Droplet Tn-Seq combines microfluidics with Tn-Seq for identifying complex single-cell phenotypes

Abstract

While Tn-Seq is a powerful tool to determine genome-wide bacterial fitness in high-throughput, culturing transposon-mutant libraries in pools can mask community or other complex single-cell phenotypes. Droplet Tn-Seq (dTn-Seq) solves this problem by microfluidics facilitated encapsulation of individual transposon mutants into growth medium-in-oil droplets, thereby enabling isolated growth, free from the influence of the population. Here we describe and validate microfluidic chip design, production, encapsulation, and dTn-Seq sample preparation. We determine that 1-3% of mutants in Streptococcus pneumoniae have a different fitness when grown in isolation and show how dTn-Seq can help identify leads for gene function, including those involved in hyper-competence, processing of alpha-1-acid glycoprotein, sensitivity against the human leukocyte elastase and microcolony formation. Additionally, we show dTn-Seq compatibility with microscopy, FACS and investigations of bacterial cell-to-cell and bacteria-host cell interactions. dTn-Seq reduces costs and retains the advantages of Tn-Seq, while expanding the method's original applicability.

Fig. 1. Schematic overview of droplet Tn-Seq.

a A microfluidic device encapsulates single bacterial cells into droplets containing growth medium. Bacteria are allowed to grow within droplets, genomic DNA (gDNA) is isolated at the start of the experiment (t1) and after growth (t2). Importantly, while growth for each transposon mutant takes place in isolation, gDNA is isolated from the pooled population, enabling screening of all mutants simultaneously. b gDNA is then amplified with DNA polymerase phi29, digested with MmeI, an adapter is ligated, a ~180 bp fragment is produced which contains ~16 nucleotides of bacterial gDNA, defining the transposon-insertion location, followed by Illumina sequencing. Reads are demultiplexed based on the barcode in the adapter and a potential second barcode in primer 1, mapped to the genome, and fitness is calculated for each defined region.

Fig. 2. Microfluidics to generate monodisperse liquid and agarose droplets.

a The droplet microfluidic device design allows syringe pumps to deliver surfactant in a fluorinated oil through tubing to the oil inlet and culture medium containing cells to the aqueous inlet. Filters prevent debris from clogging downstream channels while resistors reduce fluctuation in liquid flow rates. Oil separates the continuous flow of the cell culture into monodisperse droplets at the flow-focus junction. Droplets exit the device through the droplet outlet and are collected. b–e Depending on the size of the channel at the flow-focus junction, and the flow rates of the oil and aqueous phases, uniformly sized liquid or agarose droplets can be formed ensuring each encapsulated cell grows in the same volume of culture media. b, d With optimized flow rates, the average diameter of the droplets is 65–67 µm when using a 40 × 40-µm channel. c Liquid droplets in fluorinated oil plus surfactant. e Gelled agarose droplets with oil removed. Source data are available in the Source Data file.

Fig. 3. Culturing of different bacterial species in liquid and agarose droplets.

a Both Gram-positive and -negative bacteria grow robustly in liquid droplets (red circles), and in a similar fashion to a 8 ml liquid batch culture (blue squares) (each culture was grown at least three times, error bars are standard error of the mean). b Agarose droplets generated by adding low melting agarose to growth medium. Agarose provides structural support and results in bacteria growing in compact microcolonies for both Gram-positive and -negative bacteria. White arrows indicate developed microcolonies. Source data are available in the Source Data file.

Fig. 4. Unbiased whole-genome amplification of low-quantity genomic DNA.

a, b gDNA was prepared by two different methods for transposon sequencing. For the WGA sample, 10 ng of gDNA was amplified first with DNA polymerase phi29 before MmeI digestion and adapter ligation. For the standard sample, 1 μg of gDNA was digested with MmeI, followed by adapter ligation. There is a strong correlation between fitness values obtained from WGA preparation compared with standard Tn-Seq library preparation a, and WGA preparation is highly reproducible b.

Fig. 5. LytB affects gene expression and cell death in a confined space.

a The volcano plot shows genome-wide fitness changes from a dTn-Seq experiment between batch and droplet growth with glucose. Fitness difference is shown on the x-axis along with the associated p-value on the y-axis (one sample t test with Bonferroni correction). All significant genes are highlighted with color, which represent each gene’s functional category shown in the “Gene Category” figure key. b The wild-type (orange) has shorter cell-chain lengths compared with ΔlytB (green) in batch culture, but in droplets chain lengths are similar (c). d Loss of lytB causes a reduced fitness in droplets compared with batch culture. The live cell (CFU) expansion between wt and ΔlytB is similar in batch culture, however, in droplets ΔlytB grows less well then wt (n = 6; p-value is based on a one-way ANOVA with Bonferroni correction for multiple testing, * < 0.05). Shorter-chain lengths and less growth of ΔlytB in droplets could either be caused by slower growth or a higher death rate. e The expression of each gene relative to the control gene SPT_2222. f The change in expression of each gene when comparing droplet to batch. The expression of competence genes comD, comE, comM, and comX are downregulated in wt when grown in droplets, while they become upregulated in ΔlytB. In addition, the cell wall hydrolases cbpD and lytA, which are associated with increased cell lysis and fratricide become highly upregulated in ΔlytB, while only cbpD becomes upregulated in wt. Collectively, this suggests a role for LytB in suppressing the expression of competence-related genes in a confined environment. Each expression experiment consists of at least three biological replicates and three technical replicates each, error bars are standard error of the mean. p-values are based on a one-way ANOVA with Bonferroni correction for multiple testing. ns = not significant, * < 0.05, ** < 0.005, *** < 0.0005, **** < 0.0001. Source data are available in the Source Data file.

Fig. 6. Mutants sensitive to host-specific factors can be compensated by co-culture.

a The volcano plot shows genome-wide fitness changes from a dTn-Seq experiment comparing growth in droplets with either glucose or alpha-1-acid glycoprotein as the main carbon source. Validated genes nagB, SP_1674, nanE, and nagA are circled in red. b While ΔnanE (orange) grows similarly to wt (blue) in glucose, there is a growth defect for the mutant (green) compared with wt (red) when AGP is the sole carbon source. c Individual and mixed culturing combined with bacterial cell enumeration on agar plates containing antibiotics enabling differentiation between the wt and mt were used to determine relative fitness of the mt. While mutants grow just as well as the wt when glucose is the main carbon source (orange bars), they grow significantly slower than the wt when grown independently (green bars). However, their fitness is significantly improved when grown in the presence of the wt (blue bars), indicating that wt is providing “community support” (each growth experiment was performed at least four times, significance was determined through a one-way ANOVA with Bonferroni correction for multiple testing). d Volcano plot comparing growth in droplets in the absence and presence of elastase. e Deletion mutant ΔmscL (green) has reduced survival in two different concentrations of elastase compared with wt (orange) when cultured as single strains. However, ΔmscL survival in elastase is improved, and indistinguishable from wt, when the mutant is cultured in the presence of the wt. “Percent survival” was calculated relative to an untreated control (each experiment was performed at least three times and significance was determined through a two-way ANOVA). All error bars are standard error of the mean; ns = not significant, * < 0.05, ** < 0.005, *** < 0.0005, **** < 0.0001. Source data are available in the Source Data file.

Fig. 7. Capsule genes cpsC and cpsD are expendable in agarose droplets.

a The volcano plot shows genome-wide fitness changes from a dTn-Seq experiment between droplet growth in glucose and droplet growth in 1% agarose. Significant genes are highlighted in color, which represent each gene’s functional category shown in the “Gene Category” figure key. b Wild-type (orange), ΔcpsC (green), and ΔcpsD (blue) were cultured in either liquid batch culture or agarose droplets. In each condition, the relative growth expansion was determined for each strain by counting the number of cells in the population at the beginning of the experiment and at the end after 5 h of culture. Growth expansion of the mutants are relative to the expansion of the wt strain within each growth environment (each experiment was performed at least four times). Growth is improved when the mutants are embedded in agarose (Kruskal–Wallis with Dunn’s test for multiple testing; ns = not significant, ** < 0.005, error bars are standard error of the mean). c Agarose droplets in oil with encapsulated S. pneumoniae, which have developed into compact microcolonies. ΔcpsC and ΔcpsD mutants develop similarly sized microcolonies compared to WT after 5 h. White arrows indicate microcolonies. Source data are available in the Source Data file.

Fig. 8. Bacterial cell–cell and cell–host interaction models.

a–d S. pneumoniae strain sfCSPr that expresses GFP in response to CSP-1 was mixed with strain ADP112 that produces CSP-1 upon IPTG induction. A 40:1 mixture of ADP112:sfCSPr was encapsulated into agarose droplets, the oil removed, and then fluorescence and brightfield microscopy images were captured after 3 -h culture in the absence (a) and the presence (b) of 1 mM IPTG (white arrows highlight GFP-expressing sfCSPr microcolonies). Subsequent FACS analysis of agarose droplets accurately represents the predicted GFP signal frequency in the absence (c) and the presence (d) of 1 mM IPTG. e, f Gelled agarose droplets can be re-encapsulated into another droplet to form a second layer of agarose. Agarose droplets containing sfCSPr were re-encapsulated into media containing 1% agarose in the absence (e) or presence (f) of 560 ng/ml CSP-1. After 2 h of culture, the oil was removed and brightfield and fluorescence imaging revealed no GFP expression for the untreated sample, but positive GFP expression for the CSP-1 treated sample (black arrow highlights non-induced sfCSPr, white arrow highlights GFP-expressing sfCSPr, white asterisk indicates the inner droplet, black asterisk inidcates the outer droplet layer). g–j Agarose/hydrogel droplets containing Yersinia pseudotuberculosis (Yptb) were exposed to murine bone marrow-derived macrophages (BMDMs). Yptb strain IP2666 GFP + was grown in droplets overnight and visualized by fluorescence (g) and brightfield (h) microscopy. After a 1-h incubation, BMDMs can attach to oil-free empty droplets (i) or droplets containing Yptb cells (j) (white arrows indicate Yptb cells and black arrows indicate BMDMs).