Amyloid Fibrils Trigger the Release of Neutrophil Extracellular Traps (NETs), Causing Fibril Fragmentation by NET-associated Elastase*

Background: Amyloid fibrils are ubiquitous structures present in amyloid diseases, but the mechanism by which they exert toxicity is unclear. Results: Neutrophil-derived extracellular DNA traps decorated with elastase are present in amyloidotic human tissue and induce amyloid fragmentation into toxic oligomers. Conclusion: Neutrophil-derived elastase participates in amyloid fragmentation (amyloidolysis). Significance: Understanding how amyloid exerts toxicity is important for the design of therapies for incurable, mostly fatal, amyloid diseases. The accumulation of amyloid fibrils is a feature of amyloid diseases, where cell toxicity is due to soluble oligomeric species that precede fibril formation or are formed by fibril fragmentation, but the mechanism(s) of fragmentation is still unclear. Neutrophil-derived elastase and histones were found in amyloid deposits from patients with different systemic amyloidoses. Neutrophil extracellular traps (NETs) are key players in a death mechanism in which neutrophils release DNA traps decorated with proteins such as elastase and histones to entangle pathogens. Here, we asked whether NETs are triggered by amyloid fibrils, reasoning that because proteases are present in NETs, protease digestion of amyloid may generate soluble, cytotoxic species. We show that amyloid fibrils from three different sources (α-synuclein, Sup35, and transthyretin) induced NADPH oxidase-dependent NETs in vitro from human neutrophils. Surprisingly, NET-associated elastase digested amyloid fibrils into short species that were cytotoxic for BHK-21 and HepG2 cells. In tissue sections from patients with primary amyloidosis, we also observed the co-localization of NETs with amyloid deposits as well as with oligomers, which are probably derived from elastase-induced fibril degradation (amyloidolysis). These data reveal that release of NETs, so far described to be elicited by pathogens, can also be triggered by amyloid fibrils. Moreover, the involvement of NETs in amyloidoses might be crucial for the production of toxic species derived from fibril fragmentation.

The accumulation of amyloid fibrils formed by specific misfolded proteins is a feature shared by more than two dozen varieties of amyloidoses that afflict humans (1). Among these often fatal diseases are Alzheimer and Parkinson diseases, type 2 diabetes, and transthyretin (TTR) 4 -related amyloidoses (2). The Amyloidosis Foundation estimates that each year 50,000 people are diagnosed with systemic amyloidoses worldwide.
The toxic protein species in many amyloidoses appears to be small prefibrillar aggregates on the pathway to fibril formation; they are toxic possibly because they interact with and disrupt cell membranes or they associate with other macromolecules inside the cell leading to cell demise (3)(4)(5). The mature amyloid fibril, assumed to be innocuous, may act as a reservoir of smaller oligomeric, toxic species that spread after being formed by fibril disintegration (5). However, up to now there is very little evidence showing that fragmentation of fibrils (amyloidolysis) indeed occurs in vivo.
It has been shown that amyloid fibrils can induce an inflammatory response in which cytokines are secreted by macrophages and glial cells (6,7). This suggests a new function for the immune system in amyloidoses. As the first immune cells to reach a damaged site, neutrophils are the main mediators of systemic inflammation (8). A novel death mechanism has been described for neutrophils and other granulocytes in which chromatin is released to the extracellular milieu, forming a web called neutrophil extracellular traps (NETs). This structure is formed by a DNA-histones scaffold that is decorated with granule proteins such as elastase and myeloperoxidase, as well as with antimicrobial peptides (9,10). This mechanism, known as NETosis, can be triggered by lipopolysaccharide, phorbol 12-myristate 13-acetate (PMA), bacteria, fungi, and protozoa (9 -13) and is one mechanism used by neutrophils to trap and kill microbes and avoid their spreading (9,13). Remarkably, NETs do not seem to be limited to fighting pathogens; they have also been found in autoimmune diseases and in thrombosis (14 -17). However, it is still unknown whether amyloid fibrils induce NET release.
A potential association between amyloid fibrils and neutrophils was suggested when neutrophil-derived proteins such as elastase and histones, but no whole neutrophil cells or NETs, were found in amyloid deposits from different systemic amyloidoses such as light chain amyloidosis, secondary amyloidosis, systemic senile amyloidosis, and ␤ 2 -microglobulin amyloidosis (18 -22). Additionally, aberrant formation or persistence of NETs in tissue or serum can aggravate diseases and promote tissue damage (17), although the relevance of neutrophils in amyloidoses is unclear.
Here, using an in vitro approach, we asked whether NET release can be triggered by amyloid fibrils and whether this mechanism occurs in the tissues of amyloidosis patients. Evidence for the involvement of NETs, and especially their protease constituents, in modulating the cytotoxicity of amyloid fibrils would contribute to the understanding of the etiology of amyloid diseases. To answer these essential questions, we produced amyloid fibrils using three different proteins, namely A25T TTR, ␣-synuclein (␣-syn), and Sup35. Although Sup35 is a yeast prion (23), A25T is associated with oculoleptomeningeal amyloidosis (24) and ␣-syn with Parkinson disease (25). Here, we show that amyloid fibrils from all three proteins induced NADPH oxidase-dependent NET release, whereas the amorphous, nonfibrillar aggregates composed of the same proteins did not. The elastase that decorates the DNA traps was able to degrade the amyloid fibrils into a short, toxic oligomeric species, thereby providing the first evidence for a possible physiological mechanism leading to fibril fragmentation and aggravation of the disease. Immunohistochemical analyses of amyloidotic tissues from patients with systemic amyloidoses revealed the presence of NETs, strengthening the evidence for the participation of neutrophils in amyloid diseases. Taken together, our findings suggest that NET formation operates physiologically and should be considered a new player in the etiology of these diseases. As far as we know, this is the first time that amyloid fibrils have been shown to trigger NET formation and the first evidence that amyloid fibrils are a target for the NET-associated proteases.

EXPERIMENTAL PROCEDURES
Neutrophil Purification-Human neutrophils from buffy coats of healthy blood donors were isolated by density gradient centrifugation (Histopaque, Sigma) as described (12). All procedures and human biological samples used in this work were performed in accordance with Institution regulations and approved by the Institutional Review Board for Human Subjects (Hospital Universitário Clementino Fraga Filho, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil). NET DNA from human neutrophils was quantified as described (12).
Intracellular ROS Measurement-Intracellular ROS production was assessed by using the probe dihydrorhodamine 123 (DHR-123; Molecular Probes). Neutrophils were incubated with DPI (10 g/ml; Sigma). After 30 min of incubation, neutrophils were stimulated with fibrils of A25T (A25T-F, 6 M) or ␣-synuclein (␣-syn-F, 20 M) and incubated with DHR-123 (2 M) for 15 min. Neutrophil subsets were determined on the basis of size and granularity, and the fluorescence was monitored by FACSCalibur flow cytometer (BD Biosciences). Data analysis was performed on CellQuest software.
Nuclear Area Quantification-Neutrophils were incubated for 90 min with designated samples, and the cells were fixed and stained with the DNA marker DAPI (10 g/ml). The nuclear area of 150 -400 cells from at least three different donors was quantified using ImageJ software (National Institutes of Health), and the mean nuclear area per m 2 of a given field was plotted using GraphPad Prism 5.0 software.
NET-associated Elastase Digestion Efficiency-Neutrophils were incubated for 90 min with 100 nM PMA, and the supernatant of PMA-activated neutrophils containing NETs (SPAN) was further incubated up to 120 min at 37°C with A25T-F (10 M) or during 16 h with ␣-syn-F (10 M) or V30M-F (10 M). All SPAN were quantified to confirm that neutrophils had formed NETs, using extracellular DNA quantification as described previously (12). To confirm the presence of elastase, a Western blot using an anti-human neutrophil elastase (1:1,000; Calbiochem) was performed. In the case of A25T-F suspension, aliquots were withdrawn at different time points, and the fibrils that remained intact (seen as monomers and dimers) were examined by SDS-12% PAGE followed by Western blotting with anti-human TTR antibody (1:1,000; Dako Cytomation). The band intensity was quantified using ImageJ software (National Institutes of Health). To determine the specificity of A25T-F digestion by NET-associated elastase, 10 g/ml of an elastase-selective inhibitor (EI; MeOSuc-AAPV-CMK, Calbiochem) was preincubated with SPAN for 20 min at 37°C before adding 10 M of A25T-F, which was then incubated for 90 min at 37°C. SPAN alone and A25T-F alone served as controls under the same conditions. Aliquots were withdrawn, and A25T-F degradation was monitored by 18% SDS-PAGE followed by Western blotting using anti-human TTR antibody (1:1,000; Dako Cytomation) and CR binding assays to quantify the remaining intact amyloid fibrils. For the CR binding assays, 40 l of fibrils were incubated with 9 l of 230 M CR for 5 min, centrifuged, and resuspended in 400 l of 5 mM phosphate buffer with 150 mM NaCl at pH 7.4. CR binding to undigested fibrils was measured at 488 nm in a spectrophotometer, and the percentage of amyloid fibrils was compared with 10 M of undigested A25T-F (referred as 100%). To analyze the morphology of the digestion products, 20 M of A25T-F was incubated with SPAN for 90 min at 37°C or with SPAN pretreated with 10 g/ml of EI and total fraction was analyzed by TEM (31).
Cell Toxicity Assays-MTT reduction assay (Affymetrix) was performed on BHK-21 cells as described (34). Lactate dehydrogenase release assay was quantified in HepG2 cells using the Cytotox-ONE assay (Promega) according to manufacturer's protocol. Live/Dead assay (Molecular Probes) was used according to manufacturer's instructions. Cells were visualized using an EVOSfl microscope (Advanced Microscopy Group), and the percentage of live cells was measured.
Statistical Analysis-Unless otherwise specified, data were analyzed by ANOVA using GraphPad Prism 5.0 software. Each experiment was performed at least three times, with replicates, on independent occasions. p Ͻ 0.05 was considered significant.

Amyloid Fibrils Induced NET Release from Human
Neutrophils-It is unknown whether amyloid fibrils were able to activate human neutrophils in vitro and trigger the release of NETs. Initially, to answer these questions, we produced amyloid fibrils using three different proteins as follows: A25T TTR, ␣-syn, and Sup35. Fig. 1 shows that the fibrils formed from Sup35 (Sup35-F), A25T-TTR (A25T-F), and wild-type ␣-syn (␣-syn-F) presented the typical amyloid structure (A-C), contrary to ␣-synuclein amorphous aggregates (␣-syn-A, Fig. 1D) as seen by TEM. These results were confirmed by thioflavin-T (ThT; Fig. 1E) and Congo Red (CR; Fig. 1F) binding. These two compounds are amyloid-specific dyes that change their spectroscopic behavior when bound to the cross-␤ fold present in amyloid fibrils (36). As positive and negative controls, ThT and CR binding to IAPP fibrils (IAPP-F), BSA, and soluble A25T were evaluated.
To evaluate whether each type of fibril could trigger NET release (quantified as extracellular DNA in nanograms/ml of suspension), they were incubated with human neutrophils. Fig.  2A shows that all amyloid fibrils tested (3-10 M; black bars) promoted a significant neutrophil DNA release similar to that caused by PMA (dotted bars), a classical NET inducer (12). It is interesting to note that the extruded DNA had a web-like appearance (red in the inset of Fig. 2A), as expected when NETs are formed. More interesting was the observation that the amyloid fibrils (blue in the inset of Fig. 2A) co-localized with the DNA webs released by the neutrophils. We also have observed that NET induction by amyloid fibrils was time-(data not shown), dose-, and donor-dependent (Fig. 2B). However, the molecular basis for the differences in the extent of DNA release observed among neutrophils collected from different donors is still unknown. The supplemental Fig. 1 shows a nuclear area quantification as another means of characterizing NET formation (17). Again, the areas of DAPI-stained DNA were greater only when amyloid fibrils were added to the neutrophils, suggesting that DNA was extruded from the cells only under these conditions. As expected, unstimulated neutrophils did not exhibit great nuclear areas (supplemental Fig. 1) or released significant amounts of DNA (data not shown). To assess whether the ability to release NETs was determined by the amyloid architecture, neutrophils were incubated with soluble proteins, namely bovine serum albumin (BSA), soluble A25T, and ␣-syn or with amorphous aggregates composed of A25T (A25T-A) or ␣-syn (␣-syn-A). As an example, the amorphous morphology of this latter aggregate is shown in Fig. 1D (compare with Fig. 1C). Note also that, as expected, neither the amorphous aggregates nor the soluble proteins were able to bind ThT or CR (Fig. 1, E and F). Critically, none of these proteins was able to induce neutrophil DNA release (see bars in supplemental Fig. 1 and hatched and white bars in Fig. 2A), suggesting that the fibril architecture was a necessary prerequisite for this process.
To rule out the possibility that the extracellular DNA was released by cell disintegration, we searched for the presence of proteins known to be co-released with DNA when NETs are formed, such as elastase. Unstimulated neutrophils presented the typical lobulated nuclei ( Fig. 2O and inset), whereas neutrophils exposed to all amyloid fibrils tested released DNA, which exhibited the typical web appearance (Fig. 2, C-N, DNA in red decorated with elastase in green). Note that the amyloid fibrils (blue in Fig. 2, E, I, and M) were captured by NETs, as revealed by their co-localization (white arrows in Fig. 2, C-N) in some areas where NETs were present. Similar observations regarding the interaction between NETs and its target have been reported for neutrophils challenged with pathogens such as Leishmania (12). It has to be emphasized that the amorphous aggregates composed of ␣-syn, although incompetent to induce NET release, appeared to bind to the neutrophils (Fig. 2P).
In addition, to gain insight into the origin of the extruded DNA (mitochondrial or nuclear), an anti-histone antibody (green) was applied. It revealed an abundance of histones associated with this DNA, suggesting that the majority of extruded DNA is of nuclear origin (supplemental Fig. 2).
Are ROS Involved in Fibril-induced NET Release?-We analyzed whether amyloid fibril-induced NET release was dependent on NADPH oxidase activation and ROS formation as seen before for pathogen-induced NET release (11). To pursue this, we evaluated whether the amyloid fibrils composed of A25T (A25T-F) or ␣-syn (␣-syn-F) were able to elicit NET release in the presence of DPI, a NADPH oxidase inhibitor (Fig. 3). As seen, in the presence of DPI both fibrils were unable to induce NET formation (Fig. 3, C and D), contrary to what happened in the absence of DPI (Fig. 3, A and B). Note in Fig. 3, insets, that the nuclei of neutrophils retained their lobulated aspect when  OCTOBER 26, 2012 • VOLUME 287 • NUMBER 44 JOURNAL OF BIOLOGICAL CHEMISTRY 37209 fibrils were added in the presence of DPI. Next, we quantified the extent of NET release when human neutrophils were pretreated with an antioxidant compound (N-acetylcysteine), and two NADPH oxidase inhibitors (DPI and apocynin), before ␣-syn-F and A25T-F addition. As seen in Table 1, under these conditions, fibril-induced NET release was inhibited by ϳ40 -50%, suggesting that NADPH oxidase and consequently ROS formation contribute significantly to this phenomenon.

NET-associated Elastase Induces Amyloidolysis
Finally, to corroborate our data, we evaluated whether ROS was released by human neutrophils stimulated by amyloid fibrils, using DHR (Molecular Probes), a cell-permeable fluorogenic probe that is useful for the detection of ROS. This probe is not fluorescent until oxidized by ROS, when it is converted into the highly fluorescent product, rhodamine 123. As seen in Fig.  3, E-G, although ROS production was very low in unstimulated neutrophils, addition of PMA, A25T-F, or ␣-syn-F led to a significant increase in ROS formation. Also, the pretreatment of neutrophils with DPI diminished ROS formation in PMA-and fibril-stimulated neutrophils.
NET-associated Elastase Digested A25T-F into Toxic Species, Implication for Amyloid Diseases-Although elastase has been found in amyloid deposits from patients with different amyloid diseases (20), NETs were unknown at that early date, and it is still unclear whether the amyloid fibrils from these deposits are a target for elastase. To determine whether elastase from NETs digests amyloid fibrils, we incubated 10 M A25T-F with the Fig. 3, SPAN, a suspension rich in NET-associated elastase. The presence of NET components in SPAN was assessed by quantifying the amount of DNA released (data not shown) as well as the presence of elastase (supplemental Fig. 3).
As seen in Fig. 4A, lane 1, A25T-F added to SPAN (0 min) was visualized immediately as two bands, the monomer (14 kDa) and the dimer (28 kDa) of A25T. This occurs because A25T-F is SDS-sensitive and readily dissociates into monomers and dimers. Over a period of 120 min in the presence of SPAN (Fig.  4A), the population of monomers and dimers diminished, and after 90 -120 min, monomers and dimers were no longer visible in the Western blot (Fig. 4A, lanes 5 and 6), showing that A25T-F was being digested by the proteases present in SPAN, among them elastase. It has to be emphasized that the kinetics of SPAN-induced fibril digestion (amyloidolysis) were donordependent, as expected. Fig. 4B illustrates two extreme examples (circles and squares). SPAN recovered from donor 1 (Fig.  4B, circles) was very fast in digesting A25T-F and in the first 10 -20 min, and ϳ50% of the fibrils underwent amyloidolysis, with no major increase at longer incubation times. However, SPAN collected from donor 2 was slower initially but able to digest all A25T-F in 120 min. Fig. 4B also shows as a control that a suspension of A25T-F incubated under the same conditions in the absence of SPAN (no SPAN) did not undergo amyloidolysis, and the amount of dimers and monomers remained nearly the same throughout the experiment.
To test whether A25T-F was a target for elastase, we added MeOSuc-AAPV-CMK, an elastase inhibitor (EI), to the SPAN and then added A25T-F, leaving it for 90 min. The inset of Fig.  4A shows that in the presence of EI the digestion of A25T-F did not occur. Densitometry of these blots showed that 100% of A25T-F (monomers plus dimers) remained intact when SPAN was pretreated with EI, although less than 25% of A25T-F persisted in EI-untreated SPAN (Fig. 4C, hatched bars). The digestion of A25T-F by elastase was also monitored by CR binding (Fig. 4C, white bars). After 90 min in the presence of SPAN, almost 75% of the fibrils were digested, although in the presence of SPAN pretreated with EI, almost none of them were digested. Our supplemental Fig. 3B (lane 3) shows as control that SPAN was not recognized by our anti-TTR antibody and that EI alone did not interfere with fibril integrity (supplemental Fig. 3B,  compare lanes 1 and 2).
Next, we tested whether the elastase released from NETs by DNase treatment was as effective as when the elastase that was associated with NETs. Thus, we repeated the same experiments reported in Fig. 4A using a DNase-treated SPAN. Interestingly, free and NET-associated elastase was equally efficient in digesting A25T-F as seen by the decrease in CR binding after fibril digestion (supplemental Fig. 3C).
Because the amyloid fibrils were a target for the elastase associated to the NETs, we evaluated the morphology of A25T-F before and after SPAN digestion (Fig. 4, D-F). TEM revealed that elastase-induced fragmentation of amyloid fibrils produced smaller and less clustered aggregates (Fig. 4E). However, when SPAN was pretreated with EI and afterward incubated with A25T-F, fragmentation was no longer evident (Fig. 4F). We also asked whether these fibril fragments presented the epitope found in prefibrillar oligomeric species originating from different amyloid-forming proteins (37). Oligomers with different constitutions and morphologies have been shown to be toxic to several cell lines and are considered by several groups to be the culprits responsible for the cytotoxicity associated with the amyloidoses (3). To answer this question, we probed fibrils digested with SPAN or with purified human neutrophil elastase using A11, an anti-oligomer antibody (Fig. 4G). Interestingly, the fragments derived from A25T-F digestion with SPAN or with purified elastase (Ela) exhibited A11 immunoreactivity as strong as that displayed by hen egg white lysozyme oligomers (HEWL-O) (38), suggesting that they share structural features with the typical prefibrillar oligomers.
We further tested whether other amyloid fibrils could be targets for the NET-associated elastase. Thus, we produced fibrils composed of ␣-syn or of V30M, another variant of TTR associated with familial amyloidotic polyneuropathy (39). CR binding showed that both fibrils were partially digested after 16 h in the presence of SPAN (supplemental Fig. 4), reinforcing the idea that amyloid fibrils, regardless of their composition, are targets for the NET-associated elastase. Inset shows A25T fibrils (blue) bound to the extruded DNA (red). The amount of DNA released by A25T-F, ␣-syn-F, and Sup35-F was statistically greater than that released on average by unstimulated neutrophils from all 10 -12 donors tested (348 ng/ml of extracellular DNA, not displayed in the graph). This background value was subtracted from all measurements shown here. Statistical analysis was performed using ANOVA and Tukey's post-hoc analysis: n Ն4, *, p Ͻ 0.05; **, p Ͻ 0.01, and ***, p Ͻ 0.001 relative to the average DNA released by unstimulated neutrophils from all donors tested. Values are means Ϯ S.E. B, A25T fibril-induced DNA release is dependent on fibril concentration and neutrophil donors. A dose-dependent response was observed in neutrophils from all six donors presented here, although the amount of DNA released was variable. C-P, representative immunofluorescence images of neutrophils incubated with amyloid fibrils showing that the extruded DNA has a web-like structure and is decorated with elastase, which suggests the formation of NETs. DNA is pseudocolored in red, elastase in green, and fibrils or amorphous aggregates in blue. In merged images (F, J, and N), the co-localization is shown in purple, and yellowish areas are denoted by white arrows. C-F, ␣-syn-F; G-J, A25T-F; K-N, Sup35-F; O, unstimulated neutrophils; P, neutrophils stimulated with ␣-syn-A. Note that ␣-syn-A (P) did not trigger NET formation. Scale bars, 100 m in C-N, and 20 m in insets for A and P. The inset of O was amplified three times.
The next experiment was designed to answer the critical question of whether the fragments generated by elastase digestion would be toxic to cells in culture. Fig. 5 shows the results obtained with Live/Dead assays in BHK-21 cells. Cells treated with medium (Fig. 5A) were alive as shown by the calcein green staining, whereas cells treated with toxic HEWL-O (Fig. 5B) were dead, as indicated by the ethidium homodimer red staining. Addition of intact A25T-F did not cause cell death (Fig. 5C), reinforcing the idea that intact fibrils are innocuous (3), and the addition of SPAN alone was also harmless to these cells (Fig.  5D). However, only dead cells were observed when BHK-21 was treated with A25T-F incubated for 90 min in SPAN (Fig. 5E). To confirm that the toxic species were indeed the elastase digestion products of A25T-F, purified elastase was used to digest  intact A25T-F. Again the product of this digestion was very toxic to BHK-21 cells (Fig. 5G), although the purified elastase alone did not cause cell death (Fig. 5F). These data are quantified in Fig. 5H (ANOVA, p Ͻ 0.001 for SPAN-treated A25T-F and p Ͻ 0.01 for elastase-treated A25T-F). These results were also confirmed by MTT reduction assay (supplemental Fig. 5A). We also tested the toxicity of the digestion products of SPANtreated A25T-F in a human cell line, HepG2, and again, these species were also significantly toxic to these cells in culture (supplemental Fig. 5B). However, when these toxic oligomers were added to neutrophils at different concentrations (1-6 M), they were innocuous, as assessed by lactate dehydrogenase release assay (data not shown). Besides, the digestion products of A25T-F were not as efficient as the intact fibrils in eliciting NET release (supplemental Fig. 5C).

NETs in Amyloid Deposits from Human Patients with Primary Systemic
Amyloidosis-Although our data indicate that amyloid fibrils induce NETs in vitro, it is still unknown whether NETs are found in amyloid diseases. Thus, we searched for NETs in post-mortem tissues from two patients diagnosed with primary systemic amyloidosis, a plasma cell dyscrasia characterized by an autonomous proliferation of plasma cells with an overproduction of a monoclonal immunoglobulin protein that deposits throughout the body as amyloid fibrils (40). We confirmed the presence of amyloid fibrils in the skin and lungs of patient 1 using alkaline CR staining (Fig. 6), which, when bound to amyloid fibrils, exhibits an apple-green birefringence under polarized light (see white arrows in Fig. 6, B and D). To evaluate the presence of NETs in these tissues, we searched for the colocalization of the classical NET markers as follows: a neutro-  (A25T-F). A, representative kinetics of A25T-F digestion by SPAN is shown on a Western blot using anti-TTR antibody to quantify the amount of undigested A25T-F (monomers plus dimers). Lane 1, A25T-F immediately after mixing with SPAN (0 min); lanes 2-6, A25T-F incubated with SPAN for 15, 30, 60, 90, and 120 min, respectively. Note that after 90 min in the presence of SPAN, there were no dimers or monomers of TTR, suggesting that A25T-F was completely digested by the proteases present in SPAN. The inset shows that when SPAN was pretreated with 10 g/ml of EI, digestion of A25T-F was inhibited. B, Western blot densitometry of the kinetics of A25T-F digestion of two different donors (#1, circles, and #2, squares). Note that there is a clear variation between donors, which modifies the A25T-F digestion kinetics. A25T-F incubated in PBS (no SPAN) was used as loading control for the gels. C, percentage of intact A25T-F remaining after SPAN digestion was evaluated by densitometry of the Western blots (ImageJ software) from the inset of A (hatched bars) or by Congo Red binding to the same samples (white bars). D-F, TEM images of A25T-F before digestion with SPAN (D), after digestion for 90 min with SPAN (E), or after digestion for 90 min with SPAN pretreated with EI (10 g/ml) (F). G, dot blots of A25T using anti-oligomer antibody (A11), which recognizes specifically an epitope present in the amyloid oligomeric species. Note that A11 recognizes HEWL-O but not SPAN alone or A25T-F. However, the species derived from the digestion of A25T-F with SPAN (90 min) or with purified neutrophil elastase (5 g/ml for 90 min) were recognized by A11. Scale bars (D-F), 5 m. Statistical analysis was performed using ANOVA and Tukey's post-hoc analysis: n Ն 4, ***, p Ͻ 0.001; **, p Ͻ 0.01. Values are means Ϯ S.E. OCTOBER 26, 2012 • VOLUME 287 • NUMBER 44 phil-specific elastase, DNA, and DNA-histone complex (9). We observed co-localization of elastase and DNA-histone forming web-like structures with morphological NET characteristics in the lung (Fig. 7, A-D) and skin (Fig. 7, E-H) of patient 1 and in a skin-punch biopsy from patient 2 (supplemental Fig. 6), which strongly suggests the presence of NETs in these tissues. To determine whether NETs found in human tissues were in close contact with amyloid fibrils, we used thioflavin-S, which binds to amyloid and emits a green fluorescence. In Fig. 8, we were able to see specific areas in the lung (A-E) and skin (F-J) of patient 1 where amyloid fibrils (thioflavin-S emission in C and H) were localized exactly where NET-associated elastase was found (see yellow-orange areas in D and I and zooms in E and J). Interestingly, in areas where amyloid fibrils were absent or scarce, we could not observe the presence of NETs using two NET markers, namely human neutrophil elastase and DNA (supplemental Fig. 7). These results suggest that amyloid could be a novel NET inducer in vitro and in vivo.

NET-associated Elastase Induces Amyloidolysis
Finally, because we have shown that NET-associated elastase was able to digest amyloid fibril into oligomers, we searched for the presence of these species in the tissues of patient 1 (Fig. 9). Interestingly, A11-positive oligomers (Fig. 9, B and G) were found in areas of the lungs (Fig. 9, A-E) and skin (Fig. 9, F-J) where NET markers were observed. These data suggest that these toxic species are a side product of fibril-induced NET formation with novel, unknown consequences for the development of amyloid diseases.

DISCUSSION
Here, we showed for the first time the presence of NETs in tissue from patients with amyloid diseases (Figs. 7-9). Also, we found that amyloid fibrils induced NETs regardless of which protein formed the fibrils (Fig. 2) and that digestion of A25T-F by NET-associated elastase generated oligomers, which showed to be toxic to cell lines in culture (Fig. 5 and supplemental Fig. 5, A and B). These toxic oligomers were also observed in tissue sections from these patients in close proximity with NETs, suggesting strongly that they are formed physiologically as a side product of amyloidolysis induced by NET-associated elastase. The induction of NET by the amyloid fibrils depends largely on the NADPH oxidase system (Fig. 3 and Table 1) as shown previously for pathogen-induced NET release (11).
Earlier studies have shown that neutrophil proteases such as elastase are associated with amyloid fibrils isolated from patients with systemic amyloidosis (20). However, no granulocytic cells were observed in these tissues (20). At the time, the NET mechanism was unknown, and the role of neutrophils in amyloidoses was not even considered. Nevertheless, the lack of whole granulocytic cells reported by these authors agrees with our current knowledge of NETs, which are associated with the release of chromosomal DNA and cell death.
We observed that fibrils that had the ability to stimulate neutrophils to release NET were in close contact with the webs. It is possible that histones present in the NET mediate this contact with the fibrils through electrostatic interactions. Indeed, histones have been widely found in amyloid deposits of amyloid A in cattle and humans diagnosed with secondary amyloidosis (19,21,22). In all these studies, the question of neutrophil involvement in amyloidosis was largely ignored. The presence of histones in the neutrophil extracellular DNA observed herein may indicate that the DNA's origin is primarily nuclear. However, we cannot exclude that mitochondrial DNA may be a part of NETs observed in our study.
We suggest that the architecture of the amyloid fibril may be crucial to NET release, because the amorphous aggregates did not induce NETosis. This may mean that amyloid-induced NET release is receptor-mediated. One possibility is that the membrane receptor for advanced glycation end products, which binds amyloid fibrils (41), may mediate amyloid fibrilinduced NADPH oxidase activation and NET release. The presence of amorphous aggregates of ␣-syn inside the neutrophil cells (Fig. 2P) suggests that perhaps neutrophils also attempt to phagocytose the fibril structures. This frustrated phagocytosis might lead to neutrophil death and DNA release. Neutrophils treated with ␣-syn-F in the presence of cytochalasin-D, an actin polymerization inhibitor, were also able to release NETs (data not shown), suggesting that phagocytosis is not necessary for fibril-induced NET formation.
The smaller oligomeric species that we observed after elastase-promoted A25T-F digestion were toxic and caused mitochondrial dysfunction as seen by MTT reduction assay (supplemental Fig. 5) and cell death (Fig. 5). This suggests that NET-associated elastase found in tissues may catalyze in vivo fibril proteolysis, enhancing its toxicity. Interestingly, the oligomers formed in vitro and after amyloid fragmentation were not toxic to neutrophils (data not shown) and were not as effective as the intact fibril in inducing NET release (supplemental Fig. 5C). Possibly, the fibrils that remained intact, even after elastase digestion, were responsible for inducing NET release by neutrophils.
NET formation is beneficial to the organism because it confines the pathogens and avoids their spreading. The NET-associated enzymes and antimicrobial proteins are potent weapons used by the neutrophils to kill the pathogens. However, NETs cannot be allowed free rein; if they persist too long, NET-associated proteases such as elastase, cathepsin G, and others, might attack and destroy the surrounding tissues (42). Moreover, it has been shown that persistence of NETs induces an immune response (17,42), which has been associated with autoimmune diseases such as lupus erythematosus (17,(42)(43)(44). The elimination of NETs depends on the presence of DNase, and several pathogens, such as Streptococcus pneumoniae and Staphylococcus aureus, escape the web by secreting this enzyme (45,46). In humans, Hakkim et al. (17) have shown that the serum endonuclease DNase I is essential for NET disassembly, and in a subset of systemic lupus erythematosus patients, this enzyme was dysfunctional, leading to NET persistence with a direct correlation with the immunopathogenesis of this disease. Therefore, the observation of NETs in post-mortem tissues of amyloidotic patients poses the question of whether these patients might not have a dysfunctional DNase I, which could compromise the control of NET persistence. Persistence of NETs might aggravate the condition of the amyloid patients either by massively digesting the amyloid fibrils into toxic species, as shown here, or by eliciting an autoimmune response. In the future, it would be interesting to investigate the DNase I serum content and activity in amyloidotic patients to see if there is any correlation with disease progression and aggressiveness.
Another deleterious effect attributed to NETs is deregulation of the fibrinolytic system, leading to hemorrhage (47). Interestingly, in several amyloid diseases, bleeding symptoms and coagulation abnormalities such as hyperfibrinolysis have been  described (48). Thus, it will also be important to investigate whether fibril-induced NET formation as described here is associated with the disruption of hemostasis.
In summary, our data provide evidence that neutrophils and NET release play an important role in amyloidoses and thus should also be investigated in other amyloid-related conditions. Also, the observation of fibril fragmentation (amyloidolysis) in vitro (49) and in vivo (50) along with our data showing that the products of fibril-induced elastase digestion are toxic to cells in culture reinforces the idea of the amyloid fibrils as a reservoir of small noxious species. The proteases present in NETs may be the enzymatic system responsible for amyloidolysis in amyloidotic patients. As far as we know, this is the first report describing amyloid-induced NET formation in vitro with a clear implication for the etiology of the amyloid diseases.