α2δ-4 and Cachd1 Proteins Are Regulators of Presynaptic Functions

The α2δ auxiliary subunits of voltage-gated calcium channels (VGCC) were traditionally regarded as modulators of biophysical channel properties. In recent years, channel-independent functions of these subunits, such as involvement in synapse formation, have been identified. In the central nervous system, α2δ isoforms 1, 2, and 3 are strongly expressed, regulating glutamatergic synapse formation by a presynaptic mechanism. Although the α2δ-4 isoform is predominantly found in the retina with very little expression in the brain, it was recently linked to brain functions. In contrast, Cachd1, a novel α2δ-like protein, shows strong expression in brain, but its function in neurons is not yet known. Therefore, we aimed to investigate the presynaptic functions of α2δ-4 and Cachd1 by expressing individual proteins in cultured hippocampal neurons. Both α2δ-4 and Cachd1 are expressed in the presynaptic membrane and could rescue a severe synaptic defect present in triple knockout/knockdown neurons that lacked the α2δ-1-3 isoforms (α2δ TKO/KD). This observation suggests that presynaptic localization and the regulation of synapse formation in glutamatergic neurons is a general feature of α2δ proteins. In contrast to this redundant presynaptic function, α2δ-4 and Cachd1 differentially regulate the abundance of presynaptic calcium channels and the amplitude of presynaptic calcium transients. These functional differences may be caused by subtle isoform-specific differences in α1-α2δ protein–protein interactions, as revealed by structural homology modelling. Taken together, our study identifies both α2δ-4 and Cachd1 as presynaptic regulators of synapse formation, differentiation, and calcium channel functions that can at least partially compensate for the loss of α2δ-1-3. Moreover, we show that regulating glutamatergic synapse formation and differentiation is a critical and surprisingly redundant function of α2δ and Cachd1.


Introduction
Voltage-gated calcium channels (VGCC) are multi-subunit protein complexes consisting of a pore-forming α 1 and the auxiliary β and α 2 δ subunits, which are traditionally regarded as modulators of biophysical channel properties [1][2][3]. In the human genome, four genes (CACNA2D1-4) encode for four α 2 δ proteins (α 2 δ-1 to α 2 δ-4), which all are composed of a large glycosylated α 2 peptide that is likely linked to a GPI membrane-anchored δ peptide [1,4]. In recent years, a number of studies have linked human α 2 δ genes to neurological and neuropsychiatric disorders, such as epilepsy, autism, schizophrenia, and anxiety disorders (reviewed [5,6]). Moreover, α 2 δ proteins were recently recognized as important regulators of synaptic functions, some of which may be independent of calcium channels. For example, α 2 δ-1 has been identified as a postsynaptic receptor for the glia-secreted thrombospondins which mediate excitatory synapse formation [7,8]. Loss of α 2 δ-2 was shown to affect cerebellar climbing fiber synapses [9] and postsynaptic calcium signaling function of cerebellar Purkinje neurons [10][11][12], as well as the structure and function of auditory hair cell synapses [10]. Moreover, in CNS neurons, α 2 δ-2 functions as a trans-synaptic organizer, recruiting postsynaptic GABA A receptors [11]. α 2 δ-3 is important for the formation of auditory nerve fiber synapses in mice [12], while invertebrate α 2 δ-3 homologues are required for the presynaptic development of motoneurons [13,14]. In contrast to α 2 δ-1, -2 and -3, which are highly expressed in the brain, the fourth isoform, α 2 δ-4, only shows very low expression levels in the cortex and hippocampus [15]. However, α 2 δ-4 is specifically expressed in the photoreceptor cells of the retina, where it mainly interacts with the Ca V 1.4 presynaptic L-type channel, and, importantly, is required for the proper organization of rod and cone photoreceptor synapses [16][17][18][19]. Recent studies also suggested a functional role of lowly expressed α 2 δ-4 in the brain. mRNA levels of α 2 δ-4 were increased in human hippocampal biopsies obtained from epileptic patients [20], and a recent study on knockout mice identified a role for α 2 δ-4 in cognitive and motor behavior [21].
A putative cache (Ca 2+ channel and chemotaxis receptor) domain containing protein 1 (CACHD1) has been described as an α 2 δ-like protein, displaying structural similarities to members of the α 2 δ family and showing a wide expression in the CNS [22,23]. An expression study using rat and zebrafish illustrated that Cachd1 increases N-type calcium currents and surface expression and competes with α 2 δ-1 for binding to Ca V 2.2 [24]. Interestingly, upon co-expression, human CACHD1 also increased cell-surface localization of the Ca V 3.1 T-type channel [22], although T-type channels are not known to interact with auxiliary β and α 2 δ subunits [1]. Furthermore, overexpression of CACHD1 in hippocampal neurons caused a pronounced increase in T-type current-mediated spike firing [23]. The question of whether and how Cachd1 and α 2 δ-4 contribute to specific functions in neurons, which simultaneously express all α 2 δ isoforms, or whether they share functional redundancy with other α 2 δ isoforms, remains to be answered. Hence, considering that all α 2 δ subunits, including α 2 δ-4, have been linked to synapse formation, it is tempting to speculate that organizing synapse formation is a general feature of α 2 δ proteins.
By establishing an α 2 δ triple knockout/knockdown (TKO/KD) neuronal model system, we showed that the loss of presynaptic α 2 δ proteins leads to a severe defect in glutamatergic synapse formation and synaptic transmission. This striking phenotype could be rescued by the individual expression of the endogenous CNS isoforms α 2 δ-1, -2, and -3 [25]. In the present study, we aimed to examine the synaptic functions of α 2 δ-4 and the α 2 δ-like protein Cachd1. An analysis of surface expression revealed that α 2 δ-4 and Cachd1 can be targeted to the synaptic membrane, similar to α 2 δ-1 and α 2 δ-2, suggesting their involvement in specific synaptic functions. To investigate if mediating synaptogenesis is a general feature of α 2 δ proteins, we next examined the synaptogenic potential of α 2 δ-4 and Cachd1 in α 2 δ TKO/KD neurons. The expression of α 2 δ-4 and Cachd1 rescued synapse formation, presynaptic calcium channel clustering and presynaptic calcium transients in TKO/KD neurons. However, α 2 δ-4 and Cachd1 displayed subtle differences in their rescue abilities, suggesting distinct protein-specific synaptic functions. Indeed, α 2 δ-4 and Cachd1 differentially affected presynaptic calcium transients in wildtype neurons, where α 2 δ-4 showed reduced calcium signaling in comparison to Cachd1. Taken together, we provide the first evidence that α 2 δ-4 and Cachd1 have important synaptic roles in mediating glutamatergic synapse formation and differentially affecting presynaptic calcium transients.

Epitope-Tagged α 2 δ-4 and Cachd1 Target to Presynaptic Terminals
The α 2 δ-like protein Cachd1 is strongly expressed in multiple brain areas including the hippocampus, cerebellum, and thalamus [22]. In contrast, CNS expression of α 2 δ-4 is very low in the hippocampus [15,20]. While the subcellular localization and synaptic roles of Cachd1 are largely unknown, the synaptic actions of α 2 δ-4 have been studied in photoreceptor cells using knockout mice [18]. Because neuronal α 2 δ proteins are critical regulators of synaptic transmission and synapse formation, in the present research, we studied whether Cachd1 and the retinal α 2 δ-4 subunit also are potential regulators of synaptic functions. As synaptic functions of proteins are normally associated with a synaptic localization, we first analyzed the subcellular distribution of Cachd1 and α 2 δ-4 by expressing HA-epitope-tagged proteins in primary cultured hippocampal neurons, as previously established [11,[25][26][27]. To visualize the neuronal morphology, neurons were co-transfected with soluble eGFP [11,[25][26][27]. Life cell labeling revealed that, similar to α 2 δ-1 and α 2 δ-2 isoforms [11,25], Cachd1 and α 2 δ-4 were strongly expressed on the surface of the cell soma ( Figure 1A) and the dendrites including dendritic spines, opposite functional synapses which were identified by immunolabeling the presynaptic synapsin protein ( Figure 1B, arrowheads). Quantification showed somato-dendritic expression levels of Cachd1 and α 2 δ-4 between the levels of α 2 δ-1 and α 2 δ-2, which showed the weakest and strongest surface expression, respectively, of all α 2 δ proteins ( Figure 1C, soma and dendrite).

Epitope-Tagged α2δ-4 and Cachd1 Target to Presynaptic Terminals
The α2δ-like protein Cachd1 is strongly expressed in multiple brain areas including the hippocampus, cerebellum, and thalamus [22]. In contrast, CNS expression of α2δ-4 is very low in the hippocampus [15,20]. While the subcellular localization and synaptic roles of Cachd1 are largely unknown, the synaptic actions of α2δ-4 have been studied in photoreceptor cells using knockout mice [18]. Because neuronal α2δ proteins are critical regulators of synaptic transmission and synapse formation, in the present research, we studied whether Cachd1 and the retinal α2δ-4 subunit also are potential regulators of synaptic functions. As synaptic functions of proteins are normally associated with a synaptic localization, we first analyzed the subcellular distribution of Cachd1 and α2δ-4 by expressing HA-epitope-tagged proteins in primary cultured hippocampal neurons, as previously established [11,25,26,27]. To visualize the neuronal morphology, neurons were co-transfected with soluble eGFP [11,25,26,27]. Life cell labeling revealed that, similar to α2δ-1 and α2δ-2 isoforms [11,25], Cachd1 and α2δ-4 were strongly expressed on the surface of the cell soma ( Figure 1A) and the dendrites including dendritic spines, opposite functional synapses which were identified by immunolabeling the presynaptic synapsin protein ( Figure  1B, arrowheads). Quantification showed somato-dendritic expression levels of Cachd1 and α2δ-4 between the levels of α2δ-1 and α2δ-2, which showed the weakest and strongest surface expression, respectively, of all α2δ proteins ( Figure 1C, soma and dendrite). The somatic expression pattern of the anti-HA labelling revealed that all α2δ subunits and Cachd1 can be expressed on the neuronal surface. (B) Anti-HA and synapsin immunofluorescent labelling revealed that all α2δ subunits and Cachd1 show synaptic and dendritic membrane targeting (white arrowheads). Neurons overexpressing α2δ-1 displayed a weaker expression pattern on the somatic, dendritic, and axonal surface compared to α2δ-2, α2δ-4, and Cachd1. (C) Quantification of the average HA fluorescent intensities illustrated that the surface expression of α2δ-1 was reduced compared to those of the other subunits. α2δ-2 expression in soma, dendrites, and axons was higher compared to α2δ-1, α2δ-4, and Cachd1. Quantification data are shown as values for individual cells (dots) and means ± SEM. ANOVA with (B) Anti-HA and synapsin immunofluorescent labelling revealed that all α 2 δ subunits and Cachd1 show synaptic and dendritic membrane targeting (white arrowheads). Neurons overexpressing α 2 δ-1 displayed a weaker expression pattern on the somatic, dendritic, and axonal surface compared to α 2 δ-2, α 2 δ-4, and Cachd1. (C) Quantification of the average HA fluorescent intensities illustrated that the surface expression of α 2 δ-1 was reduced compared to those of the other subunits. α 2 δ-2 expression in soma, dendrites, and axons was higher compared to α 2 δ-1, α 2 δ-4, and Cachd1. Quantification data are shown as values for individual cells (dots) and means ± SEM. ANOVA with Tukey's multiple comparison test (C) 16-22 cells per condition from three independent culture preparations. F (3,60) = 53.6, p < 0.0001. Significances of post hoc test (*** p < 0.001, ** p < 0.01, * p < 0.05) in comparison to α 2 δ-1, unless otherwise indicated, are indicated with asterisks. Scale bars, 5 µm.
In general, the labelling intensities on the axonal surface were in line with the overall expression levels for all proteins ( Figure 1C), although the axonal expression of Cachd1 was lower compared to those of soma and dendrites ( Figure 1C, compare Cachd1 expression levels in soma, dendrite, and axon). Importantly, Cachd1 and α 2 δ-4 were also expressed in the presynaptic membrane ( Figure 2). This is particularly evident in the upper panel of Figure 2, displaying HA labeling in the outlined presynaptic membrane (white dotted lines drawn around the eGFP signal). Remarkably, and like α 2 δ-2 [25], α 2 δ-4 showed a distinct peri-synaptic localization in the presynaptic membrane around the centrally located synapsin cluster (Figure 2, upper panel and linescan). Together, these data suggest that HA-tagged Cachd1 and α 2 δ-4 target presynaptic terminals upon expression in cultured hippocampal neurons, in line with previous suggestions of the CNS functions of both proteins [20][21][22].
In general, the labelling intensities on the axonal surface were in line with the overall expression levels for all proteins ( Figure 1C), although the axonal expression of Cachd1 was lower compared to those of soma and dendrites ( Figure 1C, compare Cachd1 expression levels in soma, dendrite, and axon). Importantly, Cachd1 and α2δ-4 were also expressed in the presynaptic membrane ( Figure 2). This is particularly evident in the upper panel of Figure 2, displaying HA labeling in the outlined presynaptic membrane (white dotted lines drawn around the eGFP signal). Remarkably, and like α2δ-2 [25], α2δ-4 showed a distinct peri-synaptic localization in the presynaptic membrane around the centrally located synapsin cluster (Figure 2, upper panel and linescan). Together, these data suggest that HA-tagged Cachd1 and α2δ-4 target presynaptic terminals upon expression in cultured hippocampal neurons, in line with previous suggestions of the CNS functions of both proteins [20][21][22]. α2δ-4 and Cachd1 can target presynaptic membranes. An analysis of axonal varicosities illustrated a presynaptic localization pattern for all α2δ subunits and Cachd1. This can be inferred from the color overlay and the co-localization of the linescan peaks of HA-α2δ (red), synapsin (blue) and eGFP (green). Axonal varicosities were identified by their eGFP expression and are outlined by a dashed line. α2δ-4 specifically accumulated in the perisynaptic membrane around a central synapsin cluster. Overall, α2δ-2 and α2δ-4 displayed higher axonal expression compared to α2δ-1 and Cachd1 (for a quantitative analysis, see Figure 1C). Scale bar, 1 µm.

α 2 δ-4 and Cachd1 Restore Presynaptic Calcium Signaling in Triple Knockout Neurons
Restoring presynaptic differentiation in α 2 δ TKO/KD neurons by the expression of α 2 δ-4 and Cachd1 suggests a presynaptic rescue of the severe phenotype. Hence, we next analyzed whether this rescue effect was also reflected in other synaptic functions. To this end, we used calcium imaging with the genetically encoded and presynaptically targeted calcium sensor SynGCaMP6f [25,28] to test if synaptic calcium signaling could be completely or partially rescued by α 2 δ-4 and Cachd1. In line with previous results, a significant fraction of TKO/KD synapses failed to show calcium transients in response to stimulations with 1 action potential (AP), 3 APs, and 10 APs (Figure 4, black lines and symbols). This was particularly evident when plotting the cumulative frequency distribution of single synaptic responses ( Figure 4B), which illustrated that 56, 37, and 34% of TKO/KD synapses showed no response to 1, 3, or 10 AP stimulations, respectively. In contrast, only 26% (1 AP), 12% (3 APs), and 5% (10 APs Figure 4B, cumulative frequency distribution,). Furthermore, the fraction of non-responding synapses was significantly reduced (Chi-square test 1 AP: χ 2 = 130.9, p < 0.0001; 3 AP: χ 2 = 34.1, p < 0.0001; 10 AP: χ 2 = 38.4, p < 0.0001) upon expression of Cachd1. For example, upon 1 AP stimulation, only 14% of synapses expressing Cachd1 were non-responding, while in synapses expressing α 2 δ-1 and α 2 δ-4, the non-responding fraction accounted for 37% and 45%, respectively ( Figure 4B). A similar albeit not significant trend was seen upon stimulation with 3 and 10 APs. Taken together, these results demonstrate that all α 2 δ subunits and α 2 δ-like proteins could rescue presynaptic protein expression and calcium signaling in TKO/KD neurons, suggesting that they share redundant functions in synapse formation and differentiation. Moreover, the subtle differences between Cachd1 and α 2 δ-4 in terms of rescuing presynaptic calcium transients may be related to differences in their abilities to interact with presynaptic calcium channels.

α 2 δ-4 and Cachd1 Do Not Increase Presynaptic Ca V 2.1 Abundance
The above data suggest that expression of individual α 2 δ subunits and Cachd1 is sufficient for the proper differentiation and function of glutamatergic synapses. Nevertheless, the differential effects on presynaptic calcium transients may suggest a subtle functional diversity. Therefore, we next studied the effect of ectopic expression of α 2 δ subunits and Cachd1 on presynaptic calcium channel abundance in wildtype hippocampal neurons. Overexpression of α 2 δ-1 induced a more than two-fold increase in the intensity of presynaptic endogenous Ca V 2.1 labelling compared to eGFP-only expressing control neurons ( Figure 5A,B, eGFP and α 2 δ-1). This was in line with α 2 δ-1-mediated increases in Ca V 2.1 currents upon co-expression [29], and is further supported by the previous observation that the expression of α 2 δ-1 in hippocampal neurons increased the synaptic labelling intensity of Ca V 2.1 channels [30]. Contrary to α 2 δ-1, the overexpression of α 2 δ-4 or Cachd1 did not increase presynaptic clustering of endogenous Ca V 2.1 ( Figure 5A,B). The α 2 δ-1-induced increase of presynaptic Ca V 2.1 was not caused by an overall effect of α 2 δ-1 on presynaptic bouton size, as no concomitant increase of synapsin labelling could be observed ( Figure 5C). These results are consistent with previous observations showing the predominant interaction of α 2 δ-4 with Ca V 1.4 [17] instead of Ca V 2.1, and that Cachd1 had no effects on currents of heterologously expressed Ca V 2.1 [24]. Because Cachd1 was previously shown to increase Ca V 2.2 currents and cell surface expression upon heterologous co-expression [24], we next analyzed the synaptic expression levels of endogenous Ca V 2.2. In comparison to control (eGFP-only) neurons, we observed neither significant increases in presynaptic Ca V 2.2 nor synapsin abundance upon overexpression of α 2 δ-1, α 2 δ-4, and Cachd1 ( Figure 6). Together, this shows that α 2 δ proteins differentially modulate the abundance of presynaptic calcium channels. less, the differential effects on presynaptic calcium transients may suggest a subtle functional diversity. Therefore, we next studied the effect of ectopic expression of α2δ subunits and Cachd1 on presynaptic calcium channel abundance in wildtype hippocampal neurons. Overexpression of α2δ-1 induced a more than two-fold increase in the intensity of presynaptic endogenous CaV2.1 labelling compared to eGFP-only expressing control neurons ( Figure 5A,B, eGFP and α2δ-1). This was in line with α2δ-1-mediated increases in CaV2.1 currents upon co-expression [29], and is further supported by the previous observation that the expression of α2δ-1 in hippocampal neurons increased the synaptic labelling intensity of CaV2.1 channels [30]. Contrary to α2δ-1, the overexpression of α2δ-4 or Cachd1 did not increase presynaptic clustering of endogenous CaV2.1 ( Figure 5A,B). The α2δ-1-induced increase of presynaptic CaV2.1 was not caused by an overall effect of α2δ-1 on presynaptic bouton size, as no concomitant increase of synapsin labelling could be observed ( Figure 5C). These results are consistent with previous observations showing the predominant interaction of α2δ-4 with CaV1.4 [17] instead of CaV2.1, and that Cachd1 had no effects on currents of heterologously expressed CaV2.1 [24]. Because Cachd1 was previously shown to increase CaV2.2 currents and cell surface expression upon heterologous co-expression [24], we next analyzed the synaptic expression levels of endogenous CaV2.2. In comparison to control (eGFP-only) neurons, we observed neither significant increases in presynaptic CaV2.2 nor synapsin abundance upon overexpression of α2δ-1, α2δ-4, and Cachd1 ( Figure 6). Together, this shows that α2δ proteins differentially modulate the abundance of presynaptic calcium channels. pression (outlined with a dashed line) and presynaptic synapsin labelling along untransfected den-drites (see also qualitative linescan analysis). Contrary to neurons overexpressing α2δ-1, which display two times higher CaV2.1 intensity, neurons overexpressing α2δ-4 and Cachd1 exhibited CaV2.1 levels similar to those of the eGFP control. Quantification of CaV2.1 (B) and synapsin 1 (C) shows values for individual cells (dots) and means ± SEM. Cells were obtained from three independent culture preparations. ANOVA with Tukey's multiple comparison test. (B) 38 cells per condition, F(3, 148) = 41, p < 0.0001. (C) 38 cells per condition, F(3, 148) = 1.4, p = 0.25. Significance of post hoc test in comparison to α2δ-1 is indicated by asterisks (*** p < 0.001). Scale bar, 1 µm.

α2δ-4 and Cachd1 Do Not Affect Glutamatergic Synapse Composition
Because α2δ proteins differentially modulate the abundance of presynaptic calcium channels, and because specific α2δ isoforms and splice variants have previously been shown to strongly affect postsynaptic receptor abundance [10,31], we next sought to determine whether α2δ-1, α2δ-4, and Cachd1 would differentially modulate the molecular composition of glutamatergic synapses. Since the overexpression of a splice variant of α2δ-2 in excitatory glutamatergic neurons induces aberrant axonal wiring associated with a striking induction of mismatched synapses by recruiting exclusively postsynaptic GABAA-receptors subunits opposite glutamatergic presynaptic nerve terminals [11], we investigated the abundance of both glutamatergic and GABAergic proteins in glutamatergic synapses. To this end, we quantified the expression levels of pre-and postsynaptic proteins of excitatory glutamatergic (vGlut1, GluR1) and inhibitory GABAergic (vGAT, GABAA-R, Gephyrin) synapses upon overexpression of α2δ-1, α2δ-2-ΔE23, α2δ-4, and

Differential Effects on Presynaptic Calcium Transients
So far, our experiments have shown that α 2 δ-1, α 2 δ-4, and Cachd1 differentially affect Ca V 2.1 channel density but have no effect on the molecular composition of glutamatergic synapses. Nevertheless, rescue experiments in α 2 δ TKO/KD neurons revealed subtle differences in the peak amplitudes of presynaptic calcium transients (Figure 4), even between conditions (α 2 δ-4 vs. Cachd1) showing similar effects on presynaptic calcium channel density. Hence, α 2 δ-1, α 2 δ-4, and Cachd1 may differentially modulate the function of presy-naptic calcium channels beyond an exclusive effect on synaptic channel density. To gain insights into isoform differences in synapse-specific functions, we next analyzed the presynaptic calcium signals in wildtype neurons overexpressing the individual proteins. In this experimental setting, the overexpressed proteins compete with endogenously expressed α 2 δ subunits (α 2 δ-1 to -3) and Cachd1. We observed that α 2 δ-1, in contrast to a previous report [30], and Cachd1 overexpression led to a significant increase in the peak calcium amplitudes compared to control (1 AP) and α 2 δ-4 (1, 3, and 10 APs; Figure 8A). Overexpression of α 2 δ-4, however, resulted in the opposite effect, reducing calcium transients in response to all stimulation patterns ( Figure 8A) in comparison to α 2 δ-1 and Cachd1 expressing neurons. These differences were noticeable in the mean traces ( Figure 8A; ∆F/F0) and in the frequency distribution histogram ( Figure 8B; left-shift of blue α 2 δ-4 curve along the X-axis), as well as in the means of all individual cells analyzed ( Figure 8C). Taken together, our analysis of presynaptic calcium transients revealed differential effects of neurons over-expressing α 2 δ-1 and Cachd1 compared to α 2 δ-4. The observed effects are in line with the rescue potential of the severe synaptic phenotype in α 2 δ TKO/KD neurons ( Figure 4). However, the effect on calcium transients cannot be fully explained by the modulation of the abundance of presynaptic Ca V 2.1 and Ca V 2.2 channels (Figures 5 and 6), even though smaller transients in α 2 δ-4 expressing neurons match the slight, although not significant, tendency to reduce presynaptic calcium channel expression. Overexpression of α2δ-4 led to a reduction in presynaptic calcium signals in response to stimulation with 1 AP, 3 AP and 10 AP (blue) compared to control (green). In contrast, overexpression of α2δ-1 (orange) and Cachd1 (purple) increased calcium transients in all stimulation paradigms, as indi- Figure 8. α 2 δ subunits and Cachd1 differentially affect presynaptic calcium transients. Mean sample traces (A), cumulative frequency distribution blots (B), and quantification (C) of presynaptic calcium signals (SynGCaMP6f) in wildtype neurons overexpressing either α 2 δ or Cachd1 constructs. Overexpression of α 2 δ-4 led to a reduction in presynaptic calcium signals in response to stimulation with 1 AP, 3 AP and 10 AP (blue) compared to control (green). In contrast, overexpression of α 2 δ-1 (orange) and Cachd1 (purple) increased calcium transients in all stimulation paradigms, as indicated in the mean sample traces and the maximal responses. Quantification shows values for individual cells (dots) and means ± SEM. 37-53 cells were obtained from four independent culture preparations (number of synapses analyzed: SynGCaMP6f, 1257; α 2 δ-1, 1122; α 2 δ-4, 1806; Cachd1, 1650). ANOVA with Tukey's multiple comparison test: 1 AP: F (3, 244) = 6.4, p = 0.0004; 3 AP: F (3, 185) = 4.7, p = 0.0036; 10 AP: F (3, 183) = 8.7, p < 0.0001. Significances of post hoc tests between conditions are indicated in the graphs by asterisks (*** p < 0.001, ** p < 0.01, * p < 0.05).

Structural Determinants for Protein-Protein Interactions of α 2 δ Subunits and Cachd1 with Pore-Forming α 1 Subunits
Our experiments have shown, on the one hand, that α 2 δ-1, α 2 δ-4, and Cachd1 can all rescue the severe synaptic defect in α 2 δ TKO/KD neurons, and on the other hand, that they differentially modulate presynaptic calcium channel abundance and calcium transients. Some of the effects on presynaptic calcium transients may be related to the effect on regulating channel density (e.g., α 2 δ-1). However, the positive (Cachd1) and negative (α 2 δ-4) effects on presynaptic calcium transients cannot be consistently explained by the regulation of the density of the most abundant presynaptic calcium channels Ca V 2.1 and Ca V 2.2. Hence, α 2 δ-4 and Cachd1 may preferentially associate with other pore-forming subunits or mediate a negative effect by competing with endogenous α 2 δ subunit partners.
To better understand the structural features that might account for differences in the functionality between distinct α 2 δ isoforms and Cachd1, we performed structural homology modelling between the pore-forming Ca V 2.1 α 1 subunit and α 2 δ-1, α 2 δ-4, and Cachd1. Using the previously published cryo-EM structure of human Ca V 2.2 complexed with α 2 δ-1 as a reference ( Figure 9A, [32]) we modelled the interaction between the predicted AlphaFold structures of Ca V 2.1 with α 2 δ-1, α 2 δ-4, and Cachd1 ( Figure 9B). Amino acid residues previously identified in the Ca V 2.2 complex, which are important for α 1 -α 2 δ interactions, were further analyzed for their conservation and interaction potential in the obtained Ca V 2.1 models. Ca V 2.2 forms a total of eight hydrogen bonds with α 2 δ-1. The interacting residues in Ca V 2.2 (Asp120, Asp122, Asn636, Phe637, Glu1314, and Arg1339) are conserved in Ca V 2.1 (Asp125, Asp127, Asn643, Phe644, Glu1360, and Arg1385) and, importantly, are also predicted interaction sites in all modelled complexes with α 2 δ-1, α 2 δ-4, and Cachd1. When Ca V 2.1 is complexed with α 2 δ-1, two additional hydrogen bonds between the backbone amides of Met131 and Pro124 of Ca V 2.1 are expected to form with Glu366 and S263 of α 2 δ-1. Also, the models predicted one additional hydrogen bond between the backbone oxide of V1380 of Ca V 2.1 and Lys221 of Cachd1. Taken together, there seems to be a conserved interaction pattern of Cav α 1 subunits together with α 2 δ proteins. However, structural homology modelling suggests that the interaction strength of different α 1 -α 2 δ complexes may be fine-tuned by additional interaction sites (hydrogen bonds). Hence, the models for the specific interactions tested (exclusively α 2 δ-1, α 2 δ-4, and Cachd1 were investigated) predicted that Ca V 2.1 would form the strongest interaction with α 2 δ-1, followed by Cachd1 and, finally, α 2 δ-4. seems to be a conserved interaction pattern of Cav α1 subunits together with α2δ proteins. However, structural homology modelling suggests that the interaction strength of different α1-α2δ complexes may be fine-tuned by additional interaction sites (hydrogen bonds). Hence, the models for the specific interactions tested (exclusively α2δ-1, α2δ-4, and Cachd1 were investigated) predicted that CaV2.1 would form the strongest interaction with α2δ-1, followed by Cachd1 and, finally, α2δ-4.

Discussion
Previously, it has been shown that the neuronal α 2 δ subunits α 2 δ-1, -2, and -3 regulate and modulate synapse formation, differentiation, and synaptic transmission [7,9,11,25]. Furthermore, the largely retinal α 2 δ-4 has been identified as an organizer of synaptic ribbons in photoreceptor cells [18]. Here, we have shown that regulating synapse formation and modulating synaptic functions is a general feature of all α 2 δ isoforms and of the α 2 δ-like protein Cachd1. First, α 2 δ-4 and Cachd1 can localize to presynaptic terminals in hippocampal neurons. Second, and similar to the most abundant neuronal α 2 δ proteins, the expression of both proteins can rescue the severe synaptic defect of α 2 δ TKO/KD neurons in terms of synapse differentiation, presynaptic calcium channel localization, and synaptic function. Third, rescuing presynaptic function by the expression of individual α 2 δ proteins resulted in subtle differences in the amplitude of presynaptic calcium transients. Similar differences were observed upon over-expression of these proteins in wildtype hippocampal neurons. These differences did not correlate with the α 2 δ-mediated increase in presynaptic calcium channel density. Hence, our data suggest that α 2 δ isoforms and Cachd1 differ in their ability to interact with presynaptic calcium channels. This was also reflected by distinctly matching charge distributions at the protein interaction surface between α 1 and α 2 δ subunits and the number of predicted interaction sites. Taken together, all α 2 δ and α 2 δ-like proteins can regulate synapse formation, synapse differentiation, and synaptic function via a presynaptic mechanism. However, these effects can only partially be explained by calcium-channel dependent mechanisms.

Functional Redundancy of α 2 δ Proteins
To study the functions of individual channel subunits, it is pivotal to have a system that allows analyzing the effects of individual α 2 δ isoforms. However, this is complicated by the fact that hippocampal neurons simultaneously express three different isoforms [15]. Hence, here, we employed previously established α 2 δ subunit TKO/KD cultured hippocampal neurons as an expression system for α 2 δ isoforms and the α 2 δ-like protein Cachd1. α 2 δ TKO/KD neurons displayed a severe defect in presynaptic function and the formation and differentiation of glutamatergic synapses, as evidenced by strongly reduced presynaptic calcium transients, presynaptic calcium channel localization, presynaptic vesicleassociated proteins, and postsynaptic differentiation [25]. Expression of α 2 δ-4 and Cachd1 in TKO/KD neurons rescued the severe phenotype by restoring presynaptic Ca V 2.1 channel localization (Figure 3) and calcium signaling (Figure 4). Hence, α 2 δ-4 and Cachd1 can mediate presynaptic synapse formation and differentiation to a similar degree as the most abundant neuronal isoform, α 2 δ-1. The observations that each α 2 δ subunit (α 2 δ-1, -2, -3; [25]) and α 2 δ-4 and Cachd1 (this study) can rescue glutamatergic synapse formation and function in TKO/KD neurons, and that they can locate to presynaptic nerve terminals (Figure 2), strongly suggests that α 2 δ proteins mediate redundant presynaptic functions. Principally, the rescue of presynaptic function could be explained either by calcium channel-dependent or calcium channel-independent mechanisms. The facts that presynaptic calcium transients could effectively be restored in α 2 δ TKO/KD neurons by all α 2 δ proteins, and that the residues involved in the interaction with the α 1 subunit of Ca V 2.1 were highly conserved between α 2 δ isoforms and Cachd1 (Figure 9) point to a channeldependent mechanism. Furthermore, all expressed proteins also restored presynaptic Ca V 2.1 localization in α 2 δ TKO/KD neurons. In contrast to the rescue experiments, the consequences of α 2 δ protein over-expression on presynaptic calcium transients and calcium channel localization in wildtype hippocampal neurons seem to be more complex. The overexpression of α 2 δ-1 strongly increased presynaptic Ca V 2.1 abundance, while that of Cachd1 and α 2 δ-4 had no effect. Importantly, these differential effects on presynaptic Ca V 2.1 clustering were not related to the observed differences in neuronal surface expression (confer Figure 1). Presynaptic Ca V 2.2 expression, on the other hand, was not affected by any of the over-expressed proteins. Even though the over-expression of the proteins differentially modulated presynaptic calcium transients (see below), they altered neither the pre-and postsynaptic protein composition nor presynaptic bouton size (Figure 7). Together, this suggests that in addition to channel-dependent mechanisms, other channel-independent mechanisms are also involved in presynaptic functions.
The involvement of channel-independent mechanisms is also supported by other recent observations. First, Cachd1 preferentially complexes with T-Type channels [22], while α 2 δ-1/-2/-3 are likely associated with N-, P/Q-, and L-type channels in the CNS. α 2 δ-4 is the primary partner of the L-type channel Ca V 1.4 in retinal photoreceptor cells; however, its preferred α 1 subunit in hippocampal neurons is elusive. Nevertheless, all investigated proteins were shown to be able to rescue presynaptic calcium transients ( Figure 4) and Ca V 2.1 channel localization ( Figure 3). Second, our previous findings suggest that the transsynaptic regulation of synaptic wiring and postsynaptic GABA A -receptor accumulation by α 2 δ-2 depends on the lack of alternatively spliced exon 23, but not on the number of presynaptic calcium channels [11]. This was supported by our present observation that none of the expressed proteins (which, in contrast to α 2 δ-2, all include exon 23 or a similarly organized protein sequence (Cachd1)) affected postsynaptic GABA A -receptor abundance. Third, we have previously shown that glutamatergic synapse formation in α 2 δ TKO/KD neurons could be rescued by mutant α 2 δ constructs lacking the MIDAS motif, which is necessary for calcium current enhancement and channel trafficking [25]. Finally, a recent study demonstrated that synapse assembly is independent of presynaptic Ca V 2 channels and Ca 2+ entry [33].

Differential Presynaptic Effects of α 2 δ-4 and Cachd1
In contrast to their apparent redundancy in rescuing glutamatergic synapse formation and function (see above), we observed α 2 δ isoform-specific differences. First, α 2 δ isoforms displayed differences in their overall expression on the neuronal surface ( Figure 1). Second, they differentially affected presynaptic calcium transients, both in the rescue (Figure 4) and the over-expression (Figure 8) condition, even though the observed differences were rather subtle. Third, α 2 δ subunits differed in their ability to increase presynaptic Ca V 2.1 expression ( Figure 5). Fourth, they differed slightly in their surface charges and the number of interacting residues with the α 1 subunit (Figure 9).
All studied HA-tagged proteins are located in axons and synapses. Notably, Cachd1 and α 2 δ-4 also showed strong somato-dendritic surface expression ( Figure 1C), which may reflect their preferential interaction with T-type or L-type channels, respectively. In presynaptic boutons, Cachd1 showed a similar distribution pattern to α 2 δ-1, whereas α 2 δ-4 was predominantly located in the perisynaptic membrane, as also observed for α 2 δ-2. Ultimately, and provided suitable antibodies or experimental tools are available, these expression patterns need to be confirmed by high resolution microscopy of the endogenously expressed proteins. α 2 δ-4 reduces presynaptic calcium transients upon over-expression in wildtype neurons (Figure 8) and shows a slightly reduced efficiency in rescuing the calcium transients in α 2 δ TKO/KD (Figure 4). In contrast, Cachd1 increases presynaptic calcium transients and rescues the TKO/KD phenotype, similarly to α 2 δ-1 (Figures 4 and 8). Together, this may indicate that, like α 2 δ-1, Cachd1 specifically modulates active zone calcium channels upon over-expression, resulting in the observed increase of presynaptic calcium transients. In this context, it is noteworthy that Cachd1 can compete with α 2 δ-1 for binding to Ca V 2.2 upon heterologous expression [24]. On the other hand, the rescue experiments in the TKO/KD neurons excluded competition with endogenous α 2 δ isoforms. In these experiments, Cachd1 rescued presynaptic calcium transients, like α 2 δ-1. The perisynaptic expression of α 2 δ-4 indicated a less likely interaction with active zone channels, resulting in smaller calcium signals in the rescue experiments. The dominantnegative effect of α 2 δ-4 on the calcium signal in the over-expression condition may either result from structural interference with the function of wildtype proteins in and around the active zone or from competition with endogenous α 2 δ subunits for presynaptic channels. These mechanisms may induce a shift in the contribution of the different presynaptic calcium channels to the calcium transients. This hypothesis can be tested in the future via the sequential application of specific calcium channel antagonists.

Synaptic Roles of α 2 δ-4 and Cachd1
Our results provide a clear indication that, on the one hand, α 2 δ subunits are redundant regulators of synapse differentiation and function (see also [25]), and that, on the other hand, they differ in their modulation of presynaptic calcium channel abundance and calcium signaling. Until recently, α 2 δ-4 was considered to be an exclusive subunit of Ca V 1.4 channels in retinal photoreceptor cells. However, mutations in CACNA2D4 in patients with neurodevelopmental disorders [34][35][36] and increased α 2 δ-4 mRNA expression levels in the hippocampus after status epilepticus [20] suggest previously unrecognized functions in the brain. This is further supported by the fact that α 2 δ-4 knockout mice exhibit behavioral impairments in terms of cognitive and motor functions [21]. Given the extremely low expression of α 2 δ-4 in the hippocampus [15], α 2 δ-4 might be restricted to a specific and very small subpopulation of neurons in the brain [5]. The present study supports the possibility of more diverse neuronal functions of α 2 δ-4, as it proves, for the first time, that it can indeed modulate synaptic and calcium-channel dependent functions in classical CNS neurons. Unlike the well-studied synaptic functions of α 2 δ-4 in the retina, the synaptic functions of Cachd1 remain unclear. However, Cachd1 has recently been described as an in vivo target for the Alzheimer's protease BACE1 and potentially also a γ-secretase [37]. Considering our present findings, and that Cachd1 was identified as a modulator of Ca V 3 activity [22], it may serve as a potential target in diseases such as epilepsy and pain.

Conclusions
The present study, together with recent findings ( [11,22,25] and others), suggests redundant critical and specific modulatory functions of α 2 δ proteins. Hence, the question arises: why does our nervous system express several distinct proteins with partially overlapping functions? Evolutionarily, it has been suggested that auxiliary subunits were gradually added to the calcium channel complex [38]. Most probably, they evolved to add to the complexity of electrical signaling. This seems to account for their specific and crucial regulatory function, as, contrary to other auxiliary subunits, α 2 δ proteins were not lost during evolution. It is hence tempting to speculate that the number of α 2 δ proteins increased to keep up with the complexity of neuronal signaling and synaptic transmission. Therefore, the fine-tuning of synaptic transmission may depend on the parallel action of more than one type of α 2 δ protein interacting with more than one calcium channel isoform in single synapses. Even if it is experimentally challenging from today's perspective, revealing these parallel and specific functions will ultimately contribute to our understanding of the origin of neurological diseases, particularly neurodevelopmental disorders.

Cell Culture and Transfection Procedures Primary Cultures of Hippocampal Neurons for Fluorescence Imaging
Low-density hippocampal cultures were generated from 16.5-to 18-d-old embryonic BALB/c mice and p0-1 α 2 δ-2/3 double knockout mice as described previously [26,27,42]. Briefly, 2-8 hippocampi were dissected in cold Hank's balanced salt solution (HBSS) following dissociation by 2.5% trypsin treatment and subsequent trituration. Dissociated neurons were plated at a density of~100,000 cells/60 mm culture dish, on five 18 mm glass coverslips (#1.5; GML) coated with poly-l-lysine (Sigma-Aldrich, St. Louis, MO, USA). After attachment of neurons for 3-4 h, coverslips were transferred neuron-side down into a 60 mm culture dish containing a monolayer of glia. Neuros were maintained in serum-free neurobasal medium supplemented with Glutamax and B-27 (NBKO, all ingredients from Thermo Fisher Scientific, Waltham, MA, USA) changed weekly by a replacement of 1/3 of the volume. Glial proliferation was inhibited 3 days post plating with Ara-C (5 µM). On day in vitro (DIV) 6 neurons were transfected with plasmids using Lipofectamine 2000 (Thermo Fisher Scientific) as described previously [26]. The maximal DNA amount used for co-transfections totaled 1.5 µg of DNA at a 1:1:1 molar ratio. For presynaptic calcium measurements, neurons were used between DIV14 and 17, whereas immunolabelling experiments were conducted on neurons between DIV20 and 24.
pβA-Cachd1: The Cachd1 cDNA derived from mouse cortex was cloned into the pβA vector. Primer sequences were selected according to GenBank NM_198037.1. Briefly, the cDNA (nt 1-3867) of Cachd1 was amplified by PCR in three fragments: fragment 1 (nt 1-1366), fragment 2 (nt 1147-2892), and fragment 3 (nt 2707-3867). The forward primer used for amplifying fragment 1 introduced a HindIII site and the Kozak sequence (CC-TACC) upstream the starting codon, while the reverse primer used for amplifying fragment 3 introduced a RsrII site after the stop codon. Fragments 1 and 2 were digested with HindIII/Bsu36I and Bsu36I/BamHI, respectively, and co-ligated into the corresponding HindIII/BamHI sites of the pβA vector, generating an intermediate construct. pβA-2HA-Cachd1: Since it was predicted that the signal protein cleavage would be present at position 49 using SignalP 4.1, the 2HA tag followed by a TEV cleavage site was introduced into the pβA-Cachd1 between the third and fourth amino acid after the predicted signal peptide cleavage site. Briefly, the cDNA sequence of Cachd1 (nt 1-864) was PCR amplified with overlapping primers introducing the double HA tag and the TEV sequence in separate PCR reactions using pβA-Cachd1 as a template. The two separate PCR products were then used as templates for a final PCR reaction with flanking primers to connect the nucleotide sequences and the resulting fragment was then HindIII/SalI digested and ligated into the corresponding sites of pβA-Cachd1, resulting in a pβA-2HA-Cachd1 construct.
pβA-2HA-α 2 δ-4: Since a putative signal peptide was not predicted for α 2 δ-4, we used Signal P 4.0 to discriminate signal peptides from transmembrane regions [45]. The strongest prediction was that the signal peptide would comprise residues 1-68; therefore, the 2HA tag was introduced into pβA-α 2 δ-4 after residue I71. Briefly, the cDNA sequence of α 2 δ-4 (nt 1-1565) was PCR amplified with overlapping primers introducing the double HA tag in separate PCR reactions using pβA-α 2 δ-4 as a template. The two separate PCR products were then used as templates for a final PCR reaction with flanking primers to connect the nucleotide sequences. The resulting fragment was then NotI/BglII digested and ligated into the corresponding sites of pβA-α 2 δ-4, yielding pβA-2HA-α 2 δ-4.
Sequence integrity of all newly generated constructs was confirmed by sequencing (MWG Biotech, Martinsried, Germany).

High-Resolution Fluorescence Microscopy
Immunolabeling of permeabilized or live-cell-stained neurons was performed as described previously [11,[25][26][27]43,46]. Neurons were fixed with 4% paraformaldehyde/4% sucrose (pF) in PBS for 20 min at room temperature, washed, and incubated for 30 min in 5% normal goat serum in PBS containing 0.2% bovine serum albumin (BSA) and 0.2% Triton X-100 (PBS/BSA/Triton), enabling membrane permeabilization. Primary antibodies (Table 1) were diluted in PBS/BSA/Triton, applied in a humidified chamber overnight at 4 • C and detected by fluorochrome-conjugated Alexa secondary antibodies incubated for 1 h at room temperature. For surface staining of HA-tagged α 2 δ constructs, transfected neurons were incubated with rat-anti-HA antibody in glia conditioned neurobasal medium for 10 min at 37 • C following rinsing in warm HBSS and fixation with pF for 10 min at room temperature. Subsequent washing and blocking steps as well as 1 h incubation with fluorochrome-conjugated secondary goat anti-rat Alexa Fluor 594 antibody were conducted with PBS and PBS/BSA, respectively. After washing and postfixation of cells in pF for 5 min, neurons were permeabilized by blocking with PBS/BSA/Triton and incubated with primary mouse-anti-synapsin antibody overnight at 4 • C and detected with goat-anti-mouse Alexa Fluor 350 antibody. Coverslips were mounted with DABCO (Agilent Technologies, Santa Clara, CA, USA) glycerol fluorescence mounting medium. Hippocampal cultures were typically viewed with an Olympus BX53 microscope (Olympus, Tokyo, Japan) using a 60 × 1.42 numerical aperture (NA) oil-immersion objective lens and 14-bit grayscale images were recorded with a cooled CCD camera (XM10; Olympus) using cellSens Dimension software (Olympus) and further analyzed in MetaMorph software (Molecular Devices, San Jose, CA, USA). Images of randomly selected well differentiated positively transfected cells and transfected axons forming synapses with untransfected dendrites were acquired with the same exposure and gain settings for all conditions within an experiment for quantification analysis. Overly saturated neurons (based on eGFP and mCherry levels in presynaptic calcium measurements) were excluded from analysis and only cells with medium to low eGFP or mCherry expression were selected for further analysis. Figures were assembled in Adobe Photoshop CS6 using linear adjustments to correct black level and contrast.

Antibodies
Details on antibodies used within this study are given in Table 1. To determine the synaptic localization of HA-tagged α 2 δ subunits and Cachd1, as well as of presynaptic (synapsin, vGAT, vGLUT1, Ca V 2.1) and postsynaptic proteins (GABA A R subtypes, gephyrin, GLUR1), the "line scan" function was applied in MetaMorph (Molec-ular Devices) [47]. Average fluorescence intensities of respective signals (green-A488, blue-A350 and red-A594) were measured along a line of 3 µm length, followed by background subtraction, and plotted in Microsoft Excel.

Quantification of Fluorescent Clusters in Single Boutons
To analyze the effects of α 2 δ subunits or Cachd1 expression in glutamatergic synapses on synapse composition and protein expression, images from triple-fluorescence labeled neurons were acquired in the eGFP (green), GABA A R α1 (red), and vGlut1 (blue) channels as described [11]. Images were analyzed with a custom programmed and semi-automated MetaMorph journal (Molecular Devices), as described previously [11,25], with a slight modification. Briefly, eGFP and vGlut1 images were superimposed, eGFP/vGlut1-positive varicosities (putative glutamatergic synapses) were randomly chosen as regions of interest (ROIs), and a background region was selected for background subtraction. Axonal varicosities were defined as prominent swellings with higher fluorescence signals compared to the adjacent axonal stem after thresholding the image. Subsequently, GABA A R, and vGlut1 grayscale images were measured without thresholding to avoid potential cut-off of the fluorescent signal. By applying the "shrink region to fit" tool, automatic boundaries were drawn according to the threshold enabling only colocalized clusters to be analyzed and selected ROIs were then transferred from the eGFP image to the GABA A R and vGlut1 images to measure fluorescent intensities. For the individual synapses in each of the channels, the following parameters were detected in a blinded manner: eGFP threshold area as a measure for bouton size and average and integrated fluorescence intensities providing information on the size and intensity of clusters. In the same manner, the expression of synapsin and Ca V 2.1 was conducted.

Quantification of Live Cell Surface Expression
To measure the intensity of live cell-stained HA-tagged constructs, the fluorescent intensities were determined by measuring the fluorescent intensity along a thresholded axon, dendrite, and the cell soma.

Quantification Analysis
Analyses of all experiments were conducted with Microsoft Excel. Mean values of individual cells from two to three independent culture preparations were calculated and plotted in Graphpad Prism 8. The Ca V 2.1 and synapsin values in the rescue experiment were additionally normalized to the rescue control and log-transformed.

Data Analysis
Recordings were quantified in FIJI/ImageJ (National Institute of Health, Bethesda, MA, USA) as described previously [25]. Briefly, 22 ROIs per recorded cell were drawn around mCherry positive presynaptic boutons using the plugin 'Time Series Analyzer V3 with an AutoROI diameter of 10 pixels. The regions were subsequently used in the green SynGCaMP6f recordings to measure the fluorescence changes, removing the background signal with the commonly used "Subtract Background" tool of ImageJ (employing a "rolling ball" algorithm with a radius of 20 pixels ≈ 4.3 µm). The mean of the four highest fluorescence pixels was calculated for each ROI at each frame by applying a self-made macro [28], and further analysis was done in Microsoft Excel. To obtain mean sample traces, the changes in fluorescence as ∆F/F0 were calculated for each ROI, and the 22 synapses per cell were averaged. Statistical analysis was performed on the maxima of cells in Graphpad Prism 8, as indicated. Cumulative frequency distributions were obtained by calculating the maximal response (∆F/F0) by averaging five frames of the peak signal for every single synapse in Microsoft Excel. For each condition and stimulation, a cumulative histogram with defined class sizes was calculated, and values were normalized to the number of synapses per condition. Data were plotted as peak response (log) against the frequency (%) starting at 0.01, as signals below this threshold were indistinguishable from noise and, therefore, considered as non-responding.

Informed Consent Statement: Not applicable.
Data Availability Statement: Data is contained within the article and raw data presented in this study are available upon request.