A deep generative model of 3D single-cell organization.

We introduce a framework for end-to-end integrative modeling of 3D single-cell multi-channel fluorescent image data of diverse subcellular structures. We employ stacked conditional ß-variational autoencoders to first learn a latent representation of cell morphology, and then learn a latent representation of subcellular structure localization which is conditioned on the learned cell morphology. Our model is flexible and can be trained on images of arbitrary subcellular structures and at varying degrees of sparsity and reconstruction fidelity. We train our full model on 3D cell image data and explore design trade-offs in the 2D setting. Once trained, our model can be used to predict plausible locations of structures in cells where these structures were not imaged. The trained model can also be used to quantify the variation in the location of subcellular structures by generating plausible instantiations of each structure in arbitrary cell geometries. We apply our trained model to a small drug perturbation screen to demonstrate its applicability to new data. We show how the latent representations of drugged cells differ from unperturbed cells as expected by on-target effects of the drugs.

Efficient Front-Rear Coupling in Neutrophil Chemotaxis by Dynamic Myosin II Localization.

Efficient chemotaxis requires rapid coordination between different parts of the cell in response to changing directional cues. Here, we investigate the mechanism of front-rear coordination in chemotactic neutrophils. We find that changes in the protrusion rate at the cell front are instantaneously coupled to changes in retraction at the cell rear, while myosin II accumulation at the rear exhibits a reproducible 9-15-s lag. In turning cells, myosin II exhibits dynamic side-to-side relocalization at the cell rear in response to turning of the leading edge and facilitates efficient turning by rapidly re-orienting the rear. These manifestations of front-rear coupling can be explained by a simple quantitative model incorporating reversible actin-myosin interactions with a rearward-flowing actin network. Finally, the system can be tuned by the degree of myosin regulatory light chain (MRLC) phosphorylation, which appears to be set in an optimal range to balance persistence of movement and turning ability.

Cytoplasmic Flow and Mixing Due to Deformation of Motile Cells.

The cytoplasm of a living cell is a dynamic environment through which intracellular components must move and mix. In motile, rapidly deforming cells such as human neutrophils, bulk cytoplasmic flow couples cell deformation to the transport and dispersion of cytoplasmic particles. Using particle-tracking measurements in live neutrophil-like cells, we demonstrate that fluid flow associated with the cell deformation contributes to the motion of small acidic organelles, dominating over diffusion on timescales above a few seconds. We then use a general physical model of particle dispersion in a deforming fluid domain to show that transport of organelle-sized particles between the cell periphery and the bulk can be enhanced by dynamic deformation comparable to that observed in neutrophils. Our results implicate an important mechanism contributing to organelle transport in these motile cells: cytoplasmic flow driven by cell shape deformation.

Disentangling Random Motion and Flow in a Complex Medium.

We describe a technique for deconvolving the stochastic motion of particles from large-scale fluid flow in a dynamic environment such as that found in living cells. The method leverages the separation of timescales to subtract out the persistent component of motion from single-particle trajectories. The mean-squared displacement of the resulting trajectories is rescaled so as to enable robust extraction of the diffusion coefficient and subdiffusive scaling exponent of the stochastic motion. We demonstrate the applicability of the method for characterizing both diffusive and fractional Brownian motion overlaid by flow and analytically calculate the accuracy of the method in different parameter regimes. This technique is employed to analyze the motion of lysosomes in motile neutrophil-like cells, showing that the cytoplasm of these cells behaves as a viscous fluid at the timescales examined.

Human Polymerase-Associated Factor complex (PAFc) connects the Super Elongation Complex (SEC) to RNA polymerase II on chromatin.

The Super Elongation Complex (SEC), containing transcription elongation activators/coactivators P-TEFb, ELL2, AFF4/1, ENL, and AF9, is recruited by HIV-1 Tat and mixed lineage leukemia (MLL) proteins to activate the expression of HIV-1 and MLL-target genes, respectively. In the absence of Tat and MLL, however, it is unclear how SEC is targeted to RNA polymerase (Pol) II to stimulate elongation in general. Furthermore, although ENL and AF9 can bind the H3K79 methyltransferase Dot1L, it is unclear whether these bindings are required for SEC-mediated transcription. Here, we show that the homologous ENL and AF9 exist in separate SECs with similar but nonidentical functions. ENL/AF9 contacts the scaffolding protein AFF4 that uses separate domains to recruit different subunits into SEC. ENL/AF9 also exists outside SEC when bound to Dot1L, which is found to inhibit SEC function. The YEATS domain of ENL/AF9 targets SEC to Pol II on chromatin through contacting the human Polymerase-Associated Factor complex (PAFc) complex. This finding explains the YEATS domain's dispensability for leukemogenesis when ENL/AF9 is translocated to MLL, whose interactions with PAFc and DNA likely substitute for the PAFc/chromatin-targeting function of the YEATS domain.