Interfacial relationships between biomaterials and tissues strongly influence the success of implant materials and their long-term functionality. Owing to the inhomogeneity of biological tissues at an interface, in particular bone tissue, two-dimensional images often lack detail on the interfacial morphological complexity. Furthermore, the increasing use of nanotechnology in the design and production of biomaterials demands characterization techniques on a similar length scale. Electron tomography (ET) can meet these challenges by enabling high-resolution three-dimensional imaging of biomaterial interfaces. In this article, we review the fundamentals of ET and highlight its recent applications in probing the three-dimensional structure of bioceramics and their interfaces, with particular focus on the hydroxyapatite–bone interface, titanium dioxide–bone interface and a mesoporous titania coating for controlled drug release.
Implant materials are frequently employed in reconstructive surgery, both for temporary fixation during healing or for permanent anchorage to restore lost function. In the case of load-bearing applications, such as dental implants , hip arthroplasty  and more recently, major limb amputation prosthesis , titanium and its alloys are most frequently used owing to their excellent mechanical and corrosion-resistant properties and the biocompatibility offered by their titanium dioxide surface layer. Other frequently used materials are pure ceramics, including a wide range of calcium phosphates, mostly employed in non-load-bearing applications or purely compressive or articulate situations.
One of the key factors determining long-term implant success is firm anchorage to the surrounding bone tissue. Understanding and improving the adaptation of bone tissue to the implant material is of utmost importance. It is widely known that many surface properties are of importance for the biological response of a material, such as surface topography, surface chemistry and surface charge [4,5]. With the present implementation of nanobiotechnology in implant design, the possibility to control specific nano-features on the implant surface places a higher demand on the resolution level for interfacial analysis. The transition from light to electron microscopic evaluation of biological interfaces took place many years ago, and with it, increased resolution capabilities. However, important structural interfacial information in three dimensions is still missing. Recent development in transmission electron microscopy now enables viewing information in a third dimension using automated tomography techniques.
In this review, the three-dimensional analysis of bioceramics, namely titanium dioxide and hydroxyapatite (HA), are addressed with electron tomography (ET). Visualizing interfaces with nanometre precision in three dimensions exposes new fundamental information on material behaviour and interaction. Moreover, the extension of ET to other biological interfaces and materials of every nature is indeed possible.
2. Tomographic techniques
Tomography, derived from the Greek words tomos (to section) and graphe (to write), is the most common three-dimensional analysis method of biomaterials interfaces. Tomograms (sections or slices through the thickness of a sample) are generally created by the reconstruction of data obtained from two-dimensional projections of the object. The main principles of tomography include acquisition, reconstruction and visualization, and are illustrated in figure 1.
Tomographic reconstruction is possible from a number of media, including photons, X-rays, electrons and atom probes. These techniques fall into one of two categories: non-destructive or destructive . If we exclude some form of destructiveness required during transmission electron microscopy (TEM) sample preparation from bulk biomaterial–tissue interfaces, then we can consider purely projection-based techniques such as electron and X-ray tomography as non-destructive forms, in such a way that they obtain projections without destroying the sample simultaneously. Of course, destruction in the form of beam damage is always a concern with tissue samples. However, assuming a sufficiently low beam dosage, sample integrity is maintained. Destructive forms of tomography include the focused ion beam (FIB) method, where sections are literally milled away to create the surfaces imaged for three-dimensional information. Moreover, if we assume that high-resolution refers to the ability to resolve features on the nanometre scale, then, in fact, only electron and atom probe tomography meet this constraint. Even synchrotron source X-ray computed tomography, regularly employed in the characterization of biomaterials and their tissue interfaces, generally provides resolution in the micrometre range [7,8], albeit recent advances and specialized instrumentation have pushed the resolution into the nanometre regime . Perhaps the highest resolution tomographic technique, atom probe tomography, enables atomic resolution in three dimensions. However, the restrictions on sample size (limited to a few nanometres) and type (highly conductive surfaces) have constrained the wide-scale use of this technique to applications in materials science . Herein, we focus on the characterization of biomaterials and their interface to bone with ET.
(a) Electron tomography
ET had its beginnings in the biological sciences. The first three-dimensional reconstruction performed by DeRosier and Klug in 1968 used a single projection of bacteriophages, oriented in a sufficient number of angles owing to their symmetry, to reproduce the three-dimensional structure . Around the same time, Hart, forming the foundation for modern day electron tomographic tilt-series, introduced his concept of a ‘polytrophic montage’ . Today, over 40 years later, we see an abundance of ET techniques, further reviewed in other works that are adaptable to almost any application in the materials and life sciences [6,10,13]. Bridging these two fields, we find the area of biomaterials science relatively unexplored with ET, but, nevertheless, an area that offers opportunities for its utilization.
As previously mentioned, the basic concept of ET is the reconstruction of a three-dimensional object from acquired two-dimensional projections over an angular range – a tilt series. The transmission electron microscope has a number of operating modes, and most of these can be used for the collection of tomographic tilt series. In standard TEM, bright-field (BF) images can make up the tilt series; these types of images reveal shape-sensitive information in three dimensions . While BF TEM tomography is ideally suited for purely biological specimens and samples containing polymer-based biomaterial interfaces , the introduction of a crystalline material (such as some biomaterials) results in diffraction contrast and Fresnel fringes that severely degrade the accuracy of the reconstruction . Another form of microscopy, scanning transmission electron microscopy (STEM), which uses a focused probe rastering the sample, may alleviate these problems. In particular, high-angle annular dark-field (HAADF) STEM tomography presents a variety of advantages for the study of biomaterial interfaces. Collecting the electrons that have been incoherently scattered to high angles, the HAADF detector suppresses diffraction contrast while enhancing the Z-contrast in images. The result is a three-dimensional reconstruction highly dependent on the atomic number of the constituents in the sample. This high-contrast imaging mode has been applied to the study of some biomaterials, for example in understanding the distribution of calcite crystals in an agarose gel matrix . In another example by Okuda et al.  on HA-coated magnetite nanoparticles for cancer treatment, Z-contrast ET enabled the differentiation between the magnetite core and plate-like HA surrounding. STEM ET presents other advantages such as the lack of image-forming lenses after the specimen, which reduces resolution losses . Furthermore, the scanning electron probe may be focused anywhere on the sample, even at high tilt angles, which will generally have areas out of focus with conventional TEM imaging .
Also available in STEM mode, axial BF imaging, as demonstrated by Sousa et al.  in the investigation of pancreatic beta cells, has the ability to extend the thickness of investigated biological specimens up to 1 μm. Similar outcomes were noted by Hohmann-Marriott et al.  when they used axial BF-STEM tomography to study eukaryotic cells. This study also made use of dual-axis ET to improve resolution of the reconstructed data. By collecting data from two perpendicular tilt axes, the unsampled area referred to as the missing wedge is decreased into a missing pyramid (figure 2), improving the accuracy of the reconstructed data [13,17,19].
For biological and beam-sensitive samples, a cryogenic sample holder that maintains the specimen at liquid nitrogen temperatures throughout acquisition is routinely used to reduce beam damage [20–22]. Recently, cryo-ET has played an important role in uncovering mechanisms of mineralization in in vitro and in vivo examples. In the work by Dey et al. , cryo-ET was used to determine the formation of amorphous calcium phosphate pre-nucleation clusters, prior to formation of carbonate HA in an in vitro simulated body fluid solution. Nudelman et al.  also harnessed cryo-ET to locate the site of amorphous calcium phosphate infiltration into collagen and its subsequent orientation into mineralized HA. Other alternatives, such as high-voltage electron microscopy offer the possibility to analyse thicker specimens with high resolution, as used by Landis in the study of the mineral and collagen arrangement in the turkey tendon [25–27]. As long as the imaging technique meets the projection requirement, as described by Hawkes, that the projections are a monotonic function of a physical property, acquiring the tilt series in any TEM mode is, in theory, feasible .
The foundations for three-dimensional image reconstruction have been laid out in detail elsewhere, and as such will only be mentioned here [11,29–32]. The main methods employed are weighted-back projection  and a simultaneous iterative reconstruction technique , in which two-dimensional projections are smeared onto a three-dimensional volume, and two-dimensional projections are re-projected iteratively to create a difference tomogram, respectively. The maximum attainable resolution in a three-dimensional reconstruction was laid out in the groundwork paper by Crowther et al. . In his criterion, Crowther approximates the resolution to be proportional to the diameter of the object in view, and inversely proportional to the number of projections acquired of that object . Therefore, improved resolution depends not only on using an instrument with an overall high resolution, but it is also dependent on the size of the features to be reproduced and decreasing the angular increment between projections .
Techniques exist to decrease artefacts in reconstruction, covered in more detail in some of the key reviews on this subject [6,32]. As mentioned previously, dual-axis collection may improve the accuracy of reconstructed data. In every acquisition scenario, post-acquisition aligning of all images along the correct tilt axis is a main concern. Sample shift due to goniometer instabilities and specimen drift, among others, all contribute to shifts in the images and apparent tilt axis. To overcome this problem, fudicial markers, generally gold nanoparticles, may be added to the sample to ease in recognition and tracking of a distinct alignment feature throughout all images [34,35]. In the case of the biomaterials–tissue interfaces presented later in this paper, we were able to align images in the whole stack with no fudicial markers by the use of cross-correlation methods. The sharp interface between tissue and biomaterial provided a distinct feature for tracking throughout all tilt angles.
While ET has had roots in the biological and materials science fields dating back over 40 years, the interdisciplinary domain of biomaterials, in particular biomaterials in contact with biological matter, has not seen its use until very recently .
3. Structure of biomaterials with Z-contrast electron tomography
Z-contrast ET (tomographic tilt series collected using the HAADF detector in STEM mode) has been used in our research for studying the interface between implant materials and human tissue. Three-dimensional structural information at the interface of biomaterials with bone may aid in a deeper understanding of biocompatibility and bone-bonding mechanisms. In the following cases, we present examples of a bioceramic and metallic implant interfacing to human bone, as well as a preliminary model for a functionalized drug-deliverable coating. All TEM samples have been prepared using an FIB instrument with an in situ lift-out technique [37–39] to maintain the integrity between implant and tissue, or implant and coating.
In these works, single-axis Z-contrast ET was performed using an FEI Titan 80–300 transmission electron microscope (FEI Company, The Netherlands) equipped with a Schottky field-emission gun and a CEOS (Corrected Electron Optical Systems GmbH, Heidelberg, Germany) hexapole-based aberration corrector for the image-forming lens. The microscope was operated at an acceleration voltage of 300 kV. Images were recorded on a model 3000 in-column HAADF detector (Fischione Instruments, PA, USA) with an inner semi-angle of 40 mrad. Automated focusing, image shifting and acquisition of HAADF STEM images over an angular range of ±75° (for the biomaterial–tissue interfaces) or ±60° (for the mesostructured coating) were achieved using Inspect3D software (FEI Company, The Netherlands). The Advanced Tomography Holder model 2020 (Fischione Instruments, PA, USA) was used in a linear tilt scheme, with image acquisition increments of 2° up to angles of ±60°, and 1° for further angles up to ±75°. However, for the mesostructured coating for drug delivery, a linear tilt scheme with image acquisition every 1° up to angles of ±60° was used. The three-dimensional reconstructions were computed using a simultaneous iterative reconstruction technique, with 15–20 iterations, in Inspect3D. Models for three-dimensional visualization were created in Amira Resolve RT FEI (Visage Imaging Inc., USA).
(a) Hydroxyapatite–bone interface
HA is one of the most employed biomaterials on the market today owing to its composition being similar to that of human bone. Bone itself is considered an organic/inorganic composite material, formed of type I collagen and HA crystals arranged in a hierarchical structure. HA implantables, common in the form of granules, blocks, scaffolds and coatings, display a high degree of biocompatibility in vivo. That is, they ‘perform their desired function, without eliciting undesirable local or systemic effects’ . Moreover, HA is considered to be an osteoconductive material, in that it conducts and directs the growth of bone along its surface . Composition, however, is not the only factor in determining the biocompatibility of HA with bone tissue. Implant design, such as pore morphology, fraction and size, largely influences the affinity for bone growth onto and into scaffolds; therefore, creating a three-dimensional environment similar to the physiological situation is critical [42–44]. Surface topography and chemistry also play an important role [5,45]. In addition, crystallinity has been shown to affect bone growth . To further complicate the situation, at least 11 biologically or chemically driven processes have been identified to occur at the interface of HA in vivo . In such a complicated model as that of humans, a complete understanding of the mechanisms that govern bone growth at implantation sites would benefit greatly from the study of the HA interface, with the additional information provided by a third dimension.
Prior to the advent of ET, conventional imaging and elemental analysis have been used to study the HA–bone interface with electron microscopy in great detail. Numerous studies report the formation of an apatite layer between bone and implant, up to 1000 μm in thickness [36,47–49], commonly accepted to occur through a dissolution–reprecipitation process . Composition and orientation of this layer is debated to date [50–55].
In our recent work, we employed ET for the study of the first biomaterial–bone interface in nanoscale three-dimensional resolution . Using Z-contrast ET, we obtained a three-dimensional structure that clearly indicated the orientation of HA crystals in the bone matrix and at the implant surface. This information, not visible from the two-dimensional projections, gives us insight into bone formation mechanisms at an interface. Figure 3 outlines the findings in this work.
ET established the alignment of HA crystals precipitated on the surface of the HA scaffold implant as perpendicular to the scaffold surface. It also confirmed that the orientation of HA crystals in the bone matrix is parallel to the implant surface, which is consistent with the collagen-banding pattern being visible perpendicular to the surface in STEM projections. Full videos of the tilt series and three-dimensional reconstruction are available in our previous work .
(b) Titanium–bone interface
Titanium and its alloys comprise another highly investigated implant material for applications in load-bearing implants, owing to their high mechanical strength, superior biocompatibility, corrosion and wear resistance, and ability to osseointegrate . The biocompatibility of titanium is usually attributed to its native surface oxide, titanium dioxide. Bone has an affinity to bond to and grow from this surface, and as such, it is also considered an osteoconductive material. This oxide layer can further be modified to enhance bone bonding by altering its surface topography. Both micro and nanostructured surfaces have been shown to improve bone growth and cellular activity on the surface of titanium [58–60]. In work with our collaborators, laser modification of a titanium screw (Biohelix, Brånemark Integration AB, Sweden) has resulted in a titanium dioxide surface with a unique nanotopography (figure 4a) . Similar to HA, titanium dioxide induces the precipitation of an HA layer on its surface to facilitate bonding to bone. With ET, visualization of the complete intermixing and encapsulation of laser-modified TiO2 and HA with one another is achieved, figure 4c,d. From this model, we truly have evidence of osseointegration between implant and bone occurring on the nanometre scale. The complete three-dimensional reconstructions are available in the electronic supplementary material.
(c) Titania mesostructures for interfacial drug delivery
As mentioned already, the biocompatibility of titanium is attributed to its surface oxide layer—titanium dioxide. Clearly, there has been much research in the field of tailoring titanium dioxide coatings to improve bone integration by means of mechanical methods, chemical (alkaline, acidic, sol–gel, hydrogen peroxide) treatments, physical methods (plasma spraying, evaporation, sputtering) and heat treatment , among others . In addition to a desirable chemical composition, a suitable coating morphology may increase bone growth while increasing functionality. By functionality, we refer to the ability to accomplish more than one objective, for example, improve bone growth while reducing inflammation and infection, etc. In the case of titania, incorporating porosity into the coating assists in creating functionality in two ways: one, by providing a nano-topographical surface that improves bone attachment, and two, by acting as a site-specific carrier for active macromolecules or drug agents.
In our recent work, the morphology of mesoporous titanium dioxide coatings, produced on titanium by an evaporation induced self-assembly method , has been studied with ET. This method has enabled a deeper understanding of the pore size, volume and interconnectivity. The open three-dimensional porous network is clearly visualized, providing the prospect to load the coating with drugs for site-specific release in vivo. A sustained drug delivery may be achieved if drug molecules can penetrate through the porous network, rather than solely adsorb on the surface. Site-specific drug release offers the advantage of lower dosage, and the avoidance of a systemic antibiotic regimen. A wide range of drug types, in addition to antibiotics, could potentially be loaded into the pores, such as bisphosphonates, generally used in the treatment of osteoporosis, which suppress osteoclastic activity and enhance osteoblastic activity responsible for laying down new bone [64–66]. The suggested drug loading and release mechanisms via capillary force and diffusion are illustrated in figure 5, along with the three-dimensional reconstruction. This three-dimensional model provides the basis for further investigations and simulations of drug delivery and release mechanisms. A complete video of the tilt series and reconstruction is provided in the electronic supplementary material.
In the quest for controlled interfacial interactions, such as drug release, ET becomes an important tool. It provides a wealth of information for the properties associated with the three-dimensional structure of biomaterials, providing values for surface area, pore size and geometry. Once extracted, these values enable, for example, simulations to determine drug flow and release profiles, and bone ingrowth, providing a better understanding of the mechanisms of local drug delivery and osseointegration.
4. Future perspectives
The use of high-resolution three-dimensional techniques to probe the structure of biomaterial interfaces at the nanometre scale is vital to the advancement of biomaterials development. ET, traditionally heavily used for the investigation of biological structures, can be adapted for the study of biomaterials and their interfaces to tissue, as we have demonstrated with the use of Z-contrast ET.
Nevertheless, a shift from acquisition of three- to four-dimensional information should still be considered. It can be debated whether the fourth dimension is that of time-resolved, or spectroscopic data. In any case, the addition of either of these dimensions supplements our knowledge of biomaterials as seen to date in only three dimensions. Already a dominant technique in the materials science and biological fields, ultrafast electron microscopy as pioneered by Zewail enables both high spatial and temporal resolution . By sending discrete single-electron packets, spaced merely femtoseconds apart, a temporal resolution exceeding that of normal TEM systems by 10 orders of magnitude is achieved, enabling imaging of dynamic processes . Furthermore, the combination of this discretized imaging with angular tilt results in the collection of a time-resolved tilt series. Coupling this technique with an in situ wet cell could enable visualization of in vitro processes such as adsorption of molecules, and dissolution/reprecipitation mechanisms on biomaterial surfaces. The four-dimensional tomograms of multi-walled carbon nanotubes , among other examples [68,70], and initial steps towards complete four-dimensional tomography of biological cells , lay the foundation for application to biomaterials in the future. Alternatively, spectroscopic data can be regarded as the fourth dimension of information and is regularly paired with ET in the materials sciences in the technique of energy-filtered ET [10,72]. In the biological sciences, Leapman et al.  have demonstrated the ability to locate and quantify phosphorus in three dimensions by acquisition of energy-filtered transmission electron microscopy (EFTEM) tomography of a nematode organism.
While a valuable tool on its own, ET will not reach its full potential when used in isolation. By combining ET with the plethora of characterization techniques in the transmission electron microscope's arsenal, we can probe not only the structure of biomaterials in three dimensions, but also, the chemical composition. In addition, advancements in aberration-corrected electron microscopy may enable three-dimensional atomic resolution lattice imaging of biomaterials in the near future.
Without a doubt, forthcoming developments in the field of ET will simplify our ability to obtain three-dimensional structural, chemical and time-resolved data on biomaterials interfaces. Access to this information will thereby increase our understanding of mechanisms governing, for example, bone growth, functionalized drug release and interactions of all types, bridging the gap between synthetic materials production and biological human tissues.
The members of the Institute for Biomaterials and Cell Therapy (IBCT), Gothenburg, Sweden and the Canadian Centre for Electron Microscopy, Hamilton, Ontario, Canada are gratefully acknowledged for stimulating discussions, provision of samples or assistance with data acquisition. Financial support has been provided by the VINNOVA VinnVäxt Programme of Sweden, IBCT and the Natural Sciences and Engineering Research Council of Canada.
One contribution of 11 to a Theme Issue ‘Structure and biological activity of glasses and ceramics’.
- This journal is © 2012 The Royal Society