Pneumonia is a respiratory disease that can be caused by many different types of microbes, including bacteria such as Streptococcus pneumoniae and viruses like influenza. The diversity of microbes that can cause pneumonia poses a problem for treatment, as the exact cause is difficult to detect and treating viral influenza with antibiotics would be ineffective, as well as contribute to antibiotic resistance. A team of researchers at MIT have developed new nanoparticle sensors that could help distinguish between viral and bacterial pneumonia, and lead to better treatment for patients while limiting unnecessary antibiotic use.
Rather than attempt to identify the exact species causing the infection, the researchers took the approach of evaluating the body’s response to different types of infection. Viral and bacterial infections provoke distinct immune responses, including the activation of proteases used by cells like T cells, neutrophils and natural killer (NK) cells to fight infection. To identify proteases associated with pneumonia, the team collected 33 publicly available datasets of genes that are expressed during respiratory infections and identified 39 proteases that appear to respond differently to different types of infection. Using this data, the researchers then developed 20 different activity-based nanosensors (ABNs) that can interact with these enzymes, consisting of nanoparticles coated with labeled peptides that can be cleaved by particular proteases. When the peptides are cleaved by the target proteases, their reporter molecules are freed and can eventually be excreted in urine. Urine samples can then be tested for the presence of these mass-encoded reporter molecules via liquid chromatography-tandem mass spectrometry (LC-MS/MS).
The researchers tested their nanosensors using five different mouse models of pneumonia. This included infections caused by Streptococcus pneumoniae, Klebsiella pneumoniae, Haemophilus influenzae, influenza virus and pneumonia virus of mice. The mice were injected with ABNs and their urine was subsequently tested by LC-MS/MS, with the results being analyzed through machine learning methods. Using this approach, the team was able to train algorithms that could differentiate between pneumonia-infected individuals and health controls, and also distinguish between viral and bacterial infections. The researchers found that proteases secreted by neutrophils were most prominently seen in mice with bacterial infections, while protease activity from T cells and NK cells was more prevalent in mice with viral infections. In particular, the protease granzyme B, which triggers programmed cell death, was found to be highly active in the lungs of mice with viral infections. The results of this study were published in the Proceedings of the National Academy of Sciences.
While the ABNs were administered to mice through injections directly into the trachea, the researchers are now developing a method to administer the sensors to humans through a nebulizer or inhaler system. The team also found that while the sensors could distinguish between viral and bacterial infections with high accuracy, the accuracy was lower when attempting to identify specific pathogens, and the researchers are considering the possibility of improving their algorithms to better identify the specific class of microbes causing infection. In the future, the researchers believe their nanosensor system could be used for the development of more rapid urine tests performed at the point-of-care. They have already developed a reduced panel of five sensors that could help facilitate this application, and plan to test this panel further to determine its efficacy in humans.