Diabetes mellitus is a chronic and heterogenous metabolic disorder affecting millions of patients worldwide. Type 1 diabetes is characterized by absolute insulin deficiency due to viral or autoimmune destruction of pancreatic beta cells, while the major feature of the more common type 2 variant is obesity-linked impairment of intracellular insulin signaling [1–3]. Apart from its well-known effect on blood sugar levels, insulin is known to regulate the growth, differentiation and metabolism of its target cells at multiple levels [1]. Insulin signaling pathways have been shown to converge on and modulate the transcription of a plethora of genes [2]. In light of this, it is not surprising that gene expression microarrays revealed dramatic alterations in global gene expression profiles of several organs such as skeletal muscles and adipose tissue [3], intestine [4] and the liver [5] both in type 1 and type 2 diabetes.

Although the brain does not count as a classical target organ of insulin, it has recently been shown that this polypeptide hormone plays a crucial role in human neurophysiology, and dysregulation of insulin receptor signaling in various mental illnesses [6].

It has long been known that insulin can pass the blood brain barrier by receptor mediated endocytosis [7]. Moreover, it turned out that several brain regions are capable of producing insulin in situ [8]. The insulin receptor and insulin receptor substrate-1 (IRS1) are expressed in vegetative nuclei of the hypothalamus, in amygdala, hippocampus and in the neocortex [9]. Based on this expression pattern, cerebral insulin signaling has been implicated in the regulation of neurotransmission, feeding and cognitive functions [10].

Along with leptin, insulin seems to be a negative feedback signal in well-fed state due to its ability to reduce appetite and body weight. It might be assumed that obesity and hyperinsulinism lead to desensitization of insulin receptors situated in the blood brain barrier, giving rise to central insulin resistance [11].

There are several lines of mostly indirect evidence supporting the role of insulin signal transduction in learning and long-term memory. The first observations date back to the famous Rotterdam study, revealing that type 2 diabetes doubled the risk of patients to develop Alzheimer-type dementia, while individuals suffering from type 1 diabetes and receiving insulin therapy had four times the risk [12]. These results were corroborated by more recent studies showing that subjects with elevated body mass index, obesity, insulin resistance and diabetes have an increased risk of dementia and cognitive impairment, suggesting a causal link between decreased insulin secretion and the progression of mental decline [13]. Subsequently, post-mortem brain studies unveiled that cerebral insulin, insulin receptor and IGF levels are inversely proportional with the progression of Alzheimer's disease [14]. On the other hand, intranasal and intravenous insulin administration has reportedly improved the cognitive functions of patients suffering from memory disorders, while intracerebroventricular insulin enhanced memory formation in rodents [15, 16]. Moreover, intracerebral administration of streptozotocin, a drug known to induce type 1 diabetes by impairing pancreatic β cells when added intravenously, also led to insulin depletion in the brain with subsequent neurodegeneration [17].

The interrelationship between diabetes and Alzheimer's disease seems to be mutual as neurotoxins termed amyloid beta-derived diffusible ligands have been shown to compromise cerebral insulin signaling [18]. On the other hand, oxidative stress elicited by reactive advanced glycation end products (RAGEs) that are characteristic of diabetes might accelerate neuronal damage in memory disorders [19].

Based on these observations, a group of researchers have recently defined Alzheimer's as a neuroendocrine disorder and coined the terms "type 3" or „brain-type" diabetes [20], pointing out that this condition can simultaneously be characterized both by central insulin deficiency and insulin resistance. Their work highlighted the importance of impaired insulin signaling in the dysfunction and apoptotic death of cortical neurons.

Although global transcriptome profiling has already been carried out in Alzheimer's disease [21], according to our best knowledge this is the first study aiming to analyze whole genome gene expression profiles of various cerebral areas in streptozotocin-induced and spontaneously diabetic Goto-Kakizaki rats as model animals for type 1 and type 2 diabetes, respectively. Our results demonstrated an altered expression pattern in the hippocampus and prefrontal cortex of type 2 diabetes model, while no such changes were found in the corresponding brain areas of the type 1 model animals.

Next, we wished to assign biological relevance to our gene lists by ordering them in biochemical pathways. The Biological Process domain of the Gene Ontology database provided the most extensive pathway assignment. 64 genes from the hippocampus and 36 from the prefrontal cortex were found to be members of certain pathways (Table 1).

Finally, gene expression changes fulfilling the criteria of mathematical-statistical selection and pathway analysis were validated by real time PCR using TaqMan Low Density Arrays. It should be noted that only genes with commercially available TaqMan probes could be analyzed. Therefore, 42 out of the 64 hippocampal and 27 out of the 36 prefrontal genes were subject to validation. Finally, 30 genes from the hippocampus (71%) and 22 genes from the prefrontal cortex (82%) were validated (Table 2; for detailed gene lists, see Additional File 2). According to our results, 9 genes showed changes both in the hippocampus and in prefrontal cortex in the type 2 diabetes model (for a detailed list, see Additional File 2). Finally, pathway analysis revealed that most genes with altered expression patterns in the hippocampus are involved in oxidative stress and DNA damage signaling, cell cycle regulation, development and lipid metabolism of the central nervous system as well as in the regulation of feeding behavior (Table 2 and Figure 1).

GO Biological processes	HIPPOCAMPUS	Validated	Not validated
Insulin/GH secretion	GO:30073: insulin secretion	Gal
	GO:30252: growth hormone secretion	Gal
Oxidative stress DNA damage cell cycle	GO:6950: response to stress	Gal
	GO:305: response to oxygen radical	Cxcl4(Pf4)
	GO:303: response to superoxide	Akap3
	GO:302: response to reactive oxygen species	Gal	Nudt15_predicted
	GO:15992: proton transport	Ucp2
	GO:6977: DNA damage response, signal transduction by p53 class mediator resulting in cell cycle arrest	Ptprv
	GO:42770: DNA damage response, signal transduction	Ftcd
	GO:7346: regulation of progression through mitotic cell cycle	Snf1lk
	GO:6269: DNA replication, synthesis of RNA primer	NM_001008768 (Prim1)
	GO:7089: traversing start control point of mitotic cell cycle	Cdk10
Lipid metabolism	GO:1573: ganglioside metabolism	Gm2a
	GO:6695: cholesterol biosynthesis	Acaa2	Acaa2
Eating/feeding behavior	GO:7631: feeding behavior	Gal, Agrp
	GO:42755: eating behavior	Agrp	Stat3
Development of the nervous system	GO:7399: nervous system development	Gal, Mobp, Mobp, Cntn3	Ednrb, RGD1311340_predicted, Stat3, XM_242005
	GO:7422: peripheral nervous system development	Sncg	Ednrb
Others	GO:50776: regulation of immune response	Gal, Il22ra2
	GO:6952: defense response	Mx2
	GO:7194: negative regulation of adenylate cyclase activity	Grm2
	GO:6032: chitin catabolism	Chi3l1
	GO:42572: retinol metabolism	Retsat
	GO:45123: cellular extravasation	Itgam
	GO:19637: organophosphate metabolism	Pter
	GO:6928: cell motility	Akap3, Grm2	Stat3
	GO:9615: response to virus	Mx2, Oas1	XM_215121

Data were obtained using the GO pathway analysis software; specific GO pathway identification numbers are provided in the second column. "Validated genes" were confirmed by quantitative PCR analysis. Validated genes found in more than 2 significant pathways are shown in bold.

Insulin regulates gene expression via a set of transcription factors including the FOXO family [24]. As insulin and its receptors are both known to be expressed and to govern important functions in the brain, it seemed reasonable to search for altered gene expression patterns in animal models of type 1 and type 2 diabetes characterized by absolute or relative insulin deficiency. Here we demonstrated a substantial difference in the gene expression pattern of type 2 diabetic rats vs. control animals. The genetically determined, spontaneously diabetic Goto-Kakizaki rats exhibited profound gene expression alterations suggesting that long-standing impairment of insulin signaling has a well detectable effect on the central nervous system. On the other hand, we could hardly detect any alterations in the streptozotocin-induced diabetic animal model (Table 1), suggesting that acute insulin deficiency and/or elevated blood sugar levels do not influence significantly the cerebral gene expression pattern, or at least it is undetectable four weeks after the streptozotocin treatment in a microarray based experiment. It is tempting to speculate that streptozotocin-induced diabetic rats might successfully compensate peripheral insulin deficiency by increased cerebral insulin production. However, this presumption seems to contradict the fact that activation of the ins2 gene was not detected - maybe due to low sensitivity of the whole genome custom array.

Three main brain regions have been studied here: the prefrontal cortex and hippocampus were analyzed due to their well-known roles in learning and memory formation, while the striatum seemed to be an easily dissectable control region where no insulin action had been presumed. It is also interesting to note that streptozotocin-treted rats exhibited some gene expression alterations in the hippocampus only. These observations are in a good agreement with the findings of Agrawal et al., showing that insulin and its receptor are mostly expressed in this brain region, and intracerebroventricular administration of streptozotocin induced memory deficit in rats [25].

Streptozotocin has been proven to induce insulin deficiency and hyperglycemia (≥ 15 mM) within 72 hours in treated animals, and they were alive for 4 weeks following beta-cell destruction. In our opinion, this time window should have been enough to alter gene expression profiles in the brain as there are several reports highlighting the early effects of streptozotocin on gene expression in various organs [26]. The major drawback of the global microarray method is its minor sensitivity compared to that of TaqMan-based quantitative reverse transcription PCR assays. However, the high RT-PCR validation rate of microarray data in Goto-Kakizaki rats (71% in the hippocampus and 82% in the prefrontal cortex, respectively) convinced us of the reasonably good reliability of the chip hybridization technique. Theoretically, some minor gene expression alterations in the brains of type 1 diabetic model animals might have been left undetected by the chip hybridization technique, therefore, we are committed to validate the "non-changed" status of a set of genes which were significantly altered in type 2 diabetic animals using open-array real-time PCR assays.

Analyzing the specific genes, the mRNA levels of galanin, an inhibitory neuropeptide with pleiotropic roles were substantially upregulated in the hippocampus. Notably, galanin were identified in almost all perturbed pathways of the hippocampus (Table 2). Our results corroborated the findings of Mei et al. who detected elevated galanin expression in the celiac ganglion in diabetic rats [27]. Intracerebroventricular administration of galanin or its overexpression in transgenic mice was shown to compromise hippocampus-dependent learning processes [28, 29]. Galanin has been proposed to play a role in depression-like behavior [30]. On the other hand, improvement of cognitive functions has been reported in animals treated with galanin receptor antagonists [28]. As cerebral insulin deficiency presents with similar symptoms, it is tempting to speculate that impairment of cerebral functions in diabetes might be mediated at least in part by elevated galanin levels. This assumption is supported by the fact that plasma galanin levels have been found to be significantly elevated in patients with type 2 diabetes [31], and increased plasma galanin levels were measured following oral glucose load in a healthy population [32]. If we managed to find a causal relationship between cerebral insulin deficiency and galanin overexpression, we might be able to ameliorate cerebral symptoms of diabetes via pharmacological modulation of galanin receptors and to slow down the progression of type 3 diabetes [20].

The role of galanin receptors is also highlighted by our results which demonstrated altered galanin receptor 2 expression levels in the prefrontal cortex (Table 3). Type 2 galanin receptors are mostly expressed in the perikaryon of neurons, mediating calcium signals and promoting the survival of neurons [33], and their stimulation reportedly elicited antidepressive effects [34].

Apart from galanin and its receptor, there are several other validated genes as well, which have already been implicated in the pathogenesis of both diabetes and psychiatric disorders in some respect. For instance, Chi3l1 (YKL-40, chitinase 3-like 1) has recently been shown to represent an obesity-independent novel marker of type 2 diabetes [35]. On the other hand, Chi3l1 has been regarded as a schizophrenia susceptibility gene, a mediator of stress-induced cellular responses [36]. SNCG (synuclein gamma) has recently been termed an adipocyte-neuron gene that is coordinately expressed with leptin in human obesity and might promote adipocyte differentiation [37]. Apart from its well-known role in the development of neurodegenerative diseases [38], SNCG has also been implicated in depression [39], dopamine release [40] and as an interacting partner of the dopamine transporter in rats [41].

Perturbation of brain signaling pathways could also be a very important hallmark of type 2 diabetes. Here we identified three genes of cerebral signaling (protein kinase C gamma and epsilon, and the RET tyrosine kinase) with altered cerebral expression profiles in Goto-Kakizaki rats. They have been shown to play a pathophysiological role in brain dysfunction previously. For instance, expression of the neuron-specific gamma isoform of protein kinase C (Prkcc) that has been implied in the regulation of learning and memory formation (Additional File 2) was more than twofold upregulated in the prefrontal cortex of Goto-Kakizaki rats (Additional File 2). Schlaepfer et al. demonstrated that certain polymorphisms of the Prkcc gene are associated with behavioral disinhibition and attention deficit hyperactivity disorder (ADHD) in humans, while PKC-gamma deficient mice exhibited impulsivity, anxiety and increased ethanol consumption [42]. Importantly, the epsilon isoform of PKC (Prkce) is also overexpressed in the type 2 diabetic model (Additional File 2). This kinase is reportedly involved in neuronal ion channel activation, apoptosis and insulin exocytosis. Recently, Prkce has been implicated in the loss of insulin secretory responsiveness during the development of type 2 diabetes [43], while others highlighted its role in the pathomechanism of drug dependence and addiction [44]. Shelton et al. revealed decreased Prkce protein levels in post mortem brain specimens of patients with major depression [45]. Finally, we demonstrated changes in the expression level of the RET protooncogene, a receptor tyrosine kinase containing cadherin-like repeats in its extracellular domain, that plays a pivotal role in neural crest development. Mutations in this gene might elicit multiple endocrine neoplasia type 2B with diabetes [46]. Interestingly, RET activity has been shown to modulate and shape the brain dopaminergic systems which are known mediators of several personality traits [47].

As far as the theory of type 3 diabetes is concerned, our microarray data revealed a couple of genes which might provide a link between diabetes and neurodegeneration. Apart from the already mentioned synuclein gamma, uncupling protein 2 (UCP2), the ABC-transporter ABCA1 and the cell surface antigen CD47 should also be mentioned in this context. UCP2, a well-known inner mitochondrial membrane protein, responsible for energy dissipation and heat production, has been found to associate with obesity, diabetes and regulation of insulin secretion [48]. On the other hand, the UCP2 gene is induced in a ghrelin-dependent fashion and protects from neurodegeneration [49]. UCP2 expression was significantly downregulated in the hippocampus of our type 2 diabetic rat model (Additional File 1), implying that its neuroprotective effect might be absent from the diabetic brain.

Mutations in the cholesterol efflux pump ABCA1 have been associated with Tangier's disease. Beyond that, ABCA1 has been implicated in insulin secretion from pancreatic beta cells [50], and some single nucleoide polymorphisms (SNPs) of this gene have been demonstrated to associate with dementia (rs2230805) [51] and Alzheimer's disease (rs1800977 and rs2422493) [52]. We found significant downregulation of ABCA1 levels in the prefrontal cortex of Goto-Kakizaki rats (Additional File 2); hence it seems logical to assume that elevated cytosolic cholesterol levels might impair the viability of neurons via affecting membrane fluidity.

The gene for CD47 encodes a membrane protein which is involved in the increase in intracellular calcium concentration that occurs upon cell adhesion to the extracellular matrix. There is ample evidence supporting the role of CD47 in pancreatic insulin secretion [53]. Moreover, CD47 has been shown to interact with amyloid beta peptide in Alzheimer's disease [54]. We measured elevated CD47 mRNA levels both in the hippocampus and in the prefrontal cortex of type 2 diabetes model animals, providing a plausible link between central insulin resistance and Alzheimer-type neurodegeneration.

In conclusion, our study shed light on the seminal role of insulin in maintaining the functions of the central nervous system by unveiling characteristic perturbations in cerebral gene expression profiles in type 2 diabetic rats. We identified several cerebral expression changes in genes which were previously assumed to play a role in pancreatic insulin secretion, implying that these genes might mediate insulin production and exocytosis in the brain as well. Our results should prompt further investigations to decipher insulin signaling pathways in the brain and a detailed analysis of the transcriptional regulation of diabetes-associated genes having been identified in this study.