Discovery and Preclinical Profiling of GSK3839919, a Potent HIV-1 Allosteric Integrase Inhibitor

Abstract

Allosteric HIV-1 integrase inhibitors (ALLINIs) have been of interest recently because of their novel mechanism of action. Strategic modifications to the C5 moiety of a class of 4-(4,4-dimethylpiperidinyl)-2,6-dimethylpyridinyl ALLINIs led to the identification of a tetrahydroisoquinoline heterocycle as a suitable spacer element to project the distal hydrophobic aryl ring. Subsequent optimization of the aryl substitutions identified 12 as an ALLINI with single-digit nanomolar inhibitory potency and low clearance across preclinical species. In preclinical toxicology studies with 12 in rats, lipid hepatocellular vacuolation was observed. Removal of the C6 methyl group resulted in GSK3839919 (22), which exhibited a reduced incidence and severity of lipid vacuolation in both in vitro assays and in vivo studies while maintaining the potency and pharmacokinetic (PK) properties of the prototype. The virology, PK, and toxicology profiles of 22 are discussed.

Allosteric HIV-1 integrase inhibitors (ALLINIs) have recently been a focus of interest because of their promising and unique mechanism of action.¹ Multiple functions that are important in the viral lifecycle are disrupted by ALLINIs. These include disruption of proper viral core condensation, which results in immature noninfectious particles; blocking of lens epithelium-derived growth factor (LEDGF) binding to integrase (IN), which alters nuclear localization of the integrase complex; and inhibition of transcription when immature virions enter the cells in the next cell cycle.² A number of ALLINIs have progressed through preclinical studies, and Figure 1 depicts examples of known potent ALLINIs that have been evaluated in toxicology studies.

Figure 1.

Examples of ALLINIs evaluated in toxicology studies.

ALLINIs as a class have demonstrated a mixed record in safety evaluation. To date no ALLINIs have been approved for clinical use, but two members of this mechanistic class have advanced into human clinical trials. BI224436 (1) was reported to be safe and well-tolerated in a human phase I once-daily (qd) oral (po) single ascending dose study in which up to 200 mg of drug was admininistered,³ while STP0404 (2) was reported to be safe and well-tolerated in both rat and dog toxicology studies and has advanced into human phase I studies.⁴ Conversely, several ALLINIs have caused adverse findings in preclinical toxicology studies, halting further progression of these molecules. GS-9822 (3) was reported to cause dose-dependent vacuolation of the urothelium of kidney, bladder, and ureter in cynomolgus monkey toxicology studies.⁵ Similarly, GSK3739936 (4) was found to cause lipid vacuolation in the liver and kidneys of rats at a dose of 500 mg kg^–1 day^–1.^1f These adverse findings necessitated the development of an in vitro lipid vacuolation assay that could triage compounds for assessment in in vivo safety studies. Herein is described the exploration of structurally diverse analogues that were pursued in parallel with the development of the in vitro lipidosis assay in an effort to identify an ALLINI with a profile sufficient to support progression into clinical studies.

One challenge in the design of potent ALLINIs is discovering an approach that can tolerate the genetic variation of amino acid residues that form the binding pocket. Many polymorphic combinations result from the intrinsic genetic variation within the ALLINI binding pocket, especially at residues 124 and 125 of the integrase protein. The frequencies of the six most prevalent B-subtype virus integrase polymorphisms at these positions in order of prevalence are T/T (36.2%), A/T (14.9%), T/A (11.9%), N/T (10.6%), N/A (7.9%), and A/A (7.3%).⁶ On the basis of an evaluation of previous ALLINIs against the large panel of polymorphs,^1f an NL_4–3 virus containing the 124N/125A (N/A) polymorphism was selected as a representative sentinel polymorph (as it was generally the least sensitive of the six polymorphs to these inhibitors) and was used along with the 124T/125T (T/T) polymorphism (present in the wild-type (WT) NL_4–3 screening virus) in the initial screening assays to assess the potency of new HIV-1 ALLINIs.

Expanding on the work described elsewhere that led to the discovery of 4,^1f research efforts began with structural modifications to the p-fluorophenethyl C5 moiety. In this class of ALLINIs, the C5 substituent serves as a spacer that projects a hydrophobic group in an orientation that allows it to engage with the α3-helix residues 124–133 comprising the hydrophobic shelf.^1f The first approach explored was to restrict the rotation of the distal p-fluorophenyl group of 4 and thus intentionally control the orientation of the distal hydrophobic group (Figure 2).

Figure 2.

Initial C5 modification strategy. Selected HIV-1 integrase residue interactions for compound 4 are noted.^1f

The substituted chromane that resulted from this design concept was prepared, and both stereoisomers were tested in the primary cell-culture antiviral assay. The S and R stereoisomers (5 and 6, respectively) demonstrated similar potency profiles, with the potency against T/T being ∼5-fold greater than that against N/A (Table 1). Compared with 4, both 5 and 6 exhibited inferior potency against the N/A polymorph, which prompted further C5 modification work.

Table 1. C5 Chromane Comparison.

^a124/125 HIV-1 integrase polymorphisms.

^b50% effective concentration. The cell-culture antiviral assay that evaluated the ability of compounds to inhibit viral outgrowth of the NLRepRluc virus in MT-2 cells was performed in the presence of 10% FBS.⁷ Values are reported as mean ± standard deviation (N ≥ 2).

^cAll of the MT-2 cell CC₅₀ values were ≥22 μM.

Subsequently, replacement of the chromane moiety with a tetrahydroisoquinoline (THIQ) heterocycle was explored. This enabled the p-fluorophenyl ring to be projected in a similar manner to the chromane, albeit at a different angle, while removing the complication of the inherent chirality of the chromane scaffold. The resulting analogue 7 incurred a modest loss in potency against the T/T polymorph (3.7-fold) and was also associated with reduced metabolic stability when incubated in human liver microsomes (HLMs), with a >2-fold shorter half-life compared with 4 (Table 2). Additionally, a diverse set of C5 spacers were prepared and evaluated. Replacement of the C5 proximal phenyl ring of 4 with either a tetrahydropyridine (8), pyrazole (9), or pyrimidine (10) heterocycle resulted in both a loss of T/T polymorph inhibitory potency (between 4.8-fold and 7.7-fold) and a reduction of metabolic stability in HLMs. Pyridine 11, while showing good metabolic stability in HLMs (>120 min), exhibited a 7.2-fold reduction in potency toward the T/T-containing virus.

Table 2. Additional C5 Modifications.

^a124/125 HIV-1 integrase polymorphisms.

^b50% effective concentration. The cell-culture antiviral assay that evaluated the ability of compounds to inhibit viral outgrowth of the NLRepRluc virus in MT-2 cells was performed in the presence of 10% FBS.⁷ Values are reported as mean ± standard deviation (N ≥ 2).

^cAll of the MT-2 cell CC₅₀ values were ≥22 μM.

The THIQ heterocycle exemplified by 7 provided an intriguing scaffold with which to explore a wide range of substituents that could readily be introduced on the ring nitrogen atom and were not easily accessible in the previous series. A total of 229 N-benzyl-substituted THIQ analogues were prepared in an initial survey designed to extensively explore the structure–activity relationship (SAR). Of the compounds prepared, 46 (20%) had promising anti-HIV-1 potencies and were subsequently assessed in a panel of in vitro assays to evaluate their metabolic stability in liver microsomes (LMs) as well as for their potential drug–drug interaction (DDI) liabilities. Of these, the five analogues 12–16 showed both a low potential for DDI liabilities and were metabolically stable in LMs and therefore were evaluated in rat pharmacokinetic (PK) studies. The antiviral and PK data are summarized in Table 3. Substitution of the phenyl ring with F, CH₃, and Cl groups resulted in analogues with comparable potency against both the T/T and N/A polymorphs and low to moderate clearance (CL) in rats after intravenous (iv) dosing. Compounds containing an o-methyl (13, 14) or o-chloro (12) substituent provided the lowest observed CL in this cohort. The lower clearance of 12–14 could be due to greater shielding of the THIQ nitrogen by the bulky ortho substituents compared with analogues 15 and 16, which demonstrated moderate CL.

Table 3. Antiviral Profile and Rat PK Data for Selected Substituted N-Benzyl-THIQ Analogues.

^a124/125 HIV-1 integrase polymorphisms.

^b50% effective concentration. The cell-culture antiviral assay that evaluated the ability of compounds to inhibit viral outgrowth of the NLRepRluc virus in MT-2 cells was performed in the presence of 10% FBS.⁷ Values are reported as mean ± standard deviation (N ≥ 2).

^cAll of the MT-2 cell CC₅₀ values were ≥24 μM.

^dMale Sprague-Dawley rats were dosed at 1 mg/kg iv and 5 mg/kg po as solutions in 90:10 PEG-400/EtOH for iv dosing and 90:5:5 PEG-400/EtOH/TPGS for po dosing.

The versatility of the THIQ scaffold enabled the exploration of a range of substituents in addition to the benzyl moiety. One promising series that was identified comprised heterocycles directly linked to the THIQ nitrogen atom. The most potent compounds from this series were evaluated in iv/po PK studies in rats, and the antiviral and PK data are summarized in Table 4. Benzofuropyrimidine 17 demonstrated low-single-digit nanomolar inhibitory potency against the T/T and N/A polymorphs and also exhibited low CL in the rat with good oral bioavailability (%F). While both pyrazolopyrimidine 18 and pyrazinylpyrimidine 19 had low CL in rat after iv dosing and single-digit nanomolar inhibitory potency against the T/T and N/A polymorphs, 17 exhibited a superior profile compared with 18 and 19. Also, thiazolopyridine analogue 20 had low-single-digit nanomolar inhibitory potency against the T/T and N/A viruses and demonstrated moderate CL in rat after iv dosing. Compounds 12, 13, and 17 showed a promising balance of antiviral potency and rat PK profile that prompted subsequent evaluation in dog iv PK studies (male beagle dogs dosed at 1 mg/kg as a solution in 90:10 PEG-400/ethanol) where it was found that 12 had a lower CL (0.5 mL min^–1 kg^–1) compared with 13 (3.4 mL min^–1 kg^–1) and 17 (10.1 mL min^–1 kg^–1). On the basis of this data, 12 was selected for further progression.

Table 4. Antiviral Profile and Rat PK Data for Selected N-Substituted THIQ derivatives.

^a124/125 HIV-1 integrase polymorphisms.

^b50% effective concentration. The cell-culture antiviral assay that evaluated the ability of compounds to inhibit viral outgrowth of the NLRepRluc virus in MT-2 cells was performed in the presence of 10% FBS.⁷ Values are reported as mean ± standard deviation (N ≥ 2).

^cAll of the MT-2 cell CC₅₀ values were ≥17 μM.

^dMale Sprague-Dawley rats were dosed at 1 mg/kg iv and 5 mg/kg po as solutions in 90:10 PEG-400/EtOH for iv dosing and 90:5:5 PEG-400/EtOH/TPGS for po dosing.

More extensive in vitro and in vivo studies were performed to characterize the virology, metabolic stability, DDI potential, and PK profile of 12. First, 12 was evaluated in an expanded polymorph virology panel, where it demonstrated broad-spectrum inhibitory potency (Table 5). Additionally, 12 demonstrated a low DDI potential and good metabolic stability across human, rat, dog, and cynomolgus monkey LMs (half-lives of >120, >120, >120, and 106 min, respectively). The PK data for 12 were evaluated in additional species, and 12 was found to exhibit low CL across all of the preclinical species (Table 5). A low human CL of 0.76 mL min^–1 kg^–1 after iv dosing was predicted on the basis of allometric scaling, and an oral bioavailability of 58% was predicted on the basis of the average oral bioavailability across the preclinical species evaluated. This approach was used to predict that an estimated qd dose of between 210 and 480 mg would be needed in order to maintain the minimum efficacious plasma concentration of 0.85 μM (3 times the protein-binding-adjusted EC₉₀ of the N/A 124/125 polymorph) at steady state.

Table 5. Antiviral and PK Data for 12.

124/125 polymorphism	EC50 (nM)a	n
T/Tb	3.5 ± 2.3	303
A/T	5.8 ± 3.2	83
T/A	5.4 ± 1.6	35
N/T	5.8 ± 2.6	36
A/A	6.5 ± 3.2	87
N/A	7.5 ± 5.1	173
T/V	3.2 ± 1.2	35

species	iv CL (mL min–1 kg–1)g	po AUC0–∞ (μM h)h	%Fh
mousec	3.9	30.3	72
ratd	3.5	14.9	35
doge	0.5	103.6	90
cynof	3.3	5.4	33
human (predicted)	0.76	60.9–139	58

^a50% effective concentration. The cell-culture antiviral assay that evaluated the ability of compounds to inhibit viral outgrowth of the NLRepRluc virus in MT-2 cells was performed in the presence of 10% FBS.⁷ Values are reported as mean ± standard deviation (N ≥ 2).

^bThe MT-2 cell CC₅₀ value was 22.5 μM.

^cMale CD-1 mice.

^dMale Sprague-Dawley rats.

^eMale beagles.

^fMale cynomolgus monkey.

^gDosed at 1 mg/kg iv as a solution in 90:10 PEG-400/EtOH.

^hDosed po as a solution in 90:5:5 PEG-400/EtOH/TPGS at 5 mg/kg for mouse and rat and 2 mg/kg for dog and monkey.

The promising virology and PK data for 12 along with the structural diversity compared with 4 prompted evaluation of 12 in rat toxicology studies. In a 2 week toxicology study conducted in rats where 50, 100, and 200 mg/kg doses of 12 were administered po once daily, a dose-related increase in liver weight with a microscopic correlative increase in incidence and severity (minimal to marked) of hepatocellular vacuolation containing neutral lipids was observed. There were no accompanying changes in clinical pathology parameters or other histological changes. While the translation of this finding in rats to higher species is unknown, the result highlighted the need for an in vitro assay that could potentially be used to help triage compounds for this liability. A subsequent breakthrough on this front resulted in the identification of such an assay. Rat H-4-II-E and human Huh7 liver cells were treated with 10 μM test compound for 72 h and stained with boron–dipyrromethene (BODIPY), and the numbers of lipid droplets within the cells were determined through quantitation by high-content screening (HCS) fluorescence and compared with a DMSO control.⁸ Once the assay became available, 12 was profiled alongside 4. Both 4 and 12 exhibited high fold changes in lipid droplet count in the rat liver cell line (Table 6), which is consistent with the in vivo observations in their respective rat toxicology studies.^1f

Table 6. Lipid Vacuolation Fold Changes in Lipid Droplet Count for C6 Methyl and C6 Hydrogen Analogues⁸.

Of additional concern, high fold changes in lipid droplet count were also observed for both 4 and 12 in a human liver cell line (Table 6). While the mechanism for vacuolation remains unknown, an extensive survey in the lipid-droplet assay of compounds bearing a range of structural changes revealed that analogues lacking the C6 methyl group found in both 4 and 12 consistently appeared to show a lower propensity to cause vacuolation. Compounds lacking the C6 methyl substituent exhibited an average of 2.4 times lower fold change in lipid droplet count in this in vitro assay compared with the matched C6 methyl analogues in both the human and rat liver cell lines. Specifically, removal of the C6 methyl group in 4 and 12 afforded compounds 21 and 22, respectively, both of which exhibited a lower lipid droplet count in the rat and human cell lines (Table 6). Compound 22 showed more robust relative reduction compared with 21. Two other apparent differences that resulted from the C6 substituent change from methyl to hydrogen are that the nitrogen atom of the basic pyridine heterocyclic core is more exposed and the atropisomerism associated with 12 is eliminated. An unrealized but desired goal is to understand how these differences, or other differences that may be less apparent, play a role in reducing the amount of lipid vacuolation observed in the in vitro assay.

As a next step, the broader consequence of removing the C6 methyl group was more extensively characterized. Because of the superior profile in the lipid droplet assay, compound 22 was evaluated in the expanded virology polymorph panel, and the results were comparable to those for 12 (Table 7), with only small differences between the most and least sensitive polymorph. Similarly, 22 exhibited low potential for DDIs and demonstrated excellent metabolic stability in LMs in vitro (half-lives of >120 min in LMs of human, rat, dog, and cynomolgus monkey). Because of the promising virology profile and reduced potential for lipid vacuolation, the PK profile of 22 was evaluated in four preclinical species following iv and po dosing. Compound 22 exhibited low CL across the four preclinical species that was modestly improved compared with 12. The combination of the favorable virology and PK properties of 22 resulted in a predicted human qd oral dose of 62–200 mg to maintain a minimum efficacious plasma concentration of 0.34 μM (3 times the protein-binding-adjusted EC₉₀ of the N/A 124/125 polymorph) at steady state, which was lower than the dose predicted for 12.

Table 7. Antiviral and PK Data for 22.

124/125 polymorphism	EC50 (nM)a	n
T/Tb	2.8 ± 1.5	152
A/T	4.5 ± 1.1	6
T/A	5.0 ± 1.5	6
N/T	3.1 ± 0.6	8
A/A	4.6 ± 1.8	48
N/A	5.7 ± 2.0	47
T/V	1.6 ± 0.8	8

species	iv CL (mL min–1 kg–1)g	po AUC0–∞ (μM h)h	%Fh
mousec	1.0	43.1	47
ratd	3.8	21.6	71
doge	2.2	24.2	89
cynof	1.7	10.7	51
human (predicted)	1.0	18.2–76.1	65

^a50% effective concentration. The cell-culture antiviral assay that evaluated the ability of compounds to inhibit viral outgrowth of the NLRepRluc virus in MT-2 cells was performed in the presence of 10% FBS.⁷ Values are reported as mean ± standard deviation (N ≥ 2).

^bThe MT-2 cell CC₅₀ value was 19.5 μM.

^cMale CD-1 mice.

^dMale Sprague-Dawley rats.

^eMale beagles dogs.

^fMale cynomolgus monkey.

^gDosed at 1 mg/kg iv as a solution in 90:10 PEG-400/EtOH.

^hDosed po as a solution in 90:5:5 PEG-400/EtOH/TPGS at 5 mg/kg for mouse and rat and 2 mg/kg for dog and monkey.

To contribute to a further understanding of 22, an X-ray cocrystal structure of 22 with the IN catalytic core domain (CCD) dimer was obtained (Figure 3). Compound 22 was observed to bind to the IN dimer interface and make several key contacts with the IN monomers. The carboxyl moiety makes two hydrogen-bonding contacts with the backbone amide NHs of residues E170 and H171. In addition, the tert-butyl ether oxygen atom and one of the carboxyl oxygen atoms engage in two additional hydrogen-bonding contacts with the hydroxy moiety of T174. The 4,4-dimethylpiperidine and tert-butyl groups engage the protein through van der Waals interactions by occupying the hydrophobic pocket deep within the dimer interface. Furthermore, the 2-chloro-6-methylbenzyl group occupies the A128–W131 hydrophobic shelf and makes a π–edge contact with the W131 residue. The orientation of the THIQ moiety avoids close contact with residues 124 and 125, thus potentially accounting for the small variation in 124/125 polymorphic potency observed.

Figure 3.

X-ray cocrystal structure of 22 in the HIV-1 integrase CCD dimer. The monomers of HIV-1 integrase are colored distinctly. The X-ray protein–ligand complex was prepared and minimized using the Protein Preparation Wizard (Schrödinger, LLC). The coordinates have been deposited in the Protein Data Bank under PDB code 7UOQ. Side chains of E170 and H171 have been omitted for clarity. The 2D structure of 22 is shown in the inset for reference. The figure was prepared using the PyMOL Molecular Graphics System, ver. 2.2.0 (Schrödinger, LLC).

The overall superior profile of 22 versus 12 and 4, including the reduction in the number of lipid droplets observed in the in vitro lipid vacuolation assay, prompted evaluation of 22 in an in vivo toxicology study. Specifically, 22 was evaluated in a one-week rat toxicity study conducted at qd po doses of 20 and 50 mg/kg. The compound was clinically tolerated at both doses, and there were no compound-related findings at the 20 mg/kg dose. At the 50 mg/kg dose, minimally higher alkaline phosphatase (ALP) levels (1.3×) and minimal hepatocyte vacuolation were observed compared to control animals. The exposure of 22 in this study exceeded that achieved with 12 in the rat toxicology study where adverse events were observed. At a qd po dose of 50 mg/kg, the C_max and AUC values achieved with 22 were 488 μM and 7535 μM h, respectively, compared with 32 μM and 335 μM h for 12 at the same dose. This represents a 15-fold higher exposure of 22 over 12 with respect to C_max and 22-fold higher exposure with respect to the AUC. Additionally, the C_max and AUC for 22 at a qd po dose of 50 mg/kg were 8-fold higher than those achieved for 12 at a qd po dose of 200 mg/kg. For 12 at a qd po dose of 200 mg/kg, the level of hepatocellular vacuolation ranged from mild to marked, whereas with 22 at the qd po dose of 50 mg/kg there was a noticeable improvement with only minimal vacuolation noted. As a result, 22 was evaluated in a 1 week dog toxicology study at qd po doses of 20 and 60 mg/kg and was clinically tolerated at both doses. Compound-related findings at the 60 mg/kg dose were limited to mildly elevated alanine aminotransferase (ALT) levels (3.6× pretest or 1.8× upper limit of the normal range) and mild hepatocyte vacuolation in one of the two dogs at the 60 mg/kg dose. There were no significant compound-related findings in the other dog at this dose level despite similar systemic exposure. On the basis of the relatively favorable results observed in both rat and dog 1 week toxicology studies, development of 22 was continued.

Further work was performed to understand the in vitro to in vivo translation of the lipid droplet assessment dependent on compound concentration. In the in vitro lipid vacuolation assay, a 22 concentration of 10 μM showed a lower fold change in the rat liver cell line compared with 12 (Table 6). This appears to have translated into a lower incidence and severity of liver findings in the 1 week rat toxicology studies. However, the relatively high exposures observed in the in vivo toxicology studies prompted an evaluation of both compounds at higher concentrations in the in vitro assay. The fold change in lipid droplets rose to 3.7-fold for 22 at 20 μM (from 1.8-fold at 10 μM). By comparison, at 20 μM, 12 demonstrated a 9.1-fold change (vs 3.7-fold at 10 μM). Both compounds displayed cytotoxicity in the in vitro assay at concentrations of 40 and 80 μM. This information suggested that compounds with a low fold change at higher concentrations in the in vitro assay might demonstrate an improved profile in the in vivo toxicology studies. Moreover, the in vivo observations in dogs highlight the need for an in vitro dog liver cell line in addition to the rat and human cell lines to help assess the relative potential for vacuolation in the dog. The findings associated with this class of compounds remain a challenge because of the differences seen between species and the lack of translational knowledge to humans. However, the development of an in vitro lipid vacuolation assay provided a valuable tool that can be used to help triage compounds before they are evaluated in in vivo toxicology studies.

In summary, the THIQ moiety was identified as an alternate C5 spacer for the class of 4-(4,4-dimethylpiperidinyl)-2,6-dimethylpyridinyl-based allosteric inhibitors of HIV-1 integrase. SAR exploration identified 12 as a compound exhibiting broad-spectrum antiviral potency and favorable PK characteristics across preclinical species, which resulted in a low predicted human dose. Evaluation in rat toxicology studies highlighted a potential lipid vacuolation liability. Subsequent development and use of an in vitro lipid droplet assay to triage compounds revealed that removal of the C6 methyl substituent, a seemingly small structural change,⁹ reduced the lipidosis liability in the in vitro assay. Evaluation of the resulting compound 22 in toxicology studies showed reduced incidence and severity of lipid vacuolation at significantly higher exposure levels compared with 12, thus validating the in vitro/in vivo correlation. Ultimately, after evaluation of 22 in a 1 month GLP toxicology study, the development of 22 was halted because of the risk of progression or worsening of the lipid vacuolation upon chronic dosing. It is our hope that the lessons learned in this work can be used by others to further the discovery of ALLINIs with clinical potential to treat HIV infection.

Glossary

Abbreviations

ALLINI

allosteric integrase inhibitor

ALP

alkaline phosphatase

ALT

alanine aminotransferase

AUC

area under the curve

CCD

catalytic core domain

CL

clearance

C_max

maximum concentration

DDI

drug–drug interaction

DMSO

dimethyl sulfoxide

EtOH

ethanol

HCS

high-content screening

H₂O

water

HIV

human immunodeficiency virus

HLM

human liver microsomes

IN

integrase

INSTI

integrase strand transfer inhibitor

iv

intravenous

LEDGF

lens epithelium-derived growth factor

LMs

liver microsomes

PEG-400

poly(ethylene glycol) 400

PK

pharmacokinetic

THIQ

tetrahydroisoquinoline

TPGS

d-α-tocopheryl PEG-1000 succinate

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.2c00115.

• Synthetic schemes, experimental procedures, analytical data, biological and DMPK methods, and authors’ contact information (PDF)

The authors declare no competing financial interest.

Author Status

^§ M.G.S. and C.C. are retired from Bristol Myers Squibb; the work was carried out in Wallingford, CT.