Ramanathan Lab

Modern high throughput approaches allow us to measure the levels of thousands of genes in single cells of the embryo, or the activity patterns from thousands of neurons in freely moving animals. Such data provides novel challenges to conventional statistics because of low data density. Thus,  identifying the small number key nodes whose dynamics reflect and control the process of decision making from these data is major challenge. Our paper published in Bioinformatics explains and overcomes some of these challenges.

Our approach has been (see our manuscript led by Furchtgott and Melton) to learn patterns of gene expression from known lineage decisions and exploit this pattern to analyze single cell gene expression data. Using a probabilistic approach, we thus identify key genes whose dynamics reflect the dynamics of decision making (watch multi-potent cells making decisions). This approach in turn allowed us to infer the sequence of lineage decisions giving rise to both excitatory and inhibitory neurons during early human brain development as reported in two manuscripts led by Yao and Close.

While our recent approaches allow us to detect important patterns in large data sets, a fundamental challenge that remains is to gain an understanding of the underlying networks from these large data sets. We want to develop causal predictive models of the underlying network from these data.

The ability to watch cells make decisions has allowed us to as fundamental questions about how cells receive signals, when and how they commit to a decision, how this decision is co-ordinated with the cell cycle and the fate choices of surrounding cells and how the geometry of the embryo affects patterning.

We want to understand human embryonic stem cells and multipotent cells in the mouse embryo receive integrate signals. We discovered that all the BMP and TGF beta receptors in the embryo are basolaterally localized, and the LTA amino acid motif in their sequence governs this localization. In this study led by Zhechun Zhang and Steven Zwick, we also found that mislocalization of the receptors leads to mispatterning. The localization of the receptors in epithelial sheets of cells naturally leads to the coupling of the embryo geometry to patterning. This study leads to the question of how epithelial mesenchymal transition, signaling ligand diffusion and cell movement affect patterning by changing the geometry of the embryo.

The fate decision of a cell depends not only on what signals the multi-potent cells senses, but also when they sense it. The same BMP and TGF beta signals can cause pluripotent cells of the epiblast to adopt a mesendodermal, ectodermal or neural fate depending on when the cells sense the signal. Our ability to read the mind of the cell as it makes developmental decisions along the developmental trajectory, we can prospectively predict what fate each cell will adopt in response to signal. In this recent work led by Jim Valcourt, we exploit this predictive ability using RNA-seq and ATAC-seq to identify and validate candidate transcription factors that can modulate mesendoderm competence. These factors exert their effects by controlling the cell’s progression along the developmental trajectory, by tuning its competence to form mesendoderm at any given point along that trajectory, or by altering both of these aspects. In the classical picture of a Waddington landscape, these effects correspond to changing the cell’s location on the landscape and altering the location of the barrier between fates, respectively.

The ability of the underlying gene regulatory network to modulate these two aspects of the developmental landscape could allow separate control of the dynamics of differentiation and tissue size proportions. We are now exploiting microfluidic devices as in our prior work , human stem cell lines and microscopy to further understand how patterning and tissue proportions are established.

Are there novel progenitor cell types associated with human and in particular cortical development? If so, how are these cell types generated and what do they give rise to? Can we use comparative genomics to rewire the underlying networks generate novel phenotypes?

To determine the molecular identities of progenitor cells in developing human fetal tissue, we developed single cell sequencing methods to extract intact RNA from fixed antibody stained cells from tissue (see Thomsen et al.). This was previously not possible because traditional methods of fixing, permeabilizing and antibody staining cells led to extensive RNA degradation.Through these studies we identified molecular markers for outer radial glial (oRG) progenitors that are enriched in the developing human cortex. We are now working to determine how these cell types arise and which neuronal and glial cell types they give rise to. Following previous work where we could exploit comparative analysis across multiple species to determine how to engineer signaling pathways (see Mody et al.), we are working towards determining how the underlying molecular networks that govern oRG generation.

Finding the key nodes of a complex network that govern specific behaviors is like finding a needle in a haystack. The conventional approach has been to look for these nodes one by one, which is challenging because such a search requires as many measurements as there are nodes, and because specific genetic tools that perturb only one node are often not available. We asked if we could search for the key nodes much more efficient with a novel experimental approach based on Compressed Sensing.

In a first such work led by Jeff Lee and Abdullah Yonar, we could infer the identity key interneurons from among the hundred interneurons in C.elegans, that controlled the speed of locomotion of the animal using just twenty eight measurements. In previous work by Zengcai Guo, we developed microscopy methods to optically perturb and measure calcium activity in animals that were confined. We extended this approach to by building a new microscope that could perturb neural activity in freely moving animals while monitoring their neural activity for more than an hour. Using this microscope. we could validate our inferences.

Our goal is to understand the nervous system sufficiently well to be able to take control of the animal’s behaviors. In work led by Askin Kocabas, we have been able to make the animal think that there is a chemo-attractive gradient in the environment and track this gradient (watch a short YouTube video of C. elegans searching for food). Using a combination of computation, microscopy and optogenetics, we are now trying to understand the different behaviors of the nematode and identifying neurons underlying learning and memory.

We have build microfluidics (Hersen and McClean, Nachman, Zhang  and Zwick), microscopy (Guo, Lee and Yonar, Kocabas) and image processing tools to be able to perform the experiments we want. Such tool building has become an integral part of our lab. We are currently collaborating with the Liu, Weitz and Mooney labs at Harvard to build new and improved tools.

Our past work has involved modeling physical and biological systems using techniques ranging from those employing renormalization group methods to building predictive models to stochastic modeling to understand how noise is filtered both in communications and biological signaling networks. We are increasingly interested in using high dimensional data to build predictive models of the underlying systems using both mechanistic models and those motivated by machine learning.