Histopathological changes in insulin, glucagon and somatostatin cells in the islets of NOD

mice during cyclophosphamide-accelerated diabetes: A combined immunohistochemical

and histochemical study

Shiva Reddy1,2,*, Praneeti Pathipati1, Yan Bai1, Elizabeth Robinson3 & Jacqueline M. Ross4

1School of Biological Sciences, University of Auckland, Private Bag, 92019, Auckland, New Zealand2Department of Paediatrics, University of Auckland, Private Bag, 92019, Auckland, New Zealand3Department of Community Medicine, University of Auckland, Private Bag, 92019, Auckland, New Zealand4Department of Anatomy with Radiology, University of Auckland, Private Bag, 92019, Auckland, New Zealand

*Author for correspondence (e-mail: s.reddy@auckland.ac.nz)

Received 22 March 2005 and in revised form 16 May 2005

Summary

The cyclophosphamide model of accelerated diabetes in the NOD mouse is a useful model of insulin-dependentdiabetes mellitus (IDDM). Knowledge on the progressive destruction of beta cells and the fate of other islet endocrinecell-types in this model is sparse. We employed immunohistochemistry and histochemistry, to study temporal changesin islet cell populations, insulitis and glucose transporter-2 expression during cyclophosphamide administration.Cyclophosphamide was administered to day 95 female NOD mice and the pancreas studied at days 0 (=day 95), 4, 7,11 and 14 after treatment and in age-matched control mice. At day 0, a majority of the endocrine cells were insulin-positive. Glucagon and somatostatin cells were mostly in the islet periphery and also internally. In the cyclophos-phamide group, insulitis was moderate at day 0, declined at day 4 but increased progressively from day 7. The extentof insulitis in treated mice which were diabetes-free at day 14 was comparable to age-matched control mice. From day11, the marked increase in insulitis correlated with a reciprocal decline in the extent of insulin immunostained isletarea. At day 14, the mean insulin area per islet was markedly less in diabetic mice than in age-matched non-diabetictreated and controls. At diabetes, some islets showed co-expression of glucagon and insulin. Our studies suggest thatthe mean number of glucagon or somatostatin cells per islet does not vary during the study. Glucose transporter-2immunolabelling was restricted to beta cells but declined in those adjacent to immune cells. We conclude that in thecyclophosphamide model, there is specific and augmented destruction of beta cells immediately prior to diabetesonset. We speculate that the selective loss of glucose transporter-2 shown in this study suggests the existence of adeleterious gradient close to the immune cell and beta cell surface boundary.

Introduction

Insulin-dependent diabetes mellitus (IDDM) resultsfrom the progressive destruction of insulin-producingbeta cells (Eisenbarth et al. 1987). During the disease,beta cell loss is mediated by immune processes origi-nating from the intra-islet immune cells which infiltratethe islets during a prolonged prediabetic period (Eisen-barth et al. 1987, Kikutani & Makino 1992). Multiplemechanisms including various cytokines, free radicalsand the perforin-granzyme pathway have been proposedas effectors of beta cell destruction (Rabinovitch 1992,1998, Mathis et al. 2001).

In humans with IDDM and in animal models of thedisease, such as the non-obese diabetic (NOD) mouseand the low-dose streptozotocin (stz) mouse, the islet

immuno-inflammatory processes are beta cell specificsince the closely located intra-islet non-beta cells escapethe destructive process (Fujita et al. 1982, Foulis et al.1984, Harada 1987, Reddy et al. 1995, Li et al. 2000).Insulin-dependent diabetes mellitus consists of a benignimmune phase termed insulitis which may persist over aprotracted period lasting several months to years inhumans and over several weeks in the NOD mouse,without significant beta cell destruction. This is thenfollowed by an overt phase of rapid beta cell lossresulting in clinical disease.

In the NOD mouse model of spontaneous diabetes,the onset of clinical disease is asynchronous (usuallybetween 90 and 250 days). Cyclophosphamide was firstemployed by Harada and Makino to synchronise andaccelerate diabetes in the NOD mouse and was

Journal of Molecular Histology (2005) 36: 289–300 � Springer 2005DOI 10.1007/s10735-005-7330-4

subsequently proposed to have an immunosuppressiveeffect also (Harada & Makino 1984, Yasunami & Bach1988). Numerous investigators have since employed thismodel to study the immunopathogenesis of IDDM andto evaluate a range of preventive therapies (Bacelj et al.1989, Kay et al. 1991, Zhang et al. 1993, Rothe et al.1994, Cailleau et al. 1997, Reddy et al. 1999, 2000,2001a, b, Maron et al. 2001, Krueger et al. 2003). In thecyclophosphamide model, a single injection of the drugto female NOD mice during established insulitis, resultsin clinical diabetes approximately 14 days later in mostanimals (Reddy et al. 1999). Immune cell infiltrationinto islets lessens at 4 days after cyclophosphamide butincreases rapidly from day 7 to 14 (Kay et al. 1991,Reddy et al. 1999). The marked intra-islet infiltrationobserved after day 7 is also associated with a rise inmacrophage expression of inducible nitric oxide syn-thase (iNOS) and various cytotoxic cytokines such asinterleukin-1b (IL-1b) (Reddy et al. 1999, 2000, 2001b).However, in this accelerated model, only limited studieshave examined the temporal destruction of beta cellswith advancing disease. In addition, little informationexists on the fate of glucagon, somatostatin and pan-creatic polypeptide cells within the islets during cyclo-phosphamide-accelerated diabetes.

In the present study, we have employed dual- andtriple-label immunohistochemistry in conjunction withhistochemical techniques to examine the temporal lossof beta cells during cyclophosphamide-accelerated dia-betes in the NOD mouse and to determine the fate andintra-islet distribution of glucagon and somatostatincells during disease progression. We have also studiedchanges in the expression of glucose transporter-2 (glut-2)on the beta cell surface during cyclophosphamide accel-erated diabetes.

Materials and methods

Non-obese diabetic mice and assessment of diabetes

A colony of NOD mice was established under conven-tional conditions at the Animal Resources Unit of theAuckland School of Medicine. The spontaneous diabe-tes rate is approximately 50% among females and lessthan 1% among males between the ages of 120 and250 days. Diabetes was defined as the presence of heavyglycosuria on three consecutive days and confirmed by anon-fasting blood glucose level exceeding 12 mM. Non-fasting blood glucose levels ranged from 4 to 6 mM innon-diabetic NOD mice.

Study groups, administration of cyclophosphamide andtissue collection

Female NOD mice were obtained from several breedingpairs, weaned at day 21 and allocated at random intotwo groups (18 mice/group).

Group 1: Standard mouse chow (Diet 86; Reddy et al.2001a) plus drinking water (control group)

Group 2: Standard mouse chow plus drinking water(cyclophosphamide group)

At day 95, all but three mice from group 2 were injectedwith cyclophosphamide (Sigma, St. Louis, Mo., USA;300 mg/kg body weight; Reddy et al. 2001a) by theintraperitoneal route while mice from group 1 wereinjected with diluent (sterile water). Three mice fromeach group were then killed at days 0, 4, 7 and 11, andthe pancreas along with a small piece of the adjoiningspleen was harvested (see below). The remaining 6 micefrom each group were followed until the development ofdiabetes when they were killed and pancreas excised(irrespective of diabetes development). Each pancreaswas divided into two halves immediately after removaland was either snap-frozen (splenic end) or fixed inBouin’s solution (duodenal portion).

Tissues fixed in Bouin’s fixative were embedded inparaffin and sectioned at 5 lm thickness. They werethen prepared for haematoxylin and eosin (H&E)staining and for immunohistochemistry (see below).Cryosections (6 lm) were fixed in acetone and stored at)20 �C until required.

Primary antibodies, non-immune IgG and normal sera

Guinea pig anti-bovine insulin serum was availablein this laboratory and has been previously employed inimmunohistochemical staining of mouse insulin inpancreatic sections (Reddy et al. 1988). Rabbit anti-glucagon and rabbit anti-somatostatin sera were fromICN Pharmaceuticals, Inc, Aurora, Ohio, USA. Mousemonoclonal antibodies specific for glucagon andsomatostatin were kindly provided by Dr P Mackay,Novo-Nordisk, Bagsvaerd, Copenhagen, Denmark andemployed recently in immunohistochemical studies inthis laboratory (Reddy et al. 2004). Rabbit anti-mouseglucose transporter-2 (glut-2) was supplied by Dr BThorens, University of Lausanne, Switzerland, as affin-ity-purified rabbit IgG. This antibody has been previ-ously employed for the immunolocalization of glut-2 inmouse islet beta cells and does not immunostain gluca-gon, somatostatin or pancreatic polypeptide hormonecells (Thorens 1992, Reddy et al. 1998). Rat monoclonalantibodies to mouse macrophages (clone M/170) werepurchased from Serotec, Oxford, United Kingdom andto CD3 (clone KT3, detects all mouse CD4 and CD8 Tcells) were kindly provided by Dr H Georgiou, Walterand Eliza Hall Institute of Medical Research, Mel-bourne, Australia. The immunohistochemical specifici-ties of the antibodies employed to detect mousemacrophages and T cells have been reported previously(Reddy et al. 1998, 2003).

In the immunohistochemical procedure, all primary anti-sera were titrated to give maximal immunohistochemical

290 S. Reddy et al.

reactivity. Normal sera from a variety of mammalian specieswere available in this laboratory and were employed inimmunohistochemical specificity studies and as blockingreagents.

Immunolabelling of islet endocrine cells by single-,dual- and triple-label immunohistochemistry

Each of the three serial sections per slide were immu-nolabelled for insulin, glucagon or somatostatin asreported previously (Reddy et al. 1988, 2004). Briefly,sections were deparaffinised with xylene, dehydrated inincreasing concentrations of ethanol, washed in waterand equilibrated in phosphate-buffered saline (PBS),pH 7.4. In the immunohistochemical protocol, PBS wasused as the diluent and also acted as the wash buffer.Following blocking of sections with 5% v/v normaldonkey serum, they were incubated with either guinea-pig anti-insulin (1:100), rabbit anti-glucagon (1:50) orrabbit anti-somatostatin (1:50) for 1 h at 37 �C. Fol-lowing washing, sections were incubated with eitherdonkey anti-guinea pig IgG-FITC (1:50; JacksonImmunoresearch Laboratories, West Grove, PA, USA)or donkey anti-rabbit IgG-Texas Red (1:100; JacksonImmunoresearch Laboratories) for 1 h at 37 �C.

Sections immunolabelled for glucagon and somato-statin were dual-labelled subsequently for either insulin(section previously labelled for glucagon) or glucagon(section previously labelled for somatostatin; usingmouse anti-glucagon monoclonal antibodies 1:50). Thesecond set of primary antibodies was applied for 1 h at37 �C. After washing, sections were reacted with eitherdonkey anti-guinea pig IgG-FITC (1:50) or donkeyanti-mouse IgG-FITC (Jackson ImmunoresearchLaboratories; 1:50) for 1 h at 37 �C. Sections were thenwashed and prepared for microscopic examination witha Zeiss Axioskop 2 Plus microscope equipped with anAxiocam digital camera or a Leica DMRE microscopeequipped with a Leica DC500 digital camera.

Selected sections were also triple-labelled sequentiallyfor insulin (FITC secondary antibody), glucagon (don-key anti-rabbit IgG-Texas Red) and finally forsomatostatin (mouse anti-somatostatin followed bydonkey anti-mouse IgG-cyanin-5, 1:100; JacksonImmunoresearch Laboratories). Triple-labelled sectionswere examined by confocal microscopy.

Evaluation of insulitis

The degree of insulitis was evaluated in pancreatic sec-tions following haematoxylin and eosin staining usingZeiss Axiovision software. Briefly, the area occupied byendocrine cells and insulitis within the same islet wasdetermined separately and the total islet area wasexpressed as the sum of the two islet compartments. The

area of insulitis for each islet was expressed as a per-centage.

Evaluation of beta cell area

Following immunohistochemical staining for insulin inparaffin sections, the beta cell stained area of each isletand the corresponding islet area were determined asabove following examination by fluorescence micros-copy. The beta cell area for each islet was then calcu-lated as a percentage.

Enumeration of glucagon and somatostatinimmunolabelled cells

Sections dual-labelled for glucagon and somatostatinat day 0 and day 14 (cyclophosphamide-treated dia-betic and non-diabetic) and day 14 controls wereenumerated separately for glucagon and somatostatin-positive cells in each islet by either direct microscopicexamination and counting (20· objective) or followingimaging. The mean number of glucagon or somato-statin cells per islet ±SD for each experimental groupwas then calculated.

Immunolabelling of glucose transporter-2and co-localization

Immunohistochemical labelling of mouse glucosetransporter-2 (mouse glut-2) was carried out in selectedcryosections from various experimental groups asreported previously (Reddy et al. 1998). Briefly, sectionswere fixed in cold acetone for 10 min, equilibrated inPBS and incubated with blocking solution (5% normaldonkey serum in PBS) for 1 h at 37 �C. After washing,sections were incubated with a mixture of rabbit anti-mouse glut-2 (final dilution: 1:5000) and guinea pig anti-insulin (final dilution 1:100) for 18 h at 4 �C. Sectionswere washed and incubated with donkey anti-rabbitIgG-biotin (1:100, Jackson Immunoresearch) for 1 h at37 �C. They were then incubated with streptavidin-Texas Red (1:100, Jackson Immunoresearch) for 1 h at37 �C. After washing sections were incubated withdonkey anti-guinea pig IgG-FITC (1:50) for 1 h at37 �C before microscopic examination.

Adjacent sections were also immunolabelledsequentially for mouse glut-2 followed by immunola-belling for either CD3 cells or macrophages with ratprimary monoclonal antibodies as reported previously(Reddy et al. 1998, 2003). Immune cells were thenvisualized using goat anti-rat IgG-cyanin-2 (1:100,Rocklands, Gilbertsville, PA, USA) or donkey anti-ratIgG-cyanin-5 secondary antibodies (1:100, JacksonImmunoresearch Laboratories). Some sections weretriple-labelled sequentially for glut-2 (Texas Red

291Histopathological changes in insulin, glucagon and somatostatin cells

fluorochrome), insulin (FITC fluorochrome) and thenfor CD3 cells (cyanin-5 fluorochrome).

Microscopy and imaging

Immunostained sections and sections stained by H&E(including counterstained sections) were examined witheither a Zeiss or Leica fluorescence light microscope. Atleast 10 islets from each pancreas and the entire exocrineregion were analysed and representative islets photo-graphed. Images were recorded with a digital cameraattached to Zeiss or Leica microscopes. Final imageswere assembled with Adobe Photoshop (Adobe Sys-tems, Mountain View, CA, USA).

Immunofluorescent sections were also examined byconfocal laser scanning microscopy (Leica, Heidelberg,Germany), as described previously (Reddy et al. 1999,2001b, 2002, 2004). Separate as well as merged imageswere saved as TIFF files and assembled in AdobePhotoshop.

Controls for immunohistochemistry

In the immunohistochemical procedure, all primaryantibodies were substituted with the corresponding non-immune sera or IgG at equivalent dilutions. In addition,primary (for two-step staining) and secondary steps (forthree-step staining) were omitted. In the double- andtriple-immunolabelling protocols, species-incompatiblesecondary antibodies were also tested to confirmimmunohistochemical specificity.

Statistical analysis

Generalised linear models were used to investigate thedifferences in changes in the percentage area of insulitisand insulin over time in the control and cyclophospha-mide groups. Arc sine transformations were used on the

percentages so that the data conformed to assumptions ofnormality. Analysis of variance (ANOVA) was used toinvestigate differences in the number of glucagon andsomatostatin cells at day 0 and day 14 in the variousgroups.

Insulitis scores and insulin areas were expressed as themean and 95% confidence limits for each experimentalgroup. Somatostatin and glucagon cell numbers wereexpressed as the mean ± SD for day 0, day 14 (cyclo-phosphamide diabetic and non-diabetic) and day 14controls.

Islet analysis

The number of animals and islets employed foranalysis of insulitis, insulin area and glucagon andsomatostatin cell numbers at each time-point duringadministration of cyclophosphamide (Cy) or diluentare shown below:

Results

Development of cyclophosphamide-induced diabetes

Following administration of cyclophosphamide, diabetesdeveloped in 3 out of 6 mice by day 14 (Group 2) asreported previously (Reddy et al. 2003). Diabetes didnot develop in NOD mice from Group 1 (withoutcyclophosphamide treatment) during the same period.

Insulitis

The variation in the mean insulitis scores and the 95%confidence limits during cyclophosphamide and diluenttreatment of NOD mice are shown in Figure 1. In thecyclophosphamide group, the degree of insulitis declinedsharply at day 4 to almost zero compared with day 0 and

Study groups Insulitis: number of animals

studied and total number

of islets examined ()

Insulin area: number of animals studied and

total number of islets examined ()

Glucagon and somatostatin

cell numbers: number of

animals studied and total

number of islets examined ()

Day 0 3 (40) 2 (20) 2 (20)

Day 4 control 2 (29) 2 (20) –

Day 4 Cy 2 (27) 2 (17)

Day 7 control 3 (46) 3 (31) –

Day 7 Cy 3 (41) 3 (30)

Day 11 control 2 (22) 3 (23) –

Day 11 Cy 3 (65) 3 (29)

Day 14 control 3 (22) 4 (34) 3 (21)

Day 14 Cy diabetic 3 (36) 2 (24) 2 (17)

Day 14 Cy non-diabetic 3 (37) 3 (40) 2 (19)

Cy: cyclophosphamide.

292 S. Reddy et al.

then increased at day 7. It rose sharply at day 11 whichwas maintained at day 14. At this stage (day 14),cyclophosphamide-treated mice which did not developdiabetes, showed a significantly lower degree of insulitiscompared with mice from day 11 and day 14 (diabetic;results not shown). In mice that did not receive cyclo-phosphamide, the insulitis scores were similar at days 0,4, 7, 11 and 14 but remained considerably lower thancyclophosphamide-treated animals at days 11 and 14(diabetic). The change over time in the mean insulitisscores differed between the treatment and controlgroups and reached statistical significance (p = 0.04).

Immunohistochemical analysis of islet endocrine cellsduring cyclophosphamide-accelerated diabetes

Representative merged photomicrographs of isletsshowing the relative distribution of insulin and glucagoncells and somatostatin and glucagon cells in the sameislet from adjacent sections during the study period areshown in Figures 2–4. Sections previously dual-labelledfor insulin and glucagon and counterstained by H&Eare also shown to depict the pathological features of theimmunostained islets (Figures 2 and 3).

In the control groups at days 0 and 4, several isletswith and without insulitis were present in the samesection. In islets devoid of insulitis, the majority of isletendocrine cells were insulin-positive, co-existent withglucagon and somatostatin cells, which were locatedmostly in the islet periphery. However, islets which weredevoid of insulitis also contained several glucagon cellsin intra-islet regions (Figures 2a–c and 4a). Islets withsignificant insulitis (Figure 2d–f) were not only associ-ated with a reduction in beta cell area (see below) butalso with a marked re-distribution of glucagon and

somatostatin-positive cells located in the internalregions of the islet. A similar pattern of immunolabel-ling was also observed in the control groups at days 4, 7,11 and 14 (results not shown).

At day 4 after cyclophosphamide administration, theabsence of significant insulitis was associated withpreservation of beta cells and the location of glucagonand somatostatin cells mostly in the periphery of islets(Figure 2g–i). However, at this stage, islets with someinsulitis also showed an enhanced degree of internali-zation of glucagon and somatostatin cells (Figure 2j–l).

In the cyclophosphamide group, at days 7, 11 and 14(diabetic), the marked increase in insulitis was accom-panied by a significant decline in beta cell area and by aninternalized re-distribution of glucagon and somato-statin cells (Figure 3a–l).

In the day 14 cyclophosphamide group (diabetic),some islets showed a significant number of insulin-posi-tive cells with co-existing insulitis (Figure 4b). However,several islets were also atrophied and devoid of insulitis.Such islets had a considerably reduced number of insulincells while glucagon and somatostatin cells were muchmore numerous (Figure 4c–f). In some of the atrophiedislets, a proportion of intra-islet endocrine cells also co-expressed insulin and glucagon (Figure 4c). At the dia-betic stage at day 14, co-localization of glucagon andsomatostatin was not observed (Figure 4e, f).

Changes in islet insulin area

The variation in the mean insulin area per islet and the95% confidence limits at different stages of cyclophos-phamide and diluent-treated groups are shown inFigure 5. At days 4 and 7, the mean insulin area per isletwas similar in both the cyclophosphamide and controlgroups. However, there was no evidence of a differentpattern over time (p = 0.8). There was no difference inthe main effect of group (p = 0.3) but the percentageinsulin area did change over time in both the groups(p = 0.03). For both the control and cyclophosphamidegroups, the percentage area occupied by insulin cellsdecreased over time. In the diabetic group at day 14 (2animals) the mean insulin area at day 14 diabetic waslower (20%) than the day 14 non-diabetic mice (n = 3,mean insulin area: 50%). However, statistical signifi-cance between the two groups was not tested due toinsufficient sample points.

Glucagon and somatostatin cell numbers

Changes in the mean number of glucagon and somato-statin cells per islet at day 0 and day 14 (cyclophos-phamide-treated diabetic and non-diabetic mice andcontrols) are shown in Figure 6a. In the diabetic groupat day 14, glucagon cell numbers per islet decreasedslightly from day 0 but remained similar to days 14 non-

Figure 1. The mean percent insulitis scores and 95% confidence limits

at various time-points following administration of cyclophosphamide

(hatched bars) or diluent (open bars) to day 95 female NOD mice.

Note: day 14 cyclophosphamide-treated group includes pooled data

from 3 diabetic and 3 non-diabetic cyclophosphamide-treated mice.

293Histopathological changes in insulin, glucagon and somatostatin cells

diabetic and control groups. Changes in glucagon cellnumbers between day 0 and day 14 (cyclophosphamide-treated and control groups) did not show statisticalsignificance. There was no statistical difference betweenthe mean number of somatostatin cells per islet betweenday 0 and day 14 (cyclophosphamide-treated and con-trol groups, Figure 6).

Glucose transporter-2 expression

Representative islets depicting glut-2 immunolabellingduring cyclophosphamide administration are shown inFigure 7. In islets with minimum insulitis such as seen at

day 0, some of the beta cells close to the area of insulitiswere devoid of glut-2 (Figure 7a–c) whereas almost allbeta cells in islets without insulitis co-expressed thetransporter (Figure 7d,e). In islets with slightly moreadvanced insulitis such as seen in a proportion of isletsat day 4 cyclophosphamide group, there was a furtherdecrease in the expression of glut-2 in beta cells in closeproximity to the insulitis region. (Figure 7g–i). At day11, when most islets showed advanced insulitis, severalbeta cells adjacent to intra-islet CD3-positive cells alsoshowed attenuated expression of glut-2 (Figure 7f). Atday 14 (diabetic and non-diabetic), a more pronouncedloss of glut-2 immunolabelling was observed in beta cells

Figure 2. Representative merged images of adjacent pancreatic sections showing islets from days 0 and 4 (treated and controls) female NOD mice

dual-labelled for insulin and glucagon (left panel) and glucagon and somatostatin (middle panel) and the corresponding images following H&E

counterstaining. Left panel: insulin (green), glucagon (red); middle panel: glucagon (red) and somatostatin (green); right panel: H&E counter-

stained images of islets from left panel. Scale bars: (a–i) 50 lm; (j–l) 100 lm. INS: insulin; GLU: glucagon; SOM: somatostatin; Cy: cyclophos-

phamide.

294 S. Reddy et al.

adjacent to intra-islet CD3 cells and macrophages(Figure 7j–l).

Discussion

The cyclophosphamide model of accelerated diabetes inthe NOD mouse is being employed extensively forstudies into the immunopathogenesis and prevention ofIDDM. Although several studies have examined theunderlying immune cellular and molecular changesduring disease in this model, investigations on the tem-

poral changes in islet endocrine cell populations withadvancing disease and the resultant augmentation ininsulitis have been limited.

Our studies show that in the day 95 NOD mouse notgiven cyclophosphamide, despite variable insulitis, anoverall moderate level of insulitis is maintained duringthe 14 day study period and approximately 60–70% ofthe islet area is occupied by insulin immunoreactivecells. In the cyclophosphamide group, the level of insu-litis declined at day 4 but rose at day 7 when it wassimilar to control animals. During this initial phase(until day 7) the extent of insulin immunostaining in the

Figure 3. Representative merged images of adjacent pancreatic sections showing islets from cyclophosphamide-treated (days 7, 11, 14) female

NOD mice dual-labelled for insulin and glucagon (left panel) and glucagon and somatostatin (middle panel) and the corresponding images

following H&E counterstaining. Left panel: insulin (green) and glucagon (red); middle panel: glucagon (red) and somatostatin (green); right panel:

H&E counter-stained images of islets from left panel. Note almost complete absence of endocrine cells in the region of insulitis except in j–l where

some are interspersed among intra-islet immune cells. Scale bars: 50 lm. INS: insulin; GLU: glucagon; SOM: somatostatin; Cy: cyclophosphamide.

295Histopathological changes in insulin, glucagon and somatostatin cells

treated and untreated groups was similar. However,from day 11 onwards, the marked increase in the degreeof insulitis in the cyclophosphamide group was associ-ated with a significant loss of insulin cells and with apronounced intra-islet re-distribution of insulin, glucagonand somatostatin immunoreactive cells. Thesepathological features in the islet became more marked atonset of diabetes. However, insulin immunoreactive cellswere not totally absent in several islets at the onset of

cyclophosphamide-accelerated diabetes. The failure ofsome mice not to develop diabetes 14 days after cyclo-phosphamide can be explained by the persistence of agreater extent of insulin immunoreactive cells, as sug-gested in this study. The reason why some mice arerefractory to cyclophosphamide-induced diabetes isunclear. Our immunohistochemical findings show thatin this accelerated model, beta cell loss occurs rapidlyfrom day 11 onwards. These results are in accord with

Figure 4. Merged confocal images of pancreatic sections from day 0 and cyclophosphamide-treated day 14 diabetic mice showing islets either

triple-labelled for insulin, glucagon and somatostatin or dual-labelled for insulin and somatostatin and glucagon and somatostatin. (a) Day 0: Islet

shows preservation of a majority of beta cells (green) with peripheral and internalized of glucagon (red) and somatostatin cells (blue). (b–f)

Cyclophosphamide-treated day 14 diabetic islets. (b) Some loss of beta cells (green) with re-distribution of glucagon (red) and somatostatin cells

(blue) within the islet. (c) Atrophied islet shows insulin (green) and glucagon (red). Note a proportion of endocrine cells co-labelled for insulin and

glucagon (arrows); separate insulin cells are absent. (d) Compressed islet with several somatostatin cells (red) and only a few weakly-stained insulin

cells (green). (e) Same islet as in (d) from an adjacent section shows glucagon (red) and somatostatin (green) cells. Note separate intra-islet glucagon

and somatostatin cells. (f) An islet shows glucagon (red) and somatostatin (green) cells; a part of a second islet is also present. Scale bars: (a, b)

30 lm; (c–f) 40 lm. Cy: cyclophosphamide.

296 S. Reddy et al.

those of others who showed that during spontaneousdiabetes in the NOD mouse, beta cell loss occursabruptly just before onset of clinical diabetes (Shimadaet al. 1996). The rapid loss of beta cells from day 11observed in the present study is also supported by ourprevious immunohistochemical data which showedaugmented expression of beta cell toxic cytokines andinducible nitric oxide synthase (iNOS) within the isletsduring the late phase of cyclophosphamide-induceddiabetes in this model (Reddy et al. 1999, 2000, 2001a,b). Although we have not addressed the molecularmechanisms of beta cell death in this study, evidence forboth apoptotic and necrotic beta cell death duringIDDM has been proposed (O’Brien et al. 1997, Augsteinet al. 1998, Saldeen 2000, Fehsel et al. 2003, Reddyet al. 2003).

The present immunohistochemical analysis suggeststhat during cyclophosphamide-accelerated diabetesglucagon and somatostatin cell numbers are unaffectedalthough their distribution within the islets prior todisease and at diabetes onset change markedly. We didnot conduct a detailed enumeration of the two endo-crine cell-types at all the study periods. However, ourdata suggest that both the islet cell-types escape thedestructive process in this model despite the presence ofa cytotoxic islet milieu at the later stages of cyclophos-phamide-induced diabetes. Others have shown anexpansion of glucagon cells at onset of diabetes in themultiple low dose streptozotocin mouse model (Li et al.2000). Alpha cell neogenesis was also observed at onsetof spontaneous diabetes in the NOD mouse (O’Reillyet al. 1997). These results suggest that there may bediscrete differences in the processes involving islet cellsurvival and selective cell expansion in the three animal

models. At diabetes-onset, a small proportion of isletcells which co-express insulin and glucagon, shown inthis study, is significant and implies that an extremediabetic environment may recapitulate some of theembryonic developmental processes governing islethormones (Bonner-Weir 2000). During pancreaticdevelopment in the fetal rat, co-expression of insulin andglucagon was observed in a proportion of pancreaticcells from day 12.5 to 20 of gestation (Hashimoto et al.1988). Mechanisms which re-activate the molecularprocesses involved in the co-expression of glucagon andinsulin around the onset of cyclophosphamide-induceddiabetes in the NOD mouse are unclear and requirefurther studies.

The progressive decline in glut-2 immunostaining inselective beta cells, shown here, is in accord with previousresults in other mouse models of diabetes. Forexample, islets from the db/db mouse when transplantedin a diabetic environment showed reduced expression ofglut-2 which is normalised by euglycaemia (Thorenset al. 1992). Thus the expression of the transporter can

Figure 5. The mean percent insulin area per islet and 95% confidence

limits at various time-points following administration of cyclophos-

phamide (hatched bars) or diluent (open bars) to day 95 female NOD

mice. Note: day 14 cyclophosphamide-treated group includes pooled

data from 2 diabetic and 3 non-diabetic cyclophosphamide-treated

mice.

Figure 6. The mean number ± SD of glucagon (top graph) and

somatostatin (lower graph) cells per islet at day 0 and day 14 control

mice and day 14 cyclophosphamide-treated diabetic and non-diabetic

mice. Controls: open bars; cyclophosphamide-treated: hatched bars.

Som: somatostatin, D: diabetic; ND: non-diabetic.

297Histopathological changes in insulin, glucagon and somatostatin cells

Figure 7. Light and confocal photomicrographs of pancreatic sections showing islets co-labelled for glucose transporter-2, insulin cells, CD3 cells

and macrophages during cyclophosphamide administration. Selected micrographs following counterstaining by H&E are also included. (a–c) Day

0: An islet dual-labelled for insulin (a) and glut-2 (b); (c) shows the same islet counterstained by H&E. Majority of insulin cells are glut-2 positive

while some insulin-positive cells are glut-2 negative (arrows). Top two arrows in (a) and (b) correspond to a region of limited insulitis adjacent to

reduced immunolabelling of glut-2 (c). (d,e) Day 0: An islet dual-labelled for insulin (d) and glut-2 (e); note almost complete co-localization. (f) Day

11, cyclophosphamide-treated: An islet dual-labelled for glut-2 (red) and CD3 T cells (green); arrows point to reduced glut-2 labelling in cells

adjacent to T cells (arrows). (g–i): Day 4, cyclophosphamide-treated: An islet dual-labelled for insulin (g) and glut-2 (h). Arrows point to reduced

glut-2 labelling in selective beta cells adjacent to the area of minimum insulitis (i). (j) Day 14 cyclophosphamide-treated diabetic: An islet triple-

labelled for insulin (green), glut-2 (red) and macrophages (blue). Note a few beta cells close to macrophages are negative for glut-2 immunolabelling

(arrows). (k, l) Day 14 cyclophosphamide-treated non-diabetic: An islet (k) triple-labelled for insulin (green), glut-2 (red) and CD3 cells (blue).

Arrows point to insulin cells adjacent to CD3 cells which are negative for glut-2. (l) Same islet from an adjacent section dual-labelled for glut-2 (red)

and macrophages (blue). (a–i) are light micrographs while (j–l) are confocal micrographs. INS: insulin; GLUT2: glucose transporter-2; CD3: CD3

T lymphocytes; MU: macrophages Scale bars: (a–i) 50 lm; (j–l), 20 lm.

298 S. Reddy et al.

be influenced by blood glucose levels. Underexpressionof this transporter in the BB rat, in the low dose strep-tozotocin and NOD mouse models has also beenreported (Orci et al. 1990, Wang & Gleichmann, 1995,Reddy et al. 1998). However, in the NOD mouse, wehave previously shown that prior to the onset of insulitis(before day 35) and at the stage of minimum insulitis(day 60), glut-2 immunolabelling is seen in almost allbeta cells (Reddy et al. 1998). With increasing insulitis,the expression of the transporter was reduced markedly,particularly in those beta cells which were locatedadjacent to the region of insulitis. In the present study, asimilar attenuation in glut-2 labelling at the immune-beta cell boundary was observed and became moremarked at onset of cyclophosphamide-induced diabetes.The mechanisms responsible for the selective loss ofglut-2 expression in beta cells close to the region of in-sulitis are unclear. We speculate that toxic molecules,including free radicals such as nitric oxide and theinflammatory cytokines, IL-1b and interferon-c (IFN-c),previously shown to be expressed locally and inincreasing levels during the later stages of cyclophos-phamide-accelerated diabetes in the NOD mouse maybe responsible for the selective loss of beta cell glut-2.

In conclusion, the present studies demonstrate pro-gressive and rapid destruction of beta cells in the latterphase of cyclophosphamide-induced diabetes whileglucagon and somatostatin cells may be spared from thedestructive events. Thus, use of this accelerated model ofdiabetes in the NOD mouse may continue to contributesignificantly towards further studies on the mechanismsof beta cell destruction and serve as a valuable animalmodel for evaluating potential therapies for the pre-vention of IDDM in humans.

Acknowledgements

This study was supported by the Auckland MedicalResearch Foundation and the Paykel Trust of NewZealand. We are grateful to Ms Beryl Davy and MsLorraine Rolston for histological assistance. We thankProfessor B Thorens, University of Lausanne, Switzer-land, for providing highly specific antibodies againstmouse glucose transporter-2.

References

Augstein P, Elefanty AG, Allison J, Harrison LC (1998) Apoptosis

and beta-cell destruction in pancreatic islets of NOD mice with

spontaneous and cyclophosphamide-accelerated diabetes. Diabeto-

logia 41: 1381–1388.

Bacelj A, Charlton B, Mandel TE (1989) Prevention of cyclophospha-

mide-induced diabetes by anti-Vb8 T-lymphocyte-receptor monoclo-

nal antibody therapy in NOD/Wehi mice. Diabetes 38: 1492–1495.

Bonner-Weir S (2000) Islet growth and development in the adult.

J Mol Endocrinol 24: 297–302.

Cailleau C, Diu-Hercend A, Ruuth E, Westwood R, Carnaud C (1997)

Treatment with neutralizing antibodies specific for IL-1b prevents

cyclophosphamide-induced diabetes in nonobese diabetic mice.

Diabetes 46: 937–940.

Eisenbarth GS, Connelly J, Soeldner JS (1987) The ‘‘natural’’ history

of type 1 diabetes. Diabetes Metab Rev 3: 873–891.

Fehsel K, Kolb-Bachofen V, Kroncke KD (2003) Necrosis is the

predominant type of islet cell death during development of insulin-

dependent diabetes mellitus in BB rats. Lab Invest 83: 549–559.

Foulis AK, Stewart JA (1984) The pancreas in recent-onset type 1

(insulin-dependent) diabetes mellitus: insulin content of islets,

insulitis and associated changes in the exocrine acinar tissue.

Diabetologia 26: 456–461.

Fujita T, Yui R, Kusumoto Y, Serizawa Y, Makino S, Tochino Y

(1982) Lymphocytic insulitis in a ’non-obese diabetic (NOD)’ strain

of mice: an immunohistochemical and electron microscope investi-

gation. Biomed Res 4: 429–443.

Harada M, Makino S (1984) Promotion of spontaneous diabetes in

non-obese diabetes-prone mice by cyclophosphamide. Diabetologia

27: 604–606.

Harada M (1987) Immune disturbance and pathogenesis of non-obese

diabetes-prone (NOD) mice. Exp Clin Endocrinol 89: 251–258.

Hashimoto T, Kawano H, Daikoku S, Shima K, Taniguchi H, Baba S

(1988) Transient co-appearance of glucagon and insulin in the pro-

genitor cells of the rat pancreatic islets. Anat Embryol 178: 489–497.

Kay TWH, Campbell IL, Harrison LC (1991) Characterization of

pancreatic T lymphocytes associated with beta cell destruction in the

non-obese diabetic (NOD) mouse. J Autoimmun 4: 263–276.

Kikutani H, Makino S (1992) The murine autoimmune diabetes

model: NOD and related strains. Adv Immunol 51: 285–322.

Krueger T, Wohlrab U, Klucken M, Schott M, Seissler J (2003)

Autoantigen-specific protection of non-obese diabetic mice from

cyclophosphamide-accelerated diabetes by vaccination with den-

dritic cells. Diabetologia 46: 1357–1365.

Li Z, Karlsson FA, Sandler S (2000) Islet loss and alpha cell expansion

in type 1 diabetes induced by multiple low-dose streptozotocin

administration in mice. J Endocrinol 165: 93–99.

Maron R, Guerau-de-Arellano M, Zhang X, Weiner HL (2001) Oral

administration of insulin to neonates suppresses spontaneous and

cyclophosphamide induced diabetes in the NODmouse. J Autoimmun

16: 21–28.

Mathis D, Vence L, Benoist C (2001) b-Cell death during progression

to diabetes. Nature 414: 792–798.

O’Brien BA, Harmon BV, Cameron DP, Allan DJ (1997) Apoptosis is

the mode of b-cell death responsible for the development of IDDM

in the nonobese diabetic (NOD) mouse. Diabetes 46: 750–757.

Orci L, Unger RH, Ravazzola M, Ogawa A, Komiya I, Baetens D,

Lodish HF, Thorens B (1990) Reduced beta-cell glucose transporter

in new onset diabetic BB rats. J Clin Invest 86: 1615–1622.

O’Reilly LA, Gu D, Sarvetnick N, Edlund H, Phillips JM, Fulford T,

Cooke A (1997) a-Cell neogenesis in an animal model of IDDM.

Diabetes 46: 599–606.

Rabinovitch A (1992) Free radicals as mediators of pancreatic islet b-cell injury in autoimmune diabetes. J Lab Clin Med 119: 455–456.

Rabinovitch A (1998) An update on cytokines in the pathogenesis of

insulin-dependent diabetes mellitus. Diabetes Metab Rev 14:

129–151.

Reddy S, Bibby NJ, Elliott RB (1988) Ontogeny of islet cell antibodies,

insulin autoantibodies and insulitis in the non-obese diabetic mouse.

Diabetologia 31: 322–328.

Reddy S, Wu D, Swinney C, Elliott RB (1995) Immunohistochemical

analyses of pancreatic macrophages and CD4 and CD8 T cell subsets

prior toand followingdiabetes in theNODmouse.Pancreas11: 16–25.

Reddy S, Young M, Poole CA, Ross JM (1998) Loss of glucose

transporter-2 precedes insulin loss in the nonobese diabetic and the

low dose streptozotocin mouse models: a comparative immunohis-

tochemical study by light and confocal microscopy. Gen Comp

Endocrinol 111: 9–19.

299Histopathological changes in insulin, glucagon and somatostatin cells

Reddy S, Yip S, Karanam M, Poole CA, Ross JM (1999) An

immunohistochemical study of macrophage influx and the

co-localization of inducible nitric oxide synthase in the pancreas of

non-obese diabetic (NOD) mice during disease acceleration with

cyclophosphamide. Histochem J 31: 303–314.

Reddy S, Karanam M, Poole CA, Ross JM (2000) Dual-label

immunohistochemical study of interleukin-4 and interferon-c-expressing cells within the pancreas of the NOD mouse during

disease acceleration with cyclophosphamide. Autoimmunity 32: 181–

192.

Reddy S, Karanam M, Robinson E (2001a) Prevention of cyclophos-

phamide-induced accelerated diabetes in the NOD mouse by

nicotinamide or a soy protein-based infant formula. Int J Exp

Diabetes Res 1: 299–313.

Reddy S, Young M, Ginn SD (2001b) Immunoexpression of interleu-

kin-1b in pancreatic islets of NOD mice during cyclophosphamide-

accelerated diabetes: co-localization in macrophages and endocrine

cells and its attenuation with oral nicotinamide. Histochem J 33:

317–327.

Reddy S, Bradley J, Ginn S, Pathipathi P, Ross JM (2003) Immuno-

histochemical study of caspase-3-expressing cells within the pancreas

of non-obese diabetic mice during cyclophosphamide-accelerated

diabetes. Histochem Cell Biol 119: 451–461.

Reddy S, Lau EM, Ross JM (2004) Immunohistochemical demon-

stration of leptin in pancreatic islets of non-obese diabetic and CD-1

mice: co-localization in glucagon cells and its attenuation at the

onset of diabetes. J Mol Histol 35: 511–519.

Rothe H, Faust A, Schade U, Kleemann R, Bosse G, Hibino T,

Martin S, Kolb H (1994) Cyclophosphamide treatment of female

non-obese diabetic mice causes enhanced expression of inducible

nitric oxide synthase and interferon-gamma, but not of interleukin-

4. Diabetologia 37: 1154–1158.

Saldeen J (2000) Cytokines induce both necrosis and apoptosis via a

common Bcl-2-inhibitable pathway in rat insulin-producing cells.

Endocrinology 141: 2003–2010.

Shimada A, Charlton B, Taylor-Edwards C, Fathman CG (1996) b-Cell destruction may be a late consequence of the autoimmune

process in nonobese diabetic mice. Diabetes 45: 1063–1067.

Thorens B (1992) Molecular and cellular physiology of GLUT-2 a

high-Km facilitated diffusion glucose transporter. Int Rev Cytol 137:

209–238.

Wang Z, Gleichmann H (1995) Glucose transporter 2 expression:

prevention of streptozotocin-induced reduction in beta cells with 5-

thio-D-glucose. Exp Clin Endocrinol Diabetes, Suppl 2(103): 83–87.

Yasunami R, Bach J-F (1988) Anti-suppressor effect of cyclophos-

phamide on the development of spontaneous diabetes in NOD mice.

Eur J Immunol 18: 481–484.

Zhang ZL, Georgiou HM,Mandel TE (1993) The effect of cyclophospha-

mide treatment on lymphocyte subsets in the nonobese diabetic mouse: a

comparison of various lymphoid organs. Autoimmunity 15: 1–10.

300 S. Reddy et al.