Quantitative Morphology in Kidney Research
February 13-14, 2012 Conference Videos

Overview: Principles of Stereology and Applications to Research Settings
John Bertram, Monash University, Australia

Video Transcript

00:00:00,000 --> 00:00:08,033
JEFFREY KOPP: We’ve selected as our first speaker John Bertram to give us an overview of principles of stereology, and this was a fairly

00:00:08,033 --> 00:00:16,499
easy choice. John, obviously to those of you who know, is a central figure in this field. If you don’t know, he’s Head of the Department of

00:00:16,500 --> 00:00:25,166
Anatomy and Developmental Biology at Monash University in Australia. His research focuses on exactly this: kidney development and particularly

00:00:25,166 --> 00:00:32,432
on the regulation of nephron number during development and the implications of that low nephron number for kidney and cardiovascular

00:00:32,433 --> 00:00:41,366
disease, which is a topic that will be picked up by our second speaker, Wendy Hoy. He’s also Chair of the Healthy Start in Life Initiative, suggesting a

00:00:41,366 --> 00:00:50,699
public health dimension to the kinds of things that he’s doing. So with that, I’ll welcome him to the podium.

00:00:50,700 --> 00:00:59,233
JOHN BERTRAM: Thank you very much, Jeffrey, for your introduction and thank you to you and to Kevin for the invitation to speak at this meeting. I

00:00:59,233 --> 00:01:09,366
think the timing is very important and I’m really looking forward to where we get to over the next 24 hours. So, I’ve been asked to talk on this

00:01:09,366 --> 00:01:19,599
topic—principles of stereology and applications to research settings—and here’s an outline of my talk. I want to start with a definition or, in fact, a

00:01:19,600 --> 00:01:28,566
few definitions of stereology and then move on to the issue of particles and profiles and problems therein, and we’ll hear about this a fair bit over

00:01:28,566 --> 00:01:39,432
the next day. I’ll then consider briefly the fact that stereology provides estimates and think about some of the estimators and then reasonably

00:01:39,433 --> 00:01:47,699
quickly go over some of the traditional kidney parameters of interest—whether they give us absolute values or relative values—and we’ll then

00:01:47,700 --> 00:01:55,433
consider some traditional stereological approaches and the so-called model-based methods for estimating number. Stereology

00:01:55,433 --> 00:02:04,833
changed in 1984 with the development of new design by stereological approaches, and so I’ll then consider the disector approach, the Cavalieri

00:02:04,833 --> 00:02:12,833
Principle, and the fractionator experimental design. I think there are some very exciting developments happening with the use of

00:02:12,833 --> 00:02:21,333
magnetic resonance imaging for counting and sizing glomeruli, and I’ll touch upon that briefly. I’ll talk quickly about dealing with rare events and

00:02:21,333 --> 00:02:32,499
rare cells in the kidney, in particular, our new efforts to try to quantify numbers of glomerular stem cells, and at the end I’ll really try to do the

00:02:32,500 --> 00:02:40,700
impossible and suggest we might think about trying to move towards a consensus for how to do kidney stereology. The lung people have done

00:02:40,700 --> 00:02:48,733
it in the last two years and I think it would be wonderful if we could do that also in the nephrology community. I’ll say at the outset I’m not

00:02:48,733 --> 00:02:59,999
going to mention every kidney stereological parameter that’s possible to measure. I may not mention your favorite parameter; I’m sorry. I may

00:03:00,000 --> 00:03:08,733
not mention your favorite technique; apologies for that, too. Please don’t take it any way personal. I’m trying to give an overview of this pretty

00:03:08,733 --> 00:03:18,266
extensive list here that hopefully will provide some foundation for the rest of the workshop. So, I want to start with definitions and this issue

00:03:18,266 --> 00:03:29,699
of particles, profiles, and problems. So, it was Ewald Weibel in his book in 1979 who gave us this definition of stereology. The underlinings are

00:03:29,700 --> 00:03:38,066
mine, the words are his. He said stereology was a “body of mathematical methods relating three-dimensional parameters defining the structure to

00:03:38,066 --> 00:03:47,399
two-dimensional measurements obtainable on sections of the structure.” There are the words. If we look at that schematically, perhaps he was

00:03:47,400 --> 00:03:55,633
saying something like this. This is the real world of three-dimensional tissue. Think of it as a kidney with glomeruli in it or think of it as a glomerulus

00:03:55,633 --> 00:04:05,766
with cells in it, whatever you like, and in stereological-speak we call these three-dimensional objects particles. But of course, to

00:04:05,766 --> 00:04:13,199
understand something about this tissue, what we’ve been doing for more than a century is cutting sections and producing two-dimensional

00:04:13,200 --> 00:04:22,766
samples, if you will, or two-dimensional sections where now the particles are seen as two-dimensional samples which we call profiles. So,

00:04:22,766 --> 00:04:32,066
I’m not going to talk down sectioning as a bad thing to do. It’s still the histological disectioning, histopathology still relies on this to a tremendous

00:04:32,066 --> 00:04:39,832
degree and we can take these sections and use them and look at them with our favorite microscope or microscopes and we develop

00:04:39,833 --> 00:04:49,533
fantastic information from that, but we also lose a lot of information. This section really doesn’t look much like this three-dimensional sample at all.

00:04:49,533 --> 00:04:57,833
We’ve only actually picked up three of the particles, they’re of different sizes; we really can’t see the relationships between them at all.

00:04:57,833 --> 00:05:04,699
Stereology, according to Weibel, is where we make mathematical methods here in two-dimensional flatland. We use stereological

00:05:04,700 --> 00:05:12,400
formulary and we recover some of this lost information. It’s not serial section reconstruction—we don’t recover everything--but we certainly

00:05:12,400 --> 00:05:24,133
recover some of that lost information. But those diagrams are a little simplistic; this isn’t the real world. The real world of tissues and organs is

00:05:24,133 --> 00:05:31,699
particles have different shapes, they’ve got different sizes, they’ve got different orientations, and so when we take sections we finish up with

00:05:31,700 --> 00:05:39,266
more complex images of different sizes and so forth. I could have made this much more complicated, but for example, this pink particle

00:05:39,266 --> 00:05:48,199
here is actually hit by the section twice, so we see two profiles. So, things start to get more complicated and stereology starts to get a little bit,

00:05:48,200 --> 00:05:59,933
perhaps, more complex, and recognizing that and with the changes that occurred in stereology in the 1980s, my abbreviated definition was that

00:05:59,933 --> 00:06:06,533
stereology is the “discipline concerned with the quantitative analysis of three-dimensional structures.” I’m sure there are better definitions.

00:06:06,533 --> 00:06:14,999
What I was trying to get away from was stereology is just mathematics on sections. It’s about sampling, it’s about experimental design, it

00:06:15,000 --> 00:06:23,233
may be thin sections or even thick sections and we saw a move away from thin sections and so forth, so it’s much more than just making

00:06:23,233 --> 00:06:34,299
measurements in sections. Okay. I now want to just talk briefly about particles, profiles, and problems therein. As I’ve said, a particle in

00:06:34,300 --> 00:06:43,233
stereology-speak is a three-dimensional object such as a glomerulus, a cell or a nucleus and a profile is the two-dimensional representation of

00:06:43,233 --> 00:06:50,666
that particle once we have done some sectioning, whether it be sectioning with a microtome, whether it be optical sections with a confocal

00:06:50,666 --> 00:06:59,866
microscope, whether it be CT slices and the like, and I want to put it to you that we don’t see glomeruli in sections. What we see are glomerular

00:06:59,866 --> 00:07:09,699
profiles; bits and pieces of glomeruli. We don’t see podocyte nuclei in sections, we see podocyte nuclear profiles; we see bits and

00:07:09,700 --> 00:07:20,700
pieces of podocyte nuclei. This brings us to the issue of counting—counting glomeruli, counting podocytes—which is the major focus of this

00:07:20,700 --> 00:07:30,400
workshop, and I want to get across the point that most of the time when we’re talking about counting things or a number, most of the

00:07:30,400 --> 00:07:37,933
time—not all of the time—we’re interested in total number of things: the total number of glomeruli in the kidney, the total number of podocytes in a

00:07:37,933 --> 00:07:46,399
glomerulus. But there are other parameters that we could also come up with. The number for example, if we just stick with glomeruli, the

00:07:46,400 --> 00:07:53,766
number of glomerular profiles per unit area of section; wouldn’t it be lovely if that’s all we had to do? We wouldn’t probably need this workshop.

00:07:53,766 --> 00:08:03,199
We’d just go one, two, three, four, five, six, seven, eight, nine. There’s nine glomeruli in the kidney: publish, you know? We could really

00:08:03,200 --> 00:08:13,133
crank it out, couldn’t we? Obviously, that’s ridiculous. This is a tiny, tiny sample of the kidney but what’s more, as I’ve said, I call these

00:08:13,133 --> 00:08:22,766
glomerular profiles, not glomeruli. So the point is that the number of these profiles per unit area does not equal the number of glomeruli per unit

00:08:22,766 --> 00:08:32,332
volume of three-dimensional cortex or kidney and that doesn’t equal total number, and the literature in our field as well as in other disciplines is

00:08:32,333 --> 00:08:43,199
replete of reports of these parameters which are often interpreted—misinterpreted—as reports of total number. So, I asked the question: what’s

00:08:43,200 --> 00:08:51,466
wrong with N(A) and what’s wrong with N(V)? I can answer that in three ways. I’m going to talk about N, I’m going to talk about area and volume,

00:08:51,466 --> 00:09:00,332
and then I’m going to tell you about kind of the problems; well, more problems. If we’re doing N(A), the number of glomerular profiles seen in a

00:09:00,333 --> 00:09:07,833
section, we just saw nine and I said, well, that would be great and I think you’d all agree, but the number of profiles seen in a section does depend

00:09:07,833 --> 00:09:16,599
on how many glomeruli there are in the kidney, so that’s the good news. But it also depends on the size of the glomeruli because big glomeruli have a

00:09:16,600 --> 00:09:25,033
much better chance of being hit by a single section than small glomeruli. It depends on glomerular shape and it depends on section

00:09:25,033 --> 00:09:35,199
thickness. So the number per area, in terms of number, is not just about how many glomeruli there happen to be, unfortunately. What’s wrong

00:09:35,200 --> 00:09:43,066
with area and volume? This raises what we call in stereology the reference trap because, of course, these parameters N(A) and N(V) are

00:09:43,066 --> 00:09:52,532
ratios; it’s the number per area or the number per volume. So, the value for that parameter may change if the numerator N changes, which is

00:09:52,533 --> 00:10:01,099
what we’re mostly interested in, but it can also change if the denominator changes. So area and volume can change due to tissue shrinkage

00:10:01,100 --> 00:10:08,366
caused by fixation, histological processing, imbedding and the like; it can change thanks to disease through edema, tubular interstitial

00:10:08,366 --> 00:10:18,866
fibrosis, tubular hypertrophy, tubular loss; it can change through technical things such as section compression and so forth. So, we’ve got to

00:10:18,866 --> 00:10:26,399
remember this is a ratio. Ratios are very difficult, generally, to interpret. So as I’ve said, N(A) and N(V) are densities or ratios, they’re frequently

00:10:26,400 --> 00:10:37,300
misinterpreted, they tell us nothing about the total number of glomeruli in the kidney. I should have probably made a separate slide of this and some

00:10:37,300 --> 00:10:46,400
schematics to make it a bit easier, perhaps, to talk through or to follow but what I’ve said here is if this ratio, N(A) or N(V) increases, N may have

00:10:46,400 --> 00:10:56,600
increased—so maybe that’s a good day—but not necessarily to the same extent. A 30% increase in N(A) doesn’t necessarily mean there are 30%

00:10:56,600 --> 00:11:06,500
more glomeruli. If N(A) or N(V) increase, the total number may have actually decreased. N has decreased but the area or volume has decreased

00:11:06,500 --> 00:11:14,100
even further, and therefore the ratio has gone up. So you do your experiment, you conclude there’s a decrease that’s highly statistically significant

00:11:14,100 --> 00:11:26,066
and it’s perfectly incorrect. And finally, N(A) or N(V) may have increased but N may actually not have changed, so it’s really not helping us at all.

00:11:26,066 --> 00:11:33,066
We might get the right answer by luck but more often than not we’re going to get it wrong, and these same issues apply when we’re counting

00:11:33,066 --> 00:11:42,266
podocytes. So maybe to summarize a little bit what I’ve said here about particles and profiles, a sample of three-dimensional particles such as

00:11:42,266 --> 00:11:51,566
glomeruli and podocyte nuclei hit by a single section is not a uniform sample. As Hans Jorgen Gundersen told us 20-odd years ago, “The cells

00:11:51,566 --> 00:12:01,932
or particles seen in a single section is a biased, non-representative sample of all cells,” and in my own language I’ll put it this way and I wasn’t

00:12:01,933 --> 00:12:09,966
taught this at university and I don’t know if you’re taught it at medical school or training to be pathologists: histological sections are biased

00:12:09,966 --> 00:12:23,166
with respect to number. Okay. I now want to go on and talk about how stereology provides estimates and talk about absolute and relative

00:12:23,166 --> 00:12:32,166
parameters. So, rarely in stereology do we measure an entire population; that’s the case in most of science, right? Rather, we obtain a

00:12:32,166 --> 00:12:40,566
sample, hopefully it’s representative of the population, we analyze the sample and then estimate a population value. So in my lab we

00:12:40,566 --> 00:12:45,932
estimate the total number of glomeruli in the kidney and I can assure you, when there’s something like maybe two million glomeruli in a

00:12:45,933 --> 00:12:56,033
human kidney, we don’t count them all. Estimation is about smart experimental design, it’s not guessing, and it’s the same kind of thing that, you

00:12:56,033 --> 00:13:02,966
know, opinion polls do prior to elections. I understand there might be an election in this country this year; I’m not 100% sure but we read

00:13:02,966 --> 00:13:09,199
a little bit about it even on the other side of the world. And so, we see these opinion polls about how many people are going to vote for

00:13:09,200 --> 00:13:16,233
Mr. Romney or Mr. Gingrich or whatever. They don’t ask everybody in Florida, they asked a few thousand and it’s carefully sampled and they

00:13:16,233 --> 00:13:24,199
come up with an estimate and it has a precision, etc., etc. That’s what we do in stereology. So, let’s think about the estimates. Let’s think about

00:13:24,200 --> 00:13:34,666
the estimators or the instruments or the stereological techniques we might want to use. Most of us, I think we would all agree, we would

00:13:34,666 --> 00:13:46,166
like our estimators to be accurate and precise. So if truth in this experiment is the bulls-eye and the stars there are our estimates, here is a nice

00:13:46,166 --> 00:13:55,166
estimator. They’re close to the truth, the average value is the bulls-eye and there is high precision. Alternatively, we might have an instrument or a

00:13:55,166 --> 00:14:06,499
measuring technique such as this. Again, the values are averaging to the truth, the bulls-eye, but there’s less precision. We may have an

00:14:06,500 --> 00:14:17,333
estimator like this: very precise but clearly there is a bias here towards the top right-hand corner of the dart board, so it is inaccurate, but it is

00:14:17,333 --> 00:14:27,233
precise. And finally, there’s this outcome: inaccurate, there is a bias to the right-hand side of the dart board here and the estimates are also

00:14:27,233 --> 00:14:37,199
imprecise. I think we’d agree this is our preference: accuracy and precision. And I’d put it to you but you may disagree, perhaps this is the

00:14:37,200 --> 00:14:47,533
least favorite outcome because of all approaches we can guarantee with this technique, we won’t get the right answer. None of the targets, none of

00:14:47,533 --> 00:14:57,099
the darts, hit the true value. So, I just want to talk briefly about some kidney parameters of interest, and first of all just talk about absolute

00:14:57,100 --> 00:15:06,066
values. And again, this is a short list—I could have 5 or 10 slides here—but the things we can do in terms of volume; volume of kidney;

00:15:06,066 --> 00:15:13,432
obviously, main glomerular volume or individual glomerular volume, that’s something that I’ve been doing with Wendy Hoy in our collaboration for a

00:15:13,433 --> 00:15:21,699
number of years; or perhaps you’re interested in length, the total length of peritubular capillaries in the cortex, the total length of capillaries in an

00:15:21,700 --> 00:15:29,766
average glomerulus. To my mind these are all totally measurable; the total surface area of capillaries in an average glomerulus or the total

00:15:29,766 --> 00:15:37,832
surface area of all glomerular capillaries in the kidney, possible. And number of done: total number of glomeruli in the kidney, total number of

00:15:37,833 --> 00:15:46,833
podocytes in an average glomerulus, total number of podocytes in the kidney. I don’t know why you’d want to measure the last one, necessarily,

00:15:46,833 --> 00:15:56,433
but maybe that’s a good thing to do, it’s possible. There are the absolute values. What about relative values or densities? These are kind of the

00:15:56,433 --> 00:16:05,699
often termed the “traditional” stereological parameters. They’re all ratios and they’re all a measurement relative to a unit volume. It might be

00:16:05,700 --> 00:16:13,433
a volume to a volume such as the volume density of glomeruli in a kidney. In other words, what proportion of the kidney is made up by glomeruli?

00:16:13,433 --> 00:16:21,299
Maybe we’re interested in the length of something, a length density, length per volume, length of capillaries per unit volume of glomerulus;

00:16:21,300 --> 00:16:30,400
or a surface density, surface area of capillaries per unit volume of glomerulus. I’ve already touched upon numerical density for glomeruli and

00:16:30,400 --> 00:16:41,966
podocytes. So, there are ratios to a unit of volume. And these are the traditional stereological formulae for estimating these four

00:16:41,966 --> 00:16:50,732
densities, and you can see many of these formulae have been around for a long time and going back to the middle of the 19th century with

00:16:50,733 --> 00:16:59,766
Delesse, who gave us this formula or a version of it anyway, for volume density estimation, simple algebraic equations for length density and

00:16:59,766 --> 00:17:09,432
surface density are still totally valid today with some caveats. And then there are these traditional methods for estimating number or

00:17:09,433 --> 00:17:19,566
numerical density, so-called or nowadays called “model-based” methods because they require knowledge or assumptions of the geometry of the

00:17:19,566 --> 00:17:27,166
particles you’re interested in, hence the geometric model. So, they’re the methods of Wicksell, and Dehoff and Rhines, of Abercrombie, of Floderus,

00:17:27,166 --> 00:17:35,966
and Weibel and Gomez. Now, I just want to talk briefly about the method of Floderus and Weibel and Gomez. So, this is the Floderus approach or

00:17:35,966 --> 00:17:45,932
equation for estimating numerical density. It’s the number of profiles per unit area divided by three factors: the mean caliper diameter of the

00:17:45,933 --> 00:17:56,966
particles, the section thickness minus two times a correction for lost polar caps. So, let’s just think about the terms here on the right-hand side. N(A)

00:17:56,966 --> 00:18:06,066
and problems therein, I’ve already discussed. Mean caliper diameter is not the diameter of the profiles you measure on the section; it’s actually

00:18:06,066 --> 00:18:17,566
the mean diameter of the particles in three-dimensional space. That’s difficult to determine for a heterogeneous population of particles, and

00:18:17,566 --> 00:18:28,232
for particles with complex shapes it’s very difficult indeed; this is kind of the killer with this technique. Section thickness as I’ve said here is

00:18:28,233 --> 00:18:37,433
not easy to accurately measure. I probably should have used more delicate words, perhaps. It’s not easy to measure well but there are

00:18:37,433 --> 00:18:46,466
certainly approaches, and “h” is a correction for particles that are in the section but they’re only just a tiny little bit in and you actually miss them

00:18:46,466 --> 00:18:55,132
and so therefore you undercount. So without that correction, you will underestimate. And of course, the method still leaves the question, well,

00:18:55,133 --> 00:19:04,066
that’s number per volume; I still don’t have total number or “N.” What about Weibel and Gomez, the formula dating back to---what is that?—50

00:19:04,066 --> 00:19:13,599
years ago this year. Here is their formula. Number per volume is equal to K over beta—that I’ll come back to—and you also need to measure number

00:19:13,600 --> 00:19:21,800
per area and volume density. Well, these are easy to measure, easy and quick, and I’ve already talked about N(A). Beta is a shape

00:19:21,800 --> 00:19:30,866
coefficient. It is a factor that we need to have knowledge or assumptions about the shape of the particles we’re trying to count. It happens to

00:19:30,866 --> 00:19:43,366
be 1.382 for spheres and it’s easy to come up with shape factors for particles with known geometric shapes. “K” is a size distribution

00:19:43,366 --> 00:19:53,332
coefficient. It’s close to 1 for particle size distributions with a smaller standard deviation but then increases reasonably quickly. The concern

00:19:53,333 --> 00:20:01,733
many of us have with the Weibel and Gomez approach is that values for beta and K are often assumed; they’re not measured because that’s

00:20:01,733 --> 00:20:10,533
kind of difficult. So, we in many cases just assume that the value from the study I published five years ago or somebody published 1.38 and

00:20:10,533 --> 00:20:19,033
1.01 all used that. We know these values change with age and disease, and therefore you can see if you make assumptions and those assumptions

00:20:19,033 --> 00:20:31,066
are wrong, your estimate will be biased. This is some data that I published with Wendy Hoy a few years ago. I just show it to illustrate changes in

00:20:31,066 --> 00:20:38,032
the glomerular size distribution. I won’t talk about the biology here; I just want to talk about the size distribution. So, this is glomerular volume on the Y

00:20:38,033 --> 00:20:49,933
axis in million cubic micrometers. Each dot on this slide illustrates the volume of a single glomerulus and each column shows 30 glomeruli per subject;

00:20:49,933 --> 00:20:59,333
there are 720 glomeruli on this slide. So, you can see this subject here has a very tight size distribution. The smallest glomerulus is about 2

00:20:59,333 --> 00:21:12,066
and the largest is about 4. Compare that to this person. The smallest is about 4, the highest is 16. We can see others here ranging from about 1 up

00:21:12,066 --> 00:21:22,199
to 7 or so. So, huge differences in the glomerular size distribution which we simply cannot see using histological sections; this required detailed

00:21:22,200 --> 00:21:34,366
Cavalieri estimation. And so if we were to just assume the same size distribution with the Weibel and Gomez formula, we have a problem. So,

00:21:34,366 --> 00:21:42,032
Luis Cruz-Orive and Ewald Weibel wrote in 1990 considering these traditional model-based methods. None of these methods is unbiased.

00:21:42,033 --> 00:21:54,199
They should no longer be used because a new very simple method—one could argue that point—is now available: the disector method. So, I now

00:21:54,200 --> 00:22:03,166
want to move on and start to think about the disector principle and these other new design-based approaches. So, this is the title of the

00:22:03,166 --> 00:22:11,999
disector paper published in the Journal of Microscopy in May 1984. The author was D.C. Sterio—that I’ll come back to—and the title of the

00:22:12,000 --> 00:22:22,233
paper is “The Unbiased Estimation of Number and Sizes of Arbitrary Particles Using the Disector.” By arbitrary, this means any shape, any size, any

00:22:22,233 --> 00:22:31,033
size distribution, any orientation; it doesn’t matter, it doesn’t affect the answer. The disector is a probe which samples particles with uniform

00:22:31,033 --> 00:22:41,099
probability. I call it “equal opportunity stereology” and I think it’s fair to say the disector revolutionized stereology. For the first time, we

00:22:41,100 --> 00:22:50,900
could deal with arbitrary particles of any size and shape and so forth. No geometric assumptions suddenly were required. We suddenly started to

00:22:50,900 --> 00:23:01,700
use pairs of sections where we’d always used single sections, hence, “di-sector.” We started to use, in some situations, thicker sections whereas

00:23:01,700 --> 00:23:07,866
in stereology up until this point the whole push was thinner sections as you could possibly cut. When I looked on PubMed a few weeks ago, I

00:23:07,866 --> 00:23:17,632
see this paper as being cited over 1,700 times since it was first published in ’84. So, the idea with the disector—and I usually describe this in

00:23:17,633 --> 00:23:26,899
an atrociously bad way—is we look at pairs of sections. So let’s say section 3 and section 4 and the question is, we’re going to count a particle at

00:23:26,900 --> 00:23:34,266
some unique moment, some unique event; let’s say we’re going to count its bottom. The idea is it doesn’t matter how big or small or the shape,

00:23:34,266 --> 00:23:44,099
each particle there only ends once or only has one bottom, if you will. So in section 3 we see this particle is being cut, we then look in section

00:23:44,100 --> 00:23:53,266
4, it’s still there. So between those two sections, it did not end; its bottom is not there, it’s actually down here. We see this particle that’s present in

00:23:53,266 --> 00:24:03,332
3 but it’s gone in 4, so we know somewhere in that space it disappeared; we count it, one. We can also go back the other way. We can say,

00:24:03,333 --> 00:24:14,933
what about particles present in 4 but weren’t in 3? This particle’s in 4, it’s not in 3; we count it. This particle is in 4, it’s also in 3; we don’t count it.

00:24:14,933 --> 00:24:26,066
This particle we don’t even look for in 3 because we don’t even see it in 4. So, it’s a matter of now you see it, now you don’t counting. You might ask

00:24:26,066 --> 00:24:35,199
who D.C. Sterio was. Well, you’ll notice this asterisk here and the asterisk refers to a footnote on the bottom of the first page of the article, and

00:24:35,200 --> 00:24:43,933
the asterisk says D.C. Sterio is the nom de plume of an eminent stereologist who does not want his name associated with this technique in perpetuity.

00:24:43,933 --> 00:24:52,433
I’ve never seen that in a scientific paper. This is a journal of the Royal Microscopical Society. D.C. Sterio is, in fact, Hans Jørgen Gunderson from

00:24:52,433 --> 00:25:01,833
Denmark and his point was that the disector had kind of been invented a number of times in the previous century in part and he felt that he

00:25:01,833 --> 00:25:09,599
actually brought those different parts together to make the final technique, and he managed to convince the editors of the Journal of Microscopy

00:25:17,366 --> 00:25:19,999
to let him do that. Why D.C. Sterio? Well, if you look down in the bottom right corner—those of you who can see it—you’ll see DC Sterio comes

00:25:20,000 --> 00:25:31,366
out like that, okay? Cute. Okay. So, how do we do this in the laboratory when we’re counting glomeruli? We use pairs of microscopes. You

00:25:31,366 --> 00:25:38,666
can’t see those terribly well. They’re like microscopes fitted with projection arms. We look at the same fields on two sections and we do the

00:25:38,666 --> 00:25:45,666
now you see it, now you don’t comparison. So, how do we do this in the human kidney, for example, where we’ve been working for some

00:25:45,666 --> 00:25:56,299
time? We obtain a kidney, we cut it in half, we slice it at 4mm slices and we sample every fourth; here they are. We remove most of the

00:25:56,300 --> 00:26:04,533
medulla, we chop these into smaller pieces, we chop these smaller pieces into smaller pieces, and you can’t see but there are asterisks here

00:26:04,533 --> 00:26:13,266
showing where we sample every 20th of those blocks, and these are the blocks that we’re ultimately going to count glomeruli in. We also

00:26:13,266 --> 00:26:22,366
sample a spare set just in case things go wrong in the laboratory with fixation or sectioning or whatever; that’s that set there. We take those

00:26:22,366 --> 00:26:31,966
samples and embed in glycolmethacrylate, we exhaustively section at 20 micrometers, we pick up every 10th and 11th section stained with PAS,

00:26:31,966 --> 00:26:39,666
and we estimate total glomerular number with a physical disector/fractionator combination, and we aim to count at least 100 of these

00:26:39,666 --> 00:26:48,032
disappearing events that I talked about. For example, if you look in this section here there is a glomerular profile here and if you look over here

00:26:48,033 --> 00:26:57,399
in this look-up section as it’s called, it’s gone; so that glomerulus is counted. All of the others are still there so they’re not counted; they haven’t

00:26:57,400 --> 00:27:05,766
ended in that space, if you will. That only gives us number per volume. How do we get total number? Well, Gunderson also described the

00:27:05,766 --> 00:27:14,499
fractionator approach which I think is a very neat experimental design. So for us to estimate total number of glomeruli in the human kidney, we

00:27:14,500 --> 00:27:24,333
count these disappearing events, we call that Q minus, and we have five sub-sampling fractions. We cut the kidney in half, we take every fourth

00:27:24,333 --> 00:27:34,933
4mm slice, we take 1/20 of the blocks, we take 1/10 of the sections and we count in a known proportion of the fields. So, it’s just a matter of

00:27:34,933 --> 00:27:43,499
keeping track of those fractions and then just doing simple algebra to get the total number. This approach is not influenced by technical artifacts.

00:27:43,500 --> 00:27:51,900
We don’t have to measure section thickness. It doesn’t matter if the tissue shrinks or swells or distorts as long as we know the fractions; that’s

00:27:51,900 --> 00:28:01,200
kind of the nice thing about it. So, how does this technique work? Well, here are a number of studies of estimates of nephron number in normal

00:28:01,200 --> 00:28:09,033
human kidneys from a Danish study, a French study, a German study, and our studies of black and white Americans, Australian white and

00:28:09,033 --> 00:28:17,433
Aborigines, and also our studies in Senegalese, and here is the average count down here for these studies and the average count is around

00:28:17,433 --> 00:28:28,799
900,000-950,000 in most of those studies, except the Danish study which was an elderly group of 37 Danes and except in the Australian Aborigines

00:28:28,800 --> 00:28:37,466
where we find they have fewer nephrons than whites from the same geographic location. Just out of interest, it’s interesting to note that in the

00:28:37,466 --> 00:28:45,799
studies with 20-40 kidneys each—these ones and these ones down here—the range in nephron number is around three- to four-fold

00:28:45,800 --> 00:28:55,366
whereas in our study in the American groups—in the whites and black Americans—much bigger numbers; the range, of course, has really blown

00:28:55,366 --> 00:29:04,366
out with the increased sample size. I just want to point before moving on that we can count glomerular cells…oh, sorry, we can also use the

00:29:04,366 --> 00:29:13,199
disector at the EM level and there may well be some presentations on that here at this workshop. So, some years ago we were

00:29:13,200 --> 00:29:18,133
counting podocytes, for example, using pairs of sections. Here’s a podocyte nucleus; now you see it in the look-up section, now you don’t, so

00:29:18,133 --> 00:29:26,299
you count it. Here’s another one here; now you see it, now you don’t so you count it. So, you can do this at the EM level but it’s a bit trickier for

00:29:26,300 --> 00:29:34,666
obvious reasons. There’s one other approach I also want to point out for getting total number. I talked about disector/fractionator; we can also

00:29:34,666 --> 00:29:43,232
combine the disector with the Cavalieri approach, which is a method for estimating the volume of anything. So if we get the numerical density from

00:29:43,233 --> 00:29:54,166
disector we can multiply by kidney volume from Cavalieri to get total number, and here in the rat kidney, for example, we take a kidney, we slice it

00:29:54,166 --> 00:30:01,899
at 1mm—use a little razor blade slicing device—we pick up, in this case, every second of these slices, we measure the area of all of those slices

00:30:01,900 --> 00:30:10,966
using simple point counting, and we calculate the volume of the kidney as the reciprocal of the slice sampling fraction which is one over a half which

00:30:10,966 --> 00:30:20,599
is two, the total number of points on the slices, the area of each grid square which is a constant, and slice thickness which in this case was 1mm,

00:30:20,600 --> 00:30:33,933
and that’s quite a very robust and quick and easy method for estimating volume. So, I painted a rosy picture of the physical disector. Now, let’s maybe

00:30:33,933 --> 00:30:46,199
think a little harder about this. It requires alignment of physical sections. That’s easy for glomeruli; it’s more difficult for cell nuclei. I’m not

00:30:46,200 --> 00:30:55,933
saying it’s impossible and some of you use this approach for counting glomerular cells, but it is tougher. The counting is slow when it comes to

00:30:55,933 --> 00:31:06,066
counting glomeruli with this approach. In our studies of the rat kidney, after the sampling, fixation, embedding, sectioning, staining,

00:31:06,066 --> 00:31:14,966
cover-slipping—I think I got the order right—it takes my best technicians six hours to do the glomerular counting; that’s on top of everything

00:31:14,966 --> 00:31:27,399
else, and if we do this in the human kidney it takes 10 hours, so that kind of rules many labs uninterested due to the cost-efficiency and the

00:31:27,400 --> 00:31:35,200
labor-intensiveness of the approach. One development here was the optical disector approach that came out about five years after

00:31:35,200 --> 00:31:45,966
that first disector paper recognizing the difficulty of counting nuclei at that point with physical disectors. This is a single section technique and

00:31:45,966 --> 00:31:53,666
the idea is we cut thick sections—maybe 20 micrometers, maybe 70 micrometers—and then those sections are optically sectioned either with

00:31:53,666 --> 00:32:03,266
an appropriately deck-out in terms of lenses light microscope or perhaps optically sectioned with a confocal microscope and the idea is, again, we

00:32:03,266 --> 00:32:12,132
want to count some unique event, maybe when the nucleus ends, maybe when we first see it; as long as it’s consistent. We find that it’s easiest to

00:32:12,133 --> 00:32:23,066
count nuclei at a unique point, namely when the nucleus first comes into focus. Can we have the lights, Armand? Thank you. So, we’re using this

00:32:23,066 --> 00:32:32,066
approach to count podocytes with optical disectors. We’re cutting serial sections at 14 micrometers, we’re immunostaining the sections

00:32:32,066 --> 00:32:42,232
for WT1 to see podocytes, von Willebrand factor for endothelial cells, DAPI for all the nuclei. We then take those 14 micron sections to a Leica SP5

00:32:42,233 --> 00:32:56,866
confocal microscope suitably fitted out and we obtain images in the z-axis at 1 micrometer increments. It’s so dark in here I can’t see the

00:32:56,866 --> 00:33:08,432
keyboard. I’m hoping…oh, that was lucky. So, this is now a little movie. We’ll put the lights on in just a second but it’s a little movie summarizing

00:33:08,433 --> 00:33:21,033
what we do. So, here are the 14 micron sections we’re cutting from a glomerulus. We take every second, hence the gaps between them. This is

00:33:21,033 --> 00:33:31,299
showing they’ve already been immunostained. We then take each one of these 14 micron sections to the confocal microscope. Here are

00:33:31,300 --> 00:33:42,200
the 1 micron perfectly aligned z-series of sections and then we go through this manually to try to identify podocytes and endothelial cells

00:33:42,200 --> 00:33:54,166
counting with the optical disector approach. Thanks, Armand. Okay. That makes it much easier here. So, how do we then get the total podocyte

00:33:54,166 --> 00:34:02,066
number? We use this formula. The number of podocytes to glomerulus is 1 over the reciprocal of the section sampling fractions; remember, I

00:34:02,066 --> 00:34:10,499
said we do every second, so this is 1 over ½ or two. We count in 8 of the 14 micron section height, so this is another sampling fraction and Q

00:34:10,500 --> 00:34:19,166
minus is simply the number of podocytes we count. We haven’t published any of this data yet but I’m hoping to start to do that in the next half

00:34:19,166 --> 00:34:30,599
year or so. Here’s some preliminary data looking at kidneys of lean and obese American white men, and I’ve got data here for 6 glomeruli from

00:34:30,600 --> 00:34:40,433
each of 6 men, so that’s only 36 glomeruli so it’s very early days. Our podocyte count is, on average for these 36 glomeruli, 560 per

00:34:40,433 --> 00:34:49,933
glomerulus, but what struck me was the range. Our lowest count was 250 and our highest count is 950 just in a small sample. For non-epithelial

00:34:49,933 --> 00:35:01,566
cells—this was mesangial endothelial cells and any other non-podocytes on the tuft—the range is 600 to nearly 4,000. What also surprised us

00:35:01,566 --> 00:35:10,032
were the parietal epithelial cells which we were counting: the lowest is just over 100 and the highest is over 1,200 which is something I would

00:35:10,033 --> 00:35:18,366
not have guessed in many years, I don’t believe. I’ll just share you a little bit of data here. We’re finding glomeruli from obese subjects are larger—t

00:35:18,366 --> 00:35:25,966
here’s no surprise—and contain more cells than those from lean subjects and actually the glomeruli of the obese contain more podocytes,

00:35:25,966 --> 00:35:34,599
more non-epithelial cells, and more parietal epithelial cells. And so, we’re starting to be able to correlate glomerular size with podocyte

00:35:34,600 --> 00:35:44,866
number and we’re finding in both, obese subjects in red and lean subjects in blue, in both cases larger glomeruli contain more podocytes than

00:35:44,866 --> 00:35:52,799
smaller glomeruli, which raises in my mind the whole issue of: where did those podocytes come from, is that how the glomeruli were born,

00:35:52,800 --> 00:36:00,733
were they added as the people got older and bigger, where did they come from? I put this slide up and I’m sure Kevin will show you something

00:36:00,733 --> 00:36:11,666
similar to this tomorrow. These are some data from six previous reports of podocyte number in normal human kidneys and the point is that there’s

00:36:11,666 --> 00:36:20,566
a big range, too, in these reports. Kevin’s mean value back here from this paper was 300 podocytes per glomerulus, the Steffes paper

00:36:20,566 --> 00:36:28,699
here is over 900, and there are other high counts here. So, is this really about differences in the subjects between these studies or has it got

00:36:28,700 --> 00:36:36,033
more to do or something to do with the techniques used to count the podocytes: the embedding material, the method used to estimate

00:36:36,033 --> 00:36:45,566
glomerular volume, the method used to count the cells? I think we’d all agree that most likely technique had something to do with this variation

00:36:45,566 --> 00:36:55,332
and maybe it had something to do with why this workshop’s being held here today. Okay. I’m nearly done. Three topics to go: counting and

00:36:55,333 --> 00:37:06,133
sizing glomeruli with MRI. To paint less of a rosy picture, I pointed out limitations of counting and sizing glomeruli with stereology already. All the

00:37:06,133 --> 00:37:15,633
studies I’ve talked about, from my lab at least, require a kidney obtained at autopsy. This limits our study design very much and clearly rules out

00:37:15,633 --> 00:37:23,399
longitudinal studies, and in my lab we do all of this in glycolmethacrylate. The current methods are time-consuming, as I’ve told you; they’re tedious;

00:37:23,400 --> 00:37:31,633
they’re labor-intensive; and therefore they’re expensive and therefore they’re not actually used by that many laboratories around the world and I

00:37:31,633 --> 00:37:41,599
feel this is a real problem. Most labs are therefore still reporting number per area or number per volume. So, I suggest we need a new method

00:37:41,600 --> 00:37:53,100
that’s faster, that’s accurate, it’s precise, ideally we could count and size every glomerulus in a kidney, and ideally again it would be non-invasive,

00:37:53,100 --> 00:38:03,666
allow longitudinal studies, reducing the numbers of animals used in our experimental labs, and ultimately be applicable to patients. It sounds like a

00:38:03,666 --> 00:38:11,799
dream but I think we’re on the way because last year two, I think very significant papers were published where MRI was used for the first time,

00:38:11,800 --> 00:38:21,466
to my knowledge, to count and size glomeruli in the kidney. These reports came out of Kevin Bennett’s lab at Arizona State University and

00:38:21,466 --> 00:38:29,766
Kevin will be talking about this work tomorrow, and out of Norbert Gretz’s lab at the University of Heidelberg in Mannheim. In both studies rats were

00:38:29,766 --> 00:38:37,132
perfused IV with cationic ferritin that bound to anionic charges on the glomerular basement membrane and on podocytes. The rats were

00:38:37,133 --> 00:38:45,433
perfusion fixed and kidneys removed and placed in either glutaraldehyde or agarose, depending on the study, then the ex-vivo kidneys were imaged

00:38:45,433 --> 00:38:57,699
with MR and Norbert’s estimation was that this work could be done in 1/6 of the hands-on time that currently it takes to do in my lab, and more

00:38:57,700 --> 00:39:05,100
importantly, this approach not only gave us glomerular number, it gave us the volume of every glomerulus in the kidney and therefore the

00:39:05,100 --> 00:39:14,233
glomerular size distribution, which I had never seen before. So, here is an image from Norbert Gretz’s lab of the glomeruli labeled with the

00:39:14,233 --> 00:39:23,933
cationic ferritin and the glomerular distribution. These kidneys in both cases were sent to my lab at Monash for us to use the disector/fractionator

00:39:23,933 --> 00:39:31,699
to see if we got the same answer and the agreement in both cases was very good. Here from the Heilmann study from Norbert Gretz’s lab,

00:39:31,700 --> 00:39:44,133
our mean glomerular count for six rat kidneys was 35,000 with stereology with a standard deviation of 3,000 MR, 33,000 with an SD of

00:39:44,133 --> 00:39:54,466
3,000, so I think we’re on the way and Kevin will tell us more about this, I think very exciting, opportunity tomorrow. In biology and science

00:39:54,466 --> 00:40:03,699
dealing with a rare event is doubly difficult and I’m interested, as many of you are I know, in multipotent glomerular stem cells described a few

00:40:03,700 --> 00:40:13,333
years ago by Paolo Romagnani from Florence and Marcus Moeller’s group from Aachen, and the idea is—we can debate how strong this

00:40:13,333 --> 00:40:21,133
evidence is and where this field’s going to go—that there are a population of multipotent glomerular stem cells that reside on the

00:40:21,133 --> 00:40:30,299
Bowman’s capsule here up near the urinary pole and these express a number of stem cell markers. In good times, perhaps these cells might

00:40:30,300 --> 00:40:39,533
break down across the basement membrane, start to express some podocyte markers and ultimately stop expressing stem cell markers and

00:40:39,533 --> 00:40:48,099
now only express podocyte markers, and down towards the vascular pole there’s some evidence that they can reflect back up onto the glomerular

00:40:48,100 --> 00:40:58,700
tuft and turn into podocytes; a very exciting idea and a revolutionary idea in nephrology. So, we’re starting to develop approaches to be able to

00:40:58,700 --> 00:41:07,333
quantify these cells, to be able to quantify the subpopulations of these cells in human kidneys. This is an image from Victor Puelless, a Ph.D.

00:41:07,333 --> 00:41:15,633
student in my lab. You can see the double-labeled stem cells down here towards the urinary pole, you can see some white-labeled cells that

00:41:15,633 --> 00:41:24,466
express podocyte markers as well, and here is a beautiful parietal podocyte as some people call them on the Bowman’s capsule not expressing

00:41:24,466 --> 00:41:33,866
stem cell markers at this point but expressing podocalyxin. We’re interested in the numbers of these cells, how the proportions of these cells

00:41:33,866 --> 00:41:44,099
change with life and with BMI and other risk factors for CKD. I want to finish with a bit of a plea, perhaps, for a consensus for kidney

00:41:44,100 --> 00:41:53,033
stereology. As I said at the start, the lung people have done this. Here is the title of a paper that was published in the American Journal of

00:41:53,033 --> 00:41:59,766
Respiratory and Critical Care Medicine two years ago, entitled “An Official Research Policy Statement of the American Thoracic

00:41:59,766 --> 00:42:09,099
Society/European Respiratory Society: Standards for Quantitative Assessment of Lung Structure,” and you’ll notice Ewald Weibel is one

00:42:09,100 --> 00:42:19,866
of the lead authors on this article. Basically, this article calls for quantitative methods to be accurate, to allow meaningful interpretation of

00:42:19,866 --> 00:42:29,199
results, efficient, to yield adequate precision with reasonable effort, of adequate statistical power to encompass inherent variability, and adherent to

00:42:29,200 --> 00:42:38,500
uniform standards to facilitate comparisons among experimental groups and also across different studies. So, I’ve covered a fair bit of

00:42:38,500 --> 00:42:46,966
material here today. I hope it provides us a satisfactory foundation for the rest of the workshop and I’m sure there will be some issues

00:42:46,966 --> 00:42:56,766
there that you all want to discuss. I just want to acknowledge Victor Puelles, a Ph.D. student who’s doing most of the podocyte counting work

00:42:56,766 --> 00:43:05,032
and developing the stem cell approach, and Rebecca Douglas-Denton here who counted the glomeruli in over 400 human kidneys and probably

00:43:05,033 --> 00:43:14,799
deserves a medal or something for that. My main collaborators…I’ve mentioned these people already today and some of my collaborators at

00:43:14,800 --> 00:43:23,866
Monash. I haven’t spoken in Bethesda before but I just want to take the opportunity to thank NIDDK for the support of our autopsy study over a

00:43:23,866 --> 00:43:42,532
number of years. Thanks very much.
KEVIN BENNETT: Kevin Bennett, Arizona State

00:43:42,533 --> 00:43:51,099
University. You mentioned this variability in podocyte number in your patients and the potential that it’s either sort of the technique or

00:43:51,100 --> 00:43:59,400
the patient population. How much variability is there inside of a single kidney and does it matter where you take the sections from?

00:43:59,400 --> 00:44:09,000
JOHN BERTRAM: The best way I can answer that is from our own study with just the 36 glomeruli—we’ve done substantially more than

00:44:09,000 --> 00:44:20,233
that now—and I showed an average count of about 540 or 550, I think, for those 36 glomeruli; our lowest count was around 250 and the

00:44:20,233 --> 00:44:31,166
highest counts up around 1,000 or so. Kevin and others have counted, and John Bergin has counted, podocytes in human glomeruli, and

00:44:31,166 --> 00:44:40,832
others in this room I’m not aware of…from my memory they tend to publish mean NSDs, I’m not exactly sure of, perhaps, on a glomerulus-

00:44:40,833 --> 00:44:50,666
by-glomerulus basis and I’d be very happy for them to comment on that. For me it raises the question, and I don’t know that we know the

00:44:50,666 --> 00:44:57,966
answer to this, is: if some glomeruli are about four or five times the podocytes of other glomeruli, when did that happen? Did that happen

00:44:57,966 --> 00:45:10,766
when glomerular genesis or podocytogenesis finished? Nephrogenesis is finished at 36 weeks in humans. We’ve lost our S-shaped bodies then;

00:45:10,766 --> 00:45:20,632
I’m not sure when podocyte mitosis finishes but it probably finishes sometime after that. So, is there that variation then or does that variation kind of

00:45:20,633 --> 00:45:30,599
come as we go through adult life as we get bigger and various other things happen in our life, and that raises the issue: where do they come

00:45:30,600 --> 00:45:37,533
from? Because the dogma until recently was they couldn’t divide, but of course, that idea has kind of been turned on its head a little bit in the last 10

00:45:37,533 --> 00:45:51,566
years or so. So, I don’t know if others want to comment who’ve done that work on the variability within kidneys, but it’s there in my opinion.

00:45:51,566 --> 00:45:58,632
KEVIN LEMLEY: Well, I don’t have good data for the human glomeruli because the little we did was a long time ago and I don’t remember it all, but I’ve

00:45:58,633 --> 00:46:09,533
been counting podocytes in mouse glomeruli a lot in the last few years and I think the best example is in this one mouse kidney where we serially

00:46:09,533 --> 00:46:21,733
sectioned nine glomeruli and we were able to follow every podocyte nucleus through the glomeruli, so there was no variability. We weren’t

00:46:21,733 --> 00:46:26,599
estimating; we know what the truth was for those nine glomeruli.

00:46:26,600 --> 00:46:31,266
JOHN BERTRAM: You counted every podocyte?
KEVIN LEMLEY: And counted every podocyte,

00:46:31,266 --> 00:46:40,032
yes. I’m going to show that data tomorrow and I’m trying to remember what the numbers were, but I think it was ranging from 72 podocytes to 135

00:46:40,033 --> 00:46:53,033
podocytes in those 9 glomeruli from 1 normal rat kidney. Of course, the human is going to vary more than the normal rat, so there’s a lot of

00:46:53,033 --> 00:46:58,999
variability. Well, just with respect to the podocyte number, we switched after, as you

00:46:59,000 --> 00:47:07,500
noticed, the very low number of podocytes and that was because of paraffin shrinkage. Now

00:47:07,500 --> 00:47:17,766
we’re doing the volume estimates and the podocyte densities all in Epon and it’s kind of eliminated that; we now get 700, strangely, kind

00:47:17,766 --> 00:47:25,966
of right where everybody is. But when we switched that, we started doing five profiles per individual, whereas we had traditionally done

00:47:25,966 --> 00:47:35,766
two, and what I find is…we stopped doing that after about a year because the time consumption to do five at EM level is unbelievable, but we

00:47:35,766 --> 00:47:47,232
found the inter-individual variants in podocyte density, which is what we primarily measure, is four times what the inter-individual is. So, it

00:47:47,233 --> 00:47:51,666
seems to be relatively tight in an individual.
JOHN BASGEN: I’ve been told to mention…my

00:47:51,666 --> 00:48:00,366
name is John Basgen. Is it possible to go back to the slide that has the bulls-eye on it?

00:48:00,366 --> 00:48:24,132
JOHN BERTRAM: Sure. I’ll have to do it this way.
JOHN BASGEN: Yeah. So, I want to wrestle with

00:48:24,133 --> 00:48:33,499
you a little bit here that you made the comment you want to do the studies that are going to be in the upper left corner there that are accurate and

00:48:33,500 --> 00:48:51,000
precise. To make things more precise takes more work and I just want to mention that maybe the upper right-hand corner is good enough that in

00:48:51,000 --> 00:48:59,266
some instances you want to be accurate and not quite as precise and it’ll save time and money if there’s…

00:48:59,266 --> 00:49:05,066
JOHN BERTRAM: That’s a fair point. We won’t have much of a wrestle on this.

00:49:05,066 --> 00:49:08,399
JOHN BASGEN: All right.
JOHN BERTRAM: There’s a pretty famous

00:49:08,400 --> 00:49:20,333
stereological paper with the words, “do more less well,” because early in stereology some of us had a propensity to totally over-count and

00:49:20,333 --> 00:49:26,066
over-count fields or micrographs or whatever, where the better effort would have been to do more animals or more patients or more glomeruli,

00:49:26,066 --> 00:49:38,632
whatever. Yes, I think depending on the question and what you’re trying to do, yes, in many situations it might be that to use something like

00:49:38,633 --> 00:49:47,533
this is going to take double or triple or quadruple the time of doing this, and so perhaps one would be better. You’re still accurate but you can do two

00:49:47,533 --> 00:49:57,699
or three or four times the samples or maybe you don’t need to because of whether this is a biopsy, whether it’s a mouse or whatever. So

00:49:57,700 --> 00:50:00,966
yeah, I take your point.
JEFFREY KOPP: Jeffrey Kopp. Thank you. That

00:50:00,966 --> 00:50:08,166
was very nice, John. You had a slide where you listed glomerular numbers in different populations in different countries and you pointed out that, as

00:50:08,166 --> 00:50:16,866
I recall, European Americans and African Americans had a larger variance than the other populations and you said something about maybe

00:50:16,866 --> 00:50:28,032
it had to do with the numbers that you counted, although I didn’t see that as it went by. So is it a population…the number of cases or methods or

00:50:28,033 --> 00:50:34,299
could there be greater heterogeneity in either genetics or environment that might be driving that or some combination?

00:50:34,300 --> 00:50:47,066
JOHN BERTRAM: I think the mean values for the African white Americans were very similar and they were very similar to a number of the other

00:50:47,066 --> 00:50:56,466
studies with the exception being the Danish study, which is an old group, and the Australian Aborigines. I believe in both of those cases—I

00:50:56,466 --> 00:51:05,532
may have gone through it too fast—that the difference was we had 150 or 200 subjects in both the American groups, whereas all the other

00:51:05,533 --> 00:51:14,966
studies had a maximum of 47 subjects. So, it’s three or four times the sample size and I think you just get the extremes.

00:51:14,966 --> 00:51:19,632
JEFFREY KOPP: Because you’ve got more.

00:51:19,633 --> 00:51:25,966
DIANE ROSIN: Hi. Diane Rosin, University of Virginia. I have a question of methodology. If we’re going to try to come to some consensus,

00:51:25,966 --> 00:51:34,199
could you care to comment on the type of fixative used and the embedding method and what you choose and why?

00:51:34,200 --> 00:51:45,866
JOHN BERTRAM: No. [laughs] It just gets too hard. No, the serious answer is, in that lung consensus article that I referred to and is on the

00:51:45,866 --> 00:51:57,799
CD that you got in your packs, depending on the parameter will depend on the fixative will depend on whether you’re doing it by light microscopy,

00:51:57,800 --> 00:52:05,600
bright field fluorescence, transmission EM, whatever. Do you need to perfuse or do you not need to perfuse? So, it’s depending on the

00:52:05,600 --> 00:52:11,966
parameter you want to do. Now, you might say, “Well, I want to measure 10 parameters.” Well then, you’re going to have to come up with some

00:52:11,966 --> 00:52:22,066
kind of a trade-off as to how you’re going to do that. So no, I think depending, and you know, can we do everything from a laboratory animal where

00:52:22,066 --> 00:52:28,332
everything’s under your control and you can perfuse and have total control and do it brilliantly, or you’ve got a biopsy, you know, with 10

00:52:28,333 --> 00:52:43,166
glomeruli in it? It’s a totally different world? So yeah, sorry. Have a look at that lung paper and it’ll give you an idea on how they approached it.

00:52:43,166 --> 00:52:50,932
SUSANNE NICHOLAS: Hi, John. Susanne Nicholas, Los Angeles. I wanted to get your opinion. You stated that the MRI studies that were

00:52:50,933 --> 00:52:58,533
recently published have pretty good consistency when you compare it with the disector/fractionator method, which as we know,

00:52:58,533 --> 00:53:09,233
is time-consuming, expensive, etc., and requires expertise and resources. I wonder whether you would advocate for us to be spending our efforts

00:53:09,233 --> 00:53:19,099
trying to expand using MRI specifically for counting and typing glomeruli rather than the disector/fractionator. How much energy should

00:53:19,100 --> 00:53:26,233
we put in either? Do we need to continue to do both?

00:53:26,233 --> 00:53:33,499
JOHN BERTRAM: I’m trying to push MR in my university to develop this method and our machine arrived two weeks ago and I can’t wait

00:53:33,500 --> 00:53:45,800
to start to see what it can do. To my knowledge, it’s an N=2 papers that have done this; there’s much to be learned, although I think the early

00:53:45,800 --> 00:53:53,933
steps are very exciting. Now I must say, those of you who have heard me talk at ASN or World Congress or whatever, every time I get up and

00:53:53,933 --> 00:54:02,166
talk about counting glomeruli in human kidneys the first or second question is, “Why aren’t you using some kind of new imaging device?” and I used to

00:54:02,166 --> 00:54:13,866
say, “I’d really love to,” and now it’s starting to happen. So, I think it’s got a lot of potential. How much? I don’t know. I think that at some point in

00:54:13,866 --> 00:54:24,732
the not-to-distant future I imagine we’ll hopefully be doing this in patients. I don’t know. Is that five years away? Is that 15 years away? I don’t

00:54:24,733 --> 00:54:32,566
know; others would have a better handle on that, I’m sure. But I think we should push it as hard as we can because I think it’s got tremendous

00:54:32,566 --> 00:54:38,266
AGNES FOGO: Thank you for a wonderful talk.

00:54:38,266 --> 00:54:48,699
Agnes Fogo from Vanderbilt University. I have two points to make. First, I’d like to come back to your beautiful bulls-eye target. It wouldn’t be so

00:54:48,700 --> 00:54:53,600
bad to be inaccurate and precise if we knew exactly how far off the target we were and we could put a little correction factor; we might even

00:54:53,600 --> 00:55:02,633
be able to deal with that, too. My other point is that much of the data that you’ve talked about deal with normal populations, normal non-

00:55:02,633 --> 00:55:10,266
diseased humans and animals and our interest in this is, of course, geared to disease settings where we’re trying to understand something

00:55:10,266 --> 00:55:20,399
about the pathophysiology of disease settings, and many of these techniques will have to be re-thought when we’re thinking about podocyte

00:55:20,400 --> 00:55:27,133
number and we have glomeruli that are segmentally sclerosed. I promise you, there are no podocytes recognizable or at least not

00:55:27,133 --> 00:55:37,033
recognizable by the usual immunostaining in those areas. Globally sclerosed glomeruli may become indistinguishable by usual, at least thick,

00:55:37,033 --> 00:55:49,266
disector sections to be counted, they may become continuous with the surrounding area, and many of these processes will be challenging

00:55:49,266 --> 00:55:57,032
to translate into understanding what’s going on in a disease setting, whether we’re looking at podocyte PEC transitions or a response to

00:55:57,033 --> 00:56:06,033
intervention; we’re trying to measure a response to an intervention. Can you tell us your thoughts about how we can translate from these sampling

00:56:06,033 --> 00:56:14,633
and approaches that work well when you’re dealing with the normal variability of developmental issues to the extraordinary

00:56:14,633 --> 00:56:18,133
variability that happens with most kidney diseases and disease settings?

00:56:18,133 --> 00:56:25,299
JOHN BERTRAM: I must admit that’s a very good question and I don’t have probably a very good answer. Most of my work has been in reasonably

00:56:25,300 --> 00:56:36,666
normal human kidneys or in normal laboratory animals or knockout mice or heterozygous mice where we suspect there is a low nephron

00:56:36,666 --> 00:56:45,066
endowment or a high nephron endowment and we’ve identified all of those things, but histologically they look okay. So, certainly when

00:56:45,066 --> 00:56:54,232
we start to deal with pathology, you know, the whole range of issues of focal lesions or segmental lesions which suddenly raises all kinds

00:56:54,233 --> 00:57:02,733
of issues about sampling, about that the podocyte or whatever cell we happen to be interested in is no longer expressing a certain marker that

00:57:02,733 --> 00:57:10,033
perhaps I can rely upon now, or some of them are and some of them aren’t or they’re expressing it at a level that we can’t pick up with

00:57:10,033 --> 00:57:22,333
immunofluorescence, all of these issues are real and can’t be ignored and are tough and I think as we start, particularly with that stem cell project,

00:57:22,333 --> 00:57:31,566
try to start to look in various interesting scenarios we’re going to run into all of those things. But no, you’re quite right. Pathology just makes this an

00:57:31,566 --> 00:57:47,166
order of magnitude more difficult. All of those issues of sampling, of replicates, of non-uniform distribution of whatever’s going on. Yeah. I’m not

00:57:47,166 --> 00:57:56,732
new to stereology but I didn’t have a chance to get into them today but they’re important issues, otherwise you’re going to get problematic data.

00:57:56,733 --> 00:58:01,499
AGNES FOGO: And you don’t have good…obviously, if there was no pathology in the kidney, we wouldn’t be having this meeting probably.

00:58:01,500 --> 00:58:06,066
JOHN BERTRAM: Yeah, probably.
BEHZAD NAJAFIAN: I’m Behzad Najafian,

00:58:06,066 --> 00:58:14,799
University of Washington. Thank you very much for the very nice talk, John. Part of what I wanted to ask, I think, was asked and was answered, at

00:58:14,800 --> 00:58:25,300
least partially. I was wondering about the pictures that you showed and the wonderful pictures giving an understanding of the podocytes parietal

00:58:25,300 --> 00:58:41,933
epithelial cells and endothelial cells. So in your experience, since at least I don’t know any good marker for mesangial cells and by those markers

00:58:41,933 --> 00:58:55,266
that you stained for probably you can’t recognize also other cell types such as inflammatory cells, circulating cells. In optical sections, I wonder that

00:58:55,266 --> 00:58:59,866
probably it will be difficult just by morphology and location of the cells—whether they are outside of

00:58:59,866 --> 00:59:09,332
the glomerular base membrane or not—get some help that whether this cell which, especially if it’s in the middle of the tuft, is the podocyte or not.

00:59:09,333 --> 00:59:21,066
How often do you see cells that they do not have any cytoplasmic staining—WT1—or glomerular factor and how reliable can you tell that they are

00:59:21,066 --> 00:59:29,166
not podocytes or they might be mesangial cells or other things? The other question would be that, in the same biopsies…they were biopsies or

00:59:29,166 --> 00:59:31,732
JOHN BERTRAM: No, that was autopsy tissue.

00:59:31,733 --> 00:59:40,799
BEHZAD NAJAFIAN: Autopsies. So, that would make it easier, actually. Did you do counting of podocytes using other methodologies that you

00:59:40,800 --> 00:59:49,966
can rely on, morphology, to see that you get to the same ballpark?

00:59:49,966 --> 00:59:56,099
JOHN BERTRAM: So, the answer to the second question is, no, we didn’t use another approach, we used this approach. We wanted to count

00:59:56,100 --> 01:00:08,666
podocytes, endothelial cells, mesangial cells and parietal cells separately. Why not? But this is autopsy tissue and we found in, not a high

01:00:08,666 --> 01:00:17,699
proportion but maybe in a third or so of the cases, that the von Willebrand factor wasn’t helping us pick out the endothelial cells from the

01:00:17,700 --> 01:00:28,066
mesangial cells, and so what I published there, what I showed there today, I just called them non-epithelial glomerular tuft cells or something which

01:00:28,066 --> 01:00:37,832
is everything except podocytes. Some of them we can count the endos quite nicely separately. The nice thing with the optical disector approach

01:00:37,833 --> 01:00:47,099
is, of course, you’ve got all of these 1 micron sections through each cell. Okay, at one point you have to make the decision, the nucleus is now in

01:00:47,100 --> 01:00:55,666
focus, I’ll count it, but you’ve got the sections above that and the sections below that—the optical sections above and below—to help you

01:00:55,666 --> 01:01:03,532
with your identification. Where is the nucleus relative to the basement membrane? Is it WT1-positive? Is it von Willebrand factor-positive, etc.?

01:01:03,533 --> 01:01:12,599
So, you’ve got lots of opportunities to make the cell identification call, which we’re reasonably confident we can do. The moment one of my

01:01:12,600 --> 01:01:22,500
people come in and say, “Here is a nucleus outside a capillary that there’s no WT1 cytoplasm,” then we’ve got a problem in that

01:01:22,500 --> 01:01:27,500
sample. Strictly, we’re counting WT1-positive cells.

01:01:27,600 --> 01:01:33,266
BEHZAD NAJAFIAN: How often do you see cells outside capillary that they do not stain for WT1?

01:01:33,266 --> 01:01:35,799
JOHN BERTRAM: We haven’t yet but we’ve only just started this.

01:01:35,800 --> 01:01:41,566
BEHZAD NAJAFIAN: Okay. JOHN BERTRAM: We do see WT1 cells in parietal

01:01:41,566 --> 01:01:44,932
MICHAEL MAUER: Mike Mauer. Thank you very

01:01:44,933 --> 01:01:54,799
much for a wonderful introduction. A couple of comments; one is regarding accuracy and precision. It’s very much going to depend on what

01:01:54,800 --> 01:02:05,866
the question is and what we’re trying to answer. So if, for example, we’re trying to see whether a treatment slows down the rate of progression of

01:02:05,866 --> 01:02:17,566
a very slowly progressive disease over a five-year period of time, if we get a little bit imprecise it means we have to have more subjects that are

01:02:17,566 --> 01:02:26,332
going to have baseline and follow-up research kidney biopsies; that’s far more expensive and time-consuming than measuring things a little bit

01:02:26,333 --> 01:02:41,333
better. It would be very different if it were a rat study than a human study. As far as the pathology is concerned, this is of course the real

01:02:41,333 --> 01:02:55,333
issue and it’s very disease-dependent, and so the studies will need to differ in design based on a disease. So for example, in diabetes, in Type I

01:02:55,333 --> 01:03:11,399
diabetes where the study of three glomeruli provides an explanation of 65% of their variability in proteinuria and glomerular filtration rate, you’re

01:03:11,400 --> 01:03:21,966
not getting a huge amount of noise from the imprecision that pathology measurement provides, whereas in other diseases that may not

01:03:21,966 --> 01:03:36,332
work at all. So, I think that we need to look at diabetes and hypertension and vasculitis separately and use different strategies to unravel

01:03:36,333 --> 01:03:41,166
them; they won’t all respond to the same strategy, I don’t think.

01:03:41,166 --> 01:03:51,932
JOHN BERTRAM: Look, I agree. I hope I didn’t come across as trying to say, you know, there’s a one technique per whatever. I think it’s going to

01:03:51,933 --> 01:04:01,899
depend very much…the first divide is really clinical biopsies versus, you know, your golden beautifully prepared mouse or rat; it’s totally

01:04:01,900 --> 01:04:09,333
different. Then you’ve got your different pathologies and the kinds of things as I answered in response to Agnes’ question. In some disease

01:04:09,333 --> 01:04:20,133
states, yeah, 3 glomeruli might do the trick, others maybe it’s going to be 9 or 10 or more and the same to do with, you know, we have been talking

01:04:20,133 --> 01:04:32,966
about quantifying interstitial fibrosis or capillaries and so forth. So no, it’s not a one…I didn’t try to get across the message with the dart board that

01:04:32,966 --> 01:04:36,766
there’s one perfect answer to all of this.
MICHAEL MAUER: No, that didn’t come across to

01:04:36,766 --> 01:04:40,266
me at all.
JOHN BERTRAM: Okay. Case-by-case or

01:04:40,266 --> 01:04:43,932
disease-by-disease basis. Absolutely.
MICHAEL MAUER: It wasn’t coming across at all

01:04:43,933 --> 01:04:54,866
from what you were saying but I think it’s more the focus of the impact of pathology and disease and that’s going to demand different levels of

01:04:54,866 --> 01:05:04,532
accuracy, precision, different numbers of subjects, depending on the disease and the question.

01:05:04,533 --> 01:05:08,299
JOHN BERTRAM: And the efficiency of the whole process.

01:05:08,400 --> 01:05:16,066
ROBERT BACALLAO: Lovely talk. Bob Bacallao from Indiana University. I just have a couple of technical questions for my own edification so I

01:05:16,066 --> 01:05:24,166
can think about this differently. I spent a fair amount of time in the early days of confocal volumetric reconstruction working out problems

01:05:24,166 --> 01:05:36,099
associated with fixation to make sure that fixation techniques left you with an accurate spatial representation of what goes on in the cell. So for

01:05:36,100 --> 01:05:48,000
your rat studies, are these perfused fixed or are these kidneys sized and then placed in a fixative? Then for your human studies I imagine, since

01:05:48,000 --> 01:05:58,100
these are autopsies, was there a perfusion event that occurred or was there an excision of the organ and placed in just straight paraffin?

01:05:58,100 --> 01:06:07,600
JOHN BERTRAM: In the rat studies, some of them I showed, they were perfusion fixed, yes. I can’t remember…well, I think the study where we

01:06:07,600 --> 01:06:20,100
counted with electromicroscopy we did some cell counting. I showed you that disector pair; boy, that was 20 years ago. I think we kind of…I’m not

01:06:20,100 --> 01:06:30,833
exactly sure what the initial perfusate was but then the minute we finished that we put them in different fixatives, whether we were going to EM

01:06:30,833 --> 01:06:39,733
or to LM; it was kind of unusual because we wanted everything out of the one kidney. The human material is autopsy, it’s perfusion-fixed;

01:06:39,733 --> 01:06:45,733
the shrinkage and swelling could be very variable.

01:06:45,733 --> 01:06:48,333
JOHN BERTRAM: It doesn’t matter if you’re after

01:06:48,333 --> 01:06:56,833
total number, there’s a million glomeruli in the kidney or 500 podocytes in the glomerulus. I don’t care if the glomerulus gets this big, as long as I

01:06:56,833 --> 01:07:05,633
can still recognize those 500 podocytes, so it doesn’t actually matter with the fractionator design. We just count the number in a known

01:07:05,633 --> 01:07:09,833
fraction so it can shrink, it can distort, it can swell; it doesn’t matter.

01:07:09,833 --> 01:07:17,133
ROBERT BACALLAO: But what if it’s a homogenous process? So, you’ll have regions where you had shrinking and other regions,

01:07:17,133 --> 01:07:27,033
because we have a tremendous gradient of electrocytes that can really change the geometry locally.

01:07:27,033 --> 01:07:37,233
JOHN BERTRAM: I guess I haven’t studied that particularly. If we sample carefully enough and my answer’s going to be—which probably

01:07:37,233 --> 01:07:46,033
doesn’t convince you and isn’t convincing me entirely—but if we sample…what we are doing is sampling uniformly the kidney across the whole

01:07:46,033 --> 01:07:57,633
kidney when we’re taking out pieces, so whether it’s upper pole, hilar region, in a cortex, out of cortex, lower pole. Every point, if you will, in that

01:07:57,633 --> 01:08:08,633
kidney has to have an equal chance of being sampled. So, you had an equal chance, I guess, of getting well perfused, less perfused, shrunk a

01:08:08,633 --> 01:08:17,233
bit, swelled a bit, etc. and the same with the glomerulus. Every piece should have the same chance. But yes, if those variations in

01:08:17,233 --> 01:08:24,033
dimensional changes is happening, even so, that will contribute to your variants; there’s no question.

01:08:24,033 --> 01:08:29,033
KEVIN BENNETT: Thanks, John. In your counting

01:08:29,033 --> 01:08:35,733
technique when you measure glomeruli or pieces of glomeruli that are next to each other—and I’m sure you’ve thought a lot about this—are there

01:08:35,733 --> 01:08:43,033
cases where you accidentally call something a glomerulus, like part of one glomerulus that is actually another glomerulus? In other words, they

01:08:43,033 --> 01:08:46,933
look connected because of the way that you sliced?

01:08:46,933 --> 01:08:51,533
JOHN BERTRAM: You mean with histology and with stereology?

01:08:51,533 --> 01:08:53,533
JOHN BERTRAM: Well, I wouldn’t say every

01:08:53,533 --> 01:08:59,666
glomerulus we’ve ever counted we got it 100% right, I suppose. I actually don’t count them, as you’d appreciate, so maybe I’m not the best

01:08:59,666 --> 01:09:10,766
person to ask. I will say what we do. We work out on quality control, like this study of the human kidney took 10 or more years with Wendy and

01:09:10,766 --> 01:09:18,666
Mike Hughson and so forth. We had one woman do the first 50 and decided there was more to life and then somebody else came along and did the

01:09:18,666 --> 01:09:28,466
rest. So, I was concerned that, first of all, those two counters counted the same and I was also concerned that Rebecca, who did the last

01:09:28,466 --> 01:09:37,466
370-odd did not change with time and did not change that counting approach. So, we’d go back and measure standard kidneys—every 20-odd

01:09:37,466 --> 01:09:44,966
kidneys—to make sure she wasn’t drifting, and we do the same in our rat and our mouse. We have laboratory standards that when we train a

01:09:44,966 --> 01:09:52,066
new person we always go back to that and make sure they get it right, and when they think they’ve got it right we say, “That’s fantastic, do it again,”

01:09:52,066 --> 01:09:59,666
and they love that, and then I always slip that into the study as they’re going along. So, we’re constantly trying to make sure. So now, okay,

01:09:59,666 --> 01:10:05,766
there might be some miscounting—there has to be—but I don’t think it’s confounding our findings.

01:10:05,766 --> 01:10:21,666
JULIE INGELFINGER: Hi. This was the most wonderful talk on the subject I’ve ever heard. I’m Julie Ingelfinger and I had some questions not

01:10:21,666 --> 01:10:37,466
about glomeruli but whether in doing this work you have or anyone has tried to look at whether some glomeruli are atubular and whether there’s

01:10:37,466 --> 01:10:48,866
any way to count those and also whether you’ve stored images and are able to go through any of these glomeruli dynamically, which would be

01:10:48,866 --> 01:10:53,766
visually interesting and might be informative.
JOHN BERTRAM: Thank you for your nice

01:10:53,766 --> 01:11:06,066
comment; that means a lot. We’ve just counted glomeruli to this point in all of our studies whether they’re tubular or atubular and we haven’t noticed

01:11:06,066 --> 01:11:14,966
the latter, but I know they’re not easy to spot, necessarily. In a lot of the work we’ve done in the last four or five years on the individual glomerular

01:11:14,966 --> 01:11:22,366
volume—you know, where we estimated the volumes of 30 glomeruli in each kidney—we have serial sections and more recently we’ve captured

01:11:22,366 --> 01:11:32,266
those with using virtual microscopy. We only have every second section from memory; the 10micrometer glycolmethacrylate sections. So,

01:11:32,266 --> 01:11:41,566
they are available to go through and look at those stacks of sections to see whether there is a patent tubule or not. So, we haven’t done that.

01:11:41,566 --> 01:11:53,399
I’ve been talking to Bob Chevalier who’s over here trying to do that. It’s not a trivial task, even with virtual microscopy, but it is something we’d

01:11:53,400 --> 01:12:06,400
like to do. The material’s kind of there to do it, and if we haven’t already virtually captured it, we’ve certainly got the sections archived.

01:12:06,400 --> 01:12:09,600
MICHAEL MAUER: If I could just comment on

01:12:09,600 --> 01:12:23,233
Julie’s question which is a very important one, actually Behzad Najafian did this work, but in diabetic patients we saw virtually no atubular

01:12:23,233 --> 01:12:36,833
glomeruli prior to them becoming overtly proteinuric. And then in proteinuric subjects, glomerular tubular junction abnormalities were

01:12:36,833 --> 01:12:50,933
extremely common with various levels of abnormalities including adhesions at that level, and about 15 or 20 percent of the glomeruli were

01:12:50,933 --> 01:13:06,666
atubular, they were not sclerosed, and another similar percent were totally obstructed by the tip lesion, so about 30 or 40 percent of the glomeruli

01:13:06,666 --> 01:13:16,866
that looked functional were not functional by virtue of either having the tubular pole completely obstructed or being atubular. So, that’s a very

01:13:16,866 --> 01:13:24,332
important contributor to disease progression, we think, at least in diabetes.

Date Last Updated: 10/3/2012

General Inquiries may be addressed to:
Office of Communications and Public Liaison
Building 31, Rm 9A06
31 Center Drive, MSC 2560
Bethesda, MD 20892-2560
Phone: 301.496.3583