Phosphatidylinositol-4-phosphate-5-kinase alpha deficiency alters dynamics of glucose-stimulated insulin release to improve glucohomeostasis and decrease obesity in mice.

Abstract

Phosphatidylinositol-4-phosphate-5-kinase (PI4P5K) has been proposed to facilitate regulated exocytosis and specifically insulin secretion by generating phosphatidylinositol-4,5-bisphosphate (PIP(2)). We sought to examine the role of the α isoform of PI4P5K in glucohomeostasis and insulin secretion.
The response of PI4P5Kα(-/-) mice to glucose challenge and a type 2-like diabetes-inducing high-fat diet was examined in vivo. Glucose-stimulated responses and PI4P5Kα(-/-) pancreatic islets and β-cells were characterized in culture.
We show that PI4P5Kα(-/-) mice exhibit increased first-phase insulin release and improved glucose clearance, and resist high-fat diet-induced development of type 2-like diabetes and obesity. PI4P5Kα(-/-) pancreatic islets cultured in vitro exhibited decreased numbers of insulin granules docked at the plasma membrane and released less insulin under quiescent conditions, but then secreted similar amounts of insulin on glucose stimulation. Stimulation-dependent PIP(2) depletion occurred on the plasma membrane of the PI4P5Kα(-/-) pancreatic β-cells, accompanied by a near-total loss of cortical F-actin, which was already decreased in the PI4P5Kα(-/-) β-cells under resting conditions.
Our findings suggest that PI4P5Kα plays a complex role in restricting insulin release from pancreatic β-cells through helping to maintain plasma membrane PIP(2) levels and integrity of the actin cytoskeleton under both basal and stimulatory conditions. The increased first-phase glucose-stimulated release of insulin observed on the normal diet may underlie the partial protection against the elevated serum glucose and obesity seen in type 2 diabetes-like model systems.

Full text

Page 1

Phosphatidylinositol-4-Phosphate-5-Kinase a Deficiency
Alters Dynamics of Glucose-Stimulated Insulin Release to
Improve Glucohomeostasis and Decrease Obesity in Mice
Ping Huang,1,2Oladapo Yeku,1,3,4Haihong Zong,2Phyllis Tsang,1,2Wenjuan Su,1,3Xiao Yu,1,2
Shuzhi Teng,6Mary Osisami,1,5Yasunori Kanaho,7Jeffrey E. Pessin,2and Michael A. Frohman1,2,3,5
OBJECTIVE—Phosphatidylinositol-4-phosphate-5-kinase (PI4P5K)
has been proposed to facilitate regulated exocytosis and specifi-
cally insulin secretion by generating phosphatidylinositol-4,5-
bisphosphate (PIP2). We sought to examine the role of the a
isoform of PI4P5K in glucohomeostasis and insulin secretion.
RESEARCH DESIGN AND METHODS—The response of
PI4P5Ka2/2mice to glucose challenge and a type 2-like diabetes-
inducing high-fat diet was examined in vivo. Glucose-stimulated
responses and PI4P5Ka2/2pancreatic islets and b-cells were
characterized in culture.
RESULTS—We show that PI4P5Ka2/2mice exhibit increased
first-phase insulin release and improved glucose clearance, and re-
sist high-fat diet-induced development of type 2-like diabetes and
obesity. PI4P5Ka2/2pancreatic islets cultured in vitro exhibited
decreased numbers of insulin granules docked at the plasma
membrane and released less insulin under quiescent conditions,
but then secreted similar amounts of insulin on glucose stimula-
tion. Stimulation-dependent PIP2depletion occurred on the plasma
membrane of the PI4P5Ka2/2pancreatic b-cells, accompanied by
a near-total loss of cortical F-actin, which was already decreased in
the PI4P5Ka2/2b-cells under resting conditions.
CONCLUSIONS—Ourfindings suggestthatPI4P5Ka playsacom-
plex role in restricting insulin release from pancreatic b-cells
through helping to maintain plasma membrane PIP2levels and
integrity of the actin cytoskeleton under both basal and stimula-
tory conditions. The increased first-phase glucose-stimulated re-
lease of insulin observed on the normal diet may underlie the
partial protection against the elevated serum glucose and obesity
seen in type 2 diabetes-like model systems. Diabetes 60:454–463,
2011
F
metabolized to increase cytosolic ATP levels, which then
promote closure of ATP-sensitive K+(KATP) channels,
causing membrane depolarization. Membrane depolar-
ization triggers opening of L-type Ca2+channels, influx of
Ca2+, and exocytosis. The first phase of exocytosis entails
fusion of primed insulin granules predocked at the plasma
membrane (2). A second phase involving mobilization of
distal insulin vesicles occurs after approximately 10 min
(3), preceded by actin cytoskeletal reorganization (4–6)
and generation of lipid second messengers (7,8). The ma-
jority of type 2 diabetes is only weakly associated with
specific genetic defects and is characterized by inadequate
release of insulin, in addition to insulin resistance exhibi-
ted by fat and muscle target cells. Prolonged stimulation of
b-cells by elevated levels of glucose, such as is encoun-
tered in the typical western-style high-fat diet, eventually
suffices to trigger changes in insulin secretion in many
individuals with otherwise seemingly normal physiology
and genetics.
Lipid kinases and their phosphoinositide products play
important roles in secretory vesicle trafficking (9). Type I
phosphatidylinositol-4-phosphate-5-kinases (PI4P5Ks) a,
b, and g generate the signaling lipid phosphatidylinositol-
4,5-bisphosphate (PIP2). Elegant studies in neurons and
neuroendocrine cells on the role of PI4P5Kg (10,11) and
PIP2 (12–14) have revealed that PIP2 generation at the
plasma membrane is critical during regulated exocytosis.
PIP2 directly facilitates some types of Ca++signaling
(15,16), recruits proteins that facilitate the fusion process
(13,14), and is required for docked secretory vesicles to
undergo priming to become part of the ready-releasable
pool and then fuse into the plasma membrane (10,11).
Although less well understood mechanistically, decreased
levels of PIP2have also been shown to inhibit insulin se-
cretion in pancreatic b-cell model systems (7,17–19).
However, PIP2also carries out functions that potentially
oppose regulated exocytosis (20). First, PIP2supports the
open state of the KATP channel (21,22); thus, because it is
ATP-driven closure of the KATP channel that triggers se-
cretion, PIP2deficiency might be anticipated to decrease
K+efflux, triggering membrane depolarization and thus
increasing Ca++currents, resulting in increased secretion
(23–25). In support of this model, expression of a dominant-
negative isoform of PI4P5K to lower levels of PIP2changes
the responsiveness of mutant KATP channels with decreased
ailure of pancreatic b-cells to release adequate
amounts of insulin contributes to the onset of
type 2 diabetes and obesity (1). Elevated serum
glucose transported into pancreatic b-cells is
From the1Center for Developmental Genetics, Stony Brook University, Stony
Brook, New York; the2Department of Pharmacology, Stony Brook Univer-
sity, Stony Brook, New York; the3Program in Molecular and Cellular Phar-
macology, Stony Brook University, Stony Brook, New York; the4Medical
Scientist Training Program, Stony Brook University, Stony Brook, New
York; the
New York; the6Department of Medical Oncology, Dana-Farber Cancer In-
stitute, Boston, Massachusetts; and the7Department of Physiological Chem-
istry, Graduate School of Comprehensive Human Sciences and Institute of
Basic Medical Sciences, University of Tsukuba, Tsukuba, Japan.
Corresponding author: Michael A. Frohman, michael@pharm.stonybrook.edu.
Received 28 April 2010 and accepted 10 November 2010.
DOI: 10.2337/db10-0614
This article contains Supplementary Data online at http://diabetes.
diabetesjournals.org/lookup/suppl/doi:10.2337/db10-0614/-/DC1.
P.H. and O.Y. contributed equally to this work.
P.H. is currently affiliated with the Department of Pediatrics, Harvard Medical
School, Boston, Massachusetts.
H.Z. and J.E.P. are currently affiliated with the Department of Molecular Phar-
macology, Albert Einstein College of Medicine, Bronx, New York.
X.Y. is currently affiliated with the Department of Biochemistry, Indiana Uni-
versity, Bloomington, Indiana.
? 2011 by the American Diabetes Association. Readers may use this article as
long as the work is properly cited, the use is educational and not for profit,
and the work is not altered. See http://creativecommons.org/licenses/by
-nc-nd/3.0/ for details.
5Program in Genetics, Stony Brook University, Stony Brook,
454 DIABETES, VOL. 60, FEBRUARY 2011diabetes.diabetesjournals.org
ORIGINAL ARTICLE

Page 2

ATP sensitivity. However, expression of the dominant-
negative PI4P5K does not alter function of wild-type (WT)
KATP channels, suggesting that under normal physiologic
conditions, the level of PIP2on the plasma membrane is
not high enough to strongly affect KATP channel activity
(26). PIP2has also been suggested to restrain fusion of
docked vesicles by inhibiting SNARE complex function
(14,27), with the restraint being alleviated through se-
questration of the PIP2by Syntaxin-1 (14) or Ca2+-triggered
PIP2destruction (28).
Taken together, the action of PIP2is complex, and both
its synthesis and its turnover are required at different steps
in the fusion process.
Another function undertaken by PIP2is to promote as-
sembly of actin filaments (F-actin) (29,30). F-actin at the
periphery of the cell (cortical F-actin) has also been pro-
posed to have both positive and negative regulatory
functions in exocytosis. Cortical F-actin has been pro-
posed to act as a barrier to block access of undocked se-
cretory vesicles to the plasma membrane; consistent with
this model, cortical F-actin is disassembled during regu-
latory exocytosis events (31), and pharmacologic agents
that disassemble F-actin enhance movement of insulin
granules to the plasma membrane and insulin release,
whereas agents that stabilize the actin cytoskeleton de-
crease insulin release (4,18,32–34). F-actin may also affect
SNARE complex function by binding to and inhibiting
Syntaxin-4 function in a glucose stimulation-relieved
manner (32,33). To further complicate matters, F-actin can
also undertake a positive role in regulated exocytosis by
mediating translocation of more internal secretory vesicles
to the periphery, particularly in poorly granulated or re-
cently degranulated cells (4,35). Thus, dynamic regulation
of the actin cytoskeleton is also important in the pro-
gression of regulated exocytosis and can play positive or
negative roles depending on the setting.
Changes in PIP2and cortical F-actin can have positive
and negative effects on the secretory process, making it
difficult a priori to predict the outcomes of their physio-
logic and experimental manipulation. Critically, individual
PI4P5K isoforms may generate subpools of PIP2that reg-
ulate distinct components of the fusion process. Deletion of
PI4P5Kg markedly decreases levels of PIP2at the plasma
membrane in resting neurons and neuroendocrine cells,
and inhibits secretion at the stage of vesicle priming and
fusion without notable effect on the actin cytoskeleton
(10,11). In contrast, deletion of PI4P5Ka increases mast
cell degranulation triggered by cross-linking of the IgE re-
ceptor, accompanied by minor decreases in cellular PIP2,
but significantly increased levels of Ca2+signaling and de-
creased levels of total F-actin (36). PI4P5Ka knockdown
using RNAi has also been reported to alter Ca2+signaling,
disrupt F-actin, and affect insulin release in a pancreatic
b-cell line (37).
This report examines insulin secretion in PI4P5Ka2/2
mice and finds that first-phase insulin release is aug-
mented, and on a high-fat diet, fasting and stimulated se-
rum insulin levels are even more elevated, conferring
faster glucose clearance and resistance to the develop-
ment of obesity. In this setting, K+and Ca++signaling is
seemingly normal; in contrast, the determinative factor
seems to be a dramatic stimulation-dependent loss of PIP2
at the plasma membrane that leads to reorganization of the
actin cytoskeleton resulting in a near-total loss of cortical
F-actin and decreased numbers of insulin granules docked
at the plasma membrane.
RESEARCH DESIGN AND METHODS
Animals. PI4P5Ka2/2male mice backcrossed to C57BL/6 were genotyped per
Sasaki and colleagues (36). Animal experiments were performed in accor-
dance with the rules and regulations of the Stony Brook University Animal
Care Committee. For high-fat diet, 12-week-old mice were switched from
normal chow to 45% high fat (Research Diets, Inc., New Brunswick, NJ) for 8
weeks (38).
Intraperitoneal glucose and insulin tolerance tests. Mice fasted for 14 h
were injected intraperitoneally with 1 g/kg 25% D-glucose in saline. Blood
(5–10 mL) was collected from the tail. Blood glucose was determined by
a glucometer and insulin by ELISA kit (Crystal Chem Inc., Downers Grove, IL).
For insulin tolerance tests, mice were fasted for 6 h before being injected
intraperitoneally with insulin (0.85 units/kg of Novolin R; Novo Nordisk,
Bagsværd, Denmark). Blood glucose levels were monitored as above imme-
diately before injection and at 15, 30, 45, 60, and 90 min after injection.
Isolation of pancreatic islets. Pancreatic tissue from 8-week-old male mice
was dissolved in Hank’s balanced salt solution (HBSS) containing Liberase
(Roche 11815032001) and DNase I (Roche 10104159001) at 37°C for 40–50
min; digestion was stopped with HBSS/10% FBS and two washes with HBSS/
2% FBS. The tissue was passed through a cell strainer and applied to a Poly-
sucrose (Cellgrow 61196RO) gradient of 25, 23, 20, and 11% layers to isolate
the islets that were aspirated from the interfaces between the 11 and 20% and
the 20 and 23% gradients, washed, and picked under a dissecting microscope.
Islets were used immediately or cultured in Roswell Park Memorial Institute
medium/10% FBS, Pen/Strep, and glutamate.
Insulin secretion via static incubation. Pancreatic islets (n = 10) were
incubated in 250 mL of Krebs-Ringer bicarbonate HEPES buffer for 30 min at
37°C, and aliquots were taken as basal samples. The glucose was increased to
20 mmol/L, and the cultures were incubated for another 30 min. Samples were
placed on ice or stored at 220°C. To determine total insulin content, the islets
were dissolved in 70% ethanol and 0.18 mol/L HCl overnight, sonicated, and
centrifuged before samples were taken for analysis using an ultrasensitive
murine insulin ELISA (Mercodia: 10115010) and BioRad plate reader. Total
protein concentrations were determined with a BioRad Bradford protein
assay.
Immunostaining, electron microscopy, and morphologic analysis of
islets. Pancreatic islets were fixed in 4% paraformaldehyde for 20 min and
permeabilized in 0.1% Triton X-100 for 20 min. Immunofluorescence staining
for insulin and glucagon was performed with rabbit anti-insulin (Santa Cruz
Biotechnology Inc., Santa Cruz, CA) and murine anti-glucagon (Sigma, St. Louis,
MO), followed by fluorescein isothiocyanate-conjugated anti-rabbit and Alexa
546-conjugated anti-mouse antibodies (Molecular Probes, Eugene, OR). For
actin cytoskeleton staining, fixed and permeabilized islets were stained with
Rhodamine-phalloidin, and the fluorescence intensity was acquired from the
top to the bottom of the islets using continuous xyz-scan mode on the Leica
confocal microscope. All isletswere scanned atthe samegainintensity,and the
three-dimensional series was assembled using confocal imaging software.
For electron microscopy analysis, islets were incubated in Krebs-Ringer
bicarbonate HEPES buffer containing 2.8 mmol/L glucose at 37°C for 30 min
(nonstimulated islets), and some were further incubated in 20 mmol/L glucose
for 30 min (stimulated islets). The islets were fixed in 2% paraformaldehyde/2%
electron microscopy grade glutaraldehyde in 0.1 mol/L phosphate buffer, pH
7.4 for 15 min, postfixed in 2% OsO4 in PBS for 1 h, dehydrated, infiltrated, and
embedded in plastic resin. Ultra-thin sections were stained with uranyl acetate
and lead citrate before examination using a FEI BioTwinG2transmission
electron microscope (FEI Company, Hillsboro, OR). b-cell size and insulin
granules were analyzed using Image J software (National Institutes of Health;
http://rsb.info.nih.gov/ij/).
Statistics. Results are expressed as mean 6 SD unless otherwise stated in
Figs. 1–7. Statistical differences between datasets were assessed using the
unpaired Student t test.
RESULTS
Enhanced insulin secretion and improved glucose
tolerance in PI4P5Ka2/2mice. PI4P5Ka2/2(KO) mice
fed on normal mouse chow for 8–12 weeks were visually
indistinguishable from age- and sex-matched WT littermate
mice but weighed significantly less (26 vs. 28 g, P = 0.03,
n = 16, Fig. 1A). Glucose tolerance tests revealed that the
fasting serum glucose levels were significantly lower in the
KO mice than in the WT mice (86 6 21 mg/dL vs. 126 6 34
mg/dL, respectively) and that the KO mice cleared glucose
faster (Fig. 1B). Serum insulin levels were increased in the
P. HUANG AND ASSOCIATES
diabetes.diabetesjournals.orgDIABETES, VOL. 60, FEBRUARY 2011455

Page 3

FIG. 1. PI4P5Ka KO mice exhibit improved glucohomeostasis on normal murine chow and resist development of type 2-like diabetes responses on
a high-fat diet. A: Body weights of WT and KO mice (n = 16) raised on normal murine chow. Age-matched male mice were used for all experiments. B:
Serum glucose levels for WT and KO mice (n = 9) in fasting mice (0 min time point) and after 1 g/kg body mass intraperitoneal injection of 25%
glucose (intraperitoneal glucose tolerance test) (*P < 0.04). C: Serum insulin levels measured in parallel (*P < 0.05; #P < 0.1). D: Insulin tolerance
test performed on WT and KO mice (n = 5; *P < 0.05). E: Body weights of WT and KO mice (n = 10) after 8 weeks on a high-fat diet. F: Intraperitoneal
glucose tolerance test results on high-fat diet (*P < 0.03; #P < 0.06). G: Serum insulin levels measured in parallel (*P < 0.05, n = 7). Statistical
significance determined by Student t test. Representative of three experiments. ITT, insulin tolerance test; GTT, glucose tolerance test.
PI4P5Ka DEFICIENCY IMPROVES GLUCOHOMEOSTASIS
456 DIABETES, VOL. 60, FEBRUARY 2011diabetes.diabetesjournals.org

Page 4

KO mice at 15–30’ post glucose challenge (Fig. 1C), and
total insulin secretion, calculated as area under the curve,
was significantly increased (P , 0.04) from 67.8 6 5.5 ng/
mL/120 min to 76.2 6 6.3 ng/mL/120 min, suggesting a basis
for the improved glucose clearance. An insulin tolerance
test did not reveal increased peripheral glucose utilization
for PI4P5Ka KO mice in comparison with WT mice (Fig. 1D),
ruling out the possibility of increased insulin sensitivity by
target tissues as a mechanism.
PI4P5Ka KO mice resist developing type 2 diabetes-
like symptoms on a high-fat diet. C57BL/6 mice placed
on high-fat diets develop hyperglycemia and obesity (39).
Age- and sex- matched WT and KO mice generated and
maintained on the C57BL/6 background were placed on
high-fat chow for 8 weeks. The mice exhibited a sub-
stantial gain in weight in comparison with mice on normal
chow (Fig. 1E vs. Fig. 1A). However, the PI4P5Ka KO mice
gained 30% less weight than the WT mice (Fig. 1E). Fasting
serum glucose levels were elevated in both the WT and
PI4P5Ka KO mice, although the levels remained higher in
the WT mice (Fig. 1F). Moreover, whereas WT mice
exhibited prolonged and higher elevation of serum glucose
levels, the PI4P5Ka KO mice exhibited a more rapid return
to euglycemia (Fig. 1F). Fasting and stimulated serum in-
sulin levels were elevated in the WT mice (Fig. 1G).
However, insulin levels in the PI4P5Ka KO mice were even
further elevated under fasting (3.01 6 0.40 vs. 1.59 6 0.44
ng/mL for the WT mice, P , 0.05) and glucose-stimulated
conditions.
By using the hyperglycemic clamp technique, glucose
levels were clamped at $250 mg/dL (Fig. 2A) and insulin
levels were measured at frequent intervals. Elevated
FIG. 2. First-phase GSIR and L-arginine potentiation are increased in absence of PI4P5Ka. A: WT and KO mice received a bolus of 20% glucose to
increase serum glucose levels >250 mg/dL and were then maintained at that level by intravenous infusion adjusted for the change in serum glucose
levels over time. B: Insulin secretion during the hyperglycemic clamp. Blood samples were taken at the intervals shown and assayed for serum
insulin levels. At min 80, a bolus of 20% L-arginine and then an L-arginine infusion were added. Each experiment group was composed of four age-
matched male mice. *Significant P values as indicated. Statistical significance determined by Student t test.
P. HUANG AND ASSOCIATES
diabetes.diabetesjournals.orgDIABETES, VOL. 60, FEBRUARY 2011457

Page 5

insulin secretion (P , 0.01) in KO mice was observed at
5 min after the start of the glucose infusion (Fig. 2B), but
the levels of release were subsequently similar to those of
WT mice. Addition of L-arginine, a potentiator of glucose-
stimulated insulin release (GSIR), resulted in increased
insulin release in both WT and KO mice, but the effect was
substantially greater for the KO mice.
Taken together with the results shown in Fig. 1, these
findings suggest that deletion of the PI4P5Ka gene results
in increased first-phase glucose-dependent insulin secre-
tion that counters the development of the type 2 diabetic-
like phenotype.
Plasma membrane PIP2is not maintained in KO islets
on stimulation. The availability of PIP2at the plasma
membrane plays critical roles in many cell signaling pro-
cesses (40). Total cellular PIP2 levels in cells from
PI4P5Ka KO mice were examined previously (36) and
found to be decreased only modestly (90% of WT levels).
However, PIP2exhibits rapid turnover and resynthesis at
the plasma membrane on glucose-triggered b-cell stimu-
lation as a consequence of Ca2+-mediated phospholipase C
activation (28). The signaling pathway leading to PIP2
turnover can also be initiated by elevating levels of ex-
tracellular K+to depolarize the b-cell, thereby triggering
the same downstream effects (18). We used this approach
to examine whether there were differences in basal or
stimulated levels of accessible PIP2at the plasma membrane
of isolated pancreatic b-cells as reported by a fluorescent
PIP2sensor based on fusion of EGFP to the PIP2-binding
PH domain of PLCd1 (PLCd1-PH-EGFP). Primary pan-
creatic b-cells were transfected in vitro with a reporter
plasmid expressing PLCd1-PH-EGFP, with a transfection
frequency of 80–85% (Supplementary Fig. 1). Under basal
conditions, PIP2was readily imaged at the plasma mem-
brane in both the WT and KO cells (Fig. 3A, rows 1 and 3),
although it should be noted that this observation demon-
strates only that the minimal threshold for sensor re-
cruitment is in place, not necessarily that the levels of
accessible plasma membrane PIP2are quantitatively iden-
tical. However, differences were observed on K+-elicited
depolarization. Whereas no change was observed for WT
b-cells, indicating that the rates of resynthesis of PIP2
compensated for the depolarization-triggered turnover,
relocalization of the sensor from the plasma membrane to
the cytoplasm was observed for the PI4P5Ka KO cells. A
mutant sensor, PLCd1-R40L-PH-EGFP, which lacks the ca-
pability to bind PIP2, distributed uniformly in the cytoplasm
under both basal and stimulated conditions (Fig. 3B). There
were no obvious changes in localization of the insulin
FIG. 3. Plasma membrane PIP2 is not maintained in KO b-cells on de-
polarization. A: b-cells isolated from WT and KO islets were transfected
with a plasmid expressing the PIP2sensor PLCd1-PH-EGFP, cultured
for 2 days, depolarized with extracellular K+for 5 min, fixed, and im-
aged using confocal microscopy. Pancreatic b-cells were identified us-
ing an anti-insulin antibody in combination with a red fluorescent
secondary antibody. B: KO b-cells expressing a mutant PIP2sensor
(PLCd1-PH-R40L-EGFP) that does not exhibit PIP2 binding activity
were processed as in A. Images were taken through the midpoints of the
cells and are representative of the typical transfected, insulin-positive
cells observed in the experiments (n > 100). The experiment was re-
peated at least three times using two animals each time (n = six mice),
and similar outcomes were obtained. Greater than 90% of the cells
examined responded as shown in the images in each condition. (A high-
quality color representation of this figure is available in the online issue.)
FIG. 4. PI4P5Ka KO islets secrete insulin with altered kinetics on
glucose stimulation. WT and KO islets (10 islets per experimental
sample, experiment performed six times) were isolated from age- and
sex-matched mice, preincubated in nonstimulatory culture media (2.8
mmol/L glucose), and challenged with 20 mmol/L glucose. Six mice were
used for each genotype. Incubation times were 30 min in succession for
each condition. Stimulated insulin release is presented after sub-
traction of the basal release. Data are presented as mean 6 SEM.
PI4P5Ka DEFICIENCY IMPROVES GLUCOHOMEOSTASIS
458DIABETES, VOL. 60, FEBRUARY 2011diabetes.diabetesjournals.org