An Ultrasensitive Multiplexing Approach via Integrating Boosted Nanozyme Activity with Multidimensional Information for Antibiotic Recognition

Abstract

Most studies focus on improving the activity of nanozymes to build a nanozyme-based “chemical tongue” for the sensitive discrimination of multiple analytes. However, the construction of sensing units with multidimensional signals is often overlooked, which greatly limits the ultrasensitive identification of antibiotics and even their subtypes. Herein, we have for the first time designed the high performance of a “chemical tongue” by integrating high peroxidase (POD)-like activity Prussian blue analogues (PBAs) with multidimensional information from multiple oxidation products of 3,3′,5,5′-tetramethylbenzidine (TMB) as sensing units. Initially, the bimetallic PBA with boosted POD-like activity is achieved via combining FeRu metal doping with an acid-etching morphological reconstruction strategy. More importantly, the catalytic activity for this hollow-structure nanozyme with a specific activity of 73.3 U/mg is increased by approximately 10 times compared to the pristine Prussian blue (PB) through the synergistic reconstruction of redox couples and exposure of more active sites. And then, the one-electron, one/two-electron, and two-electron oxidation products of TMB are achieved by rationally controlling the concentration of the nanozyme, yielding three chromogenic products as three sensing units with six-wavelength channels. Therefore, the developed array is successfully applied to the ultrasensitive detection of seven antibiotics with down to 0.1 μM and even the discrimination of three subtypes including aminoglycosides, penicillins, and tetracyclines. The sensing mechanism depends on the variable inhibition efficiency of multiple antibiotics on the nanozyme catalytic activity. Furthermore, it enables identification of molar ratios in binary mixtures and also exhibits quantitative analysis of individual antibiotics over a broad concentration range from 0.1 to 200 μM. Moreover, with the assistance of optimizing various machine learning, the concentration-independent recognition model is built against this array, which enormously improves the discrimination accuracy of seven antibiotics from 73.03% by Decision Tree (DT) to 96.63% by the Gaussian Process Classification (GPC) algorithm. The outstanding discrimination capability of this array enables accurate identification of seven antibiotics in three real food samples (honey, milk, and chicken breast). This study provides a new perspective for configuring high-feature sensor arrays in the application of food and environmental samples.