PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Cell. Author manuscript; available in PMC Feb 5, 2012.
Published in final edited form as:
PMCID: PMC3262169
NIHMSID: NIHMS317576
Recruitment of actin modifiers to TrkA endosomes governs retrograde NGF signaling and survival
Anthony W. Harrington,1* Coryse St. Hillaire,1* Larry S. Zweifel,1,3 Natalia O. Glebova,1,4 Polyxeni Philippidou,2,5 Simon Halegoua,2 and David D. Ginty1,6
1The Solomon H. Snyder Department of Neuroscience and Howard Hughes Medical Institute, The Johns Hopkins University School of Medicine, Baltimore, MD 21205
2Department of Neurobiology and Behavior, Center for Nervous System Disorders and Program in Neuroscience, Stony Brook University, Stony Brook, NY 11794-5230
6 Corresponding author: dginty/at/jhmi.edu
*equal contributions
3Present address: Department of Pharmacology, University of Washington Medical Center, Seattle, WA 98195
4Present address: Department of Surgery, The Johns Hopkins University School of Medicine, Baltimore, MD 21205
5Present address: Howard Hughes Medical Institute, Smilow Neuroscience Program, New York University School of Medicine, New York, NY 10016
NGF and NT3 collaborate to support development of sympathetic neurons. Although both neurotrophins activate TrkA-dependent axonal extension, NGF is unique in its ability to promote retrograde transport of TrkA endosomes and retrograde survival. Here, we report that actin depolymerization is essential for initiation of NGF/TrkA endosome trafficking and that a Rac1–cofilin signaling module associated with TrkA early endosomes supports their maturation to retrograde transport-competent endosomes. Moreover, the actin-regulatory endosomal components are absent from NT3-formed TrkA endosomes, explaining the failure of NT3 to support retrograde TrkA transport and survival. The inability of NT3 to activate Rac1-GTP–cofilin signaling is likely due to the labile nature of NT3/TrkA complexes within the acidic environment of TrkA early endosomes. Thus, TrkA endosomes associate with actin-modulatory proteins to promote F-actin disassembly enabling their maturation into transport-competent signaling endosomes. Differential control of this process explains how NGF in final targets, but not NT3 from intermediate targets, supports retrograde survival of sympathetic neurons.
Nerve Growth Factor (NGF), the prototypical member of the neurotrophin family, is essential for development of select populations of neurons in the peripheral nervous system. One of the many functions attributed to NGF is the promotion of sympathetic neuron survival. Indeed, as predicted by the neurotrophic hypothesis, NGF is expressed in target fields of sympathetic neurons and supports their survival via a long distance signaling mechanism (Levi-Montalcini, 1987). This ultimately ensures that the number of surviving neurons is commensurate with the size and demands of the end organ, a process referred to as “systems matching”.
NGF survival signaling is initiated within distal axons and communicated in a retrograde manner to neuronal cell bodies where it prevents apoptosis. This retrograde signal is activated by NGF binding to its receptor, TrkA, which promotes canonical receptor tyrosine kinase pro-survival signaling (Miller and Kaplan, 2001), enhanced sensitization to NGF, punishment of neighboring neurons lacking NGF/TrkA signaling, and protection from p75-mediated apoptotic signals (Deppmann et al., 2008). A major mechanism by which this long-distance retrograde signal is carried from distal axons to the cell body is through NGF-induced internalization of NGF/TrkA complexes and the formation of neurotrophin “signaling endosomes”. It is now appreciated that TrkA signaling endosomes mediate retrograde control of neuronal survival, growth, gene expression, and synaptogenic signaling events (Howe and Mobley, 2005; Pazyra-Murphy et al., 2009; Sharma et al., 2010). Although several proteins that associate with and function during formation and transport of TrkA signaling endosomes have been identified (Delcroix et al., 2003; Shao et al., 2002; Wan et al., 2008), proteins that mediate internalization, sorting, axonal transport, signaling, and disassembly of the signaling endosome are largely unknown; therefore, identification of signaling endosome constituents is likely to lend insight into each of these processes.
While NGF is commonly considered the primary ligand for TrkA, the receptor can also be bound and activated by the related neurotrophin NT3 (Davies et al., 1995). In contrast to NGF's role as a target-derived survival factor, NT3 is produced and secreted by intermediate targets of sympathetic neurons, including the vasculature (Francis et al., 1999). Interestingly, ~50% of sympathetic neurons are lost in NT3-/- mice, and those that remain possess shorter and thinner axons than their wild type littermates (Davies et al., 1995; Kuruvilla et al., 2004). This loss of neurons likely reflects the requirement of NT3/TrkA signaling for local axon extension along intermediate targets, a prerequisite for final target field innervation and acquisition of target-derived NGF, as opposed to a direct effect of NT3 on cell survival. Indeed, while both NGF and NT3 can support TrkA activation and axonal extension (Belliveau et al., 1997; Kuruvilla et al., 2004), NT3 is incapable of eliciting retrograde survival signaling, presumably due to its inability to form retrogradely transported TrkA signaling endosomes. This raises the intriguing question of how NGF and NT3, acting through a common receptor, TrkA, have such divergent functions in the same neuron. Here, we present evidence that target-field-derived growth factor, NGF, and not an intermediate-target-derived neurotrophin, controls TrkA endosome maturation through an actin-based endosome sorting mechanism to support retrograde survival signaling that underlies systems matching.
NGF and NT3 similarly activate TrkA, but only NGF supports retrograde TrkA trafficking and retrograde survival in sympathetic neurons
Development of postganglionic sympathetic neurons of the superior cervical ganglia (SCG) is critically regulated by two neurotrophins, NT3 and NGF, acting through a common receptor, TrkA (Belliveau et al., 1997; Davies et al., 1995; Kuruvilla et al., 2004). NT3 supports axonal extension along an intermediate target, the vasculature (Kuruvilla et al., 2004), whereas NGF secreted from final targets is required for target field innervation, retrograde control of survival and the formation of synapses between pre- and postganglionic neurons (Levi-Montalcini and Booker, 1960; Sharma et al., 2010). Remarkably, although NGF and NT3 can promote both TrkA autophosphorylation and downstream signaling events (Figure 1B), NGF is unique in its ability to promote retrograde TrkA survival signaling to the cell soma (Kuruvilla et al., 2004; Figure 1A). This is consistent with the notion that the final target but not the intermediate target field defines the amount of survival within a neuronal population. We hypothesized that disparities in the trafficking of NGF/TrkA and NT3/TrkA complexes within distal axons of sympathetic neurons account for the differences in NGF and NT3's retrograde signaling capabilities. In order to visualize internalization, sorting, and trafficking of endogenous TrkA in sympathetic neurons, we generated a mouse in which the coding determinants of the Flag epitope tag are knocked into the endogenous TrkA locus (Sharma et al., 2010). In cultured SCG neurons dissected from these TrkAFlag knock-in mice, Flag-TrkA protein on the cell surface is labeled using an anti-Flag monoclonal antibody (M1), allowing us to monitor the internalization, intracellular trafficking and subcellular localization of TrkA endosomes. Combining this technique with compartmentalized microfluidic chambers enables an assessment of the formation of TrkA endosomes and their movements from distal axons to cell bodies. As expected, NGF application to distal axons led to robust internalization and retrograde transport of Flag-TrkA endosomes (Figure 1C, D). NT3 also promoted internalization of TrkA when applied to distal axons (Figure 1D) or when applied to cell bodies (Figure S1B, C). These NT3-formed Flag-TrkA+ punctae are likely endosomes as the anti-Flag signal remained following acid/salt stripping, which removes all cell surface antibody (Sharma et al., 2010). Moreover, NT3/TrkA punctae co-localized with the early endosome marker Rab5 (Figure S2). However, the internalized NT3/TrkA complexes in distal axons failed to undergo retrograde transport to the cell soma (Figure 1C). This is remarkable because both NT3 and NGF promoted comparable levels of TrkA activation and ERK, PLC-γ, and Akt signaling (Figure 1B). Thus, while NGF and NT3 similarly activate TrkA on the axonal surface and promote TrkA-mediated axonal extension and TrkA internalization, these two ligands differ in their abilities to generate transport-competent signaling endosomes and retrograde survival signaling.
Figure 1
Figure 1
NGF and NT3 similarly activate and promote internalization of TrkA, but only NGF supports retrograde TrkA trafficking and retrograde survival of sympathetic neurons
A proteomic screen for TrkA endosome-associated proteins reveals an essential role for actin modulation during signaling endosome maturation and transport
The mechanisms of signaling endosome formation, sorting, trafficking, and signaling are poorly understood. To address these processes and identify key differences between NGF/TrkA and NT3/TrkA signaling, we performed a proteomic analysis of biochemically-isolated TrkA endosomes. The endosome purification procedure was designed to isolate TrkA+ early endosomes from a stably transfected PC12 cell line that expresses an epitope-tagged Trk receptor. This purification procedure yielded endosomes with sizes and appearances that were consistent with early endosomes based on EM analysis (Figure S3). Protein from purified endosome preparations was extracted and subjected to mass spectroscopy, resulting in the identification of approximately 250 endosome-associated proteins (Supp. Table 1). In addition to a large number of proteins of unknown function, many identified proteins have known roles in vesicle trafficking, scaffolding, signaling and metabolism. Interestingly, several proteins implicated in actin filament modulation were identified, including actin itself, cofilin, moesin, the Rac1 and Rab5 GEF ALS-2, and a component of the myosin actin motor. The association of several of these proteins with the TrkA endosome was confirmed by immunoblot analysis of TrkA endosome extracted protein (Figure 2A). The presence of a large number of proteins implicated in the control of the actin cytoskeleton led us to focus on its role during TrkA signaling endosome formation, sorting, trafficking and retrograde transport. Moreover, based on our findings that NT3/TrkA internalizes but fails to initiate long distance TrkA signaling endosome transport, we hypothesized that differential control of the actin cytoskeleton may represent a point of divergence between NGF/TrkA and NT3/TrkA signaling. Consistent with this idea, NGF and NT3 exhibit dramatic differences in their capacities to evoke filipodial protrusions from their growth cones (Figure S4), which is an actin-dependent process (Faix and Rottner, 2006).
Figure 2
Figure 2
Actin depolymerization in distal axons is essential for retrograde trafficking of TrkA signaling endosomes
The role of actin during internalization and trafficking of endosomes in yeast is well documented, and its regulation is critical for driving the key sequential steps of invagination, scission and post-scission movement of endosomes (Robertson et al., 2009). In mammalian cells, a role for actin modulation during endocytosis and endosome trafficking is much less clear and may vary depending on cell type and context (Fujimoto et al., 2000). We used the actin modifying drugs Latrunculin A (LatA) and Jasplakinolide (Jasp) to assess the roles of F-actin and actin depolymerization, respectively, during internalization and transport of TrkA endosomes in sympathetic neurons. Surprisingly, the complete loss of F-actin in distal axons or cell bodies and proximal axons of sympathetic neurons exposed to the actin disrupting compound LatA had no discernable effect on the formation or retrograde transport of TrkA endosomes as assessed using the Flag-TrkA endosome monitoring assay (Figure 2B). In stark contrast, stabilization of the actin cytoskeleton with Jasp led to a complete loss of TrkA endosome retrograde transport, suggesting that F-actin breakdown is required for either formation or transport of TrkA endosomes (Figure 2C). In Jasp-treated distal axons, however, internalization of TrkA and formation of TrkA endosomes were unaffected (Figure 2C, right panels). Thus, actin depolymerization is required within distal axons at a stage following TrkA internalization but prior to long-distance microtubule-based retrograde transport of TrkA signaling endosomes.
To determine whether the signaling endosome itself can modulate actin dynamics, we next performed an in vitro actin disassembly assay using NGF/TrkA endosomes purified from PC12 cells. Indeed, purified TrkA endosomes accelerated the disassembly of F-actin filaments in vitro (Figure 2D). These results show that TrkA endosomes are associated with actin modifying proteins, they have an innate ability to promote actin depolymerization, and the depolymerization of filamentous actin is essential for retrograde NGF/TrkA endosome transport at a step following TrkA internalization and the formation of TrkA early endosomes within distal axons.
Actin depolymerization mediated by cofilin is required for TrkA endosome retrograde transport and survival
We next sought to determine which, if any, of the endosome-associated proteins identified in our proteomic screen modulates NGF/TrkA endosome-associated F-actin and whether any of these endosomal components account for the differential capacities of NGF and NT3 to support TrkA endosome retrograde transport. Two prime candidates are cofilin, a well-known actin filament severing protein, and the small G-protein Rac1, which has many and varied effects on actin networks (Hall, 2005). Indeed, complementary immunoblot and immunocytochemical analyses revealed that both cofilin and Rac1 are associated with NGF/TrkA endosomes (Figure 2A, ,3B,3B, ,4B).4B). Interestingly, experiments aimed at characterizing the association between cofilin and TrkA endosomes revealed that this association is a defining feature of NGF, as opposed to NT3 formed TrkA endosomes. In sympathetic neurons deprived of NGF for 24 hours and then stimulated with NGF for 15 minutes, cofilin was found to be abundantly associated with early endosome preparations (Figure 3A). In contrast, relatively little cofilin was associated with endosomes harvested from NT3-treated neurons. Moreover, cofilin associated with NGF-formed endosomes was predominantly the active, Serine 3 (Ser3)-unphosphorylated form of the protein, whereas cofilin associated with endosomes following NT3 treatment was more highly phosphorylated on Ser3, which renders it catalytically inactive (Figure 3A). When total phospho-cofilin levels in whole neuron extracts were analyzed, a modest decrease was detected following NGF treatment while NT3 had no effect on cofilin phosphorylation (Figure 3A lower panels). In addition, although NGF and NT3 treatments led to the formation of comparable amounts of internalized TrkA complexes in distal axons of sympathetic neurons, the majority of NGF/TrkA endosomes were colocalized with cofilin while only 23% of NT3/TrkA endosomes colocalized with cofilin (Figure 3B). These findings indicate that a unique ability of NGF is to promote the recruitment of the catalytically active form of the actin-severing protein cofilin to the TrkA signaling endosome.
Figure 3
Figure 3
The actin-severing protein cofilin is associated with TrkA signaling endosomes and is required for their retrograde transport
Figure 4
Figure 4
Rac1 is differentially activated by NGF and NT3 and is associated with NGF/TrkA-containing endosomes
The finding of an association between cofilin and NGF/TrkA signaling endosomes led us to ask whether this endosome-associated protein plays a role in retrograde TrkA endosome trafficking in axons and retrograde survival signaling. We found that retrograde transport of NGF/TrkA endosomes from distal axons to cell bodies was dramatically impaired in neurons expressing an shRNA directed against cofilin (Figure 3C), which efficiently knocked down cofilin levels in sympathetic neurons (Figure S4C). Cofilin deficiency did not affect TrkA internalization nor did it perturb the extent of TrkA endosome colocalization with the early endosome marker Rab5 (Figure S4B). Moreover, in compartmentalized microfluidic chambers, retrograde survival signaling was compromised in neurons in which cofilin levels were reduced (Figure 3D). In contrast, knockdown of cofilin did not result in death of neurons in which NGF was applied directly to cell bodies (Figure 3D), demonstrating a requirement for cofilin for long distance TrkA endosome-based survival signaling but not for NGF/TrkA signaling emanating from TrkA on the surface of the soma. Thus, the actin-severing protein cofilin is required for TrkA signaling endosome maturation at a step following internalization and the formation of Rab5+ early endosomes, but prior to initiation of long-distance retrograde transport.
We next asked whether the loss of cofilin-mediated actin depolymerization was responsible for the defect in retrograde TrkA transport in neurons lacking cofilin. As noted above, exposure of distal axons to LatA, which results in an undetectable amount of filamentous actin within axons (data not shown), does not impair TrkA endosome formation nor does it compromise retrograde TrkA endosome transport (Figure 2B). Remarkably, when added to distal axons of neurons in which cofilin was eliminated, LatA almost completely rescued the deficit in retrograde TrkA endosome transport (Figure 3E), indicating that pharmacological disassembly of F-actin can substitute for cofilin. Thus, while the actin cytoskeleton is dispensable for NGF-dependent internalization of TrkA, cofilin-mediated disassembly of F-actin is essential for retrograde transport of TrkA signaling endosomes from distal axons to the cell body and thus retrograde survival signaling.
Rac1 is differentially activated by NGF and NT3 and is associated with NGF/TrkA-containing endosomes
Like cofilin, the Rho family GTPase, Rac1, was found to be associated with purified TrkA endosomes (Figure 2A, ,3A).3A). Moreover, Rac1 is a major regulator of the actin cytoskeleton. Under certain conditions, Rac1 is necessary for cofilin activation through the promotion of dephosphorylation of cofilin Ser3 (Pandey et al., 2009) and it is implicated in Trk receptor endocytosis and transport (Philippidou et al., 2011; Valdez et al., 2007). Additionally, regulators of Rac1 activity are implicated in the control of endosomal trafficking (Sun et al., 2006). Therefore, we next asked whether Rac1 mediates NGF/TrkA signaling endosome formation, maturation and/or trafficking in sympathetic neurons and if Rac1 signaling controls cofilin activity and/or its association with the TrkA endosome.
We first assessed whether both NGF/TrkA and NT3/TrkA signaling activate the Rac1 signaling pathway in sympathetic neurons or whether, like cofilin, Rac1 activation is more prominently associated with NGF/TrkA. To do this, sympathetic neurons were treated for 20 minutes with either NGF or NT3, and lysates were subjected to precipitation using GST-PAK, which binds with high affinity to the active, GTP-bound form of Rac1 and its relative Cdc42. Immunoprecipitants were then subjected to immunoblot analysis using an antibody specific for Rac1. Interestingly, while NGF treatment of sympathetic neurons led to an increase in the level of Rac1-GTP, NT3 was unable to activate Rac1 (Figure 4A), despite these neurotrophins triggering comparable amounts of TrkA phosphorylation, activation of the canonical effectors (Figure 1B, S1A), and TrkA internalization (Figure 1D, S1B-C). GST-PAK was also used to assess the subcellular distribution of Rac1-GTP in sympathetic neurons that were either untreated or treated with NGF or NT3. These Rac1-GTP localization experiments were done using sympathetic neurons from TrkAFlag mice; double labeling with the Flag-TrkA endosome monitoring assay allowed for an assessment of active Rac1 and TrkA endosome colocalization. The GST-PAK immunolabelling assay specifically detects Rac1-GTP in these immuno-localization experiments since we observed a near complete loss of GST-PAK binding sites in neurons pre-treated with a Rac1 shRNA virus that significantly reduces expression of Rac1 (Figure S5A). Consistent with findings of the Rac1 pull-down/immunoblot experiments and the endosome purification experiments (Figure 2A, ,4A),4A), NGF treatment of distal axons of compartmentalized TrkAFlag sympathetic neurons led to a robust increase in the number of active Rac1-GTP punctae (Figure 4B). Remarkably, Rac1-GTP punctae were observed throughout the neuron, including the entire axon and cell body, and these Rac1-GTP punctae were colocalized with Flag-TrkA endosomes (Figure 4B). Although NT3 led to a robust accumulation of Flag-TrkA endosomes in distal axons, these NT3/TrkA endosomes were not observed in proximal axons and cell bodies and they were rarely, if ever, associated with Rac1-GTP (Figure 4B). These findings indicate that NGF, but not NT3, activates Rac1 in sympathetic neurons and, as is the case for cofilin, Rac1-GTP is associated with NGF/TrkA endosomes, while NT3-formed endosomes are devoid of Rac1-GTP.
Rac1 is necessary and sufficient for retrograde TrkA trafficking and retrograde survival signaling
To test whether Rac1 is required for formation or transport of NGF/TrkA endosomes, compartmentalized sympathetic neurons were infected with a virus expressing a dominant negative form of Rac1 (RacN17), and the Flag-TrkA endosome transport assay was performed following treatment of distal axons with NGF. Although Rac1 was found to be dispensable for NGF-dependent internalization of TrkA (Figure S6A), Rac1 is required for the retrograde transport of TrkA endosomes (Figure 5A-B). This conclusion is supported by experiments that used either a Rac1 shRNA to reduce Rac1 levels or the small molecule Rac1 inhibitor, EHT 1864 (data not shown), which blocks activation of all Rac family members by maintaining them in their GDP-bound state (Shutes et al., 2007). Moreover, as was observed following inhibition of cofilin, inhibition of Rac1 with RacN17 led to a complete loss of retrograde survival signaling in neurons in which NGF was added exclusively to distal axons (Figure 5C). Parallel to our findings with cofilin inhibition, Rac1 inhibition did not perturb survival of neurons in which NGF was applied directly to cell bodies (data not shown).
Figure 5
Figure 5
Rac1 is necessary and sufficient for retrograde TrkA trafficking and retrograde survival in sympathetic neurons
We next asked whether NT3's inability to support retrograde transport of TrkA endosomes and retrograde survival signaling is due to its inability to activate Rac1 and form TrkA endosomes that associate with Rac1-GTP. This possibility was tested in experiments in which a constitutively active form of Rac1 (RacV12) was expressed in TrkAFlag sympathetic neurons grown in compartmentalized microfluidic chambers. Expression of RacV12 did not affect retrograde transport of Flag-TrkA endosomes following NGF treatment of distal axons. Remarkably, application of NT3 to distal axons of RacV12-expressing neurons resulted in retrograde accumulation of Flag-TrkA (Figure 5D) to an extent comparable to that seen in NGF-treated neurons (Figure 5E). Moreover, expression of RacV12 in sympathetic neurons enabled NT3 to support retrograde survival (Figure 5F). These findings indicate that Rac1 is essential for retrograde NGF/TrkA trafficking and retrograde survival signaling, and that the inability of NT3/TrkA endosomes to activate Rac1 accounts for their failure to support retrograde TrkA survival signaling.
To determine whether Rac1 activation mediates TrkA endosome formation or maturation, we asked whether it is required for internalization of the NGF-TrkA complex. We found that TrkA internalization is unaffected by RacN17 in sympathetic neurons (Figure S6A). These findings are consistent with the observations that NT3, which does not activate Rac1 (Figure 3), promotes TrkA internalization (Figure 1D, S1B,C). Thus, like cofilin, Rac1 functions at a step post-internalization and most likely during maturation of transport-competent NGF/TrkA signaling endosomes. In further support of this idea, blocking dynamin, a molecule required for scission of early endosomes from the plasma membrane, using either a pharmacological inhibitor (Dynasore) or a virus that expresses a dominant-negative form of dynamin (K44A dynamin), resulted in attenuation of the levels of Rac1-GTP following NGF treatment (Figure S6B,C). Thus, internalization of TrkA is necessary but not sufficient for Rac1 activation.
Rac1 activation mediates cofilin colocalization with signaling endosomes
NGF activates Rac1 and promotes the association of both Rac1-GTP and cofilin with TrkA endosomes, whereas NT3 does not. Moreover, Rac1 and cofilin are both dispensable for TrkA internalization; however, they are required for NGF/TrkA signaling endosome maturation, retrograde TrkA endosome transport, and retrograde survival. While cofilin catalyzes actin disassembly, Rac1 is implicated in the control of both actin assembly and disassembly depending on cell type and context (Hall, 2005). What is the basis of the physical and functional relationship between Rac1 and cofilin in the context of the TrkA signaling endosome? In neutrophils and platelets, Rac1 is necessary for cofilin activation, actin depolymerization, and actin free barbed end formation (Pandey et al., 2009; Sun et al., 2007). To ask whether Rac1 also functions upstream of cofilin in the context of TrkA signaling endosomes in sympathetic neurons, we tested the requirement of Rac1 activity for the colocalization of cofilin and NGF/TrkA endosomes. To do this, the extent of colocalization between cofilin and Flag-TrkA endosomes in distal axons of sympathetic neurons was quantified for control and RacN17-expressing neurons. Interestingly, RacN17 prevented NGF-dependent colocalization of cofilin with TrkA endosomes in distal axons (Figure 6), indicating that Rac1 initiates a signaling pathway that leads to colocalization of cofilin and NGF/TrkA endosomes. Taken together, these findings indicate that NGF, but not NT3, activates a TrkA–Rac1–cofilin signaling module associated with the TrkA endosome that mediates F-actin severing and the formation of retrograde transport-competent TrkA signaling endosomes.
Figure 6
Figure 6
Rac1 activation is required for localization of cofilin to signaling endosomes
Differential sensitivity to endosomal acidification accounts for differences in NGF/TrkA and NT3/TrkA endosomes and retrograde survival signaling
NT3, like NGF, promotes TrkA phosphorylation, activation of canonical TrkA signaling pathways, survival signaling when applied directly to cell bodies, and TrkA internalization within distal axons. Why, then, is NT3 unable to promote the formation of TrkA endosomes associated with the actin modulators Rac1-GTP and cofilin, retrograde TrkA trafficking, and retrograde survival signaling? One intriguing possibility is that the NGF/TrkA and NT3/TrkA complexes are differentially labile following endocytosis, and continual ligand–receptor association within the context of TrkA endosomes may be essential for the maturation of transport-competent signaling endosomes. We postulated that the stability of neurotrophin–TrkA interactions within the mildly acidic environment of the early endosome (Overly and Hollenbeck, 1996) determines the state of activity of endosome-associated TrkA receptors. Indeed, differential pH sensitivities of ligand-receptor interactions within early endosomes have been described for EGF family ligands and their receptors (Roepstorff et al., 2009). On the cell surface, TrkA binds with higher affinity to NGF than to NT3; it is conceivable that NT3 dissociates more readily from TrkA in the lower pH environment of the early endosome rendering it “inactive” with respect to TrkA signaling. To test this idea, we first asked whether NGF and NT3 exhibit differential pH sensitivity for TrkA activation. PC12 cells were incubated with media buffered to different pH values ranging from pH 5.5 to 7 and stimulated with either NGF or NT3. Cell lysates were subjected to immunoblotting for the phosphorylated, active form of TrkA (P-TrkA). Interestingly, although NGF and NT3 similarly stimulated TrkA phosphorylation at pH 7, NGF but not NT3 promoted phosphorylation of TrkA at pH 6 and below (Figure 7A). To ask if this difference is due to an inability of NT3 to bind to TrkA at low pH, we next performed radiolabeled neurotrophin binding assays using media buffered at different pH values. 125I-labeled neurotrophins were incubated with COS cells transfected with an expression vector encoding full length TrkA or control vector-transfected COS cells, which served as a measure of non-specific binding. Comparable amounts of 125I-NGF bound to TrkA at pH 7.4, 6.0 and 5.5 (Figure 7B). In dramatic contrast, the ability of 125I-NT3 to bind to TrkA was markedly sensitive to acidic conditions (Figure 7B). 125I-NT3 bound strongly to TrkA at neutral pH, but binding of 125I-NT3 to TrkA at pH 5.5 was nearly undetectable. Thus, NGF and NT3 differ in their abilities to bind and activate TrkA under the low pH conditions typically found within early endosomes.
Figure 7
Figure 7
Blocking endosome acidification enables NT3-mediated retrograde transport
We next asked whether the differential pH sensitivity of TrkA activation by NGF and NT3 accounts for differences in signaling endosome formation and signaling. We thus determined whether prevention of endosome acidification enables NT3/TrkA endosomes to activate Rac1, form mature Rac1-GTP and cofilin-associated endosomes, and support their retrograde transport to cell bodies. Distal axons of compartmentalized TrkAFlag sympathetic neurons were incubated with ammonium chloride, which becomes ion-trapped within intracellular organelles and prevents their acidification, or bafilomycin, a pharmacological inhibitor of the vacuolar H+-ATPase (V-ATPase), a pump that mediates acidification of early endosomes (Marshansky and Futai, 2008) and which we identified as a TrkA signaling endosome component (Supplementary Table 1). The Flag-TrkA retrograde transport assay was then performed in neurons in which distal axons were treated with either NGF or NT3. Remarkably, following addition of either ammonium chloride or bafilomycin to distal axons, NT3 promoted retrograde transport of TrkA endosomes to cell bodies comparably to NGF (Figure 7C, D). Moreover, NT3 promoted colocalization of cofilin with the TrkA endosome (Figure 7E) and activation of Rac1 (Figure 7F) in neurons in which the V-ATPase was inhibited. Additionally, expression of the dominant negative RacN17 abolished retrograde transport of TrkA induced by NGF and NT3 in the presence of bafilomycin (Figure 7D). Taken together, these findings support a model in which NGF/TrkA endosomes employ a TrkA–Rac1–cofilin signaling pathway that mediates actin severing, a necessary step for the maturation of early endosomes into retrograde transport-competent signaling endosomes. Moreover, NGF and NT3 differ in their capacities to produce Rac1-GTP/cofilin-containing endosomes and, thus, their abilities to support retrograde survival signaling due to their differential sensitivities to endosomal acidification. NT3/TrkA complexes are labile within the context of early endosomes, and as a result, mature, transport competent NT3/TrkA endosomes fail to form. On the other hand, NGF/TrkA complexes are stable under the acidic conditions of the early endosome enabling endosomal recruitment of Rac1-GTP and cofilin, severing of F-actin, and maturation of TrkA endosomes that are retrogradely transported to the soma.
Here, we demonstrate that the NGF/TrkA endosome is intimately associated with the actin cytoskeleton and that endosome constituents control actin severing, which is an obligate step for maturation of transport-competent signaling endosomes. Moreover, NGF/TrkA endosomes, but not NT3/TrkA endosomes, are differentially associated with key modulators of the actin cytoskeleton, explaining why NGF/TrkA endosomes are transported retrogradely to the cell body to support survival whereas NT3/TrkA endosomes are not. Our findings support a model in which the actin cytoskeleton imposes a physical barrier that restricts maturation of Rab5+ TrkA early endosomes into retrograde transport competent signaling endosomes, and that the actin-severing activity associated with NGF/TrkA early endosomes is essential for overcoming this barrier thereby enabling endosome maturation, association with the microtubule transport machinery, retrograde transport, and retrograde survival and synaptogenic signaling. Finally, we found that NGF and NT3 differ in their capacity to support formation of mature, transport-competent TrkA endosomes associated with actin modulators because of the differentially labile nature of the neurotrophin-TrkA interactions within the context of early endosomes. These findings define a novel function of actin modulation in the control of TrkA endosome maturation and signaling, and they explain how NGF produced by final target fields is the sole neurotrophin responsible for retrograde survival of developing sympathetic neurons.
Endosomal signals control actin dynamics
How do NGF/TrkA endosomes promote actin disassembly during their maturation into transport-competent signaling endosomes? We found that active forms of Cofilin and Rac1 are associated with and required for retrograde transport of NGF/TrkA endosomes. Moreover, both Rac1 and cofilin are necessary for neuronal survival when NGF is acting exclusively on distal axons, but not for survival signaling in response to NGF applied directly to cell bodies. Thus, Rac1 and cofilin are essential for propagation of retrograde signaling and not TrkA survival signaling. Furthermore, Rac1 inhibition and TrkA endosome/cofilin colocalization experiments place Rac1 upstream of cofilin. Therefore, an early endosome-associated TrkA–Rac1–cofilin–actin severing signaling module is required for NGF/TrkA endosome maturation and retrograde transport.
The precise temporal and spatial patterns of activation of Rac1 and cofilin are likely crucial to their functions during endosome maturation and signaling. We found that Rac1-GTP and cofilin are both required at a post-endocytic step for the formation of transport-competent TrkA endosomes. Recent findings in HeLa cells suggest that activation of Rac1 follows, and likely requires, clathrin-mediated endocytosis of receptor tyrosine kinases (Palamidessi et al., 2008). Consistent with this, we found that Rac1 co-purifies with TrkA endosomes and that PAK binding sites are associated with Flag-TrkA+ endosomes, whereas little or no PAK binding is associated with the plasma membrane. Moreover, inhibition of TrkA endocytosis in sympathetic neurons using either a dominant negative form of dynamin or the dynamin inhibitor, dynasore, prevented Rac1 activation (Figure S6B-C). Conversely, inhibition of Rac1 did not prevent internalization of TrkA. Thus, TrkA internalization is required for NGF-dependent production of Rac1-GTP, which is tethered to the NGF/TrkA endosome. On the other hand, internalization of TrkA is not sufficient for Rac1 activation since NT3 promotes formation of TrkA endosomes that are devoid of Rac1-GTP. Our results showing a requirement of Rac1 for TrkA signaling endosome maturation and retrograde transport are in agreement with previous studies in which perinuclear accumulation of TrkA following NGF treatment in PC12 cells (Valdez et al., 2007) and EGF-dependent retrograde accumulation of an EGFR-TrkB chimeric receptor expressed in sympathetic neurons (Philippidou et al., 2011) were blocked by RacN17. Thus, activation of Rac1 is essential for trafficking of Trk signaling endosomes.
Several questions remain as to how Rac1-GTP promotes recruitment of cofilin to the NGF/TrkA early endosome and how cofilin phosphorylation and activity are controlled. This is potentially broadly relevant since in other cell types Rac1 is also required for dephosphorylation of cofilin Ser3 following receptor activation (Pandey et al., 2009; Sun et al., 2007). In at least one known case, Rac1 promotes activation of the cofilin phosphatase, Slingshot (SSH), leading to dephosphorylation of cofilin Ser3 and catalytic activation (Kligys et al., 2007). These findings also implicate a relief to inhibition of cofilin activity triggered by SSH and mediated by the binding of 14-3-3 to cofilin. Interestingly, our proteomic analysis identified both SSH and several isoforms of 14-3-3 proteins as TrkA signaling endosome-associated proteins (Supp. Table 1). We propose that Rac1 controls the actin-severing activity of the NGF/TrkA endosome and actin depolymerization by endosomal recruitment and activation of cofilin, possibly through a mechanism involving SSH and 14-3-3 proteins.
Differential effects of NGF and NT3 on TrkA-mediated transport and survival
Although both NGF and NT3 promote TrkA autophosphorylation, activation of TrkA signaling events, TrkA-dependent axonal extension and, when applied directly to cell bodies, neuronal survival, it is remarkable that NT3 is completely incapable of supporting retrograde transport of TrkA and retrograde survival of sympathetic neurons. We found that NT3 does lead to internalization of Flag-TrkA within distal axons, to a similar extent as NGF. Since prior to visualization of Flag-TrkA endosomes the neurons were treated with a salt-acid wash to remove cell surface Flag antibody, the anti-Flag-labeled punctae detected in these experiments represents TrkA endosomes that have undergone a sufficient degree of endocytosis as to render them invulnerable to the surface antibody stripping conditions. Interestingly, the NT3/TrkA complexes that are internalized associate with the early endosome protein Rab5 but they are neither associated with Rac1-GTP or cofilin nor are they retrogradely transported to cell bodies. It is possible that NT3/TrkA is incapable of activating pathways required for either late stages of endocytosis, such as scission or translocation, or maturation of Rab5+ early endosomes into endosomes that associate with the microtubule transport machinery. In any case, NT3/TrkA complexes internalize but fail to form transport-competent endosomes associated with actin cytoskeleton modulators that enable endosome maturation and long-range retrograde survival.
The pH sensitive nature of NT3-mediated TrkA signaling
How do NGF and NT3 acting on the common receptor TrkA lead to differential activation of the Rac1-GTP–cofilin–actin signaling module? Our findings suggest that a key difference between NGF/TrkA and NT3/TrkA endosomes is the diminished ability of NT3 to bind to and support TrkA activity immediately following early endosome formation and acidification. We found that NT3 is incapable of activating TrkA and displays markedly reduced binding to TrkA at pH values below 7.0. Conversely, NT3 is capable of supporting the formation of TrkA endosomes associated with modulators of the actin cytoskeleton and retrograde transport of TrkA endosomes under conditions in which endosome acidification is prevented. We thus propose that differential maintenance of TrkA signaling within the context of the early endosome accounts for the distinct roles of NT3 and NGF during sympathetic neuron development (Figure S7). The labile nature of NT3/TrkA complexes following endosome formation and acidification ensures a transient, local mode of action of NT3. This accounts for NT3's ability to support axonal extension but not retrograde survival (Belliveau et al., 1997; Kuruvilla et al., 2004; Figure 1A). In contrast, acid-stable NGF/TrkA complexes promote activation of an endosomal TrkA–Rac1–cofilin–actin signaling module, which enables maturation of transport-competent signaling endosomes that propagate retrogradely to the cell body where they support survival and the formation of synapses with preganglionic partners. Interestingly, the ability of NT3 to promote retrograde transport of TrkA and retrograde survival in neurons expressing a constitutively active Rac1 suggests that there may be sufficient ligand-receptor engagement under acidic endosomal conditions to maintain the low levels of downstream signaling in the cell body that are necessary for survival. An alternate possibility is that RacV12 may enable survival signaling from NT3-formed endosomes. Ultimately, the differential sensitivity to endosomal pH accounts for the unique ability of NGF, and not NT3, to establish proper matching between the size of the neuronal population and the size and demands of the target field.
In summary, our findings indicate that NGF/TrkA endosomes employ a signaling pathway composed of Rac1 and cofilin that directs the breakdown of F-actin, a necessary step for maturation of retrograde transport-competent TrkA signaling endosomes. Moreover, the formation of TrkA endosomes associated with Rac1 and cofilin represents a divergence point for NGF and NT3 signaling in sympathetic neurons. Indeed, NT3/TrkA endosomes are devoid of Rac1-GTP and catalytically active cofilin and are incapable of maturing into retrograde transport endosomes. This is reconciled by NT3's inability to remain engaged with TrkA to support the recruitment of Rac1-GTP and cofilin to NT3/TrkA endosomes under the acidic conditions of the early endosome. We propose that differential activation of endosomal signaling pathways that culminate in actin severing and relief of an “actin block” on maturation of transport-competent TrkA signaling endosomes accounts for the differences between the local-axon growth promoting effects of the intermediate target-derived factor, NT3, and the long-distance, retrograde-mediated effects of final target-derived NGF.
Sympathetic neuron culture and the Flag-TrkA endosome transport assay
Sympathetic neurons from superior cervical ganglia (SCG) were cultured as previously described (Deppmann et al., 2008; Park et al., 2006). The Flag-TrkA endosome transport assay was performed as described previously (Sharma et al., 2010).
Actin disassembly assay
Purified non-muscle actin (Cytoskeleton, Inc.) was polymerized in F-buffer (5 mM Tris-HCl pH 8.0, 0.2mM CaCl2, 50 mM KCl, 2 mM MgCl2 and 1mM ATP). F-actin (2 μM) was then incubated with TrkB/A endosome-coated Dynabeads or control Dynabeads at room temperature for 15 min. Addition of 4 μM DBP (Sigma) initiated the actin disassembly reaction. At 15 min following DBP addition, beads were magnetically removed from samples and polymerized and monomeric actin was separated by ultracentrifugation at 100,000 × g for 20 min. Protein was visualized by SDS-PAGE and Coomassie staining.
Cell fractionation for immunoblot, electron microscopy, and mass spectroscopy
A detailed procedure for the purification of TrkA endosomes and their preparation for immunoblot, electron microscopy and mass spectroscopy can be found in the Extended Experimental Procedures.
GST-PAK immunoassay and pulldown assay
Sympathetic neurons were grown in either mass culture or microfluidic chambers, treated as indicated, and fixed for 10 min using 4% paraformaldehyde. Cells were then washed and blocked in PBS containing 1%BSA/0.1%TritonX-100 for 30 min at room temperature. This was followed by incubation with a solution containing 5 μg/ml GST-PAK-PBD for 4h to overnight at 4°C. Next, cells were washed and incubated with anti-GST-Alexa488 secondary antibody for 30 min at room temperature, mounted, and examined using confocal microscopy to visualize sites of Rac1-GTP. For pulldown assays, PC12 6-24 cells that overexpress TrkA (Stephens et al., 1994) or sympathetic neurons from 3 litters of WT mice were grown for 2 DIV and serum-deprived (PC12 cells) or serum- and NGF-deprived with anti-NGF and BAF (sympathetic neurons) for 16h. Cells were treated as described in the legends and pulldowns were done as described in Extended Experimental Procedures.
Highlights
Proteomic analysis by mass spectroscopy of TrkA endosomes indicates that actin modulatory proteins are components of the endosome.
Retrograde transport of TrkA endosomes is dependent on F-actin breakdown.
Sequential recruitment and activation of Rac1 and cofilin to the TrkA endosome governs maturation of transport-competent TrkA endosomes.
Differences in sensitivity to endosomal pH by NGF/TrkA and NT3/TrkA accounts for the unique ability of NGF to form endosomes that associate with cofilin and Rac1 and promote retrograde TrkA endosome transport.
Supplementary Material
Acknowledgments
We thank Alex Kolodkin, Christopher Deppmann and members of the Ginty laboratory for helpful discussion and comments on this manuscript. We thank William Mobley and Janice Valletta for advice about cell fractionation, Robert Cole and Marjan Gucek at the JHU Mass Spectroscopy and Proteomics Facility for performing the mass spectroscopy proteomic analysis, Ann Taylor and Huy Vo for advice and help with microfluidic chamber devices, and Douglas Robinson for advice on actin depolymerization assays. This work was supported by the NIH grants NS34814 (DDG), NS18218 (SH), and The Silvio Conte Center for Neuroscience Research (DDG). DDG is an investigator of the Howard Hughes Medical Institute.
Footnotes
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.
  • Belliveau DJ, Krivko I, Kohn J, Lachance C, Pozniak C, Rusakov D, Kaplan D, Miller FD. NGF and neurotrophin-3 both activate TrkA on sympathetic neurons but differentially regulate survival and neuritogenesis. J Cell Biol. 1997;136:375–388. [PMC free article] [PubMed]
  • Davies AM, Minichiello L, Klein R. Developmental changes in NT3 signalling via TrkA and TrkB in embryonic neurons. EMBO J. 1995;14:4482–4489. [PubMed]
  • Delcroix JD, Valletta JS, Wu C, Hunt SJ, Kowal AS, Mobley WC. NGF signaling in sensory neurons: evidence that early endosomes carry NGF retrograde signals. Neuron. 2003;39:69–84. [PubMed]
  • Deppmann CD, Mihalas S, Sharma N, Lonze BE, Niebur E, Ginty DD. A model for neuronal competition during development. Science. 2008;320:369–373. [PMC free article] [PubMed]
  • Faix J, Rottner K. The making of filopodia. Curr Opin Cell Biol. 2006;18:18–25. [PubMed]
  • Francis N, Farinas I, Brennan C, Rivas-Plata K, Backus C, Reichardt L, Landis S. NT-3, like NGF, is required for survival of sympathetic neurons, but not their precursors. Dev Biol. 1999;210:411–427. [PubMed]
  • Fujimoto LM, Roth R, Heuser JE, Schmid SL. Actin assembly plays a variable, but not obligatory role in receptor-mediated endocytosis in mammalian cells. Traffic. 2000;1:161–171. [PubMed]
  • Hall A. Rho GTPases and the control of cell behaviour. Biochem Soc Trans. 2005;33:891–895. [PubMed]
  • Howe CL, Mobley WC. Long-distance retrograde neurotrophic signaling. Curr Opin Neurobiol. 2005;15:40–48. [PubMed]
  • Kligys K, Claiborne JN, DeBiase PJ, Hopkinson SB, Wu Y, Mizuno K, Jones JC. The slingshot family of phosphatases mediates Rac1 regulation of cofilin phosphorylation, laminin-332 organization, and motility behavior of keratinocytes. J Biol Chem. 2007;282:32520–32528. [PMC free article] [PubMed]
  • Kubler E, Riezman H. Actin and fimbrin are required for the internalization step of endocytosis in yeast. EMBO J. 1993;12:2855–2862. [PubMed]
  • Kuruvilla R, Ye H, Ginty DD. Spatially and functionally distinct roles of the PI3-K effector pathway during NGF signaling in sympathetic neurons. Neuron. 2000;27:499–512. [PubMed]
  • Kuruvilla R, Zweifel LS, Glebova NO, Lonze BE, Valdez G, Ye H, Ginty DD. A neurotrophin signaling cascade coordinates sympathetic neuron development through differential control of TrkA trafficking and retrograde signaling. Cell. 2004;118:243–255. [PubMed]
  • Levi-Montalcini R. The nerve growth factor 35 years later. Science. 1987;237:1154–1162. [PubMed]
  • Levi-Montalcini R, Booker B. Destruction of the Sympathetic Ganglia in Mammals by an Antiserum to a Nerve-Growth Protein. Proc Natl Acad Sci U S A. 1960;46:384–391. [PubMed]
  • Marshansky V, Futai M. The V-type H+-ATPase in vesicular trafficking: targeting, regulation and function. Curr Opin Cell Biol. 2008;20:415–426. [PubMed]
  • Miller FD, Kaplan DR. Neurotrophin signalling pathways regulating neuronal apoptosis. Cell Mol Life Sci. 2001;58:1045–1053. [PubMed]
  • Overly CC, Hollenbeck PJ. Dynamic organization of endocytic pathways in axons of cultured sympathetic neurons. J Neurosci. 1996;16:6056–6064. [PubMed]
  • Palamidessi A, Frittoli E, Garre M, Faretta M, Mione M, Testa I, Diaspro A, Lanzetti L, Scita G, Di Fiore PP. Endocytic trafficking of Rac is required for the spatial restriction of signaling in cell migration. Cell. 2008;134:135–147. [PubMed]
  • Pandey D, Goyal P, Dwivedi S, Siess W. Unraveling a novel Rac1-mediated signaling pathway that regulates cofilin dephosphorylation and secretion in thrombin-stimulated platelets. Blood. 2009;114:415–424. [PubMed]
  • Pazyra-Murphy MF, Hans A, Courchesne SL, Karch C, Cosker KE, Heerssen HM, Watson FL, Kim T, Greenberg ME, Segal RA. A retrograde neuronal survival response: target-derived neurotrophins regulate MEF2D and bcl-w. J Neurosci. 2009;29:6700–6709. [PMC free article] [PubMed]
  • Philippidou P, Valdez G, Akmentin W, Bowers WJ, Federoff HJ, Halegoua S. Trk retrograde signaling requires persistent, Pincher-directed endosomes. Proc Natl Acad Sci U S A. 2011;108:852–857. [PubMed]
  • Reichardt LF. Neurotrophin-regulated signalling pathways. Philos Trans R Soc Lond B Biol Sci. 2006;361:1545–1564. [PMC free article] [PubMed]
  • Robertson AS, Smythe E, Ayscough KR. Functions of actin in endocytosis. Cell Mol Life Sci. 2009;66:2049–2065. [PubMed]
  • Salmon P, Trono D. Production and titration of lentiviral vectors. Curr Protoc Hum Genet. 2007 Chapter 12, Unit 12 10. [PubMed]
  • Shao Y, Akmentin W, Toledo-Aral JJ, Rosenbaum J, Valdez G, Cabot JB, Hilbush BS, Halegoua S. Pincher, a pinocytic chaperone for nerve growth factor/TrkA signaling endosomes. J Cell Biol. 2002;157:679–691. [PMC free article] [PubMed]
  • Sharma N, Deppmann CD, Harrington AW, St Hillaire C, Chen ZY, Lee FS, Ginty DD. Long-distance control of synapse assembly by target-derived NGF. Neuron. 2010;67:422–434. [PMC free article] [PubMed]
  • Shutes A, Onesto C, Picard V, Leblond B, Schweighoffer F, Der CJ. Specificity and mechanism of action of EHT 1864, a novel small molecule inhibitor of Rac family small GTPases. J Biol Chem. 2007;282:35666–35678. [PubMed]
  • Stephens RM, Loeb DM, Copeland TD, Pawson T, Greene LA, Kaplan DR. Trk receptors use redundant signal transduction pathways involving SHC and PLC-gamma 1 to mediate NGF responses. Neuron. 1994;12:691–705. [PubMed]
  • Sun CX, Magalhaes MA, Glogauer M. Rac1 and Rac2 differentially regulate actin free barbed end formation downstream of the fMLP receptor. J Cell Biol. 2007;179:239–245. [PMC free article] [PubMed]
  • Sun YJ, Nishikawa K, Yuda H, Wang YL, Osaka H, Fukazawa N, Naito A, Kudo Y, Wada K, Aoki S. Solo/Trio8, a membrane-associated short isoform of Trio, modulates endosome dynamics and neurite elongation. Mol Cell Biol. 2006;26:6923–6935. [PMC free article] [PubMed]
  • Valdez G, Philippidou P, Rosenbaum J, Akmentin W, Shao Y, Halegoua S. Trk-signaling endosomes are generated by Rac-dependent macroendocytosis. Proc Natl Acad Sci U S A. 2007;104:12270–12275. [PubMed]
  • Wan J, Cheung AY, Fu WY, Wu C, Zhang M, Mobley WC, Cheung ZH, Ip NY. Endophilin B1 as a novel regulator of nerve growth factor/ TrkA trafficking and neurite outgrowth. J Neurosci. 2008;28:9002–9012. [PubMed]
  • Yarar D, Waterman-Storer CM, Schmid SL. A dynamic actin cytoskeleton functions at multiple stages of clathrin-mediated endocytosis. Mol Biol Cell. 2005;16:964–975. [PMC free article] [PubMed]