High-throughput tracking of single yeast cells in a microfluidic imaging matrix.

High-throughput Analysis of MAPK Signaling in Yeast We demonstrated the throughput, control, and precision of our approach in the analysis of the yeast pheromone mating response. The pheromone response is a well-characterized signaling pathway that serves as an archetypical model of the highly conserved mitogen activated protein kinase (MAPK) signaling cascades in eukaryotic cells. MAPK signaling governs cellular response to a staggering range of stimuli including growth factors, cytokines, hormones, cellular adhesion, stress, and nutrient conditions. 39 In yeast cells MAPK signaling is used for sexual reproduction between haploid cells of type a and α . Mating response is induced by the binding of a soluble pheromone peptide via the membrane-localized G-protein-coupled receptor Ste2 which initiates a phosphorylation cascade, ultimately activating the expression of about 200 genes and culminating in growth arrest and formation of a pointed extension (a shmoo) towards the pheromone gradient. Although this pathway has been the subject of intense study and is likely the most well-characterized of all MAPK networks, the vast majority of analysis has been qualitative and there remain many open questions on how network architecture gives rise to emergent properties including cross-talk and specificity, filtering, adaptation, memory, and cellular variability. 40 - 43 Using a single device run we monitored the response of over 60,000 individual yeast cells from 8 strains (wildtype and 7 deletion mutants) exposed to 16 different α-factor concentrations ( Supplementary Table 2 ). Brightfield and fluorescent images were collected at 20 min time resolution to capture response dynamics and while avoiding unnecessary photo-damage of cells. To correlate mating-pathway activity with morphological phenotype we transformed the strains with a GFP reporter under control of a mating-specific promoter ( Supplementary methods ). During a 24h experiment over 40,000 bright field differential interference contrast and fluorescence images were recorded, representing ∼4 million cell measurements stored in 100 Gb of raw image data. To deal with this volume data we built a custom image analysis pipeline in MATLAB (The Mathworks, Inc., Natick, MA) implementing the following tasks: (i) segmentation and tracking of cells, (ii) calculation of total fluorescence in each cell, (iii) calculation of statistics across all cells and experimental conditions and, (iv) generation of figures, movies, and tables to present time-course and steady-state data ( Supplementary Methods ). In particular, the generation of movies for each chamber enables rapid qualitative analysis of each experiment in order to guide subsequent data analysis and hypothesis testing ( Supplemental Videos 1 , 2 , 3 ). Parallel analysis of multiple strains and conditions allows for the rapid collection of unified data sets that facilitate the precise comparison of response kinetics and noise within a population, between different experimental conditions, and across varying genotypes. Such analysis is particularly important for studying subtle differences that may be obscured by systematic variations between experiments. For example, data from a single experimental run comparing the response of various genotypes to sustained pheromone stimulation at varying concentration is shown in Figure 3 . The kinetics of GFP expression response in wild type cells (GFP normalized to the cell volume) is shown in Figure 3a . Approximately 300 min after induction the population reached a steady state GFP expression with approximately a 12 fold increase over basal levels. We observe pronounced noise in network response 3 , 7 , 44 - 49 with a coefficient of variation of 0.3. The time-dependent dose response of wild type cells is represented by a two-dimensional plot of GFP expression under 16 pheromone concentrations over 600 minutes ( Figure 3b ). Cells show a subtle response at alpha-factor concentrations as low as 1nM, and exhibit a saturated response under stimulation at concentrations in the range of 22-30 nM. Below saturation, the response was found to be monotonically graded with pheromone concentration. The parallel analysis of multiple strains under identical experimental conditions allows for precise characterization of the effect of genetic perturbations to network response ( Figures 3c-i ). We observe that under constant stimulation that ptp2Δ, msg5Δ and kss1Δ deletion strains are hyper-sensitive (in increasing order) while ste50Δ, far1Δ, slt2Δ and fus3Δ deletion strains are hypo-sensitive (in decreasing order) with respect to wild-type, consistent with previous high-throughput and focused studies. 3 , 4 , 18 , 50 In addition, analysis under constant and finely varied concentrations of pheromone reveals different response saturation thresholds amongst mutants. For example, slt2Δ saturates at alpha-factor concentrations as low as 10 nM while ste50Δ does not reach steady state within the time of the experiment.

Experimental Chip fabrication Poly-dimethylsiloxane (PDMS, RTV615 manufactured by General Electric, CT) microfluidic devices were fabricated by replica molding from micromachined masters using multilayer soft lithography as previously described. 24 , 53 , 54 Devices feature 2-layers with the top layer containing channels used for pneumatic valving and the bottom layer containing flow channels and imaging chambers. Device design was completed using AutoCAD software (Autodesk, Inc., San Rafael, CA) and printed on 20,000 dot per inch resolution transparency masks (Cad/Art Services, Bandon, Oregon). Master negative molds were fabricated by standard photolithography techniques on 4 inch (101.6 mm) silicon wafers (Silicon Quest International, Santa Clara, CA). The flow layer consisted of two feature types: 4.4 μm high rectangular cell microchambers, and 9 μm high rounded flow channels. The rounded channel cross-sections were obtained with by placing the wafer on a 130°C for 30 min. Each cell microchamber had a volume of 0.71 nL with dimensions 684 × 260 × 4.4 μm 3 . The 4.4 μm and 9 μm layers were made using negative (SU8-5, Microchem Corp., Newton, MA) and positive photoresist (SPR220-7, Microchem Corp) respectively. The control master was a single layer mold consisting of 25 μm high squared features made with SU8-2025 negative photoresist (Microchem Corp.). Resist processing was performed according to the manufacturer's specifications. The bottom layer of the device was sealed by covalent bonding to a 0.7 mm low auto-fluorescence glass slides (borofloat 33, S.I. Howard Glass Co., Inc. MA) by a 10 s O 2 -plasma surface activation followed by baking for 15h at 80°C.

Segmentation and tracking algorithm

Cell segmentation was performed using only bright-field images. Focus was adjusted during image capture to ensure yeast cell walls were clearly visible as dark continuous borders. The local mean and variance were calculated for each pixel of the image using a small local neighborhood, and those pixels for which the local mean was below a threshold and the local variance was above a threshold were marked as cell wall pixels. 55 Subsequently, a mask of cell areas was obtained by thresholding the local variance image with a threshold found with the help of Otsu's method, 56 followed by a series of operations based on mathematical morphology. Initially recognized cell walls were then removed from the mask and cells that were grouped together were separated by a watershed method ( Supplementary Methods ). The cell tracking was based on minimizing the sum of Euclidean distances between cells from each image frame at time point t with cells from the image frame at time point t-1. If the number of cells at each time point was the same, the solution to this assignment problem giving the smallest total cost (sum of distances between cells) could be readily found by the classic Hungarian algorithm. 57 When this was not the case, due to cell division or segmentation errors, we applied a modified assignment method based on the Hungarian algorithm (see supplementary notes for details).

Additional methods

Description of the yeast constructs as well as further details on chip fabrication and operation are available in Supplementary Methods .

Additional methods

Description of the yeast constructs as well as further details on chip fabrication and operation are available in Supplementary Methods .

Supplementary Material Sup data Video 1 Video 2 Video 3