Chemical arrays were employed to screen ligands for HtpG, the prokaryotic homologue of Hsp (heat-shock protein) 90. We found that colistins and the closely related polymyxin B interact physically with HtpG. They bind to the N-terminal domain of HtpG specifically without affecting its ATPase activity. The interaction caused inhibition of chaperone function of HtpG that suppresses thermal aggregation of substrate proteins. Further studies were performed with one of these cyclic lipopeptide antibiotics, colistin sulfate salt. It inhibited the chaperone function of the N-terminal domain of HtpG. However, it inhibited neither the chaperone function of the middle domain of HtpG nor that of other molecular chaperones such as DnaK, the prokaryotic homologue of Hsp70, and small Hsp. The addition of colistin sulfate salt increased surface hydrophobicity of the N-terminal domain of HtpG and induced oligomerization of HtpG and its N-terminal domain. These structural changes are discussed in relation to the inhibition of the chaperone function.
Hsp (heat-shock protein) 90 is a molecular chaperone widespread in bacteria and eukaryotes. It is a homodimer. Each monomer consists of three domains: an N-terminal domain, a middle domain and a C-terminal domain. Hsp90 is a very weak ATPase, and its N-terminal domain possesses an ATP-binding site . The middle domain harbours the major binding site for a protein substrate in the case of a cyanobacterial homologue . The C-terminal domain is essential for the constitutive dimerization [3,4]. The ATPase reaction causes large conformational/domain rearrangements of Hsp90 that are thought to drive structural changes of a substrate protein and its release [5–7]. Like other major molecular chaperones, Hsp90 can recognize and bind nonnative proteins, thereby preventing their non-specific aggregation .
In eukaryotes, Hsp90 plays a central role in cellular signal transduction pathways since it is essential for maintaining the stability and function of a number of signalling proteins, including steroid receptors and protein kinases such as Plk1 (Polo-like kinase 1), Raf-1, Akt and Cdk4 (cyclin-dependent kinase 4) [5,7,9–11]. In cancer cells, ‘client’ proteins that are chaperoned by Hsp90 are multiply mutated, chimaeric and overexpressed signalling proteins that promote cancer cell growth and/or survival [12–14]. Hsp90 inhibitors, by interacting specifically with a single target, Hsp90, cause simultaneous destabilization and eventual degradation of Hsp90 client proteins in multiple signalling pathways. Importantly, Hsp90 inhibitors kill cancer cells preferentially since, cancer cells keep Hsp90 with higher affinity for the inhibitors compared with Hsp90 in normal cells . The passage into neurodegenerative diseases has many similarities to malignant transformation . In fact, administration of an Hsp90 inhibitor/anticancer agent, 17-allylamino-17-demethoxygeldanamycin, also ameliorated polyglutamine-mediated motor neuron degradation by reducing the total amount of mutant androgen receptor . Thus Hsp90 has emerged as a promising target for drug discovery and treatment of cancer/neurodegenerative diseases. This is the reason that the discovery/development of small-molecule Hsp90 inhibitors has attracted great interest.
One of the most efficient methods to search ligands/inhibitors for a protein is a high-throughput screening with chemical arrays . We have developed a unique photo-cross-linking approach for immobilization of small molecules on solid surfaces in a functional-group-independent manner . With this approach, photo-cross-linked chemical arrays were constructed , and used successfully to screen novel ligands for a human protein . In the present study, we searched for ligands for the prokaryotic homologue of Hsp90, HtpG, from the freshwater cyanobacterium Synechococcus elongatus PCC 7942 using the chemical arrays.
HtpG has physicochemical/structural properties that are similar to its eukaryotic homologues [20,21]. Although eukaryotes require a functional cytoplasmic Hsp90 for viability under all conditions , bacterial HtpG proteins are dispensable under normal growth conditions and even under heat stress in heterotrophic bacteria [23–25]. However, we showed an indispensable role of HtpG for the survival of S. elongatus PCC 7942 under heat stress . Recently, we found that LR30 (30 kDa rod linker polypeptide of phycobilisome), the major light-harvesting apparatus of cyanobacteria and red algae [27,28], is an in vivo protein substrate for HtpG in the cyanobacterium . The linker polypeptide is highly unstable to aggregate. HtpG can interact with the polypeptide to suppress its aggregation at a molar ratio of one HtpG dimer to one LR30. These studies indicate that the cyanobacterium can be used as a model organism to study the role of Hsp90 in prokaryotes. In the present study, we screened ligands for HtpG and found that several cyclic lipopeptide antibiotics, including colistin and polymyxin B, interact physically with HtpG. These are cyclic positively charged peptide antibiotics linked to a fatty acid residue . They are active against selected Gram-negative bacteria including Pseudomonas aeruginosa and Enterobacter spp.  and have become important therapeutic agents in many medical centres [29,30]. We found that the above antibiotics changed the structure of HtpG to abolish its chaperone function. The results of the present study provide useful information about small molecules that regulate the function of HtpG as well as its eukaryotic homologue.
MATERIALS AND METHODS
MDH (malate dehydrogenase) from porcine heart and bovine α-crystallin were obtained from Sigma–Aldrich and Stressgen respectively. CSS (colistin sulfate salt) was obtained from Sigma–Aldrich. Other peptide antibiotics were obtained from NPDepo (RIKEN Natural Products Depository). As reported previously , the two major components of colistin and CSS were colistin A and colistin B, which differ only in their fatty acid chains (Figure 1C). According to HPLC–MS analysis, the sample of colistin A contained some amount of colistin B. HPLC–evaporative light-scattering detector analysis revealed the ratio of colistins A and B in the samples used in the present study was as follows: colistin (2:8), CSS (3:7) and colistin A (8:2). S. elongatus PCC 7942 and its htpG mutant strain were cultured photoautotrophically in BG-11 inorganic liquid medium as described previously . Unless indicated otherwise, cultures were grown at 30 °C under a light intensity of 30 μE/m2 per s.
Identification of the compounds binding to HtpG by chemical array screening
Purification of His6-tagged HtpG protein, its truncated derivatives, His6-tagged DnaK2 and His6-tagged LR30 from S. elongatus PCC 7942
Construction of a strain which overexpresses His6-tagged HtpG protein, its truncated derivatives, or His6-tagged LR30 from S. elongatus PCC 7942 and their purification methods were as described previously . Details of the construction of a strain which overexpresses C-terminally His6-tagged DnaK2 from S. elongatus PCC 7942  are available upon request from H. N.
High-throughput screening of HtpG-binding compounds using photo-cross-linked chemical arrays
The chemical arrays were prepared as described previously [18,19]. Solutions of the 6788 compounds (2.5 mg/ml in DMSO) from NPDepo chemical library were arrayed on to four separate photoaffinity-linker-coated glass slides with a chemical arrayer developed at RIKEN. The full-length HtpG was labelled by a fluorescence probe using an Alexa Fluor® 532 protein-labelling kit (Molecular Probes). The slides were incubated with TBST [Tris-buffered saline with Tween 20: 10 mM Tris/HCl (pH 8.0), 150 mM NaCl and 0.05% Tween 20] containing 1% (w/v) dried skimmed milk powder for 1 h at room temperature (25 °C). Then the slides were probed at 30 °C for 1 h with 100 μg/ml Alexa Fluor® 532-labelled HtpG in TBST containing 1% (w/v) dried skimmed milk powder or Alexa Fluor® 532-labelled HtpG denatured at 95 °C for 10 min as a negative control. The probed slides were scanned at a resolution of 10 μm per pixel with a GenePix 4100A scanner (Molecular Devices) by using the Cy3 (indocarbocyanine) channel (excitation wavelength of 532 nm and emission wavelength of 575 nm). The fluorescence signals were quantified with GenePix 5.0 software with local background correction. The images from two slides treated with active HtpG and denatured HtpG were shown using Photoshop 5.5 software (Figures 1A and 1B).
Binding assay between HtpG domains and small molecules by SPR (surface plasmon resonance)
The PGS (photoaffinity-linker-coated gold substrate) was prepared as described previously [33,34]. DMSO solutions of compounds (5 mg/ml each) were spotted in triplicate on to the PGS using a MultiSprinter automated spotter (Toyobo). The chip obtained was dried in vacuo to evaporate the solvent, and then irradiated at 365 nm under a UV transmission filter (Sigma-Koki) with a CL-1000L ultraviolet cross-linker (UVP). The irradiation energy on the chip amounted to 4 J/cm2. The chip was rinsed repeatedly with DMSO, water and ethanol. The compound-immobilized PGS was placed into a MultiSprinter SPR imaging instrument (Toyobo) and incubated with BSA (0.1%) in running buffer (10 mM Hepes/NaOH, pH 7.4, and 150 mM NaCl) for 5 min. After washing with the running buffer, 0.1 mg/ml of the protein samples [full-length HtpG, HtpGΔC (C-terminal domain-truncated HtpG), HtpG_N (N-terminal domain of HtpG) and HtpG_C (C-terminal domain of HtpG)] in running buffer were injected on to the discrete PGS surfaces at 0.1 ml/min and incubated for 10 min. All SPR experiments were performed at 30 °C. The SPR image and signal data were collected with an SPR analysis program (Toyobo). The SPR difference image was constructed by using a Scion Image program.
ATPase activity of HtpG was measured at 37 °C with an ATP-regeneration system as described previously , except that the concentrations of MgCl2 and ATP were 1 mM and 0.2 mM respectively.
To measure thermal aggregation of LR30 or MDH, each of the protein substrates was combined with HtpG, its truncated derivatives, DnaK2 from S. elongatus PCC 7942 or bovine α-crystallin dissolved in 50 mM Hepes/KOH buffer at pH 8.0 (total volume, 1 ml) pre-warmed at 45 °C for 3 min in a thermostatically controlled quartz cell. When thermal aggregation of LR30 was measured, 20 mM NaH2PO4/Na2HPO4 buffer (pH 6.4) was used instead of 50 mM Hepes/KOH buffer, and the final concentration of urea was set to 80 mM in the reaction mixture. LR30, which is very unstable, was solubilized, purified and kept in 4 M urea. Samples were incubated at 45 °C in the thermostatically controlled cell and their apparent absorbance at 360 nm was monitored continuously. The absorbance indicates light scattering due to aggregation of denatured polypeptides.
Measurement of bis-ANS (4,4′-dianilino-1,1′-binaphthyl-5,5′-disulfonic acid) fluorescence intensity
Fluorescence spectra of bis-ANS (Sigma–Aldrich) were measured at room temperature using a Hitachi 850 fluorescence spectrophotometer. The emission fluorescence was scanned from 400 to 600 nm with the excitation wavelength set at 390 nm. A sample contained 1 μM bis-ANS, 1 μM HtpG or HtpG_N, and various amounts of CSS in 50 mM Hepes/KOH buffer at pH 8.0 (total volume, 1 ml).
SDS/PAGE using a 12% acrylamide gel was performed as described previously . For non-denaturing PAGE, a protein sample was dissolved in 50 mM Tris/HCl (pH 6.8) containing 0.005% Bromophenol Blue and 10% glycerol, and then electrophoresed in a 10 or 12% acrylamide gel containing no SDS.
High-throughput screening of small molecules that bind to HtpG by chemical array analysis
To explore novel small molecules that regulate the function of HtpG of S. elongatus PCC 7942, we carried out the screening of small molecules that bind to HtpG using NPDepo chemical arrays of 6788 compounds immobilized in duplicate on glass slides by photoaffinity-cross-linking . The chemical arrays were treated with Alexa Fluor® 532-labelled HtpG protein or heat-denatured Alexa Fluor® 532-HtpG as a negative control. We identified small molecules binding to only active HtpG as hit compounds (Figures 1A and 1B, spots indicated by arrows). Among them are cyclic lipopeptide antibiotics that include colistin A (Figure 1C) and polymyxin B.
SPR imaging analysis to identify a domain of HtpG to which peptide antibiotics bind
To identify a domain of HtpG which interacts with the cyclic lipopeptide antibiotics, we performed SPR imaging analysis  with various constructs of truncated HtpGs . Whereas the full-length HtpG, HtpGΔC and HtpG_N strongly bound to CSS, colistin, colistin A and polymyxin B (Figures 2B, rows a–c), HtpG_C did not bind to any of these antibiotics (Figure 2B, row d). CSS and colistin are lipopeptide mixtures containing colistin A and colistin B as the major components (see the Materials and methods section). These results showed that these cyclic lipopeptide antibiotics bind to the N-terminal domain of HtpG specifically.
Detection of the interaction between HtpG or truncated HtpGs and cyclic lipopeptide antibiotics
Influence of peptide antibiotics on the ATPase activity of HtpG
Small molecules such as radicicol and geldanamycin bind to the Hsp90's N-terminal domain to inhibit its ATPase activity . Thus we examined the effect of the above peptide antibiotics on the ATPase activity of HtpG at 37 °C. CSS up to 100 μg/ml did not inhibit the activity of HtpG (2 nmol/ml) at all, whereas 20 μg/ml radicicol, a control, totally inhibited the activity (results not shown). Colistin, colistin A and polymyxin B at 10 μg/ml showed no effect on the activity.
Influence of peptide antibiotics on the chaperone function of HtpG
We have shown that both the N-terminal and the middle domains of cyanobacterial HtpG contain a binding site(s) for a substrate protein . Hsp90 has been shown to contain a chaperone site at the N-terminal domain  as well. Thus we investigated effect of the peptide antibiotics on the chaperone activity of HtpG since the antibiotics interacted with the N-terminal domain of HtpG.
One of the characteristic features of molecular chaperones including Hsp90 is their ability to suppress the aggregation of proteins. Protein aggregation is usually monitored by light scattering. In the present study, we monitored aggregation of LR30 from S. elongatus PCC 7942 by measuring the increase in the apparent absorbance at 360 nm. We have shown that LR30 is one of the in vivo substrates for the cyanobacterial HtpG . When 0.2 μM LR30 solution was heated at 45 °C, it denatured and formed aggregates (Figure 3A, ‘LR30’). As we reported previously , HtpG suppressed the thermal aggregation of LR30 to the maximum extent at a molar ratio of one HtpG dimer to one LR30 (Figure 3A, ‘LR30 + HtpG’). In the presence of CSS (10 μg/ml), the anti-aggregation activity of HtpG was inhibited (Figure 3A, ‘LR30 + HtpG + CSS’). Experiments with colistin, colistin A and polymyxin B (10 μg/ml) gave similar results (results not shown).
Influence of peptide antibiotics on the suppression of LR30 aggregation by HtpG as analysed by light-scattering assay
To examine whether CSS exerts an influence on pre-formed HtpG–LR30 complexes, CSS was added 150 s after the heat treatment (Figure 3B). This resulted in a sudden increase in the light-scattering intensity (Figure 3B, ‘LR30 + HtpG + CSS’), indicating that the soluble chaperone–substrate complexes decreased immediately. Jakob et al.  observed a similar phenomenon, a sudden increase in the light-scattering signal, when MgCl2 was added to pre-formed Hsp90–citrate synthase complexes. They explained that the increase was due to immediate release of the substrate protein from Hsp90 and its aggregation. This absorbance increase took place in a CSS-concentration-dependent manner. The aggregation process of LR30 itself was not affected by the presence of CSS (Figure 3B, ‘LR30 + CSS’), whereas the intensity of the light scattering of a solution containing HtpG increased slightly by the addition of CSS (Figure 3B, ‘HtpG + CSS’).
In the following studies, we focused on CSS among the antibiotics since it is commercially available. In addition, CSS is practically important since CSS as well as sodium colistin methanesulfonate are the two forms of colistin that are available for clinical use .
Influence of CSS on the chaperone function of the N-terminal and middle domains of HtpG
The SPR analysis showed that the cyclic lipopeptide antibiotics bind to the N-terminal domain of HtpG, although they may also bind to the middle domain (Figure 2). As described above, the middle domain also has a chaperone site. Thus we studied further whether the peptides affect only the chaperone function of the N-terminal domain, but not that of the middle domain by examining effect of CSS on the anti-aggregation activity of HtpG_N, HtpGΔC and HtpGΔN (N-terminal domain-truncated HtpG). To this end, we employed light-scattering assays to measure the chaperone function of these constructs (Figure 4).
Influence of CSS on the suppression of LR30 aggregation by HtpG_N (A), HtpGΔC (B) or HtpGΔN (C)
CSS showed much stronger inhibitory effect on the anti-aggregation activity of HtpG_N than that of the full-length HtpG. The activity was almost completely diminished in the presence of 10 μg/ml CSS (Figure 4A, compare ‘LR30 + HtpG_N’ with ‘LR30 + HtpG_N + CSS’), while the chemical at the same concentration was not enough to totally inhibit the anti-aggregation activity of the full-length HtpG (Figure 3B). The intensity of the light scattering of a solution containing HtpG_N increased slightly by the addition of CSS (Figure 4A, compare ‘HtpG_N’ with ‘HtpG_N + CSS’). Results shown in Figure 4(A) were obtained using 1.6 nmol of HtpG_N per 1 ml reaction mixture. The amount is higher than those used in other experiments whose results are shown in Figures 4(B) and 4(C). To rule out the concentration-dependent effect, experiments were repeated using 0.4 nmol of HtpG_N. The results (not shown) were basically similar to those shown in Figure 4(A).
The addition of CSS also almost completely abolished the chaperone activity of HtpGΔC (Figure 4B, compare ‘LR30 + HtpGΔC’ with ‘LR30 + HtpGΔC + CSS’). On the other hand, it showed no significant effect on the anti-aggregation activity of HtpGΔN (Figure 4C). HtpGΔN retains the middle and the C-terminal domain, but lacks the N-terminal domain. These results can be predicted if we assume that the peptide antibiotic binds to the N-terminal domain and affects the interaction of the target protein with the N-terminal domain of HtpG specifically.
Influence of CSS on the interaction of HtpG with a substrate protein other than LR30
We performed experiments similar to those shown in Figure 4 with a protein substrate for HtpG other than LR30. MDH has been used as a substrate for many chaperone assays. The protein solution was heated at 45 °C. The time course of the aggregation of MDH was monitored by the apparent increase in the absorbance at 360 nm. HtpG suppressed the thermal aggregation of the protein at a molar ratio of one HtpG dimer to one MDH monomer (Figure 5A, compare ‘MDH’ with ‘MDH + HtpG’). CSS (10 μg/ml) inhibited the anti-aggregation activity of HtpG greatly (Figure 5A, ‘MDH + HtpG + CSS’). The aggregation process of MDH itself was not affected by the presence of CSS (Figure 5A, ‘MDH + CSS’). Similar experiments with citrate synthase were also performed. CSS inhibited the anti-aggregation activity of HtpG, although higher concentration of CSS was necessary in order to detect a significant effect of CSS (results not shown). The results indicate that CSS inhibits the anti-aggregation activity of HtpG regardless of a kind of the protein substrate.
Influence of CSS on the suppression of thermal aggregation of MDH by HtpG (A), DnaK2 (B) or α-crystallin (C)
Influence of CSS on the chaperone function of DnaK2 and α-crystallin
We examined further whether CSS can inhibit the anti-aggregation activity of other molecular chaperones. DnaK2, the heat-induced one of the three DnaK homologues in S. elongatus PCC 7942 , completely suppressed the aggregation of MDH at a molar ratio of two DnaK2 monomers to one MDH (Figure 5B, compare ‘MDH’ with ‘MDH + DnaK’). In contrast with HtpG, the anti-aggregation activity of DnaK2 was not inhibited by 10 μg/ml CSS at all (Figure 5B, ‘MDH + DnaK + CSS’). CSS showed no effect on the intensity of the light scattering of a solution containing DnaK2 only (Figure 5B, ‘DnaK + CSS’).
α-Crystallin, a eukaryotic member of the small Hsp family  completely suppressed the aggregation of MDH at a molar ratio of two monomers to one MDH (Figure 5C, compare ‘MDH’ with ‘MDH + Crystallin’). The anti-aggregation activity of α-crystallin was not inhibited by 10 μg/ml CSS at all (Figure 5C, ‘MDH + Crystallin + CSS’), indicating the specificity of the interaction of colistin with the molecular chaperone HtpG.
Influence of CSS on the surface-exposed hydrophobicity of HtpG
It is generally thought that a molecular chaperone binds to a substrate protein via hydrophobic interaction. The results shown above indicate that the peptide antibiotics affect the interaction of HtpG with a substrate protein. Thus we examined whether they affect surface hydrophobicity of HtpG by measuring the fluorescence intensity of bis-ANS. The fluorescence intensity of bis-ANS greatly increases upon binding to hydrophobic sites and thus bis-ANS has been used widely to assess the hydrophobic surface of proteins .
CSS increased the fluorescence intensity of bis-ANS bound to HtpG (Figure 6A), indicating that CSS changes the structure/conformation of HtpG, which results in the increase in its surface hydrophobicity. The bis-ANS-binding analysis was repeated using HtpG_N. The results were basically similar to those obtained with full-length HtpG (Figure 6B), indicating that CSS binds to the N-terminal domain to change the conformation of the N-terminal domain.
Effect of CSS on the surface-exposed hydrophobicity of HtpG (A) and HtpG_N (B)
Influence of CSS on the oligomer state of HtpG
It is rather unexpected that CSS increases surface hydrophobicity of the full-length and N-terminal domain of HtpG since it inhibited the interaction between HtpG or its N-terminal domain and a protein substrate. We hypothesized that the increase in the hydrophobic region of HtpG or its N-terminal domain may enhance the interaction between themselves rather than interaction with a denatured substrate. This means that HtpG or its N-terminal domain may form higher oligomers or aggregates.
In order to test this, we examined the oligomer state of HtpG and HtpG_N by native PAGE which is often used to examine the oligomerization state of Hsp90. First, we analysed HtpG and HtpG_N that were incubated in the reaction mixture used for the chaperone assay. A 0.4 nmol amount of HtpG or HtpG_N per 1 ml solution was incubated at 4 or 45 °C for 15 min in the presence or absence of 10 μg of CSS. As shown in Figure 7(A), HtpG formed a 144 kDa dimer as a major oligomer form at 4 and 45 °C. CSS decreased the level of the HtpG dimer at 45 °C. Interestingly, regardless of the incubation temperature used, a ladder-like pattern of oligomerized HtpG_N was detected on native PAGE. Owing to the ladder formation, the staining intensity of each protein band was weak. However, it was still evident that many of the oligomers disappeared in the presence of CSS at 45 °C.
Effect of CSS on the oligomer/aggregation states of HtpG and its N-terminal domain (HtpG_N)
We carried out similar experiments using a high concentration of HtpG or HtpG_N in order to ascertain oligomer/aggregate formation. A 0.2 nmol amount of HtpG per 15 μl solution was incubated at 4 or 45 °C for 15 min in the presence or absence of 10 μg of CSS. Under the experimental conditions, HtpG formed higher oligomers when temperature increased from 4 to 45 °C (Figure 7B). This is consistent with previous studies with Hsp90 . CSS decreased the level of HtpG dimer/tetramer, and enhanced oligomerization of HtpG at 4 °C. The pattern appeared to be similar to that obtained with HtpG at 45 °C in the absence of the chemical. At 45 °C, the dimer decreased greatly and tetramer and higher oligomers were not detected. This is likely to be due to the formation of very-high-molecular-mass oligomers or aggregates of HtpG which could not enter into a separation gel of the PAGE. We detected slight precipitation of the HtpG aggregates at both 4 and 45 °C in the presence of CSS (Figure 7C). This result is consistent with a slight increase in light scattering by the addition of CSS to a solution containing HtpG only (Figures 3B and 5A).
A 0.4 nmol amount of HtpG_N per 13 μl solution was incubated at 4 or 45 °C for 15 min. A ladder-like pattern of oligomerized HtpG_N was clearly detected in these experiments (Figure 7D). Regardless of the incubation temperature, all of these oligomers disappeared in the presence of CSS. This is due to precipitation of the domain that was observed after addition of CSS (results not shown). CSS showed a much stronger effect on the structure of HtpG_N than that of HtpG.
Influence of CSS on the heat-sensitivity of S. elongatus PCC 7942
The wild-type S. elongatus PCC 7942 strain can grow at 45 °C, whereas its htpG mutant cannot . Thus HtpG plays an essential role for the thermal tolerance of the cyanobacterium. We hypothesized that CSS makes the wild-type strain high-temperature-sensitive like the mutant if it inhibits the function of HtpG in vivo. First, we searched for an appropriate concentration of CSS which is low enough not to cause strong negative effects on the cyanobacterial growth under normal conditions at 30 °C. CSS is known to have antibacterial activity [29,30]. We found that 15 μg/ml CSS did not inhibit growth of both the wild-type and the htpG mutant (Ω-1) strains for 24 h after the addition of CSS, although a slight inhibitory effect was detected after 48 and 72 h (Figure 8A). In order to test the above hypothesis, we measured growth of the wild-type at 45 °C in the presence of 15 μg/ml CSS. The wild-type growth in the presence of CSS as well as the htpG mutant growth in the absence of CSS was greatly inhibited at this temperature, whereas the wild-type could grow in the absence of CSS (Figure 8B).
Effect of CSS on the heat-sensitivity of S. elongatus PCC 7942 and its htpG mutant
In the present paper, we found that several cyclic lipopeptide antibiotics including colistin and polymyxin B bound to the N-terminal domain of HtpG physically (Figures 1 and 2). The binding did not affect the ATPase activity of HtpG, suggesting that the antibiotic-binding site in the N-terminal domain is not located at the ATP-binding pocket. However, the binding caused inhibition of the chaperone function of HtpG that suppresses thermal aggregation of a substrate protein (Figure 3). The antibiotics interacted with the N-terminal domain of HtpG specifically since neither they interacted with other domains of HtpG nor they influenced the chaperone function of other molecular chaperones such as DnaK2 (Hsp70) and small Hsp (Figures 4 and 5). CSS, one of the antibiotics, increased the surface-exposed hydrophobicity of HtpG and its N-terminal domain (Figure 6), and induced their oligomerization/aggregation (Figure 7). The binding of CSS to the N-terminal domain may lead to exposure of a hydrophobic oligomerization site in the N-terminal domain that is absent in the absence of CSS, resulting in ‘uncontrolled’ association of the N-terminal domain(s) of an HtpG dimer with that of the other dimer (see Figure 7D). The structural changes may disrupt the chaperone site(s) and cause the inhibition of the anti-aggregation activity of HtpG by the antibiotics.
Our study with CSS suggests a role of the N-terminal domain in the chaperone function of HtpG. Previously, we showed that both the N-terminal and middle domains of HtpG prevented the aggregation of denatured proteins, indicating the potential chaperone sites in these domains . The middle domain appeared to be the major one that interacts with a target, since the middle domain showed much higher activity in suppressing the aggregation of a target than the N-terminal domain. The results of the present study showed that CSS inhibited the chaperone function of the full-length HtpG, although it did not influence that of the middle domain (Figure 4C). Thus the N-terminal domain either by itself or together with the middle domain is involved in the chaperone function of the HtpG dimer.
Polymyxin B nonapeptide that is structurally related to polymyxin B except lacking an extension of aliphatic chain also interacted with HtpG to inhibit the chaperone function (results not shown). Furthermore, gramicidin S, a cyclodecapeptide, interacted with HtpG physically (results not shown). Thus we postulate that the circular peptide structure binds to the N-terminal domain of HtpG, and eventually inhibit the chaperone function of HtpG.
The influence of peptide antibiotics on the structure and function of HtpG is reminiscent of that of high temperature and bivalent cations, which affect the structure and function of Hsp90. Hsp90 self-oligomerized at high temperatures and acquired concomitantly a chaperone activity to bind/keep protein substrates in a folding-competent state . This high-temperature-induced oligomerization of Hsp90 is enhanced by molybdate, vanadate and Nonidet P40 which are thought to increase the surface hydrophobicity of the protein . Bivalent cations such as Mg2+ and Mn2+ also caused oligomerization/aggregation of both Hsp90 and HtpG . In contrast with the effect of high temperature, the change in the oligomerization state caused by the bivalent cations was accompanied by a rapid loss of the chaperone activity of Hsp90/HtpG to suppress the aggregation of heat-denatured substrates. Thus the influence of these bivalent cations on the structure and function of Hsp90/HtpG is quite similar to effect of the peptide antibiotics on those of HtpG.
As far as we know, this is the first report that shows modulation of structure and function of HtpG by lipopeptide antibiotics. Our studies also proved usefulness of the photo-cross-linked chemical array for ligand screening. We are searching for other novel ligands for HtpG as well as Hsp90.
colistin sulfate salt
C-terminal domain of HtpG
N-terminal domain of HtpG
C-terminal domain-truncated HtpG
N-terminal domain-truncated HtpG
30 kDa rod linker polypeptide of phycobilisome
photoaffinity-linker-coated gold substrate
surface plasmon resonance
Tris-buffered saline with Tween 20
Shun Minagawa and Keigo Sueoka performed the experiments except the chemical array analysis. Yasumitsu Kondoh and Hiroyuki Osada designed and conducted the chemical array analysis. Hitoshi Nakamoto directed the research and wrote the paper.
This work was supported in part by a grant-in-aid for Scientific Research (C) [grant number 21580083] to H.N. from the Ministry of Education, Science, Sports and Culture of Japan.