Biochemical Comparison of Anopheles gambiae and Human NADPH P450 Reductases Reveals Different 2′-5′-ADP and FMN Binding Traits

NADPH-cytochrome P450 oxidoreductase (CPR) plays a central role in chemical detoxification and insecticide resistance in Anopheles gambiae, the major vector for malaria. Anopheles gambiae CPR (AgCPR) was initially expressed in Eschericia coli but failed to bind 2′, 5′-ADP Sepharose. To investigate this unusual trait, we expressed and purified a truncated histidine-tagged version for side-by-side comparisons with human CPR. Close functional similarities were found with respect to the steady state kinetics of cytochrome c reduction, with rates (k cat) of 105 s−1 and 88 s−1, respectively, for mosquito and human CPR. However, the inhibitory effects of 2′,5′-ADP on activity were different; the IC50 value of AgCPR for 2′, 5′ –ADP was significantly higher (6–10 fold) than human CPR (hCPR) in both phosphate and phosphate-free buffer, indicative of a decrease in affinity for 2′, 5′- ADP. This was confirmed by isothermal titration calorimetry where binding of 2′,5′-ADP to AgCPR (K d = 410±18 nM) was ∼10 fold weaker than human CPR (K d = 38 nM). Characterisation of the individual AgFMN binding domain revealed much weaker binding of FMN (Kd = 83±2.0 nM) than the equivalent human domain (Kd = 23±0.9 nM). Furthermore, AgCPR was an order of magnitude more sensitive than hCPR to the reductase inhibitor diphenyliodonium chloride (IC50 = 28 µM±2 and 361±31 µM respectively). Taken together, these results reveal unusual biochemical differences between mosquito CPR and the human form in the binding of small molecules that may aid the development of ‘smart’ insecticides and synergists that selectively target mosquito CPR.


Introduction
The mosquito Anopheles gambiae 'is the principal vector for malaria in sub-Saharan Africa, a disease that affects over 500 million people worldwide. Insecticides have been the mainstay of disease control programmes in disease endemic countries for many years. However, these are threatened by the rapid evolution of insecticide resistance in disease vectors [1]. An important mechanism of resistance is enhanced metabolic inactivation of insecticides by P450s [2], a diverse superfamily of heme-containing monoxygenases that catalyse a diverse range of chemical reactions important in developmental processes and for detoxification of foreign compounds. It is well known that inhibiting P450 activity can help overcome resistance by potentiating insecticidal activity. Indeed piperonyl butoxide, a broad spectrum P450 inhibitor, is being increasingly used to extend the lifetime of pyrethroids, the only class of insecticide that can be safely used for insecticide treated bednets, in areas where resistance is undermining malaria control [3]. P450s are located in the endoplasmic reticulum where they require electrons supplied by NADPH cytochrome P450 oxidoreductase (CPR) for catalysis [4] placing this protein in a critical path in metabolism-based insecticide resistance, and a novel target for the development of new chemical synergists.
Despite its great medical importance, there is relatively little known about the biochemistry and function of the individual components of the P450 complex of A. gambiae. Most recently, CPR from the mosquito A. minimus has been characterised [12,13]. The mosquito enzyme shared close biochemical similarities with other CPR family members (Figure 1), although weak binding FMN and FAD cofactors [12] points to potential species differences. We have mapped the tissue distribution of CPR in A. gambiae and discovered high levels of expression in specialised mosquito cells (oenocytes) that suggests key physiological roles for CPR in metabolic processes and pheromone production/metab-olism [14]. As expected, CPR gene knockdown by RNAi greatly increases the susceptibility of A. gambiae to permethrin, a widely used pyrethroid insecticide, emphasising the important chemoprotective role of the P450 monooxygenase complex in this  [11]. Conserved residues associated with 29-phosphate binding (S596, R597, K602 and R634 [17,32]) are shown with an asterisk. The FMN binding domain spanning amino acids E69 to G254 that was expressed in E. coli is underlined. organism [14], and validating its potential as a target for the development of chemical inhibitors that might enhance insecticidal activity.
Here we have expressed AgCPR and the equivalent human isoform in E.coli in order to allow direct comparisons of cofactor binding and steady-state kinetics. This has identified significant differences in the binding of 29, 59-ADP and FMN that suggest a potential target for the design of new insecticides or synergists to combat malaria.

Materials and Methods
Cloning, expression and purification of D63AgCPR The cloning of AgCPR cDNA was previously described [14,15]. The membrane anchor sequence was deleted by removal of amino acids 2-63 by PCR, using PFU polymerase (Stratagene) and the following oligonucleotides; forward primer: 59-CGCG GAT CCG ATG ACG ATG ACG ATG GTG GAG ACC -39and reverse primer: 59-TTC GGA TCC TTA GCT CCA CAC GTC CGC CGA -39 (BamHI sites are underlined and the respective start and stop codons are indicated in bold). For molecular dissection of D 63AgCPR into its individual FMN-binding domain, PCR was used to amplify the amino acid residue region E69 to G254 using the following oligonucleotides: forward primer: 59-CCATGGA-GAACTCGTTCATCAAGAAGC -39 (NcoI site is underlined) and revese primers: 59-CCATGGTTACTCGCCGGTGTA-GATGC -39 (KpnI site underlined). The FMN -binding region expressed is underlined in Figure 1. The PCR product was digested with Kpn1 and Nco1 and subcloned into the expression vector pETM-11 which contained a TEV protease cleavage site. Constructs were confirmed by DNA sequencing. All proteins were expressed in E. coli strain BL21-CodonPlus(DE3)-RP strain at 30uC and purified using standard Ni 2+ -affinity chromatography and elution with 250 mM imidazole [16]. Thrombin (2000:1(w/w) protein:thrombin for AgCPR) and TEV protease (20:1(w/w) protein:TEV protease for the FMN binding domain) were, respectively, used to remove the His-tag from D63AgCPR. For thrombin cleavage, the sample was incubated for 2 hr at room temperature, whereas for TEV cleavage, overnight incubation at room temperature was required. The samples were reapplied to a Ni 2+ -affinity column to removed free his-tags and uncut protein.
The samples were desalted in the presence of an excess of FMN and FAD and further purified using a 5 mL HiTrap Q FF anion exchange column (GE Healthcare) followed by MonoQ 5/5 anion exchange column (GE Healthcare). For final purification and buffer exchange, size exclusion chromatography was performed using a HiLoad 26/60 Superdex 75 column (50 mM Tris-HCl (Melford), 50 mM NaCl; pH 7.0). The human form, hCPR, was purified as described by Dohr [17]. Protein purity was .95% as assessed by SDS-PAGE and all samples were verified by mass spectrometry. Protein concentration was calculated by Bradford assay.

Spectral Analysis
Absorption spectra of the purified CPR were carried out with a Cary 300 Bio and Cary 4000 spectrophotometers [18]. The sample buffer was 100 mM Tris, pH 8.0.

Flavin content determination
Flavin cofactors (FMN and FAD) were released from AgCPR by heat denaturation (Paine et al, 1999). Briefly, 100 mL AgCPR (0.1 mg/mL) was incubated at 95uC for 5 min, centrifuged at room temperature at 20, 0006g for 5 min and the supernatant analysed for flavin content by HPLC analysis. 10 mL sample (or standards) were loaded into a mobile phase of 50 mM ammonium acetate; pH 4.5 with 20% (v/v) acetonitrile for separation on a 250 mm C 18 column (AcclaimH120, Dionex) at 23uC. The flowthrough was analysed by absorption spectroscopy and fluorescence (excitation 450 nm, emission 525 nm). Quantification was determined with reference to authentic FMN and FAD standards (Sigma).
Cytochrome c and NADPH kinetics. The rate of change of absorbance of horse heart cytochrome c (Sigma-Aldrich, UK) at 550 nm was measured at 25uC using a Cary 4000 UV-Visible spectrophotometer essentially as described [17]. For cytochrome c kinetics, 0.75 pmol purified AgCPR or hCPR was pre-incubated with 0-110 mM cytochrome c (dissolved in 0.3 M potassium phosphate buffer; pH 7.7 or 0.1 M Tris-HCl, pH 7.7/0.1 M KCl) in a total volume of 500 ml for 2 min at 25uC. Reactions were initiated by the addition of NADPH to a final concentration of 50 mM and rates measured in duplicate for 2 mins, a linear reaction range. For determination of NADPH kinetic parameters, the cytochrome c concentration was kept constant at 50 mM, and reactions were initiated with 0-150 mM NADPH.

Inhibition measurement
Cytochrome c. Measurement of cytochrome c reduction was carried out at 25uC with 50 mM cytochrome c and 0.75 pmol purified AgCPR or hCPR essentially as described [17]

Isothermal Titration Calorimetry (ITC)
ITC experiments were performed using a ITC200 microcalorimeter (Microcal Inc/GE Healthcare). Protein samples were dialysed into 100 mM BES; pH 7.0 from 50% glycerol stocks stored at 220uC. The concentrations of all the proteins were determined by using Bradford assay. Samples were centrifuged at 13,0006g for 10 min and degassed at 23uC (2uC below the temperature at which the experiments were performed). All experiments were performed at 25uC. Typically, the protein concentration in the cell was 10-20 mM whilst ligand concentrations were in the range of 100-1000 mM. The data analysed using Origin 7.0 (Microcal) after subtracting the heats of dilution obtained from parallel experiments performed by injecting the nucleotide into the buffer. Thermodynamic parameters n (stoichiometry), K d (1/K a , the association constant), and DuH (enthalpy change) were obtained by nonlinear least-squares fitting of experimental data using the single-site binding model. The free energy of binding (DGu) and entropy change (DSu) were obtained using eqs 1 and 2.

Expression and purification of AgCPR and flavin-binding domains
Initial attempts to purify full-length AgCPR expressed in E. coli using standard ion exchange and 2, 59-ADP-Sepharose affinity chromatography [19,20] failed, which suggested that AgCPR might have different nucleotide binding properties. Therefore, to examine this further N-terminally histidine tagged soluble forms, lacking the N-terminal membrane anchor, of AgCPR was expressed in E. coli and affinity purified over nickel agarose ( Fig. 2A). Although CPR lacking the amino-terminal membrane anchor loses its ability to interact with P450 it is otherwise fully functional and capable of reducing a range of electron acceptors [11]. Equivalent preparations of human anchorless CPR were also produced [17], and comparative enzyme activities measured via the reduction of cytochrome c, the surrogate electron acceptor most commonly used for measuring diflavin reductase activity [21].
Purified AgCPR was yellow, indicating the binding of flavin cofactors. AgCPR contained 0.7260.01 mol of FMN and 0.8060.01 mol of FAD per mol of enzyme. This compared with 0.8860.03 mol of FMN and 0.9260.02 mol of FAD per mol of hCPR enzyme purified under the same conditions, and somewhat higher than those for A. minimus (0.51 and 0.63 mol FMN and FAD per mol enzyme respectively) [12]. The optical spectrum of the oxidised enzyme was typical for a diflavin reductase with absorbance maxima at 379 nm and 454 nm (Fig. 2 B). The addition of a stoichiometric amount of NADPH produced a characteristic increase in the 500-650 nm region, associated with the formation of an air-stable flavin semiquinone as for human [22] and insect CPRs [12,23]. The appearance of a peak in the 340-350 nm region in the semi-quinone form associated with NADPH oxidation similar to humans [24] was also evident.

Comparison of Nucleotide Binding
Since AgCPR appeared to bind weakly to 29, 59-ADP the binding properties of this molecule were examined in more detail. 29, 59-ADP is the adenosine-ribose moiety of NADPH, which binds through a bi-partite mode with the nicotinamide moiety, to separate binding pockets of CPR [17]. Thus to delineate the roles played by the nicotinamide and ribose moieties, we used isothermal titration calorimetry (ITC) to determine the binding affinities of the adenosine-ribose fragments ( Table 1). As previously reported for hCPR [25], this was done in a phosphate-free buffer to prevent competitive interactions with free phosphate. Figure 3 represents the ITC binding isotherms resulting from the titration of AgCPR with 29, 59-ADP, NADP+ and 29-AMP. The binding isotherms were exothermic and fit to a single-site model. The  observed dissociation binding constant for the AgCPR -2959-ADP interaction was 410618 nM, with a binding stoichiometry n = 1.03. This is substantially weaker than hCPR under the same reaction conditions (K d of 3863.8 nM; this study and [25]). The binding affinity of NADP+ was ,5-fold weaker for AgCPR (K d = 363617 nM) versus its human equivalent hCPR (K d = 69612 nM), while 29-AMP binding was analysed to confirm the importance of the 29-phosphate to nucleotide interaction with CPR. As expected the binding strength was orders of magnitude lower than 29-59-ADP and NADP+. Nevertheless a modest twofold decrease in the apparent affinity for 29 AMP in the mosquito protein (K d of 460.18 m M vs human 1.660.44 m M), was measurable which compared to 2959-ADP suggest that interactions with the 59 phosphate group make significant contributions to the binding affinity of the adenosine ribose moiety.
The enthalpy change (DH) for hCPR binding to 29,59-ADP was 220.36 kcal mol 21 was higher than for A. gambiae CPR binding (DH = 213.13 kcal mol 21 ) ( Table 1). This is consistent with the lower dissociation constant value, and a further indication that the energetics of 29,59-ADP binding are more favourable with the human enzyme. Likewise, the DH for NADP + binding to hCPR was approximately 2 fold lower than AgCPR (219.46 kcal mol 21 vs 211.97 kcal mol 21 ). Finally, differences were minimal with respect to the energetics of 29-AMP binding.
In summary, AgCPR binds all the nucleotide analogues more weakly than hCPR. Secondly, for both proteins, the small difference in affinities of each protein for NADP + and 29, 59-ADP, suggests that in both the human and Anopheles gambiae proteins, the nicotinamide moiety makes only minimal contribution to the binding surface. From a structural viewpoint, it is likely that the conformations around the NADPH binding site in Anopheles gambiae CPR are similar to those already discovered for the mammalian proteins [10].

Comparison of FMN Binding
The binding affinity of FMN to the isolated FMN-binding domains of both A. gambiae and hCPR were also examined by titration of the apo-FMN binding domains ( Figure 4). Both isotherms show reasonable binding and saturation of the apoproteins is achieved with excess flavin. The K d for FMN binding to the isolated apo-human FMN-binding domain was calculated to be 2360.9 nM which is almost 4-fold stronger than for the isolated apo-A. gambiae FMN-binding domain which was calculated to be 8362.0 nM.   Table 1 The thermodynamic characteristics (

Kinetic comparisons between AgCPR and hCPR
To compare activities, standard phosphate buffer conditions that give optimal hCPR activity were initially used [17]. The specific activity of purified AgCPR was 23.9 mmol/mg/min compared with 18.3 mmol/mg/min for hCPR. Both AgCPR and hCPR catalysed NADPH-dependent cytochrome c reduction following Michaelis-Menton kinetics with respect to substrates cytochrome c and NADPH. The K m cytc values for human and mosquito enzymes were 19 mM and 23 mM respectively, indicating similar binding affinities for cytochrome c. There was a slight decrease in affinity of the mosquito CPR for the NADPH cofactor relative to the human protein (K M NADPH 30.0 mM vs 16.0 mM respectively). Rates of cytochrome c reduction were alike, characterized by turnover numbers (k cat ) of 105 s 21 and 88 s 21 respectively for mosquito and hCPRs and near the range of 50-100 sec 21 noted for most CPRs [20]. Interestingly, the turnover rate for AgCPR is ,4 fold higher than that reported for A. minimus CPR (23.8 s 21 ) [12], possibly attributable to the low stoichiometry of FMN in A. minimus CPR. Overall, however, close functional similarities were found with between human and A. gambiae CPR with respect to the steady state kinetics of cytochrome c reduction.
Comparison of cytochrome c reductase inhibition by 29, 59-ADP 29,59-ADP, 29-AMP and NADP+ were used as inhibitors of CPR catalyzed reduction of cytochrome c to determine whether the observed difference in ITC dissociation constants (K d ) for complexes formed with 29,59-ADP were manifest as variations in inhibition (IC 50 ) of catalytic activity (Table 3). Consistent with the lack of binding to 29, 59-ADP Sepharose, the major difference observed in phosphate buffer was a ten-fold decrease in affinity for 29, 59 -ADP, which inhibited AgCPR with an IC 50 value of 262 mM compared with 28 mM for hCPR. There was also a twofold increase in affinity for 29 AMP (mosquito 468 mM vs human 1085 mM), while no significant differences were observed for NADP (mosquito 129 mM vs human 105 mM).
Since the binding affinity for NADPH is reduced in the presence of high phosphate concentrations [26], which may have influenced the data, we also investigated IC 50 reactions in a phosphate-free Tris buffer (Table 3). Under these conditions we observed a ,3 fold decrease in K M NADPH values, relative to phosphate buffer, for human and mosquito enzymes (5.1 mM and 12.4 mM respectively), as well as corresponding decreases in K cat to 46.7 sec 21 and 57.5 sec 21 respectively, and a general decrease in IC 50 values (Table 3). However, the 29, 59-ADP affinity for mosquito CPRs remained significantly lower (6 fold) than human. Differences in 29-AMP and NADP IC 50 values were negligible.
Finally, we also compared the inhibitory effects of diphenyliodonium chloride, a widely used inhibitor of CPR and related enzymes [27], on cytochrome c reductase activity. A difference of the same magnitude as that observed with 29, 59-ADP but in the opposite direction was also discovered: AgCPR is one order of magnitude more sensitive than hCPR to diphenyliodonium chloride (Table 3). Taken together, these results reveal significant biochemical differences between mosquito CPR and the human form in the binding of the small molecules 29, 59-ADP and diphenyliodonium chloride.

Discussion
In view of the paucity of public health insecticides there is a pressing need to understand the mechanisms of detoxification and identify targets for new insecticides against disease vectors, in particular A. gambiae. Shutting down P450 activity is an attractive option since it can block resistance thereby extending the lifetime of insecticides already in use, as proven with the P450 inhibitor piperonyl butoxide [3]. Being the obligate redox partner for microsomal P450s, undermining Anopheline CPR activity is an attractive target provided it can be selectively inhibited. Although numerous CPRs have been characterised, few if any side-by-side comparisons have been made. Our direct comparative biochemical analysis of A. gambiae and human CPRs has revealed substantial differences in co-factor binding which may aid design of species specific CPR inhibitors. Of particular note was the weak affinity of AgCPR for 29, 59-ADP, reflected in the lack of binding to 29, 59-ADP affinity matrix and, ,6-10 fold higher IC 50 value compared to hCPR for cytochrome c reduction and ,10 fold higher dissociation binding constant (K d = 410 nM vs 38 nM for hCPR). This is significant since, for other CPRs, 29,59-ADP binding exerts conformational changes affecting electron transfer rates [25,28,29], while 29-phosphate is the major contributor to the high affinity binding of NADPH to CPR [25], 25].
The results here also show that the A. gambiae protein binds the nucleotide analogues in ways similar to the human one (hCPR), with the nicotinamide moiety making only modest contributions to the binding surface. In the crystal structure of the rat reductase, NADP + was found to bind in multiple conformations. It is, therefore, difficult to precisely identify the residues that might be important for nucleotide binding for the human protein which could provide clues as to why the binding affinities to Ag CPR are different. At this stage, we postulate that the considerably weaker binding of 29,59-ADP to AgCPR could have a significant influence on enzyme function Differences also extend to FMN binding, where AgCPR again shows ,4 fold weaker affinity against hCPR (K d 23 nM vs 83 nM respectively). Interestingly, AgCPR is much more (,10 fold) sensitive to diphenyliodonium chloride inhibition than hCPR. While there appears no obvious effect on cytochrome c reductase activity (the steady state kinetics AgCPR being comparable to hCPR), we have not compared activity with other redox partners such as P450s and heme oxygenase, where functional differences may be more pronounced. Indeed, given that heme degradation in mosquitoes is vital for survival with a blood feeding habit, and heme oxygenase is closely involved in heme degradation [30], it is feasible that biochemical differences in CPR may reflect differences in redox partner interactions. This hypothesis is supported by the fact that methionine synthase reductase, a diflavin reductase closely related to CPR, has a low affinity binding mode for 29, 59 -ADP (,500 nM) [31]. An important question arising from this work is to what extent the observed differences in affinity of AgCPR for the adenosine, flavin and diphenyliodonium chloride molecules reflect intrinsic structural and/or mechanistic differences with respect to CPRs from human and other species. Since interactions involving 29phosphate are the major contributor to the high affinity binding of NADPH to CPR [26] we might expect differences in 29, 59-ADP to be associated with this phosphate molecule. However, the amino acids proposed to make up the 29-phosphate binding motif including serine 596, arginine 597, lysine 602 and arginine 634 [17,32] are conserved in A. gambiae CPR. Furthermore, these may prove difficult to resolve through X-ray crystallography due the inherent flexibility of this region in the CPR enzyme family [11].
The three-dimensional structures of the rat [11] and yeast [9] CPR and the rat nNOS FAD domain [33] have collectively provided details of the cofactor binding sites for the oxidoreductases. FMN binds in the region 139 TYGEGPD and 175 NKTYEHFN (rat CPR numbering), with Y140 and Y178 sandwiching the FMN isoalloxazine ring, and F181 stabilising the cofactor binding [34]. We have performed a multiple sequence alignment of AgCPR with A. minimus CPR, rat CPR, S. cerevisiae CPR and rat nNOS reductase domain (Fig. S1). As far as AgCPR is concerned, the residues mentioned above are all conserved with the exception of F181 (in rat CPR) replaced by Y184 (in AgCPR). However, it had previously been reported that mutation of F181 to Y181 in hCPR had minimal effects on FMN binding and enzymatic activity [34]; hence, replacing the tyrosine residue in AgCPR protein is of little consequence for FMN affinity. Therefore, despite the availability of good structures of homologous proteins, these have so far proved inadequate for understanding the cofactor binding characteristics of the AgCPR. Determination of the FMN-binding domain of AgCPR is in progress.
Finally, the greatly increased sensitivity of AgCPR for diphenyliodonium chloride may be related to increased rates of internal electron transfer, possibly linked to differences in redox potentials. Diphenyliodonium chloride inhibition is proposed to occur through an NADPH dependant reductive mechanism whereby diphenyliodonium chloride receives an electron from a reduced flavin to produce a reactive phenyl radical that recombines with FMN in its semiquinone state, creating an inactive phenylated FMN adduct [27].

Conclusion
Overall, these results reveal significant biochemical differences between mosquito CPR and the human form in the binding of the small molecules 29, 59-ADP and diphenyliodonium chloride. Importantly, the low affinity of mosquito AgCPR for 29, 59-ADP and FMN and high affinity for diphenyliodonium chloride distinguishes it from human CPR, which offers the possibility of designing molecules to specifically inhibit mosquito CPR function. A potential advantage is that being encoded by a single gene that has a house-keeping role in furnishing electrons to all microsomal P450s, its inhibition is likely to be lethal to early development stages, giving it potential insecticidal properties in its own right. Notwithstanding the obvious challenges of designing molecules based on selective recognition of the NADPH binding pocket, these results suggest that detailed analysis of the structure and function of mosquito CPR may help in the development of 'smart' molecules that could specifically target mosquito vectors of major human diseases such as malaria. Finally, the importance of comparing AgCPR co-factor binding properties with other lentic invertebrate species must be emphasized as close similarities could limit the efficacy of inhibitors by collateral damage to non-target species. Figure S1 Multiple sequence alignments of the FMN binding region of A. gambiae CPR, A. minimus CPR, rat CPR (rCPR), S. cerevisiae CPR (yCPR) and the reductase domain of rat nNOS (nNOS). Residues involved in FMN binding (light grey) and stabilisation (dark grey) are highlighted. Sequences were aligned using ClustalW. (TIF)