Localisation and protein-protein interactions of the Helicobacter pylori taxis sensor TlpD and their connection to metabolic functions

The Helicobacter pylori energy sensor TlpD determines tactic behaviour under low energy conditions and is important in vivo. We explored protein-protein interactions of TlpD and their impact on TlpD localisation and function. Pull-down of tagged TlpD identified protein interaction partners of TlpD, which included the chemotaxis histidine kinase CheAY2, the central metabolic enzyme aconitase (AcnB) and the detoxifying enzyme catalase (KatA). We confirmed that KatA and AcnB physically interact with TlpD. While the TlpD-dependent behavioural response appeared not influenced in the interactor mutants katA and acnB in steady-state behavioural assays, acetone carboxylase subunit (acxC) mutant behaviour was altered. TlpD was localised in a bipolar subcellular pattern in media of high energy. We observed a significant change in TlpD localisation towards the cell body in cheAY2-, catalase- or aconitase-deficient bacteria or in bacteria incubated under low energy conditions, including oxidative stress or respiratory inhibition. Inactivation of tlpD resulted in an increased sensitivity to iron limitation and oxidative stress and influenced the H. pylori transcriptome. Oxidative stress, iron limitation and overexpressing the iron-sulfur repair system nifSU altered TlpD-dependent behaviour. We propose that TlpD localisation is instructed by metabolic activity and protein interactions, and its sensory activity is linked to iron-sulfur cluster integrity.

generation of the unmarked tlpBC double mutant using the described procedure 5 , tlpA (HP0099) was inactivated by insertion of the aphA3 cassette as described before 5 . Isogenic allelic exchange insertion mutants in katA (HP0875), acxC (HP0697), acnB (HP0779) in H. pylori (strain N6) were created by allelic exchange mutagenesis in an analogous manner. Briefly, the gene sequences were amplified from genomic DNA using oligonucleotides adding restriction enzyme sites to the gene sequence (oligonucleotides are listed in Supplementary Table S3). The gene portions were cloned into the respective restriction sites of pUC18 or pUT18. Using either natural restriction sites or inverse amplification, the aphA3 cassette was introduced into the central portion of each gene 7 . The resulting plasmid constructs pCJ1306, pCJ1308, and pCJ537 were used in the transformation of naturally competent H.
pylori. Clones integrating the antibiotic resistance cassette into their respective chromosomal loci were selected on selective antibiotic plates. Correct introduction of the cassette was confirmed with PCR using at least two different primer combinations and excluding single cross-over recombination events.
Complementation of isogenic tlpD mutants with TlpD was accomplished either in cis or trans. The shuttle vector system of pHel2 (Cm R ) 8 was used to express TlpD from a plasmid. We have previously reported a TlpD-V5 fusion construct in the shuttle plasmid pHel2 (pCJ522) 5 . Cis complementation with a tlpD-V5 gene fusion in the chromosomal rdxA locus was described before 1 . In addition, strains combining allelic exchange mutants in other loci (see above) with one copy of tlpD-V5 in the rdxA locus were generated by recombination-mediated allelic exchange. These strains still express endogenous TlpD in addition to the V5-tagged version. To exclude effects of this specific expression approach for TlpD localisation and function, two control strains, with (N6 rdxA::tlpD-V5) and without (N6 tlpD rdxA::tlpD-V5) endogenous TlpD, were tested in parallel and showed similar results in all experiments (controls see Supplementary Fig. S1). To express TlpD as a C-terminal fusion with a hexa-histidine tag (Hisx6) in H. pylori, the complete pHel2 expression plasmid containing the TlpD-V5 construct (pCJ522) was reverse-amplified (pCJ522_His_fw/pCJ522_His-rev2), omitting the V5 tag sequence. Primer sequences were then designed to include the Hisx6 sequences, which, upon self-ligation of the PCR product, resulted in a circular plasmid carrying the potential tlpD promoter and an in-frame fusion of the complete tlpD gene, a short linker sequence as previously used for expressing a C-terminal MCP fusion 9 , and the Hisx6-tag (pCJ545). To enable self-ligation, the PCR product was incubated with T4 polynucleotide kinase (NEB, Bad Schwalbach, Germany).
To generate a Hisx6-TlpD expression plasmid for recombinant protein purification from E. coli, the complete tlpD gene (only leaving out the start codon) was amplified from strain 26695. The fragment was inserted into the pET28a expression vector using BamHI/NotI restriction sites, resulting in an N-terminal inframe fusion of the hexa-histidine tag with tlpD (pCJ1341). For an acnB-V5 expression construct for H. pylori, acnB and 160 nucleotides upstream of the start codon were amplified from H. pylori 26695. The V5 tag, its linker region and a pHel2 backbone were amplified from a previous construct pCJ607 5 in an inverse PCR, resulting in an opened plasmid. During both amplification processes, SpeI restrictions sites were added and allowed to generate a circular pHel2 vector with in-frame fusion of acnB, linker and C-terminal V5 tag (GKPIPNPLLGLDST) 9 (pCJ1314). Due to the cloning strategy, the linker sequence was slightly modified and elongated by one nucleotide triplet (linker coding for: GGLVSAAG) compared to previously used linkers 13 . To generate NifSU expression clones, the nifSU gene cluster and 300 bp of the upstream region were amplified from 26695 and cloned into pHel2 (pCJ1350). All bacterial strains and plasmids are listed in Table 2.
H. pylori clones that had taken up plasmids or incorporated DNA by recombination were selected on antibiotic plates. DNA of recombinants was validated as described above and the presence of proteins in knockout and expression clones, respectively, were verified by Western blots. The strains were further characterised for growth and motility in growth curves and live microscopy. The intact behavioural phenotype of the TlpD-V5 fusion in H. pylori was verified before 5 . We also confirmed that the functionality of TlpD-Hisx6 (in the N6 tlpD mutant) was maintained in comparison to non-tagged TlpD using single cell tracking analysis (data not shown).
All mutants that were chosen for further analyses did not show major differences in growth or motility compared to the parental strain.

Protein methods.
Standard procedures were followed for the determination of proteins amounts and separation of proteins on SDS gels (see also Supplementary Methods below for details of antibodies used for protein detection).
Pull-down assay for protein-protein interaction and mass spectrometric analyses.
Cleared cell lysates of H. pylori N6 tlpD::aphA3 (pHel2::tlpD-hisx6) and H. pylori N6 tlpD::aphA3 (negative control) were prepared for a protein pull-down approach. For this purpose, plate-grown bacteria were resuspended in RIPA lysis buffer (100 mM NaCl, 25 mM Tris/HCl pH 7.5, 20 mM imidazole, 10% glycerol, 1% Nonidet P-40, Complete Mini protease inhibitor cocktail (Roche) and lysed by incubation on ice for 15 min and subsequent sonication. Insoluble cell debris was removed by centrifugation (10,000 x g, 20 min, 4°C). The cleared lysates (about 700 µg of total protein containing about 3.5 µg of TlpD-Hisx6) were mixed with TALON matrix (immobilised Co 2+ ions, binding up to 5 µg of Hisx6-tagged protein per µl) (BD Biosciences, Erembodegem, Belgium), which was pretreated by washing three times with cold phosphate-buffered saline (PBS) (3,000 × g, 1 min, 4°C). Talon beads and bacterial lysates were co-incubated for 2 h at 4°C with gentle rotation. Affinity matrix and bound proteins were washed three times with RIPA buffer (see above) without Nonidet P-40 and imidazole, and two times with PBS (3,000 × g, 1 min, 4°C). Since we did not want to lose any of the native bead-bound material, we directly separated the final sediment of matrix and bound proteins on denaturing SDS gels after boiling in SDS sample buffer (including β-mercaptoethanol), and subsequently stained it with Coomassie blue. The pull-down assay was carried out in three independent replicates. The precipitated proteins were further analysed using mass spectrometry as follows. After visualising the captured proteins of N6 tlpD (TlpD-Hisx6) and N6 tlpD in SDS gels, protein bands that only occurred in N6 tlpD (TlpD-Hisx6) but not in N6 tlpD were cut out of the gels and subjected to mass spectrometry to identify potential TlpD-Hisx6 protein interaction partners. Before mass spectrometry, proteins were trypsin-digested over night 10 . For tandem mass spectrometry, peptides were dissolved in 2% ACN, 0.1% formic acid, and reverse phase chromatography using acetonitrile as an eluent was performed on a nanoACQUITY UPLC system (Waters) equipped with an analytical column (Waters, BEH130C18, 100 μm × 100 mm, 1.7 μm was -10*Log(P), and P was the probability that the observed match was a random event. Protein scores greater than 81 were considered significant (p<0.05). Since proteins from strain N6 that we used for the pull-down are not specifically included in the NCBInr database, we expected not to detect all possible peptides by theoretical mass, due to frequent amino acid differences between orthologous proteins of evolutionarily unrelated H. pylori strains. Two independent mass spectrometry analyses from two separate biological pull-down experiments were performed and analysed.
We purified recombinant Hisx6-TlpD (N-terminally fused Hisx6 tag) from E. coli BL21(DE3) upon expression from a pET28a based plasmid (pCJ1341). Briefly, bacteria were resuspended in lysis buffer (50 mM Tris/HCl pH 7.5, 150 mM NaCl, 2 mM DTT, 5 mM MgCl 2 ) and lysed by sonication (5 x 2 min, 4°C). After separation into soluble and insoluble fractions (9,000 x g, 20 min, 4°C), Hisx6-TlpD was purified from the TlpD-enriched insoluble fraction using urea-containing solubilisation buffer Packed Column (Macherey-Nagel, Düren, Germany). The column and proteins were washed three times (20 mM Tris/HCl pH 6.3, 100 mM NaH 2 PO 4 , 6 M urea) before the bound Hisx6-TlpD was eluted by decreasing the pH to 5.8 and finally to 4.5. Urea was removed by dialysing three times against 1 liter of lysis buffer (see above) in a Slide-A-Lyzer cassette (Thermo Fisher Scientific, Waltham, USA), in the absence of DTT, due to the large volume needed for the dialysis. In later steps, 2 mM DTT was added again to the buffer. We also established a non-denaturing purification protocol for TlpD-Hisx6, which used larger culture volumes and the same buffer but without urea. Elution, in this case, was performed by adding imidazole (100 mM) to the elution buffer. However the TlpD obtained by non-denaturing purification was much less soluble and less pure, therefore we did not use it for the protein interaction analyses. Native TlpD-V5 and AcnB-V5 were purified from H. pylori N6 tlpD (pHel2::tlpD-V5) and H. pylori N6 (pHel2::acnB-V5) soluble fractions using the V5tagged Protein Purification Kit (MBL International Corporation, Woburn, USA) according to the manufacturer's protocol. Briefly, soluble fractions were co-incubated with -V5 tag beads on a mixing wheel for 1 h at 4°C. After several washing steps with the supplied washing buffer (all centrifugation steps were carried out at 4,000 x g for 10 s), V5-tagged proteins were eluted upon addition of a V5 elution peptide.
The amount of purified proteins was determined in stained gels or Western Blots using -V5 antibody (monoclonal mouse; Invitrogen Life Technologies, Darmstadt, Germany) or -His antiserum (polyclonal rabbit; Rockland). The purity of the eluate was verified by SDS gels and Coomassie or silver staining. The purity of the proteins was estimated to be at least 95% pure or higher by silver staining. H. pylori catalase was recombinantly expressed and purified as previously described 11 .

Testing of direct interactions of TlpD and interaction partners using biolayer interferometry.
To test for direct protein-protein interaction, biolayer interferometry (BLI) was performed on the Octet RED96 System (ForteBio, Menlo Park, USA). BLI is an optical biosensing method that yields kinetic and affinity information in a manner similar to surface plasmon resonance (SPR) 12 . Purified recombinant H. pylori catalase (KatA) 11 and V5-purified native H. pylori aconitase (AcnB-V5) were used as interaction partners and analytes for sensor-coupled Hisx6-TlpD (ligand). Coupled sensors were first dipped into ForteBio kinetics buffer (PBS, 0.1% BSA, 0.02% Tween20 and 0.05% sodium azide, pH 7.4) to establish a stable baseline, then into different concentrations of purified proteins (AcnB or KatA, respectively, diluted in kinetics buffer) and finally in pure kinetics buffer again to monitor dissociation. Both A fixed exposure time specific for each dye was set to be able to compare different strains in some experiments.

RNA preparation from H. pylori and cDNA synthesis.
Bacteria were grown in liquid culture as indicated above to an OD 600 of ~1.0 and harvested by centrifugation at 22,000 x g for 1 min at 4°C. The cell pellet was immediately frozen at -80°C until further use. Total RNA was prepared using a modified RNeasy spin column protocol (Qiagen, Hilden, Germany) as described elsewhere 14 . The amount and purity of the isolated RNA was determined photometrically and on agarose gels. PCR was used to check for DNA contamination with the primer pair OLHPFlaA_4/OLHPFlaA_9 15 . To eliminate DNA contamination, DNase treatment with DNase I (Roche) and subsequent RNeasy column clean-up (Qiagen) were performed at least once.
In order to generate cDNA for microarray analyses, random hexamer primers (Invitrogen) were mixed with 15 to 25 µg of total RNA. After preincubation at 65°C for 10 min, and 10 min at room temperature, reverse transcription was carried out at 42°C for 2 h using the Superscript III system (Invitrogen).

Statistics.
Differences in motile behaviour between various H. pylori strains in temporal assays were statistically evaluated using Student's t test (unpaired, two-sided). Levels of significance for numbers of cells with polarly localised TlpD clusters were calculated using Fisher's exact test (Fig. 3C). The statistical evaluation of altered distribution of TlpD in intact bacterial cells was performed as follows: ImageJ densitometry 21 of TIFF images, obtained using the same exposure and lighting settings for each strain, was used to generate intensity histograms of TlpD for each cell (visualised with for average intensity profiles of the unfitted data). As an additional measure for differences in TlpD distribution between the polar and the cell body region, the maximum amplitude of intensity from the absolute maximum to the relative minimum within one cell was calculated (values shown in Table 3). The amplitude is indicated as a percentage of the absolute maximum value of each cell. For these calculations, the average intensity profile depicted in Supplementary Fig. S2 was used (Table 3) Supplementary Table S1 tlpD mutant including technical duplicates of each gene on each array were included to calculate the mean expression ratios (and standard deviation) of parental strain vs. tlpD mutant. Ratios equal or higher than 2.0 (wt/tlpD) are depicted. Significantly altered transcript amounts (SAM) are highlighted in bold. SAM settings: delta 0.32022, False Discovery Rate 5.3%, q value 0.0 to 5.3% (as for Table S1).  (A) TlpD-V5 was imaged by immunfluorescence microscopy in intact H. pylori cells (strain N6) with (N6 rdxA::tlpD-V5) or without (N6 tlpD::aphA3 rdxA::tlpD-V5) the native chromosomal copy of untagged TlpD. The former strain carries two chromosomal copies of TlpD in different genomic locations, while the latter strain was used as a control strain which has only one copy of TlpD (V5-tagged). Green: TlpD-V5, detected using α-V5 antibody (primary, mouse, at 1:1,000 dilution), combined with α-mouse IgG (coupled to Alexa-488, at 1:5,000); Red: SynaptoRed FM4-64 (1:5,000) membrane-intercalating dye as counterstain. All samples were prepared from 20 h plate-grown bacteria directly resuspended in fixing agent. (B) The subcellular distribution of TlpD-V5 (fluorescence intensity) in both strains was first quantitated as sum of pixel intensities along the longitudinal axis and then subdivided into polar and non-polar pixel intensities as percent of total intensities per strain (averaged over 30 bacteria per strain, Supplementary Methods). Statistical differences between both strains were calculated to be non-significant (Student's t test), except for the comparison of the polar intensity, which was moderately but significantly higher for the N6 rdxA::tlpD-V5 strain. (C) Quantification of polar TlpD clusters per cell for ≥ 30 bacteria per strain (in percent of evaluated bacteria of each strain). Levels of significance between the two strains were calculated by Fisher's exact test for clusters at 1 vs. 2 poles. ns: not significant. Four separate parameters: mean total TlpD-V5 fluorescence intensity, mean nonpolar intensity, mean non-polar intensity distribution of TlpD-V5, and occurrence and distribution of polar TlpD clusters, that were all averaged over 30 bacteria, were not significantly different between N6 rdxA::tlpD-V5 and the control strain N6 tlpD::aphA3 rdxA::tlpD-V5 (Student's t test and Fisher's exact test). Densitometry of Western blots of soluble and insoluble fractions also showed only minor differences in the distribution of TlpD-V5 between the two strains (not shown). Detection of TlpD in crude fractions (insoluble, soluble) of cell lysates of H. pylori strain N6 expressing TlpD or TlpD-V5 (chromosomal integration) and respective mutants in the same strain lacking potential TlpD interaction partners (see Supplemental Methods for crude fractionation and detection of proteins). Total lysates of tlpD mutant and N6 parental strain were used as controls. (A) TlpD was detected using α-TlpD or (B) α-V5 antisera in Western blot (red arrows point at monomeric TlpD and TlpD-V5 respectively). The green arrow in panel (A) points at a non-specific band. Fractionation controls were performed using (C) α-FlhA (membrane-associated), (D) α-AcnB (cytoplasmic), (E) α-HopZ-II (outer membrane, grey arrow), (F) α-KatA antisera. is: insoluble cell fraction (membrane-enriched), s: soluble cell fraction. Quantitation of TlpD protein band intensities was performed with ImageJ and normalised against common protein bands detected with antiserum against whole H. pylori bacteria. Quantification of TlpD in the reference strain showed a distribution of about 5% and 95% in the insoluble and soluble fraction, respectively. Total quantity and fraction distribution of TlpD were not significantly different between the different strains as determined by chi-square test and differed maximally by 9%.    Supplementary Tables S2 and S3, since their mean ratios were below the cut-off of two-fold regulated. In single biological array experiments, these genes showed more than twofold altered transcript amounts.

Fig. S5: Growth inhibition of H. pylori wild type and NifSU overexpression strain under conditions of iron-depletion.
Defined numbers of bacteria (strain N6) were plated on BHI/yeast + 5% horse serum plates and incubated with paper discs soaked with 10 µl 10, 20 and 40 mM 2,2-dipyridyl for iron depletion. The zones of growth inhibition around the discs were determined at 72 h of incubation. Results from one representative experiment (technical triplicates) are depicted. Significance levels were calculated with Student's t test (* p<0.05; ** p<0.01). Inhibition zones of the NifSU + strain are smaller than those of the parent under these assay conditions, indicating that the NifSU + strain is less susceptible to iron depletion.