Thiodipeptides targeting the intestinal oligopeptide transporter as a general approach to improving oral drug delivery

The broad substrate capacity of the intestinal oligopeptide transporter, PepT1, has made it a key target of research into drug delivery. Whilst the substrate capacity of this transporter is broad, studies have largely been limited to small peptides and peptide-like drugs. Here, we demonstrate for the first time that a diverse range of drugs can be targeted towards transport by PepT1 using a hydrolysis resistant carrier. Eleven prodrugs were synthesized by conjugating modified dipeptides containing a thioamide bond to the approved drugs ibuprofen, gabapentin, propofol, aspirin, acyclovir, nabumetone, atenolol, zanamivir, baclofen and mycophenolate. Except for the aspirin and acyclovir prodrugs, which were unstable in the assay conditions and were not further studied, the prodrugs were tested for affinity and transport by PepT1 expressed in Xenopus laevis oocytes: binding affinities ranged from approximately 0.1 to 2 mM. Compounds which showed robust transport in an oocyte trans-stimulation assay were then tested for transcellular transport in Caco-2 cell monolayers: all five tested prodrugs showed significant PepT1-mediated transcellular uptake. Finally, the ibuprofen and propofol prodrugs were tested for absorption in rats: following oral dosing the intact prodrugs and free ibuprofen were measured in the plasma. This provides proof-of-concept for the idea of targeting poorly bioavailable drugs towards PepT1 transport as a general means of improving oral permeability.


Introduction
The oral bioavailability of a compound is a crucial factor in its success or failure as a therapeutic agent, particularly given the convenience of this route of administration. There are two main mechanisms of absorption from the GI tract: passive diffusion [1] and carrier mediated transport [2]. The oral bioavailability of poorly absorbed drugs can be improved either by modifying their physicochemical properties to aid passive diffusion and/or by targeting of the compounds towards carrier mediated transport [3e5].
PepT1 is a proton coupled oligopeptide transporter expressed principally in the small intestine and the proximal tubule of the kidney [6]. It has a broad substrate specificity including most diand tripeptides, b-lactam antibiotics and ACE inhibitors [7].
There are many examples of targeting PepT1 to improve the oral bioavailability of pharmacologically active compounds, usually by modifying them so that they resemble the natural di-or tripeptide substrates [8e13]. We have patented [14] a set of thiodipeptide substrates (such as A and B) that we hope can act as "carriers" for drug transport by PepT1 generally, and have previously published our work on model systems demonstrating that a variety of linkers can be employed [15,16]. The basic premise is illustrated in Fig. 1 in which drugs are conjugated directly or by a linker to our thiodipeptides, converting them into prodrugs that are PepT1 substrates.
In this paper, we apply the results of our previously reported characterisation of the structure-transport relationships for PepT1 [15] to drug delivery challenges and report proof-of-concept studies that validate the use of our thiodipeptide carriers as a general approach for targeting a variety of drugs towards PepT1 mediated transport. We focused on two major areas that we felt could benefit from our thiodipeptide drug delivery technology: i) Drugs with GI side effects. A common class of such drugs are the NSAIDs, as exemplified by aspirin and ibuprofen [17]. Whilst these drugs have high oral bioavailability, they also can cause severe gastric side effects. If a prodrug strategy could be developed so that bioavailability was retained, but active drug was not released close to the GI tract, such side effects might be significantly reduced. Prodrugs 1, 4 and 6e7 of ibuprofen, aspirin and nabumetone respectively ( Fig. 2) were synthesized to explore this area. ii) Drugs with poor oral bioavailability. This is a major challenge in drug development. A search of ChEMBL [18] identified several marketed drugs with low, highly variable or no oral bioavailability [17]: gabapentin (an anticonvulsant and analgesic); baclofen (a GABA receptor agonist); propofol (chemotherapeutic nausea and intractable migraine); zanamivir (treatment and prophylaxis of influenza) and mycophenolic acid (an immunosuppressant). Prodrugs 2, 3, 5, and 8e11 ( Fig. 2) were synthesized to prove our concept in this important area.

Chemistry
The synthesis of the protected serine and aspartate carrier thiodipeptides (12 and 13), nabumetone prodrugs 6e7 and ibuprofen prodrug 1 have been reported previously [15,16]. Our chosen drugs could readily be attached to the appropriate carrier using standard coupling reagents, except for the aspirin prodrug 4 (Table 1). This was synthesized by first using concentrated Mitsunobu conditions [19] with sonication to esterify the salicylic acid with triethylene glycol to give 22, then coupling this glycol ester to the aspartate carrier using standard coupling conditions (Scheme 1) to give 23. This indirect route was chosen because we were unable to accomplish direct esterification of aspirin with the serine carrier using a variety of coupling conditions. Deprotection was usually achieved in >85% yield using either a 33% solution of TFA in DCM or neat formic acid, except for 5, for which decomposition was avoided by using phenol as solvent [20]. Since the NMR [15] of carriers 12 and 13 show no signs of epimerisation, and rotamers observed in the NMR of some final compounds have spectral characteristics consistent with cis/trans rotamers around the thioamide bond as we have previously reported [21], we do not believe epimerisation occurred during synthesis.

Results and discussion
The results of binding studies, trans-stimulation and Caco-2 monolayer assays are summarised in Table 2. The binding affinities of all prodrugs for PepT1 were determined by measuring the concentration at which they inhibit uptake of radiolabelled D-Phe-L-Gln in Xenopus laevis oocytes expressing rabbit PepT1. Inhibition constants were calculated from standard Michaelis-Menten kinetics [22,23]. PepT1 is a low affinity, high capacity transporter and compounds with an affinity <1 mM are generally classed as high affinity binders of the transporter. Fig. 3 shows the data for prodrugs 1 and 3, which are representative of those determined for all the prodrugs. Prodrugs 4 and 5 had limited stability in the pH 5.5 assay buffer (multiple HPLC peaks), and so no reliable affinity or transport data could be generated.
As binding studies only show affinity for PepT1 and do not provide information as to whether the compound is a substrate or an inhibitor, further transport experiments were undertaken. Trans-stimulation assays were performed using radiolabelled [ 3 H]-D-Phe-L-Gln efflux from rabbit PepT1 expressing oocytes in the   presence of 10 mM pro-drug. As controls, 10 mM Gly-L-Gln (a standard PepT1 substrate) or buffer lacking a substrate (negative control) were used. Fig. 4 shows the efflux data for the compounds summarised in Table 2. Prodrugs 1e3, 6e7 and 11 induced statistically significant transstimulation efflux in oocytes, thereby demonstrating PepT1 mediated transport [22,23]. Most of these induced similar or greater efflux than GlyGln, however prodrug 11 only weakly triggered efflux in comparison to GlyGln.
Prodrugs that generated robust trans-stimulated efflux in oocytes were further in a Caco-2 monolayer assay to investigate further the extent and rate of trans-epithelial transport. Caco-2 cells were chosen as they are widely accepted as a good overall model for the small intestinal epithelium [24], although it has been suggested that Caco-2 cells may underestimate the in vivo trans-epithelial rate of transport [25]. Apical to basolateral transport of 2 mM pro-drug, applied to the apical side, was monitored by high performance liquid chromatography (HPLC) after 1 h. The presence or absence of excess Gly-Gln allowed us to determine both the overall and PepT1 specific permeability ( Table 2). The remaining pro-drugs were significantly transported in both oocyte and Caco-2 monolayer assays. The PepT1 mediated Papp values are of similar magnitude to known PepT1 (pro)drug substrates [10,16].
Based on these encouraging in vitro results, we elected to study 1 and 3 in vivo (Table 3), as simple examples of the two areas of interest to us. Administration of both 1 and 3 to rats resulted in intact prodrug being observed in the blood, with C Max of 0.2 and 16.7 mg/mL respectively observed. Release of ibuprofen was also observed upon administration of 1, with a relative bioavailability of 2% [26]. The shift in the T Max observed for ibuprofen following administration of 1 when compared to free ibuprofen (from 1 h for free ibuprofen to 3.5 h for 1) is indicative of a change in absorption mechanism. The low bioavailability of ibuprofen can be explained by the relative stability of the prodrug 1 to metabolism, as estimated by its in vitro half-life upon incubation with rat liver homogenate of over 17 h.
Regrettably, we were unable to detect free propofol released following administration of prodrug 3 because HPLC conditions to quantify free propofol could not be found. However, the relatively high C Max for the intact prodrug 3 is encouraging and despite the relative metabolic stability of 3 (7-h half-life in rat liver homogenate) it is likely some free propofol would be available systemically. This is notable given the fact that propofol itself had no oral bioavailability in either rats or humans [27].
The low bioavailability of ibuprofen following administration of prodrug 1, combined with fact that both prodrug 1 and 3 were relatively resistant to liver metabolism indicates that further work is required to design effective prodrugs suitable for therapeutic application. Nevertheless, we believe 1 and 3 serve as promising preliminary proof-of-concept to the idea that targeting PepT1 using our thiodipeptides can be used as a general strategy to overcoming oral bioavailability issues in drug discovery and development.

Conclusions
PepT1 is described in the literature as having a broad substrate capacity but, in reality, it has been limited to date to small peptides and peptide-like drugs. To better harness the capacity of this transporter as a drug delivery target, a rational and general targeting approach is required. We report here our data supporting a thiodipeptide prodrug as such a general targeting approach in vitro and in vivo.
We were excited to find that nine out of eleven of our rationally designed PepT1 targeting prodrugs displayed high affinity binding towards PepT1, and six of them triggered trans-stimulation. Additionally, prodrugs 1e3, as well as 6e7 (as previously reported) [16],  with at least three monolayers. The normalised figure is to the FSA value recorded in that experiment. b Prodrug was unstable to assay buffer (pH 5.5). c Previously reported data [16].  were all significantly transported in Caco-2 monolayers, with prodrugs 1 and 3 showing evidence of intact absorption in vivo. This provides proof-of-concept that diverse drug types can be delivered via a PepT1 mediated pathway using thiodipeptide carriers, with implications for future drug design strategies. Preliminary in vivo data also supports the use of thiodipeptide prodrugs to confer oral bioavailability.
The drugs exemplified represent examples of several classes of drugs for which oral delivery could be therapeutically interesting. These include thiodipeptide prodrugs of NSAIDs (e.g. 1, 4, 6e7), which have high oral activity but also suffer from significant GI side effects. Examples of drugs for which oral activity is absent (e.g. 3), low (e.g. 9), or highly variable (2, 10) have also been successfully modified using our thiodipeptide approach, to target the PepT1 transporter.
There is much future work to conduct before we can confidently say that rationally targeting PepT1 is a general strategy for oral drug delivery. In particular, the complete in vivo DMPK profile of our prodrugs needs to be established and future prodrugs need to be optimized for rapid liver and/or plasma esterase metabolism and release of free drug, or indeed potentially slower, sustained release if desired. Optimisation of the stability of the linker is also required, as evidenced by the instability of prodrugs 4 and 5. However, our previously reported work suggests that the transporter can accommodate a wide variety of linkers [15,16], allowing scope to tailor the DMPK properties of a specific prodrug.
The oral delivery of drugs is a major challenge in pre-clinical development and leads to significant shelving of promising lead candidates in drug discovery. Our prodrug approach may allow many such biologically active compounds to be re-evaluated by administration as PepT1 targeting thiodipeptide prodrugs. Our in vitro and preliminary in vivo data is highly encouraging and warrants further work. In particular, our recent report that large peptide drugs such as cyclosporine A [28] can be rationally targeted towards PepT1 using the same approach offers the tantalising possibility of PepT1 targeting as a solution to the delivery of both small and peptidic molecules.

In vitro biological studies
The K i , trans-stimulation efflux and Caco-2 assays were performed as described previously [15,16].
Fresh rat liver homogenate was prepared by isolating liver from euthanized male rats, according to approved Home Office procedures. The liver was chopped with scissors and rinsed with medium (0.25 M sucrose, 25 mM KCl, 5 mM MgCl 2 and 50 mM Tris/HCl, pH7.5) to remove trapped blood. The liver was then homogenized in fresh medium with a loose-fitting dounce-type homogenizer and kept on ice. Liver homogenate was incubated with either 0.5 mM compound 1 or 3, or L-Trp-L-Ala as a positive control, at 37 C. 250 ml aliquots of the homogenate were taken at 0, 0.25, 0.5, 1, 2, 6 and 24 h. The samples were precipitated by addition an equal volume of 3% perchloric acid and centrifugation at 17000g for 5 min. The perchloric acid was neutralized with 250 ml of 1 M KOH, and the sample subjected to a freeze/thaw cycle to precipitate the KClO 4 salt before again being centrifuged. The supernatant was then analyzed by HPLC as for the Caco-2 permeability studies [15,16], and the halflife of the compounds calculated according to the method of Vig et al. [29]. Each test compound was administered orally as solutions in distilled water (ibuprofen, 1) or polypropylene glycol (3) to three adult male Sprague-Dawley weighing 250e300 g, which were housed singly following jugular vein cannulation prior to administration of compound. Animals were given free access to food and water throughout the study and maintained under a 12-h light/ dark cycle with temperature and humidity controlled according to Home Office regulations. All compounds were well-tolerated and no-adverse events were reported. Blood samples were centrifuged to obtain the plasma, which was transferred to a separate labelled container. Aliquots from the individual time points for the three animals were analyzed singly. 80 mL of plasma was diluted with 20 mL of 1:1 ACN:water, then 800 mL chilled ACN was added, samples briefly vortex mixed and the centrifuged at 13000 rpm for 5 min at 4 C. 500 mL of the resultant supernatant was further diluted with 500 mL water. 20 mL sample was analyzed by LC-MS/MS using a C18 5 mm

In vivo studies
Gemini UHPLC column running a gradient of 90% 0.04% acetic acid in water to 90% ACN over 3 min at a flow rate of 0.5 mL min À1 . MS data was acquired under multiple reaction monitoring conditions using a turbo spray ion source. The concentration in the plasma was determined by comparison to standard curves of the administered compound prepared in blank plasma matrices and treated in an identical manner to the samples.

Synthetic chemistry
Anhydrous solvents and reagents were obtained as follows: DMF was dried three times over molecular sieves (3 Å). THF was dried by distillation from sodium benzophenone ketyl, DCM and toluene by distillation from calcium hydride. All reactions were conducted at room temperature in dry glassware under a nitrogen atmosphere, unless otherwise stated. All chemicals were used directly from suppliers' (Sigma-Aldrich) vessel without further   (14) Gabapentin (85 mg; 0.5 mmol) and di-tert-butyl dicarbonate (130 mg; 0.6 mmol) were suspended in 1 mL DMF. TEA (0.07 mL; 0.5 mmol) was added and the suspension stirred for four days at room temperature. A solution was formed after the first 3 h of stirring. The DMF was removed in vacuo. HBTU (245 mg; 0.6 mmol) and DIPEA (0.11 mL; 0.6 mmol) were added to the residue, which was then redissolved in 1 mL DMF. The resultant solution was stirred at room temperature for 30 min 12 (132 mg; 0.4 mmol) in 1 mL DMF was then added and the solution stirred for three days at room temperature. The DMF was removed and the residue purified by flash column chromatography (4:1 hexane:EtOAc / 1:1 hexane:EtOAc), followed by semi-preparative HPLC (2:1 hexane:EtOAc) to give title compound as a colourless oil (52 mg; 22%). HPLC R T : 4

2-(2-Oxo-propyl)-benzoic acid 2-[2-(2-hydroxy-ethoxy)ethoxy]-ethyl ester (21)
A suspension of aspirin (345 mg; 1.9 mmol), triphenylphosphine (545 mg; 2.1 mmol) and triethylene glycol (0.27 mL; 2.0 mmol) in 0.7 mL THF was sonicated to give a viscous solution. DIAD (0.40 mL; 2.0 mmol) was then added over 5 min to give a yellow solution. This was sonicated for 15 min at room temperature. The solution was purified by flash column chromatography (4:1 hexane:EtOAc / EtOAc) to give crude product. This was further purified by flash column chromatography (    (2-amino-6-oxo-1,6-dihydro-purin-9-ylmethoxy)-ethyl] ester (5) A mixture of 16 (51 mg; 0.1 mmol) and phenol (179 mg; 1.9 mmol) was heated to 45 C at which point the now liquid phenol had dissolved. Trifluoroacetic acid (0.03 mL; 0.4 mmol) was then added and the solution stirred at 45 C for 1 h. The solution was diluted with 2 mL EtOAc and washed three times with 3 mL water. The EtOAc was removed in vacuo and the residue sequentially taken up in DMF and water, both of which were removed under vacuum. The resultant residue, which was free from phenol contamination, was redissolved in 5 mL water and lyophilised to give di-TFA salt of  (17) S (À)-Atenolol (100 mg, 0.37 mmol) was dissolved in water (10 ml) containing (59 mg, 0.56 mmol) of sodium carbonate and the mixture was cooled to 0 C with stirring. Then, the solution of ditert. butyl dicarbonate (125 mg, 0.57 mmol) in 10 ml of 1,4-dioxane was slowly added at the same temperature. The mixture was stirred overnight at room temperature, evaporated in vacuo to get white solid, diluted with 20 ml of water, and extracted with three portions of ethyl acetate. The ethyl acetate solution was washed with brine, dried (MgSO 4 ) and evaporated in vacuo to afford white solid (131 mg). A solution of this in 10 mL MeCN was slowly added over 20 min to a solution of 13 (150 mg, 0.39 mmol) in MeCN (15 ml) containing (117 mg, 0.71 mmol) of 1,1 0 -carbonyl diimidazole and (14 mg, 0.2 mmol) of imidazole. The mixture was stirred for 3 days at room temperature under the inert atmosphere. The precipitated solid was filtered off and the filtrate was evaporated in vacuo to get thick brown gel, which can further diluted with 30 ml of EtOAc, and washed with 2 M HCl (2 Â 10 ml), saturated aqueous NaHCO 3 (3 Â 10 ml), water (2 Â 10 ml) and finally with the brine (20 ml), dried (MgSO 4 ) and evaporated in vacuo to afford white crude solid. The residue was purified by flash column chromatography (3:7 petrol:EtOAc / EtOAc) to give desired product as a white solid (130 mg; 36%

(S)-4-((S)-3-(4-(2-amino-2-oxoethyl)phenoxy)-1-(isopropylamino)propan-2-yloxy)-2-((S)-2aminopropanethioamido)-4-oxobutanoic acid (8)
17 (100 mg, 0.138 mmol) was dissolved in 3 mL 97% formic acid. The solution was refluxed at 100 C for 3 h, followed by room temperature for overnight. The excess formic acid was then removed under high vacuum and the residue taken up in 2 mL distilled water. The fine suspension was filtered through a pipette plugged with glass wool and lyophilised to give the formate salt of 8 as a brown solid (53 mg; 76%  was added and stirring was continued for another 3 days at room temperature. The reaction mixture were filtered off and the filtrate, plus a DMF washing, was evaporated in vacuo to get crude oil, which was further purified by chromatography, eluting with neat EtOAC to 1:1 (MeOH: EtOAC) to give desired product as an off white solid (20 mg, 11%