Epigallocatechin Gallate Remodelling of Hfq Amyloid-Like Region Affects Escherichia coli Survival

Hfq is a pleiotropic regulator that has key roles in the control of genetic expression. The protein noticeably regulates translation efficiency and RNA decay in Gram-negative bacteria, due to the Hfq-mediated interaction between small regulatory noncoding RNA and mRNA. This property is of primary importance for bacterial adaptation and virulence. We have previously shown that the Hfq E. coli protein, and more precisely its C-terminal region (CTR), self-assembles into an amyloid-like structure. In the present work, we demonstrate that epigallocatechin gallate (EGCG), a major green tea polyphenol compound, targets the Hfq amyloid region and can be used as a potential antibacterial agent. We analysed the effect of this compound on Hfq amyloid fibril stability and show that EGCG both disrupts Hfq-CTR fibrils and inhibits their formation. We show that, even if EGCG affects other bacterial amyloids, it also specifically targets Hfq-CTR in vivo. Our results provide an alternative approach for the utilisation of EGCG that may be used synergistically with conventional antibiotics to block bacterial adaptation and treat infections.


Introduction
Antibiotic resistance occurs when a compound has lost its ability to kill bacteria. This problem is particularly critical for public health and many efforts have been done to modify and improve existing antibiotics. Indeed, antibacterial agents use four main mechanisms; inhibition of cell wall synthesis (β-lactams), inhibition of protein synthesis (aminoglycosides, tetracyclines), inhibition of nucleic acid synthesis (quinolones) or inhibition of metabolic pathways (sulfonamides). However, the main problem with this therapeutic approach is the development of new acquired resistances. These usually involve mutations in genes of the antibiotic target or acquisition of foreign DNA, allowing antibiotic modification, destruction or efflux [1]. Therefore, the search for novel antibacterial-targets is of utmost importance.
Recent research has aimed at exploring potential targets within proteins involved in the regulation of the genetic expression. This includes proteins involved in bacterial adaptation to their host

EGCG Remodels Hfq CTR Amyloid Fibrils
To investigate the possibility that Hfq functional amyloids can be remodeled in vitro, we screened the effect of EGCG on Hfq-CTR fibrils using transmission electron microscopy (TEM, see Section 4.3). In order to allow a statistical analysis on the sample, we used a negative staining procedure and not cryoTEM to reduce the length of acquisition. Note that thioflavin T staining is a widely used method to follow amyloids self-assembly. Nevertheless in the case of Hfq-CTR, thioflavin inhibits fibril formation and cannot be used [41]. As a model we first used the 11 amino acid residues peptide corresponding to the Hfq amyloid region only, referred to as Hfq CTR 11 [42] (see sequence in Materials and Methods, Section 4.2). The effect of the EGCG's potential amyloid interference was tested on preformed CTR 11 fibrils (50 mg/mL). Fibrils were incubated in EGCG solutions (from 1 to 5 mM) for 24 h and examined by TEM (Figure 1). In the absence of the compound, the fibrils were abundant (Figure 1a), while incubation with 1 mM EGCG significantly reduced the amount of fibrils and led to formation of diffuse aggregates close to the fibrils (Figure 1b). At higher concentrations only rare fibrils can be observed (Figure 1c,d). This effect was confirmed with the 38 amino acid residues peptide corresponding to the full Hfq-CTR region ( Figure 2). The corresponding peptide is referred to as Hfq CTR 38 [34] (sequence in Materials and Methods, Section 4.2). TEM analyses did not allow for observing the total disappearance of CTR 38 pre-formed fibrils when EGCG was added, in contrast to Hfq CTR 11 . Nevertheless, we clearly see lower amount of fibrils, which confirms that EGCG also disrupts CTR 38 fibrils, even if less efficiently than in the case of CTR 11 (Figure 2a,b) Note that spontaneous self-assembly of CTR 38 is significantly lower than for CTR 11 (weeks vs days) and that CTR 38 peptide has been incubated for a few weeks to assure self-assembly before adding EGCG for 24 h [34,43]. To know whether EGCG can inhibit the formation of new fibrils and not only disrupt pre-formed ones, its effect on the conversion of Hfq monomers into amyloid fibrils was also investigated. This analysis was not possible with CTR 11 , as this peptide forms fibrils within a few hours and EGCG disrupts pre-formed fibrils in the same timeframe. We thus used Hfq CTR 38 fibrils, which are more stable, for our analysis. To induce fibrillation of CTR 38 within a shorter time (few days) we added DNA, which has been shown to accelerate fibril formation [43]. TEM analyses showed EGCG inhibition of the formation of Hfq CTR 38 (Figure 2c,d). This indicates that EGCG does not only disrupt Hfq CTR preformed fibrils, but also inhibits their fibrillation.

Characterization of Hfq Amyloid Region Stability in the Presence of EGCG
We then analysed quantitatively the effect of EGCG on Hfq CTR stability. For this goal we used circular dichroism (CD). Precisely, we used synchrotron radiation circular dichroism (SRCD) for our analysis, which allows the wavelength range to be extended down to 170 nm for the identification and distinction of amyloid signals [44] (see Section 4.4.). Aggregation into β-sheets in an amyloidal structure implies significant SRCD spectral changes: inversion of the band at~200 nm from a positive to a negative value (Figures 3 and 4); in parallel, a negative band decreases and shifts from~225 nm tõ 210 nm ( Figure 3). The inversion of the band at~200 nm and the blue-shift at 225 nm observed upon EGCG addition (5 mM, Figure 3) proves that the amyloid structure is disrupted by the compound, while the same peptide in the absence of the compound remains assembled. Precisely, in the presence of EGCG, relaxed antiparallel β-sheets (referred to as β anti 2, [45]) convert into right-handed twisted antiparallel β-sheets (referred to as β anti 3, [46]) (  Dissociation of the amyloid structure is observed by a shift of the negative peak from 226 to 212 nm [47]. The decrease of peak intensities also confirms disassembly. Path length = 1 µm, peptide concentration 12.5 mg/mL. Note that a similar effect is also observed with Hfq CTR 38 , even if in this case the blue shift is less noticeable (not shown). To quantify this effect on CTR 11 , we then analysed SRCD melting curves of preformed fibrils of CTR 11 in the presence of EGCG ( Figure 4).
The respective secondary structure compositions at different temperatures were determined using Bestsel and are reported in Table 2. As expected, at room temperature the peptide is partially folded into β-sheets, mainly in relaxed antiparallel β-sheets (β anti 2) [47]. Note that adding EGCG immediately results in the β anti 3 structure melting at 20 • C ( Table 2).   Table 2). The melting temperatures measured from SRCD values at 198 nm are Tm = 62.0 ± 1 • C for CTR 11 alone (c) and Tm = 55.6 ± 1 • C in the presence of EGCG (d). (e,f) Van't Hoff's plots giving access to ∆H • and ∆S • of CTR 11 melting. The isosbestic point indicates a two state model and a Boltzmann sigmoid equation has been used for melting curves fitting. When temperature increases, relaxed antiparallel β-sheets (β anti 2) convert into right-handed twisted antiparallel β-sheets (β anti 3) [47] ( Table 2). The signature of this conversion is mainly observed at 198 nm (Figure 4a,b) [47]. We thus analysed the melting temperature of the assembly based on the SRCD signal at 198 nm. We confirmed the disrupting effect of epigallocatechin-3-gallate (EGCG) as melting temperature decreases significantly from Tm = 62 ± 1 to Tm = 55.6 ± 1 • C in the presence of EGCG (Figure 4b,d). The Tm in absence of EGCG was confirmed using differential scanning calorimetry (DSC) (Tm = 63.9 ± 1 • C, Supplementary Figure S1). DSC analysis in the presence of EGCG did not yield interpretable data as EGCG alone gives a DSC signal close to that of the peptide. Van't Hoff's plots were drawn from melting curves (Figure 4e,f), giving access to ∆H • and ∆S • of CTR 11 fibrils disassembly in the absence or presence of EGCG. For CTR 11 fibrils alone, ∆H • = 75.06 ± 4.5 kJ·mol −1 and ∆S • = 0.225 ± 0.13 kJ·mol −1 K −1 , while for CTR 11 in the presence of EGCG ∆H • = 61.9 ± 4.9 kJ·mol −1 and ∆S • = 0.193 ± 0.015 kJ·mol −1 K −1 . In both cases, we confirm disassembly is an endothermic process. Furthermore, we confirm that fibrils are significantly less stable in the presence of EGCG.

EGCG Affects Bacterial Survival Due to Its Interaction with Hfq CTR
Finally, the effect of EGCG on bacterial survival has been evaluated. For this goal, different bacterial strains have been used, namely MG1655 WT cells [48], MG1655 ∆hfq and MG1655 hfq72 (deleted for the last 90 hfq nucleotides giving a truncated protein with only the first 72 amino acids, this strain can also be referred to as hfq∆ctr, see Section 4.5.) [43]. As EGCG is a highly absorbing and fluorescent molecule with spectral properties depending on solvents [49], the use of bacterial viability kits based on fluorescence or colorimetric assay was not possible for this analysis [50]. The plate count method has thus been used to evaluate the concentration of the antimicrobial agent, here EGCG, which inhibits bacterial growth. Precisely, colony forming units (CFU) have been measured and the C 50 , the concentration of EGCG that decreases CFU by 50%, has been determined (see Section 4.6.). The C 50 was 0.67 ± 0.02 mM for the WT strain vs 0.81 ± 0.05 mM for hfq72. Thus, the E. coli strain devoid of Hfq-CTR is less sensitive to EGCG than the WT strain. Note that due to the pleiotropic effect of the protein, the total absence of Hfq has a major effect on bacterial survival in the presence of EGCG (C 50~0 .4 ± 0.1 mM). This was expected as Hfq and sRNA establish resistance to various drugs, including antibiotics, in bacteria [9,18,51].

Discussion
Green tea polyphenolic catechins, including EGCG, have been shown to have antibacterial properties [52]. These properties arise from different mechanisms, including damage to bacterial cell membrane, inhibition of fatty acid synthesis, inhibition of enzyme activities (DNA gyrase, dihydrofolate reductase, etc) or act as an antibiofilm agent [53][54][55][56]. In this work we show that in addition to these roles, the EGCG catechin is also able to block the adaptability of bacteria when colonizing its host. Precisely, EGCG acts on the E. coli Hfq protein and inhibits its self-assembly [34]. Note that antibacterial effects reported previously for EGCG may be partially attributed to the inhibition of Hfq function, precisely in membrane and DNA-related processes or in biofilm formation [56][57][58].
In this work, we fully characterize the effect of EGCG on Hfq amyloid region self-assembly. As expected, we observe that Hfq-CTR amyloid disassembly is an endothermic process and we show that adding EGCG significantly decreases the enthalpy of the reaction (by~20%). This suggests that EGCG acts on CTR 11 oligomers mainly through disruption of hydrogen bonds. Hydrophobic interactions (entropy-driven) do not seem to be significantly involved. This is however not surprising taking into account the low number of hydrophobic amino acid residues present in the amyloidogenic region of Hfq [59]. Even if EGCG, the natural product, emerges as a promising new antibacterial agent against Gram-negative bacteria and exhibits alone antibacterial activity, this compound may be improved to more specifically and efficiently target the Hfq amyloid region. Catechins from other plants may alternatively be more specific. Otherwise, the galloyl moiety of EGCG could be chemically modified with the perspective of having more specific anti-Hfq properties [60]. Note that the in vitro inhibition concentration and C 50 values we measured (in the mM range) are close to the minimal inhibitory concentration (MIC) reported for other Gram-negative bacteria (about 1 mM, [61,62]). MICs in Gram-negative bacteria are significantly higher than that in Gram-positive bacteria, probably due to protection by the outer membrane [62]. This C 50 concentration is significantly higher than the equilibrium dissociation constant (K D ) of the EGCG/CTR 11 complex, which is in the µM range (Supplementary Figure S2). This suggests that the property of EGCG that must be optimised is its disassembling efficiency, and not its binding to CTR 11 . Nevertheless, the precise 3D structure of this region of the protein is still unknown [23], and rational design of modified catechins to target Hfq-CTR is thus challenging.

Chemicals
All chemicals including EGCG were from Sigma-Aldrich. EGCG was prepared in milliQ water at 10 mM.

TEM Imaging of Protein Fibrils
EGCG was added to fibrils (final concentration from 1 to 5 mM). Except when specified, samples were incubated for 24 h and visualized by TEM. To perform negative staining, 5 µL of peptide sample was deposited on a glow-discharged carbon-coated electron microscopy copper grid (continuous carbon film on 200 mesh square copper grid, EMS). After 2 min, the excess of sample was blotted out using a Whatman filter paper. Then, 5 µL of contrasting agent solution (Gadolinium salt, uranyl-less) was applied onto the grid with peptide. After 1 min incubation, the excess of contrasting agent was blotted out and then the grids were kept in a dry dark dust-free environment until observation with the electron microscope. For sample observation, the electron microscopy grid was mounted onto a room temperature holder and introduced into a JEOL 2200FS electron microscope (JEOL, Tokyo, Japan). The TEM images (2048 × 2048 pixels) were acquired using a Gatan UltraScan894 US1000 slow scan CCD camera at 40,000× nominal magnification (corresponding pixel size was 0.32 nm). The images presented in this work are representative of several tens of images that have been collected on 2 independent samples.

Synchrotron Radiation Circular Dichroism (SRCD)
For SRCD analysis, measurements and data collection were carried out on DISCO beam-line at the SOLEIL Synchrotron (proposal # 20180165) [63]. Samples of 2-4 µL were loaded into circular demountable CaF 2 cells of 1 micron path length and analysed at different incubation time [64]. Separated data collections were carried out to ensure repeatability. Spectral acquisitions of 1 nm steps at 1.2 integration time, between 320 and 180 nm were performed in triplicate for the samples as well as for the baselines. (+)-camphor-10-sulfonic acid (CSA) was used to calibrate amplitudes and wavelength positions of the SRCD experiment. Data-analyses including averaging, baseline subtraction, smoothing, scaling and standardization were carried out with CDtool. Secondary structure content was determined using BestSel [47,65]. Normalized root-mean-square deviation (NRMSD) indicated the most accurate fit for each spectrum; values of <0.15 were considered significant. For melting curves, SRCD spectra were acquired every 5 • C between 20 • C and 95 • C (in 1 micron path length cell) for CTR 11 ,CTR 11 in the presence of EGCG and EGCG alone. SRCD values at 198 nm are presented as a function of temperature to measure the melting point (Tm). A Boltzmann sigmoid equation, which assumes a two state model, has been used for fitting of melting curves: y = Bottom + (top − bottom)/(1 + e((Tm − x)/slope)). The energetics involved in self-assembly was determined by a van't Hoff plot. Fractions of folded (f n ) or unfolded (f u ) structures were measured from melting curves. The equilibrium constant K was calculated as f u /f n . ∆G • the effective change in standard free energy of reaction is defined as ∆G • = ∆H • -T∆S • = -RT ln K, where ∆H • is the change in standard enthalpy of reaction, ∆S • the change in standard entropy of reaction, R the gaz constant and T the absolute temperature (expressed in Kelvin). ∆S • and ∆H • are extracted directly from the van't Hoff plot where R lnK is plotted against 1/T (R lnK = ∆S • − ∆H • × 1/T), with −∆H • the slope and ∆S • the intercept at origin of the straight-line. Note that thermodynamic parameters may have an incertitude of~10% due to the uncertainty for the lower and upper limit of the melting curves.

Construction of E. coli Strains
Strains were constructed with the λ-red recombination technique, as described in Malabirade et al. [43].

Effect of EGCG on E. coli Survival
The plate count method has been used to evaluate the efficiency of EGCG on bacterial EGCG. This method involves the incorporation of different concentrations of the antimicrobial substance into a LB agar medium followed by the application of a standardized number of cells to the surface of the agar plate [66]. Briefly in our case E. coli MG1655 strains were grown overnight in LB medium (37 • C at 120 rpm), diluted, and grown to an OD 600 of 0.6. Diluted samples (from 10 −5 to 10 −7 ) were plated on agar plates in triplicate after overnight incubation. The experiment was repeated twice with independent cultures. Colonies were counted to determine CFU, an estimate of viable bacteria. To ensure no colonies grow slowly, plates were kept at 37 • C for 2 extra days. The concentration of the assayed antimicrobial agent (here EGCG) that decreases CFU by 50% has been measured in each case to evaluate the activity of EGCG as an antimicrobial agent. This value is referred as C 50 in this manuscript.

Conclusions
Natural or modified EGCG are already known to work synergistically with other antibiotics against E. coli, K. pneumoniae or A. baumanii isolates. Here we show that they may be used efficiently with conventional antibiotics (such as β-lactams or chloramphenicol) to block bacterial adaptation [67][68][69][70]. This should allow extending the spectrum of agents against multi-resistant and biofilm forming bacteria, and in general provide an efficient anti-infective compound, especially as EGCG has also antifungal and antiviral properties [71].