Natural Transformation Protein ComFA Exhibits Single-Stranded DNA Translocase Activity

ABSTRACT

Natural transformation is one of the major mechanisms of horizontal gene transfer in bacterial populations and has been demonstrated in numerous species of bacteria. Despite the prevalence of natural transformation, much of the molecular mechanism remains unexplored. One major outstanding question is how the cell powers DNA import, which is rapid and highly processive. ComFA is one of a few proteins required for natural transformation in Gram-positive bacteria. Its structural resemblance to the DEAD box helicase family has led to a long-held hypothesis that ComFA acts as a motor to help drive DNA import into the cytosol. Here, we explored the helicase and translocase activity of ComFA to address this hypothesis. We followed the DNA-dependent ATPase activity of ComFA and, combined with mathematical modeling, demonstrated that ComFA likely translocates on single-stranded DNA from 5′ to 3′. However, this translocase activity does not lead to DNA unwinding under the conditions we tested. Further, we analyzed the ATPase cycle of ComFA and found that ATP hydrolysis stimulates the release of DNA, providing a potential mechanism for translocation. These findings help define the molecular contribution of ComFA to natural transformation and support the conclusion that ComFA plays a key role in powering DNA uptake.

IMPORTANCE Competence, or the ability of bacteria to take up and incorporate foreign DNA in a process called natural transformation, is common in the bacterial kingdom. Research in several bacterial species suggests that long, contiguous stretches of DNA are imported into cells in a processive manner, but how bacteria power transformation remains unclear. Our finding that ComFA, a DEAD box helicase required for competence in Gram-positive bacteria, translocates on single-stranded DNA from 5′ to 3′, supports the long-held hypothesis that ComFA may be the motor powering DNA transport during natural transformation. Moreover, ComFA may be a previously unidentified type of DEAD box helicase—one with the capability of extended translocation on single-stranded DNA.

INTRODUCTION

Natural transformation, or the uptake and integration of extracellular DNA by competent bacteria, is one of the main mechanisms for sharing genetic information among bacterial populations (1, 2). Natural transformation is unique among the methods of horizontal gene transfer because the process is entirely controlled by the DNA recipient—the DNA donor organism need not be present (3). Competence is relatively common in the bacterial kingdom, having been demonstrated in at least 80 species, and the competence proteins and basic mechanism of transformation have been defined in both Gram-positive and Gram-negative bacteria (3–5). However, key aspects of the molecular mechanism of DNA uptake still remain a mystery.

One such mystery is the source of power for DNA transport across the cell membrane. The speed and processivity of DNA uptake strongly suggest that the cell uses energy to actively transport DNA. Indeed, magnetic tweezer experiments that tracked the movement of Bacillus subtilis along a DNA substrate demonstrated that uncoupling agents could halt the DNA uptake process, confirming that the process is dependent in some way upon the consumption of ATP (6). One hypothesis is that a DNA uptake motor helps power the translocation of DNA. However, a motor coupling ATP hydrolysis to DNA uptake has yet to be identified.

The process of natural transformation begins with the binding of double-stranded DNA (dsDNA) outside the cell (7, 8). Gram-positive and Gram-negative bacteria differ in the early steps to bring dsDNA across the cell wall and outer membrane (if present), but then in all naturally competent bacteria, ComEA guides the dsDNA toward a pore in the cell membrane (7, 8). At this point, a single strand of DNA is taken up into the cell, while the other is degraded (9–11). ComEC is presumed to be the channel protein that allows for the passage of DNA through the cell membrane, based on its large size, its conservation in every competent species identified to date, and the fact that it is indispensable for transformation (3, 12, 13). In Gram-positive bacteria, membrane-associated ComFA is also necessary for the transport of DNA across the cell membrane (9–11, 14, 15). The precise role of ComFA has not been identified, but researchers have put forth the hypothesis that the protein could act as a DNA uptake motor in Gram-positive bacteria (1, 16–19).

The idea that ComFA may be the DNA uptake motor, in part, stems from its importance in natural transformation. Transformation efficiency is decreased by as much as 10,000-fold in B. subtilis strains lacking ComFA (16–18). Similarly, competence in Streptococcus pneumoniae strains lacking ComFA is dramatically reduced or eliminated (20, 21). In addition to its critical role in transformation, however, the structural similarity of ComFA to the DEAD box family of helicases, which belongs to helicase superfamily 2, led to the hypothesis that ComFA may push or pull DNA into the cell. ComFA contains a DEAD box helicase-like domain, including conserved Walker A and Walker B motifs (17, 18). Mutation of either of these domains in B. subtilis ComFA significantly reduces transformation efficiency, while having no effect on DNA binding to the cell surface (17, 18). Our previous findings that S. pneumoniae ComFA binds single-stranded DNA (ssDNA) and that the protein exhibits ssDNA-dependent ATPase activity bolster the hypothesis that ComFA is the DNA uptake motor (19). However, DEAD box helicases are known principally as RNA remodelers that bind and melt nucleic acids locally, rather than unwinding DNA via translocation (22–24). This raises the question, by what mechanism could ComFA couple ATP hydrolysis with DNA uptake?

Here, we present biochemical data supporting the model that ComFA acts as a DNA translocase in Gram-positive bacteria, making it unique among DEAD box helicases. We use the ComFA protein from S. pneumoniae to demonstrate that ComFA does not bind dsDNA and that it does not require a free 5′ or 3′ end to bind ssDNA. ATPase assays and mathematical modeling reveal that ComFA likely translocates on DNA in a 5′ to 3′ direction. We used a strand displacement assay to determine that the translocase activity of ComFA does not give it the ability to unwind DNA under the conditions we tested. Further, we analyzed the coupling of the ATPase cycle to ssDNA interaction. Based on these data, we propose a mechanism for DNA translocation of ComFA based on ATP and ADP binding dynamics.

RESULTS

ComFA preferentially binds ssDNA.

As we reported previously, recombinant full-length ComFA from S. pneumoniae was purified, and we showed that the purified protein exhibits ATPase activity in the presence of DNA (19; Fig. 1A). A Walker B mutant (D205A), however, showed no DNA-stimulated ATPase activity, demonstrating that there were no contaminating ATPases of significance in our purified protein samples (Fig. 1A). We showed previously that ComFA ATPase activity is not stimulated by dsDNA (19; Fig. 1C). However, to further elucidate the role of ComFA in DNA uptake, we sought to determine whether ComFA is capable of binding dsDNA. We tested the dsDNA-binding ability of ComFA using an Electrophoretic Mobility Shift Assay (EMSA) and found that ComFA associates with dsDNA only very weakly, with a K_d of >1 μM, as opposed to 138 ± 86 nM for ssDNA which we reported previously (Fig. 1B; 19). To further test the preference for ssDNA, we performed a competition experiment, examining the previously demonstrated ssDNA-stimulated ATPase activity of ComFA (19) in the presence of increasing concentrations of dsDNA. We measured the ATPase activity of ComFA while keeping the concentration of ssDNA constant at 500 nM and increasing the concentration of dsDNA to see whether high concentrations of dsDNA could compete away the interaction of ComFA with ssDNA. The presence of dsDNA failed to decrease the ATPase activity of ComFA even at a concentration 20 times that of the ssDNA, confirming that ComFA exhibits a strong preference for ssDNA (Fig. 1C).

FIG 1.

ComFA binds weakly to dsDNA and binds internally on ssDNA. (A) SDS-PAGE showing purified ComFA from S. pneumoniae (red triangle indicates band containing ComFA) and ATPase activity of wild type ComFA with and without ssDNA (10 μM 50T), as well as the ATPase activity of a Walker B mutant (D205A) in the presence of 10 μM 50T. ATPase activity was evaluated using an NADH-coupled 96-well plate assay with and without nucleic acid substrate as indicated (B) A representative Electrophoretic Mobility Shift Assay (EMSA; see Materials and Methods “EMSA B”) using 5 nM dsDNA with increasing concentrations of ComFA (dsDNA gel on the left: 0, 10, 20, 40, 60, 80, 100, 200, 400, 600, 800, 1200 nM; ssDNA gel on the right: 0, 20, 40, 60, 80, 100, 200, 400, 800, 1200 nM); (red triangles indicate 100 nM). dsDNA = 50MB/50MB-rev, ssDNA = 50MB (see Table 2 for sequences). Contrast and brightness were adjusted slightly for presentation. (C-D) Mean ATPase activity of WT ComFA with indicated substrates. (B) n = 2 all conditions. ssDNA = 50T, dsDNA = 50MB/50MB-rev. (D) [DNA] = 250μg/mL. Due to high apparent background ATPase activity of two replicates, these data are shown relative to the ‘no DNA’ control. Error bars represent standard deviation. N = 4. (A, C) A one-way ANOVA with Sidak’s multiple-comparison test was used and for (D) an unpaired t test was used to determine significance, denoted by **P ≤ 0.01.

ComFA binding does not require a free end of DNA.

Many ssDNA binding proteins bind or load exclusively on either the 5′ end or the 3′ end of their respective DNA substrates, and this binding preference can be critical for the role they play in the cell (25). Determining whether or not ComFA requires a free DNA end could shed light on when during the uptake process ComFA interacts with DNA. To this end, we measured ComFA stimulation by single-stranded, circular M13mp18 bacteriophage DNA. Since M13mp18 DNA is much longer than ssDNA oligonucleotides, we used the equivalent number of bases of either M13mp18 DNA or 50MB; we added 250 μg/mL of either M13mp18 DNA or ssDNA oligonucleotide. Although M13mp18 DNA did not stimulate ATPase activity to the degree of single-stranded oligonucleotides, this difference was not significant and may be attributed to the fact that M13mp18 DNA is more likely to fold into complex secondary structures than the single-stranded oligonucleotide we designed specifically to limit secondary structure (Fig. 1D). By quantifying the linear and circular DNA in a gel provided by NEB, we estimate that around 20% of M13mp18 ssDNA is linear (Fig. S1A), which would provide ComFA with a certain number of free ends. However, given that M13mp18 ssDNA is 7,249 bases in length, even with 20% linearized, this would result in >700-fold fewer free ends in the M13mp18 ssDNA mixture than in the mixture with 50-base oligonucleotides. The nearly 3 order of magnitude reduction in free ends would be predicted to significantly reduce ATPase activity if ComFA required a free end. Yet, the ATPase activity in the presence of M13mp18 ssDNA was only reduced by half. These data suggest that ComFA likely does not require an unobstructed end of DNA, and therefore, is capable of binding internally of ssDNA.

To test this conclusion using another method, we tested the binding of ComFA across a range of concentrations with gapped substrates—stretches of poly-dT ssDNA ranging in length from 5 nucleotides (5T gap) to 30 nucleotides (30T gap) flanked on either side by dsDNA (Fig. S1B). As we showed previously, ComFA binds poorly to dsDNA (Fig. 1B; Fig. S1C). Binding improves with the 5T and 10T gapped substrates, but ComFA binds best to 20T and 30T gapped substrates (Fig. S1C). In fact, ComFA showed equivalent binding to the 20T and 30T gapped substrates as it did to the ssDNA substrate 50 bases in length (ssDNA-50; Fig. S1C). This not only confirms that ComFA does not need a free end to bind but also suggests that the preferred binding site size of ComFA is at least 20 bases. It is worth noting that there is little to no binding to the 20-base ssDNA (ssDNA-20) substrate (Fig. S1C). It may be that, with such a short substrate, steric hinderance from the fluorescent tag on the single DNA strand reduces binding (Fig. S1C). Consistent with this idea, although 30-base sequences stimulate significant ATPase activity (Fig. S2A-B), the binding efficiency of ComFA to fluorescently tagged 30-base oligonucleotides is reduced, compared with longer substrates (Fig. S2C).

ComFA ATPase activity is DNA length dependent.

Helicases that translocate on nucleic acid substrate frequently show a dependence on substrate length (26, 27). That is, longer substrates stimulate higher ATPase activity, since longer substrates allow for further ATP-stimulated “travel” when the proteins are sufficiently processive. We examined the ATPase activity of ComFA in response to increasing lengths of ssDNA (5T to 100T) to determine whether ComFA might translocate on its substrate. Lengths of DNA less than or equal to 20 bases failed to significantly stimulate ATPase activity (Fig. 2A; Fig. S2A-B), but with longer substrates up to 80 bases in length, ComFA exhibited ATPase activity that increased with DNA length (Fig. 2A). Even over a range of substrate concentrations from 1–100 μM, longer oligonucleotides consistently stimulated higher ATPase activity (Fig. 2B and C).

FIG 2.

ATPase activity of ComFA is dependent on DNA length. (A) Mean ATPase activity of WT ComFA with poly-dT substrate of various lengths at a concentration of 10 μM. no DNA n = 20, 20T n = 2, 40T n = 7, 60T n = 7, 80T n = 5, 100T n = 4 (B) Mean ATPase activity of ComFA with poly-dT substrate of various lengths and concentrations. ATPase activity was evaluated using an NADH-coupled 96-well plate assay with indicated concentration and length of poly-dT ssDNA. 40T n = 3, 50T n = 4, 60T n = 3, 80T n = 1, 100T n = 1. (C) Mean V_max (1 to 100 μM) for indicated lengths of DNA from the data in (B). (D-F) ComFA ATPase activity with the concentration of DNA nucleotide plotted as dots alongside Scheme III (D), Scheme II (E), or Scheme I (F) of the model (with minimum binding site of 20 bases). Model is shown as lines of various shades of gray. Length of DNA is indicated by dot size and shade of gray, with lighter shades of gray and smaller dots indicating shorter lengths of DNA. Error bars represent standard deviation. For (A) and (C), a one-way ANOVA with Tukey’s multiple-comparison test was used to determine significance, denoted by *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. (D-F) An F-test was used to determine significance of data fit to Scheme III or II models relative to the null model of Scheme I. F-test values can be found in Table 1.

To rule out the possibility that the increase in ATPase activity stimulated by longer oligonucleotides is simply due to the increase in the number of nucleotide residues added with longer oligonucleotides, we plotted the ATPase activity of ComFA in response to different lengths of oligonucleotides against the concentration of nucleotides. If the DNA length is indeed critical, rather than the total number of nucleotides, we would expect, for instance, that the addition of 100T at 25 μM would stimulate a higher ATPase rate than 50T at 50 μM 50T, even though both yield 2.5 mM nucleotide bases. This is, in fact, what we observe: longer oligonucleotides stimulate higher ATPase rates even when the number of nucleotides is taken into consideration (Fig. S3).

Helicase dependence on substrate length has previously been exploited to distinguish between translocation and nontranslocation mechanisms (26–29). In fact, mathematical models have been developed for ATPase data generated from nontranslocating and translocating proteins, such that we were able to fit the ATPase data from ComFA to these models. In brief, we fit our data to three different models, based on those previously presented (26, 27). Scheme I (null model) represents a nontranslocating protein, with dependence on DNA concentration, but not DNA length. Scheme II represents a simple translocating protein that has the same dissociation rate constant in the middle of the substrate as at the end of the substrate. Scheme III represents a complex translocating protein that has two different dissociation rate constants—dissociation within the substrate molecule and dissociation at the end of the molecule (26, 27).

The model requires the input of a minimum DNA binding length. Our gapped substrate experiment suggests that ComFA has a minimum binding site of about 20 bases (Fig. S1). Further, while the ATPase activity stimulated by 20-base sequences (20T) trended up compared to that stimulated by 5T and 10T sequences, 30-base sequences stimulated significantly higher ATPase actity than 5T, 10T, and 20T sequences (Fig. S2A-B). Based on these data, we assigned the minimum binding length of ComFA to be 20 bases, such that the model considered ATPase data stimulated by oligonucleotides longer than 20 bases. However, we tested multiple sizes of binding sites in the model, and all led to similar results (Table 1).

TABLE 1.

Model output with different binding site sizes^a

	R2 of model fit
Binding site (bases)	Scheme III	Scheme II	Scheme I	F-test III versus I	F-test II versus I
5	0.93	0.80	0.55	1.14 × 10−13	8.59 × 10−7
10	0.95	0.82	0.55	4.89 × 10−15	1.54 × 10−7
15	0.95	0.85	0.55	4.44 × 10−16	1.65 × 10−8
20	0.96	0.88	0.55	0	8.71 × 10−10
25	0.96	0.90	0.55	0	2.58 × 10−11

^aR² and F-test values comparing Schemes I, II, and III when data (found in Fig. 2D to F) are fit to the models.

Scheme III provided the best fit for the ATPase data of ComFA (Fig. 2D; R² = 0.96), followed by Scheme II (Fig. 2E; R² = 0.88). On the other hand, fitting the data to Scheme I yielded the worst fit (Fig. 2F; R² = 0.55). The F-test finds that the data fit either of the translocating models (Scheme III or Scheme II) significantly better than the non-translocating model (null model; Scheme I; p < 0.0001), suggesting that the ATPase activity of ComFA in response to oligos of different lengths more closely resembles that of a translocating protein than a non-translocating protein. In addition, the closer fit of our data to Scheme III indicates that ComFA may have two different dissociation constants depending on whether it is dissociating in the middle or at the end of the DNA molecule.

The fact that increasing length past 80 bases did not increase ATPase activity (Fig. 2A to C; Fig. S3), suggests that ComFA has limited processivity. Because ComFA binds internally on DNA and therefore will, presumably, bind in the middle of a strand of DNA on average, we can conclude that ComFA can translocate up to 30–40 bases before dissociation.

ATPase activity of ComFA suggests 5′ to 3′ directionality of translocation.

Proteins that translocate on single-stranded nucleic acid often exhibit a preferred directionality (i.e., 5′ to 3′ or 3′ to 5′; 30, 31). While testing the enzymatic activity of ComFA in response to different nucleic acid substrates, we discovered an experimentally useful trait of ComFA: ComFA reproducibly exhibits roughly 3 times greater ATPase activity with poly-dT substrates than with mixed-base substrates (Fig. 3B). These findings, combined with the data demonstrating that ComFA binds internally to DNA, provided the means to test the directionality of translocation. We thus designed two different oligos: an oligo with 30 dTs on the 5′ end and 20 mixed bases on its 3′ end (30T20MB) and an oligo with 20 mixed bases on its 5′ end and 30 dTs on the 3′ end (20MB30T; Fig. 3A). If ComFA travels from 3′ to 5′, 30T20MB will stimulate higher ATPase activity, as ComFA will bind internally on the DNA and therefore, on average encounter more poly-dT than mixed-base sequence (see model in Fig. 3A). On the other hand, if ComFA travels from 5′ to 3′, 20MB30T will stimulate higher ATPase activity. We found that ComFA ATPase activity was stimulated to a significantly greater degree by 20MB30T (which stimulated ATPase activity equivalent to 50T) than by 30T20MB or by a 50 mixed-base oligo (50MB; Fig. 3B), consistent with the interpretation that ComFA likely translocates from 5′ to 3′.

FIG 3.

ComFA exhibits 5′ to 3′ directionality. (A) Model of expected outcomes with respective oligonucleotides if ComFA travels in a 3′ to 5′ or 5′ to 3′ direction. Gray line indicates poly-dT sequence, while the red line indicates mixed-base sequence. Black oval represents ComFA. (B) Mean ATPase activity of ComFA in the presence of indicated oligonucleotides (for DNA sequences, see Table 2). 50T n = 4, 50MB n = 3, 20MB30T n = 4, 30T20MB n = 4. Error bars represent standard deviation. A one-way ANOVA with Tukey’s multiple-comparison test was used to determine significance, denoted by *P ≤ 0.05, **P ≤ 0.01.

ComFA lacks helicase activity.

We next examined ComFA DNA binding and unwinding activities to determine whether ComFA has helicase activity. First, we tested the binding of ComFA without ATP to 50 and 30 poly-dT sequences (50T, 30T) and to two annealed double/single-stranded oligo combinations, one containing a 5′ dsDNA sequence 20 bp in length with a 30 base 3′ dT overhang (3′ dT overhang), and the other with a 20 bp 3′ dsDNA sequence with a 30 base 5′ dT overhang (5′ dT overhang; Fig. 4A). We found that ComFA bound either the 3′ dT overhang complex or the 5′ dT overhang complex with equal affinity and with similar affinity as it did 50T or 30T (<40 nM; Fig. 4B). The K_d of 138 ± 86 nM, reported in Diallo et al. (19) was higher than what we report here, but we attribute this to the use of the altered protein purification protocol combined with optimization of the EMSA conditions for these experiments.

FIG 4.

ComFA lacks helicase activity. (A) Diagram of single-stranded and double-stranded oligonucleotides used. Red lines indicate mixed-base sequences, gray lines indicate poly-dT sequences. 5′ dT overhang =30T20MB/20MB-Cy5 annealed, 3′ dT overhang=Cy5-20MB/20MB30T. See Table 2 for sequences. (B) Representative EMSAs (Materials and Methods “EMSA A”) with substrates (40 nM DNA, 1.1, 2.2, 4.4, 8.8, 17.5, 35, 70, 140, 280, 560, 0 nM ComFA). Red triangles indicate 70 nM ComFA (C) Representative DNA gels from strand displacement assay (40 nM DNA, lanes are: 1.1, 2.2, 4.4, 8.8, 17.5, 35, 70, 140, 280, 560, 0 nM ComFA, 560 nM ComFA without ATP, unfolded substrate, and folded substrate without ComFA). Red triangles indicate 70 nM ComFA. Gray triangles indicate unfolded DNA. Gray triangles indicate unfolded DNA, black triangles indicate folded (ds) DNA. Some protein did apparently stay complexed with the DNA in spite of the addition of a stop solution containing sodium dodecyl sulfate and proteinase k because, at higher concentrations of ComFA, a larger slow-mobility complex can be seen. Contrast and brightness were adjusted slightly for presentation. (D) Quantification of unwinding activity (intensity of unfolded band relative to ‘no ComFA’). Error bars represent standard deviation (n = 3). A two-way ANOVA with Tukey’s multiple-comparison test was used to determine significance, denoted by *P ≤ 0.05.

We then used a strand displacement assay to test for DNA unwinding ability of ComFA. We incubated various concentrations of ComFA with the ss-dsDNA complexes in the presence of ATP to allow strand displacement to occur. We found that ComFA does not unwind dsDNA in our experimental conditions (Fig. 4C and D). While the amount of unwound DNA trended up very slightly with the highest concentrations of ComFA, no difference between the 3′ dT overhang and 5′ dT overhang complexes could be detected, and the slightly higher values for 280 nM and 560 nM ComFA were not significantly different from the control (no ComFA). Therefore, the negligible amount of apparent unwinding is likely an artifact of the higher concentrations used.

ATP hydrolysis reduces DNA binding.

We showed previously that binding of ComFA to DNA does not require ATP hydrolysis because a Walker B (ATP hydrolysis) mutant bound DNA comparably to wild-type protein. This finding left the possibility that ATP hydrolysis could either fuel a power stroke for movement along DNA or cause ComFA to release the DNA (19). Thus, we next investigated ComFA binding and release in the presence or absence of nucleotide analogs. We examined the DNA binding ability of ComFA in the apo state or with AMP-PNP (a non-hydrolyzable ATP analog), ADP, or ATP to determine whether ATP hydrolysis impacts DNA binding. We found that, although ComFA still bound DNA in the presence of 5 μM ATP, its DNA binding was more than 7-fold weaker, with a K_d of 171.8 ± 14 nM, compared with 23.2 ± 0.9 nM, the K_d of the apo state (Fig. 5A to C). Five μM AMP-PNP also decreased DNA binding, although not as significantly, with a K_d of 108.4 ± 25 nM, similar to that of 5 μM ADP (100.4 ± 45 nM; Fig. 5A to C). Of note, the addition of 50 μM or 5 mM ADP reversed the effects of adding 5 μM ADP. The K_d of ComFA with 50 μM ADP was significantly reduced compared to the apo state (9.4 ± 1.1 nM; Fig. S4) and appeared to promote a multimeric or aggregated state, as most of the protein and DNA amassed in the wells at concentrations of protein 50 nM and above (Fig. S4).

FIG 5.

DNA binding of ComFA is weakened in the presence of nucleotide. (A) Representative EMSAs (Materials and Methods “EMSA B”) with and without 5 μM ATP, AMP-PNP, or ADP illustrate ComFA binding to 50T DNA substrate (40 nM DNA, 0, 4.8, 9.5, 19, 38, 76, 152, 306, 611 nM ComFA; red triangles indicate 76 nM ComFA on the EMSA). Contrast and brightness were adjusted slightly for presentation. (B-C) Binding curves and mean K_d of ComFA in the absence or presence of indicated nucleotide. Apo n = 3, ATP n = 4, AMP-PNP n = 4, ADP n = 3. Error bars represent standard deviation. A one-way ANOVA with Sidak’s multiple-comparison test was used for statistical analysis. *P ≤ 0.05, **P ≤ 0.01, ****P ≤ 0.0001.

DISCUSSION

Here, we provide the first experimental evidence that the competence protein, ComFA, translocates on ssDNA in a 5′ to 3′ direction in an ATP-dependent manner. This evidence strongly supports the idea that ComFA acts as a DNA uptake motor in Gram-positive bacteria and provides a possible mechanism for coupling the ATPase cycle of ComFA to DNA translocation during import.

Our results suggest that ComFA is unusual among DEAD box helicases, since few DEAD box helicases interact with DNA, and those with unwinding capabilities primarily melt strands of nucleic acid locally without translocation (25, 32–34). A translocating DEAD box helicase is not entirely without precedent, since the DEAD box helicase DDX43 was found to have weak translocase activity on DNA (35). However, the processivity of this protein was limited to fewer than 16 bases making it significantly less processive than ComFA, which we show can travel at least 30–40 bases before DNA release (35). ComFA also shares homology with PriA, which belongs to the closely related DExH-box helicases—another member of the helicase superfamily 2. PriA translocates on forked DNA substrates (36, 37), and in fact, may play a role in DNA uptake in Neisseria gonorrhoeae, a Gram-negative bacterium that lacks ComFA (38). Thus, although ComFA contains the DEAD box motif, functionally it may be more closely related to the translocating DExH family of helicases. Future studies may examine what features give ComFA translocation abilities, setting it apart from other DEAD box helicases.

Our results that ComFA exhibits a 5′ to 3′ directionality seem to contradict a previous study reporting that DNA uptake occurs in a 3′ to 5′ direction in S. pneumoniae (39). These conclusions were based on the authors’ findings that a 3′-labeled strand of DNA was more frequently detected inside the cells than a 5′-labeled strand. One way to reconcile our findings with these data are that ComFA may pull on the degraded strand of DNA from 5′ to 3′, while the transformed strand enters the cell in a 3′ to 5′ manner. However, this seems improbable, given recent data demonstrating that ComEC has an extracellular nuclease domain that likely degrades the nontransformed strand of DNA (40). As a more probable explanation, this initial report of DNA uptake directionality may have been incorrect in its conclusion that less internalized 5′ label implies a 3′ to 5′ directionality. The protein NucA is thought to cleave one end of the DNA before uptake in B. subtilis (41). It is possible that a similar protein in S. pneumoniae cleaves off the 5′ end of the DNA prior to uptake in a 5′ to 3′ direction. This would account for the greater quantity of 3′ label detected in the cells. Further, these results have not been replicated in 30 years, and researchers were unable to determine the directionality of uptake in B. subtilis using the same method (42). Clearly, the directionality of DNA uptake is a topic that needs further investigation.

We found that ComFA could not unwind dsDNA with the standard helicase substrates we tested, which would suggest that ComFA may act as a motor to drive uptake but likely does not participate directly in unwinding DNA as it enters the cell. This is consistent with the current model of DNA transformation in which strand separation occurs prior to entry into the cell whereupon it would encounter ComFA (1). However, it is possible that ComFA, as with some other helicases, may require a specific conformation of DNA, such as forked DNA, in order to unwind base pairs (43). We have not tested the unwinding ability of ComFA exhaustively with other conformations of DNA, so this leaves open the possibility that ComFA may have some unwinding capability not detected by our assay. However, given the likelihood that the nontransformed strand is degraded prior to entry into the cell (40), it is doubtful that any DNA unwinding detected by ComFA would be biologically relevant.

Our finding that ATP hydrolysis causes the release of DNA from ComFA is consistent with the mechanism of other DEAD box helicases (44). This finding, combined with the discovery that ComFA translocates in a direction-specific manner, allows us to propose a model for the ATPase cycle and DNA interaction, which could aid in understanding the role of ComFA in the DNA uptake complex and guide future experiments. A Brownian ratchet mechanism of translocation seems unlikely, since ComFA exhibits a strong directionality without exogenous forces, such as other competence proteins or a complementary strand of DNA that might force ComFA to move in one direction (45, 46). Therefore, we propose that ComFA uses an “inchworm” mechanism for translocation (47, 48), utilizing either multiple DNA binding sites on a single ComFA or a ComFA oligomer to move forward (Fig. 6). Although we previously reported that ComFA can oligomerize, it is unclear whether the protein is active as a monomer or multimer (19). For the sake of illustration, we will assume that ComFA acts as a monomer with two binding sites (BS 1 and BS 2; labeled as 1 and 2 Fig. 6), but BS 1 and BS 2 could also represent different ComFA proteins in an oligomer. We demonstrated that ComFA has a higher binding affinity for DNA in its apo state (Fig. 5A to D), suggesting that the initial step of DNA binding may be achieved prior to binding nucleotide (Fig. 6A). Upon binding to DNA (K_d = 23 nM), ComFA binds ATP, leading to a conformational change and higher K_d (108 nM; Fig. 6B). ATP hydrolysis forces BS 1 to release, yielding a less stable complex (K_d = 172 nM; Fig. 6C). After ATP hydrolysis, the presence of ADP allows for rebinding of BS 1 and release of BS 2, resulting in the binding state demonstrated by ADP-bound ComFA (K_d = 100 nM; Fig. 6D). The ADP is then released, and BS 2 rebinds, and the process repeats (Fig. 6A). Our finding that a higher concentration of ADP results in protein-DNA aggregates or higher order multimers suggests that ADP or DNA release is the rate-limiting step of the process.

FIG 6.

Potential mechanism for ComFA translocation. Maroon proteins labeled 1 and 2 represent either two binding sites on a single ComFA or distinct ComFA proteins in an oligomer bound to ssDNA. (A) ComFA binds ssDNA without nucleotide. (B) ComFA binds ATP, changing conformation leading to (C) ATP hydrolysis and the release of binding site 1. (D) The presence of ADP causes the rebinding of binding site 1 and release of binding site 2. Release of ADP leads to the release of DNA or rebinding of binding site 2 and restoration of original strong binding. The mean K_d of ComFA at each step is indicated in red. Graphic was created with BioRender.com.

The data we present here strongly support the idea that ComFA may contribute to powering DNA import by translocating along ssDNA, but much remains to be determined. No other DEAD box helicase has exhibited processive translocation to the extent of ComFA, to our knowledge, and further studies are necessary to confirm the translocase activity of ComFA and examine what sets ComFA apart from other DEAD box helicases. Further, our data raise questions regarding the directionality of the imported DNA strand. Future studies may also examine the oligomeric state of active ComFA, as well as its behavior within the DNA uptake complex. Protein-protein interactions, particularly with ComEC and ComFC, likely impact multiple parameters of ComFA’s activity, including the speed of translocation and processivity. Finally, we have proposed a model of translocation, but more experiments are necessary to expound on the mechanism of translocation of this unique protein essential to natural transformation in Gram-positive bacteria.

MATERIALS AND METHODS

Protein purification.

Protein purification was initially completed as described in Diallo et al. (19). Protein purified by this method was used in experiments shown in Fig. 1, 2A, and 3, as well as Supplemental Fig. S2 and S6. When the lab moved from Massachusetts to Wisconsin, however, we were unable to purify functioning ComFA using this method. Thus, we further optimized the protocol, including lowering the pH to 6.2 which is more appropriate for purification of the SUMO-ComFA fusion protein. Then, following cleavage of the SUMO tag, we raised the pH for ComFA alone. These preparations of protein were used in Fig. 2B–F, 4, 5 and Fig. S1, S3, and S4. A gel containing purified protein from both the original and revised protocol can be found in Fig. S5A. Figure S5B demonstrates that both protein preps exhibited ATPase activity that was lost in a Walker B (D205A) mutant. While generally, ComFA from both protein preps behaved similarly, we did find that ComFA purified with the original protocol had ATPase activity that was 4–10 times higher than that of ComFA purified with the revised protocol (Fig. S5; Fig. 2A versus 2B). Importantly, however, we found that the dependence of ATPase activity on DNA length was consistent between protein preparation methods and that the fit of the models followed the same trend with data collected using the original protein preparation protocol (Fig. S6).

The optimized protocol proceeds as follows. A single colony of bacteria was inoculated in LB broth with corresponding antibiotics and grown at 37°C overnight. Next, the overnight starting culture was diluted to 5 mL of LB broth with antibiotics to reach 0.1 OD600. Bacteria were grown at 37°C to 0.5 OD600, and then were further diluted to 30 mL of LB broth with antibiotics to reach 0.05 OD600. Bacteria were then grown at 42°C to 0.5 OD600, followed by induction with a final concentration of 0.5 mM IPTG and incubation at 16°C overnight. Cells were pelleted and resuspended in 3 mL lysis buffer (25 mM Bis-Tris pH 6.2, 700 mM NaCl, 5 mM MgCl₂, 0.13 CaCl2, 5 mM imidazole, 1 mM TCEP, 10% glycerol) and then frozen until use. Prior to purification, 3 mL lysis buffer were added to the thawed cells. Cells were disrupted twice using a One-Shot disrupter (Constant Systems). The lysate was then centrifuged for 15 min at 20,000 rcf, 8°C. Supernatant was incubated for 20–40 min with 1 mL pre-equilibrated beads (Co-NTA XPure Agarose Resin, precharged ion: Cobalt; equilibration buffer: 25 mM Bis-Tris pH 6.2, 700 mM NaCl, 5 mM imidazole, 1 mM TCEP, 10% glycerol). Beads were washed twice with 20 mL wash buffer (25 mM Bis-Tris pH 6.2, 700 mM NaCl, 20 mM imidazole, 1 mM TCEP, 10% glycerol) and protein was eluted using 4 mL 400 mM imidazole elution buffer (25 mM Bis-Tris pH 6.2, 700 mM NaCl, 400 mM imidazole, 1 mM TCEP, 10% glycerol). Protein was then dialyzed overnight (dialysis buffer: 20 mM Bis-Tris pH 6.2, 500 mM NaCl, 1 mM TCEP, 10% glycerol) with 100 μl SUMO protease (concentration 0.41 mg/mL) per 1 mL eluate. For proteins to be used in electromobility shift assays (EMSAs), an additional dialysis step (20 mM Bis-Tris pH 7.5, 500 mM NaCl, 1 mM TCEP, 5% glycerol) was used to increase the pH.

Electrophoretic mobility shift assay A (Fig. 4).

ComFA at 1.09–560 nM was incubated with 40 nM Cy5-labeled oligos, acquired from IDT, in 50 mM Tris-Cl pH 7.5, 50 mM NaCl, 1 mM DTT, 0.1 mg/mL BSA, and 5 mM MgCl₂ in 25 μl final volume for 30 min at room temperature. 3.3% glycerol was added to samples, and 5 μl of each sample was loaded onto a 5% acrylamide 1.5-mm gel in TBE buffer. Gels were prerun at 75 V for 20 min before loading protein/DNA complexes and running at 75 V for 1 h at 4°C in 1× TBE running buffer. Gels were imaged on the Azure c600 (Azure Biosystems). For analysis, ImageJ was used to measure integrated density of folded and unfolded DNA.

Electrophoretic mobility shift assay B (Fig. 1 and 5, Fig. S2, and S4).

Protein was diluted into EMSA buffer (2.5 mM CaCl2, 12.5 mM MgCl₂, 75 mM NaCl, 10 mM Tris pH 7.5, and 12.5% glycerol) with 0.1 mg/mL BSA and 5 nM Cy5-labeled DNA, acquired from IDT, in a total volume of 20 μl. Samples were incubated for 8–10 min, and 16 μl was loaded onto a 5% acrylamide 1.5 mm gel in Tris-Borate-EDTA (TBE) buffer. Gel was prerun at 85 V in 0.5× TBE for 20 min then protein/DNA was added, and the gel was run at 100 V for 45–60 min and imaged using a GE Healthcare Typhoon FLA 9000. dsDNA for Fig. 1B was created by combining in equal parts 50MB and 50MB-rev in STE buffer (200 mM Tris-HCl, 100 mM EDTA) at a concentration of 1 mM. This mixture was incubated for 5 min at 95°C then slowly cooled to room temperature in the heating block over the course of an hour. For dsDNA in Fig. S1, oligonucleotides were diluted to 10 μM then mixed with a slight excess of unlabeled oligo. For gapped substrates in Fig. S1, oligos were diluted to 10 μM then mixed with a slight excess of Cy5-labeled 20MB. Oligos were then annealed using a PCR machine, starting at 98°C and stepping down 1°C per 30 s. Oligos were then diluted in water to 100 nM. For analysis, ImageJ was used to measure integrated density bound DNA and free DNA in order to calculate the proportion of bound DNA. The dissociation constant (K_d) was considered that concentration at which 50% of the DNA was bound. The gel analysis tool from ImageJ was used to analyze the EMSAs with gapped substrates (Fig. S1). This tool was shown to yield the same results as when the integrated density was used for analysis (https://imagej.nih.gov/ij/docs/examples/dot-blot/index.html), and made background subtraction quite straightforward.

ATPase activity assay.

ATPase activity was measured using an NADH-coupled plate assay (49) as described in Diallo et al. (19).

Oligos were obtained from IDT or Sigma-Aldrich. dsDNA for Fig. 1 was created by combining in equal parts 50MB and 50MB-rev in STE buffer (200 mM Tris-HCl, 100 mM EDTA) at a concentration of 1 mM. This mixture was incubated for 5 min at 95°C then slowly cooled to room temperature in the heating block over the course of an hour. 20MB30T and 30T20MB were used to demonstrate directionality of ComFA (Table 2).

TABLE 2.

Oligos used in this study^a

Oligo ID	Sequence
50MB	CCGACCCACCATGCGTTTGTAACCCGACCCACCGTTCGTTCCTACCCACC
20MB30T	TGGCGACGGCAGCGAGGGACTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT
30T20MB	TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTCAGGGAGCGACGGCAGCGGT
50MB-rev	GGTGGGTAGGAACGAACGGTGGGTCGGGTTACAAACGCATGGTGGGTCGG
30MB	CCGACCCACCATGCGTTTGTAACCCGACCC
40MB	CCGACCCACCATGCGTTTGTAACCCGACCCACCGTTCGTT
20MB30T-B	CAGGGAGCGACGGCAGCGGTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT
20MB	ACCGCTGCCGTCGCTCCCTG
20MB-Cy5	ACCGCTGCCGTCGCTCCCTG-Cy5
Cy5−20MB	Cy5-ACCGCTGCCGTCGCTCCCTG
5T gap	CAGGGAGCGACGGCAGCGGTTTTTTCAGGGAGCGACGGCAGCGGT
10T gap	CAGGGAGCGACGGCAGCGGTTTTTTTTTTTCAGGGAGCGACGGCAGCGGT
20T gap	CAGGGAGCGACGGCAGCGGTTTTTTTTTTTTTTTTTTTTTCAGGGAGCGACGGCAGCGGT
30T gap	CAGGGAGCGACGGCAGCGGTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTCAGGGAGCGACGGCAGCGGT

^aOligos were ordered from Sigma-Aldrich or IDT. Labeled oligonucleotides were HPLC-purified by the company.

Modeling.

Kinetic models have been developed for ATPases that bind and translocate along DNA that differ based on the underlying mechanism. A set of models that consider a segment of DNA as a lattice of binding sites that allows the ATPase to bind to the lattice and subsequently translocate across the lattice and fall off with different rate constants were proposed by Young et al. (26, 27). Below are the two kinetic models presented by Young et al. and tested in this work that depend on the length and concentration of the DNA substrate (Scheme II and Scheme III). We compare the fit of the data to these models relative to a null kinetic model (Scheme I) of a nontranslocating ATPase that depends only on the concentration of DNA, and not on the length of the lattice. We use an F-test to perform a statistical comparison of these models.

Scheme I:

$${\text{v}}_{\text{ATP}}=\frac{{\text{v}}_{\text{Max},\text{ATP}}\times [\mathbf{DNA}]}{{\text{K}}_{\text{ATP}}+[\mathbf{DNA}]}$$

where, v_{Max, ATP} is the saturating ATPase rate and K_ATP is the [DNA] half-maximal ATPase rate. [DNA] corresponds to the nucleotide concentration of lattice binding sites, approximated as the molarity of nucleotides. This is the null model that is independent of DNA length (DNALength).

Scheme II: Scheme I with K_ATP being dependent on DNA length.

$${\text{v}}_{\text{ATP}}=\frac{{\text{v}}_{\text{Max},\text{ATP}}\times [\mathbf{DNA}]}{\frac{{\text{A}}_{\text{uni}}}{(\mathbf{DNALength}-\text{bindingSite})}+{\text{K}}_{\text{diss}}+[\mathbf{DNA}]}$$

where, V_Max,ATP is the saturating ATPase rate, K_diss is the dissociation constant of ATPase from DNA and A_uni is a composite parameter describing ATPase binding to DNA. bindingSite is the number of nucleotides required for DNA binding by the protein, and (DNALength-bindingSite) is the number of unique binding sites available on the DNA. [DNA] corresponds to the nucleotide concentration of lattice binding sites, approximated as the molarity of nucleotides.

Scheme III: Scheme I with V_{Max, ATP} being dependent on DNA length.

$${\text{v}}_{\text{ATP}}=\frac{{\text{v}}_{\text{Max},\text{ATP},\text{DNALength}}\times \left[\mathbf{DNA}\right]\times (\mathbf{DNALength}-\text{bindingSite})}{\left({\text{K}}_{\text{ATP}}+\left[\mathbf{DNA}\right]\right)\times ({\text{K}}_{\text{Length}}+(\mathbf{DNALength}-\text{bindingSite})}$$

where, V_{Max, ATP, DNALength} is the saturating ATPase rate, K_ATP and K_Length are composite parameters that depend on the dissociation constant and translocation rate of ATPase on the DNA lattice. [DNA] corresponds to the nucleotide concentration of lattice binding sites, approximated as the molarity of nucleotides.

$$\text{F}-\text{statistic}=\frac{\raisebox{1ex}{$\left({\text{SS}}_1-{\text{SS}}_2\right)$}\!\left/ \!\raisebox{-1ex}{${\text{df}}_1-{\text{df}}_2$}\right.}{\raisebox{1ex}{${\text{SS}}_2$}\!\left/ \!\raisebox{-1ex}{${\text{df}}_2$}\right.}$$

$$\text{p}-\text{value}=1-\text{fcdf}(\text{F}-\text{statistic},\left({\text{df}}_1-{\text{df}}_2\right),{\text{df}}_2)$$

which can be used in hypothesis testing of an alternate model (2) against the null model (1), given that model (1) is a nested model of model (2). SS_x represents the sum-squared error of data given a model with df_x representing the degrees-of-freedom of a model (number of unique data points – number of fitted parameters). The P value is estimated as the probability that the null model cannot be discarded given the F-statistic.

Modeling fitting and statistical testing was performed in MATLAB (MATLAB 2018b, The MathWorks, Inc., Natick, Massachusetts, United States.) using functions from the core and curve fitting toolboxes.

Strand displacement assay.

Duplex DNA (dsDNA) structures Cy5-20MB/20MB30T or 30T20MB/20MB-Cy5 were annealed in 10 nM Tris-Cl pH 7.5 and 100 mM NaCl to a final concentration of 5 μM by heating to 95°C in a heat block and slowly cooling to room temperature over several hours. ComFA at 1.09–560 nM was incubated with 40 nM Cy5-labeled oligos, acquired from IDT, in 50 mM Tris-Cl pH 7.5, 50 mM NaCl, 1 mM DTT, 0.1 mg/mL BSA, 5 mM MgCl₂, 4 mM ATP, and 4 μM trap oligonucleotide (unlabeled 20MB) in 25 μl final volume for 15 min at 37°C. Melted controls were heated to 95°C for 10 min. Reactions were quenched using 5 μl stop buffer (2% Sodium dodecyl sulfate, 5 μg/mL Proteinase K, 20% glycerol, and 0.1 M EDTA), and 5 μl of each sample were loaded on 15% acrylamide 1.5-mm gels in Tris-Borate-EDTA (TBE) buffer. Gels were run for 1 h at 75 V at 4°C in 1× TBE running buffer. Gels were imaged using the Azure c600 (Azure Biosystems).

Data availability.

MATLAB code and raw data are made available at (https://github.com/TheBurtonLab/Foster-et-al.-2022.).