Small regulatory noncoding RNAs (here broadly defined as sRNAs) play critical roles in regulating genes involved in almost all cellular processes, including development, apoptosis, stress responses, tumorigenesis, infection, and immunity. Due to their specificity and versatility, sRNAs have inspired broad applications including new therapies for human diseases (e.g., small hairpin RNAs) and genome engineering (e.g., CRISPR/Cas systems), among other applications. That said, a precise quantitative description of the fundamental mechanism of sRNA-mediated regulation and interference can largely benefit the further improvement of the efficiency and robustness of these applications by providing critical models and parameters. Our research aims at providing a quantitative description at the molecular, cellular, and the systems levels.Using bacteria as model systems,our missions are to understand the molecular mechanisms by which sRNAsmodulate messenger RNA (mRNA) translation and degradation, as well as physiological response caused by sRNA-mediated regulation in the context of pathogenic bacteria-host interactions.

Riboswitches are cis-regulatory RNA elements located upstream of messenger RNAs (mRNA). Binding of ligands to the aptamer domain of the riboswitches can lead to conformational changes in the expression platform of the riboswitches, hereby affecting the transcription or translation of the controlled mRNAs. T-box riboswitches are widely found in Gram-positive bacteria, including pathogens, and are regulating essential genes, including aminoacyl tRNA synthetases, and enzymes or components involved in amino acid biosynthesis and transport. Therefore, T-box riboswitches can potentially be a target for developing new antibiotics. In addition, different from most small molecule binding riboswitches, T-box riboswitches recognize tRNA molecules as ligand, serving an excellent paradigm to understand RNA-based molecular interactions. The regulation mechanism by the T-box riboswitch involves transcription read-through or pre-mature transcriptional termination depending on the aminoacylation status of the bound tRNA. Using single-molecule FRET, we are investigating the binding dynamics of T-box and its tRNA ligand.

Eukaryotic cells are significantly compartmentalized. Dynamic RNA localization to these subcellular compartments profoundly impacts gene expression and other vital cellular activities, and can provide a novel way for stress response and adaptation. In particular, multivalent interactions among certain RNA and protein species can drive the formation of membraneless condensates. Nuclear speckles represent one type of such membraneless bodies in the nucleus of higher eukaryotes, enriched in snRNP species, splicing factors, polyadenylated RNAs (polyA RNAs) and certain long noncoding RNAs (lncRNAs). Formation of nuclear speckle requires the scaffold proteins SON and SRRM2. Eukaryotic pre-mRNA after transcription undergoes a series of processing steps to become mature mRNA, including 5’ capping, splicing to remove introns and ligate exons, 3’ polyadenylation, and export to cytoplasm. Nuclear speckle are suggested to play roles in several mRNA processing steps, including transcription enhancement, splicing quality control and RNA export; and have been implicated in neurodegenerative diseases, infectious diseases and cancers. We are currently investigating why certain RNAs are preferentially localized to nuclear speckles over others, and what are the functional impacts of speckle localization.

The center of a single fluorophore can be very accurately determined in the diffraction limited spot. Taking the advantage of photophysical properties of certain fluorophores that can stochastically blink between the bright and dark states, in single-molecule localization based super-resolution imaging, only a small fraction of fluorophores are activated and localized each time. By repeating the reactivating-and-imaging cycle many times and combining many of such frames together, a super-resolution image can be reconstructed that usually has ~10-fold enhancement in the resolution compared to diffraction-limited fluorescence microcopy. Super-resolution imaging provides a powerful tool for investigating subcellular localization, higher-order architectures of sub-compartments and inter-molecular interactions as well as tracking the motions of individual molecules inside the cell.

Single-molecule fluorescence microscopies are widely nowadays in biophysical research. By fluorescently labeling different molecules or multiple domains on the same molecule, association/dissociation of factors, critical conformational dynamics, and the temporal order of various events can be simultaneously measured, and heterogeneities in the kinetic pathways can be revealed.In particular, single-molecule fluorescence resonance energy transfer (smFRET) is especially powerful for probing conformational changesfor biomolecules, as the energy transfer efficiency is inverse proportional to the sixth power of the distance between a pair ofthe donor and the acceptor dyes, making FRET a very sensitive microscopic ruler at nanometer scale.

We have employed a density based clustering analysis algorithm, DBSCAN, to analyze RNA copy numbers. Spots corresponding to individual localization events in a reconstructed super-resolution image are segregated into clusters based on their spatial density. Clustered data are then superimposed to the DIC image and the boundaries of individual cells were identified using MATLAB code such that clusters are allocated into individual cells. After clustering analysis and cluster allocation, information derived includes: (1) total number of clusters in each cell, which approximates the total number of RNAs in low copy number cases; (2) number of localization spots in each cluster, which we use to build the characteristic distribution of number of spots per RNA; (3) total number of clustered spots in each cell, which is the product of (1) and (2) and is used for estimating the copy number of RNA per cell; (4) average radius of individual clusters; (5) center coordinates of individual cells.

In order to correctly correlate a genotype or phenotype to a specific cell from our imaging experiments, images containing a population of cells must first be properly segmented. We have developed an image analysis package, Seg-3D, for the segmentation of bacterial cells in three-dimensional (3D) images, based on local thresholding, shape analysis, concavity-based cluster splitting, and morphology-based 3D reconstruction. Seg-3D enables a proper segmentation with minimal user input, even when cells are clustered or overlapping in three dimensions. The reconstructed cell volumes allow us to directly quantify the fluorescent signals from biomolecules of interest within individual cells. Seg-3D is an efficient and simple program that can be used to analyze a wide variety of single-cell images, especially for biological systems involving random 3D orientation and clustering behavior, such as bacterial infection or colonization.

In collaboration with Prof. Chris Wiggins group (Applied Physics and Applied Mathematics, Columbia University), during my graduate research, we have developed a variational Bayesian approach that allows us to generalize the concept of maximum likelihood to determining the most likely model (e.g. the number of conformational states in the biological system) as well as the most likely parameters (e.g. the transition rates between conformational states) that best describe the single-molecule time trajectories. We have coded this algorithm into a widely available opensource software package called vbFRET (http://vbfret.sourceforge.net/).