Liver-targeted delivery of insulin-loaded nanoparticles via enterohepatic circulation of bile acids

Abstract Liver is the primary acting site of insulin. In this study, we developed innovative nanoparticles for oral and liver-targeted delivery of insulin by using enterohepatic circulation of bile acids. The nanoparticles were produced from cholic acid and quaternary ammonium modified chitosan derivative and hydroxypropyl methylcellulose phthalate (HPMCP). The nanoparticles had a diameter of 239 nm, an insulin loading efficiency of 90.9%, and a loading capacity of 18.2%. Cell culture studies revealed that the cholic acid groups effectively enhanced the transport of the nanoparticles through Caco-2 cell monolayer and greatly increased the absorption of the nanoparticles in HepG-2 cells via bile acid transporter mechanism. Ex vivo fluorescence images of ileum section, gastrointestinal tract, and liver demonstrated that the HPMCP increased the mucoadhesion of the nanoparticles in ileum, and the cholic acid groups facilitated the absorptions of the nanoparticles in both ileum and liver by use of bile acid transporters via enterohepatic circulation of bile acids. The therapy for diabetic mice displayed that the oral nanoparticle group could maintain hypoglycemic effect for more than 24 h and its pharmacological availability was about 30% compared with the insulin injection group. For the first time, this study demonstrates that using enterohepatic circulation of bile acids is an effective strategy for oral delivery of insulin.


Introduction
Diabetes mellitus is a worldwide chronic disease and an estimated 415 million people had diabetes worldwide as of 2015 (IDF Diabetes Atalas, 2017). Insulin, secreted by pancreas, is the only hormone in human body that can reduce BGL (blood glucose level) directly (Saltiel & Kahn, 2001, Edgerton et al., 2006. As the primary acting site of insulin, liver plays a major role in carbohydrate metabolism and takes the responsibility for the balance of BGL by means of glycogenogenesis and glycogenolysis (Pessin & Saltiel, 2000;Arbit, 2004;Edgerton et al., 2006). After being transported to liver cells, endogenous insulin stimulates the synthesis of glycogen that transforms the glucose in the blood into the glycogen in the liver and inhibits the breakdown of the glycogen (Pessin & Saltiel, 2000). By now, as the most effective treatment, subcutaneous injection of insulin every day is still the best choice for both type 1 and 2 diabetics (Li et al., 2018). However, subcutaneously injected insulin enters into the general circulation directly, which exposes all tissues to the same insulin concentration and the liver only receives a small fraction of the injected dose; thus, muscles and adipocytes can react to the insulin without hepatic monitoring, and side effects such as atherosclerosis, hypoglycemia and weight gain may occur (Arbit, 2004;Geho et al., 2009).
For insulin therapy, oral administration is most acceptable by diabetics, but the oral bioavailability of insulin is very low due to the physiological barriers in gastrointestinal (GI) tract, including chemical, enzymatic, and absorption barriers (Lopes et al., 2014). The orally administrated insulin undergoes denaturation due to the acidic environment in the stomach and is broken down in the GI tract by the proteolytic enzymes. Furthermore, the absorption of the insulin through the mucus covered intestinal epithelia is limited due to the high molecular weight and hydrophilicity of insulin (Mo et al., 2014). To overcome gastric acid and enzyme damages as well as facilitate the absorption in GI tract, many microand nano-sized systems have been fabricated for oral delivery of insulin (Mo et al., 2014). For example, the nanoparticles fabricated from chitosan and its derivatives are mucoadhesive (Zambito et al., 2013), which can enhance the absorption of the loaded insulin in GI tract (Sheng et al., 2016). Hydroxypropyl methylcellulose phthalate (HPMCP), a widely used enteric coating material in pharmaceutical industry (Singh et al., 2015), can protect insulin from degradation and denaturation in the harsh environments of stomach (Du et al., 2015). Makhlof et al. (2011) reported that the acid stability, and the intestinal mucoadhesion and penetration of insulin-loaded chitosan/HPMCP nanoparticles were significantly improved compared with the chitosan/tripolyphosphate nanoparticles.
It was reported that the oral delivery systems conjugated with bile acid can deliver the drugs to the liver directly by use of enterohepatic circulation mechanism (HO, 1987;Swaan et al., 1996;Zhang et al., 2016a). Enterocytes in ileum express apical Na þ -dependent bile acid transporter (ASBT) and cytosolic ileal bile acid-binding protein (IBABP) (Gong et al., 1994;Kolhatkar & Polli, 2012;Fan et al., 2018). Hepatocytes express bile acid transporters such as Na þdependent taurocholate cotransporting polypeptide (NTCP) and Na þ -independent organic anion transporting polypeptides (OATPs) (Schadt et al., 2016). These bile acid transporters are involved in the enterohepatic circulation of bile acids. The high capacity of the bile acid transporters and high efficacy of both intestinal and liver absorptions make the enterohepatic circulation of bile acids be utilized in oral delivery systems to increase the therapeutic concentration in the liver and reduce the general toxicity of the drugs (Swaan et al., 1996;Zhang et al., 2016a). For insulin delivery, it was evidenced that the bile acid conjugated insulin was taken up by the ileal bile acid transporters after infusion into the small intestine (McGinn & Morrison, 2016). Very recently, Fan et al. (2018) reported deoxycholic acid-modified nanoparticles produced by self-assembly of insulin, deoxycholic acid-modified chitosan, and poly (c-glutamic acid). The nanoparticles could overcome multiple barriers of the intestinal epithelium by using ASBT-mediated endocytosis and IBABP-guided intracellular trafficking, and facilitated the basolateral release of free insulin. To the best of our knowledge, no system for oral and liver-targeted delivery of insulin by using enterohepatic circulation of bile acids was reported by now.
To improve oral bioavailability of insulin, in this study, we designed and fabricated a novel delivery system. We used cholic acid and N-(2-hydroxy)-propyl-3-trimethylammonium chloride modified chitosan (HTCC-CA) and HPMCP to produce insulin-loaded nanoparticles INS/HTCC-CA/HPMCP. The nanoparticles were expected to prevent the loaded insulin from denaturation and degradation in GI tract as well as to improve the intestinal mucoadhesion. The nanoparticles were also expected to facilitate the intestinal and liver absorptions of the loaded insulin by utilizing the enterohepatic circulation of bile acids. We performed a series of experiments to reveal the absorption mechanism of the nanoparticles in intestinal and liver, and to prove that the nanoparticles can greatly improve the oral bioavailability of insulin.

Preparation of nanoparticles
HTCC (N-(2-hydroxy)-propyl-3-trimethylammonium chloride modified chitosan) and HTCC-CA (cholic acid modified HTCC) were synthesized, purified, and characterized as reported previously (Zhang et al., 2016b). The quaternary ammonium degree was 35.8% and CA (cholic acid) conjugation degree was 5.7% of the glycosyl units of chitosan. HTCC, HTCC-CA, and HPMCP stock solutions were prepared by dissolving the polymers in deionized water respectively and adjusting the solutions to pH 7.4. Insulin stock solution was prepared by dissolving insulin in 0.01 M HCl solution and adjusting the solution to pH 7.4.
INS/HTCC-CA nanoparticles were prepared by dropwise adding 1 mL of 1 mg/mL insulin solution into 2 mL of 1 mg/mL HTCC-CA solution with gentle stir. Successively, 3 mL of the INS/HTCC-CA nanoparticle solution was added dropwise into 2 mL of 1 mg/mL HPMCP solution with gentle stir to produce INS/HTCC-CA/HPMCP nanoparticles. Similarly, INS/HTCC/HPMCP nanoparticles were prepared. The final concentrations of insulin, HTCC or HTCC-CA, and HPMCP in INS/ HTCC/HPMCP and INS/HTCC-CA/HPMCP nanoparticles were 0.2, 0.4, and 0.4 mg/mL, respectively. Both the final concentrations of insulin and HTCC-CA in INS/HTCC-CA nanoparticles were 0.2 mg/mL.

Characterization of the nanoparticles
Z-Average hydrodynamic diameter (D h ), polydispersity index (PDI), and f-potential of the nanoparticles were measured on a laser light scattering instrument (Zetasizer Nano ZS90, Malvern Instruments, Malvern, UK) as reported previously (Zhang et al., 2016b). Transmission electron microscopy (TEM) images of the nanoparticles were acquired on a transmission electron microscope (Philips CM120 electron microscope, Philips, Amsterdam, Netherlands). Free insulin in the nanoparticle solutions was separated using centrifugal filter (cutoff molecular weight 100 kDa, Millipore, Billerica, MA) and the insulin concentrations in the filtrates were analyzed using BCA assay. The insulin loading efficiency (LE) and loading capacity (LC) of the nanoparticles were calculated using the following equations: LE %; wt wt ¼ total insulin À free insulin total insulin Â 100% LC %; wt wt ¼ total insulin À free insulin total polymers þ total insulin Â 100% In vitro release of insulin from the nanoparticles Insulin releases from the nanoparticles were investigated by dialysis of 1 mL of the nanoparticle solution (cutoff molecular weight 100 kDa, Spectrum Laboratories Inc., Piscataway Township, NJ) against 4 mL of pH 2.0 HCl solution or pH 7.4 PBS (0.01 M phosphate buffer containing 0.15 M NaCl) solution at 37 C with shaking. At predetermined intervals, 1 mL of the release medium was taken out and the same volume of fresh medium was added. The insulin concentration in the release medium was determined using BCA assay.

FITC-labeled insulin (FITC-INS) and
Cy5-labeled insulin (Cy5-INS) were synthesized and purified as described in the literature (Wang et al., 2016;Zhang et al., 2016b). The fluorescence-labeled nanoparticles were prepared as described above

Permeability of the nanoparticles across Caco-2 cell monolayer
Caco-2 cells were seeded in transwell inserts at a density of 5 Â 10 3 cells/well and were cultured as reported in the literature (Sheng et al., 2016) for 14 -21 d until their trans-epithelial electrical resistance (TEER) values were higher than 900 XÁcm 2 . The cell monolayer was washed with PBS thrice, then 0.2 mL DMEM containing insulin or insulin-loaded nanoparticles with insulin concentration of 50 lg/mL, or containing individual polymer with the concentration of 100 lg/mL, or containing free CA molecules with the concentration of 100 lM was added into the apical side; 0.6 mL DMEM without sample was added into the basolateral side. After 1 h incubation, the medium was removed. The cell monolayer was washed with PBS thrice and then was cultured in fresh DMEM for 9 h. At predetermined intervals, TEER of the Caco-2 cell monolayer was measured using an electrical resistance system (ERS-2, Millipore, Billerica, MA).
Apparent permeability coefficient (P app ) of insulin was measured as follows. After washing the Caco-2 cell monolayer with PBS, 0.2 mL DMEM containing FITC-INS or FITC-INS loaded nanoparticles with insulin concentration of 50 lg/mL was added into the apical side; 0.6 mL DMEM was added into the basolateral side. At predetermined intervals, sample was collected from the basolateral side and the same volume of fresh DMEM was added. The FITC-INS concentration in the sample was measured on a fluorescence microplate reader (Cytation3, BioTek, Winooski, VT). The P app of insulin was calculated using the following equation: where Q is the total amount of insulin permeated (ng), A is the diffusion area of the cell monolayer (cm 2 ), c is the initial concentration of insulin in the donor compartment (ng/cm 3 ), and t is the total time of the experiment (s).

Cellular uptake of insulin by HepG-2 cells
HepG-2 cells were seeded in special Petri dishes at a density of 1 Â 10 5 cells/well and cultured as reported previously (Zhang et al., 2016b). Subsequently, the cells were incubated with the culture medium containing FITC-INS or FITC-INSloaded nanoparticles at insulin concentration of 50 lg/mL. After 4 h incubation, the cells were washed with PBS thrice and the cell nuclei were stained with DAPI for 5 min, and then the cells were observed on a confocal laser scanning microscope (CLSM, C2þ, Nikon, Tokyo, Japan). The cellular uptakes of insulin were determined quantitatively using flow cytometry analysis. After 4 h incubation with the culture medium containing FITC-INS or FITC-INS loaded nanoparticles at insulin concentration of 50 lg/mL, the cells were washed with PBS thrice and then analyzed on a flow cytometer (FACSCalibur, BD, Franklin Lakes, NJ).

In vivo biocompatibility
Female ICR mice (25 ± 2 g) were from Sino-British SIPPR/BK Lab Animal Ltd, Shanghai, China. The animal experiments of this study were performed at Experimental Animal Center of School of Pharmacy of Fudan University in full compliance with the guidelines approved by Shanghai Administration of Experimental Animals. Healthy mice were separately administrated by gastric gavage with 0.2 mL mixed solution of HTCC-CA (10 mg/mL) and HPMCP (10 mg/mL) once daily for 15 and 30 d continuously. After the mice were sacrificed, the organ sections were prepared as reported previously (Zhang et al., 2016b). Histological images of the organ sections were acquired on a microscope (BX53, OLYMPUS, Tokyo, Japan).

Ex vivo fluorescence imaging of ileum section
Healthy mice were fasting for 12 h with freedom to water. Cy5-INS/RhB-HTCC-CA, Cy5-INS/RhB-HTCC/FITC-HPMCP, and Cy5-INS/RhB-HTCC-CA/FITC-HPMCP nanoparticles were separately administrated by gastric gavage at insulin dose of 30 IU/kg. The mice were sacrificed after 4 h of the administration. The ileum segments were taken out and washed with PBS thrice. The ileum segments were frozen in cryoembedding medium followed by cryostat section. The section was loaded on a microscope slide and fixed with DAPI fluoromount-G TM . The images of the section were acquired on the CLSM.

Ex vivo fluorescence imaging of organs
Healthy mice were fasting for 12 h with freedom to water. Cy5-INS/HTCC-CA, Cy5-INS/HTCC/HPMCP, and Cy5-INS/HTCC-CA/HPMCP nanoparticles were orally administrated at insulin dose of 30 IU/kg. The mice were sacrificed at 0, 2, 6, 12, and 24 h post-administration. The organs were excised and washed with PBS thrice. Ex vivo fluorescence images of the organs were observed on a small animal imaging system (In Vivo Xtreme, Bruker, Billerica, MA) and the sum fluorescence intensities of the organs were measured.

Antidiabetic efficacy
Healthy mice were intraperitoneally injected with alloxan solution at a single dose of 200 mg/kg to induce type 1 diabetes as reported previously (Zhang et al., 2016b). The blood from caudal vein was sampled and the BGL was measured using a glucometer (ACCUCHEK Active, Roche). The diabetic mice with average fasting BGL of 21.7 ± 3.5 mM were divided into five groups with five in each group. The mice were fasting for 10 h with freedom to water prior to administration. Insulin solution was injected subcutaneously into the mice at insulin dose of 3 IU/kg. Physiological saline, INS/HTCC-CA, INS/HTCC/HPMCP, and INS/HTCC-CA/HPMCP nanoparticles were separately administrated by gastric gavage at an insulin dose of 30 IU/kg. At predetermined intervals, the BGL was measured. At 4 h post-administration, about 0.2 g standard chow was provided for each of the mice. Insulin pharmacological availability (PA) of the nanoparticle (NP) groups were calculated according to the area above the relative BGL-time curve (AAC) using the following equation: For repeated administrations, diabetic mice with average fasting BGL of 17.5 ± 7.6 mM were divided into four groups with five in each group. Insulin solution was injected subcutaneously at insulin dose of 2 IU/kg once daily. Physiological saline, insulin solution, and INS/HTCC-CA/ HPMCP nanoparticles were separately administrated by gastric gavage at insulin dose of 30 IU/kg once daily. Rat chow was provided at 6 -12 h post-administration. Water was provided at all times. During the experiment, two mice in the insulin injection group and one mouse in the INS/HTCC-CA/ HPMCP oral group were died of hypoglycemia.

Statistical analysis
The data were expressed as mean ± SD (standard deviation). Statistical analysis was performed using independent samples-t test (OriginPro 8.0 software, SAS Inc., Cary, NC), and a p value <.05 was considered to be statistically significant.

Results and discussion
Preparation and characterization of insulin-loaded nanoparticles INS/HTCC-CA/HPMCP nanoparticles were prepared after mixing insulin with HTCC-CA and then HPMCP in pH 7.4 solution by means of electrostatic and hydrophobic interactions. For comparison, INS/HTCC/HPMCP and INS/HTCC-CA nanoparticles were prepared using the same process. In pH 7.4 solution, INS/HTCC-CA had D h and f-potential of 168 nm and 19.5 mV, respectively, as shown in Table S1 (Table S1 of Supplemental data), indicating that all the three systems can effectively encapsulate insulin.

In vitro insulin release
Insulin releases from the nanoparticles were investigated using a dialysis method in pH 2.0 HCl and pH 7.4 PBS media to mimic the pH environments in stomach and intestine.

Transport through Caco-2 cell monolayer
Caco-2 cells express bile acid transporters ASBT and IBABP, and Caco-2 cell monolayer can be used to mimic the enterocytes in studying the transport of the delivery system in intestinal barrier in vitro (Alam et al., 2014;Fan et al., 2018). The decrease of TEER value is considered as an open indication of the tight junctions between Caco-2 cells (Hsu et al., 2013). All the three polymers as well as individual CA reduced the TEER values significantly as shown in Figure 2(A). HTCC-CA had stronger impact on the TEER change than the others. The TEER changes were reversible after remove of the samples, suggesting that the cell monolayers recovered their integrity gradually. Figure 2 (Fan et al., 2018). To further prove this transport mechanism, the cell monolayers were pretreated with 100 lM free CA molecules for 30 min to block the bile acid transporters as reported in the literature (Khatun et al., 2014). Although free CA induced perturbation in the cell monolayer (Figure 2 monolayer via the bile acid transporters. This result can be explained by the facts that INS/HTCC-CA was unstable as proved in Figure 1(B, C) and the released insulin had very low P app value as demonstrated in Figure 2(C).

HepG-2 cellular uptake
Liver is the primary acting site of insulin (Arbit, 2004). HepG-2 cells express bile acid transporters, such as NTCP and OATPs (Swaan et al., 1996;Zhang et al., 2016a). Therefore, in this study, HepG-2 cells were used as an in vitro hepatocyte model to investigate hepatocyte uptake of the nanoparticles. In vivo biocompatibility Figure S1 of Supplemental data shows hematoxylin À eosinstained histological images of heart, liver, spleen, lung, kidney, stomach, and intestine of the mice after oral administration with HTCC-CA and HPMCP. The daily polymer dose in biocompatibility study was 30-fold higher than the dose of INS/HTCC-CA/HPMCP in the hypoglycemic study. Compared with the control group which was denoted as 0 d in Figure S1, the polymers did not induce significant morphological changes in the tissues after 15 and 30 d of continuous administrations, verifying that HTCC-CA and HPMCP, the carriers of insulin in this study, were biocompatible.
Distribution of the nanoparticles in histological section of ileum    The oral insulin group had no significant hypoglycemic effect compared with the saline group ( Figure 5(B)), which is the same as reported in the literature (Chuang et al., 2015). The injection group had rapid and short-acting effect after each administration of free insulin, the BGL reduced to about 40% of the initial level and recovered within 6 h. The oral INS/ HTCC-CA/HPMCP group had stable hypoglycemic effect, the BGL was kept at a steadily low level during the treatment, and the BGL value was lower than 50% of the initial level at each 24 h post-administration. The PA of the oral INS/HTCC-CA/HPMCP group was 36.7% compared with the injection group. The result in Figure 5(B) further confirms that INS/ HTCC-CA/HPMCP had prolonged and effective hypoglycemic effect after oral administration.
Oral delivery of insulin has been studied for many years, but the onset time, BGL control and PA of oral insulin were not satisfied yet, and no delivery system was applicable (Mo et al., 2014). As a hydrophilic polypeptide, the orally administrated insulin undergoes multiple biological barriers (Lopes et al., 2014). Stable nanoparticles can protect the insulin from denaturation and degradation in GI tract (Lopes et al., 2014;Fan et al., 2018). In this study, INS/HTCC-CA/HPMCP was more stable than INS/HTCC-CA and INS/HTCC/HPMCP as shown in Figure 1(B, C). Therefore, INS/HTCC-CA/HPMCP protected the loaded insulin in GI tract better than the others. The results in Figure 4 demonstrate that the INS/HTCC-CA/ HPMCP group had higher ileum distribution than the other groups that increased the interaction of INS/HTCC-CA/HPMCP with the epithelium. Most importantly, the enterohepatic circulation of bile acids was utilized in this study to deliver the loaded insulin to liver. As demonstrated by the results shown in Figures 2-4, INS/HTCC-CA/HPMCP utilized ASBT-mediated endocytosis, IBABP-guided intracellular trafficking and NTCP/ OATPs-mediated endocytosis to go through the epithelium, reach the liver, and be internalized by the hepatocytes. As mentioned above, liver is the primary acting site of insulin. INS/HTCC-CA/HPMCP increased accumulation and prolonged retention time of the insulin in the liver; therefore, INS/HTCC-CA/HPMCP had much better hypoglycemic effect than the other nanoparticles as shown in Figure 5 and Table 1. For the first time, this study demonstrates that using enterohepatic circulation of bile acids can effectively deliver the loaded insulin to liver.

Conclusions
In this study, we developed innovative INS/HTCC-CA/HPMCP nanoparticles for oral and liver-targeted delivery of insulin. This is the first study of using enterohepatic circulation of bile acids to deliver the loaded insulin to liver after oral administration. This study demonstrates that INS/HTCC-CA/ HPMCP protected the loaded insulin from denaturation and degradation in GI tract, the HPMCP increased the mucoadhesion of INS/HTCC-CA/HPMCP in ileum, and the CA groups greatly enhanced the absorptions of INS/HTCC-CA/HPMCP in both ileum and liver. INS/HTCC-CA/HPMCP increased oral PA of the loaded insulin to about 30% and could maintain hypoglycemic effect for more than 24 h. This study demonstrates that using enterohepatic circulation of bile acids to deliver the loaded insulin to liver is an effective strategy for oral  26.9 a The area above the curve shown in Figure 5(A) during 0-24 h.
insulin delivery, and HTCC-CA/HPMCP is a suitable carrier for oral insulin delivery.