Abstract
ABSTRACT. Podocyte loss and glomerular hypertrophy are associated with development of glomerulosclerosis, suggesting that there may be a maximal area for each podocyte in terms of its capacity to support and maintain the glomerular filter. This study hypothesized that exceeding this maximal threshold will result in mesangial expansion and glomerulosclerosis. It may therefore be useful to measure podocyte number, glomerular volume, and glomerular volume per podocyte in clinical biopsy samples. An approach that uses thick and thin histologic sections cut from paraffin-embedded tissue to measure Wilms’ tumor-1 protein–positive podocyte nuclear number and glomerular tuft area was studied. A rat model of aging has been used to track changes in glomerular podocyte number, glomerular volume per podocyte, and glomerular volume. Implications for clinical use of these variables are discussed. E-mail: rwiggins@umich.edu
There is increasing evidence from human biopsy material and experimental models that podocyte injury and reduction in the number of podocytes in a glomerular tuft may be an important factor determining the propensity for glomerulosclerosis (1–24⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓). If this is true, then it may be useful to routinely quantitate the number of podocytes in renal biopsy material and in experimental models to predict susceptibility to development of glomerulosclerosis and efficacy of treatment. Measurement of podocyte number has proved quite difficult because of the complexity of both podocyte and glomerular structure (18–26⇓⇓⇓⇓⇓⇓⇓⇓).
Although podocyte number appears to be a key variable, it is likely that podocyte number in relation to the glomerular basement membrane (GBM) area to be served by the podocyte may be more important. The relationship between GBM area and the podocyte can be approximated by relating glomerular tuft volume to podocyte number, and has been previously expressed as the “density” of podocytes or the glomerular volume per podocyte (GV/P) (21–28⇓⇓⇓⇓⇓⇓⇓). There are extensive data to support the relationship between large glomerular tuft size and glomerulosclerosis (reviewed in [29]), an observation that could well be related to a relative deficiency of podocytes.
We have attempted to develop a method that might be applicable for regular clinical use in renal biopsy material and experimental models of glomerular disease. We then used the method to examine changes in glomerular volume, podocyte number, and GV/P in the renal cortex of Brown Norway rats during normal maturation and aging.
Materials and Methods
Experimental Animal Model
Rats at a range of ages were obtained from the Aging Rodent Colony maintained by the National Institute for Aging. For these experiments, we used Brown Norway rats killed at 2 mo (adolescent), 5 to 7 mo (young adult), 15 to 17 mo (midage), and 32 mo (old). Rats were anesthetized by injection of xylazine and ketamine before perfusion of kidneys with phosphate-buffered saline for 2 min, followed by 2% PLP paraformaldehyde, lysine, and periodate in PBS for 8 min at a perfusion pressure of 125 mmHg. After perfusion, kidneys were quickly removed, and 3-mm slices of kidney were cut for fixation in formalin. All animal experimentation was conducted in accord with the NIH Guide for the Care and Use of Laboratory Animals.
Immunoperoxidase Histochemistry
Paraffin-embedded, formalin-fixed sections were cut at different thicknesses with a Reichert-Jung 820-II microtome (Cambridge Instruments, Nussloch, Germany) and mounted on polylysine-coated slides (Vector Laboratories, Burlingame, CA). After paraffin was removed from the sections and they were hydrated, they were treated with Retrieve-All-1 (Signet Laboratories, Dedham, MA) for 4 h at 90°C. The sections were then incubated with a polyclonal antibody to Wilms’ tumor-1 protein (WT-1 antibody; Santa Cruz Biotechnology, Santa Cruz, CA). Immunoperoxidase staining was performed according to the Vectastain ABC kit (Vector Laboratories). The immunoperoxidase detection system used diaminobenzidine (DAB; Sigma Chemical, St. Louis, MO). Two slides for each animal (one 3 μm and one 9 μm thick) were developed with WT-1 antibody and immunoperoxidase for podocyte counts and glomerular area measurements.
Morphometric Analysis
Images from 50 consecutive glomerular cross sections were collected for each histologic section by means of the Spot Advanced Software (Diagnostic Instruments; Silicon Graphics, Mountain View, CA). The area of every glomerular profile was measured manually by tracing the glomerular outline on a computer screen and calculating that area by computerized morphometry with the Metamorph Image System (Metamorph, Universal Imaging, Downington, PA). For each glomerular profile, the number of WT-1–positive nuclear profiles that met criteria for counting was evaluated. The criterion for counting a WT-1 nuclear profile was according to a designated size threshold. The Metamorph program allows each manually counted object to be assigned a number on the screen as it is counted.
Podocyte Nucleus Identification and Rationale for the Use of WT-1 as a Marker
The identification of podocytes for the purpose of counting cannot be reliably made by defining the cytoplasmic limits of the podocyte because of the complex processes and interdigitations between cells superimposed on the complexity of glomerular structure that make it difficult to identify where one cell begins and another ends. Therefore, one is limited to counting podocyte nuclei.
The fact that podocyte nuclei are WT-1 positive (30,31⇓), whereas nuclei of mesangial cells and endothelial cells are WT-1 negative, is potentially useful as a marker for podocyte nuclei. However, a complicating factor is the fact that under some circumstances, podocyte nuclei are not WT-1 positive (31,32⇓). Examples are in HIV nephropathy and other forms of collapsing glomerulosclerosis. Thus, any method that uses WT-1–positive nuclei as a marker system must keep in mind that some podocytes may be missed, although at the present time, there are no data to suggest that podocytes in the common forms of glomerulosclerosis such as diabetes and hypertension are WT-1 negative. In fact when considering the question of whether to use WT-1 as a marker, it is reasonable to conclude that WT-1 is an excellent marker of the normal mature differentiated podocyte. If a podocyte is WT-1 negative, then it is most likely not a normal, mature differentiated podocyte and therefore may well not be performing its job normally. Thus, by using WT-1 as a marker for podocyte nuclei, one can identify and quantitate normal mature podocytes and even podocytes with abnormal foot processes, such as is seen in minimal change disease. The population identified in this way may be different from those identified by means of an ultrastructural definition of a podocyte on the basis of its relationship to the GBM. Because we want to identify and count the population of normal mature podocytes, we have elected to use WT-1 as a convenient marker for further studies.
In counting WT-1–positive nuclear profiles, it is important to have some systematic method to help decide which nuclear profiles to count. We elected to use a large size cutoff because this eliminated small confounding variables, such as granules, as potentially countable events. In practice, we used a size restriction method where if the number 6 (1.2 × 2 μm) from the Metamorph Image analysis system could be fitted within the nuclear profile, it was counted, and if the number 6 was larger than the podocyte nuclear profile, then it was not counted. Thus, shaving nuclear sections or small transsections of podocyte nuclear profiles were not counted because doing so tended to result in a wider range of variation between individual observers who had different criteria for inclusion.
Peroxidase immunostaining that used the DAB substrate gives a brown deposit. However, we found that when viewed by phase contrast, the brown deposit became a golden color, which facilitated WT-1–positive nuclear identification and counting (Figure 1). Therefore, for the experiments described, we photographed glomeruli under phase contrast at ×20 magnification and stored the image on disk. For purposes of nuclear counting and measurements, we enlarged and then counted WT-1–positive nuclei and measured glomerular circumferences with the calibrated Metamorph program.
Figure 1. Photomicrograph of a 3-μm section showing a glomerular tuft developed by using Wilms’ tumor-1 protein (WT-1) immunoperoxidase and viewed by phase-contrast microscopy. The immunoperoxidase product appears as a gold refractile material that makes podocyte nuclear profiles easier to count.
Measurement of Actual Section Thickness
We measured actual section thickness by cutting five sections at each microtome setting and then layering them together in melted paraffin to form a sandwich-like arrangement with sections separated by a layer of paraffin to form a new paraffin block. This block was then mounted and resectioned at right angles to the plain of the previous sections. The thickness of these hematoxylin-and-eosin–stained sections was then measured by the calibrated Metamorph system.
Estimation of What Proportion of Nuclei Were Double Counted
The podocyte nuclei and glomerular tuft profile of each serial section were traced from the computer screen onto a clear plastic sheet (Highland Transparent Film, Austin, TX). The plastic sheets were then superimposed on one another so that the profiles could be matched to assess which individual podocyte nuclei were present in more than one glomerular tuft profile. The podocyte nuclei represented in more than one slice were marked and then counted to estimate the number of previously uncounted nuclei on each section.
Results
Serial Section Experiment
As a first step toward measuring podocyte number per glomerulus, we performed serial section analysis of renal cortex from 2-mo-old Brown Norway rats. The microtome was set to cut 3-μm sections, which were then developed for WT-1 immunoperoxidase staining to identify podocyte nuclei, as described above. Glomerular tuft area was measured by morphometry, and podocyte nucleus number was counted in each section for all of the serial sections for 12 individual glomeruli from a single kidney. The results for the 12 serially sectioned glomeruli were assembled together so that each section was arranged symmetrically around the midglomerular section (designated as section 0). The neighboring sections on either side of the midglomerular section were designated −1, −2, −3, etc., or +1, +2, +3, etc.
The mean glomerular tuft area from 12 serially sectioned glomeruli is shown in Table 1, column B, with data derived from the midglomerular section shown in bold. These data (mean ± 1 SD) are also plotted in Figure 2 and show a symmetrical distribution of diminishing areas as one moves away from the midglomerular section, as would be expected for serial sectioning through a spherical object. We know that the glomerulus can be modeled as a sphere, although individual glomeruli may not be spherical (27,28⇓). Table 1, column C, shows the mean glomerular radius derived from these areas for each glomerular section away from the midglomerular section. The mean maximal radius of the tuft for the 12 glomeruli was 41.9 μm. Column D shows the calculated volume of a hypothetical spherical glomerular tuft derived by using the mean maximal radius.
Table 1. Mean values from 12 serially sectioned glomerulia
Figure 2. The serial section slices numbered consecutively as they move away from the midglomerular slice are shown on the x-axis. The glomerular tuft area or calculated value for glomerular volume per podocyte (GV/P) ± 1 SD is shown on the y-axis. The glomerular area forms a parabolic curve, as expected for slices made through a sphere. The values for GV/P are relatively constant across the glomerular profile, with somewhat higher values tending to occur toward the midglomerular sections and lower values toward the periphery of the glomerulus, probably reflecting a tendency of podocytes to be arranged around the periphery of the glomerular tuft. However, this difference was not statistically significant.
The mean WT-1–positive nuclear counts for the 12 glomeruli in the serial section experiment are shown in Table 1, column E. The sum of the mean number of WT-1–positive nuclei counted in each section gives a total WT-1–positive nuclear number of 204 per glomerular tuft (mean value of two independent investigator counts). These data are shown at the foot of column E.
We now have measured the mean area of each glomerular cross section (Table 1, column B). We know the designated thickness of each section cut at 3 μm on the microtome, and we know the mean number of WT-1–positive nuclei counted in each glomerular cross section (column E). We can therefore calculate a value for glomerular volume per WT-1–positive nucleus for each cross section (column F), designated by GV/P. These data are shown in Table 1, column F. Because we know the mean glomerular volume (column D), we can also calculate the mean number of podocytes per glomerulus by dividing the glomerular volume by the value for GV/P. This result is shown in two ways in Table 1. The podocyte number per glomerulus value derived for the midglomerular section is shown in column G as 316 podocytes per glomerulus.
A second way of calculating the number of podocytes per glomerulus by using the mean GV/P value derived from all of the sections (−9 to +9; column F) gave a value of 337 podocytes per glomerulus (shown at the foot of column G). This is the method used for calculating podocyte number per glomerulus in a previous publication (19) and gives a value that is significantly larger than the value of 204 podocytes per glomerular tuft obtained by adding up the mean number of WT-1–positive nuclei in every section (column E). There is therefore a discrepancy between the number of counted WT-1–positive cells and the derived value.
In assessing the values for GV/P (Table 1, column G, and Figure 2), it appears that the values are fairly uniform across the glomerular tuft, although they tend to be higher toward the midpoint of the glomerulus. This is most likely because in the rat glomerulus, the podocytes tend to congregate on the periphery of the glomerular tuft. Thus, a cross section at the midpoint of the glomerulus would tend to have relatively fewer podocytes per unit volume and therefore a relatively higher value for GV/P than would a section taken toward the periphery of the glomerular tuft. This emphasizes the importance of evaluating all glomerular cross sections rather than just the midpoint sections to calculate mean podocyte numbers. Therefore, it would be optimal to use the mean value for GV/P for all sections through the glomerular tuft. This concern may be less in the human glomerulus where the structure is more convoluted and podocytes are more evenly distributed throughout the glomerular tuft. It is likely to be even more important in the mouse where podocytes are distributed at the periphery of the glomerular tuft.
Effect of Measuring Actual Section Thickness
One potential explanation for the discrepancy noted above would be that the actual section thickness was different from the assumed section thickness. We therefore measured the actual thickness of sections cut at various different microtome set points (see Materials and Methods). The results for the microtome setting and the actual section thicknesses were as follows: setting 3 μm measured 3.9 ± 0.2 μm; setting 5 μm measured 5.5 ± 0.2 μm; setting 7 μm measured 7.8 ± 0.1 μm; and setting 9 μm measured 10.2 ± 0.3 μm. Therefore, each microtome setting underestimated the measured section thickness by between 0.5 and 1.2 μm. This result is very similar to the previously reported data, which found that the microtome setting underestimated the actual thickness of a paraffin-embedded section by approximately 1 μm (25).
From the above analysis, we find that a microtome setting of 3 μm cuts a section that is actually 3.9 μm thick. Therefore, all calculations of numbers of WT-1–positive nuclei per glomerulus and GV/P developed from cross-sectional data will need to take this into account. For the GV/P calculations, the volume of each glomerular disc is calculated by multiplying the measured tuft area by the value for slice thickness and then dividing this number by the counted number of WT-1–positive nuclear profiles. The calculated value for glomerular tuft volume (Table 1 column D) divided by GV/P gives the value for podocyte number per glomerular tuft. In the serial section experiment, we learned that the slice thickness was actually 3.9 μm rather than the 3-μm value we had previously assumed. We therefore recalculated the values for GV/P and podocyte number per glomerulus with this information. The data are shown in Table 1, columns H and I. Again, the data for the WT-1–positive nucleus number are calculated with the midglomerular section (column I, giving a value of 243 WT-1–positive nuclei per glomerular tuft) and with the mean value for GV/P (shown at the foot of column I, to give a value of 259 WT-1–positive nuclei per glomerular tuft). These new values for WT-1–positive nuclei per glomerular tuft are significantly lower than the previous values as a result of this correction for the real slice thickness, but is still significantly higher than the value of 204 WT-1–positive nuclei per tuft obtained by actually counting the WT-1–positive nuclei in every section (column E).
Problem of Missing Tissue
It required a mean number of 17.8 sections to cut completely through the glomerular tuft for the 12 glomeruli shown in Table 1. The mean maximal diameter of the glomeruli sectioned was 41.9 × 2 = 83.8 μm. Thus, we can estimate that the glomeruli were sliced at an average thickness of 4.7 μm in this set of serial sections (83.8/17.8). There was therefore a discrepancy between the measured section thickness (3.9 μm) and the estimated thickness (4.7 μm), which would have an important effect on the calculation of podocyte number per glomerular tuft (approximately a 20% error). We therefore examined the potential causes for an average loss of 0.8 μm of tissue thickness with each section, which included the following: (1) some sections were not collected and were lost during the serial sectioning process, and therefore they are not accounted for in the calculation (we think this did not happen); (2) serial sections during the serial sectioning process were actually thicker than the 3.9 μm we measured (we remeasured the average thickness in another experiment and found it to be 3.9 ± 0.2 μm [n = 7], as described previously); (3) the process of cutting the section resulting in loss of a 0.8-μm layer of tissue with each slice; and (4) the glomerular tuft was asymmetrical so that the above calculation gives an incorrect conclusion. It is possible that a combination of factors cause this discrepancy. However, this is a real issue that must be taken into account when calculating WT-1 nuclear number per tuft.
Therefore, in Table 1, column J, we have recalculated the values for GV/P by using a value for each section thickness of 4.7 μm. This new calculation has then been used to calculate the values for WT-1–positive nuclei per glomerular tuft for the midglomerular section (202 WT-1–positive nuclei) and the mean GV/P value (215 WT-1–positive nuclei). These values are similar to the actual counted total number of WT-1–positive nuclei of 204 per tuft (column E).
The Podocyte Nucleus Overcounting Problem
The above values for podocyte nuclear counts assume that each nucleus is counted only once. However, the reality is that podocyte nuclei are relatively large (8 to 12 μm long in the 2-mo-old Brown Norway rat). This means that when sections are cut at intervals of 4.7 μm and are 3.9 μm thick and nuclei are arranged in a random orientation, many nuclei will be transected more than once, and nuclear profiles will appear in more than one serial section. When this occurs, podocyte nuclei will be counted more than once, and this will result in an inflated number of estimated podocytes per glomerulus.
We therefore estimated what proportion of podocytes were counted twice by a manual method (see Materials and Methods). The total number of nuclei in each profile was counted and added together to give a mean value 201 ± 12 podocyte nuclear profiles per glomerulus. However, when we added up the total number of nuclei that had not already been counted on the previous section, this gave a mean value of 120 ± 8 podocyte nuclear profiles per glomerulus. In this way, the true mean number of podocyte nuclei per glomerulus was estimated for the 12 serially sectioned glomeruli from an individual animal. Thus, from this experiment, we found that the proportion of podocyte nuclei counted twice was 40.2% ± 3.2% for 2-mo-old male Brown Norway rats in serial sections that were designated to be 3.0 μm thick but were actually 3.9 μm thick. If the section thickness were to change or the size of a podocyte nucleus were to change, then this correction factor would no longer be applicable.
Let us return to the serial section experiment shown in Table 1. We now know that we can correct the podocyte counts for double counting as outlined above. In Table 1, column J, we have multiplied the original podocyte nuclear counts by a correction factor of 0.598 (derived from finding that 40.2% of podocyte nuclei were double counted). We have calculated a corrected value for the mean volume of each 4.7-μm-thick disc sectioned through the glomerular tuft (column L). We have then recalculated the values for GV/P by using the corrected podocyte values (column M). By use of the value for glomerular volume (column D) and the GV/P data, we have calculated the podocyte number per glomerulus. Again, we have calculated a value by using the midglomerular section (giving a value of 120 podocytes per glomerulus, as shown in column N) as well as by using the mean GV/P values for all sections (shown at the foot of column K) to derive a value of 128 podocytes per glomerulus (shown at the foot of column N). This calculation gives values similar to the corrected value for the counted number of podocytes per glomerular tuft (column J). Thus, we believe that most of the variables have been accounted for by these calculations and adjustments.
We conclude from this serial section experiment that there are a number of potential pitfalls for estimating podocytes number per glomerulus. Factors such as true section thickness, lost tissue, and nucleus double counting may result in overestimation of podocyte numbers by substantial amounts. Limiting the analysis to midglomerular sections may result in underestimation of podocyte numbers, even if the above factors have been taken into account. A further concern is the fact that the extent of nuclear double counting might vary from individual to individual or in pathologic processes if nuclear size were to change. We know that podocyte size changes remarkably with age (25), and therefore it seems likely that podocyte nuclear size could also change under some circumstances. If that were the case, then a single correction factor for nuclear double counting could not be used between samples. We therefore need simple and more reliable methods for calculating both mean glomerular volume as well as GV/P so that more accurate estimates for podocyte numbers can be reliably obtained.
Estimating Maximal Glomerular Tuft Diameter and Mean Glomerular Tuft Volume
Although glomeruli are not all spherical, a spherical model gives a reasonable approximation of geometric aspects of the glomerulus, such as diameter and volume (27,28⇓). To derive the mean glomerular volume, we must estimate mean maximal glomerular tuft diameter. One approach is to serially cut through renal tissue to identify the maximum glomerular diameter for a representative number of glomeruli, as was done in the above serial section analysis. This process is laborious. We have therefore tested a stereologic approach that was based on the Weibel monograph (33). The stereologic formula assumes that randomly arranged spheres of equal diameter are transected by a plane. The equation used to derive the formula and proven in the Weibel monograph states that the maximum sphere diameter is equal to 4/π × the mean measured diameter of identical spheres randomly transected by the plane. To validate this approach in our hands for glomeruli, we used the data generated by serial sectioning (above) as the gold standard.
In this experiment, we measured the average tuft diameter of 12 glomeruli by serial sectioning (Table 1). The measured mean maximal diameter was 83.8 μm. To test the Weibel formula, we assumed that the glomerular tufts are of equal diameter and that the histologic section cuts through these tufts randomly as would a random plane. We measured the diameter of 129 consecutive glomeruli in the same histologic sections used for the serial section experiment and calculated values for mean glomerular diameter. The measured mean glomerular tuft diameter was 67.2 μm. Thus, applying the formula, 4/π × 67.2 = 85.6 μm. This value was similar to the value of 83.8 μm that we obtained by direct measurement of the mean maximal diameter of 12 glomerular tufts by serial sectioning. Thus, we conclude that this formula works reasonably well for glomeruli in histologic sections and can be used to calculate the mean glomerular tuft volume in a sample of renal cortex.
Approach to Calculating GV/P by Use of Different Thicknesses of Histologic Section
The number of podocyte nuclei present in a glomerular tuft cross section increases as the slice thickness increases. By making measurements of podocyte nucleus number and glomerular tuft area in two sections from the same piece of tissue but of different known thickness (for example, cut at microtome settings of 3 and 9 μm), one can derive a value for GV/P. The difference in podocyte count between thin and thick sections is directly proportional to the difference in slice thickness.
There are several possible ways to derive GV/P from these data. The method we have used is as follows. The data are expressed as the relationship between the two directly measured variables podocyte number and glomerular area to derive the mean podocyte number per glomerular tuft area (P/GA) for both thin and thick sections. The difference between the P/GA of the thick and thin sections (P/GAthick − P/GAthin = P/GAΔ) is directly related to the actual difference in section thickness (6.3 μm). By dividing the known difference in section thickness by the value for P/GAΔ, we can derive GV/P.
In essence, the potential problems caused by changing podocyte nuclear size, loss of tissue, and double counting are nullified because these problems are the same for thin and thick sections and can therefore be eliminated by subtraction. Because every glomerular profile is assessed, there is no bias of selection, and there is no bias caused by examining only midglomerular profiles.
Derivation of Podocytes per Glomerular Tuft Volume
In the above experiment, 50 consecutive glomerular profiles were assessed for tuft area and podocyte nuclear number. From these data, the mean glomerular volume is derived with the Weibel formula from the areas of the 50 consecutively measured glomerular profiles. The value for mean GV/P obtained from the 50 glomerular profiles was also derived as outlined above. The mean podocyte number per tuft is equal to the mean glomerular volume divided by GV/P. The values obtained by using the thick- and thin-slice method were similar to those obtained by using the serial section method for the same piece of kidney. Respectively, the thin/thick method and serial section methods resulted in values as follows: GV/P of 2488 and 2403; podocytes per glomerular tuft of 128 and 128; and a glomerular volume of 318,464 and 307,397 μm3.
Reproducibility of the Estimates for Glomerular Volume, GV/P, and Mean Podocyte Number per Glomerular Tuft Measured in Inner and Outer Cortex
If the method is to be useful, then it must be reproducible. To examine this point, we cut both a thin (designated 3 μm) and a thick (designated 9 μm) section from five different tissue blocks all derived from the same pair of kidneys. These pairs of thin and thick sections were each cut on different days but on the same microtome so as to simulate realistic conditions used in a single laboratory from day to day. The immunoperoxidase staining of podocyte profiles was also done on a different day for each pair of sections. Because the sections were all from the same animal, we would expect that the results would be similar or identical within normal biologic variation. We expressed the data for the superficial cortical (SC) and juxtamedullary (JM) glomeruli separately. The data shown below are the mean ± 1 SD with the SD shown as a percentage of the mean value in parentheses (coefficient of variation). The results were as follows for SC and JM glomeruli, respectively, with P values shown for comparison between the SC and JM values for each parameter: glomerular volume 348,239 ± 36,860 (10.6%), 522,930 ± 53,281 (10.2%), P < 0.01; GV/P 2696 ± 492 (18.3%), 3066 ± 590 (19.3%), P = 0.3; and podocyte number per glomerular tuft 133 ± 31 (23.0%), 175 ± 32 (18.2%), P = 0.07. The coefficient of variation for the glomerular tuft volume measurement was quite reproducible at 10%, whereas that for podocyte number and GV/P were less reproducible at approximately 20%. It is likely that the ranges of reproducibility represent a combination of methodologic and biologic variation within renal cortex.
Comparison of Glomeruli from Inner and Outer Cortex
We evaluated glomeruli from the inner cortex and the outer cortex separately because of previous reports showing that these two glomerular populations differ in size (25,26⇓). The results are shown above. We confirmed that glomeruli in the inner cortex were statistically larger than glomeruli from the outer cortex by a factor of 1.5 (P < 0.01). The values for podocyte number were also different by a factor of 1.3, although this difference did not quite reach statistical significance (P = 0.07). Estimates of GV/P were very similar between inner and outer cortex. Thus, larger glomeruli in the inner cortical area tended to have relatively more podocytes, as has previously been reported (25,26⇓). Therefore, the relationship between a podocyte and the territory it serves as represented by the value of GV/P tended to remain the same within a single kidney at one point in time. This result suggests that a focus on GV/P would not be dependent on the sampling site and therefore might be a useful approach in practice.
Number of Glomeruli Required to Make Reproducible Measurements of Glomerular Volume, Podocyte Number, and GV/P
If the method is to be useful for renal biopsy samples, then it would need to be usable with relatively few glomeruli (15 to 20 per biopsy sample). We therefore evaluated the variability of the method by arbitrarily assessing data for different numbers of glomeruli (50, 25, or 12) for ten different samples. The data are expressed with the SD as a percentage of the mean value. For 50 glomeruli, the variation was as follows: GV 12.5%, GV/P 19.5%, and podocyte number 18%. For 25 glomeruli, the variation was as follows: GV 15%, GV/P 21.5%, and podocyte number 22.5%. For 12 glomeruli, the variation was as follows: GV 18%, GV/P 26.5%, and podocyte number 24.5%. The most reproducible value was for glomerular volume, whereas the variation for podocyte number and GV/P were higher. When there were only 12 glomeruli to be evaluated, the numbers were 18% for the glomerular volume and approximately 25% for the podocyte number and GV/P. Thus, as expected, number of glomeruli evaluated is important for obtaining a reproducible result, although reasonable reproducibility can still be obtained with relatively few glomeruli.
The potential utility of these estimates can be seen when comparing the error of the method to the wide ranges found in normal kidney during aging (Figure 3). For example, glomerular volume varied from 0.2 to 1.2 × 106 μm3 in normal glomeruli from adolescence to old age. A variation of 10% in the mean glomerular volume measurement is quite acceptable in this context. In contrast, the podocyte number per glomerulus varied between 124 and 173 per glomerulus. This age range of approximately 40% is relatively small. A variation of 20% in the measurement of podocyte number is quite large in comparison. Finally, GV/P varied between 2.8 and 7.8 × 103 μm3 (three- to fourfold) in normal aging kidneys. A variability of 20% to 25% in this setting is also potentially acceptable. We conclude that a focus on glomerular volume and GV/P may be most useful given the level of reproducibility of the assay system, although podocyte numbers could be estimated if the number of glomeruli available for evaluation was large.
Figure 3. Values for mean glomerular volume (Glom Vol) (A), podocyte number per mean glomerular tuft (B), and glomerular volume per podocyte (GV/P) (C) in Brown Norway rats as they age from 2 to 32 mo while kept on an ad libitum diet. The values shown are the mean values ± 1 SD calculated from assessing superficial cortical and juxtamedullary glomeruli (n = 5). Note the wide range of values for glomerular volume and GV/P with age.
Changes in Podocyte Number, Glomerular Volume, and GV/P During Aging in the Brown Norway Rat
We used the above approach to evaluate adolescent (2 mo), young (6 mo), midaged (18 mo), and old (32 mo) Brown Norway rats (five animals per group) to determine the variation associated with differences in age. Glomeruli in the inner and outer cortical regions were separately assessed and the mean data presented in Figure 3.
We found that glomerular volume increased markedly with age, as has been well documented previously (25–28,34–37⇓⇓⇓⇓⇓⇓⇓). The podocyte number increased significantly from 2 to 6 mo of age and then trended up to peak at 18 mo (middle age). There was a suggestion that podocyte numbers decreased in older rats, although this trend did not reach statistical significance in these experiments. The values for GV/P steadily increased into old age. The continued increase in GV/P in older rats was a result of the combination of a continued increase in glomerular tuft volume and a reduction in podocyte number.
Discussion
There are increasing data to support the concept that podocyte depletion itself or in association with increased glomerular volume may be a key element determining development of progressive glomerulosclerosis and thereby, progression to end-stage renal disease (1–24⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓). Recent observations in humans describe a striking relationship between podocyte depletion and progression to end-stage renal disease in types 1 and 2 diabetic glomerulosclerosis and in IgA nephropathy (20–24⇓⇓⇓⇓). These reports are based in turn on an extensive body of literature in experimental animals that provides a strong experimental underpinning for this conclusion (2–19⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓). These experimental studies are well supported by a long-standing clinical literature describing a correlation between glomerulosclerosis and podocyte injury (1). Thus, taken together, there is a growing body of literature that supports the conclusion that progressive podocyte depletion is associated with glomerulosclerosis and progression to end-stage renal disease.
Podocytes are complex epithelial cells that are particularly susceptible to oxidant injury (38–43⇓⇓⇓⇓⇓). Mature differentiated podocytes have limited ability to divide and therefore, once lost, may not be readily replaced (14,16,44,45⇓⇓⇓). The interdigitating foot processes require large amounts of lipid membrane and actin-mediated cytoskeleton to create the foot process structure that counteracts hydrostatic pressure and facilitates filtration (12,16,46–48⇓⇓⇓⇓). In this sense, a reduction in podocyte “density” below a critical number in relation to GBM surface area to be served could be a key factor that might trigger changes in behavior of mesangial and other glomerular cells resulting in glomerulosclerosis (20). It is difficult to estimate GBM area in two-dimensional histologic sections, although it is relatively easy to measure glomerular tuft area. Glomerular tuft area can be used to calculate glomerular tuft volume, which can act as a surrogate for GBM area, although as the glomerular architecture becomes more disturbed, this relationship may not hold up well, as has been pointed out for diabetic glomerulosclerosis (23). However, in general, we would expect that a variable that incorporated the relationship between both podocyte number and GBM surface area such as GV/P would potentially be a useful measure of the degree of podocyte reserve. We have sought a method that would allow these variables to be measured on paraffin-embedded renal biopsy samples, keeping in mind that the values derived are affected by processing artifacts such as shrinkage during paraffin embedding (49) that render them relative rather than absolute. Nevertheless, they may prove to be clinically useful.
As outlined in the Results above, we elected to use WT-1 immunoperoxidase staining as a marker for podocyte nuclei. A complicating factor is the concern that under some circumstances podocyte nuclei are not WT-1 positive. Examples are in HIV nephropathy and other forms of collapsing glomerulosclerosis (31). Thus, any method that uses WT-1–positive nuclei must contend with the concern that some podocytes that have differentiated to become WT-1 negative may be missed. When considering the question of whether to use WT-1 as a marker, we concluded that WT-1 is an excellent marker of the normal, mature, differentiated podocyte. The population of podocytes identified in this way may be different from those identified by an ultrastructural definition of a podocyte on the basis of its relationship to the GBM. For the purposes of discussion, we have used the terms “WT-1–positive nuclei” and “podocyte nuclei” interchangeably, but this requires us to keep firmly in mind what we mean by the term “podocyte” in this setting (a glomerular cell with a WT-1–positive nucleus).
There are some potential advantages to the method of evaluating podocytes outlined in this report. These include the following. (1) Variations in podocyte nuclear size and shape are likely to occur at different age and pathologic conditions. Thus, any method that assumes podocyte nuclear size is constant will probably result in errors. For the current methodology, differences in podocyte nuclear size are not critical to the estimation because this factor will be neutralized by the subtraction approach. (2) All glomeruli are evaluated, thereby eliminating glomerular selection bias that might occur were particular glomeruli or parts of glomeruli selected for analysis. (3) A bias toward overcounting, which occurs in some methods as a result of underestimating effective section thickness, is not a problem with the subtraction methodology. (4) A large number of glomerular tuft cross sections can be rapidly evaluated. (5) The method does not require complex formulae that are difficult to understand and where limitations of their application are not immediately obvious.
The variability of the method was quite large. It was approximately 10% for the glomerular volume calculations and approximately 20% for the podocyte counts and GV/P estimation when 50 glomeruli were evaluated. When small numbers of glomeruli are used (n = 12), these values were higher. However, when viewed in the context of the changes in glomerular volume and GV/P that occur in the rat during its normal lifespan, these variations are within a usable range for the assay. Thus, the normal glomerulus varies in volume by fivefold, and GV/P also ranges over three- to fourfold with normal aging, with similar changes probably occurring during pathologic processes. In this setting, an experimental variation of 25% is probably acceptable. In contrast, the podocyte counts varied much less with age (approximately 40%), so an experimental variation of 20% is large in this context. Under pathologic conditions, there may be even larger differences in glomerular volume, GV/P, and podocyte numbers.
The results from Brown Norway rats reinforce the striking increase in glomerular size with age documented by many other investigators (25–28,34–37⇓⇓⇓⇓⇓⇓⇓). The fact that podocyte number does not increase proportionately to glomerular volume raises the question of whether this may lead to a critical paucity of podocytes per glomerular volume to be served (“density”) at some point in the aging process. This relationship might play a role in the remarkable increase in chronic renal failure associated with the aging population in humans (50).
We previously used a method for estimating podocyte number that examined only midglomerular profiles and that did not correct podocyte nuclear counts for double counting and where we assumed the section thickness to be 3 μm (19). By means of this approach, we defined a value for glomerular area per podocyte of 1500 to 1600 μm2 to be the approximate threshold above which glomerulosclerosis would be expected to occur. This threshold value would correspond to a GV/P value of approximately 10 to 12,000 μm3 when the current methodology is used. The maximal estimate for mean GV/P seen even in 32-mo-old Brown Norway rats in the current experiment was approximately 8000 μm3 and therefore did not reach the magnitude where glomerulosclerosis might be expected to occur on the basis of our previous report. Correspondingly, we did not observe glomerulosclerosis, even in the 32-mo-old Brown Norway rats. The above studies were performed in rat kidneys. Application to human tissues would require validation of the methodology in that setting.
Acknowledgments
This work was performed with the support of United States Public Health grants P60 DK20572, DK39255, and DK46073. Rats were provided through the Core for Aging Rodents, supported by grant NIH AG08808. We thank Richard Miller, PhD, MD, for advice and support. This work was supported by grants from the NIH (DK39255 and DK46073).
- © 2003 American Society of Nephrology
References
Jump to section
More in this TOC Section
Cited By...
- Systemic AAV10.COMP-Ang1 rescues renal glomeruli and pancreatic islets in type 2 diabetic mice
- PEPITEM/Cadherin 15 Axis Inhibits T Lymphocyte Infiltration and Glomerulonephritis in a Mouse Model of Systemic Lupus Erythematosus
- Podocyte-Specific Loss of Krüppel-Like Factor 6 Increases Mitochondrial Injury in Diabetic Kidney Disease
- Thioredoxin-Interacting Protein Deficiency Protects against Diabetic Nephropathy
- The Metabolic Syndrome and Microvascular Complications in a Murine Model of Type 2 Diabetes
- Valsartan slows the progression of diabetic nephropathy in db/db mice via a reduction in podocyte injury, and renal oxidative stress and inflammation
- Estimating Podocyte Number and Density Using a Single Histologic Section
- Podocyte-Specific GLUT4-Deficient Mice Have Fewer and Larger Podocytes and Are Protected From Diabetic Nephropathy
- Mammalian Target of Rapamycin Regulates Nox4-Mediated Podocyte Depletion in Diabetic Renal Injury
- Estimation of Glomerular Podocyte Number: A Selection of Valid Methods
- Reversibility of Structural and Functional Damage in a Model of Advanced Diabetic Nephropathy
- DGKE Variants Cause a Glomerular Microangiopathy That Mimics Membranoproliferative GN
- Growth-Dependent Podocyte Failure Causes Glomerulosclerosis
- AMP-activated Protein Kinase (AMPK) Negatively Regulates Nox4-dependent Activation of p53 and Epithelial Cell Apoptosis in Diabetes
- BTBR Ob/Ob Mutant Mice Model Progressive Diabetic Nephropathy
- Mechanisms of Podocyte Injury in Diabetes: Role of Cytochrome P450 and NADPH Oxidases
- Urine Podocyte mRNAs Mark Progression of Renal Disease
- Podocyte-Specific Overexpression of the Antioxidant Metallothionein Reduces Diabetic Nephropathy
- Differential Expression of Nephrin According to Glomerular Size in Early Diabetic Kidney Disease
- {alpha}-Actinin-4 Is Required for Normal Podocyte Adhesion
- Podocyte Depletion Causes Glomerulosclerosis: Diphtheria Toxin-Induced Podocyte Depletion in Rats Expressing Human Diphtheria Toxin Receptor Transgene
- Podocyte Hypertrophy, "Adaptation," and "Decompensation" Associated with Glomerular Enlargement and Glomerulosclerosis in the Aging Rat: Prevention by Calorie Restriction
- Permanent Genetic Tagging of Podocytes: Fate of Injured Podocytes in a Mouse Model of Glomerular Sclerosis
- Urinary Podocyte Loss Is a More Specific Marker of Ongoing Glomerular Damage than Proteinuria
- Development of Late-Stage Diabetic Nephropathy in OVE26 Diabetic Mice