Chemico-Biological Interactions 142 (2002) 135 /154
`www.elsevier.com/locate/chembiont
`
`The metabolic activation of abacavir by human
`liver cytosol and expressed human alcohol
`dehydrogenase isozymes
`
`John S. Walsh *, Melinda J. Reese, Linda M. Thurmond
`
`Division of Drug Metabolism and Pharmacokinetics, GlaxoSmithKline, Research Triangle Park,
`NC 27709-3398, USA
`
`Abstract
`
`Abacavir (ZIAGEN†) is a reverse transcriptase inhibitor marketed for the treatment of
`HIV-1 infection. A small percentage of patients experience a hypersensitivity reaction
`indicating immune system involvement and bioactivation. A major route of metabolism for
`abacavir is oxidation of a primary bg unsaturated alcohol to a carboxylic acid via an aldehyde
`intermediate. This process was shown to be mediated in vitro by human cytosol and NAD,
`and subsequently the aa and g2g2 human isoforms of alcohol dehydrogenase (ADH). The aa
`isoform effected two sequential oxidation steps to form the acid metabolite and two isomers,
`qualitatively reflective of in vitro cytosolic profiles. The g2g2 isozyme generated primarily an
`isomer of abacavir, which was minor in the aa profiles. The aldehyde intermediate could be
`trapped in incubations with both isozymes as an oxime derivative. These metabolites can be
`rationalized as arising via the aldehyde which undergoes isomerization and further oxidation
`by the aa enzyme or reduction by the g2g2 isozyme. Non-extractable abacavir protein residues
`were generated in cytosol, and with aa and g2g2 incubations in the presence of human serum
`albumin (HSA). Metabolism and residue formation were blocked by the ADH inhibitor 4-
`methyl pyrazole (4-MP). The residues generated by the aa and g2g2 incubations were analyzed
`by SDS-PAGE with immunochemical detection. The binding of rabbit anti-abacavir antibody
`to abacavir-HSA was shown to be dependent on metabolism (i.e. NAD-dependent and 4-MP
`sensitive). The mechanism of covalent binding remains to be established, but significantly less
`abacavir-protein residue was detected with an analog of abacavir in which the double bond
`was removed, suggestive of a double bond migration and 1,4 addition process.
`# 2002 Elsevier Science Ireland Ltd. All rights reserved.
`
`* Corresponding author. Tel.: /1-919-469-4719; fax: /1-919-315-6003
`E-mail address: John.S.Walsh@GSK.com (J.S. Walsh).
`
`0009-2797/02/$ - see front matter # 2002 Elsevier Science Ireland Ltd. All rights reserved.
`PII: S 0 0 0 9 - 2 7 9 7 ( 0 2 ) 0 0 0 5 9 - 5
`
`1 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`136
`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`Keywords: Abacavir; HIV-1; Carboxylic acid; Glucuronide
`
`1. Introduction
`
`Adverse drug reactions represent an area of increasing clinical concern for both
`prescribing physicians and pharmaceutical companies [1,2]. Approximately 25% of
`these are designated as idiosyncratic, or hypersensitivity reactions, in that they do
`not occur in most patients, and do not involve the known pharmacological
`properties of the drug [3]. There is substantial evidence of immune system mediation,
`which in turn may be initiated through metabolic activation of the drug to reactive
`species which covalently bind to proteins [4,5].
`Abacavir (ZIAGEN†, Fig. 1) is a nucleoside reverse transcriptase inhibitor
`marketed in 1999 for the treatment of HIV-1 infection. Approximately 4% of
`
`Fig. 1. Major routes of metabolism of abacavir in humans. Following a 600 mg oral dose of 14C-abacavir,
`83% of the radioactivity was excreted in the urine, and 16% in the feces. In urine, the acid metabolite
`(2269W93) accounted for 30% of the dose and the glucuronide for 36%, with 1% unmetabolized abacavir.
`The remaining 16% of dose was comprised of numerous minor metabolites [8]. A number of these have
`been characterized, and include two isomers of 2269W93. The fecal component was comprised primarily of
`2269W93 and abacavir.
`
`2 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`137
`
`patients experience a hypersensitivity reaction. The symptoms are varied, but fever
`and rash are the most common [6], and a small number of fatalities have also been
`reported. The clinical presentation is consistent with immune mediation, and hence
`bioactivation is likely to play a role in its origination.
`As part of broader investigations to understand the etiology of this hypersensi-
`tivity, metabolic investigations were performed to identify potential bioactivation
`pathways for abacavir, and the enzyme systems involved. While many factors are
`likely to play a role in idiosyncratic toxicities [7], genetic polymorphisms in drug
`metabolizing enzymes are factors that can potentially be identified.
`The major routes of metabolism of abacavir in humans are conjugation to an ether
`glucuronide, and oxidation to a carboxylic acid metabolite designated as 2269W93
`[8], (Fig. 1). Approximately 15% of the dose is comprised of a number of minor
`metabolites, but from structural considerations, none of those identified to date
`indicate the potential for reactive intermediates. In contrast, the metabolism of
`abacavir to 2269W93 involves a two step oxidation process via an aldehyde
`intermediate. This aldehyde has not been observed directly as a metabolite of
`abacavir, and attempts to synthesize it have been unsuccessful due to its apparent
`instability.
`A number of aldehyde metabolites have been previously implicated as reactive and
`capable of covalent binding to proteins, and this in turn has been suggested to
`underlay clinical adverse events of the parent compounds. In these cases, the
`underlying reactivity can be ascribed to one of two mechanisms: Schiff base
`formation, or a 1,4 addition process. Schiff base formation has been proposed for
`the aldehyde metabolite of Sorbinil, which shows an incidence of hypersensitivity [9],
`and for ethanol, in which acetaldyde formation is proposed to underlay the immune
`hepatitis observed with chronic alcohol consumption [10]. The reversibility of this
`Schiff base formation has been proposed to be limited either by reduction by
`ascorbate [11], or for acetaldehyde, cyclization with an amide nitrogen of the protein
`to form a stable imidazolidinone [12]. Thus mechanisms may exist in vivo for the
`stabilization of these somewhat unstable adducts. A more widely encountered
`mechanism is observed for ab unsaturated aldehydes (Michael type acceptors) and
`numerous examples exist in the literature of proposals for this type of reactive
`intermediate being associated with covalent binding and adverse clinical events [13 /
`15].
`As discussed below, both mechanisms might be possible for abacavir. The
`purposes of the studies described here were to utilize in vitro methods to determine
`if oxidation of abacavir to the carboxylic acid via an aldehyde intermediate could
`lead to protein covalent binding, to investigate the mechanism, and to identify the
`enzyme systems involved.
`
`3 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`138
`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`2. Materials and methods
`
`2.1. Chemicals
`
`Human liver microsomes and cytosol were obtained from Xenotech, (Kansas City,
`KS). NADP, glucose-6-phosphate, glucose-6-phosphate dehydrogenase, NAD, 4-
`MP, methoxylamine, horse liver alcohol dehydrogenase (HLADH), HSA, triethy-
`lamine, toluene, isobutylchloroformate, and 7-ethoxycoumarin were obtained from
`Sigma-Aldrich (St Louis, MO). Laemmli sample buffer was obtained from Bio-Rad
`(Hercules, CA). Ultima-Flo M was obtained from Packard (Meriden, CT). Gelcode
`Blue Reagent, SuperBlock Blocking Buffer, SuperSignal West Pico Chemilumines-
`cent Substrate, horseradish peroxidase-conjugated goat anti-rabbit IgG (H/L
`chains), BupH Tris Buffered Saline Packs, Blue Ranger pre-stained protein
`standards, and keyhole limpet hemocyanin (KLH) were obtained from Pierce
`(Rockford, IL). Immobilized recombinant protein A IPA-400HC was purchased
`from RepliGen (Needham, MA). Abacavir, 14C- abacavir, 2269W93, and dihydro-
`abacavir were synthesized at GlaxoSmithKline.
`Human ADH isozymes were obtained as a gift from Dr.Tom Hurley, Dept of
`Biochemistry and Molecular Biology, Indiana University Medical School.
`Isozymes were analyzed by SDS-PAGE and each shown to be comprised of a
`single major protein. The identities of the aa and g2g2 isozymes were independently
`confirmed by LC/MS/MS analysis of tryptic digests using MASCOT protein
`database searching [16].
`
`2.2. Microsomal incubations
`
`14C-Abacavir (3 mg/ml; 0.5 mCi/ml) was incubated with 1 mg/ml pooled human
`liver microsomes in 100 mM potassium phosphate buffer, pH 7.4, containing 1 mM
`NADP, 5 mM MgCl2, 10 mM glucose-6-phosphate, and 0.5 Units/ml glucose-6-
`phosphate dehydrogenase. Incubations containing no NADP were used as negative
`controls. After incubating the samples in duplicate for 1-20 h in a 37 8C shaking
`incubator, 50 ml of each incubation was removed and added to 50 ml of cold
`acetontrile. The samples were vortexed and centrifuged in an Eppendorf model 5314
`microcentrifuge at 10,000 rpm for 5 min. An aliquot of 20 ml was analyzed by HPLC
`for metabolite profiling. The remainder of the incubations were analyzed for non-
`extractable residue formation as described below. Metabolic viability of
`the
`microsomes was confirmed using 7-ethoxycoumarin.
`
`2.3. Cytosol and expressed ADH incubations
`
`14C-Abacavir (3 mg/ml; 0.5 mCi/ml) was incubated with 1 mg/ml of pooled human
`liver cytosol in 50mM sodium pyrophosphate buffer, pH 7.4 or pH 8.8, containing
`7.5 mM NAD. Incubations containing no NAD were used as negative controls.
`When used, 4-MP was added as an aqueous solution at the start of the incubation, to
`give a final concentration of 0.6 mM. Incubations were run in duplicate for 2-20
`
`4 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`139
`
`hours in a 37 8C shaking incubator, and then placed on ice for 5 min prior to
`filtering a 100 ml aliquot through a 0.45 mm AcroPrep GHP filter plate using a
`Waters Alliance filtration system. Aliquots of /20 ml were analyzed by HPLC and
`LC/MS for metabolite profiling.
`Incubations with human ADH isozymes were performed as described for cytosol,
`in 50 mM sodium pyrophosphate buffer, pH 7.4 and pH 7.8 (aa isozyme), or pH 8.8.
`Screening for metabolic activity with abacavir was performed with the different
`ADH isozymes using constant units of activity (0.025 Units/ml each) with variable
`enzyme protein level, and at constant enzyme protein level (17 mg/ml) with a variable
`number of units. Incubations were analyzed at 1, 2, 3, 4 and 20 hours by HPLC.
`When present, HSA was at a final concentration of 20 mg/ml. Incubations with
`dihydro-abacavir were performed under the same conditions. ADH inhibition by 4-
`MP was performed as described for cytosolic incubations. For trapping experiments
`with methoxylamine, an aqueous solution was added at the start of each incubation
`such that the final concentration was 10 mM.
`All human ADH isozymes were assayed for viability using either ethanol (aa,
`b1b1, b2b2, g2g2, p and s) or pentanol (x) as substrates. Metabolism was assessed
`by measuring initial rates of NADH formation spectrophotometrically, and
`expressed as units of activity (mmol NADH formed/min) as previously described [17].
`
`2.4. Measurement of Non-Extractable Protein Residues
`
`Incubations were placed on ice for 5 min and the proteins precipitated by adding 3
`ml cold acetonitrile. The samples were vortexed and centrifuged at 2500 rpm for 5
`min at 4-8 8C. The supernatants were removed and the pellets resuspended in 3 ml
`cold acetonitrile, vortexed, and centrifuged as before for a total of 5 washes or until
`the radioactivity counted in the washes wereB/2 times background. The pellets were
`dissolved in 1 ml 1%SDS and 0.8 ml was counted in a liquid scintillation counter.
`Some samples were further analyzed by SDS-PAGE/Western blot analysis.
`
`2.5. HPLC and LC/MS analyses
`
`Incubation supernatants were analyzed using a Waters Alliance 2690 HPLC
`system. The eluate was monitored using a Waters 996 photo diode array detector
`(l/295nm) and a Berthold Model LB 507A radioactivity flow monitor (EG & G
`Berthold, Nashua, NH), equipped with a 1 ml flow cell. The HPLC eluate was mixed
`with 3 volumes of Ultima-Flo M scintillation cocktail using a Waters 510 pump prior
`to entering the flow cell. HPLC analyses were performed on a Kromasil C18, 5 mm
`column (3.2/150mm; Phenomenex, Torrance, CA) with a mobile phase consisting
`of 0.025 M ammonium acetate buffer (pH 4) containing 5% methanol (A) and
`acetonitrile (B) delivered as a linear 50 min gradient of 0-13% B, at a flow rate of 0.7
`ml /min.
`While optimum chromatographic resolution of metabolites was achieved with the
`chromatographic conditions described above, this gave poor mass spectral response.
`For LC/MS analyses, the same column was used with a mobile phase consisting of
`
`5 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`140
`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`5% methanol/0.1% formic acid/water (A) and 0.1% formic acid/acetonitrile (B)
`delivered as a linear 50 min gradient of 0-13% B, at a flow rate of 0.7 ml /min.
`Samples were analyzed using a Quattro II triple quadrupole mass spectrometer
`(Micromass Ltd, Manchester, UK) equipped with a Z-spray ion source operated in
`positive ion mode. Nitrogen was used as the nebulizing and drying gas. Source and
`desolvation temperatures were 100 and 350 8C, respectively. Capillary and cone
`voltages were 2.5 kV and 15 V, respectively. The multiplier voltage was 650 V. Full-
`scanning mass spectra were acquired between 100-500 amu. Data were processed
`using MassLynx 3.3 software.
`
`2.6. Production of rabbit anti-abacavir antibodies
`
`Two abacavir analogs with spacer and carboxyl groups attached to different parts
`of the molecule (analogs A and B, Fig. 9) were conjugated to KLH using the mixed
`anhydride reaction. Mixed anhydrides were formed by mixing triethylamine and
`isobutylchloroformate under nitrogen in an ice bath for 45 min, then added dropwise
`to 25 ml of 0.8 mg/ml KLH in 0.1 M sodium bicarbonate buffer. The reaction was
`allowed to proceed 3 hr at ambient room temperature, then overnight at 48C. The
`reaction mix was dialyzed against phosphate buffer saline and concentrated by
`ultrafiltration to approximately 2 mg/ml in sterile water.
`Conjugation ratios were determined by hapten absorbance at 250 nm corrected for
`KLH, and by protein assay;
`the extinction coefficient
`for each hapten was
`determined empirically. The abacavir:KLH ratios for the KLH conjugates were
`approximately 264 (analog A) and 502 (analog B). Analog A: KLH immunoconju-
`gate was selected as the immunogen, and was diluted 1:1 with either complete
`(primary immunization) or incomplete (boosting immunizations) Freund’s adjuvant;
`New Zealand rabbits were immunized over a period of 6 weeks. Serum antibodies
`were characterized for binding to abacavir analogs using surface plasmon resonance
`(SPR) on a Biacore 2000 (Fig. 9). The KLH conjugates (or KLH alone; control
`surface) were immobilized on a carboxymethyldextran-coated gold chip using amine
`coupling. Serum samples were diluted 1/10 in Hepes-buffered saline, and injected at a
`flow rate of 20 ml/min for 2 min prior to determination of resonance units (RU)
`binding signal. The biosensor surface was regenerated with a 30-sec wash of glycine-
`HCl (pH 1.5) followed by a 30-sec wash with 250 mM NaOH containing 0.1% SDS.
`Antibody specificity was further analyzed by binding to the analog B: KLH
`conjugate. As shown in Fig. 9, the antibody recognized both analog A: and analog B:
`KLH conjugates, demonstrating broad specificity of the antibody. Antibody was
`purified by affinity chromatography on immobilized recombinant protein A
`(RepliGen IPA-400 HC), eluted with 0.2 M glycine buffer (pH 2), and neutralized
`with Tris-HCl buffer (pH 8). The purified antibody was used in western blotting
`experiments to detect abacavir conjugated to HSA.
`
`6 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`141
`
`2.7. SDS-PAGE with Immunoblotting
`
`Samples were prepared for SDS-PAGE analysis by diluting 1:1 in Laemmli sample
`buffer with 0.7 M 2-mercaptoethanol and boiling for 7 min. SDS-PAGE was
`performed using a Mini-PROTEAN II gel system (Bio-Rad) with either 10% or 4 /
`15% gradient Tris-HCl pre-cast gels (Bio-Rad). For immunoblot analysis, the
`resolved proteins were transferred electrophoretically to nitrocellulose membranes
`for 1 hour at 100 V using a Mini Trans-Blot cell (Bio-Rad). The following steps were
`conducted at room temperature with constant shaking. The membranes were
`blocked for 1 hr with blocking buffer (Super Block blocking buffer containing
`0.05% Tween 20), and then incubated with the rabbit anti-abacavir antibodies
`diluted 1:2000 in blocking buffer. Unbound antibodies were removed by washing the
`membranes in wash buffer [Tris Buffered Saline (TBS) containing 0.05% Tween-20]
`(5 washes/5 min each). The membranes were then incubated for 1 hr with
`horseradish peroxidase-conjugated goat anti-rabbit IgG (H/L chains) diluted
`1:50,000 in blocking buffer, and then washed with wash buffer (6 washes/5 min
`each). The blots were developed by incubating with SuperSignal West Pico
`Chemiluminescent Substrate for 5 min, and exposing the nitrocellulose membranes
`to ECL film under safe-light conditions.
`
`3. Results
`
`3.1. Microsomal and cytosol incubations
`
`All incubations were performed at physiologically relevant concentrations (/
`clinical Cmax). No metabolites were detected in incubations of 14C-abacavir with
`human liver microsomes at pH 7.4 from 1 /20 h. In the human liver cytosolic
`incubations containing NAD, polar metabolites were detected at 20 h at pH 7.4,
`including a peak of the same retention time as the major acid metabolite observed in
`vivo, 2269W93 (Fig. 2). Cytosolic NAD-dependent oxidation of alcohols is
`indicative of ADH involvement. This enzyme system has in vitro pH optimums
`higher than 7.4 [18] and is inhibited by 4-MP. As indicated in Fig. 2, significantly
`more metabolism was observed in cytosol at pH 8.8, and metabolism at this pH was
`inhibited completely with 4-MP.
`Following incubations of abacavir with cytosol or microsomes, proteins were
`precipitated, extensively washed, and non-extractable residues determined radio-
`chemically (Fig. 3). No NADPH-dependent non-extractable residues were observed
`in microsomal incubations at 2 h (at which point CYP450 activity is typically
`decreasing). Similarly, no NAD-dependent residues were observed in cytosolic
`incubations at pH 7.4 after 2 h, but at 20 h, NAD-dependent residues were clearly
`observed, and these residues were substantially increased at pH 8.8. Both
`metabolism and residue formation were also blocked in cytosol by 4-MP. These
`results implicate ADH involvement in metabolism and residue formation, rather
`than CYP450.
`
`7 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`142
`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`Fig. 2. Radiochemical profiles of 14C-abacavir after incubating 20 h with human liver cytosol. UV
`detection was used for the 2269W93 reference standard. The retention time of 2269W93 was identical to
`the major metabolite in the cytosolic UV profile.
`
`3.2. Expressed Human ADH Incubations
`
`Some of the properties of the known human ADH isozymes are summarized in
`Table 1. All human isozymes were examined except the g1g1 and b3b3 isoforms,
`which were unavailable. While metabolic turnover with ethanol or pentanol was
`observed at pH 7.4, all isozymes showed significantly higher metabolic activity at pH
`8.8.
`14C-Abacavir was incubated with the different human ADH isoforms at pH 8.8
`and the incubations analyzed by HPLC for evidence of metabolism from 1 to 20 h.
`At 2 h, only the aa and g2g2 isoforms showed any evidence of product formation,
`and further metabolism was observed out to 20 h. No metabolites were detected with
`
`8 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`143
`
`Fig. 3. Non-extractable residues for microsomal (mx) and cytosolic incubations of 14C-abacavir.
`
`Table 1
`
`Gene Alleles Subunit Classa In vitro pH opti-
`mum
`
`Distribution Specific activity (Units/mg pro-
`teinc)
`
`ADH1 ADH1 a
`2 b1
`ADH2 ADH1
`2 b2
`ADH2
`2 b3b
`ADH3
`3 g1b
`ADH3 ADH1
`3 g2
`ADH2
`ADH4 ADH4 p
`ADH5 ADH5 x
`ADH7 ADH7 s
`
`I
`I
`I
`I
`I
`I
`II
`III
`IV
`
`10.5
`10.5
`8.5
`7.0
`10.5
`10.5
`10.5
`10.5
`10.5
`
`1.8d
`Hepatic
`0.013e
`Hepatic
`6.4d
`Hepatic
` /
`Hepatic
` /
`Hepatic
`0.34d
`Hepatic
`0.07e
`Hepatic
`Many tissues 0.37f
`9.2d
`Stomach
`
`a Only class I are sensitive to inhibition by 4-MP.
`b Isozymes not examined in this study.
`c Assayed as described previously [17], at pH 8.8. Mean of 2 /4 determinations.
`d 30 mM ethanol.
`e 600 mM ethanol.
`f 300 mM pentanol.
`
`any of the other isoforms out to 20 h at pH 8.8. For the different isozymes, specific
`activities (i.e. units of activity/mg of protein) were variable, and so incubations were
`run both at constant units of activity with variable enzyme protein levels, and at
`constant enzyme protein levels with variable units of activity. Under both sets of
`conditions, only the aa and g2g2 enzymes showed metabolic activity with abacavir.
`Fig. 4 shows the radiochemical profiles at 20 h for the aa and g2g2 incubations run
`at pH 8.8. The aa isozyme gave a profile similar to that seen in cytosol, with a peak
`of the same retention time of the acid metabolite, and at least one other polar
`product. The profile for the g2g2 isoform was quite different. No acid metabolite or
`polar peaks were observed. Instead a single slightly less polar product was generated,
`which was minor in the aa incubations. All of these peaks were NAD and enzyme
`dependent.
`
`9 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`144
`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`Fig. 4. Radiochemical profiles of 14C-abacavir after incubating 20 h with expressed human aa and g2g2
`ADH isozymes.
`
`Because high pH optimums are required for alcohol oxidation by ADH, the effect
`of pH on abacavir metabolism by the aa isozyme was examined at lower, more
`physiological pHs. Incubations at pH 7.4 showed only trace metabolism. At pH 7.8,
`metabolite peaks were clearly observed, giving a similar qualitative profile to that
`observed at pH 8.8 (Fig. 4). Insufficient quantities of the g2g2 isozyme were available
`to perform these studies with this isozyme.
`
`10 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`145
`
`Fig. 5. Selected ion traces for incubations of 14C-abacavir with expressed human aa and g2g2 isozymes.
`The aa isozyme traces (a, b) show the protonated molecular ion for the 14C isotope of the carboxylic acid
`metabolite (m/z 303) (a), and the 12C isotope (m/z 301) when co-injected with 2269W93 (b). The
`protonated molecular ion for the 14C isotope of abacavir (m/z 289) is shown for the g2g2 isozyme (d), and
`in the absence of enzyme (e).
`
`LC/MS profiles for the aa and g2g2 incubations run at pH 8.8 are shown in Fig. 5.
`Analysis was done under different chromatographic conditions than used for the
`radiochemical profiling, so relative retention times are shifted. The aa isozyme trace
`(Fig. 5, a) shows the protonated molecular ion for the 14C isotope of 2269W93 (m/z
`303) which was the major isotope in these incubations due to the high specific
`activity of abacavir used. The protonated molecular ion for reference 2269W93 is m/
`z 301 (12C isotope). In the aa incubations, three discrete peaks with apparent
`molecular ions corresponding to carboxylic acid were detected, and one of these co-
`eluted with 2269W93 in the m/z 301 trace for the incubation (Fig. 5, b). These peaks
`are absent in incubations containing abacavir but no enzyme (Fig. 5, c). For the g2g2
`
`Fig. 6. Non-extractable 14C residues of incubations of 14C-abacavir with aa and g2g2 human ADH
`isozymes, in the presence of human serum albumin.
`
`11 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`Fig.7.Radiochemicalprofilesofaaandg2g2incubationsof14C-abacavirinthepresenceof10mMmethoxylamine.
`
`146
`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`12 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`147
`
`incubations, these acid metabolites were detected in only trace amounts by LC/MS.
`In contrast, the less polar peak, which was minor in the aa incubations, was observed
`as the only major metabolite and had the same apparent molecular weight as parent
`(m/z 289 for the 14C isotope, Fig. 5, d). This peak is absent in incubations with no
`enzyme (Fig. 5, e).
`Incubations of both aa and g2g2 ADH enzymes in the presence of HSA gave rise
`to non-extractable protein residues, which were inhibited by 4-MP. For the aa
`isozyme, a lower level of 4-MP sensitive residue was detected at pH 7.4 (Fig. 6).
`While the aldehyde intermediate was not observed directly in these incubations, its
`presence was confirmed by trapping in situ with methoxylamine. Fig. 7 shows the
`radiochemical profiles of such incubations,
`in which the previously observed
`metabolites are substantially reduced, and a less polar peak is now detected as the
`major product with both isozymes. LC/MS analysis indicated this to have an
`apparent MH/ of 316/314 (for 14C/12C isotopes) consistent with an oxime
`derivative. The mass spectrum for the oxime is shown in Fig. 8.
`
`3.3. Anti-abacavir antibody characterization
`
`The specificity of the anti-abacavir antibody was evaluated by biosensor analysis
`(Fig. 9). Even in the absence of steady-state binding during the association phase,
`significant interaction with the abacavir analog A: KLH immunoconjugate can be
`seen from 170 /350 s (the dissociation phase) compared to the KLH protein alone.
`There was also significant binding to the analog B: KLH conjugate, confirming that
`
`Fig. 8. Mass spectrum of non-polar product observed in ADH incubations with methoxylamine. The
`spectrum shown was obtained from incubation using commercial horse liver alcohol dehydrogenase,
`where higher metabolic turnover was seen. The protonated molecular ions (316/314) for both carbon
`isotopes for the oxime were detected in incubations with the aa and g2g2 human isozymes.
`
`13 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`148
`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`Fig. 9. Rabbit serum antibody binding to immobilized abacavir analogs A (immunogen) and B, coupled
`to KLH. Biosensor data are shown in resonance units (RU), and convey relative antibody binding activity
`to different surfaces [35]. The antibody was suitable for immunoblotting as it was able to recognize
`abacavir when covalently bound to protein either through the purine or cyclopentene ring.
`
`the rabbit antibody was able to recognize abacavir when conjugated to KLH from
`different sites in the molecule.
`
`3.4. Western blot analyses of expressed human ADH incubations
`
`The covalent nature of the non-extractable residues was confirmed by Western
`blot analysis of the incubations. Abacavir incubations were run with the aa and g2g2
`isozymes in the presence of HSA and NAD. Fig. 10 shows immunoblots for these
`incubations together with those for incubations without NAD, and with NAD and
`4-MP. Also shown are the total protein (Coomassie) stains. While a major protein
`band corresponding to HSA was readily detected in all incubations with Coomassie
`staining, only those incubations with NAD and without 4-MP were observed as
`immunoreactive.
`
`3.5. Metabolism of dihydro-abacavir by human aa ADH isozyme
`
`The dihydro analog of abacavir was incubated with both aa and g2g2 isoforms
`using the same conditions as for abacavir. A comparison of the non-extractable
`
`14 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`149
`
`Fig. 10. Immunoblots using the rabbit serum anti-abacavir antibody and Coomassie-stained SDS-PAGE
`gels for abacavir incubations with aa and g2g2 isozymes in the presence of human serum albumin. ABC:
`Abacavir. MWS: Molecular weight standards
`
`Fig. 11. A comparison of non-extractable 14C residue levels and extent of metabolism, for abacavir and
`dihydro-abacavir with the aa human ADH isozyme. Percent metabolism was determined from
`radiochemical profiling, and refers to the percent conversion to total metabolites. Incubations were also
`analyzed by LC/MS. Two isomeric carboxylic acid metabolites were identified as the major products with
`dihydro-abacavir.
`
`15 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`150
`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`residue levels and extent of metabolism by ADH aa of abacavir and the dihydro
`analog are shown in Fig. 11. Despite the fact that the dihydro analog was
`metabolized more extensively than abacavir, the level of non-extractable residue
`was substantially reduced.
`
`4. Discussion
`
`The metabolism of primary alcohols to carboxylic acids is most commonly
`mediated by cytosolic ADH or microsomal CYP450 [19,20]. No metabolites could be
`detected radiochemically in incubations of abacavir with pooled human liver
`microsomes, but good conversion to the acid metabolite was observed with human
`liver cytosol. An aldehyde would be the initial product of such metabolism, and the
`formation of the acid directly in cytosol might involve a role for aldehyde
`dehydrogenase. However, ADH is well known to be capable of oxidizing aldehydes
`further to carboxylic acids via gem diol formation, and /or dismutation [21,22].
`Incubations of 14C-abacavir with cytosol also lead to NAD dependent non-
`extractable protein residues. The pH and 4-MP sensitivity for both metabolism and
`residue formation implicate ADH in both processes.
`The ADH system in humans has been extensively reviewed [23,24] (Table 1). While
`methodology for phenotyping with expressed human CYP450s is well established,
`previous reports of such studies with expressed human ADH isozymes are very
`limited [25]. The high in vitro pH optimums for oxidation of alcohols by ADH
`enzymes can pose a problem for these studies if pH sensitive substrates or
`intermediates are involved. Minimal oxidation of alcohols may be observed at pH
`7.4, and under the conditions used here, low activity with ethanol or pentanol as
`substrates was seen. As a balance between optimizing turnover and minimizing base
`catalyzed side reactions, ADH activity can be examined at an intermediate pH [26 /
`28]. Metabolic turnover of ethanol or pentanol was significantly enhanced at pH 8.8
`(Table 1). When abacavir was incubated with the ADH isozymes at pH 8.8,
`metabolism was observed only with the aa and g2g2 isoforms. For the aldehyde
`intermediate in abacavir oxidation, the enolizable nature of the chiral center adjacent
`to the aldehyde functionality would suggest the potential for pH dependence on
`isomerization processes at this center. To ascertain if isomers of 2269W93 were
`arising solely due to pH effects, some studies were also conducted at lower pH. As
`shown in Fig. 2, in cytosol, qualitatively similar profiles are observed at pH 7.4 and
`8.8. In addition, non-extractable, cofactor-dependent residues could also be
`measured at pH 7.4. When abacavir was incubated with the aa isoform at pH 7.4,
`only very low levels of metabolism could be detected, but a low level of NAD-
`dependent non-extractable protein residue was observed (Fig. 6). At pH 7.8, isomeric
`acid metabolites are clearly detected in incubations with the aa enzyme (Fig. 4).
`Taken collectively, these results suggest that the isomerization processes are not
`solely artifactual due to pH, but are directly related to enzyme activity. The
`observation of two isomers of 2269W93 as minor metabolites in human urine
`following dosing with abacavir supports this proposal.
`
`16 of 20
`
`PENN EX. 2039
`CFAD V. UPENN
`IPR2015-01836
`
`

`
`J.S. Walsh et al. / Chemico-Biological Interactions 142 (2002) 135 /154
`
`151
`
`Radiochemical profiling and LC/MS analysis of the ADH incubations showed
`that the major pathways of metabolism for abacavir with the two isozymes are
`different. The aa isozyme generated a peak of the same retention time as 2269W93,
`and two apparent isomers of this. The formation of acid metabolites in these
`incubations indicates the capability of the aa isozyme to oxidize the aldehyde
`intermediate,
`in addition to abacavir itself. While this does not rule out the
`possibility of aldehyde dehydr