Page 1 of 64 Diabetes
Temporal Transcriptomic and Proteomic Landscapes of
Deteriorating Pancreatic Islets in Type 2 Diabetic Rats
Junjie Hou1,*, Zonghong Li1,3,*, Wen Zhong1,4*, Qiang Hao1, Lei Lei1, Linlin Wang1,2, Dongyu Zhao1, Pingyong Xu1, Yifa Zhou3, You Wang1,¶, and Tao Xu1,2,4¶
1 National Laboratory of Biomacromolecules, CAS Center for Excellence in Biomacromolecules, Institute of Biophysics, Chinese Academy of Sciences, Beijing 100101, China 2 College of Life Sciences, University of Chinese Academy of Sciences, Beijing 100049, China 3 School of Life Sciences, Northeast Normal University, Changchun 130024, China 4 College of Life Science & Technology, HuaZhong University of Science and Technology, Wuhan 430074, China * These three authors contributed equally to this work.
¶ Correspondence: You Wang, E-mail: [email protected]; Tao Xu, E-mail: [email protected]
1
Diabetes Publish Ahead of Print, published online May 30, 2017 Diabetes Page 2 of 64
Abstract
Progressive reduction in β-cell mass and function comprise the core of the pathogenesis mechanism of type 2 diabetes (T2D). The process of deteriorating pancreatic islets, in which a complex network of molecular events was involved, is not yet fully characterized. Here we employed RNA Sequencing and tandem mass tag
(TMT) based quantitative proteomics technology to measure the temporal mRNA and protein expression changes of pancreatic islets in Goto-Kakizaki (GK) rats from 4 to
24 weeks of age. Our omics dataset outlined the dynamics of molecular network during the deterioration of GK islets as two stages: the early stage (4-6 weeks) is characterized by anaerobic glycolysis, inflammation priming, and compensation for insulin synthesis, whereas the late stage (8-24 weeks) is characterized by inflammation amplification and compensation failure. Further time-course analysis allowed us to reveal 5551 differentially expressed genes, a large portion of which have not been reported before. Our comprehensive and temporal transcriptome and proteome data offer a valuable resource for the diabetes research community and quantitative biology as well.
2
Page 3 of 64 Diabetes
Introduction
Type 2 diabetes (T2D) mellitus is a major public health issue characterized by
pancreatic islet β-cell failure in the presence of insulin resistance. Accumulating
evidence suggest that progressive deterioration of pancreatic β-cell function and
gradual loss of β-cell mass could be the core events during T2D development,
regardless of therapy status (1-4). Recent genome-wide association (GWAS) and
sequencing studies identified multiple risk variants for T2D, the majority of which
appeared to have a primary role in β-cell function rather than an impact on insulin
resistance, further highlighting the importance of β-cells in the pathogenesis of T2D
(5).
T2D is a complex disease, and β-cell failure is likely caused by altered expression of
many genes and their products. Therefore, it is critical to employ system-oriented
strategies to investigate the complex changes that occur in β-cells or pancreatic islets,
which are primarily composed of β-cells. Hence, large-scale and unbiased ‘‘omics’’
technologies, particularly microarray-based transcriptomics and mass
spectrometry-based proteomics, have been employed to analyze islets isolated from
various T2D animal models and human cadaver donors to elucidate the mechanisms
underlying β-cell failure (summarized in Supplementary Table 1). It is evident that
β-cell failure during diabetes progression is a gradual process undergoing different
stages (6,7), wherein different molecule events occurred in chronological order.
However, current studies have primarily focused on single time points at relatively
late stages of the disease, therefore it is impossible to map out the order in which
3
Diabetes Page 4 of 64
these events occur and distinguish causal molecular events (leading to diabetes) from those occurring as a consequence of glucolipotoxicity associated with diabetic conditions. For this reason, it is of great interest to carry out prospective studies investigating the evolution of molecular events in islet β-cells at different stages of
T2D.
Unfortunately, the study of β-cells in diabetic humans has often been hindered by the limited accessibility of human islets and ethical considerations. In this context, appropriate rodent models are essential tools for the identification of diabetic mechanisms (8). The Goto–Kakizaki (GK) rat, one of the best-characterized animal models of spontaneous T2D (9), shares many characteristics with human diabetic patients (10). Similar to human T2D, the core cause underlying hyperglycemia in GK rats is β-cell failure (11,12).
In this study, to understand the process of deteriorating pancreatic islets at the molecular level, we employed RNA Sequencing and tandem mass tag (TMT) based quantitative proteomics technology to generate integrated transcriptomic and proteomic profiles of pancreatic islets in GK rats after the establishment of hyperglycemia (from 4 to 24 weeks). Subsequent bioinformatics analysis in a time-course fashion revealed the chronological order of T2D-related molecular events during the deterioration of pancreatic islets. Our large quantitative dataset represents a valuable resource that provides a comprehensive picture of the mechanisms responsible for islet dysfunction and will allow us to identify potential interventions to prevent β-cell failure and deterioration in human T2D.
4
Page 5 of 64 Diabetes
Research Design and Methods
Brief descriptions of key experimental procedures are provided below. More details
are given in the Supplemental Experimental Procedures.
Animals
Founders of GK/Jcl diabetic rats were purchased from RIKEN BioResource Center
(Ibaraki, Japan). All GK/Jcl diabetic rats and Wistar (WST) rats were maintained
under specific-pathogen-free (SPF) conditions and were used between 4-24 weeks of
age in accordance with the animal experimental guidelines set by the Institutional
Animal Care and Use Committee of the Institute of Biophysics, CAS.
Preparation of pancreatic islets from GK and WST rats
Pancreatic islets from male GK and age-matched, control WST rats were isolated via
collagenase digestion. After separation on a Ficoll density gradient, the islets were
hand-picked in Hank’s buffer under a dissection microscope.
N-acetyl-L-cysteine (NAC) Treatment Experiments
A total of 16 male GK rats were used in the following experiments. Littermates of GK
rats were randomly divided into NAC and control groups. 4-week-old rats were orally
administered NAC (200 mg/kg of body weight, 616-91-1, Sigma) or drinking water
by gavage once per day for 12 weeks. Random blood glucose assay, Glucose
tolerance test (GTT), Insulin tolerance test (ITT) and Glucose-stimulated insulin
secretion (GSIS) test were performed as described in the Supplemental Experimental
Procedures.
RNA-Seq analysis
RNA-Seq was performed using a MAPS (Multiplex Analysis of PolyA-linked
Sequences) approach as previously described (13).
TMT-based proteomics analysis 5
Diabetes Page 6 of 64
Proteins extracted from isolated islets were digested, and labeled by 6-plex TMT reagents (14) according to the instructions from the manufacture (Thermo Fisher).
TMT-labeled peptide mixtures were equally pooled, separated by off-line high-pH reversed-phase chromatography, and technique-repeatedly analyzed by nanoLC-MS/MS (Supplementary Fig. 1A). The raw MS data were processed with
Proteome Discovery (version 1.4). The peptide confidence value was set as 0.01. At the protein level, a precursor intensity fraction (PIF) of 50% was selected as an optimal trade-off value for both identification and quantification (Supplementary Fig.
1B). A “pseudocount”, representing relative protein abundance, was calculated using the TMT ratio and the normalized spectral abundance factor (NSAF) (15).
Bioinformatics analysis
Both mRNA raw counts and protein pseudocounts were normalized by employing the
RUV approach (16) (Supplementary Fig. 1C). Differentially expressed (DE) genes were assessed by ANOVA with an FDR of less than 0.01. The Database for
Annotation, Visualization and Integrated Discovery (DAVID) Web Service API Perl
Client (17,18) was utilized to perform GO functional enrichment analysis (with an
FDR of less than 0.05). The k-means clustering algorithm was used to classify dynamic gene expression patterns. KEGG signaling pathway enrichment analysis was carried out using the GAGE (19) package in R software (with a adjusted p value of less than 0.05).
Data resources
92.4 GB sequencing data were generated for this work. All RNA-seq data were deposited in the NCBI Gene Expression Omnibus (GEO) under accession number
GEO: GSE81811. The mass spectrometry data were deposited in the Proteome
Xchange Consortium via the PRIDE (20) partner repository with the dataset identifier
6
Page 7 of 64 Diabetes
PXD004709 (Account details for referees, username:[email protected],
password: KebbybEk).
7
Diabetes Page 8 of 64
Results
Transcriptomic and proteomic profiles of rat pancreatic islets over time
To investigate the global molecular dynamics of T2D islets, we analyzed the transcriptomes and proteomes of pancreatic islets isolated from male GK rats and age- and sex-matched WST rats at five consecutive time points (weeks 4, 6, 8, 16 and 24,
Fig. 1A). Transcriptomes and proteomes of islets were measured using the
MAPS-based RNA-Seq technique and TMT labeling-based proteomic method, respectively. Combined analysis of all samples yielded the identification of a total of
15101 mRNAs and 8362 proteins, of which 7395 overlapped (Fig. 1B). Furthermore,
13866 mRNAs (minimal counts of 10 detected in at least 3 samples) and 5631 proteins (identified in at least two biological replicates by peptides of PIF no more than 50%) were considered a quantifiable dataset, of which 5015 overlapped
(information for all identified genes provided in Supplementary Table 2). We estimated relative protein abundance, which was noted as the protein pseudocount in this study, based on the TMT ratio and NSAF. Compared with TMT ratios, protein pseudocounts resulted in the generation of more reasonable clusters that matched the experiments (Supplementary Fig. 2A). Protein pseudocounts positively correlated with mRNA counts, with a mean Spearman correlation coefficient of 0.37
(Supplementary Fig. 2B), similar to reported values in other biological systems
(21,22). These results demonstrated that our method of estimating protein pseudocounts was reasonable and unbiased.
First, we carried out unsupervised hierarchical clustering analysis of the transcriptomes and proteomes from GK and WST islets (Fig. 1C). At the mRNA level,
WST islets clustered together and clearly separated from GK islets as expected, representing different pathophysiological states of islets in control rats vs diseased
8
Page 9 of 64 Diabetes
rats. Overall, the transcriptomes of GK and WST rats demonstrated appropriate
clustering at different time points, representing the developmental stages of islets.
Interestingly, GK islets at 4, 6 and 8 weeks formed one branch that was distinct from
the branch comprising 16 and 24 weeks, likely reflecting two different diabetic stages
of islets in GK rats. When the proteome clustering results were examined, GK islets at
4 and 6 weeks were separated from other GK islets and instead clustered together with
WST islets, suggesting minimal changes in protein expression at the early stages of
T2D, even when there were significant changes at the mRNA level.
Next, we also performed principal component analysis (PCA) to investigate
transcriptome and proteome dynamics over time (Fig. 1D). PC2 primarily reflected
age-related developmental changes, whereas PC1 represented the differences between
normal WST and diabetic GK rats. Interestingly, similar to the above results obtained
for unsupervised hierarchical clustering, PC1 highlighted two notable diabetic stages
in GK islets. At the mRNA level, islets at 4, 6, and 8 weeks clustered together as one
stage, and islets at 16 and 24 weeks represented another stage. However, at the protein
level, islets at 4 and 6 weeks clustered as one stage, and the remaining time points
clustered together as a separate stage. GK islets progressively develop into
disorganized structures exhibiting pronounced fibrosis separating strands of endocrine
cells (23). Interestingly, these changes are not present or rare in islets at 4 to 6 weeks,
but become prominent at later ages (8-24 weeks) (Supplementary Fig. 2C), correlating
well with our proteome-defined two stages (Fig. 1C and 1D).
Taken together, global profiling at the mRNA and protein levels roughly characterize
the deterioration of islets in GK rats from 4 to 24 week of age into two stages: an
early stage at 4-6 weeks and a late stage at 8-24 weeks, with a turning point at
approximately 8 weeks.
9
Diabetes Page 10 of 64
Analysis of DE mRNAs and proteins
DE genes between GK and WST at each time point were first assessed by one-way
ANOVA (analysis of variance, p < 0.05) (Supplementary Table 2). The results revealed a remarkable increase of DE mRNAs over time, whereas there was only a mild increase in the number of DE proteins, suggesting a more dynamic regulation of gene expression at the mRNA level than at the protein level during the development of GK diabetes (Supplementary Fig. 3A). When we compared the fold changes in mRNA and protein levels, the Pearson’s correlation value (0.11) was relatively poor at 4 weeks but increased to 0.23-0.28 at later time points (Supplementary Fig. 3B).
Benefiting from our time-resolved expression dataset, we analyzed the temporal significance of gene expression changes using two-way ANOVA, considering weeks
(5 different time points) and rats (GK vs WST) as the two statistical factors. In total, we identified 5551 DE genes, including 3910 mRNAs and 2387 proteins
(Supplementary Table 2), of which 746 were identified at both the mRNA and protein levels. The correlations between these 746 DE genes varied from anti-correlation to full accordance, with an average Pearson’s coefficient of 0.39 (Fig. 2A,
Supplementary Table 3), which is larger than the correlation coefficients at individual time points. DAVID functional clustering analysis indicated carbon metabolism and ribosome enrichment among genes with concordant mRNA and protein levels (Fig.
2B), potentially representing the set of genes exhibiting stable mRNA and protein expression (24). In contrast, no significant functional clustering was identified for
10
Page 11 of 64 Diabetes
negatively correlated DE genes.
For the confirmation of expression data, we randomly selected six DE genes to
measure their mRNA expression by qRT-PCR in independent GK/WST islets samples.
The results demonstrated that the expression patterns of these genes measured by
RNA-seq and qRT-PCR were very similar (Supplementary Fig. 3C). Moreover, by
comparing our GK/WST dataset with published dataset of islets from normal and T2D
individuals (Supplementary Table 1, Ref. 10), we found that on average 68.9% of DE
genes were consistent between diabetic GK rat and human (Supplementary Fig. 3D).
This indicates high relevance of our study in GK rat to human islets in T2D.
Dynamic mRNA and protein expression patterns over time in GK islets
The primary purpose for the generation of the time course dataset in this study is to
reveal the temporal properties of biological pathways relevant to the development of
GK diabetes at the system level. Therefore, we performed time-course pattern
analysis for all DE genes using the k-means clustering method and successfully
identified 12 mRNA expression patterns (m1-m12) and 9 protein expression patterns
(p1-p9) (Fig. 2C). To explore the biological functions of these expression patterns, we
carried out DAVID analysis and organized the identified networks with enriched
functions using EnrichmentMap (25) (Supplementary Table 4). Based on the
time-course expression patterns of clustered genes, we were able to classify the
biological events into the following categories: 1) Constant up (m2, p8) and down (p2,
m3, m10) genes, which were either up- or down-regulated in all the time points. The
11
Diabetes Page 12 of 64
constant up genes were highly enriched for functions including cell redox homeostasis, translation elongation, cytoskeleton organization, anti-apoptosis and so on. In contrast, the constant down genes were mainly associated with mitochondrion, metabolism, lysosome, protein transport and so on. 2) Up early genes (p1, p6, m11) that were up-regulated at 4-8 week, which include those participating in the glucose metabolism and innate immune response. In addition, many proteasome proteins were dramatically up-regulated at 4 week. 3) Up late genes (p9, m8, p7, m1, m12) that were up-regulated at late stage of 8-24 week. These genes are highly enriched for cell adhesion and cytoskeleton organization at protein level, probably associating with the development of islets fibrosis. While at mRNA level, apoptosis was significantly up-regulated at 24 weeks. 4) Down early genes (p3, m6) that were down-regulated at
4-8 week. The most noticable feature is the down-regulation of oxidative phosphorylation and TCA cycle at protein level, probably suggesting the insufficient energy supply and oxidative stress in GK islets at early stage (26-28). At mRNA level, the genes assoicated with cell cycle and nuclear lumen were highly enriched, likely contributing to the loss of β-cells in GK islets. 5) Down late genes (p5, m9, p4, p7, m5, m4) that were down-regulated at the 8-16 week. The GO functions of insulin secretion, lysosome and secretory granule were representitively enriched.
Such time-course clustering analysis from genes to biological functions suggested that the progression of deteriorating of GK islets was stage-based and dynamically regulated. Furthermore, since genes exhibiting similar expression patterns generally share functional relationships, clustering analyses also allow us to predict genes that
12
Page 13 of 64 Diabetes
share similar temporal expression patterns with previously validated diabetes-related
genes as potential new candidates for further investigation.
Pathway dynamics in GK islets during diabetes progression
To gain a deeper understanding of temporal pathway sequences during the
deterioration of GK islets, we performed GAGE analysis with the quantitative
transcriptomic and proteomic datasets, and identified 161 KEGG pathways
significantly enriched for at least one-time point (Benjamini-Hochberg method
adjusted p value less than 0.05) (Supplementary Table 5). Consistent with the gene
expression clustering analysis results described above, these KEGG pathways also
roughly comprised of down- and up-regulation-dominated temporal classes (Fig. 3).
For example, insulin secretion, and SNARE interactions in vesicular transport were
down-regulated at the late stage with the same temporal pattern (the represented genes
were illustrated in Supplementary Fig. 4). These two pathway was partially
down-regulated at 4 and 6 weeks with no significant change in statistics, and became
significantly down-regulated after 8 weeks. This clearly demonstrated the dynamic
features of various pathways involved in insulin secretion during the progression of
T2D. We also noticed the pathway of glycolysis/gluconeogenesis was gradually
up-regulated since 6 week, probably indicating in GK islets the anaerobic metabolism
turned to be the dominant approach for supplying the energy, due to the defect of
oxidative phosphorylation and TCA cycle.
In addition, we found signaling pathways associated with inflammation were
13
Diabetes Page 14 of 64
up-regulated at mRNA level, including NOD-like receptor signaling pathway, TNF signaling pathway and NF-kappa B signaling pathway. This is consistent with previous reports that islets inflammation plays an important role in the pathogenesis of GK diabetes (26,29), and aslo provided more comprehensive details of pathway dynamics at both mRNA and protein levels.
Mitochondrial signatures in GK islets
We identified 311 DE genes as mitochondria-related genes using GO terms for cellular component that contained the keywords “mitochondrion” or “mitochondrial”, and divided them into four groups based on unsupervised hierarchical clustering analysis of their temporal profiles (Supplementary Table 6). Further manual annotation revealed the details regarding mitochondrial dysfunction during the progression of T2D (Fig. 4). We found that OXPHOS complexes, mitochondrial ribosome proteins, translocase outer/inner membrane complex (TOM/TIM) and some metabolite transporters were down-regulated early at protein level, and most of corresponding mRNAs were conversely up-regulated at 4 and 6 week, indicating transcription compensation at the early stage of T2D. However, this compensation ability was eventually lost at the later stage during β-cell deterioration. In addition, several proteins responsible for protein assembly and quality control, mitochondrial biogenesis, and mitochondrial DNA transcription in mitochondria were also down-regulated, such as Hspd1, Grpel1, Lonp1, Letm1 and Pmpcb, Tfam, Mtfr1l,
Mfn2, the tRNA ligases (Rars2, Nars2, Dars2 and Vars2), and 12 mitochondrial RNAs
14
Page 15 of 64 Diabetes
(Supplementary Fig. 5A). Taken together, mitochondria dysfunction were considered
as the one of earliest panthogentic events in islets of GK rat.
Overview of metabolism in GK islets
To gain a deeper understanding how the metabolism in GK islets changed with the
development of diabetes, we mapped our quantitative omics data to KEGG metabolic
pathways (Fig. 5). It turns out that the most notable change of metabolism in GK
islets was the up-regulation of glycolysis metabolism (26,27) and the down-regulation
of the TCA cycle, OXPHOS and fatty acid metabolism. This might suggest that the
primary metabolism of GK islets switches from aerobic metabolism to anaerobic
metabolism (the so-called Warburg-like effect). Furthermore, we found that besides
FAD-dependent glycerol-3-phosphate dehydrogenase (Gpd2) reported previously (30),
Got1, Got2, Mdh1 and Mdh2 were also down-regulated as well, representing a
comprehensive picture of defective malate-aspartate shuttle and glycerol phosphate
shuttle in GK islets (Supplementary Fig. 5B). Such defection could cause the increase
of the cytoplasmic NADH/NAD+ ratio, enhance the formation of lactate from
pyruvate, and further disrupt the link between cytosolic glycolysis and mitochondrial
metabolism.
In addition, some genes participating in some certain amino acid metabolism were
also down-regulated, partially contributing the defect of GSIS (31-33). However,
interestingly, several enzymes involved in glutathione metabolism were up-regulated
at both the mRNA and protein levels in GK islets, including Gclc, Ggct, Ggt1, Gsta3,
15
Diabetes Page 16 of 64
and Gsto1, which may reflect an adaptive mechanism to combat increased ROS in GK islets (34,35).
Reduced neogenesis and βββ-cell proliferation
Insufficient insulin secretion is caused by either impaired GSIS or reduced β-cell mass. β-cell mass is regulated by a balance between β-cell neogenesis, proliferation and apoptosis. In our data, consistent with previous studies (26,36) several genes that are functionally associated with neogenesis were down-regulated in GK islets, including Pdx1, Nkx2-2, Nkx6-1, Mafa and Fev, at the mRNA and/or protein levels
(Fig. 6). Moreover, several genes specific for α cells and PP cells, i.e., Arx, Isl1, Pax6,
Pou3f4, and Ppy, were also down-regulated at early stage. Certain genes required for the differentiation of pancreatic progenitors into endocrine progenitors, i.e., Foxa1,
Gata6, Sox9 and Onecut1, were significantly up-regulated at the late stage of GK diabetes.
Low levels of β-cell proliferation in GK rats constitute another factor contributing to decreased β-cell mass (11,26,36). In our dataset, 38 genes among cluster m6, including Aurkb, Ccna2, and Kifc1, were associated with cell cycle and nuclear lumen, and exhibited very similar expression patterns at early stage (Supplementary Fig. 6A).
These genes gradually decreased over time in WST rats, reflecting the ageing-dependent reduction in proliferation over time (37). In contrast, all of these genes were dramatically down-regulated at 4 weeks in GK islets and then progressively decreased to even lower levels at 24 weeks (Supplementary Fig. 6B).
16
Page 17 of 64 Diabetes
Consistently, Ki67-positive β-cells were significantly decreased in GK islets at 4, 6
and 8 weeks (Supplementary Fig. 6C), suggesting reduced β-cell proliferation
(11,12,26). Conversely, the apoptosis pathway was only elevated at 24 weeks at the
mRNA level, implying apoptosis was not responsible for β-cell loss during the early
phase of the disease.
Two-stage inflammation in GK islets
Chronic inflammation in GK islet has recently been demonstrated and considered as a
pathophysiological contributor in T2D (29,38). In our study, temporal expression of
pro-inflammatory cytokines revealed two distinct stages: early priming and late
amplification. During the priming stage, interleukin-1β (IL-1β) and IL-6 were only
elevated to low levels between 4-8 weeks, followed by a rapid increase to much
higher levels (84-fold increase in GK at 24 weeks for IL-6) during the late
amplification stage (between 16-24 weeks) (Fig. 7). As immune cell infiltration was
hardly detectable in GK islets prior to 8 weeks (38), the early priming stage is likely
induced by metabolic dysfunction in GK islets. The identity of specific sensors that
are triggered to produce this priming of inflammation was not fully understood.
The NLRP3 (NLR family, pyrin domain containing 3) inflammasome activates both
NF-kappaB (to induce pro-IL-1β production) and caspase-1 (to process pro-IL-1β into
its mature active form). ASC (apoptosis associated speck-like protein containing a
CARD, also called PYCARD), which interacts with NLRP3 during inflammasome
assembly, was up-regulated in GK islets, indicating possible activation of the NLRP3
17
Diabetes Page 18 of 64
inflammasome during the initiation of sterile inflammation. The NLRP3 inflammasome is also activated by thioredoxin-interacting protein (TXNIP), a key node linking glucotoxicity and ER stress to NLRP3 inflammasome activation (39).
We observed TXNIP was gradually up-regulated in GK islets after 6 weeks.
Invasive ROS contributing to the deterioration of islets in GK
Oxidative stress is a pathogenic factor caused by chronic hyperglycemia in GK pancreatic islets (40). However, sources of ROS in T2D have not been clearly defined
(41). Our data suggest multiple sources of ROS accumulation in GK islets at early stage, including dysfunctional mitochondria, nitric oxide synthase (Nos2/iNOS),
NADPH oxidase (Nox4), cyclooxygenase (Ptgs1/Cox1 and Ptgs2/Cox2) and cytochrome P450 monooxygenase (Cyp2s1, Cyp7b1 and Cyp4f5) (Fig. 7). We also observed the increased expression of several antioxidants, such as Sod1, Prdx4 and
Gpx2 (Fig. 7), in GK islets, consistent with previous reports (27,35,42).
To validate the involvement of ROS in the pathogenesis of T2D, we treated
4-week-old GK rats with an antioxidant, N-acetylcysteine (NAC), for 10 weeks. NAC treatment significantly reduced the random blood glucose of GK-NAC rats (n = 8,
NAC of 200 mg/kg) compared with that of age-matched GK-control rats (sham, n = 8)
(Supplementary Fig. 7A-D), and GK-NAC rat exhibited greater tolerance to high glucose in the oral glucose tolerance test. No significant differences in insulin sensitivity between GK-NAC and GK-control rats were observed. The islets in
GK-NAC groups have higher GSIS, as compared with those in GK-control groups
18
Page 19 of 64 Diabetes
(Supplementary Fig. 7E). Furthermore, we measured the mRNA expressions in islets
between GK-NAC and GK-control using RNA-Seq. The results showed that insulin
secretion pathway was up-regulated, whereas ROS and inflammation-related genes
(Tnf, Ggt1, Pycard, Cxcl1, etc.) and signaling pathways (NOD-like receptor signaling
pathway, TNF signaling pathway, metabolism of xenobiotics by cytochrome P450,
etc.) were down-regulated in GK islets after NAC treatment (Supplementary Fig. 7F).
Thus, the neutralization of increased ROS by NAC ameliorates its impact on GSIS
and inflammation and protects GK β-cell function.
19
Diabetes Page 20 of 64
Discussion
In this study, we carried out a large-scale analysis of gene and protein dynamics in pancreatic islets of GK rats at different stages of T2D. Combined transcriptome and proteome analysis revealed sufficient depth of coverage and quantitative accuracy to generate functional portraits of healthy and diseased pancreatic islets with unprecedented detail. Many DE genes and proteins identified previously were confirmed in our study, such as those associated with OXPHOS, mitochondrial function, metabolism, insulin secretion, oxidative stress and inflammation, thus validating the analytic methods employed here. More importantly, the construction of time course-based gene expression and protein profiles allowed us to identify the chronological order of biological events contributing to the pathogenesis of T2D.
Our data suggest two early events that likely contribute T2D in GK islets: a reduction in β-cell mass and a shift in metabolism. Many transcription factors required for the specification of endocrine cells were down-regulated early in GK islets, whereas those required for trunk and exocrine cells were not altered at 4 weeks but increased at later time points (Fig. 6). Indeed, GK rats from the Paris colony exhibit a significant reduction in β-cell mass at the fetal stage that precedes the onset of hyperglycemia at approximately 4 weeks after birth (11), similar to our GK colony. Besides neogenesis, defective proliferation also causes reduction in β-cell mass as has been suggested for
GK Paris colony(11,26). In our GK rats, many genes required for the cell cycle and proliferation were significantly down-regulated at early stage (4-6 weeks)
(Supplementary Fig. 6C and 6D), suggesting an early defect in proliferation.
20
Page 21 of 64 Diabetes
Another notable feature of our data is the observation of an early shift in metabolism.
As early as 4 weeks, the primary metabolism in the islets of GK rats switched from
aerobic metabolism to anaerobic metabolism (Warburg-like effect). Although
metabolism switch has been proposed in previous research (34), our time
course-based quantitative data allow us to detect this metabolic shift as an early event
of T2D in GK rats and permit us to gain a deeper insight into the mechanism
underlying this shift. Insufficient reduction in β-cell mass alone may not necessarily
cause T2D, as autopsy studies of patients with T2D have revealed a ~50% decrease in
β-cell mass compared to body mass index-matched controls (43,44). Furthermore, a
reduction in β-cell mass of approximately 50% is required for dogs and rats to
develop diabetes (45). This metabolic shift is likely caused by mitochondrial
dysfunction. Many proteins associated with the OXPHOS system, metabolite
transporters, and the TCA cycle were down-regulated starting at 4 weeks (Fig. 3).
Thus, the mitochondrial limitation of glucose oxidation in GK islets occurred during
the early stage. To compensate for this energy deficiency, GK islets improved the rate
of glycolysis and up-regulated the expression of Ldha, which converted pyruvate to
lactate and further disrupted the link between cytosolic glycolysis and mitochondrial
metabolism (Fig. 4). It was demonstrated that anaerobic glucose metabolism with
NADH accumulation in the β cell of mitochondrial diabetes, caused by ethidium
bromide (EtBr) treatment that can impair the transcription of mitochondria DNA,
halted TCA cycle and impacted on GSIS (46). Therefore, this Warburg-like metabolic
shift may contribute to the early impairment of GSIS in GK β-cells because GSIS
21
Diabetes Page 22 of 64
requires the production of both a triggering signal (ATP) and amplifying signals (i.e., cAMP, short-chain acyl-CoA compounds, and NADPH) produced during aerobic metabolism (47).
Despite the early reduction in β-cell mass, insulin and the enzymes required for proinsulin processing were comparable or even higher at both the mRNA and protein levels at 4-6 weeks, suggesting compensation during the early phase in response to hyperglycemia. Absolute insulin insufficiency only occurred during the late stage of
T2D. Consistently, we also observed compensation for OXPHOS complexes and mitochondrial machinery at the mRNA level at 4-8 weeks, despite reduced protein levels. Thus, the progression of T2D in GK rats from 4-24 weeks can be divided into two stages: compensation (4-6 weeks) and compensation failure (8-24 weeks). Further data mining to examine temporal expression patterns will help to elucidate the mechanisms underlying compensation and decompensation.
Mitochondrial dysfunction and the Warburg effect generate greater ROS invasion, which in turn induces chronic, low-grade inflammation (48). Interestingly, pro-inflammatory cytokines also exhibit two distinct stages (Fig. 7): a priming stage at 4-6 weeks and an amplification stage after 8 weeks. Given that no inflammatory cell infiltration was observed in GK islets before 8 weeks (38), the priming stage of inflammation was likely induced by intracellular signals generated by metabolic stress, such as the presence of increased ROS or free fatty acids. However, the second stage of inflammation amplification may be induced by a complex combination of intracellular and extracellular (i.e., macrophage infiltration) inducers. Our data
22
Page 23 of 64 Diabetes
provide clues to unravel the mechanism underlying the initiation and amplification of
sterile inflammation; understanding this mechanism is necessary to develop novel
anti-inflammation therapies to treat T2D. Islet inflammation is undoubtedly an early
event during T2D pathogenesis, but it is not likely a causal event because IL-1β and
TXNIP were not significantly expressed at 4 weeks, although they gradually increased
during later stages of the disease.
In summary, our data reveal two stages during the progression of T2D in GK islets.
The early stage (4-6 weeks) is characterized by anaerobic glycolysis, inflammation
priming, and compensation for insulin synthesis, whereas the late stage (8-24 weeks)
is characterized by inflammation amplification and compensation failure (as depicted
in Supplementary Fig. 8). We did not observe significant apoptosis during the early
stage. The apoptosis pathway was only significantly elevated at 24 weeks at the
mRNA level. Our time course transcriptome and proteome datasets for GK rat islets
depict a comprehensive landscape of dynamic changes in gene expression at different
stages of diabetes, representing a valuable resource for the research community to
further explore the molecular etiology and progression of diabetes. In-depth
exploration of this resource will aid in the discovery of potential diagnostic and
therapeutic targets for human T2D.
Acknowledgments We thank the staff of the Institute of Biophysics Core Facilities, in particular, Yan
Teng for her technical support with confocal imaging, Dr. Jifeng Wang for MS
operation, and Zhen Fan and Xiaowei Chen for RNA-seq design and data collection.
23
Diabetes Page 24 of 64
This work was supported by grants from the National Key Basic Research Project of
China (2015CB910303), the Strategic Priority Research Program of the Chinese
Academy of Sciences (XDA12030000), the National Science Foundation of China
(31421002, 31400703, and 31400658), the National Key Basic Research Project of
China (2014CB910503).
Author contributions T.X and Y.W conceived the project. Y.W carried out the islet preparation and animal experiments. L.L and W.L performed qRT-PCR and GSIS experiments. J.H performed the proteomic experiments and MS data analysis. Z.L performed the RNA-Seq experiments and immunohistochemistry imaging. W.Z, D.Z and J.H carried out the bioinformatics analyses. J.H and W.Z prepared the figures. Q.H carried out rat breeding. J.H, Y.W and T.X wrote the manuscript with help from all of the authors.
All authors read the manuscript and discussed the interpretation of results. T.X is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Competing financial interests
The authors declare no competing financial interests.
24
Page 25 of 64 Diabetes
References
1. Saisho, Y. (2015) beta-cell dysfunction: Its critical role in prevention and management of type 2
diabetes. World journal of diabetes 6, 109-124
2. Turner, R. C. (1998) The U.K. Prospective Diabetes Study. A review. Diabetes care 21 Suppl 3, C35-38
3. Cnop, M., Welsh, N., Jonas, J. C., Jorns, A., Lenzen, S., and Eizirik, D. L. (2005) Mechanisms of
pancreatic beta-cell death in type 1 and type 2 diabetes: many differences, few similarities. Diabetes
54 Suppl 2, S97-107
4. Del Prato, S., Bianchi, C., and Marchetti, P. (2007) beta-cell function and anti-diabetic
pharmacotherapy. Diabetes/metabolism research and reviews 23, 518-527
5. Brun, T., Li, N., Jourdain, A. A., Gaudet, P., Duhamel, D., Meyer, J., Bosco, D., and Maechler, P. (2015)
Diabetogenic milieus induce specific changes in mitochondrial transcriptome and differentiation of
human pancreatic islets. Human molecular genetics 24, 5270-5284
6. Fonseca, V. A. (2009) Defining and characterizing the progression of type 2 diabetes. Diabetes care
32 Suppl 2, S151-156
7. Weir, G. C., and Bonner-Weir, S. (2004) Five stages of evolving beta-cell dysfunction during
progression to diabetes. Diabetes 53 Suppl 3, S16-21
8. King, A. J. (2012) The use of animal models in diabetes research. British journal of pharmacology
166, 877-894
9. Goto Y, K. M., Masaki N. (1975) Spontaneous diabetes produced by selective breeding of normal
Wistar rats. Proc Jpn Acad. 51, 5
10. Akash, M. S., Rehman, K., and Chen, S. (2013) Goto-Kakizaki rats: its suitability as non-obese
diabetic animal model for spontaneous type 2 diabetes mellitus. Current diabetes reviews 9, 387-396
25
Diabetes Page 26 of 64
11. Movassat, J., Saulnier, C., Serradas, P., and Portha, B. (1997) Impaired development of pancreatic
beta-cell mass is a primary event during the progression to diabetes in the GK rat. Diabetologia 40,
916-925
12. Plachot, C., Movassat, J., and Portha, B. (2001) Impaired beta-cell regeneration after partial
pancreatectomy in the adult Goto-Kakizaki rat, a spontaneous model of type II diabetes.
Histochemistry and cell biology 116, 131-139
13. Fox-Walsh, K., Davis-Turak, J., Zhou, Y., Li, H., and Fu, X. D. (2011) A multiplex RNA-seq strategy to
profile poly(A+) RNA: application to analysis of transcription response and 3' end formation.
Genomics 98, 266-271
14. Dayon, L., Hainard, A., Licker, V., Turck, N., Kuhn, K., Hochstrasser, D. F., Burkhard, P. R., and
Sanchez, J. C. (2008) Relative quantification of proteins in human cerebrospinal fluids by MS/MS
using 6-plex isobaric tags. Analytical chemistry 80, 2921-2931
15. Zhang, Y., Wen, Z., Washburn, M. P., and Florens, L. (2010) Refinements to label free proteome
quantitation: how to deal with peptides shared by multiple proteins. Analytical chemistry 82,
2272-2281
16. Risso, D., Ngai, J., Speed, T. P., and Dudoit, S. (2014) Normalization of RNA-seq data using factor
analysis of control genes or samples. Nature biotechnology 32, 896-902
17. Huang da, W., Sherman, B. T., and Lempicki, R. A. (2009) Systematic and integrative analysis of large
gene lists using DAVID bioinformatics resources. Nature protocols 4, 44-57
18. Huang da, W., Sherman, B. T., and Lempicki, R. A. (2009) Bioinformatics enrichment tools: paths
toward the comprehensive functional analysis of large gene lists. Nucleic acids research 37, 1-13
19. Luo, W., Friedman, M. S., Shedden, K., Hankenson, K. D., and Woolf, P. J. (2009) GAGE: generally
26
Page 27 of 64 Diabetes
applicable gene set enrichment for pathway analysis. BMC bioinformatics 10, 161
20. Vizcaino, J. A., Csordas, A., del-Toro, N., Dianes, J. A., Griss, J., Lavidas, I., Mayer, G., Perez-Riverol,
Y., Reisinger, F., Ternent, T., Xu, Q. W., Wang, R., and Hermjakob, H. (2016) 2016 update of the PRIDE
database and its related tools. Nucleic acids research 44, D447-456
21. Zhang, B., Wang, J., Wang, X., Zhu, J., Liu, Q., Shi, Z., Chambers, M. C., Zimmerman, L. J., Shaddox,
K. F., Kim, S., Davies, S. R., Wang, S., Wang, P., Kinsinger, C. R., Rivers, R. C., Rodriguez, H.,
Townsend, R. R., Ellis, M. J., Carr, S. A., Tabb, D. L., Coffey, R. J., Slebos, R. J., Liebler, D. C., and Nci,
C. (2014) Proteogenomic characterization of human colon and rectal cancer. Nature 513, 382-387
22. Gry, M., Rimini, R., Stromberg, S., Asplund, A., Ponten, F., Uhlen, M., and Nilsson, P. (2009)
Correlations between RNA and protein expression profiles in 23 human cell lines. BMC genomics 10,
365
23. Guenifi, A., Abdel-Halim, S. M., Hoog, A., Falkmer, S., and Ostenson, C. G. (1995) Preserved beta-cell
density in the endocrine pancreas of young, spontaneously diabetic Goto-Kakizaki (GK) rats.
Pancreas 10, 148-153
24. Schwanhausser, B., Busse, D., Li, N., Dittmar, G., Schuchhardt, J., Wolf, J., Chen, W., and Selbach, M.
(2011) Global quantification of mammalian gene expression control. Nature 473, 337-342
25. Merico, D., Isserlin, R., Stueker, O., Emili, A., and Bader, G. D. (2010) Enrichment map: a
network-based method for gene-set enrichment visualization and interpretation. PloS one 5, e13984
26. Portha, B., Giroix, M. H., Tourrel-Cuzin, C., Le-Stunff, H., and Movassat, J. (2012) The GK rat: a
prototype for the study of non-overweight type 2 diabetes. Methods in molecular biology 933,
125-159
27. Portha, B., Lacraz, G., Chavey, A., Figeac, F., Fradet, M., Tourrel-Cuzin, C., Homo-Delarche, F., Giroix,
27
Diabetes Page 28 of 64
M.-H., Bailbé, D., Gangnerau, M.-N., and Movassat, J. (2015) Islet Structure and Function in the GK
Rat. in Islets of Langerhans (Islam, M. S. ed.), Springer Netherlands, Dordrecht. pp 743-765
28. Giroix, M. H., Saulnier, C., and Portha, B. (1999) Decreased pancreatic islet response to L-leucine in
the spontaneously diabetic GK rat: enzymatic, metabolic and secretory data. Diabetologia 42,
965-977
29. Homo-Delarche, F., Calderari, S., Irminger, J. C., Gangnerau, M. N., Coulaud, J., Rickenbach, K., Dolz,
M., Halban, P., Portha, B., and Serradas, P. (2006) Islet inflammation and fibrosis in a spontaneous
model of type 2 diabetes, the GK rat. Diabetes 55, 1625-1633
30. Ostenson, C. G., Abdel-Halim, S. M., Rasschaert, J., Malaisse-Lagae, F., Meuris, S., Sener, A., Efendic,
S., and Malaisse, W. J. (1993) Deficient activity of FAD-linked glycerophosphate dehydrogenase in
islets of GK rats. Diabetologia 36, 722-726
31. Gheni, G., Ogura, M., Iwasaki, M., Yokoi, N., Minami, K., Nakayama, Y., Harada, K., Hastoy, B., Wu, X.,
Takahashi, H., Kimura, K., Matsubara, T., Hoshikawa, R., Hatano, N., Sugawara, K., Shibasaki, T.,
Inagaki, N., Bamba, T., Mizoguchi, A., Fukusaki, E., Rorsman, P., and Seino, S. (2014) Glutamate acts
as a key signal linking glucose metabolism to incretin/cAMP action to amplify insulin secretion. Cell
reports 9, 661-673
32. Sener, A., and Malaisse, W. J. (2002) The stimulus-secretion coupling of amino acid-induced insulin
release. Insulinotropic action of L-alanine. Biochimica et biophysica acta 1573, 100-104
33. Yang, J., Dolinger, M., Ritaccio, G., Mazurkiewicz, J., Conti, D., Zhu, X., and Huang, Y. (2012) Leucine
stimulates insulin secretion via down-regulation of surface expression of adrenergic alpha2A
receptor through the mTOR (mammalian target of rapamycin) pathway: implication in new-onset
diabetes in renal transplantation. The Journal of biological chemistry 287, 24795-24806
28
Page 29 of 64 Diabetes
34. Sasaki, M., Fujimoto, S., Sato, Y., Nishi, Y., Mukai, E., Yamano, G., Sato, H., Tahara, Y., Ogura, K.,
Nagashima, K., and Inagaki, N. (2013) Reduction of reactive oxygen species ameliorates
metabolism-secretion coupling in islets of diabetic GK rats by suppressing lactate overproduction.
Diabetes 62, 1996-2003
35. Lacraz, G., Figeac, F., Movassat, J., Kassis, N., Coulaud, J., Galinier, A., Leloup, C., Bailbe, D.,
Homo-Delarche, F., and Portha, B. (2009) Diabetic beta-cells can achieve self-protection against
oxidative stress through an adaptive up-regulation of their antioxidant defenses. PloS one 4, e6500
36. Chavey, A., Bailbe, D., Maulny, L., Renard, J. P., Movassat, J., and Portha, B. (2013) A
euglycaemic/non-diabetic perinatal environment does not alleviate early beta cell maldevelopment
and type 2 diabetes risk in the GK/Par rat model. Diabetologia 56, 194-203
37. Wang, P., Fiaschi-Taesch, N. M., Vasavada, R. C., Scott, D. K., Garcia-Ocana, A., and Stewart, A. F.
(2015) Diabetes mellitus--advances and challenges in human beta-cell proliferation. Nature reviews.
Endocrinology 11, 201-212
38. Calderari, S., Irminger, J. C., Giroix, M. H., Ehses, J. A., Gangnerau, M. N., Coulaud, J., Rickenbach,
K., Gauguier, D., Halban, P., Serradas, P., and Homo-Delarche, F. (2014) Regenerating 1 and 3b gene
expression in the pancreas of type 2 diabetic Goto-Kakizaki (GK) rats. PloS one 9, e90045
39. Devi, T. S., Lee, I., Huttemann, M., Kumar, A., Nantwi, K. D., and Singh, L. P. (2012) TXNIP links innate
host defense mechanisms to oxidative stress and inflammation in retinal Muller glia under chronic
hyperglycemia: implications for diabetic retinopathy. Experimental diabetes research 2012, 438238
40. Ihara, Y., Toyokuni, S., Uchida, K., Odaka, H., Tanaka, T., Ikeda, H., Hiai, H., Seino, Y., and Yamada, Y.
(1999) Hyperglycemia causes oxidative stress in pancreatic beta-cells of GK rats, a model of type 2
diabetes. Diabetes 48, 927-932
29
Diabetes Page 30 of 64
41. Cantley, J., and Biden, T. J. (2013) Sweet and sour beta-cells: ROS and Hif1alpha induce
Warburg-like lactate production during type 2 diabetes. Diabetes 62, 1823-1825
42. Ehses, J. A., Lacraz, G., Giroix, M. H., Schmidlin, F., Coulaud, J., Kassis, N., Irminger, J. C., Kergoat,
M., Portha, B., Homo-Delarche, F., and Donath, M. Y. (2009) IL-1 antagonism reduces hyperglycemia
and tissue inflammation in the type 2 diabetic GK rat. Proceedings of the National Academy of
Sciences of the United States of America 106, 13998-14003
43. Yoon, K. H., Ko, S. H., Cho, J. H., Lee, J. M., Ahn, Y. B., Song, K. H., Yoo, S. J., Kang, M. I., Cha, B. Y.,
Lee, K. W., Son, H. Y., Kang, S. K., Kim, H. S., Lee, I. K., and Bonner-Weir, S. (2003) Selective beta-cell
loss and alpha-cell expansion in patients with type 2 diabetes mellitus in Korea. The Journal of
clinical endocrinology and metabolism 88, 2300-2308
44. Butler, A. E., Janson, J., Bonner-Weir, S., Ritzel, R., Rizza, R. A., and Butler, P. C. (2003) Beta-cell
deficit and increased beta-cell apoptosis in humans with type 2 diabetes. Diabetes 52, 102-110
45. Matveyenko, A. V., Veldhuis, J. D., and Butler, P. C. (2006) Mechanisms of impaired fasting glucose
and glucose intolerance induced by an approximate 50% pancreatectomy. Diabetes 55, 2347-2356
46. Noda, M., Yamashita, S., Takahashi, N., Eto, K., Shen, L. M., Izumi, K., Daniel, S., Tsubamoto, Y.,
Nemoto, T., Iino, M., Kasai, H., Sharp, G. W., and Kadowaki, T. (2002) Switch to anaerobic glucose
metabolism with NADH accumulation in the beta-cell model of mitochondrial diabetes.
Characteristics of betaHC9 cells deficient in mitochondrial DNA transcription. The Journal of
biological chemistry 277, 41817-41826
47. Prentki, M., Matschinsky, F. M., and Madiraju, S. R. (2013) Metabolic signaling in fuel-induced insulin
secretion. Cell metabolism 18, 162-185
48. Talero, E., Garcia-Maurino, S., Avila-Roman, J., Rodriguez-Luna, A., Alcaide, A., and Motilva, V. (2015)
30
Page 31 of 64 Diabetes
Bioactive Compounds Isolated from Microalgae in Chronic Inflammation and Cancer. Marine drugs
13, 6152-6209
49. Khadra, A., and Schnell, S. (2015) Development, growth and maintenance of beta-cell mass: models
are also part of the story. Molecular aspects of medicine 42, 78-90
31
Diabetes Page 32 of 64
Figure legends
Fig. 1 Transcriptomic and proteomic analysis of pancreatic islets in diabetic GK rats over time. (A) Experimental workflow. Pancreatic islets isolated from GK and control, age-matched WST rats of five different ages (4, 6, 8, 16 and 24 weeks) were subjected to MAPS-based RNA sequencing and TMT-labeling-based proteomics analysis. After performing data quality control and normalization, differentially expressed mRNAs and proteins were analyzed by ANOVA, followed by integrated bioinformatics analysis and biological validation. (B) Venn diagram of identifiable and quantifiable mRNAs and proteins in this study. (C, D) Unsupervised hierarchical clustering (C) and PCA (D) analysis of all quantifiable mRNAs and proteins from GK and WST islets indicated the reproducibility of biological replicates; however, islet gene expression in both GK and WST rats was highly variable between time points at both the mRNA and protein levels.
Fig. 2 Temporal gene expression patterns during GK diabetes progression. (A) Pearson's correlation analysis of the temporal mRNA and protein expression of 746 overlapping DE genes. In total, 76.8% of genes were positively correlated, of which 41.3% were significantly correlated (adjusted p value < 0.01).
The mean value of Pearson’s correlation coefficient was 0.388. (B) DAVID analysis of positively correlated DE genes revealed two enriched functional annotations primarily associated with carbon metabolism and ribosomes. No functional annotation was enriched for negatively correlated DE genes. (C) Time course dynamic
32
Page 33 of 64 Diabetes
expression clustering analysis of DE genes. First, fold-changes in DE genes were
transformed into Z-scores. Next, the k-means clustering method was utilized to
classify the genes into 12 mRNA and 9 protein patterns, displayed as a Circos figure.
Functional enrichment analyses of the genes within each pattern were carried out
using “EnrichmentMap” software. The enriched GO functional groups are selectively
highlighted with transparent pink (up-regulated) and blue (down-regulated) circles.
Fig. 3 Heatmap of KEGG pathway enrichment analysis. Normalized
counts/pseudocounts of the DE genes were subjected to GAGE analysis employing
the Bioconductor package “gage”. Pathways with adjusted p values (Benjamini &
Hochberg method) of less than 0.05 are indicated by “*”. The “Stat.mean” values
represent the averaged magnitude and direction of fold-changes at the gene-set level
corresponding to the color-coded up- (red) and down-regulated (blue) changes. KEGG
pathway maps were used to perform classifications.
Fig. 4 Mitochondrial signatures in GK islets. Based on the locations or
biological functions of GO annotations, the mitochondria-related DE genes were
mapped to the outer and inner membrane, membrane transporter, OXPHOS complex,
ribosomal proteins, protein quality control, transcription, translation, biogenesis and
antioxidant. The heatmaps for mRNA and protein expression at the five time points
are color-coded according to the log2-fold-changes for GK vs WST.
33
Diabetes Page 34 of 64
Fig. 5 Metabolic overview of GK islets. DE enzymes were mapped on the
KEGG global metabolism map. Metallic gold lines represent differentially expressed metabolic enzymes. Metabolic pathways are selectively highlighted with pink and green bold lines representing up- and down-regulation, respectively. The heatmaps for mRNA and protein expression at five time points are displayed and grouped as glycolysis, TCA cycle, fatty acid biosynthesis, fatty acid degradation, amino acid metabolism and glutathione metabolism.
Fig. 6 βββ-cell neogenesis defects in GK rats. The cartoon depicts an overview of pancreatic endocrine cell development, which was re-constructed by adapting a figure from a reference paper (49) with minor modifications. The master transcription factors within each type of cell are listed; up- and down-regulation are colored in red and green, respectively. Unless noted otherwise, gene expression line chart data are displayed as the mean ± SEM of repeated experiments. Adjusted ANOVA p values are labeled as ‘*’, ‘**’ and ‘***’, representing less than 1.0E-02, 1.0E-03, and 1.0E-04, respectively. “ns” indicates no significant difference between GK and WST.
Fig. 7 ROS and inflammation contribute to the pathogenesis of islet dysfunction in GK rats. The cartoon depicts ROS signaling flux as a core hub linking metabolic dysfunction with islet inflammation and fibrosis. The time course gene expression data for several key genes involved in the generation of ROS, antioxidants, inflammation and fibrosis are displayed as a line chart with the same
34
Page 35 of 64 Diabetes
figure annotations as described in Fig. 6.
35
Diabetes Page 36 of 64
Transcriptomic and proteomic analysis of pancreatic islets in siabetic GK rats over time.
139x110mm (300 x 300 DPI)
Page 37 of 64 Diabetes
Temporal gene expression patterns during GK diabetes progression.
195x225mm (300 x 300 DPI)
Diabetes Page 38 of 64
Heatmap of KEGG pathway enrichment analysis.
182x385mm (300 x 300 DPI)
Page 39 of 64 Diabetes
Mitochondrial signatures in GK islets.
111x76mm (300 x 300 DPI)
Diabetes Page 40 of 64
Metabolic overview of GK islets.
206x229mm (300 x 300 DPI)
Page 41 of 64 Diabetes
β cell neogenesis defects in GK rats.
207x234mm (300 x 300 DPI)
Diabetes Page 42 of 64
ROS and inflammation contribute to the pathogenesis of islet dysfunction in GK rats.
206x244mm (300 x 300 DPI)
Page 43 of 64 Diabetes
Supplementary figure legends:
Supplementary Fig. 1 Experimental design of proteomic analysis and
data processing. (A) The experimental design of the proteomic workflow,
including protein extraction from isolated GK and WST rat islets, tryptic digestion,
6 plex TMT peptide labeling, offline peptide separation via high pH reversed phase
chromatography, and repeated 1D LC MS/MS analysis of each fraction. (B) The
accumulation curve for quantifiable proteins with different PIF cut off values. The
vertical blue broken line indicates 50% PIF. (C) Box plotting comparisons of raw
mRNA counts and protein pseudocounts before and after RUV normalization. The
y axis indicates the log2 transformation of counts.
Supplementary Fig. 2 Extension of data analysis and morphological
analysis of islets. (A) Comparison of hierarchical clustering results between TMT
ratios and protein normalized pseudocounts. (B) Spearman correlation analysis
between mRNA normalized counts and protein normalized pseudocounts for each
sample. (C) Immunohistochemistry micrographs of GK and WST rat islet paraffin
sections co stained for insulin (INS, red) and glucagon (GCG, green) at different time
points. Scale bar, 50 m.
Supplementary Fig. 3 Transcriptome and proteome comparisons
between GK and WST rats at different time points. (A) Scatter plots of
quantifiable genes from the transcriptomic and proteomic data at each time point. Up Diabetes Page 44 of 64
and down regulated genes are highlighted in red and blue. (B) Bar plot of Pearson’s correlation coefficients from the transcriptome and proteome at each time point. (C)
Random selection of 6 genes for the validation of our omics data by qRT PCR methods (n = 3 5). The annotation of “*” is the same as described in Fig. 6. (D) The histogram of overlapping DE genes both in our mRNA dataset and recent publically available dataset of islets from normal and T2D individuals (Supplementary Table 1,
Ref. 10). The genes either up regulated or down regulated in two datasets are considered as consistent ones, otherwise as inconsistent ones. The line indicates the percentage of the consistent genes at five different time points.
Supplementary Fig. 4 Insulin secretion was impaired in GK βββ-cells over the duration of the evaluated time course. The cartoon depicts the primary signal flow of GSIS by pancreatic β cells, including glucose transport, glucose metabolism, Ca2+ flux, insulin secretion granule docking and release, and insulin biosynthesis. The time course gene expression data for several key genes are selectively displayed as a line chart with the same figure annotations as in Fig. 6.
Supplementary Fig. 5 Mitochondrial dysfunction in GK islets. (A) Box plot of log2 fold changes (GK vs WST) from 12 mitochondria encoded mRNAs at each time point. (B) The malate aspartate shuttle and glycerol phosphate shuttle were abnormal in GK rat islets. Got1, Got2, Mdh1, Mdh2 and Gpd2 were significantly down regulated during diabetes development in GK rats, resulting in an increased Page 45 of 64 Diabetes
NAD+/NADH ratio.
Supplementary Fig. 6 Defective βββ-cell proliferation in GK islets. (A)
Heatmap of the mRNA expression of 38 genes functionally related to the cell cycle.
The mRNA expression data for each gene were normalized to a range of 0 1 via the
min max normalization method. (B) Box plot of the mRNA expression data in (A). A
two tailed Student’s t test was applied to the pairwise comparisons, revealing
significantly different mRNA expression levels between GK and WST islets at 4, 6,
and 8 weeks. (C) The β cell proliferation rate was measured based on Ki67 (cell
proliferation marker) and insulin staining. All data are presented as the mean ± SEM.
“*” indicates p < 0.05. n = 3 independent experiments.
Supplementary Fig. 7 NAC treatment experiment. (A-D) Body weights (A),
random blood glucose (B), GTT (C) and ITT blood glucose (D) were measured (n =
8). (E) After NAC treatment for 2 weeks, GSIS of isolated islets in GK NAC and
GK control was evaluated by calculating the fold change of secreted insulin amount
under high (16.7 mM) and low (2.8 mM) glucose stimuli (n=7). All data are presented
as the mean ± SEM. “*” indicates p < 0.05. “**” indicates p < 0.01. “***” indicates p
< 0.001. (F) Selected plotting of enriched KEGG pathways analyzed by GAGE
comparing the transcriptome between GK NAC and GK control (q < 0.05). The
“Stat.mean” values represent the averaged magnitude and direction of fold changes at
the gene set level. Diabetes Page 46 of 64
Supplementary Fig. 8 Time course of pancreatic islet deterioration in GK rats. The early stage (4 6 weeks) is characterized by up regulated anaerobic glycolysis, inflammation priming, down regulation of the TCA cycle, defective insulin secretion due to impaired mitochondrial metabolism, and compensation for insulin synthesis and OXPHOS gene expression at the transcription level. The late stage (8 24 weeks) is characterized by inflammation amplification, reduction in insulin production and compensation failure. No significant apoptosis occurred during the early stage, but apoptosis was elevated at 24 weeks at the mRNA level.
Neogenesis and cell proliferation were impaired starting during the early stage, resulting in β cell mass reduction.
Page 47 of 64 Diabetes
Table Legends *
Supplementary Table 1: The reported transcriptomes and proteomes of islets from various T2D animal models and human cadaver donors. Supplementary Table 2: All the information about identification and quantification of mRNAs and proteins. Supplementary Table 3: Pearson's correlation analysis of the temporal mRNA and protein expression of 746 overlapping DE genes, related to Figure 2A. Supplementary Table 4: Time course dynamic expression clustering analysis of DE genes, related to Figure 2C. Supplementary Table 5: KEGG pathway enrichment analysis. Supplementary Table 6: Unsupervised hierarchical clustering analysis of mitochondria related genes Supplementary Table 7: DE genes that were identified in NAC experiment by RNASeq, and the result of KEGG pathway enrichment analysis.
* All the tables can be accessed via the website of http://www.ibp.cas.cn/ibplwfjtj/xutao/201611/t20161117_4697642.html. Diabetes Page 48 of 64
Experimental design of proteomic analysis and data processing.
194x224mm (300 x 300 DPI)
Page 49 of 64 Diabetes
Extension of data analysis and morphological analysis of islets.
225x297mm (300 x 300 DPI)
Diabetes Page 50 of 64
Transcriptome and proteome comparisons between GK and WST rats at different time points.
158x141mm (300 x 300 DPI)
Page 51 of 64 Diabetes
Insulin secretion was impaired in GK β cells over the duration of the evaluated time course.
231x308mm (300 x 300 DPI)
Diabetes Page 52 of 64
Mitochondrial dysfunction in GK islets.
194x215mm (300 x 300 DPI)
Page 53 of 64 Diabetes
Mitochondrial dysfunction in GK islets.
138x112mm (300 x 300 DPI)
Diabetes Page 54 of 64
NAC treatment experiment.
203x243mm (300 x 300 DPI)
Page 55 of 64 Diabetes
Time course of pancreatic islet deterioration in GK rats
108x69mm (300 x 300 DPI)
Diabetes Page 56 of 64
Supplemental Experimental Procedures
Preparation of pancreatic islets from GK and Wistar rats
Pancreatic islets from male diabetic GK and control Wistar rats were isolated by collagenase digestion. Briefly, pancreases were inflated by instilling 5 ml of 0.5 mg/ml collagenase P (Roche Applied Science, Mannheim, Germany) in Hanks’ buffered saline solution (HBSS) through the pancreatic duct, dissected out, and incubated in a water bath at 37 °C for 20 min. The digested pancreases were rinsed with HBSS, and islets were separated on a Ficoll density gradient. After three washes with HBSS, islets were carefully hand-picked under a dissection microscope to remove extra-islet tissues as much as possible. The purity of islets was evaluated by checking the expression of Amy1a (a marker of acinar cells) and Krt19 (a marker of ductal cells). In our omics dataset, Amy1a was slightly higher at protein level (only at
16 weeks) in GK islet, but not statistically different at mRNA level at any time points.
For Krt19, it was only detected at mRNA level with no statistical difference. This indicates that the contaminants in islets were fairly minimized.
Random blood glucose assay
Random-fed blood glucose levels were measured weekly from the tail vein with an
Accu-Chek glucometer (Roche Diagnostics, Mannheim, Germany).
Glucose tolerance test (GTT)
Rats were fasted overnight (14 h) prior to the administration of glucose (1.5 g/kg body weight) by oral gavage. Glucose measurements were taken at 0, 15, 30, 60, 90 and
120 min post-administration using an Accu-Chek glucometer (Roche Diagnostics,
Mannheim, Germany). Blood was collected from the tail vein at each time point during the glucose tolerance test. Rats were denied access to food during the study.
Insulin tolerance test (ITT) Page 57 of 64 Diabetes
After 6-h fasting, rats were injected with regular human insulin (NovolinR, Novo
Nordisk) at a dose of 0.75 U/kg of body weight. Blood samples were collected from
the tail vein at 0, 15, 30, 60, 90 and 120 min post-injection to determine blood glucose
concentrations.
Glucose-stimulated insulin secretion (GSIS)
The insulin secretion induced by glucose in primary islets was assessed as previously
described (1). The primary islets of rats were isolated through collagenase (Roch)
perfusion method. The islets were cultured in RPMI160 medium containing 100 U/ml
penicillin, 100 µg/ml streptomycin and 10% fetal bovine serum overnight, and then
incubated for 1 h in Krebs-Ringer buffer (KRB) containing 2.8 mM and 16.7 mM
glucose serially. The insulin concentration in the collected supernatant was measured
using an insulin ELISA kit (Millipore). The amount of protein was determined using a
BCA kit.
RNA-Seq library preparation, sequencing and data processing
Isolated islets (the number was around 100-150 at 4-6 weeks and 100-120 after 8
weeks) were immediately placed in TRIzol reagent (Invitrogen, cat. no. 15596018).
RNA was extracted by adding chloroform, followed by isopropanol precipitation. The
RNA pellet was washed twice with 70% ethanol, dried, and resolved in RNase-free
water. A260/A280 and 28s/18s rRNA ratios were determined to ensure that the RNA
samples were highly purified and not degraded.
The MAPS (Multiplex Analysis of PolyA-linked Sequences) protocol has been
published previously (3). Briefly, TRIzol-isolated RNA (1 µg) was enriched for
Poly(A+) RNA with biotinylated oligo(dT) and converted to cDNA with Superscript
III (Invitrogen, cat. no. 18080-051). Next, a terminal transferase (NEB, cat. no.
M0315S) was used to add a ddNTP to the 3’-end of the cDNA. After second-strand Diabetes Page 58 of 64
cDNA synthesis, 23 PCR cycles were performed to amplify the cDNA. PCR products in the range of 200-400 nt were subjected to deep sequencing on a HiSeq 2500
(Illumina).
The analysis was performed using scripts generated in-house. Fastq files were aligned
(4) to merged transcript sequences obtained from the UCSC RefGene database (5) with the mapping program Bowtie (v1.1.1) using the following parameter: “-l25 -n2
-M3 -e 200 --best --strata --phred33-quals --trim5 4”. The unmapped tags in the first round were then mapped to the rat genome (rn6). For each gene, only tags that were uniquely mapped and localized in exons or exon-exon junctions were counted.
Due to limit amount of islets from one animal, the islets for transcriptomics and proteomics (see below) were not from the same animal, but from the same offspring to minimize inter-animal variations.
Protein sample preparation for proteome analysis
Rat islets (the number was 150-200 at 4-6 weeks and around 100-150 after 8 weeks) were lysed in a buffer containing 8 M urea/100 mM NH4HCO3 (Sigma, cat. no.
A6141), pH 8.3, and a protease inhibitor cocktail (Sigma, cat. no. P8340-1ML).
Proteins (50 µg) were first reduced by adding a final concentration of 10.0 mM dithiothreitol for 30.0 min at 37 ºC and were then immediately alkylated by incubating with a final concentration of 20 mM iodoacetamide for 45 min at 37 ºC in the dark. Resulting mixtures were diluted to less than 2 M urea with 100 mM
NH4HCO3 prior to digestion overnight with sequencing-grade trypsin (Promega, cat. no. V5111) at a substrate/enzyme ratio of 100:1 (w:w). Digestion was quenched via acidification with formic acid (FA). Peptide desalting was subsequently performed via solid phase extraction (Sep-pack Vac C18 cartridges, Waters, cat. no. Wat020515), followed by vacuum-drying. Page 59 of 64 Diabetes
The resulting peptide samples were re-suspended in 100 mM triethylammonium
bicarbonate (TEAB) buffer. An aliquot of 50% of each sample was chemically labeled
with TMT reagents (Thermo Fisher, cat. no. 90066) according to the manufacturer’s
instructions. To accommodate all 10 samples in the analysis of a single biological
replicate, we created a reference sample by mixing 1/5 of the amount of each sample
(10 in total) and then labeling with TMT reagent 131. The remainder of the TMT
reagents were used to label the other samples: 126 for GK, WST_4w, 127 for GK,
WST_6w, 128 for GK, WST_8w, 129 for GK, WST_16w and 130 for GK, WST_24w.
The reference sample was used for data normalization and dataset combination. To
ensure that each of the samples contained the same amount of protein, a small
1:1:1:1:1:1 aliquot was prepared and analyzed by MS. Summed reporter ion ratios
informed mixing ratios of the remaining labeled digests. In this way, two sets of
6-plex TMT-labeled peptide mixtures were equally pooled and were treated in parallel
throughout the following steps.
Prior to MS analysis, TMT-labeled peptide mixtures were fractionated using high-pH
reversed-phase chromatography. Briefly, the samples were first desalted using
Sep-Pak Vac C18 SPE cartridges (Waters, Massachusetts, USA) and dried in a
vacuum concentrator. Desalted peptides were dissolved in solution A (2% Acetonitrile,
pH 10, pH adjusted with ammonium hydroxide) and were then loaded automatically
onto a YMC-Triart C18 basic reversed-phase liquid chromatography column (250 ×
4.6 mm, 5 µm particles) (cat. no. TA12S05-2546WT, YMC, Kyoto, Japan). Peptides
were separated in a binary buffer system of solution A and solution B (98%
Acetonitrile, without pH adjustment, solution B) in an Ultimate-3000 LC system
(Thermo Scientific, Massachusetts, USA). The gradient of buffer B was set as follows:
0-5% for 1 min, 5-15% for 5 min, 15-26% for 32 min, 26-40% for 22 min, 40-95% Diabetes Page 60 of 64
for 2 min, and 95% for 3 min. The first fraction was collected starting at 5 min, and the remaining eluates were collected at intervals of 84 s. A total of 45 fractions were obtained, and these were concatenated into 15 fractions by merging fractions 1, 16, 31; fractions 2, 17, 32; and so on. Then, all 15 peptide fractions were dried in a vacuum concentrator and stored at -20 °C until subsequent nanoLC-MS/MS analysis was performed.
NanoLC-MS/MS analysis
NanoLC-MS/MS experiments were performed on a Q Exactive mass spectrometer
(Thermo Scientific) coupled to an Easy-nLC 1000 HPLC system (Thermo Scientific).
The peptides were loaded onto a 100-µm id × 2-cm fused silica trap column packed in-house with reversed-phase silica (Reprosil-Pur C18 AQ, 5 µm, Dr. Maisch GmbH) and then separated on a 75-µm id × 20-cm C18 column packed with reversed-phase silica (Reprosil-Pur C18 AQ, 3 µm, Dr. Maisch GmbH). The loaded peptides were eluted with a 78-min gradient. Solvent A consisted of 0.1% FA in water solution, and solvent B consisted of 0.1% FA in acetonitrile solution. The following segmented gradient was used at a flow rate of 280 nl/min: 4-12% B, 5 min; 12-22% B, 50 min;
22-32% B, 12 min; 32-95% B, 1 min; and 95% B, 7 min.
The mass spectrometer was operated in data-dependent acquisition mode, and full-scan MS data were acquired in the Orbitrap with a resolution of 70,000 (m/z 200) across a mass range of 300-1600 m/z. The target value was 3.00E+06 with a maximum injection time of 60 ms. After the survey scans, the top 20 most intense precursor ions were selected for MS/MS fragmentation with an isolation width of 2 m/z in the HCD collision cell and an optimized normalized collision energy of 32%.
Subsequently, MS/MS spectra were acquired in the Orbitrap with a resolution of
17,500 (m/z 200) and a low-mass cut-off setting of 100 m/z. The target value was set Page 61 of 64 Diabetes
as 5.00E+04 with a maximum injection time of 80 ms. The dynamic exclusion time
was 50 s. The following nanoelectrospray ion source settings were used: spray voltage
of 2.0 kV, no sheath gas flow, and a heated capillary temperature of 320 °C. Each
fraction was repeatedly analyzed.
MS data processing
The raw MS data were processed with Proteome Discovery (version 1.4, Thermo
Scientific). Briefly, peptide identification was performed with both Sequest HT and
the Mascot 2.3 search engine comparing against a UniProt database (version 5.62)
supplemented with all frequently observed MS contaminants. The following
parameters were used for database searching: 10 ppm precursor mass tolerance, 0.02
Da fragment ion tolerance, up to two missed cleavages, carbamidomethyl cysteine,
TMT modification on amino (N)-term and lysine as fixed modifications, and oxidized
methionine as a variable modification. The peptide confidence was set to a high level
(q-value < 0.01) for peptide filtering.
To improve the accuracy and confidence of protein quantification, optimized data
processing was developed using freely accessible tools and in-house written scripts
(available upon request). 1) msconvert (http://proteowizard.sourceforge.net) was first
used to perform a deconvolution of the high-resolution MS2 spectra, in which all
fragment ion isotopic distributions were converted to an m/z value corresponding to
the monoisotopic single charge. The signals of TMT reporter ions were extracted with
the following requirements: maximum mass accuracy of 15 ppm, detection of all 6
TMT reporter ion channels required. 2) The summed reported ion intensity from each
channel for all acquired MS2 spectra was used for sample normalization. The
intensities of 131 channels representing reference samples in two 6-plex TMT
experiments were utilized as conjunct factors for data normalization such that Diabetes Page 62 of 64
comparison between different time point samples from GK and WST rats achieved the same level. 3) To minimize ratio distortion due to the presence of more than one peptide species within a precursor ion isolation width, we also rejected the quantification of MS/MS spectra based on the precursor intensity fraction (PIF). For our dataset, a PIF of 50% was selected as the optimal trade-off value for both identification and quantification. 4) If a specific peptide was quantified multiple times, the peptide with the lowest PIF was selected to produce the representative quantitative
TMT ratio. The median values of the TMT ratios of peptides from the same protein were calculated as the protein ratios. 5) To maintain experimental design power for additional statistical analyses, the protein TMT ratio was further processed with the following step. The normalized spectral abundance factor (NSAF) for each protein was calculated using an in-house Perl script. After multiplying by an amplification factor (1.0E+06 was adopted in this study), NSAF values were transformed into natural integer numbers, which were considered summed 6-sample pseudo-abundances that were distributed to each sample based on individual TMT ratios, hence the name “pseudocount”. ANOVA was adopted to analyze differentially expressed proteins between GK and WST islets.
Hierarchical clustering and principal component analysis (PCA)
To assess the similarity among all samples using the transcriptomic and proteomic datasets, unsupervised hierarchical clustering was performed using the “pheatmap” package in R. PCA analysis was performed using the built-in R function “prcomp()”.
Differential gene expression analysis
Both mRNA raw counts and protein pseudocounts were normalized by implementing the RUV (remove unwanted variation) approach. Differential gene expression analysis was carried out using the two-way ANOVA method with the built-in R Page 63 of 64 Diabetes
function “anova()”. Two variables, time (5 different time points) and type (GK and
WST), were taken into consideration, and genes with FDRs (adjusted p value with
“Benjamini & Hochberg” method) of less than 0.01 were considered differentially
expressed genes.
Gene ontology (GO) enrichment analysis
The NCBI ENTREZ gene ID list was processed in batches using the DAVID Web
Service API Perl Client to identify enriched Gene Ontology categories. Rattus
norvegicus was set as the background. KEGG category terms with FDRs of less than
0.05 were considered significant hits.
Dynamic gene expression patterns
To investigate time course patterns of gene expression at both the mRNA and protein
levels, we first calculated the log2 ratios of all differentially expressed mRNAs and
proteins. To maintain the expression scale, we created ten series of values, including
five ‘0’ values (4, 6, 8, 16, and 24 weeks, representing expression values in WST rats)
and the log2 ratios of GK/WST (4, 6, 8, 16, 24 weeks, representing expression values
in GK rats), and then performed data scaling via Z-transformation for the following
clustering analysis.
Different expression patterns of mRNAs and proteins were clustered using the built-in
R function “kmeans()”. The optimal results for clustering analysis were obtained by
iteratively performing k-means clustering with a range of cluster numbers, followed
by DAVID functional annotation analysis to assess their biological relevance.
KEGG signaling pathway analysis
The Generally Applicable Gene-set Enrichment (GAGE) package in R was utilized to
perform pathway analysis. GAGE was generally applicable to gene expression
datasets with different sample sizes and experimental designs. The KEGG Pathway Diabetes Page 64 of 64
Database was used as a reference. Pathways with q-values of less than 0.05 were considered significantly different between GK and WST rats.
Immunohistochemistry, antibodies and microscopy
Pancreases from 4- to 24-week-old rats were fixed in 4% paraformaldehyde and were embedded in paraffin. The primary (i) and secondary (ii) antibodies (Ab) used in this study were (i) guinea pig anti-insulin (prepared by our lab, 1:200), mouse anti-glucagon (Abcam, ab10988, 1:200), and rabbit anti-Ki67 (Abcam, ab15580,
1:1000 dilution) and (ii) goat anti-guinea pig IgG Rhodamine (Jackson, 1:200), donkey anti-rabbit IgG Alexa Fluro 488 (Invitrogen, 1:200), donkey anti-mouse IgG
DyLight 488 (Jackson, 1:200). Fluorescent images were acquired with a confocal laser scanning microscope (Olympus, FV 1200) operated in multitrack mode using the following objective: UPLSAPO40X2 40×/0.95. The microscope system was drived by Olymbus Fluoview Ver.2.0a software.
1. Lei, L., Liu, Q., Liu, S., Huan, Y., Sun, S., Chen, Z., Li, L., Feng, Z., Li, Y., and Shen, Z. (2015) Antidiabetic potential of a novel dual-target activator of glucokinase and peroxisome proliferator activated receptor-gamma. Metabolism: clinical and experimental 64, 1250-1261 2. Avall, K., Ali, Y., Leibiger, I. B., Leibiger, B., Moede, T., Paschen, M., Dicker, A., Dare, E., Kohler, M., Ilegems, E., Abdulreda, M. H., Graham, M., Crooke, R. M., Tay, V. S., Refai, E., Nilsson, S. K., Jacob, S., Selander, L., Berggren, P. O., and Juntti-Berggren, L. (2015) Apolipoprotein CIII links islet insulin resistance to beta-cell failure in diabetes. Proceedings of the National Academy of Sciences of the United States of America 112, E2611-2619 3. Fox-Walsh, K., Davis-Turak, J., Zhou, Y., Li, H., and Fu, X. D. (2011) A multiplex RNA-seq strategy to profile poly(A+) RNA: application to analysis of transcription response and 3' end formation. Genomics 98, 266-271 4. Langmead, B., Trapnell, C., Pop, M., and Salzberg, S. L. (2009) Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome biology 10, R25 5. Karolchik, D., Hinrichs, A. S., Furey, T. S., Roskin, K. M., Sugnet, C. W., Haussler, D., and Kent, W. J. (2004) The UCSC Table Browser data retrieval tool. Nucleic acids research 32, D493-496