Identification of Bioactivating Enzymes Involved in the Hydrolysis of Laninamivir Octanoate, a Long-Acting Neuraminidase Inhibitor, in Human Pulmonary Tissue

Laninamivir octanoate (LO) is an octanoyl ester prodrug of the neuraminidase inhibitor laninamivir. After inhaled administration, LO exhibits clinical efficacy for both treatment and prophylaxis of influenza virus infection, resulting from hydrolytic bioactivation into its pharmacologically active metabolite laninamivir in the pulmonary tissue. In this study, we focused on the identification of LO-hydrolyzing enzymes from human pulmonary tissue extract using proteomic correlation profiling—a technology integration of traditional biochemistry and proteomics. In a single elution step by gel-filtration chromatography, LO-hydrolyzing activity was separated into two distinct peaks, designated as peak I and peak II. By mass spectrometry, 1160 and 1003 proteins were identified and quantitated for peak I and peak II, respectively, and enzyme candidates were ranked based on the correlation coefficient between the enzyme activity and the proteomic profiles. Among proteins with a high correlation value, S-formylglutathione hydrolase (esterase D; ESD) and acyl-protein thioesterase 1 (APT1) were selected as the most likely candidates for peak I and peak II, respectively, which was confirmed by LO-hydrolyzing activity of recombinant proteins. In the case of peak II, LO-hydrolyzing activity was completely inhibited by treatment with a specific APT1 inhibitor, palmostatin B. Moreover, immunohistochemical analysis revealed that both enzymes were mainly localized in the pulmonary epithelia, a primary site of influenza virus infection. These findings demonstrate that ESD and APT1 are key enzymes responsible for the bioactivation of LO in human pulmonary tissue.


Introduction
Laninamivir octanoate (LO), an octanoyl ester prodrug of the neuraminidase inhibitor laninamivir, demonstrates a long-lasting antiviral effect in contrast to oseltamivir and zanamivir, and is currently used in clinical practice in Japan for treatment and prophylaxis of influenza virus infection. In confirmatory clinical trials, a single inhalation of LO demonstrated a therapeutic efficacy in both adult and pediatric patients (Sugaya et al. 2010, Watanabe et al. 2010, Watanabe et al. 2013). In addition, LO was recently reported to be efficacious in the post-exposure prophylaxis of influenza in household contacts (Kashiwagi et al. 2013).
Hydrolytic activation of LO into the active form laninamivir in the pulmonary tissue is a critical factor for exerting its in vivo pharmacological effect. After a single intranasal/intratracheal administration in mice and rats, LO was efficiently hydrolyzed to laninamivir, and then it was highly retained in the pulmonary tissue (Koyama et al. 2009, Koyama et al. 2010; with a sufficient concentration in the epithelial lining fluids (ELF), a possible site of action of neuraminidase inhibitors (Koyama et al. 2013). In humans as well, after a single inhalation of LO (40 mg), laninamivir was generated and highly maintained in the ELF (Ishizuka et al. 2012), exceeding the in vitro 50% inhibitory concentrations for influenza viral neuraminidases over 10 days (Yamashita et al. 2009). A slow elimination of laninamivir was also observed in the systemic circulation, with an elimination half-life (t 1/2 ) of approximately 3 days (Ishizuka et al. 2009, Yoshiba et al. 2011). These favorable pharmacokinetic characteristics are considered to result in the long-lasting effect.

DMD #57620
6 In drug development, the prodrug approach has been widely used as a strategy for improving the physicochemical, biopharmaceutical or pharmacokinetic properties of pharmacologically active agents, and approximately 10% of the drugs approved worldwide are reported to be classified as prodrugs (Rautio et al. 2008, Zawilska et al. 2013). Prodrugs require the bioactivation into their active forms by metabolizing enzymes in order to be therapeutically effective. On the other hand, the prodrug-activating enzymes may have a potential risk to influence interindividual variability in drug exposure and response, and possibly to induce drug-drug interactions. Therefore, it is very important to identify the prodrug-activating enzymes.
For the molecular identification of endogenous enzymes from biological samples, biochemical purification has been used for decades as a general methodology, and a series of sequential separation by different chromatographic methods (ion-exchange, affinity, gel filtration, etc.) is usually required to achieve the purification. On the other hand, we have recently extended an advanced methodology named proteomic correlation profiling to identify drug metabolizing enzymes (Sakurai et al. 2013), whose basic concept was previously reported by Kubota et al (2009). In this methodology, the biological material is fractionated by column chromatography, followed by the calculation of each protein's correlation coefficient between the enzyme activity and proteomic profile of the fractions.
Subsequently, possible enzyme candidates are chosen among proteins with a high correlation value, based on the fundamental assumption that protein quantity correlates with protein activity. This streamlined method enables us to require fewer purification steps with minimal starting material, and also to resolve the challenges associated with physicochemical stability and solubilization, as compared to the conventional biochemical DMD #57620 7 purification. Therefore, the proteomic correlation profiling is considered to be a powerful and efficient tool for enzyme identification.
In the present study, from the human pulmonary tissue extract, we identified S-formylglutathione hydrolase (also known as esterase D; ESD) and acyl-protein thioesterase 1 (APT1) as most likely candidates for LO-hydrolyzing enzymes, using the proteomic correlation profiling. We also performed further analysis to confirm the enzyme identification and to investigate the enzyme localization in the pulmonary tissue. We believe that this study can provide fundamental information on the LO bioactivation process in the target site. This article has not been copyedited and formatted. The final version may differ from this version. DMD Fast Forward. Published on March 28, 2014as DOI: 10.1124 at ASPET Journals on July 27, 2023 dmd.aspetjournals.org 2004) or AQUA (Gerber et al. 2003). Two missed cleavages were allowed, along with carbamidomethylation of cysteine as a fixed modification; variable modifications were oxidation of methionine and acetylation of N-term amino acid of the protein. Mass tolerance for precursor ions was 7 ppm, mass tolerance for fragment ions was 0.5 Da, and false discovery rates at peptide and protein levels were less than 0.01. The proteins identified by more than one single peptide were used for further correlation analysis. processed on the tissue sections in a single step using Dako PT Link (EnVision FLEX TRS Low pH), where the optimal incubation was set at 97°C for 20 min. Endogenous peroxidase activity was quenched with peroxidase-blocking solution (Dako) for 5 min, and nonspecific antibody binding was blocked with protein block, serum-free (Dako) for 30 min. Then, anti-ESD antibody or anti-APT1 antibody (2 μ g/ml) was applied as a primary antibody on the sections and incubated for 45 min, and isotype antibody (2 μ g/ml) was used as a negative control. EnVision+System-HRP Labelled Polymer Anti-Rabbit was reacted as a secondary antibody for 30 min, and antibody complex was visualized after the addition of 3,3'-diaminobenzidine tetrahydrochloride (DAB)+, liquid (5 min × 2; Dako).
The section was counterstained with hematoxylin.

Results
Enzymatic LO hydrolysis in human pulmonary subcellular fractions. LO-hydrolyzing activity in human pulmonary S9, cytosol, and microsomes is shown in Fig. 2A. At 10 μM LO, activity was observed in all of the subcellular fractions tested, with almost the same levels per mg protein basis (1.2-1.9 pmol/min/mg). Furthermore, for the inhibitory effect of typical esterase inhibitors, LO-hydrolyzing activity in pulmonary S9 was strongly inhibited by a potent serine esterase inhibitor DFP at 0.1 mM or more. DTNB, known as an inhibitor of SH-containing esterase, exhibited weaker inhibitory effects than DFP, whereas much weaker inhibition was observed in the treatment with BNPP (carboxylesterase inhibitor). Eserine (cholinesterase inhibitor) showed no enzymatic inhibition (Fig. 2B).

Selection of possible candidate proteins for LO-hydrolyzing enzymes.
A schematic diagram for this study is illustrated in Fig. 3. As shown in Fig. 2A, LO-hydrolyzing activity in pulmonary S9 was substantially derived not only from the soluble fraction (cytosol) but also from the insoluble fraction (microsomes). Therefore, non-ionic detergent NP-40 was added to the S9 fraction in order to solubilize membrane proteins without altering biological activity, and the centrifuged soluble fraction (overall recovery: 73.8%) was subjected to a single-step gel filtration chromatography. From enzyme assay for the chromatographic fractions collected, LO-hydrolyzing activity was separated into two distinct peaks (Fig. 4), which were designated as peak I (fraction No. 26-32) and peak II (fraction No. 33-39), respectively. The contribution rate to total activity was estimated to This article has not been copyedited and formatted. The final version may differ from this version. be approximately 40% and 30% for peak I and peak II, respectively, based on the calculation of each peak area (%) of LO-hydrolyzing activity in the gel filtration fractions.
Then, in order to seek proteins that showed a high correlation with LO-hydrolyzing activity, all proteins in the fractions were identified and quantified by LC-MS/MS analysis.
Theoretically protein of a higher correlation coefficient has higher probability of a responsible protein to the activity profile and our previous experiences (Kubota et al. 2009, Sakurai et al. 2013, andunpublished results) are in agreement with this concept. In total, 2381 proteins were identified, however there was no single protein that had the bimodal chromatographic peaks highly correlated with the enzyme activity profile (data not shown).
Therefore, it was speculated that the active peaks consisted of two individual enzymes, which were independently derived from peak I and peak II. Possible candidate proteins were listed for each peak based on the correlation coefficient with the enzyme activity profile. In respect to peak I, there were 1160 proteins identified, and the top 30 most correlated proteins were listed in Table 1. In a similar manner, with regard to peak II, 1003 proteins were identified and the top 30 proteins were also placed in descending order of correlation coefficient, as shown in Table 2. Among the proteins that were annotated as "hydrolase", ESD (R 2 =0.9432) and APT1 (R 2 =0.9765) were finally selected as the most possible candidates for the peak I-and peak II-derived proteins, respectively. ESD and APT1 proteins were clearly detected in the peak I and peak II fractions, respectively, by Western blot analysis (data not shown). their recombinant proteins (rESD and rAPT1) were subjected to the determination of LO-hydrolyzing activity. As shown in Table 3, the enzyme activities in rESD and rAPT1
On the other hand, BSA (negative control) had a negligible activity.

Inhibitory effect of palmostatin B on chromatographic LO-hydrolyzing activity.
Palmostatin B is known as a specific APT1 inhibitor (Dekker et al. 2010, Rusch et al. 2011), although no specific inhibitor has been reported for ESD. LO-hydrolyzing activity profiles were measured for the fractions separated by gel filtration chromatography, in the absence or presence of a specific APT1 inhibitor, palmostatin B (Fig. 5). In the absence of palmostatin B, there were two active peaks, peak I and peak II, on the chromatogram. On the other hand, in the presence of palmostatin B, peak II-derived activity was mostly inhibited without any significant effect in peak I-derived activity, supporting APT1 as a major enzyme in peak II.
Immunohistochemical localization of ESD and APT1 in human pulmonary tissue. Fig.   6 shows representative immunohistochemical images of ESD and APT1 in human pulmonary tissue sections stained with anti-ESD antibody (A) and anti-APT1 antibody (B), respectively, using isotype antibody (C) as a negative control. As shown in Fig. 6A, ESD was highly expressed in the pulmonary epithelia with a strong dot-like or granular stain being localized in the cellular cytoplasm on the side of the luminal membrane. No obvious staining of ESD was observed in other tissue components. In Fig. 6B, APT1 was also This article has not been copyedited and formatted. The final version may differ from this version. highly expressed in the pulmonary epithelia, with overall cellular staining including the cellular cytoplasm. Positive staining of APT1 was observed in the alveolar macrophages as well. In the case of the isotype antibody (Fig. 6C), there was little or no cross-reactivity with cell surface antigens on the tissue section.
This article has not been copyedited and formatted. The final version may differ from this version.

Discussion
In this study, we have successfully identified two hydrolyzing enzymes, ESD and APT1, which would be primarily involved in the bioactivation of the anti-influenza prodrug LO in human pulmonary tissue, using a versatile methodology named proteomic The proteomic correlation profiling has a feature to facilitate rapid identification of the responsible protein from crude complex lysate with fewer purification steps using a smaller starting material, as compared to the conventional methodology. Actually, in single-step gel filtration chromatography using only approximately 2.5 mg of the human pulmonary S9 fraction, we could identify two highly possible candidates of LO hydrolases, ESD for peak I and APT1 for peak II (Fig. 4). has an O-octanoyl ester linkage on its sialic acid-like structure which is very similar to that of Neu5, 9Ac2. This structural feature supports the reasonability of LO hydrolysis by ESD.
In our experiment, recombinant ESD had distinct LO-hydrolyzing activity (Table 3), suggesting a responsible enzyme for LO hydrolysis. In addition, it was consistent with ESD as one of the LO hydrolases that the LO-hydrolyzing activity in pulmonary S9 was strongly inhibited by DFP, a serine hydrolase inhibitor (Fig. 2).

Similarly to ESD, sialate O-acetylesterase (SIAE) is known as an enzyme
hydrolyzing O-acetyl ester linkage of Neu5, 9Ac2 (Schauer 2000, Angata et al. 2002. Moreover, ESD has been reported to show significant carboxylesterase activity against the following model substrates: α-naphthyl acetate and p-nitrophenyl acetate (Degrassi et al. 1999). A partial contribution of carboxylesterase on the LO hydrolysis was also supposed from the data that carboxylesterase inhibitor BNPP showed a weak inhibition against LO-hydrolyzing activity with 42.3% inhibition at the highest concentration (1 mM; Fig. 2).
Considering these situations, we additionally investigated the potential of LO-hydrolyzing activities of SIAE and CES1, using their recombinant proteins. However, there was little or no LO-hydrolyzing activity in either protein (data not shown), indicating that these two enzymes were not responsible for the LO bioactivation although they were identified as candidate proteins for peak I and placed in a higher rank than ESD (Table 1). No LO-hydrolyzing activity was clearly observed in LTA4H as well, using its recombinant protein (data not shown). Nevertheless, the involvement of other enzyme(s) cannot be completely ruled out, since some proteins with hydrolase annotation still remain as candidate proteins for peak I, such as PGAM1, RCL, DDX39B, and IAH1. However, from enzyme activity using recombinant ESD (Table 3), it is highly plausible that ESD is at This article has not been copyedited and formatted. The final version may differ from this version.  et al. 1999). Furthermore, APT1 was also reported as an oxyesterase that cleaves the octanoyl group of ghrelin (Shanado et al. 2004, Satou et al. 2010. From the similarity of chemical structure with ghrelin, APT1 was hypothesized to be the most promising candidate for peak II, and this hypothesis was supported by LO-hydrolyzing activity of recombinant APT1 (Table 3). Moreover, as shown in Fig. 5, LO-hydrolyzing activity of peak II was almost inhibited by palmostatin B, a specific APT1 inhibitor (Dekker et al. 2010, Rusch et al. 2011. Also, in the same way as ESD, a strong enzyme inhibition by DFP (Fig. 2) was consistent with the fact that APT1 is categorized in the serine hydrolase class of enzymes. These results demonstrated that APT1 consisted mostly of active peak II.
Immunohistochemical analysis showed that ESD was highly localized in the cytoplasm and/or cytoplasmic vesicles of pulmonary epithelia (Fig. 6A), whereas APT1 was highly expressed in the overall pulmonary epithelia and alveolar macrophages (Fig.   6B). According to our previous reports, active metabolite laninamivir was highly maintained in the ELF and alveolar macrophages after inhaled administration of LO in healthy volunteers (Ishizuka et al. 2012). In addition, in mice intranasally administered LO, This article has not been copyedited and formatted. The final version may differ from this version. the laninamivir was also deposited in the pulmonary tissue, especially in the epithelia.
These results indicated that LO could be hydrolyzed into laninamivir in the pulmonary epithelia and/or alveolar macrophages, in which ESD and APT1 were mainly localized.
Prodrugs require bioactivation by the metabolizing enzymes in order to demonstrate its therapeutic effect. Genetic polymorphism in the prodrug-activating enzymes can cause a crucial impact on the interindividual variability in the drug exposure and response, as reported in some drugs, such as clopidogrel (Shuldiner et al. 2009), codeine (VanderVaart et al. 2011, and tramadol (Gan et al. 2007). Regarding the LO-hydrolyzing enzymes, the genetic polymorphism has been identified for ESD in a previous report (Ebeli-Struijk et al. 1976), in which the ESD*2 allele was commonly observed with a different gene frequency depending on ethnic groups. Additionally, it has been reported that the enzyme activity associated with ESD*2 was estimated to be 40% lower than that associated with ESD*1, using a typical substrate 4-methylumbelliferyl acetate (Horai and Matsunaga 1984).
However, in the case of LO, this genetic factor would be unlikely to make a huge influence on the interindividual variability, since at least the two enzymes (ESD and APT1) were involved in the LO bioactivation and the contribution of multiple enzymes in the same reaction is considered to dilute the potential impact on the interindividual variability. Further assessment on this issue should be performed more precisely in the future.
In conclusion, this study demonstrated that both ESD and APT1 would be key  After inhaled administration of LO in humans, the active metabolite laninamivir is generated by an enzymatic hydrolysis of the octanoyl ester moiety. When dissolved in water, LO is equilibrated at 9:1 (3-acyl form:2-acyl form) and therefore defined as a mixture of the 3-acyl form (major) and the 2-acyl form (minor).

Figure 2. Enzymatic LO hydrolysis in human pulmonary subcellular fractions.
A, shows LO-hydrolyzing activity in human pulmonary subcellular fractions (S9, cytosol, and microsomes). Each data represents the mean±S.E. in triplicate determinations. B, shows the inhibitory effect of typical esterase inhibitors on LO-hydrolyzing activity in human pulmonary S9. Each data represents the mean of triplicate determinations.    Human pulmonary tissue sections were stained with anti-ESD antibody (A), anti-APT1 antibody (B), and isotype control (C). After treatment with HPR-labeled anti-rabbit IgG, the antibody complex was visualized as a dark brown precipitate by using DAB as a substrate.
This article has not been copyedited and formatted. The final version may differ from this version.  Among the 1003 proteins identified, the top 30 candidates with high correlation coefficient between relative intensity of the identified proteins and LO-hydrolyzing active peak II (fraction No. 33-39) on gel filtration chromatogram are listed above. The asterisk represents hydrolase domain which was annotated by Uniprot. Identified peptides for each protein are listed in Supplemental Table 2.
This article has not been copyedited and formatted. The final version may differ from this version.