Temporally dynamic antagonism between transcription and chromatin compaction controls stochastic photoreceptor specification

Stochastic mechanisms diversify cell fates during development. How cells randomly choose between two or more fates remains poorly understood. In the Drosophila eye, the random mosaic of two R7 photoreceptor subtypes is determined by expression of the transcription factor Spineless (Ss). Here, we investigated how cis-regulatory elements and trans factors regulate nascent transcriptional activity and chromatin compaction at the ss gene locus during R7 development. We find that the ss locus is in a compact state in undifferentiated cells. An early enhancer drives ss transcription in all R7 precursors to open the ss locus. In differentiating cells, transcription ceases and the ss locus stochastically remains open or compacts. In SsON R7s, ss is open and competent for activation by a late enhancer, whereas in SsOFF R7s, ss is compact and repression prevents expression. Our results suggest that a temporally dynamic antagonism, in which transcription drives decompaction and then compaction represses transcription, controls stochastic cell fate specification.

Conformational flexibility of EptA driven by an interdomain helix provides insights for enzyme–substrate recognition

Many pathogenic gram-negative bacteria have developed mechanisms to increase resistance to cationic antimicrobial peptides by modifying the lipid A moiety. One modification is the addition of phosphoethanolamine to lipid A by the enzyme phosphoethanolamine transferase (EptA). Previously we reported the structure of EptA from Neisseria, revealing a two-domain architecture consisting of a periplasmic facing soluble domain and a transmembrane domain, linked together by a bridging helix. Here, the conformational flexibility of EptA in different detergent environments is probed by solution scattering and intrinsic fluorescence-quenching studies. The solution scattering studies reveal the enzyme in a more compact state with the two domains positioned close together in an n-dodecyl-β-D-maltoside micelle environment and an open extended structure in an n-dodecyl-phosphocholine micelle environment. Intrinsic fluorescence quenching studies localize the domain movements to the bridging helix. These results provide important insights into substrate binding and the molecular mechanism of endotoxin modification by EptA.

Interrogation of the Dynamic Properties of Higher-Order Heterochromatin Using CRISPR/dCas9

Eukaryotic chromosomes feature large regions of compact, repressed heterochromatin hallmarked by Heterochromatin Protein 1 (HP1). HP1 proteins play multi-faceted roles in shaping heterochromatin, and in cells, HP1 tethering to individual gene promoters leads to epigenetic modifications and silencing. However, emergent properties of HP1 at supranucleosomal scales remain difficult to study in cells due to lack of appropriate tools. Here, we develop CRISPR-Engineered Chromatin Organization (EChO), combining live cell CRISPR imaging with inducible large-scale recruitment of chromatin proteins to native genomic targets. We demonstrate that human HP1α tiling across kilobase-scale genomic DNA forms novel contacts with natural heterochromatin, integrates two distantly targeted regions, and reversibly changes chromatin from a diffuse to compact state. The compact state exhibits delayed disassembly kinetics and represses transcription across over 600 kilobases. These findings support a polymer model of HP1α-mediated chromatin regulation and highlight the utility of CRISPR-EChO in studying supranucleosomal chromatin organization in living cells.

Playing Atari with few neurons

We propose a new method for learning compact state representations and policies separately but simultaneously for policy approximation in vision-based applications such as Atari games. Approaches based on deep reinforcement learning typically map pixels directly to actions to enable end-to-end training. Internally, however, the deep neural network bears the responsibility of both extracting useful information and making decisions based on it, two objectives which can be addressed independently. Separating the image processing from the action selection allows for a better understanding of either task individually, as well as potentially finding smaller policy representations which is inherently interesting. Our approach learns state representations using a compact encoder based on two novel algorithms: (i) Increasing Dictionary Vector Quantization builds a dictionary of state representations which grows in size over time, allowing our method to address new observations as they appear in an open-ended online-learning context; and (ii) Direct Residuals Sparse Coding encodes observations in function of the dictionary, aiming for highest information inclusion by disregarding reconstruction error and maximizing code sparsity. As the dictionary size increases, however, the encoder produces increasingly larger inputs for the neural network; this issue is addressed with a new variant of the Exponential Natural Evolution Strategies algorithm which adapts the dimensionality of its probability distribution along the run. We test our system on a selection of Atari games using tiny neural networks of only 6 to 18 neurons (depending on each game’s controls). These are still capable of achieving results that are not much worse, and occasionally superior, to the state-of-the-art in direct policy search which uses two orders of magnitude more neurons.

Stability of non-linear filter for deterministic dynamics

<p style='text-indent:20px;'>This papers shows that nonlinear filter in the case of deterministic dynamics is stable with respect to the initial conditions under the conditions that observations are sufficiently rich, both in the context of continuous and discrete time filters. Earlier works on the stability of the nonlinear filters are in the context of stochastic dynamics and assume conditions like compact state space or time independent observation model, whereas we prove filter stability for deterministic dynamics with more general assumptions on the state space and observation process. We give several examples of systems that satisfy these assumptions. We also show that the asymptotic structure of the filtering distribution is related to the dynamical properties of the signal.</p>

Mixing and average mixing times for general Markov processes

Yuval Peres and Perla Sousi showed that the mixing times and average mixing times of reversible Markov chains on finite state spaces are equal up to some universal multiplicative constant. We use tools from nonstandard analysis to extend this result to reversible Markov chains on compact state spaces that satisfy the strong Feller property.

Marked Difference in the Conformational Transition of DNA Caused by Propanol Isomer

We measured the changes in the higher-order structure of DNA molecules (λ phage DNA, 48 kbp) at different concentrations of 1- and 2-propanol through single-molecular observation. It is known that 2-propanol is usually adapted for the procedure to isolate genomic DNA from living cells/organs in contrast to 1-propanol. In the present study, it was found that with an increasing concentration of 1-propanol, DNA exhibits reentrant conformational transitions from an elongated coil to a folded globule, and then to an unfolded state. On the other hand, with 2-propanol, DNA exhibits monotonous shrinkage into a compact state. Stretching experiments under direct current (DC) electrical potential revealed that single DNA molecules intermediately shrunk by 1- and 2-propanol exhibit intrachain phase segregation, i.e., coexistence of elongated and compact parts. The characteristic effect of 1-propanol causing the reentrant transition is argued in terms of the generation of water-rich nanoclusters.