Accumulation of storage proteins in plant seeds is mediated by amyloid formation

Amyloids are protein aggregates with a highly ordered spatial structure giving them unique physicochemical properties. Different amyloids not only participate in the development of numerous incurable diseases but control vital functions in archaea, bacteria and eukarya. Plants are a poorly studied systematic group in the field of amyloid biology. Amyloid properties have not yet been demonstrated for plant proteins under native conditions in vivo. Here we show that seeds of garden pea Pisum sativum L. contain amyloid-like aggregates of storage proteins, the most abundant one, 7S globulin Vicilin, forms bona fide amyloids in vivo and in vitro. Full-length Vicilin contains 2 evolutionary conserved β-barrel domains, Cupin-1.1 and Cupin-1.2, that self-assemble in vitro into amyloid fibrils with similar physicochemical properties. However, Cupin-1.2 fibrils unlike Cupin-1.1 can seed Vicilin fibrillation. In vivo, Vicilin forms amyloids in the cotyledon cells that bind amyloid-specific dyes and possess resistance to detergents and proteases. The Vicilin amyloid accumulation increases during seed maturation and wanes at germination. Amyloids of Vicilin resist digestion by gastrointestinal enzymes, persist in canned peas, and exhibit toxicity for yeast and mammalian cells. Our finding for the first time reveals involvement of amyloid formation in the accumulation of storage proteins in plant seeds.


INTRODUCTION
Amyloids represent protein aggregates having an unusual structure formed by intermolecular beta-sheets and stabilized by numerous hydrogen bonds 1 . Such a structure called 'cross-β' 2 gives amyloids the morphology of predominantly unbranched fibrils and unique physicochemical properties including (i) resistance to treatment with ionic detergents and proteinases; (ii) binding amyloid-specific dyes like Thioflavin T (ThT); (iii) apple-green birefringence in polarized light upon binding with Congo Red (CR) dye 1,3 .
The biological significance of amyloids is based on two aspects: pathological and functional. Amyloid deposition leads to the development of more than 40 incurable human and animal diseases including various types of amyloidoses and neurodegenerative disorders 4,5 . Nevertheless, amyloids may not be only pathogenic but also functional 6 . A growing number of studies demonstrate that amyloids play vital roles in Archaea 7 , Bacteria and Eukarya including humans 8 . Amyloids of prokaryotes fulfill mostly structural (biofilm and sheaths formation) and storage (toxin accumulation) functions 9 . In fungi, infectious amyloids called prions control heterokaryon incompatibility, multicellularity and drug resistance [10][11][12] . In animals, amyloid formation is important for different functions including the long-term memory potentiation, melanin polymerization, hormone storage, and programmed necrosis 13 . Compared to other groups of organisms, plants remain to be poorly studied in the field of amyloid biology.
However, several plant proteins or their regions were shown to form fibrils with several properties of amyloids in vitro (after the proteolytic digestion or other treatments 14,15 ) suggesting that plants might form bona fide amyloids in vivo 16 .
Previously, we performed a large-scale bioinformatic analysis of potentially amyloidogenic properties of plant proteins including all annotated proteomes of land plant species 17 . This screening demonstrated that seed storage proteins comprising the evolutionary conservative β-barrel domain Cupin-1 were rich in amyloidogenic regions in the majority of analyzed species 17 . Such proteins belonging mainly to 11S and 7S globulins 18 represent key amino acid sources for the growing seedlings, important components of human diet and major allergens 19 . We hypothesize that the amyloid formation could occur at seed maturation to stabilize storage proteins, thus preventing their degradation and misfolding during the seed dormancy. In order to test this hypothesis, we have analyzed whether amyloid proteins are present in seeds of an important agricultural crop and Mendel's genetic model, garden pea Pisum sativum L.

The storage proteins form aggregates resistant to treatment with ionic detergents in pea seeds
Resistance to treatment with ionic detergents is a typical feature of amyloids discriminating them from the majority of other non-amyloid protein complexes 20 . We decided to study whether storage proteins form detergent resistant aggregates in plant seeds. For this purpose, we have used mature seeds of garden pea P. sativum L.
genetic line Sprint-2 collected at 30 days after the pollination when the accumulation of storage proteins is expected to reach its maximum 21 . We have extracted protein complexes resistant to treatment with ionic detergent sodium dodecyl sulfate (SDS, 1 %) from pea seeds using a previously published method called PSIA-LC-MALDI 22 with several modifications (see Materials and Methods). Next, the obtained detergentresistant protein fractions have been solubilized with formic acid and subjected to trypsinolysis followed by the reversed-phase high performance liquid chromatography and mass-spectrometry (see Materials and Methods). As a result, we have identified all three major classes of seed storage proteins (Vicilins, Convicilins and Legumins) and several other proteins including heteropolymeric iron-binding ferritin, biotin-containing protein SBP65 and drought stress response protein dehydrin as detergent resistant components of pea seeds (Table S1). Among identified proteins, 7S storage globulins Vicilins have demonstrated the highest mass-spectrometric scores suggesting for their prevalence in detergent resistant fraction of pea seeds (Table S1). We have selected 47 kDa Vicilin for a detailed analysis of its amyloid properties in vivo and in vitro since its Nand C-terminal regions have been identified in MS/MS analysis indicating that full-length protein participates in the formation of detergent resistant polymers ( Figure S1).

Recombinant Vicilin and its domains, Cupin-1.1 and Cupin-1.2, form detergent-resistant fibrils in vitro
To check ability of Vicilin to aggregate in vitro, we produced the full-length Cterminally 6x-His tagged Vicilin in the E. coli cells, extracted and purified it. Since Vicilin contains two evolutionary conserved Cupin-1 domains called Cupin-1.1 and Cupin-1.2 ( Figure 1a) belonging to the Cupin β-barrel domain superfamily, which have been bioinformatically predicted to be amyloidogenic 17 , we have also analyzed the aggregation of these domains in vitro to estimate their contributions to the aggregation of the full-length Vicilin. We have found that incubation of Vicilin, Cupin-1.1 and Cupin-1.2 proteins in phosphate buffer (pH 7.4) for one day at room temperature (25°C) with the constant stirring has caused the formation of typical amyloid fibrils only in the Cupin- To confirm the assumption of the Vicilin, Cupin-1.1 and Cupin-1.2 fibril formation we have studied the interaction of the obtained aggregates with ThT 23 . A unique feature of this dye is a very weak fluorescence in the free state in an aqueous solution and an intense fluorescence in the bound to fibrils state. Since the samples contained both ones bound to fibrils and free ThT, we have used an equilibrium microdialysis for preparing test solutions and the separation of photophysical characteristics of two different dye fractions 24 . The use of this approach and other special techniques taking into account the contribution of the aggregates light scattering to the recorded absorption spectra 24 and the correction of the recorded fluorescence intensity to the primary inner filter effect 25 , have made it possible to calculate the fluorescence quantum yield of free ThT in the samples (the value of which turned out to be close to 10 -4 , which suits the literature data 26 ) and the dye associated with the studied aggregates ( Figure   1d). Respective molecular weights (kDa) are shown. (d) Fluorescence quantum yield of amyloidspecific probe ThT in a free state in buffer solution (left bars) and bound to Vicilin, Cupin-1.1 and Cupin-1.2 aggregates (right bars) have been determined by using the equilibrium microdialysis for the sample preparation.
We have found that ThT binds to all the tested samples, however, an increase in the fluorescence quantum yield of the bound dye in comparison to that for free ThT varies considerably in different samples (Figure 1d). The greatest increase in the fluorescence quantum yield is observed in the case of 'typical' Cupin-1.2 fibrillar aggregates obtained in the phosphate buffer at 25°C (more than 2 orders of magnitude). At the same time, ThT fluoresces significantly less when it is binding to Vicilin and Cupin-1.1 aggregates obtained in the same conditions ( Figure 1d). It can be assumed that the main fraction of the bound dye in these samples interacts with aggregates (apparently with less compact and ordered compared to fibrillar ones) not specifically, which leads only to an insignificant restriction of the intramolecular mobility of ThT fragments relative to one another, and, therefore, to a small increase in its fluorescence quantum yield. ThT bound to Vicilin, Cupin-1.1 and Cupin-1.  Next, we have stained Vicilin, Cupin-1.1 and Cupin-1.2 aggregates with Congo red dye. When Congo red binds to amyloids it leads to the so-called 'apple-green birefringence' in polarized light 30 . This effect is highly specific and for a long time has considered to be the 'gold standard' in the amyloid diagnostics 4

Fibrillogenesis of the full-length Vicilin can be efficiently seeded by the preincubated Vicilin or Cupin-1.2 fibrils
Characteristic feature of many amyloids is their ability to 'seed' the aggregation of monomers with small quantities of fibrils 33 . The process of such 'seeding' is mostly sequence-specific. Therefore, we have decided to investigate the ability of the Vicilin, Cupin-1.1 and Cupin-1.2 aggregates to induce the fibrillation of the full-length Vicilin.
For this purpose, we have used 'seeds' prepared from pre-incubated Vicilin, Cupin-1.     Table 1). Immunofluorescent microscopy carried on unfixed cryosections has shown an almost total overlapping of the anti-Vicilin antibody and ThT signals indicating the presence of the Vicilin amyloid aggregates (Figure 5a-c). The anti-Vicilin antibody signal has been localized predominantly in the central vacuole and has revealed the presence of little granular compartments that have been proposed to be protein bodies, the membrane compartments where Vicilin typically accumulates 35 . To confirm this observation, we have isolated protein bodies using sucrose cushion sedimentation assay (see Materials and Methods) and have analyzed their staining with anti-Vicilin antibody and ThT. We have found that unfixed, air-dried protein bodies exhibit total overlapping of the anti-Vicilin and ThT signals suggesting that the amyloid Vicilin aggregates are located in protein bodies (laser scanning confocal microscopy, Figure   5d-f, and fluorescent microscopy, Figure S2). Finally, we have stained protein bodies with CR and have demonstrated that protein bodies are CR-positive and exhibit applegreen birefringence confirming their amyloid properties (Figure 5g).
Next, we have studied the accumulation of the amyloid, detergent-resistant Vicilin aggregates in pea seeds using total protein lysates obtained at the different stages of the maturation or germination. Protein lysates were treated with 0.5 % SDS and 0.5 Tween20 for 10 min at RT, boiled for 10 min in the sample buffer containing 2% SDS (final concentration) and subjected to SDS-PAGE followed by western-blot with polyclonal anti-Vicilin antibody. The results of this experiment have confirmed aforementioned proteomic (Table 1) and histological (Figure 5a-g) data and demonstrated that amyloid aggregates of Vicilin are present in pea seeds (Figure 5h).
We have found that these aggregates tend to accumulate during the seed maturation (from 10 to 30 days after pollination) and reach their maximum in mature seeds ( Figure   5h). Notably, these aggregates rapidly disassemble in germinating seeds resulting in the formation of robust proteolytic band (Figure 5h). Since aggregates of Vicilin are highly stable, we have decided to analyze their presence in the commercial canned peas produced by 'Bonduelle' (Bonduelle Group, France) and 'Heinz' (H.J. Heinz, USA).
The results of this experiment have shown that Vicilin amyloids persist canning and retain in these food products (Figure 5h, Figure S3).
We have analyzed the resistance of Vicilin amyloids to treatment with proteases.
For this purpose, we have used in vitro protein digestibility assay (IVPD) that imitates gastrointestinal protein digestion 36 . Total protein lysates have been isolated from pea seeds, sprout cotyledons and canned peas. Next, lysates have been consequently treated with pepsin and pancreatin, boiled in the sample buffer with 2% SDS and subjected to SDS-PAGE and western-blot with polyclonal anti-Vicilin antibody. Results of this experiment have demonstrated that in contrast to monomers, Vicilin amyloids from pea seeds resist proteolytic digestion demonstrating high stability (Figure 5i).
Notably, in vitro obtained Vicilin fibrils were not resistant to pepsin and pancreatin ( Figure S4a); nevertheless, they demonstrated resistance to trypsin treatment ( Figure   S5), prolonged boiling with 2% SDS for 1 h ( Figure S4b) and completely dissolved only by concentrated formic acid ( Figure S6).
Taken together, we have demonstrated that Vicilin amyloids are present in pea seeds in vivo accumulating during the seed maturation, disassembling after germination and retaining in food products like canned peas.

Vicilin amyloid fibrils exhibit toxicity for yeast and mammalian cells
Since we have found Vicilin forms amyloids both in vivo and in vitro, we have decided to analyze the effects of Vicilin amyloid formation. Vicilins are known to have carbohydrate-binding lectin activity resulting in their toxicity for fungi and exhibit functional dualism being not only storage but also pathogen-defense proteins 37 . In this study we have found that Vicilin can form at least two types of amyloid aggregates in vitro: less structured ones that are formed at the initial point of incubation and fibrils (Figures 1-3). To discriminate their effects, we tested toxicity of Vicilin, Cupin-1.1 and

Cupin-1.2 non-fibrillar aggregates and fibrils for yeast cells at the same concentrations.
The data obtained has demonstrated that Vicilin fibrils were very toxic for yeast culture resulting in significant yeast cells growth reduction after 48 h incubation (Figure 6a).
Cupin-1.2 fibrils were extremely toxic resulting in almost complete death of cells ( Figure   6a). In contrast to fibrils, non-fibrillar aggregates of Vicilin and Cupin-1.2 did not exhibit toxicity, while in the case of Cupin-1.1 both non-fibrillar aggregates and fibrils were toxic ( Figure 6.a).
To evaluate potential toxicity of Vicilin amyloids towards mammalian cells, we   Not only storage but also defense from pathogens (primarily, fungi and insects) function is typical for Vicilins 49 and associated with lectin (carbohydrate-binding) properties of these proteins 50,51 and their oligomerization 52,53 . Several lectins are known to form amyloid-like fibrils in vitro affecting their activity 54,55 . We have found out that the formation of amyloid fibrils significantly increases the toxicity of the full-length Vicilin for yeast in comparison with its unstructured aggregates ( Figure 6) presupposing the role of amyloid formation in the defense function of this protein. Another property of lectins, in particular, Vicilins, is their allergenicity. Vicilin is one of the major plant-derived allergens, and its allergenic properties were found to be associated with its proteaseresistant fragments 56,57 . Our finding that Vicilin forms amyloids in vivo explains its protease resistance ( Figure 5) and suggests that these amyloids may represent major source of food allergy according to the data that in vitro generated lectin amyloids are phagocytized by macrophages and elicit an immune response 55 . Vicilin fibrils in high concentrations are toxic for mammalian cells ( Figure 6) but concentrations of Vicilin amyloids in food products are many times lower, thus, their toxic effect for humans seems unlikely. Nevertheless, the presence of Vicilin amyloids in canned peas ( Figure   5) can decrease food quality and increase allergic properties of seeds due to incomplete gastrointestinal digestion. Germination significantly reduces the amounts of amyloids ( Figure 5), therefore, it is recommended to soak the seeds before eating to induce germination. Finally, the creation of novel plant varieties with decreased amyloid formation of storage proteins may represent a promising strategy for future agriculture to improve nutritional value and reduce allergenicity of plant seeds.
Overall, in this study we have identified plant protein that forms amyloids under native conditions in vivo, have found that these amyloids mediate protein storage in plant seeds and have demonstrated dynamics of their accumulation and disassembling, high stability, resistance to proteolytic digestion and canning as well as toxicity for fungal and mammalian cells.

P. sativum L. genetic lines and growing conditions
The pea (P. sativum L.) line Sprint-2 from the collection of ARRIAM (St. Petersburg, Russia) with determinate stem growth and early seed maturation was used 58 . To obtain seeds at various stages of maturation, pea seeds were sown in pots Harvested dry pea seeds were surface-disinfected for 6 min in 98% sulfuric acid, then thoroughly rinsed with sterile water and placed on Petri dishes containing sterile 1% agar-agar. After the seed germinating for 3 days at 27°C in the dark seedlings without signs of microbial contamination were selected and their cotyledons were cut off. Plant seeds and cotyledons collected were immediately frozen in liquid nitrogen and then stored at -80°C.

Microbial strains, plasmids and cultivation conditions
The plasmids for analysis of aggregation of Vicilin, Cupin-1.1 and Cupin-1.2 fused with Yellow fluorescent protein (YFP) in yeast S. cerevisiae were obtained by the insertion of PCR amplified fragments of the corresponding genes amplified with primer pairs containing inserted BamHI (reverse) and HindIII (forward) restriction sites (Table   S2) and pea seed cDNA fragment as template into the pRS315-CUP1-SIS1-YFP plasmid 60 by the BamHI and HindIII sites, respectively. Total RNA from pea seeds was extracted using Trizol (Invitrogen, USA) and cDNA was prepared with SuperScript III reverse transcriptase (Invitrogen, USA). E. coli strain DH5α 61 was used for plasmid amplification. The insertions of the corresponding genes were confirmed by the sequencing by using the CUP1 primer 60 .
To construct plasmids for Vicilin, Cupin-1.1, and Cupin-1.2 export by curlidependent amyloid generator (C-DAG) system the fragments of interest were amplified by PCR using pairs of primers flanked with Not1 (forward) and Sal1 (reverse) restriction sites (Table S2) and pea seed cDNA as template. The PCR products and pExport vector 34 were digested by Not1 and Sal1 restriction enzymes and then ligated. Plasmids encoding yeast Sup35NM (aa 2-253) and Sup35M (aa 125-253) fused with CsgA signal sequence were constructed previously 34 .

C-DAG assay
Analysis of amyloid properties of Vicilin, Cupin-1.1, and Cupin-1.2 with the usage of curli-dependent amyloid generator (C-DAG) system was performed as described earlier 34 . To export proteins on the cell surface E. coli strain VS39 34 was transformed with pExport-based plasmids, encoding Vicilin, Cupin-1.1, and Cupin-1.2 fused with CsgA signal sequence. Export of amyloid-forming Sup35NM and non-amyloidogenic Sup35M proteins was used as positive and negative control of amyloid formation respectively. Birefringence analysis was performed with a usage of Axio Imager A2 transmitted light microscope (Zeiss) equipped with a 40x objective and cross-polarizers.
For transmission electron microscopy analysis incubation on inducing plates without Congo red dye was used.

Mammalian cell lines and toxicity assay
Human DLD1 cells were seeded at 5*10 4 cells per well density on 24-well plates and exposed to protein samples analyzed. At indicated time points, medium was replaced with 0,5 mg/ml MTT solution for 1h at 37°C. Then formazan was dissolved in DMSO, and optical densities were measured at 572 nm wavelength using the Multiscan Ex spectrophotometer (Thermo, USA). Data were presented as the mean of three independent experiments ± the standard error of the mean.

PSIA-LC-MALDI proteomic assay
The PSIA-LC-MALDI proteomic assay for the identification of protein complexes resistant to the treatment with ionic detergents has been published previously 22

Protein expression and purification
To express the Vicilin, Cupin1-1, Cupin1-2 terminally fused with a 6x-His tag, an Alicator kit (Thermo Scientific, USA) was used. The fragments were PCR-amplified using respective primer pairs (Table S2)

In vitro protein fibrillation
For initiation of Vicilin and its domains aggregation in vitro, different buffers and incubation conditions were used. At the first stage the proteins in concentration 0.5 mg/ml were incubated in phosphate buffer (pH 7.4) at room temperature and at 50°C with constant stirring for 2 weeks. Furthermore, the proteins in the same concentration were dissolved in 50% HFIP (Sigma-Aldrich, USA) and incubated for 7 days.
Afterwards, the HFIP was evaporated under a stream of nitrogen, and the samples were stirred for an additional 7 days. These conditions were also used for experiments with seeding. 'Seeds' that were prepared on the basis of pre-incubated Vicilin, Cupin-1.1 or Cupin-1.2 aggregates were added to the samples at the beginning of fibrillogenesis in 1% (v/v) concentration.

Pea seed protein extraction, in gel separation and transfer
A standard protocol was used for total protein isolation from pea seeds. Three

In vitro protein digestibility (IVPD)
Pea seed proteins were digested in vitro as described previously 36 . After each step of enzyme treatment samples were collected and inactivated by heating at 100°C for 5 min. Pepsin (Roche, Germany) and pancreatin from porcine pancreas (Sigma, USA) were used. Samples were checked with SDS-PAGE followed by western-blot with polyclonal anti-Vicilin antibodу (Imtek, Russian Federation). All analyses were performed in quadruplicate.

Immunochemical analysis
To obtain polyclonal anti-Vicilin antibodies from serum of a healthy rabbit, immunization with the antigen of the protein received according to the protocol described above was performed. Rabbit immunization and purification using affinity chromatography with rabbit antisera on a sorbent with immobilized recombinant Vicilin protein were carried out in Imtek company (Imtek, Russian Federation). Dilution 1:1 000 was used. Further goat anti-rabbit IgG (H+L) secondary antibody (Thermo Scientific, USA) was used in dilution 1:33 000. Also, 6x-His epitope tag antibody (4A12E4, Invitrogen, USA) was used in dilution 1:5 000, then goat anti-mouse IgG (H+L) secondary antibody (Invitrogen, USA) was used in dilution 1:5 000.

Protein body isolation
Protein bodies were extracted from fresh pea seeds of 30 days after pollination according to modified protocol for Zea mays protein bodies 63  Longer staining times were avoided due to rapid washing-off of immunofluorescent label. A control for nonspecific staining was performed by replacing the primary antibody produced in rabbit with 10% normal rabbit serum. Sections were analyzed with the fluorescence microscope Leica DM5500 and scanning confocal microscope Leica TCS SP5 (Leica, Germany).

Transmission electron microscopy
Micrographs were obtained using a transmission electron microscope Libra 120 (Carl Zeiss, Germany). The samples were placed on nickel grids coated with formvar films (Electron Microscopy Sciences, USA). To obtain electron micrographs, the method of negative staining with a 1% aqueous solution of uranyl acetate was used.

Polarization microscopy
Congo red staining of amyloid samples was performed using saturated Congo red (Sigma, USA) solution filtered through 45 µm filter (Millipore, USA). Samples stained with Congo red were air-dried and rigorously washed with distilled water. Birefringence was analyzed using Zeiss Axio Imager A2 (Carl Zeiss, Germany) polarization microscope equipped with cross polarizers.

ThT staining and confocal microscopy
Thioflavin T (ThT) UltraPure Grade (AnaSpec, USA) without after-purification was used. ThT-fibrils tested solutions were prepared by equilibrium microdialysis using a Harvard Apparatus/Amika device (USA). Equilibrium microdialysis was performed with a concentration of aggregates ~0.5 mg/ml and initial concentration of ThT ~32 μM.
Spectroscopic study of the sample and reference solutions prepared by proposed approach allowed us to determine the photophysical characteristics ThT bound to tested amyloids 24 .
For obtaining the fluorescence images of the ThT-stained fibrillar structures confocal laser scanning microscope Olympus FV 3000 (Olympus, Japan) was used. We used the oil immersion objective with a 60x magnification, numerical aperture NA 1.42 and laser with excitation line 405 nm.

Spectral measurements
The absorption spectra of the samples were recorded using a U-3900H spectrophotometer (Hitachi, Japan). The absorption spectra of proteins aggregates and ThT in their presence were analyzed along with the light scattering using a standard procedure 64  CD spectra in the far UV-region were measured using a J-810 spectropolarimeter (Jasco, Japan). Spectra were recorded in a 0.1 cm cell from 260 to 200 nm. For all spectra, an average of three scans was obtained. The CD spectrum of the appropriate buffer was recorded and subtracted from the samples spectra. It has turned out that the recorded CD spectra have a clear distortion in the region of 250-260 nm (Figure 2a).
According to literature light scattering of macromolecules can substantially distort the CD spectra (see, for example, 65 ). We have shown that the degree of the observed distortion (difference of recorded values from 0) actually correlates with the turbidity of the samples, which, in turn, is determined by the size of the studied aggregates ( Figure   2a, Inset). We attempted a quantitative analysis of the secondary structure by the CDPro program using three different regression methods (Selcon, Contin, and CDSSTR) and several basic sets of proteins with a known secondary structure (the sets include from 37 to 56 soluble, membrane, and denatured proteins with different content of the secondary structure). Since such results could be arbitrary because the standard basic sets of proteins used to estimation of the secondary structure content are not representative for the analysis of the spectra of protein aggregates, we have also conducted a visual analysis of the recorded spectra with the use of CD spectra of proteins and peptides with representative secondary structures 28 .

Time-resolved fluorescence measurements
Fluorescence decay curves were recorded by a spectrometer FluoTime 300 (PicoQuant, Germany) with the Laser Diode Head LDH-C-440 ( ex = 440 nm). The fluorescence of ThT was registered at  em = 490 nm. The measured emission decays were fit to a multiexponential function using the standard convolute-and-compare nonlinear least-squares procedure 66 . In this method, the convolution of the model exponential function with the instrument response function (IRF) was compared to the experimental data until a satisfactory fit was obtained. The fitting routine was based on the nonlinear least-squares method. Minimization was performed according to Marquardt 67 .

Data analysis
All experiments in this work were performed in at least three repeats. Data in figures is presented as the mean ± the standard error of the mean. In human cell toxicity assay, a dependence of the toxicity on the concentration was considered as significant if p-value of the regression coefficient in generalized linear model (glm function in R language) was lower than 0.05 and therefore the regression coefficient was significantly different from zero.