Gene Transcription and Splicing of T-Type Channels Are Evolutionarily-Conserved Strategies for Regulating Channel Expression and Gating

T-type calcium channels operate within tightly regulated biophysical constraints for supporting rhythmic firing in the brain, heart and secretory organs of invertebrates and vertebrates. The snail T-type gene, LCav3 from Lymnaea stagnalis, possesses alternative, tandem donor splice sites enabling a choice of a large exon 8b (201 aa) or a short exon 25c (9 aa) in cytoplasmic linkers, similar to mammalian homologs. Inclusion of optional 25c exons in the III–IV linker of T-type channels speeds up kinetics and causes hyperpolarizing shifts in both activation and steady-state inactivation of macroscopic currents. The abundant variant lacking exon 25c is the workhorse of embryonic Cav3 channels, whose high density and right-shifted activation and availability curves are expected to increase pace-making and allow the channels to contribute more significantly to cellular excitation in prenatal tissue. Presence of brain-enriched, optional exon 8b conserved with mammalian Cav3.1 and encompassing the proximal half of the I–II linker, imparts a ∼50% reduction in total and surface-expressed LCav3 channel protein, which accounts for reduced whole-cell calcium currents of +8b variants in HEK cells. Evolutionarily conserved optional exons in cytoplasmic linkers of Cav3 channels regulate expression (exon 8b) and a battery of biophysical properties (exon 25c) for tuning specialized firing patterns in different tissues and throughout development.


Introduction
Ca v 3 channels are known for gating ''transient'' currents at low voltages near the resting membrane potential, and often depolarize cells to threshold in a cyclical manner to promote rhythmic firing (for reviews see Senatore et al., 2012 [1] and [2]). Animals with Ca v 3 channels appear in relatives of extant multicellular organisms without tissues or organs (e.g. Trichoplax), and within Cnidarians, the simplest phlyum to harbor a nervous system (e.g. sea anemone Nematostella; Figure 1A). Soft-bodied invertebrates, precursors to various animal phyla including molluscs, likely first appeared in the Late Vendian Period (650 to 543 mya), prior to the divergence of the single ancestral Ca v 3 channel gene into three mammalian genes CACNA1G (Ca v 3.1 or a 1 G), CACNA1H (Ca v 3.2 or a 1 H) and CACNA1I (Ca v 3.3 or a 1 I) [1]. Recently, we have cloned and expressed the first non-vertebrate Ca v 3 channel in vitro, LCa v 3, from the pond snail Lymnaea stagnalis [3]. Here, we describe two optional exons in the I-II and III-IV cytoplasmic linkers of LCa v 3 that are evolutionarily conserved with vertebrate Ca v 3 channels and likely play critical roles in regulating membrane expression and an array of biophysical properties during development. The evolutionarily distant LCa v 3 channel highlights key and fundamental features for T-type channels, providing an important perspective for understanding Ca v 3 channel regulation.

Structural conservation in Ca v 3 channels
Comparisons between Ca v 3 channels reveal that mammalian genes cluster more closely in overall sequence similarity amongst themselves than to the more evolutionarily distant and solitary Ca v 3 gene in snails (Figure 1 A, Figure 1B). Much of the divergence from the snail sequence lies in the tethered, cytoplasmic loops between transmembrane domains, which also bear surprisingly conserved islands of conservation ( Figure 1C). A signature helix-loop-helix in the proximal I-II cytoplasmic linker forms a ''gating brake'' that is unique to all Ca v 3 channels, which when deleted augments characteristic features by shifting lowvoltages of activation to even more hyperpolarized potentials, and increases kinetics of channel opening and closure [4,5]. Downstream of the gating brake in vertebrate Ca v 3.1 and invertebrate Ca v 3 channels is a region that contains a large cluster of histidine residues, followed by an isolated and conserved ''APRASPExxD/ E'' motif that is surrounded by unconserved sequences ( Figure 1D).

Conserved splicing of exons 8b and 25c
Interestingly, the ''APRASPE'' motif is contained within an optional portion of exon 8 that normally spans from Domain I segment 6 across ,70% of the cytoplasmic I-II linker. The , is similarly spliced out in snail and mammalian [6] Ca v 3 channel genes by use of alternative, upstream, intron donor splice sites within exon 8 that truncate the I-II linker coding sequences of LCa v 3 and Ca v 3.1 by 603 and 402 bp, respectively ( Figure S1). Exon 8b occurs downstream of the gating brake, and its omission shortens the I-II linker of LCa v 3 channels by 201 aa (,50%) and Ca v 3.1 by 134 aa (,39%; Figure 2A). A second conserved region of alternative splicing corresponds precisely with the middle of the III-IV cytoplasmic linker ( Figure 2B), which is similarly short in closely related Na v and Ca v channels (54+/21 aa; Figure 2C). Splicing at alternative and more upstream, phase 1 intron donor splice sites shortens the III-IV linker by between 7 and 11 aa in different snail Ca v 3 and mammalian Ca v 3.1 and Ca v 3.2 channels ( Figure 2B, Figure S2, Figure S3). Ca v 3.1 and Ca v 3.2 genes also possess downstream, similarly short optional cassette exons, termed exon 26, that code for between 6 and 19 aa, and that may be included in lieu of exon 25c or appear in tandem with it (Figures 2B and Figure S4). A consensus amino sequence or a number of positively or negatively charged residues is not a consistent feature of 25c exons. The only consistent feature is the first residue coding serine (coded by AGT) that completes the consensus kinase phosphorylation site (KKRKS for LCa v 3), which also contributes to the consensus upstream 59 donor splice site (GTRAGT; Figures 2B and Figure S4).
A simple evolutionary pattern has many invertebrate and mammalian Ca v 3.3 channel genes having ''DD'' isoforms lacking exons 25c or 26. Snail LCa v 3 possesses exon 25c besides DD, but analyses of over 48 independent RT-PCR products from snail embryonic and adult RNA did not uncover an optional cassette exon 26 for LCa v 3 ( Figure S5). Vertebrate Ca v 3.1 and Ca v 3.2 possess both exons 25c and 26, and have correspondingly larger intron sizes spanning these regions than genes lacking either exon 26 (e.g. snail Ca v 3), or both exons 25 and exons 26 (e.g. Ca v 3.3; Figure 2D). Larger intron sizes suggest more extensive regulation of alternative splicing for these short exons.
LCa v 3 channel expression patterns are consistent with those in mammals mRNA transcript levels measured by quantitative RT-PCR suggest that LCa v 3 channels are most abundant in the brain, with intermediate expression in the heart and secretory glands (albumen and prostate), and almost non-detectable levels in buccal and foot musculature of adult snails ( Figure 3A). The overall profile of expression closely matches that of the Ca v 3.1, which is more abundant in the CNS, but not exclusively expressed in the adult brain (i.e. Ca v 3.3), nor is it expressed more widely outside the brain (i.e. Ca v 3.2) [7].
A general precipitous decline in mRNA transcript levels of LCa v 3 occurs from mid-embryo stage (50-75% embryonic development) to near hatching (100% embryonic development) to juvenile snails ( Figure 3B and Figure 3D), corresponding to a similar decline of mammalian Ca v 3 calcium channel gene expression during development (reviewed in Senatore et al., 2012 [1]). There is a continued, slight decline in Ca v 3 channel expression from juvenile to adult animals except for a spike in expression in the albumen gland, likely associated with sexual maturation, and a dramatic decline in the heart, that also parallels the precipitous fall in LCa v 3 expression during mammalian heart development ( Figure 3C) [8].

Conserved regulation of exon 8b and 25c splicing
LCa v 3 transcripts with and without 8b exons are of approximately equal abundance in the central nervous system and secretory glands (such as albumen gland; Figure 3A), which approximates the findings in rat brain where there is significant mRNA expression of Ca v 3.1 with and without 8b exons [6]. We observe that the +8b isoform is less associated with the snail heart (16%) and buccal/foot musculature, compared to the higher levels in the brain and secretory glands ( Figure 3A). Exon 25c has a more striking developmentally-regulated pattern, with a precipitous decline in LCa v 3 transcripts lacking exon 25c from embryo to adults ( Figure 3B and Figure 3D), especially in the heart ( Figure 3C). The continued down-regulation of Ca v 3 channels lacking exon 25c from juveniles to adults gives the appearance of a switch with an increasing relative expression of the plus exon 25c isoform in the adult brain and secretory albumen and prostate glands ( Figure 3C). Exon 25c-containing isoforms of LCa v 3 predominate in the adult heart (79%; Figure 3A) and enhance in expression in most adult snail tissues ( Figure 3C). The relative absence of exon 25c before and at birth and its predominance in adults is consistent between snail LCa v 3 and mammalian Ca v 3.1 [9][10][11] and Ca v 3.2 [12,13] ( Figure 3E). Correlative analysis of normalized qPCR data for plus and minus 8b and 25c variants, across all juvenile and adult tissues tested, reveals that +8b and 225c variants share similar variability in expression between the different tissues (correlation coefficient R 2 = 0.885), and that 28b and +25c also share similar expression patterns (R 2 = 0.898), while +8b/+25c and 28b/225c have much lower correlation coefficients ( Figure S6).

Exon 25c selectively alters biophysical properties of LCa v 3
Using whole cell patch clamp technique we examined the biophysical consequences of the absence or presence of exons 8b and 25c in cloned LCa v 3 variants heterologously expressed in HEK-293T cells, and performed one-way analysis of variance to asses statistical significance (Table 1). Surprisingly, the absence or presence of the large I-II linker 8b exon has comparatively little influence on the biophysical properties of LCa v 3 (Figures 4 and Figure 5) when compared to the small 25c insert. The peaks of the rapidly-activating and inactivating calcium currents were measured in response to 5 mV steps in the presence of 2 mM extracellular calcium ( Figure 4A and Figure 4B). Inclusion of exon 25c induces statistically significant 23.7 mV (+8b) and 23.8 mV (28b) hyperpolarizing shifts in the half-maximal activation (V 0.5 ) of LCa v 3 calcium currents, extrapolated from the fitted Boltzmann of the plot of the fraction of maximal conductance at each voltage step ( Figure 4C and Table 1). Exon 25c also causes parallel 23.9 and 26.9 mV hyperpolarizing shifts in half-maximal inactivation for +/2 exon 8b LCa v 3 variants, respectively ( Figure 4C and Table 1). Boltzmann-fitted inactivation curves were generated by measuring residual peak currents at 235 mV following a series of inactivating, pre-pulse voltage steps ( Figure 4C). Similar hyperpolarizing shifts in the half maximal values for activation and inactivation curves are apparent for exon 25c inserts, regardless of the difference in their APRASPE motif vary from 0 to 89 aa. Accession/IDs: ( sequence or gene isoform type for snail LCa v 3, or mammalian Ca v 3.1 [10,14] or Ca v 3.2 channels [12,13,15] ( Figure 4D and Table 1). The effect of exon 25c on the hyperpolarizing shift is greater than the differences in voltage-sensitivities between different T-type channels ( Figure 4E). Optional exon 26 also causes shifts in the voltage-sensitivities of activation and inactivation for Ca v 3.1and Ca v 3.2 ( Figure 4D), although generally the differences are much less dramatic than those imposed by exon 25c.
Exon 25c also promotes a significant speeding up of channel kinetics, most apparent in currents elicited by small voltage steps (260 or 255 mV; Figure 5A and Table 1). Activation kinetics are significantly faster in the presence of exon 25c, especially when exon 8b is also present, as measured as the delay to time to peak current ( Figure 5B), as are inactivation kinetics, as measured by single exponential tau curve fits ( Figure 5C). A role of exon 25c in promoting faster channel activation and inactivation is common to both snail and mammalian Ca v 3 channels ( Figure 5D). Deactivation kinetics are slowed by exon 25c, which corresponds to a slower rate of closure of Ca v 3 channels from the open state, especially those currents elicited from voltage steps down to negative voltages such as resting membrane potential or more hyperpolarized than rest (2100 to 265 mV; Figure 5E). A slowing of deactivation kinetics in the presence of exon 25c is also a shared feature of snail and mammalian [14,15] Ca v 3 channels ( Figure 5F). Exon 25c also promotes a slowing of the recovery rate from inactivation at the earliest time points of recovery (,0.2 seconds; Figure 5G, inset and Table 1), reminiscent to the slowing of inactivation promoted by exon 25c in mammalian Ca v 3 channels [14]. It should be noted that LCa v 3 is relatively slow to recover from inactivation compared to mammalian Ca v 3 channels, even in the absence of exon 25c ( Figure 5G). Exon 8b can fine tune the biophysical changes imparted by exon 25c, such as influencing kinetics at depolarized potentials (i.e. above 245 mV; Figure 5B, Figure 5C, Table 1), or speeding up deactivation, more pronounced in the presence of exon 25c, near resting membrane potential (i.e. 260 to 270 mV; Figure 5E and Table 1). In summary, snails and mammals possess highly variable short 25c exons that exert near identical biophysical changes to Ca v 3 channels during development.

Exon 8b selectively alters LCa v 3 expression
The large optional exon 8b has a major role likely associated with controlling the expression of Ca v 3 channels, since transfection of equal molar quantities of cloned LCa v 3 vectors into HEK-293T cells produces approximately 2-fold increases in current density recordings when variants lack 8b ( Figure 6A, Table 1). In contrast, current density does not change in the absence or presence of exon 25c. Increased currents were also reported for mammalian Ca v 3.1 lacking exon 8b [6], suggesting that analogous regulatory mechanisms might act on snail and mammalian 8b exons to control membrane expression. The only obvious similarity between exon 8b amino acid sequences of LCa v 3 and Ca v 3.1 is an APRASPE motif ( Figure 1D), which, when deleted, surprisingly has no effect on LCa v 3 channel current density ( Figure 6A).
The doubling of current densities in the absence of exon 8b, attributable to an increase in the number of channels present at the membrane, could arise from increases in total protein expression or strictly increased trafficking to the membrane [6]. A possible change in membrane trafficking of LCa v 3 associated with exon 8b was assessed by separating and quantifying biotinylated, membrane-delimited channel variants expressed in HEK-293T cells, relative to channels present in whole cell fractions on immunoblots labeled with polyclonal LCa v 3 antibodies [16]. Antigen specificity of the polyclonal antibodies for these experiments was confirmed by immunolabeling of HEK cells transfected with either the T-type channel cDNA (+8b 225c), or that of snail LCa v 1 calcium channel ( Figure 6D), and further tested on Western blots using expressed and purified LCa v 3 I-II linker peptides ( Figure S7). Biotinylation experiments revealed dramatic increases in both total and membrane-expressed fractions of transfected LCa v 3 variants lacking exon 8b ( Figure 6B and Figure 6C), indicating that the doubling of current densities is likely due to an increase in protein expression, and not specifically increased membrane trafficking. We also inserted hemagglutinin (HA)-epitope tags in the extracellular Domain I s5-s6 loop of LCa v 3 variants, to quantify the chemiluminescent signals of labeled, membrane-delimited epitope by luminometry [6]. Surprisingly, luminometry experiments were inconclusive, since there was a doubling of signal in permeabilized transfected cells compared to non-permeabilized conditions, regardless of the treatment, including cells transfected with untagged LCa v 3 (i.e. +8b +25c) ( Figure S8). In addition, all HA-tagged channels, regardless of their insert, produced ,20-fold smaller recordable currents than their wild type channel counterparts ( Figure S9).

Evolution and development of Ca v 3 channels
Invertebrates possess only a single Ca v 3 channel gene which provides a reference point for evaluating fundamental features of T-type channels. We illustrate here two highly conserved and developmentally regulated optional exons in cytoplasmic linkers shared between invertebrates and mammals, that provide insights into the fundamental roles that alternative splicing has played in the early evolution of Ca v 3 channels.
Ca v 3 channels likely first appeared in early multicellular organisms, since single-celled animals, such as the coanoflagellates, have a single calcium channel homolog, an L-type (Ca v 1), but no Ca v 3 channel [1]. Likely, gene duplication of the L-type calcium channel gene generated a synaptic Ca v 2, ''N-type like'' channel gene and a Ca v 3 channel gene, in a close ancestor of primitive multi-cellular organisms (i.e. Trichoplax) or within an animal phylum with the most primitive nervous system, the Cnidarians, which both possess a full complement of single Ca v 1, Ca v 2 and Ca v 3 channel genes ( Figure 1A; [17]). The snail homolog from  Lymnaea stagnalis closely matches the mammalian Ca v 3.1 and Ca v 3.2 channels in the quintessential features of Ca v 3 channels, which appear optimized for generating rhythmic firing patterns, with a low voltage range of gating and rapid kinetics to drive membrane depolarization from resting membrane potential to threshold quickly, and a property of slow deactivation kinetics, which keeps Ca v 3 channels open to maximize their effectiveness, if they are not in a refractory, inactivated state [1]. Conserved features in invertebrates also extend to their developmental expression profile. Ca v 3 channel mRNA levels fall precipitously (80%) from embryo to juvenile snails in comparison to a more gradual decline of related channels, such as Ca v 2 and NALCN (,20%) or L-type calcium channels (60%) ( Figure 3D). This decline in Ca v 3 channel transcripts continues from juvenile to adults in most tissues and is most dramatic in the heart compared to the brain. The higher density of Ca v 3 expression correlates well with the faster embryonic heart rate, and drops sharply with the slower heart rate after the rapid phase of growth (embryo/neonate to juvenile animals). Ca v 3 channel expression is also highest in animals of smaller sizes [8], which have faster heart and metabolic rates associated with allometric scaling. A prominent Ca v 3 current remains in adult snails [18], but this is diminished in the pacemaking cells of increasingly large mammals, to a level that may be imperceptible in the adult human heart [8]. In the tissues we studied, only albumen and prostate glands, which grow dramatically in size from juvenile to adult snails, exhibit increases in Ca v 3 channel expression, consistent with organ maturation and emergent properties for Ca v 3 channels in secretory roles of sexually-mature animals.
A high embryonic level of Ca v 3 channel expression is associated with expanded roles in the early proliferative states such as myoblast fusion [19], and recapitulated to high levels in disease states such as cancer [20] and ventricular hypertrophy [8]. Extracellular calcium contributes to contraction in immature muscle, which lacks the elaborate calcium delivery system in adult muscle involving transverse tubules signalling to intracellular calcium release units via coupling to membranal L-type calcium channels [21]. Adult invertebrate muscle is also primitive, lacking tetradic organization and striations [21], where Ca v 3 channels can serve as the only calcium source for muscle contraction (Polyorchis jellyfish muscle) [22], are a major contributor to muscle action potentials (nematode) [23], provide an alternative to sodium spikes in giant motor neurons (Aglantha jellyfish) [24] or are a prominent source of calcium for the adult heart [18].

Evolution of cytoplasmic linkers
The fall in mRNA expression from embryo to adult snails is almost exclusively with Ca v 3 channels that lack exon 25c in the cytoplasmic III-IV linker ( Figure 3B), leading to a change in mode of T-type channel activity in the transition from embryo to adult. Remarkably, III-IV linkers have been restricted to a discrete size of 54+/21 aa in all human calcium and sodium channels, whereas the other cytoplasmic linkers substantially vary in size ranging from ,100 to .500 aa ( Figure 2C). The shortness constrains the III-IV linker so that it is tightly coupled to the cytoplasmic end of the pore, preventing it from protruding too deeply into it. The III-IV linker is a primary agent for fast inactivation of Na v channels, serving as a manhole cover with a hydrophobic latch (IFM) flanked by an alpha helix on either side that pivots on a flexible hinge to occlude the cytoplasmic pore [25]. The ''inactivation particle'' of the III-IV linker of Na v channels is conserved down to single-celled coanoflagellates, is absent in all Ca v channels [17], but a parallel role is likely played by the similar sized III-IV linker of Ca v 3 channels. Evolution of an alternative 59 donor splice site expands the middle of the III-IV linker to retain intron sequence as a 7 to 11 aa coded exon, extending exon 25 (dubbed exon 25c) in molluscs and vertebrate Ca v 3.1 and Ca v 3.2 channels beyond the tightly-regulated size of 54+/1 aa. Corresponding with III-IV linker exon variants, there is an added complexity of the regulation of alternative splicing, with a dramatic increase in intron size from non-molluscan invertebrates and vertebrate Ca v 3.3 channel genes that lack optional exons in the III-IV linker, to the molluscan Ca v 3 channel with an exon 25c, to vertebrate Ca v 3.1 and Ca v 3.2 channels which contain two differentially-regulated optional exons 25c and 26 ( Figure 2D). Splicing factors in spliceosomes have varying compositions in different developmental stages and cell types, to generate unique mRNAs from heteronuclear pre-mRNA with the guidance of specific nucleotide sequences within introns and adjacent exons [26]. Larger intron sizes suggest extensive regulation of alternative splicing for these short exons in different tissues [26]. Interestingly, +8b and 225c variants have a significantly correlated expression pattern amongst the juvenile and adult tissues tested by qPCR (correlation coefficient R 2 = 0.885), as do 28b and +25c (R 2 = 0.898), while +8b/+25c and 28b/225c have lower R 2 values ( Figure S6). This suggests that there is a co-ordinated splicing of LCa v 3 isoforms containing either +8b with 225c, or 28b with +25c in different tissues, and that these are the most physiologically relevant isoforms in juvenile and adult snails. However, alternative splicing can be somewhat stochastic [27,28], and given the presence of all four splice variants together in many different tissues (Figure 3), it is probable that all four possible configurations are present in the animal, at least to some degree.

Biophysical consequences of 25c and 26
Modeling studies suggest that there are general truisms associated with Ca v 3 channel splicing, although Ca v 3 channel behavior will vary considerably with background cellular context such as other ionic conductance and the resting membrane potential, as well as with subcellular localization [29]. Exon 25c imparts a hyperpolarizing shift in the activation and inactivation curves of Ca v 3 channels (by a few to 10 mV), that restricts their activity because they are unavailable and inactivated at rest. If the membrane potential does not change significantly throughout development, enrichment of exon 25c in adults would serve to dampen the contribution of these channels to excitability. These variants would be more adept at driving post-inhibitory rebound excitation after strong hyperpolarizing input, which together with the faster activation and inactivation kinetics reported for plus exon 25c variants, would generate calcium spikes with a faster onset and faster attenuation after hypepolarization, reminiscent of low threshold spikes (LTS) in thalamocortical neurons [30]. Within action potential bursts that sometimes ride over LTS, the depolarizing contribution of Ca v 3 calcium currents during action potential repolarization is maximized by the long delay of channel closure from the open state (slower deactivation kinetics) imparted by exon 25c [14,31]. Exon 25c, the more prominent isoform in   [15]). (E) Voltages of half-activation and inactivation of LCa v 3 resulting from inclusion or exclusion of exon 25c, compared to adults, promotes inactivation to limit excitability, and also slows the recovery from complete inactivation at the earliest time point of recovery. Given that Ca v 3 channels play important roles in setting the rhythmicity of oscillatory firing, the above features suggest that the enrichment of exon 25c in adults, along with the developmental down-regulation of Ca v 3 channel expression, serve to slow down oscillatory firing or diminish the contribution of Ca v 3 channels to excitability in general.
Embryonic Ca v 3 channels lacking exon 25c have properties more similar to Ca v 3.3, with slower activation and inactivation kinetics, as well as right-shifted activation and steady-state inactivation curves [31]. DD variants conduct less calcium into the cell during a single burst, and are better suited for prolonged spiking at higher frequencies (e.g. burst firing in nRT neurons [26,32]) due to a lower propensity for cumulative inactivation during high frequency firing. Rather than dampen, Ca v 3.3 channel currents actually facilitate through the first few spikes in a train. Since DD variants are the predominant isoform in the embryo of both snails and mammals, the default state is likely one which favours a contribution of Ca v 3 channels to more rapid firing patterns in embryonic cells relative to adult ones. In addition, Ca v 3 channels without exon 25c are less inactivated at rest, with a more positive-shifted activation and inactivation curves, and are more readily available to open in response to depolarization, rather than relying on hyperpolarizing input. Ca v 3.3 channels generally have a brain-specific, somato-dendritic localization where they extend further into the dendritic arbour than Ca v 3.1 and Ca v 3.2 [33].
Dendritic T-type channels are implicated in synaptic integration, where they serve to amplify post-synaptic inputs to the soma [34], and the slower activation and inactivation kinetics of Ca v 3.3 might facilitate this role by allowing the channels to overcome the high input resistance of dendrites. Interestingly, modelling suggests that dendritic T-type channels are subject to a hyperpolarizing shift in their activation during somatic depolarization [29], which for Ca v 3.3 channels would approximate their depolarized activation curves to match those of the more hyperpolarized and somatic Ca v 3.1 and Ca v 3.2.
Exon 25c sets the framework for the major biophysical differences between Ca v 3 channels. Snail LCa v 3 and mammalian Ca v 3.1 and Ca v 3.2 utilize exon 25c to differentiate themselves from Ca v 3.3, with shifted activation and inactivation gating to significantly more hyperpolarized potentials, not achievable by merely switching to the expression of a different mammalian channel gene (i.e. Ca v 3.1, Ca v 3.2 or Ca v 3.3; Figure 4E). Indeed, systematic replacement of different trans-membrane and cytoplasmic regions of Ca v 3.1 into Ca v 3.3 creates chimeric channels that resemble Ca v 3.1 only when an exon 25c-containing III-IV linker is inserted into Ca v 3.3 [35]. These changes (measured in Xenopus oocytes) include dramatically shifted activation and inactivated curves in the hyperpolarizing direction (29.5 mV and 28.6 mV, respectively), and increases in the kinetics of activation and inactivation [35].

Structure of exon 25c and exon 26
Sequence comparisons do not elucidate a consistent picture for the structural requirements of exon 25c, since the number of charged residues can vary even amongst closely related species (e.g. freshwater and sea snail Ca v 3 channels; Lymnaea SEGP-KASSSE vs. Lottia SEGISTKKAG), size can vary considerably (e.g. between frog and human Ca v 3.2 channels; SKALPVAVA-VAE vs. SKALYMSE respectively) and a requirement for the first position being a serine residue to completing a consensus phosphorylation site is lacking when comparing human exon 25c and 26 inserts (SKEKQMA vs NLMLDDVIASGSSASAASE respectively; Figure 2B). An effective, threshold size of 8 aa may be required for III-IV inserts, since the slightly shorter exon 26 insert of Ca v 3.2 channels (STFPSPE), imparts only small biophysical changes such as a shift in the curve for steady-state inactivation ( Figure 4D), an no significant changes in kinetics of activation and inactivation ( Figure 5D). A role for residue charge would be consistent with Na v channels, where clusters of charged residues in the III-IV linker differentially regulate the kinetics of fast inactivation [36].

Role of exon 25c in gating
Dynamic clamp simulations suggest that minimal changes in the biophysical properties of the Ca v 3 channel currents, especially in the voltage range corresponding to the base of the current-voltage curve, drastically alter the contribution of Ca v 3 channels to calcium spikes [30], such as the low threshold calcium spikes in the thalamus characterized in states such as physiological sleep or pathological states such as epilepsy. LTS are often crowned with sodium channel-dependent action potential spike trains, whose frequency and longevity highly depends on the shape of the underlying calcium spike and thus Ca v 3 channel activity [2]. While the suite of biophysical changes associated with inclusion of exon 25c provides Ca v 3.1 and Ca v 3.2 channels for fast postinhibitory depolarizing responses with calcium spikes crowned by relatively fast but short lived spike trains (e.g. thalamocortical LTS), omission of exon 25c (and 26) creates variants better suited for LTS with a more delayed onset, less dependent on hyperpolarization, and crowned by longer lasting sodium channel spike trains (e.g. nRT LTS) [31]. Clearly even subtle differences may drastically influence the contribution played by Ca v 3 channels under different conditions. These include differences in the channel genes (e.g. recovery from inactivation of Ca v 3.1 vs Ca v 3.2), other splice variants that modulate gating and trafficking (such as exons 8b and 38b for Ca v 3.1, exon 35a for Ca v 3.2, and exon 14 for Ca v 3.1/Ca v 3.2) [10,12,14], G protein modulation and phosphorylation (largely a capacity of Ca v 3.2 channels [37]), potentiation by glutamate receptors (Ca v 3.1 channels in cerebellar Purkinje fibers) [38], association with K + channels [39], as well as being influenced by the shape of the trigger (e.g. synaptic input) [30].

Role of exon 8b in expression
Exon 8b has only a minor influence on biophysical properties, while its omission dramatically increases the membrane expression of snail LCa v 3 and mammalian Ca v 3.1 (by ,2-fold) in transfected human cells ( Figure 6A). A conserved APRASPE motif is an obvious, shared feature of exon 8b in invertebrate and Ca v 3.1 channels, but its absence has little effect on the current densities of snail LCa v 3 in mammalian cell lines. It may indicate that the APRASPE motif is not critical for protein and/or membrane expression, or that its function depends on tissue-specific factors not present in HEK-293T cells. Exon 8b is abundant in the snail brain and secretory organs and mostly lacking in the heart. It is contained in the largest cytoplasmic linker of Ca v 3 and Na v channels ( Figure 2C), which is also the primary domain for regulating expression of all Ca v and Na v channels. Expression of closely related high voltage-activated (HVA) Ca v 1 and Ca v 2 channels is promoted by assembly with an accessory b subunit binding to the proximate I-II linker (in the homologous position of the gating brake of Ca v 3 channels) [40]. Increases in the expression of HVA channels have recently been shown to depend on the b subunit interaction that protects the channels from targeted degradation via the endoplasmic reticulum-associated protein degradation (ERAD) pathway [40]. It has been proposed that misfolding in the I-II linker of Ca v 2.1 is responsible for pathological ER retention and ERAD of mutant channels associated with episodic ataxia [41].
Subunit assembly with accessory subunits is not a likely control mechanism for Ca v 3 channel expression or protein stability, but the I-II linker of snail and mammalian Ca v 3 channels is a key region for its regulation. Large deletions in the I-II linker (downstream of the gating brake) for snail LCa v 3 and mammalian Ca v 3.1, Ca v 3.2 and Ca v 3.3 moderately enhances, moderately enhances, dramatically enhances, and depresses membrane expression respectively [42]. Lower current densities observed in the presence of exon 8b could be attributed to regulation of channel trafficking to the cell membrane, or the depression of total protein expression of Ca v 3 channels, or a combination of these factors. Our results indicate that inclusion of exon 8b, regardless of  the presence or absence of exon 25c, causes dramatic decreases in both total protein and membrane expression of LCa v 3 channels ( Figure 6B and Figure 6C).

Conclusions from snail work
Work with snails provides a unique perspective for defining the fundamental features of Ca v 3 channels. We show that the snail and mammalian channels operate within tightly regulated biophysical constraints for supporting rhythmic firing in the brain, heart and secretory organs, and that there are many remarkable parallels in expression patterns between respective Ca v 3 channels and exon 8b and 25c splice isoforms. The presence of exon 25c in the III-IV linker, a region whose length is highly invariable in sodium and calcium channels of the 4-domain superfamily, suggests that this locus was exploited during T-type channel evolution to provide splice variants with markedly different biophysical properties. We suggest that the snail channel is more akin to Ca v 3.1 because of the common regulation of membrane expression with exon 8b, which is enriched in the brain of both snails and mammals. If indeed Ca v 3.1 is more reminiscent of the ancestral channel type, then Ca v 3.2 deviated from Ca v 3.1 less in biophysical terms, but rather in its greater capacity for modulation. Ca v 3.3 became the most divergent of all T-type channels with a restricted tissue expression profile in the brain, and lacking an exon 25c. In the embryo, T-type channels are highly abundant and lack exon 25c, which supports accelerated rhythmic firing, and high channel density might also serve expanded roles such as the calcium delivery for contraction of immature muscle. In adults, there is an upregulation of T-type channels in secretory glands, coinciding with sexual maturation and active secretion of vesicular components into reproductive tracts. Further insights into T-type channels are facilitated in snails which have only a single Ca v 3 channel gene, and a highly tractable and accessible preparation for studying its associated functions in brain, heart and secretory organs.

Materials and Methods
Expanded materials and methods are provided as a supplement (Methods S1).

Source of animals
Giant pond snails, Lymnaea stagnalis were raised in-house in a snail vivarium and breeding facility in B1-177, Department of Biology, University of Waterloo.

Identification of splice variants
During the initial sequencing and cloning of LCa v 3 [3], multiple variable sequences were identified, including conserved optional exons 8b in the I-II linker and 25c in the III-IV linker. PCRscreening attempts with Lymnaea cDNAs did not reveal exon 26, found in the III-IV linkers of mammalian Ca v 3.1 and Ca v 3.2 channel genes ( Figure S5). Genomic sequence surrounding optional exons 8b and exon 25c were obtained by PCR and compared to other available snail and mammalian sequences (Figures S1, Figure S2, Figure S3, Table 2). PCR was also used to generate the APRASPEQSD sequence deletion, HAtagged channels and 2/+ exon 8b and 2/+ exon 25c splice variants ( Table 2).

RNA extraction and PCR
mRNA for cDNA analyses was extracted [43] from 50-75% and 100% embryos grouped according to morphological features of egg capsules [44] and the shell length of juvenile and adult snails [45].
Lymnaea transcripts were amplified by quantitative RT-PCR (qPCR) with primers designed against LCa v 1 [46], LCa v 2 [47], LCa v 3 and 2/+ exons 8b and 25c splice isoforms of LCa v 3 ( Table 3). qPCR transcripts were normalized against standards, actin, SDHA and HPRT1. Cycle threshold (CT) values for the HPRT1 gene were found to produce the lowest stability value (i.e. 0.098) using NormFinder software [48], indicating its suitability as a reference gene. Expression levels of genes/isoforms were normalized relative to HPRT1 using the DDCT method [49] where ratio = (E target gene ) DCTtarget gene /(E HPRT1 ) DCTHPRT1 . Amplicons ranged from 119 to 145 bp, producing single products, with PCR efficiencies (E) ranging from 86 to 110%, with templates generated with 1:5 serial dilutions of pooled cDNA as templates ( Table 3). qPCR reactions were carried out in quadruplicate, and standardized between 96 well plate samples with primers against HPRT1.
We use an optimized cell culture and CaPO 4 transfection strategy for heterologous expression and patch clamp recording of ion channel cDNAs in HEK-293T cells, previously outlined in video protocol format http://www.jove.com/video/2314 [50]. Whole cell patches were sealed with pipette resistances of 2-5 MV, and with typical access resistance maintained after breakthrough between 4 and 6 MV [3,50]. Series resistance was compensated to 70% (prediction and correction; 10-ms time lag). Only recordings with minimal leak (,10% of peak) and small current sizes (,500 pA) were used for generating peak currentvoltage relationship curves due to loss of voltage clamp above 500 pA; all other recordings were typically maintained below 1.5 nA. Offline leak subtraction was carried out using the Clampfit 10.1 software (Molecular Devices). Voltage-command protocols and the curve fitting of data are described previously [3,50].

Measurement of membrane expression
Current densities (pA/pF) were measured from transfections of 6 mg of plasmid of the LCa v 3 splice variants [50], quantified to equal amounts by plasmid linearization, electrophoresis and densinometric analysis.

Luminometry
Single haemagglutinin (HA) epitope tags were introduced into the Domain I s5-s6 extracellular loops of the four full-length LCa v 3 variant constructs. Equimolar amounts of mock, epitopetagged and un-tagged LCa v 3 were transfected into HEK-293T cells, and after incubation cells were labeled with rat a-HA monoclonal antibody under membrane permeabilized and nonpermeabilized conditions, and luminometry quantified using a FilterMax F5 Multi-Mode Microplate Reader (Molecular Devices) with a chemiluminescence reaction catalyzed with a goat a-rat HRP-conjugated secondary antibody.

Biotinylation
Surface-expressed channels were measured by biotinylation and identified in Western blots with snail Ca v 3 channel-specific polyclonal antibodies raised in rabbits using as antigen a large portion of the LCa v 3 I-II linker peptide lacking exon 8b expressed and purified from bacteria using a pET-22b(+) protein expression vector ( Table 2). Antibodies were pre-tested for immune reactivity in Western blotting and immunolabeling, as reported previously for antibodies produced strictly against the exon 8b peptide sequence [3]. HEK-293T cell were transfected with equimolar amounts of LCa v 3 channel variants and biotinylated with Sulfo-NHS-SS-Biotin (Pierce). Surface-labelled, biotinyated fractions were isolated using NeurAvidinH Agarose columns, with the Ca v 3 channel surface and total protein pools quantified on Western blots using HRP-activated chemiluminescence from a goat a-rabbit secondary antibody.  Figure S6 Scatter matrix analysis of normalized mRNA qPCR values for the various alternative splice sites of LCa v 3 reveals that +8b and 225c variants tend to have similar expression patterns amongst the various adult and juvenile tissues tested (correlation coefficient R 2 of 0.885); 28b and +25c also have a high R 2 value of 0.898. Confidence regions of 95% for the correlated values is depicted by the red ellipses. Analysis was carried out using Origin 8.5 software (OriginLab). (TIF) Figure S7 Confirmation of specificity of snail LCav3 polyclonal antibodies for bacterially-expressed epitope peptide using Western blotting. Polyclonal antibodies raised against a 17.6 kDa peptide corresponding to the I-II linker of LCav3 lacking exon 8b, detect bacteria-expressed and Histidine tag-purified I-II linker protein on western blots (white arrow), and do not detect a similarly expressed 23 kDa protein corresponding to exon 8b (black arrow). (TIF) Figure S8 Membrane-expression of HA-tagged LCav3 variants could not be measured using luminometry. More than a doubling of ELISA signal was recorded when permeabilized transfected cells were compared to non-permeabilized conditions, regardless of the treatment, including cells transfected with HA-epitope tags or untagged (wt) LCa v 3 channels. HA epitopes were introduced into the IS5-S6 extracellular loops of the LCav3 variants, and identified in transfected homogenates of HEK-293T cells with labelled antirat HA monoclonal antibody. Secondary goat anti-rat HRP (horse radish peroxidase) catalyzed the chemilluminescence quantified using a FilterMax F5 Multi-Mode Microplate Reader (Molecular Devices). (TIF) Figure S9 HA-tagged LCav3 variants expressed .20 fold less than un-tagged LCav3 variants in HEK-293T cells. Above is the largest-sized currents of three LCav3 variants (+8b225c, 28b225c, and +8b+25c) transfected and recorded in HEK-239T cells by whole cell patch clamp technique and generated by voltage-steps to 240 mV from 2110 mV holding potential. These currents for HA-tagged channels were carried out under optimal transfection efficiency and culturing conditions. Usually for untagged clones, peak currents were usually greater than 2000 pA, recorded three three days post-transfection, and we selected for smaller currents. No recorded HA-tagged channel was as large as 100 pA. (TIF) Methods S1 Supplementary Materials and Methods. (DOCX)