Yanxian Zhang,Yijing Tang,Dong Zhang,Yonglan Liu,Jian He,Yung Chang,Jie Zheng,*

1 Department of Chemical,Biomolecular,and Corrosion Engineering,The University of Akron,OH,USA

2 Department of Chemical Engineering,R&D Center for Membrane Technology,Chung Yuan Christian University,Taoyuan,Taiwan,China

ABSTRACT Amyloid cross-seeding of different amyloid proteins is considered as a highly possible mechanism for exacerbating the transmissible pathogenesis of protein misfolding disease (PMDs) and for explaining a molecular link between different PMDs,including Alzheimer disease (AD) and type 2 diabetes (T2D),AD and Parkinson disease(PD),and AD and prion disease.Among them,AD and T2D are the most prevalent PMDs,affecting millions of people globally,while Aβ and hIAPP are the causative peptides responsible for AD and T2D,respectively.Increasing clinical and epidemiological evidences lead to a hypothesis that the cross-seeding of Aβ and hIAPP is more biologically responsible for a pathological link between AD and T2D.In this review,we particularly focus on(i)the most recent and important findings of amyloid cross-seeding between Aβ and hIAPP from in vitro, in vivo,and in silico studies,(ii) a mechanistic role of structural compatibility and sequence similarity of amyloid proteins(beyond Aβ and hIAPP)in amyloid cross-seeding,and (iii) several current challenges and future research directions in this lessstudied field.Review of amyloid cross-seeding hopefully provides some mechanistic understanding of amyloidogenesis and inspires more efforts for the better design of next-generation drugs/strategies to treat different PMDs simultaneously.

Keywords:Amyloid peptide Amyloid aggregation Amyloid cross-seeding Amyloid-β hIAPP Protein misfolding

1.Introduction of amyloid cross-seeding

The misfolding and aggregation of amyloid proteins/peptides into polymorphic aggregates with β-sheet-rich structures are considered as a common neurodegeneration mechanism for different protein misfolding diseases (PMDs).Generally speaking,a specific PMD is pathologically associated with the aggregation of specific amyloid proteins,e.g.,Alzheimer disease (AD),type 2 diabetes(T2D),and Parkinson disease (PD) are linked to the aggregation of amyloid-β (Aβ),human islet polypeptide (hIAPP),and αsynuclein (α-syn),respectively.More importantly,the copresence of amyloid aggregates of different sequences in the same tissues/organs has also been observed and linked to the cooccurrence of different PMDs in the same individual with the higher risk of aggression and progression of these diseases.This suggests that molecular cross-talk between different PMDs is possibly attributed to the interactions between their respective disease-causative proteins,leading to amyloid cross-seeding [1,2].

While the exact amyloid cross-seeding mechanisms are still under investigation,some underlying sequence-structureaggregation relationships could be revealed from different viewpoints.(1)From a broader viewpoint,amyloid cross-seeding occurs not only between the naturally-occurring amyloid proteins,including Aβ and hIAPP[3–6],Aβ and α-synuclein[7],Aβ and tau[8],Aβ and transthyretin [9],hIAPP and insulin [10],but also between non-amyloid-proteins (i.e.bacterial-produced curli) and amyloid proteins(i.e.Aβ,hIAPP)[11].These important findings suggest that amyloid proteins/aggregates can be transported between different types of cells,which also increases a possibility not only for crossseeding between different amyloid proteins/aggregates,but also for the spread of disease-related amyloids to different tissues/organs involved in different PMDs [12].(2) From a structural viewpoint,different amyloid aggregates of distinct sequences,origins,and biological functions contain β-sheet-rich structures,which provide a common structural basis for amyloid cross-seeding.It is intuitive to speculate that amyloid cross-seeding requires compatible structure selection and association [13].Amyloid cross-seeding produces hybrid fibrils containing conformationally similar cross-β-sheet structures to pure amyloid fibrils [13–16].Amyloid aggregates of one species with dominant β-structures could serve as a structural template to recruit another amyloid monomers or aggregates with less populated β-structure to form hybrid amyloid aggregates.Alternatively,different amyloid proteins could mutually adjust and optimize their conformations to achieve binding preference,leading to cross-seeding.Both amyloid cross-seeding scenarios will alter the aggregation propensity(promotion or inhibition or both at different aggregation stages)of different amyloid proteins.(3)From a free energy viewpoint,different from homologous-seeding that always results in the acceleration of amyloid aggregation,not any two amyloid proteins always enable cross-seeding;even if occurring,amyloid cross-seeding could lead to the acceleration of amyloid formation,the inhibition of amyloid formation,or a more complex scenario of the co-existence of crossamyloid acceleration and inhibition behaviors at different aggregation stages.All these scenarios suggest the existence of crosssequence barriers along the folding and binding pathways of different amyloid proteins.

Despite fundamental and biological significance,amyloid crossseeding is still a subject poorly explored,as compared to the continuously increased research on homo-amyloid aggregation(Fig.1).The early studies of amyloid cross-seeding can be tracked back to 90s [17],there are lack of continuous and consistent studies on amyloid cross-seeding since then (Fig.1).Much less is known about the amyloid cross-seeding mechanism and its underlying sequence-structure-aggregation relationship.In this review,we aim to provide a different perspective to summarize all possible aspects of amyloid cross-seeding between Aβ and hIAPPin vitro,in vivo,andin silicoat different levels.Selection of Aβ-hIAPP cross-seeding system is simply because of (i) biomedical significance of both AD and T2D in terms of the pathology of each disease and a potential pathological link between both diseases and(ii)our recent studies on amyloid cross-seeding between Aβ and hIAPP,between hIAPP and rIAPP,and between tau fragments [3–6,18–20],all related to AD and T2D.In Table 1,we summarize all typical amyloid cross-seeding between Aβ and hIAPP over the past decade.This review mainly covers fundamental principles of cross-seeding phenomena fromin vitro,in vivo,andin silicoaspects,selectively highlights some typical and interesting Aβ-hIAPP cross-seeding systems,and finally presents some personal opinions about current challenges and future directions in amyloid cross-seeding research beyond Aβ-hIAPP systems.Hopefully,this review will stimulate more research efforts to explore different aspects of amyloid cross-seeding phenomena,which will offer the better mechanistic understanding of amyloidogenesis,transmission,and prevention of PMDs.

Fig.1.Publications,searching for keywords of “amyloid aggregation”or “amyloid cross-seeding”from Web of Science.

2.Molecular and structural basis of amyloid cross-seeding between Aβ and hIAPP

AD and T2D are the two most common PMDs [36,37],both of which have affected millions globally.Clinical and epidemiological studies have showed a potential link between AD and T2D[38,39].It is reported that～81% of AD patients had either T2D or glucoserelated disorders [40],while T2D patients also had a higher risk to develop AD pathogenesis than healthy individuals [41,42].Specifically,each of AD and T2D is identified as a risk factor to elevate the incidence of another disease,and both diseases induce some overlapping pathological symptoms including insulin resistance and deficiency,impaired insulin receptor,impaired insulin growth factor signaling,reduced cerebral glucose metabolism,cerebrovascular injury,excess oxidative stress,and vascular inflammation [31,43–59].Some risk factors including advancing age,obesity,and a sedentary life style are known as common factors for the development of both diseases [60].

Considering that both AD and T2D are PMDs,whose pathological hallmarks are associated with the accumulation and deposition of protein aggregates of Aβ and tau in the brain in AD and human islet amyloid polypeptide(hIAPP)in pancreatic islets in T2D.While it is not clear how the two diseases are connected,several lines of evidences at cell and protein levels appear to support the hypothesis that the AD-T2D link could arise from amyloid cross-seeding between Aβ and hIAPP (Fig.2).(1) Aβ and hIAPP are co-existence in blood vessels and cerebrospinal fluids with similar nanomolar concentrations [31];(2) Aβ is found to co-localize with hIAPP in pancreatic islet amyloid deposits of T2D patients [27];(3) hIAPP,normally co-secreted with insulin,is also expressed by the sensory neurons and found at hindbrain[44,52,61],(4)hIAPP and Aβ show high degrees of sequence identify (25%) and similarity (50%)[31,62],and both can misfold and self-aggregate into similar Ubend fibrillar structures[31,39],(5)Our experimental and computational studies[3–6]have found that full-length Aβ42and hIAPP37proteins can cross-interact with each other to form hybrid amyloid fibrils containing morphologically similar β-sheet-rich structures to pure Aβ and hIAPP fibrils.While the precise mechanism for the co-occurrence of AD and T2D and their relevant links between the two PMDs is still unclear,[1,2,63,64]these evidences at least provide a fundamental basis to reveal some clues for amyloid cross-seeding between Aβ and hIAPP.

Generally speaking,amyloid cross-seeding of different amyloid proteins(not only limited to Aβ and hIAPP)is presumably built on a primary principle of conformational compatibility between different amyloid aggregates,in which both partners of different amyloid species need to dynamically and mutually adjust their conformations to achieve co-aggregation.The most common conformational feature of amyloid aggregates,irrespective of their sequences,is a unique β-sheet structure[65],which acts as a general structural motif and interaction template for amyloid crossseeding.Conceptually,two cross-β sheets of different amyloid aggregates enable a possibility to co-aggregation through the conformational selection and association in a bidirectional or unidirectional manner,depending on aggregation differences in the folding kinetics and populations of β-sheet structures of different amyloid proteins.Also,similar β-structures from different amyloid aggregates could effectively lower cross-species barriers,and thus greatly increase a possibility for amyloid cross-seeding even if they have large differences in their sequences.Amyloid cross-seeding process is an outcome of the intermolecular interactions between self-seeded and cross-seeded aggregates in both competitive andcooperative manners,thus leading to different cross-seedinginduced slower or faster aggregation scenarios at different aggregation stages (Fig.3).In some cases,different amyloid proteins,due to their conformational complexity and polymorphism [66–69],will select conformationally similar seeds from vast conformational samplings for co-aggregation(Fig.3a).Usually,such searching of conformational similar aggregates formed by different amyloid proteins requires exhaustive traverse of a huge sampling space and across large energy barriers,thus often leading to the slower amyloid cross-seeding or the lower efficiency of amyloid cross-seeding,as compared to homo-seeding.This further implies the existence of cross-species barriers [64,70].On the other hand,if one type of amyloid proteins exhibits the faster selfaggregation property than another amyloid protein,it is likely that the preformed aggregates serve as a structural template to recruit similar or dissimilar amyloid monomers to form larger amyloid cross-seeds along the energy-downhill aggregation pathways(Fig.3b) [13].In both amyloid cross-seeding scenarios,conformational compatibility between different amyloid proteins is a key for amyloid cross-seeding.From a different viewpoint,the high structural complexity and polymorphism of amyloid proteins is considered as a key energetic and physical barrier to greatly impede the cross-seeding ability,because the diverged structural forms would decrease effective templates for hybrid amyloid fibril growth.In summary,if different amyloid aggregates have significant differences in their conformational distributions and populations,there is no structural basis for initiating amyloid crossseeding.On the contrary,if the dominant conformations of two amyloid species are sufficiently similar to each other,crossseeding barriers become low enough to drive them to interact with each other and grow into hybrid fibrils,realizing amyloid crossseeding.

Table 1 Summary of amyloid cross-seeding between Aβ and hIAPP

3.Cross-seeding of Aβ and hIAPP

The presence and interaction of Aβ and hIAPP have been considered as a main cause to induce,spread,and explain the pathophysiology and co-existence of AD and T2D in the same individual.Different from other pairs of amyloid cross-seeding proteins,Aβ and hIAPP show high degrees of sequence identify (25%) and sequence similarity (50%),especially these identical and similar sequences are mainly located at the β-strand forming region [62]While the exact role of amyloid sequence similarity/identify in amyloid cross-seeding still remains unclear,the cross-seeding possibility and efficiency is likely to be higher if amyloid sequences are mainly located in amyloid β-structure region.As a result,Aβ and hIAPP adopt similar U-bend fibrillar structures [31,39],which may provide a common structural basis and interaction template for amyloid cross-seeding.While the exact correlation between the Aβ-hIAPP interactions and the AD-T2D link is still under investigation [1,2,64],we summarize some important studies of amyloid cross-seeding between Aβ and hIAPPin vivo,in vitro,andin silico,which are still a less-explored subject.

3.1.Cross-seeding of Aβ and hIAPP in vivo

Fig.2.Evidences(1–5)to support Aβ-hIAPP cross-seeding.(1)Co-existence of Aβ and hIAPP in blood vessels and cerebrospinal fluids.(2)Localization of Aβ in pancreatic islet amyloid deposits of T2D patients.(3)Expression and existence of hIAPP in sensory neurons and hindbrain of AD patients.(4)High sequence identity(red color)and similarity(green color) between Aβ and hIAPP.(5) Capability of Aβ and hIAPP to form hybrid fibrils containing morphologically similar β-sheet-rich structures to pure Aβ and hIAPP fibrils.

Fig.3.Mechanistic models of “template-assisted”and “conformational selection”for amyloid cross-seeding.

Clinical and epidemiological studies have reported (i) the cooccurrence of both AD and T2D in the same individuals [39],(ii)the increased risk for developing either PMD if another one is presented [38,41,42],(iii) the colocalization of Aβ and hIAPP plaques in human AD brain and T2D pancreas tissues [26,27],and (iv) the co-presence of Aβ and hIAPP in human plasma and cerebrospinal fluid (CSF) [21,71].All these evidences suggest a possible amyloid cross-seeding between AD-causative Aβ and T2D-causative hIAPPin vivo.

Evidently,the early reports in rat model have shown a widespread distribution of hIAPP immunoreactivity in central nervous system [53,72].More recently,in the Tg2576 transgenic mouse model of AD,immunostaining of both endogenous rodent IAPP and Aβ exhibited a overlap between both Aβ and IAPP amyloid plaques in serial sections of brain tissues,suggesting a closely associated accumulation of IAPP with Aβ [21].Later,similar hIAPP plaques were observed in the temporal lobe gray matter from T2D patients and the brain parenchyma of AD patients without T2D,where hIAPP depositions were identified not only next to or embedded into Aβ plaques in brain parenchyma,but also in blood vessels of AD patients without T2D(Fig.4a),suggesting a potential circulation influx induced cross-interaction between hIAPP and Aβ in human brain of AD patients[26].Apart from the fact that hIAPP indeed colocalized with Aβ plaques in the brain models,Aβ accumulation was also found to elevate hIAPP deposition levels.In 8-month diabetic rats model,an increase of Aβ accumulation caused more severed neuronal loss and islet β-cell death,indicating a possible interplay between hIAPP and Aβ [23].Another interesting study reported that intravenous injection of preformed hIAPP or Aβ fibrils into human IAPP transgenic mice presented an enhanced hIAPP amyloid formation,as determined by the increased number of islets with amyloids from 2.7%to 24.1%for homologous seeding of pure hIAPP and 15.2%for cross-seeding of hIAPP with Aβ,respectively.This observation demonstrated thein vivoco-existence of homologous seeding of pure hIAPP and cross-seeding of hIAPP with Aβ [22],but the cross-seeding of hIAPP with Aβ was less efficient than homologous seeding of pure hIAPP,consistent within vitroobservations[3]and supporting the existence of cross-species barriers.In human sample study,upon examining a total of 1,092 elderly population,hIAPP contents in human plasma exhibited a positive correlation with both Aβ1-42(r=+0.20,p<0.0001) and Aβ1-40(r=+0.12,p<0.0001) [71].An ELISA analysis of hIAPP contents in different human cerebrospinal fluids(CSFs) acquired from 7 normal and 7 AD cases in a brain bank presented a significantly elevated hIAPP contents in AD’s CSFs (～1.9 × 10-12mol·L-1),as compared to normal CSFs (～1.5 × 10-12mol·L-1,p=0.013) [21],and further support the cross-seeding ability of Aβ with hIAPPin vivo.

In parallel to the cross-seeding of Aβ with hIAPPin vivo,amyloid cross-seeding of hIAPP with Aβin vivowas also studied.In a double transgenic mice model that can express both IAPP and β-amyloid precursor protein(APP)human genes,it was found that Aβ deposition in brain was a 4-fold higher than that in single APP transgenic mice.As a result,both proteins were detected in amyloid plaques,suggesting a cross-seeding effect between hIAPP and Aβ.Furthermore,intracerebral inoculation of hIAPP aggregates to APP transgenic mice showed a significantly higher Aβ deposits in cortex and the hippocampus areas as compared to the control mice group.Consequently,the cross-seeding of hIAPP with Aβ accelerated amyloid aggregation and caused more severe cognitive impairment[25].Another study of hIAPP transgenic mice showed an increased Aβ deposition in hippocampi[73].These observations suggest that the over-expression of hIAPP is closely implicated in Aβ deposition in central nervous system,leading to AD-like pathology and behavior.Mutually,the cross-seeding of hIAPP with Aβ also occurs in pancreas.In a double transgenic mouse model that co-expresses Aβ and hIAPP amyloids,an increased amyloid deposition that composed of both hIAPP and Aβ was found in islets.And the Aβ amyloids in pancreas was only observed in the presence of hIAPP,suggesting the cross-seeding capability of hIAPP for Aβ[24].Additional evidence included that pancreas section extracted from T2D patients presented a combined deposition of Aβ and hIAPP in islets as recognized by anti-amylin and anti-Aβ antibodies(Fig.4b),suggesting that amyloid plaques in T2D pancreas are the mixture of hIAPP and Aβ [27].

Fig.4. In vivo cross-seeding of Aβ and hIAPP.(a) Colocalization of hIAPP and Aβ in brain sections of AD patients by co-staining with anti-hIAPP and anti-Aβ antibodies.Distinct patches of hIAPP and Aβ are shown in blood vessels(upper panel)and brain parenchyma(bottom panel).(Reprinted with permission[26],Copyright 2013 John Wiley &Sons) (b) Co-immunostaining of pancreas section of T2D patients by antiamylin monoclonal antibody (red),polyclonal antibody to the C terminus of Aβ40(green),and the co-localization of amylin and Aβ (orange).Scale bar=120 μm.(Reprinted with permission [27],Copyright 2010 Elsevier) (c) Schematic of in vivo cross-seeding between Aβ and hIAPP linked to both AD and T2D pathologies.Both Aβ and hIAPP transport through peripheral circulation to cross-seed with each other in brain and pancreas.

Despite the growing evidences to support the cross-seeding between Aβ and hIAPPin vivo,little is known about how AβhIAPP interactions occur,whether hIAPP found in brain and Aβ deposited in pancreas are locally produced or recruited from peripheral circulation,or in a combined way.Several studies have shown that small hIAPP and Aβ aggregates can not only cross the blood brain barrier (BBB) [74],but also appear inside vessel walls and pericapillary space [22,75–77].There also exists several evidences on the expression of APP in human pancreas[27]and IAPP in rat and primates brain [78,79].All of these intracellular places provide a peripheral circulation environment for Aβ-hIAPP coexpression and cross-seedingin vivo(Fig.4c).As a less-studied topics,more and in-depthin vivostudies should be conducted to examine the effects of production and circulation of both Aβ and hIAPP on the initiation and progression of cross-seeding between Aβ and hIAPP.

3.2.Cross-seeding of Aβ and hIAPP in vitro

Fig.5. In vitro cross-seeding of Aβ and hIAPP.Time-dependent ThT fluorescence curves for(a)Aβ40monomers in the absence and presence of hIAPP37fibrillar seeds.(Adapted from Fig.2 with permission [29],Copyright 2014 2014 Elsevier B.V.) (b) Pure Aβ42,pure hIAPP37,and cross-seeding of Aβ-hIAPP at different concentrations.(Adapted from Fig.5 with permission [3],Copyright 2015 American Chemical Society) (c) Pure Aβ40,pure hIAPP37,and cross-seeding of Aβ40-hIAPP37.(Reprinted with permission [30],Copyright 2007 WILEY-VCH Verlag GmbH &Co.KGaA,Weinheim) (d) hIAPP37in the presence of Aβ42at different molar ratios (Adapted from Fig.1 with permission [33],Copyright 2020,Springer Nature).(1 M=1 mol·L-1).

In parallel to cross-seeding of Aβ and hIAPPin vivo,here we aim to summarize allin vitrostudies of amyloid cross-seeding of Aβ and hIAPP [80,81].From an interaction viewpoint,amyloid crossseeding may favor some directional interactions and specific conformations,showing a directional effect,i.e.the interaction between different amyloid proteins may work in both directions or in a single direction.For example,the first report of amyloid cross-seeding of hIAPP with both Aβ40and Aβ42accelerated hIAPP fibrillization by shortening the lag phase of hIAPP aggregation but hIAPP seeds were almost inert for Aβ aggregation.This suggested that Aβ fibrils acted as good seeds for promoting hIAPP aggregation,while hIAPP fibrils had little or no effect on Aβ fibrillization[28]showing unidirectional cross-seeding barriers.However,in another study,adding hIAPP fibrils to fresh Aβ40/Aβ42monomers enabled to promote the Aβ aggregation process(Fig.5a)[29]which is consistent with the following study showing that the addition of preformed hIAPP seeds to Aβ40monomers accelerated Aβ40aggregation [25].

Different from amyloid cross-seeding between preformed seeds and monomers that often exhibit a directional effect,amyloid cross-seeding between Aβ and hIAPP monomers exhibits some controversial observation,strongly depending on experimental conditions (seeding concentrations,sequence specificity,even batch quality of amyloid proteins).For instance,monomeric Aβ and hIAPP can cross-seed with each other to form hybrid amyloid fibrils containing morphologically similar β-sheet-rich structures to pure Aβ and hIAPP [3],indicating a bidirectional cross-seeding effect.But it should be noted that bidirectional cross-seeding does not necessarily imply either acceleration or inhibition of the coaggregation of different amyloid proteins.The cross-seeding of Aβ and hIAPP monomers showed the coexistence of the retarded process at the initial nucleation stage and the accelerated process at the fibrillization stage (Fig.5b).This suggests that the crossseeding of Aβ and hIAPP is less efficient than homologous seeding of either Aβ or hIAPP alone,but such cross-seeding does not necessarily prevent either Aβ or hIAPP aggregation [30].The coincubation of equimolar Aβ40and hIAPP monomers enabled to form hybrid amyloid aggregates ranging from soluble oligomers,protofibrils,and fibrils,but in a less efficient way,as evidenced by the prolonged lag phase,slowly increased growth phase,and the reduced total amount of hybrid fibrils (Fig.5c),as compared to homogenous aggregation of Aβ or hIAPP alone [30].Differently,anotherin vitrostudy showed that co-incubating Aβ42monomers with hIAPP monomers at different molar ratios led to a marked increase in Aβ42aggregation as compared to Aβ42aggregation alone [33](Fig.5d).All of thesein vitrocross-seeding of Aβ40/42and hIAPP37monomers underwent conformational transition from random structures to α-helix to β-sheet [3,32,33].Taken together,while amyloid cross-seeding indicates the intermolecular interactions between different amyloid proteins,such interactions could be either competitive to promote cross-seeding or cooperative to delay cross-seeding,thus leading to a more complex scenario of the cross-seeding-induced acceleration and inhibition of amyloid formation.In a broader view,any difference in aggregation kinetics between amyloid cross-seeding and homo-seeding reflects the existence of cross-seeding barriers,which depend on structural compatibility between conformations of two co-existing sequences in the ensemble.Any subtle changes in structural compatibility will dramatically determine amyloid cross-seeding fate.Similarly,controversial results regarding the toxicity of Aβ-hIAPP assemblies were also reported,i.e.,Aβ-hIAPP assemblies were found to promote [33]or inhibit [30]cell toxicity as compared to pure Aβ-or hIAPP-induced cell toxicity.

Amyloid cross-seeding between Aβ and hIAPP also occurred at both raft-like and isolated β-cell lipid membranes.It was reported that Aβ and hIAPP can consistently co-aggregate into form hybrid fibrils on different cell membranes.Interestingly,the resultant Aβ-hIAPP fibrils exhibited similar structural morphologies to pure hIAPP fibrils but were quite different from Aβ counterpart.Meanwhile,aggregation kinetics of Aβ-hIAPP cross-seeding was slower than that of pure hIAPP,but faster than that of pure Aβ[32].Consistently,Aβ-hIAPP aggregates adsorbed on and penetrated into isolated β-cell membranes slower than pure hIAPP,but faster than pure Aβ[82].The cross-seeding and their effects on cell membranes are very important for better understanding the pathology of each PMD and pathological links between different PMDs.While (1)extensive studies have revealed different amyloid-induced membrane disruption mechanisms,including ion-permeable,transmembrane pores,membrane thinning and softening,carpet models,and detergent models and(2)several studies have showed that amyloid cross-seeding causes the increase of membrane leakage and cell toxicity as compared to pure amyloid aggregates,there are lack of direct experimental evidence to examine the interactions between amyloid cross-seeding species and cell membranes.It remains a great challenge to isolate amyloid cross-seeding species from their co-incubated pure amyloid species,thus it is unclear to truly identify a role of amyloid cross-seeding in membrane disruption by excluding the influence from pure amyloid species.It is highly possible for amyloid cross-seeding species to adopt similar membrane disruption mechanisms to pure amyloid species and antimicrobial peptides.Amyloid cross-seeding species could (i)penetrate into cell membranes to form transmembrane,ion leakage pores,(ii) adsorb onto cell membranes to induce membrane thinning and curvature,(iii) bind to membrane receptors to interfere with synaptic functions of cells,and (iv) grow on cell membranes to cause detergent-like membrane dissolution[83–86].

Beyond the typical amyloid cross-seeding of Aβ and hIAPP as described above,a few interesting studies were also reported to study Aβ-hIAPP amyloid cross-seeding by implementing different strategies or techniques.First,several amyloid protein fragments derived from full-length Aβ and hIAPP,including Aβ19-22,Aβ27-32,Aβ35-40,hIAPP8-18,and hIAPP22-28,were found to cross-seed with both Aβ and hIAPP [31].This finding demonstrates a new strategy to identify hotspot sequences involving in amyloid cross-seeding.Second,different from the structural characterization of large AβhIAPP assemblies at macroscopic or low-resolution levels,electrospray ionization-ion mobility spectrometry-mass spectrometry(ESI-IMS-MS) was used to identify the formation of small AβhIAPP oligomers from dimer to hexamer.The results demonstrated that Aβ-hIAPP oligomers were different from either Aβ or hIAPP oligomers in isolated culture.Third,Surface-enhanced Raman spectroscopy (SERS) was used to detect the secondary structures of hIAPP-Aβ oligomers at very low concentration of nM.SERS results showed that hIAPP-Aβ oligomers exhibited an unique core–shell structure,with a hIAPP-rich core and a Aβ-rich shell [87].Finally,uncovering the key residues and fragmental sequences of both Aβ and hIAPP is of high importance for the better understanding of amyloid cross-interaction and the development of possible intervention strategies and therapeutics against both diseases.Using of a systematic alanine-scanning mutation,several hotspot segments including F15,L16,F23and I26from hIAPP were determined and mutation of these residues by Ala caused a significant loss of binding affinity between Aβ and hIAPP and a structural destabilization of Aβ-hIAPP assemblies [88].

3.3.Cross-seeding of Aβ and hIAPP in silico

Despite experimental studies of Aβ-hIAPP cross-seeding,lack of atomic-structures of Aβ-hIAPP cross-seeding pose a great challenge to derive and understand the molecular mechanism of Aβ-hIAPP cross-seeding.Classical molecular dynamics (MD) simulation is a powerful tool to gain a better understanding of the molecularlevel details associated with amyloid cross-seeding,but it remains infeasible or extremely difficulty to direct simulate the whole process of amyloid cross-seeding due to its short timescale of nanoseconds-microseconds and small lengthscale of nanometersmicrometers.To circumvent this challenge,some special techniques including coarse-grained models and enhanced sampling have been proposed to study either very small cross-seeding oligomers at a very early stage or mature cross-seeding fibrils at a final equilibrium stage,both of which still offer critical missing pieces of Aβ-hIAPP cross-seeding at atomic details.

Surprisingly,there are no any simulation reports on the crossseeding of Aβ and hIAPP monomers to mimic the early coaggregation of both peptides,largely because of lack of an efficient sampling method to capture the essential aggregation events between disordered Aβ and hIAPP monomers in a huge conformational space.Instead,several MD simulations from different labs have been conducted to study the stability and dynamics of the preformed Aβ-hIAPP cross-seeds at different oligomeric sizes.All of these computational studies used the fibrillar models to represent the U-shaped β-strand–turn -β-strand structures of Aβ and hIAPP oligomers,then explored different β-structure association patterns for possible Aβ-hIAPP cross-seeding,and finally determined the most structurally stable and energetical favorable AβhIAPP cross-seeds.Specifically,all-atom MD simulations have shown that Aβ15-40and hIAPP10-35octamers favored to associate together by aligning their similar U-shaped conformations along a fibrillar axis(Fig.6a)[34].Later,a more advanced,efficient computational strategy has been developed by a combination of coarse-grained(CG)replica-exchange molecular dynamics(REMD)and all-atom molecular dynamics (MD) simulations to explore a much larger sampling space for Aβ-hIAPP cross-seeding[4,5].Four different cross-seeding modes between Aβ17-42and hIAPP1-37pentamers,including double-layer,elongation,tail–tail,and block modes (Fig.6b),were identified,which represent not only a general polymorphic nature of amyloid cross-seeding,but also different cross-seeding pathways to form different cross-seeds.Among them,the most populated double-layer model showed that Aβ and hIAPP pentamers were laterally stacked on the top of each other to form an interface between the N-terminal β-sheet of Aβ and C-terminal β-sheet of hIAPP.The Aβ17-42-hIAPP1-37elongation model,as the second populated cross-seeding assembly,was similar to the Aβ15-40-hIAPP10-35cross-seed [34],where both Aβ and hIAPP oligomers were associated together via stable end-to-end interactions,presenting the addition of the same or different peptides to the edge of existing amyloid aggregates.Different from the above-mentioned computational studies that did not use fulllength Aβ or hIAPP peptides for cross-seeding,another MD simulations used full-length Aβ1-42and hIAPP1-37to study their crossseeding behaviors [35].It was found that both Aβ1-42and hIAPP1-37hexamers can associate with each other in an elongation way to form single-layer cross-seeds and in an lateral stacking way to form double-layer cross-seeds (Fig.6c).

Different from the abovementioned MD simulations of AβhIAPP cross-seeding in bulk solution,MD simulations of the cross-seeding of Aβ17-42and hIAPP1-37on lipid membranes presented more biological impacts on possible cross-seedinginduced cell membrane disruption linked to the cell toxicity of each disease and the pathological link between AD and T2D.Different double-layer Aβ-hIAPP cross-seeds,obtained from previous works [4,5],were also found to be adsorbed on both zwitterionic POPC and anionic POPC/POPG mixed bilayers via a specific orientation of the N-terminal strands of Aβ in contact with lipid bilayer[6].This orientation with hIAPP being exposed to the external environment explains the experimental observation that the Aβ-hIAPP cross-seeds has similar topologies to the pure hIAPP aggregates,not Aβ aggregates.Aβ-hIAPP cross-seeds interacted more strongly with and located more closely to anionic POPC/POPG lipids than zwitterionic POPC lipids due to the presence of Ca2+bridges to connect both negatively charged groups of Glu residues from hIAPP andfrom lipids (Fig.6d).Ca2+bridges may also provide additional explanation to Ca2+homeostasis responsible for neuronal dysfunction and death.

Fig.6.Different molecular structures of amyloid cross-seeding between(a)Aβ15-40and hIAPP10-35octamers in a elongation mode (Reprinted with permission [34],Copyright 2013 American Chemical Society),(b)Aβ17-42and hIAPP1-37pentamers in four different double-layer,elongation,tail–tail,and block mode (Reprinted with permission [4],Copyright 2015 American Chemical Society),(c) Aβ1-42and hIAPP1-37hexamers in elongation and double-layer mode(Reprinted with permission[35],Copyright 2016 Royal Society of Chemistry),and (d) Aβ17-42and hIAPP1-37pentamers in a double-layer mode on a POPC/POPG bilayer with a Glu--Ca2+-salt bridge (Reprinted with permission [6],Copyright 2016 American Chemical Society).

Taken together,these MD simulations used different force fields(Gromacsvs.CHARMM),oligomer sizes (pentamer,hexamer,ocatomer),sequence lengths (Aβ15-40,Aβ17-42,Aβ1-42and hIAPP10-35,hIAPP1-37),working environments (in solutionvs.lipid bilayers)the resultant Aβ-hIAPP cross-seeds,on one hand,adopt a wide range of polymorphic structuresviadifferent combinations of βsheet associations and orientations.On the other hand,different Aβ-hIAPP cross-seeds associate Aβ and hIAPP oligomers together in a very similar wayvia(i)a elongation mode by adding peptides along the fibril axis and (ii)a lateral model by stacking two different β-sheets on the top of each other [89–94].Such similar crossseeding organizations are likely attributed to similar U-shaped βstrand–turn-β-strand core structures in both Aβ and hIAPP,which again highlights the importance of structural compatibility particularly β-structure in amyloid cross-seeding.Highly populated βsheet conformation,irrespective of sequences,allows to lower the cross-species barriers and promote mutual binding and recognition between different species,possibly facilitating amyloid cross-seeding via different complex interaction pathways [13].

4.Conclusions and perspectives

Amyloid cross-seeding between different amyloid proteins is considered as a main mechanism for the spread of overlapping pathologies and symptoms across cells and tissues in the same individual with different PMDs.While this review summarizes the cross-seeding of Aβ and hIAPPin vitro,in vivo,andin silico,below we strive to offer some personal opinions about amyloid cross-seeding that could be generally applied to other amyloid proteins.

Conceptually,amyloid cross-seeding strongly depends on their conformational compatibility between different amyloid proteins,which create different cross-seeding barriers to determine the fate of cross-seeding.From a structure viewpoint,a common β-structure as exclusively existed in different amyloid proteins serve as a basic structural motif and a determinant factor for amyloid cross-seeding,i.e.,different amyloid aggregates containing highly β-sheet-rich structures initiate “l(fā)ike-interact-like”co-aggregation via β-sheet-to-β-sheet interactions.Similar β-sheet interactions between different amyloid aggregates lower the energy barriers for co-aggregation pathways including template-assisted co-aggregation and conformational selection co-aggregation pathways.Different β-structure-based motifs(e.g.steric zipper,β-strand-loop-β-strand) resembles those of shorter peptides,illustrating the advantage of the tight packing.The combination of different protein sequences in the cross-β structure explains the polymorphism of amyloid cross-seeding.

From a protein–protein interaction viewpoint,structuraldependent interactions between different amyloid aggregates does not necessarily indicate the acceleration of amyloid cross-seeding,but it could also lead to the inhibition of amyloid cross-seeding.Such intermolecular interactions between different amyloid proteins could be either competitive to promote cross-seeding or cooperative to delay cross-seeding,and subtle changes in intermolecular interactions will cause significant differences in structural conversions and mismatches between different amyloid aggregates,leading to a more complex or even contrasting scenario of amyloid cross-seeding.Amyloid cross-seeding interactions are also supported by the bidirectional or unidirectional co-aggregation scenarios.More importantly,it is also worth to hypothesize that the β-sheet structures could be transmissible amyloid species,which travel between different cells and tissues via cerebrospinal fluids and blood vessels to induce the overlapping amyloidogenesis of different PMDs and to accelerate the progress of PMDs.

However,challenges still remain.Highly diverse aggregate structures and complex aggregation pathways in different healthcare conditions make amyloid cross-seeding studies extremely difficult.This again requires on-going efforts to address many fundamental issues for amyloid cross-seeding,e.g.,If crossspecies barriers exist,what are sequence and structural characteristics critical for amyloid cross-seeding?Why subtle structural differences can lead to different outcomes? What are the driving forces behind the occurrence of amyloid cross-seeding? Whether is amyloid cross-seeding depended more on intrinsic properties of both amyloid proteins (e.g.,sequence or structural similarity)or extrinsic conditions (e.g.protein concentrations,preformed seeds,temperature,pH,or,even agitation),or both?Dose amyloid cross-seeding has a directional effect,i.e.the interaction between different amyloid proteins may work in both directions or in a single direction?Whether or how does amyloid cross-seeding induce cytotoxicity to different hosting cells associated in different PMDs?

To address these fundamental issues,a large scale of research efforts from different interdisciplinary is required to tackle amyloid cross-seeding from different aspects and levels.First,a major challenge is to determine the recognition sequences and structures of amyloid proteins that possibly participate in amyloid crossseeding.“Seeing is understanding”.Due to the highly dynamic nature and strong self-and co-aggregation propensities of any two amyloid proteins,characterization of their hetero-assemblies at an atomic level via NMR or X-ray crystallography has not been yet possible.Recent advances in cryogenic electron microscopy(cryo-EM) hold a great promising to possibly solve the atomicresolution structures of amyloid cross-seeds,without a need for preparing crystalline specimen.Also,most of amyloid crossseeding studies including ours mainly focus on the aggregation of different full-length amyloid proteins.But,given that any amyloid protein and its aggregates always exist as a complex conformational ensemble containing a vast number of polymorphic structures of different sizes,conformations,and morphologies,and populations [66–69],use of full-length amyloid proteins introduces system complexity and uncertainty,making it difficult to identify structural and sequence features for cross-amyloid aggregation and to address many fundamental questions.Thus,it is more feasible and testable to use different amyloid fragments to study amyloid cross-seeding so as to systemically map the key cross-seeding sequences of amyloid proteins.

Second,more advanced and transferable computational models and algorithms,compatible with parallel computing and graphic process units,are highly demanded to simulate an entire amyloid-crossing process from monomers to oligomers,protofibrils,and fibrils.While the timescale and lengthscale appears to not permit such brute-force simulations,use of graph theory and kinetic MD simulations may provide a possible solution to put this puzzle together.Conceptually,a series of cross-seeding aggregates are first generated and used a node in a free energy graph.The relative conformational energy differences between any two nodes(two aggregates)determine their directional aggregation pathways(i.e.graph edge).In this way,it is possible to construct a kinetic graph to represent multiple co-aggregation pathways between different amyloid cross-seeds by assembling a series of MD simulations.Also,computational studies of amyloid cross-seeding in a more complex environment (i.e.cell membranes) is also critically important to illustrate their biological impacts on cell functions.More interestingly,recent advances in data-driven artificial intelligence (AI) models provide a powerful tool to rapidly screen and identify the sequences and structures of amyloid proteins and to predict the cross-seeding possibility of any two amyloid proteins.The AI-driven model can also be iteratively trained and optimized so as to achieve high prediction accuracy.

While amyloid cross-seeding is an exciting area of research,increasing research efforts and funding are very necessary and important to stimulate the studies of amyloid cross-seeding,which would help to decipher a possible molecular cross-talk and disease pathophysiology between and within different PMDs,develop diverse and effective pharmaceutical strategies and drugs to reduce amyloid cross-seeding and treat PMDs simultaneously,and to reduce medical and healthcare burdens globally.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

J.Z.thanks financial supports,in partial,from previous NSF grants (CBET-1510099,DMR-1806138,and CMMI-1825122) to support our amyloid research.