Abstract
In drug discovery, establishing a correlation between in vitro potency and in vivo activity is critical for the validation of the selected target and for developing confidence in the in vitro screening strategy. The present study developed a competition equilibrium dialysis assay using a 96-well dialysis technique to determine the intrinsic Kd for 13 inhibitors of human liver glycogen phosphorylase a (GPa) in the presence of liver homogenate to mimic the physiological environment. The results provided evidence that binding of an inhibitor to GPa was affected by extra cofactors present in the liver homogenate. A good correlation was demonstrated between the in vitro Kd determined under liver homogenate environment and free liver concentration of an inhibitor at the minimum efficacious dose in diabetic ob/ob mice. This study revealed important elements (such as endogenous cofactors missing from the in vitro assay and free concentration at the target tissue) that contributed to a better understanding of the linkage between in vitro and in vivo activity. The approach developed here may be applied to many drugs in pharmacology studies in which the correlation between in vitro and in vivo activities for the target tissue (such as solid tumors, brain, and liver) is critical.
In the drug discovery process within the pharmaceutical industry, initial lead compounds are usually identified from high-throughput in vitro biological screens, for example, an inhibition assay against a target enzyme. It is often hoped that the in vitro potency (such as IC50) can be used to predict the in vivo pharmacological activity (such as EC50). However, discovery scientists often face a question: why doesn't an in vitro potent inhibitor work in vivo or only work at a much higher in vivo concentration? Establishment of a correlation between in vitro potency and in vivo activity is crucial for validation of the target enzyme and for achieving confidence in an in vitro screening strategy.
According to the fundamental free ligand hypothesis, the average free efficacious concentration at the steady-state in vivo should correlate with the intrinsic (unbound) potency determined from an in vitro assay. This hypothesis has been supported by many researchers (Wagner et al., 1965; Wagner, 1976; DeGuchi et al., 1992; Wright et al., 1996). In practice, however, this relationship is often obscured or confounded because of a variety of factors. For example, non-physiological conditions, involvement of nonspecific binding within in vitro systems, or a combination may yield an inaccurate estimate of the true intrinsic potency. In addition, complex pharmacokinetic/pharmacodynamic relationships arising from indirect effects or target site disequilibrium may result in the inappropriate determination of in vivo potency.
Several of the variables stated above confounded the initial establishment of a correlation between in vitro potency and in vivo activity for human liver glycogen phosphorylase a (GPa) inhibitors. Inhibition of liver GPa blocks the glycogenolysis pathway thereby leading to reduction of hepatic glucose production (Martin et al., 1998; Treadway et al., 2001). This approach may have potential to be useful for the treatment of type 2 diabetes mellitus. In this study, 13 reversible GPa inhibitors were tested against the human liver GPa and displayed IC50 values in the range of 23 to 460 nM. All these GPa inhibitors except one showed blood glucose lowering following an acute oral dose in fed diabetic ob/ob mice. However, no obvious correlation was observed between in vitro GPa enzyme IC50 values and any in vivo measurements such as the minimum efficacious dose (MED), or plasma or liver concentrations at the MED. One hypothesis to explain the poor correlation is that the IC50 values were determined under nonphysiological conditions because of missing or insufficient amounts of important cofactors or modulators (such as AMP, ATP, and so on, in this case) present in vivo, which have been shown to affect the binding affinity of a GPa inhibitor to GPa (Monanu and Madsen, 1987; Ercan-Fang et al., 2002a,b, 2005). As such, the initial focus of this work was to determine whether a good correlation could be revealed by obtaining GPa inhibitor binding affinity to GPa under more physiological conditions.
Theoretically, the in vivo activity should be driven by the amount of liver GPa bound to a GPa inhibitor, which is determined by the intrinsic Kd and the free GPa inhibitor concentration at the target site (liver). Accurate measurement of Kd under physiological conditions can be complicated and time-consuming (Romer and Bickel, 1979; Wright et al., 1996). In the present study, a robust method was developed to determine the intrinsic Kd of 13 GPa inhibitors to purified human liver GPa in the presence of liver tissue homogenate using a previously validated 96-well equilibrium dialysis apparatus (Cory Kalvass and Maurer, 2002; Banker et al., 2003). These conditions should mimic those in vivo better than the conventional assay (Martin et al., 1998) because most, if not all, cofactors are present in the liver homogenate. Good in vitro-in vivo correlation was found between the Kd determined in vitro in the presence of liver homogenate and the in vivo free liver concentration of a GPa inhibitor at the MED. The MED was used because the study design was aimed to detect in vivo active GPa inhibitors from a large panel of in vitro potent GPa inhibitors while minimizing the use of experimental animals. Determination of ED50 or ED90 requires many more experiments and thus is usually conducted only on the lead compound(s). Because very few GPa inhibitors had ED50 or ED90 determinations, the correlation with these parameters was not investigated in the present study.
Materials and Methods
Materials
All GPa inhibitors were obtained from Pfizer's proprietary sample bank (Groton, CT); and their identities by chemical names, including stereochemistry, and cLogP (calculated log octanol-to-water partition coefficient) are given in Table 1. References to the preparation and characterization of these compounds arel isted in Table 1, except for compounds 6, 9, and 13. Compounds 6, 9, and 13 are previously unreported, but they are close analogs, which were prepared by closely analogous procedures. All of these compounds are neutral and not ionizable at physiological pH. Human liver GPa was prepared as described previously (Martin et al., 1998). Solvents and other reagents were obtained from commercial sources and were of reagent grade or better.
Phosphorylase Enzyme Inhibition Assay (GPa IC50)
The assay for GPa IC50 was performed as described previously (Martin et al., 1998). In brief, human liver GPa (85 ng; ∼0.85 pmol) activity was measured in the direction of glycogen synthesis by the release of phosphate from glucose-1-phosphate at 22°C in 100 μl of buffer containing 50 mM HEPES, pH 7.2, 100 mM KCl, 2.5 mM EGTA, 2.5 mM MgCl2, 0.5 mM glucose-1-phosphate, and 1 mg/ml glycogen. Phosphate was measured at 620 nm, 20 min after the addition of 150 μl of 1 M HCl containing 10 mg/ml ammonium molybdate and 0.38 mg/ml malachite green. Test compounds were added to the assay in 5 μl of 14% dimethyl sulfoxide.
In Vivo Acute Plasma Glucose Lowering in the ob/ob Mouse
The acute hypoglycemic activity of a GPa inhibitor was determined using 7- to 8-week-old male C57BL/6J-ob/ob mice (The Jackson Laboratory, Bar Harbor, ME) housed five mice per cage under standard animal care practices. After a 1-week acclimation period, the animals were weighed, and 25 μl of blood was collected from the retro-orbital sinus before treatment. The blood sample was immediately diluted with 100 μl of saline containing 0.025% sodium heparin and held on ice for bioanalysis. Animals were assigned to treatment groups such that each group (n = 5 mice) had a similar mean for baseline plasma glucose concentration, and then animals were dosed orally with the vehicle alone or with a GPa inhibitor at dose ranging from 1 to 25 mg/kg. The vehicle consisted of 10% dimethyl sulfoxide/0.1% Pluronic P105 block copolymer surfactant (BASF Corporation, Parsippany, NJ) in 0.1% saline without pH adjustment or neat polyethylene glycol 400 (Sigma-Aldrich, St. Louis, MO). Mice were then bled from the retro-orbital sinus 1 and 3 h postdose for determination of plasma glucose levels and exposure. Mice were terminated following the last bleeding, and liver samples were collected for exposure. Freshly collected blood samples were centrifuged for 2 min at 10,000g at room temperature. The supernatant was analyzed for glucose by the Abbott VP (Abbott Laboratories, Diagnostics Division, Irving, TX) and VP Super System Autoanalyzer (Abbott Laboratories), using the A-Gent Glucose-UV test reagent system (Abbott Laboratories). Alternatively, plasma glucose was measured using a Roche/Hitachi 912 Clinical Chemistry Analyzer (Roche Diagnostics, Indianapolis, IN). Hypoglycemic activity of the test compounds was determined by statistical analysis (unpaired t test) of the mean plasma glucose concentration between the test compound group and vehicle-treated group. The MED was determined by the lowest dose at which significant plasma glucose lowering was observed in the test compound group. Aliquots of plasma samples were also analyzed for exposure. All animal procedures involved the humane care and use of animals, were performed within an Association for Assessment and Accreditation of Laboratory Animal Care International-accredited facility, and were approved by the Pfizer Global Research and Development-Groton/New London Laboratories Institutional Animal Care and Use Committee.
Liver and Plasma Sample Preparation for Concentration Determination
The liver tissues from studied animals were homogenated by adding ∼0.2 g of liver tissue to ∼1 ml of the ice-cold buffer (1:5 dilution), composed of 50 mM BES, 4 mM NaCl, 1 mM EDTA, 0.5 mM dithiothreitol, and 0.02% sodium azide, in a glass tube sitting on ice followed by sonication via a probe homogenizer (twice for 10–15 s each time at maximum speed). Ten microliters of liver homogenate (0.2 g/ml) was precipitated using 200 μl of methanol-acetonitrile (1:1). The calibration curve was prepared by mixing 10 μl of a control tissue homogenate (0.2 g/ml), 10 μl of a standard in methanol-acetonitrile, and 190 μl of methanol-acetonitrile. The standard samples (nine concentrations per standard curve; dilution factor, 5) were processed as the unknowns. The plasma samples from studied animals were diluted 5-fold with saline upon bleeding. Fifty microliters of the plasma sample was precipitated by adding 100 μl of methanol-acetonitrile. The calibration curve was prepared by mixing 10 μl of control plasma, 40 μl of saline, and 10 μl of a standard in methanol-acetonitrile followed by precipitation with 90 μl of methanol-acetonitrile. The standard samples (nine concentrations per standard curve) were processed exactly as the unknowns. All samples were then vortexed and centrifuged, and the supernatant was analyzed by a liquid chromatography with tandem mass spectrometry (LC/MS/MS) assay. The calibration curve was obtained by fitting linear least-squares regression.
Equilibrium Dialysis A
A validated two-chamber 96-well equilibrium dialysis apparatus was used (Banker et al., 2003) (Fig. 1, top). The buffer used in dialysis and liver homogenate was the same as described above. The two chambers (150 μl each) were separated by a preconditioned Spectra-Por Number 2 membrane with molecular mass cut-off of 12 to 14 kDa. One chamber contained purified human liver GPa at 0.2 mg/ml (∼2 μM, molecular mass = ∼97 kDa for the monomeric form (Treadway et al., 2001), and the other chamber contained an ob/ob mouse liver homogenate at 0.2 g/ml (prepared as described above). To facilitate rapid equilibrium, the compound was added to both chambers at a concentration of 250 ng/ml (∼0.5 μM). The sealed apparatus was incubated in a 37°C water bath with a gentle shaking at 60 rpm. Equilibrium was achieved between 4 and 7 h without detectable degradation for tested compounds (recovery ranged from 80 to 110%). At the end of dialysis, 90 μl of GPa sample and 10 μl of liver homogenate sample were taken from the dialysis apparatus to separate HPLC vials containing 100 μl of methanol-acetonitrile (1:1). Ten microliters of control liver homogenate (0.2 g/ml) and 90 μl of unfortified GPa solution (0.2 mg/ml) were added to the GPa sample vial and the liver homogenate sample vial, respectively, to yield an identical matrix. All samples were then vortexed, centrifuged, and the supernatant was analyzed by an LC/MS/MS assay to determine the total concentration of the inhibitor in the GPa chamber ([I]t-GPa) and the total concentration of the inhibitor in the liver homogenate chamber ([I]t-liver).
Equilibrium Dialysis B
The assay was carried out as described above in Equilibrium Dialysis A, with the exception that the GPa solution was replaced by the buffer (Fig. 1, middle). The compound was added to the liver homogenate chamber at 500 ng/ml (∼1 μM). Equilibrium was achieved by 7 h of incubation. After dialysis, 10 μl of the liver homogenate sample and 90 μl of the buffer sample were taken from the dialysis apparatus to separate HPLC vials containing 100 μl of methanol-acetonitrile (1:1). Ninety microliters of unfortified buffer and 10 μl of control liver homogenate (0.2 g/ml) were added to the liver homogenate sample vial and the buffer sample vial, respectively, to yield an identical matrix. All samples were then vortexed, centrifuged, and the supernatant was analyzed by an LC/MS/MS assay. The concentrations in the buffer chamber [equal to the free concentration, ([I]f)] and in the liver homogenate chamber [equal to the total concentration, [I]t-liver] were determined, respectively.
Equilibrium Dialysis C
The assay was carried out as described above for equilibrium dialysis A, with the exception that the liver homogenate was replaced by the buffer (Fig. 1, bottom). The compound was added to the GPa chamber at 500 ng/ml (∼1 μM). Equilibrium was achieved by 7 h of incubation. After dialysis, 10 μl of the GPa sample and 90 μl of the buffer sample were taken from the dialysis apparatus to separate HPLC vials containing 100 μl of methanol-acetonitrile (1:1). Ninety microliters of unfortified buffer and 10 μl of unfortified GPa solution (0.2 mg/ml) were added to the GPa sample vial and the buffer sample vial, respectively, to yield an identical matrix. All samples were then vortexed, centrifuged, and the supernatant was analyzed by an LC/MS/MS assay. The concentrations in the buffer chamber ([I]f) and in the GPa chamber ([I]t-GPa) were thus determined, respectively.
Plasma Protein Binding
The assay was carried out as described previously using the two-chamber 96-well equilibrium dialysis apparatus (Tracey et al., 2004). In brief, fresh plasma samples from ob/ob mice were spiked with the test compound to achieve a concentration of 500 ng/ml. An aliquot of a fortified plasma sample (150 μl; n = 4–6) was loaded into one chamber and dialyzed against 150 μl of sodium phosphate buffer, pH 7.4, in another chamber at 37°C for 5 h. At the end of the dialysis period, 20 μl of the dialyzed plasma and 90 μl of the buffer were transferred to HPLC vials containing 100 μl of methanol-acetonitrile (1:1). Control buffer (90 μl) was added to the vial containing the plasma sample, and 10 μl of control plasma was added to the vial containing the buffer sample. All samples were then vortexed and centrifuged, and the supernatant was analyzed by an LC/MS/MS assay. The plasma free fraction was estimated by the ratio of drug concentration in the buffer chamber to the drug concentration in the plasma chamber.
LC/MS/MS Assay
Two Shimadzu LC-10ADVP binary pumps with a 10-μl static mixer and a PE-Sciex API3000 triple quadrupole mass spectrometer (PerkinElmer Sciex, Concord, ON, Canada) were used in all experiments. An aliquot of sample (5 μl) was injected onto a Phenomenex 40 × 2mm5-μm C18 column maintained at 37°C with a run time of ∼3 min. The analyte was eluted at 0.5 ml/min flow rate with a linear gradient program consisting of methanol (pump A, 5–95% ramping) and 10 mM ammonium acetate (pump B, 95–5% ramping). The column effluent was analyzed using a Turbo Ionspray source at 500°C of API3000. Compounds were measured using multiple reaction monitoring with positive ionization and retention time between 1 and 4 min.
Theoretical
Equilibrium Dialysis. In the equilibrium dialysis assay A, the ratio of the fortified GPa concentration to the liver homogenate concentration was 1:1000 (w/w). Under this ratio, the amount of endogenous mouse GPa in the liver homogenate chamber was approximately equal to the amount of fortified human GPa in the GPa chamber (∼0.1% of whole wet liver tissue; J. L. Treadway and R. K. McPherson, unpublished data). In addition, small cofactors (smaller than 12 kDa) released from the liver tissue can pass through the dialysis membrane into the GPa chamber. At equilibrium, the [I]f should be same in both chambers. Therefore, the [I]t-liver in the liver homogenate chamber times the free fraction of the inhibitor in the liver homogenate (fuliver) should be equal to the total inhibitor concentration ([I]t-GPa-liver) in the GPa chamber times the free fraction of the inhibitor in the GPa solution (fuGPa-liver; see eq. 1): The [I]t-GPa-liver and [I]t-liver were directly measured from samples collected from the equilibrium dialysis assay A as described above. The fuliver was estimated from the equilibrium dialysis assay B by the ratio of the inhibitor concentration in the buffer chamber to the inhibitor concentration in the liver homogenate chamber (see eq. 2). Therefore, the fuGPa-liver can be estimated using eq. 3: In the equilibrium dialysis assay C, the inhibitor is bound to the GPa in the buffer without presence of cofactors released from the liver homogenate. The free fraction of an inhibitor, fuGPa-buffer, is estimated by the ratio of the inhibitor concentration in the buffer chamber to the inhibitor concentration in the GPa chamber (see eq. 4).
Kd Determination. The Kd of an inhibitor to GPa under the condition of equilibrium dialysis A (KdGPa-liver) is defined as in eq. 5, where [GPa]f is the free GPa concentration and [I]b-GPa is the inhibitor concentration bound to GPa. Because under this condition [I]b-GPa is equal to [I]t-GPa-liver times (1–fuGPa-liver) and the ratio of [I]f to [I]t-GPa-liver is equal to fuGPa-liver, the KdGPa-liver can be derived by eq. 6. The [GPa]f is derived by eq. 7 where [GPa]t and [GPa]b are the total and bound GPa concentrations, respectively. The [GPa]b should be the same as the [I]b-GPa; therefore, introducing eq. 7 into eq. 6 gives eq. 8. Thus, KdGPa-liver can be estimated by using eq. 8.
Estimation of the Inhibitor Concentration Bound to GPa ([I]b-GPa)
The [I]b-GPa in a liver homogenate can be estimated based on eq. 9. Assuming the ratio of ([I]b-GPa)/([I]t-liver) determined from the in vitro dialysis assay is same as in vivo at the MED, [I]b-GPa in vivo can be estimated based on the in vitro ratio times the total inhibitor concentration in the liver in vivo (, eq. 10), which can be measured ex vivo after administration of an inhibitor to the animal as described above.
Determination of Free Fraction of an Inhibitor in Undiluted Liver Tissue (uFuliver) and Free Liver Concentration. The free fraction in an undiluted liver tissue (uFuliver) was estimated using eq. 11 previously described by Kalvass and Mauer (2002). The fuliver was determined as described in equilibrium dialysis B, and D was the dilution factor (D = 5 in the present study). The free liver concentration of a GPa inhibitor in vivo was estimated by the liver concentration times uFuliver.
Results
In Vitro and In Vivo Activities. The in vitro IC50 values of the panel of GPa inhibitors tested in a purified human liver GPa assay are summarized in Table 2. The MED was determined following a sequential oral administration in diabetic ob/ob mice starting from a dose of 25 mg/kg to a dose approximately 2-fold lower than the MED. The total in vivo liver and plasma concentrations of GPa inhibitors at MED are recorded in Table 2. The plasma concentration listed in Table 2 was either at 1 or 3 h postdose, whichever the lower concentration associated with the glucose lowering at the time. The IC50 values ranged from 0.023 to 0.46 μM. No correlation was found between in vitro IC50 and in vivo activities determined by MED (Table 2) or total plasma or liver concentrations (R2 values were all less than 0.05; Fig. 2). There was no obvious correlation either between the IC50 and the free plasma or liver concentrations (R2 are 0.497 and 0.184, respectively; Fig. 3).
Determination of the Dissociation Constant of a GPA Inhibitor to GPa in Presence of Liver Homogenate (KdGPa-liver). The KdGPa-liver in the liver homogenate environment (equilibrium dialysis A) was estimated based on eq. 8 for all 13 GPA inhibitors after obtaining values for fuGPa-liver (Table 3), [GPa]t (0.2 mg/ml ≈ 2 μM), fuliver, and [I]t-liver. The estimated KdGPa-liver values are present in Table 3 and ranged from 0.05 to 0.92 μM. The relationship between the KdGPa-liver and the free liver concentration of a GPa inhibitor in vivo at the MED is illustrated in Fig. 4. However, there was no obvious correlation between fuGPa-liver and fuGPa-buffer for the 13 GPa inhibitors (Table 3), suggesting the binding of a GPa inhibitor to GPa in the presence of liver homogenate differs from the binding in the buffer.
Inhibitor Bound to GPa versus Nonspecific Binding to Liver Tissue. The fuliver was determined for all compounds (data not shown), and the extrapolated uFuliver is listed in Table 3. The measured ratios of total drug concentration in the GPa chamber to that in the liver homogenate chamber ([I]t-GPa/[I]t-liver) are summarized in Table 3 and ranged from 0.076 to 0.94. The concentration ratio of an inhibitor bound to GPa over total inhibitor concentration in the liver homogenate ([I]b-GPa/[I]t-liver) estimated from eq. 9 was listed in Table 3. The [I]b-GPa in vivo in the liver of an ob/ob mouse for all of the GPa inhibitors at their respective MED (estimated by eq. 10) was found to be relatively constant with mean ± coefficient of variation of 1.42 ± 37% μM. The relationship between the [I]b-GPa/[I]t-liver ratio and the inhibitor concentration in the liver of an ob/ob mouse 3 h postdose at the MED is shown in Fig. 5.
Discussion
The techniques used in the present study combined a published method for distribution dialysis (Bickel et al., 1987; Clausen and Bickel, 1993) and a 96-well equilibrium dialysis apparatus (Banker et al., 2003). These techniques have been used extensively to determine drugs binding to plasma protein, tissue protein, or any macromolecules (Pacifici and Viani, 1992; Cory Kalvass and Maurer, 2002). Many of the drug-binding studies were aimed at predicting drug distribution in vivo (Khalafallah and Jusko, 1984; Barre et al., 1988; Clausen and Bickel, 1993) or to understand pharmacokinetic consequences (Wilkinson, 1983; Fichtl et al., 1991). Similar studies were conducted to understand pharmacokinetics/pharmacodynamics and to predict drug effect based on receptor binding at the target (Proost et al., 1996) or to determine free drug concentration at the target site (Cory Kalvass and Maurer, 2002). However, few publications have revealed a correlation between in vitro enzyme binding potency and in vivo drug activity for multiple compounds in a drug discovery setting.
It was hypothesized that the previous method to determine GPa IC50 (Martin et al., 1998) was under nonphysiological conditions and that the binding potency of an inhibitor to GPa might be influenced by the assay conditions. A human hepatocyte (SK-HEP1) assay was used to determine glycogenolysis inhibition in the cells for GPa inhibitors. A marked right shift in IC50 values was observed for all GPa inhibitors compared with values determined from the GPa enzyme assay. No obvious correlation to in vivo potency was established (data not shown). One plausible explanation is that the protein binding of the compounds in this hepatocyte assay was not determined, and this may differ between the two systems. The free IC50, if measured, might improve the correlation. Another possibility is that all transporters, which may mediate uptake and efflux of GPa inhibitors (vide infra) may not be expressed and function the same in SK-HEP1 cells as in vivo. To fully understand factors in hepatocytes that contribute to the discordance to the in vivo potency is important and deserves additional study.
The effect of the presence of liver homogenate on the binding of a GPa inhibitor to GPa was clearly evident from the present study. As shown in Table 3, the free fraction of a GPa inhibitor in the dialysis buffer (fuGPa-buffer), the condition used in the previous method for IC50 determination, does not correlate with the free fraction of a GPa inhibitor in the presence of liver homogenate (fuGPa-liver). The latter should be closer to the in vivo physiological conditions. Ercan-Fang et al. (2005) also reported that the inhibition potency of a GPa inhibitor to GPa was modulated by endogenous effectors and that the magnitude of shift of in vitro IC50 in the presence of all effectors varied for different GPa inhibitors. Therefore, to accurately measure the Kd of GPa inhibitor to GPa enzyme that predicts in vivo activity, all cofactors at physiological concentrations should be included in the in vitro assay. The KdGPa-liver was obtained under the conditions designed to be close to those in vivo and correlated very well for the in vivo activity determined by free inhibitor concentration in the liver of an ob/ob mouse 3 h postdose at their respective MED for all the GPa inhibitors tested in the present study (Fig. 4).
The Kd determines the binding affinity of a ligand to an enzyme theoretically independent of the enzyme concentration. In the present study, the KdGPa-liver was determined using a single GPa and liver homogenate concentration. The liver homogenate concentration (200 mg/ml) used in the competition dialysis assay was under a nonsaturation condition (data not shown). The concentration ratio of GPa to liver homogenate was selected for equilibrium dialysis A based on the estimated ratio in a mouse liver (GPa ∼0.1% of wet liver weight). A good correlation (R2 = 0.776) shown in Fig. 4 between the free inhibitor concentration in the liver of an ob/ob mouse 3 h postdose at the MED and the KdGPa-liver supports the validity of the KdGPa-liver measurement used in the present study.
It is worthwhile to point out that the measured free inhibitor concentrations in the liver at the MED were all lower than the KdGPa-liver (approximately 20% of the KdGPa-liver) for all tested GPa inhibitors, consistent with estimated low percentage of GPa bound to GPA inhibitor (1.4 μM; see Results) relative to total GPa concentration (∼10 μM; J. L. Treadway and R. K. McPherson, unpublished data) at the MED. This is expected because only minimum glucose lowering (∼20% of Emax) was observed at the MED. To have the maximum effect on GPa inhibition, the free inhibitor concentration in liver should be ≥10 times Kd wherein >90% GPa is bound. This hypothesis was supported by a separate study, in which compound 4 administered via portal vein infusion to normal male Sprague-Dawley rats blocked 69 and 100% of glucagon-induced glucose excursion in rats at steady-state free plasma concentrations of 1.5 and 2 μM (7.5 and 10 times KdGPa-liver, respectively; T. Checchio and L. J. Yu, unpublished data).
Theoretically, the free liver concentration should be equal to free plasma concentration under equilibrium conditions if the liver distribution is driven by passive diffusion and no significant metabolism occurs. However, a good correlation was observed only for the KdGPa-liver with the free liver concentration but not with the free plasma concentration at the MED. This discrepancy may be explained either by nonequilibrium conditions at the sampling times (1 or 3 h postdose) from the in vivo study or by some of GPa inhibitors possibly serving as substrates for liver transporter proteins (Fichtl et al., 1991; Cory Kalvass and Maurer, 2002). These two possibilities may also contribute to the lack of a good agreement between free liver and free plasma concentrations (Fig. 3). Both possibilities could not be rigorously examined in the present study because of the practical limitations of a drug discovery setting and available tools for all liver transporters, although some GPa inhibitors were found to be P-glycoprotein substrates (data not shown). Therefore, this complexity may be avoided by using free concentration in liver to correlate with an in vitro activity.
Last, the present study demonstrates an excellent inverse correlation (R2 = 0.862; Fig. 5) between the [I]b-GPa/[I]t-liver ratio and the liver concentration of a GPa inhibitor at the MED. The [I]b-GPa/[I]t-liver ratio represents the fraction of a GPa inhibitor bound to GPa over the total GPa inhibitor concentration in the liver of an ob/ob mouse after oral administration of the GPa inhibitor. A high [I]b-GPa/[I]t-liver ratio of a GPa inhibitor suggests a high selectivity of the GPa inhibitor to GPa over all other liver proteins and low nonspecific binding. Therefore, to achieve the same amount of GPa bound to the inhibitor at the respective MED, the GPa inhibitor with a high [I]b-GPa/[I]t-liver ratio requires a lower liver concentration than one with a low [I]b-GPa/[I]t-liver ratio. The very poor [I]b-GPa/[I]t-liver ratio for compound 8 may be the reason, at least in part, for its poor in vivo activity.
In summary, the present study has demonstrated a useful in vitro model for determining KdGPa-liver that correlates well with the in vivo activities of a series of GPa inhibitors. This model may be particularly useful for studying compounds where binding to the target enzyme in vitro is dependent upon multiple endogenous cofactors present in the target tissue. A selectivity parameter determined by the [I]b-GPa/[I]t-liver ratio can also be very useful for understanding in vivo potency and disposition. Improving the selectivity of a GPa inhibitor to GPa over nonspecific protein binding in the liver will reduce the exposure required to show efficacy and may help to avoid an undesired effect. The simplicity and usefulness of this model, along with its adaptation to a 96-well format, make it a useful new tool applied to many drugs in pharmacology studies in which the correlation between in vitro and in vivo activities for the target tissue (such as solid tumors, brain and liver) is critical.
Acknowledgments
We gratefully acknowledge J. C. Kalvass for scientific discussion and information sharing and T. Checchio, W. Zavadoski, and P. Genereux for technical support. We are indebted to Drs. T. Maurer and D. Plowchalk for critical review of the manuscript and for valu-able comments. We also acknowledge Drs. S. Wright and R. Gammill for contribution on making some of the GPa inhibitors for the study.
Footnotes
-
This work was supported by Pfizer Inc.
-
doi:10.1124/jpet.105.100545.
-
ABBREVIATIONS: GPa, glycogen phosphorylase a; MED, the minimum efficacious dose; BES, N,N-bis(2-hydroxyethyl)2-amino-ethanesulfonic acid; LC/MS/MS, liquid chromatography with tandem mass spectrometry; HPLC, high-performance liquid chromatography; fuGPa-buffer, free fraction of an inhibitor in the GPa solution when dialyzed again the buffer; fuGPa-buffer, free fraction of an inhibitor in the GPa solution when dialyzed against the liver homogenate; fuliver, free fraction of an inhibitor in a diluted liver homogenate; uFuliver, extrapolated free fraction (see eq. 11 of an inhibitor in an undiluted liver homogenate); [GPa]b, GPa concentration bound to an inhibitor; [GPa]f, free GPa concentration; [GPa]t, total GPa concentration; [I]b-GPa, concentration of an inhibitor bound to GPa; [I]f, the free concentration of an inhibitor; [I]t-GPa, total concentration of an inhibitor in the GPa chamber of the equilibrium dialysis; [I]t-GPa-liver, total concentration of an inhibitor in the GPa chamber of the equilibrium dialysis against liver homogenate; [I]t-liver, total concentration of an inhibitor in the liver homogenate; KdGPa-liver, dissociation constant of a GPa inhibitor to GPa in presence of liver homogenate.
- Received December 22, 2005.
- Accepted March 13, 2006.
- The American Society for Pharmacology and Experimental Therapeutics