Dietary Restriction and Medical Therapy Drives PPARα-Regulated Improvements in Early Diabetic Kidney Disease in Male Rats

The attenuation of diabetic kidney disease (DKD) by metabolic surgery is enhanced by pharmacotherapy promoting renal fatty acid oxidation (FAO). Using the Zucker Diabetic Fatty and Zucker Diabetic Sprague Dawley rat models of DKD, we conducted studies to determine if these effects could be replicated with a non-invasive bariatric mimetic intervention. Metabolic control and renal injury were compared in rats undergoing a dietary restriction plus medical therapy protocol (DMT; fenofibrate, liraglutide, metformin, ramipril, and rosuvastatin) and ad libitum-fed controls. The global renal cortical transcriptome and urinary 1H-NMR metabolomic profiles were also compared. Kidney cell type-specific and medication-specific transcriptomic responses were explored through in silico deconvolution. Transcriptomic and metabolomic correlates of improvements in kidney structure were defined using a molecular morphometric approach. The DMT protocol led to ~20% weight loss, normalised metabolic parameters and was associated with reductions in indices of glomerular and proximal tubular injury. The transcriptomic response to DMT was dominated by changes in fenofibrate- and PPARα-governed peroxisomal and mitochondrial FAO transcripts localizing to the proximal tubule. DMT induced urinary excretion of PPARα-regulated metabolites involved in nicotinamide metabolism and reversed DKD-associated changes in the urinary excretion of TCA cycle intermediates. FAO transcripts and urinary nicotinamide and TCA cycle metabolites were moderately-to-strongly correlated with improvements in glomerular and proximal tubular injury. Weight loss plus pharmacological PPARα agonism is a promising means of attenuating DKD.


SUPPLEMENTARY FIGURES
S1 Raw spectra and spectral processing of selected urinary 1 Figure S1. Raw spectra and spectral processing of selected urinary 1 H-NMR spectroscopy peaks in the ZDSD DMT preclinical study.
A, D, and G: Raw 1 H-NMR spectra of selected peaks: hippurate, 1-methylnicotinamide, and 2-oxoglutarate. Each line corresponds to the spectrum of a single sample. B, E, and H: Detection of peaks within the spectra using a Mexican hat wavelet transformation. Each dot represents peak intensity for a single sample. C, F, and I: Alignment of peaks by grouping to account for shifts in peaks between spectra due to differences in sample environment and/or experimental conditions. Each dot represents peak intensity for a single sample. Manual inspection of the raw spectra and spectral processing in Speaq was performed as outlined here for peaks identified as important to classification of ZDSD rats by multivariate models to ensure that between-group differences in the identified peaks were not artefactual and could be reliably identified in the spectra. Healthy, n=11; untreated mild, n=11; untreated severe, n=8; DMT, n=6. 1 H-NMR, proton nuclear magnetic resonance spectroscopy; DMT, dietary restriction plus medical therapy; PPM, parts per million chemical shift relative to TSP-d4; PQN, probabilistic quotient normalisation; ZDSD, Zucker Diabetic Sprague Dawley.

2/11
Supplementary Figure S2. Comparison of mitochondrial morphological characteristics between the pars convoluta and pars recta sections of the proximal tubule in the ZDSD study. Supplementary Figure S3. Volcano plots of differentially expressed transcripts between experimental groups in the ZDF and ZDSD DMT preclinical studies.
Volcano plots of differentially expressed renal cortical transcripts between experimental groups in the ZDF and ZDSD studies [1]. Log 2 fold change is presented on the x-axis with −log 10 transformation of the multiplicity-corrected p-value for differential expression on the y-axis. Horizontal and vertical lines demarcate thresholds used to define strongly expressed transcripts: log 2 fold change values of −2 and 2, as well as a −log 10 p-value of 1. qRT-PCR validation of expression changes in PPARα-responsive transcripts, both peroxisomal (Acox1, Ehhadh) and mitochondrial (Acaa2, Pdk4), in both liver (n=6 SD, n=7 SHAM, n=6 DMT) and epididymal fat tissue (n=5 SD, n=6 SHAM, n=4 DMT) in the ZDSD experiment. Statistical significance of between-group differences derived from multiplicity-corrected Wilcoxon rank-sum tests is denoted as follows: ns=not significant; *=p<0.05; **=p<0.01; ***=p<0.001; ****=p<0.0001. DMT, dietary restriction plus medical therapy; RQ mRNA, relative quantification of messenger ribonucleic acid; ZDSD, Zucker Diabetic Sprague Dawley.

6/11
Supplementary Figure S6. In silico deconvolution of the predicted cellular source of transcripts differentially expressed between DMT and SHAM rats in a human DKD snRNA-seq dataset.
DMT vs SHAM DEGs commonly changed and sharing directionality between both models were intersected with a human diabetic kidney snRNA-seq dataset [4].

7/11
Supplementary Figure S7. Relational network map of the correlation structure between transcripts, metabolites, and structural parameters in the renal cortical response to DMT.
The network depicts Pearson correlation r values between selected renal cortical transcripts, urinary metabolites, and kidney structural parameters. Regularized log-transformed gene expression counts were used for gene-structure correlations. Selected transcripts which resulted in enrichment of peroxisomal and mitochondrial lipid metabolism pathways in ZDSD DMT rats relative to SHAM rats are plotted. PQN-normalised urinary 1 H-NMR peaks from samples obtained at 4 weeks after intervention were used for metabolite-structure correlations. Correlations for TCA cycle intermediates and PPARα biomarker metabolites involved in nicotinamide metabolism, many of which were differentially abundant between DMT-treated and untreated ZDSD rats, are plotted. Each node represents a variable and node labels are coloured according to variable type. Node size is scaled by the node degree, reflecting the number of edges connected to that node. Clustering of node positions is based on multidimensional scaling of absolute correlation values.   Multiple testing corrections were applied using the Benjamini-Hochberg method [5].