Amyloidogenic amyloid-β-peptide variants induce microbial agglutination and exert antimicrobial activity

Amyloid-β (Aβ) peptides are the main components of the plaques found in the brains of patients with Alzheimer’s disease. However, Aβ peptides are also detectable in secretory compartments and peripheral blood contains a complex mixture of more than 40 different modified and/or N- and C-terminally truncated Aβ peptides. Recently, anti-infective properties of Aβ peptides have been reported. Here, we investigated the interaction of Aβ peptides of different lengths with various bacterial strains and the yeast Candida albicans. The amyloidogenic peptides Aβ1-42, Aβ2-42, and Aβ3p-42 but not the non-amyloidogenic peptides Aβ1-40 and Aβ2-40 bound to microbial surfaces. As observed by immunocytochemistry, scanning electron microscopy and Gram staining, treatment of several bacterial strains and Candida albicans with Aβ peptide variants ending at position 42 (Aβx-42) caused the formation of large agglutinates. These aggregates were not detected after incubation with Aβx-40. Furthermore, Aβx-42 exerted an antimicrobial activity on all tested pathogens, killing up to 80% of microorganisms within 6 h. Aβ1-40 only had a moderate antimicrobial activity against C. albicans. Agglutination of Aβ1-42 was accelerated in the presence of microorganisms. These data demonstrate that the amyloidogenic Aβx-42 variants have antimicrobial activity and may therefore act as antimicrobial peptides in the immune system.

Gram staining. The same cultures analysed by flow cytometry were used for the gram staining. After incubating the bacteria and C. albicans with the peptides, 10 μ l of the cell suspensions were air-dried on microscope slides and flame-fixated. The smears were flooded with crystal violet solution for 2 min. The slides were rinsed and then incubated for 2 min with iodine solution. The slides were then flooded with 96% ethanol for approximately 10 seconds and washed with distilled water. The slides were flooded with safranin for 1 min, washed with distilled water and examined on an Olympus IX 70 microscope (Olympus, Hamburg, Germany) with NIS Elements BR software (Nikon, Duesseldorf, Germany) using a 100× oil immersion objective (1000× magnification).
Immunocytochemistry. The immunocytochemistry of E. faecalis and S. aureus was examined using the Gram preparations specified above. To achieve the robust adherence of pathogens on the glass slides during the staining procedure, cells were incubated with 2% paraformaldehyde (Sigma-Aldrich, Munich, Germany) on the slides and air dried for 60 min at room temperature, followed by washing with PBS (Biochrom, Berlin, Germany). Slides were blocked with 1% bovine serum albumin/5% goat serum (Sigma-Aldrich, Munich, Germany) in PBS for 1 h at room temperature before incubation with the Aβ peptide-specific monoclonal antibody 6E10 (Covance, Princeton, New Jersey, USA) at a concentration of 4 μ g/ml for 60 min at room temperature. Labeling with the secondary AF488-conjugated goat anti-mouse antibody (Life Technologies, Darmstadt, Germany) at a concentration of 2 μ g/ml was performed for 1 h at room temperature. Slides were mounted with Roti ® -Mount FluorCare DAPI (Roth, Kralsruhe, Germany), and photographs were taken on an automated screening epifluorescence microscope (Eclipse Ti-E, Nikon, Duesseldorf, Germany) equipped with a 100× objective (Plan Apo VC 100× oil, N.A.: 1.4, W.D.: 0.13 mm, Nikon, Duesseldorf, Germany), a motorized stage (TI-S-ER Motorized Stage, Nikon, Duesseldorf, Germany), an sCMOS camera (Neo, Andor, Belfast, UK), a 100-W lamp for illumination (Lambda LS Xenon Arc, Sutter Instrument, Novato, CA, USA) and a filter set (Nikon, Duesseldorf, Germany). Gain and scaling were kept constant throughout all measurements.

Quantification of Aβ peptides and cell aggregation on the immunocytochemistry slides. All
images were visually quality controlled. Images that did not fulfill pre-defined criteria based on image acquisition in the adequate focal plane and adequate cell density were discarded. Of the 2,304 images acquired, 2,072 (~90%) were used for further analysis. An automated quantitative phenotypic image analysis was performed using a custom-adapted version of the image analysis software DetecTiff 27 . Pericellular Aβ peptide levels were quantified from the AF488 channel within donut-shaped masks (see Fig. 1B), which were generated from the DAPI channel in a three-step procedure. First, the images of DAPI-stained bacteria were automatically segmented by combined dynamic intensity thresholding and size-dependent particle filtering. In the second step, the binary masks of bacteria were iteratively dilated using a morphological operator to cover the area of the pericellular AF488 signal. Finally, the donut-shaped masks were computed by subtracting the binary masks generated in step one from the masks generated in step two. A total of 57,249 individual donut-shaped analysis masks were constructed from all images. The Aβ peptide level in an individual cell was calculated as the arithmetic mean of all pixel values within the area of a donut-shaped mask and reported as the mean pixel intensity. Cell aggregation was quantified in an automated manner from segmented images of DAPI-stained nuclei by counting the number of bacteria adjacent to a selected individual bacterium within a rectangular 50-pixel 2 region.
Scanning electron microscopy of Enterococcus faecalis. The specimens were attached to aluminum holders and sputtered with platinum in 15-nm layers using a LEICA EM SCD-500 (Leica, Wetzlar, Germany). The surface morphology of the platinum sputter-coated samples was investigated using a JSM 6 610 scanning electron microscope (JEOL, Peabody, MA, USA) at a working distance of 10 mm and an acceleration voltage of 10 kV at room temperature.

Assessment of proliferation and viability.
An inoculum of 5 × 10 5 cells/ml was dispensed into 96-well plates containing Mueller-Hinton broth growth medium and 3.13 μ g/ml or 6.25 μ g/ml Aβ -peptides or LL-37 (Anaspec, Fremont, CA, USA). Bacterial plates were incubated aerobically at 37 °C for 6 h, and C. albicans was cultured for 20 h at 37 °C. To assess proliferation, 10 μ l Alamar Blue ® reagent (Life Technologies, Darmstadt, Germany) was added to each well containing a 100-μ l sample, followed by incubation for 60 min at 37 °C in the dark. The resulting absorbance was measured with a SpectraMax 340 PC 384 microplate reader (Molecular Devices, Biberach, Germany). The proliferation of C. albicans was assessed by measuring turbidity at 570 nm.
The viability of E. faecalis, E. coli and C. albicans was further examined by plating Aβ -treated cultures on agar plates. Cultures of E. coli, E. faecalis and C. albicans were prepared as described above. After incubation with 50 μ g/ml Aβ 1-40 , Aβ 1-42 or Aβ 3p-42 for 6 h for bacteria and 20 h for C. albicans, the cell suspensions were diluted in Mueller-Hinton broth, and the microbial load was determined using an Eddy Jet Spiral Plater (IUL Instruments, Germany) by depositing 50 μ l of sample on a rotating agar plate. Bacteria were grown on Mueller-Hinton agar plates, and C. albicans was grown on Sabouraud dextrose agar plates. After 24 h of incubation, colonies were counted and translated to colony-forming units (CFU)/ml. Assessment of Aβ peptide aggregation. Thioflavin T was used to monitor the aggregation of Aβ 1-42 in the presence of microorganisms. In a black 96-well plate, 160 μ g/ml of Aβ 1-42 in PBS/0.4% NH 4 OH were mixed with heat inactivated E. coli, S. aureus or C. albicans at a final density of McFarland 0.5. Leaving away the microorganisms or adding 10 μ g/ml carboxylated microbeads (Micromod, Rostock, Germany) with a diameter of 0.5 μ m served as control. After adding Thioflavin T (Sigma-Aldrich, Munich, Germany) in final concentration of 0.2 mM, kinetic measurement was performed at 37 °C with excitation at 450 nm and emission at 484 nm in a CLARIOstar ® microplate reader (BMG labtech, Ortenberg, Germany). Measurements containing the respective microorganisms but no Aβ 1-42 served as blank.
Statistical analysis. The statistical analysis was performed using GraphPad Prism ® 6.0 software (GraphPad, La Jolla, CA, USA). Although normality tests could not be calculated, due to the small sample sizes, a Gaussian distribution of the data can be assumed. No pairing of the experiments was assumed. Therefore, one-way ANOVA, followed by the Dunnett's post-test for multiple comparisons against the control were calculated.

Results
Aβ peptides bind to bacterial surfaces. E. faecalis and S. aureus were cultivated in the presence of the Aβ peptide variants Aβ 1-40 , Aβ 2-40 , Aβ 1-42 and Aβ 3p-42 at a concentration of 50 μ g/ml. After 6 h, the bacteria were smeared on glass slides, labeled with the Aβ peptide-specific mouse monoclonal antibody 6E10 and counterstained with DAPI. Immunofluorescence microscopy revealed Aβ peptide-specific staining on the surfaces of both, E. faecalis and S. aureus (Fig. 1A). Automated quantification of experiments with E. faecalis and with S. aureus including more than 1,000 bacteria per condition revealed that Aβ x-42 bound more strongly to the surface of the microorganisms than did Aβ 1-40 (Fig. 1A,C). Aβ -reactive material was also observed in between the bacteria after incubation with Aβ 1-42 and Aβ 3p-42 , but the intensity was strongest directly around the bacteria.
Aβ peptides agglutinate microorganisms. Automated analysis of the immunocytochemistry slides further revealed that treatment with Aβ 1-42 and Aβ 3p-42 led to agglutination of E. faecalis and S. aureus. By contrast, incubation with Aβ 1-40 resulted in almost no agglutination (Fig. 1A,D).
Scanning electron microscopy of E. faecalis incubated with the different Aβ peptide variants confirmed the agglutinating activity of Aβ x-42 . Additionally, bacteria treated with Aβ x-42 were dysmorphic, and large amounts of amorphous material were present within the aggregates. In cultures of E. faecalis incubated with Aβ 2-40 , there were larger aggregates of bacteria that retained their vital morphology. Only minor aggregates were observed in the absence of Aβ peptides and in cultures incubated with Aβ 1-40 . Taken together, these results suggest that Aβ x-42 peptides induce agglutination of microorganisms (Fig. 3).
Aβ peptides exert antimicrobial activity. Flow cytometry analysis of the microorganisms treated with Aβ x-42 peptides revealed a population characterized by increased autofluorescence at 525 nm and/or reduced forward scatter (AF+ /FSC− ). Increased autofluorescence and reduced forward scatter are common features of damaged cells 25,26 . This population was therefore categorized as damaged microorganisms. DAPI staining was performed to differentiate the microorganisms from the background, and only DAPI-positive events were gated for further analysis. The microbial aggregates observed via immunocytochemistry, Gram staining and scanning electron microscopy were only partially detected by flow cytometry; Compared with the size of untreated and consequently unaggregated microorganisms Aβ peptide treated microorganisms did not show increased forward or side-scatter characteristics.
Aβ  exerted antimicrobial activity on all tested microorganisms. As characterized by reduced forward scatter and increased autofluorescence, damaged microorganisms represented up to 70% of cells in the analyzed samples. N-terminal truncation and pyroglutaminylation of Aβ 1-42 further increased microbicidal activity against all microorganisms except S. aureus (Fig. 4). No microbicidal activity was observed for Aβ  or Aβ 2-40 . Identical concentrations of Aβ peptides did not agglutinate human THP-1 cells and were not toxic to those cells within 24 h.
The antimicrobial activities of Aβ 1-42 and Aβ 3p-42 were further confirmed by the Alamar blue test for bacteria, by turbidimetry for Candida and by seeding the bacterial and fungal cultures on agar plates (Fig. 5). Incubating E. coli and E. faecalis with 50 μ g/ml Aβ 1-42 or Aβ 3p-42 reduced the number of colony forming units by 50% compared to untreated cultures (Fig. 5A,B). The number of colony forming units was not reduced by Aβ 1-40 . The toxic effects of the Aβ peptides were even stronger against C. albicans. Incubation with Aβ 1-42 or Aβ 3p-42 reduced the number of colony forming units by 85% (Fig. 5C). Aβ 1-40 also had a moderate antimicrobial effect against C. albicans in culture.

Discussion
We observed that the more amyloidogenic Aβ x-42 peptides led to the agglutination and death of microorganisms. For Aβ 1-40 only an antifungal activity was seen.
Microbial agglutination induced by Aβ x-42 was observed by Gram staining, immunocytochemistry and scanning electron microscopy. Microbial agglutination was accompanied by binding of the Aβ peptide to the microbial surface. Only in cultures of L. monocytogenes and C. albicans no agglutinated bacteria were observed after the treatment with Aβ 2-42. As there can still be seen this red amorphic material, it is well possible that the bacteria already disintegrated and left only debris. This is supported by the fact that a very strong microbicidal effect for Aβ 2-42 is also observed by flow cytometry. Also, when looking at the gram stains of Aβ 3p-42 treated L. monocytogenes it seems that there are already several bacteria "missing" within the agglutinate. The antimicrobial peptides eosinophil cationic protein and salivary agglutinin 23,24 also induce agglutination. The agglutination of pathogens by Aβ -peptides may contribute to antimicrobial control via several mechanisms. First, agglutination may prevent the distribution of microorganisms by causing physical immobilization. Second, agglutination facilitates phagocytosis 23,24 . Third, Aβ peptides act as opsonins for phagocytosis 28 .Thereby, the Aβ peptide variants with the highest microbicidal activity are also the most effective in inducing phagocytosis 28 . Finally, in addition to bacterial agglutination, Aβ x-42 -peptides have direct antimicrobial activity. Microorganisms exposed to Aβ x-42 exhibited reduced forward scatter and increased autofluorescence in flow cytometry analyses. A loss of forward scatter is a common feature of damaged cells and bacteria, but increased autofluorescence of bacteria exposed to bactericidal agents has only recently been observed 25 . Renggli et al. suggested that this autofluorescence is caused by a change of cell morphology 25 .
Autofluorescence after excitation with ultraviolet and blue light was also observed in amyloid plaques containing full length Aβ peptides and after aggregation of synthetic Aβ peptides [29][30][31] . As can be seen from the IgG staining control in Fig. 1, the detected signals do not result from autofluorescence. In the flow cytometry experiments, the reported autofluorescence after treatment with Aβ peptides also occurred after treatment with antibiotics (data not shown). Therefore we suppose, that autofluorescence in our experiments is due to cell death. However, we cannot fully exclude, that part of the observed fluorescence is due to autofluorescence of Aβ peptides.
The antimicrobial activity of amyloidogenic Aβ peptides was confirmed by reduced Alamar blue turnover and a reduction in colony forming units in our study. In case of E. faecalis, both assays showed a consistent reduction of bacterial growth by the Aβ x-42 peptides. In contrast, the decrease in E. coli CFU by these peptides was paralleled by a weaker effect on Alamar blue turnover (Fig. 5B,E), indicating that agglutination of E. coli may lead to overestimation of the killing activity in the CFU assay. On the other hand, the Alamar blue test and turbidimetry might have underestimated the antimicrobial effect, as the microbial aggregates could disturb the optical path, leading to heterogeneous and false high readings. This could explain, why we miss the cytotoxic effect of Aβ 1-40 and Aβ 2-40 on C. albicans with the Alamar blue assay.
C-and N-terminal modifications of the Aβ peptides greatly affected their aggregation and antimicrobial activity, such that Aβ x-42 was much more effective than Aβ x-40 . N-terminal truncation and pyroglutaminylation further enhanced the antimicrobial activity. The differences in the effects of the Aβ peptide variants are most likely due to their physicochemical characteristics. Aβ x-42 is much more hydrophobic than Aβ x-40 . The truncated and pyroglutaminylated N-terminus further enhances this hydrophobicity [32][33][34][35] . The increased hydrophobicity of Aβ x-42 peptides is also reflected by their increased binding to the surface of the microorganisms. Therefore, there is a good correlation between Aβ -peptide hydrophobicity and the binding/agglutination of microorganisms. While the exact mechanism by which Aβ -peptides kill microorganisms remains to be elucidated, it seems as if the heparin  binding domain of the Aβ peptide sequence is involved. Mannan and glucan are suggested to bind to Aβ peptides via its heparin-binding site [36][37][38] . By competitive binding of the heparin binding site by mannan and glucan, the agglutination and binding of Aβ peptides to microbes was effectively inhibited 36 . Furthermore, binding of the heparin binding site of Aβ peptides promotes their aggregation and the formation of β -sheet structures 39,40 . This explains, why we found an accelerated aggregation of Aβ 1-42 when incubated together with E. coli, S. aureus and C. albicans. Binding of proteoglycans to the heparin-binding site of Aβ peptides has been suggested as an early step in plaque formation 39 . Eosinophil cationic protein is also an amyloid-like protein. Its antimicrobial activity depends on its ability to agglutinate bacteria and form amyloid fibrils on the bacterial surface 24 . The eosinophil cationic protein mediated agglutination of bacteria is followed by membrane leakage and cell death 24 .
The antimicrobial activity of Aβ x-42 -peptides demonstrated in this study is consistent with previous reports 16,36 . However, Soscia et al. found, that Aβ 1-40 is not as effective as Aβ 1-42 but still possesses antimicrobial activity against bacteria and fungi. This effect strongly differed between the investigated microorganisms and was in several bacteria near the limit of detection. In respect to the colony forming units, we only observed an antimicrobial activity of Aβ 1-40 against C. albicans. In accordance with the report of Kumar et al., this activity was weaker as for The proliferation of the bacteria incubated with the indicated Aβ peptides at a concentration of 6.25 μ g/ml (black) or 3.13 μ g/ml (grey) was assessed by the Alamar Blue assay after 6 h (D,E). The proliferation of C. albicans was assessed by turbidimetry after incubation for 24 h (E). The results are expressed relative to untreated cultures and are depicted in a scatter plot together with the median and interquartile range. Asterisks indicate significant differences calculated by the Friedman test followed by Dunn's post-hoc test. *p < 0.05, **p < 0.01, ***p < 0.001. the Aβ x-42 peptides. The discrepancy concerning the effect against bacteria might be due to the selection of different bacterial strains or differences in Aβ peptide pretreatment. Aβ peptides are prone to self-aggregation, and the method of synthesis, the solvents used and the time until application strongly affect their conformation and, consequently, their biological activity. Thus, while an antifungal effect of Aβ 1-40 was consistently shown, its impact against bacteria needs to be further investigated. Having observed increased phagocytotic activity of monocytes after stimulation with Aβ 2-40 , we previously suggested, that the more soluble Aβ x-40 peptides, may also act as autoor paracrine factors regulating immune activity 28 .
The physiological relevance of Aβ peptide antimicrobial activity is supported by several lines of evidence.
(a) In an experimental model of meningitis, survival was reduced in mice that were not able to produce Aβ peptides due to a knockout of its precursor. Reciprocally, mice overexpressing Aβ peptides showed reduced mortality in this model 36 . (b) Acute infections of the brain result in reduced Aβ 42 levels (but not Aβ 40 ) in cerebrospinal fluid but increased deposition in brain tissue 3,41-44 . (c) Aβ peptide deposition has also been reported in chronic infections of the central nervous system, such as neuroborreliosis, neurosyphilis, HIV or Herpes simplex encephalitis 22,[45][46][47] . Recently, Diana Pisa and her colleagues observed Candida species in brain regions affected by Alzheimer's disease pathology but not in normal controls 48 . In our study, Candida was particularly sensitive to Aβ -peptide-induced aggregation and cell death. (d) The γ -secretase blocker DAPT impairs recovery from lipopolysaccharide-induced inflammation in the rat brain 49 . (e) Monocytes, microglia and astrocytes increase the expression of amyloid precursor protein and release Aβ peptides upon activation by lipopolysaccharide [50][51][52] . In several mouse models, lipopolysaccharide induces amyloid precursor protein expression and subsequent Aβ plaque deposition 22 .
These findings suggest that the secretion of Aβ -peptides is part of the innate immune defense in the CNS. Aβ -lowering therapies might therefore hamper the resistance of the brain to infections and malignant disorders. The increased neonatal mortality of BACE1/BACE2 knock-out mice only in animals housed under non-sterile conditions supports this assumption 53 . Special attention should therefore be given to these events during clinical tests of Aβ -lowering therapies. Others have even suggested that an infectious agent may be involved in Alzheimer's disease pathogenesis 22,46,54 . Our finding of accelerated Aβ 1-42 agglutination in the presence of microorganisms may support this hypothesis. Further research is needed to determine whether Aβ plaque deposition is the consequence of a microbial infection or a mechanism of CNS immune defense that is misled in Alzheimer's disease.