Discussion

Induction of drug-metabolizing enzymes is an important mechanism underlying some drug-drug interactions. Increases in expression levels of important enzymes, such as CYP3A4, can cause increases in the intrinsic clearance of many drugs, resulting in decreases in their plasma concentrations and in some cases a loss of efficacy. Thus, it is important to be able to find agents capable of causing induction early in the drug research process to identify those compounds most prone to this activity and select alternate agents that will lack this activity. Several types of assays have been described for this purpose, such as PXR binding assays and reporter gene approaches using engineered systems (Sinz et al., 2006). However, the use of human hepatocytes in testing for induction remains the most widely accepted approach for making in vitro assessments of induction, especially for informing clinical drug interaction study strategies and inclusion of data in drug registration dossiers and product labels. Previously, there were practical challenges with the use of human hepatocytes for induction studies due to inconsistent supplies of cells that yielded acceptable responses and no manner of conducting cross-compound comparison of large numbers of drugs in individual sources. With the recent achievement of hepatocyte cryopreservation techniques, these limitations have been overcome because reliable supplies of cells that can be thawed and used over periods of time are now available. Cells from individual donors can be tested for induction response using known positive control compounds and those providing the most robust responses can be selected for routine testing of new compounds.

To measure induction of drug-metabolizing enzymes in vitro, several endpoints can be chosen. Activation of PXR yields increases in transcription of several genes for drug-metabolizing enzymes and drug transporters, and these increases in mRNA result in increased biosynthesis of the corresponding proteins. Because they are under the control of this receptor, it might be expected that the levels of these various messages and proteins would behave similarly to each other. However, this is not necessarily the case. In some cases, the changes in transcription and activity may not run in parallel because the drug-activating PXR could also directly affect the activity of the protein. For example, some drugs can cause an increase in transcription of CYP3A4 and several other drug-metabolizing enzymes, yet also be mechanism-based inactivators of CYP3A4. Thus, if CYP3A4 activity were used as the marker for induction activity, an erroneous conclusion could be made that such a compound will not cause clinical induction, yet the other induced but noninactivated enzymes will still have increased expression and activity. To avoid such a pitfall, measurement of the activities or transcripts of all inducible drug-metabolizing enzymes and transporters could be individually measured. Alternately, a more streamlined approach would be to measure the increase in mRNA of one of the induced enzymes as a marker for PXR activation and induction potential. In this report, a comparison of the induction of CYP3A4 mRNA, CYP3A4 catalytic activity, and the transcription of four other PXR-regulated drug-metabolizing P450s were measured in three lots of cryopreserved human hepatocytes to better define the most sensitive marker (CYP3A4 mRNA) for detecting induction in vitro and determining how well that marker distinguishes clinically meaningful inducers from noninducers.

The data clearly show that measurement of CYP3A4 mRNA levels represents the most sensitive marker for detecting induction in human hepatocytes. Several compounds were shown to cause marked increases in CYP3A4 mRNA, whereas the increases in catalytic activity were much lower. For example, aprepitant caused 7.6- to 16-fold increases in CYP3A4 mRNA, but activity levels actually decreased. This finding suggests that aprepitant does induce expression of drug-metabolizing enzymes and that one might expect a clinical drug interaction between aprepitant and other drugs. In the clinic, aprepitant has been shown to cause a decrease in midazolam clearance but induction of ethinyl estradiol clearance [(Majumdar et al., 2003) and (New Drug Application 021549 to the Food and Drug Administration)].

In addition, compared with mRNAs for other drug-metabolizing enzymes (CYP3A5, 2B6, 2C9, and 2C19), the mRNA for CYP3A4 increases more in all cases. This result suggests that if one measurement should be chosen for testing for induction, that measurement should be CYP3A4 mRNA. Genes for enzymes proposed to also be regulated by the CAR, such as CYP2B6 or CYP2C19, demonstrate increased transcription upon treatment with known CAR mechanism inducers such as phenobarbital and phenytoin. However, these inducers still caused an even greater increase in mRNA levels for CYP3A4, supporting the fact that just the measurement of CYP3A4 mRNA is adequate to determine the possibility that a compound will be an inducer. Thus, measurement of mRNA for these other drug-metabolizing enzymes should not be necessary to identify inducers, even if a compound were to only activate CAR.

Finally, the utility of in vitro induction measurements in the identification of clinically meaningful inducers was tested. As opposed to more sophisticated approaches for quantitatively predicting clinical drug interactions caused by induction (Persson et al., 2006; Ripp et al., 2006; Sinz et al., 2006; Fahmi et al., 2008a,b, 2009; Hariparsad et al., 2008; Kanebratt and Andersson, 2008; Shou et al., 2008), in this report a simple binary classification of inducers was attempted. Simpler approaches such as this are more amenable to early drug research efforts when hundreds of compounds must be assessed. In addition, an approach whereby a simple comparison of induction responses for new compounds with that of rifampin as a positive control had been previously proposed (Bjornsson et al., 2003; U.S. Food and Drug Administration, http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ucm072101.pdf). As with any pharmacological effect, an induction response will be dictated not only by the maximum possible effect but also by the relationship between the concentration in vivo and the intrinsic binding potency (Smith et al., 2007). Thus, the use of single, high concentrations of test compounds in in vitro induction assays may yield a high rate of false-positive results, because test concentrations far exceed those that are achieved in vivo. Nevertheless, testing of this binary classification was attempted. The results demonstrate that there are high false-negative rates if the cutoff based on fold induction relative to the positive control is applied. The best classification was observed when mRNA data for which the cutoff (4-fold) is based on absolute fold induction (Table 6) relative to the vehicle control are used, as no false-negatives were observed in two of the three lots tested. However, terbinafine, which is a weak clinical inducer suggested by the observed decrease of midazolam exposure of 24.5%, was not identified as an in vitro inducer in one of the lots tested. This observation confirms the need to evaluate new chemical entities with three different human hepatocytes lots as recommended by the U.S. Food and Drug Administration (http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ucm072101.pdf). It is also worth mentioning that notable differences were observed with the three human lots of hepatocytes used in this study. For example, lot RCP seemed to give rise to less false-positives based on mRNA with fold induction and the 20% cutoff criteria approaches.

Even though using the cutoff of 4-fold based on mRNA shows no false-negatives, there is a relatively high percentage of false-positive results (approximately 43%) (Table 7), if absolute fold induction is considered regardless of the in vivo plasma concentration. Some of the clinical noninducers such as rosiglitazone, nifedipine, omeprazole, nitrendipine, pioglitazone, and fluvoxamine show induction in vitro (Table 2). However, a drug such as rosiglitazone shows induction in vitro only at concentrations of 30 μM and above, which far exceed the actual C_max of 1.4 μM (Fahmi et al., 2008a). Another example is omeprazole, for which induction in vitro is observed at concentrations of 25 μM and above, which far exceed the plasma concentration of 0.68 μM. Therefore, it is crucial to design the in vitro induction study around expected in vivo plasma concentration.

In conclusion, the measurement of CYP3A4 mRNA as a marker for PXR-based induction serves as the most sensitive indicator for the potential for induction. This endpoint has the advantage over measurement of CYP3A4 activity in that compounds that activate PXR but also inhibit/inactivate CYP3A4 will not be missed. The induction of transcription of other drug-metabolizing enzymes including CYP2B6 and CYP2C occurred every time that CYP3A4 mRNA levels increased, albeit to a lesser extent, indicating that measurement of CYP3A4 mRNA is a sensitive marker for the induction of these other enzymes. This was the case even for enzymes and inducers that are known to also act via the CAR pathway. Although a binary approach to assignment of inducers versus noninducers from CYP3A4 mRNA tends to overpredict the frequency of clinically relevant inducers, this is probably due to the use of single, high concentrations that far exceed clinically relevant exposure levels.