Demonstration of lanthanum in liver cells by energy-dispersive X-ray spectroscopy, electron energy loss spectroscopy and high-resolution transmission electron microscopy


D. Schryvers. Tel: +32 3 2653247; fax: +32 3 2653257; e-mail:


The appearance of lanthanum in liver cells as a result of the injection of lanthanum chloride into rats is investigated by advanced transmission electron microscopy techniques, including electron energy loss spectroscopy and high-resolution transmission electron microscopy. It is demonstrated that the lysosomes contain large amounts of lanthanum appearing in a granular form with particle dimensions between 5 and 25 nm, whereas no lanthanum could be detected in other surrounding cellular components.


In patients with chronic renal failure, high phosphate levels develop in the blood as a consequence of insufficient renal phosphate excretion. The concomitant slight decrease of ionized serum calcium and decrease of vitamin D synthesis by the sick kidney results in an overactive parathyroid gland, and hence parathyroid hormone secretion. These changes, in turn, may result in particular types of bone disease and favour the formation of vascular calcifications, and hence cardiovascular mortality (Llach, 1998; Block et al., 1998).

In order to control serum phosphate, patients with chronic renal failure have to take phosphate-binding agents. Although very effective, the use of high doses of aluminium-containing phosphate binders causes severe side-effects at the level of the bone and the brain, and the currently used calcium-containing compounds have been associated with hypercalcaemia and vascular calcifications (Block et al., 1998; Goodman, 1985). Consequently, there is a need for safe, well-tolerated alternatives. In the search for new agents, lanthanum carbonate has been shown to be an effective phosphate binder in patients with chronic kidney disease (Joy et al., 2003; D’Haese et al., 2003). As demonstrated in preclinical and clinical studies, lanthanum (La) is poorly absorbed by the gastrointestinal tract, and low-level deposition mainly occurs in bone and liver (Behets et al., 2004). Recent clinical and ultrastructural localization data of La in bone showed the element to be present at both active and quiescent sites of bone mineralization, independent of the type of bone disease, and to be not associated with aluminium-like alteration (Behets et al., 2005).

As demonstrated in rats after intravenous injection of La, hepatobiliary excretion is the main elimination route (Damment & Gill, 2003). A recent study indicated in addition that, as opposed to the situation in other organs, liver La concentrations of renal failure rats treated with oral La were significantly higher than those seen in rats with normal renal function that had received the same doses (Slatopolsky et al., 2005).

In view of this, it is obvious that insight into the subcellular localization of the element in the liver is an essential first step in studying the potential hepatotoxicity of La. In the present work, the presence and localization of La in the liver was demonstrated by spectroscopy and imaging in different transmission electron microscopes (TEMs). Conventional imaging was applied to recognize dark precipitates in particular cellular structures, and energy dispersive X-ray (EDX) and electron energy loss spectroscopy (EELS) analysis were used to determine the occurrence of La. The first spectroscopy technique allows for quick determination of the La, and the second technique is more elaborate but yields a better lateral as well as energy resolution. The nanoscale morphology of the La-containing precipitate was investigated by high-resolution TEM.

Materials and methods

La was administered by daily intravenous injection of lanthanum chloride into rats with normal renal function at a 0.3 mg kg−1 dose over 4 weeks. This resulted in total liver La concentrations varying between 30 and 50 µg g−1.

Liver fragments of treated as well as untreated animals were fixed in 4% formaldehyde in phosphate buffer and postfixed in reduced OsO4. Formaldehyde was preferred over a more robust fixation with glutaraldehyde, because the same block was also studied by light microscopy, and also by periodic acid–Schiff staining, which is incompatible with glutaraldehyde fixation. After dehydration and embedding in Epon, 100-nm and 500-nm sections were made and used with or without counterstaining with uranyl acetate and lead citrate. Although the 500-nm sections show considerable detail at low magnifications in a TEM and ensure the observation of many features in a single section, the results presented in the present article were all obtained using 100-nm sections, stained or unstained.

Conventional imaging was performed in an FEI Technai 100 (at 80 kV) equipped with a tungsten filament, at magnifications between 1800 and 85 000. For the EDX analytical work, an atmospheric thin window Oxford QX2000 instrument with an energy resolution of 138 eV was used, attached to a Philips CM20 TEM instrument equipped with an LaB6 single-crystal filament and operating at 200 kV, 120 kV or 80 kV. The EDX spectra were acquired with an energy dispersion of 20 eV channel−1. For the EELS, a postcolumn Gatan Imaging Filter (GIF2000) instrument with an energy resolution of 0.7 eV and a detection limit of 4 p.p.m. was used, attached to an Ultratwin Philips CM30 TEM instrument equipped with a Shottky field emission gun and operating at 300 kV. For the atomic resolution imaging, a 400-kV high-resolution LaB6 JEOL 4000EX TEM with a Scherzer resolution of 0.17 nm and a top-entry sample holder for mechanical stability at high magnifications was used.

As regular sections did not withstand the local heating and charging by the electron bombardment in the higher-voltage instruments, an amorphous carbon coating with a thickness of approximately 10–20 nm was applied onto the samples by thermal evaporation after deposition of the sections on the copper grids. This procedure successfully avoided immediate damage by the intense electron beams.


Conventional TEM

Unstained liver sections of La-treated animals (n = 2) showed precipitates displaying a granular-like substructure as shown in Fig. 1, and an electron density comparable to that of the cerium reaction product, but lower than that of the lead phosphate reaction product; see, for example, Figure 4 in Roels et al. (1993). La salt used as an extracellular tracer in vitro also produces a weaker density than lead (Friend & Gilula, 1972; Rassat et al., 1982). The ultrastructural detail obtained after formaldehyde fixation is due to the postfixation in reduced osmium tetroxide (Roels et al., 1995).

Figure 1.

Unstained liver sections of La-treated animals obtained with the FEI Technai 100 instrument and showing electron-dense precipitates displaying a granular substructure in a lysosome at ‘B’ (‘A’ is an example of a dense body without precipitates, and ‘M’ is a mitochondrion).

Precipitates were predominant in parenchymal cells in dense bodies preferentially localized near the bile canaliculi, and known to be lysosomes. The amount of granular material varied strongly between individual organelles; some dense body profiles seemed to contain none (e.g. ‘A’ in Fig. 1), whereas others were entirely covered with dense granules (e.g. ‘B’ in Fig. 1). In addition, precipitates were observed inside a minority of bile canaliculi, suggesting that lysosomes excrete their contents into the bile. The observed sizes for the lysosomes imply that in most cases the lysosomes will run completely through the 500-nm sections and certainly through the 100-nm ones. A Mann–Whitney nonparametrical test further indicated that precipitate-rich lysosomes are larger than the precipitate-free ones (D’Haese et al., 2006).

In macrophages (Kupffer cells) of La-treated animals, very little, if any, granular precipitates were seen, although lysosomes were large, as expected in normal liver. Granular precipitates were not observed over other organelles (nuclei, mitochondria, endoplasmic reticulum, peroxisomes). Glycogen rosettes in the cytosol had lower density than the La precipitates.

In untreated control livers, parenchymal lysosomes were well recognized as dense bodies (i.e. similar to ‘A’ in Fig. 1), but they did not contain dense granular precipitates.

An example of the granular structure as observed in lead/uranium-stained samples is shown in Fig. 2, from which the location of the lysosomes with respect to two cell nuclei is also clear. Although the overall contrast has increased due to the counterstaining, the granules in the lysosomes are still well recognizable.

Figure 2.

View of lysosomes in a stained section obtained with the 200-kV CM20 instrument and revealing the relative location of the precipitate-rich lysosomes with respect to other cell organelles (N, nuclei). Most profiles seen in the low-magnification photograph are mitochondria.

Recognizing the subcellular structures in unstained sections, especially those without any La content, proved very hard in the high-voltage instruments. Therefore, for all of the EDX, EELS and high-resolution TEM experiments discussed in the following chapter, lead- and uranium-stained sections were used as in Fig. 2.


From the EDX spectra shown in Fig. 3(a) and obtained from the lysosomes containing dark granular matter, it can be concluded that the La Lα peak at 4.65 keV and the Lβ peak at 5.04 keV are faintly but clearly observed in between the peaks from Fe and the omnipresent osmium from the fixation and lead and uranium from the counterstaining, which do not interfere with the La peaks. Iron storage in hepatic lysosomes is not unexpected, as known from human liver; it is a breakdown product of, among others, haem proteins. This spectrum was obtained with a focused electron beam with a diameter of around 0.2 µm, which implies that in principle it is indeed originating from the material within the respective lysosome. In Fig. 3(b), a spectrum from a precipitate-free lysosome (e.g. region ‘A’ in Fig. 1) is shown, lacking the La peaks but still revealing a weak iron peak. Spectra collected in cell regions sufficiently far away from those lysosomes in order to avoid interference did not reveal any La or iron, as shown in Fig. 3(c). All spectra have been scaled to have the same amount of total counts over the entire window and are presented with the same y-scale. The number of counts for the La Lα peak in precipitate-rich lysosomes is around 2000 for an acquisition time of 120 s. Comparing the three spectra in Fig. 3, it can further be concluded that P (phosphorus) with a Kα peak at 2.01 keV also primarily appears in lysosomes containing dark granular matter. Standardless quantification, in which the computer software performs a Gaussian fit for and calculates the area under the elemental peaks for the given atomic number Z (less accurate than the more elaborate alternative of working with standards but sufficient for the present purpose) and corrects for absorption (A) and fluorescence (F) of the X-rays for a given thickness in a so-called ZAF correction procedure (Heinrich & Newbury, 1991), yields an average of 20/80 for the La/P ratio in these lysosomes, with some probability of overestimating the P amount due to overlapping with the osmium peak.

Figure 3.

EDX spectra of (a) a precipitate-rich lysosome revealing La L and Fe K peaks, (b) a precipitate-free lysosome only showing a weak Fe peak, and (c) a region far away from any lysosome, lacking the La peak. The insets clearly show the essential differences between the three spectra. Pb and U peaks (counterstaining), Os (fixation) and Cu (grid) are obvious.

EELS and Energy Filtered TEM(EFTEM)

As no information on the internal distribution of La in the lysosomes could be obtained with the EDX, and as the La M5 and M4 edges are known to be very well detectable by EELS, the same samples were investigated with an electron energy-loss detector. These excitation edges are produced by inelastic collisions of the incident electrons with the 3d electrons of the M shell and have onsets at 832 and 849 eV, respectively. The fact that for the postcolumn GIF only the forward-scattered electrons are used also improves the lateral information in comparison to the EDX experiment. As expected, La was readily detected in the precipitate-rich lysosomes in 500-nm as well as 100-nm sections. As the 100-nm ones yield the best signal-to-noise ratio, all following quantification work was done on these samples.

In Fig. 4 (a) the relevant part of the EELS spectrum obtained by probing a 50-nm-diameter region of a precipitate-rich lysosome in a 100-nm section is shown (top line), together with that of a nearby region outside the lysosome (bottom line). As the coated sample is still radiation sensitive when an intense electron beam focused to the diameter indicated above is used, the EELS spectra are acquired in the imaging mode using an entrance aperture to select the respective region. The first spectrum clearly shows the fingerprint of La M5 and M4 edges on top of the decreasing background, whereas no edges can be observed in the second one. Lysosomes in untreated animals (negative controls) show a profile similar to the bottom line in Fig. 4(a), as expected.

Figure 4.

(a) EELS spectra revealing the La M5,4 edge inside the precipitate-rich lysosomes (top line) and no La nearby the lysosomes (bottom line). (b) The omnipresent C K edge. The dashed window in (a) shows the energy range for EELS quantification and EFTEM mapping.

The number of atoms NA per unit area for an element A in the material can be calculated from the EELS spectra by (Egerton, 1996, p. 280),


where IA(β, Δ) is the measured ionization edge intensity integrated within an energy range Δ and inside a collection semiangle β, Ilow(β, Δ) is the intensity of a window of equal β and Δ containing the zero loss, and σA(β, Δ) is the partial ionization cross-section. For the present measurement, a collection semiangle of about 17 mrad was used, and the Hartree–Slater model, available in the Digital Micrograph package, was used to calculate the atomic cross-section σ. The EELS spectra were treated following the standard route (Egerton, 1996, p. 245): the spectra are integrated over an energy window Δ = 25 eV wide, starting at the onset of the respective edge as indicated in Fig. 4(a), and taking into account the local background as a power law. The effect of multiple scattering was removed for each core-loss spectrum by deconvolution with the corresponding low-loss spectrum using the Fourier ratio method. Averaging over four precipitate-rich lysosomes, a La areal density of 40 ± 7 atoms nm−2, i.e. projected along the investigated column of material, is found.

Unfortunately, no good measure of the thickness of a particular lysosome can be obtained and thus, when looking for absolute densities, no thickness variation between the different lysosomes can be taken into account. However, the effects of thickness variation can be corrected for when using relative ratios, and for this reason the La content was compared with the omnipresent carbon, for which an example of the carbon K edge is shown in Fig. 4(b), clearly revealing the expected π* and σ* peaks. As a result, about five La atoms are found for each 100 carbon atoms in the central region of a precipitate-rich lysosome, whereas this number decreases towards two or three La atoms at a spot located towards one of the edges. In all other positions, i.e. outside the lysosome, the quantification yields a ratio below 10−5 or 10 p.p.m. With an instrumental detection limit of 4 p.p.m. under ideal experimental conditions (e.g. diffraction mode), it is safe to conclude that no La was detected outside the lysosomes.

As well as spot probe EELS, conventional three-window EFTEM can also be applied (Egerton, 1996, p. 330). An energy window of 25 eV was used when acquiring the images. The postedge window was positioned right at the threshold of the M5,4 edges of La, as shown in Fig. 4(a). Relative drift between successive images was corrected by a standard cross-correlation technique when computing the elemental intensity map. In Fig. 5, an example of an EFTEM image together with a two-dimensional trace from the edge of the granular material within a lysosome is shown, again indicating the gradual increase of La towards the centre of the lysosome. The averaged increase of La is spread over a distance of approximately 150 nm, leaving a central region of about 200 nm (for this particular lysosome) with a maximum La content.

Figure 5.

EFTEM obtained with the 300-kV CM30 instrument and revealing the gradual increase of La from the edge towards the centre of the lysosome: (a) zero loss image; (b) La EFTEM intensity map; and (c) two-dimensional averaged trace over rectangle in (b).

High-resolution TEM

Because of the strong contrast of the La granules in the lysosomes, the potential crystallographic nature was investigated using a dedicated atomic resolution instrument with a Scherzer lattice resolution of 0.17 nm. The 500-nm sections did not reveal sufficient resolution contrast, due to absorption and overlapping. In the 100-nm sections, however, atomic resolution images could be obtained. In Fig. 6, two examples of different nanoparticles revealing lattice resolution and obtained on the edge of a lysosome of a treated animal are shown. Although the overlapping of the particles with surrounding tissue does not permit direct interpretation of the actual crystal structure, the contrast still allows for proper measurement of lattice spacings [Van Dyck (2002) and references therein]. The sizes of the observed particles range from 5 to 25 nm, and they can thus be considered as the origin of the granular image obtained at lower magnifications, such as seen in Figs 1 and 2. The present examples show interplanar lattice spacings of 0.30 ± 0.02 nm and 0.55 ± 0.02 nm, respectively, both omnipresent distances when looking at the ensemble of images. Moreover, the alternating contrast of lattice fringes in the high-resolution TEM image of the particle in Fig. 6(b) indicates the existence of a superlattice structure formed by different atomic species. Surrounding these large nanoparticles, smaller but still crystalline nanoparticles with a maximum diameter of 3 nm and typical interplanar lattice spacings of 0.27 and 0.3 nm are found. These latter particles are also very abundant in regions outside the precipitate-rich lysosomes, as well as in stained but untreated samples, and are thus considered to be related to the lead staining.

Figure 6.

Examples of high-resolution TEM images of nanoparticles of approximately 12 nm (a) and 9 nm (b) in diameter and with interplanar lattice spacings of 0.3 nm (a) and 0.55 nm (b) in a granular lysosome obtained with the 4000EX instrument.


The occurrence of La in the lysosomes was readily demonstrated by both EDX and EELS spectroscopic methods. Moreover, both techniques also indicate that no La appears in other subcellular compartments. The lateral resolution of EDX is not sufficient to determine any structural morphology in the La area, but the nanoscopic EELS and EFTEM do yield more detailed information on the La distribution inside the lysosomes. In order to quantify the relative amount of La, the ratio with respect to the omnipresent carbon was determined, yielding maxima of five La atoms per 100 carbon atoms in the centre of the lysosomes over widths up to several hundred nanometers. However, when interpreting these numerical ratios, one should keep in mind the possibility of partial extraction of La during fixation and processing, as well as the unknown amount of carbon added by embedding and coating; a comparison with the La concentration in this liver as assayed chemically would thus not be appropriate, and the quantification only serves to obtain relative information on the microscopic distribution of the La inside the lysosomes. Spot-probe EELS analyses as well as EFTEM indicate a gradual increase of La from the outside of the lysosomes towards the centre. It is less clear whether this is due to a structural feature of an actual La gradient inside the lysosome or whether it is caused by viewing a sphere-like volume for which the projection of the edges automatically contains less La. However, the fact that the lysosome measured in Fig. 5 is at least half a micrometre in diameter and that the EFTEM is performed on 100-nm-thick sections would indicate that a real gradient is indeed present.

As biliary elimination is the main excretion route of La, some concern has been raised as to whether the long-term therapeutic use of La might be associated with hepatotoxic effects. However, the observations of the present study showing that the La is uniquely present in the lysosomes is in line with the absence of any hepatotoxicity noticed so far in both clinical and experimental studies. The actual form of the molecular appearance might be relevant as well. Under the extreme magnifications obtained in a high-resolution TEM instrument, it is seen that the precipitates in the lysosomes reveal a crystallographic nature. Two distinct lattice parameters could be determined: the appearance of the 0.55-nm distance as well as the fact that this is observed as part of a superlattice excludes the possibility of pure La in the room temperature hcp structure. The values found are in the range of some of the known lanthanum phosphate structures, such as orthorhombic LaP3O9 or hexagonal LaPO4, with the 0.55-nm spacing particularly pointing towards the (200) spacing of the former. Still, the obtained precision does not enable one to make a definite decision on the structure, as many systems display candidate spacings for the 0.3-nm case, but when including the quantification performed from EDX spectra as in Fig. 3(a), the La/P ratio seems to favour the orthorhombic phase. The size of the particles as seen from the high-magnification images implies that the trace taken in Fig. 5 is averaged over several particles in the lateral direction for each position along the trace, explaining the observation of a gradual rather than a stepwise increment. However, it is unknown to what extent the sample preparation and/or any radiation effect in the TEM could have affected the morphology and structure of these particles. Still, the visible La precipitates and EELS spectra did not change or weaken during several minutes of observation in a parallel electron beam, so there is no direct indication of any radiosensitivity. Moreover, in living material inorganic salts are mostly linked to proteins, improving stability after fixation, as in bone matrix and dental enamel (Bloom & Fawcett, 1994).

The finding that La is concentrated in the lysosomes of hepatocytes is compatible with the assumption that the element strongly binds to proteins, among them transferrin, by which the element enters the hepatocyte by transferrin receptor-mediated endocytosis, and fusion of the endosomes with the hepatocellular lysosomes (phagolysosomes), which release their content into bile after fusion of the lysosomal membrane with the membrane of the biliary canaliculus (Larusso & Fowler, 1979). La follows hereby the classic transcellular pathway of serum protein-bound metals. The role of the paracellular pathway and of ductular secretion of this highly protein-bound trivalent cation appears minimal (Ballatori, 1991). As long as the concentration in the lysosomes of metals is below a certain level, this type of compartmentalization is considered to be protective against possible cellular toxic effects of metals (Fuentealba et al., 1989). However, progressive accumulation of metals into the lysosomes may in the long term, for several reasons, interfere with the structural integrity of these organelles. Hence cellular damage may occur, as is well documented for iron. In the case of La, clinical studies over more than 5 years in several hundreds of patients have never disclosed any hepatotoxicity as evaluated by the classic liver enzymes determined in the serum of La-treated patients (D’Haese et al., 2006).


From this work, it can be concluded that La is present in some but not all lysosomes of animals treated with a daily injection of La chloride. The La appears as nanoscale particles yielding views of granular precipitates with a gradual increase of La towards the centre of the lysosome. To what extent the observed crystallinity of these particles can be related to the in vivo situation is unclear at present. Near the lysosomes no La is detected, which implies that the La is well confined to the lysosomes, thereby reducing potential toxic effects.