Phosphate (Pi)-regulated heterodimerization of the high-affinity sodium-dependent Pi transporters PiT1/Slc20a1 and PiT2/Slc20a2 underlies extracellular Pi sensing independently of Pi uptake

Extracellular phosphate (Pi) can act as a signaling molecule that directly alters gene expression and cellular physiology. The ability of cells or organisms to detect changes in extracellular Pi levels implies the existence of a Pi-sensing mechanism that signals to the body or individual cell. However, unlike in prokaryotes, yeasts, and plants, the molecular players involved in Pi sensing in mammals remain unknown. In this study, we investigated the involvement of the high-affinity, sodium-dependent Pi transporters PiT1 and PiT2 in mediating Pi signaling in skeletal cells. We found that deletion of PiT1 or PiT2 blunted the Pi-dependent ERK1/2-mediated phosphorylation and subsequent gene up-regulation of the mineralization inhibitors matrix Gla protein and osteopontin. This result suggested that both PiTs are necessary for Pi signaling. Moreover, the ERK1/2 phosphorylation could be rescued by overexpressing Pi transport–deficient PiT mutants. Using cross-linking and bioluminescence resonance energy transfer approaches, we found that PiT1 and PiT2 form high-abundance homodimers and Pi-regulated low-abundance heterodimers. Interestingly, in the absence of sodium-dependent Pi transport activity, the PiT1-PiT2 heterodimerization was still regulated by extracellular Pi levels. Of note, when two putative Pi-binding residues, Ser-128 (in PiT1) and Ser-113 (in PiT2), were substituted with alanine, the PiT1-PiT2 heterodimerization was no longer regulated by extracellular Pi. These observations suggested that Pi binding rather than Pi uptake may be the key factor in mediating Pi signaling through the PiT proteins. Taken together, these results demonstrate that Pi-regulated PiT1-PiT2 heterodimerization mediates Pi sensing independently of Pi uptake.

demonstrate that P i -regulated PiT1-PiT2 heterodimerization mediates P i sensing independently of P i uptake.
Phosphorus is the sixth most abundant element in the human body, constituting ϳ1% of total body weight (1). About 85% of total phosphate can be found in the skeleton, where it is a major constituent of hydroxyapatite crystals deposited on the extracellular organic matrix during the mineralization process. The remaining 15% is found mainly in cells from soft tissues and in extracellular volume, where it represents Ͻ1% of total phosphate (2)(3)(4). In plasma, ϳ16% of circulating phosphate is found as organic phosphate bound to proteins and lipids, whereas the major part (84%) is orthophosphate, or free inorganic phosphate (P i ), 5 that can be filtered by the kidney (1). At physiological pH, the monovalent H 2 PO 4 Ϫ and the divalent HPO 4 2Ϫ forms are present at a 1:4 molar ratio (5). Although this plasma P i represents a small fraction of total body phosphorus, it serves as an exchange pool between the various P i -containing and -regulating organs, and disturbances in P i homeostasis can affect almost all organ systems.
In addition to the widespread structural and metabolic functions of P i , it has become increasingly apparent during the past 15 years that extracellular P i can act as a signaling molecule directly altering gene expression and cell phenotype (6 -9). The abundance of P i in the skeleton has led to early studies describing the effects of extracellular P i in this organ. Exposing cultured chondrocytes to a high level of extracellular P i leads to their terminal maturation and subsequent matrix mineralization (10 -13). Similarly, the apoptosis of terminally differentiated hypertrophic chondrocytes was shown to be dependent upon the circulating plasma P i levels in vivo (14). In both of these in vitro and in vivo approaches, the P i -mediated apoptosis of chondrocytes was dependent upon the activation of the MAPK ERK1/2 pathway (15)(16)(17), but not of other mitogen-activated protein kinases, such as p38 or c-Jun N-terminal kinase. Interestingly, the P i -dependent activation of the ERK1/2 pathway up-regulated the gene expression of the mineralization inhibitors matrix Gla protein (Mgp) and osteopontin (Opn), most likely setting off a feedback mechanism to control P i -induced mineralization (7,16,18,19). The elevated extracellular P i level was also shown to be important in osteoblast proliferation and differentiation (16, 20 -24), cementoblast formation (25), odontoblast differentiation (26,27), and osteoclast differentiation (28 -30).
The P i -mediated signaling underlies the notion that cells must possess a P i -sensing mechanism on the surface of or within the cell that is able to detect and respond to the variation of extracellular P i levels. The ability of organisms to detect changes in extracellular levels of other metabolites (such as Ca 2ϩ , glucose, or amino acids) has already been described (31)(32)(33), and emerging evidence suggests that similar events are at work to mediate the cellular effects of P i (8, 34 -36). Although the identity of the molecules involved in these mechanisms is still unknown in mammals, P i -sensing machineries have been characterized in prokaryotic and eukaryotic unicellular organisms (37). In Escherichia coli, the phosphate transporter PstS and other periplasmic proteins (PstC, Pst, and PstB/PhoU) detect the variation of external P i concentrations. In case of a low extracellular P i level, this system increases the efficiency of P i retention in the bacteria (38). In Saccharomyces cerevisiae, a low extracellular P i level resulted in induction of the P i transporter Pho84, now identified as the essential component of the P i -sensing system (38,39). Interestingly, following P i restriction, it was demonstrated that Pho84 could trigger the rapid activation of protein kinase A without transporting P i (40).
In mammals, the Slc20a1/PiT1 and Slc20a2/PiT2 proteins are expressed at the plasma membrane and have been described as mediating the intracellular uptake of P i with a high affinity (41)(42)(43). PiT1 and PiT2 have a wide tissue distribution, being the only P i transporters expressed in bone (44,45). Interestingly, their expression can be modulated by extracellular P i (21,43,44,46,47), and previous studies have suggested that they can mediate downstream effects of extracellular P i . In bone, the elegant study of Kimata et al. (48) suggests that the chondrocyte response to extracellular P i is mediated by a PiT1-dependent up-regulation of cyclin D1 through ERK1/2 pathway activation. The authors hypothesize that P i -driven conformational changes of PiT1 could be involved in the P i -sensing mechanism. In parathyroid cells, PiT1 was suggested to act as a P i sensor to modulate the secretion of the phosphaturic parathyroid hormone (49). On the other hand, based on its property of oligomerizing upon extracellular P i variation, PiT2 was also proposed to serve as a P i sensor (50). Although these data support a possible role for PiT1 or PiT2 as P i sensors, little is known about the underlying mechanisms. Because PiT1 and PiT2 have very close P i transport characteristics (51), they may also share P isensing properties and thus have interconnected roles in P i sensing. Moreover, because P i -independent functions have been highlighted recently for PiT1 (52)(53)(54)(55)(56), the involvement of P i transport in the P i sensing by PiT1 or PiT2 remains to be investigated.
In this report, we investigated the role of PiT1 and PiT2 as P i sensors in osteoblastic and chondrocytic cell lines. We show that both PiT1 and PiT2 are required for mediating P i -dependent signaling. We demonstrate that PiT1 and PiT2 could interact together and that extracellular P i modulates this interaction. Finally, we show that cellular P i uptake is not required to mediate P i signaling through the PiT proteins.

Requirement of both PiT1 and PiT2 for P i -mediated signaling
We first investigated whether PiT1 and/or PiT2 were involved in the P i -dependent up-regulation of Mgp and Opn expression. To this aim, using RNA interference, we established stably transfected osteoblastic MC3T3-E1 clones in which PiT1 or PiT2 expression was knocked down. In MC3T3-E1 shPiT1 clones, PiT1 gene expression showed a 63% reduction, together with a significant up-regulation of PiT2 (Fig. 1A). Similarly, the MC3T3-E1 shPiT2 clones displayed a 62% decrease in PiT2 mRNA level, together with a significant up-regulation of PiT1 (Fig. 1A). Comparable results were observed when cells were incubated with 10 mM extracellular P i concentration (Fig. S1A). Interestingly, the sodium-dependent P i uptake was similar in MC3T3-E1 shPiT1 and shPiT2 clones and control MC3T3-E1 cells (Fig. 1B), suggesting that a depletion of either PiT may be compensated by the remaining PiT, as was previously suggested (56). Consistent with this possibility, MC3T3-E1 clones stably transfected with both shPiT1 and shPiT2 resulted in a 52% reduction of both PiTs (Fig. 1A), resulting in a similar decrease in sodium-dependent P i uptake (Fig. 1B). In contrast to wildtype differentiated MC3T3-E1 cells in which Mgp and Opn expression was up-regulated following stimulation with 10 mM extracellular P i for 24 h, the up-regulation of Mgp and Opn expression in PiT-depleted MC3T3-E1 clones was blunted (Fig.  1C). The defect in P i -dependent Mgp and Opn up-regulation arose despite a normal P i transport in the shPiT1 or shPiT2 MC3T3-E1 clones, suggesting that a variation in intracellular P i content is unlikely to account for defects in P i -dependent signaling in the absence of either PiTs. Because the ERK1/2 signaling pathway was shown to be required for P i -dependent regulation of Mgp and Opn expression (16,19), we investigated the P i -dependent ERK1/2 activation in differentiated PiT-depleted MC3T3-E1 clones. We showed that following a 30-min (Fig. S1B) or 24-h ( Fig. 1D) stimulation with 10 mM extracellular P i , the activation of ERK1/2 pathway was blunted in shPiT1, shPiT2, or shPiT1/shPiT2 clones, as compared with untransfected and shScramble-transfected cells. Similar data were obtained in three separate PiT-depleted MC3T3-E1 clones (Fig. S1, B-D) and in transiently transfected MC3T3-E1 cells (Fig. S2). Interestingly, the effect of PiT depletion on the activation of ERK1/2 pathway following stimulation with 10% fetal bovine serum (FBS) was not as pronounced as P i stimulation, arguing for a specificity of PiT1 and PiT2 in the P i -dependent activation of the ERK1/2 pathway.
Moreover, we performed similar experiments in the MC615 chondrogenic cell line. We used a transient transfection approach leading to a 50 and 56% deletion of PiT1 and PiT2 mRNA levels, respectively ( Fig. 2A). A similar extent of PiT Pivotal role of PiT1 and PiT2 in bone phosphate sensing deletion was obtained at the protein level, as shown by immunofluorescence (Fig. S3). MC615 cells were also used to rescue PiT deletion by overexpressing human PiT1 and PiT2 in PiT1and PiT2-depleted cells, respectively (Fig. 2B). Similar to what was observed in MC3T3-E1 cells, depletion of either PiT1 or PiT2 in MC615 cells blunted the activation of the ERK1/2 pathway by 10 mM extracellular P i , despite the up-regulation of the remaining PiT (Fig. 2C). When human PiT1 was overexpressed in PiT1-depleted MC615 cells, we could rescue the P i -dependent ERK1/2 phosphorylation (Fig. 2C). Similar results were obtained when human PiT2 was overexpressed in PiT2-depleted MC615 cells or when both human PiT1 and PiT2 were overexpressed in PiT1-PiT2-depleted MC615 cells (Fig. 2C). This further illustrated the requirement of both PiT1 and PiT2 for the P i -dependent ERK signaling in cell lines of skeletal origin.

PiT1 and PiT2 form hetero-oligomers upon variation of extracellular P i concentrations
The requirement of both PiT1 and PiT2 for P i -dependent ERK1/2 signaling may indicate the existence of a functional protein complex comprising both PiTs. This possibility is also reinforced by the presence in PiT1 and PiT2 protein sequences of a conserved and highly hydrophobic 127-amino acid domain that was suggested to be important for determining the quaternary structure of the protein (57). To investigate the formation of hetero-and homo-oligomers, we used HEK293T cells that are easy to transfect, allow a robust expression of PiTs at the plasma membrane, and have been shown to have similar P i -mediated ERK1/2-specific activation (58). In addition, because PiT proteins are often difficult to detect using total cell extracts in Western blot experiments, we used a crude cell fractionation . B, sodium-dependent P i uptake was measured in untransfected or stably transfected MC3T3-E1 cells, as indicated. Data are means Ϯ S.E. (n ϭ 3). C, RT-qPCR analysis of Mgp and Opn mRNA levels in untransfected or stably transfected MC3T3-E1 cells, as indicated. Cells were incubated in low-serum (0.5%) medium for 24 h and stimulated with 1 mM (white bars) or 10 mM (black bars) extracellular P i concentration for 24 h. Data are means Ϯ S.E. (error bars) (#, p Ͻ 0.05; ##, p Ͻ 0.01 versus 1 mM P i control; and *, p Ͻ 0.05; **, p Ͻ 0.01 versus shScramb; n ϭ 3) D, Western blot analysis of ERK1/2 phosphorylation (P-ERK 1/2) in untransfected or stably transfected MC3T3-E1 cells, as indicated. Cells were incubated in low-serum (0.5%) medium for 24 h and stimulated for another 24 h with 1 mM or 10 mM extracellular P i concentration or with 10% FBS used as a positive control for ERK1/2 phosphorylation. Total ERK1/2 proteins were used as a loading control.

Pivotal role of PiT1 and PiT2 in bone phosphate sensing
approach to analyze an enriched plasma membrane protein fraction revealed by the specific expression of the Na/K-AT-Pase, as shown in Fig. 3A. When analyzing the enriched plasma membrane fraction from PiT1-and/or PiT2-transfected cells by Western blotting, we could observe protein complexes after cell surface cross-linking using BS 3 , a membrane-impermeable cross-linker (Fig. 3B). The apparent molecular mass of 142-165 kDa that we detected from cells transfected with hPiT2 alone recapitulated the results obtained by Salaün et al. (50) demonstrating the formation of PiT2 homodimers. Similarly, the 151-188 kDa apparent molecular mass band detected from cells transfected with hPiT1 is consistent with the association of two PiT1 molecules. When cells were co-transfected with both hPiT1 and hPiT2, no change was observed in the band profile, apart from a less intense signal due to transfection with 50% fewer hPiT1 or hPiT2 plasmids. In this condition, we could not detect a distinct band at the theoretical PiT1-PiT2 heterodimer molecular mass, most likely due to the expected similar molecular weights of PiT homo-and heterodimers.
To study further the PiT oligomerization, we then used a BRET approach. To this aim, we constructed hPiT1 and hPiT2 chimeric proteins expressing the eYFP acceptor or Rluc donor. Because structure-function studies have excluded a role of the large intracellular loop (iLoop) in P i transport and retrovirus binding (59 -62) and showed no overlapping between iLoop and the highly hydrophobic domain (57), we substituted the iLoop with eYFP and Rluc sequences (Fig. 3C). When expressed in HEK293T, the chimeric hPiT1-eYFP or -Rluc and hPiT2-eYFP or -Rluc proteins could be visualized at the plasma membrane, as shown by confocal microscopy (Fig. 3D), enabling us to study their role in detecting the variation of extracellular P i levels. We performed saturation BRET experiments in living cells to investigate their potential hetero-and homo-oligomerization. As shown in Fig. 3E, we obtained typical BRET-saturable curves when using hPiT1-eYFP and hPiT2-Rluc proteins, together with a high BRET ratio. In contrast, when the BTN3A2 protein was used instead of either PiT, no saturation could be achieved, together with a weak BRET ratio (Fig. S4, A and B). These data support the notion that PiTs can form hetero-and homo-oligomers specifically. We confirmed the specificity of the hetero-oligomers by a competition assay (Fig. S4C) whereby overexpression of untagged PiT1 or PiT2 reduces the BRET ratio, whereas unrelated BTN3A2 expression does not. We next determined whether the interaction of PiTs could be modulated by the variation of extracellular P i concentration. We therefore performed the same saturation BRET experiments after a 10-min stimulation with 1, 3, or 10 mM extracellular P i . Results reported in Fig. 3 (F and G) and Fig. S5A showed that saturation curves were different upon extracellular P i concentration only for hPiT1-hPiT2 hetero-oligomers. Indeed, calculating the BRET 50 values from these curves, we showed a significant decrease at 10 mM extracellular P i , suggesting a stronger interaction between PiT1 and PiT2 in this condition, whereas no variation was observed for homo-oligomers. In all conditions, the BRET max values did not vary significantly (Fig. S5B).

activation by extracellular P i and regulation of PiT hetero-oligomerization are P i transport-independent
Our results illustrated that both PiT proteins are important for P i -dependent ERK signaling and that P i can modulate PiT hetero-oligomerization. To elucidate whether the P i transport function of PiTs was important to mediate P i -dependent ERK1/2 signaling, we used the previously reported hPiT1 S128A and hPiT2 S113A P i transport-deficient mutants (56,63,64). When hPiT1 S128A was overexpressed in PiT1-depleted cells in which P i -dependent ERK1/2 activation was lost, we could rescue the activation of ERK1/2 signaling (Fig. 4A). Similarly, we could rescue the P i -dependent ERK1/2 activation in PiT2-depleted cells by overexpressing hPiT2 S113A mutant. This was also true when cells depleted from both PiTs were transfected by both hPiT1 S128A and hPiT2 S113A (Fig. 4A). These data demonstrated that the P i transport function of PiTs was dispensable for the P i -dependent ERK1/2 activation. To further study whether the sodium-dependent P i transport function of PiTs was important to mediate P i -dependent PiT hetero-oligomerization, we performed BRET experiments in the absence of Na ϩ . In this condition where P i transport was blunted, the variation of extracellular P i concentrations was still able to modulate PiT interaction ( Fig. 4 (B and C) and Fig. S5C). Because P i was able to modulate PiT interaction without being transported, this suggests that the binding of P i to PiT proteins rather than its actual uptake into the cell may be involved in the modulation of PiT interaction. We next generated mutated versions of the chimeric hPiT1-eYFP and hPiT2-Rluc proteins in which Ser 128 or Ser 113 was replaced by an alanine residue. As expected, overexpression of chimeric hPiT1 S128A -eYFP and hPiT2 S113A -Rluc resulted in decreased sodium-dependent P i transport compared with transporting chimeric PiT proteins (Fig. 4D). Interestingly, using a BRET approach, we showed that although hPiT1 S128A -eYFP and hPiT2 S113A -Rluc were still able to interact together, this interaction was not modulated anymore by extracellular P i variations ( Fig. 4E (left) and Fig. S5C), consistent with a role for Ser 128 or Ser 113 in P i binding. When BRET experiments were performed with a transport-deficient mutant (hPiT1 S128A -eYFP or hPiT2 S113A -Rluc) and a transporting chimeric PiT protein (hPiT1-eYFP and hPiT2-Rluc), we could recover the modulation of PiT interaction by P i (Fig. 4E (middle  and left) and Fig. S5C), further supporting a role of Ser 128 or Ser 113 in the P i -dependent interaction of PiTs.

Discussion
The ability of a cell to detect changes in extracellular P i levels is paramount for its adequate response to environmental fluctuations and critical for the appropriate modulation of P i homeostasis and skeletal mineralization. In this work, we provided mechanistic insights into the molecular events leading to the detection of changes in extracellular P i concentrations using skeletal cell lines as a model.
Skeletal cells are constantly exposed to high local extracellular P i levels, mainly due to the continual resorption and formation of the mineralized extracellular matrix of bone and the need for tremendous quantities of P i for mineralization purposes (65). This has made bone a model of choice to study P i signaling, where it has been shown in early studies to regulate the programmed cell death of hypertrophic chondrocytes (10,11,13). In subsequent studies, we and others have shown that P i could regulate skeletal mineralization by controlling the expression of the mineralization inhibitors Mgp and Opn through the activation of the ERK1/2 pathway (7,15,16,18,19). Our present study brings evidence for a role of PiT1 and PiT2 in transmitting the P i signal to the cell by showing that the P i -dependent up-regulation of Mgp and Opn and ERK1/2 phosphorylation were blunted in PiT1-or PiT2-depleted cells. A role for PiT1 in mediating ERK1/2 signaling has been reported earlier (48,55); however, a similar role for PiT2 has never been illustrated before.
A critical aspect in studying the molecular events involved in P i signaling is whether or not P i needs to be transported within the cell to fulfill its role. This question is particularly relevant in view of the role of PiT1 and PiT2 as mediators of P i signaling because these two proteins have well-described P i transport functions (41)(42)(43). By using P i transport-deficient mutants of PiT proteins, we could rescue the ERK1/2 signaling, demonstrating that PiT1 and PiT2 can mediate a P i signal without transporting the ion. Furthermore, the loss of ERK1/2 signaling in the absence of PiT1 or PiT2 was not associated with a change in cellular P i uptake, further indicating that the transport of P i into the cell is not necessary for P i signaling.
These data have major implications for the understanding of the P i -signaling mechanism. The absence of functional compensation by PiT1 in PiT2-depleted cells, and vice versa, together with the requirement of both PiTs for P i signaling, suggested that they could interact functionally and/or physically to mediate the P i signal. Consistent with this hypothesis, using a BRET approach, we showed that PiT proteins could Figure 3. Specific interaction between PiT1 and PiT2 varies upon extracellular P i concentration. A, Western blotting analysis of Na/K-ATPase, hPiT1, and hPiT2 expression in pmaxGFP-transfected (Ctrl) or hPiT1 and hPiT2 co-transfected HEK293T cells after crude cellular fractionation, as indicated (see "Experimental procedures" for details). Overexpression of hPiT1 and PiT2 allowed a better signal than Ctrl. B, Western blotting analysis (IB) of hPiT1 (left) and hPiT2 (right) expression in HEK293T cells untransfected (UT) or transfected with hPiT1, hPiT2, or both after cell surface cross-linking using bis(sulfosuccinimidyl)suberate and cellular fractionation. Enriched plasma membrane fraction (F1) was analyzed. C, schematic representation of chimeric hPiT1-eYFP or -Rluc and hPiT2-eYFP or -Rluc proteins used for BRET assays. The DNA region encoding for the large internal loop of PiT1 and PiT2 (iLoop1 and iLoop2, respectively) was replaced by coding sequences of eYFP or Rluc. Transmembrane domains (black rectangles) are shown. D, confocal images obtained from transfected HEK293T cells, as indicated. Rluc was detected after immunofluorescence and nuclei were stained using Hoechst (blue). Scale bar, 10 m. E, representative BRET saturation curves of co-transfected HEK293T cells, as indicated. F, representative BRET saturation curves (left panel) and BRET 50 index (right panel) of HEK293T co-transfected with hPiT1-eYFP and hPiT2-Rluc obtained following a 10-min incubation with 1, 3, or 10 mM extracellular P i concentration, as indicated. BRET 50 data are means Ϯ S.E. (error bars) (**, p Ͻ 0.01 versus 1 mM P i condition, n ϭ 8). G, the BRET 50 index was measured from BRET saturation curves obtained from HEK293T co-transfected with hPiT1-eYFP and hPiT1-Rluc (left) or hPiT2-eYFP and hPiT2-Rluc (right) following a 10-min incubation with 1, 3, or 10 mM extracellular P i concentration. BRET 50 data are means Ϯ S.E., and no statistically difference was observed (n ϭ 4 -8).

Pivotal role of PiT1 and PiT2 in bone phosphate sensing
form homo-and heterodimers. More strikingly, we could illustrate that the formation of PiT1-PiT2 heterodimers only was affected by the variation of extracellular P i levels. We could not quantify the relative importance of heteroversus homodimers, but the cross-linking data supported the idea that the P i -sensitive PiT1-PiT2 heterodimers were present at much lower quantities. It is possible that a low-abundance P i -sensitive PiT1-PiT2 heterodimer may be more effectively tunable than a highly abundant P i sensor. This abundance may also be consistent with the on/off P i effect on ERK1/2 signaling that we have observed when deleting the PiT proteins.
Although a P i -sensitive PiT1-PiT2 heterodimer is likely to represent an important component of the P i -sensing machinery, deciphering the detailed functioning of such a sensor requires additional work. Nevertheless, our data provide several important mechanistic insights that may give clues to the understanding of the P i -signaling cascade. We showed that in the absence of Na ϩ , which blunts the P i transport activity of PiT proteins, the formation of PiT1-PiT2 heterodimers was still sensitive to extracellular P i variations. This is consistent with the idea that binding of P i , rather than transport itself, is important for PiT1-PiT2 heterodimerization. This also suggests that Ser 128 and Ser 113 residues that were demonstrated to be critical for P i transport by PiT1 and PiT2 (56, 63) may indeed be involved in P i binding, a prerequisite step for P i transport (51). In line with this possibility, when BRET experiments were performed using PiT1 and PiT2 chimeric proteins in which Ser 128 and Ser 113 were substituted with alanine residues, the PiT1-PiT2 heterodimer was not responsive anymore at extracellular P i variations. The substitution of only one serine, however, rescued P i sensitivity of the PiT1-PiT2 heteroduplex, illustrating the complex relationship between the structural arrangement Overexpression of P i transport-deficient hPiT1 (PiT1 S128A ) and/or hPiT2 (PiT2 S113A ) was performed in PiT1-, PiT2-, or PiT1-PiT2-depleted cells, respectively. Cells were incubated in low-serum (0.5%) medium for 24 h and stimulated for 30 min with 1 or 10 mM extracellular P i concentration. Total ERK1/2 proteins were used as a loading control. B, sodium-dependent and -independent P i uptake was measured in HEK293T cells transfected as indicated. Data are means Ϯ S.E. (error bars) (*, p Ͻ 0.05 versus sodium-dependent, n ϭ 4). C, BRET 50 index was measured from BRET saturation curves obtained after a 10-min stimulation with 1, 3, or 10 mM extracellular P i concentration from HEK293T co-transfected with hPiT1-eYFP and hPiT2-Rluc in an Na ϩ -free medium. Data are means Ϯ S.E. (*, p Ͻ 0.05 versus 1 mM P i condition, n ϭ 4). D, sodium-dependent P i uptake was measured in HEK293T cells transfected with pcDNA6A plasmid (Ctrl) or with plasmids containing normal (hPiT1 and hPiT2) or P i transport-deficient mutants (hPiT1 S128A and hPiT2 S113A ) with or without BRET chimeric acceptor or donor, as indicated. Data are means Ϯ S.E. (*, p Ͻ 0.05 versus P i -transporting PiTs, n ϭ 4). E, BRET 50 index was measured from BRET saturation curves obtained after 1, 3, or 10 mM extracellular P i concentration stimulation for 10 min from HEK293T co-transfected with the indicated plasmids. Data are means Ϯ S.E. (*, p Ͻ 0.05 versus 1 mM P i condition, n ϭ 4 -5).

Pivotal role of PiT1 and PiT2 in bone phosphate sensing
of the P i binding site and P i sensitivity. A determination of the crystal structure of PiT proteins is therefore necessary to help determine the identity of the P i binding site and its role in P i transport and sensing.
An interesting consequence of our work is that PiT proteins should not be considered anymore as P i transporters solely but also as P i receptors able to mediate P i signaling by activating specific downstream pathways. Such a hybrid transporter-receptor behavior may indicate that the PiT1-PiT2 heteroduplex could behave as a P i transceptor, whereby conformational changes during the transport cycle (including the P i -binding step) affect a signal transduction component that triggers a downstream signaling pathway (66). The difference between a transceptor and a pure receptor is that the transceptor can also transport the ligand into the cell. In recent years, evidence for transporters functioning as transceptors has been obtained in several eukaryotic systems (67). Interestingly, in yeast, the Pho84 phosphate carrier that is considered as an essential component of the P i -sensing system was characterized as a P i transceptor (40). In prokaryotic or lower eukaryotic organisms, true transceptors could be functionally characterized by specific mutants that either lack the signaling capacity and retain normal transport or have lost transport but retain signaling. In the case of PiT proteins, the signaling capacity is lost when either PiT1 or PiT2 is deleted, indicating that the possible transceptor function may only be revealed by heterodimerization and may not be associated with structural changes of either PiT. The gain of function provided through PiT heterodimerization implies that a unique protein complex mediates P i signaling. This underlies the idea that specific PiT1 and PiT2 protein partners may be involved in this process. If this is the case, the identification of PiT-specific partners in the future may provide unique targets to modulate P i signaling in specific organs or specific physiological conditions.
In summary, this study provided mechanistic insights into the P i -signaling cascade in skeletal cell line models by unraveling PiT1-PiT2 heterodimers as essential components of a P isensing machinery. Although in vivo studies are required to strengthen the physiological relevance of these findings, our work may help in deciphering the mechanisms underlying the ability of the organism to respond to the serum P i level variations, which is the necessary first step in the P i homeostasisregulating cascade.

Cells and culture conditions
Murine preosteoblastic MC3T3-E1 cells were seeded at 10 4 cells/cm 2 and cultured for 10 days in ␣-minimum essential medium GlutaMAX TM (catalogue no. 32751, Thermo Fisher Scientific, Saint-Aubin, France) supplemented with 10% FBS and 1% penicillin/streptomycin. Murine chondrogenic MC615 cells were seeded at 10 4 cells/cm 2 and maintained in a medium consisting of Dulbecco's modified Eagle's medium (DMEM) high glucose GlutaMAX TM /Ham's F-12 (1:1) (catalogue no. 31966 and 31765, respectively, Thermo Fisher Scientific) supplemented with 10% FBS and 1% penicillin/ streptomycin. HEK293T cells were maintained in DMEM high glucose GlutaMAX TM supplemented with 10% FBS, 10 mM HEPES, and 50 g/ml gentamicin. Cells were cultured at 37°C in a humidified atmosphere of 5% CO 2 in air, and media were renewed every 2-3 days. When indicated, cells were incubated in low-serum (0.5%) medium for 24 h and stimulated with various concentrations of P i for 30 min or 24 h. P i was added as a mixture of NaH 2 PO 4 and Na 2 HPO 4 (pH 7.4), as described previously (10).

RNA interference
Inactivation of PiT1 or PiT2 were performed by cloning an shPiT1 (56) or shPiT2 (see sequences in Table 1) into pSUPER vector (68) expressing a puromycin resistance gene. A scramble sequence cloned into pSUPER was used as a negative control (56). Stable knockdown of PiT1 or PiT2 was obtained by transfecting MC3T3-E1 cells with 5 g of the pSUPER-shRNAs using the T-20 program of the Amaxa nucleofector system (Cell Line Nucleofector TM Kit V VCA-1003, Lonza, Bâle, Switzerland). Cells were plated at limiting density, and puromycinresistant clones were picked, expanded, and tested for PiT expression. Experiments were performed with three independent stable transfectants, and the data presented illustrate a representative clone. Transient inactivation of PiT was also performed in MC3T3-E1 and MC615 cells using transfection as described above. These transient experiments were stopped 72 h after transfection.

Pivotal role of PiT1 and PiT2 in bone phosphate sensing Gene expression analysis
Total RNA was isolated from cells using the NucleoSpin RNA II kit (Macherey-Nagel, Hoerdt, France) according to the manufacturer's instructions. RNA was reverse transcribed using Affinity Script (Agilent Genomics, Santa Clara, CA) as per the manufacturer's recommendations. Real-time PCR was performed on a Mx3000P system (Stratagene, San Diego, CA) using Brilliant III Ultra-Fast SYBR QPCR Master Mix (Agilent Genomics). The following temperature profile was used: denaturation at 95°C for 3 min, amplification during 40 cycles of 5 s at 95°C and 20 s at 60°C, followed by a step at 95°C for 1 min and 65°C for 30 s. Expression of target genes were normalized to GAPDH expression levels and were calculated as described previously (69). Primer sequences are listed in Table 1.

Immunofluorescence and confocal microscopy
One day after transient transfection, MC615 cells were fixed/ permeabilized in methanol at Ϫ20°C for 5 min and blocked in 1% bovine serum albumin for 1 h at room temperature. Immunodetection of PiT1 and PiT2 was performed using rabbit antimouse PiT1 or PiT2 antibody, respectively (generously provided by Dr. G. Friedlander), at a 1:200 dilution overnight at 4°C, and goat anti-rabbit Alexa488 secondary antibody (catalogue no. A11008, Thermo Fisher Scientific) at a 1:1,000 dilution for 1 h at room temperature. The nuclei were counterstained using 1 g/ml TO-PRO 3 iodide solution (Thermo Fisher Scientific). The stained cells were mounted in Prolong Antifade mounting medium (Thermo Fisher Scientific). Images were acquired with a Nikon Eclipse TE2000E confocal microscope (Nikon, Badhoevedorp, The Netherlands) equipped with a ϫ60 oil immersion objective. The averaged intensity for PiT1 and PiT2 staining was determined using Metamorph version 7.5 software.

Phosphate uptake measurements
HEK293T cells were seeded at 10 5 cells/cm 2 in poly-L-lysine (Sigma-Aldrich)-precoated 24-well plates. Cells were transfected with 0.5 g/well of plasmid using JetPrime (Polyplus transfection) according to the manufacturer's instructions. P i uptake in HEK293T cells was measured 24 h after transfection. To perform uptake experiments in MC3T3-E1 cells, PiT-depleted MC3T3-E1 clones were seeded at 1.5 ϫ 10 5 cells/cm 2 , and uptake was performed 24 h later. P i uptake was performed as described previously (70). Briefly, after three washing steps, cells were incubated in an uptake medium containing 100 M [ 32 P]KH 2 PO 4 (0.5 Ci/ml) in the presence of 137 mM NaCl (total P i transport) or N-methyl-D-glucamine (sodium-independent transport) at 37°C for 10 min. sodium-dependent P i transport was calculated as the difference between total and sodiumindependent P i transports. Cells were washed three times and lysed with 0.1 M NaOH solution, and aliquots of cell lysates were taken for the determination of protein content (Pierce BCA protein assay kit, Thermo Fisher Scientific) and the radioactivity by liquid scintillation counting (Ultima Gold LLT, PerkinElmer Life Sciences) in a Hidex 300 SL ␤ counter.

Crude cellular fractionation and cross-linking
HEK293T cells were seeded at 10 5 cells/cm 2 in 6-well plates and transfected with 1 g/well each of hPiT1 and hPiT2 plasmids using JetPrime (Polyplus transfection) according to the manufacturer's instructions. Twenty-four hours post-transfection, cells were scraped and lysed in 1 ml of buffer A/well (10 mM Hepes, pH 7, 10% sucrose, 5 mM EDTA, Phosphatase Inhibitor Mixture 3 (catalogue no. P0044), and Protease Inhibitor

Pivotal role of PiT1 and PiT2 in bone phosphate sensing
Mixture (catalogue no. P8340) from Sigma-Aldrich. Lysates were passed 10 times through a 22-gauge 1-ml syringe and centrifuged at 800 ϫ g for 5 min at 4°C. Pellet (fraction 1; F1) was resuspended in 20 l of radioimmune precipitation assay buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 20 mM EDTA, Phosphatase Inhibitor Mixture 3, and Protease Inhibitor Mixture). The supernatant was centrifuged at 20,000 ϫ g for 45 min at 4°C, and the resulting supernatant (Fraction 2; F2) was reserved. The pellet (Fraction 3; F3) was resuspended in 20 l of radioimmune precipitation assay buffer and submitted to two freeze/thaw cycles, a 10-min incubation at 100°C, and a 30-s sonication. All fractions were incubated 1 h at 4°C with gentle mix. F3 was centrifuged at 12,000 ϫ g for 10 min at 4°C, and supernatant was used for analysis.
For cross-linking experiments, transfected HEK293T cells were washed three times in cold PBS and cross-linked 30 min at 4°C with 1 mM bis(sulfosuccinimidyl)suberate (catalogue no. 21580, Thermo Fisher Scientific). This cell surface cross-linker was blunted with ice-cold 50 mM Tris-HCl, pH 8, for 15 min and washed twice with cold PBS. Cells were then lysed in Buffer A and processed for cell fractionation as described above.

Construction of chimeric and transport-deficient PiT plasmids
The hPiT1 and hPiT2 sequences were previously cloned into pcDNA6A plasmid (56). The V5-His 6 fusion tag sequence present in pcDNA6A was excluded by introducing a stop codon at the end of the hPiT coding sequence using site-directed mutagenesis (QuikChange, Agilent Genomics) (see primers in Table 1). The iLoop of hPiT1 and hPiT2 was then replaced by eYFP or RLuc proteins using FastCloning (71). Briefly, using Phusion high-fidelity DNA polymerase (New England Biolabs) and overlapping specific primers (Table 2), we PCR-am-plified the pcDNA6A-hPiT1 and pcDNA6A-hPiT2 vectors with the exception of the iLoop regions and the coding regions of eYFP and Rluc from pEYFP-N1 (Clontech, Mountain View, CA) and phRluc-C1 (PerkinElmer Life Sciences) plasmids, respectively. After digestion of the DNA templates by DpnI, PCRamplified overlapping sequences were reassembled by E. colimediated recombination-ligation following transformation in high-efficiency NEB 10-␤ competent cells (New England Biolabs, Ipswich, MA). In the final hPiT1-eYFP or -Rluc chimeric constructs, the eYFP or Rluc coding sequences were inserted in place of iLoop1 (amino acids 268 -492). Similarly, the eYFP or Rluc coding sequence was inserted in place of iLoop2 (amino acids 256 -450) in the hPiT2-eYFP or -Rluc constructs. The integrity of the constructs was verified by sequencing. To serve as a control for BRET experiments, the BTN3A2 DNA sequence (generously provided by Dr. Scotet, INSERM UMR1232, Centre de Recherche en Cancérologie et Immunologie, Nantes-Angers, France), encoding a small cell surface-expressed protein, was fused to the N terminus of eYFP or Rluc coding sequences using the same strategy (primers used are reported in Table 2).
To generate transport-deficient mutant versions of hPiT1-eYFP and hPiT2-Rluc chimeric constructs, we replaced the Ser 128 of hPiT1 or Ser 113 of hPiT2 by an alanine using sitedirected mutagenesis (QuikChange, Agilent Genomics), as reported previously (56,63). The sequences of the primers used are listed in Table 1.

BRET saturation assays
HEK293T cells were seeded at 5 ϫ 10 4 cells/cm 2 in 12-well plates. The next day, cells were co-transfected using JetPrime (Polyplus transfection) with a fixed amount of hPiT1-Rluc (50 ng/well), hPiT2-Rluc (100 ng/well), or BTN3A2-Rluc (10 ng/well) plasmids (encoding BRET donors) and variable amounts of hPiT1-eYFP or hPiT2-eYFP (from 12.5 to 1,500 ng/well) or BTN3A2-eYFP (from 1.56 to 50 ng/well) plasmids (encoding BRET acceptors). The pcDNA6A empty vector was used to compensate for the variable amounts of transfected DNA and to ensure equivalent transfection conditions in each well. Twenty-four hours later, transfected cells were detached using 0.5 mM EDTA solution and seeded at 1.5 ϫ 10 5 cells/cm 2 in white flat bottom 96-well plates in duplicate. BRET experiments were performed 48 h post-transfection. Cells were washed once with 0.9% NaCl solution and stimulated with various concentrations of P i for 10 min. P i was added as a mixture of NaH 2 PO 4 and Na 2 HPO 4 (pH 7.4). When indicated, cells were previously starved of P i by incubating cells overnight with DMEM high glucose no phosphates (catalogue no. 11971, Thermo Fisher Scientific) supplemented with 10% FBS, 10 mM HEPES, and 50 g/ml gentamicin before P i stimulation. The coelenterazine h substrate (UPR3078, Interchim Uptima, Montluçon, France) was added at a final concentration of 5 M by automated injection in the Mithras LB940 plate reader (Berthold Technologies, Versailles, France), and 485-and 530-nm light emissions were measured consecutively several times. The BRET ratio was calculated as the ratio of light emitted by the acceptor fusion protein at 530 nm over the light emitted by the donor fusion protein at 485 nm. Values were

Pivotal role of PiT1 and PiT2 in bone phosphate sensing
corrected with the background signal calculated from a well without donor fusion protein. The BRET 50 was calculated as the eYFP/Rluc value at which the BRET ratio is half of the maximum BRET ratio achieved at saturating substrate concentration.

Statistical analysis
Data are expressed as mean Ϯ S.E. GraphPad version 5.0 software was used to perform Mann-Whitney tests. A p value of Ͻ0.05 was considered statistically significant. Unless otherwise stated, experiments were repeated at least three times.