Repurposing p97 inhibitors for chemical modulation of the bacterial ClpB–DnaK bichaperone system

The ClpB–DnaK bichaperone system reactivates aggregated cellular proteins and is essential for survival of bacteria, fungi, protozoa, and plants under stress. AAA+ ATPase ClpB is a promising target for the development of antimicrobials because a loss of its activity is detrimental for survival of many pathogens and no apparent ClpB orthologs are found in metazoans. We investigated ClpB activity in the presence of several compounds that were previously described as inhibitor leads for the human AAA+ ATPase p97, an antitumor target. We discovered that N2,N4-dibenzylquinazoline-2,4-diamine (DBeQ), the least potent among the tested p97 inhibitors, binds to ClpB with a Kd∼60 μM and inhibits the casein-activated, but not the basal, ATPase activity of ClpB with an IC50∼5 μM. The remaining p97 ligands, which displayed a higher affinity toward p97, did not affect the ClpB ATPase. DBeQ also interacted with DnaK with a Kd∼100 μM and did not affect the DnaK ATPase but inhibited the DnaK chaperone activity in vitro. DBeQ inhibited the reactivation of aggregated proteins by the ClpB–DnaK bichaperone system in vitro with an IC50∼5 μM and suppressed the growth of cultured Escherichia coli. The DBeQ-induced loss of E. coli proliferation was exacerbated by heat shock but was nearly eliminated in a ClpB-deficient E. coli strain, which demonstrates a significant selectivity of DBeQ toward ClpB in cells. Our results provide chemical validation of ClpB as a target for developing novel antimicrobials. We identified DBeQ as a promising lead compound for structural optimization aimed at selective targeting of ClpB and/or DnaK.


Introduction
Upon infection of a host, pathogens experience heat shock and oxidative stress, and their survival depends on molecular responses to these external conditions. The pathogen stress-response has emerged as a critical mechanism for the development of novel antimicrobials (1)(2)(3). However, no successful inhibition of pathogen stress-response machinery has been developed to date, mostly due to sequence conservation among heat-shock proteins (Hsps) across different domains of life.
The Hsp100 family offers a unique opportunity for inhibitor development among Hsps. The ATPdependent Hsp100 chaperones are essential for the survival of bacteria (where such proteins are called ClpB), lower eukaryotes (fungi and protozoa), and plants under stressful conditions. Unlike other heat-shock families, Hsp100s are not found in animals or humans (4,5). Hsp100 chaperones, in cooperation with Hsp70 and Hsp40 (bacterial DnaK and DnaJ), are uniquely responsible for reactivation of aggregated proteins in microbial cells (6)(7)(8).
Hsp100 chaperones are a sub-group of the AAA+ family of ATPases associated with different cellular activities (28). Like most AAA+ ATPases, Hsp100s form cylinder-shaped hexamers and use energy from ATP hydrolysis to induce conformational rearrangements in protein substrates (29). Hsp100-mediated reactivation of aggregates is linked to a forced extraction of polypeptides from aggregated particles and their subsequent unfolding by translocation through a narrow channel at the center of the Hsp100 hexamer (30,31). Small-molecule inhibitors have been recently developed for the human AAA+ ATPase, p97, a promising antitumor target (32)(33)(34), which led to clinical evaluation and validated the drugability of AAA+ proteins. We reasoned that inhibitors of AAA+ ATPases distantly related to Hsp100 might serve as prototype scaffolds for the development of Hsp100-selective ligands. In this study, we investigated the effects of several p97 inhibitor leads on the activity of bacterial ClpB. We discovered that one of the known p97 inhibitors, DBeQ is a promising candidate compound for targeting Hsp100 chaperones.

DBeQ inhibits the casein-activated ATPase activity of ClpB.
We investigated the ATPase activity of E. coli ClpB in the presence of three previously described p97 inhibitors: N 2 ,N 4 -dibenzylquinazoline-2,4diamine (DBeQ) (32), its p97-optimized derivative ML240 (35), and an alkylsulfanyl-1,2,4-triazole NMS-873 (36). We also tested two ClpB inhibitor candidates previously identified through a highthroughput ClpB interaction screen: C3 and C6 (37) (Supplementary Figure 1). None of the above compounds at 100 μM, except C3, showed a significant inhibition of the basal ClpB ATPase ( Figure 1A). However, only C3 and DBeQ strongly inhibited the ClpB ATPase in the presence of casein, a known pseudo-substrate of ClpB and an activator of the ClpB ATPase ( Figure  1A, B). We also identified DBeQ, but not NMS-873, as a candidate ClpB ATPase inhibitor in the presence of casein in the pilot screen of 388 bioactive compounds (Supplementary Figure 2). We therefore focused the following studies on DBeQ as a novel ClpB inhibitor candidate.
The solubility of DBeQ in aqueous buffers is limited (32), but we found that the DBeQ absorbance obeys the Beer-Lambert law and the solution turbidity does not significantly increase for concentrations of up to 100 μM DBeQ in the presence of 1% DMSO (Supplementary Figure 3). DBeQ suppressed the ClpB ATPase in the presence of casein to the basal level with an apparent IC50 ~ 5 μM ( Figure 1B), but did not inhibit the ATP-dependent binding of casein to ClpB ( Figure 1C). We introduced E-to-Q substitutions in the Walker B motif in each of the two ATP-binding domains of ClpB (D1 and D2, see Supplementary Figure 4) to disable ATP hydrolysis activity in each domain. Of the two ATP-binding domains of ClpB, only the Cterminal one (i.e. D2) showed casein-induced activation and was affected by DBeQ ( Figure 1D, E).
ClpB contains two Trp residues, W543 located between D1 and D2 and W462 in the coiled-coil middle domain (See Supplementary Figure 4). Remarkably, the ClpB variants with each of the two Trp residues substituted with Phe became sensitive to inhibition by DBeQ even in the absence of casein ( Figure 1F). The sedimentation coefficient distributions for wt ClpB and the W/F variants (Supplementary Figure 5) showed a dominant ~13-15S molecular species, which corresponds to a hexameric ClpB in equilibrium with monomers (38). Importantly, no shifts of the sedimentation coefficient distribution maximum towards a lower s20,w were observed in the presence of DBeQ (Supplementary Figure 5), which indicates that DBeQ does not induce dissociation of the ClpB hexamers. Overall, the different parameters tested above showed that DBeQ directly affects ClpB, but does not compromise its structural integrity, with the basal ATPase unaffected by DBeQ even at 45 ºC (Supplementary Figure 6).

DBeQ inhibits the aggregate-reactivation activity of ClpB/DnaK.
We tested the effect of DBeQ on the reactivation of aggregates prepared from two model substrates: firefly luciferase and bacterial glucose-6phosphate dehydrogenase (G6PDH), whose native activity was not affected by DBeQ (Supplementary Figure 7). Either substrate can be reactivated by DnaK/DnaJ/GrpE (KJE), albeit with a significantly lower efficiency than by ClpB with KJE ( Figure 2A, C), which allowed us to test the effect of DBeQ on both components of the ClpB/DnaK bi-chaperone system. Unexpectedly, we found that DBeQ inhibited both ClpBdependent and ClpB-independent reactivation of aggregated luciferase with a similar IC50 value ( Figure 2B), which suggests that the rate of luciferase reactivation is limited by the activity of KJE. For aggregated G6PDH, however, ClpBdependent reactivation was inhibited by DBeQ with a several-fold lower IC50 than ClpBindependent reactivation by KJE alone ( Figure  2D). These results indicate that DBeQ inhibits the activity of not only ClpB, but also DnaK. However, we did not detect inhibition of the DnaK ATPase activity by DBeQ (Supplementary Figure  8A) or suppression of DnaK binding to a substrate-mimicking peptide (Supplementary Figure 8B).

DBeQ interacts with ClpB and DnaK.
We used surface-plasmon resonance (SPR) to verify a direct interaction between DBeQ and ClpB ( Figure 3) or DnaK (Supplementary Figure  9). The binding isotherms for DBeQ and ClpB showed positive cooperativity ( Figure 3B) with an apparent Kd ~ 60 μM and a Hill coefficient ~ 2.3 (Supplementary Table 1). The DBeQ binding isotherms were virtually identical for wt ClpB and its two W/F variants ( Figure 3). In contrast to ClpB, the DBeQ interaction with DnaK was noncooperative (Supplementary Figure 9A-C) with an apparent Kd ~ 100 μM (Supplementary Table 1). The above apparent binding affinity of DnaK for DBeQ should be considered approximate, as it is close to the ligand solubility limit.
We determined that DBeQ interacted with the isolated nucleotide-binding domain (NBD) of DnaK with a similar affinity as with the full-length protein (Supplementary Figure  9D-F, Supplementary Table 1). As observed before (39), the ATPase activity of the DnaK NBD is similar to that of the full-length DnaK (Supplementary Figure 8A). We found that DBeQ did not inhibit the ATPase activity of the isolated NBD of DnaK (Supplementary Figure 8A).

DBeQ suppresses the proliferation and survival of E. coli.
ClpB is produced in E. coli cultured at 37 ºC and is strongly upregulated during heat shock (Supplementary Figure 10). The ΔclpB strain, which does not produce ClpB (40) does not show growth defects at 37 ºC or during a mild heat shock at 45 ºC, but rapidly loses viability at 50 ºC ( Figure 4B, D, F). DBeQ inhibited the growth of E. coli at 37 ºC in a concentration-dependent manner ( Figure 4A), but the DBeQ effects were eliminated in the ΔclpB strain ( Figure 4B). The viability of E. coli at 45 ºC (but not the ΔclpB strain) was strongly inhibited by DBeQ ( Figure  4C, D and Figure 5). Upon expression of a ClpB-YFP fusion protein in the ΔclpB strain (for the experiments in Fig. 6), the cells' susceptibility to DBeQ was restored and was exacerbated at higher levels of ClpB expression (Supplementary Figure  11). A critical role of ClpB in E. coli survival under severe stress manifested at 50 ºC, as shown by a loss of viability of the ΔclpB strain even without DBeQ ( Figure 4F). E. coli with an intact clpB gene survived the heat-shock at 50 ºC for several hours, but lost viability at the lowest DBeQ concentrations tested ( Figure 4E). These experiments demonstrate that a treatment of E. coli with DBeQ is toxic to the cells, but the DBeQ toxicity at a moderate stress of 45 ºC is more severe than that caused by a loss of ClpB in the ΔclpB strain.
To test the effects of DBeQ on the cross-talk between ClpB and DnaK, we investigated the effects of DBeQ on viability of two E. coli strains with defective DnaK: dnaK103, which produces a truncated inactive DnaK that does not bind to its substrates (41,42) and dnaK756, in which substrate release from DnaK is inefficient (43). We found that DBeQ suppressed viability of dnaK103, but surprisingly, viability of dnaK756 was not affected by DBeQ ( Figure 5).
To further explore the effects of DBeQ on the function of ClpB in E. coli cells, we tracked the ClpB localization with confocal fluorescence microscopy using a ClpB construct with a Cterminally-fused YFP. As demonstrated earlier (42,44), heat-induced aggregates of thermosensitive proteins, such as Photinus pyralis luciferase accumulate at the poles of E. coli cells (Supplementary Figure 12). ClpB and DnaK colocalize with the aggregates (ref. (44) and Figure  6). Importantly, the absence of a functional DnaK in the dnaK103 strain prevents ClpB from accumulating with the aggregates (Figure 6 and ref. (42)), but in the presence of DBeQ, ClpB colocalizes with the aggregates even in dnaK103 ( Figure 6). The heat-shock-induced localization of ClpB at the cellular poles in the dnaK756 strain with or without DBeQ was indistinguishable from that in wt E. coli (Supplementary Figure 13). The results with dnaK103 E. coli show that DBeQ affects localization of ClpB in cells exposed to heat shock. The fluorescent foci observed in E. coli in the presence of DBeQ are more diffuse than those found in the absence of the ligand, an effect exacerbated in the dnaK103 strain (Fig. 6). Since the foci localization was attributed to nucleoid occlusion (44), an altered foci appearance in the presence of DBeQ may reflect a modified size of the nucleoid-free space in cells whose proliferation was challenged with DBeQ and/or a loss of a functional DnaK.

Discussion
We hypothesized that small-molecule ligands of one AAA+ ATPase, p97, might also interact with another one, ClpB. We tested three inhibitors that show an increasing potency towards p97 (45): DBeQ (p97 IC50~3 μM), ML240 (p97 IC50~0.1 μM), and NMS-873 (p97 IC50~0.02 μM). We discovered that among the three ligands, only DBeQ, the least potent one towards p97, affected ClpB ( Figure 1) with an apparent IC50 close to that displayed towards p97. ML240 is a p97-optimized DBeQ derivative and its loss of potency towards ClpB suggests that a structural diversity among AAA+ ATPases is sufficient to allow discrimination between different ligands from the same chemical family. Thus, it may be feasible to orthogonally modify DBeQ and achieve an enhanced selectivity towards ClpB with a lower potency towards p97 and the other AAA+ ATPases.
Antimicrobial activity of 2,4-diaminoquinazolines has been reported before (46,47), but without a clear identification of their cellular targets. We have now shown that ClpB is the main target of DBeQ in E. coli under both permissive conditions and during heat stress ( Figure 4 and 5). It is remarkable that the DBeQ-induced inhibition of E. coli viability depends on production of a single protein, ClpB. While the ΔclpB strain does not respond to DBeQ, it becomes susceptible upon expression of ClpB (Supplementary Figure 11), which indicates that ClpB is required for the sensitivity of E. coli towards DBeQ. p97, the known DBeQ target, is not produced in bacteria. Interestingly, eukaryotic parasites, including Plasmodium and Leishmania produce p97 as well as ClpB. An anti-parasitic activity of DBeQ has been reported (48,49) and it remains to be determined if the compound's preferred in vivo target is p97 or ClpB.
As shown by SPR, DBeQ binds to ClpB with a positive cooperativity (Hill coefficient >2), which suggests multiple binding sites ( Figure 3, and Supplementary Table 1). Detection of multiple binding sites for DBeQ is consistent with the oligomeric structure of ClpB. The apparent IC50 for DBeQ is an order of magnitude lower than the apparent Kd ( Figure 1, 2, 3, and Supplementary Table 1), which suggests that partial saturation of the DBeQ sites in the ClpB hexamer is sufficient for a strong inhibition. Indeed, it has been observed that incorporation of a single inactive subunit into a hexameric ClpB blocks aggregate reactivation (50).
ClpB-mediated reactivation of protein aggregates depends on several molecular processes: the ClpB hexamer assembly, ATP-dependent substrate binding, DnaK-dependent substrate engagement, and ATP-hydrolysis-dependent substrate unfolding/translocation (4). Importantly, there is an allosteric linkage between substrate binding and the ATPase engine of ClpB, as demonstrated by activation of the ATPase in the presence of substrates (7). We found that DBeQ does not inhibit the ClpB hexamer assembly, substrate binding, or the basal ATPase of ClpB, but it does affect the linkage between substrate binding and the ATPase of the C-terminal D2 module ( Fig. 1). Indeed, whereas substrate binding primarily occurs at the N-terminal region of ClpB (51,52), a partial insertion of the substrate into the ClpB channel and its contact with the D2 module have been observed even without ATP hydrolysis (53). Moreover, D2 is fully responsible for the caseininduced activation of the ClpB ATPase because the activation is lost upon disabling D2 ( Figure  1E).
DBeQ-mediated modulation of the allostery within ClpB is further demonstrated by the apparent uncoupling of the DBeQ effects from substrate binding in the two Trp-to-Phe ClpB variants ( Figure 1F). Trp462 is located within the mobile coiled-coil domain of ClpB (Supplementary Figure 4), which controls the ATPase activity of the ClpB hexamer (54,55). Trp543 is located at the interface between D1 and D2 (Supplementary Figure 4). Both Trp462 and Trp543 side chains are exposed at the inter-subunit interface within the ClpB hexamer (Supplementary Figure 14). A similarity between the DBeQ binding isotherms for wt ClpB and its W/F variants ( Figure 3 and Supplementary Table 1) indicates that the W/F substitutions do not produce additional DBeQ binding sites in ClpB. Altogether, the enhanced sensitivity of the W/F ClpB variants towards DBeQ suggests that the ligand binding site(s) may be located between the hexamer subunits and in the vicinity of the D1/D2 junction.
We found that aggregate reactivation with the ClpB/DnaK bi-chaperone system becomes inefficient in the presence of DBeQ ( Figure 2). Since the energy generated from ATP in the AAA+ engine is directly coupled with substrate translocation/unfolding (56,57), our results suggest that DBeQ affects a linkage between the ClpB ATPase and substrate translocation by suppressing the acceleration of ATPase in response to a substrate, which decelerates substrate translocation and makes its reactivation inefficient.  Figure  8A). Our results are consistent with a recent report on the binding of amino-quinazolines within the ATP site of Hsp70 (58), but may also indicate a nonspecific interaction due to a predominance of hydrophobic groups in DBeQ. Interestingly, C3 also inhibits KJE (37) as well as ClpB and even luciferase ( Figure 1A, Supplementary Figure 7A). These results demonstrate a challenge in designing selective inhibitors for different families of molecular chaperones.
In vitro binding of DBeQ to DnaK notwithstanding, ClpB appears the main target of DBeQ in E. coli, because the compound's effects manifest even in the absence of functional DnaK in the dnaK103 strain, but do not manifest in the ΔclpB strain ( Figure 5). Remarkably, at 45 ºC, inhibition of ClpB with DBeQ is more detrimental for bacterial survival than a lack of ClpB in the ΔclpB strain. Thus, the phenotype of chemical inhibition of ClpB transcends a loss of function, at least under moderate stress.
Toxicity of ClpB in bacteria and its orthologue Hsp104 in yeast has been observed for "hyperactive" protein variants with mutations within the coiled-coil middle domain (55,59,60). However, unlike the DBeQ-treated ClpB, the toxic hyperactive variants display elevated ATPase and disaggregase activities.
The apparent toxic gain-of-function of the DBeQinhibited ClpB in bacterial cells could be blamed on DBeQ-induced aggregation of ClpB. However, we have not detected a propensity of ClpB to misfold and/or aggregate in the presence of DBeQ in vitro (Supplementary Figure 5), which is also supported by a lack of DBeQ effects on the basal ATPase activity of ClpB (Figure 1 and Supplementary Figure 6). Moreover, the survival of dnaK756 E. coli upon treatment with DBeQ does not support a premise that the compound's toxicity in cells is linked to misfolding or aggregation of its main target, ClpB because in such a case a viability rescue should not be expected in a strain with a nonproductive DnaK chaperone (43).
Chaperone-deficient strains of E. coli, ΔclpB and dnaK103 survive moderate heat stress of 45 ºC thanks to other protein quality control factors: chaperones and proteases whose highly promiscuous and redundant activities are sufficient for maintaining proteostasis in those strains ( Figure 5). Since DBeQ apparently targets ClpB and no other chaperones in E. coli, the ΔclpB strain is therefore resistant to DBeQ under moderate stress conditions. Physiologically, DnaK recruits ClpB to the aggregates and hands the substrates over to ClpB for disaggregation. In wt E. coli cells, a recruitment of ClpB to the aggregated proteins located at the cell poles strictly depends on DnaK (42). A striking polar localization of ClpB in dnaK103 cells in the presence of DBeQ suggests that the compound allows ClpB to overcome a deficiency of its recruiter co-chaperone ( Figure 6). However, ClpB interactions with protein aggregates become nonproductive in the presence of DBeQ (Figure 2). A nonproductive binding of ClpB to protein aggregates may become dominant and could suppress viability of E. coli because it may hinder access of other protein quality control factors to their aggregated substrates and irreversibly disturb cellular proteostasis. The above explanation of the DBeQ-induced phenotype in E. coli is corroborated by an unexpected rescue of cellular viability in the dnaK756 strain ( Figure 5). The DnaK756 variant binds to aggregates, but cannot release them (43). Apparently, a substrate-trapping capability of DnaK756 counteracts the nonproductive interactions of ClpB with the aggregates in the presence of DBeQ and is sufficient to preserve cellular viability. Thus, the apparent gain-offunction effect of DBeQ may be due to disturbing the balance of cellular proteostasis by stimulating nonproductive interactions of ClpB that suppress effectiveness of the remaining chaperones. Importantly, regardless of its exact mode of function in bacterial cells, DBeQ is a potent and highly selective molecular probe that targets and disrupts protein quality control in E. coli.
Molecular chaperones have not been previously explored as targets for novel antimicrobials. In this work, we demonstrated that the AAA+ disaggregase ClpB can be selectively targeted with a small-molecule ligand in bacterial cells and that such a treatment could produce a loss of bacterial viability. This result is significant because mammalian cells do not contain orthologues of ClpB (5), whereas many pathogenic microorganisms require the ClpB activity for infectivity and survival. We have also shown that due to a complexity of the cellular protein control machinery, understanding the mechanism of action of chemical inhibitors and the cellular phenotypes they produce requires multiple orthogonal biochemical and biological tests, such as those employed in this study.

DNA constructs
The nucleotide sequence encoding the full-length E. coli DnaK (residues 1-638) was amplified from plasmid pTTQ19dnaK + (61) using the primers: DnaK-NcoI (5'-ATATACCATGGGTAAAATAATTGGTATCG  ACC-3') and DnaK-XhoI (5'-TATATCTCGAGTTTTTTGTCTTTGACTTCTT C-3'), where the engineered restriction sites are underlined. The PCR product did not contain the dnaK STOP codon. Subsequently, the dnaK sequence was cloned into the NcoI/XhoI restriction site of pET28a. The STOP codon was introduced between the dnaK coding sequence and the Cterminal His-tag sequence in pET28a by sitedirected mutagenesis with the primers: and DnaK_STOPr (5'-GGTGGTGGTGGTGGTGCTCTTATTTTTTGT CTTTGAC-3'), where the STOP codon is underlined. The final construct (plasmid pET28-DnaK) was verified by DNA sequencing and was used for production of DnaK.
Single transformants were used to prepare overnight culture in LB media supplemented with kanamycin (50 µg/ml). The overnight culture was diluted 50-fold in 1 l of LB with kanamycin and incubated at 30 °C. When absorbance at 600 nm reached 0.4, IPTG was added to 1 mM and the culture was incubated for 3 h at 30  Fractions containing NBD with a purity greater than 95% were collected and concentrated to ~5 mg/ml. Protein concentration was determined by the Bradford method. The aliquots of NBD were stored at -20°C.

Bacterial strains
The following strains of E. coli were used: MC4100, MC4100ΔclpB::kan (40), dnaK103 (41), and dnaK756 (66). Candidate inhibitor compounds DBeQ, ML240, and NMS-873 were purchased from Sigma-Aldrich (St. Louis, Missouri) and used directly without further purification; C3 and C6 were obtained from TimTec (Newark, Delaware). The compounds in a powder form were dissolved in DMSO at 10 mM and used in biochemical studies after further dilutions.

ATPase activity assays
To determine the ATPase activity, ClpB was diluted to 28 µg/ml (49 nM hexamer) in buffer C (50 mM Tris-HCl pH 7.4, 20 mM MgCl2 1 mM EDTA, 0.5 mM TCEP). In some experiments, the ClpB solution was supplemented with 17.4 µg/ml (10 µM) κ-casein. The ClpB aliquots (18 µl) were mixed with 1 µl of the investigated compounds at different concentrations in 100% DMSO or with DMSO as a control. A sample without ClpB was used as a baseline control. After a 10-min preincubation at 37 °C or 45 °C, the ATP hydrolysis reaction was initiated by adding 1 µl of 100 mM ATP. The samples were incubated for 60 min (basal ClpB activity in the absence of κ-casein) or 15 min (in the presence of κ-casein) at 37 °C, or 30 min at 45 °C in the absence of casein. For the E678Q ClpB variant, the incubation time in the presence of κ-casein was 60 min. After incubation, 15 µl of each sample was mixed with 200 µl of the ammonium molybdate/malachite green reagent (67) dispensed into a 96-well plate, followed by an addition of 30 µl of 34% sodium citrate (68). The plate was agitated inside a Synergy H1 reader (BioTek, Winooski, Vermont) for 15 min at room temperature and the samples' absorbance was measured at 630 nm. The readouts for ClpBcontaining samples were corrected for absorbance of samples without ClpB, to account for nonenzymatic production of inorganic phosphate. A standard curve obtained with different inorganic phosphate concentrations was used to determine the amount of phosphate produced from ATP in the presence of ClpB.
To determine the ATPase activity of DnaK, 18 µl of DnaK solution, or DnaK NBD solution prepared in buffer C were mixed with 1 µl of the investigated compounds or DMSO as a control. Samples without DnaK were used as control blanks. The samples were pre-incubated for 10 minutes at 37 °C. The reaction was initiated by adding 1 µl of 100 mM ATP bringing the final concentration of of DnaK to 1.1 µM, or DnaK NBD to 3.4 µM. The samples were incubated at 37 °C for 70 min for DnaK or 75 mins for NBD. The concentration of inorganic phosphate produced from ATP in the presence of DnaK was determined as described above for ClpB. ). Then, 17 µl of a chaperone solution was mixed with 1 µl of DBeQ in DMSO or pure DMSO as a control, and pre-incubated for 10 min at 25 °C. Next, 1 µl of the heat-denatured luciferase (0.7 µM) and 1 µl of ATP (100 mM) were added to the samples, followed by incubation for 20 min at 25°C. To determine the luciferase activity, 5 µl of each sample was transferred into 100 µl of the luciferase substrate (Promega), previously dispensed into a white 96-well plate. Luminescence was measured using Synergy H1 plate reader (BioTek, Winooski, Vermont).

Surface plasmon resonance
ClpB, DnaK, and NBD of DnaK were extensively dialyzed against the SPR buffer (20 mM HEPES pH 7.4, 150 mM NaCl, 20 mM MgCl2, 1 mM EDTA, 1 mM DTT). After dialysis protein concentration was determined by Bradford method. The DBeQ-protein interactions were studied using a Biacore T200 instrument (GE Healthcare) at 25 °C. The proteins were diluted to 300 µg/ml in 10 mM sodium acetate buffer (pH 5 for ClpB DnaK or pH 4 for NBD) and immobilized on HC1500M sensor chips (Xantec) by standard amine coupling chemistry as described previously (70). A reference flow cell was created by ethyl(dimethylaminopropyl) carbodiimide/NHS activation followed by quenching with 1 M ethanolamine (pH 9.0). Solvent correction curves were obtained at the beginning, end, and after To test the structural integrity of ClpB after immobilization on the sensor chips, we performed control titrations with ATPγS and adenosine. The dissociation constants determined from those experiments were ~20 µM for ATPγS and ~2 mM for adenosine. These preliminary experiments indicated that the immobilization on the SPR chip preserved the oligomeric structure of ClpB because monomeric ClpB has low affinity towards ATP (Kd > 1 mM (62)). Moreover, the immobilized ClpB discriminated between a nucleotide analog and a nucleoside, which indicated an adequate structural integrity of ClpB on the sensor chips.
The resulting sensorgrams were analyzed using Biacore T200 Evaluation Software v3.0 according to the following procedure. Solvent correction curves were generated and applied to all datasets, and all sensorgrams were double-referenced by subtracting the most recent buffer blank injection. The signal immediately prior to injection stop of these corrected sensorgrams was treated as the binding response. GraphPad Prism was used for nonlinear least-squares fitting of the model assuming either non-cooperative or cooperative binding of DBeQ to DnaK and ClpB, respectively.

Bacterial viability
Escherichia coli MC4100, MC4100ΔclpB, dnaK103, and dnaK756 strains were used to determine an influence of DBeQ on bacterial growth and viability. Bacteria were maintained in LB media and in the case of the MC4100ΔclpB strain LB was supplemented with 30 µg/ml kanamycin. All experiments were initiated by preparing overnight cultures inoculated from single colonies and grown at 37 °C. The dnaK103, and dnaK756 overnight cultures were grown at 30 °C. On the following day, the cultures were diluted 100-fold in 10 ml of LB without antibiotics and incubated at 37 °C. The culture optical density was monitored at 600 nm. When absorbance at 600 nm reached 0.4, the culture was divided into 1 ml aliquots and supplemented with 10 µl of DMSO or different concentrations of DBeQ in DMSO. The samples were immediately transferred into an incubator/shaker with the temperature either 37 °C, 45 °C, or 50 °C. Bacteria were cultured for up to 4 h with shaking (200 RPM). At specific time points, 100 µl of each culture was withdrawn and serially diluted in sterile 0.9% NaCl up to 10 6 -fold dilution. To estimate the number of viable cells, 5 µl or 10 µl of the diluted cultures were spread on LB-agar (LA) plate without an antibiotic. After a complete adsorption of liquid on LA surface, the plates were incubated overnight at 37 °C. On the following day, the bacterial colonies were counted and the colony formation units (CFU)/ml were calculated.

Chemical library screening
An absorbance-based assay for ClpB ATPase activity was developed by using BioMol Green Reagent (Enzo Life Sciences, Farmngton, New York)) for screening in a 384-microplate based format. The optimized assay conditions were as follows: ClpB (150 nM), ATP (200 mM) and κcasein (25 µM) in 100 mM Tris/HCl pH 8, 10 mM MgCl2, 1 mM DTT and 1 mM EDTA. The samples were incubated for 10 min at 37 0 C and the ATP hydrolysis reaction was terminated by adding 80 µl of BioMol Green. Absorbance at 620 nm was measured with a Perkin Elmer Enspire. The optimized assay was used to screen compounds drawn from Selleck Bioactive library (1900 compounds. Selleck Chemicals, Houston, Texas) in the presence or absence of κ-casein.

Analytical ultracentrifugation
Protein samples (0.5 ml) of wt ClpB and its variants W462F and W543F were dialyzed against 3 changes of 1 l of buffer AUC (50 mM Tris-HCl pH 7.4, 200 mM KCl, 20 mM MgCl2, 1 mM EDTA, 0.5 mM TCEP). After dialysis, the samples were centrifuged (15 min, 13,000 RPM, 4 °C). The supernatants were collected and protein concentration was determined with Bradford assay. Sedimentation velocity experiments were performed in Beckman Optima XL-I analytical ultracentrifuge equipped with a 4-cell rotor and 2sector cells. The ClpB samples were diluted to 1 mg/ml in AUC buffer supplemented with DMSO (5% (v/v) final concentration) with 1 mM ATPɤS, with or without 50 µM DBeQ. Centrifugation was performed at 48,000 RPM at 20 °C. The sedimentation profiles were collected using the interference detection system and analyzed with the Origin software supplied by the instrument manufacturer using the time-derivative method of Stafford (71). Observed sedimentation coefficients were corrected to values corresponding to the density and viscosity of water (s20,w) using Sednterp software (www.jphilo.mailway.com) with additional corrections for the presence of 5% DMSO (72).

Fluorescence anisotropy
To monitor interactions between casein and ClpB, we employed FITC-casein and the substrate trapping ClpB E279Q/E678Q variant (52,73). The titration of FITC-casein with ClpB E279Q/E678Q in the presence of ATP or ADP and the fluorescence anisotropy measurements were performed as previously described (65).
To monitor interactions between the peptide B2 and DnaK, the DnaK stock was diluted in buffer C

Western blotting
The ClpB protein level in E. coli cells was analyzed by Western blotting. Overnight culture of each strain was diluted 100-fold and incubated at 37 °C with shaking (200 RPM). When the absorbance at 600 nm reached 0.4, the culture was split into two: for one, the incubation at 37 °C continued, while the second one was transferred to 45 °C. After 2 h of incubation at the target temperatures, the cells were collected by centrifugation. Supernatant was discarded and cell pellets were suspended in 160 µl of 2x Laemmli sample buffer. 60 µl of each sample were resolved in duplicates in 8% SDS-PAGE gel. The gel was cut into two parts, one part was used for blotting onto a nitrocellulose membrane and the second part of gel was stained in Coomassie R-250. The membrane was blocked with 5% non-fat dried milk dissolved in TBST buffer (19 mM Tris-HCl pH 7.4, 137 mM NaCl, 2.7 mM KCl, 0.2% Tween-20) overnight at 4°C. Next, the membrane was incubated with rabbit polyclonal anti-ClpB IgG (1:50,000 in milk) for 1 hour at room temperature. After incubation with the primary antibody, the membrane was washed in TBST buffer (4 times, 10 min). Next, the membrane was incubated with anti-rabbit HRP-conjugated secondary anti-bodies for 1 hour at room temperature and washed in TBST (4 times, 10 min) and TBS (10 min). Signal detection was performed using SuperSignal West Pico Chemiluminescent Substrate (Pierce) and an Azure c500 digital imaging system (Azure Biosystems, Dublin, California).

Confocal microscopy
ClpB-YFP and Luciferase-YFP were expressed at low levels in E. coli strains ΔclpB, dnaK103, and dnaK756 as described before (44). The expression of ClpB-YFP and YFP-Luciferase was initiated by addition of 200 μM IPTG or 0.02% (w/v) arabinose, respectively. The bacterial cells were cultured at 30 °C with 100 μM DBEQ or DMSO only. Protein translation was stopped by addition of erythromycin (30 μg/ml) and each culture was divided into two samples. One sample remained at 30 °C (control), while the other sample was transferred to 45 °C for 30 minutes (heat-shock conditions).
A thin layer of 1% Agarose (TopVision Low Melting Point Agarose, Thermo Scientific) was prepared on a clean microscope glass slide. A small drop of 1% agarose was placed near the frosted end of the glass slide while bringing a second glass slide at a 30-45° angle from above, allowing the agarose drop to spread along the contact edge. The top slide was moved gently across the bottom slide to produce an evenly spread thin layer of agarose. An E. coli culture sample (1 μL) was immediately placed on the agarose-covered glass slide, covered with a cover slip, and used for fluorescence imaging.
Fluorescence imaging was performed using a Carl Zeiss LSM 880 Airyscan confocal microscope equipped with a Plan-apochromat 63x/1.40 oil DIC M27 objective coupled with a YFP Filter (excitation at 514 nm, emission at 561nm, laser strength 0.98%). Zen 2.3 lite software was used for image capture and processing.
Data Availability: All relevant data are available within this manuscript and the associated Supporting Information.