1. RESEARCH ARTICLE
Analysis of mouse skin reveals proteins that are altered
in a diet-induced diabetic state:
A new method for detection of type 2 diabetes
Edward O. List1
, Darlene E. Berryman2
, Amanda J. Palmer1, 2
, Linghua Qiu1
,
Sudha Sankaran1
, Doug T. Kohn1
, Bruce Kelder1
, Shigeru Okada1
and John J. Kopchick1, 3
1
Edison Biotechnology Institute, Ohio University, Athens, OH, USA
2
Human and Consumer Sciences, Ohio University, Athens, OH, USA
3
Department of Biomedical Sciences, College of Osteopathic Medicine, Ohio University, Athens, OH, USA
In this study, proteomic analysis was performed on the skin of C57BL/6J mice with type 2 dia-
betes and compared to nondiabetic controls. To induce obesity and subsequent diabetes, mice
were placed on a high-fat diet for 16 wk. After 16 wk, both diabetic and nondiabetic control mice
were sacrificed and their skin removed for analysis. Following 2-DE, proteomic profiles from the
skin samples were quantified using PDQuest software. Out of more than 1000 distinct protein
spots, 28 were shown to be significantly altered with 6 being decreased and 22 increased in the
diabetic state compared to controls. The 28 protein spots were removed from the gels and ana-
lyzed by MALDI-TOF and MS/MS analyses. Protein identifications revealed that 17 of the 28
proteins were involved in energy metabolism (60.7% of changes observed). Collectively, none of
the significantly altered proteins had been shown previously to be altered in diabetic skin. This
study not only helps to identify proteins found in skin samples of obese mice with type 2 dia-
betes, but also shows that skin biopsies coupled with proteomic analysis may be useful as a
noninvasive method for the diagnosis of hyperinsulinemia and diabetes.
Received: August 21, 2006
Revised: December 1, 2006
Accepted: December 30, 2006
Keywords:
Creatine kinase / Diabetes / Obesity / Skin
1140 Proteomics 2007, 7, 1140–1149
1 Introduction
Obesity rates have increased significantly in the United
States and worldwide. In the US, 64% of adults are now
overweight, and approximately one-half (30% of adults in the
US) of them are estimated to be obese [1]. This represents an
obesity prevalence more than double that observed 40 years
ago. Children and adolescents are not immune to this trend
with rates having more than tripled in the same timeframe
[2]. Thus, it comes as no surprise that obesity has been
described as a true epidemic and public health crisis [3].
One outcome of the obesity epidemic has been a dra-
matic increase in type 2 diabetes, a hyperglycemic condi-
tion characterized by pathological resistance to insulin [4]
and glucose toxicity [4, 5]. According to the Centers for
Disease Control and Prevention (CDC) as of 2002, 18.2
million people, which represents 6.3% of the U.S. popula-
tion, had diabetes and 1.3 million new cases of diabetes are
diagnosed each year [6]. These figures represent only clini-
cally diagnosed diabetes, and many more cases of diabetes
Correspondence: Dr. Edward O. List, Edison Biotechnology Insti-
tute, Ohio University, 101 Konneker Research Laboratories, The
Ridges, Athens, OH 45701
E-mail: edlist@yahoo.com
Fax: 11-740-593-4975
Abbreviations: Apo A-1, apolipoprotein A-1; FFA, free fatty acids;
G3PDH, glyceraldehyde-3-phosphate dehydrogenase; TBP, tri-
butylphosphine
DOI 10.1002/pmic.200600641
© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

2. Proteomics 2007, 7, 1140–1149 Animal Proteomics 1141
(approximately 35%) remain undiagnosed and untreated.
In addition, approximately one-quarter of adults from
western societies have impaired glucose tolerance, repre-
sentative of a prediabetic state [6]. Of concern are the
complications associated with diabetes, e.g. retinopathy,
nephropathy, and neuropathy, that begin to develop well
before the clinical diagnosis of type 2 diabetes [7]. Since an
individual usually progresses from a normal, to obese, to
insulin-resistant/hyperinsulinemic to diabetic condition,
end organ damage is usually well underway by the time
diabetes is diagnosed. Thus, it is of great interest to estab-
lish early markers of obesity-associated diabetes and to use
those markers to allow for preventative measures that can
either slow down, stop, or even reverse the disease pro-
gression.
The C57BL/6J strain of mice has been well documented
to develop obesity and type 2 diabetes when exposed to a
high-fat diet [8–11]. Thus, they represent a useful animal
model to mimic several aspects of diabetes progression as
seen in humans. While other models of obesity-linked dia-
betes exist, such as leptin receptor gene disrupted mice (db/
db) and leptin gene disrupted (ob/ob) mice, the majority of
these models are monogenic, meaning that they are the
result of a single mutated gene. Given that type 2 diabetes in
humans is largely thought to be a polygenic disease in com-
bination with poor lifestyle choices, the use of monogenic
mouse models is limiting.
In order to detect type 2 diabetes at an earlier stage than
conventional methods, alternate forms of diagnosis must be
developed. This study explores the use of proteomics as a
tool for the diagnosis in whole skin samples. Skin tissue
was selected for this study because skin, through the use of
a punch biopsy, is one of the few tissues besides blood that
can be easily obtained from patients. In humans, a punch
biopsy is a quick and relatively easy procedure that is per-
formed in an outpatient environment and requires no spe-
cific surgical skills and minimal surgical equipment [12]. In
this study, several proteins are shown to be altered in the
skin of diabetic mice versus controls. If the same proteins
are found to be altered in the skin of human diabetic indi-
viduals, these proteins may well be of value for the diag-
nosis of diabetes.
2 Materials and methods
2.1 Animals
Male C57BL/6J mice were purchased from Jackson Labora-
tory (Bar Harbor, ME). Obese and type 2 diabetic mice were
generated by feeding 3-wk-old animals a high-fat diet
(#F1850, Bioserve, Frenchtown, NJ) in which 17% of the cal-
ories were provided by protein, 27% by carbohydrates, and
56% by fat. Control mice were placed on a standard rodent
chow diet at 3 wk of age (Prolab RMH 3000, PMI Nutrition
International, St. Louis, MO) in which 26% of the calories
were provided by protein, 60% by carbohydrates, and 14%
by fat.
Mice were housed two per cage in a temperature con-
trolled room (227C) on a 14 h light, 10 h dark cycle. All mice
were allowed ad libitum access to water and food. They were
weighed once every 2 wk and were sacrificed after 16 wk on
the diet by cervical dislocation. These methods are consistent
with the recommendations of the Panel on Euthanasia of the
American Veterinary Medical Association and ensure that
the animals will endure minimal distress and discomfort.
These protocols also have been approved by the Ohio Uni-
versity Institutional Animal Care and Use Committee and
conform to local, state, and federal laws.
2.2 Glucose, insulin, and free fatty acids (FFA)
Fasting blood glucose, plasma insulin, and serum FFA con-
centrations were determined at 2, 4, 8, and 16 wk while the
mice were on the high-fat diet. For all serological measure-
ments, the mice were fasted for 8 h starting at 7 am. After
fasting, blood was obtained by tail bleeding. The first drop of
blood collected was used to determine glucose levels
employing a OneTouch glucometer from Lifescan (Milpitas,
CA). Approximately 100 mL of blood was then collected using
heparinized capillary tubes followed by the collection of
100 mL of blood using nonheparinized tubes. Using the
plasma collected with the heparinized tubes, insulin con-
centrations were determined using the rat insulin ELISA kit
and rat insulin standards (ALPCO, Windham, NH). Values
were adjusted by a factor of 1.23, as recommended by the
manufacturer, to correct for the species difference in cross-
reactivity with the antibody. Using the serum collected in
nonheparinized tubes, FFA levels were determined using the
NEFA-C kit (Wako Chemicals USA, Richmond, VA) as per
the manufacturer’s instructions.
2.3 Tissue collection and processing
At 16 wk, nonfasted mice were sacrificed in the late morning
between 10 am and 12 pm via cervical dislocation. Electric
clippers were used to remove the hair prior to skin removal.
Skin was dissected from the trunk region of the mice by
making a sagittal cut down the center of the ventral side fol-
lowed by a transverse cut around the thoracic region imme-
diately posterior to the front limbs and a transverse cut
around the lumbar region immediately anterior to the hind
limbs. The skin was immediately frozen in liquid nitrogen,
transferred to a freezer, and kept at 2807C until processing.
Skin samples were weighed to determine the amount of
solubilization buffer to be used and then (still frozen) freeze
fracture homogenized using a biopulverizer (Fisher Scien-
tific, Pittsburgh, PA). The pulverized skin was transferred to
a tube containing solubilization buffer (7 M urea; 2 M thio-
urea; 3% CHAPS; 1% SB3-10; 0.1% Biolytes 3–10 (BioRad,
Hercules, CA); 2 mM tributylphosphine (TBP); 1.5% pro-
tease inhibitor cocktail (Sigma, St. Lewis, MO)) which was
© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

3. 1142 E. O. List et al. Proteomics 2007, 7, 1140–1149
made fresh just prior to removing frozen tissues. The ratio of
the solubilization buffer used per skin mass was 4 mL/g tis-
sue. The samples were then homogenized using a mechan-
ical homogenizer followed by a third homogenization via
sonication. The samples were transferred to ultracentrifuge
tubes and centrifuged for 45 min at 150 0006g. Following
centrifugation, the supernatant was collected and aliquotted
into new tubes and stored at 2807C.
2.4 2-DE
Solubilized protein samples were removed from 2807C
storage and protein concentrations were determined using
the Bradford method [13]. One milligram of protein was
added to freshly prepared solubilization buffer to a total
volume of 400 mL followed by the addition of 6 mL of
200 mM TBP and 8 mL of 1 M Tris–HCl, pH–8.8. The sam-
ples were incubated at room temperature for 2 h to allow for
reduction of the proteins. After protein reduction, the pro-
teins were alkylated by the addition of 6 mL of freshly pre-
pared iodoacetamide (160 mg/mL) and incubated at room
temperature for 3 min, a process that was repeated two ad-
ditional times. Following alkylation, the samples were
transferred to individual wells of disposable 17 cm IPG trays
(BioRad) with 17 cm IPG strips pH 3–10 (BioRad). The
trays containing the samples and IPG strips were then
wrapped with plastic wrap and incubated overnight at 207C
to allow for passive rehydration of the strips. Following
rehydration, the IPG strips were removed from the dispos-
able trays, briefly blotted onto filter paper to remove excess
moisture, placed into a focusing tray, and covered with
mineral oil. The focusing tray was then placed into a PRO-
TEAN IEF cell (BioRad) and the proteins were separated via
IEF at 4000 V for 60 000 V?h. Once the IEF was complete,
the IPG strips were removed from the focusing tray, briefly
blotted onto filter paper to remove excess mineral oil and
placed into disposable IPG trays containing 1.5 mL freshly
prepared equilibration buffer (6 M urea; 2% SDS; 375 mM
Tris–HCl, pH 8.8; 20% glycerol). The samples were incu-
bated at room temperature for 25 min. Following equilibra-
tion, 4.5 cm was cut from each end of the 17 cm IPG strips
leaving the center 8 cm, which was then placed on top of
15% polyacrylamide gel containing 4% stacking gels for the
second dimension. The proteins were separated via SDS-
PAGE at 25 mA per gel for 250 V?h.
2.5 Imaging/protein identification
Following SDS-PAGE, the gels were incubated with a
SYPRO Orange fluorescent stain (Molecular Probes, Eugene,
OR) using a modified protocol by Malone et al. [14]. SYPRO
Orange-stained gels were imaged using a VersaDoc 1000
Imaging System (BioRad). Spot detection and densitometry
were performed using the Discovery Series PDQuest 2-DE
analysis software package version 7.0 that accompanied the
VersaDoc 1000 Imaging System. Proteins that were differ-
entially expressed were removed from the polyacrylamide gel
and sent to the proteomics facility at the University of
Michigan for identification by MALDI-TOFMS.
2.6 Concentration and spotting of gel digest
extracts (performed at the Michigan Proteome
Consortium)
For MS and MS/MS analysis, 5 mL of CHCA (5 mg/mL in
50% ACN, 0.1% TFA, 2 mM ammonium citrate) matrix was
added to 30 mL of digest extract for each well of the extraction
plate. The samples were taken to dryness and 5 mL of 50%
ACN/0.1% TFA was added back into the extraction well. This
solution (0.5 mL) was hand-spotted on a 192-well MALDI tar-
get and allowed to dry in atmosphere.
2.7 MS analysis (performed at the Michigan
Proteome Consortium)
Mass spectra were acquired on an Applied Biosystems 4800
Proteomics Analyzer (TOF/TOF). MS spectra were acquired
in Reflector Positive Ion mode. Peptide masses were
acquired for the range from 800 to 3500 Da. MS spectra were
summed from 2000 laser shots from an Nd-YAG laser oper-
ating at 355 nm and 200 Hz. Internal calibration was per-
formed using a minimum of three trypsin autolysis peaks.
2.8 Manual protein database searching with
MS-generated peak lists (performed at Ohio
University)
MS peaks were submitted to Matrix Science website (http://
www.matrixscience.com) where MASCOT peptide mass fin-
gerprint searches were performed using the following search
parameters (database: NCBInr; taxonomy: Mus musculus;
enzyme: trypsin; missed cleavage: 1; fixed modifications:
none specified “default”; variable modifications: none speci-
fied “default”; protein mass: not specified “default”; peptide
tolerance: 6 1 Da “default”; mass value: MH1 “default”;
monoisotopic: selected “default”).
2.9 MS/MS analysis (performed at the Michigan
Proteome Consortium)
MS/MS spectra were acquired in MS/MS 2 kV positive
mode. Spectra were acquired for 6000 laser shots or until five
peptide fragment ions reached an S/N of 100, whichever was
less. Fragmentation of the peptides was induced by the use of
atmosphere as a collision gas with a pressure of ,6610–7
Torr and a collision energy of 2 kV. Database searching was
performed using Applied Biosystems GPS Explorer v.3.6,
with MASCOT v.2.1. Spectra were subjected to a 7-point
Gaussian smooth prior to peak picking. For the automated
MS/MS database search, peaks from a minimum mass of 20
to 60 Da below the precursor mass were used. A maximum
of 65 peaks were submitted for the search with a minimum
© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

4. Proteomics 2007, 7, 1140–1149 Animal Proteomics 1143
S/N of 10, and a maximum peak density of 50 peaks per
200 Da.
2.10 Manual protein database searching with MS/MS
peaks (performed at Ohio University)
MS/MS peaks were submitted to Matrix Science website
(http://www.matrixscience.com) where MASCOT MS/MS
ion searches were performed using the following search pa-
rameters: (database: NCBInr; taxonomy: Mus musculus; en-
zyme: trypsin; missed cleavage: 1; fixed modifications: none
specified “default”; variable modifications: none specified
“default”; protein mass: never specified “default”; ICAT: not
selected “default”; peptide tolerance 6 2.0 Da “default”; MS/
MS tolerance 6 0.8 Da “default”; data format: MASCOT
generic; monoisotopic: selected “default”; precursor m/z: not
specified; instrument: default).
2.11 Statistical analysis
Mouse physiological data were presented as mean 6 SEM.
Physiological data were analyzed by single factor ANOVA
(Microsoft Excel), while protein expression data were ana-
lyzed by Student’s t-test. Differences were considered statis-
tically significant if p,0.05. For MS/MS MASCOT peptide
mass fingerprint searches, probability-based MOWSE scores
were used. Since these scores vary for each fragment, the
MOWSE score needed for significance (p,0.05) needed to be
greater than a MOWSE score significance level that was
reported for each fragment. Thus, if the score was more than
the indicated score, then p,0.05. A minimum of two signif-
icant MS/MS peptide fragments was considered sufficient to
assign an identification for a spot.
3 Results
High fat feeding in C57BL/6J mice resulted in obesity as
evidenced by weight gain. Weight increased continuously
throughout the study as compared to control mice fed
standard chow, resulting in obese mice that weighed ap-
proximately 50% more by the end of the study (Fig. 1A). The
difference in weight became statistically significant
(p,0.001) at the 4 wk time-point and continued throughout
the duration of the 16 wk diet study. These results were
similar to those reported previously [11].
The increased weight of high-fat fed mice was accom-
panied by an increase in blood glucose (Fig. 1B), serum FFA
(Fig. 1C), and plasma insulin (Fig. 1D) levels. Moreover, at
the time of sacrifice and tissue collection, all of the physio-
logical measurements tested (blood glucose, FFA, and insu-
lin) were significantly higher in high-fat fed mice as com-
pared to nondiabetic control mice. This indicated that these
high-fat fed mice were suffering from a condition compara-
ble to type 2 diabetes in humans.
Figure 1. Progression of obesity and type 2 diabetes in C57BL/6J
mice fed a high-fat diet. (A–D) Solid lines represent control mice
fed a standard chow diet (n = 15), while the dashed lines repre-
sent mice fed a high-fat diet (n = 43). All measurements were
taken at 2, 4, 8, and 16 wk on the diet except for body weight,
which was measured every 2 wk. Since the diet was initiated at
weaning, which occurred at 21 days of age and represents the
zero time-point, the 2, 4, 8, and 16 wk time-points actually repre-
sent 5, 7, 11, and 19 wk of age, respectively. (A) Body weights of
mice fed a high-fat diet became significantly increased after 4 wk
on the diet as compared to controls and remained significant
throughout the 16 wk duration of the study. (B) Fasting blood
glucose was significantly higher in mice on the high-fat diet at all
time-points measured with maximal hyperglycemia being
achieved at the final 16 wk time-point. (C) Fasting serum FFA and
(D) fasting plasma insulin levels increased over time with signifi-
cant differences at both the 8 and 16 wk time-points. Error bars
represent the SEM. Statistical analysis was performed using
ANOVA. *p,0.05 and **p,0.001.
In order to identify proteins that are altered in the skin of
mice with type 2 diabetes, whole skin samples were har-
vested from diabetic (n = 4) and nondiabetic (n = 4) control
mice. Physiological measurements of these 8 mice are
shown in Fig. 2. Solubilized proteins from all eight mouse
skin samples were separated in triplicate using 2-DE result-
ing in a total of 24 gels used for imaging, quantification, and
spot selection. Using a VersaDoc 1000 imaging system to
capture the gel images, and the accompanying PDQuest
2-DE analysis software version 7.0.0, more than 1000 distinct
protein “spots” were consistently detected from the whole
skin (Figs. 3, 4, and 5). The 2-D gels were extremely repro-
ducible, with a similar pattern and number of spots even
when comparing gels from animal to animal (Fig. 3). From
the detected protein spots, 28 were shown to be significantly
altered (Figs. 4 and 5) with 6 being decreased (Table 1) and 22
being increased (Table 2) in the diabetic state compared to
controls. In order to identify the altered proteins, the 28 dif-
ferential protein spots were removed from the gels and ana-
lyzed by both MALDI-TOFand MS/MS analyses (Supporting
Information). MS/MS peak data were used to determine the
© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

5. 1144 E. O. List et al. Proteomics 2007, 7, 1140–1149
Figure 2. Physiological measurements in the C57BL/6J mice
used for tissue collection. (A–D) Gray bars represent control mice
fed a standard chow diet starting at 3 wk of age for 16 wk (n = 4),
while the striped bars represent mice fed a high-fat diet starting
at 3 wk of age for 16 wk (n = 4). All measurements were taken
1 day prior to tissue collection, which occurred after 16 wk on the
diet and at 19 wk of age. (A) Body weights of mice fed a high-fat
diet were significantly different as compared to controls. Fasting
levels of blood glucose (B), serum FFA (C), and plasma insulin (D)
were significantly higher in mice on the high-fat diet as com-
pared to controls. Error bars represent the SEMs. Statistical
analysis was performed using ANOVA. *p,0.05, **p,0.01, and
***p,0.005.
Figure 3. Animal-to-animal reproducibility of skin samples
resolved by 2-DE. The eight independent gels labeled with letters
A–H represent skin proteins isolated from eight individual ani-
mals that were sacrificed at 19 wk of age after 16 wk on a stand-
ard chow (A–D) or high-fat diet (E–H). Gels A–D on the left side of
the figure are from four separate nondiabetic control mice, while
gels E–H on the right side of the figure are from four separate
diabetic mice.
identities of 26 of the protein spots, while two of the protein
spots were only weakly identified. Of the six protein spots
shown to decrease in the diabetic state (Figs. 4 and 5), two
separate spots were determined to be creatine kinase chain
M and the remaining spots were identified as aldolase 1,
transferrin, calpactin I light chain, and a protein similar to
glyceraldehyde-3-phosphate dehydrogenase (G3PDH). The
percent decrease with high fat diabetic (HF-D) samples as
compared to control samples, of the two creatine kinase
spots was 90 and 56% for spots B5 and G12, respectively. Of
the four other decreased proteins, the percent decrease in
order of largest decrease to smallest was 89% for G3PDH,
84% for aldolase 1, 66% for calpactin I light chain, and 51%
for transferrin.
The 22 protein spots identified as increased in diabetic
skin (Figs. 4 and 5) are listed below in the order of the largest
increase to the smallest increase in HF-D/control: malate
dehydrogenase (EC 1.1.1.37) 1262%, protein disulfide-isom-
erase 1190%, creatine kinase chain M (spot B6) 837%, type II
Keratin Kb39 808%, creatine kinase chain M (spot B3) 483%,
peroxiredoxin 6 415%, creatine kinase chain M (spot B4)
413%, fatty acid-binding protein (spot F6) 403%, apolipo-
protein A-1 Precursor 385%, nucleoside-diphosphate-kinase
2 350%, fatty acid-binding protein (spot F4) 276%, fatty acid-
binding protein (spot F5) 267%, keratin 2 – type I hair 226%,
phosphoglycerate mutase-1 208%, proteasome (prosome,
macropain) subunit alpha type 1 195%, 14-3-3 protein beta
164%, prohibitin 156%, calpactin I light chain 153%, lectin –
galactose-binding soluble-7 120%, apolipoprotein E Pre-
cursor 83%, nucleoside-diphosphate kinase 64%, vacuolar
protein sorting-29 63%.
4 Discussion
It is widely known that skin is the largest organ of the body
providing a physical, chemical and biological barrier be-
tween an organism and the outside environment. Skin also
performs a multitude of other important physiological
functions ranging from thermoregulation to immune sys-
tem surveillance. Structurally, skin is a complex tissue being
formed from both mesoderm and ectoderm and in its
mature state is a tissue that contains glands and muscle, as
well as fat, and is highly innervated with nerves and vascu-
lature.
In the diabetic state, microvascular complications play a
major role in the pathophysiology of this disease ultimately
resulting in end organ damage such as blindness and
nephropathy. In addition to eyes and kidneys, another vas-
cularized organ that is highly susceptible to diabetic damage
is skin. Reported cases of cutaneous manifestations in
patients suffering from diabetes ranges from 30 to 71% [15–
17]. Since skin is an organ that is known to be affected by
diabetes, and next to blood, is one of the easiest tissues to
biopsy, the idea of using a skin biopsy to screen for diabetes
was a logical choice. The use of skin biopsy in the diagnosis
of diabetic neurological damage has already been shown to
be useful at the histological level [18]. In one study [18],
© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

6. Proteomics 2007, 7, 1140–1149 Animal Proteomics 1145
Figure 4. 2-D gel from normal mouse
skin. This SYPRO Orange-stained gel
contains proteins isolated from normal
mouse skin at 19 wk of age following
16 wk on a standard chow diet. The gel
was imaged using a VersaDoc 1000 Im-
aging System. The approximate pI and
molecular weights are labeled along the
top and left-hand side borders of the gel,
respectively. Spot detection and densi-
tometry were performed using the Dis-
covery Series PDQuest 2-DE analysis
software package version 7.0 that
accompanied the VersaDoc 1000 Imag-
ing System. Spots labeled in red repre-
sent proteins that were increased in the
diabetic state while spots labeled in
green represent proteins that were
decreased in the diabetic state as com-
pared to control skin samples. All
labeled spots were removed from the
polyacrylamide gel and analyzed by
both MALDI-TOF and MS/MS MS at the
Michigan Proteome Consortium.
Table 1. Proteins found to be decreased in the skin of high-fat fed diabetic mice following 16 wk of feeding starting
at 3 wk of age as compared to protein level found in nondiabetic control mice fed standard rodent chow
for 16 wk starting at 3 wk of age
Spot Proteina)
MASCOT MS/MS
accession no.
% Decrease
(HF/C)b,c)
B5 Creatine kinase (EC 2.7.3.2) chain M A23590 90%
B10 Aldolase 1, A isoform gi)42490830 84%
C7 Similar to G3PDH gi)6679937 89%
C12 Transferrin gi)17046471 51%
G12 Creatine kinase (EC 2.7.3.2) chain M A23590 56%
H8d)
Calpactin I light chain (protein S100-A10) S10AA_MOUSE 66%
a) A minimum of two significant MS/MS peptide fragments was considered sufficient to assign an ID for a spot.
More detailed information is provided as Supporting Information.
b) The mean decreases in spot intensities were determined using PDQuest 7.0.0 software on 24 gels from 8 mice.
c) Only differences that were statistically significant p,0.05 as determined by Student’s t-test as part of the
PDQuest software are reported.
d) Indicates a weak identification by matching one significant fragment with at least one fragment that
approached significance (p = 0.0013 and 0.056).
© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

7. 1146 E. O. List et al. Proteomics 2007, 7, 1140–1149
Table 2. Proteins found to be increased in the skin of high-fat fed diabetic mice following 16 wk of feeding starting
at 3 wk of age as compared to protein level found in nondiabetic control mice fed standard rodent chow
for 16 wk starting at 3 wk of age
Spot Proteina)
MASCOT MS/MS
accession no.
% Increase
(HF/C)b,c)
A12d)
Calpactin I light chain (protein S100-A10) S10AA_MOUSE 153%
B3 Creatine kinase (EC 2.7.3.2) chain M A23590 483%
B4 Creatine kinase (EC 2.7.3.2) chain M A23590 413%
B6 Creatine kinase (EC 2.7.3.2) chain M A23590 837%
C2 14-3-3 Protein beta gi)3065925 164%
C6 Peroxiredoxin 6 gi)6671549 415%
C8 Phosphoglycerate mutase 1 gi)10179944 208%
C11 Nucleoside-diphosphate kinase 2 gi)6679078 350%
D1 Keratin complex 1, acidic gene 4 gi)13386238 226%
D2 Apolipoprotein A-1 precursor gi)109571 385%
D8 Lectin, galactose binding, soluble 7 gi)31543120 120%
D9 Nucleoside-diphosphate kinase 1 gi)37700232 63%
E1 Vacuolar protein sorting 29 gi)9790285 64%
E4 Apolipoprotein E precursor gi)114041 83%
E5 Prohibitin gi)6679299 156%
E7 Malate dehydrogenase, cytoplasmic gi)92087001 1262%
E12 Keratin 1b gi)38565071 808%
F1 Protein disulfide isomerase associated 3 gi)112293264 1190%
F2 Proteasome (prosome, macropain) subunit,
alpha type 1
gi)33563282 195%
F4 Fatty acid-binding protein PC4011 276%
F5 Fatty acid-binding protein, adipocyte FABPA_MOUSE 267%
F6 Fatty acid-binding protein, adipocyte FABPA_MOUSE 403%
a) A minimum of two significant MS/MS peptide fragments was considered sufficient to assign an ID for a spot.
More detailed information is provided as Supporting Information.
b) The mean decreases in spot intensities were determined using PDQuest 7.0.0 software on 24 gels from 8 mice.
c) Only differences that were statistically significant p,0.05 as determined by Student’s t-test as part of the
PDQuest software are reported.
d) Indicates a weak identification by matching one significant fragment with at least one fragment that
approached significance (p = 0.014 and 0.089).
Lauria et al. used a 3 mm diameter punch device to biopsy
skin from trunk and chest. Through histology, they found a
reduction in both epidermal and dermal nerve fibers in
affected areas in patients with diabetes. Although not deter-
mined, these histological changes are accompanied by
changes in protein expression, which may be detectable by
conventional proteomics methodology.
In the current study, proteomic analysis was performed
on skin from mice fed a high-fat diet and subsequently
developed diabetes. Proteomic analysis revealed 28 altered
protein spots. Database searches of MS/MS generated peak
lists obtained from these spots revealed that most of the
changes observed were to proteins that are involved in
energy metabolism (17 proteins or 60.7% of changes
observed). This finding is not surprising since type 2 diabetes
is a disease that is ultimately defined by insulin resistance
and elevated glucose; thus, individuals who are afflicted with
this disease are likely to have altered energy metabolism. The
other categories include metal-binding proteins (3 or 10.7%),
proteins involved in protein processing (3 or 10.7%), struc-
tural proteins (2 or 7.1%), proteins involved in signal trans-
duction (2 or 7.1%), and finally, oxidative stress proteins (1 or
3.6%). While some of these proteins have been shown to be
altered in a diabetic state, none of these proteins have been
reported previously to be altered in diabetic skin.
The 17 proteins involved in energy status can be further
divided into three categories; ATP converting enzymes
(seven protein spots or 25% of changes), proteins involved in
fat metabolism (five protein spots or 17.9% of changes), and
proteins involved in carbohydrate metabolism (five protein
spots or 17.9% of changes). For the ATP converting enzymes,
five forms of creatine kinase, which is discussed in more
detail later in the discussion, and two forms of nucleoside-
diphosphate kinase, which are involved in the conversion of
nucleoside diphosphates to nucleoside triphosphates were
increased in the diabetic mouse model. The five changes in
proteins involved in fat metabolism were to apolipoprotein E,
apolipoprotein A-1 (Apo A-1) precursor, and 3 forms of fatty
© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

8. Proteomics 2007, 7, 1140–1149 Animal Proteomics 1147
Figure 5. 2-D gel from diabetic mouse
skin. This SYPRO Orange-stained gel
contains proteins isolated from the skin
of diabetic mice at 19 wk of age follow-
ing 16 wk on a high-fat diet. The gel was
imaged using a VersaDoc 1000 Imaging
System. The approximate pI and molec-
ular weights are labeled along the top
and left-hand side borders of the gel,
respectively. Spot detection and densi-
tometry were performed using the Dis-
covery Series PDQuest 2-DE analysis
software package version 7.0 that
accompanied the VersaDoc 1000 Imag-
ing System. Spots labeled in red repre-
sent proteins that were increased in the
diabetic state while spots labeled in
green represent proteins that were
decreased in the diabetic state as com-
pared to control skin samples. All
labeled spots were removed from the
polyacrylamide gel and analyzed by
both MALDI-TOF and MS/MS MS at the
Michigan Proteome Consortium.
acid-binding protein, all of which were increased in the high-
fat fed mice. Although these changes could be due to the
increased dietary fat or subsequent development of obesity in
these mice, hyperinsulinemia has been shown to increase
Apo A-1 by activating the Apo A-1 promoter [19], suggesting
that the diabetic state could also play an important role.
The five changes observed in proteins involved in carbo-
hydrate metabolism included two decreased proteins and
three increased proteins. The two proteins that were found to
be decreased were aldolase 1 A isoform, and a protein called
“similar to G3PDH”, which contains exactly the same 226
amino acids as G3PDH, but contains additional 23 and 84
amino acid sequences on the N and C portions of the protein,
respectively. Both of these proteins play key roles in glycol-
ysis, and have been shown to be decreased in a diabetic state:
aldolase in the pancreas [20, 21] and G3PDH in the heart and
whole embryos [22, 23]. The three proteins involved in car-
bohydrate metabolism that were increased in the diabetic
mouse model were phospoglycerate mutase, lectin galactose-
binding soluble 7 (also called galectin-7), and malate dehy-
drogenase. While no previous studies report a connection
between phospoglycerate mutase levels and diabetes, this
protein has been shown to increase with aging in dermal
fibroblasts cultured in vitro from human subjects [24] and it
has been suggested that diabetes resembles accelerated
aging [25, 26]. Galectin-7 also has not been shown to be di-
rectly linked to diabetes; however, other members of this
protein family, like galectin-3, have been shown to be
increased in diabetes [27]. In contrast to this study, malate
dehydrogenase has been shown to be decreased in a diabetic
state in dogs and cats [28] although this was only observed in
peripheral leucocytes of animals suffering from type 1 dia-
betes.
The other 11 proteins shown to be altered in this study
have various functions, ranging from signal transduction to
oxidative stress. Of this group, several of these proteins have
been indirectly linked to diabetes although none have been
shown to be altered in diabetic skin. For example, 14-3-3
family members are widely expressed in eukaryotes, and
while no single conserved function has been linked to 14-3-3
© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

9. 1148 E. O. List et al. Proteomics 2007, 7, 1140–1149
proteins, hundreds of binding partners are reported in the
literature linking 14-3-3 proteins to many cellular processes.
In experimental diabetes, it has been shown that inactivation
of 14-3-3 protein exacerbates cardiac hypertrophy by increas-
ing the expression of protein kinase C beta 2 [29]. Peroxi-
redoxin 6 is also linked to diabetes. It functions as a peroxi-
dase and is involved in the removal of reactive oxygen species
[30]. Oxidative stress has been linked to both insulin resis-
tance as well as many diabetic complications [31]. The higher
peroxiredoxin 6 protein levels observed is consistent with an
increase in the oxidative stress in the diabetic state.
In this study, of all proteins that were found to be
increased or decreased in skin, creatine kinase displayed the
most interesting profile and proved to be a great example of
the type of PTMs that can only be determined by proteomic
analysis and not by gene chip analysis. Two separate charge
trains with at least two charge variants of creatine kinase
were found to be altered. Both of the creatine kinase charge
variants in the upper charge train (spots B5 and G12, Figs. 4
and 5) were decreased in the diabetic state. Conversely, all
three of the creatine kinase charge variants in the lower
charge train (spots B3, B4, and B6) were increased in the
diabetic state. As this shift was from a higher charge train of
proteins in control mouse skin to a lower charge train of
proteins in the skin of mice in the diabetic state, it appears
that an overall change that occurred in the creatine kinase
variants did not represent a simple change in charge (such as
a change in phosphorylation state) but rather a shift in pro-
tein size. Moreover, this increase in the size of creatine
kinase in the skin of diabetic mice was constant in all 24 gels
analyzed.
Creatine kinase is an important enzyme involved in
energy metabolism; thus, it is not surprising that this protein
was found to be altered in the skin of diabetic mice since
diabetes has global effects on nutrient metabolism. While
not previously reported to be altered in skin, creatine kinase
has been shown to be altered in diabetic and obese states in
other tissues. More specifically, Popovich et al. [32] found that
the total creatine kinase activity was decreased by 35% in the
hearts of streptozotocin-induced diabetic rats with a 61.1%
decrease in mRNA. They also demonstrated that creatine
kinase mRNA levels and activity were both restored with
insulin therapy. In another study using streptozotocin-
induced diabetic rats, creatine kinase activity in serum, heart,
skeletal muscle, and brain were significantly lower [33]. Cu-
riously, Zhao et al. found the opposite to be true in bladder as
a higher activity of creatine kinase was observed in this tis-
sue. Zhao et al. suggested that the higher creatine kinase ac-
tivity in diabetic bladder may reflect a functional compensa-
tory mechanism. With obesity, creatine kinase activity was
shown to increase 30% in obese as compared to nonobese
women [34]. While this study reported that creatine kinase
activity was increased in obese women, no change was
observed at the protein level. Hittel et al. [34] also proposed
that creatine kinase was post-translationally modified with
obesity, which agrees with the data presented in this study.
While this still remains speculative, it would seem worth-
while in future studies to assay these alternative forms for
activity.
Unfortunately, this change in creatine kinase could be
due to obesity, high-fat feeding, diabetes or a combination of
these since all were present in the mouse model used in this
study. Thus, the specificity of this change exclusively to dia-
betes is difficult to determine. Since type 2 diabetes is highly
correlated with obesity, the obese and diabetic model still
seems to be clinically significant and an appropriate first step
in proteomic analysis of skin. Future analysis using different
diabetic mouse models that do not exhibit obesity, such as
the nonobese diabetic mice or streptozotocin-treated mice
that both develop type 1 diabetes, or conversely mouse mod-
els that develop obesity but not diabetes may help distin-
guish between the impacts of obesity versus that of diabetes.
By examining two or three independent models of diabetes
and obesity, future studies could address the specificity of
this change in creatine kinase.
Previously, our laboratory has reviewed proteomic meth-
odologies [35] and utilized these methods to study the dia-
betic pancreas [11]. The purpose of this study was two-fold:
first, to determine the feasibility of using skin for proteomic
analysis in a diabetic mouse model and second, to determine
which proteins are significantly altered in diabetic skin. Al-
though skin is a complex tissue made up of many cell types,
the reproducibility in the protein profiles and the ability to
resolve more than 1000 protein spots, suggest that skin
biopsy coupled with proteomic technologies is definitely
feasible as a diagnostic technique for the study of diabetes as
well as other disease states. In total, 28 proteins were shown
to be significantly altered when compared to nondiabetic
control mice. While these skin proteins were altered in a
high-fat fed mouse model of type 2 diabetes, a more com-
prehensive analysis of individual proteins is required to
determine if these proteins may serve as potential bio-
markers for disease diagnosis or treatment.
This work was supported in part by the State of Ohio’s Emi-
nent Scholar program, which includes a gift from the Milton and
Lawence Goll family and by DiAthegen, LLC. D. B. is supported
by a mentored career development award from NIH-NIDDK
(DK064905). Proteomics data were provided by the Michigan
Proteome Consortium (www.proteomeconsortium.org) which is
supported in part by funds from The Michigan Life Sciences Cor-
ridor.
5 References
[1] Flegal, K. M., Carroll, M. D., Ogden, C. L., Johnson, C. L.,
JAMA 2002, 288, 1723–1727.
[2] Ogden, C. L., Flegal, K. M., Carroll, M. D., Johnson, C. L.,
JAMA 2002, 288, 1728–1732.
[3] Wyatt, H. R., Prim. Care 2003, 30, 267–279.
© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

10. Proteomics 2007, 7, 1140–1149 Animal Proteomics 1149
[4] Resnick, H. E., Valsania, P., Halter, J. B., Lin, X., J. Epidemiol.
Community Health 2000, 54, 596–602.
[5] Kim, J. H., Sen, S., Avery, C. S., Simpson, E. et al., Genomics
2001, 74, 273–286.
[6] CDC, U. S. Department of Health and Human Services,
Centers for Disease Control and Prevention 2004.
[7] Spijkerman, A. M., Dekker, J. M., Nijpels, G., Adriaanse, M. C.
et al., Diabetes Care 2003, 26, 2604–2608.
[8] Surwit, R. S., Feinglos, M. N., Rodin, J., Sutherland, A. et al.,
Metabolism 1995, 44, 645–651.
[9] Surwit, R. S., Kuhn, C. M., Cochrane, C., McCubbin, J. A.,
Feinglos, M. N., Diabetes 1988, 37, 1163–1167.
[10] Surwit, R. S., Wang, S., Petro, A. E., Sanchis, D. et al., Proc.
Natl. Acad. Sci. USA 1998, 95, 4061–4065.
[11] Qiu, L., List, E. O., Kopchick, J. J., Mol. Cell. Proteomics 2005,
4, 1311–1318.
[12] Rice, J. C., Zaragoza, P., Waheed, K., Schofield, J., Jones, C.
A., Eye 2003, 17, 478–481.
[13] Bradford, M. M., Anal. Biochem. 1976, 72, 248–254.
[14] Malone, J. P., Radabaugh, M. R., Leimgruber, R. M., Ger-
stenecker, G. S., Electrophoresis 2001, 22, 919–932.
[15] Braverman, I. M., Med. Clin. North Am. 1971, 55, 1019–1029.
[16] Romano, G., Moretti, G., Di Benedetto, A., Giofre, C. et al.,
Diabetes Res. Clin. Pract. 1998, 39, 101–106.
[17] Yosipovitch, G., Hodak, E., Vardi, P., Shraga, I. et al., Diabetes
Care 1998, 21, 506–509.
[18] Lauria, G., McArthur, J. C., Hauer, P. E., Griffin, J. W., Corn-
blath, D. R., J. Neurol. Neurosurg. Psychiatr. 1998, 65, 762–
766.
[19] Mooradian, A. D., Haas, M. J., Wong, N. C., Diabetes 2004,
53, 513–520.
[20] Shih, D. Q., Heimesaat, M., Kuwajima, S., Stein, R. et al.,
Proc. Natl. Acad. Sci. USA 2002, 99, 3818–3823.
[21] Lupi, R., Marselli, L., Dionisi, S., Del Guerra, S. et al., Dia-
betes Metab. Res. Rev. 2004, 20, 246–251.
[22] Leyva, F., Wingrove, C. S., Godsland, I. F., Stevenson, J. C.,
Metabolism 1998, 47, 657–662.
[23] Wentzel, P., Ejdesjo, A., Eriksson, U. J., Diabetes 2003, 52,
1222–1228.
[24] Boraldi, F., Bini, L., Liberatori, S., Armini, A. et al., Proteom-
ics 2003, 3, 917–929.
[25] Petersen, K. F., Befroy, D., Dufour, S., Dziura, J. et al., Science
2003, 300, 1140–1142.
[26] Tatar, M., Bartke, A., Antebi, A., Science 2003, 299, 1346–
1351.
[27] Iacobini, C., Amadio, L., Oddi, G., Ricci, C. et al., J. Am. Soc.
Nephrol. 2003, 14, S264–270.
[28] Magori, E., Nakamura, M., Inoue, A., Tanaka, A. et al., Res.
Vet. Sci. 2005, 78, 39–44.
[29] Gurusamy, N., Watanabe, K., Ma, M., Zhang, S. et al., Biol.
Pharm. Bull. 2005, 28, 957–962.
[30] Manevich, Y., Fisher, A. B., Free Radic. Biol. Med. 2005, 38,
1422–1432.
[31] Evans, J. L., Goldfine, I. D., Maddux, B. A., Grodsky, G. M.,
Endocr. Rev. 2002, 23, 599–622.
[32] Popovich, B. K., Boheler, K. R., Dillmann, W. H., Am. J. Phy-
siol. 1989, 257, E573–E577.
[33] Zhao, X., Bassirat, M., Zeinab, K., Helme, R. D., Chin. Med. J.
(Engl) 1999, 112, 1028–1031.
[34] Hittel, D. S., Hathout, Y., Hoffman, E. P., Houmard, J. A., Dia-
betes 2005, 54, 1283–1288.
[35] Kopchick, J. J., List, E. O., Kohn, D. T., Keidan, G. M. O. et al.,
Endocrinology 2002, 143, 1990–1994.
© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com