Detection and Characterization of a Biochemical Signature Associated with Diabetic Nephropathy Using Near-infrared Spectroscopy on Tissue Sections

Abstract

Histological evaluation of renal biopsies is currently the gold standard for acquiring important diagnostic and prognostic information in diabetic nephropathy (DN) patients. Nevertheless, there is an unmet clinical need for new biomarkers that allow earlier diagnosis and risk stratification. As biochemical changes in tissues must precede any symptomatic or morphological expression of a disease, we explored the potential of near-infrared (NIR) spectroscopy in the detection of a biochemical signature associated with DN. Kidney tissue sections were investigated using NIR spectroscopy, followed by principal component analysis and soft independent modelling of class analogy. A biochemical signature indicative of DN was detected, which enabled perfect discrimination between tissue sections with normal histological findings (n = 27) and sections obtained from DN patients (n = 26). Some spectral changes related to carbamoylation and glycation reactions appeared to be similar to the ones obtained in patients with DN. In addition, treatment with the deglycating enzyme fructosamine-3-kinase resulted in partial to pronounced restorations of the spectral pattern. Significant relationships were found between spectral features and laboratory parameters indicative of glycemic and uremic load, such as hemoglobin A1c, urea, creatinine, estimated glomerular filtration rate, and proteinuria. The presented method could be a useful tool to complement histopathological analysis in order to prevent or delay further disease progression, especially in the setting of post-transplant surveillance kidney biopsies.

Keywords: diabetic nephropathy; near-infrared spectroscopy; post-translational modifications; renal tissue.

Figure 1

Schematic representation of the near-infrared (NIR) spectroscopy-based setup. The basic components include a computer, NIR-spectrometer, optical fiber, integrating sphere, and a white reference tile layered with immersion oil.

Figure 2

Discriminative spectral features in diabetic nephropathy (DN) patients. (A) First derivative of the median spectra from the control group (green line, n = 27) and DN patients (blue line, n = 26). Statistically significant differences (Mann-Whitney U test) in peak intensities between both groups are indicated with an asterisk. In first derivative, resolution is enhanced since the rate of change of absorbance (A) with respect to wavelength (λ) is examined (dA/dλ). (B) Box and whisker plots showing decreased intensities at 1468 nm, 1949 nm, and 2279 nm in the group of DN patients, compared to the controls. In contrast, the intensities at 2082 nm and 2209 nm showed significant increases.

Figure 3

Spectral signature of (de)glycated and carbamoylated tissue sections. (A) Score plot showing clear clustering of three groups: baseline samples (green dots), deglycated samples (yellow dots), and a separate group with the glycated and carbamoylated samples (red and blue dots, respectively) based on the 1700–2165 nm spectral range. (B) First derivative of the mean spectra (full spectral range) from the baseline samples (green line), carbamoylated (blue line), glycated (red line), and deglycated samples (yellow line). Close-up on the zones of interest at 1468 nm (C), 1949 nm (D), 2082 nm (E), 2209 nm (F), and 2279 nm (G).

Figure 4

(A) First derivative spectra of lysine (green line) and homocitrulline powder (blue line). The peaks of interest are indicated with an asterisk. (B) First derivative spectra of non-post-transplant patients without DM (n = 2) and non-post-transplant patients with DM (n = 3).

Figure 5

Correlation of spectral markers with routine laboratory parameters. (A) Scatterplots illustrating Spearman’s correlations between first derivative intensities at 1879 nm, 1949 nm, 2209 nm, 1987 nm, 2222 nm, and hemoglobin A1c (HbA1c) (%). (B) Scatterplots illustrating correlations between first derivative intensities at 1403 nm, 1732 nm, and urea (mg/dL). The solid and dashed lines represent the 95% prediction interval and 95% confidence interval, respectively.