Introduction

Top Abstract Introduction Experimental Procedures Results Discussion References

The prediction of relative contributions of individual CYP isoforms to the overall metabolic rate of a drug biotransformation in human liver microsomes is important because this information allows the prediction of drug-drug interactions that may result upon coadministration with CYP isoform-selective inhibitors or inducers. This process is called quantitative reaction phenotyping and is generally accomplished using a combination of inhibition studies, correlation studies, and studies using heterologously expressed CYP isoforms (Clarke, 1998; Rodrigues, 1999). Due to the recent advances in molecular cloning and expression technology and the commercial availability of cDNA-expressed CYP isoforms, these heterologously expressed enzymes are being increasingly used as in vitro models of human liver microsomes in quantitative phenotyping (Crespi, 1995; Crespi and Penman, 1997; Crespi and Miller, 1999). When multiple enzymes catalyze a drug biotransformation, the utility of correlation analyses is limited unless they are used in a multivariate analysis format that generally requires a large sample set (Clarke, 1998) and may thus not be of value from a practical standpoint. Thus, the complementary approaches of inhibition studies and studies using heterologously expressed CYP isoforms are generally used in combination, and the utility of this combined approach of integrated quantitative reaction phenotyping has been demonstrated in numerous studies.

In the present investigation, we have characterized the biotransformation of the tricyclic antidepressant amitriptyline using lymphoblast-expressed CYP isoforms and applied the relative activity factor (RAF) approach (Crespi, 1995; Crespi and Penman, 1997; Crespi and Miller, 1999; Störmer et al., 2000; Venkatakrishnan et al., 2000a) in mathematically reconstructing the biotransformation rates in microsomes from a panel of 12 human livers. In addition, we have estimated the relative contributions of individual CYP isoforms to the overall rate of amitriptyline metabolism via the parallel pathways of N-demethylation and E-10 hydroxylation using chemical inhibition studies and by application of the RAF approach. The results using both approaches have been compared, and the possible sources of discrepancies between the two methods are discussed. These studies provide further validation of the RAF approach in scaling metabolic rates from heterologous expression systems to human liver microsomes, and the subsequent application of this approach in quantitative reaction phenotyping of multienzyme biotransformation pathways.

	Experimental Procedures

Top Abstract Introduction Experimental Procedures Results Discussion References

Materials. Liver samples, obtained from the International Institute for the Advancement of Medicine (Exton, PA), or the Liver Tissue Procurement and Distribution Service (University of Minnesota) were from 12 different transplant donors (L1-L12) with no known liver disease. The tissue was partitioned and kept at 80°C until the time of microsome preparation as described previously (von Moltke et al., 1993, 1994). Microsomal protein concentrations were determined using the bicinchoninic acid method (Pierce, Rockford, IL). Two of the 12 livers (L11 and L12) were CYP2C19-deficient, and one liver was CYP2D6-deficient. The phenotypic properties of the 12 livers with respect to the activities of CYPs 1A2, 2B6, 2C9, 2C19, 2D6, and 3A have been previously described (Venkatakrishnan et al., 1998b, 1999, 2000a).

Measurement and Prediction of Amitriptyline Biotransformation Rates. Incubations of amitriptyline with human liver microsomes and lymphoblast-expressed CYPs were performed using previously described methods (Schmider et al., 1995, 1996; Venkatakrishnan et al., 1998a). Reactions were performed for 20 min at 37°C in 50 mM KH₂PO₄ (pH 7.4) containing 5 mM MgCl₂, 0.5 mg/ml -NADP⁺, and an isocitrate/isocitric dehydrogenase NADPH regenerating system, in a total volume of 250 µl. Microsomal protein concentrations and reaction times were chosen to be in the linear range and to minimize substrate consumption. Incubations were terminated by the addition of 100 µl of acetonitrile and desipramine was added as the internal standard. The incubates were centrifuged and the concentrations of nortriptyline and E-10 hydroxy amitriptyline in the supernatants were measured by high performance liquid chromatography with UV detection at 214 nm. A 30-cm × 3.9-mm steel reverse phase C₁₈ µBondapak column (Waters, Milford, MA) was used and the mobile phase was a 70:30 mixture of 50 mM KH₂PO₄ and acetonitrile, at a flow rate of 1.5 ml/min.

(1)

In Vitro Profiling of Oxidative Amitriptyline Metabolism. The fraction of total oxidative metabolic rate that was attributable to N-demethylation was predicted as the ratio of predicted N-demethylation rate to the sum of the predicted N-demethylation and E-10 hydroxylation rates for each liver sample, under the assumption that N-demethylation and E-10 hydroxylation account for the total oxidative metabolism of amitriptyline. That is, the contributions of the minor pathways of Z-10 hydroxylation, 2-hydroxylation, and N-oxidation were ignored. The predicted N-demethylation fractions were compared with the observed values using correlation analyses (SigmaStat software; SPSS Inc.).

Prediction of Pharmacokinetic Clearance. The unbiased kinetic parameters of lymphoblast-expressed CYP isoforms were used to calculate estimates of in vitro intrinsic clearance for amitriptyline N-demethylation. Because CYP3A4-mediated amitriptyline N-demethylation was characterized by Hill enzyme kinetics, the intrinsic clearance was calculated based on the assumption of maximal autoactivation of the enzyme in vivo, using the method proposed by Houston and Kenworthy (2000). For all other CYPs, the intrinsic clearance was calculated as the V_max/K_m ratio (Houston, 1994). The net in vitro human liver microsomal intrinsic clearance was calculated as a linear combination of the intrinsic clearance terms contributed by each CYP isoform, weighted by their respective lymphoblast RAF estimates. This was then scaled up to in vivo intrinsic clearance using previously published values of scaling factors: 50 mg of microsomal protein per gram of liver, and 20 g of liver per kilogram of body weight (Carlile et al., 1999). The resulting estimated intrinsic clearance of amitriptyline via N-demethylation was used in conjunction with estimates of human hepatic blood flow (20 ml/min/kg) and the free fraction of amitriptyline in human plasma (0.057; Schulz et al., 1985) to predict intravenous clearance via N-demethylation using the well stirred and parallel-tube models (Obach et al., 1997; Thummel et al., 1997; Obach, 1999). Predicted N-demethylation clearance estimates using these liver models were compared with the in vivo N-demethylation clearance of amitriptyline (calculated as 60% of the total intravenous clearance) of 4 ml/min/kg (Schulz et al., 1985).

Chemical Inhibition Studies. Chemical inhibition studies were performed at substrate (amitriptyline) concentrations of 1.5 and 15 µM, at inhibitor concentrations specified below. The choice of inhibitor concentrations was based on the existing data in the literature (Newton et al., 1995; Bourrié et al., 1996; Ono et al., 1996; Ko et al., 1997; Koyama et al., 1997) and from personal experience in our laboratory. The objective was to maximize the extent of inhibition of the target CYP with minimal nonspecific inhibition of other CYP isoforms. With the exception of TAO, all other inhibitors were coincubated with substrate, microsomes, and the NADPH regenerating system. TAO being a mechanism-based inhibitor was preincubated with human liver microsomes and the NADPH regenerating system at 37°C for 20 min before initiation of the reaction with substrate. The selectivities of omeprazole (10 µM) and S-mephenytoin (500 µM) as CYP2C19 inhibitors were evaluated using individual lymphoblast-expressed CYP isoforms with amitriptyline as the substrate, at an amitriptyline concentration of 10 µM. At the concentration used, both omeprazole and S-mephenytoin inhibited CYPs 2C9 and 1A2 in addition to CYP2C19 (under Results). Although omeprazole inhibition studies were performed simply by coincubation with human liver microsomes and amitriptyline, the experimental design was modified as follows with S-mephenytoin as inhibitor. Control incubations were performed in the presence of -naphthoflavone (1 µM) and sulfaphenazole (10 µM), whereas CYP2C19-inhibited reactions contained these inhibitors in addition S-mephenytoin (500 µM). The fractional decrement of reaction velocity (FDV) by S-mephenytoin determined using this method should in principle be a better reflector of CYP2C19 contribution to the overall rate of amitriptyline N-demethylation due to the lack of confounding effects of CYP1A2 and CYP2C9 inhibition.

Prediction of Relative Contributions and Integrated Reaction Phenotyping. The relative contributions (f_i) of each CYP isoform to the overall rate of amitriptyline N-demethylation and E-10 hydroxylation were predicted using the kinetic parameters of lymphoblast-expressed enzymes and RAF estimates (eq. 2), as described earlier for rate predictions:

(2)

The model-predicted relative contributions were compared with the fractional decrement in reaction velocity by isoform-selective chemical inhibitors for each liver using correlation analyses (SigmaStat software; SPSS Inc.).

	Results

Top Abstract Introduction Experimental Procedures Results Discussion References

Kinetics of Amitriptyline Biotransformation by Lymphoblast-Expressed CYP Isoforms. Amitriptyline was N-demethylated by CYPs 1A2, 2B6, 2C8, 2C9, 2C19, 2D6, and 3A4, whereas CYPs 2A6 and 2E1 did not show any detectable activity. E-10 hydroxylation was observed in microsomes containing CYPs 2B6, 2D6, and 3A4, whereas the other isoforms did not show any detectable activity. All the identified amitriptyline N-demethylases and E-10 hydroxylases were characterized by enzyme kinetics (Figs. 1 and 2, respectively) and the parameters are given in Table 1. CYP1A2- and 2C9-mediated amitriptyline N-demethylation showed inhibition by substrate at high concentrations of amitriptyline (Fig. 1). However, models describing substrate inhibition did not yield physiologically acceptable parameter estimates. Thus, a Michaelis-Menten model was used with the data points at amitriptyline concentrations greater than 167 µM excluded from the nonlinear regression analysis.

View larger version (22K): [in this window] [in a new window]   Fig. 1.   Kinetics of amitriptyline (AMI) N-demethylation by lymphoblast-expressed CYP isoforms. Closed circles are experimental data points, and lines are fitted functions. Data points for CYPs 1A2 and 2C9 with unbound amitriptyline concentrations greater than 167 µM were excluded in nonlinear regression analysis for estimation of kinetic parameters, and are indicated as open circles. See Table 1 for kinetic parameters.

View larger version (12K): [in this window] [in a new window]   Fig. 2.   Kinetics of amitriptyline (AMI) E-10 hydroxylation by lymphoblast-expressed CYP isoforms. Closed circles are experimental data points, and lines are fitted functions. See Table 1 for kinetic parameters.

Rate Measurements and Predictions, and in Vitro Profiling of Oxidative Amitriptyline Metabolism. Strong statistically significant correlations were observed between the predicted and observed rates of amitriptyline N-demethylation and E-10 hydroxylation, and the N-demethylation fraction in the panel of 12 human livers. The predicted and observed N-demethylation rates were almost identical (Fig. 3, A and D), whereas the E-10 hydroxylation rates were generally underpredicted with a less than 2-fold error (Fig. 3, B and E). This resulted in a small overprediction of the N-demethylation fraction (calculated as the ratio of N-demethylation rate divided by the sum of N-demethylation and E-10 hydroxylation rates), especially at the lower amitriptyline concentration of 1.5 µM (Fig. 3, C and F).

View larger version (37K): [in this window] [in a new window]   Fig. 3.   Comparison of predicted (x-axis) and observed (y-axis) human liver microsomal rates of amitriptyline N-demethylation (A and D) and E-10 hydroxylation (B and E), and the N-demethylation fraction (C and F) at substrate concentrations of 1.5 µM (A-C) and 15 µM (D-F). Closed circles represent individual human liver samples and the open circle represents SUPERMIX. The dotted line is the line of identity, and the area between the dashed lines is the area with a less than 2-fold prediction error. When statistically significant (p < 0.05), R2 values indicating the strength of the correlation are provided, and the regression line is depicted as a solid line. The SUPERMIX data point was not included in correlation analyses.

Prediction of Pharmacokinetic Clearance. The RAF model-predicted value of human liver microsomal intrinsic clearance for amitriptyline N-demethylation was 35.7 µl/min/mg microsomal protein. The resulting scaled up estimated intrinsic clearance of amitriptyline via N-demethylation was 35.7 ml/min/kg of body weight. Predicted N-demethylation clearance estimates using the well stirred and parallel-tube liver models were 1.8 and 1.9 ml/min/kg, respectively.

Quantitative Reaction Phenotyping. At amitriptyline concentrations of 1.5 and 15 µM, the model-predicted relative contributions of each CYP isoform to overall N-demethylation and E-10 hydroxylation rate are provided in Table 3. These predicted relative contributions can then be compared with and used in conjunction with the results of chemical inhibition studies (Table 4) for integrated quantitative reaction phenotyping.

View larger version (25K): [in this window] [in a new window]   Fig. 4.   Inhibition of amitriptyline N-demethylation catalyzed by lymphoblast-expressed CYP isoforms, by omeprazole (A and B) and S-mephenytoin (C and D). A and C, effects of a fixed concentration of inhibitor (10 µM omeprazole and 500 µM S-mephenytoin, respectively) on amitriptyline N-demethylation by CYPs 1A2, 2B6, 2C8, 2C9, 2C19, 2D6, and 3A4 were examined. B and D, inhibition of lymphoblast-expressed CYP2C19 was studied, and IC50 analyses were performed using varying concentrations of the inhibitor. In all experiments the amitriptyline concentration was 10 µM.

View larger version (30K): [in this window] [in a new window]   Fig. 5.   Comparison of model-predicted relative contributions (x-axis) of individual CYP isoforms to amitriptyline N-demethylation rate at a substrate concentration of 1.5 µM, and the FDV by an isoform-selective chemical inhibitor of that CYP (y-axis). See legend to Fig. 3 for meanings of symbols and lines, and text for details of chemical inhibition methods.

View larger version (29K): [in this window] [in a new window]   Fig. 6.   Comparison of model-predicted relative contributions (x-axis) of individual CYP isoforms to amitriptyline N-demethylation rate at a substrate concentration of 15 µM, and the FDV by an isoform-selective chemical inhibitor of that CYP (y-axis). See legend to Fig. 3 for meanings of symbols and lines, and text for details of chemical inhibition methods.

View larger version (32K): [in this window] [in a new window]   Fig. 7.   Comparison of model-predicted relative contributions (x-axis) of individual CYP isoforms to amitriptyline E-10 hydroxylation rate at substrate concentrations of 1.5 µM (A and B) and 15 µM (C-E), and the FDV by an isoform-selective chemical inhibitor of that CYP (y-axis). See legend to Fig. 3 for meanings of symbols and lines, and text for details of chemical inhibition methods.

View larger version (22K): [in this window] [in a new window]   Fig. 8.   Comparison of the averaged results of the RAF-based method using heterologously expressed CYP enzymes (model-predicted relative contribution of individual CYP isoforms on the x-axis) and of chemical inhibition studies (FDV by chemical inhibitors of the respective CYP isoforms on the y-axis) as approaches to quantitative phenotyping of amitriptyline N-demethylation (A and B) and E-10 hydroxylation (C and D), at substrate concentrations of 1.5 µM (A and C) and 15 µM (B and D). CYP-inhibitor pairs shown are 1A2--naphthoflavone (1), 2C9-sulfaphenazole (2), 2C19-omeprazole (3), 2C19-S-mephenytoin (4), 2D6-quinidine (5), 3A4-TAO (6), and 3A4-ketoconazole (7). Data points are averaged values of the nine livers containing functional CYPs 2C19 and 2D6, for N-demethylation; and of the 11 livers containing functional CYP2D6 for E-10 hydroxylation. See Tables 3 and 4 for standard deviations and range of values. The dotted line is the line of identity, and the area between the dashed lines is the area with a less than 2-fold prediction error.

	Discussion

Top Abstract Introduction Experimental Procedures Results Discussion References

We have established using amitriptyline as a model substrate that the human liver microsomal rates of multienzyme drug biotransformations can be mathematically reconstructed using the enzyme kinetic parameters of heterologously expressed CYPs and scaling factors that incorporate the relative hepatic abundance of individual CYP isoforms and the differences in turnover number between the cDNA-expressed enzymes and their human liver microsomal counterparts. Amitriptyline was chosen as the model substrate considering the multiplicity of CYP enzymes that mediate its oxidative biotransformation.

The prediction of human liver microsomal rates using kinetic parameters of individual cDNA-expressed CYPs has been reported for omeprazole 5-hydroxylation (Yamazaki et al., 1997), 7-ethoxycoumarin O-deethylation (Shimada et al., 1999), chlorzoxazone 6-hydroxylation (Shimada et al., 1999), and for cyclophosphamide and ifosfamide 4-hydroxylation (Roy et al., 1999) and N-dechloroethylation (Huang et al., 2000). Our approach is similar to the approaches of Roy et al. (1999) and Huang et al. (2000) that used the relative substrate-activity-factor-based method (Roy et al., 1999; Huang et al., 2000) that is identical in principle to the RAF approach (Crespi, 1995; Crespi and Penman, 1997; Crespi and Miller, 1999). On the other hand, Yamazaki et al. (1997) and Shimada et al. (1999) used immunoquantified levels of individual CYPs to predict biotransformation rates in liver microsomes (Yamazaki et al., 1997; Shimada et al., 1999). As demonstrated previously, immunoquantified CYP levels may not be valid scaling factors to predict human liver microsomal rates from rates measured in heterologous expression systems, due to differences in turnover number of the enzyme in the two situations (Venkatakrishnan et al., 2000a).

As illustrated here for amitriptyline, this scaling approach can be used for in vitro metabolite profiling. The N-demethylation fraction for amitriptyline predicted using this approach (0.80 ± 0.09 and 0.84 ± 0.06 at substrate concentrations of 1.5 and 15 µM, respectively) was similar to that observed in human liver microsomes (0.71 ± 0.11 and 0.82 ± 0.06 at substrate concentrations of 1.5 and 15 µM, respectively). These values are greater than the average in vivo demethylation fraction of 0.6 after a single amitriptyline dose in humans (Rollins et al., 1980; Mellström et al., 1983, 1986), most likely explained by the failure to measure the minor oxidative pathways of Z-10 hydroxylation, 2-hydroxylation, and N-oxidation, and the direct conjugative pathway of N-glucuronidation in our in vitro system.

A potential application of the RAF approach is the prediction of pharmacokinetic clearance from intrinsic clearance measurements in heterologous expression systems (Ito et al., 1998). For amitriptyline, there was a 53% underprediction of in vivo N-demethylation clearance (predicted value of 1.9 ml/min/kg, compared with literature average value of 4 ml/min/kg). Amitriptyline is lipophilic and undergoes extensive hepatic uptake in humans, with a liver/plasma ratio of 36.4, based on autopsy studies (von Moltke et al., 1998). Thus, although a basic tenet of the widely used liver models for prediction of pharmacokinetic clearance is that free drug concentrations in plasma are in equilibrium with hepatocyte concentrations, this may not be true for drugs that are actively taken up and/or concentrated in the liver (Thummel et al., 1997). The clearance of lipophilic drugs such as amitriptyline, diphenhydramine, and diltiazem has been underpredicted in previous in vitro-in vivo scaling studies using human liver microsomes (Obach, 1999). Thus, the existing liver models may not adequately describe the relationship between in vitro intrinsic clearance and hepatic drug clearance in vivo, and the assumption that unbound plasma drug concentrations are reflective of hepatic enzyme-available concentrations is not generally valid.

The relative contribution of a CYP isoform to the overall rate of a multienzyme drug biotransformation pathway is a function of substrate concentration. CYP2C19 appears to be the major determinant of amitriptyline N-demethylation at therapeutically relevant drug concentrations, with added contributions of CYPs 2C8, 2C9, 1A2, and 2D6, whereas the contributions of CYPs 2B6 and 3A4 are predicted to be negligible (Table 3). However, at high substrate concentrations, the importance of CYP3A4 increases owing to its high hepatic abundance and sigmoidal kinetics. With E-10 hydroxylation, CYP2D6 is predicted to be the major catalyst at therapeutically relevant concentrations. However, CYPs 2B6 and 3A4 assume greater importance at higher substrate concentrations (Table 3). Thus, the mechanism of amitriptyline disposition after a therapeutic dose is likely to be different from that after a toxic overdose.

The predicted substrate concentration-dependent differences in relative contributions of CYPs 2C19 and 3A4 to amitriptyline N-demethylation are confirmed by the results of inhibition studies. Inhibition by the CYP2C19 inhibitors omeprazole and S-mephenytoin was greater at a substrate concentration of 1.5 µM compared with that observed at 15 µM, whereas the reverse was true for the CYP3A4 inhibitors TAO and ketoconazole (Table 4). These findings emphasize the importance of selecting low therapeutically relevant substrate concentrations in inhibition studies.

In general, good correlations were observed between the extent of inhibition of amitriptyline N-demethylation or E-10 hydroxylation by an isoform-specific inhibitor and the RAF model-predicted relative contribution of the target CYP isoform (Figs. 5-7). The scatter plots in Figs. 5 through 7 reveal some interesting properties of the inhibitors used in this study. Figures 5C and 6C examine the association of the model-predicted relative contribution of CYP2C19 with the extent of inhibition by 10 µM omeprazole. Although a significant positive association was noted, suggesting the inhibition of CYP2C19 by this inhibitor, a statistically significant positive intercept of the regression line was also noted, suggesting the lack of complete specificity of omeprazole toward CYP2C19. In the CYP2C19-deficient livers L11 and L12, omeprazole produced 25% inhibition of amitriptyline N-demethylation. This may be explained by the inhibition of CYPs 1A2 and 2C9 by omeprazole (Fig. 4).

Even S-mephenytoin, a selective CYP2C19 substrate, inhibited lymphoblast-expressed CYP1A2 and 2C9-catalyzed amitriptyline N-demethylation (Fig. 4). Although we measured inhibition by S-mephenytoin with CYPs 1A2 and 2C9 inhibited, this approach did not completely eliminate nonspecific inhibition, as is reflected by the nonzero y-intercept in Figs. 5D and 6D, and the approximately 20% inhibition of amitriptyline N-demethylation by livers L11 and L12. Contrary to our current findings, no nonspecific effects of S-mephenytoin were noted in human liver microsomes up to a concentration of 750 µM with imipramine as the substrate (Koyama et al., 1997). These data suggest that the effects of chemical inhibitors may be substrate-dependent.

The incomplete specificity of quinidine toward CYP2D6 at a concentration of 5 µM is demonstrated in Fig. 7, A and C. In liver L8 lacking CYP2D6, quinidine inhibited amitriptyline E-10 hydroxylation by 40 and 25% at substrate concentrations of 1.5 and 15 µM, respectively. In this liver, CYP3A4 should account for a major fraction of the hydroxylation rate. The substrate concentrations used are much lower than the CYP3A4 K_m of 69 µM for amitriptyline E-10 hydroxylation. Thus, quinidine, an alternative substrate of CYP3A4 (Nielsen et al., 1999), is expected to inhibit the CYP3A4 component of amitriptyline E-10 hydroxylation. In fact, amitriptyline E-10 hydroxylation and N-demethylation catalyzed by lymphoblast-expressed CYP3A4 were both inhibited by 30 to 40% by 5 µM quinidine at 10 µM amitriptyline. Thus, there is no absolute window of selectivity for a competitive inhibitor, and the extent of nonspecific inhibition will depend on the inhibitor concentration in relation to the K_m of the nontarget CYP(s).

Ketoconazole is a selective CYP3A4 inhibitor at concentrations less than 5 µM based on its effects on index reactions at their K_m values (Newton et al., 1995). At 1.5 µM amitriptyline, the specificity of ketoconazole may be questionable, explaining the overprediction of the CYP3A4 contribution to overall metabolic rate (Fig. 8), and the large y-intercept in Figs. 5G and 6G. In fact, at a low substrate concentration of 5 µM, ketoconazole (2.5 µM) significantly inhibited amitriptyline N-demethylation by lymphoblast-expressed CYPs 2C8, 2C9, and 2C19. Although less evident, even with TAO, the extent of inhibition was greater than the model-predicted relative contribution of CYP3A4 in some livers (Figs. 5F and 6F). The reason for this discrepancy is not clear and may be related to the sigmoidicity in CYP3A4 kinetics.

With the exception of ketoconazole, Fig. 8 suggests that the methods of chemical inhibition and the RAF approach using cDNA-expressed CYPs yield comparable estimates of the relative contributions of individual CYP isoforms to the rate of amitriptyline biotransformation via N-demethylation and E-10 hydroxylation. Another widely used method in reaction phenotyping is the use of selective antibody inhibitors of specific CYP isoforms (Shou et al., 2000). When designed carefully, inhibitory antibodies are often much more selective than chemical inhibitors, although selective inhibitory antibodies of CYP2C19 were not commercially available when this study was performed. A combination of the complementary approaches of selective enzyme inhibition and the use of heterologously expressed CYPs should be a useful approach to reaction phenotyping and should provide important information for the prediction of drug interactions with inhibitors or inducers of specific CYP isoforms. The advantages and pitfalls associated with both approaches need to be recognized in interpreting experimental data and in the inference of a reaction phenotype.