Skip to main content
PMC home page
eLife logoLink to eLife
. 2021 Sep 7;10:e62091. doi: 10.7554/eLife.62091

Activity-dependent regulation of mitochondrial motility in developing cortical dendrites

Catia AP Silva 1, Annik Yalnizyan-Carson 2, M Victoria Fernández Busch 1, Mike van Zwieten 1, Matthijs Verhage 3, Christian Lohmann 1,3,✉
Editors: Marla B Feller4, Gary L Westbrook5
PMCID: PMC8423438  PMID: 34491202

Abstract

Developing neurons form synapses at a high rate. Synaptic transmission is very energy-demanding and likely requires ATP production by mitochondria nearby. Mitochondria might be targeted to active synapses in young dendrites, but whether such motility regulation mechanisms exist is unclear. We investigated the relationship between mitochondrial motility and neuronal activity in the primary visual cortex of young mice in vivo and in slice cultures. During the first 2 postnatal weeks, mitochondrial motility decreases while the frequency of neuronal activity increases. Global calcium transients do not affect mitochondrial motility. However, individual synaptic transmission events precede local mitochondrial arrest. Pharmacological stimulation of synaptic vesicle release, but not focal glutamate application alone, stops mitochondria, suggesting that an unidentified factor co-released with glutamate is required for mitochondrial arrest. A computational model of synaptic transmission-mediated mitochondrial arrest shows that the developmental increase in synapse number and transmission frequency can contribute substantially to the age-dependent decrease of mitochondrial motility.

Research organism: Mouse

Introduction

Newborns can interact with their environment soon after birth, without any previous experience of sensory input. This ability relies on extensive preparation of the developing nervous system before the onset of sensory experience. Young networks are initially established by molecular guidance cues and refined by activity-driven synaptic plasticity. Before the onset of sensory processing, developing neuronal networks generate neuronal activity spontaneously that strengthens well-targeted synapses and weakens others to prepare the brain for sensory processing. Later, learning adjusts synaptic circuits to the prevalent environmental conditions (Katz and Shatz, 1996; Sengpiel and Kind, 2002; Sanes and Yamagata, 2009; Kilb et al., 2011; Kirkby et al., 2013; Leighton and Lohmann, 2016).

The development of synapses and synaptic transmission are highly energy-demanding processes. A substantial amount of this energy is supplied by mitochondria, the main energy providers in neurons (Harris et al., 2012). Imaging experiments showed that neuronal mitochondria can be highly motile in intact tissue (Misgeld et al., 2007; Plucińska and Misgeld, 2016). For example, mitochondria are generated at the soma and transported to distal dendrites and axons via the microtubule network (Sheng and Cai, 2012). This motility allows for energy provision at high-energy-demanding sites, in particular, synapses. Defects in mitochondrial motility have been shown to lead to impaired neurotransmission, further linking mitochondrial motility and synaptic function (Sheng and Cai, 2012). In addition, previous studies reported that experimentally enhancing neuronal activity (with high extracellular potassium, the voltage-gated sodium channel activator veratridine, glutamate, or electrical stimulation) stops mitochondria at synapses, whereas blocking action potential firing using tetrodotoxin (TTX) increases mitochondrial motility and reduces the number of stationary mitochondria at synapses (Rintoul et al., 2003; Li et al., 2004; Chang et al., 2006; MacAskill et al., 2009). MIRO1, a calcium-sensitive protein-linking mitochondria to the microtubule network, can mediate mitochondrial arrest in dendrites and axons (MacAskill et al., 2009; Wang and Schwarz, 2009): upon calcium binding, MIRO1 releases mitochondria from motor proteins (kinesins or dyneins), thus interrupting their motility.

In contrast, other evidence suggests that mitochondrial motility in neuronal dendrites is not affected by activity (Beltran-Parrazal et al., 2006; Faits et al., 2016). In retinal explants, neither spontaneously occurring nor stimulus-evoked activity affect mitochondrial motility (Faits et al., 2016). Moreover, mitochondrial motility remains high in an hyperactive retina with immature synapses (Morrow et al., 2005; Tran et al., 2014; Faits et al., 2016). These observations suggest that high mitochondrial motility may not be the consequence of low activity in immature tissue, but rather a characteristic of very young neurons (Faits et al., 2016). Thus, activity levels may co-vary with mitochondrial arrest rather than causing it.

To address the role of natural activity patterns in mitochondrial arrest, we investigated here whether spontaneous activity affects mitochondrial motility in the developing visual cortex both in vivo and in organotypic slice cultures. We found that mitochondrial motility decreased over the first 2 postnatal weeks while the frequency of spontaneous activity increased. Global spontaneous calcium transients did not affect mitochondrial motility; however, spontaneous activity at the synaptic level preceded mitochondrial motility arrest and pharmacological stimulation of synaptic vesicle release, but not focal glutamate application alone, was sufficient to stop mitochondrial motility. A computational model of synaptic activity-mediated control of mitochondrial motility suggests that the developmental increase in synapse number and transmission frequency contributes substantially to the age-dependent decrease of mitochondrial motility.

Results

We investigated the relationship between spontaneous activity and mitochondrial motility in vivo and in organotypic slice cultures of the developing mouse primary visual cortex during the second postnatal week before eye opening at postnatal day (P) 14 (Figure 1A). We used in utero electroporation at embryonic day 16.5 to express the calcium indicator GCaMP6s and mitochondrial-DsRed in pyramidal neurons of layer II/III (Figure 1B–C). Time-lapse recordings were performed to reveal the spatio-temporal relationship between mitochondrial motility and calcium signaling in developing dendrites (Figure 1D–E).

Figure 1. Simultaneous imaging of dendritic calcium transients and mitochondrial motility in vitro and in vivo.

Figure 1.

Open in a new tab

(A) Timeline of in vivo and in vitro experiments: in utero electroporation (IUE) was performed at embryonic day (E) 16.5 to deliver GCaMP6s (calcium indicator) and Mito-DsRed (mitochondrial marker) to pyramidal neurons of layer II/III in the visual cortex. In vivo experiments: acute imaging of transfected dendrites in pups between postnatal day (P) 5 and P12 using a two-photon microscope. In vitro experiments: imaging of transfected dendrites using a confocal microscope in organotypic cortical slices cultured for 3–7 days after slice preparation from P5 or P8 pups. (B) GCaMP6- and Mito-DsRed-expressing layer II/III pyramidal neurons in vivo (P16). (C) GCaMP6- and Mito-DsRed-expressing layer II/III pyramidal neurons in vitro (P5 + DIV4). (D) Dendrite of layer II/III pyramidal neuron in vivo and kymograph (right) representing dendritic calcium transients (green) as well as motile and stationary mitochondria (red). Immobile mitochondria appear as horizontal lines (no change in position over time) and moving mitochondria as diagonal lines. Below, graphic representation of the percentage of moving mitochondria and global calcium transients. The percentage of moving mitochondria was calculated as the number of moving mitochondria over the total number of mitochondria, binned for every second. The mean percentage of moving mitochondria across the duration of this recording was 8.5%. Vertical green lines show spontaneously occurring global calcium transients, most likely resulting from back-propagating action potentials. (E) Dendrite of the layer II/III pyramidal neuron shown in C. The mean percentage of moving mitochondria across the duration of the recording was 2.4%.

Previous studies reported that neuronal activity and calcium signaling reduce mitochondrial motility in dendrites in vitro (Li et al., 2004; Chang et al., 2006), but this idea has not been tested in vivo. Therefore, we first investigated the interaction between spontaneous network activity and mitochondrial motility in neonatal mice. Overall, we observed an anti-correlation between the frequency of spontaneous global calcium transients and the percentage of moving mitochondria in awake (unanesthetized) animals (Figure 2A). Upon closer inspection, it became clear that these parameters were linked systematically to the age of the animal: in older animals (≥P8) activity levels were consistently higher and mitochondrial motility was low (Figure 2A–D). We observed a similar relationship between neuronal activity, mitochondrial motility, and age in anesthetized mice (0.8% isoflurane, Figure 2—figure supplement 1A-D). Therefore, we combined both groups for the analyses shown below (Figure 2E–G).

Figure 2. Mitochondrial motility and spontaneous activity are anti-correlated during in vivo early postnatal development.

(A) Anti-correlation between the frequency of spontaneous global calcium transients and the percentage of moving mitochondria in imaging experiments of awake pups (n = 13 pups, Spearman’s rank correlation; 1192 mitochondria in 131 dendrites). (B) The frequency of spontaneous global calcium transients increased until postnatal day (P) 9 (but not significantly for the entire age range, Spearman’s rank correlation) and the percentage of moving mitochondria decreased over P6–11 in vivo (Spearman’s rank correlation). (C-D) When comparing awake animals younger than P8 to P8 and older, the frequency of spontaneous global calcium transients increased (t-test, n = 5 vs. n = 8, p = 0.02) and the percentage of moving mitochondria decreased (t-test, n = 5 vs. n = 8, p = 0.045). (E-F) Application of tetrodotoxin (TTX, 2 µM) on the surface of the cortex (n = 7 pups, 1625 mitochondria in 160 dendrites) completely abolished spontaneously occurring global calcium transients (paired t-test, p = 6*10–4) and increased the percentage of moving mitochondria (paired t-test, p = 0.035). (G) Higher baseline frequency of spontaneous global calcium transients was correlated with a larger effect of TTX on the percentage of moving mitochondria (n = 7 pups, Pearson correlation, r = 0.85, p = 0.015). (H-I): Mean mitochondrial motility time-locked to the onset of single global calcium transients. The percentage of moving mitochondria did not change significantly between the 2 minutes before and after spontaneously occurring global calcium transients in awake animals (n = 136 transients, paired t-test, p = 0.33).

Figure 2—source data 1. Source data for Figure 2A-D.
Figure 2—source data 2. Source data for Figure 2E-G.
Figure 2—source data 3. Source data for Figure 2H.
elife-62091-fig2-data3.xlsx (125.6KB, xlsx)
Figure 2—source data 4. Source data for Figure 2I.

Figure 2.

Figure 2—figure supplement 1. Relationship between neuronal activity, mitochondrial motility, and age in vivo.

Figure 2—figure supplement 1.

(A) Anti-correlation between the frequency of spontaneous global calcium transients and the percentage of moving mitochondria in animals under isoflurane anesthesia (Spearman’s rank correlation: rs = –0.47, p = 0.04). (B) The frequency of spontaneous global calcium transients increased with age (Spearman’s rank correlation). (C-D) When comparing animals younger than postnatal day (P) 8 to P8 and older, the frequency of spontaneous global calcium transients increased (t-test, n = 5 vs. n = 14, p = 3*10–6) and the percentage of moving mitochondria decreased (t-test, n = 5 vs. n = 14, p = 0.002) in animals under isoflurane anesthesia. (E-F) Mean mitochondrial speed time-locked to the onset of single global calcium transients. The speed of moving mitochondria did not change between the 2 min before and after a single spontaneously occurring calcium transient in vivo (right, Student’s t-test, n = 1029 mitochondria, p = 0.07).
Figure 2—figure supplement 1—source data 1. Source data for Figure 2—figure supplement 1A-D.
Figure 2—figure supplement 1—source data 2. Source data for Figure 2—figure supplement 1E.
Figure 2—figure supplement 1—source data 3. Source data for Figure 2—figure supplement 1F.

Figure 2—figure supplement 2. Synaptic calcium transients in vivo.

Figure 2—figure supplement 2.

(A, B) Two examples of local calcium transients (green) in spines of a layer II/III pyramidal neuron dendrite at postnatal day (P) 13 (red: mitochondrial-DsRed). Arrow heads mark two spines that are activated in A or B.
Open in a new tab

Since overall calcium signaling correlated with mitochondrial motility, we asked whether neuronal activity could directly affect mitochondrial motility. First, we replicated previous experiments performed in cell cultures (DIV14–17) that showed an increase of mitochondrial motility after blocking action potential firing (Li et al., 2004; Chang et al., 2006). Application of the sodium channel blocker TTX (2 µM) to the surface of the brain (P5–12) abolished global calcium transients (Figure 2E) and, as expected, led to a significant increase in mitochondrial motility (Figure 2F, see also Materials and methods for an extended discussion on statistics). Furthermore, the effect of TTX on mitochondrial motility was highly proportional to the frequency of baseline activity (r2 = 0.81; Figure 2G), suggesting that natural patterns of neuronal activity efficiently constrain mitochondrial motility.

We then examined whether spontaneously occurring single global calcium transients affected mitochondrial motility. We compared mitochondrial motility before and after global calcium transients across all recordings by aligning the occurrence of global calcium transients in time and plotting the percentage of moving mitochondria around this time point (Figure 2H). We found that spontaneous global calcium transients did not precede a change in mitochondrial motility (Figure 2H–I) or mitochondrial speed (Figure 2—figure supplement 1E, F). Together, these experiments showed that while neuronal activity modulated mitochondrial motility, global calcium transients – most likely reflecting single back-propagating action potentials and bursts of back-propagating action potentials – were ineffective in doing so. We therefore speculated that synaptic transmission, rather than postsynaptic action potential firing, might regulate mitochondrial motility. To address this possibility we aimed at analyzing the relationship between synaptic activity and mitochondrial motility. Our in vivo recordings showed transmission events at individual synapses (Figure 2—figure supplement 2), but we detected these events too rarely to quantify any possible effect of synaptic activity on mitochondrial motility.

Therefore, we moved to organotypic slice culture preparations, which allow higher signal-to-noise ratio imaging and more stable recordings to investigate the role of transmission at synapses. We obtained cortical slices at P5 or P8 and kept them in culture for at least 3 days before imaging. Slices obtained from older animals showed a trend toward higher spontaneous activity levels (Figure 3A, Figure 3—figure supplement 1A, B) and significantly lower mitochondrial motility than slices obtained from younger animals (Figure 3B, Figure 2—figure supplement 1A, B). As in vivo, spontaneous global calcium transients did not precede changes in mitochondrial motility (Figure 3C–D, Figure 3—figure supplement 1C, D) or speed (Figure 3—figure supplement 1E, F). Thus, mitochondrial motility and its independence of spontaneous global calcium signaling were maintained in slice cultures (Figure 3—figure supplement 1A, B). Together, we reproduced our in vivo observations on mitochondrial motility in slice cultures and, thus, found them suitable to investigate the role of synaptic activity in regulating mitochondrial motility.

Figure 3. Mitochondria stop at synapses after synaptic transmission events.

(A-B) Frequency of global calcium transients and mitochondrial motility in slices obtained from postnatal day (P) 5 and P8 pups. The frequency of spontaneous global calcium transients did not change significantly (n = 15 vs. n = 5 cells, Student’s t-test, p = 0.37). The percentage of moving mitochondria was significantly decreased in slices from older animals (n = 15 [252 mitochondria] vs. n = 5 [85 mitochondria], Student’s t-test, p = 0.02), similar to the in vivo results. (C-D) The percentage of moving mitochondria did not change significantly between the 2 min before and after spontaneously occurring calcium transients in P5 (n = 158 transients, paired t-test, p = 0.07) or P8 slices (paired t-test, n = 101 transients, p = 0.3). (E) Dendritic segment and kymograph showing a mitochondrion approaching and passing an inactive synapse (arrow). (F) Same dendritic segment as in A. A mitochondrion arrived near the same synapse (arrow) after a synaptic calcium transient occurred and stopped within its vicinity (Δx: distance to synapse, Δt: time after synaptic calcium transient). (G-H) Mitochondria moving toward a synapse can show one of two behaviors: they may continue moving (left) or stop near the synapse (right). We compared the percentage of approaching mitochondria that stopped within a specific distance range before individual synaptic calcium transients occurred (G) with that of mitochondria that reached a synapse after a transient (H) within a specific time interval. (I) There was a significant increase in the percentage of stopping mitochondria after a single local calcium transient occurred (distance ≤ 1 µm; interval ≤ 120 s; *p = 6*10–5, chi-squared test). (J) We compared the effect size of mitochondrial arrest at active synapses to a distribution generated by shuffling the time points at which synaptic calcium transients occurred (1000 runs). The observed effect size was within the top 5 percentile of those generated from shuffled data for distances ≤ 1 µm and intervals ≤ 120 s. (K) Quantitative estimation of the spatio-temporal characteristics of mitochondrial arrest (chi-squared test for each distance/interval pair Bonferroni-corrected; distance ≤ 1 µm; interval ≤ 80 s, p = 0.0035; interval ≤ 100 s, p = 0.0016; interval ≤ 120 s, p = 0.0025). (L) Matrix showing the individual chi-squared test p-values from each distance/interval pair. Roughly, p < 0.05 for intervals between 80 and 120 s and distances of up to 5 µm. (Number of observations for K,L: see Figure 3—figure supplement 2B.) (M) Distribution of mitochondrial arrest durations after single spontaneous synaptic events. Shown in dark gray are underestimated durations for data points where mitochondria remained immotile until the end of the recording.

Figure 3—source data 1. Source data for Figure 3A,B.
Figure 3—source data 2. Source data for Figure 3C.
Figure 3—source data 3. Source data for Figure 3D.
Figure 3—source data 4. Source data for Figure 3I.
Figure 3—source data 5. Source data for Figure 3J.
Figure 3—source data 6. Source data for Figure 3K.
Figure 3—source data 7. Source data for Figure 3L.
Figure 3—source data 8. Source data for Figure 3M.

Figure 3.

Figure 3—figure supplement 1. Relationship between neuronal activity, mitochondrial motility, and age in organotypic slice cultures.

Figure 3—figure supplement 1.

(A-B) In slices obtained from postnatal day (P) 5 (A) or P8 pups (B), the frequency of spontaneous global calcium transients increased slightly over days in vitro whereas the motility of mitochondria did not change significantly (Spearman’s rank correlation). (C-D) Mean mitochondrial motility time-locked to the onset of single global calcium transients for slices obtained from P5 (C) and P8 pups (D). (E-F) The speed of moving mitochondria did not change between the 2 min before and after a single spontaneously occurring calcium transient in cells from P5 (E, Student’s t-test, n = 1592 mitochondria, p = 0.1) or P8 pups (F, paired t-test, n = 769 mitochondria, p = 0.5).
Figure 3—figure supplement 1—source data 1. Source data for Figure 3—figure supplement 1A,B.
Figure 3—figure supplement 1—source data 2. Source data for Figure 3—figure supplement 1C.
Figure 3—figure supplement 1—source data 3. Source data for Figure 3—figure supplement 1D.
Figure 3—figure supplement 1—source data 4. Source data for Figure 3—figure supplement 1E.
Figure 3—figure supplement 1—source data 5. Source data for Figure 3—figure supplement 1F.

Figure 3—figure supplement 2. Number of observations for mitochondrial arrest at individual synapses.

Figure 3—figure supplement 2.

(A) Significance levels (Figure 3L) and (B) number of observations for mitochondrial arrest at synapses after synaptic transmission events for each interval and distance bin.
Figure 3—figure supplement 2—source data 1. Source data for Figure 3—figure supplement 2A.
Figure 3—figure supplement 2—source data 2. Source data for Figure 3—figure supplement 2B.
Open in a new tab

In slice cultures, visual cortex layer II/III neurons frequently showed spontaneous calcium transients in spines representing synaptic transmission events at excitatory synapses, as shown previously in the developing visual cortex and hippocampus (Kleindienst et al., 2011; Winnubst et al., 2015; Niculescu et al., 2018). In nine cells (P5 + 3–7 DIV), we identified 157 spines of which 71 (45%) showed spontaneous synaptic calcium transients (376 transients). We asked whether synaptic activity affected the motility of passing mitochondria. We observed that mitochondria typically passed by inactive synapses (Figure 3E), but frequently halted when they reached a synapse that had just been active (Figure 3F). Therefore, we specifically determined whether synaptic calcium transients affected the likelihood that incoming mitochondria stopped at or passed by synapses. To quantify this effect, we compared the percentage of approaching mitochondria that stopped at a synapse before and after the occurrence of a synaptic calcium transient (Figure 3G–H). When we compared the percentage of stopping mitochondria during a 120 s interval before a single synaptic calcium transient occurred with an interval of the same duration after that calcium transient, we found that the percentage of stopping mitochondria increased significantly after the transient (Figure 3I). To answer whether the observed effect size (the difference between the arrest rates before and after a local calcium transient) was likely to occur by chance or not, we performed a bootstrap analysis where we randomized the time points of synaptic calcium transients in our recordings and determined the resulting effect size for a total of 1000 runs. We found that the observed effect size was above the 95 percentile of the randomized effect size distribution (Figure 3J) demonstrating that this effect was unlikely to be observed by chance.

Next, we quantified the effect size for different distances from the synapse and different time intervals after a synaptic calcium transient and found that mitochondrial arrest was most prevalent within distances of up to 5 µm around a synapse and for intervals of 80–120 s after the synaptic event (Figure 3K and L). On average, synaptic activity was associated with an interruption of mitochondrial movement for about 1 min (68 s; Figure 3M). However, this number underestimated the duration of arrest, since one-third of the stopping mitochondria were still immobile at the end of a recording (mean 131 s at a recording duration of 350 s) preventing an exact estimate of the time point when they started moving again. Together, our observations at individual synapses suggested that spontaneous synaptic transmission can capture moving mitochondria in postsynaptic dendrites.

To address the potential mechanism of mitochondrial arrest at active synapses, we first tested whether membrane depolarization leads to mitochondrial arrest. Consistent with previous studies (Li et al., 2004; MacAskill et al., 2009; Faits et al., 2016), we found that an increase of extracellular potassium to 50 mM decreased mitochondrial motility by approximately 50% (Figure 4A and B). This result demonstrated that long-lasting depolarization arrests mitochondria. However, our finding that global calcium transients, which are most likely the consequence of depolarization-induced opening of voltage-gated calcium channels, suggest that depolarization alone is insufficient to stop mitochondria. To test whether synaptic transmission is sufficient to interrupt mitochondrial motility and whether this effect is dependent or independent of action potential firing, we pharmacologically triggered synaptic release while action potential generation was prevented with TTX (Figure 3C–E). After three baseline recordings we applied TTX (1 µM), which blocked all global calcium transients as expected. Next, we stimulated release of synaptic vesicles by applying latrotoxin (Deak et al., 2009) to the bath. The molecular mechanism of latrotoxin-induced transmitter release is unknown. Nevertheless, at the concentration used here (1 nM), latrotoxin specifically triggers synaptic vesicle release through activation of its receptors latrophilin and neurexin that are located at presynaptic terminals (Valtorta et al., 1984; Matteoli et al., 1988; Südhof, 2001). After application of latrotoxin to the bath, the intracellular calcium concentration increased within a few minutes and mitochondrial motility was either entirely suppressed or largely inhibited (Figure 3C–E). Mitochondrial motility recovered only partly after around 1 hr, whereas calcium levels returned to baseline levels within 15–20 min.

Figure 4. Mechanism of activity-induced mitochondrial arrest.

Figure 4.

Open in a new tab

(A, B) Perfusing layer II/III pyramidal neurons (postnatal day [P] 5 + 3–7 DIV) with high-potassium medium [50 mM] triggered a massive influx of calcium and significantly reduced mitochondrial motility within 2 min (n = 9 cells; 107 mitochondria, paired t-test, p = 0.04). (C-E) Stimulating synaptic vesicle release with latrotoxin (LTX) interrupted mitochondrial motility entirely. (C) Example kymographs from recordings during baseline, in the presence of tetrodotoxin (TTX), TTX and LTX, and after washout of LTX. Basal calcium levels were elevated and mitochondrial motility was absent during the presence of LTX. (D) Averaged time course of mitochondrial motility and GCaMP6 ΔF/F0 for the duration of the experiments. Shaded areas and horizontal bars indicate SEMs of values and time points, respectively. (E) Percentage of moving mitochondria across all conditions (p = 0.0058, repeated measures ANOVA, *p = 0.028 (baseline vs. LTX + TTX), *p = 0.022 (TTX vs. TTX + LTX), post hoc t-test with Bonferroni multi-measures correction, n = 5 cells, 92 mitochondria). (F-G) Triggering calcium transients with focal application of glutamate (100 µM) in the presence of TTX did not affect mitochondrial motility significantly (P5 + 3–7 DIV, n = 74 transients from 13 cells, 146 mitochondria, paired t-test, before vs. after, 10.32 ± 2.09 vs. 7.12 ± 1.12, p = 0.1).

Figure 4—source data 1. Source data for Figure 4B.
Figure 4—source data 2. Source data for Figure 4D,E.
Figure 4—source data 3. Source data for Figure 4G.

These experiments indicated that single synaptic transmission events have the capacity to stop mitochondria for 1 to a few minutes and that massive synaptic activation interrupts mitochondrial motility almost entirely for periods of tens of minutes. Finally, we asked whether the transmitter glutamate is responsible for presynaptic release-mediated mitochondrial arrest. We applied single puffs of glutamate (100 µM) to individual dendrites using a pico-spritzer. Focal glutamate delivery triggered calcium increases in the dendrite extending 7–59 µm (27 ± 14 µm; mean ± standard deviation). We analyzed mitochondrial motility before and after glutamate application in the dendritic stretch that responded with a calcium rise. We found that mitochondrial motility was slightly, but not significantly, reduced after glutamate puffs (Figure 4F and G), demonstrating that glutamate is not sufficient to cause vesicle release-mediated mitochondrial arrest.

While the factor that mediates mitochondrial arrest remains unknown, our experiments showed that synaptic vesicle release interrupts mitochondrial transport locally. Since synaptic density (De Felipe et al., 1997) as well as network activity (Rochefort et al., 2009) and thus synaptic vesicle release increase dramatically in the cortex during the second postnatal week, we hypothesized that the temporary recruitment of mitochondria to synapses by spontaneous synaptic activity could contribute to the overall decrease in mitochondrial motility we observed during in vivo development.

To investigate this idea, we designed a computational model for estimating mitochondrial motility in a developing dendrite at different synaptic input frequencies. Since we established a lower bound for the mean duration of immobilization of approximately 70 s, we modeled the effect of synaptic inputs on mitochondrial motility for mean arrest durations of 1–5 min using Gaussian distributions (µ = 1–5 min, σ = 2.5 min). We found that the distributions for 1–3 min were comparable with our observed duration distributions (Figure 5A, Figure 3M). The model showed that changes in synaptic activity could affect mitochondrial motility critically: while low input frequencies (e.g. 0.035 Hz in a 100 µm dendritic segment) reduced mitochondrial motility only marginally from the default state (approximately 10%), higher input frequencies showed clear effects (Figure 5B). For example, at 0.35 Hz, mitochondrial motility was reduced by approximately 50% at steady state. We determined mitochondrial motility for increasing input frequencies and different durations of mitochondrial arrest (Figure 5C). To determine the putative effect of an increase in synaptic activity between the first and the second postnatal week on mitochondrial motility, we estimated the frequency of synaptic inputs received by a stretch of dendrite of 100 µm length. In our previous in vivo study, we found a mean density of 36 active synapses per 100 µm dendrite in visual cortex layer II/III pyramidal neurons at the end of the second postnatal week (P10–15) and transmission occurred 0.6 times per minute at each synapse (Winnubst et al., 2015). We estimated, therefore, that a 100 µm dendrite receives synaptic inputs at a frequency of approximately 0.36 Hz. Anatomical studies showed a 5–10 times increase of synaptic density in the developing sensory cortex between P5 and P11 (De Felipe et al., 1997) and we found here that neuronal activity doubled from the first to the second postnatal week (Figure 2C). Assuming that release probabilities do not change dramatically during this period, synaptic input frequencies should be about 0.02–0.04 Hz at P5. Similarly, we find in in vivo patch-clamp and calcium imaging experiments a 16 times increase in synaptic transmission events in dendrites of V1 layer II/III pyramidal cells from P8 to P13 (AH Leighton et al. 2021, in preparation). As Figure 5C shows, an increase in synaptic input frequency in this range would reduce mitochondrial motility by 30–60% depending on the actual arrest times at synapses. Since we observed a 70% decrease in mitochondrial motility between the first and second postnatal week, our model showed that temporary immobilization of mitochondrial motility through synaptic signaling can mediate a large proportion of this effect. Together, our data suggest a primary role of synaptic vesicle release, but not action potential firing, in the reduction of mitochondrial motility with dendrite maturation.

Figure 5. Model of synaptic input-mediated modulation of mitochondrial motility.

Figure 5.

Open in a new tab

(A) Distribution of mitochondrial arrest durations generated by the model for mean durations of 1–5 min. (B) Changes of mitochondrial motility after onset of simulated synaptic inputs for a mean arrest duration of 1 min. Input frequencies are given as total synaptic inputs along a 100 µm stretch of dendrite. Low input frequencies hardly changed overall mitochondrial motility. Higher input frequencies reduced motility substantially. Steady state was reached after a few minutes. Colored lines: individual simulations, black lines: average of 10 simulations. (C) Relationship between synaptic input frequency and mitochondrial motility for different arrest durations at steady state. The expected increase of synaptic activity from postnatal day (P) 5 to P12 reduced mitochondrial motility by 30–60%, depending on the actual duration of mitochondrial arrest after synaptic transmission.

Discussion

Mitochondrial motility and positioning are fundamental for axon and dendrite development and synaptic plasticity (Courchet et al., 2013; Kimura and Murakami, 2014; Fukumitsu et al., 2015; López-Doménech et al., 2016; Vaccaro et al., 2017; Divakaruni et al., 2018). Moreover, mitochondrial dynamics are affected in many neurological disorders (Chen and Chan, 2009; Deheshi et al., 2013; Misgeld and Schwarz, 2017). Thus, regulation of mitochondrial motility is important for neuronal function; however, to what degree neuronal activity determines mitochondrial motility in intact neuronal circuits has been unclear. By directly observing mitochondrial motility and neuronal activity simultaneously in developing dendrites, we provide evidence that synaptic vesicle release, but not postsynaptic action potential firing, constrains mitochondrial motility and stabilizes mitochondria with increasing age.

Imaging neuronal activity and dendritic mitochondria in vivo demonstrated a dramatic motility reduction of visual cortex layer II/III pyramidal cells during the second postnatal week. Motility reduction progressed in parallel with a strong increase in overall neuronal activity. Given that dendritic mitochondria in retinal explants and axonal mitochondria in the visual cortex show similar decreases in motility (Chang and Reynolds, 2006; Faits et al., 2016; Lewis et al., 2016; Smit-Rigter et al., 2016), our results confirm a general progression towards more stationary mitochondria in intact tissue.

Since there have been conflicting reports on the regulation of mitochondrial motility by natural activity patterns, we set out to investigate their relationship directly. We found that global calcium transients reflecting back-propagating action potentials were unrelated to changes in mitochondrial motility. Considering that manipulations of neuronal action potential firing did trigger changes in mitochondrial motility in the present study and many studies in cell cultures (Li et al., 2004; Chang et al., 2006; MacAskill et al., 2009; Wang and Schwarz, 2009; but see: Beltran-Parrazal et al., 2006), this finding was surprising at first (but see below).

As global calcium transients appeared to be ineffective, we studied the role of transmission at individual synapses and discovered that the likelihood for a passing mitochondrion to stop at a synapse increased significantly when this synapse was active within 2 min before it arrived. That synaptic activation, but not action potential firing, arrests mitochondrial motility is in fact consistent with the majority of observations of activity-dependent regulation of mitochondrial motility in hippocampal and cortical neurons (Li et al., 2004; Chang et al., 2006; MacAskill et al., 2009; Wang and Schwarz, 2009). Since activity manipulations, such as depolarizing neurons by increasing extracellular potassium, blocking action potential firing with TTX, and electric field stimulation affect both action potential firing and synaptic transmission, it is possible that synaptic release changes, but not firing alone, altered mitochondrial motility in these studies.

How can local synaptic activation stop moving mitochondria when their motility is unaffected by spontaneously occurring global calcium transients? Our observations that pharmacological activation of synaptic vesicle release arrests mitochondria, but that neither spontaneous global activation nor focal glutamate application stops moving mitochondria, suggest that a local factor, either by itself or together with glutamate, mediates mitochondrial arrest. A possible candidate is ATP: it is present in synaptic vesicles and released simultaneously with glutamate at excitatory cortical synapses (Khakh, 2001; Burnstock, 2007; Lalo et al., 2016). ATP receptors of the P2X and P2Y families are expressed in cortical pyramidal cell dendrites (Guzman and Gerevich, 2016), trigger postsynaptic depolarizations (Pankratov et al., 2002) and calcium increases (Lalo et al., 1998; Lalo and Kostyuk, 1998), activate CaMKII (Pougnet et al., 2014) and mediate synaptic plasticity (Pankratov et al., 2009; Lalo et al., 2016). Thus, a local factor co-released with glutamate, such as ATP, is most likely required for mitochondrial arrest at active synapses. A co-released factor could, in principle, provide single synapse specificity by generating or – together with glutamate – boosting a local signal that stops mitochondria at active synapses.

Our analysis of changes in mitochondrial motility in relation to spontaneously occurring synaptic transmission events allowed us to determine the spatio-temporal characteristics of this effect quantitatively. We found that mitochondrial arrest was restricted to a segment of dendrite of roughly 5–10 µm distally and proximally from the insertion point of a spine. Mitochondria were stopped when they arrived within 2 min after a synaptic transmission event and remained stationary for 1 min on average. Interestingly, several molecular signaling cascades at the synapse have been described that act on very similar scales in time and space. In particular, several small GTPases become activated within less than a minute after single spine stimulation in short stretches of dendrite (5–10 µm) and stay active for several minutes (e.g. Ras and RhoA; Harvey et al., 2008; Murakoshi et al., 2011). These and other small GTPases, including DRP-1 and Miro-1, which regulate mitochondrial activity and motility, are controlled by intracellular calcium rises and CaMKII activation (MacAskill et al., 2009; Wang and Schwarz, 2009; Fukumitsu et al., 2016; Divakaruni et al., 2018). CaMKII expression increases dramatically in the visual cortex during the second postnatal week (2008 Allen Institute for Brain Science. Allen Developing Mouse Brain Atlas. Available from: https://developingmouse.brain-map.org/). Therefore, synaptic transmission-induced CaMKII phosphorylation requiring, for example, ATP receptor activation may stop mitochondria more frequently with increasing age by activating small GTPases for a few minutes and several micrometers along the dendrite.

To estimate whether mitochondrial arrest through synaptic activity can explain the progressive demobilization of mitochondria in dendrites during development, we employed a computational model. This model indicates that the estimated increase in synaptic activity from the first to the second postnatal week can reduce mitochondrial motility by 30–60%. These numbers are in line with our in vivo observation that blocking synaptic activity with TTX increased mitochondrial motility by 60%. Together, these data show that developmental increases in synaptic activity can explain a large proportion of the motility decrease observed during this period. We speculate that the here described reduction of mitochondrial motility through synaptic activity with increasing age is complemented by a shift in the number of potentially mobile mitochondria toward a pool of stationary mitochondria during development. While for example Miro1 controls temporary mitochondrial arrest, there is no molecular mechanism known for retaining mitochondria permanently at a location in dendrites. Theoretical models suggest that mitochondria are stationary in the absence of Miro1 (MacAskill et al., 2009); however, in Miro1 knockout neurons mitochondrial motility is only mildly affected (Saotome et al., 2008; MacAskill et al., 2009; López-Doménech et al., 2018). Furthermore, to our knowledge a reduction in Miro1 expression or function during development has not been reported. Alternatively, increased tethering of mitochondria may reduce the pool of potentially mobile mitochondria with increasing age. For example, myosin V anchors mitochondria (Pathak et al., 2010), has been proposed to keep mitochondria in a stationary state (Schwarz, 2013; Misgeld and Schwarz, 2017), and is enriched in dendrites (Wang et al., 2008; Konietzny et al., 2019).

The regulation of mitochondrial motility through synaptic activity we describe here may serve developing synapses to efficiently meet their energy demands and calcium handling. In addition, synaptic regulation of mitochondrial trafficking can account to a large degree for the reduction of mitochondrial motility during development and is probably a fundamental process in wiring the developing brain.

Materials and methods

Key resources table.

Reagent type (species) or resource Designation Source or reference Identifiers Additional information
Strain, strain background (Mus musculus, male/ female) C57Bl/6 J, male and female Janvier Labs;
Transfected construct GCaMP6s (species: Rattus norvegicus) Add gene plasmid 40753; Douglas Kim RRID:Addgene_40753 Cloned into pCAG vector
Transfected construct Mitochondrial-DsRed (species: Homo sapiens) Gift from Thomas Misgeld Mitochondrial targeting
sequence from subunit
VIII of human cytochrome
c oxidase causing
mitochondrial localization
as previously described;
Rizzuto et al., 1995;
Li et al., 2004;
MacAskill et al., 2009
Cloned into pCAG vector
Chemical compound, drug TTX 1078, Bio-Techne, Minneapolis, MN
Chemical compound, drug LTX ALX-630–027 C040, Enzo Life Sciences b.v., Farmingdale, NY
Chemical compound, drug Glutamate G1626, Sigma
Software, algorithm MitoMotil This study https://github.com/annikc/MitoMotil (copy archived at swh:1:rev:a4cfb2b4fd66579f63ea5a150a0f9b1b21b89a83, Yalnizyan-Carson, 2021)
Software, algorithm MATLAB The MathWorks https://mathworks.com
Software, algorithm NormCorre Flatiron Institute, Simons Foundation https://github.com/flatironinstitute/NoRMCorre, Pnevmatikakis and Giovannucci, 2021
Software, algorithm Python Python Software Foundation https://www.python.org/
Software, algorithm Elephant library Human Brain Project https://elephant.readthedocs.io/en/latest/
Open in a new tab

Plasmids

To investigate the relationship between neuronal activity and mitochondria, we used the genetically encoded calcium indicator GCaMP6s (Addgene plasmid 40753; Douglas Kim) in combination with mitochondrial-DsRed (mitochondrial targeting sequence from subunit VIII of human cytochrome c oxidase causing mitochondrial localization as previously described; Rizzuto et al., 1995; Li et al., 2004; MacAskill et al., 2009). These plasmids were cloned into pCAGGS, to enable delivery to neurons via in utero electroporation.

Animals and in utero electroporation

All experimental procedures were approved by the institutional animal care and use committee of the Royal Netherlands Academy of Arts and Sciences. To sparsely deliver the plasmids of interest to pyramidal neurons of layer II/III of the visual cortex, pregnant C57Bl/6 J female mice at 16.5 days gestation underwent in utero electroporation surgery. Pregnant females were anesthetized using 3% isoflurane mixed with 1 l/min oxygen and kept under anesthesia with 1.5–2% isoflurane. A midline incision was made and uterine horns were exposed. Plasmid DNA (mitochondrial-DsRed: 0.1 µg/µl, GCaMP6s: 2 µg/µl) was dissolved in 10 mM Tris and 0.05% Fast Green. Approximately 1 µl of this mixture was injected through a pulled capillary pipette in the lateral ventricle of each embryo using a picospritzer (PLI-100, BTX Harvard Apparatus, Holliston, MA). A custom-made square wave isolated pulse generator (voltage of 50 V, 5 pulses, pulse width 50 ms, and 150 ms interval) was used for electroporation. After electroporation, the uterine horns were carefully placed back in the abdomen cavity and the abdomen was sutured. During the surgery, embryos were kept moist with warm saline and the mothers were kept warm using a euthermic pad. Pregnant females were allowed to recover after Lidocaine ointment was applied on the wound for local analgesia and Metacam (1 mg/kg s.c.) was administered for post-operative analgesia. Once the pups were born they were checked before P2 for expression and targeting of V1.

Organotypic slice cultures

Organotypic slice cultures of transfected visual cortex were prepared as follows: at P5 or P8, animals were decapitated quickly, and brains were placed in ice-cold Gey’s balanced salt solution under sterile conditions. Coronal slices (400 μm for P5 and 250 μm for P8) were cut using a tissue chopper (McIlwain) and incubated with serum-containing medium on Millicell culture inserts (Millipore, Merck, New York, NY). Slices were kept in culture for 3–7 days before imaging.

Confocal microscopy of organotypic slice cultures

For confocal imaging, slices were excised from their membrane supports and placed in a flow-through chamber. Slices were continuously perfused with heated (35°C) Hank’s Balanced Salt Solution (HBSS, Fisher Scientific, Waltham, MA, supplemented with in mM: 4.2 NaHCO3, 2.6 CaCl2, 0.1 Trolox). Slices were imaged on a SP5 Leica confocal microscope with a 63× objective (0.9 NA, Leica, Wetzlar, Germany). For imaging we selected neurons that showed the following characteristics: soma localized in upper layer II, apical dendrite pointing to layer I, low basal GCaMP6s fluorescence as well as long and dim mitochondria. Preference was given to isolated cells, to minimize background fluorescence. Apical dendrites (at least 50 µm from the soma) were imaged using an argon laser at 488 nm and power levels between 0.3% and 1%. Time-lapse image stacks (up to six optical sections, 1.2 µm z-spacing), at 0.23 µm per pixel, 350 ms per stack were collected for 350 s, every 10 min, for a total of 10 times per cell. We observed no changes in fluorescence intensity, cell activity levels, or mitochondrial motility levels with time under these conditions. At the end of the experiment, low magnification image stacks (0.23 µm pixel size and 1 µm z-spacing) were collected to localize the recorded dendrite within the dendritic arborization.

In vivo two-photon microscopy

For in vivo imaging, transfected neonatal mice (P5–12) were pre-anesthetized using 3% isoflurane mixed with 1 l/min oxygen and kept under anesthesia with 1–2% isoflurane. A head bar with an opening (4 mm Ø) was attached to the skull above the visual cortex (0–2 mm rostral from lambda and 0–2 mm lateral from the midline) with superglue (Henkel, Düsseldorf, Germany) and dental cement (Heraeus, Hanau, Germany). A small craniotomy above the visual cortex (approximately 1–2 mm Ø) was performed with a needle and forceps and care was taken not to damage the dura mater. The exposed cortical surface was kept moist with cortex buffer (in mM: 125 NaCl, 5 KCl, 10 glucose, 10 HEPES, 2 MgSO4, 2 CaCl2, pH 7.4). For additional stability, a thin layer of 1.5% high electroendosmosis agarose (Biomol, Hamburg, Germany) was applied to the cortical surface. Before imaging, isoflurane was decreased to 0.8% (under anesthesia condition) or 0% (awake condition). A pulsed titanium sapphire laser (Chameleon Vision II, Coherent, Palo Alto, CA) at 900 nm and power up to 30% was used with a 25× water-immersion objective (1.10 NA, Nikon). Time-lapse image stacks (up to five optical sections, 2 µm z-spacing) were obtained at a pixel size of 0.13–0.17 µm and stack rate of 5–10 Hz. Throughout the entire experiment, physiological parameters such as heartbeat and body temperature were monitored, and temperature was controlled using a heating pad.

Pharmacological manipulations

High extracellular potassium in vitro: cells were imaged as described above for at least 20 min and then the imaging medium was replaced by one supplemented with KCl to a final concentration of 50 mM. After 20 min of high potassium incubation and imaging, normal medium was restored and cells were imaged for at least 20 more minutes. Cells did not show dendritic blebbing and in most cells spontaneous activity reappeared, suggesting that they were healthy until the end of the experiment.

To test the role of presynaptic release on mitochondrial motility in dendrites, we imaged the effects of latrotoxin, a synaptic vesicle release stimulator (Deak et al., 2009), on mitochondrial motility and calcium levels in slice cultures during the following conditions sequentially: baseline, TTX, TTX+ LTX, and after LTX washout. TTX was applied (1078, 1 µM, Bio-Techne, Minneapolis, MN) through the bath perfusion. Then LTX (ALX-630–027 C040, 1 nM, Enzo Life Sciences b.v., Farmingdale, NY) was added to the bath and the perfusion was stopped for 10 min. Subsequently, perfusion resumed with TTX containing solution.

Focal glutamate application: TTX (1 µM; No. 1078, Bio-Techne, Minneapolis, MN) was applied through the bath perfusion. A glass pipette with a resistance of approximately 4 MΩ containing glutamate (100 µM) in bath solution was inserted into the slice, approximately 50 µm from the dendrite, and glutamate was applied focally with a Picospritzer at 20 psi (PLI-100, BTX Harvard Apparatus, Holliston, MA). Pulse duration was chosen between 1 and 20 ms to evoke local calcium transients. After placing the glutamate-containing pipette, adjusting the pulse duration and a wait period of at least 10 min, 1–3 single pulses were applied during each recording of 350 s duration.

To block action potential firing in vivo, TTX (2 µM) was prepared in cortex buffer and in agarose. After baseline imaging, the agarose was carefully removed from the top of the brain and the TTX solution in cortex buffer was applied to the surface of the brain for 2 min. Then, this cortex buffer was removed and agarose with TTX was applied to the surface of the brain. Imaging continued as previously. This procedure blocked neuronal activity for the entire imaging period, while the pups’ physiological parameters did not change.

Image analysis

All images were processed using ImageJ software. Images were filtered using a median filter (radius one pixel). Maximal intensity projections of image stacks were generated. All stacks recorded at one dendrite were corrected for motion artifacts due to drift as well as aligned with respect to each other using NoRMCorre (Pnevmatikakis and Giovannucci, 2017).

From the resulting stacks, two-dimensional projections of time (x-axis) vs. displacement (y-axis) were generated for individual dendrites to examine spontaneous global calcium transients as well as mitochondrial motility. Global calcium transients appeared as vertical lines, as there was an increase in intracellular calcium levels throughout the entire dendrite. Immotile mitochondria appeared as horizontal lines, and mitochondrial motility as diagonal lines. The percentage of moving mitochondria was calculated as the number of moving mitochondria divided by the total number of present mitochondria, for each second.

For the analysis of local calcium transients, ΔF/F0 images were calculated where F0 was the average fluorescence of the first 200 frames without apparent calcium transients of the first recording for each cell. Custom-made Matlab scripts aided the manual identification of synaptic events: signals had to last for more than the duration of two frames, did not spread from other sites, and were localized to the spine head.

Statistics

Calcium transients per minute and percent moving mitochondria per 1 s bins are shown in all figures where we compare global calcium transient activity or mitochondrial motility across time or different experimental conditions, respectively. Spearman’s rank correlation was used to detect correlations across time. For single comparisons t-tests (two-tailed, paired, or unpaired) and for multiple comparisons repeated measures ANOVA with post hoc t-tests and Bonferroni multi-measure correction were used. Since for the in vivo measurements (Figure 2) the initial percentages of moving mitochondria were very low, the observed effect may be susceptible to discretization. To test whether age and TTX do indeed affect mitochondrial motility in vivo, we performed additional analyses. The number of moving and stable mitochondria were counted in 2 min bins (MacAskill et al., 2009) and then summed across all recordings for both conditions, respectively, and the resulting contingency tables (see Source Data Tables) were used to perform Fisher’s exact test. We found that the number of observed mitochondria moving was significantly decreased in animals that were P8 or older compared to younger animals (p < 0.00001) and that TTX increased the number of moving mitochondria significantly (p = 0.0034). To test whether there is a significant relationship between the occurrence of individual synaptic calcium transients and the arrest of mitochondria, we used chi-squared tests and performed a bootstrap analysis as described in the Results section.

Modeling mitochondrial motility modulation by synaptic inputs (MitoMotil)

The model of mitochondrial motility was written with Python 3.6 (Python Software Foundation). First, a population of mitochondria (n = 500) was generated where each mitochondrion was initialized with a recovery time drawn from a normal distribution. We ran simulations varying recovery time distribution means over 1–5 min (σ = 2.5 min in each condition). All mitochondria were in the motile pool at the beginning of the simulation run. Synaptic transmission events were generated from a homogeneous Poisson process (from the Elephant library, https://elephant.readthedocs.io/en/latest/) with synaptic input frequencies ranging from 0.001 to 0.5 Hz. We ran each simulation for 1500 s to allow the percentage of immobile mitochondria to reach steady state. Each synaptic transmission event immobilized a variable number of mitochondria randomly selected from the total pool. The proportion of affected mitochondria was drawn from a normal distribution (μ = 0.05, σ = 0.01). These values were based on our observation that single synaptic transmission events affected approximately 5–10 μm of a 100 µm stretch of dendrite. This variable proportion of affected mitochondria was used to calculate the number of mitochondria from the population for immobilization. Affected mitochondria remained immobilized for the duration of the recovery time variable with which they were initialized. If a mitochondrion was already immobilized and selected from the total pool for immobilization by a subsequent event, the immobilization time was extended by the second event.

Acknowledgements

We thank Ginny Farias, Koen Kole, Thomas Misgeld, and Rajeev Rajendran for critically reading the manuscript. Johan Winnubst, Juliette Cheyne, and Alexandra Leighton for custom-made Matlab scripts and Matthew Self for advice on statistics, Thomas Misgeld for the original mitochondrial-DsRed plasmid, and Christiaan Levelt for the pCAGGS construct. In addition, we thank Christiaan Levelt for teaching us the in utero electroporation surgery.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Christian Lohmann, Email: c.lohmann@nin.knaw.nl.

Marla B Feller, University of California, Berkeley, United States.

Gary L Westbrook, Oregon Health and Science University, United States.

Funding Information

This paper was supported by the following grants:

  • Nederlandse Organisatie voor Wetenschappelijk Onderzoek 819.02.017 to Christian Lohmann.

  • Nederlandse Organisatie voor Wetenschappelijk Onderzoek 822.02.006 to Christian Lohmann.

  • Nederlandse Organisatie voor Wetenschappelijk Onderzoek ALWOP.216 to Christian Lohmann.

  • Nederlandse Organisatie voor Wetenschappelijk Onderzoek 865.12.001 to Christian Lohmann.

  • Nederlandse Organisatie voor Wetenschappelijk Onderzoek OCENW.KLEIN.535 to Christian Lohmann.

  • Stichting Vrienden van het Herseninstituut 822.02.006 to Christian Lohmann.

Additional information

Competing interests

None.

none.

Author contributions

Conceptualization, Formal analysis, Investigation, Writing – original draft, Writing – review and editing.

Formal analysis, Software.

Investigation.

Investigation.

Conceptualization, Supervision, Writing – review and editing.

Conceptualization, Investigation, Project administration, Supervision, Writing – original draft, Writing – review and editing.

Ethics

All experimental procedures were approved by the institutional animal care and use committee of the Royal Netherlands Academy of Arts and Sciences. License number: AVD801002015249.

Additional files

Transparent reporting form

Data availability

All data generated or analyzed during this study are included in the manuscript and supporting files. Source data files have been provided for all figures and figure supplements.

References

  1. Beltran-Parrazal L, López-Valdés HE, Brennan KC, Díaz-Muñoz M, de Vellis J, Charles AC. Mitochondrial transport in processes of cortical neurons is independent of intracellular calcium. Am J Physiol Cell Physiol. 2006;291:C1193–C1197. doi: 10.1152/ajpcell.00230.2006. [DOI] [PubMed] [Google Scholar]
  2. Burnstock G. Physiology and pathophysiology of purinergic neurotransmission. Physiological Reviews. 2007;87:659–797. doi: 10.1152/physrev.00043.2006. [DOI] [PubMed] [Google Scholar]
  3. Chang DTW, Honick AS, Reynolds IJ. Mitochondrial trafficking to synapses in cultured primary cortical neurons. The Journal of Neuroscience. 2006;26:7035–7045. doi: 10.1523/JNEUROSCI.1012-06.2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Chang DTW, Reynolds IJ. Differences in mitochondrial movement and morphology in young and mature primary cortical neurons in culture. Neuroscience. 2006;141:727–736. doi: 10.1016/j.neuroscience.2006.01.034. [DOI] [PubMed] [Google Scholar]
  5. Chen H, Chan DC. Mitochondrial dynamics--fusion, fission, movement, and mitophagy--in neurodegenerative diseases. Human Molecular Genetics. 2009;18:R169–R176. doi: 10.1093/hmg/ddp326. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Courchet J, Lewis TL, Lee S, Courchet V, Liou DY, Aizawa S, Polleux F. Terminal axon branching is regulated by the lkb1-nuak1 kinase pathway via presynaptic mitochondrial capture. Cell. 2013;153:1510–1525. doi: 10.1016/j.cell.2013.05.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. De Felipe J, Marco P, Fairen A, Jones EG. Inhibitory synaptogenesis in mouse somatosensory cortex. Cerebral Cortex. 1997;7:619–634. doi: 10.1093/cercor/7.7.619. [DOI] [PubMed] [Google Scholar]
  8. Deak F, Liu X, Khvotchev M, Li G, Kavalali ET, Sugita S, Sudhof TC. {alpha}-Latrotoxin Stimulates a Novel Pathway of Ca2+-Dependent Synaptic Exocytosis Independent of the Classical Synaptic Fusion Machinery. The Journal of Neuroscience. 2009;29:8639–8648. doi: 10.1523/JNEUROSCI.0898-09.2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Deheshi S, Pasqualotto BA, Rintoul GL. Mitochondrial trafficking in neuropsychiatric diseases. Neurobiology of Disease. 2013;51:66–71. doi: 10.1016/j.nbd.2012.06.015. [DOI] [PubMed] [Google Scholar]
  10. Divakaruni SS, Van Dyke AM, Chandra R, LeGates TA, Contreras M, Dharmasri PA, Higgs HN, Lobo MK, Thompson SM, Blanpied TA. Long-Term Potentiation Requires a Rapid Burst of Dendritic Mitochondrial Fission during Induction. Neuron. 2018;100:860–875. doi: 10.1016/j.neuron.2018.09.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Faits MC, Zhang C, Soto F, Kerschensteiner D. Dendritic mitochondria reach stable positions during circuit development. eLife. 2016;5:e11583. doi: 10.7554/eLife.11583. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Fukumitsu K, Fujishima K, Yoshimura A, Wu YK, Heuser J, Kengaku M. Synergistic action of dendritic mitochondria and creatine kinase maintains atp homeostasis and actin dynamics in growing neuronal dendrites. The Journal of Neuroscience. 2015;35:5707–5723. doi: 10.1523/JNEUROSCI.4115-14.2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Fukumitsu K, Hatsukano T, Yoshimura A, Heuser J, Fujishima K, Kengaku M. Mitochondrial fission protein Drp1 regulates mitochondrial transport and dendritic arborization in cerebellar Purkinje cells. Molecular and Cellular Neurosciences. 2016;71:56–65. doi: 10.1016/j.mcn.2015.12.006. [DOI] [PubMed] [Google Scholar]
  14. Guzman SJ, Gerevich Z. P2Y Receptors in Synaptic Transmission and Plasticity: Therapeutic Potential in Cognitive Dysfunction. Neural Plasticity. 2016;2016:e1207393. doi: 10.1155/2016/1207393. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Harris JJ, Jolivet R, Attwell D. Synaptic energy use and supply. Neuron. 2012;75:762–777. doi: 10.1016/j.neuron.2012.08.019. [DOI] [PubMed] [Google Scholar]
  16. Harvey CD, Yasuda R, Zhong H, Svoboda K. The Spread of Ras Activity Triggered by Activation of a Single Dendritic Spine. Science. 2008;321:136–140. doi: 10.1126/science.1159675. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Katz LC, Shatz CJ. Synaptic activity and the construction of cortical circuits. Science. 1996;274:1133–1138. doi: 10.1126/science.274.5290.1133. [DOI] [PubMed] [Google Scholar]
  18. Khakh BS. Molecular physiology of P2X receptors and ATP signalling at synapses. Nature Reviews. Neuroscience. 2001;2:165–174. doi: 10.1038/35058521. [DOI] [PubMed] [Google Scholar]
  19. Kilb W, Kirischuk S, Luhmann HJ. Electrical activity patterns and the functional maturation of the neocortex. The European Journal of Neuroscience. 2011;34:1677–1686. doi: 10.1111/j.1460-9568.2011.07878.x. [DOI] [PubMed] [Google Scholar]
  20. Kimura T, Murakami F. Evidence that dendritic mitochondria negatively regulate dendritic branching in pyramidal neurons in the neocortex. The Journal of Neuroscience. 2014;34:6938–6951. doi: 10.1523/JNEUROSCI.5095-13.2014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Kirkby LA, Sack GS, Firl A, Feller MB. A role for correlated spontaneous activity in the assembly of neural circuits. Neuron. 2013;80:1129–1144. doi: 10.1016/j.neuron.2013.10.030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Kleindienst T, Winnubst J, Roth-Alpermann C, Bonhoeffer T, Lohmann C. Activity-dependent clustering of functional synaptic inputs on developing hippocampal dendrites. Neuron. 2011;72:1012–1024. doi: 10.1016/j.neuron.2011.10.015. [DOI] [PubMed] [Google Scholar]
  23. Konietzny A, González-Gallego J, Bär J, Perez-Alvarez A, Drakew A, Demmers JAA, Dekkers DHW, Hammer JA, Frotscher M, Oertner TG, Wagner W, Kneussel M, Mikhaylova M. Myosin V regulates synaptopodin clustering and localization in the dendrites of hippocampal neurons. Journal of Cell Science. 2019;132:3790–3798. doi: 10.1242/jcs.230177. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Lalo U, Kostyuk P. Developmental changes in purinergic calcium signalling in rat neocortical neurones. Brain Research. Developmental Brain Research. 1998;111:43–50. doi: 10.1016/s0165-3806(98)00120-5. [DOI] [PubMed] [Google Scholar]
  25. Lalo U, Voitenko N, Kostyuk P. Iono- and metabotropically induced purinergic calcium signalling in rat neocortical neurons. Brain Research. 1998;799:285–291. doi: 10.1016/s0006-8993(98)00486-7. [DOI] [PubMed] [Google Scholar]
  26. Lalo U, Palygin O, Verkhratsky A, Grant SGN, Pankratov Y. ATP from synaptic terminals and astrocytes regulates NMDA receptors and synaptic plasticity through PSD-95 multi-protein complex. Scientific Reports. 2016;6:33609. doi: 10.1038/srep33609. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Leighton AH, Lohmann C. The wiring of developing sensory circuits-from patterned spontaneous activity to synaptic plasticity mechanisms. Frontiers in Neural Circuits. 2016;10:71. doi: 10.3389/fncir.2016.00071. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Lewis TL, Turi GF, Kwon SK, Losonczy A, Polleux F. Progressive Decrease of Mitochondrial Motility during Maturation of Cortical Axons In Vitro and In Vivo. Current Biology. 2016;26:2602–2608. doi: 10.1016/j.cub.2016.07.064. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Li Z, Okamoto K, Hayashi Y, Sheng M. The importance of dendritic mitochondria in the morphogenesis and plasticity of spines and synapses. Cell. 2004;119:873–887. doi: 10.1016/j.cell.2004.11.003. [DOI] [PubMed] [Google Scholar]
  30. López-Doménech G, Higgs NF, Vaccaro V, Roš H, Arancibia-Cárcamo IL, MacAskill AF, Kittler JT. Loss of dendritic complexity precedes neurodegeneration in a mouse model with disrupted mitochondrial distribution in mature dendrites. Cell Reports. 2016;17:317–327. doi: 10.1016/j.celrep.2016.09.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. López-Doménech G, Covill-Cooke C, Ivankovic D, Halff EF, Sheehan DF, Norkett R, Birsa N, Kittler JT. Miro proteins coordinate microtubule- and actin-dependent mitochondrial transport and distribution. The EMBO Journal. 2018;37:321–336. doi: 10.15252/embj.201696380. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. MacAskill AF, Rinholm JE, Twelvetrees AE, Arancibia-Carcamo IL, Muir J, Fransson A, Aspenstrom P, Attwell D, Kittler JT. Miro1 is a calcium sensor for glutamate receptor-dependent localization of mitochondria at synapses. Neuron. 2009;61:541–555. doi: 10.1016/j.neuron.2009.01.030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Matteoli M, Haimann C, Torri-Tarelli F, Polak JM, Ceccarelli B, Camilli PD. Differential effect of alpha-latrotoxin on exocytosis from small synaptic vesicles and from large dense-core vesicles containing calcitonin gene-related peptide at the frog neuromuscular junction. PNAS. 1988;85:7366–7370. doi: 10.1073/pnas.85.19.7366. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Misgeld T, Kerschensteiner M, Bareyre FM, Burgess RW, Lichtman JW. Imaging axonal transport of mitochondria in vivo. Nature Methods. 2007;4:559–561. doi: 10.1038/nmeth1055. [DOI] [PubMed] [Google Scholar]
  35. Misgeld T, Schwarz TL. Mitostasis in neurons: Maintaining mitochondria in an extended cellular architecture. Neuron. 2017;96:651–666. doi: 10.1016/j.neuron.2017.09.055. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Morrow EM, Furukawa T, Raviola E, Cepko CL. Synaptogenesis and outer segment formation are perturbed in the neural retina of Crx mutant mice. BMC Neuroscience. 2005;6:5. doi: 10.1186/1471-2202-6-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Murakoshi H, Wang H, Yasuda R. Local, persistent activation of Rho GTPases during plasticity of single dendritic spines. Nature. 2011;472:100–104. doi: 10.1038/nature09823. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Niculescu D, Michaelsen-Preusse K, Güner Ü, van Dorland R, Wierenga CJ, Lohmann C. A BDNF-mediated push-pull plasticity mechanism for synaptic clustering. Cell Reports. 2018;24:2063–2074. doi: 10.1016/j.celrep.2018.07.073. [DOI] [PubMed] [Google Scholar]
  39. Pankratov Y, Lalo U, Krishtal O, Verkhratsky A. Ionotropic P2X purinoreceptors mediate synaptic transmission in rat pyramidal neurones of layer II/III of somato-sensory cortex. The Journal of Physiology. 2002;542:529–536. doi: 10.1113/jphysiol.2002.021956. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Pankratov Y, Lalo U, Krishtal OA, Verkhratsky A. P2X receptors and synaptic plasticity. Neuroscience. 2009;158:137–148. doi: 10.1016/j.neuroscience.2008.03.076. [DOI] [PubMed] [Google Scholar]
  41. Pathak D, Sepp KJ, Hollenbeck PJ. Evidence that myosin activity opposes microtubule-based axonal transport of mitochondria. The Journal of Neuroscience. 2010;30:8984–8992. doi: 10.1523/JNEUROSCI.1621-10.2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Plucińska G, Misgeld T. Imaging of neuronal mitochondria in situ. Current Opinion in Neurobiology. 2016;39:152–163. doi: 10.1016/j.conb.2016.06.006. [DOI] [PubMed] [Google Scholar]
  43. Pnevmatikakis EA, Giovannucci A. NoRMCorre: An online algorithm for piecewise rigid motion correction of calcium imaging data. Journal of Neuroscience Methods. 2017;291:83–94. doi: 10.1016/j.jneumeth.2017.07.031. [DOI] [PubMed] [Google Scholar]
  44. Pnevmatikakis EA, Giovannucci A. NORMCORRE: Non-rigid motion correction. 1b39f82Github. 2021 https://github.com/flatironinstitute/NoRMCorre
  45. Pougnet JT, Toulme E, Martinez A, Choquet D, Hosy E, Boué-Grabot E. ATP P2X receptors downregulate AMPA receptor trafficking and postsynaptic efficacy in hippocampal neurons. Neuron. 2014;83:417–430. doi: 10.1016/j.neuron.2014.06.005. [DOI] [PubMed] [Google Scholar]
  46. Rintoul GL, Filiano AJ, Brocard JB, Kress GJ, Reynolds IJ. Glutamate decreases mitochondrial size and movement in primary forebrain neurons. The Journal of Neuroscience. 2003;23:7881–7888. doi: 10.1523/JNEUROSCI.23-21-07881.2003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Rizzuto R, Brini M, Pizzo P, Murgia M, Pozzan T. Chimeric green fluorescent protein as a tool for visualizing subcellular organelles in living cells. Current Biology. 1995;5:635–642. doi: 10.1016/S0960-9822(95)00128-X. [DOI] [PubMed] [Google Scholar]
  48. Rochefort NL, Garaschuk O, Milos RI, Narushima M, Marandi N, Pichler B, Kovalchuk Y, Konnerth A. Sparsification of neuronal activity in the visual cortex at eye-opening. PNAS. 2009;106:15049–15054. doi: 10.1073/pnas.0907660106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Sanes JR, Yamagata M. Many paths to synaptic specificity. Annual Review of Cell and Developmental Biology. 2009;25:161–195. doi: 10.1146/annurev.cellbio.24.110707.175402. [DOI] [PubMed] [Google Scholar]
  50. Saotome M, Safiulina D, Szabadkai G, Das S, Fransson A, Aspenstrom P, Rizzuto R, Hajnoczky G. Bidirectional Ca2+-dependent control of mitochondrial dynamics by the Miro GTPase. PNAS. 2008;105:20728–20733. doi: 10.1073/pnas.0808953105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Schwarz TL. Mitochondrial trafficking in neurons. Cold Spring Harbor Perspectives in Biology. 2013;5:A011304. doi: 10.1101/cshperspect.a011304. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Sengpiel F, Kind PC. The role of activity in development of the visual system. Current Biology. 2002;12:R818–R826. doi: 10.1016/S0960-9822(02)01318-0. [DOI] [PubMed] [Google Scholar]
  53. Sheng ZH, Cai Q. Mitochondrial transport in neurons: impact on synaptic homeostasis and neurodegeneration. Nature Reviews. Neuroscience. 2012;13:77–93. doi: 10.1038/nrn3156. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Smit-Rigter L, Rajendran R, Silva CAP, Spierenburg L, Groeneweg F, Ruimschotel EM, van Versendaal D, van der Togt C, Eysel UT, Heimel JA, Lohmann C, Levelt CN. Mitochondrial dynamics in visual cortex are limited In vivo and not affected by axonal structural plasticity. Current Biology. 2016;26:2609–2616. doi: 10.1016/j.cub.2016.07.033. [DOI] [PubMed] [Google Scholar]
  55. Südhof TC. alpha-Latrotoxin and its receptors: neurexins and CIRL/latrophilins. Annual Review of Neuroscience. 2001;24:933–962. doi: 10.1146/annurev.neuro.24.1.933. [DOI] [PubMed] [Google Scholar]
  56. Tran NM, Zhang A, Zhang X, Huecker JB, Hennig AK, Chen S. Mechanistically distinct mouse models for crx-associated retinopathy. PLOS Genetics. 2014;10:e1004111. doi: 10.1371/journal.pgen.1004111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Vaccaro V, Devine MJ, Higgs NF, Kittler JT. Miro1-dependent mitochondrial positioning drives the rescaling of presynaptic Ca2+ signals during homeostatic plasticity. EMBO Reports. 2017;18:231–240. doi: 10.15252/embr.201642710. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Valtorta F, Madeddu L, Meldolesi J, Ceccarelli B. Specific localization of the alpha-latrotoxin receptor in the nerve terminal plasma membrane. The Journal of Cell Biology. 1984;99:124–132. doi: 10.1083/jcb.99.1.124. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Wang Z, Edwards JG, Riley N, Provance DW, Jr, Karcher R, Li X-D, Davison IG, Ikebe M, Mercer JA, Kauer JA, Ehlers MD. Myosin Vb mobilizes recycling endosomes and AMPA receptors for postsynaptic plasticity. Cell. 2008;135:535–548. doi: 10.1016/j.cell.2008.09.057. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Wang X, Schwarz TL. The mechanism of Ca2+ -dependent regulation of kinesin-mediated mitochondrial motility. Cell. 2009;136:163–174. doi: 10.1016/j.cell.2008.11.046. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Winnubst J, Cheyne JE, Niculescu D, Lohmann C. Spontaneous activity drives local synaptic plasticity in vivo. Neuron. 2015;87:399–410. doi: 10.1016/j.neuron.2015.06.029. [DOI] [PubMed] [Google Scholar]
  62. Yalnizyan-Carson A. MitoMotil. Software Heritage. 2021 https://archive.softwareheritage.org/swh:1:rev:a4cfb2b4fd66579f63ea5a150a0f9b1b21b89a83

Decision letter

Editor: Marla B Feller1

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Thank you for submitting your article "Synaptic inputs, but not action potentials, regulate motility of dendritic mitochondria in the developing visual cortex" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing editors, and the evaluation has been overseen by Gary Westbrook as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission. The editors and reviewers have judged that your manuscript is of interest, but as described below additional experiments are required before it could be published.

We would like to draw your attention to changes in our revision policy in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option.

Overview of discussion between Editors and Reviewers:

This study explores the relationship between calcium signaling and mitochondrial motility in dendrites of developing visual cortex neurons. The manuscript provides interesting new insights into the regulation of mitochondrial movements in dendrites. Specifically it explains apparent previous discrepancies in studies on the role of activity in regulating mitochondrial motility by identifying differences between the impact of synaptic activity vs. action potential firing. The experiments appear carefully conducted, the findings are well illustrated, and communicated clearly in the text.

The reviewers agree that there are two interesting aspects of the study.

1) The in vivo demonstration that mitochondrial motility decreases in developing dendrites as activity increases and that this correlation holds for the effects of TTX on motility.

2) The in vitro finding that synaptic activity, rather than spikes, are what matters. We think if both these points were made convincingly, the impact of the study would be appropriate for eLife. However, the reviewers think these claims were not fully supported by the data.

The reviewers think the following needs to be done:

(a) cleaning up the analysis of the effects of local events on passing mitochondria to take into account of low overall motility and clarify effects at short intervals;

(b) showing that local calcium transients occur in vivo and have the same effect on mitochondrial motility as they do in vitro;

and

(c) locally activating synapses (e.g., glutamate uncaging or puffing) to show that this is indeed what drives mitochondrial arrests. To be clear -- (a) is required and either (b) or (c) should be sufficient.

Below is a consolidation of the original comments from individual reviewers that led to the above discussion and decision-- in some cases it may seem repetitive with the above, but it may be helpful to see how the reviewers struggled with some analysis as presented.

Summary:

In this study the authors explore the relationship between neural activity and mitochondrial motility in the dendrites of cortical neurons during development. The authors conduct simultaneous imaging of calcium and fluorescently tagged mitochondria motion both in vivo and in an organotypic slice preparation. They show that there is an increase in the frequency of global calcium transients with age and a reduction in motility. However, there is essentially no correlation between motility and spontaneous global calcium transients on a dendrite-by-dendrite level. Rather, they argue that mitochondria motility is influenced by synaptic activity. The data to support this are two-fold: first, mitochondria are more likely to stall near synapses if those synapses have been recently active; second, latrotoxin (which induces exocytosis) but not TTX leads to complete cessation of mitochondria movement. Finally, the authors construct a model to simulate how changes in synaptic activation impact motility making a few assumptions of underlying the arrest duration of mitochondria. The model suggests that given the observed local effects of synaptic activity on mitochondrial movements, the developmental decline in mitochondrial motility could be accounted for by the simultaneous increase in the density of synapses.

Essential revisions:

1. Given there appear to be very few motile mitochondria for any given dendrite, the authors need to be careful as to how they quantify their data. The example shown in Figure 1 appears to have perhaps 1 or 2 motile mitochondria our of 8-10. The quantification of the data is “percent of motile mitochondria”. As seen in Figure 1D, this corresponds to “bins” of percentile changes of 5-10%. Yet, the entire range used to establish the correlation in Figures 2A and 2B is 12%! Hence there is not much confidence that the resulting correlation is meaningful. There is even less confidence that the 1% change in percent observed in TTX (Figure 2F).

2. In figure 2 the authors quantify global Ca events and mitochondrial motility in dendrites in vivo over a range of postnatal days. As has been demonstrated by others (e.g. Faits et al., 2016), the authors see a progressive decrease in motility of mitochondria. They further demonstrate a negative correlation between global Ca events and mitochondria motility. The authors do not present whether they are able to detect spine-specific spontaneous events in-vivo and how spine-specific events change during this period of development. This seems important given the distinction made in vitro.

3. Figure 4 is quite critical to the study but several aspects were confusing. The authors argue that mitochondria are halted near a synapse after the synapse was active. This quantification depends on the length scale that means “near” and the time scale that means “after”. The authors need to clarify this quantification much more. Some questions:

– Figure 4E is based on 2 microns and 120 seconds compared to “before”. Does “before” mean less than 120 seconds or compared only to the time prior to spine activation?

– Figure 4F: the terms of the bootstrap analysis need to be clearly stated – is the hypothesis that seeing a reduction in motility >120 seconds is more than you would expect by chance if all the time points between 0 and 120 seconds are included?

– Figure 4G: I am quite confused here. Let’s take the lightest pink plot. Does this mean if you look at the interval 20 second after the synaptic activation that there are more mitochondria stopping prior to calcium transient than after?

Given all of these questions, the authors must justify 120 seconds as the most relevant time scale. Particularly if they get an opposite sign effect if you look at 20 seconds!

4. The primary manipulation in the paper is the application of LTX in the presence of TTX. This manipulation demonstrates that release of neurotransmitter in the absence of Aps can induce the stopping of mitochondria. However, it seemed unsatisfying that this manipulation was global in nature and not more local (e.g glutamate uncaging/ glutamate puffing/ stimulation of local axons) given that the authors make the distinction earlier between global and local calcium measurements. The authors discuss the potential mechanism by which a local (synaptically induced) calcium transient and a global (backpropagating AP induced) Ca transient could differentially regulate mitochondrial trafficking briefly I the discussion. Mechanistic findings of these differences would certainly elevate our understanding and the paper.

5. The latrotoxin effect is quite dramatic. Though it is true that latrotoxin induces exocytosis, my understanding is that latrotoxin does this causing a massive increase in intracellular calcium and influx of water. Hence latrotoxin may impact mitochondrial motility in a manner independent of synaptic release. Given that the TTX is also likely to impact synaptic release, this seems like the most likely explanation.

6. The authors argue synaptic activity not global cellular activation stops mitochondria. Hence TTX has a small effect and latrotoxin has a big one. But TTX also impact synaptic events as well as global calcium events. So why is there not a bigger impact of TTX on mitochondria? Do they authors argue that most of the synaptic activity is independent of evoked release? This point can be clarified.

7. In Figure 4, the authors present data on the effects of local calcium transients on mitochondrial motility. Panel G indicates that motility is enhanced shortly after the calcium transient and decreases after longer time intervals. This observation appears to approach or reach statistical significance. The authors should clarify whether this is a consistent observation and discuss what mechanisms may account for it.

8. It is not clear for many of the figures (e.g. Figure 2, Figure 3) what the size/ content of the dataset that is being analyzed. In particular, how many mitochondria are being tracked from how many dendrites from how many neurons from how many slices/animals. In figure 4 the text reads “In nine cells (P5 + 3‐7 DIV), we identified 157 spines of which 140 71 (45%) showed spontaneous synaptic calcium transients(376 transients).” This was very helpful to the reader to give an idea of the dataset and it would be helpful to include similar statements for other datasets. This is particularly important given that much of the data is presented as normalized data (e.g. percent moving mitochondria).

9. The authors develop a model that suggests that the local stopping of mitochondria in response to synaptic activity can in large part account for the age-dependent decline in mitochondrial mobility observed in vivo (~70%). The authors’ model suggests that this is only true if the mean arrest duration of mitochondria is around 5 minutes. The data in Figure 4I suggests that the mean arrest duration of mitochondria is about 1 min, but as the authors point out, this is likely an underestimate due to the fact that ~8/20 mitochondria remain at rest when their imaging session ended. Given the importance of this parameter in their model, longer imaging sessions would be necessary to determine mean arrest time more accurately. The data looks like in fact there may be a multimodal distribution of mitochondrial arrest time. As it stands, I don’t feel that the model provides much additional understanding.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Activity-dependent regulation of mitochondrial motility in developing cortical dendrites" for further consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing editors, and the evaluation has been overseen by Gary Westbrook as the Senior Editor.

The manuscript has been greatly improved and two the reviewers felt their concerns were adequately addressed. There remains a major issue regarding the quantification of the "%-motile" data brought up by Reviewer #1 that has not been satisfied. After discussion with all the reviewers it was deemed that this was something that needs to be addressed. We suggest that you consult with a statistician concerning this issue. We will not be able to make a final decision until that issue is adequately addressed.Reviewer #1:

Most of my concerns have been addressed in the revision. There is still one major concern remaining – and I have repeated it here.

The remaining concern was the first one I raised in first review and had to do with quantification of % motile mitochondria. Repeating what I said in this first review --

The example shown in Figure 1 appears to have perhaps 1 or 2 motile mitochondria out of 8-10. The quantification of the data is "percent of motile mitochondria". As seen in Figure 1D, this corresponds to "bins" of percentile changes of 5-10%. Yet, the entire range used to establish the correlation in Figures 2A and 2B is 12%! Hence there is not much confidence that the resulting correlation is meaningful. There is even less confidence that the 1% change in percent observed in TTX is meaningful (Figure 2F).

The only answer offered by authors is that this is the same method used in a previous study (MacAskill et al., 2009) but with shorter time windows. However, this is not a question of methods, this is a question of statistics and how reliable the effects are. As noted by authors in this manuscript, the longer time windows in the previous study led to a larger range of percent motile mitochondria, ranging from 0-50%. However this larger range is less susceptible to discretization errors, which is my concern here.

At a minimum, comparisons across conditions (Figure 2) should not be based on t-tests, which assume a normal distribution around a mean. With discretized date like this, the assumption of a normal distribution is incorrect – evidence of this can be seen in Figure 2B where the variance at P8 and P10 actually goes to zero.

An example of the implications of this is Figure 2F – their claim that TTX effects motility in vivo. They find the percent motile goes from 2% to 3% – again a finding based on individual measurements that are binned at >5%. Though they find this incredibly small effect significant using a t-test, this sort of small effect is susceptible to discretization errors.

The authors are requested to perhaps use longer time periods for their session so like the previous paper, it won't as susceptible to this error. At a minimum, the authors should perform a Fisher's exact test rather than a t-test to make statements regarding whether their effects on motility are significant.

Reviewer #2:

The authors have satisfactorily addressed my previous concerns.

Reviewer #3:

The authors have responded to my concerns regarding the clarity of the paper and analysis. They have addressed my concern regarding the lack of a local manipulation with new experiments (glutamate puffing). This led to a surprising finding that they discuss. Other concerns were addressed through more careful discussion in the manuscript. I feel the manuscript is suitable for publication.

eLife. 2021 Sep 7;10:e62091. doi: 10.7554/eLife.62091.sa2

Author response

Essential revisions:

1. Given there appear to be very few motile mitochondria for any given dendrite, the authors need to be careful as to how they quantify their data. The example shown in Figure 1 appears to have perhaps 1 or 2 motile mitochondria our of 8-10. The quantification of the data is “percent of motile mitochondria”. As seen in Figure 1D, this corresponds to “bins” of percentile changes of 5-10%. Yet, the entire range used to establish the correlation in Figures 2A and 2B is 12%! Hence there is not much confidence that the resulting correlation is meaningful. There is even less confidence that the 1% change in percent observed in TTX (Figure 2F).

Our analysis is essentially identical to those previously established and accepted (e.g. MacAskill et al., 2009). The only difference is that we determine the percentage of mitochondria for a relatively short interval (bin, 1 second) compared to MacAskill (2 minutes). We intentionally chose 1 second, because it allowed us to compare mitochondrial motility across experimental conditions and for any durations, which was important to infer the exact temporal characteristics of the relationship between synaptic transmission and mitochondrial motility (New Figure 3). As a consequence, our motility rates are lower than reported in previous studies. To confirm that our analysis is equivalent to previous analyses we reanalyzed some of our data as described by MacAskill et al. and found that we get roughly 3x higher values for mitochondrial motility when we chose a 2-minute interval, and these values are in line with previously reported mitochondrial motility rates. Thus, we feel that we use an established analysis approach and we are convinced that this approach allows comparing mitochondrial motility across different conditions, in particular considering that we tracked large numbers of mitochondria (we present these numbers now for all data, see also Point 8 below).

2. In figure 2 the authors quantify global Ca events and mitochondrial motility in dendrites in vivo over a range of postnatal days. As has been demonstrated by others (e.g. Faits et al., 2016), the authors see a progressive decrease in motility of mitochondria. They further demonstrate a negative correlation between global Ca events and mitochondria motility. The authors do not present whether they are able to detect spine-specific spontaneous events in-vivo and how spine-specific events change during this period of development. This seems important given the distinction made in vitro.

We indeed observe local calcium transients in spines in vivo as we do in slices. We mention this now in the Results section and show examples in our new Supplementary Figure 2. Unfortunately, however, we detect these events very rarely, probably because of the lower signal-to-noise ratio and higher incidence of movements in the in vivo recordings. Therefore, we cannot make statements about the relationship between mitochondrial motility and synaptic signaling or the development of synaptic activity in vivo. From a different study in our lab that we prepare for publication currently, we know that synaptic activity increases dramatically during the second postnatal week in vivo. Combined voltage-clamp and calcium imaging recordings at different ages in vivo show that the number of transmission events along dendrites increases 16 times during the second postnatal week. We reference this study in the revised manuscript.

3. Figure 4 is quite critical to the study but several aspects were confusing. The authors argue that mitochondria are halted near a synapse after the synapse was active. This quantification depends on the length scale that means “near” and the time scale that means “after”. The authors need to clarify this quantification much more. Some questions:

We are sorry that the description of our analysis has been confusing. We added new analyses, expanded their description, and made the figure more intuitive. We hope that the reviewers will find this section much clearer now. We addressed the specific points as follows:

– Figure 4E is based on 2 microns and 120 seconds compared to “before”. Does “before” mean less than 120 seconds or compared only to the time prior to spine activation?

For the analysis shown in this panel (now Figure 3I) we compared the percentage of stopping mitochondria during a 120 second interval after a calcium transient occurred with the percentage of stopping mitochondria during a 120 second interval before the transient. When we compare intervals of different durations (e.g. 80, 100, 120 seconds) we always compare an interval of the same duration before and after the occurrence of a calcium transient. We have made this clearer now in the text and by labeling interval durations as “< = interval” within the Figure to indicate that this includes the entire period from the occurrence of a calcium transient until the end of that period.

– Figure 4F: the terms of the bootstrap analysis need to be clearly stated – is the hypothesis that seeing a reduction in motility >120 seconds is more than you would expect by chance if all the time points between 0 and 120 seconds are included?

We performed this bootstrap analysis to test whether the effect size that we observed in the previous panel (previously Figure 4E, now 3I), is higher than one would expect by chance. To test this, we randomized the time points at which local calcium transients occurred and determined the effect size for 1000 randomized data sets. The result shows that the actual effect size is above the 95 percentile of the randomizations, indicating that this is not likely to occur by chance. We have made this clearer in the manuscript now.

– Figure 4G: I am quite confused here. Let’s take the lightest pink plot. Does this mean if you look at the interval 20 second after the synaptic activation that there are more mitochondria stopping prior to calcium transient than after?

We are sorry that this plot was confusing. We have updated it and hope it is clearer now. As the reviewer points out the most confusing aspect was that the 20 second line appeared to suggest a reduction in mitochondrial arrest for this short interval after the occurrence of a calcium transient. This decrease was not significant and was actually based on a very low number of observations (< 10, bottom row in Figure 3 supplement 2, Figure B). Therefore, we removed this line from the plot. We believe that the effect ramps up until pprox.. 1 minute after a calcium transient and becomes more significant with somewhat longer intervals since they include more observations (see Figure 3 supplement 2).

Given all of these questions, the authors must justify 120 seconds as the most relevant time scale. Particularly if they get an opposite sign effect if you look at 20 seconds!

We hope that our new manuscript and the answers to the reviewers’ questions above help making clearer how we derived the spatio-temporal characteristics of mitochondrial arrest from our data. In particular, we find that mitochondria stop moving within 2 minutes after a synaptic calcium transient occurs, probably as a consequence of a second messenger cascade that operates in this time frame. The effect becomes more significant with increasing interval duration (up to 2 minutes) because more observations are included. Therefore, we started with the 120 second interval at short distances to show that mitochondrial arrest is robust and statistically significant in panels I and J of our new Figure 3 and then moved ahead to characterize the spatio-temporal profile of the effect in panels K, L and M of the same Figure.

4. The primary manipulation in the paper is the application of LTX in the presence of TTX. This manipulation demonstrates that release of neurotransmitter in the absence of Aps can induce the stopping of mitochondria. However, it seemed unsatisfying that this manipulation was global in nature and not more local (e.g glutamate uncaging/ glutamate puffing/ stimulation of local axons) given that the authors make the distinction earlier between global and local calcium measurements. The authors discuss the potential mechanism by which a local (synaptically induced) calcium transient and a global (backpropagating AP induced) Ca transient could differentially regulate mitochondrial trafficking briefly I the discussion. Mechanistic findings of these differences would certainly elevate our understanding and the paper.

The reviewer asks for additional evidence that the observed effect of synaptic transmission on mitochondrial arrest is local and a more in-depth analysis of the mechanisms underlying this phenomenon. We have now performed the experiments the reviewer proposed: we puffed glutamate locally and measured both the postsynaptic increase in calcium triggered by glutamate as well as the effect on mitochondrial motility. We observed that glutamate increased the intracellular calcium concentration locally; however, we find only a small, insignificant decrease in mitochondrial motility during a 2 minute period after local glutamate puffing. This finding indicates that glutamate is not causing release-mediated arrest, at least not in the absence of other factors. Thus, we have constrained the possible mechanisms, but unfortunately, could not identify the main factor that mediates mitochondrial arrest at synapses. In the new manuscript we describe a potential mechanism for synaptic transmission induced mitochondrial arrest based on these findings in the discussion.

5. The latrotoxin effect is quite dramatic. Though it is true that latrotoxin induces exocytosis, my understanding is that latrotoxin does this causing a massive increase in intracellular calcium and influx of water. Hence latrotoxin may impact mitochondrial motility in a manner independent of synaptic release. Given that the TTX is also likely to impact synaptic release, this seems like the most likely explanation.

We agree with the reviewer’s notion that the mechanism of latrotoxin induced release of synaptic vesicles is not entirely clear. However, latrotoxin’s action is highly restricted to presynaptic terminals and very specific to synaptic vesicle release. First, the receptors of latrotoxin (latrophilin/CIRL and neurexins) are specifically localized to presynaptic terminals (Valtorta et al., 1984). Furthermore, at the low concentration we used in these experiments (1 nM), latrotoxin triggers vesicle release independently of calcium and affects exclusively readily releasable synaptic vesicles and the reserve pool, but no other cellular organelles (Südhof, 2001) and it does not trigger release of other release-competent organelles, such as dense core vesicles (Matteoli et al., 1988). Therefore, presynaptic vesicle release most likely mediates the strong effect of latrotoxin on dendritic mitochondria. As the reviewer points out, TTX will diminish synaptic release as a consequence of the absence of action potential firing and thus, our observation that TTX increases mitochondrial motility in vivo (Figure 2F, G) is in perfect agreement with our interpretation that vesicle release triggers mitochondrial arrest.

6. The authors argue synaptic activity not global cellular activation stops mitochondria. Hence TTX has a small effect and latrotoxin has a big one. But TTX also impact synaptic events as well as global calcium events. So why is there not a bigger impact of TTX on mitochondria? Do they authors argue that most of the synaptic activity is independent of evoked release? This point can be clarified.

The reviewer wonders why TTX, which, as they point out, affects not only action potential firing, but also reduces synaptic vesicle release, has a relatively small effect on mitochondrial motility, whereas latrotoxin blocks it entirely. TTX and latrotoxin affect vesicle release in opposite directions: TTX leads to a reduction of release (but not an entire block as the reviewers note) and increases mitochondrial motility in vivo by 60% (Figure 2F), whereas latrotoxin stimulates vesicle release and blocks motility entirely in slices. The effect size of TTX should depend on the level of spontaneous activity in a given cell before TTX application: if activity is low and mitochondrial motility mostly unhindered by synaptic blockade, TTX will have a small effect. Conversely, when activity levels are high, TTX should have a large effect. This is exactly what we observe (Figure 2G). Furthermore, the 60% increase in mitochondrial motility after TTX application in vivo is perfectly in line with the estimate of our model for the percentage of moving mitochondria that halt in response to synaptic activity. In slice cultures, we see a smaller effect of TTX. This is probably a consequence of the fact that the slice cultures were a bit less mature and showed lower levels of spontaneous activity.

7. In Figure 4, the authors present data on the effects of local calcium transients on mitochondrial motility. Panel G indicates that motility is enhanced shortly after the calcium transient and decreases after longer time intervals. This observation appears to approach or reach statistical significance. The authors should clarify whether this is a consistent observation and discuss what mechanisms may account for it.

As pointed out above, we agree with the reviewers that our original Figure 4G was confusing. The line for the 20 second interval indeed appeared to indicate that mitochondrial motility was increased during this interval. However, this is not the case: we do not find a significant effect for this interval and the number of observations for this interval was very low, indicating that the downward deflection was spurious. To avoid the false impression that mitochondrial motility is increased for short intervals we have removed the 20 second interval data from this figure (New Figure 3K).

8. It is not clear for many of the figures (e.g. Figure 2, Figure 3) what the size/ content of the dataset that is being analyzed. In particular, how many mitochondria are being tracked from how many dendrites from how many neurons from how many slices/animals. In figure 4 the text reads “In nine cells (P5 + 3‐7 DIV), we identified 157 spines of which 140 71 (45%) showed spontaneous synaptic calcium transients(376 transients).” This was very helpful to the reader to give an idea of the dataset and it would be helpful to include similar statements for other datasets. This is particularly important given that much of the data is presented as normalized data (e.g. percent moving mitochondria).

We have added the number of experiments, animals and mitochondria tracked to the figure legends. In addition, we made a new supplemental Figure (sFigure 4, Figure in response to Point 3, above) to give an overview of the number of observations for each distance/interval bin in our new Figure 3K, L (previously 4G, H).

9. The authors develop a model that suggests that the local stopping of mitochondria in response to synaptic activity can in large part account for the age-dependent decline in mitochondrial mobility observed in vivo (~70%). The authors’ model suggests that this is only true if the mean arrest duration of mitochondria is around 5 minutes. The data in Figure 4I suggests that the mean arrest duration of mitochondria is about 1 min, but as the authors point out, this is likely an underestimate due ’o the fact that ~8/20 mitochondria remain at rest when their imaging session ended. Given the importance of this parameter in their model, longer imaging sessions would be necessary to determine mean arrest time more accurately. The data looks like in fact there may be a multimodal distribution of mitochondrial arrest time. As it stands, I don’t feel that the model provides much additional understanding.

The reviewer points out that we modeled the effect of synaptic activity on mitochondrial motility for several mean arrest durations and that we find a maximal effect of 70% at a mean arrest duration of 5 minutes. In fact, we believe that this is an upper bound of the effect. The real mean arrest duration is between 1 and 5 minutes and we estimate in the manuscript that the increase in synaptic transmission between the first and the second postnatal week would reduce mitochondrial motility by 30-60%. Since we find mitochondrial motility to be reduced by 70% between the first and the second postnatal week, our model suggests that 43-86% of the developmental effect can be explained by synaptic transmission induced arrest. Thus, in our mind the model demonstrates biological relevance of this mechanism for the frequently observed, but so far unexplained, observation that mitochondrial motility decreases during development. The reviewer argues that it would be necessary to image for longer periods of time to get an even more precise estimate of the real arrest duration. Unfortunately however, we are not comfortable having longer time-lapse recordings since we need to avoid any risk of photo-damage on the cells. Finally, the reviewer suggests that there is a bimodal distribution of arrest durations. While we believe that this could indeed be a possibility, we find that a unimodal distribution as plotted in Figure 5A can look like a bimodal distribution, simply because of the constraint that there are no negative arrest durations, causing a relatively high peak in the lowest bin of the histogram. This is now stated more clearly in the text.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #1

Most of my concerns have been addressed in the revision. There is still one major concern remaining – and I have repeated it here.

The remaining concern was the first one I raised in first review and had to do with quantification of % motile mitochondria. Repeating what I said in this first review --

The example shown in Figure 1 appears to have perhaps 1 or 2 motile mitochondria out of 8-10. The quantification of the data is "percent of motile mitochondria". As seen in Figure 1D, this corresponds to "bins" of percentile changes of 5-10%. Yet, the entire range used to establish the correlation in Figures 2A and 2B is 12%! Hence there is not much confidence that the resulting correlation is meaningful. There is even less confidence that the 1% change in percent observed in TTX is meaningful (Figure 2F).

The only answer offered by authors is that this is the same method used in a previous study (MacAskill et al. 2009) but with shorter time windows. However, this is not a question of methods, this is a question of statistics and how reliable the effects are. As noted by authors in this manuscript, the longer time windows in the previous study led to a larger range of percent motile mitochondria, ranging from 0-50%. However this larger range is less susceptible to discretization errors, which is my concern here.

At a minimum, comparisons across conditions (Figure 2) should not be based on t-tests, which assume a normal distribution around a mean. With discretized date like this, the assumption of a normal distribution is incorrect – evidence of this can be seen in Figure 2B where the variance at P8 and P10 actually goes to zero.

An example of the implications of this is Figure 2F – their claim that TTX effects motility in vivo. They find the percent motile goes from 2% to 3% -- again a finding based on individual measurements that are binned at >5%. Though they find this incredibly small effect significant using a t-test, this sort of small effect is susceptible to discretization errors.

The authors are requested to perhaps use longer time periods for their session so like the previous paper, it won't as susceptible to this error. At a minimum, the authors should perform a Fisher's exact test rather than a t-test to make statements regarding whether their effects on motility are significant.

We have now addressed the concern of Reviewer 1 as follows:

We agree with the reviewer that the statistics we used for comparisons in Figure 2 do not entirely rule out the possibility that the effects we observe are spurious. In consultation with Dr. Matthew Self, an expert in biomedical statistics, we reanalyzed the relevant data in a way that allowed us to follow the advice of Reviewer 1. We summed the number of moving versus stable mitochondria for every 2 minute bin from all recordings during the respective conditions to generate a contingency table (see Author response Table 1). We performed a Fisher’s Exact test on these contingency tables. We find that there are indeed highly significant differences between the investigated conditions. We feel thus, that the new analyses support our conclusions that (1) mitochondrial motility decreases with age and that (2) TTX increases mitochondrial motility significantly. We added these analyses and the statistical tests to the Methods section as part of an extended discussion on the statistics. In the new manuscript, we still show the %-moving mitochondria values, because they can be compared directly to all other measurements in the manuscript, but refer the reader to the Statistics section in Materials and methods. In addition, we include the respective contingency tables in the Source Data Tables. We hope that the reviewers and editors agree with us that the new analyses clearly support our conclusions regarding the factors that regulate mitochondrial motility in vivo.

Author response table 1. Contingency table related to data shown in Figure 2F. The Fisher’s exact test statistic value is 0.0034.

Moving mitochondria Stable mitochondria
Baseline 70 599
TTX 172 946
Open in a new tab

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    Figure 2—source data 1. Source data for Figure 2A-D.
    elife-62091-fig2-data1.xlsx (9KB, xlsx)
    Figure 2—source data 2. Source data for Figure 2E-G.
    elife-62091-fig2-data2.xlsx (10.2KB, xlsx)
    Figure 2—source data 3. Source data for Figure 2H.
    elife-62091-fig2-data3.xlsx (125.6KB, xlsx)
    Figure 2—source data 4. Source data for Figure 2I.
    elife-62091-fig2-data4.xlsx (14.3KB, xlsx)
    Figure 2—figure supplement 1—source data 1. Source data for Figure 2—figure supplement 1A-D.
    elife-62091-fig2-figsupp1-data1.xlsx (9.2KB, xlsx)
    Figure 2—figure supplement 1—source data 2. Source data for Figure 2—figure supplement 1E.
    elife-62091-fig2-figsupp1-data2.xlsx (679.2KB, xlsx)
    Figure 2—figure supplement 1—source data 3. Source data for Figure 2—figure supplement 1F.
    elife-62091-fig2-figsupp1-data3.xlsx (37.9KB, xlsx)
    Figure 3—source data 1. Source data for Figure 3A,B.
    elife-62091-fig3-data1.xlsx (10.4KB, xlsx)
    Figure 3—source data 2. Source data for Figure 3C.
    elife-62091-fig3-data2.xlsx (14.8KB, xlsx)
    Figure 3—source data 3. Source data for Figure 3D.
    elife-62091-fig3-data3.xlsx (12.7KB, xlsx)
    Figure 3—source data 4. Source data for Figure 3I.
    elife-62091-fig3-data4.xlsx (10.8KB, xlsx)
    Figure 3—source data 5. Source data for Figure 3J.
    elife-62091-fig3-data5.xlsx (34.9KB, xlsx)
    Figure 3—source data 6. Source data for Figure 3K.
    elife-62091-fig3-data6.xlsx (11.5KB, xlsx)
    Figure 3—source data 7. Source data for Figure 3L.
    elife-62091-fig3-data7.xlsx (12.1KB, xlsx)
    Figure 3—source data 8. Source data for Figure 3M.
    elife-62091-fig3-data8.xlsx (10.4KB, xlsx)
    Figure 3—figure supplement 1—source data 1. Source data for Figure 3—figure supplement 1A,B.
    elife-62091-fig3-figsupp1-data1.xlsx (10.4KB, xlsx)
    Figure 3—figure supplement 1—source data 2. Source data for Figure 3—figure supplement 1C.
    elife-62091-fig3-figsupp1-data2.xlsx (149.7KB, xlsx)
    Figure 3—figure supplement 1—source data 3. Source data for Figure 3—figure supplement 1D.
    elife-62091-fig3-figsupp1-data3.xlsx (122.4KB, xlsx)
    Figure 3—figure supplement 1—source data 4. Source data for Figure 3—figure supplement 1E.
    elife-62091-fig3-figsupp1-data4.xlsx (53.6KB, xlsx)
    Figure 3—figure supplement 1—source data 5. Source data for Figure 3—figure supplement 1F.
    elife-62091-fig3-figsupp1-data5.xlsx (31.7KB, xlsx)
    Figure 3—figure supplement 2—source data 1. Source data for Figure 3—figure supplement 2A.
    elife-62091-fig3-figsupp2-data1.xlsx (12.1KB, xlsx)
    Figure 3—figure supplement 2—source data 2. Source data for Figure 3—figure supplement 2B.
    elife-62091-fig3-figsupp2-data2.xlsx (9.7KB, xlsx)
    Figure 4—source data 1. Source data for Figure 4B.
    elife-62091-fig4-data1.xlsx (15.3KB, xlsx)
    Figure 4—source data 2. Source data for Figure 4D,E.
    elife-62091-fig4-data2.xlsx (13.1KB, xlsx)
    Figure 4—source data 3. Source data for Figure 4G.
    elife-62091-fig4-data3.xlsx (11.6KB, xlsx)
    Transparent reporting form
    elife-62091-transrepform1.docx (246.3KB, docx)

    Data Availability Statement

    All data generated or analyzed during this study are included in the manuscript and supporting files. Source data files have been provided for all figures and figure supplements.

    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES