The E. coli CsgB nucleator of curli assembles to β-sheet oligomers that alter the CsgA fibrillization mechanism

Abstract

Curli are extracellular proteinaceous functional amyloid aggregates produced by Escherichia coli, Salmonella spp., and other enteric bacteria. Curli mediate host cell adhesion and invasion and play a critical role in biofilm formation. Curli filaments consist of CsgA, the major subunit, and CsgB, the minor subunit. In vitro, purified CsgA and CsgB exhibit intrinsically disordered properties, and both are capable of forming amyloid fibers similar in morphology to those formed in vivo. However, in vivo, CsgA alone cannot form curli fibers, and CsgB is required for filament growth. Thus, we studied the aggregation of CsgA and CsgB both alone and together in vitro to investigate the different roles of CsgA and CsgB in curli formation. We found that though CsgA and CsgB individually are able to self-associate to form aggregates/fibrils, they do so using different mechanisms and with different kinetic behavior. CsgB rapidly forms structured oligomers, whereas CsgA aggregation is slower and appears to proceed through large amorphous aggregates before forming filaments. Substoichiometric concentrations of CsgB induce a change in the mechanism of CsgA aggregation from that of forming amorphous aggregates to that of structured intermediates similar to those of CsgB alone. Oligomeric CsgB accelerated the aggregation of CsgA, in contrast to monomeric CsgB, which had no effect. The structured β-strand oligomers formed by CsgB serve as nucleators for CsgA aggregation. These results provide insights into the formation of curli in vivo, especially the nucleator function of CsgB.

Curli are highly aggregated, thin (2–6 nm diameter) amyloid fibers expressed on the surface of Escherichia coli, Salmonella, and many other Enterobacteriaceae (1–6). Curli can bind a variety of host proteins, can mediate host cell adhesion and invasion, and are involved in colonization and biofilm formation (7–11). Bacterial biofilms constitute a protected mode of growth against environmental stresses and immune defense, causing high tolerance for antimicrobials, which results in persistent infections (12–16).

Curli formation involves a complex molecular machinery that is encoded by the divergently transcribed csgBA and csgDEFG operons (7, 17). Curli fibers consist of two subunits: CsgA, which constitutes the major portion of the fiber, and CsgB, the minor component. Both CsgA and CsgB are secreted proteins with similar molecular weights of ∼13 kDa, and their interaction triggers wild-type curli formation in bacteria (5). The csgA gene product (18) is a soluble, unstructured protein that requires the presence of CsgB to assemble into fibers in vivo (4). In the absence of CsgB, CsgA is secreted from the cell as monomers, and no fibrils are formed (4–6). The csgB gene product (19) is considered to be the nucleator for the assembly of curli (5). Unlike CsgA, overexpressed CsgB can self-assemble into short polymers on the bacterial surface when expressed alone (5). These data suggest that CsgA and CsgB have different aggregation properties and play different roles in curli formation.

Previous studies have shown that CsgA and CsgB are highly similar in terms of biochemical and biophysical properties; they show high sequence homology, with about 50% similarity and 30% identity. The predicted structures of monomeric CsgA and CsgB are similar, consisting of five conserved, 18-residue tandem strand-loop-strand motifs, each containing conserved glycine, glutamine, and asparagine residues (7, 20, 21). In vitro, purified CsgA is able to form fibers upon prolonged incubation that are indistinguishable from wild-type curli as analyzed by EM (6). CsgA aggregation shows a distinct lag, as detected by thioflavin T (ThT) fluorescence changes, followed by a rapid increase in fluorescence to a final plateau value (22). A C-terminal truncate of CsgB, in which the C-terminal 19 amino acids have been removed (CsgB_trunc), can also assemble into fibers in vitro that bind to the amyloid-specific dyes Congo red and ThT (23). However, CsgB_trunc is less efficient than full-length CsgB in mediating CsgA nucleation in vivo (23). X-ray diffraction of fibrils formed by purified CsgA and CsgB (WT, full length) showed the distinctive spacings (∼4.7 and ∼9 A), indicative of cross-β structure that is commonly found in disease-associated amyloid (24). It remains unclear how CsgA depends on the presence of the nucleator protein CsgB to form functional amyloid fibrils in vivo.

In this study we compared the aggregation kinetics of both WT CsgA and CsgB in vitro using multiple techniques. The results reveal that CsgA and CsgB aggregate by different mechanisms and suggest that the formation of curli fibers in vivo is regulated by the ability of CsgB to rapidly form oligomeric structures that alter the mechanism of aggregation of CsgA.

Results

ThT Fluorescence Assay of Csg Aggregation.

The aggregation of CsgA and CsgB_trunc has previously been reported to show a lag phase, as measured by ThT fluorescence change, followed by an exponential rise in fluorescence to a final plateau value (22, 23, 25). Seeds made from CsgB_trunc (in which the C-terminal domain has been truncated by 19 residues) can nucleate CsgA aggregation in vitro, but CsgB_trunc is different from WT in nucleation of curli formation in vivo (23). It is therefore important to examine the properties of CsgB. Fig. 1A shows the time dependence for CsgB aggregation as a function of CsgB concentration. Similar to CsgB_trunc, CsgB exhibits a lag phase of ∼100 min, which is shorter than that observed for CsgA (∼200 min) under the same conditions.

Fig. 1.

Aggregation/fibrillation of CsgB and its interaction with CsgA monitored by ThT fluorescence. (A) The time course of CsgB (1–7 μM) aggregation measured by fluorescence change at 482 nm. (B) Comparison of the time course of ThT fluorescence change of CsgA (4 μM, dashed line), CsgB (4 μM, solid line), and the mixture of CsgA and CsgB (4 μM each, circles). The asterisk (*) and caret ( ^ ) highlight the differences in the lag phase of CsgB and CsgA, respectively. Assays were performed using a plate reader. The standard conditions for the fibrillations were in 50 mM KPO₄ (pH 7.2) at room temperature.

The effect of monomeric CsgB on the aggregation of CsgA was investigated using the ThT assay. A 1:1 mixture of monomeric CsgA and CsgB (4 μM each) was observed to have a lag phase similar to that of CsgB alone over the first 100 min (Fig. 1B). Thus, the lag phase typical of CsgA alone disappeared (Fig. 1B). Changing the CsgA concentration from 2 to 10 μM had no obvious effect on the time course of the lag phase, suggesting that the aggregation of CsgB was not affected by CsgA in the lag phase. However, after 100 min, the presence of CsgB decreased the long lag phase of CsgA. During this time, CsgB forms aggregates as indicated by Fig. 1A. Seeds (oligomers) of CsgB abolished the lag phase of CsgA and of CsgB (Fig. 2 A and B). The results suggest that CsgB oligomers accelerate CsgA aggregation. Thus, the absence of the CsgA lag phase when CsgA is mixed with CsgB in Fig. 1B is likely due to the formation of CsgB oligomers during the first 100 min that subsequently accelerate CsgA aggregation.

Fig. 2.

ThT fluorescence assay of (A) CsgA and (B) CsgB in the presence of different seeds/oligomers made from preformed aggregates/fibers of either CsgA or CsgB. Both CsgA and CsgB were 4 μM in 50 mM KPO₄ (pH 7.2) with or without 1% seeds (by weight). Experiments were performed using a PTI spectrofluorometer.

Circular Dichroism Changes.

Far-UV CD changes were used to follow the secondary structural changes of CsgA and CsgB during aggregation (Fig. 3). Immediately after removal of Gdn and initiating the aggregation process, the CD spectra of both CsgA and CsgB are characteristic of a random coil with a minimum close to 200 nm and little signal elsewhere. The two proteins, however, show significant time-dependent CD differences at 216 nm, a wavelength that is characteristic for β-sheet formation (Fig. 3 A and B). For CsgA during the first 24 h, the CD signal exhibits little or no change. Over longer times, the CD signal at 200 nm decreases (from 2 d to 13 d) with only a slight change in the signal at 216 nm (Fig. 3C). At very long times (9–13 d), the CD spectrum of CsgA shows a weak minimum signal at ∼220 nm, rather than at 216 nm, but probably characteristic of some β-sheet structures (Fig. 3 C and D). The slow change of the CD signal at 200 nm after 1,000 min has a half-time of ∼4,800 min (Fig. 3E).

Fig. 3.

Far-UV CD spectra of (A) CsgA and (B) CsgB as a function of time. (C) CD spectra of CsgA over 13 d. (D) CD spectrum of CsgA after 13 d, showing a minimum at 220 nm. The time is shown in minutes (m), or hours (h). (E) Comparison of CD signal change over time at 200 and 216 nm, for CsgA and CsgB. All spectra were determined in a 0.1-cm path cell using 10 μM of protein in 50 mM KPO₄ (pH 7.2). Similar CD behavior was observed with a 1-cm cell using a protein concentration of 3 μM in 10 mM KPO₄ (pH 7.2). Table 1 shows the kinetic parameters obtained from these data.

In contrast, the CD signal of CsgB rapidly increased at 200 nm and decreased at 216 nm, indicating the appearance of β-sheet structure (Fig. 3B). The CD change of CsgB at either 200 nm or 216 nm could be fit by a single exponential function with a half-time of 180 min (Fig. 3E and Table 1).

Table 1.

Kinetic parameters of structural change during Csg protein aggregation

Protein	CD, nm	Amplitude1	k1, m−1	Amplitude2	k2, m−1
CsgB*	200	−26.32 ± 0.29	0.0043 ± 0.00012	N/A	N/A
	216	8.71 ± 0.15	0.0042 ± 0.00018	N/A	N/A
CsgA:CsgB = 20:1†	200	−1.47 ± 0.25	0.0045 ± 0.0015	−12.53 ± 0.55	0.00011 ± 0.000013
	216	1.22 ± 0.42	0.016 ± 0.01	2.86 ± 0.55	0.00012 ± 0.000048
CsgA:CsgB = 10:1‡	200	−3.58 ± 0.49	0.002 ± 0.0006	−15.45 ± 2.82	0.00007 ± 0.00003
	216	1.70 ± 0.53	0.0093 ± 0.0064	2.78 ± 0.30	0.00021 ± 0.000077
CsgA:CsgB = 5:1§	200	−4.98 ± 0.33	0.006 ± 0.00093	−11.92 ± 0.33	0.00015 ± 0.000015
	216	5.87 ± 1.72	0.02 ± 0.0054	2.63 ± 0.15	0.00051 ± 0.00008

All experiments were carried out in 50 mM KPO₄ (pH 7.2). Amplitudes for CD are given as millidegrees (mdeg). The parameters were obtained by fitting the CD data using a single-exponential function for CsgB, and a two-exponential function for the mixture of CsgA and CsgB.

*The protein concentration of CsgB used was 10 μM.

^†20:1 ratio of CsgA 7.8 μM to CsgB 0.39 μM.

^‡10:1 ratio of CsgA 6.9 μM to CsgB 0.69 μM.

^§5:1 ratio of CsgA 5.6 μM to CsgB 1.12 μM.

Of interest is the comparison of CD and ThT over the first 1,000 min as shown in Fig. 4. For CsgB (Fig. 4A), the CD signal change corresponds to the increase in ThT fluorescence, although no lag phase is observed, which suggests that CsgB forms aggregates with ordered β-sheet structure that can bind to ThT and produce the characteristic ThT fluorescence. In contrast, no correlation was observed between the CD signal change and that monitored by ThT for CsgA (Fig. 4B). Although ThT fluorescence increases for CsgA over time, there is no corresponding change in CD, suggesting that at early times, CsgA forms aggregates with little or no β-sheet structure. The difference between the behavior of CsgA and CsgB reflects a difference in the mechanism of aggregation, which is discussed below.

Fig. 4.

Comparison of normalized CD₂₀₀ and ThT fluorescence signals during the aggregation of (A) CsgB and (B) CsgA. CsgA and CsgB were 4 μM for the ThT assay and 10 μM for the CD measurements. Data obtained in 50 mM KPO₄ (pH 7.2). For CsgB, the CD₂₀₀ was inverted from that shown in Fig. 3 for comparison with the ThT change.

Role of CsgB in Affecting CsgA Aggregation.

Because CsgA and CsgB alone show significant differences in the CD signal change at 216 nm, we tested the effect of CsgB on the aggregation of CsgA as monitored by CD. Fig. 5 shows the CD change over incubation time using concentrations of CsgA to CsgB at ratios of 20:1, 10:1, and 5:1. In contrast to CsgA alone, the time-dependent spectra of the mixtures showed apparent changes at 216 nm, characteristic of β-sheet structure at all CsgA:CsgB ratios. At the concentrations of CsgB used, from 0.39 to 1.12 μM, the CD₂₁₆ signal of all of the mixtures was larger than that expected for CsgB alone (Table 1; amplitudes). In these experiments the CD changes were fit by a two-exponential function (Table 1). Interestingly, the rate constant for the first phase is near that of CsgB alone (Table 1), suggesting that this fast phase is mostly due to the rapid polymerization of CsgB. Thus, it is clear that CsgB rapidly forms oligomers that change the aggregation mechanism of CsgA. Because CsgA alone shows little change at 216 nm, the large change at 216 nm in the presence of substoichiometric amounts of CsgB indicates that the aggregation of CsgA into β-strand structures was induced by oligomeric CsgB. Therefore, CsgB rapidly self-assembles to oligomers/fibers with ordered β-sheet structure, which serves as a nucleator for CsgA aggregates/fibers formation with more-ordered β-sheet structures. At higher CsgB concentrations, the rate of the change at 216 nm increases, as would be expected for more extensive CsgB aggregation. These results are consistent with the in vivo observation that the formation of curli requires CsgB serving as a nucleator (5).

Fig. 5.

Far-UV CD change of the mixture of CsgA and CsgB at different ratios. (A and B) 20:1 ratio of CsgA (7.8 μM) to CsgB (0.39 μM). (C and D) 10:1 ratio of CsgA (6.9 μM) to CsgB (0.69 μM). (E and F) 5:1 ratio of CsgA (5.6 μM) to CsgB (1.12 μM). (A, C, and E) Comparison of CD spectra at different incubation time for 3 d. (B, D, and F) Time course of CD signal change over 9 d of incubation. The time is shown in minutes (m). Data obtained using a 0.1-cm path cell in 50 mM KPO₄ (pH 7.2).

Intrinsic Fluorescence Measurements.

Changes in the intrinsic fluorescence of a protein reflect changes in solvent accessibility of the side chains of aromatic amino acids. CsgA and CsgB differ in tryptophan content. CsgA contains a single tryptophan (W106), four tyrosines (Y26, 48, 50, and 151), and three phenylalanines, whereas CsgB contains no tryptophan, six tyrosines (Y24, 31, 88, 92, 109 and 129), and two phenylalanines.

For CsgA, using an excitation wavelength of 280 nm, the protein exhibits two emission maxima: one at 303 nm due to the four tyrosine residues and one at 350 nm due to the single tryptophan (Fig. 6A). At an excitation wavelength of 295 nm, only a single emission maximum, due to the tryptophan, is observed (Fig. 6B). For CsgB, a single-emission maximum at 303 nm is observed using an excitation wavelength of 280 nm (Fig. 6A), whereas no fluorescence is observed when using an excitation wavelength of 295 nm (Fig. 6B). Thus, using an excitation wavelength of 280 nm, the fluorescence of the mixture (CsgA + CsgB) at 350 nm is only due to CsgA, whereas that at 303 nm measures both CsgA and CsgB (as shown in Fig. 6C). When excited at 295 nm, the fluorescence of a CsgA/CsgB mixture measures only that of CsgA, as shown in Fig. 6D.

Fig. 6.

Intrinsic fluorescence change of CsgA and CsgB during aggregation. (A and B) Emission scans of CsgA alone and CsgB alone with the excitation wavelength of (A) 280 nm and (B) 295 nm. (C and D) Emission scans of the mixture of CsgA and CsgB with excitation at (C) 280 nm or (D) at 295 nm. (E–H) The fluorescence intensity changes as a function of time. (E) CsgA alone and (F) CsgB alone both using 280 nm excitation. (G and H) The mixture of CsgA and CsgB excitation at (G) 280 nm and (H) 295 nm. The protein concentration for CsgA or CsgB alone was 1 μM, and 1 μM each for CsgA and CsgB mixtures.

The intrinsic fluorescence of CsgA alone, excited at 280 nm, shows a small linear decrease over time, monitored by either tyrosine or tryptophan signal (Fig. 6E), which is probably due to the loss of soluble protein during the aggregation process. However, the intrinsic fluorescence of CsgB alone, excited at 280 nm, shows an exponential decrease as a function of time (Fig. 6F) with a rate constant similar to the change in CD signal (Fig. 3E), confirming that the structure of CsgB rapidly changes with a single exponential rate during its aggregation.

The tyrosine fluorescence change [Fig. 6G; excitation (Ex) = 280 nm, emission (Em) = 303 nm] of the mixture of CsgA and CsgB exhibits both the exponential decrease of CsgB and the linear decrease of CsgA. The tryptophan fluorescence change of the mixture is similar to that of CsgA with a linear decrease (Fig. 6G; Ex = 280 nm, Em = 350 nm). When excited at 295 nm (Fig. 6H), the mixture shows only the linear decrease seen for CsgA alone. Thus, the intrinsic fluorescence results show that CsgB undergoes a structural change either alone or when coincubated with CsgA.

Discussion

Curli are examples of functional amyloid fibers that have been observed in a wide variety of sources, including bacteria, fungi, insects, invertebrate, and humans (6, 7, 26). Although Csg proteins had been extensively studied in recent years, many questions remain, particularly with regard to the structures and interactions that occur at early steps in the aggregation process and lead to subsequent fibrillization.

CsgB Aggregation in Vitro.

For CsgB, the CD signal at 216 nm changes rapidly, as does its intrinsic fluorescence, suggesting early formation of β-sheet structure. Fibril formation, as measured by ThT fluorescence change, is coincident with changes in both CD₂₀₀ and CD₂₁₆.

CsgA Aggregation in Vitro.

The behavior of CsgA alone differs dramatically from that of CsgB. Although CsgA forms visible aggregates within 4–6 h, and ThT fluorescence increases after a lag time of ∼200 min, there is no change in the CD signal at either 200 or 216 nm. Instead, any CD signal change is very slow, occurring over a period of days. Hence, any CsgA aggregate formed at early times has little or no β-sheet structure. Atomic force microscopy (AFM) data (Fig. 7) obtained after several days revealed amorphous aggregates of CsgA. Both normal fibers and amorphous aggregates were found, as shown in Fig. 7A. Fibers could often be seen growing from these amorphous aggregates (Fig. 7B). Therefore, to explain the paradoxical results between the ThT and CD data, we propose that CsgA does not undergo fibril formation at early times of CsgA aggregation but rather forms amorphous aggregates; thus, the ThT fluorescence at these times is a measurement of the amorphous aggregates rather than fibrils. This observation is not unprecedented, as discussed by Groenning (27). Also, when investigating the aggregation of β-lactoglobulin, Carrotta et al. (28) suggested that the structural motif recognized by ThT can appear in protein aggregates. Some intrinsically disordered proteins may initially form amorphous aggregates, and the amyloid assemblies (fibers) emerge exclusively from within the aggregates, as reported by others (29). Polymorphism of amyloid has been widely observed (30, 31).

Fig. 7.

Representative AFM tapping mode amplitude image of CsgA. The protein was 10 μM in 50 mM KPO₄ (pH 7.2) and incubated at 4 °C for 24 d before imaging. The boxed region in A is shown magnified in B. (Scale bars: A, 1 μm; B, 0.1 μM.)

Over a long period (9–13 d), the CD spectrum of CsgA shows a minimum ∼220 nm (Fig. 3 C and D), indicating CsgA eventually formed aggregates with β-strand–like structures. The slight difference in the minimum of CD signal between CsgA (∼220 nm) and CsgB (∼216 nm) suggests that CsgA and CsgB aggregates/fibers may differ somewhat structurally.

It has been previously noted that ThT fluorescence and CD signal changes are not correlated for CsgA. For example, Wang et al. (22) showed that the CD signal at 200 nm changed dramatically within 2 d, whereas the minimum at 220 nm increased only slowly over a period of 2–15 d. In measurements over a period of 6 h, Dueholm et al. (25) observed large CD changes at 197 nm but little signal change at 216 nm, and that signal changes in CD preceded those in ThT fluorescence. Differences in the kinetics of aggregation between different studies may reflect differences in protein preparation or experimental conditions.

CsgA Aggregation in the Presence of CsgB in Vitro.

The aggregation of CsgA in the presence of full-length CsgB has not previously been reported. Our CD data show that, at substoichiometric concentrations, CsgB changes the nature of the aggregation of CsgA. Under conditions where both CsgA and CsgB are present, CsgB rapidly self-assembles to oligomers with β-sheet structure, which then serve as nucleators for CsgA aggregation. Our current data, however, cannot determine the actual number of CsgB monomers required to form a functional nucleator.

CsgA and CsgB Aggregation in Vivo.

Wild-type curli are CsgA/CsgB heteropolymers. The major subunit is CsgA, and CsgB is a minor subunit. The fibers are ∼5–12 nm in diameter with irregular thin branches. CsgA is unable to polymerize in the absence of CsgB; thus, in the absence of either CsgA or CsgB, no surface structures are observed (5). However, when overexpressed in the absence of CsgA, CsgB forms large quantities of cottony short polymers lacking the filamentous structure of wild-type curli (5). CsgB fused to maltose-binding protein also triggered the assembly of curli that were distinctly curved and loosely aggregated, 10–15 nm thick (5). These morphological differences can be explained by the differences in structure and polymerization kinetics that we observed between CsgA and CsgB. The differences in the length of the fibers should relate to the aggregation rate and the relative concentration of subunits available. The fast aggregation rate of CsgB limits the diffusion region of the molecules and results in shorter polymers, whereas CsgA with a slower rate can form longer fibers. Moreover, different structural properties of the oligomers affect the packing mode of the fibers, resulting in differences in shape and thickness.

In vivo, CsgA is the major subunit and is present in much greater amounts than CsgB. Our data show that CsgB rapidly forms aggregates with β-sheet structure. These structures alter the formation of CsgA aggregates. In vivo and in the absence of CsgB, CsgA monomers diffuse away from the surface of the cells, which we propose is a consequence of two factors: the relatively slow aggregation of CsgA and the fact that CsgA intermediates are not structured. We have shown that CsgA aggregation is accelerated and that the aggregates formed are structurally changed in the presence of CsgB. Thus, our data argue for a mechanism whereby CsgB rapidly assembles to aggregates/fibers that nucleate the formation of CsgA fibrils. Due to the rapid CsgB aggregation kinetics, CsgB is capable of forming short polymers condensed on the cell surface when overexpressed, as described in a previous report (5). In wild-type E. coli, the stoichiometry of CsgA to CsgB precludes CsgB fiber formation. In sum, our data argue that the formation of curli in vivo uses the different aggregation kinetics and structural differences of CsgA and CsgB to form curli fibers.

Curli proteins CsgA and CsgB share many biochemical and structural properties with disease-associated amyloids (6, 22, 23). Wang et al. (22) observed that CsgA formed transient intermediates that bind to A11, an antibody that specifically recognizes intermediate oligomers, but not soluble monomers or mature amyloid fibers of many disease-associated amyloids, such as Aβ, islet amyloid polypeptide, polyglutamine, prion peptide, and Sup35p (32–34). Therefore, curli biogenesis may provide an excellent model for understanding the polymerization mechanism of disease-associated amyloids. The structural and kinetic mechanism observed in Csg proteins also provides insights into the aggregation/fibrillization mechanism of other amyloids.

Conclusion

Curli biogenesis is a directed amyloid aggregation process requiring specific molecular machinery to not only secrete subunits to the extracellular surface but to also ensure that amyloid fibril formation is restricted to the outer surface of the bacterium. In this paper, we explored the aggregation of CsgA and CsgB both alone and together in vitro. We found that CsgA and CsgB self-associate to form aggregates/fibrils with different aggregation kinetics and structural properties. For CsgA, the mechanism appears to be the formation of amorphous aggregates from which fibrils appear after long times. For CsgB, the aggregation appears to follow a more common path, with fibrils occurring early in the aggregation process. Substoichiometric concentrations of CsgB change the aggregation mechanism of CsgA to that characteristic of CsgB. Our results indicate that the formation of wild-type curli requires a combination of both CsgA and CsgB, with CsgB directing the aggregation process.

Materials and Methods

Protein Expression and Purification.

Wild-type CsgA and CsgB with 6 His-tag at the C-terminal end (designated here as CsgA and CsgB) were cloned into pET11d (New England Biolabs) and overexpressed in E. coli NDH471 (NEB3016 SlyD::aph). Both proteins were expressed in inclusion bodies. The inclusion bodies (∼1 g) were suspended in 50-mL 8 M guanidine hydrochloride (Gdn), 50 mM KPO₄ (pH 8), stirred for 1 d, and then centrifuged. The supernatant solution in 8 M Gdn was then passed through a 0.45-uM bottle-top filter (Corning) and loaded onto a column packed with TALON metal affinity resin (Clonetech). The protein was eluted with 200 mM imidazole in 6 M Gdn, 50 mM KPO₄ (pH 7.2) and stored at 4 °C. Immediately before spectroscopy or other experiments, the protein was desalted and changed to buffer of 10 or 50 mM KPO₄ (pH 7.2) using a Sephadex G25 Desalt column (HiTrap; GE Healthcare). The start time of protein aggregation was considered to begin from the time that Gdn was removed. In the experiments using a mixture of CsgA and CsgB, CsgA and CsgB were refolded for 5–15 min after removal of Gdn. The start time of the aggregation of the mixture was considered to be the time that Gdn was removed (15 min).

Preparation of Csg Protein Seeds.

Following removal of Gdn, CsgA or CsgB (5–20 μM) was incubated in 50 mM KPO₄ (pH 7.2) in a glass tube at 4 °C. Csg protein seeds were made from the samples incubated >30 d. The aggregates/fibers were centrifuged, resuspended in water (1 mg/mL), and sonicated for 30 s. The visible aggregates/fibers in the samples disappeared and the solution turned milky. This material was used as seeds at a 1:100 dilution.

ThT Assay.

Aggregation was followed by the change in fluorescence of ThT (35). When using the PTI fluorometer (Photon Technology International), the excitation wavelength was 438 nm and the fluorescence emission was recorded (from 450 to 550 nm) every 10 min after mixing the sample. The measurements were made in 50 mM KPO₄ (pH 7.2) containing 25 μM ThT in a total volume of 3 mL at 25 °C. When using the Tecan Infinite 200 plate reader with the ThT assay, proteins were mixed with 25 μM ThT in a total of 125 μL in 96-well plates in duplicate and incubated at room temperature. Samples were measured every 10 min using an excitation wavelength of 438 nm and emission >495 nm, using a 475-nm cutoff filter. Samples were shaken for 10 s (Orbital shaker; amplitude 2 mm) before each reading. Varying concentrations of protein (2–20 μM) were used in this study. Signal intensities are in arbitrary units and differed between data collected on the PTI fluorometer and the plate reader. All ThT assays were repeated 3×, and representative data are shown. The lag phase for aggregation of CsgA was similar to that of previous reports (22, 25). We observed, however, that at high protein concentrations the solution turned cloudy at the end of the lag phase, and the fluorescence signal became noisy, especially at protein concentrations >4 μM. Thus, to minimize this problem, a protein concentration ∼4 μM was used in most ThT assays.

Circular Dichroism.

Far-UV CD spectra were recorded on a Jasco-J715 spectropolarimeter with a 1-cm or 0.1-cm path cell, depending on the protein concentration. Spectra were recorded at 25 °C as an average of 5 or 10 scans from 190 to 240 nm. The scan rate was 50 nm/min with a response time of 1 s. All samples were mixed before every reading. Different CD experiments were repeated with varying protein concentrations (2–15 μM).

Fluorescence Spectroscopy.

The intrinsic fluorescence emission of CsgA and CsgB were measured on a PTI fluorometer (Photon Technology International) using an excitation wavelength of 280 or 295 nm at 25 °C. The protein concentration was 1 μM in 50 mM KPO₄ (pH 7.2). Because CsgB does not contain tryptophan, an excitation wavelength of 280 nm was used.

Atomic Force Microscopy.

Aggregates/fibers of CsgA from the incubation of 10 μM protein in 50 mM KPO₄ (pH 7.2) at 4 °C were diluted in 1:20 in Milli-Q H₂O and placed on a mica disk to adhere for 1 min. The sample was removed by pipetting, and the mica was gently rinsed with Milli-Q H₂O twice and dried under a nitrogen stream. AFM images were acquired by tapping mode using the Asylum Research MFP-3D-BIO AFM.