Voltage-gated calcium channels of Paramecium cilia

Cilia are slender organelles that protrude from eukaryotic cell surfaces. Depending on the cell type, there may be a solitary primary cilium or multiple cilia on a cell surface. These cilia have in common internal microtubule structures called axonemes, and those cilia with additional motor protein machinery in their axoneme can be motile. Primary cilia generally are immotile. The cilia of multi-ciliated cells usually are motile and contribute to the flow of spinal fluid or mucous in the airways. When cilia are dysfunctional, they contribute to a class of diseases called ciliopathies that are complex syndromes of developmental abnormalities. Regardless of superficial differences, all cilia are antennae for eukaryotic cells to sense many kinds of environmental stimuli (Berbari et al., 2009; Bloodgood, 2010; Brenker et al., 2012; Pazour and Witman, 2003; Singla and Reiter, 2006; Valentine et al., 2012).

As antennae, cilia detect stimuli such as odorants or mechanical forces, and convert these stimuli to intra-ciliary signals (reviewed in Bloodgood, 2010; Kleene and Van Houten, 2014; Lishko and Kirichok, 2015). Although the identities of channels and the roles of ciliary versus cellular sources of Ca²⁺ are now under re-examination for mammalian primary and ependymal and nodal cilia (DeCaen et al., 2013; Delling et al., 2013, 2016; Blum and Vick, 2015; Lee et al., 2015; Doerner et al., 2015), for protists and sperm, the influx of Ca²⁺ into cilia is clearly through the ciliary voltage-gated Ca²⁺ (Ca_V) channels in response to chemical signals, mechanosensation or depolarization. Calcium levels in the cilia control signaling output, such as the beat form and force of motile cilia or cell proliferation (Bloodgood, 2010; Kleene and Van Houten, 2014).

Whether cilia are solitary or numerous, motile or non-motile, ion channels control the intra-ciliary calcium levels, which play important roles in sensory transduction (Bloodgood, 2010). For example, cyclic nucleotide-gated Ca²⁺ channels that are found only in the immotile cilia of olfactory neurons open as part of the signal transduction pathway initiated by odorant binding to its receptor (Pifferi et al., 2006; Kleene, 2008). Motile cilia similarly have ion channels that participate in sensory transduction. For example, the Ca_V channel of the Chlamydomonas flagellum is necessary for the change in waveform in response to light or mechanical stimulation (Fujiu et al., 2011, 2009). The TRP family member polycystin-2 (PKD2) in Chlamydomonas flagella is crucial for the mating process that is dependent upon a Ca²⁺ influx (Huang et al., 2007). The CatSper Ca²⁺ channels of the sperm flagellum are responsible for the change in waveform in the vicinity of the egg (Brenker et al., 2012).

Because cilia are so highly conserved, it is possible to use a model system such as the ciliate Paramecium tetraurelia to provide insights into the ciliary calcium compartment and ciliary ion channel function. Paramecium tetraurelia, which is a cell covered in cilia, responds to sensory stimuli with changes in swimming behavior that result from changes in ciliary beating (Kung and Saimi, 1982; Machemer, 1988). Ciliary beat frequency and form are controlled by intra-ciliary Ca²⁺, which, in turn, is controlled by action potentials of Ca_V channels. The Ca_V channels, which are necessary for the action potential and increase in intra-ciliary calcium, are found exclusively in the cilia (Dunlap, 1977). Depolarizing stimuli, such as mechanical stimulation and high K concentrations, activate this channel and initiate the calcium action potential. The Ca²⁺ entering the cilium through the Ca_V channels affects the ciliary beating, reversing the power stroke and transiently sending the cell swimming backward. As ciliary Ca²⁺ levels return to normal, the cell pivots in place and then swims forward in a new direction. Thus, the activation of the voltage-gated calcium current [I_Ca(v)] of the Ca_V channels causes cells to swim backward transiently and make a turn in their swimming paths (Eckert, 1972; Machemer, 1988).

For almost 50 years, the physiological characteristics of the Paramecium ciliary Ca_V channels have been known (reviewed in Eckert, 1972; Machemer, 1988), but the proteins were not identified until recently, when the P. tetraurelia ciliary membrane proteome was determined. The three Ca_Vα1 subunits that were identified among the proteins (Yano et al., 2013) were subsequently cloned and confirmed to be expressed in the cilia (Valentine et al., 2012; present study). Here we demonstrate using RNA interference (RNAi) that these channels function as the ciliary Ca_V channels that are responsible for the reversal of ciliary beating. Furthermore, we have turned to the important mutants of Paramecium called Pawns (pw), which are named for the chess piece because they cannot swim backward for lack of ciliary Ca_V channel activity (Satow and Kung, 1974). Here we show that pw mutants do not express the Ca_V1a, b or c channels in their ciliary membrane. However, when the pw mutants are rescued from the mutant phenotype (i.e. swim backward and turn) by expression of the wild-type PW gene, the Ca_V1a, b and c channels can be found in the ciliary membrane as in the wild type, reinforcing the correlation of these three Ca_V channels with the calcium action potential that causes backward swimming through ciliary reversal.

Paramecium tetraurelia wild-type (51s sensitive to killer), nd6 (non-trichocyst discharge mutant, courtesy of Dr Jean Cohen, Centre de Génétique Moléculaire, Gif-sur-Yvette, France), pwA (d4-94), pwA-nd6 double mutant (d4-94-nd6) and pwB (d4-95) cells were used. The cells were cultured in wheat grass infusion medium inoculated with Aerobactor aerogenes (Sasner and Van Houten, 1989; Wright and Van Houten, 1990).

Previous physiological studies indicated that the Ca_V channels of the action potential were exclusively in the cilia (Dunlap, 1977; Machemer and Ogura, 1979), which led our search for Ca_V channels to the cilia. We describe three putative Ca_V α1 subunit genes as identified in our proteomic study of P. tetraurelia ciliary membrane: GSPATG00010323001 Ca_V1a, GSPATG00033414001 combined with GSPARG00033415001 Ca_V1b, and GSPATG00017333001 Ca_V1c (Yano et al., 2013) (see Fig. 1, Table S1). The genes for Ca_V1a and b are paralogs and are closely related to the gene for Ca_V1c.

We used OMIGA 2.0 and DS GENE (Accelrys, San Diego, CA, USA) for multiple alignments of the P. tetraurelia sequences with the primary structures of Ca_V α-subunits from all three mammalian subfamilies (Ca_V1–3) (McRory et al., 2001; Mikami et al., 1989; Starr et al., 1991; Tyson and Snutch, 2013). We analyzed the sequences of the Ca_V1a–c genes for expected conserved domains (Fig. 1). We used TMHMM Server v. 2.0 (http://www.cbs.dtu.dk/services/TMHMM/) and SOSUI (http://bp.nuap.nagoya-u.ac.jp/sosui/sosuimenu0.html) for domain analysis. Calmodulin Target Database (http://calcium.uhnres.utoronto.ca/ctdb/ctdb/sequence.html) was used to locate putative calmodulin binding domains.

Segments of 2055, 1688 and 1514 bp corresponding to the sequences between putative domains II and IV in Ca_V1a, 1b and 1c genes, respectively, were amplified by PCR using genomic macronuclear DNA as template (Yano et al., 2003) (for a diagram of these channels, see Fig. 1). PCR primers for this process are shown in Table S2. The 23-mers that could theoretically be generated by RNAi were compared with other genes using the RNAi off-target tool in ParameciumDB (http://paramecium.cgm.cnrs-gif.fr/cgi/tool/alignment/off_target.cgi). Any off-target overlap from the RNAi constructs and other channels is shown in Table S3.

The PCR products were ligated to the vector L4440 (AddGene, Cambridge, MA, USA), using restriction enzymes (Xho1 and Xba1, New England BioLabs, Ipswich, MA, USA) and a ligation kit (Ligate IT, Affymetrix, Santa Clara, CA, USA). The plasmid (empty or with insert) was transformed into strain HT115 of Escherichia coli following the manufacturer's instructions. The feeding RNAi experiments were performed following a published protocol (Valentine et al., 2012).

All solutions for testing swimming behavior contain a base buffer of 1 mmol l⁻¹ Tris and 1 mmol l⁻¹ calcium citrate. 100 mmol l⁻¹ Tris was used to adjust the pH of solutions to pH 7.0. Cells from the control and test RNAi cultures were transferred to depression slides with a resting solution, 4 mmol l⁻¹ KCl in base buffer, pH 7.0, for 20 min. Cells were then transferred one by one to depressions with the test solution (30 mmol l⁻¹ KCl in base buffer, pH 7.0). The control cells first whirled or jerked and then swam backward. At the end, they whirled again and started swimming forward. The durations of backward swimming, measured using a stopwatch, were analyzed using Mann–Whitney U-tests. We also used 8 mmol l⁻¹ BaCl₂ in base buffer, pH 7.0, as the depolarization solution to induce backward swimming in pw mutants that were injected with wild-type sequences of the PWA or PWB genes.

The wild-type PWA (AF050753) and PWB (AF179276) genes were amplified by PCR using macronuclear genomic DNA. The primers used for amplification and restriction enzymes used for cloning into the expression plasmids are shown in Table S4. The PCR products were ligated into the vector pPXV (Haynes et al., 1995), or into pPXV with 3×Myc or 3×FLAG sequences for C-terminus tagging of the expressed protein (Valentine et al., 2012). Full-length sequences from the Ca_V channels Ca_V1a and 1b were amplified by PCR using TaKaRa LA Taq DNA polymerase (Takara/Clontech, Mountain View, CA, USA) and inserted into the pPXV with a 3×FLAG tag at the N terminus following our previously published description of inserting Ca_V1c into pPXV-3×FLAG for an N-terminal tag (Valentine et al., 2012; Yano et al., 2013). Inserted sequences were confirmed by the DNA sequencing in the Vermont Cancer Center Advanced Genome Technologies Core.

Following established procedures (Haynes et al., 1995; Yano et al., 2013), a solution of linearized plasmid (5 µg µl⁻¹) was injected into the macronuclei of cells; the individual injected cells were cultured. The plasmid with the PWA or tagged PWA or PWB sequence was injected into the macronucleus of pwA-nd6 or pwA (d4-94) cells, and the plasmid with PWB or tagged PWB into the macronucleus of pwB (d4-95) cells. When FLAG-Ca_V1s and PWA wild-type or PWB wild-type genes (with or without tags) were to be co-expressed, the FLAG-Ca_V1 plasmids were the first plasmids injected into wild-type or pw mutant cells.

Genomic DNA from the cell lines was used as template in PCR to determine which lines had the highest concentration of exogenous gene copies for a second injection of a different plasmid or further experiments. To do so, the endogenous and exogenous (plasmid) DNAs were separately amplified by PCR, using a forward primer from the coding region, and a reverse primer from its 3′ un-transcribed region (for endogenous sequences) or calmodulin 3′ un-transcribed region of pPXV plasmid (for exogenous plasmid sequences). The intensity of PCR bands from endogenous and exogenous mRNA were analyzed using ImageJ (National Institutes of Health, Bethesda, MD, USA) to compare indirectly the relative amounts of plasmid copies in cell lines. Cells injected with plasmids containing the epitope tag sequence only were used as a control to match the experimental cells in each experiment.

Two to three milliliters of packed cells were collected from 3 to 6 liters of control and test cultured cells. The cells were washed twice in Dryl's solution (1 mmol l⁻¹ Na₂HPO₄, 1 mmol l⁻¹ NaH₂PO₄, 2 mmol l⁻¹ trisodium citrate, 1.5 mmol l⁻¹ CaCl₂) and once in cold LAP200 (50 mmol l⁻¹ HEPES, pH 7.4, 200 mmol l⁻¹ KCl, 1 mmol l⁻¹, EGTA, 1 mmol l⁻¹ MgCl₂) with protease inhibitors [1 mmol l⁻¹ phenylmethylsulfonyl fluoride (PMSF, Sigma-Aldrich, St Louis, MO, USA), 0.1 µg ml⁻¹ Pepstatin (Research Products International, Mt Prospect, IL, USA), 0.1 µg ml⁻¹ Leupeptin (Research Products International) and protease inhibitor cocktail (Sigma-Aldrich) at a final concentration of 0.1%]. The washed cells were resuspended in 5 ml of LAP200 buffer with the protease inhibitors and maintained at 4°C while being homogenized until 95% of the cells were ruptured.

For work flow, see Fig. 2. The pellicles (cell surface membranes with cytoskeleton) were prepared as described previously (Wright and Van Houten, 1990). In this preparation, cells were washed three times in cold HM buffer (20 mmol l⁻¹ maleic acid, 20 mmol l⁻¹ Tris, 1 mmol l⁻¹ Na₂EDTA, pH 7.8) instead of Dryl's solution. The cells were homogenized in HM buffer with the protease inhibitors at 4°C. The pellet was washed, collected as the ‘pellicle’ and further treated with detergent and centrifugation to produce a supernatant for immunoprecipitation (IP). The first supernatant from the pellicle preparation was vortexed vigorously for 5 min and centrifuged at 2000 g for 10 min at 4°C. The supernatant was then centrifuged at 19,800 g for 30 min at 4°C to produce the pellet P2. The resulting supernatant was centrifuged at 100,000 g for 1 h at 4°C to produce the pellet P10. Intracellular membranes with an enzyme signature (glucose-6-phosphatase) of the endoplasmic reticulum (ER) are reported to be enriched in the P2 and P10 pellets (Haga et al., 1984; Wright and Van Houten, 1990; S. Lodh, Characterization of PWA and PWB proteins in Paramecium, PhD dissertation, University of Vermont, 2012). The pellicle, P2 or P10, were resuspended in membrane buffer (10 mmol l⁻¹ Tris buffer, pH 7.4, 50 mmol l⁻¹ KCl, 5 mmol l⁻¹ MgCl₂, 1 mmol l⁻¹ EGTA) or LAP200 buffer with the protease inhibitors (1 mmol l⁻¹ PMSF, 0.1 µg ml⁻¹ Pepstatin, 0.1 µg ml⁻¹ Leupeptin and 0.1% protease inhibitor cocktail) at 4°C (Haga et al., 1982, 1984; Wright and Van Houten, 1990).

Cytoplasmic factors that cure the three pw mutants and restore action potentials and backward swimming are thought to be in the subcellular membrane fraction; the fraction that we refer to as P10 is the equivalent of the ‘P2’ fraction that Haga and others used to cure P. tetraurelia pw mutants by injection (Haga et al., 1982, 1984).

The cilia from the control and test cells were isolated from 2–3 ml of packed cells collected from 3–6 liters of cell culture following the protocol of Yano and co-workers (Yano et al., 2013). The isolated cilia were suspended in 1 ml of membrane buffer or LAP200 with protease inhibitor at the same concentration previously described at 4°C.

The protein concentration of each sample was measured using the Pierce Protein Assay (Thermo Fisher Scientific, Waltham, MA, USA). The cilia samples from the control and test were adjusted to the same volume and protein concentration for further experiments.

Triton X-114 for solubilizing Ca_V1s, calcium ATPases (PMCA) or Pw proteins, or Triton X-100 for solubilizing Pw proteins (as indicated in each specific figure) was added to the whole-cell lysate, pellicle, P2, P10 or cilia suspension to achieve a final concentration of 1%. Each sample was agitated by rocking at 4°C for 1 h followed by centrifugation at 48,400 g (pellicle and cilia) or 100,000 g (whole-cell lysate and P10) at 4°C for 30 min. The supernatant was incubated with 20 µl Protein A beads (Amersham Pharmacia/GE HealthCare, Pittsburgh, PA, USA) at 4°C for 1 h to clarify the supernatant. Prior to use, the Protein A beads had been washed in membrane buffer or LAP200 buffer with 1% Triton and 1% (w/v) bovine serum albumin (BSA). After removing the beads by centrifugation at 48,400 g at 4°C for 30 min, the clarified supernatant was incubated with 20 µl of anti-FLAG M2 affinity agarose (Sigma-Aldrich), or anti-c-Myc (polyclonal antibody) affinity agarose (Sigma-Aldrich) at 4°C for 1 h, which had been pre-washed in the membrane buffer or LAP200 with 1% Triton and 1% BSA.

Next, the antibody-conjugated beads were washed three times by centrifugation at 8000 g in the membrane buffer or LAP200 with 1% Triton and 1% BSA and three times in the buffer without Triton and BSA. The beads were suspended in 50 µl of 2× sodium dodecyl sulfate (SDS) buffer (6.25 mmol l⁻¹ Tris, 1.5% SDS, 1% glycerol, 0.001% Bromophenol Blue, pH 6.8) with or without 3% β-mercaptoethanol, and boiled for 10 min. After centrifuging at 14,000 g at 4°C, the supernatant was loaded onto a 4–18% or 7–18% gradient SDS-polyacrylamide gel (SDS-PAG) and run at the constant current of 20 mA. To confirm that approximately the same amounts of protein went into the control and test IPs, we removed 20 µl of the test and control samples, clarified the Triton-treated supernatants and analyzed the proteins on western blots using α-tubulin as a protein that should not vary between control and test samples.

For the IP of the Ca_V channels from the ciliary membrane, plasma membrane calcium ATPases (PMCAs) were used to assess gel loading. The same IP protocol as described above for cells expressing FLAG-Ca_V1c was used. In addition, the supernatant that had been clarified with Protein A beads was incubated with 5 µg of rabbit anti-calmodulin binding domain antibody of P. tetraurelia PMCA2 (anti-CBD2 antibody) (Van Houten, 1998) at 4°C for 1 h followed by the incubation with Protein A beads for IP.

We examined the subcellular localization of expressed epitope-tagged PwA and PwB proteins by western blot analysis. A total of 100 µg of protein from the pellicle, P2, P10, or cilia suspension that was prepared from both the test and control cells were run on 12% SDS-PAG. The proteins separated in the SDS-PAGE (SDS-PAG electrophoresis) were transferred to BioTrace nitrocellulose blotting membrane (PALL Life Sciences, Pensacola, FL, USA). The western blots were treated with blocking buffer, incubated with primary antibody, followed by development with alkaline phosphatase or enhanced chemiluminescence, as previously described (Yano et al., 2003). The blots were probed for the protein of interest with the proper primary antibody: rabbit or mouse anti-FLAG antibodies (Sigma-Aldrich, F3165-5MG, F7425-0.2MG), 1:2500 dilution; rabbit or mouse anti-Myc antibodies (GenScript, Piscataway, NJ, USA, A00704), 1:2000 dilution; rabbit anti-CBD2 (for the plasma membrane calcium ATPases), 1:5000 dilution; or mouse anti-α-tubulin (loading control) (Sigma-Aldrich, T6199), 1:10,000 dilution. Secondary antibodies were goat anti-mouse or rabbit conjugated to alkaline phosphatase or horse radish peroxidase used in a 1:10,000 dilution.

When the rabbit antibody was used for IP, the precipitated proteins were detected on western blot with the appropriate mouse primary antibody. When the mouse antibody was used for IP, a rabbit antibody was used for western blot analysis. For re-probing, the blots were incubated in the stripping buffer (50 mmol l⁻¹ dithiothreitol, 50 mmol l⁻¹ Tris HCl, 70 mmol l⁻¹ SDS, pH 7) at 70°C for 30 min before washing in TBS-T (16 mmol l⁻¹ Tris HCl, 4 mmol l⁻¹ Tris, 137 mmol l⁻¹ NaCl, 0.1% Tween 20, pH 7.5) and re-blocking and re-probing the blot as just described.

We used mass spectrometry (MS) to confirm that the bands precipitated with anti-FLAG M2 affinity agarose from whole-cell lysates of PWA-FLAG- and PWB-FLAG-expressing cells were FLAG-tagged PwA and PwB proteins, respectively. The precipitated proteins were separated on 12% SDS-PAGE. The gels were silver stained using the FAST Silver Kit (G-Biosciences, St Louis, MO, USA). The region corresponding to PwA- or PwB-FLAG was cut from the gel for MS analysis (see immediately below).

We examined whether the Ca_V1c channel could be immunoprecipitated with PwA from the P10 pellet in Fig. 2. Cultures of cells expressing both FLAG-Ca_V1c and PWA-Myc could not be cultured in sufficient quantities to analyze the results of IPs by western blots. Therefore, we used cells that grew better, i.e. expressing FLAG-Ca_V1c that had been transformed with untagged PWA expression plasmid. The P10 fraction was prepared from the cells expressing PWA and FLAG-Ca_V1c or PWA and the control FLAG epitope. The P10 fractions were normalized for protein concentration and solubilized with 1% Triton X-114 before centrifugation at 100,000 g.

After centrifugation, we performed the IP from the resulting supernatant using anti-FLAG M2 affinity agarose using the method described above. The precipitated proteins were separated on 4–18% gradient SDS-PAGE. The resulting gel was silver stained and the regions corresponding to Ca_V1c (200–270 kDa) and PWA (20–30 kDa) were cut out. Each band was diced, destained in 30 mmol l⁻¹ K₃Fe(CN)₆ and 100 mmol l⁻¹ Na₂S₂O₃ in distilled water, and subjected to in-gel digestion with trypsin buffer (5% CH₃CN, 25 mmol l⁻¹ NH₄HCO₃, 6 ng µl⁻¹ trypsin) overnight at 37°C. The resulting peptides were analyzed by LC-MS/MS in an LTQ-XL linear ion trap mass spectrometer (Thermo Fisher Scientific) following the protocol of Yano and co-workers (Yano et al., 2013). The resulting protein data were searched simultaneously against the Paramecium tetraurelia forward (target) and reverse (decoy) peptide database (http://paramecium.cgm.cnrs-gif.fr/download/fasta/Ptetraurelia_peptides_v1.99.14.fasta) using Scaffold 4 (Proteome Software, Portland, OR, USA) with a precursor tolerance of 2 Da and a fragment ion tolerance of 0.5 Da. In our experience, the majority of cysteine residues following reducing conditions and SDS-PAGE are identified with an acrylamide adduction. For increased throughput and simplicity, we conducted searches with a static increase in 71.0 Da for acrylamide adduction. Differential modification of 16.0 Da on methionine residues was permitted. The search results were filtered using a delta correlation (dCn) score of 0.1 and cross-correlation (Xcorr) values of 1.9, 2.6, 3.2 and 3.4 for singly, doubly, triply and quadrupally charged ions, respectively. Proteins on these filtered lists that had two or more peptides were retained, and the false-positive ratio was zero in this list.

Total RNA, first-strand cDNA and PCR were carried out as in Yano et al. (2003) and Valentine et al. (2012). Instead of serial cDNA dilution, PCR cycles were changed to 10, 15 and 20. See Fig. 3 for an example of the RT-PCR analysis that we carry out to determine whether the RNAi process is decreasing the mRNA of interest. We do not use this as a quantitative method. Note that the mRNA (cDNA) is not completely eliminated by RNAi.

The backward swimming duration (s) values are shown as means±s.d. The Mann–Whitney U-test with two-tailed distribution was used for statistical analysis.

In our proteomic analysis of the P. tetraurelia ciliary membrane (Yano et al., 2013), peptides from the putative Ca_V1 α-subunit were identified. These Ca_V peptide sequences correspond to Ca_V1a, 1b and 1c. Ca_V1a and Ca_V1b are 87% identical at the nucleotide level and are likely to be derived from a recent whole genome duplication (WGD) (Aury et al., 2006). Their amino acid sequences are so close that the peptides we identified did not distinguish between Ca_V1a and 1b, but the peptides allowed us to establish that one or both of them are in the ciliary membrane. While Ca_V1c is 75% identical at the nucleotide level to Ca_V1a and 1b, Ca_V1c can be distinguished from these other Ca_V proteins through peptides that we identified in our mass spectrometry analysis. The gene for Ca_V1c probably separated from the more ancient paralogs at the intermediate WGD and survived after the more recent WGD that created Ca_V1a and 1b (Aury et al., 2006). A FLAG-tagged sequence for Ca_V1c was previously used to confirm the presence of the protein in cilia (Valentine et al., 2012; Yano et al., 2013).

Blast searches showed that Ca_V1a, 1b and 1c are most closely related to the Ca_Vα1 subunits from the mammalian subfamily called Ca_V1. When the Paramecium database was searched using the mouse and rat sequences for Ca_V1.1 (NP_055008 for mouse and NP_446325 for rat) that are α1 subunits of typical high voltage-activated (L-type) channels, the Paramecium Ca_V1a, 1b and 1c have Expect (E) values of 6E⁻⁹⁵, 8E⁻¹⁰⁶ and 4E⁻⁹⁸ to mammalian Ca_V1.1, respectively (see Table S1).

The P. tetraurelia ciliary Ca_V α-subunits share conserved domains with the sequences of the three vertebrate Ca_V subfamilies Ca_V1–3. Our analyses of predicted structure show that P. tetraurelia Ca_V1a, 1b and 1c all have the expected four copies of the ion transporting domain comprising six transmembrane domains (s1 to s6) and a pore loop in each unit (Tyson and Snutch, 2013) (Fig. 1). These four channel domains come together to form a highly charged selectivity filter that confers specificity for Ca²⁺ on the channel. The sequences found in the P. tetraurelia sequences have glutamic acids (E) in a critical position in each pore loop, giving these Ca_V α subunits the identity of the ‘EEEE’ motif in common with vertebrate Ca_V1 and 2, which are associated with high voltage-activated calcium channels. All CaV subgroups have transmembrane segments S3 and S4 that generally have an NxxD sequence and an R/KxxRxxxRxxR/K voltage sensor motif, respectively. The P. tetraurelia sequences likewise conserve these sequences (Fig. 1). The C termini of vertebrate Ca_V1 and 2 have a calmodulin binding motif and an IQ motif. Upon examination of the C termini, we found a putative calmodulin binding site in the C-terminal cytoplasmic region of P. tetraurelia Ca_V1a, 1b and 1c (Fig. 1). The cytoplasmic loop between domains I and II of P. tetraurelia Ca_V1a–c is very long (around 680 amino acids) as compared with that of the mammalian Ca_V1 and 2.

Backward swimming is known to depend upon an I_CaV in the cilia. The Ca²⁺ conductance is proportional to the duration of backward swimming induced by depolarization with high potassium (Haga et al., 1984; Hiwatashi et al., 1980). We used RNAi to examine whether reduction of Ca_V1a, 1b, 1c, or all three affected backward swimming, and, indirectly, whether their channel activities participate in the backward swimming behavior.

Segments of the Ca_V1a, 1b and 1c sequences (see Materials and methods) were amplified by PCR and sub-cloned into the RNAi vector L4440 for feeding RNAi. Paramecium tetraurelia cells were fed bacteria with the RNAi vector with the Ca_V insert or the empty RNAi vector (L4440) as a control. Cells were first tested in 30 mmol l⁻¹ KCl in buffer to induce backward swimming after 24, 48 and 72 h of feeding on the RNAi bacteria, but changes in the backward swimming duration were most dramatic at 72 h of feeding. Therefore, we present here the data collected at 72 h of feeding RNAi. The RNAi with each Ca_V1a, 1b or 1c sequence individually caused significantly shorter backward swimming in high KCl compared with the control fed the empty vector (Mann–Whitney U-test; Fig. 4). Moreover, the RNAi for the mixture of Ca_V1a, 1b and 1c showed the shortest backward swimming as compared with the control or with RNAi for each individual calcium channel (Fig. 4).

Off-target effects by the RNAi sequences developed to downregulate Ca_V1a–c showed that the potential 23-mer nucleotide products from RNAi processing of the double-stranded RNA for Ca_V1a could bind many sequences within the mRNA for Ca_V1b and vice versa, marking these sequences for degradation (Table S3). However, there are very few potential off-target effects of Ca_V1a or 1b on Ca_V1c and of Ca_V1c on either Ca_V1a or 1b.

As a negative control for our RNAi of Ca_V1a–c, we carried out RNAi for gene sequence GSPATG00005636001, which has been identified previously (Ben-Johny et al., 2014; Taiakina et al., 2013) as a putative Ca_V channel α1-subunit based on sequence homology to mammalian Ca_V1.1. This putative Ca_V α-subunit has not been found in the proteomic analysis of the ciliary membrane. RNAi for this sequence does not reduce backward swimming (Fig. 5) as seen for the RNAi of Ca_V1a, 1b or 1c shown in Fig. 4.

Haynes and others have shown that injection of the wild-type sequence for PWA or PWB into the mutant cell nucleus will rescue the wild-type phenotype (Haynes et al., 2000, 1998). In order to carry out the present study, we reproduced the outcome that the wild-type PW sequences could rescue the wild-type phenotype, i.e. restore the ability of pw cells to swim backward in depolarizing solutions. Table S5 shows that injection of the wild-type PWA or PWB sequence restores the ability of pwA or pwB mutants, respectively, to swim backward in 8 mmol l⁻¹ BaCl₂, although the duration is not as long as the backward swimming of wild-type controls. The PWA or PWB sequences with epitope tags similarly restore the ability of pwA or pwB mutant cells to swim backward in 8 mmol l⁻¹ BaCl₂ solutions (Table S5).

In contrast, expression of FLAG-Ca_V1a, 1b or 1c in pwA or pwB cells does not result in the restoration of backward swimming tested in 30 mmol l⁻¹ KCl solutions (Table S5). Even with these additional exogenous sequences for Ca_V1a, 1b or 1c injected into the pw cells, the wild-type phenotype is not rescued in pw mutants.

The FLAG-tagged Ca_V1c can be immunoprecipitated from the ciliary membrane of wild-type cells (Fig. 6A) and its expression in wild-type cells increases their backward swimming (Table S5). Similarly, FLAG-Ca_V1a or 1b were immunoprecipitated from the ciliary membrane of wild-type cells expressing FLAG-Ca_V1a or 1b, respectively (Fig. S1). However, the cells expressing tagged Ca_V1a or 1b showed the same backward swimming duration as the control cells (Table S5). The expression of the tagged Ca_V1s' proteins in wild-type cilia made it possible for us to investigate the possibility that the pw mutants do not activate the Ca_V1s to initiate backward swimming because the Ca_V1s are not in their cilia.

FLAG-Ca_V1c was expressed in mutant pw cells or in pw cells whose phenotypes had been rescued with wild-type PW sequences (Fig. 6A). In each case, the cells were tested for restoration of backward swimming before the cilia were isolated from transformed cells. The isolated cilia were treated with 1% Triton X-114 and the proteins were precipitated using anti-FLAG M2 affinity agarose, then analyzed by western blot. Note that in lane C of both pwA and pwB cells shown in Fig. 6A, the FLAG-Ca_V1c is not immunoprecipitated from the cilia of pwA or pwB cells co-expressing only the pPXV plasmid. However, once the pw mutants are rescued with the wild-type sequence for PWA or PWB, the FLAG-Ca_V1c protein can be immunoprecipitated from the cilia (lane T of pwA and pwB cells in Fig. 6A).

As a concentration control for Fig. 6A, the plasma membrane calcium ATPases 2, 3 and 4 (PMCA) (CR932147, CR932150 and CR933346) were precipitated from the cilia with anti-CBD2 antibody (Van Houten, 1998). The western blot showed that the intensity of bands corresponding to the PMCAs were almost the same between the cilia of cells expressing the FLAG-Ca_V1c and control vector pPXV, and the cilia co-expressing FLAG-Ca_V1c and the PWA or PWB wild-type gene (Fig. 6B).

Similar results were obtained when the tagged channel genes for Ca_V1a and 1b were co-expressed with the wild-type PWA in pwA cells or PWB gene in pwB cells. Under these conditions, FLAG-Ca_V1a and 1b proteins could be immunoprecipitated from the cilia of pwA or pwB cells (see Fig. S1).

The injection of cytoplasm, especially fraction P10 (see Fig. 2 and Materials and methods), from the wild-type cells into the pwA or pwB cells caused the mutants to regain the voltage-activated calcium conductance without new protein synthesis (Haga et al., 1984). Therefore, we included in our determination of the subcellular localization of Pw proteins the P10 fraction used by Haga and co-workers to cure pw phenotypes (Haga et al., 1984) and also the P2 fraction, because both fractions were reported to be enriched in ER (Haga et al., 1984; Wright and Van Houten, 1990).

We used western blots to locate Pw proteins in cell fractions. Fig. 7A shows western blots of the P2 fraction and the pellicle, and western blots of IPs from the cilia of pwA cells expressing PwA-FLAG (test, T lane) or FLAG (control, C lane) (see Materials and methods). We considered bands to be from PwA-FLAG only if they were found in T lanes and not C lanes, as only the T lanes should have the expressed tagged protein. The blots of the P2 fraction show three bands (30, 27 and 25 kDa) in the T lane and one (30 kDa, blue arrow) in the C lane. We discounted the 30 kDa band from further analysis because it was in both the T and C lanes. The P2 band at 27 kDa (upper black arrow, T lane only) matches the expected size of the PwA-FLAG protein. The P2 band of 25 kDa (lower black arrow, T lane only) matches the PwA-FLAG protein without the signal sequence.

Blots of the pellicles of cells expressing PwA-FLAG and FLAG show three bands of 25, 27 and 28 kDa in the T lane only. The pellicle band of 28 kDa (arrowhead) matches a glycosylated form of the PwA-FLAG protein, which has two putative glycosylation sites. The lower band in the pellicle blot at 25 kDa matches the predicted size of PwA-FLAG without the signal sequence. In the blot from the IP of PwA-FLAG from cilia, two bands of 25 and 28 kDa were detected in the T lane only, similar to the pellicle western blot.

Proteins were immunoprecipitated with anti-FLAG M2 from the Triton-114 extracts of P2, P10, pellicle and cilia from pwB cells expressing PWB-FLAG (T lane) or FLAG (C lane; Fig. 7B). Western blots show bands at 35 kDa in the T lane for the P2 and P10 fractions (Fig. 7B, black arrow), which are consistent with PwB-FLAG protein. The PwB-FLAG protein was not found in the pellicle or cilia IPs.

To examine whether the PwB protein interacts (directly or indirectly) with Ca_V1c in the P10 fraction, we carried out reciprocal IPs of FLAG-tagged Ca_V1c and PwB-Myc proteins expressed in wild-type cells. Fig. 8A,B shows the western blot results of reciprocal IPs of FLAG-Ca_V1c and PwB-Myc proteins from the P10 fraction. After the IP with the anti-Myc beads, a FLAG-Ca_V1c band of 250 kDa was detected in lane T (Fig. 8A). As expected, the PwB-Myc protein of 35 kDa was also detected. From the IP with anti-FLAG M2 affinity agarose, the PwB-Myc protein was detected at 35 kDa (Fig. 8B). The loading control is shown in Fig. 8C,D.

To examine whether the PwA protein interacts with FLAG-Ca_V1c in the P10 fraction, FLAG-Ca_V1c and PwA-Myc (test) and empty vectors containing FLAG and Myc (control) were co-expressed in wild-type cells. We could not grow the PWA-Myc transformed cells to sufficient density for the IPs, which led us to look for the presence of the PwA protein by overexpressing the untagged version and using MS to examine the gel of the IP products. IP with anti-FLAG was carried out, and the immunoprecipitated proteins were analyzed by separating the precipitate into a sample for a blot to confirm the successful IP of Ca_V1c and a sample for a silver-stained gel that would be analyzed by MS/MS. Two regions, 200–270 kDa and 20–30 kDa, corresponding to Ca_V1c and PwA proteins, respectively, were cut out from the silver-stained gel (Fig. 9) and analyzed by MS. The same experiments were repeated three times. Although the peptides of Ca_V1c were detected from the regions corresponding to FLAG-Ca_V1c, no peptides from the PwA protein were detected.

The power stroke of the cilia of P. tetraurelia is controlled by the activity of the Ca_V channels in the ciliary membrane. Through proteomic analysis of the ciliary membrane, we found three proteins and their corresponding genes that could potentially be responsible for the action potential that controls the ciliary beat form (Yano et al., 2013). The cloning and epitope tagging of these large genes was challenging, but allowed us to demonstrate that the proteins from the expression vectors were located in the ciliary membrane. The first channel, Ca_V1c, was the first of three that we cloned and expressed (Valentine et al., 2012). The expressed Ca_V1a and 1b channels are shown in the present study. The epitope tags allowed us to surmount the challenge of immunoprecipitating these large proteins to concentrate them for definitive identification in wild-type and rescued pwA and pwB mutants.

Using RNAi, we were able to downregulate the expression of the Ca_V1a, 1b and 1c genes singly and together to demonstrate that these Ca_V α-subunits contribute to the action potential that causes the cells to turn. RNAi generally does not cause a complete removal of a protein as a null mutation would. The mRNA for these channels is not eliminated by RNAi, as shown by RT-PCR (Fig. 3). However, the extreme reduction of backward swimming by RNAi treatment, especially that seen in the cells depleted of all three Ca_V1a–c channels, gives us confidence that these channels are the major, if not the only, contributors to the calcium action potential.

Another related sequence not found in the cilia by proteomics was used as an RNAi control. Downregulation of the expression from this gene had no effect on the depolarization-induced backward swimming (Fig. 5).

We turned to the pw mutants for additional evidence that the presence of the Ca_V1a–c proteins in cilia correlates with the action potential and depolarization-induced backward swimming. In addition, the techniques of tagging and immunoprecipitation of the large channels allowed us to address the very old question of the cause of the failure of Pawn pwA and pwB mutants to show a calcium conductance upon depolarization that would normally elicit an action potential and backward swimming in the wild type.

Pawn mutants, pwA, pwB and pwC, were first described in 1969 (for a history, see Kung, 1971). Electrophysiological studies showed that Pawns lack the I_Ca(V) current, but otherwise have normal K⁺ and other conductances (Satow and Kung, 1974, 1980). Between 1998 and 2000, the PWA and PWB genes of the pwA and pwB mutants were cloned by complementation, but neither appeared to code for a Ca_V α or ancillary subunit (Haynes et al., 2000, 1998) (the gene for the pwC mutant remains unidentified). It had been shown that if the ciliary membrane were bypassed, and Ca²⁺ had access to the axoneme, pwA and pwB cells could beat their cilia with the reversed power stroke. However, when the ciliary membrane was intact, they could not (Kung and Naitoh, 1973). These experiments demonstrated that the axonemes of pw cells were functional. More recent experiments showed that the Ca²⁺ in cilia that enters through the Ca_V channels of wild-type cells does not spill into the cell body to trigger the Ca²⁺-dependent exocytosis of trichocysts, but that strong chemically induced exocytosis that is dependent upon a burst of Ca²⁺ released from intracellular stores could elicit the reversal of the ciliary beat, that is, reach the axoneme (Husser et al., 2004). While pw cells could not reverse their ciliary beating with depolarization, the strong chemically induced exocytosis could sweep enough Ca²⁺ into the cilia to cause a ciliary reversal. These results are consistent with the observations of Kung and Naitoh (1973) that the pw cell's axonemes were functional if Ca²⁺ could reach them.

It was not known whether intact pwA and pwB mutants did not reverse their beat because they lacked the Ca_V channels in their cilia or whether the pw mutants had Ca_V channels in their ciliary membranes but these channels could not be activated with depolarization. In the present study, we could not find the Ca_V1a–c expressed channels in the ciliary membranes of Pawn mutants pwA or pwB, suggesting that these mutants cannot swim backward for lack of the Ca²⁺ current from the channels. Expressing the genes for these channels in the mutants did not restore the ability to reverse ciliary beating and these expressed channels could not be found in the ciliary membrane. However, restoring the wild-type versions of the PWA or PWB mutant genes not only restores the ability to reverse swimming, but also restores the Ca_V1a–c channels in the ciliary membrane. These results reinforce our contention that the presence of one of the channel types Ca_V1a, b or c is necessary and perhaps sufficient for backward swimming.

The Pawn proteins do not resemble vertebrate Ca_V channel α1 or other subunits that are thought to assist in trafficking of the vertebrate Ca_V α-subunits to the cell surface (Dolphin, 2012). However, they appear to be involved in trafficking the P. tetraurelia Ca_V1a–c proteins to the ciliary membrane. Because the PwB protein appears to be limited to the ER, its role in trafficking may be to assist the channels to reach the proper ER–Golgi pathway. The PwB and Ca_V1c proteins can be reciprocally immunoprecipitated, implicating an interaction (direct or indirect) between the two proteins that is sufficiently strong to survive immunoprecipitation. The PwA protein, while found in the ER, surface membranes and cilia, does not have a similarly strong interaction. Nonetheless, it appears to be crucial in guiding the Ca_V1a–c channels to the cilia.

Previously, we showed that other channels of the P. tetraurelia ciliary membrane, a calcium-activated K⁺ channel (SK1a) and polycystin-2 (PKD2), require Bardet–Biedel syndrome (BBS) proteins to traffic into the cilia (Valentine et al., 2012). In that same study, we found that Ca_V1c did not require the BBS8 protein of the BBSome complex to successfully reach the ciliary membrane.

We thank Dr Y. W. Lam for analyzing the mass spectrometry data.