Identification of a vacuolar proton channel that triggers the bioluminescent flash in dinoflagellates

In 1972, J. Woodland Hastings and colleagues predicted the existence of a proton selective channel (HV1) that opens in response to depolarizing voltage across the vacuole membrane of bioluminescent dinoflagellates and conducts protons into specialized luminescence compartments (scintillons), thereby causing a pH drop that triggers light emission. HV1 channels were subsequently identified and demonstrated to have important functions in a multitude of eukaryotic cells. Here we report a predicted protein from Lingulodinium polyedrum that displays hallmark properties of bona fide HV1, including time-dependent opening with depolarization, perfect proton selectivity, and characteristic ΔpH dependent gating. Western blotting and fluorescence confocal microscopy of isolated L. polyedrum scintillons immunostained with antibody to LpHV1 confirm LpHV1’s predicted organellar location. Proteomics analysis demonstrates that isolated scintillon preparations contain peptides that map to LpHV1. Finally, Zn2+ inhibits both LpHV1 proton current and the acid-induced flash in isolated scintillons. These results implicate LpHV1 as the voltage gated proton channel that triggers bioluminescence in L. polyedrum, confirming Hastings’ hypothesis. The same channel likely mediates the action potential that communicates the signal along the tonoplast to the scintillon.


Introduction
The first postulation that a depolarization-activated, proton selective channel (H V 1) should exist was published in 1972 by J. Woodland Hastings and colleagues [1]. A decade later, Thomas and Meech reported the first voltage-gated proton conductance to be identified by voltage-clamp studies, in snail neurons [2]. Subsequent electrophysiological studies have elucidated the defining characteristics of a family of voltage-gated proton-selective channels, H V 1, which have been found in amphibia [3], rat [4], human [5][6][7][8], insects [9], and both multicellular [10] and unicellular marine species [11,12]. H V 1 are exquisitely selective for protons [13], a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 resulting from a critical Asp residue in the S1 transmembrane helix [12,14] that is thought to interact with an arginine in S4 [15,16]. H V 1 have a single-channel conductance 10 3 smaller than most ion channels [17], reflecting the 10 6 lower concentration of permeant ions. They open with depolarization but their voltage-dependence is strongly influenced by both external and internal pH, pH o and pH i [18], such that a one unit change in either pH o or pH i (or in the pH gradient ΔpH = pH o -pH i ) shifts the g H -V relationship by 40 mV [19]. Predicted H V 1 genes are nearly ubiquitous in eukaryotic genomes and the protein has multiple demonstrated functions in various eukaryotic cells [20].
Dinoflagellate bioluminescence is a striking phenomenon, producing brilliant blue flashes in the ocean water when the organisms are stimulated mechanically [21] in the form of shear stress [22,23]. The biochemical basis of dinoflagellate light production was elucidated in seminal work over several decades by Hastings and colleagues, who focused mainly on Lingulodinium polyedrum (formerly Gonyaulax polyedra). This work established that the light originates in organelles called scintillons arising from evaginations of the central vacuole membrane [24][25][26][27][28]. Scintillons contain a breakdown product of chlorophyll [29] called luciferin that is bound to luciferin binding protein (LBP) at pH > 7, while at pH < 7 LBP releases luciferin, thereby enabling it to interact with the enzyme luciferase (LCF) that catalyzes the reaction with oxygen and luciferin to produce light [1,24,26,[30][31][32]. Low pH activates the LCF in bioluminescent dinoflagellates [26,33] providing a second mechanism by which protons trigger the flash.
Fogel and Hastings [1] reasoned that a hypothetical voltage-sensitive proton channel should respond to a stimulus-induced depolarization of the L. polyedrum vacuole membrane [34], and that such a channel would transport protons from the acidic vacuole into the scintillons, thus providing the pH change that triggers bioluminescence. H V 1 proton channels have been reported in several non-bioluminescent unicellular marine species [11,12] where they function in calcium fixation (coccoliths) and possibly in feeding (dinoflagellates). Here we identify a bona fide H V 1 in the bioluminescent species studied by Hastings and colleagues, L. polyedrum, and demonstrate its scintillon localization. In addition to confirming a longstanding prediction we authenticate a unique mode for H V 1 in the vacuole membrane as the control step in the signal transduction pathway that leads to dinoflagellate bioluminescence.

Results
Sequence similarity searches of RNA-seq assemblies [35,36] from the bioluminescent organism L. polyedrum revealed a sequence (gi: 346282507) with the signature sequence pattern [12] that has proven to be diagnostic of H V 1 (longest open reading frame from this assembled transcript is shown in S3 Table.) cDNA libraries prepared from L. polyedrum populations sampled at mid-light and mid-dark and probed with PCR primers designed using the RNA-seq data produced PCR products with expected sizes from this putative H V 1 (S1 Table). qRT-PCR shows that the RNA for the putative H V 1 was expressed in L. polyedrum (data not shown).

LpH V 1 is a bona fide voltage-gated proton channel
We ordered the synthesis of a mammalian codon-optimized gene (sequence shown in S3 Table) corresponding to the predicted L. polyedrum H V 1 gene. When the gene was expressed in a human cell line (HEK-293) the gene product produced voltage-and time-dependent currents in voltage-clamped cells (Fig 1). Non-transfected HEK-293 cells sometimes had small native proton currents, but otherwise had no significant conductances that exhibited time dependent activation or tail currents under our recording conditions. The proton current during depolarizing voltage pulses turned on more rapidly than in mammalian species in which activation time constants may be seconds at room temperature [4,5,6]. LpH V 1 current also differed from most species studied to date in turning on much more rapidly at more positive voltages (discussed below). Similar to H V 1 in all other species, LpH V 1 exhibited strong sensitivity to pH. As in the cell in Fig 1A-1C, when pH o was decreased, the voltage range over which the g H was activated shifted positively.
LpH V 1 can produce inward current. In the vacuole or scintillon membrane, the proton channel is expected to be oriented with its external side facing the vacuole. In order to mediate the flash-triggering action potential in the vacuole membrane and enable H + influx into the scintillon to activate luciferase, the channel in situ must be able to conduct inward current. At symmetrical pH, activation occurred well positive to 0 mV (e.g., Fig 1B) and thus only outward current was produced, as in all other species except for K. veneficum [12]. In fact, the average "threshold" voltage at which LpH V 1 was clearly activated was 46 ± 1.8 mV (mean ± SEM, n = 38; 23 cells and 15 patches), well positive to +23 mV reported for a variety of native proton currents, mostly in mammalian cells, or to -10 to +10 mV in hH V 1 ( Table 3 in [20]). However, Fig 1D shows that when there was a large inward pH gradient (simulating the low pH vacuole and high pH cytoplasm that exist in L. polyedrum), inward current was activated. In vivo, LpH V 1 is exposed to an enormous inward pH gradient (ΔpH is 3.5), because the flotation vacuole has very low pH 4.5 [28], compared with cytoplasmic pH~8. HEK-293 cells did not tolerate such a large gradient, but with a moderate inward gradient (ΔpH 2.0), we observed inward H + current ( Fig 1D). Inward currents were detected in ten cells with 1.0 to 3.0 U gradients. Extrapolated to in vivo conditions, LpH V 1 should conduct inward current when activated.
LpH V 1 is proton selective. Measurement of the reversal potential (V rev ) of the current in cells expressing LpH V 1 over a wide range of pH o and pH i confirmed that these currents were proton selective, because V rev was close to the Nernst potential for H + (E H ) (dashed green line in Fig 2). When tetramethylammonium + was replaced by other small cations V rev did not change, confirming H + selectivity. The mean change in V rev when TMA + in the bath was exchanged with Na + or K + , respectively, was 0.4 ± 0.4 mV (mean ± SEM, n = 4) or 1.2 ± 1.8 mV (n = 4), after correction for the measured liquid junction potentials. The expressed gene product is clearly a highly proton-selective voltage-gated channel, so we named it LpH V 1, and explored its similarities and differences from H V 1 in other species.
LpH V 1 exhibits ΔpH dependent gating. In all species studied so far, H V 1 exhibits ΔpH dependent gating, in which the voltage range of channel opening is defined by the pH gradient, ΔpH (pH o -pH i ). The position of the proton conductance-voltage, g H -V, relationship generally shifts 40 mV/unit change in either pH o or pH i (positively for increasing pH i and negatively for increasing pH o ) [13,18,19]. Fig 3 shows g H -V relationships in a cell studied in whole-cell configuration with changes in pH o (A) and in an excised, inside-out patch of membrane in which pH i was changed (B). It is evident from Fig 3A that increasing pH o shifts the voltage dependence of LpH V 1 opening negatively, as observed in H V 1 from all species studied to date. It is further evident that there was little shift between pH o 8 and pH o 9, indicating that this effect saturates at high pH o , as reported recently in human H V 1, hH V 1, kH V 1, and EhH V 1 [37]. The data in Fig 3B show that changing pH i similarly shifts the g H -V relationship, with decreasing pH i shifting the curve negatively, as occurs in all other H V 1. The effects of pH o and pH i on the position of the g H -V relationship were assessed quantitatively by plotting the voltage at which the g H was 10% of its maximal value, g H,max , as a function of pH o or pH i (Fig 4). This parameter was chosen because it is clearly defined, it does not require forcing data to fit a Boltzmann distribution, and it does not require estimating the threshold of activation, which has been used for this purpose previously by us and others, but is imprecise and arbitrary. These considerations are discussed at greater length elsewhere (see Methods of [37]). When pH o or pH i was varied, the g H -V relationship shifted by~40 mV/unit change in pH, matching the reference line showing this slope (Fig 4). For both pH o and pH i the shift tended to saturate above pH 8, similar to H V 1 from rat [19], human, the dinoflagellate Karlodinium veneficum, and the coccolithophore Emiliania huxleyi [37]. The saturation occurs above typical environmental pH, and likely reflects the approach of pH to the pK a of one or more pH sensing sites.
LpH V 1 channel opening kinetics. The rising current during a depolarizing pulse was fitted with a single exponential function to determine the time constant of activation, τ act . At symmetrical pH 8 (pH o = pH i = 8), τ act ranged 45-600 msec at +60 mV (n = 6); at symmetrical pH 7, τ act ranged 30-470 msec at +80 mV (n = 8). Examples of the dependence of τ act on pH are shown for whole-cell measurements ( Fig 5A) in which pH o was varied, and for inside-out patches of membrane ( Fig 5B) in which pH i was varied. In both configurations, changes in pH appeared to simply shift the τ act -V relationship along the voltage axis. As in all species, current activation (turn-on) became faster with greater depolarization (Fig 1A-1C). However, in LpH V 1 this property was markedly exaggerated. Activation became much faster (smaller τ act ) with larger depolarizing pulses; typically, τ act was 100 times faster at voltages 60 mV more positive.

LpH V 1 stains multiple membranes in intact cells
If LpH V 1 functions to allow protons across the vacuole into the scintillons as predicted, it should be localized in scintillon membranes. The other bioluminescence proteins LCF and LBP were previously demonstrated to localize to scintillons in L. polyedrum [24,27,30]. We immunostained PFA-fixed, methanol-dehydrated L. polyedrum cells with chicken anti-LCF, rabbit anti-LBP, or rabbit anti-LpH V 1. We visualized their localization with organism-specific secondary antibodies, each labeled with a different fluorophore. Western blotting of L. polyedrum whole cell lysates and purified recombinant LCF, LBP, and LpH V 1 probed with these primary antibodies detected proteins of the expected sizes and no cross-reactivity (data not shown). The confocal microscopy images in Fig 6 demonstrate that LCF and LBP are distributed in a punctate pattern as previously observed in dark-harvested L. polyedrum [25]. Fluorescence intensities of labeled cells were significantly different than both pre-serum and noprimary-antibody controls (see also S1 Fig for negative control images). As expected from previous studies [30,38], total fluorescence from antibody-labeled LCF and LBP decreases significantly in cells fixed at mid-light phase compared to mid-dark phase (data not shown).
The proposed function of H V 1 in bioluminescence requires localization in scintillon membranes (evaginations of the vacuole membrane); nonetheless, no cellular membrane could be excluded a priori as a potential site for H V 1. As seen in Fig 6, a significant fraction of the LpH V 1 localizes around the periphery of the organism, consistent with an additional plasma membrane location. The g H -V relationships calculated from LpH V 1 currents in whole cell measurements (A) or inside-out patches (B) are strongly affected by pH o or pH i , respectively. The proton conductance, g H , was usually calculated from the measured reversal potential, V rev , and the amplitude of a single rising exponential fitted to the current. In some cases, for example with test pulses near V rev , the amplitude of the tail current was used, after appropriate scaling. doi:10.1371/journal.pone.0171594.g003

LpH V 1 co-localizes with LCF and LBP in isolated scintillons
To further explore the cellular localization of H V 1, we isolated scintillons from L. polyedrum using density gradient centrifugation [30,39]. Scintillon isolation was confirmed initially by a bioluminescence activity assay [30]. Western blots of proteins extracted from isolated scintillons and probed with antibodies to LBP and LCF demonstrate the presence of these known scintillon markers and also show no detectable cross-reactivity of these antibodies with recombinant LpHv1 protein ( Fig 7B). As has been described previously, presumed proteolysis products for LBP and LCF are visible in these Western blots. As seen in Fig 7A, Western blotting of proteins extracted from isolated scintillon preparations shows that the antibody to LpH V 1 detects a scintillon protein with an apparent size of about 30 kDa; this antibody detects purified recombinant LpH V 1 at the expected size of 37 kDa. Many possibilities, including different post-translational processing in bacteria compared to dinoflagellates, could explain the apparent size difference between recombinant and native LpH V 1. Both in purified recombinant protein preparations and in preparations of scintillon protein, our antibody to LpH V 1 frequently detects protein bands at~60 and at~80 kDa, which are consistent with the size of a truncated (60 kDa) or full length (80 kDa) LpH V 1 dimer. LpH V 1 has a strongly predicted coiled-coil region in its C-terminus, so it likely dimerizes like H V 1 from several other species [40][41][42][43]. Our antibodies to scintillon proteins do not cross react ( Fig 7B). Western blotting from separate preparations consistently shows more LpH V 1 (~2 fold) in scintillons isolated during the day phase than the night phase (as in Fig 7A), although the difference is not statistically significant.
Scintillon isolation was further confirmed by confocal microscopy of fixed scintillon preparations, using native luciferin fluorescence [44] to positively identify scintillons [25] (Fig 7C). As previously described [25], we observed a low level of contaminating chlorophyll fluorescence. On the slide, chlorophyll fluorescence rarely overlapped with luciferin fluorescence, and many structures with luciferin fluorescence but no nearby chlorophyll fluorescence were visible. Immunostaining of fixed scintillon preparations confirms the presence of LBP and LCF in isolated scintillons ( Fig 7C). Immunostaining of fixed scintillon preparations show that LpH V 1 localizes to luciferin-containing structures that also contain LCF ( Fig 7C). That both luciferin and LCF are known markers of the scintillon structure is strong evidence that LpH V 1 localizes to scintillon membranes.

Mass spectrometry detects LpH V 1 peptides in isolated scintillon preparations
To further confirm the presence of LpH V 1 in scintillons, we subjected proteins from scintillon preparations to tandem mass spectrometry (MS/MS) analysis. We excised five prominent bands from an Imperial stained gel: one at 33 kDa, close to the size expected for LpH V 1, and others ranging from~25 to~50 kDa. We detected 17 different peptides from the 33 kDa band, and 2 different peptides from the 50 kDa band, each with more than one independent peptide spectrum, that met the probability threshold for matching the predicted sequence of LpH V 1. The sequences of these peptides are mapped onto the predicted sequence of LpH V 1 in Fig 8. Many distinct peptides from the excised bands met the probability threshold for scintillon marker proteins LBP and LCF. In addition, a large number of peptides with significant matches to known chloroplast proteins (e.g., ribulose-bisphosphate carboxylase, glyceraldehyde-3-phosphate dehydrogenase, peridinin-chlorophyll a-binding protein, etc.) were detected, consistent with the known chloroplast contamination of the scintillon preparation.

Zn 2+ inhibits LpH V 1 proton currents and the flash induced by acid in isolated scintillons
Few H V 1 inhibitors have been identified; the most potent in many species is Zn 2+ [45,46]. Fig 9 shows that Zn 2+ inhibited LpH V 1 currents detectably at 10 μM, and substantially at 100 μM. At similar concentrations, Zn 2+ inhibited luminescence elicited by acid exposure in scintillons isolated from L. polyedrum (Fig 9B and 9C). No flash was elicited when detergent was present, suggesting that intact scintillons were required for the response. These results strongly support the hypothesis that LpH V 1 is the proton channel that triggers the bioluminescent flash in L. polyedrum.

Discussion
In 1972, J. Woodland "Woody" Hastings predicted the existence of voltage-activated protonselective channels that trigger the bioluminescent flash in Gonyaulax polyedra (now Lingulodinium polyedrum) [1]. An action potential in the vacuole membrane was known to initiate the flash [34]. The scintillon contains luciferase, its substrate luciferin, and luciferin-binding protein. A drop in pH both activates luciferase [33] and releases luciferin from its binding protein [47]. Because the vacuole pH was 3.5-4.5 [28,48], a proton-permeable channel that was opened by depolarization during the action potential would allow rapid proton flux into the scintillon, releasing luciferin, activating luciferase, and triggering the flash [1]. The present results strongly support this hypothesis.
Luciferin, LBP, and LCF from several species cross react [26,49] supporting a common role of pH, and presumably H V 1, in the signal transduction pathway. We used RNA-seq data for L. polyedrum to identify a putative H V 1 gene. We confirmed that the organism expresses the RNA (S1 Table) and protein (Fig 7) predicted from this gene.

Structural comparison of LpH V 1 with other proton channels
LpH V 1 is the tenth H V 1 gene to be identified and confirmed by voltage-clamp studies in a heterologous expression system. Globally, LpH V 1 resembles all other H V 1 in having four transmembrane helical segments. Compared with hH V 1, it has a short N terminus,~33 amino acids vs.~100 in hH V 1. LpH V 1 has a very long S1-S2 linker of~100 residues vs.~10 in hH V 1. S1-S2 linkers over 70 residues long appear in two electrophysiologically confirmed H V 1 from coccolithophores, and five of the twelve high confidence dinoflagellate H V 1 shown in S3 Fig. These long linkers do not share significant similarity to each other, and BLAST searches of the linker region reveal little similarity to proteins of known function. A region of about 20 amino acids (351-373) in the LpH V 1 linker is predicted to be helical by several secondary structure prediction programs. LpH V 1 most likely exists as a dimer, because it has a strongly predicted coiled-coil region in its C-terminus. H V 1 from several other species are thought to dimerize, in large part enforced by C terminal coiled-coil interactions [40][41][42][43]50].
The sequence of LpH V 1 includes the amino acids thus far identified to play critical roles in the function of the molecule in other species [51]. LpH V 1 shares the "signature sequence" of all other H V 1 that includes an Asp in S1 and the motif RxWRxxR in S4 [12,14]. The Asp in the middle of the S1 segment is crucial to establishing proton selectivity [9,12,14,16], the Arg in S4 are thought to confer voltage sensing as in other VSDs, and the Trp affects multiple properties of H V 1 [37]. LpH V 1 has residues (Phe 171 , Leu 42 and Ile 199 ) that we propose form the hydrophobic gasket that has been identified in other voltage gated ion channels [52][53][54][55]. These three hydrophobic residues are aligned horizontally near the middle of the membrane, where they separate internally and externally accessible aqueous vestibules. The position occupied by Phe 171 , identified as a delimiter of the charge transfer center [56], is conserved almost universally among VSD-containing molecules.
A number of acidic amino acids are thought to stabilize the channel in closed, open, and intermediate states by electrostatic interactions with the cationic charges in S4. Likely The sequence and predicted secondary structure acid of LpH V 1 is shown (Johns S.J., TOPO2, Transmembrane protein display software, http://www.sacs.ucsf.edu/TOPO2/). Acidic residues in the transmembrane helices are shown in red, basic residues in dark blue, and aromatic residues in gray. Brown diamonds indicate the overlap of peptide sequences found by mass spectrometry analysis of isolated scintillons and the epitope to which the antibody against LpH V 1 was raised; otherwise peptide sequences are shown in orange squares and the epitope is shown in green stars.
doi:10.1371/journal.pone.0171594.g008 countercharges in the S1 helix in hH V 1 include Asp 112 , Glu 119 , Asp 123 , and possibly Lys 125 [51]. The corresponding positions in LpH V 1 are Asp 45 , Ser 52 , Glu 56 , and His 58 . When externally-accessible acidic groups are neutralized by mutation, the g H -V relationship shifts positively, because the open state is disfavored [51]. The somewhat more positive activation of LpH V 1 may reflect the absence of Glu 119 and Lys 125 . On the other hand, the other identified dinoflagellate H V 1, kH V 1 activates 60 mV more negatively than all other species, despite kH V 1 lacking Glu 119 (with Gly instead). The important countercharges in S2 and S3, are conserved in all identified H V 1 including LpH V 1, with its Glu 174 and Asp 195 . Another acid in S3 unique to H V 1 is Asp 185 (hHv1 numbering), which is absent in other VSDs [12,51]. In LpH V 1 this position is occupied by the conservative substitution Glu 206 . In summary, with the exception of one acidic residue in S1, the charges in the transmembrane region of LpH V 1 are quite similar to those in other H V 1. The electrophysiological properties of LpH V 1 are consistent with its proposed function in triggering the flash We measured several biophysical properties of the LpH V 1 channel to determine whether the LpH V 1 gene product has properties consistent with its proposed function of triggering the bioluminescent flash. These properties include 1) proton selectivity, 2) activation by depolarization, 3) opening kinetics comparable to that of the flash, and 4) the ability to conduct inward current (from vacuole to cytoplasm), all of which should occur in vivo. Our experiments were done at room temperature (20-25˚C) which is within the range of oceanic temperatures, but other possible differences between our experimental conditions and those in vivo cannot all be evaluated so easily. The mammalian cells used as an expression system (HEK-293 cells) may process proteins differently than dinoflagellates and likely have a different membrane composition than L. polyedrum. Salt concentrations appropriate for mammalian cells,~300 mOsm, are about half those of sea water. A critical factor for the H V 1 protein is the pH gradient. The cytoplasmic pH in dinoflagellates is estimated to be pH 8.0 [1], but the vacuolar pH in bioluminescent species L. polyedrum [28] and Noctiluca [48] is 4.5 and 3.5, respectively. HEK-293 cells did not survive such a large pH gradient. The membranes of mammalian cells tolerate neither extreme pH per se, nor extreme pH gradients. Our experiments address the posed questions with these constraints.
LpH V 1 meets the first two requirements: it is clearly highly proton selective (Fig 2) and it opens with depolarization (Fig 1). The next question is whether the kinetics of LpH V 1 match that of the bioluminescent flash. The flash recorded from individual L. polyedrum had a latency of 15-22 msec [57] and the flash in individual scintillons isolated from L. polyedrum or from entire organisms had a time-to-peak of~100-200 msec [25,58]. LpH V 1 opens faster than H V 1 of mammalian species (where τ act is measured in seconds), but slower than that of snail neurones with τ act of a few milliseconds [18]. As in all species, LpH V 1 channels open faster with increasing depolarization. Differing sharply from other species, the activation kinetics of LpH V 1 depends quite steeply on voltage. As evident in Fig 5, τ act was~1 s just above V threshold but became~100 times faster within 50-60 mV, changing e-fold in just~10 mV. In contrast, in several mammalian or amphibian cells, the voltage required to change τ act e-fold ranges 40-80 mV [3,[59][60][61][62][63]. As a result of the steep voltage dependence of τ act , a large depolarization could activate LpH V 1 within~10 msec. The kinetics of LpH V 1 thus seems consistent with the kinetics of the bioluminescent response. Density of LpH V 1 expression in the native membrane and the presence of other ion channels would also modulate the response.
We also addressed whether LpH V 1 could produce inward H + current during an action potential in situ. At symmetrical pH (e.g., pH o 7, pH i 7) inward current was not observed. With an inward pH gradient of 1-2 units (e.g., pH o 7, pH i 9), however, inward current was observed. An example is shown in Fig 1D. Once inward H + current is activated in the tonoplast, the resulting H + influx would further depolarize the membrane regeneratively, opening more channels and driving the membrane potential toward E H , which for pH 4.5//8.0 is in the vicinity of +200 mV. Correspondingly, action potential peaks of 200 mV have been recorded in situ in Noctiluca [21]. We were at first surprised that LpH V 1 activated relatively positively, compared with other H V 1 [20] and especially when compared with kH V 1 that activates well negative to E H [12]. Perhaps, given the enormous inward H + gradient across the L. polyedrum tonoplast, it would be perilous for the cell to allow LpH V 1 activation except when triggering a flash. K. veneficum is not bioluminescent, and the function of kH V 1 is uncertain, but is likely different from that in bioluminescent species.

The cellular localization of LpH V 1 is consistent with its proposed role in bioluminescence
Three lines of evidence demonstrate convincingly that LpH V 1 is expressed in the vacuole membrane surrounding the scintillons, confirming its predicted role in bioluminescence: 1) Western blots of isolated scintillons show antibody staining of a protein with the predicted size and no detectable cross-reactivity with LCF or LBP (Fig 7); 2) immunostaining of isolated scintillons with the same antibodies shows that LpH V 1 colocalizes with LCF, a marker of the scintillon organelles (Fig 7); and 3) proteomics analysis of proteins extracted from isolated scintillons shows the expected presence of LBP and LCF proteins (S2 Table) and the presence of LpH V 1 protein (Fig 8) in these structures. Further confirmation that LpH V 1 is present in scintillons is the inhibition of the flash in isolated scintillons by Zn 2+ at concentrations that inhibit LpH V 1 proton currents (Fig 9).
Upon stimulation by shear stress, mechanosensor (probably stretch activated) channels at the surface of L. polyedrum [58,64] are thought to relay a signal through intracellular calcium signaling via G-proteins [65,66], but the molecular identities of the signaling components that result in the action potential at the vacuole membrane and subsequent luminescence are unknown. It is likely that LpH V 1 mediates the action potential in the vacuole membrane that triggers the flash, because H + is the only ion with a sufficiently positive Nernst potential to generate an action potential that peaks at +200 mV [21]. Voltage gated Na + channels have been reported in the outer membrane of Noctiluca, but their reversal potential in sea water is only +33 mV [67]. Although Hastings' original proposal required a proton channel to open only during the action potential in scintillon membranes, it was later realized that the same channel could also mediate the action potential in the tonoplast [28,68]. We have identified additional putative H V 1 sequences in RNA-seq data from two other bioluminescent dinoflagellates, Alexandrium tamarense and Noctiluca scintillans, excellent evidence that H V 1, like LCF and LBP, is a conserved component of the signal transduction pathway that leads from shear stress at the organism surface to the light flash.
Immunostaining of whole cells suggests that membranes other than those surrounding the scintillons may contain LpH V 1 (Fig 6). The non-bioluminescent dinoflagellate K. veneficum expresses a H V 1 in feeding populations at night [12]; we have identified putative H V 1 genes in several additional dinoflagellates (S3 Fig) based on the presence of an Asp in the middle of S1 crucial to proton selectivity [9,12,14] and a signature sequence in S4 involved in gating (10). In other organisms H V 1 functions in many processes [20], raising the likelihood that in L. polyedrum LpH V 1 also serves purposes in addition to the control of bioluminescence. The dinoflagellates (bioluminescent or not) in which we found H V 1 gene sequences span a large fraction of the dinoflagellate phylogenetic tree [69,70]. Taken together, these data suggest that primordial H V 1 functions have been co-opted by the bioluminescent species for light production. Intriguingly, antibody to LCF stained trichocysts in L. polyedrum [28], a tantalizing hint of possible non-bioluminescent functions for the bioluminescence enzyme.

Sequence searching and alignments
The K. veneficum voltage-gated proton channel (kH V 1, NCBI accession JN255155) was used as a BLAST probe with e-value cutoff of 10 −2 against the Marine Microbial Eukaryote Transcriptome Sequence Project (MMETSP; [36]) which yielded a single contig (26874) in MMETSP1032. BLAST searches of the non-redundant database at NCBI using the translation of this contig yielded H V 1 from diatoms and coccolithophores with e-values of 10 −11 . Using various H V 1 probes, searches of MMETSP also yielded full and partial sequences for putative H V 1s from Akashiwa sanguinea, Alexandrium monilatum, Alexandrium tamarense, Amphidinium carterae, Azadinium spinosum, Karenia brevis, Noctiluca scintillans, Symbiodinium, and Scrippsiella trocoida. We included a recently described putative H V 1 from Prorocentrum minimum [71]. These sequences were aligned with a set of high confidence animal H V 1s using MSAprobs [72] (S3 Fig). The predicted secondary structure for LpH V 1 was drawn using TOPO2 (Johns S.J., TOPO2, Transmembrane protein display software, http://www.sacs.ucsf. edu/TOPO2/).

Gene and antibody synthesis
Genes for LpH V 1 (based on longest open reading frame of RNA-seq contig 19215 from the L. polyedrum library) and LCF (NCBI accession AF085332.1), codon-optimized for mammalian (LpH V 1) or E. coli (LCF) expression, were synthesized by Genscript Corp. and subcloned into pcDNA and pEGFP (LpHv1) or pQE-30 (LCF). Peptides corresponding to chosen epitopes from LpH V 1 (CDAGRQLSSDGDQ) and LCF (CLDYPKKRDGWLEKN) were synthesized by Genscript and polyclonal antibodies to these peptides were raised and affinity purified from rabbit (LpH V 1, final protein concentration 1.1 mg/ml) or chicken (LCF, final protein concentration 0.88 mg/ml). Preserum from animals was also provided by Genscript. The native gene for LBP in pGEX-4T, and rabbit antibody raised to LBP, were the generous gifts of Dr. David Morse (University of Montreal).
Electrophysiology HEK-293 cells were grown to~80% confluence in 35 mm cultures dishes. Cells were transfected with 0.4-0.5 μg of cDNA using Lipofectamine 2000 (Invitrogen) or PEI (polyethylenimine, Sigma). After 6 h at 37˚C in 5% CO 2 , cells were trypsinized and re-plated onto glass cover slips at low density for patch clamp recording the following day. We selected green cells under fluorescence for recording. Whole-cell or excised inside-out patch configurations of the patch-clamp technique were carried out as described in detail previously [73]. Bath and pipette solutions were used interchangeably. They contained (in mM) 2 MgCl 2 , 1 EGTA, 80-100 buffer, 75-120 TMA + CH 3 SO 3 -(to bring the osmolality to~300 mOsm), titrated using TMAOH. Buffers with pK a near the desired pH were used: Homopipes for pH 4.5-5.0, MES for pH 5.5-6.0, BisTris for pH 6.5, BES for pH 7.0, HEPES for pH 7.5, Tricine for pH 8.0, CHES for pH 9.0, and CAPS for pH 10. Experiments were done at 21˚C or at room temperature (20-25˚C). Current records are shown without leak correction. Reversal potentials (V rev ) in most cases were determined from the direction and amplitude of tail current relaxation over a range of voltages, following a prepulse that activated the proton conductance, g H . When the g H activated negative to V rev the latter could be determined directly from families of currents. Currents were fitted with a single exponential to obtain the activation time constant (τ act ) and the fitted curve was extrapolated to infinite time to obtain the "steady-state" current amplitude, from which the g H was calculated. The voltage at which g H was 10% of g H,max (V gH,max/10 ) was determined after defining g H,max as the largest g H measured.
Cell culture L. polyedrum (CCMP 1932, obtained from National Center for Marine Algae and Microbiota-Bigelow Laboratory for Ocean Sciences) cultures were grown in three locations with minor differences in conditions. Cultures were grown in L1 minus Si [74] or F/2 minus Si [75] medium prepared in artificial seawater (Instant Ocean, Blacksburg) and maintained in 12:12 or 14:10 light:dark cycle (photon flux 100 μmoles/m 2 /s) at 18-20˚C. Cultures were allowed to grow until reaching a cell density of 4,000-10,000 cells/ml at which point they were collected at middark and mid-day time points.

RNA extraction, qPCR analysis, and cDNA cloning
Primer sets covering the entire predicted LpHv1 coding sequence (S1 Table) were used for qPCR analysis performed on cells harvested at mid-light or mid-dark. Cells from 50 ml aliquots of culture were collected by centrifugation. The pellet was used immediately or flash frozen in an ethanol and dry ice bath and stored at -80˚C. Samples from each timepoint were resuspended and Dounce homogenized in 1 ml TRI reagent (Zymoresearch), and RNA was extracted according to the manufacturer's protocol. RNA was reverse transcribed using Superscript II reverse transcriptase (Invitrogen by Life Technologies) with random primers (Invitrogen) according to the manufacturer's protocol. Quantitative real-time PCR was performed in triplicate using an Applied Biosystems (Life Technologies) Fast 7500 thermal cycler with primers listed in S1 Table. Thermal cycling conditions consisted of an initial denaturation at 95˚C for 2 minutes followed by 40 cycles of denaturation at 95˚C for 15 seconds, annealing and fluorescent data collection at 60˚C for 15 seconds, and extension at 72˚C for 30 seconds. Cycle thresholds and baselines were determined manually and cycle thresholds were averaged and compared across time points.

Recombinant protein expression and purification
Recombinant LBP and LpH V 1 proteins tagged with glutathione s-transferase (GST) were expressed in E. coli by induction with isopropyl b-d-1 thiopyranogalactoside (IPTG) at 23˚C (LBP) or 17˚C (LpH V 1) for 8-24 hours. Recombinant LCF protein tagged with 6-His was expressed in E. coli by induction with IPTG at 17 C for 16-24 hours. Proteins were purified from E. coli according to the manufacturer's instructions (GE-Healthcare for LBP and LpH V 1, Qiagen for LCF). GST tag was cleaved from LpH V 1 by digestion with Prescission protease according to the instructions of the vendor. In one experiment, GST tag was cleaved from LBP by digestion with thrombin for 16 hours; the insoluble LBP precipitate was pelleted by centrifugation and solubilized in SDS-PAGE loading buffer.

Scintillon isolation, luminescence assays, and gel analysis
Scintillons were isolated from cultures grown in two different locations and isolated using a sucrose [39] or a Percoll [30] density gradient. For luminescence assays to test metal sensitivity, the 0.5 ml fraction from a sucrose gradient with the highest luminescence was kept separate and considered to comprise "pure scintillons". This fraction was pelleted, washed with buffer, resuspended and then diluted to about 1 ml in the extraction buffer. 100 μl of the diluted scintillon preparation was added to wells of a 96 well plate. 50 μl of 5x metal ion/drug solution was added to the same well. The reaction was started by injecting 150 μl of 0.05 M acetic acid and luminescence was measured immediately using pClamp software (Molecular Devices). No flash was detected in the presence of detergents (0.03-0.1% TWEEN-20 or 0.1% SDS).
For gel analysis, total scintillon protein was extracted by heating for 10 min at 95˚C in SDS-PAGE sample buffer. Total scintillon proteins, and preparations of purified recombinant proteins, were separated by SDS-PAGE, Western blotted onto PVDF, blocked with 5% milk or Licor blocking buffer (Licor) and probed with diluted antibodies to LCF (1:1000), LBP (1:10000), and LpH V 1 (1:3000). Blots were visualized with secondary antibodies to appropriate animal IgG conjugated either to HRP or to Licor fluorescent tags. Western blotting results from different scintillon sources were directly comparable.

Fixation and immunostaining of cells and isolated scintillons
We used the procedure of [76] with modifications. Preliminary experiments indicated that low speed centrifugation was optimum to preserve cell integrity so all centrifugation steps were performed at 300 x g. Cells were pelleted from 50-100 ml of culture by centrifugation, washed with seawater, and fixed in 4% Paraformaldehyde in seawater at room temperature. Fixed cells were washed with phosphate buffered saline (PBS), permeabilized in 100% methanol at 4˚C, washed at room temperature with PBS supplemented with 0.1% tween and 1% BSA (PBST-BSA), then blocked by incubation in PBST-BSA at room temperature. Primary antibodies, or serum controls diluted to the same protein concentrations, or vehicle controls, were added and incubated overnight at 4˚C. Cells were collected by centrifugation and washed at room temperature with PBST-BSA. For detection, Alexa555-Fluor-conjugated goat anti-rabbit IgG (H+L) and/or Alexa555-Fluor-conjugated goat anti-chicken IgG (ThermoFisher, Waltham MA) was added and incubated in the dark at room temperature. Cells were washed with PBST-BSA at room temperature, incubated with 4',6-diamidino-2-phenylindole (DAPI) 5μg/ mL in PBS, and washed again. Cells were incubated in 2 drops of Vectashield overnight at 4˚C in the dark and 15 to 50 μL were mounted on slides. Fixation and immunostaining of isolated scintillons were performed as described, but without the methanol permeabilization step. Cells and scintillons were visualized using a Zeiss LSM 700 confocal microscope, equipped with a 20x (whole cells) and 40x (scintillons) 1.2 NA C-Apochromat objective. For immunofluorescent localization, all channel pinholes were set to 1 Airy Unit. Isolated scintillons identified by luciferin fluorescence were scored for the presence or absence of secondary antibody fluorescence. Confocal slices or maximum intensity projections of the Z-stack were rendered using Zeiss Zen software, and processed using Adobe Photoshop.

Preparation of scintillon proteins for mass spectrometry
Purified scintillon preparations were concentrated by ultra-centrifugation at 4˚C. The pellet was re-suspended in SDS sample buffer, heated for 5 minutes at 95˚C, and cleared by centrifugation at 10,000 × g. 50 μl of sample was loaded and run on NovexNuPAGE 4-12% bis-tris gels according to the manufacturer's protocol. Gel bands corresponding to the location of the presumptive proteins were excised with a clean scalpel. Samples were processed using the Thermo Scientific In-Gel Tryptic Digestion Kit according to manufacturer's protocol. Gel bands were destained twice with 200 μl destaining solution (~25 mM sodium bicarbonate in 50% acetonitrile) and incubated at 37˚C with shaking for 30 minutes. Samples were reduced by incubation at 60˚C for 10 min in 50mM TCEP (tris(2-carboxyethyl)phosphine) in 25 mM ammonium bicarbonate buffer. Free sulfhydryl groups were alkylated by incubation in 100 mM iodoacetamide at room temperature for 1 hour in the dark. Gel pieces were shrunk in acetonitrile. For the initial proteomics run samples were treated overnight with 100 ng trypsin at 30˚C. The second proteomics run samples were treated with 100ng typsin and digestion was performed at 50˚C at high pressure using the PBI Barocyler (Pressure Biocsciences Inc.) according to manufacturer's protocol. Samples were dried in a SpeedVac.

Mass spectrometry analysis and data processing
Scintillon samples were analyzed by electrospray ionization on an Elite tandem orbitrap mass spectrometer (Thermo Scientific Inc). Nanoflow HPLC was performed by using a Waters NanoAcquity HPLC system (Waters Corporation). Peptides were trapped on a fused-silica pre-column (100 μm i.d. 365 μm o.d.) packed with 2 cm of 5 μm (200 Å) Magic C18 reversephase particles (Michrom Bioresources, Inc). Subsequent peptide separation was conducted on a 75 μm i.d. x 180 mm long analytical column constructed in-house and packed with 5 μm (100 Å) Magic C18 particles, using a Sutter Instruments P-2000 CO 2 laser puller (Sutter Instrument Company). The mobile phase A was 0.1% formic acid in water and mobile phase B was 0.1% formic acid in acetonitrile. Peptide separation was performed at 250 nL/min in a 95 min run, in which mobile phase B started at 5%, increased to 35% at 60min, 80% at 65min, followed by a 5 min wash at 80% and a 25 min re-equilibration at 5%. Ion source conditions were optimized by using the tuning and calibration solution recommended by the instrument provider. Data were acquired by using Xcalibur (version 2.8, Thermo Scientific Inc.). MS data was collected by top-15 data-dependent acquisition. Full MS scan of range 350-2000 m/z was performed with 60K resolution in the orbitrap followed by collision induced dissociation (CID) fragmentation of precursors in iontrap at normalized collision energy of 35. The MS/MS spectra of product ions were collected in rapid scan mode.
Acquired tandem mass spectra were searched for sequence matches against UniprotKB database using COMET. The following modifications were set as search parameters: peptide mass tolerance at 10 ppm, trypsin digestion cleavage after K or R (except when followed by P), one allowed missed cleavage site, carboxymethylated cysteines (static modification), and oxidized methionines (variable modification/differential search option). PeptideProphet and Pro-teinProphet, which compute a probability likelihood of each identification being correct, were used for statistical analysis of search results. PeptideProphet probability ! 0.9 and ProteinProphet probability ! 0.95 were used for positive identification at an error rate of less than 1%. Only proteins identified by more than one unique peptide sequence were included in the study.
Supporting information S1 Table. Primers and predicted product sizes for qPCR. (DOCX) S2 Table. Assignments of peptides via MS/MS analysis of protein from isolated scintillons of Lingulodinium polyedrum. Tabs show assignments of proteins from two independent scintillon preparations. Samples shown in 'MS results 1' tab were initially treated with 100 ng trypsin at 30˚C; those in 'MS results 2' tab were initially treated with 100 ng trypsin at 50˚C at high pressure using the PBI Barocyler (Pressure Biocsciences Inc.) according to manufacturer's protocol. (XLSX) S3 Table. Sequence of longest open reading frame of contig 26784 in MMETSP1032 assembly, and sequence of codon-optimized gene of LpH V 1. Sequences were found by BLAST searches of RNA-seq projects [35,36] and were aligned with MSA-Probs [72]. Dinoflagellate sequences are shown aligned with sequence logos of individual transmembrane helices obtained from an alignment of animal H V 1s [12]. The sequence logos are numbered for hH V 1. Overlapping partial sequences from Alexandrium monilatum are not shown. (TIF)