作者
Ana Espinosa,Alejandra Rojo-Sánchez,Ada Carmona-Martes,Natally Vidal-Figueroa,Antony A. Cardozo-Puello,Lisandro Pacheco-Lugo,Elkin Navarro Quiroz,Eloina Zarate-Peñata,Lorena Gómez Escocia,Nataly J. Galán‐Freyle,Samuel P. Hernández‐Rivera,Joe Villa-Medina,Gustavo Aroca,Elvin S. Lebrón-Ramírez,Maximiliano Méndez‐López,Leonardo C. Pacheco‐Londoño
摘要
Systemic lupus erythematosus (SLE) is an autoimmune disorder whose most severe complicationnephritis in lupus patientsdrives much of the disease's morbidity and mortality. NL is still confirmed by a renal biopsy, an invasive test unsuitable for frequent monitoring. Here, we present a rapid, minimally invasive alternative based on surface-enhanced Raman spectroscopy (SERS) of urine, amplified with citrate-reduced gold nanoparticles (AuNPs) and interpreted by partial least-squares discriminant analysis (PLS-DA). We acquired SERS spectra from 235 urine samples: 72 healthy controls (C), 70 SLE patients without nephritis (L), and 93 biopsy-confirmed Nephritis in Lupus patients (NL). Three binary PLS-DA models were trained and validated by cross-validation (CV) and external prediction through the measures of sensitivity (Sen), specificity (Spe), and accuracy (Acc): C vs L + NL: SenCV = 91.0%, SpeCV = 87.3%, AccCV = 89.8%, and external prediction accuracy = 94.8%; C + L vs NL: SenCV = 84.3%, SpeCV = 85.0%, AccCV = 84.7%, and prediction accuracy = 76.3%; and C vs NL: SenCV = 97.1%, SpeCV = 96.3%, AccCV = 96.8%, and prediction accuracy = 97.6%. Receiver-operating characteristic areas under the curve exceeded 0.96 for the best-performing models, underscoring the robust discriminatory power. Variable-importance-in-projection analysis highlighted Raman bands (260-1430 cm-1) associated with urea, creatinine, amino acid, and lipid vibrations, suggesting metabolic pathways perturbed during LN. Longitudinal assessment of three patients who progressed from SLE to biopsy-confirmed LN demonstrated that the SERS-PLS-DA pipeline detected nephritic signatures up to two years before clinical diagnosis. These results establish AuNP-enhanced SERS of urine, combined with chemometric modeling, as a sensitive and cost-effective platform for differentiating SLE phenotypes and for early, noninvasive detection of lupus nephritis. Deploying this approach in larger prospective cohorts could facilitate routine NL screening and guide timely therapeutic intervention.