Background
Beclin 1 is a component of the type III phosphatidylinositol-3-kinase (PI3K) complex. In yeast, beclin 1 (Atg6/Vps30) regulates both autophagy [
1] and protein sorting [
2‐
4] between the Golgi and vacuole. In this vacuolar protein sorting (vps) pathway, Atg6/Vps30 recruits the retromer complex (composed of Vps35, Vps29, Vps26 and a pair of sorting nexins) to mediate recycling of receptors back to the Golgi. While the role of beclin 1 in autophagy is well established in mammalian cells, we only recently demonstrated a role for mammalian beclin 1 in protein sorting. In microglia, beclin 1 recruits the retromer complex to phagosomes to regulate recycling of the phagocytic receptor CD36 [
5], indicating conservation of the sorting function of beclin 1.
Beclin 1 is highly expressed in the nervous system and is essential for neuronal survival. While beclin 1 knockouts are not viable, mice that are heterozygous deficient for beclin 1 experience age-dependent neurodegeneration. Furthermore, beclin 1 deficiency exacerbates amyloid β (Aβ) pathology in a mouse model of Alzheimer’s Disease (AD) [
6]. Conditional knockout of beclin 1 in either hippocampal or cerebellar neurons was recently shown to result in rapid neurodegeneration [
7]. While basal levels of autophagy are required to prevent neurodegeneration [
8,
9], previous studies have not examined a potential role of beclin 1-mediated protein sorting in neuronal survival.
Trophic factors both direct neuronal development and support neuronal survival. Transforming growth factor - β (TGF-β) is a pleiotropic cytokine that mediates diverse effects in the nervous system. During development, TGF-β is both necessary and sufficient for axon specification [
10], and is required for synaptic pruning during development of ocular dominance in the lateral geniculate nucleus [
11]. In adults, TGF-β signaling is required for the development of immature neurons during neurogenesis [
12]. TGF-β also participates in establishment of long-term potentiation (LTP), a correlate of memory formation [
13,
14]. In addition to these developmental functions, TGF- β1 is released and exerts neuroprotective effects in response to a variety of injuries, including stroke [
15], hypoxia [
16], excitotoxicity [
16], and Aβ exposure [
17,
18]. Inhibition of TGF-β signaling in mice is sufficient to cause age-dependent neurodegeneration [
19]. Indeed, the TGF-β pathway is dysregulated in several neurodegenerative diseases including AD [
16]. For example, levels of the type II TGF- β receptor are decreased [
19], and the downstream Smad proteins are mislocalized in AD [
20,
21]. Taken together, these data indicate a critical role for TGF-β signaling in many aspects of nervous system homeostasis.
Signaling through the TGF-β pathway requires interaction of the type I receptor (also known as activin-like kinase 5, or ALK5) and the type II receptor (TBRII) (Additional file
1: Figure S1). Binding of TGF-β by TBRII induces formation of a heterotetrameric complex with ALK5 and results in ALK5 activation [
22,
23], which then transduces the signal by phosphorylating Smad2 and Smad3 [
24]. In contrast to many other receptor types that are endocytosed only after ligand binding, the TGF-β receptors undergo both constitutive (ligand-independent) and ligand-induced endocytosis [
25]. Endocytosed receptors can then either be sorted to the lysosome for degradation and signal attenuation [
26] or recycled back to the plasma membrane in a Rab11-dependent manner to maintain signaling competency [
25]. Much of our knowledge on TGF-β receptor trafficking and signaling comes from studies on epithelial cells, and receptor trafficking studies in neurons largely focus on the other major receptor types (G-protein coupled receptors, and receptor tyrosine kinases). Given the importance of TGF-β signaling in the nervous system, understanding how TGF-β receptor trafficking is regulated in neurons is critical to our understanding of its developmental and neuroprotective functions. Here we show a novel, autophagy-independent role of beclin 1 in regulating TGF-β signaling in neurons. We demonstrate that beclin 1 recruits the retromer to ALK5 and regulates its recycling, and that loss of beclin 1 results in neuronal death.
Discussion
Despite the importance of TGF-β signaling in neuronal development and homeostasis, little is known about the regulation of its receptors and signaling in these cells. We demonstrate here a novel, autophagy-independent role for beclin 1 in regulating TGF-β signaling in both primary mouse neurons and fibroblasts. Our data show that beclin 1 facilitates localization of ALK5 with both the retromer complex and Rab11 in neurons, and provides direct evidence that beclin 1 regulates ALK5 recycling in COS7 cells. The conservation of a beclin 1 function in ALK5 sorting and TGF-β signaling in multiple cell types suggests beclin 1 is fundamental to the regulation of this pathway.
Our results are consistent with a recent report that identified ALK5 as one of over a hundred receptors down-regulated at the surface of cells depleted for either SNX27 or VPS35 [
35]. Additionally, an independent report from Yin et al. recently demonstrated a role for the retromer in polarized distribution of TBRII in MDCK cells [
36]. TBRII is endocytosed from the apical membrane and recycled through a Rab11-positive recycling endosome to the basolateral membrane in a retromer-dependent manner. While this study did not find a direct interaction between ALK5 and the retromer component VPS26, the observations on the effect of retromer knockdown on ALK5 levels is consistent with our data. These authors found that although ALK5 does not lose its polarized localization upon retromer knockdown (in contrast to TBRII), levels of ALK5 at the basolateral membrane appear to decrease. This is consistent with a role of the retromer in ALK5 recycling, though not polarized localization of this receptor.
The fact that we see only a partial decrease in the colocalization of ALK5 with VPS35 and Rab11 upon beclin 1 knockdown may be due either to incomplete knockdown or to the presence of additional, beclin 1-independent retromer recruitment mechanisms. Indeed, Rojas et al. have previously proposed a model where both Rab5 and Rab7 are required for retromer recruitment to endosomes [
37]. Rab5 recruits the PI3K complex to generate PI3P on endosomes, which in turn recruits the sorting nexin subcomplex of the retromer. These authors suggest that the interaction between the sorting nexin and VPS subcomplexes is weak, and an additional interaction of the retromer with Rab7 is required for its localization at the endosome. Our data localizing beclin 1 to both Rab5 and Rab7-positive endosomes make it likely that beclin 1 is indeed part of the PI3K complex that Rab5 recruits. However, additional mechanistic studies are required to test if Rab proteins are sufficient to recruit the retromer to endosomes in the absence of beclin 1.
A broader role for beclin 1 in protein sorting?
In addition to its function in receptor recycling, beclin 1, as well as UVRAG, has been implicated in downregulation of the EGF receptor (EGFR) via recruitment of the C-VPS/HOPS complex [
7,
38]. The recruitment of either the retromer or the C-VPS/HOPS complex by beclin 1 may therefore represent a decision point that determines whether a receptor is recycled or degraded. There are several potential mechanisms by which this may occur. For example, the composition of the beclin 1 complex, or post-translational modifications of its components, may determine which trafficking complex is recruited. Alternatively, modification of the receptor itself (e.g., by ubiquitination), or occupancy of the receptor by its ligand may influence trafficking complex recruitment. Identification of the mechanism by which beclin 1 can recruit multiple protein trafficking complexes and direct receptors into distinct trafficking pathways is an exciting area for future research.
The sorting functions of beclin 1 are all the more intriguing in light of the recent discovery and characterization of beclin 2 [
39]. Beclin 2 is present only in mammals, and like beclin 1, functions in both autophagy and protein sorting. Beclin 2 interacts with the GPCR-associated sorting protein, GASP1, to downregulate GPCRs. Future studies are needed to delineate the receptor repertoire for both beclin 1 and beclin 2, as well as the ultimate fates of sorted receptors, and to determine the precise mechanisms by which each member of the beclin family regulates sorting.
Our lab has previously reported that levels of beclin 1 are decreased in the brains of AD patients, and that heterozygous deficiency of beclin 1 in mice results in neurodegeneration [
6]. Since decreased beclin 1 levels have been shown by our lab and many others to impair autophagy, and that autophagy inhibition in neurons is sufficient to cause their degeneration [
8,
9], the neuronal death associated with decreased beclin 1 levels was ascribed to autophagy impairment. Recent work from McKnight et al. finds that genetic deletion of beclin 1 in Purkinje cells of the cerebellum or in hippocampal neurons impairs endosomal maturation and results in rapid neurodegeneration [
7]. However, this study does not provide a direct link between a role for beclin 1 in the endosomal system and signaling through neurotrophic or neuroprotective pathways.
Our results showing that reduced beclin 1 levels impair signaling in the TGF-β pathway reveal that beclin 1-mediated protein sorting, in addition to its function in autophagy, may be critical for neuronal survival. These data may link the neurodegeneration observed in beclin 1-deficient mice with the age-dependent neurodegeneration observed in mice expressing a kinase-dead TBRII [
19]. Furthermore, given that beclin 1 knockdown results in decreased levels of both ALK5 and TBRII, it is intriguing to hypothesize that impaired beclin 1-mediated sorting may cause the decrease in TBRII levels previously observed in AD brain tissue [
19]. Future studies should address whether the protein sorting functions of beclin 1 directly contribute to neuronal survival.
TGF-β signaling in AD is particularly interesting for the protection it may provide against some of the negative effects of Aβ exposure. Treatment of primary cultures with TGF-β protects against Aβ-induced neurotoxicity [
17,
18], while pharmacological inhibition of ALK5 in rats exacerbates neurotoxicity of Aβ oligomers injected into the hippocampus [
17]. Inhibition of beclin 1-mediated TGF-β signaling in AD may therefore sensitize neurons to Aβ exposure and contribute to overt neuronal death. Prior to neuronal demise, TGF-β may also protect against cognitive changes induced by Aβ. Aβ exposure impairs LTP in hippocampal slices [
40], which may lead to the deficits in learning and memory observed in mouse models of AD [
41]. TGF-β, however, can positively modulate LTP. Treatment of hippocampal slices with TGF-β prior to a weak stimulus converts early LTP to late-LTP, while pharmacological inhibition of ALK5 impairs LTP and cognitive performance in mice [
14]. A decrease in beclin 1 levels and subsequent inhibition of TGF-β signaling early in disease may therefore contribute to impaired LTP and decreases in cognitive function prior to neuronal death. Future studies should address whether decreased levels of beclin 1 impair LTP in both wild type and AD mouse models, as well as the potential therapeutic benefit of increased TGF-β signaling in protecting against Aβ-induced cognitive impairments. Given that beclin 1 reduces levels of the TGF-β receptors, therapies that target TGF-β signaling downstream of the receptors, rather than simply increasing ligand levels, may prove more efficacious.
Our work presented here fits with an emerging literature that links dysregulation of protein trafficking pathways with neurodegeneration. Levels of VPS35 are decreased in AD [
42], and mutations in VPS35 have been identified in patients with Parkinson’s Disease (PD) [
43,
44]. Mutations in the VPS10 receptor family proteins SorLA and SorCs, well-known retromer cargo that mediate trafficking of the amyloid precursor protein, have also been linked to AD [
45‐
47]. Additionally, variants of CHMP2B, a component of the ESCRT machinery that functions in multivesicular body formation, are associated with frontotemporal dementia [
48]. Looking more broadly within the endosomal system, mutations in Rab7 are associated with Charcot-Marie-Tooth Disease [
49]. In addition to alterations in the trafficking machinery itself, variants in the microglial phagocytic receptor Trem2 that have been associated with neurodegenerative diseases including AD, PD, amyotrophic lateral sclerosis, and frontotemporal dementia impair maturation of this receptor [
50]. These studies together strongly suggest that the proper endosomal trafficking of receptors in multiple cell types in the brain is critical for the maintenance of this organ and the prevention of neurological disease.
Conclusions
Neurodegenerative diseases such as Alzheimer’s Disease are associated with alterations in multiple pathways important for neuronal homeostasis, including autophagy, growth factor signaling, and protein sorting. Because beclin 1 functions in both autophagy and protein sorting, it represents an intriguing therapeutic target for neurodegenerative disease. Indeed, lentiviral delivery of beclin 1 in a mouse model of Parkinson’s Disease was shown to protect against the loss of neuronal markers [
51], and in an AD model reduced amyloid pathology [
6]. While these effects were attributed to increased autophagy, it is possible that enhanced protein sorting also contributed to this rescue. Our work presented here links beclin 1 with the TGF-β pathway. A decrease in beclin 1 levels, as previously observed in AD [
6], leads to impaired ALK5 recycling and deprives neurons of neuroprotective TGF-β signaling. Future studies should determine the full extent of beclin 1-mediated protein sorting, and its contribution to neuronal survival. In particular, identification of the full receptor repertoire beclin 1 regulates may reveal additional trophic factor or other signaling pathways and provide a deeper insight into the consequences of altered beclin 1 levels. Knowledge of both the autophagy and protein sorting functions of beclin 1 should inform strategies to target beclin 1 in neurodegenerative disease.
Methods
Cell culture and primary neuron isolation
Single cell-suspensions of primary hippocampal and cortical neurons were isolated at E16.5 from CF1 pregnant mothers (Charles River) following the protocol described in Fath et al. [
52]. Cells were seeded onto 24-well plates with or without coverslips coated with 0.1 mg/mL poly-L-lysine in boric acid pH 8. Neurons were maintained in neurobasal medium + B27. Hippocampal neurons were used for microscopy (50,000 cells/well) only due to their limited numbers, while cortical neurons were used for western blotting (200,000 cells/well). COS7 and mouse fibroblast (MFB) F11 cells were cultured in DMEM media supplemented with 10 % fetal bovine serum. For TGF-β treatments in F11 cells, cells were washed 2× in PBS and incubated with 1 ng/mL TGF-β1 (R&D Systems) or TGF- β3 (National Institute for Biological Standards and Control #98/608) in serum-free media for 1 h.
Plasmids and lentiviruses
The lentivirus knockdown plasmids contain shRNA targeting mouse beclin 1 at nucleotides 405–423 [
5] or a scrambled control shRNA in the pSIH-H1 vector from SBI System Biosciences. This plasmid also contains a copepod GFP (copGFP) to monitor infection efficiency. All lentiviruses were generated by the Stanford Neuroscience Gene Vector and Virus Core. Plasmids for VPS34, UVRAG, ATG7, and Atg14 shRNA were obtained from Santa Cruz Biotechnology Inc. Plasmids were provided as pools of three 19–25 bp target shRNA. The Rab5-GFP and Rab7-GFP plasmids were generously provided by Dr. Craig Garner.
Cell transfection and viral transduction
Primary cells were infected at 7 DIV at an MOI of 20 in neurobasal medium + B27. Virus was removed 16–20 h later and cells were analyzed 7 days post infection.
F11 or COS7 cells were plated in media containing 8 μg/mL polybrene for 1 h prior to virus addition at an MOI of 50. Virus was removed after 16–20 h and cells were analyzed 48 h post infection. For transfection, cells were transfected using lipofectamine 2000 (Thermo Fisher).
Antibodies for western blotting (WB) and immunocytochemistry (ICC)
Beclin 1 (WB 1:500, BD Biosciences #612113), Beclin 1 (ICC 1:500, Anaspec #54229), Phospho-Smad2 (WB 1:500, Millipore # AB3849), Smad 2/3 (WB 1:500, Cell Signaling Technology # 3102S), TGF beta receptor 1 (ALK5) (WB 1:500, Abcam # ab31013), TGF beta receptor 1 (ALK5) (ICC 1:100, R&D Systems # MAB5871), TGF beta receptor 2 (TBRII) (WB: 1:500, Santa Cruz Biotechnologies # sc-220), VPS35 (ICC 1:500, Abcam # ab10099), Rab11 (ICC: 1:500, Cell Signaling Technology #3539S), Neuron Specific Enolase (WB 1:1000, Thermo Scientific Pierce #MA1-16696), Golgin 97 (ICC 1:250, Thermo Fisher #A-21270), MAP2 (ICC 1:1000, Abcam #ab32454), HA.11 Clone 16B12 (Recycling 1:100, Fisher #NC9693348).
Immunocytochemistry and confocal microscopy
CF1 primary hippocampal cells were grown on coverslips coated with 100ug/ml poly-L-lysine in 100 mM boric acid (pH8). Cells were then fixed in 4 % paraformaldehyde and permeabilized with 0.02 % triton X-100. Coverslips were blocked for 1 h in 5 % milk in TBST, and then incubated with primary antibodies overnight at 4 °C in a humidified chamber. After washing, coverslips were incubated with species-specific Alexa-dye conjugated secondary antibodies for 1 h at RT. Cell nuclei were stained with Hoechst’s dye (Thermo Fisher). All microscopy was performed on a Zeiss LSM 700. Images were collected using the Zeiss Zen software. Mander’s colocalization coefficients were generated using the Zen software. Fluorescence integrated density measurements were made in ImageJ software (version 1.48, NIH).
Western blotting
Cells were lysed in RIPA buffer containing Halt Protease and Phosphatase Inhibitor Cocktail (Roche) and total protein concentrations were determined with a BCA Protein Assay Kit (Thermo Scientific). 10 μg total protein/sample was loaded into precast 4–12 % bis-tris gels and run with MES buffer (Invitrogen). Gels were transferred onto 0.2 μm nitrocellulose (BioRad) and incubated with antigen-specific primary antibodies at 4 °C overnight. For LiCor detection, membranes were incubated with species-specific IR Dye antibodies (LiCor) and scanned on an Odyssey Infrared Imager. Bands were quantified using ImageJ software (version 1.48, NIH).
Quantitative PCR
RNA was isolated from CF1 cortical cells using the Qiagen RNeasy Mini Kit and treated with DNase I (Thermo Fisher) to degrade genomic DNA. cDNA was generated from 1 μg RNA using the SensiFast cDNA Synthesis Kit. Equal amounts of cDNA were used for RT-PCR. The following primers were used:
-
ALK5: F – CAGCTCCTCATCGTGTTGGT
-
TBRII: F -CCTCACGAGGCATGTCATCAG
-
GAPDH: F – AGGTCGGTGTGAACGGATTTG
RT-PCR was carried out on a Roche LightCycler 480 Instrument II using the SYBR Green I Master Mix (Roche) to detect DNA levels. The ΔΔCt method was used to quantify relative mRNA amounts in each sample.
Neuron survival assay
CF1 hippocampal neurons (7 DIV) cultured on coverslips were infected with control or beclin shRNA lentivirus and cultured either for an additional 2 or 3 weeks. Cells were fixed as above and stained for MAP2. Coverslips were imaged using a 10× objective by confocal microscopy. Three randomly selected fields were imaged per coverslip (3 coverslips/condition) and the number of MAP2-positive neurons were counted (9 fields/condition). Results are expressed as relative number of cells/field.
SEAP assay
MFB-F11 cells stably expressing the SBE-SEAP reporter construct [
32] were plated at a density of 40,000 cells/well in 96-well plates and allowed to adhere overnight. Cells were washed twice with PBS and treated with or without 1 ng/mL TGF-β1 (R&D Systems) overnight. Media was collected and assayed for SEAP activity using the Roche chemiluminescent SEAP Reporter Gene Assay.
Receptor recycling assay
Receptor recycling assay was performed as previously described [
25]. COS7 cells were plated onto poly-L-lysine coated coverslips and infected with lentivirus as described above. Forty-eight hours after infection, cells were transfected with a plasmid encoding HA-tagged ALK5. Twenty-four hours after transfection, cells were incubated in 1 % goat serum (the source of the secondary antibody) in DMEM for 15 min on ice to block. To measure recycling, cells were incubated with anti-HA antibody (1:100 in 1 % goat serum + DMEM) for 1 h at 37C. Cells were then acid washed on ice with cold DMEM at pH 2.0, then with cold PBS. Cells were then incubated with secondary antibody for 1 h at 37C. Cells were again acid washed with cold DMEM at pH 2.0 and washed with cold PBS. Cells were then fixed with 4 % PFA, permeabilized with 0.2 % triton in PBS, and washed. After permeabilization, cells were blocked for 1 h in 5 % milk, and then stained for total ALK5 (1:250 in 5 % milk) as described above. Image z-stacks were collected by confocal microscopy. Infected (GFP+) cells were outlined and measurements were made for all GFP+ cells in an image using ImageJ software (version 1.48, NIH). Fluorescent signals of anti-HA (recycled) and anti-ALK5 (total) were thresholded and the fluorescent integrated density was measured for each image in the stack for every cell. The ratio of surface or recycled to total ALK5 for each cell was then taken, and summed for all images in the stack. Only cells with ALK5 intensity density over 10/cell area and HA intensity density over 1/cell area were analyzed.
Statistics
All statistical analyses were conducted using the Prism6 (GraphPad Software). Differences between control and treatment conditions were calculated using a Student’s unpaired t test, one-way ANOVA with Dunnet’s post-test to compare test conditions to control, or a two-way ANOVA with Sidak’s post-test for groups with multiple variables. All bar graphs are presented as mean ± SEM and p values less than 0.05 were considered significant. Statistical details for each experiment are indicated in the figure legends.
Acknowledgements
The authors thank Dr. Jian Luo and Dr. Vidhu Mathur for critical review of this manuscript, Dr. Ed Plowey and Dr. Gayathri Swaminathan for helpful discussions, Michael Lochrie, Javier Fernandez-Alcudia and the Stanford Neuroscience Gene Vector and Virus Core for production of the lentiviruses used in this study and for useful discussions on viral infection protocols, and Dr. Craig Garner for providing the Rab5-GFP and Rab7-GFP plasmids. Funding for these studies was provided by the National Institutes of Health Institute on Aging (R01 AG030144 T.W.-C.), the Veterans Administration (BX001319), and a National Institutes of Health Institute of Neurological Disorders and Stroke Ruth L. Kirschstein NRSA predoctoral fellowship (1F31NS078865, C.E.O.).
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
C.E.O. designed and carried out the experiments. L.B. helped run beclin localization experiments. H.Z. prepared primary neuronal cultures. T.W.C carefully reviewed the manuscript. All authors read and approved the final manuscript.