Two Highly Conserved Glutamate Residues Critical for Type III Sodium-dependent Phosphate Transport Revealed by Uncoupling Transport Function from Retroviral Receptor Function* □ S

Type III sodium-dependent phosphate (NaP i ) cotrans- porters, Pit1 and Pit2, have been assigned housekeeping P i transport functions and suggested involved in chon- droblastic and osteoblastic mineralization and ectopic calcification. Both proteins exhibit dual function, thus, besides being transporters, they also serve as receptors for several gammaretroviruses. We here show that it is possible to uncouple transport and receptor functions of a type III NaP i cotransporter and thus exploit the retro- viral receptor function as a control for proper processing and folding of mutant proteins. Thus exchanging two putative transmembranic glutamate residues in human Pit2, Glu 55 and Glu 575 , with glutamine or with lysine severely impaired or knocked out, respectively, P i transport function, but left viral receptor function undisturbed. Both glutamates are conserved in type III NaP i cotransporters, in fungal NaP i cotransporters PHO-4 and Pho89, and in other known or putative phosphate permeases from a number of species and are the first residues shown to be critical for type III NaP i cotransport. Their putative transmembranic positions to- gether with the presented data are consistent with Glu 55 and Glu 575 being parts of a cation liganding site or playing roles in conformational changes associated with substrate transport. Finally, the results also show that Pit2 retroviral receptor function per se is not dependent on Pit2 P i transport function. transport function Pit2, retroviral receptor function retroviral receptor function of Pit2, Pit1, control for processing and folding of mutant proteins out transport function. two

Inorganic phosphate (P i ) 1 is essential for cellular metabolism and skeletal mineralization. Moreover, it serves as the source of phosphate for organic cell constituents, e.g. nucleotides and a variety of phosphorylated metabolic intermediates. Two proteins that show the same transport characteristics as P i uptake across the plasma membrane in animal cells have been identified (1)(2)(3), namely the sodium-dependent phosphate (NaP i ) cotransporters, Pit1 (human Pit1 formerly GLVR1 (4)) and Pit2 (human Pit2 formerly GLVR2 (5)). Both proteins are characterized as type III NaP i cotransporters (6) and show a broad tissue distribution being expressed in all investigated human tissues albeit at different levels (7). Furthermore, low extracellular P i levels result in up-regulated Pit1 and Pit2 expression in mammalian cells (1,8). These observations strongly suggest that the major cellular P i demand in mammalian cells is handled by type III NaP i cotransporters (1). However, recent results also point at type III transporters as playing specific roles in chondroblastic and osteoblastic mineralization (9, 10) as well as being critically involved in vascular calcification under hyperphosphatemic conditions, which are often present in diabetic patients and individuals with renal failure (11). The mechanisms underlying the bone-forming roles of type III NaP i transporters are presently not known. Recent results, however, showed that high extracellular P i levels can induce expression of the gene for osteopontin and that the induction is dependent on Na ϩ -dependent P i uptake across the plasma membrane (12). Interestingly, osteopontin is involved in normal bone development and present in calcified arterial plaques (13).
Despite the important roles of type III NaP i cotransporters in cellular P i uptake and increasing evidence for their critical roles in normal and pathologic calcification, nothing is known about what determines their transport function. The two known type III NaP i cotransporters, Pit1 and Pit2, show about 60% amino acid identity (5), whereas orthologs of Pit1 and Pit2 exhibit more than 90% amino acid identity (5,14,15). A putative topological model for both proteins based on hydropathy plots predicts 10 transmembrane (TM) regions and 5 extracellular loops (5,14,16). No significant overall sequence similarity exists between type III cotransporters and members of the two other NaP i cotransport systems referred to as types I and II (1).
Both Pit1 and Pit2 were originally identified as receptors for retroviruses belonging to the gammaretrovirus genus, and they are currently the two cellular receptors targeted in human gene therapy trials employing retroviral vectors. The human gene SLC20A1 encoding Pit1 was cloned as a receptor for gibbon ape leukemia virus (4) and feline leukemia virus subgroup B (17), and the Pit2-encoding gene SLC20A2 from human and rats was cloned as receptor for amphotropic murine leukemia virus (AM-MLV) (5,18); an AM-MLV-related isolate, 10A1 MLV, utilizes both transporters as receptor (19). In recent years, substantial insight into the receptor functions of Pit1 and Pit2 for their cognate viruses has been obtained (reviewed in Refs. 20 and 21).
We speculated that the retroviral receptor function might provide a functional assay for proper processing and folding of mutant type III NaP i transporters with knocked out NaP i transport function thus allowing for identification of amino acids critical for transport function. We here show that it is possible to uncouple human Pit2 retroviral receptor function from NaP i transport function and to exploit the dual function of Pit2 as NaP i transporter and retroviral receptor for identification of amino acids critical for NaP i transport. Doing this, we identified two Pit2 glutamate residues Glu 55 (E55) and Glu 575 (E575) as critical for Na ϩ -dependent P i transport function. The glutamates are positioned in putative transmembranic domains and are conserved in type III NaP i cotransporters, in NaP i cotransporters PHO-4 from Neurospora crassa (22) and Pho89 from Saccharomyces cerevisiae (23), and in a number of known and putative phosphate permeases from other species (see on-line supplemental material) strongly suggesting that they also are critical for the physiological function of these transporters.

Construction of Expression Plasmids and Mutagenesis-
The pcDNA1A R tkpA-derived expression plasmid pOJ74 (Wyeth-Ayerst Research, Pearl River, NY) encoding human Pit2 has been described previously (24). Mutations were introduced in human Pit2 using the Altered sites II kit (Promega) and p-ALTER-1-derived vectors harboring the 5Ј-and 3Ј-ends of the SLC20A2 gene. Mutated fragments were moved from pALTER to pOJ74 with appropriate restriction enzymes, and their authenticities were confirmed by sequencing.
Cell Culture-AM-MLV (4070A) and 10A1 pseudotypes of the ␤-galactosidase-encoding vector G1BgSvN (25) were obtained from the producer cell lines PA317GBN and PT67GBN, respectively. PA317GBN was derived from PA317 cells (26) and was a kind gift from M. V. Eiden. The PT67GBN producer cell line was made by transducing PT67 cells (27) with a retroviral vector carrying the G1BgSvN transfer vector and selecting for a high producer clone. PA317GBN cells, Chinese hamster ovary (CHO) cells (ATCC CCL-61), and dog osteosarcoma (D17) cells (ATCC CCL-183) were cultivated as described (24); PT67GBN cells were cultivated as PA317GBN cells. Filtered (0.45-m pore size) vector supernatants were stored at Ϫ80°C.
Transient Transfection and Infection Assay-Transient transfectioninfection assays were performed essentially as described previously (24). CHO cells were seeded at 8 ϫ 10 4 cells/60-mm-diameter dish and transfected the following day using calcium phosphate-DNA precipitates. Each precipitate contained 10 g of a CsCl-purified expression plasmid and 5 g of CsCl-purified pUC19 plasmid as a carrier in 1 ml, and aliquots of 200 l, corresponding to 2 g of expression plasmid, were added per dish. Three independent precipitates were made per construct to be tested. Forty-eight hours after transfection, 1.5 ml of supernatants (undiluted or diluted to give the indicated vector loads) from PA317GBN and PT67GBN, adjusted to 8 g of Polybrene per ml, was added per dish. PT67GBN supernatants were always diluted 1:1 with CHO-conditioned medium to specifically block the endogenous 10A1 receptors on CHO cells (19). Four hours later, 4.5 ml of fresh media or CHO-conditioned media (10A1 vector-exposed cells) were added, and the cells were incubated for another 2 days, at which time they were fixed and stained for ␤-galactosidase activity as described (24). The number of ␤-galactosidase-positive (blue) cells per dish was counted under a light microscope. Numbers of vector pseudotypes used in the experiments were calculated from the number of blue-forming colonies per ml obtained on D17 cells as described (24). Detailed information on this method and appropriate controls can be provided upon request. 2.4 mM NaHCO 3 , 10 mM HEPES-Tris, pH 7.5, 100 IU/ml penicillin, 100 g/ml streptomycin) containing 5% dialyzed fetal bovine serum at 18°C, until they were used for injection the following day.
CsCl-purified DNA preparations used in receptor studies were linearized with NotI and m 7 G (5Ј) ppp (5Ј) G-capped complementary RNAs (cRNAs) were prepared using the T7 mMESSAGE mMACHINE kit (Ambion). cRNA (25 ng) or H 2 O in volumes of 50.6 nl were injected into the cytoplasm of oocytes, and these were incubated 2 days at 18°C in MBS with 5% fetal bovine serum and 1 day at 18°C in MBS before being used for uptake studies.
Phosphate uptake was measured using a protocol modified from Tatsumi et al. (28). For NaP i uptake measurements, oocytes were incubated briefly 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) followed by a 60-min incubation in uptake solution containing 0.1 mM KH 2 32 PO 4 (1 Ci/mmol, PerkinElmer Life Sciences) at 18°C. After uptake, oocytes were washed in ice-cold uptake solution containing 5 mM KH 2 PO 4 , and each was transferred to a scintillation vial containing 0.2 ml of 10% (w/v) SDS. The radioactive decay was determined after addition of 5 ml of Optiphase Polysafe scintillation liquid (Wallac, Finland). For cationdependent 32 P i transport (see Table I), NaCl was substituted for 100 mM LiCl in the uptake solution.
Surface Biotinylation of Xenopus laevis Oocytes-For investigation of the presence of wild-type and mutant proteins in the outer membrane of oocytes, five cRNA-injected oocytes were washed three times in ice-cold PBS buffer at pH 8.0. The oocytes were then incubated 30 min at room temperature in 0.5 ml of PBS containing 0.5 mg/ml sulfo-NHS-SS-biotin (Pierce), and excess biotinylation reagent was removed by four washes with ice-cold PBS. The oocytes were subsequently dissolved in lysis buffer (150 mM NaCl, 20 mM Tris-HCl, pH 8.0, 1% Triton X-100) for 30 min on ice. Solubilized oocytes were centrifuged for 15 min at 13,000 rpm, and the supernatants containing the biotinylated proteins were incubated with 50 l of streptavidin-coupled Dynabeads (Dynal) for 1 h at 4°C. The bead complexes were washed four times by suspension in ice-cold lysis-buffer, and the final pellets were resuspended in a 1/5 volume of 5ϫ sample buffer (125 mM Tris-HCl, pH 6.8, 5% SDS, 12.5% ␤-mercaptoethanol, 50% Glycerol, 0.25% Bromphenol blue) and incubated for 8 min at room temperature. The samples were analyzed by SDS-PAGE (4 -15% acrylamide gel) followed by Western blotting to a polyvinylidene difluoride membrane (Immobilon-P, Millipore). The blot was probed with an antibody raised against the large intracellular domain between the sixth and seventh TM regions of human Pit2 made in a rabbit (kind gift from Dr. V. Garcia) (16) and developed using swine anti-rabbit immunoglobulin conjugated to horseradish peroxidase (Dako) and chemiluminescence according to the manufacturer's protocol (ECL Western blot kit, Amersham Biosciences).
Statistical Analysis-32 P i uptake values and receptor functions were expressed as means Ϯ S.E. The hypothesis that two mean values were identical was tested by a two-tailed Student's t test; values were considered different at a 95% confidence level.

RESULTS
Candidate Residues for Mutagenesis-The objective of the present study was to identify amino acids critical for human Pit2 NaP i transport function and investigate whether it was possible to knock out NaP i transport function leaving retroviral receptor function undisturbed. Conserved acidic residues in TM regions are prominent candidates for amino acids liganding monovalent cations (29 -33) and, as such, are potentially critical for NaP i transport function. Moreover, transmembranic amino acids might be less likely to directly interact with viral proteins involved in entry compared with amino acids positioned in extracellular loops. According to the prevailing topological model of Pit1 and Pit2 (5,14,16), two transmembranic glutamate residues are widely conserved in type III transporters and in related proteins from a number of species (see on-line supplemental material). They are positioned in the putative second (E55, human Pit2 numeration) and ninth (E575, human Pit2 numeration) TM regions of Pit1 and Pit2 proteins. We analyzed the possible roles of E55 and E575 in human Pit2 transport function by initially replacing them with glutamine. Moreover, in these studies we also included a non-conserved glutamate residue only present in Pit2 orthologs (Glu 91 (E91) in the putative third TM region of human Pit2) (see supplemental material). Mutant Pit2 proteins are designated with a threeletter code, where the first, second, and third letters show the identity of the amino acids in positions 55, 91, and 575, respectively; thus, wild-type Pit2 would be designated EEE.
Exchange of Glutamate Residues E55 and E575 with Glutamine Severely Impaired NaP i Transport of Human Pit2-The Na ϩ -dependent 32 P i transport function of wild-type Pit2 and the seven Pit2 mutants QEE, EQE, EEQ, QQE, QEQ, EQQ, and QQQ were analyzed in Xenopus oocytes (Fig. 1). In the experiment shown in Fig. 1A, oocytes injected with cRNA encoding human Pit2 supported a 32 P i uptake of 521 Ϯ 36 pmol/h per oocyte at pH 7.5 or about 25-fold above the 32 P i uptake in H 2 O-injected oocytes (p Ͻ 0.001). Depending on the oocyte preparation, human Pit2 supported 32 P i uptake varied from about 250 to 500 pmol/h per oocyte (Figs. 1A and 2A and Table  I) (unpublished data).
The 32 P i transport activity of the EQE mutant was indistinguishable from that of Pit2 (p Ͼ 0.8) (Fig. 1A). However, exchanging the highly conserved 2.TM or 9.TM glutamate residues with glutamine (QEE and EEQ) resulted in 4-and 7-fold decreases in 32 P i uptake, respectively, compared with Pit2 (p Ͻ 0.001). Introducing the E91Q mutation together with the E55Q mutation (QQE) further reduced the 32 P i uptake (compared with QEE p Ͻ 0.001), whereas introducing the E91Q mutation together with the E575Q mutation (EQQ) led to an increased 32 P i uptake (compared with EEQ p Ͻ 0.001) (Fig. 1A). Noticeably, exchanging both the highly conserved E55 and E575 residues with glutamine reduced the 32 P i uptake level ϳ10-(QEQ) and 8-fold (QQQ) compared with Pit2 (p Ͻ 0.001) (Fig.  1A). In the experiment shown in Fig. 1A, all mutants did support 32 P i uptake levels significantly different from that of H 2 O-injected oocytes (p Ͻ 0.002). The same relative Na ϩ -dependent 32 P i uptake pattern was found in an independent experiment (unpublished data); however, in this experiment as well as in a separate experiment (Table I) the 32 P i uptake levels supported by QEQ or QQQ did not differ significantly from that of H 2 O-injected oocytes. Thus, the QEQ and QQQ mutants do indeed have very low P i transport activities.
The cell surface expression of the QEQ and QQQ mutants was assessed by cell surface biotinylation using the membraneimpermeant reagent, sulfo-NHS-SS-biotin, followed by SDS-PAGE and Western blotting (Fig. 3). The predominant form of the membrane-associated Pit2 and mutant proteins was dimeric (marked with arrowhead in Fig. 3) in agreement with recent results from Salaü n and coworkers (36), which suggest that the active transporting form of human Pit2 is a dimer. The QEQ and QQQ mutants were present at the oocyte cell surface in approximately half the amounts of Pit2 and in amounts comparable to that of the mutant EQE, which shows wild-type transport function (Fig. 3). Thus the severely impaired transport function of the QEQ and QQQ mutants cannot solely be explained by impaired processing of the mutant proteins to the oocyte cell surface. The presence of the mutant proteins in the outer membrane of the oocyte does, however, not exclude that the impaired transport functions are due to improper folding of the mutant proteins. We therefore investigated whether we could exploit the additional function of Pit2 as retroviral receptor to assess proper processing and folding of the mutant proteins in mammalian cells.

Viral Receptor Function of Mutant Pit2 Proteins with Gln for Glu Substitutions-Using a transient transfection-infection as-
say, we analyzed whether the neutral conserved Gln for Glu substitutions, which in some mutants severely impaired transport activity, affected viral receptor functions of the Pit2 mutants. Expression plasmids encoding Pit2 and mutant proteins were transfected into CHO cells, which are non-permissive and largely non-permissive for infection by AM-MLV and 10A1, respectively (Fig. 1, B-D) (19,37,38). The ability of these proteins to support infection by AM-MLV and 10A1 was analyzed by exposing the transfected cells to retroviral vectors encoding ␤-galactosidase and carrying viral surface proteins responsible for receptor recognition; vectors carrying AM-MLV and 10A1 surface proteins are referred to as AM-MLV and 10A1 vector pseudotypes, respectively. The infection levels were evaluated as the number of ␤-galactosidase-positive (blue) cells per 60-mm-diameter dish. Dishes with CHO cells transfected with the same precipitates were chosen at random and exposed to either 20,000 AM-MLV vector pseudotypes (Fig.  1B), 85,000 10A1 vector pseudotypes (Fig. 1C), or 850,000 10A1 vector pseudotypes (Fig. 1D). CHO cells expressing human Pit2 were permissive for infection by both vector pseudotypes   3. Presence of mutant proteins at oocyte cell surface. X. laevis oocytes injected with cRNAs encoding the indicated mutants proteins and Pit2 or water were biotinylated using the membraneimpermeant reagent sulfo-NHS-SS-biotin. Equal amounts of purified biotinylated protein fractions were loaded and separated by SDS-PAGE. The proteins were transferred to polyvinylidene difluoridemembrane and visualized by an antibody toward the large intracellular hydrophilic domain of Pit2. The predominant form of the proteins is dimeric (arrowhead); a fragment of ϳ40 kDa was also observed as previously found in mammalian tissue preparations (34,35). The data were digitized on a flatbed scanner, and the bands representing dimeric proteins were quantified using Bio-Rad Quantity One software; the ratios of mutant proteins to Pit2 are as follows. EQE, 0.4; QEQ, 0.7; QQQ, 0.5; KEE, 1.0; EEK, 1.1; KEK, 0.9. (Fig. 1, B-D) in agreement with the well described receptor function of Pit2 for AM-MLV and 10A1 (5,19,24,38). Increasing the 10A1 vector load 10-fold resulted in an increase of only about 4-fold in the number of infected cells in Pit2-expressing cultures, showing that, with this combination of Pit2 protein amount and virus load, we were approaching saturating conditions for infection (Fig. 1, compare C and D legends). The mutants QEE, EQE, and QEQ supported infection with both viruses at wild-type receptor levels (Fig. 1, B-D) and behaved as Pit2 upon increased 10A1 vector load confirming their wildtype receptor functions. The remaining mutants EEQ, QQE, EQQ, and QQQ exhibited slightly impaired receptor functions for both pseudotypes (Fig. 1, B and C). Upon increasing the 10A1 vector load, the infection levels obtained with mutants EEQ, QQE, and QQQ furthermore decreased compared with that obtained with Pit2, reaching levels of ϳ35%, 48%, and 43% of Pit2, respectively, although only those of EEQ and possibly QQQ were significantly impaired (EEQ, p ϭ 0.05; QQE, p ϭ 0.1; and QQQ, p ϭ 0.06 compared with Pit2) (Fig. 1D).
The validity of the conditions employed in the transfectioninfection experiment shown in Fig. 1 was further confirmed using the mutants QQQ and QEQ, which showed impaired and wild-type receptor functions, respectively. This was done by expressing high and low receptor amounts and using high and low 10A1 vector loads (unpublished data). For analysis of QQQ receptor function, Pit2 infection levels ranging from no detectable infection to ϳ40,000 blue cells per 60-mm-diameter dish were employed. At low vector load, higher amounts of QQQ encoding plasmid DNA, compared with Pit2 encoding plasmid DNA, were needed to obtain 10A1 infection. Moreover, the infection via the QQQ mutant never reached the levels obtained via Pit2, and it exhibited significantly impaired 10A1 receptor function compared with Pit2 (p Ͻ 0.05) (unpublished data). In an independent experiment, the infection levels obtained with QEQ and Pit2 ranged from ϳ600 to 20,000 (QEQ) and 500 to 15,000 (Pit2) blue cells per 60-mm-diameter dish, respectively, again confirming that QEQ was a fully functional 10A1 receptor.
In summary, these results strongly indicate that it is possible to uncouple transport function from retroviral receptor function in human Pit2. Thus, although some of the mutants carrying Gln for Glu substitutions showed severely impaired Na ϩ -dependent 32 P i transport function, all retained receptor function albeit some about 2-fold impaired compared with Pit2. Specifically, exchanging the highly conserved E55 and E575 in Pit2 with glutamines, mutant QEQ, had no influence on receptor function, whereas the Na ϩ -dependent 32 P i uptake was severely impaired indicating a role of E55 and E575 in Pit2 NaP i transport. On the other hand, the mutant EQE showed wildtype transport and receptor functions, thus no pivotal role in Pit2 transport function can be assigned to the Pit2-specific E91 residue alone.
Cation-dependent P i Uptake of Pit2 Mutant Proteins Carrying Glutamine Instead of the Highly Conserved E55 and E575-The observation that the QEQ and QQQ mutants showed severely impaired NaP i transport function but retained viral receptor function, although that of QQQ was impaired, led us to investigate whether QEQ and QQQ exhibited a different cation-dependent P i transport activity than Pit2 (Table I).
Oocytes injected with human Pit2 encoding cRNA were found to support Li ϩ -dependent 32 P i uptake (compared with H 2 Oinjected oocytes p Ͻ 0.05) (Table I), in agreement with the fact that hamster Pit2 has been shown to sustain Li ϩ -dependent 32 P i uptake when overexpressed in hamster cells (3). The QQQ mutant exhibited severely impaired Li ϩ -dependent 32 P i transport, whereas no Li ϩ -dependent 32 P i uptake above background could be detected in oocytes expressing the QEQ mutant (Table  I). Thus, because QEQ showed wild-type receptor function, these results suggest that E55 and E575 in general are important for cation-dependent P i transport of Pit2.
Exchange of Either of the Glutamate Residues in Positions 55 and 575 of Pit2 for Lysine Knocked Out NaP i Transport of Human Pit2-The observation that the QEQ mutant harboring E55Q and E575Q substitutions had retained wild-type viral receptor functions encouraged us to replace these amino acids with lysine, in an attempt to completely knock out Na ϩ -dependent P i transport function of Pit2. Thus glutamate was replaced with lysine in positions 55 and/or 575 in wild-type Pit2 resulting in the mutants KEE, EEK, and KEK (Fig. 2). No increase in Na ϩ -dependent 32 P i uptake could be detected in oocytes injected with cRNAs encoding Pit2 with E55 and/or E575 exchanged for lysine compared with H 2 O-injected oocytes ( Fig. 2A). All three mutant proteins were present at the oocyte cell surface in amounts comparable to wild-type Pit2 (Fig. 3), thus their lack of transport activity is not due to impaired processing to the oocyte cell surface. However, we introduced rather dramatic changes in these proteins, and, as mentioned above, the presence of the mutant proteins in the outer membrane of the oocyte does not exclude that their impaired transport functions are due to improper folding. We therefore investigated whether the additional function of Pit2 as retroviral receptor could be employed to assess proper processing and folding of these mutant proteins in mammalian cells.
Pit2 Proteins Carrying E55K and/or E575K Substitutions Retained Viral Receptor Function-The effect of E55K and E575K substitutions on Pit2 AM-MLV and 10A1 receptor function is shown in Fig. 2 (B and C), respectively; notice that the infection levels in this experiment were in general higher than those shown in Fig. 1 (B-D) due to a more efficient batch of transfection buffer. All mutants exhibited AM-MLV receptor function at wild-type Pit2 levels (for all mutants p Ͼ 0.1 compared with Pit2) (Fig. 2B). Similar results were obtained for the 10A1 receptor functions of the mutants (for all mutants p Ͼ 0.1 compared with Pit2) (Fig. 2C). However, although not significantly different from the receptor function of Pit2, the mutant KEE supported infection levels of about 60% of that obtained with Pit2 (Fig. 2C). We therefore chose to subject the 10A1 receptor functions of KEE to a more detailed analysis with independent DNA preparations, using high and low receptor amounts and high and low 10A1 vector loads. In this experiment, the Pit2 infection levels ranged from 300 to 14,000 blue cells per 60-mm-diameter dish, and no impairment in receptor function of the mutant protein could be observed (unpublished data), in agreement with the data in Fig. 2C. We conclude that the KEE, EEK, and KEK mutants have knocked out P i transport functions, while exhibiting wild-type receptor functions; thus, in these mutants, the receptor function is uncoupled from the transport function. Moreover, the observation that introduction of lysine instead of either E55 or E575 knocked out human Pit2 Na ϩ -dependent P i transport function is in agreement with the results obtained on Pit2 mutant proteins harboring Gln for Glu substitutions, and we conclude that the two highly conserved Pit2 glutamate residues E55 and E575 are crucial determinants of human Pit2 P i transport function. DISCUSSION Type III NaP i cotransporters, Pit1 and Pit2, exhibit dual function. Here we have shown that it is possible to knock out P i transport function of human Pit2, leaving its retroviral receptor function undisturbed. Thus, it is possible to exploit the retroviral receptor function of Pit2, and possibly of Pit1, as a control for processing and folding of mutant proteins with knocked out transport function. Doing this, we identified two glutamate residues, E55 and E575 in human Pit2, as critical for Pit2 P i transport function. This is the first time amino acids critical for P i transport of a type III NaP i cotransporter have been identified. The two glutamates are both highly conserved in type III NaP i cotransporters and in a number of known and putative phosphate permeases from other species (see supplemental material) strongly suggesting that they also are critical for the physiological function of these transporters.
The strength of employing retroviral receptor function as a control for correct processing and folding of mutant Pit2 proteins rather than relying on protein expression in outer membranes is illustrated with the mutant QQQ. It exhibited severely impaired transport function, which could not be explained by the protein not being present in the oocyte outer membrane, but it did, however, also show significantly impaired receptor function. It is possible that impaired receptor function can be due to involvement in retroviral entry of the amino acids analyzed; however, in this case, it is not likely, because mutant proteins like QEE, EQE, QEQ, KEE, EEK, and KEK were fully functional receptors, rather we suggest that the QQQ mutant proteins in the plasma membrane exhibit disturbed processing and/or folding.
The question arises whether viral receptor function will be applicable as control when studying involvement of extracellular loop positioned amino acids in transport function of type III transporters. It is in this connection noteworthy that the amino acids so far identified as critical for retroviral receptor function of Pit1 and Pit2 orthologs are those highly variable between Pit1 and Pit2 and the NaP i transporter PHO-4 from N. crassa (19, 24, 37, 39, 40 -43), whereas the amino acids critical for NaP i uptake are expected to be found among the residues conserved in Pit1, Pit2, and related proteins from other species as shown in the present work. In line with this, we have recently shown that a Pit1 mutant with abolished receptor functions showed no impairment in P i transport function as analyzed in Xenopus oocytes. 2 Thus, it is possible that there in general is no overlap between Pit sequences critical for P i transport function and Pit sequences critical for receptor function.
We do not know the exact role of Pit2 residues E55 and E575 in P i transport. However, the observation that changing either E55 or E575 for lysine was sufficient to knock out P i transport function leaving receptor function undisturbed is indeed in agreement with these residues being involved in NaP i transport function, rather than fulfilling structural roles. It is possible that E55 and E575 are involved in function-dependent conformational changes of the transporter. Our data are, however, also consistent with these glutamates being parts of a cation liganding site created by residues from different transmembrane regions as has been implicated for negatively charged residues of the melbiose permease of Escherichia coli, the rabbit Na ϩ /dicarboxylate cotransporter NaDC-1, Na ϩ /H ϩ exchangers, the c subunit of the Propionigenium modestum F 1 /F 0 -ATP synthase and Na ϩ ,K ϩ -ATPases (29 -33, 44). However, a Na ϩ -liganding site may contain only one acidic residue, the other residues being, e.g. glutamine and serine (44). In this context it is worthy to note that at least some of the mutant Pit2 proteins with Gln for Glu substitutions in only one of the positions E55 or E575 (mutants QEE, EEQ, and EQQ) still supported low levels of NaP i uptake (Fig. 1A).
In recent years, a number of transport proteins have been identified that have dual function as transporters and retroviral receptors (4,5,18,(45)(46)(47)(48). For murine CAT1, which is a Na ϩ -dependent cationic amino acid transporter, it has also been shown that a glutamate residue in a putative transmembrane region was critical for the transport function but not for retroviral receptor function of the protein (49). Based on the data presented here and that of Wang et al. (49), we suggest that for proteins exhibiting dual function as transporters and retroviral receptors, their viral receptor function might in general be exploited for identification of amino acid residues critical for transport function. Using a transient transfection-infection assay, which is a fast and reliable method, for analyzing the viral receptor function as described here, indeed makes it a highly feasible approach to analyze processing/overall topology of mutant proteins by investigating their receptor function.
The presented data also have implications for understanding the role of Pit2 in retroviral entry. Addition of sulfhydryl reagents to cells was previously shown to impair Pit2 P i transport and AM-MLV infection but not virus binding (50). According to the authors, these data suggest that P i -induced conformational changes in Pit2 are involved in AM-MLV entry via Pit2 (50). Our data are not in direct conflict with this interpretation, but they do show that viral receptor function is not dependent on P i transport function per se. Indeed, comparison of the data in Figs. 1 and 2 reveals that there is no correlation between Na ϩ -dependent P i transport function and retroviral receptor function of Pit2. We suggest that further important insight into both functions of the type III NaP i transporters can be achieved by combining studies on transport and receptor functions as presented here.