Mechanistic investigations into the species differences in pinometostat clearance: impact of binding to alpha-1-acid glycoprotein and permeability-limited hepatic uptake

Sherri A. Smith, Sandra Gagnon & Nigel J. Waters

To cite this article: Sherri A. Smith, Sandra Gagnon & Nigel J. Waters (2016): Mechanistic investigations into the species differences in pinometostat clearance: impact of binding to alpha-1-acid glycoprotein and permeability-limited hepatic uptake, Xenobiotica, DOI: 10.3109/00498254.2016.1173265
To link to this article:

Published online: 10 May 2016.

Submit your article to this journal

Article views: 70
View related articles View Crossmark data

Full Terms & Conditions of access and use can be found at

ISSN: 0049-8254 (print), 1366-5928 (electronic)

Xenobiotica, Early Online: 1–9
! 2016 Informa UK Limited, trading as Taylor & Francis Group. DOI: 10.3109/00498254.2016.1173265


Mechanistic investigations into the species differences in pinometostat clearance: impact of binding to alpha-1-acid glycoprotein and permeability-limited hepatic uptake

Sherri A. Smith1, Sandra Gagnon2, and Nigel J. Waters1

1Epizyme Inc, Cambridge, MA, USA and 2Charles River Laboratories, Montreal, QC, Canada

1. The plasma clearance of the first-in-class DOT1L inhibitor, EPZ-5676 (pinometostat), was shown to be markedly lower in human compared to the preclinical species, mouse, rat and dog.
2. This led to vertical allometry where various interspecies scaling methods were applied to the data, with fold-errors between 4 and 13. We had previously reported the elimination and metabolic pathways of EPZ-5676 were similar across species. Therefore, the aim of this work was to explore the mechanistic basis for the species difference in clearance for EPZ- 5676, focusing on other aspects of disposition.
3. The protein binding of EPZ-5676 in human plasma demonstrated a non-linear relationship suggesting saturable binding at physiologically relevant concentrations. Saturation of protein binding was not observed in plasma from preclinical species. Kinetic determinations using purified serum albumin and alpha-1-acid glycoprotein (AAG) confirmed that EPZ-5676 is a high affinity ligand for AAG with a dissociation constant (Kd) of 0.24 mM.
4. Permeability limited uptake was also considered since hepatocyte CLint was much lower in human relative to preclinical species. Passive unbound CLint for EPZ-5676 was estimated using a correlation analysis of logD and data previously reported on seven drugs in sandwich cultured human hepatocytes.
5. Incorporation of AAG binding and permeability limited hepatic uptake into the well-stirred liver model gave rise to a predicted clearance for EPZ-5676 within 2-fold of the observed value of 1.4 mL min—1 kg—1. This analysis suggests that the marked species difference in EPZ- 5676 clearance is driven by high affinity binding to human AAG as well as species-specific hepatic uptake invoking the role of transporters.

DOT1L inhibitor, EPZ-5676, fraction unbound, IVIVE, non-linear kinetics

Received 30 January 2016
Revised 28 March 2016
Accepted 29 March 2016
Published online 9 May 2016

Acute leukaemias bearing MLL rearrangements are aggres- sive diseases with current treatment options limited to chemotherapy and allogeneic hematopoietic stem cell trans- plantation; however, these have significant side effects and outcome remains poor. As a result, there is intense interest in developing novel therapeutic strategies for this disease. MLL rearrangements result in the loss of the carboxyterminal methyltransferase domain and an in-frame fusion of the amino-terminal region of MLL to 1 of more than 60 potential fusion partners, typically sequences derived from AF4, AF9, AF10 and ENL, which interact with DOT1L directly or indir- ectly in complexes that promote transcriptional elongation (Biswas et al., 2011). DOT1L is a histone methyltransferase

Address for correspondence: Sherri A. Smith, Drug Metabolism and Pharmacokinetics, Epizyme Inc, 400 Technology Square, 4th Floor, Cambridge, MA 02139, USA. Tel: 617 229 7557. Fax: 617 349 0707.
E-mail: [email protected]

(HMT) enzyme that targets lysine 79 in the globular domain of histone H3 (H3K79) for mono-, di-, or trimethylation (H3K79me1, me2, or me3). As a result, MLL fusion proteins gain the ability to recruit DOT1L to MLL target genes where the resulting hypermethylation at H3K79 leads to aberrant expression of a characteristic set of genes including HOXA9 and MEIS1 that drive leukaemogenesis (Bernt et al., 2011).
Hit and lead optimization efforts led to the discovery of the aminonucleoside analog, EPZ-5676 (pinometostat; (2R,3R,4S,5R)-2-(6-amino-9H-purin-9-yl)-5-((((1r,3S)-3-(2-
(5-(tert-butyl)-1H-benzo[d]imidazol-2-yl)ethyl)cyclobutyl) (isopropyl)amino)methyl)tetrahydrofuran-3,4-diol), a potent and selective inhibitor of DOT1L with a inhibition constant (Ki) of 80 pM and > 37,000-fold selectivity over a panel of other HMTs (Chesworth et al., 2013; Daigle et al., 2013). Administration of EPZ-5676 in rat xenograft models of MLL- rearranged leukemia via continuous intravenous (IV) infusion, caused complete tumor regressions that were sustained after cessation of dosing, with no significant weight loss or signs of

toxicity (Daigle et al., 2013). EPZ-5676 was the first member of the novel HMT inhibitor class to enter clinical development and is currently under investigation in Phase 1 studies of both adult and pediatric leukemia patients bearing the MLL rearrangement (Stein et al., 2014; Waters et al., 2014). Administered as a continuous IV infusion, EPZ-5676 has shown an acceptable safety profile up to 90 mg m—2 day—1 with clinical activity including complete responses and clonal differentiation in a heavily pretreated adult patient population (Stein et al., 2014). Pharmacokinetic (PK) data from this study revealed a markedly lower plasma clearance (CL) than observed in preclinical species. The median CL in man was determined as 1.4 mL min—1 kg—1 (Stein et al., 2014), roughly 7% hepatic blood flow. We recently reported on the preclin- ical PK and in vitro metabolism of EPZ-5676 in mouse, rat and dog from our early characterization of the compound, which showed high CL in preclinical species (Basavapathruni
et al., 2014), ranging from 61 to >100% hepatic blood flow. The aim of this work was to explore the mechanistic basis for this species difference. The metabolism of EPZ-5676 has previously been shown to be similar across species (Waters et al., 2015), thus we focused on other aspects of disposition
that could lead to the marked species difference in CL.

Materials and methods
EPZ-5676 was synthesized by Epizyme (Cambridge, MA). [14C]-EPZ-5676 was synthesized by Selcia (Essex, UK). All other reagents were purchased from sources as described below.

Prediction of human pharmacokinetics using preclinical species
Human PK prediction by interspecies scaling approaches was based on data for EPZ-5676 reported in Basavapathruni et al. (2014), with the exception of plasma protein binding which was generated as part of this study (see below). PK endpoints of CL in mouse, rat and dog were used to scale to the corresponding parameters in human using a variety of allometric and interspecies scaling approaches as previously reported (Boxenbaum, 1982; Jones et al., 2011; Lombardo

et al., 2013a,b; Mahmood, 2000; Mahmood, 2006; Mahmood, 2007; Mahmood & Balian, 1996; Ring et al., 2011; Tang & Mayersohn, 2006; Tang & Mayersohn, 2005, Tang et al., 2007; Wajima et al., 2002). Table 1 provides a summary of previously reported data used in part to generate the various clearance prediction values.

Drug–protein binding assays and blood: plasma partitioning
Prior to drug binding assays, the stability of [14C]-EPZ-5676 was established in Dulbecco’s phosphate buffered saline (DPBS) and each lot of plasma. Nonspecific binding to the equilibrium dialysis apparatus and the time to equilibration were determined to ensure the study design was appropriate. An incubation time of 4 h was used to perform the protein binding assays for all concentrations of [14C]-EPZ-5676 in plasma for all species. The protein binding assay was performed using healthy donor plasma from human, rat or dog (Bioreclamation, Westbury, NY), in duplicate, to achieve nominal final concentrations of 0.3, 1, 3 and 10 mM contain- ing 0.017–0.1 mCi mL—1 [14C]-EPZ-5676. In addition, the equilibrium dialysis was performed using DPBS containing human serum albumin (HSA) (40 mg mL—1; Sigma-Aldrich, St. Louis, MO) or human a-1-acid glycoprotein (AAG) (1.8 mg mL—1; Sigma-Aldrich, St. Louis, MO), in duplicate, to achieve the same nominal final concentrations of [14C]- EPZ-5676. Portions of the spiked plasma solutions (1 mL) were placed in one compartment of the equilibrium dialyzer cell (Spectrum 5-Cell Equilibrium Dialyzer, New Brunswick, NJ) with a membrane MWCO of 12–14 kDa and subjected to equilibrium dialysis against DPBS in the other compartment, in a warm room set to maintain 37 ◦C, with rotation. Following an incubation period of 4 h (previously shown to be sufficient to reach equilibrium), the contents of duplicate cell halves were removed and aliquots were taken to determine the radioactivity concentration, in each compart- ment, by LSC (Packard 2900TR liquid scintillation counter, Shelton, CT). The percentage of binding to plasma proteins of
1.45 mM [14C]-salicylic acid (0.2 mg mL—1) was assessed as a positive control, in duplicate over an incubation time of 5 h using pooled human plasma. The binding values determined for the positive control were compared with those published

Table 1. Summary of previously reported EPZ-5676 data.

Parametera Mouse Rat Dog Human
Clearance (mL min—1 kg—1)
Scaled CLH liver microsomes 78 45 20 17
Scaled CLH hepatocytes 26 7 21 53
CL in vivob 77 68 18.7 1.4
Microsomal binding
Fraction unbound, liver microsomes (fumic) 0.80 0.77 0.72 0.77
Permeability (Papp × 10—6 cm s—1)
Cell line A — B B — A Efflux ratio
MDCK – mock 0.09 0.30 3.3
MDCK – MDR1 transfected 50.06 0.22 3.8
aAs reported previously (Basavapathruni et al., 2014; Stein et al., 2014; Waters et al., 2014).
bIV dose: 5 mg kg — 1 mouse, 1 mg kg — 1 rat and dog.
cPermeability apical to basolateral (A — B) and basolateral to apical (B — A).

in the literature to demonstrate the conditions suitable during the assay. The pre- and post-dialysis samples were analyzed for radioactivity concentration by LSC. The stability of

permeability-limited hepatic uptake, as described previously (Paine et al., 2008; Webborn et al., 2007).

radiolabeled compound was assessed by HPLC with radio- metric detection. Radioactivity measurements were conducted by LSC where each sample was counted for 5 min or to a two-

CLint, app

¼ CL

int, met

CLint, pass
× CLint, met þ CLint, pass


sigma error of 0.1%, whichever occurred first. All counts were converted to absolute radioactivity (dpm) by automatic quench correction based on the shift of the spectrum for the external standard. The appropriate background dpm values were sub- tracted from all sample dpm values. The percentage of binding was calculated as follows: % bound = 100 (Cp Cu) Cp—1. The percentage recovery, in plasma and buffer, was calculated as follows: % Recovery = 100 (Cp + Cu) Ct—1; where Cp is concentration in donor compartment after dialysis (bound and unbound), Cu is concentration in receiver compartment (unbound) and Ct is initial concentration prior to dialysis. The blood:plasma partitioning assay was performed in dupli- cate using healthy donor blood from human, rat or dog (Charles River, Montreal, CA) at the same final [14C]-EPZ-5676 concentrations indicated above. Equilibrium and stability were established in blood as part of the optimization process. Samples were incubated for 2 h in a 37 ◦C shaking water bath (60 oscillations min—1). At time 0 and the end of incubation, spiked blood aliquots were decolored with hydrogen peroxide, treated with Soluene 350 for solubilization, and mixed with liquid scintillation fluid. Additional blood aliquots were centrifuged and the resultant plasma was mixed with liquid scintillation fluid. Samples were analyzed for radioactivity concentration by LSC as described above. Blood:plasma (Kwb P—1) was calculated as follows: Kwb P—1= Cct Cp —1, where Cct =dpm was detected in control blood and Cp =dpm was detected in plasma.

In vitro–in vivo extrapolation (IVIVE) approaches for clearance
IVIVE scaling was performed using the well-stirred venous equilibration model as previously reported (Houston, 1994; Obach, 1999; Pang & Rowland, 1977). The scaled intrinsic clearance values were scaled to predicted in vivo clearance values, using the well-stirred venous equilibration model as shown below:

The unbound passive intrinsic clearance (CLint) was derived
from a correlation analysis as detailed below, and together with the metabolic CLint (liver microsomal CLint) was transformed to the apparent CLint.
Using data generated previously by Jones et al. (2012), a compelling correlation between passive CLint and logD7.4 was established with an R2 of 0.89. This was prospectively applied to the investigation of EPZ-5676 and IVIVE of human hepatic CL with permeability limited uptake. The logD7.4 for EPZ- 5676 was 1.8 and derived an unbound passive CLint of 28 mL min—1 million—1 cells.
Interspecies PK scaling
A summary of human CL predictions by various interspecies scaling methods are shown in Table 2. Exemplar allometric plots are illustrated in Figure 1. Using three species, simple allometry gave rise to an exponent outside the generally accepted criterion for simple allometry of 0.555b50.7, and in this case invoked the use of the maximum life potential (MLP) CL product term, as part of the rule of exponents approach (Mahmood & Balian, 1996). Simple allometry with or without factoring unbound CL terms gave rise to human CL values in the high range of 10 mL min—1 kg—1. Consistent predictions for human CL were obtained using multiple linear regression (MLR), rule of exponents and fu corrected intercept method (FCIM) approaches, all in the 3–7 mL min—1 kg—1 range. All interspecies scaling methods gave rise to what has been termed vertical allometry, and an over-prediction in CL with fold errors ranging from 3.9 to 13. The FCIM method was the most accurate with a fold error of 1.8.

In vitro plasma protein binding and blood: plasma partitioning
The plasma protein binding data across species and the

QH · CLint

QH · fub · CLint

relationship between fu and EPZ-5676 concentration is


þ CLint


þ fub

· CLint


illustrated in Figure 2 and Table 3. At clinically relevant drug
concentrations (≤ 3 mM EPZ-5676), the fu in rat is ~3–5-fold

where QH is blood flow, fub is fraction unbound in blood and CLint is scaled intrinsic clearance. The well-stirred model was applied with and without inclusion of the free fraction (fub) in blood. Blood:plasma ratios used in the scaling methods were 1.28, 0.73 and 0.53 in rat, dog and human, respectively. The previously reported blood partitioning and plasma free fraction in mouse were used in this analysis (Basavapathruni et al., 2014). The fub values used in the predicted clearance calculations were 0.06, 0.14, 0.05 and 0.08 in mouse, rat, dog and human, respectively, derived by correcting the fu in plasma by blood:plasma ratio. The CLint value used in the model was either the metabolic CLint derived from liver microsomes or the apparent CLint as derived below which included a distributional clearance term for

higher than in human, which could explain the superior predictive accuracy of the FCIM method. The other key finding shown in Figure 2(A) is the concentration dependence observed for human plasma protein binding that is not apparent in rat and less apparent in dog. This is further supported in Figure 2(B), the corresponding Scatchard analysis, which shows a clear correlation in human plasma indicative of saturable binding. The Kd estimated from the human plasma data was 0.52 mM, suggestive of a high affinity binding site for EPZ-5676 in human plasma.
The blood partitioning of EPZ-5676 did not show a marked concentration dependent relationship in rat, dog or human at clinically relevant drug concentrations ( 3 mM EPZ-5676). However, at 10 mM EPZ-5676, the blood:plasma ratio

Table 2. Human clearance projections for EPZ-5676 using various interspecies scaling approaches.

Scaling method, equationa and number of speciesb

Predicted human CL (mL min—1 kg—1)

errorc Exponentd Reference


2 Simple allometry CL = a*Wb
Simple allometry, CL unbound 3f

3f 13.3

18.2 9.5

13 0.76 Mahmood 2007 Tang and Mayersohn 2006 Tang
et al. 2007
Mahmood 2000

3 CLu = a*Wb
Rule of exponents, maximum life potentiale
Tang et al. 2007

4 fu corrected intercept method (FCIM)
CL = 33.35*(a Rfu—1)0.77 3 2.5 1.8 Tang and Mayersohn 2005
Mahmood 2006

5 Multiple linear regression (MLR) 2 6.9 4.9 Lombardo et al. 2013a

logCLhuman = 0.4*logCLrat + 0.4*logCLdog – 0.4
aEquation Key: a = coefficient of allometric equation, b = exponent of allometric equation, W = body weight, fu = fraction unbound in plasma, CLu = unbound clearance (CL fup — 1), Rfu = ratio of rat to human fu.
bWhen number of species is 2, rat and dog factored; when number of species is 3, rat, dog and mouse factored.
cFold error from measured human CL (1.4 mL min — 1 kg — 1). dExponent for simple allometry acceptable range 0.55–7.
eRule of exponents, MLP equation used when exponent 0.715b50.99.
fRepresented in Figure 1.

(A) 1000


100 H-obs



(B) 100000

0.01 0.1 1 10 100
BW (kg)
(C) 100000








100 R

100 R

10 10 M

1 1

0.01 0.1 1 10 100
BW (kg)

0.01 0.1 1 10 100
BW (kg)

Figure 1. Allometric plots for the scaling of EPZ-5676 clearance in human plasma using body weight (BW) and clearance values measured in mouse (M), rat (R) and dog (D), (A) based on simple allometry; (B) MLP correction; (C) and simple allometry of unbound clearance. H-pred and H-obs refer to predicted and observed clearance in human respectively.

increased 20% in dog and human. The fub values used in the predicted clearance calculations were 0.06, 0.14, 0.05 and
0.08 in mouse, rat, dog and human, respectively. Data generated at 1 mM EPZ-5676 were used in the scaling methods, except for mouse where 5 mM data was used due to lack of a complete data set.
The concentration dependence and binding affinity of EPZ-5676 was explored using physiologically relevant

concentrations of HSA and AAG, prepared separately in aqueous buffer. The saturation binding curves are shown in Figure 3(A) (HSA) and Figure 3(B) (AAG). For HSA, a nonlinear relationship was less clear and suggested a low affinity binding site with a Kd of 33 mM. The value of maximal binding (Bmax) for HSA was undetermined. For AAG, there was a nonlinear relationship between free and bound EPZ-5676 concentration in the nominal 0.3–10 mM

0.25 30






0 2 4 6 8 10
total EPZ-5676 concentration (M)

0 2 4 6 8 10
bound EPZ-5676 concentration (M)

Figure 2. (A) Relationship between EPZ-5676 concentration and free fraction (fu) in rat ( ), dog (*) and human (m) whole plasma; (B) Scatchard plots for binding of EPZ-5676 in rat, dog and human plasma.

Table 3. Summary of EPZ-5676 protein binding data.

Parameter [14C]EPZ-5676 (mM) Mousea Rat Dog Human
Fu plasma 0.3 0.19 0.04 0.04
1 0.18 0.04 0.04
3 0.14 0.20 0.03 0.06
10 0.20 0.09 0.12
Fu blood 0.3 0.13 0.06 0.08
1 0.14 0.05 0.08
3 0.06 0.15 0.04 0.11
10 0.15 0.09 0.19
Blood:Plasma 0.3 1.41 0.70 0.51
1 1.28 0.73 0.53
3 2.15 1.30 0.81 0.54
10 1.35 1.0 0.63
Kd mM Plasma 0.52
AAG 0.24
Albumin 33
aMouse data previously reported (Basavapathruni et al., 2014), 5 mM EPZ-5676.

range. A Kd value of 0.24 ± 0.02 mM (mean ± SE) suggested a single high affinity binding site for EPZ-5676 on human AAG with a Bmax of 11.8 ± 0.35 mM.

IVIVE approaches
A number of IVIVE approaches were explored across species in order to establish the methods that were most predictive and therefore provide some mechanistic insights (Table 4). Scaling CL using the well-stirred liver model (perfusion limited, without incorporation of plasma free fraction) led to high CL estimates in all species. These estimates were very closely aligned with in vivo CL values in preclinical species (within 35%), but showed a marked over-prediction in human, with a fold error of 12, analogous to interspecies scaling approaches. Further refinement with inclusion of fu in blood improved the scaling accuracy in human to a 3.4-fold error, however, this resulted in marked under-prediction in preclin- ical species (3- to 6-fold error).
The in vitro permeability of EPZ-5676 was previously determined in MDCK and Caco-2 cell lines with mean

apparent permeability (Papp) of 50.09 10—6 cm s—1 (Basavapathruni et al., 2014) (Table 1). In addition, there was a large disconnect in the scaled CLint between human liver microsomes and human hepatocytes, indicative of potential permeability-limited hepatic uptake. The physico- chemical properties of EPZ-5676 also suggested low perme- ability, which may be important in the hepatic CL process. Previous reports have shown the importance of permeability in the hepatic CL of a number of drugs, and in doing so generated in vitro estimates of passive CLint. Using data generated previously by Jones et al. (2012), a compelling correlation between passive CLint and logD7.4 was established with an R2 of 0.89 (Figure 4). This was prospectively applied to the investigation of EPZ-5676 and IVIVE of human hepatic CL with permeability limited uptake. The logD7.4 for EPZ- 5676 was 1.8 and derived an unbound passive CLint of 28 mL min—1 million—1 cells. Using the well-stirred model with permeability limited uptake, gave rise to a predicted human hepatic CL of 13 mL min—1 kg—1 (9-fold error). The combined inclusion of fu in blood and permeability limited uptake, further improved the predictive accuracy with a CL value of

6 10




0 1 2 3 4 5
free EPZ-5676 concentration (M)

0.0 0.2 0.4 0.6 0.8 1.0
free EPZ-5676 concentration (M)

Figure 3. Saturation binding curves for the binding of EPZ-5676 to (A) human serum albumin at 40 mg mL—1 and (B) a-1-acid glycoprotein at 1.8 mg mL—1. Nonlinear regression shown in main plot with scatchard plot shown as inset. For AAG, Kd was determined as 0.24 mM with a Bmax of 11.8 mM. For HSA, Kd was defined as 33 mM.

Table 4. Summary of cross species IVIVE predictions of clearance for EPZ-5676.

CLint mL min—1 kg—1
Scaled intrinsic clearance Mouse Rat Dog Human
Metabolica 617 135 72 72
Passiveb 314 143 168 71
Apparentc 208 70 51 36
Scaled clearance CL mL min—1 kg—1 (fold change from in vivo)
Mouse Rat Dog Human
Perfusion limited without fubd 78 (1.0) 45 (0.66) 20 (1.1) 17 (12)
with fubd 28 (0.36) 15 (0.22) 3.4 (0.18) 4.7 (3.4)
Permeability-limited hepatic uptake without fube 19 (0.25) 16 (0.24) 15 (0.79) 13 (9.3)
with fube 12 (0.16) 8.7 (0.13) 2.5 (0.13) 2.6 (1.9)
Clearance, in vivof 77 68 19 1.4
Clearance, in vivo as % HBFg 86 123 61 6.7
aDerived from liver microsome data (Basavapathruni et al., 2014).
bDerived from correlation analysis (Figure 4).
cCalculated using Equation (2).
dCalculated using metabolic CLint.
eCalculated using apparent CLint, fub = fraction unbound in plasma corrected for blood to plasma partitioning.
fAs reported previously (Basavapathruni et al., 2014; Stein et al., 2014; Waters et al., 2014).
gHepatic blood flow (HBF) derived from physiological data (Davies & Morris, 1993).

2.6 mL min—1 kg—1, and a 52-fold error from the observed CL in man. Further refinements with inclusion of fu and permeability limited uptake led to lower CL estimates in the preclinical species, and abrogated the predictive accuracy that was achieved with the simple perfusion-limited well-stirred model. No unifying, singular IVIVE model was able to recapitulate the observed CL across species.

EPZ-5676 (pinometostat) is a potent, selective DOT1L inhibitor that is currently under clinical investigation for MLL-r leukaemias in adult and pediatric populations. During early development, the observed CL in human was shown to be markedly lower than that determined in preclinical species. The predictive accuracy of interspecies PK scaling has received much attention recently with several large initiatives to identify the most predictive methods in estimating human

PK endpoints from preclinical species (Jones et al., 2011; Lombardo et al., 2013a,b; Ring et al., 2011). As shown in Table 2, we analyzed the EPZ-5676 PK data in several ways using a multitude of approaches recommended in these recent analyzes, giving careful consideration to cases of congruence or disparity in prediction. With the exception of the FCIM method, all interspecies scaling and allometric methods over- predicted human CL of EPZ-5676 with fold errors ranging from 3.9 to 13. This phenomenon has been coined ‘vertical allometry’ by some investigators and was observed with compounds such as UCN-01, diazepam, tamsulosin, valpro- ate, warfarin, susalimod and antipyrine (Mahmood & Boxenbaum, 2014). These authors highlighted the difficulty in identifying vertical allometry with any certainty and that no drug properties have been demonstrated to date that univer- sally characterize this behavior. In the case of EPZ-5676, the 3–5-fold difference in fu between rat and human likely provides the basis for the improved prediction using the FCIM





1.0 0.5 0.0 0.5 1.0 1.5 2.0 2.5

Figure 4. Correlation analysis of the unbound passive diffusion clear- ance (CLint, u, passive) in sandwich cultured human hepatocytes with logD7.4, based on data from pravastatin, cerivastatin, bosentan, fluvastatin, rosuvastatin, valsartan and repaglinide (Jones et al, 2012). Linear regression is described by y = 0.7396x + 0.1179 with an R2 of
0.89. EPZ-5676 is represented by the solid circle.

(molecular weight of ca. 67 kDa, circulating concentration in plasma of 35–50 mg mL—1 or 522–746 mM) (Fournier et al., 2000; Putnam, 1984). AAG has a very low isoelectric point (pI 2.8–3.8), and therefore binds mostly to basic drugs. The concentration of AAG increases in acute states in response to systemic tissue injury, inflammation or infection, as well as disease states, such as cancer (Wright et al., 1996). Drug binding to AAG can be much more variable, compared to albumin, within and across species, and is susceptible to nonlinear behavior (i.e., saturation) because of the relatively low abundance, variable protein levels in disease states, and genetic polymorphism (Fournier et al., 2000). The Kd for EPZ-5676 binding to human AAG was measured as 0.24 mM indicating a high affinity interaction. By comparison, proto- typical AAG ligands such as UCN-01, dipyridamol, disopyr- amide and thioridazine have Kd values of 803, 15.5, 1.0 and 63 mM, respectively. Vismodegib is a recent, well character- ized example in terms of understanding the impact of AAG on its PK. Vismodegib was reported to have human Kd values of

method. This is in agreement with recent global analyses of human CL prediction where different groups have highlighted the predictive accuracy of the FCIM method (Lombardo et al., 2013a; Ring et al., 2011). However, a disproportionate free fraction between rat and human is not a singular parameter that ensures predictive accuracy with the FCIM approach. The prospective application of allometry and interspecies scaling in drug development still requires a judicious review of the data and assumptions of each method used in clearance projection.
Our previous work had indicated the stark species difference in CL that was not related to metabolic pathways or routes of elimination (Waters et al., 2015). In rat and dog, fecal excretion was the major route of elimination, represent- ing 80% of the total dose, and in all species including human, renal excretion of both EPZ-5676 and EPZ-5676- related material was low. EPZ-5676 underwent extensive oxidative metabolism with the major metabolic pathways being hydroxylation of the t-butyl group and N-dealkylation of the central nitrogen. Unchanged EPZ-5676 was the predominant circulating species in rat, dog and man (Waters et al., 2015), suggesting efficient hepatobiliary elimination of EPZ-5676 metabolites and low systemic burden to metabol- ites. The aim of this work was to investigate other disposition factors that could explain the striking difference in CL, and thereby provide insight into additional considerations for a priori human PK prediction in the future. Although there were some species differences in the extent of plasma protein binding e.g. between rat and human, this was not a singular explanation for the species difference. More striking was the concentration dependence in protein binding observed in human plasma, over a relevant EPZ-5676 concentration range, which was less apparent in the preclinical species. This, along with the in vitro kinetic determinations, suggested the saturable binding of EPZ-5676 to AAG. AAG is an acute phase protein with a molecular weight of ca. 42 kDa and is present at a much lower circulating concentration in plasma (0.55–1.8 mg mL—1 or ~8–30 mM) relative to HSA

13 and 120 mM for AAG and albumin, respectively (Gianetti
et al., 2011). In rat, the AAG Kd value could not be determined due to weak binding, whereas the Kd for albumin (140 mM) was similar to the human Kd. Human CL of vismodegib exhibited non-linear PK and was substantially over-predicted by preclinical models, in large part attributed to the low human AAG Kd (Graham et al., 2011,2012).
The lack of commercially available isolated AAG protein from rat and dog precluded Kd estimates of EPZ-5676 in these species. Notwithstanding, there is a disproportionately higher expression of AAG in human plasma (0.55–1.8 mg mL—1) relative to mouse (0.1 mg mL—1, Azuma et al. (2007)), rat (0.1–0.3 mg mL—1, Chauvelot-Moachon et al., (1988); Stolina et al., (2009)) and dog (0.3–1 mg mL—1, Belpaire et al., (1987); Dello et al., (1987); Rikihisa et al., (1994)), which is likely a contributing factor alone, irrespective of potential species-specific differences in AAG affinity. It should also be stated that the accurate determination of fu for compounds with affinity for AAG can be markedly affected by the blood collection method (Butler et al., 2015). This appears to be caused by a direct effect of certain plasticizers on the binding of drugs to AAG (Bowen & Remaley, 2014; Devine, 1984; Sager & Little, 1989; Shah et al., 1982,) and as a result the fu for EPZ-5676 in human plasma may be overestimated.
IVIVE-based predictions for EPZ-5676 human CL were much improved by the inclusion of fu, yet the prediction error remained relatively high at 3–4-fold. It is a common observation for liver microsomes to over-predict CL, espe- cially for compounds with low passive membrane permeabil- ity. This was largely exemplified with the previously reported scaled hepatocyte CL values for EPZ-5676 in which low turnover was observed for mouse, rat and human (CLint54 mL min—1 million—1 cells) (Basavapathruni et al, 2014). Further refinement of the IVIVE model utilizing a distributional clearance term was explored (Paine et al., 2008; Webborn et al., 2007). A correlation analysis was performed in order to derive a passive CLint for EPZ-5676. Jones and coworkers have previously presented in vitro data on the hepatic uptake of seven drugs with logD ranging from 0.88 (valsartan) to
2.1 (repaglinide) in sandwich cultured human hepatocytes. On
review, we noted a compelling correlation between logD and

unbound passive CLint that was deemed suitable for predicting passive CLint for new compounds particularly in cases of interpolation within the training set data. This approach was used for EPZ-5676 and led to a projected unbound passive CLint of 28 mL min—1 million—1 cells. Incorporation of permeability limited uptake as well as free fraction correction into the IVIVE approach was able to recapitulate the observed human CL within a 2-fold window. This finding was consistent with our previous PBPK modeling of the human PK data for EPZ-5676 where inclusion of permeability limited hepatic uptake led to a superior fit of the concentra- tion–time profile as well as the PK parameters, Css and CL (Rioux and Waters, 2016; Waters et al., 2014).
Despite the improved IVIVE human CL projection by
factoring in permeability and AAG binding considerations, there was no universal IVIVE method that was able to accurately predict CL across species. This suggests that there may be a sinusoidal active uptake process in preclinical species that is absent or less active in man. Alternatively, there may be a sinusoidal active efflux mechanism in man that is absent or less active in preclinical species. Transporter phenotyping studies have indicated that BCRP, OCT1/2, OATP1B1/3, CNT1/2/3 and ENT1/2 are not involved in the transport of EPZ-5676 (data not shown). Furthermore, MDR1 and MATE1 are likely involved in canalicular efflux (internal data). Radiolabeled ADME and QWBA studies in rat demonstrate high levels of radioactivity uptake into liver with a large proportion of dose-related material excreted into bile. The molecular basis for the apparent species difference in hepatic uptake remains to be elucidated, but could invoke EPZ-5676 as a more sensitive substrate for the rat isoforms of sinusoidal transport proteins such as OCTs, CNTs or ENTs for example. Species-specific active transport and hepatic dis- position has recently been implicated in the differential mitochondrial toxicity of the nucleosidic anti-viral com- pounds, including fialuridine, with hENT1 attributed as the molecular determinant. Another, as yet unexplored, hypoth- esis would be the role of human MRPs in the sinusoidal efflux of EPZ-5676.
In summary, the mechanistic basis for the over-prediction of EPZ-5676 CL in human was demonstrated to be caused by species-specific differences in binding to AAG and perme- ability limited uptake. The predictive accuracy of IVIVE can be markedly improved by including these factors into a priori human CL projections, and highlights the importance of understanding permeability and the identity of the plasma proteins involved in drug binding. In addition, the correlation analysis between lipophilicity and passive unbound CLint provides a useful means to estimate the role of permeability in hepatic CL during drug discovery. The molecular basis for the species differences in EPZ-5676 hepatic uptake remain to be elucidated and future efforts will be enabled with the increased availability of transfected cell lines expressing transporter isoforms for the preclinical species.

Declaration of interest
The authors S.A.S. and N.J.W. are/were employees of and/ or hold equity in Epizyme, Inc. All efforts were funded by Epizyme, Inc.

Azuma M, Nishioka Y, Aono Y, et al. (2007). Role of a1-acid glycoprotein in therapeutic antifibrotic effects of imatinib with macrolides in mice. Amer J Resp Crit Care Med 176:1243–50.
Basavapathruni A, Olhava EJ, Daigle SR, et al. (2014). Nonclinical pharmacokinetics and metabolism of EPZ-5676, a novel DOT1L histone methyltransferase inhibitor. Biopharm Drug Dispos 35: 237–52.
Belpaire FM, DeRick A, Dello C, et al. (1987). a1-acid glycoprotein and serum binding in healthy and diseased dogs. J Vet Pharmacol Ther 10: 43–8.
Bernt KM, Zhu N, Sinha AU, et al. (2011). MLL-rearranged leukemia is dependent on aberrant H3K79 methylation by DOT1L. Cancer Cell 20:66–78.
Biswas D, Milne TA, Basrur V, et al. (2011). Function of leukemogenic mixed lineage leukemia 1 (MLL) fusion proteins through distinct partner protein complexes. Proc Natl Acad Sci Unit States Am 108: 15751–6.
Bowen RAR, Remaley AT. (2014). Interferences from blood collection tube components on clinical chemistry assays. Biochem Med 24: 31–44.
Boxenbaum H. (1982). Interspecies scaling, allometry, physiological time, and the ground plan of pharmacokinetics. J Pharmaco and Biopharm 10:201–27.
Butler P, Frost K, Barnes K, et al. (2015). Impact of blood collection method on human plasma protein binding for compounds preferen- tially binding to a1-acid glycoprotein. Poster presented at the ISSX 20th North American meeting; 2015 Oct 18–22; Orlando, FL, USA. Chauvelot-Moachon L, Delers F, Pous F, et al. (1988). Alpha-1-acid glycoprotein concentrations and protein binding of propranolol in Sprague-Dawley and Dark Agouti rat strains treated by phenobarbital.
J Pharmacol Exp Ther 244:1103–8.
Chesworth R, Olhava EJ, Kuntz KW, et al. (2013). From protein to candidate: discovery of EPZ-5676, a potent and selective inhibitor of the histone methyltransferase DOT1L. Abstracts of Papers, 246th ACS National Meeting & Exposition; 2013 Sept 8–12; Indianapolis, IN, United States.
Daigle SR, Olhava EJ, Therkelsen CA, et al. (2013). Potent inhibition of DOT1L as treatment for MLL-fusion leukemia. Blood 122:1017–25.
Davies B, Morris T. (1993). Physiological parameters in laboratory animals and humans. Pharm Res 10:1093–5.
Dello CP, Belpaire FM, Kint JA, Freyman NH. (1987). Dog alpha-1-acid glycoprotein: purification and biochemical characterization. J Pharmacol Met 18:335–45.
Devine JE. (1984). Drug-protein binding interferences caused by the plasticizer TBEP. Clin Biochem 17:345–7.
Fournier T, Medjoubi NN, Porquet D. (2000). Alpha-1-acid glycopro- tein. Biochim Biophys Acta 1482:157–71.
Gianetti AM, Wong H, Dijkgraaf GJP, et al. (2011). Identification, characterization, and implications of species-dependent plasma pro- tein binding of the oral Hedgehog pathway inhibitor vismodegib (GDC-0449). J Med Chem 54:2592–601.
Graham RA, Lum B, Cheeti S, et al. (2011). Pharmacokinetics of Hedgehog pathway inhibitor vismodegib (GDC-0449) in patients with locally advanced or metastatic solid tumors: the role of alpha-1-acid glycoprotein binding. Clin Cancer Res 17:2512–20.
Graham RA, Hop CE, Borin MT, et al. (2012). Single and multiple dose intravenous and oral pharmacokinetics of the hedgehog pathway inhibitor vismodegib in healthy female subjects. Br J Clin Pharmacol 74:788–96.
Houston JB. (1994). Utility of in vitro drug metabolism data in predicting in vivo metabolic clearance. Biochem Pharmacol 47: 1469–79.
Jones HM, Barton HA, Lai Y, et al. (2012). Mechanistic pharmacoki- netic modeling for the prediction of transporter-mediated disposition in humans from sandwich culture human hepatocyte data. Drug Metab Dispos 40:1007–17.
Jones RD, Jones HM, Rowland M, et al. (2011). PhRMA CPCDC initiative on predictive models of human pharmacokinetics, Part 2: comparative assessment of prediction methods of human volume of distribution. J Pharm Sci 100:4074–89.
Lombardo F, Waters Nigel J, Argikar Upendra A, et al. (2013a). Comprehensive assessment of human pharmacokinetic prediction

based on in vivo animal pharmacokinetic data, part 2: clearance. J Clin Pharmacol 53:178–91.
Lombardo F, Waters Nigel J, Argikar Upendra A, et al. (2013b). Comprehensive assessment of human pharmacokinetic prediction based on in vivo animal pharmacokinetic data, part 1: volume of distribution at steady state. J Clin Pharmacol 53:167–77.
Mahmood I. (2007). Application of allometric principles for the prediction of pharmacokinetics in human and veterinary drug development. Adv Drug Deliv Rev 59:1177–92.
Mahmood I. (2006). Prediction of human drug clearance from animal data: application of the rule of exponents and ‘fu corrected intercept method’ (FCIM). J Pharm Sci 95:1810–21.
Mahmood I. (2000). Interspecies scaling: role of protein binding in the prediction of clearance from animals to humans. J Clin Pharmacol 40: 1439–46.
Mahmood I, Balian JD. (1996). Interspecies scaling: predicting clearance of drugs in humans. Three different approaches. Xenobiotica 26: 887–95.
Mahmood I, Boxenbaum H. (2014). Vertical allometry: fact or fiction? Reg Toxicol Pharmacol 68:468–74.
Obach RS. (1999). Prediction of human clearance of twenty-nine drugs from hepatic microsomal intrinsic clearance data: an examination of in vitro half-life approach and nonspecific binding to microsomes. Drug Metab Disp 27:1350–9.
Paine SW, Parker AJ, Gardiner P, et al. (2008). Prediction of the pharmacokinetics of atorvastatin, cerivastatin, and indomethacin using kinetic models applied to isolated rat hepatocytes. Drug Metab Dispos 36:1365–74.
Pang KS, Rowland M. (1977). Hepatic clearance of drugs. I. Theoretical considerations of a ‘‘well-stirred’’ model and a ‘‘parallel tube’’ model. Influence of hepatic blood flow, plasma and blood cell binding, and the hepatocellular enzymatic activity on hepatic drug clearance. J Pharmaco Biopharm 5:625–53.
Putnam FW. (1984). Progress in plasma proteins. In: The plasma proteins, Vol. IV, structure, function and genetic control. Orlando, FL: Academic Press, 1–44.
Rikihisa Y, Yamamoto S, Kwak I, et al. (1994). C-reactive protein and a-1-acid glycoprotein levels in dogs infected with Ehrlichia canis. J Clin Microbio 32:912–7.
Ring BJ, Chien JY, Adkison KK, et al. (2011). PhRMA CPCDC initiative on predictive models of human pharmacokinetics, part 3: comparative assessment of prediction methods of human clearance. J Pharm Sci 100:4090–110.

Rioux N, Waters NJ. (2016). Physiologically-based pharmacokinetic modeling in pediatric oncology drug development. Drug Metab Dispos. DOI: 10.1124/dmd.115.068031.
Sager G, Little C. (1989). The effect of the plasticizers TBEP (tris-(2- butoxyethyl)-phosphate) and DEHP (di-(2-ethylhexyl)-phthalate) on beta-adrenergic ligand binding to alpha 1-acid glycoprotein and mononuclear leukocytes. Biochem Pharmacol 38:2551–7.
Shah VP, Knapp G, Skelly JP, Cabana BE. (1982). Interference with measurements of certain drugs in plasma by a plasticizer in vacutainer tubes. Clin Chem 28:2327–8.
Stein E, Garcia-Manero G, Rizzieri D, et al. (2014). The DOT1L inhibitor EPZ-5676: Safety and activity in relapsed/refractory patients with MLL-rearranged leukemia. Blood 124:387.
Stolina M, Schett G, Dwyer D, et al. (2009). RANKL inhibition by osteoprotegerin prevents bone loss without affecting local or systemic inflammation parameters in two rat arthritis models: comparsion with anti-TNFa or anti-IL-1 therapies. Arthritis Res Ther 11:R187.
Tang H, Hussain A, Leal M, et al. (2007). Interspecies prediction of human drug clearance based on scaling data from one or two animal species. Drug Metab Dispos 35:1886–93.
Tang H, Mayersohn M. (2006). A global examination of allometric scaling for predicting human drug clearance and the prediction of large vertical allometry. J Pharm Sci 95:1783–99.
Tang H, Mayersohn M. (2005). A novel model for prediction of human drug clearance by allometric scaling. Drug Metab Disp 33: 1297–303.
Wajima T, Fukumura K, Yano Y, Oguma T. (2002). Prediction of human clearance from animal data and molecular structural parameters using multivariate regression analysis. J Pharm Sci 91:2489–99.
Waters NJ, Thomson B, Gardner I, et al. (2014). Pediatric dose determinations for the phase I study of the DOT1L inhibitor, EPZ- 5676, in MLL-r acute leukemia: leveraging early clinical data in adults through physiologically-based pharmacokinetic modeling. Blood 124:3619.
Waters NJ, Smith SA, Olhava EJ, et al. (2016). Metabolism and disposition of the DOT1L inhibitor, pinometostat (EPZ-5676), in rat, dog and human. Cancer Chemother Pharmacol 77:43–62.
Webborn PJ, Parker AJ, Denton RL, Riley RJ. (2007). In vitro-in vivo extrapolation of hepatic clearance involving active uptake: theoretical and experimental aspects. Xenobiotica 37:1090–109.
Wright JD, Boudinot FD, Ujhelyi MR. (1996). Measurement and analysis of unbound drug concentrations. Clin. Pharmacokinetics 30: 445–62.

Leave a Reply

Your email address will not be published. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>