Exponential phase cells express high levels of ribosomal protein genes

To determine whether T3SS expression is sufficient to stall bacterial growth to the point that cells lose exponential phase markers, we sought to generate a fluorescent reporter that would differentiate between rapidly-growing exponential and slower-growing stationary phase cells. To identify candidate promoter regions for an exponential phase transcriptional reporter, we utilized RNA-seq to determine the transcriptional profile of exponential phase Yersinia pseudotuberculosis as compared to stationary phase cells. Wild-type (WT) Y. pseudotuberculosis cells were cultured overnight (stationary phase) and subcultured in fresh media (exponential phase) at 26^o C to avoid spontaneous virulence plasmid loss, which can occur at a relatively high frequency during prolonged culture at 37^o C [36–38]. We found 894 transcripts increased more than 2-fold in exponential phase, 806 genes increased more than 2-fold in stationary phase, and 2255 protein coding genes were unchanged (S1 and S2 Tables). KEGG pathway analyses indicated that transcripts for genes associated with the ribosome, biosynthesis of secondary metabolites, and aminoacyl-tRNA biosynthesis were heightened in exponential phase (Fig 1A), while transcripts for flagellar assembly and two-component systems were heightened in stationary phase (Fig 1B). Additional groups of genes expressed in stationary phase included genes involved in tyrosine metabolism, ABC transporters, degradation of aromatic compounds, and quorum sensing (Fig 1B). Enrichment of these KEGG pathways was somewhat expected, based on RNA-seq analyses in other Y. pseudotuberculosis strains [39]. Of note, we found transcript levels of 56 genes were greater than 10-fold more abundant in exponential compared to stationary phase, and 22 of these 56 genes were ribosomal protein genes (Table 1). This set included rpsJ, which encodes for the 30S ribosomal protein S10, and whose transcripts were ~13 fold more abundant in exponential phase. rpsJ is the first gene expressed in a large operon of small and large ribosomal subunit protein genes (S1 Fig), suggesting that activity of the promoter upstream of rpsJ could serve as a reporter for transcription of the operon. Consistent with this assumption, all genes in this operon (encoding for: S10, L3, L4, L23, L2, S19, L22, S3, L16, L29, S17) were among the top 56 most highly upregulated genes in exponential phase, with more than 10-fold increased transcript abundance in exponential phase cells (S1 Fig, Table 1). Additionally, rpsJ/S10 expression is transcriptionally regulated by DksA and ppGpp in stationary phase, indicating a reporter containing the rpsJ leader and promoter sequences could reflect DksA and ppGpp-dependent transcriptional regulation of other ribosomal genes as well [40]. For these reasons, we chose the leader and promoter sequences upstream of the S10 ribosomal protein gene (rpsJ) for our ribosomal reporter construct to mark rapidly growing cells.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Genes with highest fold change in exponential compared to stationary phase. Genes with >10 fold enrichment are shown, ribosomal genes are highlighted with bolded text. rpsJ/S10 is highlighted with a black box.

https://doi.org/10.1371/journal.ppat.1012548.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Exponential phase cells express high levels of ribosomal protein genes.

RNA-seq was performed on exponential and stationary phase WT Y. pseudotuberculosis cells. The DESeq2 method of pairwise comparisons was used to determine significant differences in transcript levels. KEGG pathway analyses were utilized to determine the biological pathways overrepresented in (A) exponential and (B) stationary phase cells.

https://doi.org/10.1371/journal.ppat.1012548.g001

Ribosomal reporter expression is heightened in exponential phase cells

To visualize S10 (rpsJ) expression dynamics throughout the cell cycle, an ectopic reporter construct was made using the S10 leader and promoter sequences to drive expression of a ssrA-tagged, destabilized GFP [21,41,42]. The half-life of this destabilized version of GFP is approximately 59 minutes which will allow for visualization of dynamic changes in reporter expression [21]. It is important to note this reporter was generated to detect rapidly growing exponential phase cells based on S10 ribosomal protein expression, in contrast with other approaches that detect ribosomal RNA, which may better represent ribosome abundance [43,44]. After confirming ectopic GFP reporter expression did not affect growth rate (Fig 2A), P_S10::gfp-ssrA cell populations were sampled at defined timepoints to represent different growth stages, and GFP fluorescent intensity within individual bacterial cells was quantified by fluorescence microscopy (Fig 2B). GFP expression significantly increased as cells transitioned from stationary phase (0h) to exponential phase (4h), and reporter signal significantly decreased at 8h (Fig 2B) as cells began to transition back into stationary phase. These results suggest that the S10 reporter can dynamically detect rapidly growing cells.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Ribosomal reporter expression is heightened in exponential phase cells.

(A) Growth curve of WT and P_S10::gfp-ssrA Y. pseudotuberculosis strains at 26^o C. Absorbance (A_600nm) detected by plate reader at the indicated timepoints (hours, h). Four biological replicates shown with mean and standard deviation. (B) Single cell fluorescence microscopy used to quantify the fold change in fluorescence of the P_S10::gfp-ssrA strain at the indicated timepoints, sampled from cultures in (A). Fold change in signal is relative to the mean GFP fluorescence value at 0h, which is shown as a dotted line at a value of 1. Horizontal lines: median values. Data represents 3 biological replicates for each condition. (C-D) qRT-PCR to detect S10, gfp, and 16S transcript levels during culture of the P_S10::gfp-ssrA strain at (C) 26^o C or (D) 37^o C. RNA was isolated at the indicated timepoints. Data represents 6 biological replicates. (E) Representative images used to quantify single-cell fluorescence corresponding with 0h and 4h timepoints in panel B. Statistics: (B) Kruskal-Wallis one-way ANOVA with Dunn’s post-test, ****p < .0001.

https://doi.org/10.1371/journal.ppat.1012548.g002

To confirm changes in GFP fluorescence reflect changes in S10 ribosomal protein transcript levels, qRT-PCR was performed to quantify transcript levels of S10 and gfp, compared to ribosomal RNA (16S). Transcripts of the housekeeping gene, rpoC, were used here for normalization to detect any 16S fluctuation. P_S10::gfp-ssrA cells were cultured overnight, sampled in stationary phase (0h), sub-cultured, and samples were collected at 2, 4, and 8 hours for RNA isolation. As mentioned above, 26^o C is typically used for Y. pseudotuberculosis culture, since incubation of Yersinia at mammalian body temperatures (37^o C) triggers expression of mammalian cell-targeting virulence factors, and prolonged culture can result in loss or inactivation of virulence genes [15,36]. To determine whether the S10 reporter would also detect faster-growing exponential phase cells during culture at 37° C, cells were grown in parallel at 26° C and 37° C. At 26° C, gfp and S10 transcript levels increased between 0 and 2 hours, remained steady at 4h, and began to decline at 8h, while 16S transcripts remained relatively constant (Fig 2D). At 37° C, while 16S transcript levels stayed constant, both gfp and S10 transcript levels peaked at 2h then declined between 4h and 8h, consistent with when bacteria were growing more rapidly (Fig 2E). gfp and S10 transcript levels correlated well at each temperature and throughout the sampled timepoints, suggesting the P_S10::gfp-ssrA reporter is accurately depicting S10 transcript levels. Overall, these results indicate the S10 reporter can be utilized to detect rapidly growing exponential phase cells at either temperature.

Multiple bacterial subpopulations are present within splenic microcolonies

We then sought to utilize the S10 reporter to determine if expression of critical virulence factors, specifically the type-III secretion system (T3SS) alters ribosomal protein expression in individual cells based on S10 reporter expression. In the mouse model of systemic infection, we previously showed that Y. pseudotuberculosis at the periphery of microcolonies are in direct contact with neutrophils and express high levels of the T3SS [20]. Bacteria at the center of microcolonies lack host cell contact, and have an intermediate level of T3SS expression, induced by growth at mammalian body temperatures (37^o C) [22,34]. We hypothesized that the high levels of T3SS expression within peripheral cells would impact their fitness [14], and could result in decreased S10 reporter expression.

To test this hypothesis, the S10 reporter construct was moved into a strain containing a fluorescent yopE::mCherry reporter to detect T3SS expression [20,24]. This is a transcriptional reporter with an internal ribosomal binding site and stable mCherry fluorescent protein gene inserted downstream of yopE in the Yersinia virulence plasmid. To assess relative S10 reporter expression, we also moved the reporter construct into a strain constitutively expressing mCherry (mCherry⁺) [45]. C57BL/6 mice were infected intravenously, and spleens were harvested at day 2 and day 3 post-inoculation to quantify bacterial colony-forming units (CFUs) and visualize reporter expression by fluorescence microscopy. Based on previous studies, we expected significant bacterial growth between these timepoints, allowing us to differentiate between slower-growing and faster-growing individual cells [21,35]. While the CFUs with the yopE::mCherry P_S10::gfp-ssrA strain did not significantly increase, CFUs of the mCherry⁺ P_S10::gfp-ssrA did increase between timepoints (Fig 3A). More microcolonies were detected in cryosections at day 3 (Fig 3B). Microcolonies were significantly smaller in the mCherry⁺ P_S10::gfp-ssrA-infected tissues at day 3, suggesting we captured a timepoint where larger microcolonies were dissociating to form smaller microcolonies (Fig 3B). Similar day 2 and day 3 microcolony areas in the yopE::mCherry P_S10::gfp-ssrA infection may suggest smaller microcolonies were formed and had grown by day 3. We expected both infections to progress similarly, but these data may indicate the mCherry⁺ infection is progressing more slowly.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Multiple bacterial subpopulations are present within splenic microcolonies.

C57BL/6 mice were inoculated intravenously with the mCherry⁺ P_S10::gfp-ssrA (constitutive mCherry), yopE::mCherry P_S10::gfp-ssrA, or yopE::mCherry gfp-ssrA (constitutive GFP) strains. (A) CFU/spleen was quantified at days 2 and 3 post-inoculation. Dots: individual mice. (B) Microcolony areas from spleens were quantified by fluorescence microscopy at the indicated timepoints. Dots: individual microcolonies, representing all bacteria visualized within tissues (including single cells). (C) Representative images of microcolonies: day 2 and day 3 mCherry⁺ P_S10::gfp-ssrA. (D) Reporter expression (ratio GFP/mCherry signal intensity) was quantified at the indicated timepoints and spatial locations (centroid: average of 4 measurements clustered at the geometric centroid of each microcolony, periphery: average of 4 measurements taken every 90^o around the microcolony periphery). Horizontal dotted line represents a value of 1, equal amounts of GFP and mCherry signal. (E) Linear regression data indicating slope and R2 value of best fit lines for S10 centroid values compared to microcolony areas at the indicated timepoints. (F) Representative images of microcolonies: day 2 and day 3 yopE::mCherry P_S10::gfp-ssrA. (G, H) Linear regression data indicating slope and R2 value of best fit lines for GFP compared to mCherry signal of microcolonies at the indicated spatial locations at (G) day 2 and (H) day 3. Horizontal dotted line at y = 250 separates low and high GFP signal. (H) Microcolonies with low levels of yopE::mCherry signal (<100) were analyzed separately (open circles). Dots represent individual mice (A) or individual microcolonies (all other panels). Statistics: (A-B) Mann-Whitney; (D) Wilcoxon matched-pairs; ***p < 0.001, *p < .05, ns: not-significant. (E, G, H) Significantly non-zero slope indicates correlation between values.

https://doi.org/10.1371/journal.ppat.1012548.g003

To analyze S10 reporter expression, we first used the mCherry⁺ P_S10::gfp-ssrA strain to directly compare S10 expression to constitutive mCherry and calculated relative S10 expression based on GFP/mCherry values. The constitutive mCherry signal was used to account for higher bacterial cell density at the geometric centroid of microcolonies vs. the periphery, and baseline cell-to-cell variability in transcriptional and translational activity. Peripheral cells had increased relative S10 expression compared to the centroid at day 2, however there was no spatial difference in S10 expression at day 3 (Fig 3C, 3D). This patterning indicates several things: S10 expression differs from constitutive expression, S10 expression becomes more spatially uniform at day 3, and these data hint that peripheral cells can express higher levels of S10. Interestingly, we also observed a strong positive correlation between centroid S10 levels at the microcolony centroid and areas at day 2, while at day 3, there was an inverse correlation between S10 signal and microcolony areas (Fig 3E). This is consistent with our previous results that larger microcolonies begin to run out of nutrients by day 3 [21,35], and suggests that smaller, newly founded microcolonies exhibit increased S10 expression at day 3 (Fig 3B, 3E).

To explore the relationship between yopE and S10 expression, we generated correlation plots of the mCherry and GFP fluorescent signals at the centroid and periphery of individual microcolonies. At day 2, the data suggest that microcolonies express both S10 and yopE, which contradicts our hypothesis that high T3SS expression will correlate with low S10 expression (Fig 3F, 3G). However, given the smaller sample size of microcolonies at this timepoint, there was not a correlation between signals. At day 3, the larger number of microcolonies allowed us to also analyze values separately based on levels of yopE expression (greater or less than mean fluorescent intensity of 100). yopE high cells (MFI ≥ 100) at the centroid and periphery had a weak positive correlation between yopE and S10 levels (Fig 3H). In contrast, yopE low cells (MFI < 100) displayed some higher levels of S10 expression (MFI > 250), as evidenced by a best-fit line with a negative slope (Fig 3H). These results indicate that multiple populations emerge within the mouse spleen: a surprising population with higher yopE expression that retain S10 expression, bacteria with lower yopE expression that retain higher S10 expression, consistent with our initial hypothesis, and bacteria with low expression of both. We also noted the high S10 values in this infection set compared to mCherry⁺ (Fig 3E, 3H), which again suggest active growth of microcolonies (Fig 3B).

T3SS induction slows growth, but cells can support high levels of S10 expression

Outside of a mammalian host, T3SS expression can be induced in Yersinia cultures by magnesium oxalate (MgOx) treatment resulting in slowed growth [15,38]. This induction strategy involves adding equimolar amounts of MgCl₂ and sodium oxalate, which dissociate to form NaCl and MgOx [46]. This strategy will add Na⁺ and Mg²⁺ cations in the absence of added calcium (Ca²⁺), and calcium oxalate precipitation of residual calcium will result in low calcium conditions trigger high levels of T3SS expression [15,47]. It is important to note incubation at 37^o C induces a low level of T3SS expression, and MgOx addition heightens expression to a growth-inhibited, full induction state. To explore if growth inhibition results in decreased S10 expression, the yopE::mCherry P_S10::gfp-ssrA reporter strain was grown at 37^o C in the presence or absence of MgOx to induce T3SS expression. As expected, growth curves demonstrated significant growth inhibition at 3h and 4h post-MgOx treatment (Fig 4A). T3SS expression was significantly higher in +MgOx treated samples compared to –MgOx between 2h-4h of treatment, correlating well with growth inhibition (Fig 4B, 4C). However, these data suggest T3SS expression may precede growth inhibition, based on the timing of reporter signal detection (2h) and growth inhibition (3h) (Fig 4A, 4C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. T3SS induction slows growth, but cells can support high levels of S10 expression.

WT and yopE::mCherry P_S10::gfp-ssrA strains were cultured at 37^o C in the presence (+, T3SS-induced) or absence (-) of MgOx for the indicated timepoints (hours, h). (A) Growth curve of strains with and without MgOx. Absorbance (A_600nm) detected by plate reader at the indicated timepoints. Statistics compare the yopE::mCherry P_S10::gfp-ssrA strain in the presence (+) or absence (-) of MgOx. Mean and standard deviation are shown. (B) Representative fluorescence microscopy images of bacteria from (A) immobilized on 1% agarose pads. Fold change in (C) mCherry (yopE::mCherry) reporter signal and (D) GFP (P_S10::gfp-ssrA) reporter signal in the absence (-) or presence (+) of MgOx. Values quantified in individual bacteria by fluorescence microscopy. Thresholding was used to select individual cell events, and single cell fluorescence was normalized to the average fluorescent value at 0h (value of 1, represented by dotted line). Each dot: individual cell, horizontal lines: median values. (E) Quantification of single cell bacterial areas (µm2) from samples in panels (C) and (D). Horizontal lines: median values. (F) Correlation plot of fold change in single cell mCherry and GFP fluorescence from bacteria in panels (C-E) cultured in the presence (+) of MgOx. Black arrows denote trends in fluorescence across timepoints. (G) Linear regression data indicating slope, R2, and significance of the lines of best fit shown in (F). All data represent 3 biological replicates for each strain and condition in this figure. Statistics: (A) Two-way ANOVA with Tukey’s multiple comparison test; (C-E) Kruskal Wallis one-way ANOVA with Dunn’s post-test; ****p < 0.0001, **p < .01, ns: not-significant. (F,G) Significantly non-zero slope indicates correlation between values.

https://doi.org/10.1371/journal.ppat.1012548.g004

Interestingly, while S10 expression generally correlated with growth, the link between S10 expression T3SS expression varied (Fig 4A, 4C, 4D). At the 1h timepoint, untreated cells exhibited higher median T3SS expression, due to temperature, and lower median S10 expression. Over time, MgOx treated cells increased expression of the T3SS and decreased in S10 signal, specifically at 3h post-treatment when growth was also arrested (Fig 4A, 4C, 4D). In contrast, at 4h we observed the median S10 reporter signal was higher in treated cells, largely due to a subset of bacterial cells that retained high S10 expression (Fig 4D). Additionally, MgOx treatment resulted in a smaller median cell area, suggesting a decreased division rate and decreased metabolic activity (Fig 4E) [48]. The difference in median cell size was lost at the 4h timepoint, suggesting untreated and treated cells may be approaching stationary phase.

At the 4h timepoint, we noticed subsets of cells that retained high S10 expression and cells with lower T3SS expression (Fig 4C, 4D). To determine if these were one subpopulation or represented distinct sets of cells, we plotted the fold change of each fluorescent marker for MgOx treated cells. In the first hour of culture S10 expression is strongly induced in all cells, likely capturing their transition out of stationary phase, as seen by a shift up the y-axis (in red) relative to the 0h timepoint (in gray) (Fig 4F). As also seen with transcript levels in Fig 2D, S10 expression levels appear highest at 1h-2h of culture (red, green), and gradually decrease (Fig 4F). After the initial increase, S10 levels generally decreased as T3SS expression increased; evidenced by a shift in the population towards the bottom right of the graph (Fig 4F).

To our surprise, S10 levels did not decrease universally. At 4h post-treatment (in orange) two populations of cells emerged; a population with low S10 and lower T3SS expression and a population with high S10 and high T3SS expression (Fig 4F). This contrasts with our initial hypothesis, and indicates a subpopulation of cells is capable of expressing high levels of both T3SS and ribosomal protein genes (S10). The linear regression analysis was consistent with a positive correlation between the two reporter signals at 4h, based on a significantly non-zero positive slope and R² value of 0.5234 (Fig 4G). There was very little correlation between S10 and T3SS levels in the absence of MgOx based on the slope and R² values of the best-fit lines, likely because T3SS expression was low under these conditions (S2 Fig).

Our experiments in culture indicate that a subset of bacteria exhibit high T3SS activity and high levels of S10 expression, contrary to our initial hypothesis. Subpopulations of bacteria with high T3SS expression and high translational activity have also been observed in cultured Pseudomonas aeruginosa [49].

MgOx treatment is not sufficient to alter S10 reporter activity

We next asked whether MgOx treatment alone, independent of temperature, is sufficient to impact bacterial growth and S10 reporter expression. Bacteria were incubated in the presence or absence of MgOx at 26^o C, absorbance measurements were taken, and cells were collected for fluorescence microscopy. The yopE::mCherry strain (lacking P_S10::gfp-ssrA) and non-fluorescent WT strains were used as controls to ensure growth was not impacted by expression of either reporter (yopE::mCherry or P_S10::gfp-ssrA). There were no significant differences, indicating the reporters did not impact growth and that MgOx treatment, without a temperature shift to 37^o C, was not sufficient to inhibit bacterial growth (S3A Fig). Fluorescence microscopy indicated MgOx treatment resulted in low levels of yopE::mCherry expression (S3 B, C Fig). Levels of T3SS induction were very low relative to MgOx-treated cells at 37^o C, and treated cells at 26^o C no longer had T3SS induction at 4h post-treatment (S3C Fig). In the absence of significant growth inhibition, S10 levels were minimally impacted by treatment, however MgOx-treated cells had lower S10 expression at 3h post-treatment (S3 D Fig), and we observed a decrease in cell size 2h-4h after MgOx treatment (S3 E Fig). The decreased cell size could suggest that MgOx treatment has other impacts on cell physiology (S3E Fig).

Lack of T3SS expression was associated with lower S10 expression

To further assess the possible link between T3SS expression and S10 reporter expression, we transformed the P_S10::gfp-ssrA reporter plasmid into a plasmid-cured Y. pseudotuberculosis strain lacking the virulence plasmid (P(-)) [36,47], a 68kb extrachromosomal plasmid that encodes for the T3SS structural proteins and associated effector proteins [36,50]. It is important to note there are additional genes encoded on the virulence plasmid, notably yadA, a temperature-regulated adhesin and known virulence factor that promotes injection of T3SS effector proteins into host cells [31]. We first utilized the P(-) strain for culture-based experiments to assess the role of T3SS expression, since we expected the impact of YadA to be minimal under these conditions.

WT P_S10::gfp-ssrA and P(-) P_S10::gfp-ssrA strains were treated with MgOx and grown at 37^o C. As expected, the P(-) P_S10::gfp-ssrA strain had significantly higher absorbance values than WT P_S10::gfp-ssrA after 4h culture, indicating the presence of the virulence plasmid impaired growth (Fig 5A). To determine if differences in growth correlated with changes in S10 levels, we detected S10 reporter expression by fluorescence microscopy. The WT P_S10::gfp-ssrA and P(-) P_S10::gfp-ssrA strains had similar levels of S10 expression at 0h, although the WT strain was slightly higher (Fig 5B). Surprisingly, S10 expression levels continued to increase in the WT strain, and WT exhibited significantly higher S10 levels than the P(-) strain at both 2h and 4h, despite significantly slower growth (Fig 5A, 5B). To better assess the impact of T3SS expression on S10 expression, we completed experiments with ∆lcrF and ∆iscR strains, which, respectively, lack and exhibit lower T3SS expression under these conditions [51,52]. We observed the expected growth patterns with these strains at 37^o C + MgOx, where the WT strain exhibited significantly slower growth than both mutants, and the ∆lcrF strain, which lacks T3SS expression, exhibited the most rapid growth (Fig 5C). Surprisingly once again, we detected significantly higher S10 levels in the WT strain compared to the ∆lcrF strain lacking T3SS expression, and no difference at the 2h and 4h timepoints between the WT and ∆iscR strains (Fig 5D). To determine if this apparent discrepancy in the growth curve and S10 levels was due to differences in cell morphology, specifically if smaller cells were exhibiting higher S10 signal because the GFP molecules are condensed in a smaller space, we quantified the single cell areas of each strain. While WT cells were significantly smaller than the mutants at 2h, there was no difference at 4h, a timepoint that also showed different S10 values (S4 Fig). These data strongly indicate that T3SS expression can result in both growth arrest and retention of high S10 expression levels, and provides some explanation as to why we detected cells expressing both T3SS and S10 in the mouse model and in culture.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Lack of T3SS expression was associated with lower S10 expression.

WT P_S10::gfp-ssrA, and P(-) P_S10::gfp-ssrA (virulence plasmid-cured) strains were cultured at 37^o C in the presence (+) of MgOx for the indicated timepoints (hours, h). (A) Growth curve of strains, absorbance (A_600nm) was detected by plate reader at the indicated timepoints. Mean and standard deviation are shown. (B) Fluorescence microscopy to quantify single cell fold change in GFP (P_S10::gfp-ssrA) reporter signal. Single cell fluorescence was normalized to the average fluorescent value of the WT strain at 0h (value of 1, represented by dotted line). Each dot: individual cell, horizontal lines: median values. (C) WT P_S10::gfp-ssrA, ∆lcrF P_S10::gfp-ssrA, and ∆iscR P_S10::gfp-ssrA were cultured at 37^o C in the presence (+) of MgOx. Growth curve of strains, absorbance (A_600nm) was detected by plate reader at the indicated timepoints. Statistics compare the WT strain to other strains. Mean and standard deviation are shown. (D) Fluorescence microscopy to quantify single cell fold change in GFP reporter signal. Single cell fluorescence was normalized to the average fluorescent value of the WT strain at 0h (value of 1, represented by dotted line). Each dot: individual cell, horizontal lines: median values. All data represent 3 biological replicates for each strain and timepoint in this figure. Statistics: (A, C) Two-way ANOVA with Tukey’s multiple comparison test; (B, D) Kruskal Wallis one-way ANOVA with Dunn’s post-test; ****p < 0.0001, **p < .01, *p < .05, ns: not-significant.

https://doi.org/10.1371/journal.ppat.1012548.g005

Droplet-based microfluidics can be used to model clustered bacterial growth

One limitation of in vivo infection models is the inability to track growth and reporter expression profiles over time. In contrast, culture-based experiments can enable kinetic analyses, but lack the clustered spatial architecture of bacteria seen within the host environment. To circumvent these issues, droplet-based microfluidics can be used to generate local agar environments that force bacteria to grow in tightly-clustered colonies, as they would within host tissues [20,53]. Here, we employed microfluidic droplets to observe growth and reporter expression over time during incubation at 26^o C and 37^o C, and at 37^o C in the presence and absence of MgOx. Agarose-based microfluidics droplets were generated as previously described, where liquid and oil phases intersect to generate oil-encapsulated spherical droplets, and an appropriate bacterial density in the liquid phase was used to ensure a majority of droplets were seeded with a single bacterial cell [53]. Oil was removed from the exterior of droplets to allow free diffusion of molecules into and out of the droplets during culture [53]. We hypothesized that MgOx addition during culture at 37^o C would result in high levels of T3SS induction, growth inhibition, and may result in either higher or lower S10 levels. Based on results in culture, we hypothesized that incubation at 37^o C would not significantly delay droplet growth relative to 26^o C, since this should result in intermediate T3SS induction.

We first explored if temperature impacted growth by encapsulating the yopE::mCherry P_S10::gfp-ssrA strain into droplets and culturing at 26^o C or 37^o C. Cultures were sampled every two hours to quantify bacterial growth, T3SS reporter signal, and S10 reporter signal by fluorescence microscopy. Bacterial clusters grew at both temperatures, but were significantly larger at 37^o C (Fig 6A). While microcolonies had significant T3SS induction at 4h and 6h of growth at 37^o C (Fig 6B), there was no change in S10 reporter signal between conditions (Fig 6C). These results corroborate planktonic cell culture experiments where growth at 37^o C moderately induces T3SS expression but does not reduce growth rate (Fig 4A-C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Droplet-based microfluidics can be used to model clustered bacterial growth.

The yopE::mCherry P_S10::gfp-ssrA strain was encapsulated into droplets and cultured at the indicated temperatures in the presence (+) or absence (-) of MgOx. At the indicated times aliquots were taken and bacterial growth and reporter expression were quantified by fluorescence microscopy. Growth at 26^o C was compared to 37^o C, and (A) area (µm2), (B) fold change in mCherry signal (relative to 0h), and (C) fold change in GFP signal (relative to 0h) were quantified. Horizontal dotted lines represent a value of 1, the initial average value at time 0h. Growth at 37^o C in the presence (+) or absence (-) of MgOx was compared, and (D) area (µm2), (E) fold change in mCherry signal (relative to 0h), and (F) fold change in GFP signal (relative to 0h) were quantified. (G, H) Linear regression data indicating slope and R2 value of best fit lines for GFP compared to mCherry signal of microcolonies grown (G) 4h or (H) 6h at 37^o C, (+) or (-) MgOx. (I) mCherry/GFP signal ratios are shown at the indicated spatial locations for microcolonies grown -/ + MgOx. (J) Representative images of microcolonies grown within droplets under the indicated conditions, all images represent 6h timepoints. Differential Interference Contrast (DIC) shown alongside GFP, mCherry, and merged fluorescent channels. Dotted line outlines the periphery of each droplet. Scale bar: 20µm. All data represent 2 biological replicates for each condition in this figure. Mean and standard deviation are shown. Each dot represents an individual microcolony. Statistics: (A-F) Two-way ANOVA with Tukey’s multiple comparison test; (G, H) Linear regression, significantly non-zero slope indicates correlation between values; (I) Wilcoxon matched-pairs; ****p < 0.0001, ***p < .001, **p < .01, *p < .05, ns: not-significant.

https://doi.org/10.1371/journal.ppat.1012548.g006

While temperature would be sensed by the entire population during infection, local chemical mediators, such as those released from surrounding neutrophils or monocytes [20], can create chemical gradients experienced most strongly by peripheral cells in the microcolony. In the droplet system, we would expect MgOx to result in uniform low calcium conditions. During MgOx treatment at 37^o C, we observed significant droplet growth inhibition at 4h and 6h post-treatment (Fig 6D) and high levels of T3SS induction (Fig 6E) after 6h of culture. These results mirror experiments in Fig 4 in bacterial media, where strong T3SS expression results in growth inhibition (Fig 4A, 4C). As also seen in previous figures, S10 expression did not directly correlate with growth inhibition. At the 4hr timepoint S10 expression was significantly lower in the growth inhibited +MgOx group, but S10 levels were similar at the 6h timepoint (Fig 6F). Interestingly, correlation plots clearly showed that T3SS expression increased between 4h and 6h without an impact on S10 levels, again uncoupling T3SS expression, growth arrest, and S10 levels (Fig 6G,6H).

To determine if the local position in the droplet community impacts reporter expression, we quantified the centroid and peripheral fluorescence values within individual droplets at the 6h timepoint, when T3SS reporter signal was highest. While we have consistently seen a significant increase in T3SS signal at the periphery of microcolonies within host tissues (Fig 3F) [20], we did not observe spatial differences in reporter signal within droplets (Fig 6I, 6J). This ready diffusion of compounds could be beneficial for future applications of this technology, where researchers could generate uniform microenvironments within a single droplet or even droplet configurations. These results also suggest that contact with the agarose matrix of droplets, which could mimic surface contact, is not sufficient to mimic neutrophil contact within host tissues.

T3SS induction reduces gentamicin susceptibility, but S10 expression predicts survival

We hypothesized that the growth arrest observed with MgOx treatment may be sufficient to decrease antibiotic susceptibility, based on known associations between slowed growth, reduced metabolic activity, and decreased antibiotic susceptibility [54,55]. To test this hypothesis, we cultured the yopE::mCherry P_S10::gfp-ssrA strain in the presence and absence of MgOx, then exposed bacteria to either gentamicin or ciprofloxacin (Fig 7A). These antibiotics were chosen because they are bactericidal, albeit with distinct drug targets, the ribosome and DNA gyrase, respectively. MgOx treatment significantly reduced the susceptibility of Y. pseudotuberculosis to gentamicin, as evidenced by increased survival during gentamicin exposure (Fig 7B). MgOx treatment was also sufficient to promote tolerance to ciprofloxacin (Fig 7C). However, it has been well-established that the presence of metal cations, including Mg²⁺, can reduce ciprofloxacin activity [56]. To determine whether altered antibiotic susceptibility was linked to T3SS expression, we performed additional experiments with a ∆lcrF strain, which lacks T3SS induction [52]. While differences between the groups didn’t reach statistical significance, the ∆lcrF strain had similar growth kinetics in the presence and absence of MgOx and ∆lcrF gentamicin sensitivity was similar in the presence or absence of MgOx (Fig 7D), suggesting that the reduced sensitivity of the WT strain to gentamicin was linked to T3SS induction. In contrast, ∆lcrF had similar ciprofloxacin sensitivity compared to the WT strain in the absence of MgOx, and in the presence of MgOx, the ∆lcrF continued to replicate during ciprofloxacin exposure, consistent with resistance (Fig 7E). This suggests that reduced ciprofloxacin sensitivity was instead due to cation interaction with the antibiotic, and not T3SS expression.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. T3SS induction reduces gentamicin susceptibility, but S10 expression predicts survival.

Strains were grown in the presence (+) or absence (-) of MgOx for 2h at 37o C. Antibiotics were added (gentamicin: gent, ciprofloxacin: cipro) and strains were cultured an additional 4h. Samples were taken at 0h (addition of antibiotics), 2h, and 4h to quantify CFUs or quantify fluorescence by microscopy. (A) Experimental design. Image Created in BioRender. Cotten, K. (2025) https://BioRender.com/b41q462 (B) yopE::mCherry P_S10::gfp-ssrA strain + /- MgOx, gent treatment. (C) yopE::mCherry P_S10::gfp-ssrA strain + /- MgOx, cipro treatment. Fold change in CFUs is calculated by normalizing the number of viable bacteria at 2h and 4h to the number of viable bacteria at 0h. (D) Comparison between non-fluorescent WT and ∆lcrF strains, + /- MgOx, gent treatment. Raw CFU values are shown. LOD: limit of detection (horizontal dotted line). (E) Comparison between non-fluorescent WT and ∆lcrF strains, + /- MgOx, cipro treatment. Raw CFU values are shown. Fold change in (F) mCherry (yopE) and (G) GFP (S10) signal from cells represented in (B) was calculated by normalizing the single cell fluorescence to the average fluorescent untreated (unt) value at 0h (value of 1, dotted line). Untreated 0h, 2h, and gentamicin 2h are shown. Each dot: individual cell, horizontal lines: median values. Dead cells were excluded using the fixable blue dead cell stain. (H) Linear regression data indicating slope and R2 value of best fit lines for GFP compared to mCherry signal of bacteria in the presence of MgOx, either untreated (unt) 2h or gent treatment 2h. Statistics: (B-E) Two-way ANOVA with Tukey’s multiple comparison test, statistics shown represent comparisons + /- MgOx; (F, G) Mann-Whitney; (H) Linear regression, significantly non-zero slope indicates correlation between values; ****p < 0.0001, ***p < .001, *p < .05.

https://doi.org/10.1371/journal.ppat.1012548.g007

To eliminate the impact of magnesium cations on antibiotic activity, we also performed experiments in M9 minimal media, which are conditions known to induce T3SS expression [51]. While we observed T3SS expression during culture in M9, we did not observe WT growth arrest relative to the ∆lcrF strain (S5A Fig), and consistent with this, we also did not observe changes to antibiotic susceptibility (S5B Fig). It is important to note we prepared our M9 media using the protocol in Miller JH 1972 [57] instead of the recipes used in Cheng LW et al. [58] and Miller HK et al. [51], and the calcium in our media explains why we did not observe growth arrest.

To determine the reporter expression profile of bacterial survivors following gentamicin treatment, we performed fluorescence microscopy with the yopE::mCherry P_S10::gfp-ssrA strain after 2h of gentamicin treatment. 2h was chosen because we observed significant changes in viability (Fig 7B) and because this timepoint correlates with the 4h MgOx timepoint of Fig 4 and Fig 5. We quantified fluorescent reporter expression of surviving, viable bacterial cells by incorporating a live/dead fluorescent dye to exclude non-viable cells from our analyses (see Methods). When compared with untreated samples (2h), bacteria that survived gentamicin treatment had significantly lower mCherry (Fig 7F) and lower S10 (Fig 7G) reporter signal. A comparison of the two reporter signals within individual cells indicated a specific loss of cells expressing high levels of S10, some of which also had high yopE::mCherry expression (Fig 7H).

Gentamicin treatment selects for surviving bacteria with reduced S10 expression in the mouse spleen

To determine if gentamicin selects for surviving bacterial cells with lower S10 levels during infection, we set-up a gentamicin treatment mouse model. Experiments were performed with the yopE::mCherry P_S10::gfp-ssrA strain, and in parallel, we performed experiments with a yopE::mCherry-ssrA P_S10::gfp-ssrA strain containing a destabilized version of the mCherry T3SS reporter. The reporter expression patterns were very similar with the two strains, although the overall magnitude of signal was lower with the destabilized reporter, and fewer cells seemed to express very high levels of the T3SS (Fig 8A). Given the similar reporter patterns, we tested gentamicin sensitivity with the stable T3SS reporter (yopE::mCherry P_S10::gfp-ssrA) strain. We harvested untreated spleens at the 48h timepoint, then treated a subset of mice at 48h with a single intraperitoneal injection of 40mg/kg gentamicin, which should result in rapid systemic delivery of inhibitory concentrations of gentamicin, meeting or exceeding 10µg/ml (MBC₉₉) [59]. Spleens were harvested 4h post-treatment and we found gentamicin treatment significantly reduced the bacterial load at this timepoint (Fig 8B). We processed tissues for fluorescence microscopy and specifically analyzed surviving bacteria based on retention of nucleic acid (Hoechst) signal. We observed significantly higher mCherry/GFP ratios in treated tissues (Fig 8C), which was not due to a change in mCherry (yopE) MFI (Fig 8D), but instead was specifically due to lower GFP (S10) MFI (Fig 8E). Microcolonies were significantly disrupted by gentamicin treatment based on images before (Fig 8A) and after (Fig 8F) treatment. High levels of mCherry expression were detected within surviving bacteria (Fig 8F), however the image quantification shows yopE (mCherry) levels in treated mice (+gent 4h) are similar to untreated (48h), and instead surviving bacteria specifically have low S10 (GFP) expression.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Gentamicin treatment selects for surviving bacteria with reduced S10 expression in the mouse spleen.

C57BL/6 mice were inoculated intravenously with the yopE::mCherry P_S10::gfp-ssrA or yopE::mCherry-ssrA P_S10::gfp-ssrA strains. At 48h p.i., spleens were harvested from a subset of mice to represent the time of treatment (48h). Additional sets of mice were injected intraperitoneally with 40mg/kg gentamicin, and spleens were harvested 4h post-treatment (+gent 4h) (A) Representative images at 48h p.i. (untreated). (B) CFU/spleen at 48h compared to 4h gent treatment. Dots: individual mice. (C) Reporter expression (ratio mCherry/GFP signal intensity) was quantified at the indicated timepoints. Dots: individual microcolonies. (D) mCherry mean fluorescent intensity (MFI) across individual microcolonies. Dots: individual microcolonies. (E) GFP mean fluorescent intensity (MFI) across individual microcolonies. Dots: individual microcolonies. (F) Representative image at 4h post-treatment. Scale bars: 20µm. Statistics: (B-E) Mann-Whitney; **p < .01, *p < .05, ns: not-significant.

https://doi.org/10.1371/journal.ppat.1012548.g008

Ethics statement

All animal studies were approved by the Institutional Animal Care and Use Committee of Johns Hopkins University, protocol number: MO22H322. Animal studies and procedures were performed in compliance with Animal Welfare Act regulations and Public Health Service (PHS) Policy.

Bacterial strains & growth conditions

The WT Y. pseudotuberculosis strain, IP2666, was used throughout [20,69]. For in vitro experiments with fluorescent reporters, bacteria were grown in LB broth overnight (16h) at 26^o C with rotation, sub-cultured 1:100, and then incubated at either 26^o C or 37^o C with rotation in the absence (-MgOx) or presence of magnesium oxalate (+MgOx: 20mM MgCl₂ and 20mM sodium oxalate [15,33]) to induce T3SS expression for the indicated timepoints. Equivalent volume (200µl) dH₂O was added to -MgOx tubes. Bacterial growth was assessed based on absorbance at 600nm using a Synergy H1 microplate reader (Biotek). For mouse infection experiments, bacteria were grown overnight (16h) to post-exponential phase in 2xYT broth (LB, with 2x yeast extract and tryptone) at 26^o C with rotation as previously described [20]. The P_S10::gfp-ssrA reporter was expressed from the low copy pMMB67EH plasmid; all cultures with strains containing this construct were supplemented with carbenicillin (100ug/ml). Our M9 media was prepared using the Miller JH 1972 protocol [57]: 6g/L Na₂HPO₄, 3g/L KH₂PO₄, 0.5g/L NaCl, and 1g/L NH₄Cl were dissolved in distilled H₂O and autoclaved, then supplemented with filter sterilized components at final concentrations of: 1mM MgSO₄, 0.2% glucose, 0.2% casamino acids, and 0.1mM CaCl₂.

RNA isolation and qRT-PCR

Bacterial cells were grown in LB broth for the indicated time points, pelleted, resuspended in Buffer RLT (QIAGEN) + ß-mercaptoethanol, and RNA was isolated using the RNeasy kit (QIAGEN). DNA contamination was eliminated using the on-column DNase digestion kit (RNase-free DNase set, QIAGEN). RNA was reverse transcribed using the Protoscript II First Strand cDNA synthesis kit (NEB). qRT-PCR reactions were prepared with 0.5 µM of forward and reverse primers added to cDNA samples and amplification using SYBR Green (Applied Biosystems). rpoC was used as an endogenous control since 16S transcript levels were detected as a target gene. Control samples were prepared that lacked reverse transcriptase to confirm DNA was eliminated from samples. Reactions were carried out using the StepOnePlus Real-Time PCR system (Applied Biosystems), and relative comparisons were obtained using the ∆∆C_T or 2^-∆Ct method. Kits were used according to manufacturers’ protocols.

RNA-seq

Overnight cultures (16h, stationary phase) were sub-cultured 1:100 into fresh LB and grown an additional 4h at 26^o C with rotation to generate exponential phase samples (A_600nm=0.2). Between 3 x 10⁸-1 x 10⁹ bacterial cells were collected from each culture and RNA was isolated as described above. Purified RNA was shipped to Novogene (Novogene Corporation Inc), for library preparation, sequencing, and bioinformatic analysis. Sequencing reads were aligned to IP2666 genome assembly GCA_003814345.1. The DESeq2 method of pairwise comparisons between treatment groups [70] was used to determine significant differences in transcript levels between stationary and exponential phase cells. An adjusted p-value of 0.05 was considered significant, and hits were determined based on ≥2 fold change, and transcripts with less than 100 mean reads under both conditions were excluded. Full datasets are in S1 and S2 Tables. KEGG pathway analyses were utilized to determine the biological pathways overrepresented in exponential phase compared to stationary phase.

Generation of reporter strains The following Y. pseudotuberculosis strains were previously described: the strain lacking the virulence plasmid (P(-)) [71], the yopE::mCherry strain [20, 24], the mCherry⁺ strain [45],and the ∆lcr and ∆iscR strains [51,52]. The P_S10::gfp-ssrA reporter was constructed by fusing the P_S10 promoter to gfp-ssrA by overlap extension PCR and ligating this fragment into the multiple cloning site of the low copy pMMB67EH plasmid. P_S10::gfp-ssrA strains were constructed by transformation of this plasmid into Y. pseudotuberculosis strains using a previously described protocol [72]. The constitutive GFP construct in pACYC184 was used as a template for mutagenesis to generate the gfp-ssrA construct [20], which was constructed by inserting a ssrA-tag at the 3’ end of gfp between the last coding amino acid and the stop codon [21].

Fluorescence microscopy: bacteria

Bacteria were pelleted, resuspended in 4% paraformaldehyde (PFA) and incubated overnight at 4^o C for fixation. PFA was removed and bacteria were resuspended in PBS prior to imaging. Agarose pads were prepared to immobilize bacteria for imaging by solidifying a thin layer of 25µl 1% agarose in PBS between a microscope slide and coverslip [73]. Once solidified, coverslips were removed, bacteria were added, coverslips were replaced, and bacteria were imaged with a 63x oil immersion objective, using a Zeiss Axio Observer 7 (Zeiss) inverted fluorescent microscope with XCite 120 LED boost system and an Axiocam 702 mono camera (Zeiss). Volocity image analysis software was used to quantify the fluorescent signal associated with individual bacterial cells, as recently described [73]. Briefly, individual bacterial cells were selected as objects using the green fluorescent channel, previewed to view the objects selected, then thresholding was used to ensure only individual cells were included in the analysis. WT non-fluorescent cells were grown in parallel for all experiments to detect baseline background fluorescence.

Antibiotic susceptibility experiments

Overnight cultures were diluted 1:100 into fresh LB treated with MgOx (20mM MgCl + 20mM NaOx) or the same volume of H₂O (-MgOx) and grown at 37°C to induce the T3SS. After 2h antibiotics were added (gentamicin: 10µg/mL (MBC₉₉), ciprofloxacin: 0.1µg/mL (MBC₉₀)); this represents the 0h timepoint. Samples were taken at 0h, 2h, and 4h to plate for CFUs and for staining with the Live/Dead Fixable Blue Dead Cell Stain Kit (Invitrogen). After washing in PBS, bacterial samples were stained for 30 minutes, protected from light, and washed in PBS before fixation in 4% PFA. Fixed samples were imaged as described above on agarose pads, and cells with detectable blue fluorescence (non-viable) were excluded from single cell analyses. Experiments completed in M9 media followed the same experimental outline of dilution into M9, and antibiotics were either added after 2h or 3h culture, as denoted in the Figure. For these experiments, the yopE::mCherry strain was grown in parallel as a control to determine when cells began expressing the T3SS.

Murine model of systemic infection

Six to eight-week old female C57BL/6 mice were obtained from Jackson Laboratories (Bar Harbor, ME). Mice were injected intravenously into the tail vein with 10³ bacteria for all experiments. At the indicated timepoints post-inoculation (p.i.), spleens were removed. ½ of each spleen was homogenized and plated to determine CFU/spleen. Intact tissue (remaining ½ spleen) was processed for fluorescence microscopy as described below.

Murine model of gentamicin treatment

Inoculations were performed as described above, and at 48 hours p.i., mice were either injected intraperitoneally with 40mg/kg gentamicin or sacrificed to take an untreated timepoint at the point of treatment. This dosage and route was chosen because it should result in a serum concentration between 10–20µg/ml within 20 minutes after injection [59]. 4 hours after treatment, treated mice sacrificed, and spleen were removed. As described above, ½ of each spleen was homogenized and plated to determine CFU/spleen. Intact tissue (remaining ½ spleen) was processed for fluorescence microscopy.

Fluorescence microscopy

host tissues Spleens were harvested and immediately fixed in 4% PFA by incubating overnight at 4^o C. Tissues were frozen-embedded in Tissue Plus O.C.T. compound (Fisher Scientific) and cut by cryostat microtome into 10µm sections. Two sections were analyzed for each spleen. To visualize reporters, sections were thawed in PBS at room temperature, stained with Hoechst at a 1:10,000 dilution to detect host cell nuclei, washed in PBS, and coverslips were mounted using ProLong Gold (Life Technologies). Tissue was imaged as described above, with an Apotome.2 (Zeiss) for optical sectioning.

Image analysis

Volocity image analysis software was used to quantify microcolony areas and fluorescence as previously described [3,20]. Two cross-sections were analyzed for each spleen, and every bacterial cell or clustered microcolony was imaged and analyzed within these sections. Spatial analyses were completed using Volocity software by sampling 4, ~ 4µm² regions of interest (ROI) at both the centroid and periphery of each microcolony, and calculating the ratio of mCherry/GFP signal intensity in each ROI. Centroid measurements were captured in a cluster of 4 ROIs around the geometric centroid of each microcolony. Peripheral measurements were taken approximately every 90^o around each microcolony, sampling bacterial cells in contact with host cells. Each mCherry/GFP ratio was averaged to generate an average peripheral or centroid measurement for each microcolony, and these values were plotted. Individual fluorescent channel or spatial location values were also plotted using these ROIs. Fluorescent signals were quantified the same way from agarose droplets.

Droplet generation

CAD-designed microfluidics chips with 38 independent devices controlled by two input ports were constructed according to published protocols [74,75] at the FabLab Micro and Nano fabrication laboratory at the University of Maryland, and devices were used as previously described [53]. Prior to droplet generation, each device was primed with HFE-7500 Novec oil (Oakwood Chemical) by connecting a 1ml syringe fitted with a 26-gauge needle to each input port via polyethylene (PE 20) tubing (BD). A third tube was inserted into the droplet collection port and waste was collected in a 1.5 ml microcentrifuge tube. After priming, the syringe connected to the oil-phase port was replaced with a syringe containing 1.5% FluoroSurfactant in HFE7500 Novec oil by removing the tubing from the needle of the priming syringe and placing the tubing onto the 1.5% FluoroSurfactant syringe. The syringe connected to the aqueous phase port was replaced with a 1 ml syringe containing bacteria diluted 1:2000 in 1% ultra-low-melt agarose, yielding approximately 5 x 10⁶ cells/ml. The oil-phase and aqueous-phase syringes were loaded onto separate programmable syringe pumps (Harvard Apparatus 11 Elite Series) set to 700 µl/hour and 200 µl/hour respectively. A heat lamp (VWR) was placed at a short distance (~ 0.5 m) from the syringe pump and tubing to prevent premature solidification of the agarose. Droplets were collected in a 1.5 ml microcentrifuge tube and incubated at 4° C for 30 minutes to solidify.

Oil removal

Oil was removed from the droplets as previously described [53]. To remove oil from 200 µl of droplets, aliquots of 100 µl of droplets were moved into two separate 1.5 ml microcentrifuge tubes and 500 µl of 10% 1H, 1H, 2H, 2H-Perfluoro-1-octanol (PF) (Sigma) in Novec 7500 oil was added to each. The tubes were shaken vigorously and centrifuged for 30 seconds at 250 RCF. 400 µl PBS was added and the droplets were flicked into suspension then centrifuged for 30 seconds at 250 RCF. The PBS/droplet layers from each tube were combined in a new 1.5 ml microcentrifuge tube, centrifuged for 30 seconds at 250 RCF, and washed with 1 ml PBS. The remaining droplets were resuspended in 700 µl 2xYT broth.

Magnesium oxalate droplet experiments

 ~ 400 µl of droplets were generated and separated as 200 µl aliquots resuspended in 1.5 ml microcentrifuge tubes containing 700 µl 2xYT. Both tubes were shaken gently then rotated at 37^o C. At 2 hours, MgCl₂ (20mM) and sodium oxalate (20mM) were added to one tube while an equivalent amount of ddH₂O (200µl) was added to the untreated tube. Samples were taken at 2-hour intervals by removing the tubes from the incubator, aliquoting 100 µl of each 2xYT-droplet mixture into new microcentrifuge tubes, centrifuging the tubes for 30 seconds at 250 RCF, and resuspending the droplets in 100 µl of 4% PFA.