydrated samples at high resolution in cases where the potential artefacts of conventional SEM need to be avoided, and use of the solidity of frozen samples to fracture and view internal surfaces. The first of these overlaps somewhat with potential applications of ESEM, but it can be seen that there is a clear division between when each technique would be used. Cryo-SEM provides high-resolution images so is more applicable for viewing fine details, but is more time consuming and costly than ESEM, so where larger features are to be viewed, ESEM is likely to be preferable. Cryo-immobilization followed by SEM at low temperature can be used to trap dynamic events at individual time points. It cannot be used to follow any dynamic processes or to visualize surface liquids and exudates on samples – the ‘living features’ of the material – ESEM may be used to image these features. However, for fracturing and viewing internal surfaces, cryo-SEM is of course the method of choice. Therefore, though there is some overlap between the two techniques, in most cases they are likely to be complementary.

It should be noted that recent work exists in which cryo-microscopy is used in combination with the ESEM. Frozen samples are imaged at low temperature in a gaseous atmosphere, which should not be water and is most commonly nitrogen. The paper by Stokes et al. (2004b) compares cryo-ESEM with conventional and low voltage ESEM and shows how the elimination of the need for coating can extend the potential applications of cryo-microscopy. In particular, the potential to study sublimation in situ is investigated in order to look at ice crystal formation, in this case in ice cream. This study confirms the viability of this technique and shows that it offers advantages over low voltage techniques. They find that although the ability to observe fine features is obviously decreased relative to traditional high-vacuum methods, the potential for dynamic experiments means this is an area that merits further investigation.

Principles of ESEM

The ESEM is an adaptation of the SEM which eliminates the need for many of the sample preparation treatments discussed previously. Samples are imaged in a partial pressure of gas (Danilatos, 1988), which is achieved using a series of pressure-limiting apertures and differential pumping along the column's length, maintaining the electron source at ultra high vacuum and most of the beam pathway at high vacuum. The path through the low vacuum region in the sample chamber is minimized by bringing the sample close to the final aperture and attaching a cone (Fig. 1A).

Figure 1.

(a) Schematic of different pressure regions in the VPSEM instrument. (b) Cascade amplification in the VPSEM.

The beam does undergo some scattering in the sample chamber, and the amount will vary depending on beam gas path length, chamber pressure, accelerating voltage and gas type. However, the majority of the beam remains focused, producing the same probe diameter as would be available in an SEM, meaning essentially no resolution is lost. However, the scattered electrons will affect the signal to noise ratio of the image, generating an electron ‘skirt’ distributed over the sample surface. This does not significantly compromise the image quality except at high pressures, but is problematic in the case of EDX, making any results purely qualitative. The beam skirting means the interaction volume for X-rays is significantly broadened (Sigee & Gilpin, 1994). Although programs do exist to correct for the alteration, their effectiveness is questionable (Doehne, 2005).

Beam electrons enter the sample and generate secondary electrons as in a SEM (Shimizu & Ding, 1992). Upon leaving the sample, the secondary electrons are accelerated by a detector field. As these electrons have a much lower energy than the primary electron beam, they interact more strongly with the gas atoms and undergo ionizing collisions, each of which generates an extra electron. Both these electrons are then accelerated and in turn undergo collisions, doubling the number of electrons with each step. This process is called gas cascade amplification and allows the signal to be amplified before detection (Fig. 1B) (Thiel, 2005). The standard Everhardt–Thornley detector (Everhart & Thornley, 1960) cannot be used in the ESEM as the large electric fields required would cause dielectric breakdown in the gas. The signal can instead be detected in a variety of ways, depending upon the instrument design. Amplified SE signal can be detected, as can photons generated in collisions, or flux of gas ions.

The gas ions then drift towards the sample surface where they compensate for the negative charge which can build up on the surface of insulators, which makes it possible to image uncoated insulators with few charging artefacts. Different gases have different amplification properties and will give stronger signals at different pressures (Fletcher et al., 1997). Water is one of most commonly used gases as it provides strong amplification at around 5 torr (Fletcher et al., 1997) and also can act to maintain hydration in the case of wet samples, and it is therefore the gas of choice when viewing biological specimens.

Wet scanning transmission electron microscopy, is an alternative imaging mode in the ESEM which enables transmission images to be produced in the same environment. The beam passes through the sample and is detected on the underside of the sample, usually by a solid state detector, generating both dark and bright field images if a multi-segment detector. This technique allows access to the internal features of the sample, providing complimentary information to the surface image created in ESEM mode, and the two modes can function concurrently.

However, this technique can only be used on a limited range of samples because specimens must be thin enough to allow transmission of a significant number of electrons. The actual thickness this corresponds to will vary depending on the stopping potential of the material being viewed and the beam voltage used.

This newly emerging technique has been shown by Bogner et al. (2005, 2007) to offer potential advantages over ESEM in terms of resolution and high-quality images of samples, including organic material, suspended in liquids are shown. These images could not be produced using ESEM, as it is a surface imaging technique, or using approaches which require samples to be dried.

Comparing ESEM and SEM: artefacts and resolution

Clearly, the ability to images samples untreated, thereby avoiding the possible artefacts created during SEM sample preparation is one of the main potential advantages of the ESEM. This is illustrated by a number of studies in which the use of ESEM has changed the fundamental understanding of the structure or behaviour of a material. One such example is the investigation of the gluten network in dough (Bache & Donald, 1998) in which it was found that the network structure previously observed by SEM was an artefact of dehydration and the underlying network structure was much smaller than could be observed by SEM. Similarly, it was found that the mechanism of adhesion of spores of the marine alga Enteromorpha was via a gel-like adhesive pad which was visible in ESEM rather than the fibrillar structure seen in SEM as previously thought (Callow et al., 2003). Consequently, there have been a number of studies comparing ESEM and SEM, spanning a range of sample types including cells, microorganisms, tissue samples and plant material. We investigate this, focusing on human monocyte-derived macrophages (MDMs), which are particularly attractive test subjects as they exhibit many delicate membrane specializations including surface ruffles and fine filopodia, all of which are vulnerable to artefact and collapse during specimen preparation.

Cell culture

Monocytes were isolated from buffy coat residues by percoll gradient centrifugation then resuspended in medium and seeded onto cover slips and incubated for 1 h at 37°C before washing twice with phosphate buffered saline (PBS) to remove non-adherent cells (Muller et al., 1998). Culture medium (Mϕ-SFM with 1% penicillin and streptomycin) was replaced and cells were cultured for 5–7 days. The differentiated macrophages were then imaged either in the natural state, or following 1-h fixation in 4% glutaraldehyde, in the FEI XL30 FEG ESEM. For SEM imaging, cells were either freeze dried or critical point dried as outlined previously and gold coated. A range of images were taken using both the ESEM and SEM.

Imaging conditions

In SEM, imaging was performed at a range of accelerating voltages; however, 5–10 kV was found to give the best surface localization of the SE signal and hence the best image quality. Beam currents of 0.1–2 nA were used and the images were taken using the standard Everhardt–Thornley detector in the FEI XL30 FEG instrument (FEI Company, Hillsboro, Oregan, USA).

For ESEM, imaging conditions were optimized by varying the voltage and spot size to find the settings that maximized contrast between surface features of the sample. It was found that the optimum conditions for imaging fixed human MDMs were approximately 0.4–1.6 nA, 4–7 kV, as illustrated in Figs 2 and 3. However, the judgement of image quality is inevitably subjective, and it should be noted that optimized conditions will vary depending on sample type.

Figure 2.

Human MDMs fixed in 4% glutaraldehyde/PIPES, stored in HEPES and rinsed in deionised water before imaging in FEI XL30 FEG VPSEM. Cycled six times from 7.5–10 torr then imaged at, 2.4 torr, 3°C (42% RH) magnification 1466x. Sample voltage varied as follows: (a) −2 kV, (b) −3 kV, (c) −4 kV, (d) −5 kV, (e) −7 kV, (f) −10 kV, (g) −15 kV, (h) −20 kV.

Figure 3.

Human MDMs fixed in 4% glutaraldehyde/PIPES, stored in HEPES and rinsed in deionised water before imaging in FEI XL30 FEG VPSEM. Cycled six times from 7.5–10 torr then imaged at 5 kV, 3.5 torr, 5°C (53% RH), magnification 1954x. Beam current varied as follows (a) −5 pA, (b) −25 pA, (c) −98 pA, (d) −398 pA, (e) −1.57 nA, (f) −6.27 nA, (g) −24.3 nA.

Resolution

One of the main limitations cited by biological ESEM users is that effective resolution is reduced in ESEM compared to SEM (Bergmans et al., 2005; Muscariello et al., 2005). Actually, the intrinsic resolution is unchanged as it is dependent on probe size, which is the same in both instruments (given the same beam voltage and current settings), and the same instrument was used for both ESEM and SEM in this investigation meaning direct comparisons can be made. The signal to noise is degraded slightly in the ESEM due to the beam skirt, but the key reason for the loss of fine detail is the nature of the samples themselves. In high vacuum SEM coating is used for two reasons, the most important being to make the sample electrically conductive and eliminate surface charging. The coating also acts to reduce the penetration of the primary electron beam, minimizing the interaction volume that secondary electrons are emitted from, thus improving resolution.

In their native state, many biological samples are covered in fluid obscures details. These secretions are composed of low atomic number materials. They have a low stopping power, so the beam interaction with the sample produces fewer SEs as the incident electrons penetrate deeper within the sample or even pass completely through it, so signal is reduced. If we model electron interactions using Monte Carlo simulation (Joy, 1991) we find that the mean free path of a 30 kV electron in a low atomic number material such as carbon is around 10 μm. Also, secondary electrons (SEs) will come from a greater range of depths as escape depth is increased for lower density materials, degrading visible surface detail. Gold is a far superior emitter of electrons than low atomic number biological materials.

Some of these issues can be combated by the use of stains (Priester et al., 2007), but for completely untreated samples, they are inherent challenges to imaging. For example, it is difficult to distinguish small cell features from surface topography when observing cell attachment to biomaterial surfaces in ESEM (McKinlay et al., 2004). Mucus and other surface secretions also make it extremely difficult to image the underlying topography, as shown in work imaging at jejunal villi of rats (Habold et al., 2003). However, the ability to look at wet features of samples such as liquid secretions is also stated as an advantage of the technique.

We show that despite these challenges, many of the finer cellular features of human MDMs can still be resolved in the ESEM. In Fig. 4 we can see that although the filopodia are more clearly defined in the SEM image, much fine detail is shown by ESEM. In some cases ESEM images are easier to interpret as it can be difficult to distinguish fine sample features from coating artefacts in the SEM. Variations in coating thickness are more significant on the finest features.

Figure 4.

Human MDMs imaged in SEM and VPSEM. Samples (a) and (c) imaged by SEM. (a) shows gold coated, critical point dried samples imaged at 10 kV, beam current 538 pA, scale bar shows 10 μm. (c) is freeze dried, carbon coated and imaged at 5 kV, 98 pA with scale bar showing 2 μm. (b) and (d) fixed in 4% glutaraldehyde and rinsed in deionised water before imaging in VPSEM. All have been cycled from 5.4–9.8 torr eight times before imaging at 1°C, 5 kV, 1.57 nA. In (b), imaging is at 1.9 torr (39% RH) and the scale bar denotes 5 μm, in (d) imaging is at 2.4 torr (49% RH) and the scale bar shows 5μm.

Many users suggest that the ESEM better preserves some fine features, such as conidia and sporangia in fungal samples (Collins et al., 1993) and the microstructure of potato samples which were found to be damaged by freeze substitution and destroyed by chemical fixation (Uwins et al., 1993). Therefore, though imaging fine structure may be more difficult in ESEM, it is likely that structure may be better preserved.

Shrinkage

As discussed previously, shrinkage due to drying processes can be as much as 60% for critical point dried samples and 15% in the worst case for freeze dried samples (Boyde, 1978). Previous work has shown shrinkage in SEM images of a range of samples from mammalian cells (Stokes et al., 2003) and biofilms (Surman et al., 1996) to rice pollen (Tai & Tang, 2001). This is further demonstrated in our own images, as in Fig. 5 we can see that in the SEM images, the cell body has separated from the membrane protrusions at the cell periphery as a result of contraction during drying. This is not seen in the corresponding ESEM images where the cell body is intact. It should be noted that this is an extreme example of shrinkage although lesser examples abound, and subtler artefacts of shrinkage may be missed. For example, the relative positions of sample features may have shifted during drying, as shrinkage is known to be non-uniform (Priester et al., 2007).

Figure 5.

Human MDMs imaged in SEM and VPSEM. Image (a) taken using SEM at 5 kV, 1.57 nA. Samples critical point dried and gold coated. Image (b) taken using VPSEM at 5 kV, 1.57 nA, 4.9 torr, 7°C, (65% RH) cycled eight times from 7–10 torr. Samples are fixed in 4% glutaraldehyde and rinsed with deionised water before imaging. Scale bars on images shows 10 μm.

Cracking

Cracking (Fig. 6) is also likely to be the result of contraction during the drying process. We find this effect to be more prevalent in freeze dried samples, probably as a result of thermal stress (Boyde, 1978). Though these features are easy to identify and avoid during imaging, they may again point to other changes also being present within the sample which are not so easily identified. Fine features such as filopodia are particularly vulnerable to this artefact (Stokes et al., 2003; McKinlay et al., 2004). Cracking is an almost inevitable consequence of processing cells that shrink even slightly on rigid substrates such as glass. Therefore improvements can be made by the use of more flexible substrates.

Figure 6.

Human MDMs imaged in SEM at 5 kV, 98 pA. Scale bars denote 2 μm. Both samples freeze dried, (a) is carbon coated, (b) is gold coated.

Surface texture

Comparing the ESEM and SEM images in Fig. 7 we can see there is a substantial change in the apparent surface texture between them. The ESEM images show a much smoother texture, the complex ruffling and folding which is characteristic of MDMs, and which can be seen in the SEM image, not being present. This could be for a number of reasons. Possibly the cell surfaces are covered with some residual water which is obscuring the finer features of the cells surface. Alternatively, partial dehydration in the ESEM may have altered morphology. This is supported by Fig. 7B where we see a protrusion such as a tertiary lysosome or other dense structure in the cytoplasm which has been exposed by shrinkage. This suggests that ESEM is not free from artefacts and we will discuss this possible mode of damage in detail later. Alternatively, the extreme roughness of surface structure seen in the SEM image may be a desiccation artefact as observed previously by Muscariello et al., (2005). However, this cell type is known to have a highly ruffled structure, so it is more likely that the difference is a result of conditions in the ESEM in this case.

Figure 7.

Human MDMs imaged in VPSEM and SEM. (a) in SEM at 10 kV, 538 pA with samples critical point dried and gold coated. (b) VPSEM image of samples fixed in 4% glutaraldehyde, rinsed in deionised water. (b) taken at 1.57 nA and 5 kV, after cycling eight times from 7–10 torr at 4.9 torr, 7°C (65% RH).

Other users have found that the lack of coating gives access to the internal structure of samples, which allowed Collins et al., (1993) to observe cytoplasmic structures of algae in addition to surface morphology.

Fixation

All previous ESEM images in this paper have shown samples that have been fixed in glutaraldehyde. Cells can be imaged in their native, hydrated state as shown in Fig. 8. The morphology seen in the unfixed cells is very similar to that seen in the fixed case. However, it is much more difficult to obtain these images of untreated samples as cells are easily damaged by the chamber environment in the ESEM. Human MDMs become easily damaged when imaged at high magnifications and beam voltages. Under these conditions damage manifests as membrane blebbing and mass loss as shown in Fig. 9. Previous studies have observed similar beam damage in biofilms (Surman et al., 1996), though this kind of damage is not seen at moderate magnifications in a wide range of samples (Tai & Tang, 2001). Chemical fixation significantly reduces this artefact as shown previously (Collins et al., 1993). Fixed samples appear to be stable in the microscope for extended periods [the exact length of time will depend on the microscope parameters used, particularly relative humidity (RH)] and though beam damage can be observed at high beam voltages and magnifications (as shown in Fig. 9) the samples are resilient to imaging at 5–7 kV and magnifications of around 3000×. The morphology of fixed and unfixed samples appear very similar, indicating that light fixation introduces minimal artefacts, as shown previously (Stokes et al., 2003). Therefore, it appears preferable to image mammalian cells fixed where possible.

Figure 8.

VPSEM images of human MDMs. (a) and (b) fixed in 4% glutaraldehyde, rinsed with deionised water, imaged in VPSEM at 5 kV, 1.57 nA. (a) cycled 6–10 torr eight times, imaged at 10°C, 3.9 torr (42% RH). (b) cycled 7–10 torr eight times, imaged at 4.9 torr, 7°C (65% RH). (c) and (d) unfixed, untreated and imaged by VPSEM at 3°C, 4.1 torr (72% RH), 1.57 nA, 5 kV. All scale bars show 10 μm.

Figure 9.

Human MDMs imaged in VPSEM. (a) shows effect of repeatedly imaging a small region at high magnification. (b) was repeatedly imaged at 15kV at slow scan retes resulting in membrane blebbing and mass loss.

We have shown here that although imaging in ESEM can confer advantages, there is some evidence that the imaging process can damage samples through a variety of mechanisms. To explore this further, we present a study of cell viability in the ESEM.

Sample damage: cell viability in ESEM

Very few studies have looked at the effect of ESEM imaging on biological samples. Misirli et al. (2007) looked at Saccharomyces cerevisiae cells in low vacuum (0.1–4 torr, where charging in insulators can be eliminated) and wet mode (up to 20 torr, where hydration can be maintained) and assessed viability by analysing shape change in cells. Their results suggest that viability is much increased in wet mode compared to low vacuum mode as expected, and suggest that the cells can be maintained live in wet mode. However, no further viability analysis was carried out after imaging, so if damage was inflicted without change in cell shape this would not be detected, and mechanisms of damage are not investigated. Early ESEM users, particularly Danilatos, also looked at imaging live biological samples in the ESEM, but under high/atmospheric conditions and at room temperature (Danilatos, 1979, 1981a, 1982, 1991). They observed that some plant samples continued to grow after ESEM imaging, though closer observation showed that the areas subject to the electron beam exhibited cell damage (Danilatos, 1981a). Ants were also found to resume motion after imaging in the ESEM (although they were immobilized while in the chamber), with those subjects also exhibiting increased damage when subject to the electron beam (Danilatos, 1982). These studies suggest that imaging samples in the live state is possible, at least for certain plant and insect species, but further quantitative studies would be needed to establish which conditions minimize the observed sample damage. Here we explore the significance of a range of environmental factors in sample viability in the case of mammalian cells, specifically human MDMs and 3T3 fibroblasts, highlighting the key steps which may lead to sample damage. We work in the lower pressure, cooled regime which offers clear advantages in resolution over comparable light microscope images.

Cell culture

The 3T3 fibroblasts were prepared from frozen stock by culturing onto flasks in D-MEM with 1% pen-strep, 10% fetal bovine serum (FBS) and 1% glutamine. Cells were extracted using trypsin and seeded onto glass cover slips at a low density around 2500–3000 cells per cm2 and left for 2–3 days until around 50% confluent. Samples more than 70% confluent such that cells were strongly overlapping and could not be distinguished from each other, were rejected. Human MDMs were isolated from buffy coat by percoll gradient centrifugation as outlined previously and seeded to create similar cell densities.

Viability analysis

Cells were exposed to a range of different conditions in and outside the microscope (FEI FEG XL30 ESEM), and then the cell viability was studied by uptake of calcein-AM (Sigma-Aldrich Company Ltd., Dorset, UK) and changes in cell morphology relative to a control sample using an epi-flourescence microscope. The ability to sequester and retain calcein was used as an indicator of cell viability. Cells were also stained with Hoescht-33342 (Invitrogen, Paisley, UK) in order to locate the nucleus and act as a guide to total cell numbers in quantitative studies. In quantitative studies the number of cells with calcein uptake matching that of a control sample was counted in a number of fields of view crossing the sample in two perpendicular directions and using a minimum of 10 images per sample. The total number of cells was counted using the Hoescht staining and the proportion of live cells taken as the number with calcein uptake divided by the total number of cells. Each data point results from at least three samples giving the error bars shown. Stains were applied as a 0.1% solution in PBS and incubated for 15–20 min at room temperature before imaging. Morphology of calcein uptake was also studied to look for changes in membrane structure as a result of adverse imaging conditions.

Choice of imaging liquid

Usually, to produce the cleanest image, samples were rinsed with distilled water before imaging. However, as seen in Fig. 10, cells rapidly swell and detach when placed in water, and in some cases burst. Figure 11 shows the survival rates for 3T3 cells submerged in water and in buffer solutions Hanks' Balanced Salt Solution (HBSS) and PBS. These are used as a control for effects resulting from changes in pH and osmolarity, and we can see cells are stable in both at room temperature for up to 45 min. We see that 3T3 cells are stable for at least 45 min in both the buffer solutions, but that survival rates in water decline rapidly with time. Similar results are found for MDMs which are seen to swell when placed in water (Fig. 12).

Figure 10.

(b) 3T3 cell morphology in the light microscope after 5 mins in water, image area 1000 × 1500 μm, compared with control (a).

Figure 11.

Plot of survival rate of 3T3 cells submerged in liquids for up to 45 mins. There are large error bars on the water data as samples tend to detach and clump making it difficult to evaluate total cell numbers.

Figure 12.

(a) Normal MDM morphology compared to (b) showing MDMs submerged in water for 1 minute and imaged in the light microscope. Image area 1000 × 1500 μm.

Temperature

As outlined by Danilatos and co-workers (Danilatos, 1979, 1981b, 1983, 1985), it is possible to image in ESEM under a range of gas pressures up to atmospheric pressure. This requires significant adaptation of the machine to include additional pumps in order to maintain the gun at high vacuum. However, many commercial ESEM instruments now offer capabilities up to 20 torr without modification. A pressure of 20 torr is sufficient to work at 100% RH at around room temperature. However, the compromise made in all high-pressure work is signal to noise. As gas pressure is increased, so is scattering of the beam, resulting in a reduction in the signal to noise of the instrument. Indeed, much of the work conducted at high pressures barely offers any advantage in terms of the level of magnification over the light microscope, which must be a minimum standard for most imaging applications. Therefore, to improve signal to noise, in most applications the ESEM operates at reduced pressures of around 5 torr. Here, amplification in water is maximized and scattering is relatively low. However, the choice of pressure conditions used will vary depending on sample type and application and it is worth considering that though most ESEM users operate at these higher magnifications, reduced pressure conditions, gas pressures from high vacuum to atmospheric pressure are theoretically available.

For most commercially available microscopes, the partial pressures of gas under which good quality images can be produced are low, around 5 torr. As stated previously, most standard ESEM gaseous secondary electron detectors start to drastically lose image quality above 5 or 6 torr and by 10 torr (∼1.3 KPa) the signal to noise is too poor to generate an image at magnifications significantly higher than those used in the light microscope. However, a new system, termed the ‘needle’ detector has been developed in the Cavendish Laboratory (Baker, Toth & Baker, 2004), which enables pressures of 15 torr and above to be used while still preserving a good level of image quality (Toth et al., 2007). This allows imaging to occur at temperatures close to room temperature and it has been used to image bacteria under these conditions (Stokes, 2004a). The needle detector uses a fine tip to generate a highly localized electric field improving the collection of secondary electrons. This means that scattering losses are less significant using this type of detector, and it certainly has potential for pushing out the range of pressure and temperature conditions that can be used while still maintaining a high-quality image. However, this has only been demonstrated for a limited number of samples at relatively low magnifications so far, and it is likely that higher magnifications will be particularly difficult to achieve for low contrast biological samples.

Considering the water vapour pressure curve shown in Fig. 13, it is apparent that to view samples at pressures of around 5 torr while maintaining a high level of RH, the sample will need to be cooled to below room temperature. In the microscope this is achieved using a Peltier stage which cools the sample by passing a continuous flow of water through below the base of the sample holder which acts as a heat sink. This can cool down to below 0°C, meaning it is possible to maintain 100% RH. However, to improve image quality by raising gas pressure or to increase temperature, it is often desirable to work at humidities below 100% RH. Therefore, it is worth considering the thermodynamics and kinetics of the interaction between the gas and the sample as explored in detail in the work by Stokes et al. (2003). They argue that as the composition of the sample being imaged, in most cases, will not be pure water, but instead some solution containing salts and other molecules, operating below 100% RH is likely to be desirable. The ideal RH will vary on a sample to sample basis, but it has been shown that an RH in the region of 65–75% gives favourable imaging conditions with no visible evidence of sample dehydration for a range of biological samples because of the equilibrium vapour pressure of the samples (Tai & Tang, 2001; Stokes et al., 2003; Bergmans et al., 2005).

Figure 13.

Plot showing pressure and temperature conditions required to produce different temperature. Upper curve represents 100% RH - below this conditions are dehydrating and evaporation will occur, above it water will condense onto the sample. The other curves represent the crystallisation points of the imaging liquids considered – 75% RH for PBS, 65% RH for D-MEM medium and 55% RH for HBSS.

Kinetics is also significant, as when cooled to a few degrees centigrade, most processes in biological systems are considerably slowed down. Therefore, dehydration under these conditions is likely to be a slow process meaning it may be possible to work at lower pressures for a short period of time without considerable sample dehydration, enabling lower pressures to be used (Stokes, 2003).

In order to help control the hydration state of cells in the ESEM, they are usually cooled. To investigate this, cooling experiments were conducted. Cells grown on cover slips were placed on a glass Petri dish in a water bath and covered in the imaging liquid used. The temperature inside the Petri dish was checked using a thermocouple and the water bath temperature adjusted accordingly.

Figure 14 shows that 3T3 cells are relatively stable in both medium and buffers for up to 45 min at 3°C. The cells will be adopting a lower metabolic state, enabling them to survive these conditions by functioning more slowly. If this is the case, it may prove problematic if we wish to perform dynamic experiments under these conditions at a later time. When 3T3 were immersed in water (Fig. 14) viability rapidly decreased. The same trend was seen with MDMs (Fig. 15). Both cell types showed little reduction in viability in medium or either buffer when cooled to 3°C. Therefore, cooling is not found to be a significant cause of cell death. Cells suffered high levels of damage in water regardless of temperature, and were found to be stable at low temperatures in the other liquids.

Figure 14.

Plot of survival rate of 3T3 samples submerged in liquids for up to 45 mins at 3°C. As before, there are large error bars on the water data as samples tend to detach and clump making it difficult to evaluate total cell numbers.

Figure 15.

Plot showing survival rates of MDMs cooled to 3°C in a range of liquids for up to 45 mins.

Reaching imaging pressure

Once cooled, samples are pumped down to ESEM pressure for imaging following an optimized process as outlined previously (Cameron & Donald, 1993), in which the pressure is cycled between high and low values eight times in order to replace the air in chamber with water vapour for imaging. RH inside the microscope chamber was investigated using a humidity sensor (Sensirion SHT15, Sensirion, Staefa, Switzerland) held as close as possible to the Peltier stub. The sensor incorporated a temperature sensor so temperature at the sensor and at the stub were also recorded during pumpdown so the humidity measured could be adjusted to allow for the temperature difference due to the small air gap between sensor and stub.

As shown in Fig. 16 the sample is exposed to low humidities of around 25% during the initial stages of pumpdown, which could be damaging. This could be avoided by use of an airlock to transfer samples into the chamber as described by Danilatos (1988). However, no evidence of sample damage in 3T3 cells or human MDMs during pumpdown was found, provided several droplets of distilled water were placed in the base of the chamber. As the water droplets are not cooled, they evaporate preferentially to the cooled liquid in the vicinity of the cells, which protects the cells during the initial stages of pumpdown.

Figure 16.

Line A shows humidity predicted from set pressure and temperature during pumpdown. As the environment is not pure water vapour, the actual relative humidity follows line B as measured by a humidity sensor placed next to the Peltier stub.

Exposing the sample surface

After pumpdown, the chamber pressure must be reduced in order to remove surface liquid and expose the cell for imaging. As ESEM is a surface imaging technique, cells covered with a surface layer of liquid cannot be imaged until this layer is removed by the application of evaporating chamber conditions. This is achieved by setting the chamber pressure to some fixed value, the evaporating pressure, then allowing liquid to evaporate off the sample due to the reduced humidity which generates evaporating conditions.

Different rates of evaporation can be achieved by varying the pressure. If the pressure is more substantially reduced, the chamber humidity is lower and the liquid will be removed from the sample more rapidly. However, as illustrated in Fig. 13, using low humidities to expose the sample will cause the buffer or medium to crystallize, damaging cells and obscuring images. This restricts the range of evaporating pressures available depending on the liquid used.

It was found that the best imaging conditions were produced by intermediate rates of evaporation. At slower evaporation rates it proved impossible to remove all surface liquid and fully expose surface features as shown in Fig. 17. However, exposing 3T3 fibroblasts at faster rates completely destroys all cell morphology, leaving only cell fragments as a result of crystallization of the medium as shown in Fig. 18. This explains why damage to cells is more severe at faster evaporation rates, and cells exposed at faster rates show no uptake of calcein after imaging, with only cell fragments remaining. Placing samples in water would avoid this, and this approach could be applied to other sample types, but as shown previously mammalian cells are rapidly damaged via osmotic shock when placed in water. Evaporating at a slower rate still damages cells, producing a distinctive morphology as shown in Fig. 19. In these fluorescent microscope images, the cells appear to have little uptake of calcein in the centre, suggesting that this area has been damaged. As this raised central area of the cell will be exposed to the dehydrating cell conditions first, this suggests that the mechanism of death is cell membrane collapse and outgassing due to exposure to dehydrating chamber conditions. As the raised centre of the cell is the first region exposed, it exhibits damage at slowly evaporating conditions. At a lower evaporating pressure and hence lower RH the whole cell will become exposed and hence completely collapse. This conclusion is supported by the work by Stokes et al. (2003) in which they imaged osteoblast-like cells both by SEM and ESEM. They found that the ESEM image had a more flattened appearance with internal features such as the nuclear envelope and nucleoli visible at the surface suggesting membrane shrinkage or collapse.

Figure 17.

ESEM images of 3T3 cells showing same field of view after stabilising at a range of pressures at 3°C. Chamber pressure set to the following: (a) 5.0 torr; (b) 4.6 torr; (9c) 4.2 torr and (d) 3.8 torr and sample allowed to stabilise for around 15 minutes. Liquid seems to accumulate close to cells and is not completely removed at 3.8 torr, corresponding to 71% RH. At 5 torr (94% RH), cells are completely covered with liquid, and even after 30 mins in the chamber barely any of the cell surface is exposed, supporting previous suggestions that humidity varies across the chamber.

Figure 18.

3T3 fibroblasts imaged in ESEM. Images show same field of view after stabilising at a range of pressures at 3°C. Chamber pressure set to the following: (a) 3.8 torr; (b) 3.6 torr; (c) 3.4 torr and (d) 3.4 torr. In (a) and (b) sample is allowed to stabilise for around 15 minutes. (c) shows image immediately after pressure reduced to 3.4 torr and (d) shows same field of view several minutes later. These samples were imaged in medium, which crystallises at around 65% RH. Similar behaviour is seen for PBS and HBSS at 75\% and 55\% RH respectively.

Figure 19.

3T3 cells stained with calcein-AM and viewed with epi-fluorescence microscope. (b) was pumped down with water in base of chamber and a surface layer of medium on cells to 5 torr at 3°C (88% RH). After 22 mins, the upper cell surface became visible, though cells were never fully exposed. Although cells still show some calcein uptake, morphology is significantly changed relative to control (a). Image sizes 1000 × 1500 μm.

As shown previously, we can take images of untreated MDMs which appear to be hydrated and fully swollen, by contrast to the collapsed and flattened morphology seen in 3T3 cells and the osteoblast-like cells in previous work (Stokes et al., 2003). This suggests that MDMs have some additional tolerance of the liquid removal step.

Evaporating conditions were investigated for MDMs, and it was found that the conditions in which images of rounded cells could be produced were very limited. In the range 4.1–4.5 torr at 3°C, corresponding to 72–79% RH, images of cells with normal morphology such as those shown earlier could be obtained, with the cell surface becoming exposed approximately 5 min after the end of the pumpdown sequence. At lower evaporating pressures, cells were found to collapse and desiccate. At higher pressures, the cell surface could not be fully exposed as for 3T3 cells.

Subsequent staining of samples revealed that though a rounded morphology could be maintained, it was not possible to image these samples live as they dehydrated at 4.1–4.5 torr and did not take up calcein after imaging. Cells that had been subject to slower liquid removal did exhibit patchy calcein uptake but the morphology was changed relative to a control sample as shown in Fig. 20. It is suggested that this cell type is more resistant to mass loss and dehydration than 3T3 cells. Therefore, either this mechanism of damage proceeds at a slower rate, such that it is less noticeable in the microscope with no obvious cell collapse but still leading to cell damage and death over time, or there is another mechanism of cell damage that supersedes the dehydration process. It is possible that differences in cell size and shape could account for differences in dehydration between the cell types as 3T3 cells spread more on glass cover slips than MDMs, giving a greater surface area from which moisture could be lost.

Figure 20.

Samples stained for 15 mins in calcein, (a) shows control compared to (b) which has previously been dehydrated at 5 torr, 3°C (88% RH) for 25 mins following pumpdown in VPSEM. Image area 1000 × 1500 μm in both cases. (a) still shows calcein uptake but morphology is drastically changed compared to (a) and sample seems fragmented.

Beam damage

It is necessary to irradiate the sample in order to observe when the sample surface is exposed, so beam damage could be partly responsible for the cell damage. Cells which remain coated with liquid are found to show little evidence of cell damage regardless of time spent in the chamber or 5 kV beam irradiation, but the beam might not penetrate through the liquid coating to the cell. However, the mechanisms of death and damage observed do not correspond to the usual modes of beam damage such as mass loss seen as bubbling of the surface, or instantaneous membrane rupture. Attempts have been made to dehydrate the cells without irradiation by recording the pressures and time periods required to expose cellular features and duplicating them without the beam. However, the point at which cells become exposed depends also on the initial amount of surface liquid which cannot be precisely controlled, making it more difficult to determine whether cells are uncovered. Despite this, these studies reproduce the characteristic morphology in 3T3 cells confirming cell damage is unlikely to be a result of beam irradiation. Whether beam irradiation would also act as a source of damage cannot be established in this case as it is superseded by environmental factors.

Previously studies of polypropylene (Kitching & Donald, 1998) and of radiolysis products of water (Royall et al., 2001), have focused on beam damage, and found that that the level of damage was severely exacerbated by the presence of surface water and established that low beam voltage conditions are likely to be least damaging. These studies were primarily for polymeric materials, so though the studies are not for biological samples, the findings are likely to apply as the basic molecules present are similar. Also, previous work in higher/atmospheric pressure imaging conditions by Danilatos (1981a, 1982) showed that biological samples exhibited clear evidence of beam damage, so where environmental conditions are less significant, this is likely to require further consideration.

Dynamic experiments

Despite this evidence of sample damage in the FEI XL30 FEG ESEM, the technique has been used successfully in a variety of dynamic experiments, exploiting the fact that samples can be imaged untreated and the control of sample environment afforded by the ESEM chamber. However, given the challenges in sample damage outlined above, it is important that appropriate consideration is given to the likely impact on samples of the imaging process, which is likely to be highly sample dependant. We also need to consider methods that can be used to minimize this damage, such as working at lower voltages and magnification where possible and limiting the time samples spend in the microscope.

The most obvious dynamic experiments which can be carried out in the ESEM are hydration studies, adjusting the RH of the chamber by controlling the temperature and pressure. This has been used in several investigations (Bache & Donald, 1998; Callow et al., 2003) including work observing collagen and atelocollagen sponges (Ruozi et al., 2007). Through exposing the sponges to hydrating and dehydrating conditions they established that though they have substantial differences in pore size, both sponges readily adsorb water which is then very difficult to remove indicating that the pore structure does not solely govern the water binding capacity. The ability to follow dynamic dehydration processes to also allows users observe changes in liposome morphology of poorly water soluble drugs during dehydration (Mohammed et al., 2004). This method has established that ibuprofen incorporation improved stability of the liposomes of the drug studied, showing an increased resistance to coalescence during dehydration.

Another area which has not yet been substantially explored is the ability to use the ESEM detector at high temperatures, which is not possible with the Everhardt–Thornley detector. This means samples can be imaged in situ on a heating stage, which has been used to observe the pyrolysis of hull-enriched by-products from the scarification of hulled barley, candidates for future biofuels (Boateng et al., 2007). The use of ESEM in low-vacuum mode, combined with a 1000°C heating stage, allowed them to study in situ the gas evolution patterns and char microstructure development during thermal degradation. The challenge in applying this to biology is hydration, as at high temperatures and viable imaging pressures the chamber humidity will be very low. Therefore this technique is only appropriate for samples which are stable in low vacuum mode.

Samples can also be observed while subject to mechanical tests in their natural state. Stages exist in which shear and tension experiments can be carried out and this has been used to look at failure modes and behaviour under strain of a number of materials, mostly food stuffs. Examples include the study of slicing, tension and compression of fully hydrated carrots (Thiel & Donald, 1998) and the fracture of plant tissues and walls, specifically onion epidermis (Donald et al., 2003). In both cases a specially designed mechanical testing stage is implemented in which a Peltier chip is installed to enable cooling so that full hydration can be achieved during testing. This makes it possible to probe the true modes of deformation under mechanical stress, taking images in situ during measurements. Here the aim is not necessarily to take precise quantitative measurements but to observe the failure mechanisms and identify critical features in the sample microstructure.

There have been some studies, largely in the ondontic field, of the use of a modified micromanipulator to inject liquids onto samples in the chamber and observe the interaction with the sample. Previous work has looked at acid etching of enamel in situ in the ESEM (Cowan et al., 1996). The ESEM micromanipulator was adapted to include an intrachamber extension with a tip which could be loaded with etchant, allowing the observation of etching of dental enamel in real time. This method established that the majority of etching happened within the first 10 s of application and showed the formation of distinctive etching patterns.

Finally, an area which shows potential is the semi-dynamic study of ESEM samples. A number of locations are imaged on the sample and the coordinates of these points are stored in the microscope memory. The door of the chamber is then opened and the sample subject to some process and then the same locations imaged again. This method can be used to take time points in the same way as samples would be taken for SEM at different stages during a process, but has the advantage that the same region on the same sample is being used at each stage so that direct comparisons can be made. This approach has been used to study the use of laser irradiation to kill bacteria found in root canals, observing the numbers and changes in morphology of the bacteria after a number of laser cycles (Bergmans et al., 2006). It should be noted that as repeated imaging is required, this study requires samples to be in the microscope for some length of time and so may not be suitable for use with more delicate samples.

For most of these studies it would be valuable to know the extent of sample damage in the instrument and whether samples can be kept alive. We know for mammalian cells that samples cannot be imaged live in standard ESEM conditions, but it is as yet not established whether samples which are more resilient to dehydration, such as plant material, can withstand imaging. If we are to conduct these interesting experiments on samples in their native state in situ in the microscope, it is important that more consideration is given to the question of sample viability in a wider range of sample types.

Conclusions

ESEM opens up a wide range of new applications in SEM which rely on being able to image samples in, or close to, their native state. However, it is important to be cautious about the conditions when using ESEM. Careful control of pressure is required in order to achieve the desired chamber humidities when working with more delicate samples, and in the case of mammalian cells it is recommended that fixation is used where possible to stabilize samples against damage and open up a wider range of possible imaging conditions.

We have shown that it is possible to produce images of mammalian cells in the ESEM revealing fine detail of delicate features such as filopodia and membrane ruffles. We have confirmed a range of artefacts are generated during fixation and dehydration of cells for SEM including cracking and gross shrinkage. The extremes of these are not seen in ESEM images. The two techniques complement one another both providing different information. Given the reduced time and effort in sample preparation, it may be appropriate to use the ESEM where very fine detail observation is not necessary. Where fine features are damaged in the ESEM by beam interactions and partial shrinkage it will be more appropriate to use chemical fixation, dehydration and imaging by SEM. In most cases a combination of ESEM and SEM may be most informative.

Furthermore, we have shown that it is not possible to image 3T3 fibroblasts or human MDMs live in the ESEM, and suggest this is likely to be the case for other mammalian cell types. Both cell types are found to survive the initial stages of cooling and pumpdown but experience damage during the dehydration stage in which the chamber pressure is reduced to remove surface liquid and expose the cells for imaging. This damage is most severe in the case of 3T3 cells which exhibit a distinctive morphology at slower dehydration rates which suggests that the mechanism of cell death may be via dehydration and membrane collapse. This would suggest that it may be possible to image samples with a stronger cell membrane or wall, such as bacteria and plant material, in the living state. However, beam damage would still need to be investigated, as environmental factors supersede any damage resulting from irradiation in this case.

Further studies of the effect of ESEM imaging on biological samples are required in order to develop further the promising field of dynamic experiments currently being established in ESEM.

Ancillary