Visualizing dynamic competence pili and DNA capture throughout the long axis of Bacillus subtilis

Prior studies of natural competence associated pili, including the type IV competence pili of Vibrio cholerae and S. pneumoniae, produced fluorescently labeled pili by using cysteine substitution variants of major structural pilins (20, 42). Since B. subtilis competence pilus biogenesis likely involves conversion of an intramolecular disulfide to an intermolecular disulfide between ComGC pilin subunits, we first assessed how the introduction of cysteine substitution variants affected transformability (22, 43). When comGC^Cys alleles were ectopically expressed as the sole copy of comGC, transformation efficiency dramatically decreased by multiple orders of magnitude compared to wild type (Fig. S1). When comGC^WT was expressed in the same strain background from the ectopic locus, however, a similar decrease in transformation efficiency was observed (Fig. S1E). This indicated that our ectopic expression construct had an intrinsic flaw that prevented proper complementation of the endogenous comGC deletion present in each strain. Importantly, this also implied that any strain ectopically expressing comGC^Cys that maintained transformation efficiency near that of the isogenic strain ectopically expressing comGC^WT was likely completely or mostly functional. Two alleles, comGC^E56C and comGC^S65C, allowed for transformation efficiency near that of comGC^WT; therefore, we decided to focus on comGC^E56C and comGC^S65C going forward (Fig. S1C and D).

To prevent any problems arising from the deletion of endogenous comGC, we opted to ectopically express comGC^E56C or comGC^S65C without altering the native comG operon, resulting in merodiploid strains expressing both comGC^WT and comGC^E56C or comGC^S65C. We surmised that any potential pilus would be able to incorporate both ComGC^WT and ComGC^Cys, allowing for pilus labeling via ComGC^E56C or ComGC^S65C subunits. These strains also carried an inducible comK allele to increase the fraction of the population that would express late competence gene products and thus facilitate microscopic analysis (44). Both comGC^E56C and comGC^S65C had limited negative effects on transformation efficiency when co-expressed with native comGC. The strains expressing comGC^E56C and comGC^S65C under control of the native comG promoter were cultured until 1 hour from maximum competence, at which point comK expression was induced. Donor DNA carrying a spectinomycin resistance gene at a separate ectopic locus was transformed after 1 hour of induction. The merodiploid strains producing the cysteine variants had transformation efficiencies within an order of magnitude of the matched wild type as well as a strain expressing only the endogenous copy of comGC (Fig. 1A). These transformation efficiencies were consistent with a functional apparatus for DNA translocation across the cell wall, which further encouraged us to continue using comGC^E56C and comGC^S65C in future experiments.

Next, we assessed ComGC pilin levels in the merodiploid strains as well as an isogenic control strain lacking an ectopic comGC copy. Previous investigations into type II secretion system pseudopili have demonstrated a direct relationship between pilin levels and pilus length (45). Therefore, we sought to verify that any pili observed would be biologically relevant, rather than artifacts caused by increased levels of pilin present in comGC^Cys mutants. Whole-cell lysates of each strain were produced after inducing competence for 1 hour as described above and prepared for immunoblots with antiserum to detect ComGC (Fig. 1B). No ComGC was detected for an uninduced ΔcomK culture, confirming ComGC-specific signal. ComGC levels for both strains carrying comGC^Cys at lacA were within tenfold of the isogenic control strain carrying only the endogenous comGC copy (Fig. S2). We note that although there was much higher variability in ComGC levels in the merodiploid strains expressing comGC^E56C or comGC^S65C compared to strains with either a single or two copies of wild-type comGC, these measurements support that any extended pilus identified is not simply due to excess ComGC subunits in the cell but is physiologically relevant.

With multiple transformable comGC^Cys alleles identified, we next wanted to determine if strains expressing either allele produced pili observable in real-time via epifluorescence microscopy. To this end, we employed the widely used maleimide labeling method that has successfully identified natural competence-associated type IV pili from both Gram-negative and Gram-positive species (20, 42). This method utilizes the highly selective reactivity of maleimide toward thiol moieties. If a pilus is produced, and the major constituent pilin subunit contains an unpaired cysteine that faces the solvent, maleimide conjugates in the solvent will covalently bond with the free thiol group on each available ComGC^Cys cysteine residue. If the maleimide is conjugated to a fluorophore, epifluorescence microscopy can be used to identify fluorescent pili and observe their dynamic production and depolymerization in real-time.

To perform maleimide labeling, we first cultured comGC^WT, comGC^E56C, and comGC^S65C to competence as described above. Samples of each culture were incubated with Alexa Fluor 488 C₅ maleimide in the dark, washed, and deposited on agar pads for microscopic analysis. Both comGC^E56C and comGC^S65C produced stubby, filamentous structures emanating from the cell periphery that resembled short pili (Fig. 2A). Critically, there was not a single instance of such structures being produced by comGC^WT treated with AF488-Mal (Fig. 2A; Fig. S3). This demonstrates that observation of these filaments depends on the presence of the ComGC^Cys monomers, lending credence to the idea that ComGC is incorporated into filaments that project into the extracellular space. Due to this finding, and the wealth of information indicating ComGC’s homology to other known pilins in terms of both form and function, we will henceforth refer to these filamentous structures as “ComGC pili.”

The pilus production capacity of both strains was assessed quantitatively (Fig. 2B). Both intact pili that remained co-localized with the cell body and regions of bright fluorescence near the periphery of the cell that we surmised were sheared pilus fragments (see Discussion) were identified for 300 individual cells of each strain. While both strains produced both classes of signal, comGC^S65C consistently produced a greater number of pili and pilus-like features compared to comGC^E56C. We decided to move forward in our analysis using comGC^S65C based on these results. Qualitatively, these ComGC pili were notably shorter than competence pili analyzed for other species across the bacterial domain (20, 42). Therefore, we assessed the distribution of pilus lengths for comparison (Fig. 2C). The pili produced by both comGC^E56C and comGC^S65C strains were indeed quite short, with mean pilus lengths of 0.32 µM and 0.33 µM and standard deviations of 0.11 µM and 0.17 µM, respectively. The implications of this size distribution will be expanded upon in the Discussion section.

One of the outstanding questions for B. subtilis DNA internalization is the mechanism by which transforming DNA is translocated across the cell wall (14). The prevailing hypothesis, that a pilus-like structure is formed which binds DNA and retracts to convey DNA to the membrane-localized DNA receptor ComEA, can now be tested directly with the aid of ComGC pilus maleimide labeling (14). First, we sought to address whether binding to DNA by ComGC pili occurs in the extracellular space during natural competence. To this end, an ~4.5 kb PCR fragment from the ycgO locus was produced and covalently labeled with Cy5 fluorophores along the length of DNA. This labeled PCR product can be co-incubated with maleimide-labeled cells to probe interactions between ComGC pili and extracellular DNA in real-time via time lapse epifluorescence microscopy (20, 42).

After silica column purification of the labeled PCR product, comGC^S65C was maleimide labeled as described previously. To ensure that we could visualize all possible pili around the periphery of the cell, we collected Z-stacks with a small step size (0.2 µM). Collecting the z-stacks also generated a time lapse of the cells with roughly 5-second intervals between images. In numerous instances, labeled PCR product was already co-localized with pili at the onset of imaging, and in other cases migrated toward pili and subsequently remained co-localized with the pili (Fig. 3; Movie S1). Even subtle movements of pili were mirrored by co-localized PCR product over time. The persistent binding of PCR product through time, the coordinated movement of co-localized PCR product and pili, as well as the real-time migration and co-localization of PCR product to pili, would be highly unlikely to occur unless a bona fide binding interaction was being formed between pili and PCR product. These data provide direct evidence of the binding of DNA to B. subtilis ComGC pili.

Once a ComGC pilus has formed and bound to extracellular DNA, a translocation step must ensue to move the bound DNA across the cell wall, where it can then presumably interact with the membrane-bound DNA receptor ComEA (24). Given that type IV pili are dynamic structures that can retract back into the membrane by the sequential disassembly of pilin monomers into the membrane, the most parsimonious mechanism of translocating DNA across the cell wall is simply depolymerizing the ComGC pilus (17). Maleimide labeling of ComGC pili allows us the opportunity to study the dynamics of pilus production and depolymerization, which allows us to verify whether DNA translocation is driven by ComGC pilus retraction.

To assess ComGC pilus retraction, comGC^S65C was maleimide labeled and deposited on agar pads as described previously. Time lapse imaging was performed using 10-second intervals for 1–2 minutes to search for pili in the process of retraction (Fig. 4A; Movie S2). ComGC pilus retraction events were documented, wherein the length of a pilus would steadily decrease until only a strong focus of signal would persist at the cell periphery. For each pilus retraction event documented, the difference in pilus length between sequential frames was determined (44 total intervals across 10 pili). Interestingly, we observed two populations of retractile pili, one being considerably faster than the other. The median retraction rate for the “slow” group was 3 nm/second, while the “fast” group was 12 nm/second. These retraction rates were both notably slower than those calculated for competence pili of either V. cholerae or S. pneumoniae, which are much more in line with other observed type IV pilus systems (20, 42). All members within the fast population did not have statistically significant differences in the average retraction rate nor did all members within the slow population (ANOVA, P = 0.36 and P = 0.69, respectively). However, a strong statistically significant difference was observed when the entire fast population and the entire slow population set of retraction rates were compared (t-test, P = 3.4 × 10⁻⁷). These data will be examined further in the Discussion section.

Our ability to identify the sites of pilus biogenesis presents the opportunity to reassess the spatiotemporal dynamics of DNA uptake during natural competence. DNA uptake has been previously shown to occur primarily at or near the cell poles, perhaps taking advantage of a natural structural weakness in the cell wall region at the interface of the lateral and polar cell wall sections (40, 41). To probe this hypothesis, we analyzed the localization of ComGC pilus biogenesis in comGC^S65C as well as the localization of stably bound fluorescent PCR product. If DNA uptake occurs predominantly at or near the cell poles, we should expect the majority of assembled pili, and the majority of bound DNA, to localize close to the cell poles during natural competence.

The analysis of ComGC pilus localization, employing approximately 230 individual pilus biogenesis events, demonstrated that pili emanate along the long axis of competent cells at the cell periphery, with relatively few events occurring at or near the cell poles (Fig. 5A). In agreement with these data, fluorescent DNA that was stably bound to pili or the cell periphery primarily localized along the long axis of competent cells (Fig. 5B). To further assess this model of DNA uptake, we compared the distribution of ComGC pili to subcellular localization of functional GFP-ComEA and GFP-ComFA. Since ComEA serves the vital role of binding incoming DNA at the cell membrane, the protein should localize at the sites of initial DNA entry, which presumably coincides with the location of pili biogenesis. In accordance with this hypothesis, GFP-ComEA localizes along the long axis of competent cells, just like ComGC pili and bound DNA (Fig. 5C). There are two possibilities for the next phase of DNA entry. The hand off from ComEA to the ComEC channel could occur near the site of pilus biogenesis, or it could occur at a secondary location. ComFA, which powers DNA translocation across the cell membrane, thus should serve as a proxy for the cytoplasmic internalization step. Indeed, GFP-ComFA, like all the other competence proteins and DNA assessed above, localizes near the cell periphery along the long axis of competent cells (Fig. 5D). It would appear, therefore, that DNA translocation across the cell wall, and entry into the cytoplasm, need not occur proximal to a cell pole.

General methods for strain construction were performed according to published protocols (62, 63). Molecular cloning was performed using the Gibson assembly method with HiFi assembly enzyme mix (NEB) (64). PCR amplification templates were derived from either chromosomal DNA isolated from the prototrophic domesticated B. subtilis strain PY79 or plasmid pKRH83. Introduction of DNA into PY79 derivatives was conducted by transformation (65). The bacterial strains, plasmids, and oligonucleotide primers used in this study are listed in Tables S1 to S3.

For general propagation, B. subtilis strains were grown at 37°C in Lennox lysogeny broth (LB) medium (10 g tryptone per liter, 5 g yeast extract per liter, 5 g NaCl per liter) or on LB plates containing 1.5% Bacto agar. Where indicated, B. subtilis strains were grown in the nutrient-limiting medium, medium for competence, with 2% fructose [MC-Fru; 61.5 mM K₂HPO₄, 38.2 mM KH₂PO₄, 2% (wt/vol; 110 mM) D-fructose, 3 mM Na₃C₆H₅O₇ · 2H₂O, 80 µM ferric ammonium citrate, 0.1% (w/v) casein hydrolysate, 11 mM L-Glutamic acid potassium salt monohydrate] substituted for 2% glucose to prevent catabolite repression of P_xyl promoter (62). When appropriate, antibiotics were included in the growth medium as follows: 100 µg mL⁻¹ spectinomycin, 5 µg mL⁻¹ chloramphenicol, 5 µg mL⁻¹ kanamycin, 10 µg mL⁻¹ tetracycline, and 1 µg mL⁻¹ erythromycin plus 25 µg mL⁻¹ lincomycin (mls). When required, 0.5% (wt/vol; ~30 mM) D-xylose was added to the cultures to induce protein expression.

A single colony of the B. subtilis strain bBB050 (Cm^R) was inoculated into LB medium and incubated at 37°C with 250 rpm shaking for 3 hours. The OD₆₀₀ of a 1:10 dilution of the culture was measured, and then 1 mL of culture was pelleted at 21,000 × g for 2 minutes and the supernatant removed completely. The pellet was resuspended to an OD₆₀₀ = 10 in Bacillus protoplasting buffer (50 mM tris pH 8.0, 50 mM NaCl, 5 mM MgCl₂, 25% (wt/vol) sucrose, 0.2 mg/mL lysozyme) and incubated in a 37°C water bath for 30 minutes to protoplast cells. The sample was removed from the water bath and left at room temperature until transformation, at which time protoplasts were pelleted at 10,000 × g for 5 minutes, the supernatant was completely removed, and then an equal volume of ddH₂O was added. The protoplasts were lysed by resuspension in the ddH₂O by repeated pipetting.

Single colonies of B. subtilis strains of interest were inoculated into MC-Fru and cultured at 37°C with 250 rpm shaking until an OD₆₀₀ = 0.2–0.5 was reached. Cells were pelleted at 10,000 × g for 2 minutes and resuspended in ~15% of residual supernatant to concentrate cells, OD₆₀₀ of a 1:20 dilution of resuspended cells was measured, and the resuspensions were diluted into fresh MC-Fru to an OD₆₀₀ = 0.05. Cultures were incubated at 37°C with 250 rpm shaking for 2 hours (1 hour prior to max natural competence induction in these conditions), and strains containing P_xyl-comK were induced with 0.5% xylose to maximize the proportion of competent cells in the populations. The strains were continued to be cultured at 37°C with 250 rpm shaking for 1 hour to allow for maximal natural competence induction, and then ~10⁵ lysed Cm^R protoplasts per µL competent cells were added to the cultures (31). Transformation was allowed to proceed under the same culturing conditions for 2 hours. Tenfold serial dilutions of each sample were made down to 10⁶-fold diluted in PBS, and appropriate dilutions were plated onto LB and LB–Cm⁵ agar plates and incubated for 16–20 hours at 37°C to allow for colony growth. The number of transformants and total cells were calculated from the single colonies on LB–Cm⁵ and LB, respectively, and the transformation efficiency was calculated as the ratio of transformants to total cells in a given sample.

B. subtilis strains of interest were cultured according to the same protocol noted for transformation efficiency assays, but 1 mL of the culture was centrifuged at 21,000 × g for 2 minutes to pellet cells after 3-hour incubation post-dilution, and supernatant was completely removed. The OD₆₀₀ of a 1:10 dilution of each culture was measured, and each cell pellet was resuspended in cell lysis buffer [25 mM Tris pH 8.0, 25 mM NaCl, 3 mM MgCl₂, 1 mM CaCl₂, 0.2 mg/mL lysozyme, 0.1 mg/mL DNase I] to an OD₆₀₀ = 10 based on the previous measurements. Resuspended cells were incubated in a 37°C water bath for 20 minutes to lyse cells and degrade genomic DNA. Cell lysates were mixed with an equal volume of 2× reducing tricine sample buffer [200 mM tris pH 6.8, 40% (vol/vol) glycerol, 2% (wt/vol) SDS, 0.04% (wt/vol) Coomassie Blue G-250, 2% (vol/vol) beta-mercaptoethanol] and heated to 37°C for 30 minutes for protein denaturation. Five microliters of each preparation were added to the wells of a tris-tricine mini gel [stacking gel: 1M tris pH 8.45, 4% (wt/vol) acrylamide/bis-acrylamide (29:1); resolving gel: 1M tris pH 8.45, 15% (wt/vol) glycerol, 10% (wt/vol) acrylamide/bis-acrylamide (29:1)] and electrophoresed [running buffer: 0.1 M tris-Cl, 0.1 M tricine, 0.1% (wt/vol) SDS] at 100 V until loading dye exited the gel (typically 1.75–2 hours). Separated proteins were Western transferred to a 0.2 µm PVDF membrane using the semi-dry transfer method at 15 V for 20 minutes with Towbin transfer buffer (66). The membrane was washed 3× in ddH₂O for 5 minutes with gentle shaking to remove transfer buffer; then, the membrane was stained with 0.1% Ponceau S solution to detect total protein for loading normalization. The membrane was blocked with 5% (wt/vol) milk in TBS-T (Tris-buffered saline with Tween 20) for 1 hour with gentle shaking, and then the membrane was incubated with 1:2,000 rabbit-derived antisera containing 1° ComGC antibody [TBS-T + 1% (wt/vol) milk] overnight at 4°C (37). The membrane was washed 3× in TBS-T for 5 minutes with gentle shaking, and then the membrane was incubated with 1:20,000 Abcam goat anti-rabbit 2° HRP antibody [TBS-T + 1% (wt/vol) milk] for 1 hour at room temperature with gentle shaking. The membrane was washed as described previously and then developed using Clarity ECL substrate (Bio-Rad) and imaged for chemiluminescence using a Bio-Rad ChemiDoc Touch imaging system.

All cover glasses used in microscopy experiments were pre-cleaned prior to use. 22 mm × 22 mm #1.5 borosilicate coverslips (DOT Scientific) were placed in a Wash-N-Dry coverslip rack (Sigma) and submerged in ~80 mL of 1 M NaOH in a 100-mL glass beaker. The beaker was then placed into an ultrasonic cleaning bath (frequency = 40 kHz, power = 120 W) and sonicated for 30 minutes to remove the thin grease layer present on the coverslips. The coverslip rack was submerged into a fresh 250-mL glass beaker filled with ddH₂O, the ddH₂O was removed, and then the coverslips were washed 3× with 250 mL of ddH₂O. The coverslips were then either air-dried overnight or dried immediately with compressed air.

For the preparation of agar pads, pre-cleaned borosilicate glass microscope slides were first rinsed free of detritus using ddH₂O and were then either air-dried overnight or dried immediately with compressed air. Working in a fume hood, half of the total rinsed slides were dipped into a 2% (vol/vol) solution of dichlorodimethylsilane (in chloroform) so that their entire surface was contacted by the solution. The solution was allowed to drip off the slides into the original container, and the remaining organic solvent on the slides was allowed to evaporate in the fume hood for 30 minutes. The slides were then thoroughly rinsed in ddH₂O and dried as mentioned previously to generate dry glass slides with extremely hydrophobic surfaces. Two pieces of lab tape (VWR #89098-074) were placed on top of one another, running lengthwise across an unrinsed microscope slide. For every agar pad to be produced, two of these taped slides were made. Approximately 15 minutes prior to imaging, conditioned MC-Fru medium (0.22 µM PES filter sterilized) from the cultures grown to competence via the transformation efficiency assay protocol was heated to 37°C, added in the ratio of 1:1 to 90°C 2% (wt/vol) molten LE agarose (SeaKem) in ddH₂O, and then vortexed to make molten 1% (wt/vol) agarose in 0.5× conditioned MC-Fru medium. Twenty microliters of this mixture were applied to the surface of a hydrophobic glass slide, and a solidified pad was produced according to a previously established protocol, using a rinsed (but untreated and hydrophilic) glass slide to form the top of the pad (67). This specific setup allows for the agar pad to stick to the bottom slide and easily slide out from the top slide, leaving an unmarred and flat surface for imaging.

For all experiments involving ComGC pilus labeling, B. subtilis strains of interest were cultured according to the same protocol noted for transformation efficiency assays. After 3-hour incubation post-dilution, 100 µL of each culture was transferred to a 37°C pre-warmed 13 mm glass test tube, and 25 µg/mL Alexa Fluor 488-maleimide was added to each culture aliquot. The aliquots were incubated in the dark at 37°C on a rolling drum for 20 minutes to allow for ComGC^Cys pilin labeling. The aliquots were transferred to centrifuge tubes, centrifuged at 5,000 × g for 30 seconds to gently pellet the cells, and all supernatant was removed. The cell pellets were washed by gently resuspending in conditioned MC-Fru medium (0.22 µM PES filter sterilized) via gentle pipetting. The cells were centrifuged again at 5,000 × g for 30 seconds, the supernatant removed, and the cells gently resuspended in one-tenth the original volume of conditioned MC-Fru medium.

An ~4.5 kb PCR product was amplified from genomic DNA of a B. subtilis strain bearing P_hysp-bdbDC at the ycgO locus (bJZ185) using LongAmp Taq DNA Polymerase (NEB) and oligonucleotide primer pair (oJZ436 + oJZ437). This PCR product was purified using the E.Z.N.A. Cycle Pure Kit (Omega Bio-Tek) following the manufacturer’s instructions. One microgram of the PCR product was fluorescently labeled using the Label IT-Cy5 Nucleic Acid Labeling Kit (Mirus) following the manufacturer’s instructions, with a sufficient quantity of Label IT-Cy5 reagent to covalently link a fluorophore to approximately 1.5% to 5% of basepairs (75–230 Cy5 molecules per DNA molecule). The Cy5-labeled PCR product was purified using the E.Z.N.A. Cycle Pure Kit as mentioned previously. The final product was electrophoresed on a 1% agarose in TAE mini gel at 100 V for 45 min, stained using SYBR Safe dye (Invitrogen) according to the manufacturer’s instructions, and quantified by densitometric analysis of the product band compared to a reference band of known quantity.

For the initial identification of ComGC pili, quantification of pilus production, pilus length measurements, and the determination of ComGC pilus retraction rates, cells were labeled with Alexa Fluor 488-maleimide as previously described. Resuspended, labeled cells of 0.5 µL volume were applied to the center of a conditioned MC-Fru agar pad, and a pre-cleaned 22 mm × 22 mm #1.5 borosilicate coverslip was applied to the drop of culture. The coverslip was gently compressed with a gloved finger to ensure cells made contact with the agar pad, and then the space between the coverslip and glass slide was sealed by applying molten (60°C) VaLAP sealant (1:1:1 petroleum jelly:lanolin:paraffin) to the coverslip edges. The cells were imaged using a Zeiss Axio Observer.Z1 inverted epifluorescence microscope equipped with a Zeiss Plan-Apochromat 100×/1.4 Oil PH3 objective lens and a Teledyne Photometrics CoolSNAP HQ² CCD camera. Cells were typically exposed to light from a Zeiss Colibri 469 nm LED module with Zeiss filter set 38 HE for 2 seconds at 100% LED power to observe ComGC pili. Cell bodies were imaged using phase-contrast microscopy (50 ms exposures). Imaging was performed at 37°C, achieved by a PECON Heater S objective heater calibrated via thermocouple. Time lapse microscopy was performed with the above conditions, with exposures occurring every 5–11 seconds depending on the experiment.

For co-localization experiments on ComGC pili and PCR products, cells were labeled with Alexa Fluor 488-maleimide as previously described. Just prior to the deposition of labeled cells onto a conditioned MC-Fru agar pad, Cy5-labeled PCR product was added to the cell suspension to a final concentration of 120 pg/µL. The new mixture of Alexa Fluor 488-labeled cells and Cy5-labeled PCR product was applied to a conditioned MC-Fru agar pad as described above. Imaging was performed using a Zeiss Axio Imager.Z2 upright epifluorescence microscope equipped with a Zeiss Plan-Apochromat 100×/1.4 Oil PH3 objective lens and a Teledyne Photometrics Prime 95B sCMOS camera. Cells were typically exposed to light from a Zeiss Colibri 469 nm LED module with Zeiss filter set 38 HE for 200 ms at 20% LED power to observe ComGC pili, while Label IT-Cy5-labeled PCR product was observed using light from a Zeiss Colibri 631 nm LED module with Zeiss filter set 90 HE for 500 ms at 20% LED power. Cell bodies were imaged using phase-contrast microscopy (100 ms exposures). Z-stacks were acquired in 0.2 µm increments starting 1 µm above the midcell plane and progressing until 1 µm below the midcell plane. Due to a fortuitous imaging delay during phase-contrast acquisition, the time interval between epifluorescent image acquisition was ~5 s, producing a time lapse.

Fluorescence microscopy was performed as previously described (68, 69). Exposure times were typically 500 ms for GFP-ComEA and 250 ms for GFP-ComFA. Membranes of gfp-comEA cells were stained with TMA-DPH [1-(4-trimethylammoniumphenyl)-6-phenyl-1,3,5-hexatriene p-toluenesulfonate] (Molecular Probes), at a final concentration of 0.01 mM, and imaged with exposure times of 200 ms. Cell bodies of gfp-comFA cells were imaged using phase-contrast microscopy (20 ms exposures). Fluorescence images were analyzed, adjusted, and cropped using Metamorph v 6.1 software (Molecular Devices).

All epifluorescence microscopy images presented in Fig. 2 through 4 were deconvoluted prior to publication. First, point-spread-functions (PSFs) were computationally estimated for Alexa Fluor 488 and Cy5 signals using the PSF Generator plugin (EPFL Biomedical Imaging Group) on Fiji (NIH) (70). For the Alexa Fluor 488 signal captured on the Zeiss Axio Observer.Z1 microscope configured as described above, the following parameters were entered into the PSF Generator: optical model = Born and Wolf 3D Optical Model; refractive index immersion = 1.518 (corresponding to Immersol 518F); accuracy computation = Best; wavelength = 516 nm (corresponding to Alexa Fluor 488 emission maximum); numerical aperture =1.4; pixelsize XY = 65 nm; Z-step = 250 nm; sizes XYZ—X = 256, Y = 256, Z = 65; display = Linear, 16-bits, Fire. The Z-stack composing the PSF was Z-projected to a single 256 × 256 pixel image by averaging the pixel intensities of each individual pixel in the 256 × 256 pixel array across the 65 Z-slices of the Z-stack, resulting in a 2D PSF. For Alexa Fluor 488 signal captured on the Zeiss Axio Imager.Z2 microscope configured as described above, the same procedure was performed to estimate the PSF, only changing Pixelsize XY to 110 nm. For Cy5 signal captured on the Zeiss Axio Imager.Z2 microscope configured as described above, all parameters were kept constant to estimate the PSF, only changing wavelength = 666 nm (corresponding to Cy5 emission maximum).

Epifluorescence microscopy images were deconvolved using the DeconvolutionLab2 plugin (EPFL Biomedical Imaging Group) for Fiji (NIH). Individual epifluorescence microscopy images were opened in Fiji, and the DeconvolutionLab2 plugin was run. The PSF generated as above, corresponding to a particular epifluorescence signal, was used to deconvolve that signal with DeconvolutionLab2 (algorithm = Richardson Lucy, 100 iterations).

Images of Alexa Fluor 488-maleimide-labeled cells from each merodiploid strain, acquired as described previously, were opened in the Fiji image processing package (ImageJ2, NIH). The plugin ObjectJ was first used to identify 300 individual cells of each strain from phase-contrast images. Narrow filaments of length greater than 0.5 µM that were directly connected to the cell bodies of these 300 cells, which were surmised to be ComGC pili, were identified and counted in the green epifluorescence channel. The extracellular space immediately adjacent to the 300 cell bodies was scanned in the green epifluorescence channel for foci, which were likely sheared ComGC pili, and foci within 0.5 µM of the cell body were counted. These epifluorescence data were graphed using Microsoft Excel.

Images of Alexa Fluor 488-maleimide-labeled cells of the comGC^E56C and comGC^S65C merodiploid strains, acquired as described previously, were opened in Fiji (NIH). The images were scaled up fivefold using bilinear interpolation, and the phase-contrast and green epifluorescence image channels were split for each pilus-producing cell. Each channel was converted into a binary image using Fiji’s default thresholding parameters, and then outlines of the signal present in each channel were produced. The outline view of the phase-contrast channel demarcated the cell boundary and extracellular space, while the green epifluorescence channel outline divided a pilus from the extracellular space. These outlines were merged together to simultaneously display both the cell boundary and pilus boundary. Pilus length was manually measured by placing the start of a segmented line on the cell boundary, approximately where the medial axis of the pilus would cross the cell boundary, and creating a line that roughly followed the pilus medial axis to the extreme tip of the pilus.

Time lapse microscopy images of Alexa Fluor 488-maleimide-labeled comGC^S65C cells, acquired as described previously, were visually scanned for potential ComGC pilus retraction events using Fiji (NIH). Once identified, cells with retracting pili were isolated, and pilus length was measured for each time point of the time lapse as described previously. Frame-to-frame retraction rates were calculated by dividing the change in pilus length between frames by the time interval of the time lapse. These data were collected for 10 individual pilus retraction events, which included 44 instances of frame-to-frame retraction. Microsoft Excel was used to perform the t-test and ANOVA (using the Real Statistics Resource Pack release 8.6.3) referenced in the Results section.

Images acquired as previously described for Alexa Fluor 488-maleimide-labeled comGC^S65C merodiploid, gfp-comEA, and gfp-comFA strains were opened in Fiji (NIH). For labeled ComGC^S65C pili, individual pilus-producing cells were isolated and identified in the phase-contrast channel, and a 4 × 4 pixel white square was added to the green epifluorescence channel where the pilus medial axis would cross the cell boundary to mark the base of pilus production. Because maleimide labeling also resulted in cell membrane staining, cell bodies were checked against the green fluorescence channel to identify cell septa not visible in phase-contrast images. If a septum was identified, a 1-pixel wide white line was added onto the phase-contrast image across the cell body to allow MicrobeJ to detect multiple cells in the image; this process was repeated for 232 individual pilus production events. MicrobeJ (version 5.11x) was used to define cell bodies and medial axes from phase-contrast images, and pili bases were identified from the green epifluorescence images by adjusting the plugin’s sensitivity parameters (tolerance and Z-score) until only the added white square was detected as a fluorescent focus (71). The identified cells and pili boundaries were associated with one another, and a heatmap of pilus production localization on a size-normalized rod-shaped cell was made within the “Heatmap(s)” tab of the results window. This heatmap was scaled up tenfold using bilinear interpolation to create the final version included in Fig. 5.

For localization of GFP-ComEA foci, cell bodies were identified by TMA-DPH epifluorescence signal, and the outline of each cell was filled in using a white rounded rectangle with a 6-pixel wide black border to enhance MicrobeJ detection of cells. GFP-ComEA foci were identified in the green epifluorescence channel and marked with a 4 × 4 pixel white square to enhance contrast. Microbe J (version 5.11x) was used as described above to generate a heatmap of GFP-ComEA localization. For localization of GFP-ComFA foci, cell bodies of isolated single cells were identified using phase-contrast images, and GFP-ComFA foci were identified in the green epifluorescence channel. A localization heatmap was generated as described previously. For localization of Cy5-labeled PCR product binding to Alexa Fluor 488-maleimide-labeled cells, time course images acquired as described previously were opened in Fiji (NIH). A binding event was defined as the co-localization of a Cy5 focus with either the cell body or a labeled ComGC filament for at least three consecutive frames (≥15 seconds of co-localization). Cell bodies were identified as described previously, and Cy5-labeled PCR product foci were identified in the red epifluorescence channel and contrast-enhanced with a 4 × 4 white square as described previously. A localization heatmap was then generated as described previously.