ComE/ComE∼P interplay dictates activation or extinction status of pneumococcal X-state (competence)


  • Bernard Martin,

    1. Centre National de la Recherche Scientifique, LMGM-UMR5100, Toulouse, France
    2. Université de Toulouse, UPS, Laboratoire de Microbiologie et Génétique Moléculaires, Toulouse, France
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Anne-Lise Soulet,

    1. Centre National de la Recherche Scientifique, LMGM-UMR5100, Toulouse, France
    2. Université de Toulouse, UPS, Laboratoire de Microbiologie et Génétique Moléculaires, Toulouse, France
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Nicolas Mirouze,

    1. Centre National de la Recherche Scientifique, LMGM-UMR5100, Toulouse, France
    2. Université de Toulouse, UPS, Laboratoire de Microbiologie et Génétique Moléculaires, Toulouse, France
    3. INRA, UMR1319 Micalis (Microbiologie de l'Alimentation au service de la Santé), Jouy-en-Josas cedex, France
    Search for more papers by this author
  • Marc Prudhomme,

    1. Centre National de la Recherche Scientifique, LMGM-UMR5100, Toulouse, France
    2. Université de Toulouse, UPS, Laboratoire de Microbiologie et Génétique Moléculaires, Toulouse, France
    Search for more papers by this author
  • Isabelle Mortier-Barrière,

    1. Centre National de la Recherche Scientifique, LMGM-UMR5100, Toulouse, France
    2. Université de Toulouse, UPS, Laboratoire de Microbiologie et Génétique Moléculaires, Toulouse, France
    Search for more papers by this author
  • Chantal Granadel,

    1. Centre National de la Recherche Scientifique, LMGM-UMR5100, Toulouse, France
    2. Université de Toulouse, UPS, Laboratoire de Microbiologie et Génétique Moléculaires, Toulouse, France
    Search for more papers by this author
  • Marie-Françoise Noirot-Gros,

    1. INRA, UMR1319 Micalis (Microbiologie de l'Alimentation au service de la Santé), Jouy-en-Josas cedex, France
    Search for more papers by this author
  • Philippe Noirot,

    1. INRA, UMR1319 Micalis (Microbiologie de l'Alimentation au service de la Santé), Jouy-en-Josas cedex, France
    Search for more papers by this author
  • Patrice Polard,

    1. Centre National de la Recherche Scientifique, LMGM-UMR5100, Toulouse, France
    2. Université de Toulouse, UPS, Laboratoire de Microbiologie et Génétique Moléculaires, Toulouse, France
    Search for more papers by this author
  • Jean-Pierre Claverys

    Corresponding author
    1. Centre National de la Recherche Scientifique, LMGM-UMR5100, Toulouse, France
    2. Université de Toulouse, UPS, Laboratoire de Microbiologie et Génétique Moléculaires, Toulouse, France
    • For correspondence. E-mail; Tel. (+33) 561 33 59 11; Fax (+33) 561 33 58 86.

    Search for more papers by this author


Since 1996, induction of competence for genetic transformation of Streptococcus pneumoniae is known to be controlled by the ComD/ComE two-component regulatory system. The mechanism of induction is generally described as involving ComD autophosphorylation, transphosphorylation of ComE and transcriptional activation by ComE∼P of the early competence (com) genes, including comX which encodes the competence-specific σX. However, none of these features has been experimentally established. Here we document the autokinase activity of ComD proteins in vitro, and provide an estimate of the stoichiometry of ComD and ComE in vivo. We report that a phosphorylmimetic mutant, ComED58E, constructed because of the failure to detect transphosphorylation of purified ComE in vitro, displays full spontaneous competence in ΔcomD cells, an that in vitro ComED58E exhibits significantly improved binding affinity for PcomCDE. We also provide evidence for a differential transcriptional activation and repression of PcomCDE and PcomX. Altogether, these data support the model of ComE∼P-dependent activation of transcription. Finally, we establish that ComE antagonizes expression of the early com genes and propose that the rapid deceleration of transcription from PcomCDE observed even in cells lacking σX is due to the progressive accumulation of ComE, which outcompetes ComE∼P.


Bacterial transformation, which changes genotypes via homologous recombination, was discovered in the pneumococcus, Streptococcus pneumoniae (Griffith, 1928). In several species including S. pneumoniae, transformation depends on achieving a specialized state, competence, during which the transformasome is assembled. This dynamic machine involves both membrane and cytosolic proteins, and internalizes, protects, and processes transforming DNA (Claverys et al., 2009). Competence or X-state (Claverys et al., 2006) is induced in exponentially growing pneumococcal cultures by a secreted competence-stimulating peptide (CSP), encoded by comC (Håvarstein et al., 1995). CSP constitutes, together with the membrane-bound histidine kinase (HK) ComD and the response regulator (RR) ComE (Pestova et al., 1996), the master competence switch.

The ComD–ComE pair is a classical two-component signal transduction system (TCS) (Gao and Stock, 2009) with ComE belonging to the LytTR subfamily of RRs (Nikolskaya and Galperin, 2002; Galperin, 2006). These RRs, like PlnC, the regulator of bacteriocin production in Lactobacillus plantarum (Diep et al., 2003), and AgrA, the regulator of Staphylococcus virulence (Novick and Geisinger, 2008; Thoendel et al., 2011), bind a target site consisting of a 9 bp direct repeat (DR) separated by a stretch of 12 nucleotides (Sidote et al., 2008) resulting in a spacing between repeated nucleotides of 20 bp, i.e. exactly two helix turns (Nikolskaya and Galperin, 2002). The identification of potential ComE binding site (CEbs) exhibiting this structure in the comCDE promoter region (Fig. 1A) and the observation that in vitro ComE specifically bound to this region led to a model of direct transcriptional activation of comCDE by ComE in response to CSP (Ween et al., 1999). According to this model, at a critical CSP concentration, CSP interacts with its receptor ComD, which is activated. ComD autokinase (to produce phospho-ComD, ComDP) and phosphotransfer (from ComDP to ComE) reactions would produce ComEP, which then binds and activates transcription from its own promoter. ComE∼P is also believed to activate transcription of the comAB operon, which encodes the machinery required for maturation and export of the comC-encoded pre-CSP (Hui et al., 1995). Thus, by increasing CSP production, export and signal transduction capacities, ComE∼P creates a positive feedback loop, amplifying the signal and co-ordinating competence throughout the population. ComE∼P also induces expression of comX, which encodes an alternative sigma factor, σX (Lee and Morrison, 1999) and comW, which encodes a protein required for both the stabilization and activation of σX (Sung and Morrison, 2005). σX directs the transcription of the so-called late com genes (Dagkessamanskaia et al., 2004; Peterson et al., 2004), which include those encoding components of the transformasome.

Figure 1.

Expression of pneumococcal com genes.

A. Putative regulatory sites at early com genes. Alignment of early com promoter sequences incorporating the newly defined comW, comM, def and spr1962 (R6 genome annotation) promoters (Supporting results and discussion; Fig. S1). Bases matching or differing from the CEbs consensus in blue and red respectively. Boxed residues in the comCDE regulatory region indicate previously studied substitutions in LR (Ween et al., 1999) and in −10 (Martin et al., 2010). sp genes (TIGR4 genome annotation) with CEbs and/or −10 differing from those previously proposed (Peterson et al., 2004) are underlined. The consensus was generated using Weblogo (; Crooks et al., 2004).

B. Comparison of com gene expression in R800 and CP1250 derivatives. (Left panels) Activity of early PcomCDE, PcomX and PcomW, and late PssbB promoters was monitored using transcriptional fusions with the Firefly luciferase gene (Experimental procedures). CSP (200 ng ml−1) was added after 12 min incubation at 37°C to cultures inoculated in CTM pH 7 (Martin et al., 2000) at OD = 0.005 (from culture growing in CTM to OD = 0.2–0.3) and luminescence and OD492 were recorded at the indicated time points. (Right panels) Apparent rates of transcription of com genes in R800 and CP1250 derivatives (expressed as RLU increase per arbitrary time unit; calculated at a given time point, t1, by substracting RLU value at t0 from RLU value at t1). Strains used (with promoter under investigation indicated between parenthesis) as follows. Upper panel: R800 derivatives R2227 (comC), R2271 (comX), R2353 (comW) and R1313 (ssbB). Lower panel: CP1250 derivatives, TCP1257 (comC), TCP1265 (comX), TCP1273 (comW) and TCP1253 (ssbB).

The model of ComE∼P-dependent X-state induction has received indirect support over years. Thus, substitution of four conserved positions in the left repeat of comCDE putative CEbs (Fig. 1A) resulted in abolition of in vitro binding of ComE (Ween et al., 1999). Electrophoresis mobility shift assays (EMSA) of two other early com promoter fragments, harbouring presumptive CEbs (those of the comAB and qsrAB operons) provided further evidence that DRs are part of the sequence elements primarily recognized by ComE (Ween et al., 1999). Furthermore, the location of comCDE and comAB transcription start sites on RNA extracted from competent pneumococci (Halfmann et al., 2007), and the abolition of basal as well as CSP-induced ComE-dependent transcription of the comCDE operon by a mutation of the CEbs-associated −10 (Martin et al., 2010) were fully consistent with transcription initiating from the proposed ComE-dependent regulatory site. Finally, the finding that basal comCDE expression was strongly reduced in cells harbouring the ComED58A mutation, which is predicted to abolish transphosphorylation of ComE by ComD∼P, supported the view that ComE∼P is required for basal expression of comCDE (Martin et al., 2010).

Nevertheless, more than 15 years after the characterization of CSP and the ComD/ComE TCS, three key features of the model, namely ComD autophosphorylation, transphosphorylation of ComE, and the dependence of CSP-induced transcriptional activation of early com genes on ComE∼P have still not been experimentally established. The hierarchy of early com promoters is also not documented. It is thus unknown whether the PcomX and PcomW promoters are weaker or stronger than PcomCDE. This situation prompted us to investigate these aspects. Here, we document in vitro the autokinase activity of wild-type ComD and the constitutive mutant, ComDT233I (Martin et al., 2000), and provide an estimate of the stoichiometry of ComD and ComE in vivo. We report the in vivo and in vitro characterization of the ComED58E phosphorylmimetic mutant, constructed due to our failure to detect transphosphorylation of purified ComE in vitro. After redefining some of the early com promoters, we provide evidence for a differential transcriptional activation of the PcomCDE, PcomX and PcomW promoters, as well as a differential repression of PcomCDE and PcomX. Finally, we report the unexpected finding that while ComE∼P acts as an activator, ComE controls extinction of early com gene expression, and that a rapid deceleration of transcription from PcomCDE occurs even in the absence of σX. As unphosphorylated ComE antagonizes transcriptional activation, we propose that its progressive accumulation outcompetes ComE∼P and accounts for the observed deceleration of PcomCDE transcription, despite the presence of fully inducing CSP concentrations in the medium. We suggest that together with the rapid decay of comCDE mRNA (Alloing et al., 1998), this mechanism accounts at least in part for the shutoff of competence occurring in pneumococcal cultures despite the presence of extracellular concentrations of CSP far above the inducing threshold.


Re-examination of early com promoters

A previously published alignment of the early com promoters (Peterson et al., 2004) suggested some heterogeneity in the spacing between left (LR) and right (RR) repeats (from 10 to 12 bp) as well as in RR to −10 box distances (from 13 to 42 bp). Distance heterogeneity for the proposed ComE-dependent promoters was puzzling since the conservation of the DR to −10 spacing is supposed to reflect the existence of close contacts between LytTR RRs and the RNA polymerase. This situation contrasted with promoters regulated by two members of the LytTR subfamily of RRs, AgrA and PlnC, for which we calculated rather homogeneous distances between RR and −10; 31 bp for the AgrA-regulated P2 and P3 promoters (Koenig et al., 2004), and 32 bp for all PlnC-dependent promoters except PM (31 bp) (Diep et al., 2003). This prompted us to re-examine the early com promoters looking for potential CEbs, with special focus on comW. We identified alternative CEbs in front of comW, comM, def and spr1962 (Fig. 1A). The new CEbs for comW was supported both by evolutionary arguments and experimentally (Supporting results and discussion; Fig. S1).

Together with the new CEbs for comW, the proposed corrections for comM, def and spr1962 CEbs (Supporting results and discussion) result in an identical 12 bp spacing between LR and RR (except for def, 13 bp) and a more homogeneous RR to −10 distance, 31 or 32 bp (Fig. 1A), i.e. an organization much similar to that of promoters controlled by other LytTR RRs. It is of note that PcomW, Pdef and PcomM contain the most divergent LRs, while the least conserved RR is that of PcomX (Fig. 1A).

Hierarchical expression of early com promoters

To explore the kinetics of com gene expression, com::luc transcriptional fusions were constructed at the early com loci, comC, comX and comW, and the late gene ssbB (Experimental procedures). In every case, the luc gene was inserted at a very similar position (next to the 13th or 14th codon of the targeted com gene) with luc translation relying on an identical RBS (Fig. S2). Luciferase activities were then compared in the same experiment and in two different genetic backgrounds, R800 (our laboratory strain) and CP1250 (in use in D. Morrison's laboratory). Expression was recorded every 23 s after CSP addition over ∼ 40 min (Fig. 1, left). PcomCDE, PcomX and PcomW displayed similar kinetics with maximal transcription rates attained 9–10 min after CSP addition in both R800 and CP1250 derivatives (Table 1). However, maximal expression rate of comC appeared 2.2- and 3.7-fold higher than that of comX and comW, respectively, in R800. Although maximal transcription rates were 2.2–2.4-fold higher in CP1250 derivatives, a similar hierarchy was observed among early com promoters (Table 1).

Table 1. Kinetics of expression of early and late com genes in R800 and CP genetic backgrounds
  Genetic backgroundaPromoter
Early com geneLate com gene
  1. aR800 and CP derivatives used are indicated in Fig. 1 legend.
  2. bmin post-CSP addition.
  3. c2.0 and 5.0 min delay between early and late com gene expression in R800 and CP backgrounds respectively.
  4. dmin.
  5. eRLU per time unit; calculated from luciferase expression measurements (e.g. Fig. 1C); values represent average of six independent experiments (including those in Fig. 1) ± standard error.
Time to maximal transcription rate (min)b R8009.0 ± 0.39.6 ± 0.69.6 ± 0.711.4 ± 0.6c
 CP10.0 ± 0.610.1 ± 0.49.6 ± 0.914.9 ± 0.9c
CP difference to R800d 1.1 ± 0.50.5 ± 0.50.0 ± 0.53.5 ± 0.6
Maximal transcription rate (arbitrary units)eabsoluteR80017.1 ± 3.58.0 ± 1.65.4 ± 0.921.6 ± 3.0
  CP38.8 ± 4.516.9 ± 2.612.7 ± 2.039.4 ± 3.7
 relative to PcomXR8002.2 ± ± 0.22.8 ± 0.4
  CP2.3 ± ± 0.12.4 ± 0.4
 CP relative to R800 2.4 ± 0.62.2 ± 0.52.3 ± 0.31.8 ± 0.3

In contrast, a larger delay between early and late com gene expression was observed in CP1250 compared with R800 (Fig. 1B). ssbB transcription thus initiated in CP1250 ∼ 8 min after CSP addition versus 4 min in R800 (Fig. 1, left panels). As a consequence, the maximal rate of transcription of ssbB was attained about 3.5 min later in CP1250 than in R800, and the delay between the maximal rates of transcription of early and late genes was respectively 5 and 2 min (Table 1). The much shorter delay between expression of early and late com genes in R800 is fully consistent with the previously observed expression overlap between the two classes (Dagkessamanskaia et al., 2004). The larger delay in CP1250 remains unexplained particularly since σ70 replacement by σX can be expected to occur similarly in the two backgrounds in view of the striking similarity in comX and comW expression kinetics. Nevertheless, despite the existence of this longer lag and of differences in final transcription levels, the hierarchy among early com promoters, PcomCDE, PcomX and PcomW, is very similar in R800 and CP1250. We attribute the weaker transcription of PcomW and PcomX compared with PcomCDE to divergence from the consensus in the LR and RR respectively (Fig. 1A).

ComER120S derepresses PcomX

A single-residue change in comE leading to synthesis of the ComER120S mutant protein was previously found to confer a competence-upregulated (cup) phenotype (Martin et al., 2000). The mutant strain was shown to display an extended competence period, as evaluated through direct measurement of transformation, but the underlying mechanism remained unknown. Differences observed between PcomCDE and PcomX activity prompted us to analyse the impact of ComER120S on these promoters. Expression profiles clearly showed that while PcomC transcription was only marginally affected, PcomX transcription was prolonged in comER120S mutant cells (Fig. 2). Conversion of expression profiles into transcription rates confirmed this conclusion and revealed that 15 min after CSP addition, PcomX became even more active than PcomC in cells harbouring the comER120S mutation (Fig. 2, compare upper and lower panels). We conclude that the primary effect of the comER120S mutation is to boost transcription from PcomX, which accounts for both the cup phenotype and the previously documented extended competence period exhibited by mutant cells.

Figure 2.

Impact of the comER120S mutation on comCDE and comX expression. Apparent rates of transcription of PcomCDE and PcomX in comE+ (black symbols) and comER120S (red symbols) cells were calculated from expression data as described in the legend of Fig. 1. Expression was monitored in cells harbouring a comC::luc fusion (top panel: R1960, black open symbols; R2329, red symbols) and a comX1::luc (triangles) or comX2::luc (squares) fusion (bottom panel: R2218 and R2200, black open symbols; R2313 and R2314, red symbols). CSP was added after 10 min incubation in C + Y (Experimental procedures).

These observations provided further evidence for a qualitative difference between PcomCDE and PcomX. They suggest that the design of early com promoters tends to limit comX transcription, which constitutes a hitherto ignored layer of control of σX, in addition to the previously documented post-transcriptional control exerted by ComW, involving both stabilization and activation of ComX (Sung and Morrison, 2005).

ComD–ComE cellular amounts, HK autophosphorylation and transphosphorylation of RR

Our desire to document the binding of ComE∼P to PcomCDE and PcomX promoters in vitro necessitated the phosphorylation of ComE. Unlike the pneumococcal RR CiaR (Halfmann et al., 2011), we could not detect phosphorylation of purified ComE by acetylphosphate (data not shown). This prompted us to attempt transphosphorylation of ComE by its cognate HK in vitro. The C-terminal cytoplasmic domain of wild-type ComD (ComD-Cter) was therefore purified and its autophosphorylation investigated in vitro (Experimental procedures and Supporting Information). In parallel, we studied the cup mutant ComDT233I previously shown to exhibit near wild-type transformation levels in the absence of CSP (Martin et al., 2000), which suggested increased autophosphorylation of the mutant protein. Consistent with this expectation, 32P labelling was detected with both proteins but ComDT233I incorporated ∼ 40-fold more label than the wild-type protein. A phosphorylated oligomeric form, presumably a dimer, representing ∼ 1/10th of the label in the monomer was observed with ComDT233I (Fig. S3).

Next, before assaying the transphosphorylation of ComE proteins by phosphorylated ComDs in vitro, we determined the stoichiometry of ComE and ComD through Western blot analysis (Fig. S4). Our calculations resulted in an estimate of ∼ 87 000 monomers per cell for ComE and ∼ 39 000 for ComD (Supporting results and discussion), i.e. a ComE/ComD ratio close to 2, in contrast to the situation for the archetypical EnvZ/OmpR (30-fold molar excess of OmpR RR over EnvZ HK) and the WalK/WalR (14-fold excess of the RR WalR) (Wayne et al., 2010). We also deduced the basal level of ComE through measurement of the induction factor by comparing CSP induced and non-induced R1501 cell extracts (Fig. S4D). An induction factor of ∼ 30-fold, in the range routinely observed with comC::luc transcriptional fusions, was obtained, leading to an estimate of ∼ 2900 ComE monomers per cell. About 1450 monomers of ComD should thus be present per cell before CSP addition, assuming there is no change in the ComE/ComD ratio upon induction. With this information in hand, despite the use of various concentrations as well as combinations of wild-type proteins and cup variants (ComDT233I and ComER120S), we failed to detect any transphosphorylation of ComEs (Fig. S3; data not shown). Whatever the reason for this failure (Supporting results and discussion), this led us to construct the ComED58E mutant in which Asp58, the putative acceptor site for transphosphorylation by ComD∼P, was changed into Glu to mimic the phosphorylated state of Asp (Klose et al., 1993).

Characterization of the phosphorylmimetic ComED58E protein

ComED58E mutant strains were constructed in a ΔcomC background by site-directed mutagenesis of comE as described in Experimental procedures. The first observation of very low transformation frequencies in the absence of CSP (Experiment A in Table 2) suggested that the phosphorylmimetic mutant could not activate transcription of the early com genes. A second observation, that addition of CSP resulted in significant (137–694-fold) increase in transformation (Experiment A in Table 2), was also unexpected as it suggested that CSP could still trigger ComD-dependent activation of ComE, despite mutation of the phosphate-accepting residue in the latter. Nevertheless, CSP-induced transformation frequencies remained 10–50-fold lower than average frequencies observed with wild-type cells. Monitoring of comC and comX expression in comED58E cells revealed that CSP stimulated PcomCDE more efficiently than PcomX (Fig. 3A). In any case, stimulation of expression was reduced compared with wild-type cells, thus accounting for the lower transformation frequencies.

Figure 3.

Impact of the comED58E mutation on comCDE and comX expression.

A. (Left panels) Expression of comC::luc and comX1::luc fusions was monitored in ΔcomC comE+ (Grey symbols; Top, R1960; Bottom, R2218) and ΔcomC comED58E (Red and green symbols; Top, R2446 and R2247; Bottom, R2403 and R2404) cells following addition of CSP after 12 min incubation at 37°C in C + Y medium (Experimental procedures). (Right panels) Apparent rates of transcription of com genes in comE+ and comED58E cells. Rates were calculated from data in the left panels as described in the legend of Fig. 1.

B. Constitutive high level expression of PcomX in ΔcomD comED58E cells. Expression of a comX1::luc fusion was compared in ΔcomC ΔcomD comED58E (R2588, green triangles; R2590, red triangles), ΔcomC comD+ comED58E (R2403, light grey triangles; R2404, dark grey triangles) and ΔcomC comD+ comE+ cells (R2218, black triangles) cells following addition of CSP (100 ng ml−1) at time 0 in C + Y medium.

Table 2. Transformability of comED58E cells: effect of CSP and ComD
ExperimentStrains (all ΔcomC comED58E)Clonecfu ml−1% SmRRatio (+CSP/−CSP)
  1. a. Not detectable.
  2. b. Calculated on the basis of 1 SmR colony per plate (instead of 0) at the lowest dilution plated (c) not done.
  3. d. Scoring for NovR transformants; % NovR transformants multiplied by 5 to take into account the fact that transformation efficiency of the nov1 marker is ∼ fivefold lower than that of str41 because of Hex-dependent mismatch repair.
  4. cfu, colony-forming units.
AR2401a4.6 × 107nda0.15> 694b
b7.4 × 107nda0.50> 371b
R2403a2.2 × 1080.00100.29303
b2.9 × 1080.000710.10137
BR2588 (comD::kan105C)a5.7 × 1074.05.11.3
 b1.3 × 1084.25.71.3
R2631 (comD::spc105C)a3.3 × 1077.89.71.2
 b3.6 × 1074.75.21.1
CR2929 1.6 × 108c0.70dc
R2977 (ΔcomD) 4.7 × 107c5.5dc

The discovery of a direct interaction between ComD and ComED58E in yeast two-hybrid assays (Fig. S5) provided an explanation for these puzzling observations. It led us to hypothesize that ComED58E was sequestered by ComD and that a conformational change in ComD subsequent to its CSP-induced autophosphorylation could release ComED58E. To check this hypothesis, we investigated the effect of comD inactivation using both mariner transposon insertions (Experiment B in Table 2) or a complete deletion of the gene (Experiment C in Table 2). In the absence of ComD, ComED58E mutants displayed high spontaneous transformation frequencies, similar to those in fully induced wild-type cells, and CSP no longer stimulated transformation. High constitutive comX expression was observed in ΔcomD comED58E cells in the absence of CSP; comX expression was at least as high as in wild-type cells fully induced by CSP (Fig. 3B). After 40 min incubation, comX expression was also more than twofold higher in ΔcomD comED58E cells than in comED58E cells which had received CSP (Fig. 3B), while as expected comX expression remained very low in the latter without CSP (Fig. 1B, 0 min). In good agreement with transformation data (Table 2), Western blot experiments confirmed the presence in ΔcomD comED58E cells of ComE amounts similar to those in fully induced wild-type cells, whereas significantly lower amounts were detected in CSP-induced comD+ comED58E cells (data not shown).

We conclude that the effect of CSP in comD+ comED58E cells and the restoration of full competence in comED58E cells through comD inactivation are both reflections of the sequestration of ComED58E by ComD (see Discussion). In any case, the finding that comED58E cells exhibited wild-type transformation frequencies in the absence of both CSP and ComD demonstrated that the phosphorylmimetic protein triggers its own production and activates transcription of the early com promoters. Taken together, these observations provide strong though indirect support for the autokinase and phosphotransfer model of ComE activation.

Oligomerization potential of ComE proteins and in vitro binding to PcomCDE and PcomX

To document the binding of ComE to its promoter targets in vitro, the ComE, ComED58E, ComER120S, ComED58A and ComED58E-R120S proteins were purified (Experimental procedures; Fig. S7A and data not shown). First, their oligomerization status was investigated by gel filtration (Supporting experimental procedures; Fig. S6). All ComEs, except the phosphorylmimetic ComED58E protein, behaved as monomers in solution (Table 3). It is of note that the dimerization effect (in solution) of the comED58E mutation was abolished when combined with the comER120S mutation.

Figure 4.

EMSA of ComE and ComED58E binding to PcomCDE and PcomX promoters.

A. ComE binding to PcomCDE promoter. End-labelled promoter or control fragments were incubated with C-ter His-tagged proteins at concentrations indicated in the figure, as described in Experimental procedures. C1, Type 1 complex; C2, Type 2 complex; FP, free probe; no P, control fragment without promoter. See also Fig. S7.

B. ComED58E binding to PcomCDE promoter.

C. ComED58E binding to PcomX promoter.

Table 3. Oligomerization potential and promoter-binding affinities of ComE proteins
  1. aThe oligomeric state of purified ComE proteins in solution was determined by gel filtration (Fig. S6).
  2. bApproximate equilibrium dissociation constant was deduced from the protein concentration required to shift 50% of the corresponding labelled promoter fragment; average value (based on six experiments for ComED58E and ComED58A; five for ComE and ComER120S; and four for ComED58E-R120S) with standard error calculated from densitometer tracing of EMSA (e.g. Figs 4 and S7).
  3. cC1 and C2 retarded bands are defined in Fig. 4A, B and C for PcomCDE and PcomX promoter fragments respectively; C1 and C2 amounts are expressed as % of total retarded probe present in C1 + C2; average of values measured at protein concentrations flanking the apparent Kd (except for ComER120S and ComED58E-R120S binding to PcomX, see e and f); calculated from densitometer tracings of EMSA.
  4. dSee footnote b; average value based on four experiments for ComE; five for ComED58E, ComER120S and ComED58A; and three for ComED58E-R120S.
  5. e34.9 ± 17.3% of the labelled promoter fragment shifted at 1000 nM ComER120S protein (data not shown).
  6. f32.2 ± 2.9% of the labelled promoter fragment shifted at 1000 nM ComED58E-R120S protein (Fig. S7E).
  7. gObserved only with 700 nM or more ComER120S (data not shown).
  8. hOnly C2 complex observed with 100–400 nM ComED58E-R120S (Fig. S7E).
MW (KDa)a 34.6 ± 2.938.0 ± 8.059.7 ± 9.633.3 ± 4.434.6
PcomCDE promoterKdb (nM) 449 ± 35254 ± 12197 ± 86487 ± 6363 ± 14
Major complexc:
C10.75 ± 0.050.66 ± 0.09   
C2  0.64 ± 6.50.72 ± 0.061.00
PcomX promoterKdd (nM)916 ± 90906 ± 271881 ± 219(> 1000)e(> 1000)f
Major complexc:
C10.85 ± 0.190.58 ± 0.070.53 ± 0.03  
C2   (0.72 ± 0.12)g(1.00)h

The binding of purified ComEs to DNA fragments containing early com promoters was then analysed by EMSA as described in the Experimental procedures. Examples of binding to PcomCDE are shown for ComE (Fig. 4A), ComED58E (Fig. 4B), ComED58A (Fig. S7B), ComER120S (Fig. S7C) and ComED58E-R120S (Fig. S7D); and to PcomX for ComED58E (Fig. 4C) and ComED58E-R120S (Fig. S7E). A compilation of apparent Kds is presented in Table 3. These data established that the non-phosphorylable mutant ComED58A binds the PcomCDE promoter with a Kd similar to that of ComE and ComER120S. The phosphorylmimetic mutant ComED58E displayed a significant reduction (at least fourfold) of the apparent Kd for PcomCDE, including when combined with the comER120S mutation. This differential behaviour of ComED58E was fully consistent with a stimulation of binding by phosphorylation. On the other hand, no large differences were observed between the apparent Kd for PcomX among the various ComE proteins including ComED58E. Interestingly, all ComEs exhibited higher Kds for PcomX than for PcomCDE, an observation in good agreement with conclusions from our in vivo expression data.

The various ComE proteins did not only display differences in Kds but also in the type of retarded complexes that were formed (Fig. 4 and S7; Table 3). Thus, whatever the promoter fragment, the ComER120S protein formed mainly C2 complexes migrating more slowly than the major species (C1) observed with other ComEs (Figs 4 and S7). ComED58E also formed a majority of C2 complexes with PcomCDE (Fig. 4B). It is also of note that only C2 complexes were observed with both promoters upon incubation with the ComED58E-R120S double mutant, suggesting that each mutation contributed to the reinforcement of oligomerization (Table 3). While the phosphorylmimetic mutation results in the oligomerization of purified ComED58E in solution, presumably phosphorylation-induced oligomerization of ComE could increase its affinity for its promoters, as observed here with ComED58E and PcomCDE. It is also interesting that ComER120S exhibits DNA-dependent oligomerization. We propose that this feature accounts for its cup phenotype (see Discussion). Altogether, these data are consistent with a crucial role of ComE oligomerization for transcriptional activation of the early com genes.

Surge then intrinsic subsidence of PcomCDE transcription

In all experiments investigating the kinetics of com gene expression, apparent rates of transcription peaked about 9 min after CSP addition, then decreased despite the presence of high concentrations of CSP in the medium (Figs 1 and 2; Table 1). We wished to establish whether the subsidence in transcription of the early com genes resulted from σX competing with σA. comCDE expression was therefore compared in wild-type and comX cells (Fig. 5A). Calculation of apparent transcription rates revealed that subsidence of PcomCDE transcription occurred similarly in wild-type and comX mutant cells; while transcription rate peaked 8.5 min after CSP addition, it decreased by 3.5-fold within the next 5 min (Fig. 5B). This experiment ruled out a role for σX in the deceleration of PcomCDE transcription and established that this deceleration occurred independently of any late com gene product.

Figure 5.

Slowing down of PcomCDE transcription occurs independently of late com gene expression while ComD–ComE cellular levels remain constant.

A. Kinetics of CSP-induced PcomC transcription in comX mutant cells. Expression of a comC::luc fusion was monitored in ΔcomC comX+ (R1960) and ΔcomC comX (R2002) cells incubated at 37°C in C + Y medium. CSP was added at 0 min.

B. Apparent rate of transcription from PcomC. Calculated as described in the legend of Fig. 1 using data in (A).

C. Stability of ComD and ComE proteins. The levels of ComD and ComE were analysed with a BioImager (LAS-4000, Fujifilm) after Western blot using extracts from R1205 cells induced with CSP (4 ng ml−1) and harvested at the indicated time post-CSP addition.

Western blot experiments clearly indicated that ComD and ComE levels remained essentially constant while transcription decelerated, which ruled out TCS-targeted proteolysis as a possible explanation (Fig. 5C). We tentatively concluded that a mechanism intrinsic to ComD–ComE slowed down PcomCDE expression. In light of the above-documented binding of non-phosphorylated ComEs, particularly ComED58A, to the PcomCDE promoter we hypothesized that accumulation of non-phosphorylated ComE antagonizing the binding of ComE∼P to PcomCDE could be responsible for the observed subsidence of comCDE transcription.

ComE antagonizes comCDE expression, while ComD is limiting

The binding of the non-phosphorylable ComED58A mutant to PcomCDE and PcomX promoters (Table 3) together with the failure to detect phosphorylation of ComE by acetylphosphate strongly suggested that present and previous in vitro ComE-binding experiments (Ween et al., 1999) could no longer be considered as supporting a direct transcriptional activation of PcomCDE by ComE∼P. Instead, these results establish the capacity of non-phosphorylated ComE to bind its targets, a capacity which would be consistent with an antagonistic role of the RR on early com gene expression. To document this role in vivo, an ectopic copy of comE or comD was placed at the chromosomal expression platform, CEP (Guiral et al., 2006b), under the control of a maltose-inducible promoter (Experimental procedures). It was previously observed that maltose-driven expression of both comD and comE, resulting in ComE cellular amounts slightly lower than in CSP-induced wild-type cells, impeded spontaneous as well as CSP-induced competence (Guiral et al., 2006a). Maltose-induced expression of comE alone was sufficient to inhibit competence (Fig. 6A) providing support to the hypothesis of an antagonistic role of ComE on competence. In contrast, maltose-induced comD expression stimulated spontaneous competence development (Fig. 6B), suggesting that ComD is limiting in wild-type cells.

Figure 6.

ComE antagonizes comCDE expression, whereas ComD is limiting.

A. Impact of comE overexpression on spontaneous competence development. Competence and growth of cells harbouring the ssbB::luc fusion and the CEPM-comE construct (R1565; competence, blue circles; OD, open circles) or an empty CEP (R1567; competence, grey triangles; OD, open triangles) was monitored during growth at 37°C. Cultures were inoculated at OD = 0.004 in C + Y medium containing 0.3% sucrose (upper panel), 0.2% glucose (centre panel) or 0.3% maltose (lower panel), which results in full competence repression, intermediate induction and full induction of PM respectively (indicated as Low, Intermediate and High [ComE]).

B. Overexpression of comD upregulates competence revealing that ComD is limiting. Cells harbouring the CEPM-comD construct (R1563, green symbols) were grown in parallel with cells harbouring an empty CEP (R1567). Same experimental design as in panel A.

C. Impact of late comE and comED58A expression on PcomCDE and PcomX promoter activity. Expression of a comC::luc (top) and a comX1::luc (bottom) fusion was monitored at 37°C in ΔcomC cells harbouring CEPX-comE (red symbols; top and bottom panel strains respectively R2007 and R2236), CEPX-comED58A (green symbols; top and bottom panel strains respectively R2006 and R2235) or no CEP (open symbols; top and bottom panel strains, R1960 and R2218 respectively). CSP was added at 0 min. Expression profiles essentially identical to those with comX1::luc strains were obtained with comX2::luc strains (R2200, no ectopic comE; R2239, CEPX-comE; R2238, CEPX-comED58A; data not shown).

To further document the effect of non-phosphorylated ComE, we investigated the impact of the late expression of comE and comED58A (i.e. expression driven by a ComX-dependent promoter) on PcomCDE and PcomX promoter activity in otherwise comE+ cells (Fig. 6C). Consistent with our hypothesis, late extra-synthesis of both ComE and the non-phosphorylable variant ComED58A accelerated the blockage of PcomCDE transcription. In contrast, extra-synthesis of ComED58A had the same inhibitory effect on PcomX transcription, whereas extra-synthesis of ComE resulted in a significant increase in PcomX transcription (near doubling of the time during which luciferase accumulated linearly, leading to about 35% overall increase in luciferase activity). We conclude that freshly produced ComE∼P still activates transcription from PcomX at a time when PcomCDE can no longer be activated, implying that PcomX is less sensitive to extinction by ComE than PcomCDE.


Competence activation by ComE∼P

The original observation that the RR ComE, which constitutes together with the HK ComD and the comC-encoded CSP the master switch for induction of pneumococcal X-state (competence), could specifically bind to the comCDE promoter region in vitro was taken as strong support for a direct transcriptional activation of PcomCDE by ComE∼P produced through transphosphorylation by ComD, in response to CSP (Ween et al., 1999). However, present data establish that in vitro binding to PcomCDE is not in itself sufficient to demonstrate the transcriptional activation of this promoter by ComE∼P (see below). With the aim of providing support to the direct activation model, we investigated the ComD autokinase and ComD-to-ComE phosphotransfer reactions. While autophosphorylation of the cytoplasmic domain of ComD proteins could be detected, transphosphorylation of ComE proteins failed (Fig. S3). Whatever the explanation for this failure (see Supporting results and discussion), it led us to construct a phosphorylmimetic mutant, ComED58E. Results of in vivo and in vitro analyses with this mutant strongly support the idea of a canonical ComD kinase-dependent mechanism of activation of ComE triggering pneumococcal competence. Thus, comED58E cells exhibited near wild-type transformation frequencies in the absence of both CSP and ComD (Table 2) indicating that the phosphorylmimetic protein activates its own production. Interestingly, the phosphorylmimetic ComED58E displayed a significantly increased affinity for the PcomCDE promoter in vitro (Table 3), providing indirect support to the model of ComE∼P-dependent activation of transcription from this promoter. This model is also fully consistent with the location of the 5′ extremity of CSP-induced comCDE transcripts (Halfmann et al., 2007) with respect to CEbs and the abolition of basal as well as CSP-induced ComE-dependent transcription by a mutation of the presumptive PcomCDE −10 (Martin et al., 2010).

Biological significance of the binding of ComD to ComED58E

Interestingly, the presence of ComD appeared to reduce spontaneous competence of comED58E cells (Table 2) suggesting that ComD had some sort of activity against ComED58E. This is reminiscent of a previous observation with the essential VicRK (also named WalRK) TCS of S. pneumoniae. Cells containing both the phosphorylmimetic vicRD52E RR mutation and a deletion of vicK grew markedly better than the single vicRD52E mutant, leading to the suggestion that the VicK kinase either phosphorylated VicRD52E (necessarily on residue(s) other than the blocked canonical phosphate acceptor site) to further decrease its activity or bound and sequestered the phosphorylmimetic mutant (Ng et al., 2003). The former hypothesis would not account for our observations with ComED58E as the mutant protein could not activate competence in the absence of CSP, i.e. in a situation in which ComD is not expected to display any kinase activity. On the other hand, the hypothesis of a sequestration of the phosphorylmimetic RR mutant by its cognate HK readily accounts for the data. In full agreement with this hypothesis, yeast two-hybrid assays established that ComD interacts with ComED58E but not with other ComE proteins including wild-type ComE (Fig. S5). We propose that ComD sequesters ComED58E thus preventing transcriptional activation of early com promoters.

What could be the significance of these parallel observations? It is possible that this binding reflects a functional interaction during phosphotransfer. It is of note that the addition of CSP was observed to reduce the inhibitory effect of ComD on ComED58E resulting in a 137–694-fold increase in transformation frequency (Table 2). We suggest that CSP-induced autophosphorylation of ComD triggers a conformational change leading to the release of ComED58E. Release of ComE∼P from ComD could be required to allow activation of early com promoters. Alternatively, the preferential binding of ComED58E by ComD could reflect a functional interaction between non-phosphorylated HK and its phosphorylated RR necessary for HK-dependent dephosphorylation of the latter in vivo.

ComE oligomerization, DNA binding and the activation of early com promoters

In contrast to wild-type ComE, ComED58E formed dimers in solution independently of DNA (Fig. S6) and exhibited an increased affinity for its target promoters (Table 3). For this member of the LytTR family, this suggests a mechanistic model similar to that of the OmpR/PhoB family (Gao and Stock, 2010) in which phosphorylation mediates dimerization of the receiver domain, which then promotes transcriptional activation (Fig. 7A). Phosphorylation was recently reported to induce oligomerization of the ComE paralogue of Streptococcus mutans but was concluded not to affect DNA-binding affinity (Hung et al., 2012), in contrast to our findings with the S. pneumoniae phopshorylmimetic ComED58E. The hypothesis that oligomerization of pneumococcal ComE plays a crucial role in the transcriptional activation of early com promoters is further supported by the finding that the cup ComER120S protein, which is monomeric in solution (Fig. S6C and Table 3), formed oligomers upon binding to its DNA targets in vitro (Fig. S7C).

Figure 7.

ComD, ComE and the control of pneumococcal X-state.

A. Transcriptional control of the early PcomCDE and PcomX promoters. Diagrammatic representation of the differential extinction and activation of PcomCDE and PcomX by ComE and ComE∼P respectively. Presumably, the receiver domain (black or green part) is tethered to the DNA-binding domain (grey triangle) through a flexible linker (black line) and does not impose structural constraints on the DNA-binding domain. Following transphosphorylation by ComD, dimerization of ComE∼P (green receiver domain), as suggested by in vitro data, allows binding of the two DRs (black arrows) in front of comCDE with high affinity, whereas a lower affinity of ComE∼P for PcomX results in a weaker binding. Similarly, the reduced affinity of ComE for PcomX is illustrated by the binding of a monomer to LR only, whereas ComE monomers are drawn as interacting with each repeat at PcomCDE.

B. The X-state regulatory cascade. (1) to (5) refers to features directly or indirectly documented during this work (see Discussion) as follows. (1) ComE-dependent extinction of competence through binding to PcomCDE and PcomX promoters (see panel A). (2) Autophosphorylation of ComD. (3) Transphosphorylation of ComE by ComD∼P. (4) Differential transcriptional activation of PcomCDE and PcomX promoters by ComE∼P (see panel A). (5) ComE dephosphorylation. In features (1) and (4), the thickness of the arrows represents strong (thick) or weak (thin) binding of ComE/ComE∼P to the indicated promoter regions.

Molecular basis for the cup phenotype of ComER120S and ComDT233I mutants

Interestingly, the ComER120S mutation, referred to as cup10, conferred a cup phenotype only in the presence of CSP (Martin et al., 2000). This requirement for CSP as well as for ComD (Martin et al., 2000) strongly suggested that ComER120S required phosphotransfer from ComD to activate comCDE expression. However, in contrast to wild-type ComE, ComER120S exhibited DNA-dependent oligomerization (Fig. S7C). The R120S mutation affects a residue located in the middle of α5 in the predicted structure of ComE (Fig. S8). This α-helix belongs to the α4-β5-α5 face of the receiver domain, which contains residues highly conserved in the OmpR/PhoB family. This face is the site of the greatest differences between the inactive and active conformations, and likely plays a key role in RR dimerization (Gao and Stock, 2010). Despite limited residue conservation, the predicted secondary structure of S. pneumoniae ComE also contains an α4-β5-α5 arrangement (Fig. S8). Interestingly, R120 is a conserved residue in the middle of α5 and could exchange intermolecular salt bridges with another residue in the opposite partner, D106, also conserved at the appropriate position in β5 (Fig. S8), as documented in the case of WalR (also named MicA) and ArcA (Toro-Roman et al., 2005). Thus, although ComE belongs to a different RR subfamily, its receiver domain could form a dimer with a twofold rotational symmetry with the entire α4-β5-α5 face participating in contacts with the opposite partner, as established for the OmpR/PhoB subfamily (Gao and Stock, 2010). Intriguingly, however, the very same substitution in Escherichia coli OmpR (OmpRR115S) was reported to result in an oligomerization defect and to severely affect in vitro DNA binding while phosphorylation and dephosphorylation remained unchanged (Nakashima et al., 1991). As a consequence, ompRR115S mutant cells were no longer able to respond to change in medium osmolarity. The opposite effects of the very same substitution in ComE and OmpR could be due to the fact that the C-terminal DNA-binding domain of ComE need to dimerize with translational symmetry, to fit the symmetry of the DR DNA binding sites. However that may be, we propose that DNA-facilitated formation or stabilization of ComER120S oligomers accounts for its previously documented cup phenotype. Possibly, ComER120S multimerization reduces the rate of dephosphorylation, which in turn alters the transcriptional shutoff of early com promoters, particularly PcomX (Fig. 2).

The present work allows us to also propose a molecular explanation for the effect of another previously isolated cup mutation, comDT233I (Martin et al., 2000). In contrast to comER120S, this mutation described as cup3 resulted in near wild-type transformation levels in the absence of CSP. We initially envisioned two possible effects of the mutation, on ComD itself (i.e. affecting autophosphorylation) or on ComE (i.e. affecting phosphotransfer and/or RR dephosphorylation). Exacerbation of autokinase activity in vitro (Fig. S4) strongly suggests an effect on autophosphorylation. Examination of the predicted structure of the C-terminal domain of ComD indicated that the altered residue, T233, is positioned in a 54-residue α-helix, α1, which is part of the conserved DHp (dimerization and histidine phosphotransfer) cytoplasmic domain and contains the phospho-accepting His (H248 in ComD) (Gao and Stock, 2009; Stewart, 2010). T233 is located in a region believed to be important for transmission of the activation signal from the sensory input domain to the DHp domain and belongs to a series of residues predicted to have high propensity for coiled-coil transition (see Fig. S1 in Thoendel et al., 2011). In view of the activation of competence in comDT233I cells in the absence of CSP (i.e. in the absence of signal), we propose that the mutation locks α1 in an active (signal ON) conformation and favours the association between the catalytic and ATP-binding domain and the DHp domain (Stewart, 2010) thus allowing transfer of the phosphoryl group from ATP to the phospho-accepting His248.

Differential activation of early com promoters by ComE∼P

The hierarchical expression of early com promoters (Fig. 1B) presumably depends on the sequence of their DRs. Thus, a lower affinity of ComE∼P for PcomX than for PcomCDE was first suggested by a ∼ threefold lower transcription rate for the former (Table 1). The difference between the two promoters was even larger in ComED58E cells (Fig. 3A). In addition, an apparent Kd ninefold lower of ComED58E for PcomCDE than for PcomX was measured through EMSA (Fig. 4 and Table 3). We suggest that the reduced affinity of ComED58E for PcomX is a direct consequence of the poor match of PcomX RR to the consensus (Fig. 1A) affecting ComED58E binding to this site (Fig. 7A).

Similarly, expression of the comAB operon was previously shown to be limiting for competence development (Martin et al., 2000), despite the presence of a BOX element (Martin et al., 1992) inserted in between the −10 and the start of comA translation (Fig. 1A) enhancing expression of the two genes (Knutsen et al., 2006). Examination of the PcomAB promoter indicates that a conserved A residue is missing at the first position of the RR repeat (Fig. 1A). This suggests that this residue is important for promoter recognition, binding and activation by ComE∼P.

A third example is provided by the PcomW promoter validated in this study (Fig. S1). Although PcomW was not investigated in great detail, the lower transcription rate of PcomW compared with PcomCDE (Fig. 1B) indicates it is a weak early com promoter. We suggest that the presence of a degenerate LR (Fig. 1A) limits comW expression. As the ComW protein is crucial for σX stability (Luo and Morrison, 2003; Luo et al., 2004) and activation (Sung and Morrison, 2005), comW expression directly impacts the amount of active σX. The weak activity of both PcomX and PcomW promoters can thus be regarded as a first-level control of late com gene expression in pneumococcus (Fig. 7B).

A mechanism intrinsic to ComDE slows down PcomCDE transcription

It was previously concluded that comAB and comCDE were not sufficient for the shutoff of CSP-induced gene expression nor for the subsequent refractory period, and that these phenomena depended directly or indirectly on ComX (Lee and Morrison, 1999). However, our data clearly show that the shut-off of comCDE expression readily takes place in comX mutant cells (Fig. 5). Thus, the apparent transcription rate from PcomCDE peaked 8 min after CSP addition, then decreased ∼ 3.5-fold over the next 4.5 min in the absence of any late com gene (Fig. 5B). This deceleration could be due to the activity of a non-induced protein, of an early gene product or be intrinsic to the ComDE TCS. In light of a previous observation that increased expression of both comD and comE impeded spontaneous and CSP-induced competence (Guiral et al., 2006a), we favour the latter interpretation.

Let us first consider the hypothesis that ComD exhibits strong phosphatase activity, turning down transcription of the early com genes through dephosphorylation of ComE∼P. ComD accumulation and stability (Fig. 4C) would be fully compatible with this hypothesis. Then, however, overexpression of ComD would be expected to antagonize competence. The opposite observation (Fig. 6B) led us to conclude that ComD did not play a major role in the rapid reversal of the response to CSP. In contrast, overexpression of ComE antagonized spontaneous competence (Fig. 6A). We therefore propose that while ComE∼P activates comCDE transcription, ComE antagonizes it and is directly responsible for subsidence of PcomCDE transcription (Fig. 7). It is of note that ComE should not be regarded as a ‘repressor’ sensus stricto, since we previously showed that comCDE expression is very low in cells lacking ComE (Martin et al., 2010), indicating that there is no strong σ70-dependent promoter that would be ‘repressed’ by ComE.

PcomCDE transcription extinction by non-phosphorylated ComE

Several lines of evidence support our proposal of an active role of non-phosphorylated ComE in the rapid reversal of the response to CSP. First, the finding that ComED58A protein, which harbours a mutation blocking the phosphate-acceptor site of ComE, binds in vitro a PcomCDE promoter fragment at least as efficiently as wild-type ComE (Table 3) establishes that the ComE DNA recognition (LytTR type) domain is readily available for binding, in contrast to other RRs in which the corresponding region of the effector domain is occluded by the unphosphorylated regulatory domain (Gao and Stock, 2009). This observation removes any ambiguity as concerns the ability of non-phosphorylated ComE to bind early com promoters. Second, the rapid accumulation of ComE, reaching amazingly high amounts (Fig. S4), and ComE intrinsic stability (Fig. 5; Ween et al., 1999) would readily compensate for the observed difference in Kd betweeen ComE and the phosphorylmimetic ComED58E (Table 3), allowing ComE to outcompete ComE∼P for binding to early com promoters. Third, the acceleration the PcomCDE transcription blockage by ectopic expression of comE or comED58A as a late com gene (Fig. 6C) is fully consistent with an antagonistic role of ComE.

We propose that once the response to CSP is initiated, either spontaneously or after CSP addition, the progressive accumulation of ComE through neosynthesis (as well as presumably through dephosphorylation of ComE∼P) antagonizes ComE∼P, resulting in subsidence of PcomCDE transcription. It is of note that any mechanism modifying the equilibrium between ComD kinase and phosphatase activities in favour of the latter would further reduce ComE∼P levels and contribute to the shutoff of transcription.

Differential extinction of PcomCDE and PcomX promoters by ComE

Converging evidence indicates a differential antagonization of PcomX and PcomCDE transcription by non-phosphorylated ComE. Thus, additional expression of comE (but not comED58A) as a late com gene prolonged only PcomX transcription (Fig. 6C) suggesting that this promoter could still be activated (presumably by ComE∼P) whereas PcomCDE was already blocked (by ComE). Similarly, the observation that the comER120S mutation led preferentially to overexpression of PcomX (Fig. 2) could be explained by a lower affinity of non-phosphorylated ComER120S for PcomX than PcomCDE resulting in reduced antagonization of the former. Finally, EMSA revealed a ∼ two-, > two- and ∼ fourfold lower apparent Kd of respectively ComE, ComER120S and ComED58A proteins for PcomCDE than for PcomX (Table 3). We attribute the reduced affinities to the presence of a degenerate RR in PcomX and propose that ComE binds as a monomer (Fig. 7A). C1 complexes as observed in Figs 4 and S7 may thus correspond to the binding of a monomer, although we cannot exclude that C1 and C2 complexes both involve dimers but with different topologies. It is of note that it has been hypothesized that AgrA could bind its target as a monomer (Koenig et al., 2004) and interestingly the structure of the LytTR DNA-binding domain of S. aureus AgrA was obtained through crystallization of a monomer bound to a single 9 bp repeat (Sidote et al., 2008), unambiguously establishing the binding of this RR as a monomer.

A previous study suggested that ComE could recognize secondary sites in the comCDE region. Thus, recovery of fragments bound by ComE through in vitro binding of biotinylated ComE C-terminal domain to a library of pneumococcal DNA fragments led to the identification of a fragment internal to the comCDE operon, in addition to the PcomCDE fragment (Zahner et al., 2002). We therefore looked for the presence of degenerate CEbs in that region but could find none. Then, in line with the hypothesis of ComE binding as a monomer, we searched for the presence of only one of the two repeats. While nothing striking appeared in the fragment identified by Zahner and coworkers, our scan identified four repeats sharing 7/9 nt with the consensus LR or RR within a 260 bp region overlapping the PcomCDE promoter. In contrast, flanking segments of the same size on the right and left sides contained respectively 1 and 0 degenerate repeat (data not shown). This puzzling accumulation of degenerate repeats ∼ 100 bp away from the PcomCDE CEbs could result in the formation of complex ComE oligomeric structures contributing to the reinforcement of PcomCDE sequestration. Further experiments are required to establish whether this accumulation of single repeats is biologically relevant.

Concluding remarks

A number of TCS positively autoregulate their own expression like ComDE. These include AgrCA of staphylococci, PlnBC of L. plantarum, the vancomycin-responding VanRS of Streptomyces coelicolor (Hutchings et al., 2006) and the PhoQP regulator of Salmonella pathogenicity (Shin et al., 2006). These autoregulated systems presumably share a similar induction pattern with an initial surge in phosphorylation and transcription, followed by subsidence to new steady state levels of expression (Shin et al., 2006). However, they probably diverge as concerns the mechanisms of subsidence. It was suggested that the PhoQ protein governs the levels of phospho-PhoP and that there is a ‘temporal change’ in the kinase and phosphatase activities of PhoQ after activation of this TCS (Shin et al., 2006). Similarly, VanS control of VanR∼P levels appeared important for regulation of the van genes (Hutchings et al., 2006). ComDE of S. pneumoniae illustrates a different strategy. Although a ComD-dependent control of ComE phosphorylation is by no means excluded, presumably antagonization of ComE∼P binding by accumulating non-phosphorylated ComE plays a prominent role in the deceleration of transcription. This antagonistic role of ComE could in part account for the unexpectedly high amounts of ComE synthesized in response to CSP.

Finally, an interesting parallel can be made with the Pln system of L. plantarum, in which the PlnA peptide pheromone activates transcription of a regulatory unit, plnABCD, including two antagonizing RRs, PlnC and PlnD (Diep et al., 2003). While the activity of both RRs is significantly enhanced upon phosphorylation, the latter acts as a negative regulator, which would readily account for transcription deceleration. The dual role we propose for ComE, which combines both response activation and extinction abilities within a single protein, depending on its phosphorylation status, constitutes an alternative economical regulatory design.

Experimental procedures

Bacterial strains, culture and transformation conditions

All the strains and plasmids constructed in this work are listed, together with primers, in Table S1. Stock cultures were routinely grown at 37°C in Todd-Hewitt medium +0.5% Yeast extract (THY) or C + Y medium to OD550 = 0.4; after addition of 15% glycerol, stocks were kept frozen at −70°C. For the study of spontaneous or CSP-induced competence, cells were incubated in C + Y medium (except in Fig. 1 experiment). Unless otherwise indicated, fresh cultures were first grown to OD = 0.2–0.3, then diluted to OD550 = 0.04–0.07 (depending on the experiment) and synthetic CSP1 (100 ng ml−1, unless otherwise indicated) was added after 10 or 12 min incubation depending on the experiment.

For chromosomal transformation, DNA was added 10 min after CSP and cells were incubated for 20 min at 30°C. Transformants were selected by plating on CAT-agar supplemented with 4% horse blood, followed by selection using a 10 ml overlay containing chloramphenicol (Cm; 4.5 μg ml−1), erythromycin Ery; (0.05 μg ml−1), kanamycin (Kan; 250 μg ml−1), spectinomycin (Spc; 100 μg ml−1) or streptomycin (Sm; 200 μg ml−1), after phenotypic expression for 120 min at 37°C.

Construction and use of com::luc transcriptional fusions

The monitoring of com gene expression using transcriptional fusions with the luc gene involved the cloning of an S. pneumoniae DNA fragment upstream of the luc gene, followed by homology-dependent integration of the non-replicative recombinant plasmid (pR424 derivative) into the pneumococcal chromosome, with selection for chloramphenicol resistance. At each locus, the luc gene was inserted at a very similar position, next to the 13th or 14th codon of the targeted com gene and luc translation relied on the same SD (Fig. S2). It is of note that insertion-duplication of the recombinant plasmid placing luc under the control of the com promoter maintains a fully functional copy of the com gene downstream of the integrated plasmid (Fig. S2).

Construction of comX::luc and comW::luc fusions

To generate the comX1::luc and comX2::luc reporter plasmids pR473 and pR474, respectively comX1fusF-comXfusR (473 bp) and comX2fusF-comXfusR (401 bp) PCR fragments were digested with HindIII and BamHI and ligated to HindIII-BamHI-digested pR424. comW::luc reporter plasmids pR476 and pR484 were similarly generated by ligating, respectively, MP208comW-MP210comW (501 bp) and MP211comW-MP210comW (202 bp) PCR fragments digested with HindIII and BamHI and ligated to HindIII-BamHI-digested pR424.

For the monitoring of luc expression (and growth where indicated), light emission, which results from luciferase activity, was directly measured in cultures of pneumococci actively growing in luciferin-containing medium as previously described (Prudhomme and Claverys, 2007). RLU (relative luminescence unit) (and OD) values were recorded throughout incubation at 37°C into a 96-well white microplate with clear bottom in a LucyI luminometer (Anthos) or a Varioskan Flash luminometer (Thermo 399 Electron Corporation). RLU should therefore be compared within but not between figures.

Ectopic expression of wild-type and mutant ComE proteins, and of ComD

Construction of CEPM-comD and CEPM-comE

Chromosomal ectopic maltose-driven expression of comD and comE was achieved through amplification of the genes with primer pairs comDNco-BM92 and comENco-BM93, using as template R800 chromosomal DNA. After digestion by NcoI/BamHI, comD and comE fragments were ligated to NcoI/BamHI-digested pCEP plasmid (Guiral et al., 2006b). Transformation of strain R1556 with the ligation mixture and selection for KanR transformants generated strains R1563 (CEPM-comD) and R1565 (CEPM-comE).

Construction of CEPX-comED58A and CEPX-comE

pR454, which carries the CEPX-comED58A construct, was generated by ligation of NcoI-HindIII-digested pCEPcin and pR447 fragments. pR455, which carries the CEPX-comE construct, was generated by ligation of NcoI-ClaI-digested pR454 (9969 bp) and pR447 (177 bp) fragments.

Mutations of comE

Site-directed mutagenesis of comE involved the use of the PCR-based gene splicing by overlap extension (gene SOEing) method (Horton et al., 1990). Briefly, a couple of PCR reactions (with primer pairs A1–A2 and B2–B1) are used to generate two fragments A and B that incorporated a mutant primer (A2) at one extremity of A and its complement (B2) at the other extremity of B. A third PCR reaction with primer pair A1–B1 then produces a unique fragment with the mutant sequence in the middle. Primer pairs A1–A2, B2–B1 and A1–B1 to generate comED58E were pQEfor-comED2 (template for PCR, pR438; 254 bp fragment), comED1-pQErev (template for PCR, pR438; 664 bp fragment) and pQEfor-pQErev (888 bp fragment generated).

To construct a comED58E strain, the pQEfor-pQErev PCR fragment was digested with NcoI and BglII generating a 752 bp fragment which was ligated to NcoI-BglII-digested pQE-60 plasmid DNA. The comE region in the resulting plasmid, pR486, was sequenced to rule out the presence of unwanted mutation(s). Transformation of strain R1501 (ΔcomC) with pR486 plasmid DNA was followed by phenotypic expression and segregation of transformed cells in liquid culture, and plating on CAT-agar without selection. As the GAT to GAG mutation corresponding to the D58E change inactivated an EcoRV site (GATATC) present in wild-type comE, individual colonies were picked up, the corresponding comE region was amplified and digested with EcoRV to identify comED58E transformants (strains R2401 and R2402).

To transfer the comER120S allele, the mutation was amplified from strain R682 with the VH5-BM93 primer pair. After NcoI-BglII digestion, this fragment was cloned into NcoI-BglII-digested pQE-60 generating plasmid pR437. The comE region in pR437 was sequenced to rule out the presence of unwanted mutation(s) and plasmid DNA was then used as donor in transformation of strain R1501. Following phenotypic expression, segregation of transformed cells in liquid culture and plating on CAT-agar without selection, individual colonies were picked up and the comE region was amplified and sequenced to identify a comER120S transformant (strain R2199).

Construction of a ΔcomC ΔcomD comED58E strain

A deletion of comD in otherwise comED58E cells was generated by taking advantage of strain R2957 in which comC was substituted with Janus. Janus is a kan-rpsL+ cassette that confers KanR and dominant sensitivity to streptomycin in an SmR background; its replacement by an arbitrary segment of DNA during a second transformation restores SmR (and sensitivity to Kan) (Sung et al., 2001). To delete comD (and comC), comDdelta1-BM113L (comD upstream region; 705 bp) and comDdelta3-comDdelta5 (comD downstream region; 794 bp) PCR fragments were first generated by using respectively R800 and R2401 (i.e. comED58E template) chromosomal DNA as template. The two PCR fragments were then mixed and used as template to generate a unique comDdelta1-comDdelta5 1454 bp fragment (lacking comCD), which was then BamHI-EcoRI-digested and ligated to BamHI-EcoRI-digested pGBDU-C1 generating plasmid pR499. After transformation of strain R2957 with pR499 and selection of SmR transformants, a control PCR with primer pair BM47–BM123 was used to confirm comD deletion and strain R2977 was retained after sequencing the relevant chromosomal region.

Cloning, expression and purification of soluble ComE and ComD proteins

To construct expression plasmids combining two comE mutations, a 1784 bp SpeI-XmaI fragment from pKHS-comER120S was subcloned into a 4266 bp SpeI-XmaI fragment from pKHS-comED58E (Table S1) thus generating pKHS-comED58E-R120S. The E. coli BL21-Gold(DE3) strain (Stratagen) was then successively transformed with a pKHS plasmid containing a comE insert and pGKJE3. The latter allows expression of chaperones to prevent ComE aggregation.

Purification of soluble ComE proteins was then carried out as described (Boudes, 2011). Briefly, E. coli cells were grown at 37°C in 800 ml of 2× YT medium (BIO101) up to an OD600 of ∼ 0.3 and then shifted at 15°C. At an OD600 of ∼ 0.6, chaperones expression from pGKJE3 was induced by addition of l-arabinose (0.2% final concentration). At an OD600 of ∼ 0.8, ComE expression was induced with 0.5 mM IPTG (Sigma) and cells were grown ∼ 16–20 h. Cells were harvested by centrifugation, resuspended in 40 ml of buffer E (50 mM MES pH 6.5, 0.5 M NaCl, 5 mM β-Mercaptoethanol, 5% glycerol). Cell lysis was completed by sonication (with a Branson probe-tip sonicator). Soluble His-tagged ComE proteins were retained on a Ni-NTA column (Qiagen), eluted with the same buffer complemented with imidazole (200 mM), concentrated on a centrifugal concentrator Vivaspin 5000 (Vivascience) and loaded onto a Superdex Hiload 16/60 column (Amersham Pharmacia Biotech) equilibrated with buffer L. ComE proteins eluted as a single homogeneous species from the gel filtration column. Fractions containing ComE were pooled and dialysed against buffer L supplemented with 50% glycerol for 5 h and then stored at −20°C.

To express and purify the cytosoluble autophosphorylation domain of ComD under the control of the IPTG-inducible (T5 promoter/lac operator) promoter, a 722 bp DNA fragment containing residues 208–441 of ComD was amplified from strain R800 with the VH8–VH9 primer pair. After NcoI-BglII digestion, this fragment was cloned into NcoI-BglII-digested pQE-60 generating plasmid pR440. The E. coli XL1-Blue strain (Stratagene) was transformed with this plasmid. Cells were grown at 37°C in 1 l of Luria–Bertani medium (BIO101) up to OD600 of ∼ 0.3 and then shifted at 30°C. At OD600 of ∼ 0.8, ComD expression was induced with 1 mM IPTG (Sigma) and cells were grown ∼ 2 h. Cells were harvested by centrifugation, resuspended in 20 ml of buffer D (50 mM Tris-HCl pH 7.5, 0.2 M KCl, 10 mM MgCl2, 0.1 mM EDTA, 1 mM DTT, 0.05% Triton, 5% glycerol). Cell lysis was completed by sonication (with a Branson probe-tip sonicator). Soluble His-tagged ComD was retained on a Ni-NTA column (Qiagen), eluted with the same buffer complemented with imidazole (200 mM), concentrated on a centrifugal concentrator Vivaspin 5000 (Vivascience) and loaded onto a Superdex Hiload 16/600 column (Amersham Pharmacia Biotech) equilibrated with buffer L. ComD eluted as a single homogeneous species from the gel filtration column. Fractions containing ComD were pooled and dialysed against buffer D supplemented with 50% glycerol for 5 h and then stored at −20°C.


The upstream regulatory regions of comCDE (255 bp) and comX (248 bp) as well as control (i.e. fragment devoid of CEbs, 250 bp) were amplified with primer pairs Oligo-LE/Oligo-RE, Oligo-LX/Oligo-RX and BM57/Oligo-R0 (Table S1) respectively. PCR fragments were end-labelled with [γ-33P]-ATP (GE Healthcare) using T4 polynucleotide kinase (NE Biolabs) according to standard procedures (Sambrook et al., 1989). The PCR fragments were purified on MicroSpin G-50 columns (GE Healthcare). DNA-binding assays were carried out in a total volume of 20 μl containing 50 mM NaCl, 50 mM Tris/HCl pH 7.5, 5% (v/v) glycerol, ∼ 5 nM end-labelled PCR fragments, 1 mM MgCl2, 0.15 μg Poly(dI-dC) (as non-specific competitor), and increasing concentrations of ComEHis proteins. Reaction assays were incubated at 37°C for 30 min. Complexes were resolved by electrophoresis in native TBE polyacrylamide gels [5% (w/v)]. After 110 min of electrophoresis at constant voltage (20 V cm−1) in 0.5× TBE buffer, gels were dried at 80°C for 30 min and exposed to a PhosphorImager screen (Life Science Systems – Fujifilm Global). Quantification of free DNA and protein-bound DNA was performed using Multi Gauge 3.0 Analysis Software (Life Science Systems-Fujifilm Global).


We thank Nathalie Campo and Calum Johnston for editing this manuscript. We are grateful to Marion Boudes and Sophie Quevillon-Cheruel for sharing ComE plasmids (pKHS-comE, pKHS-comED58A, pKHS-comED58E and pKHS-comER120S) and protocol for ComEHis protein purification with us. We thank Donald Morrison for providing us with strains CP1250 and CPM11. This work was supported in part by grants from the Ministère Délégué à la Recherche et aux Nouvelles Technologies (Programme Microbiologie 2003-2004; ref: RB/CD/2003/09/001; to J.-P. C. and P. N.) and from the Agence Nationale de la Recherche (Projet 2010 Blanc SVSE 3: PneumocoX; to J.-P. C. and P. P.). N. M. was the recipient of a PhD thesis fellowship accompanying the Programme Microbiologie 2003-2004 grant (September 2003–September 2006) and from the Association pour la Recherche sur le Cancer (2007).