Uncovering phenotypic inheritance from single cells with Microcolony-seq

Abstract

Uncovering phenotypic heterogeneity is fundamental to understanding processes such as development and stress responses. Due to the low mRNA abundance in single bacteria, determining biologically relevant heterogeneity remains a challenge. Using Microcolony-seq, a methodology that captures inherited heterogeneity by analyzing microcolonies originating from single bacterial cells, we uncover the ubiquitous ability of bacteria to maintain long-term inheritance of the host environment. Notably, we observe that growth to stationary phase erases the epigenetic inheritance. By leveraging this memory within each microcolony, Microcolony-seq combines bulk RNA sequencing (RNA-seq) with whole-genome sequencing and phenotypic assays to detect the distinct subpopulations and their fitness advantages. Applying this directly to infected human samples enables us to uncover a wealth of diverse inherited phenotypes. Our observations suggest that bacterial memory may be a widespread phenomenon in both Gram-negative and Gram-positive bacteria. Microcolony-seq provides potential targets for the rational design of therapies with the power to simultaneously target the coexisting subpopulations.

Keywords: EPEC; S. aureus; UPEC; UTI; adhesion factor; bacterial differentiation; bet-hedging; bloodstream infection; division of labor; epigenetic inheritance; pathogens; single-cell; single-cell heterogeneity; virulence factors.

Graphical abstract

Figure S1

Determination of appearance time of microcolonies for Microcolony-seq methodology, related to Figure 1 (A) Illustration of the ScanLag experimental setup (image adapted with permission from Levin-Reisman et al.⁸³). Bacteria are plated on agar plates, placed on office scanners, and incubated in a temperature-controlled room. The software controls the scanners and allows the monitoring of colony appearance by time-lapse imaging. Automated image analysis extracts the appearance time of each colony, as well as its doubling time. (B) A typical LB agar plate with EPEC microcolonies of the size picked for Microcolony-seq methodology (diameter of approximately 500 μm). (C) Growth of individual EPEC microcolonies of LB agar plates monitored by ScanLag. EPEC bacteria were grown in liquid for 3 h in host-mimicking conditions (DMEM; 37°C) and plated on LB agar plates at 32°C. The plot shows the earlier appearance and faster growth of the Avir^EPEC (blue) vs. the Vir^EPEC subpopulations (pink). Each growth curve represents one microcolony (n = 27 for Avir^EPEC and m = 12 for Vir^EPEC microcolonies). (D–G) Distribution of appearance time of EPEC microcolonies measured by ScanLag. Growth of EPEC WT bacteria, as described in (C), and plated on agar plates with different conditions: (D) growth on LB agar plates at 32°C (aerobic), (E) growth on LB agar plates supplemented with butyrate (50 mM) at 32°C, (F) growth on LB agar plates supplemented with infant stool metabolites at 32°C, and (G) anaerobic LB plates at 37°C. The bimodality in appearance time is robust to conditions, but the relative fractions of the Vir^EPEC and Avir^EPEC populations may vary between experiments. The average doubling time of all microcolonies (Vir^EPEC and Avir^EPEC) is indicated for each growth condition, i.e., the time needed for each microcolony to grow from 30 to 60 pixels. (H) Identification of variable genes in technical replicates allows ruling out transcripts prone to technical noise. Dispersion vs. normalized mean expression for the sequenced reads calculated on tested microcolonies, representing technical replicates. Each dot represents a gene. High dispersion outlier genes, i.e., the variable genes between these replicates, were obtained by dividing a mix of four microcolonies into four separate samples, each undergoing separate extraction, and RNA-seq analyses. The technical variable genes are highlighted in red. Identification of the genes that show inherent variability in technical replicates enabled to exclude them from the list of variable genes in the biological replicates (microcolonies) and, therefore, to better identify the genes that are truly variable due to biological heterogeneity. For example, some tRNAs were found to be technically variable and were excluded from the Microcolony-seq analysis. The technically variable genes can be found in Table S2. (I) PCA plot of Vir^EPEC (dark pink, n = 4) and Avir^EPEC (dark blue, n = 4) biological microcolonies, along with their technical replicates (light pink, n = 4, and light blue, n = 4, respectively). The PCA was generated using variable genes identified in the biological replicates of Vir^EPEC and Avir^EPEC microcolonies. All samples were normalized together using DESeq2. (J and K) ScanLag analysis of the regrowth of the individual EPEC microcolonies taken for the aerobic Microcolony-seq experiment (Figure 1). Regrowth of Vir^EPEC microcolonies (J) results in mostly Vir^EPEC colonies, and regrowth of Avir^EPEC microcolonies (K) results in mostly Avir^EPEC ones. Results are presented for three Vir^EPEC and for four Avir^EPEC microcolonies. The percentage of Vir^EPEC and Avir^EPEC colonies was determined according to the growth time.

Figure 1

Microcolony-seq identifies phenotypic heterogeneity in EPEC microcolonies (A) Experimental workflow of Microcolony-seq: bacteria are plated on solid medium, microcolonies are sampled as soon as they appear and individually resuspended, with a fraction stored at −80°C for further analyses and the remainder subjected to bulk RNA-seq. Created in BioRender. Romm, R. (2025) https://BioRender.com/u87gclv. (B–D) Microcolony-seq identifies a previously known heterogeneity without relying on colony morphology. EPEC wild-type (WT) bacteria were grown in liquid for 3 h in host-mimicking conditions (DMEM) and plated on LB agar plates, incubated at 32°C for 15 h, and subjected to the experimental pipeline shown in (A) (n = 4 Vir^EPEC; n = 4 Avir^EPEC). (B) Dispersion vs. normalized mean expression for the sequenced reads calculated on tested microcolonies, representing biological replicates. Each dot represents a gene. High dispersion outlier genes, i.e., the variable genes between the biological replicates, are highlighted in yellow. (C) Principal-component analysis (PCA) plot of the microcolonies using only the variable genes identified in (B) clearly distinguishes between the Vir^EPEC and Avir^EPEC subpopulations, which correspond to the two morphotypes. (D) Volcano plot illustrating the differential expression analysis between Vir^EPEC and Avir^EPEC subpopulations identified in (C), highlighting statistically significant upregulation of EPEC virulence machinery and downregulation of flagellar and chemotaxis genes. Genes colored in pink or blue are statistically significantly increased or decreased, respectively (|log₂fold change| > 2 and padj < 0.1). (E) Identification of pilW as a gene upregulated in Vir^EPEC microcolonies. The coverage of the gene expression reads mapped to the region on the pEAF plasmid between perABC (on the plus strand) and the pilW (on the minus strand) in Vir^EPEC is presented. Below are the values of the differential expression as in (D). (F and G) Impaired attachment of ΔpilW to HeLa cells. (F) Microscopy of HeLa cells (red:actin) infected with WT EPEC or ΔpilW strains. EPEC bacteria marked in green (anti-EPEC antibody). Although the WT strain formed clusters on HeLa cells (marked with an arrow), in the ΔpilW strain, clusters were formed on the glass (marked with white circles). Bar size is 10 μm. (G) Quantification of bacterial attachment by flow cytometry analysis of HeLa cells. An inducible plasmid for GFP expression was introduced to WT, ΔpilW and ΔpilW, pPilW EPEC strains. The bacteria were induced to quantify the number of bacteria attached to HeLa cells. For each sample, the GFP intensity for 60,000 events (corresponding to 60,000 HeLa cells) was measured. The fraction of infected HeLa cells was significantly lower in the ΔpilW mutant (p = 8.76 × 10⁻⁶) and restored by complementation of the ΔpilW with a plasmid carrying the pilW gene (ΔpilW, pPilW, p = 0.02). Statistical analysis by Student’s t test. See also Figures S1, S2, S3, and S4 and Tables S1 and S2.

Figure S2

Characterization of PilW: A key virulence gene in EPEC, related to Figure 1 (A) The amino acid sequence of PilW in EPEC. (B) PilW protein is expressed at the expected size in exponentially growing EPEC bacteria. Both the WT EPEC strain and the ΔbfpA EPEC strain, in which PilW was FLAG-tagged, were grown to exponential phase in DMEM. PilW was highly expressed in both strains, indicating that PilW is indeed expressed and that its expression is independent of BfpA. (C) Predicted secondary structures for BfpA and PilW were obtained by AlphaFold. (D) Bacterial attachment to host cells is reduced in the ΔpilW strain. Attachment was assessed by flow cytometry using HeLa cells infected for 2.5 h with EPEC strains containing a GFP plasmid (NY11113, NY12134, and NY12136): WT, ΔpilW, and ΔpilW complemented with a PilW-expressing plasmid. GFP levels from 60,000 infected HeLa cells were measured for each sample using the CytoFLEX X5 flow cytometer. The experiment was performed in triplicate, and data were analyzed and visualized using the Floreada.io platform. A significant reduction in bacterial attachment was observed in the ΔpilW strain compared with the WT (p = 8.76 × 10⁻⁶). Attachment was rescued in the ΔpilW strain complemented with a PilW-expressing plasmid (p = 0.02). The experiment was repeated three times. Statistical significance was determined using a Student’s t test.

Figure S3

Application of Microcolony-seq on EPEC microcolonies under various host-resembling growth conditions, related to Figure 1 Microcolony-seq was applied to EPEC microcolonies grown under three additional host-resembling conditions. WT EPEC bacteria were grown for 3 h in DMEM at 37°C and plated on LB agar plates supplemented with: (A) butyrate (50 mM), (B) infant stool metabolites at 32°C, or (C) under anaerobic conditions at 37°C. In each condition, n = 4 for sampled Vir^EPEC and Avir^EPEC microcolonies. Left: dispersion vs. normalized mean expression for the sequenced reads in microcolonies grown under each condition. Each dot represents a gene. Highly variable genes, identified as the variable between biological replicates, are highlighted in yellow. Right: PCA of variable genes used to cluster the microcolonies. A clear separation along PC1 was observed between Vir^EPEC (pink) and Avir^EPEC (blue) microcolonies, accounting for more than 83% of the variance across the three growth conditions. The plots for the aerobic condition appear in Figures 1B and 1C.

Figure S4

Identification of common as well as distinct virulence programs with Microcolony-seq on EPEC microcolonies under various host-resembling growth conditions, related to Figure 1 (A) A heatmap using the log₂fold change values of the differential expression analysis between the Vir^EPEC and Avir^EPEC microcolonies showing the genes that have been changed (|log₂fold change| > 1; padj < 0.1) under at least three growth conditions (aerobic, butyrate, infant stool metabolites, and anaerobic). Genes with positive (increased) and negative (decreased) log₂fold change values are presented in pink and blue colors, respectively. The group of increased genes includes the T3SS, perABC, and bfp operons, whereas the group of decreased genes includes motility and chemotaxis genes. Genes marked with an asterisk are putative virulence genes and the pilW gene identified in this study. (B) Condition-specific differentially expressed genes between Vir^EPEC vs. Avir^EPEC microcolonies. The heatmap uses log₂fold change values of significantly changed genes in the comparison between Vir^EPEC and Avir^EPEC microcolonies (|log₂fold change| > 1.5; padj < 0.1, DESeq2⁵⁴). Only genes differentially expressed in at most two conditions were included. Positive (pink) and negative (blue) values of log₂fold change are shown. Genes with low mean expression (below 10 normalized reads) or with non-significant padj (padj > 0.1) are shown in white. T2SS genes (gsp operon), sulfate metabolism genes (cys operons), and taurine metabolism (tauABC operon) are marked to show their increase, predominantly in the Vir^EPEC vs. Avir^EPEC microcolonies in the anaerobic condition.

Figure 2

Characterization of the Vir^EPEC/Avir^EPEC differentiation and its reset mechanism at stationary phase (A) The Avir^EPEC morphotype is more motile than the Vir^EPEC one. Motility comparison between Vir^EPEC (n = 4) and Avir^EPEC (n = 4) microcolonies phenotypic analyses on the same microcolonies kept at −80°C from the Microcolony-seq experiment shown in Figure 1. Bacteria from the Avir^EPEC (blue) or Vir^EPEC (pink) microcolonies were inoculated in the middle of aerobic soft agar motility plates and imaged using the ScanLag setup. The motility rate was statistically significantly higher in the Avir^EPEC microcolonies (p = 0.007, by Student’s t test), in agreement with the gene expression results (Table S2). (B) Growth advantage of the Avir^EPEC subpopulation under high salt conditions. The analysis was done on fresh microcolonies from a strain bearing the perABC-GFP reporter for the Vir^EPEC (GFP-ON) and Avir^EPEC bacteria (GFP-OFF). Microscopy images of phase-contrast, GFP fluorescence, and merging of the two channels for EPEC grown on either standard LB plates or on LB plates with high salt concentration. Both Avir^EPEC and Vir^EPEC morphotypes were able to grow on standard LB plates (images after 2.5 h of growth). However, under high salt conditions, only the Avir^EPEC morphotype (GFP-OFF) grew (images after 8 h of growth). The Vir^EPEC morphotype (GFP-ON) either switched to the Avir^EPEC morphotype (GFP-OFF) or failed to survive. The biofilm-like structures on high salt suggest the secretion of extracellular material (see Video S1). Bar size is 10 μm. (C and D) per controls the Vir^EPEC/Avir^EPEC reset at stationary phase. Western blot (C) and ScanLag (D) analysis were carried-out on the same samples of an EPEC strain with FLAG-tagged PerB grown to exponential phase (3 h growth in DMEM) or to stationary phase (25 h growth in DMEM). (C) PerB is highly expressed during the exponential growth phase and vanishes at stationary phase. Two biological replicates for each growth phase. (D) Bimodal growth quantified in microcolonies plated from exponential phase, corresponding to the Avir^EPEC and Vir^EPEC subpopulations. Reset to unimodal Avir^EPEC growth when plated from stationary phase. (E) Overexpression of PerABC an EPEC strain impairs the reset at stationary phase. The strain with the control plasmid (top) shows the reset at stationary phase as in (D) (bottom), whereas the PerABC overexpression strain (bottom) maintains bimodality at stationary phase without resetting to Avir^EPEC. See also Figure S5.

Figure S5

GFP compatibility with high salt growth of EPEC bacteria and motility differences in anaerobic Vir^EPEC subpopulations, related to Figures 2, 3, and 4, Video S2, and STAR Methods (A) GFP fluorescence is compatible with growth on high salt. Control showing that GFP fluorescence is compatible with biofilm-like growth on high salt. Constitutively GFP expressing strain EPEC pZS^∗11-GFP was grown on LB agar pads with 0.7 M NaCl at 37°C. Microscopy images of phase contrast, GFP fluorescence, and merging of both confirm the ability of bacteria to produce GFP and biofilm-like growth under high salt conditions. Micro-Manager software was utilized for time-lapse imaging. Imaged 8 h after plating. Scale bar: 10 μM. (B) Weak switching of the Vir^EPECfim OFF state at stationary phase: frozen samples of the same microcolonies analyzed in Figure 3E were grown for 40 h in DMEM at 37°C without shaking, diluted 1:100, and grown for 2 h in DMEM at 37°C without shaking. A comparable number of bacteria of the frozen and after re-growth was used for the PCR. Colonies that were initially only OFF (Figure 3E) now contained a small fraction of bacteria in the ON direction. (C) The difference in motility between the Hyper-flagellated (n = 6) and Non-flagellated (n = 6) subpopulations in the anaerobic Vir^EPEC microcolonies was functional and maintained during the exponential phase (p = 0.002 by Wilcoxon rank sum test) as well as in the stationary phase (p = 0.002 by Wilcoxon rank sum test). Bacteria from the original microcolonies from the anaerobic condition (Figure 3A) were grown in DMEM in standing conditions at 37°C for 1.5 h (exponential) or 60 h (stationary) and plated on soft agar motility plates. Motility soft agar tests were carried out under aerobic conditions at 37°C. Plates were imaged every 30 min using the ScanLag setup. The motility area in each time point for each growth phase was quantified using ImageJ software. The data demonstrated a statistically significant difference in motility between Hyper-flagellated and Non-flagellated subpopulations in both growth phases. (D) A schematic representation (left) of the agar-patterned microwells prepared to observe the motility of the Hyper-flagellated and Non-flagellated EPEC microcolonies at the single-cell level (Video S2). Agar pools were prepared by pouring LB agar on top of silicon wafer patterned with microwells in which bacteria can freely swim. The pattern was done using photolithography with SU-8 photoresist (MicroChemCorp, MA), resulting in a pattern of microwells (100 μm depth and 450 μm diameter). A volume of ∼2 μL of bacteria was placed on a slide and covered by the agar pools. (Right) A representative image of a single pool. Scale bar: 450 μm.

Figure 3

Phenotypic heterogeneity in fim expression within the anaerobic Vir^EPEC microcolonies (A) EPEC WT bacteria were grown for 3 h in DMEM, plated on anaerobic LB plate, and incubated at 37°C for 12 h. Microcolony-seq was performed on microcolonies of the Vir^EPEC morphotype (n = 12) with uniform size and morphology. (B) PCA of anaerobic Vir^EPEC microcolonies using only the variable genes. Separation between microcolonies with high fim expression (red, n = 5) and low fim (black, n = 7) are identified along PC1. (C) Volcano plot of the differential expression analysis between the two subpopulations based on PC1 separation. Colored gene names present the statistically significant upregulated genes (log₂fold change > 1; padj < 0.1), revealing higher expression of all fim genes in the red subpopulation compared with the black one. (D) Schematic representation of the fim operon, highlighting the region undergoing reversible phase variation at the fimA promoter. Created in BioRender. Romm, R. (2025) https://BioRender.com/w7a4fjg. (E) fim operon state maintenance in anaerobic EPEC microcolonies: PCR results with two sets of primers for the ON and OFF states demonstrated that microcolonies were predominantly either in the fim ON or OFF state (two biological replicates of each are shown). Analyses were performed on the same microcolonies analyzed by Microcolony-seq and kept at −80°C. (F) Partial reset of the fim ON state at stationary phase: frozen samples of the same microcolonies analyzed in (E) were grown overnight. Colonies that were initially fim ON (E) now also contained bacteria in the fim OFF direction. See also Figure S5 and Table S3.

Figure 4

Genetic heterogeneity in flagellar repressor revealed in anaerobic Vir^EPEC microcolonies (A) PCA plot of anaerobic Vir^EPEC microcolonies using only the variable genes, presenting the PC2 and PC3 axes, color coded by PC2 values. The Vir^EPEC Hyper-flagellated (n = 6) and Vir^EPEC Non-flagellated (n = 6) subpopulations are depicted in green and yellow, respectively. (B) Volcano plot of the differential expression analysis between the two subpopulations based on PC2 separation. Colored gene names present the statistically significant upregulated genes (|log₂fold change| > 1; padj < 0.1), revealing major motility gene upregulation in the green subpopulation compared with the yellow one. (C) Transcriptional differences in flagellar genes translate to pronounced differences in flagella expression between the Vir^EPEC Hyper-flagellated and Vir^EPEC Non-flagellated subpopulations. Bacteria from both subpopulations were used for infection of HeLa cells. Flagella were stained with an anti H6 antibody (green). Notably, the Vir^EPEC non-flagellated bacteria had a very low expression of flagella. Bar size is 10 μm. (D) Motility differences between the Vir^EPEC Hyper-flagellated and Vir^EPEC Non-flagellated subpopulations. Aerobic soft agar motility plates were imaged with ScanLag and quantified using ImageJ. Analyses were performed on the same microcolonies analyzed by Microcolony-seq and kept at −80°C. The motility rate was statistically significantly higher in the Vir^EPEC Hyper-flagellated (n = 3) vs. Vir^EPEC Non-flagellated (n = 3) microcolonies (p = 1.8 × 10⁻⁶, by Student’s t test), in agreement with the gene expression results (Table S3). (E) The Hyper-flagellated variants arise from high-frequency mutations in a flagellar repressor. WGS results uncover mutations in the coding sequence of lrhA, a repressor of FlhDC. (F) Schematic representation of all coexisting subpopulations detected by Microcolony-seq in EPEC microcolonies. Three types of heterogeneity (Vir^EPEC/Avir^EPEC, fim ON/OFF, and Vir^EPEC Hyper-flagellated/Non-flagellated) generated five bacterial phenotypes. Stability of inheritance was mapped. Created in BioRender. Romm, R. (2025) https://BioRender.com/tmtvfy0. See also Figure S5, Table S3, and Video S2.

Figure 5

Unraveling bacterial heterogeneity in a clinical UTI using Microcolony-seq (A) Urine obtained from a UTI patient infected with UPEC was spread on LB agar and growth was monitored by ScanLag. Microcolonies (n = 20 for biological and n = 4 for technical) were picked as soon as they appeared, resuspended, and analyzed by Microcolony-seq. Created in BioRender. Romm, R. (2025) https://BioRender.com/ah5epdp. (B) PCA utilizing the variable genes (Figure S6C) reveals two UTI subpopulations: those that lost the antibiotic resistance cluster encoding for resistance genes to streptomycin, sulfonamide, and tetracycline (n = 4, purple) and those expressing it (n = 16, brown), Table S4. WGS performed on the same microcolonies (Table S5) attributes these differences to genetic changes. (C) E tests with trimethoprim/sulfamethoxazole and streptomycin antibiotics performed on the same UTI microcolonies analyzed by Microcolony-seq and kept at −80°C provided confirmation of the heterogeneous resistance levels predicted from the genetic analyses for the Antibiotic resistance⁺ and Antibiotic resistance⁻ subpopulations. (D) PCA utilizing the variable genes (Figure S6C), including only the Antibiotic resistance⁺ subpopulation. PC1 differentiated between two additional subpopulations: Vir^UTI (n = 7, orange) and Avir^UTI (n = 8, turquoise). One microcolony indicated in gray could not be attributed to any subpopulation. The Vir^UTI/Avir^UTI separation was shown to be due to phenotypic changes (Table S5). (E) Differential expression analysis between Vir^UTI and Avir^UTI subpopulations. The heatmaps display genes with |log₂fold change| > 1 and padj < 0.1, excluding uncharacterized genes. Upregulated genes (orange) include virulence factors like iron acquisition, acid response, T2SS, toxins, and colicins. Downregulated genes (turquoise) include sulfate metabolism and chemotaxis processes. The upregulation of acid response genes suggests that Vir^UTI microcolonies maintained memory of the host’s acidic urine (pH 5.0). See also Figure S6 and Tables S4 and S5.

Figure S6

Additional information on the human urine sample (UTI) analyzed by Microcolony-seq, related to Figures 5 and 6 (A) Urine of a patient with a UTI of UPEC bacteria was plated on LB agar plate and incubated at 37°C for 8 h. Appearance time of UTI microcolonies was monitored by ScanLag. Unimodal appearance time distribution was observed. (B) An agar plate with UTI microcolonies displaying the typical size of microcolonies used for Microcolony-seq. (C) Identification of variable genes in UTI microcolonies (n = 20). Variable genes were identified based on gene dispersion (according to Equation 1) and normalized mean expression. Each dot on the graph represents a gene, with variable genes highlighted in yellow. (D) Identification of variable genes in technical replicates allows ruling out transcripts prone to technical noise. Dispersion (according to Equation 1) vs. normalized mean expression for the sequenced reads calculated on technical replicates. These technical replicates were obtained by picking and resuspending four UTI microcolonies together in a single Eppendorf tube. The mixture was then divided into four Eppendorf tubes, each of which was subjected to RNA-seq extraction and analysis separately. Each dot represents a gene. Technical variable genes are highlighted in red. (E) A PCA plot of UTI microcolonies, including technical replicates (indicated in black). The PCA was performed using the variable genes identified across biological replicates. All microcolony counts were normalized together using DESeq2. Colors (orange, turquoise, and purple) correspond to subpopulations identified in the PCA plots (Figures 5B and 5D). Orange denotes Vir^UTI, turquoise represents Avir^UTI, and purple signifies antibiotic resistance cluster loss. (F) A phylogenetic tree, constructed from WGS data of UTI microcolonies analyzed by Microcolony-seq, illustrates genetic relationships. Colors as in (E). All microcolonies belong to the same clone. The shortest edge represents one genetic difference. For example, UTI_9 and UTI_20 differ by only one single nucleotide polymorphism (SNP). All detected genetic changes were included in Table S5. Note that the phenotypic clusters (orange and turquoise) are not due to genetic clustering. (G) Antibiotic susceptibility testing obtained from the clinic on the same UTI sample used for the Microcolony-seq experiment. The infecting UPEC bacteria were classified as resistant to ampicillin and to trimethoprim/sulfamethoxazole. Our results showed that a clinical test typically performed on a single colony may not reflect the resistance pattern of the sample (Figure 5C). Only part of the bacteria within the sample were actually resistant to trimethoprim/sulfamethoxazole. S, sensitive; R, resistant. (H) A heatmap based on log₂fold change values of the DESeq2 analysis highlights differential expression (log₂fold change > 1; padj < 0.1) between Vir^UTI and Avir^UTI microcolonies (Figure 5D). The upregulated genes were marked in orange. We found a significant overlap with a gene set identified in a study comparing bulk RNA-seq data from urine samples of UTI patients to bacteria grown in LB medium. Our analysis indicated that the Vir^UTI microcolonies, despite being grown on LB, kept a memory of the host environment. Only genes with matching names in both datasets were included in this comparison.

Figure 6

Probing the stability of inheritance of the different subpopulations in the UTI sample All analyses were performed on the same microcolonies analyzed by Microcolony-seq (Figure 5) and kept at −80°C. (A–C) Phenotypic consequences of iron pathway expression heterogeneity in Vir^UTI (n = 4, high iron expression, orange) and Avir^UTI (n = 4, low iron expression, turquoise) microcolonies. Microcolonies were plated on standard LB and on IL LB agar. All microcolonies grew similarly on standard LB (A), but only Vir^UTI grew under iron limitation (B), revealing the hypersensitivity of Avir^UTI to IL. Created in BioRender. Romm, R. (2025) https://BioRender.com/b43y573. (C) The distinction between Vir^UTI and Avir^UTI phenotypes was lost after growth to stationary phase on standard LB, as both subpopulations could grow on IL conditions, indicating a phenotypic switch. (D) Hypersensitivity to iron limitation was observed in the original urine but not after the bacteria were grown to stationary phase, as quantified by the ratio of CFUs that grew under IL vs. CFUs on LB. The experiment was repeated twice. Mean values of three biological replicates were included. The standard deviation is indicated. (E) Schematic representation of all coexisting subpopulations detected by Microcolony-seq in the UTI clinical sample, showing epigenetic heterogeneity in virulence. This diversity provided a fitness advantage under IL conditions and reset at stationary phase. Genetic and epigenetic heterogeneity resulted in eight different subpopulations. Created in BioRender. Romm, R. (2025) https://BioRender.com/b32o015. See also Figure S6.

Figure 7

Microcolony-seq reveals antigen expression heterogeneity in a low bacterial load bloodstream S. aureus infection (A) Microcolony-seq was performed on blood from a patient with S. aureus BSI. 800 μL of blood was plated on tryptic soy agar (TSA), resulting in microcolonies (n = 23). (B) Hierarchical clustering of the variable genes in the RNA-seq data identified three subpopulations in the BSI: Vir^BSI (n = 9), Intermediate^BSI (n = 11), and Avir^BSI (n = 3). (C) Heatmap of significantly changed genes (|log₂fold change| > 1; padj < 0.1) between Vir^BSI vs. remainder. Genes upregulated in Vir^BSI (pink) include capsule expression, arginine, and histidine pathways (related to acidic environment) and clfA adhesion factor. Intermediate^BSI and Avir^BSI (blue) showed increased spa and mnaT expression, with SpA enabling immune evasion. (D) The sensitivity of Intermediate^BSI and Avir^BSI to acid conditions is lost after stationary phase. Growth rate measured in a plate reader in trypticase soy broth (TSB) at acidic pH (5.5). Cultures were inoculated either directly from the original BSI microcolonies or after growth to stationary phase in standard TSB. Acid sensitivity, indicated by a lower growth rate, was observed in the Intermediate^BSI (n = 11, p = 0.005) and Avir^BSI (n = 3, p = 0.002) subpopulations before stationary-phase reset (black) compared with Vir^BSI (n = 9). However, no significant differences were observed after the stationary phase reset (gray) (p = 0.4 and 0.3 when comparing Intermediate^BSI and Avir^BSI, respectively, to Vir^BSI). Statistical analysis by Student’s t test. (E) Three distinct phenotypic states were detected in the BSI, with acid sensitivity heterogeneity reset at stationary phase. (A) and (E) were created in BioRender. Romm, R. (2025) https://BioRender.com/z60b158. See also Table S6.