Despite some success with liver microsomes, the DMPK field was in need of a more suitable model system for the structural and functional characterization of individual P450 enzymes (Poulos et al., 1982). This role would be filled by the microbial cytochrome P450CAM, a camphor-metabolizing monooxygenase that was identified in Pseudomonas putida (Gunsalus and Wagner, 1978; Poulos et al., 1982, 1987) (Table 1). This enzyme was purified for study prior to the advent of rDNA technologies and instead was isolated from bacterial culture according to a procedure optimized by Gunsalus and Wagner (Gunsalus and Wagner, 1978). P450CAM was the first CYP enzyme that had been purified in both soluble and crystalline forms, making it an ideal biologic model to study P450 structure and function (Gunsalus and Wagner, 1978; Poulos et al., 1982, 1985). High-resolution crystal structures were generated of P450CAM both bound to camphor and in the absence of its substrate, providing valuable insight into the substrate binding site and the structure-function relationship of the enzyme (Poulos et al., 1986, 1987). This crystallography data also elucidated the secondary structures and overall conformation of the polypeptide, illustrating important structural features of the enzyme such as the oxygen/heme binding pocket as well as a metal cation interaction that may stabilize active site formation (Poulos et al., 1987).

Expression of some recombinant P450 monooxygenases for structural or functional studies

These advancements in P450CAM visualization were also paired with amino acid sequencing techniques, further improving the detail of structural and functional characterization and enabling the comparison of P450CAM to eukaryotic P450 sequences (Poulos et al., 1987). Compared P450 sequences were found to be fairly homologous, with residues that form the oxygen binding area being highly conserved between species (Poulos et al., 1987). Amino acid sequencing also revealed that eukaryotic P450s uniquely possess a hydrophobic NH2 terminus, a domain that was speculated at the time to contribute to the membrane solubility observed in mammalian P450s (Poulos et al., 1987). While bacterial cytochrome enzymes like P450CAM are easily solubilized in the cytoplasm, eukaryotic P450 enzymes reside in the membranes of the endoplasmic reticulum and to a lesser extent the mitochondria (Lee et al., 1981). The functional role that this hydrophobic NH2 terminus plays in mammalian P450 membrane association and its effect on catalytic activity would later be elucidated using rDNA technologies (Oeda et al., 1985; Sakaguchi et al., 1987). P450CAM has also served as an effective model system to assess the role of accessory proteins, with crystal structures of the enzyme complexed to its redox partner putidaredoxin providing insight into the structural changes induced by putidaredoxin and how they may facilitate electron transfer (Hiruma et al., 2013; Tripathi et al., 2013). Even today, P450CAM is used as a model system to further investigate the allosteric roles of accessory proteins and how they may impact conformational dynamics (Poulos and Follmer, 2022).

Although P450CAM was purified from a gram-negative bacterium, its extensive study aided our understanding of mammalian P450 enzymes by enabling the thorough characterization of well-conserved domains and functions (White et al., 1984; Black and Coon, 1987; Poulos et al., 1987; Gonzalez et al., 1988a,b; Umeno et al., 1988; Gonzalez and Gelboin, 1992). With isolation of pure P450 enzymes from human tissue being an arduous and time-consuming process, this P450CAM model system was invaluable in establishing initial structure-function relationships and elucidating the basic catalytic mechanism conserved across nearly all P450s (Guengerich, 2020; Poulos and Follmer, 2022). P450CAM was also one of the first P450 enzymes cloned for recombinant expression, utilizing a plasmid-derived vector under strict promoter control to produce catalytically active P450 monooxygenase in P. putida and E. coli host systems (Koga et al., 1985; Unger et al., 1986). This was not the only microbial P450 enzyme investigated utilizing rDNA technologies, as the crystal structure of the prokaryotic form of 14α-demethylase (CYP51) present in Mycobacterium tuberculosis was also elucidated this way (Podust et al., 2001) (Table 1). While CYP51 exists in most species as a membrane-bound enzyme that is localized to microsomes, the Mycobacterium tuberculosis genome possesses a uniquely soluble form of the enzyme that was expressed in E. coli for easier determination of the crystal structure (Bellamine et al., 1999; Podust et al., 2001). This newfound structural information would inform rational drug design of improved antifungal agents that target CYP51.

Not long after the crystal structure and amino acid sequence of P450CAM were elucidated, rDNA strategies gained widespread usage in the study of mammalian CYP enzyme structure and function in drug metabolism (Fig. 1) (Sakaguchi et al., 1984; White et al., 1984; Oeda et al., 1985; Gonzalez et al., 1988a,b; Umeno et al., 1988; Larson et al., 1991b; Khan et al., 2002; Wester et al., 2002; Yu et al., 2002; Scott et al., 2003; Williams et al., 2003, 2004; Schoch et al., 2004) (Table 1). Although some P450 amino acid sequences had been elucidated through methods discussed earlier, the development of cloning, nucleotide sequencing, and DNA manipulation technologies would equip researchers with powerful tools to generate libraries of P450 cDNA sequences (Poulos et al., 1987; Gonzalez and Gelboin, 1992; Poulos and Follmer, 2022). Complementary DNA sequences for human P450 enzymes were generated using mature mRNA from liver and lung tissue lysates (Gonzalez and Gelboin, 1992). Purified mRNA was then used to generate cDNA strands via reverse transcriptase, which then underwent amplification by polymerase chain reaction to generate double-stranded DNA fragments (Patil et al., 2022). These gene segments were then ready be digested with complementary endonucleases and ligated into a vector for transformation of the amplified cDNA into a suitable host organism (Unger et al., 1986; Gonzalez and Gelboin, 1992; Patil et al., 2022). Due to observed sequence homology between species, human P450 sequences could be identified from these libraries through the use of cDNA probes and antibodies against rodent forms of P450 (Gonzalez et al., 1988b; Gonzalez and Gelboin, 1992). Once cDNA sequences of interest were identified and cloned using this approach, researchers set out to optimize the expression of these recombinant gene products in model organisms so their catalytic activities, substrate specificities, and structure-function relationships could be characterized (White et al., 1984; Oeda et al., 1985; Gonzalez et al., 1988a,b; Gonzalez and Gelboin, 1992; Schoch et al., 2004; Guengerich, 2020).

One important example that used cDNA-directed expression to illuminate P450 structure was the investigation of rabbit microsomal P450 and its cotranslational insertion into rough microsomes from dog pancreas (Sakaguchi et al., 1984, 1987) (Table 1). This would serve as an effective model of mammalian P450 integration into membranes of the endoplasmic reticulum, allowing researchers to interrogate the role of key structural features responsible for initiating cotranslational insertion and for conferring membrane solubility. Initial research demonstrated that cotranslational insertion of P450 could not occur without the presence of a ribonucleoprotein complex referred to as the signal recognition particle (SRP) (Sakaguchi et al., 1984). However, further investigation was required to identify the way in which microsomal P450 recognizes the SRP to influence P450 function and membrane topology. Hybrid cDNAs were generated to express chimeric forms of the P450 protein for study, with each mutant possessing unique modifications impacting the length and identity of the N-terminal amino acid chain (Sakaguchi et al., 1987). Of the several chimeric P450 proteins generated, only those possessing an NH2 terminus at least 29 AAs in length were able to cotranslationally insert into the membrane (Sakaguchi et al., 1987). This demonstrated that the insertion signal that enables SRP-mediated cotranslation to the membrane was contained within this 29 AA segment of the enzyme’s NH2 terminus (Sakaguchi et al., 1987). This research broadened the understanding of the NH2 terminus present in eukaryotic P450 forms, demonstrating the impact that NH2 truncation has on their topogenic function and membrane solubility (Sakaguchi et al., 1987).

Interestingly, recombinant research efforts expressing rat P450MC in Saccharomyces cerevisiae also generated a truncated form of the enzyme that lacked an extended NH2 terminus (Oeda et al., 1985). Spectral and catalytic analysis revealed almost identical properties between the truncated and nontruncated forms of the enzyme, demonstrating that this NH2 terminus is not required for P450 catalytic activity (Oeda et al., 1985) (Table 1). This knowledge would be widely applied to solubilize mammalian P450 through deletion of the NH2 terminus, allowing researchers to more easily purify and crystallize mammalian P450 enzymes for structural and functional characterization (Larson et al., 1991b; Hsu et al., 1993; Richardson et al., 1995; von Wachenfeldt et al., 1997; Williams et al., 2000; Podust et al., 2001; Wester et al., 2002, 2003, 2004; Scott et al., 2003; Williams et al., 2003, 2004; Schoch et al., 2004; Rowland et al., 2006; Hsu et al., 2018). This cDNA-directed approach would soon be used to express truncated rabbit CYP2E1 in E. coli using a pKKHC vector, an optimized derivative of the pBR322 plasmid discussed earlier (Larson et al., 1991a,b) (Table 1). These research efforts further supported the notion that the NH2 terminus is not necessary to retain P450 monooxygenase function, with recombinant truncated CYP2E1 possessing identical spectral and catalytic properties compared with full-length CYP2E1 isolated from rabbit liver microsomes (Larson et al., 1991a,b).

Similar success was also achieved in the recombinant expression of rabbit CYP2C3 as a soluble dimer in E. coli. The CYP2C3 gene insert with a modified NH2 terminus was digested with endonucleases NdeI and EcoRI and transformed into an optimized pCWOri+ vector that had been linearized with the same enzymes (von Wachenfeldt et al., 1997) (Fig. 2). This general procedure is highlighted in Fig. 2, serving as an example of heterologous P450 expression in E. coli. Additionally, the key components of this vector that allow for strong transgene expression and inducible control are summarized in Fig. 2. This pCWOri+ derived expression vector would also later prove useful in the recombinant study of CYP2C5 (von Wachenfeldt et al., 1997; Wester et al., 2002, 2003).

Not long afterward, two mammalian P450 enzyme crystal structures (CYP2B4, CYP2C5) were identified through recombinant expression in E. coli (Wester et al., 2002, 2003; Scott et al., 2003) (Table 1). Both enzymes were expressed in soluble form through truncation of the NH2 terminus, allowing for their crystallization and structural characterization. Crystallization in the presence and in the absence of substrate allowed investigators to examine conformational differences that can determine substrate recognition and catalytic function (Wester et al., 2003). These experiments were also one of the first to add a histidine tag to the C terminus of the recombinant protein to aid the purification process (von Wachenfeldt et al., 1997). This strategy would also be employed in the study of human CYP2C8, which was uniquely expressed in E. coli as stable homodimer complexed by two palmitic acid molecules (Schoch et al., 2004). These peripheral fatty acid binding sites can alter characteristics of the active site and have potential to modulate substrate binding as well as contribute to DDIs (Schoch et al., 2004) (Table 1). This observed dimerization of CYP2C8 in addition to its significantly larger active site cavity suggest a distinct metabolic function of this enzyme when compared with previously elucidated P450s. This makes sense when considering that the primary substrates of CYP2C8 are comparatively larger molecules such as taxol (Schoch et al., 2004). Further research crystallized CYP2C8 with a variety of different substrates to examine the role of protein flexibility and ligand-induced conformational changes in accomplishing substrate oxidation (Schoch et al., 2008).

The crystal structure of human CYP2C9 was also determined through heterologous expression of a NH2 truncated mutant, which revealed a similarly large active site capable of accommodating multiple warfarin molecules or an additional substrate of similar size (Williams et al., 2003) (Table 1). Given that CYP2C9 can be activated by some exogenous compounds including warfarin, it was hypothesized that the binding of some ligands may serve an allosteric function that promotes enzyme activity (Williams et al., 2003). Compared with CYP2C9, the crystal structure of recombinantly expressed human CYP3A4 revealed a slightly smaller active site as well as a characteristic cluster of phenylalanine residues above it, some of which are known to cooperate in enzyme selectivity and function (Khan et al., 2002; Williams et al., 2004). X-ray crystallography data were generated for CYP3A4 in the absence of ligand, in the presence of an inhibitor metyrapone, and with its substrate progesterone (Williams et al., 2004) (Table 1).

To determine how the properties and structures of CYP3A4 and CYP3A5 differ, both enzymes were recombinantly expressed in E. coli and complexed with a high-affinity substrate ritonavir (Hsu et al., 2018). Crystallography data for these enzymes revealed differences in the shape and plasticity of their active sites upon substrate binding (Hsu et al., 2018) (Table 1). This structural data may reflect notable differences in the substrate dynamics and catalytic accessibility of these enzymes, potentially leading to divergent metabolic pathways or varying enzymatic efficiencies for shared metabolic pathways (Hsu et al., 2018). Additionally, enzymes of the CYP3A family can be differentially expressed across ethnic populations, with allelic variations having a significant effect on enzyme abundance and metabolic capability (Hirota et al., 2004; Yamaori et al., 2005). Given that the CYP3A family is the largest single contributor to drug metabolism, illuminating the structural features and processes that govern their catalytic activity and specificity is critically important in designing safe therapies, predicting DDIs, and understanding individual differences in drug metabolism (Zanger et al., 2008).

One recent advancement in the structural understanding of this enzyme family is the solved crystal structure of neonatal CYP3A7, a critical enzyme in hormone homeostasis and pediatric drug metabolism (Li and Lampe, 2019; Sevrioukova, 2021). Structural characterization of CYP3A7 was previously lacking, as the enzyme proved difficult to crystallize in its wild-type form. Using site-directed mutagenesis, a recombinant CYP3A7 variant capable of crystallization was generated to allow for structural characterization and comparison with other CYP3A enzymes (Sevrioukova, 2021) (Table 1). In many ways rDNA technologies have provided unprecedented insight into the catalytic structure, conformational dynamics, and substrate specificities of the most integral drug-metabolizing enzymes. Even today we continue to build upon our understanding of P450 monooxygenases, with recent investigations providing additional context regarding the nature of active site plasticity and redox partner selectivity (Poulos and Follmer, 2022).

Another cytochrome P450 enzyme that performs a significant share of xenobiotic biotransformations is human CYP2D6, as it is estimated to metabolize up to 20% to 30% of commonly prescribed drugs. The crystal structure was solved once again through heterogenous expression of a soluble recombinant CYP2D6 in E. coli utilizing a pCWOri+ derived vector (Rowland et al., 2006) (Table 1). This experiment revealed a structure most closely resembling that of CYP2C9, and the crystallography data enabled further research efforts toward building a more detailed understanding of the structural basis of ligand binding and specificity (Rowland et al., 2006). This information proved especially critical in understanding the many biotransformations mediated by CYP2D6 and CYP2C9, as their allelic variants between ethnic groups and polymorphisms between individuals can have a dramatic effect on metabolic activity, drug disposition, and therapeutic outcomes (Higashi et al., 2002; Yu et al., 2002; Ingelman-Sundberg, 2005; Zhou, 2009a,b). The rDNA platform has been adapted to interrogate the impact of specific CYP polymorphisms present in the population, uncovering mechanistic rationale for unique phenotypes of drug metabolism. The ability to recombinantly express specific P450 polymorphisms has allowed researchers to quantify enzymatic differences in the presence of any substrate, providing critical knowledge in our understanding of pharmacogenomic variability between individuals and populations. This knowledge has proven invaluable in drug design and clinical practice, as CYP polymorphisms and genetic variants can result in altered ADME processes, DDIs, and unexpected adverse effects. One key example of this is observed in warfarin-treated individuals with CYP2C9 polymorphisms, who are at increased risk of major bleeding events and excessive anticoagulation associated with increased drug exposure (Higashi et al., 2002). CYP2C9 polymorphisms have also been shown to promote genotype-dependent DDIs when warfarin or other major substrates are coadministered with an inducer or inhibitor of CYP2C9 (Higashi et al., 2002; Gardin et al., 2019; Cheng et al., 2022). Research utilizing rDNA technologies has revealed many genotype-specific differences in PK function and individual drug disposition, providing understanding that is critical to rational drug design, clinical dose management, and DDI prediction. Elucidating these pharmacogenomic differences has been key in developing and maintaining safe therapies, and this knowledge will prove increasingly useful as pharmacogenomic testing of patients becomes more commonplace and cost effective (Goh et al., 2017; Sayer et al., 2021).

In addition to providing valuable insight into the most predominant drug metabolizing P450s, rDNA technologies have also been used to characterize P450 enzymes that biosynthesize critical endogenous compounds. One predominant example of this is human CYP17A1, which is known to catalyze the synthesis of androgens and other important steroids (Miller and Auchus, 2011). Because of this, CYP17A1 inhibition represents an attractive strategy to prevent androgen synthesis and treat castration-resistant prostate cancer. The crystal structures of CYP17A1 bound to two inhibitors were determined by recombinant expression of a truncated in version in E. coli (DeVore and Scott, 2012) (Table 1). These initial findings elucidated the structural features integral to CYP17A1’s catalytic activity, further enhanced soon afterward by crystal structures of CYP17A1 with all four of its major endogenous substrates and clinically relevant inhibitors (Petrunak et al., 2014, 2017). Structural characterization of CYP17A1 enabled by recombinant expression provided critical information in the rational drug design of CYP17A1 inhibitors. Additionally, rDNA technologies allowed investigators to develop more selective CYP17A1 inhibitors through assessment of off-target interactions with other steroidogenic P450 enzymes such as CYP21A2 (Fehl et al., 2018).

Another enzyme that mediates critical biosynthetic reactions is cytochrome P450 aromatase (CYP19A1), which is known to synthesize estrogens from androgens. While the crystal structure of CYP19A1 was first determined using aromatase purified from human placenta, subsequent research efforts recombinantly expressed truncated CYP19A1 to further its structural characterization and interrogate the functional roles of specific amino acid residues (Ghosh et al., 2009, 2010; Lo et al., 2013) (Table 1). This insight into the structure of CYP19A1 and its androgen-specific active site aided the intelligent design of aromatase inhibitor drugs used to treat hormone receptor-positive breast cancers (Ghosh et al., 2012). Indeed, this same approach has proven similarly useful in the characterization of many P450 enzymes that catalyze the biosynthesis of important endogenous compounds (Brixius-Anderko and Scott, 2019, 2021; Liu et al., 2022). These investigations have supported the rational design of selective inhibitor drugs, with recent examples being the characterizations of P450 11B1, P450 11B2, and P450 8B1, which catalyze the production of cortisol, aldosterone, and bile acid, respectively (Brixius-Anderko and Scott, 2019, 2021; Liu et al., 2022) (Table 1).