Introduction

Alzheimer’s Disease (AD) is a debilitating neurodegenerative disease characterized by extracellular plaques of Amyloid-β (Aβ) peptides and intracellular neurofibrillary tangles of tau1,2. It is widely recognized that the ϵ4 allele of apolipoprotein E (apoE) is the strongest genetic risk factor for late-onset AD3,4,5, while apoE3 is considered normal and apoE2 is protective6,7. The molecular mechanism underpinning the pathological role of apoE4 in AD is still unclear, but numerous evidence points to the role of apoE4 in accelerated deposition of Aβ in the brain compared to the other isoforms of apoE8,9,10. For example, positron emission tomography (PET) imaging using amyloid tracer Pittsburgh compound B (PiB) showed that the burden of Aβ deposits in the brain, even in cognitively normal individuals, was correlated with apoE genotype11,12,13. The order of amyloid burden was found to be ε4 homozygotes > ε4 heterozygotes > ε4 noncarriers10,14,15. By introducing apoE inducible mouse model, a study showed that the effect of apoE4 on amyloid deposition in the brains of the amyloid model mice was highest when it was overexpressed in the seeding phase of amyloid development16. Moreover, apoE4 is found to increase the levels of both soluble and insoluble Aβ aggregates in PDAPP/TR mice at a relatively early stage compared to the other isoforms of apoE17. The pathogenic effects of apoE4 in AD are believed to occur via both impairing the clearance10,18 and promoting the aggregation of Aβ in vivo16,19. However, the molecular mechanism of the apoE-Aβ interaction remains largely unknown20.

In vitro, apoE isoforms interact with Aβ and influence aggregation of Aβ21,22,23,24. Most of these studies have shown that the apoE isoforms delay the fibrillization of Aβ, possibly via interacting with the oligomers and the fibrillar intermediates of Aβ. The structure of the apoE-Aβ complex is not known, but crosslinking studies suggest that both the N-terminal and C-terminal domains of apoE are involved in the interactions25. Using Forster Resonance Energy Transfer (FRET) and fluorescence correlation spectroscopy (FCS), we have previously shown that apoE interacts poorly with the monomers but with high affinity with the oligomeric forms of Aβ via multivalent interactions23. While the effect of apoE on the aggregation of Aβ has been studied by various groups, the effect of apoE on the microscopic reaction mechanism of the fibrillization of Aβ has not been investigated yet.

The canonical pathway of self-assembly of soluble monomeric Aβ to the insoluble fibrils consists of three essential micro steps: an initial slow process of primary nucleation, growth of the nuclei via elongation and proliferation of the fibrils via secondary nucleation (Fig. 1A)26,27. We and others have previously shown that all three isoforms of apoE delay fibrillization of Aβ21,22,23. However, the effect of apoE on the individual reaction steps of Aβ aggregation is not known (Fig. 1A). Moreover, the isoform specificity of the apoE-Aβ interaction is still unclear due to a lack of quantitative estimation of the rate constants of the fibrillization of Aβ in the presence and absence of the isoforms of apoE.

Fig. 1: Isoform-dependent effect of apoE on the elongation of the Aβ42 fibrils.
figure 1

A Schematic representation of the important reaction steps involved in Aβ42 aggregation. Aggregates are formed by primary nucleation or introduced through sonicated preformed fibrils. Fibril mass increases by monomer addition to the fibril ends or by replication through secondary nucleation and elongation. B Single fibril growth was monitored in TIRFM (left), and growth rates were monitored by tracking (Cyan) of single fibrils (right). Extreme right panel: fast-growing end is colored with red and slow-growing end with blue to show polarised growth of fibrils. (Data from seeded growth of 2 µM Aβ42). C Aggregation of 2 µM Aβ42 in 1x PBS pH 7.4, 25 ⁰C, in presence of 1 nM monomer equivalent seeds with no apoE (top 0 h, 3 h, 10 h); 150 nM apoE4 (middle 0 h, 6 h, 10 h); and 150 nM apoE3(bottom 0 h, 6 h, 10 h) monitored in real-time in TIRFM. scale bar 5 µm.D Tracking time traces of single fibril growth with no apoE(black), in presence of 150 nM apoE4(red) and with 150 nM apoE3(blue). The rate of elongation is slowed down when apoE is present. E The rate of elongation of the fibrils in absence and presence of apoE2, apoE3 and apoE4. Each dot reresents a single fibril. The error bar represents mean ± SD; *P < 0.05, **P < 0.01 and ***P < 0.001 (two-sample Student’s t test)) (n = 127 for Aβ42 only; n = 92, n = 121, n = 113 for apoE4, apoE3 and apoE2 containing samples respectively). F Histogram of the fibril length of Aβ42 with no apoE at t = 10 h. Panels (G) and (H) show the distribution of the fibril length at t = 15 h in presence of 150 nM apoE4 and apoE3 respectively. Fitting with a Gaussian (red) shows a shift in the centroid fibril length from 15.8 µm in Aβ42 (F) to ≈7.5 µm in Aβ42+apoE4 (G) and ≈ 2μm in Aβ42+apoE3 samples (H).

Full size image

We set out to investigate apoE-Aβ interactions by monitoring the growth of the Aβ fibrils in real-time with single fibril resolution by using Total Internal Reflection Fluorescence Microscopy (TIRFM) equipped with an autofocus system28. We have previously used this approach to estimate the rate constants of elongation and secondary nucleation of the Aβ fibrils at the level of single fibrils28. Here, we report that apoE inhibits elongation of the Aβ42 fibrils in an isoform-dependent manner, with apoE2 and apoE3 possessing the strongest inhibitory effects. Furthermore, using an array of single-molecule fluorescence techniques such as single-molecule colocalization, Stochastic Optical Reconstruction Microscopy (STORM), and single-molecule photobleaching assay, we find that oligomers of apoE bind extensively both to the surface and the ends of the fibrils in an isoform dependent manner. We observed that apoE3 binds preferentially to the growing ends of the fibrils, whereas the binding of apoE4 is less specific. Our results provide mechanistic insights into how apoE can modulate the early events of aggregation of Aβ42 in AD.

Results

The widely recognized mechanism of fibrillization of Aβ42 involves primary nucleation, elongation, and secondary nucleation (Fig. 1A)27,29. Here, we used TIRFM to visualize the fibrils directly 28 and investigated the effects of the apoE isoforms on the growth of the single fibrils of Aβ42 (Fig. 1B). Thioflavin T (ThT) has been used as the reporter dye to visualize the fibrils, thereby eliminating the need for covalent linking of a fluorophore to the Aβ42 peptide. In most of the experiments described below, we have used 2 µM Aβ42 added to 1 nM (monomer equivalent concentration) preformed fibrillar seeds of Aβ42. The addition of the seeds bypasses the lag phase associated with the primary nucleation. Such a low concentration of seeds was chosen to ensure clear visualization of the growth of the individual fibrils in TIRFM (See SI, for further discussion on the seeds, and Fig. S1). The apoE proteins used here were lipidated with 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) (See methods, SI, Fig. S2), keeping in mind that in vivo apoE exists primarily in the lipidated forms30. Moreover, the automation of the imaging of the fibrils as a function of time allowed us to increase the sampling size by 10- 100 fold than the previously reported studies31,32,33.

Isoform-dependent effect of apoE on the elongation of the Aβ42 fibrils

First, we recorded videos of the growth of the fibrillar seeds upon addition of 2 µM monomeric Aβ42 in the absence or presence of 150 nM lipidated-apoE4 or apoE3 or apoE2 using TIRFM for several hours at discrete time points over multiple fields of view to increase the fibril statistics. The videos (Movie S1) revealed that in the absence of apoE, Aβ42 fibrils grew rapidly to about ~45 µm in 10 h. However, the addition of 150 nM apoE slowed down the growth of the fibrils considerably (Fig. 1C). In order to quantify the effect of apoE on the rate of elongation of the fibrils, we used a semiautomated approach to track the growth of the individual fibrils, similar to what we have used in our earlier work (see methods, SI) (see Fig. 1B, Movie S1, for fibril tracking)28. Briefly, first the images in the subsequent frames were stabilized, and then the length of the individual fibrils was tracked using the “Neuron J” plugin in ImageJ. Tracking of the fibrils shows that fibrils grow following the ‘stop and go’ model, as we and others have reported previously28,33 (Fig. 1D). Moreover, we found that after each pause, the rate of increase in length is higher in Aβ42-only samples compared to apoE-containing samples (Fig. S3A, B). In the apoE-containing samples, the pause frequency increases slightly; for Aβ42, the value is 0.15 ± 0.01 \({{\rm{pauses}}}\) \({{{\rm{hr}}}}^{-1}\); for 150 nm apoE4, apoE3, and apoE2 the values are 0.19 ± 0.01, 0.18 ± 0.01and 0.18 ± 0.01 \({{\rm{pauses}}}\) \({{{\rm{hr}}}}^{-1}\) respectively (P < 0.05) (Fig. S3C). However, an increase in pause duration in the presence of apoE is more evidently notable; for example, Aβ42 (1.3 ± 0.14 h) < apoE4 (2.01 ± 0.14 h) < apoE3 (3.07 ± 0.24 h) ~ apoE2 (3.14 ± 0.24 h) (Fig. S3D). The elongation rates of the individual fibrils are measured from the rate of increase of the contour length of the fibrils (Fig. 1D). The average elongation rates are estimated to be 0.73 ± 0.32, 0.25 ± 0.13, 0.06 ± 0.03, and 0.07 ± 0.04 \({{\rm{nm}}}\) \({\sec }^{-1}\) in the absence of apoE and the presence of apoE4, apoE3, and apoE2, respectively (Fig. 1E). The measured growth rates are found to be of a similar order to those published previously using the Thermophoretic trap technique34. Moreover, we found that there is considerable heterogeneity in the growth rate of the fibrils both in the absence and in the presence of apoE (Figs. 1E,S4B–E); despite that, differences between no apoE, apoE4, and apoE3 (or apoE2) are significant (p < 0.001). However, the difference between apoE3 and apoE2 is not significant (Fig. 1E). Therefore, all the three isoforms of apoE delay the elongation of the fibrils of Aβ42. However, apoE3 and apoE2 exhibit about 4 -times higher inhibitory effects than apoE4.

The effects of the apoE isoforms on the elongation of the fibrils can also be quantified by analyzing the distribution of the lengths of the fibrils at a specified time point. The histogram of the lengths of the fibrils (n ~ 100) at t = 10 h showed that in the absence of apoE, the Aβ42 fibrils grow to an average length of about 15.8 ± 0.6 µm (Fig. 1F). However, in presence of apoE4, apoE3, and apoE2, even at t = 15 h, the fibrils were found to be considerably shorter (Fig. 1C), with the centroids of the distributions observed at 7.5 ± 1.0, 1.9 ± 0.3, and 2.7 ± 0.1 µm, respectively (Figs. 1G, H and S4A). Measurement of the rates of elongation in presence and absence of apoE at different time points showed that the growth rates remained largely conserved during the growth phase of aggregation (Fig. S4F–I), consistent with the results reported earlier28. Next, we opted for measurements on a larger number of fibrils (n\(\sim\) 1000) that were achieved by using the “Ridge Detection” plugin in ImageJ (Fig. S4J, see methods, SI). The mean length of the Aβ42 only fibrils at t = 1.5 h was measured to be 3.9 ± 1.9 μm. However, in presence of apoE4 and apoE3, the mean lengths, even at t = 3 hr were found to be 2.3 ± 1.2 μm and 0.6 ± 0.3 μm, respectively (Fig. S4K). Therefore, the elongation rates measured from the snapshots of the fibrils at the above-mentioned time points are 0.7 ± 0.3 \({{\rm{nm}}}\) \({\sec }^{-1}\) for Aβ42 only and 0.2 ± 0.1 \({{\rm{nm}}}\) \({\sec }^{-1}\) and 0.05 ± 0.03 \({{\rm{nm}}}\) \({\sec }^{-1}\), respectively, for the apoE4 and apoE3 containing samples (Fig. S4L). These values are in agreement with the elongation rates measured from tracking of the single fibrils. The results presented here are consistent with the observations reported by Fagan et al.35. These authors have shown that Aβ deposition in the APPV717F+/− model mice is delayed in presence of apoE in an apoE isoform-dependent manner (No apoE > apoE4 > apoE3 > apoE2).

Direct observation of secondary nucleation at single fibril level in the presence and absence of the apoE isoforms

Next, we focus on the characterization of the rate of formation of secondary nuclei of the Aβ42 fibrils. Aβ42 aggregation kinetics can gain rapid acceleration through the fibril surface-catalyzed secondary nucleation processes (Fig. 1A)27. Therefore, studying the effect of the apoE isoforms on secondary nucleation is crucial, as the inhibitors of secondary nucleation are believed to hold better promise as therapeutic agents36. Here, we used single fibril data to examine the effects of the apoE isoforms on the secondary nucleation.

Ideally, secondary nucleation leads to branching of the fibrils (Figs. 2A, S5A, Movie S2). However, in an earlier study, we reported that only a small fraction of the branches remained attached to the parent fibrils, while most of the nuclei go into the solution28. Therefore, we decided to quantify the rate of secondary nucleation by counting the secondary nuclei, i.e., the new fibrils \({N}_{2}\left(t\right)\), and normalizing it with respect to the total fibrillar mass M(t) (note that M(t) is proportional to the total contour length of the fibrils) using the relationship. \({N}_{2}\left(t\right)={r}_{2}{\int }_{{t}_{0}}^{t}M\left(\tau \right)d\tau\) (for details see Eq. S6)28, where \({r}_{2}\) is the rate of secondary nucleation. Measurement of the secondary events and the cumulative fibril mass over time ‘t’ shows that they are in agreement with each other, supporting that the new fibrils are generated primarily via secondary nucleation (Fig. 2B). We note here that if the newly detected fibrils were simply preformed fibrils or formed via primary nucleation, then the rate of appearance of such fibrils would decrease with time and be independent of the total mass of the existing fibrils28. A similar trend for the formation of the new fibrils via secondary nucleation is observed in the apoE-containing samples as well (Figs. 2C,S5B,C). In absence of apoE, the Aβ42 fibrils undergo rapid growth both in terms of the fibrillar length and the secondary nucleation events, while the samples containing apoE exhibit not only shorter but also fewer fibrils (Figs. 2A,S5A). This is also reflected in the observed slower growth in the overall fibrillar mass and number concentration (Figs. 2C, S5B,C). The presence of the shorter fibrils indicates inhibition of elongation by apoE. However, fewer fibrils in the presence of apoE do not necessarily indicate a reduced rate of secondary nucleation since the generation of the secondary nuclei is dependent not only on the associated rate constant but also on the total fibrillar length/mass (see Eqs. S4 and S6). Therefore, apoE can affect the formation of the secondary nuclei by affecting the elongation of the fibrils. To quantify this effect, the rate of secondary nucleation (r2) of the Aβ42 fibrils in the absence and presence of the apoE isoforms estimated using Equation S6 from the data presented in Figs. 2B, C and S5B, C are found to be nearly the same (r2 ~ 10−9 sec−1 hence, k2 ~ 103 M−2 sec−1) for all the samples. Therefore, our data suggest that the effect of the apoE isoforms on the secondary nucleation of Aβ42 is not significant. We note here that the value of k2 obtained from the analysis of the single fibril data is one order of magnitude less than that of the previously reported data obtained from the bulk experiments27.

Fig. 2: Direct observation of secondary nucleation in the absence and presence of the apoE isoforms and quantification of elongation properties of secondary fibrils.
figure 2

A representative TIRFM images of the secondary fibrils generated on the primary fibril surface (indicated by yellow arrow) in absence of apoE (top panel), in presence of 150 nM apoE4 (middle), and 150 nM apoE3 (bottom). Scale bar = 2 μm. Panels (B, C) show the time evolution of the integrated fibril mass \(M(t)\) (black) and the total number of new fibrils \({N}_{2}\left(t\right)\) (considered as 2⁰ events) (blue) in absence and presence of 150 nM apoE4, respectively. \({N}_{2}\left(t\right)\) increases proportionately with an increase in \(M(t)\). D, E Single fibril tracking data of the primary (black) and secondary fibrils (red) in absence and presence of the apoE3, respectively. F, G Rate of elongation of the primary and secondary fibrils in absence and presence of apoE3, respectively. Rate of elongation of each fibril is represented as a dot here. In (F, G) the error bars represent mean ± SD, *P < 0.05, **P < 0.01 and ***P < 0.001 (from Student’s t test)).

Full size image

Next, we ask whether apoE affects the growth of the secondary fibrils. This is a unique advantage of the single fibril measurements over bulk measurements, which cannot distinguish the kinetics of growth of the secondary fibrils from the primary fibrils. The secondary events observed in our experiments at/near the fibril surface are demonstrated in Figs. 2A,S5A, and in Movies S3–S6 (Aβ only, +apoE4, +apoE3 and +apoE2, respectively). Hence, we tracked the growth of the secondary fibrils in a similar way we measured the primary fibrils (Figs. 2D,E,S5D,E). Quantification of the elongation of the secondary fibrils shows that in absence of apoE, the secondary nuclei elongate at the same rate as the primary fibrils as presented in Fig. 2A (top panel), Movie S3 and quantified in Fig. 2F, consistent with the observations reported earlier28. Expectedly, all the three isoforms of apoE inhibit elongation of the secondary fibrils. However, the inhibitory effect on the secondary fibrils was ~1.6–2 times stronger than its effect on the primary fibrils (Figs. 2G,S5F,G). A greater rate of inhibition of the secondary fibrils might indicate that the secondary fibrils grown in the presence of apoE are structurally different from the primary fibrils, which are grown from the seeds that are prepared in absence of apoE. Overall, our data suggest that all the three isoforms of apoE clearly delay elongation but not the secondary nucleation of the Aβ42 fibrils. Notably, the effects of apoE on the rate of elongation are isoform-specific. ApoE may play a role in altering the structure of the nascent fibrils. We note that some variations in the intensity profile along the length of the fibrils are observed in the kinetic experiments (Fig. 2A). This may occur due to movement of the fibrils, strong depth dependence of TIRF intensity, and slight fluctuations in the autofocus system. However, this intensity variation does not affect our study in elongation and secondary nucleation characterization.

Estimation of the kinetic rate parameters from bulk measurements and comparison with single fibril data

Then, we performed bulk kinetic experiments in a plate reader to compare the rate parameters extracted from the bulk and single fibril data. Bulk fluorescence of ThT was monitored continuously from the samples prepared under identical conditions as that of the TIRFM experiments described earlier, i.e., samples containing fibrillar seeds + 2 µM soluble Aβ42 \(\pm\)150 nM lipidated-apoE (Fig. S6A). To extract the microscopic rate parameters from the normalized bulk aggregated mass concentration, we fit the data with integrated rate law (Eq S7) using AmyloFit37. As described in Eq. S7, the bulk kinetics is dependent on the primary nucleation (kn), elongation (k+), and secondary nucleation (k2). Since we used a seeded growth approach, kn is ignored here. It may be noted that the kinetics of the bulk fluorescence is more sensitive to the parameter (κ), where \(\kappa \propto ({{k}_{+}{k}_{2}})^{0.5}\) (see Eq. S8), rather than k+ or k2 individually. Hence, to fit the bulk data we kept one of the rate constants (i.e., k+ or k2) fixed while the other is allowed to be free. Figure 3A shows the kinetic data from the aggregation of 2 µM Aβ42 in the absence or presence of 150 nM apoE4, apoE3, and apoE2, and the global fits with k+ being allowed to be free but k2 kept fixed at 104 M−2 s−1. The value of k2 was chosen based on previously published results28. It may be seen that the kinetic data fit quite well (mean residual errors obtained are less than 0.001 for all the curves). The values of k+ obtained from the fitting are 2.9 × 106 M−1s−1, 7.8 × 105 M−1s−1, 2.1 × 105 M−1s−1 and 1.4 × 105 M−1s−1 for Aβ42 only, Aβ42 + apoE4 or apoE3 or apoE2 respectively (Table 1). Figure 3B shows the global fits of the same kinetic data with k+ is kept fixed at 106 M−1 s−1 but k2 is allowed to be free. In this case poor fits are observed particularly for Aβ42 + apoE3 and apoE2 samples (mean residual errors >0.006). Therefore, the bulk aggregation data are consistent with isoform-dependent inhibition of the elongation of the Aβ42 fibrils without affecting the secondary nucleation. Furthermore, to verify if the k+ values extracted from the single fibril measurements are consistent with the bulk data, we performed fitting of the bulk data with k+ fixed at 2.9 × 106 M−1s−1 for Aβ42 only and 7.8 × 105 M−1s−1, 2.1 × 105 M−1s−1, 1.4 × 105 M−1s−1 for the Aβ42 + 150 nM apoE4, apoE3 or apoE2 containing samples respectively. The values mentioned above were estimated by converting the r+ in nm/sec (e.g., 0.73 \({{\rm{nm}}}\) \({\sec }^{-1}\) for Aβ42 only, and 0.2 \({{\rm{nm}}}\) \({\sec }^{-1}\), 0.053 \({{\rm{nm}}}\) \({\sec }^{-1}\), and 0.035 \({{\rm{nm}}}\) \({\sec }^{-1}\), Aβ42 + 150 nM apoE4 or apoE3 or apoE2 samples respectively) to k+ in M−1s−1 (see Eq. S21). Figure 3C shows that global fitting of the data is consistent with the values of k+ obtained from the single fibril measurements. The values of k2 obtained from the global fitting of the data are 0.8 × 104 M−2s−1 for Aβ42 only, and 0.6 × 104 M−2s−1, 0.7 × 104 M−2s−1 and 0.4 × 104 M−2s−1 for Aβ42 + apoE4 or apoE3 or apoE2 respectively (Fig. 3C, Table 1). It may be seen that the values of k2 that were thus obtained are quite similar. The rate of secondary nucleation r2 comes out to be ~10−8 s−1, similar to the value reported earlier for Aβ42 only27,28. While in Fig. 3A–C, we presented the data using only 150 nM apoE, we also performed the experiments using 75 and 300 nM apoE (Fig. S7A–C). Fitting of the kinetic data across all three apoE isoforms with a range of apoE concentrations revealed that the secondary nucleation rate constant (k2) remains almost the same while the elongation rate constant (k+), a key determinant of overall fibril mass38, was significantly reduced in the order (apoE4 > apoE3 ~ apoE2) (Figs. S7, S13C, Table S1).

Fig. 3: Estimation of the kinetic rate parameters from bulk aggregation kinetics and comparison with the single fibril data.
figure 3

A, B, C represents bulk aggregation kinetics of 2 µM Aβ42, in 1x PBS pH 7.4 buffer in the presence of 1 nM monomer equivalent seeds at 25 °C. The symbols represent normalized ThT fluorescence from Aβ42 only (gray), and Aβ42 + 150 nM apoE4 (Blue), or apoE3 (Green), or apoE2 (Cyan) (data represents: mean ± SD, n = 3). In (A) and (B), the solid line represents the best global fit to the integrated rate law when we varied the \({k}_{+}\) keeping k2 fixed and varied \({k}_{2}\) keeping k+ fixed, respectively. Primary nucleation rate constant (\({k}_{n}\)) is set to zero. The best fit is achieved in (A). C We fixed the \({k}_{+}\) to the values obtained from the single fibril measurements in Fig. 2C with \({k}_{2}\) as free. All four curves yielded similar \({k}_{2}\) values varying between 0.8–0.4 × 104 M−2s−1. Panels(D),(E), (F), and (G) show examples of time traces of oligomer detection above a threshold (green dotted line) in the absence and presence of 150 nM apoE4, apoE3, and apoE2, respectively. Samples were taken at respective \({t}_{1/2}^{\prime} s\) of each samples, centrifuged to pellet down insoluble aggregates, and the supernatant was measured to assess the soluble oligomer population. Panel(H) shows the absence of oligomers in the lag phase, both in the absence (red, t = 40 min) and presence of apoE4 (Blue, t = 1.5 h), apoE3 (green, t = 2 h) and apoE2 (gray, t = 2 h). Panel(I) represents quantification of oligomer population in the presence and absence of 150 nM apoE isoforms at \({t}_{1/2}\). Each dot represents number of oligomers in a 30-s trace. Data represents mean ± SD,* indicates P < 0.05, ** indicates P < 0.01, and *** indicates P < 0.001 (using two-sample Student’s t test).

Full size image
Table 1 Rate constants obtained from the bulk kinetic data and single fibril data
Full size table

Moreover, agreement between the bulk and single fibril data is also reflected when a comparison is made between the ratio of the elongation rate constants, viz, (k+(Aβ))/k+(ApoE))0.5 obtained from the single fibril measurements and the ratio of the half-time, i.e. \({t}_{1/2\left({ApoE}\right)}/{t}_{1/2\left(A\beta \right)}\) obtained from the bulk data (Fig. S6B). In both the measurements, the ratios obtained are 1.0, 2.1, 4.4, and 5.3 from the bulk and 1.0, 1.7, 3.6, and 3.2 from the single fibril data for Aβ42 and apoe4, apoE3, and apoE2 containing samples, respectively.

Next, we tried to examine whether apoE affects the concentration of the soluble oligomers. To measure the soluble oligomers, we aggregated WT-Aβ42 spiked with Atto425-Aβ42 in the presence and absence of 150 nM apoE isoforms under seeded conditions similar to that of bulk aggregation kinetics. At different times, the aggregate mixture was collected and centrifuged. The soluble fraction containing both monomers and soluble aggregates was isolated from the supernatant39 (See methods, SI) and examined in a fluorescence correlation spectroscopy setup40. In this technique, the soluble oligomers, being considerably brighter than the monomers, give rise to occasional fluorescent bursts while diffusing through the confocal volume39. It has been previously shown that the oligomers of Aβ42 could be generated via secondary nucleation, which reaches maximal rate near half-time (\({{{\rm{t}}}}_{1/2}\)) of the aggregation kinetics29. Hence, we quantified the oligomers of the Aβ42, Aβ42+apoE4, Aβ42+apoE3, and Aβ42+apoE2 samples at their respective \({{{\rm{t}}}}_{1/2}\) time points, i.e., 2 h 35 min, 6 h 25 min, 11 h 22 min and 13 h 30 min respectively. This revealed an isoform-dependent decrease in oligomers (Aβ42 only > Aβ42+apoE4 > Aβ42+apoE3 ~ Aβ42+apoE2) (Fig. 3D–G, I). The reduction in the concentration of the soluble oligomers is consistent with the observed reduction in the number of secondary fibrils (\({N}_{2}(t)\)) in the apoE2 and apoE3-containing samples compared to the apoE4 sample in the TIRFM images (Figs. 2C and S5B, C). We think that a lower population of the oligomers in the presence of apoE2 and apoE3 is observed due to a lower value of the total aggregated mass (M(t)), which is evident from the TIRFM data and the unnormalized bulk fluorescence data (Figs. S6A, 1C, 5A, S13A, B). Furthermore, the absence of detectable oligomers during the lag phase also highlights the role of fibril-catalyzed secondary nucleation (Fig. 3H) in generating these oligomers.

Binding of apoE to the growing ends of the fibril is isoform-dependent

To investigate the binding of single molecules of apoE with the single fibrils of Aβ42, we utilized fluorescently labeled apoE. We examined the colocalization of the apoE molecules with preformed Aβ42 fibrils using TIRFM. In these experiments, apoE4(W210C) is labeled with alexa647, and the fibrils were prepared using unlabeled 2 µM Aβ42 spiked with 2.5% atto425-Aβ42. Since apoE3 contains a native cysteine residue (at position 112), we used a mutant of apoE3(C112I/W210C), which behaves more like apoE3 than apoE441,42. We observed that even 300 pM of apoE4 or apoE3(C112I) bind readily to both the ends and the surface of the fibrils (Figs. 4A,B, S8A,B). The binding of apoE to the fibrils, even at picomolar concentration, suggests that apoE binds to the fibrils with high affinity. To compare the affinities of the apoE isoforms towards binding to the ends versus the surface of the fibrils, we examined the colocalization of apoE at the ends versus the surface of the individual fibrils. First, we defined the terminal 600 nm of the fibrils as the ends and the rest as the surface. Then, we estimated the fraction of the end surface area or the fraction of fibril surface bound by apoE molecules. In presence of apoE4, we observed that only 40% of the fibrillar end surface area is covered by apoE4 molecules, whereas in the rest of the fibrillar surfaces, the coverage is 43%. However, in the presence of apoE3(C112I), 75% of the fibril end surface area is found to be covered by apoE3(C112I), and outside of the fibrillar ends, only ~13% is covered by apoE3(C112I) (Fig. 4C). Therefore, apoE3(C112I) binds more preferentially to the fibril ends than apoE4 (Figs. 4B,S8A). This result, in turn, can explain the stronger inhibitory effect of apoE3 over apoE4 on the elongation of the fibrils.

Fig. 4: Binding of apoE to the Aβ42 fibrils by colocalization microscopy.
figure 4

Colocalization of 300pM of alexa647-apoE4 (A) or alexa647-apoE3(C112I) (apoE3 equivalent) (B) mixed with preformed Aβ42 fibrils prepared using 2 μM WT-Aβ42 spiked with 2.5% atto425-Aβ42. scale bar 2 μm. C Quantification of the fraction of the fibril surface area covered by apoE near (within 600 nm) the fibril end versus the rest of the fibril surface. ApoE3(C112I) shows a sixfold higher propensity to bind to the fibril ends (75%) than the fibril surface (13%). ApoE4 shows nearly equal propensity (≈40%) of binding to the ends versus anywhere else on the fibrils. Each Dot represents the average value from a field of view. Data represents mean ± SD from n ≥ 30 fibrils. D TIRFM (i) and STORM (ii) imaging of WT Aβ42 fibrils upon addition of 150 nM alexa647-apoE4 shows that apoE4 binds extensively to the fibrils. (iii) zoomed STORM image of the yellow inset box in (ii)) shows dark patches of apoE4 free fibril surface denoted by the white arrow. (iv) comparison of the TIRFM and STORM image of the fibrils. E Distribution of localization uncertainty shows overall spread of ≈ 20 nm and FWHM = 9.76 ± 0.05 nm. Calculated from a cropped region in Fig. 4D (ii) (Blue box) (F) and (G) show colocalization of apoE4 and apoE3(C112I) on the Aβ42 fibrils, respectively, after 8 h of incubation of 1 nM seed, 2 μM Aβ42 spiked with 50 nM atto425-Aβ42 and 100 nM of apoE. The right panel shows a zoomed image, the blue arrowheads denote apoE bound fibril end. The white arrows denote apoE free fibril ends. (scale bar = 2 μm) quantified in (H) error bars represent ±SD from n ≥ 80 fibrils. I single-molecule photobleaching of alexa647-apoE4 (of position 1: blue box, in the panel (i)) bound to the Aβ42 fibrils shows multi-step bleaching (bottom panel(ii)), suggesting multimers of apoE4 are involved in binding to fibrils. (blue line is a guide to eye for step count). J The quantification of bleaching steps of the fibril bound apoE4 molecules shows that about 80% are multimers of apoE4. *P < 0.05, **P < 0.01 and ***P < 0.001 (two-sample Student’s t test).

Full size image

Next, we investigated whether apoE4 binds uniformly or preferentially to specific sites on the fibrils. Since the resolution of optical microscopy is limited to about 250 nm, we decided to use superresolution microscopy, viz, STORM43. In these experiments, we used alexa647-apoE4 and preformed Aβ42 fibrils labeled with ThT (Fig. 4D). The resolution of STORM imaging in our experiments is found to be less than 20 nm (Fig. 4E, for details: see SI and Fig. S9), which is a factor of 15 better than the diffraction-limited resolution. The STORM images also show that apoE binds to both the ends and the surface of the fibrils. However, there are several places on the fibril surface where apoE is found to be missing, giving rise to bright and dark patches along the length of the fibrils (Fig. 4D, D iii white arrow, S10G). ApoE3(C112I) also binds to the fibril surfaces with occasional apoE-free dark patches (Fig. S8A). Therefore, apoE appears to bind to specific sites rather than uniformly on the fibrillar surface. Further, studies are required to understand the conformational or morphological variations that may exist along the surface of the amyloid fibrils.

Next, we incubated Aβ42 seeds with 2 μM Aβ42 monomer and 50 nM atto425-Aβ42 with 100 nM apoE3 or apoE4 spiked with 10 nM alexa647-apoE (see methods, SI). Unlike in the preceding section, here, apoE is added at t = 0 rather than on the preformed fibrils. After 8 hours, we performed the colocalization measurements to examine how apoE4 and apoE3 differ in terms of blocking the growing ends of the fibrils (Fig. 4F, G). We analyzed more than 80 fibrils and counted the number of fibrillar ends that are bound or free (Fig. 4F, G, see the arrowheads blue or white). Following quantification of the bound versus the free ends, it is clear that apoE3 binds more specifically (82% bound ends) to the ends than apoE4 does (54% bound ends) (Fig. 4H). Moreover, similar to the STORM images, TIRFM colocalization of apoE also shows apoE-free patches on the fibril surface (Fig. 4F, G, right panel; S8A bottom panel).

To further assess the binding preferences of the apoE isoforms on the growing fibrils, we incubated 1 nM fibrillar seeds with 2 µM Aβ42 monomers at varying apoE concentrations (Fig. S10A, B). Quantification of the binding to the fibrillar ends showed that apoE3 preferentially binds to fibril ends more than apoE4, with binding percentages of 70.7 ± 4.1%, 86.5 ± 1.5%, and 91.8 ± 2.4% for apoE3, compared to 42.5 ± 2.9%, 59.2 ± 1.92%, and 70.3 ± 3.5% for apoE4, at apoE concentrations 75 nM, 150 nM, and 300 nM, respectively (Fig. S10C). Surface coverage also increased with apoE concentration: at 75 nM, apoE4 covered 18.86 ± 3.46% of the fibril surface, while apoE3 covered 13.5 ± 2.0%; at 150 nM, coverage was 27.8 ± 2.8% for apoE4 and 26.14 ± 5.8% for apoE3; at 300 nM, coverage reached 36.2 ± 4.1% for apoE4 and 45.4 ± 7.9% for apoE3 (Fig. S10D). However, even at 300–500 nM, apoE-free patches remained visible (Fig. S10A, B). At 300 nM apoE concentration for apoE4 ca. 64% and for apoE3 ca. 55%, the fibril surface remains unoccupied (Fig. S10D). This suggests that apoE binds to certain hotspots on the fibrillar surface. Why the binding of apoE to the fibrillar surface does not retard the secondary nucleation of Aβ is unclear. Further investigations are required to clarify this behavior.

To confirm that the dark patches, i.e., apoE-free patches on the fibrils, are not artifacts caused by strong depth dependence of the fluorescence signal in TIRFM, we performed the imaging of the fibrils further by variation of depth using HiLo microscopy. The apoE-free dark patches were observed even in HiLo microscopy images (see Fig. S10F). Furthermore, STORM imaging in the apoE channel shows several apoE free patches, whereas visualization of the same fibrils by ThT fluorescence did not exhibit such dark patches (Fig. S10G). Moreover, STORM imaging of the fibrils using alexa647-Aβ42 did not exhibit Aβ free patches (Fig. S10G). Thus, the observed apoE hotspots are not TIRFM depth-dependent imaging artifacts.

Then, we performed single-molecule photobleaching experiments to examine the oligomeric status of the apoE molecules binding to the fibrils. Single-molecule photobleaching is a powerful technique for counting fluorescently labeled molecules within a biomolecular complex if the complex is small44. Real-time monitoring of the alexa647-apoE4 fluorescence intensity trace upon continuous exposure to excitation light shows multiple discrete steps of photobleaching, indicating that apoE4 binds to the fibrils mainly as oligomers (Figs. 4I,S8C, Movie S10). Quantification of the bleaching events shows that almost 80% are multi-step (n = 2, 3, or 4), with highest number of steps being n = 4 (Fig. 4J). ApoE3 also exhibited multi-step photobleaching. However, apoE3 exhibited a shift toward larger oligomers, ranging from dimers to heptamers (Fig. S8D, E).

In Fig. 1B, we showed the growth of Aβ42 fibrils are polarized. Here, we examined the effect of apoE on the polarized growth of Aβ42 fibrils. First, two-color TIRFM experiments confirmed the highly polarized growth of Aβ42 fibrils (Fig. S10E, Movies S7–9: recorded in ThT channel with pseudo-colored seeds). Anisotropic growth of the fibrils was observed in the presence of apoE as well (Movies S8 and S9 for samples with 150 nM apoE3 and apoE4). Three-color imaging revealed that apoE can bind to both ends of the fibrils (Fig. S10E, middle and right). Further, quantitative analysis of the growth rate of both ends showed that apoE4 significantly reduced the elongation rate of the fast-growing end from 0.73 ± 0.04 \({{\rm{nm}}}\) \({\sec }^{-1}\) to 0.23 ± 0.09 \({{\rm{nm}}}\) \({\sec }^{-1}\) and the slow-growing end from 0.08 ± 0.02 \({{\rm{nm}}}\) \({\sec }^{-1}\) to 0.01 ± 0.03 \({{\rm{nm}}}\) \({\sec }^{-1}\) (Fig. S11A). In the presence of apoE3 and apoE2, no detectable growth of the slow-growing end was observed even after several hours (Movies S8, 9). Moreover, apoE3 shows higher coverage at both ends than apoE4 (Fig. S11B). Furthermore, both isoforms show higher coverage at the fast-growing compared to the slow-growing end (Fig. S11B).

Finally, we examined whether apoE-Aβ interactions are predominantly ionic or hydrophobic in nature. We found that the effect of ApoE-Aβ42 interactions on elongation remains unaffected by the charge of the lipids. For example, a similar inhibitory effect was observed in the presence of DMPS-apoE3 (Fig. S12A). The common hydrophobic interaction disruptor 1,6-hexanediol (HXD) disrupted apoE binding to fibrils, while 2 M NaCl had no impact, confirming hydrophobic forces as the key driver (Fig. S12B, C)40.

ApoE concentration-dependent elongation inhibition kinetics at single fibril level and determination of isoform-dependent apoE binding constant to the fibril ends

Then, we set out to determine the binding constants of the apoE-fibril interactions following the approach taken by Arosio et al.45, assuming that inhibition of the elongation rate is proportional to the binding of apoE to the terminal ends of the fibrils, apoE is in large excess, and the binding is under equilibrium45,46. Although these authors used bulk kinetic data, we are using single fibril data. Therefore, we are using the concentration-dependent inhibition of the elongation to estimate the Keq,(end). The advantage of the single fibril analysis is that we can measure the effect of apoE purely on the elongation of the fibrils. Figures 5A and S13B demonstrate that the Aβ fibrils become progressively shorter in the presence of increasing concentrations of apoE2 and apoE3, respectively. Expectedly, the effect of apoE4 is lesser than apoE2 or apoE3 (see Fig. S13A). From the measured elongation rates \({k}_{+}^{{app}}/{k}_{+}\) is calculated (Fig. S13C), where \({k}_{+}\) and \({k}_{+}^{{app}}\) indicate elongation rate constants in absence and in presence of apoE, respectively (see methods, SI). Figure 5B shows that the plots of the \({k}_{+}^{{app}}/{k}_{+}\) versus [apoE] agrees well with Eq. 1, which contains only one free parameter, viz, Keq,(end).

$${k}_{+}^{{app}}/{k}_{+}=(2+{K}_{{eq}({end})}{C}_{i})/(2+{2K}_{{eq}({end})}{C}_{i}+2{K}_{{eq}({end})}^{2}{C}_{i}^{2})$$
(1)

In Eq. 1, Ci represents the concentration of apoE relative to Aβ42, and the numeric factor ‘2’ arises due to the two growing ends of the fibrils. Fitting of the \({k}_{+}^{{app}}/{k}_{+}\) data presented in Fig. 5B with Eq. 1 yields the values of \({K}_{{eq}({end})}\) for apoE4, apoE3 and apoE2 are 107 M−1, 5.5 × 107 M−1 and 4 × 107 M−1 respectively. Therefore, the affinity constant of the apoE isoforms to bind to the fibril ends turns out to be 100 nM for apoE4, 18 nM for apoE3, and 25 nM for apoE2. Thus, apoE2 and apoE3 bind more strongly to the fibrils than apoE4, giving rise to stronger inhibitory effects by apoE2 and apoE3. The cys-cys bond formation in apoE3 and apoE2 could be the reason behind the higher inhibitory effect compared to apoE447. In fact, the data presented in Fig. 4H shows that the fraction of fibril ends bound to apoE agrees well with the concentration-dependent reduction in \({k}_{+}^{{app}}/{k}_{+}\) (Fig. 5B). For instance, in the presence of 100 nM apoE4, the percentage of the bound fibrillar ends is \(\sim\) 53%, which agrees with \({k}_{+}^{{app}}/{k}_{+}\) = 0.5, similarly, for apoE3(C112I) the values are 82% and 0.1(i.e. 90% bound end) respectively.

Fig. 5: Determination of the apoE-fibril (ends) binding constants from the apparent rate of elongation of the Aβ42 fibrils in presence of apoE.
figure 5

A Real-time monitoring of TIRFM images of Aβ42 fibrils formed under seeded condition (1 nM monomer equivalent seed with 2 µM Aβ42) with increasing concentration of lipidated-apoE2 (from top to bottom 0 nM, 75 nM, 150 nM and 300 nM of apoE2). From Left to right, the representative images are at t = 2 h, 10 h, 20 h, and 35 h, respectively. Scale bar = 2 μm. B The elongation rate constant, normalized by the elongation rate constant in the absence of apoE, is plotted as a function of apoE concentration relative to Aβ42. The error bars represent \(\pm\) SEM. The solid lines are fits using Eq.( 1). Elongation rates of the Aβ42 fibrils were measured by tracking single fibrils.

Full size image

Discussions and conclusions

Despite its high importance in the development and progression of the pathology of AD, the effect of the apoE isoforms on the microscopic rate constants of fibrillization of Aβ has not been determined yet. Here, using single-molecule fluorescence techniques (TIRFM/STORM, see schematic Fig. S14), we report that the apoE inhibits elongation of the Aβ42 fibrils in an isoform-dependent manner, with apoE4 exhibiting the lowest inhibitory effects. Our results, summarized in the schematic form in Fig. 6, show that the binding affinity of apoE4 is lower towards the fibril ends compared to apoE3 and apoE2. Thus, attachment of the Aβ42 monomers to the fibril ends is also inhibited in an isoform-dependent manner, leading to slower growth of the fibrils. However, Aβ42 monomer binding to the catalytic (i.e., secondary nucleation) sites remains possibly unaffected in presence of the apoE isoforms. The values of the rate constants obtained from our experiments are summarized in Table 1. Since apoE binds to both the ends and surface of the fibrils, our observation that apoE does not affect secondary nucleation appreciably is somewhat surprising. Our results suggest that apoE is a weaker competitor of Aβ for binding to the secondary nucleation sites. Recently, Curk et al.48 showed that only a small fraction of the catalytic sites are necessary for the self-replication of Aβ42. In fact, the movie represented in movie S2 (Aβ42 + 75 nM apoE4) shows only a fraction of the fibril surface capable of self-replication to generate new fibrils, i.e., secondary nucleation. Therefore, in the presence of apoE, secondary nucleation could potentially occur at the apoE-free patches on the fibrillar surface (Fig. 4F, G, right panel; S4A bottom panel; S10).

Fig. 6: Schematic model for interaction of apoE with the Aβ42 fibrils.
figure 6

Top panel shows apoE binds to the growing ends of the fibrillar seeds in an isoform-dependent manner with apoE3 and apoE2 binding with higher affinity than apoE4 \(({apoE}3 \sim {apoE}2 > {apoE}4)\). Binding of ApoE inhibits binding of the Aβ42 monomers to the fibril ends slowing down elongation of the fibrils, denoted by the broken arrow. Secondary nucleation is unaffected in presence of apoE. Bottom panel shows that in absence of apoE, Aβ42 fibrils can grow unhindered.

Full size image

Furthermore, using single-molecule photobleaching analysis, we found out that apoE binds to the fibrils primarily as oligomers with an oligomeric status of 2 to 4 mer for apoE4 and 2 to 7 mer for apoE3 (Movie S10, Figs. 4J, S8E). Oligomerization of the apoE is known to be isoform dependent, with apoE2 exhibiting the highest and apoE4 the lowest oligomeric status in the lipidated particles47,49. Therefore, differences in the oligomeric status and their binding ability, along with structural differences, could be responsible for the isoform-dependent effect on Aβ aggregation. However, further studies are required to understand the structural differences between apoE isoforms bound to Aβ fibrils. Our results also suggest weaker binding of apoE4 compared to apoE3 with the aggregates of Aβ. In vivo, the microdialysis studies in PDAPP/TRE mice and APP/PS1 mice expressing human apoE isoforms show that apoE4 impairs the clearance of the soluble Aβ from the brain interstitial fluid (ISF), i.e., higher half-life of soluble Aβ compared to apoE3 and apoE2 isoforms10,16. Alternatively, weaker binding of the aggregates by apoE4 could affect its ability to facilitate microglia-mediated removal of the apoE-associated Aβ plaques50. Several receptors (e.g., Triggering Receptor Expressed on Myeloid Cells-2 (TREM2), Low-density lipoprotein receptor-related protein 1 (LRP1), Low-density lipoprotein receptor (LDLR), Very low-density lipoprotein receptor (VLDLR), Apolipoprotein E Receptor-2 (apoER2)) are known for binding to apoE-Aβ complexes and redirecting them for clearance. ApoE4 is known to disrupt these processes51,52,53,54,55. One can speculate that the higher amyloid burden in the brains of the apoE4 individuals could be multifactorial, such as its poor inhibitory effects on the aggregation of Aβ, its weaker binding affinity towards the nascent fibrils, the higher half-life of the soluble Aβ in ISF, and the impaired microglial mediated clearance of the apoE associated plaques10,14.

Mouse model studies expressing human apoE isoforms have observed the highest extent of amyloid deposition in the presence of apoE4 and the lowest in the presence of apoE217. Our results are consistent with these observations. Although we observed that apoE isoforms do not inhibit secondary nucleation, they inhibit elongation of the secondary fibrils more effectively than the primary fibrils. Since secondary nucleation is believed to be a major source of generating the soluble oligomers29,56, which are believed to be highly cytotoxic56,57,58,59, we hypothesize that blocking the secondary nucleation and targeting the apoE receptor-mediated Aβ clearance could be pivotal in developing therapeutic strategies for AD. In fact, a human monoclonal antibody-based FDA-approved drug, Aducanumab, has shown good clinical promise60 with some drawbacks, like amyloid-related imaging abnormality (ARIAs) and failure of microglial response at a later stage61,62. Aducanumab specifically binds to soluble and insoluble aggregates of Aβ60. It is also known for its inhibitory effects on the secondary nucleation of Aβ4263, as well as helping microglial receptor-mediated clearance of Aβ aggregates through phagocytosis of the antibody-Aβ complexes60. Lecanemab, an antibody selective for soluble Aβ protofibrils64,65, has recently demonstrated promising and consistent clinical efficacy results66, unlike Aducanumab67, and is known to protect neurons from Aβ-induced cell death68. However, ARIAs-related adverse effects in patients treated with Aβ monoclonal antibodies raised concerns against these drugs, and it has been proposed that targeting plaque-associated apoE could be a more promising approach for next-generation AD therapeutics69,70.

Microglia are known to engulf and clear Aβ aggregates more in apoE3 mice compared to the apoE4 case18, and these could be linked to the nature of the apoE-Aβ complexes, which forms in an apoE isoform-dependent manner in vivo. We propose that these microglial phagocytosis processes might resemble chaperone-mediated fibril breakdown mechanism initiated at fibril ends71, and efficient binding of apoE3 to Aβ fibril ends may potentially accelerate these processes compared to apoE4, which binds unspecifically to fibrils. Our findings underscore the isoform-dependent binding patterns of apoE to fibrils, which may correlate with better microglial clearance of Aβ aggregates in apoE3 mice compared to apoE4 mice, as discussed above. These results might open up new avenues of research in AD diagnosis and treatment. Based on the previous studies and our findings with the apoE-Aβ interactions, we propose that next-generation AD drugs should be developed with the following features: enhanced efficacy in halting secondary nucleation63, small molecule modifiers to convert apoE4 to behave more like apoE3/apoE272, and agents that increase microglial and astrocytic receptor-mediated clearance of apoE-Aβ complexes54. Therefore, future studies should compare the biophysical properties, biological toxicity, and the clearance of the aggregates prepared in the absence and the presence of apoE.

In conclusion, we have determined the rate constants of elongation and secondary nucleation of the Aβ fibrils in the presence of all the three isoforms of apoE using both single molecule/fibril and bulk fluorescence techniques. Our results suggest that the effect of apoE on the elongation of Aβ fibrils is isoform-specific, with apoE3 (and apoE2) showing the highest and apoE4 the weakest effect. While secondary nucleation is not affected, the inhibitory effects of the apoE isoforms are stronger towards the growth of the secondary fibrils than the primary fibrils. The poor ability of apoE4 to suppress fibrillization and poor binding with the fibrils of Aβ may be responsible for its pathogenic roles in AD.

Materials and methods: Detailed Materials and methods are included in Supplementary Information (SI).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.