E. coli chaperones DnaK, Hsp33 and Spy inhibit bacterial functional amyloid assembly

DnaK inhibits CsgA amyloid formation.

The conversion of soluble amyloidogenic monomeric proteins to an ordered amyloid fiber in vitro can be measured using the amyloid specific dye thioflavin T (ThT). When freshly purified CsgA was mixed with ThT, fluorescence emission increased rapidly after an approximately 2 h lag phase (Fig. 1A). The rapid growth phase of CsgA fibers was followed by a stationary phase, consistent with previous findings (Fig. 1A).⁷ When comparing CsgA polymerization profiles in the presence or absence of chaperone proteins, we measured two parameters. The first kinetic parameter was calculated as the time period preceding rapid fiber growth, called lag phase. The second parameter, called the elongation phase, was calculated as the time period encompassing the fiber growth phase from initiation of rapid polymerization to its completion.^7,67

To determine if DnaK affects CsgA amyloid formation in vitro, ThT fluorescence was measured in the presence or absence of DnaK. DnaK increased the lag phase 3- and 8-fold at molar ratios of 1:20 and 1:5 (DnaK:CsgA), respectively (Fig. 1A). However, once CsgA started to aggregate, neither the rate of fiber formation nor the maximal ThT fluorescence were affected by the presence of the chaperone (Fig. 1A). To determine if the ability of DnaK to inhibit CsgA polymerization was dependent on the intact DnaK substrate-binding domain, the DnaK^V436F mutant was tested in the ThT assay. The DnaK^V436F mutant has reduced substrate affinity, but similar substrate specificity.⁶⁸ When added to soluble CsgA, DnaK^V436F was significantly less efficient at inhibiting CsgA fiber formation compared with wild-type DnaK (Fig. 1B). Wild-type DnaK prevented CsgA polymerization for greater than 30 h, while an equivalent concentration of DnaK^V436F only increased the lag phase from 2–7 h (Fig. 1B).

To verify that the observed effect was due to the inhibition of CsgA polymerization, and not due to an indirect effect such as interference of the chaperone with ThT fluorescence we measured CsgA solubility in the presence and absence of DnaK. As an amyloid, curli are highly resistant to denaturation and require treatment with a harsh denaturant such as formic acid (FA) or hexafluoro-2-propanol (HFIP) to liberate the monomeric species.^12,38 After 13 h of incubation, polymerization reactions were centrifuged to separate the soluble and insoluble CsgA prior to analysis by SDS-PAGE. Duplicate samples were treated with 80% (w/w) FA to depolymerize CsgA aggregates.^12,69 CsgA was present only in the FA treated pellet samples after 13 h of incubation in the absence of DnaK indicating that CsgA was entirely in the polymerized form (Fig. 1C and lanes 1–4). After 13 h in the presence of DnaK, however, CsgA was still detected in the non-FA treated supernatant samples suggesting that DnaK maintains CsgA in a soluble state (Fig. 1C and lanes 5–8). These results were in agreement with transmission electron micrograph (TEM) images. CsgA formed unbranched fibers in the absence of DnaK; however, no such fibers were observed when CsgA was incubated in the presence of DnaK for 13 h (Fig. 1D and E).

Oxidized Hsp33 inhibits CsgA polymerization in vitro.

Hsp33 is a homodimeric molecular chaperone which, when activated by oxidative stress, functions as a potent holdase. To determine if Hsp33 influences amyloid formation, we measured CsgA fiber formation in vitro using ThT fluorescence.^70,71 Reduced, chaperone-inactive, Hsp33red and oxidized, chaperone-active, Hsp33ox, were assayed for the ability to inhibit CsgA polymerization in vitro.⁵⁵ The presence of Hsp33ox at a 1:80 (Hsp33ox dimer:CsgA) molar ratio increased the lag phase of CsgA fiber formation to 8 h (Fig. 2A). The presence of Hsp33ox at a molar ratio of 1:20 increased the lag phase by 13 h, while a molar ratio of 1:4 prevented CsgA polymerization for a time that was greater than the duration of the experiment (Fig. 2A). To verify that Hsp33 interfered with CsgA fiber formation and not with the ThT fluorescence, a CsgA solubility assay was conducted. All CsgA-containing samples had one band corresponding to CsgA in the pellet fraction treated with FA (Fig. 2B, lane 4). The majority of CsgA was present in the supernatant fraction of samples that contained Hsp33ox and CsgA at a 1:2 molar ratio signifying that CsgA was maintained in an SDS-soluble form (Fig. 2B, lanes 5–6). Fibers were observed by TEM in samples containing only CsgA, while no fibers were observed in samples containing both CsgA and Hsp33ox (Fig. 2C and D).

In contrast to Hsp33ox, Hsp33red had only marginal effects on CsgA polymerization in vitro (Fig. 2E). Consistent with the ThT results, CsgA was mostly present in the pellet fraction of samples containing Hsp33red and CsgA at a 1:2 molar ratio (Fig. 2F, lane 8). Significantly smaller amounts of CsgA were present in the supernatant fraction along with large quantities of Hsp33red (Fig. 2F, lanes 5–6). CsgA polymerization was also monitored by TEM. Regular, unbranched fibers were observed when CsgA was incubated alone or with Hsp33red for 13 h (Fig. 2G and H). Together, these results demonstrate that oxidized, but not reduced, Hsp33 maintains CsgA in a soluble state.

Hsp33ox and DnaK prolong the presence of a transient folding intermediate common to amyloids.

The conformation-specific antibody A11 recognizes a conserved, transient folding intermediate that is present during the polymerization of many functional and disease-associated amyloids.^7,72,73 Although the exact molecular structure that A11 recognizes is unresolved, it is proposed to be an ordered intermediate that precedes the mature amyloid fiber.⁷² A11 transiently recognizes freshly purified CsgA during the first part of its transition to an amyloid fiber. However, once polymerization reaches stationary phase, no A11 reactive species can be observed.⁷ The A11 antibody was therefore used to test if Hsp33ox and DnaK act on CsgA in a manner that impacts progression from an A11-reactive conformation to an amyloid conformation. Freshly purified CsgA monomers were incubated with substoichiometric concentrations of DnaK or Hsp33ox. Samples were removed at the indicated time points (Fig. 3A, arrows) and spotted onto a nitrocellulose membrane. An A11-reactive species was detectable up to approximately 8 h when CsgA was incubated in the absence of chaperone (Fig. 3A and B). In the presence of DnaK, the lag phase is increased from 2 h to 10 h and the A11-reactive species was detectable for up to 15 h (Fig. 3A and B). When CsgA was incubated with Hsp33ox the A11-reactive species did not disappear for more than 24 h after the start of the reaction (Fig. 3B). The persistence of the A11 reactive species was not due to cross-reactivity with Hsp33ox, as the A11 antibody did not react with the above assayed concentration of purified Hsp33ox (Fig. 3C). The small amount of cross-reactivity observed for the A11 antibody with DnaK (Fig. 3C) is consistent with previous findings that the A11 antibody reacts with Hsp70 and particularly with the substrate binding domain of Hsp70; however, Lotz et al. did not observe reactivity with full-length DnaK.^45,74

The inhibitory effects of Hsp33ox and DnaK are abolished by the addition of CsgA seed.

CsgA amyloid formation can be accelerated or “seeded” by the addition of preformed CsgA fibers.⁷ To determine whether DnaK and Hsp33ox alter CsgA seeding by preformed fibers, polymerization assays were performed in the presence of different CsgA seed concentrations. CsgA seed concentrations as low as 3% (weight/weight relative to soluble monomer) eliminated the lag phase of CsgA amyloid formation as has been previously demonstrated (Fig. 4A).⁷ The inhibitory effects of DnaK and Hsp33ox, as well as the lag phase of CsgA fiber formation, were reduced by the addition of 3% (w/w) CsgA seed and abolished by 12% (w/w) CsgA seed for both chaperones (Fig. 4B and C; note the time scales are different in A from B and C). In the presence of 3% seed there was an initial burst of fluorescence signal, a phase of minimal fluorescence increase and a more rapid growth phase that concluded in a stationary phase (Fig. 4B and C). Hsp33 reduced the maximum ThT fluorescence by a similar amount in the presence and absence of CsgA seed (Fig. 4C). The observation that the final ThT fluorescence remains reduced in the presence of Hsp33 and 12% CsgA seeds suggests that Hsp33 is sequestering a population of CsgA that cannot be incorporated into an amyloid (Fig. 4C, compare solid and dotted lines).

Spy inhibits CsgA polymerization in vitro.

Spy is a periplasmic chaperone that is regulated by the Cpx and Bae extracellular stress responses, both of which are induced by conditions that stress the proteome.^63–66 Spy exists as a homodimer and is structurally homologous to CpxP, the adaptor protein of the Cpx two-component stress response system. CpxP has also been proposed to act as a chaperone.^62,75,76 However, Spy has two hydrophobic patches in the concave surface of its structure that may be involved in substrate binding and chaperone activity; similar hydrophobic patches are less well-defined in CpxP.^62,76 We chose to test CsgA as a substrate for Spy since CsgA passes through the periplasm and, as a functional amyloid, is prone to ordered aggregation. No amyloid aggregates have been detected in the periplasm, therefore, we hypothesized that chaperone proteins like Spy may help to keep CsgA soluble as it passes through the periplasm to the cell surface.

To test the ability of Spy to prevent CsgA aggregation, we mixed purified Spy with freshly purified, monomeric CsgA and monitored polymerization in vitro by ThT fluorescence. Spy inhibited CsgA polymerization in a concentration dependent manner (Fig. 5A). Like DnaK and Hsp33, inhibition by Spy could also be overcome by the addition of preformed fibers (Fig. 5B). We found that Spy inhibited CsgA polymerization but the addition of comparable concentrations of CpxP to soluble CsgA had no effect on the polymerization of CsgA (data not shown). Comparison of these seemingly similar proteins will likely reveal their distinct functions in envelope stress responses.

Deletion of spy from the curli-expressing strain BW25113 had no effect on curli biogenesis (data not shown). It has been shown that some periplasmic chaperones are functionally redundant and can compensate for the loss of a single chaperone.^77,78 It is therefore not surprising that the spy deletion did not produce an observable phenotype. However, when Spy was overexpressed, we observed reduced Congo red binding (Fig. 6A) consistent with reduced CsgA protein levels by protein gel blot (Fig. 6B). One possible explanation for this result is that stabilization of unfolded CsgA by Spy in the periplasm allows for more efficient proteolytic degradation of CsgA and therefore less secretion and curli assembly. Alternatively, the excessive production and secretion of Spy may have titrated resources that are required for the production and secretion of curli subunits. CsgG levels are not as dramatically reduced as CsgA levels, suggesting that CsgA is more sensitive to Spy overexpression. Because CsgG levels remain similar to WT when Spy is overexpressed, we concluded that there was not competition for secretion or significant feedback repression of the curli operons (Fig. 6B).

Next we tested whether reduced curli assembly was due to an increased chaperone capacity in the periplasm. Periplasmic protein extracts (PEs) were harvested and added back to freshly purified CsgA monomers, polymerization was monitored by ThT fluorescence in vitro (Fig. 6C). We harvested periplasmic proteins from WT BW25113 with empty vector (vector) or overexpressing Spy (pSpy). PEs from cells with empty vector showed a minimal effect on CsgA polymerization (Fig. 6C and long dashes). However, PEs from cells overexpressing Spy inhibited CsgA polymerization (Fig. 6C, short dashes). After 20 h, CsgA was found to be in an SDS-insoluble state after incubation with periplasmic extracts from WT cells with empty vector only but primarily in an SDS-soluble state after incubation with periplasmic extracts from cells overexpressing Spy (Fig. 6D). These results demonstrate that the periplasmic chaperone activity was likely due to an increase in Spy protein in the periplasm.

CsgA purification.

CsgG and CsgA-6XHis were overexpressed in LSR12 (C600::Δcsg) and CsgA-6x His was purified from the growth media as previously described in references 7 and 12. To remove the imidazole, eluates were run over a gel filtration column (Sephadex G-25 fine, Amersham Biosciences) into either 50 mM potassium phosphate buffer pH 7.2 (KPi) or 50 mM KPi pH 7.2 that was supplemented with 50 mM potassium chloride (KCl). CsgA-6x His polymerizes into an amyloid fiber with similar kinetics to wild-type CsgA in vitro and will be referred to as “CsgA” in this paper.¹²

Hsp33 purification.

Hsp33 was purified as previously described in reference ⁵⁵. The reduction of Hsp33 was performed as previously described in reference ⁵⁹. To oxidize Hsp33, it was treated with 2 mM H₂O₂ at 43°C for 1 h prior to buffer exchange into 40 mM KPi pH 7.5, as previously described in reference ⁶⁰.

DnaK purification.

Wild-type and mutant DnaK (V436F) was purified as previously described in reference ⁸⁶ with slight modifications. Mutant DnaK was created by site-directed mutagenesis using the QuikChange kit (Stratagene).

Spy purification.

Spy was purified as previously described in reference ⁶².

Periplasmic extract preparation.

WT BW25113 cells were made chemically competent and transformed with pCDFTrc (vector) or pCDFTrckanBamHI-spy (pSpy)⁶² were grown in LB at 37°C overnight and then plated on YESCA plates supplemented with kanamycin and grown for 2 d at 26°C. It was unnecessary to induce Spy expression as increased Spy levels could be observed by coomassie stained SDS-PAGE (data not shown). Cells were harvested in 20 mM Tris (pH 8.0). Cells were collected by centrifugation at 8,000 rcf for 10 min followed by incubation in osmotic shock buffer (30 mM Tris, 40% sucrose 2 mM EDTA, pH 7.3) for 10 min. Cells were pelleted at room temperature and resuspended in 2 mM MgCl₂, incubated on ice for 3 min and pelleted again. The supernatant was kept as the periplasmic extract (PE).

ThT fiber formation assay.

Each experiment shown is a representative of at least three replicates. Freshly purified CsgA was mixed with 20 µM ThT and incubated in the presence of DnaK or Hsp33 in a black 96-well flat bottom plate (Corning Inc.), which was placed in a Spectramax M2 plate reader (Molecular Devices; Figs. 1–4). The plate reader was maintained at 25°C for all experiments. 50 mM KPi pH 7.2 was used as reaction buffer for all experiments involving Hsp33. For experiments involving DnaK, 50 mM KPi 50 mM KCl pH 7.2 was used as reaction buffer. Unless stated differently the reaction volumes were 100 µl. Molar ratios are reported in terms of monomeric DnaK and dimeric Hsp33. Fluorescence was monitored in 10 min intervals at 495 nm after excitation at 438 nm. The cut off of the plate reader was set to 475 nm. The plate was shaken for 5 sec prior to each read. CsgA polymerization in the presence of Spy and periplasmic extracts were conducted in 50 uL reaction volumes in 50 mM KPi (pH 7.3) with 20 uM ThT. ThT readings were taken in a Infinite 200 Pro NanoQuant plate reader (Tecan Group LTD.; Figs. 5 and 6) every 10 min after 2 sec of shaking at 2.5 mm amplitude with an excitation wavelength of 438 nm and emission wavelength of 495 nm. Molar ratios are reported in terms of Spy dimer. At least three replicates of each assay were conducted.

CsgA solubility assay.

100 µl CsgA samples supplemented with the respective chaperones were incubated in 96-well plates in a Spectramax M2 plate reader at 25°C. Identical samples were supplemented with 20 µM ThT to monitor amyloid formation. ThT fluorescence was monitored as described above. After 13 h 30 µl aliquots were taken from the CsgA samples and centrifuged at 70,000x g for 1 h at 4°C in 1.5 ml polyallomer tubes (Beckman Instruments) using a TLA-55 rotor (Beckman Coulter) in an Optima TLX ultracentrifuge (Beckman Coulter). After centrifugation, the supernatant fractions (∼30 µl) were transferred into a fresh tube. Pellet fractions were treated with 50 µl 95% (w/w) formic acid (Sigma-Aldrich, F0507), while 95% formic acid was added to the supernatant fraction yielding a final formic acid concentration of 76%. Immediately after addition of formic acid the samples were placed in a SPD111V SpeedVac (Thermo Fisher Scientific) and all solvent was removed. The formic acid treated samples were resuspended in 4x SDS sample buffer (125 mM Tris pH 6.8, 10% 2-Mercaptoethanol, 20% Glycerol, 6% sodium dodecyl sulfate, 0.02% bromophenol blue), sonicated for ten minutes in a Solidstate/Ultrasonic FS-14 bath sonicator (Fisher Scientific) before incubating at 95°C for 10 min. Formic acid treated and untreated samples were then analyzed by SDS-PAGE and Coomassie staining. The solubility of CsgA after incubation with periplasmic extracts was determined similarly. Soluble and insoluble CsgA were separated by centrifugation in an Eppendorf 5415R centrifuge in duplicate. Pellets were treated with 40 µL 4x SDS sample buffer or 50 µL hexafluoro-2-propanol (HFIP; Sigma Aldrich) for 10 min and then dried in a SPD111V SpeedVac (Thermo Fisher Scientific) before resuspension in 40 µL of 4x SDS sample buffer. Supernatants were treated with and equal volume of 4x SDS sample buffer or with a final concentration of 80% HFIP for 10 min, dried and then resuspended in 40 µL 4x SDS sample buffer. Samples were boiled for 10 min prior to loading on a 15% SDS-PAGE gel and analyzed by Coomassie staining.

Dot-Blot Assay with A11.

The blot assay was performed as previously described in reference ⁷. CsgA samples were taken from fiber formation reactions in 96-well plates that were simultaneously monitored with ThT fluorescence in separate wells at 25°C, as described above.

Whole cell protein gel blot.

BW25113 with pCDFT (vector) or pCDFTrckanBamHI-spy (pSpy) were grown overnight in LB supplemented with kanamycin at 37°C. Cultures were diluted to OD = 1.0 in 50 mM KPi (pH 7.3) and 4 µL were spotted on a YESCA plate and grown for 48 h at 26°C. Cells were harvested in 1 mL of 50 mM KPi (pH 7.3). Two 100 µL samples were taken from a 1.0 OD suspension and pelleted at 16,000 rcf for 10 min. Samples were either treated with 40 µL of 2x SDS loading buffer or 50 µL of hexafluoro-2-propanol (Sigma-Aldrich, 52512), dried in a Thermo Savant SPD SpeedVac (Thermo Fisher Scientific) and then treated with 40 µL of 2x SDS loading buffer. Seven microliters were loaded in a 15% SDS PAGE gel, transferred to 0.45 µm PVDF transfer membrane (Thermo Fisher Scientific) and blocked overnight in TBS-T 5% skim milk. Blots were probed with anti-CsgA (1:10,000) and anti-CsgG (1:70,000) polyclonal antibodies followed by goat-anti-rabbit (1:10,000; Sigma-Aldrich, A0545).

Electron microscopy.

Electron microscopy sample preparation and image acquisition was performed as previously described in reference ⁷.