Pancreas specimens from ten non-diabetic and ten T2D subjects were studied. Their clinical characteristics can be found in .
Clinical characteristics of type 2 diabetic and control subjects.
Approach to LCM and islet anatomy
It is important to point out that LCM was used selectively in that islets were chosen that had relatively well defined patches of beta-cells with endogenous immunofluorescence. It was important to do this because human islets are known have clumps of α and δ cells in the islet core 
, which means that great care must be taken with LCM to avoid the non beta-cells. Thus, there are some populations of beta-cells that cannot be collected for geographical reasons. We expect that these selected beta-cells are representative of beta-cells throughout the pancreas but must be cautious about this assumption. It cannot be excluded that beta-cells in islets of different size and locations may be subjected to different paracrine and environmental influences that could alter gene expression. Another point is that the endogenous immunofluorescence of the beta-cells may emanate from lipofuscin, which is thought to accumulate as beta-cells age 
. Thus, it is possible that the beta-cells collected with LCM are older than the overall population. However they would be comparable in the two groups since no differences in the intensity of immunofluoresence could be discerned between beta-cells of control versus T2D pancreases.
The 10 T2D pancreases had the range of pathology expected for T2D. Amyloid could be found in some islets in most of the pancreases, as was expected from previous observations 
. Moreover, much islet variation was seen, which included islets with fibrosis, islets with high non-beta-/beta-cell ratios, and a range of islet size. Extracellular lipid deposits were not particularly noticeable. We did not stain for alpha-cells, so we do not know the alpha-cell/beta-cell ratio specifically, but we had the impression that non-beta-cell/beta-cell ratio is up, as is well described by many including MacLean and Ogilvie in 1955 
and then more recently by Clark et al 
and Yoon et al 
. As was also done for the control pancreases, relatively well defined patches of beta cells were chosen for LCM. Thus, for the T2D pancreases, this means that care was taken to avoid islets with amyloid and/or fibrosis.
Expression profiles in samples obtained from T2D donors compared to control donors
By using the lower confidence bound (LCB) parameter at the cutoff of 1.2, DNA-Chip Analyzer (dChip) software identified 2,062 differentially expressed probe sets, corresponding to 1,920 transcripts. Among them, 1,203 transcripts were upregulated and 717 were downregulated in samples obtained from T2D subjects as compared to non-diabetic controls. Using the p value parameter, 7,164 probe sets, corresponding to 6,384 transcripts were differentially expressed at p<0.05 (3,464 transcripts upregulated and 2,920 downregulated), whereas 2,133 probe sets, corresponding to 1,870 transcripts were different at p<0.01 (1006 upregulated and 864 downregulated). Hierarchical clustering analysis of differentially expressed genes showed a clear separation between diabetic and non-diabetic samples ().
Hierarchical clustering of differentially expressed genes.
As reported in , the Principal Component Analysis showed that samples run at the Genomic Core facilities of the Joslin Diabetes Center and the Massachusetts General Hospital, at two different times, clustered in two different groups. All data are MIAME compliant and the row data have been deposited in a MIAME compliant database (GEO, accession numbers: GSE20966).
Approach to analysis
Most of the analysis was done with dChip software. We first looked at groups of genes thought to be important for the beta-cell abnormalities known to occur in T2D. They were grouped into categories of function. Other genes were included because they seemed to be good candidates to be important; some were differentially expressed, others were not. Clusters of genes with common function were assessed with GSEA. In assessing the signal strength of arrays as determined by dChip, the strength can be a general indicator of the abundance of the mRNA, but caution must be used for interpretation. Two major variables: the amplification process will change the abundance of the transcripts and low signal strength may results from inefficient probes. We know that some of the high strength signals are incorrect. For example, the array signal strength for insulin and glucagon was similarly high, yet when unamplified mRNA obtained with LCM was assessed with quantitative PCR, the insulin transcript levels were far higher than glucagon 
. Nonetheless, the signals are very consistent for comparing groups such as samples from T2D versus controls. Moreover, the general strength of the signals provides insights into the relative strength of various isoforms. For example creatine kinase, brain (CKB), which may be important for glucose stimulated insulin secretion (GSIS), was highly expressed while the muscle form (CKM) was hardly expressed at all (Table S1
Islet hormones and receptors
The array signals for insulin (INS), glucagon (GCG), somatostatin (SST) and IAPP were all very high and only IAPP was differentially expressed, downwards (Table S2
). These signals clearly do not provide accurate information about the actual abundance of message, as we have shown previously using PCR on samples collected by LCM 
. It seems this is particularly a problem when the message is very highly expressed. Nonetheless, the reduced expression of IAPP is not surprising because similar reduction has been found in rodents with diabetes, which is probably a result of glucose toxicity 
. No differential expression was found for either the receptors for glucagon (GCGR) or GLP1 (GLP1R); the signals for the gastric inhibitory polypeptide receptor (GIPR) were very weak.
Islet transcription factors
In our previous studies with the rat partial pancreatectomy model, with multiplex PCR on isolated islets we found downregulation of pdx1, nkx6.1, pax6, neurod and hnf1alpha 
. In the present study there was down regulation of HNF1A (TCF1) but no change in the expression of NKX6-1 or NEUROD1 (Table S2
). PAX6 was upregulated. Unexpectedly, PDX1 was upregulated but the signal strength was low compared to other transcription factors. This might be due to low amplification efficiency of PDX1 transcript or inefficient binding of the transcript to the corresponding probe set. Other transcription factors not changed included NKX2-2, PAX4, HNF4A, neurogenin 3 (NEUROG3), and MAFB. PAX6 and ARX, which are normally found in alpha-cells 
, were upregulated, which raises the question as to whether the beta-cells of T2D developed some alpha-cell characteristics or whether the LCM of the T2D tissue somehow collected more alpha-cell tissue. It is possible that an increase in the ratio of alpha-cells to beta-cells might have made it more likely that a few more alpha-cells were contained in the material captured from the T2D pancreases. However, great care was taken to restrict the LCM to fluorescent beta-cells. If there were some inadvertent inclusions, one might have expected variability of PAX6 and ARX expression among the T2D pancreases, but the consistency was similar to that seen in the controls. Another possibility is that beta-cells in T2D express alpha-cell genes more highly. Others and we have found that some single beta-cells obtained by flow cytometry have a low level of glucagon gene expression 
. This becomes especially interesting with the recent report by the group of Herrera that alpha-cells may be able to transdifferentiate into beta-cells during regeneration 
Because beta-cells in a diabetic environment have profoundly abnormal GSIS, components of glucose metabolism pathways that are linked to secretion can be expected to be abnormal in T2D. The expression results of noteworthy genes are shown in Table S1
. With regard to glucose transport, GLUT2 was downregulated in the T2D sample while GLUT1 was unchanged. This is of interest because while GLUT2 is thought to be important for rodents, evidence suggests that GLUT1 is largely responsible for glucose transport in human beta-cells 
. For the glycolytic enzymes, no change was seen in the expression of glucokinase (GCK). Three forms of phosphofructokinase (PFKM, P and L) were highly but not differentially expressed. Aldolase B (ALDOB) was highly expressed and upregulated while enolase 1, alpha (ENO1) was modestly expressed and downregulated. Lactate dehydrogenase A (LDHA), which normally has very low expression in beta-cells 
was upregulated. Pyruvate carboxylase (PC), which is highly expressed in rodent islets 
, had low signal strength and no differential expression.
For gluconeogenesis, which is thought to be negligible in normal beta-cells, phosphoenolpyruvate carboxykinase 1 (PCK1) was upregulated, while the Fructose-1, 6-bisphosphatase (FBP2) and glucose-6-phosphatase (G6PC and G6PC3) enzymes genes had low signal strength and no differential changes. The mitochondrial enzyme dihydrolipoamide dehydrogenase (DLD), which is the E3 component of pyruvate dehydrogenase complex, was downregulated. Of enzymes of the tricarboxylic acid cycle, the only potentially interesting changes were a modest reduction in aconitase 2 (ACO2) and an increase of mitochondrial succinate dehydrogenase complex, subunit C (SDHC). There was no differential expression of the uncoupling protein genes 1, 2, or 3. Notable changes in mitochondrial shuttle genes were down regulation of both mitochondrial glycerol-3-phosphate dehydrogenase 2 (GPD2) and cytosolic malic enzyme 1 (ME1). ATP citrate lyase (ACLY), a cytosolic gene important for the production of acetyl CoA was clearly downregulated. These enzymes are thought to be important for GSIS 
MacDonald et al recently reported reduced expression of GPD2 and ACLY in isolated islets from subjects with T2D 
, which agrees with our findings. They also found reductions in isocitrate dehydrogenase 1 (NADP+), soluble (IDH1), propionyl Coenzyme A carboxylase, beta polypeptide (PCCB), malate dehydrogenase 2, NAD (mitochondrial, MDH2) and 3-oxoacid CoA transferase 1 (OXCT1). None of these changes were confirmed in our study.
These data may provide some clues to the mechanisms responsible for glucotoxicity. We have extensively studied changes of islet gene expression 4 wk after 90% pancreatectomy in rats and found marked changes in glucose metabolism genes 
. The general trend was that genes highly expressed in beta-cells such as GLUT2 and GCK were downregulated while normally suppressed genes such as LDHA were upregulated. There was only some agreement between these and the present data. The combined rodent and human data provide a valuable partial picture, but the ideal study remains to be done to better understand beta-cells in T2D. Results obtained with isolated islets from cadaver donors have their limitations. LCM is an important advance, but the amplification step brings potential artifact and arrays of the future will provide better information. Other weaknesses of the present approach are the serious illness of the donors prior to death and the pancreases being subjected to hours of cold ischemia time. We tried to exclude a history of premorbid hyperglycemia in the nondiabetic controls, but with the use of steroids, and the presence of brain damage and stress, there may been periods of hyperglycemia that could have caused some glucotoxicity changes. A goal for the future should be to obtain pancreatic tissue from well-characterized patients at surgery. Frozen sections can be subjected to LCM, and enough RNA should be obtained to avoid the amplification step.
With regard to lipid metabolism (Table S1
), fatty acid synthase (FASN) expression was modestly upregulated. It is interesting that L-3-hydroxyacyl-Coenzyme A dehydrogenase (HADH), short chain, which is part of the beta oxidation pathway in mitochondria, was downregulated because mutations in this gene have been linked to hyperinsulinemic hypoglycemia 
. Another downregulated gene of potential interest was stearoyl-CoA desaturase (SCD), which contrasts with the increase found in Zucker fatty (ZF) rats after partial pancreatectomy 
Channels and adrenergic receptor
In the evaluation of channels important for GSIS (Table S1
), KIR6.2 (KCNJ11) had weak signal strength and no differential expression, but SUR1 was highly expressed and downregulated in diabetes. The Calcium channel, voltage-dependent, L type, alpha 1D subunit (CACNA1D) was also downregulated. Inositol 1,4,5-triphosphate receptor, type 3 (ITPR3), which could be important for inducing calcium release, was downregulated. Potassium voltage-gated channel, KQT-like subfamily, member 1 (KCNQ1), which acts to repolarize cells, was clearly upregulated in the samples from diabetic subjects. Another change that could influence secretion is downregulation of the adrenergic, alpha-2A-, receptor (ADRA2A) that mediates the adrenergic suppression of insulin secretion 
Reports of downregulation of exocytosis molecules and gene expression in islets from human T2D and GK rats, include syntaxin-1A, SNAP-25, VAMP-2, nSec1 (Munc18), Munc 13-1, synaptotagmin V and synaptophysin 
. However, in our study (Table S2
), syntaxin 1A (STX1A), synaptobrevin 2 (VAMP2), cellubrevin (VAMP3), Synaptophysin (SYP) and Syntaxin binding protein 1 (STXBP1, Munc18-1) had high signal strength but no differential expression. Synaptobrevin 1 (VAMP1) had much lower signal strength. Synaptosomal-associated protein, 25 kDa (SNAP25) was modestly downregulated and Unc-13 homolog B (C. elegans) (UNC13B) was clearly downregulated. Synaptotagmin V (SYT5) had little signal strength but Synaptotagmin IV (SYT4) was highly expressed. Both N-ethylmaleimide-sensitive factor (NSF) and N-ethylmaleimide-sensitive factor attachment protein, alpha SNAP (NAPA) were highly, but not differentially, expressed.
In contrast to the findings of Gunton et al that were performed with isolated islet preparations of variable purity from subjects with T2D 
, we found increased expression of the insulin receptor (INSR) and the insulin-like growth factor receptor (IGF1R) in the beta-cell tissue from T2D (Table S1
). Increased expression was also found for inositol polyphosphate-5-phosphatase (INPP5D, SHIP1) and inositol polyphosphate phosphatase-like 1 (INPPL1, SHIP2). Other upregulated genes included v-Ha-ras Harvey rat sarcoma viral oncogene homolog (HRAS) and v-raf-1 murine leukemia viral oncogene homolog 1 (RAF1). Of note, Forkhead box 01 (FOXO1) was downregulated. The AKTs, 1, 2 and 3 all had strong but no differential expression.
ARNT, ARNT2 and HNF-1alpha
A major decrease in the expression of aryl hydrocarbon receptor nuclear translocator (ARNT) was reported in isolated islets of T2D subjects by Gunton et al 
. This decrease was accompanied by reductions in HIF 1 alpha (HNF1A), the insulin receptor (INSR), several glycolytic enzymes including glucose-6-phospate isomerase (GPI), phosphofructokinase (PFK), aldolase (ALDO), and phosphoglucomutase (PGM). We had good to excellent signal strength for all of these genes but found no differential expression, except for an increase in ALDOB (Table S1
). Gunton et al also reported a decrease in HNF4 alpha, but our signal strength for this gene was too low to be meaningful.
Cell cycle genes
There were few changes of significance in the expression of cell cycle genes (Table S2
). Of the cyclins, cyclin D2 (CCND2) and D3 (CCND3) were the most highly expressed, but there was no differential expression. Of the cyclin-dependent kinases, CDK4 and CDK6 were more highly expressed than CDK2. Of the cell cycle inhibitors, only two tended to be upregulated, p16 (INK4a), p<0.06, and p21 (Cip1). There were no changes of the E2F family transcription factors. Retinoblastoma 1 (RB1) and RBL2: retinoblastoma-like 2 (RBL2, p130) were well expressed but not differentially. The signal for retinoblastoma-like 1 (RBL1, p107) was very weak. Of interest, the CDK5 regulatory subunit associated proteins (CDK5RAP) 1, 2 and 3 were all differentially upregulated. They have been associated with neuronal development and spindle check point function 
. Because so much of cell cycle control is exerted at the posttranslational level, these few modest changes in gene expression must not be over interpreted. The suggestion of change for p16 and p21 need to followed up; p16 in particular appears to be an important marker of aged cells that have restricted capacity for regeneration 
. There has been interest in the epigenetic control of replication 
, but there was no differential expression of the following candidate genes EZH2, MLL, PRC1, JMJD3, or BMI1.
The ER stress genes were notable for their lack of change (Table S3
). The few changes of interest included some mannosidases; isoforms MAN1B1 and MAN2B2 were upregulated. ER degradation enhancer, mannosidase alpha-like 3 (EDEM3) was downregulated. Endoplasmic oxidoreductin-1-like protein B (ERO1LB) was highly expressed and downregulated. Endoplasmic reticulum protein 27 kDa (ERP27), a noncatalytic member of the protein disulfide isomerase, was modestly upregulated as were DnaJ (Hsp40) homolog, subfamily C, member 3 (DNAJC3) and protein disulfide isomerase family A, member 4 (PDIA4).
Other genes that were highly expressed and had no change included: X-box binding protein 1 (XPB1), activating transcription factor 6 (ATF6), eukaryotic translation initiation factor 2A (EIF2A), UDP-glucose ceramide glucosyltransferase-like 2 (UGCGL2), DNA-damage-inducible transcript 3 (DDIT3, CHOP), heat shock 70 kDa protein 5 (HSPA5, BIP), heat shock protein 90 kDa beta (Grp94), member 1 (HSP90B1), and mitogen-activated protein kinase 8 (MAPK8, JNK). Wolfram syndrome 1 (WFS1), the loss of which leads to apoptosis of beta cells through ER stress 
, was highly, but not differentially, expressed.
There has been great interest in the possibility that ER stress contributes to the loss of beta-cells in T2D, and data indicate that parts of the process are activated in beta-cells of people with T2D 
. One must be cautious about interpreting data from isolated islets because of the trauma of isolation, which could influence the expression of ER stress genes. It is possible that our own data was influenced by severe premorbid illness of the cadaver donors. One must remember that the unfolded protein response of ER stress has a number of features that protect cells from excessive demand for folding 
, so it is possible that the few changes we found were beneficial.
There was no differential expression of the major genes implicated in apoptosis (Table S3
). These include the anti-apoptotic genes B-cell CLL/lymphoma 2 (BCL2) and BCLX, and the pro-apoptotic genes BAD, BAK1, BAX and BIM. FAS had very low signal strength and was upregulated with the probe set (g4507582_3p_a_at) covering the 10%, and recognizing the 3′ end of the transcript. Likewise, no changes were found for caspases 3, 6 or 7.
With regard to other death mechanisms, genes found upregulated in a study using isolated human islets exposed to cytokines were BIRC3, BCL2A1, TNFAIP3, CFLAR 
– most were unchanged in ours but other interesting gene changes were found. In particular, survivin (BIRC5) was modestly upregulated, although with a less specific probe set. Survivin enhances proliferation and survival of cancer cells and seems to have similar effects upon beta-cells 
. Other upregulated genes that could be important are TRAF3, TRADD and TRAF7. Interestingly, BOK and TRAF3IP2 were downregulated.
Expression of some key genes related to oxidative stress are shown in Table S3
. The antioxidant thioredoxin (TXN) was highly expressed and downregulated. The mitochondrial thioredoxin 2 (TXN2), which regulates mitochondrial membrane potential and protects against oxidant-induced apoptosis 
was upregulated with a probe set with weaker signals but not with a probe set with stronger signals. Thioredoxin interacting protein (TXNIP), which in models of diabetes is overexpressed and seems to have deleterious effects on beta-cells 
, was strongly but not differentially expressed. Likewise, thioredoxin reductase 1 (TXNRD1), which may enhance ROS generation 
, was not differentially expressed.
Other genes thought to be important for protection against oxidative stress 
were not differentially regulated including the glutathione peroxidases 3 and 4 (GPX3,GPX4) and catalase. Heme oxygenase 1 (HMOX1) was not different and had a weak signal. The mitochondrial superoxide dismutase 2 (SOD2) was upregulated; the extracellular form of superoxide dismutase 3 (SOD3) was as well but less significantly.
There were notable changes in the expression of a variety of metallothionein genes with a number being clearly upregulated and others tending in that direction. These include the metallothioneins 1E, 1G, 1M, 1X and 2A. In addition, metallothionein 1 pseudogene 2 was upregulated. Metallothioneins are involved in the regulation of metal metabolism, especially zinc, which is important for insulin biosynthesis and storage, and they have potent antioxidant effects because their cysteine residues can capture harmful oxidant radicals 
. Their expression can be increased with oxidative stress.
Another interesting highly expressed downregulated gene is protein phosphatase 1E (PPM1E, PP2C domain containing), which is a member of a family implicated as negative regulators of cell stress response pathways and cell growth 
. In summary, these data, taken in their entirety, suggest that beta-cells in T2D are chronically subjected to oxidative stress. Other changes of interest were up regulation of GADD45 variants, which contribute to DNA repair and are upregulated by the stress of growth arrest.
Data on inflammation from this study have been reported in earlier publications 
A variety of genes potentially involved in pancreatic islet development and regeneration were evaluated (Table S3
). Genes belonging to the regenerating islet gene (REG) family were upregulated in samples of T2D subjects (Table S3
); among them, REG1A and REG3G were those with the highest differential expression, and REG3G was the gene with the highest expression variability among diabetic subjects (Table S3
). The expression of REG1A, REG1B, REG3A, and REG3G was quantified by real-time PCR, performed on samples from four T2D subjects and four controls, which, confirmed the differential expression between the two groups (). The roles of Reg family of proteins continue to be intriguing but poorly understood. Reg was first found in the islets after partial pancreatectomy in rats 
. Reg1 has been implicated in beta-cell growth by association and by administration of the peptide 
, and a knock-out resulted in reduced islet mass 
. Reg has also been found in beta-cells in mice with new onset type 1 diabetes 
. Reg3A (HIP/PAP) is expressed in multiple tissues and has been associated with liver regeneration; it has been found to have mitogenic and anti-apoptotic functions 
. The peptides human proIslet peptide (HIP) and islet neogenesis associated protein (INGAP) are related to sequences in REG3A, and are candidates for inducing neogenesis 
. We also found that the transcription factor SOX9 was upregulated in samples from T2D donors (Table S3
), and this differential expression of SOX9 was confirmed by real-time PCR (). This finding is of interest because SOX9 is known to be present in pancreatic duct cells and has been implicated in islet development by maintaining pancreatic progenitors 
. The metalloproteinase MMP7 was upregulated in samples of T2D subjects (Table S3
). The real-time PCR performed in samples obtained from four T2D subjects and four controls confirmed the differential expression of MMP7 between the two groups (). MMP-7 is involved in development, cancer invasion and sending epithelial signals to stromal cells 
Real-time PCR analysis of the regenerating islet genes Reg 1 alpha, Reg 1 beta, Reg 3 alpha, and Reg 3 gamma.
Real-time PCR analysis of Sox9 and matrix metalloproteinase 7 (MMP7) genes.
Other growth and development genes
Because of indications of a regeneration process taking place, WNT signaling genes were examined (Table S2
). Most of the WNT genes had no differential regulation except for down regulation of WNT6. Of interest, dishevelled (dsh homolog 2, DVL2), which is linked to WNT signaling and expressed in some cancers 
, was upregulated. In addition, dickkopf homologs (DKK) 3 and 4, which are important for development 
, were upregulated. Another upregulated gene was sulfatase 1 (SULF1), which is important for development and survival 
. Frizzled homolog 4 (FZD4) was downregulated. Secreted frizzled-related proteins (SFRP) 2 and 5 were upregulated. We have previously found increases of SFRP expression in differentiating human pancreatic duct cells in vitro 
. There was no differential expression of transcription factor 7-like 2 (TCF7L2).
Notch signaling is important for differentiation of pancreatic precursor cells to islet cells 
. There was no differential expression of notch 1, 2 and 3; notch 3 was most highly expressed (Table S2
). There was, however, a large differential increase of notch homolog 2 (Drosophila) N-terminal like (NOTCH2NL). No differential expression was found with delta-like (DLL) 1, 3 or 4, nor of delta-like 1 homolog (DLK1, pref1), but this was very highly expressed. Jagged 1 (JAG1) was not differentially expressed, but jagged 2 (JAG2) was upregulated. Hairy and enhancer of split 1 (HES1) showed no difference, nor did presenilins (PSEN) 1 and 2, or MFNG O-fucosylpeptide 3-beta-N-acetylglucosaminyltransferase (MFNG, manic fringe).
A variety of other genes have been suggested to be involved in islet development 
. There was downregulation of SMAD3 and SMAD5, which are downstream targets of TGF beta. There was modest upregulation of BMP1, isoform 3, and BMP5. Other downregulated genes included cellular retinoic acid binding protein 1 (CRABP1) and ubiquitin carboxyl-terminal esterase L (UCHL) 1 and 3. BAMBI, which functions as a negative regulator of TGF-beta signaling during development 
, was upregulated. Inhibin, beta E (INHBE), a member of the activin beta family, is thought to be involved in pancreatic exocrine cell growth and proliferation 
and was upregulated as was LASS1, a member of the TGF-beta superfamily, involved in cell growth and differentiation 
Endothelin 3 (EDN3) was very highly expressed but markedly downregulated in T2D, while the expression of its receptor (EDNRB) was unchanged. This secreted factor was found to enhance proliferation and survival of intestinal goblet cells 
. Ephrin-A3 (EFNA3) was upregulated, which is of interest because this tyrosine kinase, has been found play a role in development and also influence insulin secretion 
We wondered if there were a regeneration process that was held in check by increased expression of tumor suppressors (Table S2
). However, there was no differential expression of RB1, TP53, PTEN or APC. Other genes have been found to suppress metastases including MED23, DRG1, CD81, NME 1 and 2, TIMP2 and BRCA1; none were differentially changed.
Genes associated with type 2 diabetes
Recent genome wide association studies (GWAS) have identified a number of gene loci associated with type 2 diabetes, many of which are related to beta-cell function or development 
. As shown in , the genes with upregulated expression include: IGF2BP2, TSPAN8, and HNF1B (TCF2). Downregulated expression was found for JAZF1 and SLC30A8. Then there were genes with clear signal strength and little to no suggestion of differential expression: CDKAL1, TCF7L2, CDKN2A/2B, FTO, WFS1, THADA, ADAMTS9, HHEX, IDE, and CDC123. Finally there were genes with little or no expression: CAMK1d NOTCH2, KCNJ11, MC4R, and PPARG. While these data are unique, it is not clear whether the genes with differential expression have pathophysiological importance. Some of the gene polymorphisms must cause problems with function and development, but there is no particular reason to have expected their level of expression to be altered in this group of non-genotyped subjects.
Gene expression of molecules associated with type 2 diabetes according to genome wide association studies (7).
Functional and ontologies gene sets analysis
After collapsing of the probe sets by using the gene symbols, 20,648 genes were selected; out of these genes, 9,718 (47.1%) were markers of the control phenotype and 10,930 (52.9%) were markers of the T2D phenotype. Using the phenotype-based permutation on the functional gene set analysis (C2 collection, canonical pathways, chemical and genetic perturbations), two gene sets were found enriched in samples obtained from T2D subjects and one was enriched in samples from non-diabetic controls (Table S4
). In the C5 GO Biological Process collection, 13 gene sets were enriched in samples from T2D subjects and 5 were enriched in samples from controls (Table S4
). In the GO Molecular Function and GO Cellular Component, 7 gene sets and 3 gene sets were respectively enriched in beta-cell samples from T2D subjects (Table S4
). Changes of particular interest include three sets indicating altered JNK activity, which may have implications for susceptibility to apoptosis. Also noteworthy was the differential expression of fatty acid oxidation pathways, especially considering the findings linking altered fat metabolism with insulin secretory dysfunction 
These data provide unique new information about gene expression in beta-cells of subjects with T2D. Some of the results obtained previously from preparations of isolated human islets agree with ours, but much does not, which is not surprising considering the lack of purity of these samples and the trauma to which they were exposed. Our approach also has some drawbacks, but the use of LCM on frozen sections eliminates many for the artifacts of islet isolation. Some changes were found in genes implicated in glucotoxicity, which complement existing rodent data, but the premorbid illness of the subjects may have obscured changes that might have been found with pristine specimens that might be obtained by either surgery or biopsy.
Little was found in genes associated with beta-cell replication and death, including the categories of cell cycle, apoptosis and ER stress. An important point is that there is heterogeneity in beta-cells, which is poorly understood, such that cells primed for division and death might be a small minority of the population. This fits the small number of beta-cells immunostained for CHOP (DDIT) in T2D islets 
. Thus, it is easy to see how relevant changes of gene expression in a minority population could be missed. It is of considerable interest that upregulation of a number of metallothionein genes was found; the genes are known to be sensitive to alterations in oxidative stress.
The changes in the expression of genes associated with pancreatic regeneration were unexpected findings from the samples of these subjects with T2D. There is little to support the presence of active regeneration in T2D, but there may be a process that is somehow activated and then arrested. These findings certainly require further study.
Finally, differential gene expression was found in some of the genes linked by GWAS studies to the pathogenesis of T2D. These finding may contribute to the intense efforts currently underway to understand how these genes make their contributions to the development of T2D.