The unique cytoarchitecture of human pancreatic
islets has implications for islet cell function
Over Cabrera*, Dora M. Berman*, Norma S. Kenyon*, Camillo Ricordi*, Per-Olof Berggren*†‡, and Alejandro Caicedo*§
*Diabetes Research Institute, Miller School of Medicine, University of Miami, Miami, FL 33136; and †The Rolf Luft Center for Diabetes Research,
Department of Molecular Medicine, Karolinska Institutet, SE-171 76 Stockholm, Sweden
Communicated by Rolf Luft, Karolinska Institute, Stockholm, Sweden, December 21, 2005 (received for review July 5, 2005)

The cytoarchitecture of human islets has been examined, focusing
on cellular associations that provide the anatomical framework for
paracrine interactions. By using confocal microscopy and multiple
immunofluorescence, we found that, contrary to descriptions of
prototypical islets in textbooks and in the literature, human islets
did not show anatomical subdivisions. Insulin-immunoreactive ␤
cells, glucagon-immunoreactive ␣ cells, and somatostatin-containing ␦ cells were found scattered throughout the human islet.
Human ␤ cells were not clustered, and most (71%) showed associations with other endocrine cells, suggesting unique paracrine
interactions in human islets. Human islets contained proportionally
fewer ␤ cells and more ␣ cells than did mouse islets. In human
islets, most ␤, ␣, and ␦ cells were aligned along blood vessels with
no particular order or arrangement, indicating that islet microcirculation likely does not determine the order of paracrine interactions. We further investigated whether the unique human islet
cytoarchitecture had functional implications. Applying imaging of
cytoplasmic free Ca2ⴙ concentration, [Ca2ⴙ]i, we found that ␤ cell
oscillatory activity was not coordinated throughout the human
islet as it was in mouse islets. Furthermore, human islets responded
with an increase in [Ca2ⴙ]i when lowering the glucose concentration to 1 mM, which can be attributed to the large contribution of
␣ cells to the islet composition. We conclude that the unique
cellular arrangement of human islets has functional implications
for islet cell function.
␣ cell 兩 ␤ cell 兩 cytoplasmic free Ca2⫹ concentration 兩 insulin 兩 glucagon

I

n the last three decades, hundreds of individuals with type 1
diabetes mellitus have received allogeneic transplants of endocrine pancreas, the islets of Langerhans, to cure their chronic
condition. In these patients, diabetes is reversed by transplanting
cells capable of physiologically regulating insulin secretion.
Determining the quality of islets obtained from cadaveric pancreata should be indispensable in this context. However, it is not
known which physiological parameters correlate best with a fully
functional islet capable of reversing diabetes after transplantation. There is a wealth of information about the physiology of
rodent islets, but the biology of human islets remains poorly
understood. As assays for determining islet quality are being
developed by many laboratories in the field of islet transplantation, a reassessment of the structure and function of human
islets is warranted.
The islets of Langerhans are small organs located in the
pancreas that are crucial for glucose homeostasis. Islets typically
consist of four types of secretory endocrine cells, namely, the
insulin-containing ␤ cells, the glucagon-containing ␣ cells, the
somatostatin-containing ␦ cells, and the pancreatic polypeptideproducing (PP) cells. In rodent islets, the vastly predominating
␤ cells are clustered in the core of a generally round islet,
surrounded by a mantle of ␣, ␦, and PP cells. Thus, in rodent
islets, there is a clear segregation of cell types to different regions
of the islet, suggesting that there are anatomical subdivisions in
the islet (1). Although this is the prevailing description found in
textbooks in endocrinology (e.g., refs. 2 and 3), large interspecies
2334 –2339 兩 PNAS 兩 February 14, 2006 兩 vol. 103 兩 no. 7

variabilities in cell composition and structure of the islet have
been reported (4).
Detailed quantitative studies assessing the cell composition of
human islets are sparse (5–7). These studies have reported that
islets are composed of ⬇70% ␤ cells, 20% ␣ cells, ⬍10% ␦ cells,
and ⬍5% PP cells. More recent studies, however, have provided
lower numbers of ␤ cells and higher numbers of ␣ cells (8–10).
More importantly, no studies have addressed how the different
cell types are distributed within human islets and how they relate
to the islet’s microcirculation and islet function. Although mentioned in the literature (1, 11), the notion that human or monkey
islets lack the organization seen in rodent islets has largely been
dismissed.
To address this question, we have examined the cell composition of islets in human pancreatic sections by using confocal
microscopy and multiple immunofluorescence labeling. We enumerated the proportional contribution of the different cell types
to the islet and the associations among endocrine cells and
between endocrine cells and blood vessels. We were particularly
interested in comparing the cytoarchitectural pattern of the
human islet with that of the mouse islet, which has been
considered to be a prototypical islet. In addition, we conducted
imaging studies of cytoplasmic free Ca2⫹ concentration, [Ca2⫹]i,
to establish whether there were functional differences among
mouse, monkey, and human islets. Our results show striking
species differences with regard to both cytoarchitecture and
function and emphasize that human islet structure and function
need to be reevaluated.
Results
As a first step toward a characterization of human islets of
Langerhans, we investigated whether human islets belong to the
prototypical islet type described in many textbooks. We compared the cellular composition of human islets to that of the islets
of three other mammalian species by using confocal microscopy
and multiple immunofluorescence. The presence of cells expressing insulin, glucagon, and somatostatin was detected in
sections of human, monkey, pig, and mouse pancreata (Fig. 1).
In the islets of all species examined, insulin-expressing cells were
the most abundant. However, mouse islets were unique in that
they were mainly composed of insulin-expressing cells clustered
in the core of the islet (Fig. 1C). The few glucagonimmunoreactive cells in mouse islets were localized to the
periphery of the islet (Fig. 1C). In human and monkey islets, by
contrast, glucagon-containing cells were numerous and were
found scattered throughout the islet (Fig. 1 A and B). SomaConflict of interest statement: No conflicts declared.
Abbreviations: [Ca2⫹]i, cytoplasmic free calcium concentration; PP, pancreatic polypeptide.
‡To

whom correspondence may be addressed at: The Rolf Luft Center for Diabetes Research, Department of Molecular Medicine, Karolinska Institutet, Karolinska University
Hospital Solna, SE-171 76 Stockholm, Sweden. E-mail: per-olof.berggren@molmed.ki.se.

§To whom correspondence may be addressed at: Diabetes Research Institute, Miller School

of Medicine, University of Miami, 1450 NW 10th Avenue, Miami, FL 33136. E-mail:
acaicedo@med.miami.edu.
© 2006 by The National Academy of Sciences of the USA

www.pnas.org兾cgi兾doi兾10.1073兾pnas.0510790103

Table 1. Homotypic and heterotypic contacts of cells within
the islet

Cell type

␤ cells
␣ cells

Fig. 1. Islets of Langerhans show striking interspecies differences. (A–D)
Confocal micrographs (1-␮m optical sections), showing representative immunostained pancreatic sections containing islets of Langerhans from human (A),
monkey (B), mouse (C), and pig (D). Insulin-immunoreactive (red), glucagonimmunoreactive (green), and somatostatin-immunoreactive (blue) cells were
all found randomly distributed in human and monkey islets. By contrast,
insulin-containing cells were located in the core, and glucagon- and somatostatin-containing cells in the mantle of mouse islets. Pig islets seemed to be
formed of smaller units (three, in this case) showing a core–mantle organization. (Scale bar, 50 ␮m). (E) Quantitative enumeration of the contribution of
insulin-, glucagon-, and somatostatin-immunoreactive cells to the composition of islets in four different regions of the human pancreas (n ⫽ 5 subjects;
mean ⫾ SEM). (F) Comparison of the cell composition of human islets with that
of mouse islets. Human islets had more glucagon-immunoreactive cells and
fewer insulin-immunoreactive cells (n ⫽ 3 mice and 5 humans; mean ⫾ SEM).

tostatin- and PP-expressing cells constituted a minority in islets
from all species. In mouse islets, but not in those of the other
species, somatostatin-immunoreactive cells were preferentially
located in the periphery (Fig. 1C). Pig islets seemed to be
composed of several smaller subunits that resembled mouse
islets (Fig. 1D). Our results confirmed previous studies on the
Cabrera et al.

Human

Heterotypic,
%

Homotypic,
%

Heterotypic,
%

Homotypic,
%

28 ⫾ 6
98 ⫾ 1

71 ⫾ 4
2⫾1

71 ⫾ 5
90 ⫾ 3

29 ⫾ 5
10 ⫾ 3

composition of mouse islets but pointed to striking interspecies
differences in the structure of islets of Langerhans, in particular
between primate and rodent islets (see also ref. 4).
There are few detailed studies on the cellular composition of
human islets (5–7, 10). Using our approach (see above), we
decided to enumerate the relative contribution of cells containing insulin, glucagon, and somatostatin to the composition of
islets located in four different regions of the human pancreas.
Islets in the four regions examined had a similar composition,
although the neck of the pancreas showed islets with a somewhat
higher proportion of glucagon-immunoreactive cells (Fig. 1E).
Depending on the region, insulin-containing cells contributed
between 48% and 59% of the total cell population, glucagoncontaining cells between 33% and 46%, and somatostatincontaining cells between 8% and 12%. There was variability in
the cell composition between islets within each region; some
islets were mainly composed of insulin-immunoreactive cells
(range, 32–77%) and other islets mainly of glucagonimmunoreactive cells (range, 15–50%). The average cell composition of human islets differed from that of mouse islets (Fig.
1F). The proportion of insulin-containing cells was higher in
mouse than in human islets (77% versus 55%; ANOVA, followed
by Bonferroni multiple comparisons test, P ⬍ 0.05), and the
proportion of glucagon-containing cells was lower (18% versus
38%; P ⬍ 0.05). Because we used the same approach to quantify
the different cell types in both mouse and human islets, we
conclude that our results reflect major differences in the cell
composition between the two species.
To estimate the degree of clustering or segregation of cell
types within the islet, we further enumerated the proportion of
insulin- and glucagon-containing cells exclusively apposed to
cells of the same type (homotypic associations) versus the
proportion of cells closely apposed to cells of other cell types
(heterotypic associations). In mouse islets, 71% of the insulincontaining cells showed exclusively homotypic associations,
whereas, in human islets, only 29% showed these associations
(Table 1; ANOVA, followed by Bonferroni multiple compari-

Fig. 2. Confocal micrographs (1-␮m optical sections) of two series of five
consecutive human pancreatic sections (10 ␮m apart), showing that the
cytoarchitecture did not change within a given islet. Similar proportions of
insulin- (red), glucagon- (green), and somatostatin- (blue) immunoreactive
cells could be seen in all sections. Two islets are shown; the equatorial plane
for A is shown third from the left and, for B, fourth from the left. (Scale bar,
50 ␮m.)
PNAS 兩 February 14, 2006 兩 vol. 103 兩 no. 7 兩 2335

MEDICAL SCIENCES

Mouse

Fig. 3. Endocrine cells in human islets were closely but randomly associated
with vascular cells. Most insulin- (red), glucagon- (green), and somatostatin(cyan) immunoreactive cells were in close proximity to vascular cells immunoreactive for both smooth muscle cell actin (vascular smooth muscle cells, blue)
and CD34 (endothelial cells, blue). Note that many blood vessels were cut
along their main axis, allowing for inspection of long tracts (B–D and F–H).
Endocrine cells were aligned along the blood vessels in a random order;
insulin-immunoreactive cells were bordered on both sides by glucagon- (B–D)
or somatostatin- (F–H) immunoreactive cells and vice versa. Cells from different cell types were seen facing each other across the lumen of the blood vessel.
The image shown in C is a higher magnification of a region in A; the image
shown in H is a higher magnification of a region in E. (Scale bars, 50 ␮m.)

sons test, P ⬍ 0.001). Conversely, 71% of the insulin-containing
cells in human islets showed associations with other cell types. In
both species, ⬎90% of the glucagon-containing cells were in
close proximity to cells of other cell types (P ⬎ 0.05). These
results demonstrate that insulin-containing cells are more clustered in mouse islets than in human islets and that human
insulin-containing cells intermingle with other cell types within
the islet. To rule out the possibility that our results on human
islets were due to biased sampling of islets, we examined serial
sections of the same islet. We found that all sections from the
same islet showed a similar cell composition and cytoarchitecture, independent of whether the sections were from the pole or
equatorial regions of the islet (Fig. 2).
In contrast to the mouse islet, insulin-, glucagon-, and somatostatin-containing cells were found distributed throughout the
human islet. Because endocrine cells are generally associated
with blood vessels, we investigated whether this apparently
random cellular organization reflected the vascularization pattern in human islets. We found that most of the cells, irrespective
of cell type, were located close to islet blood vessels (Fig. 3).
Eighty-six percent of the glucagon-immunoreactive and 77% of
the insulin-immunoreactive cells were closely apposed to immunostained vascular endothelial and smooth muscle cells. The
different cell types were aligned along the vessels with no
particular sequence, i.e., glucagon- or somatostatin-immunoreactive cells were seen on both sides of insulin-immunoreactive
cells and vice versa (Fig. 3). We were not able to determine
whether cells were upstream or downstream of each other. The
random arrangement, however, was independent of whether we
2336 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0510790103

Fig. 4. Human, monkey, and mouse islets showed functional differences that
correlated with structural differences. Different contributions of glucagonimmunoreactive cells were seen for the three species; in human islets (A), the
contribution was ⬇38%, in monkey islets (B), ⬇25%, and, in mouse islets (C),
⬇18%. (Scale bar, 50 ␮m.) [Ca2⫹]i responses (Fura 2-AM) elicited by low glucose
(1 mM, 1), high glucose (11 mM, 11), and high KCl (30 mM, KCl) showed that
human (D) and monkey islets (E), but not mouse islets (F) responded to low
glucose. Bars under the traces indicate the duration of the stimulus. The peak
amplitudes of these responses were quantified for human (G), monkey (H),
and mouse (I) islets (n ⱖ 10 islets).

examined the sequence from one side or the other (Fig. 3).
Furthermore, many insulin-immunoreactive cells were seen facing glucagon- or somatostatin-immunoreactive cells across the
blood vessel lumen (Fig. 3).
To examine whether the morphological differences between
mouse and human islets had functional correlates, we performed
imaging experiments of [Ca2⫹]i in whole islets. Human, monkey,
and mouse islets loaded with the Ca2⫹ indicator Fura-2 showed
an increase in [Ca2⫹]i in response to an elevation of the
extracellular glucose concentration (11 mM; Fig. 4). Furthermore, human and monkey islets responded with an increase in
[Ca2⫹]i when the extracellular glucose concentration was decreased from 3 mM to 1 mM (Fig. 4 D, E, G, and H). These
responses were absent in mouse islets (Fig. 4 F and I). By using
[Ca2⫹]i imaging in dissociated human and monkey islet cells, we
found that reducing the glucose concentration elicited [Ca2⫹]i
transients in a subset (⬇20%) of islet cells that was glucagonimmunoreactive (data not shown) (for activation of ␣ cells at low
glucose concentrations, see also ref. 12). We conclude that
[Ca2⫹]i responses to low glucose levels were detectable in whole
human and monkey islets because of the large contribution of ␣
cells to the islets in these species.
Do the differences in cellular organization have other consequences for islet physiology? It has been widely reported for
rodent islets that the activity of ␤ cells, in particular their
oscillatory behavior, is synchronized throughout the islet, thus
forming a functional syncytium (e.g., refs. 13 and 14). We found
that mouse islets showed oscillations in [Ca2⫹]i in their response
to prolonged glucose elevations (11 mM; Fig. 5A; n ⫽ 6). By
contrast, this type of response could not be elicited in human or
monkey islets (Fig. 5B; n ⱖ 70 islets from ⱖ10 preparations for
human and monkey islets). When smaller regions were examined
within the whole islet, however, oscillatory responses in [Ca2⫹]i
could also be discerned in human or monkey islets (Fig. 5 C and
Cabrera et al.

D). Isolated human or monkey ␤ cells also showed oscillatory
responses to elevated glucose concentrations (Fig. 5E; 48 of 66
cells from four human preparations and 18 of 26 cells from five
monkey preparations). These results suggest that, whereas individual ␤ cells display an oscillatory response, these oscillations
are not synchronized throughout the human or monkey islet and
become neutralized by the out-of-synch activity from other islet
regions. In human and monkey islets, ␤ cell activity seems to be
discontinuous, consistent with our morphological results showing that ␤ cells in human and monkey islets are not clustered as
they are in mouse islets.
Discussion
This study has revealed five findings relevant to the biology of
human islets of Langerhans: (i) human islets contained proportionally fewer ␤ cells and more ␣ cells than mouse islets, the
prototypical islets described in the literature; (ii) ␤ cells intermingled with ␣ and ␦ cells throughout the human islet; (iii) all
endocrine cell types were closely associated with the islet’s
microcirculation; (iv) there was no particular order in the
alignment of endocrine cells along the blood vessels; and (v)
human islets have a distinct physiology that is consistent with the
morphological findings.
The cellular composition of human islets described in our
study is consistent with studies showing that human islets are
composed of ⬇60% ␤ cells and ⬇30% ␣ cells (8–10, 15).
However, islet descriptions in endocrinology textbooks (e.g., ref.
3) reflect results from other studies that have reported a cell
composition similar to that of mouse islets, in which ⬎70% of the
cells are ␤ cells, and ⬍20% are ␣ cells (5, 6). Cell numbers may
inevitably differ between studies because cell composition varies
strongly from islet to islet and from subject to subject (see refs.
5 and 10). In addition, the different immunohistochemical
methods and quantification approaches used in the different
studies may produce discrepant results. Nevertheless, we are
Cabrera et al.

PNAS 兩 February 14, 2006 兩 vol. 103 兩 no. 7 兩 2337

MEDICAL SCIENCES

Fig. 5. [Ca2⫹]i responses of ␤ cells were not synchronized in human islets.
Oscillations in the [Ca2⫹]i response to high glucose (11 mM) could be readily
detected when recording from whole mouse islets (A) (n ⫽ 6 islets). By
contrast, oscillatory responses could not be seen when recording from
whole human islets (B) (representative trace of ⱖ70 islets from 10 preparations). Oscillations in the [Ca2⫹]i response were discernible only in small
areas of the islet that corresponded in size to single cells or small groups of
cells (C). (D) A magnification of C. Isolated human ␤ cells showed oscillatory
[Ca2⫹]i responses to high glucose concentrations (E) (11 mM; n ⫽ 48 of 66
cells from four preparations). The time scale in A also applies to B and C.
Arrows indicate the time point when the bath solution was switched from
3 mM to 11 mM glucose.

confident that our technical approach yielded an accurate
estimate of the cell proportions in the islet. By using confocal
microscopy and multiple immunofluorescence to enumerate the
proportional contribution of the different cell types in one and
the same section, we have avoided the potential inaccuracies of
counting cells in consecutive sections. Our immunohistochemical data are in line with results of recent reports using laser
scanning confocal microscopy on whole isolated islets (10) and
laser scanning cytometry on dispersed islet cells (16). Finally, the
cell composition of mouse islets described in our study is
consistent with that reported in the literature (e.g., refs. 17–19).
The conclusion that human islet composition differs quantitatively from that of mouse islets thus seems inescapable. As
discussed below, the relative contributions of the different
endocrine cells to the human islet, in particular the larger
proportion of ␣ cells, likely have an impact on islet function.
The islets of the mammalian classes examined in this study had
very distinct and unique cytoarchitectures (for a comparative
study, see also ref. 4), suggesting that a prototypical mammalian
islet type may not exist. The wide species variability in islet cell
composition and topography might reflect evolutionary adaptations to different dietary habits or other environmental constraints. Additional comparative studies are needed to address
these hypotheses. Nevertheless, the prevailing view is that there
is a common mammalian scheme in islet architecture. Human
islets have been considered to have a modified core–mantle
structure. In this view, ␤ cells are contiguous and arranged either
into smaller clusters surrounded by non-␤ cells or along invaginations of the exterior surface of the islet (20, 21). Our results,
however, indicate that human islets do not show obvious subdivisions. There was no clustering of ␤ cells, and ␤ cells
intermingled freely with other endocrine cells throughout the
islet. ␤, ␣, and ␦ cells had equivalent and random access to blood
vessels within the islet, ruling out the possibility that the different
endocrine cells are organized in layers around blood vessels.
Our results further indicate that cellular interactions in the islet
may not be determined by the direction of blood flow (22, 23). It
has been proposed, on the one hand, that the blood first flows
through ␤ cell regions and then through non-␤ cell regions. On the
other hand, authors have suggested that non-␤ cell regions are
perfused before ␤ cell regions (for a review, see ref. 24). These
models hold that the direction of blood flow in the islet establishes
a signaling order, where one particular endocrine cell influences the
other endocrine cells. Common to these models is the assumption
that there is specific localization of endocrine cells in the islet (e.g.,
a mantle-core pattern). However, we found no evidence for subdivisions in human islets, and all endocrine cells examined were
randomly associated with blood vessels. Thus, an anatomical basis
for an order in paracrine signaling as determined by the direction
of blood flow likely does not exist for human islets. Our results
support a model in which there is no set order of islet perfusion and
in which any given cell type can influence other cell types, including
its own cell type (24).
The unique cellular topography of human islets had measurable consequences for islet function. Our [Ca2⫹]i recordings
from whole human islets did not show the typical oscillatory
patterns described for rodent islets (13, 14). In rodent islets, ␤
cells are clustered, and their oscillations in membrane potential
and [Ca2⫹]i in response to high glucose concentrations are
coordinated. As a result, the whole islet displays a synchronous
oscillatory response that may constitute the molecular basis for
the typical pulsatility in insulin release. Although we were able
to measure oscillatory [Ca2⫹]i responses in individual ␤ cells or
in distinct regions within the islet, ␤ cell activity in terms of
[Ca2⫹]i oscillations was not synchronized in human islets, suggesting that human ␤ cells are functionally segregated, which is
in line with our anatomical findings showing that they are not
clustered. Interestingly, mice that lack connexin 36 have islets

that do not display regular [Ca2⫹]i oscillations, because ␤ cell
synchronization is lost (25). In these mice, high glucose concentrations elicit [Ca2⫹]i responses in whole islets that resemble
those of human islets. An important functional consequence is
that islets from these mice lose the ability to release insulin in a
pulsatile fashion. How the unique structural and functional
features of human islets affect insulin release remains to be
determined.
Our results also demonstrate the importance of the relative
contributions of the different endocrine cells for islet function.
Species differences in islet cell composition have functional
consequences, as suggested by our results showing that human
and monkey islets, but not mouse islets, respond with [Ca2⫹]i
increases to low glucose concentrations. ␣ cells are believed to
be activated by hypoglycemia, and individual ␣ cells in mouse
islets (26) and isolated human ␣ cells (this study) show [Ca2⫹]i
responses to low glucose concentrations. In mouse islets, ␣ cells
represent a small fraction, and their responses are not likely to
be detected at the level of the whole islet. By contrast, [Ca2⫹]i
responses to low glucose concentrations can be recorded from
whole human and monkey islets because these islets have a large
proportion of ␣ cells, and their individual responses add to give
a measurable signal. Given their large contribution to the islet
and their close association with ␤ cells, ␣ cells may exert a
stronger influence on the overall activity of the human islet than
in rodent islets.
Taken together, our results have provided a picture of
human islets that differs from that of prototypical islets
described in textbooks. Our anatomical and functional data
point to the uniqueness of human islets but also emphasize that
human islet biology is still poorly understood. An arrangement
in which most ␤ cells contact other endocrine cells and vice
versa predisposes human islets for strong paracrine interactions. This same arrangement may not allow ␤ cell oscillatory
activity to be coordinated throughout the islet. How human
islet architecture affects cell-to-cell interactions that lead to
regulated and concerted hormonal secretion remains to be
determined.

with DAPI (Molecular Probes). Slides were mounted with
ProLong Anti Fade (Molecular Probes) and coverslipped. As a
negative control, we substituted the primary antibody with the
serum of the animal used to raise this antibody. No staining was
observed under these conditions. We could never see an overlap
in the staining patterns for the four endocrine hormones,
indicating that there was no crossreactivity between antisera and
that the different immunofluorescent signals could be readily
discriminated.
Data Analysis. Pancreatic sections containing islets were exam-

Immunohistochemistry. Blocks of human, monkey, pig, or mouse

ined for the expression of the different endocrine and vascular
markers by using a Zeiss LSM 510 scanning confocal microscope.
We chose an optical section of 1 ␮m for all image acquisitions.
All images were digitally acquired and not further processed.
Sections were viewed at magnifications ⫻20 and ⫻40.
Islets (2–5 per section) were selected randomly by using the
nuclear staining (DAPI) as a guide in at least three sections per
pancreatic region from five human subjects and three mice.
Cells were counted on digitized images, pooled from all
examined sections in an individual human or mouse, and the
average from all of the pancreata was calculated for each
species. The proportional contribution of a cell type was
calculated as the number of immunoreactive cells for that
endocrine marker divided by the total number of cells immunoreactive for all of the markers. Only cells that had a clearly
labeled nucleus were counted. Because we focused on pancreatic regions with few cells containing pancreatic polypeptide (⬍2%), we did not include these cells in the analyses.
Results are expressed as mean ⫾ SEM.
We further established whether cells contacted cells of other
cell types (heterotypic associations) or cells of the same type
(homotypic associations). To be counted as a contact, the edges
of two cells had to be closer than 3 ␮m on our confocal images.
Cells were included only if they had clearly visible nuclei and a
strongly stained cytoplasm. These criteria were also applied to
evaluate contacts between endocrine and vascular cells. The
number of cells having homotypic associations, heterotypic
associations, or associations with vascular cells was expressed as
a percentage of the total number of that particular cell type.
These values were pooled from all of the sections and averaged
as described above. We further examined whether there was a
particular sequence of cells aligned along blood vessels. We took
advantage of the fact that at least one blood vessel per islet was
cut along its main axis in our sections. Only long stretches of
vessels with a visible lumen spanning a row of at least five cells
were inspected. We examined vessels in 24 sections from three
human subjects.
For statistical comparisons, we used one-way ANOVA, followed by Bonferroni multiple comparisons test.

pancreas (0.5 cm3) were fixed in Bouin’s fixative solution (Sigma)
for 4–6 h, dehydrated in 70% ethanol, and embedded in paraffin.
Sections (5 ␮m) were cut on a microtome, air dried overnight,
deparaffinized, and rehydrated. After a rinse with OptiMax
Wash Buffer (Biogenex, San Ramon, CA), sections were incubated in Universal Blocker Reagent (Biogenex) for 5–10 min,
rinsed again in OptiMax Wash Buffer, and incubated in Protein
Block (Biogenex) for 20 min. Thereafter, sections were incubated overnight with anti-insulin (1:500, Accurate Chemical &
Scientific, Wesbury, NY), anti-glucagon (1: 2,000, Sigma), antisomatostatin (1:500, Serotec), and anti-pancreatic polypeptide
(1:100, Serotec) antibodies. To visualize blood vessels, we incubated the sections in a mixture of anti-CD34 (1:500, Biogenex)
and anti-smooth muscle actin (1: 2,000, Sigma) together with
antisera against two of the endocrine hormones. Immunostaining was visualized by using Alexa Fluor conjugated secondary
antibodies (1:400, Molecular Probes). Cell nuclei were stained

[Ca2ⴙ]i Imaging. Islets from human pancreata were isolated at the
Human Islet Cell Processing Facility of the Cell Transplant
Center at the Miller School of Medicine, University of Miami.
The glands were cold-preserved in University of Wisconsin
solution (27). Islets were isolated by using a modification of the
automated method (28) with seven different lots of the enzyme
Liberase HI (Roche) and a standard purification step, as described in ref. 28. Monkey and mouse islet isolation was performed by using modifications of the automated method for
human islet isolation (28), adapted for isolation of monkey islets
(29) and mouse islets (30). Human islets were dissociated into
single cells by using enzyme-free cell-dissociation buffer (Invitrogen). Islets from all three species were cultured identically
(37°C and 5% CO2) in Connaught Medical Research Laboratories (CMRL) medium 1066 (Invitrogen), niacinamide (10 mM,
Sigma), ITS (BD Biosciences), Zn2SO4 (15 ␮M, Sigma), Gluta

Experimental Procedures
Human pancreata were obtained from multiorgan cadaveric
donors (n ⫽ 5, age, 48 ⫾ 7). Monkey islets were isolated from
cynomolgus monkeys (Macacca fascicularis; n ⫽ 3) ⬎4 years of
age at the time of pancreas procurement. The pig pancreas was
procured from the local slaughterhouse. Mice (C57BL兾6; n ⫽ 5)
were killed by exposure to a rising concentration of CO2,
followed by cervical dislocation. All experimental protocols
using mice were approved by the University of Miami Care and
Use Committee.

2338 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0510790103

Cabrera et al.

Max (Invitrogen), Hepes (25 mM, Sigma), FBS (10%, Invitrogen), and penicillin–streptomycin (Invitrogen).
For [Ca2⫹]i imaging, islets or dispersed islet cells were
immersed in Hepes-buffered solution (125 mM NaCl, 5.9 mM
KCl, 2.56 mM CaCl2, 1 mM MgCl2, 25 mM Hepes, and 0.1%
BSA, pH 7.4). Glucose was added to give a final concentration
of 3 mM. Islets or dispersed islet cells were incubated in Fura
2-AM (2 ␮M for 1 h) and placed in a closed small-volume
imaging chamber (Warner Instruments, Hamden, CT). Stimuli
were applied with the bathing solution. Islets loaded with Fura
2-AM were excited alternately at 340 and 380 nm with a
monochromator light source (Optoscan Monochromator,
Cairn, Faversham, U.K.). Images were acquired with a
Hamamatsu camera attached to a Zeiss Axiovert 200 microscope. Changes in the 340兾380 f luorescence-emission ratio

We thank the members of the Human Cell Processing Facility, Translational
Research Laboratory of the Cell Transplant Center, Clinical Islet Transplant Center, Organ Procurement Organizations, the Imaging Core at the
Diabetes Research Institute, Administrative Offices at the Diabetes Research Institute, and Kevin Johnson for excellent technical assistance. This
work was supported, in part, by National Institutes of Health Grants
M01RR16587 and 1R01-DK55347-IU42RR016603 (for the General Clinical Research Center) and DK-58508, Islet Cell Resources Grant
5U42RR016603, Juvenile Diabetes Research Foundation International
Grant 4-2004-361, the Diabetes Research Institute Foundation, the Swedish
Research Council, the Novo Nordisk Foundation, the Swedish Diabetes
Association, and the Family Stefan Persson Foundation.

1. Orci, L. & Unger, R. H. (1975) Lancet 2, 1243–1244.
2. Nussey, S. S. & Whitehead, S. A. (2001) Endocrinology: An Integrated Approach
(BIOS Scientific, Oxford, U.K.).
3. Greenspan, F. S. & Gardner, D. G. (2001) Basic & Clinical Endocrinology
(McGraw–Hill, New York).
4. Wieczorek, G., Pospischil, A. & Perentes, E. (1998) Exp. Toxicol. Pathol. 50,
151–172.
5. Stefan, Y., Orci, L., Malaisse-Lagae, F., Perrelet, A., Patel, Y. & Unger, R. H.
(1982) Diabetes 31, 694–700.
6. Rahier, J., Goebbels, R. M. & Henquin, J. C. (1983) Diabetologia 24, 366–371.
7. Clark, A., Wells, C. A., Buley, I. D., Cruickshank, J. K., Vanhegan, R. I.,
Matthews, D. R., Cooper, G. J., Holman, R. R. & Turner, R. C. (1988) Diabetes
Res. 9, 151–159.
8. Sakuraba, H., Mizukami, H., Yagihashi, N., Wada, R., Hanyu, C. & Yagihashi,
S. (2002) Diabetologia 45, 85–96.
9. Street, C. N., Lakey, J. R., Shapiro, A. M., Imes, S., Rajotte, R. V., Ryan, E. A.,
Lyon, J. G., Kin, T., Avila, J., Tsujimura, T. & Korbutt, G. S. (2004) Diabetes
53, 3107–3114.
10. Brissova, M., Fowler, M. J., Nicholson, W. E., Chu, A., Hirshberg, B., Harlan,
D. M. & Powers, A. C. (2005) J. Histochem. Cytochem. 53, 1087–1097.
11. Klöppel, G., In’t Veld, P. A., Stamm, B. & Heitz, P. U. (1998) in Functional
Endocrine Pathology, eds. Kovacs, K. & Asa, S. L. (Blackwell Science, Malden,
MA), pp. xiv, 1138, and [8] of plates.
12. Quesada, I., Nadal, A. & Soria, B. (1999) Diabetes 48, 2390–2397.
13. Santos, R. M., Rosario, L. M., Nadal, A., Garcia-Sancho, J., Soria, B. &
Valdeolmillos, M. (1991) Pflügers Arch. 418, 417–422.
14. Valdeolmillos, M., Santos, R. M., Contreras, D., Soria, B. & Rosario, L. M.
(1989) FEBS Lett. 259, 19–23.
15. Orci, L., Baetens, D., Ravazzola, M., Stefan, Y. & Malaisse-Lagae, F. (1976)
Life Sci. 19, 1811–1815.

16. Ichii, H., Inverardi, L., Pileggi, A., Molano, R. D., Cabrera, O., Caicedo, A.,
Messinger, S., Kuroda, Y., Berggren, P.-O. & Ricordi, C. (2005) Am. J.
Transplant 5, 1635–1645.
17. Starich, G. H., Zafirova, M., Jablenska, R., Petkov, P. & Lardinois, C. K. (1991)
Acta Histochem. 90, 93–101.
18. Gomez Dumm, C. L., Console, G. M., Luna, G. C., Dardenne, M. & Goya,
R. G. (1995) Pancreas 11, 396–401.
19. Yukawa, M., Takeuchi, T., Watanabe, T. & Kitamura, S. (1999) Anat. Histol.
Embryol. 28, 13–16.
20. Brelje, T. C., Scharp, D. W. & Sorenson, R. L. (1989) Diabetes 38, 808–814.
21. Weir, G. C. & Bonner-Weir, S. (1990) J. Clin. Invest. 85, 983–987.
22. Bonner-Weir, S. & Orci, L. (1982) Diabetes 31, 883–889.
23. Stagner, J. I., Samols, E., Koerker, D. J. & Goodner, C. J. (1992) Pancreas 7,
26–29.
24. Brunicardi, F. C., Stagner, J., Bonner-Weir, S., Wayland, H., Kleinman, R.,
Livingston, E., Guth, P., Menger, M., McCuskey, R., Intaglietta, M., et al.
(1996) Diabetes 45, 385–392.
25. Ravier, M. A., Guldenagel, M., Charollais, A., Gjinovci, A., Caille, D., Sohl, G.,
Wollheim, C. B., Willecke, K., Henquin, J. C. & Meda, P. (2005) Diabetes 54,
1798–1807.
26. Nadal, A., Quesada, I. & Soria, B. (1999) J. Physiol. 517, 85–93.
27. Kuroda, Y., Kawamura, T., Suzuki, Y., Fujiwara, H., Yamamoto, K. & Saitoh,
Y. (1988) Transplantation 46, 457–460.
28. Ricordi, C., Lacy, P. E., Finke, E. H., Olack, B. J. & Scharp, D. W. (1988)
Diabetes 37, 413–420.
29. Kenyon, N. S., Chatzipetrou, M., Masetti, M., Ranuncoli, A., Oliveira, M.,
Wagner, J. L., Kirk, A. D., Harlan, D. M., Burkly, L. C. & Ricordi, C. (1999)
Proc. Natl. Acad. Sci. USA 96, 8132–8137.
30. Berney, T., Molano, R. D., Cattan, P., Pileggi, A., Vizzardelli, C., Oliver, R.,
Ricordi, C. & Inverardi, L. (2001) Transplantation 71, 125–132.

Cabrera et al.

PNAS 兩 February 14, 2006 兩 vol. 103 兩 no. 7 兩 2339

MEDICAL SCIENCES

over time were analyzed in individual islets and dispersed cells
by using the program AQM ADVANCE (Kinetic Imaging, Richmond, VA). Peak changes in the f luorescence ratio constituted
the response amplitude.