, . 181: 235–242 (1997)

ULTRASTRUCTURAL AND CONFOCAL LASERSCANNING MICROSCOPIC EXAMINATION OF

TUNEL-POSITIVE CELLS

*†, †, *, *, *, ‡, ‡, *, * †

*Third Department of Internal Medicine and †Department of Pathology, Tohoku University School of Medicine and‡Department of Pathology, Tohoku University Hospital, Sendai, Miyagi, Japan

SUMMARY

TdT-mediated dUTP-biotin nick end labelling (TUNEL) has been widely used for detecting cells with DNA fragmentation orapoptotic cells. However, since the concept of apoptosis is based on cellular ultrastructure, it is important to identify the morphologicalfeatures of TUNEL-positive cells. In this study, we performed TUNEL and electron microscopic observation on serial semithin andultrathin sections of pancreas from bilaterally adrenalectomized rats with caerulein-induced pancreatitis. TUNEL-positive cells wereidentified with two different ultrastructural patterns. One was characteristic of apoptosis, with condensed nuclei, intact mitochondria,and zymogen granules. The other pattern was one of marked cellular degeneration, possibly representing the end stage of cell death.Cells which did not demonstrate these ultrastructural patterns were not labelled by the TUNEL method. The three-dimensional structureof TUNEL-positive cells was also investigated by confocal laser scanning microscopy (CLSM), which showed the apoptotic nucleiexhibited various three-dimensional structures. These results confirm the utility of the TUNEL method in detecting apoptosis;application of the technique reported in this study will contribute to the further characterization of individual TUNEL-positive cells.

KEY WORDS—apoptosis; TUNEL; electron microscopy; serial sections; CLSM; pancreas

INTRODUCTION

Programmed cell death (PCD) is considered an activephysiological mechanism of cell death or a selectivephysiological process of cell deletion.1–6 PCD is alsoimportant in various pathological conditions, such asthe response to irradiation,7 stimulation by glucocorti-coid,8 withdrawal of glucocorticoid,9 and changes insome kinds of growth hormones.10,11 Morphologically,PCD has also been termed ‘apoptosis’. PCD or apop-tosis has been characterized ultrastructurally by shrink-age of cytoplasm, condensation and fragmentation ofthe nucleus, and membrane blebbing. It has beenbiochemically characterized by increased endogenousendonuclease activity, chromatin cleavage, and DNAfragmentation.8,12 Based on these characteristics, apop-totic cells have been identified mainly by biochemicalmethods and conventional electron microscopic obser-vations. Biochemical detection of DNA cleavage intooligonucleosomal fragments of multiples of 180–200base pairs, or the ladder pattern of nucleosomal DNAfragments on agarose gel electrophoresis, has beengenerally recognized as the hallmarks of apoptosis.8,12,13However, as expected, this method cannot identify theindividual cells that are undergoing apoptosis. On theother hand, conventional electron microscopy canrecognize apoptotic cells by the ultrastructural charac-teristics associated with PCD, but these features areobserved only for a limited period in the course ofapoptosis.12,14,15

In 1992, Gavrieli et al.16 demonstrated that DNAfragmentation can be detected in situ by labelling 3*-OHends with biotinylated deoxyuridine triphosphate(dUTP) through the action of terminal deoxynucleotidyltransferase (TdT). Since then, this method, subsequentlytermed 3*-OH nick end labelling, or TdT-mediateddUTP-biotin nick end labelling (TUNEL), has beenwidely used to detect cells with DNA fragmentation.The correlation of results obtained by the TUNELmethod and DNA ladder formation by agarose gelelectrophoresis has been demonstrated,16,17 but theultrastructural features of TUNEL-positive cells havenot been described in detail. This distinction is especiallyimportant when considering recent reports that theTUNEL method is not specific for the detection ofapoptosis;17 the technique may also detect cellswhich are committed to, but not yet in the process ofapoptosis.13,17In this study, we investigated the morphological

features of TUNEL-positive cells in a rat pancreatitismodel by the following two methods: (1) ultrastructuralexamination of TUNEL-positive cells in continuousmirror image semithin sections of Epon–Araldite-embedded tissue and (2) examination of FITC-labelledTUNEL and PI stained cells by confocal laser scan-ning microscopy (CLSM), which provides a three-dimensional image of TUNEL-positive nuclei.

MATERIALS AND METHODS

Animal treatmentInduction of severe oedematous pancreatitis by

caerulein—Male Wistar strain rats weighing

Addressee for correspondence: Kenji Kimura, MD, Third Depart-ment of Internal Medicine, Tohoku University School of Medicine,1-1 Seiryo-machi, Aoba-ku, Sendai, Miyagi 980-77, Japan.

CCC 0022–3417/97/020235–08 Received 15 December 1995? 1997 by John Wiley & Sons, Ltd. Accepted 23 July 1996

approximately 250 g underwent bilateral adrenalectomy(ADx) and received 5 ìg/kg per h caerulein intra-venously for 6 h. Details of the procedure have beenreported previously by our group.18

Tissue preparation—Rats were anaesthetized byintraperitoneal injection of pentobarbital (20 mg/kg)(Tokyo-Kasei, Co., Ltd., Tokyo, Japan) and underwentlaparotomy and thoracotomy. They were perfusedthrough the left ventricle with 4 per cent paraformal-dehyde (PFA) in 0·1 phosphate buffer (PB), pH 7·4(1000 ml/kg), to fix the pancreas. Preliminary exper-iments revealed that this method of fixation yielded themost satisfactory results in both TUNEL labelling andmorphological preservation of the DNase-treated speci-mens. The pancreas was removed quickly and trimmedfree of adherent fat and lymph nodes. Part of thepancreas close to the spleen was minced to 1 mmcubes and was post-fixed for electron microscopy. Theremaining portion of the pancreas was post-fixed forCLSM.

Post-fixation

Post-fixation for electron microscopy (EM)–TUNELserial sections—The pancreatic specimens were fixedfurther by immersing in ‘half’ Karnovsky solution (2 percent PFA, 2·5 per cent glutaraldehyde, 0·1 PB, pH 7·4)for 18 h at 4)C and rinsed in 15 per cent sucrose in0·01 PB. The specimens were subsequently immersedin 1 per cent osmium tetroxide (OsO4) in 0·1 PB for2 h at room temperature. Our preliminary experimentsdemonstrated that the post-fixation described above,especially immersion in 1 per cent OsO4, is crucial inthe preservation of subcellular structures, especiallymitochondria (Mt).

Post-fixation for confocal laser scanning microscopy(CLSM)—The pancreatic specimens were fixed byimmersing in 4 per cent PFA for 18 h at 4)C.

Embedding

Embedding for EM–TUNEL serial sections—Afterpost-fixation, the specimens described above werepolymerized in an Epon–Araldite mixture (Epon-812;17 ml; DDSA; 33 ml CY-212: 9 ml DMP-30: 1·2 ml)(TAAB, U.K.) for 24 h at 40)C then for 48 h at 60)C.

Embedding for CLSM—After post-fixation, thepancreatic specimens were embedded in regular paraffinwax.

Preparation of tissue sections

Sectioning for EM–TUNEL serial sections—Epon–Araldite-embedded tissues were trimmed, sectioned at athickness of 1 ìm, and stained with toluidine blue forlight microscopic examination. Ultrathin sections of80–85 nm thickness were cut on an Ultracut micro-tome (MT-5000, Dupont Company, Newtown, CT,U.S.A.) using a diamond knife (Diatome Ltd., Biel,

Switzerland). Serial adjacent semithin sections of 2 ìmthickness were subsequently cut using a sapphire knife(Sumitomo Kagaku, Co., Ltd., Tokyo, Japan) andmounted on a silane-coated slide glass (DAKO Japan,Tokyo, Japan). Ultrathin sections 80–85 nm thick weredouble-stained by uranyl acetate and lead citrate andexamined and photographed in an H-600 electronmicroscope (Hitachi Seisakujo, Co., Ltd., Tokyo,Japan). TUNEL was performed on the semithinsections.

Sectioning for CLSM—Paraffin-embedded tissue wassectioned at 10 ìm thickness and mounted on a silane-coated slide glass. TUNEL was performed on thesesections.

TUNEL

TUNEL was performed according to the method ofGavrieli et al.16 For the positive control, TUNEL wasperformed after the DNase treatment. On the otherhand, for the negative control, TUNEL was performedwithout the addition of TdT and/or dUTP.

TUNEL on serial tissue sections—The removal ofplastic carried out according to the method of Hoganand Smith,19 with some modification. The slides wereimmersed in a 0·5 per cent potassium hydroxide,benzene, and acetone mixture (1:1:1) for 18 h at roomtemperature and then neutralized in 1 per cent methanolacetate for 1 min. OsO4 was removed from the sectionsby immersing in 10 per cent periodic acid for 5 min.20The conditions for the removal of plastic and OsO4 fromthe tissue sections described above yielded the mostsatisfactory TUNEL labelling results in a preliminarystudy using DNase-treated tissue sections. The slideswere then incubated with 1·5 ìg/ml proteinase K (PK)(Boehringer Mannheim, Mannheim, Germany) for20 min at room temperature and washed three times indistilled water (DW) for 5 min. The slides were thenimmersed in TdT buffer (0·1 potassium cacodylate,2 m CoCl2, 0·2 m DTT, pH 7·2) (GIBCO BRL,Gaithersburg, MD, U.S.A.) and incubated in labellingmixture [5#TdT buffer, 10 ìl; TdT (GIBCO BRL),0·3 e.u./ìl; dUTP (Boehringer Mannheim), 0·04 nmol/ìl;DW, 37 ìl/slide] at 37)C for 120 min.21 The reaction wasterminated by immersing the slides in TB buffer (3 msodium citrate, 30 m sodium chloride) at room tem-perature for 30 min. The reacted slides were then rinsedwith DW, immersed in 0·01 PBS for 3 min, coveredwith 10% bovine serum albumin (BSA) for 30 min, andrinsed with PBS. The slides were covered with alkalinephosphatase (ALP) conjugated streptavidin (DAKO,Copenhagen, Denmark) at a concentration of 6 ìg/mlfor 20 min at room temperature and then rinsedthree times with 0·01 PBS for 5 min. After blockingendogenous ALP with 0·01 per cent levamisol (Sigma,St Louis, MO, U.S.A.), the slides were stained using anew fuchsin kit (DAKO, Copenhagen, Denmark) andwashed with PBS. The slides were counterstained withhaematoxylin, then dried and mounted.

236 K. KIMURA ET AL.

DNase treatment

The slides were processed with PK and treated withDN buffer (30 m Trizma base, pH 7·2, 140 mpotassium cacodylate, 4 m MgCl2, 0·1 m DTT).DNase I (2750 U/mg) (AMRESCO, OH, U.S.A.) dis-solved in DN buffer (final concentration 2 U/ìl) wasadded to the tissue sections. Following a 10 minincubation at room temperature, the slides were washedextensively with DW. These slides were then used forthe procedures described above.

TUNEL for CLSM—The tissue sections were de-paraffinized in xylene and immersed in 20 ìg/ml PK asdescribed above. The slides were then immersed in TdTbuffer and incubated with the labelling mixture for120 min at 37)C as described above. The slides werecovered with fluorescein-isothiocyanate (FITC)-conjugated avidin (DAKO) at a concentration of 20 ìg/ml for 20 min in the dark and stained with DNAfluorochrome propidium iodide (PI) (Sigma) at a con-centration of 50 ìg/ml in the dark for 10 min. The slides

were then mounted in mowiol (Calbiochem, La Jolla,CA, U.S.A.) prior to performing observations byCLSM.

CLSM observationAn LSM-GB 200 laser scanning microscope

(Olympus Co., Ltd., Tokyo, Japan) equipped with a25 mW multiline random polarization argon laser (488and 514·5 nm wavelengths) was used. The microscopewas fitted with 10#[numerical aperture (NA) 0·4],20#(NA 0·7), 40#(NA 0·95), 60#(NA 1·4), and100#(NA 1·35) objective lenses in connection witha BX-FLA-LSM fluorescent microscope (Olympus).Digitized fluorescent images excited by the argon laser(488 nm for FITC and 514·5 nm for PI) were acquiredat a scanning rate of 40 s per image and the scanningmode was XYZ. The X–Y images were acquired at 1 ìmintervals by Z-axial scanning. Digitized images werestored on the hard disk of the host IBM OS-2 computer(IBM Japan, Tokyo, Japan) and were photographed on35 mm film (Fuji Film Co., Ltd., Tokyo, Japan).22

Fig. 1—(a) A TUNEL-positive nucleus of an acinar cell in anEpon–Araldite semithin section. The arrow indicates an intenselylabelled acinar cell nucleus. The other nuclei are stained byhaematoxylin. (b) Ultrastructural features of the same acinus in aserial adjacent ultrathin section. The arrow indicates condensedchromatin with crescentic clumping. Bar=5 ìm. (c) At highermagnification, mitochondria (large arrow) and zymogen granules(small arrow) are observed to be almost intact. Rough ERrevealed a vesicular pattern (arrow-head). Bar=2 ìm

237ULTRASTRUCTURAL AND CLSM EXAMINATION OF TUNEL-POSITIVE CELLS

RESULTS

Ultrastructure of TUNEL-positive cellsSome nuclei of pancreatic acinar cells were intensely

labelled by the TUNEL procedure in semithin sectionsof Epon–Araldite blocks (Fig. 1a, arrow). The cyto-plasm of these cells was negative for TUNEL reactionproducts. Only condensed nuclei were labelled, theother nuclei remaining unlabelled in the tissue sectionsexamined.In the serial sections, we observed two ultrastructural

patterns in the TUNEL-positive cells. One is shown inFigs 1a, 1b, and 1c. TUNEL-positive acinar cells havenuclei with condensed chromatin, with crescentic clump-ing under the nuclear membrane. The condensed nucleus

was much smaller than the other intact nuclei (Fig. 1b).The cytoplasm was relatively compact and the cell wassmaller than other acinar cells, unlabelled by TUNEL.Ultrastructural examination of these TUNEL-positivecells at higher magnification revealed that mitochondriaand zymogen granules were almost intact, but roughendoplasmic reticulum (rER) was vesicular (Fig. 1c).The second ultrastructural pattern is illustrated in

Figs 2a and 2b. A TUNEL-positive cell is present insidea vacuole in the centre of the field, the other cells beingunlabelled. On a serial section, although this cell couldnot be definitely identified, it was considered a degener-ate acinar cell. The round electron-dense body containedcondensed rER and zymogen granules, but no nucleusand was not labelled by the TUNEL procedure. The cell

Fig. 2—(a) A TUNEL-positive nucleus in an intracellularvacuole of an acinar cell in an Epon–Araldite semithinsection. The large arrow indicates an intensely labelledacinar cell nucleus within the vacuole. The structuresindicated by the small arrow and arrow-head are notlabelled by TUNEL. (b) The ultrastructural features of thesame acinus in a serial adjacent ultrathin section at lowermagnification. The cell within the vacuole seems to be adegenerating acinar cell (large arrow). The structuresindicated by the small arrow and arrow-head are recog-nized as condensed cytoplasm. Bar=5 ìm. (c) Theultrastructure of the same acinus shown in a serial adja-cent ultrathin section at higher magnification. The struc-ture in the vacuole indicated by the large arrow is celldebris. The round structure indicated by the small arrowcontains condensed rER and zymogen granules, but nonucleus. Bar=3 ìm

238 K. KIMURA ET AL.

on the periphery of the acinus was not labelled by theTUNEL procedure. In DNase-treated slides, all thenuclei, including condensed apoptotic nuclei, wereintensely labelled. In contrast, negative control slides didnot have any TUNEL reactivity.

Confocal laser scanning of TUNEL-positive cells

Figures 3, 4 and 5 illustrate serial images at 2 ìmintervals, showing TUNEL-positive cells stained by

FITC and DNA staining by PI of the same field.TUNEL-positive nuclei labelled with FITC showedmarked condensation and fragmentation (Fig. 3a). Bothnuclei were stained intensely by PI (Fig. 3b). Whenobserved in the serial plane at 2 ìm intervals, both nucleiexhibited more intense labelling by FITC and PI; frag-mentation was more clearly observed in this section(Figs 4a and 4b). However, in the scan of the next serialplane at a 2 ìm interval from the plane in Fig. 4 (Figs 5aand 5b), one nucleus was not available for examination

Fig. 3—(a) A confocal image of TUNEL-positive cells labelled with FITC. The arrow and arrow-head showmarkedly condensed and fragmented nuclei. (b) PI staining of the same field. The nuclei which are positive forTUNEL are also stained intensely by PI

Fig. 4—(a) A confocal image of TUNEL-positive cells labelled with FITC, showing the same nuclei as inFig. 3a at a 2 ìm-interval. Both nuclei are stained more intensely by FITC and PI. (b) Nuclear fragmentationis more clearly observed

239ULTRASTRUCTURAL AND CLSM EXAMINATION OF TUNEL-POSITIVE CELLS

and the other had taken on a round shape, rather thanthe fragmentation pattern observed in other sections.

Effects of various histochemical procedures on patternsof TUNEL

The intensity of labelling by the TUNEL procedure inDNase-treated Epon–Araldite tissue sections can beinfluenced by various histochemical reactions, especiallythe following; (i) plastic removal; (ii) concentration ofPK; and (iii) osmium removal. Low concentrations ofpotassium hydroxide (0·1 per cent) did not completelyremove the plastic from tissue sections and labelling ofthe apoptotic cells was very faint. When plastic removalwas performed in 0·5 per cent potassium hydroxide for18 h and a PK concentration of 15 ìg/ml was used, thetissue was damaged and non-specific binding increased(data not shown). We observed intense staining after0·5 per cent potassium hydroxide for 18 h and a PKconcentration of 1·5 ìg/ml. The staining was markedlyweaker in the sections in which osmium had not beenremoved. After osmium removal by 10 per cent periodicacid, apoptotic cells were intensely stained.

DISCUSSION

The concept of ‘apoptosis’ which Kerr and Wyllieproposed was based on morphological and ultrastruc-tural examination.1–3 While TUNEL is an excellentmethod for examining DNA fragmentation in situ,especially in human tissues using surgical materials,23 itis still important to study the morphological features ofTUNEL-positive cells in greater detail.

This study has revealed that all ultrastructurallyapoptotic cells were positive by the TUNEL technique,but degenerate nuclei which may represent the end stageof cell death, either programmed or accidental, are alsostrongly labelled by the TUNEL method. Ansari et al.suggested that TUNEL may identify not only the apop-totic nuclei, but also the nuclei in necrotic tissues; thestaining of necrotic cells could be explained either by theactivation of an apoptotic pathway in damaged cells orby the degradation of DNA by lysozomal enzymes.17However, these two patterns can be distinguished by thedistribution of TUNEL-positive nuclei; apoptotic nucleiare usually scattered, without localized aggregation.17 Inthis study, since the TUNEL-positive nucleus illustratedin Fig. 2 is isolated and the adjacent nuclei exhibitnormal structure, it is presumed that this TUNEL-positive nucleus is in an end stage of the apoptoticprocess. Several investigators have proposed the possi-bility that cells which have not yet begun the process ofapoptosis and which lack its morphological featurescan also be detected by the TUNEL method,15,24 whichmay therefore be able to detect apoptotic cells muchearlier than electron microscopy. In the present study,however, nearly all TUNEL-positive cells exhibited thecharacteristic ultrastructural features of apoptosis.Very recently, Migheli et al.24 also applied the

TUNEL method at the electron microscopic level, usingan immunogold staining technique in embryonic mousedorsal root ganglia. They reported that nuclear localiz-ation of immunogold was observed not only in typicalapoptotic cells at various stages of cell death, but alsoin some apparently normal cells, which were structur-ally indistinguishable from adjacent unstained cells.Morphologically viable TUNEL-positive cells, including

Fig. 5—(a) A confocal image of TUNEL-positive cells labelled with FITC, showing the same nuclei asin Fig. 4a at the next 2 ìm-interval. The nucleus indicated by the arrow is not available for examination, whilethe nucleus indicated by the arrow-head shows a round shape

240 K. KIMURA ET AL.

the ‘pre-apoptotic cells’ described above,24 may repre-sent experimentally-induced artefactual DNA damage,leading to TUNEL positivity. The possibility that therelatively rapid process of apoptosis in pancreaticacinar cells in this model made it difficult to detect‘pre-apoptotic cells’ with TUNEL cannot be com-pletely ruled out, and awaits further investigation. Ouruse of serial sections has the following advantages overthe immunogold technique used by Migheli et al.:24much more ultrastructural detail is obtained than in asingle section; TUNEL-positive cells can be identifiedin a much wider area by light microscopy and aremuch more clearly visualized due to the colorimetricreaction; and the method is much less cumbersome anddifficult than immuno-electron microscopy. However,the method appears to be less sensitive than immuno-gold and further studies are required to improvethis.In addition, it is also important to note that labelling

is easily influenced by the techniques used to removeplastic and osmium from the sections. A lower concen-tration of potassium hydroxide did not fully remove theplastic and the labelling was very faint. Treatment with0·5 per cent potassium hydroxide for 18 h provided themost satisfactory results, indicating that completeremoval of the plastic is crucial for a successful TUNELassay. The detailed effects of plastic embedding on theTUNEL reaction are not known, but several groupsstudying the immunohistochemistry of growth hormoneand prolactin on plastic embedded tissue sectionsdemonstrated alteration of antigenic determinants byprotein bonding to resin monomers and shielding ofantigenic determinants from antibodies by the resinused for plastic embedding.19,25,26 Alcoholic sodiumhydroxide, which is known to break diester bonds inpolymerized resin mixtures, has been demonstrated tounmask antigenic determinants by removing attachedepoxy monomers and by eliminating the cross-linkedresin molecular lattice.27When OsO4 removal was not performed or was

incomplete, TUNEL positivity was markedly weak. Thegroups of Griegee,28 Wolman,29 and Munger30 haveproposed that OsO4 can react directly with nucleic acids,based on the observation that tissues reacted poorly withhaematoxylin following osmium treatment. In addition,Munger30 suggested that peracetic acid solubilizes thebound osmium and possibly liberates stainable phos-phate groups. Thus, we postulate that Epon–Aralditeplastic and osmium interrupt to some extent the bindingof dUTP to nick.CLSM yields better resolution and contrast than

conventional light microscopy and allows direct three-dimensional studies on tissue sections. The present studydemonstrated that the TUNEL-positive cells which didnot have the characteristic nuclear morphology ident-ified by PI staining of DNA in one plane, but which didexhibit nuclear features consistent with apoptosis indifferent optical levels, did not necessarily have thecharacteristic nuclear structure of apoptosis in all theplanes examined by CLSM. In addition, it is alsoimportant to note that apoptotic nuclei had a variablethree-dimensional structure.

In conclusion, the results of our present study indicatethat the ultrastructural features of apoptosis areunequivocally associated with DNA fragmentation andsupport the close association of DNA fragmentationdetected by TUNEL with ultrastructurally identifiedapoptosis. The methods used in this study provideimportant information on the interpretation ofindividual TUNEL-positive cells, possibly allowingdifferentiation between programmed and accidentalcell death.

REFERENCES

1. Kerr JFR, Wyllie AH, Currie AR. Apoptosis: basic biological phenomenonwith wide-ranging implications in tissue kinetics. Br J Cancer 1972; 26:239–257.

2. Wyllie AH, Kerr JFR, Currie AR. Cell death: the significance of apoptosis.Int Rev Cytol 1980; 68: 251–306.

3. Wyllie AH. Cell death: a new classification separating apoptosis fromnecrosis. In: Bowen ID, Lockshin RA, eds. Cell Death in Biology andPathology. London: Chapman & Hall, 1981; 9–34.

4. Smith CA, Williams GT, Kingston R, Jenkinson EJ, Owen JJT. Antibodiesto CD3/T-cell receptor complex induce death by apoptosis in immature Tcells in thymic cultures. Nature 1989; 337: 181–184.

5. Russel JH. Internal disintegration model of cytotoxic lymphocyte-mediatedcytotoxicity. Biol Rev 1981; 56: 153–197.

6. Duvall E, Wyllie AH. Death and the cell. Immunol Today 1986; 7:115–119.

7. Yamada T, Ohyama H. Radiation-induced interphase death of rat thymo-cytes is internally programmed (apoptosis). Int J Radiat Biol 1998; 53:65–75.

8. Wyllie AH. Glucocorticold-induced thymocyte apoptosis is associated withendogenous endonuclease activation. Nature 1980; 284: 555–556.

9. Kyprianou N, Isaacs JT. Activation of programmed cell death in the ratventral prostate after castration. Endocrinology 1988; 122: 552–562.

10. Solviter RS, Sollas AL, Dean E, Neubort S. Adrenalectomy-inducedgranule cell degeneration in the rat hippocampal dentate gyrus: characteriz-ation of an in vitro model of controlled neuronal death. J Compl Neurol1993; 330: 324–336.

11. Wyllie AH. Apoptosis (The 1992 Frank Rose Memorial Lecture). Br JCancer 1993; 67: 205–208.

12. Arends MJ, Morris RG, Wyllie AH. Apoptosis. The role of endonuclease.Am J Pathol 1990; 136: 593–600.

13. Wyllie AH, Morris RG, Smith AL, Dunlop D. Chromatin cleavagein apoptosis: association with condensed chromatin morphology anddependence on macromolecular synthesis. J Pathol 1984; 142: 67–77.

14. Hockenbery D. Review: Defining apoptosis. Am J Pathol 1995; 146:16–19.

15. Gorczyca W, Gong J, Darzynkiewicz Z. Detection of DNA strand breaksin individual apoptotic cells by the in situ terminal deoxynucleotidyltransferase and nick translation assays. Cancer Res 1993; 53: 1945–1951.

16. Gavrieli Y, Sherman Y, Ben-Sasson SA. Identification of programmed celldeath in situ via specific labeling of nuclear DNA fragmentation. J Cell Biol1992; 119: 493–501.

17. Ansari B, Coates PJ, Greenstein BD, Hall PA. In situ end-labeling detectsDNA strand breaks in apoptosis and other physiological and pathologicalstates. J Pathol 1995; 170: 1–8.

18. Abe R, Shimosegawa T, Kimura K, et al. The role of endogenousglucocorticoids in rat experimental models of acute pancreatitis. Gastro-enterology 1995; 109: 933–943.

19. Hogan LH, Smith GH. Unconventional application of standard lightand electron immunocytochemical analysis to aldehyde-fixed, araldite-embedded tissues. J Histochem Cytochem 1982; 30: 1301–1306.

20. Knight DP. Cytological Staining Methods in Electron Microscopy:Practical Methods in Microscopy. 5th edn. New York: North-Holland,1977: 36–37.

21. Sasano H, Imatani A, Shizawa S, et al. Cell proliferation and apoptosis innormal and pathological human adrenal. Mod Pathol 1995; 8: 11–17.

22. Sasano H, Date F, Itakura Y, Goukon Y, Nishihara T, Nagura H.Confocal laser scanning microscopy in cytopathology. Mod Pathol 1993; 6:625–629.

23. Sasano H. In situ end labeling and its applications to the study of endocrinedisease: how can we study programmed cell death in surgical pathologymaterials? Endocrine Pathol 1995; 6: 87–89.

24. Migheli A, Attanasio A, Shiffer D. Ultrastructural detection of DNA strandbreaks in apoptotic neural cells by in situ end-labeling techniques. J Pathol1995; 176: 27–35.

25. Vogt A, Takamiya H, Kim WA. Some problems invoved in postembeddingstaining. First International Symposium on Immunoenzymatic Techniques,INSERM Symposium No. 2. New York: American Elsevier, 1976; 109–116.

241ULTRASTRUCTURAL AND CLSM EXAMINATION OF TUNEL-POSITIVE CELLS

26. Fisch W, Hofmann W, Koshikallio J. The curing mechanism of epoxy resin.J Appl Chem 1956; 6: 429.

27. Baskin DG, Erlandsen SL, Parsons JA. Immunocytochemistry withosmium-fixed tissue. I. Light microscopic localization of growth hormoneand prolactin with the unlabeled antibody–enzyme method. J HistochemCytochem 1979; 27: 867–872.

28. Griegee R. Osmiumsaure-ester als Zwischenprodukte bei Oxydationen. AnnChem 1936; 522: 75.

29. Wolman M. The reaction of osmium tetroxide with tissue components. ExpCell Res 1957; 12: 231–240.

30. Munger BI. Staining methods applicable to sections of osmium-fixed tissuefor light microscopy. J Biophys Biochem 1961; 11: 502–506.

242 K. KIMURA ET AL.