X-ray nano-diffraction on cytoskeletal networks

Human adrenal cortex carcinoma-derived SW-13 cells (ATCC CCL-105) [28], stably transfected with DNA encoding for fluorescent human keratin hybrids (HK8-CFP, HK18-YPF) [29–31], are generously provided by Rudolf Leube (RWTH Aachen, Germany). These so-called SK8/18-2 cells are grown in culture dishes in high glucose (4.5 g l⁻¹) Dulbecco's modified eagle medium (E15-810, PAA Laboratories GmbH, Pasching, Austria) with 10% fetal calf serum (Invitrogen GmbH, Darmstadt, Germany), 100 units ml⁻¹ penicillin and 0.1 mg ml⁻¹ streptomycin (Sigma-Aldrich, Munich, Germany) at 37 °C in a water-saturated atmosphere with 5% CO₂.

Silicon nitride (Si₃N₄) membranes (frame size: 5 × 5 mm², frame thickness: 200 μm, membrane size: 1.5 × 1.5 mm², membrane thickness: 200–1000 nm, from Silson Ltd, Blisworth, England) are placed with the flat side facing up in a culture dish and covered with cell culture medium. Cells are detached from a confluently covered culture dish with 0.25% (v/v) trypsin (Sigma-Aldrich) and 0.02% (w/v) EDTA (Roth, Karlsruhe, Germany) in phosphate buffered saline (PBS), suspended in medium and spread over the Si₃N₄ membranes [32]. After 1–2 days of incubation at 37 °C in a water-saturated atmosphere with 5% CO₂, the samples are briefly washed with PBS, fixed by adding 4% paraformaldehyde solution (Sigma-Aldrich) for 15 min at room temperature and washed three times with PBS. At this stage, phase contrast and epifluorescence microscopy images of all samples are taken using an inverted microscope (IX 71, Olympus, Hamburg, Germany) equipped with 10 ×  and 20 ×  objectives. Fixation of the cells with paraformaldehyde solution prevents changes in cellular morphology during further sample preparation and therefore allows for the comparison of microscopy images of single cells taken at different steps of the sample preparation.

The samples are washed in ultrapure water to avoid salt precipitation on the membrane and plunge-frozen by injection into liquid ethane (Leica EM GP, Leica, Vienna, Austria) to prevent crystallization of water in the sample [33]. All samples are stored for at least 24 h in liquid nitrogen and subsequently lyophilized in a home-built freeze-drier. Afterwards, the samples are warmed up to room temperature and imaged again by phase contrast microscopy. The keratin network could not be imaged with fluorescence microscopy in the dry state, because the fluorophores are destroyed during sample preparation. Samples are stored over silica gel in a desiccator.

X-ray diffraction experiments are performed at the experimental hutch III of the beamline ID13 at the European Synchrotron Radiation Facility (ESRF, Grenoble, France). The beam is pre-focused by refractive beryllium lenses and monochromatized by a channel-cut Si(111) monochromator to a photon energy of 15.25 keV. Figure 1 shows a sketch of the set-up downstream of the monochromator. The beam is focused on the sample (S) by nano-focusing parabolic refractive x-ray lenses (L) [34, 35] and cleaned by an electron microscopy aperture (A) and a pin hole (P), yielding a spot sizes of about 140 × 110 and 200 × 125 nm² (horizontal × vertical) at two different beamtimes and a primary beam intensity of about 3 × 10⁹ cps. The samples are mounted on a scanning stage comprising a hexapod for coarse sample positioning and a piezo stage for fine translations during scans. Behind the sample, the primary beam is blocked by a beamstop (BS) and the scattered intensity is recorded using a Maxipix detector (ESRF) with a resolution of 516 × 516 pixels and a pixel size of 55 × 55 μm² at a sample-to-detector distance of about 0.9 m. For sample alignment prior to the x-ray measurements, a visible light microscope (M) is moved into the beam path. The focus of the microscope is adjusted to coincide with the x-ray focus, which allows for finding the focal plane of the x-ray beam for all samples easily. After alignment with the visible light microscope, coarse mesh scans with a step size of 1–2 μm and 0.1 s exposure time are performed on a large sample area to assure precise sample positioning. X-ray dark-field images in real space are calculated by integrating the scattered intensity outside a chosen dark-field radius around the primary beam for each diffraction pattern in one scan. These values are then arranged in a two-dimensional array with dimensions corresponding to the number of scan points in the lateral directions. Therefore, each pixel in an x-ray dark-field image corresponds to the total scattered intensity at the corresponding scan position. The dark-field radius is chosen to just cover the primary beam on the detector. Figure 2(b) shows an x-ray dark-field image of a coarse mesh scan on the sample. Scans with smaller step sizes and longer exposure times are performed on selected regions of interest, as shown in figures 2(c) and (d). Here, we present results from three scans for which the experimental parameters are listed in table 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Based on the fluorescence microscopy images of the keratin network, as shown in figure 2(a), we particularly select cells with a distinct network morphology allowing for a correlation of results from x-ray diffraction experiments with structural information obtained from visible light microscopy. One interesting example is the cell in the center of figure 2(a) exhibiting a long extension (red arrow), which contains several thick keratin bundles. The corresponding x-ray dark-field image in figure 2(b), reconstructed from a coarse mesh scan with a step size of 2 μm and an exposure time of 0.1 s, shows a high contrast between cellular material and substrate, even though biological samples are generally weak scatterers. In particular, the nuclei of the imaged cells show high scattered intensity, but also the periphery and the cellular extension pointed out in figure 2(a) can be clearly identified. Note that not all cells express keratin and therefore some cells, which are visible in x-ray dark-field images (and also in visible-light phase contrast images, not shown here), are not visible in the corresponding fluorescence microscopy images.

On the keratin-rich extension, a region of interest scan with a smaller step size of 250 nm and a longer exposure time of 10 s is performed. Again, the dark-field image in figure 2(c) shows a high contrast between the cellular extension and the substrate. Further, the integral intensity is not homogeneous over the whole extension, but varies from scan point to scan point indicating a substructure in the sample. When looking at individual diffraction patterns, shown as composite image for a small region of this scan in figure 2(d), highly anisotropic scattering signals are visible, which indicate an oriented substructure in the sample. We can therefore not only distinguish between cellular material and empty regions, as shown by the dark-field images, but we can also use x-ray scattering patterns in reciprocal space to obtain information about the local sample structure.

Individual diffraction patterns from the sample exhibit highly anisotropic scattering signals, as can be seen from the raw data diffraction pattern in figure 3(a). The anisotropy of the scattering signal, and accordingly the anisotropy in the sample structure, provides additional information to the x-ray dark-field contrast images. To capture the signal anisotropy we determine the overall degree and direction of signal orientation in each diffraction pattern by approximating the thresholded intensity distribution by an ellipse. For this analysis, we first process the raw data by masking dead or hot pixels and pseudo pixels without sensitivity, filtering the image with a 3 × 3 median filter to reduce the noise and thresholding the image with a threshold value of 0.1 cps. To be able to treat diffraction patterns from empty regions, which show nearly no intensity in the thresholded diffraction patterns, with the same method, we add a circular intensity distribution in the central region of the diffraction pattern where the intensity is blocked by the beamstop. This assures that diffraction patterns from empty regions yield an approximation by a circle, which corresponds to no anisotropy in the scattering signal. The thresholded intensity distribution is then approximated by an ellipse, which has the same second moments as the intensity distribution. Figure 3(b) shows a thresholded diffraction pattern along with the approximated ellipse and the principle axes. From the ellipses we determine the direction of the minor axis in reciprocal space, which indicates the average direction of oriented structures in the sample in real space at the measured position, and the eccentricity, which gives a measure for the average degree of orientation. Together, the degree and direction of orientation in the sample can be visualized as a vector field, where the arrow length and direction are determined by the eccentricity and the direction of the minor axis of the ellipse, respectively. An overlay of the vector field, where the arrowheads of the vectors are omitted due to no physical relevance in this analysis, along with the corresponding dark-field image for the scan on the cellular extension is shown in figure 3(c). The vector field indicates structures in the sample that are oriented in parallel with the direction of the cellular extension, which is in-line with the fluorescence microscopy images of the keratin network in figure 2(a).

Zoom In Zoom Out Reset image size

The same analysis is applied to a mesh scan with a step size of 100 nm covering an area of 10 × 10 μm² on a different cell of the same type. An overlay of the corresponding dark-field image and the orientation map is presented in figure 4 and a fluorescence microscopy image of the keratin network, where the scanned region in the periphery of the cell is marked by a red box, is shown as an inset. The dark-field image shows a network-like intensity distribution with apparent feature thicknesses of about 200–400 nm. This is at the resolution limit due to the used beam size. The local orientation, determined from scattering signal anisotropy, agrees very well with the direction of the elongated structures in the dark-field image. The feature thickness in the x-ray dark-field image along with the structure orientation indicate that the observed structures correspond to the keratin network in the sample. Differences between the x-ray dark-field image in combination with the orientation map, and the fluorescence microscopy image might be attributed to different factors. The fluorescence images have comparatively low resolution and therefore thin bundles are invisible. Further, the fluorescence microscopy images are taken on hydrated samples in buffer, whereas the diffraction experiments are performed on freeze-dried samples and artifacts due to plunge-freezing and freeze-drying cannot be excluded.

Zoom In Zoom Out Reset image size

A typical method to determine the anisotropy in the scattering signal is by radial integration of the scattered intensity in a ring in reciprocal space and fitting the obtained azimuthal intensity curve. Here, the amplitude yields a measure for the degree of sample orientation. Further, specific structures with characteristic length scales in the sample can be visualized by selecting a certain region of momentum transfer. Recently, this method has been also applied to small angle scattering from biological samples, i.e. a mouse soleus muscle [26] or the dentinal collagen network in human teeth [25, 36, 37]. In contrast to these samples with well-defined and well-known real space structures, the structure of keratin bundles in our samples is probably not perfectly homogeneous and would therefore lead to scattering in different q-regions. Hence, in our case, it is important that the analysis allows for the determination of the anisotropy of the complete diffraction pattern and not solely of a selected region of momentum transfer in reciprocal space corresponding to a specific range of structure sizes in real space.

At each scan point a scattering pattern in reciprocal space is recorded. The scattering signal is analyzed by azimuthal integration of single and averaged diffraction patterns. To account for anisotropy in the scattering signal, the reciprocal space is divided into eight angular segments and the first and fifth segments are centered symmetrically around the major axis of the ellipse approximating the scattering signal. Average diffraction patterns recorded on the cellular extension and on the empty region are shown in figures 5(a) and (b) and the angular segments are indicated by dashed, white lines. For azimuthal integration of the average diffraction pattern from the empty region, the same positions of the angular segments as for the cell region are used. The insets indicate from which regions of the scan the diffraction patterns are used to obtain an average diffraction pattern from the cellular extension and the empty region. Only diffraction patterns recorded at the black positions are used for averaging. Azimuthal integration of the average diffraction pattern from cell and background regions yield the radial intensities shown in figure 5(c) and segment-wise background subtraction results in the radial intensity profiles shown in figure 5(d). As expected intuitively, the radial intensity profiles in figure 5(d) can be classified into three groups: the intensity profiles in the direction of the major and minor axes of the ellipse show highest and lowest scattering intensities, respectively, and the intensity profiles in between the principle axes show intermediate scattering intensity. All radial intensity profiles can be well described by a power-law decay of the form f(q) = aq^b + c, as shown in figure 6 exemplarily for one segment. The power-law exponents obtained from the fits are in the range from −4 to −4.25. Here, leaving out data points in the lowest q-region for fitting slightly changes the power-law exponents, which stay, however, all below −4. Similar diffraction experiments on freeze-dried bacterial cells yielded exponents −3 ⩽ b ⩽ − 4 for the decay of the radial intensity [38]. Corresponding to the well-known Porod law [39], compact two-phase systems with a sharp interface exhibit a power-law decay with b = −4 and further, power-law exponents smaller than −4 have been attributed to structures with diffuse boundaries [40, 41]. In our case, the power-law exponents are still close to the classical Porod exponent of −4, but consistently deviate to smaller numbers.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The radial intensities obtained from single diffraction patterns do not show a smooth decay, but a structured signal with weak scattering maxima and minima. Figure 7(a) shows a diffraction pattern recorded on the edge of the cellular extension along with dashed white lines indicating the segmentation for azimuthal integration. In figure 7(c), the corresponding radial intensities after segment-wise background subtraction are shown. For background subtraction the radial intensities of the averaged empty diffraction pattern (see figure 5(b)) are calculated with the segmentation of the single diffraction pattern and subtracted for each segment separately. Particularly, the radial intensities in the first and fifth segment, which are perpendicular to the local structure direction, show scattering maxima as indicated by arrows in figure 7(c). This intensity distribution might be explained as the form factor of a cylinder with a diameter of about 42 nm, which could correspond to a keratin bundle or sub-bundles. Deviations of the radial intensity from the form factor of a cylinder (fit data are shown in the supplementary information, available from stacks.iop.org/NJP/14/085013/mmedia) might be attributed to the rough bundle surface due to protein side chains and the substructure of the bundle, which is made of several filaments with a diameter of about 10 nm.

Zoom In Zoom Out Reset image size

In contrast to the above observation, the same analysis applied to an average of 2 × 2 adjacent diffraction patterns, including the diffraction pattern shown in figure 7(a), reveals a smoother decay of the radial intensity profiles without pronounced maxima or minima in the scattering signal (see figures 7(b)–(d)). This finding indicates small changes in the peak positions, which result in a smooth intensity decay after averaging. Therefore, the detailed intensity distribution in the scattering patterns in reciprocal space and correspondingly also the local sample structure in real space, sensitively depend on the position on the sample. Taking into account the beam size and step size during the scan, the average of 2 × 2 diffraction patterns covers an area of about 400 × 400 nm² on the sample and thus the maximum beam size for probing the local structure in this type of samples needs to be smaller than about 400 × 400 nm².

Due to the small beam size and the high x-ray flux in our experiments, damage of the samples caused by the radiation cannot be avoided. Following the approach by Howells et al [42], we estimate the radiation dose based on the assumption that all cellular material consists of protein with the empirical average formula H₅₀C₃₀N₉O₁₀S and mass density ρ = 1.35 g cm⁻³. The average radiation dose during the scan is calculated using

$$\begin{eqnarray} &&D = \frac{I_0\,T h \nu}{d \rho \Delta_x \Delta_y}, \end{eqnarray} \tag{ 1 }$$

with I₀ = 3 × 10⁹ s⁻¹ being the primary beam intensity, T the exposure time, hν = 15.25 keV the photon energy, d = 4.964 mm the attenuation length in the sample as obtained for the assumed average cellular material [42] (http://www.cxro.lbl.gov/), ρ the mass density, and Δ_x and Δ_y the step sizes in the lateral directions. The estimates of the radiation dose in gray (Gy = J kg⁻¹) for the three scans as well as the scan parameters are listed in table 1. For fine scans with long exposure times and small step sizes the radiation dose is of the order of 10⁸ Gy. After these scans, sample thinning and shrinkage at the edges was observed, indicating visible damage of the sample due to the radiation. The visible effects of radiation damage on the sample might be reduced by cooling the sample with a cryo jet and by using frozen-hydrated instead of freeze-dried sample. However, this imposes further geometrical limitation to the set-up and especially the temporal stability might decrease due to thermal drifts.