Biochemical and Biophysical Research Communications 492 (2017) 397e403

Contents lists available at ScienceDirect

Biochemical and Biophysical Research Communications

journal homepage: www.elsevier.com/locate/ybbrc

Metformin protects against retinal cell death in diabetic mice

Yoon Sook Kim a, Minjun Kim a, Mee Young Choi a, Dong Hoon Lee a, Gu Seob Roh a,
Hyun Joon Kim a, Sang Soo Kang a, Gyeong Jae Cho a, Seong-Jae Kim b, Ji-Myong Yoo b,
Wan Sung Choi a, *
a
Department of Anatomy and Convergence Medical Science, Institute of Health Science, Gyeongsang National University School of Medicine, Jinju,
Gyeongnam, South Korea
b
Department of Ophthalmology, Gyeongsang National University School of Medicine, Jinju, Gyeongnam, South Korea

a r t i c l e i n f o a b s t r a c t

Article history: Retinal degeneration is an early feature of diabetic retinopathy, the major cause of blindness in the
Received 9 August 2017 developed world. Here we investigated how the widely used antidiabetic drug metformin reduces retinal
Accepted 22 August 2017 injury in diabetic mice. Metformin was orally administered to control mice or mice with streptozotocin-
Available online 23 August 2017
induced diabetes. Western blot analysis showed that levels of O-linked b-N-acetylglucosamine (O-
GlcNAc) transferase (OGT) and other related proteins such as carbohydrate-responsive element-binding
Keywords:
protein (ChREBP) and thioredoxin-interacting protein (TXNIP) were significantly increased, and nuclear
ChREBP
factor kappaB (NF-kB) and poly (ADP-ribose) polymerase (PARP) were activated in the diabetic retinas or
NF-kB
OGT
retinal pigment epithelial (RPE) cells exposed to high glucose compared to controls. More importantly,
PARP RPE cells exposed to high glucose and treated with thiamet-G had higher levels of those proteins,
TXNIP demonstrating the role of elevated O-GlcNAcylation. Double immunofluorescence analysis revealed
increased co-localization of terminal deoxynucleotide transferase-mediated dUTP nick-end labelling
(TUNEL)-positive ganglion cells and OGT, ChREBP, TXNIP, or NF-kB in diabetic retinas compared to control
retinas. Co-immunoprecipitation analysis showed that interaction between OGT and ChREBP or NF-kB
was increased in diabetic retinas compared to control retinas, and this was accompanied by more cell
death. Notably, metformin attenuated the increases in protein levels; reduced co-localization of TUNEL-
positive ganglion cells and OGT, ChREBP, TXNIP, or NF-kB; and reduced interaction between OGT and
ChREBP or NF-kB. Our results indicate that OGT inhibition might be one of the mechanisms by which
metformin decreases retinal cell death.
© 2017 Elsevier Inc. All rights reserved.

1. Introduction repressing endogenous glucose production and enhancing insulin

Diabetic retinopathy (DR) is the most common complication of
diabetes and is a major cause of blindness in the western world [1].
Neurodegenerative changes such as neural apoptosis and ganglion
cell loss occur in the earliest stages of DR [2]. Although hypergly-
cemia is the primary pathogenic factor in DR development [3], the
mechanisms by which hyperglycemia causes retinal injury remain
elusive.
Metformin (N,N-dimethylbiguanide) is a widely used antidia-
betic drug that primarily improves metabolic function by
* Corresponding author. Department of Anatomy and Convergence Medical Sci-
ence, Institute of Health Science, Gyeongsang National University School of Medi-
cine, 15, 816 Beongil, Jinju-daero, Jinju, Gyeongnam 52727, South Korea.
E-mail address: choiws@gnu.ac.kr (W.S. Choi).
sensitivity [4,5]. Metformin activates AMP-activated protein kinase
(AMPK), and its antidiabetic effect is very likely via this mechanism
[6]. Further, metformin protects beta cells by attenuating oxidative
stress-induced apoptosis [7]. Although metformin has been used to
treat hyperglycemia for decades, the exact molecular mechanisms
of its therapeutic action remain unclear. Here, we examined how
metformin protects the ganglion cells of the inner retinal layers
from cell death in DR.
Modification of intracellular proteins with the O-linked mono-
saccharide N-acetylglucosamine (O-Glc-NAc) is involved in
glucose-induced apoptosis [8]. Previous studies showed that
carbohydrate-responsive element-binding protein (ChREBP) is a
transcriptional regulator of glucose metabolism [9], and ChREBP
activity is regulated through multiple post-translational modifica-
tions including O-GlcNAc modification. O-GlcNAc modification

http://dx.doi.org/10.1016/j.bbrc.2017.08.087
0006-291X/© 2017 Elsevier Inc. All rights reserved.
398 Y.S. Kim et al. / Biochemical and Biophysical Research Communications 492 (2017) 397e403
increases ChREBP protein levels and transcriptional activity in the
liver [10]. Notably, O-GlcNAc transferase (OGT) facilitates the
interaction and O-GlcNAc modification of key insulin signaling
regulators [11]. Moreover, thioredoxin-interacting protein (TXNIP),
which is transcriptionally regulated by ChREBP [12], is a crucial
mediator of glucotoxicity-induced b-cell apoptosis [13]. The
mechanism by which glucose activates ChREBP is complex [14], and
its role in DR has not yet been elucidated.
Nuclear factor kappaB (NF-kB) has been implicated in inflam-
matory processes associated with diabetes [15]. The NF-kB family of
transcription factors consists of five members: p105/p50, p100/p52,
p65 (RelA), c-Rel, and RelB [16]. Many compounds with neuro-
protective effects are strongly associated with NF-kB inhibition [17],
and NF-kB is an important regulator of programmed cell death [18].
Of note, elevated O-GlcNAc levels enhance NF-kB signaling through
increasing the binding of p65/RelA to its target promoters [19], and
O-GlcNAc modification of NF-kB is involved in in glucotoxicity [11].
We therefore presumed that O-GlcNAc modification of the p65
could activate NF-kB and affect retinal cell death in diabetic mice.
Of additional note, poly (ADP-ribose) polymerase (PARP) is acti-
vated by hyperglycemia-induced mitochondrial superoxide, which
causes strand breaks in nuclear DNA, leading to PARP activation and
hexosamine pathway flux [20] that may increase O-GlcNAc modi-
fication and diabetic complications [21].
In this study, we assessed whether metformin decreases retinal
neuronal death in diabetic mice, in part driven by post-translational
modification of ChREBP and NF-kB, with the aim of developing
novel agents for protecting against retinal damage in DR.
2. Materials and methods
2.1. Animals
Diabetes was induced in male C57BL/6 mice (KOATEC, Pyeong-
taek, Korea), as previously described [22]. All animal experiments
were carried out in accordance with the National Institutes of
Health Guide for the Care and Use of Laboratory Animals (NIH
Publications No. 8023). Metformin was purchased from Enzo Life
Sciences (ENZO-ALX-270-432, Farmingdale, NY, USA) and orally
given to the mice at 200 mg/kg of body weight. All mice were
sacrificed at 2 months after the final injection of 2-deoxy-2-(3-
(methyl-3-nitrosoureido)-D-glucopyranose (STZ) or saline. Blood
was obtained by tail puncture, and diabetes induction was verified
weekly after STZ injection by evaluating blood glucose concentra-
tions using a Precision glucometer (Abbott Laboratories, Alameda,
CA, USA). Mice with a blood glucose concentration
250 mg/dL
were considered diabetic.
2.2. Cell culture and treatments
The ARPE-19 human retinal pigment epithelial (RPE) cell line
was obtained from American Type Culture Collection (Manassas,
VA, USA), and grown at 37  C in Dulbecco's modified Eagle medium
supplemented with 10% foetal bovine serum (Invitrogen, Carlsbad,
CA, USA), 100 g/mL streptomycin, and 100 units/mL penicillin
(Invitrogen). Cells were treated with low glucose (LG, 5 mM), high
glucose (HG, 25 mM) ± thiamet-G (TMG, 15 mM), HG ± metformin
(MET, 80 mM), or HG plus MET ± TMG.
2.3. Metformin administration
Metformin was administered daily by oral gavage. When con-
ventional antidiabetic doses are used in mice, the equivalent dose is
250 mg/kg of body weight/day [5]. Based on a previous report [20],
we orally administered 200 mg/kg of body weight/day metformin
once a day for 8 weeks after the final STZ or saline injection. Control
and diabetic mice were gavaged daily with saline. Blood glucose
levels and body weights were measured weekly.
2.4. Antibodies
The following antibodies were used: OGT (sc-74546, Santa Cruz
Biotechnology, Santa Cruz, CA, USA; ab96718, Abcam, Cambridge,
UK), ChREBP (NB400-135, Novus), TXNIP (sc-166234, Santa Cruz
Biotechnology), NF-kB p65 (sc-8008, Santa Cruz Biotechnology),
phospho-NF-kB p65 (Ser536) (MA5-15160, Thermo Fisher Scienti-
fic, Waltham, MA, USA), PARP (#9632, Cell Signaling, Danvers, MA,
USA), cleaved PARP (#5625, Cell Signaling), b-Actin (A5441, Sigma,
St. Louis, MO, USA), secondary horseradish-peroxidase-conjugated
goat anti-mouse IgG (#31430, Pierce Biotechnology, Waltham, MA,
USA), and goat anti-rabbit IgG (#31460, Pierce Biotechnology).
2.5. Western blotting
Protein extraction and western blotting were performed as
described previously [22].
2.6. Immunoprecipitation
Immunoprecipitation was performed as described previously
[23].
2.7. Immunohistochemistry analysis
Immunohistochemistry was performed on frozen retinal sec-
tions (5-mm thick), as described previously [24].
2.8. Immunofluorescence analysis
Immunofluorescence analysis was performed as described
previously [23].
2.9. Statistical analysis
Quantitative analyses were performed using ImageJ analysis
software (National Institutes of Health, Bethesda, MD, USA) and
GraphPad Prism 5 (GraphPad Software, San Diego, CA, USA). Data
are representative of three independent experiments and are pre-
sented as the means ± standard error of the mean (SEM). The
statistical significance of differences was determined using one-
way analysis of variance followed by Bonferroni's post hoc anal-
ysis to compare groups. Results were considered significant when P
was less than 0.05.
3. Results
3.1. Metformin lowers blood glucose levels and promotes weight
gain in diabetic mice
Blood glucose levels were significantly increased in diabetic
mice compared to control mice (Fig. 1A, P < 0.0001). However,
metformin treatment significantly lowered the blood glucose levels
of diabetic mice compared to diabetic mice not treated with met-
formin steadily from 1 week to 2 months after diabetes induction,
while control mice remained normoglycemic throughout the study
(Fig. 1A, P < 0.0001). Diabetic mice showed significant weight loss
compared to control mice (Fig. 1B, P < 0.0001), but metformin
moderately increased the body weights of diabetic mice compared
to those not treated with metformin (Fig. 1B, P < 0.05 or P < 0.005).
 Y.S. Kim et al. / Biochemical and Biophysical Research Communications 492 (2017) 397e403
Fig. 1. Metformin lowers blood glucose levels and promotes weight gain in diabetic
mice. Blood glucose levels (A) and body weights (B) of control (CTL, n ¼ 5) and diabetic
mice with (DM þ MET, n ¼ 5) or without (DM, n ¼ 5) metformin treatment.
***P < 0.0001 vs. CTL; yP < 0.05, yyP < 0.005, yyyP < 0.0001 vs. DM.
3.2. Metformin decreases OGT, ChREBP, TXNIP, and phospho-NF-kB
p65 levels in diabetic retinas
Protein O-GlcNAc modification contributes to diabetes mellitus
and neurodegeneration [21]. Notably, OGT facilitates O-GlcNAc
modification of ChREBP, and O-GlcNAcylation increases ChREBP
levels [10], which plays a potential role in glucolipotoxicity and
apoptosis [14]. We therefore assessed whether metformin affects
OGT levels in the diabetic retina. Levels were significantly elevated
in diabetic retinas compared to controls (Fig. 2A, P < 0.0001).
However, metformin treatment significantly lowered OGT levels in
the diabetic retinas, compared to diabetic retinas not treated with
metformin (Fig. 2A, P < 0.0001). To confirm the role of elevated O-
GlcNAcylation, we checked the effect of the O-GlcNAcase (OGA,
which removes O-GlcNAc from target proteins) inhibitor thiamet-G
on OGT in RPE cells exposed to high glucose. Indeed, levels of OGT
in RPE cells exposed to high glucose with thiamet-G were signifi-
cantly increased compared to those without thiamet-G (Fig. 2B,
P < 0.005). However, metformin significantly lowered OGT levels in
RPE cells exposed to high glucose compared to RPE cells not treated
with metformin (Fig. 2C, P < 0.0001), while thiamet-G treatment
reversed these changes (Fig. 2C, P < 0.0001). We next tested
whether metformin affects ChREBP activation in the diabetic ret-
inas. Western blotting showed that ChREBP levels were signifi-
cantly increased in diabetic retinas compared to controls (Fig. 2A,
P < 0.05). However, metformin reversed this increase in ChREBP in
both diabetic retinas (Fig. 2A, P < 0.05) and RPE cells exposed to
high glucose (Fig. 2C, P < 0.0001), and this could be due to O-
GlcNAc modification of ChREBP as demonstrated by thiamet-G
treatment (Fig. 2C, P < 0.0001). Furthermore, pro-apoptotic TXNIP
is induced by diabetes and transcriptionally regulated by ChREBP
[12], but whether metformin affects TXNIP in the diabetic retinas
remains unclear. We found that TXNIP levels were significantly
399
increased in diabetic retinas compared to the controls (Fig. 2A,
P < 0.0001). However, metformin treatment greatly reversed the
diabetes-induced increase in TXNIP (Fig. 2A, P < 0.05). Thiamet-G
treatment of RPE cells exposed to high glucose with or without
metformin demonstrated the role of elevated O-GlcNAcylation,
showing significantly higher TXNIP levels in RPE cells exposed to
high glucose compared to those in high glucose plus metformin not
treated with thiamet-G (Fig. 2B, P < 0.05 or Fig. 2C, P < 0.0001,
respectively).
NF-kB is associated with insulin resistance, obesity, and diabetes
[20]. We therefore tested whether metformin affects NF-kB acti-
vation in the diabetic retina. Indeed, phospho-NF-kB p65 levels
were significantly increased in diabetic retinas compared to con-
trols (Fig. 2A, P < 0.0001). However, metformin treatment greatly
reversed the increase in NF-kB p65 activation (Fig. 2A, P < 0.005).
Moreover, levels of phospho-NF-kB in RPE cells exposed to high
glucose were significantly increased compared to those exposed to
low glucose (Fig. 2B, P < 0.0001). They were further increased in
thiamet-Getreated RPE cells exposed to high glucose (Fig. 2B,
P < 0.0001), showing the role of elevated O-GlcNAcylation in NF-kB
p65 activation in RPE cells exposed to high glucose. However,
opposite effects were observed in metformin-treated cells (Fig. 2C,
P < 0.005). Increased superoxide produced by mitochondria in
response to hyperglycemia activates PARP [20], which may activate
the hexosamine biosynthetic pathway (HBP) [22]. We therefore
tested whether metformin affects PARP activation in the diabetic
retina. As expected, cleaved PARP levels were significantly
increased in the diabetic retina compared to controls (Fig. 2A,
P < 0.005), while metformin treatment greatly attenuated PARP
activation (Fig. 2A, P < 0.005). Cleaved PARP in RPE cells exposed to
high glucose was significantly increased compared those exposed
to low glucose (Fig. 2B, P < 0.0001), and treatment of RPE cells with
thiamet-G further enhanced cleaved PARP levels (Fig. 2B,
P < 0.0001), demonstrating the effect of elevated O-GlcNAcylation
on PARP activation in RPE cells exposed to high glucose. More
importantly, this increase was largely reversed by metformin
(Fig. 2C, P < 0.0001).
3.3. Metformin inhibits OGT interaction with ChREBP in the diabetic
retinas
ChREBP interacts with OGT and is subjected to O-GlcNAcylation,
which stabilizes the ChREBP protein and increases its transcrip-
tional activity toward its target glycolytic genes [10]. We therefore
assessed the interaction between OGT and ChREBP to test whether
metformin affects ChREBP O-GlcNAc modification in diabetic ret-
inas. Immunoprecipitation assays showed that the OGT and ChREBP
interaction was significantly increased in the diabetic retina
compared to control (Fig. 3A, P < 0.005). However, metformin
treatment mostly reversed this change (Fig. 3A, P < 0.05), sug-
gesting that metformin lowered O-GlcNAc-modified ChREBP levels
in the diabetic retina. Consistently, we found that ChREBP and OGT
were mostly co-localized in the ganglion cell layer (GCL) of diabetic
retinas (Fig. 3B). Metformin decreased the extent of co-localization
(Fig. 3B). To test whether metformin affects OGT-related cell death,
we assessed OGT immunoreactivities in TUNEL-positive ganglion
cells of diabetic retinas. We found that TUNEL-positive ganglion
cells were significantly increased in diabetic retinas compared to
controls, but metformin decreased the numbers of dying ganglion
cells (Fig. 3C). Immunofluorescence analysis showed significant co-
localization of TXNIP, which is affected by ChREBP [12], and the
TUNEL signal in the GCL of the diabetic retinas (Fig. 3D), suggesting
its pro-apoptotic action. Again, metformin greatly attenuated this
co-localization (Fig. 3D).
400 Y.S. Kim et al. / Biochemical and Biophysical Research Communications 492 (2017) 397e403

Fig. 2. Metformin lowers levels of OGT, ChREBP, TXNIP, phospho-NF-kB p65, and cleaved PARP in the diabetic retina. (A) Representative western blot and quantification of OGT,
ChREBP, TXNIP, phospho-NF-kB p65, and cleaved PARP in the retinas of control (CTL, n ¼ 5) and diabetic mice with (DM þ MET, n ¼ 5) or without (DM, n ¼ 5) metformin treatment,
2 months after diabetes induction. Band intensity was normalized to that of b-actin. Data are presented as the mean ± SEM (n ¼ 3). *P < 0.05, **P < 0.005, ***P < 0.0001 vs. CTL;
y
P < 0.05, yyP < 0.005, yyyP < 0.0001 vs. DM. (B and C) Representative western blot and quantification of OGT, ChREBP, TXNIP, phospho-NF-kB p65, and cleaved PARP in RPE cells
treated with low glucose (LG, 5 mM), high glucose (HG, 25 mM) ± thiamet-G (TMG), HG ± metformin (MET), or HG plus MET ± TMG. Band intensity was normalized to that of b-
actin. Data are presented as the mean ± SEM (n ¼ 3). **P < 0.005, ***P < 0.0001 vs. LG; yP < 0.05, yyP < 0.005, yyyP < 0.0001, xxP < 0.005, xxxP < 0.0001 vs. HG; #P < 0.05, ###P < 0.0001
vs. HG þ MET.

3.4. Metformin inhibits OGT-NF-kB interaction in diabetic retinas
Because NF-kB p65 activation by O-GlcNAc modification could
be linked to the metabolic syndrome [15] and NF-kB is an impor-
tant regulator of programmed cell death [18], we examined the
interaction between OGT and NF-kB p65 to determine whether
metformin affects O-GlcNAc modification of NF-kB p65 in diabetic
retinas. Indeed, immunoprecipitation assays showed that the OGT
and NF-kB p65 interaction was significantly increased in diabetic
retinas compared to control (Fig. 4A, P < 0.005). More importantly,
metformin mostly reversed this change (Fig. 4A, P < 0.05), sug-
gesting that metformin lowered O-GlcNAc-modified NF-kB p65
levels in diabetic retinas. Consistently, double immunofluorescence
analysis showed that OGT and NF-kB p65 were mostly co-localized
in the GCL of diabetic retinas (Fig. 4B), while metformin treatment
reduced co-localization compared to diabetic retinas not treated
with metformin (Fig. 4B). Notably, we observed significant co-
localization of phospho-NF-kB p65 and the TUNEL signal in the
GCL of the diabetic retinas (Fig. 4C), suggesting that NF-kB con-
tributes to retinal injury in DR. However, metformin treatment
attenuated this co-localization (Fig. 4C), indicating reduced cell
death related to NF-kB activation in the GCL of the diabetic retina.
4. Discussion
Hyperglycemia promotes the rapid and reversible accumulation
of O-GlcNAc [25], and OGT triggers apoptosis in diabetes [26].
Although metformin is commonly used to treat hyperglycemia, the
mechanisms remain unclear. In this study, we demonstrated that
metformin reversed the diabetes-associated OGT increase and
protected the retinas of diabetic mice from neuronal cell death in
the GCL.
Metformin reduced ChREBP levels in diabetic retinas, a protein
that is correlated with insulin sensitivity in early type 2 diabetes
[25] and is known to play a crucial role in apoptosis [14]. Because
ChREBP is involved in metabolic adaptation to changing glucose
 Y.S. Kim et al. / Biochemical and Biophysical Research Communications 492 (2017) 397e403 401

Fig. 3. Effect of metformin on the interaction of OGT with ChREBP in diabetic retinas. (A) Binding of OGT to ChREBP. Representative immunoblots and quantification of ChREBP co-
immunoprecipitated with OGT in the retinas of control or diabetic mice with or without metformin treatment. Tissue lysates were subjected to immunoprecipitation (IP) with an
anti-OGT or -ChREBP antibody and immunoblotted with an anti-ChREBP or -OGT antibody, respectively. The same blots were re-probed with the immunoprecipitating antibody to
confirm equal protein loading. The densitometry signal ratio corresponding to co-immunoprecipitated ChREBP or OGT to that of OGT or ChREBP was normalized to IgG, respectively.
Data are presented as means ± SEM (n ¼ 5e7 per group). **P < 0.005, vs. CTL; yP < 0.05 vs. DM. (B) Representative images of double immunofluorescence staining for OGT and
ChREBP along with DAPI staining for nuclear localization in the retinas of control (CTL) or diabetic mice with (DM þ MET) or without (DM) metformin treatment. The arrowheads
indicate OGT-positive cells that were stained for ChREBP in the retinas of diabetic mice. Scale bar, 50 mm. Representative images of immunofluorescence staining for OGT (C) or
TXNIP (D) plus the TUNEL assay in the retinas of control (CTL, n ¼ 5) or diabetic mice with or without metformin treatment. To confirm the death of cells in conjunction with OGT or
TXNIP expression, immunofluorescent staining of OGT or TXNIP using Alexa Fluor 488 goat anti-rabbit IgG (green) and TUNEL (red) were performed consecutively in the same
sections. Arrowheads indicate TUNEL-positive cells that were stained for OGT or TXNIP in the retinas of diabetic mice. GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner
plexiform layer; ONL, outer nuclear layer; OPL, outer plexiform layer. Scale bar, 50 mm. (For interpretation of the references to colour in this figure legend, the reader is referred to
the web version of this article.)

levels, and alterations in glucose homeostasis are associated with
metabolic diseases such as type 2 diabetes [14], reducing ChREBP is
very intriguing, revealing a novel metformin target. Specifically,
metformin downregulated TXNIP in the diabetic retina, similar to a
recent study where AMPK was found to decrease TXNIP protein
levels [27], adding another target of metformin action in diabetic
mice. Moreover, a previous study showed that increased beta cell
TXNIP expression due to high glucose plays a major role in
pancreatic beta cell loss and diabetes pathogenesis [13]. We found
that TXNIP was increased and co-localized with TUNEL-positive
ganglion cells in the GCL of the diabetic retina, and these changes
were attenuated by metformin. Furthermore, we found that
ChREBP interacts and co-localizes with OGT, as well as with TUNEL-
positive GCLs in the diabetic retinas. However, metformin
treatment reduced the degree of ChREBP-OGT interaction and co-
localization in the GCL of the diabetic retinas, suggesting that
metformin decreases GCL cell death in the diabetic retina by
inhibiting the O-GlcNAc modification of ChREBP, consequently
decreasing TXNIP levels.
Notably, increased intracellular glucose generates increased
reactive oxygen species in the mitochondria, which may activate
PARP, leading to decreased glyceraldehyde-3 phosphate dehydro-
genase activity and ultimate activation of the HBP and glycosylation
by OGT [21,28]. Our finding that metformin treatment decreased
cleaved PARP levels in the diabetic retina or RPE cells exposed to
high glucose suggests metformin's role in the HBP or O-GlcNAc
modification inhibition, as well as in inhibition of the cleavage of
caspase-3 substrates. Moreover, decreased interaction between
402 Y.S. Kim et al. / Biochemical and Biophysical Research Communications 492 (2017) 397e403

Fig. 4. Metformin decreases the interaction between OGT and NF-kB in the diabetic retinas. (A) Binding of OGT to NF-kB. Representative immunoblots and quantification of OGT co-
immunoprecipitated with NF-kB in the retinas of control or diabetic mice with (DM þ MET) or without (DM) metformin treatment. Tissue lysates were subjected to immuno-
precipitation (IP) with an anti-NF-kB antibody and immunoblotted with an anti- OGT antibody. The same blots were re-probed with the immunoprecipitating antibody to confirm
equal protein loading. The densitometry signal ratio corresponding to co-immunoprecipitated OGT to that of NF-kB was normalized to IgG. Data are presented as means ± SEM
(n ¼ 5e7 per group). **P < 0.005 vs. CTL; yP < 0.05 vs. DM. Representative images of double immunofluorescence staining for OGT, NF-kB, and TUNEL (B) or phospho-NF-kB p65 and
TUNEL (C) along with DAPI staining for nuclear localization in the retinas of control (CTL) or diabetic mice with (DM þ MET) or without (DM) metformin treatment. To confirm the
death of cells in conjunction with OGT and NF-kB p65 expression, immunofluorescent staining of OGT or NF-kB p65 using Alexa Fluor 488 goat anti-rabbit IgG (green), NF-kB p65
(violet), and TUNEL (red) were performed consecutively in the same sections. Arrowheads indicate TUNEL-positive cells that were stained for OGT and/or NF-kB p65, or phospho-
NF-kB p65 in the retinas of diabetic mice. GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; ONL, outer nuclear layer; OPL, outer plexiform layer. Scale bar,
50 mm. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

OGT and NF-kB was paralleled by less retinal cell death in the GCL of
metformin-treated diabetic retinas.
In conclusion, our results demonstrate the neuroprotective ac-
tion of metformin in the retinas of diabetic mice. It can inhibit O-
GlcNAc modification to attenuate retinal injury in DR.
Conflict of interest
The authors declare no conflict of interest.
Funding
This work was supported by the Basic Science Research Program
of the National Research Foundation (NRF) of Korea funded by the
Ministry of Science, ICT, and Future Planning (2014049413) and
NRF-2015R1A5A2008833.
References
[1] R.B. Caldwell, M. Bartoli, M.A. Behzadian, A.E. El-Remessy, M. Al-Shabrawey,
D.H. Platt, R.W. Caldwell, Vascular endothelial growth factor and diabetic
retinopathy: pathophysiological mechanisms and treatment perspectives,
Diabetes/metabolism Res. Rev. 19 (2003) 442e455.
[2] H.W. van Dijk, F.D. Verbraak, P.H. Kok, M. Stehouwer, M.K. Garvin, M. Sonka,
J.H. DeVries, R.O. Schlingemann, M.D. Abramoff, Early neurodegeneration in
the retina of type 2 diabetic patients, Invest. Ophthalmol. Vis. Sci. 53 (2012)
2715e2719.
[3] M. Lu, M. Kuroki, S. Amano, M. Tolentino, K. Keough, I. Kim, R. Bucala,
A.P. Adamis, Advanced glycation end products increase retinal vascular
endothelial growth factor expression, J. Clin. Invest. 101 (1998) 1219e1224.
[4] C.J. Bailey, R.C. Turner, Metformin, N. Engl. J. Med. 334 (1996) 574e579.
[5] M. Foretz, B. Guigas, L. Bertrand, M. Pollak, B. Viollet, Metformin: from
mechanisms of action to therapies, Cell metabol. 20 (2014) 953e966.
[6] G. Zhou, R. Myers, Y. Li, Y. Chen, X. Shen, J. Fenyk-Melody, M. Wu, J. Ventre,
T. Doebber, N. Fujii, N. Musi, M.F. Hirshman, L.J. Goodyear, D.E. Moller, Role of
AMP-activated protein kinase in mechanism of metformin action, J. Clin.
Invest. 108 (2001) 1167e1174.
[7] P. Marchetti, S. Del Guerra, L. Marselli, R. Lupi, M. Masini, M. Pollera,
M. Bugliani, U. Boggi, F. Vistoli, F. Mosca, S. Del Prato, Pancreatic islets from
type 2 diabetic patients have functional defects and increased apoptosis that
are ameliorated by metformin, J. Clin. Endocrinol. Metabol. 89 (2004)
5535e5541.
[8] K. Liu, A.J. Paterson, E. Chin, J.E. Kudlow, Glucose stimulates protein modifi-
cation by O-linked GlcNAc in pancreatic beta cells: linkage of O-linked GlcNAc
to beta cell death, Proc. Natl. Acad. Sci. U. S. A. 97 (2000) 2820e2825.
[9] C. Postic, R. Dentin, P.D. Denechaud, J. Girard, ChREBP, a transcriptional
regulator of glucose and lipid metabolism, Annu. Rev. Nutr. 27 (2007)
179e192.
[10] C. Guinez, G. Filhoulaud, F. Rayah-Benhamed, S. Marmier, C. Dubuquoy,
R. Dentin, M. Moldes, A.F. Burnol, X. Yang, T. Lefebvre, J. Girard, C. Postic, O-
GlcNAcylation increases ChREBP protein content and transcriptional activity
 Y.S. Kim et al. / Biochemical and Biophysical Research Communications 492 (2017) 397e403
in the liver, Diabetes 60 (2011) 1399e1413.
[11] J.A. Hanover, M.W. Krause, D.C. Love, Bittersweet memories: linking meta-
bolism to epigenetics through O-GlcNAcylation, Nat. Rev. Mol. Cell Biol. 13
(2012) 312e321.
[12] H. Cha-Molstad, G. Saxena, J. Chen, A. Shalev, Glucose-stimulated expression
of TXNIP is mediated by carbohydrate response element-binding protein,
p300, and histone H4 acetylation in pancreatic beta cells, J. Biol. Chem. 284
(2009) 16898e16905.
[13] J. Chen, G. Saxena, I.N. Mungrue, A.J. Lusis, A. Shalev, Thioredoxin-interacting
protein: a critical link between glucose toxicity and beta-cell apoptosis, Dia-
betes 57 (2008) 938e944.
[14] G. Filhoulaud, S. Guilmeau, R. Dentin, J. Girard, C. Postic, Novel insights into
ChREBP regulation and function, Trends Endocrinol. Metabol. 24 (2013)
257e268.
[15] P. Dandona, A. Aljada, A. Chaudhuri, P. Mohanty, R. Garg, Metabolic syndrome:
a comprehensive perspective based on interactions between obesity, diabetes,
and inflammation, Circulation 111 (2005) 1448e1454.
[16] M.S. Hayden, S. Ghosh, Shared principles in NF-kappaB signaling, Cell 132
(2008) 344e362.
[17] Z.H. Qin, L.Y. Tao, X. Chen, Dual roles of NF-kappaB in cell survival and im-
plications of NF-kappaB inhibitors in neuroprotective therapy, Acta Pharma-
col. Sin. 28 (2007) 1859e1872.
[18] A. Denk, T. Wirth, B. Baumann, NF-kappaB transcription factors: critical reg-
ulators of hematopoiesis and neuronal survival, Cytokine Growth Factor Rev.
11 (2000) 303e320.
[19] Y.R. Yang, D.H. Kim, Y.K. Seo, D. Park, H.J. Jang, S.Y. Choi, Y.H. Lee, G.H. Lee,
K. Nakajima, N. Taniguchi, J.M. Kim, E.J. Choi, H.Y. Moon, I.S. Kim, J.H. Choi,
H. Lee, S.H. Ryu, L. Cocco, P.G. Suh, Elevated O-GlcNAcylation promotes colonic
inflammation and tumorigenesis by modulating NF-kappaB signaling, Onco-
target 6 (2015) 12529e12542.
403
[20] M. Brownlee, The pathobiology of diabetic complications: a unifying mecha-
nism, Diabetes 54 (2005) 1615e1625.
[21] X. Yang, P.P. Ongusaha, P.D. Miles, J.C. Havstad, F. Zhang, W.V. So, J.E. Kudlow,
R.H. Michell, J.M. Olefsky, S.J. Field, R.M. Evans, Phosphoinositide signaling
links O-GlcNAc transferase to insulin resistance, Nature 451 (2008) 964e969.
[22] Y.H. Kim, M.Y. Choi, Y.S. Kim, J.M. Han, J.H. Lee, C.H. Park, S.S. Kang, W.S. Choi,
G.J. Cho, Protein kinase C delta regulates anti-apoptotic alphaB-crystallin in
the retina of type 2 diabetes, Neurobiol. Dis. 28 (2007) 293e303.
[23] Y.S. Kim, M.Y. Choi, D.H. Lee, B.T. Jeon, G.S. Roh, H.J. Kim, S.S. Kang, G.J. Cho,
W.S. Choi, Decreased interaction between FoxO3a and Akt correlates with
seizure-induced neuronal death, Epilepsy Res. 108 (2014) 367e378.
[24] Y.H. Kim, S.Y. Park, J. Park, Y.S. Kim, E.M. Hwang, J.Y. Park, G.S. Roh, H.J. Kim,
S.S. Kang, G.J. Cho, W.S. Choi, Reduction of experimental diabetic vascular
leakage and pericyte apoptosis in mice by delivery of alphaA-crystallin with a
recombinant adenovirus, Diabetologia 55 (2012) 2835e2844.
[25] K. Liu, A.J. Paterson, E. Chin, J.E. Kudlow, Glucose stimulates protein modifi-
cation by O-linked GlcNAc in pancreatic b cells: linkage of O-linked GlcNAc to
b cell death, Proc. Natl. Acad. Sci. 97 (2000) 2820e2825.
[26] S.H. Shin, D.C. Love, J.A. Hanover, Elevated O-GlcNAc-dependent signaling
through inducible mOGT expression selectively triggers apoptosis, Amino
acids 40 (2011) 885e893.
[27] T. Kawaguchi, K. Osatomi, H. Yamashita, T. Kabashima, K. Uyeda, Mechanism
for fatty acid “sparing” effect on glucose-induced transcription: regulation of
carbohydrate-responsive element-binding protein by AMP-activated protein
kinase, J. Biol. Chem. 277 (2002) 3829e3835.
[28] S. Banerjee, V. Sangwan, O. McGinn, R. Chugh, V. Dudeja, S.M. Vickers,
A.K. Saluja, Triptolide-induced cell death in pancreatic cancer is mediated by
O-GlcNAc modification of transcription factor Sp1, J. Biol. Chem. 288 (2013)
33927e33938.