Phosphate uptake by the phosphonate transport system PhnCDE

Background Phosphate is a fundamental nutrient for all creatures. It is thus not surprising that a single bacterium carries different transport systems for this molecule, each usually operating under different environmental conditions. The phosphonate transport system of E. coli K-12 is cryptic due to an 8 bp insertion in the phnE ORF. Results Here we report that an E. coli K-12 strain carrying the triple knockout ΔpitA Δpst Δugp reverted the phnE mutation when plated on complex medium containing phosphate as the main phosphorus source. It is also shown that PhnCDE takes up orthophosphate with transport kinetics compatible with that of the canonical transport system PitA and that Pi-uptake via PhnCDE is sufficient to enable bacterial growth. Ugp, a glycerol phosphate transporter, is unable to take up phosphate. Conclusions The phosphonate transport system, which is normally cryptic in E. coli laboratory strains is activated upon selection in rich medium and takes up orthophosphate in the absence of the two canonical phosphate-uptake systems. Based on these findings, the PhnCDE system can be considered a genuine phosphate transport system. Electronic supplementary material The online version of this article (10.1186/s12866-019-1445-3) contains supplementary material, which is available to authorized users.


Background
Phosphorus is a macronutrient of utmost importance to all living beings. It is thus not surprising that bacteria developed several different mechanisms of phosphorus acquisition. Particularly, the PHO regulon, a set of genes involved in the acquisition and metabolism of phosphorus containing molecules in response to phosphate (Pi)-shortage in the medium is present across bacterial phyla. The PHO regulon of E. coli comprises more than 30 genes [1]. The most well characterized genes of the PHO regulon are phoB-phoR, the operons pstSCAB-phoU, ugpBAEC, phnCDEFGHIJKLMNOP and phoA that respectively encode the two-component system that controls the transcription of the regulon, an ABC-type Pitransport system, an ABC-type glycerol phosphate transport system, a 14-gene operon involved in phosphonate (Pn) transport and assimilation and a periplasmic alkaline phosphatase (AP). Genes belonging to the PHO regulon are synchronously induced by Pi-shortage [2]. For the sake of simplicity, the operons pstSCAB-phoU, ugpBAEC, phnCDEFGHIJKLMNOP will be respectively shortened to pst, ugp and phn.
Most, but not all available phosphorus in nature is in its most oxidized state (+5), also known as phosphate. Phosphonates (P valence +3) are phosphorus-containing organic molecules in which the P atom is linked directly to C in a stable chemical bond. Though less common in nature than organic phosphates, many Pn molecules, such as the antibiotic fosfomycin and the herbicide glyphosate are of clinical and environmental importance [3,4]. The first three genes of the phn operon -phnCDE, encode an ABC-type Pn transport system. The polypeptide products of phnC, phnD and phnE respectively are an ATP-binding subunit, a periplasmic binding protein and a Pn permease. The next gene, phnF encodes, based on sequence similarity, a negative regulatory protein that represses Pn transport via PhnCDE. The remaining ten genes of the operon (phnGHIJKLMOP) code for enzymes involved in Pn catabolism, such as a C-P lyase complex.
Owe to the presence of an 8 bp insertion in phnE E. coli K-12 strains do not express the PhnE permease and consequently are unable to take up phosphonates [5]. However, phnE + revertants that arise by spontaneous slippage [5,6] are relatively easy to select by growing bacteria with Pn as the sole phosphorus source or by growing bacteria in Pilimited medium for several days [5][6][7]. The vast majority of E. coli strains and isolates carry a functional phnE gene.
E. coli possesses three Pi-transport systems -PstSCAB (Pst), PitA and PitB. Pst is an ABC-type high-affinity system formed by four proteins: the Pi-binding periplasmic protein PstS, the integral membrane proteins PstC and PstA, and the ATPase PstB. The pst operon belongs to the PHO regulon and responds to Pi-limitation. In addition to its role in Pi uptake, the Pst system also acts as a repressor of the PHO regulon under Pi-excess conditions [8].
PitA and PitB are metal-phosphate symporters that share 81% sequence identity. The former is likely constitutive, while PitB is inhibited by the PHO regulon [9] and is not functional under Pi-starvation or in a PHOconstitutive background, i.e., in a pst mutant [9][10][11].
Based on the ability to grow in media containing different phosphorus sources, Metcalf and Wanner [12] suggested that Phn is able to take up Pi, phosphite and Pi-esters. However, the bacterial strains used in that study were not fully characterized and the role of the PhnCDE system in Pi transport was not further confirmed.
By characterizing a spontaneous Phn + revertant in E. coli K-12 carrying knockouts in genes related to Pi transport we unequivocally show that Phn has the ability to take up Pi in the absence of canonical Pi-transport systems and to support the growth of bacteria lacking another functional Pi-transport system. It is also shown that despite previous suggestions, the Ugp system is unable to transport Pi.

Selection of a phnE + revertant in E. coli K-12
PHO-constitutive mutants can be isolated by plating wildtype bacteria on glycerol-2-phosphate (G2P) as the sole C source. We have previously observed that the frequency of such mutants in bacteria carrying both pitA and ugp knockouts is extremely low [13]. This is because as most PHO-constitutive mutations occur in one of the five genes of the pst operon, a pitA pst strain would lack a functional Pi-uptake system and will thus not be able to grow with Pi as a phosphorus source. Furthermore, most G2P, which could in principle be utilized as a Pi source [14], is hydrolyzed to glycerol and Pi in the periplasm by the constitutively expressed AP, while the remaining intact G2P cannot be taken up due to the absence of Ugp or another G2P-transport system. Nevertheless, a small number of pitA ugp spontaneous pst mutants could be isolated on G2P, suggesting that the triple mutant acquired a fourth mutation that enabled growth in this medium.
To further investigate this matter, we attempted constructing a pitA ugp pst triple mutant in the wildtype K-12 strain MG1655 by means of recombineering and transduction. Individual deletions obtained by λred mediated recombineering were transduced one after another into MG1655 (see Methods). While all pairwise combinations could be easily obtained, attempts to originate the triple mutant were mostly unsuccessful. In one of these attempts, pst was transduced into the pitA ugp double mutant and the selective plate (L-agar containing kanamycin) was inadvertently kept at 37 • C for 72 h. A single colony emerged after 72 h which, once isolated was able to grow in LB and in minimal medium containing Pi as the sole P source. Genome sequencing of this bacterium (strain RI57) revealed the deletion of 8 bp (5'-GCTGGCGT) at position 407-414 of phnE ORF.
In addition to the three knockouts ( pst pitA ugp) and the phnE 8 bp deletion, other point mutations in the genome of strain RI57 were observed ( Table 1). One of them is a G insertion in glpR ORF, which encodes the repressor of the glycerol-3-phosphate regulon [15]. Some variants of strain MG1655 are glpRwhile others, like the one used in this study carry a wild-type copy of this gene [16]. With the exception of the 1 bp insertion in glpR none of these substitutions is apparently connected to the PHO regulon or to Pi metabolism. The 8 bp deletion in phnE was subsequently confirmed by Sanger sequencing.

PhnCDE supports growth with Pi in the absence of canonical Pi-transport systems
The ability of the triple knockout pitA pst ugp phnE + (strain RI57) to grow in LB medium that contains 2.2 mM Pi as the main phosphorus source [17] suggests that the PhnCDE system might be involved in the uptake of Pi in the absence of a viable Pi-transport system. To investigate this assumption a set of bacterial constructs was generated. First, the phnE + allele was transduced from BL21, an E. coli B strain to MG1655 (strain RS03), then the pitA, ugp and pst deletions were sequentially transduced to MG1655 phnE + originating strain RS07 (phn + 3 ). In parallel, a phnCDE deletion was constructed and transferred to each one of the following double mutants: ugp pst, pitA pst and pitA ugp resulting in strains RS04 (pitA + 3 = pitA + pst ugp phn), RS05 (ugp + 3 = ugp + pitA pst phn) and RS06 (pst + 3 = pst + pitA ugp phn), respectively.
The new constructs were tested for overnight growth in minimal medium with 1 mM KH 2 PO 4 as the sole phosphorus source (Fig. 1a). As expected, bacteria carrying either PitA or Pst grew almost as well as the wild-type strain. In the absence of both PitA and Pst, only the strains that possessed a functional PhnCDE system managed to grow. This includes the spontaneous pitA ugp pst phnE + revertant (strain RI57) and the phnE + 3 transductant (strain RS07). The ugp + 3 triple mutant (strain RS05) carrying a functional Ugp system did not grow in this medium. To further investigate the ability of the PhnCDE system in supporting growth in the presence of Pi as the only phosphorus source, the phn + 3 triple knockout (strain RS07) and the wild-type strain MG1655 were set to grow in medium TGP (1 mM Pi) for 24 h (Fig. 1b). It can be observed that the growth rate of the phn + triple mutant at the exponential phase (μ = 0.52 h -1 ) was as fast as that of the wild-type strain (μ = 0.46 h -1 ), suggesting that PhnCDE can take up Pi at a rate that enables maximal growth rate.
The only triple deletion that did not grow in medium TGP was strain RS05 (ugp + 3 ), which suggests that the Ugp system is unable to transport Pi. To exclude the possibility that the ugp operon in this strain has acquired a spontaneous null mutation, the wild-type strain and the triple mutant ugp + 3 were grown in minimal medium supplemented with either Pi (TGP) or G3P (TGG3P) as phosphorus sources. Figure 1c shows that the growth yield of strain RS05 (ugp + 3 ) in medium TGG3P was slightly higher than that of the wild-type strain, while in the TGP medium the ugp + 3 mutant grew very poorly. This indicates that strain RS05 (ugp + 3 ) carries a functional Ugp system.
We also investigated whether the pitB gene, that encodes a PitA-like Pi-transporter that is normally non-functional in wild-type E. coli or in a PHO-constitutive mutant [9] was spontaneously activated enabling the growth of the phn + 3 triple mutant (strain RS07) in TGP. To test this possibility the pitB gene was deleted from strain phn + 3 resulting in the quadruple mutant phn + pitA pitB pst ugp (strain RS08 = phn + 4 ). Figure 1d shows that both strain RS07 (phn + 3 ) and RS08 (phn + 4 ) grew in medium TGP, suggesting that PitB is not functional and confirming that the PhnCDE system is able to support growth with Pi as a phosphorus source.

The PhnCDE system transports Pi
Next, Pi uptake assays were performed in the wildtype strain MG1655, and in the triple knockouts phn + 3 (strain RS07) and ugp + 3 (strain RS05). Bacteria grown in medium TGP or TGG3P to an OD 600 of 1.0 were washed to remove any phosphorus leftovers and exposed to 0.2 mM 32 P(KH 2 PO 4 ). Figure 2 shows that the rate of Pi uptake in strain RS07 ( phn + 3 ) was similar to that of the wild-type strain, while the ugp + 3 (strain RS05) failed to take up significant amounts of Pi. To further characterize the PhnCDE system the kinetic parameters of Pi uptake via Phn and PitA were obtained. The triple mutants phn + 3 (RS07) and pitA + 3 (RS04) were exposed to increasing concentrations of 32 Pi and assayed for Pi- uptake. The kinetic parameters of Pi-uptake for strains RS07 (phn + 3 ) and RS04 (pitA + 3 ) were calculated from a Lineweaver-Burk double reciprocal plot (Additional file 1: Figure S1). The uptake of Pi by PhnCDE was also validated by competing Pi with aminoethylphosphonate (AePn). Bacteria were exposed to 10 μM 32 Pi and 50 s later an excess concentration of AePn (4 mM) was added. Additional file 1: Figure S2 shows that addition of AePn strongly inhibited Pi uptake when compared to the control which received no AePn. Lower concentrations of AePn resulted in less transport inhibition (data not shown).
Finally, we assayed Pi consumption throughout a bacterial growth curve. Strains MG1655 and RS07 (phn + 3 ) were grown in medium TGP containing 1 mM Pi as the sole phosphorus source. At 1 h intervals, samples were taken for growth assessment (OD 600 ) and residual Pi in the medium. Additional file 1: Figure S3 shows that the patterns of growth and Pi-consumption were very similar in both strains throughout the entire curve. In both cases, almost all Pi was exhausted after 24 h. These results indicate that Pi uptake via the PhnCDE system continues for several hours as efficiently as in the wild-type strain.

Discussion
The data presented here show that in the absence of a canonical Pi-transport system, PhnCDE confers on bacteria the ability to take up Pi and to grow as fast as the wild-type strain. The functional redundancy of three different Pi-transport systems (PitA, Pst and Phn) attests for the importance of Pi as a vital macronutrient. However, it should be noticed that each one of the aforementioned transport systems play, at least one additional role or function. PitA is a symporter of Zn 2+ and other divalent cations [19] and Pst is a repressor of the PHO regulon [8]. Thus it is not surprising that Phn possesses more than one physiological role, namely transport of phosphonates, Pi and perhaps of other phosphorus-containing molecules [12]. The PhnCDE transport system in E. coli K-12 is normally cryptic, due to an 8 bp insertion in phnE ORF [5]. Deletion of the 8 bp insertion restores the reading frame and activates phosphonate transport via PhnCDE. Thus the 8 bp deletion observed in our pitA pst ugp transductant was likely to cause the reactivation of PhnCDE. Selection of phnE + revertants has been reported before in bacteria plated on minimal medium with phosphonate as the sole phosphorus source [5,6] and by serial transfer of a wild-type strain growing in low Pi minimal medium [7,20].
Of the four tested transport systems, only Ugp was unable to support growth and take up Pi. In a previous study, we have shown that the emergence of PHO-constitutive mutants on plates containing glycerol-2-phosphate (G2P) as the sole carbon source partially depends on the presence of Ugp [13]. The hypotheses raised to explain this phenomenon were that Ugp was required for the uptake of either G2P or Pi. The finding that Ugp is unable to take up Pi indicates that the contribution of Ugp to the emergence of PHO-constitutive mutants is likely to be associated with the uptake of G2P [14]. It should be noted, however, that the ability of Ugp in transporting G2P is still unsettled, as genetic evidences showed that Ugp transports G2P [14], while in vitro, UgpB, the periplasmic binding protein, did not bind this molecule [21].
An early report suggested that E. coli's PhnCDE is able to take up Pi [12]. This assertion was based on the growth of strain BW4794, which when plated on phosphite or phosphonate acquired the ability to grow on both phosphorus sources and also on Pi. However BW4794 was a poorly defined mutant that carried a partial pst deletion and another undefined linked mutation. The fact that this strain was not clearly shown to be pitA-negative adds uncertainty to the identity of the transport system used for Pi uptake. To the best of our knowledge, since that publication the ability of the PhnCDE system of E. coli to transport Pi was not further tested and confirmed. There were, however, some hints that the PhnCDE system might be involved in Pi transport. Rizk et al. [22] reported that E. coli PhnD (the periplasmic binding protein) displays a relatively small affinity towards Pi (Kd = 50 μM). The PhnDCE system of Mycobacterium smegmatis was shown to transport Pi but not phosphonates [23] and the PhoCDET Pi transport system of Sinorhizobium meliloti transports both Pi and phosphonates [24].

Conclusions
By sequentially deleting all known Pi transport-related genes we clearly demonstrated that Phn takes up Pi in sufficient amounts to support growth with Pi as the sole phosphorus source. It was also shown that reversion of the phnE mutation could be selected by plating the triple knockout pitA pst ugp on rich media, such as L-agar. This medium contains 2.2 mM free Pi [17] and possibly limiting concentrations of organic phosphates. Under these conditions, the triple knockout pitA pst ugp cannot grow unless by reverting the phnE insertion.

Gene and operon knockouts
Gene and operons were deleted using the λ-red recombinase system as originally described by Datsenko and Wanner [27] and Murphy et al. [28]. Briefly, the cat or kan resistance cassettes were amplified from plasmids pKD4 or pKD3, respectively, using the following pairs of DNA oligos: ugp350 and ugp4480 for deletion of ugpBAEC; pitA_mut_Fow and pitA_mut_Rev for deletion of pitA; pitB_mut_Fow and pitB_mut_Rev for deletion of pitB; pst_mut_Fow and pst_mut_Rev for deletion of the pstSCAB-phoU operon; and phn_mut_Fow and phn_mut_Fow for deletion of the phnCDE genes. The amplicons containing the cat or kan cassettes were electrotransformed in strains KM44 or KM32.
To induce the λ-red genes the bacterial cultures were supplemented with 1 mM of IPTG. Selection of recombinants was done by plating on L-agar supplemented with chloramphenicol or kanamycin. Knockouts of pitA, pitB, phn, pst and ugp were confirmed by PCR using primers pitA_ver_Forw/pitA_ver_Rev, pitB_ver_Forw/pitB_ver_Rev, phn_ver_Forw/phn_ver_Rev, pstS_S_Fow/phoU_S_Rev and ugp858/ugp52101 and, when applicable, by phenotypic assays, such as AP activity [29] or growth on phosphonates. The deletions were transferred to strain MG1655 by P1 transduction. The antibiotic resistance marker (kan or cat) was eliminated from the recombinant chromosomes by transformation with the temperature-sensitive plasmid pCP20 as described [27].

P1 transduction
Transfer of chromosomal markers using P1 transduction was performed as described [25]. Knockout transductants were selected on L-agar plates containing either kanamycin or chloramphenicol. Selection of the phnE + allele transductant (from strain BL21 to MG1655) was in minimal medium plates containing 2-aminoethylphosphonic acid as a phosphorus source.

Genome sequencing
Genomic DNA from strain RI57 was extracted using the Wizard Genomic DNA purification kit (Promega) following the manufacturer instructions. solutions were used to start the uptake assay. At several intervals after the addition of 32 P(KH 2 PO 4 ), 500 μl of the bacterial cultures were withdrawn, filtered under vacuum through a cellulose acetate filter (0.22 μm) and washed with 5 ml of a TP solution (T-salts + 10 mM KH 2 PO 4 ) to remove the unincorporated 32 Pi. The filters were then move to vials filled with scintillation cocktail (Optiphase Hisafe 3, Perkin Elmer) and the radioactive content of each sample was measured in a β-counter.

Growth curves and determination of Pi
The wild-type strain MG1655 and strain RS07 (phn + 3 ) were grown in medium TGP containing 1 mM KH 2 PO 4 . Samples were taken hourly for assessing the OD 600 and residual Pi in the medium. Pi concentration in the culture medium was measured as described [30]. The supernatants of the bacterial cultures were collected, diluted 100 times and mixed with the same volume of the working reagent (1 volume of 167 mM H 2 SO 4 , 1 volume of 2.5% (NH 4 ) 2 MoO 4 , 1 volume of 10% ascorbic acid and 2 volumes of deionized water) and incubated at 37 • C for two hours. Following the incubation period, samples were measured in a spectrophotometer at 820 nm. Calibration solutions were prepared with the following concentrations of NaH 2 PO 4 : 0.16 mM, 0.08 mM, 0.04 mM, 0.02 mM and 0.01 mM. Minimal medium containing glycerol-3-phosphate as a P source; TGP: Minimal medium containing phosphate as a P source;