Kidney International, Vol. 32 (1987), pp. 736—741

Observations of glomerular epithelial cell structure in patientswith type I diabetes mellitus

EILEEN N. ELLIS, MICHAEL W. STEFFES, BLANCHE CHAVERS, and S. MICHAEL MAUER

Department of Pediatrics and Laboratory Medicine and Pathology, University of Minnesota, Minneapolis, Minnesota, and University ofArkansas for Medical Sciences and Arkansas Children's Hospital, Little Rock, Arkansas, USA

Observations of glomerular epithelial cell structure in patients withtype I diabetes mellitus. Overt proteinuria is a hallmark of diabeticnephropathy while microalbuminuria is thought to be a predictor oflater onset of diabetic nephropathy. Yet the mechanisms for abnormalurinary protein leak in diabetes have not been defined. We studied 28patients with type I diabetes for 7 to 33 years. Creatinine clearance,urinary albumin excretion rate (UAE), and multiple blood pressureswere obtained in each patient. A renal biopsy was performed in eachpatient and in 28 normal subjects, Quantitative stereology was used todetermine foot process (FP) width, filtration slit length density (FSLv)and filtration slit length/glomerulus (FSLG). FP width was slightly widerthan normal in diabetic patients with UAE <250 mg124 hr while FP wassignificantly wider than both of these groups in diabetics with UAE >250 mg/24 hr. FSLv and FSLG were similar in normals and diabeticswith UAE < 250 mg/24 hr but both were reduced in diabetics with UAE> 250 mg/24 hr. UAE correlated with FP width (P < 0.05), FSL0 (P <0.01) and most precisely and FSLv (P < 0.001). Diabetics withmicroalbuminuria had values for all the structural parameters measuredhere not different from diabetics with UAE in the normal range.Perturbations of epithelial cell structure are present in diabetes mellitusespecially in patients with nephropathy. The exact relationships be-tween albuminuria and epithelial cell structure remains to be elucidated.

Overt diabetic nephropathy is typically ushered in by thedevelopment of dipstick—positive proteinuria and then pro-gresses through hypertension and declining glomerular filtrationrate (GFR) to terminal uremia [1—3]. Microalbuminuria (in-creased urinary albumin excretion below the range of detect-ability by standard laboratory tests) is a strong predictor of thelater development of overt diabetic nephropathy [4—7]. Thus,proteinuria is an important subject for study in the diabeticpatient. Yet, the mechanisms for microalbuminuria and overtproteinuria in diabetes remain unclear and precise renal struc-tural correlates of the abnormal albumin leakage have not beendefined.

Since protein leakage in patients with minimal change ne-phrotic syndrome [8] and in animals with experimental ne-phrotic syndrome [9] has been associated with demonstrablewidening of the glomerular epithelial foot processes, and sinceGFR is inversely correlated to foot process width in minimalchange nephrotic patients [8], we investigated whether epithe-

Received for publication November 21, 1986and in revised form April 2, 1987

© 1987 by the International Society of Nephrology

hal cell structural parameters were related to albuminuria andGFR in type I diabetes mellitus.

Methods

Twenty-eight patients aged 20 to 55 years (31 7 years, XSD) with type I diabetes mellitus for 7 to 33 years (19 6 years)were studied in the Clinical Research Center at the Universityof Minnesota. All patients were evaluated as possible candi-dates for pancreas transplantation. These patients represent asubset of patients previously reported by us [10, 11]; in thecurrent study we evaluated all patients who had adequatenumbers of glomeruli available on renal biopsy for glomerularvolume determination as well as urine specimens available forfluorometric albumin assay. All studies were approved by theCommittee on the Use of Human Subjects in Research at theUniversity of Minnesota and each subject gave informed con-sent. Each patient had two or three 24-hour urine collectionscompleted for measurement of creatinine clearance (Car) bystandard laboratory techniques and urinary albumin excretionrate (UAE) measured by fluorometric immunoassay [12]. Thesemeasurements were averaged and expressed as ml/min for Ccrand mg/24 hr for UAE. Each patient had multiple bloodpressure determinations taken during their CRC hospitaliza-tion. Patients were considered hypertensive if their bloodpressures were regularly greater than 140/90 or if they weretaking anti-hypertensive medication; 9 patients were labeled ashypertensive. Creatinine clearance ranged from 35 to 157mllmin while UAE ranged from 3 to 3600 mg/24 hours. Serumcreatinine was less than 2.4 mg/dl in all patients. Patients weredefined as having microalbuminuria if their UAE was 30 to 250mg/24 hr; 5 patients were so defined. Patients were defined ashaving clinical diabetic nephropathy if they had two or more ofthe following: Cr less than 95 mI/mm; UAE greater than 250mg/24 hr; or hypertension. Eight patients fulfilled the criteria forclinical nephropathy (Table 1). One patient had mild hyperten-sion but no other findings of clinical nephropathy.

Each diabetic patient underwent a percutaneous renal bi-opsy. In addition, 28 normal people aged 24 to 52 years (33 8years) serving as living—related kidney donors had kidneybiopsies performed at the time of donation for transplantation[131. Other structural values for these normal people are re-ported elsewhere [13]. All biopsy specimens were immediatelyexamined under a dissecting microscope to insure that glomer-uli were present. One part of the tissue was placed in Zenker's

736

fixative, washed, embedded in paraffin, and processed by usuallaboratory techniques for light microscopy. The other part ofthe tissue was cut into approximately 1 mm cubes and placed in2.5% glutaraldehyde in 0.17 M cacodylate buffer and preparedfor electron microscopy as previously described [10, 111.

Thick (1 Lm) sections were cut from each EM block, stainedwith toluidine blue, and used for selection of the centermostglomerulus of each block in which the entire profile was at least1 tubular diameter from the edge of the block. Three suchglomeruli were selected from each biopsy for thin sectioning.The thin sections were stained with uranyl acetate and leadcitrate for examination with a JEOL 100 CX electron micro-scope (JEOL, Tokyo, Japan). Each glomerulus was enteredrandomly and electron micrographs were taken systematicallythroughout the glomerulus. A carbon replica grating (28,800lines/inch, E.F. Fullam, Inc., Schenectadey, New York, USA)was photographed with each glomerulus to determine finalmagnification. Foot process width, filtration slit length density,and glomerular basement membrane width were measured onelectron micrographs with an approximate final magnification ofx 18,000 while all other measurements were completed onelectron micrographs with an approximate final magnification ofx 6000.

Average glomerular volume was determined on light micro-scopic sections using the method of Hirose and coworkers [14]where average glomerular area is first measured by pointcounting and mean glomerular volume is then calculated. Toinsure that each glomerulus was measured only once, onlysections that were at least 100 m apart were examined. Each

biopsy had at least 20 glomeruli available for examination andthe number examined averaged 50 glomeruli (range: 23 to 126glomeruli).

Glomerular basement membrane (GBM) width and foot proc-ess width were determined by the orthogonal intercept method[15]. The width of the epithelial foot processes was determinedalong the peripheral GBM only. Each foot process which fellwithin an unbiased counting frame was classified using a logreciprocal scale [16]. The number of foot processes in eachclass was used to calculate the harmonic mean foot processwidth and multiplied by 8/(3ir) to give the actual mean footprocess width in nanometers [15, 16]. All filtration slits (Q)thatfell within the same unbiased counting frame were counted todetermine filtration slit length density (Lv) [8, 17]:

where

2xQL =k2 x

k = distance between points on the counting frame (in m)

magnificationand P = number of points failing on the glomerulus.

The filtration slit length density is a mathematical representa-tion of the length of filtration slit available per referencevolume, in this case the glomerulus.

The volume density of the mesangium and the surface densityof the peripheral capillary filtration surface were determined bypoint and intercept counting using standard stereologic tech-

Glomerular epithelial cell in diabetes 737

Table 1. Patient characteristics

AgeDuration of

diabetes Ccr UAEFoot

processYears Years mI/mm mg/24 hr Hypertension width nm

Filtrationslit length

densityFiltration

slit length

-x l03.tm

26 16 59 3600 yes 366 0.20 24723730 14 70 1476 yes 429 0.0937138 22 81 1276 yes 308 0.15

31026 22 60 831 yes 338 0.22

39930 20 135 691 yes 405 0.25

16935 28 35 652 yes 391 0.08

18430 12 91 578 yes 492 0.12

69130 23 154 203 no 307 0.34

22 11 113 179 no 371 0.29

24 23 61 136 yes 356 0.29

48637 21 78 73 no 364 0.36

29 18 94 40 no 297 0.23 316

31 21 102 27 no 279 0.35

27 10 122 27 no 354 0.4023 17 157 26 no 399 0.26 736

24 14 125 24 no 289 0.41 691

20 9 136 21 no 284 0.43 645

25 IS 112 19 yes 288 0.34 65640941 21 134 14 no 383 0.28

40 22 154 12 no 298 0.36 805

26 22 85 12 no 327 0.31 351

33 20 124 10 no 363 0.40 707

36 27 122 8 no 299 0.51 1013

31 20 115 7 no 382 0.35 550

26 7 113 7 no 379 0.38 285

55 33 95 6 no 297 0.34 521

39

34

2214

109

99

53

nono

374

289

0.41

0.62

412

1061

738 Ellis et al

niques [10, 11, 171. The per glomerulus mesangial volume,peripheral capillary filtration surface area, and filtration slitlength were calculated from the glomerular density measuresand mean glomerular volume [11]. The individual values formesangial volume/glomerulus and capillary filtration surface!glomerulus are identical to those reported previously [111.

Statistical analysisRelationships between variables were assessed by linear

regression using the method of least squares and by multipleregression analysis. Differences among groups of patients wereassessed by analysis of variance. Results are expressed a mean

standard deviation. Values of P < 0.01 were consideredsignificant while values of P < 0.05 were considered borderline,these criteria taking into account the number of analysis per-formed.

Results

Epithelial foot process width, filtration slit length density andtotal filtration slit lengthlglomerulus in the normal subjectsaveraged 304 29 nm, 0.41 0.10 jtmItm3, and 576 240 Xl0 m respectively (Fig. 1). Foot processes were slightly widerthan the normals in the diabetic patients with UAE < 250 mg/24hours averaging 332 41 nm, although this difference did notreach statistical significance (P <0.05). Diabetic patients withUAE > 250 mg124 hours had even wider foot processesaveraging 390 61 nm (P < 0.001) compared to either normalsor diabetic subjects with UAE <250 mg/24 hr (Fig. 1). Diabeticpatients with UAE < 250 mg/24 hr did not differ from normalsubjects in filtration slit length density or total filtration slit

length. Diabetic patients with UAE > 250 mg/24 hr had signif-icantly reduced filtration slit length density (P < 0.001) and totalfiltration slit length (P < 0.001) compared to diabetics withUAE < 250 mg124 hour or normal subjects. Diabetics withmicroalbuminuria (that is, UAE 30 to 250 mg/24 hr) had valuesof foot process width, filtration slit length density, and filtrationslit length which were no different from diabetics with normalUAE (Fig. 1).

In the diabetic patients. the log of UAE tended to correlatedirectly with foot process width (r = +0.41, P < 0.05) and didcorrelate significantly with total filtration slit length/glomerulus(r = —0.53, P < 0.01) (Fig. 2, 3). However, the relationshipbetween filtration slit length density and albuminuria was muchmore precise (r = —0.80, P < 0.001, Fig. 4). Creatinineclearance correlated directly with filtration slit length density (r

+0.49, P < 0.01) and total filtration slit length/glomerulus (r+0.58, P < 0.001) but not with foot process width (r =

— 0.21, NS). As clearly expected, filtration slit length!glomerulus was highly correlated with peripheral capillaryfiltration surface/glomerulus (r +0.77, P < 0.001). Thus,similar to our previous studies with peripheral capillary filtra-tion surface [11], filtration slit length!glomerulus correlatedhighly with both total mesangiuin/glomerulus and glomerularvolume together (r = +0.81, P < 0.001).

Discussion

We have previously considered [10, 11] that the diabeticpatients we have studied may not be a precise cross—sectionalrepresentation of renal complications in type I diabetes sincegenerally they have had difficulty in diabetes management or

510 -

490

470

450

430

410

2 390

i 3700C-)

2 350a330

310

290

270

0.70

0.60

0.50

— 0.40-Ca)

CC)

C00.20-

U-

0.10

0.00

t r

1100$

E 1000

900

800

• 700a)E 600• 0

• )• . 500

o •4 CI 0) 400

300 IDI1 0tC I•tI 200

Fig. 1. Foot process width, filtration100 slit—length density, and filtration slit

tengihiglomerutus in normal subjects, diabetic

•patients with UAE < 250 nzg/24 hr anddiabetic patients with UAE > 250 mg/24 hr.

Normal UAE UAE Normal UAE UAE Symbol is: El, those diabetic subjects with<250 >250 <250 >250 microalbuminuria, The lines represent meanDiabetics Diabetics SD, Statistical differences between the groups

L.___iL___J are shown at the bottom. NS is notNS P<0.001 NS P<0.001 significant.

250

0cNormal UAE UAE

<250 >250

DiabeticsLJLJP<0.05 P<0.001

Glomerular epithelial cell in diabetes

I-

Ea)a)

0a)

xa)CE.0a)

a)C

739

have manifested one or more of the major complications ofdiabetes. However, these patients do represent a broad spec-trum of renal structural and functional parameters [10, 11] andthus, there is no reason to doubt the validity of the relationshipsbetween renal function and structure in these studies. Thefluorometric method of measuring urinary albumin is reproduc-ible over a wide range of UAE and is both sensitive and precise[12]. Creatinine clearances in diabetic patients with GFR in therange reported here when carefully performed, provide resultswhich are highly correlated with GFR as measured by inulinclearance [11,18] and plasma disappearance or5 'Cr-EDTA [19].

The foot process width in our normal subjects averaged 30429 nm, larger than the value of 224 nm previously reported byGundersen and coworkers using the same stereologic methods[16]. However, we studied 28 normals while only 5 wereevaluated in the previous work [16]. The five normals evaluatedby Gundersen, Seefelt and Osterby were aged 12 to 28 years,while our normal patients averaged 28 years of age (range 24 to52 years). Further, in another study with 11 patients designatedas normal who had kidneys surgically removed for tumors, footprocess width using these same stereologic methods averaged296 nm [20]. Thus, mean foot process width appears to be about300 nm in normal adult man. It should be pointed out thatassumptions made in the quantitation of foot process width are

clearly inexact. The expression of a mean value for foot processwidth belies the fact that the epithelial cell has primary,secondary and tertiary extensions whose size by qualitativescanning electron microscopy are orders of magnitude different[21]. The harmonic mean used here and elsewhere reduces butdoes not eliminate the problem that foot process width is notnormally distributed.

Nonetheless, qualitative expansion of foot processes is welldocumented in a variety of proteinuric human and experimentalrenal diseases. It has been argued that the negative charge ofthe sialic acid coat of the epithelial cell, together with intracel-lular actin filaments probably maintain normal structure of thefoot processes, and that loss of this negative charge results infusion of foot processes and obliteration of slit diaphragms [221.Epithelial cell foot process widening has been quantitativelydocumented in minimal change nephrotic syndrome [8, 23] andmembranous nephropathy [24] in man, as well as in aminonu-cleoside nephrosis [9], uranyl acetate—induced acute renal fail-ure [251, and experimental diabetes [261 in animals. In minimalchange nephrotic syndrome in man, foot process widening hasbeen correlated with serum albumin [8] and the degree ofproteinuria [23] while urinary albumin excretion in aminonucle-oside nephrosis in rats correlates with foot process width [9]. Inmembranous nephropathy, the frequency of filtration slits cor-

100,000

10,000

1000

100 :

10 :

1— . .0 250 300 350 400 450 500

Foot process width, nm

Fig. 2. The relationship between foot process width and UAE in thediabetic subjects (r = +0.41, P < 0.05).

.S

S.S

100,000

10,000

1000

100

10

...u.uuu I r

0 200 400 600 800 1000

Filtration slit length/glomerulus, x 1O m

Fig. 3. The relationship between filtration slit length/glomerulus andUAE in the diabetic subjects (r = —0.53, P < 0.01).

• S

•

S

aS

S

a SS

S

S

•

• S

740 Ellis et al

between diabetic patients with and without microalbuminuria.However, diabetic patients with overt nephropathy were clearlydifferent from all other groups, and with almost no overlap.Careful inspection of Figure 4 shows that the statisticallysignificant correlation between UAE and filtration slit lengthdensity is largely driven by the patients with clinical proteinuriaand that, in the main, two clusters of patients exist. Absolutefiltration slit length per glomerulus shows a less precise rela-tionship to UAE. This is not surprising since a second param-eter, that of peripheral capillary filtration surface enters into thecalculation of this measure.

Thus, clearcut changes in epithelial cell structure occur intype I diabetes only in patients with severe diabetic nephropa-thology characterized by marked mesangial expansion anddistortion of glomerular capillary architecture, including de-creased glomerular peripheral capillary filtration surface [11].We have previously hypothesized that this decrease in surfacecould signal adaptive renal responses resulting in hemodynamicand, possibly, other glomerular alterations which could thenlead to further glomerular injury. It is uncertain as to whetherour measure of filtration slit length per glomerulus or capillaryfiltration surface is a better measure of the true surface acrosswhich most of the water flux occurs. Even so, we suppose thatthis reduction(s) could be responsible for the epithelial cell

•, changes and the proteinuria seen in advanced diabetic nephrop-0.7 athy. The exact relationships between proteinuria and epithelial

cell structure, however, remain to be elucidated. Nonetheless,our work here would argue against glycosylation of glomerularconstituents [27], changes in proteoglycan metabolism [28—30],or other parallel processes as the sole and direct cause of majorpermselectivity defects and epithelial changes in diabetes.Rather, we argue this is more likely to be consequent to themechanical influence of mesangial expansion on filtration sur-face and the renal adaptations to these perturbations, althoughany mechanism alluded to above could play a role in themesangial expansion. This argument is consistent with theobservation that the fractional clearances of IgG and albumin aswell as the dextran sieving curves remain normal in diabeticpatients until GFR begins to fall [31]. Then, with progressiveloss of GFR [31] (and thus, filtration surface [11]) there isprogressive disruption of the permselectivity bather.

AcknowledgmentsThis work was supported in part by National Institutes of Health

grants AM07087 and AMO1OI4 and the Viking Children's Fund. Theauthors thank Thomas Groppoli and Frank Strgar for expert technicalassistance, Rai Smith for preparation of the manuscript, and MarshallHoff for production of the figures.

Reprint requests to Eileen N. Ellis, M.D., Arkansas Children'sHospital, 800 Marshall Street, Little Rock, Arkansas 72202, USA.

References

1. KNOWLES HC, GUEST GM, LAMPE J, KESSLER M, SKILLMAN TG:The course of juvenile diabetes treated with unmeasured diet.Diabetes 14:239—273, 1965

2. MARKS HH: Longevity and mortality of diabetes. Am J PublicHealth 55:416—423, 1965

3. MOROENSEN CE: Renal functional changes in diabetes. Diabetes 25(Suppl. 2):872—879, 1976

4. PARVING HH, OXENBALL B, SvEND5EN PA, C1-rnt5TIAN5EN IS,ANDERSON AR: Early detection of patients at risk of developing

100,000

10,000

1000

100

10

0 0:1 0:2 0:3 0:4 0:5 0:6

I-C

Filtration slit length density, sm4am3

Fig. 4. The relationship between filtration slit length density and UAE inthe diabetic subjects (r = —0.80, P < 0.001).

relates with fractional IgG and albumin clearances [24]. Instreptozotocin—induced diabetes in rats, an imprecise relation-ship between urinary albumin excretion rate and foot processwidth was noted [26] although the magnitude of the albuminuriawas relatively trivial. We demonstrate here in patients withdiabetes mellitus there is also an imprecise relationship betweenfoot process width and urinary albumin excretion, while albu-minutia is more precisely related to filtration slit length/glome-rulus and especially filtration slit length density. Patients withtype I diabetes whose UAE is normal or in the microalbuminu-na range have, at most, only marginally different foot processwidth—measurements compared to normal individuals, therebeing much overlap. Clearly, there are no differences in thismeasure comparing diabetic patients with and without microal-buminuria. The group of diabetic patients with overt proteinuriado have significantly increased measures of foot process widthbut, again, there is substantial overlap with normals. It may bethat different approaches to foot process width measurementcould reveal clearer, although most likely subtle, differencesbetween normals and nonproteinuric diabetic people.

The length density of the filtration slit is an indirect estimateof epithelial cell structural integrity which avoids the problemsof width measurement discussed above. Our concept is that adecrease in this length density represents a simplification ofepithelial cell structure which can be closely estimated indepen-dent of the complexities of epithelial cell shape. No cleardifferences in this estimate of epithelial structure were seenbetween normals and diabetic patients without nephropathy or

Glomerular epithelial cell in diabetes 741

diabetic nephropathy. A longitudinal study of urinary albuminexcretion. Acta Endocrinol 100:550—555, 1982

5. VIBERTI GC, JARRETT RJ, MOHMUD U, HILL RD, ARGYROPOULOSA, KEEN H: Microalbuminuria as a preditor of clinical nephropathyin insulin—dependent diabetes mellitus. Lancet 1:1430—1432, 1982

6. MOGENSEN CE: Microalbuminuria predicts clinical proteinuria andearly mortality in maturity onset diabetes. N Engl J Med 310:356—360, 1984

7. MOGENSEN CE, CHRISTENSEN CK: Predicting diabetic nephrop-athy in insulin-dependent patients. N Engi J Med 311:89—93, 1984

8. BOHMAN SO, JAREMKO G, BOHLIN AB, BERG U: Foot processfusion and glomerular filtration rate in minimal change nephroticsyndrome. Kidney mt 25:696—700, 1984

9. NEvINs TE, GASTON T, BASGEN JM: Quantitative indexes ofaminonucleoside—induced nephrotic syndrome. Am J Pathol 117:30—36, 1984

10. MAUER SM, STEFFES MW, ELLIS EN, SUTHERLAND DER,BROWN DM, GOETZ FC: Structural—functional relationships indiabetic nephropathy. J C/in Invest 74:1143—1155, 1984

11. ELLIS EN, STEFFES MW, GOETZ FC, SUTHERLAND DER, MAUERSM: Glomerular filtration surface in type I diabetes mellitus.Kidney mt 29:889—894, 1986

12. CHAVERS BM, SIMONSON J, MICHAEL AF: A solid phase fluores-cent immunoassay for the measurement of human urinary albumin.Kidney mt 25:576—578, 1984

13. ELLIS EN, MAUER SM, SUTHERLAND DER, STEFFES MW: Gb-merular capillary morphology in normal man (submitted for publi-cation)

14. HIROSE K, OSTERBY R, NOZAMA M, GUNDERSON HJG: Develop-ment of glomerular lesions in experimental long—term diabetes inthe rat. Kidney In! 21:689—695, 1982

15. JANSEN EB, GUNDERSEN HJG, OSTERBY R: Determination ofmembrane thickness distribution from orthogonal intercepts. JMicroscopy 115:19—33, 1979

16. GUNDERSEN HJG, SEEFELDT T, OSTERBY R: Glomerular epithelialfoot processes in normal man and rats. Cell Tissue Res 205:147—155, 1980

17. ELLIS EN, BASGEN JM, MAUER SM, STEFFES MW: Kidney biopsytechnique and evaluation in diabetes mellitus, in Methods ofDiabetes Research (vol. II) edited by J LARNER, WL CLARKE, SLPOHL, New York, John Wiley and Sons, 1986, pp 633—647

18. BROCKNER—MORTENSEN J, GEISE J, R055ING N: Renal inulin

clearance versus total plasma clearance of 51Cr-EDTA. Scand JC/in Lab Invest 23:301—305, 1969

19. PARVING HH, ANDERSEN AR, SMIDT UM: Monitoring kidneyfunction in insulin dependent diabetes with diabetic nephropathy.Diabetic Nephropathy 3:130—134, 1984

20. SEEFELDT T, BOHMAN SO, GUNDERSEN HJG, MAUNSBACH AB,PETERSEN VP, OLSEN 5: Quantitative relationship between glomer-ular foot process width and proteinuria in glomerulonephritis. LabInvest 44:541—546, 1981

21. TISHERCC: Anatomy of the kidney, in The Kidney (2nd ed.), editedby BM BRENNER, FC RECTOR, Philadelphia, WB Sanders, 1981, pp3—75

22. KANWAR YS: Biophysiology of glomerular filtration and protein-uria. Lab Invest 51:7—21, 1984

23. POWELL HR: Relationship between proteinuria and epithelial cellchanges in minimal lesion glomerulopathy. Nephron 16:310-317,1976

24. SHEMESH 0, Ross JC, DEEN WM, GRANT GW, MYERS BD: Natureof the glomerular capillary injury in human membranous glomeru-lopathy. JCI 77:868—877, 1986

25. KOBAYASHI 5, NAGASE M, HONDA N, HISHIDA A: Glomerularalterations in uranyl acetate induced acute renal failure in rabbits.Kidney Int 26:808—815, 1984

26. STEFFES MW, LEFFERT JD, BASGEN JM, BROWN DM, MAUERSM: Epithelial cell foot process width in intact and uninephrectom-ized diabetic and nondiabetic rats. Lab Invest 43:225—230, 1980

27. COHEN MP, URDANIVIA E, SURMA ML, CIBOROWSKI CJ: Nonen-zymatic glycosylation of basement membranes. Diabetes 30:367—371, 1981

28. ROHRBACH DH, HASSELL JR, KLEINMAN HK, MARTIN GR: Alter-ations in the basement membrane (heparin sulfate) proteoglycan indiabetic mice. Diabetes 31:185—188, 1982

29. BROWN DM, KLEIN DJ, MICHAEL AF, OEGEMA TR: 355-glyco-saminoglycan and 35S-glycopeptide metabolism by diabetic glomer-uli and aorta. Diabetes 31:418—424, 1982

30. COHEN MP, SURMA ML: Effect of diabetes on in vivo metabolismof [35S]-labelled glomerular basement membrane. Diabetes 33:8—12,1984

31. MYERS BD, WINETZ JA, CI-IUI F, MICHAELS AS: Mechanisms ofproteinuria in diabetic nephropathy: A study of glomerular barrierfunction. Kidney In! 21:633—641, 1982