Electrostatic lipid–protein interactions sequester the curli amyloid fold on the lipopolysaccharide membrane surface

Curli is a functional amyloid protein in the extracellular matrix of enteric Gram-negative bacteria. Curli is assembled at the cell surface and consists of CsgA, the major subunit of curli, and a membrane-associated nucleator protein, CsgB. Oligomeric intermediates that accumulate during the lag phase of amyloidogenesis are generally toxic, but the underlying mechanism by which bacterial cells overcome this toxicity during curli assembly at the surface remains elusive. Here, we elucidated the mechanism of curli amyloidogenesis and provide molecular insights into the strategy by which bacteria can potentially bypass the detrimental consequences of toxic amyloid intermediates. Using a diverse range of biochemical and biophysical tools involving circular dichroism, fluorescence, Raman spectroscopy, and atomic force microscopy imaging, we characterized the molecular basis of the interaction of CsgB with a membrane-mimetic anionic surfactant as well as with lipopolysaccharide (LPS) constituting the outer leaflet of Gram-negative bacteria. Aggregation studies revealed that the electrostatic interaction of the positively charged C-terminal region of the protein with a negatively charged head group of surfactant/LPS promotes a protein–protein interaction that results in facile amyloid formation without a detectable lag phase. We also show that CsgB, in the presence of surfactant/LPS, accelerates the fibrillation rate of CsgA by circumventing the lag phase during nucleation. Our findings suggest that the electrostatic interactions between lipid and protein molecules play a pivotal role in efficiently sequestering the amyloid fold of curli on the membrane surface without significant accumulation of toxic oligomeric intermediates.

Amyloid fibrils are exquisite nanoscopic aggregates of misfolded proteins and are associated with a plethora of deadly neurodegenerative diseases such as Alzheimer's, Parkinson's, Huntington's, and prion diseases (1)(2)(3)(4)(5). They are characterized by an ordered supramolecular architecture consisting of a common fold possessing a protease-resistant cross-␤ spine. An enlarging body of study has revealed that amyloid formation is an intrinsic property of a polypeptide chain regardless of initial structures (6,7). Extensive studies have been carried out to understand the molecular mechanism of amyloid formation. However, the molecular basis for amyloid propagation and cytotoxicity remains elusive (8 -10). A body of evidences has indicated that the oligomers are the key toxic species in the disease progression. Recent findings have revealed the existence of a variety of amyloid systems that play important functional roles in a wide range of organism from bacteria to humans (11,12). These amyloids are termed as functional amyloids. Functional amyloids are known to be involved in biofilm formation in bacteria, non-Mendelian inheritance traits in yeast, melanin biosynthesis in mammals, and in long-term memory in Aplysia, Drosophila, and possibly in mammals (12). Understanding functional amyloidogenesis will provide insights into how organisms have evolved to control amyloid propagation and minimize the associated cytotoxicity.
Curli is a functional amyloid expressed in the extracellular matrix of enteric bacteria such as Escherichia coli and Salmonella spp. (13,14). Curli biogenesis is involved in biofilm formation that mediates cell adhesion and interactions with host proteins that induce inflammatory responses (15)(16)(17). Unlike disease-associated amyloids, curli biogenesis has a tightly regulated pathway. The proteins encoded by csgBAC operon (curli-specific genes, csg) 2 along with four accessory proteins encoded by the csgDEFG operon are essential for the curli assembly (18,19). Curli is composed of two subunits, CsgA and CsgB. The solubility of these subunits inside the cell is maintained by CsgC and their secretion to the outer membrane is regulated by the accessory proteins encoded by csgDEFG operon (20 -22). Both CsgA and CsgB are secreted as intrinsically disordered proteins (14,23). On the cell surface, CsgA, the major subunit of curli is nucleated by membrane-anchored minor subunit CsgB (24). The first domain forms the secretary signal peptide that is responsible for targeting the proteins to the periplasm and is cleaved off in the periplasm (Fig. 1a) (25). Thus the mature CsgA and CsgB are devoid of secretary signal peptide. The subsequent 22 amino acids form the N22 domain, which helps CsgA to get transported onto the outer membrane; however, this domain is not essential for CsgB secretion (21,26). The major amyloidogenic domain consists five imperfect repeats (R1-R5) comprising of 19 -24 amino acids that forms the amyloid core (26,27). In vivo studies have shown CsgB is necessary for CsgA polymerization, whereas in vitro, CsgA aggregates even in the absence of CsgB (27,28).
The mechanism by which the bacteria bypass the toxic events during curli amyloidogenesis is poorly understood. The mechanistic studies of a variety of disease-associated proteins have revealed that amyloid formation typically follows a nucleation-dependent polymerization with a characteristic lag, elongation, and saturation phases (29,30). The transient oligomeric intermediates formed during lag phase are known to be highly toxic compared with the mature fibrils (31,32). Additionally, amyloid formation on the lipid membrane can disrupt the membrane integrity and can cause cell death (32)(33)(34)(35). Nevertheless, how enteric bacteria allow the curli assembly on its cell surface and bypass the toxic events remain poorly understood. We hypothesize that the bacterial membrane surface might be involved in the recruitment of the amyloidogenic precursors and thereby resulting in the formation of amyloid fibrils without accumulation of toxic intermediates. Outer membrane surface of Gram-negative bacteria is composed of lipopolysaccharide (LPS) containing negatively charged glycophospholipids (36). In this work, we aimed at dissecting the key molecular events of the intermolecular interactions between anionic lipids and curli subunits that allow the sequestration and subsequent polymerization into matured amyloid fibrils. Using a diverse range of tools, we first show that the electrostatic modulation between the protein and a model anionic detergent accelerates the aggregation, markedly leading to the disappearance of the typical lag phase during nucleation. Then we demonstrate that negatively charged LPS plays a pivotal role in promoting the heteronucleation during curli amyloid assembly.

A switch in the polymerization mechanism in the presence of an anionic detergent
A line of evidences suggested that CsgB gets tethered onto the membrane, which nucleates the aggregation of CsgA (24,26). Lipid A, a negatively charged glycophospholipid, forms the outer membrane of Gram-negative bacteria (36). To understand the molecular basis of CsgB membrane interactions, we began our studies using an anionic detergent namely SDS, which is a well-known anionic lipid mimetic (37). Subsequently, we carried out our experiments in the presence of LPS.
The function of curli nucleator CsgB is directed by its C-terminal. Without R4 and R5 repeats, CsgB is secreted away from the bacterial cell surface. Hence, we speculated that both R4 and R5 might have membrane interaction. To verify whether R4 and R5 interact with membrane, we carried out experiments using CsgB trunc or CsgB⌬R5 (19 amino acids are deleted from C-terminal of CsgB) and CsgB. In this study we refer to CsgB⌬R5 as CsgB t . C-terminal His-tagged CsgB t , CsgB, and CsgA were expressed and purified (see "Experimental procedures"). The purity was checked by SDS-PAGE (supplemental Fig. S1a) and the proteins of interest were confirmed by probing against anti-His antibody (supplemental Fig. S1b). First we monitored the aggregation kinetics of curli subunits using thioflavin T (ThT). CsgA, CsgB t , and CsgB exhibited typical nucleation-dependent polymerization kinetics with lag times of ϳ120, ϳ100, and ϳ30 min, respectively (supplemental Fig.  S1c). However, in the presence of preformed CsgB t seeds, a shortening of the lag phase in the CsgA aggregation was observed, which suggested that preformed CsgB t oligomers can nucleate the CsgA polymerization (supplemental Fig. S1c). These results are in good agreement with previous reports (23,38). Next, we investigated whether SDS at non-denaturing submicellar concentrations can modulate the course of CsgB t aggregation. A shortening in the lag time of CsgB t aggregation was observed with an increase in the SDS concentration ranging from 100 to 500 M. Interestingly, in the presence of 500 M SDS, CsgB t bypassed the nucleation phase as indicated by the absence of any apparent lag time (Fig. 1b). A complete elimination in the lag phase was also observed for CsgB in the presence of 500 M SDS (Fig. 1c). This observation suggests that the aggregation propensity of CsgB t /CsgB is accelerated in the presence of submicellar concentrations of SDS. This aggregation might follow a non-nucleation or an isodesmic polymerization mechanism as there is no strong concentration dependence on half-time (t1 ⁄2 ), when different protein concentrations were titrated with 500 M SDS (supplemental Fig. S1, d and e).
We next monitored the secondary structural changes by using far-UV circular dichroism (CD) spectroscopy. CsgB t adopts a random-coil structure with minima at 200 nm. As a function of time, CD spectra showed an increase in the ␤-sheetrich secondary structure (216 nm) at the expense of random coil (200 nm) and within 3 h there was a complete transition to the ␤-sheet (Fig. 1d). On the contrary, upon addition of SDS, CsgB t showed a rapid conformational conversion and the attained ␤-sheet structure within 30 min (Fig. 1e). The CD kinetics monitored at 200 nm validates that SDS induces spontaneous secondary structural changes during the fibrillation of CsgB t (Fig. 1f). We also monitored the backbone conformation of the amyloid state using vibrational Raman spectroscopy. The amide I Raman band at 1670 cm Ϫ1 exhibited by both (with and without SDS) fibrils indicated the formation of a cross-␤ structure, which is a hallmark of an amyloid (Fig. 1f) (39). Therefore, our ThT fluorescence assays together with the CD and Raman spectroscopic data show that the SDS-CsgB t interaction facilitates the conformational change and accelerates amyloid fibrillation by minimizing or eliminating the lag phase.

SDS induces rapid oligomerization of CsgB t
Next, we embarked upon studies aimed at elucidating the mechanism of CsgB t polymerization in the presence of SDS. To do this we have used 9-(2,2-dicyanovinyl)julolidine (DCVJ), a molecular rotor, which shows an enhanced fluorescence upon binding to the early oligomers of protein aggregates (40). DCVJ is weakly fluorescent in buffer, and as a function of CsgB t aggregation, a slow increase in DCVJ fluorescence was observed that was saturated by ϳ4 h (Fig. 2, a and b). Upon addition of SDS to CsgB t , there was a rapid increase in DCVJ fluorescence and reached a plateau within ϳ30 min. The DCVJ fluorescence Electrostatic modulation in curli assembly results indicated that SDS facilitates rapid oligomerization of CsgB t . Next, we performed the glutaraldehyde cross-linking assay to monitor the formation of soluble oligomers. We observed that the SDS-treated sample has much a lower content of monomeric protein compared with the untreated sample (supplemental Fig. S2a). The depletion of the monomer/ dimer population in the SDS-treated sample provided an additional evidence of early oligomerzation (supplemental Fig.  S2a). To gain insights into the nanoscale morphology of the transient intermediates, we next utilized atomic force microscopy (AFM). At early time points, CsgB t forms protofibrils of ϳ5 nm height (Fig. 2c). These protofibrils convert into matured fibrils upon prolonged incubation (Fig. 2d). Interestingly, the fibrils exhibited two distinct height distribution, 8 -10 and 20 -25 nm indicating that the fibrils are laterally associated to form bundles (supplemental Fig. S2b), as described previously for the CsgA fibrils (41). The bundle formation might be the possible reason for the high stability of curli and require formic acid to disintegrate these fibrils (14). In the presence of SDS, CsgB t formed spherical oligomers with an average height of 8 -10 nm (Fig. 2e). The oligomers convert into amyloid fibrils upon overnight incubation (Fig. 2f). Therefore, both our AFM and cross-linking results corroborated our DCVJ data, which suggested that the SDS-CsgB t interaction leads to the formation of higher order species. The absence of the lag phase observed in the kinetics of ThT fluorescence, CD, and DCVJ fluorescence indicate that the SDS-CsgB t interaction leads to the rapid formation of ␤-rich obligatory oligomers and allows the protein to bypass the nucleation phase. We next aimed at delineating the physicochemical basis of the SDS-CsgB t interaction.

The C-terminal end of CsgB interacts electrostatically with the anionic head groups
It was shown previously that the C-terminal region of CsgB containing the R4 and R5 are essential for anchoring and proper localization of the protein onto the membrane and without these repeats CsgB is secreted away from the cell (26). To monitor whether the C-terminal segment is responsible for the initiation of aggregation, we incorporated a tryptophan (Trp) at the C-terminal end (Y129W). CsgB t is devoid of Trp and therefore we envisioned that the single Trp mutant at R4 will report the early events of the aggregation. This Y129W mutant of CsgB t displayed aggregation kinetics similar to that of CsgB t indicating that the mutation did not alter the aggregation propensity of the protein (supplemental Fig. S3a). We next recorded several fluorescence readouts to monitor the aggrega-

Electrostatic modulation in curli assembly
tion process. Trp is an environmently-sensitive fluorescent probe (42). Before aggregation, Y129W showed an emission maxima at ϳ345 nm, which suggested that the Trp is solvent exposed. However, during the course of aggregation, there was a progressive blue shift of ϳ8 nm indicating that the C-terminal segment harboring Trp gets buried (supplemental Fig. S3b). Interestingly, in the presence of SDS, Y129W exhibited a blue shift in the emission maxima (ϳ334 nm) that did not undergo any further shift upon aggregation (supplemental Fig. S3c). This result indicated that the structure formation at the C-terminal segment of CsgB t was very rapid in the presence of SDS during the oligomerization (Fig. 3a).
Next, we monitored the fluorescence anisotropy that provides the information about the rotational mobility of the fluorophore (42). Without SDS, Y129W exhibited a low fluorescence anisotropy of ϳ0.06 (Fig. 3b). The low anisotropy value indicates that Trp resides in a highly flexible region of disordered CsgB t . During the aggregation, the anisotropy showed a typical nucleation-dependent kinetics and increased to ϳ0. 16. The increase in the fluorescence anisotropy is due to the disorder-to-order transition during aggregation that causes a restriction in the mobility of Trp. In the presence of SDS, the initial Trp fluorescence anisotropy is much higher (ϳ0.13) and reached a plateau (ϳ0.18) during the aggregation without a lag phase (Fig. 3b). The initial increase in the Trp anisotropy upon addition of SDS indicates the increase in the rotational hindrance arising due to the rapid oligomerization of CsgB t . Taken together, our Trp fluorescence results show that the C-terminal region of CsgB t harboring the R4 repeat interacts with anionic SDS. This region of CsgB t contains positively charged residues such as lysine (Fig. 1a). Therefore, we conjectured that the interaction between the C-terminal segment of CsgB t and SDS

Electrostatic modulation in curli assembly
is driven by the electrostatic interaction. To test this, we next carried out aggregation experiments in the presence of salt. The kinetics of aggregation in the presence of salt is much slower than in the absence of salt, indicating that higher ionic strength decelerates the aggregation of CsgB t even in the presence of SDS (Fig. 3c). Interestingly, with the same concentration of salt there is no change in the aggregation kinetics of CsgB, which indicates that the interaction of SDS with the R5 repeat is stronger than the R4 segment (supplemental Fig. S3d). As expected, CsgB t failed to aggregate in the presence of a positively charged detergent such as cetyltrimethylammonium bromide (supplemental Fig. S3e). This set of experiments revealed that the ionic interaction between CsgB t and SDS plays a key role in triggering the aggregation process.

C-terminal segment of CsgB t is involved in the formation of critical oligomers
The ionic interaction of proteins with the membranes or detergents can neutralize the charge and facilitate the proteinprotein interaction by increasing the local protein concentration on the surface. We next studied the formation of early oligomers that are formed by the interaction between the Cterminal segments of CsgB t . To monitor the protein-protein interaction we performed intermolecular fluorescence resonance energy transfer (FRET). For this set of experiments, we chose two C-terminal variants of CsgB t , Y129W (Trp as a donor) and Y129C was labeled with an acceptor dye, IAEDANS (5-((((2-iodoacetyl)amino)ethyl)amino)naphthalene-1-sulfonic acid)). Trp and IAEDANS are known to be an efficient FRET pair (42). We used FRET efficiency as the readout of the intermolecular interaction between CsgB t molecules. In the absence of SDS, CsgB t exhibited a low FRET efficiency (ϳ14%), whereas upon addition of SDS, the FRET efficiency increased sharply (ϳ36%) indicating the formation of early oligomers that bring together the C-terminal segments of CsgB t within the range of Förster distance (Fig. 3d). To establish further that the C-terminal contacts are critical for the oligomerization, we carried out aggregation of CsgB t Y129C in the presence and absence of a reducing agent, DTT (dithiothreitol). Upon addition of DTT, CsgB t Y129C aggregated similar to wild-type CsgB t , as expected (Fig. 3e). However, in the absence of DTT, the aggregation kinetics was much faster indicating that dimerization of the C-terminal segments through disulfide formation facilitates the oligomerization. In other words, the C-terminal dimer might be an on-pathway intermediate during the aggregation process. Taken together, our data suggested that the electrostatic interaction between SDS and the C-terminal segment of CsgB t results in the charge neutralization and leads to the formation of oligomeric precursors, which mature into amyloid fibrils. Next, we asked whether or not the SDS-

Electrostatic modulation in curli assembly
induced aggregates of CsgB t are capable of nucleating CsgA polymerization that is critical for curli biogenesis.

SDS-induced oligomers of CsgB t nucleates CsgA polymerization
CsgA aggregated with a lag time of ϳ2 h, however, in the presence of SDS, the lag time was shortened but the aggregation followed a typical nucleation mechanism (Fig. 4a). This result is in sharp contrast to SDS-induced aggregation of CsgB t , which exhibited a non-nucleation or an isodesmic polymerization mechanism. Even upon addition of a catalytic amount of CsgB t to CsgA (CsgA:CsgB t , 5:1) the aggregation kinetics was similar to that of CsgA (Fig. 4b). Interestingly, upon addition of SDS, CsgA:CsgB t aggregated rapidly without a lag phase and exhibited kinetics similar to CsgB t seed-induced aggregation of CsgA (Fig. 4b). We also characterized the structural transition from the disordered to the ordered ␤-rich state by monitoring the secondary structural changes. CsgA alone takes several days to undergo a complete transition to ␤-sheet-rich fibrils (28). However, we observed a relatively faster disordered-to-ordered transition of CsgA (supplemental Fig. S4, a and b). Even in the presence of CsgB t , CsgA undergoes conformational transition to ␤-sheet within ϳ3 h (Fig. 4, c and e), which is similar to CsgA alone. Upon addition of CsgB t and detergent, CsgA aggregates through a more facile alternate pathway akin to seeded polymerization kinetics (Fig. 4, d and e). This set of results indi-cated that at substoichiometric concentrations of CsgB t , SDS accelerates the polymerization rate of CsgA. Our next aim was to verify that even in the presence of catalytic amounts of CsgB t , SDS-induced CsgB t oligomers act as seeds for CsgA polymerization.

CsgB oligomerizes faster than CsgA
CsgA has a single Trp at the 109th position in the R3 region of the polypeptide chain. We monitored Trp fluorescence of CsgA in the presence of CsgB t . During the course of aggregation, there was a blue shift in the emission maxima from 345 to 337 nm indicating that the R3 region gets progressively buried (supplemental Fig. S5a). However, in the presence of SDS, the emission maxima underwent blue shift much faster and then remained constant during the late kinetics of the aggregation process (supplemental Fig. S5b). This result indicated that the solvent-exposed Trp in the disordered state gets buried at a much faster rate in the presence of SDS. Therefore, the time course of the spectral shift suggested that CsgA undergoes fibrillation relatively faster in the presence of SDS and CsgB t (supplemental Fig. S5c), but slower than SDS-induced aggregation of CsgB t (Fig. 3a). Next, we monitored the Trp anisotropy, without SDS the early fluorescence anisotropy of CsgA:CsgB t (5:1) was low (ϳ0.07), which increased to ϳ0.2 during aggregation. Even in the presence of SDS, CsgA exhibited low anisotropy (ϳ0.07) at 0 h, which then increased and plateaued at ϳ0.2  (Fig. 4f). On the contrary, upon addition of SDS, Y129W CsgB t showed a rapid change in the emission maxima and anisotropy of Trp at the onset of aggregation (Fig. 3, a and b). Therefore, our Trp fluorescence data validates that the rapid aggregation and the change in the secondary structural content of CsgA even in presence of the substoichiometric ratio of CsgB t is indeed due to the initial SDS-induced oligomers of CsgB t . These experiments suggested that SDS imparts an amyloid fold in the C-terminal end of CsgB t that leads to rapid oligomerization of the latter, which nucleates the polymerization of CsgA. Similarly, upon addition of substoichiometric amounts of CsgB and submicellar concentration of SDS to CsgA, the latter aggregated without exhibiting lag phase (Fig. 5a). The preformed SDS-induced oligomers of CsgB, even after diluting, are capable of nucleating aggregation of CsgA, suggesting that these obligatory oligomers are stable even upon dilution (supplemental Fig. S5d). Fig. 5, b-d, summarizes the kinetic parameter halftime (t1 ⁄ 2 ) recovered from the aggregation of curli subunits in the absence and presence of SDS obtained from the time course of different probes. Overall, our studies indicated that the electrostatic interaction between detergent and protein molecules facilitates curli assembly.

Lipopolysaccharide modulates curli assembly
Next, we asked whether curli assembly proceeds through a similar mechanistic control in the presence of LPS, the outer leaflet of Gram-negative bacteria. LPS is composed of three domains: lipid A, core oligosaccharide, and O-antigen. Lipid A

Electrostatic modulation in curli assembly
is a glycophospholipid consisting of a ␤-1,6-linked disaccharide of glucosamine, which is phosphorylated and fatty acylated (36). The phosphate groups impose negative charge on lipid A. To monitor the CsgB t -LPS interaction, LPS vesicles of ϳ100 nm (supplemental Fig. S6a) were used in concurrence with the other studies on LPS-protein interactions (43). First, we monitored ThT fluorescence to follow the aggregation of CsgB t in the presence of different concentrations of LPS vesicles. There is an increase in the rate of aggregation as a function of LPS concentrations, as expected from our SDS experiments. Upon addition of 50 g (per 1.5 ml) of LPS, both CsgB t and CsgB aggregated without a lag phase (Fig. 6a and supplemental Fig.  S6b), similar to what we observed with SDS. AFM imaging revealed that CsgB t undergoes spontaneous oligomerization in the presence of LPS (Fig. 6b) and eventually formed fibrils of different heights (Fig. 6c and supplemental Fig. S6c). The spherical oligomers formed in the presence of LPS is 8 -10 nm, which is similar to the size of SDS-induced oligomers of CsgB t (Fig.  2e). Thus, the ThT assay and AFM results in the presence of LPS are in line with those obtained from SDS-induced aggregation. Next, we monitored aggregation of CsgA:CsgB t (5:1) in the presence of LPS that exhibited a much faster kinetics compared with that of without LPS (Fig. 6d). CsgA:CsgB (5:1) aggregates were faster in the presence of LPS, suggesting that the interaction between the repeats and LPS promote the aggregation process without an apparent lag phase (Fig. 6e). Taken together, our results show that in the presence of LPS, the aggregation of curli-forming proteins is highly accelerated.  Fig. S6c. d, ThT assay for CsgA without (magenta) and with (green) LPS, 5:1 of CsgA:CsgB t (black), CsgA:CsgB t upon addition of 50 g/1.5 ml of LPS (red), CsgA in the presence of 10% CsgB t (blue) and CsgB t seeds were made from the aggregates of CsgB t formed in the presence of 50 g/1.5 ml of LPS. e, ThT fibrillation kinetics of CsgA:CsgB t (5:1) without (black) and with 50 g/1.5 ml of LPS (red) and CsgA:CsgB(5:1) (magenta) and upon addition of 50 g/1.5 ml of LPS vesicles (blue).

Discussion
Curli is a functional amyloid formed on the surface of Gramnegative bacteria. Within the bacteria, curli subunits exist as intrinsically disordered proteins and avoid aggregation by interacting with CsgC (20). In this work, we dissect the role of membrane in curli formation. Our aggregation studies on CsgB t using an array of biophysical and biochemical tools indicated that SDS, a membrane mimetic, binds to the protein through electrostatic interaction and promotes amyloid formation without a lag phase. We also showed that LPS, a primary component in the outer membrane surface of Gram-negative bacteria, also modulates the aggregation kinetics in a similar way. Binding to SDS/LPS neutralizes the positive charge of the protein molecules that facilitates intermolecular association leading to amyloid formation. Proteins upon binding to a surfactant such as SDS, at submicellar concentration, can induce the formation of clusters that can further grow into micelles and drive protein association (44). This ionic interaction with a detergent or a lipid is known to increase the local protein concentration, which can also promote protein aggregation (45,46). Our FRET measurements using the Y129C mutant of CsgB t , and cross-linking assay suggested that SDS promotes intermolecular interaction in the C-terminal segment of CsgB t and accelerates the aggregation of the protein, thereby minimizing the possibility of accumulation of toxic intermediates (supplemental Figs. S2a and 5b). In vivo study has previously demonstrated that the C-terminal of CsgB, especially R4 and R5 repeats are essential for proper localization on the membrane and curli assembly (26). R5 peptide fails to aggregate in vitro thus it was speculated that the membrane-bound C-terminal end might be devoid of amyloid fold thus the membrane integrity is not compromised (26). The ionic interaction between the R4/R5 repeat of CsgB t and SDS/LPS induced the formation of structured oligomers, which allow efficient polymerization of CsgA (Figs. 5b and 6, d and e). The binding of the positively charged C-terminal segment comprising R4 and R5 to SDS/LPS neutralizes the polypeptide chains via electrostatic interactions and therefore promotes critical chain-chain association that was otherwise hindered because of charge repulsion. Our findings allowed us to depict a simplistic model that is shown in Fig.  7. The facile recruitment of curli subunits on the LPS surface and subsequent fibrillation can potentially prevent the prolonged incubation of intermediates on the membrane: a mechanism by which bacteria might overcome the cytotoxicity during the course of aggregation of curli subunits.
Functional amyloid proteins might have evolved to undergo facile aggregation without significant accumulation of toxic oligomeric intermediates. For instance, Pmel17, a human functional amyloid involved in melanin synthesis undergoes a rapid pH-induced amyloidogenesis that is proposed to preclude the formation of its toxic intermediates (47). On the contrary, ␣-Synuclein, A␤ (amyloid-␤), and tau, which are implicated in neurodegenerative diseases, exhibit slow aggregation kinetics with a long lag phase (48). Mounting evidences suggest that the accumulation of amyloid precursors is the cause of cytotoxicity (32)(33)(34)(35). Doughnut-shaped prefibrillar oligomers of many disease-related proteins are known to be membrane active species (32). The oligomeric intermediates formed by ␣-Synuclein, A␤, islet amyloid polypeptide, polyglutamine, etc. are recognized by A11 antibody (49). CsgA oligomers are A11 active indicating that these oligomers possibly share structural similarity with the disease-associated amyloids (38). The minimum size of the transient intermediate of CsgA recognized by the A11 antibody is a monomer or dimer. CsgA dimers have been shown to be very stable and are observed even after treating the CsgA fibrils with formic acid (14). Our results indicate that in the presence of SDS the minimum oligomeric size of CsgB is a dimer that is consumed rapidly to form higher order fibrils. Taken together, our results show that the electrostatic interaction between CsgB and SDS/LPS-membrane promotes facile proteinprotein interaction leading to the sequestration of amyloid fold that finally nucleates the polymerization of CsgA. This membrane-assisted process is likely to prevent the accumulation of toxic species during the course of aggregation. Plausibly, bacterial membranes provide a platform that accelerates the dimer formation of the nucleator protein and facilitates the aggregation of curli subunits by sequestering the toxic intermediates.

Electrostatic modulation in curli assembly
tific. NEB3016 strain was procured from New England Biolabs. Nickel-nitrilotriacetic acid resin was obtained from Qiagen and PD-10 procured from GE Healthcare Life Sciences.

Expression, purification, and labeling
His-tagged CsgA, CsgB t , and CsgB were overexpressed in NEB3016 strain. CsgA and CsgB t were purified as described previously with slight modifications (50). A cell pellet from 500 ml of culture was lysed with 50 ml of 8 M GdnHCl in 50 mM potassium phosphate buffer, pH 7.3, and kept under stirring for 1 day. The lysate was centrifuged and the supernatant was sonicated for 2 min (pulse on: 30 s, pulse off: 20 s) and incubated with 1.5 ml of nickel-nitrilotriacetic acid resin at room temperature for 1 h. After 1 h the lysate was loaded onto a polypropylene column and the beads were washed with 7 column volumes of 50 mM potassium phosphate buffer, pH 7.3 (buffer A). Followed by another wash with 7 column volumes of buffer A containing 12.5 mM imidazole. The protein was eluted with buffer A containing 125 mM imidazole and passed through 30-kDa (Millipore Amicon) filter. The filtrate was subjected to PD-10 desalting column to remove imidazole. For the Y129C mutant of CsgB t , 10 M ␤-mercaptoethanol was added in all steps except during the PD-10 gel filtration. CsgB was purified as described previously (41) with slight modifications. The wash and elution buffers had 8 M urea and 150 mM NaCl in 50 mM potassium phosphate buffer, pH 7.3, and the eluent was directly subjected to PD-10 spin desalting column. The protein concentrations were estimated using ⑀ 280 of 10,810, 7,680, and 7,680 M Ϫ1 cm Ϫ1 for CsgA, CsgB, and CsgB t , respectively. Mutants of CsgB t were created by site-directed mutagenesis (primers sequences used are given in supplemental Table S1) and the concentration was determined using ⑀ 280 ϭ 12,090 M Ϫ1 cm Ϫ1 for Y129W and ⑀ 280 ϭ 6,400 M Ϫ1 cm Ϫ1 for Y129C. The molar extinction coefficients were obtained using Protein Calculator 3.4 (The Scripps Research Institute).
For labeling CsgB t Y129C with IAEDANS, the protein was purified using native buffer containing 10 M ␤-mercaptoethanol. The PD column was done in 6 M GdnHCl, 30 M DTT, buffer A, and the protein concentration was estimated and DTT was further added so that the final concentration for protein:DTT was 1:2. The resulting solution was stirred at room temperature for 30 min at 10 rpm. Then 10 eq of IAEDANS (according to protein concentration) were added from 200 mM stock prepared in DMSO and allowed to stir for 1 h at room temperature. Ten eq of IAEDANS were further added and incubated for 1 h. PD-10 column chromatography was done in buffer A. The protein concentration was estimated by measuring the absorbance at 280 and 340 nm. The concentration of labeled protein was calculated by using absorbance of IAEDANS at 340 nm (⑀ 340 ϭ 6100 M Ϫ1 cm Ϫ1 ).

Steady-state fluorescence
All aggregation reactions were carried out at room temperature in 50 mM potassium phosphate buffer, pH 7.3, and the protein concentration used was 8 M unless otherwise specified. All fluorescence experiments were performed on Fluoromax-4 spectrofluorometer (Horiba YJ). For the ThT fluorescence assay, the samples were mixed with 20 M ThT and incubated at room temperature. The samples were excited at 450 nm and emission was collected at 485 nm after every 10 min. The aggregation kinetic traces were fitted using the nucleation-dependent polymerization model (52). The excitation wavelength used for Trp was 295 nm and emission spectra were collected in the range of 310 -400 nm. The collected spectra were buffer corrected and normalized. The samples containing 5 M DCVJ were excited at 450 nm and emission spectra were collected in the range of 470 -550 nm. The steady-state fluorescence anisotropy of Trp was measured at 345 nm. The steadystate fluorescence anisotropy is given by the following equation, where, I and I Ќ are the parallel and perpendicular fluorescence intensities, respectively, with respect to the excitation polarizer. The perpendicular components were always corrected using a G-factor. The reaction mixture was stirred only during the measurements. The protein concentration used for all experiments was 8 M in 50 mM potassium phosphate, pH 7.3, except for CsgA:CsgB t (5:1) in which the concentration of CsgA was 8 M and CsgB t used 1.6 M. The seeds of CsgB t were prepared from 8 M CsgB t fibrils (50 mM potassium phosphate, pH 7.3), which were prepared by incubating at 25°C for 1 day under quiescent conditions. The fibrils were then subjected to bath sonication (37 kHz) for 1 min.
For FRET measurements, a mixture of 6 M CsgB t Y129W (donor) and 2 M AEDANS-labeled protein (acceptor) was excited at 295 nm and the spectra were recorded in the range of 320 -580 nm. The FRET spectra were corrected with respect to the direct excitation of AEDANS at 295 nm. The FRET efficiency was estimated using the following relationship (42), where, F DA and F D are the donor (Trp) fluorescence intensities in the presence and absence of acceptor (AEDANS), respectively.

CD experiments
CD experiments were performed on an Applied Photophysics Chirascan CD spectrophotometer. The CD spectra were recorded every one-half hour and scanned from 190 to 250 nm using 1-mm path length cuvette. Three spectra were averaged, blank subtracted, and smoothened using Pro-Data viewer and then plotted using Origin software. Single point CD at 200 nm was collected at 10-min intervals to monitor the kinetics. The concentrations used were 8 M CsgB t , CsgA:CsgB t (5:1) ϭ 8 M, and 1.6 and 500 M for SDS in 50 mM potassium phosphate buffer, pH 7.3. The samples were mixed properly before every measurement.

Cross-linking assay and Western blot analysis
To 8 M CsgB t in 50 mM potassium phosphate buffer, pH 7.3 (with and without 500 M SDS), glutaraldehyde was added so that the final concentration of the latter was 0.01%. The reaction was incubated at room temperature for 10 min. SDS-PAGE loading dye with ␤-mercaptoethanol was added to the reaction mixture and heated for 5 min at 100°C. The samples were run Electrostatic modulation in curli assembly on a 15% SDS-PAGE and transferred to a PVDF membrane (Amersham Biosciences) followed by blocking with 3% BSA in PBST (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 0.05% Tween 20) for 1.5 h at room temperature. The samples were washed three times with PBST and then probed with 1:3000 diluted anti-His primary antibody for 1.5 h at room temperature. After washing, the blots were incubated with 1:10,000 HRP-conjugated rabbit anti-mouse secondary antibody for 1 h at room temperature. The blots were developed using metal enhanced 3,3Ј-diaminobenzidine substrate.

AFM imaging
Ten l of sample was deposited on freshly cleaved and Milli-Q water-washed mica (muscovite grade V-4 mica from SPI, PA) and incubated for 5 min followed by washing with water. CsgB t sample (0 h) was diluted 100 times with water and deposited on mica. Then mica was dried under a stream of nitrogen gas for 15 min. The images were acquired using Innova (Bruker) AFM in tapping mode. The collected images were processed and analyzed using WSxM 5.0 (51).

Raman spectroscopy
The Raman spectra were recorded on an inVia TM Raman microscope (Renishaw, UK). The aggregation reaction incubated overnight at 25°C was centrifuged at 13,000 rpm for 15 min and the supernatant was discarded keeping ϳ100 l of solution. The resultant solution was deposited on a glass slide and air dried. The sample was focused by 50ϫ objective lens (Nikon, Japan) and was excited using 785 nm HPNIR laser. A 1200 lines/mm grating was used to disperse the scattered light. Data were acquired using Wire 3.1 software provided with the instrument. The spectra were baseline corrected using the cubic spline interpolation method for eliminating the tilt and smoothed using Wire 3.1 software and plotted in Origin 8.5 (39).

Preparation of LPS vesicles
One mg/ml of LPS was dissolved in 50 mM potassium phosphate buffer, pH 7.3. The mixture was sonicated in bath sonicator (37 kHz) for 15 min at room temperature. The sizes of the vesicles were determined using Malvern Zetasizer Nano-ZS90 instrument.