Islet neoformation in tissue culture

11

Molecular und Cellulur Endocrinology, 48 (1986) 11-20 Elsevier Scientific Publishers Ireland, Ltd.

MCE 01538

Islet neoformation in tissue culture Heinz Popiela ‘, Tatsuo Tomita 2, Orion Hegre 3 and Wayne Moore 1 ’ Department of Pediatrics, ’ Department of Pathology, Umoersity of Kunsus Medical Center, Kansus City, KS und ’ Depurtment of Anatomy, University of Minnesota,Minneapolis, MN (U.S.A.) (Received

Key words: islet neoformation;

tissue culture;

30 April 1986; accepted

18 June 1986)

rat pancreas.

Summary We have devised a tissue culture system that permits de novo formation of islets. Neonatal rat pancreata are enzymatically dissociated into single cells. The cell suspension is filtered through polyester cloth with 20 pm pores to exclude cell aggregates as well as preformed islets and a single cell suspension is then plated into tissue culture dishes at a density precluding reaggregation. Pancreatic cells proliferate forming numerous colonies of epithelioid cells. After a confluent monolayer, cells proliferate into a third dimension, the space occupied by the culture medium. Third dimensional proliferation occurs from basal monolayers of epithelioid colonies. At about 9 days in culture, numerous hillocks are visible that are spaced at about 1 mm from one another. Islets are observed to bud from the hillock surfaces. In 1 pm-thick sections, secretion granules are detected with the light microscope in some islet cells. With the electron microscope three basic cell types are seen. One peripherally located cell type is sparsely granulated and appears to be a precursor cell. The other peripherally located cell type shows a homogenous population of secretion granules characteristic of A-cells. The third cell type is found in the interior of islets containing granules characteristic of B-cells. Islet cells, but not hillock cells, react immunocytochemitally for insulin and glucagon. The cultures secrete 2 to lo-fold the amounts of glucagon present in fresh medium. It is concluded that differentiation of A- and B-cells occurs in neoformed islets.

Introduction A preparative, nonenzymatic method for the collection of pure islets of Langerhans has been described (Hegre et al., 1979; Hellerstrom et al., 1979). After culture of a slurry of neonatal pancreatic tissue fragments for about 1 week, islets became visible allowing collection of pure islets. These islets were either already present in the

Send correspondence Kansas Medical Center, City, KS 66103, U.S.A. 0303-7207/86/$03.50

to: Heinz Popiela, University of MRRC Bldg. 37, Room 124, Kansas

0 1986 Elsevier Scientific

Publishers

Ireland,

tissue but were subsequently released during culture due to degeneration of encasing acinar tissue and liberation of surviving islets, or they may have formed de novo. Both processes, release of preexisting islets and neoformation may have occurred concomitantly. After culture of enzymatic digests of neonatal (Orci et al., 1973; Chick et al., 1976; Meda et al., 1982; Yoshida et al., 1984; Romanus et al., 1985) or adult (Hilwig and Schuster, 1971) pancreata, several cell types were identified. Initially, exocrine, endocrine, and fibroblastoid cells were visible but soon the exocrine cells died or dedifferentiated leaving only endocrine and fibroblastoid Ltd.

12

13

cells (Orci et al., 1973). When pancreatic digests were cultured in a collagen gel matrix they were reported to reorganize into a third dimension leading to the formation of islet-like aggregates (Montesano et al., 1983). Others (Dudek et al., 1980; Mourmeaux et al., 1985) reported formation of aggregates in a standard culture medium in the absence of a gel matrix. In these prior studies, tissue fragments and preexisting islets were not removed prior to plating of tissue digests, thus, doubts concerning de novo formation of islets persist. In order not to confound surviving islets with neoformed islets we eliminated tissue fragments and preexisting islets prior to plating and then investigated the capacity of single pancreatic cells to form islets de novo. We report proliferation of single cells into monolayers followed by growth into a third dimension leading to the formation of hillocks from which islets bud de novo. Materials and methods Cell culture Fischer F-344 rats (Harlan-Sprague-Dawley, Indianapolis, IN) were bred at the University Animal Care Facility. Neonatal, l-day-old, animals were killed by cervical dislocation and pancreata were then dissected and collected in calcium and magnesium-free Puck’s Saline G (Puck, 1958). After removal of most of the nonendocrine tissue under the dissecting microscope, lo-12 pancreata were transferred to a dry tissue culture dish and minced for 5 min with small curved scissors. The tissue mince was then transferred to 0.25% trypsin (bovine type III, Sigma, St. Louis, MO) in Saline G and incubated for 20 min at 37°C. Following incubation, a 4-fold excess of Saline G was added and the tissue slurry was centrifuged for 5 min at 200 x g. The supernatant was aspirated, the pellet resuspended in 10 ml of Saline G and then tri-

turated for 5 min to dissociate the tissue. The tissue dissociate was filtered through three layers of cheese cloth to remove tissue fragments, and then centrifuged again for 5 min at 200 x g. The second pellet was resuspended in 3 ml of Saline G, triturated to suspend the cells, and then filtered through Swiss polyester cloth (20 pm pores, TetCo, Elmsford, NY). The final cell suspension contained mostly single cells, which were counted with a hemacytometer. 5 x lo6 or 1 x lo6 cells were plated into 35 mm tissue culture dishes (Corning, Corning, NY) in a volume of 2 ml of medium. The culture medium was modified from a method described previously (Popiela and Ellis, 1981; Popiela et al., 1984). It consisted of Ham’s F-12 (Ham, 1965) supplemented with 15% fetal calf serum (KC Biologicals, Lenexa, KS), 1 pg/ml porcine insulin (Lilly Research Laboratories, Indianapolis, IN), 10 pg/ml human transferrin (Research Plus Laboratories, Bajonne, NJ), 30 nM selenium (Alfa, Danvers, MA), and an antibioticantimycotic mixture (Gibco Laboratories, Chagrin Falls, OH) consisting of 1 mU/mI penicillin, 1 ng/ml streptomycin, and 0.25 pg/ml fungizone. Since high concentrations of glucose have been reported to stimulate islet cell growth and insulin secretion (King and Chick, 1976; Logothetopoulos et al., 1983; Milner et al., 1984; Swenne, 1985) the glucose concentration of the medium was increased to 30 mM. Cells were incubated at 37°C in 5% CO, and air. The medium was renewed every second day. Microscopy Cultures were washed twice with Dulbecco’s phosphate-buffered saline (Gibco, Grand Island, NY) and then processed for light and electron microscopy as previously described (Popiela, 1976). Washed cultures were fixed for 1 h with 3% glutaraldehyde (Sigma, St. Louis, MO) in 0.1 M cacodylate (Polysciences, Warrington, PA) buffer,

Fig. 1. Development of islets from single cells in culture. Following enzymatic dissociation of neonatal rat pancreata, single cells were plated at densities of 5 x 10’ cells/35 mm tissue culture dish. One day after plating, dispersed single cells are still visible except for a few ‘epithelioid’ cells that appear to have divided (Id, arrow). After 3 days in culture, proliferation of cells is readily visible and colonies of ‘epithelioid’ cells are seen (3d, arrows). Seven days after plating, pancreatic cells have proliferated to a confluent monolayer and numerous colonies of ‘epithelioid’ cells are visible (7d, arrows). From about 9 days in culture, hillocks are seen and islets are budding out. At 14 days, mature hillocks and islets are visible (14d). Micrograph I4d show a mature hillock (H) with a budding islet (open arrows). Numerous large vacuolated cells emanate from the hillock (diamonds). Photomicrographs of live cells seen under phase contrast. Calibration bars indicate 100 pm.

Fig. 2. Secretion granules in islet cells but few in hillock ceils, A: Hillock and islet appearing after 10 days in culture are shown in cross-section. B shows the islet visible in A at a greater magnification to reveal secretion granules (arrows). Photomicrograph of a preparation fixed, transversely sectioned at 1 pm, and stained with Azure 3. The culture dish surface is visible at the hottoti nf both micrographs adjacent to the lower surface of the hillock. The curvature of the dish surface is an artifact of infiltration with plastic. In A, part of a cell monolayer leading to the hillock is visible. Calibration bars indicate 10 pm.

15

pH 7.2. After a 10 min wash with 0.1 M cacodylate buffer, the cultures were postfixed for another hour with 2% osmium tetroxide (Polysciences, Warrington, PA) in 0.1 M cacodylate buffer. The cultures were then dehydrated in a graded series of ethanol and stained en bloc for 10 min with 2% uranyl acetate (Polysciences, Warrington, PA) in 100% ethanol. The cultures were then infiltrated with Spurr’s plastic (Polysciences, Warrington, PA), followed by polymerization overnight at 60 o C. For light microscopy, sections were cut at 1 pm with an LKB ultramicrotome (LKB Instruments, Rockville, MD), mounted on glass slides, and stained with Azure B (Polysciences, Warrington, PA). For electron microscopy, 500-700 A-thin sections were cut with the same ultramicrotome, mounted on copper grids, and stained with 0.25% lead citrate (Polysciences, Warrington, PA). The sections were viewed with a Philips electron microscope. Histology and immunocytochemistty Cultures were fixed for 2 h in Bouin’s fixative followed by dehydration in a graded series of ethanol. They were infiltrated for 4 min with Histo-Clear (National Diagnostics, Somerville, NJ) and then with paraffin. After solidification of the paraffin, the cultures were separated from the dishes and blocks of embedded cultures were sectioned at 5 pm. Deparaffinized sections were immunocytochemically reacted for insulin, glucagon, and somatostatin. Immunocytochemistry was performed according to the unlabelled peroxidase-antiperoxidase method of Sternberger (1979) using a reagent kit (Dako Pap Kit, Dako Corp., Santa Barbara, CA). After photography of the immunocytochemically reacted sections, they were counterstained with Gill’s hematoxylin. Radioimmunoassays Insulin was iodinated with chloramine T following the procedure of Gavin et al. (1972). A stock solution of 1 mg/ml porcine insulin (courtesy of Dr. Mary Root, Lilly, Indianapolis, IN) in 0.01 M HCl was prepared. 5 ~1 of the insulin solution was then mixed with 10 ~1 of 1 M phosphate buffer, pH 7.4. 1251(Amersham, Arlington Heights, IL) was diluted with phosphate buffer, pH 7.4 to 100 pCi/pl. 1 ml of distilled water was added to

0.3 mg of chloramine T (Eastman Kodak, Rochester, NY) and 0.6 mg of sodium metabisulfite (Fisher, Fair Lawn, NJ). For iodination, 20 ~1 of 1251-solution was reacted with 10 ~1 of the chloramine T solution for 30 s. 10 ~1 of sodium metabisulfite solution was then added followed by 200 ~1 of 5% bovine serum albumin (Sigma, St. Louis, MO) in water. The iodination mixture was then loaded onto a 15 X 670 mm Sephadex G-100 column and eluted with 0.03 M phosphate buffer containing 1% bovine serum albumin. Typically, insulin was iodinated at 75-85% with a specific activity of 250-265 pCci/pg. In preparation for measurement of insulin content, cells were washed 3 times with phosphatebuffered saline (Gibco, Grand Island, NY) and extracted with acid ethanol. To measure insulin secretion, supernatant media were not extracted. For the assay, 100 ~1 of sample or rat insulin standard (courtesy Lilly, Indianapolis, IN) were mixed with 100 ~1 of guinea pig antiserum to porcine insulin (Biotek, Lenexa, KS; 1: 20 dilution) and 150 ~1 of assay buffer. Assay components were diluted with assay buffer, a phosphate buffer, pH 7.6 containing 0.025 M EDTA and 1% bovine serum albumin (Sigma, St. Louis, MO). Following incubation at 4°C for 24 h, 100 ~1 of diluted ‘251-insulin (Amersham, Arlington Heights, IL; 1.5 nCi) were added and incubation continued for another 16 h at 4OC. A subsequent 2 h incubation period followed after addition of 100 ~1 guinea pig normal serum (1 : 80 dilution) and a final 2 h incubation period at 4°C after addition of goat anti-guinea pig gamma globulin (1: 20 dilution). After centrifugation for 30 min at 4°C followed by aspiration of the supernatant, the pellet was counted in a gamma counter. For glucagon radioimmunoassay the procedure previously published was followed (Tomita, 1980). Briefly, aprotinin and diluted rabbit anti-pork glucagon were mixed with standard beef-pork glucagon (Lilly, Indianapolis, IN) or sample and incubated at 4°C for 24 h. Normal diluted rabbit serum and ‘251-glucagon were then added and incubation continued for another 72 h. Bound hormone was separated from free hormone by incubation with diluted goat anti-rabbit serum followed by sedimentation of the immune complex. Sedimented pellets were counted with an

6

17

automatic gamma counter. pared in 0.05 M phosphate ing bovine albumin.

All solutions were prebuffer, pH 7.4 contain-

Results After plating a suspension of single pancreatic cells, they attach to the tissue culture dishes and begin division as early as 1 day of culture (Fig. 1: Id, arrow). Cell aggregates or islets are absent. Cellular reaggregation of adherent cells is not observed and nonadherent cells that possibly could reaggregate are removed when the culture medium is changed every second day. Proliferation of adherent cells continues until small colonies of epithelioid cells are visible at 3 days (Fig. 1: 3d, arrows). Besides epithelioid cell colonies, fibroblastoid cells are visible. These cells, however, do not form colonies. By 7 days in culture, the cells have formed a confluent monolayer and numerous colonies of epithelioid cells are seen (Fig. 1: 7d, arrows). Following the formation of a confluent monolayer, cell growth continues from the epithelioid colonies into a third dimension, into the space occupied by the medium. At about 9 days, hillocks become visible and numerous hillocks begin to show budding islets. Hillocks and islet buds continue to grow until mature islets are visible at about 14 days in culture (Fig. 1: 14d). When islets reach about 30 pm in diameter they apparently are severed from the hillocks and float into the medium. Since the medium is renewed every second day, islets are removed from the cultures at that time. Centrifugation of the aspirated medium reveals islets in the pellet. Islet production continues until about 6 weeks and then ceases. Hillocks continue to survive and they grow by lobular extension as long as cultures are maintained. Eventually, the cultures are accidentally contaminated with microorganisms and are then discarded. However, we have maintained hillock cul-

tures for as long as 3 months. A hillock and islet bud formed after 10 days in culture is shown in a high-resolution micrograph (Fig. 2). Secretion granules are seen in 1 pm-thick sections in some cells of the islet (arrow) but none or very few in hillock cells (Fig. 2B). To investigate the type of granule, thin sections are viewed with an electron microscope. Three basic islet cell types are seen with the electron microscope. One cell type shows few secretion granules (Fig. 3: Ia, arrows) but is rich in rough endoplasmic reticulum and Golgi complexes. Another cell type shows homogeneous populations of secretion granules (Fig. 3: Ib, solid arrows) and is found mostly in the periphery of islets. The granules are round and have very little if any space between the granular membrane and its contents. The third cell type is predominantly located in the islet interior and shows a more variable granular morphology (Fig. 3: Ic, arrows). Besides granules with little or no space between the membrane and its contents, most granules show a variable space between the membrane and its core. Some granules show small, dense cores characteristic of beta-granules (Fig. 3: Ic). Glycogen particles are distributed throughout the cytoplasm in all islet cell types. Hillock cells, on the other hand, rarely show granules (Fig. 3: H). Lobulated nuclei with large nucleoli and an extensive network of rough endoplasmic reticulum are characteristically seen. The insulin content measured by RIA in mature hillock cultures is low. The supernatant medium contains less insulin than initially present in fresh medium. In fact, 73.5% + 7 of the insulin added to the hillock culture medium disappears after 2 days. In control muscle cultures, 90% of the insulin disappears. It is conceivable that the difference between the two cultures represents cellular secretion of insulin into the hillock culture medium. Presence of B-cells in neoformed islets is demonstrated immuno-

Fig. 3. Ultrastructure of secretion granules in islet cells. A hillock cell is shown in FI: a lobulated nucleus (N) is visible and a large nucleolus (Nu). Numerous cisternae of rough endoplasmic reticulum, and glycogen particles are visible. Zcr shows a sparsely granulated cell frequently seen at the periphery of islets. A nonlobulated nucleus (N) is visible with a smaller nucleolus (Nu). Rough endoplasmic reticulum (E) as well as Golgi complexes (G) are seen. Two secretion granules (arrows) are visible in the cell that occupies most of the micrograph la. I/I shows a cluster of three, moderately granulated (arrows) islet cells. A portion of a nucleus of a fourth cell is visible in the lower left corner. The open arrow indicates a cell junction. Ic shows the cytoplasmic portion of a heavily granulated cell found in the islet interior. Arrows indicate morphologically dissimilar secretion granules. Calibration bars indicate 1 pm.

18

Fig. 4. Immunocytochemical reaction for insulin in islet cells. Pancreatic cells were cultured for 14 days, then fixed and infiltrated with paraffin, Cultures were sectioned at 5 pm perpendicular to the culture dish surface. Deparaffinjzed hillock sections showing an islet bud were reacted with antibody to insulin (I) according to the uniabeled antibody method of Sternberger (1979) using a kit (Dako, Santa Barbara, CA). After p~otograpby, the same sections were counterstained with Gill’s hematox~lin (H). islet bud celis show a reddish-brown reaction product only in the cytoplasm (not visible in black and white micrograph I); bud nuclei remain unreacted. Hillock cells show virtually no reaction product. Calibration bar indicates 10 @m.

19

types consistent with the observation production ceases after 6 weeks.

cytochemically (Fig. 4), Fig. 4(H) shows a hillock section stained with hematoxylin-eosi~ and Fig. 4(I) the same section immunoche~caIly reacted with antibody to insulin. ImmunochemicaIly reacted hillock sections show a reddish-brown reaction product only in the cytoplasm of islet bud cells (color not visible in black and white micrograph Fig. 4(I)); nuclei remain unreacted. Hillock cells show virtually no reaction product. Reaction with antibody to glucagon indicates distribution of A-cells mainly in the periphery of islets; however, some reacting cells are also located in the islet interior. Hillock cells, likewise, do not react with antibody to glucagon. Afthough secreted insulin is not measurable by RIA presumably due to the exogenous insulin in the medium, secreted glucagon (Table 1) is assayed. Whereas control muscle cultures secrete little or no glucagon, hillock cultures secrete 2 to 13-fold the amounts initially present in fresh medium. Surprisingly, medium supplemented with insulin, chicken ovotransferrin, and 10% fetal calf serum cause secretion of larger amounts than medium supplemented with human transferrin, human growth hormone, and other serum combinations. A decline in secretory activity in older cultures is observed in the presence of two medium

TABLE

that

islet

Discussion After removal of tissue fragments and preexisting islets by filtration, single cells are plated. Since the culture medium is changed every second day, nonadherent cells are removed; hence, formation of islets by cellular reaggregation does not occur. Adherent cells do not reaggregate. Single, adherent cells proliferate, forming a monolayer followed by formation of hillocks and finally islet buds. Islet buds growing from hillocks in this study are reminiscent of islet development in vivo (Pictet and Rutter, 1972). In fact, the similarities of in vitro-generated islet buds in this study to the buds shown in vivo by Pictet and Rutter (1972) are remarkable. In our in vitro system, morphology of secretion granules as well as immunocytochemitally localized insulin and glucagon conjointly with the considerable amount of glucagon released into the culture medium indicates B- and A-cell differentiation. Whereas most previous reports on cultured pancreatic digests describe survival of endocrine cells in ‘monolayer’ cultures (Orci et al., 1973; Chick et al., 1976; Meda et al., 1982:

1

GLLJCAGON

SECRETION

BY ISLET CULTURES

Pancreatic cells were plated in the centrifuged to sediment cells and CM/FM: Ratios of CM over fresh Medium containing 10% fetal calf serum, insulin and human transfer& growth hormone. Cell type

Culture age (days)

Muscle CM/FM

38

Islet CM/FM

38

Islet CM/FM

59

Fresh medium -

fresh media indicated. After exposure for 3 days, the supernatant medium (CM) was aspirated, debris, and then assayed for glucagon. The data are expressed as pg glucagon per ml of CM. medium (FIM). P(n): Medium containing 15m Oi horse serum, no insulin, but chicken transferrin. P: serum, insulin and chicken transferrin. H: Medium containing 10% fetal calf serum, 5% horse GH: Medium containing 10% fetal calf serum, 5% horse serum, insulin, human transferrin, and Type of medium P(n)

P

II

GH

60.93 1.13 147.05 12.98 123.48k1.18 I.58 77.97

156.56 * 24.27 2.90

133.32 2.74

74.06

95.21

6.27

I .76

141.75 2.91

11.33

53.95

48.72

20

Yoshida et al., 1984; Romanus et al., 1985) our data indicate growth of cells into a third dimension, formation of hillocks, and neoformation of islet buds. The precursor cell for budding islets may be an undifferentiated cell (Fig. 3: H) with a similar morphology as the ‘protodifferentiated’ duct cell described by Pictet and Rutter (1972). Detection of mitotic figures in the superficial hillock cell layer in the vicinity of buds suggests proliferation of undifferentiated hillock cells followed by preferential adhesion to forming islets and subsequent differentiation into A- and B-cells. Differentiated endocrine cells are not seen external to islet buds and mitotic figures are not detected in the buds, thus, proliferation of differentiated endocrine cells is unlikely the mechanism of bud formation. Sparsely granulated cells as well as presumptive A-cells are found in the bud periphery and well-granulated presumptive B-cells in the interior suggesting that younger, relatively undifferentiated cells as well as A-cells are located in the periphery and more mature B-cells in. the interior. In addition, many of the cells located in the interior of islet buds react immunocytochemitally for insulin allowing for B-cell identification. Glucagon-reactive cells are located mainly in the bud periphery. Thus, the neoformed islets in this study generally conform to the islet configuration in vivo; the A-cells are located in the periphery and the B-cells in the interior. The most striking difference between in vivo pancreatic development and that in vitro is the absence of acinar cells in lo-day-old or older cultures. Even with the electron microscope, cells with zymogen granules are never detected. Orci et al. (1973) also described poor survival of exocrine cells in culture noting that cells containing zymogen granules are rarely detected after 60 h. Quite likely, acinar cells are present in our initial cell suspension as they were shown to be present by Orci et al. (1973) but subsequently die, do not adhere, or dedifferentiate several days after plating. In short, we have demonstrated neoformation of rat islets from single neonatal pancreatic cells and plan to generate human islets from abortuses for transplantation to diabetic recipients.

References Chick, W.L., King, D.L. and Lauris, V. (1976) In: Diabetes Research Today, Ed.: E. Lindenlaub (F.K. Schattauer Verlag, New York) pp. 11-23. Dudek, R.W., Freinkel, N., Lewis, NJ., Hellerstrom, C. and Johnson, R.C. (1980) Diabetes 29, 15-21. Gavin, J.R., Roth, .I.. Jen, P. and Freychet, P. (1972) Proc. Natl. Acad. Sci. U.S.A. 69, 747-751. Ham, R.G. (1965) Proc. Natl. Acad. Sci. U.S.A. 53, 288-293. Hegre, O.D.. Marshall, S., Schulte, B.A., Hickey, G.E., Williams, F., Sorenson, R.L. and Serie, J.R. (1983) In Vitro 19, 611-620. Hellerstrom, C., Lewis, N.J., Borg, H., Johnson, R. and Freinkel, N. (1979) Diabetes 28, 769-776. Hilwig, I. and Schuster, S. (1971) In: Handbuch der experimentellen Pharmakologie, Vol. 32/l, Eds.: 0. Eichler, A. Farah, H. Herken and A.D. Welch (Springer Verlag, New York) pp. 109-121. King, D.L. and Chick, W.L. (1976) Endocrinology 99, 1003-1009. Logothetopoulos, J., Valiquette, N. and Cvet, D. (1983) Diabetes 32. 1172-1176. Meda, P., Kohen, E., Kohen, C., Rabinovitch, A. and Orci, L. (1982) J. Cell Biol. 92, 221-226. Milner, R.D.G., Cser, A. and Cope, G.H. (1984) J. Endocrinol. 100, 155-160. Montesano, R., Mouron, P., Amherdt, M. and Orci, L. (1983) J. Cell Biol. 97, 9355939. Mourmeaux, J.L., Remacle, C. and Henquin, J.C. (1985) Mol. Cell. Endocrinol. 39, 237-246. Orci, L., Like, A.A., Amherdt, M., Blondel, B., Kanazawa, Y., Marliss, E.B., Lambert, A.E. and Renold, A.E. (1973) J. Ultrastruct. Res. 43, 270-297. Pictet, R. and Rutter, W.J. (1972) In: Handbook of Physiology, Sect. 7, Vol. 1, Eds.: R.O. Greep and E.B. Astwood (American Physiological Society, Washington, DC) pp. 25-66. Popiela, H. (1976) J. Exp. Zool. 198, 57-64. Popiela, H. and Ellis, S. (1981) Dev. Biol. 83, 266-277. Popiela, H., Taylor, D., Ellis, S., Beach, R. and Festoff, B. (1984) J. Cell Physiol. 119, 234-240. Puck, T.T., Cieciura, S.J. and Robinson, A.J. (1958) J. Exp. Med. 108, 945-956. Romanus, J.A., Rabinovitch, J.A. and Rechler, M.M. (1985) Diabetes 34, 696-702. Sternberger, L.A. (1986) In: Immunocytochemistry (John Wiley, New York). Swenne, J. (1985) Diabetes 34, 803-807. Tomita, T. (1980) Diabetologia 19, 154-157. Yoshida, K., Kagawa, S., Murokoso, K. and Matsuoka, A. (1984) In Vitro 20, 756-762.