Kidney International, Vol. 45 (1994), pp.

1668—1672

Estimating glomerular number in situ using magnetic resonance

imaging and biopsy
JOHN M. BASGEN, MICHAEL W. STEFFES, ARTHUR E. STILLMAN, and S. MICHAEL MAUER

University of Minnesota Medical School, Minneapolis, Minnesota, USA

Estimating glomerular number in situ using magnetic resonance imag- dogs to test whether a useful estimate of glomerular number
ing and biopsy. In the past researchers have used an estimate of one could be obtained from combining MRI [18] with renal biopsy.
million as the number of glomeruli in each human kidney. However,
recent work on excised kidneys has demonstrated a large variation in Methods
giomerular number from one person to another (330,000 to 1,400,000)
per kidney. Theoretically an in situ estimate of giomerular number Five mongrel dogs each weighing approximately 20 kilograms
could be obtained if renal cortical volume, volume density of glomeruli
were obtained from the Department of Surgery, University of
per cortex [Vv(glom/cortex)] and mean glomerular volume are known.
We used a dog model to demonstrate that an accurate estimate of Minnesota. The dogs had previously been used for surgical
cortical volume could be obtained in situ using magnetic resonance procedures, but the kidneys were not affected. The University
imaging (MRI). Vv(glomlcortex) and mean glomerular volume were of Minnesota's Animal Care and Usage Committee gave per-
obtained from needle biopsies. An independent and more direct method mission to use dogs for this project. The dogs were anesthetized
(the fractionator) was used to validate the estimate of glomerular with pentobarbital, placed on their left side and introduced into
number obtained using MRI and renal biopsy. On average there was
very good agreement between the fractionator method (379,000 the MRI magnet.
40,000) and the MRI/renal biopsy method (376,000 108,000) for the 10
dog kidneys measured; however we found up to a 36% difference Magnetic resonance imaging
between the two methods in an individual kidney. Nonetheless, the All images were obtained on a 1.5 Tesla imaging system
estimate from the MRI/renal biopsy method has more precision than the
assumption that there are one million glomeruli per human kidney. (Magnetom SP; Siemens, Iselin, New Jersey, USA) using a
Helmholtz coil. Five millimeter Tl-weighted images were ob-
tained in the axial plane utilizing the following parameters: field
of view (FOV) of 220 mm; matrix of 192 x 256 elements;
Quantitative studies of glomerular structure [1—71 and studies repetition time (TR) of 700 msec; echo time (TE) of 22 msec and
of glomerular structure-functional relationships [8—161 have 3 excitations. Gradient motion rephasing was utilized to mini-
added to our understanding of several renal diseases. Often mize motion-related artifact. The number of films per kidney
these works have assumed a constant number of glomeruli per varied from 8 to 12 depending on the length of the kidney. Films
kidney among different individuals. However, by careful objec- were printed at one-half actual size.
tive measurements glomerular number per kidney in normal
humans ranges as widely as from 330,000 to 1,400,000 [17]. The Kidney biopsy
ability to estimate, even with moderate precision, glomerular
Following MRI the kidneys were excised, and each kidney
number in a given patient could substantially improve our was biopsied at six sites using a TRU-CUTTM biopsy needle
understanding of human renal disease. (Baxter-Travenol, Inc., Chicago, Illinois, USA). Two biopsy
Development of a method to estimate glomeruli number from cores were taken from both poles entered laterally. Two addi-
small biopsy specimens obtained in the study of living humans tional cores were taken from the mid portion of the lateral side
remains a difficult challenge. One approach to this problem is to entered caudally. The biopsy cores were placed in 10% buffered
obtain an estimate of glomerular number by multiplying the formalin for several days. After fixation each biopsy sample
cortical volume of a kidney by the fraction of renal cortex made was dehydrated through a series of ethanol solutions to 95%
up of glomeruli [Vv(glomlcortex)} and dividing this result by the ethanol and embedded in Historesin® (Leica, Inc., Deerfield,
mean glomerular volume for that kidney. Vv(glomfcortex) and Illinois, USA). One 5 m thick section was cut from each
mean glomerular volume could be estimated from a renal biopsy using a Reichert-Jung 2050 Supercut microtome (Leica,
biopsy, while cortical volume per kidney in situ could be Inc.) with Ralph knives and stained with toluidine blue.
obtained by magnetic resonance imaging (MRI). We studied
Structural parameters
Kidney and cortical volumes. Kidney and renal cortical
Received for publication September 24, 1993 volumes were estimated from the MRI films using the Cavalieri
and in revised form January 6, 1994 principle which states that the volume of a body can be
Accepted for publication January 6, 1994 estimated without bias, and independent of shape and size, by
© 1994 by the International Society of Nephrology utilizing a series of slices through that body, and knowing the

1668
 Basgen et a!: Glomerular number from MRI and biopsy 1669

level of the tissue in micrometers. P is the number of points

average thickness of the slices [19]. To estimate volume, one
measures area of the body on each slice and calculates the hitting glomerular profiles; N is the number of glomerular
volume using the equation: profiles hit by points; 1.38 and 1.01 are correction factors
related to shape and size variability of the glomeruli [20]. Since
n—I
mean glomerular volume was measured on tissue not under
volume = t * area1 (Eq. 1) ambient arterial pressure, we divided the calculated glomerular
i= I volumes by 0.69 which is the ratio of the kidney volume after
excision to the in vivo kidney volume (see Results section). We
where t is the average slice thickness; area, is the area of the used the results of these calculations as estimates of mean in
body on a slice; and n is the number of slices. situ glomerular volumes.
Kidney and renal cortical volumes were estimated from the Glomerular number. The number of glomeruli per kidney was
MRI films by placing the films on a horizontal viewing box. A calculated using the
tessellation of points was placed over each kidney image, and
the number of points overlying kidney and the number of points glomerular number =
overlying cortex were noted. The areas of whole kidney and renal cortical volume * Vv(glom/cortex)
renal cortex of each slice were estimated by the equation: (Eq. 5)
mean in situ glomerular volume
f5Ø\2
area = * P mm2
(Eq. 2) Glomerular number for a kidney was calculated using only the
estimates of Vv(glomlcortex) and mean glomerular volume
where 5 is the distance between grid points in millimeters; 0.5 is from one of the biopsy cores taken from the lower pole.
the reduction factor of the films; and P is the number of points
hitting kidney or cortex. Volume of kidney or renal cortex was Validation of parameters
calculated using the Cavalieri equation. Kidney and cortical volume. To validate the estimate of
Volume density of glomeruli per cortex. An Olympus BM-2 kidney and cortical volumes obtained from the MRI films each
microscope (Olympus Optical Co., Ltd., Tokyo, Japan), fitted excised kidney was cut into several (range 7 to 14) five
with a mirror attached to the monocular, was used to estimate millimeter thick slices. The Cavalieri method was used to
Vv(glomlcortex). Microscopic images of the biopsy sections estimate kidney and cortical volume on these freshly cut slices.
were projected at a final image magnification of 150 X onto a The same grid used on the MRI films was placed over the left
laboratory bench top on which a double lattice grid was placed. cut surface of each kidney slice and the number of points hitting
The number of fine points hitting glomeruli and the number of kidney and hitting cortex were counted. After the slices were
coarse points hitting cortex in the entire cortical area of a measured, they were placed in 10% buffered formalin for at
section were counted. Vv(glom/cortex) was calculated as fol- least one week.
lows: Glomerular number. The fractionator technique [17, 21] was
used to obtain an independent estimate of glomerular number.
Pglomeruli This method samples, without bias, a known fraction of a whole
Vv(glom/cortex) (Eq 3) organ; all of the particles in the fraction are counted, and finally
1cortex * 16
the number of particles counted is divided by the sample
fraction (see below).
1gIomeru1us is the number of fine points overlying glomeruli and For these dogs one-third of the kidney slices was selected;
'>cortex is the number of coarse points overlying cortex. There the medulla was cut out and discarded; and the remaining
were 16 fine points (0.5 mm apart) for each coarse point (2.0 mm cortex cut into 5 millimeter thick strips. One-fifth of the strips
apart) on the grid. Biopsies were not obtained from the kidneys was selected and cut into 5 x 5 x 2.5 millimeter blocks. One-
of dog 2, therefore for that dog, the sections used for the fifteenth of the blocks was selected. Several blocks of tissue
fractionator technique (see below) were used to measure this were embedded in one block of Historesin®. Fifteen microme-
parameter. ter thick sections were cut from the block until all tissue was
Mean glomerular volume. Mean glomerular volume was exhaustively sectioned. One-tenth of the sections was selected
estimated by the method of Weibel and Gomez [20]. The same for counting. The microscopic images of the sections were
biopsy section and magnification were used as described above, projected onto the bench top and the stage controls advanced
but with a different grid having points 6.35 mm apart. The such that 0.4 of the tissue on a slide was used for counting
number of grid points falling on each glomerular profile was glomeruli. All the glomeruli within this fraction were counted
noted, and mean glomerular volume was calculated as: using the disector technique, which allows one to estimate
mean glomerular volume = particle number from tissue sections independent of particle
size or shape [21, 22]. The total number of glomeruli in a kidney
was estimated using the
Ii' 6350 \2* 1.38
* m3 (Eq. 4) Q-
[magnitication) glomerular number = (Eq. 6)
Sl * S2 * S3 * S4 * S5
The first factor of the equation converts the 6.35 millimeters where S 1, S2, S3, S4 and S5 are known sampling fractions (for
between grid points to the area each grid point represents at the example, 1/3, 1/5, 1/15, 1/10, 0.4) and Q is the number of
1670 Basgen et a!: Glomerular number from MRI and biopsy

60

50

40
a)

'4
E
0> 30

t
00 20

I
Fig. 1. MRI i,nage of dog kidney. Cortical region corresponds to the
10

darker peripheral area (arrows) in this transverse section. Magnification

is 1.20 x.
1 2 3 4 5 6 7 8 9 10
glomeruli counted. The fraction of kidney used for counting was Kidney number
1/5625. Fig. 2. Comparison ofcorrected cortical volume estimated from MRJ
films (open circles) and from slices of the excised kidneys (closed
Results circles).
Kidney and cortical volumes
MRI films permitted reasonably precise estimates of kidney
volume in situ using the Cavalieri technique; however, renal
volumes estimated from the kidney slices, not under ambient — 31%. Again there was no systematic difference from site to
perfusion pressure, were consistently 31 4% (mean SD) site within a kidney. There was agreement between the two
smaller than the renal volumes estimated from MRI films. The kidneys from an individual dog; however, there was consider-
peripheral darker areas of the kidney on the MRI films were able variation in glomerular volume from one dog to another.
interpreted as cortex (Fig. 1). The estimates of renal cortical
volume from the kidney slices were also on the average 31 6% Glomerular number
smaller than the estimates from the MRI films. The calculated estimate of glomerular number using MRI and
Assuming the diminished kidney and cortical volumes in the one biopsy core from the lower pole (376,000 108,000)
excised kidney slices resulted from the consequences of the loss compared well to the estimate of glomerular number obtained
of physiologic perfusion pressure and blood, the cortical vol- using the fractionator technique (379,000 40,000). However
ume obtained from the slices was corrected by dividing by 0.69 the estimates for an individual kidney may differ by as much as
(1.00 — 0.31). On average there was excellent agreement for 36% (Fig. 3).
cortical volume (44 6 for MRI and 44 9 for slices), although
for an individual kidney there may be up to a 16% disagreement Discussion
(Fig. 2). We used MRI and kidney biopsy in dogs to estimate the
number of glomeruli within a kidney in situ and have compared
Volume density of glomeruli per cortex this technique to studies performed on kidneys from these same
Variability of Vv(glom/cortex) throughout the kidney was dogs for independent estimates of cortical volume and glomer-
determined on six biopsies taken from different regions of the ular number using established stereologic methods. MRI images
kidney. There was considerable variation of Vv(glom/cortex) produced good demarcation of the kidney cortex as indicated
from one biopsy site to another as demonstrated by the coeffi- by the correspondence to direct measures on the excised
cients of variation which ranged from 14 to 36%. However, kidneys. Measuring the kidney and cortical areas on these
there was no systematic difference in Vv(glomlcortex) from one images and knowing the distance between the images, we could
biopsy site to another within a kidney. Using all six measure- calculate the volumes of the kidney and the renal cortex.
ments in each kidney the average Vv(glom/cortex) was consis- Miller and Meyer [23] have shown that glomerular size
tent from one kidney to another (0.042 0.005). decreases in rats when renal perfusion pressure is reduced.
Presumably this is the reason the kidney and cortical volumes
Mean glomerular volume measured on the kidney slices were 31% smaller than the
There was also considerable variability in the estimates of volumes measured on the MRI films. We therefore divided the
mean glomerular volume from the multiple biopsies; coeffi- cortical volumes from the kidney slices by 0.69 (1.00 — 0.31) to
cients of variation among biopsies within a dog ranged from 10 obtain a corrected cortical volume corresponding to the in situ
 Basgen et a!: Glomerular number from MRI and biopsy 1671

600000 data from two sites, therefore the data from only one lower pole
biopsy site was used to calculate glomerular number.
Mean glomerular volume also varied from site to site within a
500000 kidney, and there was considerable variation from dog to dog.
These variations could be due to imprecision of our measuring
technique, true biological variation from dog to dog or both and
400000 our studies do not discriminate between these possibilities. In
the past we have found that measuring this parameter on the

I !!!
E
limited number of glomerular profiles which may be available
300000 from needle biopsies can result in an imprecise estimate of
glomerular volume [27]. The biopsies used in this study con-
tained an average of 17 profiles. If more glomeruli were avail-
200000 able per kidney, it would be possible to obtain a more precise
estimate of glomerular volume and therefore a more precise
estimate of kidney number per kidney.
100000 Often an estimate of 1,000,000 glomeruli per human kidney
has been used by researchers. However, recent studies by
Nyengaard and Bendtsen [17] have shown that in human
0 kidneys, glomerular number may vary from 330,000 to
2 3456 9 10 1,400,000. Thus, using 1 million can result in a 300% over
estimation of glomerular number in some kidneys. The MRI
Kidney number with biopsy technique underestimated glomerular number by an
average of only 1% in the 10 kidneys studied; however, there
Fig. 3. Comparison of glomerular number estimated using MRI films
and kidney biopsies (open circles) and from the fractionator technique was a discrepancy of up to 36% for an individual kidney. The
(closed circles) in ten dog kidneys. imprecision in the estimate of glomerular number by our
method is likely to be related to the imprecision of our estimates
of Vv(glomlcortex) and glomerular volume. In addition the need
to multiply and divide these estimates compounds the variabil-
values. Whether this same correction factor is applicable to ity in the final result.
humans or is applicable to diseased as well as healthy kidneys This technique may be applicable to human kidneys. In
needs to be determined. However, the usual ratio of cortical patients without advanced renal scarring we believe there
volume to total renal volume is maintained in diseased kidneys would be approximately the same precision in estimating gb-
[24]. On average the MRI technique estimated cortical volume merular number as found in the normal dog kidneys. In human
precisely, although in an individual there can be a discrepancy kidneys with more advanced pathology and, especially in
of as much as 16%. Thus, a good estimate of cortical volume in diseases with focal changes, there may be a decrease in the
situ can be obtained from measurements on MRI films. precision of the estimate. Nevertheless, a large number of
The biopsy specimens were embedded in Historesin® to glomeruli may be lost with advanced renal disease, permitting
reduce the shrinkage that occurs when tissue is embedded in the loss to be detected even with a decrease in precision of the
paraffin [25]. The extent of tissue shrinkage that occurred method. Assuming these same methods when applied to human
during tissue processing in our studies was not measured. kidneys would provide a similar maximal intraindividual error
However, if shrinkage occurred, and assuming this occurred of approximately 36%, this technique might greatly improved
equally in glomeruli and non-glomerular cortical structures, the ability to estimate the number of glomeruli in kidneys in situ
then Vv(glom/cortex) would remain constant, mean glomerular over the currently accepted assumptions which can subsume
volume would be reduced and a systematic overestimation of errors in excess of 300% [17]. In this regard we performed a
glomerular number would be expected. In fact, we did not power estimate to determine the number of subjects per group
observe a systematic overestimation of glomerular number needed to detect differences in the number of glomeruli between
compared to the fractionator method, a technique independent two groups of people. Presuming a normal group would have a
of changes in tissue dimensions. This suggests that tissue fixed distribution of values for glomerular number as per Nyengaard
in formalin and embedded in Historesin® does not undergo and Bendtsen [17], then approximately 20 subjects per group
significant shrinkage. would detect a 20 to 25% difference in glomerular number.
The considerable variation of Vv(glomlcortex) within a kid- Studies are underway in humans to confirm these assumptions.
ney is due to the small sample size provided by a needle biopsy
and to the heterogeneous distribution of glomeruli in the cortex Acknowledgments
[261. Nonetheless, the mean value for this parameter derived Preliminary results of the work were presented at the Eighth Inter-
from six biopsy cores was consistent from one kidney to national Congress for Stereology, Irvine, California, USA, August
another. However, since the ultimate goal was to develop a 25—29, 1991 and the Twenty-Sixth Annual Meeting of the American
technique applicable to patients undergoing percutaneous renal Society of Nephrology, Boston, Massachusetts, November 14—17,
1993. Financial support was provided by the National Institutes of
biopsies we compared estimates of Vv(glomlcortex) using mea- Health (DK43605) and the American Diabetes Association, Minnesota
surements from one and then from both lower pole biopsy sites. Affiliate, Inc. and the Juvenile Diabetes Foundation. The authors are
There was no increase in the precision of the estimate using the grateful to Bruce Wigness and Dale Arens for providing surgical help,
1672 Basgen et a!: Glomerular number from MRI and biopsy
and Dan Busian and the MRI technologists, Department of Radiology,
University of Minnesota, for obtaining the MRI images of the dog
kidneys. Dr. James Walker's suggestions regarding the manuscript are
also appreciated.
Reprint requests to John M. Basgen, Department of Pediatrics,
University of Minnesota, Box 73 UMHC, 320 Delaware Street SE,
Minneapolis, Minnesota 55455, USA.
References
1. STEFFES MW, BARBOSA J, BASOEN JM, SUTHERLAND DER,
NAJARIAN JS, MAUER SM: Quantitative glomerular morphology of
the normal human kidney. Lab Invest 49:82—86, 1983
2. HAYNES WDG: The normal human renal glomerulus. Virchows Arc
B Cell Pathol 35:133—158, 1981
3. ØSTERBY R, GUNDERSEN HJG, NYBERO 0, AURELL M: Advanced
diabetic glomerulopathy: Quantitative structural characterization
of nonoccluded glomeruli. Diabetes 36:612-619, 1987
4. MCLACHLAN MSF, GUTHRIE JC, ANDERSON CK, FULKER MJ:
Vascular and glomerular changes in the ageing kidney. J Patho!
121:65—78, 1977
5. DUNNILL MS, HALLEY W: Some observations on the quantitative
anatomy of the kidney. JPathol 110:113—121, 1973
6. VIMTRUP BJ: On the number, shape, structure, and surface area of
the glomeruli in the kidneys of man and mammals, Am J Anat
41:123—151, 1928
7. BRENNER BM, GARCIA DL, ANDERSON S: Giomeruli and blood
pressure less of one, more the other? Am J Hypertens 1:335—347,
1988
8. HIROSE K, TSUCHIDA H, ØSTERBY R, GUNDERSEN HJG: A strong
correlation between glomerular filtration rate and filtration surface
in diabetic kidney hyperfunction. Lab Invest 43:434—437, 1980
9. BADER R, BADER H, GRUND KE, MACKENSEN-HAEN S, CHRIST H,
BORLE A: Structure and function of the kidney in diabetic glomer-
uloscierosis: Correlations between morphological and functional
parameters. Pathol Res Practice 167:204—216, 1980
10. MAUER SM, STEFFES MW, ELLIS EN, SUTHERLAND DER,
BROWN DM, GOETZ FC: Structural-functional relationships in
diabetic nephropathy. J Clin Invest 74:1143—1155, 1984
11. SHEMESH 0, Ross JC, DEEN WM, GRANT GW, MYERS BD: Nature
of the glomerular capillary injury in human membranous glomeru-
lopathy. J Clin Invest 77:868—877, 1986
12. ELLIS EN, STEFFES MW, GOETZ FC, SUTHERLAND DER, MAUER
SM: Glomerular filtration surface in type I diabetes mellitus.
Kidney mt 29:889—894, 1986
13. ØSTERBY R. PARVING H-H, NYBERG G. HOMMEL E. JØRGENSEN
HE, LØKKEGAARD H, SVALANDER C: A strong correlation between
glomerular filtration rate and filtration surface in diabetic nephrop-
athy. Diabetologia 31:265—270, 1988
14. ØSTERBY R, PARVING H-H, HOMMEL E, JØRGENSEN HE, LØiutiao-
AARD H: Glomerular structure and function in diabetic nephropa-
thy: Early to advanced stages. Diabetes 39:1057—1063, 1990
15. WALKER JD, CLOSE CF, JONES SL, RAFFTERY M, KEEN H,
VIBERTI 0, ØSTERBY R: Glomerular structure in type-i (insulin-
dependent) diabetic patients with normo- and microalbuminuria.
Kidney mt 41 :741—748, 1992
16. HAYHSHI H, KARASAWA R, INN H, SAITOU T, UENO M, NISHI S.
SUZUKI Y, OGIN0 S, MARUYAMA Y, KOUDA Y, ARAKAWA M: An
electron microscopic study of glomeruli in Japanese patients with
non-insulin dependent diabetes mellitus. Kidney mt 41:749—757,
1992
17. NYENGAARD JR, BENDTSEN TF: Glomerular number and size in
relation to age, kidney weight, and body surface in normal man.
Anat Rec 232:194—201, 1992
18. HRICAK H, NEWHOUSE JH: MR imaging of the kidney. Radio! Clin
North Am 22:287—296, 1984
19. GUNDERSEN HJG, JENSEN EB: The efficiency of sampling in
stereology and its prediction. J Microsc 147:229—263, 1987
20. HIRosE K, ØSTERBY R, NOZAWA M, GUNDERSEN HJG: Develop-
ment of glomerular lesions in experimental long-term diabetes in
the rat. Kidney Int 21:689—695, 1982
21. GUNDERSEN HJG: Stereology of arbitrary particles: A review of
unbiased number and size estimators and the presentation of some
new ones, in memory of William R. Thompson. J Microsc 143:3—45,
1986
22. STERIO DC: The unbiased estimation of number and sizes of
arbitrary particles using the disector. I Microsc 134:127—136, 1984
23. MILLER PL, MEYER TW: Effects of tissue preparation on glomer-
ular volume and capillary structure in the rat. Lab Invest 63:862—
866, 1990
24. WILTRAKIS MG: Correspondence between glomerular size and
renal volumes in atrophic, normal, and hypertrophic human kid-
neys. Acta Path Microbiol Scand (Section A) 80:801—811, 1972
25. HANSTEDE JG, GERRITS P0: The effects of embedding in water-
soluble plastics on the final dimensions of liver sections. J Microsc
131:79—86, 1983
26. TISHER CC, MADSEN KM: Anatomy of the kidney (Chapter 1), in
The Kidney, edited by BRENNER BM, RECTOR FC (3rd ed) Phila-
delphia, W. B. Saunders Company, 1986, pp 3—60
27. BILOUS RW, MAUER SM, BASGEN JM, STEFFES MW: Estimation
of mean glomerular volume in patients with insulin-dependent
diabetes mellitus. Kidney mt 32:930—932, 1987