Characterization of insulin cross-seeding: the underlying mechanism reveals seeding and denaturant-induced insulin fibrillation proceeds through structurally similar intermediates

Insulin rapidly fibrillates in the presence of amyloid seeds from different sources. To address its cross-reactivity we chose the seeds of seven model proteins and peptides along with the seeds of insulin itself. Model candidates were selected/designed according to their size, amino acid sequence, and hydrophobicity. We found while some seeds provided catalytic ends for inducing the formation of non-native insulin conformers and increase fibrillation, others attenuated insulin fibrillation kinetics. We also observed competition between the intermediate insulin conformers which formed with urea and amyloid seeds in entering the fibrillogenic pathway. Simultaneous incubation of insulin with urea and amyloid seeds resulted in the formation of nearly similar insulin intermediate conformers which synergistically enhance insulin fibrillation kinetics. Given these results, it is highly likely that, structurally, there is a specific intermediate in different pathways of insulin fibrillation that governs fibrillation kinetics and morphology of the final mature fibril. Overall, this study provides a novel mechanistic insight into insulin fibrillation and gives new information on how seeds of different proteins are capable of altering insulin fibrillation kinetics and morphology. This report, for the first time, tries to answer an important question that why fibrillation of insulin is either accelerated or attenuated in the presence of amyloid fibril seeds from different sources.


Introduction
Fibrillary deposits cause serious problems during commercial insulin production, in implantable insulin pumps and some diabetic patients. [1][2][3] Despite numerous efforts, elucidation of the exact mechanism of protein brillation has eluded the investigators so far. Amyloid brillation is a process in which native soluble protein becomes misfolded into insoluble bril, rich in cross b-sheet structure. [4][5][6] Beginning in the lagphase, partially unfolded proteins convert into oligomers and then protolaments both of which are mostly nondetectable by the current methods. Thereaer, the bril population increases rapidly during the elongation phase. 7 The elongation issue gets more complicated when seeding molecules or foreign surfaces inuence the kinetics of protein brillation. 8,9 In our previous study, it was found that hydrophobicity and electrostatic forces on the external surfaces have distinct effects on insulin brillation. 10 Also, during brillation of amyloid precursor proteins, b2-microglobulin, and insulin the lag-phase is reduced in the presence of preexisting bril seeds, [11][12][13] however, no sufficient description has been presented about elongation kinetics. Although the seeding-or secondary nucleation-induced brillation is an ambiguous phenomenon, it becomes even more complicated when the kinetics of brillation is affected through crossreactivity between seeds originating from a different protein.
Many efforts have been made to elucidate the cross-reactivity between the amyloidogenic peptides and proteins. 11,14 In general, it is accepted that the added seeds will provide catalytic end(s), accelerating the rate of protein brillation. 14,15 By considering this assumption, one may speculate that the kinetic of brillation is always speeded up in the presence of foreign protein seeds. However, it has been recently indicated that the amyloid seed of pro-islet amyloid polypeptide (proIAPP) is capable of inhibiting amyloid formation by asynuclein. 11 This then indicates cross-seeds might inuence the structure of intact monomers of brillating protein.
Generally, on the bases of cases examined yet, the rate of protein bril formation increases upon the addition of its seed (secondary nucleation). 11,14,16,17 A study on cross-interaction between b-amyloid (Ab) and IAPP suggested that identity and similarity between the two protein sequences could affect the rate of their brillation. But it is not always the case. For example, seeds of huntingtin are capable to efficiently brillate the RNA-binding protein (TIA-1) although the two proteins have a low sequence similarity. 9 Similarly, it was reported that brillation of insulin is signicantly reduced in the presence of the seed of its B-chain 16 at neutral pH. Likewise, it has been shown that amyloid brils from different proteins with low sequence similarity exhibited structural differences that control the seeding efficiencies. 18 Yet another study suggested a relationship between the number of amino acids in the monomeric subunit and the size of protein bril. 19 It thus follows that the interactions between soluble proteins and pre-formed seeds at the molecular resolution are complex, elusive, and more studies are needed in understanding this process.
In addition to the efforts made so far in the secondaryseeding/cross-seeding experiments, our current knowledge about insulin brillation also comes from the study of chaotropic additives such as urea. Under monomeric condition and in the presence of low urea concentration, it has been shown that the rate of insulin brillation decreases, while the rate is signicantly faster at the higher levels of this denaturing agent indicating the presence of several intermediate conformers with different susceptibilities toward entering on-or off-brillation pathway. 20 It, therefore, becomes interesting to compare different scenarios. If seeding does affect the kinetics i.e., accelerates or inhibits the rate of insulin brillation then it becomes compelling to explore if seeds induce structural modications in the insulin molecules akin to chatropes and any other similarities in the mechanism thereof. To address these questions and to characterize the cross-seeding effect on the brillation of human recombinant insulin, herein, we used seven model proteins and peptides that vary based on their size, sequence, and hydrophobicity. The model proteins and peptides were human aB-crystallin (aB-Cry), human aAcrystallin (aA-Cry), N-terminal domain of aB-Cry (aB-NTD), a fusion protein of aB-Cry and insulin A-chain (aB-AC), the fusion protein of aB-Cry and insulin B-chain (aB-BC), insulin A-chain, and insulin B-chain. The seeds of intact human insulin served as a control to compare the results. Using several biophysical methods, in vitro, our data suggested that while the seed of aB-AC, insulin A-and B-chain and aB-Cry provided catalytic ends for accelerating insulin brillation, those of aA-Cry, aB-BC and aB-NTD were capable of inhibiting insulin brillation. Investigating the insulin brillation in the presence of insulin seed (secondary nucleation) revealed that the reaction rate was also signicantly augmented. There was a similarity in the structural modication induced in insulin by seed type or denaturant. Overall, this study provides mechanistic insight into insulin brillation and how seeds from different proteins can alter the nature and kinetics of insulin aggregation. These studies will help us better understand the seed-induced insulin brillation process at the molecular level. The results will also help in the better formulation of this antidiabetic hormone.

Preparation of the proteins and peptides
In the current study, the chain combination method was used for producing recombinant human insulin. 21 In brief, the plasmids (pET28b+) containing aB-BC (aB-Cry + B-chain) and aB-AC (aB-Cry + A-chain) fusion genes were constructed. Isopropyl b-D-1-thiogalactopyranoside (IPTG)-mediated expression of the fusion genes were done in Escherichia coli BL21 (DE3) pLysS cells (Invitrogen, Carlsbad, CA). The produced fusion proteins were puried by Ni-NTA (for aB-AC) and diethylaminoethyl (DEAE) anion exchange chromatography (for aB-BC). Owing to the presence of methionine residue between insulin chains and the fusion partner (aB-Cry), CNBr-mediated cleavage strategy was used for detaching the peptides from aB-Cry. For masking thiol groups of cysteine residues located in the chains of insulin sultolysis experiment was done, then due to the signicant difference between sizes of fusion partner and the peptides, G50-Sephadex (Merck, Darmstadt, Germany) gel ltration was performed for separation/purication of the Aand B-chain of human insulin. Aer combination of these two chains (see ref. 21), natively folded insulin was obtained using phenyl Sepharose hydrophobic column (GE Healthcare, Chicago IL). To produce a mutant form of aB-Cry, Quick-change site-directed mutagenesis kit (Stratagene, La Jolla, CA) was used. 22 As a hydrophobic part, the N-terminal domain of aB-Cry (aB-NTD) was obtained by CNBr cleavage at methionine 68 of wild type aB-Cry. Purication of the produced NTD was done by Sephadex G25 (90 Â 1 cm).

Kinetics of monomeric insulin brillation
A solution of acetic acid (20%, pH 2.0) containing 50 mM NaCl in varying concentrations of urea (at 60 C) was selected as the brillogenic condition. 20 The insulin sample (2 mg mL À1 ) was freshly prepared immediately before the experiments using a molar extinction coefficient of 1.08 at 276 nm for 1.0 mg mL À1 . 4 The growth of amyloid brils was monitored by measuring ThT uorescence intensity at 484 nm. To perform this experiment, 10 mM ThT was added to each protein solution and then incubated in the black 96-well plate with the nal volume of 100 mL. The uorescence intensity at 482 nm was plotted against time, and the kinetic proles were analyzed by curve tting using GraphPad Prism V7.0 (GraphPad Soware, Inc La Jolla, California, USA). For each sample, three replicates were measured to reduce well to well variation.

Circular dichroism (CD) assessment
The circular dichroism (CD) spectra were taken on a Jasco J-810 spectropolarimeter using a cell with a 0.1 mm optical path length at 60 C. The measurements were done in acetic acid (20%, pH 2.0) containing 50 mM NaCl and various urea concentrations ranging from 0.5 to 7.0 M. In each experiment, the concentration of insulin was set at 2.0 mg mL À1 . To study the reversibility of the induced secondary structures under brillogenic condition, a serial dilution of urea concentrations was applied. Therefore, the protein solutions were individually dissolved with 7.0-1.0 M urea. Aer 15 min incubation, the solutions were diluted to a nal concentration of 2.0 mg mL À1 for insulin and desired concentration of urea (6.0, 5.0, 4.0, 3.0, 2.5, 2.0, 1.5, 1.0 and 0.5 M). To analyze the a-helix transition during each step, molar ellipticities at 209 nm were plotted versus ascending and descending urea levels.

Fourier transform infrared (FTIR) spectroscopy
At the end of brillation, for analyzing the secondary structure of brils, 100 mL of the aliquots (2.0 mg mL À1 ) were taken aer 6.0 h of the incubation under brillogenic conditions. The samples were then centrifuged, separately. The generated brils were also centrifuged (20 min, 13 000 g at ambient temperature) and washed with D 2 O to exchange hydrogens and buffers by deuterons. 23,24 This procedure was repeated at least three times for each sample. Finally, the brils were spread uniformly on the surface of an attenuated total reectance (ATR) diamond cell. ATR-Fourier transform infrared (ATR-FTIR) spectra were recorded on a Tensor II instrument (Bruker, Germany) from 1700 to 1600 cm À1 (amide I), using a resolution of 4.0 cm À1 and an accumulation of 256 scans. Curve tting of raw data and deconvolution analyses with Gaussian function were carried out by Origin 2018b soware using the positions of ve peaks. These positions have been previously used for the analysis of protein amyloid: 1600-1615 cm À1 (b-sheet, extended chain, side chain), 1615-1630 cm À1 (b-sheet), 1630-1650 cm À1 (a-helix), 1650-1690 cm À1 (turn, disordered), 1690-1700 cm À1 (antiparallel b-sheet) and random coil (1648 cm À1 ). 24 Transmission electron microscopic analyses A 5-fold dilute solution of the brils in each sample was deposited on the formvar carbon-coated copper grids and negatively stained with 1.0% aqueous uranyl acetate. 25 The specimens were examined at 100 kV excitation voltages, with a Philips CM10 transmission electron microscope (Philips, The Netherlands).

Fluorescence microscopy measurements
At the end of brillation, to nd out the ber cross-connections and/or lateral interactions between them (bundles of the brils), the generated brils were mixed with ThT at the nal concentration of 20 mM. Then, 50 mL of the samples were spread homogeneously onto a microscopic slide and incubated for 30 min at a dark place. Using excitation/emission lters (469/ 525 nm), the uorescence images were obtained.

The seeding experiments
In the present work, 2.0 mg mL À1 solution of the brillated proteins and peptides were collected by centrifuging the specimen (13 000g for 20 min at ambient temperature). The pellet of brils was dialyzed in double-distilled water and gently sonicated. 26 Aer that, lyophilizing was done and residual bril samples were weighted. Based on the images taken from the transmission electron microscope, the seeds were short, 40 nm. A 5% (w/w) of the seeds/insulin was prepared to study the kinetics of brillation. 14,16,26 The dened brillogenic condition and the desired urea concentration were considered for the seeding experiments. Monitoring of the brillation was done as described before.

Measuring the protein concentration
The concentrations of recombinant human insulin, insulin Achain, insulin B-chain and the fusion proteins (aB-AC and aB-BC) were estimated based on previous publications. 6,21,25 Also, following previous literature, 27 aA-Cry concentration was determined by UV absorption at 280 nm (an extinction coefficient of 0.72, mL mg À1 cm À1 , for 1.0 mg mL À1 ).

Gel electrophoresis experiment
The purity of the proteins and peptides were analyzed by reducing SDS-PAGE (gel 18%). The gel stained using Coomassie Brilliant Blue (CBB) method. 28

Statistical analyses
The data were analyzed in GraphPad Prism and presented as mean AE S.E.M. The signicance was statistically analyzed by one-way ANOVA using SigmaPlot 12.0 soware. The statistical signicance among the groups was determined using analyses of variance, and P < 0.05 was considered signicant. 21,27

Designing seeding proteins and peptides
In this study, brillation kinetics of human recombinant insulin was examined in the presence of the bril seeds from seven different proteins or peptides and the seed from insulin itself. Scheme 1 summarizes variation in the size, sequence, and hydrophobicity of the selected proteins or peptides.
The scheme exhibits intact human insulin and its A-and Bchain and each was studied separately. Next, aAand aBcrystallins were included in this group. These two chaperone proteins have a mass of $20 kDa, share signicant sequence identity and both possess three distinct structural regions; the hydrophobic N-terminal domain (NTD), a-crystallin domain (ACD) and hydrophilic C-terminal domain (CTD). The purpose of this selection was to investigate the impact, if any, of the sequence variation in aAand aB-crystallins on human insulin brillation during cross nucleation. Additionally, the NTD fragment of human aB-crystallin (aB-NTD) was prepared with the aim to assess how seed of a hydrophobic protein will affect insulin brillation behavior. We also included a variant of human aB-Cry (M68I, P130V) that was individually fused to A-or B-chain of human insulin (aB-AC or aB-BC, respectively). Taken together, we prepared eight proteins and peptides, possessing the molecular masses of 24, 23.3, 20, 7, 5.8, 3.4, and 2.4 kDa, and their purity was checked by SDS-PAGE (Fig. 1). The bril seeds were generated individually from all the above mentioned proteins and hybrid proteins (aB-AC and aB-BC) and were then used in secondary or cross nucleation to assess their impact on insulin brillation. The effect of nucleation on brillation kinetics, morphology, and oligomerization of insulin was characterized and compared.

Primary structure plays an important role in brillation behavior of the model proteins and peptides
We started by studying the brillation of selected proteins by themselves. It has been suggested that protein sequences and hydrophobicity play an important role in the amyloidogenic nature of protein. 19,29 Bis-ANS is a classical probe for the assessment of protein surface hydrophobicity 4,29,30 but due to the controversial results with insulin and it's B-chain, 16 it was not possible to accurately compare the bis-ANS-based measurement of surface hydrophobicity among our different seeding proteins and peptides. Therefore, to have a better picture of surface hydrophobicity of the selected proteins and peptides, they were applied to reverse-phase high-performance liquid chromatography (RP-HPLC) column at ambient temperature ( Fig. 2A). In RP-HPLC, the retention of a protein of interest reects its overall hydrophobicity; the more hydrophobic protein needs a longer time for the elution. 31 As shown in Fig. 2A, A-chain with the least hydrophobicity among the model proteins and peptides was detected to elute as a peak at about 2.1 min, while aB-BC as the most hydrophobic protein eluted at Scheme 1 The schematic illustration of the seed-forming proteins and peptides. Selected eight protein and peptide models on the fibrillation of human insulin. As depicted, these models vary in their molecular masses and the number of residues. Human aA-Cry and aB-Cry share over 50% sequence similarity (multiple sequence alignment was done using Clustal Omega tool), also, both possess three structural regionshydrophobic N-terminal domain (NTD), a-Cry domain (ACD) and C-terminal domain (CTD). Human insulin A-and B-chain were purified by gel filtration chromatography. Human aB-Cry, human aA-Cry, human aB-BC, and human aB-AC were purified as described in published literature. 21 7.2 min. Based on our result, the surface hydrophobicity of these proteins and peptides is in the following order: aB-BC > aA-Cry > aB-AC > aB-Cry > aB-NTD > B-chain > insulin > A-chain. It should be noted that the bril formation by these proteins and peptides was conducted at high temperatures (60 C) so we also characterized surface hydrophobicity of these proteins and peptides at 60 C (Fig. 2B). Interestingly, when they were studied at high temperature, the retention times for insulin, the A-and B-chain did not alter signicantly, suggesting no important changes occurring in their surface hydrophobicity under thermal stress. The signicant delay in the elution of aB-Cry and aB-AC hybrid protein suggests an important increment of their surface hydrophobicity upon thermal stress. The new elution peaks could be due to their intrinsic ability for oligomerization. Likewise, under this condition, the shis of 1.0, 2.8 and 2.0 min were respectively realized for aB-NTD, aA-Cry and aB-BC proteins, suggesting the increment of their solventexposed hydrophobic surfaces or oligomerization state upon thermal stress. To gain an insight into the hydrophobic prole of the samples, the sequence hydrophobicity index of the samples was predicted using the parameters described by Kyte & Doolittle (Fig. 2C). In this type of prediction a value, GRAVY (grand average hydrophobicity), is associated with the hydrophobicity of proteins. 32 High positive values suggest high hydrophobicity, whereas negative values indicate high solubility. Generally, regions in the N-terminal domain (residues 1-68) of aBand aA-crystallins frequently exhibit positive GRAVY index, indicating these segments of the sequence are more hydrophobic than other segments in the proteins. In contrast, the 28 residues of the ACD have negative GRAVY values, indicating these areas are probably not hydrophobic. Remarkably, surface hydrophobicity assessments of A-and B-chain indicate high positive values. Although the theoretical data showed a signicant difference in the hydrophobic nature of the samples, they were very different from the experimental point of view.
To analyze the kinetics of brillation and to prepare the seeds, these proteins and peptides were incubated under acidic conditions (pH 2.0) containing 50 mM NaCl at high temperature for 360 min. During incubation, the progress in bril formation was monitored using a ThT-based assay (Fig. 3).
As shown in Fig. 3, ThT uorescence emission intensities of A-chain and aB-AC increased rapidly, showing a lag time of lower than 140 min (Table 1). A-chain alone and A-chain fused to human aB-Cry revealed differences in the propensity for brillation under thermal stress. Comparing the brillation kinetics data of A-chain, aB-Cry and aB-AC, it can be concluded that insulin A-chain reinforces its strong brillation behavior on the fusion protein, although, A-chain forms about one-tenth of the primary structure of the fusion protein (aB-AC). The brillation propensity of the intact insulin molecules (lag time 185 min) was ranked signicantly lower than A-chain alone but higher than B-chain by itself. Judging from these kinetic studies Fig. 2 RP-HPLC analysis of the seed forming proteins and peptides at 25 C and 60 C. A 20 mL of each protein and peptides (2.0 mg mL À1 ) in the starting buffer (24% acetonitrile in water) was subjected to the column and run gradient over 15 min (1.0 mL min À1 ) using 24-60% acetonitrile in water. The hydrophobicity was analyzed by the Knauer HPLC system. The absorbance signals were recorded at 214 nm, using DAD 2.1 UV-Visible detector (Knauer, Germany). The experiments were conducted at 25 C (A) and 60 C (B). The hydrophobicity index versus amino acid residues is shown for insulin A-chain, insulin B-chain, human aA-Cry and human aB-Cry (C). Monitoring the kinetics of protein and peptide models fibril formation by ThT fluorescence assessment. The fibril formation by different proteins and peptides was monitored by ThT fluorescence intensity. The concentration of each sample was fixed at 2.0 mg mL À1 with 10 mM ThT in 20% acetic acid at pH 2.0, containing 50 mM NaCl. The samples were incubated at 60 C for 360 min to induce fibril formation. Also, at the desired time intervals, the analyses were conducted. The excitation wavelength of the protein/peptide samples was fixed at 450 nm, while the emission was recorded at 484 nm. The results were average of three independent experiments.
This journal is © The Royal Society of Chemistry 2020 RSC Adv., 2020, 10, 29885-29899 | 29889 of ThT uorescence assessment under thermal stress aB-NTD reveals signicantly higher aggregation propensity than its parent molecule. The recombinant lens proteins aA-Cry and aB-Cry exhibited the least propensity for aggregation under thermal stress (Table 1). According to the data presented in Fig. 2B and 3, it is evident that poor aggregation of aA-Cry does not relate to the solvent-exposed hydrophobic surface indicated by this protein under similar thermal stress, suggesting the contribution of additional factors in brillation process of aA-Cry.
Next, the morphology and structure of the aggregates/brils were characterized by TEM imaging and FTIR spectroscopy (Fig. 4).
As shown in Fig. 4 (top panel), intact insulin, insulin Bchain, human aB-Cry and human aA-Cry formed the long straight brils while insulin A-chain, aB-AC, aB-BC and aB-NTD produced short brils that clumped with each other. The width of brils in all these conditions was between 20-45 nm. This morphology is similar to the amyloid or brillar aggregates reported for other proteins. 25,33,34 In FTIR spectroscopy, the amide I region (1600-1700 cm À1 ) is obtained by backbone C]O and C-N stretching/vibration of the proteins and gives information about the secondary structure elements of the proteins. 23,35 FTIR spectroscopy of the generated brils shows that brillation of all samples is accompanied by an increase in the b-sheet structure as indicated by a strong peak at 1628 cm À1 (ref. 36) in each sample. The other structural elements (e.g., turns 1680 cm À1 and helical structure 1658 cm À1 were observed to populate in varying extents in each case). Interestingly, in the brils of aB-NTD, a peptide composed of 68 residues, there was no indication of the turn structures (absence of a peak at 1680 cm À1 ). 36,37 Taken together, the ThT, TEM and FTIR data establish without doubt the production of brils from all the selected proteins under these conditions.
Mature brils of the investigated proteins and peptides were broken down by ultrasound waves using previously dened setup 26 and TEM images of resulting mixtures were recorded (Fig. 5). These images show a uniform eld of brillar aggregates in each picture. The average length of all seeds was approximately 50 nm. All of these seeds were used in the seeding and cross-seeding experiments as mentioned below.
Only particular amyloid seeds are capable of enhancing the rate of insulin brillation There are abundant shreds of evidence in the literature indicating enhancement of protein brillation is generally induced through the generation of partially folded intermediates by environmental stresses. 38 Given this assumption, it will be important to know if the seeds from different proteins and peptides enhance brillation especially in cross seeding cases and also if seeds induce any structural changes in monomeric insulin to catalyze the bril conversion. Therefore, for reference, we used urea to induce structural alteration of human insulin and subsequently studied brillation kinetics of insulin in the presence of both an increasing concentration of urea (Fig. 6A) and in seeding/cross-seeding conditions (Fig. 6B).
The lag-phase for insulin in 0 M urea was 144 minutes. As urea concentration was increased stepwise to 1.5 M urea, a remarkable decrease in the lag time of insulin brillation (nucleation) was observed as indicating faster kinetics under these conditions (Table 2).
Between 2.0 to 3.0 M urea the lag time decrease indicated an increase in brillation rate under these conditions. Above 3.0 M urea, the brillation rate of insulin again increased indicating a slowdown in the bril formation under these conditions. Based on the results it appears that insulin in the presence of 0-2 M, 2-3 M, and >3 M urea adopts structurally distinct conformations, leading to a non-uniform propensity toward brillation. These experiments were followed by investigating the brillation of monomeric insulin in the presence of the protein and peptide seeds (Fig. 6B). The control of the human insulin monomer (symbol Fig. 6B) exhibited a lag time of 144 minutes. Secondary nucleation of insulin by the seed of insulin itself totally diminished the lag time of insulin brillation (red circle symbol Fig. 6B). Interestingly, the kinetics of brillation increased signicantly by cross seeding of insulin itself by the seeds of aB-AC, aB-Cry, A-chain and B-chain although to different extents. On the other hand, seeds of aB-BC, aA-Cry and aB-NTD reduced brillation kinetics of human insulin.
More details on the bril morphology of human insulin under these stress conditions were inspected by TEM and uorescence microscopy. Compared to insulin by itself, TEM images (Fig. 7, top panel) indicated brils that are generated in 2.0-3.0 M urea have straight and long morphology while the insulin brils generated in the 0.5-2.0 M urea revealed short and disordered and at times diffused shape. Further, by increasing urea concentration up to 4.0 M, the amorphous aggregates were observed along with minor brillar amounts. Highly aggregated monomeric insulin was seen in the presence of 6.0 and 7.0 M urea. Also, the zoom out imaging analyses of the insulin bril populations by the microscopic uorescence agreeably indicated comparable results (Fig. 7, down panel). Although the interaction between ThT molecules and amyloid b-sheets excites uorescence property of the dye, some other studies have indicated the interactions between ThT dye and large amorphous aggregates by stacking in between the aggregates, leading to increased uorescence. 39 Likewise, the insulin brils made in the presence of the seeds were also examined by TEM and uorescence microscopy. Judging from the results presented in Fig. 8, top panel, brils obtained in the presence of seeds of aB-AC, insulin, aB-Cry, A-chain, and aB-BC, possessed long, disperse and distinguishable morphology. However, seeds of insulin B-chain, aA-Cry and aB-NTD were observed to induce the formation of short and clumped brils of human insulin. Fluorescence microscopic  images displayed comparable results, indicating separate and more notable brils of human insulin obtained from the seeds of aB-AC, insulin, aB-Cry, A-chain, and aB-BC in contrast to the short and clumped brils in other cases. Overall, our results suggested a specicity in bril induction, while seeds of some proteins and peptides enhance the rate of insulin brillation others do not. Comparing Fig. 5 and 8, no evidence of residual seeds is apparent in the mixtures of the mature brils, indicating the investigated seeds were consumed in the brillation process.
Also, our data from urea-induced insulin brillation suggested that the kinetics of bril formation in 2.0-3.0 M urea were faster than the kinetics of brillation in the other concentrations of this chaotropic agent (Fig. 6). Insulin appears to be in the brillation prone conformation in this range. In contrast, at concentration 7.0 M urea, insulin conformation is not very susceptible to amyloid brils ( Fig. 6A and 7). Therefore, we wanted to measure the synergistic effects of the seeds and urea on the kinetics of insulin brillation under these two different scenarios. Seeds (5%) from different proteins were added to the insulin in the presence of 2.5 or 7.0 M urea and brillation monitored. The results are summarized in Fig. 9.
When compared to brillation of insulin by itself at 2.0 M urea concentration (lag time 110 min, Fig. 6A), in the copresence of 2.5 M urea and seeds from different proteins, a signicant enhancement (but to different degrees) in brillation kinetics was observed in all cases except for aB-NTD seeds (Fig. 9A). A similar enhancing effect was observed in the co-presence of 7 M urea and seeds from different proteins (Fig. 9B) when compared to 7.0 M urea (lag time 154 min, Fig. 6A) only. Analyzing Fig. 9A and B, we see the overall acceleration of brillation kinetics in the co-presence of urea and seeds but individual seeds did not exhibit the exactly comparable effect. These data indicate that urea at both concentrations alters insulin conformation so that it is highly prone to association with the seeds. We need to recall here that concerning control 0 M urea, urea at 2.0 M is accelerating and at 7.0 M is retarding the brillation kinetics of insulin but the presence of seeds always lead to increased brillation rate. These ndings indicated that seeds might be capable to induce structural modifying effect on insulin under these conditions.
To realize the seed effect on insulin conformation, the monomeric insulin was pre-incubated with 2.5 or 7.0 M urea for 120 min, respectively. Based on the data in Fig. 9A, the mentioned time is sufficient for 2.5 and 7.0 M urea to induce partially unfolded conformations of human insulin. 40 The seeds were added to the mixtures aer the incubation and insulin bril formation was monitored by ThT uorescence intensity ( Fig. 9C and D). For the set of samples where insulin was preincubated with 2.5 M urea (Fig. 9C), adding the seeds increased the rate of brillation except for B-chain seeds that exhibited a reduced rate. Interestingly, the set of samples where insulin pre-incubated with 7.0 M urea a dramatic effect was observed. Each seed had a conspicuous observable effect on the insulin brillation kinetics. While seeds form B-chain and aA-Cry reduced the brillation compared to insulin at 7.0 M urea (Fig. 6A), aB-AC, insulin, aB-Cry, A-chain, and aB-Cry exhibited brillation accelerating effect in decreasing order, respectively. aB-BC and aB-NTD showed similar kinetic in these stepwise sequential incubation experiments as compared to 7.0 M urea only. Taken together Fig. 9 thus establishes that seeds can induce conformational modication of insulin. In other words, when starting the conformation of insulin is more constrained as in Fig. 9D and aer prolonged incubation in urea, the kinetic rate of brillation is slow making it difficult for seeds to recruit the conformationally relevant monomers onto their growing  changes can be categorized in three phases; a structure inducing phase from 0.5-1.5 M, an unfolding phase (associated with oligomerization) from 2.0-7 M and a transition phase from 1.5 to 2 M urea. Next, we investigated whether the structural alterations of human insulin at the lower urea concentration are reversible. To do this, a xed concentration of human insulin was incubated in different concentration of urea and then dilution to lower urea concentration was done in such a way that insulin concentration remains xed at 2 mg mL À1 (Fig. 10B). As indicated, when urea was diluted from 7.0 to 4.0 M, no signicant change in the unfolded insulin population was observed. However, reducing the urea levels in the range of 3.0 to 2.0 M, a remarkable recovery effect on the secondary structures of human insulin was observed. This nding indicates the dynamic and reversible uctuation of the generated insulin intermediates in this range of urea concentrations. Finally, urea concentrations between 2.0-0.5 M did not have a signicant effect on the secondary structure of insulin. These data when compared to brillation kinetics in Fig. 6 conclusively establish the contribution of the structural aspect of protein molecule in its brillation kinetics.
To study the seed-induced structural alteration of human insulin, the far-UV CD spectra of insulin were recorded in the presence of different seeds. Insulin was pre-incubated for 15 min in the presence of corresponding seeds at pH 2.0 (20% acetic acid, 60 C) and then the CD data was collected. Insulin concentration was 2 mg mL À1 and the seed concentration was calculated to be 0.1 mg mL À1 . Besides, the CD signal from each seed was subtracted from the nal insulin CD spectra. Each CD prole in Fig. 11, therefore, represents the structure of insulin only.
As shown in Fig. 11, co-incubation of insulin by aB-NTD, Bchain, aB-BC and aA-Cry induced the structural transitions in human insulin, generating conformations which were similar to those changes observed in urea concentrations ranged from 4.0 to 7.0 M. Interestingly, the structural transitions induced by amyloid seeds of insulin, the A-chain, aB-AC and aB-Cry were comparable to those induced by 2.0-3.0 M urea. Although the dichotomy in this cohort of seeds in terms of structural modi-cations was evident in CD analysis their effect on brillation kinetics was much disperse.
If we summarily arrange the effect of various conditions on the T 1/2 of the brillation in each condition (Fig. 11) we can perform a comparative analysis of the addition of each seed on insulin brillation. The analysis shows that insulin and aB-AC seeds are brillation inducers in all the cases whereas B-chain and aB-BC seeds either have no effect or decrease the brillation rate. The effect of A-chain and aA-Cry seed was observed to be indiscriminate. A-chain exhibited most random effect and varied from extreme inducer (7.0 M urea) to extreme inhibitor (2.5 M urea + seed) of the brillation and everything in between. aB-NTD showed an interesting overall effect. As seed itself or along with low concentrations of urea, aB-NTD inhibited the brillation of insulin as opposed to when the aB-NTD seeds along with urea or at high concentrations of urea only where aB-NTD induced the brillation. Fig. 11 also shows the order of proteins based on T 1/2 values is different in the presence of seeds, urea, or in the absence of either indicating toward the contribution of conformational dispersion in different situations. It can also be observed that brillation pattern in the presence of seed and 2.5 M urea are signicantly similar indicating convergence of underlying conformational states prone to brillation under these two different conditions. The summary suggests non-universality of seeding effect on brillation and the process is too complex to be categorized into an exclusively effect caused by hydrophobicity, size or origin of the seeds.

Discussion
Accumulating evidence support cross-seeding of amyloid brillation. For example, the seeds of Ab40, Ab42, and aS have been reported to enhance the aggregation within themselves and among each other. 41 Ren et al. 42 have populated a list of different proteins and their counterparts exhibiting the crossseeding behavior. Cross seeding has been termed "bidirectional" when two seeds from two proteins trigger each other's brillation (as in ref. 41); "unidirectional" (for example when Ab seeds promoted hIAPP aggregation but hIAPP seeds did not for Ab 18 ); and "inhibitory" (when seeds from hIAPP derived peptides slowed down the brillation of insulin 11,43 ). Further, two mechanisms have been proposed for cross-seeding; the 'template-assisted' 15,44 and 'conformation selection and population shi' mechanism. Cross-seeding has been suggested to exist in human diseases. For example, the association of Ab and aSyn has been known for Lewy body dementia but relevant to our study in the pancreatic islet amyloid deposits of type-II diabetes patients Ab, tau and hIAPP were found co-localized. 45 Given the above background, our studies provide different aspects of the brillation and cross-seeding induced brillation phenomenon. For example, based on several assumptions a popular model for amyloid brillation is based on the contribution of hydrophobic interaction among the partially unfolded states of the amyloidogenic proteins. 29,30,34 Our results (Fig. 6-10) suggest this might not be the case and it is likely factors other than hydrophobicity may play an important role in altering the protein brillation kinetics. In this case, the human aA-Cry and aB-Cry studied here provide suitable examples. Although these proteins revealed a signicant amount of solvent-exposed hydrophobic surfaces, they indicated a poor Fig. 9 Kinetics of insulin fibrillation at pH 2.0 monitoring by ThT fluorescence assay. Co-incubation of the monomeric insulin (2 mg mL À1 ) with the seeds of different proteins and peptides (5% W/W) at 2.5 M (A) and 7 M urea (B). Also, before adding the seeds, insulin was pre-incubated at 2.5 M (C) and 7 M (D) urea and after that, the seeds were added to the mixtures.
propensity for the thermal-induced brillation compared to the other proteins and peptides ( Fig. 2 and 3). Specically, these proteins indicated more hydrophobicity than insulin A-and Bchain at 25 C and 60 C yet their brillation lag time was meaningfully higher than that of insulin chains. Another aspect is the contribution of size. Our results in Fig. 1 and 3, suggested that the size of the proteins and peptides was not the determining factor associated with their susceptibility for brillation under these conditions. For example, the size of fusion proteins aB-AC (25.7 kDa) and aB-BC (26.7 kDa) is similar (4% difference) but the differences in the brillation propensity obtained from lag times (Table 1) are signicant. aB-AC exhibited 30% faster brillation kinetics than aB-BC. While aB-AC indicated only 3% faster brillation compared to A-chain alone; aB-BC exhibited 15% faster brillation than B-chain alone. Between themselves, A-chain showed 42% faster kinetics than B-chain alone (at 30% size difference). Moreover, in the comparative seeding effect (Fig. 12) aB-AC exhibited increased brillation induction under all the cross-seeding conditions studied here but aB-BC was placed at the bottom with low cross-seeding brillation induction capacity. Our results indicated that size of the proteins has no signicant effect on the kinetic of crossseeding brillation. Perhaps, the primary structure can be considered to have the most remarkable inuence on the brillation behavior of these proteins and peptides in conjunction with some previous studies. 5,13,19 The marked increase in aB-fusion proteins compared to aB-Cry itself appears to t this assumption well. Each constituent of the fusion proteins converts to brils individually so it cannot be ascertained from this data if the more aggressive aggregating constituent solely drives the faster kinetics of fusion protein or there is a more intricate mechanism involved.
Here we also want to clarify some choices that had to be adopted for technical reasons and feasibility of reactions. In the fusion forms of aB-Cry, the M68I mutation was made to prevent the formation of additional unwanted cleavage products in the presence of CNBr. The Asp-Pro sequence (residues 129 and 130) in aB-Cry is sensitive to the cleavage by formic acid and this reagent is needed for releasing of the insulin chains from the fusion proteins in the cleavage step. Therefore, Pro 130 was mutated to valine. Also, fusion forms of aB-Cry were used rather than aA-Cry because of the following reasons; the A-and B-chain of insulin has 3 and 2 cysteine residues, respectively. The aA-Cry has 2 cysteines in its sequence (residues 131 and 142). Due to the lack of cysteine in the sequence of aB-Cry it was considered a better choice for constructing the fusion protein to eliminate the effect of cysteine interaction in the structure. In the seed of fusion proteins, the accessibility of A-and B-chain for possible interactions with insulin is important.
Focusing on the cross-seeding studies, so far, several efforts have been reported on studying the effect of the seed of one protein on brillation of another protein. 9,16,18,19,42 In this context, we examined the possibility of cross-seeding of insulin amyloid formation with pre-formed amyloid seeds from many different proteins and peptides. We observed that the amyloid seeds of insulin, aB-AC, aB-Cry, A-chain and B-chain were capable of speeding up the kinetics of human insulin brillation (Fig. 9B). Also, our results indicated that seeds from aB-BC, aA-Cry, or aB-NTD signicantly attenuated the brillation rate of human insulin. According to the results of our study, herein, Before the dilution, insulin was incubated for 15 min. Then, upon dilution, the protein was incubated with the lower concentrations of urea for 15 min to equilibrate. The experiments were conducted at 60 C, in 20% acetic acid, pH 2.0. Fig. 11 The seed-induced structural transition of human insulin. The far UV-CD analysis of human insulin (2.0 mg mL À1 ) in the presence of different seeds was done. Given 5% (w/w) seeds to insulin concentration, the final concentration of the seeds was 0.1 mg mL À1 . Before collecting the spectra, insulin was incubated for 15 min in the presence of each seed. The experiments were conducted in 20% acetic acid containing 50 mM NaCl, 60 C, at pH 2.0.
we propose a new mechanism for the seed-induced insulin brillation. Insulin can be converted to various intermediates in the presence of different seeds. For example, Fig. 11 shows brillation inducing seeds like insulin, aB-AC or aB-Cry all have similar intensity at 209 and 224 nm but the aB-NTD or aA-Cry that have stronger intensities at 224 nm and correspondingly their seeds inhibit insulin brillation. Thus, some of the seeds with high molecular dynamics exhibit more susceptibility to form long and distinguishable brils. Although many different models for cross-seeding have been proposed 14,15,42,46 our studies here provide the pivotal data in support of the templateassisted brillation model of cross-seeding. 42 Further, we have thus different scenarios in the presence and absence of urea to compare. First, as human insulin cross-seeded by the low hydrophobic proteins and peptides, mimicking the narrow range of urea concentrations, the produced insulin conformers will be converted to the long brils. Second, non-brillar insulin aggregates will be formed in the presence of the seeds with high hydrophobic surfaces, mirroring the high urea levels (Scheme 2).
The results in Fig. 6, 7 and 9 suggested that those non-native insulin conformers which were produced in the presence of 2.0-3.0 M urea were much more susceptible to brillation. Besides, these intermediate species can recover their structures by reducing the urea concentrations, suggesting a high molecular dynamic propensity of the species.
Also, urea in the range of 4.0-7.0 M is capable of producing amorphous insulin aggregates (Fig. 7). Again, these data are in agreement with the previous studies 14,15,46 the seeds of one protein can alter the conformational state of other proteins especially the template assembly (TA) models. 15 In these two models, amyloid seeds act as a template to dictate a vulnerable state for soluble/folded proteins to brillate. Retrospectively, our data from RP-HPLC and far UV-CD studies (Fig. 2, 10 and 11) indicated that the seeds with a low degree of hydrophobicity (seeds of insulin, aB-AC, aB-Cry and A-chain) (Fig. 2) were capable to mimic the role of urea at the concentration range of 2-3 M, producing intermediate insulin conformers which were susceptible for the formation of long amyloid brils. However, the seeds of aB-BC, aA-Cry and aB-NTD with a higher level of hydrophobicity induce the entire unfolding of human insulin via hydrophobic interaction, resulting in the formation of amorphous aggregates under brillogenic condition. From these observations, if therefore is most likely that the insulin intermediate species induced by the urea and/or the amyloid seeds have a competitive relationship. Therefore, the core insulin molecule that contributes to bril assembly is similar irrespective of being induced by the cross-seeding or denaturant. Fig. 12 Comparative seeding effect on human insulin. The seeding effect was compared based on T 1/2 value obtained in each condition and arranged in the decreasing order of T 1/2 obtained in each seeding or cross seeding sample. The colors correspond to the seed used.
Scheme 2 A proposed mechanism for the cross-templating insulin fibrillation. Native insulin in the presence of 2-3 M urea forms the intermediate conformers which subsequently produce ordered aggregates (amyloid fibrils), while urea at the higher concentration (4-7 M) induces the amorphous aggregation. The produced seeds with low hydrophobicity, similar to low urea level, are capable of inducing insulin intermediates-conformers that form long and distinguishable fibrils. Also, the seeds with high hydrophobicity, similar to high urea levels, were able to induce the entire unfolding of insulin, producing amorphous aggregates.
This journal is © The Royal Society of Chemistry 2020 RSC Adv., 2020, 10, 29885-29899 | 29897 Last but not least, it is worthwhile considering that the presented brillation condition was conducted at acidic pH (pH 2.0). At this pH, on account of pI of insulin (5.5) positive charges would be expected, 7 likewise other involved in seed samples have positive charges. However, the interaction between insulin and preexisted seeds to form mature insulin bril brings about the importance of hydrophobic interaction during seedmediated bril formation. When it comes to explaining the effects of electrostatic charges during the brillation, relay on our previous study, 10 at physiological pH (7.4) the effect of seeding is expect to be similar as acidic pH due to negative charge of insulin and other seed-generated proteins (theoretically, pIs 6.82 and 5.84 were calculated for aB-Cry and aA-Cry, respectively).
Look over the study, we have focused on insulin which is an unstable protein with a high degree of conformational exibility for extended periods in aqueous buffers. 47,48 As a monomer, in the brillogenic condition (20% acetic acid), insulin may meet Leviathan's paradox, giving it too many accessible conformations to nd a stable fold in a reasonable time frame. 49 Here, through a combination of observations obtained from the seedmediated and urea-induced insulin brillation, we have elucidated the similarities in the mechanism by which mild hydrophobic seeds of proteins and peptides are capable to provide the catalytic ends, inducing the formation of non-native insulin conformers.

Conclusion
Through a combination of urea and seed-induced insulin brillation, we have identied a mechanism that monomeric insulin is efficiently incorporated into mature brils via crosstemplating. These results indicate the core conformation of brillation prone intermediate is similar in either case. We also showed that the seeds of some proteins and peptides are capable to mimic the role of urea in inducing formation of the intermediate insulin conformation with the signicantly increased rate of brillation kinetics. Finally, the seed-mediated insulin brillation conducted here will provide new insight into the specic role of seeds and the impact of cross-seed of different proteins and peptides in insulin brillogenesis.

Conflicts of interest
The authors declare that they have no conicts of interest with the contents of this article.