A Drug Delivery Strategy: Binding Enkephalin to Asialoglycoprotein Receptor by Enzymatic Galactosylation

Glycosylation of biopharmaceuticals can mediate cell specific delivery by targeting carbohydrate receptors. Additionally, glycosylation can improve the physico-chemical (drug-like) properties of peptide based drug candidates. The main purpose of this study was to examine if glycosylation of the peptide enkephalin could facilitate its binding to the carbohydrate receptor, asialoglycoprotein. Firstly, we described the one-pot enzymatic galactosylation of lactose modified enkephalin in the presence of uridine-5′-diphosphogalactose 4-epimerase and lipopolysaccharyl α-1,4-galactosyltransferase. Stability experiments using human plasma and Caco-2 cell homogenates showed that glycosylation considerably improved the stability of enkephalin (at least 60% remained stable after a 2 hr incubation at 37°C). In vitro permeability experiments using Caco-2 cells revealed that the permeability of mono- and trisaccharide conjugated enkephalins was 14 and 28 times higher, respectively, than that of enkephalin alone (Papp 3.1×10−8 cm/s). By the methods of surface plasmon resonance and molecular modeling, we demonstrated that the enzymatic glycosylation of enkephalin enabled binding the asialoglycoprotein receptor. The addition of a trisaccharide moiety to enkephalin improved the binding of enkephalin to the asialoglycoprotein receptor two fold (KD = 91 µM). The docking scores from molecular modeling showed that the binding modes and affinities of the glycosylated enkephalin derivatives to the asialoglycoprotein receptor complemented the results from the surface plasmon resonance experiments.


Introduction
The clinical use of peptide-based therapeutics is mainly limited by their poor membrane permeability and stability in biological systems. Different strategies have been used to improve the physico-chemical characteristics of peptides to make them more amenable as therapeutics [1,2]. These include chemical modifications such as cyclization and peptide bond reduction [3], use of unnatural amino acids [4,5], co-administration with penetration enhancers such as bile salts and surfactants [6], and conjugation with bacterial and viral proteins [7], lipids [8,9], and carbohydrates [10][11][12].
The coupling of carbohydrate moieties such as glucose or lactose to peptides improved their solubility, stability and permeability [13][14][15][16][17]. Additionally, glycosylation enhanced bioavailability and passage through the blood brain barrier [18][19][20]. The addition of carbohydrates can also enable cellular uptake via sugar transporters [11,21] and may be used to target specific cells via carbohydrate receptors [17,22]. The transport of glucose and galactose across the intestinal membranes occurs primarily via the sodium-glucose co-transporter 1 (SGLT1) or the facilitated diffusion glucose transporter protein 2 (GLUT2) [23]. Although these transporters are primarily responsible for the transport of sugars, some studies have implicated them in the movement of other carbohydrate-modified compounds such as glycopeptides [11,19,21].
Several studies used carbohydrate receptor binding to target drug molecules, DNA and peptides to specific cells types [17,22,24,25]. In this study, we focused on the asialoglycoprotein receptor (ASGPR) located on hepatic cells [26] and in discrete areas of the brain, such as the cerebellum and brain stem [27]. ASGPR belongs to the C-type lectin family [28] and is targeted by glycoproteins and liposaccharides that contain terminal galactose or N-acetylgalactosamine, for example asialoorosomucoid [29] and gonococcal lipooligosaccharide [30,31].
Leu-enkephalin (Tyr-Gly-Gly-Phe-Leu) is an opioid peptide with known pain-regulating activity and the potential to be used as a therapeutic when delivered to the central nervous system. Several previous attempts have been made to improve the delivery of enkephalin to the central nervous system, including lipidation [8,9] and glycosylation [9,32]. Various glycosyltransferases were found to be an efficient tool for the addition of carbohydrate moieties to biologically active peptides [33][34][35]. We previously galactosylated the enkephalin peptide chemically by solid phase peptide synthesis and enzymatically using galactosyltransferase as described by Simerska et al. [35].
Preparative RP-HPLC was conducted on a Waters Delta 600 system (Milford, USA) at a flow rate of 20 mL/min and detection at 230 nm using PicoLog software. A Vydac C18 column (10 mm, 22 mm6250 mm; Grace Davison Discovery Sciences, Deerfield, USA) was used for all purifications.
Electrospray ionization mass spectrometry (ES-MS) and liquid chromatography mass spectrometry (LC-MS) were conducted on a Perkin-Elmer-Sciex API 3000 mass spectrometer (Applied Biosystems/MDS Sciex, Toronto, Canada). Solvents used for ES-MS and LC-MS consisted of solvent A: 0.1% acetic acid

Chemo-enzymatic Synthesis of Enkephalin and its Glycopeptides
Enkephalin and carbohydrate-modified enkephalins, Gal-Enk, Lac-Enk and Gal-Lac-Enk, were prepared according to Simerska et al. [35]. Peptides and glycopeptides were assembled using standard solid phase peptide synthesis on Rink Amide-MBHA resin using Fmoc chemistry protocols. Crude peptide and glycopeptides were purified by preparative RP-HPLC, a gradient of 0-100% solvent B over 60 min was used for enkephalin and a gradient of 20-35% solvent B over 60 min was used for the glycosylated enkephalins. Fractions that contained pure peptide/ glycopeptide were analyzed using analytical RP-HPLC and ES-MS and corresponded with those previously published [35].
The galactosyltransferase LgtC was recombinantly expressed in Escherichia coli AD202 and purified using TALON metal affinity resin [35]. The purified protein was dialyzed into 20 mM Na 2 HPO 4 buffer at pH 7.5, concentrated to 2 mg/mL and stored

Plasma Stability Assay
Blood sample collection from a healthy and consenting human adult for this study was approved by and conducted according to the guidelines set by The University of Queensland Medical Research Ethics Committee (approval number: 2009000661). Written consent was obtained prior to the collection of blood samples. Plasma was isolated from fresh-heparin treated blood by centrifugation at 6606g for 15 min. A 1 mg/mL solution of each test compound (enkephalin, Gal-Enk, Lac-Enk, and Gal-Lac-Enk) was prepared in phosphate buffered saline (PBS). Compounds (300 mL) were mixed with 300 mL of the plasma, pre-warmed at 37uC and incubated at 37uC. Samples (50 mL) were collected at predetermined time points (0, 5, 10, 20, 30, 40, 60, 90 and 120 min), and added into 75 mL ACN to precipitate the plasma proteins. Samples were then centrifuged at 73006g for 10 min. The resultant supernatant (30 mL) was analyzed by analytical RP-HPLC for the amount of test compound remaining in the solution using an elution gradient of 20-35% (enkephalin) or 0-70% solvent B (glycosylated enkephalins) over 30 min on a C18 Vydac column.

Caco-2 Cell Stability Assay
Caco-2 cells cultured for 21 to 28 days were washed 3 times with 5 mL of 0.02% ethylenediaminetetraacetic acid solution. They were detached using a cell scraper and re-suspended in 4 mL of Hank's buffered salt solution (HBSS) buffered with 25 mM HEPES to pH 7.4 (HBSS-HEPES). Cells were lysed by sonication on a Branson Sonifier, four times for 5 sec, with an output of 2 and 20% duty cycle. The cell lysate was then centrifuged at 4306g for 5 minutes to remove cell debris. The protein content of the supernatant was determined using the BioRad protein assay (Gibbons et al. 1990) and adjusted to 0.5-0.8 mg/mL with HBSS-HEPES. The test compounds were solubilized in HBSS-HEPES buffer and 100 mL samples (100 mM) were mixed with 100 mL of the Caco-2 cell homogenate in a TPP 96-well plate and incubated at 37uC with shaking. Each compound was tested in triplicate at 5, 10, 20, 30, 40, 50, 60 and 90 minutes. At each time point, 10 mL of sample was mixed with ACN/water solution containing 5.5% TFA to stop enzymatic degradation. Collected samples were analyzed using LC-MS in positive ion electrospray mode and selective ion monitoring. A Phenomenex Luna C18 column with a linear gradient of 0-80% solvent B was used for enkephalin and a linear gradient of 0-70% solvent B for all other compounds over 5.5 min at a 0.5 mL/min flow rate. A standard curve was generated for each compound and used to calculate the compound's concentration.

Caco-2 Cell Permeability Assay
A cell suspension of Caco-2 cells in the cell culture medium was adjusted to a cell density of 1610 6 cells/mL before being aliquoted (100 mL) into Transwell polycarbonate cell culture inserts in a 24well plate, while 600 mL of media was added to the basolateral chamber. The culture medium was changed every other day for 21-28 days. TEER values were measured before and after the assay, to determine the integrity of tight junctions of the cell monolayers. TEER values of the monolayers were between 730 -975 Vcm 2 . The integrity of the monolayers was further monitored by measuring the permeability of radiolabelled [ 14 C]-D-mannitol (0.09 mCi/mL in 90% ethanol in water) by adding a 100 mL solution of 1.80 mCi, 32.73 nmol/4 mL solution in HBSS-HEPES buffer to the apical chamber of three wells. Cell  The apparent permeability (P app , cm/s) of each compound was calculated using the following equation: Where dC/dt is the steady-state rate of change in the test item concentration (M/s) or radiochemical concentration (dpm mL/s) in the receiver chamber, Vr is the volume of the receiver chamber (mL), A is the surface area of the cell monolayers and C0 is the initial concentration in the donor chamber (M or dpm/mL). Permeability experiments (as described above) were conducted to investigate whether the compounds used the sugar transporters SGLT1 or GLUT2 for their transport across cell membranes. Samples containing 200 mM test compound and 100 mM phlorizin (inhibitor for SGLT1) or 100 mM phloretin (inhibitor for GLUT2) in HBSS-HEPES buffer were added to the apical chamber. Solution (0.4 mL) was collected from the basolateral chambers and replaced with the same volume of buffer after 30, 90, 120, and 150 min incubations. P app values for each compound were then calculated as above.
The average P app values (6S.D.) from three replicates is presented. Unpaired t-tests (p,0.05) were performed using Prism software (Version 6, Graphpad Software Inc., La Jolla, USA) to determine whether the P app values between different compounds were statistically significant.

SPR Analysis of ASGPR Binding
The affinities of enkephalin derivatives for ASGPR were tested using SPR methodology on a Biacore T100 at 25uC. Recombinant human ASGPR was immobilized onto a series S sensor chip CM5 using the NHS capture kit, where ASGPR was covalently linked to the surface of the chip via free amine groups. Compounds were tested at 10-1000 mM concentrations using multi-cycle kinetics with at least three experiments performed for each interaction. Single cycle kinetics was applied to optimize concentrations prior to completion of multi-cycle kinetics as described by Dumont et al. [36]. D-Galactose and D-lactose at the same concentrations as the test compound were used as positive controls. The running buffer for all SPR experiments was 20 mM HEPES, pH 7.4, 150 mM NaCl, 5 mM CaCl 2 , and 0.005% TWEEN 20.

Molecular Modeling Studies Using MolDock
In silico docking was used to investigate the binding affinity of Dgalactose, D-lactose and enkephalin derivatives to ASGPR based on energy minimization calculated by Molegro Virtual Docker software [37]. The crystal structure of ASGPR (PDB code: 1DV8, [38]) was used as the target in this study. Binding affinities of enkephalin, Gal-Enk, Lac-Enk, and Gal-Lac-Enk for ASGPR were displayed by docking scores [39].

Chemo-enzymatic Synthesis of Enkephalin and its Glycosylated Derivatives
The chemo-enzymatic synthesis of Gal-Lac-Enk was previously conducted using the lipopolysaccharyl galactosyltransferase LgtC with UDP-Gal as a donor [35]. However, the high cost of UDP-Gal limited the large-scale use of this reaction (despite its high efficiency). On the other hand, UDP-Glc is available for a lower cost and is readily converted to UDP-Gal by galactose epimerase (Figure 1). Previous studies successfully used galactose epimerase for the synthesis of UDP-Gal from UDP-Glc for other one-pot reactions [40][41][42]. A similar one-pot synthesis of Gal-Lac-Enk was attempted with UDP-Glc and galactose epimerase to synthesize UDP-Gal, which in turn acted as a donor for the LgtC-catalyzed transfer of a galactose unit to Lac-Enk to produce Gal-Lac-Enk ( Figure 1). Reaction progress was monitored by RP-HPLC and ES-MS (Figure 2), which showed product formation in the reaction catalyzed by galactose epimerase and LgtC after 48 hrs of incubation at 37uC. The control reaction using UDP-Glc and LgtC only (without galactose epimerase) did not show any product formation.

Stability of Carbohydrate-modified Enkephalins
Enkephalin is susceptible to degradation by enzymes such as aminopeptidases, carboxypeptidases and endopeptidases, leading to rapid degradation in vitro [43,44]. In this study, we used human plasma and Caco-2 cells to test the stability of the glycosylated enkephalins. Plasma stability experiments were conducted to determine the stability of the compounds upon entry into the circulatory system. Compounds with low plasma stability generally showed rapid clearance and consequently low efficacy in vivo [45]. Caco-2 cells are derived from human carcinoma cells and have been used to assess enzymatic stability upon oral or systemic delivery [46]. Our results from plasma and Caco-2 stability experiments showed that enkephalin had a half-life of 2.1 min and 7.1 min, respectively. These values are similar to previously reported values for enkephalin [47,48]. In contrast, the glycosylated enkephalins were considerably more stable. Around 60, 70 and 100% of Gal-Enk, Lac-Enk and Gal-Lac-Enk, respectively, remained in human plasma even after incubation at 37uC for 120 min ( Figure 3A). The stability profiles of the glycopeptides (Gal-Enk and Lac-Enk) plateaued after a 40 min incubation in human plasma. This is most likely to have occurred because the test compounds saturated the peptidases found in plasma. The saturation effect was observed for the glycosylated enkephalins Table 3. Calculated binding affinities using MolDock for enkephalin and its derivatives with ASGPR.

Enkephalin Positive
Gal-Enk Positive

D-Galactose 234
D-Lactose 245 # The binding affinity was calculated by energy minimization and given a MolDock score, which was derived from the piecewise linear potential scoring function [53]. doi:10.1371/journal.pone.0095024.t003 even when a lower concentration of test compound (0.3 mg/mL) was used (data not shown). We were therefore unable to obtain an accurate measure of the half-life of glycosylated enkephalins in plasma. In Caco-2 cell homogenates, the glycosylated compounds remained stable throughout the course of the experiment ( Figure 3B). The increased stability observed for the glycosylated peptides over the parent peptide can be attributed to increased protection of the peptide from enzymes such as aminopeptidase. It is likely that the presence of the carbohydrate moiety sterically hindered the peptidases from accessing the enkephalin peptide. Peptides such as endomorphin [11], luteinizing hormone-releasing hormone [16] and enkephalin modified at the N-terminus [48], also showed a greatly increased stability in comparison to the parent peptides in vitro. For example, Wang et al. [48] observed that the modification of the N-terminus of enkephalin with a lipidic moiety increased the half-life of the parent peptide from 6.7 min to ,193 hrs.

Permeability of Carbohydrate-modified Enkephalins
The permeability of enkephalin and glycosylated enkephalins was determined in vitro using Caco-2 cells. The apparent permeability of the negative control 14 C mannitol was 4.4610 29 (63.86) cm/s and the apparent permeability for the positive control propanolol was 1.34610 24 (60.009) cm/s. Enkephalin displayed an apparent permeability of 3.1610 28 (61.1) cm/s. The presence of a mono-or a tri-saccharide moiety in the glycosylated enkephalin structure increased the apparent permeability of enkephalin by 28 and 14 fold, respectively (p, 0.05, Figure 4A, Table 1). The permeability of Lac-Enk (P app 1.72610 26 (61.24) cm/s) was also higher than that of enkephalin alone. However, this increase was not statistically significant due to the high standard deviation observed for the Lac-Enk sample ( Figure 4A). Although all the glycosylated enkephalins were more permeable than the parent peptide, we did not observe any significant difference between them. Wong et al. [12] reported that enkephalin was degraded on the apical side of the cells in Caco-2 permeability experiments and consequently reduced the apparent permeability of the compound. The increased permeability of glycosylated enkephalins may result from their increased metabolic stability ( Table 1).
The higher permeability observed for the glycosylated enkephalins could also be due to improved transport across the Caco-2 cell membrane. The compounds may be transported into cells via passive diffusion, transporter molecules or by transcytosis. Some reports showed that small glycosylated peptides moved across the cell membrane via sugar transporters [21]. Transport by absorptive transcytosis was also suggested for glycosylated opioid peptides such as enkephalin [18,49]. The apparent permeability of these compounds was measured in the presence of SGLT1 and GLUT2 sugar transport inhibitors; phlorizin and phloretin, respectively [23,[50][51][52], to test if the carbohydrate-derived enkephalins were able to utilize sugar transporters to enter cells. Our experiments demonstrated that none of the glycosylated enkephalins were inhibited either by phlorizin or phloretin ( Table 1, and Figure 4B) suggesting that transport across membranes was not mediated by SGLT1 or GLUT2. Adsorptive transport is therefore the most likely mechanism by which the carbohydrate-derived enkephalin compounds crossed the membrane.

SPR Experiments to Determine Binding between ASGPR and Enkephalin Derivatives
Peptide glycosylation could facilitate their binding to specific carbohydrate receptors [17,22]. All of our glycosylated products contained a terminal galactose moiety making them potential targets for the lectin receptor, ASGPR. SPR experiments with ASGPR protein immobilized onto the sensor chip were performed to test whether the glycosylated enkephalins could bind to ASGPR. D-Galactose and D-lactose were used as controls to show that ASGPR could actively bind its native substrate while bound to the sensor chip. Enkephalin, Gal-Enk, and Lac-Enk showed similar binding to ASGPR with K D values of ,200 mM ( Table 2). Our results indicated that the increasing length of the glycan residue resulted in higher binding affinity to the receptor. Consequently, Gal-Lac-Enk, which contained a trisaccharide moiety, displayed a two-fold increase in binding affinity (K D 91.1 mM) compared to the binding affinity of the peptide alone. The interaction kinetics was measured to confirm that binding to ASGPR was biologically relevant. Kinetic analysis demonstrated that enkephalin modified with longer glycan chains (Lac-Enk and Gal-Lac-Enk) had a slower on-rate (K on ) and a slower off-rate (K off ) ( Table 2). However, the off-rate of Lac-Enk for dissociating from ASGPR was twice that of Gal-Lac-Enk resulting in the lower affinity (K D 211.3 mM) reported for Lac-Enk. This indicated that it took longer for Gal-Lac-Enk to enter the binding pocket of ASGPR but once Gal-Lac-Enk was bound it was more stable than Lac-Enk, Gal-Enk, or enkephalin. Together the data suggested that only Gal-Lac-Enk binding to ASGPR was biologically relevant and therefore glycosylation of enkephalin with at least a trisaccharide moiety was necessary to achieve effective ASGPR binding.

Molecular Modeling Studies for Enkephalin Derivatives Binding to ASGPR
We also conducted molecular modeling experiments for the enkephalin derivatives using the crystal structure of ASGPR (PDB 1DV8, [38]). The molecular docking scores obtained showed a good correlation to the surface plasmon resonance data (Table 2). Enkephalin, Gal-Enk and Lac-Enk showed very low binding affinities to the active site of ASGPR with positive MolDock scores (Table 3). However, Gal-Lac-Enk bound to ASGPR with a lower total interaction energy (a negative docking score of 29 kcal/mol) indicating a higher binding affinity to the receptor ( Table 3).
The similar outcomes from both the SPR experiment and the molecular modeling studies indicated that the attachment of enkephalin most likely hindered the terminal galactose moiety on Gal-Enk and Lac-Enk from accessing the active site of ASGPR. In the case of Gal-Lac-Enk that contained a trisaccharide moiety, the greater distance between peptide and the terminal galactose enabled binding to the active site of the receptor (Figure 5 and 6).

Conclusions
We successfully used a one-pot enzymatic reaction that contained UDP-Glc, galactose epimerase and LgtC galactosyltransferase to galactosylate Lac-Enk to Gal-Lac-Enk. In comparison to the parent peptide, all chemo-enzymatically glycosylated enkephalins showed improved stability and permeability in vitro. The interaction between the glycosylated enkephalins and the carbohydrate receptor ASGPR was examined by SPR experiments and molecular modeling analysis. The SPR systems are widely used as a standard tool in areas such as pharmaceutical drug discovery. Here, SPR in conjunction with molecular modeling, allowed the rapid in vitro evaluation of glycosylated peptide-receptor interactions. The SPR results showed that the addition of a glycan moiety to the enkephalin peptide enabled binding of the peptide to ASGPR. The binding affinity between ASGPR and glycosylated enkephalins increased when the number Figure 5. Docking between Gal-Lac-Enk and the ASGPR receptor. A) Cartoon representation of Gal-Lac-Enk with the highest negative docking score conformation, in the active site of ASGPR. B) Surface view of ASGPR with Gal-Lac-Enk, docked into the active site. For clarity, atoms have been colored as follows: blue-nitrogen, red-oxygen, gray-carbon, green-Ca 2+ (on ASGPR) and pink-carbon (on Gal-Lac-Enk). doi:10.1371/journal.pone.0095024.g005 of sugar moieties increased from one (Gal) to three (Gal-Lac). The molecular docking analysis also showed the binding of a trisaccharide (Gal-Lac) modified enkephalin to the active site of ASGPR. The chemo-enzymatic conjugation of oligosaccharides to therapeutic peptides can potentially be applied to cell-specific targeting via carbohydrate receptors. However, further experiments, including in vivo experiments will be required to confirm this observation. Given the wide array of glycosyltransferases available, we envisage that this technique may also be more widely applied to modify therapeutic peptides to contain different oligosaccharide moieties for targeting other carbohydrate receptors. Figure 6. Details of the interaction between Gal-Lac-Enk in the conformation that gave the highest negative docking score and the ASGPR receptor. (1) A hydrogen bond between the oxygen of the -OH group on C4 of the terminal galactose moiety in Gal-Lac-Enk and the -NH group of His256 of ASGPR (O-N distance 3.4 Å ), (2) a hydrogen bond between the oxygen of the -OH group on C4 of the terminal galactose of Gal-Lac-Enk and the -OH of a carboxyl group of Asp266 of ASGPR (O-O distance 2.6 Å ), (3) a hydrogen bond between the oxygen of the -OH group on C3 of the terminal galactose of Gal-Lac-Enk and the NH of Asn264 of the receptor (O-N distance 2.3 Å ), (4) a hydrogen bond between the oxygen of the -OH group on C6 of the glucose moiety of Gal-Lac-Enk and the -NH of Arg236 of ASGPR (O-N distance 3.4 Å ) and (5) a hydrogen bond between the -NH group of Phe in Gal-Lac-Enk and the oxygen of the carbonyl group of Asp241 of ASGPR (N-O distance 2.7 Å ). Atom colors are as follows: bluenitrogen, red-oxygen, gray-carbon, green-Ca 2+ (on ASGPR) and pink-carbon (on Gal-Lac-Enk). Hydrogen bonds are represented in cyan. doi:10.1371/journal.pone.0095024.g006