Label-free microscopy enables high-throughput identification of genes controlling biofilm development

Abstract

The biofilm mode of growth plays a critical role in microbial ecology and in the persistence of human pathogens. Yet, much remains unknown regarding the molecular determinants of biofilms in human pathogens. In this study, we present label-free analysis of biofilms (LFAB), an imaging approach that combines time-lapse, low-magnification brightfield microscopy with regional optical density measurements to quantify biofilm biomass. Unlike other approaches to biofilm biomass quantification, LFAB enables real-time, non-perturbative, and high-throughput monitoring of biofilms. We validated LFAB in diverse microbes and found that our measurements strongly correlate with traditional biofilm assays. We then used LFAB to identify and characterize critical factors mediating biofilm formation in Streptococcus pneumoniae, a major human pathogen whose biofilm lifecycle is known to be intimately related to colonization and infection. Initial characterization revealed that S. pneumoniae microcolonies form by radial expansion of attached cells, displaying reproducible morphology and growth dynamics. Screening of a transposon mutant library revealed that genes spanning carbohydrate metabolism, signaling, surface binding, cell wall synthesis, and adhesion impinge on the biofilm lifecycle of S. pneumoniae. We performed follow-up investigations of choline binding protein A (CbpA) and its adjacently encoded two-component system regulator, which we find are critical for the dynamics of microcolony biofilms in S. pneumoniae. Overall, this work establishes LFAB as a powerful approach for identifying and characterizing biofilm determinants across bacteria and uncovers key regulators of the biofilm lifecycle in a major human pathogen.

Introduction

To colonize environmental and host niches, microbes often form surface-attached multicellular communities called biofilms.¹ Biofilms are formed via the secretion of extracellular components that facilitate cell-to-cell and cell-to-surface adhesion.² The biofilm matrix sterically restricts interactions between the cells and threatening agents such as bacteriophages and antibiotics.³ In addition, resident cells are protected from dislocation by fluid flow, and the high cell density that emerges from cell-to-cell attachments facilitates intercellular interactions such as the exchange of metabolites, cell-cell communication molecules, and genetic material.¹ Biofilm forming bacteria are notorious for causing difficult-to-treat infections, as the biofilm lifestyle allows bacteria to evade the immune system and overcome clinical interventions.⁴ Thus, over the past several decades, the identification and characterization of molecular mechanisms and ecological principles underlying biofilm formation have become major areas of focus within microbiology. Ongoing efforts aim to manipulate the biofilm lifecycles of notorious pathogens as a novel strategy for treating recalcitrant infections.^5,6

Given the ubiquity and importance of biofilms in medical and industrial settings, there is a broad interest in assessing the characteristics of biofilms formed by diverse microbial communities, species, and strains in a non-perturbative fashion. Furthermore, in genetically tractable organisms, there is an interest to identify the genes controlling biofilm development, which often requires high-throughput genetic screens. Existing methods of measuring biofilm characteristics are generally invasive, low throughput, and/or are unable to capture temporal dynamics. Perhaps the most widely used method to assess the propensity of a bacterial culture to form biofilms is the crystal violet (CV) assay.^7,8 In this approach, bacteria are cultured, non-adherent cells are washed away, and remaining biofilm cells are stained using the CV dye. “Biofilm biomass” – a proxy for the number of biofilm cells in the community – can then be inferred by measuring the absorbance of solubilized CV using a spectrophotometer. While this approach has enabled countless discoveries and has been instrumental in the biofilm field, it imposes significant limitations: (1) the staining process kills cells, making real-time measurements impossible, (2) the numerous washing steps can dislodge biofilm structures, and (3) CV staining does not provide any information about biofilm morphology. Another common approach for measuring biofilm properties is confocal, light-sheet or super-resolution fluorescence microscopy, which, combined with advanced image analysis techniques, can yield reliable quantitative information on both biofilm dynamics and architecture.^9–14 However, fluorescence-based imaging generally requires genetic manipulation of organisms of interest to introduce fluorescent probes, and samples are often subject to photodamage and photobleaching during image acquisition. The drawbacks of these and other assays for biofilm development have contributed to a diversity of approaches that are employed in biofilm literature, which in turn has led to inconsistent empirical standards for the objects that are qualified as biofilms.

Previous studies by members of our group began to pursue label-free imaging (e.g., brightfield microscopy) as a simple, generic approach for assessing biofilm dynamics.^15,16 Label-free microscopy has the advantage of being minimally invasive, does not require genetic manipulation to introduce probes, and can potentially be used to assess the visible phenotypes of any bacterial sample, due to the optical contrast intrinsically generated by bacterial cells and biofilm matrix components. Moreover, the simple and affordable optics required for brightfield microscopy are readily available to researchers investigating microbial group behaviors across the world. In this work, we demonstrate that low-magnification brightfield imaging of bacterial culture growth in microtiter plates, together with simple image analysis steps, can be used to reliably quantify microcolony biofilm lifecycles in a sample-agnostic manner. We refer to this approach as “label-free analysis of biofilms” (LFAB). We show that LFAB measurements of biofilm biomass correlate strongly with ground truth confocal microscopy and CV staining measurements across diverse bacterial species. A user-friendly image analysis application is provided, which automates the steps required to extract quantitative data from this assay, so that it can be readily adopted by the field.

As proof of the utility of the LFAB approach, we used high-throughput imaging to discover genes implicated in the biofilm lifecycle of the human pathogen Streptococcus pneumoniae (S. pneumoniae). S. pneumoniae is a major global pathogen, responsible for ~750,000 annual deaths worldwide, as estimated in 2019.¹⁷ It forms biofilms on the host epithelium during chronic colonization of the nasopharynx^18–22 and of the middle ear²³, as well as during chronic heart disease.^24,25 These biofilms are sites for strain evolution by horizontal gene transfer^18,26 and provide S. pneumoniae protection from antimicrobials^27,28 and host immune cells.^18,29,30 Despite the importance of the biofilm lifecycle for S. pneumoniae ecology, few studies have sought to characterize its molecular determinants.³¹ To address this gap, we characterized the microcolony biofilm dynamics of S. pneumoniae and identified genes implicated in biofilm development. We found that initial cell inoculum and presence/absence of the cell-surface polysaccharide capsule strongly determined the propensity of an S. pneumoniae population to form microcolony biofilms. Moreover, using a transposon screen, we identified novel genes associated with biofilm development. Of these, we validated the role of a locus encoding the cell-surface protein CbpA^32–35, and the two-component system that regulates it (TCS06)^36–38, in biofilm formation.^31,39 In summary, in this study we present LFAB across multiple bacterial species and validate its utility by revealing novel genetic determinants of the S. pneumoniae biofilm lifecycle.

Results

Low-magnification brightfield microscopy of culture growth yields accurate microcolony biofilm quantifications

In the pathogen Vibrio cholerae, individual founding cells develop into discrete biofilm communities at the solid-liquid interface (i.e., the bottom of the well)^10–12. We refer to these as microcolony biofilms. To assess whether microcolony biofilms could be visualized by brightfield microscopy, we inoculated dilute bacterial cultures (10⁴ cells/mL) into polystyrene- or glass-bottom microtiter dishes and performed time-lapse imaging of static culture growth using a 10x or 4x magnification objective lens (see Methods). Consistent with previous reports, we observed Vibrio cholerae microcolony biofilms as areas where light was scattered to a greater degree than their surroundings, yielding darkened regions relative to the background (Fig. 1A). We confirmed the verticalization of these communities using confocal microscopy (Fig. 1B). To verify that regional darkening in brightfield images is dependent on verticalized biofilm growth, we captured time-lapses of a mutant lacking an exopolysaccharide biosynthesis gene (vpsL), known to be required for biofilm formation.⁴⁰ As expected, these cultures exhibited homogenous growth across the field of view with no regional heterogeneity in transmittance (Fig. 1A). Notably, we found that other biofilm forming bacteria, including S. pneumoniae, Klebsiella pneumoniae, and Pseudomonas species also exhibited microcolony biofilm formation, with distinct morphologies, at the microplate-liquid interface when grown under similar conditions (Fig. 1C).

Figure 1. Low-magnification brightfield microscopy captures microcolony biofilm dynamics through regional pixel darkening.

(A) Time series of brightfield images acquired with a 10x objective for Vibrio cholerae. Timepoints reflect the number of hours post-inoculation and scale is as indicated. Top panels represent biofilm formation by a strain carrying Pbad-vpvC^W240R induced with 0.1% w/v arabinose, which drives expression of a constitutively active diguanylate cyclase, in turn activating biofilm formation⁴¹. Bottom panels represent the ΔvpsL strain, which lacks a Vibrio exopolysaccharide gene known to be required for biofilm development. (B) Confocal fluorescence images of the induced Pbad-vpvC^W240R strain after 24 hours of growth, acquired with a 20x magnification objective. Cells were stained with 10μM of the lipophilic dye MM4-64 and images are colored in a magenta-hot lookup table. Top: side-on x-z-projection. Bottom: single x-y plane. Scales and axes are as indicated. (C) Brightfield images of the indicated species (strains: P. aeruginosa PAO1ΔwspF, P. fluorescens SBW25ΔwspF, K. pneumoniae KPPR1ΔwcaJ, S. pneumoniae SV36 unencapsulated), acquired with a 10x magnification objective at the microplate-liquid interface after 24 hours of culture growth. Scale bar is the same for all species in C and is shown in the bottom right panel.

Our next goal was to exploit the transmittance information present in our low magnification brightfield images to quantify microcolony biofilm biomass, and to compare these measurements to ground-truth biofilm biovolume measured by confocal fluorescence microscopy. We set up an image analysis pipeline towards this goal. As noted above, brightfield images of bacterial cultures exhibited low intensity regions where microcolonies scattered light to a greater degree than the surrounding areas, which were comprised of comparatively sparse planktonic or surface-attached cells (Fig. 2A). As a result, we found that microcolony biofilms in our images could be reliably thresholded and their transmittance properties could be measured using simple image analysis techniques (depicted schematically in Fig. 2A; for details, see Methods). First, to identify biofilm regions, a local background subtraction was applied to raw images, blurring was performed to eliminate intrinsic intensity variation inside biofilms, and a constant intensity threshold was applied to isolate biofilm regions (depicted as a binary mask on the image). Separately, to convert pixel-wise intensity values in the raw images to optical density (OD) values, a log10-based transformation was applied to the transmittance, which accounted for the maximum transmittance (a blank) and the camera noise of the optical configuration. Finally, the binary biofilm mask was applied to the optical density image, yielding a segmented biofilm image, where pixel intensities represent the degree of light scattering within biofilm structures. As a measure of the overall biofilm biomass across the field of view (accounting for biofilm surface area coverage and biofilm transmittance), the average optical density of the segmented biofilm image was computed, which we hereafter refer to as brightfield (BF)-biofilm biomass. We provide a free, user-friendly web application, (see https://github.com/BridgesLabCMU/Brightfield-biofilm-assay), so that our brightfield image analysis pipeline is accessible to all biofilm researchers.

Figure 2. Proof-of-principle for using low-magnification brightfield microscopy of culture growth as a species-agnostic and quantitative biofilm assay.

(A) Schematic of the pipeline for analyzing brightfield images of culture growth in microtiter plates. Images are from a V. cholerae strain carrying a chromosomal Pbad-vpvC^W240R construct, induced with 0.1% w/v arabinose⁴¹. Cultures were grown for 24 hours from an initial density of ~10⁴ cells/mL. Left: raw image. Middle, top: raw image with a binary mask showing biofilm regions overlayed in cyan. The mask was computed as indicated in the main text and Methods. Middle, bottom: optical density image calculated based on the indicated transformation, applied pixel-wise to the raw image. The following represent pixel intensities of respective images: “I” = raw image; “I_min” = image acquired with the camera shutter closed; “I_max” = image of fresh media (i.e., a blank). (B) Time series of biofilm biomass values, calculated from images of the indicated strains over a 20-hour lifecycle. Lines and scatter points represent means and shading represents standard deviations. (C) Correlation of biofilm biomasses, calculated by brightfield microscopy (x-axis) and by spinning-disc confocal fluorescence microscopy (y-axis), for the indicated species (strains: P. aeruginosa PAO1 WT andΔwspF, P. fluorescens SBW25 WT andΔwspF, K. pneumoniae KPPR1 WT andΔwcaJ, S. pneumoniae D39 and SV36 encapsulated and unencapsulated, V. choleraeΔvpsL, and Pbad-vpvC^W240R with arabinose concentrations 0%, 0.025%, 0.0375%, 0.05%, and 0.1%). For confocal microscopy, cells were stained with 10 μM (or 20 μM for S. pneumoniae) of the lipophilic dye MM4-64. Each point represents the mean biofilm biomass of a strain/arabinose condition. Line represents the best-fit orthogonal distance regression to the data. Error bars represent standard deviation. Inset displays the data points in the boxed region, with re-scaled axes. For brightfield microscopy, N = 3 biological with n = 3 technical replicates each. For confocal microscopy, N = 3 biological replicates. a.u.: arbitrary units. OD: optical density. BF: brightfield. CF: confocal fluorescence. Ara: arabinose. Scale bar as shown.

A major advantage of our approach is that it is non-perturbative, and measurements can be made in real-time throughout biofilm growth. To demonstrate this utility, we captured brightfield images over culture growth and measured biofilm biomass at each timepoint (Fig. 2B). Under biofilm inducing conditions in V. cholerae, we observed that microcolony biofilm biomass increased steadily post-inoculation, eventually plateauing at ~18 hours (presumably when nutrients become limited). By contrast, the exopolysaccharide deficient strain exhibited no appreciable microcolony biofilm biomass over the entire growth cycle (Fig. 2B). These results suggested that brightfield microscopy of culture growth in microtiter plates, combined with simple image analysis steps, can be used to quantify entire biofilm development cycles in real time.

Having established a pipeline to measure biofilm biomass from brightfield images, we validated our approach by comparing the brightfield quantifications to biofilm biovolume measurements from confocal fluorescence microscopy. Further, we wanted to assess the utility of this pipeline for other biofilm forming bacteria. To this end, we analyzed brightfield images for various biofilm formers under a range of biofilm-inducing conditions (V. cholerae, P. fluorescens, P. aeruginosa, K. pneumoniae, and S. pneumoniae). Concurrently, we measured the total cellular biofilm biovolume for the same strains/conditions using volumetric spinning-disc confocal fluorescence microscopy, in which biofilm cells were labeled with the lipophilic dye, MM4-64, as a ground-truth for comparison. We found a strong correlation (r² = 0.96 based on ordinary least squares) between brightfield and confocal biofilm biomass across all strains and species (Fig. 2C). Qualitatively, the correlation between brightfield and confocal biofilm biomass was independent of the objective lens used (as long as sufficiently low magnification objectives lenses were used) (Fig. S1A) or microscope (Fig. S1B), thus demonstrating that the analyses are agnostic to microscope hardware (see supplemental text and Fig. S2 for additional discussion of optical considerations for BF-biofilm biomass measurements). Additionally, we found a strong correlation between brightfield biofilm biomass measurements and absorbance due to crystal violet staining (Fig. S1C; r² = 0.84 based on ordinary least squares). Together, our results show that analyses of biofilm images captured by brightfield microscopy can recapitulate the results of confocal fluorescence microscopy or crystal violet staining and can enable tracking of biofilm lifecycles at high temporal resolution.

Characterization of S. pneumoniae microcolony biofilms using LFAB

To demonstrate the utility of LFAB, we chose to further interrogate the biofilm lifecycle of S. pneumoniae. S. pneumoniae forms biofilms during chronic colonization of the upper respiratory tract^18–22, the middle ear²³, and the heart.^24,25 Our goal was to resolve the phases of S. pneumoniae biofilm development and to identify molecular determinants controlling the lifecycle. S. pneumoniae expresses a polysaccharide capsule on the cell surface, which is a major virulence determinant. Multiple studies have reported an inhibitory role for the capsule in in vitro biofilm development.^{18,26,28,39,42,43} In vivo, the capsule is downregulated when cells adhere to host epithelial cells, such that low capsule conditions are relevant to colonization^18,44. Thus, we began by searching for an optimal seeding density in unencapsulated variants of two S. pneumoniae strains (model strain D39 and clinical strain SV36). Our aim was to identify conditions that allowed us to assess the ability of individual founder cells to grow into verticalized microcolony biofilms. We captured brightfield time-lapses and observed that when seeding these unencapsulated strains at low density (10⁴-10¹ cells/mL), distinct cell clusters were observable 5-6 hours post-inoculation and grew radially into microcolonies that plateaued at a peak biomass around 12 hours after inoculation (Figs. 3A, B and S4A, B). In contrast, when seeded at high densities (10⁵ cells/mL and above), unencapsulated strains grew into a confluent structure covering the entirety of the field of view. On the other hand, encapsulated strains did not grow into microcolony biofilms at any density, suggesting that their propensity to form verticalized biofilms from single founder cells is reduced (Fig. S3). Previous work suggests that encapsulated strains form confluent adherent biofilms at high densities, but we are unable to assess this using the LFAB imaging approach. Finally, while the microcolony phenotype was observed in two phylogenetically distant strains (D39 and SV36), this mode-of-growth may not be universal to S. pneumoniae, as the model strain TIGR4 displayed only the confluent phenotype, even when using an unencapsulated strain at low seeding density (Fig. S5). We note that the microcolony biofilms examined here are grown under conditions that are quite distinct from the convention in the field. S. pneumoniae biofilms have traditionally been examined after inoculation at comparatively high densities using a diverse set of strains, many expressing capsular genes, generally yielding confluent biofilms.^{28,42,43,45–47} These biofilms are often quantified using CV assays^31,46, or by confocal microscopy.^42,48 Thus, the S. pneumoniae microcolony biofilms analyzed here represent a mode of biofilm growth understudied in this organism. To compare our findings with LFAB to established methods, we performed confocal microscopy and CV assays at low seeding densities with the encapsulated D39 and SV36 strains and their respective unencapsulated mutants. The CV assays and confocal microscopy mirrored the phenotypes observed by LFAB, both for D39 (Fig. 3C, D) and for SV36 (Fig. S4C, D). We conclude that the S. pneumoniae microcolony biofilms captured by LFAB faithfully represent biofilm forming capacity as defined by established assays, while also unlocking new opportunities to study biofilm dynamics in real time and at scale.

Figure 3. Absence of capsule and low seeding density enable microcolony biofilm formation in S. pneumoniae (model strain D39).

(A) Brightfield microscopy images (4x magnification) of model strain D39 WT (type 2 capsule; left) and isogenic unencapsulated (D39Δcps) mutant (right) at 24 hours post seeding, grown from ~10³ cells/mL. Time series plot (bottom) shows microcolony biofilm biomass for WT (purple) and Δcps (green) strains, quantified by LFAB from 0-24 hours post seeding at 30 min intervals. Line shows mean and shaded region shows standard deviations. Bar plot (bottom right) shows the 24 hour timepoint. (B) Time series brightfield images (left) of D39Δcps at various initial cell seeding densities, showing highest seeding density at the top. Points on the boxplot (right) show the peak microcolony biofilm biomass for respective seeding densities across a 24 hour time series. Box plots show median and inter-quartile range, and whiskers show min and max values. (C) Confocal microscopy images (showing Z-stack sum projection) of D39 WT (left) and Δcps (right) strains. Biofilms were stained with 20 μM of the lipophilic dye MM4-64 and imaged at 24 hours post seeding. Bar plot shows quantified cellular biovolume of microcolonies. (D) Crystal violet (CV) assay for D39 WT and Δcps at 24 hours post seeding at 20x magnification. Biofilms were washed 3 times with phosphate buffered-saline (PBS), stained with 0.1% crystal violet, and excess stain was removed with 3 more PBS washes. Images show representative wells after staining. CV was then quantified by solubilizing in 70% ethanol and measuring absorbance at 600 nm on the spectrophotometer. For LFAB and crystal violet, N = 3 biological replicates with n = 3 technical replicates each; each point on the bar plot is a tech. repl., and each biol. repl. is shown by a unique symbol. For confocal microscopy, N = 3 biological replicates; each point on the bar plot is a biological replicate. Scale bars are as indicated. (C and D) Bars show mean and error bars show standard deviation. Student’s two-tailed t-test; *** p < 0.005. BF: brightfield. a.u.: arbitrary units.

High-throughput screening for molecular determinants of S. pneumoniae microcolony biofilms using LFAB on a transposon library

A major advantage of the LFAB methodology is that it is high-throughput, i.e., it can easily be scaled to image and analyze hundreds to thousands of samples in a day using robotic microscopy. Having established a robust protocol yielding microcolony biofilms that can be reliably quantified using LFAB, we turned towards utilizing the platform’s high-throughput capability to identify genetic determinants of microcolony biofilm formation in S. pneumoniae. A previous study reported genes associated with biofilm formation in S. pneumoniae using a CV-based screen of a transposon library³¹. Here we revisited this approach using LFAB. To this end, we prepared an indexed library of transposon (Tn) insertion mutants in the unencapsulated model strain D39Δcps. We selected this background for two reasons. First, D39 displays high transformation efficiency (>10-fold higher than most clinical strains, including SV36). Second, as a widely used model strain, there is extensive literature in this strain background, allowing for curation of our hits with published data.^49–51

To screen for biofilm phenotypes, we seeded low-density cultures (diluted to ~10³ cells/mL using a liquid-handling robot) from each of the 2366 Tn-mutant clones in our library, into 96-well microtiter plates. These cultures were grown for 24 hours, after which they were imaged, and biomass was quantified using LFAB. To set the baseline for D39Δcps (parental strain of the library) and account for inherent variability between replicates, we imaged and quantified 288 replicates of the parental strain in the same manner (representative images in Fig. 4B). We then plotted distributions of biofilm biomass for the parental strain (Fig. 4A, yellow) and the transposon clones (Fig. 4A, green) in rank order. Using low-stringency thresholds on this distribution, we shortlisted 213 mutant clones that displayed higher or lower microcolony biofilm biomass than the parental strain and performed time-lapse imaging of these shortlisted strains. Based on the time-lapses, we then updated the thresholds to get our final (more stringent) upper and lower significance thresholds (dotted lines in Fig. 4A), outside of which biofilm biomass was considered different from the parental strain. This yielded a set of 139 hits (5.9% of the total library size), 69 below and 70 above the parental biofilm biomass threshold. Using Sanger sequencing, we identified the transposon insertion locations in each of these mutants and found that they corresponded to insertions in 123 unique open reading frames. Table S1 provides a list of all these hits, and Fig. 4C displays images of the corresponding microcolony biofilms at 24 hours post-seeding. Visual inspection of the images reveals that the transposon mutants sample the phenotypic space of S. pneumoniae microcolony biofilms. They display several distinct microcolony morphologies, with differences in features such as microcolony biofilm size, shape, edge structure, spacing between microcolonies, and distribution of interstitial non-microcolony cells (Fig. 4C), which will be quantified in future work. The genes that our screen identified as modulators of S. pneumoniae biofilm formation included components with established roles in regulating S. pneumoniae biofilms (which validated our approach), as well as genes that have never been characterized (Table S1).

Figure 4. A brightfield microscopy screen identifies novel genes implicated in biofilm development in S. pneumoniae.

(A) Ranked distribution of all Tn-mutant clones screened for microcolony biofilm phenotype at 24 hours post-seeding (green dots). The y-axis displays the biomass quantified by LFAB, and the x-axis shows the rank of each Tn-mutant clone when arranged in decreasing order of biofilm biomass. Yellow points display replicates of the parental strain (D39Δcps) imaged in identical fashion to the Tn-library, forming a “null” distribution. Grey dotted lines show the thresholds (0.3651 and 0.1070), on either side of which we considered Tn-mutants to be different from parental strain in microcolony biofilm biomass. (B) Representative images (at 24 hours) of parental strain (D39Δcps), capturing variability in inoculum due to high-throughput processing by a liquid-handling robot. (C) Endpoint (24 hours) images of all 139 shortlisted clones (69 below and 70 above the parental biofilm biomass threshold) from the Tn-library. Details of clones are given in Table S1. Blue downward arrow indicates lower microcolony biomass than parental and red upward arrow indicates higher microcolony biofilm biomass than parental. Position n10 is blank and shows the scale bar for all images in panels B and C. BF: brightfield. Tn: transposon.

We identified multiple hits in carbohydrate transporters and modifiers. This class of genes is central to S. pneumoniae, which relies exclusively on carbohydrates as a carbon source, a feature reflected in the high number of sugar transporters encoded in its genome.^52,53 In addition to its critical role as a nutrient, some carbohydrates are also structural components of the capsule⁵⁴ or of the cell wall.⁵⁵ This functional category has been highlighted in previous work³¹, but our screen revealed multiple new components. For example, we found that an insertion in the galactose catabolism gene (galT) led to an enhancement in biofilm production (Fig. 4C position k5). We also found that an insertion in the cellobiose importer celB reduced biofilms (Fig. 4C position c6). Most striking was the observation that 7 clones with insertions in a mannitol importer locus displayed a dramatic drop in biofilm mass by 24 hours (Fig. 4C positions d2-d8). Overall, the number of hits in carbohydrate-related genes suggests that S. pneumoniae links metabolic information to its biofilm lifestyle.

We also observed a dramatic loss of biofilms at 24 hours for insertion within the cell wall biosynthetic enzyme murM and its operon (Fig. 4C positions e5, e6). This tRNA-dependent aminoacyl transferase plays a critical role in the assembly of branched peptidoglycan on the cell wall.^56,57 MurM is required for resistance to penicillin^57–59, highlighting its importance on the cell surface. More recently, MurM has also been shown to buffer entry into the stringent response.⁶⁰ The genes encoding murM, as well as the peptidoglycan biosynthesis enzymes murB and murE, have been previously observed in a screen for biofilm determinants.³¹ Interestingly, we also found that insertion within another cell wall biosynthetic enzyme, the transglycosylase Pbp2a, resulted in increased biofilm biomass (Fig. 4C position l5). Together, these findings suggest that genes coordinating cell wall synthesis are linked to biofilm development. Similarly, our screen and a previous screen³¹ both identified cardiolipin synthase, suggesting the composition of the cell membrane may also contribute to biofilm development. For further discussion of hits from our screen, and comparisons to previous results, see Supplemental Discussion. Taken together, applying LFAB screening to S. pneumoniae revealed a systems-level connection between fundamental S. pneumoniae processes and its biofilm lifecycle.

Characterization of cell surface protein CbpA and its transcriptional regulators as determinants of microcolony biofilm formation in S. pneumoniae

Our final goal was to use LFAB to interrogate a representative pathway identified in our screen. We chose to focus our efforts of characterizing the role of choline binding protein A (cbpA) in microcolony biofilm formation because this adhesin is known to play a critical role in S. pneumoniae adhesion to host cells, virulence,^32–35 and has previously been implicated in S. pneumoniae biofilms using traditional approaches.^31,39 Our screen identified 6 mutants with Tn-insertions inside/near cbpA, all of which displayed an almost complete loss of microcolony biofilms at 24 hours post-inoculation (Fig. 4C, positions m6-m10, n1). In addition, we repeatedly hit genes encoded adjacent to cbpA, including a two-component system (TCS06 or cbpRS) and a putative peptide, SPD_2021 (see locus schematic in Fig. 5A). To test the role of cbpA in microcolony biofilms, we generated a deletion (ΔcbpA) and a complement (ΔcbpA::cbpA^OE) in the D39Δcps (parental) background. The ΔcbpA strain was severely impaired in microcolony biofilm formation (Fig. 5C), consistent with the result for the transposon insertions. Analyses of brightfield time-lapses revealed that prior to ~8-9 h post-inoculation, the ΔcbpA and parental strain displayed similar dynamics (Fig. S6). However, the ΔcbpA strain subsequently deviated from the parental phenotype, resulting in a significantly lower peak biomass value (Fig. 5C), and altered biofilm morphology (Fig. S6). This phenotype could be complemented by overexpressing cbpA from a different genomic location. In fact, the ΔcbpA::cbpA^OE overexpressor strain displayed greater microcolony biofilm biomass than the parental strain and a decrease in interstitial non-microcolony cells (Fig. 5C). This phenotype mimicked the parental strain overexpressing cbpA (cbpA^OE) (Fig. 5C) and demonstrated that elevated expression of cbpA gives rise to enhanced microcolony biofilm formation. Together, these data show that CbpA is required for microcolony biofilm formation in S. pneumoniae.

Figure 5: Cell surface protein CbpA and its transcriptional regulation are important determinants of microcolony biofilms in S. pneumoniae.

(A) Schematic of the locus encoding cbpA, TCS06 and clpC in S. pneumoniae model strain D39Δcps. Black arrowheads show locations of hits from Tn-screen. SPD_2018 & cbpA form an operon³⁷, and rr06 & hk06 form a separate operon³⁶, both separate from ctsR & clpC.⁶¹ The operon status of SPD_2021 is unknown. (B) Schematic showing regulation of cbpA expression by the two-component system TCS06. The response regulator (RR06) in unphosphorylated form binds to the cbpA promoter region and activates its expression, which in turn promotes microcolony biofilm formation. Once phosphorylated by the histidine kinase (HK06), phosphorylated RR06 can no longer activate cbpA expression.^36–38 (C) Left: Brightfield biofilm images (4x magnification) of the parental strain (D39Δcps), ΔcbpA, an overexpressor (OE) complement ΔcbpA::cbpA^OE and a parental cbpA OE, at 24 hours post seeding. Middle: Time series of microcolony biofilm biomass for the above strains, as quantified by LFAB. Right: Peak microcolony biofilm biomass for each strain. (D) As in C, but for Δhk06 and Δhk06 ΔcbpA strains. Data of parental and ΔcbpA controls is replotted from A for comparison. (E) As in C, but for a Δrr06 strain and a cbpA OE strain in a Δrr06 background. Data of parental, ΔcbpA and cbpA^OE controls is replotted from C for comparison. For all strains, N = 3-10 biological replicates, with n = 3 technical replicates each. Scale bars are the same for all images, as shown in D. In time series plots, lines show the mean of all biological and technical replicates, and shaded areas show standard deviations. In bar plots, each data point represents the peak biomass of an individual time series. One way ANOVA, p < 0.0001 for each set. Šídák’s multiple comparisons test; **** p < 0.0001; ns: not significant. BF: brightfield. a.u.: arbitrary units.

Next, we turned our focus to the two-component system encoded adjacent to cbpA. TCS06 has been previously shown to transcriptionally regulate cbpA.^36–38 Briefly, this work demonstrated that phosphorylation of the response regulator (RR06) by its cognate histidine kinase (HK06) leads to repression of cbpA expression. By contrast, dephosphorylation of RR06 leads to cbpA activation (Fig. 5B).^36–38 Therefore, knocking out hk06 is expected to increase cbpA expression (as RR06 is dephosphorylated in this background), whereas knocking out rr06 is expected to decrease cbpA expression.³⁶ Consistent with this notion, in our screen a Tn-insertion mapping proximal to the response regulator rr06 displayed a complete abrogation of microcolony biofilms at 24 hours (Fig. 4C, position n4), whereas a clone with insertion inside the histidine kinase hk06 displayed increased microcolony biofilms and lower background cell density (Fig. 4C, position n2). To validate the role of the histidine kinase (hk06) and response regulator (rr06) in microcolony biofilm formation, we generated clean deletions within this operon in the parental D39Δcps background. The phenotypes of these mutants were all consistent with the existing model of cbpA regulation. We found that deletion of hk06, which increases cbpA expression, led to greater microcolony biofilm biomass than the parental strain (Fig. 5D). This increase in biomass was dependent on cbpA expression, as knocking out cbpA in the Δhk06 background decreased biomass to the same level as the ΔcbpA single deletion (Fig. 5D). In contrast, we found that deletion of rr06, which decreases cbpA expression, led to reduced biofilm biomass, displaying an intermediate phenotype between the parental strain and ΔcbpA (Fig. 5E). These results suggest that the absence of rr06 does not completely abrogate cbpA gene expression. Moreover, overexpression of cbpA in the Δrr06 background rescued the reduced microcolony biomass phenotype, suggesting that the decrease in microcolony biomass in the Δrr06 strain is due to a decrease in cbpA expression (Fig. 5E). Further, deletion of both the response regulator and the histidine kinase (ΔTCS06), which does not influence cbpA levels, did not significantly change the biofilm phenotype (Fig. S7A, B, C). Finally, we tested two more hits in this genomic region, a small putative peptide (SPD_2021) encoded upstream of rr06 and the clpC protease, using directed mutations. These phenocopied the wild-type strains, suggesting polar effects of Tn-insertion could be responsible for the observed defects in biofilm development (Fig. S7A, B, C). Overall, our studies in the genes surrounding cbpA revealed that both hk06 and rr06 affect microcolony biofilm formation via their control of cbpA expression levels.

In summary, we found that the absence of cbpA led to defects in microcolony biofilm formation as well as early disruption of these microcolonies, while high levels of cbpA increased microcolony biofilm size and definition. Further, changes in the levels of phosphorylated RR06, which modulates levels of cbpA, also drove differences in microcolony biofilm phenotypes. Taken together, using LFAB, we established that CbpA levels are directly associated with microcolony morphology and can be modulated by the individual components of the upstream two-component system TCS06.

Discussion

Given the limitations of existing methods, here, we asked whether label-free imaging could provide a broadly accessible means to quantify biofilm developmental dynamics. We developed LFAB, which uses low-magnification brightfield microscopy to exploit the intrinsic optical contrast of microcolony biofilms without the need for probes, stains, or genetic manipulation. In LFAB, bacterial populations are initiated at low seeding densities such that biofilm development from founder cells (or small, well-separated groups of cells) can be tracked before complete surface coverage occurs. Because confluence is not reached when microcolonies grow from low seeding densities, the observed biofilm outcomes reflect intrinsic founder-cell biofilm forming capacity rather than homogeneous cell packing in dense populations. Further, since brightfield microscopes are widely accessible to researchers across the world, LFAB can be readily implemented across laboratories and microbial systems. Notably, LFAB occupies a methodological space between crystal violet staining and confocal microscopy: it combines the scalability and affordability of bulk assays with the ability to capture morphological features throughout biofilm development. Strong correlations of LFAB measurements with both crystal violet and confocal benchmarks across diverse species underscores its robustness, and the accompanying automated analysis pipeline makes the approach accessible for broad adoption in studies of microbial community behavior.

As a demonstration of the utility of LFAB, and to further characterize biofilm formation in a notorious pathogen, we applied LFAB to S. pneumoniae. We found that, consistent with previous studies, biofilm formation is negatively correlated with capsule production in this organism.^{18,26,28,39,42,43} We then performed a high-content imaging screen which identified 123 candidate genes impacting the biofilm lifecycle of S. pneumoniae. The predicted functions for these genes include the transport and metabolism of carbohydrates, the synthesis and autolysis of the cell wall, and cell adhesion, in particular including the cell surface protein CbpA^32–35 and its regulatory cascade.^36–38 We show that while CbpA is not required in the early stages of microcolony biofilms, it plays a critical role in the maturation of these communities. CbpA is the most abundant surface protein in S. pneumoniae³² and is well characterized as an adhesin to host cells.^32–35 The mechanisms by which CbpA influences biofilms in the absence of host factors remain to be defined. We hypothesize that it facilitates bacterial cell-cell adhesion. Like many surface proteins, CbpA displays a wide range of allelic variants and future studies could test whether and how these variations influence cell-cell contacts, and consequently biofilm maturation and structure.^62,63 Overall, we present a list of candidate genes involved in S. pneumoniae microcolony biofilms, validate the role of CbpA, and demonstrate that LFAB serves as an effective tool to characterize S. pneumoniae biofilm determinants.

A previous transposon screen for S. pneumoniae biofilm determinants, conducted using the crystal violet assay by the Camilli lab³¹, offers an opportunity to compare our findings to those obtained using traditional biofilm assays. The original study screened both encapsulated and unencapsulated strains seeded at high densities in the TIGR4 strain background (for which we did not measure appreciable microcolony biofilm formation). In the encapsulated strain, aside from capsule biosynthesis genes, this study only identified lytC as potentiating biofilms³¹, which we also identified in our screen. In the unencapsulated background, we identified 11 genes common to both sets: choline binding protein A (CbpA)^32–35, its neighboring histidine kinase^36–38 and Clp protease⁶¹, another choline binding protein CbpF⁶⁴, a cardiolipin synthase, ribonuclease Y⁶⁵, an ATP-binding cassette (ABC) transporter, an aldose epimerase, the cell wall synthesis protein MurM^56–59, and two hypothetical proteins. In addition, we observed a mild biofilm reduction for a nanA insertion and the Camilli lab’s study³¹ yielded nanB⁶⁶ as a hit. Together, we take the consistency between these screens as proof of principle for the LFAB approach. Our screen also identified 112 unique hits. These include the cell wall peptidoglycan synthesis enzyme MurN⁵⁶, the mannitol transporter system, 3 surface-localized proteins (PavB⁶⁷, NanA⁶⁸ and Pbp2a⁶⁹), the surface charge modifying protein DltA⁷⁰, multiple enzymes involved in carbohydrate processing^52,53 and general metabolism (e.g., CelB, GalT, GalE, GlgA, PdxS, NrdH, etc.), several carbohydrate transporters in the ABC and phosphotransferase system (PTS) families⁵², a bacteriocin regulator (BlpC)^71,72 and an immunity protein, an Rgg/SHP cell-cell communication system Rgg1518⁷³, the endonuclease EndA involved in DNA uptake⁷⁴, the DNA repair protein RadA⁷⁵, and the cysteine synthase CysK⁷⁶, and numerous proteins with uncharacterized functions.

In summary, studies on S. pneumoniae biofilms have revealed that capsule, cell-cell communication, fratricide molecules, nutrient transporters, cell wall synthesis, and adhesins form the genetic toolkit that drives S. pneumoniae biofilms, and likely chronic colonization. Yet, a clear playbook of when and how these components are combined to drive formation, maturation, and dispersal of S. pneumoniae remain to be determined. Further, it is unclear whether there is one major pathway driving these processes, or whether multiple different pathways achieve similar outcomes. Finally, questions remain as to whether microcolony biofilms or high-density confluent biofilms most closely model the biofilms associated with S. pneumoniae carriage, chronic middle ear infections, and heart disease. We propose that LFAB provides a valuable platform to continue studies on the molecular determinants and the pathways that drive biofilms across multiple human pathogens that employ biofilms for colonization and/or infection of the host.

Methods

Bacterial growth

All strains used in this work are reported in Table S2. V. cholerae and P. fluorescens strains were propagated on lysogeny broth (LB) plates supplemented with 1.5% agar or in liquid LB with shaking at 30°C. P. aeruginosa, and K. pneumoniae strains were propagated on LB plates or in liquid LB with shaking at 37°C. S. pneumoniae strains were propagated on tryptic soy agar (TSA)-II plates supplemented with 5% sheep blood (BD, BBL, New Jersey, USA; Ref. no. 221261) for overnight growth at 37°C in 5% CO₂, or in Columbia broth (CB; Remel Inc.; Ref. no. R452972) statically at 37°C in 5% CO₂.

For LFAB and confocal fluorescence measurements V. cholerae strains were grown in M9 minimal medium containing dextrose and casamino acids (1× M9 salts, 100 μM CaCl₂, 2mM MgSO₄, 0.5% w/v dextrose, 0.5% w/v casamino acids), P. fluorescens strains were grown in M63 minimal medium containing dextrose and casamino acids (1× M63 salts, 2 mM MgSO₄, 0.5% w/v dextrose, 0.5% w/v casamino acids), P. aeruginosa strains were grown in M63 minimal medium containing arginine (1× M63 salts, 1 mM MgSO₄, 0.4% w/v L-arginine), K. pneumoniae strains were grown in M9 minimal medium containing dextrose (1× M9 salts, 100 μM CaCl₂, 1 mM MgSO₄, 0.4% w/v dextrose), and S. pneumoniae strains were grown in CB. For K. pneumoniae, the M9 medium containing dextrose (but no additional salts) was chelated for 3 hours using Chelex 100 chelating resin (BioRad catalog no. 1421253). After filter sterilization, salts were added.

Biofilm growth

For V. cholerae, P. fluorescens, and P. aeruginosa, unless otherwise indicated, strains were grown overnight in LB and subsequently back-diluted to ~10⁴ cells/mL in minimal medium. For K. pneumoniae, overnight LB cultures were washed 1x in PBS, then back-diluted to ~10⁴ cells/mL in minimal medium. For S. pneumoniae strains, cultures were grown overnight on TSA-II plates with 5% sheep blood, at 37°C in 5% CO₂. In the morning, cultures were inoculated into CB at OD₆₀₀ ~0.02-0.03, and grown to OD₆₀₀ = 0.05, i.e. early log phase (corresponding to ~10⁷ cells/mL). Cultures were then back-diluted in fresh CB to ~10³ cells/mL unless otherwise indicated. 200 μL of diluted cultures were subsequently grown statically in 96-well microtiter plates (Costar; Ref. no. 3370). For experiments pertaining to Fig. 1, V. cholerae, P. fluorescens, and P. aeruginosa biofilms were grown for 24 hours and subsequently washed 6x in PBS media prior to imaging. The procedure was identical for K. pneumoniae but without wash steps to avoid biofilm disruption. For S. pneumoniae, the images were acquired at 10 h post seeding without washing, to capture the peak stage of microcolony biofilm growth.

Microscopy

Brightfield time-lapse microscopy images were acquired at 30-min intervals at 30 °C (V. cholerae, P. fluorescens, P. aeruginosa, and K. pneumoniae) or 37 °C (S. pneumoniae) on an Agilent Biotek Cytation 1 imaging plate reader using either a 10x air objective (Olympus Plan Fluorite, NA 0.3) or a 4x air objective (Olympus Plan Fluorite, NA 0.13) driven by Biotek Gen5 (Version 3.12) software. In Fig. S1, images were additionally acquired on an EVOS M5000 imaging system (Invitrogen, Thermo Fisher Scientific Inc.), using an EVOS Plan phase-contrast 10x air objective (0.30NA/7.13WD), or using a Nikon Ti-E inverted epifluorescence HCS imager, equipped with a Photometrics Evolve EMCCD camera and a 10x air objective, as indicated. Samples were grown in 96-well polystyrene microtiter plates (Costar; Ref. no. 3370) or in Mattek glass bottom 96-well microtiter plates (MatTek Corporation; Part no. P96G-1.5-5-F). Spinning disc confocal images were acquired using a motorized Nikon Ti-2E stand outfitted with a CREST X-Light V3 spinning disk unit, a back-thinned sCMOS camera (Hamamatsu Orca Fusion BT) and a 20x air objective (Nikon Plan Apochromat, NA 0.75) driven by Nikon Elements software (Version 5.42.02). The source of illumination was an LDI-7 Laser Diode Illuminator (89-North). Cells were stained with 10 μM (or 20 μM for S. pneumoniae) of the lipophilic dye MM4-64 (AAT Bioquest) for 1 hour and were imaged with an excitation wavelength of 561 nm. V. cholerae, P. fluorescens, P. aeruginosa, and K. pneumoniae strains were grown in 18-well glass-bottom chamber slides (ibidi GmbH). S. pneumoniae strains were grown in a Mattek glass-bottom 96-well microtiter plate (MatTek Corporation; Part no. P96G-1.5-5-F).

LFAB image analysis and web application

All code for image analysis were written in the Julia programming language (v1.11.2) and are publicly available on GitHub (https://github.com/BridgesLabCMU/Brightfield-biofilm-assay). All brightfield images were subjected to the same image analysis pipeline. A web application interface for performing image analysis can be used, and the instructions are provided at the GitHub page.

To generate a mask (i.e., a spatial indicator) of biofilms in a brightfield image or timeseries of images, a corresponding “background” image is generated through the application of a large-kernel Gaussian filter (default side length = 100 pixels in x and y; for S. pneumoniae and K. pneumoniae, we used side length = 200 pixels in x and y due to large biofilm sizes). The background image is then subtracted from the raw image, normalizing the local contrast within an image and, for time-lapse imaging, normalizing the background intensity across timepoints. At this point, time-lapses are registered (aligned) using the single-step discrete Fourier transform algorithm described by Guizar-Sicairos and colleagues to correct for frame-to-frame jitter.⁷⁷ A small-kernel Gaussian filter (side length = 2 pixels in x and y) is then applied to control for intensity variation within biofilms, and a constant threshold is applied to generate a segmented mask of biofilms. In our study, the following thresholds were used for respective organisms: 0.03 (S. pneumoniae and K. pneumoniae) and 0.04 for all other species. If dust correction is selected (time-lapses only), all pixels in the mask that are 1’s at the beginning of the time-lapse and transition to 0’s at any point are taken to be dust and are set to 0 throughout. If closed-shutter and blank images (I_min and I_max, respectively) are not supplied, the mask is directly applied, pixel-wise, to the raw image, and the average value in the resulting image yields the “biofilm biomass” value. All S. pneumoniae data was analyzed in this manner, without supplying I_min and I_max images. If I_min and I_max images (or just I_min if the first image of a time-lapse is taken to be a blank image) are supplied, an optical density image is calculated as

$$I_{OD}=-lo{g}_{10}\left(\frac{I_{raw}-I_{min}}{I_{max}-I_{min}}\right).$$

The mask is then applied, pixel-wise, to the optical density image, and the average value for all pixels in the resulting image yields a “biofilm biomass” value. For display purposes (in particular, for time-lapses, where the background intensity might change dramatically due to growth of planktonic cells), local contrast and global (temporal) background intensity are normalized using a local averaging filter based on integral arrays.

Confocal fluorescence microscopy image analysis

Confocal fluorescence images were acquired as described above. For each image, we reduced the contribution of Poisson shot noise in the images using the Denoise.ai module available in Nikon Elements software (Version 5.42.02). We then performed top-hat filtering of each x-y slice to reduce the contribution of out-of-focus light along the optical axis. A z-projection was computed and then blurred using a Gaussian filter (side length = 10 pixels). A bounded automatic threshold was then applied to the z-projection using Otsu’s method, generating a 2D mask. In effect, this procedure isolated regions in the x-y plane that contained biofilms (regions of locally greater thickness than the surroundings). Then, for each slice in the original tiled z-stack (corrected for anisotropy along the optical axis), a mask of cells was computed using a bounded automatic threshold, again using Otsu’s method. The z-projection mask was then applied to the cell mask, and the sum of the pixels in the resulting binary image yielded a value for the cell area occupied by biofilms. The sum of these areas over all slices yielded a total biofilm biomass for the 3D image.

Crystal violet assay

For S. pneumoniae, biofilms were seeded at 10³ cells/mL into 12-well microtiter plates (VWR USA; catalog no. 10062-894) using CB as growth media, and grown for 24 h. At endpoint, supernatant was removed and biofilms were gently washed 3 times with phosphate buffered-saline (PBS). Biofilms were then incubated with 0.1% crystal violet (CV) solution for 10 min at room temperature. Excess stain was removed with 3 more PBS washes. CV was then quantified by solubilizing in 70% ethanol and measuring absorbance at 600 nm on the spectrophotometer. The protocol for V. cholerae was identical except that cultures were grown in M9, as described above, in 24-well plates. Readings were taken at 590 nm on the spectrophotometer.

Bacterial construct generation and transformation

For generating S. pneumoniae mutants, we used a two-step transformation approach to generate clean deletions using the fluorescent removable antibiotic cassette (FRANC) containing spectinomycin-resistance aad9, sucrose-sensitivity sacB and a fluorescent mCardinal marker.⁷⁸ Briefly, for the first transformation FRANC was inserted to replace the region desired for deletion. Colonies were screened for spectinomycin resistance and mCardinal fluorescence. In the second step, the FRANC was removed using a construct with only the two flanking regions. Colonies were selected for sucrose resistance.

In brief, around 2 kb of flanking regions up- and downstream of the gene/region of interest, and FRANC were amplified by PCR using Q5 2x Master Mix (NEB, USA). PCR fragments were verified by gel electrophoresis and subsequently cleaned up via DNA Clean & Concentrator-5 kit (Zymo Research, USA). Fragment assembly was carried out using an in-house GIBSON assembly mix. All primers used for construct generation are listed in Table S3.

Strains were inoculated in CB and grown to an OD₆₀₀ of 0.05. Then 1 ml culture was incubated with the previously assembled DNA construct and 125 ng CSP-1 peptide (EMRLSKFFRDFILQRKK, GenScript, USA). After a 2 hours incubation period, cultures were plated on respective CB agar spectinomycin plates (100 μg/mL spectinomycin) or CB agar sucrose plates (10% w/v sucrose). After overnight incubation, resistant colonies were picked, regrown in CB media with 100 μg/mL spectinomycin or 10% sucrose (w/v) and frozen stocks were generated. Mutants were verified via PCR and Sanger sequencing.

Isolation/generation of unencapsulated S. pneumoniae strains

The unencapsulated version of the SV36 strain was isolated as a spontaneous mutant in the lab. SV36, which expresses the type 3 capsule, makes visibly thick mucoidal colonies on TSA+blood plates. Spontaneous mutants that do not express the capsule make visibly distinct non-mucoidal colonies. One such non-mucoidal colony was picked, propagated and its genome was sequenced. We found a single nonsense mutation in the cpsA gene. This strain was used as the unencapsulated SV36 in this study.

The unencapsulated D39 was engineered by making a markerless gene deletion using the FRANC-based process described above. The deletion spanned multiple capsule genes (cpsA-H and cpsT) and mirrored the natural deletion found in the unencapsulated lab strain R6. This engineered strain was used as the unencapsulated D39 in this study.

Transposon library generation and screening

The indexed transposon (Tn) library was prepared using genomic DNA from a pre-existing pooled Tn-mutant library in the D39 WT (encapsulated) background, by transforming it into D39Δcps and picking single colonies off a selection plate.⁷⁹ Each colony was picked into an individual well of a 96-well plate containing Columbia Broth (CB), grown for 8 h at 37°C in 5% CO₂, and frozen stocks were generated. The library consisted of 25 such 96-well plates.

To grow the library subsequently for screening, an autoclaved PCR plate was used as a “stamp” to inoculate from each frozen stock plate into a fresh 96-well plate containing CB. These were grown for 8 h at 37°C in 5% CO₂, and a liquid handling robot (Opentrons) was used to dilute down to 10⁻⁵ for seeding biofilms into another 96-well plate. The biofilms were grown for 24 h, and imaged at endpoint on the Agilent Biotek Cytation 1 imaging plate reader using a 4x magnification objective.

Identification of transposon insertion location by arbitrary PCR and Sanger sequencing

To identify location of transposon inserts, a PCR reaction was run using Q5 2x Master Mix (NEB, USA), with a forward primer inside the Tn sequence (P1) at 0.5 μM concentration, a random reverse primer mixture containing a defined overhang (adapter) sequence (P2) at 10x concentration of P1 (5 μM), and a cell suspension of the Tn-mutant as template. The following PCR conditions were used: 98°C for 30 s + {98°C for 15 s + 30°C for 30 s + 72°C for 30 s} x 5 cycles + {98°C for 15 s + 55°C for 30 s + 72°C for 30 s} x 30 cycles + 72°C for 2 min. The product of this first PCR was used as a template to run a second PCR (conditions: 98°C for 30 s + {98°C for 15 s + 55°C for 30 s + 72°C for 30 s} x 30 cycles + 72°C for 2 min), using another forward primer (P3) inside the Tn sequence (downstream of where P1 binds), and a reverse primer that binds to the overhang adapter sequence in P2. Following the second PCR, the amplified fragment was purified by DNA Clean & Concentrator-5 kit (Zymo Research, USA). It was Sanger sequenced using primer P3, by Genewiz from Azenta Life Sciences, USA.

Upon receipt of the Sanger sequencing read, the part of the read that matched the end of the Tn was trimmed off. The trimmed read was used as a query to run nucleotide BLAST against the S. pneumoniae D39 genome on NCBI (RefSeq accession GCF_000014365.2). The region of best match (start coordinate of BLAST hit) was recorded as the insertion location of the transposon. The locus was inspected visually, the nearest gene in the RefSeq annotation was identified, and the gene IDs (new NCBI ID “SPD_RSxxxxx” and old ID “SPD_xxxx” were recorded. Corresponding R6 IDs “sprxxxx” were identified using PneumoBrowse.⁴⁹