Introduction

Podocytes are glomerular epithelial cells that play critically important roles in glomerular filtration (Jefferson et al. 2011). A landmark study by Wharram et al. (2005) showed that podocyte depletion is sufficient for the development of glomerular scarring (glomerulosclerosis); both podocyte depletion and glomerular scarring are found in a wide range of glomerular diseases (Wiggins 2007). If a sufficient number of glomeruli become non-functional because of sclerosis, chronic kidney disease and end-stage kidney disease can occur. As a consequence, the podocyte depletion hypothesis has been proposed as a unifying principle of glomerular pathology and is currently the focus of much research.

An important aspect to note is that podocytes are post-mitotic cells that are unable to re-enter the cell cycle and successfully complete cytokinesis (Lasagni et al. 2013). For decades, lost podocytes were believed to be irreplaceable and it was universally accepted that, at the completion of nephrogenesis, kidneys contained their final complement of both nephrons (nephron endowment) and podocytes (podocyte endowment; Puelles et al. 2011). During postnatal life, physiological and pathological stressors have been shown to lead to podocyte depletion in rodents and humans (Puelles and Bertram 2015). These findings have been interpreted as further evidence that podocytes cannot be replaced, although recent studies suggest that podocyte gain may occur in early postnatal life (Berger et al. 2014) and in settings of podocyte depletion (Angelotti et al. 2012; Peired et al. 2013; Pichaiwong et al. 2013; Pippin et al. 2013; Zhang et al. 2013; Wanner et al. 2014; Eng et al. 2015; Lasagni et al. 2015; Roeder et al. 2015).

Given the central role of podocytes in glomerular disease, they are considered potential targets for the development of future therapeutic strategies (Allison 2015), with a focus on reducing or preventing podocyte depletion and/or enhancing podocyte repair/gain. The ability to quantify podocyte depletion reliably is a central factor in the development of therapeutic strategies. To date, no consensus exists concerning the best approach for quantifying podocyte depletion, although all methods involve the counting of podocyte nuclei. Two recent reviews discussed the advantages and disadvantages of most available methodologies (Lemley et al. 2013; Puelles and Bertram 2015). In this review, we analyse existing methods, highlight new developments and consider practical issues that may influence the choice of method for podocyte quantification.

Definition of podocyte depletion

The current literature describes two forms of podocyte depletion (Puelles and Bertram 2015): (1) absolute podocyte depletion, which involves the loss of podocytes; (2) relative podocyte depletion, which occurs, for example, when glomeruli hypertrophy and a decrease in podocyte density (number per volume) occurs. The initially proposed podocyte depletion hypothesis also considered podocyte phenotypic switch or foot process effacement, which represents a transient loss of functional podocytes, to be a form of podocyte depletion (Wiggins 2007). However, in order to simplify the present analysis, we only focus on absolute and relative podocyte depletion.

A few years ago, Wharram et al. (2005) and Fukuda et al. (2012) showed that both absolute and relative podocyte depletion are sufficient for the development of glomerulosclerosis. In both studies, a threshold of podocyte depletion was found to be required before the development of pathological changes. For example, in the case of absolute podocyte depletion, loss of less than 20 % of podocytes led to mesangial expansion but no changes in renal function, whereas loss of more than 40 % of podocytes led to overt focal segmental glomerulosclerosis and reduced renal function (Wharram et al. 2005). This finding clearly demonstrated the value of being able to quantify podocyte depletion accurately and precisely quantify in order to obtain important insights into glomerular pathophysiology.

Recent studies have also suggested that podocyte size, for example, average podocyte volume, or podocyte nuclear and cytoplasmic volumes reflect settings of absolute and relative podocyte depletion (Venkatareddy et al. 2014; Puelles et al. 2016b). When podocytes are facing stress, for example, following podocyte loss or increases in glomerular volume, increases in podocyte size may reflect an attempt to maintain an intact glomerular filtration barrier and thereby glomerular health (Kriz 2012). For this reason, podocyte volume may serve as an indirect measure of podocyte depletion in at least some circumstances.

Methods to quantify podocyte depletion

Podocyte number per glomerular cross-section

This two-dimensional approach remains the most commonly used method for quantifying podocyte depletion, because it is easy and fast. The method simply requires the counting of the number of podocyte nuclei in a histological section taken through a glomerulus. Unfortunately, it has many major limitations.

First, a histological section through a glomerulus (glomerular cross-section) represents only a small fraction of a whole glomerulus. For example, for a mouse glomerulus (approximately 80 μm in diameter), a 4-μm-thick paraffin section represents ∼5 % of the glomerular thickness. Thus, this section may not provide information on the total number of podocytes in the glomerulus. The second limitation is that the likelihood of podocyte nuclei appearing in histological sections may not necessarily be identical to the frequency of the same podocyte nuclei being present in three-dimensional glomerular tissue, because podocyte nuclei may have various sizes and shapes. A third limitation is that larger glomeruli are preferentially sampled compared with smaller glomeruli in single histological sections (Bertram 1995). A fourth limitation is that section thickness influences the number of podocyte nuclei present, with thicker sections having more podocyte nuclei than thinner sections.

The issue of variability in glomerular size and the effects on glomerular sampling in single histological sections deserves further consideration, because heterogeneity in glomerular size is present in normal kidneys and non-uniform hypertrophy of glomeruli is a feature of both normal kidney growth and many kidney diseases (Puelles et al. 2011). To illustrate this problem of glomerular sampling, let us imagine that we have a box containing 10 watermelons and 10 apples. Given that watermelons are much larger and have a different shape from apples, the chance of sampling a watermelon with a single slice (section) through the box is significantly higher than sampling an apple, even though 10 of each are present in the box (Fig. 1a; Bertram 2001). Because podocyte nuclei are large in relation to section thickness, the number of podocyte nuclei present in a glomerular cross-section depends not just on the number of podocyte nuclei in the glomerulus but also on nuclear size, nuclear shape and section thickness (Lemley et al. 2013). Importantly, if glomerular size (reference space) increases but podocyte number remains constant, then the number of sampled nuclei is reduced (Fig. 1b).

Fig. 1
figure 1

Problems with two-dimensional sampling. a Glomerular sampling: apples (red) and watermelons (green) in a box. b Podocyte nuclear (grey) sampling in two reference spaces of differing size; the dotted lines and ticks represent two-dimensional sampling. In both cases, while the number of apples and watermelons are the same (a), and the number of podocyte nuclei is the same in both reference spaces (b), two-dimensional sampling favours larger structures (3:1 ratio, when it should be 1:1) and the larger reference space reduces the likelihood of sampling the same number of nuclei (3:1 ratio, when it should be 3:3)

Full size image

Podocytes are not all the same size and shape, especially in the context of glomerular stress (Venkatareddy et al. 2014; Hodgin et al. 2015). Furthermore, glomerular size changes in various pathophysiological settings (Puelles et al. 2011, 2012). These caveats are illustrated by a recent study (Puelles et al. 2016b) comparing three- and two-dimensional data obtained from whole glomeruli of PodCreiDTR mice (Guo et al. 2012; Wanner et al. 2014). The mice were injected with various doses of diphtheria toxin, which led to different levels of podocyte depletion and showed that two-dimensional quantification of podocyte number did not identify the subtle differences that were detected by three-dimensional methods (Puelles et al. 2016b; Fig. 2).

Fig. 2
figure 2

Glomerular cross-sections versus whole glomerular analysis. a Control. Whole mouse glomerulus labelled with p57 (red podocyte nuclei) and synaptopodin (green podocyte cytoplasm) in which all podocytes (77) were counted by using exhaustive enumeration. b, b’ Two confocal optical sections through the same glomerulus showing variability in numbers of podocyte nuclei (red). c-e Quantitative differences between two- and three-dimensional analyses of podocyte depletion. With permission from the Journal of the American Society of Nephrology

Full size image

Thus, the reporting of the numbers of podocyte nuclei per glomerular cross-section does not provide an accurate estimation of the total number of podocytes per glomerulus. In this context, values will be influenced by changes in podocyte nuclear dimensions; subtle and important changes in podocyte numbers may therefore be missed. Furthermore, glomerular hypertrophy will result in the preferential sampling of large glomeruli and the under-sampling of smaller glomeruli. For these reasons, this approach should be avoided. Fortunately, more reliable methods are now available.

Stereological methods

The word stereology comes from the Greek roots “stereo” and “logos”, which together mean “science of studying solids”. Originally, stereology provided practical approaches for the three-dimensional quantification of structures based on two-dimensional measurements. However, in the 1980s and 1990s, stereological methods utilizing thick sections and/or tissue slices were developed and the traditional reliance on two-dimensional measurements was therefore reduced (Gundersen et al. 1988). Stereological methods can be divided into two broad categories: model-based and design-based. Model-based stereology, as the name implies, requires either knowledge of the geometry of the object of interest or assumptions regarding the geometry. For example, when counting and sizing podocytes with some model-based stereological methods, glomerular size and shape must be known. If these are not known, then assumptions are required, for example, that all glomeruli are spherical and have the same diameter. To the extent that these assumptions are incorrect, the estimates obtained will be biased (Lemley et al. 2013; Puelles and Bertram 2015). In contrast, design-based stereology is based on the application of probes and sampling schemes that are “designed” or “defined a priori”, so that the methods are independent of spatial orientation, distribution, size or shape. If applied correctly, these approaches are considered “unbiased”, given that their conception minimizes/removes the introduction of assumptions and thus of systematic bias.

Model-based stereology

The most commonly used model-based methods for the determination of podocyte numbers are the methods of Weibel and Gomez (1962) and a modification of the method of Abercrombie (1946) recently described by Venkatareddy et al. (2014). These methods are particularly useful when limited amounts of tissue are available (e.g., renal biopsies). In addition to several measurements on sections, the Weibel and Gomez method (1962) requires knowledge of the podocyte nuclear size distribution and nuclear shape. These values are difficult to obtain and are therefore typically assumed, potentially introducing bias.

The method described by Venkatareddy et al. (2014) relies on the direct measurement of the mean calliper diameter of podocyte nuclei, which is used to determine a correction factor for section thickness. This represents an important development because, previously, the mean calliper diameter was usually assumed rather than measured, providing a likely source of bias. Importantly, the method of Venkatareddy et al. (2014) also allows an estimation of podocyte size, including nuclear and cytoplasmic volumes. This method represents a significant advance in the field of podometrics, because it requires no geometric assumptions for podocyte nuclei and only requires a single section, not multiple sections or serial sections, as is the case with some other methods (see below). However, despite its many advantages, this method only provides an estimate of the average number of podocytes per glomerulus in a kidney. This is an important limitation given that podocyte depletion is focal in multiple disease settings; moreover, the use of a mean value per kidney does not provide any indication of the variation in podocyte numbers between glomeruli.

Reports of the reliability of model-based methods to estimate podocyte numbers acurately differ between studies. Steffes et al. (2001) and White and Bilous (2004) showed that podocyte numbers obtained by using model-based and design-based stereology in human biopsies are similar, especially when electron microscopy is used to identify and count podocytes. In whole mouse glomeruli, Basgen et al. (2006) found that the Weibel and Gomez method overestimates glomerular cell number and densities by 10 % compared with exhaustive enumeration (see below) and design-based methods. In contrast, Nichols et al. (2011) reported that model-based stereology overestimates podocyte number by 57 % and glomerular volume by 17 % in mice. Similar to the method of Venkatareddy et al. (2014), the method of Weibel and Gomez (1962) is easy to establish and reproduce but can provide data on individual glomeruli, thereby allowing the assessment of the variability between glomeruli within a specimen for the detection of focal injury. This is a potential advantage that should be taken into account.

Although model-based stereological methods have limitations, they can provide reliable and reproducible quantitative data and are especially useful when dealing with limited amounts of tissue. The direct measurement of the mean nuclear diameter of podocytes in the method of Venkatareddy et al. (2014) represents an important development. Indeed, recent reports from the Wiggins laboratory on podocyte depletion in human disease and ageing clearly demonstrate the immediate impact that this method is having on the field (Hodgin et al. 2015; Wickman et al. 2016).

We recommend that an effort to minimize the introduction of systematic bias should always be made when using model-based stereological methods, e.g., Venkatareddy et al. (2014) measured mean calliper diameter rather than relying on assumptions. Limitations and potential sources of bias should be identified, acknowledged and discussed.

Design-based stereology

Unlike model-based stereological methods, design-based stereological methods do not require knowledge or assumptions of the geometry of the objects of interest (Bertram 1995; Nyengaard 1999). However, these methods are often time-consuming, complex and difficult to implement.

We will use the disector/fractionator method as an example. Disector refers to the analysis of “two sections”. A disector can be “physical” when two histological sections are compared or “optical” when two optical sections (possibly confocal sections), typically from one thick section, are compared (Puelles and Bertram 2015). The fractionator principle relies on the systematic sampling of events (e.g., podocyte nuclei) based on a known fraction of the tissue (Nyengaard 1999). A disector/fractionator method for estimating podocyte number has recently been described for the analysis of human autopsy tissue (Puelles et al. 2014) and combines serial sectioning of paraffin-embedded tissue, immunofluorescence labelling for podocyte identification, confocal microscopy, optical disectors for podocyte counting (Fig. 3; Sterio 1984) and the Cavalieri estimator for estimating glomerular volume (Gundersen 1986). This approach provides estimates of the total numbers of podocytes per glomerulus, individual glomerular volumes and podocyte density per glomerulus, without any geometric assumptions. This method has been used to identify higher podocyte numbers in adults than children (Puelles et al. 2015) and mild podocyte depletion in association with older age and hypertension (Puelles et al. 2016a).

Fig. 3
figure 3

Use of the optical disector method to identify and estimate podocyte number. When cell nuclei are counted with optical disectors, they are counted at a unique moment (i.e., when they first “come into focus”). We start counting on Panel A (a–d). The nuclear stain 4′,6-diamidino-2-phenylindole (DAPI) was used to label all nuclei blue. Optical section 1 is used first to identify those cell nuclei that are in focus (stars). These nuclei are excluded from further analysis. Consecutive sections 1 μm apart (Optical sections 2, 3 and 4) are then used to identify nuclei that “come into focus”: these are the nuclei that are counted. In this case, we identified four nuclei that “come into focus” in optical sections 2–4; the numbers in yellow indicate the moment when the nucleus comes into focus, i.e., the moment when it is counted. The second step is to “turn on” the fluorescent channels to define whether these “new nuclei” are podocytes defined as Wilms’ Tumour 1 (WT-1)-positive cells and appearing in green in Panel B (e-h) or as other glomerular cells, e.g., endothelial cells that express von Willebrand Factor (vWF) and appear in red in Panel C (i-l). In this sequence, we identified only two podocytes (cells with nuclei 2 and 4). All images are shown at the same magnification with a as the reference. With permission from Annals of Anatomy

Full size image

Design-based stereology is considered to be the gold standard for estimating total nephron number in the kidney (Puelles et al. 2011; Puelles and Bertram 2015). Similarly, the disector/fractionator method has been identified as the preferred method for the evaluation of podocyte depletion when sufficient tissue is available for analysis (Lemley et al. 2013; Puelles and Bertram 2015). Unfortunately, the use of this approach has many disadvantages, including the need for serial sectioning of thick paraffin sections (approximately 14-μm-thick sections), extensive use of confocal microscopy and significant time for counting performed by a highly trained observer.

Whole glomerular analysis

So far, we have considered methods that provide “estimates” of podocyte depletion by using single or multiple sections through glomeruli. An alternative approach is exhaustive enumeration (Basgen et al. 2006; Lemley et al. 2013) in which every podocyte nucleus in whole glomeruli is counted. Traditionally, this method has involved the serial sectioning (at a thickness of 1–2 μm) of glomeruli embedded in Epon-araldite. Podocytes are identified based on their morphology and location outside capillaries, with each podocyte being seen in multiple sections further assisting their identification (Basgen et al. 2006; Nicholas et al. 2011). This method requires a high level of technical expertise and resources.

Significant progress has been achieved recently with development of a new podocyte counting method that employs a combination of optical clearing, immunolabelling and confocal microscopy to determine podocyte loss in whole glomeruli (Puelles et al. 2016b). Unlike exhaustive enumeration, which relies on physical sections, optical clearing enables the use of serial optical sections obtained by confocal microscopy to identify and count every podocyte nucleus in whole glomeruli. In addition to podocyte counts, this method provides estimates of average podocyte volume per glomerulus, podocyte nuclear and cytoplasmic volumes and an analysis of spatial relationships between podocyte nuclei within each glomerulus. Major advantages of this method include data accuracy and precision and the high speed of analysis. However, this method requires the significant employment of a confocal or two-photon microscope and the use of optical clearing in combination with immunofluorescence of thick slides of tissue (approximately 800 μm). A recent review is available on the use of optical clearing in kidney morphometry (Puelles et al. 2017).

Practical suggestions

Major points for deciding which of the above-mentioned methods to use are presented below. Problems to consider include the amount of kidney tissue available and access to imaging hardware and software. However, before we consider these practical issues, we need to contemplate the most important point when counting podocytes, namely their unambiguous identification.

Unambiguous identification of podocytes

Podocytes can be identified by using several approaches including morphology, e.g., light and electron microscopy (Bertram et al. 1992; White and Bilous 2004), immunolabelling, e.g., Wilms’ Tumour 1 ( WT-1; Puelles et al. 2015), TLE-4 (Venkatareddy et al. 2014) and p57 (Eng et al. 2015) and constitutive/inducible genetic labelling, e.g., podocyte-specific reporter lines (Berger et al. 2014; Wanner et al. 2014; Lasagni et al. 2015). All of these approaches have potential limitations.

As discussed above, the ability to identify podocytes with a high level of certainty also relies on section thickness and the imaging technique employed. Transmission electron microscopy enables easy podocyte identification, at least in normal glomeruli (Bertram et al. 1992). Confocal microscopy allows the visualization of thin optical sections without the need for physical sectioning and the use of multiple cell-specific markers (Richardson and Lichtman 2015). Furthermore, confocal microscopy facilitates the analysis of multiple, perfectly aligned, serial optical sections through each podocyte; this assists cell identification and also facilitates the generation of three-dimensional datasets, again enhancing our capacity to identify podocytes (Puelles et al. 2014).

Detecting mild levels of depletion

A recent study showed that conventional two-dimensional approaches are able to detect moderate/large levels of podocyte depletion but may miss mild levels of depletion (Puelles et al. 2016b). This illustrates an important point: the decision as to which method to use relies heavily on the experimental model (expected level of podocyte depletion) and potential impact of interventions (expected level of prevention or recovery). In our opinion, when high levels of accuracy and precision are essential in order to answer the research question, methods that minimize or negate the introduction of systematic bias should be used, including model-based stereology (e.g., Venkatareddy et al. 2014) and design-based stereology (e.g., disector/fractionator; Puelles et al. 2014) or whole glomerular analysis (Puelles et al. 2016b).

Available resources

Most laboratories physically cut 3–4 μm paraffin sections. However, cutting serial sections with a defined or even greater than normal thickness (14 μm) requires significant technical expertise. This favours the use of glomerular cross-sections and model-based stereology, whereas design-based stereology becomes more tedious and less appealing. The use of thick kidney slices (1–2 mm) bypasses the need for serial physical sectioning, which makes optical clearing an exciting possibility.

A critical point is access to imaging hardware and software. Whereas model-based stereological methods can usually be implemented with standard laboratory equipment, design-based and optical clearing methods may require access to expensive imaging hardware and software. Model and design-based stereological methods can be implemented by using light, electron, confocal and two-photon microscopy. The newly developed optical clearing approaches may require more specialized equipment (i.e., BABB compatible objective lens; Puelles et al. 2016b).

Towards automated quantitation of podocyte depletion

We are perhaps not too far from automating, or at least semi-automating, assessments of podocyte depletion. Image segmentation algorithms can now detect, count and size immunolabelled nuclei in histological sections (Venkatareddy et al. 2014). With the novel clearing/immunolabelling/confocal strategy that we developed to count podocyte nuclei in whole mouse glomeruli (Puelles et al. 2016b), we manually counted all podocytes in a single glomerulus within about 2 min. The use of commercially available three-dimensional–rendering software also allows the semi-automated quantification of podocyte number per glomerulus with remarkably similar results. A recent study by Klingberg et al. (2016) provided an example of the way in which current data analysis can be used for the automation of large datasets and the automation of morphological quantification (e.g., total glomerular number) in a whole kidney. Further refinement and development in this area will accelerate experimental studies and perhaps have utility in the analysis of renal biopsies for diagnostic purposes.

Conclusions

Podocytes have emerged as a crucial cell type in the pathogenesis of glomerular diseases. Several methods are available for quantifying podocyte depletion. Careful considerations should be made when deciding which method to utilize based on their theoretical and practical advantages and disadvantages. In the near future, automated podocyte counting may accelerate experimental studies and allow the inclusion of podocyte quantification in renal biopsies for diagnostic proposes. Finally, we encourage researchers to use proper terminology when referring to the analysed parameters in order to facilitate reproducibility and comparisons between studies.