Genetic Analysis of the Neuronal and Ubiquitous AP-3 Adaptor Complexes Reveals Divergent Functions in Brain

INTRODUCTION

Membrane proteins reach their resident organelles by means of vesicle carriers that selectively sequester membrane protein cargo (Bonifacino and Glick, 2004). Central to membrane protein sorting and vesiculation processes is a family of cytosolic adaptor (AP) complexes that recognize sorting signals present on cargo proteins (Boehm and Bonifacino, 2001; Bonifacino and Traub, 2003; Robinson, 2004). Adaptors are made of four subunits or adaptins: a large α, γ, δ, or ϵ adaptin; a large β, a medium μ, and a small σ subunit. Numbers appended to the β, μ, or σ adaptin denote the adaptor complex to which the subunit belongs. Thus, for example, the adaptor complex 3 (AP-3) is assembled by a single copy of δ, β3, μ3, and σ3 adaptins. Diverse vesiculation mechanisms are accounted only in part by multiple adaptors. In mammalian cells, four adaptor complexes are localized to specific subcellular locations providing a first layer of diversity in the generation of distinct vesicle carriers (Boehm and Bonifacino, 2001; Bonifacino and Traub, 2003; Robinson, 2004). However, multiple isoforms in the subunits constituting individual adaptor complexes likely provide additional diversification to the membrane protein sorting and vesiculation mechanisms (Takatsu et al., 1998; Folsch et al., 1999, 2001, 2003).

The AP-3 complex is unique among adaptors because three of its subunits are encoded by two alternative genes: β3A/β3B, μ3A/μ3B, and σ3A/σ3B (Dell'Angelica et al., 1997a,b; Simpson et al., 1997). Of these subunits, β3B and μ3B are exclusively expressed in neurons, suggesting that they assemble AP-3 complexes devoted to neuron-specific sorting and vesiculation (Pevsner et al., 1994; Newman et al., 1995). Our knowledge about AP-3 function has been greatly illuminated by spontaneous AP-3 mouse mutations affecting two independent loci, mocha (Kantheti et al., 1998, 2003) and pearl (Feng et al., 1999). Loss of AP-3 in these mice leads to defective biogenesis of lysosomes and specialized secretory organelles, such as melanosomes, platelet-dense granules, lymphocyte cytotoxic granules (Dell'Angelica et al., 2000; Clark et al., 2003), and in neurons, synaptic vesicles (Kantheti et al., 1998, 2003; Salazar et al., 2004a,b).

Mocha affects δ adaptin, a unique AP-3 subunit expressed in all tissues and an obligatory component to all AP-3 complexes (Kantheti et al., 1998, 2003). Its absence leads to degradation of all the AP-3 subunits both in neuronal and nonneuronal tissues (Kantheti et al., 1998, 2003). In contrast, the pearl allele perturbs the β3A subunit, thus depleting AP-3 in all tissues except neurons, which still assemble AP-3 complexes carrying the neuronal-specific β3B isoform (Feng et al., 1999). Pearl and mocha mice share all their lysosomal and specialized secretory phenotypes except for those that depend on the assembly of synaptic vesicles, an organelle exclusively perturbed in the mocha allele. This divergence on the synaptic phenotypes associated with AP-3 deficiencies may derive from the fact that pearl neurons possess functional β3B-containing AP-3 complexes. Yet, this alone does not explain whether neurons selectively require β3B-containing AP-3 complexes (neuronal AP-3), or whether β3A- and β3B-containing AP-3 complexes are functionally interchangeable in neuronal tissue as is the case in nonspecialized fibroblastic cell types (Peden et al., 2002). In vitro reconstitution of synaptic-like microvesicle biogenesis supports the first hypothesis because formation of these vesicles requires neuronal AP-3 complexes (Faundez et al., 1998; Blumstein et al., 2001). However, whether neuronal β3B-containing AP-3 fulfills unique functions in the biogenesis of brain synaptic vesicles and regulates presynaptic functions remains largely unexplored. In this manuscript, we investigated whether β3B-neuronal AP-3 complexes mediate unique neuron-specific functions that cannot be accomplished by β3A-containing AP-3. We tested this hypothesis by using genetically engineered mouse strains lacking neuronal or ubiquitous AP-3 generated by ablation of the β3B and β3A subunits, Ap3b2^-/- and Ap3b1^-/-, respectively. Analysis of the subcellular distribution of AP-3, the synaptic zinc stores, and the targeting of AP-3–dependent synaptic vesicle cargoes ZnT3 and CIC-3 (Salazar et al., 2004a,b) indicated that the functions of both AP-3 complexes were distinct and divergent. Our results suggest that concerted nonredundant functions of neuronal and ubiquitous AP-3 provide a mechanism to control levels of membrane proteins in synaptic vesicles.

MATERIALS AND METHODS

Animals

The Ap3b1^-/- mouse line congenic on C57BL/6J was kindly donated by Dr. S. Mansour (University of Utah, Salt Lake City, UT) (Yang et al., 2000) and then bred in-house. Ap3b2^-/- animals were generated by directly targeting C57BL/6 and continued breeding on the same genetic background. Most animals for our experiments were obtained from matings of heterozygous males and females. Control animals used were Ap3b2^+/+ littermates of Ap3b2^-/- animals. Mocha mice were originally obtained from The Jackson Laboratory (Bar Harbor, ME) and then bred in-house, also as heterozygotes. All animal procedures were approved by the University of Michigan and Emory University Committees on Use and Care of Animals.

Ap3b2 Targeting and Ap3b2^-/- Mouse Generation

To generate Ap3b2^-/- mice on a C57BL/6 (B6) genetic background for comparison to Ap3b1^-/- mice, we targeted Ap3b2 in B6-derived (Bruce4) embryonic stem (ES) cells (Kontgen et al., 1993) as well as in 129S1/SvImJ (129)-derived (CJ7) ES cells (Swiatek and Gridley, 1993). Targeting frequency and germ line transmission from both ES cell lines have been previously compared for this construct (Seong et al., 2004).

The targeting vectors for both 129 and B6 targeting were prepared using the pflox vector (Chui et al., 1997). To target a large region of Ap3b2, we first screened 129 and B6 bacterial artificial chromosome (BAC) libraries (CITB-CJ7 and RPCI-23, respectively; Research Genetics, Huntsville, AL). From resulting BACs identified in each library, two identical targeting vectors were created by subcloning the target region, exons 5–12 of Ap3b2, from each of the coisogenic BACs into the pflox vector. After electroporation of the targeting vector, ES cells were screened by genomic polymerase chain reaction (PCR) followed by Southern blot analysis of PCR-positive clones. Further Southern blot analyses and PCR from both ends were performed to verify that the selected ES cells have a properly targeted allele without affecting neighboring genomic regions. Ap3b2^-/- animals were obtained on three genetic backgrounds: 129, B6, and (129 × B6)F2. However, to compare the most similar genetic backgrounds, only Ap3b2^-/- mice on C57BL/6 were used for the data presented here.

Northern blots were prepared as described previously (Bomar et al., 2003). ³²P-labeled antisense RNA probes were prepared by in vitro transcription using Maxiscript (Ambion, Austin, TX) from Ap3b1 and Ap3b2 reverse transcription (RT)-PCR fragments with the additional following T7 sequence at the 5′ of reverse primers: TAA TAC GAC TCA CTA TAG GGA G. Ap3b1 and Ap3b2 RT-PCR primers are Ap3b1mF3161 (ACT TCA CTC CCT CCA TGA TCC TC), Ap3b1mR3498 (TGC CAG ATG GAA GGG CTA TTA TT), Ap3b2mF2874 (CAG CCA ACT TCC AGC TGT GC), and Ap3b2mR3143 (TCC TCC CTG CAA ACC TGT ACT C).

Open Field Test

Individual mice were placed into a white nontransparent chamber 33.5 × 36.5 cm in size, and activity was recorded for 100 min. Horizontal locomotion of each mouse was determined using Ethovision software (Noldus, Amsterdam, The Netherlands). Statistical analysis was performed with SPSS 11.0. Statistical analyses were performed using the unpaired t test and analysis of variance followed by Bonferroni correction.

Antibodies

Monoclonal antibody (mAb) to synaptophysin (SY38) and polyclonal antibody to MAP2 were purchased from Chemicon International (Temecula, CA). mAb to SV2 (10H4) was a gift of Dr. R. Kelly (University of California, San Francisco, San Francisco, CA). Anti-116-kDa subunit of the vacuolar ATPase was purchased from Synaptic Systems (Göttingen, Germany). Affinity-purified antibodies directed against ClC-3 were a gift from Dr. D. J. Nelson (University of Chicago, Chicago, IL) (Huang et al., 2001). Affinity-purified polyclonal antibody against ZnT3 has been described previously (Salazar et al., 2004b). mAb against AP-3 δ (SA4) was obtained from Developmental Studies Hybridoma Bank at the University of Iowa (Iowa City, IA) (Peden et al., 2004). mAb to β3B (βNAP) was purchased from BD Transduction Laboratories (Lexington, KY). Polyclonal pan-β3 antibody that recognizes both β3A and β3B, and polyclonal pan-σ3 have been described previously (Faundez et al., 1998; Salem et al., 1998). Polyclonal antipeptide antibodies against β3B were described in Blumstein et al. (2001).

Primary Cultures of Cortical Neurons

Cultures were generated from 16-d-old mouse embryos. Ap3b1^-/-, Ap3b2^-/-, mocha, and control mating cages were set up at the same time, and embryos were processed simultaneously. The morning that the vaginal plug occurred was defined as day 1. Because homozygous Ap3b2^-/- and mocha animals are subfertile, homozygous males were mated to heterozygous females to obtain homozygous embryos. Ap3b1^-/- and mocha embryos were identified by the delayed eye pigmentation phenotype. Ap3b2^-/- embryos were morphologically indistinguishable; thus, entire litters were individually processed and genotyped later.

Primary cultures of cortical neurons were performed according to the method described by Meberg and Miller (2003) with some modifications. Briefly, both cortical hemispheres from each embryo were isolated and dissociated with trypsin, and cells were plated at 10,000 cells/cm² in Neurobasal medium with 10% fetal bovine serum (FBS), 1× 100 U/ml penicillin and 100 μg/ml streptomycin (PS). After 12 h, the FBS-containing medium was replaced with Neurobasal medium supplemented with B27, l-glutamine, 1× PS. 5-Fluorouracil deoxyribonucleoside (1 μM) was added 4 d after plating, and cells were fed twice weekly thereafter with the same B27-supplemented Neurobasal medium. Cortical neurons were grown at 37°C in 5% CO₂ on coverslips coated with poly-d-lysine (Sigma-Aldrich, St. Louis, MO) in 24-well plates. All cell culture reagents were purchased from Invitrogen (Carlsbad, CA). Cells were used after 4–19 d in vitro (DIV).

Immunocytochemistry and Confocal Microscopy

Cells were washed twice in phosphate-buffered saline (PBS) for 5 min. After washing, cells were fixed for 20 min in 4% paraformaldehyde and processed for immunofluorescence. Fixed cells were permeabilized in blocking buffer (0.02% saponin in PBS, 2% bovine serum albumin, and 1% fish skin gelatin) for 30 min, and then incubated with primary antibodies for 90 min. After washing three times with blocking buffer, each time 15 min, cells were incubated with secondary antibody for 60 min. Secondary antibodies used were Alexa-conjugated goat anti-mouse 488 and/or goat anti-rabbit 594. Cells were then rinsed and mounted in Prolong Gold (Invitrogen). Specimens were viewed using a FluoView 500 confocal microscope (Olympus, San Diego, CA) coupled to HeNeG and Argon ion lasers. Alexa-488 and Alexa-594 were excited with the Argon laser tuned at 488 nm and the HeNeG laser tuned at 543 nm, respectively. Images were acquired using FluoView 4.3 software (Olympus). The emission filters used were BA505–525 and BA610IF. Images were viewed and acquired using UplanApo 20×/0.7, 60×/1.4 oil, or 100×/1.4 oil objective. To compare Ap3b1^-/-, Ap3b2^-/-, mocha, and control neurons, all images were captured at the same time and threshold values.

Timm's Staining

Adult mocha, Ap3b1^-/-, Ap3b2^-/-, and control mice were used for Timm's staining. Timm's staining was performed according to the method by Sloviter (1982) with some modifications. After an anesthetic overdose (pentobarbital, 60 mg/kg i.p.), all animals were transcardially perfused as follows: 1× PBS, pH 7.2 (Invitrogen) for 3 min, 0.37% sulfide solution, pH 7.2, for 5 min, and 4% paraformaldehyde in 1× PBS for 3 min. The brains were removed, postfixed for 3–5 h, and left overnight in 25% sucrose in 0.1 M phosphate buffer, pH 7.4. Coronal vibratome sections (40 μm in thickness) were mounted on slides and allowed to dry. Sections were developed in the dark for 45 min in a solution consisting of 250 ml of 50% gum arabic, 40 ml of sodium citrate buffer, 120 ml of 5.7% hydroquinone, and 2 ml of 17% silver nitrate. Slides were then rinsed in distilled water for 5 min, dehydrated, and coverslipped. Sections were viewed at 6.3× under transillumination by using a Leica MZ FL III stereomicroscope (Leica Microsystems, Bannockburn, IL). Images were captured on a DKC5000 digital camera (Sony, Tokyo, Japan) and acquired using Adobe Photoshop (Adobe Systems, San Jose, CA).

Immunohistochemistry

Animals were treated similarly, except that sulfide was omitted from the perfusion solution. Brains were then sliced coronally and embedded in paraffin. To ensure identical processing, slices containing control, Ap3b1^-/-, and Ap3b2^-/- hippocampi were included in the same block and sectioned together (8 μm in thickness). Antigens were retrieved by microwave treatment in 10 mM citrate buffer, pH 6, for 10 min. All stainings were performed in duplicate in at least two independent experiments. Primary antibody incubations were carried out overnight at 4°C, and immunocomplexes were detected by species-specific Vectstain avidin-biotinylated enzyme complex kit (Vector Laboratories, Burlingame, CA) according to manufacturer's instructions. ZnT3 antibody was used at 1/200; serial dilutions up to 1/1000 gave identical results.

Morphometric Analyses

Fluorescence intensity determinations were performed using MetaMorph software (Universal Imaging, Downingtown, PA). Tiff files of single optical sections obtained with FluoView 4.3 software were imported directly into MetaMorph. Processes and cell bodies were identified in the MAP-2 channel, outlined, and the average delta fluorescence intensity of at least four randomly selected processes per neuron was measured. Images were collected from two independent experiments totaling 36 wild-type, 33 Ap3b1^-/-, 34 Ap3b2^-/-, and 11 mocha neurons. Specific δ antibody fluorescence intensity was determined by subtracting the average fluorescence intensity obtained in mocha neurons from wild-type, Ap3b1^-/-, and Ap3b2^-/- values.

Timm's staining images were captured in color using Photoshop software. Tiff files were converted to gray scale and inverted to obtain “negatives” of the brain section images by using Adobe Photoshop 7. Timm's deposits showed up as a white signal, which was quantified using MetaMorph software. Average intensities were obtained from two areas of the brain cortex and from the stratum oriens and radiatum of the hippocampi. A total of 19 wild-type (3 mice), 20 Ap3b1^-/- (3 mice), 29 Ap3b2^-/- (4 mice), and 13 mocha (2 mice) brain sections obtained in three independent experiments were analyzed.

Cell Fractionation, Immunoprecipitation, and Western Blot Analysis

Frozen brains of Ap3b1^-/-, Ap3b2^-/-, and control animals were pulverized to a fine powder by using porcelain mortars under a continuous supply of liquid nitrogen. Extracts were thawed at 4°C in 5 volumes of buffer A (150 mM NaCl, 10 mM HEPES, pH 7.4, 1 mM EGTA, and 0.1 mM MgCl₂ plus Complete antiprotease mixture; Salazar et al., 2004a). Homogenates were sedimented at 1000 × g for 10 min to generate S1 supernatants. S1 were further fractionated at 27,000 × g for 45 min to obtain high-speed supernatants, which were resolved in glycerol velocity gradients (5–25%) prepared in intracellular buffer at 218,000 × g for 75 min in a SW55 rotor (Beckman Coulter, Fullerton, CA). Gradient fractions were analyzed by immunoblot, and immunoreactivity was revealed by enhanced chemiluminescence. Immunoreactive bands were quantified using NIH Image 1.62 software as described previously (Salazar et al., 2004b).

Immunoprecipitation was performed with polyclonal pan-σ3 or monoclonal δ antibodies prebound to protein G-Sepharose, as described previously (Faundez and Kelly, 2000; Salazar et al., 2004b).

Other Procedures

Protein concentration was determined by Bradford reagent (Bio-Rad, Hercules, CA). Data are expressed as average ± SE. Statistical analysis was performed using nonpaired, two-tailed t test.

RESULTS

β3B-deficient Mice Generated by Gene-targeting of Ap3b2

Neurons abundantly express AP-3 complexes assembled with either the ubiquitous or the neuronal-specific β3 isoforms (Newman et al., 1995; Dell'Angelica et al., 1997b). However, it has not been determined whether the function of β3A- and β3B-containing AP-3 complexes is redundant in neurons, a specialized secretory cell. To test this hypothesis, we targeted Ap3b2 exons 5 through 12 and removed one-third of the Ap3b2 coding region near the N terminus (Figure 1A) to create a β3B-deficient mouse model. The normal β3B protein consists of 1082 amino acids, whereas the targeted gene encodes an aberrantly short polypeptide consisting of the first 120 amino acids followed by 63 extra amino acid residues. The Ap3b2 gene was disrupted in both CJ7 (129S1/SvImJ-derived) and Bruce4 (C57BL/6-derived) ES cell lines (Seong et al., 2004). Properly targeted ES cell lines and their germline transmission in animals were screened and confirmed by both PCR (our unpublished data) and Southern blot analysis (Figure 1B). CJ7 ES cell-derived Ap3b2^-/- lines were bred to C57BL/6J, and Ap3b2^-/- animals were on a mixed F2 genetic background of 129S1/SvImJ (129) and C57BL/6 (B6) strains. By targeting in the Bruce4 ES cell line, a different Ap3b2^-/- line was created on a B6 background. Although the major phenotypes of Ap3b2^-/- mice were consistent regardless of the genetic background, we used Ap3b2^-/- mice on B6 for comparisons with Ap3b1^-/- mice on the same genetic background. Ap3b1^-/- mice were initially generated in ES cells from two different 129 substrains (Yang et al., 2000), but they have been repeatedly (>8 generations) backcrossed to B6, resulting in an almost pure B6 genetic background. We used Ap3b1^-/- instead of pearl mice because they carry a null allele for β3A, whereas pearl mice carry a hypomorphic allele (Yang et al., 2000; Peden et al., 2002). Disruption of the Ap3b1 or Ap3b2 loci was demonstrated either by the absence of transcripts or by the generation of aberrant transcripts, respectively. Transcripts of the targeted Ap3b1 allele were absent (Figure 1C, lane 4). However, the targeted Ap3b2 mRNA was transcribed into a smaller mRNA of ∼2.7 kb, with exons 5 through 12 replaced with the neo cassette from the vector (Figure 1C, lane 7). Although the aberrant mRNA is as stable as the wild-type, RT-PCR analysis confirmed that its sequence did indeed contain an early stop codon due to a frame shift mutation as well as the deletion; as predicted from our targeting vector. Importantly, the amount of Ap3b1 transcript was not altered in Ap3b2^-/- mice, and likewise, Ap3b1^-/- mice had the normal amount of Ap3b2 transcript (Figure 1C, compare lanes 5 and 8, and 1 and 3, respectively). These results demonstrate that there is no dramatic compensatory increase of either transcript in the absence of the other.

AP-3 Complex Assembly in β3A and β3B Adaptin Deficiencies

We further explored the effects of engineered β3B and β3A deficiencies upon brain β3 expression levels by using a panel of antibodies that selectively recognize either the β3B subunit, both β3 subunits, or AP-3 subunits common to both AP-3 complexes, δ and σ3A-B (Figure 2). Brain homogenates from wild-type (control), β3A- (Ap3b1^-/-), and β3B-deficient mice (Ap3b2^-/-) were probed with monospecific anti-β3B antibodies (Figure 2A, lanes 1–6) or with an antibody recognizing both β3 subunits (Figure 2A, lanes 7–9). β3B adaptin was readily detectable in both control and β3A-deficient (Ap3b1^-/-) mouse brain homogenates (Figure 2A, lanes 1–2 and 4–5). In contrast, both β3B antibodies fail to detect β3B protein in β3B-deficient extracts (Ap3b2^-/-; Figure 2A, lanes 3 and 6). We did not detect compensatory changes of β3B protein expression levels in β3A-deficient brains (Figure 2A, compare lanes 1 and 2, 4 and 5). Importantly, the remaining β3A adaptin present in β3B-deficient mice was detected with the β3 antibody recognizing both β3 subunits (Figure 2A, lane 9); thus, confirming the specific nature of the β3 deficiencies.

To evaluate whether β3B and β3A present in Ap3b1^-/- and Ap3b2^-/- brain tissue were assembled into whole heterotetramers, we tested whether β3B- and β3A-containing AP-3 complexes could be immunoprecipitated with antibodies against δ and σ3 adaptins. These two AP-3 subunits do not directly bind β3 (Peden et al., 2002). However, δ and σ3 adaptins indirectly associate with β3 in the context of holotetramers (Peden et al., 2002). AP-3 complexes containing β3B were immunoprecipitated with σ3 antibodies, from both control and Ap3b1^-/- brain extracts (Figure 2B, lanes 1–2). These complexes were absent from Ap3b2^-/-homogenates (Figure 2B, lane 3). Similarly, δ antibodies immunoprecipitated β3B present in control and Ap3b1^-/- homogenates (Figure 2B, lanes 4–5), demonstrating that the neuronal-specific β3 subunit was incorporated into AP-3 complexes. β3A-containing complexes were still present in β3B-deficient brain extracts (Figure 2B, lane 9). Thus, the lack of either β3A- or β3B-AP-3 does not appreciably perturb the other adaptor isoform. In summary, our results demonstrate that Ap3b1^-/- and Ap3b2^-/- mouse brains selectively lack β3A- and β3B-containing AP-3 complexes, without detectable compensatory changes.

Distinct Phenotypes Distinguish Ap3b1^-/- and Ap3b2^-/- Mice

The absence of ubiquitous and neuronal AP-3 in mocha leads to pigment dilution and neurological disorders such as locomotor hyperactivity (Kantheti et al., 1998). To assess the tissue penetrance of these phenotypes in isoform-specific deficiencies, we analyzed pigmentation and open field activity in Ap3b1^-/- and Ap3b2^-/- mice. As predicted from its selective expression in neurons (Newman et al., 1995), removal of β3B did not affect coat color (Figure 3A). Ap3b2^-/- mice possessed normal black coat color, indistinguishable from the control. In contrast, Ap3b1^-/- mice display a light gray coat color thus confirming that only ubiquitous AP-3 is involved in skin melanosome biogenesis. On the other hand, horizontal locomotor activity in response to handling was significantly increased in Ap3b2^-/-, whereas it was normal in Ap3b1^-/- mice (Figure 3B). During the first 30 min in the open field, Ap3b2^-/- mice exhibited twofold higher locomotion than control (Figure 3B), which gradually receded to nearly control values. Multiple behavioral and electroencephalographic analyses further demonstrated a normal neurological phenotype in Ap3b1^-/- mice but complex neurological and behavioral impairments, including tonic-clonic seizures, in Ap3b2^-/- mice (Seong and Burmeister, unpublished data). These observations indicate that the hyperactivity observed in mocha is due to the absence of neuronal AP-3 complexes. Overall, these results indicate that neuronal and nonneuronal phenotypes are selectively triggered by isoform-specific adaptor deficiencies.

Neuronal and Ubiquitous AP-3 Complexes Possess Distinct Subcellular Distribution in Neurons

Functionally distinct adaptors possess characteristic and unique subcellular distribution both in polarized and nonpolarized cells (Gan et al., 2002; Folsch et al., 2003; Robinson, 2004). Thus, we hypothesized that divergence in the function of ubiquitous (β3A) and neuronal (β3B) AP-3 could be manifested as differential distribution of isoforms within neurons. We took advantage of primary cultures of embryonic neurons derived from Ap3b1^-/- or Ap3b2^-/- mouse forebrains as cellular systems exclusively containing neuronal or ubiquitous AP-3, respectively. These neuronal cultures possess a branched and polarized morphology where subcellular domains can be easily identified by cytoskeletal markers (Caceres et al., 1986).

Primary neuronal cultures were immunostained with δ antibodies to highlight the neuronal AP-3 complexes present in Ap3b1^-/- cells or the ubiquitous AP-3 present in Ap3b2^-/- neurons. We selected a single antibody to avoid differences that could emerge from disparity in antigen–antibody complex formation by using the isoform specific β3 antibodies. Moreover, and in contrast with the δ mAb, both β3 antibodies were unsatisfactory in fixed cells. We first determined the specificity of the δ antibody staining (Peden et al., 2004) by using wild-type and mocha brain primary cultured neurons. Mocha neurons do not express functional δ subunits, making them an ideal control. Delta antibody decorated abundant AP-3 puncta present in neuronal cell bodies as well as in all processes (Figure 4A). In contrast, specific fluorescent signal was ablated in mocha cell bodies and process (Figure 4B), demonstrating the specificity of this antibody.

AP-3 has been described in cell bodies, neuronal processes, and nerve terminals by using a variety of antibodies, neuronal cell types, and species (Darnell et al., 1991; Newman et al., 1995; Simpson et al., 1996; Zakharenko et al., 1999). Thus, we explored whether AP-3 selectively decorated axons and/or dendrites in primary cultured mouse neurons. Cultured neurons develop polarized axons and dendrites, cell domains, which can be distinguished by the presence of MAP-2 in dendrites (Caceres et al., 1986). Neurons were cultured for 19 DIV to allow the development of extensive and interconnected processes (neuropil) and double labeled with δ and MAP-2 antibodies (Figure 4, C–E). AP-3–positive organelles were detected in cell bodies, MAP-2–positive (dendrites) and MAP-2–negative (axons) processes. There were no appreciable differences in the AP-3 content and distribution between axons and dendrites (Figure 4, E1–E3), suggesting that AP-3 was equally segregated between these neuronal domains. To explore whether neuronal and ubiquitous AP-3 were differentially distributed in neurons, we double labeled wild-type, Ap3b1^-/-, and Ap3b2^-/- neurons with δ and MAP-2 antibodies and imaged cells by confocal microscopy (Figure 5). Wild-type and β3B-containing neurons (Ap3b1^-/-, Figure 5, D–F) showed prominent AP-3 labeling in cell bodies, MAP-2–positive and–negative processes. In contrast, neurons lacking neuronal AP-3 (Ap3b2^-/-, Figures 5, G–I, and 6A) consistently showed a dramatically decreased staining of all processes despite the fact that the cell body AP-3 content seemed normal. We confirmed these observations by quantitative analysis of AP-3 fluorescence intensity by using MetaMorph software (Figure 6B). There were no significant differences in fluorescence intensity among cell bodies, irrespective of both their genetic background and the time that cells were differentiated in vitro (DIV) (Figure 6B). In contrast, neuronal AP-3–deficient cells (Ap3b2^-/-) contained 2.6 times less AP-3 in their processes when differentiated for 7 d (p < 0.001 n = 98 wild-type and 95 Ap3b2^-/- processes analyzed). This phenotype was even more pronounced in cells differentiated for 15 d (15 DIV). Ap3b2^-/- neuronal processes contained 4.2-fold less AP-3 than their wild-type counterpart (Figure 6B, p < 0.001 n = 38 wild-type and 56 Ap3b2^-/- processes analyzed). Thus, these results not only confirm our observations in 7 DIV neurons, but they also rule out that the lack of AP-3 in Ap3b2^-/- cell processes is due to a developmental delay in the export of ubiquitous AP-3 to cell processes.

In summary, these results indicate that, at steady state, mature neurons preferentially target neuronal AP-3 to cell processes in a nonpolarized manner. The distinct targeting of neuronal and ubiquitous AP-3 strongly supports the hypothesis that AP-3 isoforms perform divergent functions in neurons.

Presynaptic Mechanisms Are Impaired in Neuronal AP-3–deficient Brain

The presence of neuronal AP-3 in axons (Figures 4, C–E, and 6A), nerve terminals, and its reported function in the biogenesis of synaptic vesicles (Newman et al., 1995; Faundez et al., 1998; Zakharenko et al., 1999; Blumstein et al., 2001; Salazar et al., 2004b) prompted us to explore whether presynaptic mechanisms were affected in Ap3b1^-/- and Ap3b2^-/- brains. To test this hypothesis, we explored the content of histochemically reactive ionic zinc. Ionic zinc is stored exclusively in synaptic vesicles (Frederickson, 1989) by the action synaptic-specific zinc transporter ZnT3 (Palmiter et al., 1996; Cole et al., 1999), an AP-3–interacting molecule (Salazar et al., 2004b). Moreover, ZnT3 (Cole et al., 1999) or AP-3 (mocha) (Kantheti et al., 1998; Kantheti et al., 2003) genetic deficiencies severely compromise synaptic zinc stores, thus making histochemically reactive zinc an unparalleled reporter of synaptic AP-3 function.

We performed Timm's staining in control and mutant mice brain sections to assess the histochemically reactive presynaptic zinc pools. We used mocha brain sections as controls. As previously described, Timm's staining was prominent in discrete cortex layers and in different regions of the hippocampus (Figure 7A) (Frederickson, 1989; Frederickson and Danscher, 1990). Histochemically reactive zinc was substantially reduced in all these regions in mocha brain (Figure 7D), thus validating Timm's staining as a tool to assess neuronal AP-3 synaptic phenotypes. We focused our analysis in the brain cortex and the stratum oriens (Figure 7A, area 3) and radiatum of the hippocampal formation (Figure 7A, area 4). Timm's staining deposits were reduced in all brain areas of Ap3b2^-/- brains, although the phenotype was not as pronounced as in mocha brain. Quantification of the Timm's staining by using MetaMorph software showed that in all areas of Ap3b2^-/- mouse brain there was on average a 25% reduction in Timm's staining intensity (n = 27 brain sections in three independent experiments). The most dramatic differences were observed in the CA1 stratum oriens with 62 ± 5% of the control values found in Ap3b2^-/- brains (n = 12). Surprisingly, in the absence of ubiquitous AP-3 (Ap3b1^-/-), we observed a significant increase of 12% in the histochemically reactive zinc pools (p < 0.045), in the cortex as well as the hippocampus (n = 20), thus supporting the hypothesis that the functions carried out by the ubiquitous and neuronal AP-3 are divergent.

Decreased synaptic ionic zinc content in mocha synapses is associated with reduced ZnT3 protein levels in brain tissue as well as in synaptic vesicles (Salazar et al., 2004b). To explore whether the changes in Timm's staining observed in Ap3b1^-/- and Ap3b2^-/- were paralleled by changes in ZnT3 levels, we determined the ZnT3 content by immunocytochemistry in hippocampal brain sections. As we described previously (Salazar et al., 2004a), there was a drastic reduction in the total ZnT3 protein levels in mocha mouse hippocampus compared with control brain sections (Supplementary Figure 1, compare G and H). In contrast, we did not detect differences in the ZnT3 levels and distribution in Ap3b1^-/- and Ap3b2^-/- hippocampi (Supplementary Figure 1, A–F). Identical results were obtained with serial dilutions of ZnT3 antibody (our unpublished data). This suggests that rather than global changes in ZnT3 protein content, the changes in synaptic zinc observed in Ap3b1^-/- and Ap3b2^-/- hippocampi may represent modifications not detectable by immunocytochemical techniques. These results suggest that in the subcellular distribution of membrane proteins involved in the synaptic vesicle zinc uptake is differentially affected in Ap3b1^-/- and Ap3b2^-/- mouse models.

ZnT3 and ClC-3 Targeting to Small Vesicles Are Differentially Modified in Ap3b1^-/- and Ap3b2^-/- Brains

Synaptic vesicle ionic zinc uptake is critically dependent on two AP-3–trafficked transmembrane proteins, ZnT3 (Cole et al., 1999) and the synaptic vesicle chloride channel ClC-3 (Salazar et al., 2004a). The targeting of both proteins into synaptic vesicles is impaired after pharmacological or genetic (mocha) perturbation of AP-3 function (Salazar et al., 2004a,b). Thus, we hypothesized that the changes in synaptic zinc observed in Ap3b1^-/- and Ap3b2^-/- brain were associated with corresponding modifications in the content of ZnT3 and ClC-3 in Ap3b1^-/- and Ap3b2^-/- synaptic vesicles. To address this question, we analyzed the targeting of these two proteins to synaptic vesicles from control, Ap3b1^-/-, and Ap3b2^-/- brains. Synaptic vesicles can be isolated from high-speed supernatants fractionated in velocity glycerol gradients. Velocity sedimentation discriminates vesicles by size, resolving large membranes from small synaptic vesicles, which sediment as a symmetric peak in the middle of the gradient (fractions 7–9) (Salazar et al., 2004a,b). We monitored synaptic vesicle marker content and its distribution along gradients to assess both targeting and whether these proteins were present in synaptic vesicles, respectively. We selected membrane proteins whose targeting to synaptic vesicles is affected by the mocha allele (ZnT3, ClC-3) and contrasted their behavior to AP-3–independent cargoes (synaptophysin, SV2, and the 116-kDa subunit of the vacuolar ATPase) (Salazar et al., 2004a,b). All of membrane proteins analyzed cosedimented as a symmetric peak around fraction 8 in control brain membranes (Figure 8). On the contrary, the AP-3 cargoes did not cosediment with the AP-3–independent markers in Ap3b2^-/- brains (Figure 8). One-fifth of the ZnT3 protein contained in the membranes resolved by glycerol sedimentation was present at the peak synaptic vesicle fraction from wild-type membranes (fraction 8 = 20.5 ± 1.63%, n = 5), whereas we observed a 40% decrease in Ap3b2^-/- membranes (p < 0.001, n = 5). These results are consistent with the hypothesis that ZnT3 targeting to synaptic vesicles was impaired in the absence of neuronal AP-3. Remarkably, the absence of ubiquitous AP-3 (Ap3b1^-/-) rather than reducing ZnT3 in synaptic vesicles doubled its content (ZnT3 in fraction 8 = 41.2 ± 15.4%, p < 0.001, n = 5). The increased content of an AP-3 cargo in Ap3b1^-/- synaptic vesicles was not restricted to ZnT3. Similarly, we observed a 1.8-fold increase in the ClC-3 synaptic vesicle content in ubiquitous AP-3–deficient Ap3b1^-/- membranes (control 19.2 ± 3.5%, Ap3b1^-/- 34.6 ± 5.2%, n = 4, p < 0.05). ClC-3 targeting to small vesicles was also altered in the absence of neuronal AP-3. ClC-3 content in Ap3b2^-/- small vesicles was increased, although the membranes containing ClC-3 migrated into the gradient faster than synaptic vesicles (fraction 6). In wild-type membranes, only 9.9 ± 2.3% (n = 4) of the total ClC-3 contained in the overall gradient is present in fraction 6, yet in neuronal AP-3 deficiency this fraction contains 33.1 ± 5.5% (n = 4, p < 0.01), providing evidence of defective ClC-3 targeting in Ap3b2^-/- neurons. The changes in the content and distribution of AP-3 cargoes (ZnT3 and ClC-3) observed in the isotype-selective AP-3 deficiencies were specific. The distribution and synaptic vesicle content of the AP-3–independent cargoes, synaptophysin (Sphysin), SV2, and the vacuolar ATPase subunit remained unaffected in Ap3b1^-/- and Ap3b2^-/- brains.

Collectively, these data are consistent with the hypothesis that selective defects in AP-3 cargo targeting contribute to the synaptic zinc phenotypes observed in Ap3b1^-/- and Ap3b2^-/- brains. Moreover, they indicate that the functions of the neuronal and ubiquitous AP-3 complexes in neurons are divergent, despite the fact that both regulate synaptic vesicle protein composition.

DISCUSSION

Neurons possess AP-3 adaptors carrying either β3A or β3B (Newman et al., 1995; Dell'Angelica et al., 1997b). Absence of both β3 subunits, such as in the mocha allele (Kantheti et al., 1998), leads to neurological manifestations, a phenotype absent in mice deficient just in β3A (pearl) (Feng et al., 1999). An attractive hypothesis to explain this phenotypic difference is that β3B-AP-3 performs a specific role in neuronal physiology. However, mocha and pearl alleles do not discriminate whether β3B-containing AP-3 complexes perform functions unique to neurons, or whether β3A- and β3B-containing AP-3 complexes are functionally interchangeable in neuronal tissue. We addressed this question by using mouse strains selectively lacking the ubiquitous AP-3 subunit β3A (Ap3b1^-/-) or the neuronal-specific β3 subunit β3B (Ap3b2^-/-). Herein, we demonstrate that AP-3 complexes assembled with β3A or β3B perform preferentially distinct and divergent roles in neuronal-specific sorting and vesiculation mechanisms. Four lines of evidence support our conclusion. First, we have successfully segregated the pigment dilution and hyperactivity phenotypes observed in the mocha mouse (Kantheti et al., 1998) with deficiencies of the ubiquitous and neuronal-specific AP-3, respectively. Second, β3A- and β3B-containing AP-3 complexes are differentially distributed in neuronal cell bodies and processes of cultured neurons. Ubiquitous and neuronal AP-3 distribute similarly in neuronal cell bodies. Nevertheless, β3B-containing AP-3 is approximately fourfold more abundant than β3A-AP-3 in axons and dendrites. Third, histochemically reactive synaptic vesicle zinc is selectively decreased in β3B deficiency, yet it increases in β3A-null cells. Finally, targeting of two AP-3 synaptic vesicle cargo proteins, ZnT3 and ClC-3, is selectively impaired in β3B-deficient neurons. In contrast, ZnT3 and ClC-3 content are increased at least twofold in synaptic vesicle fractions from β3A-deficent neurons.

Interpretation of the neuronal phenotypes in Ap3b1^-/- and Ap3b2^-/- rests on the assumption that the changes induced by each genetic ablation are constrained to just one AP-3 isoform. Our data provide evidence supporting this notion. Neuronal AP-3 present in Ap3b1^-/- mouse brain is assembled into complexes containing δ and σ3 subunits. These heterotetramers can be recruited to membranes because β3B-AP-3 is clearly discernible in puncta rather than the cytoplasm of neurons. This line of thought is further supported by the increased content of synaptic vesicle specific AP-3 cargoes (ZnT3 and ClC-3) in small vesicle populations from Ap3b1^-/- neurons. Thus, based on the quaternary structure, presence on membranes, and effects upon synaptic vesicle protein targeting, neuronal complexes present on the Ap3b1^-/- neurons are capable of sorting and vesiculation. Together, these results support the hypothesis that specialized sorting and vesiculation mechanisms fulfilled by neuronal AP-3 cannot be efficiently replaced by β3A-AP-3 adaptors.

Histochemically reactive ionic zinc detected with Timm's staining mainly reports zinc pools present in synaptic vesicles (Frederickson, 1989). This vesicular pool requires the presence of a synaptic vesicle-specific transporter, ZnT3 (Palmiter et al., 1996), a membrane protein trafficked to synaptic vesicles by AP-3 (Kantheti et al., 1998; Salazar et al., 2004b). ZnT3^-/- brains completely lack histochemically reactive vesicular zinc (Cole et al., 1999), thus demonstrating that this transporter is central to generate synaptic zinc pools. Although the decreased Timm's staining in β3B null as well as mocha brain was predicted by our hypothesis, the increased level of synaptic zinc in β3A-deficient brain was a revelation. This change in synaptic zinc likely is mechanistically linked with the concomitant increment in the synaptic vesicle ZnT3 and ClC-3 content. Both membrane proteins control vesicular zinc uptake (Cole et al., 1999; Salazar et al., 2004a). How can β3A and β3B deficiencies trigger opposing synaptic zinc phenotypes? In nonneuronal cells, ClC-3 targeting to late endosome/lysosome compartments is sensitive to the mocha allele (Salazar et al., 2004a), thus indicating that ClC-3 is recognized by ubiquitous β3A-containing AP-3. Therefore, if β3A-containing adaptors are incapable of distinguishing synaptic vesicle from nonsynaptic vesicle cargoes, then a reasonable hypothesis to explain the phenotypes observed in Ap3b1^-/- mouse brain is that both AP-3 isoforms compete for synaptic vesicle membrane proteins (Figure 9). Thus, ZnT3 or ClC-3 may have two pathways to follow. Either they would be included in a β3A-AP3–generated vesicle in route to late endosomal compartments, or they would be sorted into a β3B-AP-3–generated vesicle to be delivered to synaptic vesicle pools (Figure 9). These divergent fates could be particularly prominent in neuronal cell bodies where β3A and β3B are abundantly represented, but less prominent in axons and dendrites where neuronal AP-3 is the predominant adaptor isoform (Figure 9). This model is consistent with the phenotypes observed in Ap3b1^-/-, Ap3b2^-/-, and mocha mice.

A puzzling observation is why ZnT3 and ClC-3 phenotypes are dissimilar in the Ap3b2^-/- background, with ClC-3 being targeted to larger membranes and its levels increased whereas ZnT3 levels are decreased, yet the transporter is present in normal size vesicles. A possible explanation may derive from the fact that in neurons ClC-3 is targeted to synaptic vesicles and likely to lysosomes (Stobrawa et al., 2001; Salazar et al., 2004a), whereas in nonneuronal cells it is routed to lysosomes (Li et al., 2002). In contrast, ZnT3 has been reported only in synaptic vesicles (Wenzel et al., 1997). Thus, in the absence of neuronal AP-3, ClC-3 may accumulate in organelles distinct to those where ZnT3 may be present. To properly answer this question, additional factors need to be considered, such as the participation of the other neuronal AP-3 isoform μ3B (Blumstein et al., 2001), and the possibility that hierarchical and differential interactions may occur between other nonAP-3 adaptors, multiple sorting signals present in cargo proteins, and post-translational modifications affecting the sorting of some of these cargoes (Salazar and Faundez, unpublished observations).

An intriguing question is how β3 isoforms contribute to selective sorting and vesiculation. Tyrosine-based sorting signal recognition is carried out by μ3 subunits (Bonifacino and Traub, 2003), whereas the dileucine sorting signal is recognized by δ-σ3 (Janvier et al., 2003). μ3A and μ3B preferentially bind tyrosine-based sorting motifs (Ohno et al., 1998), suggesting that this subunit could contribute to the sorting selectivity of the neuronal and ubiquitous AP-3. Because β3 subunits assemble into β3/μ3 dimers (Peden et al., 2002), selective sorting of synaptic vesicle cargo could result from a preferential association of β3B/μ3B. However, two-hybrid analysis indicates that μ3B bind β3A but not β3B (Peden et al., 2002). Alternatively, the hinge-ear domains of both β3 subunits could play a role in recruiting accessory molecules or cargo-specific modular adaptors as in the case of β-arrestin (Laporte et al., 2000; Kim and Benovic, 2002; Laporte et al., 2002) or ARH (He et al., 2002; Mishra et al., 2002) in AP-2–mediated vesiculation. The relevance of these β3 domains is corroborated by β3A chimeras where the ear domain of β3A is replaced by the β2 adaptin ear. This hybrid adaptor complex is unable to rescue AP-3–deficient phenotypes (Peden et al., 2002). A complementary mechanism to explain β3A- or β3B-differences in cargo recognition and vesiculation could rely on the adaptors being in different domains of the neuron. This possibility is particularly attractive in light of the differential distribution of neuronal and ubiquitous AP-3 among cell bodies and axo-dendritic compartments. Interestingly, the AP-1 β1 subunit ear domain interacts with the motor kif13A modulating subcellular distribution and sorting function of AP-1 (Nakagawa et al., 2000). Although direct interactions of AP-3 with cytoskeletal motors have not been reported, recent evidence suggests that intermediate filaments regulate AP-3 function (Styers et al., 2004) and that AP-3 might regulate positioning of organelles in cytotoxic lymphocytes (Clark et al., 2003). Thus, a viable yet not exclusive mechanism to account for specialized functions of neuronal AP-3 may in part derive from spatial segregation of the adaptor complexes into distinct neuronal domains.

Taken collectively, our data indicate that in neurons ubiquitous and neuronal AP-3 complexes participate in distinct and divergent sorting and vesiculation processes. We propose that concerted opposing functions of neuronal and ubiquitous AP-3 provide a mechanism to control synaptic vesicle protein composition.

While this article was under review, Nakatsu et al. (2004) reported the characterization of an Ap3m2 knockout mouse, which is also hyperactive and prone to seizures. Nakatsu et al. (2004) also reported electrophysiological changes attributed to missorting of the vesicular GABA transporter. Whether Ap3m2 and Ap3b2 knockout mice are otherwise similar in behavior or in cellular distribution of AP-3 complexes remains to be seen.

FOOTNOTES

FOOTNOTES

ACKNOWLEDGMENTS