Calfacilitin is a calcium channel modulator essential for initiation of neural plate development

Calcium fluxes have been implicated in the specification of the vertebrate embryonic nervous system for some time, but how these fluxes are regulated and how they relate to the rest of the neural induction cascade is unknown. Here we describe Calfacilitin, a transmembrane calcium channel facilitator that increases calcium flux by generating a larger window current and slowing inactivation of the L-type CaV1.2 channel. Calfacilitin binds to this channel and is co-expressed with it in the embryo. Regulation of intracellular calcium by Calfacilitin is required for expression of the neural plate specifiers Geminin and Sox2 and for neural plate formation. Loss-of-function of Calfacilitin can be rescued by ionomycin, which increases intracellular calcium. Our results elucidate the role of calcium fluxes in early neural development and uncover a new factor in the modulation of calcium signalling.

A graft of a special region of the embryo, known as Spemann's organizer [1][2][3] (Hensen's node in amniotes [4][5][6][7][8][9] can trigger the entire process of neural induction in all vertebrate classes, generating a fully patterned central nervous system. For a long time it was thought that a single signal, emitted from the organizer, might account for its neural inducing activity. Consistent with this, a large body of research, mainly in amphibian embryos, suggested that BMP inhibition is a sufficient signal to trigger the entire process of neural induction 1,10-14 . However, development of the neural plate takes a relatively long time; for example, in the chick embryo expression of the earliest definitive neural plate marker, Sox2, requires about 12 h exposure of competent ectoderm to neural inducing signals from a grafted organizer and a morphologically recognizable neural plate only appears several hours later [15][16][17][18] . Also, experiments in a variety of systems have suggested that signals other than BMP inhibition are required for neural induction, including fibroblast growth factor (FGF) [19][20][21][22][23][24] , Wnt inhibition 25,26 and calcium/ protein kinase-C signalling [27][28][29][30][31][32][33][34][35][36][37] . However, how these signals are integrated, from which tissues they emanate and in which order, is not yet understood. FGF signalling is partly integrated with BMP signalling through MAP kinase-dependent phosphorylation of the linker region of the BMP effector Smad1 38 , but FGF inhibition experiments 19,20 suggest that FGF signals are also required independently of BMP inhibition.
Timed organizer transplantation and removal experiments suggested that a minimum of 5 h signalling from a grafted Hensen's node is required for the responding epiblast to become sensitive to BMP inhibition [39][40][41][42] . To uncover the events that occur during these initial 5 h of neural induction (upstream of BMP inhibition), we conducted a differential screen between chick epiblast cells that had or had not been exposed for 5 h to neural inducing signals from a graft of the organizer, Hensen's node 21,[43][44][45] . Twelve differentially expressed genes were identified, encoding proteins involved in transcriptional regulation (ERNI 17,21 , Churchill 43 , Sox3, Otx2 46 ), known and putative receptors (TrkC and Asterix 45 ), a putative RNA-binding protein (Obelix 45 ), the retinoid regulator Cyp26A1 46 and proteins with pro-and anti-apoptotic functions (Dad1, Fth, HCF 44 ). These genes are expressed at characteristic times following a graft of the organizer, suggesting that the organizer induces a hierarchical succession of states ('epochs') 45 . In addition to the above genes, one further gene with differential expression, initially designated C3, was identified in the screen. It encodes a protein of unknown function and has not yet been investigated. This is the subject of the present study.
Here we show that C3 encodes a multi-pass transmembrane protein, which facilitates calcium signalling through Ca V 1.2 L-type channels and therefore named it Calfacilitin. We further show that Calfacilitin is required for neural induction (a requirement that can be bypassed by forcing calcium entry into the responding tissue). Calfacilitin is induced by FGF and is required for induction of Geminin and Sox2, two key effectors of neural induction. Our findings uncover a new regulator of calcium signalling as well as a new link between calcium and other signals during neural induction.

Isolation of Calfacilitin.
In situ hybridization reveals that expression of C3 mRNA starts weakly in the epiblast before gastrulation (Fig. 1a), then increases in the future neural plate and thereafter remains expressed throughout the central nervous system (Fig. 1b- Fig. 1g). This is mimicked by FGF8 (19/26 in 3 h, 13/13 in 5 h; Fig. 1h) but not by other secreted proteins including BMP antagonists (Chordin, Noggin), Cerberus, Dkk1 or Crescent (not shown). Its normal expression and the dynamics of its regulation by organizer signals and by FGF therefore suggest a possible early role for the C3 product in the cascade of events leading to neural plate formation.
Sequence analysis predicts a protein of 252 aminoacids with several putative transmembrane regions (Fig. 2a), containing a TLC ('TRAM, LAG1, CLN8') domain, present in many proteins with important roles including lipid synthesis, transport and sensing [49][50][51][52] . This appears to be highly conserved across the vertebrates ( Supplementary Fig. S1). To determine the topology of the protein within the membrane, we performed epitope  ARTICLE mapping experiments. A myc epitope was introduced into seven different locations between the predicted transmembrane domains, each construct introduced into either COS or HEK293T cells, which were then fixed and stained with anti-Myc either in the presence of detergent (to detect both intracellular and extracellular epitopes) or without detergent (for extracellular epitopes) (Fig. 3). The results suggest a 6-pass transmembrane topology with both the N-and C-termini outside the cell (Fig. 2b). This topology is reminiscent of some ion channels, including Ca 2 þ channels 53 , which have been implicated in early neural development in Xenopus [29][30][31]33,34,36,37,54,55 .
Calfacilitin is a facilitator of Ca V 1.2 channels. These observations raised the question of whether the C3 protein functions as an ion channel. We transfected the cDNA into HEK293T cells but detected no currents in whole-cell patch clamp electrophysiological recordings. However, when co-transfected with L-type Ca V 1.2 calcium channels, I-V relationships were shifted towards a negative potential ( Fig. 2c; Po0.025, Student's t-test) and the steady-state inactivation towards a positive potential ( Fig. 2d; Po0.02, Student's t-test), therefore evoking a larger window current. Furthermore, C3 slows inactivation in the presence of either Ba 2 þ or Ca 2 þ as the charge carrier ( Fig. 2e,f). To test its specificity for Ca V 1.2 channels, we examined its effects on another major L-type channel, Ca V 1.3, and found no significant effect ( Supplementary Fig. S2). To determine whether these channels are expressed appropriately in the embryo, we performed in situ hybridization: Ca V 1.2 mRNA is expressed in the prospective neural plate with a pattern and dynamics of expression very similar to those of the novel protein ( Fig. 4a- Calfacilitin-myc2 Calfacilitin-myc3 Calfacilitin-myc4 Calfacilitin-myc5 Calfacilitin-myc6 Calfacilitin-myc7 ( Fig. 2i). Together, these data suggest that although the novel protein is not itself an ion channel, it enhances calcium influx into the cell through the L-type calcium-channel Ca V 1.2, to which it can bind within the cell membrane and with which it is co-expressed in the embryo. We therefore named this protein Calfacilitin.
The abnormal development of the host could be due to nonspecific toxicity, but Sox2 induction by the node can be rescued in nicardipine treated embryos with beads soaked in the calcium ionophore ionomycin, which drives Ca 2 þ ions into the cell independent of any channels (16/19 expressing; Fig. 5d, arrow). Dramatically, induction can be rescued even when the host neural plate is severely affected (Fig. 5d). These results demonstrate that induction of the neural plate by the organizer requires Ca 2 þ influx, normally through nicardipine-sensitive, L-type calcium channels.
To test this further we used various Ca 2 þ indicators that fluoresce in proportion to the free intracellular Ca 2 þ concentration (Rhod-2, Fura2 or Fura-Red with Fluo4). Whole embryos present particular challenges to Ca2 þ -imaging, which are further complicated by combining in experiments using fluorescein-labelled morpholinos and GFP-labelled reporter constructs (see below). The main problem is imaging large areas of the embryo with enough sensitivity and discrimination of Ca 2 þ levels over a long period of time (hours) without loss of dye from the cells. We explored Fura2 (ratio-imaged with IR excitation in a multi-photon microscope or with UV excitation by standard fluorescence microscopy), the ratio between Fura-Red þ Fluo4 in a confocal or conventional fluorescence microscope, and Rhod-2 (conventional fluorescence). The latter was most stable and allowed combining with FITC-labelled morpholino or GFP-labelled constructs. Although Rhod-2 tends to accumulate in mitochondria, the emitted signal is still sensitive to cytosolic Ca2 þ levels; for example, this can be seen in ionomycin control experiments (see below). With the microscopy methods used, in whole embryos, we were unable to detect fast Ca 2 þ transients either in the normal embryo or in experimentally manipulated regions. However, we were able to detect slow changes over several hours.
Organizer grafts into embryos loaded with these dyes cause an increase in the Ca 2 þ signal, starting after about 3-4 h in the adjacent host epiblast (Fig. 6a-e). Interestingly, this is also the time required for the node to induce expression of Calfacilitin in host epiblast. We also tested whether Calfacilitin is sufficient to cause a rise in intracellular Ca 2 þ in vivo by electroporating dyeloaded embryos with an expression plasmid driving either Calfacilitin-IRES-GFP or just GFP (Calfacilitin-IRES-RFP or DS-Red in some experiments). Calfacilitin-transfected cells display elevated Ca 2 þ as compared with neighbouring nontransfected cells or to control GFP-or DS-Red-transfected cells (Fig. 6f-m). Therefore, both misexpression of Calfacilitin and neural induction by the organizer are accompanied by a rise in intracellular Ca 2 þ .
Calfacilitin and Ca 2 þ are not sufficient for neural induction.
To determine whether Calfacilitin is sufficient to induce Sox2 in competent epiblast, we electroporated Calfacilitin-IRES-GFP as a line extending from the prospective neural plate to the lateral, non-neural epiblast. This construct (26/32; Fig. 5e,f; arrow), but not GFP alone (0/11; Fig. 5g,h), expands Sox2 expression into the embryonic non-neural epiblast. When these embryos are examined by in situ hybridization for the neural plate border marker Dlx5, this border is found shifted towards the non-neural ectoderm (Calfacilitin: 6/10 shifted, Fig. 5i,j, arrow; Control: 0/6 shifted, Fig. 5k,l)  ARTICLE beads induce Sox2 in the more peripheral area opaca epiblast (not shown). These effects resemble the activity of BMP antagonists, which can only expand the neural plate and its border into the adjacent non-neural ectoderm when misexpressed along a continuous line of cells extending from the neural plate 40,42,56 . These observations suggest that neither Calfacilitin nor an increase in intracellular Ca 2 þ is sufficient for neural induction, which involves additional signals.
Calfacilitin is required for neural induction. To establish whether Calfacilitin is required for neural plate development, we electroporated morpholino oligonucleotides (MO) targeting internal splice sites of Calfacilitin into the prospective neural plate of early primitive streak stage embryos. When electroporated into the embryo, this causes exon skipping, generating a truncated Calfacilitin protein (Supplementary Fig. S4). First, we measured the effect of the knockdown on intracellular Ca 2 þ levels in the prospective neural plate. Control-MO electoporated cells showed no significant difference to neighbouring non-electroporated tissue (ratio: 1.02; two-tailed t-test P ¼ 0.25; n ¼ 12), whereas Calfacilitin-MO significantly lowers Ca 2 þ with respect to neighbouring cells (1.6-fold reduction; two-tailed t-test P ¼ 2.32 Â 10 À 5 , n ¼ 12; in these experiments, rectangles of 224 Â 170 pixels in the electroporated area were measured). Next, we determined whether this treatment affects expression of the neural plate marker Sox2. Indeed it does (17/22 affected; Fig. 5m,n), unlike control-MO (0/9; Fig. 5o,p). To test the fate of the cells that had lost Sox2 expression we examined the border marker Dlx5: this was found to shift into the neural plate, expanding the non-neural ectoderm territory (Calfacilitin-MO: 5/6 shifted, Fig. 5q,r ; Control-MO: 0/ 6, Fig. 5s,t). We also tested whether neural induction by a grafted organizer requires Calfacilitin: we electroporated MOs into competent epiblast and then grafted a node onto the electroporated site. Calfacilitin-MO completely blocked Sox2 induction in 21/31 embryos and substantially reduced it in a further 6/31 (Fig. 7a,b), as compared with control-MO (22/26 with normal expression; Fig. 7c,d). Calfacilitin-MO did not affect the expression of earlier pre-neural markers (Sox3 and ERNI; 19/22; Fig. 7e,f). These results show that Calfacilitin is required for neural induction, acting downstream of Sox3 and ERNI but upstream of Sox2.
As a further test of the specificity of the Calfacilitin-MO we tested whether the phenotype can be rescued with Calfacilitin cDNA (lacking the targeted splice site). This is the case (36/42 embryos with rescued Sox2 expression; Fig. 7g,h). As a more dramatic test of the importance of Calfacilitin in relation to its calcium-regulating functions, we explored whether the effect of the MO can be rescued by ionomycin. When a node is grafted onto Calfacilitin-MO-transfected epiblast along with ionomycin-soaked beads, Sox2 expression is restored (14/21; Fig. 7i,j). This is not due to induction of Sox2 by ionomycin, as no ectopic expression is seen when ionomycin beads are grafted alone (see above). This predicts that the MO lowers Ca 2 þ in cells adjacent to the grafted node; to test this, we imaged Ca 2 þ using Rhod-2. The MO prevented the increase in Ca 2 þ caused by node grafts (see above and with respect to neighbouring regions of the host within 3-4 h, with continued decline over at least 6 h (Fig. 8). Together, these results strongly suggest that Calfacilitin is necessary for neural induction, through its role in Ca 2 þ signalling.
Calfacilitin regulates Geminin expression. Calcium channels have been proposed to be required for the expression of Geminin in Xenopus 54 . As Geminin was recently implicated in the acquisition of neural fate and initiation of Sox2 expression in chick and mouse 17 , we tested whether Calfacilitin can induce Geminin. Although Calfacilitin alone is not sufficient to induce Geminin in competent area opaca epiblast (not shown), Calfacilitin-MO (3/20 with expression; Fig. 7k,l), but not control-MO (8/8; Fig. 7m,n), inhibits upregulation of Geminin by a node graft. This effect can be rescued by Calfacilitin cDNA lacking splice sites (39/43; Fig. 7o,p) as well as by ionomycin beads (7/10; Fig. 7q,r). Together, these results suggest that induction of Geminin is downstream of Calfacilitin.

Discussion
Our results uncover Calfacilitin as a previously undiscovered regulator of Ca 2 þ signalling and as a new factor in neural induction. It is required early, downstream of initial FGF signals and defines an essential step for the later acquisition of Sox2 expression and neural plate formation. We demonstrate that its action in these processes is through modulation of intracellular Ca 2 þ levels. As FGF induces expression of both Geminin 17 and Calfacilitin, this could explain the finding that FGF can activate calcium channels in Xenopus ectoderm. A possible mechanism might involve arachidonic acid and TRPC channels 37 , similar to the lipid-regulating properties displayed by other members of the TLC family [49][50][51] . This could represent the long-sought mechanism for how L-type calcium channels are involved in neural induction.

Methods
Embryo manipulations and in situ hybridization. Brown Bovan Gold strain chick eggs (obtained from Henry Stewart & Co., UK) were used for most experiments. Quail eggs were obtained from Potter Poultry Farm, UK. All animal experiments conform to UK Home office regulations and were restricted to embryos during the first 2 days of incubation. Embryos were staged according to Hamburger and Hamilton 57 for primitive streak and later stages and following Eyal-Giladi and Kochav 58 for earlier stages. They were explanted and cultured by the method of New 59 but using a ring of square cross-section and the embryo grown in a Petri dish 60 . Node grafts 4 were performed using chick or quail donor nodes and chick host embryos. FGF8 was administered by soaking heparin-acrylic beads (Sigma) in FGF8 (R&D) diluted in Pannett-Compton saline, briefly washed in the same saline and implanted into the desired region of the host 17,21,43 . In situ hybridization was performed using DIG-labelled cRNA probes and detected using NBT-BCIP 61 .
Calcium imaging. To visualize Ca 2 þ levels, embryos were incubated in Rhod-2 (Invitrogen R-1245MP, 0.5 mg ml À 1 in 0.1% DMSO in Pannett-Compton saline), or a mixture of Fura-Red and Fluo4 (Invitrogen; each at 2 mg ml À 1 in 0.4% DMSO and 0.08% pluronic in Pannett-Compton saline) or Fura2 (Invitrogen; 2 mg ml À 1 in DMSO/Pluronic as above) at 37°C for 30 min, rinsed in saline and left to recover for 30 min-1 h in new culture before grafting a node and/or electroporation of either Control-MO or Calfacilitin-MO. Time-lapse imaging was performed (image frame size ¼ 1,344 pixels Â 1,024 pixels). Fura2 was imaged in a conventional fluorescence microscope as the ratio between emission at 510 nm generated by excitations at 340 and 380 nm or in a confocal microscope by excitation at 351 and 364 nm. Fluo4 and Fura-Red were excited at 488 nm and the ratio between green and red fluorescence generated was imaged by conventional microscopy. Rhod-2 was imaged by conventional fluorescence, using TRITC filters. For quantification in Rhod-2 experiments with electroporation of a morpholino or GFP construct, Ca 2 þ levels (red signal) were measured in fixed areas (a rectangle or circle) of electroporated and contralateral control tissue, measuring the total intensity of red pixels; Student's t-tests were used for analysis. In Ca 2 þimaging experiments, implantation of a AG1X2 bead soaked in 2 mM ionomycin (Sigma I9657) in DMSO caused an explosive increase in the calcium signal with all dyes-this was performed as a control in time-lapse calcium imaging experiments. For inhibition of calcium channels, embryos were incubated in nicardipine (Sigma N7510, 100 mg ml À 1 in 0.1% DMSO in Pannett-Compton saline) at 37°C for 30 min and rinsed in saline before incubation of the embryo. To drive Ca 2 þ into cells, AG1X2 beads were incubated in 2 mM ionomycin in DMSO, rinsed in Pannett-Compton and implanted into embryos.

Molecular characterization of Calfacilitin.
To study the topology of the transmembrane domains of Calfacilitin, a myc-tag was introduced after aminoacids 3, 42, 74, 107, 163, 201 and 252 of Calfacilitin to produce seven different tagged versions (with a myc epitope in the N-terminus, in one of the five loops between the predicted transmembrane domains or in the C-terminus of the protein) and cloned into pcDNA3.1. The constructs were transfected into COS cells using Lipofectamine-2000 40 or into HEK293T cells using PEI 62 and the cells obsserved by fluorescence microscopy.
Electrophysiology. The electrophysiological properties of Calfacilitin were studied by patch clamp techniques after transfection into HEK293 using calcium phosphate 64 . For IP experiments, HEK293 cells were transfected with Ca V 1.2 a1, b 2a -GFP, a 2 d-subunits, TAG and C3-myc using lipofectamine 2000 (Invitrogen). Control cells were transfected with an equal amount of pcDNA3 vector. 24 h later, the cells were harvested and lysed in 100 ml Hepes lysis buffer containing: 20 mM Hepes, 137 mM NaCl, 1% Triton X-100, 10% glycerol, 1.5 mM MgCl 2 and 1 mM EGTA. After incubation for 1 h at 4°C, 50 ml of protein lysate were taken as input and the remaining for IP. Eight microgram anti-Ca V 1.2 antibody (Alomone ACC-003) were incubated with the protein lysate and protein A beads (Roche) for 3 h. After washing three times with lysis buffer containing 0.1% Triton X-100, the protein was separated by SDS-PAGE. Calfacilitin-myc was detected with anti-myc antibody (Sigma C3956) and Ca V 1.2 protein by anti-Ca V 1.2 antibody.