Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
Imaging applications of synchrotron X-ray phase-contrast microtomography in biological morphology and biomaterials science. I. General aspects of the technique and its advantages in the analysis of millimetre-sized arthropod structure
Zoologisches Institut der Universität, Abteilung Evolutionsbiologie der Invertebraten, Auf der Morgenstelle 28E, D-72076 Tübingen, Germany
Synchrotron-generated X-rays provide scientists with a multitude of investigative techniques well suited for the analysis of the composition and structure of all types of materials and specimens. Here, we describe the properties of synchrotron-generated X-rays and the advantages that they provide for qualitative morphological research of millimetre-sized biological organisms and biomaterials. Case studies of the anatomy of insect heads, of whole microarthropods and of the three-dimensional reconstruction of the cuticular tendons of jumping beetles, all performed at the beamline ID19 of the European Synchrotron Radiation Facility (ESRF), are presented to illustrate the techniques of phase-contrast tomography available for anatomical and structural investigations. Various sample preparation techniques are described and compared and experimental settings that we have found to be particularly successful are given. On comparing the strengths and weaknesses of the technique with traditional histological thin sectioning, we conclude that synchrotron radiation microtomography has a great potential in biological microanatomy.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
A detailed knowledge of the interior of biological structures and organisms is crucial for a better understanding of their function and evolution. Traditionally, in both biological morphology and anatomy, the internal structures of whole organisms or parts of them are accessed by dissecting or histological serial sectioning. The resulting situs or sections thereof can then be viewed in their natural state, whereas histological sections can be stained and viewed by light microscopy (LM) at a spatial resolution of about 300 nm or by transmission electron microscopy (TEM) at a spatial resolution of about 1 nm. In addition to imaging, both techniques offer a wide range of analytical techniques such as histochemistry, immunogold labelling, autoradiography, freeze-fracturing and negative staining.
From these methods, biologists can choose the best suited to answer a particular question in an attempt to understand better the composition, structure or function of a whole organism, a tissue, an individual cell or the components of a cell (e.g. Harris, 1991; Bozzola & Russel, 1998; Murphy, 2001).
Whereas LM and TEM provide information mainly in two dimensions (2D), scanning electron microscopy (SEM) allows the surface of animals (Arens, 1998) or plants (Barthlott, 1990) to be viewed not only at high spatial resolution (∼5 nm), but also as three-dimensional (3D) objects, because of the extensive depth of focus. With an environmental SEM (ESEM), many biological samples can now even be investigated in their natural or wet state (Füting, 2003). Confocal laser scanning microscopy (CLSM) has been successfully used not only for visualizing cellular structures, but also to produce 3D images of larger anatomical structures of arthropods such as genitalia, legs or internal organs with submicron resolution (e.g. Heinstra & Thörig, 1982; Zill et al., 2000; Klaus et al., 2003). However, this method can only be applied to transparent structures, which need to be fluorescent. Moreover, the visualization of thick specimens (more than several hundred micrometers) might be affected by aberration artefacts (Klaus et al., 2003), and the x-, y-dimensions of the sample to be scanned are restricted by the object field diameter, which is about 1.5–2 mm.
A nondestructive technique enabling both a qualitative and a quantitative analysis of the complete structure and anatomy of a nontranslucent organism is thus much desired, especially if a time-consuming sample preparation method can be avoided.
True representations of the internal 3D structure of small-scale objects in the range of a few millimetres or smaller are traditionally reconstructed from dissections or from 2D LM or TEM images of thin sections. However, these reconstructions are the result of an elaborate process that requires great artistic skill and patience, as the impressive results of several insect anatomists (Snodgrass, 1935; Weber, 1966; Weber & Wenk, 1969) well illustrate. Research that relies on these traditional methods for 3D imaging, such as comparative studies of several developmental stages, individuals or species, is highly time-consuming both in sample preparation and in interpretation and is thus expensive and sometimes impossible. Another disadvantage is that these traditional techniques are destructive, so that true 3D morphometry of internal anatomical structures is virtually impossible.
Synchrotron-based radiography and X-ray microtomography
One method that fulfils all of these requirements is X-ray tomography. This technique has been used for large scale biological samples since the 1980s, when it was developed primarily for medical applications with a resolution of about 1 mm (Kalender, 2006). Since the arrival of synchrotron radiation, the resolution of tomography has improved dramatically, so that samples can now be imaged at submicrometer resolution, with effective pixel sizes down to about 0.25 μm.
Absorption was, for a long time, the only contrast mechanism that could be used for the imaging and reconstruction of 3D structures. This meant that X-ray tomography could be applied successfully only to samples of materials that had either a high X-ray attenuation coefficient or that consisted of components with sufficiently large differences in their attenuation coefficients. The arrival of phase-contrast imaging (Snigirev et al., 1995; Cloetens et al., 1996), developed specifically for the imaging of materials with low X-ray attenuation coefficients such as polymers and many biological materials, changed this state of affairs shortly before the turn of the millennium. The radiographic observation and quantification of tracheal volume changes connected with respiratory ventilation in insects (Westneat et al., 2003) represented one of the first experiments in which synchrotron-based phase-enhanced X-ray imaging was applied to biological structures and materials and demonstrated the extent to which the bio-sciences can benefit from this method. Only recently, there has been published a review on real-time phase-contrast X-ray imaging by Socha et al. (2007).
Below, we focus on the application of synchrotron-based X-ray microtomography (SR-μCT) with phase-contrast to millimetre-sized biological samples and materials and illustrate, with a range of case studies, the wealth of information that this nondestructive 3D imaging technique provides.
The main reasons that synchrotron X-rays are often preferred to desktop-generated X-rays are that they offer a significantly higher resolution, a better signal-to-noise ratio, short acquisition times and quantitative reconstructions and, last but not least, provide phase contrast in addition to absorption contrast imaging. The goal of this contribution is to introduce SR-μCT to a broader community of biologists and biomaterial scientists and to provide practical recommendations for sample preparation and parameter settings at synchrotron facilities. We focus exclusively on millimetre-sized samples that have negligible absorption contrast. First, we explain the underlying physical and technical principles and evaluate and compare the effects of various beamline settings and sample preparation methods. This part also includes a discussion on the principles and practice of data acquisition, including data reconstruction and rescaling. We illustrate the above-mentioned points with case studies from our own research, which comprises biological materials and structures ranging from whole microarthropods, such as mites, to insect heads and legs. After a comparison of SR-μCT with conventional histological techniques we finally introduce the practical application of some currently available software systems for 3D visualization and voxel data analysis.
Principles of the technique
X-rays cover a part of the spectrum of electromagnetic radiation with wavelengths of approximately 10−9–10−11 m, which is about 1000 times shorter than that of visible light (4–7·10−7 m). The shorter wavelength permits the achievement of a significantly higher resolution with X-rays than with light. Another property that makes X-rays interesting for the investigation of structures and materials is that they have, in contrast to visible light, which is reflected at the surface of most objects, a high penetration depth. Moreover, the interaction with matter is weak and, as a result, X-rays travel predominantly in straight lines through matter. These properties make it possible to image the internal structures of materials and organisms.
X-ray tube generated X-rays
Since Röntgen's discovery of X-rays in 1895 and up until the arrival of cyclotron and synchrotron radiation, X-ray tubes were the most common source of X-rays. Because of their small size, they are still the standard in medical computed tomography (CT) scanners and desktop microscanners. The X-rays are generated by accelerating electrons in a strong electric field from a cathode source onto a metallic anode, which acts as a target in a collision experiment. When the electrons collide with the metal target (typically a tungsten block on a copper support), two types of interaction occur. Electrons of sufficient energy knock out electrons from the inner shells of the target atoms so that, subsequently, electrons from higher energy levels can fill up the vacancies, causing the so-called characteristic X-rays, typical of any chemical element, to be emitted. The others are decelerated as they are scattered by the strong electric field near the nuclei of a target material with high atomic number Z; this causes a continuous spectrum of electromagnetic radiation, the so-called bremsstrahlung, to be emitted.
In a synchrotron, only the equivalent of bremsstrahlung is generated because of the transverse acceleration of electrons travelling in a magnetic field. The setup and functional principle of a synchrotron facility is depicted in Fig. 1(A), taking the European Synchrotron Radiation Facility (ESRF) in Grenoble (France) as an example. Electrons are produced and emitted by an electron gun. They are accelerated first in a linear accelerator (linac) and then in a circular accelerator (booster ring) to reach the final energy level, which is 6 GeV at the ESRF. These high-energy electrons are then injected into a large storage ring, where they circulate without gaining further energy and with a velocity close to that of the speed of light. Booster and storage rings are not perfectly round but consist of straight sections connected by curved ones in which dipole magnets bend the path of the electrons into the next straight section. To produce X-rays by means of the circulating electrons, they need to be decelerated or accelerated. This happens in the bending magnets in the curved sections and, for modern third generation storage rings, in the straight sections by so-called insertion devices. The latter are periodic arrays of alternating dipole magnets that cause small undulations in the electron path. They can be separated into wigglers and undulators. Whereas the radiation produced in bending magnets and wigglers can cover a broad spectrum from microwaves to hard X-rays, that produced by undulators (which cause, in comparison with wigglers, only small undulations in the electron path) exhibits sharp peaks at certain energies provoked by interference effects of the emitted radiation on the electron beam path. In all cases, the radiation is emitted in a well-defined direction, tangent to the electron storage ring, and is used in tangentially arranged experimental setups (beamlines) for diffraction, scattering or imaging experiments.
The imaging beamline ID19 (ID refers to an insertion device as source) of the ESRF, used for the majority of the experiments described here, is depicted in Fig. 1(B). Its setup was optimized for the use of a spectrally and spatially homogeneous, highly coherent, X-ray beam with maximum dimensions of 45 by 15 mm at the sample position, a high photon flux and a tunable photon energy ranging from 6 to 120 keV. The 145-m-long beamline has three insertion devices: two undulators and a wiggler. The long source-to-sample distance and the small source dimensions (∼0.1 mm) ensure a highly spatially coherent beam. The optical hutch of ID19 is located just behind the front-end, where the beam leaves the storage ring, and contains diaphragms, slits, filters and a shutter to control the bandwidth, photon flux and beam dimensions of the X-rays. The satellite laboratory houses the monochromators and the experimental hutch and is joined to the first hutch by a tunnel of ∼100 m in length.
In addition to the ESRF, most synchrotrons worldwide have dedicated imaging beamlines with microtomography capabilities. Although most of these facilities do not have the advantage of the small source size and the long beamline as the ESRF ID19, X-ray CT is routinely performed with spatial resolutions of a few micrometers. Major efforts in biomedical imaging are ongoing at the Elettra (Italy), NSLS (USA), PLS (Korea) and SLS (Switzerland).
Absorption contrast imaging versus phase-contrast imaging
The contrast mechanism traditionally exploited in X-ray imaging, such as radiography or tomography, is X-ray absorption (Fig. 2: above). This is based on the fact that the intensity of the beam decreases exponentially with increasing propagation distance within homogeneous matter. The attenuation coefficient of the material determines the speed at which this exponential decrease occurs.
For a given wavelength or energy of the X-rays, the absorption coefficient increases with the electron density of the material and with the atomic number of its constituting chemical elements: the higher the atomic number and atomic weight, the stronger the absorption, and thus the larger the difference in the atomic numbers, the larger the contrast. It is therefore not surprising that absorption contrast is ideal for the medical imaging of high-density bone surrounded by low density tissue, such as muscle and skin.
However, absorption contrast is not well suited for the characterization of materials composed of constituents with low absorption and only small difference in atomic number, such as polymers and nonmineralized biological materials. Such materials are best imaged with phase contrast. Whereas phase contrast tomography enhances the contrast for the fine features within a sample, pure absorption tomography at the micron and submicron scale often does not show much fine detail. This is connected with the lack of contrast and with limitations in resolution. Another important issue is that in fragile samples like ours, the speed of acquisition can become crucial to avoid sample drift during acquisition and the associated blurring.
Since phase-contrast imaging (Fig. 2: below) requires a homogeneous X-ray beam with spatial coherence, synchrotron radiation is particularly well suited for this technique. Phase contrast takes advantage of the fact that the X-ray beam is not only absorbed when it penetrates matter, but also that the phase of the wave is affected. This effect is determined by the refractive index of the material. Except in the vicinity of absorption edges, the change in refractive index is proportional to the electron density or, to a good approximation, to the mass density. However, unlike absorption, phase changes do not influence the wave amplitude of the beam, and thus cannot be measured directly behind the sample. The image is instead created by the interference of the diffracted components of the beam with itself and the nondiffracted part of the beam. At the detection plane, these interferences are expressed in lateral intensity variations. Thus, a lateral coherence is required for this imaging technique; this property increases with a decrease in the source size of the beam and/or an increase in the source-sample distance. In general, within an applicable range, the contrast and the width of the interference fringes increase with increasing distance between sample and detector.
Hence, phase-contrast imaging enables the mapping of the changes in the refractive index. Interfaces between materials with different densities correspond to discontinuities in the refractive index, which in turn lead to phase jumps causing Fresnel diffraction to occur, appearing as white–black fringes in the images (Cloetens et al., 1997). The great advantage of phase-contrast imaging is that it produces edge-enhancement effects, which improve the visibility of small structures and which make it possible to image materials with small variations in mass density and absorption contrast. For example, at 25 keV, the maximum absorption contrast of a 100-μm diameter tube of air in water (e.g. trachea) is only 0.3%. In comparison, with phase-contrast imaging, the edge contrast of the tube can be more than 50%. Thus, even though there is essentially no absorption differences, the edges of the tube can be clearly visualized.
One disadvantage of this technique is that the usual tools for quantitative analysis of structures are adapted for absorption-based CT datasets and that they cannot be applied to phase-contrast datasets. Quantitative phase-contrast techniques, whereby the actual density of every point in the sample is recovered, have been demonstrated (Nugent et al., 1996; Cloetens et al., 1999), but the quantitative density recovery process is tedious and requires additional datasets. For the study of small animal morphology, a qualitative visualization of the structures is usually sufficient; and so, in this study, we utilized the standard absorption-based reconstruction software even though our datasets clearly show phase-contrast. As a result, a certain amount of caution has to be taken with regards to the interpretation of a particular reconstructed image. Nonetheless, we emphasize that without phase-contrast, in most of the images and 3D renderings shown in this paper, the recognition and distinction of minute adjacent structures with similar grey values (e.g. tendons, muscle attachment sites and cuticular sclerites) would have been much less straightforward.
Although in this paper we present phase-contrast data taken at a third generation synchrotron (ESRF) using highly spatially coherent monochromatic X-rays, it is important to point out that monochromaticity and spatial coherence requirements for qualitative phase-contrast imaging are very minimal (Margaritondo & Tromba, 1999). In fact, X-ray phase-contrast imaging has been demonstrated at second generation synchrotrons (Elettra Highlights 1998–1999; Kowalski et al., 1999), and with polychromatic beams using a microfocus tube source (Wilkins et al., 1996). It should also be pointed out that the technique described in this paper is sometimes called propagation-based phase-contrast, to distinguish it from interferometer-based phase-contrast (Ando & Hosoya, 1972) and analyser-based phase-contrast (Chapman et al., 1997).
Experimental setup and data acquisition
Imaging with X-rays in 2D is carried out by a technique known as radiography, viz. the form of imaging that Röntgen used in his experiments and that is widely employed for medical imaging and the nondestructive testing of components. The resulting radiographs provide no depth information, since spatial features at different penetration depths are superimposed in the image, because, according to the Beer-Lambert law, radiographs depict the projection of the linear attenuation coefficient (μ) integrated along the propagation path of the X-ray beam.
To avoid the superposition of information and to allow the reconstruction of 3D information from an object, computed tomography (CT) was developed in the 1960s. The principle of CT involves several steps (Fig. 3) (Zollikofer & Ponce de León, 2005), which are explained in the following for the parallel-beam case, as frequently encountered at synchrotron imaging setups. First, 2D radiographs of an object are taken at equally spaced angular positions when the sample is rotated from 0° to 180°. For each layer, the set of projections results in a so-called sinogram, which comprises the total sequence of attenuation profiles of one 180° turn around the object. In a second step, the 3D structure of the sample is reconstructed numerically layer by layer from these sinograms. This is usually performed via so-called back-projection algorithms, which are based on Radon's theorem (Radon, 1917). The original object information can then be recovered from the set of line integrals taken around the object. In a third step, the tomographic dataset, consisting of voxels (3D pixels), is visualized by using specialized volume graphics software.
The μCT reconstruction algorithms require the entire sample (including possible appendages) to be sampled at each angle of rotation. Hence, structures such as the antennae of insects, which protrude a long way out of the sample under study, should be removed prior to the experiment to avoid their temporary disappearance from the field of view. Since the sample rotates about the axis of the aluminium pin (Fig. 4), only the dimension that is orthogonal to the pin axis must fit into the field of view (cf. hatched line in Fig. 4). As the detectors have a limited number of pixels (typically 2048), this requirement imposes a compromise between the obtainable spatial resolution and the required field of view (see also Table 1).
Table 1. Overview of the currently available fields of view and corresponding spatial resolutions for μCT studies at ID 19 of the ESRF (Grenoble). In the binning mode, two vertical and two horizontal pixels are grouped to one pixel to reduce noise. The resolution of the optics is specified as the full width at half maximum of the measured point spread function.
Optics – spatial resolution (μm)
Pixel size (μm)
Field of view (mm)
Pixel size (μm)
Field of view (mm)
Spatial resolution attainable in X-ray tomography
The optical paths of the X-rays in a synchrotron versus a desktop μCT facility are shown in Fig. 5. Because of the long distance between source and sample, the resulting parallel geometry of the synchrotron beam ensures that the image on the charge-coupled device (CCD) detector has the same scale for all sample detector distances but does not provide a means for geometrical magnification. In contrast, we can see that the strongly divergent cone-beam geometry of the microfocus source allows significant magnification of the radiograph projected on the detector by variation of the distances source-sample and sample-detector. Cone-beam geometries involve some intrinsic distortions of the 3D images but they allow standard microfocus sources (X-ray tubes) combined with flat-panel or CCD-based detectors to achieve practical spatial resolutions down to approximately 5–10 μm.
The present limitations of cone-beam systems are mainly attributable both to the source sizes (i.e. the area from which the X-rays are emitted) and to the photon flux (i.e. the number of photons per unit time and per unit area on the detector): at reasonable photon flux, the source size, which is responsible for image blurring at high magnifications, cannot be further reduced, primarily because of heat-load problems. Drifts and instabilities related to the long exposure times and Fresnel fringes intrinsic to the projection geometry further deteriorate the spatial resolution.
Modern SR-μCT facilities achieve spatial resolutions of less than 1 μm, which is limited, in parallel beam geometry, by the detector. The sinograms are recorded with a detector system that consists of an X-ray to light converter (scintillator) and with an optical system (mirror and lenses) to project the image onto the sensory chip of a CCD camera. The sensory chip consists of an array of coupled capacitors that are sensitive to light. The number of capacitors (e.g. 2048 by 2048) on the chip determines the size of the resulting data file. The choice of the thickness of the scintillator screen is a compromise between adequate detection efficiency (favouring thick screens) and high spatial resolution (requiring thin screens).
The achievement of spatial resolution in X-ray tomography below 500 nm (the current limit of scintillator screens) requires some kind of X-ray magnification and is still a domain of active research. On the one hand, the projection or cone-beam geometry has been further developed both in lab facilities by using, for example, a field-emission scanning electron microscope to produce a tiny X-ray source (Mayo et al., 2003), and in synchrotron facilities by using mirror based X-ray optics to focus the beam and to create an intense monochromatic divergent beam (Pereiro et al., 2005). On the other hand, X-ray lenses are becoming available to magnify the image in the same way as an objective lens does in a visible light microscope (Yin et al., 2006). This approach is particularly successful for tomography with softer X-rays, where it allows the achievement of sub-100-nm spatial resolution on minute (<10 μm) specimens such as single cells (Schneider et al., 2002; Larabell & Le Gros, 2004). For the studies presented in this work, the spatial resolution, field of view and contrast mechanisms offered by parallel beam synchrotron μCT are best suited.
For all the tomographic experiments at ID19 of the ESRF, we have used a FRELON (Fast REadout LOw Noise) CCD camera with visible light and scintillator optics (designed and produced at the ESRF). This camera has a 14-bit dynamic range and a readout time of 0.12 s for 2048 by 2048 pixels and of 0.06 s for 1024 by 1024 pixels (obtained by binning the pixels two by two). By using a range of combinations of scintillators and optics, effective pixel sizes ranging from 0.28 to 60 μm can be achieved by magnification via microscopy or by demagnifying optical setups (Table 1). The whole sample-camera ensemble sits on a one-meter precise translation mechanism to enable the sample-to-detector distance to be changed; this parameter is of great importance with respect to taking advantage of the coherence of the beam and to choosing between absorption or phase-contrast dominated imaging.
The time required to acquire a tomographic dataset strongly depends on the settings. It can be attained as quickly as 10 s, thereby allowing dynamic studies. For a 2048 by 2048 pixels dataset with 1500 angular positions, the scan duration is typically 20 min, whereas this reduces to typically 5 min for a 1024 by 1024 pixels dataset with 900 angular positions. At the full camera resolution, a tomographic dataset recorded with a 2048 by 2048 pixels camera results in a reconstructed 8-bit dataset of 8 Gigabytes. Powerful computational tools are thus required for the 3D reconstruction, which currently takes 2–3 h at the ESRF, for visualization and for segmentation.
Advantages of synchrotron-based CT to desktop-based CT
Laboratory- or desktop-based X-ray CT machines are becoming increasingly common and available at most research institutions. In addition, many institutions provide an avenue of access to these machines, even for external researchers. Laboratory- or desktop-based CT systems typically operate with the cone geometry using the raw spectrum from the X-ray source whereas most synchrotron-based CT measurements are performed using monochromatic (ΔE/E ∼ 10−2–10−4) X-rays in a parallel beam geometry (cf. Fig. 5). Synchrotron-based CT offers three significant advantages: (1) wide energy range (4–200 keV), (2) high flux and (3) small source size and beam divergence. Thanks to these properties, thick dense samples can be imaged within a short time period; in addition, the spatial coherence properties allow for phase-contrast techniques. The high photon flux allows for monochromatic beam imaging with very narrow energy bandwidths (ΔE/E ∼ 10−2–10−4) that is advantageous for quantitative absorption and phase imaging and for element-sensitive absorption imaging near absorption edges of specific chemical elements. Acquisition times of synchrotron-based systems, for comparable spatial resolutions, are usually 10–100 times faster. Furthermore, despite claims of micron to submicron capabilities of desktop-based systems, most X-ray μCT publications that report spatial resolutions below 5 μm have been synchrotron-based. For soft-tissue morphology of millimetre-sized insects, where phase-contrast and submicron-level spatial resolutions are required, synchrotron-based CT is currently the only option.
Anatomy of insect heads
Morphological studies on insects often focus on a detailed analysis of the head or parts of it (Snodgrass, 1928; Matsuda, 1965; Beutel, 1999; Hörnschemeyer et al., 2002; Beutel & Weide, 2005; Waloszek et al., 2005). The insect head shows a broad spectrum of structures, such as sensory organs, endo- and exo skeletal features, muscles, brain or mouthparts, whose specific organization can act as an aid to understanding the evolution, phylogeny, development and ecology of the insect under study. The head of an insect can enclose up to 60 muscles that move the antennae and mouthparts. In addition, several muscles are associated with the fore gut (von Kéler, 1963). Among the different taxa, the head muscles may not only vary in terms of their dimensions, but also in their points of origin and insertion, both of which have significant functional consequences for the kinematics and force exertion of the mouthparts.
In this example, we demonstrate that synchrotron μCT (SR-μCT) can be successfully applied to the study of the head anatomy of small insects such as rove beetles (Coleoptera, Staphylinidae) with head diameters of about 0.5 mm.
Sample preparation and experimental setup The fore bodies of alcohol-fixed or freshly killed beetles were cut between the pro- and mesothorax, stepwise dehydrated in ethanol/acetone and critical-point dried (CPD 020, Balzers). We used dried specimens to avoid shrinking artefacts caused by water loss during the tomography procedure, which might affect fresh samples exposed to the dry atmosphere in the experimental hutch.
The posterior end of each prothorax was attached to the tip of a plastic or aluminum pin (1.2 cm long; 3.0 mm in diameter), so that the head was oriented upright (cf. Fig. 4). For tomography, the pin was mounted onto the goniometer head of the sample stage.
High-resolution X-ray tomography was performed at beamline ID19 (ESRF, Grenoble) at an energy of 20.5 keV (wavelength of 6 × 10−11 m). The radiographs were recorded with a cooled CCD (ESRF FReLoN camera) with a 14-bit dynamic range, 2048 × 2048 pixels and an effective pixel size of 0.28 μm. This resulted in a corresponding field of view of 0.60 by 0.60 mm. Over the 180° sample rotation, 1500 projections were recorded with an exposure time of 1 s each. The detector-to-sample distance was 10 mm, which made it possible to utilize moderate phase-contrast effects. The filtered back-projection algorithm was employed (Cloetens et al., 1997, 2002) for the reconstruction of the 3D datasets from the 2D radiographs, entailing the conversion from 32-bit to 8-bit data representation. The conversion of the 32-bit data to 8-bit data involves a compromise between image intensity dynamic range and resolution. This downward bit conversion results in a loss of information and there is not a generally applicable optimal solution. This step is performed after careful inspection of the 32-bit data so as to maximize the visibility of the structures of interest. That is, for the same dataset, it might be possible to use multiple solutions in order to visualize minute differences both between and within different anatomical compartments, i.e. for example, the depiction of small differences between cuticular structures or differences in soft tissue such as muscles and nerves.
The resulting voxel datasets were visualized with VGStudio Max (Volume Graphics, Heidelberg, Germany). To separate and visualize individual structures, such as the internal tentorium, we used the segmentation tools provided by the visualization software tool amira™ (Mercury Computer Systems, Berlin, Germany).
Results We obtained complete series of virtual thin sections through the heads of selected rove beetles (Coleoptera, Staphylinidae); these could be used both for the inspection of planar sections in all three major sectional planes (i.e. transverse, sagittal and horizontal) (Fig. 6) and for 3D reconstructions of the entire head (Fig. 7). The contrast between the different tissues in the μCT images was sufficient to be able to distinguish between cuticular structures and individual soft tissue, e.g. between muscles and cuticular apodemes (Fig. 6). Muscle tissue could easily be distinguished from nerve tissue by its bundle- or fibre-like bright appearance. The nerve mass of the brain usually appeared in a slightly darker and mostly uniform grey. Although histological details could not be revealed from these sections, they were well suited for comparative studies of the general anatomy of complex body structures. In such studies, the 3D visualization tools of the relevant software packages make it possible to reconstruct the spatial organization of complex anatomical features. For instance, for identification purposes or functional morphological analyses, the origin and insertion and the course of the various muscles can easily be determined by moving back and forth within the 2D image stacks or by cutting into the reconstructed 3D volume models as shown in Fig. 7.
Complex internal structures such as the tentorium (the inner skeleton of the insect head as formed by invaginations of the exocuticle of the head capsule) and the hypopharynx-prementum sclerites can be reconstructed by the segmentation tools of current software packages for visualizing voxel data. The tentorium and the hypopharyngeal sclerites ensure the stability of the head capsule and serve as an attachment structure for a number of head muscles. The main components of the tentorium are the laminatentorium with its anterior, dorsal and posterior arms. The last-mentioned can fuse with the gular suture to build the so-called posterior tentorial wall. Further posterior, this wall is spanned by the tentorial bridge. The complex structure of the internal head and hypopharynx skeleton of insects is illustrated here for the rove beetle, Gyrophaena fasciata (Marsham) (Fig. 8).
Anatomy of whole microarthropods
Because of its high resolution, SR-μCT techniques are also particularly useful for the study of the internal anatomy of microarthropods, which are arthropods ranging from 0.2 to 3 mm in size, such as springtails and mites (Acari). Even those organs that extend through the entire body can be imaged in their natural state and reconstructed with a resolution close to that of LM. The resulting 3D datasets should help to answer questions regarding the anatomy of the digestive tract, the locomotory system, the reproduction apparatus and even the finely structured nervous system.
Oribatid mites (Acari, Oribatida) are important decomposers whose densities can reach several hundred thousand individuals per square meter in acidic soils of northern boreal forests. In oribatid mites, 10 times more species than in other metazoan taxa reproduce via parthenogenesis (i.e. self-fertilising) (Norton et al., 1993); some parthenogenetic lineages presumably have existed for more than 100 million years (Heethoff et al., 2007a). The feeding strategy of oribatid mites is unique among arachnid arthropods and involves the use of particulate rather than liquid food. These features make them suitable model organisms for both ecological and evolutionary studies. However, most morphological studies are limited to the ultrastructure of their internal anatomy because of their small size (∼0.2–1.5 mm body length) and the strong sclerotization of their cuticle. The available 3D reconstructions are therefore schematic, being based on dissections, and restricted to specific organ systems (Alberti & Coons, 1999). Here, we present the first attempt to analyse the internal anatomy of whole microarthropods by using SR-μCT techniques, exemplified by 3D renderings of the digestive system of an oribatid mite.
Sample, sample preparation and experimental setupArchegozetes longisetosus Aoki reproduces parthenogenetically and is the most-studied oribatid mite species under laboratory conditions (Smrz & Norton, 2004). Both the ultrastructure of the digestive system and the gnathosoma have been studied by electron microscopic techniques (Alberti et al., 2003). Its internal anatomy has moreover been previously examined with respect to the reproductive system (Heethoff et al., 2007b) by using SR-μCT.
The specimens of A. longisetosus used in this study were derived from a culture initiated by R. A. Norton in 1993 with a single female sampled from decomposing coconut debris at Luquillo (Puerto Rico). Fresh adult specimens were taken from the culture, cleaned with a fine brush and placed in 2.5% glutaraldehyde for 60 h. Specimens were dehydrated in a graded ethanol series and then critical point dried (CPD 020, Balzers). The posterior end of the opisthosoma was superglued onto the tips of plastic pins (12 mm long; 3.0 mm in diameter). X-ray tomography and visualization were performed as described above for insect heads with the exception that the effective pixel size was 0.7 μm and the field of view was 1.43 by 1.43 mm.
Results The digestive tract comprises (1) a cuticular foregut (mouth, pharynx, oesophagus), (2) a midgut (ventriculus, preventricular glands, a pair of caecae, a colon, an intercolon and a postcolon) and (3) a cuticular hindgut terminating with the anus (Fig. 9; cf. Alberti et al., 2003 for the functions of the digestive parts).
All these components are provided with cross-striated muscles (Figs 9 and 10). Two rows of anal muscles are connected to the anus (iam and oam in Fig. 10). Initially, similar muscles observed in related oribatid mites were thought to be anal-closing muscles (Hoebel-Mävers, 1967) but were later suggested to be ‘anal dilators’ (Akimov & Yastrebstov, 1991). The outer anal muscles insert ventrolaterally on the opisthosoma and the inner cuticular fold of the adanal plate and have a horizontal orientation. The inner anal muscles connect the inner cuticular folds of the adanal to the anal plates and act ventrolaterally. Although the actual function of these muscles remains to be uncovered, the angles of the two muscle rows support the hypothesis of a closing rather than an opening mechanism. The use of SR-μCT data should now allow the generation of a functional model of the anus including the natural orientation and insertion of involved muscles and the associated parts of the cuticle. With this model, the reaction of the anus (opening vs. closing) can be studied in silico.
This example illustrates the high-resolution structural information with which μCT aids in the analysis of the anatomy and functional morphology of microarthropods. A longer-term aim, as more data become available, is to construct and calculate functional models of whole animals to answer questions regarding functional interactions between the various muscles and between the muscles and cuticle. This will be of great value in the fields of arthropod morphology and biomechanics.
Three-dimensional rendering of cuticular tendons in jumping beetles
The ability to jump in order to escape from predators has evolved in many insect groups. Most insects use the rapid extension of the tibia in the hind legs for this sudden movement (Chapman, 1998). In beetles, jumping has independently evolved in several families (Furth & Suzuki, 1992) and has been extensively investigated in the leaf beetle tribe Alticini (‘flea beetles’) (Barth, 1954; Furth et al., 1983; Schmitt, 2004). In contrast to slow movements, the mechanism of jumping requires the sudden release of energy, which cannot be obtained from direct muscle contraction but requires the rapid release of energy previously stored in elastic elements of the muscles and cuticle (Chapman, 1998).
In the Alticini (Chrysomelidae), energy storage involves the metafemoral extensor tendon, viz. ‘Maulik's organ’ named after its discoverer (Maulik, 1929), a complicated three-dimensionally convoluted structure that provides attachment points for muscles (Maulik, 1929; Furth, 1980; Schmitt, 2004). The exact mechanism of energy storage and release is still unknown but the extensive tibia extensor muscles, which are housed in the distinctly swollen metafemur and that insert at this apodeme, in concert with a catch-like sclerite (‘Lever's triangle’) (cf. Lt in Fig. 11) connected to the tibial flexor muscles might function as a mechanical spring (Barth, 1954; Furth, 1982). Such a spring mechanism involves the cocontraction of both the tibial extensor muscles and the tibial flexor system but, for the moment, without causing extension of the tibia. This is because ‘Lever's triangular plate’ is pressed against the distal margin of the posterior femoral wall (action 3 in Fig. 11A), which is slightly curved inwards at this point (cf. fab in Fig. 11). This locking of ‘Lever's triangular plate’ makes it possible for the tensile strain energy produced by the contracting tibia extensor muscles to be stored in the elastic elements of the system, especially by distorting the convoluted ‘Maulik's organ’. Only upon release of the catch can the previously stored energy be used rapidly to extend the tibia and thus exert the required mechanical power at the tibio-femoral joint.
In this example of the chrysomelid beetle Altica sp., we present a 3D reconstruction of the metafemoral extensor apodeme together with the sclerite connected with the tibial flexor system. As representatives of the tribe Alticini, these beetles are capable of abruptly jumping away when disturbed. Our reconstruction aims at a better understanding of the functional morphology of the energy storage and mechanical catch mechanism behind the extraordinary jumping performance of these beetles.
Sample preparation and experimental setup The hind legs of freshly caught specimens of Altica sp. Geoffroy were fixed with a thread, so that the tibia was almost maximally flexed against the femur. After preservation in alcohol, the legs were stepwise dehydrated in increasing ethanol concentrations, critical-point dried (CPD 020, Balzers) and superglued onto the tips of aluminum pins (1.2 cm long; 3.0 mm in diameter). For μCT, we used the ID19 at the ESRF (Grenoble) at an energy of 15.5 keV (wavelength of 8·10−11 m) and an effective detector pixel size of 0.7 μm with a corresponding field of view of 1.43 by 1.43 mm. Over the 180° rotation, 1200 projections were recorded with an exposure time of 1 s per projection. The detector-to-sample distance was 12 mm. In other respects, our settings and reconstruction algorithm were identical to those described above for the investigation of insect heads. To reconstruct the 3D structure of the various elements of the spring mechanism, we used the segmentation tools of the volume graphics visualization software amira®.
Results The obtained series of virtual thin sections allowed us to reconstruct the spatial structure of the various elements involved in the metafemoral jumping mechanism of the investigated Altica beetles (Fig. 12). These elements are (1) the convoluted tendon of the metafemoral extensor of the tibia termed ‘Maulik's organ’ (cf. Mo in Figs 12A and B), (2) the sclerite (‘Lever's triangular plate’) connected to the tibial flexor apodeme (cf. Lt in Figs 12A, C–D, F–G) and the connection of this sclerite to (3) the distal internal projection at the posterior femoral wall (cf. fab and ri in Fig. 12E–G), which might function as a locking device for ‘Lever's triangular plate’ during the contraction of the tibial extensor muscles. Moreover, we were able to reconstruct (4) parts of the actual articulation between both the femur and the tibia (Figs 12B and E).
Metafemoral extensor tibiae tendon (‘Maulik's organ’) (Figs 12A and B): The following description uses the terminology of Furth (1980, 1982). The large tendon (‘Maulik's organ’) traverses almost half the longitudinal axis of the femur. In the transverse section, its proximal half is shaped like an ‘S’, i.e. it forms two lobes (a dorsal and a ventral lobe) separated by two furrows (the central and ventral furrow according to Furth, 1982). The convoluted shape of the tendon facilitates large deformations and results in the ability to store large amounts of strain energy during the contraction of the attached muscles, whereas the tibia is retained in its flexed position prior to the actual jump. The dorsal lobe extends distally, narrowing down into an ‘extended arm’ (cf. ea in Fig. 12A) and finally merging into an elongated ligament that forms the actual insertion of the tibial extensor system with the tibia. The point of insertion is situated close to the base of the tibia, i.e. proximal of the femoral-tibial articulation (cf. Figs 11, 12A and C).
Catch-like sclerite connected to the metafemoral flexor tibiae tendon (‘Lever's triangular plate’) (cf. Lt in Fig. 12A, C–D, F–G): The tendon of the flexor system of the tibia merges proximally into a distinct sclerotized plate termed ‘Lever's triangular plate’. This plate mediates the connection to the tibia via two short lateral ligaments (cf. li in Fig. 12D). The actual points of insertion of both these ligaments at the tibia are located slightly distal of the point of articulation between both the femur and the tibia (Fig. 12C). The plate itself has the shape of a triangle whose lateral margins are swollen (Fig. 12D). The base of its posterior side is modified into a series of protruding ridges (cf. ri in Fig. 12D), which closely fit into a corresponding set of clefts and ridges at the inner side of the distal margin of the posterior wall of the femur (Fig. 12E–G). The close interaction between both Lever's triangular plate and the femoral wall is facilitated by the posterior femoral wall being slightly curved inwards at this point, forming an abutment for ‘Lever's triangular plate’ (cf. fab in Figs 11A and B). The contraction of the flexor tibiae muscle together with the close interlocking of both these systems via transeversely running cuticular ridges probably constitute the claimed catch, which for the moment prevents the extension of the tibia, whereas the cocontracting extensor tibiae muscles build up the energy necessary for the jump. Probably, only upon relaxation of the tibial flexor muscles is the stored energy suddenly released, detaching Lever's triangular plate from its close engagement with the femoral wall and rapidly extending the tibia.
Articulation between both femur and tibia (Figs 12C and E): The axis of articulation between both the femur and the tibia lies perpendicular to the anterior–posterior axis, the latter being shown in Figs 11(A) and 12(A). The articulation is mediated by a dorsal and a ventral condylus at the distal margin of the femur, fitting into corresponding joint cavities at the proximal end of the tibia [cf. Figs 12(C) and (E) for both the condylus and the joint cavity of the dorsal point of articulation].
Comparison with conventional histological techniques
Synchrotron microtomography versus histology in biological objects
Initially, SR-μCT data might be less informative than colour-stained histological serial sections (cf. Fig. 13), because no differential staining protocols for the distinction of the various types of tissue are as yet available for this technique. On the other hand, compared with light microscopy, the inspection of virtual μCT serial sections is much easier, since several tools for semiautomated data analyses are available (e.g. VGStudio Max, amira™). Both VGStudio Max (Fig. 14) and amira™ simultaneously display different windows on screen. In VGStudio Max, one is a 3D window, whereas the other three windows are 2D sectional views of a selected plane in, for instance, a frontal, horizontal and sagittal orientation (Fig. 14). A special feature in VGStudio Max is the ‘Scene relative mode’, which allows the 3D object to be sliced along any arbitrary axis. The simultaneous display of all three orthogonal sectional planes, through which one can easily scroll forth and back, significantly facilitates orientation within an object under study. In addition, by using the so-called 3D cursor tool, a particular spot within a structure in one window can be marked and the actual position of this spot is then also shown in all the other 2D views (windows). This tool is helpful, for example, when muscle courses and their points of origin and insertion are investigated.
Table 2 contrasts both traditional histological thin sectioning versus synchrotron X-ray microtomography. The advantages and disadvantages of both the methods need to be considered when a decision regarding one or the other method is to be made. Often, a combination of both is recommended for optimal results (cf. Fig. 13).
Table 2. Gains and shortcomings of synchrotron X-ray microtomography versus conventional histological serial sectioning techniques.
Synchrotron X-ray microtomography
Histological serial sectioning techniques
Noninvasive, free of mechanical artefacts and complete sequence of virtual sections without loss of single slices
Risk of mechanical artefacts and loss of single slices
Differential coloured staining of tissues and chemical substances less straightforward than standard histology
Availability of a diversity of histological staining and histochemical and light microscopic techniques
Time-saving data acquisition, inspection and visualization
Time consuming data acquisition and inspection of serial sections and difficult orientation throughout the object; only one sectional plane available per sequence of sections
Easy preparation of unbiased self-illustrating 3D reconstructions of the entire object or parts of it
3D reconstructions hard to obtain by difficult manual labour
Depending on object size and selected field of view, spatial resolution down to the submicron level (cf. Table 1) but still inferior to that from light microscopy
Maximum spatial resolution about 0.2 μm
Free-of-charge beamtime for nonprofit scientific organizations available upon peer-reviewed proposals
Easy orientation within the three dimensions of the sample and opportunity to evaluate any desired sectional plane within one dataset
Opportunity to perform 3D morphometric analyses (e.g. geometric morphometry)
Possibility to combine 3D-renderings with element-sensitive absorption imaging
Availability of phase contrast for contrast enhancement
In conclusion, SR-μCT techniques are of great benefit for microanatomical investigations. They offer the possibility of attaining comparative datasets within a short time, significantly facilitate orientation within an object under study and allow 3D reconstructions to be produced that can be used for quantitative studies (Nickel et al., 2006a,b; Wegst et al., in preparation). Although the maintenance of Synchrotron beamlines is expensive, facilities like the ESRF provide shifts of beamtime each year upon peer-reviewed applications, which are free of charge for institutes of the member countries.
Although the resolution of μCT data comes close to that of the light microscope, traditional bright-field light microscopic techniques are still irreplaceable in histological studies relating to different types of tissue or cells and with respect to the abundance of staining methods (Romeis, 1989). Traditional histological section techniques remain useful also as a verification for μCT results and as an additional means of identification for difficult structures.
Effect of osmium ‘staining’
One histological technique that can be used to fix fatty acids and to stabilize unsaturated lipids is to cross-link them by means of osmium tetroxide, which binds to the C-C double bonds in tissues thereby forming osmium acid ester (Hayat, 1981). Since X-rays are strongly absorbed by heavy metals, one can use osmium tetroxide as a contrasting agent in μCT scans. This might be especially advantageous in soft tissue, such as nerves or muscles, which might otherwise show only weak absorption compared with denser material such as cuticle.
To test the effect of osmium ‘staining’, insect samples were fixed in 2.5% glutaraldehyde (overnight), washed in sodium-cacodylate-buffer, rinsed in 2% osmium tetroxide (16 h) as a secondary fixative, washed again in the buffer, stepwise dehydrated in a series of increasing isopropanol concentrations and finally critical-point dried (cf. above section on illustrative examples). On tuning the beam energy close to the absorption peak of osmium, viz. 13 keV (although higher energies yielded similar effects), regions in our samples with higher concentrations of osmium appeared brighter than others (cf. Fig. 15) and thus facilitated the recognition and identification of soft tissue, which would otherwise have appeared darker because of its lower absorption.
The applicability of other TEM fixatives and combinations thereof as staining agents in SR-μCT remains to be tested.
Software packages and hardware recommendations
Once the SR-μCT voxel data are obtained, a variety of scientific questions can be addressed by using various computer programs. The major objective may be to produce volume renderings, which can be segmented to reveal individual structures. Numerous software packages are available for these purposes. Here, we briefly introduce two widespread commercial products, which we used in our illustrative examples: VGStudio Max (Volume Graphics, Heidelberg, Germany) and amira™ (Mercury Computer Systems, Inc., Berlin, Germany). Both are available for 64-bit computer systems and can handle both voxel datasets and image stacks.
VGStudio Max deals with the representation of both object surfaces and interiors. Thereby, 3D renderings of the voxel data can provide information on the actual interior structure without reduction to triangular surfaces. The 3D rendering is calculated in real time with appropriate hardware configurations (see below). Additionally, quantitative measurements can be made with this software. The spatial distribution of any interior material can be visualized quickly and efficiently and the visual appearance can be improved by the available light and shadow settings. Although VGStudio Max is also capable of surface modelling and segmentation, we suggest the use of amira™ for these kinds of analyses (Figs 8 and 12). amira™ offers user-friendly tools to represent 3D objects as both triangular surfaces and volumetric tetrahedral grids, suitable for numerical simulations.
The minimum hardware recommendations should be obtained from the software manufacturers for the most recent versions upon need. According to our experience with both these programs and synchrotron voxel data, we suggest the following minimum hardware-configuration. A multiprocessor system, which addresses enough memory (e.g. x64 system), is highly recommended. High processor speeds are important for the analysis of large datasets; however, the main memory and the graphics board are also crucial components for 3D renderings. The graphics board must provide hardware-accelerated OpenGL 3D graphics and hardware texture mapping and should have enough memory and multimonitor support. With respect to the main memory of the system, a rule of thumb is ‘the more the better’. Main memory of more than 4 GB cannot be addressed by 32-bit computer systems without using physical address extension and, even then, the main memory available for a single task remains restricted to 4 GB; hence, we recommend 64-bit computer systems, which have no such restrictions. The computer should have large and fast hard-drives or, even better, a RAID 0 configuration with two fast S-ATA drives to allow fast access to the data. The hardware configuration given here represents a ‘minimum’ configuration. Since the development of faster processors and the availability of main memory will increase in the future, the optimal configuration should be discussed with the software company and the SR-μCT dataset providing facility.
Both the software and the hardware recommendations reflect our personal experience thus far. However, a variety of combinations of other software and hardware systems may perform just as well as the systems described here. For example, ImageJ (Abramoff et al., 2004) is a public domain Java-based image processing program developed at the National Institutes of Health (U.S.A.). It is widely used by the life science community and allows to visualize the reconstructed 2D cross-sections as a stack. It also has the advantage of having many plug-ins available for image analysis.
The present contribution is the first one of a two-part publication on the application of synchrotron X-ray microtomography (SR-μCT) in biological morphology and biomaterial science (U. G. K. Wegst et al., in preparation). Its goal is to introduce this novel method in biological morphology to a broader community of biologists. SR-μCT has the potential to revolutionize biological microanatomy and to open new research opportunities for the biomaterial scientist. Classical anatomical techniques are costly in terms of time and effort and usually involve the destruction of the object under study. They are often associated with technical difficulties, especially when dealing with small organisms in the range of a few millimetres. The reconstruction of the 3D appearance of anatomical structures from dissections or sectional series requires particular scientific drawing skills and is always a matter of the author's interpretation. In contrast, SR-μCT is nondestructive and provides the investigator with complete sets of perfectly aligned virtual sections throughout the object at submicron resolution. The data can be uploaded on a customary desktop computer, viewed at any sectional plane and used to produce self-illustrating 3D renderings that are unbiased with respect to the drawing skills and opinion of the investigator. Together with the rapid acquisition of raw data, SR-μCT rivals classical methods (though without replacing them) and must be considered a methodological milestone in microanatomical research. Another aspect of the outstanding scientific potential of SR-μCT is that it can be employed to obtain comparative anatomical datasets of different species or developmental stages within a short time frame. This makes it an important tool in comparative disciplines, such as evolutionary, developmental biology, taxonomy, systematics and the environmental sciences. The possibility of obtaining nondestructive views inside fossils also makes it an interesting technique for paleontologists.
The fields of view available at synchrotron facilities are currently limited by the dimensions of the accessible CCD chips. In future, new developments in the digital camera sector might enable larger fields of view without compromising resolution. For tomographic imaging of regions of interests in flat but laterally extended specimens, modified scanning schemes such as computed laminography with synchrotron radiation (Helfen et al., 2005) might be beneficial, since the requirements concerning the field of view of the CCD have only to be accounted for in one specimen dimension (i.e. the through-plane dimension of flat objects). In general, for all tomographic techniques, future developments in hardware and software are expected greatly to improve the handling of large datasets. Volume rendering of SR-μCT data might also become an important branch in the production of electronic teaching material for the life sciences, similar to the use of CT data in medical education. The potential of quantitative phase-contrast tomography for the visualization and reconstruction of biological microstructure (especially that of soft tissue) remains to be further explored (cf. Cloetens et al., 2006).
We thank Sebastian Schmelzle and Isabel Koerner for their segmentation work on Figs 8 and 12, Dr. Heike Betz for the creation of Fig. 11(A) and Dr. Steffen Orso, Dr. Muriel Veron, Prof. Michel Schlenker and Christian Schreiber for experimental assistance. Dr. Jan Michels provided helpful information on the CLSM technique. This paper was written during Ulrike Wegst's stay at the Lawrence Berkeley National Laboratory, whose hospitality and support are gratefully acknowledged. Our beam time sessions at the ESRF were funded by the European Union. Oliver Betz received funding from the Deutsche Forschungsgemeinschaft. Dr. Theresa Jones corrected the English of the manuscript. We are grateful to three anonymous reviewers for providing useful comments on previous versions of our manuscript.