DeLTA 2.0: A deep learning pipeline for quantifying single-cell spatial and temporal dynamics

Abstract

Improvements in microscopy software and hardware have dramatically increased the pace of image acquisition, making analysis a major bottleneck in generating quantitative, single-cell data. Although tools for segmenting and tracking bacteria within time-lapse images exist, most require human input, are specialized to the experimental set up, or lack accuracy. Here, we introduce DeLTA 2.0, a purely Python workflow that can rapidly and accurately analyze images of single cells on two-dimensional surfaces to quantify gene expression and cell growth. The algorithm uses deep convolutional neural networks to extract single-cell information from time-lapse images, requiring no human input after training. DeLTA 2.0 retains all the functionality of the original version, which was optimized for bacteria growing in the mother machine microfluidic device, but extends results to two-dimensional growth environments. Two-dimensional environments represent an important class of data because they are more straightforward to implement experimentally, they offer the potential for studies using co-cultures of cells, and they can be used to quantify spatial effects and multi-generational phenomena. However, segmentation and tracking are significantly more challenging tasks in two-dimensions due to exponential increases in the number of cells. To showcase this new functionality, we analyze mixed populations of antibiotic resistant and susceptible cells, and also track pole age and growth rate across generations. In addition to the two-dimensional capabilities, we also introduce several major improvements to the code that increase accessibility, including the ability to accept many standard microscopy file formats as inputs and the introduction of a Google Colab notebook so users can try the software without installing the code on their local machine. DeLTA 2.0 is rapid, with run times of less than 10 minutes for complete movies with hundreds of cells, and is highly accurate, with error rates around 1%, making it a powerful tool for analyzing time-lapse microscopy data.

Fig 1. Segmentation and tracking of cells within a microcolony.

(A) DeLTA pipeline consists of segmentation, tracking, and lineage reconstruction. (B) Segmentation example with phase contrast image containing an E. coli microcolony, which is input into a U-Net convolutional neural network to obtain segmentation results. (C) Histogram of cell lengths. Inset shows a zoomed-out version with outliers included. (D) Cell tracking between frames. Representative examples of cell tracking with and without division are shown with a phase contrast image of the ‘previous frame’ on the left, a phase contrast image of the ‘current frame’ in the middle, and a greyscale image of the ‘prediction’ on the right. The ‘current frame’ also shows the tracking prediction overlayed. The ‘prediction’ shows the U-Net output with the ground truth overlayed (S1 Fig). (E) Lineage reconstruction keeps track of cell lineages and records pole age. (F) Plot of cell lengths over time. Black line is a representative example of one cell’s length as it grows and divides; all cells in the microcolony are shown in grey.

Fig 2. Resistant and susceptible strains of E. coli on agarose pads containing an inhibitory concentration of tetracycline.

(A) Phase contrast images and associated fluorescence overlays. RFP expressing cells contain a tetracycline resistance gene and GFP expressing cells do not. The magenta and green cell outlines in the fluorescence overlay represent the resistant and susceptible cells, respectively. Region of interest boxes show the areas represented in (B). (B) Representative examples of antibiotic resistant and susceptible cells tracked over time. (C) RFP and GFP fluorescence tracked for individual cells over time. (D) GFP fluorescence versus RFP fluorescence for single cells plotted against growth rate. Fluorescence values are the averages over all the frames for that cell. For growth rate calculations, only cells that were present at t = 150 min were tracked, which is a time point mid-to-late in the movie. The analysis omits those cells that enter the field of view after t = 150 min since the growth rates become noisier with less data. Three resistant cell outliers with growth rates of ~1.4 1/hr are omitted from this view.

Fig 3. Pole age and its impact on growth rate.

(A) Schematic showing how poles are passed down during a division. When a cell divides, the newly formed poles are defined as the ‘new’ poles (white dot) whereas the poles that were passed down from the mother are defined as the ‘old’ poles (black dot). Scale bar, 2 μm. (B) Pole assignment schematic. When the mother cell with known poles divides, the daughter cell that inherits the mother’s old pole is denoted ‘O’ whereas the daughter that inherits the mother’s new pole is ‘N.’ For each generation, either an O or an N is appended to the pole history. (C) Growth rate within each generation. The growth rate of an individual cell is calculated for the period right after the mother’s division until right before the cell divides again. To reduce noise, only cells present for at least three frames were included in the analysis. Daughters (n = 11,246 cells; two tailed unpaired t-test; p-value *** ≤ 0.001), granddaughters (n = 10,726 cells; a one-way ANOVA with post hoc Tukey test used for statistical analysis. Statistical significance: ‘OO’ and ‘NO’ versus ‘NN’ and ‘ON’; p-value ** ≤ 0.01), and great granddaughters (n = 10,217 cells; a one-way ANOVA with post hoc Tukey test used for statistical analysis. Statistical significance: ‘OOO’ and ‘NOO’ versus ‘ONN’,’NNN’,’NON’, and ‘OON’; p-value * ≤ 0.05). Error bars show standard error of the mean.