The spermidine acetyltransferase SpeG regulates transcription of the small RNA rprA

Spermidine N-acetyltransferase (SpeG) acetylates and thus neutralizes toxic polyamines. Studies indicate that SpeG plays an important role in virulence and pathogenicity of many bacteria, which have evolved SpeG-dependent strategies to control polyamine concentrations and survive in their hosts. In Escherichia coli, the two-component response regulator RcsB is reported to be subject to Nε-acetylation on several lysine residues, resulting in reduced DNA binding affinity and reduced transcription of the small RNA rprA; however, the physiological acetylation mechanism responsible for this behavior has not been fully determined. Here, we performed an acetyltransferase screen and found that SpeG inhibits rprA promoter activity in an acetylation-independent manner. Surface plasmon resonance analysis revealed that SpeG can physically interact with the DNA-binding carboxyl domain of RcsB. We hypothesize that SpeG interacts with the DNA-binding domain of RcsB and that this interaction might be responsible for SpeG-dependent inhibition of RcsB-dependent rprA transcription. This work provides a model for SpeG as a modulator of E. coli transcription through its ability to interact with the transcription factor RcsB. This is the first study to provide evidence that an enzyme involved in polyamine metabolism can influence the function of the global regulator RcsB, which integrates information concerning envelope stresses and central metabolic status to regulate diverse behaviors.


Introduction 48
SpeG, a member of the Gcn5-related N-acetyltransferase (GNAT) family, is a 49 bacterial spermidine N-acetyltransferase that acetylates spermidine and spermine. 50 These polyamines are toxic to bacteria at high concentrations and acetylation 51 neutralizes this toxicity [1,2]. Studies indicate that SpeG plays an important role in 52 virulence and pathogenicity of many bacteria, which have evolved SpeG-53 dependent strategies to control polyamine concentrations and survive in their hosts 54 [3][4][5][6]. Kinetic and structural analyses have demonstrated that SpeG from both 55 Escherichia coli and Vibrio cholerae can acetylate spermidine [7][8][9]. These studies 56 also showed that SpeG from V. cholerae is an allosteric protein; when spermidine 57 binds to its allosteric site, SpeG exhibits a symmetric closed dodecameric structure 58 [7,9]. Finally, in the absence of spermidine binding, V. cholerae SpeG can adopt 59 a unique asymmetric dodecameric structure with an open conformational state 60 [10]. 61 During the course of this study, we found that SpeG also regulates the small 62 RNA rprA, whose transcription strictly requires the phosphorylated isoform of the 63 two-component response regulator RcsB [11,12]. The canonical two-component 64 signal transduction system is composed of two proteins. The first is a sensor 65 kinase that detects a signal and, in response, autophosphorylates a conserved 66 histidine residue using ATP as the phosphoryl donor. The second is a response 67 regulator that autophosphorylates a conserved aspartate residue using the 68 phosphorylated sensor kinase as the phosphoryl donor [for reviews, see [13][14][15]]. 69 A more complex variant of the basic two-component system is the phosphorelay, 70 such as the Rcs phosphorelay, which consists of five proteins (RcsC,RcsD,RcsF,71 IgaA, and RcsB). The first four proteins are involved in controlling the 72 phosphorylation status of the response regulator RcsB in response to diverse 73 extracytoplasmic stimuli. The phosphorylation status of RcsB is set by an ATP-74 dependent protein-protein interaction chain whose core consists of the cytoplasmic 75 membrane-associated sensor kinase/phosphatase RcsC and its cognate histidine Rcs phosphorelay components [17][18][19][20][21][22][23][24]. RcsB also can become phosphorylated in 82 response to central metabolic changes via the central metabolite acetyl phosphate 83 [25]. Both mechanisms (RcsC-dependent and acetyl phosphate-dependent) 84 regulate the phosphorylation status of RcsB and thus both control RcsB-dependent 85 processes, such as desiccation, flagellar biogenesis, capsule biosynthesis, and 86 cell division [16,[25][26][27][28]. 87 The Rcs phosphorelay is unusual, as the response regulator RcsB can form 88 both a homodimer and a variety of heterodimers. The homodimer activates 89 transcription of rprA [11,12,29], which encodes the small RNA regulator of the 90 stationary phase sigma factor RpoS, and represses transcription of flhDC, which 91 encodes the master regulator of the flagellar regulon [25,30,31]. To activate 92 synthesis of the capsular exopolysacchararide colanic acid, RcsB forms a complex 93 with a partner transcription regulator, RcsA, stabilizing the interaction between 94 RcsB and a specific DNA binding site, the "RcsAB box" [32,33]. RcsB also can 95 form protein-protein complexes with other partner transcription factors, including 96 GadE, RmpA, MatA, BglJ, and RflM; there is also evidence to suggest an 97 interaction with PhoP [34][35][36][37][38][39]. Because these protein-protein complexes form in 98 response to a variety of conditions, the Rcs system can mediate diverse responses 99 that contribute to biofilm formation, virulence, motility and antibiotic resistance in 100 pathogens [26][27][28][34][35][36]. 101 Biochemical and mass spectrometry analyses indicate that RcsB can become 102 N ε -lysine acetylated on multiple residues [29,[40][41][42]. Two mechanisms for N ε -103 lysine acetylation have been reported. One mechanism involves the direct 104 donation of the acetyl group from acetyl phosphate to a deprotonated lysine ε-105 amino group [41,43]. The other mechanism is enzymatic, relying on a lysine 106 acetyltransferase (KAT) to catalyze donation of the acetyl group from acetyl-107 coenzyme A (acCoA) to the ε-amino group of a lysine residue [44]. All known 108 bacterial KATs are members of the large family of GNATs [29,40,[44][45][46][47]. 109 One of our previous studies suggested that acetylation of RcsB diminished its 110 ability to activate rprA transcription in E. coli [29]. In an effort to identify a KAT that 111 might be responsible, we first screened 21 known E. coli genes that encode or are 112 predicted to encode GNATs, seeking those that inhibited rprA transcription. This 113 screen revealed that SpeG could inhibit rprA activity; however, we obtained no 114 evidence that SpeG functions as a RcsB lysine N ε -acetyltransferase. Instead, we 115 report here that SpeG can interact with RcsB through the latter's DNA binding 116 domain. Our findings represent the first evidence that the metabolic enzyme SpeG 117 can affect transcription by interacting with the response regulator RcsB. 118 119

120
SpeG regulates rprA promoter activity 121 While the GNAT YfiQ (also known as Pka and PatZ) can acetylate RcsB in vitro 122 [29,40], the yfiQ mutant does not affect RcsB acetylation [29]. Therefore, we 123 suspected another GNAT was responsible for RcsB acetylation and proceeded to 124 test a series of 21 known or putative GNATs. We overexpressed these GNATs and 125 measured their effect on PrprA-lacZ, a transcriptional fusion of the RcsB-126 dependent rprA promoter (PrprA) and the lacZ gene, which we had integrated as 127 a single copy into the chromosome of BW25113 to generate our reference strain 128 AJW3759 (Table 1) [12]. From this preliminary screen, we identified SpeG as an 129 inhibitor of PrprA activity. When SpeG was overexpressed from a plasmid in the 130 reference strain, PrprA activity was reduced compared to the vector control during 131 late exponential growth and during the transition into early stationary phase (OD > 132 1.0, Fig 1A, linear regression analysis t=-2.553, p=0.01472). When speG was 133 deleted, PrprA activity increased in the isogenic speG mutant compared to its wild-134 type parent (Fig 1B,   promoter activity versus WT was statistically significant (t=7.750, p= 8. 65E-12). 153 154

SpeG does not acetylate RcsB in vitro 155
Since SpeG belongs to the GNAT family of acetyltransferases known to 156 acetylate proteins, we also tested the hypothesis that SpeG regulates rprA 157 transcription by acetylating RcsB. To accomplish this, we used an in vitro 158 colorimetric enzymatic assay with purified recombinant proteins. This assay 159 measures the formation of product (CoA) indirectly via its reaction with 160 dithionitrobenzoic acid (DTNB) to produce the thioanion product thionitrobenzoate 161 (TNB 2-), which is monitored spectrophotometrically at 415 nm [7,54]. We 162 compared the acetylation activity of SpeG toward spermidine or RcsB. As 163 predicted, we detected SpeG acetylation activity on spermidine when acCoA was 164 present; however, we observed no change in RcsB acetylation status in the 165 presence of SpeG and acCoA (Fig 2). This result suggests that RcsB is not a 166 substrate for SpeG under the conditions we used to assay acetylation. The spermidine synthase SpeE transfers a propylamine from decarboxylated 178 S-adenosylmethionine to putrescine to form spermidine, which is both a substrate 179 and an allosteric activator of SpeG [7]. To explore the role of SpeE/spermidine in 180 SpeG overexpression-inhibited PrprA activity, we transformed a mutant that does 181 not synthesize spermidine (speE) and its WT parent with either the SpeG 182 overexpression plasmid or its vector control and monitored PrprA activity (Fig 3). 183 SpeG overexpression resulted in reduced PrprA activity in both the parental strain 184 (Fig 3, linear regression analysis t=-3.752, p=0.000282) and the speE mutant (Fig  185  3, linear regression analysis t=-3.470, p=0.000745). Furthermore, exposure of the 186 speE mutant to exogenous spermidine exerted no effect on PrprA activity whether 187 or not SpeG was overexpressed (S1 Fig). We conclude that SpeG can inhibit 188 PrprA activity regardless of SpeE/spermidine status. We next asked if SpeG overexpression-dependent inhibition of PrprA activity 211 requires the spermidine acetyltransferase activity of SpeG. We therefore 212 overexpressed SpeG Y135A, a predicted catalytically inactive SpeG variant, in the 213 parent (AJW3759). This tyrosine (Y) residue acts as a general acid during 214 substrate acetylation and has been shown to be critical for catalytic activity of many 215 GNAT homologs [55][56][57]. We found that the SpeG Y135A mutant retained the 216 ability to inhibit PrprA activity in the parent AJW3759 (Fig 3A, linear regression 217 analysis t= -2.456, p=0.015623). These results are consistent with a SpeG-218 dependent, but spermidine acetylation-independent mechanism of inhibition in WT 219

SpeG binds to RcsB through its C-terminal domain 222
Since SpeG does not appear to acetylate RcsB and its catalytic activity is 223 unnecessary for its ability to inhibit rprA transcription, we considered whether 224 SpeG inhibits RcsB activity through a physical interaction. We used SPR to 225 investigate whether SpeG and RcsB can form a complex. First, we immobilized 226 SpeG onto the SPR chip and evaluated whether full-length RcsB or its N-or C-227 terminal domains could bind to SpeG. Both full-length RcsB (Fig 4A) and its C-228 terminal domain (Fig 4B) bound to immobilized SpeG in a concentration-229 dependent manner. In contrast, the N-terminal domain of RcsB did not (Fig 4C). 230 These results suggest that RcsB binds to SpeG through its C-terminal domain. We 231 also performed the reverse experiment, assessing whether SpeG could bind to 232 immobilized RcsB or its domains, but we detected no signal (data not shown). 233 Perhaps RcsB binds to the chip in a manner that prevents interaction with SpeG. We next tested the effect of spermidine on the SpeG-RcsB interaction. To 254 accomplish this, we exposed the surface of the chip containing immobilized SpeG 255 to spermidine and then measured the SPR signal from binding the three separate 256 RcsB constructs (described in Materials and Methods; Fig 4D-F). By fitting the 257 sensograms to these data using the one-to-one binding model, we obtained KD 258 values of 128 and 281 μM for the RcsB full-length and its C-terminal domain, 259 respectively (Fig S2B-C). In contrast, we could not determine a KD for the RcsB 260 N-terminal domain due to a large chi-squared fitting value. Furthermore, we 261 conclude that the binding of the N-terminal domain to SpeG is weak because the 262 response signals obtained at concentrations greater than 100 μM were relatively 263 low (Fig 4F). On the basis of these data and those obtained in the absence of 264  (Fig 5). We also generated a phylogenetic tree using these 279 sequences to determine which RcsB homologs had the greatest sequence 280 similarity to its C-terminal domain and, therefore, propensity for interacting with 281 SpeG (S3 Fig). We found the most conserved RcsB C-terminal domain residues 282 across LuxR/FixJ-type homologs are S152, P153, K154, L167, V168, T169, R177, 283 S178, K180, T181, S183, S184, Q185, K186, K187, and D198. From our analysis, DNA-binding domain was colored by the degree of sequence conservation from 295 red (100% conserved residues) to blue (non-conserved residues). A search for 296 RcsB C-terminal DNA-binding domain homologs was done using the PSI-BLAST 297 server. From the list of 500 sequences against the non-redundant database a 298 random set of 30 sequences with identity from 98% to 40% were chosen. A multiple 299 sequence alignment for visualization of the sequence conservation with respect to 300 the three-dimensional structure was generated. 301 302

SpeG does not bind the LuxR/FixJ family member RcsA 303
To determine if binding to SpeG is specific for RcsB or if SpeG can bind in 304 vitro to other LuxR/FixJ transcriptional regulators that have C-terminal domains 305 similar to RcsB, we heterologously expressed and purified the E. coli RcsA 306 transcriptional regulator (an auxiliary partner with RcsB in a heterodimer that 307 interacts with a specific DNA site called the "RcsAB" box [60]) and tested RcsA 308 binding to SpeG by SPR. We found that SpeG does not bind to RcsA in the 309 absence of spermidine (S4 Fig), which highlights the binding specificity between 310 RcsB and SpeG. However, we cannot exclude that possibility that SpeG may 311 bind an RcsB-RcsA heterodimer or other LuxR/FixJ-type family members in the 312 presence or absence of spermidine. While RcsA has an HTH motif, its inability to 313 bind SpeG also suggests that other regions of RcsB within its DNA-binding 314 domain besides the HTH motif and/or its oligomeric state might be important for 315 the specificity of the SpeG-RcsB interaction. 316 317

318
We have presented evidence that the metabolic enzyme SpeG regulates 319 transcription from the rprA promoter. We also have shown that SpeG binds the 320 DNA binding domain of the transcription factor RcsB. We propose that this 321 interaction interferes with the ability of RcsB to activate transcription from the rprA 322 promoter. This represents the first report of a direct link between spermidine 323 metabolism and an envelope stress signal transduction pathway. 324 325

SpeG can inhibit rprA transcription through interactions with RcsB 326
We began this study because we had previously reported that N ε -lysine 327 acetylation regulates RcsB activity at the rprA promoter [29]. Since deletion of the 328 only known E. coli N ε -lysine acetyltransferase YfiQ had no obvious effect on the 329 acetylation state of RcsB [29], we screened the known and putative 330 acetyltransferases for regulators of rprA transcription and found that SpeG 331 inhibited rprA promoter activity: overexpression of SpeG reduced rprA promoter 332 activity (Fig 1A), while deleting speG relieved inhibition of the rprA promoter (Fig  333   1B). 334 Because we did not observe acetylation of RcsB by SpeG (Fig 2) and since we 335 did not find that SpeG activity could affect RcsB-dependent rprA inhibition (Fig 3), 336 we instead investigated the possibility of a physical interaction between RcsB and 337 SpeG. Indeed, SPR analysis showed that SpeG forms a complex with RcsB 338 through the RcsB C-terminal DNA-binding domain (Fig 4). We further report that 339 this interaction is specific, as we did not detect binding between SpeG and RcsB's 340 auxiliary transcription factor RcsA (57) (S3 and S4 Figs). These in vitro results 341 combined with the in vivo analysis support the hypothesis that SpeG and RcsB 342 interact and that the resulting complex impacts RcsB activity at the rprA promoter. 343 As rprA transcription absolutely requires RcsB, we did not test if SpeG affected 344 rprA transcription in an rcsB mutant. 345

Physiological implications of SpeG-RcsB interactions 347
It has been estimated that RcsB regulates 5% of the E. coli genome, 348 including but not limited to the colanic acid biosynthetic locus, the small RNA rprA, 349 and the operon that encodes FlhDC, the master regulator of flagellar biogenesis 350 [61,62]. The Rcs phosphorelay has also been implicated in regulating biofilm 351 formation and sensitivity to antibiotic-induced peptidoglycan damage [16,[62][63][64].

Bacterial strains, bacteriophage, and plasmids 373
All of the bacterial strains, bacteriophage, and plasmids used in this study are 374 listed in Table 1 To monitor the promoter activity of PrprA-lacZ, biological replicates were grown 396 aerobically at 37°C in TB7 overnight. The overnight cultures were diluted in fresh 397 TB7 to an OD600 of 0.05 and grown aerobically with agitation at 250 rpm at 37°C 398 until early stationary phase. At regular intervals, cells were harvested and stored 399 at 4°C in a microtiter plate. β-galactosidase activity was determined quantitatively 400 as described previously (26) using All-in-One β-galactosidase reagent (Pierce 401 Biochemical). Sterile TB7 was used as a negative control on each microtiter plate. 402 Promoter activity was monitored throughout growth and plotted against OD600.

Site-directed mutagenesis 408
Site-directed mutagenesis of SpeG to pCA24n-speG(Y135A) was conducted 409 in pCA24n-speG with the QuikChange Lightning Multi site-directed mutagenesis 410 kit (Agilent Technologies), in accordance with the manufacturer's instructions by 411 using the mutagenic primers SDMspeGY135A and SDMspeGY135A_as, as listed 412 in Table 1. 413

RcsB, RcsA and SpeG expression plasmids 415
Plasmids containing genes from Escherichia coli str. K-12 substr. MG1655 416 included the following: 1) full-length RcsB (NCBI accession code AAC75277, GI: RcsB and RcsA plasmids were grown at 37°C in a fermenter until the OD600 451 reached 0.8, whereupon they were induced with 0.6 mM IPTG. The RcsA 452 transformant was also exposed to 0.25% L-rhamnose. The RcsB constructs were 453 expressed at 25°C overnight, whereas RcsA was expressed at 22°C overnight. 454 The next day cells were harvested by centrifugation and resuspended in lysis 455 buffer (1.5 mM magnesium acetate, 1mM calcium chloride, 250 mM sodium 456 chloride, 100 mM ammonium sulfate, 40 mM disodium phosphate, 3.25 mM citric 457 acid, 5% glycerol, 5 mM imidazole, 5 mM beta-mercaptoethanol (BME), 0.08% n-458 dodecyl-beta-maltoside (DDM), 1 mM phenylmethylsulfonyl fluoride (PMSF), and 459 20 µM leupeptin) and homogenized. Cells containing the SpeG plasmid were 460 grown at 37°C in a benchtop shaker to an OD600 of 0.6, induced with 0.5 mM IPTG, 461 and expressed at 25°C overnight. Cells were harvested and resuspended in lysis 462 buffer (as stated above without PMSF and leupeptin) and sonicated. After 463 sonication, lysates were centrifuged and the supernatant was purified as follows. 464 The proteins were purified using an ÄKTAxpressÔ (GE Healthcare, 465 Piscataway, NJ) high-throughput purification system at 4°C. The crude extract was 466 covalently coupled with the surface NHS esters at a flow rate of 40 μL/min at room 503 temperature. To achieve saturation, two sequential injections of SpeG for 3 min 504 followed by 1.5 min of dissociation were performed. To block formation of residual 505 NHS esters, an ethanolamine solution was injected over the chip. To remove 506 weakly bound SpeG molecules, the chip was washed with running buffer 507 containing 10 mM HEPES at pH 8.3 and 100 mM sodium chloride. The instrument 508 was cooled and all SPR measurements were carried out at 4°C. All protein 509 solutions were prepared in running buffer. 160 μL of RcsB full-length (10, 21, 42, 510 53, 63 and 74 μM), RcsB C-terminal domain (45, 67, 91, 114, 136, and 159 μM), 511 or RcsB N-terminal domain (59, 118, and 176 μM) were injected sequentially over 512 the SpeG-chip with a flow rate of 40 μL min -1 for 30 sec followed by a 1.5 min rinse 513 and a 1 min dissociation. After each binding cycle, SpeG surfaces were 514 regenerated by injecting 0.5 M sodium chloride for 45 sec at a flow rate of 30 μL 515 min -1 and washed with running buffer. All analyte injections were performed in 516 duplicate. For each measurement, a background response recorded in the 517 reference cell was subtracted as well as the response from a blank injection with 518 the running buffer. 519 To investigate how spermidine affects binding interactions of SpeG to RcsB 520 and its individual domains, we used a Reichert4SR (Reichert Technologies, 521 Buffalo, NY) four-channel SPR system at the Keck Biophysics Facility. SpeG was 522 immobilized at a concentration of 46 μM onto cells 3 and 4 using the amine 523 coupling procedure described above. Cells 1 and 2 were used as reference cells. 524 All measurements were performed at 4°C. A solution containing 0.5 mM 525 spermidine in the running buffer was flowed over the surface of the immobilized 526 SpeG at 40 μL min -1 for 30 sec followed by a 1.5 min rinse and a 1 min dissociation. 527 The chip was then washed with running buffer until the SPR signal reached a 528 stable value. 160 μL of RcsB full-length (10, 21, 42, 53, 63 and 74 μM), RcsB C-529 terminal domain (23, 45, 91, 114, and 136 μM) or RcsB N-terminal domain (59, 530 118, 177, 236 and 295 μM) were injected sequentially over the SpeG-chip, as 531 described above, to monitor binding of RcsB constructs to SpeG in the presence 532 of spermidine. After each binding cycle of RcsB full-length and RcsB C-terminal 533 domain, SpeG surfaces were regenerated with an injection of 0.5 M sodium 534 chloride for 1.5 min at a flow rate of 30 μL min -1 and washed with the running buffer. To examine binding interactions between SpeG and RcsB's auxiliary partner 546 RcsA from E. coli, we used a four-channel SPR system at the Keck Biophysics 547 Facility following the amine coupling protocol, as described above. SpeG protein 548 at a concentration of 46 μM in 10 mM HEPES buffer at pH 8.3 containing 100 mM 549 sodium chloride was immobilized onto the chip. 160 μL of RcsA protein solution in 550 running buffer (21, 42, 64 and 85 μM) was injected consecutively over the SpeG-551 chip followed by regeneration and washing, as described above. A background 552 response and response from a blank injection that contained running buffer were 553 subtracted from each sensorgram to determine the actual binding response. Data 554 were processed using TraceDrawer software. 555 556

Linear Regression Analysis 557
To determine whether experimental results were statistically significant, a linear 558 regression was performed, comparing all experimental groups with their respective 559 vector controls. All of the regressions used were set up as follows: the calculated 560 rprA promoter activity was the response variable, the overexpressed plasmids or 561 mutant were the explanatory variable, and time was a random effect. OD was not 562 included as an effect on activity as it is already used in the calculation of activity. 563 Time as a random effect was chosen based on the question asked: Accounting for 564 the effects of time on activity does the experimental group in question significantly 565 affect overall rprA promoter activity? The significance threshold was set at 0.05. 566 The open source program R (version 3.3.2) and packages "lmerTest", "ggplot2", 567 and "moments" were used to visualize and analyze the data (76,77,78,79