The sample remains frozen during PLX4032 in vivo imaging on the cold stage, which is cooled by liquid nitrogen. Even though the frozen state stabilizes the soft material and liquids of the biofilm (which would otherwise be impossible to examine at high magnification, because

of sample movement or beam damage), the cryo-SEM images (Fig. 3) appear to be of a lower resolution compared to conventional SEM. This is partly attributable to a lower conductivity of the frozen surface compared to the dehydrated gold-sputtered surface we employ in conventional SEM. Another downside of cryo-SEM is that the frozen surface melts and cracks at high magnifications because of the heat generated by the focused electron beam. However, we were able to produce images of the biofilm that clearly show that the bacteria are enveloped in a gel-like matrix. We were not able to

obtain high-magnification images showing details of the matrix. Cryo-SEM also allows for freeze-fracture, which exposes the internal structure of the biofilm and may thus reveal how the bacteria are interconnected. The ESEM was developed in the late C59 wnt mouse 1980s (Danilatos, 1988). The ESEM retains many of the advantages of a conventional SEM, without the high vacuum requirement by varying the sample environment through a range of pressures, temperatures, and gas compositions. Wet and nonconductive samples may be examined in their natural state without modification or preparation. The ESEM offers high-resolution secondary electron imaging in a gaseous environment. The obvious advantage of ESEM is the total lack of preparation. The sample is placed directly on a stub and placed in the SEM chamber. However, with ESEM, we had problems obtaining high-resolution images of the biofilm because of the lack of conductivity in the wet sample. We experienced that close to magnifications of 10 000× and more, where the beam current is locally very out high, the focused electron beam seemed to destroy the 3D biofilm structure. We also observed that during prepumping, the sample also slightly dehydrates, but not to near the same extent as the dehydration used in conventional

SEM (Fig. 4). A superior, yet more sophisticated alternative to the conventional SEM and CLSM is the FIB–SEM. Similar to confocal scanning microscopy, it is possible with FIB–SEM to create 3D reconstructions. With a process termed ‘slice and view’, the FIB can sequentially mill away down to 10-nm-thick sections from the surface of a resin embedded specimen and subsequently record a SEM image (Fig. 5a) of the exposed block surface using a back scattered electron detector (BSED). Following acquisition of the successive image slices, the image data are processed to perform a 3D volume reconstruction (Fig. 5b and c). We were able to produce stunning 3D reconstructions of the spatial interaction of bacteria down through the 3-day-old biofilm (Supporting information, Movie S1).