Application of DNP-enhanced solid-state NMR to studies of amyloid-β peptide interaction with lipid membranes

Abstract

The cellular membrane disruption induced by the aggregation of Aβ peptide has been proposed as a plausible cause of neuronal cell death during Alzheimer’s disease. The molecular-level details of the Aβ interaction with cellular membranes were previously probed using solid state NMR (ssNMR), however, due to the limited sensitivity of the latter, studies were limited to samples with high Aβ-to-lipid ratio.

The dynamic nuclear polarization (DNP) is a technique for increasing the sensitivity of NMR. In this work we demonstrate the feasibility of DNP-enhanced ssNMR studies of Aβ₄₀ peptide interacting with various model liposomes: (1) a mixture of zwitterionic 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC) and negatively charged 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1ߣ-rac-glycerol) (POPG); (2) a mixture of POPC, POPG, cholesterol, sphingomyelin and ganglioside GM1; (3) the synaptic plasma membrane vesicles (SPMVs) extracted from rat brain tissues. In addition, DNP-ssNMR was applied to capturing changes in Aβ₄₀ conformation taking place upon the peptide insertion into POPG liposomes. The signal enhancements under conditions of DNP allow carrying out informative 2D ssNMR experiments with about 0.25 mg of Aβ₄₀ peptides (i.e. reaching Aβ₄₀-to-lipid ratio of 1:200). In the studied liposome models, the ¹³C NMR chemical shifts at many ¹³C-labelled sites of Aβ₄₀ are characteristic of β-sheets. In addition, in POPG liposomes the peptide forms hydrophobic contacts F19-L34 and F19-I32. Both the chemical shifts and hydrophobic contacts of Aβ₄₀ in POPG remain the same before and after 8 hours of incubation. This suggests that conformation at the ¹³C-labelled sites of the peptide is similar before and after the insertion process. Overall, our results demonstrate that DNP helps to overcome the sensitivity limitation of ssNMR, and thereby expand the applicability of ssNMR for charactering the Aβ peptide interacting with lipids.

1. Introduction

The amyloid cascade hypothesis suggests that the pathology in Alzheimer’s disease (AD) is a result of dysregulation in the production and clearance of Aβ peptide, which predominantly exists in its 40- and 42-residue alloforms (known as Aβ₄₀ and Aβ₄₂ respectively). Such dysregulation eventually leads to the peptide self-assembly into various amyloid aggregates, among which the fibrils represent the main constituent of amyloid plaques formed in the in human AD brain tissue. Although the deposition of insoluble fibrillar aggregates in human brain has been recognized as a crucial factor for inducing neuronal loss, the detailed mechanisms explaining how the fibrillar Aβ aggregation may cause neuronal cell death, remain unclear. One plausible cause for the cell death is the disruption of the cellular membrane induced by the aggregation of Aβ (Nagarathinam et al., 2013; Niu et al., 2018; Williams and Serpell, 2011). This is supported by many previous studies, in which the fibrillization of membrane-associated Aβ peptides leads to changes in a variety of physicochemical and physiological properties of model systems, such as the synthetic phospholipid bilayers and the neuronal plasma membranes in living cells (Cheng et al., 2018; Delgado et al., 2016; Gibson Wood et al., 2003; Hayashi et al., 2000; Kotler et al., 2014; Milanesi et al., 2012; Oshima et al., 2001; Peters et al., 2009; Sciacca et al., 2018; Vander Zanden et al., 2019; Widenbrant et al., 2006; Yip and McLaurin, 2001).

While biophysical and biological studies of the interaction between Aβ peptides and membranes are abundant, there have been much fewer experimental works on the molecular-level details of such interaction. One-dimensional (1D) ²H, ³¹P and ¹³C solid-state nuclear magnetic resonance (ssNMR) spectroscopy done by Separovic and co-workers demonstrated that the neurotoxic Aβ peptides interacted differently with phospholipid molecules when they were added externally or pre-incorporated with vesicles, and when metal ions such as Cu²⁺ or Zn²⁺ were added to the model membranes (Gehman et al., 2008; Lau et al., 2006, 2007). Yang and co-workers applied ssNMR to determine the backbone architectures of Aβ₄₀ fibrils formed in the presence of zwitterionic lipid bilayers. The structures of such fibrils showed apparent differences (especially at the C-terminal segment) compared with the Aβ₄₀ fibril structures solved in aqueous buffer (Niu et al., 2018, 2014). Additionally, solution NMR spectroscopy done by Ramamoorthy and coworkers showed that Aβ₄₀ adopted partially helical conformation within its conserved residues upon binding to zwitterionic lipid vesicles (Korshavn et al., 2016).

However, due to low sensitivity of ssNMR spectroscopy, application of this technique for studies of Aβ has been limited to simple 1D measurements (Gehman et al., 2008) or to samples with rather large or non-biological Aβ-to-lipid ratio (Cheng et al., 2018). In particular, recent ssNMR works explored the fibrillation pathways of Aβ₄₀ in the presence of lipids (Cheng et al., 2020, 2018; Qiang et al., 2014). There, the NMR signals of ¹³C-labelled sites in Aβ₄₀ were limited because lipids filled a significant fraction of the NMR sample space, thus diluting the peptides. Overall, the sensitivity of such experiments enabled 2D ssNMR measurements with Aβ₄₀-to-lipid ratio of 1:30 (Cheng et al., 2018; Qiang et al., 2014).

However, there is a need for probing the systems with lower Aβ₄₀-to-lipid ratios. In particular, at Aβ₄₀-to-lipid ratios smaller than 1:30, certain Aβ fibrillation pathways become dominant and therefore under such conditions they can be studied individually, without interference from other pathways (Akinlolu et al., 2015). The requirement for probing low Aβ₄₀-to-lipid ratios is also important for characterizing the structures of Aβ₄₀ peptides formed upon its binding and insertion into the membrane (Arce et al., 2011; Quist et al., 2005; Wong et al., 2009; Zhao et al., 2011). Acquiring a good quality ssNMR 2D spectrum of a sample containing several milligrams of Aβ₄₀ ¹³C-labelled at specific sites, typically takes a day of signal averaging. Therefore, such conventional ssNMR spectroscopy cannot be applied for probing transient processes taking place on a shorter timescale.

The dynamic nuclear polarization (DNP) solves the sensitivity problem in many domains of NMR spectroscopy and imaging (Brindle et al., 2011; Thankamony et al., 2017; Zhang and Hilty, 2018). The DNP increases NMR signals by transferring large polarization of unpaired electron spins to the coupled nearby nuclei via microwave (MW) irradiation of electron spin transitions. One area where DNP methods are especially helpful is ssNMR spectroscopy with magic angle spinning (MAS) at moderately high magnetic fields (Thankamony et al., 2017). Currently there are several commercially available MAS DNP systems that operate at magnetic fields of 9.4–18.8 T (Rosay et al., 2016). Since DNP mechanisms become effective at cryogenic temperatures, DNP-ssNMR measurements are typically carried out at temperatures of ≲100 K that can be achieved by cooling with cold N₂ or He gas (Bouleau et al., 2015; Matsuki et al., 2012; Thurber et al., 2013). The unpaired electron spins required for DNP are usually introduced in the form of biradicals or triradicals (Sauvée et al., 2013; Song et al., 2006; Thankamony et al., 2017; Thurber et al., 2010). The microwave irradiation required for saturating their electron spin transitions is typically produced by a high power source such as a gyrotron or extended interaction oscillator/klystron (Becerra et al., 1995; Kemp et al., 2016; Potapov et al., 2015; Rosay et al., 2016).

In the recent decade DNP has been widely applied to studies of various biological solids such as amyloid fibrils (Bayro et al., 2011; Debelouchina et al., 2013, 2010; Frederick et al., 2017; Potapov et al., 2013; Weirich et al., 2016), viral DNA (Sergeyev et al., 2011), membrane proteins (Bajaj et al., 2009; Kaur et al., 2015; Koers et al., 2013; Mao et al., 2014; Mehler et al., 2015; Smith et al., 2015) and to studies of protein folding kinetics (Jeon et al., 2019). In addition, DNP-enhanced ssNMR spectroscopy was used to characterize the structures of Aβ₄₀ aggregates (Potapov et al., 2015), in which the uniform ¹³C and ¹⁵N labels were introduced only at a few selected residues. There, the DNP-ssNMR provided key information about the molecular structures of fibrils, protofibrils, monomers and aggregates forming at elevated concentrations of monomeric Aβ₄₀. In particular, it was shown that β-sheet-like structures persisted in all forms of Aβ₄₀ aggregates, and individual Aβ₄₀ molecules in all aggregates had a varying propensity to fold into the U-shape as seen in many fibrils. Although some of the aggregates are metastable the cryogenic temperatures employed in DNP-ssNMR spectroscopy quench such structural transitions and allow detection of the transient states.

The main goal of this work is to demonstrate the applicability of DNP-enhanced ssNMR to studies of Aβ₄₀ peptide interaction with several lipid bilayer models. Here we focus on the attainable signal enhancements, spectral resolution and discuss the overall feasibility of extracting useful information from the chemical shifts and inter-residue cross-peaks. To this end we characterized Aβ₄₀ interacting with the following model liposomes: (1) a mixture of zwitterionic 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC) and negatively charged 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1’-rac-glycerol) (POPG); (2) a mixture of POPC, POPG, cholesterol, sphingomyelin and ganglioside GM1; (3) the synaptic plasma membrane vesicles (SPMVs) extracted from rat brain tissues. In addition, we apply the DNP technique for capturing changes in Aβ₄₀ conformation taking place upon initial binding to POPG liposomes.

2. Methods

Peptide synthesis.

All Aβ₄₀ peptides were synthesized manually using routine solid-phase peptide synthesis protocols with FMOC chemistry. The crude peptides were cleaved from the resin support using a mixture of trifluoroacetic/phenol/water/1,2-ethanedithiol/thioanisole with volume % ratio of 90:5:10:5:2.5. All peptides were purified using High-Performance Liquid Chromatography system (HPLC 1200 Series, Agilent Inc.) equipped with C18 reversed-phase column. After purification, the peptides were lyophilized and stored in a freezer at −20 °C. The purified peptides were verified with LC-MS/ESI (LCMS-2020, Shimadzu Inc.) to confirm >95% purity. Isotopic-labeling patterns of Aβ₄₀ samples for ssNMR measurements are summarized in Table 1.

Table 1.

DNP enhancements measured at various regions of the ¹³C-CP spectra of Aβ₄₀ in lipids. The values are missing for regions where the spectral intensity recorded without MW irradiation is too weak and for regions that are NMR silent due to the peptide labelling pattern.

Lipid composition and Aβ40 labelling pattern		DNP enhancement
	Aβ40-to- lipid ratio	“CO” 180-165 ppm	“Cα” 57-48 ppm	“Cβ” 38-44 ppm	“Calkyl” 12-20 ppm	“Caromatic” 130-140 ppm	“Lipid- CH2”38- 32 ppm	“Glycerol”, 80-58 ppm
preincorporated Aβ40
POPC/POPG
F19, L34, G38	1:20	40	33	33	40	40	46	-
F20, A21, G29, V36	1:100	-	-	-	-	-	22	19
F20, A21, G29, V36	1:200	-	-	-	-	-	53	54
POPC/POPG/cholesterol/sphingomyelin/ GM1
G5, K16, A21, V24, S26, M35	1:30	24	20	20	-	-	38	34
F19, L34	1:150	-	-	-	-	-	57	34
externally added Aβ40
SPMVs
F19, L34, I32, A21	1:10	-	-	-	-	-	87	137
POPG 0h incubation
F19, L34, I32, A21	1:40	20	20	41	20	20	20	31
POPG 8h incubation
F19, L34, I32, A21	1:40	32	21	35	20	24	16	31

Preincorporation of Aβ₄₀ in liposomes.

Model liposomes were composed of either: 1) 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC) and 1-palmitoyl-2-oleoyl-sn-glycero-3- phospho-(1’-rac-glycerol) (POPG) with 3:1 molar ratio; or 2) POPC/POPG/cholesterol/sphingomyelin/ganglioside GM1 with 1:1:1.33:1:0.1 molar ratio. Lyophilized Aβ₄₀ was dissolved at about 0.25 mg/ml in hexafluoroisopropanol (HFIP) to form a clear solution, that was then dried under a stream of N₂ gas. Finally, the appropriate amount of the dried peptide was redissolved in a chloroform solution containing ~10 mg of a lipid mixture. The total peptide weight there varied from ~2.5 mg in a 1:20 Aβ₄₀-to-lipid ratio sample, down to ~0.25 mg in a 1:200 ratio sample. Then, chloroform was removed using a N₂ gas stream and overnight application of vacuum. The resulting dry film was hydrated using phosphate buffer (10 ml, 10 mM, pH 7.4) up to a concentration of [Aβ₄₀]=25 uM, which is sufficiently small to prevent a formation of aggregates over the course of sample preparation. The rehydrated suspension was agitated using a shaker for 1 h at room temperature and then subjected to 5 cycles of freezing in liquid N₂ and thawing at room temperature to produce homogeneous liposomes. The suspension was then centrifuged (26000 rpm, F1010 rotor Beckman Coulter Inc.) and the remaining supernatant was removed to give a wet pellet, which hydration level is calculated based on the dry mass of the lipid and protein film used to prepare it. ¹³C-depleted glycerol-d8 (99.95% ¹²C, Cambridge Isotopes Inc.) was used as a glassing agent in the DNP experiments. Smalle aliquots of a mixture of ¹³C-depleted glycerol-d8/D₂O/H₂O (60:30:10 wt%) stock solution containing AMUPol were directly added to the membrane pellet. After adding each aliquot, the sample was stirred for 2 min using a vortex mixer, and for a further 5 min after the final aliquot. The excess of water was removed from the pellet by drying under vacuum to attain ~40% hydration and [AMUPol]~10 mM. The sample was loaded into a 3.2 mm sapphire MAS rotor, flash frozen and stored in liquid N₂ until the time of DNP-ssNMR measurements.

External addition of Aβ₄₀ to synaptic plasma membrane vesicles (SPMVs).

Previously published protocols were used to isolate SPMVs from the brain tissues of 12-month old rats (Cheng et al., 2020). The SPMVs stock solution in 4 mM HEPES buffer at 755.2 μM total lipids concentration was quantified using ³¹P solution NMR spectroscopy. About 0.5 mg of lyophilized Aβ₄₀ was dissolved in 1 ml HFIP and sonicated in a water bath for 5 minutes to obtain a clear solution. HFIP was removed by applying a N₂ gas stream and overnight drying under vacuum. The resulting peptide film was resuspended in a solution of [NaOH]=60 mM (pH~12) to a concentration of [Aβ₄₀]= 200 μM. The Aβ₄₀ aliquots were diluted in 4 mM HEPES buffer (pH~7.4) and mixed with SPMVs stocks to obtain a final concentration of [Aβ₄₀]=10 μM with Aβ₄₀-to-lipid ratio of 1:10. After 2 minutes of stirring in a vortex mixer, the solution containing Aβ₄₀ with SPMVs was incubated quiescently for 48 hours at 37°C. The lipid and peptide material was pelleted by ultracentrifugation (80 000 rpm, TLA-100 rotor Beckman Coulter Inc.) and was then used to prepare the DNP MAS sample according to a procedure already described for preincorporated Aβ₄₀.

External addition of Aβ₄₀to POPG liposomes.

Two aliquots of POPG (11.56 × 10⁻³ mmol, 8.6 mg) were dissolved in chloroform and sonicated in a water bath for 5 min. Chloroform was removed by applying a N₂ gas stream and overnight drying under vacuum. The remaining lipid film was rehydrated using 10 mM phosphate buffer (pH 7.4) up to the lipid concentration of 1 mM. The resulting suspension was agitated for 1 h at room temperature followed by 10 freeze thaw cycles and 10 cycles of extrusion with 300 nm pore size membranes. A solution of 1.25 mg Aβ₄₀ dissolved in 155 μL DMSO was added to each of the aliquots of POPG to give a Aβ₄₀-to-lipid ratio of 1:40. The first aliquot was immediately centrifuged (26000 rpm, F1010 rotor Beckman Coulter Inc.) to produce a pellet, while the other was centrifuged after quiescent incubation at 37 °C for 8 h. The DNP MAS sample was the prepared according to a procedure already described for preincorporated Aβ₄₀.

DNP-enhanced ssNMR measurements.

DNP-enhanced solid state NMR experiments were performed using a commercial DNP system (Bruker BioSpin Inc.) equipped with a 14.1 T magnet (600 MHz ¹H Larmor frequency), Avance III solid-state NMR spectrometer, 3.2 mm MAS probes and a 7.2 T gyrotron as a source of microwave (MW) irradiation at 395 GHz. The MW irradiation is delivered to the NMR probe via corrugated waveguides of about 4 m total length. In this work the power output of the gyrotron was set to 11 W (out of a maximum of 17 W) to avoid excessive heating of the NMR sample. The sample temperature is maintained by a cold N² gas used for cooling, MAS driving and MAS bearing supplied by a chiller unit. The temperature was set at ~100 K in all measurements, according to sensors installed in the MAS probe, however, the actual sample temperature may be at least ~105 K, as shown using a measurement with a MAS rotor filled with KBr (Thurber and Tycko, 2009). The NMR sample placed in a sapphire rotor was stored in liquid N₂ prior the measurements. For loading into the NMR probe the rotor was quickly cleaned using a lint free tissue to remove the condensation on the rotor walls, loaded into a sample catcher and inserted into the probe using a pressurized N₂ gas line connected to the chiller unit. The ¹H radio frequency fields for SW_f-TPPM decoupling were ~90 kHz (Thakur et al., 2006; Vinod Chandran et al., 2008),¹H-¹³C cross-polarization (CP) used radio-frequency fields of 72 kHz for ¹³C and a ramp of 78–82 kHz for ¹H, MAS spinning speed was typically set at 8.5 kHz.

The collected NMR data were processed using Bruker Topspin and nmrPipe (Delaglio et al., 1995) software, the peak analysis was done using Sparky (Goddard and Kneller, 2004). All chemical shifts were measured with respect to tetramethylsylane (TMS). Two-dimensional (2D) ¹³C-¹³C double quantum-single quantum (DQ-SQ) spectra were collected using POST-C7 dipolar recoupling sequence having a duration of 471 μs at 8.5 kHz MAS speed (Hohwy et al., 1998). Two-dimensional ¹³C-¹³C single quantum-single quantum spectra were recorded with DARR mixing (Takegoshi et al., 2001). For probing the long-range ¹³C-¹³C contacts the DARR mixing time was set to 2 s in accordance with the magnetization exchange rates reported in earlier DNP-ssNMR experiments with Aβ₄₀ (Potapov et al., 2015). The signal averaging times in 2D experiments varied in a range of 8-55 hours depending on the sample.

3. Results and discussion

Signal-to-noise and resolution

Under conditions of DNP, NMR signal strength depends on the DNP enhancement (measured as a ratio of signal intensity with and without MW irradiation) and several other factors such as temperature (Bouleau et al., 2015), paramagnetic “bleaching” (Takahashi et al., 2012; Vitzthum et al., 2011) and nuclear depolarization (Mentink-Vigier et al., 2015, 2012; Thurber and Tycko, 2012). Despite this complexity, for comparing similar samples (i.e. the same polarizing agent, glassing agent, deuteron/proton content) DNP enhancement alone can be used as a rough figure of merit of the overall signal strength. The DNP enhancements for all the samples in this work were calculated from the ¹³C-CP spectra recorded with and without MW irradiation. Figure S1A,B of the Supplementary material shows one example of such spectra for a sample of 1:20 Aβ₄₀to-lipid in POPC/POPG. All the DNP enhancements for samples with lipids vary in a range of ε~16…87 as shown in the summarizing Table 1. In contrast, the DNP enhancements approach a factor of ~130 in a sample of uniformly ¹³C,¹⁵N-labelled arginine with 10 mM AMUPOL in the glycerol/water matrix (¹³C-depleted glycerol-d8/D₂O/H₂O 60/30/10 % wt) measured using the same setup (data not shown).

Several factors may be responsible for the overall lower DNP enhancements in samples with lipids. First, a rather fast intrinsic nuclear relaxation may compete with DNP. Such fast nuclear relaxation can be caused by thermal motions of the protonated methyl groups present in the Aβ₄₀ peptide and lipids. The enhancements usually become larger in samples that had their methyl groups deuterated (Akbey et al., 2010; Lumata et al., 2013; Potapov et al., 2015; Zagdoun et al., 2013). However, DNP enhancements measured in protonated and deuterated DMPC lipids were previously found to be almost the same for several tested polarizing agents including AMUPol (Salnikov et al., 2017), thereby suggesting that fast relaxation of methyl groups is not always the main cause for smaller enhancements. Second, an incomplete mixing of a lipid pellet with the glycerol/water matrix may produce non-uniform distribution of the polarizing agent molecules. This, in turn, may lead to some parts of the sample being unenhanced by the DNP (Liao et al., 2016; Rossini et al., 2012). Finally, the polarizing agent molecules may be distributed non-uniformly due to their affinity to lipid membranes (Jakdetchai et al., 2014; Liao et al., 2016; Salnikov et al., 2017). The clustering of polarizing agent molecules near the lipid surface may affect the electron relaxation time and thereby may lead to a poorer saturation of the electron spin transitions.

The DNP enhancements in this work vary by a factor ~30% as shown by repeat measurements using a test sample that contains POPC/POPG 3:1 mol/mol without Aβ₄₀ and that was prepared according to the preincorporation procedure described in the “Methods” section. Such variance in general agrees with previous reports: in particular, it has been observed upon freeze-thaw cycles in lipid samples with the same glycerol/water cryoprotecting matrix (Fernández-de-Alba et al., 2015). There, the variance was attributed to the destabilization of a lipid bilayer in the presence of glycerol, however, DNP enhancements may vary even without any lipids (Leavesley et al., 2018) due to a polymorphism of the glass matrix upon freezing.

Given the variance of the DNP enhancements observed in our experiments, the trends in enhancements vs the lipid composition cannot be confidently established. However, such trends were previously shown to exist for many types of polarizing agents including AMUPol (Salnikov et al., 2017). While the detailed mechanism for this is unknown, specific interactions of the polarizing agent with lipids, inhomogeneous partitioning of the polarizing agent and residual molecular dynamics are the most likely factors. In particular, DNP enhancements in POPC lipids obtained with various polarizing agents were found to be consistently larger than those in POPG lipids. This earlier finding agrees in principle with our observations that demonstrate the lowest DNP enhancements of ε~16…20 for the lipid hydrocarbon chain of POPG.

The largest DNP enhancement of ε~87 is observed in a sample of SPMVs. This enhancement most likely arises due to a small size of the sample (~10 ul out of ~30 ul available in the standard 3.2 mm MAS rotor) that is positioned in the centre of the MAS rotor and that is subjected to an effectively stronger MW field (Nanni et al., 2011; Rosay et al., 2016). This demonstrates that DNP is applicable to size-limited samples despite a loss in filling factor.

As shown in Table 1, the ¹³C-CP DNP enhancements vary across different spectral regions. Although quantitative modelling of DNP enhancements in these model systems is difficult, qualitatively the differences may be attributed to several factors reflecting inhomogeneities in the samples:

• fast relaxing methyl groups act as polarization sinks producing polarization gradients around them. A polarization gradient preferentially directed along the lipid chain may be produced because of the alignment of lipid molecules in the lipid bilayer.

• polarizing agent molecules also produce a polarization gradient around them as was previously confirmed in simulations (Mentink-Vigier et al., 2017; Wiśniewski et al., 2016). A polarization gradient along the lipid chain may be produced if such molecules are distributed non-uniformly along the depth of a lipid bilayer.

• different density of ¹H and ²H nuclei along the length in the lipid bilayer (Carmieli et al., 2006) leads to a gradient of the spin diffusion rates, further influencing the contribution of the first two factors.

Since enhancements of peptide ¹³C nuclei (labelled as “Cα”,“Cβ” etc. in Table 1) and glycerol for all samples are different, the Aβ₄₀ most likely binds to the lipid bilayer. In addition, peptide enhancements also differ from the enhancements of the lipid chain (“lipids-CH₂” in Table 1), which suggests that the peptide has some specific position with respect to the headgroup and hydrophobic regions of the lipid bilayer.

The DNP enhancements of nuclei in Aβ₄₀ are significant enough for obtaining good quality 2D spectra. Double-quantum-single-quantum (DQ-SQ) ¹³C-¹³C POST-C7 correlation spectra (Hohwy et al., 1998), collected for all the samples listed in Table 1, enable the assignment of most aminoacid residues. The resolution in DQ-SQ spectra in principle allows assigning up to 6 labelled residues, however in most samples only 4 labelled residues were used to avoid ambiguities. Figure 1A shows the 2D DQ-SQ ¹³C-¹³C correlation spectrum of Aβ₄₀ preincorporated in POPC/POPG lipids at Aβ₄₀-to-lipid ratio of 1:100. DQ evolution suppresses large contribution of the natural abundance ¹³C signals, which in conventional SQ-SQ ¹³C-¹³C correlation spectra produce strong diagonal peaks and t1 noise masking rather weak peptide cross-peaks. Figure 1B shows slices through the 2D spectrum in Figure 1A done at the DQ frequencies corresponding approximately to the combined chemical shift frequencies of pairs CO and Cα, Cα and Cβ, Cβ and Cγ in V36 residue. Slices in Figure 1B illustrate the actual signal-to-noise in the 2D spectrum and the extent of inhomogeneous line broadening. The ¹³C line full width at half height (FWHH) for all the studied samples vary in the intervals of 2.2…6.6 ppm for carbonyl carbons (average value 4.2 ppm), 1.6…5.1 ppm for Cα (average value 3.0 ppm). 1.8…5.6 ppm for Cβ (average value 3.2 ppm). These linewidths are comparable to the ones previously found using DNP-ssNMR in monomers and globular oligomers (¹³C line FWHH of 4.4–7.4 ppm) and protofibrils (¹³C line FWHH of 3.0–5.2 ppm) of Aβ₄₀ (Potapov et al., 2015). In contrast, the linewidth of Aβ₄₀ in lipids observed in our work is noticeably broader than in mature amyloid fibrils (¹³C FWHH of 2.4–3.2 ppm) (Potapov et al., 2015), which reflects a greater degree of disorder in the former compared with the latter. Overall, the resolution in our spectra of Aβ₄₀ interacting with a lipid bilayer demonstrates that DNP-ssNMR technique is best suited for capturing primarily the substantial conformational changes.

Figure 1.

(A) DNP-enhanced 2D P0ST-C7 DQ-SQ spectrum of Aβ₄₀ preincorporated in POPC/POPG liposomes at Aβ₄₀-to-lipid ratio 1:100. Total experimental time ~18 hours. (B) 1D slices made at DQ chemical shifts for pairs of CO and Cα , Cα and Cβ, Cβ and Cγ of V36 residue.

The obtained sensitivity enables DNP-enhanced measurements with preincorporated Aβ₄₀ at ratios as low as 1:200 for POPC/POPG liposomes and 1:150 for POPC/POPG/cholesterol/sphingomyelin/ganglioside GM1 liposomes (see Figure S2 of the Supplementary material). While most cross-peaks can be resolved and assigned, the spectral quality for such quantities of peptide suffers due to t1 noise. This noise arises due to a slight variation of the intensity of the natural abundance peaks caused by some slight changes of the sample temperature.

Since the 1:200 Aβ₄₀to-lipid sample contains only about ~0.25 mg of Aβ₄₀, DNP can also be helpful for other ssNMR measurements in samples with a limited amount of Aβ₄₀ peptide. For example, a sample of SPMVs contains ~1 mg of lipids (compared to preincorporated Aβ₄₀ containing ~10 mg of lipids), therefore, at a nominal Aβ₄₀to-lipid ratio of 1:10 the total amount of peptide is rather small (~0.5 mg) even assuming 100% binding of the peptide to liposomes. In practice, the intensity of the peptide NMR signals in the spectra with SPMVs is somewhat lower than expected for this amount of Aβ₄₀. To estimate Aβ₄₀to-lipid content we use the ratio of Aβ₄₀ aromatic residue intensity to the total spectral intensity outside glycerol region in ¹³C-CP spectra. This ratio is almost the same for externally added Aβ₄₀ in SPMVs (nominal 1:10 Aβ₄₀to-lipid) and for preincorporated Aβ₄₀ in POPC/POPG (1:20 Aβ₄₀to-lipid), which spectra are shown in Figures S1C and S1A respectively. This similarity in intensities can be explained by an incomplete binding of Aβ₄₀ to SPMVs and therefore the actual Aβ₄₀to-lipid ratio in the sample with SPMVs can be estimated as ~1:20 and the total amount of Aβ₄₀ is ~0.25 mg.

Chemical shifts

Chemical shifts of ¹³C nuclei report on the conformation of the Aβ₄₀ peptide inserted into a lipid bilayer. The peptide secondary ¹³C chemical shifts (i.e. differences from the random coil values) differ by as much as 3 ppm from one another (see Table S1). While this variation is somewhat greater than previously seen in fibrillar structures of different morphologies, the secondary shifts at many labelled sites of Aβ₄₀ follow the pattern typical for β-sheets: the secondary chemical shifts are negative for CO and Cα and are positive for Cβ. In particular, such β-sheet-like pattern is confirmed for:

• F19 and L34 residues of Aβ₄₀ in POPG/POPC (Aβ₄₀to-lipid ratio 1:20) and POPC/POPG/cholesterol/sphingomyelin/GM1 (Aβ₄₀to-lipid ratio 1:150)

• F20, A21, V36, G29 residues of Aβ₄₀ in POPG/POPC (Aβ₄₀to-lipid ratio 1:100 and 1:200)

• K16, A21, M35 residues of Aβ₄₀ in POPC/POPG/cholesterol/sphingomyelin/GM1 (Aβ₄₀to-lipid ratio 1:150)

• A21, I32, L34 residues of Aβ₄₀ in POPG (incubated for 0 h and 8 h)

• F19, A21, I32 residues of Aβ₄₀ in SPMVs.

These findings are consistent with many structural models of Aβ₄₀, where β-sheet region span residues 11-23 and 31-40 (Bertini et al., 2011; Paravastu et al., 2008; Petkova et al., 2002; Qiang et al., 2012). Our results show that the trend for having β-sheet-like conformation does not depend on the type of lipids and concentration of Aβ₄₀ peptide. The only exception from this pattern is residue L34 of Aβ₄₀ in SPMVs, however, non-β-sheet conformation at this site was also observed in Aβ₄₀ fibrils from human brain tissue (Lu et al., 2013) and in Aβ₄₂ fibrils (Wälti et al., 2016).

While the chemical shifts at many sites have β-sheet-like character, they provide only a coarse-grain view of the conformation, because the conformation at unlabelled sites is unknown. In fact, many previous reports suggest that various components of lipid bilayers (i.e. phospholipids, cholesterol, ganglioside GM1 etc.) have their own specific effect on the Aβ₄₀ secondary structure producing conformations rich in either α-helices or β-sheets (Niu et al., 2018; Williams and Serpell, 2011). In particular for POPG liposomes studied in our work, the α-helical content of Aβ₄₀ is expected to reach ~20% (for 1:40 Aβ₄₀to-lipid ratio)(Terzi et al., 1997), whereas our results show no signature of α-helical structures.

Hydrophobic contacts

More details about the peptide conformation can be obtained from DNP-ssNMR measurements reporting on inter-residue contacts. Two-dimensional ¹³C-¹³C correlation spectra with 2 s DARR mixing in a sample of Aβ₄₀ externally added to POPG vesicles (Aβ₄₀to-lipid ratio 1:40) reveal the presence of cross-peaks between the aromatic group of F19 and aliphatic groups of L34 (~25 ppm), and between the same aromatic group and the methyl groups of I32 (~15 ppm) (Figure 2A). The 1D slices, made through F19-L34 and F19-I32 cross-peaks in Figure 2A, demonstrate the level of signal-to-noise (see Figure 2B). Even after a moderate signal acquisition time of ~11 h the signal-to-noise at these cross-peaks is high enough for DNP-ssNMR measurements in samples with Aβ₄₀to-lipid ratios lower than 1:40.

Figure 2.

(A) DNP-enhanced 2D ¹³C-¹³C correlation spectrum (2 s DARR mixing) of Aβ₄₀ externally added to POPG liposomes without incubation. Total experimental time ~ 8 h. (B) 1D slices made at chemical shifts 15 ppm (green) and 25 ppm (red). Dashed lines in (A) show the slice positions on a 2D spectrum. (C) Schematic diagram of the contacts observed in Aβ₄₀ externally added to POPG liposomes.

The presence of F19-L34 and F19-I32 cross-peaks shows that F19 is located in proximity of L34 and I32 residues. The intensity of the cross-peaks is the same for the sample with and without 8 h of incubation (see Figure S2). The F19-L34 contact has also been previously detected in many Aβ fibrils (Bertini et al., 2011; Lu et al., 2013; Paravastu et al., 2008; Qiang et al., 2012), except the Aβ₄₂ fibrils (Wälti et al., 2016), fibrils of Aβ₄₀ formed in phospholipid vesicles (Niu et al., 2014) and fibrils reported by Petkova et al., 2002. As shown by previous DNP-ssNMR measurements the same contact is present to varying degrees in: a) metastable protofibrils; b) fibrils formed from protofibrils upon their further conversion; c) oligomers, formed at elevated concentrations of Aβ₄₀; d) and monomers formed at high pH. The F19-I32 contact is somewhat less common: while some fibrillar models have it (Paravastu et al., 2008; Qiang et al., 2012), there are others that do not. Specifically, the model of Aβ₄₂ fibrils (Wälti et al., 2016), the model of Aβ₄₀ fibrils by Bertini et al., 2011, the models of Aβ₄₀ fibrils formed in human brain tissue (Lu et al., 2013) and phospholipid vesicles (Niu et al., 2014), show F19 and I32 to be too far apart from one another to provide a cross-peak. In general, the presence of F19-I32 and F19-L34 is be consistent with both L34 and I32 residues facing one side of an extended peptide strand as shown schematically in Figure 2C.

Interestingly, that neither the intensity of F19-L34 and F19-I32 cross-peaks (see Figure S2), nor the Cα, Cβ and CO chemical shifts of labelled residues of Aβ₄₀ in POPG (see Table S1) change upon incubation time. On the other hand, previous studies have indicated that after binding Aβ₄₀ inserts into lipids on a timescale of 3 h (Terzi et al., 1997), so that after 8 h of incubation the peptide should have already undergone some structural rearrangement detectable by circular dichroism. One possible explanation for the discrepancy between the results of our DNP-ssNMR measurements and the previous circular dichroism study, may be that structural changes of Aβ₄₀ are taking place at the sites that were not ¹³C-labelled. Testing this hypothesis requires measurements of Aβ₄₀ with different labelling scheme, however, that goes beyond the scope of this work.

4. Conclusions

This work demonstrates for the first time the feasibility of DNP-ssNMR measurements to characterize Aβ₄₀ peptide interacting with the cellular membrane models. The results demonstrate that informative 2D ssNMR experiments probing Aβ₄₀ conformation can be carried out in samples containing only about ~0.25 mg of the peptide at Aβ₄₀-to-lipid ratio as low as 1:200. DNP provides almost an order of magnitude improvement in sensitivity compared to the previous room temperatures ssNMR studies, in which the Aβ₄₀-to-lipid ratios were limited to 1:30 (Cheng et al., 2018). Therefore, DNP enables ssNMR measurements in systems where they were earlier precluded due to a limited amount of available peptide or lipids, in situations where the peptide binding is incomplete, or the required ₄₀to-lipid ratio is low. In addition, due to the cryogenic nature of the DNP experiments, they are a promising tool for reporting on transient species emerging in the process of Aβ peptide insertion into the lipid bilayer. DNP-enhanced ssNMR may therefore provide an atomic-level resolution picture of the kinetics and structural changes of Aβ₄₀ that cannot be obtained using other biophysical techniques.