Full text Figures and data Side by side Abstract Introduction Results Discussion Materials and methods Acknowledgements References Decision letter Author response Article and author information Metrics Abstract Mechanistic studies of axon growth during development are beneficial to the search for neuron-intrinsic regulators of axon regeneration. Here, we discovered that, in the developing neuron from rat, Akt signaling regulates axon growth and growth cone formation through phosphorylation of serine 14 (S14) on Inhibitor of DNA binding 2 (Id2). This enhances Id2 protein stability by means of escape from proteasomal degradation, and steers its localization to the growth cone, where Id2 interacts with radixin that is critical for growth cone formation. Knockdown of Id2, or abrogation of Id2 phosphorylation at S14, greatly impairs axon growth and the architecture of growth cone. Intriguingly, reinstatement of Akt/Id2 signaling after injury in mouse hippocampal slices redeemed growth promoting ability, leading to obvious axon regeneration. Our results suggest that Akt/Id2 signaling is a key module for growth cone formation and axon growth, and its augmentation plays a potential role in CNS axonal regeneration. https://doi.org/10.7554/eLife.20799.001 Introduction Developmental axon growth or axon regeneration requires active molecular machinery that regulates specific transcription factors, growth cone components, and mediators of signal transduction (Fawcett, 2001; Tanabe et al., 2003; Raivich et al., 2004; Lasorella et al., 2006). Injured axons of the adult central nervous system (CNS) do not regenerate, because the ability to activate growth genes and growth cone substantially declines as neurons mature (Fawcett, 2001; Fernandes and Tetzlaff, 2001), and the CNS environment is hostile to those processes (Filbin, 2006; Goldberg et al., 2002; Silver and Miller, 2004). Neutralization of environmental inhibition is not sufficient for axon regeneration; therefore, elucidating intrinsic growth capacity and regulation of the neuron after injury is of critical importance (Yiu and He, 2006; Lee et al., 2009; Fawcett et al., 1992). Indeed, recent studies have proposed that reactivation of the intrinsic growth ability promotes CNS axon regeneration (Rossi et al., 2001; Teng and Tang, 2006; Bouquet and Nothias, 2007; Smith et al., 2009; Leibinger et al., 2013; Watkins et al., 2013). It has been proposed that the mechanisms involved in axonal regeneration of the mature CNS have many features in common with those important in CNS development (Cui, 2006; Harel and Strittmatter, 2006). In addition to its role in neuronal survival (Ahn et al., 2004b; Ahn and Ye, 2005), Akt/PKB (protein kinase B) signaling controls a variety of neuronal responses. It regulates both axon establishment and elongation both during development and in the regeneration of mature neurons through glycogen synthase kinase 3 (GSK3). However, the mechanism of GSK3 control of peripheral axon regeneration is controversial and its function in CNS axon regeneration remains unknown (Jiang et al., 2005; Yoshimura et al., 2006; Kim et al., 2011; Saijilafu et al., 2013; Zhang et al., 2014; Gobrecht et al., 2014). Moreover, Akt links a host of signaling molecules through activation of mTORC1, which regulates cap-dependent protein translation by inhibiting TSC1/2 to allow axon development, growth, and regeneration in CNS (Ma et al., 2008; Li et al., 2008; Morita and Sobue, 2009; Park et al., 2008). However, some evidence suggested an mTORC1 independent pathway that regulates axon regrowth in phosphatase and tensin homolog (Pten) deficient neurons (Park et al., 2008; Yang et al., 2014), which causes aberrant activation of Akt signaling. Thus, although Akt signaling encompasses developmental regulation of the intrinsic neuronal growth and axon regeneration after injury, the roles and molecular mechanism of Akt signaling in the growth of CNS axons remain to be determined. Inhibitor of DNA binding 2 (Id2) is a negative regulator of basic helix-loop-helix (bHLH) transcription factors. During development, Id2 binds to bHLH transcription factors and hampers their ability to activate transcription of several growth inhibitory molecules and receptors, thus promoting axon growth (Jackson, 2006). Id2 degradation by a complex of the anaphase-promoting complex/cyclosome and its activator Cdh1 (APC/CCdh1) reduces axonal growth in the adult (Stegmüller and Bonni, 2005; Lasorella et al., 2006). Conversely, protection from Id2 degradation results in erratic growth and an abnormal distribution of parallel fibers in the cerebral cortex (Konishi et al., 2004), while enhanced Id2 expression in the dorsal root ganglion (DRG) promotes axonal growth after spinal cord injury (Yu et al., 2011). Thus, Id2 contributes to axonal growth during development and may also be involved in the intrinsic inability of the injured axons to regenerate in the adult (Lasorella et al., 2006). However, to our knowledge the specific temporal and spatial signals that may regulate the molecular changes induced by Id2 are not yet understood. In this study, we defined the role of Akt in regulating Id2 functions in axon growth during development and attempted to enhance Akt/Id2 signaling after injury to promote axon regeneration. We identified Id2, as a new binding partner and novel kinase substrate of Akt. Akt-mediated phosphorylation of serine 14 (S14) on Id2 augmented its protein stability through disruption of the association of Id2 and E3 ligase Cdh1. During neuronal differentiation, S14- phosphorylated Id2 is predominantly enriched in the growth cones at the axonal tips where it facilitates axonal growth. This contributed to the maintenance of the growth cone via interaction with radixin, one of the ezrin, radixin, and moesin (ERM) family of proteins, which links F-actin to the plasma membrane. Moreover, in organotypic hippocampal slice culture, reactivation of Akt/Id2 signaling by adeno-associated virus (AAV) two after injury, prominently increased regrowth of axons, whereas ablation of Akt-dependent phosphorylation of Id2 caused failure in axonal regeneration. Our study suggests the molecular basis of intrinsic growth regulation of Akt/Id2 signaling and delineates the potential role of Akt/Id2 in CNS axonal regeneration. Results Akt binds to Id2 and phosphorylates serine 14 Because of the multiple downstream effectors of Akt in both the neuronal soma and axon terminal, this pathway might coordinate different steps of axon growth during development. In an effort to identify downstream targets of the Akt signal that might be involved in the regulation of axon growth, we examined protein interaction profiles using proteomic analysis in PC12 cells stably transfected with a constitutively active (CA) form or a kinase-dead (KD) form of Akt. Interestingly, our proteomic analysis showed that Id2 is a potent binding partner of active Akt (Figure 1—figure supplement 1). Indeed, we found endogenous interaction between Akt and Id2 in mouse brain lysates (Figure 1A); the specific interaction was confirmed in mouse brain extract using purified glutathione S-transferase (GST)-Id2 protein (Figure 1B). Employing Flag-tagged Akt isoforms (Akt1-3), we verified that, among the three isoforms evaluated, Id2 interacts with Akt1 (Figure 1C). Our mapping analysis showed that the PH domain of Akt interacts with Id2 (Figure 1D). Reciprocal experiments with a series of purified GST-tagged Id2 fragments demonstrated that the helix domain of Id2 adjacent to the C-terminus bound to Akt (Figure 1E). Figure 1 with 2 supplements see all Download asset Open asset Akt binds to Id2 and phosphorylates Serine 14. (A) Mouse brain lysates were subjected to immunoprecipitation (IP)/immunoblotting (IB) with the indicated antibodies. (B) GST pull-down assays with purified GST-Id2 protein and P1 mouse brain lysates. (C) Flag-Akt1, 2, or three wee transfected into HEK293T cells together with GFP-Id2, and lysates were subjected to anti-Flag IP followed by IB as indicated. (D) Schematic diagram of the Akt fragments (upper). Flag-Id2 was co-transfected with mammalian GST-Akt fragments into 293T cells and lysates were subjected to GST pull-down assay and IB as indicated (bottom). (E) Schematic diagram of the Id2 fragments (upper). Purified GST-Id2 fragment proteins were pre-bound to GST-resin and reacted with lysate from PC12 cells followed by IB (bottom). Arrows indicate purified Id2 fragments protein. (F–H) In vitro Akt kinase assay was performed with purified GST-proteins and purified active Akt. GSK3β fusion and GST proteins were used as positive and negative controls, respectively. Arrows indicate purified Id2 fragments protein or phosphorylated GST-Id2 fragments proteins (E–G). (I) IB of DIV1-5 cortical neuron lysates probed on the indicated antibodies. Densitometry analysis of IB is shown in the bottom. Data are representative of at least three independent experiments. See also Figure 1—figure supplements 1 and 2. https://doi.org/10.7554/eLife.20799.002 To determine whether Id2 is a substrate of Akt kinase, we generated constructs of GST-tagged Id1-Id4, as four Id genes with highly conserved HLH regions have been identified in human cells (Figure 1—figure supplement 2A), and performed in vitro kinase assays with purified active Akt protein. Among the Id family proteins, only the Id2 protein was substantially phosphorylated by active Akt, although all Id 1–4 proteins interacted with Akt (Figure 1F, Figure 1—figure supplement 2B). We verified that the specific site of Id2 phosphorylation by Akt is located on the very end of the N-terminus, within amino acid residues 1–37 (Figure 1G). Employing anti-phospho-Ser/Thr Akt substrate sequence antibody, we supported the notion that Id2 has a putative phosphorylation site for Akt (Figure 1—figure supplement 2C). In vitro kinase assay with phospho-ablated mutant forms of Id2, revealed that Id2-S14A completely lacked phosphorylation, whereas WT-Id2 and mutation on Serine 5 of Id2, which has been shown to be phosphorylated by cyclin A/cdk2 (Hara et al., 1994), showed strong phosphorylation by Akt. This finding was in agreement with our phospho-proteomic analysis, which revealed S14 to be a putative phosphorylation site of Akt (Figure 1H, Figure 1—figure supplement 2D). To confirm specific phosphorylation at S14 on Id2, we generated phospho-specific antibody that recognized S14 (Figure 1—figure supplement 2E) and demonstrated that S14 is indeed phosphorylated in primary cultured neurons as they develop (Figure 1I). Taken together our data demonstrated that Id2 is a novel binding partner and kinase substrate of Akt in the developing neuron. Akt controls Id2 protein stability in the neuron Id2 was highly expressed in the mouse hippocampus in the embryonic stages (E14 and E17) and decreased after birth and overtime (P7-P28). The level of phospho-Akt paralleled the decrease in Id2, showing a drastic decrease after P14 in the postnatal hippocampus of mouse brain (Figure 2A). Only the level of Akt1, but not that of Akt2 or Akt3, was reduced in a time frame similar to that of Id2 (Figure 2B), suggesting that Akt1 might be relevant in the control of Id2 protein level in neurons, correlating with our observation that Akt1 specifically interacted with Id2 (Figure 1C). Based on this finding, we focused our investigation on the biological significance of this interaction using Akt1, unless otherwise specified. Figure 2 with 1 supplement see all Download asset Open asset Akt controls Id2 protein stability in the neuron. (A–B) Lysates from mouse hippocampus of the indicated days were subjected to IB with the indicated antibodies. (C) PC12 cell were transfected with the indicated combination of HA-Akt or GFP-Cdh1 and the protein level was determined by IB (left). Densitometry analysis of IB is shown on the right. (D) PC12 cells were transfected with GFP-Cdh1 together with HA-vector or HA-Akt (+: 2 μg or ++:4 μg) and probed on IB (left). Densitometry analysis of IB is shown on the right. (E) GST-Id2 was co-transfected with HA-Akt into PC12 cells. Twenty-four hours after transfection, the cells were treated with the proteasome inhibitor MG132. GST-pull down assay was performed to determine ubiqutinated Id2. (F) PC12 cells were transfected with GST-Id2, GFP-cdh1 and increasing amounts of myc-Akt (+: 2 μg/++:4 μg) and the cell lysates were subject to GST pull-down. Immunoblot is shown on the left and quantification of the interaction affinity of GFP-cdh1 and GST-Id2 by densitometry analysis is shown on the right. (G) Transfected PC12 cells were treated with cycloheximide (CHX, 10 μM) as indicated time and probed on the IB (upper). Quantification of the Id2 protein levels by densitometry analysis (bottom). (H) HA-cdh1 was co-transfected with GFP-Id2 WT or mutants into 293T cells and protein levels of Id2 was detected by anti-GFP antibody after IP with HA antibody. (I) PC12 cells were transfected with GFP-Id2 WT, S14A, or S14D with HA-Akt KD or HA-Akt CA and probed on IB (left) Quantification of protein levels is shown in the bottom. (J) PC12 cells were treated with Akt inhibitor VIII (0, 0.1, 0.5 or 1 μM) or PD184352 (1 μM). Amounts of total and phosphorylated Id2 were determined by IB. *p<0.05, **p<0.005 versus indicated (G and I). Values in this figure represent mean ± SEM from three independent experiments and image shown here is representative from at least three independent experiments. See also Figure 2—figure supplement 1. https://doi.org/10.7554/eLife.20799.005 Id2 degradation in neurons is facilitated by APC/CCdh1, which inhibits axonal growth and Cdh1 is a regulatory subunit of the E3 ubiquitin ligase APC/CCdh1 responsible for Id2 degradation (Lasorella et al., 2006). Based on our finding of a development-dependent decline in Akt/Id2 signaling, we wondered if Akt activation regulates Id2 stability by blocking its proteasomal degradation. Treatment with the proteasomal inhibitor MG132 protected against Id2 degradation, confirming that the reduction in Id2 level is facilitated by the ubiquitin-proteasome system (UPS)-dependent degradation (Figure 2—figure supplement 1). Overexpression of Cdh1 markedly reduced endogenous Id2 level; importantly, this effect was prevented by Akt expression in PC12 cells (Figure 2C). In cortical neurons, Id2 protein level was proportionally increased with increased Akt level in the presence of Cdh1, indicating that Akt prevents Id2 degradation (Figure 2D). Accordingly, polyubiquitination of Id2 was efficiently abrogated in the presence, but not in the absence of Akt (Figure 2E), indicating that Akt regulates Id2 protein stability by preventing APC/CCdh1-mediated degradation. Id2 was found to be associated with Cdh1. However, the interaction between Id2 and Cdh1 weakened with increased Akt expression, whereas the interaction of Id2 with Akt increased, suggesting that Akt competes with Cdh1 to bind Id2 (Figure 2F). The half-life of Id2 was lower in a phospho-ablated mutant (GFP-S14A) that could not be phosphorylated by Akt than in GFP-Id2-WT-expressing cells, whereas a phospho-mimetic mutant (GFP-S14D) showed more stable expression after cycloheximide (CHX) treatment (Figure 2G). This implies that Akt protects Id2 from Cdh1-mediated proteasomal degradation through phosphorylation of Id2. To further verify the importance of Id2 phosphorylation by Akt for protein stability, we introduced a phospho-ablated mutant or phospho-mimetic mutant of Id2, along with HA-Cdh1, into PC12 cells. While Id2-WT and Id2-S14D rarely bound to Cdh1, the phospho-ablated mutant form of Id2 largely showed enhanced association with Cdh1, reflecting its instability and weak detection (Figure 2H). Moreover, Id2-WT was found to be more stable in the presence of CA-Akt than in that of KD-Akt, while the Id2-S14A mutant was not stabilized by CA-Akt as it could not be phosphorylated. The protein level of phospho-mimetic Id2-S14D was highly stable regardless of whether CA-Akt or KD-Akt was expressed (Figure 2I). Furthermore, using Akt inhibitor VIII, a chemical inhibitor of Akt signaling, we showed a reduction of Id2 protein levels as its phosphorylation is decreased upon inhibition of AKT phosphorylation, whereas in the presence of PD184352, a chemical inhibitor of MAPK signaling, Id2 stability or phosphorylation is not altered (Figure 2J). These data suggest that Akt-dependent Id2 phosphorylation enhances resistance to Cdh1-mediated degradation of Id2, interrupting the interaction between Id2 and Cdh1. Phosphorylation of Id2 by akt is essential for augmentation of axon growth and branching Akt has been shown to be predominantly localized at the tip of the axon in developing hippocampal neurons (Yan et al., 2006). Furthermore, we found that the level of Akt protein and its activation state are closely related to the expression level of Id2 during development, and that Akt-mediated Id2 phosphorylation enhanced Id2 stability by preventing APC/CCdh1-mediated degradation; therefore, we wondered whether Akt regulates the function of Id2 in axonal growth. During differentiation of rat hippocampal neurons (up to in vitro day (DIV) 5), the spatial distributions of Id2 and Akt were visualized not only in the soma, but also prominently in the precursors of axons and dendrites in the early stages of differentiation (Figure 3—figure supplement 1, DIV1). Id2 and Akt were found in proximal axons, with expression tapering off along the distal axon (Figure 3—figure supplement 1, DIV2); however, the signals were strikingly intense at the distal part of the growing axon with a growth cone and axon branching points in later stages of axon growth (Figure 3A, Figure 3—figure supplement 1, DIV 4 - 5), suggesting that Id2 is potentially a downstream target of Akt in the regulation of axon growth and branching. Figure 3 with 1 supplement see all Download asset Open asset Phosphorylation of Id2 by Akt is essential for augmentation of axon growth and branching. (A) Representative merged image of localization of endogenous Akt and Id2 in the hippocampus neurons (DIV 4). The neurons stained for Id2 (red) and Akt (green). Right panel shows a higher magnification of the region indicated by a box. Scale bar, 10 µm. Image shown here is representative from at least three independent experiments. (B–C) Cultured neurons were transfected with GFP-Id2 WT, S14A, S14D or GFP vector control at day DIV one and fixed at DIV 3. Neurons were stained with anti-Tuj1(red). Representative images with a higher magnification of the region indicated by a box are shown in (B). Quantification of axon length and branching point measurements from three independent experiments is shown in (C). n = 16–24 cells. Error bars, SEM; Scale bar, 50 µm or 10 µm. *p<0.05 versus indicated. Arrows indicate axonal tip and branch points (A–B). See also Figure 3—figure supplement 1. https://doi.org/10.7554/eLife.20799.007 To delineate the roles of Akt/Id2 signaling in the regulation of axon growth, we transfected GFP-Id2 constructs into dissociated rat E18 hippocampal neurons and maintained them. Cultured hippocampal neurons at DIV three were immunolabeled with neuron specific class III beta tubulin (Tuj1) antibody to assess the extent of axon growth. Ectopic expression of Id2-WT led to considerably better axon growth and branching than that of control; S14D expressing neurons showed more abundant branching and extended length of axon than did the control or Id2-WT expressing neurons, revealing high expression at the tip of axon and branching. In contrast, the phospho-ablated mutant (S14A)-expressing neurons exhibited substantially shorter extent of axon growth and less branching in primary cultured hippocampal neurons (Figure 3B,C). Interestingly, we failed to detect GFP signal in the axonal tip of S14A expressing neurons despite no alteration of this signal in the soma, indicating lack of Id2 expression in the growth cone of the growing axon (Figure 3B, third panel). Taken together, these data imply that Akt regulates axon growth and branching by regulating Id2 phosphorylation and that this phosphorylation is essential for Id2 localization in the growth cone and branching. Akt regulates growth cone localization of Id2 in the developing neuron As we found that Akt/Id2 accumulate in the axon tip and branching points in growing hippocampal neurons, and that Akt-mediated Id2 phosphorylation is essential for axonal growth and growth cone localization of Id2, we hypothesized that there might be spatial and temporal correlations between the expression and subcellular localization of Akt/Id2 with development of the growth cone. During the differentiation of hippocampal neurons, we found that prominent endogenous-Id2 expression occurred in the central domain of growth cone and partially colocalized with phalloidin-labeled F-actin, while S14-phospho-Id2 was localized on punctate structures all-around of the growth cone area and filopodia. S14-phospho-Id2 stained strongly in the growth cone leading edge and the peripheral domain, revealing notable co-distribution with phalloidin-labeled F-actin in the growing axon (Figure 4A–C). To more accurately determine the role of Akt/Id2 signaling in the growth cone, we monitored Id2 and S14-phospho-Id2 expression as neuronal development proceeded (Figure 4D, Figure 4—figure supplement 1A). In the early stage (stage 1: DIV 1) both Id2 and S14-phospho-Id2 were observed in the filopodial and lamellipodial structure of the leading margin. However, as the axon developed (stage II~III: DIV 2 and 3), Id2 was predominantly detected in the central microtubule-containing zone, demonstrating complete co-localization with beta tubulin, while S14-phospho-Id2 displayed a more intense signal at the peripheral-domain of growth cone, where there was relatively less beta tubulin staining (Figure 4D,E, Figure 4—figure supplement 1A,B). Stage determination was performed as previously described for hippocampal neurons (Dotti et al., 1988). Moreover, quantitative analysis that determined the expression level of Id2 or S14-phospho-Id2 from soma to axonal tip, supported the notion that S14-phospho-Id2 is relatively enriched in the tip of the growth cone, with respect to that of Id2, as confirmed by fluorescence intensity analysis (Figure 4F,G). Figure 4 with 2 supplements see all Download asset Open asset Akt regulates growth cone localization of Id2 in the developing neuron. (A) Schematic diagram of growth cone, showing microtubule mostly in the central [C] region and F-actin based peripheral [P]region. (B) Representative image of Id2 or S14-phospho-Id2 (green) with phalloidin labeled F-actin (red) and beta-tubulin (Tuj1:blue) in the growth cone of hippocampal neuron (stage3:DIV3). Arrows indicate example of [P]and [C] domain. Scale bar, 5 µm. (C) Quantification of S14-phospho-Id2/ Phalloidin or Id2/ Phalloidin at [P] and S14-phospho-Id2/Tuj1 or Id2/Tuj1 at [C] domain was averaged over multiple growth cones (right and middle). The ration of S14-phospho-Id2/Id2 at [C] and [P] was shown in left. n = 35. *p<0.05. **p<0.005. [P] or [C] domain is outlined by dashed gray or white line based on immunolabeling of phalloidin or Tuj1 in (B). (D) Representative image of beta-tubulin (Tuj1:green) with Id2 or S14-phospho-Id2 (red) in DIV2 neuron. Scale bar, 5 µm. The fluorescent image of DIV 1–3 is shown in Figure 4—figure supplement 1A and the original image of neuron for this representative growth cone is placed in Figure 4—figure supplement 1B. (E) Graphs plot the fluorescence intensity of immunolabeled Id2 (red) and Tuj1 (green) or phosphor Id2 (red) and Tuj1 (green) the arrowed line in Figure 4D is shown in each growth cone image. (F) The hippocampal neuron was fixed and stained with anti-Id2 or S14-phospho-Id2 antibodies (red). The neuron was stained with the Tuj1 (green), and nuclei were counterstained with DAPI. Scale bar, left: 20 µm. Relative immunofluorescence intensity profiles of Id2 and Tuj1 along the axon from cell body to axonal tip (right). (G) Quantification of Id2 and S14-phospho-Id2 signal intensity in the soma or axonal tip respectively. *p<0.05 versus control.Data represent mean ± SEM of three independent experiments. n = 20. See also Figure 4—figure supplements 1 and 2. https://doi.org/10.7554/eLife.20799.009 To further confirm the specificity of anti-S14-phospho antibody, we depleted Id2 in the growing neuron using lentiviral-siRNA or inhibited Akt activity by chemical inhibitor of Akt. We selected Id2-siRNA that selectively reduced Id2 levels, but not the expression of unrelated protein such as beta-actin in PC12 cells (Figure 4—figure supplement 2A). Using selected Id2-siRNA, we generated and purified lentivirus that expresses Id2-siRNA and confirmed that GFP-lentiviral Id2-siRNA can specifically suppress endogenous Id2 protein expression by immunoblotting (Figure 4—figure supplement 2B). Either knockdown of Id2 or treatment of Akt inhibitor diminished the specific signal of S14-phospho-Id2 in the growth cone with abnormal feature of growth cone (Figure 4—figure supplement 2C–E). Thus, our data indicate that S14 phosphorylation of Id2 probably drives its localization in the peripheral region of growth cone. Akt/Id2 signaling promotes axon growth by regulating growth cone development We analyzed the localization of Id2 and S14-phospho-Id2 with phalloidin-labeled F-actin in the growth cone of neurons, we next asked whether Akt/Id2 signaling in the growth cone is involved in Id2 growth cone formation and function. ERM proteins link the actin cytoskeleton to the plasma membrane and play prominent roles in growth cone morphology and motility (Mangeat et al., 1999; Dickson et al., 2002). While ezrin and moesin expression is strongest in the central region of growth cone, radixin is highly stained in the peripheral region with phalloidin-label F-actin (Marsick et al., 2012). When we used an anti-ERM antibody that recognizes an epitope common to all ERM family members, our immunoprecipitation assay mouse brain extract (E18) showed that endogenous Id2 interacts with the ERM proteins (Figure 5A). Interestingly, compared to the binding affinity of Id2 to ERM, brain extracts immunoprecipitated with anti-S14-phospho-Id2 antibody showed a relatively strong interaction with the ERM proteins (Figure 5B), suggesting that S14-phospho-Id2 probably binds to radixin among the ERM proteins based on the peripheral distribution of S14-phospho-Id2 and radixin. Id2 was concentrated in the central region and in the base of some filopodia, while radixin expression was relatively abundant in the peripheral region (Figure 5—figure supplement 1A). Intriguingly, S14-phospho-Id2 was highly expressed in the peripheral filopodia, revealing co-distribution with radixin, as confirmed by fluorescence analysis (Figure 5—figure supplement 1B). Figure 5 with 1 supplement see all Download asset Open asset Akt/Id2 signaling promotes axon growth by regulating growth cone development. (A–B) E18 mouse brain lysates were subject to IP with anti-Id2 or anti-S14-phospho-antibody, followed by IB with anti-ERM antibody. (C–D) GST pull-down assay using cell lysates of PC12 cells transfected with indicated constructs following by IB. (E) Hippocampal neurons were infected with lenti-GFP- si-Id2 or lenti-GFP-scramble control at stage three and fixed after 48 hr. Neurons were stained with anti-radixin antibody (red). Growth cone area is outlined by dashed line based on immunolabeling of radixin. (F) Quantification of growth cone size and number of axonal length was based on radixin fluorescence from three experiments (n = 29–50). Scale bar, 10 µm. Error bars, SEM; **p<0.005 versus control. (G and H) GFP- si-Id2 was introduced to hippocampal neurons at DIV1 along with a series of RFP-Id2-WT, Id2-S14A or Id2-S14D and determined axon length and growth cone size at DIV4. Enlargement of growth cone area was shown in inserted box. (H) Quantification of axonal length and growth cone area (n = 15–21). Scale bar, 20 µm. Error bars, SEM; *p<0.05 **p<0.005. See also Figure 5—figure supplement 1. https://doi.org/10.7554/eLife.20799.012 To further determine the importance of S14 phosphorylation on Id2 in the binding with radixin, we conducted an in vitro binding assay that demonstrated that Id2-WT interacted with radixin, whereas S14A failed to interact with GST-radixin (Figure 5C) and depletion of Akt by shRNA reduced the interaction between Id2 and radixin (Figure 5D, Figure 5—figure supplement 1C–D). Hence, Akt-mediated phosphorylation of Id2 is crucial for its association with radixin in the growing axon, which probably contributes to the growth cone function. To determine the functional consequence of Akt/Id2 signaling in the growth cone, we utilized siRNA mediated knockdown of endogenous Id2. We infected the growing axon from hippocampal neuron (stage III) with either GFP-lentiviral control or GFP-lentiviral Id2-siRNA. Control GFP-expressing lentivirus-infected neurons showed normal growth cone architecture that was visualized with expression of radixin. However, knockdown of Id2 in the growing axon dramatically disrupted growth cone shape and cytoskeletal organization detected in neurons with reduced radixin levels (Figure 5E). Quantitative analysis of neuronal response involving alteration of the growth cone indicated noticeable reduction of growth cone size and shortening of axon length, implicating the importance of Id2 expr
We present Ev-NeRF, a Neural Radiance Field derived from event data. While event cameras can measure subtle brightness changes in high frame rates, the measurements in low lighting or extreme motion suffer from significant domain discrepancy with complex noise. As a result, the performance of event-based vision tasks does not transfer to challenging environments, where the event cameras are expected to thrive over normal cameras. We find that the multi-view consistency of NeRF provides a powerful self-supervision signal for eliminating the spurious measurements and extracting the consistent underlying structure despite highly noisy input. Instead of posed images of the original NeRF, the input to Ev-NeRF is the event measurements accompanied by the movements of the sensors. Using the loss function that reflects the measurement model of the sensor, Ev-NeRF creates an integrated neural volume that summarizes the unstructured and sparse data points captured for about 2-4 seconds. The generated neural volume can also produce intensity images from novel views with reasonable depth estimates, which can serve as a high-quality input to various vision-based tasks. Our results show that Ev-NeRF achieves competitive performance for intensity image reconstruction under extreme noise conditions and high-dynamic-range imaging.
Neural networks trained with ERM (empirical risk minimization) sometimes learn unintended decision rules, in particular when their training data is biased, i.e., when training labels are strongly correlated with undesirable features. To prevent a network from learning such features, recent methods augment training data such that examples displaying spurious correlations (i.e., bias-aligned examples) become a minority, whereas the other, bias-conflicting examples become prevalent. However, these approaches are sometimes difficult to train and scale to real-world data because they rely on generative models or disentangled representations. We propose an alternative based on mixup, a popular augmentation that creates convex combinations of training examples. Our method, coined SelecMix, applies mixup to contradicting pairs of examples, defined as showing either (i) the same label but dissimilar biased features, or (ii) different labels but similar biased features. Identifying such pairs requires comparing examples with respect to unknown biased features. For this, we utilize an auxiliary contrastive model with the popular heuristic that biased features are learned preferentially during training. Experiments on standard benchmarks demonstrate the effectiveness of the method, in particular when label noise complicates the identification of bias-conflicting examples.
Causal dynamics learning has recently emerged as a promising approach to enhancing robustness in reinforcement learning (RL). Typically, the goal is to build a dynamics model that makes predictions based on the causal relationships among the entities. Despite the fact that causal connections often manifest only under certain contexts, existing approaches overlook such fine-grained relationships and lack a detailed understanding of the dynamics. In this work, we propose a novel dynamics model that infers fine-grained causal structures and employs them for prediction, leading to improved robustness in RL. The key idea is to jointly learn the dynamics model with a discrete latent variable that quantizes the state-action space into subgroups. This leads to recognizing meaningful context that displays sparse dependencies, where causal structures are learned for each subgroup throughout the training. Experimental results demonstrate the robustness of our method to unseen states and locally spurious correlations in downstream tasks where fine-grained causal reasoning is crucial. We further illustrate the effectiveness of our subgroup-based approach with quantization in discovering fine-grained causal relationships compared to prior methods.
Chemoresistance is a major cause of treatment failure in many cancers. However, the life cycle of cancer cells as they respond to and survive environmental and therapeutic stress is understudied. In this study, we utilized a microfluidic device to induce the development of doxorubicin-resistant (DOXR) cells from triple negative breast cancer (TNBC) cells within 11 days by generating gradients of DOX and medium. In vivo chemoresistant xenograft models, an unbiased genome-wide transcriptome analysis, and a patient data/tissue analysis all showed that chemoresistance arose from failed epigenetic control of the nuclear protein-1 (NUPR1)/histone deacetylase 11 (HDAC11) axis, and high Nupr1 expression correlated with poor clinical outcomes. These results suggest that the chip can rapidly induce resistant cells that increase tumor heterogeneity and chemoresistance, highlighting the need for further studies on the epigenetic control of the NUPR1/HDAC11 axis in TNBC.
Chemoresistance is a major cause of treatment failure in many cancers. However, the life cycle of cancer cells as they respond to and survive environmental and therapeutic stress is understudied. In this study, we utilized a microfluidic device to induce the development of doxorubicin-resistant (DOXR) cells from triple negative breast cancer (TNBC) cells within 11 days by generating gradients of DOX and medium. In vivo chemoresistant xenograft models, an unbiased genome-wide transcriptome analysis, and a patient data/tissue analysis all showed that chemoresistance arose from failed epigenetic control of the nuclear protein-1 (NUPR1)/histone deacetylase 11 (HDAC11) axis, and high Nupr1 expression correlated with poor clinical outcomes. These results suggest that the chip can rapidly induce resistant cells that increase tumor heterogeneity and chemoresistance, highlighting the need for further studies on the epigenetic control of the NUPR1/HDAC11 axis in TNBC.
Chemoresistance is a major cause of treatment failure in many cancers. However, the life cycle of cancer cells as they respond to and survive environmental and therapeutic stress is understudied. In this study, we utilized a microfluidic device to induce the development of doxorubicin-resistant (DOXR) cells from triple negative breast cancer (TNBC) cells within 11 days by generating gradients of DOX and medium. In vivo chemoresistant xenograft models, an unbiased genome-wide transcriptome analysis, and a patient data/tissue analysis all showed that chemoresistance arose from failed epigenetic control of the nuclear protein-1 (NUPR1)/histone deacetylase 11 (HDAC11) axis, and high Nupr1 expression correlated with poor clinical outcomes. These results suggest that the chip can rapidly induce resistant cells that increase tumor heterogeneity and chemoresistance, highlighting the need for further studies on the epigenetic control of the NUPR1/HDAC11 axis in TNBC.
Chemoresistance is a major cause of treatment failure in many cancers. However, the life cycle of cancer cells as they respond to and survive environmental and therapeutic stress is understudied. In this study, we utilized a microfluidic device to induce the development of doxorubicin-resistant (DOXR) cells from triple negative breast cancer (TNBC) cells within 11 days by generating gradients of DOX and medium. In vivo chemoresistant xenograft models, an unbiased genome-wide transcriptome analysis, and a patient data/tissue analysis all showed that chemoresistance arose from failed epigenetic control of the nuclear protein-1 (NUPR1)/histone deacetylase 11 (HDAC11) axis, and high NUPR1 expression correlated with poor clinical outcomes. These results suggest that the chip can rapidly induce resistant cells that increase tumor heterogeneity and chemoresistance, highlighting the need for further studies on the epigenetic control of the NUPR1/HDAC11 axis in TNBC.
Understanding geometric concepts, such as distance and shape, is essential for understanding the real world and also for many vision tasks. To incorporate such information into a visual representation of a scene, we propose learning to represent the scene by sketching, inspired by human behavior. Our method, coined Learning by Sketching (LBS), learns to convert an image into a set of colored strokes that explicitly incorporate the geometric information of the scene in a single inference step without requiring a sketch dataset. A sketch is then generated from the strokes where CLIP-based perceptual loss maintains a semantic similarity between the sketch and the image. We show theoretically that sketching is equivariant with respect to arbitrary affine transformations and thus provably preserves geometric information. Experimental results show that LBS substantially improves the performance of object attribute classification on the unlabeled CLEVR dataset, domain transfer between CLEVR and STL-10 datasets, and for diverse downstream tasks, confirming that LBS provides rich geometric information.
'%3e%3cg id='b'%3e%3cpath id='a' stroke='%23000' stroke-width='2' d='M-6-25H6m-12 3H6m-12 3H6'/%3e%3cuse xlink:href='%23a' y='44'/%3e%3c/g%3e%3cpath stroke='%23fff' d='M0 17v10'/%3e%3ccircle r='12' fill='%23cd2e3a'/%3e%3cpath fill='%230047a0' d='M0-12A6 6 0 0 0 0 0a6 6 0 0 1 0 12 12 12 0 0 1 0-24z'/%3e%3c/g%3e%3cg transform='rotate(-123.69)'%3e%3cuse xlink:href='%23b'/%3e%3cpath stroke='%23fff' d='M0-23.5v3M0 17v3.5m0 3v3'/%3e%3c/g%3e%3c/svg%3e)
