GAP-43 (growth-associated protein-43) is a dually palmitoylated protein, at cysteine residues at positions 3 and 4, that mostly localizes in plasma membrane both in neural and non-neural cells. In the present study, we have examined membrane association, subcellular distribution and intracellular trafficking of GAP-43 in CHO (Chinese hamster ovary)-K1 cells. Using biochemical assays and confocal and video microscopy in living cells we demonstrated that GAP-43, at steady state, localizes at the recycling endosome in addition to the cytoplasmic leaflet of the plasma membrane and TGN (trans-Golgi network). Pharmacological inhibition of newly synthesized GAP-43 acylation or double mutation of Cys3 and Cys4 of GAP-43 completely disrupts TGN, plasma membrane and recycling endosome association. A combination of selective photobleaching techniques and time-lapse fluorescence microscopy reveals a dynamic association of GAP-43 with recycling endosomes in equilibrium with the plasma membrane pool. Newly synthesized GAP-43 is found mainly associated with the TGN, but not with the pericentriolar recycling endosome, and traffics to the plasma membrane by a brefeldin A-insensitive pathway. Impairment of plasma membrane fusion and internalization by treatment with tannic acid does affect the trafficking of GAP-43 from plasma membrane to recycling endosomes which reveals a vesicle-mediated retrograde trafficking of GAP-43. Here, we also show that internalization of GAP-43 is regulated by Arf (ADP-ribosylation factor) 6. Taken together, these results demonstrate that dual acylation is required for sorting of peripheral membrane-associated GAP-43 to recycling endosome via an Arf6-associated endocytic vesicular pathway.
- Golgi complex
- growth-associated protein-43 (GAP-43)
- membrane traffic
- palmitoyl-acyl transferase
- recycling endosome
S-acylated peripheral proteins such as the small G-proteins Ras, Gα subunit, and also the neuronal proteins post-synaptic density protein-95 and GAP-43 (growth-associated protein-43)  are synthesized in the cytosol and then post-translationally modified by different lipid moieties . The subcellular distribution of these acylated proteins, which occur by vesicular transport or reversible membrane association, depends on the number and types of lipids attached, on the physicochemical properties of aa (amino acids) surrounding the lipidation site, as well as on the presence of additional domains (i.e. protein–protein and protein–lipid interaction domains) .
GAP-43 is mainly localized at the PM (plasma membrane) and contributes to the mechanism of axonal outgrowth in embryo and to axonal regeneration in the adult. This protein was also identified in peripheral and central glia cells  and developing muscle cells , which points to a fundamental role for GAP-43 in cellular processes. GAP-43 is stably bound to membranes due to the S-acylation of Cys3 and Cys4 [5,6]. Also, this protein is acetylated at its initial methionine  which, by analogy with the isoprenyl cysteine carboxyl methylation of CAAX-containing proteins (e.g. Ras, Rab and Rho families), could operate by increasing the hydrophobicity of the N-terminal region of GAP-43 or in participating in protein–protein interactions . In addition, GAP-43 contains a basic effector domain organized in an α-helical secondary structure. This domain binds anionic lipids in both model and cellular systems [9,10], operates as a CaM (calmodulin)-binding region (aa 39–55) and contains a PKC (protein kinase C) phosphorylation site at Ser41 [11,12]. Phosphorylation of the effector domain disrupts its α-helical structure, prevents CaM/GAP-43 interaction and GAP-43-dependent clustering of PIP2 [phosphatidylinositol (4,5)-bisphosphate] at the PM [9,10,13,14].
GAP-43 is axonally transported in neurons by anterograde fast axonal transport . This observation supports the idea that stable membrane association of GAP-43 is a consequence of post-translational lipid modifications that occur early in the secretory pathway. In this regard, it was suggested that the ER (endoplasmic reticulum)–Golgi intermediate compartment is the initial site of GAP-43 acylation . However, recent results [17,18] suggest that acylation of this protein might occur at the Golgi level, in agreement with previous reports proposing this association and supporting the kinetic trap model for GAP-43 in such membranes .
In addition to the PM, the Golgi localization of GAP-43 has been widely documented [5,20–22]. We have also previously described the presence of a truncated form of GAP-43, containing the acylation motif (aa 1–13, N13GAP-43) fused to YFP (yellow fluorescent protein), in RE (recycling endosomes) . Nevertheless, the mechanisms of intracellular transport (vesicular and/or adsorption/desorption) involved in the targeting of GAP-43 to the RE are not understood.
In the present study, we have investigated, in CHO (Chinese hamster ovary)-K1 cells, the membrane association, subcellular distribution and intracellular trafficking of the GAP-43full (full-length wild-type GAP-43) and the acylation motif of GAP-43 both fused to the N-terminal domain of spectral variants of GFP (green fluorescent protein). Using biochemical assays, confocal and video microscopy in living cells, we demonstrated that GAP-43full and N13GAP-43 are mainly acylated at the TGN (trans-Golgi network). At steady state, these proteins were localized at the TGN, PM and RE. The TGN pool observed at steady state represented a newly synthesized fraction of GAP-43 en route to the PM, whereas the RE-associated pool of these proteins represented a post-biosynthetic fraction originating from endocytosis through a pathway regulated by the small GTPase Arf (ADP-ribosylation factor) 6. Taken together, these results demonstrate that exocytic and endocytic vesicular carriers mediate intracellular trafficking of GAP-43 in a double acylation-dependent process, and that the pool of GAP-43 associated to the endocytic recycling compartment is dynamic and exchangeable with the PM-associated pool.
The expression vectors pECFP-N1 [where ECFP, enhanced CFP (cyan fluorescence protein)], pEYFP-N1 (where EYFP, enhanced YFP), pECFP-C1 and pEYFP-C1 were from Clontech. Expression plasmids for N13GAP-43–(CFP or YFP), (CFP or YFP)–K-RasC14, (CFP or YFP)–H-RasC20, N27UDP-GalNAc-T (GalNAc:LacCer/GM3/GD3 N-acetylgalactosaminyltransferase)–CFP and N52UDP-Gal-T2 (Gal:GA2/GM2/GD2/GT2 galactosyltransferase)–CFP were described previously [23,24]. Mutations at Cys3 and Cys4 of N13GAP-43–YFP and GAP-43full were performed by directed mutagenesis using PCR. The plasmid encoding mouse GAP-43full has also been described previously  and was used for subcloning into the pEYFP-N1 vector. The GPI (glycosylphosphatidylinositol)–YFP fusion construct was kindly supplied by Patrick Keller (Max-Planck Institute, Dresden, Germany). Iip33–CFP was obtained by subcloning the Iip33 cDNA into pECFP-N1. The plasmid encoding GFP–PH-PLCδ1 (chimeric protein containing the PIP2-specific pleckstrin homology domain of phospholipase Cδ1 fused to GFP), HA (haemagglutinin)–Arf6-Q67L and pEGFP-Rab11a (where EGFP, enhanced GFP) wild-type were received from Mark Lemmon (University of Pennsylvania School of Medicine, Philadelphia, U.S.A.), Julie Donaldson (National Heart, Lung, and Blood Institute, National Institutes for Health, Bethesda, MD, USA) and Maria Colombo (Universidad Nacional de Cuyo, Mendoza, Argentina) respectively.
Cell culture and DNA transfections
CHO-K1 cells (A.T.C.C.) were maintained at 37°C, 5% CO2, in DMEM (Dulbecco's modified Eagle's medium) supplemented with 10% FBS (fetal bovine serum) and ATB (antibiotics, 100 μg/ml penicillin and 100 μg/ml streptomycin). Cells grown on coverslips were fixed for 15 min with paraformaldehyde and used for fluorescent microscopy, or grown onto Lab-Tek II chambered coverglass (Nalge Nunc International) for live cell imaging, or grown on Petri dishes for both live cell imaging and Western blot experiments. Cells were transfected with 0.6–1.2 μg/35 mm dish of the indicated plasmid using cationic liposomes (Lipofectamine™).
Subcellular distribution of fusion proteins at steady state
At 24 h after cell transfection, the cells were washed with cold PBS (140 mM NaCl, 8.4 mM Na2HPO4 and 1.6 mM NaH2PO4, pH 7.5) and harvested by scraping for Western blot experiments as described by  or fixing for phenotypic analysis of subcellular distribution. Alternatively, cells were plated onto Lab-Tek II chambered coverglass, incubated for 36 h, and used in live cell imaging on a fluorescence confocal microscope (see below).
Synchronized expression and time course analysis of subcellular distribution of fusion proteins
For synchronized synthesis and time course analyses of subcellular distribution, CHO-K1 cells were grown in 60 mm dishes for Western blot experiments, onto coverslips for phenotypic analyses or onto Lab-Tek II chambered coverglass for live cell imaging and transfected as described above. At the moment of transfection, 50 μg/ml CHX (cycloheximide, Sigma–Aldrich) was added. Then, 5 h after transfection, FBS (10% v/v final concentration) was added to the medium and 3 h later CHX was removed by three washes with DMEM containing FBS and ATB. Where indicated, medium was supplemented with 2-BP (2-bromo hexadecanoic acid, Fluka), BFA (brefeldin A; 1μg/ml, Sigma-Aldrich) or vehicle (DMSO or methanol). Time course expression, membrane association and subcellular distribution were analysed by Western blot and fluorescence microscopy on both fixed and live cells.
Subcellular fractionation and TX-114 (Triton X-114) partition assay
Cells grown onto 60 mm dishes were washed with cold PBS and harvested by scraping in PBS-PIM (PBS containing protease inhibitors). Extracts were centrifuged at 4 °C for 5 min at 13000 g and resuspended in 400 μl of 5 mM T-PIM [buffer T (Tris/HCl, pH 7.0) in the presence of protease inhibitors (Sigma–Aldrich)]. Pellets were dispersed by repetitive pipetting and vortexing. After 30 min of incubation in T-PIM, pellets were passed 60 times through a 25-gauge needle. Nuclear fractions and unbroken cells were removed by centrifuging twice at 4 °C for 5 min at 600 g. Supernatants were then ultracentrifuged at 4 °C for 1 h at 400000 g using a TLA 100.3 rotor (Beckman Coulter). The supernatant (S1 fraction) was removed, and the pellet (P1 fraction) was resuspended in 400 μl of T-PIM. Both fractions were further ultracentrifuged at 400000 g and the supernatant of S1 fraction was removed (S fraction) and the pellet of P1 fraction was resuspended in 400 μl of T-PIM (P fraction). 100 μl of 5% (v/v) TX-114 in T-PIM was added to the S and P fractions and incubated at 4 °C for 1 h. Then, samples were incubated at 37 °C for 3 min and centrifuged at 13000 g. The aqueous upper phase, A, and the detergent-enriched lower phase, D, were separated and extracted again with detergent and aqueous solutions respectively. The resulting samples were adjusted to equal volumes and the detergent content and proteins were precipitated with chloroform/methanol (1:4, v/v) for Western blot analyses.
The stock solution of 2-BP (0.42 M) was prepared in DMSO. CHO-K1 cells expressing recombinant proteins were incubated with 2-BP (in the range 25–150 μM) or vehicle for control cells. To analyse the effect of 2-BP on newly synthesized proteins, 2-BP was added to the culture medium 30 min before CHX withdrawal to allow for an adequate 2-BP internalization to inhibit PAT (palmitoyl-acyl transferase) activity . After CHX withdrawal, cells where washed 3 times with DMEM and further incubated with 2-BP for 3 h before harvesting for Western blot experiments or confocal microscopy. All experiments, except for those of Figure 5, were performed using 2-BP at 150 μM. To analyse the effect of 2-BP on the steady state subcellular distribution of fluorescent proteins, CHX and 2-BP were added to culture medium 24 h after cell transfection (Western blot experiments) or 36 h after plating transfected cells into Lab-Tek II chambered coverglass.
Confocal microscopy and image acquisition
Confocal images were collected using a Carl Zeiss LSM5 Pascal laser scanning confocal microscope (Carl Zeiss AG) or an Olympus FluoView FV1000 confocal microscope (Olympus Latin America), the latter equipped with an AOTF (acoustical optical transmission filter) and a high speed diffraction grating in combination with a variable slit to select the different emission wavelengths to be detected. A multi-line Argon laser (458, 488 and 514 nm) and two Helium Neon lasers (543 nm and 633 nm respectively) were used in both microscopes. For the FluoView FV1000 microscope, CFP was detected by using laser excitation at 458 nm, with a 458/514 nm excitation dichroic mirror and the fluorophore emission collected between 475 and 495 nm. YFP was acquired by using laser excitation at 514 nm, with a 458/514 nm excitation dichroic mirror and emission collected between 530 and 560 nm. Alexa Fluor® 647 was acquired with a laser excitation at 633 nm, a 488/543/633 nm excitation dichroic mirror, and a 650 nm long-pass emission filter. For CFP–YFP or CFP–YFP- Alexa Fluor® 647 colocalization experiments, images were sequentially acquired in line mode. This minimizes bleed-through between channels.
For the Zeiss LSM5 Pascal microscope, fluorescence from YFP was detected by using a 514 nm excitation laser line, a 458/514 dichroic mirror and a 530–560 nm bandpass emission filter, and fluorescence from CFP was detected by using a 458 nm excitation laser line, a 458/514 dichroic mirror and a 475–490 nm bandpass emission filter.
Live cell experiments were performed at 35±2 °C (Tritech DigiTherm temperature controller and a heating adapter plate, CA, U.S.A.) on the Olympus FluoView FV1000 confocal microscope. For colocalization experiments, a 100× UPlanSApo oil immersion/1.4 NA (numerical aperture) (Olympus, Japan) objective was used with a 3× digital zoom, and using an appropriated pinhole to obtain 1 Airy unit for the shortest wavelength (optical slice 0.8 μm). Images of different cells for each experiment were taken in a period no longer than 30 min.
For spectral unmixing experiments, CHO-K1 cells expressing GFP–PH-PLCδ1 or GAP-43–YFP (N13GAP-43 or GAP-43full) were cultured in Lab-Tek II chambered coverglass and analysed on the Olympus FluoView FV1000 confocal microscope. At 60 h after transfection, two sequential images (lambda scan) were acquired by exciting fluorochromes with 488 nm laser line and the emission detector was adjusted to acquire fluorescence emission between 496 and 508 and between 516 and 528 nm respectively. Images were acquired with identical parameters: pixel time 8 μs/pixel (time acquisition 16 s), zoom 2.5, 100× UPlanSApo oil immersion/1.4 NA (Olympus, Japan) objective, 1024 pixels×1024 pixels resolution, pinhole: 1 Airy unit for 496–508 nm emission wavelengths. A (square) ROI (region of interest) around the analysed cell was drawn in each experiment to reduce time acquisition (typically 10 to 16 s). For spectral unmixing of fluorescent signals, the spectral response of each fluorochrome was evaluated in single transfected cells (GFP–PH-PLCδ1 or GAP-43–YFP) by performing a lambda scan over a 75 pixels×75 pixels ROI, selectively excited using the AOTF with identical acquisition parameters. The fluorescence intensity of each fluorochrome, at each emission band pass, was measured and normalized by its maximum value and then utilized for linear unmixing of fluorescent signals taken from a double labelled specimen. After acquisition, images of co-transfected specimens were spectral unmixed using the spectral response of each fluorochrome. The spectral linear unmixing of fluorochromes from co-transfected cells was performed on the spectral deconvolution application of the Fluoview Software, without background subtraction. The performance of the spectral deconvolution algorithm was evaluated by deconvolution of lambda scans from single transfected cells, in the conditions described above for co-transfected specimens, giving no significant signal in the GFP channel when images of single transfected GAP-43–YFP cells were linearly unmixed. The same was observed when GFP fluorescence was evaluated.
Phenotypic analyses were performed on the Carl Zeiss LSM5 Pascal laser scanning confocal microscope using a 63× Plan-Apochromat oil immersion/1.4 NA (Carl Zeiss AG, Germany) with a fully opened pinhole.
Phenotypic analyses of subcellular distribution
The phenotypic analyses were performed essentially as we have previously described , except that for GAP-43 the phenotypes at steady state were classified as TGN+PM, PM and PM+RE, according to morphological and fluorescence intensity differences and triple labelling experiments using specific markers for organelles (see Figures 3 and 4 and Supplementary Figure S3). The TGN+PM phenotype is described as perinuclear dispersed tubular structures and PM staining as average intensity fluorescence greater than cytoplasm; the PM phenotype is PM staining with no significant fluorescence from intracellular compartments; and the PM+RE phenotype is PM staining and where perinuclear structures contain both tubular dispersed structures and a circular structure characteristic of RE, which is distinguishable because of greater average fluorescence intensity than those of dispersed tubules. The average fluorescence intensity on these structures was estimated with the threshold function of the Metamorph software (Methamorph® 4.5, Molecular Devices). For exocytic transport experiments, the phenotypes were classified as TGN>PM, TGN=PM, PM>TGN and PM+RE, according to their spatial organization and average fluorescence intensities, estimated as described above, and using specific markers for organelles. Statistical significance (P) was determined using the t test on the average frequency of phenotypes.
The full-length and N-terminal domain of GAP-43 localize, at steady state, on the PM and perinuclear structures
To investigate the intracellular transport and subcellular distribution of GAP-43 in CHO-K1 cells, we constructed fusion proteins consisting of GAP-43full or the acylation motif (MLCCMRRTKQVEK) of GAP-43, N13GAP-43, fused to the N-terminal domain of CFP or YFP (Figure 1A). The membrane association and the extent of post-translational modification of these constructs were analysed by ultracentrifugation and TX-114 partition assays (Figure 1B) . More than 70% of N13GAP-43 and 60% of GAP-43full were found to be associated with the P fraction after ultracentrifugation. On the other hand, more than 90% of YFP was recovered in the S fraction, whereas the YFP fused to a GPI attachment signal, GPI–YFP, was mostly recovered in the P fraction. 83% and 24% of P-fraction-associated N13GAP-43 and GAP-43full respectively, were found enriched in the TX-114 D phase, reflecting a higher hydrophobic character for N13GAP-43 (Figure 1B). Similarly, when the S fraction was analysed, 39% and <10% of N13GAP-43 and GAP-43full respectively, were found to be enriched in the D phase. YFP was found to partition into the A (aqueous) phase and GPI–YFP into the D phase [23,26]. We next analysed the subcellular distribution of co-expressed N13GAP-43–CFP and GAP-43full–YFP, using live-cell confocal microscopy. N13GAP-43–CFP and GAP-43full–YFP were colocalized at the PM and perinuclear structures (Figure 2A), suggesting that molecular signals for localization on these compartments are present at the N-terminal polypeptide of GAP-43 (acylation, basic residues 6, 7 and 9, and/or acetylation) .
At steady state, GAP-43full and N13GAP-43 are distributed between PM, TGN and RE
It was previously demonstrated that GAP-43 transits the exocytic pathway to arrive at the PM [15,16,28]. In addition, we described the presence of N13GAP-43 both at TGN structures and RE . To extend our previous observation and to characterize further in detail the perinuclear organelles to which GAP-43 (full-length and truncate) binds at steady state, we performed an extensive in vivo colocalization analysis with markers of perinuclear organelles. No colocalization was observed between GAP-43 (GAP-43full and N13GAP-43) and the major histocompatibility complex class II invariant chain isoform–CFP (Iip-33), or for N52galactosyltransferase 2-CFP (Gal-T2), an ER and a proximal Golgi-resident protein respectively (Figure 2B). However, a fraction of perinuclear localized GAP-43full and N13GAP-43 was found colocalizing with GalNAc-T, a TGN-resident protein [23,29,30] (Figure 2B). Moreover, some of the perinuclear structures decorated with GAP-43full and N13GAP-43 were found colocalizing with endocytosed human Alexa Fluor® 647–Tf (transferrin) (Figure 2B), or Rab11 (results not shown), two established RE markers . Interestingly, we also demonstrated that the fractional concentration of GAP-43full and N13GAP-43 on membranes from RE is independent of their expression level (Supplementary Figure S1 at http://www.BiochemJ.org/bj/421/bj4210357add.htm). It should be also remarked that a similar subcellular distribution for ectopically expressed GAP-43full and N13GAP-43 was observed in the undifferentiated neuroblastoma cell line SH-SY5Y, suggesting that sorting mechanisms involved in organelle localization of GAP-43 are also operative in this cell line (Supplementary Figure S2A at http://www.BiochemJ.org/bj/421/bj4210357add.htm). Moreover, confocal microscopy analysis revealed that immunoreactivity for endogenous GAP-43 partially colocalized with endocytosed Tf in a perinuclear region from retinoic acid-treated Neuro2A neuroblastoma cells, demonstrating that a fraction of endogenously expressed GAP-43 in Neuro2A was present at the RE (Supplementary Figure S2B).
To further evaluate the steady state subcellular distribution of GAP-43, we explored in vivo the effect of BFA on the perinuclear localization of GAP-43 by confocal microscopy. BFA reversibly inhibits the guanine nucleotide exchange factors specific to the GTPase Arf1, which causes redistribution/fusion of cis-/medial-Golgi with the ER, and redistribution/fusion of TGN with the RE . In control cells, N13GAP-43 and GAP-43full decorated perinuclear structures, with some of them colocalizing with Tf but not with Gal-T2 (Figure 2C). After BFA treatment, there was a significant redistribution of Gal-T2, but not GAP-43, to the ER. On the other hand, there was a significant increase in the proportion of colocalization between GAP-43 and Tf in perinuclear structures. These results demonstrate that the perinuclear structures containing GAP-43 at steady state mainly correspond to the TGN and RE.
Newly synthesized GAP-43 is mainly associated with TGN, but not with RE; role of dual acylation
The presence of acylated GAP-43 at a given compartment could result from the balance of GAP-43 membrane adsorption/desorption and/or entry and exiting of vesicular carriers containing this protein. In addition, PAT activity on a given membrane also contributes to GAP-43 concentration, by adsorption of non-acylated protein. To discriminate between these possibilities, we first investigated the organelle/s where newly synthesized GAP-43 is primarily associated by quantifying the phenotypic subcellular distribution of N13GAP-43 and GAP-43full after synchronized protein expression (Figure 3). After CHX withdrawal (t=0 h), the subcellular distribution of N13GAP-43 and GAP-43full at different times was evaluated by subcellular phenotype classification according to morphological, fluorescence intensity differences and triple labelling experiments using specific markers for organelles. These results are shown in Figure 3(C) and Supplementary Figure S3 (at http://www.BiochemJ.org/bj/421/bj4210357add.htm), and indicate that after 1 h of CHX withdrawal more than 90% of cells expressing GAP-43 were mostly localized at the TGN (TGN>PM phenotype). At 2 and 3 h, the fraction of cells exhibiting expression of GAP-43 at the PM (TGN=PM and PM>TGN phenotypes) was significantly increased, with this being clearly evident at 7 h after CHX withdrawal. However, at this time, the PM>TGN phenotype was more frequent for N13GAP-43 (60%) than for GAP-43full (15%), suggesting that GAP-43full had been transported less efficiently from TGN to PM. On the other hand, the association of GAP-43 with membranes from the RE was clearly observed at 24 h after CHX withdrawal, with N13GAP-43 being observed early in this organelle at 7 h, demonstrating that the RE-associated GAP-43 is a post-biosynthetic pool probably resulting from endocytosis.
Additionally, we also analysed subcellular distribution of GAP-43 when vesicular traffic was reduced by incubation at 16 °C (Figure 4A). Soon after synthesis at 37 °C, N13GAP-43 and GAP-43full proteins localized more frequently at the TGN [Figures 4B and 4C, 37 °C (7 h)], which was clearly different to the phenotypic subcellular distribution observed at steady state conditions (Figure 4C, steady state). Moreover, incubation at 16 °C caused an additional increase in the number of cells expressing N13GAP-43 and GAP-43full at the TGN [Figure 4C, 16 °C (7 h)], which was later reversed by further incubation of the system at 37 °C [Figure 4C, 16 °C (7 h)+37 °C (5 h)]. The reduction of vesicular transport at 16 °C was confirmed by analysing the exocytic transport of GPI–YFP (Supplementary Figure S4 at http://www.BiochemJ.org/bj/421/bj4210357add.htm). Supporting the notion that vesicular carriers are involved in the trafficking of GAP-43 from TGN to PM, we observed by FLIP (fluorescence loss in photobleaching) experiments that incubation of cells at 20 °C significantly reduced the decay time of the TGN-associated pool of GAP-43 and prevented its targeting to the PM (Supplementary Figure S5 at http://www.BiochemJ.org/bj/421/bj4210357add.htm).
To study PAT activity in the TGN-association of newly synthesized N13GAP-43 and GAP-43full, we analysed the effect of PAT inhibition on this process by using 2-BP (Figure 5A). Treatment with 100 or 150 μM 2-BP almost completely inhibited membrane association, particularly to TGN, of both N13GAP-43 and GAP-43full (Figure 5B). Notably, 50 μM 2-BP inhibited membrane association of GAP-43full, although at this 2-BP concentration there was a significant amount of N13GAP-43 associated to TGN. Taken together, these results show that PAT activity is required for TGN association of GAP-43, and demonstrate that sequences outside the acylation motif of GAP-43full probably affect the enzymatic process of lipid modification of this protein when compared with N13GAP-43, in spite of the fact that both proteins contain the same acylation motif. Supporting the role of acylation for proper GAP-43 membrane association, it was also observed that a double mutant at Cys3 and Cys4 of N13GAP-43 and GAP-43full occurred diffusely throughout the cells, characteristic of a cytosolic protein (Figure 5C). We next analysed in vivo the subcellular distribution of single acylation mutants of N13GAP-43 and GAP-43full at steady state (Figure 5C). The point mutation at Cys3 in N13GAP-43 caused an accumulation at the cytosol and TGN and also disrupted RE association, but did not affect their PM association. On the other hand, the point mutation at Cys3 in GAP-43full and mutation at Cys4 in N13GAP-43 and GAP-43full disrupted the association with the PM, drastically reduced their TGN association, and resulted in an accumulation at the cytosol. The differences observed between GAP-43 mutants with regard to their capacity to bind membranes could be due to differences in their ability to be post-translationally lipid modified, as previously reported . These results suggest that efficiency in fatty acylation of GAP-43 at the TGN could modulate its incorporation into vesicular transport carriers directed to the PM, and also indicate that a dual acylation is necessary for its proper RE localization.
A de/reacylation cycle does not account for the accumulation of GAP-43 at the TGN
As indicated above, a fraction of GAP-43 localized at the TGN at steady state. From these and other published results [5,22,27], we thought that the presence of this TGN-associated pool of GAP-43 could be due, at least in part, to two mechanisms: the first involving a fast deacylation step at the PM, thereby disrupting the membrane association of the protein, and a later acylation by TGN-specific PAT, causing a kinetic trapping of the protein in the organelle [27,32]. The second mechanism implies a slow, rate-limiting transport in the secretory pathway after GAP-43 had been synthesized and acylated at the TGN. To analyse these possibilities, we first tested in vivo the kinetics of deacylation of GAP-43full and N13GAP-43 at steady state by treating the cells with 2-BP and CHX at different times, followed by TX-114 partition assay. As control, we analysed, at the same conditions, the TX-114 partition of newly synthesized GAP-43full and N13GAP-43 in cells treated with 2-BP. PAT inhibition prevented the partition of newly synthesized GAP-43full and N13GAP-43, but not GPI–YFP (results not shown), to the detergent phase, which indicates that the non-acylated protein was unable to partition into this fraction (Figure 6A). The analysis of the effect of 2-BP on the steady-state pool of GAP-43full and N13GAP-43 revealed that there was no significant deacylation of the expressed protein for a wide range of time periods (0, 3 and 6 h) (Figure 6B, as shown by the fraction of GAP-43 present in the aqueous phase from S fraction). These results were further confirmed by using live cell confocal microscopy (Figure 6C). Taken together, results from these experiments (Figures 3–6) demonstrate the same subcellular trafficking itinerary for N13GAP-43 and GAP-43full (TGN→PM→RE). Moreover, TGN accumulation of GAP-43 at steady state is probably due to a slow, rate-limiting transport to PM after it is synthesized in the cytosol, rather than to a process involving deacylation and later acylation at the TGN.
The pool of GAP-43 associated to the pericentriolar compartment is dynamic and exchangeable by vesicular transport with the pool associated to the PM
To address if the pool of GAP-43 present in RE exchanges with GAP-43 on the PM, FRAP (fluorescence recovery after photobleaching) experiments were performed in CHO-K1 cells previously treated with CHX to reduce the pool of TGN-associated GAP-43 (Figure 7B). To that end, the RE was photobleached by confocal laser microscopy and the region was then monitored for fluorescence recovery. A typical FRAP experiment is shown in Figure 7(C, panel i, upper row). In this cell, prior to photobleaching, N13GAP-43–YFP appears as a bright spot associated with the RE, colocalizing with Tf. Following focal irradiation, this fluorescence was considerably reduced. Subsequently, the fluorescence began to rapidly reappear in the RE recruitment of N13GAP-43–YFP. A quantification of this recovery process is shown in Figure 7(C, panel ii). Clearly, results from these experiments demonstrate the dynamic nature of GAP-43 (GAP-43full and N13GAP-43) trafficking to the RE from PM in CHO-K1 cells.
To demonstrate the role of vesicular transport in PM to RE trafficking of GAP-43, FRAP experiments in tannic acid-treated cells were performed. The impermeable fixative tannic acid avoids both vesicular fusion and endocytosis at the PM [33,34]. To check first whether tannic acid impairs internalization by vesicular transport in CHO-K1 cells, we followed Tf endocytosis. In control cells, Tf was localized in the perinuclear endosomal compartment, whereas in cells treated with tannic acid, it was only present at the cell surface (Figure 7A). As additional control, we verified that tannic acid did not significantly affect the non-vesicular redistribution of K-Ras from PM to RE after valinomycin treatment  (Figure 7C, panel ii). However, we observed that in tannic acid-treated cells, trafficking of GAP-43 to RE was drastically reduced (Figure 7C, panel ii). As control, we also investigated the effect of tannic acid on the internalization of GPI–YFP via the clathrin-independent endocytic pathway . Similar to GAP-43, RE/Golgi complex refilling of GPI-GFP was considerably inhibited in tannic acid-treated cells (Figure 7C, panel ii). Taken together, the results from these experiments clearly indicate vesicular trafficking of GAP-43 from PM to RE.
Expression of Arf6(Q67L) causes accumulation of GAP-43 into enlarged vacuolar membranes
Internalization of many GPI-anchored proteins and the double acylated H-Ras occurs by a clathrin-independent pathway regulated by the small GTP-binding protein Arf6 (Figure 8) [36,37]. It has been described that expression of the constitutively active GTPase-deficient Arf6 mutant, Arf6(Q67L), perturbs trafficking of those cargo proteins transported through the Arf6-associated pathway, resulting in cargo accumulation in vacuolar structures that are enriched in PIP2 and actin. We examined the effect of expression of Arf6(Q67L) on GAP-43 trafficking and we found that GAP-43 was at the PM and associated with vacuolar structures decorated with the fluorescent probe GFP–PH-PLCδ1 (Figure 8). These results show that GAP-43 is internalized through the Arf6-associated pathway and supports the notion of vesicular transport in the trafficking of this acylated peripheral protein from PM to endosomal compartments.
GAP-43 is acylated mainly at the TGN
The presence of GAP-43 at the TGN, early after its synthesis, allows us to identify this subcellular compartment as the place where the lipid modification of GAP-43 occurs. The pharmacological inhibition of acylation with 2-BP completely disrupts TGN (and also other membrane) association, further confirming that this adsorption is a PAT activity-dependent process. In addition, double mutation of Cys3 and Cys4 of GAP-43 completely disrupts membrane association. Cell incubation at 16 °C, which inhibits vesicular transport, induced an increase in the number of cells expressing GAP-43 at the TGN, thus confirming that the initial adsorption of newly synthesized GAP-43 mainly occurs on this organelle. However, under this condition, a minor fraction of GAP-43 was also found localized at the PM, suggesting the presence of PATs in this membrane, which were able to acylate GAP-43, although less efficiently, than the TGN-associated PAT. In this regard, it was proposed that acylation could occur in different compartments besides the TGN . In contrast with previous results , we were not able to observe in CHO-K1 cells any association of newly synthesized GAP-43 with the ER–Golgi intermediate compartment.
The requirement of acylation of GAP-43 for TGN adsorption suggests that the enzyme/s catalysing this post-translational modification (PAT) is/are present at the TGN. PATs are multiple spanning membrane proteins containing a conserved DHHC (aspartate-histidine-histidine-cysteine) domain. In all, five PATs have been so far described to catalyse the acylation of GAP-43 (DHHC2, 3, 7, 15 and 17) [18,39], with the subcellular distribution and tissue expression of these proteins having already been previously characterized [18,40]. DHHC2 was found to be observed in membranes from the ER and Golgi complex, whereas DHHC3, 7, 15 and 17 were discovered mainly associated with the Golgi complex. DHHC7 and 17 are ubiquitously expressed in tissues, with GAP-43 being highly expressed in the brain . Therefore, DHHC7 and 17 are excellent candidates for GAP-43 acylation at the TGN, for both neuronal and non-neuronal cells. The ubiquitous tissue expression of DHHC5 and 21 and their PM localization  suggest that these PATs could be also involved in the PM acylation of GAP-43, although it remains to be demonstrated.
Vesicular carriers mediate TGN→PM→RE trafficking of GAP-43
TGN to PM transport of GAP-43
Cell incubation at 16 °C, which reduces vesicular transport, induced an increase in the TGN association of GAP-43. Using FLIP experiments, we observed that incubation of cells at 20 °C significantly reduced the decay time of the TGN-associated pool of GAP-43 and prevented its targeting to PM. Moreover, fixation of PM with tannic acid, which prevents the fusion of exocytic vesicles, but not the adsorption of proteins (i.e. K-Ras), caused an increase of discrete structures dispersed throughout the cytoplasm decorated with GAP-43full and N13GAP-43 without affecting Golgi complex structure (Supplementary Figure S6 at http://www.BiochemJ.org/bj/421/bj4210357add.htm). These results support the idea that molecular signals present in the first 13 aa are sufficient for inclusion into the same vesicular carriers. Although positively charged aa at position 6, 7 (arginine-arginine) and 9 (lysine) of GAP-43 contribute to membrane binding and polarized sorting, they are not absolutely necessary for inclusion into TGN to PM vesicular carriers [7,28]. Therefore, in agreement, our results show that dual acylation, into a cluster of hydrophobic aa and/or acetylation, operates as a dominant signal for GAP-43 inclusion into vesicular carriers, whose selectivity could be modulated by the presence of Arg6, Arg7 and Lys9 residues.
We next attempted to identify the kind of carrier involved in TGN to PM transport of GAP-43. It was found that BFA had no appreciable effect on the transport of newly synthesized N13GAP-43 from TGN to PM, suggesting that in these processes, the BFA-sensitive guanine nucleotide exchange factors might not be participating (Supplementary Figure S7 at http://www.BiochemJ.org/bj/421/bj4210357add.htm), and therefore discarding an essential role of clathrin-coated vesicles in the TGN exiting of GAP-43. These results are in agreement with those recently published describing that selective knockdown of clathrin does not affect the sorting of proteins to the apical domain of MDCK cells , comparable with the axonal delivery of proteins in neurons. Thus our results suggest a clathrin-independent and double acylation-dependent process operating in the TGN exiting of GAP-43 in different cell types.
It was also observed that there are kinetic differences with respect to times of exiting from and arrival at each subcellular compartment, with N13GAP-43 being transported quicker to the PM than GAP-43full, suggesting that sequences outside the acylation motif of GAP-43 modulate its vesicular transport. In addition, we also found that single acylated mutants of N13GAP-43 were inefficient in reaching the PM and concentrated at the TGN. In particular, this phenomenon was clearly evidenced for N13GAP-43C3S. Moreover, treatment with 50 μM 2-BP did not disrupt the TGN association of N13GAP-43, but severely inhibited its PM targeting, suggesting that the TGN-associated N13GAP-43 pool is single acylated and hence unable to be efficiently transported to PM, supporting a role of dual acylation for the proper inclusion of GAP-43 into TGN to PM vesicular carriers.
PM to RE transport of GAP-43
The role of vesicular transport in PM to RE trafficking of GAP-43 was demonstrated by FRAP experiments in tannic acid-treated cells (Figure 7). This was further confirmed by the overexpression of Arf6(Q67L), which affects endocytic trafficking by inhibiting the fusion of early endocytic vesicles with sorting endosomes  (Figure 8). A similar result was also observed for double acylated H-Ras. Arf6 promotes the internalization of PM proteins via clathrin-dependent [42–44] and clathrin-independent [36,37] endocytosis through activation of phosphatidylinositol-4-phosphate 5-kinase to generate PIP2. We found that the expression of a dominant-negative mutant of the epidermal growth factor receptor pathway substrate clone 15, which selectively affects clathrin-mediated endocytosis, did not significantly affect GAP-43 trafficking to the RE, suggesting that clathrin-coated vesicles are not participating in the endocytic process of GAP-43 (results not shown). We also observed that single acylated mutants of GAP-43 did not arrive at the RE compartment, supporting a role of dual acylation for endocytic trafficking of GAP-43 (Figure 5C). Thus, these results indicate that the exocytic trafficking of GAP-43, from the TGN to PM as well as its endocytic delivery to RE, is mediated by vesicular carriers. Moreover, this supports the notion that dual acylation is required for RE delivery of acylated proteins .
Deacylation is not involved in the retrograde transport of GAP-43 to TGN
We demonstrated that, at steady state, a fraction of N13GAP-43 and GAP-43full is present at the TGN probably as a consequence of a rate-limiting TGN exit of newly synthesized protein and not through a de/reacylation mechanism [27,32], in which CHO-K1 cells could operate over a time scale of hours. Moreover, we demonstrated that the pool of TGN-associated GAP-43full was reduced after 6 h of CHX treatment (results not shown). These results discard the possibility that a retrograde transport of GAP-43 from PM (or other endomembranes) to TGN, involving vesicular carriers or a desorption facilitated process (protein mediated, without deacylation), is the principal contribution to the steady state TGN localization of GAP-43.
To sum up, in this work we have shown that the intracellular transport and main destinations (TGN, PM and RE) of GAP-43 depend on its dual acylation state. This molecular requirement is important for inclusion into exocytic and endocytic vesicular carriers. We showed that there are kinetic differences with respect to times of exiting from and arrival at each subcellular compartment, with N13GAP-43 being transported faster to PM than GAP-43full, suggesting that sequences outside the acylation motif of GAP-43 modulate its acylation state and vesicular transport. The intracellular trafficking of GAP-43 was demonstrated to be complex, resulting from a dynamic interplay between its exocytic and endocytic transport and PAT-dependent membrane adsorption. In particular, we have described for the first time the presence of GAP-43 at the RE, along with the molecular requirements necessary for this subcellular location. What is the potential role of GAP-43 at the RE? As already mentioned, the basic effector domain of GAP-43 participates in clustering of PIP2, which can be blocked by Ser41 phosphorylation . Moreover, it is known that the phosphorylated form of GAP-43 at Ser41 by classical and novel isoforms of PKC  stabilizes long actin filaments and plays an important role in actin skeleton reorganization underlying growth cone motility and adhesion to substrate [46–48]. In addition, it has been shown that classical PKCα and PKCβII can associate to membranes of the RE after phorbol ester stimulation in HEK-293 (human embryonic kidney 293) cells  and CHO-K1 cells (A. Trenchi, G. A. Gomez and J. L. Daniothi, unpublished work). Taken into consideration all these findings and those described in the present study, we hypothesize that GAP-43, probably modulated by PKC phosphorylation, could be regulating vesicle formation and transport at the RE through its interaction with polyphosphoinositides and actin, as was previously described at the cytoplasmic leaflet of the PM [14,50].
Alejandra Trenchi and Guillermo Gomez performed experiments. Alejandra Trenchi, Guillermo Gomez and Jose Daniotti conceived ideas, designed experiments, analysed results and wrote the manuscript. All authors edited and reviewed the final manuscript before submission.
A. T. and G. A. G. are recipients of CONICET (Consejo Nacional de Investigaciones Científicas y Técnicas, Argentina) Fellowships. J. L. D is a Career Investigator of CONICET. This work was generously supported by research grants to J. L. D. from Agencia Nacional de Promoción Científica y Tecnológica (ANPCyT)-Ministerio de Ciencia, Tecnología e Innovación Productiva de Argentina [grant number PICT 601], CONICET [grant number PID 5151], Secretaría de Ciencia y Tecnología de la Universidad Nacional de Córdoba [grant number 162/06] and Ministerio de Ciencia y Tecnología de la Provincia de Córdoba.
We acknowledge the technical assistance of G. Schachner, S. Deza and C. Mas. G. A. G. thanks Adolfo Alfonso Pecchio (CIQUIBIC, Córdoba, Argentina) for supplying the Iip-33-CFP construct and for helpful discussions on quantitative fluorescence microscopy.
Abbreviations: aa, amino acids; AOTF, acoustical optical transmission filter; A phase, aqueous upper phase; Arf, ADP-ribosylation factor; ATB, antibiotics; 2-BP, 2-bromo hexadecanoic acid; BFA, brefeldin A; CaM, calmodulin; CFP, cyan fluorescence protein; CHO, Chinese hamster ovary; D phase, detergent-enriched lower phase; ECFP, enhanced cyan fluorescence protein; ER, endoplasmic reticulum; EYFP, enhanced yellow fluorescent protein; CHX, cycloheximide; DMEM, Dulbecco's modified Eagle's medium; FBS, fetal bovine serum; FLIP, fluorescence loss in photobleaching; FRAP, fluorescence recovery after photobleaching; GalNAc-T, UDP-GalNAc:LacCer/GM3/GD3 N-acetylgalactosaminyltransferase; Gal-T2, UDP-Gal: GA2/GM2/GD2/GT2 galactosyltransferase; GAP-43, growth-associated protein-43; GAP-43full, full-length wild-type GAP-43; N13GAP-43, GAP-43 only containing the acylation motif (aa 1–13); GFP, green fluorescent protein; GPI, glycosylphosphatidylinositol; NA, numerical aperture; PAT, palmitoyl-acyl transferase; GFP–PH-PLCδ1, chimeric protein containing the PIP2-specific pleckstrin homology domain of phospholipase Cδ1 fused to GFP; HA, haemagglutinin; PIP2, phosphatidylinositol (4,5)-bisphosphate; PKC, protein kinase C; PM, plasma membrane; RE, recycling endosome; ROI, region of interest; Tf, transferrin; TGN, trans-Golgi network; TX-114, Triton X-114; YFP, yellow fluorescent protein
- © The Authors Journal compilation © 2009 Biochemical Society