Development of a Reagentless Biosensor for Inorganic Phosphate, Applicable over a Wide Concentration Range.

A fluorescent reagentless biosensor for inorganic phosphate (Pi), based on the E. coli PstS phosphate binding protein, was redesigned to allow measurements of higher Pi concentrations and at low, substoichiometric concentrations of biosensor. This was achieved by weakening Pi binding of the previous biosensor, and different approaches are described that could enable this change in properties. The readout, providing response to the Pi concentration, is delivered by tetramethylrhodamine fluorescence. In addition to two cysteine mutations for rhodamine labeling at positions 17 and 197, the final variant had an I76G mutation in the hinge region between the two lobes that make up the protein. Upon Pi binding, the lobes rotate on this hinge and the mutation on the hinge lowers affinity ∼200-fold, with a dissociation constant now in the tens to hundreds micromolar range, depending on solution conditions. The signal change on Pi binding was up to 9-fold, depending on pH. The suitability of the biosensor for steady-state ATPase assays was demonstrated with low biosensor usage and its advantage in ability to cope with Pi contamination.

Inorganic phosphate (Pi) is a byproduct of numerous reactions in the cell, including metabolic reactions like fatty acid metabolism, energy transducing ATPases and cell signaling, such as by GTPases and protein phosphatases. Therefore, considerable effort has been expended to develop means of measuring Pi as a generic way to monitor such reactions. Pi assays, using complex formation with molybdate are widely used,[1−3] although they are not continuous. Several coupled enzyme assays have been described, particularly using a phosphorylase. Examples include the use of a fluorescent substrate, such as 7-methylguanosine[4] and one with an absorbance change, 2-amino-6-mercapto-7-methylpurine ribonucleoside,[5] or using other coupled enzymes to produce an absorbance or fluorescence change, for example with Amplex Red.[6] Fluorescent reagentless biosensors provide an alternative method of assaying Pi: they are single molecular species that respond to the particular analyte of interest with a change in fluorescence.[7] This approach circumvents some of the complexities of coupled enzyme assays, for example, in which multiple species are required as additives in the assay mix. Reagentless biosensors require a minimum of recognition element, such as a binding protein, and a reporter, here a fluorophore, in the same molecule, so no extra components are required for measurements.

The periplasmic phosphate binding protein (PstS) from Escherichia coli, here abbreviated to PBP, is a highly specific phosphate scavenger that has been used previously as the recognition element for fluorescent biosensors for Pi.[8,9] Fluorophores, covalently bound to surface cysteines, were the reporters for Pi binding, responsive to Pi concentration in the medium. A diethylaminocoumarin-labeled version of the protein (MDCC-PBP) typically has a 7-fold signal change,[8] whereas a tetramethylrhodamine-labeled biosensor (rho-PBP) results in up to an 18-fold signal increase.[9] Both of these biosensors bind Pi at rates suggesting diffusion control and with dissociation constants in the nanomolar range. These have been particularly useful to measure Pi production or release in real time in a wide range of enzymatic systems in vitro, particularly in transient kinetic assays,[10−14] but have also been developed into other formats, such as attachment to an optical fiber.[15]

While this fast response and the tight binding of the original PBP-based biosensors enables the study of reactions that release Pi rapidly, these same properties also necessitate the sensor to be present at an excess over Pi, as essentially all Pi binds to the protein in almost all assay conditions. The biosensors typically measure Pi in tens of nanomolar to low micromolar range. The tight binding also renders them sensitive to even low levels of Pi contamination, which can be present in solutions or on surfaces and particularly occurs with such molecules as nucleotides.[16,17] The relatively high consumption of stoichiometric biosensors may also be a limiting factor for high-throughput assays and Pi contamination may require extra precautions or purifications of reagents used.

It is, therefore, desirable to have biosensors with different properties to expand the range of applications and types of assay. In particular, a variant of rho-PBP is described here with much lower affinity for Pi, but retaining a fluorescence change on Pi binding. This means that the biosensor can be used substoichiometrically relative to the Pi and that it can tolerate significant Pi contamination. In addition a biosensor with a lower affinity can be used for assays where higher micromolar amounts of Pi are released. This is a potential advantage in high-throughput and steady-state assays that are typically carried out over time scales of minutes.

Like the other members of the periplasmic binding protein superfamily, the phosphate binding protein is formed of two lobes linked by a hinge region and the Pi binding site is located in the cleft between these lobes.[18] The protein undergoes a large conformational change upon ligand binding including a bending motion of the two lobes around the hinge.[19] To produce the biosensor, fluorophores were attached to cysteines introduced on the surface so that their environment changes when this Pi-induced conformation change occurs.[9,20] In MDCC-PBP, a single diethylaminocoumarin was attached to one lobe at the top of the cleft and its interaction with the protein changes upon the conformation change. In rho-PBP, a tetramethylrhodamine (TMR) was attached on either side of the cleft such that in the apo form the rhodamine rings can stack, exhibiting very low fluorescence.[21] The Pi-induced conformation change causes partial disruption of the stacking interaction and the fluorescence increases.[9]

When introducing further mutations to weaken binding, it was important not to lose the fluorescence signal on Pi binding. In the rho-PBP construct, the main interactions of the rhodamines are with each other and not the protein, potentially making its fluorescence response less susceptible to changes to the protein. Because of this, together with its large fluorescence response and the high photostability of rhodamines, rho-PBP was chosen as the starting point to develop a weak-binding variant. As part of this development, several different strategies were explored to investigate where which parts of the PBP structure might be altered by mutation. Two approaches were to disrupt either the binding of the phosphate itself or the associated conformational change, namely cleft closure. A third approach was to mutate the hinge between the lobes. Protein variants of each type are described with their relative success. The final biosensor carries the I76G mutation in the hinge in addition to the A17C and A197C previously introduced for labeling with TMR.[9] Properties of this variant, including its interaction with Pi, phosphate analogues and nucleotides are described. Its useful range is characterized and an example assay to measure Pi production rate is shown.

Experimental Procedures

Materials

“Bacterial” purine nucleoside phosphorylase was obtained from Sigma and dissolved to 1000 U mL–1. PcrA helicase was purified as described.[22,23] Oligonucleotide dT35 and nucleotides were from Sigma. 6-Iodoacetamidotetramethylrhodamine (6-IATR) was a gift from Dr. J. Corrie (NIMR, London).[24,25]

Expression and Purification of Phosphate Binding Protein

PBP mutants were created in pET22b harboring the gene for mature E. coli (A17C,A197C)PBP between Nde1 and Xho1 sites in the MCS using a Quikchange site-directed mutagenesis kit (Stratagene) according to manufacturer’s instructions. A stop codon was inserted at the end of the PstS ORF, so that the encoded His6-tag was not added to the polypeptide chain. Plasmids were sequenced (GATC Biotech) to confirm the presence of the desired mutation(s). Previously PBP was expressed from the full gene, induced by Pi starvation, and so included the N-terminal signal peptide that was lost in the mature protein.[9,26] The pET22 vector described above produced a protein identical to the mature PBP except for an additional N-terminal methionine. It has an advantage of simple induction by IPTG. The amount of purified (A17C,A197C)PBP from this new construct was comparable to the previous method and typically 300 mg from 4 L of E. coli culture. An equivalent construct produced similar amounts of (A197C)PBP for MDCC labeling. Note that amino acid numbering is based on the natural, mature wild-type protein.

Plasmid pET22b carrying the desired mutations within the PstS ORF encoding PBP was transformed into BL21 (DE3) and used for preparing rho-PBP variants. An overnight culture was grown in LB medium containing 100 μg mL–1 ampicillin at 37 °C with aeration by vigorous shaking. This culture was diluted 50-fold into 500 mL aliquots of fresh medium and grown to an OD600 ∼ 0.8 before expression was induced with 500 μM IPTG. After 4 h induction, cells were harvested by 20 min centrifugation at 2500g and 4 °C. Cells were resuspended in 20 mM Tris-HCl pH 8.0 and stored at −80 °C.

For purification, cells from 500 mL culture were thawed and sonicated 4 × 30 s at 200 W with a 5 s on/off pulse cycle. The lysate was cleared by centrifugation at 142 000g and 4 °C for 45 min. A 5 mL HiTrapQ FF column (GE Healthcare) was equilibrated in 10 mM Tris-HCl pH 8.0, 1 mM dithiothreitol (DTT). The conductivity of the supernatant was adjusted to that of the buffer before applying it to the column. Protein was eluted in a 50 mL gradient of 0–200 mM NaCl in 10 mM Tris-HCl pH 8.0. Fractions containing PBP were pooled and concentrated in a Vivaspin 20 concentrator (MWCO 10 kDa, GE Healthcare), yielding ∼130 mg of PBP per liter culture.

To determine the quarternary structure of (A17C, I76G, A197C)PBP it was applied to a Superdex 200pg 16/60 size exclusion column equilibrated in 10 mM Tris-HCl pH 8, 150 mM NaCl, 1 mM NaN3. The protein ran as single species corresponding to the size of the monomer.

Labeling

Purified PBP was labeled with 6-IATR as described previously[9] in 10 mM Tris-HCl pH 8.0, 100 mM NaCl. The mixture was then slowly diluted to ∼3 mM NaCl and concentrated prior to separation of free label and labeled protein. Precipitate was removed from the soluble protein fraction by centrifugation at 16 000g for 10 min at 4 °C, and the supernatant filtered through a 0.2 μM polysulfone membrane (PALL Life Sciences). The protein was then applied to a 1 mL MonoQ HR 5/5 column (GE Healthcare) equilibrated in 10 mM Tris-HCl pH 8.0. The protein was eluted with a 30 mL gradient of 0–100 mM NaCl. The elution profile showed three peaks, with the major, second peak eluting at around 20 mM NaCl.

As determined by mass spectrometry and the ratio of absorbance of label (526 nm) and protein (280 nm), this fraction corresponds to the double-labeled PBP. It was concentrated and further analyzed as described below. The variant used for further study was (A17C,I76G,A197C)PBP, labeled with 6-IATR and this is termed rho-PBPw. The concentration of rho-PBP was calculated using an extinction coefficient of 108 mM–1 cm–1 at 526 nm.[9]

Absorbance and Fluorescence Measurements

Absorbance spectra were obtained on a JASCO V-550 UV/vis spectrophotometer. Fluorescence spectra and titrations were obtained on a Cary Eclipse fluorimeter with xenon lamp. Stopped-flow experiments were performed using a HiTech SF61MX apparatus with mercury–xenon lamp and HiTech Kinetic Studio software (TgK Ltd., U.K.). There was a monochromator and 4 mm slits on the excitation light (548 nm) and a 570 nm cutoff filter on the emission. The concentrations given are those in the mixing chamber unless otherwise stated, and data were fitted to theoretical curves using HiTech software and Grafit 7.[27]

Circular Dichroism

Near- and far-UV spectra were recorded in 1 mm fused silica cuvettes using a JASCO J-815 spectropolarimeter at 20 °C.

Steady-State ATPase Rate Measurements

Rho-PBPw was used to measure the steady-state ATP hydrolysis of the DNA helicase PcrA. Measurements were carried out in 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 3 mM MgCl2 containing 4 nM PcrA, 100 μM ATP, 5 μM BSA, 3 μM rho-PBPw, and dT35 at concentrations ranging from 25 to 300 nM. Reactions were followed using the fluorescence detection setting on a CLARIOstar Microplate Reader (BMG Labtech) with excitation at 546 nm and emission at 580 nm.

Results

Design Approach

In order to extend the useful range of the existing rhodamine-labeled phosphate biosensor[9] to a wider concentration range, the affinity of PBP for Pi had to be lowered by mutation. This could, in principle, be accomplished by several means, given that the two lobes are largely unchanged internally by Pi binding: only their relative position and orientation changes as they enclose the bound Pi. Most obviously and directly amino acids associated with Pi binding could be changed, disrupting binding interactions. Second, the associated conformational change could be targeted by modifying amino acids that are in the cleft and modify its closure, but are not involved in Pi binding. Third, the conformation change, and therefore Pi binding, might be affected by modifying the flexibility of the hinge between the two lobes. Several mutation sites in these three areas were identified, based on crystal structures[18,19,28] (Figure and Table ) and each approach is described in turn. Each PBP variant was labeled on the two cysteines by 6-iodoacetaminotetramethylrhodamine (6-IATR): the fluorescence of this product was used to assess the response to Pi.

Figure 1

Structure of E. coli phosphate binding protein, highlighting active site residues and mutations tested. (A) Complete structure, indicating residues altered in rho-PBPw. Apo PBP is cyan and Pi-bound PBP dark blue.[18,19] The mutation site in the hinge region (I76G) is in yellow, the sites for rhodamine attachment (A17C, A197C) are in magenta. (B) The hinge region from the same structures, showing the movement between apo (cyan) and Pi-bound PBP (dark blue). I76 is shown in yellow. (C) Binding site residues modified are shown as red sticks, and putative bonds between residues and Pi are shown as yellow dashed lines.

Table 1

Survey of Mutations to Weaken Bindinga

label sites	mutations	location	fluorescence ratio (+Pi/–Pi)	Kd [μM]
A17C, A197C		N/A	18	0.07b
	T10A	active site	2.2	32 ± 2
	S38A	active site	1.5	68 ± 9
	R135A	active site	4.5	tightc
	G140A	active site	2.8	70 ± 17
	T10A, G140A	active site	<1.1	ND
	S139A, T141A	active site	<1.2	ND
	I76G	hinge	5.9	7.4 ± 0.4
	K200A	cleft	2.1	ND
	G140A, A225Q	cleft	<1.1	ND
	G140A, A225Y	cleft	1.0	ND
	S39D, S164 K, G140A	cleft	1.0	40 ± 4
K229C, E302C	T10A	active site	1.2	ND
	G140A	active site	2.5	220 ± 11

Amino acids are numbered based on the sequence of the mature PBP. Cysteine mutations for fluorophore attachment are indicated as “label sites”, while “mutations” denote residues altered to lower Pi affinity. The location denotes the general area of the protein where the mutation was introduced. The fluorescence change is the ratio between the signal at saturating Pi concentrations to that containing Pi mop. For variants showing a significant fluorescence change, the Pi affinity was determined by titration of 1 μM rho-PBP variant with Pi in 10 mM Pipes pH 7.0. The maximum signal change and the Kd value were determined typically from three independent titrations.

Values for the “parent”, tight-binding protein are taken from Okoh et al., where the Kd was calculated from the ratio of rate constants.[9]

This protein showed tight-binding, with dissociation constant ∼1 μM or lower.

Binding-Site Mutations

The crystal structures of PBP, bound with Pi, show a large number of potential hydrogen bonds between amino acids and the four oxygen atoms of Pi.[18] Previously PBP with the binding site mutation T141D was used to obtain the crystal structure of the apo protein, as this was shown to have weakened, but highly pH-dependent, binding of Pi.[19] A study of the phosphate binding protein from the cyanobacterium Synechococcus sp. identified several active site residues affecting phosphate affinity.[29] Although sharing only 37% sequence identity, it was predicted to have high structural homology with the E. coli protein, with all but one active site residue between the species being conserved. In Synechococcus sp., replacement of active-site residues lowered phosphate affinity by up to 5 orders of magnitude.

Based on observation of the crystal structures and these data, single or double mutations were introduced into the E. coli PBP as listed in Table and shown in Figure C. With ∼1000-fold decrease compared to the original tight-binding PBP sensor, S38A (Kd 68 μM) and G140A (70 μM) showed the largest change in Pi affinity. Residue R135, forming contacts with two of the phosphate’s oxygens and other residues close to the active site,[18] was also mutated. While this R135A variant showed the largest fluorescence change between apo and Pi-bound PBP, binding remained tight. The T141D variant of MDCC-PBP was also tested, but both the fluorescence change and affinity were highly dependent on pH, as shown previously, with the Kd changing by ∼3 orders of magnitude between pH6.0 and 8.5[11,19] and so was not included in this survey. This would make application of that variant as a biosensor difficult. In addition, some pairs of mutations were also tested in combination, but failed to give a significant fluorescence response, and so their phosphate affinity could not easily be determined. Overall for these variants with binding site mutations, large changes in affinity were accompanied by considerable loss of fluorescence change on Pi binding.

Cleft Mutations

The second approach was based on targeting the cleft closure by affecting the interactions between the facing, inner surfaces of the two lobes.[30−32] This might be achieved by changing amino acids that have interactions across the closed cleft with residues on the opposite side or by changing residue size. Several residues on the lobe surfaces were mutated (Table ). In all cases the Pi-dependent fluorescence change was almost completely lost and so the affinities of these variants for Pi were not determined.

Hinge Mutations

The final approach, in which the flexibility of the hinge region was altered, proved most successful. The hinge consists of two peptides, approximately parallel and linking the two lobes. Cleft closure is achieved by a bending rotation of the hinge (Figure B). A double glycine in the sequence provides flexibility and to increase this, a neighboring isoleucine was replaced with a further glycine resulting in a triple glycine stretch. This residue is located at the point of flexion associated with the rotational movement during the conformational change. The I76G variant of the TMR-labeled (A17C, A197C)PBP responded to Pi with a large fluorescence change and showed the desired reduction in affinity. The combination of these characteristics made this variant, from here on described as rho-PBPw, the best candidate for further characterization.

Spectral Properties of rho-PBPw

On binding Pi to rho-PBPw, the absorbance spectrum of the rhodamine changed (Figure ), with the peak at 515 nm decreasing and the peak at 555 nm increasing. This is typical of two rhodamines shifting from a stacked configuration toward unstacked[33] and is similar to that observed in the original tight-binding rho-PBP.[9] The isosbestic point was 526 nm (Figure inset).

Figure 2

Absorbance spectra of rho-PBPw. Spectra of 1 μM rho-PBPw with 200 μM Pi or phosphate mop (0.1 U mL–1 purine nucleoside phosphorylase, 200 μM 7-methylguanosine) (−Pi) in 10 mM Pipes pH 7.0, 100 mM NaCl. Inset: Traces with 0–400 μM Pi, showing that the isosbestic point was 526 nm.

The fluorescence spectra showed a maximum increase of around 9-fold on binding Pi (Figure ), with excitation and emission maxima of 556 and 577 nm, respectively. This fluorescence change was used to measure Pi affinity, by titrating Pi into the apoprotein, giving a dissociation constant (Kd) of 28 μM (Figure ).

Figure 3

Fluorescence spectra of rho-PBPw. Spectra of 1 μM rho-PBPw with either 200 μM Pi (+Pi) or phosphate mop (0.1 U mL–1 purine nucleoside phosphorylase, 200 μM 7-methylguanosine) (−Pi) were obtained in 20 mM Pipes pH 7.0, 100 mM NaCl. Excitation spectra were measured by emission at 577 nm. Emission spectra were obtained by excitation at 555 nm.

Figure 4

Titration of rho-PBPw with Pi. Two μM rho-PBPw were titrated against Pi at 20 °C in 20 mM Pipes pH 7.0, 100 mM NaCl, and 5 μM BSA. Aliquots of Pi, adjusted to pH 7.0, were added, and fluorescence intensity corrected for the dilution. Data were fitted to a quadratic binding curve[38] with a Kd of 28 ± 3 μM and an 8.7-fold fluorescence change.

To confirm that there were no gross changes to secondary structure by this additional mutation, far UV circular dichroism spectra were obtained (data not shown). Both rho-PBP and rho-PBPw had spectra identical to each other and unchanged in the 200–250 nm region in the presence and absence of saturating Pi. These spectra were also identical to those reported for MDCC-PBP.[20]

Variation of Fluorescence Response with Solution Conditions

Both the affinity of rho-PBPw for Pi and the fluorescence change on binding Pi varied with pH and salt concentration (Table ). The main features are that the amplitude of the signal change decreased with increasing pH, but did not change significantly with salt concentration. The dissociation constant did not vary much with pH at low salt (8.3 μM at pH 6.5 to 46 μM at pH 8.5), but increased more at high salt. There was also a several-fold increase with salt concentration. The maximum signal change was ∼9-fold in 20 mM Pipes pH 6.5, 150 mM NaCl.

Table 2

Fluorescence Change and Affinity at Different pH Values and Salt Concentrationsa

pH	NaCl [mM]	fluorescence ratio (+Pi/–Pi)	Kd [μM]
6.5	20	6.4 ± 1.1	11 ± 2
	150	8.9 ± 1.4	39 ± 5
7.0	0	5.0 ± 0.3	4.2 ± 0.4
	20	5.5 ± 0.4	15 ± 1
	150	7.2 ± 0.8	50 ± 8
7.5	20	4.4 ± 0.9	25 ± 4
	150	5.4 ± 0.4	121 ± 16
8.0	20	2.1 ± 0.1	39 ± 9
	150	2.6 ± 0.2	120 ± 28
8.5	20	1.8 ± 0.1	33 ± 3
	150	2.0 ± 0.1	103 ± 20

Fluorescence titrations were obtained as described in Figure . Measurements at pH 6.5 and 7 were in 20 mM Pipes, and those at pH 7.5–8.5 in 50 mM Tris-HCl. The fluorescence ratio is between the signal at saturating Pi concentrations to that of the solution prior to Pi addition. The value displayed corresponds to the signal change obtained from the fit to the titrations as in Figure , rather than from spectra. Measurements were done on at least three experimental replicates and two separate protein preparations, which were then combined for fitting.

Limit of Pi Binding Kinetics

With the tight-binding variants, rho-PBP and MDCC-PBP, Pi binding was interpreted as occurring in a two-step mechanism, namely, diffusion-controlled binding to the open conformation of PBP, followed by the cleft-closing conformation change.[9,20] It is the latter that produces the fluorescence change. It is likely that the reduced affinity of rho-PBPw is accompanied by a combination of slower binding and/or faster dissociation. To test this, the time course was measured, following rapid mixing of rho-PBPw with Pi, using fluorescence stopped-flow. At all the concentrations of Pi tested, no fluorescence change occurred during the observed time course, which starts after the dead time of the stopped-flow instrument (2 ms). However, the fluorescence intensity, although constant with time, increased with Pi concentration with a similar dependence as the steady-state titrations (Figure ). This suggests that the transient fluorescence change on binding was complete within the dead time and only the final, Pi-bound fluorescence was observable.

Figure 5

Comparison of fluorescence intensities of rho-PBPw, following stopped-flow mixing with Pi and end-point titration. Pi at various concentrations was mixed with 1 μM rho-PBPw in 20 mM Tris-HCl pH 7.5, 150 mM NaCl and 5 μM BSA at 20 °C in a stopped-flow apparatus. No change in fluorescence was observed during the time courses, suggesting the reaction was complete within the dead time of the instrument (2 ms). However, the constant intensity for each trace increased with [Pi]. As a control and to get the background fluorescence, 1 μM rho-PBPw was mixed with buffer. To confirm that the fluorescence intensities were those of the final Pi-bound form, an end-point titration was done in a cuvette under exactly the same solution conditions and the fluorescence changes (circles) compared with the stopped-flow data (triangles). To enable the comparison of data obtained using the different optics, the fluorescence intensities were adjusted to have the same values at 500 μM Pi. The combined data were fitted to a quadratic binding curve (Kd of 50 ± 8 μM).

To provide further evidence for this interpretation, an attempt was made to measure the dissociation kinetics directly by rapidly mixing rho-PBPw·Pi with a large excess of unlabeled, but high affinity PBP, as previously done to measure the dissociation kinetics of high affinity rho-PBP.[9] In this case, no fluorescence change was observed (data not shown), suggesting that dissociation was rapid and complete within the dead time.

Specificity

Wild-type PBP and the original phosphate biosensors showed high selectivity for inorganic phosphate.[9,18] In order to establish whether lowering the affinity had a significant effect on specificity, rho-PBPw was titrated with a selection of Pi analogues, pyrophosphate, and nucleotides (Table ). No significant fluorescence response was obtained from addition of the main competing ligands. The Pi analogues chosen were those that did give a response with the tight-binding Pi sensor, namely, arsenate and vanadate: neither gave a response at the concentrations tested.

Table 3

Response of rho-PBPw to Pi Analogues and to Nucleotidesa

species	fluorescence ratio (+Pi/–Pi)
ATPb	1.1
ADPb	1.1
GDPb	1.5
PPib	0.9
sodium arsenate	1.1
sodium vanadate	1.0

The species tested were added to 5 μM rho-PBPw in 20 mM Pipes pH 7.0, 100 mM NaCl. Fluorescence emission spectra of the solution ±100 μM of the respective species were measured as described for Figure . Saturating Pi was subsequently added to each mixture to verify sensor response. The fluorescence ratio was calculated from the baseline fluorescence of the solution prior to substrate addition and the resultant fluorescence after addition. The highest signal change was observed for GDP, but may be explained by contaminating Pi in the nucleotide stocks.

These were treated with phosphate mop (1 U ml–1 purine nucleoside phosphorylase, 200 μM 7-methylguanosine)[16] for 15 min at room temperature to reduce contaminating Pi, prior to addition to rho-PBPw. The phosphorylase activity may be inhibited by the presence of purine nucleotides at high concentration, limiting the effectiveness of this treatment.

Comparison of Fluorescence Intensity Levels of TMR

The rhodamine fluorescence intensity of rho-PBPw was compared with the tight-binding variant and tetramethylrhodamine in solution. The fluorescence quantum yield is difficult to measure accurately when the emission spectrum extends to high wavelength, where corrections are large. To circumvent this, a simpler measure was used by dividing the corrected fluorescence emission intensity at the maximum wavelength by the absorbance at the excitation maximum. This ratio was then used to give an approximate comparison between different species containing the same tetramethylrhodamine fluorophore (Table ). This shows that the rhodamine is much less fluorescent when attached to the (A17C, I76G, A197C)PBP than in free solution or attached to a small molecule thiol, but does not change greatly on Pi binding.

Table 4

Relative Fluorescence Intensities of Tetramethylrhodaminesa

species	relative fluorescence
6-IATR in EtOH	1
6-IATR in buffer	0.42
6-IATR-MESNA	0.51
rho-PBPw, Pi	0.08
rho-PBPw, no Pib	0.03
rho-PBP, Pi	0.13
rho-PBP, no Pib	0.07

Apart from the 6-IATR sample in ethanol, all measurements were in 10 mM Pipes pH 7.0, 100 mM NaCl using solutions with a maximal absorbance of <0.05. The fluorescence spectra were then corrected for the photomultiplier profile and baseline. The fluorescence intensity at maximum emission was divided by the absorbance at maximum excitation. The relative fluorescence was normalized to that of 6-IATR in ethanol.

Solutions were treated with phosphate mop as described in Figure .

Steady-State Assay of Pi Production

To test rho-PBPw, the ATPase activity of PcrA, a DNA helicase, was measured in a multiwell plate format (Figure ). PcrA couples ATP hydrolysis to translocation along single-stranded DNA, using one ATP molecule per translocation step of one base.[34] This ATPase activity has been well characterized using the high-affinity phosphate sensor under identical solution conditions[35] and so provided a good test of the new biosensor. Both the Km (88 nM) for dT35 and kcat (32 s–1) obtained using rho-PBPw are similar to the previous measurements. Using the new biosensor a larger range of Pi could be monitored (up to 80 μM). In addition, the lowered affinity rendered the sensor less sensitive to contaminating Pi, so that no phosphate mop was required to remove such contaminant.[16]

Figure 6

Steady-state ATPase activity of PcrA. Measurements were carried out at 18 °C in 20 mM Tris-HCl pH 7.5, 150 mM NaCl and 3 mM MgCl2. Reactions contained 4 nM PcrA, 3 μM rho-PBPw, 100 μM ATP, and dT35 at various concentrations. The concentration dependence of initial rates was fitted to the Michaelis–Menten equation to give a Km for dT35 of 88 ± 13 nM and a kcat of 32 ± 1 s–1. The Inset shows a calibration under the same conditions.

Discussion

The affinity of the original Pi biosensor, rho-PBP,[9] was altered by inserting mutations in strategic positions of the PBP scaffold. Kinetic and other measurements on the fluorescent, tight-binding PBP suggested a rapid, possibly diffusion controlled initial binding of Pi to the open conformation, followed by a rate-limiting cleft closure.[9,20] In principle, the affinity could be altered by changing either step of this mechanism. Three parts of the structure were targeted for mutations that might affect the binding, namely the binding site, interlobe cleft surfaces and the hinge region between the two lobes.

The Pi-binding site is highly specific, being able to discriminate against similar molecules such as sulfate. It contains 12 residues that form hydrogen bonds with the four oxygens of the Pi.[18] While most of these residues are hydrogen donors, D56 is a hydrogen bond acceptor and plays a crucial role in substrate recognition and discrimination. On the whole for binding site mutations, a large reduction in affinity was accompanied by a large reduction in the signal change on binding Pi. This meant that mutations in the binding site did not give a suitable candidate weak-binding variant. Moreover, there were significant differences in effects of mutations in the E. coli protein, used here, from those described for Synechococcus PBP, predicted to be a close structural homologue.[29] For example, the T10A mutation, which resulted in a large loss of affinity (5 orders of magnitude) in the Synechococcus PBP, whereas there was a much smaller change in affinity, ∼ 500-fold lower, in the E. coli protein. The difference in effect of active site mutations cannot, therefore, be exclusively due to changes in the active site residues, as these are conserved between the two species, even though overall the proteins share only moderate sequence homology. Active site mutations may have varied secondary effects on the two proteins, due to differences in their amino acid composition. In addition, different signal elements were used in the two biosensors, a small fluorophore here and fluorescent protein with the Synechococcus PBP. This, as well as the different sources of the two PBPs, may provide an explanation of the differences observed.

Mutations to the cleft surface resulted in small or no signal change upon Pi addition, so the affinity could not be readily determined. Although the predicted structure of the rhodamines in the apo structure suggests little contact with surface amino acids,[21] this aspect is little understood and mutations in the cleft might affect indirectly the structural changes that disrupt the rhodamine interaction. In other words, mutations that weaken the interaction across the cleft may mean that cleft closure becomes unfavored thermodynamically. Measurements with the tight-binding variant, rho-PBP, suggest that the cleft closing conformation change on rho-PBP·Pi occurs subsequent to Pi binding itself and has an equilibrium constant of 40 in favor of closure,[20] so only a small change in this equilibrium constant will be significant in this respect.

The most successful strategy for lowering Pi affinity, while maintaining a fluorescence signal, was altering the hinge between two lobes making up PBP. Insertion of a third glycine residue at position 76 in the hinge (Figure B) lowered Pi affinity up to several hundred-fold, compared to the tight-binding sensor, but maintained a reasonably large signal change (Table ). This extra mutation is on the opposite side of the protein from the rhodamines (Figure A), so may be unlikely to affect their fluorescence directly. The fact that the replacement amino acid is glycine suggests the possibility that the hinge is now more flexible. Circular dichroism measurements show no significant changes to secondary structure due to this mutation.

Presumably the cleft-closing conformation change is still favored thermodynamically in rho-PBPw, in order to result in net cleft closure and produce the fluorescence change. So the main effect of I76G would be on the equilibrium constant of the binding step itself, either slowing binding or increasing dissociation rate constants.

Stopped-flow kinetic measurements were unable to give precise kinetic information as the binding and dissociation kinetics were too fast to measure. However, the data suggested that the sum of rate constants for dissociation must be >1800 s–1 for the reactions to be almost complete within the 2 ms dead time of the stopped-flow instrument, even at low Pi concentration. This in turn suggests that the dissociation rate constant is much increased compared to that determined for the tight-binding rho-PBP (7 s–1).[9] Because the fluorescence change is likely to depend on the closing of the cleft, an equilibrium constant favoring cleft closure must be maintained. This may in turn limit the extent of weakening Pi binding by mutation, while keeping a fluorescence change.

Table shows a measure of the relative fluorescence of various TMR adducts, including 6-IATR in different solvents, as its adduct with a small molecule, MESNA, and covalently attached to the tight-binding, as well as weak-binding Pi sensor. The relative fluorescence efficiency was determined using the ratio of fluorescence emission to absorbance, each at its maximum wavelength. By this measure, the fluorescence of TMR in aqueous solution is less than half that in ethanol. In both the weak- and the tight-binding variant of rho-PBP, TMR has relatively low fluorescence, even when Pi is bound. There is only a relatively small difference in this fluorescence ratio of rho-PBP in presence and absence of Pi, which is qualitatively in line with the expected mechanism for the fluorescence increase. On binding Pi the TMR labels become at least partially unstacked, leading to an increase in absorbance of the rhodamines. This, in turn, is accompanied by an increase at the fluorescence excitation maximum. The observed fluorescence intensity change on binding Pi is a combination of the increase in extinction coefficient at the exciting wavelength and an increase in fluorescence per unit of absorbance. However, the absorbance spectra are complex with discrete maxima at 516 and 556 nm with the latter peak showing a shoulder in the presence of Pi. The predicted structure, based on molecular modeling, suggests that the two rhodamines in the apo PBP are not exactly parallel, so the stacking is imperfect,[21] at least compared with the situation where rhodamines are freely mobile in solution,[36] and the relative size of the two rhodamine peaks suggests incomplete unstacking when Pi binds. In contrast, monomeric TMR in free solution has the 555 nm peak larger than the 515 nm one. These factors suggest that the stacking/unstacking between the two conformations of rho-PBP is more complex than seen in free solution.

The I76G variant of rho-PBP, rho-PBPw, extends the useable range of the biosensor so Pi can be readily measured in the tens of micromolar range, using substoichiometric protein. The fluorescence response to Pi was ionic strength and pH dependent with the maximum response of 9-fold. While the affinity of rho-PBPw was little affected by pH, there was a decrease with salt concentration (Table ). These findings highlight the requirement to obtain a calibration for each experimental condition in any assay based on fluorescence. Importantly, the additional, weakening mutation did not affect specificity: none of the molecules tested resulting in a significant fluorescence response (Table ).

Both previous, high-affinity sensors based on the PBP scaffold have been used in a range of in vitro studies.[10,13,34,35,37] The ability of rho-PBPw to be used to detect Pi in a steady-state assay in real time was demonstrated in a test system. The ATPase activity of the helicase, PcrA, showed comparable results using rho-PBPw to those using rho-PBP under the same solution conditions.[35]

The new biosensor has several properties differentiating it from the existing MDCC-PBP and tight-binding rho-PBP. Rhodamines are much more photostable than diethylaminocoumarin, an important property for assays relying on high intensity excitation, such as single molecule and high-throughput studies. High-throughput assays, in particular, also benefit from the extended useable range in two ways. First, the lower affinity allows for the sensor to be used at substoichiometric concentrations relative to Pi, making it more economical, whereas the tight-binding biosensors must be present in significant excess over the highest concentration of Pi. Second, the higher Pi detection range allows reactions to be measured over a longer time span, extending the assay range, for which the sensor can be used. The tight-binding MDCC-PBP and rho-PBP remain particularly suited for rapid reactions, such as single-turnover measurements of Pi release, where sensitivity and rapid response are essential. Importantly, the substoichiometric use and weaker affinity mean that rho-PBPw is likely to be little affected by typical levels Pi contamination that can be present in many biological solutions, buffers, and so forth.[17] In contrast, the tight-binding version binds contaminant Pi stoichiometrically under most assay conditions, so that a significant proportion of the biosensor can be Pi-bound at the start of the assay, increasing the background fluorescence and also decreasing the biosensor free to detect Pi formation.

The development of this biosensor, based on the phosphate binding protein, illustrates the variety of mutational approaches that might be used to achieve the desired weakened binding, while maintaining a useful signal change. The strategy that succeeded was addition of a third glycine in the hinge region to increase its flexibility. In doing so, it widens the range of applications for this type of biosensor.