Human neuronal stargazin-like proteins, γ2, γ3 and γ4; an investigation of their specific localization in human brain and their influence on CaV2.1 voltage-dependent calcium channels expressed in Xenopus oocytes.

Background Stargazin (γ2) and the closely related γ3, and γ4 transmembrane proteins are part of a family of proteins that may act as both neuronal voltage-dependent calcium channel (VDCC) γ subunits and transmembrane α-amino-3-hydroxy-5-methyl-4-isoxazoleproponinc (AMPA) receptor regulatory proteins (TARPs). In this investigation, we examined the distribution patterns of the stargazin-like proteins γ2, γ3, and γ4 in the human central nervous system (CNS). In addition, we investigated whether human γ2 or γ4 could modulate the electrophysiological properties of a neuronal VDCC complex transiently expressed in Xenopus oocytes. Results The mRNA encoding human γ2 is highly expressed in cerebellum, cerebral cortex, hippocampus and thalamus, whereas γ3 is abundant in cerebral cortex and amygdala and γ4 in the basal ganglia. Immunohistochemical analysis of the cerebellum determined that both γ2 and γ4 are present in the molecular layer, particularly in Purkinje cell bodies and dendrites, but have an inverse expression pattern to one another in the dentate cerebellar nucleus. They are also detected in the interneurons of the granule cell layer though only γ2 is clearly detected in granule cells. The hippocampus stains for γ2 and γ4 throughout the layers of the every CA region and the dentate gyrus, whilst γ3 appears to be localized particularly to the pyramidal and granule cell bodies. When co-expressed in Xenopus oocytes with a CaV2.1/β4 VDCC complex, either in the absence or presence of an α2δ2 subunit, neither γ2 nor γ4 significantly modulated the VDCC peak current amplitude, voltage-dependence of activation or voltage-dependence of steady-state inactivation. Conclusion The human γ2, γ3 and γ4 stargazin-like proteins are detected only in the CNS and display differential distributions among brain regions and several cell types in found in the cerebellum and hippocampus. These distribution patterns closely resemble those reported by other laboratories for the rodent orthologues of each protein. Whilst the fact that neither γ2 nor γ4 modulated the properties of a VDCC complex with which they could associate in vivo in Purkinje cells adds weight to the hypothesis that the principal role of these proteins is not as auxiliary subunits of VDCCs, it does not exclude the possibility that they play another role in VDCC function.


Background
The mutation underlying the absence epilepsy phenotype of the allelic stargazer (stg) and waggler (wag) mutant mice occurs in a gene, cacng2, whose product, stargazin, has been hypothesized to be a neuronal voltage dependent calcium channel (VDCC) γ subunit [1]. VDCCs are intrinsically involved in the regulation of a multiplicity of Ca 2+ dependent processes in many different cell types where they are inserted into the plasma membrane as hetero-oligomeric complexes of a pore-forming α 1 subunit with auxiliary β, α2δ and possibly γ subunits [2].
Investigation of the functional influence of these stargazin-like γ proteins upon VDCCs has yielded mixed results. Some laboratories have reported that γ 2 and its close homologue γ 4 cause small hyperpolarizing shifts in the voltage dependence of steady-state inactivation [1,14,18]. This however, might be dependent upon which other auxiliary subunits are co-expressed in the VDCC complex under investigation [18]. In contrast, Chen et al. [19] showed whole cell VDCC currents from the cerebellar granule cells of stg mice, which effectively lack the γ 2 subunit, do not have significantly altered voltage-dependence of activation or inactivation compared to wild type. Other laboratories reported that γ 2 or γ 3 can significantly reduce peak current amplitudes of N-type VDCCs expressed in Xenopus oocytes, but only when co-expressed with an α2δ 1 subunit [20], and supporting this, that thalamic relay neurons from stg mice express enhanced low and high voltage-activated VDCC currents compared to wild type [21]. Furthermore, clear biochemical evidence has been generated for a direct interaction of γ 2 with the VDCC Ca V 2.2 α 1 subunit [20,22]. Another stargazin-like protein, γ 7 , which is phylogenetically distinct from γ 2 , γ 3 and γ 4 [13,15], almost abolishes the expression of Ca V 2.2 when coexpressed in vitro, and also reduces Ca 2+ currents via Ca V 1.2 and Ca V 2.1 channels [15]. However, our data indicated that the influence of γ 7 on VDCC function is to reduce α 1 subunit protein expression, a functional prop-erty unlike anything reported for the other stargazin-like proteins, which suggests that γ 7 is not a subunit of a calcium channel complex [15].
Whilst controversy surrounds the role of the γ 2 , γ 3 and γ 4 stargazin-like proteins in relation to VDCC modulation, a clear function has been determined for these proteins as chaperones for the appropriate trafficking and receptor biogenesis of alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptors [19,[23][24][25]. Consequently, these three proteins together with their close homologue γ 8 , were recently dubbed transmembrane AMPA receptor regulatory proteins (TARPs) [25]. A primary interaction (probably via transmembrane and/or extracellular regions [26]) promotes the trafficking of the GluR subunits to the plasma membrane, and a secondary interaction of the C-terminus of stargazin with PSD-95 or a similar cytoskeletal protein via a PSD-95/DLG/ZO-1(PDZ)-binding motif facilitates the lateral relocation of the glutamate receptor complex to its correct position in the post-synaptic density [19] and hence influences the number of AMPA receptors located at this site [27].
Elucidation of the differential distribution of the stargazin-like proteins coupled with studies of the physiological abnormalities underlying the epilepsy phenotypes of the mice expressing what are effectively null mutations of γ 2 has also helped to determine some of the normal functions of these proteins in the murine CNS. The stg and wag mice display a loss of the fast component of EPSC at mossy fiber to cerebellar granule cell synapses [28,29], plus reduced synaptic transmission at parallel fiber Purkinje cell synapses in wag [29]. However, the synaptic transmission to CA1 pyramidal cells (Schaffer collateral projection) in stg is not altered [28]. In situ hybridization studies have determined that murine γ 2 is normally expressed at its highest levels in the cerebellum [1,14,19]. γ 3 and γ 4 mRNAs are also detected in mouse cerebellum but γ 3 mRNA has been detected only in the Golgi neurons of the granule cell layer and is absent from the molecular layer [19]. γ 4 mRNA is localized to the Purkinje, rather than the granule cell layer [14]. Whilst this manuscript was in preparation, Tomita et al. [25] reported that γ 2 is the only TARP expressed in rat cerebellar granule cells, but like mouse, all isoforms are detected in hippocampus [1,14,19,25]. The genetic defect in stg results in the loss of γ 2 mRNA and protein and does not appear to result in upor down-regulated expression of γ 3 or γ 4 [1,22]. Collectively, these data suggest that stg mice exhibit the loss of fast synaptic transmission in the mossy fiber to granule cell synapse because γ 2 is the major and possibly the only stargazin-like protein in the cerebellar granule cells of wild-type mice and no other stargazin-like protein is expressed at sufficient levels to rescue normal AMPA receptor trafficking and maturation. The reason that synaptic transmission to CA1 pyramidal cells in stg mice is preserved is probably because, although γ 2 is not expressed in stg, the total remaining TARP expression levels in hippocampal CA1 pyramidal cells are adequate to promote normal surface expression of mature AMPA receptors at the postsynaptic membrane in this synapse.
It was therefore of great interest to examine whether the expression patterns of the human stargazin-like proteins paralleled those of mouse and if they were differentially expressed in the various cell types of each tissue. This study presents the differential distribution of γ 2 , γ 3 and γ 4 in human brain by northern blotting and more detailed immunohistochemical analysis of their expression in human cerebellum and hippocampus. In addition, we used the results of our distribution study to investigate whether the human γ 2 and γ 4 could modulate currents gated by a VDCC complex heterologously expressed in Xenopus oocytes that was assembled from the major VDCC subunits expressed in a cerebellar Purkinje cell.

Northern Blot analysis of mRNA distribution
The overall tissue distribution of human γ 2 , γ 3 and γ 4 mRNAs was analyzed by northern blotting. Specific cDNA probes representing each gene were generated as described in the methods and hybridized against a human multiple-tissue northern blot ( Figure 1A) and two brain region blots ( Figure 1B and 1C). The γ 2 -specific cDNA probe detected two mRNAs of approximately 7 kb and 3 kb. Bands of similar size were reported in a mouse multiple-tissue blot probed with a murine γ 2 cDNA probe [1]. Like mouse, the human γ 2 mRNA transcripts were detected only in brain ( Figure 1A), and were particularly abundant in cerebellum, cerebral cortex, occipital lobe, frontal lobe and temporal lobe, hippocampus and thalamus ( Figure  1B and 1C). γ 2 transcripts were detected at somewhat lower levels in medulla, putamen, amygdala and substantia nigra but only weakly in caudate nucleus. γ 2 transcripts were absent from corpus callosum and the sub-thalamic nucleus and spinal cord.
A γ 3 cDNA probe detected one 2.0-2.1 kb γ 3 mRNA transcript which like γ 2 , was exclusively localized to the brain ( Figure 1A). However, γ 3 mRNA was detected only in cerebral cortex, including occipital lobe, frontal lobe, and temporal lobe, the putamen, caudate nucleus, amygdala and hippocampus and was absent from all of the other regions probed ( Figure 1B and 1C).
The γ 4 -specific probe identified an mRNA of approximately 4 kb detected exclusively in brain ( Figure 1A). This was widely detected throughout the brain but was most prevalent in the putamen and caudate nucleus. Unlike γ 2 or γ 3 , γ 4 mRNA was also weakly detected in spinal cord.

Generation and characterization of γ specific antisera
In order to investigate the distribution of the γ 2 , γ 3 and γ 4 proteins and to analyze their expression in different neuronal cell types we generated specific antisera against each isoform. These were generated using peptide immunogens that were selected to be subtype-specific as determined by multiple sequence alignment. To confirm activity against the corresponding holopeptide and lack of cross-reactivity with the other subtypes, each of these antisera were validated against the different subunit proteins recombinantly expressed in COS-7 cells. Cells were transiently transfected with a pMT2 expression vector containing either the human γ 2 , γ 3 or γ 4 cDNAs, and 2-3 days later were analyzed by immunocytochemistry. All three γ isoforms could be readily detected and were localized largely in the plasma membrane of the cells (Figure 2A). The specificity of the antibodies (Abs) was confirmed by incubating non-transfected cells with each primary Ab, which in each case gave no immunoreactivity ( Figure 2B). Similar results were obtained when cells expressing one of the γ isoforms were incubated with Abs directed against either of the other two γ proteins (data not shown). Cells transfected with a γ cDNA but which were not permeablized after fixation also exhibited no immunoreactivity ( Figure 2C). Since the peptide sequences chosen for immunization are in the C-terminus, these results indicate this portion of recombinantly expressed γ 2 , γ 3 and γ 4 is located intracellularly. Co-transfection of VDCC subunits Ca V 2.1 and β 4 did not alter the expression pattern of the γ proteins indicating that there are no effects on trafficking and that the anti-γ Abs did not cross-react with the Ca V 2.1 or β 4 subunits (data not shown).

Immunolocalization of γ 2 , γ 3 and γ 4 in human cerebellum and hippocampus
Immunohistochemical analysis was carried out to examine the differential expression of γ 2 , γ 3 and γ 4 in human cerebellum and hippocampus. Control experiments to measure non-specific binding to tissue sections were performed by pre-absorbing antisera overnight with 100 µM of the relevant peptide ( Figures 3E and 3I, 4C and 4D, 5C and 5D and 6B) and by omitting the primary antibody (data not shown). Figure 3 shows the pattern of staining observed in the cerebellum using γ 2 , γ 3 and γ 4 specific antisera. Moderate γ 2specific staining was seen in the molecular layer suggesting expression in the dendrites of cerebellar Purkinje neurons ( Figure 3A). Small cell bodies in the molecular layer that were most likely stellate or basket cell interneurons also stained positively. The γ 2 immunostaining was also strong in the soma of Purkinje neurons extending to the dendrites, but decreasing in intensity following the first few bifurcations ( Figure 3A). Cerebellar granule cells were Northern blot analysis of the human γ 2 , γ 3 , and γ 4 mRNA transcripts Figure 1 Northern blot analysis of the human γ 2 , γ 3 , and γ 4 mRNA transcripts. A. Human multiple-tissue northern blots revealed that the γ 2 , γ 3 , and γ 4 are exclusively detected in the brain. B & C. Brain region blots determined that the γ 2 and γ 4 are almost ubiquitously expressed, but at differential levels in the same tissues. γ 3 is more specifically localized to cerebral cortex, amygdala, caudate nucleus and hippocampus. Size markers were from an RNA ladder provided on the blot by BD Biosciences Clontech and the β-actin control probe results are displayed in the bottom panels.  Figure 3B), although assessment of immunostaining in this cell type was difficult since the majority of cell volume is comprised of the nucleus. The strongest γ 2 immunostaining in this region was actually in the interneurons. Figure 3C shows strong γ 2 immunostaining in the cell bodies of the dentate cerebellar nucleus with only weak to moderate staining in the surrounding neuropil. Very weak γ 3 staining was observed in Purkinje cell bodies and in the interneurons of the granule cell layer ( Figure  3D). Staining of the molecular layer neuropil and the granule cells is comparable to the peptide pre-absorption control ( Figure 3E). It is therefore possible that the γ 3 protein is poorly represented in these particular cell types and not completely absent from the cerebellum. No γ 3 immunoreactivity was observed in the dentate cerebellar nucleus (data not shown).

Bloo
Staining of a human cerebellar folium for γ 4 is shown in Figure 3F. The molecular layer and adjacent granule cell layer was lightly to moderately stained, but the staining of the Purkinje cell bodies and dendrites was striking. γ 4 immunostaining extended well into the Purkinje cell dendritic arbors from the cell body indicating that γ 4 expressed well throughout this cell type. The cell bodies that stained positively in the molecular layer were small interneurons. In the granule cell layer the interneurons stained strongly for γ 4 whereas the granule cells appear unstained ( Figure 3G). In the dentate cerebellar nucleus ( Figure 3H), γ 4 immunostaining was detected strongly in the perisomatic neuropil with only weak staining in the cell bodies. Interestingly, this staining pattern was almost the complete inverse of the γ 2 immunostaining in the same region ( Figure 3C).

Hippocampus and dentate gyrus
Immunohistochemistry revealed differential expression patterns for γ 2 , γ 3 and γ 4 in the different regions and cell types of human hippocampus and dentate gyrus ( Figure  4, 5 and 6). When incubated with the anti-γ 2 Ab, generalized staining of the cell layers and neuropil throughout the hippocampal formation was observed. In Figure 4A the alveus, stratum oriens most adjacent to the CA1 pyramidal layer, and the pyramidal layer all stained moderately for γ 2 . The stratum radiatum and lacunosummoleculare stained much more weakly. The CA2 and CA3 pyramidal layers stain well for the γ 2 subunit ( Figure 4B) and the intensity of staining in the alveus increased slightly in this region. It is however apparent that there is only a very weak staining of the strata oriens and radiatum in the CA2/3 regions; yet, moderate immunostaining is detected in the stratum lucidum. Weak staining is visible in the stratum lacunosum-moleculare. In the same section we also observed moderate γ 2 staining in the molecular layer of the dentate gyrus and strong immunostaining in the granule cells (examined at higher magnification on Figure 4H). The γ 2 staining was weak or absent in the polymorphic layer, but distinct fibers were stained which course across this region. Moderate staining of cell bodies in CA4 was also observed ( Figure 4B, top left portion of the panel). Pre-absorption controls using the γ 2 peptide immunogen ( Figures 4C and 4D) show the background staining in a serial section of the same hippocampus as in Figures 4A and 4B. The perisomatic staining in the stratum pyramidale, in the region of the CA1 towards the CA2 was moderate, with slightly more intense somatic staining ( Figure 4E). An apparently higher level of γ 2 staining in the CA2 and CA3 regions (Figures 4F and 4G) compared to CA1 was probably due to a higher density of pyramidal cell soma in these regions compared to the CA1 region [30].
Staining for γ 3 was absent from the alveus, stratum oriens, stratum radiatum and lacunosum-moleculare, but the pyramidal layer of the CA1, CA2 and CA3 regions of Transient expression of γ 2 , γ 3 , and γ 4 in COS-7 cells Positive staining for γ 2 , γ 3 , or γ 4 (red) is strongly localized to the membranes of permeablized cells transfected with the individual γ 2 , γ 3 , or γ 4 cDNAs. Little immunostaining is observed in the cytoplasm between the membrane and nucleus (blue). B. Non-transfected cells did not stain for γ 2 , γ 3 , or γ 4 using the Abs generated in this study. C. Cells transfected with γ 2 , γ 3 , or γ 4 cDNA, but not permeablized during the staining process do not show immunoreactivity for the appropriate anti-γ Abs. D. When co-transfected together with the Ca V 2.1 and β 4 VDCC subunit cDNAs, the expression patterns of γ 2 , γ 3 , or γ 4 were unaltered. Scale bars in all panels represent 10 µm.
Expression of the γ 2 , γ 3 , and γ 4 stargazin-like proteins in human cerebellum Figure 3 Expression of the γ 2 , γ 3 , and γ 4 stargazin-like proteins in human cerebellum. A. γ 2 immunoreactivity in the molecular layer (m) and part of the adjacent granule cell layer (g) in a cerebellar folium. Strong γ 2 immunostaining was seen in the Purkinje cell somata (p) that continued into the dendrites, observed as moderate staining of the molecular layer neuropil. Cell bodies in the molecular layer with positive γ 2 immunoreactivity are small interneurons (i). B. In the granule cell layer, the granule cells (g) are moderately stained for γ 2 although the strongest immunostaining in this region is actually in the interneurons (i). C. The dentate cerebellar nucleus (dcn), displayed strong γ 2 immunostaining in the cell bodies of this nucleus with only weak to moderate staining in the surrounding neuropil. In this particular section, cell nuclei are stained blue by Mayer's hematoxylin counterstain. D. Cerebellar folia displayed little to no γ 3 immunostaining in the central white matter (wm) and molecular layer (m). Weak immunostaining was observed in the Purkinje cell bodies and the interneurons of the granule cell layer (g). E. Pre-absorption control, with the peptide immunogen for the γ 3 Ab in the adjacent section to Fig. 3D. F. The cell bodies of the Purkinje cells (p) stained strongly for γ 4 and this extended well into the Purkinje cell dendrites (d). The surrounding molecular layer neuropil (m) displayed light-moderate staining, as did the granule cell layer (g). Cell bodies in the molecular layer immunoreactive to the γ 4 Ab were small interneurons (i). G. Granule cell layer interneurons (i) stained much more strongly for γ 4 than the granule cells (g). H. The dcn gave a strong γ 4 immunostaining signal in the perisomatic neuropil with only weak staining in the nucleus cell bodies. I. Immunizing peptide pre-absorption control for the γ 2 Ab. Pre-absorption of the γ 3 or γ 4 Abs with their immunizing peptides produced almost identical results. Scale bars in all panels represent 25 µm. γ 2 immunohistochemistry in human hippocampus Figure 4 γ 2 immunohistochemistry in human hippocampus. A. In the CA1, alveus (a), stratum oriens (o) most adjacent to the pyramidal layer, and the pyramidal layer (p) all stain moderately for γ 2 . The stratum radiatum (r) and lacunosum-moleculare (lm) stain much more weakly. B. Strong detection of γ 2 protein occurred in the alveus and pyramidal layers of CA2 and CA3 regions. Moderate immunostaining is detected in stratum lucidum (l). Staining of the strata oriens and radiatum was very weak in this region. The dentate gyrus molecular layer (ML) stains moderately for γ 2 and the granule layer strongly (GL). γ 2 staining is weak or absent in the polymorphic layer (PL). Cells in the CA4 show moderate staining. C and D. Serial sections to those displayed in panels A and B incubated with synthesis peptide pre-absorbed γ 2 Ab. E. In the region of the CA1 towards the CA2 the perisomatic staining is light to moderate with similar staining of the cell bodies. F. In the CA2, staining of neuropil and cell bodies is more intense than in CA1. G. Staining of the soma of the CA3 hippocampal neurons is strong with moderate staining of the surrounding neuropil. H. The dentate gyrus shows moderate staining of the molecular layer and dense immunostaining in the granule layer. The polymorphic layer is weakly stained. Scale bar in panels A-D represents 250 µm, in panels E-H 25 µm. The dotted lines in panels A and B represent the border between the CA4 region of the hippocampus and the polymorphic layer of the dentate gyrus. γ 3 immunohistochemistry in human hippocampus Figure 5 γ 3 immunohistochemistry in human hippocampus. A. The pyramidal cell layer (p) of the CA1 region exhibited moderate to strong staining for γ 3 . Weak staining was observed in the stratum oriens (o) and stratum radiatum (r). The alveus (a) and lacunosum-moleculare (l-m) are in general, absent of immunoreactivity. B. More intense γ 3 immunostaining was observed in the CA2 and CA3 pyramidal cell layers than in CA1. Strong staining was also detected in CA4 cell bodies but surrounding neuropil is devoid of immunoreactivity. C and D. Sections adjacent to those displayed in panels A and B incubated with immunizing peptide-pre-absorbed γ 3 Ab. E. Moderate to strong staining of CA1 pyramidal cell bodies and weaker staining in the surrounding neuropil. F. Intense staining of the pyramidal cells of the CA3 with moderate perisomatic staining. G. Cell bodies in the subiculum (Sub) are moderately stained and neuropil staining is very weak. Neuropil staining in panels E, F and G may reflect relative cell densities in each region. H. Granule cells of the dentate gyrus were stained moderately to strongly by the γ 3 Ab. The immunoreactivity is mainly in the soma of these cells, but can also be seen in the early branches of the granule cells dendritic trees, which extend into the lightly stained molecular layer. The neuropil of the polymorphic layer is almost devoid of staining. Scale bars in panel A-D represent 250 µm, in panels E-H 25 µm. Ammon's horn strained strongly, with lesser immunostaining in the subiculum and CA4 regions (Figure 5A and 5B). Closer examination revealed that the pyramidal cell bodies stained strongly for γ 3 and this staining was strongest in CA3 and became progressively weaker through the CA2 and CA1 regions and into the subiculum. High power images of CA1 ( Figure 5E), CA3 ( Figure  5F), and subiculum ( Figure 5G) determined that this was most likely to be caused by increased cell densities in the CA2/3 regions rather than more intense staining of pyramidal neurons, although perisomatic staining was more intense in the CA3 region than CA1, and absent from the subiculum. Visible in Figure 5A and 5B, and also observed at an increased magnification in Figure 5H, the granule layer of the dentate gyrus is distinctly labeled whereas the molecular and polymorphic layers display weak or no immunostaining. The granule cells stained moderately to strongly for γ 3 and the immunoreactivity was mainly in the soma of these cells, but could also be seen in the early branches of the granule cell dendritic trees, which extend into the lightly stained molecular layer ( Figure 5H).
At the macroscopic level, medium to strong γ 4 immunostaining was seen throughout the pyramidal layers of Ammon's horn and in the alveus ( Figure 6A). The alveus stained strongly for γ 4 as did the cell bodies in the pyramidal layer of all CA regions. This was observed more clearly at high power with a similar high level of staining of pyramidal cell bodies throughout the CA1-4 regions, but with the most intense perisomatic staining observed in the CA2/3 region (Figure, 6C, 6D, and 6E). Weak to moderate staining was observed throughout the neuropil of the strata surrounding the pyramidal layers in all regions. In the dentate gyrus, the granule cell layer appears to stain slightly more strongly than the adjacent molecular or polymorphic regions ( Figure 6F). This also may be an artifact of the cell density in the granule layer rather than increased expression of the γ 4 protein in the dentate granule cell bodies compared to their processes.

The effects of γ 2 and γ 4 on the biophysical properties of Ca V 2.1 calcium channels
Since the role of the stargazin-like proteins in relation to VDCC function remains controversial we examined the effects of γ 2 and γ 4 on the biophysical properties of Ca V 2.1 calcium channels. Together with the Ca V 2.1 α subunit, we co-expressed a β 4 subunit with or without an α2δ 2 subunit. These VDCC subunits are known to be highly expressed in cerebellar Purkinje cells [31][32][33][34][35][36][37][38], a cell type that, according to the immunohistochemistry data presented herein, strongly expresses both γ 2 and γ 4 . Figure 7A and 7B show the current-voltage relations of Ca V 2.1/β 4 and Ca V 2.1/β 4 /α2δ 2 VDCCs respectively, expressed in Xenopus oocytes either in the presence or absence of γ 2 or γ 4 . Neither γ 2 nor γ 4 produced a statistically significant shift in the half-maximal voltage of activation (V 50act ) of Ca V 2.1/β 4 ( Table 1). The slope factor (k act ), maximum conductance (G max ), and peak current amplitude at +10 mV were also very similar to controls on co-expression of either γ 2 or γ 4 in the absence the α2δ 2 subunit. Co-expression of the α2δ 2 subunit was sufficient to significantly increase the G max and peak current amplitude of a Ca V 2.1/ β 4 Ca 2+ channel complex ( Figure 7A and 7B). However, the increase was not affected by the additional co-expression of either γ 2 or γ 4 .

Differential distribution of human stargazin-like γ mRNAs
The northern blot analysis described herein detected the human γ 2 , γ 3 and γ 4 mRNAs only in the brain. Similar mRNA distributions and transcript sizes have been reported by in situ hybridization and northern blot for the mouse orthologues [1,14,19] and these data are supported by western blots [22]. We do not however rule out the possibility that human γ 2 , γ 3 and γ 4 may be expressed in some non-CNS tissues that are not included on the multiple tissue northern blot used in this study. Other laboratories that employed highly sensitive reversetranscriptase polymerase chain reaction expression analyses have reported that γ 4 in particular is expressed in some non-neuronal tissues [13,17]. The γ 2 and γ 4 transcripts were detected in most of the brain regions probed and mainly in the same tissues. However, in some regions where both isoforms are detected, for example the cerebellum or thalamus, γ 2 expression is higher than γ 4 , whereas in the basal ganglia regions of putamen and caudate nucleus γ 4 mRNA is the more highly represented transcript. γ 3 is much more selectively detected, but its expression is coincident with both γ 2 and γ 4 in all regions in which it was detected by northern blot.
An additional observation was that these stargazin-like γ proteins are differentially detected in some of the nuclei that comprise the basal ganglia, a region believed to be involved in the planning and programming of movement, or more broadly in the processes by which intention is converted into voluntary action. The putamen has a heterogeneous neuronal γ population with γ 4 mRNA possibly the most prevalent of the three γ transcripts investigated herein, as has been previously observed in both mouse and rat brain [14,25]. The caudate nucleus also expresses transcripts for all three stargazin-like γ proteins, however the signal detected with the γ 2 probe was extremely weak in comparison to those for γ 3 and γ 4 . On the other hand, γ 2 was the most prevalent species detected in substantia nigra, with faint detection of γ 4 and no γ 3 signal. No signals for any of the γ transcripts were detected in the subthalamic nucleus. Further immunohistochemical and electrophysiological investigation will be required to elu-cidate why γ 2 , γ 3 and γ 4 have such differential distributions in these nuclei.

In vitro expression of human stargazin-like proteins and antibody specificity
The transient expression of cloned human γ 2 , γ 3 and γ 4 cDNAs in COS-7 cells served several purposes: It deter-Influence of γ 2 and γ 4 upon Ca V 2.1/β 4 ± α2δ 2 VDCC currents expressed in Xenopus oocytes Figure 7 Influence of γ 2 and γ 4 upon Ca V 2.1/β 4 ± α2δ 2 VDCC currents expressed in Xenopus oocytes.A The peak currentvoltage relationships for Ca V 2.1/β 4 (n = 16), Ca V 2.1/β 4 /γ 2 (n = 14) and Ca V 2.1/β 4 /γ 4 (n = 11) show no significant differences in any parameters ( Table 1). The bottom panels display representative traces recorded from a single oocyte injected with each subunit combination investigated. B. The peak current-voltage relationships for Ca V 2.1/β 4 /α2δ 2 (n = 43), Ca V 2.1/β 4 /α2δ 2 /γ 2 (n = 34) and Ca V 2.1/β 4 /α2δ 2 /γ 4 (n = 20) show no significant differences in any parameters (Table 1). In both panels A and B, only representative traces from -50 to +10 mV are displayed for reasons of clarity. C. Mean steady state inactivation data for I Ba recorded in 10 mM Ba 2+ from Xenopus oocytes injected with Ca V 2.1/β 4 /α2δ 2 (n = 18) or Ca V 2.1/β 4 /α2δγ (n = 18) D. Ca V 2.1/β 4 / α2δ 2 (n = 20), or Ca V 2.1/β 4 /α2δ 2 /γ 4 (n = 20). Co-expression of either γ 2 or γ 4 produced data almost identical to the control. E. Oocytes injected with Ca V 2.1/β 4 (n = 22), Ca V 2.1/β 4 γ 2 (n = 16) and Ca V 2.1/β 4 γ 4 (n = 9) normalized to the maximum current I max and fitted with a single Boltzmann function. The numerical data for parameters defining the steady-state inactivation relationships are displayed in Table 2. A mined that the human γ 2 , γ 3 and γ 4 cDNA clones expressed in vitro; it demonstrated the specificity of three anti-γ Abs, one generated to detect each of the three isoforms; it revealed that the stargazin-like γ 2 , γ 3 and γ 4 all localize to the plasma membrane of COS-7 cells when expressed either alone or in combination with other VDCC subunits and finally it showed that COS-7 cells do not endogenously express γ 2 , γ 3 or γ 4 . Furthermore, because the antiγ Abs, all of which were designed to detect epitopes in the C-terminus after the fourth predicted transmembrane segment, failed to stain COS-7 cells that had not been permeablized, it was established that the C-terminus is localized to the membrane and inside the cell. If predictions of secondary structure as envisaged by other laboratories are correct, a tetra-spanning transmembrane conformation will also place the N-terminus on the cytoplasmic side of the membrane [12,16].

The distribution of the γ 2 , γ 3 and γ 4 in cerebellum
This study is the first immunohistochemical analysis of stargazin-like γ proteins in the human CNS. In the cerebellum, we observed that γ 2 was detected in molecular, granule and Purkinje cell layers as has been previously reported for mouse γ 2 protein [22] and mRNA [1,14,19]. A major difference between human and rodent immunostaining patterns was observed for γ 4 , which was detected as very high levels in human cerebellar Purkinje cell bodies and processes, an observation not reported for rat or mouse [22,25]. The detection of γ 3 protein expression in the human cerebellum, albeit at extremely low levels, was similar to the γ 3 mRNA detection patterns reported in mouse or rat cerebellum [19,25] but disagree with the findings on our northern blots and the in situ hybridization data of Klugbauer et al. [14]. The positive, albeit weak detection of γ 3 protein in cerebellar interneurons and low levels detected in Purkinje cell bodies suggests that γ 3 is To analyze the voltage-dependent activation, current-voltage (I-V) relationships ( Figure 7A and Figure 7B) were fitted with a modified Boltzmann equation (see methods) to calculate V 50act , the mid-point of voltage dependence of activation, k act , the slopefactor and G max , the maximum conductance. Values are expressed as mean ± S. E. M. of the number of replicates, n. One-way analysis of variance (ANOVA) tests determined no statistically significant changes in any pairs of data differing by the presence of a γ 2 or γ 4 P < 0.05 was considered significant). Statistically significant differences observed between the Ca V 2.1/β 4 and Ca V 2.1/β 4 /α2δ 2 channels are indicated by * (P < 0.05). . Values are expressed as mean ± S. E. M. of the number of replicates, n. In the presence of an α2δ 2 subunit, neither γ 2 nor γ 4 caused statistically significant modulation of the measured parameters according to an un-paired Student's t-test (P < 0.05 was considered significant). Equally, when the α2δ 2 subunit was omitted from the injection mixture, ANOVA determined no statistically significant changes in any pairs of data differing by the presence of a γ protein.
expressed in distinct types of neurons in human cerebellum at levels too low to be detected by some hybridization conditions.
The almost inverse staining patterns of γ 2 and γ 4 in the dentate cerebellar nucleus (DCN) was striking. Much of the perisomatic neuropil surrounding the DCN somata consists of afferents from the Purkinje cells and is stained particularly strongly for γ 4 while the DCN cell bodies stained strongly for γ 2 but were devoid of γ 4 staining. It is therefore a reasonable assumption that γ 4 is pre-synaptically localized in the Purkinje cell afferents to the DCN whilst γ 2 localizes to the post-synaptic regions of the DCN cell bodies. Indeed this observation holds with the finding that γ 4 expressed well throughout Purkinje cell processes. γ 4 also appears to be localized in the GABAergic neurons of the cerebellum more than the excitatory glutamatergic cell types. In addition to showing strong immunostaining throughout Purkinje cells, γ 4 is detected in the interneurons of the molecular layer and also the Golgi interneurons of the granule layer. Whilst γ 2 is also detected in all these cell types, the γ 4 immunostaining is noticeably lower, if not absent from the excitatory granule cells, and is absent from the DCN cell bodies.

The distribution of γ 2 , γ 3 and γ 4 in hippocampus
This study has determined that γ 2 , γ 3 and γ 4 show differential but overlapping expression patterns in the human hippocampus and dentate gyrus. As is the case for the cerebellum, the expression patterns of γ 2 and γ 4 more closely resemble one another than that of γ 3 . The γ 2 , and γ 4 proteins were detected throughout the hippocampus and dentate gyrus, although there were variations in the staining of cell bodies, dendrites and neuropil in the different sub-fields. γ 3 localized more specifically in the neuronal cell bodies of the hippocampus and dentate gyrus. This possibly indicates that γ 2 and γ 4 are involved in synaptic modulation of neurotransmission throughout the cell, whereas γ 3 is solely involved in functions such as regulation of VDCCs or AMPA receptors in the cell soma.

Ca V 2.1/β 4 VDCC currents are not modulated by human γ 2 or γ 4 co-expression in the presence or absence of an α2δ 2 subunit
The co-expression of γ 2 or γ 4 with Ca V 2.1/β 4 VDCCs in the absence or presence of the α2δ 2 subunit, did not significantly affect peak current amplitude or any of the activation or inactivation properties. These data agree closely with the findings of Chen et al. [19] who recorded whole cell Ca 2+ currents from stg and wild type cerebellar granule cells. They reported that absence of γ 2 neither altered the I-V relationship of the native whole cell Ca 2+ current nor did it significantly modulate the steady-state inactivation properties of VDCC current compared to wild type. Although they did not use pharmacological agents to isolate specific components of the whole cell Ca 2+ current which might have highlighted subtle changes particular to the P/Q-, N-, R-, or L-type currents present in this cell type [39,40], it is unlikely that another known γ isoform could functionally substitute for γ 2 to maintain normal VDCC function in that instance because distribution studies have shown that they are probably not expressed in this cell type [14,19,25]. This indicated that even if γ 2 was associated directly or indirectly with a VDCC complex in cerebellar granule cells [20,22], it did not modulate the high voltage activated VDCC I-V relationship or inactivation properties.
More recent patch clamp recordings from stg thalamic relay neurons showed a 45% increase in HVA VDCC peak current densities compared to wild type [21] consistent with a previous report that γ 2 inhibited high voltage activated VDCC peak current amplitude by 37-40% when expressed in Xenopus oocytes [20]. Why our data and those of Chen et al. [19] are so different from these results may be explained by differences in subunit combinations expressed in granule cells and thalamic relay neurons and between the two in vitro studies. Stargazin-like proteins might be able to directly modulate VDCC complexes consisting of particular subunit combinations, but are unable to reproduce this influence on other subunit complexes if they required additional interacting proteins not present in the Xenopus oocyte or cerebellar granule cells. The electrophysiological data therefore cannot be used as a strong argument to warrant considering these γ proteins as an integral part of high voltage-activated Ca 2+ channels because modulation of current properties has not been reproducible between different studies. Nevertheless, there is biochemical evidence for the association of γ 2 and γ 3 with the N-type Ca V 2.2 channel [20,22] and it is quite possible that whilst γ 2 , γ 3 and γ 4 readily associate with certain neuronal VDCC complexes, specific environmental conditions must be met for them to exert a measurable biophysical influence.

Conclusions
Human γ 2 , γ 3 and γ 4 stargazin-like proteins (or TARPs) are detected solely in the CNS. On the whole, their differential distributions closely parallel those of their rodent orthologues as observed by northern blot, in situ hybridization, western blot and immunohistochemistry [1,14,19,22,25] with some notable exceptions. The differential expression pattern of each isoform among the cell types present in human cerebellum and hippocampus predicts specific roles for each subtype in neuronal function, and possibly even segregated VDCC-γ or AMPA receptor-γ complexes [25]. The results of our electrophysiology experiments support the notion that γ 2 , γ 3 and γ 4 stargazin-like proteins are not "subunits" of VDCCs in the true sense of the word. Nevertheless, we do not discount the possibility that they may interact with VDCCs and possibly influence trafficking, assembly or integration of VDCCs and AMPA-receptor function in their native environment or that to modulate VDCC current they require other factors not endogenously expressed in Xenopus oocytes.

cDNA sources and synthesis
Human brain total RNA was purchased from Invitrogen (Paisley, UK) and used to generate cDNA using the Superscript Pre-amplification System (Invitrogen) primed with random hexamers according to the manufacturer's instructions.

Isolation and cloning of the γ 2 , γ 3 and γ 4 cDNAs
The complete open reading frame (ORF) of human CACNG2 cDNA was amplified from 25 ng human brain cDNA by PCR (cycling parameters: 98°C for 1 min, then 30 cycles of 98°C for 30 s, 55°C for 30 s, 72°C for 2 min, followed by a final 10 min extension step at 72°C) containing Pfu polymerase and 25 pmol each of the gene specific primers (GSPs), 5'-GCGGCCGCACCATGGGGCTGTTTGATC-3' and 5'-GCTAGCCTCGAGTTAGTGTTTATATAATGAAGAA-3'. These amplify the ORF of CACNG2, with a 5' extension of a Not I restriction site and partial Kozak sequence for initiation of translation in vertebrates (ACC) [41] and a 3' extension of Xho I and Nhe I restriction enzyme sites. Amplified fragments of the correct size were purified from agarose gels using the Qiaex II kit (Qiagen, Crawley, UK), and cloned into pCR2.1-TOPO vector (Invitrogen). Positive colonies identified by blue/white screening were confirmed by EcoR I restriction digest of purified plasmid and were sequenced on both strands using T7 (5'-TAATAC-GACTCACTAT AGGG-3') and M13R (5'-CAG-GAAACAGCTATGAC-3') universal primers and gene specific primers in an automated dye terminator sequencer (Applied Biosystems, Warrington, UK).

Antibodies
Multiple sequence alignments and basic local alignment search tool (BLAST) searches compared the protein sequences of all known stargazin-like γ proteins with one another and with other proteins in the public databases. This identified regions of lowest homology between γ 2 , γ 3 and γ 4 that were not present in any other known proteins. The following peptides, TARATDYLQASAITRIPS (γ 2 , amino acids 211-228), FHNSTPKEFKESLHNNPAN (γ 3 , amino acids 291-309) and VHDFFQQDLKEGFHVSMLN (γ 4 , amino acids 303-321), were synthesized by standard solid-phase techniques at Severn Biotech (Kidderminster, UK) to generate specific polyclonal antibodies (Abs). Each was coupled to the carrier protein tuberculin purified protein derivative (PPD) using sulpho-Succinimidyl4-[Nmaleimidomethyl] cyclohexane-1-carboxylate (SMCC) (Pierce, Tattenhall, UK) via a Cys residue added at the Nterminus during synthesis. To raise polyclonal anti-γ 2 , anti-γ 3 and anti-γ 4 Abs, the resulting conjugates were used to immunize Bacille Calmette-Guerin (BCG)-sensitized Dutch rabbits at monthly intervals [43]. The immune response was monitored by indirect enzyme-linked immunoabsorbent assay (ELISA) with free peptide-coated micro-titer plates. Immunoglobulins from the terminal bleeds were purified using immobilized peptide antigen columns (Sulfo-link, Pierce). Each anti-γ Ab was checked for specificity for its target by immunocytochemistry. COS-7 cells transfected with a single stargazin-like γ protein cDNA were examined for positive staining following incubation with the appropriate affinity purified Ab. Control slides were also examined for cross-reaction of the primary Ab with non-transfected COS-7 cells, stargazin-like γ proteins other than the target against which the primary Ab was intended to bind (data not shown) and the VDCC α 1 , α2δ, and β subunits used in this investigation. Finally, positive staining of target protein was abolished following overnight pre-incubation (at 4°C) of the primary Ab with a 10 × molar excess of the peptide against which it was raised (data not shown).
Cell culture and transfections COS-7 cell were cultured as previously described [44]. Transfection was performed using the Geneporter transfection reagent (Gene Therapy Systems, San Diego, CA). Cells were plated onto coverslips 2-3 h prior to transfection. The DNA and Geneporter reagent (2 µg and 10 µl, respectively) were each diluted in 500 µl of serum-free medium, mixed, and applied to the cells. The α 1 , β, α2δ and γ cDNAs were mixed and transfected in a 1:1:1:1 ratio by DNA weight. If a particular cDNA was absent from the transfection, substitution with blank pMT2 vector maintained correct subunit ratios. After 3.5 h, 1 ml of medium containing 20% serum was added to the cells, which were then incubated at 37°C for 3 days. Prior to staining, cells were re-plated using a non-enzymatic cell dissociation solution (Sigma, Dorset, UK) and maintained at 27°C for between 2 and 6 h.
Immunocytochemistry COS-7 cells were fixed and permeablized for immunocytochemistry essentially as previously described [44]. Primary Abs, affinity purified anti-γ 2 , anti-γ 3 and anti-γ 4 were used at 0.2 µg/ml. Secondary biotin conjugated or goat anti-rabbit (Sigma, Dorset, UK) Ab was applied at 5 µg/ ml. Texas red-conjugated streptavidin was applied at 3.3 µg/ml. The nuclear dye 4', 6-diamidino-2-phenylindole (DAPI, 300 nM, Molecular Probes) was also used to visualize the nucleus. Cells were examined on a confocal scanning laser microscope (Leica TCS SP, Milton Keynes, UK). All images were scanned sequentially to eliminate crosstalk and photomultiplier settings kept constant in each experiment.
containing (in mM): Ba(OH) 2 , 10; TEA-OH, 80; NaOH, 25; CsOH, 2; and HEPES, 5 (pH7.4 with methanesulfonic acid). In all experiments, oocytes were injected with 30-40 nl of a 100 mM solution of K3-1, 2-bis (aminophenoxy) ethane-N, N, N', N'-tetra-acetic acid (BAPTA) in order to suppress endogenous Ca 2+ -activated Clcurrents. Recording microelectrodes, were pulled from thick-walled borosilicate glass capillary tubing with the following dimensions: 1.5 mm outer diameter, 1.0 mm bore diameter and with an internal 0.1 mm fiber (Plowden and Thompson, Stourbridge, UK). The TEVC pipettes were pulled using a P-87 Flaming/Brown microelectrode puller (Sutter Instrument Company, Novato, CA). Electrodes contained 3M KCl and had resistances of 0.3-2 MΩ. The holding potential (V H ) was -100 mV. Membrane currents were recorded, amplified, low-pass filtered at 1 kHz using a Geneclamp 500 B amplifier, digitized through a Digidata 1200 interface (Axon Instruments, Foster City, CA) and stored on a PC using data acquisition software pClamp 6.02 (Axon Instruments). In all cases currents were leak subtracted on-line by a P/4 protocol. Additional analyses including calculation of means, standard error of the mean (S. E. M.), significance (unpaired Student's ttests, where applicable) and curve fitting were calculated using Origin 5.0 (OriginLab Corporation, Northampton, MA). One-way analysis of variance (ANOVA) and posthoc tests were performed using software available at http:/ /faculty.vassar.edu/lowry/VassarStats.htm. Where mean values are presented they are shown as mean ± S. E. M. (with n depicting the number of oocytes from which the mean was calculated). Statistical significance was defined as P < 0. 05. Current-voltage (I-V) relation curves generated from currents activated by a 200 ms long depolarizing pulse were fitted with a combined Boltzmann and linear fit function: I = G max (V -V rev ) / (1 + exp(-(V -V 50act ) / k)) where I is the whole cell current amplitude, G max is the maximum slope conductance, V 0.5act is voltage of the midpoint of activation, V rev is the reversal potential and k is the slope factor for activation.
Steady-state inactivation data were generated from currents activated by a 100 ms long depolarizing pulse from the holding potential (V H ) to 0 mV immediately after a 25 s conditioning pre-pulse between -100 and 0 mV. Current amplitudes were normalized to the maximum amplitude and fitted with a Boltzmann function of the form: I / I max = 1 / (1 + exp((V -V 50inact ) / k)) where I / I max is the normalized peak current, V 50inact is the voltage for the mid-point of inactivation, V is the conditioning voltage and k is the slope factor for inactivation.