|Home | About | Journals | Submit | Contact Us | Français|
mTOR is a regulator of cell growth and survival, protein synthesis-dependent synaptic plasticity, and autophagic degradation of cellular components. When triggered by mTOR inactivation, macroautophagy degrades long-lived proteins and organelles via sequestration into autophagic vacuoles. mTOR further regulates synaptic plasticity, and neurodegeneration occurs when macroautophagy is deficient. Nevertheless, whether macroautophagy is a modulator of presynaptic function was previously unknown. We find that the mTOR inhibitor, rapamycin, induces formation of autophagic vacuoles in prejunctional dopaminergic axons with associated decreased axonal profile volumes, synaptic vesicle numbers, and evoked dopamine release. Evoked dopamine secretion was enhanced and recovery accelerated in transgenic mice in which macroautophagy deficiency was restricted to dopaminergic neurons; rapamycin failed to decrease evoked dopamine release in the striatum of these mice. Macroautophagy that follows mTOR inhibition in presynaptic terminals, therefore, rapidly alters presynaptic structure and neurotransmission.
The kinase mammalian target of rapamycin (mTOR) regulates protein synthesis (Huang and Manning, 2009) and degradation (Cuervo, 2004). mTOR activity enhances protein synthesis via participation in the complex mTORC1, which phosphorylates p70, S6 kinase and 4E-BP (Huang and Manning, 2009). mTORC1 also phosphorylates Atg13, inhibiting Atg1, which is required for the induction of macroautophagy (Kamada et al., 2010). mTOR activity, therefore, both enhances protein synthesis and inhibits cellular degradation pathways.
In the nervous system, mTORC1 activity stimulates protein synthesis-dependent synaptic plasticity and learning (Huang and Manning, 2009; Long et al., 2004; Richter and Klann, 2009). Most studies on cellular and neuronal functions of mTOR use rapamycin, an inhibitor that when bound to FKBP12 interacts with mTOR’s FRB domain and prevents mTOR from binding raptor, a component of the mTORC1 complex (Dowling et al., 2010). Rapamycin blocks axonal hyperexcitability and synaptic plasticity in cellular models of injury as well as learning and memory by inhibiting protein synthesis (Hu et al., 2007; Weragoda and Walters, 2007).
Macroautophagy is a highly conserved cellular degradative process in which proteins and organelles are engulfed by autophagic vacuoles (AVs) that are subsequently targeted for degradation in lysosomes. It is possible that degradation of pre- or postsynaptic components could contribute to plasticity: for example, local mTOR inhibition might elicit autophagic degradation of synaptic vesicles, providing a means of presynaptic depression. We therefore explored whether mTOR-regulated degradation of proteins and organelles via macroautophagy alters synaptic function and morphology.
To do so, we generated transgenic mice in which macroautophagy was selectively inactivated in dopamine neurons. These neurons are deficient in expression of Atg7, an E1-like enzyme that conjugates microtubule associated protein light-chain 3 (LC3) to phospholipid and Atg5 to Atg12, steps that are necessary for AV formation (Martinez-Vicente and Cuervo, 2007). We chose to specifically delete Atg7 to abolish macroautophagy and the formation of AVs because, in contrast to Atg1, it is not thought to directly regulate membrane trafficking (Wairkar et al., 2009).
We chose to examine presynaptic structure and function in the dopamine system because: 1. In the acute striatal slice preparation, dopamine axons are severed from their cell bodies but continue to synthesize, release, and reaccumulate neurotransmitter for up to ten hours, allowing us to clearly focus on axonal autophagy. 2. Electrochemical recordings of evoked dopamine release and reuptake in the striatum provide a unique means to measure CNS neurotransmission with millisecond resolution that is independent of postsynaptic response.
We found that: 1) Chronic macroautophagy deficiency in dopamine neurons resulted in increased size of axon profiles, increased evoked dopamine release, and more rapid presynaptic recovery. 2) In mice with intact macroautophagy, mTOR inhibition with rapamycin acutely increased AV formation in axons, decreased the number of synaptic vesicles, and depressed evoked dopamine release. 3) Rapamycin had no effect on evoked dopamine release and synaptic vesicles in dopamine-neuron specific macroautophagy deficient mice. We conclude that mTOR-dependent local axonal macroautophagy can rapidly regulate presynaptic structure and function.
We generated dopamine neuron-specific macroautophagy deficient mice by crossing Atg7flox/flox mice (Komatsu et al., 2005) to a line expressing cre recombinase under the dopamine transporter (DAT) promoter (DAT Cre/+) (Zhuang et al., 2005). The progeny (Atg7flox/+;DAT Cre/+) were crossed to Atg7flox/flox to generate Atg7flox/flox;DAT Cre/+ (Atg7 DAT Cre). As the mutant mice have a single functional copy of DAT, we used DAT Cre/+ (DAT Cre) animals as controls; these animals express two copies of wild-type Atg7 and a single functional copy of DAT.
We detected Atg7 expression by non-radioactive in situ hybridization using an RNA probe designed against nucleotides 1518–1860 of the atg7 gene in 8–10 week old mice. Atg7 mRNA was detected in both the anterior and central substantia nigra pars compacta and pars reticulata in DAT Cre animals, but was absent in Atg7 DAT Cre mice. Atg7 mRNA was detected in the red nucleus (RN) and in the dentate gyrus (DG) from Atg7 DAT Cre, further indicating cellular specificity for the knocked out gene (Supplementary Figure 1). We conclude that the atg7 gene was effectively deleted in ventral midbrain dopamine neurons.
In contrast to CNS-wide macroautophagy deficient mice, which are smaller than controls, exhibit abnormal limb clasping, and begin to die at 4 weeks (Hara et al., 2006; Komatsu et al., 2006), Atg7 DAT Cre mice showed similar survival and weight gain as DAT Cre mice at 8 weeks of age (mean weights: 22.7±1.1g and 25.2±1.6g, respectively; p>0.05, n = 6 mice per group, t-test). The limb clasping reflex of Atg7 DAT Cre mice was normal (not shown). To evaluate motor behavior in tasks thought to specifically involve dopamine transmission (Crawley, 1999; Karl et al., 2003), we performed tail hang, beam walk and open field tests on mice aged 6–12 weeks. Motor performances of Atg7 DAT Cre mice were not different than DAT Cre in any of the tests (n=4 in each group: data not shown). We did not examine mice older than 3 months in this study, and motor and behavioral differences may develop in aged mice.
We examined striatal dopaminergic axonal profiles immunolabeled for tyrosine hydroxylase (TH) from 8-week old DAT Cre and Atg7 DAT Cre mice by electron microscopy (Figure 1A). There was no difference in the number of striatal TH immunoreactive axonal profiles per area in Atg7 DAT Cre mice (Figure 1B). There was, however, an increase in the fraction of total area occupied by TH+ profiles: TH+ axon profiles occupied 2.3 ± 0.2% of the total sampled area in the striatum of DAT Cre mice, but 4.7 ± 0.5% of the area in Atg7 DAT Cre mice (p < 0.005, t-test, > 6000 µm2 sampled per condition in 10 and 8 micrographs, respectively; Figure 1C). Consistently, striatal TH+ axonal profiles from Atg7 DAT Cre mice (0.42 ± 0.04 µm2, n = 84) were larger than from DAT Cre animals (0.29 ± 0.03 µm2, n = 60; p < 0.05, Mann-Whitney test; Figure 1D). We found no difference in the size of terminals unlabeled for TH between DAT Cre and Atg7 DAT Cre mice (0.24 ± 0.03 µm2, n = 26; 0.30 ± 0.03 µm2, n =27; p > 0.05, t-test; Figure 1E).
To explore the effects of mTOR inhibition and macroautophagy deficiency on the size of dopamine axonal profiles, we injected pairs of DAT Cre and Atg7 DAT Cre mice with rapamycin (2 mg/kg) or vehicle (DMSO) 36 and 12 hours prior to perfusion. Rapamycin decreased the area of TH+ striatal axon profiles by 32% in DAT Cre mice but had no effect on DA terminals of the DAT Cre Atg7 mutant line (Figure 1F) (interaction between rapamycin and genotype, F=6.72, p < 0.01, two way ANOVA).
We used cyclic voltammetry to directly measure evoked dopamine release and reuptake in the striatum. The peak amplitude of the signal is dependent on both neurosecretion and reuptake through DAT, while the the half-life (t½) is a function of DAT activity (Schmitz et al., 2001).
The amplitude of the dopamine signal evoked by a single pulse of electrical stimulation in Atg7 DAT Cre mice was 54% greater than in DAT Cre controls (n = 9 and n = 7, respectively: 4.0 ± 0.3 and 2.6 ± 0.2 nM; p < 0.005, t-test; Figure 2A,B). As DAT Cre and Atg7 DAT Cre mice express a single functional copy of DAT, the signal duration in both genotypes was longer than in wild type mice (mean t½ ~ 490 msec) (Schmitz et al., 2001), but the mean t½ of DAT signals from DAT Cre and Atg7 DAT Cre slices were not different (Supplementary Figure 2A: mean t1/2: 637 ± 51 and 662 ± 23 ms; p > 0.05, t-test), which indicates that reuptake kinetics are similar and that the increased peak amplitude in the Atg7 deficient line was due to greater dopamine release rather than decreased reuptake.
To measure the rate of presynaptic recovery, we stimulated dopamine release with pairs of pulses separated by intervals that ranged from 1 to 60 sec (Schmitz et al., 2002). Atg7 DAT Cre mice exhibited faster recovery (p < 0.05, repeated measures ANOVA; Figure 2D), suggesting that basal macroautophagy can restrict synaptic transmission.
We then examined effects of rapamycin on evoked dopamine release. Striatal slices were bisected, and one striatum was exposed to rapamycin and the other to vehicle. Rapamycin decreased dopamine release evoked by a single electrical stimulus by 25 ± 3% in DAT Cre slices (n=7) and by 6 ± 6% in Atg7 DAT Cre slices (n= 9; p < 0.05, two-way ANOVA, Newman-Keuls post-test; Figure 2E,F). Rapamycin did not significantly alter the t½ of the signals from DAT Cre (control: 718 ± 29 ms, Rapa: 675 ± 22 ms) or Atg7 DAT Cre (753 ± 23 ms; Rapa: 743 ± 32 ms) mice (Supplementary Figure S2A; p>0.05, 2-way ANOVA).
The data indicate that the bulk of rapamycin’s inhibition of evoked dopamine release is mediated by macroautophagy. To confirm that these effects were not limited to DAT Cre mutants, we repeated the recordings in slices from wild-type mice, and observed a similar rapamycin-induced reduction in dopamine secretion (Supplementary Figure S2B).
To examine the effect of acute mTOR inhibition on AV formation, we first exposed postnatally-derived ventral midbrain neuronal cultures (Rayport et al., 1992) to rapamycin (200 nM, 3.5 h). The macroautophagy-related protein LC3 exists in two forms, LC3-I and LC3-II, a phosphatidylethanolamine-conjugated form of LC3-I. LC3-I is widely distributed in the cytosol while the conjugated LC3-II form specifically associates with AV membranes (Mizushima et al., 2004). Dopamine neurons were identified by TH immunolabel and immunolabel for native LC3 was used to identify AVs. There were occasional LC3-immunolabeled puncta in the Atg7 deficient cell bodies and neurites, possibly due to non-canonical AV formation (Nishida et al., 2009). Rapamycin strikingly increased LC3-immunolabeled puncta in dopamine cell bodies and neurites in DAT Cre mice, but had no effect on puncta in DAT Cre Atg7 mutants (p < 0.01, ANOVA) (Figure 3A–C), showing that induction of AVs by rapamycin required Atg7 expression.
We then examined the induction of LC3-II by rapamycin (3 µM) in acute striatal slices by Western blotting. Rapamycin at 3.5 h produced a 56% increase in LC3-II (Fig 3D) (p < 0.001, t-test), but this response was no longer apparent at 7 h, indicating that rapamycin induced a transient increase of LC3-II, a characteristic of macroautophagy.
In electron micrographs of striatal slices, we identified AV-like organelles based on previously described criteria (Yu et al., 2004) as non-mitochondrial structures in presynaptic terminals that possessed multiple membranes, usually with luminal content. These organelles were different from multivesicular bodies (MVBs), an organelle of the autophagic-lysosomal pathway that typically displays an even distribution of vesicles in the lumen. Many AV-like organelles contained a wide range of luminal constituents, including small vesicles resembling synaptic vesicles (compare Figures 4A and B). Some multilamellar structures were devoid of obvious luminal electron dense material (Martinez-Vicente et al., 2010),possibly due to acute induction of AVs by rapamycin. It is likely that some of these multilamellar organelles include endosomes or are “amphisomes” that result from fusion of endosomes and AVs. Rapamycin in the striatal slice more than doubled the number of presynaptic terminal profiles containing AV-like structures from 15.4% of control terminal profiles (n = 65) to 35.5% in rapamycin-treated terminals (n =75, p < 0.05, Chi-square test, Figure 4C) and decreased terminal profile areas by 19% (p<0.05, t-test, Figure 4D). Striatal terminal profiles from rapamycin-treated samples, of which only a small fraction are dopaminergic, moreover contained fewer synaptic vesicles than untreated controls (49.2 ± 3.6, n = 75 vs. 70.1 ± 4.2, n = 65; p < 0.0001, respectively, t-test, Figure 4E).
Dopamine axonal varicosities typically do not display presynaptic or postsynaptic densities (Nirenberg et al., 1997), but amperometric studies demonstrate stimulation-evoked quantal transmitter release from these structures (Pothos et al., 1998) and many accumulate and secrete fluorescent dopamine analogs (Gubernator et al., 2009), confirming their identity as presynaptic terminals. As TH immunolabel obscured synaptic vesicles and other intracellular structures (see Figure 1), we examined if rapamycin reduced the number of dopaminergic synaptic vesicles by using the false neurotransmitter, 5-hydroxydopamine (5OHDA), which is selectively accumulated into these dopaminergic synaptic vesicles and produces osmophilic dense cores (Tennyson et al., 1974) (Figure 5, blue arrows). For each experiment, striatal slices were obtained from a single mouse, bisected, and individual striata were incubated in vehicle (DMSO) or rapamycin (3 µM, 6.5h), and then treated with 5OHDA (500 µM, 30 min). The numbers of synaptic vesicles in the labeled terminals were compared between slices derived from the same mouse. In a wild-type mouse, rapamycin decreased synaptic vesicles within 5OHDA labeled terminals by 18% (from 105 to 86 synaptic vesicles per µm2; p<0.02, t-test, 37 and 42 terminals rated); and in a DAT Cre mouse, rapamycin decreased synaptic vesicles within labeled terminals by 26% (from 82 to 61 synaptic vesicles per µm2; p = 0.05, t test, 31 and 27 terminals rated). In contrast, rapamycin did not decrease synaptic vesicles within labeled terminals of an Atg7 DAT Cre mouse (84 to 95 synaptic vesicles per µm2, p = 0.13, t test, 38 and 39 terminals rated), indicating that rapamycin decreased the number of dopaminergic synaptic vesicles only if Atg7 was present.
We compared the levels of a range of synaptic proteins between striatal slices of DAT Cre mice and Atg7 DAT Cre mice exposed to rapamycin (3 µM) or vehicle for 7 h. Treated and untreated Atg7 DAT Cre mice showed substantially lower levels of DAT (Supplementary Figure S3, Supplementary Table I), a small but significant decrease of TH (p< 0.05, 2 factor ANOVA), and similar levels (p>0.5) of the postsynaptic marker PSD95, and the mitochondrial proteins porin, tomm20, and tim23. While there was a transient increase in LC3-II at 3.5 h (Figure 3D), no protein examined was altered by rapamycin at 7 h. It may be that while this period provided sequestration of cellular elements in AVs, there was no measurable net degradation over this period. Note that only axons of dopamine neurons were present, and corresponding cell bodies with mature lysosomes were absent.
Our data indicate that both basal and induced macroautophagy modulates presynaptic structure and function. Mice with chronic macroautophagy deficiency in dopamine neurons had abnormally large dopaminergic axonal profiles, released greater levels of neurotransmitter in response to stimulation, and exhibited more rapid presynaptic recovery. mTOR inhibition by rapamycin administered to control mice induced AV-like structures in axons and decreased synaptic vesicles to nearly the same level as the accompanying decrease in evoked dopamine release. In contrast, rapamycin had little or no effect on the number of synaptic vesicles or neurotransmitter release in macroautophagy deficient neurons. Together, our results introduce acute presynaptic changes that depend on Atg7 expression and hence macroautophagy.
These presynaptic effects were observed in dopaminergic presynaptic terminals in slices without their cell bodies and so the critical steps in autophagy must have occurred locally in axons that typically lack mature lysosomes (Overly et al., 1995). Our data confirm that AVs can be synthesized locally in the axons (Lee et al., 2011) and are to our knowledge the first data to indicate local axonal autophagic sequestration of presynaptic components and an accompanying modulation of presynaptic function. This evidence extends studies of selective degradation of postsynaptic receptors via macroautophagy (Hanley, 2010; Matsuda et al., 2008; Rowland et al., 2006) and classic work indicating a role for lysosomal degradation in recycling synaptic vesicle turnover (Holtzman et al., 1971). Thus, in addition to well-established roles of macroautophagy in stress response and cellular homeostasis (Tooze and Schiavo, 2008), neurons have adapted this phylogenetically ancient process to modulate neurotransmitter release and remodel synapses.
Macroautophagy deficiency throughout the CNS results in decreased weight, motor deficits, and premature death (Hara et al., 2006; Komatsu et al., 2006). Purkinje cells from cell-specific autophagy deficient mice show axonal swellings and signs of neurodegeneration as early as P19 (Komatsu et al., 2007). Signs of neurodegeneration were however not observed in young DAT Cre Atg7 mice (<14 weeks), possibly due to compensation by other degradative pathways (Koga et al., 2011). It may be that further aged DAT Cre Atg7 mice model aspects of Parkinson’s-related disorders.
Chronic autophagy deficiency rather increased the size of dopaminergic synaptic terminal profiles and striatal dopaminergic innervation, consistent with studies that implicate macroautophagy in retraction of neuronal processes (Bunge, 1973) and neuritic growth in developing neurons (Hollenbeck, 1993). The results, however, contrast with studies in Drosophila, where disruption of AV formation or AV-lysosomal fusion decreases the size of the neuromuscular junction, while Atg1 overexpression or rapamycin promotes macroautophagy and increases the number of synaptic boutons and neuritic branches (Shen and Ganetzky, 2009). Some synaptic Atg1-related changes may be autophagy-independent because the loss of other autophagy related proteins does not mimic the effect of Atg1 overexpression on the number of boutons and neurite branches (Toda et al., 2008; Wairkar et al., 2009).
We further observed that chronic ATg7 deficiency in dopamine terminals led to an increase in presynaptic mitochondria (not shown), Atg7 deficiency may contribute to changes in mitochondria in multiple ways, for example via effects on presynaptic mitochondria size, shape, trafficking, fission, and fusion.
Acute induction of AVs by rapamycin in control neurons was confirmed by electron microscopy, LC3 immunolabel, and transiently elevated LC3-II. Acute exposure to rapamycin decreased synaptic terminal profile size and number of synaptic vesicles, indicating that mTOR inhibition can rapidly decrease presynaptic components. Some AV-like profiles contained cargo that resembled synaptic vesicles, although we were unable to immunolabel AV components, presumably due to the low luminal pH. Presynaptic terminals are very active in endocytosis due to the turnover and recycling of synaptic vesicles, receptors and other constituents, and it is likely that many of the multilamellar organelles we observe are products of the fusion of endosomes and AVs, sometimes called “amphisomes”. An apparently clear content of occasional AV-like organelles suggests that acute mTOR blockade may result in some “empty” early AVs (Martinez-Vicente et al., 2010).
AV-like profiles were absent in dopamine axon profiles of the Atg7 deficient mice, and while low levels of LC3 immunolabeled puncta were present in the mutant neurons, they were not enhanced by rapamycin. Thus, the increase in AVs by mTOR inhibition apparently requires Atg7, and we hypothesize that in normal neurons, rapamycin redistributed synaptic vesicle membranes into axonal AVs, endosomes, and/or amphisomes.
Chronic lack of macroautophagy enhanced evoked dopamine release and the rate of synaptic recovery. At a variety of synapses, a higher release probability can increase the peak amplitude of the first pulse followed by a relative depression of the second pulse, due to a decreased availability of release-ready vesicles, culminating in a lower paired pulse ratio (2nd pulse / 1st pulse). This situation differs from that in Atg7 DAT Cre animals, where both the initial and subsequent pulses showed increased amplitudes relative to control mice. The probability of dopaminergic synaptic vesicle fusion is regulated by the size of the recycling and readily releasable pools (Daniel et al., 2009): the enhanced release and recovery in the mutant line could be due to multiple non-exclusive effects, including a greater synaptic terminal size/density, number of synaptic vesicles, more calcium influx, or an increase in vesicle docking and fusion sites and/or rates. We measured lower total striatal DAT and TH levels in the macroautophagy deficient line, although the kinetics of dopamine release do not indicate altered activity of the proteins, which are regulated by a variety of compensatory mechanisms (Schmitz et al., 2003).
Rapamycin depressed evoked dopamine release in control mice, but had no effect in Atg7 DAT Cre mice, confirming that the rapid changes in neurotransmission evoked by mTOR inhibtion were macroautophagy-dependent and not the result of effects on protein synthesis.
Although we have focused on dopaminergic terminals, the data suggest that these effects are not specific to them. Rapamycin induced apparent AVs in both dopaminergic (TH+) and non-dopaminergic (TH−) terminals, and a decrease in synaptic vesicles was observed generally in striatal synaptic terminals, which include glutamatergic, dopaminergic, GABAergic, and cholinergic synaptic terminals. Our experiments do not address how subsets of particular presynaptic organelles, such as individual synaptic vesicles, may be specifically targeted by mTOR-dependent axonal macroautophagy. Clues might be offered if alternative modes of vesicle recycling are identified that could partake in or avoid endocytic compartments that might fuse with AVs (Voglmaier et al., 2006).
Starvation, injury, oxidative stress, toxins including methamphetamine and infection by neurotropic viruses trigger autophagy in neurons, which is further associated with protein aggregate-related disorders including Huntington’s, Parkinson’s, and Alzheimer’s diseases (Cheng et al., 2011; Koga et al., 2011; Larsen et al., 2002; Talloczy et al., 2002; Tooze and Schiavo, 2008). mTOR activity is regulated by multiple endogenous pathways involved in synaptic activity and stress, including tuberous sclerosis complex, Rheb, AKT, NF1, and PTEN (Malagelada et al., 2010). Alterations in mTOR activity are associated with neuropathological conditions such as epilepsy, tuberous sclerosis, and autism. Regulation of presynaptic function by mTOR activity and macroautophagy could thus contribute to manifestations of neurological disorders.
We thank Ana Maria Cuervo, Zsolt Talloczy, and Steven Siegelbaum for discussion. We thank Maksaaki Komatsu for providing the floxed Atg7 line. This work was funded by a Udall Center of Excellence (NINDS), the Parkinson’s Disease Foundation, and the Picower Foundation.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.