Hemocytes in Drosophila melanogaster embryos move via heterogeneous anomalous diffusion

Abstract

Drosophila melanogaster hemocytes are highly motile cells that are crucial for successful embryogenesis and have important roles in the organism’s immunological response. Here we measure the motion of hemocytes using selective plane illumination microscopy. Every hemocyte cell in one half of an embryo is tracked during embryogenesis and analysed using a deep learning neural network. We show that the anomalous transport of the cells is well described by fractional Brownian motion that is heterogeneous in both time and space. LanB1 and SCAR mutants disrupt the collective cellular motion and reduce its persistence due to the modification of laminin and actin-based motility respectively. The anomalous motility of the hemocytes oscillated in time with alternating periods of varying persistent motion. Touching hemocytes appear to experience synchronised contact inhibition of locomotion. A quantitative statistical framework is presented for hemocyte motility which provides biological insights.

Heterogeneous anomalous transport in cellular and molecular biology

“…Interestingly, local (time resolved) anomalous exponents (and also local generalized diffusion coefficients) of individual hemocytes were found to exhibit oscillatory behaviour which could be a manifestation of a positive feedback loop between actin flows and maintenance of cell polarity [299,300] and/or similar periodic shape oscillations to those observed during the run-and-stop motions of T cells [301]. Upon their contact, we found that the hemocyte motion was synchronised, although their clear sense of directional motion was lost [74]. Contact of the cells inhibits their motion (a process of contact inhibitory locomotion) and the anomalous transport of cells continues to be synchronised in terms of the anomalous exponents, generalized diffusion coefficients, displacements, and velocities.…”

“…For a large class of HAT systems (containing anomalous yet Brownian diffusion [115,116,118]), G(x, t|D) is given by a Gaussian distribution and thus non-Gaussian tails for P (x, t) are due to the distribution of diffusion coefficients, P (D). Homogeneous diffusion with a single value for the diffusion coefficient, 74,115,118,123]. The Gamma probability distribution of diffusion coefficients (γ and β are constants), p(D) ∼ D β−1 exp(−γD), produces Laplacian tails in the P (x, t) [124].…”

“…Another artifact of the calculation of MSDs in overlapping windows is the occurrence of spurious positive correlations between α L and D L , D L ∼ exp(α L ). Interestingly, α L and D L of hemocyte cells in Drosophila embryos were found to be anticorrelated [74]. One way to overcome these distortions is to avoid calculations of L-tMSDs altogether, for example by estimating anomalous diffusion and diffusion coefficient via machine learning techniques (see sections 5.14 and 5.15).…”

“…The survival function Ψ(t) can be written in terms of the probability distribution of the first passage time F (t) as Ψ(t) = 1 − F (t). Survival analysis was used to understand the switching between persistent and antipersistent motions of endosomes in human cells and hemocyte cells in drosophila embryos [71,74,160].…”

“…The non-Gaussian form of the probability distributions of the displacements suggests that the motility of hemocytes is heterogeneous. Using a deep learning neural network [71], a broad spectrum of Hurst exponents H (related to anomalous exponents as H = α/2) and generalized diffusion coefficients were found [74]. The probability distributions of Hurst exponents were characteristic of the type of mutant considered.…”