Multi-Dimensional Laser Induced Microfluidic Valve System Based Combinational Antibiotics Susceptibility Screening and Sub-5 Minute Pathogen Identification

Introduction: Overusing broad-spectrum antibiotics, due to the slow process of traditional bulk bacteria identification and antibiotic susceptibility testing (ID & AST), has raised severe public health concerns regarding antibiotic resistance. Despite innovations such as PCR and FISH, these methods still face challenges, such as a high risk of contamination and suboptimal accuracy, which hamper their effectiveness in clinical settings. In response, we herein report the Multi-Dimensional Laser-Induced Microfluidic Valve System as a promising solution for rapid and accurate bacteria ID and AST. The system combines a microfluidic valve-based automated combinational multi-sample assembling platform, a multi-channel laser-induced fluorescent single-cell resolution optical detection system, an innovative thermal permeation-based assay hybridizing fluorescent oligonucleotide probes and 16S ribosomal RNA, and a multi-dimensional Quadratic Discriminant-based machine learning bacteria ID algorithm. This unique combination allows our system to achieve automated bacteria identification with 98% accuracy among 7 different bacteria species in under 5 minutes and high-throughput combinational AST in under 1 hour with high specificity and sensitivity. Taken together, the Multi-Dimensional Laser-Induced Microfluidic Valve System represents a rapid, simple, specific, sensitive, and versatile solution for bacteria ID and AST, and portends an important role in the fight against antibiotic resistance.

Materials and

Methods: The polydimethylsiloxane (PDMS) microfluidic valve-based automated combinational multi-sample assembling chip (Fig. a) is fabricated using photolithography and soft lithography techniques on a silicon wafer. To experiment, the chip is first loaded with multiple fluorescent oligonucleotide probes (PNA probes), bacteria samples, and antibiotics before automated sample assembling by controlling pneumatic valves. (Fig. c i) Afterward, the sample goes through in-line thermal treatment for bacteria membrane permeation, and probe internalization follows single-cell detection using the multi-channel laser-induced fluorescent optical detection system. (Fig. b & c ii) The detected signal is processed and fed into a trained machine-learning algorithm for bacteria ID and AST.

The Multi-Dimensional Laser-Induced Microfluidic Valve System ID applies a combination of target-preferred and universal probes, allowing classification beyond the limit of the number of probes. When probe-targeted bacteria are detected, both probes will enter the bacteria and bind to target their 16S rRNA, resulting in a synchronized green and red fluorescent signal. Whereas for other bacteria, only the universal probe will enter the cell and bind to target 16S rRNA, resulting in only red fluorescent signals (Fig. d). System AST is achieved by quantifying the bacteria 16s rRNA proliferation revealed by the intensity of fluorescent peaks detected prior to and after a short incubation when exposed to antibiotics. An increase in peak intensity after incubation indicates the bacteria is resistant to the exposed antibiotics, whereas susceptible strains will show little detected intensity increase. (Fig. e).

Results, Conclusions, and Discussions: PNA probes are designed to target different binding sites on the 16S ribosomal RNA of different bacteria species with different binding efficiencies, leading to different fluorescent intensities when binding to different bacteria. Using the ratio between the number of fluorescent signals counted in a unit time of each 2 of the 3 fluorescent channels, we are able to generate the scatter plot showing different bacteria species shows different clusters that can be distinguished. (Fig. f) To further improve the bacteria ID accuracy, more features of collected fluorescent signals are fed into a Quadratic Discriminant machine learning algorithm and shows 98% accuracy amount 7 different bacteria species. (Fig, g) To perform bacteria AST, samples are automatically assembled with different concentrations of antibiotics and incubated before thermal permeation and optical detection. The incubation and minimum inhibition concentration result shows a significant increase in the fluorescent signal intensity of detected bacteria after incubation when no antibiotics were used, whereas suppressed proliferation when antibiotics were used, demonstrating the system’s AST capability and clinically relevant MIC. (Fig. h)

Based on the presented results, we demonstrated that the Multi-Dimensional Laser Induced Microfluidic Valve System can perform automated multi-condition bacteria ID in under 5 min and AST in under 1 hour with high accuracy. Our system can significantly accelerate and simplify conventional clinical bulk assays while improving detection sensitivity.

We recognize our thermal permeation cannot promote PNA probe internalization for gram-positive bacteria since their cell wall prevents them from permeation. In future studies, we will discover more permeation methodologies, such as electroporation enabling automated plug-and-play combinatorial AST for both gram-negative and gram-positive species.

Acknowledgements (Optional):