Introduction

In the liver microsomal system, two proteins working in tandem carry out the major step (Phase I) in redox metabolism of a wide variety of drugs and xenobiotics. The two enzymes involved are (a) cytochrome P450s (CYPs), a diverse family of heme-thiolate proteins (of ~50 kDa) and (b) cytochrome P450 reductase (CPR), a highly conserved diflavoenzyme (of ~78 kDa) [1, 2]. These two membrane proteins are distributed at CYP:CPR ratios of >20:1 in the phospholipid microenvironments of endoplasmic reticulae and microsomes [3].

The nature of interactions of these two proteins in situ (or in vitro) has been a matter of intense exploration. Coon’s group had reported that N-term portions of trypsinized CPR led to a decrease in the substrate hydroxylation, inhibition of reduced nicotinamide adenide dinucleotide phosphate (NADPH) oxidase (Nox) activity and gave perturbed Type-I binding spectra [4, 5 and several references mentioned therein]. They proposed that the inhibitory action exerted by the peptide was by an alleged preference of the CPR’s transmembrane segments (TMS) peptide for CYPLM2 (currently known as rabbit CYP2B series). Gum and Strobel had subsequently found that large excesses of the amphiphilic TMS peptide derived by steapsin treatment of CPR did not inhibit the reaction in a reconstituted system [6]. To explain their observations, they postulated that CYP and CPR’s TMS serve as a hydrophobic anchor and render the CPR catalytically active or mediate protein–protein interactions. Coon’s group also predicted the CPR’s N-term to contain a β-sheet structure, which was believed to be involved in P450 and phospholipid binding [7]. There were speculations that the N-terminal sequences of P450s differ in their amino acid composition due to evolutionary pressure as reflected by the statement “the N-terminal section becomes more viable during the elaboration of the phylogenetic tree in response to the evolutionary pressures to metabolize different types of substrate molecules” [8]. Such statements convey the impression that the P450 N-terminal region forms a part of the active site, and thus, amino acid changes in the N-term could influence drug metabolism reactions owing to altered substrate specificities of CYP450s. Müller-Enoch and Gruler [9] said that “the action of catalytic center where the chemical reaction takes place is not influenced by the membrane binding peptide, but the complexation process is influenced by the membrane binding domains.” Throughout the past four decades, majority of the workers in the area retained an impression that the N-termini of these two enzymes were crucial to the protein–protein complexation in the reaction mixture [3–13] and interpreted several experimental results to this effect.

The important hypotheses for the role of N-terminal TMS in CYP + CPR interactions/catalysis are depicted in Fig. 1, and are summarized below:

  1. 1.

    The N-termini aid in hetero-recognition/binding, which facilitates CPR-CYP protein–protein electron transfer, which is deemed essential for catalysis.

  2. 2.

    The N-termini (particularly, for CPR) form structural components of the active site in situ, which is obligatory for catalysis.

  3. 3.

    The N-termini play only passive anchors’ roles, bringing CYP and CPR in proximity in the phospholipid interface, which facilitates catalysis.

This communication adopts a comprehensive approach to explore all the hypotheses above and attempts to find support in terms of biochemical logic or direct experimental evidence to ratify each hypothesis. We arrive at a conclusion that the N-termini could function as passive anchors for localizing and bringing CYP and CPR together in the phospholipid microenvironment.

Fig. 1
figure 1

Probable hypothesis regarding the roles of N-term TMS in CYP + CPR reactions. The top-left panel (within the box) shows the fundamental components of the reaction system. The three hypotheses under investigation are appropriately labeled therein

Full size image

Materials and Methods

Analysis of the N-termini for TMS

The N-terminal sequences of most of the published major and other minor mammalian (especially human) microsomal CYPs and CPR were collected from protein databases on the internet. Thirteen different programs (available online on servers from various research groups working on transmembrane peptide biochemistry [14–26], details of which are shown in Table 1) were used to predict the N-term TMS.

Table 1 Determination of TMS of various mammalian CYPs and CPRs using programs (servers) available on internet [14–26]
Full size table

Probing Hetero-Recognitions Based on Available Phylogenetic and Structural Logic

TMhit server [27] was used for predicting N-terminal transmembrane helices “contact residue pairs” of CYP450 and CPR. This recently published methodology has been hitherto used for predicting contact residues of transmembrane helix (TMH) within a protein only. The consensus TMH sequences of CYP and CPR were taken in a single sequence pair so that its topology was in an “out–in–in–out” fashion within membrane. As a positive control, selected TMHs of aquaporin (1LDF), ATPase (1YCE) and bacteriorhodopsion (1C3W), which have known interactions, were also submitted to the same protocol. Some proteins containing helices which are known not to interact were sourced from the negatome database, and were used as negative controls (as exemplified by chains from cytochrome bc1 complex-1BE3). The results obtained from the server were then processed for better interpretation using an in-house written PERL script.

General Methodologies

Chemicals used in this study were purchased from reputed commercial suppliers and of analytical grade. Expression and purification of mammalian CYP and CPR was done as per published protocols [28, 29]. CYP (human CYP2C9 and rabbit 2E1) concentration was determined using the standard dithionite reduced carbon monoxide (CO) binding assay [30]. The intact CPR employed was a K56Q mutant to stabilize it from N-term proteolysis [31]. Flavin-based CPR quantification employed 10–20 μl of an appropriately diluted CPR stock in 10 μM ferricyanide at neutral pH. The concentration of CPR was determined by the optical density (OD) of the sample (OD 455–550 nm) divided by the molar extinction coefficient of 21,200 [32]. The four different CPR preparations (preps) were analyzed with sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) (10% polyacrylamide gel) to assess the extent of N-term scission. Densitometric analysis of the SDS-PAGE gel (Fig. A in the supplementary material) shows that the two (rabbit and human) CPR preps made from C41 and C-1a cells had only about 30 ± 2% (32 and 28%, respectively) intact CPR. In this study, they are also addressed as truncated (or broken) CPRa and CPRb, respectively. A prep of rabbit CPR from Top 3 cells had ~2/3 of the CPR intact and is referred to herein also as partly intact CPR. Another fresh prep of rabbit CPR from Top three cells showed no indication of scission of the N-term TMS, which is hereby referred to as intact CPR.

Reactions Employing Various CYP and CPR Preps

All reactions were carried out in open vials, with 100 mM potassium phosphate buffer, pH 7.4, incubated at 37°C. Unless otherwise mentioned, reconstituted systems had 10 μg/ml of 0.2 μm vesicles of dilauryl phosphatidylcholine (DLPC, Avanti Lipids). To 100 μl of the appropriate mixture of buffer stock and DLPC in distilled water, 20 μl each of the following four components was added: 1 μM human CYP2C9 (or 250 nM rabbit CYP2E1), 4 μM (or 1 μM) pertinent CPR, 2 μM (or 1 μM) cytochrome b 5 (if present, else 20 μl distilled water), 25 μM diclofenac [or 25 μM para-nitrophenol (pNP)] and the mixture was briefly equilibrated. Then, 20 μl of freshly prepared 5 mM NADPH was added and mixed to start the reaction. Therefore, CYP:cytochrome b 5 (when present):CPR was in the ratio of 1:2:4, respectively. Initial rates under steady-state conditions were monitored. Typically, 0.2–3 ml of the reaction mixture was taken initially and an appropriate volume taken out and quenched with a chilled reaction terminating solvent mixture. After centrifugation and filtration to remove any particulate matter, 20–50 μl of the sample was taken for product analysis by high performance liquid chromatography (HPLC). Details of HPLC methods to quantify 4′-hydroxydiclofenac (reaction product of CYP2C9 with diclofenac) and para-nitrocatechol (pNC, CYP2E1 reaction product of pNP) have been discussed in a recent communication from our lab [33]. Other specific details of concentrations and procedures are given in the respective legends.

Results

Determination of N-term TMS, Topology Prediction, and Preliminary Analyses

To predict the putative TMS, the sequences of the first 50 amino acids of the N-term were fed to 13 servers on the internet. As shown in Table 1, at times, the exact location of the predicted TMH varied among the different methodologies. However, they were statistically in agreement with each other. The extracted segments were predicted to have a helical topology by MPEx [26], and neither β-sheet nor β-strand was seen. A consensus segment of ~20 amino acids was thus derived for each CYP or CPR with N–C sequence in an out–in orientation. Table 2 (and Table A in the supplementary material) details the amino acid sequence and location of the predicted TMHs. CPR’s TMH was highly conserved across the species studied. However, wide variations were observed within families and subfamilies of CYPs. For example, (a) the number of predictable TMS varied—none in 1A1 to three segments within the first 100 amino acids for 2E1, (b) the specific location of the first N-term TMS, inclusive of the signal peptide, and the length of hydropathy breaks varied and (c) the distribution of residues therein varied (JALView-Muscle or JALView-ClustalW alignment analyses did not show significant consensus within first 50 amino acids of the different CYPs).

Table 2 TMHs of some mammalian CYPs and CPRs
Full size table

TMhit Predictions for Interhelical Binding (Fig. 2)

The contact residue pairs for each N-terminal TM CYP–CPR helix pair were predicted from the TMhit server. In each CYP–CPR helix pair, a probability of 1 is taken for the residue pair with the highest confidence in the prediction. As these pairs are scored on the basis of relative confidence, probability scores of <0.9 may be considered unimportant. In an ideal case, the complete interacting faces of the helices that are in contact would have high probability values which would be visualized as a diagonal in the contact map. This is clearly visible for the positive controls, shown in three panels on the top row, left side. In the case of non-interacting helices, these interactions would be sparse, and not aligned along the face of the helix, as is seen in the data presented for the negative control in the single panel on top row, right. The remaining three rows of panels are for test interaction predictions for human CYP–CPR, mouse/rabbit CYP–CPR interactions, and cross-species CYP–CPR interactions, respectively. Out of several CYP–CPR combinations studied, most of the CYP–CPR contact pair residues’ probabilities were usually <0.9, except a few exceptions. The test data show little tendency to form the diagonal pattern with high probability points (barring exceptions, like mCPR-mCYP1A2), indicating that there is little argument for a strong interaction. Apart from the CYP–CPR helical interaction predictions for major liver microsomal P450s, we had performed the same analyses for six other minor human P450s and two other rabbit and mouse P450s (Figs. B, C of the supplementary material). It is well-known that CPR works efficiently across species too. For example, rabbit CPR works quite well with human CYP2C9, a combination employed in this study. Prediction studies between human, rabbit, and mouse CPRs with alien CYPs 1A2 and 2E1 showed little scope for binding. (CYPs 1A2 and 2E1 were chosen as cross-species CYP representatives because these classes of CYPs were present in fair consistency across the species.)

Fig. 2
figure 2

TMhit results for TMH interactions between select mammalian liver microsomal CYPs and CPR. The top row shows the controls employed (three positives on the left and one negative on the right) and the remaining panels in the bottom three rows are test predictions between various mammalian CYPs and CPRs

Full size image

Mammalian CPR Preps Employed for In Vitro Studies in the Current Work

Equal amounts of four different CPR preps (as estimated by flavin spectrum) with varying extents of N-term integrity were used for carrying out reactions.

Effect of Cytochrome b 5 on Various CPR preps’ Activities with Human CYP2C9

All four CPR preps were separately employed for carrying out the hydroxylation of diclofenac in reconstituted CYP2C9 systems and the results are shown in Fig. 3. The intact N-term CPR showed highest activities in the control as well as the reaction employing cytochrome b 5. However, the point to be clearly noted was that the extent of integrity was not directly correlated to the specific reaction product in control or test reactions. In controls, the truncated CPR preps (CPRa, CPRb) and partly intact CPR, respectively, gave ~13% (lower than expected), 47% (higher than expected), and 36% (lower than expected) of the intact CPR’s activity. Incorporation of cytochrome b 5 gave surprising results—the activity of the two truncated CPR preps (CPRa and CPRb) and partly intact CPR was ~36% (roughly equal to expected, increased relative to control lacking cytochrome b 5 for the same reaction), 19% (lower than expected, decreased relative to control lacking cytochrome b 5 for the same reaction), and 26% (lower than expected, increased relative to control lacking cytochrome b 5 for the same reaction) of the intact CPR’s activity.

Fig. 3
figure 3

The effect of employing various CPR preps and cytochrome b 5 on CYP2C9-mediated hydroxylation of diclofenac. The initial concentration of reaction components were: [CYP2C9] = 100 nM, [CPR] = 400 nM, [cytochrome b 5, when present] = 200 nM, [diclofenac] = 25 μM, [NADPH] = 500 μM. The amount of specifically hydroxylated product 4′-hydroxydiclofenac formed after 8 min of reaction is plotted

Full size image

Effect of Cytochrome b 5 on Various CPR preps’ Activities with Rabbit CYP2E1

The four CPR preps were separately employed for carrying out the hydroxylation of pNP in reconstituted CYP2E1 systems. The results are shown in Fig. 4. Again, the intact CPR showed highest activities in the control as well as the reaction employing cytochrome b 5. It was clearly noted yet again that the extent of integrity was not directly correlated to the specific reaction product in control or test reactions. In controls, the truncated CPR preps (CPRa, CPRb) and partly intact CPR, respectively, gave 28% (roughly expected), 24% (roughly expected), and 43% (lower than expected) of intact CPR’s activity. Incorporation of cytochrome b 5 yet again gave surprising results—the activity of the two truncated CPR preps and partly intact CPR were 60% (higher than expected, increased relative to the control lacking cytochrome b 5 for the same reaction), 94% (higher than expected, increased relative to control lacking cytochrome b 5 for the same reaction), and 77% (roughly expected, increased relative to control lacking cytochrome b 5 for the same reaction) of the control reaction.

Fig. 4
figure 4

The effect of employing various CPR preps and cytochrome b 5 on CYP2E1-mediated hydroxylation of p-nitrophenol. The initial concentration of reaction components were: [CYP2E1] = 25 nM, [CPR] = 100 nM, [cytochrome b 5, when present] = 50 nM, [pNP] = 25 μM, [NADPH] = 500 μM. The amount of specifically hydroxylated product pNC formed after 8 min of reaction is plotted

Full size image

CPR preps’ Activities Obtained with CYP2E1 by Varying Substrate Concentration

Here, the activity of the soluble CPRa, (1/3 intact), CPRb (<1/3 intact) and partly intact CPR preps is compared with the activity of fully intact CPR. The results are shown in Fig. 5. At a substrate concentration of 5 μM, both the soluble and partly intact CPR preps gave roughly equal (~51 and 47%, respectively) activity, with respect to the control reaction incorporating the fully intact CPR. From the expected values, these values are much higher for the soluble prep and lower for the partly intact prep. At 15 μM pNP, the same preps gave 51 and 78% activity, respectively, with respect to the intact CPR reaction. These values are higher than the expected value for both preps. At 750 μM pNP, the preps showed 34% (roughly expected value) and 45% (lower than expected value) activity, respectively. At the other substrate concentration ranges, as expected, the partly intact CPR prep showed double the activity of soluble CPRa prep. However, the absolute percentages of values with respect to the control varied significantly. For example, the partly intact CPR prep gave activities ranging from 40 to 74% within the 15–750 μM substrate concentration regimes.

Fig. 5
figure 5

The effect of employing various CPR preps and varying pNP (substrate) concentration on CYP2E1-mediated hydroxlations. The initial concentration of reaction components were: [CYP2E1] = 15 nM, [CPR] = 60 nM, [cytochrome b 5] = 50 nM, [NADPH] = 500 μM. The data for the controls is shown for extended period so that it is clearly understood that the reaction regime is within steady-state conditions

Full size image

Discussion

Depending on the depth of the lamellum (formed with fatty acyl chain lengths of 12–22 carbons, translating to ~17–34 Å) and the inherent composition of the peptide backbone, a TM traversing helix’s length could vary. Usually, it ranges from 15 to nearly around 20 amino acid residues, with a typical length of 19–20 as an upright TM helix [34, 35]. Tryptophan and proline could serve as a surface anchor [36]. “Positive inside rule” says that segment of TMS having amino acids with positively charged R groups are present in the cytoplasmic side [37]. Topology prediction and modeling macromolecular interactions within the transmembrane domain remains a highly emerging and speculative field, constantly inundated with new ideas. Computational measurements for normal docking procedures involve maximizing the van der Waals and electrostatic contributions to the energy function. These are inadequate for an accurate prediction of protein–protein interactions, since the proteins may be fairly flexible. Currently, little factual knowledge about inter-protein TMH interactions is available. Since TMS are predominantly hydrophobic structures embedded in a hydrophobic phospholipid microenvironment [38], one could envisage the formation of a predictably stable complex by the chemical logic of disulfide bonds, salt bridges, hydrogen bonds, and pi–pi stacking [37]. Researchers in the area have also endorsed special cases like—serine zippers, (i, i + 4) sequences, Ile/Val/Leu patches, or small residue close packing [39, 40]. A simple qualitative survey of the TMS (as shown in Table 2) do not present strong arguments along the above lines for supporting the binding of CYPs and CPR within the same species or across species. Only, instances with scope for putative pi-stacking or hydrogen bonding could be found, a maximal of one each in a given CYP–CPR pairing.

Since crystal structures of some membrane multimeric proteins are available, we used the inter-chain binding within these proteins as a reference for inter-protein TMH-bindings. This is because the logic of binding would be similar in both the cases. For the purpose of predicting interacting residues between TMHs, we have used the TMhit server [27], which is a Kernel method (Support Vector Machine). It is built by using known interacting residues and non-interacting residues as a training data set and can assign a probability value to a contact pair. For further analysis, we plotted the results as a contact plot, from which interacting faces in the two helices could be clearly interpreted. The liver microsomal CYP–CPR binding predictions showed fewer contact points with lower probability values and a dispersion of the same (as opposed to the diagonal pattern for positive control), for an overwhelming majority of combinations studied. We also found that the results obtained for TMhit predictions for other minor liver microsomal P450s and CPR (six human CYPs-human CPR, two rabbit CYPs-rabbit CPR, and two mouse CYPs-mouse CPR combinations, Figs. B, C of the supplementary material) were similar to the predictions obtained for major P450s studied herein. This finding re-enforces the analyses drawn from the preliminary analyses of TMS of CYP and CPR that there is little evidence or reason to suppose that a conserved CPR’s TMS could rest the merit to interact selectively with the various CYPs’ TMS, within and across species. It had been noted that N-term truncated P450 expressed in bacteria and other systems could express functional P450 protein. Also, several reports affirm that the expression and activity of truncated P450s does not significantly alter their kinetic profiles or perturb their spectral properties [41–47]. In fact, the CYP2C9 employed in this study was expressed as a poly his-tag protein in E. coli for convenience in purification. A study has also revealed that human liver microsomal cytosolic fractions contained a catalytically active N-term truncated soluble CYP3A4 [48]. Therefore, the hypothesis that CYP’s N-term TMS could be involved in hetero-recognition with CPR’s N-term TMH (Hypothesis 1) is conclusively rendered invalid.

Next, we look into Hypothesis 2, which vouches for the involvement of CPR’s N-term in active site. Venkateswarlu et al. [46] had recently found that CPR’s N-term is not required for the CPR activity or reconstitution of CYP activity in situ, within a yeast system. However, this finding could be downplayed by arguments of heterologous expression incompatibilities [43]. Clearly, our in vitro experiments with the four CPR preps (Figs. 3, 4) show that the preps lacking the intact N-term could also afford significant functionality. What could be factually inferred from our experiments is that: (1) under certain experimental conditions, the presence of free N-term TMS in the reaction medium could be detrimental to activity and (2) mostly, intact CPR gave generally better activities (were better coupled) than truncated CPR. Therefore, an obligatory role for CPR’s N-term TMS can be downplayed. Therefore, Hypothesis 2 is effectively annulled.

Recently published work [49] on Nox, a membrane flavoprotein akin to CPR, is very relevant to the understanding of phenomenon seen in Figs. 3, 4, and 5. Working on Nox-4, Ralf Brandes’ group found that deletion of the E-loop or mutation of a conserved histidine within that stretch switched Nox-4 from an H2O2 generating system to an O .−2 generating system. Their work clearly implies that the E-loop, a peripheral component of the flavoprotein, is involved in diffusible reduced oxygen species (DROS, as exemplified by H2O2 and O .−2 ) dynamics. Also, some reports of perturbation of Nox activity by synthetic peptides bearing positively charged amino acid residues [50, 51] could be considered analogous to the observations of Minor Coon's group, who noted that N-term portions of trypsinized CPR leading to a decrease in the substrate hydroxylation [4, 5]. This is because the N-term proteolytically cleavable TMS of CPR possesses positively charged residues (His and Lys). We have recently invoked on DROS dynamics to explain for several novel and hitherto poorly understood phenomena in heme-enzyme reaction environments [32, 33]. We have shown that the incorporation of cytochrome b 5 and altering the concentration of the final substrate modulates the DROS dynamics in heme-enzyme reactions [32, 33]. Putting the findings on Nox and our works together, we could interpret our findings (Figs. 3, 4, 5) as an outcome of the DROS dynamics. CPR’s N-term could affect product formation because when free, it could alter the DROS dynamics in the reaction microenvironment. Incorporation of cytochrome b 5 or altering substrate concentration could potentially offset this effect, under certain reaction regimes [33]. Presence of redox-active additives in the enzyme preps could also lead to chaotic perturbances of the reaction outcome [52, 53], which could explain the difference in activity between the various preps.

To conclude, our findings and arguments support the Hypothesis 3, as shown in Fig. 1. The phospholipid bilayer, a relatively two-dimensional and constrained space, enhances the probability of catalysis/interactions by localizing and bringing the CYP and CPR in proximity to each other. We also provide indirect evidence for the impact and involvement of DROS in the reaction microenvironment. We are currently investigating the competing reactions within the microenvironment and chartering the roles of DROS in CYP–CPR-mediated catalysis.