Intrinsic excitability increased more sharply with the simulation strength elevation in APP/PS1 mouse hippocampal neuron
As a severe neurodegenerative disorder, AD has destabilized calcium homeostasis and presynaptic/postsynaptic glutamate regulations related to dysfunction of neuronal excitability [48-51]. Axonal pathology in APP/PS1 neurons, accompanied by membrane proteins and ion channel alterations, could seriously affect intrinsic neuronal excitability [52-55]. In the present study, we found that wild-type neuron and APP/PS1 neurons have normal resting potential (WT: -69.1 ± 0.98 mV, n = 77; APP/PS1: -67.1 ± 0.82 mV, n = 48, nsP value = 0.1703) (Fig. 1A). Then, we performed a current injection with small gradual increases (from 5 to 100 pA with a step of 5 pA). The current injection duration and the following rest duration were 500 ms) to investigate the excitability change in different stimulation strengths in wild-type and APP/PS1 neurons. Our results showed that APP/PS1 neurons had significantly higher rheobase under 500 ms depolarizing current injection than wild-type neurons (WT: 50.2 ± 3.7 pA, n = 52; APP/PS1: 63.6 ± 4.3 pA, n = 35, **P-value = 0.0057) (Fig. 1B).
APP/PS1 neurons were much more sensitive to the current injection than wild-type neurons and had higher AP frequency when the current injection was strong enough, showing a stimulation-strength-dependent excitability heterogeneity. As shown in Fig. 1C, compared to wild-type neurons (gray), APP/PS1 neurons (green) had lower AP frequency with a lower intensity of current injection (for example, 30 pA) but had higher AP frequency with a higher intensity of current injection (for example, 100 pA). The spiking probability of APP/PS1 neurons obviously increased when the current raised to more than 40 pA and surpassed wild-type neurons when the current was more than 65 pA (n = 52) (Fig. 1D). More specifically, two types of neurons' AP frequency showed a similar trend (Fig. 1E). Compared to the wild-type neuron, the AP frequency of APP/PS1 neurons was lower with a milder current injection, but it increased severely when the current injection intensity was over 50 pA. The AP frequency surpassed wild-type neurons when the current injection intensity was over 70 pA (Fig. 1E). As highlighted in Fig. 1F and G, the AP frequency of APP/PS1 neuron was about 60% less than wild-type under 50 pA current injection (WT: 3.49 ± 0.71 Hz, n = 52; APP/PS1: 1.32 ± 0.37 Hz, n = 35, *P value = 0.0191) (Fig. 1F), but about 60% more than the wild-type neurons under 100 pA current injection (WT: 6.78 ± 1.03 Hz, n = 52; APP/PS1: 10.8 ± 1.31 Hz, n = 35; *P value = 0.0132) (Fig. 1G).
Stimulation-strength-dependent excitability increase not only reflected in the AP frequency but also in the first AP peak latency. It decreased along with the current injection intensity increase, and the gap between the two types of neurons decreased to less than 10 ms when the current injection intensity was over 75 pA (Fig. 1G). The first AP peak latency of APP/PS1 neuron was significantly longer than wild-type neuron when neurons were stimulated by a 50 pA current injection (WT: 149.6 ± 12.4 ms, n = 28; APP/PS1: 228.1 ± 28.2 ms, n = 14, *P value = 0.0202) (Fig. 1I), while it was almost equal to wild-type neurons with 100 pA injection (WT: 109.9 ± 10.6 ms, n = 42; APP/PS1: 111.3 ± 12.0 ms, n = 32, nsP value = 0.9303) (Fig. 1J). Correspondingly, in APP/PS1 neurons, the time interval between the first and second peak with the increased stepwise current injection also decreased faster than wild-type neurons and was basically the same as wild-type neurons when current injection intensity was over 70 pA (Fig. 1K).
High spatial-temporal voltage imaging was applied to study action potential initiation and propagation in the AIS
To study the AP initiation site and propagation manners in the AIS, we performed voltage imaging in cultured hippocampal neurons from wild-type and APP/PS1 mice. We chose QuasAr2 as the GEVI in this study for its high sensitivity and fast response kinetics [43]. It has been employed to monitor AP initiation in cultured rat hippocampal neurons and investigate neuronal excitability of ALS patient-derived motor neurons [43, 56]. Compared to previous imaging studies of the AIS using dyes [40-42], QuasAr2 offers a superior signal-to-noise ratio (SNR) for its lower background signals and higher sensitivity in responding to AP events and describing the AP waveform details [36, 57].
As showed in Fig. 2A, cultured hippocampal neurons were transfected with QuasAr2-mOrange2 plasmid. They were stimulated by periodic current injections injected into the soma via the whole-cell current clamp. Meanwhile, QuasAr2 fluorescence images of the AIS and adjacent soma, axon, and dendrites were acquired at a 484 Hz camera frame rate. The transfected neurons for voltage imaging from wild-type and APP/PS1 mice showed a significant difference neither in their Full Width at Half Maxima (FWHM) of evoked APs, nor in the amplitude, resting potential, or threshold potential (Supplementary Fig. S1). After the voltage imaging, we used the standard immunofluorescence (IF) technique to visualize the recorded neurons and their AIS. The neurons were filled with biocytin (0.5% w/v) as a tracer, and the AIS-localized AnkG or NF186 indicated their AISs. (Fig. 2A and B; Supplementary Fig. S2).
We generated the spike-triggered average movies from the voltage imaging results corresponding to the current injection-induced AP trains (Fig. 2C). The optical signals of APs in the spike-triggered averaged movies had much higher SNR, and a few AP propagations could be directly observed from the movie (Fig. 2D). In our experiment results, the current injection (10 ms - 150 pA stimulation, frequency = 6.25 Hz, and total AP number = 125) generated AP trains and corresponding fluorescence of QuasAr2 changed in the soma, AIS, and axon (outlined in blue, orange and yellow, respectively) (Fig. 2D). The fluorescence intensity robustly reported AP spikes at the AIS and neighboring regions, with an SNR ranging between 11 and 36 in the outlined AIS region (Fig. 2D, center column, mean ± standard deviation: 17.2 ± 4.3). We generated the spike-triggered average spike movie to boost the SNR of the AP signal in the AIS to 111 (Fig. 2D, right column, the AIS trace). We further adapted an upsampling algorithm [42, 43] to analyze the AP timing with sub-millisecond-level temporal resolution and pixel-level spatial resolution (Fig. 2E and Supplementary Fig. S3, see METHODS). After this processing, we identified the AP initiation site and the velocity of bidirectional AP propagation in the AIS and adjacent soma, axon, and dendrites (Fig. 2F).
Voltage imaging reveals AP propagation velocity at the AIS
In our study, AIS length was defined as the distance between the proximal end and distal end of the AIS, and AIS location was defined as the distance from cell soma to the AIS proximal end [58]. Since its initiation, AP propagates bi-directionally, with forward-propagating action potentials (fpAPs) traveling along the axon and back-propagating action potentials (bpAPs) towards the soma and the dendrites (Fig. 3A). The previously described voltage imaging technique and data processing algorithm (Fig. 2E and f and Supplementary Fig. S3) generated a pixel-by-pixel AP firing timing heatmap (Fig. 3B). We extracted the mean AP propagation velocities on several compartments with high temporal-spatial resolution from our raw data (Fig. 3C). We measured AP velocities in the AIS and compared these values pairwisely. We found that the velocity of fpAP in the AIS was (1.87 ± 0.39 times) faster than in the axon (119.5 ± 18.7 μm/ms and 73.45 ± 7.76 μm/ms, respectively, *P-value = 0.0214) (Fig. 3D). We also found that the velocity of fpAPs were generally faster (1.51 ± 0.18 times) than the corresponding bpAPs (112.6 ± 14.2 μm/ms and 84.99 ± 7.59 μm/ms, respectively, nsP value = 0.0547) (Fig. 3E). Taken together, our voltage imaging method enabled us to distinguish AP propagation velocities from different subcellular regions.
It is well established that the AIS is a highly structured region and is attached by many ion channels, playing a crucial role in the AP initiation and propagation. This high spatial and temporal resolution technique allowed us to observe the changes in AP velocity during the propagation. As shown in Fig. 3F, the raw peak arrival times (gray circles) were not monotonic because of the data noise. Thus, we used the least-squares curve fitting with linear constraints to fit the raw data. Lines in Fig. 3F indicated the monotonic data; the slopes of lines on each location were the instantaneous velocities (Fig. 3G). The velocity sign represented the propagation direction; negative and positive velocities were bpAP and fpAP velocities, respectively. The velocity change per unit time was the acceleration (Fig. 3H). The sign of acceleration represented if the velocity becomes slower or faster. For bpAP, the maximum acceleration (the blue triangle) was where bpAP speeded up most, while the minimum acceleration (the blue inverted triangle) was where bpAP slowed down most. For fpAP, the maximum acceleration (the brown triangle) was where fpAP speeded up most, while the minimum acceleration (the brown inverted triangle) was where fpAP slowed down most. Here we defined SDMP as the point where AP slowed down most and SUMP as the point where AP speeded up most.
The calculation details of the acceleration are mentioned in the METHODS section. We calculated SDMP and SUMP on all neurons. In the control group, the relative distance from SDMP and SUMP of bpAP to the AIS proximal end (SDMP/SUMP - AIS proximal end) was 3.8 ± 0.6 μm (n = 64) and 4.3 ± 0.6 μm (n = 64), the location distribution was present in the histogram (Fig. 3I and J). The relative distance from SDMP and SUMP of fpAP to the AIS distal end (SDMP/SUMP - AIS distal end) was -1.3 ± 0.7 μm (n = 64) and -1.8 ± 0.6 μm (n = 64); histogram shows the location distribution (Fig. 3K and L). These results indicated that the AP velocity had a sudden change near the proximal and distal end of AIS. It suggests a specific morphological structure in the AIS, not only the selective transportation of the molecules at the terminal of the AIS, but the structure affects the AP propagation properties.
AP was initiated more distal to the soma in APP/PS1 neurons
Previously we showed that APP/PS1 mouse model had axon abnormalities with shortened AIS mediated by AnkG down-regulation [9]. We identified the AIS with AnkG as a marker (Fig. 4A). We set the highest fluorescence value of AnkG along the axon as the maximum (100%) fluorescence intensity and the lowest fluorescence value of AnkG as the minimum (0%) fluorescence intensity. We defined the region with more than 10% of the maximum fluorescence intensity as the AIS proximal and distal end [58]. Compared to wild-type neurons, the AIS of APP/PS1 hippocampal neurons had similar location (WT: 5.904 ± 0.77 μm; APP/PS1: 5.383 ± 1.05 μm, nsP-value = 0.6857) (Fig. 4B), but had significantly shorter length (WT: 38.49 ± 1.86 μm; APP/PS1: 24.07 ± 1.07 μm, ***P-value < 0.001) (Fig. 4C).
To obtain the detailed AP initiation and propagation information, we extracted the AP propagation process from raw data and defined the location with the earliest action potential peak time as the AP initiation site (red dot in Fig. 3C). We first normalized each AIS length from proximal end (0%) to distal end (100%) and calculated the relative AP initiation site. We found that in wild-type neurons, the highest probability of AP initiation site was at 50%-70% of total AIS length, which was coincided not only with the spatial distribution of voltage-gated sodium channel Nav1.6 in the AIS, but also with the findings of the previous study that AP is initiated in the slightly distal end of the AIS [30, 59]. However, in APP/PS1 neurons, APs were initiated more distally (70%-90% of total AIS length), even one-third of neurons-initiated AP in the axon area distal to the AIS (Fig. 4D). Nevertheless, APs were nearly initiated at the same absolute distance to the soma in wild-type and APP/PS1 neurons (Fig. 4E and F). In addition, the correlation analysis showed that in wild-type neurons, when the AIS was located distal to the soma, AP further tended to the soma (n = 17) (Fig. 4G). This positive correlation was more apparent between the AP absolute initiation site and AIS length (n = 17) (Fig. 4H). However, in APP/PS1 neurons, although the absolute AP initiation site was similar to wild-type neurons, the correlations between absolute AP initiation site and AIS location or length were both disturbed (n = 11) (Fig. 4I and J).
The alteration of AIS length and sodium channel decreased AP velocities in APP/PS1 neuron
The velocity of AP propagation significantly affects the approach timing of the electrical signals, which is the key to trigger neurotransmitter release at the axon terminals and integrate the information at the soma [23-28]. Since APP/PS1 neurons showed apparent AP initiation alteration, the AP propagation patterns in APP/PS1 neurons might be different from those in wild-type neurons. To study whether AIS structural alteration in APP/PS1 neurons also affected AP propagation, we directly recorded AP propagation velocity in the AIS via the high spatial-temporal voltage imaging. We measured AP velocities in the AIS and found that both bpAP (WT: 132.5 ± 10.33 μm/ms; APP/PS1: 77.49 ± 11.55 μm/ms, **P-value = 0.0016) and fpAP (WT: 90.42 ± 8.45 μm/ms; APP/PS1: 66.42 ± 6.51 μm/ms, *P-value = 0.0254) velocities in APP/PS1 neurons were lower than the wild-type neurons (Fig. 5A and B). The correlation analysis of the AP initiation site and AP velocities showed that in wild-type neurons, neither bpAP (n = 18) nor fpAP velocity (n = 13) were correlated with its initiation site, so did APP/PS1 neurons (n = 11 and n = 9, respectively) (Fig. 5C and D), suggesting the AP initiation site fluctuation did not affect its propagation velocities.
To study whether the shorter AIS in APP/PS1 neurons reduced the AP propagation velocities in the AIS, we analyzed the correlation between the AIS location/length and bpAP/ fpAP velocities. In wild-type neurons, the bpAP velocity (n = 18) was not affected by the AIS location (Fig. 5E), but was slightly reduced in the longer AIS (Fig. 5F). In contrast, the velocity of fpAP (n = 12) significantly decreased in the AIS located far from the soma or in the shorter AIS (Fig. 5G and H). According to these observations, the APP/PS1 neuron has a shorter AIS with a similar location than the wild-type neuron. The lower fpAP velocity was consistent with the correlation between fpAP velocity and AIS length, not the bpAP. These results suggest that in APP/PS1 neurons, the decreased AP velocities in the AIS and neighbor compartments were not uniquely related to the AIS length but might also be contributed to other parameters. We also analyzed the correlation in APP/PS1 neurons, but those correlations were not appeared in APP/PS1 neurons (bpAP: n = 12 and fpAP: n = 9, respectively) (Fig. 5I-L).
The AIS was shorter in APP/PS1 neurons, and Nav1.6 was distributed along the entire axon and showed a lower proportion in the AIS region. We constructed two multi-compartment models to verify whether these changes lead to the bpAP and fpAP velocities decrease in APP/PS1 neurons, one simulated wild-type neuron (Fig. 5M), and the other simulated APP/PS1 neuron (Fig. 5N). Both models were composed of several sections: dendrite, soma, hillock, AIS, and axon. We adjusted the model's morphological and physical parameters according to the published studies [9, 45-47, 60]. Compared with the wild-type neuron model, the length of the AIS segment in the APP/PS1 neuron model was set shorter, the sodium channel density in the AIS segment was set lower, and the sodium channel density in the axon segment was set higher than wild-type neuron model. The specific parameters are listed in Supplementary Table S2. We injected the same current (1 nA, 1 ms) into the soma section. Each section contained some compartments, and the length of each compartment was 1 μm. Fig. 5M and N showed the AP peak arrival time of each compartment. The slopes of the dotted arrows showed the mean velocity of bpAP and fpAP in the AIS. In the simulation, the bpAP and fpAP velocities in APP/PS1 neuron model (bpAP: 125.0 μm/ms; fpAP: 166.7 μm/ms) were both lower than wild-type neuron model (bpAP: 144.7 μm/ms; fpAP: 185.2 μm/ms). These simulation results reproduced the experiment conclusions, indicating that the shorter AIS and the abnormal location of sodium channels together decreased AP velocities in the APP/PS1 neuron.
fpAP propagation velocity was influenced by the AIS length and location
Our study uncovered that fpAP velocity significantly correlated with AIS location and length (Fig. 5G and H). To further investigate the relationship between fpAP velocity and AIS location, we pharmacologically perturbed the AIS location in wild-type neurons. A previous study showed that treating mature AIS (after DIV7) with 15 mM KCl for 48 hours significantly shifted the AIS distally without affecting the AIS length [58]. Therefore, we employed this method to induce an AIS location shift.
In the untreated neuron, the velocity of bpAP (n = 30) slightly decreased in AIS either was located far from the soma or longer. However, neither of them showed statistical significance (Fig. 6A and B). In contrast, the velocity of fpAP (n = 19) was negatively correlated to the AIS location and positively correlated to the AIS length (Fig. 6C and D). After 15 mM KCl treatment from DIV10 to DIV12, the resting potential of these neurons did not change (Fig. 6E), while the rheobase current significantly increased when the neurons were moved back to the 2.5 mM K+ extracellular medium (Fig. 6F).
As envisioned, the AIS location was considerably moved away from the soma (Fig. 6G and H), while the AIS length did not change (Fig. 6G, I). Consistent with the above observations (Fig. 6A-D), bpAP velocity barely changed (101.3 ± 8.87 μm/ms and 83.74 ± 13.89 μm/ms, respectively, nsP value = 0.2676) (Fig. 6J), and fpAP velocity was much lower (138.8 ± 20.1 μm/ms and 65.33 ± 9.30 μm/ms, respectively, **P-value = 0.0028) (Fig. 6K), when compared to the untreated neurons. Moreover, in the KCl-treatment group, the velocity of bpAP (n = 15) was not significantly influenced by AIS location and length (Fig. 6L and M). In contrast, the velocity of fpAP (n = 10) was negatively correlated with AIS location and positively correlated to AIS length (Fig. 6N and O), consistent with the correlations in the untreated neurons (Fig. 6A-D). Taken together, the AIS distally shifting did not significantly influence the bpAP velocity in the AIS, but significantly affected the fpAP velocity in the AIS. These results demonstrated that the AIS location and length influenced the fpAP velocity.
Sodium channel distribution determined the correlation between AIS location/length and AP propagation velocities
To quantitatively study neuronal propagation velocity, we constructed a multi-compartment model. Our model consisted of several sections: dendrite, soma, hillock, AIS, and axon. For simplicity, we modeled each section as a cylinder (Fig. 7A). The model included four types of ionic currents: INav, INav1.2, INav1.6, and IKv. The AIS contained INav1.2, INav1.6, and IKv, but not INav; all other sections contained INav and IKv. Furthermore, the AIS had higher current densities than the other sections, consistent with previous observations. The stimuli injected in the middle of the soma was a pulse current (1 ms, 1 nA).
To examine the mechanisms underlying the correlation between AIS location/length and AP propagation velocities in our experiments (Fig. 6A-D), we explored many parameters in our model. We found parameter regimes where the model qualitatively reproduced experimentally observed trends. Among all the parameters, we found that the spatial distribution of sodium channels most robustly affected the trends. Here we demonstrated the effect by comparing two specific types of sodium channel distribution in the AIS. The default, Type 1 (left panel in Fig. 7B), was a uniform distribution. However, as described in previous studies, the densities of Nav1.2 and Nav1.6 were not uniform. Therefore, in Type 2 (right panel in Fig. 7B), we modeled Nav1.2 to have a higher density in the proximal AIS and Nav1.6 to higher density in the distal AIS. As shown in Fig. 7C, when the distribution of sodium channels was set to be uniform (Type 1), bpAP propagated slightly lower in the AIS located distally, consistent with our experimental results. However, bpAP propagated faster in longer AIS, which contrasts with the experimental results. Moreover, the dependencies of fpAP velocities in the AIS location and length were all inconsistent with our experimental observations.
However, if sodium channels were distributed as Type 2, we could reproduce many experimental trends. For example, in Type 2 sodium channel distributions simulations, bpAP propagated slightly lower in the AIS located distally. Furthermore, modeling results demonstrated that the longer the AIS was, the faster the fpAP propagated. These conclusions were consistent with the experimental results. Compared to Type 1, in Type 2 AISs, the fpAP velocity was not higher in the AIS located distally, and the bpAP velocity had no uniform correlation with AIS length. These conclusions are similar to the experimental trends in Type 2 than in Type 1. In addition to these two types of sodium channel distributions, we tried other types of sodium channel distributions, as shown in Supplementary Fig. S4A-D. Their corresponding simulation results were similar to Type 2 (Supplementary Fig. S5).
Type 2 simulation results could not completely reproduce the experimental trends due to the complexity of soma-to-dendritic morphology in neurons. Soma-to-dendritic morphology influences neuronal physiology, weakening the correlation between AIS plasticity and AP velocities. Furthermore, in our models, as long as the proximal AIS had higher Nav1.2 densities than the distal AIS and distal AIS had higher Nav1.6 densities than the proximal AIS, simulation results can reproduce many experimental results. Therefore, we concluded that sodium channel distribution significantly influenced the interaction between AIS length/location and AP propagation velocities. These simulations suggested that the change in the correlation between AIS location/length and AP propagation velocities in APP/PS1 neurons to wild-type neurons might due to the sodium channel disorganization attributed to axon pathology.
Finally, to study the effect of ion channel densities on AP propagation velocities, we adjusted gNav1.2, gNav1.6, and gK, respectively, keeping to a Type 2 sodium distribution (Supplementary Fig. S6). We found that densities of Nav1.2 and Nav1.6 hardly affected neither the AP velocities nor the trends of how bpAP and fpAP velocities varied with AIS location and length. However, the density of Kv affects the AP velocities the most (generally, a higher Kv density led to lower bpAPs and faster fpAPs) but hardly affected how bpAP and fpAP velocities varied with AIS location and length. These results suggested that the ion channels' spatial distribution, but not its densities on specific sites, more strongly influenced the interaction between AIS length/location and AP propagation velocities.