We report renal genome-wide gene expression changes induced by prolonged hyperglycaemia in models of experimentally-induced (STZ-WKY) or spontaneous (GK) diabetes, widely used in diabetes research and known to develop renal alterations [18, 20, 30]. Two well-established transcription profiling technologies (Affymetrix and Illumina) provided highly concordant information on the effects of severe (STZ-WKY) or moderate (GK) hyperglycaemia, as well as GK-specific diabetes susceptibility genes, on the expression of over 15,000 known and predicted genes. Differentially expressed genes, which reflect mechanisms responsive to hyperglycaemia, provide improvements in functional annotations of known and predicted genes and potentially novel targets in DN genetics.
Renal histological changes in GK and STZ-WKY rats are relatively mild in these models [21, 30–32], which may therefore be useful for investigating early molecular adaptations to hyperglycaemia in DN progression. Genome-wide gene expression profiling, which allows unbiased analysis of thousands of genes regardless of their function and chromosomal position, is a practical approach to comprehensively assess perturbed mechanisms in GK and STZ-WKY rats maintained in strictly identical conditions and uncover renal molecular adaptations to impaired glucose homeostasis in a context where environmental influences and genetic polymorphisms are minimized, as inbred WKY and GK strains share extensive genetic similarities outside GK-specific diabetes variants [19, 32]. Gene expression results from two well-established and robust microarray technologies designed to investigate the transcription of largely overlapping series of genes , which provide systems for assessing data replication, showed remarkable concordance in transcription regulation patterns. Some discordant results may be explained by the design of the probesets (Affymetrix) or the oligonucleotides (Illumina) in different gene isoforms or the existence of polymorphisms in the sequences arrayed on the chips .
Transcriptomic alterations were more profound in STZ-WKY than in GK and involved largely non-overlapping sets of genes, thus underlining the importance of investigating gene expression in very different, but complementary, contexts of diabetes aetiology and pathogenesis. Age may have a confounding impact on these processes as, for obvious reasons, the models used were not age-matched, but the duration of spontaneous (GK) or experimentally-induced (STZ-WKY) hyperglycaemia was identical (3 months). Along the same line, gene expression differences between STZ-WKY and WKY, which compare pre- and post-drug treated groups and test molecular responses to prolonged and severe hyperglycaemia, may also reflect to some extent "normal" biological adaptations to ageing described in WKY . Similar model-specific patterns of gene expression regulation was observed in renal transcriptomes of mouse models of STZ-induced and spontaneous (db/db) diabetes . Diabetes-predisposing variants in the GK and db/db strains may therefore affect renal changes through specific mechanisms not necessarily related to hyperglycaemia itself. The existence of such specific polymorphisms  and a genetic predisposition to salt-induced hypertension  are documented in the GK strain. Metabolic and hormonal factors, as well as renal structural alterations secondary to impaired glucose regulation, may also explain the extent of renal transcriptomic differences between STZ-WKY and GK. Insulin deficiency is the leading cause of hyperglycaemia in STZ-treated animals , whereas the GK strain develops insulin resistance, which is associated with DN  and may play a specific role on renal gene expression in GK rats.
Combining transcriptomic results in GK and STZ-WKY rats with published data in animal models of diabetes [31, 39, 40]. and in humans [41, 42] contribute to the enrichment of renal functional pathways that may be involved in DN. Of note, upregulation of Grem1, a bone morphogenetic protein antagonist, and strong downregulation of cadherins, Egf and Tff3, a protein promoting epithelial cell restitution, may contribute to renal damage in STZ-WKY, [43, 44]. In GK rats, altered transcription of proliferation and differentiation factors (Bmp3, Bmp6, Bmp7)  suggests the involvement of mechanisms protecting the kidney against interstitial fibrosis , that upregulation of transcripts increasing extracellular matrix production (Ctgf, Mmp9) may initiate . Downregulation of aldo-keto reductases (Akr1b8, Akr1c12, Akr7a3) and upregulation of Sord (in STZ-WKY) may reflect the activation of compensatory mechanisms preventing the accumulation of sorbitol and reducing polyol pathway flux . Stimulated expression of glutathione synthase in GK may reflect cytoprotective mechanisms, whereas its downregulation in STZ-WKY may be balanced by increased expression of glutathione transferases, therefore contributing to protect the cells against oxidative stress . However, expression of genes encoding protein known to play a role in DN pathogenesis, including TGFB, heme oxygenase (HMOX1), osteopontin (SPP1), were not significantly altered in diabetic rats, despite a 30–40% upregulation.
The primary aim of the study was to identify genes that are differentially expressed in diabetic models and map to regions of the rat genome that show evidence of synteny conservation with human loci associated with DN. As largely overlapping series of genes are represented on Illumina and Affymetrix arrays, replicated gene transcription differences with the two technologies provide robust information for selecting positional and functional DN candidate genes. Candidates include genes and ESTs that have no apparent functional relevance to DN, but their renal transcription patterns indicate a possible role in kidney structural and metabolic alterations. Of note, hepatic transcription profiling data in the same animals show that only a small proportion (6%–11%) of genes are consistently differentially expressed in kidney and liver (data not shown). Comparative genomic analysis of DN positional candidates mapped to 3q, 7q, 18q and 20p [8–11] revealed several differentially expressed genes that have not been tested in genetic studies of DN [13, 27, 28] and may therefore be important in genetic and clinical investigations of DN. Among relevant genes, Slc14a2 encodes a vasopressin-dependent urea transporter expressed in the collecting duct. Its upregulation in STZ-WKY may contribute to nitrogen conservation in response to glucosurea . In these rats, downregulation of Cldn16, a major structural component of tight junction in the ascending limb of Henle, may result in ions wasting and renal glomerular and tubular alterations [49, 50]. Finally, altered renal transcription of Rbp1 and Mep1b has been reported in mouse models of experimentally-induced diabetes  and tubular fibrosis .
Positional candidates differentially expressed in diabetic rats also include protein coding sequences predicted by bioinformatic models which nevertheless represent important functional DN candidates. Even though genome annotations and microarray technologies keep improving, potentially important DN positional candidate genes are not represented on gene expression arrays. Systematic renal expression analysis of ESTs mapped to the most significant regions of DN loci identified transcripts differentially regulated between GK and STZ-WKY rats. Of note, using quantitative RT-PCR, we found evidence of significant changes in expression for transcripts predicted to encode a further aldo-keto reductase (Akr1b10) and the zinc finger protein 236, which has been tested as a DN candidate . In the vast majority of cases, confirmation of gene organization and biological functions, including altered protein abundance, is required. Although we prioritized our study to gene expression studies in total kidney to take into account interactions between renal cell types in the regulation of organ function, further investigations in tubular and mesangial cells will establish tissue specific gene expression patterns.