Protease shaving of Mycobacterium tuberculosis facilitates vaccine antigen discovery and delivery of novel cargoes to the Mtb surface.

Tuberculosis (TB), caused by Mycobacterium tuberculosis (Mtb), is the leading cause of infectious disease death and lacks a vaccine capable of protecting adults from pulmonary TB. Studies have shown that Mtb uses a variety of mechanisms to evade host immunity. Secreted Mtb proteins such as Type VII secretion system substrates have been characterized for their ability to modulate anti-Mtb immunity; however, studies of other pathogens such as Salmonella Typhi and Staphylococcus aureus have revealed that outer membrane proteins can also interact with the innate and adaptive immune system. The Mtb outer membrane proteome has received relatively less attention due to limited techniques available to interrogate this compartment. We filled this gap by deploying protease shaving and quantitative mass spectrometry to identify Mtb outer membrane proteins which serve as nodes in the Mtb-host interaction network. These analyses revealed several novel Mtb proteins on the Mtb surface largely derived from the PE/PPE class of Mtb proteins, including PPE18, a component of a leading Mtb vaccine candidate. We next exploited the localization of PPE18 to decorate the Mtb surface with heterologous proteins and deliver these surface-engineered Mtb to the phagosome. Together, these studies reveal potential novel targets for new Mtb vaccines as well as facilitate new approaches to study difficult to study cellular compartments during infection.

Sequence-to-function deep learning frameworks for engineered riboregulators.

While synthetic biology has revolutionized our approaches to medicine, agriculture, and energy, the design of completely novel biological circuit components beyond naturally-derived templates remains challenging due to poorly understood design rules. Toehold switches, which are programmable nucleic acid sensors, face an analogous design bottleneck; our limited understanding of how sequence impacts functionality often necessitates expensive, time-consuming screens to identify effective switches. Here, we introduce Sequence-based Toehold Optimization and Redesign Model (STORM) and Nucleic-Acid Speech (NuSpeak), two orthogonal and synergistic deep learning architectures to characterize and optimize toeholds. Applying techniques from computer vision and natural language processing, we 'un-box' our models using convolutional filters, attention maps, and in silico mutagenesis. Through transfer-learning, we redesign sub-optimal toehold sensors, even with sparse training data, experimentally validating their improved performance. This work provides sequence-to-function deep learning frameworks for toehold selection and design, augmenting our ability to construct potent biological circuit components and precision diagnostics.