Microarray, RT-qPCR and statistical analysis

Total RNA was extracted from peripheral blood mononuclear cells (PBMCs) since we focused on the circulating lymphocytes and monocytes gene expression signature. After extraction and purification of RNA samples, only those in the λ(260/280) and λ(260/230) ratios between 1.8 to 2.2 were used. Details regarding the aforementioned methodologies can be found in the Supplementary Materials and Methods. Microarray data (U133 Plus 2.0, Affymetrix Inc., Santa Clara, CA, USA) was generated from patients T2DMpoorly-DL-P (n = 5), T2DMwell-DL-P (n = 7), DL-P (n = 6), P (n = 6) and H (n = 6), after considering greater homogeneity regarding biochemical, lipid and clinical periodontal parameters⁶.

The raw.CEL files were preprocessed using the Robust Multichip Average (RMA) strategy, as implemented in the Affy package⁴⁷, which performs background correction through a normal-exponential convolution model, quantile normalizes the probe intensities and summarizes them into probeset-level quantities using an additive model fit through the median-polish strategy. The log2-expression quantities resulted from the RMA method were further processed by the RankProd package⁴⁸. Probesets that presented |log2FC|>1 and percentage of false prediction (pfp)<0.01 were called up- or downregulated, depending on the sign of the log-FC. Pairwise comparisons were made between the H (systemically and oral healthy subjects) and each of the T2DMpoorly-DL-P, T2DMwell-DL-P, DL-P and P groups (T2DMpoorly-DL-P versus H; T2DMwell-DL-P versus H; DL-P versus H and P versus H). The significant probesets from the aforementioned workflow were analysed with the Ingenuity Pathway Analysis software (IPA; Build 470319 M, Version 43605602, Qiagen, Redwood City, CA) as well as the Gene Set Enrichment Analysis software (GSEA; version 4.0.3 http://www.broad.mit.edu/gsea). Additional methods details from processing and analyses can be found at the Supplementary Materials and Methods section.

Independent validation of selected DEGs was performed by RT-qPCR analysis in a total of 150 patients (n = 30 each group, including the patients chosen for microarray analysis). Statistical analyses were performed in the GraphPad Prism software (version 5.0) using a significance level of 0.05⁶.