Biochemical Journal

Research article

An anti-Aβ (amyloid β) single-chain variable fragment prevents amyloid fibril formation and cytotoxicity by withdrawing Aβ oligomers from the amyloid pathway

Marta Marín-Argany , Geovanny Rivera-Hernández , Joaquim Martí , Sandra Villegas

Abstract

Aβ (amyloid β) immunotherapy has been revealed as a possible tool in Alzheimer's disease treatment. In contrast with complete antibodies, the administration of scFvs (single-chain variable fragments) produces neither meningoencephalitis nor cerebral haemorrhage. In the present study, the recombinant expression of scFv-h3D6, a derivative of an antibody specific for Aβ oligomers, is presented, as well as the subsequent proof of its capability to recover the toxicity induced by the Aβ1–42 peptide in the SH-SY5Y neuroblastoma cell line. To gain insight into the conformational changes underlying the prevention of Aβ toxicity by this antibody fragment, the conformational landscape of scFv-h3D6 upon temperature perturbation is also described. Heating the native state does not lead to any extent of unfolding, but rather directly to a β-rich intermediate state which initiates an aggregation pathway. This aggregation pathway is not an amyloid fibril pathway, as is that followed by the Aβ peptide, but rather a worm-like fibril pathway which, noticeably, turns out to be non-toxic. On the other hand, this pathway is thermodynamically and kinetically favoured when the scFv-h3D6 and Aβ1–42 oligomers form a complex in native conditions, explaining how the scFv-h3D6 withdraws Aβ1–42 oligomers from the amyloid pathway. To our knowledge, this is the first description of a conformational mechanism by which a scFv prevents Aβ-oligomer cytotoxicity.

  • aggregation
  • Alzheimer's disease
  • amyloid β (Aβ) peptide
  • immunotherapy
  • single-chain variable fragment (ScFv)
  • worm-like fibril

INTRODUCTION

Since the first report of neurotoxicity induced by a fragment of APP (amyloid precursor protein) (which included the C-terminal 105 residues) [1], increasing lines of evidence suggest that soluble Aβ (amyloid β) oligomers are the cause of the synapse loss and neuronal injury characteristic of AD (Alzheimer's disease) [2]. Consistently, it has been seen that certain cognitive impairments appear before amyloid deposits are formed [3]. Additionally, the severity of cognitive deficits correlates with the levels of soluble Aβ and not with the presence of amyloid plaques [4]. These observations indicate that neurodegeneration may begin prior to, and is not the result of, amyloid deposition. Accordingly, the need exists for new therapies capturing soluble Aβ oligomers rather than amyloid fibrils.

Naturally occurring antibodies against the Aβ peptide (39–43 amino acids long) are present in human cerebrospinal fluid and in the plasma of healthy individuals, but are significantly lower in AD patients, suggesting that this disorder may have an immunodeficient basis [5]. The first descriptions regarding the therapeutic potential of anti-Aβ peptide antibodies were based on the inhibition of in vitro fibrillation, and on the prevention of neurotoxicity in cell culture of the Aβ1–40 peptide [6]. The demonstration of the effectiveness of Aβ immunotherapy in vivo was obtained in the PDAPP transgenic mouse, which overexpresses mutant human APP (V717F) [7]. Active immunization with the Aβ1–42 peptide drastically reduced the amyloid burden in the cortex and in the hippocampus. Subsequently, the effectiveness of passive Aβ immunotherapy, i.e. administering Aβ-directed antibodies, was shown in the same mouse model [8]. Once clinical trials with animal models were successfully achieved, clinical trials with humans actively vaccinated with Aβ1–42 were developed [9]; however, the so-called AN-1792 clinical trial was halted because meningoencephalitis complications arose [10]. Although the treatment was stopped, those patients that were initially immunized continued to be monitored and a significant slowdown of the expected cognitive decline was reported [11]. At this time, and after several studies to prevent undesirable effects, the most advanced clinical trials are those for the ACC-001 vaccine, a derivative of the Aβ peptide, and antibody AAB-001, or Bapineuzumab, a murine mAb (monoclonal antibody) against the N-terminal 1–5 residues of the Aβ peptide (mAb-h3D6) which is specific for Aβ oligomers [1214].

Apart from the aforementioned meningoencephalitis, one of the side-effects of the administration of full mAb is the exacerbation of CAA (cerebral amyloid angiopathy) [15]. Passive immunization with scFvs (single-chain variable fragments), that is, antibodies devoid of Fc, is an attractive therapeutic strategy, since they do not activate microglia, do not produce cerebral haemorrhage, and have been demonstrated to be as potent and specific as the parent mAb [1618]. Nevertheless, the yield after recombinant expression and purification of scFv fragments remains, in general, too low, especially when the isolated (non-tagged) molecule is intended [19]. Production of scFv variants is usually performed in phage-display systems [16,17] or in the periplasm of Escherichia coli [18,20], and just a few cases of soluble expression of tagged-scFvs within the cytoplasm of Origami 2 (DE3), a defective strain for the two main redox pathways in E. coli, are reported [19,21].

In the present study, the sequence of the mAb-h3D6.v2 has been used to construct a synthetic gene consisting of the VH (variable domain of the heavy chain) and the VL (variable domain of the light chain) fused by a (Gly4Ser)3 linker (scFv-h3D6). Expression of thioredoxin- and NusA-tagged precursors was performed in E. coli Origami 2 (DE3), and the tags were afterwards efficiently removed by TEV (tobacco etch virus) protease. Demonstration of the effectiveness of scFv-h3D6 on reversing Aβ-induced toxicity was assessed in the SH-SY5Y neuroblastoma cell line. The subsequent description of the conformational landscape of scFv-h3D6 upon temperature perturbation permitted our group to gain insight into the molecular mechanisms behind its capability to capture Aβ1–42 oligomers. Heat treatment of scFv-h3D6 initiates an aggregation pathway characterized by a β-rich intermediate state which aggregates in the form of WL (worm-like) fibrils. These fibrils are different from the amyloid fibrils formed by the Aβ peptide, especially in terms of the toxicity of their oligomeric precursors. In addition, the formation of WL fibrils by scFv-h3D6 is kinetically and thermodynamically favoured upon binding Aβ1–42 oligomers, explaining how scFv-h3D6 withdraws Aβ1–42 oligomers from the amyloid pathway and, consistently, how it prevents cytotoxicity.

EXPERIMENTAL

Large-scale expression and purification of scFv-h3D6

For large-scale production, the intracellular expression in pETtrx-1a allowed for the purification of both soluble and insoluble fractions. Induction with 0.5 mM IPTG (isopropyl β-D-thiogalactopyranoside) was performed at D=0.7 and incubation in the shaker at 20 °C for 15 h. The cellular pellet was then washed three times with cold PBS (pH 7.4) and resuspended in Ni2+-binding buffer (20 mM sodium phosphate, 0.5 M NaCl and 0.5 mM EDTA, pH 7.4) containing a cocktail of protease inhibitors [1 μg/ml leupeptin, 1 μg/ml benzamidine, 1 μg/ml BPTI (basic pancreatic trypsin inhibitor) and 1 mM PMSF]. After two freeze–thaw cycles, the sample was sonicated for 5 cycles of 45 s, at 50% duty cycle and output 9 (Sonifier 450, Branson). The soluble and the insoluble fractions were fractionated at 43700 g. The soluble fraction, containing a 40.6 kDa precursor (His6–thioredoxin–TEV target–scFv), was purified by 5 ml Histrap HP columns (GE Healthcare). The presence of 0.5 mM EDTA was necessary to preclude undesirable proteolysis within the time-course of the IMAC (immobilized metal-ion-affinity chromatography). This precursor was also obtained by solubilizing the insoluble fraction in denaturing buffer (100 mM Tris/HCl, 10 mM GSH, pH 8.5, and 8 M urea) and refolding by dilution (1:10) in ice-cold refolding buffer (100 mM Tris/HCl, 100 mM L-arginine and 0.15 mM GSSG, pH 8.5) for 48 h. The precursor was proteolized at 30 °C with TEV protease for 4 h at a precursor/protease ratio of 50:1 (w/w) in 20 mM Tris/HCl, 100 mM NaCl, 0.5 mM EDTA, 0.3 mM GSSG and 3 mM GSH, pH 8.3 [22]. TEV protease was recombinantly obtained as described previously [23]. The scFv was fractionated from the initial fusion by binding the His-tagged proteins (thioredoxin and TEV protease) to 1 ml Histrap HP columns (GE Healthcare). Finally, a Superdex-75 gel-filtration chromatography (Hiload 26/60, GE Healthcare) in PBS (pH 7.4) at a flow rate of 2 ml/min was used to both completely purify and to assess the degree of dimerization of the isolated scFv-h3D6.

Because the yield of the protein fused to thioredoxin was not very high, and also because some undesirable proteolysis within the course of the initial IMAC was observed, an 83.4 kDa precursor containing NusA (His6–NusA–TEV target–scFv) was expressed. The protocol for purifying from the soluble fraction was similar to that described for the thioredoxin fusion, with the only modification being that of the Ni2+-binding buffer (20 mM Tris/HCl and 0.5 M NaCl, pH 8.0).

Finally, and in order to perform functional assays {i.e. MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide] assays}, lipopolysaccharides were removed from the protein by using Detoxi-Gel Endotoxin Removing columns (Thermo Scientific). Lipopolysaccharides are the major endotoxins of Gram-negative bacteria, and they could elicit a cellular response (up to 300% of viability in the MTT assay was detected; results not shown).

The buffer was changed to PBS using PD-10 Desalting Columns (GE Healthcare), and protein aliquots (500 μl) were stored at −20 °C until use. The molecular mass of the different batches of purified protein was confirmed by MALDI–TOF (matrix-assisted laser-desorption ionization–time-of-flight) spectrometry, and the protein concentration was determined from its absorbance at 280 nm using an absorbance coefficient of 1.5 absorbance-units/cm for a 1 mg/ml native scFv-h3D6 solution, as determined by the method of Gill and von Hippel [24].

Secondary structure determination

CD and FTIR (Fourier-transform infrared)

Protein secondary structure was monitored at different temperatures by far-UV CD spectroscopy from 260 nm–190 nm in a Jasco J-715 spectrophotopolarimeter. The protein concentration was 20 μM, and 20 scans were recorded in cuvettes of a 0.2 cm pathlength at 50 nm/min (response 2 s).

scFv-h3D6 at 100 μM (2.6 mg/ml) was dialysed at 4 °C against deuterated-PBS using 15000 MWCO (molecular mass cut-off) MINI Dialysis Units (G-Biosciences) for FTIR analysis, and the completeness of the H→2H exchange was monitored by the disappearance of the amide II band (1545 cm−1). Spectra were acquired at 25 °C, 37 °C and 60 °C in a Variant Resolutions Pro spectrometer using excavated cells with a 50 μm path (Reflex Analytical) and the series software licensed under OMNIC (Thermo Scientific). Typically, 1000 spectra recorded at a scan rate of 95 cm−1/min were averaged, and the series obtained was corrected against a background, the buffer was subtracted, and a vapour-control spectrum was finally subtracted. Data treatment and band decomposition of the original amide I band have been described elsewhere [25]. Briefly, the number and position of the component bands are obtained by deconvolution and derivation [25]. Bandwidths are estimated from the derivatives, and bandshape is set to a Gaussian curve. The fitting is obtained by iteration in two steps: band positions are fixed in a first iteration and are free in the second.

Thermal denaturation

Thermal denaturation was followed up by far-UV CD spectroscopy, tryptophan fluorescence and turbidity of 20 μM samples (tryptophan fluorescence was also assayed at 2 μM). Experiments were carried out from 25 °C to 90 °C, at a rate of 60 °C/h, by following the ellipticity at 218 nm, tryptophan fluorescence at 338 nm (excitation 290 nm, slits set at 5) and optical absorbance at 350 nm. CD and tryptophan fluorescence-emission spectra were recorded initially at 90 °C, and again after cooling the samples to 25 °C.

Limited protease digestion studies

To characterize the intermediate state present in the heat-induced aggregation pathway, limited thermolysin digestion was performed at 60 °C for 2 h in 20 mM Tris/HCl and 10 mM CaCl2 (pH 7.8), and aliquots removed at specific times had the reactions stopped with SDS-loading buffer [26]. A 1 mg/ml sample of scFv-h3D6 was pre-incubated at 60 °C for 10 min and a ratio of thermolysin to scFv of 1:400 (w/w) was used. Thermolysin hydrolyses peptide bonds on the N-terminal side of valine, leucine, isoleucine and phenylalanine residues [27].

Preparation of the Aβ peptide and aggregation

An Aβ1–42 synthetic lyophilized peptide, purified with HCl as the counter-ion (Caslo Laboratories ApS), was dissolved at 200 μM in HFIP (1,1,1,3,3,3-hexafluoro-2-isopropanol), a pre-treatment that breaks down β-sheet structures and disrupts hydrophobic forces, leading to monodispersed peptide preparations [28]. Aliquots of 150 μl were prepared, and HFIP was removed by drying in vacuum in a SpeedVac (Savant instruments), and the resulting peptide film was then stored at −80 °C until use. The peptide film was dissolved with 6 μl of dry DMSO (5 mM peptide) and subsequently diluted to 300 μl with PBS (100 μM peptide). Oligomerization was induced by incubation at 37 °C for different periods of time, whereas fibrillation was induced by heating the samples at 90 °C for 5 min.

Because of the method of preparation of the Aβ peptide, 2% (v/v) DMSO remained in the initial solution, and all of the samples to assess the effect of scFv-h3D6 on Aβ-peptide aggregation and cytotoxicity, including control samples, contained the same percentage of DMSO during incubation. It is worth noting that a 10-fold dilution was performed after incubation when analysing cytotoxicity, ThT (thioflavin T) and ANS (1,8-anilinonaphthalenesulfonate) binding, and TEM (transmission electron microscopy) (see below), so 0.2% DMSO remained in the final samples. Additionally, we have found that 2% DMSO does not affect the far-UV CD spectrum of scFv-h3D6 (see Supplementary Figure S1 at http://www.BiochemJ.org/bj/437/bj4370025add.htm).

Cytotoxicity assays

The SH-SY5Y human neuroblastoma cell line was grown in serum-supplemented medium in 5% CO2 at 37 °C. The medium contained 50% MEM (minimal essential medium) (Invitrogen), 50% Ham's modification of F-12 (Invitrogen), and was supplemented with 10% fetal bovine serum (Sigma), 1% MEM non-essential amino acids (Gibco) and a 1% mix of antibiotics: penicillin, streptomycin and anti-fungal amphotericin (Gibco). Cells were plated in 96-well tissue-culture-treated plates (Corning) at 3×104 cells/well in 100 μl of medium, and incubated for 24 h to allow for the attachment to the bottom of the wells. Medium was removed by aspiration and replaced with 100 μl of serum-free medium. We first determined the optimal concentration of the Aβ1–42 peptide to generate maximum cell toxicity by assaying samples ranging from 0 to 200 μM which were induced for oligomerization (37 °C, 3 h) before adding 10 μl to each well; cells were incubated with the peptide for 24 h. To assess the effect of scFv-h3D6 on the recovery of Aβ1–42-induced cell toxicity, samples of 100 μM Aβ1–42 peptide with increasing concentrations of scFv-h3D6 (0, 1, 10, 50 and 100 μM), were treated and assayed similarly. The MTT assay was performed. The MTT reagent was reconstituted in PBS to 5 mg/ml. MTT labelling reagent (10 μl) was added to each well, and the plate was incubated at 37 °C for 4 h. After removal by aspiration, 100 μl of solubilization solution (DMSO) was added to each well, and the plate was shaken for 10 min at room temperature (25 °C). The absorbance of the samples was measured at 540 nm (absorbance at 620 nm was afterwards subtracted) in a Microplate Photometer (Packard Instrument). Each condition consisted of six replicas per experiment, and four independent experiments were statistically analysed for significance (Wilcoxon signed-rank test, SPSS) after standardizing the data using concentration 0 (buffer) as the reference.

TEM

1–42 peptide (100 μM), 100 μM scFv-h3D6 and the combination of both molecules at the same concentration were induced for oligomerization (37 °C, 3 h and 48 h) and for amyloid fibril formation (90 °C, 5 min) and then analysed. Incubation at 60 °C for 10 min was also studied to obtain information about the heat-induced intermediate state (see the Results section). The samples were diluted 1:10 in PBS and quickly adsorbed on to glow-discharged carbon-coated grids. The material was stained using the method of uranyl acetate described previously [29], and the samples were visualized with an Hitachi H-7000 microscope.

ThT and ANS fluorescence-based aggregation assays

ThT and ANS fluorescence were used to quantify the aggregation extent of the samples visualized by TEM (see above). A 10 μl sample was added to 90 μl of a 27.78 μM ThT solution (final concentration 25 μM) or of a 56 μM ANS solution (final concentration 50 μM), and monitored at 25 °C. The excitation wavelengths were 450nm for ThT (emission 470–530 nm) and 388 nm for ANS (emission 400–600 nm); both slits were set to 10, and 10 spectra were averaged for each measurement. The intensities were quantified at 482 nm for ThT and at 470 nm for ANS. Each condition consisted of three replicas per experiment, and six independent experiments were averaged after standardizing the data using time 0 as the reference. Kinetics of ThT and ANS binding upon thermal denaturation (heating rate of 60 °C/h) were followed at 482 nm and 470 nm respectively.

RESULTS

Expression and purification of scFv-h3D6

A 750 bp NcoI–NotI fragment coding for the sequence of the VH domain of the humanized mAb h3D6.v2, a (Gly4Ser)3 linker, and the sequence of the VL domain of the same antibody, was designed and synthesized (GenScript). The fragment was subcloned into different pETM-derived vectors containing a His-tag, a signal peptide (DsbA, DsbC or PelB) or a protein tag [MBP (maltose-binding protein), Ztag, GB1 (Protein G B1 domain), GST (glutathione transferase), thioredoxin or NusA], as well as the target sequence for the TEV protease. The E. coli BL21 (DE3) strain was used to secrete the recombinant protein to the periplasmic space, whereas the Origami 2 (DE3) strain was used to produce correct folding in the cytoplasm. None of the signal peptides gave a high yield of recombinant protein, in agreement with previously published results for other scFv sequences secreted to the periplasmic space [18,30]. Origami 2 (DE3), a strain carrying mutations in the major intracellular disulfide-bond reduction systems (the glutathione/glutaredoxin and thioredoxin pathways), allowed for correct folding in the cytoplasm of this disulfide-containing protein (see below). The protein was N-tagged expressed (His6–thioredoxin–TEV target–scFv; 40.6 kDa) and digestion with the TEV protease released the isolated scFv-h3D6 (26.4 kDa). Digestion of the precursory fusion obtained from the soluble fraction was quite efficient, but it was revealed to be rather inefficient when using the refolded one (results not shown). At the end of the process, the yield of purified scFv-h3D6 from the soluble fraction was approximately 1 mg/l, whereas that obtained by refolding from inclusion bodies was approximately 4 mg/l.

Nevertheless, some undesirable proteolysis was observed within the time course of the initial IMAC, and 0.5 mM EDTA was mandatory in the binding buffer to preclude the generation of a scFv with two extra residues in its N-terminus (as determined by Western blotting and subsequent N-terminal sequencing). The proteolysis during IMAC might be indicative of an improper packing between the thioredoxin domain and the scFv, which could permit some E. coli metalloprotease to bind. For this reason, together with the aim of increasing the final yield, the scFv-h3D6 was expressed as a fusion with NusA (His6–NusA–TEV target–scFv; 83.4 kDa), and the precursor obtained did not show any undesirable proteolysis. Additionally, the high level of expression in the soluble form makes the refolding from the insoluble fraction unnecessary and permits a complete cleavage from the precursory form by the TEV protease. The final yield obtained for the isolated scFv-h3D6 was 7 mg/l.

The tendency for scFv molecules to dimerize is known, presumably by domain swapping [31]. These dimers appear to be kinetically trapped species produced during recombinant expression. The final step of purification consisted of a Superdex-75 gel filtration in PBS and rendered approximately 15% of dimerized scFv. The purified monomeric form of scFv-h3D6 was analysed at the highest concentration used in the present study (200 μM) by the same method and proved to be 100% monomeric (results not shown).

Secondary structure of scFv-h3D6

Before performing functional assays, the folding of scFv-h3D6 was characterized at different temperatures by CD and FTIR spectroscopy. Figure 1(A) shows two ellipticity minima (218 nm and 230 nm), an ellipticity maximum (200 nm) and a positive shoulder (237 nm) in the CD spectrum of the native state recorded at 25 °C. The spectrum recorded at 37 °C was identical (results not shown). The ellipticity minimum at 218 nm, together with the ellipticity maximum at 200 nm, shows the typical Ig β-sheet secondary structure. The CD contributions from the aromatic and/or cystinyl side-chains can interfere in the far-UV CD spectra and could explain the second ellipticity minimum at 230 nm and the positive shoulder at 237 nm [32]. This particular shape of the spectra is lost at 60 °C (Figure 1A); indeed, the spectra recorded at 60 °C and at 90 °C showed a canonical β-sheet conformation (a single ellipticity minimum at 215 nm) that was maintained after returning to 25 °C (results not shown), indicating that the thermal denaturation of scFv-h3D6 follows an irreversible process.

Figure 1 Secondary structure and thermal denaturation of scFv-h3D6

(A) Far-UV CD spectra at different temperatures. The spectrum at 25 °C shows two ellipticity minima (218 nm and 230 nm), an ellipticity maximum (200 nm) and a positive shoulder (237 nm). This particular shape of the spectra is lost at 60 °C in favour of a canonical β-sheet conformation (ellipticity minimum at 215 nm), which is maintained at 90 °C and after renaturation (results not shown). (BD) Band decomposition of the FTIR amide I spectra at relevant temperatures. (B) 25 °C. (C) 37 °C. (D) 60 °C. Decomposition at 25 °C and 37 °C generated four components, whereas at 60 °C, eight components contributed to the spectrum (see Table 1). The dotted line indicates the 1626 cm−1 component. (E) Thermal denaturation followed by CD (ellipticity at 218 nm). (F) Thermal denaturation followed by tryptophan fluorescence (338 nm) and turbidity (A350). ●, protein fluorescence at 20 μM; ░, protein fluorescence at 2 μM; ○, protein turbidity at 20 μM.

Because the particular shape of the initial CD spectrum actually precludes deconvolution, analysis by FTIR was performed at different temperatures (Figures 1B–1D and Table 1). Band decomposition of the amide I spectrum at 25 °C generated three main bands located at 1681 cm−1, 1660 cm−1 and 1636 cm−1, and a minor band located at 1612 cm−1 (Figure 1B). The low- and high-frequency components of the antiparallel β-sheet, characteristic of an Ig fold, were located at 1636 cm−1 and 1681 cm−1 respectively, and its β-turns/loops were located at 1660 cm−1 [33,34]. At 37 °C, the spectrum was very similar (Figure 1C) although the component for turns/loops slightly increased (Table 1). However, band decomposition of the spectrum acquired at 60 °C generated four additional bands located at 1671 cm−1, 1653 cm−1, 1644 cm−1 and 1626 cm−1 (Figure 1D). The 1671 cm−1 band is assigned to β-turns/loops in the bibliography, as is the band at 1660 cm−1 [33]. The band at 1653 cm−1 corresponds to α-helices, the 1644 cm−1 to random coil, and the 1626 cm−1 to WL fibrils [35]. Although also accompanied by an increase of the 1615 cm−1 band, assigned to amyloid fibrils [36], the different β-conformation detectable by far-UV CD upon heating scFv-h3D6 mainly corresponded to the presence of WL fibrils (Table 1) (see below). Amyloid and WL fibrils populate different aggregation pathways: amyloid fibrils are straight and long, and form following a nucleated-dependent kinetics, whereas WL fibrils are curved and short, and form following non-nucleation-dependent kinetics [37].

View this table:
Table 1 Band decomposition of FTIR amide I band of scFv-h3D6 acquired at different temperatures

Thermal denaturation of scFv-h3D6

To delve into the conformational transition from the native state to that observed at 60 °C, thermal denaturation was followed by CD, tryptophan fluorescence and turbidity (Figures 1E and 1F). Ellipticity at 218 nm decreases with the temperature starting at 45 °C, such that the fold enriches its β-sheet content (Figure 1E). At 60 °C, however, there is a transition that leads to the loss of part of this newly acquired β-content and the stabilization of the signal at above 65 °C. The fact that the β-content initially increases, without previously suffering from an evident unfolding, shows a direct rearrangement of the native β-sheets. This rearrangement leads to partial aggregation, as can be observed, and subsequent partial loss of the signal. Turbidity also corroborates that thermal denaturation leads to aggregation (Figure 1F). Also, the CD spectrum at 60 °C was clearly different from the initial spectrum (Figure 1A), and thermal denaturation was revealed as irreversible. This conformational reorganization leads to the appearance of an intermediate state, which is prone to aggregation and generates the WL-fibril FTIR-component described above.

Although it is known that intrinsic tryptophan fluorescence decreases with temperature and, in consequence, no quantification upon varying the temperature can be made, thermal denaturation followed by tryptophan fluorescence has been used to determine the effect of protein concentration in the aggregation process (Figure 1F). When comparing the thermal denaturation at protein concentrations of 2 μM and 20 μM, a transition at approximately 60 °C is observed in both cases. Although the transition region is sharper at the higher concentration, the slopes before and after this point remained unaffected, which implies that the aggregation of scFv-h3D6 follows non-nucleation kinetics, as has previously been reported as being a main feature of WL fibrils [37].

In order to characterize the intermediate state in the aggregation pathway, limited proteolysis with thermolysin was performed at 60 °C and analysed by SDS/PAGE (see Supplementary Figure S2 at http://www.BiochemJ.org/bj/437/bj4370025add.htm). No apparent band was accumulated during the time-course of the digestion, and after 2 h the scFv-h3D6 was completely proteolysed. This indicates a conformation where both domains are equally susceptible to digestion.

Once we learned that the scFv-h3D6 is properly folded at temperatures below 60 °C, we were ready to test its effect on the toxicity of the Aβ peptide at physiological conditions.

Prevention of Aβ1–42 toxicity by scFv-h3D6

We first determined the optimal concentration of the Aβ1–42 peptide to generate maximum cell toxicity in the human SH-SY5Y neuroblastoma cell line, as determined by the MTT method. Figure 2(A) shows 10 μM and 20 μM of the Aβ1–42 peptide as the most effective concentrations in the culture to reduce viability. Although, at first glance, the fact that 20 μM Aβ1–42 is not more toxic than 10 μM is surprising, it makes sense that at higher concentrations the aggregation pathway is shifted towards the formation of amyloid fibrils instead of towards the formation of oligomeric cytotoxic species, because amyloid fibrils form following nucleation-dependent kinetics. Since we are interested in reversing the toxicity of the Aβ oligomers, 10 μM of the Aβ1–42 peptide in the MTT assay shown in Figure 2(B) was used. Samples of Aβ1–42 with different concentrations of scFv-h3D6 were induced for aggregation by incubation at 37 °C for 3 h. The treatment with 10 μM Aβ1–42 reduced cell viability to 60%, 10 μM scFv-h3D6 alone had no effect, and the combination of both molecules recovered cell viability in a dose-dependent manner. A ratio of 2:1 (10 μM Aβ1–42/5 μM scFv-h3D6) was enough to observe a significant effect, with the equimolar ratio being the one that almost completely abolished the toxicity induced by the Aβ1–42 peptide.

Figure 2 MTT toxicity assays with the SH-SY5Y human neuroblastoma cell-line

(A) Toxicity induced by different concentrations of the Aβ1–42 peptide. *P<0.068 compared with 0 μM Aβ1–42. (B) Recovery of cell viability in the presence of 10 μM of Aβ1–42 peptide by adding different concentrations of scFv-h3D6. S, 10 μM scFv-h3D6 alone (without Aβ peptide); B, buffer (without Aβ peptide and scFv). *P<0.068 compared with 0 μM scFv-h3D6. Results are means±S.E.M. Significance was calculate using the Wilcoxon signed-rank test for four independent experiments (six replicas for each condition per experiment).

Because of the therapeutic potential of scFv-h3D6, it is encouraging to gain insight into the conformational changes underlying the prevention of Aβ toxicity by this antibody fragment. Given that the Aβ peptide follows the amyloid fibril pathway and that scFv-h3D6 follows the WL fibril pathway (see above), it becomes necessary to study these misfolding pathways separately to further study what occurs when both molecules are combined.

Misfolding of the Aβ1–42 peptide

Since mAb-h3D6 preferentially recognizes Aβ oligomers [1214], which are the precursors of amyloid fibrils and the cytotoxic species, we purposely used mild conditions (PBS, 3 h at 37 °C) to aggregate the Aβ1–42 peptide before adding it to the cell cultures. To demonstrate that in these conditions oligomers are actually formed, TEM was performed (Figure 3A). Oligomers are visualized at the moment of dilution of the Aβ-peptide film (see the Experimental section) and remain after incubation for different periods at 37 °C, without the occurrence of amyloid fibrils. This observation does not mean we are out of the amyloid fibril pathway, because heat treatment (60 °C and 90 °C) induces the formation of amyloid fibrils.

Figure 3 Regulation of formation of amyloid fibrils and WL fibrils by Aβ1–42, scFv, DMSO and temperature

TEM of 100 μM Aβ1–42 (A), 100 μM scFv-h3D6 (B), 100 μM scFv-h3D6 in the absence of DMSO (see the Experimental section) (C), and 100 μM Aβ1–42 plus 100 μM scFv-h3D6 (D), at different temperatures. (A) The cytotoxic oligomers of the Aβ peptide are visualized at 37 °C, and heating at a higher temperature is mandatory to obtain amyloid fibrils. (B) ScFv-h3D6 does not form WL fibrils in the presence of DMSO. (C) ScFv-h3D6 in the absence of DMSO forms small oligomers at 37 °C, and at 60 °C (where the thermal-induced intermediate state of scFv-h3D6, I-state, is more populated) initiates the formation of WL fibrils; these WL fibrils are better structured after treatment at 90 °C. (D) The Aβ1–42–scFv-h3D6 complex directly forms WL fibrils at 37 °C and heating disrupts them, allowing amyloid fibrils and oligomers to form.

The different temperature treatments performed for TEM visualization were also performed before quantifying ThT and ANS binding at 25 °C (Figures 4A and 4B). ThT is considered to be a specific dye for amyloid fibrils [38]; whereas ANS has been extensively used as a probe of hydrophobic binding sites in proteins [39]. In the conditions used in the present study, the Aβ1–42 peptide initially binds some ThT, and treatment at 90 °C increases the fluorescence signal 2-fold (Figure 4A), whereas a very low level of ANS binding is detected, even upon heating (Figure 4B). This indicates that we are, effectively, on the amyloid pathway.

Figure 4 Fluorescence-based aggregation assays

(A and B) Different temperature treatments were used to study binding to ThT (482 nm) (A) and ANS (470 nm) (B). Results are means±S.E.M. (C and D) Binding upon thermal denaturation to ThT (482 nm) (C) and ANS (470 nm) (D).

Misfolding of scFv-h3D6

As mentioned above, heat treatment of scFv-h3D6 initiates an aggregation pathway characterized by a β-rich intermediate state which aggregates in the form of WL fibrils. Unexpectedly, these WL fibrils were not visualized by TEM when assaying scFv-h3D6 alone, instead, oligomers appeared in all of the conditions tested (Figure 3B). It is important to note that these sets of experiments, as well as all of the experiments to assay the effect of scFv-h3D6 in Aβ-peptide cytotoxicity and aggregation, contained 2% DMSO during incubation (0.2% in the final samples) in order to properly be compared with those in the presence of the Aβ peptide, which inevitably contained these DMSO traces (see the Experimental section). As expected, when DMSO is not present, WL fibril formation by the isolated scFv-h3D6 is initiated upon incubation at 60 °C, and this becomes more evident upon treatment at 90 °C (Figure 3C). This is in accordance with the FTIR analysis discussed above, where a characteristic band for WL fibrils appeared at 60 °C (Figure 1D and Table 1).

ScFv-h3D6 alone does not bind ThT, and temperature induces just some binding, especially at the temperature where the heat-induced intermediate state is present (Figure 4A). ANS is bound to scFv-h3D6 in all of the conditions, although also preferentially at the temperature where the heat-induced intermediate state is present (Figure 4B). The behaviour is similar when no traces of DMSO are present (results not shown), revealing that neither ThT nor ANS are capable of distinguishing between the oligomers and the WL fibrils formed by scFv-h3D6.

In order to confirm the relevance of the heat-induced intermediate state in the aggregation pathway of scFv-h3D6, the ThT and ANS binding kinetics upon heating were assayed without stirring (Figures 4C and 4D). Effectively, at 60 °C, both dyes are bound to the intermediate state, and at higher temperatures the signal is lost because insoluble aggregation occurs.

What occurs with misfolding pathways when the Aβ1–42 peptide and scFv-h3D6 are combined?

The addition of an equimolar ratio of scFv-h3D6 to the Aβ1–42 peptide prevented the formation of Aβ oligomers and, in its place, initiated the formation of WL fibrils immediately after mixing (Figure 3D). These WL fibrils were present in the samples used for the cytotoxicity assays (3 h at 37 °C), where the toxicity of Aβ oligomers was prevented (Figure 2B), and became more evident at a longer period of incubation (48 h at 37 °C). When the WL fibrils of the Aβ1–42–scFv-h3D6 complex are heated at 60 °C (the temperature where the heat-induced intermediate of scFv-h3D6 is observed), they begin to disassemble, and at 90 °C both amyloid fibrils and oligomers appear. Taking into account that WL fibrils are kinetically trapped species, the temperature has depopulated this state, allowing the actual amyloid-fibril pathway by the Aβ1–42 peptide to proceed.

In contrast with the behaviour of the scFv-h3D6 alone, when mixing Aβ1–42–scFv-h3D6, the binding of ThT and ANS is apparent from zero time and becomes more evident upon heating (especially at 90 °C) (Figures 4A and 4B). This agrees with the presence of WL fibrils, which are known to bind both dyes [37,40], before heating and the presence of amyloid fibrils and oligomers upon heating.

Hence, it appears that, although the formation of WL fibrils is an intrinsic property of the scFv-h3D6, those made by the Aβ1–42–scFv-h3D6 complex are more stable than those made by the isolated scFv-h3D6, because they occur without the requirement of heat treatment and even occur in the presence of DMSO traces.

DISCUSSION

In the present study, the production and conformational characterization of a scFv designed from the sequence of a mAb of therapeutic interest for AD is described. Passive immunization with scFv is an appealing therapeutic strategy, since it does not produce the undesirable side-effects of complete antibody administration while maintaining the avidity of the parent mAb [1618].

Expression of scFv-h3D6 and chaperone activity

The effective recombinant production of scFv fragments remains a challenging matter, mainly because of the difficulty of properly folding both cystinyl-containing domains. In the present study, acceptable yields have been obtained from both the soluble and the insoluble-refolded fractions by expressing, in the cytoplasm of E. coli Origami 2 (DE3), a thioredoxin-tagged scFv that afterwards was released by TEV proteolysis. Thioredoxin has also been useful in producing other functional scFvs, and its capability has been related to its chaperone activity rather than to its disulfide-isomerase activity [21]. A fusion to NusA, a protein known to mediate termination of transcription [41] and to display chaperone activity [42], was also tried. To our knowledge, the highest yield for a soluble scFv in the cytoplasm of Origami 2 (DE3) has been obtained by expressing a NusA-tagged precursor (3 mg/l) [19]. In the present study, the fusion to this chaperone allowed us to obtain a soluble precursor that generated, after TEV protease digestion, a yield of 7 mg/l of scFv-h3D6. Thus it is likely that, in order to properly fold cystinyl-containing domains in the cytoplasm of Origami 2 (DE3), it is advisable to improve chaperone activity.

Folding landscape of scFv-h3D6 upon temperature perturbation

ScFv-h3D6 shows the typical secondary structure of an Ig fold, with around 60% of antiparallel β-sheet and 30% of loops/turns components, as determined by FTIR. The CD spectrum of the native protein is affected by contributions from the aromatic and/or cystinyl side-chains within the far-UV [32], featuring a second ellipticity minimum at 230 nm and a positive shoulder at 237 nm. Although other scFv molecules in the literature show ordinary β-sheet spectra [4346], the second ellipticity minimum shown in the present study is very similar to that described at 235 nm for a VL domain [47]. This ellipticity minimum was attributed to the interaction of the aromatic residues with the conserved Trp35.

Thermal denaturation of scFv-h3D6 leads to aggregation, as seen by CD, tryptophan fluorescence and turbidity. The CD spectrum in the mid-point of the transition (60 °C) shows the presence of an intermediate state with a conformation similar to that at 90 °C; this intermediate state is generated by direct reorganization of the native state, and no unfolding is required in order to enter the aggregation funnel. When submitted to thermolysin proteolysis, a defined pattern of digestion was not obtained, suggesting that both domains are equally susceptible to proteolysis. This β-rich intermediate state accumulates in a concentration-independent fashion, following non-nucleation-dependent kinetics. Also, heat-induced aggregation of scFv-h3D6 leads to the formation of WL fibrils rather than to amyloid fibrils, as indicated by TEM. These WL fibrils are known to follow nucleation-independent kinetics and to bind to ThT and ANS [37,40]. Additionally, the FTIR spectrum of these WL fibrils shows a band at 1626 cm−1, similar to the 1622 cm−1 band characteristic for the WL fibrils of β2-microglobulin [35]. Despite the fact that both folds correspond to the same class (Ig fold), it has been demonstrated that other folds can also form WL fibrils (i.e. [48]).

The general aggregation tendency of scFv molecules upon thermal denaturation has previously been reported in the literature [44,49], but the conformational nature of these aggregates is described for the first time in the present study. Other similar folds that are known to aggregate in vivo do form amyloid fibrils, as is the case for the light-chains of immunoglobulins, especially of their isolated variable domains (VL) [47], and the above-mentioned β2-microglobulin (a 99-residue, all β-sheet protein from the major histocompatibility class I complex) [40]. Acidic pH and high ionic strength are necessary to observe WL fibrils in vitro of β2-microglobulin [37], whereas those shown in the present study for the scFv-h3D6 alone are formed in a physiological buffer (PBS) after heat induction.

Effect of scFv-h3D6 on Aβ1–42 peptide fibrillation and toxicity

Evidence suggests that therapeutic interventions that reduce Aβ fibrils at the cost of augmenting non-fibrillar Aβ assemblies, including Aβ oligomers, could be harmful [50]. The mAb from which scFv-h3D6 is derived specifically recognizes Aβ oligomers [12].

When samples induced for oligomerization were added to cell cultures, 10 μM scFv alone had no effect, 10 μM Aβ1–42 reduced cell viability to 60%, and the previous addition of scFv-h3D6 to 10 μM Aβ1–42 annulled the effect of the peptide on cell viability in a dose-dependent manner. In the conditions used in the present study for the cytotoxicity assays, Aβ1–42 oligomers are formed, as visualized by TEM. The fact that WL fibrils are formed when mixing the Aβ1–42 peptide and scFv-h3D6, without the requirement of the isolated scFv of incubation at 60 °C and in the presence of DMSO, implies that these WL fibrils are both kinetically and thermodynamically favoured when scFv-h3D6 traps Aβ1–42 oligomers. Upon heating, the WL state becomes depopulated by dissociation of the complex and, in turn, some amyloid fibrils and oligomers appear. This would imply that temperature pulls the Aβ peptide out of the WL pathway and into the amyloid pathway, whereas the scFv-h3D6 remains in the WL pathway in the form of oligomers (Figure 5). As has been demonstrated in the present study, in the presence of DMSO traces, inherent to the presence of the Aβ peptide, the oligomers of scFv-h3D6 cannot assemble into WL fibrils.

Figure 5 Energy diagrams for the aggregation pathways of the Aβ1–42 peptide, scFv-h3D6, and the Aβ1–42–scFv-h3D6 complex

(A) At 37 °C (→), the Aβ peptide follows the amyloid pathway through the formation of cytotoxic oligomers (O) and scFv-h3D6 (S) follows the WL pathway through the formation of non-cytotoxic oligomers (sO). Temperature treatment (·····▶) is necessary to convert oligomers of Aβ into amyloid fibrils (AF) and sO of scFv-h3D6 into WL fibrils, in the latter case in the absence of DMSO traces. (B) The Aβ1–42–scFv-h3D6 complex directly forms WL fibrils, and its disruption by temperature (·····▶) generates sO of scFv-h3D6, whereas the Aβ peptide ‘jumps’ into the amyloid fibril pathway, forming oligomers and AF.

Apart from the details of the aggregation pathways, the most relevant conclusion of the present study is that, in native conditions, scFv-h3D6 inhibits Aβ-peptide amyloid fibril formation and cytotoxicity by pulling its oligomers towards the WL pathway. The description of the mechanism by which scFv-h3D6 protects against Aβ-peptide toxicity opens a new way to hereafter improve its therapeutic potential.

AUTHOR CONTRIBUTION

Geovanny Rivera-Hernández obtained the recombinant scFv and performed its aggregation studies. Marta Marín-Argany performed all of the studies to assess the effect of scFv-h3D6 on Aβ peptide fibrillation and toxicity. Joaquim Martí supervised the toxicity assays. Sandra Villegas designed and supervised all of the experimental work, and wrote the manuscript.

FUNDING

This work was funded by Instituto de Salud Carlos III [grant numbers FIS-PI07-0148 and FIS-PI10-00975], Fundación Mutua Madrileña [grant number FMM-08], and Generalitat de Catalunya [grant number SGR 2009-00761]. G.R.-H. is supported by a MAEC-AECI (Ministerio de Asuntos Exteriores, Spain) fellowship and M.M.-A. is supported by a PIF (UAB) fellowship.

Acknowledgments

We thank Professor Luis Serrano and Dr Jose C. Martínez for their critical comments on the manuscript and Dr Alicia Roque for her FTIR technical assistance.

Abbreviations: Aβ, amyloid β; AD, Alzheimer's disease; ANS, 1,8-anilinonaphthalenesulfonate; APP, amyloid precursor protein; FTIR, Fourier-transform infrared; HFIP, 1,1,1,3,3,3-hexafluoro-2-isopropanol; IMAC, immobilized metal-ion-affinity chromatography; mAb, monoclonal antibody; MEM, minimal essential medium; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide; scFv, single-chain variable fragment; TEM, transmission electron microscopy; TEV, tobacco etch virus; ThT, thioflavin T; WL, worm-like

References

View Abstract