Bacteria respond to nutritional stresses by producing an intracellular alarmone, guanosine 5′-(tri)diphosphate, 3′-diphosphate [(p)ppGpp], which triggers the stringent response resulting in growth arrest and expression of resistance genes. In Escherichia coli, upon fatty acid or carbon starvation, SpoT enzyme activity switches from (p)ppGpp degradation to (p)ppGpp synthesis, but the signal and mechanism for this response remain totally unknown. Here, we characterize for the first time a physical interaction between SpoT and acyl carrier protein (ACP) using affinity co-purifications and two-hybrid in E. coli. ACP, as a central cofactor in fatty acid synthesis, may be an ideal candidate as a mediator signalling starvation to SpoT. Accordingly, we show that the ACP/SpoT interaction is specific of SpoT and ACP functions because ACP does not interact with the homologous RelA protein and because SpoT does not interact with a non-functional ACP. Using truncated SpoT fusion proteins, we demonstrate further that ACP binds the central TGS domain of SpoT, consistent with a role in regulation. The behaviours of SpoT point mutants that do not interact with ACP reveal modifications of the balance between the two opposite SpoT catalytic activities thereby changing (p)ppGpp levels. More importantly, these mutants fail to trigger (p)ppGpp accumulation in response to fatty acid synthesis inhibition, supporting the hypothesis that the ACP/SpoT interaction may be involved in SpoT-dependent stress response. This leads us to propose a model in which ACP carries information describing the status of cellular fatty acid metabolism, which in turn can trigger the conformational switch in SpoT leading to (p)ppGpp accumulation.
Bacteria monitor their nutritional state and respond in the event of starvation by an elevation of the guanosine 5′-(tri)diphosphate, 3′-diphosphate [(p)ppGpp] nucleotide level that in turn regulates a large set of genes to stop growth and induce a stress response (Magnusson et al., 2005). In Escherichia coli, two homologous enzymes are involved in regulation of the (p)ppGpp level, RelA and SpoT. RelA synthesizes (p)ppGpp whereas SpoT can either degrade or synthesize it (Cashel et al., 1996). However, most bacteria contain only one protein, homologous to SpoT, which displays both degradation and synthesis activities (Mittenhuber, 2001). This bifunctional enzyme has been characterized in a variety of bacteria (Mechold et al., 1996; Avarbock et al., 1999), and has been shown to be crucial for long-term survival or virulence of numerous pathogens (Godfrey et al., 2002). The two catalytic sites are localized in the N-terminal domain of these proteins and the C-terminal domain is involved in regulation or protein binding (Gentry and Cashel, 1996; Gropp et al., 2001; Mechold et al., 2002). The three-dimensional structure of the N-terminal domain of the Rel/Spo bifunctional enzyme from Streptococcus equisimilis has been determined (Hogg et al., 2004). It confirmed that two separate catalytic sites in the N-terminal domain are responsible for synthesis and degradation of (p)ppGpp. However, the two catalytic sites share a central bundle of α-helices that participates in allosteric transitions between two conformational states suggesting that the regulations of the two activities are interconnected (Hogg et al., 2004). In E. coli, amino acid starvation triggers the RelA-dependent stringent response. RelA detects uncharged tRNA stalled on the ribosome and as a result activates its (p)ppGpp synthesis while hopping to other stalled ribosomes (Wendrich et al., 2002). However, the (p)ppGpp level increases in response to several other nutritional stresses such as fatty acid (Seyfzadeh et al., 1993; Gong et al., 2002), carbon (Xiao et al., 1991) or iron (Vinella et al., 2005) starvation. In these cases, SpoT-dependent (p)ppGpp synthase activity has been suggested to be necessary for the response. In contrast to RelA, there is currently no indication of how the signal is detected by SpoT or of the identity of the regulator that induces the shift in SpoT activity between synthesis and degradation. A few hypotheses have been proposed. For example, the energetic state of the cell at the membrane (Tetu et al., 1980) or the uncharged tRNA levels in the cytoplasm (Murray and Bremer, 1996) could be detected by SpoT. However, another exciting possibility is that SpoT would sense variations in the fatty acid metabolic status of the cell (DiRusso and Nystrom, 1998). Indeed, it has been shown that (p)ppGpp accumulates in a SpoT-dependent manner in response to fatty acid synthesis inhibition in the presence of glucose and amino acids (Seyfzadeh et al., 1993; Gong et al., 2002). Interestingly (p)ppGpp itself inhibits the fabHDG operon involved in fatty acid synthesis (Podkovyrov and Larson, 1996) and (p)ppGpp directly inhibits PlsB, the first-step enzyme of phospholipid synthesis (Heath et al., 1994). Therefore, there is a strong regulational link existing between the stringent response and lipid metabolism. The level of a critical intermediate of fatty acid synthesis could be the signal detected by SpoT (DiRusso and Nystrom, 1998).
In previous studies, SpoT protein was repeatedly detected among proteins co-purified by tandem affinity purification (TAP) of acyl carrier protein (ACP) in E. coli (Gully et al., 2003; Butland et al., 2005; Gully and Bouveret, 2006). To be functional, the 9 kDa ACP protein must be modified post-translationally by a 4′-phosphopantethein prosthetic group (4′-PP) enabling the binding of fatty acid intermediates through a thioester linkage (Rock and Cronan, 1996). The modified ACP (holo-ACP) acts as a cofactor in lipid metabolism by carrying the fatty acid intermediates through the successive enzymatic reactions of the fatty acid synthesis elongation cycle. It also delivers completed acyl chains to acyltransferase enzymes involved in envelope biogenesis or in synthesis of other acylated molecules. Thus, we hypothesized that if ACP, a central cofactor of fatty acid and phospholipid synthesis, interacted with SpoT, then ACP might well be the presumed soluble molecule responsible for triggering the SpoT-dependent response to fatty acid synthesis inhibition.
In the present study, we first demonstrated that SpoT, but not RelA, interacts with a functional form of ACP in vivo and then, we identified the domains of SpoT responsible for the interaction. Further, mutated SpoT proteins unable to interact with ACP were obtained. Plate tests of (p)ppGpp levels revealed that these mutants still display (p)ppGpp synthesis and degradation activities, but that the balance between the enzymatic activities was affected. We finally showed that the mutants fail to respond to fatty acid synthesis inhibition, suggesting that the ACP/SpoT interaction may be involved in the regulation of SpoT-dependent stress response. This allowed us to propose a model hypothesizing how the ACP/SpoT interaction could be involved in SpoT-dependent (p)ppGpp synthesis in response to fatty acid synthesis inhibition.
SpoT interacts with ACP
As it has been described in yeast (Gavin et al., 2002), purifications in E. coli using TAP method repeatedly contain common contaminants such as ribosomal and heat shock proteins but also other types of proteins such as SecA (Gully et al., 2003; Gully and Bouveret, 2006). Protein co-purification identified by a TAP proteomic approach is therefore not sufficient to conclude in the existence of a physiological interaction. As for interaction screenings using the two-hybrid technique (Brent and Finley, 1997), good practice prescribes that any new interaction presumed from a TAP purification should be confirmed by: (i) checking that it is not an unspecific binding to the columns, (ii) testing the reverse co-purification by tagging the candidate interactant and looking for co-purification of the initial TAP-tagged protein and (iii) using an independent method. We first present the tests performed to confirm that the SpoT protein, found in the purification of ACP (Gully et al., 2003; Butland et al., 2005), is indeed a physiological partner of ACP.
(i) For co-purification experiments, we used the CBP (calmodulin-binding peptide) tag that binds calmodulin beads in the presence of calcium. We first tagged ACP at its N-terminus with CBP (plasmid pBAD24-CBP-ACP) and verified that endogenous wild-type SpoT co-purified specifically with CBP-ACP on calmodulin beads (Fig. 1A). To ascertain that the band revealed with the anti-SpoT serum was indeed SpoT, the same experiment was performed with CBP-ACP in the ΔrelAΔspoT strain CF1693 lacking SpoT protein, and detected no band (Fig. 1A). This confirmed the identification of SpoT by mass spectrometry in the ACP-TAP purifications (Gully et al., 2003; Butland et al., 2005).
(ii) Then, we constructed a CBP-SpoT recombinant protein by fusing the CBP tag at the N-terminus of SpoT (plasmid pBAD24-CBP-SpoT). The functionality of this fusion protein was verified by expressing it in the ΔrelAΔspoT strain CF1693, devoid of (p)ppGpp and therefore unable to grow on minimal medium (MM) without amino acids (Xiao et al., 1991). Expression of CBP-SpoT in this strain restored growth on MM (data not shown). In order to detect ACP, we constructed a strain in which the chromosomal acpP gene is tagged on the chromosome at its 3′ end with the Flag tag (see Experimental procedures). Thus, the only ACP present in this strain is tagged at its C-terminus with the Flag epitope. ACP being essential (De Lay and Cronan, 2006), this demonstrated the functionality of the ACP-Flag recombinant protein and assured the physiological relevance of the interactions we detected. When performing a purification of CBP-SpoT on calmodulin beads in this strain, ACP-Flag was specifically co-purified with CBP-SpoT (Fig. 1B). Therefore, SpoT and ACP interact under the conditions of our purification assay and in total soluble extracts of wild-type E. coli.
(iii) We used the bacterial two-hybrid system based on the reconstitution of adenylate cyclase activity in an E. coli cya mutant (Karimova et al., 1998) to assess the interaction by an independent method. ACP and SpoT were fused to the C-terminus of the T18 and T25 domains of Bordetella pertussis adenylate cyclase using, respectively, the two compatible plasmids pUT18Clink and pKT25link (see Table 3). In the cya strain BTH101, adenylate cyclase activity was restored when the T18-ACP and T25-SpoT proteins were produced together (Fig. 1C). Therefore, SpoT and ACP proteins interact in vivo.
Shaded rows indicate the cloning vectors for each type of construction.
Antibiotic resistance, origin of replication and promoter are given for the cloning vectors. Description of the construction (oligonucleotides and restriction sites used) is given for the new plasmids. Oligonucleotide sequences are given in Table 4 and the numbering of mutants and truncated constructs refers to amino acid residues.
The ACP/SpoT interaction is specific and related to the function of both proteins
Specificity relative to SpoT. Because ACP plays a crucial role of cofactor in fatty acid synthesis, we hypothesized that ACP may carry the signal detected by SpoT in order to trigger the SpoT-dependent stress response resulting from fatty acid synthesis inhibition. In this case, the interaction should be specific for SpoT and be absent with RelA. This is because RelA does not respond to the same signals as SpoT, despite similarities of both sequence and domain organization (34% identity, 50% similarity; Fig. 2A). Unlike the T18-ACP/T25-SpoT interaction, no interaction was detected between T18-ACP and T25-RelA by two-hybrid (Fig. 1C). Also, endogenous RelA was not co-purified with CBP-ACP (data not shown) and conversely ACP-Flag was not co-purified with CBP-RelA (Fig. 1B). These results are consistent with the fact that RelA was never found in the TAP purifications of ACP (Gully et al., 2003; Butland et al., 2005). The lack of interaction between ACP and RelA demonstrates that ACP interacts specifically with SpoT.
Specificity relative to ACP. If SpoT could detect a signal carried by ACP through the ACP/SpoT interaction, it was important to determine which form of ACP was able to bind SpoT, i.e. apo-ACP (not modified by the 4′-PP group), holo-ACP (carrying a free 4′-PP group) or an ACP carrying a fatty acid intermediate bound by a thioester link to the 4′-PP group. The ACP residue modified post-translationally by 4′-PP is Serine 36. We first checked the interaction of SpoT with the ACP(Ser36Thr) mutant that cannot be modified by 4′-PP. The BTH101 cya strain was co-transformed with the two hybrid plasmids pT18-ACP(Ser36Thr) and pT25-SpoT. The interaction between the two constructs was nearly abolished (Fig. 1C). Therefore, only a post-translationally modified and hence functional form of ACP, either holo- or a thioester derivative, is able to interact with SpoT. To rule out the possibility that SpoT might interact with any 4′-PP-bearing protein, we performed the two-hybrid test with the EntB protein that also bears a 4′-PP group. No interaction was detected between T18-EntB and T25-SpoT (Fig. 1C). The lack of interaction between SpoT and either ACP(Ser36Thr) or EntB demonstrates that SpoT binds specifically to a functional form of ACP.
ACP interacts with the TGS domain
To further understand how ACP might regulate SpoT activities, we determined which domain(s) of SpoT were involved in the interaction with ACP. Indeed, the organization of Spo/Rel proteins into distinct domains has been well documented (Gentry and Cashel, 1996; Gropp et al., 2001; Mechold et al., 2002). The N-terminal half of the protein carries the enzymatic activities, whereas the C-terminal part is involved in the regulation of these activities, supposedly by binding ligands, by modification of the folding of the protein, or by both.
We constructed several truncated SpoT proteins fused to the T25 domain of B. pertussis adenylate cyclase. We first chose the limits of the N-terminal and C-terminal domains based on the predicted SpoT domains (Aravind and Koonin, 1998; Wolf et al., 1999) that are in accordance both with the experimental data mapping the activities of SpoT (Gentry and Cashel, 1996) and the protease-resistant domain of the S. equisimilis SpoT homologue used for structural experiments (Mechold et al., 2002). Both T25-Nterm-SpoT and T25-Cterm-SpoT recombinant proteins were correctly expressed as shown by Western blot with Anti-SpoT (Fig. 2B). ACP interacted with the C-terminal domain of SpoT (Fig. 2A, lane 4), which is consistent with a role of ACP in SpoT regulation. In order to map the region of interaction more precisely, we constructed two additional truncated proteins by separating SpoT protein in two parts, using the TGS domain boundaries. This domain has been named based on three families of proteins containing it: threonyl-tRNA synthetase, Obg family of GTPases and SpoT (Wolf et al., 1999). It can be precisely defined by sequence alignment and corresponds to a structural domain as shown by the three-dimensional structure obtained for the threonyl-tRNA synthetase (Sankaranarayanan et al., 1999). Both T25-Nterm+TGS-SpoT and T25-CtermΔTGS-SpoT recombinant proteins were correctly expressed as shown by Western blot with Anti-SpoT (Fig. 2B). T25-Nterm+TGS-SpoT interacted with ACP, whereas T25-CtermΔTGS-SpoT did not interact with ACP (Fig. 2A, lanes 5 and 6). To recapitulate, ACP interacted with the N-terminal domain of SpoT containing the TGS, whereas it did not with the N-terminal domain deleted for TGS (Fig. 2A, lanes 5 and 3). T18-ACP did not interact with the C-terminal domain deleted for TGS, whereas it interacted with the C-terminal domain containing the TGS (Fig. 2A, lanes 6 and 4). Thus, the TGS domain is the only common region of the two constructions interacting with ACP (N-terminal+TGS and C-terminal). However, we were not able to express a T25-TGS construction (containing only the TGS domain of SpoT), so measurement of the interaction between ACP and this domain alone was not possible.
That ACP apparently interacts with the TGS domain of SpoT when it is fused to either the N-terminal or C-terminal domain of SpoT is perhaps surprising given that this domain is the most conserved region between SpoT and RelA (51.6% identity and 75.0% similarity) and yet ACP does not interact with RelA. Either this very low divergence in the TGS sequences of RelA and SpoT is enough for ACP to discriminate between the two proteins, or each flanking region of the RelA protein could be incompatible with ACP binding. In order to decide between these two hypotheses, we constructed several chimeric proteins between SpoT and RelA proteins in the pKT25link plasmid as shown in Fig. 2A. Every recombinant protein was correctly expressed as shown by Western blot using Anti-SpoT or Anti-RelA serum (Fig. 2C). A T25-RelA protein with the TGS sequence of SpoT did not interact with T18-ACP, whereas a T25-SpoT protein with the TGS sequence of RelA did (Fig. 2A, lanes 7 and 8). This demonstrates that the TGS domain was not the discriminatory determinant permitting ACP binding to SpoT and not to RelA. It also shows that the nature of the N-terminal or C-terminal domain influenced the binding of ACP to the TGS. As the constructs containing the N-terminal domain of SpoT (T25-SpoRelSpo, T25-SpoRel and T25-SpoTGS-Rel) interacted with T18-ACP independently of the source of the C-terminal domain (Fig. 2A, lanes 7, 9 and 11) and those containing the N-terminal domain of RelA did not (Fig. 2A, lanes 8, 10 and 12), we reasoned further that the specificity was provided by the N-terminal domain of SpoT and that the N-terminal domain of RelA prevented the interaction of ACP with the TGS.
N-terminal and N-terminal+TGS domains of SpoT display different activities correlated with ACP binding ability
Having shown that ACP interacts specifically with SpoT and binds to the TGS domain, we wanted to correlate this interaction with the regulation of SpoT. We first tested the activities of the two constructions containing the N-terminal domain expressed from the pUT18Clink plasmid (pT18-Nterm+TGS-SpoT and pT18-Nterm-SpoT). The (p)ppGpp synthesis and hydrolysis activities of the T18-SpoT derivatives can be evaluated, respectively, in the standard reporter strains CF1693 and CF4943.
As already mentioned, the (p)ppGpp-deficient strain CF1693 (ΔrelAΔspoT) is unable to grow on MM without addition of amino acids (Xiao et al., 1991). Therefore, the ability of a construct to complement growth of CF1693 on MM indicates that (p)ppGpp is synthesized. CF1693 transformed with pT18-SpoT grew on MM as well as the single-copy spoT+ strain CF1652 (ΔrelAspoT+), whereas CF1693 transformed by the control plasmid pT18 did not (Fig. 3A). This demonstrates the ability of the recombinant T18-SpoT protein to synthesize (p)ppGpp in these conditions.
Strain CF4943 (ΔrelAspoT203) displays high basal (p)ppGpp levels due to a defect in (p)ppGpp degradation, and is only viable thanks to the mutation of the relA allele (Sarubbi et al., 1988). CF4943 grows slowly even on rich media [as Luria–Bertani (LB) or MM+amino acids] due to its constitutive high (p)ppGpp level. CF4943 transformed with pT18-SpoT grew as well as the single-copy spoT+ strain CF4941 on MM+amino acids, whereas CF4943 transformed with the control pT18 plasmid grew poorly owing to the lack of hydrolase activity (Fig. 3B). This demonstrates that the recombinant T18-SpoT protein lowers (p)ppGpp basal levels in CF4943, thanks to an excess of hydrolase over synthetase activity in these conditions, like a wild-type SpoT protein.
The T18-SpoT fusion protein displayed detectable (p)ppGpp synthesis and hydrolase activities in reporter strains with phenotypes similar to those of a single-copy spoT+ allele. It was therefore possible to follow variations in these activities for the two N-terminal domain constructions in the pUT18Clink plasmid.
Strain CF1693/pT18-Nterm-SpoT grew on MM without amino acids, indicating the presence of at least basal (p)ppGpp synthesis (Fig. 3A). On the contrary, strain CF4943/pT18-Nterm-SpoT grew as slowly as the CF4943/pT18 control strain on MM+amino acids, indicating the absence of detectable hydrolase activity in the T18-Nterm-SpoT construct (Fig. 3B). We also tested the growth of strain CF1693/pT18-NTerm-SpoT on a MM plate containing 1 mM each of Serine, Methionine and Glycine (SMG) that induce an isoleucine starvation. On this SMG plate, single-copy chromosomal spoT+ in the absence of relA does not produce enough (p)ppGpp to induce amino acid synthesis and permit growth, whereas a wild-type relA+spoT+ does (Uzan and Danchin, 1978). The CF1693/pT18-Nterm-SpoT strain grew on SMG plate, whereas CF1693/pT18-SpoT did not; this demonstrates that T18-Nterm-SpoT produces higher (p)ppGpp basal levels than the wild-type T18-SpoT construction (Fig. 3C). This was confirmed by (p)ppGpp labelling: under normal growth conditions, no (p)ppGpp was detected in strain CF1693/pT18-SpoT whereas high and constant amounts of (p)ppGpp were detected in strain CF1693/pT18-Nterm-SpoT (data not shown). In contrast, T18-Nterm+TGS-SpoT displayed a reversed activity compared with T18-Nterm-SpoT: it did not synthesize even enough (p)ppGpp to complement CF1693 for growth on MM (Fig. 3A) and instead was able to degrade (p)ppGpp as shown by growth complementation of CF4943 on MM+amino acids (Fig. 3B).
Therefore, T18-Nterm+TGS-SpoT appeared to be ‘locked’ in a state where hydrolysis is favoured over synthesis, whereas for T18-Nterm-SpoT the activity balance was reversed and shifted towards high (p)ppGpp levels favouring synthesis at the expense of hydrolysis. As we have shown that ACP binds Nterm+TGS-SpoT (via the TGS) and not Nterm-SpoT, the inverted balance of activities of T18-Nterm-SpoT and T18-Nterm+TGS-SpoT correlates with ACP binding ability. These results also highlight the importance of the TGS domain for the regulation of the catalytic sites localized in the N-terminal domain of SpoT.
ACP/SpoT interaction is involved in SpoT-dependent (p)ppGpp accumulation in response to fatty acid synthesis inhibition
To demonstrate that the ACP/SpoT interaction is involved in SpoT-dependent (p)ppGpp regulation, we wanted to test SpoT-dependent (p)ppGpp synthesis in response to fatty acid synthesis inhibition when the SpoT/ACP interaction is abolished. To isolate point mutants of SpoT impaired in ACP interaction, we followed two approaches.
First, we performed random mutagenesis on the pT25-SpoT plasmid and screened for T25-SpoT mutants unable to interact with T18-ACP by the bacterial two-hybrid technique (see Experimental procedures). We obtained one mutant, T25-SpoT(S587N), with a unique point mutation mapping in the C-terminal domain of SpoT that impaired the interaction with T18-ACP (Fig. 4A). The expression and stability of the mutated protein did not differ from the one of the wild-type construct (Fig. 4B). Furthermore, T18-SpoT(S587N) was still able to synthesize and hydrolyse (p)ppGpp as shown by complementation of, respectively, the reporter strains CF1693 and CF4943 (Fig. 4C and D).
Then, in a second approach, we looked for SpoT mutants affected in stringent response already described in the literature to check their interaction with ACP. The A404E mutation described to affect (p)ppGpp levels in the cell (Johansson et al., 2000) was particularly interesting to us because it was located in the TGS domain of SpoT (Fig. 4A). We tested the interaction of T25-SpoT(A404E) with T18-ACP by two-hybrid and the enzymatic activities of T18-SpoT(A404E) in the reporter strains CF1693 and CF4943. As shown in Fig. 4A, T25-SpoT(A404E) was impaired in its interaction with T18-ACP. The expression and stability of the mutated protein did not differ from the one of the wild-type construct (Fig. 4B). Surprisingly, T18-SpoT(A404E) did complement CF1693 for growth on MM indicating (p)ppGpp synthesis (Fig. 4C), although the spoT(A404E) mutation has been isolated in a strain described to be devoid of (p)ppGpp (Johansson et al., 2000). T18-SpoT(A404E) also complemented the degradation defect of CF4943 showing that T18-SpoT(A404E) is able to degrade (p)ppGpp (Fig. 4D).
In conclusion, two point mutations in the C-terminal domain of SpoT were isolated that impaired the interaction with ACP. The two mutated proteins were correctly expressed and still able to synthesize and degrade (p)ppGpp. Then, we asked whether their regulation and the SpoT-dependent response were affected.
A first hint was obtained from complementation tests of strain CF4943 on variable aminotriazole (AT) concentrations that provoke a histidine starvation. Constitutive high (p)ppGpp basal levels enable strain CF4943 to induce enough histidine synthesis to compete with high concentration of AT and permit growth, whereas the low basal levels of a ΔrelAspoT+ strain do not allow growth in AT (Rudd et al., 1985). CF4943 transformed with pT18-SpoT did not grow on 7 mM AT, whereas CF4943 transformed with the pT18 control plasmid did (Fig. 5A). The restoration of the AT sensitivity phenotype reaffirms that the balance of hydrolase over synthetase of the T18-SpoT recombinant protein is comparable to the balance of the wild-type SpoT protein. CF4943/pT18-SpoT(A404E) grew better than CF4943/pT18-SpoT on 7 mM AT (Fig. 5A, left), demonstrating higher basal levels of (p)ppGpp. On the contrary, CF4943/pT18-SpoT(S587N) did not grow on 5 mM AT, whereas CF4943/pT18-SpoT still grew at this concentration, demonstrating lower basal levels of (p)ppGpp in CF4943/pT18-SpoT(S587N) (Fig. 5A, right). Therefore, both mutations seemed to provoke changes in the balanced activities of the T18-SpoT protein (although opposite ones), showing a correlation between ACP binding and SpoT activity balance control. However, these tests were performed in steady state on plate and did not reflect the SpoT-dependent response to nutritional stress.
It was previously reported that SpoT responds specifically to fatty acid synthesis inhibition (Seyfzadeh et al., 1993). ACP being a central component of fatty acid synthesis, the interaction was likely to be directly involved in this response and we chose to test the behaviour of the two SpoT mutants for this SpoT-specific response.
Strain CF1693ΔrelAΔspoT was transformed with pT18-SpoT, pT18-SpoT(S587N) or pT18-SpoT(A404E) plasmids. In the resulting strains (p)ppGpp production can be only the result of the activities of the T18-SpoT derivatives. We performed global (p)ppGpp labelling and followed the response to fatty acid synthesis inhibition triggered by addition of cerulenin antibiotic. At the beginning of the experiment corresponding to steady-state growth, no (p)ppGpp could be detected from the background (Fig. 5B, lanes 0 min of the TLC plates). Following cerulenin addition (p)ppGpp became visible on TLC plate in strain CF1693 transformed with pT18-SpoT plasmid (Fig. 5B, first row). (p)ppGpp level increased in a time-course that is very similar to what was reported by Seyfzadeh et al. (1993) for a wild-type strain, with a total (p)ppGpp increase of approximately threefold (Fig. 5B). Once again, this comforted us in the use of T18-SpoT heterologous expression that behaves as the wild-type spoT+ allele. In strain CF1693 transformed with pT18-SpoT(S587N) or pT18-SpoT(A404E), no (p)ppGpp accumulation was visible after cerulenin addition (Fig. 5B, second and third rows), whereas GTP levels were comparable in the three strains (Fig. 5B, quantification not shown). However, from the complementation tests on plates, we knew that both mutants are able to synthesize (p)ppGpp (see above, Fig. 4C). Therefore, the two mutants do not respond to cerulenin. We cannot rule out that a low accumulation of (p)ppGpp might take place in the strains producing the T18-SpoT(S587N) or T18-SpoT(A404E) proteins, but with a level too low to be detected from the background. However, this response would still be significantly lower than those triggered by the wild-type T18-SpoT protein.
In conclusion, our results support the fact that ACP/SpoT interaction is needed for correct SpoT regulation in response to fatty acid synthesis inhibition.
It has been proposed previously that a key molecule of fatty acid metabolism such as a thioester derivative of coenzyme A (CoA) or of ACP could regulate SpoT (Seyfzadeh et al., 1993; DiRusso and Nystrom, 1998). In this article, we confirm the link between fatty acid metabolism and SpoT regulation and we present a body of arguments that led us to propose that ACP might be the crucial intermediate molecule signalling fatty acid synthesis inhibition to SpoT: we show that ACP interacts specifically with SpoT, not with RelA, and that SpoT interacts only with a functional form of ACP. This interaction is shown to involve the TGS domain present in the C-terminal regulating region of SpoT. Using SpoT truncated and point mutants, we found a correlation between SpoT/ACP interaction and the hydrolase/synthetase activity balance of SpoT. Finally, we show that SpoT point mutants impaired in ACP interaction fail to accumulate (p)ppGpp in response to fatty acid synthesis inhibition.
We conclude that ACP must interact via the TGS domain of SpoT as it can interact with the N-terminal domain+TGS and the C-terminal domain of SpoT (Fig. 2A). However, while the TGS domain is very conserved between SpoT and RelA proteins, it does not determine the specificity of the interaction for SpoT. Instead, the experiments with chimeric proteins revealed that the N-terminal domain of RelA prevented ACP binding. A weak influence of the C-terminal domain on ACP binding is also possible given the diminished intensity of the interaction with ACP for T25-SpoRel and T25-SpoTGS-Rel proteins (Fig. 2A). Yet, the presence in the chimera of the C-terminal domain of SpoT is not strictly required for ACP binding. Therefore, we can hypothesize that SpoT adopts a fold where the TGS comes near the N-terminal domain containing the catalytic activities to be regulated. The drastic effect on SpoT activities of the sole addition of the TGS domain to the N-terminal truncated domain of SpoT strengthens this hypothesis (Fig. 3). Beside Rel/Spo proteins, the TGS domain is found in tRNA synthetases and GTPases of the Obg family (Wolf et al., 1999). Based on the functions of the TGS containing proteins, it has been suggested that this domain may be involved in binding nucleotides (Wolf et al., 1999). We suggest that all these proteins may also be involved, as SpoT, in global stress response mechanisms in which the TGS may play a role.
Acyl carrier protein has been shown to interact with proteins not directly involved in fatty acid or phospholipid synthesis (Gully et al., 2003; Butland et al., 2005; Gully and Bouveret, 2006) and SpoT has been shown to interact with CgtA (Wout et al., 2004), a small GTPase whose function is controversial, from DNA replication checkpoint to ribosome maturation (Foti et al., 2005; Sato et al., 2005). It will be important to check the relation of ACP and SpoT with these various partners that could participate in a SpoT network involved in monitoring the metabolic state of the cell. Furthermore, it seems that Spo/Rel proteins display different activities depending on the host: the SpoT homologue of S. equisimilis displays different basal activities in E. coli or in S. equisimilis (Mechold et al., 1996). Such a case of host-specific activities could be explained by breakage of specific protein–protein interactions, such as with ACP, and it would be interesting to test heterologous protein interactions.
The structure of the N-terminal domain of the SpoT homologue of S. equisimilis suggests that the two activities of the Rel/Spo bifunctional proteins are interconnected and regulated allosterically as a balance between two states, which is logical to avoid futile cycle of (p)ppGpp synthesis and degradation (Hogg et al., 2004). In various Spo/Rel proteins (E. coli SpoT and RelA proteins and S. equisimilis SpoT homologue), it has been shown that the truncation of the C-terminal domain led to a drastic shift towards high (p)ppGpp synthesis levels (Gentry and Cashel, 1996; Gropp et al., 2001; Mechold et al., 2002). This is consistent with what we obtained with the T18-Nterm-SpoT construct that corresponds more or less to the SpoT(Δ376–702) truncated protein described previously (Gentry and Cashel, 1996). Based on these results, it has been proposed that the C-terminal domain of Rel/Spo proteins might exert a repressor role on (p)ppGpp synthesis that could be regulated by ligand or protein binding (Mechold et al., 2002).
We first studied the steady-state balance between (p)ppGpp synthesis or hydrolysis activities of SpoT and its derivatives in reporter strains. The results are summarized in Table 1. We used SpoT constructions in fusion with the T18 domain of adenylate cyclase and expressed from the pUT18Clink plasmid, i.e. with high copy number and basal Plac promoter expression. It has been reported that the activity balance in a strain overexpressing SpoT is not the same as those of a wild-type strain (Sarubbi et al., 1989). However, the phenotypic tests that we performed revealed no detectable difference between a strain producing T18-SpoT expressed from the pT18-SpoT plasmid and a single-copy spoT+ strain. Furthermore, we found that the T18-SpoT protein was functional for both (p)ppGpp synthesis and degradation, allowing us to study variations in these activities. The N-terminal domain of SpoT seems locked in a state of (p)ppGpp synthesis, together with an inability to bind ACP, whereas the N-terminal+TGS domain that binds ACP seems locked in a degradation state. This would be consistent with the TGS domain alone acting as a strong repressor of synthesis activity, as it was suggested for the whole C-terminal domain (Mechold et al., 2002). ACP bound to the TGS domain may act as a modulator of this repression depending on growth conditions.
Table 1. Summary of the phenotypes observed in reporter strains.
+ means growth of CF1693 with plasmid on M9 glucose medium.
+ means complementation of the slow growth phenotype of CF4943.
+ means growth of CF1693 on SMG plate.
+ means growth of CF4943 on aminotriazole (AT). The AT concentration permitting growth is indicated.
+ means interaction of the T25-SpoT derivative with T18-ACP by two-hybrid.
nd, not determined.
< 7 mM
Locked in synthesis
Locked in degradation
< 5 mM
Modified balance of activities
> 7 mM
Modified balance of activities
SpoT(A404E) and SpoT(S587N) mutants that are impaired for interaction with ACP are not blocked in a synthesis state, as it was the case for the N-terminal truncated domain. Instead, they seem to allow both (p)ppGpp synthesis and degradation in steady-state conditions. This shows that ACP binding is not the cause of repression of the ppGpp synthesis activity. However, the two mutants display a modified balance between the two activities compared with wild-type SpoT, resulting in a lower (p)ppGpp basal level for SpoT(S587N) and a higher (p)ppGpp basal level for SpoT(A404E) as visible by their behaviour in strain CF4943 on AT plates (Fig. 5A). This suggests that impairment of ACP binding affects the regulation performed by the TGS domain, resulting in an alteration of the activity balance. Yet, these differences are too low to detect with ppGpp labelling in steady-state growth conditions (Fig. 5B, lanes 0 min). Finally, when faced to fatty acid synthesis inhibition, both mutants fail to induce SpoT-dependent (p)ppGpp accumulation, suggesting that the lack of ACP/SpoT interaction prevents correct response from SpoT. The mutation S587N is not in the TGS domain and we showed that the C-terminal part of SpoT is not necessary for ACP binding (the Nterm+TGS construction interacted with ACP). We suggest that this mutation does not affect a residue directly involved in ACP binding but rather provokes conformational changes in SpoT mimicking a normally ACP-driven conformational modification, and the inability of ACP to bind SpoT(S587N) would be an indirect effect of this blockage.
Acyl carrier protein is central to fatty acid metabolism, being the required cofactor for all syntheses in the cell involving acyl chains. Subtle changes in fatty acid availability may therefore be transmitted by ACP. It has been proposed that the signal sensed by SpoT during fatty acid synthesis inhibition may be synthesized from fatty acid derivatives such as fatty acyl-CoA (Seyfzadeh et al., 1993; McDougald et al., 2002). ACP is as good a candidate as CoA for this role because ACP is involved for syntheses while CoA is involved in fatty acid degradation. Fatty acid metabolism could also be the relay for carbon source starvation, another nutritional stress sensed by SpoT. Indeed, carbon deprivation would lead to fatty acid starvation through shrinkage of the acetyl-CoA pool produced during glycolysis (DiRusso and Nystrom, 1998). In this regard, we intend to test whether SpoT-dependent response to glucose starvation is also affected when the ACP/SpoT interaction is impaired.
If ACP transduces the lipid metabolic status of the cell to SpoT via protein–protein interaction, it remains to be understood whether the signal transmission involves binding or release of ACP, or whether ACP is always bound to SpoT. In preliminary experiments, we did not detect any changes in the intensity of the SpoT/ACP interaction (using two-hybrid or co-purification methods) under experimental starvation conditions such as cerulenin addition or glucose starvation (A. Battesti and E. Bouveret, unpublished), suggesting that ACP may be permanently bound to SpoT. This would be reminiscent of RelA protein being bound to ribosomes under normal growth conditions. Then, a specific modification in the form of bound ACP [apo-ACP, holo-ACP (4′-PP) or specific thioester derivatives] could induce structural changes in ACP that would be transduced to SpoT. The known flexibility and dynamic of ACP structure (Kim and Prestegard, 1989) is compatible with this hypothesis. Demonstrating that SpoT binds the holo-ACP form does not rule out that very subtle changes in the proportion of the different ACP forms could be sensed. Therefore, it seems that it will be necessary to purify the SpoT-bound ACP during fatty acid synthesis inhibition or to reconstitute in vitro the system to understand the mechanism by which ACP regulates SpoT activity.
In conclusion, we propose a model in which SpoT, complexed with ACP, would sense the lipid metabolic state of the cell, like RelA, bound to the ribosomes where it detects uncharged tRNA that triggers its synthetic activity (Fig. 6). This parallel functions as well with respect to the ratio of the protagonists: hundreds of RelA or SpoT molecules per cell (Pedersen and Kjeldgaard, 1977), and, respectively, 70 000 ribosomes during optimal growth or 60 000 molecules of ACP (Rock and Cronan, 1996). This model explains the maintenance of two Rel/Spo protein homologues in E. coli, each one specialized in a specific pathway of activation. This is consistent with the specificity of the binding of ACP to SpoT only (and not to RelA) despite the high sequence similarities between the two proteins. During favourable growth conditions, RelA bound to ribosomes maintains a basal activity. Similarly, during favourable growth conditions, ACP bound to the TGS would keep SpoT in a basal hydrolase activity leading to low (p)ppGpp levels. In case of fatty acid or carbon starvation, modification of the proportion of the different ACP forms in the cell would be monitored by SpoT, via the ACP proteins bound to SpoT. Through structural changes of ACP, this would result in an allosteric transition of SpoT turning the balance towards synthesis, finally resulting in SpoT-dependent response to nutritional starvation.
Strains, media and plasmids
All strains used were derivatives of E. coli K-12 and are listed in Table 2. Cells were grown at 37°C in LB medium unless stated otherwise (Miller, 1992). Plasmids were maintained with ampicillin (100 μg ml−1), kanamycin (50 μg ml−1) or chloramphenicol (30 μg ml−1). Wild-type strain MG1655 was used for co-purification experiments. Strain MC4100 was used for expression tests. W3110/ACP-Flag strain was constructed by gene replacement using the pKO3 suicide vector (Link et al., 1997): wild-type strain W3110 was transformed with a pKO3 derivative plasmid containing an acpP-Flag allele (pEB770; Table 3). Insertion of the plasmid in the chromosome was selected on chloramphenicol plates at 42°C. Then, plasmid sequences removal was selected on 5% sucrose plates without antibiotic and acpP-Flag recombinant strains were screened by Western blotting. CF1693, CF1652 (Xiao et al., 1991), CF4941 and CF4943 (Gentry et al., 1993) strains were used to test SpoT activities. BTH101 strain (Karimova et al., 1998) was used for two-hybrid analyses. SpoT(A404E) mutation was obtained from JGJ301 strain (Johansson et al., 2000). M9 MM glucose plates, AT plates and SMG plates were prepared as described (Rudd et al., 1985; Miller, 1992). Plasmids and their construction are described in Tables 3 and 4.
Plasmid pT25-SpoT was subjected to random mutagenesis through propagation in the XL1-Red strain from Stratagene. BTH101/pT18-ACP strain was transformed with the resulting pT25-SpoT mutated plasmids library. After selection on X-Gal indicator plates, clones that did not interact with ACP were subcloned in pUT18Clink and pKT25link plasmids and sequenced.
From a 400 ml culture grown at 37°C in LB and induced during 60 min with 0.01% arabinose to a final OD600 of approximately 2, an extract was prepared by sonication in 10 ml of calmodulin-binding buffer (10 mM Tris-Hcl pH 8.0, 150 mM NaCl, 0.1% NP40, 1 mM Mg-acetate, 1 mM imidazole, 2 mM CaCl2). After centrifuging for 30 min at 27 000 g, glycerol was added to 15% final concentration and the extract was frozen in liquid nitrogen. For each co-purification assay, 1.5 ml of extract was incubated on 30 μl of calmodulin beads washed in calmodulin-binding buffer. After 60 min of incubation at 4°C, the beads were washed four times in 1 ml of calmodulin-binding buffer, resuspended in 30 μl of Laemmli loading buffer and heated for 10 min at 96°C.
We used the adenylate cyclase-based two-hybrid technique (Karimova et al., 1998). Pairs of proteins to be tested were fused to the two catalytic domains T18 and T25 of adenylate cyclase using, respectively, plasmids pEB355 and pEB354 (Table 3). After co-transformation of BTH101 strain with the two plasmids expressing the fusions, plates were incubated at 30°C for 2 days. Three millilitres of LB medium supplemented with ampicillin, kanamycin and 0.5 mM IPTG was inoculated and grown at 30°C overnight. β-Galactosidase activity was determined as described by Miller (1992). The values presented are the mean of three independent assays.
Measurements of (p)ppGpp
Measurements of (p)ppGpp were performed accordingly to Cashel (1994). In brief, cultures in low-phosphate medium (about 0.4 mM phosphate) were continuously labelled for at least two generations with 100 μCi of [32P]orthophosphate per ml starting at DO600 = 0.05. Cerulenin was then added at a final concentration of 200 μg ml−1, and 20 μl of samples were taken at time 0, 7, 15, 30 and 60 min. Samples were immediately mixed with 20 μl of 16 M formic acid on ice. Equal volumes of the acid formic extracts were chromatographed in one dimension on 20 × 10 cm polyethyleneimine cellulose TLC plates (JT Baker). TLC plates were developed using a FLA5100 Fuji phosphorimager. (p)ppGpp spots and GTP spots were quantified. The graphs present the ratio of (p)ppGpp/GTP.
We are very grateful to M. Cashel for the gift of antibodies, strains, plasmids, and for his constant support. We are grateful to D. Vinella and D. Schneider for the gift of antibodies, strains and plasmids, and to B.E. Uhlin for the JGJ301 strain. We also thank J. Sturgis, D. Gully, D. Leduc, and people from Lloubès and Sturgis groups for discussion and reading of the manuscript. A. Magalon helped us with the two-hybrid technique and N. Philippe trained us to the ppGpp assay; thanks to both of them. This work was funded by an ACI grant of the French Ministry and A.B. is recipient of a French Ministry fellowship.