Stargazin-related protein γ7 is associated with signalling endosomes in superior cervical ganglion neurons and modulates neurite outgrowth

The role(s) of the newly discovered stargazin-like γ-subunit proteins remains unclear; although they are now widely accepted to be transmembrane AMPA receptor regulatory proteins (TARPs), rather than Ca2+ channel subunits, it is possible that they have more general roles in trafficking within neurons. We previously found that γ7 subunit is associated with vesicles when it is expressed in neurons and other cells. Here, we show that γ7 is present mainly in retrogradely transported organelles in sympathetic neurons, where it colocalises with TrkA–YFP, and with the early endosome marker EEA1, suggesting that γ7 localises to signalling endosomes. It was not found to colocalise with markers of the endoplasmic reticulum, mitochondria, lysosomes or late endosomes. Furthermore, knockdown of endogenous γ7 by short hairpin RNA transfection into sympathetic neurons reduced neurite outgrowth. The same was true in the PC12 neuronal cell line, where neurite outgrowth was restored by overexpression of human γ7. These findings open the possibility that γ7 has an essential trafficking role in relation to neurite outgrowth as a component of endosomes involved in neurite extension and growth cone remodelling.


Introduction
moved primarily in a retrograde direction, although anterograde movement was also observed. Fig. 1B depicts several frames from regions of a time-series, pseudo-coloured and merged to emphasise the movement. The image shows that there is a mix of stationary (white) and non-stationary structures (blue, green and red). The average speed of retrograde movement of the  7 -YFP particles was 1.19±0.08 m/second (n7), and the maximum speed was 2.47±0.09 m/second (n7).  7 -containing particles were also densely clustered within the growth-cone bulb of SCG neurites (Fig. 1D), as previously described for the styryl dye FM4-64, following application to growth cones (Bonanomi et al., 2008).

Colocalisation of  7 -CFP with organelle markers in neurites
The retrogradely moving  7 -CFP particles in neurites did not colocalise with the ER marker dsRed2-ER ( Fig. 2A) or with a Golgi marker, dsRed2-Golgi (Fig. 2B). Furthermore,  7 -CFP did not associate with lysosomes labelled with Lysotracker Red (Fig.  2C) or LAMP1 (data not shown), or with mitochondria labelled with Mitotracker Red (Fig. 2D).  7 -CFP also did not colocalise with CD63-positive organelles in neurites (Fig. 2E). CD63 is a tetraspanin protein that is a marker of late endosomes and multivesicular bodies (Pols and Klumperman, 2009). The percentage of particles that colocalised with  7 -positive particles was determined to be 22.2±3.6% (n3) for Mitotracker Red, 19.8±4.8% for Lysotracker Red (n5) and 14.6±10.7% for CD63 antibody (n4) (see Fig. 6B for summary). This level of colocalisation is likely to represent the baseline that occurs by chance in motile particles.

Colocalisation of  7 -CFP with endosome markers
When TrkA-YFP was coexpressed with  7 -CFP, they were substantially colocalised in retrogradely moving particles (Fig.  3A). These generally originated from neurites that terminated in a growth cone (data not shown).  7 -CFP is shown in Fig. 3Ai and TrkA-YFP in Fig. 3Aii. The merged images (Fig. 3Aiii) clearly show colocalisation in a large proportion of particles, calculated to be 46.5±7.9% of  7 -CFP clusters (n5, see Fig. 6B for summary). Furthermore, the colocalisation was continuous over time and during movement, as shown by the kymograph (Fig. 3B). These results suggest that  7 -CFP is associated with signalling endosomes, which are a form of early endosome containing activated TrkA (Cui et al., 2007).
In addition,  7 -CFP particles also partially colocalised with particles expressing the early endosome marker EEA1. This was present in particles both within SCG cell bodies (Fig. 4A), where there was 61.3±4.4% (n8) colocalisation with  7 -CFP, and within the SCG neurites ( Fig. 4B), where there was 73.9±5.6% (n6) colocalisation (see Fig. 6B for summary). We also found that both the number (Fig. 4C) and average size (Fig. 4D) of the EEA1containing particles was increased when  7 -CFP was coexpressed.
Further evidence that  7 is associated with early endosomes comes from our finding that  7 -CFP was associated with EGF receptors in Cos-7 cells visualised after incubation of live cells with EGF-TR (Fig. 5A). In these experiments, 39.2±3.2% of  7 -CFP clusters colocalised with EGF-TR (n5). These particles were also colocalised in SCG neuronal somata (Fig. 5B), where 55.3±5.6% of  7 -CFP clusters colocalised with EGF-TR (n5). However, in neurites, very few EGF-TR-containing particles were observed (Fig. 5C), and only 14.0±7.1% of  7 -CFP clusters colocalised with EGF-TR (n5). This is to be expected because although EGF receptors have been reported in the soma of sympathetic neurons, they have not been observed in neurites (Jia et al., 2007).
N-type Ca 2+ channels are widely expressed in SCG neurons, and are the main Ca 2+ channel subtype responsible for transmitter release at presynaptic terminals in these neurons (Brock and Cunnane, 1999). We therefore wished to examine whether Ca V 2.2 would colocalise with  7 . Interestingly, we observed that 2050 Journal of Cell Science 124 (12) Fig. 1.  7 is present on intracellular organelles subject to transport in neurites. (A) 7 -YFP was microinjected into an SCG neuron, and images taken after 24 hours. The cell body is located to the left of the frame. Multiple  7 -YFP-containing organelles are observed in the neurites, some of which are stationary and some of which are motile. The image represents a single frame from a time series movie. Scale bar: 20m. (B)Consecutive images, from the field depicted in A, obtained at 3 second intervals, were pseudo-coloured blue (image 1), green (image 2) and red (image 3) and superimposed. Stationary structures appear white, whereas the moving particle appears blue, green and red. Scale bar: 20m. (C)Region of neurite from a cell expressing  7 -YFP (top) and free CFP (middle). The merged image is shown at the bottom. The free CFP is uniformly distributed throughout the neurites. The neurite is outlined with a dashed line in the top panel. Scale bar: 10m. (D)Image of a growth cone from a cell expressing  7 -YFP (left), and CFP (middle). The merged image is shown on the right. Scale bar: 10m.  7 -YFP is present as discrete particles in the bulb of the growth cone. coexpression of  7 -YFP with CFP-Ca V 2.2 resulted in its colocalisation with 58.4±8.0% (n4) of  7 -YFP clusters (Fig.  6A,B).

Endogenous  7 is partially colocalised with EEA1
To determine whether endogenous  7 , which we had previously shown to have a punctuate distribution (Ferron et al., 2008), was also colocalised with endosomal markers, we examined its colocalisation with EEA1 in SCG neurons (Fig. 7). Colocalisation was observed in both somata and neurites (Fig. 7A), such that about 45% of  7 -containing particles colocalised with an EEA1 particle (Fig. 7B). This suggests that the colocalisation of  7 -CFP with endosomal markers was not an artifact of overexpression.

Endogenous  7 influences neurite outgrowth
We had previously observed an indication of altered neurite morphology following  7 overexpression in peripheral neurons (Moss et al., 2002). Furthermore, in our previous study we showed that it was possible to knock down endogenous  7 in cultured neurons with shRNA (Ferron et al., 2008). Therefore, to examine whether endogenous  7 has a role in neurite outgrowth and morphology in sympathetic neurons, we transfected SCGs with a mixture of three  7 shRNAs (Ferron et al., 2008), and found that this resulted in a reduction in levels of endogenous  7 (Fig.  8A,B). It also caused a reduction in neurite outgrowth compared with control shRNA (Fig. 9A,B). Cells were transfected with the shRNAs and GFP, and cultured for 4-6 days to allow  7 knockdown to occur, then re-plated, and neurite outgrowth was measured 24 hours later (Fig. 9A). The total neurite length per cell was markedly reduced 4, 5 and 6 days after transfection with  7 shRNA, compared with control shRNA, and the pooled data are shown in Fig. 9B.
We then examined in more detail the effect of  7 knockdown on neurite outgrowth, using the PC12 neuronal cell line, which represents a model for these peripheral neurons. As before, we used a mixture of three shRNAs specifically complementary to mRNA encoding rat  7 , and used Drosophila gnu shRNA as a control (Ferron et al., 2008). We found that transfection of rat  7 shRNAs markedly decreased the length of PC12 neurites, compared with the control shRNA, when measured 9 days after transfection and 3 days after the start of differentiation with NGF ( Fig. 9C,D). Importantly, the effect was reversed by cotransfection of fulllength human  7 with  7 shRNA (Fig. 9D). No increase in apoptosis was observed following transfection with 7 shRNA, using the nuclear dye Hoechst 33258 (data not shown).

Discussion
We have previously described the identification of two genes that encode  5 and  7 , by their homology with the mouse stargazin gene (Cacng2), and have cloned and expressed the cDNA for both human and mouse  7 (Moss et al., 2002). Together with the other -like proteins,  7 belongs to the claudin superfamily, members of which have diverse roles in cellular physiology (Sanders et al., 2001). Some members of the family of -subunits or TARPs have 2052 Journal of Cell Science 124 (12) been shown to have a role in AMPA receptor function (Tomita et al., 2003). Despite not having a classical C-terminal PDZ motif,  7 has also recently been shown to have effects on AMPA receptor trafficking (Kato et al., 2007), and  5 has been found to affect Ca 2+permeable AMPA receptors in Bergmann glia (Soto et al., 2009). Nevertheless, in a recent proteomic study, only a small proportion of AMPA receptors were found to be associated with  subunits, with  5 and  7 being particularly rare (Schwenk et al., 2009), despite the fact that  7 is widely expressed in brain (Moss et al., 2002). This suggests that these -subunits have other roles. Knockout of  7 has recently been found to have little effect on cerebellar function (Yamazaki et al., 2010), unlike knockout of  2 in Stargazer mice (Letts et al., 1998); however, the double-knockout ( 2 and  7 ) mouse is more severely affected than  2 -knockout mice (Yamazaki et al., 2010), suggesting that there is redundancy in subunit function.

Presence of  7 in organelles within neurites that are retrogradely transported
In this study, we have found that when  7 -YFP or  7 -CFP are expressed in SCG neurons, they are localised in endosome-like organelles that are transported primarily in a retrograde direction, away from growth cones. Many studies have shown that growth cones contain vesicles (Cheng and Reese, 1988), and both constitutive and stimulated endocytosis have been demonstrated (Bonanomi et al., 2008;Diefenbach et al., 1999). Constitutively formed endocytotic vesicles have also been shown to undergo retrograde transport (Diefenbach et al., 1999). We have examined the properties of  7 -containing vesicles in neurites, and showed that they do not colocalise with mitochondria, lysosomes, ER or late endosomes. However, we found that the  7 -containing organelles do share some characteristics with signalling endosomes (Zweifel et al., 2005), because the transport is primarily retrograde and the organelles colocalise with TrkA and EEA1. We also found that endogenous  7 -containing particles were partially colocalised with EEA1 in somata and neurites.

2053
Stargazin-like  7 in signalling endosomes  The speed of  7 -CFP-containing vesicles in SCG neurites is in close agreement with the rate of movement of quantum-dot-labelled NGF in DRG neurites, which showed an average rate of 1.3 m/second, with a speed during active movement of 2.1 m/second (Cui et al., 2007). Signalling endosomes are thought to transport activated growth factor receptors to the soma, and are essential for maintaining neurite outgrowth (Cosker et al., 2008). Furthermore, we know that SCG neurons contain endogenous  7 (Ferron et al., 2008), and we have observed here that knockdown of endogenous  7 with shRNA reduced neurite outgrowth both in SCGs and in PC12 cells, where we showed that neurite outgrowth is restored by expression of human  7 .
YFP-Ca V 2.2 was also observed to colocalise with  7 -CFP in retrogradely transported vesicles, and it is possible that this represents Ca V 2.2 that was inserted into the growth cone membrane and then endocytosed during the extensive remodelling of the growth cone cell surface that occurs in the early phase of axon outgrowth (Bonanomi et al., 2008).
In our previous study (Ferron et al., 2008), we identified by coimmunoprecipitation from a PC12 cell line stably transfected with  7 -HA, that the RNA binding protein hnRNP A2 coimmunoprecipitates with  7 . This protein has been shown to be involved in the stability, trafficking and localisation of mRNAs containing a specific binding motif termed A2RE (Ainger et al., 1997;Shan et al., 2003). We showed that mRNA encoding Ca V 2.2 contained an A2RE motif and was immunoprecipitated with hnRNP A2. However, other mRNAs containing A2RE motifs are also likely to be similarly affected, and with Ca V 2.2, might be involved in the observed consequences of knockdown of  7 on neurite outgrowth.
These results provoke the hypothesis that  7 is a protein that is involved more generally in intracellular transport, which possibly has a role in transport of mRNAs encoding transmembrane proteins, or perhaps a more global role. There are indications that Ca V 2.2 mRNA is subject to transport, because it has been identified in dendritic growth cones (Crino and Eberwine, 1996). Furthermore, N-type Ca 2+ channels are localised to the presynaptic terminals of peripheral neurons, such as DRG neurons, and local synthesis of transmembrane proteins has recently been demonstrated in axons and axonal growth cones (Brittis et al., 2002;Lin and Holt, 2007). Our finding that shRNA knockdown of endogenous  7 markedly 2054 Journal of Cell Science 124 (12)  decreases neurite outgrowth might indicate that the presence of  7 in growth cones maintains a particular level of expression of the products of certain mRNAs.

Microinjection
 7 -YFP or  7 -CFP in pcDNA3.1, together with other plasmids, were injected into SCG neurons usually in a 1:1 ratio, 18-24 hours after they were placed in culture. Microinjection was performed using an Eppendorf microinjection system on a Zeiss Axiovert 200M microscope using the following settings: 100-150 hPa injection pressure, an injection time of 0.2 seconds and constant pressure of between 40 and 50 hPa.  7 -XFP cDNA was injected at 25-50 ng/l diluted in 200 mM KCl.

Transfection of PC12 cells
PC12 cells were grown in Dulbecco's modified Eagle's medium, containing 7.5% foetal bovine serum and 7.5% horse serum. Differentiation was with serum-free medium containing nerve growth factor (NGF; 100 ng/ml murine 7S; Invitrogen, Paisley, UK), which was replenished every 48 hours. Cells were used after 5-7 days of differentiation. Six days before differentiation, PC12 cells were transfected using an Amaxa Nucleofector, according to manufacturer protocol (Lonza), DNA mixes contain GFP (0.5 g) and either control gnu shRNA or  7 shRNA (1.5 g), as previously described (Ferron et al., 2008).

Transfection of SCG neurons
SCG neurons were transfected using a biolistic PSD-1000/He unit, after 1 day in culture, according to the manufacturer's protocol (Bio-Rad). DNA mixes contained GFP (1 g) and either control gnu shRNA or  7 shRNA (3 g).

Measurement of neurite outgrowth
Neurons and PC12 cells were replated, 4-6 days after transfection, and cultured in the presence of NGF for a further 24 hours, before measurement of neurite length.

2055
Stargazin-like  7 in signalling endosomes  in PC12 cells by transfection with human  7 . Neurite outgrowth was measured following transfection with the constructs stated, and expressed as a percentage of the neurite length obtained with the negative control shRNA (Ctrl, black bar, n43), for the mixture of shRNAs against rat  7 (white bar, n79), rat  7 shRNAs plus human  7 (grey bar, n36) and human  7 alone as a control (hatched bar, n74). ***P<0.001 (One-way ANOVA and post-hoc Bonferroni test).
Cells were observed using a Zeiss 200M fluorescent microscope. The neurite length of transfected cells was determined using NeuronJ software (Meijering et al., 2004).

Immunocytochemistry and staining
Cells were fixed and permeabilised for immunocytochemistry essentially as previously described (Brice et al., 1997). The primary antibodies used were CD63 (Abcam), EEA1 (BD Biosciences), tau (Millipore), microtubule-associated protein 2 (MAP2, EnCor Biotechnology) and  7 (Moss et al., 2002). Endogenous  7 was detected using a cyanine 5-tyramide system kit (Perkin Elmer). Secondary goat antimouse (Molecular Probes, Eugene, Oregon) or goat anti-rabbit (Sigma) antibodies conjugated to Texas Red or biotin were applied at 10 g/ml and 5 g/ml, respectively. When used, Texas-Red-conjugated streptavidin was applied at 3.33 g/ml. For Epidermal Growth Factor (EGF) receptor localisation, live cells were incubated with EGF conjugated to Texas Red (EGF-TR, Invitrogen) for 1 hour at 37°C. Mitotracker Red (Invitrogen) and Lysotracker Red (Invitrogen) were used to visualise mitochondria and lysosomes in the neurites using the manufacturer's recommended staining protocol. In some experiments, apoptosis was assessed using the dye Hoechst 33258. Cells were mounted in Vectashield (Vector Laboratories, Burlingame, CA) to reduce photobleaching, and examined on a confocal laser-scanning microscope (Ziess LSM 510 Meta), using a 40ϫ (1.3 NA) or 63ϫ (1.4 NA) oil-immersion objective.

Live-cell confocal imaging
Time-series imaging of live cells was performed 24 hours after cDNA injection. Cells were maintained at 37°C, and each plate was imaged for no more than 30 minutes. Cells were selected if they exhibited xFP fluorescence in the processes and the soma. The confocal microscope was used at 512ϫ512 pixel resolution. The 458 and 514 nm lines of an argon laser were used to excite the specimen, together with 543 HeNe diode laser for CFP, YFP or dsRED excitation, respectively. Z-stacks were generated by taking slices with a thickness of 1 m.

Kymograph generation
Neurons were imaged in L15 medium containing 10 mM HEPES and incubated at 37°C throughout. Imaging was performed using the laser lines: 405 nm (4.5%) and 488 nm (2.0%). No bleed-through was present. Images were obtained using 40ϫ objective and 1.5ϫ zoom. A region of interest (500ϫ50 pixels in size) was drawn on a section of neurite between 50 and 200 m from the soma. Imaging was performed every 385 ms and images were obtained from the CFP and YFP channel with a pixel dwell time of 3.15 s. The time-series was performed over 10 minutes.

Image analysis
Image analysis for colocalisation was performed using ImageJ. Images were split into channels and adjusted to 0.15 m pixel resolution The different colour channels were then balanced for intensity and range and the non-specific noise reduced using a Gaussian filter. The 'Find maxima' function was then applied to images to count the local maxima points. The resultant maxima map was then enlarged by two pixels by using the 'dilate' function twice. The numbers of independent particles were then counted in the separate channels and in the colocalised channel. Percentage colocalisation was calculated by taking the number of colocalised particles divided by the total number of particles in the  7 channel.
Analysis of the image time-series was performed using ImageJ. Kymographs were produced using the Multiple Kymograph plug-in. A line was drawn across the neurite length and a kymograph was generated using a bin of 3 pixels. A median (2.0) filter was applied to the kymographs generated in each channel and the image intensity and range matched for each image channel. Kymographs were overlaid using the 'Projection' function set to 'maximum'.
Particle sizes were calculated from image stacks of sympathetic neurons stained for EEA1. A threshold was applied to EEA1 staining and the number and size of the particles counted in each image slice using the 'Analyze Particles' function of ImageJ. The mean size and frequency was calculated for each stack.