Mapping of the minimal inorganic phosphate transporting unit of human PiT2 suggests a structure universal to PiT-related proteins from all kingdoms of life

Background The inorganic (Pi) phosphate transporter (PiT) family comprises known and putative Na+- or H+-dependent Pi-transporting proteins with representatives from all kingdoms. The mammalian members are placed in the outer cell membranes and suggested to supply cells with Pi to maintain house-keeping functions. Alignment of protein sequences representing PiT family members from all kingdoms reveals the presence of conserved amino acids and that bacterial phosphate permeases and putative phosphate permeases from archaea lack substantial parts of the protein sequence when compared to the mammalian PiT family members. Besides being Na+-dependent Pi (NaPi) transporters, the mammalian PiT paralogs, PiT1 and PiT2, also are receptors for gamma-retroviruses. We have here exploited the dual-function of PiT1 and PiT2 to study the structure-function relationship of PiT proteins. Results We show that the human PiT2 histidine, H502, and the human PiT1 glutamate, E70, - both conserved in eukaryotic PiT family members - are critical for Pi transport function. Noticeably, human PiT2 H502 is located in the C-terminal PiT family signature sequence, and human PiT1 E70 is located in ProDom domains characteristic for all PiT family members. A human PiT2 truncation mutant, which consists of the predicted 10 transmembrane (TM) domain backbone without a large intracellular domain (human PiT2ΔR254-V483), was found to be a fully functional Pi transporter. Further truncation of the human PiT2 protein by additional removal of two predicted TM domains together with the large intracellular domain created a mutant that resembles a bacterial phosphate permease and an archaeal putative phosphate permease. This human PiT2 truncation mutant (human PiT2ΔL183-V483) did also support Pi transport albeit at very low levels. Conclusions The results suggest that the overall structure of the Pi-transporting unit of the PiT family proteins has remained unchanged during evolution. Moreover, in combination, our studies of the gene structure of the human PiT1 and PiT2 genes (SLC20A1 and SLC20A2, respectively) and alignment of protein sequences of PiT family members from all kingdoms, along with the studies of the dual functions of the human PiT paralogs show that these proteins are excellent as models for studying the evolution of a protein's structure-function relationship.


Background
Phosphate is needed by any living cell for structural and metabolic purposes. Inorganic phosphate (P i ) has to be actively transported across the cell membrane against a chemical and electrical gradient. In mammalian cells this task is managed by the type III sodium-dependent P i (NaP i ) symporters, PiT1 and PiT2, which utilize the free energy provided by the Na + concentration gradient as the driving force for uphill import of P i [1][2][3], reviewed in [4].
The mammalian type III transporters are part of the P i transport (PiT) family (SLC20 [5]; TC #2.A.20 [6]), but several members were originally identified as receptors for different retroviruses belonging to the gamma-I 53 -L 127 http://prodom.prabi.fr ( Figure 1). In an attempt to narrow down a PiT family trait, Saier aligned the Nterminal protein sequences from 17 members representing all kingdoms [6,30]. The author noted the existence of an 11-amino-acid-long sequence in the N-terminal region containing the conserved core sequence [GAND-VANA] and proposed it to be a signature sequence for the PiT family [6,30]. However, refined studies of the Nand C-termini of 109 protein sequences representing PiT family members from all kingdoms revealed that these proteins harbor a 12-amino-acid-long PiT family signature sequence -with the common core consensus sequence [GANDVANA] -within each of the PD001131 ProDom domains proposed in 2004 [18]. Furthermore, D 28 and D 506 shown to be critical for PiT2 P i transport are placed in either of the PiT family signature sequences [18].
To further investigate the importance of the PiT family signature sequences, we have analyzed the human PiT2 histidine, H 502 , located in the C-terminal PiT family signature sequence, and we show that it is indeed critical for the P i transport function but dispensable for infection by PiT2 cognate gamma-retroviruses. The human PiT2 H 502 is the second amino acid in this sequence to be identified as critical for P i transport function. In addition, we also show that the human PiT1 glutamate, E 70 , located in the PD001131 ProDom domain, is critical for the P i transport function but dispensable for infection by PiT1 cognate gammaretroviruses.
We have, moreover, combined studies of the gene structure of the human PiT genes (SLC20A1 and SLC20A2), alignment and TM domain prediction of protein sequences of PiT family members from all kingdoms of life, and studies of the dual functions of the human PiT paralogs as P i transporters and gamma-retroviral receptors, and we found that these proteins are excellent as models for studying the evolution of protein structure-function relationship. Specifically based on the observation that bacterial and archaeal PiT family members are substantially smaller than eukaryotic members [18] and our alignment (Additional File 1 Figure A), we analyzed truncation mutants of human PiT2. Our results clearly show that the large intracellular domain of human PiT2 is dispensable for P i transport function, and that a fully functional P i -transporting unit can be created by the 10 TM domains and the small loop sequences connecting them (human PiT2ΔR 254 -V 483 ). A further truncated human PiT2 protein with the 5 th and 6 th TM domains and the large intracellular domain removed resembles the structures of as well a putative phosphate permease from archaea as of PiTA from bacteria (Archaeoglobus fulgidus (A. fulgidus) and E. coli, respectively). This mutant (human PiT2ΔL 183 -V 483 ) was an excellent gamma-retroviral receptor [31], and we here show that it can support low levels of P i transport. Altogether, these results suggest that the overall structure of the P i -transporting unit of the PiT family proteins has remained unchanged during evolution.

Sequence alignment
Protein sequence alignment of nine PiT family members representing all kingdoms was made using the ClustalW alignment program version 2.0.12 available at the European Bioinformatics Institute server (URL: http://www. ebi.ac.uk/clustalw2/) [32]. The Swiss-Prot protein sequences were retrieved from the NCBI Protein server (URL: http://www.ncbi.nlm.nih.gov/protein/). Accession ]. All sequences encompass two 12-aminoacid-long sequences, which based on comparison of 109 sequences, were identified in PiT proteins and related proteins and suggested to be PiT family signature sequences [18]. We, however, observed that the C-terminal PiT family signature sequence of E. coli PiTA did not group together with the C-terminal PiT family signature sequences of the eight other species in the alignment (data not shown). In order to group all the Cterminal PiT family signature sequences together, the alignment was adjusted manually after an alignment of   [8,20]. The numbers of the TMs are indicated above the model. Other membrane topology models have been proposed for PiT1 [22,23] and PiT2 [24], which suggested diverging topology for the two paralogs; the alternative PiT2 model is shown in Figure 2. The amino acids previously identified in human PiT2 as being critical for P i transport function are highlighted with black filling and pointed out with arrows; D 28 , E 55 , S 113 , D 506 , E 575 , and S 593 [18,27,28]. In human PiT1, the amino acids S 128 (PiT2 S 113 ) and S 621 (PiT2 S 593 ) have previously been identified as being critical for PiT1 P i transport function [29]. In the present study, human PiT2 H 502 (situated in the PiT family signature sequence) and human PiT1 E 70 (equivalent in position to human PiT2 E 55 ) are also identified as critical for P i transport function (see Figure 3). The grey-filled sequences (L 11 -L 161 and V 492 -V 640 ), represent the Nand C-terminal, respectively, ProDom domains (PD001131) published in 2004 defining the PiT family members [27].  [33], and DAS is based on lowstringency dot-plots of the query sequence against a collection of non-homologous membrane proteins using a previously derived special scoring matrix [34].   [24]; the TMs are shown as grey-filled sequences and their numbers are indicated with roman numbers above the model. This model shares some similarity to a membrane topology model for PiT1 proposed in 2002 [22]. Based on the cellular location of C-terminal tags, the Cterminal ends of PiT1 and PiT2 were predicted to be extracellular [22,24]. And based on the cellular location of an N-terminal tag on PiT2 and glycosylation of a site in human PiT1 and partly glycosylation of the same site in human PiT2 although oddly not in hamster PiT2, the N-termini of PiT1 and PiT2 were suggested to be extracellular [22,24]; due to a suggested additional TM after TM3 in Figure 1 (TMIV in this figure), this did not influence the orientation of the large intracellular domain in these models compared to the model in Figure 1. The PiT2 model shown in Figure 1 and this figure, respectively, and the PiT1 model proposed in 2002 [22] were later compared by us [18]. In 2009, Farrell and coworkers proposed a modified model of human PiT1 based on substituted cysteine accessibility mutagenesis [23]. The recent model of PiT1 shows more resemblance to the PiT2 models shown in this figure and in Figure  In general, the predictions using both servers correspond well to each other when compared (data not shown), however, the DAS server tends to predict shorter TM domains in agreement with the tendency for prokaryotic TM domains to be shorter in length when compared to the length of eukaryotic TM domains [35]. Therefore, we chose to use the DAS server over the TMHMM server when predicting TM domains in the prokaryotic protein sequences for E. coli PiTA and A. fulgidus putative phosphate permease. The predicted TM domains are shown in Additional File 1 Figure A.

Intron-exon border analysis of human PiT genes SLC20A1 and SLC20A2
The SPIDEY mRNA-to-genome DNA alignment program version 1.40 available from the NCBI homepage (URL: http://www.ncbi.nlm.nih.gov/spidey/index.html) [36] was used to determine the location of intron-exon borders in the human PiT genes. SPIDEY takes as input an mRNA sequence and the corresponding genomic sequence, and it generates an alignment that establishes the gene structure. The GenBank mRNA sequences were retrieved from the NCBI nucleotide server (http:// www.ncbi.nlm.nih.gov/nuccore/). Accession numbers are: H. sapiens
The plasmid encoding the human PiT2 H 502 A mutant was made by using the QuickChange ® XL site-directed mutagenesis kit (Stratagene, La Jolla CA, USA) according to the manufacturer's instructions. Besides the mutations creating H 502 A, the forward primer 5'-TTCGGGTCCTTTGCTGCCGGCGGCAATGACGT-3' and reverse primer 5'-ACGTCATTGCCGCCGGCAG-CAAAGGACCCGAA-3' also generated, by introduction of a silent mutation, an NgoM IV restriction enzyme cleavage site in pOJ74, which was used for screening. The plasmid encoding the human PiT1 E 70 K mutant was made by using the Altered sites II kit (Promega, Madison WI, USA) according to the manufacturer's instructions. A mutation creating E 70 K as well as a Dra I restriction enzyme cleavage site was introduced into a pAlter-1 vector (Promega) harboring the Pst 1 -Hind III fragment of pOJ75 (the nucleotide sequence encoding the N-terminal part of the human PiT1 protein) using the primer 5'-GACAGAGCCCACTGTTTTAAA-GATGCTAGCTAG-3'. Finally, this construct was digested with Kpn I and Hind III generating a fragment, which was used to replace the corresponding fragment in pOJ75 resulting in the desired plasmid.
The plasmid encoding the human PiT2ΔL 183 -V 483 mutant has previously been described [31]. The plasmid encoding the human PiT2ΔR 254 -V 483 mutant was made using a pAlter-1 vector harboring the Pst I -Hind III fragment of pOJ74 (the nucleotide sequence encoding the N-terminal part of the human PiT2 protein) as template in a polymerase chain reaction (PCR) with the forward primer 5'CTATAGGGAGACCCAAGCTTTGTT TATTTAA3' and the reverse primer 5'GAGGACCTG-GAGGAAATGGAACAGGAGGTGTGATAAAGCACC TTCTTTTTG3'; the latter primer was used to create the link between the 5' sequence encoding KEGALS 253 and the 3' sequence encoding H 484 LLFH ( Figure 1). The amplification product was digested with Sse 8387 I and Hind III and used to replace the corresponding fragment in pOJ74 resulting in the desired plasmid.
The authenticities of all the nucleotide sequences were confirmed.
The plasmids were purified using either cesium chloride (CsCl) according to the protocol described by Maniatis and coworkers [38], or using Nucleobond (Macherey-Nagel, Düren, Germany) or Qiagen maxiprep (Qiagen GmbH, Hilden, Germany) according to the manufacturer's instructions.
Vectors were harvested as supernatants from confluent producer cells, and the vector containing supernatants were filtered (0.45-μm pore size) and stored at -80°C.

Transient transfection and infection assay
Transient transfection-infection assays were performed essentially as described [37]. Briefly, CHO K1 cells seeded in 60-mm-diameter dishes at 8 × 10 4 cells per dish were transfected with 2 μg per dish of plasmid DNA encoding human PiT2 (pOJ74), human PiT1 (pOJ75), human PiT2 H 502 A, human PiT1 E 70 K, or equimolar amounts to human PiT2 of human PiT2ΔR 254 -V 483 or human PiT2ΔL 183 -V 483 . Mock treated cells were transfected with empty vector DNA (pcDNA1A R tkpA). Three independent precipitates were made per construct. Forty-eight hours after transfection, approx. 4 to 8 × 10 4 10A1 MLV or A-MLV pseudotypes carrying the G1BgSvN transfer vector were added per dish in the presence of Polybrene. Forty-eight hours later, the dishes were stained and evaluated. Infection was analyzed by counting the number of β-galactosidase-positive (infected) cells per dish. Analyses for FeLV-B and GALV receptor functions were performed on MDTF cells using 1.5 × 10 4 cells and approx. 1.5 to 3.0 × 10 4 vector pseudotypes per dish. Numbers of vector pseudotypes used in the experiments were calculated from the number of β-galactosidase-positive colonies per mL obtained on D17 cells as described [37].

P i transport assay
Female Xenopus laevis (X. laevis) frogs were obtained from Nasco (Nasco, Modesto CA, USA) and kept and handled according to guidelines from the Danish Animal Experiments Inspectorate. Oocytes were isolated from frogs anesthetized in a 0.1-0.2% MS.222 (3-aminobenzoic acid ethyl ester) (Sigma, St. Louis MO, USA) solution for 10-30 minutes. A 1-1.5 centimeters incision was made in the abdomen and several ovaries were removed surgically by authorized personnel. The oocytes were manually dissected and subsequently collagenase (Sigma, St. Louis MO, USA) treated and maintained in modified Barth's solution [88 mM NaCl, 1 mM KCl, 0.82 mM MgSO 4 , 0.4 mM CaCl 2 , 0.33 mM Ca(NO 3 ) 2 , 2.4 mM NaHCO 3 , 10 mM HEPES-KOH, pH 7.5, 100 IU per mL penicillin, 100 μg per mL streptomycin] at 18°C as described [28]. The following day, the oocytes were used for cRNA injection and subsequent analyses of 32 P i uptake essentially as described previously [28]. Briefly, cRNAs were prepared from Apa 1 ( Figure 3A) or Bln 1 ( Figures 3B and 6) linearized plasmid preparations applying the mMESSAGE mMACHINE kit (Ambion, Austin TX, USA). Stage V-VI oocytes were microinjected with 12.5 ng of cRNA (or H 2 O as negative control) and incubated at 18°C. After two to three days, the oocytes were washed in phosphate-free uptake solution [100 mM NaCl, 2 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES-Tris pH 7.5], and hereafter incubated in uptake solution containing 0.1 mM KH 2 32 PO 4 (2 mCi per mL, New England Nuclear, Boston MA, USA) at RT for 1 hour. The oocytes were washed in ice-cold uptake solution containing 5 mM KH 2 PO 4 and the 32 P i uptake of each oocyte measured in a liquid scintillation counter as described previously [28]. It should be noted that factors coupled to the health and husbandry of the female X. laevis frogs can influence the oocyte batches. These factors include nutrition, season of the year (light cycle), water temperature, salinity and hardness of the water, water contaminants or toxins, and diseases [44], and the impact is that different batches of oocytes injected with cRNAs encoding the same proteins exhibit different average transport capacities.

Statistical analysis
The null hypothesis that two mean values are identical was tested by a two-tailed Student's t-test. The test compares the actual difference between two mean values in relation to the variation in the data (expressed as the standard error of the difference between the mean values). The null hypothesis was rejected, e.g., the mean values were considered different when P<0.05.

Results and discussion
Human PiT1 E 70 and human PiT2 H 502 are critical for P i transport function but dispensable for gamma-retroviral receptor function In a former study, we identified the putative 2 nd -TM domain-positioned human PiT2 E 55 as being critical for PiT2 P i transport function (Figure 1) [28]. The human PiT2 paralog, human PiT1, harbors a corresponding glutamate in position 70, E 70 . To investigate whether this conserved residue was important for PiT1 P i transport function, it was mutated to a lysine generating the mutant human PiT1 E 70 K. In the experiment shown in Figure 3A, oocytes injected with cRNA encoding human PiT1 supported a 32 P i uptake of 119.86 ±28.16 pmol/ oocyte-hour at pH 7.5 in agreement with previous results obtained addressing the Na 32 P i uptake function of human PiT1 in X. laevis oocytes [45]. The P i transport function of human PiT1 E 70 K was severely impaired when compared to that of wildtype PiT1 (P = 0.002, 2.78 ±0.74 pmol/oocyte-hour (PiT1 E 70 K)) ( Figure  3A); see Additional File 2 for data and statistics to Figure 3.
Besides being P i -transporting proteins, the mammalian PiT proteins also serve as gamma-retroviral receptors, and this dual-function allows for analyzing whether a mutated PiT protein is properly processed, folded and translocated to the cell surface [18,28]. The human PiT1 E 70 K mutant was therefore analyzed for gamma-retroviral receptor function using a transient transfectioninfection assay [37]. For the infection assay, retroviral vectors harboring a β-galactosidase encoding transfer vector and carrying viral surface proteins responsible for receptor recognition were used; vectors carrying, e.g., 10A1 MLV surface proteins are referred to as 10A1 MLV vector pseudotypes. Eukaryotic expression plasmids encoding human PiT1 and human PiT1 E 70 K mutant protein were transfected into CHO K1 cells non-permissive for infection by 10A1 MLV vector pseudotypes ( Figure 3C) [28]. The abilities of these proteins to support infection by 10A1 MLV vector pseudotypes were analyzed; the infection levels were evaluated as the number of β-galactosidase positive (blue) cells per 60mm-diameter dish. CHO K1 cells expressing human PiT1 were permissive for infection by 10A1 MLV vector pseudotypes ( Figure 3C) in agreement with PiT1's welldescribed receptor function for 10A1 MLV [10]. Moreover, the human PiT1 E 70 K mutant supported wildtype PiT1 levels of 10A1 MLV infection (884 ±146 blue cells per dish (PiT1), 767 ±42 blue cells per dish (PiT1 E 70 K), P = 0.48) ( Figure 3C). Besides being a receptor for 10A1 MLV, PiT1 is also a receptor for GALV [7] and for FeLV-B [13]. The human PiT1 E 70 K protein was analyzed in parallel for receptor function for vector psedotypes of these two viruses in non-permissive Mus dunni tail fibroblasts and found to sustain wildtype PiT1 infection levels of GALV (2087 ±780 blue cells per dish (PiT1), 1992 ±273 blue cells per dish (PiT1 E 70 K), P = 0.91) and FeLV-B (1424 ±346 blue cells per dish (PiT1), 1715 ±527 blue cells per dish (PiT1 E 70 K), P = 0.67).
The wildtype receptor functions of PiT1 E 70 K confirm that the overall membrane topology is preserved and that the processing to the cell surface was unaffected by the E 70 K-mutation.
The glutamate E 70 in human PiT1 is conserved in eukaryotic PiT family members as are the other two human PiT1 residues (S 128 and S 621 ) (Additional File 1 Figure A) previously shown to be critical for P i transport function [29]. Since the corresponding glutamate and serine residues in human PiT2 have already been identified as being critical for P i transport function [27,28], this demonstrate that equivalent glutamate or serine residues in the human PiT paralogs both are critical for their P i transport functions. These observations illustrate that it is highly likely that the other conserved amino acids identified in human PiT2 as being critical for P i transport function also are important for the transport function of human PiT1 and other PiT family members.
The histidine residue, human PiT2 H 502 is positioned in the 7 th TM domain according to the Johann topology model (Figure 1) [20]. It is, moreover, located in the Cterminal PiT family signature sequence and conserved in eukaryotic PiT family members [18] (Additional File 1 Figure A). Moreover, analysis of 60 sequences of bacterial PiT family members revealed only 5 sequences without the histidine residue illustrating that this residue is also highly preserved in the C-terminal PiT family signature sequence of PiT family members belonging to this kingdom [18] (Additional File 1 Figure A). Since the conserved aspartic acid in the C-terminal PiT family signature sequence, that is human PiT2 D 506 , is critical for P i transport of PiT2 [18], we hypothesized that other conserved amino acids in this motif might be critically involved in P i transport function of human PiT2 and other members of the PiT family as well. Mutation of human PiT2 H 502 to alanine created the mutant denoted PiT2 H 502 A. This mutant was analyzed for 32 P i transport function in X. laevis oocytes ( Figure 3B) and 10A1 MLV and A-MLV receptor functions in CHO K1 cells (Figures 3D-E).
In the experiment shown in Figure 3B, oocytes injected with cRNA encoding human PiT2 supported a 32 P i uptake of 44.96 ±0.46 pmol/oocyte-hour at pH 7.5 in agreement with former studies addressing the Na 32 P i uptake function of human PiT2 in X. laevis oocytes [18,28,45]. The Pi transport function of human PiT2 H 502 A was severely impaired when compared to that of wildtype PiT2 (P = 0.002, 2.36 ±0.56 pmol/oocyte-hour (PiT2 H 502 A)) ( Figure 3B).
To analyze whether the human PiT2 H 502 A mutant is properly folded and processed to the cell surface, it was also analyzed for gamma-retroviral receptor function using the transient transfection-infection assay [37].  Figure 3E). These results demonstrate that the overall membrane topology of human PiT2 H 502 A is preserved, and that the processing of human PiT2 H 502 A to the membrane surface is unaffected by the mutation. Thus, histidine 502 in the 7 th TM domain is the second amino acid -besides D 506 -in the C-terminal PiT family signature sequence [HGANDVQNAIGP], which has been shown to be essential for human PiT2 P i transport function. While the exact role of the histidine residue in the C-terminal signature sequence still needs to be revealed, its critical role for human PiT2 P i transport function emphasizes the importance of the C-terminal PiT family signature sequence in the physiological function of the PiT proteins.
Besides human PiT1 E 70 and human PiT2 H 502 , six conserved amino acids in human PiT2 and two corresponding positions in human PiT1 have previously been identified as being critical for P i transport function [18,[27][28][29]. All these amino acids are located in the Pro-Dom domains (PD001131) suggested in 2004 to define members of the PiT family (Figure 1) [27]. Therefore it is likely that sequences outside these two domains might be dispensable for the P i transport function of the PiT proteins, and that a minimal P i -transporting unit of the PiT proteins can be identified.

Alignment of protein sequences of PiT family members from all kingdoms
A previously published alignment of human PiT1 and human PiT2 protein sequences shows that the L6 loopthe large intracellular domain -is the region where these sequences diverge the most [8] (Additional File 1 Figure A). Moreover, alignment of human PiT1 and N. crassa Pho-4 + shows that the large intracellular domain (L6) is smaller in Pho-4 + , whereas the rest of the Pho-4 + protein sequence aligns well with the protein sequence of human PiT1 [20] (Additional File 1 Figure A). To further address this, we counted the number of amino acids in the large intracellular domain (L6) of nine different PiT family members and plotted the lengths according to their phylogenetic relationship in Figure  4A. The figure shows that PiT family members from archaea and bacteria harbor the shortest L6 loops whereas the PiT-proteins from chordates harbor the longest L6 loops ( Figure 4A, see also Additional File 1 Figure A). Note that the L6 loop of the C. elegans putative phosphate permease is unexpectedly short (73 amino acids), and according to the plot we would have expected a L6 loop length for this protein in the interval between 175 and 232 amino acids ( Figure 4A). The observed differences in the L6 loop lengths of PiT family members from different species thus suggest that the L6 loop evolved from being a regular loop to become a regular domain during evolution. In order to address this issue, we counted the number of amino acids in all loops (L1 to L9) in the nine PiT family members and plotted the average loop lengths ±SEM in Figure 4B. The figure shows that the L6 loop in average is much larger than all other loops (L6: >131.7 ±32.8 amino acids, Figure 4B); see Additional File 2 for data to Figure  4B. The figure also shows that the variation in the L7 loop lengths is substantial (42.9 ±14.7 amino acids), see Figure 4B legend for discussion.   maximum length of 42 amino acids, see legend to Figure  4B for discussion. The definition of the maximum length of a loop also has the impact that the L7 of E. coli PiTA consisting of 160 amino acids (Additional File 1 Figure A) has to be considered a domain. In summary, analysis of the sizes of the loop sequences L1 to L9 in nine PiT family members from all kingdoms led to the determination of a limit of maximum 42 amino acids in a regular loop sequence -and sequences longer than 42 amino acids are highly likely domains. In support of our calculations of the maximum loop length for PiT-proteins is a previous study of 243 transmembrane domaincontaining sequences, with 146 sequences being multitransmembrane spanning, showing that~90% of the loops are shorter than 40 amino acid residues [46]. Another study supporting our finding is the analysis of loops in 79 existing 3D structures of transmembrane proteins showing that the majority of loops connecting transmembrane domains are shorter than 50 amino acid residues [47]. The proteins in Figure 4A with L6 loop sizes smaller than 42 amino acids are the archaeal putative phosphate permease and the bacterial PiTA protein, implying that single cell organisms without nuclei that rarely harbor membrane-bound organelles cope without the large intracellular domain, whereas single cell animals (protozoan's) with nuclei and membrane-bound organelles have distinct L6 domains as shown for the T. brucei putative phosphate permease ( Figure 4A). Altogether this suggest a role(s) for the large intracellular domain, which is not directly related to P i transport per se, and it also suggest that the large intracellular domain (L6) may have increased in length during the evolution from archaea to chordata as a consequence of adaptation to more complex environments.
Besides a difference in the lengths of L6, a difference in the number of TM domains in the PiT family members was observed (Additional File 1 Figures A and B). The illustration of TM domain conservedness (black boxes) and TM domains, which are suggested by us to be present but not predicted by protein sequence analysis using the TMHMM server (red boxes, see argumentation in legend to Additional File 1 Figure A), shows the following conservedness of TMs: TM 4, TM 8, TM 10 (fully conserved) > TM 5, TM 6 (fully conserved in eukaryotes) > TM 1, TM 2, TM 3 > TM 9 > TM 7 (least conserved) (Additional File 1 Figure B). The most prominent observation is that E. coli PiTA and A. fulgidus putative phosphate permease both lack the 5 th and 6 th TM domains (Additional File 1 Figures  A and B). This in addition to the previous observation that these two proteins also lack the L6 domain (Figure 4A), suggest that the 5 th and 6 th TM domains and the L6 domain are dispensable for P i transport function, and that a basic P i -transporting unit of the PiT family members can be identified. This unit would consist of regions flanking the large intracellular domain (L6) but highly likely also be devoid of the 5 th and 6 th TM domains. Interestingly, in support of this theory, drawing of the putative topology models for human PiT2, E. coli PiTA, and A. fulgidus putative phosphate permease based on the alignment in Additional File 1 Figure A, shows that the bacterial and archaeal proteins have a predicted eight TM backbone where the N-terminal PiT-family signature sequence is placed in the 1 st extracellular loop (L1) and the Cterminal PiT family signature sequence is placed in the 3 rd extracellular loop (L7) ( Figure 5). In comparison, the drawing of the putative topology model for human PiT2 shows a backbone of 10 TM domains where the Nterminal and C-terminal PiT-family signature sequences are placed in the 1 st extracellular loop (L1) and the 4 th extracellular loop (L7), respectively ( Figure 5). An interpretation of these drawings could be that the intraprotein locations of the N-terminal and C-terminal PiT-family signature sequences are of importance, and that TM 1 to TM 4 and TM 7 to TM 10 constitute a core sustaining the P i -transporting function whereas TM 5 and TM 6 and the large intracellular domain (L6) constitute a regulatory unit. Finally, the amino acids identified as being critical for P i transport function are located in the ProDom domains suggested in 2004 (TM 1 to TM 4 and TM 7 to TM 10) [27] (Figure 1) in agreement with the 5 th and 6 th TM domains and the large intracellular domain (L6) might be dispensable for the P i transport function.

Design of human PiT2 truncation mutants
To identify the minimal P i -transporting unit, two human PiT2 truncation mutants were analyzed. They were designed to address the P i transport function and the gamma-retroviral receptor functions of: 1) A human PiT2 mutant protein, which consists of the 10 TM domains and a L6 loop of 18 amino acids (human PiT2 P 236 -S 253 ) creating the mutant human PiT2ΔR 254 -V 483 . The human PiT2ΔR 254 -V 483 mutant does not resemble a naturally occurring homolog found in lower species, and it is merely designed to address if the large intracellular domain is dispensable for Na + -dependent P i -uptake (Figure 1), and 2) A human PiT2 mutant protein that resembles an archaeal and bacterial homolog with respect to protein composition, i.e., lacking the 5 th and 6 th TM domains and the large intracellular domain (L 183 -V 483 ) (human PiT2ΔL 183 -V 483 ) (Figure 1). Note that in the Salaün model the 5 th and 6 th TM domains correspond to TMVI and TMVII (Figure 2).
The large intracellular domain (R 254 -V 483 ) of human PiT2 is dispensable for P i transport function whereas the fragment L 183 -V 483 is more critical for P i transport function The Na + -dependent 32 P i transport function of wildtype human PiT2 and the human PiT2-derived truncation mutants PiT2ΔL 183 -V 483 and PiT2ΔR 254 -V 483 (Figure 1) were analyzed in X. laevis oocytes ( Figure 6).
The 32 P i transport activities of the PiT2 mutant lacking the major part of the large intracellular domain, human PiT2ΔR 254 -V 483 , (47.38 ±6.59 pmol/oocyte-hour ( Figure  6A) and 38.74 ±3.73 pmol/oocyte-hour ( Figure 6B)) were indistinguishable from those of PiT2 (P = 0.119) ( Figure  6A) and P = 0.553 ( Figure 6B)); see Additional File 2 for data and statistics to Figure 6. Thus, the large intracellular domain of human PiT2 but 18 amino acids (fragment R 254 -V 483 ) is dispensable for its P i transport function.
The 32 P i transport activity of the human PiT2 mutant lacking the large intracellular domain as well as the 5 th and 6 th TM domains, PiT2ΔL 183 -V 483 (Figure 1), was severely impaired (3.93 ±0.44 pmol/oocyte-hour ( Figure  6A) and 8.33 ±2.85 pmol/oocyte-hour ( Figure 6B)) when compared to the P i transport function of wildtype PiT2

C N
A. fulgidus putative phoshate permease  PiT2 PiT2 H 2 O L 183 -V 483 R 254 -V 483 Figure 6 Na 32 P i uptake mediated by human PiT2 and truncation mutants analyzed in X. laevis oocytes. Oocytes were injected with H 2 O or cRNA of the indicated constructs. Two (experiment A) or three (experiment B) days later, a 32 P i uptake assay was performed and the 32 P i uptake in individual oocytes was measured. Data are the mean value of (n) numbers of oocytes ±SEM, see Additional File 2 for data and statistics.

Viral receptor function of mutant PiT2 proteins
Using the transient transfection-infection assay, we analyzed whether the deletions in human PiT2 affected their viral receptor functions for A-MLV and 10A1 MLV. Eukaryotic expression plasmids encoding human PiT2 and the mutant proteins were transfected into CHO K1 cells. As expected, human PiT2 transfected cells were permissive for infection by both 10A1 MLV and A-MLV vector pseudotypes (Table 1). While the human PiT2 truncation mutant lacking the large intracellular domain, human PiT2ΔR 254 -V 483 (Figure 1) was a fully functional P i transporter ( Figure 6), it only supported low levels of PiT2 cognate gamma-retroviral infection ( Table 1). Note that human PiT2ΔR 254 -V 483 was tested once for A-MLV receptor function and twice for 10A1 MLV receptor function. The A-MLV study was done in parallel to a 10A1 MLV receptor function study using the same set of plasmid precipitates. Interestingly, the human PiT2 truncation mutant lacking the 5 th and 6 th TM domains in addition to the large intracellular domain, human PiT2ΔL 183 -V 483 (Figure 1), supported substantial levels of PiT2 cognate gammaretroviral infection (Table 1) [31] showing that its low levels of P i transport function were not due to incorrect processing of this mutant to the cell surface.
PiT2 regions directly involved in receptor function for 10A1 MLV and A-MLV have also been identified by expression of chimeric proteins in CHO K1 cells and were found to be located in the putative extracellular loops 2 (L3) and 4 (L7) (Figure 1) [26,37,48,49]. Both of the human PiT2 mutants, PiT2ΔR 254 -V 483 and PiT2ΔL 183 -V 483 , harbor extracellular loops 2 (L3) and 4 (L7) according to the Johann PiT2 model (Figure 1). Based on their -here identified -P i transport abilities, it is unlikely that PiT2ΔR 254 -V 483 is less expressed at the cell surface than PiT2ΔL 183 -V 483 , and the observation that the less truncated human PiT2 mutant protein is a worse gamma-retroviral receptor than a more heavily truncated human PiT2 mutant protein might instead reflect a disturbance of the folding and/or conformation of the extracellular loops 2 (L3) and 4 (L7) due to the sole presence of the extracellular loop 3 (L5) without the large intracellular domain in PiT2ΔR 254 -V 483 .

Intron-exon borders of the human PiT genes SLC20A1 and SLC20A2
The human PiT proteins are encoded by genes that localize to different chromosomes. The human gene, SLC20A1, encoding the PiT1 protein is located on chromosome 2 at position q13 [50,51], and the human gene, SLC20A2, encoding the PiT2 protein is located on chromosome 8 at position p11.2 [8,52,53].
To analyze the gene structure of SLC20A1 and SLC20A2, the intron-exon borders in each of the genes were determined using the SPIDEY mRNA-to-genome DNA alignment as described in "Methods". The intronexon borders are marked with stars (✰) and vertical lines in the PiT1 and PiT2 protein sequences in the alignment of nine PiT family members in Additional File 1 Figure A.
Eight out of nine intron-exon borders (labeled ✰ a to e and ✰ g to i on PiT1 and PiT2 in Additional File 1 Figure A) in SLC20A1 and SLC20A2 are predicted to be homologous. One intron-exon border (labeled ✰ f 1 (SLC20A2) and f 2 (SLC20A1)) are displaced giving a gap corresponding to 12 amino acids (~36 nucleotides). These two borders are placed in the middle of the genome sequences, which encode the large intracellular domain (L6) of the human PiT proteins. As seen from Additional File 1 Figure A, the alignment between the human PiT proteins in this region is poor and the gap highly likely reflects this, and not a significant difference in intron-exon structure between SLC20A1 and SLC20A2.
Interestingly, in support of the theory that the 5 th and 6 th TM domains can be dispensable for P i transport function, is the observation that these TM domains are encoded by two different exons, see Additional File 1 Figure A (✰ labeled c to d, and ✰ labeled d to e), and therefore the possibility exists that the sequences in these two exons have entered later in evolution.

Specialized functions of the mammalian PiT proteins
Mammalian PiT proteins are expressed in all tissues investigated and due to their broad expression profiles, they have been suggested to accommodate house-keeping functions, i.e., supplying cells with P i to maintain basic cellular functions [2,20,54]. However, in recent years additional specialized functions of the PiT proteins have been reported. These include roles for PiT2 in proximal tubule phosphate reabsorption [55], and for PiT1 in regulation of parathyroid gland PTH production [56,57], cell proliferation [29,58,59], and in tumor necrosis factor (TNF) induced apoptosis [60]. Recent studies also indicate that both the PiT proteins function as P i sensors [27,56], reviewed in [61]. Interestingly, some of these functions, that is, PiT2's suggested role in P i sensing [27] and PiT1's role in cell proliferation and TNFinduced apoptosis [29,59,60] have been shown to be independent of the P i transport functions of the proteins.
PiT1 has also been implicated in normal chondroblastic and osteoblastic differentiation and mineralization processes [62][63][64][65][66], as well as trans-differentiation of vascular smooth muscle cells to cells with characteristics of chondro-/osteoblasts in the pathologic process of vascular calcification at hyperphosphatemia [67]. More rodent in vivo models have been used to study the role of PiT1 in normal bone formation and/or embryonic development. Rats with transgenic overexpression of PiT1 showed no major bone deformity during skeletal development [57]. However, these rats displayed a slight but significant decrease in the bone mineral content of the whole skeleton together with a reduction albeit non-significant in the total bone area [57]. The role of PiT1 during embryonic mouse development has been studied by two different groups employing early conditional excision of SLC20A1 Exons 3-4 [68] and SLC20A1 Exon 5 [59], which resulted in homozygous embryonic lethality. Both studies find that the embryos are anemic and do not survive past E12.5, at which stage the morphology shows reduced growth [59,68]; the anemia was found to be due to severe defects in liver development [59]. Comparison of wildtype mice to mice with low (15%) expression of PiT1 mRNA showed that some of the latter mice displayed impaired bone mineralization at birth, while 15-days old mice showed no major differences in mineralization [59]. Interestingly, in embryos (E11.5) lacking PiT1 expression Beck and coworkers found an upregulated PiT2 expression, which however could not rescue the embryos past E12.5, and the authors therefore suggest that the critical non-redundant role of PiT1 in development is not P i -uptake [59]. Altogether, the in vivo studies do not exclude a role for PiT1 in normal bone formation, although they imply that PiT1 is not critical for the early skeletal developmental processes.
The alignment and analyses of exon structure together with the observed P i transport functions of the PiT2 deletion mutants presented here might suggest that the regions of the PiT proteins involved in the P i -transport independent functions map to sequences in the 5 th and 6 th TM domains and/or in the large intracellular domain. In line with this, we are currently investigating the function of the large intracellular domain of the human PiT2 protein and our results support the hypothesis that the large intracellular domain has other functions than P i transport.

Conclusions
Investigation of the P i transport and retroviral receptor functions of the human PiT proteins has allowed for identification of a histidine residue (human PiT2 H 502 ) in the C-terminal PiT family signature sequence as being critically involved in P i transport function. Moreover, we show that a PiT1 glutamate residue (human PiT1 E 70 ) positioned in the 2 nd TM domain is critical for P i transport function in agreement with the former identification of the equivalent glutamate in human PiT2 (human PiT2 E 55 ) as being critical for P i transport function [28].
We have shown that a human PiT2 mutant consisting of the 10 TM domains and minor loops (human PiT2ΔR 254 -V 483 ) transports P i as wildtype PiT2, proving that the large intracellular domain (L6) is dispensable for P i transport function. A further truncated human PiT2 mutant consisting of the 1 st to 4 th TM domains linked to the 7 th to 10 th TM domains and the minor loop sequences connecting the TMs (human PiT2ΔL 183 -V 483 ), and which resembles archaeal and bacterial homologs, sustained low levels of P i transport. This protein harbors the ProDom domains defining the PiT family members and, moreover, harbors all the amino