Lung stereology has a long and successful tradition. From mice to men, the application of new stereological methods at several levels (alveoli, parenchymal cells, organelles, proteins) has led to new insights into normal lung architecture, parenchymal remodelling in emphysema-like pathology, alveolar type II cell hyperplasia and hypertrophy and intracellular surfactant alterations as well as distribution of surfactant proteins. The Euler number of the network of alveolar openings, estimated using physical disectors at the light microscopic level, is an unbiased and direct estimate of alveolar number. Surfactant-producing alveolar type II cells can be counted and sampled for local size estimation with physical disectors at a high magnification light microscopic level. The number of their surfactant storage organelles, lamellar bodies, can be estimated using physical disectors at the EM level. By immunoelectron microscopy, surfactant protein distribution can be analysed with the relative labelling index. Together with the well-established classical stereological methods, these design-based methods now allow for a complete quantitative phenotype analysis in lung development and disease, including the structural characterization of gene-manipulated mice, at the light and electron microscopic level.
The main function of the lung is gas exchage. According to Fick's diffusion law, the rate of oxygen uptake is structurally limited by the surface area and the thickness of the diffusion barrier. The internal structure of the lung is optimized to fulfil this function. In humans, the gas exchange surface area of the alveolar epithelium is 120–140 m2 (roughly the size of a tennis court), while the diffusion barrier has a harmonic mean thickness of only 0.6 µm (about 50 times thinner than a sheet of air-mail paper) (Weibel, 1984a).
The air–liquid interface at the alveolar epithelium generates the surface tension that contributes to the retractive forces that oppose overdistension of the delicate lung parenchyma. However, surface tension has to be lowered and moderated to prevent alveolar collapse. The surface active agent that is necessary to keep lung alveoli open is termed pulmonary surfactant (comprehensively reviewed in Clements, 1997; Hawgood, 1997; Wright, 1997; Notter, 2000). Pulmonary surfactant is composed of around 90% lipids, mainly saturated phospholipids, and around 10% proteins, including the surfactant apoproteins SP-A, -B, -C and -D (Crouch & Wright, 2001; Hawgood & Poulain, 2001; McCormack & Whitsett, 2002; Whitsett & Weaver, 2002; Wright, 2005). The hydrophobic SP-B and -C are mainly involved in the biophysical surfactant functions (thus contributing to alveolar stability), while the hydrophilic SP-A and -D belong to the protein family of collectins involved in immunomodulation (thus contributing to alveolar sterility). Surfactant is synthesized, stored, secreted and, to a large extent, recycled by alveolar type II cells (Mason & Shannon, 1997; Fehrenbach, 2001). Specific organelles, termed lamellar bodies, represent the intracellular storage form of surfactant (Fig. 1). Within the alveolus, intra-alveolar surfactant is present as ‘fresh’ surface active forms (ultrastructurally largely corresponding to tubular myelin) and as ‘spent’ inactive forms (ultrastructurally largely corresponding to unilamellar vesicles). Thus, different intra-alveolar surfactant subtypes that can be distinguished morphologically correspond to different stages in surfactant metabolism.
The quantification of lung structure by means of stereology has a long and successful tradition (Weibel, 1996; Weibel, 2001). Many stereological parameters that describe the structural composition of the lung at the light and electron microscopic level have been well established for decades (usually derived from ratios, e.g. component volumes of alveolar air and alveolar septal tissue, alveolar surface area and blood–air barrier thickness [see Weibel (1963) and Gehr et al., (1978) for applications to the human lung]. In many cases, these parameters are still sufficient to provide the necessary quantitative data on lung structure – when applied properly [see Weibel (1979) and Weibel (1980)]. A typical example is the assessment of emphysematous alterations in animal models of chronic obstructive pulmonary disease. Although widely used for this purpose, the presence of emphysema-like pathology cannot be concluded from the mean linear intercept length of distal airspaces or radial alveolar counts. Both parameters only indicate a relative increase in distal airspace size, expressed as a zero- or one-dimensional parameter obtained on single histological sections that does not differentiate between alveolar airspace and alveolar duct airspace. This is certainly not sufficient to prove true destructive emphysema, which is characterized by severe alterations of acinar architecture including a loss of alveoli and gas exchange surface. Therefore, total alveolar surface area should be estimated instead (Heemskerk-Gerritsen et al., 1996; Wiebe & Laursen, 1998; Fehrenbach, 2002–03). With new design-based methods available (see below), even total alveolar number and mean alveolar size can now be estimated, and allow for an unambiguous determination of loss (or even recovery) of gas exchange tissue (Massaro & Massaro, 1996; Massaro et al., 2002; Ochs et al., 2004a,b). Recommended parameters for the stereological assessment of emphysematous alterations are given in Table 1.
Table 1. List of stereological parameters recommended for the analysis of emphysematous alterations.
Method to estimate the parameter
After estimation of total lung volume with the Cavalieri method, which can easily be integrated into the sampling procedure, point counting in a cascade sampling design will generate all necessary subcomponent volumes (Fig. 3). Loss of alveolar septal tissue in emphysema will be reflected by a decreased total alveolar epithelial surface area, a decreased total number of alveoli, and an increased number-weighted mean alveolar volume. If, in addition, the volume-weighted mean alveolar volume is estimated, information on the size distribution of alveoli can be derived since . An increased mean alveolar septal thickness can be used as an indicator of fibrotic alterations.
total lung volume
total parenchymal volume
total parenchymal air volume
total parenchymal tissue volume
total alveolar volume
total alveolar septal volume
total alveolar epithelial surface area
total number of alveoli
Euler number using physical disector
number-weighted mean volume of alveoli
volume-weighted mean volume of alveoli
mean alveolar septal thickness
2 · [V(alvsep)]/[S(alvepi)]
Classical stereological methods were also applied to study the surfactant system. The analysis of the volume fractions of surface active and inactive intra-alveolar surfactant forms in a rat lung model of ischemia/reperfusion injury revealed alterations that closely correlated with postischemic lung function (Fehrenbach et al., 2000; Ochs et al., 2000), thus emphasizing the clinical importance of surfactant preservation in lung transplantation (Lewis et al., 1997; Ochs, 2001). Moreover, stereology showed that the intra-alveolar surfactant alterations in ischemia/reperfusion injury are a cause rather than a mere result of intra-alveolar oedema formation (Ochs et al., 1999). An increased alveolar surface tension results in an increased pressure gradient from the capillary to the alveolar lumen, and thus in fluid fluxes across the blood–air barrier. The physiological function of surfactant is therefore not only to keep alveoli open and clean, but also to keep them dry.
The importance of these steps cannot be overemphasized. It is obvious that, although theoretically unbiased stereological methods are available for first order properties like volume, surface area, length, particle number and particle size, other sources of bias might still remain, e.g. due to improper tissue processing or structure recognition. Therefore, adequate lung fixation, tissue processing, embedding and staining methods, as well as sufficient observer experience in lung morphology, are extremely important. After all, even with the well-deserved appreciation of stereology that led several established journals to require state-of-the-art techniques for counting of particles (Coggeshall & Lekan, 1996; West & Coleman, 1996; Madsen, 1999; Kordower, 2000), we should always bear in mind that ‘stereologists […] should identify the shapes of structures before they measure them’ (Elias, 1972).
Every fixation for light and electron microscopy produces an artifact, in the best sense of the word, which represents some aspects as faithfully as possible but at the expense of others (Weibel et al., 1982). We therefore have to live with the ‘failed dream of the physiological fixation’ (Gil, 1990). In principle, whole lungs of any size can be fixed chemically by instillation via the airways or by perfusion via the vasculature. Both routes of fixation have their merits and limitations (Bachofen et al., 1982; Weibel, 1984b; Gil, 1990; Fehrenbach & Ochs, 1998). Physical fixation of small lung pieces by rapid freezing is usually not suited for stereological analysis because the whole lung should be available for systematic uniform random sampling. At present, this can only be achieved by chemical fixation by instillation or perfusion, which also fulfils the standard criteria of compatibility and consistent reproducibility (Weibel et al., 1982).
The choice of the chemical composition of the fixative depends on whether light and/or transmission electron microscopic analysis will be performed and whether structural preservation alone or also preservation of antigenicity has to be achieved. Various methods for conventional and immunoelectron microscopy are reviewed in Griffiths (1993), Hayat (2000) and Skepper (2000). Protocols for the lung are given in Weibel (1984b) and Fehrenbach & Ochs (1998).
Modern microscopy goes far beyond simple qualitative description of structures. Raw data generation at the microscope is followed by data analysis, e.g. by image analysis or stereology with a computerized system. In practice, quantification in microscopy by stereology is basically a two-step procedure: the first step is to take a sample, the second is to make a measurement by probing the sample with a test system (see Fig. 2). The first step is often underestimated outside the field of stereology. The generation of quantitative data must start with the biological system of interest and not with images captured at the microscope. Therefore, the image on the computer screen is not only a starting point for analysis but also the result of a sampling process (chosen individuals, tissue blocks, sections, fields of vision). Adequate sampling procedures are indispensable to assure that whatever is analysed at the microscope is representative and therefore meaningful with respect to the biological system. Stereology offers highly efficient methods to achieve this. Systematic uniform random sampling guarantees that all parts of an organ are represented equally well (Cruz-Orive & Weibel, 1981; Gundersen & Østerby, 1981; Gundersen & Jensen, 1987; Mayhew, 1991; Mayhew & Gundersen, 1996; Nyengaard, 1999). Starting from outside, the first item in the sample is chosen randomly, but then determines the position of all other items. It is obvious that the sampling chain has to be complete at all levels of the sampling cascade. To be efficient (i.e. to achieve a sufficient precision within a reasonable amount of time), more effort should be put into the higher levels of the sampling cascade, a principle known as ‘do more less well’ (Weibel, 1981; Gundersen & Østerby, 1981).
In lung stereology, estimates are typically obtained as ratios that are then transformed into absolute values via a cascade sampling at different levels of light and electron microscopic magnification, leading to a ‘density × reference volume’ design (see Weibel, 1963; Weibel, 1979; Weibel, 1980; Cruz-Orive & Weibel, 1981). In this cascade or multilevel sampling design (Fig. 3), the compartment of interest (numerator) of one level becomes the reference compartment (denominator) of the next level. It is essential that each compartment is defined identically in the denominator of each ratio and in the numerator of the preceding one. This approach requires special care to avoid technical bias due to tissue shrinkage during processing. In general, either methods that prevent shrinkage have to be used or shrinkage has to be measured and corrected for (for strategies to handle tissue shrinkage in stereology, see Dorph-Petersen et al., 2001). While global tissue shrinkage (i.e. all components of the tissue block are subject to the same degree of shrinkage) can be corrected for, differential shrinkage (i.e. certain components of the tissue block behave differently than others) is hard to manage. For light microscopy, paraffin embedding is widely used. While well suited for immunohistochemistry, its major disadvantage is the very high and unpredictable degree of shrinkage that makes correction factors necessary. Whenever possible, embedding in plastic (glycol methacrylate) is recommended instead to avoid these problems (Gerrits & Horobin, 1996). However, especially for lung tissue, shrinkage should always be checked since differential shrinkage might still occur even when plastic embedding is used. In our experience, osmication and prolonged staining of tissue blocks with uranyl acetate prior to dehydration and embedding in plastic prevented differential shrinkage. Osmication makes membranes permeable and thus insensitive to osmotic changes (Wangensteen et al., 1981). At the electron microscopic level, tissue dimensions are supposed to remain rather stable after embedding in Epon, Araldite, Lowicyls or London Resins.
Another approach is fractionator sampling (Gundersen, 1986), now in its smooth version (Gundersen, 2002), leading to a ‘total counts × sampling fraction’ design for number estimation (see below). Since the fractionator is independent of changes in tissue dimensions, it is the method of choice for number estimation in the lung whenever applicable.
In contrast to number and global volume, surface area and length as well as local size estimators are sensitive to tissue orientation. Although lung parenchyma, in contrast to larger airways and blood vessels, can generally be regarded as isotropic, methods to account for anisotropy during sampling and processing for LM or EM studies are available: vertical sections (Baddeley et al., 1986), the orientator (Mattfeldt et al., 1990) or the isector (Nyengaard & Gundersen, 1992). Vertical sections have been, for example, used to estimate pleural surface area in the rabbit lung (Michel & Cruz-Orive, 1988).
Whenever applicable, absolute quantities (i.e. ‘per lung’ data) should be obtained. Any misinterpretations due to the use of ratios alone (termed the ‘reference trap’, see Braendgaard & Gundersen, 1986; Casley-Smith, 1988) are thereby avoided. Determination of total lung volume is therefore extremely important and, in a ‘density × reference volume’ design, influences all final data. This can be done either by fluid displacement or using the Cavalieri principle. While the classical method to determine organ volume in the past was fluid displacement (Scherle, 1970), the Cavalieri principle (Gundersen & Jensen, 1987; for lung applications, see Michel & Cruz-Orive, 1988; Fehrenbach & Ochs, 1998; Yan et al., 2003; Ochs et al., 2004a) offers several advantages. (a) Organ volume is estimated after cutting and is therefore closer to the final dimensions of the tissue. Lung tissue recoil forces are not completely eliminated after aldehyde fixation (Bachofen et al., 1982; Oldmixon et al., 1985; Yan et al., 2003). Total lung volume determined by the Cavalieri method therefore more closely represents the state of the tissue to be used for subsequent stereological analysis, especially for larger lungs. (b) Subcomponent volumes can be estimated already at a macroscopic level without physically seperating them, e.g. coarse airway and vascular nonparenchyma. (c) Data sets for the estimation of organ volume can be generated without physically cutting the organ, e.g. using noninvasive CT or MRI scans (for an application to the lung, see Pache et al., 1993). Finally, the Cavalieri method does not require much additional effort when integrated into the sampling scheme (e.g. using lung slices of appropriate thickness for the Cavalieri estimator as well as for sampling of tissue blocks for light microscopy; see Fig. 3).
If total lung volume is known, the total volume of any subcomponent within the lung can easily be estimated by point counting in a cascade sampling design. This approach has been used successfully to study pulmonary oedema formation. In contrast to the lung wet/dry ratio, which is commonly used as an indicator of fluid accumulation in the lung, stereology allows for distinction between intra-alveolar, interstitial and peribronchovascular oedema. In experimental ischemia/reperfusion injury, stereological analysis of pulmonary oedema has also been shown to better reflect the functional status of the lung (Ochs et al., 2000; Fehrenbach et al., 2001). It is thus the method of choice to evaluate the functional degree of oedema formation in acute lung injury.
Particle number and size
The formal introduction of the disector method in stereology (Sterio, 1984) completed the set of probes necessary to estimate first order stereological parameters (Fig. 2). Since the disector method can be used not only to count particles but also to sample them for subsequent estimation of particle size (see below), it is also a prerequisite for many local stereological estimators.
The smallest gas exchange units in the mammalian lung are alveoli. Many lung diseases, like emphysema, are characterized by a loss of alveoli. The classical method to estimate alveolar number (Weibel & Gomez, 1962) is based on assumptions, e.g. on the shape and size of alveoli. In principle, an assumption-free stereological method to estimate alveolar number should use disectors (see Fig. 2). However, alveoli are not isolated particles but rather a network of ‘incomplete’ structures, connected via their openings. A rigorous topological definition of the appearance of an alveolus in a disector is therefore necessary. Each alveolus has only one ‘true’ opening (into an alveolar duct or a respiratory bronchiole). With their entrance rings at the free edges of the alveolar septae, alveolar openings form a two-dimensional network in three-dimensional space. The Euler number of this network can be estimated. Named after the Swiss mathematician Leonhard Euler (1707–1783), the Euler number is an integer-valued measure for the connectivity of an object (Gundersen et al., 1993; Nyengaard & Marcussen, 1993; Kroustrup & Gundersen, 2001). Thus, the Euler number of the network of alveolar openings depends on the number of connections in the network. Topological changes of this network in physical disectors can be counted and indicate the appearance of new alveolar openings (Fig. 4a).
From a practical perspective, disector heights beween 3 µm (for mice) and 9 µm (for humans) and primary magnifications between 5 (for mice) and 2.5 (for humans) have been found convenient (Fig. 4b). To distinguish ‘true’ alveolar openings (that are counted) with their elastic fibre rings from interalveolar pores of Kohn (that do not contribute to the network of alveolar openings and are thus not counted), a stain that demonstrates elastic fibres like orcein or Van Gieson is recommended.
In a first application to the human lung (Ochs et al., 2004a), a total alveolar number of 480 million has been estimated, with a close correlation with total lung volume. Thus, in humans larger lungs are built by increasing the number of alveoli, not their size. In Wistar rats, around 20 million alveoli have been found (Hyde et al., 2004) while in C57Bl6/129Sv:CD-1 mice, around 12–13 million alveoli were estimated (Ochs et al., 2004b; Jung et al., 2005). It is, however, important to notice that these numbers might vary considerably between different mice strains due to differences in total lung volume.
When total alveolar number (from the Euler number estimation) and total alveolar volume (from point counting in a cascade design) are known, the number-weighted mean alveolar volume can also be estimated. The volume-weighted mean alveolar volume can be estimated with the point-sampled intercepts method (Gundersen & Jensen, 1985; for applications to lung alveoli, see Fehrenbach et al., 1999; Hawgood et al., 2002; Yang et al., 2002). If both number- and volume-weighted alveolar volumes are known, information on the size distribution of alveoli can be obtained (Table 1). A possible alternative is to estimate mean alveolar size with the selector (Cruz-Orive, 1987) and calculate total alveolar number indirectly from total alveolar volume and mean alveolar size (Massaro & Massaro, 1996).
To predict the coefficient of error (CE) of the new Euler number estimate of alveolar number, two shortcut formulae are available at present, one developed for a ratio estimator (Kroustrup & Gundersen, 1983), and one developed specifically for a connectivity estimator (Gundersen et al., 1993). However, these formulae lead to results that might differ considerably, so the biomedical researcher is left with the problem of which CE predictor to trust when optimizing the sampling design. In gene-manipulated mouse models as well as in human disease, lung alterations might be very heterogeneous, thus increasing the variability between sections and fields of view. The problem of CE prediction is even more pronounced when gene-manipulated mice are studied. Here, genetic variability is kept as low as possible, which is necessary to attribute differences in the phenotype to the manipulated gene. Usually, inbred, fully backcrossed mice of the same sex are used. Thus, the biological variation between animals is extremely low in these models. This means that, in some cases, it might be necessary to increase the sampling effort to ensure that the total observed variation is largely dominated by the biological variation between individuals and not by the variation due to stereological sampling.
Type II cells
Surfactant-producing alveolar type II cells can be counted using physical disectors at the light microscopic level. To make the recognition of type II cells easier, tissue blocks should be osmicated and stained with uranyl acetate before embedding in glycol methacrylate. The cell's lamellar bodies are then well preserved and show a characteristic metachromasia when stained with methylene blue or toluidine blue. Alternatively, semithin sections from Epon blocks can be used. In general, consecutive sections should be used for the physical disector to faciliate particle recognition from one section to another. When counting type II cells with their low density in thin plastic sections, it is possible to make an exception to this rule as long as the disector height is kept considerably smaller than the smallest type II cells so that no cells are missed. We usually use a disector height of 3 µm (by using the first and fourth section of a consecutive row of 1 µm sections) and a primary magnification of 63 or 100 (high numerical aperture oil immersion lenses) (Fig. 5).
In the human lung, around 24 billion type II cells have been estimated, i.e. around 50 per alveolus (Ochs et al., 2001). In Sprague–Dawley rats, around 86 million type II cells were found (Waizy et al., 2001), while the lungs of C57Bl6/129Sv:CD-1 mice contain around 13–15 million type II cells, i.e. around 1 per alveolus (Ochs et al., 2004b; Jung et al., 2005).
Since the disector is not only a method for unbiased counting of particles but also for unbiased sampling of particles, the number-weighted mean volume of type II cells can then be estimated with local stereological methods like the nucleator (Gundersen, 1988) or planar rotator (Vedel Jensen & Gundersen, 1993).
Within type II cells, surfactant-storing lamellar bodies can be counted using physical disectors at the electron microscopic level (Ochs et al., 2001). Two consecutive ultrathin sections of 100 nm thickness are mounted in parallel on one slot grid. Corresponding type II cell profiles are recorded at a primary magnification of 3000–5000 (Fig. 6). The total volume of lamellar bodies per lung can be regarded as a morphological measure of the size of the intracellular surfactant pool.
In the human lung, one type II cell contains around 200–500 lamellar bodies, and the total lamellar body volume per lung is around 1900 mm3 (Ochs et al., 2001). A single type II cell in Sprague–Dawley rats contains around 60–120 lamellar bodies, while the total volume of lamellar bodies per lung is around 2.6 mm3 (Waizy et al., 2001). In C57Bl6/129Sv:CD-1 mice, one type II cell contains 50–100 lamellar bodies, with a total lamellar body volume per lung of about 0.7 mm3 (Ochs et al., 2004b; Jung et al., 2005).
The number-weighted mean volume of lamellar bodies can be calculated from their total volume (estimated by point counting in a cascade design) and their total number. The volume-weighted mean volume of lamellar bodies can be estimated with the point-sampled intercepts method (Gundersen & Jensen, 1985). If both number- and volume-weighted volumes are known, information on the size distribution of lamellar bodies can be obtained. Recommended parameters for the stereological assessment of type II cells and their lamellar bodies are given in Table 2. These estimators allow the important differentiation between alterations in the number of surfactant-producing cells (type II cell proliferation) and alterations in surfactant content per cell (intracellular surfactant metabolism).
Table 2. List of stereological parameters recommended for the analysis of type II cells and lamellar bodies.
Method to estimate the parameter
Type II cell hyperplasia and hypertrophy will be reflected by an increase in the total number of type II cells and their number-weighted mean volume, respectively. As a morphological measure of intracellular surfactant pool size, the total volume of lamellar bodies can be expressed per type II cell or per lung. Information about the way how this surfactant amount is handled by type II cells can be obtained by estimating the number of lamellar bodies per type II cell and their number-weighted mean volume. If, in addition, the volume-weighted mean volume of lamellar bodies is estimated, information on the size distribution of lamellar bodies can be derived because .
total lung volume
total parenchymal volume
total number of type II cells
number-weighted mean volume of type II cells
planar rotator or nucleator
V(lb, type II)
total volume of lamellar bodies per type II cell
total volume of lamellar bodies per lung
V(lb,type II) · N(type II)
N(lb, type II)
number of lamellar bodies per type II cell
number-weighted mean volume of lamellar bodies
V(lb,type II)/N(lb, type II)
volume-weighted mean volume of lamellar bodies
Owing to inherent limitations, immunohistochemical data obtained at the light microscopic level have to be interpreted with caution when it comes to conclusions on subcellular localization or possible colocalization. This has been clearly shown by careful immunoelectron microscopical investigations (Griffiths et al., 1993; Keller & Eggli, 1998). For subcellular localization of gene products with high spatial resolution, immunoelectron microscopy is the method of choice (Fig. 7). Since the introduction of colloidal gold as a particulate marker for immunolabelling at the EM level (for review, see Griffiths, 1993; Bendayan, 2001), immunoelectron microscopy can be combined with stereology to study the distribution of labelled molecules. The principles of ‘molecular morphometry’ were already outlined about 25 years ago (Weibel, 1981). Principles and techniques for quantifying a colloidal gold labelling with stereological methods are reviewed in Griffiths (1993) and Lucocq (1994). A new and efficient method to test for preferential labelling of cell compartments has been developed recently: relative labelling index (RLI) (for the first description, see Mayhew et al., 2002; for further developments, see Mayhew et al., 2003; Lucocq et al., 2004; Mayhew et al., 2004). The RLI is used to compare an observed intracellular distribution of gold particles over defined cell compartments (whose volume fraction is obtained by point counting) with an expected distribution that would occur if the same total number of gold particles were scattered randomly over the cells (i.e. according to the volume fraction of the compartments). The RLI for each compartment is determined as the observed value divided by the expected value. Thus, an RLI of 1 for all compartments would indicate a pure random labelling. If, on the other hand, the RLI of a certain compartment is significantly higher than 1, as tested by χ2-analysis, and if the compartment's partial χ2-value contributes substantially to the total χ2-value of all compartments, this indicates preferential (or, if the necessary criteria for antibody specificity are met: specific) labelling of this compartment.
In an immunoelectron microscopic study, the distribution of SP-A within human type II cells has been analysed with the RLI. Despite only around 2 gold particles per cell over lamellar bodies, RLI detected this weak signal to be due to preferential labelling (Ochs et al., 2002). The RLI has also been used to compare surfactant protein distribution in different age groups during postnatal lung development in the rat (Schmiedl et al., 2005).
Conclusions and some future prospects
Stereology provides methods to study the lung under physiological and pathological conditions, including the complete quantitative lung phenotype analysis of gene-manipulated mice that are used to analyse the in vivo function of selected genes. However, the potential of the combination of electron microscopy and stereology in cell biology has still not been fully exploited (Griffiths, 2001).
The current limitations of rapid freezing as a fixation method for stereological studies at the electron microscopic level might be overcome with the use of high-pressure freezing devices (Studer et al., 2001) in combination with microbiopsy systems (Vanhecke et al., 2003) that allow for a more representative tissue sampling. However, the air-filled lung is an especially challenging organ with respect to high-pressure freezing.
Traditionally, biological objects had to be destroyed (physically cut into nearly two-dimensional thin sections) to perform stereology. Nowadays, new technologies generate optical sections through biological objects that can be analysed by stereology. At the macroscopic level, noninvasive scanning techniques like CT and MRI are ideally suited for reference volume estimation with the Cavalieri principle (Roberts et al., 2000). Micro-CT is a promising tool for analysis of emphysema-like pathology in mice (Recheis et al., 2005). However, before these methods can be used routinely, they have to be verified against the existing gold standard, i.e. against stereology applied on histological sections. Recently, it has been shown that, if applied properly, micro-CT can be used as a substitute for histological sections when investigating trabecular bone structure (Thomsen et al., 2005). At the microscopic level, confocal laser scanning microscopy provides datasets from thick sections where registered optical section planes with known distance can be used for stereology (Kubinova & Janacek, 2001). The potential of datasets from electron tomography of thick EM sections (McIntosh, 2001; Baumeister, 2002; Koster & Klumperman, 2003; Bonetta, 2005; McIntosh et al., 2005) for stereological analysis has yet to be explored.
The estimation of the Euler number in combination with the disector can be applied to estimate the number of any complex structure in the lung, e.g. functional units at the acinar or subacinar level (‘ventilatory units’; Wulfsohn et al. unpublished results) or capillaries (the vascular network of the lung).
The development and application of stereological methods to the lung has led and will lead to significant new insights into the biology of the respiratory system – from comparative physiology to cell biology. The long and successful story of lung stereology is about to continue.
This brief review is based on a presentation given at the Workshop on Variance Estimation in Stereology, Aarhus, Denmark, November 2004.
Many thanks to Barbara Krieger, Beat Haenni, and Véronique Gaschen (Bern) for their invaluable help with the figures and to my collaborators Joachim Richter (Göttingen), Frank Brasch (Bochum), Jens R. Nyengaard, Hans Jørgen G. Gundersen (both Aarhus), Sam Hawgood (San Francisco), and Thorsten Wahlers (Jena). A very special thanks to Ewald R. Weibel (Bern) for critical reading of the manuscript and many helpful comments.
Grants: Deutsche Forschungsgemeinschaft (DFG OC 23/7–3 and 8–1).