Metal- and UV- Catalyzed Oxidation Results in Trapped Amyloid-β Intermediates Revealing that Self-Assembly Is Required for Aβ-Induced Cytotoxicity

Summary Dityrosine (DiY), via the cross-linking of tyrosine residues, is a marker of protein oxidation, which increases with aging. Amyloid-β (Aβ) forms DiY in vitro and DiY-cross-linked Aβ is found in the brains of patients with Alzheimer disease. Metal- or UV- catalyzed oxidation of Aβ42 results in an increase in DiY cross-links. Using DiY as a marker of oxidation, we compare the self-assembly propensity and DiY cross-link formation for a non-assembly competent variant of Aβ42 (vAβ) with wild-type Aβ42. Oxidation results in the formation of trapped wild-type Aβ assemblies with increased DiY cross-links that are unable to elongate further. Assembly-incompetent vAβ and trapped Aβ assemblies are non-toxic to neuroblastoma cells at all stages of self-assembly, in contrast to oligomeric, non-cross-linked Aβ. These findings point to a mechanism of toxicity that necessitates dynamic self-assembly whereby trapped Aβ assemblies and assembly-incompetent variant Aβ are unable to result in cell death.


INTRODUCTION
Alzheimer disease (AD) is the most common form of dementia, and it is characterized by the deposition of amyloid-b (Ab) and Tau in extracellular plaques and intracellular neurofibrillary tangles, respectively. The amyloid cascade hypothesis implicates the pathological accumulation of Ab and its aggregation from monomers into oligomers and fibrils, as a key event in the development of AD (Hardy and Higgins, 1992). Many pieces of evidence have subsequently shown that the oligomeric form of Ab is the most toxic species, resulting in a reformulated amyloid cascade hypothesis in which Ab oligomers are proposed to be central to AD pathogenesis (Selkoe and Hardy, 2016). Indeed, accumulated evidence shows that Ab oligomers disrupt cellular function in cultured cells and animal models Lambert et al., 1998;Lacor et al., 2007;Zhang et al., 2014;Selkoe and Hardy, 2016). Numerous studies have searched for the elusive ''toxic'' species and tried to characterize its structure. For example, 12mers, *56 KDa, and hexamers have all been implicated as specific structural species that interact with particular receptors (e.g., NDMA) leading to downstream cell death (Lesne et al., 2006;Reed et al., 2011).
Oxidative stress has been proposed as a key mechanism that mediates Ab toxicity (Butterfield and Halliwell, 2019;Butterfield et al., 2013) and, it is potentially one of the earliest sources of damage in human AD (Nunomura et al., 2001). Furthermore, using a cellular model, we have shown that oxidative stress is one of the earliest events induced by Ab oligomers (Maina et al., 2018). One of the ways that oxidative stress causes cellular damage is through protein oxidation. The most common consequences of protein oxidation include amino acid side-chain modification, protein fragmentation, and protein cross-linking (e.g., via dityrosine [DiY] bond formation) (Lund et al., 2011). DiY cross-linking is mediated via carbon-carbon bonding between two proximal tyrosines, resulting in a stable, non-reversible covalent bond (Gross and Sizer, 1959). DiY cross-linking is known to provide elasticity, strength, and stability to proteins and has been shown to form within proteins involved with neurodegenerative diseases (e.g., Ab and a-synuclein) (Galeazzi et al., 1999;Souza et al., 2000). Indeed, we have previously shown the colocalization of DiY with Ab in plaques and a-synuclein in Lewy bodies in human AD and Parkinson disease brain tissue, respectively (Al-Hilaly et al., 2013,2016). In addition to DiY, other modifications such as oxidation of histidine, lysine, and methionine35 (met35) have been shown to occur in Ab (Kowalik-Jankowska et al., 2004;Ali et al., 2005;Palmblad et al., 2002;Friedemann et al., 2015), whereas hydroxylated phenylalanine has been suggested to form cross-links alongside DiY (Zhang et al., 2019). This indicates the potential relevance of these modifications in disease pathogenesis or as markers of disease progression in AD.
To learn more about the importance of DiY cross-linking in AD specifically, several studies have investigated the impact of DiY bond formation on Ab aggregation and toxicity, mostly using metal-catalyzed oxidation (e.g., Cu 2+ /H 2 O 2 ) (MCO) and peroxidase-catalyzed oxidation. Some studies have shown that DiY cross-linking of both Ab40 and Ab42 results in the generation of toxic Ab assemblies with reduced assembly speed (Barnham et al., 2004;Smith et al., 2006;Kok et al., 2013;O'malley et al., 2014Sitkiewicz et al., 2014). Others have implicated DiY in the inhibition of Ab40 assembly, especially in highly oxidative environments (Gu et al., 2018) and it has been demonstrated to be associated with the formation of non-amyloidogenic Ab42 aggregates when catalyzed with a high concentration of Cu 2+ (Smith et al., 2007). However, whether DiY cross-linking is a driver, facilitator, or inhibitor of Ab self-assembly remains unclear. Moreover, given that Ab self-assembly is rapid, the time point of the cross-linking during the assembly process may influence the nature of the cross-linked Ab assemblies. Here, using MCO and UVinduced oxidation to induce DiY cross-linking, we show that the oxidation process results in the stabilization of Ab42 assemblies and either prevents or very significantly slows further elongation. To specifically compare the influence of DiY cross-linking on Ab assembly, we compared the effect of oxidation on a non-assembly competent variant Ab (vAb)  and revealed that oxidation and DiY cross-linking does not induce or promote its assembly. We show that in the absence of H 2 O 2 , CuCl 2 at a concentration similar to the concentration found around Ab plaques (~400 mM) (Lovell et al., 1998) is sufficient to facilitate the DiY cross-linking and formation of Ab42 oligomers into a long-lived oligomer population. A cell live/dead assay revealed that, unlike the self-assembling non-cross-linked Ab, oxidized DiY containing Ab42 is non-toxic to neuron-like cells. Our results suggest that under certain conditions in vitro, oxidation can result in trapping of intermediate species, which cannot elongate further and that are non-toxic to neuroblastoma cells. Together with the observation that non-assembly-competent variant Ab is non-toxic, this reinforces the importance of a continual self-assembly process in mechanisms of Ab toxicity.

In Vitro Metal-Catalyzed Oxidation Results in the Formation of Dityrosine in Wild-Type and vAb Peptides
To investigate the influence of oxidation on Ab assembly, we compared the effect of MCO using CuCl 2 and H 2 O 2 on wild-type Ab42 and variant Ab42 (henceforth called Ab and vAb, respectively) (see Transparent Methods in Information for Authors for more details).
DiY serves as a useful measure of the levels of oxidation and was used here to follow oxidation effects. Other side chains can be oxidized such as Met35, which can be detected using matrix-assisted laser desorption ionization mass spectrometry (Friedemann et al., 2015). We have previously detected DiY cross-links in Ab and a-synuclein using a combination of techniques and shown that DiY can reliably be detected using fluorescence spectroscopy . Here, rapid formation of DiY was detected for Ab and vAb samples that were incubated with both CuCl 2 and H 2 O 2 (Ab/CuCl 2 /H 2 O 2 and vAb/CuCl 2 /H 2 O 2, henceforth called MCO), indicated by the observation of a fluorescence peak at 410 nm after only 5 min. In contrast, samples incubated with CuCl 2 alone or with buffer-only showed no peak at 410 nm ( Figure 1A). Detection of tyrosine with an excitation/emission of 280/305 nm (following quenching using EDTA) showed that the formation of DiY was matched by a decrease in tyrosine fluorescence in both the Ab and vAb MCO reactions compared with the samples incubated with CuCl 2 alone or buffer only ( Figure 1B). DiY fluorescence intensity continued to increase for the MCO reactions up to 2 h but did not increase further after 5 days ( Figure 1C). However, by 5 days, small peaks could be observed for samples incubated with CuCl 2 alone ( Figure 1D), although the DiY signal remained negligible for buffer-only samples after 5 days incubation in the dark. Standard curve estimation of DiY content ( Figure S1) revealed that 2 h MCO of Ab and vAb induced about 5 mM and~12 mM DiY, respectively. As each DiY is contributed to by two molecules of Ab, the percentage of Ab molecules where Tyr is cross-linked is approximately 20% for Ab and 48% for vAb. Incubation of Ab with CuCl 2 alone for 2 h induced~1 mM DiY, which further increased tõ 2 mM at 5 days, which equates to approximately 4% of Ab molecules participating in cross-links. Ab and vAb samples incubated with CuCl 2 showed a decrease in tyrosine fluorescence after only 5 min ( Figure 1B), indicating that the CuCl 2 rapidly induces conformational changes (Roberts et al., 2012)

Metal-Catalyzed Oxidation Influences the Assembly of Wild-Type Ab, but Does Not Affect the Structure and Aggregation Propensity of Variant Ab
We have previously shown that DiY forms in both Ab42 oligomers and fibrils . Thioflavin T (Th-T) fluorescence assay was used to investigate whether DiY formation correlates with changes in the Ab assembly (see Transparent Methods). As expected, the assembly-incompetent vAb incubated in buffer shows no increase in Th-T fluorescence with time . Wild-type Ab gave the expected Th-T spectra showing a sigmoidal curve ( Figure 2A) and the Th-T fluorescence increased further when Ab was assembled in the presence of the metal chelator EDTA, which suggests some impact of trace metals on assembly properties ( Figure S2). However, the Ab and vAb MCO reactions showed no Th-T fluorescence signal increase over the time frame of 50 h, indicating that either no self-assembly had taken place or that the assembly is significantly slow enough that the threshold of Th-T detection had not been reached. Wild-type Ab incubated with CuCl 2 shows a small fluorescence signal for DiY at~2 h ( Figure 1C) and Th-T spectrum showed a short lag phase (approximately up to 2 h) followed by a plateau at a low fluorescence signal. This appears to suggest that the Ab assemblies formed become stabilized without further elongation ( Figure 2A). vAb incubated with CuCl 2 showed no increase in Th-T fluorescence, consistent with the absence of assembly under these conditions.
We have previously shown that our Ab preparation method results in the generation of monomers that assemble into oligomers detected by the oligomer-specific antibody, NU-1 (Lambert et al., 2007), before forming fibrils and amyloid plaques ) (see Transparent Methods). As expected, dot blotting with the NU-1 antibody revealed the presence of oligomers at 2 h in buffer-incubated Ab samples, which disappeared over time ( Figure 2B). In contrast, the buffer-incubated vAb reactions showed no NU-1 reactivity, indicating the absence of oligomer formation as previously reported ( Figure 1F) . Ab oligomers were only minimally detected in the Ab MCO reaction at 5 min, which disappeared over time (Figures 2B and 2C). No NU-1 reactivity was observed for vAb MCO reaction at any of the time points measured. Interestingly, the Ab incubated with CuCl 2 formed more oligomers early on, which persisted throughout the time studied. The data are consistent with the possibility that slower oxidation by CuCl 2 facilitates Ab oligomer formation and stabilization. The formation of DiY cross-linking correlates with this time point when Ab is found in a NU-1 affinity conformation ( Figures 2B and 2C).
Circular dichroism (CD) and transmission electron microscopy (TEM) were used to study the secondary structure and apparent morphologies of the resulting assemblies after 5 days incubation (see Transparent Methods). Spectra from vAb under all three conditions showed a trough at 198 nm consistent with random Freshly prepared Ab and vAb (50 mM) were incubated at 37 C, without CuCl 2 , with CuCl 2 (400 mM), or CuCl 2 in combination with H 2 O 2 . Fluorescence spectra were collected 5 min post-incubation using fluorescent excitation wavelength of 320 nm and emission collected between 340 and 600 nm, with DiY peak signal observed between 400 and 420 nm (A). Fluorescence spectra were also collected at 5 min using an excitation wavelength of 280 nm and emission collected between 290 and 500, with peak tyrosine signal observed at 305 nm (B). Fluorescence spectra were collected again at 2 h (C) and then 5 days (D) to follow DiY formation over time. A minimum of three independent experiments was repeated to ensure the reproducibility of the findings.

OPEN ACCESS
iScience 23, 101537, October 23, 2020 3 iScience Article coil conformation. Ab in buffer showed a minimum at 218 nm consistent with an expected b-sheet conformation, typical of fibrillar Ab ( Figure 2D). Ab CuCl 2 shows a broad but weak minimum at 218 nm, but the signal from Ab MCO reaction is too weak for assignment of secondary structure. This may be consistent with loss of protein from solution, which may arise from the formation of small, amorphous oligomers.
Electron micrographs showed an extensive fibril network for buffer-incubated Ab, whereas CuCl 2 -oxidized Ab showed clumped assemblies with a significantly reduced fiber density, and the MCO Ab sample revealed amorphous structures with very few or no fibers ( Figure 2E). Buffer-incubated vAb sample showed very few assemblies of any kind. The vAb MCO and CuCl 2 reactions exhibited a few amorphous-like clumped assemblies. The CD and TEM data suggest that the MCO-catalyzed oxidation of Ab prevents the formation of b-sheet-rich amyloid fibrils, whereas CuCl 2 -catalyzed oxidation generates oligomeric Ab assemblies with some b-sheet conformation. Taken together, the results from Th-T, dot blotting, CD, and TEM revealed that the rapid DiY cross-linking of Ab induced through MCO correlates with the association and trapping of Ab as intermediates and inhibits their further assembly, whereas CuCl 2 facilitates the formation and DiY cross-linking of Ab oligomers into a long-lived oligomer population. In contrast, none of the conditions induced any assembly of vAb into amyloid fibrils. Instead, the MCO and CuCl 2 reactions resulted in association and trapping of vAb into amorphous assemblies that show no evidence of b-sheet conformation.

Photo-oxidation Induces Dityrosine Cross-Linking in Wild-Type Ab and vAb Peptides
The aforementioned results suggest that copper contributes to DiY formation. However, previous studies have suggested that metals influence the assembly of Ab (Barritt and Viles, 2015;Gu et al., 2018;Smith et al., 2006Smith et al., , 2007, and this is supported by the increased assembly of Ab in the presence of EDTA iScience Article ( Figure S2). Therefore, to avoid the complication of using metals, which may influence assembly, UV photooxidation was used to oxidize Ab and vAb ( Figure 3) (see Transparent Methods). UV oxidation of amino acids is limited to direct damage to Trp, Tyr, His, Cys Figure 7and cystine, but Met can be oxidized indirectly, usually in the presence of a sensitizer (Pattison et al., 2012). As before, the measurement of DiY formation was used to follow oxidation. Fluorescence spectra showed a small peak at 410 nm for both Ab and vAb after only 5 min incubation in UV. However, the intensity was lower than under MCO conditions (Figure 3A). After 2 h of UV exposure, a significant increase in intensity at 410 nm was observed ( Figure 3B). Calculation of DiY content using the standard curve shown in Figure S1 suggested that approximately 12% of Ab and 14% of vAb molecules were involved in DiY, which is lower than following MCO treatment.
After 2 h of UV exposure, samples were stored in the dark but despite this the intensity continued to increase for 120 h following incubation (18% DiY) ( Figure 3C). These results show that UV photo-oxidation can induce DiY cross-linking for Ab and vAb peptides. The results are compared with peptides incubated without UV exposure for reference. Oxidation of other residues by the UV irradiation can not be ruled out, but here we focus on DiY to monitor the level of oxidation.
Photo-oxidization Influences the Assembly of Wild-Type Ab, but Does Not Impact on the Structure and Assembly of vAb Th-T fluorescence was used to monitor the assembly of Ab and vAb following UV photo-oxidation for 2 h. The results are compared with peptides incubated without UV exposure for reference. A small increase in Th-T fluorescence was observed for Ab+UV after a lag phase of 25 h, which may represent a slow assembly of the Ab assemblies that escaped the less-efficient UV oxidation, whereas vAb + UV showed no fluorescence intensity ( Figure 4A). NU-1 dot blots showed no reactivity to vAb + UV ( Figure 4B). The Ab+UV sample showed lower level NU-1 reactivity at 2 h compared with the Ab control ( Figure 4B), and reactivity disappeared after 5 days post-UV incubation in the cross-linked Ab+UV sample, even though a small quantity of oligomers could still be detected in the control Ab sample ( Figure 4B). CD showed spectra consistent with random coil conformation for both vAb and Ab incubated under UV and a loss of signal that was more evident for Ab than for vAb ( Figure 4C). By TEM, Ab+UV and vAb+UV again showed small assemblies at 2 h, which are still present alongside occasionally clumped amorphous-like assemblies after 5 days (Figures 4D and 4E). Similar to the results from MCO ( Figure 2B), this suggests that the UV-induced DiY crosslinking correlates with the stabilization of Ab and vAb in a trapped, assembly-incompetent species. Other oxidation reactions may also be involved.

Co-incubation with Oxidized Ab Slows the Assembly of Freshly Prepared Ab
Our findings thus far suggest that oxidation results in the formation of DiY cross-links and results in stabilized Ab assemblies that are prevented from further elongation into amyloid fibrils. To investigate this were incubated under UV. Fluorescence spectra were collected 5 min postincubation using fluorescent excitation wavelength of 320 nm and emission collected between 340 and 600 nm, with DiY peak signal observed between 400 and 420 nm after 5 min of incubation (A), which increased following 2 h of incubation (B). Fluorescence intensity at 410 nm against time showed that incubation of the 2 h UV-exposed Ab and vAb samples in the dark resulted in further increase in DiY formation in the absence of the UV (C). The Ab and vAb samples that were not exposed to UV showed no DiY signal. A minimum of three independent experiments was repeated to ensure the reproducibility of the findings.

OPEN ACCESS
iScience 23, 101537, October 23, 2020 5 iScience Article further, DiY cross-linking was induced in Ab using 2 h UV exposure, and the sample was then incubated with an equal concentration of freshly prepared Ab (20 mM:20 mM) and compared with a 40 mM Ab sample not exposed to UV (see Transparent Methods for further detail). Fluorescence spectroscopy confirms the presence of DiY in the 40 mM Ab sample exposed to UV, and the signal reduced by half when the oxidized 40 mM Ab was diluted to 20 mM ( Figure 5A); 20 mM oxidized Ab added to 20 mM freshly prepared unoxidized Ab (Ab+UV:Ab) revealed the presence of DiY signal ( Figure 5A). Interestingly, the levels of DiY in the Ab+UV:Ab mixture was higher than that of 20 mM oxidized Ab, suggesting that the presence of the UV-incubated Ab in the environment results in the cross-linking of the freshly prepared Ab following the co-incubation. Th-T fluorescence showed that the 40 mM Ab sample showed a shorter lag phase and higher Th-T signal compared with the 20 mM Ab sample confirming the expected concentration-dependent effect. The 20 mM and 40 mM oxidized Ab samples showed no Th-T fluorescence intensity, similar to previous observations. When the Ab mixtures were incubated together (Ab+UV:Ab 1:1), the mixture showed a longer lag phase (+20 h) but reached a similar Th-T intensity signal to the 20 mM unoxidized sample at 50 h. iScience Article However, this was very significantly lower than the 40 mM unoxidized Ab sample ( Figure 5B). This suggests inhibition of self-assembly in the Ab+UV:Ab-UV mixture. At 4 days, TEM imaging revealed the presence of Ab fibrils in both 20 and 40 mM Ab samples, which were not detected in the photo-oxidized samples (Figure 5C). Only scant fibrils and smaller Ab assemblies could be detected in the Ab+UV:Ab mixture, further confirming a reduced assembly in this mixture ( Figure 5C). Taken together, this suggests that the incubation of freshly prepared Ab with photo-oxidized Ab results in an increase in DiY cross-linking and leads to the stabilization of some Ab assemblies, resulting in the slower aggregation kinetics observed and prevention of elongation. DiY-containing, oxidized Ab species may bind to Ab and inhibit further elongation.

Photo-oxidation of Pre-formed Ab Assemblies Slows Further Assembly of Ab Oligomers/ Protofibrils and Prolongs Their Half-Life
Our findings thus far strongly suggest that oxidation stabilizes and strongly slows the aggregation and further elongation of Ab assemblies. However, it is not clear whether these could also alter the assembly of preformed Ab assemblies. To investigate this, Ab (50 mM) was freshly prepared and allowed to assemble for 24 h (henceforth called aAb). aAb samples were exposed for 2 h to UV to induce DiY cross-linking (aAb+-UV) ( Figure 6) (see Transparent Methods). Unlike samples that were not exposed to UV (aAb), the aAb+UV samples showed an increasing intensity arising from DiY, which continued to increase even after incubation in the dark following UV exposure ( Figure 6A). Th-T fluorescence intensity showed that aAb continues to assemble, reaching plateau after approximately 15 h ( Figure 6B). However, DiY cross-linking induced by UV correlates with the inhibition of further assembly of the aAb+UV for~40 h, suggesting that the photo-oxidation leads to the stabilization or trapping of aAb+UV ( Figure 6B). A gradual increase in fluorescence is observed beyond 40 h, which may indicate delayed assembly. Dot blotting using NU-1 revealed a similar level of oligomers in the oxidized and unoxidized aAb samples at the starting point ( Figure 6C). Interestingly, the oligomers in the oxidized aAb+UV sample persisted beyond 4 days, unlike the unoxidized aAb sample, which showed a very low level of oligomers at the later time point ( Figure 6C). TEM imaging showed the presence of oligomeric assemblies and scant fibers in the oxidized aAb+UV sample, compared iScience Article with the unoxidized aAb, which showed extensive fibril network. Together, these data provide further evidence indicating that photo-oxidation of pre-formed Ab fibrils leads to the formation of DiY cross-links and results in stabilization of aAb assemblies, which prevent or delay further elongation. This is similar to the stabilization of Ab oligomers observed following the slower/milder DiY cross-linking in the Ab/CuCl 2 preparation, which occurred after the formation of Ab oligomers ( Figures 2B and 2C).

Self-Assembly Is Important for Ab Toxicity
Multiple pieces of evidence have shown the detrimental role of Ab on neuronal function that eventually leads to neuronal death Lambert et al., 1998;Lacor et al., 2007;Zhang et al., 2014;Selkoe and Hardy, 2016). Our Th-T, dot blot, and TEM data showed that Ab42 monomers self-assemble to form significant level of oligomers at 2 h, and eventually protofibrils, fibrils, and a network of fibrils at 4 days (Figures 2 and 6). Hence, the rate of self-assembly is high at early time points and plateaus at later time points when Ab forms fibril network and plaques . To investigate the role of selfassembly in toxicity, ReadyProbes live/dead assay was conducted on terminally differentiated SH-SY5Y neuroblastoma cells incubated for 3 days with wild-type Ab42 prepared in the following manner; Ab42 was allowed to assemble for 2, 24, 48, and 96 h to form oligomers, protofibril mixture, and fibrils and then treated for 2 h with UV (Ab+UV) to induce DiY, or untreated (Ab) (see Transparent Methods). vAb was used as an assembly-incompetent control . Our results reveal a significantly higher level of cell death in wells incubated with Ab after 2 h of preparation, compared with cells treated with Ab after 24, 48, and 96 h incubation consistent with previous observations (Figure 7) . However, the none of the oxidized Ab samples containing DiY showed any toxic effect on the cells, similar iScience Article to the assembly-incompetent oxidized and unoxidized vAb (Figure 7). This suggests a key role for self-assembly in the toxicity of Ab.

DISCUSSION
In AD, Ab self-assembles to dimers, oligomers, fibrils, and eventually amyloid plaques, one of the key hallmarks of the disease. Ab is generally accepted to play a key role in AD, but the mechanism behind its toxicity is still not completely understood. Numerous studies have searched for the elusive ''toxic'' species and attempted to characterize its structure, and this has identified supposedly toxic assemblies such as dimers, 12mers, *56 KDa and hexamers (Lesne et al., 2006;Reed et al., 2011;Benilova et al., 2012). Here, we show that oxidation results in the formation of DiY cross-links which is one of several possible oxidative modifications and this significantly slows, or halts the self-assembly of Ab42, trapping it in a specific state. We compared the self-assembly and toxicity of assembly-incompetent vAb with wild-type Ab and oxidized, trapped Ab.
MCO and photo-induced oxidation of vAb revealed that DiY forms very rapidly as early as 5 min postoxidation. However, oxidation did not lead to vAb assembly even after 5 days in the oxidative environment or post-oxidation. This demonstrates that the oxidation and the formation of DiY does not induce aggregation of the vAb, which is known not to self-assemble . DiY was also rapidly induced in wild-type Ab using both MCO and photo-induced oxidation; however, further assembly is inhibited or significantly slowed. Co-incubation of DiY-containing Ab with freshly prepared uncross-linked Ab demonstrated significantly reduced assembly. This suggests that oxidation and DiY cross-linking does not induce or facilitate the aggregation of the wild-type Ab. Instead, Ab assemblies are trapped and further elongation is delayed. This is supported by previous reports that showed that DiY-cross-linked Ab are slow to fibrilize and form long-lived soluble oligomeric aggregates (Kok et al., 2013;O'malley et al., 2014Sitkiewicz et al., 2014). Mass spectrometry studies have revealed that DiY cross-linking leads to the stabilization of Ab40 in compact oligomeric species (Sitkiewicz et al., 2014), which is in strong support of our findings.
Previous studies have suggested that DiY cross-linking can facilitate Ab assembly (Atwood et al., 2004;Yoburn et al., 2003;Barnham et al., 2003;Zhang et al., 2017) or inhibit or slow Ab self-assembly (Smith et al., 2007;Gu et al., 2018) or stabilize assemblies (Vazquez et al., 2019). Importantly, it has been shown that copper influences self-assembly in different ways depending on the concentration and ratio (Matheou Figure 7. Oxidized, DiY-Containing Ab Assemblies Are Not Toxic to Cells Differentiated SH-SY5Y neuroblastoma cells were incubated with UV oxidized or unoxidized vAb or Ab for 3 days, following which the percentage of dead cells was quantified using ReadyProbes Live/Dead Assay. The Ab and vAb samples were freshly prepared and UV oxidized for 2 h (Ab+UV/vAb+UV), or UV oxidized for 2 h followed by a further incubation on bench and in the dark for 24, 48, or 96 h, before being administered to cells. vAb and Ab samples not exposed to UV were used as reference. Only the unoxidized/assembling wild-type Ab induced significant cell death. Experiments were repeated five times. ***p < 0.001. p % 0.05 (*), <0.01 (**), <0.0001 (****) and >0.05 was not significant. Error bars are expressed as GSEM.

OPEN ACCESS
iScience 23, 101537, October 23, 2020 iScience Article et al., 2015;Sarell et al., 2010), which may go some way to explaining these inconsistencies. These discrepancies could also arise from the different methods used to induce DiY in Ab. Previous studies have used varying concentration of horseradish peroxidase and H 2 O 2 (Galeazzi et al., 1999;Yoburn et al., 2003;Ali et al., 2006;Sitkiewicz et al., 2014;O'malley et al., 2014), varying concentrations of copper and H 2 O 2 (Yoburn et al., 2003;Atwood et al., 2004;Barnham et al., 2004;Smith et al., 2007;Davis et al., 2011;Al-Hilaly et al., 2013;Williams et al., 2016), copper in combination with ascorbate (Gu et al., 2018), or photo-oxidation (Yoburn et al., 2003;Williams et al., 2016) to induce DiY cross-links in Ab. In addition, even small differences in incubation conditions (e.g., trace metals, temperature, etc.) as well as atmospheric ozone levels (Vazquez et al., 2019) would result in different levels of DiY cross-links formed, which may result in varying impact on Ab assembly. Moreover, the nature of the DiY cross-links may differ from one protocol to another. For example, other amino acids, such as phenylalanine, could modulate formation of cross-links (Zhang et al., 2019). Methodologies are likely to induce different oxidative effects on other amino acid side chains such as histidine, lysine, and met35 of Ab (Kowalik-Jankowska et al., 2004;Ali et al., 2005;Palmblad et al., 2002;Friedemann et al., 2015). For example, met35 oxidation has been shown to attenuate Ab oligomer formation and to enhance oxidation of Ab Y10 (Palmblad et al., 2002). Mapping the specific modifications induced by the oxidation methods in a time-dependent manner using mass spectrometry would help to provide clarity on the relative role for DiY on Ab assembly and whether this occurs in combination with other modifications to Ab. This is a part of further studies.
Furthermore, the method of Ab preparation used before DiY cross-linking may be critical in the outcome of the DiY cross-links on Ab. Different sources and methods are used to prepare Ab, resulting in a diverse pool of Ab aggregates (Benilova et al., 2012). Ab exists in a pool of monomers, soluble oligomers, and insoluble fibrils. Multiple studies have reported that the soluble Ab oligomers in AD are composed of dimers, trimers, tetramers, pentamers, and decamers; Ab-derived diffusible ligands (ADDLs); dodecamers; and Ab*56 (Benilova et al., 2012). A question therefore arises regarding how DiY cross-linking impacts these pools of assemblies: presumably by creating new cross-linked dimeric species. Previous molecular dynamics studies have revealed that induction of DiY in a pool of monomeric Ab42 results in conformationally altered dimers that expose hydrophobic residues that may be limited to forming trimers via hydrophobic rather than polar interactions (Zhang et al., 2017). Our previous data showed that within 2 h of preparation, Ab exists mostly as oligomers with a random coil conformation with a small b-sheet contribution . Here, our results showed that the very slow oxidation induced by CuCl 2 alone first facilitates the formation of Ab oligomers followed by DiY cross-linking of the oligomers resulting in a stabilized oligomeric population. The more rapid MCO reaction results in DiY formation as early as 5 min post-oxidation whereby the Ab becomes trapped in a pre-oligomeric conformation (as assessed using the antibody NU-1). Th-T fluorescence, CD, and TEM showed that photo-oxidation of early Ab species traps Ab in a random coil conformation and prevents or significantly delays further assembly into amyloid fibrils. UV oxidation and DiY cross-linking in preformed oligomer/protofibril assemblies similarly results in the stabilization of this state and significantly delays further elongation to fibrils. Taken together, these results suggest that the timing of oxidation of Ab critically influences its assembly, leading to the stabilization or significantly reduced assembly of the Ab assemblies, which correlates with the time of cross-linking.
Ab self-assembly is believed to be important for its toxicity , and many studies have implicated the role of oligomeric species in cytotoxic effects (Glabe and Kayed, 2006;Soura et al., 2012;Walsh et al., 1999). Here, we compared toxicity of the unoxidized Ab with Ab that had been photo-oxidized and DiY cross-linked in vitro, whereby specific species in the assembly pathway have been stabilized. Ab42 was oxidized at different time points to stabilize a series of pre-oligomer, oligomeric, protofibrillar, and fibrillar forms. We show that none of these species is able to induce cell death following 3 days of incubation with differentiated neuroblastoma cells, whereas unoxidized, oligomeric Ab remained potently toxic. This finding is in conflict with previous studies that showed that DiY Ab42 assemblies are toxic to cells (Barnham et al., 2004); DiY-cross-linked Ab40 dimers induce cell viability loss (Kok et al., 2013) and that Th-T positive, DiY-cross-linked Ab40 fibrils were able to inhibit long-term potentiation (LTP) (O'malley et al., 2014). We also observed that oxidized vAb is not toxic to cells, suggesting that the presence of DiY alone is not sufficient to induce toxicity. The oxidized and DiY-cross-linked Ab42 assemblies produced here are different from DiY Ab40 reported by others (O'malley et al., 2014;Kok et al., 2013), as the DiY Ab42 produced shows little Th-T fluorescence intensity and does not proceed to form fibrils. However, we do not rule out the possibility that DiY Ab preparation in our study and others also results in other oxidative modifications, which may explain the discrepancies between these studies.

OPEN ACCESS
iScience Article It is important to note that method of peptide preparation, peptide type, peptide concentration and aggregation state, and model system used may play a huge role in determining cell toxicity (Jana et al., 2016;Kaniyappan et al., 2017;Cecchi et al., 2008;Krishtal et al., 2015). Previous studies on DiY Ab toxicity have used LTP (O'malley et al., 2014), MTS proliferation assay (Barnham et al., 2004), MTT. and lactate dehydrogenase assay (Ono et al., 2009) with varying Ab concentrations to determine DiY Ab toxicity. These assays, although showing the presence of cell injury, do not quantify absolute cell death. Impaired spine morphology and density, accompanied by increased reactive oxygen species and intracellular calcium, without apparent cell death have been reported as a result of tau toxicity (Kaniyappan et al., 2017). We have shown that a very short, 2 h exposure to Ab oligomers of differentiated SHSY5Y cells result in oxidative and nucleolar stress without DNA damage or neuronal death (Maina et al., 2018). Thus, some discrepancies may arise from the assays and the cell model used. Moreover, if DiY Ab is toxic, then it may depend on the level of DiY. None of the previous studies that studied DiY Ab toxicity quantified the level of DiY formed on Ab. As a result, discrepancies may also arise from the differences in the quantity of DiY cross-links in the Ab used in toxicity assays. Indeed, we observed different levels DiY intensity between MCO and UV treatments.
Nonetheless, here we show that oxidation of Ab in vitro leads to formation of DiY, halts Ab self-assembly, and prevents cytotoxicity in a live-dead assay. We have previously demonstrated that assembly-incompetent vAb is not toxic to cells . We therefore conclude that continued self-assembly is important for Ab toxicity. We believe that the timing of the oxidation may be critical. For example, formation of DiY in Ab fibrils would promote its stability and formation of amyloid plaques. Indeed, we have previously shown the presence of DiY on Ab plaques in AD brain tissue and demonstrated that DiY Ab fibrils become highly insoluble and resistant to formic acid denaturation .
In conclusion, oxidation, which results in DiY cross-linking, promotes Ab stabilization and does not induce or facilitate Ab assembly. Our findings strongly suggest a role for self-assembly for Ab toxicity. We show that Ab exerts a high level of toxicity at a stage when self-assembly potential is high, compared with when the self-assembly rate is significantly diminished or abolished, as is the case for oxidized and vAb. This is observed even for those preparations wherein oligomeric Ab has been stabilized. Our work implies that the timing of DiY formation plays a key role in further assembly and stability of Ab.

Limitations of the Study
Here, we have provided evidence to show that oxidative conditions can induce the formation of DiY crosslinks in Ab42 using MCO and UV photo-oxidation in vitro. We show that oxidation under the conditions used here halts further assembly. Stabilized Ab42 following oxidation is non-toxic to differentiated neuroblastoma cells. However, our study was not able to fully characterize whether other amino acids were affected by oxidation and what impact this might have on the prevention of assembly. We confirm that DiY is a major outcome of oxidation. Our work shows that oxidation of Ab in vitro results in formation of non-toxic Ab species. However, oxidative stress is known to be an important trigger for neurodegenerative diseases and our results do not imply a protective effect of oxidative stress. Oxidation has been performed under controlled environment in vitro affecting only Ab self-assembly. Oxidation of Ab in vivo is likely to have made diverse effects that have not been addressed in this study.
were prepared in Milli-Q water. Each concentration was prepared in triplicate and dityrosine fluorescence was recorded for each concentration using excitation wavelength 280 nm and emission wavelength 410 nm. The mean values of dityrosine fluorescence intensity for each concentration were plotted against dityrosine concentration and line plot was constructed by linear regression analysis using Microsoft Excel software. The equation of this line was used to quantify the dityrosine content of Aβ samples oxidized with MCO or CuCl2 alone.

Photo-oxidation of Aβ and vAβ
Freshly prepared samples of Aβ1-42 and vAβ1-42 peptides (50 μM) in 10 mM phosphate buffer, pH 7.4 were incubated i) without UV-C in the dark, and ii) under of UV-C for 5min or 2h using a G6T5 Germicidal 9′ 6W T5 UVC lamp set to 8 J/m 2 /sec (General Lamps Ltd). A minimum of three independent experiments were conducted to ensure the reproducibility of the findings.

Fluorescence spectroscopy
The formation of dityrosine was monitored with a fluorescence spectrophotometer (Varian Ltd., Oxford, UK), using a 1 cm path length quartz cuvette (Starna, Essex, UK). The presence of dityrosine was detected using fluorescent excitation wavelength of 320 nm and emission collected between 340 -600 nm, with dityrosine peak signal observed between 400-420 nm.
Tyrosine fluorescence signal was monitored using an excitation wavelength of 280 nm and an emission wavelength of 305 nm, with the peak tyrosine emission observed at 305 nm. For experiments involving metal-catalysed oxidation, the reaction was quenched using EDTA to a final concentration of 2 mM. For all the measurements, the excitation and emission slits were both set to 10 nm, scan rate set to 300 nm/min with 2.5 nm data intervals and an averaging time of 0.5 s. The photomultiplier tube detector voltage was set at 500 V.

Thioflavin T fluorescence assay of Aβ self-assembly
Samples were incubated with 100 μM Thioflavin T (Th-T), and the rate of Th-T binding was monitored over time at 37°C using SpectraMax i3 plate reader with samples incubated in

Circular Dichroism (CD)
The secondary structure of Aβ and vAβ peptides at 50 μM concentration in 10 mM phosphate buffer (pH 7.4) incubated under different conditions was assessed using Jasco J715 CD spectrometer (Jasco, Goh-Umstadt, Germany). 140 μL of each sample was placed into a 1 mm path length quartz cuvette (Hellma) and scanned between 190 nm and 260 nm. The CD spectra were collected in triplicate at a maintained temperature of 21 °C.

Negative-stain transmission electron microscopy (TEM)
The morphology of control and cross-linked Aβ and vAβ peptides was assessed by negative stain TEM. Briefly, 4 μL of each sample was dropped onto 400-mesh carbon-coated grids (Agar Scientific, Essex, UK). After 1 min incubation, the excess sample was blotted using filter paper, and the grid was washed with 4 μL filtered Milli-Q water and blotted. The grid was then negatively stained for 40 sec using 4 μL of filtered 2% (w/v) uranyl acetate. The excess stain was blotted with filter paper and grids left to air-dry before storage. The grids were examined on a Jeol Jem1400-plus transmission electron microscope (Jeol, USA), operated at 80 kV fitted with a Gatan Orius SC100 camera (UK).

Dot-blotting
A total of 4 μl was spotted onto a 0.2 μM pore nitrocellulose membrane and allowed to dry for 10 min. The membrane was boiled with PBS for 1 min twice and then blocked with blocking buffer (5% milk in 0.05% TBS-T) for 1 hour at room temperature on a rocker. The blocking buffer was next replaced with mouse NU-1 primary antibody (1/2000) and left to bind overnight at 4°C on a rocker. The membrane was washed 6 times for 5 min with washing buffer (0.05% TBS-T), then incubated with an HRP-conjugated goat anti-mouse secondary antibody for 1 hour. The membrane was washed six times for 5 min with washing buffer, then incubated with Clarity Western ECL Substrate (Bio-Rad) for 1 min before being developed in the darkroom.
The NU-1 antibody was a gift from the William Klein lab (Lambert et al.). A minimum of three independent experiments were conducted to ensure the reproducibility of the findings. Fetal Calf Serum supplemented with 10 μM trans-Retinoic acid (Abcam) for 5 days. Next, the medium was replaced with a serum-free media supplemented with 2 nM brain-derived neurotrophic factor (BDNF) (Merck Millipore). After 2 days in the BDNF-containing media, the media was replaced with serum-free media and the cells were treated with UV cross-linked or uncross-linked vAβ or Aβ for 3 days. At the end of the incubation period, the cells were incubated with ReadyProbes reagent (Life Technologies) for 15 min. The ReadyProbes kit contains NucBlue Live reagent that stains the nuclei of all live cells and Propidium iodide that stains the nuclei of dead cells with compromised plasma membrane. The cells were imaged at 37 o C and 5% CO2 using Operetta CLS high-content analysis system (PerkinElmer) using DAPI and TRITC filters. At least 5000 dead and live cells were analysed using the Harmony software automated analysis algorithm within the Operetta CLS high-content analysis system.

Cell death assay
A minimum of three independent experiments were repeated to ensure the reproducibility of the findings.

Supplemental Results
Figure S1  assembly under MCO conditions, copper alone and in buffer with EDTA to chelate trace metals. Ab1-42 self-assembly at a higher rate than Ab1-42 in buffer suggesting that the trace metals present in the water used to make the buffer can influence the assembly rate.