Abstract
In the olfactory epithelium (OE), odorant receptor stimulation generates cAMP signals that function in both odor detection and the regulation of axon guidance molecule expression. The enzyme that synthesizes cAMP, adenylyl cyclase 3 (AC3), is coexpressed in olfactory sensory neurons (OSNs) with poly-N-acetyllactosamine (PLN) oligosaccharides determined by the glycosyltransferase β3GnT2. The loss of either enzyme results in similar defects in olfactory bulb (OB) innervation and OSN survival, suggesting that glycosylation may be important for AC3 function. We show here that AC3 is extensively modified with N-linked PLN, which is essential for AC3 activity and localization. On Western blots, AC3 from the wild-type OE migrates diffusely as a heavily glycosylated 200 kDa band that interacts with the PLN-binding lectin LEA. AC3 from the β3GnT2−/− OE loses these PLN modifications, migrating instead as a 140 kDa glycoprotein. Furthermore, basal and forskolin-stimulated cAMP production is reduced 80–90% in the β3GnT2−/− OE. Although AC3 traffics normally to null OSN cilia, it is absent from axon projections that aberrantly target the OB. The cAMP-dependent guidance receptor neuropilin-1 is also lost from β3GnT2−/− OSNs and axons, while semaphorin-3A ligand expression is upregulated. In addition, kirrel2, a mosaically expressed adhesion molecule that functions in axon sorting, is absent from β3GnT2−/− OB projections. These results demonstrate that PLN glycans are essential in OSNs for proper AC3 localization and function. We propose that the loss of cAMP-dependent guidance cues is also a critical factor in the severe axon guidance defects observed in β3GnT2−/− mice.
Introduction
The orderly convergence of sensory axons to the olfactory bulb (OB) is critical for odor discrimination and processing by the brain. Although the populations of olfactory sensory neurons (OSNs) expressing a given odorant receptor (OR) are widely dispersed in the nasal cavity, their axons converge to form synapses in glomeruli at relatively stereotyped loci in the OB. A critical role for ORs in olfactory axon guidance has long been appreciated, although the mechanism by which they influence targeting has been elusive (Mombaerts et al., 1996; Wang et al., 1998; Feinstein and Mombaerts, 2004). Recently, several studies have convincingly shown that OR-derived cAMP signals are essential for proper olfactory map formation (Imai et al., 2006, 2009; Chesler et al., 2007).
Olfactory signaling proteins are concentrated in specialized cilia that project from OSNs into the nasal lumen (for review, see Jenkins et al., 2009). Odorant binding to ORs activates the Gαs homolog Golf, which stimulates cAMP production by adenylyl cyclase 3 (AC3). Increased cAMP levels activate a heterotetrameric olfactory cyclic nucleotide gated channel leading to membrane depolarization and the propagation of action potentials (Kleene, 2008) (for review, see Kaupp, 2010). Blocking OR stimulation of AC3 in transgenic mice by disrupting G protein coupling results in a failure of axon convergence (Imai et al. 2006).
A number of molecules associated with cell adhesion and axon guidance are differentially regulated by cAMP (Imai et al., 2006; 2009). Nrp1 is mosaically expressed by subsets of OSNs in the olfactory epithelium (OE) and is required for proper glomerular targeting through interactions with semaphorin-3A (Sema3A) (Schwarting et al., 2000, 2004; Taniguchi et al., 2003). Higher Nrp1 expression is positively correlated with increased cAMP levels, and either downregulating or upregulating Nrp1 in transgenic mice shifts the position of glomeruli toward the anterior or posterior OB, respectively (Imai et al., 2006, 2009). These results are further supported by analysis of AC3 null mice, which exhibit a loss of Nrp1 expression, severely disorganized glomeruli, and anosmia (Wong et al., 2000; Trinh and Storm, 2003; Chesler et al., 2007; Dal Col et al., 2007, Zou et al., 2007).
Interestingly, the effects of AC3 loss mirror the defects we observed in null mice for β3GnT2 (Zhou et al., 1999; Shiraishi et al., 2001), a glycosyltransferase highly expressed by OSNs and other sensory neuron populations (Henion et al., 2005; Schwarting and Henion, 2007). β3GnT2−/− mice display a loss of selected OSN subsets and glomeruli, a severe axon pathfinding defect, and impaired sexual behavior (Henion et al., 2005; Biellmann et al., 2008). Here we investigate the relationship between β3GnT2 and olfactory signaling. We show that β3GnT2 has a novel function in regulating the expression of poly-N-acetyllactosamine (PLN) glycans on AC3 that are required for cAMP synthesis. The loss of AC3 activity in β3GnT2−/− mice leads to a misregulation of both Nrp1 and Sema3A expression. Our results suggest that PLN glycans are required for maintaining cAMP signaling in olfactory and perhaps other sensory neuron populations.
Materials and Methods
Animals.
β3GnT2−/− mice were established from the gene-trapped KST308 embryonic stem (ES) cell line, developed through the BayGenomics project (Mitchell et al., 2001). Details of the ES insertion event and generation of β3GnT2−/− mice have been reported previously (Henion et al., 2005). Mice were housed according to standard National Institutes of Health and institutional care guidelines, and procedures were approved by the University of Massachusetts Medical School Institutional Animal Care and Use Committee (Worcester, MA).
Antibodies and lectins.
LEA (Lycopersicon esculentum) lectin, biotinylated and Texas Red and FITC conjugated, were obtained from Vector Laboratories. Antibodies used included rabbit anti-AC3 (Santa Cruz Biotechnology), rabbit anti-activated caspase-3 (Cell Signaling Technology), goat anti-actin (Santa Cruz Biotechnology), goat anti-kirrel2 (R & D Systems), mouse anti-neural cell adhesion molecule (NCAM; Sigma-Aldrich), rabbit anti-neuropilin-1 (R & D Systems), and rabbit anti-plexin-A4 (Santa Cruz Biotechnology). Mouse anti-GAP-43 was a gift from Karina Meiri, Tufts University, Boston, MA. Rabbit anti-OCAM was a gift from K. Mori, University of Tokyo, Tokyo, Japan. Goat anti-olfactory marker protein (OMP) was a gift from Frank Margolis, University of Maryland, Baltimore, MD.
Histology and immunocytochemistry.
For most immunocytochemical procedures, tissues were prepared by transcardial perfusion using 4% paraformaldehyde fixation in 0.1 m PBS, pH 7.4. For Nrp1 labeling, tissues were fixed in 2% paraformaldehyde-lysine-periodate. Heads were subsequently removed and postfixed overnight in the same fixative solution, followed by cryoprotection in 30% sucrose. After embedding, tissue sections were prepared on a Microm HM505E cryostat at 50 μm thickness and then immediately thawed in Transwell boats filled with PBS for staining as free-floating sections. Tissues were blocked for 1 h in 2% BSA and then incubated overnight at 4°C with primary antibodies diluted in 1% BSA/PBS/0.3% Triton X-100. After washing, tissue sections were further incubated for 2 h at room temperature with species-specific secondary antibodies conjugated to either Alexa Fluor 488, Alexa Fluor 568 (1:1000; Invitrogen Corporation), or Cy3 (1:300; Jackson ImmunoResearch Laboratories). Images were captured using either a Leica SP1 laser scanning confocal microscope, or a Zeiss Axioplan photomicroscope equipped with a Spot RT camera (Diagnostic Instruments).
Cell counts.
AC3-expressing OSNs in postnatal day (PD)3 and adult OEs of β3GnT2−/− and wild-type (WT) control mice were quantified from digital images of the dorsal septum. Three mice at each time point were analyzed using five images captured per mouse from sections located near the midpoint OE along the anterior–posterior axis. The number of immunoreactive OSNs was counted for three boxes digitally placed on each image, and the results expressed as the average number of AC3+ OSNs per 50 μm segment of OE. Statistical analysis was performed using SigmaStat 2.0 software.
Biochemical analyses.
Olfactory epithelia from adult mice between the ages of 6 and 10 weeks of age were carefully removed under a dissecting microscope and homogenized in PBS containing 1% Triton X-100 and a protease inhibitor cocktail (Sigma-Aldrich). OE preparations are comprised mainly of OSNs and other OE cell types, but also contain an adjacent extracellular matrix that includes olfactory axons and blood vessels. Protein concentrations were determined colorimetrically using a BCA protein assay kit (Pierce Biotechnology). Proteins were separated by gradient (4–15%) SDS-PAGE under reducing conditions and transblotted to 0.45 μm nitrocellulose filters. The blots were exposed to primary antibody overnight at 4°C and then to horseradish peroxidase-conjugated secondary antibodies (1:500) for 2 h at room temperature. Immunoreactive bands were visualized using an Opti-4CN substrate kit (Bio-Rad). For enzyme treatment of proteins, 5 μl denaturing solution (0.2% SDS and 100 mm 2-mercaptoethanol) was added to ∼100 μg of protein in 50 μl of PBS and heated at 100°C for 10 min. Triton X-100 concentration was adjusted to 1.25%, and 5 U of peptide N-glycosidase (PNGase F) (Sigma-Aldrich) was added and incubated at 37°C for 3 h. The reaction was stopped by heating at 100°C for 5 min. Deglycosylation was assessed by change in mobility by SDS-PAGE.
Affinity purification of PLN-modified glycoproteins.
To fractionate olfactory glycoproteins, freshly dissected OEs were homogenized in 0.1 m Tris-buffered saline, pH 7.5, containing 1.0 mm CaCl2, 1.0 mm MgCl2, 0.5% Nonidet P-40, 5 mm sodium deoxycholate, and protease inhibitor cocktail (Sigma-Aldrich). The homogenate was centrifuged at 12,000 × g for 10 min and the protein concentration was adjusted to 6–10 mg/ml. LEA-agarose (200 μl) was equilibrated in homogenization buffer, and then 2 ml of protein was added followed by end-over-end mixing for 4 h at room temperature. The agarose gel was spun at low speed followed by removal of the supernatant, and the gel was washed three times in homogenization buffer without detergent. The LEA-agarose gel was then mixed end over end at room temperature in TBS containing 0.3 m lactose and protease inhibitors. The resulting supernatant was concentrated to 100 μl using a Centicon-10 spin column (Millipore) for Western blot analysis.
Real time reverse transcriptase quantitative PCR.
OEs from 3- to 6-month-old WT and β3GnT2−/− mice (n = 3 for each age and genotype) were microdissected directly into TRIzol reagent (Invitrogen). Total RNA was isolated individually from each sample according to the manufacturer's protocol, and 3 μg of each RNA sample was reverse transcribed into cDNA with random hexamers using the SuperScript II RT System (Invitrogen). Oligonucleotides for quantitative PCR (qPCR) amplification were designed using Primer3 software (version 0.4.0) and are available upon request. qPCRs were set up in triplicate with GoTaq PCR Master Mix (Promega) for amplification on a StepOnePlus Real-Time PCR System (Applied Biosystems). Relative expression levels, normalized to RNA polymerase 2, were determined from comparative CT (threshold cycle) values calculated using StepOne Real-Time PCR Software.
Adenylyl cyclase enzyme assays and cAMP immunoassay.
For the analysis of adenylyl cyclase enzymatic activity, OE tissues from PD9 and adult β3GnT2 WT and null mice were carefully microdissected and placed into 0.5 ml of chilled homogenization buffer containing 50 mm Tris-HCl, pH 7.4, 2 mm MgCl2, 1 mm EDTA, 0.5 mm DTT, and 1× protease inhibitor cocktail (Roche). Samples were ground by hand using 20 strokes with a Potter–Elverhejam tissue homogenizer (Kontes), followed by a 7 min centrifugation step at 800 × g to pellet debris. Supernatants were further centrifuged at 120,000 × g in a Beckman Ultracentrifuge using a SW55Ti rotor. Pellets were resuspended in 0.25 ml homogenization buffer with protease inhibitors and were subsequently ground again by hand using 10 strokes in a tissue homogenizer. Samples were immediately frozen on dry ice and stored at −80°C. Protein concentrations were determined by the BCA Protein Assay Kit (Pierce). For assaying adenylyl cyclase activity, 7 μg of olfactory homogenate was preincubated on ice in 25 μl of homogenization buffer supplemented with 0.5 mm 1-methyl-3-isobutylxanthine to inhibit phosphodiesterase activity. Olfactory homogenates were then supplemented with 100 μl of reaction mix containing 40 mm Tris, pH 7.4, 5 mm MgCl2, 1 mm ATP, 10 mm phosphocreatine, 40 U/ml creatine phosphokinase, 1 mm DTT, 1 mm EDTA, 0.2 mm EGTA, 0.1% BSA, and 10 μm GTP. In addition, stimulated samples were coincubated with 10 μm forskolin, while basal samples received DMSO solvent alone. Reactions were incubated at 37°C for 15 min and then were stopped by boiling for 10 min. The reaction products were homogenized in 7.5% ice-cold TCA, pelleted at 2000 × g for 10 min, extracted four times in water-saturated diethyl ether, and dried in a Speed Vac. The amount of cAMP generated in each reaction was quantified by competitive immunoassay using the cAMP Enzyme Immunoassay Kit according to the manufacturer's protocol (CA201, Sigma-Aldrich). Three WT and null mouse OE samples were assayed in duplicate. The immunoassay was repeated three times and the results, expressed as picomoles of cAMP generated per milligram per minute, were averaged for each genotype. The accumulation of cAMP in OE samples at time 0 was negligible and was omitted from further analysis.
In situ hybridization.
Tissue sections for in situ hybridization were fixed in 4% paraformaldehyde as described (see above, Histology and immunocytochemistry) and then sectioned at 14–20 μm thickness before thaw mounting on Superfrost Plus slides (Fisher Scientific). Riboprobes were transcribed with SP6 or T7 polymerase from linearized cDNAs in the presence of digoxigenin labeling mix (Roche). Tissue sections were hybridized to antisense riboprobes and localized with alkaline phosphatase-conjugated anti-digoxigenin antibody (Roche), as detailed previously (Henion et al., 2001). The color reaction was developed with NBT/BCIP (nitro blue tetrazolium and 5-bromo-4-chloro-3-indoyl phosphate) substrate (Roche).
Results
PLN expression by olfactory glycoproteins
We have previously reported that the glycosyltransferase β3GnT2 (formerly termed β3GnT1 under a prior nomenclature by Henion et al., 2005) is required in mice for establishing proper axon connectivity with the OB. The loss of β3GnT2 in OSNs leads to a postnatal delay in OB innervation, axon guidance errors, and multiple abnormalities in glomerular formation that affect all OSN subsets. Terminal N-acetyllactosamine moieties recognized by the monoclonal antibody 1B2 are one form of glycan that is decreased in subpopulations of OSNs in β3GnT2−/− mice. The widespread nature of the axon guidance defects we observed in null mice as well as the broad distribution of β3GnT2 in the OE led us to examine whether other glycan structures were affected by the absence of this glycosyltransferase.
β3GnT2 both initiates N-acetyllactosamine synthesis and extends these structures into large poly-N-acetyllactosamine chains of variable length (Fig. 1A). These oligosaccharides can be identified in tissues by the lectin Lycopersicon esculentum agglutinin, LEA, which binds with high affinity to PLN glycans bearing three or more linear N-acetyllactosamine repeats (Merkle and Cummings, 1987). Beginning in embryonic development, mature OSNs strongly bind LEA, as do their axons in the underlying lamina propria and OB (Fig. 1B,D). However, PLN glycans are absent from OSN cell bodies and OB projections of β3GnT2−/− mice (Fig. 1C,E). These results are consistent with the known activity of β3GnT2 in PLN synthesis (Shiraishi et al., 2001; Zhou et al., 1999; Togayachi et al., 2007) and with the fact that this glycosyltransferase is the only β3GnT family member that is expressed at high levels in the OE (data not shown).
Proper OB targeting is dependent on several guidance molecules expressed on axons whose transcription is regulated by cAMP levels determined by AC3 (Imai et al., 2006, 2009; Chesler et al., 2007; Dal Col et al., 2007; Zou et al. 2007). Similarities between the guidance defects observed in AC3−/− and β3GnT2−/− mice prompted us to investigate whether PLN might influence AC3 expression and/or function. For this analysis, we first examined whether AC3 and other olfactory guidance receptors were PLN modified. Homogenates from adult OE were first separated into bound and unbound fractions on LEA-agarose before being subjected to Western blot analysis. Greater than 95% of total OE protein separates into the LEA unbound fraction (data not shown). The subset of LEA bound proteins is visible as an extended smear composed of glycoproteins of widely varying molecular weights (Fig. 1F).
Prior studies have shown that AC3 expressed by OSNs is a heavily N-glycosylated protein that migrates diffusely as a high molecular weight band (Bakalyar and Reed, 1990; Wei et al., 1998). Differences in PLN chain length and composition could be one source of this size heterogeneity. Consistent with this, virtually all AC3 protein is found in the LEA bound fraction, indicating the presence of extensive PLN modifications (Fig. 1F). The alternatively spliced NCAM-140 and NCAM-180 isoforms express PLN at different levels. A portion of NCAM-180 binds LEA while NCAM-140 is unbound, suggesting that the six potential extracellular N-linked glycosylation sites shared by both isoforms are differentially modified.
We identified several additional OSN glycoproteins that are PLN modified. Kirrel2, a 74 kDa adhesion molecule implicated in homotypic axon sorting (Serizawa et al., 2006), migrates predominantly as a broad 100 kDa band that is retained on LEA-agarose (Fig. 1F). In contrast, Nrp1, the glycoprotein receptor for the secreted guidance cue Sema3A, does not bind to LEA-agarose, although a portion of its signaling coreceptor, plexin-A4, is retained. Thus, several adhesion and signaling molecules that influence olfactory axon targeting are modified with PLN glycans.
β3GnT2 modifies AC3 with PLN
We next investigated the effect of β3GnT2 loss on AC3 glycosylation and expression. In the OE of β3GnT2 WT mice, the spectrum of LEA-reactive glycoproteins is visible on Western blots as a diffuse high molecular weight smear. In contrast, LEA+ glycans are absent from the OE of β3GnT2−/− mice, confirming that PLN synthesis in OSNs is exclusively β3GnT2 dependent (Fig. 2A, lanes 1, 2). The most heavily glycosylated form of AC3 runs diffusely at an unusually high molecular weight relative to its 129 kDa polypeptide size. This low mobility may reflect the aberrant migration of highly glycosylated proteins on polyacrylamide gels due to poor binding of SDS to neutral oligosaccharides (Leach et al., 1980). Consistent with this, AC3 from the null OE runs at 140 kDa, much closer to its predicted molecular mass, as a result of the loss of PLN from residual core glycan chains (Fig. 2A, lanes 3, 4). In addition, AC3 protein is decreased in the null OE, which could reflect an effect of PLN on AC3 expression or the known loss of mature OSNs in β3GnT2−/− mice (Henion et al., 2005).Together, these data show that PLN is a major constituent of AC3 glycosylation that may be required for AC3 stability or neuronal survival.
Other olfactory glycoproteins are not as significantly altered in β3GnT2−/− mice. In the rat OE, NCAM is modified with lactosamine-based carbohydrates that have been suggested to influence axon sorting (Storan et al., 2004). The NCAM-180 isoform that is partially retained on LEA-agarose (Fig. 1F) is only moderately reduced in expression, whereas NCAM-140 is unchanged (Fig. 2A, lanes 5, 6). Neither NCAM isoform shifts position significantly on gels, indicating that PLN is a minor component of NCAM glycosylation.
To examine the nature of AC3 glycosylation in more detail, we incubated OE homogenates from WT and β3GnT2−/− OEs with PNGase F to remove N-glycans and analyzed AC3 migration by Western blotting. PNGase F treatment of WT OE lysates restored the mobility of most high molecular weight AC3 to its calculated molecular mass near 129 kDa (Fig. 2B, lanes 1, 2). Incubation of β3GnT2−/− OEs with PNGase F removes the residual N-glycans from the 140 kDa AC3 glycoform, also returning its mobility to that expected of the native polypeptide (Fig. 2B, lanes 3, 4). The PNGase F sensitivity of these glycans indicates that the heterogeneous migration of AC3 results from PLN modifications that are N-linked and not from mucin-type O-glycans. Furthermore, AC3 from β3GnT2−/− OEs still contains residual N-linked glycans that may be required for trafficking through the endoplasmic reticulum and Golgi and transport to the plasma membrane.
AC3+ OSNs are specifically lost from β3GnT2−/− OEs
We next immunolabeled AC3 neurons in WT and β3GnT2−/− OEs to investigate how PLN loss affects OSN survival. In PD3 WT OEs, AC3 accumulates in the apical compartment of OSN cell bodies and is also highly concentrated in sensory cilia, forming a dense mat of labeling at the luminal surface of the nasal cavity (Fig. 2C). In the null OE, there is a significant decrease in the number of AC3+ neurons, although the level of AC3 expression per cell appears similar to that in WTs (Fig. 2D). The decrease in AC3+ OSNs is further supported by a similar loss of AC3-expressing neurons in null OEs by in situ hybridization (Fig. 2C′,D′). Likewise, the number of neurons expressing OMP, a marker of functionally mature OSNs, is reduced to a similar degree to that of AC3 (Fig. 2E,F). Conversely, immature OSNs expressing the marker GAP-43 are moderately increased in the nulls (Fig. 2E,F), while NeuroD1-expressing neural precursors appeared unchanged (Fig. 2E′,F′). Thus, the effects of β3GnT2 ablation appear specific to differentiated neurons and are not a downstream consequence of OSN precursor loss.
The decrease in AC3+ neurons persists in adult mice. In WT animals, the number of AC3+ neurons is greatly expanded relative to PD3, as the OE becomes increasingly composed of differentiated OSNs (Fig. 2G). In adult β3GnT2−/− mice there is also a significant increase in the number of AC3+ OSNs compared to PD3, although this number is still reduced relative to WT littermates (Fig. 2H). Quantification of these decreases reveals that AC3+ OSNs in PD3 nulls are reduced 58% compared to WT controls (n = 4, p < 0.001), a ratio that is maintained in adults (Fig. 2I). Relative expression levels for AC3, OMP, and GAP-43, determined by qPCR, further confirm that it is the mature OSN population that is affected by β3GnT2 loss in adults (Fig. 2J).
AC3 activity is significantly decreased in β3GnT2−/− mice
To investigate the potential effects of PLN loss on cAMP generation, we examined basal and forskolin-stimulated adenylyl cyclase activity in WT and β3GnT2−/− OEs. Despite the fact that nearly half of AC3+ OSNs persist in null mice, adenylyl cyclase enzymatic activity was extremely low (Fig. 3A). The basal level of cAMP production was ∼6-fold higher in the PD9 WT OE (cAMP generated in pmol/mg/min ± SE; WT, 50.13 ± 5.2; null, 8.2 ± 0.4). Forskolin stimulation elevated adenylyl cyclase activity over 5-fold in the null, although the amount of cAMP generated was still <20% that of WT littermates (WT, 290 ± 30; null, 52.7 ± 7.1). Adenylyl cyclase activity in adult null mice, although higher than at PD9, was still extremely low, even with forskolin stimulation (WT, 876.7 ± 141; null, 75.8 ± 15.6). Heterozygous mice displayed consistently reduced adenylyl cyclase activity compared to WT controls, although this difference was not significant. Together, these results suggest that AC3+ OSNs that persist in the OE of mutant mice are deficient in their ability to generate cAMP.
The reason for AC3 activity loss does not appear to involve defective trafficking of AC3 to cilia. High-magnification en face scans across the surface of the OE reveals apparently normal localization of AC3 protein to the multiple cilia proximal segments that surround the dendritic knob (Fig. 3B,C). Additionally, there is no obvious decrease in labeling intensities in either cilia or OSN cell bodies, except for the lower number of labeled OSNs in the OE of null mice. It thus appears that PLN glycans influence AC3 enzyme activity without affecting protein expression or trafficking to cilia.
AC3 is absent from early postnatal β3GnT2−/− OB projections
In addition to olfactory signal transduction, cAMP levels influence the site of glomerular synapse formation in the OB by regulating transcription of axon guidance molecules (Imai et al., 2006; 2009). Recently, it has become clear that AC3 protein is expressed not only in OSN cell bodies and cilia but is also present on axons (Dal Col et al., 2007; Zou et al., 2007). In PD3 WT mice, AC3 colocalizes completely with the axonal marker NCAM in all glomeruli (Fig. 4A–C). Importantly, in β3GnT2−/− mice where glomerular formation is delayed (Henion et al., 2005), AC3 is absent from olfactory axons. Even residual fibers within the nerve layer that are strongly labeled with NCAM fail to express AC3 (Fig. 4D–F).
To visualize this loss more closely, OBs were immunolabeled with the axonal marker OCAM at PD7, after glomeruli have emerged in WT (Fig. 4G–I), but not β3GnT2−/− OBs (Henion et al., 2005). Despite the delay in glomerular formation, many OCAM-expressing axons are still present in the nerve layer of mutant mice. However, unlike their WT littermates, these axons fail to express AC3 (Fig. 4K–M). There is also a dramatic increase in activated caspase-3 reactivity associated with degenerating olfactory axons in the mutant nerve layer (Fig. 4, compare J, N). Thus, the elimination of many OSNs occurs at a late stage of development, concurrent with or shortly after axons attempt to form synaptic contacts. OSN cell death is one consequence of the loss of trophic support from the absence of OB connectivity (Carr and Farbman, 1992). These results are also consistent with the finding that some OR reporter-labeled axons initially form protoglomeruli in the β3GnT2−/− OB but are subsequently eliminated in adults (Henion et al., 2005).
Although AC3 null mice do not exhibit electroolfactogram responses to odorants, their axons still exit the nerve layer and form rudimentary glomeruli (Wong et al., 2000, Trinh and Storm, 2003; Dal Col et al., 2007; Zou et al., 2007). β3GnT2−/− mice display a similarly disorganized glomerular layer in the absence of AC3 activity that persists in adults (Henion et al., 2005; Schwarting and Henion, 2007). The ability of null mice to discriminate AC3-dependent odors is also moderately reduced relative to WT controls, although they are not anosmic (Henion et al., 2005; T. K. Knott, G. A. Schwarting, unpublished data). In adults, after an initial early postnatal delay, AC3 is eventually expressed on null olfactory axons within the nerve layer and glomeruli, although the levels remain below those of WT controls (Fig. 4O,P). Activated caspase-3 expression also remains elevated in the adult nerve layer (Fig. 4Q,R) and OE (data not shown), indicating that cell death is an ongoing process in the β3GnT2−/− OE, where AC3 activity is severely decreased despite the presence of AC3 protein.
Misregulation of cAMP-dependent guidance cues in β3GnT2−/− mice
LEA lectin blotting identified several olfactory axon guidance molecules that are also modified by PLN glycans (Fig. 1F). We investigated whether the expression of any of these is altered in β3GnT2−/− mice similarly as that of AC3 using Western blot analysis. Kirrel2, an activity-dependent immunoglobulin superfamily glycoprotein (Serizawa et al., 2006), runs on gels as a broad 100 kDa band (Fig. 5A, lanes 1, 2). In mutant OEs, kirrel2 migrated at 90 kDa due to a decrease in β3GnT2-dependent glycosylation. In addition, kirrel2 protein levels were also decreased. PNGase F treatment considerably reduces the size heterogeneity of WT kirrel2, resulting in a sharply migrating 75 kDa protein that approximates its calculated molecular mass (Fig. 5B). Kirrel2 blotted from β3GnT2−/− OEs migrates as a homogeneous 90 kDa band, consistent with the presence of residual N-glycan core-structures. PNGase F treatment of mutant kirrel2 returns its size to 75 kDa, identical to that in the WT. Thus, much like AC3, kirrel2 is modified by N-linked PLN glycans that contribute significantly to the variable migration of these glycoproteins on gels.
The receptor complex for Sema3A is a dimer composed of Nrp1 and plexin-A4 (Suto et al., 2005). Nrp1 is a 103 kDa single pass transmembrane protein with several potential N- and O-linked glycosylation sites. By Western blot analysis, Nrp1 migrates as a 140 kDa protein in WT OEs, which does not change in nulls (Fig. 5A, lanes 3, 4). Plexin-A4 is a 212 kDa single pass transmembrane glycoprotein that migrates as a major 240 kDa band and a faint 230 kDa band. In null mice, the 240 kDa band is modestly decreased while the 230 kDa band is increased from the loss of PLN (Fig. 5A, lanes 5, 6). PNGase F digestion further confirmed that these modifications were carried on N-glycan chains (data not shown).
A decrease in kirrel2 expression in mutant mice is also evident by an overall reduction in the number of OSNs expressing this adhesion molecule. In the WT OE, kirrel2 is expressed by most mature neurons but is decreased in mutants to only a few layers of cells in the most apical OE regions (Fig. 5C,D). In the OB, kirrel2 is broadly expressed at PD3 but refines in adults to a variable pattern of glomerular innervation, with the majority being positive to some degree (Fig. 5E,G). The mosaic expression of kirrel2 by different axon subsets has been proposed to mediate local axon sorting into glomeruli (Serizawa et al., 2006). However, kirrel2 expression is progressively lost with age in null mice. Although there are few glomeruli in the PD3 null OB, there is expression of kirrel2 on olfactory axons within the nerve layer and the few protoglomeruli that form (Fig. 5F). In adults, kirrel2 is completely absent from all fibers in both the mutant nerve layer and glomeruli, consistent with the decrease in AC3 activity in null mice (Fig. 5H).
Nrp1 plays a critical role in olfactory axon targeting through repulsive interactions with the secreted ligand Sema3A, expressed by glial ensheathing cells within the ventral OB nerve layer and subsets of OSNs (Pasterkamp et al., 1998; Schwarting et al., 2000; Imai et al., 2009). Nrp1 levels in OSNs are regulated by cAMP determined through AC3 (Dal Col et al., 2007; Imai et al., 2006; 2009; Miller et al., 2010). Expression of Nrp1 is decreased in the β3GnT2−/− OE by Western blot analysis, although the receptor is not directly modified by PLN (Figs. 1D, 5A). This decrease may therefore in part reflect the loss of AC3-derived cAMP signals, which could be one mechanism for targeting defects observed in the null OB.
To investigate this further, we immunolabeled OBs from early postnatal and adult null mice to examine potential alterations in Nrp1 distribution. At PD3, Nrp1 expression in posterior glomeruli is restricted to medial and lateral domains, whereas AC3 is strongly expressed by virtually all protoglomeruli (Fig. 6A,B). This patterned expression is maintained in adults, although like kirrel2 its distribution is mosaic, with some glomeruli within these domains failing to express Nrp1 while others are strongly positive (Fig. 6C,D). In PD3 null mice, Nrp1 is absent from all neuronal fibers in both the nerve layer and glomeruli (Fig. 6E,F). The loss of Nrp1 expression continues in adult mutants, despite the emergence of many small, abnormally shaped glomeruli (Fig. 6G,H).
To investigate the mechanism for this decrease, we used in situ hybridization to examine Nrp1 mRNA expression at different developmental time points. As reported previously, Nrp1 is mosaically expressed in subpopulations of OSNs in most OE regions (Fig. 6I–K, arrowheads) (Schwarting et al., 2004). This expression is highest prenatally and during early postnatal development, although Nrp1 continues to be transcribed in the adult OE. Nrp1 is also strongly expressed in blood vessels adjacent to the neuroepithelium, where it influences angiogenesis and blood vessel branching independently of Sema3A through its ligand VEGF165 (Fig. 6I, arrows). In β3GnT2−/− mice, the expression of Nrp1 is greatly reduced throughout the OE at all time points, although expression in blood vessels is unaffected (Figs. 6L–N, 7E). Together with the loss of AC3 activity and the absence of Nrp1 protein expression on axons, our data support a prominent role for cAMP in regulating Nrp1 expression levels in OSNs.
Sema3A is also expressed by OSNs in a pattern complimentary to that of the Nrp1 receptor, which has been proposed to be critical for pre-target sorting of olfactory axons within nerve bundles (Imai et al., 2009). Ablation of Sema3A leads to targeting abnormalities in the OB (Schwarting et al., 2000; Taniguchi et al., 2003; Imai et al., 2009). In contrast to Nrp1, Sema3A expression in OSNs is inversely correlated with cAMP levels (Imai et al., 2009). In β3GnT2−/− mice, we observed a notable increase in Sema3A expression by in situ hybridization at all time points examined (Fig. 7A–D). This alteration, as well as those identified for several other cAMP-dependent molecules, was confirmed by qPCR (Fig. 7E).
Thus, the loss of AC3 activity observed in β3GnT2−/− mice correlates with predictable alterations in multiple guidance cues and sorting molecules. These disruptions provide a mechanism for the axon guidance phenotype observed in β3GnT2−/− mice, as well as the loss of OR subsets through a failure to form proper connections with synaptic partners in the OB.
Discussion
Olfactory projections are patterned by multiple guidance cues that act cooperatively to target axons to stereotyped loci in the OB. The OR-dependent production of cAMP by AC3 modulates transcription of Nrp1, its ligand Sema3A, as well other genes that regulate growth cone dynamics (Imai et al., 2006, 2009). In this report, we show that β3GnT2 glycosylation influences the expression of multiple axon guidance molecules that mediate proper OB connectivity. This is achieved through the synthesis of PLN glycans on AC3, which are required for efficient cAMP production by OSNs. An additional consequence of β3GnT2 deletion is the loss of AC3 and kirrel2 from OB axons. This deficit suggests a defect in localization or maintenance of these proteins on β3GnT2−/− axons. These distinct effects of β3GnT2 loss, summarized in Figure 8, will be discussed in greater detail below.
β3GnT2 glycosylation is essential for AC3 activity
Although β3GnT2 modifies a number of glycoproteins, many of the defects associated with PLN loss appear to result from decreased AC3 activity. The phenotypes for AC3−/− and β3GnT2−/− mice, each characterized by a deficit in cAMP production, are strikingly similar (Trinh and Storm, 2003; Henion et al., 2005; Dal Col et al., 2007; Zou et al., 2007): (1) the formation of disorganized glomeruli that are largely excluded from the dorsal OB; (2) identical defects in axon guidance and OSN survival for several representative OR subsets; (3) a persistence of heterotypic glomeruli that lasts into adulthood; and (4) the loss of cAMP-dependent proteins, Nrp1 and kirrel2, and the upregulation of Sema3A. This overlapping spectrum of defects strongly suggests that β3GnT2 is necessary for maintaining the AC3-dependent cAMP levels required for olfactory map formation.
In β3GnT2−/− mice, forskolin-stimulated AC3 activity is reduced 80–90% in the OE (Fig. 3A). Although AC2, AC3, and AC4 are expressed in the OE, AC3−/− mice are functionally anosmic and fail to elicit behavioral or electrophysiological responses to a variety of odorants (Wong et al., 2000; Trinh and Storm, 2003). These studies identified AC3 as the only adenylyl cyclase that signals downstream of odorant receptor activation. Interestingly, AC3−/− mice exhibit higher residual cAMP responses to forskolin than β3GnT2−/− mice, presumably from the activity of these residual isoforms. Like AC3, AC2 is a heavily glycosylated glycoprotein (Wong et al., 2000). It is possible that PLN may be important for AC2 activity as well. All nine mammalian adenylyl cyclases share conserved N-glycosylation sites within extracellular loops 5 and/or 6, suggesting that these glycans may be important for protein expression or function (Wu et al., 2001).
It is unclear at present why PLN modifications are critical for cAMP synthesis. Complete removal of N-glycosylation can lead to protein retention in the endoplasmic reticulum (Helenius and Aebi, 2001). However, β3GnT2 acts distally on N-glycan branches, and its ablation leaves intact N-glycan core structures required for folding and transport (Fig. 2B). In addition, AC3 localizes normally to the cilia plasma membrane (Fig. 3B,C). Although it is possible that PLN could be required for AC3 enzymatic activity directly, in vitro experiments examining the effects of prematurely blocking AC3 glycan extensions would not appear to support such a role (Li et al., 2007).
A possible mechanism by which β3GnT2 could promote cAMP production would be to facilitate the formation of signaling complexes through PLN interactions with endogenous lectins in the extracellular matrix. Within cellular microdomains, specificity needs to be provided to cAMP signals to promote localized responses. The β2-adrenergic receptor, for example, associates with adenylyl cyclases and the Cav1.2 channel in a complex that promotes “fight or flight” responses (Davare et al., 2001). PLN could enhance the association of AC3 with other signaling proteins to amplify OR-derived stimulation. We have previously shown that the PLN binding proteins galectins-1 and -9 are highly expressed along axon tracts between the OE and OB (Mahanthappa et al., 1994; Schwarting et al., 2000). These galectins have functionally multivalent binding sites that could crosslink AC3 with other signaling proteins, holding these molecules in close proximity at the cell surface.
β3GnT2 maintains AC3 on olfactory axons
An issue that complicates our understanding of the role of activity in olfactory targeting is the fact that all proteins required for odorant signaling, including AC3, ORs, Gαs, and CNGA2, are expressed not only in cilia but also in axon termini, where they can mediate cAMP production and [Ca2+] influx (Maritan et al., 2009). How cAMP levels in cilia are coordinated with axonal signaling components, particularly in the absence of obvious sources of distal ligand stimulation, is not known. Intrinsic OR signaling in growth cones has been proposed as a potential mechanism for establishing cAMP levels in different axon subsets (Imai and Sakano, 2008).
It is noteworthy, therefore, that AC3 is initially absent from axons that innervate the OB in embryonic and early postnatal β3GnT2−/− mice. How this loss is related to the decreased cAMP synthesis we measured in null OSNs is unclear, but it may be critical to the targeting defects observed in mutants. It is possible, for example, that cAMP synthesized in the axonal compartment influences OSN gene expression, which could account for the loss of Nrp1 and kirrel2 from β3GnT2−/− projections. Focal exposure of olfactory growth cones to odorants has been shown to mediate cAMP production, nuclear protein kinase A translocation, and potentially gene transcription (Maritan et al., 2009). However, a definitive role for localized growth cone signaling in olfactory axon targeting remains to be shown experimentally. In adults, AC3 is detectable at low levels on null axons within the nerve layer and glomeruli, suggesting potentially a permissive role in OB reinnervation (Fig. 4O,P). Despite this, cAMP production remains dramatically reduced in β3GnT2−/− OEs, and Nrp1 and kirrel2 are absent from adult null axons. Furthermore, elevated levels of axon degeneration are detectable in the null OB throughout adulthood, making the significance of this delayed AC3 expression unclear.
There are several mechanisms by which β3GnT2 could promote axonal localization of AC3 and kirrel2. PLN may be required for axonal transport or to directly maintain glycoprotein surface expression through interactions with carbohydrate binding proteins in the extracellular matrix. Both AC3 and kirrel2 are heavily modified by PLN, and the high molecular weight glycoforms of each are specifically decreased in β3GnT2 mutants. Olfactory axons are exposed to galectin-1 and -9, which are produced by glial cells along axonal pathways between the OE and OB (Crandall et al., 2000). Thus, in addition to the requirement for PLN in AC3 activity, β3GnT2 glycosylation could also function directly to promote axon localization.
Kirrel2 expression has also been reported to be upregulated by neuronal activity that requires CNGA2 (Serizawa et al., 2006). Expression levels of activity-regulated genes may also vary depending on stimulation within the local odor environment (Bennett et al., 2010). The severe decrease in cAMP levels in the β3GnT2−/− OE suggests that odor-evoked activity would be compromised, which could account for the loss of kirrel2 from OSNs. β3GnT2−/− mice, however, appear to have only moderately affected behavioral responses to AC3-dependent odorants, suggesting significant residual olfactory function (Henion et al., 2005; T. K. Knott, G. A. Schwarting, unpublished data). This is surprising given the central role of cAMP in odor-evoked activity. The influx of [Ca2+] in stimulated OSNs occurs on timescales much faster than those for cAMP accumulation (Maritan et al., 2009), perhaps suggesting that neuronal activity could be less sensitive than cAMP-dependent gene expression to the decreased activity of AC3 in the β3GnT2−/− OE.
β3GnT2 function in axon guidance
The increased expression of Sema3A by OSNs and the loss of Nrp1 from axons that reach the OB appear to be critical aspects of the mistargeting in β3GnT2−/− mice. The importance of Nrp1-Sema3A repulsive interactions for glomerular positioning is clear from the guidance defects observed in Sema3A−/− mice (Schwarting et al., 2000, 2004; Taniguchi et al., 2003) and in transgenic models with altered Nrp1 expression (Imai et al., 2009). Manipulating levels of cAMP also alters the expression of Nrp1 in OSNs of transgenic mice (Imai et al., 2006, 2009). Furthermore, when AC3 activity is ablated, Nrp1 expression is lost from axons (Dal Col et al., 2007; Miller et al., 2010).
In early postnatal β3GnT2−/− mice with very low levels of AC3 activity, Nrp1 is likewise absent from projections to the OB. This decrease is not a direct consequence of the loss of glycosylation, since Nrp1 is not a substrate for β3GnT2. Rather, the large decrease in Nrp1 mRNA expression, detected both by in situ hybridization and qPCR at all developmental time points, points toward a failure in cAMP-dependent transcriptional activation. In the absence of β3GnT2-dependent cAMP signaling, a significant number of M72 axons mistarget to the anterior OB, consistent with AC3−/− mice and transgenic models where Nrp1 levels are decreased (Fig. 8) (Chesler et al., 2007; Imai et al., 2009).
Unlike other sensory systems where axons respond to uniform gradients of guidance cues in target tissues, olfactory axon targeting is achieved in part by the relative levels of guidance and cell adhesion molecules expressed mosaically by subsets of OSNs and their axons (Cutforth et al., 2003; Schwarting et al., 2000, 2004; Schwarting and Henion, 2008; Serizawa et al., 2006). Nrp1 itself is not expressed in a linear gradient in the OB (Schwarting et al., 2004). Even in posterior OB domains with elevated expression, Nrp1 levels are highly variable among neighboring glomeruli (Fig. 6D). Other activity-dependent guidance molecules, such as plexin-A1, have been identified that might also influence axon sorting (Imai et al., 2009). Expression of this coreceptor is upregulated by low cAMP levels, as was also found for β3GnT2−/− mice (Fig. 7). It seems likely that additional guidance cues will be required for establishing glomerular positions along the anterior–posterior axis.
Footnotes
-
This work was supported by National Institutes of Health Grants DC00953 and DC09034 to G.A.S and by the Mizutani Foundation for Glycoscience. We are grateful to Thierry Hennet for providing β3GnT2−/− mice used throughout this study, James Schwob for NeuroD1 and Ngn1 cDNAs, Karina Meiri for anti-GAP-43 antibody, Frank Margolis for anti-OMP, Kensaku Mori for anti-OCAM, Brian Nasipak and Chandrashekara Mallappa for helpful discussions regarding qPCR analysis, and Jean Underwood for confocal expertise. We also thank Daniel Storm for reading this manuscript.
- Correspondence should be addressed to either Timothy R. Henion or Gerald A. Schwarting, Department of Cell Biology, University of Massachusetts Medical School, 55 Lake Avenue North, Worcester, MA 01655, gerald.schwarting{at}umassmed.edu or timothy.henion{at}umassmed.edu