In situ genotyping of a pooled strain library after characterizing complex phenotypes

Abstract

In this work, we present a proof-of-principle experiment that extends advanced live cell microscopy to the scale of pool-generated strain libraries. We achieve this by identifying the genotypes for individual cells in situ after a detailed characterization of the phenotype. The principle is demonstrated by single-molecule fluorescence time-lapse imaging of Escherichia coli strains harboring barcoded plasmids that express a sgRNA which suppresses different genes in the E. coli genome through dCas9 interference. In general, the method solves the problem of characterizing complex dynamic phenotypes for diverse genetic libraries of cell strains. For example, it allows screens of how changes in regulatory or coding sequences impact the temporal expression, location, or function of a gene product, or how the altered expression of a set of genes impacts the intracellular dynamics of a labeled reporter.

Keywords: DuMPLING; live cell; microfluidic; single cell; strain libraries.

Figure 1. The DuMPLING strategy

(1) Pooled strain library generation. (2) Live single cell phenotyping using microscopy. (3) Genotypes recovered by in situ genotyping.

Figure 2. Three strain lac operon knockdown library: Repression network for the three different plasmids used

1. lacY knockdown (lowest LacY‐YPet expression, purple).

2. No knockdown (low LacY‐YPet expression, green).

3. lacI knockdown (high LacY‐YPet expression, blue).

Data information: Color scheme holds throughout this paper.

Figure EV1. DuMPLING oligonucleotide plasmid library design and production

Each member of the oligonucleotide library for plasmid construction contains two divergent promoters: a constitutive promoter toward the spacer and PT7 toward the barcode RNA (top). The oligonucleotide library was amplified using pooled emulsion PCR to avoid the formation of chimeras (bottom middle). The pool of amplified oligonucleotides was assembled into a functional pGuide plasmid library using pooled Golden Gate (bottom right).

Figure EV2. Sequence elements of the dual RNA barcode FISH target and sgRNA expression pGuide vector

Minus (−) signs indicate anti‐sense direction of genetic element.

Figure 3. Mapping phenotypes to genotypes

1. Examples of channels and cells in the custom‐made microfluidic device which are imaged in both phase contrast (top) and fluorescence microscopy (bottom). Phase contrast is used to segment the cells (green outlines), and single‐molecule fluorescence microscopy is used to detect gene expression (red circles in red inset box, which is a blow up of the figure as indicated by the smaller red square and has a change of levels to allow visualization of single molecules) from the lac operon.

2. In situ genotyping with six sequential rounds of FISH probe hybridization and stripping. Cropped images of two cells that are representative of all cells in the trap are shown for the first two rounds (overlay of Cy3 (green) and Cy5 (red) images). The genotype is called by summing the signal in the channel: 0 is assigned for Cy5 (red) and 1 for Cy3 (green). Rectangles indicate assigned genotype (10: lacI knockdown; 01: lacY knockdown; 11: no knockdown).

Figure EV3. Bulk growth rate and CRISPRi repression assay results of the DuMPLING screening strain with different pGuide constructs

1. Mean doubling times of the different strains under the indicated conditions.

2. Steady‐state mean fluorescence normalized by cell density (Fluo/OD₆₀₀) of the different strains under the indicated conditions.

Data information: Error bars indicate sample standard deviations (Bkg n = 3, Empty vector n = 3, P1‐lacY and P2‐lacI n = 6, P3‐control n = 5). Bkg (cell background), AU (arbitrary units), na (not applicable).

Figure 4. Phenotype data

1. Gene expression categorized by assigned genotype.

2. Single‐molecule counting of expression from the two low‐expression genotypes.

3. Top growth curves for one cell lineage (from one channel). Dashed lines indicate the end of detection of a branch. Bottom corresponding lineage tree.

Figure EV4. Reproducibility of dot detection results

1. A–D

   Normalized histograms of single‐molecule counting of expression from the two low‐expression genotypes (strain definitions are given in the Materials and Methods section “Design and construction of the CRISPRi/RNA barcode plasmid library”). (A and C) are the same results displayed in Fig 4B. (B and D) are from a repeat of the same experiment 1 week later. Note: (A and B) are the full histogram, (C and D) are zoom in on the lower frequency events. mn: mean.

Figure EV5. Schematic overview of the enzymatic steps of the probe generation protocol

Templates were amplified from the template library pool by PCR using primers specific for each FISH genotyping round. Lambda exonuclease selectively digested the 5′‐phosphorylated strand, leaving only the 5′‐phosphorothioate strand. Fluorescently labeled and phosphorothioate‐modified elongation probes were hybridized to the ssDNA template and elongated with DreamTaq polymerase. The dsDNA product was digested with restriction enzyme SchI, removing the phosphorothioate bonds from the unlabeled strand. Lambda exonuclease digestion produced the final FISH probe.

Figure EV6. Products from the different steps of the FISH probe production protocol

1. A, B

   Products were run on a 10% polyacrylamide gel and imaged in (A) Cy3 and (B) Cy5 channels. (M): Cy3‐ and Cy5‐labeled 39‐nt and 19‐nt ssDNA probes were used as size references. (1) and (2): the initial fluorescent elongation product for the two rounds of FISH probe generation, respectively. (3) and (4): SchI digestion of the elongation products for rounds one and two, respectively. (5) and (6): Lambda exonuclease treatment and gel‐purified product for rounds one and two, respectively.