Sex- and Age-Dependent Red Blood Cell Remodeling in Type 2 Diabetes Revealed by FT-IR and Raman Spectroscopy: Part II

Using Fourier transform infrared (FT-IR) spectroscopy and Raman spectroscopy, a research team identified sex- and age-specific molecular alterations in red blood cells from diabetic mice. Spectral analysis revealed protein misfolding, oxidative stress–related membrane destabilization, and lipid remodeling, establishing a non-invasive, label-free platform for molecular staging and monitoring of type 2 diabetes progression.

In the second of a three-part series, Spectroscopy spoke to authors of the resulting paper¹ about how FT-IR spectroscopy and Raman spectroscopy detect oxidative stress–related disulfide bond alterations and protein instability in diabetic RBCs, address experimental challenges in resolving membrane versus cytoplasmic protein signals, enable non-invasive molecular profiling beyond conventional assays, and, when coupled with multivariate tools such as oPLS-DA and VIP scoring, enhance robust identification of disease-specific spectral biomarkers.

How can changes in disulfide bond signatures detected by FT-IR or Raman spectroscopy be interpreted as markers of oxidative stress and protein instability in diabetes-affected RBCs?

Disulfide bonds are covalent linkages formed between cysteine residues and play a fundamental role in stabilizing protein tertiary and quaternary structures. By constraining protein folding and maintaining structural integrity, they are essential for the proper function of membrane and cytoskeletal proteins in RBCs.

In diabetes, chronic hyperglycaemia is associated with persistent oxidative stress and increased production of reactive oxygen species (ROS). These conditions promote oxidative modification of cysteine residues and disturb the redox balance between reduced sulfhydryl (–SH) groups and oxidized disulfide (–S–S–) bonds. Such redox imbalance can lead to disruption, rearrangement, or abnormal formation of disulfide bridges, thereby compromising protein stability and increasing susceptibility to misfolding or aggregation.

These molecular alterations can be detected spectroscopically through changes in disulfide-related vibrational signatures. Raman spectroscopy provides a direct probe of S–S and S–H bonds via their characteristic vibrational modes. In our study, the 544 cm⁻¹ band, assigned to disulfide bonds, served as a sensitive indicator of protein folding integrity when normalized to the phenylalanine band at 1004 cm⁻¹. Age- and sex-dependent variations in the 544/1004 intensity ratio were observed, reflecting differences in redox regulation and structural stability. However, no statistically significant differences (P > 0.05) were detected between healthy and diabetic groups, except for elevated ratios in 24-week-old diabetic females.

FT-IR spectroscopy captures the consequences of altered disulfide bonding more indirectly. Perturbations in disulfide networks affect overall protein conformation and are reflected in changes within the Amide I region, which is sensitive to secondary structure. Shifts in band position, broadening or altered intensities in this region therefore indicate downstream structural effects of disrupted disulfide bonding.

Taken together, changes in the intensity, position, and relative contribution of disulfide- and thiol-associated spectral features can serve as indicators of oxidative stress and protein instability in diabetes-affected RBCs. However, in our study, these effects were pronounced mainly in aged female mice,indicating that this parameter reflects diabetes-related protein instability only at more advanced stages of pathological change. Moreover, this ratio represents just one of multiple structural alterations and may also vary in other conditions characterized by increased oxidative stress.

What challenges arise when assigning specific spectral features to membrane proteins versus cytoplasmic proteins in intact RBC samples, and how can these challenges be mitigated experimentally?

When intact RBCs are analyzed using vibrational spectroscopy, the resulting spectra represent a superposition of signals from multiple molecular components rather than distinct protein populations. An important methodological consideration, as shown in our previous study, is the different practical applicability of FT-IR (including nano IR) and Raman spectroscopy depending on whether isolated membranes or intact RBCs are analyzed.² In the case of Raman spectroscopy applied to intact RBCs, the spectrum is overwhelmingly dominated by intense resonance Raman enhancement of hemoglobin vibrational bands. Hemoglobin’s heme chromophore exhibits strong electronic absorption in the visible range, and when Raman excitation wavelengths overlap with these absorption features, resonance enhancement occurs, selectively amplifying heme-related vibrational modes by several orders of magnitude. This strong resonance signal effectively masks weaker Raman bands arising from membrane lipids and proteins, making it challenging to extract subtle molecular changes in the membrane matrix from whole-cell Raman spectra. Therefore, Raman spectroscopy yields the most interpretable information on membrane biochemical composition when isolated membranes are measured, where hemoglobin contribution is absent or greatly reduced.

As demonstrated in the present study, as well as in our previous investigations on other cardiometabolic disease models such as heart failure³ and atherosclerosis,⁴ Raman spectroscopy is well suited for biochemical analysis, but its practical utility is predominantly in the study of isolated RBC membranes. Under these conditions, resonance effects from cytosolic hemoglobin are effectively eliminated and Raman spectra yield sensitive, chemically specific markers of membrane composition. In our recent work on diabetes, we have likewise demonstrated Raman-based clear and reproducible biochemical differences associated with disease state in RBC membranes. These and earlier findings strongly indicate that Raman spectroscopy of isolated membranes can sensitively detect pathologically relevant changes in membrane biochemistry.

On the other hand, FT-IR spectra of intact RBCs are not dominated by hemoglobin effects (as in Raman) and therefore allow more balanced access to both cytosolic and membrane molecular features. In our previous comparative study,² we showed that FT-IR spectroscopy provided broad chemical characterization of atherosclerosis-induced alterations in RBC membranes, including alterations in lipid classes and protein structural features, both in isolated membranes and in intact cells. In our current study,¹ we showed that FT-IR spectroscopy was found to be highly sensitive to biochemical changes induced by diabetes not only in isolated membranes but also in whole RBC samples. The spectral differences related to disease state in intact cells were sufficiently strong and systematic that they could be used to distinguish stages of diabetes without the need for membrane isolation, suggesting substantial diagnostic potential for FT-IR analysis of whole RBCs. Avoiding the laborious membrane isolation step enhances translational feasibility and supports the application of FT-IR spectroscopy as a practical approach for future clinical spectroscopy-based phenotyping. This study, which is based on an extensive experimental design involving 120 animals, ensures robust statistical validation. Multivariate models demonstrate high discriminatory power, with FT-IR protein signatures emerging as the most diagnostically informative domain for intact RBCs.

Why is vibrational spectroscopy particularly suited for non-invasive molecular characterization of RBCs compared to conventional hematological or biochemical assays in diabetes research?

In our work, we use two key terms to highlight the advantages of vibrational spectroscopy: label-free and non-invasive.

One of the main advantages of these spectroscopic approaches is their non-invasive nature - not in the clinical sense of measurements performed through intact skin, but in the analytical sense that the sample remains structurally intact after measurement. For whole RBC analysis, neither Raman nor FT-IR spectroscopy destroys the cells or consumes their constituents, allowing the same sample to be subsequently examined using other analytical techniques. This non-destructive property enables complementary and longitudinal studies on the same specimen and preserves valuable biological material for further investigations. It should be noted, however, that in our study only FT-IR spectroscopy provided reliable diagnostic performance for diabetes when applied to whole RBCs.

It is also important to emphasize that, although both techniques require an initial blood draw to isolate RBCs, like conventional biochemical assays, they involve minimal sample preparation and do not require chemical modification of the specimen. Raman and FT-IR spectroscopy are particularly advantageous for molecular characterization because they provide detailed biochemical information directly from the sample without the need for external probes or dyes. Consequently, these methods generate a label-free molecular “fingerprint,” as they do not rely on exogenous labels, stains or markers. The recorded spectral signals arise exclusively from intrinsic molecular vibrations within the sample. As a result, the obtained biochemical signatures reflect the true composition and structural state of proteins, lipids and other biomolecules in RBCs without altering their native properties.

In contrast, conventional hematological and biochemical assays often rely on targeted markers, enzymatic reactions or immunoassays that require reagents and are restricted to a limited set of predefined analytes. These methods typically quantify specific parameters, such as glucose concentration, HbA1c levels or cell counts, but do not provide comprehensive information on molecular alterations across multiple biomolecular classes. By directly capturing a broad range of vibrational modes associated with biochemical bonds, vibrational spectroscopy enables simultaneous monitoring of diverse molecular components. The resulting IR or Raman spectrum therefore serves as an integrated snapshot of the biochemical state of RBCs.

How do multivariate approaches such as oPLS-DA improve the identification of diabetes-related spectral markers compared to univariate band integration, and what is the role of VIP scores in biomarker discovery?

The biochemical information contained in FT-IR and Raman spectra is distributed over numerous, highly correlated wavenumbers that reflect overlapping vibrational contributions from various biomolecules. As a result, univariate band integration, which focuses on isolated spectral features, captures only a limited fraction of this information and may overlook subtle but biologically relevant spectral variations associated with disease progression.

Multivariate approaches such as orthogonal partial least squares discriminant analysis (oPLS-DA) overcome these limitations by simultaneously analysing the entire spectral dataset and exploiting its covariance structure. By modelling correlations among multiple variables, oPLS-DA enables the detection of coordinated intensity changes, band shifts and band-shape variations that collectively characterize diabetes-related molecular alterations. Importantly, oPLS-DA separates predictive variance linked to disease status from orthogonal variance arising from biological heterogeneity, experimental noise or instrumental effects. This decomposition enhances model robustness and improves the reliability of class discrimination between diabetic and healthy samples.

In addition to its strong classification performance, oPLS-DA is an interpretable machine-learning method. It allows direct examination of the spectral variables that drive group separation.It allows direct examination of the spectral variables that drive group separation in the component regression vector. Furthermore, Variable Importance in Projection (VIP) scores quantitatively describe the contribution of each wavenumber to the predictive components of the model. Variables with high VIP values are therefore identified as the most influential features for class discrimination and can be prioritized as candidate spectral biomarkers.

These high-VIP spectral regions can subsequently be correlated with specific molecular vibrations, intensity ratios, or secondary-structure–related bands, enabling mechanistic interpretation in terms of protein conformation, lipid organization or metabolic alterations. Consequently, VIP-guided variable selection facilitates the transition from purely statistical discrimination to biologically meaningful biomarker discovery.

Overall, by integrating information across the full spectral profile, separating relevant from confounding variability, and enabling systematic identification of influential variables, oPLS-DA combined with VIP analysis substantially enhances biomarker discovery compared to univariate approaches. This multivariate framework captures complex, disease-specific spectral signatures that reflect coordinated molecular changes in diabetes-affected RBCs.

Join us tomorrow for the final part of this interview series, where the authors discuss how FT-IR spectroscopy and Raman spectroscopy can track diabetes progression through spectral markers of protein aggregation and membrane rigidity, account for age- and sex-dependent variations in the db/db mouse model, and address translational challenges in adapting murine spectrochemical signatures to human type 2 diabetes diagnostics.

Acknowledgments

This research was partially funded by the Polish National Science Centre, No. UMO-2020/38/E/ST4/00197. Part of the work was carried out at the Laboratory for Biomedical Spectroscopic Applications (LBSA), Faculty of Pharmacy, Jagiellonian University Medical College. The laboratory infrastructure was co-financed by the European Regional Development Fund under the European Funds for Malopolska 2021–2027 programme (project No.FEMP.01.04-IZ.00-0269/24). The methodologies applied in this study are based on spectroscopic approaches developed by the authors and protected by intellectual property rights, including international and national patent applications: PCT/PL2020/050093, P.436054 and P.43217, as well as a registered trademark (No. Z.493251).

Read the first part of the interview here.

References:

1. Wilk, A.; Wilkosz, N.; Rugiel, M. et al. Next-Generation Diabetes Diagnostics: Spectrochemical Staging of Red Blood Cells Using Vibrational Spectroscopy. J. Adv. Res. 2026, S2090-1232 (26), 00077-9. DOI: 10.1016/j.jare.2026.01.052
2. Blat, A.; Dybas, J.; Kaczmarska, M. et al. An Analysis of Isolated and Intact rbc Membranes - A Comparison of a Semiquantitative Approach by Means of FTIR, Nano-FTIR, and Raman Spectroscopies. Anal Chem 2019, 91. DOI: 10.1021/acs.analchem.9b01536
3. Mohaissen, T.; Proniewski, B.; Targosz-Korecka, M. et al. Temporal Relationship Between Systemic Endothelial Dysfunction and Alterations in Erythrocyte Function in a Murine Model of Chronic Heart Failure. Cardiovasc Res 2022, 118, 2610–2624. DOI: 10.1093/cvr/cvab306
4. Dybas, J.; Bulat, K.; Blat, A. et al. Age–Related and Atherosclerosis–Related Erythropathy in ApoE/LDLR−/− Mice. Biochim Biophys Acta Mol Basis Dis 2020,1866. DOI: 10.1016/j.bbadis.2020.165972
5. Wilk, A.; Wilkosz, N.; Rugiel, M. et al. Next-Generation Diabetes Diagnostics: Spectrochemical Staging of Red Blood Cells Using Vibrational spectroscopy. J Adv Res 2026. DOI: 10.1016/j.jare.2026.01.052