1. Cell Biology
  2. Neuroscience
Download icon
A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Unconventional secretory processing diversifies neuronal ion channel properties

  1. Cyril Hanus Is a corresponding author
  2. Helene Geptin
  3. Georgi Tushev
  4. Sakshi Garg
  5. Beatriz Alvarez-Castelao
  6. Sivakumar Sambandan
  7. Lisa Kochen
  8. Anne-Sophie Hafner
  9. Julian D Langer
  10. Erin M Schuman Is a corresponding author
  1. Max Planck Institute for Brain Research, Germany
  2. Max Planck Institute for Biophysics, Germany
Research Article
Cited
5
Views
1,459
Comments
0
Cite as: eLife 2016;5:e20609 doi: 10.7554/eLife.20609

Abstract

N-glycosylation – the sequential addition of complex sugars to adhesion proteins, neurotransmitter receptors, ion channels and secreted trophic factors as they progress through the endoplasmic reticulum and the Golgi apparatus – is one of the most frequent protein modifications. In mammals, most organ-specific N-glycosylation events occur in the brain. Yet, little is known about the nature, function and regulation of N-glycosylation in neurons. Using imaging, quantitative immunoblotting and mass spectrometry, we show that hundreds of neuronal surface membrane proteins are core-glycosylated, resulting in the neuronal membrane displaying surprisingly high levels of glycosylation profiles that are classically associated with immature intracellular proteins. We report that while N-glycosylation is generally required for dendritic development and glutamate receptor surface expression, core-glycosylated proteins are sufficient to sustain these processes, and are thus functional. This atypical glycosylation of surface neuronal proteins can be attributed to a bypass or a hypo-function of the Golgi apparatus. Core-glycosylation is regulated by synaptic activity, modulates synaptic signaling and accelerates the turnover of GluA2-containing glutamate receptors, revealing a novel mechanism that controls the composition and sensing properties of the neuronal membrane.

https://doi.org/10.7554/eLife.20609.001

eLife digest

Information is carried around the nervous system by cells called neurons. The ability of neurons to communicate with each other relies on many proteins that are found on the surfaces of the cells. Like in all animal cells, surface proteins are made inside the cell in a compartment called the endoplasmic reticulum. During this process, one or several complex sugar molecules are usually added to newly made proteins. These sugar molecules are then modified as the proteins leave the endoplasmic reticulum and pass through another compartment called the Golgi apparatus on the way to the cell membrane. The precise number and structure of the sugar molecules attached to the protein define its glycosylation profile.

Neurons receive information from other neurons at branch-like structures called dendrites, which trigger electrical signals that travel through the rest of the cell. To directly control how dendrites generate these signals, neurons make surface proteins locally in dendrites. However, while the endoplasmic reticulum is found all over the neuron, including in the dendrites, the Golgi apparatus is generally only present in the main cell body. It is not known how surface proteins are made in the dendrites or how the proteins’ glycosylation profiles are altered in the absence of a Golgi apparatus.

Hanus et al. used microscopy and biochemical techniques to study the glycosylation profiles of surface proteins in rat neurons. The experiments revealed that immature glycosylation profiles are found on hundreds of different proteins that have been transported to the cell surface. This includes many proteins that are needed to transmit electrical signals between neurons. Next, Hanus et al. selectively blocked the modification of sugar molecules on proteins in the Golgi apparatus. This showed that dendrites are able to form and work properly even if surface proteins have primarily immature glycosylation profiles.

Further experiments suggest that immaturely glycosylated proteins might travel to the surface of neurons using a different route that bypasses the Golgi apparatus. The next step will be to investigate exactly how these proteins are delivered to the surface and how they influence the way neurons behave.

https://doi.org/10.7554/eLife.20609.002

Introduction

Most membrane and secreted proteins are N-glycosylated during their synthesis and processing in the secretory pathway (Moremen et al., 2012). During this process, as nascent proteins emerge in the lumen of the endoplasmic reticulum (ER), a mannose-rich precursor is first transferred en bloc to specific aspargine residues. These immature 'core-glycans' are then trimmed down and modified by the sequential addition of diverse monosaccharides as proteins exit the ER and progress through the Golgi apparatus before they are sent to their final destination (Moremen et al., 2012; Aebi et al., 2010). This sequential and combinatorial modification results in a huge potential diversity of N-glycans and regulates virtually every aspect of membrane protein biology, in particular protein folding, trafficking, stability, ligand-binding and interaction with the extracellular matrix (Moremen et al., 2012; Aebi et al., 2010; Scott and Panin, 2014; Miller and Aricescu, 2014). Consequently, congenital N-glycosylation defects, especially in the brain, result in severe and often lethal developmental disorders (Cylwik et al., 2013). Although the primary organelles of the secretory pathway were first described in neurons (Golgi, 1989; Nissl, 1903), little is known about the N-glycosylation of neuronal membrane proteins.

Numerous mRNAs encoding surface and secreted proteins are localized to dendrites (Cajigas et al., 2012). Yet, how dendritic secretory cargo is processed after local synthesis is still debated. In mammals, neuronal dendrites contain ER, ER-exit sites and ER-Golgi intermediate compartments (ERGIC) and occasionally Golgi outposts, but, for the most part lack canonical Golgi membranes (Torre and Steward, 1996; Gardiol et al., 1999; Krijnse-Locker et al., 1995; Horton and Ehlers, 2003; Hanus and Ehlers, 2008; Cui-Wang et al., 2012; Hanus et al., 2014). It is thus conceivable that, depending on their synthesis in the soma or dendrites, nascent cargo visits distinct sets of secretory subcompartments, and hence acquire specific types of N-glycans. For example, it is believed that nascent neurotransmitter receptors may follow multiple and specific secretory itineraries (Jeyifous et al., 2009), but whether this impacts their glycosylation is unknown.

Intriguingly, Concanavalin A (ConA), a mannose-binding lectin, has been widely used to block the desensitization of plasma-membrane localized AMPA and kainate glutamate receptors (Reiner and Isacoff, 2014; Hoffman et al., 1998; Evans and Usherwood, 1985). From a cell-biological perspective, this is puzzling as core-glycosylated (mannose-rich) N-glycans are typically not found at the cell-surface (Moremen et al., 2012; Aebi et al., 2010; Grieve and Rabouille, 2011). Indeed, a specific sensitivity to the glycosidase EndoH (which cleaves mannose-rich glycans) is commonly used to identify intracellular immature membrane proteins in total cellular lysates (Shi et al., 2010; Greger et al., 2002; Tomita et al., 2003; Sans et al., 2001; Tucholski et al., 2013a).

Here we demonstrate that hundreds of neuronal surface membrane proteins are indeed core-glycosylated, resulting in the neuronal membrane displaying atypically high and activity-dependent levels of ConA-reactive species. We found that while N-glycosylation is generally required for the proper expression of membrane proteins at the neuronal surface, 'immature' core-glycosylated proteins are sufficient to sustain dendritic development and synaptic transmission, indicating that these proteins are fully functional. Focusing on candidate neurotransmitter receptors and auxiliary subunits, we show that core-glycosylated proteins access the cell-surface in a Golgi-independent manner indicating a bypass or a hypo-function of the Golgi apparatus. This atypical N-glycosylation results in an accelerated turnover of membrane proteins and modulates synaptic signaling, revealing a novel mechanism controlling membrane protein homeostasis and function in morphologically complex cells such as neurons.

Results

Core-glycosylated proteins are atypically abundant at the neuronal surface

Their specific binding and sensitivity to distinct lectins and glycosidases distinguishes 3 basic types of N-glycans on membrane proteins: core-glycosylated (or 'immature'), hybrid and complex (Figure 1A; and see Materials and methods for details) (Moremen et al., 2012; Scott and Panin, 2014; Zielinska et al., 2010). To compare the surface levels of these three generic N-glycan types in neurons to non-neuronal cells, we labeled mixed (neuron-glia) hippocampal cultures and 3 commonly used cell-lines (COS7, BHK and CHO) with different lectin biotin-conjugates under non-permeabilizing conditions. As expected, all cell types examined (including neurons and glia) displayed both hybrid and complex N-glycans (labeled by RCA and WGA, respectively) (Figure 1B). Surprisingly, however, mature neurons (Figure 1—figure supplement 1) also displayed a high level of core-glycosylated N-glycans (strongly labeled by ConA and GNA) (Figure 1B). This observation was reinforced by the comparison of neurons and several other cell types: only neurons showed a prominent surface labeling by ConA and GNA, confirming high levels of core-glycosylated proteins at the neuronal plasma membrane (Figure 1C).

Figure 1 with 4 supplements
Core-glycosylated proteins are abundant at the neuronal surface.

(A) The binding to specific lectins and the sensitivity to distinct glycosydases distinguish different N-glycans. Hybrid and complex N-glycans detected at the plasma membrane (PM) and the Golgi …

https://doi.org/10.7554/eLife.20609.003
Figure 1—figure supplement 4
ConA-, RCA- and WGA-reactive glycans are localized in distinct organelles and display contrasting sensitivities to EndoH and PNGase in COS cells.

(A) Subcellular distribution (maximal intensity projections of confocal stacks) and (B) average levels of ConA-, RCA- and WGA-reactive glycans in COS 7 cells in control conditions (light grey) or …

https://doi.org/10.7554/eLife.20609.007
Figure 1—figure supplement 3
Intracellular ConA-, RCA- and WGA-reactive glycans are localized in distinct organelles in neurons.

Images (single confocal planes) of neurons (DIV 28) after labeling with ConA or WGA (red), RCA (green), DAPI (blue) and anti-MAP2 antibodies (cyan). ConA labels structures that have the …

https://doi.org/10.7554/eLife.20609.006
Figure 1—figure supplement 2
Blocking the maturation of core-glycans with kifunensine increases the surface levels of ConA- and GNA-reactive glycans but decreases RCA- and WGA-reactive glycans.

(A) Simplified representation of the N-glycosylation pathway and site of action of kifunensine (Kf). (B) Surface levels of ConA-, GNA-, RCA- and WGA-reactive glycans in neurons (DIV 14–18) treated …

https://doi.org/10.7554/eLife.20609.005
Figure 1—figure supplement 1
Viable hippocampal neurons can be maintained in culture for more than two months.

Labeling of dendrites (MAP2, green), synapses (bassoon, red) and nuclei (DAPI, blue) in DIV28 and DIV70 neurons documenting the viability of DIV 40 and older neurons as in Figure 1. Scale bar 10 μm.

https://doi.org/10.7554/eLife.20609.004

To confirm the specificity of the labeling, neurons were labeled with the same lectins after treatment with Kifunensine (Kf), an inhibitor of ER and Golgi type I mannosidases, which prevents the maturation of core-glycans into hybrid and complex glycans (Tulsiani et al., 1982). As expected, Kf increased the surface levels of ConA and GNA-reactive glycans and reduced the surface levels of RCA and WGA-reactive glycans (Figure 1—figure supplement 2), thus confirming the presence of core-glycosylated proteins on the neuronal surface. To verify that intracellular proteins were indeed localized in the expected secretory compartment, we labeled neurons with ConA, RCA and WGA under permeabilizing conditions together with anti-MAP2 antibodies to visualize the somatodendritic compartment. As expected, ConA marked ER-like structures (Cui-Wang et al., 2012) while RCA and WGA labeled distinct Golgi-like membranes and, for WGA, the nuclear envelope (EN) (Figure 1—figure supplement 3). We also validated the labeling of these intracellular compartments in COS7 cells, where the clear morphology of secretory organelles allows one to unambiguously identify various intracellular structures, and obtained similar results (Figure 1—figure supplement 4). Altogether, these data thus indicate that the atypical abundance of ConA-reactive species at the neuronal surface is indeed due to core-glycosylated proteins.

How abundant is the core-glycosylation of neuronal membrane proteins? To compare directly the relative expression of the different glycans at the neuronal cell surface, we fractioned surface and intracellular proteins by affinity purification after surface biotinylation (Figure 2A and Figure 2—figure supplement 1) and quantified the levels of ConA, RCA and WGA reactive-glycans in these two fractions by far-western blotting (see Materials and methods). We again found, both in cultured neurons and brain tissue, that core-glycans (ConA-reactive species) were surprisingly abundant at the neuronal surface (Figure 2B). Control experiments in total neuronal extracts treated with PNGase and EndoH showed that only ConA-reactive species were sensitive to EndoH, thus confirming that these were core-glycosylated N-glycans (Figure 2—figure supplement 2). Interestingly, quantification of the relative surface expression of ConA, RCA and WGA reactive species showed that, in neurons, core-glycosylated proteins were expressed at relative levels comparable to conventional 'mature' N-glycans (Figure 2C).

Figure 2 with 2 supplements
Core-glycosylated proteins are expressed at the neuronal surface at levels similar to hybrid and complex N-glycosylated proteins.

(A) Isolation of surface ion channel (GluA1) versus cytoplasmic proteins (RhoA) after surface biotinylation and purification with varying amounts of NHS-SS-biotin and streptavidin-beads. (B) Surface …

https://doi.org/10.7554/eLife.20609.008
Figure 2—figure supplement 2
ConA-, RCA- and WGA-reactive glycans assessed by far-western blotting display distinct and specific sensitivities to PNGAse and EndoH.

Total neuronal extracts labeled with ConA, RCA or WGA after or without pretreatment with PNGase or EndoH showing the specificity of the labeling. In contrast to ConA reactivity that is abolished by …

https://doi.org/10.7554/eLife.20609.010
Figure 2—figure supplement 1
The neuronal plasma membrane is intact after surface biotinylation.

Intrinsic fluorescence (single confocal plane image) of pHluo-TM (green), a transmembrane protein marker that selectively exhibits fluorescence at the plasma membrane (and to a lesser extent the …

https://doi.org/10.7554/eLife.20609.009

A number of surface neurotransmitter receptors display atypical N-glycosylation profiles

Where are core-glycosylated proteins localized in the neuronal plasma membrane? To address this, we imunolabeled synapses with the presynaptic marker protein bassoon after surface labeling with ConA. Core N-glycans were distributed throughout the entire neuronal surface but were particularly abundant at synapses, notably in dendritic spines (Figure 3A). To examine whether synaptic proteins themselves displayed core-glycosylation profiles, we assessed the glycosylation status of a cast of key synaptic proteins using their electrophoretic profiles after digestion by PNGase and EndoH, to identify core-glycosylated (EndoH-sensitive) proteins among the family of N-glycosylated (PNGase-sensitive) surface molecules (Figure 3B). We found that a substantial fraction of purified surface GABAA receptor or AMPA-type and NMDA-type glutamate receptor subunits were core-glycosylated (EndoH-sensitive) (Figure 3C). More specifically, these proteins were either entirely core-glycosylated (e.g. GluN1 and GABAAR β3), partially core-glycosylated (i.e. displayed both mature and core-glycosylated glycans on the same polypeptide chain, such as GluA1, and GluN2B) or were expressed as a mixture of proteins with either core-glycosylated or mature N-glycans (e.g. GluA2, GluA4). We note that while there are few reports of core-glycosylated sugars on functional proteins (Hirayama et al., 2016), these are typically associated with a few, select glycosylation sites thought to be inaccessible to Golgi enzymes. In contrast, here we show that all of the glycosylated sites of GluN1 (an obligatory subunit of NMDA receptors with 10 predicted N-glycosylation sites), GluA2 and GluA3 (4 predicted sites) are core-glycosylated. In contrast, TARP γ8, an AMPA receptor auxiliary protein that is thought to be co-trafficked with nascent receptors in the secretory pathway (Tomita et al., 2003) was completely insensitive to EndoH (Figure 3C) as is expected for a typical mature surface glycoprotein in non-neuronal cells (Figure 3—figure supplement 1). Similarly, the synaptic adhesion protein Neuroligin 1 (NLG1, Figure 3C) was also completely insensitive to EndoH. These results strongly indicate that synaptic proteins and receptors are processed and trafficked to the cell-surface through distinct mechanisms / secretory routes: a classical pathway that generates glycoproteins with mature N-glycans (e.g. TARP γ8 and NLG1), which in some instances are detected together with core-glycosylated sites on the same polypeptide (e.g. GluA1), as well as an unconventional route that generates proteins with only core-glycosylated profiles (e.g. GluN1 and GABAAR β3).

Figure 3 with 1 supplement
A number of key neuronal surface proteins display unconventional N-glycosylation profiles.

(AC) Synaptic localization of ConA reactive species and core-glycosylation of surface excitatory neurotransmitter receptors. (A) Fluorescent micrographs of cultured hippocampal neurons (DIV 26) …

https://doi.org/10.7554/eLife.20609.011
Figure 3—figure supplement 1
Standard mature surface N-glycans are typically insensitive to EndoH.

9AB) Immunoblots (A) and relative surface expression (B, surface / total protein) of thermosensitive vesicular stomatitis virus glycoprotein (VSVG-GFP) (Presley et al., 1997) in COS 7 cells at …

https://doi.org/10.7554/eLife.20609.012

Hundreds of key surface neuronal membrane proteins are core-glycosylated

While both binding to ConA and digestion by EndoH rely on a similar biochemical substrate - a sufficient number of mannose residues in a given N-glycan – these two reagents are typically used experimentally to probe molecules that should reside in different cellular compartments: ConA is used to block the desensitization of AMPA and kainate receptors on the surface (Reiner and Isacoff, 2014; Hoffman et al., 1998; Evans and Usherwood, 1985) and EndoH is used to mark immature intracellular proteins in total cellular extracts (Tucholski et al., 2013a; Rouach et al., 2005). To assess whether ConA and EndoH might interact with the same proteins in neurons, we devised a two-round purification strategy to assess the EndoH-sensitivity of surface proteins interacting with ConA (Figure 4A). Consistent with their respective electrophoretic profiles after deglycosylation (Figure 3C), GluN1 and GluA2 were purified based on their interaction with ConA, whereas TARP γ8 was not detected (Figure 4B). As expected, the binding of surface GluN1 and GluA2 to ConA was abolished by competition with free mannose (Figure 4C). Finally we observed that prior digestion with EndoH also blocked GluA2 and GluN1 binding to ConA (Figure 4D), thus demonstrating that in hippocampal neurons, the binding of surface membrane proteins to ConA and their sensitivity to EndoH are equivalent. Thus EndoH-sensitive proteins – proteins that are classically regarded as intracellular and immature - are abundant at the neuronal surface.

Figure 4 with 2 supplements
Hundreds of surface neuronal membrane proteins are core-glycosylated.

Purification scheme (A) and immunoblots (BD) showing the isolation of core-glycosylated N-glycans of the neuronal plasma membrane. (B) Presence of GluA2 and GluN1, but not TARP γ8, among proteins …

https://doi.org/10.7554/eLife.20609.013
Figure 4—figure supplement 2
Evaluation of mass spectrometry data part 2.

(A) Enrichment for secreted and transmembrane proteins among identified surface core-glycosylated proteins. Relative frequencies (solid bars) of soluble (intracellular or secreted), transmembrane …

https://doi.org/10.7554/eLife.20609.015
Figure 4—figure supplement 1
Evaluation of mass spectrometry data part 1.

Enrichment of peptides from proteins in the A (w/o EndoH) and B (w EndoH) sets as a function of repeatability across experiments and technical replicates (see Materials and methods). (A) …

https://doi.org/10.7554/eLife.20609.014

How many surface proteins fall in this category and how important are they for neuronal function? To determine whether core-glycosylation is limited to a small set of neuronal surface proteins or represents a more widespread phenomenon, we used high-resolution mass spectrometry to identify neuronal surface proteins that interact with ConA (Figure 4A,E and F). To increase the specificity of our detection and obtain a high-confidence dataset, we used a label-free quantification (LFQ) approach to identify proteins that were consistently detected across experiments and whose binding to ConA was impaired by EndoH. In brief, plasma membrane proteins were first isolated after surface biotinylation and affinity purified with ConA, after ('background' control group) or without prior treatment with EndoH (target 'core-glycosylated' group). Purified proteins were fragmented with proteases and the resulting peptides were separated by HPLC, ionized and subjected to LFQ and fingerprinting by mass spectrometry. Surface core-glycosylated proteins were identified by focusing the analysis on peptides that showed the highest reproducibility and enrichment in the target group (see Materials and methods and Figure 4—figure supplement 1).

As expected, the resulting protein dataset was markedly enriched for secreted and plasma membrane transmembrane proteins with predicted N-glycosylation sites (Figure 4—figure supplement 2). To our surprise, we found that hundreds of transmembrane or secreted neuronal proteins were core-glycosylated (n=227 protein groups, 647 protein IDs; Supplementary file 1A–1C), including major ionotropic and metabotropic glutamate and GABA receptor subunits, synaptic adhesion proteins, neurotrophin receptors and voltage-gated ion channels (Figure 4E). Protein pathway and gene ontology analysis of our dataset compared to a control dataset (a proteome obtained from total neuronal lysates) (Supplementary file 1D) showed that multiple functional classes of glycoproteins related to axon guidance, synaptic transmission, synaptic plasticity and addiction were significantly overrepresented among the family of coreglycosylated surface proteins (Figure 4F and Supplementary file 1E). Of note, multiple terms associated with autoimmune diseases and cancer were also overrepresented in our dataset (Figure 4F and Supplementary file 1E). As with any affinity purification procedure, we cannot exclude the possibility that some false positive proteins are present owing to their strong interaction with surface proteins. We note, however, that we isolated surface AMPA receptor subunits from TARP γ8 based on their distinct glycosylation profiles (Figure 4B), even though these two proteins that are typically found in the same macromolecular complexes (Schwenk et al., 2014), thus suggesting a high level of specificity.

Taken together, the above data indicate that a large number of neuronal plasma membrane proteins display unconventional N-glycosylation profiles that are typically expressed at low levels in the plasma membrane of heterologous cells, suggesting that core-glycosylation plays a major physiological role in neurons.

Mature N-glycans are not required for dendritic development and maintenance

During neuronal development, the extension and elaboration of dendrites and axons places a high demand on neuronal secretory function to provide membrane, and thus an equally high demand on protein glycosylation. To address whether the presence of core-glycans at the neuronal membrane is specific to a particular developmental stage, we examined the relative surface expression of core and hybrid N-glycans throughout neuronal and synaptic development (over a nine week time-course). We found that while the surface expression of core-glycans remains constant, the surface expression of hybrid glycans progressively increases as neurons mature (Figure 2D), documenting a development-dependent regulation of neuron surface N-glycosylation. We next examined the dependence of dendritic development on N-glycans by treatment with tunicamycin (Tm), a drug that completely blocks all N-glycosylation by preventing the transfer of the N-glycan precursor to target proteins in the ER (Figure 5A). Dendritic growth and complexity were assessed by performing a Sholl analysis (Figure 5C) or by quantifying the total dendritic length and number of tips (Figure 5D). As shown in Figure 5B, we found that dendritic growth was severely impaired by a global blockade of N-glycosylation. To distinguish the contributions of mature vs. immature glycans to dendritic growth we used swainsonine (Sw), a selective blocker of Golgi type-II mannosidase Man2b1 and Man2b2 (Tulsiani et al., 1982), the enzymes that convert EndoH-sensitive proteins into EndoH-resistant species (a prerequisite for their subsequent maturation), and two other mannosidase inhibitors: kifunensine (Kf) and deoxymannojirimycin (DMJ) which inhibit ER or Golgi mannosidases acting upstream of Man2b1/Man2b2 (Herscovics, 1999) (Figure 5A). Surprisingly, we found that the immature N-glycans that remain following treatment with Kf, DMJ or Sw were sufficient to initiate and maintain dendritic growth (Figure 5C–D and Figure 5—figure supplement 1). To verify that these drugs had the expected effects on the surface expression of specific N-glycan subtypes, we assessed the surface expression of core, hybrid and complex glycans by far-western blotting (Figure 5—figure supplement 2). As expected, Kf, DMJ and Sw increased the surface levels of core-glycans while decreasing the levels of hybrid and, at least for Kf and DMJ, complex glycans. Thus, while N-glycosylation is necessary for dendritic development and maintenance, 'immature' N-glycans are sufficient to sustain these processes, which indicates that core-glycosylated proteins on the neuronal membrane surface are fully functional.

Figure 5 with 2 supplements
Mature N-glycans are not required for normal dendritic development.

(A) Simplified representation of the N-glycosylation pathway and site of action of the glycosylation inhibitors tunicamycin (Tm), kifunensine (Kf), 1-deoxymannojirimycin (DMJ) and swainsonine (Sw). …

https://doi.org/10.7554/eLife.20609.016
Figure 5—figure supplement 2
The lack of effect of specific N-glycosylation inhibitors on dendritic development is not due to a lack of activity or to non-specific effects on the glycosylation pathways.

(A) Blots and (B) surface levels of immature/core-glycans (ConA), hybrid glycans (RCA) and complex/mature glycans (WGA) in neurons (DIV 7–10) treated for three days with Tm, Kf, DMJ or Sw as in Figur…

https://doi.org/10.7554/eLife.20609.018
Figure 5—figure supplement 1
Other examples of dendrites after treatment with the drugs described in Figure 5A.
https://doi.org/10.7554/eLife.20609.017

Core-glycosylated proteins traffic to the neuronal surface through a non-canonical secretory pathway

As stated above, the trimming of the N-glycan mannose-core and the resulting loss of glycoprotein sensitivity to EndoH occurs in the Golgi apparatus (Moremen et al., 2012), raising the possibility that the neuronal core-glycosylated proteins that we detect on the plasma membrane may be trafficked to the cell-surface via a Golgi-independent mechanism. To address this, we used Brefeldin A (BFA) – a drug that is commonly used to disrupt the Golgi apparatus (Klausner et al., 1992) (Figure 6A) – and examined its effect on the surface expression of nascent proteins. To do so, neurons were pre-treated for with BFA (2.5 μg/mL; 1 hr), and metabolically labeled by bio-orthogonal non-canonical amino acid tagging (BONCAT) with azido-homo-alanine (AHA) for 120–150 min (Dieterich et al., 2006), or as a negative control methionine (Figure 6—figure supplement 1). Surface and intracellular proteins were then separated and quantified after surface-biotinylation and immunobloting (Figure 6B and C). As a positive control, similar experiments were performed in COS 7 cells. In neurons, BFA had no detectable effect on the levels of nascent surface proteins and slightly decreased intracellular nascent proteins (Figure 6B). In contrast in COS cells, BFA had no effect on intracellular proteins but markedly reduced the levels of nascent proteins at the plasma membrane (Figure 6B), as also observed for other cell types (Davis and Mecham, 1996). Thus, while BFA strongly reduced secretory trafficking in COS 7 cells, it had no detectable effect on the accumulation of nascent proteins at the neuronal surface under these experimental conditions (Figure 6C). We cannot rule out that BFA impairs secretory trafficking in neurons but that our method is not sensitive enough to detect this effect. Yet, our results indicate that secretory trafficking in neurons is markedly less dependent on the Golgi apparatus (GA) than in other cell types, providing an explanation for the atypical prominence of core-glycosylated proteins at the neuronal surface.

Figure 6 with 1 supplement
Unconventional secretory processing of core-glycosylated surface proteins.

(A) GalT-GFP expression in neurons showing the morphology of the somatic Golgi apparatus after a 2 hr long exposure to vehicle (DMSO; control, Ct) or 2.5 μg/mL brefeldin A (BFA). Scale bar 5 μm. (BC

https://doi.org/10.7554/eLife.20609.019
Figure 6—figure supplement 1
Nascent neuronal intracellular and surface proteins (BONCAT) after a 5-hr metabolic labeling with AHA or Methionine (Met) in the presence or absence of BFA.

DIV19 neurons. Note the absence of signal in neurons incubated with Methionine (which for surface proteins demonstrates the absence of residual biotin after elution from streptavidin beads during …

https://doi.org/10.7554/eLife.20609.020

Do the complex and atypical (i.e. core-glycosylated) glycosylation profiles of candidate proteins reflect processing and lack of processing by the GA, respectively? To address this, we chose GluN1, GluA2 and TARP γ8 because of their physiological relevance and their respective complete, mixed (co-existence of EndoH sensitive and insensitive species) or absent core-glycosylation (Figure 3C). We found that exposure to BFA (5 μg/mL; 6–7 hr) significantly reduced the surface expression of TARP γ8 (Figure 6D,E)– as expected for a typical mature N-glycan (Figure 3C). The surface expression of GluA2 was also reduced by BFA, albeit to a lesser extent (Figure 6E). In contrast, the surface expression of GluN1 - a protein whose full surface complement is core-glycosylated – was insensitive to Golgi disruption (Figure 6E). Interestingly, the sensitivity to EndoH of these proteins was inversely correlated with their sensitivity to Golgi-disruption (Figure 6E). Importantly, as proteins with a faster turnover can be expected to respond more quickly to disruption of their biosynthetic pathway, protein stability must be taken into account in interpreting these results. We note, however, that previous studies have shown that GluA2 has a slower turnover (t1/2 1.95 days) than GluN1 (t1/2 1.61 days) (Hanus and Schuman, 2013), which indicates that the differential sensitivity of GluA2 and GluN1 to BFA cannot be accounted for by differences in stability. The stability of TARP γ8 is not known at present.

Thus, these data suggest that the surface expression of core-glycosylated glutamate ionotropic receptors is indeed due to a Golgi-independent secretory processing. Further supporting this view, our protein annotation analysis of the nascent proteins identified by mass spectrometry showed a significant enrichment for ER proteins, but not Golgi proteins, among the surface core-glycosylated proteins (Figure 4—figure supplement 2). Together with the relative paucity of canonical Golgi membranes as compared to other components of the secretory pathway in dendrites (Hanus and Ehlers, 2008; Hanus et al., 2014), the abundance of core-glycosylated proteins at the neuronal surface suggests that Golgi-by pass is surprisingly common for neuronal membrane proteins (Torre and Steward, 1996).

Core-glycosylation is sufficient to maintain synaptic transmission and modulates synaptic signaling

The coexistence of two forms of GluA2 on the cell surface – one form that presents only standard N-glycans and one form that presents only unconventional glycans – suggests that complex glycosylation profiles are not required for the expression of AMPA receptors on the cell-surface. To determine whether this was indeed the case, we used the glycosylation inhibitors described above to determine which glycosylation step(s) are required for the surface expression of GluA2. We found that whereas blocking N-glycosylation entirely with Tm markedly reduced GluA2 surface expression (Figure 7—figure supplement 1), blocking the maturation of core-glycans with Kf, DMJ or Sw had no detectable effect on GluA2 surface levels, despite clear effects (at least for Kf and Sw) on the glycosylation profile of the protein (Figure 7—figure supplement 1). Thus, the delivery of GluA2 to the neuronal surface requires N-glycosylation but does not require processing by Golgi glycosylation enzymes.

To determine more generally the dependence of synaptic receptors on mature glycans, we examined whether Kf impaired synaptic transmission using patch-clamp recording of spontaneous excitatory miniature postsynaptic AMPA/kainate currents. We found that blocking the maturation of core-glycans in the ER for 48 hr had no detectable effect on the frequency or the amplitude of synaptic currents (Figure 7A,B), thus strengthening the view that mature N-glycans are largely dispensable for synaptic transmission.

Figure 7 with 2 supplements
Blocking the maturation of N-glycosylation in the ER does not impair synaptic transmission and modulates the properties of postsynaptic AMPA receptors.

(A) Example traces of miniature excitatory postsynaptic currents (mEPSCs) and (B) average mEPSC amplitude and frequency and input resistance (Mean ± SEM) in control (Ct) neurons (DIV 20–21) and in …

https://doi.org/10.7554/eLife.20609.021
Figure 7—figure supplement 2
Decay plateau (fraction of peak amplitude) as a function of time to peak (s) for synapses in control neurons (Ct, blue) or in neurons treated with Kf (pink).

Note higher proportions of synapses with short time to peak and rapid decay in Kf. N=22 synapses in 3 experiments.

https://doi.org/10.7554/eLife.20609.023
Figure 7—figure supplement 1
Core-glycosylation regulates GluA2 surface expression.

Effects of ER (TM, Kf, DMJ) or Golgi (DMJ, Sw) N-glycosylation inhibitors (see Figure 5A) on the glycosylation profile and surface expression of GluA2. (A) GluA2 before or after digestion with EndoH …

https://doi.org/10.7554/eLife.20609.022

Do core-glycosylated proteins have specific functional properties? As a first step towards addressing this, we determined whether Kf, which inhibits processing beyond core-glycans, alters postsynaptic signaling by combining local glutamate uncaging and calcium imaging. We focused on AMPA-dependent signaling by monitoring postsynaptic calcium responses elicited by repetitive glutamate uncaging (~25 stimuli at ~2.5 Hz) in the presence of AP5 (an NMDA receptor blocker) (Figure 7C–G). This stimulation induced strong postsynaptic responses, which were completely blocked by the AMPA/kainate receptor blocker CNQX (Figure 7D,E). In blind recordings and analyses, we compared the responses of Kf-treated (for 48 hr prior to recording) and control neurons and discovered that, on average, the Kf-treated neurons exhibited responses with a shorter time to peak and a slight trend towards a larger decay (Figure 7F; Figure 7—figure supplement 2). It will be important for future studies to address how these parameters are mechanistically linked to increased core-glycosylation as multiple proteins including voltage-dependent calcium channels, calcium pumps and binding proteins (Rose and Konnerth, 2001) might be involve in shaping the intracellular calcium responses induced by AMPA receptor activation and postsynaptic depolarization. Nevertheless, our results show that core-glycosylation is sufficient to maintain normal synaptic transmission and may regulate the kinetics and magnitude of postsynaptic signaling.

Accelerated turnover and regulated expression of core-glycosylated surface proteins

In native AMPA-type glutamate receptor complexes, the inclusion of the GluA2 subunit prevents the direct permeation by calcium (Isaac et al., 2007). Its presence or absence in synaptic receptor complexes is highly regulated, notably by protein synthesis (Mameli et al., 2007). We thus took advantage of the co-existence of standard (mature N-glycans) and core-glycosylated GluA2 subunits at the neuronal surface to directly compare the turnover of two distinct glycosylated forms of a well-studied and functionally important synaptic protein. Because the overall lifetime of a protein is a key determinant of how fast and by which mechanisms its levels can be tuned in a compartment specific manner (Hanus and Schuman, 2013; O'Leary et al., 2013), we compared the stability of core-glycosylated and mature glycosylated GluA2 with a chase assay after surface biotinylation (Figure 8A–C). The two forms of the protein were quantified by immunoblotting after separation by treatment with EndoH. Consistent with values measured for GluA1 in spinal cord neurons (Mammen et al., 1997), the overall stability of GluA2 (core + standard glycosylated forms) was on the order of tens of hours (15.8 ± 1.5 hr) (Figure 8B). However, the core-glycosylated pool of GluA2 exhibited a substantially shorter half-life than the standard form of the protein (3.4 versus 21.1 hr, respectively) (Figure 8C), demonstrating an accelerated turnover of the core-glycosylated form of the surface receptor.

The core-glycosylation of surface proteins may accelerate their turnover and is regulated by synaptic activity.

(A) Immunoblot of biotinylated-GluA2 at various chase-times (hours) after surface biotinylation. Surface proteins were deglycosylated with EndoH to allow the distinction of mature versus …

https://doi.org/10.7554/eLife.20609.024

GluA2 surface expression is also regulated during homeostatic synaptic scaling (Gainey et al., 2009). We thus determined whether modulation of synaptic activity also impacts the surface expression of core-glycosylated N-glycans. Neuronal activity was either increased by blocking inhibitory synaptic transmission with the GABAA receptor blocker bicuculline (20 μM) or reduced with the ionotropic glutamate receptor blockers CNQX and AP5 (50 μM) for 20 hr and then the levels of ConA-reactive species among surface and intracellular proteins were analyzed by far western blotting of total surface proteins. As expected (Gainey et al., 2009), we found that decreased synaptic activity increased GluA2 surface expression (Figure 8D). This was associated with an increase in the surface expression of total ConA-reactive (core-glycosylated) N-glycans (Figure 8E), thus indicating an activity-dependent regulation of core-glycosylated membrane protein trafficking.

Discussion

It has been previously observed that, compared to other hexose-types, protein mannosylation was particularly pronounced in dendrites (Torre and Steward, 1996; Villanueva and Steward, 2001). Yet, it remained unclear whether mannose-rich N-glycosylated proteins are trafficked to the neuronal-surface. Here, we report that such 'immature' glycoproteins are abundant at the plasma membrane. We show that numerous synaptic adhesion proteins, surface neurotransmitter receptors, voltage-dependent ion channels and growth factor receptors present in the plasma membrane are core-glycosylated, thus displaying (mannose-rich) N-glycosylation patterns previously associated with nascent ion channels in the early secretory pathway (Greger et al., 2002; Tomita et al., 2003; Sans et al., 2001; Rouach et al., 2005).

While it is commonly believed that the maturation of N-glycans in the Golgi apparatus is required for the genesis of functional surface membrane proteins, we found that mature N-glycans are largely dispensable for proper dendritic development and spontaneous synaptic transmission. Interestingly, the core-glycosylation of surface proteins is regulated by synaptic activity and, as seen for GluA2, increases protein turnover. Altogether, these results point towards an important physiological role for core-glycosylation in neurons.

Although AMPA receptor subunits and TARPs are thought to be co-assembled and co-trafficked in the secretory pathway (Zheng et al., 2015), our results show that these proteins display distinct glycosylation profiles and distinct sensitivities to BFA. This indicates that these proteins are processed by distinct mechanisms and likely use distinct secretory routes (Jeyifous et al., 2009) to reach the plasma membrane. Intriguingly, the complete and selective loss of 'mature' N-glycosylated GluA2/3 in hippocampal neurons of TARP γ8 knockout mice results in a relatively modest decrease (34% ) in AMPA excitatory postsynaptic potential amplitude (Rouach et al., 2005). This suggests that core-glycosylated receptors, which account for 33% of total surface receptors, mediate 66% of synaptic transmission, which represents a relative contribution to synaptic currents (and an enrichment at synapses) that is ~4-fold higher. The present study thus challenges the notion that the core-glycosylated GluA2/3 subunits that remain unaltered in the TARP γ8 knockout are retained in the ER, and rather indicate that these receptors are localized at synapses and are functional both in wild type and in TARP γ8 knockout mice. Similarly, AMPA, kainate and GABAA receptors display abnormally increased or decreased EndoH sensitivities in schizophrenic patients (Tucholski et al., 2013a, 2013b; Mueller et al., 2014). Our results question the interpretation that this merely reflects altered intracellular levels of immature receptors and rather suggest that these glycosylation defects impact functional synaptic receptors present on the plasma membrane.

Although dendritic Golgi outposts can be found in some neurons, most dendrites contain early secretory compartments (i.e. ER and ERGICs) but lack generic Golgi membranes (Hanus and Ehlers, 2008; Hanus et al., 2014). We previously showed that multiple postsynaptic signaling pathways control the local processing of nascent secretory cargo in dendrites (Cui-Wang et al., 2012; Hanus et al., 2014). However, it remained unclear whether and how nascent membrane proteins could be N-glycosylated in segments of dendrites lacking Golgi compartments. Thus far, few mammalian proteins have been shown to be trafficked to the plasma membrane via unconventional secretory processing (USP) (Grieve and Rabouille, 2011). The atypical prevalence of core-glycosylated proteins at the neuronal surface that we report here indicates that USP (e.g. Golgi by-pass or hypo-function of Golgi enzymes) is a much more widespread phenomenon than initially anticipated, and allows neurons to modulate the properties of key membrane proteins.

The correlation that we observed between the glycosylation status and the turnover of GluA2 is particularly interesting in this regard (Hanus and Schuman, 2013). The lifetime of a protein has a critical impact on its regulation in space and time. For example, a synaptic protein that is stable for multiple days has the time to potentially explore multiple synapses and go through several rounds of internalization and recycling before being degraded (Ehlers, 2000). It is thus hard to imagine how the local dendritic synthesis of such a protein could lead to rapid and localized changes of synaptic composition and properties. On the other hand, a protein that is much less stable will likely be more efficiently regulated by a fine-tuning of its local synthesis and degradation. It is thus conceivable that the unconventional glycosylation of key proteins and the resulting decrease of their lifetime may tune the spatial length-scale over which local protein synthesis may functionalize synapses (Govindarajan et al., 2011).

In the recent years, our group and others have implemented and developed new tools and strategies to assess the distribution and translation of mRNAs (Cajigas et al., 2012; Buxbaum et al., 2014; Wu et al., 2016), the site of synthesis and the redistribution of specific nascent proteins (tom Dieck, 2015), and their dynamics in dendritic secretory organelles (Cui-Wang et al., 2012; Hanus et al., 2014). Here, we show that the glycosylation status of functional neuronal proteins is the result of distinct post-translational processing mechanisms that likely reflect the availability of secretory machinery in the different subcellular compartments that support protein synthesis. It will thus be interesting for future studies to investigate to what extent synthesis and secretory processing in the soma versus specific segments of dendrites determine protein glycosylation and hence their dynamics and function.

Previous studies in cancer cells have shown that surface proteins may be internalized, trafficked to the GA for further glycan maturation and sent back to the plasma membrane (Snider and Rogers, 1986). We cannot exclude that a similar mechanism exists in neurons and may convert core-glycosylated AMPA receptors into mature proteins. We note however that core-glycosylated GluA2 represents ~one-third of the GluA2 – an abundant neuronal protein – at the steady state. Owing to the paucity of dendritic Golgi membranes compared to endosomal compartments (Ehlers, 2000; Cooney et al., 2002), we find it unlikely that the large fraction of neuronal proteins identified here visit the Golgi following internalization. It will thus be interesting to investigate whether the rapid turnover of core-glycosylated synaptic receptors is due to a faster degradation or to another mechanism.

Given the prevalence of N-glycosylation in the brain (Zielinska et al., 2010), and its influence on membrane protein folding, trafficking, ligand-binding and ion channel conductivity (Moremen et al., 2012; Scott and Panin, 2014), it is clear that the atypical glycosylation of a large number of neuronal surface proteins is physiologically meaningful. In a broader context, it is worth noting that N-glycosylation is dysregulated in numerous human pathologies, and in particular in various cancers. Most notably, an increased branching of complex N-glycans is typically associated with a poor prognosis for breast and colon cancers in humans (Fernandes et al., 1991; Seelentag et al., 1998). Consistently, studies in mice show that Golgi-associated glycosyltransferases such as N-GlcNac, sialyl- and fucosyl-transferases are instrumental to tumor invasiveness (Granovsky et al., 2000; Tsui et al., 2008). The prevalence of core-glycosylated surface proteins in neurons may thus provide important insights on how N-glycans terminal branching is regulated and can be opposed. The immunological isolation of the brain likely plays a permissive role in surface proteins acquiring atypical glycosylation patterns in neurons, as those may otherwise trigger an immunological response. Indeed, multiple lines of evidence suggest that altered glycosylation profiles are important drivers of autoimmune diseases (Rabinovich et al., 2012; Maverakis et al., 2015). The enrichment of proteins involved in such pathologies among the core-glycoproteins that we identified thus indicates that investigating core-glycosylation in neurons may provide important cues on autoimmunity.

Materials and methods

Cell culture

Dissociated hippocampal neurons and cell lines (COS 7, BHK, CHO, L cells; obtained from the American Type Culture Collection) were prepared and maintained essentially as previously described (Cui-Wang et al., 2012; Hanus et al., 2014). Neurons were maintained for up to two months in vitro (Figure 1—figure supplement 1). Lack of cell line contamination with mycoplasma was checked by PCR (eMyco detection kit, Intron Biotechnology).

Brain slices

Acute hippocampal slices were prepared from three week-old Sprague Dawley rats and the CA1 area carefully microdissected by hand as previously described (Cajigas et al., 2012).

Plasmids and transfection

The VSVG-GFP, GalT-GFP, PSD95-mCh and pHluo-TM plasmids were described previously (Cui-Wang et al., 2012). GCaMP6-S (Chen et al., 2013) was purchased from Addgene (plasmid 40753). COS 7 and neurons were transfected with Extreme Gene 9 or Lipofectamine 2000 (Life Technologies), respectively, according to the manufacturer's instructions.

Drug treatments

All drugs were used in the neuron maintenance medium. BFA (Sigma or Tocris), bicuculline, 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and amino-5-phosphonovaleric acid (AP5) (Tocris) were used at final concentrations of 5 μg/mL, 20 μM, 50 μM, 50 μM, respectively. Tunicamycin (Tm), Kifunensine (Kf), deoxymannojirimycin (DMJ) and swainsonine (Sw) (Tocris) were used at final concentrations of 1,2 μM, 5 μM, 75 or 100 μM and 88 μM, respectively.

Glycosydases

Peptide-N-Glycosidase F (PNGase, New England Biolab) and Endoglycosidase H (EndoHf, NEB) were used according to the manufacturer's instructions. In brief, proteins were denaturated in PBS supplemented with 1% Triton, ~0.6% SDS and 50 mM DTT for 15 min at 75°C, and diluted (~1.5 fold) in sodium phosphate (50 mM pH 5.5 final) or sodium citrate buffer (50 mM pH 7.5) plus NP40 (or triton, ~1% final) for PNGase and EndoHf, respectively. Proteins were typically digested with 1000 (PNGase) or 3000 (EndoHf) units/ug protein at 37°C overnight.

Lectins, antibodies and fluorescent streptavidine conjugates

The following lectins (fluorescein or biotin conjugates) were used for cytochemistry (CC) or far-western blotting (FWB) at the indicated final concentrations. Concanavalin A (biotin-ConA, Sigma; CC, 0.33 μg/mL; FWB, 1 μg/mL), galanthus nivalis agglutinin (biotin-GNA, Galab; CC, 1 μg/mL), RCA120-fluorescein (RCA120-FITC, Vector laboratories; CC, 0.7 μg/mL) and RCA (biotin-RCA, Vector laboratories; FWB, 1 μg/mL), wheat germ agglutinin (biotin-WGA, Sigma; CC 0.4 μg/mL; FWB 1 μg/mL), Streptavidine Alexa647 (Life Technologies, CC, 1 μg/mL), IRDye-streptavidin (Licor, FWB, 1:15,000). The following antibodies were used for immunocytochemistry (ICC) or immunoblotting (IB) at the indicated dilutions. Mouse anti-βactin (Sigma, IB, 1:10,000), mouse anti-bassoon (Enzo Life, ICC, 1:1000), rabbit anti-biotin (Cell signaling, ICC, 1:1000), rabbit anti-cacng8/stargazin (Millipore, IB, 1:1000), rabbit anti-GABAA receptor β3 subunit (Synaptic Systems, IB, 1:750), rabbit anti-GABAA receptor γ2 subunit (Synaptic Systems, IB, 1:1000), chicken anti-GFP (Aves Labs, IB, 1:5000), rabbit anti-GluA1 (Abcam, IB, 1: 1000), rabbit anti-GluA2 (Abcam, IB, 1:2000), rabbit anti-GluA4 (Synaptic Systems, IB, 1:1000), mouse anti-GluN1 (BD Pharmingen, IB, 1:1000), rabbit anti-GluN2A (Millipore, IB, 1:1000), rabbit anti-GluN2B (Millipore, IB, 1:1000), guinea pig anti-MAP2 (SYSY, ICC, 1:2000), mouse anti-MAP2 (Sigma, ICC, 1:1000), mouse anti-Neuroligin 1 (Neuromab, IB, 1:200), IRDye secondary antibodies (Li-Cor, IB, 1:15,000), goat anti-guinea pig-Alexa 647 (Life Technologies, ICC, 1:750), goat anti-mouse (GAM)-RRX (Jackson Laboratory, ICC, 1:1000), goat anti-rabbit-RRX and GAM-FITC (Jackson Laboratory, ICC, 1:800).

Surface labeling and immunocytochemistry

For labeling of surface glycoproteins, cells were rinsed in ACSF (119 mM NaCl, 2.5 mM KCl, 1.3 mM MgSO4, 2.5 mM CaCl2, 1.0 mM NaH2PO4, 26.2 mM NaHCO3, and 11.0 mM glucose) or in Hibernate A without phenol red (Brain Bits) and incubated with lectin biotin conjugates diluted in ACSF or Hibernate A, for 10 min at room temperature. After washes, cells were fixed in 4% PFA, blocked in PBS supplemented with 1% fish skin gelatin (Sigma) for 15 min and incubated for 30 min with streptavidin-Alexa647 in the same buffer. For co-immunolabeling, cells were incubated live with ConA-biotin, fixed and permeabilized in 0.2% Triton in PBS for 15 min, blocked in a buffer containing 10% goat serum (Life Technologies) and 3% fish skin gelatin for 30 min and incubated with primary (anti-bassoon, anti-biotin and anti-MAP2) and secondary antibodies diluted in a 1:2 dilution of the same blocking buffer.

Image acquisition and analysis

Confocal imaging was performed with a 40x 1.4 NA objective on Zeiss LSM780 or LSM880 laser point scanning confocal microscopes. Cell average fluorescence was quantified in Metamorph (Molecular Devices) using Z-stack maximal intensity projections.

Surface-biotinylation and pulse chase experiments

Surface biotinylation was performed essentially as described previously (Cui-Wang et al., 2012; Hanus et al., 2014; Ehlers, 2000). In brief, cells were rinsed in ACSF or E4 buffer (120 mM NaCl, 3 mM KCl, 15 mM glucose, 10 mM HEPES, 2 mM CaCl2, 2 mM MgCl2 CaCl2) and incubated with 0.8 to 1 mg/mL NHS-SS-biotin (Thermo) in the same buffer for 7 min at room temperature. Cells were then rinsed and quenched in ACSF or E4 supplemented with 10–20 mM L-lysine, scraped in PBS supplemented with L-lysine and, when appropriate, protease inhibitors (Calbiochem). Cell pellets harvested after centrifugation were then either stored at −80°C or directly lyzed. Pulse chase experiments were performed essentially as described previously (Cui-Wang et al., 2012; Mammen et al., 1997). In brief, cells were biotinylated, washed in ACSF supplemented with 0.1% BSA, and put back in fresh (cell lines) or conditioned (neurons) maintenance medium and kept at 37°C for varying times. Slices were surface-biotinylated in oxygenated ACSF at 4°C for 20 min, quenched in L-lysine at 4°C for 10 min, homogenized and then lyzed as descried above. Surface (biotinylated) proteins were eluted from streptavidin agarose or magnetic beads (Thermo) by reduction of S-S-biotin with 50 mM DTT (in PBS supplemented with ~1% Triton and ~0.6% SDS) for 15 min at 75°C, resulting in the complete removal of biotin from surface proteins (Figure 4B and Figure 6—figure supplement 1).

Immunoblotting and far-western blotting

Immunoblotting was performed essentially as described previously using far-red fluorescent dyes and a Licor Odyssey scanner (Cui-Wang et al., 2012). For far-western blotting, lectin biotin-conjugates, antibodies and streptavidin were diluted in PBS-Tween without the addition of blocking agents. Protein levels in bands of interest were quantified in Image Studio (Licor) or ImageJ (NIH).

Quantification of protein surface expression after surface biotinylation

Known amounts of proteins were separated into a surface (biotinylated) and an intracellular fraction (remaining supernatant) and processed in parallel for immunoblotting or far western blotting. The total fluorescent intensity of individual bands (immunoblotting) or smear of proteins (far western blotting with lectins, or BONCAT, e.g. between ~250 and ~ 70–25KDa) was then quantified and protein relative surface expression calculated as follows: with S, I the total fluorescence intensity and 1/a, 1/b sample dilution factors in surface and intracellular fractions, protein relative surface expression was calculated as the ratio: S / (S + bI/a).

Purification of surface N-glycans

Surface (biotinylated) proteins were eluted from streptavidin-agarose beads at 70°C with 50 mM DTT for 15 min and incubated in the absence (group A) or in the presence (group B – background) of EndoHf (NEB) overnight at 37°C. EndoHf was then heat-inactivated at 80°C for 25 min. Proteins were then incubated with ConA biotin-conjugate in PBS supplemented with 1% Triton x100 and 0.1% SDS for 3 hr at 4°C in the absence, or, for control experiments, in the presence of BSA-mannose (Vector laboratories) and purified with high-capacity streptavidin-agarose or streptavidin-magnetic beads (Thermo).

Mass spectrometry

For each sample group A (target group: surface proteins binding to ConA without prior treatment with EndoH) and B (background group: surface proteins binding to ConA after treatment with EndoH), 2 independent biological replicates (surface proteome preparations from separate primary neuron preparations), 2 experimental replicates (replicate affinity purifications and digestions for each surface preparation) and 3 to 4 technical replicates (replicate LC-MS runs on identical peptide preparations) were analysed, so a total of 13–14 replicates per sample.

Proteins were incubated in 6 M urea, 2 mM DTT, alkylated using 5 mM iodoacetamide and sequentially digested overnight using LysC (1:15 protease:protein ratio). After bringing the urea concentration to a final concentration of 3 M, the samples were incubated with Trypsin (1:15) overnight. The crude peptide mixtures were purified using c18-ZipTips (Millipore), dried using a SpeedVac and stored at −80°C. Four sets of proteolytic fragments (two technical replicates for 2 independent neuronal preps) were loaded 3–4 time each on reverse phase HPLC columns (trapping column: particle size <2 µm, C18, L=20 mm; analytical column: particle size <2 µm, C18, L = 50 cm; ThermoFisher Scientific) using a nano-UPLC device (Dionex Ultimate 3000 RSLC, ThermoFisher Scientifc), and eluted in binary solvent gradients (buffer I: 5% acetonitrile, 95% water, 0.1% formic acid; buffer II: 80% acetonitrile, 20% water, 0.1% formic acid). Typically, gradients were ramped from 5% to 55% buffer II within 200 min at flow rates of 300 nl/min. Peptides eluting from the column were ionised 'online' using a FlexIonSource (Thermo) and analyzed in a hybrid ion trap mass spectrometer (Orbitrap Elite, ThermoFisher Scientific). Sequence information was obtained by computer-controlled, data-dependent switching to MS2 mode (TOP15, FT-IT-mode) using collision energies based on mass and charge state of the candidate ions. Proteins were identified by matching the derived mass lists against a NCBI or 'refseq' database (downloaded from http://www.ncbi.nlm.nih.gov) on a local Mascot server (Matrix Sciences, United Kingdom). In general, a mass tolerance of 2 ppm for parent ions and 0.5 Da for fragment spectra, two missed cleavages and oxidation of methionine as a variable modification and carbamidomethylation of cysteine as a fixed modification were selected as matching parameters in the search program. For quantitative evaluation, the dataset was processed using the MaxQuant software package, Perseus and custom scripts in Matlab (https://github.molgen.mpg.de/MPIBR-Bioinformatics/AtypicalNeuroNGlycans).

We then used a high-stringency label-free quantification (LFQ) approach to identify proteins that were significantly enriched in samples A (target, core-glycosylated) over samples B (i.e. whose binding to ConA was impaired by EndoH and thus represent “background). The efficiency of EndoH was high but not absolute, resulting in B samples consisting of missing values and low-intensity peptides. Although protein abundance was clearly higher in A than in B, this bias (zero values for B sample peptides) complicated the use of average peptide intensities as in a typical LFQ approach. We thus developed a peptide based-strategy and sorted peptides according to their repeatability across experiments (intraclass correlation (ICC) index between biological replicates x ICC index between technical replicates) and their enrichment in A or B: ([average intensity in A – average intensity in B] / max intensity in A or B). Peptides were then separated by hierarchical clustering based on Euclidian distance (0.2 threshold in the distance matrix). Peptides with the highest enrichment and repeatability were clearly separated as a 'natural' cluster (Figure 4—figure supplement 1A), and were used to define our high-confidence 'core-glycosylated' proteins: resulting in the identification of 227 protein groups and 647 protein IDs (Supplementary files 1A and 1C).

As expected, the retrospective analysis of all the peptides that corresponded to these proteins showed a higher enrichment in A and a higher repeatability than the rest of the 'background' proteins (Figure 4—figure supplement 1B and C). As an additional validation, we crossed referenced these proteins with those identified using a standard LFQ approach. Here, we selected proteins based on peptides that were detected in at least 10 out of 13–14 'A' replicates, and whose average peptide intensity was at least 1.3 times higher (Kruskal Wallis test) in A than in B samples. 88.4% of the protein groups that we identified with our peptide-based strategy were also detected by this approach (Supplementary file 1B).

To compare our dataset to the full hippocampal proteome, we selected proteins identified by both mass spectrometry (Sharma et al., 2015) and mRNA deep sequencing in mouse hippocampi (You et al., 2015), yielding 19,690 proteins. Likely owing to the double purification procedure that was used to purify surface core-glycosylated proteins, a few proteins present in our A and/or B sets were not found among these proteins and were thus added to the later list to generate our final input dataset (Supplementary file 1D).

Protein topology, N-glycosylation sites and subcellular localization were defined with Signal IP and TMHMM (Sprenger et al., 2008), NetNGlyc1.0 (http://www.cbs.dtu.dk/services/NetNGlyc/) and CELLO (Yu et al., 2006), respectively. For gene pathway and ontology analyses (Figure 4 and Supplementary file 1E), overrepresented protein functional families were determined using the Kyoto Encyclopedia of Genes and Genomes’ (KEGG) pathway database (Kanehisa et al., 2012) or Gene Ontology annotation (Ashburner et al., 2000) using full protein lists, or subclasses determined according to expected topology (soluble intracellular versus secreted + transmembrane proteins) or predicted N-glycosylation sites.

Distinction of core and hybrid N-glycans

ConA binds with high affinity to terminal α-linked mannoses, thus preferentially to core glycans but may, in theory, also bind to hybrid N-glycans. In this study, we can clearly distinguish between these two types of N-glycans: as exemplified in Figures 1A and 5A, hybrid N-glycans are typically recognized by RCA. However, as shown in Figure 2—figure supplement 2, RCA-binding proteins are EndoH insensitive in hippocampal neurons, in contrast to ConA binding proteins that are EndoH sensitive. Further, as shown in Figure 1—figure supplement 2 and Figure 5—figure supplement 1, RCA-reactive proteins respond completely differently from ConA-reactive species to Kf, DMJ and Sw. We cannot exclude that rare hybrid and yet unrecognized glycans that are Endo-H sensitive and selectively bind ConA but not RCA may be found in neurons. Yet, our data strongly indicate that core-glycans are atypically abundant at the neuronal surface.

Quantification of dendritic morphology

Hippocampal neurons were transfected with GFP at DIV7 and were treated with vehicle (control, DMSO), Tm, Kf, DMJ, Sw at DIV8, fixed at DIV11 and immunolabeled for the somatodendritic marker protein MAP2. Dendrite morphology was assessed basically as previously described (Cui-Wang et al., 2012) using the Simple Neurite Tracer plugin (Longair et al., 2011; Ferreira et al., 2014) in ImageJ/Fiji (NIH). The experimenter was blind to the experimental conditions tested both during image acquisition and analysis.

Protein metabolic labeling by bio-orthogonal non-canonical amino acid tagging (BONCAT)

Cells were preincubated for 1 hr with 2.5 ug/mL BFA (or vehicle, DMSO) and then metabolically labeled with 4 mM azido-homo-alanine (AHA) (tom Dieck, 2015) for 120 to 150 min (or for control for 5 hr with 4 mM AHA versus methionine, Figure 6—figure supplement 1) in Neurobasal without methionine supplemented with glutamax and B27. Cells were then surface biotinylated and surface (S) and intracellular proteins (I) separated as described above. S and I proteins (50 μg protein/lysate equivalent) were then precipitated in cold acetone. Precipitated samples were resuspended in 120 μL of PBS (pH 7.8) supplemented with 0.07% SDS and 0.1% triton and protease inhibitors, cleaned on desalting columns (PD Spin trap G25) and were then biotinylated by azide-alkyne cycloaddition ('CLICK' reaction) as previously described (tom Dieck, 2015), immunobloted and detected with an anti-biotin antibody.

Calcium imaging and focal glutamate uncaging

DIV10 neurons were transfected with PSD-mCh and GCaMP6 as described above and treated with Kf for 48 hr. Neurons were then washed and monitored at 37°C in standard E4 medium (150 mM NaCl, 3 mM KCl, 15 mM glucose, 10 mM HEPES, pH 7.4) supplemented with 3 mM CaCl2, 0.5 mM MgCl2 and 50 μM AP5 and 2.5 mM MNI-glutamate (Tocris). Confocal imaging was performed using a 60x 1.4 NA objective on a Zeiss Observer Z1 inverted microscope equipped with a CSUX1 spinning disk unit (Yokugawa, Inc), an EM-CCD camera (Evolve 512, Photometrics), and a custom diode-laser illumination module (3I Intelligent Imaging Innovations, Inc). MNI-glutamate was uncaged by local irradiation at diffraction limited spots with a tunable 2-photon laser set at 720 nm (Chameleon Ultra, Coherent) coupled to a high speed x,y scanner (Vector). Dendrites were stimulated at individual synapses identified by PSD-mCh fluorescence. Imaging was done in alternance with 26 uncaging pulses delivered at 2.6 Hz. The peak amplitude and time to peak of stimulated calcium responses (GCaMP6 fluorescence) were determined after normalization to baseline by detection of fluorescence maxima. Responses decay and plateau were calculated after normalizing responses to peak values and fitting of the resulting decay plots with a mono-exponential function using Prism (GraphPad). The experimentalist was blind to experimental conditions during data acquisition and analysis.

Electrophysiology

Whole-cell recordings were performed in DIV 20–21 neurons after a ~48 hr exposure to Kf. Neurons were held at −70 mV in voltage clamp and mEPSCs were recorded for at least 10 min using an Axopatch 200B amplifier. The extracellular solution contained (in mM) 140 NaCl, 3 KCl, 10 HEPES, 2 CaCl2, 1 MgSO4, 15 glucose, 0.002 TTX, 0.04 bicuculline and 0.05 AP5 (pH 7.4). Recording pipettes, with resistances between 3–8 MΩ, were filled with a solution containing (in mM) 120 Potassium gluconate, 20 KCl, 10 HEPES, 2 MgCl2, 0.1 EGTA, 2 Na2-ATP and 0.4 Na2-GTP (300 mOsm/l, pH 7.2). Data were analyzed offline with a template-matching algorithm (Guzman et al., 2014) using Stimfit (Guzman et al., 2014) or in Matlab using a custom script. Selected mEPSC events were individually screened with an amplitude threshold of >5 pA and an exponential decay. The experimenter was blind to the experimental conditions tested during both acquisition and analysis. One outlier cell in the control group had a mEPSC frequency close to 4 standard deviations higher than average and was excluded from the data set.

Statistics

Data are presented as means ± SEM unless otherwise indicated. The number of measured values and independent experiments used for quantification are indicated in the text or in the figure legends. Mann Whitney non-parametric test was used to compare two means. Data normality was assessed with Shapiro Wilk’s or Kolmogorov Smirnov’s tests. When data passed normality test, one-way ANOVAs and post hoc Tukey or Sidak’s multicomparison tests were performed to determine whether significant differences existed among all or preselected pairs of means (i.e. control versus BFA) across multiple conditions (i.e. N>2). Otherwise, multiple comparisons were assessed with Kruskal-Wallis and Dunn’s multiple comparison tests. Hypergeometric test was used to assess the significance of relative frequencies in total or subpopulation of proteins (e.g. core-glycosylated versus total proteins). See Statistical reporting in Supplementary file 1F.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
    Endosomal compartments serve multiple hippocampal dendritic spines from a widespread rather than a local store of recycling membrane
    1. JR Cooney
    2. JL Hurlburt
    3. DK Selig
    4. KM Harris
    5. JC Fiala
    (2002)
    Journal of Neuroscience 22:2215–2224.
  7. 7
  8. 8
    Congenital disorders of glycosylation. Part I. Defects of protein N-glycosylation
    1. B Cylwik
    2. M Naklicki
    3. L Chrostek
    4. E Gruszewska
    (2013)
    Acta Biochimica Polonica 60:151–161.
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
    Beta 1-6 branched oligosaccharides as a marker of tumor progression in human breast and colon neoplasia
    1. B Fernandes
    2. U Sagman
    3. M Auger
    4. M Demetrio
    5. JW Dennis
    (1991)
    Cancer Research 51:718–723.
  14. 14
  15. 15
  16. 16
    Dendritic and postsynaptic protein synthetic machinery
    1. A Gardiol
    2. C Racca
    3. A Triller
    (1999)
    Journal of Neuroscience 19:168–179.
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
    Dual modes of endoplasmic reticulum-to-Golgi transport in dendrites revealed by live-cell imaging
    1. AC Horton
    2. MD Ehlers
    (2003)
    Journal of Neuroscience 23:6188–6199.
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
    Redistribution and stabilization of cell surface glutamate receptors during synapse formation
    1. AL Mammen
    2. RL Huganir
    3. RJ O'Brien
    (1997)
    Journal of Neuroscience 17:7351–7358.
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
    Die Neuronenlehre Und Ihre Anhänger : Ein Beitrag Zur Lösung Des Problems Der Beziehungen Zwischen Nervenzelle, Faser Und Grau
    1. F Nissl
    (1903)
    Jena: G. Fischer.
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
    Synapse-associated protein 97 selectively associates with a subset of AMPA receptors early in their biosynthetic pathway
    1. N Sans
    2. C Racca
    3. RS Petralia
    4. YX Wang
    5. J McCallum
    6. RJ Wenthold
    (2001)
    Journal of Neuroscience 21:7506–7516.
  51. 51
  52. 52
  53. 53
    Prognostic value of beta1,6-branched oligosaccharides in human colorectal carcinoma
    1. WK Seelentag
    2. WP Li
    3. SF Schmitz
    4. U Metzger
    5. P Aeberhard
    6. PU Heitz
    7. J Roth
    (1998)
    Cancer Research 58:5559–5564.
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
    Protein synthesis within dendrites: glycosylation of newly synthesized proteins in dendrites of hippocampal neurons in culture
    1. ER Torre
    2. O Steward
    (1996)
    Journal of Neuroscience 16:5967–5978.
  61. 61
    Evaluating the function of matriptase and N-acetylglucosaminyltransferase V in prostate cancer metastasis
    1. KH Tsui
    2. PL Chang
    3. TH Feng
    4. LC Chung
    5. HC Sung
    6. HH Juang
    (2008)
    Anticancer Research 28:1993–1999.
  62. 62
  63. 63
  64. 64
    Swainsonine inhibits the biosynthesis of complex glycoproteins by inhibition of Golgi mannosidase II
    1. DR Tulsiani
    2. TM Harris
    3. O Touster
    (1982)
    The Journal of Biological Chemistry 257:7936–7939.
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70

Decision letter

  1. David D Ginty
    Reviewing Editor; Howard Hughes Medical Institute, Harvard Medical School, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for resubmitting your work entitled "Unconventional secretory processing diversifies neuronal ion channel properties" for further consideration at eLife. Your revised article has been favorably evaluated by Eve Marder (Senior editor), a Reviewing editor, and three reviewers.

The manuscript has been improved but there are a few remaining issues that need to be addressed before acceptance, as outlined below:

The reviews wonder why the blots of BONCAT samples in Figure 6B and Figure 6—figure supplement 1 look so different. Particularly the left panel in Figure 6B appears unusual – the strongest signal is at the edges of the lane. There is also a question about the combination of surface biotinylation and BONCAT. For the experiment shown in Figure 6B, surface proteins were biotinlylated for the separation of surface and intracellular proteins, it is assumed by affinity purification, then again biotinylated in the BONCAT procedure, and then separated by SDS-PAGE and immunodetected by Western blotting with anti-biotin antibodies. This means, that the proteins analyzed in the surface protein fraction are biotinylated irrespective of the BONCAT treatment. Does this not confound the analysis and the interpretation that nascent proteins are analyzed specifically? Please address this issue in the resubmission.

[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "Unconventional secretory processing diversifies neuronal ion channel properties" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by David Ginty as the Reviewing Editor and Eve Marder as the Senior Editor. Bruno Gold (peer reviewer) has agreed to reveal their identity decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

Summary:

In this study, the authors characterize the types and extent of N-glycosylation of cell-surface membrane proteins in neurons. Unlike in other tissues or cells, neurons express substantial amounts of "core-glycosylated" ER-type modified proteins on their cell-surface. Neurons possess vastly greater amounts of plasma membrane vs. cell volume than other cells, and significant amounts of the cellular proteins are synthesized in the dendrites, structures containing extensive ER compartment but little to no Golgi membrane compartments. The Authors suggest that neurons may heavily utilize an unconventional protein secretion pathway in which membrane proteins are synthesized in the ER and then trafficked directly to the plasma membrane. Functionally, the Authors have shown that the life-time of a specific protein GluA2, is determined by the type of N-glycosylation and that neuronal activity can determine the extent of ER-type N-glycosylation on cell-surface. A strength of the work is the use of combinatorial purification steps to isolate core-glycosylated proteins from the cell-surface followed by mass-spec identification. The use of different glycosidase enzymes to reveal that specific proteins contain core, mature or mixed N-glycosylation modifications is also convincing.

This study addresses a fundamental aspect of neuronal and cell biology. The implications of this work are important and the findings are likely to be of interest to a broad audience. However, while some of the key findings are convincing, there are major concerns with both the cell biology and functional aspects of the study. The major concerns fall into two categories:

The evidence that dendritic proteins bypass the Golgi is weak. The physiological relevance of the observations is not well developed.

Given the extensive and deep concerns expressed by the reviewers, and the length of time it would take to generate additional data in support of your conclusions, we must decline to accept this work but encourage you to continue to pursue the problem. Please do consider a new submission to eLife if you are able to address the issues raised by the reviewers.

Major concerns:

1) Figures 1 and 3A use specific lectins to show that neurons express large amounts of core glycoproteins on the cell surface and at synapses. Although presented as quantitative, it is not clear whether the flourescent labeling intensity (Figure 1C, D) provides information about the relative abundance of cell surface glycans with different structures. Also, the only control for these experiments is glycosidase treatment of permeabilized COS7 cells. The signal observed for different lectins will depend upon the biotin conjugation density of the different lectins and degree of non-specific background binding. Notably, lectin binding decreased by less than 2-fold with N-glycanase treatment (Figure 1D) suggesting that nonspecific binding is an issue. The specificity of lectin staining and of the pharmacological treatments should be further tested in experiments in which neurons are treated with tunicamycin (Tm), swainsonine (Sw), kifunensine (Kf), or deoxymannojirimycin (DMJ) and the staining intensity/patterns of surface lectin labeling is then assessed. The Authors could also compare the lectin staining patterns of surface and total (permeabilized) lectin labeling in neurons.

The far Western method used in Figure 2 shows results with better controls than surface lectin staining. Again, the specificity and effectiveness of the pharmacological treatments would be clearly demonstrated if the authors could show that Tm, Sw, Kf, DMJ treatment could alter the staining intensity/pattern in far Western. There was some debate amongst the reviewers about whether this experiment in Figure 2 is compelling. The Authors should describe more precisely how the experiments of Figure 2 were performed and interpreted/quantified.

A related issue is the assumption that ConA will only bind high mannose oligosaccharides that are exemplified by the generic "immature/core" structure shown in Figure 1A. Hybrid structures, which can also have 1 or 2 terminal α-linked mannose residues also bind to ConA with high affinity. This problem impacts Figure 1, Figure 2 and the MS analysis in Figure 4. This issue needs to be discussed in the revised manuscript.

2) Figure 1 uses cultured hippocampal neurons at 40DIV, while Figure 3A uses neurons at 26DIV, and Figure 5 uses 11DIV. For all of the other experiments, especially for the mass-spec analysis, the age of the neurons used is not specified. The Methods section regarding neuron culture is inadequate and should be expanded, as should the figure legends to make it clear to the readers how the experiments were done. It seems highly likely that the pattern and degree of different N-glycosylation shows a developmental profile, in which case the use of neurons of vastly different (or unknown) ages may complicate the interpretation of these results. The authors should either choose a specific age of neuron for their experiments or test whether the patter of N-glycosylation does indeed show a developmental profile. It is curious that the authors decided to use neurons at 40DIV for Figure 1, was there a specific reason to use such old neurons. In many culture preparations, 40DIV may be at or beyond the limits of cell viability, in which case high levels of ConA surface staining may be a result of an artifact due to poor cell health. Some indication of the health of 40DIV neurons should be included in the revision.

Related to this, Figure 5 shows the use of pharmacological inhibitors to test how different N-glycosylation contributes to dendrite growth. The methods/legend describes that drug treatment began on 8DIV and cells were imaged on 11DIV, this time frame is on the late side to examine neurite growth. In the majority of studies examining neurite growth in culture, neurons younger neurons DIV1-7 are used to examine the peak of dendrite growth. Could the authors also include data on neurite growth of 8DIV old neurons prior to drug treatment? The examples given for Tm treated neurons look like 1-2DIV neurons with almost no dendrites, far fewer than what would be expected from 8DIV. This suggests that Tm treatment not only prevented neurite outgrowth but caused considerable retraction of dendrites, perhaps due to severe toxicity of Tm treatment. The fact that Kf, Sw, and DMJ did not greatly impair neurite growth (or cause retraction) supports the conclusions of the authors, however, as mentioned above there are no data to show that Kf, Sw, and DMJ are actually effective at limiting mature N-glycosylation. Lectin labeling and far-Western could be used to show the effectiveness of these drugs.

3) The Authors' conclusion that the plasma membrane proteins use a non-canonical secretory pathway is in part based upon the treatment of cells with brefeldin A (Figure 6). A limitation of this set of experiments is that the detection method (cell surface biotinylation) cannot distinguish between proteins that were transported to the cell surface before or after drug treatment. Cell surface expression will be determined by a combination of the externalization rate, internalization rate and degradation rate. If two proteins are compared that have different half-lives, the one with the shorter half-life will show the greatest reduction when cell surface expression is blocked by any treatment (BFA, cycloheximide, tunicamycin, etc.). The authors should address this critical issue.

4) A major issue raised by all three reviewers is that the physiological relevance of the observations is not adequately developed. In fact, the title of the paper implies that unconventional processing of membrane proteins "diversifies" neuronal ion channel properties. However, there are almost no data to support this.

Given that many surface receptors appear to be core-glycosylated, one would expect that treatment with ConA would have a strong effect on post-synaptic signaling. This could be easily tested. The effects of inhibitors of glycan maturation and glycosidase treatment should also be tested for their impact on Ca++ signaling or other aspects of neuronal physiology. These or related functional experiments should be done to increase the impact of the work.

Beyond this, the authors should limit their conclusions regarding "diversification" of neuronal ion channel properties. Alternatively, since the authors specifically address GluA2 and TARP processing, these proteins could be further examined in new experiments to strengthen this aspect of the work. Brefeldin A (BFA) or Kf treatment reduces the expression of surface TARPg8 but not (or less so) for GluA2. Do these treatments result in greater amount of TARPless AMPARs on the surface/ at synapses? The authors could examine the extent of GluA2/TARP interaction using co-immunoprecipitations, with the expectation that CoIP would be reduced following BFA or Kf treatment. Kf treatment did not show any effect on mEPSC amplitude or frequency. If significantly less TARP is expressed on the surface/synapses, then there may be important changes in AMPAR channel properties. The authors could also examine mEPSC decay kinetics or sensitivity of surface AMPAR to different drugs that modify channel properties in a TARP dependent manner. Such experiments are quite involved and likely beyond the scope of the present study, however, and thus simply toning down the conclusions/title seems the best way forward.

5) An experimental problem with the 2-stage enrichment MS procedure is readily observed when one examines the supplemental tables. Neither stage of the enrichment strategy is sufficiently robust to prevent false-positives. At best 66% of the proteins listed in Table III should be biotinylated in intact cells, and this value assumes that the secreted proteins were extracellular and remained bound to the cells when media was added prior to biotinylation. The false positives included proteins that are not glycoproteins (cytosol, nucleus mitochondria) as well as potential intracellular glycoproteins (ER, Golgi, vacuole) indicating that non-specific binding of proteins to ConA beads was also an issue. Given a 33% false positive rate, it is only reasonable to question whether a 33% false positive rate also applies to the plasma membrane and secreted proteins. The Authors should discuss this issue of false-positives in their ms dataset in the revised manuscript.

6) The experiment to test whether the different forms of GluA2 (Endo H sensitive or EndoH resistant) have different half lives assumes that the disappearance of the immature form of GluA2 from the plasma membrane is diagnostic of degradation. It has been known for more than 20 years that there are retrograde trafficking pathways from the cell surface back to the Golgi that allow additional Golgi glycan processing reactions on cell-surface glycoproteins. The authors are referred to multiple papers published by Martin Snider's laboratory between 1986 and 1996 including the following: (J.C.B. 103:265; J.B.C 264: 7675; Met Cell Biol. 32:339). Conversion of the immature GluA2 to the mature GluA2 is a reasonable alternative explanation for the more rapid loss of immature GluA2. This is an important point that the authors should discuss when describing the interpretation of results shown in Figure 7.

7) The implied model, not specifically described, is that dendritic proteins synthesized at distance from Golgi elements may be expressed on the cell-surface as core-glycosylated proteins, while proteins synthesized near Golgi elements or in the soma will be expressed with mature glycosylation. Previous studies have identified numbers of mRNAs that are trafficked to dendrites for "localized" protein synthesis. Is there any correlation between dendritic mRNA targeting and core-glycosylation? For example, is TARPg8 primarily synthesized in the soma whereas GluN1 or GABAARb3 are targeted to dendrites? This point should be discussed in the revised manuscript.

8) While perhaps not essential for publication, the manuscript would be greatly strengthened by some cell biological characterization of ER/Golgi or glycosyl transferases/glycosidases. For example, are core (ER) glycosyltransferases localized in dendrites or near synapses to a greater extent than Golgi-type modifying enzymes? This point should be discussed in the revision.

Reviewer #1:

In the current study by Hanus et al., the authors characterize the types and extent of N-glycosylation of cell-surface membrane proteins in neurons. Unlike in other tissues or heterologous cells, neurons express substantial amounts of "core-glycosylated" ER-type modified proteins on their cell-surface. Neurons possess vastly greater amounts of plasma membrane vs. cell volume than other cells, and significant amounts of the cellular proteins are synthesized in the dendrites, structures containing extensive ER compartment but little to no golgi membrane compartments. The implications of the current work are that neurons may heavily utilize an unconventional protein secretion pathway in which membrane proteins are synthesized in the ER and then trafficked directly to the plasma membrane. The authors have also shown that the life-time of a specific protein GluA2, is determined by the type of N-glycosylation and that neuronal activity can determine the extent of ER-type N-glycosylation on cell-surface. A major strength of the current work is the use of combinatorial purification steps to isolate core-glycosylated proteins from the cell-surface followed by mass-spec identification. The use of different glycosidase enzymes to reveal that specific proteins contain core, mature or mixed N-glycosylation modifications is also convincing.

This work addresses a fundamental aspect of neuronal and cell biology. The implications of this work are important and the findings are likely to be of interest to a broad audience. While some of the key findings of this work are convincing, the cell biology and functional aspects of the study are underdeveloped. The work would also be strengthened by several additional controls.

1) Figure 1 and Figure 3A use specific lectins to show that neurons express large amounts of core glycoproteins on the cell surface and at synapses. The only control for these experiments is glycosidase treatment of permeabilized COS7 cells. It seems highly probable that some of the surface lectin staining may be non-specific. The authors should also attempt glycosidase treatment to eliminate surface lectin labeling of neurons. In Figure 2C, glycosidase treatment results in almost complete loss of signal using far-Western, and Figure 3C shows that glycosidase treatment results in a quantitative shift in electrophoretic mobility of specific proteins showing complete loss of glycosylation. However, the control experiment shown in 1D shows that glycosidase treatment reduces lectin binding by less than 50%, suggesting that the majority of signal from lectin staining is non-specific. This potential background staining may be even higher in neurons, hindering clear interpretation of Figures 1 and 3A. The specificity of lectin staining and of the pharmacological treatments would be further demonstrated if the authors could show that treatment of neurons with tunicamycin (Tm), swainsonine (Sw), kifunensine (Kf), or deoxymannojirimycin (DMJ) could alter the staining intensity/patterns of surface lectin labeling. It would also be interesting the compare the lectin staining pattern of surface and total (permeabilized) lectin labeling in neurons.

2) The far Western method used in Figure 2 shows more clear results with better controls than surface lectin staining. Again, the specificity and effectiveness of the pharmacological treatments would be clearly demonstrated if the authors could show that Tm, Sw, Kf, DMJ treatment could alter the staining intensity/pattern in far Western.

3) Figure 1 is described as using cultured hippocampal neurons at 40DIV, while figure 3A uses neurons at 26DIV, and Figure 5 uses 11DIV. For all of the other experiments, especially for the mass-spec analysis, the age of the neurons used is not specified. The Methods section regarding neuron culture is inadequate and should be expanded, as well as the figure legends to make it clear to the readers how the experiments were done. It seems highly likely that the pattern and degree of different N-glycosylation may show a developmental profile, in which case the use of neurons of vastly different (or unknown) ages may complicate the interpretation of these results. The authors should either choose a specific age of neuron for their experiments or test whether the patter of N-glycosylation does indeed show a developmental profile. It is curious that the authors decided to use neurons at 40DIV for Figure 1, was there a specific reason to use such old neurons? In many culture preparations, 40DIV may be at or beyond the limits of cell viability, in which case high levels of ConA surface staining may be a result of an artifact due to poor cell health.

4) Related to the previous comment, Figure 5 shows the use of pharmacological inhibitors to test how different N-glycosylation contributes to dendrite growth. The methods/legend describes that drug treatment began on 8DIV and cells were imaged on 11DIV, this time frame is on the late side to examine neurite growth. In the majority of studies examining neurite growth in culture neurons younger neurons DIV1-7 are used to examine the peak of dendrite growth. Could the authors also include data on neurite growth of 8DIV old neurons prior to drug treatment? The examples given for Tm treated neurons look like 1-2DIV neurons with almost no dendrites, far fewer than what would be expected from 8DIV. This suggests that Tm treatment not only prevented neurite outgrowth but caused considerable retraction of dendrites, perhaps due to severe toxicity of Tm treatment. The fact that Kf, Sw, and DMJ did not greatly impair neurite growth (or cause retraction) supports the conclusions of the authors, however, as mentioned above there is no data to show that Kf, Sw, and DMJ are actually effective at limiting mature N-glycosylation. Lectin labeling and far-Western could be used to show the effectiveness of these drugs.

5) The title of the paper implies that unconventional processing of membrane proteins "diversifies" neuronal ion channel properties. However, there is almost no data to support this. The authors may wish to limit their conclusions on the functional consequences of core-glycosylation in favor of further characterization of unconventional secretory processing phenomenon, such as during development or by neuronal activity. Alternatively, since the authors specifically address GluA2 and TARPg8 processing these proteins could be further examined to strengthen the functional conclusions of the work. Brefeldin A (BFA) or Kf treatment reduces the expression of surface TARPg8 but not (or less so) for GluA2. Do these treatments result in greater amount of TARPless AMPARs on the surface/ at synapses? The authors could examine the extent of GluA2/TARP interaction using co-immunoprecipitation, with the expectation that CoIP would be reduced following BFA or Kf treatment. Kf treatment did not show any effect on mEPSC amplitude or frequency. If significantly less TARP is expressed on the surface/synapses there may be important changes in AMPAR channel properties. The authors could also examine mEPSC decay kinetics or sensitivity of surface AMPAR to different drugs that modify channel properties in a TARP dependent manner. Does acute loss of surface glycosylation by glycosidase treatment alter synaptic properties?

6) Figure 7 shows that core-glycosylated GluA2 has a shorter half-life than mature glycosylated GluA2. This is very interesting data that strengthens this work. However, this is the only data showing that unconventional secretory processing diversifies neuronal ion channel properties. The authors also show using their far-Western method that blocking excitatory synaptic transmission increases the surface abundance of core-glycosylation. Can this result be reproduced using surface lectin staining? Can the authors show that core-glycosylation of GluA2 is also increased by AP5/CNQX treatment? If this is the case than it would also be anticipated that the half-life of GluA2 would be reduced following AP5/CNQX treatment. However, this seems unlikely given that previous studies show that inactivity reduces AMPAR surface turnover (Ehlers, Neuron 2000) and increases AMPAR half-life (O'Brien et al., Neuron 1998). Currently the data on neuronal activity and glycosylation is underdeveloped. Either, Figure 7E could be removed, or the effect of neuronal activity on surface glycosylation could be further characterized.

7) The implied model, not specifically described, is that dendritic proteins synthesized at distance from golgi elements may be expressed on the cell-surface as core-glycosylated proteins, while proteins synthesized near golgi elements or in the soma will be expressed with mature glycosylation. Previous studies have identified numbers of mRNAs that are trafficked to dendrites for "localized" protein synthesis. Is there any correlation between dendritic mRNA targeting and core-glycosylation? For example, is TARPg8 primarily synthesized in the soma whereas GluN1 or GABAARb3 are targeted to dendrites? This point could be discussed. Finally, while perhaps not essential for publication, the manuscript would be greatly strengthened by some cell biological characterization of ER/golgi or glycosyl transferases/glycosidases. For example, are core (ER) glycosyltransferases localized in dendrites or near synapses to a greater extent than golgi-type modifying enzymes?

Reviewer #2:

The manuscript from Hanus et al. reports that neuronal cells either have a higher proportion of cell surface glycoproteins bearing high mannose oligosaccharides than several fibroblast cell lines. The authors would like to conclude that this difference in oligosaccharide structure is explained by trafficking of neuronal proteins to the plasma membrane by a "non-canonical secretory pathway." The authors would also like to conclude that the presence of immature oligosaccharide structures on a cell surface glycoprotein (GluA2) reduces the stability of the glutamate receptor as part of a regulatory mechanism. Unfortunately, there are a number of conceptual and technical concerns with this manuscript that preclude publication in eLife.

1) The authors appear to view Golgi processing events as obligatory events for typical cell-surface and secreted proteins. Retention of some high mannose oligosaccharides on N-linked glycoproteins is not that rare, nor is it restricted to proteins synthesized by neuronal cells. HIV gp120 retains a number of high-mannose glycans. ConA capture has been used as a method to capture glycoproteins for mass spectrometry in secreted fluids (serum and urine). Golgi processing of glycans is influenced by accessibility of the core structure to the Golgi mannosidases and glycosyltransferases.

2) A major issue in this manuscript is the assumption that ConA will only bind high mannose oligosaccharides that are exemplified by the generic "immature/core" structure shown in Figure 1A. Hybrid structures, which can also have 1 or 2 terminal α-linked mannose residues also bind to ConA with high affinity. This problem with specificity impacts Figure 1, Figure 2 and the MS analysis in Figure 4.

3) Although presented as quantitative, it is not clear whether the flourescent labeling intensity (Figure 1C,D) provides information about the relative abundance of cell surface glycans with different structures. The signal observed for different lectins will depend upon the biotin conjugation density of the different lectins and degree of non-specific background binding. Notably, lectin binding decreased by less than 2-fold with N-glycanase treatment (Figure 1D) suggesting that nonspecific binding is an issue.

4) Banding patterns for surface ConA and RCA in neuronal cells seem to be different in panels B and C of Figure 2. Prominent bands that are both ConA reactive and RCA reactive are apparent in Figure 2B. The ConA lanes are overloaded in Figure 2C, while the RCA pattern differs in 2B and 2C show a different number of major bands (~5 vs. ~8). Internal samples in Figure 2B are apparently underloaded, as little signal is detected even though the final quantification (Figure 2D) indicates that only ~35% of ConA reactive glycoproteins were present in the surface fraction of neuronal cells.

5) A conceptual problem with the 2-stage enrichment procedure used for the MS analysis is well illustrated by the examples of GluA1 and GluA2. GluA1 has six glycosylation acceptor sites in the extracellular domain. Based upon the molecular weight shift caused by EndoH digestion (Figure 3C) the majority of these glycans are EndoH resistant. However, one high mannose oligosaccharide/protein is sufficient to allow ConA capture, and eventual inclusion into the "core glycosylated protein category". GluA2 is an example of a protein that was detected as a mixture of immature and mature forms. The minor form (not quantified, but clearly much less than 50%) is Endo H sensitive. This minor population is also sufficient to allow GluA2 to be categorized in the core glycosylated protein category (Figure 4 and supplemental tables) even though the majority of the protein has complex and/or hybrid oligosaccharides.

6) An experimental problem with the 2-stage enrichment MS procedure is readily observed when one examines the supplemental tables. Neither stage of the enrichment strategy is sufficiently robust to prevent false-positives. At best 66% of the proteins listed in Table III should be biotinylated in intact cells, and this value assumes that the secreted proteins were extracellular and remained bound to the cells when media was added prior to biotinylation. The false positives included proteins that are not glycoproteins (cytosol, nucleus mitochondria) as well as potential intracellular glycoproteins (ER, Golgi, vacuole) indicating that non-specific binding of proteins to ConA beads was also an issue. Given a 33% false positive rate, it is only reasonable to question whether a 33% false positive rate also applies to the plasma membrane and secreted proteins.

7) The author's conclusion that the plasma membrane proteins use a non-canonical secretory pathway is in part based upon the treatment of cells with brefeldin A (Figure 6). The limitation of this set of experiments is that the detection method (cell surface biotinylation) cannot distinguish between proteins that were transported to the cell surface before or after drug treatment. Cell surface expression will be determined a combination of the externalizaiton rate, internalization rate and degradation rate. If two proteins are compared that have different half-lives, the one with the shorter half-life will show the greatest reduction when cell surface expression is blocked by any treatment (BFA, cycloheximide, tunicamycin, etc.).

8) The experiment to test whether the different forms of GluA2 (Endo H sensitive or EndoH resistant) have different half lives assumes that the disappearance of the immature form of GluA2 from the plasma membrane is diagnostic of degradation. It has been known for more than 20 years that there are retrograde trafficking pathways from the cell surface back to the Golgi that allow additional Golgi glycan processing reactions on cell-surface glycoproteins. I refer the authors to multiple papers published by Martin Snider's laboratory between 1986 and 1996 including the following: (J.C.B. 103:265; J.B.C 264: 7675; Met Cell Biol. 32:339). Conversion of the immature GluA2 to the mature GluA2 is a reasonable alternative explanation for the more rapid loss of immature GluA2.

9) To obtain any evidence that the neuronal plasma membrane proteins are transported to the cell membrane by a "non-conventional secretory pathway" or by passage through a Golgi that is deficient in mannosidase activities it would be important to express a panel (4 or 5) of the candidate neuronal glycoproteins in a heterologous expression system to determine whether the glycosylation patterns differ. Conceptually, the observation that some neuronal proteins have mature oligosaccahrides, mixtures of mature and immature oligosaccharides, or mainly immature oligosaccharides) is inconsistent with the "non-conventional secretory pathway", as this would necessitate sorting events that allow Golgi skipping for subsets of proteins.

Reviewer #3:

In this manuscript, Hanus and colleagues report that an unexpected high number of surface proteins (hundreds), including neurotransmitter receptors, voltage-dependent ion channels, and neurotransmitter transporters are core-glycosylated (i.e. contain the core-glycan added to nascent proteins in the ER but not trimmed and processed by ER/Golgi glycosidases and glycosyltransferases) in hippocampal neurons in culture. This result was obtained by surface biotinylation, pulse-chase labeling, affinity purification and mass spectrometry experiments. Although a complete block of N-glycosylation using tunicamycin impairs dentritic growth, inhibitors of core-glycan maturation have no effect, suggesting that mature N-glycans are not required for dendritic development. The authors provide evidence that core-glycosylated proteins reach the cell surface by Golgi by-pass pathways. Finally, by studying the turn-over of the AMPA receptor subunit GluA2 that exists at the plasma membrane under mature and immature forms, they document that the core-glycosylated pool of GluA2 has a substantially shorter half-time than the pool with mature N-glycans.

This is an interesting piece of work that could have important physiological consequences. It also explains the blocking effect on the desensitization of glutamate and kainate receptors of concanavalin A (ConA), a lectin that binds to core-glycans and is classically used by electrophysiologists. The experiments are well done and generally convincing.

Specific comments:

1) Given that many surface receptors appear to be core-glycosylated, one could expect that the treatment with ConA will have a strong effect on post-synaptic signaling. This could be easily test (for instance by measuring Ca++ activity) and could improve the impact of this study.

2) Figure 1: in panel D, the authors confirm the specificity of lectin labeling using permeabilized COS-7 cells. On the other hand, in panel C, the level of core-glycosylated N-glycans is measured in non-permeabilized neuronal cells. As the treatment with EndoH and PNGase does not fully inhibit ConA or WGA signal (about 50%), it seems important to treat non-permeabilized neurons with EndoH and PNGase and to measure lectin binding.

3) Figure 6: the experiments with BFA are not fully convincing. In panel B, BFA treatment seems to decrease the surface expression of the three receptors, including GLuN1. This experiment should be quantified. There are two lanes for each condition (Ct and BFA): do they correspond to duplicates? Also, the treatment with BFA seems pretty long (6-7h) whereas a dispersion of Golgi is already observed after 2h (panel A). What is the reason? Could it affect cell viability? Finally, it is unclear to me how the results shown in panel C have been obtained.

https://doi.org/10.7554/eLife.20609.026

Author response

The manuscript has been improved but there are a few remaining issues that need to be addressed before acceptance, as outlined below:

The reviews wonder why the blots of BONCAT samples in Figure 6B and Figure 6—figure supplement 1 look so different. Particularly the left panel in Figure 6B appears unusual – the strongest signal is at the edges of the lane.

The different labeling pulse durations (~3 vs. 5 hours) and resulting low or higher signal intensities in Figures 6B and Figure 6—figure supplement 1 may give the impression that the two blots are qualitatively different. Yet, the unusual migration pattern that is particularly obvious for the surface samples in Figure 6B was repeatedly observed in these experiments.

We have now replaced the blots in Figure 6—figure supplement 1 to show a broader set of conditions and biological replicates, to show the repeated occurrence of this unusual migration pattern and to further document the lack of effect of BFA on secretory trafficking (here in contrast to Figure 6, without pre-incubation prior to metabolic labeling.)

The reason for this pattern is not totally clear. We suspect that it is due both to the exposure of surface samples to high levels of detergents during their elution after purification (~0.6% SDS, ~1% Triton and 50mM DTT) and the further precipitation and salt exchange of the samples (both surface and total) that was needed to eliminate DTT, as, in our hands, this compound inhibits CLICK reactions.

There is also a question about the combination of surface biotinylation and BONCAT. For the experiment shown in Figure 6B, surface proteins were biotinlylated for the separation of surface and intracellular proteins, it is assumed by affinity purification, then again biotinylated in the BONCAT procedure, and then separated by SDS-PAGE and immunodetected by Western blotting with anti-biotin antibodies. This means, that the proteins analyzed in the surface protein fraction are biotinylated irrespective of the BONCAT treatment. Does this not confound the analysis and the interpretation that nascent proteins are analyzed specifically? Please address this issue in the resubmission.

This is not correct. Surface proteins were labeled with NHS-SS-biotin which can be cleaved by reduction, enabling an efficient and straightforward means to elute surface proteins from streptavidin beads after purification. This reduction results in the complete removal of biotin from surface proteins as demonstrated by the lack of further interaction with streptavidin beads (see lane “0 uL ConA” in Figure 4B) or lack of detection by anti-biotin antibodies (see lanes “Met” in Figure 6—figure supplement 1). Nevertheless, we thank the reviewer(s) for pointing this out and have modified the manuscript to clarify this issue (see subsection “Surface-biotinylation and pulse chase experiments”).

[Editors’ note: the author responses to the first round of peer review follow.]

Major concerns:

1) Figures 1 and 3A use specific lectins to show that neurons express large amounts of core glycoproteins on the cell surface and at synapses. Although presented as quantitative, it is not clear whether the flourescent labeling intensity (Figure 1C,D) provides information about the relative abundance of cell surface glycans with different structures. Also, the only control for these experiments is glycosidase treatment of permeabilized COS7 cells. The signal observed for different lectins will depend upon the biotin conjugation density of the different lectins and degree of non-specific background binding. Notably, lectin binding decreased by less than 2-fold with N-glycanase treatment (Figure 1D) suggesting that nonspecific binding is an issue. The specificity of lectin staining and of the pharmacological treatments should be further tested in experiments in which neurons are treated with tunicamycin (Tm), swainsonine (Sw), kifunensine (Kf), or deoxymannojirimycin (DMJ) and the staining intensity/patterns of surface lectin labeling is then assessed. The Authors could also compare the lectin staining patterns of surface and total (permeabilized) lectin labeling in neurons.

The point of the reviewer is well taken. In contrast to immunoblotting and far western blotting where these glycosydases completely abrogate glycoprotein recognition by lectins (see Figure 2), EndoH and PNGase have only a partial effect when used for cell labeling and imaging experiments. This discrepancy strongly suggests that these enzymes (EndoH and PNGase) are only fully efficient when used with denatured, solubilized proteins. Although we tried these deglycosylation experiments for imaging using various permeabilization conditions, we did not detect stronger effects. We note, however, that this partial effect and resulting high background signal could also indicate that we may greatly underestimate the levels of surface core-glycosylated proteins in neurons versus other cell types. We also note that to avoid biases in the quantification, we did not threshold fluorescence and quantified the mean fluorescence across the entire cell, which partly contributed to high background signals.

Based on these observations, we relied primarily on far western blotting to compare the surface levels of these glycoproteins and, as suggested by the reviewer, used this assay to test a cast of glycosylation inhibitors (see response to the next point below). Nevertheless, to address reviewer’s comment, we compared the surface levels of ConA-reactive glycans in control neurons and in neurons treated with kifunensine and confirmed that this inhibitor increases the levels of surface labeling with ConA and GNA while decreasing the levels of surface RCA and WGA (Figure 1—figure supplement 2).

We now include imaging data in permeabilizing conditions and document the localization of intracellular core, hybrid, and complex N-glyans in distinct organelles (Figure 1—figure supplement 3).

We now also include additional imaging data in COS7 cells where secretory organelles are easily identifiable by their subcellular localization and their morphology. We now document a differential effect of PNGase and EndoH on specific secretory compartments. EndoH selectively abolishes the signal associated with ConA-reactive immature glycoproteins in the endoplasmic reticulum, but has no effect on the levels of RCA or WGA-reactive species in the Golgi. In contrast, PNGase strongly decreases the labeling of the Golgi by RCA or WGA or the ER by ConA (not shown) (revised Figure 1—figure supplement 4).

The far Western method used in Figure 2 shows results with better controls than surface lectin staining. Again, the specificity and effectiveness of the pharmacological treatments would be clearly demonstrated if the authors could show that Tm, Sw, Kf, DMJ treatment could alter the staining intensity/pattern in far Western.

We thank the reviewer for her/his suggestion. We performed these experiments for Kf using surface staining and imaging (Figure 1—figure supplement

2) and, for Kf, DMJ and Sw surface biotinylation and far western blotting (Figure 5—figure supplement 1). As expected, Tm decreases the surface levels of ConA, RCA and WGA reactive proteins. Kf, Sw and DMJ strongly increase the levels of ConA reactive species while markedly decreasing the levels of RCA reactive species. Kf and DMJ decrease the levels of WGA species, but these are unexpectedly increased by Sw. Although the reason for this later observation is not clear at present, these experiments clearly confirm the specificity of our quantification method and, importantly, validate our conclusions regarding the effect of these drugs on dendritic development and maintenance (see also our last response to point 1 below).

There was some debate amongst the reviewers about whether this experiment in Figure 2 is compelling. The Authors should describe more precisely how the experiments of Figure 2 were performed and interpreted/quantified.

We modified the manuscript to provide more details on these experiments and how they were quantified. See subsection “Quantification of protein surface expression after surface biotinylation” and Figure legend 2.

A related issue is the assumption that ConA will only bind high mannose oligosaccharides that are exemplified by the generic "immature/core" structure shown in Figure 1A. Hybrid structures, which can also have 1 or 2 terminal α-linked mannose residues also bind to ConA with high affinity. This problem impacts Figure 1, Figure 2 and the MS analysis in Figure 4. This issue needs to be discussed in the revised manuscript.

The reviewer raises a valid concern. The revised manuscript has been modified accordingly (see subsection “Distinction of core and hybrid N-glycans”). We note however that our data show that we can indeed distinguish core-glycosylated from hybrid N- glycans:

As exemplified in Figures 1A and 5A, hybrid N-glycans are typically recognized by RCA. But, as clearly shown in Figure —figure supplement 2, RCA-binding proteins are EndoH insensitive in hippocampal neurons, in contrast to ConA binding proteins which are EndoH sensitive;

And, as clearly shown in Figure 1—figure supplement 2 and Figure 5—figure supplement 1, RCA-reactive proteins respond completely differently from ConA-reactive species to Kf, DMJ and Sw.

Thus, although we cannot exclude that rare Endo-H sensitive hybrid N- glycans that bind ConA but not RCA exist in neurons, our data strongly indicate that core-glycans are atypically abundant at the neuronal surface.

2) Figure 1 uses cultured hippocampal neurons at 40DIV, while Figure 3A uses neurons at 26DIV, and Figure 5 uses 11DIV. For all of the other experiments, especially for the mass-spec analysis, the age of the neurons used is not specified. The Methods section regarding neuron culture is inadequate and should be expanded, as should the figure legends to make it clear to the readers how the experiments were done. It seems highly likely that the pattern and degree of different N-glycosylation shows a developmental profile, in which case the use of neurons of vastly different (or unknown) ages may complicate the interpretation of these results. The authors should either choose a specific age of neuron for their experiments or test whether the patter of N-glycosylation does indeed show a developmental profile.

Most of the experiments were done in mature cultured neurons, well beyond the stage at which synapses are forming or are in flux.

Nevertheless, we wanted to directly address this point. To address this, we monitored the surface levels of ConA and RCA reactive glycans over the course of 9 weeks and detected equally high levels of ConA reactive species throughout neuronal development and maturation. These experiments thus justify the validity of using neurons at distinct ages. Importantly, we found that in contrast to ConA-reactive glycans, the surface expression of RCA-reactive- glycans progressively increases with neuronal maturation, documenting a developmental regulation of the glycosylation profile of neuronal surface proteins. See revised Figure 2D.

It is curious that the authors decided to use neurons at 40DIV for Figure 1, was there a specific reason to use such old neurons. In many culture preparations, 40DIV may be at or beyond the limits of cell viability, in which case high levels of ConA surface staining may be a result of an artifact due to poor cell health. Some indication of the health of 40DIV neurons should be included in the revision.

We respectfully disagree with the reviewer and take pride in being able to consistently maintain cultured neurons in good health over long periods of time. As mentioned above, we can now rule out that the high expression of surface ConA reactive glycans in neurons is an artifact due to neuron “excessive” age.

As shown in revised Figures 1B, Figure 1—figure supplement 1, Figure 2A, Figure 2—figure supplement 1 and Figure 3A, the neurons that were used for these experiments showed no sign of poor health as indicated by the surface expression of neurotransmitter receptors, the morphology and density of their synapses, their MAP2 staining and the integrity of their plasma membrane.

Related to this, Figure 5 shows the use of pharmacological inhibitors to test how different N-glycosylation contributes to dendrite growth. The methods/legend describes that drug treatment began on 8DIV and cells were imaged on 11DIV, this time frame is on the late side to examine neurite growth. In the majority of studies examining neurite growth in culture, neurons younger neurons DIV1-7 are used to examine the peak of dendrite growth. Could the authors also include data on neurite growth of 8DIV old neurons prior to drug treatment? The examples given for Tm treated neurons look like 1-2DIV neurons with almost no dendrites, far fewer than what would be expected from 8DIV. This suggests that Tm treatment not only prevented neurite outgrowth but caused considerable retraction of dendrites, perhaps due to severe toxicity of Tm treatment.

We agree with the reviewer about the fact that Tm induces a retraction of dendrites and modified the text to emphasize that Tm prevents dendritic growth and maintenance(see subsection “Mature N-glycans are not required for dendritic development and maintenance”).

We however respectfully disagree with the reviewer about the non-specific toxicity of Tm. Although the exact turnover rate of dendritic proteins is still a matter of debate, their average stability is in the order of 2-3 days (Cohen et al., 2013). Because N-glycosylation is necessary for the processing of a large number of dendritic membrane proteins, it is expected that blocking the renewal of these proteins over 3-4 days will induce dendritic shrinkage.

The fact that Kf, Sw, and DMJ did not greatly impair neurite growth (or cause retraction) supports the conclusions of the authors, however, as mentioned above there are no data to show that Kf, Sw, and DMJ are actually effective at limiting mature N-glycosylation. Lectin labeling and far-Western could be used to show the effectiveness of these drugs.

We fully agree with the reviewers and did these experiments, which, as mentioned above (see our third response to point 1), confirmed that the drugs had the expected effects on neuronal surface N-glycans. These new experiments strongly support our conclusion that core-glycosylated proteins are indeed sufficient to sustain dendritic growth. See revised Figure 1—figure supplement 2 and Figure 5—figure supplement 1.

3) The Authors' conclusion that the plasma membrane proteins use a non-canonical secretory pathway is in part based upon the treatment of cells with brefeldin A (Figure 6). A limitation of this set of experiments is that the detection method (cell surface biotinylation) cannot distinguish between proteins that were transported to the cell surface before or after drug treatment. Cell surface expression will be determined by a combination of the externalization rate, internalization rate and degradation rate. If two proteins are compared that have different half-lives, the one with the shorter half-life will show the greatest reduction when cell surface expression is blocked by any treatment (BFA, cycloheximide, tunicamycin, etc.). The authors should address this critical issue.

The point of the reviewer is well-taken and we agree that this is a key issue. To address this, we selectively monitored the effect of BFA on the accumulation of nascent proteins at the surface after protein metabolic labeling with a bio-orthogonal strategy with a non-canonical amino acid (azido-homoalanine). As shown in revised Figure 8A–C, BFA completely abolishes the delivery of nascent proteins to the cell surface in COS cells. In stark contrast and as we expected, BFA has only a minor effect on the accumulation of these proteins in neurons, thus showing that unconventional secretory processing is atypically common in these cells.

In addition, we would like to stress that, in hippocampal neurons, GluN1 has a shorter half-life than GluA2 (1.61 vs. 1.95 days, respectively) (Cohen et al., 2013; reviewed in Hanus and Schuman, 2013). It is thus very unlikely that the distinct effects of BFA that we observed for these proteins is merely due to their having distinct stabilities. The stability of TARP γ 8 is unknown at present. This point is now discussed in the manuscript (see subsection “Core-glycosylated proteins traffic to the neuronal surface through a non-canonical secretory pathway”, second paragraph).

4) A major issue raised by all three reviewers is that the physiological relevance of the observations is not adequately developed. In fact, the title of the paper implies that unconventional processing of membrane proteins "diversifies" neuronal ion channel properties. However, there are almost no data to support this.

Given that many surface receptors appear to be core-glycosylated, one would expect that treatment with ConA would have a strong effect on post-synaptic signaling. This could be easily tested. The effects of inhibitors of glycan maturation and glycosidase treatment should also be tested for their impact on Ca++ signaling or other aspects of neuronal physiology. These or related functional experiments should be done to increase the impact of the work.

Beyond this, the authors should limit their conclusions regarding "diversification" of neuronal ion channel properties.

The reviewers raise an important point. However, we would like to stress that the fact that dendritic development and maintenance require N- glycosylation but does not rely on Golgi-dependent N-glycosylation is a novel functional finding and has important implications for our understanding of neuronal development and synaptic function.

Nevertheless, we performed additional experiments to address the issue. We decided to focus on the synaptic responses driven by AMPA receptors because of the co-existence of surface AMPA receptors with mature and core-glycosylated glycosylation profiles that we detected and reported in the manuscript. We thus sought to determine whether converting these proteins into mostly core-glycosylated profiles was sufficient to modulate neuron responses to glutamate. As described in revised Figure 8, we used calcium imaging and glutamate uncaging to selectively stimulate AMPA (/kainate) receptors in control conditions or after a 48h-incubation with kifunensine (which blocks the maturation of core-glycosylated N-glycans). We found that Kf regulates the kinetics of postsynaptic responses (decreased time to peak upon repeated stimulation) thus showing that core-glycosylation does modulate postsynaptic responses. These new data thus directly support the notion that unconventional secretory processing diversifies the properties of neuronal ion channels. See Revised Figures 7C–H and Figure 7—figure supplement 2.

Alternatively, since the authors specifically address GluA2 and TARP processing, these proteins could be further examined in new experiments to strengthen this aspect of the work. Brefeldin A (BFA) or Kf treatment reduces the expression of surface TARPg8 but not (or less so) for GluA2. Do these treatments result in greater amount of TARPless AMPARs on the surface/ at synapses? The authors could examine the extent of GluA2/TARP interaction using co-immunoprecipitations, with the expectation that CoIP would be reduced following BFA or Kf treatment. Kf treatment did not show any effect on mEPSC amplitude or frequency. If significantly less TARP is expressed on the surface/synapses, then there may be important changes in AMPAR channel properties. The authors could also examine mEPSC decay kinetics or sensitivity of surface AMPAR to different drugs that modify channel properties in a TARP dependent manner. Such experiments are quite involved and likely beyond the scope of the present study, however, and thus simply toning down the conclusions/title seems the best way forward.

We fully agree with the reviewer. We are interested in addressing this issue in more detail, which, as the reviewer noted, is not a trivial undertaking. We thus decided to remove the data that pertained to this from the present study as these experiments only hinted at a modulation of receptors properties without directly addressing this possibility.

5) An experimental problem with the 2-stage enrichment MS procedure is readily observed when one examines the supplemental tables. Neither stage of the enrichment strategy is sufficiently robust to prevent false-positives. At best 66% of the proteins listed in Table III should be biotinylated in intact cells, and this value assumes that the secreted proteins were extracellular and remained bound to the cells when media was added prior to biotinylation. The false positives included proteins that are not glycoproteins (cytosol, nucleus mitochondria) as well as potential intracellular glycoproteins (ER, Golgi, vacuole) indicating that non-specific binding of proteins to ConA beads was also an issue. Given a 33% false positive rate, it is only reasonable to question whether a 33% false positive rate also applies to the plasma membrane and secreted proteins. The Authors should discuss this issue of false-positives in their ms dataset in the revised manuscript.

We modified the manuscript to include a comment about possible false positives (see subsection “Hundreds of key surface neuronal membrane proteins are core-glycosylated”, third paragraph). We do however respectfully disagree with what the reviewer seems to imply about the specificity of our purification procedure:

Revised Figure 2A clearly shows our ability to selectively capture surface membrane proteins with our surface-biotinylation assay;

We included additional controls in revised Figure 2—figure supplement 1 that further show that the integrity of the plasma membrane is not compromised by this protocol;

We note that our sequential purification strategy is not properly speaking “denaturating” (SDS concentration during binding is 0.2 to 0.05%) and it is thus expected to find some of the many cytoplasmic partners of transmembrane surface proteins;

Despite this and as shown in revised Figure 4B, these two purification steps enable us to selectively purify AMPA receptors from TARP gamma8, while these proteins are typically found together in immuno-precipitation experiments (Schwenk et al., 2014);

Figure 4—figure supplement 2 clearly shows the striking enrichment of N- glycosylated proteins in our dataset;

Finally, we would like to stress that the localization terms (ER, Golgi, vacuole) that are provided in the Tables are based on annotations (mostly from studies in non-neuronal cells) and should not be taken at face value. That being said, we note that the over-representation of ER proteins vs Golgi proteins is completely consistent with our view that Golgi bypass may be atypically frequent in neurons.

6) The experiment to test whether the different forms of GluA2 (Endo H sensitive or EndoH resistant) have different half lives assumes that the disappearance of the immature form of GluA2 from the plasma membrane is diagnostic of degradation. It has been known for more than 20 years that there are retrograde trafficking pathways from the cell surface back to the Golgi that allow additional Golgi glycan processing reactions on cell-surface glycoproteins. The authors are referred to multiple papers published by Martin Snider's laboratory between 1986 and 1996 including the following: (J.C.B. 103:265; J.B.C 264: 7675; Met Cell Biol. 32:339). Conversion of the immature GluA2 to the mature GluA2 is a reasonable alternative explanation for the more rapid loss of immature GluA2. This is an important point that the authors should discuss when describing the interpretation of results shown in Figure 7.

We thank the reviewer for bringing our attention to this point. We modified the manuscript, which now mentions this alternative possibility (see Discussion, seventh paragraph) which we nevertheless still find unlikely, given the paucity of golgi membrane in the dendrites.

7) The implied model, not specifically described, is that dendritic proteins synthesized at distance from Golgi elements may be expressed on the cell-surface as core-glycosylated proteins, while proteins synthesized near Golgi elements or in the soma will be expressed with mature glycosylation. Previous studies have identified numbers of mRNAs that are trafficked to dendrites for "localized" protein synthesis. Is there any correlation between dendritic mRNA targeting and core-glycosylation? For example, is TARPg8 primarily synthesized in the soma whereas GluN1 or GABAARb3 are targeted to dendrites? This point should be discussed in the revised manuscript.

We agree with the reviewer that these aspects are important. Arguably however, delineating where receptors are synthesized, folded, assembled, glycosylated and what exact itinerary they follow to reach the plasma membrane is largely unknown, and that these questions extend far beyond the scope of a single paper. In fact, this question constitutes the major project of another postdoctoral fellow in the lab. We hope, given the scope of what is covered in the revised manuscript, that the reviewers and editors agree that the including findings are of fundamental importance and indeed motivate future studies like those suggested above.

8) While perhaps not essential for publication, the manuscript would be greatly strengthened by some cell biological characterization of ER/Golgi or glycosyl transferases/glycosidases. For example, are core (ER) glycosyltransferases localized in dendrites or near synapses to a greater extent than Golgi-type modifying enzymes? This point should be discussed in the revision.

We modified the manuscript according to reviewer’s suggestion and now specifically discuss this point. A difficulty here is that as far as we can discern, the commercially available antibodies directed against these proteins do not work in rodent neurons. All the antibodies that we have tried so far did not work, yielding either no signal at all or unspecific signal (e.g. strong nucleus staining etc.). We note however that the exclusion of Golgi-GFP (derived from a human galactosyl-transferase) and endogenous Golgi Mannosidase II (with an antibody that is no longer available) from dendrites have been extensively documented (reviewed in (Hanus and Ehlers, 2008; 2016). (See also our second response to point 1).

https://doi.org/10.7554/eLife.20609.027

Article and author information

Author details

  1. Cyril Hanus

    Max Planck Institute for Brain Research, Frankfurt, Germany
    Contribution
    CH, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    For correspondence
    cyril.hanus@brain.mpg.de
    Competing interests
    The authors declare that no competing interests exist.
  2. Helene Geptin

    Max Planck Institute for Brain Research, Frankfurt, Germany
    Contribution
    HG, Conception and design, Acquisition of data, Analysis and interpretation of data
    Competing interests
    The authors declare that no competing interests exist.
  3. Georgi Tushev

    Max Planck Institute for Brain Research, Frankfurt, Germany
    Contribution
    GT, Conception and design, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  4. Sakshi Garg

    Max Planck Institute for Brain Research, Frankfurt, Germany
    Contribution
    SG, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  5. Beatriz Alvarez-Castelao

    Max Planck Institute for Brain Research, Frankfurt, Germany
    Contribution
    BA-C, Conception and design, Acquisition of data, Analysis and interpretation of data
    Competing interests
    The authors declare that no competing interests exist.
  6. Sivakumar Sambandan

    Max Planck Institute for Brain Research, Frankfurt, Germany
    Contribution
    SS, Acquisition of data, Analysis and interpretation of data
    Competing interests
    The authors declare that no competing interests exist.
  7. Lisa Kochen

    Max Planck Institute for Brain Research, Frankfurt, Germany
    Contribution
    LK, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  8. Anne-Sophie Hafner

    Max Planck Institute for Brain Research, Frankfurt, Germany
    Contribution
    A-SH, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  9. Julian D Langer

    1. Max Planck Institute for Brain Research, Frankfurt, Germany
    2. Max Planck Institute for Biophysics, Frankfurt, Germany
    Contribution
    JDL, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  10. Erin M Schuman

    Max Planck Institute for Brain Research, Frankfurt, Germany
    Contribution
    EMS, Conception and design, Analysis and interpretation of data, Drafting or revising the article
    For correspondence
    erin.schuman@brain.mpg.de
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0002-7053-1005

Funding

European Commission (Marie Curie Career Integration Grant 303818)

  • Cyril Hanus

European Research Council (Advanced Investigator Award)

  • Erin M Schuman

Deutsche Forschungsgemeinschaft (Cluster of Excellence for Macromolecular Complexes)

  • Erin M Schuman

Deutsche Forschungsgemeinschaft (CRC 902)

  • Erin M Schuman

Deutsche Forschungsgemeinschaft (CRC 1080)

  • Erin M Schuman

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Ina Bartnik, Nicole Fürst, Dirk Vogel, Anja Staab, Christina Thum and Imke Wüllenweber for excellent technical assistance. We thank Stuart EH Moore for insightful discussions. Work in the laboratory of EMS is supported by the Max Planck Society, the European Research Council, DFG CRC 902, 1080, and the DFG Cluster of Excellence for Macromolecular Complexes. CH is supported by a Marie Curie career integration grant.

Ethics

Animal experimentation: We hereby certify that all the experiments involving animals (i.e. postmortem tissue removal as defined in the § 4(3) of German animal welfare act) that were done in relation to our manuscript entitled "Unconventional secretory trafficking diversifies the properties of neuronal ion channels" were carried out in accordance with the European directive 2010/63/EU, the German animal welfare act, and the guidelines of the Federation of Laboratory Animal Science Associations (FELASA) and the Max Planck Society.

Reviewing Editor

  1. David D Ginty, Reviewing Editor, Howard Hughes Medical Institute, Harvard Medical School, United States

Publication history

  1. Received: August 14, 2016
  2. Accepted: September 22, 2016
  3. Accepted Manuscript published: September 28, 2016 (version 1)
  4. Accepted Manuscript updated: October 5, 2016 (version 2)
  5. Version of Record published: October 24, 2016 (version 3)

Copyright

© 2016, Hanus et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,459
    Page views
  • 527
    Downloads
  • 5
    Citations
Created with Highcharts 5.0.7PageViewsPage Views(Monthly)0204060801 483 …1 483 185.6M

Toggle chartsNavigate left icon

Created with Highcharts 5.0.7Downloads(Monthly)0510152025301 483 …1 483 185.6M

Toggle chartsNavigate left icon

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Comments

Download links

Further reading

    1. Neuroscience
    Angela M Bruno et al.
    Research Article