|Home | About | Journals | Submit | Contact Us | Français|
Amyloid is associated with debilitating human ailments including Alzheimer’s and prion diseases. Biochemical, biophysical, and imaging analyses revealed that fibers produced by Escherichia coli called curli were amyloid. The CsgA curlin subunit, purified in the absence of the CsgB nucleator, adopted a soluble, unstructured form that upon prolonged incubation assembled into fibers that were indistinguishable from curli. In vivo, curli biogenesis was dependent on the nucleation-precipitation machinery requiring the CsgE and CsgF chaperone-like and nucleator proteins, respectively. Unlike eukaryotic amyloid formation, curli biogenesis is a productive pathway requiring a specific assembly machinery.
Bacteria express a variety of cell-surface proteinacious filaments that can promote colonization of an epithelial surface, entry into host cells, exchange of DNA between bacteria, and development of bacterial communities organized as biofilms, colonies, or multicellular fruiting bodies. Curli are a class of highly aggregated, extracellular fibers expressed by Escherichia and Salmonella spp. that are involved in the colonization of inert surfaces and biofilm formation (1, 2) and mediate binding to a variety of host proteins (3-5).
Polymerized curli appear as 4- to 7-nm-wide fibers of varying lengths by negative-stain electron microscopy (EM) (6). Under high-resolution EM, curli appeared as a tangled and amorphous matrix surrounding the bacteria (Fig. 1A) (7). At higher magnifications, curli fibers appeared as ~6- to 12-nm-wide fibers of varying lengths (Fig. 1, B and C).
Curli were purified from MC4100 by sequential differential centrifugation (S6) and analyzed by SDS–polyacrylamide gel electrophoresis (PAGE) (8). Resolution of CsgA, the major structural component of curli, required brief treatment with 90% formic acid (FA) to depolymerize the CsgA polymers into a ~17.5-kD protein and two minor proteins that migrated at ~30 and 32 kD (Fig. 1D). Only the 17.5- and 32-kD bands were unique to FA-treated samples, and these bands were recognized by antibodies to CsgA (anti-CsgA) (9). The migration of these proteins is consistent with monomer and dimer sizes of CsgA (10, 11). By EM, non-FA–treated S6 curli were indistinguishable from those presented naturally on the bacterial surface, appearing as aggregated fibers of varying lengths and widths (compare Fig. 1, E and F). Circular dichroism (CD) analysis indicated that these fibers were rich in β-sheet secondary structure with a minimum peak at ~218 nm (Fig. 2A).
Like other amyloid fibers, S6 curli induced a spectral change of a 10 μM Congo red (CR) solution with a maximum difference in absorbance between CR alone and CR bound to curli fibers at ~541 nm (Fig. 2, B and C) (12). The curli used in these assays contained a small amount of contaminating proteins (Fig. 1D). Pure, intact curli (called GP curli for “gelpurified”) were isolated as described (10) (Fig. 1D). GP curli retained the ability to bind CR and cause the red shift, demonstrating that curli were sufficient to augment the absorbance of CR (Fig. 2B). Addition of purified S6 curli to a 5 μM solution of thioflavin T (ThT) resulted in a fluorescence emission maximum at 482 nm (Fig. 2D), which is identical to the fluorescence induced by other amyloid proteins (13, 14).
Amyloid formation in eukaryotic cells is thought to be the result of a misguided protein-folding pathway. In contrast, E. coli possesses a specific nucleation-precipitation machinery encoded by the csgAB and csgDEFG operons to assemble curli. CsgB is thought to nucleate CsgA fibers (15). The csgDEFG genes encode CsgD, a FixJ-like transcriptional regulator, and three putative curli assembly factors, CsgE, CsgF, and CsgG. The lipoprotein CsgG localizes to the inner leaflet of the outer membrane and may serve as a curli assembly platform (16).
A nonpolar csgF− deletion mutant (MHR592) (17) resulted in aberrant CR binding properties. Wild-type bacteria stained CR-positive after 30 hours of growth on YESCA plates (9). Strain MHR592 (csgF−) was CR-negative after 30 hours of growth and only slightly CR-positive after 48 hours of growth (9). EM analysis of csgF− bacteria showed that fibers were less abundant but were otherwise indistinguishable from those produced by wild-type bacteria (Fig. 3A).
The monomeric and polymeric state of CsgA in the absence of CsgF was assessed. Very little SDS-soluble CsgA was present in extracts made from wild-type bacteria (Fig. 3B) because most of the CsgA subunits were assembled into curli as determined by the presence of a 17.5-kD band after pretreatment with FA (Fig. 3B). Similar to a csgB− mutant (Fig. 3B), most CsgA produced by a csgF− mutant remained in an SDS-soluble form after 48 hours of growth (Fig. 3B).
CsgA is secreted in a soluble, assembly-competent form by a csgB− mutant (CsgA+ donor) and can be assembled on the surface of csgA− mutant (CsgB+ recipient) bacteria in a process called interbacterial complementation (Fig. 3D) (18). CsgB+ recipient cells lacking the CsgA protein stained CR-positive whencross-streaked with either csgF− or csgF−B− double-mutant bacteria (Fig. 3D), indicating that CsgA was secreted from csgF− cells and assembled on the CsgB+ recipient cells. In contrast, csgF− and csgF−B− mutants were unable to accept CsgA from a CsgA+-donating strain (Fig. 3D). Thus, the curli assembly defect in csgF− mutants was a nucleation defect because CsgA produced by these cells was assembly competent. CsgF may work independently or in concert with CsgB to guide in vivo extracellular nucleation of CsgA.
A nonpolar csgE− deletion mutant (MHR480) (17) produced pale, non CR–binding colonies similar to those produced by a csgA mutant (9). Despite the pale-colony phenotype, MHR480 (csgE−), but not a csgE−A− double mutant, produced fibers that reacted with anti-CsgA. However, these structures were less abundant than wild-type curli and were architecturally distinct in that they tended to arrange into rings (Fig. 3C). In the csgE− cells, no SDS-soluble CsgA could be detected, and the total amount of SDS-insoluble CsgA was markedly reduced (CsgA was detected in the FA-treated sample only after extended exposure) (Fig. 3B). A csgE− mutant was unable to donate CsgA subunits when cross-streaked against the CsgB+ recipient (Fig. 3D). However, a csgE−mutant retained the ability to act as a recipient and guide CsgA polymerization, because it stained CR-positive when cross-streaked against a CsgA+ donor (Fig. 3D). This staining was weaker than that observed on a CsgB+ recipient cross-streaked against a CsgA+ donor (Fig. 3D), suggesting that in addition to the CsgA stability defect, csgE− bacteria are also partially defective in their ability to nucleate exogenous CsgA. A csgB−E− double mutant was unable to accept or donate CsgA (Fig. 3D).
To understand the requirements for subunit polymerization, we purified CsgA in a soluble form and analyzed its polymerization in vitro. A six-histidine–tagged version of csgA (CsgA-his) was cloned behind the IPTG (isopropyl-β-D-thiogalactopyranoside)-inducible trc promoter in pHL3 (17), to create pMC3 (19). Plasmid-derived CsgA-his complemented fiber formation and CR binding in a csgA− mutant (9). To produce soluble, nonpolymerized CsgA, we transformed pMC3 (csgA-his) into LSR6 (C600:ΔcsgDEFG;ΔcsgBA) (19). However, attempts to detect CsgA-his protein expressed in LSR6 were unsuccessful (Fig. 4A). When LSR6/pMC3 (csgA-his) was transformed with pMC5 (csgEFG) or pMC1 (csgG), CsgA-his could be detected in the culture supernatants (Fig. 4A). Under these growth conditions (logarithmic growth in LB broth), only CsgG, and not CsgE, was required for efficient CsgA stabilization and secretion.
CsgA-his was purified from the supernatant of LSR6/pMC3/pMC5 and LSR6/pMC3/pMC1 by standard nickel affinity chromatography (Fig. 4B). Immediately after elution from the nitrilotriacetic acid (NTA) agarose column, solutions containing purified CsgA-his were clear with no evidence of aggregation, and EM of this material revealed no fibers (9). CD analysis of this material indicated that soluble CsgA-his, unlike curli, was not rich in β-sheet secondary structure (Fig. 4C). However, after prolonged incubation (4°C for 4 to 12 hours), the CsgA-his solutions became opaque and noticeably viscous. EM analysis revealed that fibers had formed that were similar to those produced by wild-type bacteria (Fig. 4D). The in vitro–assembled CsgA-his fibers, but not the CsgA-his soluble precursors, were able to bind CR and cause a red shift (Fig. 4E), signifying that they had adopted the cross β structure conserved in all amyloid fibers. CsgA purified from cells expressing csgEFG or only csgG formed CR-binding fibers with indistinguishable kinetics (9). Thus, although CsgB and CsgEFG are required to facilitate efficient polymerization in vivo, they are not required for polymerization to proceed under these in vitro conditions. A critical function of the nucleation-precipitation assembly machinery may be to prevent CsgA polymerization within the cell and accelerate it at the cell surface.
Our demonstration that E. coli can produce extracellular amyloid-like fibers increases the recognized functional repertoire of amyloid fibers and provides a useful model system to study their formation. Purified amyloid fiber subunits associated with human diseases, such as the Aβ protein associated with Alzheimer’s disease, readily polymerize when incubated at high concentrations in vitro (20). However, the accessory proteins and conditions that facilitate in vivo polymerization of Aβ are incompletely understood. Understanding the regulation of curli subunit polymerization in E. coli will offer insight into the formation of eukaryotic amyloids. This work also raises the intriguing possibility that bacterial amyloids could play a role in certain human neurodegenerative and amyloid-related diseases. Future experiments will further examine the role of CsgB, CsgE, and CsgF during the in vivo polymerization of curli, and their function will be used as a model to understand the formation of other amyloids.
We thank members of the Hultgren lab and especially K. Dodson for helpful comments during the preparation of this manuscript. This work was supported by the Alzheimer’s Disease Research Center grant NIA P50 AG05681-17 and NIH grants AI29549, DK51406, and AI48689 (S.J.H.). S.N. acknowledges grants from the Swedish Medical Research Council (16x-10843) and Swedish Natural Science Research Council (3373-309). M.R.C. was supported by a Keck fellowship and by NIH fellowship 1 F32 AI10502-01A1.