Phosphotransferase-dependent accumulation of (p)ppGpp in response to glutamine deprivation in Caulobacter crescentus

The alarmone (p)ppGpp is commonly used by bacteria to quickly respond to nutrient starvation. Although (p)ppGpp synthetases such as SpoT have been extensively studied, little is known about the molecular mechanisms stimulating alarmone synthesis upon starvation. Here, we describe an essential role of the nitrogen-related phosphotransferase system (PTSNtr) in controlling (p)ppGpp accumulation in Caulobacter crescentus. We show that cells sense nitrogen starvation by way of detecting glutamine deprivation using the first enzyme (EINtr) of PTSNtr. Decreasing intracellular glutamine concentration triggers phosphorylation of EINtr and its downstream components HPr and EIIANtr. Once phosphorylated, both HPr∼P and EIIANtr∼P stimulate (p)ppGpp accumulation by modulating SpoT activities. This burst of second messenger primarily impacts the non-replicative phase of the cell cycle by extending the G1 phase. This work highlights a new role for bacterial PTS systems in stimulating (p)ppGpp accumulation in response to metabolic cues and in controlling cell cycle progression and cell growth.

T o face the environmental changes, organisms have developed complex regulatory mechanisms that integrate stimuli and stresses. Once activated, these signalling pathways modulate essential cellular processes such as DNA replication, cell division or cell growth. For example, upon nutrient starvation, yeast cells access to a specific quiescent state that enhances stress resistance and survival 1 . Bacteria also select many strategies to survive in challenging environments. One of the most studied bacterial adaptations to harsh conditions is certainly the formation of endospores in Bacillus subtilis, which requires asymmetric cell division and differentiation of the prespore 2 . Other bacteria take advantage of their asymmetric cell division to adapt to starvation conditions. It is the case of the aquatic a-proteobacterium Caulobacter crescentus that divides asymmetrically to give birth to two different daughter cells: a chemotactically active motile swarmer cell and a sessile stalked cell. Whereas the stalked cell grows and reinitiates DNA replication immediately at birth to ultimately divide again, the newborn swarmer cell enters first into a pre-replicative (G1) phase (Fig. 1a). In nutrient-replete conditions, the G1/swarmer cell differentiates into a stalked cell (swarmer-to-stalked transition) and concomitantly initiates chromosome replication (G1-to-S transition) 3 . Upon nitrogen starvation, C. crescentus extends its swarmer phase to promote spreading and colonization of new environments [4][5][6] . Asymmetric cell division might be a strategy commonly used by a-proteobacteria to generate daughter cells with different cell fates 7 .
Nutritional stresses are also known to be associated with the accumulation of an alarmone, the guanosine tetra-and penta-phosphate commonly called (p)ppGpp. Burst of intracellular (p)ppGpp alarmone allows cells to quickly adapt to a nutrient stress by affecting important cellular processes such as transcription, translation or DNA replication (reviewed in refs 8 and 9). For example, (p)ppGpp interferes with cell cycle steps by the direct binding of the alarmone to the DNA primase DnaG, which stops DNA replication in B. subtilis 10 . As a consequence of its pivotal role in stress adaptation, (p)ppGpp became crucial for virulence of several bacterial pathogens, long-term persistence, competence and biofilm formation 8,9,11 .
In Escherichia coli, (p)ppGpp level is regulated by two proteins, RelA and SpoT 12 . RelA is a monofunctional (p)ppGpp synthetase stimulated by amino acids starvation, in contrast to SpoT, which is a bifunctional synthetase-hydrolase enzyme responding to a wide range of nutritional stresses such as carbon, phosphate or fatty acid starvation 9 . C. crescentus possesses a single RelA/SpoT homologue 12 that was named SpoT because of its bifunctional activity 13,14 . Previous studies showed that (p)ppGpp can modulate cell cycle progression in C. crescentus by delaying simultaneously the swarmer-to-stalked differentiation and the G1-to-S transition 13,15,16 . Nitrogen or carbon starvation was described to trigger accumulation of (p)ppGpp but the regulatory networks sensing these stresses and activating SpoT remain uncovered 13,14 . Furthermore, interacting partners of SpoT Cc are largely unknown even if SpoT Cc was shown to co-fractionate with the 70S ribosomal subunit 14 .
Ammonium (NH 4 þ ) is the preferred inorganic nitrogen source for most of living cells. There are only two reactions that efficiently assimilate NH 4 þ (Fig. 1b). The first one is catalysed by the NADPdependent assimilative glutamate dehydrogenase. The other one is mediated by the ATP-dependent glutamine synthetase (GlnA). There is no NADP-dependent glutamate dehydrogenase encoded in the genome of C. crescentus, suggesting that the assimilation of inorganic nitrogen is strictly dependent on the glutamine synthetase (GlnA) activity. In most bacteria, nitrogen metabolism is tightly regulated by a well-characterized pathway, which involves the universal nitrogen sensor GlnD (Fig. 1b, reviewed in refs 17 and 18). When E. coli is grown in nitrogen-deplete ( À N) conditions, the PII uridyltransferase GlnD catalyses the transfer of uridine monophosphate (UMP) groups to PII regulatory proteins, GlnB and GlnK. GlnKBUMP no longer inhibits the ammonia channel AmtB, and GlnBBUMP stimulates deadenylylation of GlnA by the adenylyltransferase GlnE, and thereby promotes glutamine synthetase activity (Fig. 1b). In nitrogen-replete ( þ N) conditions, GlnB inhibits the transcription of glnA, by stimulating dephosphorylation of the transcriptional activator NtrC, and promotes the addition of the adenine monophosphate groups by GlnE to GlnA, which slows down the glutamine synthetase activity (Fig. 1b).
In this work, we unravel the regulatory network that stimulates (p)ppGpp accumulation in C. crescentus in response to nitrogen starvation. In particular, we uncover the essential role of the nitrogen-related phosphoenolpyruvate (PEP) phosphotransferase system (PTS Ntr ) in transducing glutamine deprivation signal to (p)ppGpp accumulation, which in turn controls the cell cycle progression. The cell cycle control described here constitutes a (a) Asymmetric cell division of C. crescentus gives birth to a non-replicative swarmer cell that goes through G1 phase before replicating and a replicative stalked cell that directly enters into S phase. Upon nitrogen starvation ( À N), swarmer cells extend their G1 phase. (b) In E. coli, ammonium can be assimilated either by the NADP-dependent glutamate dehydrogenase (GdhA) to generate glutamate (Glu) from a-ketoglutarate (a-KG) or by the glutamine synthetase (GlnA) to produce glutamine (Gln) from Glu, this latter being recycled by the glutamate synthase (GOGAT). GlnA is regulated at different levels by the GlnD/GlnB/ GlnE and NtrBC pathways, and GlnD senses intracellular pool of Gln.
new PTS Ntr -dependent regulatory role, illustrating the diversity of the cellular processes regulated by PTS systems.

Results
Glutamine deprivation signals nitrogen starvation. Previous studies showed that nitrogen starvation extends the swarmer cell lifetime in C. crescentus 4-6 . By following DNA content and cell cycle-regulated proteins (the flagellin and the stalked-associated protein StpX) in synchronous or asynchronous population of Caulobacter cells, we confirmed the specific extension of the G1/swarmer phase in response to nitrogen starvation ( Supplementary Fig. 1a-e). By contrast, stalked cells can complete DNA replication once initiated, despite the absence of a nitrogen source ( Supplementary Fig. 1f).
To understand how nitrogen starvation affects the differentiation of G1/swarmer cells, we focused our work on proteins involved in nitrogen assimilation and metabolism. First, we created an in-frame deletion of the predicted gene coding for the general sensor for nitrogen availability, glnD (CCNA_00013). In contrast to wild-type cells, DglnD cells were unable to use ammonium as a nitrogen source. Indeed, the DglnD mutant did not grow and accumulated G1/swarmer cells when ammonium was used as the sole nitrogen source ( Supplementary Fig. 2a,b). As expected, the G1 block and the growth were rescued in the presence of glutamine ( Supplementary Fig. 2). Indeed, as for glnD mutants in E. coli 19 , C. crescentus DglnD is auxotrophic for glutamine. Thus this result shows that G1/swarmer cells accumulation in DglnD is a consequence of its inability to use ammonium as a nitrogen source. Interestingly, the DglnD mutant strain cultivated in a complex peptone yeast extract (PYE) medium accumulated G1/swarmer cells (Fig. 2c,d). As a consequence, DglnD cells also exhibited (i) a slower growth than the wild-type strain and (ii) a bigger motility halo than the wild-type, despite the growth defect (Fig. 2a,b). Indeed, the overrepresentation of the G1/swarmer cells in a DglnD population increases the overall motility and the doubling time of the strain. Thus, our results indicate that G1/swarmer lifetime is extended in the absence of glnD (Fig. 2a-d). Again, addition of glutamine suppressed all these phenotypes (Fig. 2d), suggesting that defects of DglnD are a consequence of glutamine availability in the complex PYE medium. Indeed, PYE is mainly composed of yeast extract, in which glutamine is the less-abundant amino acid (r0.2%, see the 'Methods' section).
In E. coli, glutamine auxotrophy displayed by glnD mutant strains comes from the under-expression and lower activity of the glutamine synthetase GlnA. In the absence of GlnD, the PII protein GlnB is not uridylylated, and thereby constitutively stimulates (i) the dephosphorylation of transcriptional activator NtrCBP by NtrB, which subsequently inhibits the NtrCBPdependent expression of glnA, and (ii) the adenylylation ( þ adenine monophosphate) of GlnA by the adenylytransferase GlnE, thereby inhibiting the glutamine synthetase activity (Fig. 1b). C. crescentus encodes three PII protein homologues (glnB CCNA_02046, glnK CCNA_01400 and glnC CCNA_00555), one adenylytransferase homologue (glnE CCNA_02839), three GlnA homologues (glnA CCNA_02047, glnA 2 CCNA_03230 and glnA 3 CCNA_03240) and two NtrC homologues (ntrC CCNA_01815, and ntrX CCNA_01817). Single in-frame deletions of each of these genes were created and tested for growth, motility and G1 accumulation in complex PYE medium. We found that only DglnB, DntrC and DglnA recapitulated DglnD phenotypes, and that all these phenotypes could be suppressed by adding glutamine to the medium (Fig. 2d). However, it is noteworthy that the motility defect in DntrC was not completely rescued by addition of glutamine, suggesting that NtrC also controls motility independently of G1/swarmer cells accumulation (Fig. 2d). Moreover, deleting glnE alleviated the DglnD defects, supporting the fact that inactivation (adenylylation) of GlnA by GlnE is the causative effect of the glutamine auxotrophy detected in DglnD cells (Fig. 2d). Unexpectedly, neither GlnA 2 nor GlnA 3 seems to participate in glutamine synthesis, at least in the conditions used here ( Supplementary Fig. 3a,b). Finally, a catalytic mutant inactivating the glutamine synthetase activity of GlnA (GlnA R360A ) phenocopied DglnA in terms of glutamine auxotrophy, growth defect and G1 accumulation ( Supplementary  Fig. 3c,d), and all these defects can be suppressed by the addition of glutamine or complemented with a wild-type copy of glnA expressed in trans ( Supplementary Fig. 3d,e). Altogether, these data support that (i) glutamine deprivation constitutes the signal for nitrogen starvation and that (ii) intracellular levels of glutamine drive the cell cycle progression of C. crescentus.
Cell cycle response to nitrogen starvation requires (p)ppGpp.
To fish out key actors participating to the G1/swarmer extension in response to glutamine deprivation, we isolated spontaneous mutations that increase motility of the wild-type strain on PYE swarm agar plates supplemented with glutamine. Whole-genome sequencing of one candidate revealed a unique mutation (D81G) in the hydrolase domain of SpoT, the protein synthesizing (p)ppGpp ( Fig. 3a and Supplementary Fig. 4a). It has been shown that mutations abolishing, at least partially, the hydrolase activity of SpoT without altering its synthetase activity, were all found in the hydrolase domain 20 . Interestingly, the conserved aspartate, corresponding to the D81 in C. crescentus SpoT, was found to be required for the hydrolase activity of SpoT in Streptococcus dysgalactiae 20 , which suggests that the D81G mutation may reduce hydrolase activity of SpoT in C. crescentus as well. The spoT D81G strain had a growth delay in PYE and accumulated G1/ swarmer cells in exponential phase of growth ( Fig. 3b; suppressed the accumulation of G1/swarmer cells ( Fig. 3b and Supplementary Fig. 5b,c). This result highlights the role of (p)ppGpp produced by SpoT in response to glutamine deprivation to control the cell cycle. We thus checked the (p)ppGpp production in À N or þ N conditions. In agreement with previous studies 14 , we found that (p)ppGpp concentration increases on glutamine deprivation, that is, in the wild-type strain grown without nitrogen source ( À N) or in DglnD cultivated with ( þ N) or without ( À N) NH 4 þ (Fig. 3c,d; Supplementary Figs 1g and 5d). As already mentioned in previous works 13,15 , we also observed a low amount of (p)ppGpp produced by wild-type cells in non-stressed conditions ( Fig. 3c and Supplementary Fig. 5d). This (p)ppGpp steady-state level was slightly higher in the spoT D81G strain in þ N conditions ( Fig. 3c and Supplementary Fig. 5d), which would explain the phenotypes displayed by spoT D81G , that is, slower growth, bigger motility halo and G1 accumulation (Fig. 3b), even in the presence of glutamine (Supplementary Figs 5a-c and 6). Nevertheless, spoT D81G cells accumulated similar levels of (p)ppGpp than wild-type cells upon nitrogen starvation ( À N), suggesting that SpoT D81G is still sensitive to nitrogen starvation ( Fig. 3d). In contrast, the disappearance of (p)ppGpp accumulated upon nitrogen starvation is slower in spoT D81G cells than in wildtype cells ( Supplementary Fig. 5e). These results support that D81G mutation mainly affects the hydrolase activity of SpoT. It's noteworthy that an artificial increase of (p)ppGpp levels in nonstarved cells displayed similar phenotypes than spoT D81G in complex medium PYE 16 . The fact that, in non-stressed conditions, a strain producing more (p)ppGpp (spoT D81G ) accumulated G1/ swarmer cells, whereas a strain producing no (p)ppGpp at all (DspoT) contained less G1/swarmer cells (Fig. 3b,c), suggest that (p)ppGpp steady-state level might determine the time spent by swarmer cells in G1 phase. Altogether, our findings indicate that glutamine deprivation increases (p)ppGpp level, which in turn, will extend the lifetime of G1/swarmer cells.

PTS Ntr promotes (p)ppGpp accumulation on nitrogen starvation.
To identify factors that participate to the activation of SpoT in response to nitrogen starvation, we selected for transposon insertions that improve growth of spoT D81G cells on complex medium (PYE). Indeed, the accumulation of (p)ppGpp in spoT D81G cells decreases the growth rate on PYE medium ( Fig. 3b and Supplementary Fig. 5a). We identified multiple transposon insertions (34 out of 50 clones) into spoT D81G itself. The remaining 16 clones harboured a transposon insertion into the ptsP gene (CCNA_00892), coding for a nitrogen-related PEP-phosphotransferase (PTS) protein homologue, called PtsP or EI Ntr in Enterobacteria (Fig. 3a). Canonical PTS systems are composed of several components that form a phosphorylation cascade initiated by autophosphorylation of the first protein called EI, using PEP as phosphoryl donor (reviewed in ref. 21). The phosphoryl group is then transferred from EIBP to HPr and then to EIIA proteins. When the PTS system is used to take up sugars, the phosphoryl group is ultimately transferred from EIIABP to transported carbohydrates by using specific permeases (EIIB and EIIC components). In many other cases, PTS systems are dedicated to regulatory functions implying that PTS components (EI, HPr or EIIA) phosphorylate or interact with regulatory target proteins 21 . Nitrogen-related PTS (PTS Ntr ) systems are so far considered as unusual PTS systems that respond to nitrogen availability, but their regulatory roles in bacterial physiology remain poorly understood (reviewed in ref. 21).
An in-frame deletion of ptsP in the parental spoT D81G strain suppressed the spoT D81G phenotypes, confirming the genetic interaction between ptsP and spoT D81G (Fig. 3b). In addition, DptsP phenocopied DspoT in terms of motility, G1/swarmer cells accumulation ( Fig. 3b; Supplementary Figs 5a-c and 7a-c), and capability to suppress G1/swarmers cells accumulation of DglnD cells (compare DglnD DptsP to DglnD DspoT in Fig. 3b; Supplementary Figs 5a-c and 7a-c). Interestingly, we isolated another candidate than spoT D81G in the gain-of-motility screen, which harboured a mutation (L83Q) in the GAF domain of EI Ntr (ptsP L83Q ; Fig. 3a). As ptsP L83Q phenocopies spoT D81G (Fig. 3b; Supplementary Figs 5a-c and 7a-c), DptsP suppresses spoT D81G defects and DspoT suppresses ptsP L83Q defects (Fig. 3b), we wondered whether ptsP (EI Ntr ) is upstream or downstream of spoT. To test that, we measured the (p)ppGpp levels in a spoT D81G DptsP background. As shown in Fig. 3d, no (p)ppGpp accumulation was detected in spoT D81G DptsP cells starved for nitrogen ( À N). However, spoT D81G DptsP cells still produced a low amount of (p)ppGpp, whether a nitrogen source was added to the medium or not (Fig. 3c). This constitutive low levels of (p)ppGpp produced by spoT D81G DptsP cells is very close to the (p)ppGpp level detected in non-starved wild-type cells (Fig. 3c). Interestingly, there is a systematic correlation between the amount of (p)ppGpp produced by the cells and the time spent by these cells in G1 phase. Indeed, DspoT or DptsP swarmer cells do not produce detectable levels of (p)ppGpp (Fig. 3c) and have shortened G1 phase (Fig. 3b), whereas spoT D81G DptsP and wild-type cells have similar levels of (p)ppGpp and G1 lifetime (Fig. 3b,c). Altogether, these results support the role played by (p)ppGpp in determining the G1 lifetime, and show that EI Ntr regulates (p)ppGpp levels by controlling SpoT.
Glutamine inhibits EI Ntr autophosphorylation. To understand how glutamine deprivation is transduced to SpoT, we first looked at the autophosphorylation level of EI Ntr . Indeed, as described above, accumulation of G1/swarmer cells (Fig. 3b) observed in a ptsP L83Q background, are not compensated by supplying an exogenous source of glutamine ( Supplementary  Fig. 6). Moreover, it has been recently shown, in the closely related a-proteobacterium Sinorhizobium meliloti, that binding of glutamine to the conserved N-terminal GAF domain of EI Ntr inhibits its autophosphorylation 22 . To check whether the phosphorylation of EI Ntr is also sensitive to glutamine in C. crescentus, we performed in vitro autophosphorylation assays with a purified fraction containing EI Ntr using [ 32 P]PEP as a phosphoryl donor, in the presence or absence of glutamine. We found that autophosphorylation of EI Ntr was strongly reduced by glutamine ( Fig. 4a and Supplementary Fig. 8a). In contrast, autophosphorylation of the EI Ntr L83Q mutant form was not modulated by the presence of glutamine (Fig. 4a), suggesting that the mutation L83Q prevents glutamine binding to the highly conserved region of the EI Ntr GAF domain ( Supplementary Fig. 4b).
Phosphorylated PTS Ntr proteins trigger (p)ppGpp accumulation. To unravel how EI Ntr controls SpoT activity, we first searched for components that could participate to PTS Ntr phosphorelay (Fig. 5a). Besides ptsP (EI Ntr ), we found a unique HPr homologue (ptsH, CCNA_00241) and another nitrogen-related PTS Ntr component, EIIA Ntr (ptsN, CCNA_03710). We created single in-frame deletions of the two genes (DptsH and DptsN) and found that the proportion of G1/swarmer cells in DptsH (without HPr) or DptsN (without EIIA Ntr ) strain was reduced in comparison with the wild-type strain in complex medium (Fig. 5b), a phenotype already described for DptsP (without EI Ntr ) and for DspoT (Fig. 5b). Interestingly, strains expressing a non-phosphorylatable version of EIIA Ntr (EIIA Ntr H66A ) accumulated G1 cells as much as the loss-of-function mutant (DptsN), that is, less than the wild-type strain (Fig. 5b), suggesting that the PTS Ntr pathway is slightly phosphorylated in complex medium PYE and that the last component of the PTS Ntr phosphorelay, EIIA Ntr , controls SpoT activity.
To validate the conservation of PTS Ntr phosphorelay and the inhibitory effect of glutamine on this cascade, we checked the phosphorylation level of EIIA Ntr in þ N or À N conditions.
To this purpose, we performed in vivo phosphorylation assays on WT and DptsP (without EI Ntr ) strains expressing a xylose-inducible tagged version of EIIA Ntr (PxylX::3FLAG-ptsN; Fig. 4b,c and Supplementary Fig. 8b,c). In agreement with our previous data, we found that the phosphorylation of EIIA Ntr is enhanced in the absence of nitrogen sources ( À N) in comparison with the þ N conditions (Fig. 4b,c and Supplementary Fig. 8b,c). In addition, we showed that EI Ntr is required in vivo for EIIA Ntr phosphorylation, since phosphorylated EIIA Ntr was undetectable in DptsP cells starved for nitrogen (Fig. 4b,c and Supplementary Fig. 8b,c).
In addition, we measured the (p)ppGpp levels in the single PTS Ntr mutants first in þ N conditions. Consistent with the G1 accumulation in PYE, we found that PTS Ntr mutant strains (DptsP, DptsH, DptsN or ptsN H66A ) produced significantly lower amount of (p)ppGpp than the wild-type strain in the presence of a nitrogen source (Fig. 5c). Moreover, EIIA Ntr H66A partially abrogated the cell cycle and developmental defects of ptsP L83Q (EI Ntr L83Q ) supporting the fact that overphosphorylation of EIIA Ntr in ptsP L83Q cells is partially responsible for (p)ppGpp accumulation and subsequent G1-to-S transition delay (Fig. 5b). On the contrary, a strain expressing a phosphomimetic mutant of EIIA Ntr (EIIA Ntr H66E ) had increased proportion of G1 cells independently of the presence of EI Ntr (Fig. 5b). Altogether, these data suggest that the phosphorylated form of EIIA Ntr (EIIA Ntr BP) controls SpoT activity.
However, (p)ppGpp measurements in À N conditions showed that SpoT is also controlled in an EIIA Ntr BP-independent way. Indeed, in contrast to cells devoid of EI Ntr (DptsP) or HPr (DptsH), which did not accumulate (p)ppGpp upon nitrogen starvation (Fig. 5d), the absence of EIIA Ntr BP in DptsN or ptsN H66A cells did not abolish (p)ppGpp accumulation upon nitrogen starvation (Fig. 5d), showing that SpoT is still sensitive to nitrogen availability in the absence of EIIA Ntr BP. On the basis of these results, we propose a model in which HPrBP controls the intracellular levels of (p)ppGpp by at least two ways, in an EIIA Ntr BP-dependent way but also independently of EIIA Ntr BP.
Phosphorylated EIIA Ntr directly interacts with SpoT. Since most of the regulatory functions of PTS components are mediated by protein-protein interactions, we checked whether HPr and EIIA Ntr were able to interact with SpoT by performing bacterial two-hybrid (BTH) assays. To this end, T18 or T25 domains of Bordetella pertussis adenylate cyclase 23 were fused to coding sequences of HPr (ptsH and ptsH H18A ), EIIA Ntr (ptsN, ptsN H66A and ptsN H66E ) and SpoT (spoT and spoT D81G ). We found that both the wild-type EIIA Ntr (ptsN) and the phosphomimetic mutant of EIIA Ntr (EIIA Ntr H66E ) were able to interact with SpoT versions (Fig. 6a,b and Supplementary Fig. 9a), while the non-phosphorylatable mutant EIIA Ntr H66A was not (Fig. 6a,b and Supplementary Fig. 9a). Both T18-EIIA Ntr H66A and T25-EIIA Ntr H66A can, respectively, interact with T25-HPr and T18-HPr ( Fig. 6c and Supplementary Fig. 9a), showing that EIIA Ntr H66A is functional in the BTH assays. Altogether, these findings suggest that EIIA Ntr is phosphorylated in vivo in E. coli. Indeed, there are two PTS systems in E. coli, a canonical one composed of EI (ptsI), HPr (ptsH) and EIIA (ptsM), as well as a nitrogen-related one composed of EI Ntr (ptsP), NPr (npr) and EIIA Ntr (ptsN), and both pathways can cross-talk to some extent 24 (Supplementary Fig. 9b). To test whether Caulobacter EIIA Ntr is phosphorylated in vivo in E. coli, strains deleted for npr (NPr) or for both ptsP (EI Ntr ) and ptsI (EI) genes were created. As illustrated on Fig. 6a and Supplementary Fig. 9c  (a) Glutamine inhibits autophosphorylation of EI Ntr but not EI Ntr L83Q . Autophosphorylation assays of EI Ntr and EI Ntr L83Q using [ 32 P]PEP as a phosphoryl donor in the absence or presence of increasing concentration of glutamine (0, 2, 5, 10 mM). The full autoradiography is available in Supplementary Fig. 8a. (b) The EI Ntr -dependent phosphorylation of EIIA Ntr is enhanced upon nitrogen starvation. In vivo phosphorylation assays of EIIA Ntr in À N or þ N conditions supplemented with ( þ Xyl) or without ( À Xyl) xylose in wildtype (WT) or DptsP cells expressing 3FLAG-ptsN from PxylX promoter. Error bars ¼ s.d.; n ¼ 3. Statistically significant differences by Student's t-test are indicated as **Po0.01% (n ¼ 3). The full autoradiography is available in Supplementary Fig. 8b. (c) Immunoblotting of protein samples extracted from WT and DptsP cells expressing 3FLAG-ptsN from PxylX promoter, incubated 3 h in M5GG supplemented with ( þ Xyl) or without ( À Xyl) xylose. MreB was detected in all conditions, while 3FLAG-EIIA Ntr was detected only in the presence of xylose. The full blot is available in Supplementary Fig. 8c. no more b-galactosidase activity was detected in a DptsP DptsI background (without E. coli EI proteins) expressing T25-SpoT and T18-EIIA Ntr , while the interaction between SpoT and EIIA Ntr H66E remained unchanged in this background (Fig. 6b). Finally, the expression of Caulobacter ptsH (HPr) from the inducible pBAD promoter (pBAD33-ptsH Cc ) in a Dnpr  background restored the interaction between EIIA Ntr and SpoT only when arabinose was added to the medium ( Supplementary  Fig. 9c), indicating that Caulobacter HPr and EIIA Ntr proteins can be phosphorylated by E. coli PTS systems, and that only the phosphorylated form of EIIA Ntr interacts with SpoT. In contrast to EIIA Ntr , no interaction was detected between HPr (or HPr H18A ) and SpoT (or SpoT D81G ) on MacConkey maltose agar plates (Supplementary Fig. 9d). The fact that HPr interacts with EIIA Ntr H66A (Fig. 6c and Supplementary Fig. 9a) shows that HPr is functional in the BTH assays.
Altogether, these BTH data strongly suggest that (i) HPr and EIIA Ntr are both phosphorylated in E. coli and (ii) EIIA Ntr BP is the only form of EIIA Ntr able to interact with SpoT, thereby supporting a model in which SpoT activity is controlled directly by EIIA Ntr BP, and indirectly by HPrBP (Fig. 8).
Phosphorylated EIIA Ntr inhibits hydrolase activity of SpoT. Interestingly, the deletion of ptsN (EIIA Ntr ) did not abolish the G1 accumulation of spoT D81G cells in contrast to DptsP (EI Ntr ) or DptsH (HPr; Fig. 7a). The fact that SpoT D81G , which harbours a reduced hydrolase activity ( Supplementary Fig. 5e), is insensitive to the presence of EIIA Ntr suggests that EIIA Ntr BP might inhibit the hydrolase activity of SpoT rather than stimulating its synthetase activity. This could explain why DptsN or ptsN H66A are still able to accumulate high levels of (p)ppGpp on nitrogen starvation, while DptsP or DptsH cannot (Fig. 5d). To validate our hypothesis, we engineered Caulobacter strains in which the only (p)ppGpp synthetase activity was supplied by the unrelated E. coli RelA protein, and we measured the endogenous hydrolase activity of SpoT in different genetic backgrounds. To this end, we first abolished the synthetase activity of SpoT in several backgrounds (spoT D81G , ptsN H66A , ptsN H66E and DptsP), by replacing the tyrosine 323 of SpoT by an alanine (SpoT Y323A ; ref. 14). As expected, all these strains displayed a G1 accumulation similar to a DspoT strain in PYE complex medium (Fig. 7c). In a second time, we inserted a truncated version of E. coli RelA (p)ppGpp synthetase at the xylose locus, leading to an artificial (p)ppGpp accumulation in Caulobacter upon addition of xylose (PxylX:: relA-FLAG; ref. 16). Since the hydrolase domain of RelA is inactive, the only (p)ppGpp hydrolase activity in these strains was carried out by the Caulobacter SpoT protein, while the only (p)ppGpp   synthetase activity was supported by the E. coli RelA protein.
In the presence of xylose, both spoT D81G Y323A and ptsN H66E spoT Y323A displayed a growth defect and a G1 accumulation in comparison with the parental spoT Y323A strain (Fig. 7b,c and Supplementary  Fig. 10). On the contrary, neither ptsN H66A spoT Y323A nor DptsP spoT Y323A strains had a growth delay or accumulated G1/swarmer cell upon xylose induction. These results strongly suggest that the phosphorylated form of EIIA Ntr (ptsN) specifically inhibits the hydrolase activity of SpoT. In support of this, we found that upon xylose induction (Fig. 7d), spoT Y323A PxylX::relA-FLAG cells accumulated (p)ppGpp in À N conditions (that is, when EIIA Ntr is highly phosphorylated; Fig. 4b), but not in þ N conditions (that is, when EIIA Ntr is less phosphorylated; Fig. 4b). Furthermore, (p)ppGpp accumulated in spoT D81G Y323A PxylX::relA-FLAG cells even in þ N conditions (Fig. 7d), supporting again that the D81G mutation abolishes the hydrolase activity of SpoT. Finally, (p)ppGpp became undetectable in spoT Y323A PxylX::relA-FLAG strains harbouring ptsN H66A or DptsP allele (Fig. 7d), indicating that SpoT hydrolase activity is completely unlocked when EIIA Ntr is unphosphorylated. Altogether, these findings demonstrate that EIIA Ntr BP inhibits hydrolase activity of SpoT to modulate (p)ppGpp accumulation upon nitrogen availability.

Discussion
Adaptation to starvation conditions requires sophisticated regulatory mechanisms that sense an external stimulus and translate it into an internal molecular response. In this report, we uncovered how Caulobacter copes with nitrogen starvation by triggering (p)ppGpp accumulation (Fig. 8), which in turn will control the cell cycle and development by extending the G1/swarmer phase 14,16 . Increasing the time spent in the non-replicative (G1), motile phase reflects the adaptation of Caulobacter cells to their natural environment, that is, freshwater in which nutrients can rapidly be limiting 3,4 . Interestingly, G1 arrest also occurs during the intracellular trafficking of Brucella abortus, and on nitrogen and carbon starvation in Sinorhizobium meliloti 25,26 . In addition, the G1 block encountered by S. meliloti cells starved for nitrogen and carbon is also dependent on (p)ppGpp 26,27 . Therefore, (p)ppGpp-dependent mechanisms delaying DNA replication initiation could be a common feature used by a-proteobacteria in response to harsh conditions such as infection or starvation. As previously suggested in the literature 5 , our data indicate that stalked cells are able to complete replication upon nitrogen starvation, supporting that only swarmer cells are responsive to nitrogen depletion. Indeed, even if the speed of chromosome duplication is slowed down in nitrogen-starved conditions, the stalked cell seems to be unable to stop ongoing DNA replication ( Supplementary Fig. 1f). In contrast, the swarmer cell can avoid DNA replication initiation in the same conditions ( Supplementary Fig. 1a-e). One reason for this difference could be that initiating DNA replication without enough nitrogen supplies would ultimately be detrimental to the cells. In support of this, we showed that deletion of spoT is deleterious in a DglnD background, since DglnD DspoT strain displays a strong growth defect (Fig. 3b). This result highlights the importance for Caulobacter swarmer cells to delay DNA replication until reaching a critical intracellular nitrogen pool.
Our data established that glutamine deprivation constitutes the intracellular signal perceived by the cell in response to nitrogen starvation and is sufficient to mediate (p)ppGpp accumulation (Fig. 8). Intracellular glutamine concentration is known to vary in bacteria, up to 10-fold depending on nitrogen availability 28 . As a consequence, monitoring intracellular glutamine concentration is an efficient strategy to evaluate nitrogen availability, and subsequently adjust nitrogen assimilation. In E. coli, the uridylyltransferase GlnD is known to directly sense the intracellular glutamine pool, and according to it, to modify uridylylation level of regulatory PII proteins (GlnB and GlnK), which in turn will adapt nitrogen metabolism. For instance, in the absence of glutamine, GlnD will increase ammonium transport, as well as the expression and activity of the glutamine synthetase. Intriguingly, three GlnA paralogs are encoded into the genome of C. crescentus, suggesting a functional redundancy, and the presence of multiple glutamine synthetase is conserved in several a-Proteobacteria 29 . Even though we showed that only the glutamine synthetase encoded by glnA is necessary for assimilating ammonium in complex and minimal media, we do not exclude that the two other paralogs (GlnA 2 and GlnA 3 ) display a glutamine synthetase activity under specific growth conditions that remain to be determined. Glutamine synthetase activity has been shown to promote growth of the obligatory intracellular a-proteobacterium Ehrlichia chaffeensis inside human cells 30 . Moreover, this successful intracellular growth of E. chaffeensis promoted by the glutamine pool was accompanied by a rapid degradation of CtrA, a cell cycle regulator known to inhibit DNA replication initiation in several a-proteobacteria 7,31 . Furthermore, the CtrA level in Caulobacter cells was shown to be maintained upon nitrogen starvation 5 , and even increased upon (p)ppGpp accumulation 16 . These observations suggest that the asymmetrically dividing a-proteobacteria might use glutamine as a metabolic cue for nitrogen availability that controls the cell The GlnD/GlnB/GlnE pathway (in blue) regulates glutamine homoeostasis by modulating GlnA (GS) activity. Intracellular glutamine inhibits EI Ntr autophosphorylation, limiting the (p)ppGpp production in þ N conditions. Note that the GlnD activity is very likely also inhibited by intracellular glutamine. On nitrogen starvation, intracellular pool of glutamine drops, relieving inhibition of EI Ntr autophosphorylation and thereby increasing HPr and EIIA Ntr phosphorylation levels (in green). Once phosphorylated, EIIA Ntr BP interacts with SpoT to inhibit its hydrolase activity (HD), whereas HPrBP regulates indirectly SpoT synthetase activity (SD). This dual control of SpoT by HPrBP and EIIA Ntr BP leads to (p)ppGpp accumulation, which in turn delays the G1-to-S and swarmer-tostalked cell transition.
cycle thanks to (p)ppGpp alarmone. It would be interesting to check if the PTS Ntr system is used by other a-proteobacteria to relay nitrogen starvation (glutamine deprivation) to (p)ppGpp production and subsequent G1 arrest. Only a few mechanisms triggering (p)ppGpp accumulation in nutrient-limiting conditions have so far been deciphered at the molecular level 8 . When E. coli cells are starved for amino acids, the (p)ppGpp synthetase RelA is directly activated by ribosomes whose A site is occupied by an uncharged tRNA 32 , whereas the bifunctional (p)ppGpp synthetase/hydrolase SpoT is regulated by an acyl carrier protein in response to fatty acid starvation 33,34 . In this report, we discovered a new molecular mechanism stimulating (p)ppGpp accumulation in response to nutrient starvation. This mechanism involves the PTS Ntr system as an important metabolic sensor that translates a glutamine deprivation signal into a (p)ppGpp accumulation signal. Our data suggest that EIIA Ntr BP directly reduces the hydrolase activity of SpoT, while HPrBP indirectly activates (p)ppGpp production upon nitrogen starvation (Fig. 8). Historically, the PTS system was discovered as a phosphorylation cascade involved in the regulation of sugar uptake and carbon catabolite repression 21,35 . Afterwards, a second phosphotransferase system (PTS Ntr ) was proposed to be connected to nitrogen metabolism but this connection remained poorly described 21 . The direct inhibition of EI Ntr autophosphorylation by glutamine observed in E. coli and S. meliloti 22,36 , as well as now in C. crescentus (Fig. 4), reinforces the idea that nitrogen constitutes a signal for PTS Ntr systems. The fact that the GAF domain, highly conserved in all EI Ntr proteins ( Supplementary Fig. 4b), is required for binding glutamine suggests that the glutamine-dependent control of EI Ntr phosphorylation might be a common feature in PTS Ntr system.
In contrast to its EIIA paralog, the EIIA Ntr component is not associated with permeases, but rather carries out regulatory functions, by interacting with its target(s). For example, the unphosphorylated form of EIIA Ntr inhibits pyruvate dehydrogenase activity in Pseudomonas putida by interacting with the E1 subunit 37 . Our work constitutes so far the first example of regulatory functions mediated by the phosphorylated form of EIIA Ntr (EIIA Ntr BP). Indeed, our results support the conclusion that only the phosphorylated form of EIIA Ntr interacts with SpoT to inhibit its hydrolase activity. This is further supported by the fact that EI Ntr and SpoT are found in the same protein complex during stationary phase 38 . Interestingly, a direct interaction between the non-phosphorylated form of EIIA Ntr and SpoT has been recently found in the b-proteobacterium Ralstonia eutropha by BTH but no function was assigned for this connection 39 . This differential interaction between phosphorylated or nonphosphorylated form EIIA Ntr and SpoT illustrates the evolutionary plasticity of PTS Ntr components with their targets.
Besides EIIA Ntr BP, we know that phosphorylated HPr also controls (p)ppGpp accumulation on nitrogen starvation, but how this regulation works at the molecular level remains an open question. HPrBP could interact with an unknown factor (X in Fig. 8), which in turn could modulate the abundance of SpoT or activate its synthetase activity, to subsequently increase the global (p)ppGpp pool. Although we have now uncovered the pathway that stimulates (p)ppGpp accumulation in response to nitrogen starvation, understanding how (p)ppGpp affects the G1-to-S transition at the molecular level will be a challenge for future work.

Methods
Bacterial strains and growth conditions. Oligonucleotides, strains and plasmids used in this study are listed in Supplementary Tables 1, 2 and 3, altogether with construction details provided in the Supplementary Methods. E. coli Top10 was used for cloning purpose, and grown aerobically in Luria-Bertani (LB) broth (Sigma) 40 . Electrocompetent cells were used for transformation of E. coli.
All Caulobacter crescentus strains used in this study are derived from the synchronizable wild-type strain NA1000, and were grown in PYE or synthetic M2 (20 mM PO 4 3 À , 9.3 mM NH 4 þ ; þ N) or P2 (20 mM PO 4 3 À ; À N) supplemented with 0.5 mM MgSO 4 , 0.5 mM CaCl 2 , 0.01 mM FeSO 4 and 0.2% glucose (M2G or P2G, respectively) media at 28-30°C. Glutamine (Q) was used at a final concentration of 9.3 mM. Growth was monitored by following the OD (600 nm) during 24 h, in an automated plate reader (Bioscreen C, Lab Systems) with continuous shaking at 30°C. Genes expressed from the inducible vanA promoter (P vanA ) was induced with 0.5 mM vanillate. Generalized transduction was performed with phage^Cr30 according to the procedure described in ref. 41. Motility was monitored on PYE swarm (0.3% agar) plates. Area of the swarm colonies were quantified with ImageJ software as described previously in ref. 42. Motility screen was performed on PYE swarm (0.3% agar) plates supplemented with glutamine (9.3 mM) during 3-4 days at 30°C. Genomic DNA of the candidates was then sequenced by the Illumina sequencing method (Beckman Coulter Genomics). Transpositional screen was performed with himar1 transposons on PYE plates as previously described in ref. 43. The exact positions of three himar1 insertion sites into the ptsP locus (Fig. 3a) have been determined by semi-arbitrary PCR. The presence of himar1 transposons into the ptsP locus was checked by PCR with primers (926 and 927) hybridizing upstream and downstream of ptsP. For E. coli, antibiotics were used at the following concentrations (mg ml À 1 ; in liquid/solid medium): ampicillin (50/100), kanamycin (30/50), oxytetracycline (12.5/12.5). For C. crescentus, media were supplemented with kanamycin (5/20), oxytetracycline (1/2.5) where appropriate. The doubling time of Caulobacter strains was calculated in exponential phase (OD 660 : 0.2-0.5) using D ¼ (ln(2) Á (T (B) À T (A) ))/(ln(OD 660(B) ) À ln(OD 660(A) )) and normalized according to the wild-type strain. E. coli S17-1 and E. coli MT607 helper strains were used for transferring plasmids to C. crescentus by bi-and tri-parental mating, respectively. In-frame deletions were created by using pNPTS138-derivative plasmids and by following the procedure described previously in ref. 44.
Flow cytometry analysis. DNA content was measured using fluorescenceactivated cell sorting (FACS). Cells were fixed in ice-cold 70% ethanol. Fixed samples were then washed twice in FACS staining buffer (10 mM Tris pH 7.2, 1 mM EDTA, 50 mM NaCitrate, 0.01% Triton X-100) containing 0.1 mg ml À 1 RNaseA and incubated at room temperature (RT) for 30 min. Cells were then collected by centrifugation for 2 min at 8,000g, resuspended in 1 ml FACS staining buffer containing 0.5 mM Sytox Green Nucleic acid stain (Life Technologies), and incubated at RT in the dark for 5 min. Samples were analysed in flow cytometer (FACS Calibur, BD Biosciences) at laser excitation of 488 nm. At least 1 Â 10 4 cells were recorded in triplicate for each experiment. Gate for cells in G1 phase was determined with a sample of wild-type cells treated with Rifampicin (2.5 mg ml À 1 ) for 3 h. Percentage of gated G1 cells of each strain was then normalized using gated G1 cells of the wild-type strain as reference.
Synchronization of cells. For synchrony, cells were grown in 200 ml of PYE (OD 660 B0.8), collected by centrifugation for 15 min at 6,000g, 4°C; resuspended in 60 ml of ice-cold 20 mM phosphate (PO 4 3 À ) buffer and combined with 30 ml of Ludox LS Colloidal Silica (30%; Sigma-Aldrich) 45 . Cells resuspended in Ludox was centrifuged for 40 min at 9,000g, 4°C. Swarmer cells, corresponding to the bottom band, were isolated, washed twice in ice-cold PO 4 3 À buffer and finally resuspended in prewarmed PYE media for growth at 30°C. Samples were collected every 15 min for western blot, microscopy and FACS analyses.
Light and fluorescent microscopy. All strains were imaged during exponential growth phase after immobilization on 1% agarose pads 41 . Microscopy was performed using Axioskop microscope (Zeiss), Orca-Flash 4.0 camera (Hamamatsu) and Zen 2012 software (Zeiss). Images were processed with ImageJ. Supplementary Movie 1 was done with Debut Video Capture Software.
Detection of intracellular (p)ppGpp levels. (p)ppGpp levels were visualized as described previously in ref. 13. Briefly, strains were grown overnight in PYE and then diluted for a second overnight culture in M5GG (low-phosphate medium M5G supplemented with 1 mM glutamate). Then, cells were diluted a second time in M5GG and grown for 3 h to reach an OD 660 of 0.4. Cells were split into two parts and washed twice with P5G-labelling buffer (M5G without NH 4 þ and with 12.2 mM NaCl and 3.9 mM KCl instead of Na 2 HPO 4 and KH 2 PO 4 ). In all, 1 ml of cells were then resuspended in 225 ml of P5G-labelling ( À N) or M5G-labelling ( þ N) supplemented with 25 ml of KH 2 32 PO 4 at 100 mCi ml À 1 and incubated for 2 h with shaking (450 r.p.m.) at 30°C. Then, samples were extracted with an equal volume of 2 M formic acid, placed on ice for 30 min and then stored overnight at À 20°C. All cell extracts were pelleted at 18,000g for 3 min and 6 Â 2 ml of supernatant was spotted onto a polyethyleneimine plate (Macherey-Nagel). Polyethyleneimine plates were then developed in 1.5 M KH 2 PO 4 (pH 3.4) at RT. Finally, TLC plates were imaged on a MS Storage Phosphor Screen (GE Healthcare) and analysed with Cyclone Phosphor Imager (PerkinElmer). For hydrolase experiments (Fig. 7d), cells were incubated 1 h in P5G supplemented with xylose (0.1%). Then, cells were washed twice with P5G-labelling and resuspended in P5G-labelling ( À N) supplemented with KH 2 32 PO 4 , xylose (0.1%) and with ( þ N) or without ( À N) glutamine (9.3 mM).
The b-galactosidase assays were performed as described in ref. 46. Briefly, 50 ml E. coli BTH strains cultivated overnight at 30°C in LB medium supplemented with kanamycin, ampicillin and IPTG (1 mM) were resuspended in 800 ml of Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 ) and lysed with chloroform. After the addition of 200 ml ONPG (4 mg ml À 1 ), reactions were incubated at 30°C until colour turned yellowish. Reactions were then stopped by the addition of 500 ml of 1 M Na 2 CO 3 , and absorbance at 420 nm was measured. Miller units are defined as (OD 420 Â 1,000)/(OD 590 Â t Â v), where, 'OD 590 ' is the absorbance at 590 nm of the cultures before the b-galactosidase assays, 't' is the time of the reaction (min) and 'v' is the volume of cultures used in the assays (ml). All the experiments were performed with three biological replicates and Miller units of the T25-X T18-ZIP combination were used as a blank and substracted.
Autophosphorylation levels of EI Ntr and EI Ntr L83Q . [ 32 P]PEP was prepared enzymatically as described previously in ref. 48. Briefly, 50 ml reaction solution containing 100 mM triethylamine/HCl pH 7.6, 15 mM KCl, 3 mM MgCl 2 , 165 mM PEP, 1 mM pyruvate, 5 mM ATP, 60 mCi [g-32 P]-ATP (5,000 Ci mmol À 1 ) and 40 units of pyruvate kinase (Sigma) were incubated at 30°C for 2 h. Phosphorylation assays were performed in 20 ml of solution containing 10 ml of proteins extract (containing EI Ntr or EI Ntr L83Q ), 25 mM Tris/HCl pH 7.5, 10 mM MgCl 2 , 1 mM DTT, glutamine (0, 2, 5 or 10 mM) and 0.5 ml of [ 32 P]PEP solution at 37°C for 30 min. Then, 5 ml of 5 Â SDS-PAGE loading buffer were added to the samples. Proteins were subjected to electrophoresis in a 10% SDS-polyacrylamide gel. SDSpolyacrylamide gels were then dried and imaged on a MP Phosphor system (Packard) and analysed with Cyclone Phosphor Imager (PerkinElmer). Analysis of radioactive spots reveals three bands at different size (B50, 80 and 90 kDa). The band corresponding to B80 kDa, absent in protein extracts from the DptsP strain, was determined as EI Ntr or EI Ntr L83Q .
In vivo phosphorylation of EIIA Ntr . Strains containing pXMCS2-3FLAG-ptsN were grown overnight in PYE supplemented with kanamycin and then diluted for a second overnight culture in M5GG (low-phosphate medium M5G supplemented with 1 mM glutamate) supplemented with kanamycin. Then, cells were diluted in M5GG with or without xylose (0.1%), and grown for 3 h to reach an OD 660 of 0.5. In all, 1 ml of cells were washed twice with P5G-labelling buffer (M5G without NH 4 þ and with 12.2 mM NaCl and 3.9 mM KCl instead of Na 2 HPO 4 and KH 2 PO 4 ). Cells were then resuspended in 225 ml of P5G-labelling, P5X-labelling (xylose 0.1%) or P5XQ-labelling (xylose 0.1%, glutamine 9.3 mM) supplemented with 25 ml of KH 2 32 PO 4 at 100 mCi ml À 1 and incubated for 2 h with shaking (450 r.p.m.) at 30°C. Samples were collected for 2 min at 12,000 r.p.m., resuspended in 50 ml of lysis buffer (50 mM Tris pH 7.0, 80 mM EDTA, 150 mM NaCl, 4% Triton X-100) and incubated for 2 min at 4°C. Then, 900 ml of low-salt buffer (50 mM Tris pH 7.0, 100 mM NaCl, 50 mM EDTA, 2% Triton X-100) were added and samples were collected for 15 min at 4°C. Supernatants were mixed with anti-FLAG M2 magnetic beads (Sigma) previously washed three times with TBS buffer and twice with low-salt buffer. Samples were incubated on a rotating shaker for 90 min at 4°C, and beads were washed once with cold low-salt buffer and twice with cold high-salt buffer (50 mM Tris pH 7.0, 500 mM NaCl, 50 mM EDTA, 0.1% Triton X-100). Magnetic beads were then resuspended in 20 ml of 3 Â SDS loading buffer and 5 ml of 0.5 mg ml À 1 of 3FLAG peptide (Sigma) were added to each sample. After 10 min incubation with shaking (1,300 r.p.m.), proteins were subjected to electrophoresis in a 12% SDS-polyacrylamide gel. SDS-polyacrylamide gels were then dried and imaged on a MP Phosphor system (Packard) and analysed with Cyclone Phosphor Imager (PerkinElmer). Band intensities were quantified with ImageJ software by using the WT ( À Xyl, À N) as the background.