Palmitoylation controls DLK localization, interactions and activity to ensure effective axonal injury signaling

Significance

Dual leucine-zipper kinase (DLK) is essential for responses to nerve injury and subsequent neural regeneration by controlling transfer of signals from damaged distal axons to neuronal cell bodies. However, DLK is predicted to be freely diffusible, raising the question of how it conveys long-distance, directional signals. Here we report that direct modification of DLK with the lipid palmitate, a process called palmitoylation, is critical for retrograde injury signaling. At the molecular level, palmitoylation targets DLK to retrograde trafficking vesicles, assembles DLK-dependent signaling complexes, and is also essential for DLK kinase activity. This “multifunctional” palmitoylation explains how DLK mediates injury responses and may be a previously unappreciated mechanism that ensures the specificity of enzymatic signaling in diverse cell types.

Abstract

Dual leucine-zipper kinase (DLK) is critical for axon-to-soma retrograde signaling following nerve injury. However, it is unknown how DLK, a predicted soluble kinase, conveys long-distance signals and why homologous kinases cannot compensate for loss of DLK. Here, we report that DLK, but not homologous kinases, is palmitoylated at a conserved site adjacent to its kinase domain. Using short-hairpin RNA knockdown/rescue, we find that palmitoylation is critical for DLK-dependent retrograde signaling in sensory axons. This functional importance is because of three novel cellular and molecular roles of palmitoylation, which targets DLK to trafficking vesicles, is required to assemble DLK signaling complexes and, unexpectedly, is essential for DLK’s kinase activity. By simultaneously controlling DLK localization, interactions, and activity, palmitoylation ensures that only vesicle-bound DLK is active in neurons. These findings explain how DLK specifically mediates nerve injury responses and reveal a novel cellular mechanism that ensures the specificity of neuronal kinase signaling.

Peripheral nervous system (PNS) axons project long distances to their target tissues, presenting a great challenge for the relay of retrograde signals from distal locations back to neuronal cell bodies. One important requirement for axonal retrograde signaling is to activate transcription following distal nerve injury, a response that is critical for PNS axon regeneration (1, 2). One key mediator of nerve injury responses is dual leucine-zipper kinase (DLK), an evolutionarily conserved upstream activator (a MAP3K) of the mitogen-activated protein kinases (MAPKs) c-Jun N-terminal kinase (JNK) and p38 MAPK (3–6). DLK’s importance in retrograde signaling is inferred from in vivo studies, where distal axonal injury triggers DLK-dependent responses in sensory and motor neuron cell bodies (7, 8). Direct evidence for DLK’s role in retrograde signaling comes from studies using compartmented chambers, which physically isolate neuronal cell bodies from distal axons. Selective activation of DLK signaling in distal axons results in phosphorylation of the JNK transcription factor substrate c-Jun in cell bodies (9). However, it is unclear how DLK, a predicted soluble, diffusible protein, transfers signals directionally over long distances.

It is also puzzling why genetic loss or pharmacological inhibition of DLK (MAP3K12) causes such striking neuronal phenotypes (5, 7, 9, 10) that are not compensated for by the large family of homologous MAP3Ks [23 enzymes in mammals (11)]. Many of these DLK homologs are also expressed in peripheral neurons (12) and can activate JNK and p38 signaling in vitro and in transfected cells (13). These findings suggest that differential subcellular localization and regulation underlies DLK’s unique role in nerve injury signaling, but the nature of any such DLK-specific regulation is unclear.

Interestingly, although bioinformatically predicted to be soluble, DLK localizes to discrete axonal puncta across animal species, including worms, flies, and mice (8, 14, 15). We reasoned that an evolutionarily conserved mechanism might account for DLK’s unexpected subcellular localization and thus explain what makes DLK unique among MAP3Ks.

DLK axonal puncta resemble lipid vesicles, raising the possibility that DLK associates with vesicle membranes. One way to target soluble proteins to membranes is by protein–lipid modification. Four such modifications are known for intracellular proteins: palmitoylation, myristoylation, farnesylation, and geranylgeranylation, of which palmitoylation is most frequently observed for neuronal proteins (16, 17). Palmitoylation is best known to target proteins to the plasma membrane, but can also target proteins to vesicles (18–20). However, roles for palmitoylation in axonal retrograde trafficking and signaling have not been described.

Here we report that palmitoylation at a conserved cysteine residue is critical for DLK-dependent retrograde injury signaling. Mechanistically, palmitoylation targets DLK to motile trafficking vesicles and thus provides a way for DLK to traffic retrogradely. In addition, palmitoylation is essential for DLK to bind JNK-interacting protein-3 [JIP3, which couples JNK pathway proteins to microtubule retrograde motors (9, 21)], and to bind the MAP2Ks MKK4 and MKK7, DLK’s direct downstream targets in the JNK pathway (11, 22, 23). Strikingly, direct palmitoylation is also essential for DLK kinase activity; preventing DLK palmitoylation eliminates DLK-mediated kinase signaling in heterologous cells and even in vitro. These findings suggest that active, palmitoylated DLK is restricted to trafficking vesicles in neurons, providing an explanation for how DLK-dependent signals are conveyed without affecting basal, physiological JNK activity (9). In addition to providing new insights into axonal DLK signaling, these results reveal palmitoylation as a multifunctional modification that can control the localization, interactions, and activity of signaling kinases. This finding has broad implications for our understanding of how neurons convey specific signals over long times and distances.

Results

DLK Is Palmitoylated at an Evolutionarily Conserved Site.

To determine the basis of DLK’s reported vesicle-like axonal distribution, we examined DLK’s sequence for lipid binding motifs. DLK contains no predicted lipid binding domains and lacks consensus sequences for myristoylation, farnesylation, or geranylgeranylation. However, we bioinformatically identified a predicted palmitoylation site [C127 in mammalian DLK (24)] that is highly conserved across animal species (Fig. 1A).

Fig. 1.

DLK is palmitoylated at a conserved site. (A) A predicted palmitoyl-motif (yellow cysteine) is conserved in DLK orthologs from the indicated species. (B) DLK is specifically palmitoylated at C127. HEK293T cells were transfected with GFP-tagged wtDLK (wtDLK-GFP) or DLK-C127S point mutant (DLK-CS-GFP) cDNAs. Palmitoyl-proteins were isolated by ABE. Palmitoyl-DLK is robustly detected in ABE fractions (Upper), but no palmitoyl-DLK signal is seen from cells treated with 2-Br, or if Cys-127 is mutated to Ser. DLK protein levels were similar in all conditions (Lower). (C) ABE assays confirm endogenous DLK palmitoylation in rat forebrain (Upper) and cultured sensory neurons (Lower).

To determine whether DLK is palmitoylated at C127, we expressed GFP-tagged wild-type DLK (wtDLK-GFP) in HEK293T cells and used acyl biotin exchange (ABE), a nonradioactive method, to purify palmitoyl-proteins from lysates (16). WtDLK-GFP was robustly detected in ABE fractions but was absent from controls in which the essential ABE reagent hydroxylamine (NH₂OH) was omitted (Fig. 1B). In contrast, no ABE signal was seen if cells were pretreated with the palmitoylation inhibitor 2-Bromopalmitate (2-Br) (25) or if Cys-127 was mutated to nonpalmitoylatable Ser (DLK-CS-GFP) (Fig. 1B). In addition, cotransfection of a subset of DHHC family palmitoyl acyltransferases (PATs) markedly increased levels of wtDLK-GFP, but not DLK-CS-GFP, in ABE fractions (Fig. S1 A–C). DLK palmitoylation at C127 was also detected using a complementary method, metabolic labeling with the palmitate analog 17-octadecynoic acid (17-ODYA), and subsequent coupling to a fluorescent azide reporter tag (Fig. S1D) (26, 27).

Fig. S1.

Further evidence that DLK is palmitoylated at C127. (A) HEK 293T cells were transfected with wtDLK-GFP plus either control vector or one of the 23 mammalian DHHC-PATs, as indicated. ABE fractions were blotted to detect palmitoyl-DLK (Upper). Cell lysates were blotted to detect total DLK expression and HA-tagged DHHC-PAT expression as indicated. Note that DHHC-PATs are numbered as originally classified by Fukata et al., as explained in ref. 16. DHHC10 = ZDHHC11; DHHC11 = ZDHHC23; DHHC22 = ZDHHC13; DHHC23 = ZDHHC25. (B) Histogram shows palmitoyl:total DLK levels (mean ± SEM) from n = 2–5 experiments following cotransfection of the indicated PATs. (C) Follow-up ABE experiments from HEK293T cells cotransfected with the indicated PATs plus either wtDLK-GFP or DLK-CS-GFP confirmed that DLK palmitoylation by all PATs examined occurs predominantly at C127. (D) HEK 293T cells expressing the indicated GFP-tagged cDNAs were metabolically labeled with palmitate, or with the alkyne-containing palmitate analog 17-ODYA and lysed. GFP immunoprecipitates were subjected to Click chemistry using IRDye 800CW azide, denatured, subjected to SDS/PAGE gels and transferred to PVDF membranes. (Upper) Palmitoylated DLK, visualized by infrared detection of Click-reacted 17-ODYA. (Lower) Western blots of parallel samples using anti-GFP antibody.

Palmitoyl-DLK was also robustly detected in ABE fractions from rat forebrain and from cultured sensory neurons, in which DLK-dependent retrograde signaling is well described (9) (Fig. 1C). Taken together, these data suggest that DLK is palmitoylated in heterologous cells and in neurons. In contrast, homologous MAP3Ks that are expressed in sensory neurons (12) were not palmitoylated (Fig. S2). In these latter ABE assays, palmitoyl:total signals for DLK were very similar to palmitoyl:total signals of the known palmitoyl-protein PSD-95, suggesting that DLK is a bona fide palmitoyl-protein.

Fig. S2.

DLK is palmitoylated but closely related MAP3Ks are not. (A) Schematic of DLK/MLK family MAP3Ks. MAP3K9-11 (MLK1-3, respectively) kinase domains and leucine zippers are similar to DLK (MAP3K12) but MAP3K9-11 lack the consensus palmitoylation motif. (B) C-terminally GFP-tagged forms of DLK (DLK-GFP), the indicated MAP3Ks, or the known palmitoyl-protein PSD-95, were expressed in HEK293T cells. DLK and PSD-95 are detected in ABE fractions, indicative of palmitoylation, but MAP3K9-11 are not (Upper). Expression of all MAP3Ks and of PSD-95 was broadly similar (Lower). Histogram shows quantified palmitoyl-MAP3K: total MAP3K ratio, plotted relative to DLK from n = 3 experiments.

Palmitoylation Targets DLK to Axonal Trafficking Vesicles.

We next addressed whether palmitoylation accounts for DLK’s previously reported discrete localization in axons. Lentivirally expressed wtDLK-GFP localized to discrete axonal puncta in mammalian cultured sensory neurons (Fig. 2 A and B), consistent with reports from other neuron types (8, 14, 15). In contrast, only diffuse GFP fluorescent signals were seen in wtDLK-GFP–expressing neurons treated with 2-Br, or in neurons expressing DLK-CS-GFP (Fig. 2 A and B). These results suggest that palmitoylation controls DLK distribution within axons.

Fig. 2.

Palmitoylation targets DLK to motile axonal trafficking vesicles. (A) Palmitoylation-dependent localization of DLK to axonal puncta. Sensory neurons cultured in microfluidic chambers were infected with lentiviruses expressing either wtDLK-GFP or DLK-CS-GFP, plus the morphology marker mCherry. Cultures were treated with or without 2-Br, fixed, and immunostained with the indicated antibodies. (B) Quantified data confirm that DLK punctate distribution is diffused by 2-Br treatment, or by C127S mutation. Data are mean + SEM for n = 6–11 determinations per condition, ***P < 0.001 vs. wtDLK-GFP condition, ANOVA with Tukey post hoc test. (C) A GFP fusion of the C. elegans DLK ortholog DLK-1 (GFP-wt-CeDLK-1) localizes to axonal puncta (arrowheads), but the homologous point mutant (GFP-CeDLK-1-C104S) is diffuse. (D) Axonal DLK-GFP puncta are motile. Cultured sensory neurons were infected with lentivirus expressing wtDLK-GFP. GFP-fluorescent images were acquired at the indicated times. (E) Cultured sensory neurons were coinfected with lentiviruses expressing wtDLK-GFP and the retrograde trafficking vesicle marker RFP-Rab7. Representative kymographs obtained from live imaging of individual axons reveal DLK puncta that comigrate with Rab7 (arrow).

DLK’s palmitoyl-motif is conserved in the Caenorhabditis elegans ortholog DLK-1 (CeDLK-1) (Fig. 1A). Consistent with findings from mammalian neurons, GFP-wt-CeDLK-1 was punctate in C. elegans axons but the homologous Cys→Ser point mutant (GFP-CeDLK-1-C104S) was diffuse (Fig. 2C). Furthermore, overexpression of wt-CeDLK-1 in C. elegans mechanosensory neurons induced extra axonal branching. In contrast, overexpression of CeDLK-1-C104S did not show detectable effects, supporting that palmitoylation at C104 is important for in vivo activity of DLK-1 (Fig. S3). These results suggest that palmitoylation is an evolutionarily conserved mechanism that controls DLK’s axonal localization and function.

Fig. S3.

DLK-1 palmitoylation is functionally important in C. elegans sensory neurons. (A) Representative images showing axons of posterior lateral microtubule mechanosensory neurons visualized by GFP. WtCeDLK-1 or CeDLK-1-CS was overexpressed under the control of a mechanosensory neuron-specific promoter using an extrachromosomal array. Axonal branches marked with black arrowheads reached their target in the ventral cord but those marked with white arrowheads did not. (Scale bar, 20 µm.) (B) The number of extra branches of posterior lateral microtubule neurons was scored along the entire axon. Number of worms analyzed is shown. Statistics: one-way ANOVA; ***P < 0.001.

To define the DLK axonal puncta, we performed time-lapse imaging in sensory axons expressing wtDLK-GFP. A subset of DLK puncta was motile, moving both anterogradely and retrogradely in axons (Fig. 2 D and E). Dual-color imaging revealed that a subset of motile wtDLK-GFP puncta comigrate with mCherry-tagged Rab7 (mCh-Rab7, a marker of retrograde trafficking vesicles) (28) (Fig. 2E). Moreover, a subset of DLK-GFP puncta comigrated with NPY-mCh (a marker of Golgi-derived vesicles) (29) and also with Lysotracker Red, a marker of acidic vesicles (Fig. S4). Taken together, these findings suggest that axonal puncta of palmitoyl-DLK are motile trafficking vesicles.

Fig. S4.

Palmitoylation targets DLK to axonal NPY-mCherry and Lysotracker Red-positive vesicles. (Left) Cultured sensory neurons were coinfected with lentiviruses to express wtDLK-GFP and mCherry-tagged NPY signal sequence (NPY-mCh), a marker of Golgi-derived vesicles. Representative kymographs obtained from live imaging of individual axons reveal that a subset of DLK puncta comigrate with NPY-mCh⁺ vesicles. (Right) Cultured sensory neurons were coinfected with lentivirus to express wtDLK-GFP and then incubated with Lysotracker Red, a marker of acidic vesicles. Representative kymographs obtained from live imaging of individual axons reveal that a subset of DLK puncta comigrate with Lysotracker Red-positive vesicles.

Direct Palmitoylation Is Required for DLK-Dependent Axonal Retrograde Signaling.

This striking effect of palmitoylation on DLK’s localization in axons led us to examine whether this modification is required for DLK-dependent axonal retrograde signaling. To address this issue, we first used lentiviral short-hairpin RNA (shRNA) knockdown/rescue to replace endogenous DLK with shRNA-resistant wtDLK or DLK-CS (Fig. S5 A and B). We then used these lentiviruses to infect cultured sensory neurons in microfluidic chambers, which provide isolated environments for cell bodies and distal axons (Fig. S5C). We selectively axotomized distal axons of microfluidic cultures and examined subsequent retrograde signaling to cell bodies (30). Distal axotomy dramatically increased phosphorylation of the DLK-JNK pathway transcription factor target c-Jun in cell bodies of neurons in microfluidic culture (Fig. 3 and Figs. S5D and S6). Axotomy-induced c-Jun phosphorylation was almost completely abolished in DLK “knockdown” neurons, and was rescued by shr-DLKwt, but not by shr-DLK-CS (Fig. 3). These findings are consistent with reports that DLK is essential for retrograde signaling (7, 9) but, unexpectedly, reveal that only palmitoylated DLK can perform this role.

Fig. S5.

A combined shRNA knockdown/rescue and microfluidic approach to determine the importance of palmitoyl-DLK for retrograde injury signaling. (A) Schematic of shRNA knockdown/rescue experimental design. One lentivirus expresses GFP (infectivity marker) with or without DLK shRNA. A second virus expresses shRNA-resistant wtDLK or DLK-CS. (B) Effective lentiviral knockdown/rescue of DLK in sensory neurons. Cultured sensory neurons were infected with the indicated viruses. Western blotting of lysates confirms that DLK protein levels are greatly reduced by DLK shRNA and restored to endogenous levels by rescue constructs. (C) Overview of microfluidic chambers used to study axonal retrograde signaling. (Left) Microfluidic device. Colored dyes added to distal axon and cell body compartments confirm fluidic isolation. Enlarged fluorescent image of boxed area from left image confirms that GFP-expressing lentivirus infects >95% of neurons. Note cell bodies in central chamber, narrow microgrooves and axons in distal chamber. (D) Axotomy triggers retrograde signals that induce robust c-Jun phosphorylation. Sensory neurons from control (Left) or axotomized (Right) microfluidic cultures were immunostained to detect phospho-cJun (pcJun) and Neurofilament. pcJun signals overlap with NeuN (neuronal marker) and DAPI (nuclear marker) (see also Fig. S6).

Fig. 3.

Retrograde signaling following axotomy requires palmitoyl-DLK. (A) Sensory neurons cultured in microfluidic chambers were infected with the indicated lentiviruses, axotomized, and immunostained with the indicated antibodies. Images of axotomized and uninjured sister cultures for all infection conditions are shown in Fig. S5. (B) Quantified data reveal that axotomy-induced c-Jun phosphorylation is strongly DLK dependent and is rescued by shRNA-resistant wtDLK. DLK-CS fails to rescue. ***P < 0.001 versus uninfected, axotomized condition; ns, not significantly different from uninfected axotomized condition, ANOVA with Tukey post hoc test, n = 3–5 determinations per condition.

Fig. S6.

Magnified images of the indicated fluorescent markers for all conditions in Fig. 3C. Examination of these images confirms that phospho-c-Jun-positive signals overlap with NeuN (neuronal nuclear marker) and DAPI (nuclear marker).

Palmitoylation Is Critical to Assemble DLK/JNK Pathway Signaling Complexes.

We next sought to define how palmitoylation regulates DLK at the molecular level. In particular, we were struck by the similarity of DLK’s axonal vesicular localization to that previously described for the scaffold protein JIP3, which binds DLK and is essential for DLK-dependent retrograde signaling (9, 21). Indeed, wtDLK-GFP clearly colocalized and cotrafficked with mCherry-tagged JIP3 (mCh-JIP3) in axons (Fig. 4A). However, given DLK’s ability to bind and cotraffic with JIP3, it was striking that the DLK-CS-GFP mutant was diffusely localized in axons (Fig. 2A). This finding suggested that DLK might need to be palmitoylated to bind JIP3. Consistent with this notion, wtDLK strongly bound JIP3 in cotransfected cells, as previously reported (9), but C127S mutation completely disrupted DLK-JIP3 binding (Fig. 4B). Interestingly, C127S mutation did not affect DLK-DLK homodimerization (31) (Fig. S7A). Taken together, these findings suggest that palmitoylation not only targets DLK to vesicles, but also facilitates specific protein–protein interactions. We note that JIP3 still localized to motile, vesicle-like axonal structures in DLK knockdown neurons (Fig. S7B), perhaps because of JIP3’s known interactions with other axonal cargoes (32).

Fig. 4.

Palmitoylation is essential for DLK to bind MKK4 and the scaffold protein JIP3. (A) Cultured sensory neurons were coinfected with lentiviruses expressing wtDLK-GFP and mCherry-tagged JIP3 (mCh-JIP3). Representative kymographs reveal DLK-positive vesicles that comigrate with JIP3 (arrowed example). (B) HEK293T cells were transfected with the indicated constructs. Blots of GFP immunoprecipitates (IPs) reveal that C127S mutation disrupts binding of DLK to both JIP3 and MKK4.

Fig. S7.

C127S mutation does not affect DLK–DLK dimerization, but disrupts DLK–MKK7 interactions. (A) HEK293T cells were transfected with the indicated tagged DLK constructs. Blots of myc immunoprecipitates (IPs) confirm that C127S mutation does not affect dimerization of DLK-GFP with DLK-myc, suggesting that this mutation does not grossly affect DLK structure. (B) DLK knockdown does not prevent JIP3 punctate localization or motility in axons. Cultured sensory neurons were infected with lentivirus expressing GFP, with or without DLK shRNA (as in Fig. S5A), plus a second virus expressing mCh-JIP3. Representative dual-color time lapse fluorescent images confirm that DLK knockdown does not affect distribution or motility of JIP3. (C) HEK293T cells were transfected with the indicated constructs. Blots of GFP immunoprecipitates (IPs) confirm that C127S mutation greatly reduces binding of MKK7 to DLK.

We next addressed whether palmitoylation is also necessary for additional interactions of DLK, focusing on the MAP2K’s MKK4 and MKK7, which are likely DLK’s major cellular substrates (11, 22, 23, 33). In cotransfected cells, wtDLK bound MKK4, and this interaction was slightly enhanced in the presence of JIP3. In contrast, DLK-C127S bound MKK4 far more weakly than did wtDLK (Fig. 4B). In parallel assays, DLK-C127S also bound MKK7 more weakly than did wtDLK (Fig. S7C). Focusing on MKK4, we observed clear cotrafficking of mCh-MKK4 with wild-type (i.e., palmitoylated) DLK-GFP on axonal vesicles and also observed cotrafficking of GFP-MKK4 with mCh-JIP3 (Fig. S8 A and B). Taken together, these findings suggest that palmitoylation at C127 is essential for DLK to form multiprotein complexes in axons that contain both JIP3 and downstream MAP2Ks. Interestingly, when we examined the turnover of DLK palmitoylation by treating neurons with 2-Br, we observed that >50% of palmitoyl-DLK turns over within 1 h of 2-Br treatment (Fig. S8 C and D). However, the remaining pool of palmitoyl-DLK was still stable after 4 h of 2-Br treatment, sufficient time to mediate trafficking of DLK-JIP3-MKK complexes over considerable distances. We also found that both palmitoyl and total DLK levels were slightly reduced following in vitro axotomy (Fig. S8E), but again note that the pool of palmitoyl-DLK that remains postaxotomy could be critical for retrograde injury signaling.

Fig. S8.

Evidence for motile axonal complexes containing wild-type (palmitoylated) DLK, JIP3, and MKK4 and further characterization of neuronal DLK palmitoylation. (A) Cultured sensory neurons were coinfected with lentiviruses expressing wtDLK-GFP and mCherry-tagged MKK4 (mCh-MKK4). Representative kymographs reveal DLK-positive vesicles that comigrate with MKK4 (arrow). (B) As in A, except that neurons expressed mCh-JIP3 and GFP-MKK4. Note that although MKK4 puncta were readily detectable in A and B, irrespective of the fluorescent tag, the diffuse axonal signal for MKK4 was reproducibly higher than that of either wt-DLK or JIP3. A plausible explanation for this observation is that additional pools of MKK4 are involved in signaling via other cytosolic MAP3Ks. (C) Stable and rapidly regulated pools of palmitoyl-DLK in sensory neurons. Sensory neurons were treated with 2-Br for the indicated times and lysates subjected to ABE to detect palmitoyl-DLK (Upper Left) or the known palmitoyl-protein GAP43 (Upper Right). 2-Br did not affect total expression of either protein (Lower). (D) Histogram of pooled data from n = 3 determinations per condition from C confirms that palmitate turnover is rapid for the majority of cellular DLK, but a stably palmitoylated pool of DLK is also detectable. (E) Clear palmitoyl-DLK signal remains postinjury. Cultured sensory neurons were left untreated or subjected to axotomy and lysates were subsequently subjected to ABE. ABE fractions were blotted to detect palmitoyl-DLK and total lysates were blotted to detect DLK and tubulin. Similar results were obtained in three other experiments.

Palmitoylation Is Critical for DLK to Activate the JNK Pathway in Transfected Cells and in Vitro.

Importantly, MKK4 and MKK7 not only bind DLK, but are also direct substrates that are critical for DLK-dependent signaling (11, 22, 23). Our observation that DLK must be palmitoylated to recognize these MKKs (Fig. 4B and Fig. S7C) and the proximity of DLK’s palmitoylation site to its kinase domain suggested that DLK-MKK4/7-JNK signaling per se might be palmitoylation-dependent. To address this possibility, we first assessed whether the palmitoylation inhibitor 2-Br affects signaling via DLK. Consistent with a previous report (3), wtDLK markedly increased phosphorylation of cotransfected JNK3 in HEK293T cells (Fig. 5 A–C). Strikingly, 2-Br treatment abolished JNK3 activation by DLK, without affecting JNK activation by the nonpalmitoylated MAP3Ks MLK1-3 (Fig. 5 A–C). The effect of 2-Br was phenocopied by C127S mutation, which prevented DLK-dependent phosphorylation of both JNK3 and of DLK’s direct substrates MKK4 and MKK7 (Fig. 5 D and E and Fig. S9).

Fig. 5.

DLK must be palmitoylated to activate the JNK pathway. (A) Scheme of MKK4-JNK activation by DLK and related MAP3Ks. (B) HEK293T cells expressing myc-JNK3 plus wtDLK-GFP or related GFP-tagged MAP3Ks were treated with 2-Br or vehicle. Lysates were blotted to detect phospho-JNK (pJNK, Top), JNK3 (Middle), and MAP3K expression (Bottom). (C) Quantitation of multiple experiments from B confirms that 2-Br specifically prevents JNK3 phosphorylation induced by DLK. *P < 0.05; ns: not significantly different, t test, n = 3 determinations per condition. (D) C127S mutation prevents DLK-dependent JNK3 phosphorylation. Lysates of HEK293T cells expressing myc-JNK3 plus either wtDLK-GFP or DLK-CS-GFP were blotted as in B. Histogram shows pJNK:total JNK3 levels (mean ± SEM, relative to wtDLK-GFP). ***P < 0.001 vs. wtDLK-GFP, ANOVA, Tukey post hoc test, n = 3. (E) C127S mutation prevents DLK-dependent MKK4 phosphorylation. Lysates of HEK293T cells expressing myc-MKK4 plus either wtDLK-GFP or DLK-CS-GFP were blotted with the indicated antibodies. Histogram shows pMKK4:total MKK4 (mean ± SEM, relative to wtDLK-GFP). ***P < 0.001 vs. wtDLK-GFP, ANOVA, Tukey post hoc test, n = 3. (F) Palmitoyl-DLK is more active in vitro. HEK293T cells expressing wtDLK-GFP or vector were treated with 2-Br or vehicle as indicated. GFP immunoprecipitates were assayed for ability to phosphorylate GST-MKK4 in vitro. Samples were blotted and histogram plotted as in E. (G) As in F, except cells expressed wtDLK-GFP or DLK-CS-GFP.

Fig. S9.

Evidence that DLK must be also palmitoylated to phosphorylate MKK7 at activatory sites, unexpected gain-of-function effects of heterologous lipid attachment to DLK and schematic of the role of DLK palmitoylation. (A, Left) HEK293T cells were transfected to express myc-MKK7 plus either wtDLK-GFP or DLK-CS-GFP. Lysates were immunoblotted to detect phosphorylated MKK7 (pMKK7, Top), total MKK7 expression (Middle), and DLK expression (Bottom). (Right) Histogram of pMKK7:total MKK7 levels (mean + SEM, relative to wtDLK-GFP condition). ***P < 0.001 vs. wtDLK-GFP condition, ANOVA with Tukey post hoc test, n = 3 determinations per condition. (B) Adding a heterologous membrane-targeting motif to DLK has unexpected gain-of-function effects. HEK293T cells were transfected with wtDLK-GFP or with wt- or C127S DLK mutants carrying an N-terminal myristoylation motif (myr-wtDLK-GFP, myr-DLK-CS-GFP). ABE fractions were blotted to detect palmitoyl-DLK and lysates blotted to detect total DLK expression. Both myristoyl mutants are palmitoylated to a greater extent than wtDLK-GFP. For myr-DLK-CS-GFP, this additional palmitoylation must occur at sites distinct from the physiological site (C127). (C) A model of the role of palmitoyl-DLK in axonal injury signaling. Palmitoylation targets DLK to axonal trafficking vesicles and assembles a complex with JIP3 (which binds microtubule motor proteins), MKK4/7, and possibly JNK itself. In addition, palmitoylation is essential for DLK kinase activity (Upper). This latter role acts as a security feature, because if DLK is depalmitoylated and dissociates from vesicles it concomitantly loses activity, thus preventing inappropriate activation of cytosolic MKKs (Lower).

These findings suggest that palmitoylation is required for DLK-dependent signaling, irrespective of its role in retrograde trafficking. However, even in nonneuronal cells these results could be because of differential localization of wt-DLK and DLK-CS or, alternatively, could be because DLK must be membrane-localized in cells for activity. We first attempted to distinguish between these possibilities by making a DLK palmitoyl-site mutant with an N-terminal sequence that directs addition of the lipid myristate (Myr-DLK-CS-GFP). We hypothesized that Myr-DLK-CS-GFP would still localize to membranes but that the heterologous lipid attachment would not phenocopy effects of C127 palmitoylation. However, we found that Myr-DLK-CS-GFP was palmitoylated at additional, nonphysiological sites, an unexpected gain-of-function effect that led us not to pursue experiments with this mutant (Fig. S9B).

We therefore adopted an alternative approach to rule out effects of cellular localization by assessing the palmitoylation-dependence of DLK activity in an in vitro kinase assay. Strikingly, we found that wtDLK was markedly more active toward GST-tagged MKK4 in vitro, compared with wtDLK from 2-Br–treated cells (Fig. 5F), or to DLK-CS (Fig. 5G). These results suggest that palmitoylation plays a third role in DLK-dependent signaling, by directly and autonomously regulating DLK kinase activity.

Discussion

Three Complementary Roles of Palmitoylation Govern DLK-Dependent Signaling.

DLK is critical for axonal injury responses (7, 8), but how DLK mediates axonal retrograde signaling has been unclear. Here we report that direct palmitoylation exerts three complementary effects on DLK at the cellular and molecular level, which together ensure the fidelity and specificity of DLK-dependent signals.

The first role of palmitoylation is to target DLK to axonal trafficking vesicles (Fig. 2). By allowing DLK to “hitchhike” on vesicles, palmitoylation may provide a way for DLK to transfer signals directionally and more rapidly than would occur by passive diffusion (34). Palmitoylation-dependent vesicular targeting of DLK appears highly conserved across animal species, as palmitoyl-site mutation transforms DLK distribution from punctate to diffuse in both mammalian and invertebrate axons (Fig. 2).

Second, palmitoylation is necessary for DLK to form multiprotein complexes with the scaffold protein JIP3 and the downstream kinases MKK4 and MKK7 (Fig. 4 and Figs. S7, S8, and S9C). Interestingly, JIP3, which is essential for DLK-dependent retrograde signaling (9), in turn binds microtubule motor proteins (21, 35). Palmitoylation of DLK may thus serve as a master regulatory event to assemble vesicular DLK/JIP3/MKK complexes and to ensure their motility in axons. It is an intriguing possibility that these palmitoyl-DLK complexes also contain the downstream kinase JNK3, which colocalizes with JIP3 on axonal vesicles (21).

Finally, a major finding of this study is that palmitoylation controls the kinase activity of DLK. This novel role for palmitoylation would appear to be independent of any effects on DLK localization in axons, because it is readily demonstrable in heterologous cells and even in in vitro assays (Fig. 5). However, we cannot exclude the possibility that the ability of palmitoylation to increase the local concentration of DLK on membranes also contributes to effects of this modification on DLK intracellular signaling and interactions.

How might palmitoylation exert these effects at the molecular level? DLK interactions with JIP3, MKK4, and MKK7 are all palmitoylation-dependent, but DLK–DLK dimerization is not (Fig. S7A). Hence, one possibility is that palmitoylation locally alters DLK conformation without grossly affecting DLK structure. This possibility is enhanced by the observation that DLK’s noncatalytic N terminus (residues 1–164), which contains the palmitoyl-site, is Gly- and Pro-rich, and thus predicted to lack extended secondary structure and be flexible (36–38). Moreover, this N-terminal region is highly acidic (pI = 4.58), whereas DLK’s kinase domain (residues 165–400) is basic (pI = 8.70). It would thus appear possible that DLK’s N-terminal region can loop back and inhibit its kinase domain and that such autoinihibition (a common feature of related MAP3Ks, albeit via different protein–protein interaction motifs) (13) might be sterically impaired by C127 palmitoylation. If this were the case then palmitoylation might act like a phosphorylation event, but in this case the functional moiety would be a hydrophobic lipid rather than a charged phosphate group. However, although intriguing, we appreciate that this is only one possible explanation and that further studies are required to define how palmitoylation controls the interactions and activity of DLK.

We also note that, despite its clear palmitoylation (Fig. 1 and Fig. S1), DLK was not identified in prior palmitoyl-proteomic studies. However, several bona fide palmitoyl-proteins identified by ourselves and others were also not found in such studies (18, 39, 40). One reason for the lack of identification of DLK is likely tissue source; most palmitoyl-proteomic studies were performed from nonneuronal tissues, whereas DLK is largely restricted to the nervous system (41, 42). A second issue is likely detection sensitivity; DLK protein levels in neurons are likely low, in part because DLK protein expression is tightly controlled by ubiquitin-dependent degradation (10, 43, 44). These observations suggest that additional palmitoyl-proteomic studies of neuronal tissues, focusing on low abundance palmitoyl-proteins, may identify other palmitoyl-kinases and other signaling enzymes.

Palmitoylation Regulates DLK but Not Homologous MAP3Ks.

If palmitoylation indeed accounts for DLK’s unique role in axonal retrograde signaling, then one would strongly predict that related MAP3Ks expressed in sensory neurons (i.e., MLK1-3/MAP3K9-11) are not palmitoylated. Our experiments confirm that this is indeed the case and also reveal that a palmitoylation inhibitor specifically blocks DLK-dependent signaling without affecting signaling by MLK1-3 (Fig. 5 and Fig. S2). Interestingly, the closely related MAP3K13 does contain a homologous palmitoylation site, but MAP3K13 is not detectably expressed in sensory neurons (12), likely explaining its failure to compensate for loss of DLK in this cell type. However, it will be of interest to determine whether MAP3K13 is also palmitoylated and, if so, whether MAP3K13 localization and activity is palmitoylation-dependent in other types of neuron.

Palmitoylation May Explain How DLK Mediates Injury Responses Without Affecting Overall JNK Activity.

An additional key feature of DLK is that it mediates axonal retrograde signaling via JNK and c-Jun without affecting basal levels of JNK activity (9). This observation has spurred interest in DLK as a therapeutic target (45), but how DLK plays this specific role has been unclear. Our findings suggest that the triple control of DLK localization, interactions, and activity by palmitoylation holds the key to this issue. This is because active palmitoyl-DLK is tightly associated with vesicles and thus is unlikely to phosphorylate cytosolic pools of MAP2Ks. Conversely, if DLK is depalmitoylated and dissociates from vesicles, it concomitantly loses kinase activity. This elegant security feature may thus minimize aberrant phosphorylation of other JNK pathway substrates (Fig. S9C). In situations where DLK signaling drives neurodegeneration (10), then strategies to inhibit DLK palmitoylation could be used in combination with, or as alternatives to, small molecules that directly target DLK’s kinase domain. Conversely, in situations in which enhanced DLK signaling might be beneficial (e.g., to increase regeneration of injured adult peripheral nerves), strategies to inhibit the thioesterase(s) that depalmitoylates DLK hold considerable promise, particularly given DLK’s high rate of palmitate turnover (Fig. S8C). Given their potential therapeutic importance, the identity of DLK PATs and thioesterases is an exciting area for future study.

Broader Roles for Palmitoylation-Dependent Retrograde Signaling and Regulation of Kinase Activity.

Finally, the effectiveness of palmitoylation in controlling DLK localization, interactions, and activity suggests that other retrograde signaling proteins or other kinases might use this mechanism. In support of the first possibility, key retrograde signaling proteins, including the transcription factor STAT3 and importins (1), were recently proteomically identified as palmitoyl-proteins (26, 46, 47). These findings suggest that palmitoylation may be critical for retrograde transport of multiple neuronal signaling proteins.

In addition, we recently identified another kinase whose activity in neurons is dependent upon its palmitoylation (48). This raises the possibility that “multifunctional” palmitoylation is broadly used to localize kinases and perhaps other signaling enzymes, and thereby restrict their interactions and activity to specific membrane locations. Multifunctional palmitoylation may thus be critical for the spatial control of kinase signaling across an array of cell types and species.

Methods

Antibodies and cDNAs used in this study are listed in SI Methods. Protocols for well-established procedures (immunostaining and immunoprecipitation) and protocols described in prior reports are also fully documented in SI Methods.

Chemicals.

2-Bromopalmitate and S-Methyl methanethiosulfonate (MMTS) were from Sigma. All other chemicals were from Fisher Biosciences and were of the highest reagent grade.

Transfection.

HEK293T cells were transfected using a calcium phosphate-based method, as described previously (18).

Detection of Palmitoylation by ABE Assay and Metabolic Labeling.

ABE was performed as described previously (18) and metabolic labeling using 17-ODYA was performed similarly to prior studies (26, 27). Full details of both assays are provided in SI Methods.

Cultured Sensory Neurons.

All animal use protocols were approved by Temple University Institutional Animal Care and Use Committee. Dorsal root ganglia were dissected from embryonic day 14.5 rat embryos, digested with 5 mg/mL dispase, and plated on poly-l-lysine/laminin–coated glass coverslips or 12-well dishes. Neurons were cultured in Neurobasal media containing B27, GlutaGro (Mediatech), and 25 ng/mL NGF (BD). Fluorodeoxyuridine (FDU) was added on day in vitro 0 to prevent nonneuronal cell growth.

Microfluidic Cultures.

Microfluidic chambers were designed using AutoCAD software, based on a prior description (49). Master molds (fabricated at Stanford University Microfluidics Foundry and Missouri State University Jordan Valley Innovation Center) consisted of three chambers (width of middle chamber: 1,000 μm; width of adjacent chambers: 1,500 μm) connected by 450-μm-long microgrooves. PDMS devices were cast using Sylgard 184 (Dow Corning). Distal axons in microfluidic chambers were axotomized as described (30, 49). Axotomy of conventionally cultured neurons is described in SI Methods.

Lentiviral Infection and shRNA Knockdown.

Lentivirus was generated as described previously (18). Full details are in SI Methods.

Microscopy.

DLK distribution was imaged using a Nikon C2 inverted confocal microscope with an oil immersion objective (40×, 1.3 NA). Maximum intensity projections were generated from z-slices (0.25-µm spacing, 1,024 × 1,024 pixel resolution) using NIS Elements software. c-Jun phosphorylation was imaged using a Nikon 80i microscope (10×, 0.3 NA objective). Live images of DLK-GFP were acquired at 0.375-s intervals using a Zeiss Axio Observer Z1 inverted TIRF microscope with dual wavelength laser system, α-plan-apo TIRF objective (100× 1.46 NA) and Orca R2 camera (Hamamatsu). Images were analyzed using either Zeiss Axiovision software or ImageJ.

C. elegans Strains.

C. elegans strains were maintained on NGM plates at 20° C, as described previously (50). Pmec-4-GFP::CeDLK-1(wt) (pCZGY2335) and Pmec-4-GFP::CeDLK-1L(C104S) (pCZGY2336) were made using Gateway cloning (Invitrogen). Details are provided in SI Methods. Fluorescent images were collected from young adult animals using a Zeiss LSM710 confocal microscope with a 63× objective (NA = 1.4). Images are maximum-intensity projections from three z-stacks separated by 0.5 µm.

In Vitro Kinase Assay.

HEK293T cells transfected with DLK-GFP cDNAs were lysed in immunoprecipitation buffer (IPB) (18). After centrifugation, soluble supernatants were immunoprecipitated with anti-GFP antibody precoupled to protein G Sepharose. Beads were washed extensively with IPB and then with kinase assay buffer (10 mM Tris, pH 7.5, 0.2 mM EDTA, 0.1% Tx-100, protease and phosphatase inhibitors). Supernatant was removed and beads resuspended in kinase assay buffer containing 1 μM purified mouse GST-MKK4 and incubated for 10 min with 0.1 mM ATP/10 mM MgCl₂ at 30 °C. Assays were stopped with SDS sample buffer, processed for SDS/PAGE and immunoblotted with anti-GST, anti-GFP, anti-pMKK4, or anti-pMKK7 antibodies. GST fusion protein purification is described in SI Methods.

SI Methods

Antibodies and Molecular Biology.

The following antibodies, from the indicated sources, were used: anti-MAP3K12 (DLK; Sigma-Prestige), anti-HA11 16B12, anti-c-myc 9E10 (Covance); anti-dsRed (Clontech); anti–p-c-Jun (Ser63), anti-phospho-JNK (#9251), antiphospho-MKK4 (Ser257/Thr261), anti-MKK4, anti-phospho-MKK7 (Ser271/Thr275), anti-MKK7 (Cell Signaling); anti-2H3 neurofilament (Developmental Studies Hybridoma Bank); anti-NeuN A60 (Millipore); anti-GFP 3E6 (Life Technologies), anti-GFP N86_8 (Neuro mAb); anti-tubulin (Sigma); anti-GST (Bethyl Laboratories).

Mouse DLK cDNA was generated by PCR from a commercially available template (BC57572.1, Thermoscientific) and subcloned into a modified FUW lentiviral vector (18) and eGFP-N vector (Clontech), in frame with a C-terminal myc or GFP tag, respectively. C127S point mutants of both DLK cDNAs were generated by Quickchange mutagenesis (Stratagene). The ubiquitin promoter of FUW-DLK-myc, FUGW, and FUmChW cDNAs (48) were replaced with the EF1-α promoter, which ensures stronger and more specific expression in sensory neurons (51), generating FEW-DLK-myc, FEGW, and FEmChW vectors, respectively. An effective shRNA against DLK (9) was subcloned into pSuper vector and then amplified by PCR, together with its neighboring H1 promoter. The resulting fragment was subcloned into FEGW. DLK rescue constructs were generated by PCR using FEW-DLK-myc templates. shRNA resistance was confirmed in both cotransfected HEK293T cells and in neurons.

The N-terminal myc tag of mycFEW was replaced with an mCherry tag, generated by PCR, to make mCh-FEW vector. Full-length mouse JIP3 cDNA, generated by PCR from a commercially available template (Thermoscientific) was subcloned into mCh-FEW and also into the N-terminally HA-tagged mammalian expression vector HA-pRK5. Full-length mouse MKK7 cDNA generated by PCR from a commercially available template (Thermoscientific) was subcloned into myc-tagged pRK5. Full-length mouse MKK4 cDNA was gene synthesized (Genewiz) and subcloned into myc-tagged pRK5, and into FEW vectors downstream of N-terminal mCherry or GFP tags.

RFP-Rab7 cDNA was obtained from Addgene (plasmid #14436) and subcloned into FEW. NPY-mCherry cDNA was gene synthesized by Genewiz, based on a previous design of a similar NPY-GFP construct (52).

Human MAP3K9 cDNA was obtained as a C-terminal GFP fusion from ASU plasmid repository (DNASU). Human MAP3K10 and MAP3K11 cDNAs were from DNASU and were amplified by PCR. The resultant products were inserted into eGFP-N vector as C-terminal GFP fusions.

For Caenorhabditis elegans DLK-1 studies, cDNA of DLK-1 long isoform (F33E2.2a in WormBase) was cloned into pCR8 vector (Invitrogen) to make pCZGY1085 entry vector. The C104S mutation was introduced by a PCR-based method using the primers (YJ10091: CTTTTCGAGTTTTCAGCCAGTTTTTG; YJ10092: TGAAAACTCGAAAAGAGCCCATTCC) to make pCZGY2334 entry vector. pCZGY1085 and pCZGY2334 were then recombined with Pmec-4-GFP-GTW destination vector (pCZGY603) to generate pCZGY2335 and pCZGY2336 expression vectors, respectively. Transgenic worms were obtained by injecting pCZGY2335 or pCZGY2336 plasmids at 10 ng/µL with Pmec-4-tagRFP at 2.5 ng/µL and Pttx-3-mCherry at 90 ng/µL following standard procedure, as previously described (53).

Acyl Biotinyl Exchange Assay.

For ABE experiments, transfected HEK 293T cells or neurons were lysed directly in buffer [50 mM Hepes pH 7.0, 2% SDS, 1 mM EDTA plus protease inhibitor mixture (PIC, Roche)], and 20 mM methyl-methane thiosulfonate (MMTS, to block free thiols). Following lysis, excess MMTS was removed by three sequential acetone precipitations and pellets were resuspended in buffer containing 4% (wt/vol) SDS. Samples were diluted and incubated for 1 h in either 0.7 M hydroxylamine pH 7.4 (to cleave thioester bonds) or 50 mM Tris pH 7.4. Acetone precipitation was performed to remove hydroxylamine or Tris. Pellets were resuspended in 4% (wt/vol) SDS, diluted in 50 mM Tris pH 7.4 containing sulfhydryl-reactive (HPDP-) biotin and incubated for 1 h at room temperature. Acetone precipitation was performed to remove unreacted HPDP-biotin and pellets were resuspended in lysis buffer without MMTS. SDS was diluted to 0.1% (wt/vol) and biotinylated proteins in the samples were affinity-purified using neutravidin-conjugated beads. Beta-mercaptoethanol [1% (vol/vol)] was used to cleave HPDP-biotin and release purified proteins from the beads. The released proteins in the supernatant were denatured in SDS sample buffer and processed for SDS/PAGE.

Metabolic Labeling Combined with Click Chemistry.

HEK293T cells transfected as described (18) were deprived of fatty acids by incubation in DMEM supplemented with 5% (vol/vol) dextran-coated charcoal-treated FBS for 1 h. Cells were then metabolically labeled for 4 h in DMEM containing 1% fatty acid-free BSA, supplemented with either 100 μM 17-ODYA or 100 μM palmitate. Cells were washed with ice-cold PBS and lysed in IPB. Following centrifugation, lysates were incubated with protein G Sepharose that had been precoupled with anti-GFP antibody. Immunoprecipitates were washed extensively with IPB and then with PBS alone. Click reactions were performed in PBS containing 10 μM IRDye 800CW Azide, 5 mM ascorbate, 5 mM THPTA, and 1 mM CuSO₄. Reactions were allowed to proceed for 1 h at room temperature with regular flick-mixing. Samples were then denatured with SDS sample buffer, heated at 95 °C for 2 min and then immediately cooled, subjected to SDS/PAGE, and transferred to PVDF membranes. Parallel membranes were either directly visualized using an Odyssey scanner (LI-COR) or subjected to Western blotting with anti-GFP antibodies.

Lentiviral Infection and shRNA Knockdown.

HEK293T cells were cotransfected with FEGW or FEGW/shRNA plus VSV-G, pMDLg and RSV-Rev helper plasmids. Supernatant containing virus was harvested at 48 and 72 h posttransfection, concentrated by ultracentrifugation, resuspended in Neurobasal medium, and used to infect sensory neurons at day in vitro 1–7 for 5–7 d. Effective FEGW/shRNA virus titer was then determined by reduction of DLK expression (>90%) in sensory neurons. Absolute titers of FEGW control virus and FEGW/shRNA virus were determined by monitoring GFP expression in infected HEK293T cells and equal numbers of transducing units were used for experiments in neurons. Effective titers of DLK “rescue” viruses were determined by coinfecting sensory neurons with rescue virus plus FEGW/DLK shRNA virus and subsequent Western blotting of lysates to determine DLK levels. Amounts of rescue viruses that restored DLK expression to endogenous levels under these conditions were used for subsequent experiments. Lentiviruses to express fluorescently-tagged DLK, JIP3, and MKK4 were generated using a similar method.

Immunostaining.

Neurons were fixed in 4% (wt/vol) paraformaldehyde, 4% (wt/vol) sucrose in PBS. After PBS washes, neurons were permeabilized with PBS containing 0.25% (wt/vol) Triton X-100, blocked in 10% (vol/vol) normal goat serum (NGS) in PBS and incubated overnight at 4 °C with primary antibodies [in 10% (vol/vol) NGS] and then with AlexaFluor-conjugated secondary antibodies for 1 h at room temperature. Nuclei were stained with 300 nM DAPI in PBS for 10 min and coverslips mounted in Fluoromount G (Southern Biotech) before imaging.

Immunoprecipitation.

HEK293T cells were transfected with DLK cDNAs plus JIP3, MKK4, or MKK7 cDNAs. Cells were lysed in ice-cold 20 mM Tris pH 7.5, 150 mM NaCl 0.1% Triton X-100 plus protease and phosphatase inhibitors, as described previously (9). After centrifugation of lysates, supernatants were immunoprecipitated with anti-myc or anti-GFP antibodies that had been precoupled to protein G Sepharose (GE Healthcare). Beads were washed twice with lysis buffer. Proteins were eluted with SDS sample buffer and subjected to SDS/PAGE and subsequent immunoblotting.

Axotomy of Cultured Sensory Neurons.

Five-millimeter glass coverslips (#1 thickness) were placed in the center of a 12-well tissue culture plate that was subsequently coated with poly-l-lysine/laminin. Embryonic sensory neurons, dissected as described in Methods, were then plated onto the center of the coverslip. As the neurons develop, their axons extend from this central region, over the coverslip edge and onto the surface of the 12-well plate. After 11 d in vitro, the coverslip was picked up with fine forceps, the edge quickly wiped with a cotton swab to ensure that detached axons were severed, and was placed in a new 12-well plate containing 50% of the medium from the original culture well. After 90 min, medium was removed and cells processed for ABE as described above. Uninjured sister cultures remained in their original plates until medium was aspirated and cells were lysed.