European Journal of Pharmaceutical Sciences 63 (2014) 167–177

Contents lists available at ScienceDirect

European Journal of Pharmaceutical Sciences

journal homepage: www.elsevier .com/ locate/e jps

A new approach to predict human intestinal absorption using porcineintestinal tissue and biorelevant matrices

http://dx.doi.org/10.1016/j.ejps.2014.07.0030928-0987/� 2014 Elsevier B.V. All rights reserved.

Abbreviations: CYP, cytochrome P450 enzyme; FaSSIF, fasted-state simulatedintestinal buffers; FD4, fluorescein isothiocyanate-dextran MW 4000; FeSSIF, fed-state simulated intestinal buffers; GI, gastro intestinal; KRB, Krebs’ bicarbonateringer buffer; LDH, lactate dehydrogenase; Papp, apparent permeability; PhIP, 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine; SD, standard deviation; SEM,standard error of mean; TEER, transepithelial electrical resistance; TIM, TNOdynamic gastrointestinal Model.⇑ Corresponding author. Address: Utrechtseweg 48, 3704 HE Zeist, The Nether-

lands. Tel.: +31 8886 61740.E-mail address: heleen.wortelboer@tno.nl (H.M. Wortelboer).

Joost Westerhout, Evita van de Steeg, Dimitri Grossouw, Evelijn E. Zeijdner, Cyrille A.M. Krul,Miriam Verwei, Heleen M. Wortelboer ⇑The Netherlands Organization for Applied Scientific Research (TNO), Zeist, The Netherlands

a r t i c l e i n f o a b s t r a c t

Article history:Received 3 April 2014Received in revised form 12 June 2014Accepted 9 July 2014Available online 18 July 2014

Chemical compounds studied in this article:AntipyrineAtenololCaffeineCimetidineDigoxinMannitolPropranololRanitidineTestosteroneVerapamil

Keywords:InTESTine™Porcine intestineHuman intestineCaco-2Intestinal absorptionFood-effect

A reliable prediction of the oral bioavailability in humans is crucial and of high interest for pharmaceu-tical and food industry. The predictive value of currently used in silico methods, in vitro cell lines, ex vivointestinal tissue and/or in vivo animal studies for human intestinal absorption, however, is often insuffi-cient, especially when food–drug interactions are evaluated. Ideally, for this purpose healthy humanintestinal tissue is used, but due to its limited availability there is a need for alternatives. The aim of thisstudy was to evaluate the applicability of healthy porcine intestinal tissue mounted in a newly developedInTESTine™ system to predict human intestinal absorption of compounds with different chemical char-acteristics, and within biorelevant matrices. To that end, first, a representative set of compounds was cho-sen of which the apparent permeability (Papp) data in both Caco-2 cells and human intestinal tissuemounted in the Ussing chamber system, and absolute human oral bioavailability were reported. There-after, Papp values of the subset were determined in both porcine jejunal tissue and our own Caco-2 cells.In addition, the feasibility of this new approach to study regional differences (duodenum, jejunum, andileum) in permeability of compounds and to study the effects of luminal factors on permeability was alsoinvestigated. For the latter, a comparison was made between the compatibility of porcine intestinal tis-sue, Caco-2 cells, and Caco-2 cells co-cultured with the mucin producing HT29-MTX cells with biorele-vant samples as collected from an in vitro dynamic gastrointestinal model (TIM). The resultsdemonstrated that for the paracellularly transported compounds atenolol, cimetidine, mannitol and rani-tidine porcine Papp values are within 3-fold difference of human Papp values, whereas the Caco-2 Papp val-ues are beyond 3-fold difference. Overall, the porcine intestinal tissue Papp values are more comparable tohuman Papp values (9 out of 12 are within 3-fold difference), compared to Caco-2 Papp values (4 out of 12are within 3-fold difference). In addition, for the selected hydrophilic compounds a significant increase inthe permeability was observed from duodenum to ileum. Finally, this study indicated that porcine jejunaltissue segments can be used with undiluted luminal samples to predict human intestinal permeabilityand the effect of biorelevant matrices on this. In conclusion, viable porcine intestinal tissue mountedin the InTESTine™ system can be applied as a reliable tool for the assessment of intestinal permeabilityin the absence and presence of biorelevant samples. This would enable an accessible opportunity for areliable prediction of human intestinal absorption, and the effect of luminal compounds such as digestedfoods, early in drug development.

� 2014 Elsevier B.V. All rights reserved.

1. Introduction

In the development of pharmaceutical and nutritional products,a reliable prediction of the intestinal absorption in humans is cru-cial and of high interest to reach the desired effect in the humanbody. Several in silico, in vitro and in vivo animal methods are cur-rently used to predict human intestinal absorption. However, thepredictive value of these methods is often insufficient due to the

168 J. Westerhout et al. / European Journal of Pharmaceutical Sciences 63 (2014) 167–177

complexity of the human gastrointestinal (GI) tract, especiallywhen the interplay of active transport and metabolism, a food–drug interaction, or an excipient–drug interaction is involved(Lentz, 2008; Musther et al., 2014). Ideally, biorelevant matricescould be combined with in vitro permeability assays to get a betterinsight in the influence and interactions of luminal factors onin vivo permeability. Previous research, though, showed that thebarrier function of cell culture models such as Caco-2 cells wasreduced in combination with physiologically relevant luminalsolutions such as fasted-state simulated intestinal fluid (FaSSIF)or fed-state simulated intestinal fluid (FeSSIF) (Ingels et al.,2002). This is probably due to the lack of a protective mucus layer,making the Caco-2 cells much more sensitive to direct exposure tothe low pH and high osmolality of the buffers. Alternatively, aphospholipid vesicle-based permeation assay could be applied,which has been shown to be compatible with FaSSIF (Fischeret al., 2012). However, the phospholipid vesicles show limited sim-ilarity to the in vivo physiology with regards to metabolism andactive transport and therefore can only be applied for studyingthe potential effect of food on the passive diffusion.

Slightly modified FaSSIF and FeSSIF have been developed andapplied successfully for predicting the effect of food on humanintestinal absorption of poorly water-soluble drugs using Caco-2cells (Kataoka et al., 2011). However, these alternative intestinalsimulation fluids showed incompatibilities with rat intestinal tis-sue (Patel et al., 2006). Furthermore, some compounds, such asphosphate ester prodrugs, require a different approach as theyare incompatible with the high phosphate concentration in bothFaSSIF and FeSSIF (Brouwers et al., 2007). Alternatively, differentsolubilizers, such as cyclodextrins, can be added to the regularmedium for permeability studies, thereby mimicking the effect offood on the solubility (Dahan et al., 2010). However, the physiolog-ical relevance of such an approach is questionable. Therefore, thereis a need for an alternative approach which allows the combinationof biorelevant samples with an in vitro permeability model. Thein vitro permeability model for this purpose should reflect thephysiology of the in vivo intestinal epithelium, including the mucuslayer as a protective layer on the luminal side. Besides absorptiveenterocytes, intestinal epithelium also contains other cells suchas goblet cells, endocrine cells, and M cells, with the mucus secret-ing goblet cells representing the second most frequent cell type,ranging from 10% in the small intestine and 24% in the distal colon(Madara and Trier, 1987). As the Caco-2 model is composed ofsolely absorptive enterocyte cells, efforts have been made to estab-lish a co-culture with a mucus-producing goblet cell line, theCaco-2/HT29-MTX co-culture (Walter et al., 1996). The additionalpresence of mucin-secreting goblet cells in co-cultures reducesthe tightness of the tight junctions and yields a protective mucusgel covering the whole cell surface as an additional layer, thusincreasingly mimicking physiological conditions (Hilgendorfet al., 2000). However, the expression of the active transporterPgp reduced with enhancing HT29 cell fraction, indicating thatthe similarity to the in vivo situation may be dependent on theseeding ratio, composition of the culture medium and culturingtime (Hilgendorf et al., 2000; Chen et al., 2010; Béduneau et al.,2014). Furthermore, these models still lack the morphologicaland physiological features of the intestinal epithelium, includinginterplay of many processes, such as passive diffusion, carrier-mediated transport, or paracellular transport via tight junctions(Rozehnal et al., 2012). Furthermore, Caco-2 cells have only limitedmetabolism capacity in comparison to human intestinal tissue(Ungell and Karlsson, 2003), and do not allow studying regionalintestinal absorption. In case of food-drug, excipient-drug or otherluminal interactions, the combination of luminal samples within vitro permeability models is essential to investigate these com-plex processes. Consequently, there is an increased interest in

more physiologically relevant in vitro intestinal permeability mod-els, such as ex vivo intestinal tissue segments, especially when aneffect of food on the bioavailability is to be expected.

The use of human intestinal tissue to investigate preclinicalhuman intestinal permeability would be most ideal, which wasalso recognized by the Food and Drug Administration (FDA,2000) who recommended the use of ex vivo intestinal tissue forpermeation studies. For this purpose, usually small pieces ofhuman intestinal tissue are obtained from patients after bariatricsurgery, and permeability studies using human small intestinalor colon tissue has been described (Haslam et al., 2011; Rozehnalet al., 2012; Sjöberg et al., 2013). Differences in regional absorptionwere observed for both highly permeable and hydrophilic com-pounds (Sjöberg et al., 2013), indicating the need to include studiesof regional absorption, but the limited availability of sufficienthealthy human intestinal tissue hampers this type of research.Besides, the condition of the human tissue segments is variabledue to interindividual differences in diet, medication, smokingand environment, which may affect the tissue morphology andresult in up or down regulation of drug transporters andmetabolizing.

Currently, permeability studies with ex vivo (human) intestinaltissue are mostly performed using the Ussing chamber technique,recently described in detail by Sjöberg et al. (2013). A small pieceof intestinal epithelium is mounted between two chambers, bothfilled with a physiological buffer. The Ussing chamber is kept at37 �C, placed in a system on a lab table and continuously oxygen-ated with carbogen (95% O2, 5% CO2). Addition of the compound ofinterest to the mucosal side of the epithelium allows determiningthe apparent permeability of the compound from luminal to sero-sal side. A major restriction of the Ussing model, however, is its rel-atively low-throughput. Hence, there is a clear need for thedevelopment and application of methods that allow a reliable pre-diction of intestinal permeability at medium throughput (rangingup to 96 segments per system per day). For this purpose, intestinaltissue from healthy pigs, having also an omnivorous diet, couldpossibly be used based on the high similarity of the GI tract of pigscompared to humans (Patterson et al., 2008; Walters et al., 2011;Groenen et al., 2012). Furthermore, it is more widely available thanhealthy human intestinal tissue and therefore allows a more sys-tematic approach to investigate (regional) intestinal absorption.

We recently developed a medium throughput system (InTES-Tine™) in which ex vivo intestinal segments are mounted to inves-tigate intestinal permeability. The easy set-up and horizontalmounting of tissue in InTESTine™ (Fig. 1A) enables the system tobe incubated in a humidified oxygenated incubator at 37 �C on arocker platform thereby reducing the unstirred water layer, evap-oration and possible foaming. The horizontal mounting of tissuealso enables the direct contact of test compounds with the epithe-lial tissue. Multiple devices can easily be arrayed (Fig. 1B) allowingmedium throughput analyses of transport and/or secretion acrossthe epithelial tissue. In addition, the device is generated from dis-posable glass material, which reduces nonspecific binding of thecompound of interest and risk of contamination of the mounteddevice.

The aim of the present study was to evaluate whether ex vivointestinal tissue of healthy pigs mounted in InTESTine™ could beused as an alternative for human intestinal tissue for the assess-ment of intestinal permeability.

2. Materials and methods

2.1. Selection of compounds

Currently, the Caco-2 cell model is still the model of choice forhigh throughput prediction of human intestinal absorption

Fig. 1. (A) Schematic representation of measuring intestinal permeability using intestinal tissue; (B) a picture of TNO’s newly developed InTESTine™ system in a 24-setting.

J. Westerhout et al. / European Journal of Pharmaceutical Sciences 63 (2014) 167–177 169

(Haslam et al., 2011; Yazdanian et al., 1998; Yee, 1997). Papp valuesobtained in Caco-2 cells (Artursson and Karlsson, 1991; Lennernäset al., 1996; Yee, 1997; Grès et al., 1998; Yazdanian et al., 1998;Lennernäs, 2007; Parrot and Lave, 2008 and Haslam et al., 2011)were combined with literature data on human intestinal tissue Papp

values (Haslam et al., 2011; Rozehnal et al., 2012; Sjöberg et al.,2013). Literature data were averaged when more than one sourcewas available. A total of 29 compounds for which both Papp valuesare known could be selected (Table 1). From this data set a repre-sentative set of compounds was selected with different Papp values,different chemical characteristics, and preferably radiolabelledavailability, for evaluation in the permeability studies with porcineintestinal tissue and our own batch of Caco-2 cells.

Table 1Overview of compounds with reported Caco-2 Papp values, human intestinal tissue Papp vareported values or given average ± standard deviation when only one reference was availabselected for this study.

Compound Passive/active transport acrossintestinal barriera

Reported Caco-2 Papp values(�10�6 cm/s)b

PEG 4000 Passive 0.15–0.78 (N = 2)Sulfasalazine Active 0.13–0.3 (N = 3)Famotidine Passive 0.40 (N = 1)Digoxin Active 1.60 ± 0.15 (N = 1)Melagatran Active 0.10 (N = 1)Quinidine Active 20.4–23.2 (N = 2)Atenolol Passive 0.17–1.73 (N = 7)Candesartan Passive 0.30 (N = 1)Ranitidine Active 0.49–0.67 (N = 3)Mannitol Passive 0.18–1.17 (N = 5)Cimetidine Active 0.59–3.06Oxprenolol Passive 28.4 (N = 1)Acebutolol Active 0.51–1.60 (N = 2)Testosterone Active 24.9–72.3 (N = 4)Metoprolol Passive 18.0–64.0 (N = 7)Ciprofloxacin Active 2.90 (N = 1)Cetirizine Passive 15.9 (N = 1)Creatinine Passive 1.55 (N = 1)Hydrocortisone Passive 12.2–21.5 (N = 3)Ibuprofen Active 52.5 (N = 1)Propranolol Passive 15.9–47.2 (N = 7)Pindolol Passive 16.7–29.4 (N = 2)Verapamil Active 15.1–44.7 (N = 3)Antipyrine Passive 39.0–82.0 (N = 4)Salicylic acid Active 11.9–22.0 (N = 2)Indomethacin Passive 20.4–31.1 (N = 2)Diazepam Passive 27.0–71.0 (N = 3)D-Glucose Active 17.0–20.0 (N = 2)L-Leucine Active 0.3–15.5 (N = 2)

a Data obtained from drugbank.ca.b Data obtained from Artursson and Karlsson (1991), Lennernäs et al. (1996), Yee (19

(2008) and Haslam et al. (2011).c Data obtained from Haslam et al. (2011), Rozehnal et al. (2012), and Sjöberg et al. (

2.2. Chemicals

[14C]Caffeine (51.1 mCi/mmol), [3H]digoxin (29.8 Ci/mmol),[3H]mannitol (14.2 Ci/mmol), [3H]propranolol (20.9 Ci/mmol),[14C]-testosterone (51.1 mCi/mmol) and [3H]verapamil (81.4 Ci/mmol) were purchased from Perkin Elmer Inc. (Boston, Massachu-setts, United States). [14C]Antipyrine (6.2 mCi/mmol), [3H]cimetidine(80 Ci/mmol), and [14C]creatinine (58.1 mCi/mmol) were purchasedfrom American Radiolabeled Chemicals Inc. (St. Louis, Missouri, Uni-ted States). [3H]Atenolol (3.3 Ci/mmol), and [3H]quinidine (20 Ci/mmol) were purchased from Moravek biochemicals Inc. (Brea, Cali-fornia, United States). [14C]PEG-4000 was obtained from AmershamGE Healthcare Inc. (Buckinghamshire, United Kingdom). [14C]PhIP

lues, and absolute oral bioavailability in humans (data are presented as the range inle). N indicates the number of reference values. Compound names marked in bold are

Reported human intestinal tissue Papp

values (�10�6 cm/s)cHuman absolute oralbioavailability (Fabs) (%) b,c

0.88 ± 0.66 (N = 1) 11.00 ± 0.68 (N = 3) 131.12 ± 0.44 (N = 1) 381.32 ± 0.12 (N = 2) 811.51 ± 0.39 (N = 1) 63.36 ± 0.56 (N = 1) 803.47 ± 0.65 (N = 2) 503.60 (N = 1) 155.50 (N = 1) 505.56 ± 2.01 (N = 1) 167.22 ± 3.48 (N = 2) 649.50 ± 1.73 (N = 1) 9411.0 (N = 1) 4615.1 (N = 1) 10015.3 ± 1.64 (N = 3) 9415.7 (N = 1) 6919.1 (N = 1) 10020.2 ± 7.84 (N = 1) 8022.3 ± 4.22 (N = 1) 9024.7 ± 2.46 (N = 1) 9834.7 ± 2.75 (N = 2) 9435.8 (N = 1) 9836.0 ± 10.0 (N = 1) 9838.3 ± 11.5 (N = 2) 9941.1 ± 16.5 (N = 1) 10048.2 ± 12.4 (N = 1) 10089.0 (N = 1) 100113 ± 49.4 (N = 1) 100140 ± 40.8 (N = 1) 100

97), Grès et al. (1998), Yazdanian et al. (1998), Lennernäs (2007), Parrot and Lave

2013).

170 J. Westerhout et al. / European Journal of Pharmaceutical Sciences 63 (2014) 167–177

(50 mCi/mmol; 2-Amino-1-methyl-6-phenylimidazo[4,5-b]pyri-dine) was obtained from Toronto Research Chemicals Inc. (Toronto,Canada). All radiolabelled compounds were checked on purity beforeuse. Dulbecco’s Modified Eagle Medium (DMEM), Minimum EssentialMedium (MEM), L-glutamine, gentamicin, penicillin, streptomycinand heat-inactivated fetal bovine serum were purchased from Gibco(Paisley, Scotland). Androstenedione, testosterone, and 2b-, 6b-, and11b-hydroxytestosterone were obtained from Steraloids (Newport,RI, United States). Ranitidine and all other chemicals were purchasedfrom Sigma–Aldrich Chemie B.V. (Zwijndrecht, the Netherlands).

2.3. Porcine intestinal tissue

During 1 year, nineteen healthy domestic pigs (Sus scrofadomesticus, both male and female, bodyweight between 15 and25 kg) and thirteen healthy Ellegaard Göttingen minipigs (bothmale and female, between 15 and 20 kg) were used for the collec-tion of intestinal tissue. Before euthanization, pigs had free accessfor food and water. In compliance with the 3Rs principle, animalswere not sacrificed for this purpose, and were only used whendefined healthy as judged by a veterinarian. Intestinal tissue ofdomestic pigs was collected, after which the animals were usedfor educational purposes at the Utrecht University (Utrecht, TheNetherlands). Intestinal tissue of healthy adult minipigs was col-lected from animals after a wash-out period of at least one weekafter finalization of an in vivo study at TNO Triskelion BV (Zeist,The Netherlands). The study protocol to use tissue from so called‘‘left-over’’ animals was approved by the Animal Ethics Committeeof TNO (Ethics Committee permit number 2012.3328, Zeist, TheNetherlands) and all animal procedures were performed in accor-dance with Dutch laws on animal experimentation. Prior to the iso-lation of the intestine, 2000 mL Krebs–Ringer Bicarbonate Buffer(containing 10 mM glucose, 25 mM HEPES, 15 mM sodium bicar-bonate, 2.5 mM calcium chloride, pH 7.4, and saturated with oxy-gen using a 95%/5% O2/CO2 mixture by gassing for 120 min,further indicated as KRB) was divided over different small volumeflasks. After sedation, animals were euthanized and segments ofduodenal tissue (the first 25 cm from the stomach), jejunal tissue(150 cm from the stomach), and ileal tissue (first 50 cm ‘back-wards’ from the ileocecal valve) were excised, flushed with ice-coldKRB buffer, stored in ice-cold KRB, transported to the lab, andimmediately used for ex vivo preparation. Since we used domesticpigs and minipigs of the same weight, it is assumed that thelengths of the different tissue segments are comparable. Once inthe lab, the intestinal tissue segments were prepared based uponexperiences described before (Polentarutti et al., 1999; Haslamet al., 2011; Rozehnal et al., 2012; Sjöberg et al., 2013). In short,intestinal tissue was cut into pieces of 10–20 cm and cut open lon-gitudinally, continuously submerged under ice-cold oxygenatedKRB buffer during further preparation. The serosa and muscularispropria were carefully stripped from the mucosa using a forceps.With a hollow punch, segments with a diameter of 20 mm wereobtained. The basolateral sides of the epithelial segments werethen, without stretching the tissue, mounted on nylon mesh sup-ports. A rubber ring with an internal diameter of 10 mm was thenplaced on the tissue segments, resulting in a tissue exposure areaof 0.785 cm2. Subsequently, the supported segments weremounted in the InTESTine™ system with the mucosal sides facingupwards. Epithelial tissue segments were placed in basolateralcompartments containing 7.5 mL of ice-cold KRB. Next, 1 mL ofthe ice-cold KRB buffer was added to the apical compartment.The complete system was then pre-incubated for 60 min in ahumidified incubator at 37 �C with 90% O2, 5% CO2 for equilibrationof the tissue. Viability and permeability studies were further per-formed as described in Sections 2.6 and 2.7.

2.4. Cell cultures

The human colon carcinoma cell line, Caco-2, was obtainedfrom the German Collection of Microorganisms and Cell cultures(DSMZ ACC 169, Braunschweig, Germany). The mucin-secretingHT29-MTX cell line was kindly provided as a gift from Dr. TheclaLesuffleur (INSERM). Caco-2 cells were cultured in HEPES-bufferedDMEM containing 4.5 g/L glucose, supplemented with 1% (v/v)MEM non-essential amino acids, 6 mM L-glutamine, 50 mg/L gen-tamicin and 10% (v/v), heat-inactivated fetal bovine serum.HT29-MTX cells were cultured in DMEM with 10% (v/v) heat-inac-tivated fetal bovine serum, 2 mM L-glutamine, 1% non-essentialamino acids, 100 U/mL penicillin and 100 lg/mL streptomycin.Cells were grown in 75 cm2 flasks (Corning-Costar, Cambridge,Massachusetts, United States) at approximately 37 �C in a humidi-fied incubator containing a 95% air/5% CO2 mixture.

For permeability studies, Caco-2 cells and a mixture of Caco-2and HT29-MTX cells (approximately 1 � 105 cells/cm2) wereseeded on semi-permeable polycarbonate Transwell� filter inserts(12 well) with a mean pore size of 0.4 lm (Corning-Costar, Cam-bridge, Massachusetts, United States) and exposure area of1.13 cm2. Co-cultures of Caco-2 cells and HT29-MTX cells wereprepared as described by Hilgendorf et al. (2000). In short, prede-termined cell numbers of Caco-2 and HT29-MTX cells were mixedprior to seeding to yield a cell ratio of 9:1 for Caco-2 and HT29-MTX cells, respectively. Co-cultures were maintained under identi-cal conditions as the stem cultures in HT29-MTX medium. The cellson the inserts were cultured in a total volume of 2.3 mL culturemedium per well (0.5 mL apical and 1.8 mL basolateral) at 37 �Cin a humidified incubator containing 95% air/5% CO2 with mediumchanges every 2–3 days. After 18–22 days in culture, the transepi-thelial electrical resistance (TEER) of the monolayers was mea-sured using a Millicell-ERS epithelial volt-ohmmeter. Monolayerswith a TEER value below 300 O cm2 were excluded from the per-meability studies.

2.5. Dynamic gastrointestinal model (TIM)

Biorelevant samples were obtained using TNO’s dynamic com-puter-controlled in vitro gastrointestinal model of the stomachand small intestine (TIM-1; Minekus et al., 1995; Brouwers et al.,2011). This system, consisting of four compartments representingthe stomach, duodenum, jejunum and ileum, offers insight intothe solubility, release and availability for absorption of a com-pound within the GI tract. TIM-1 was used simulating the dynamicphysiological processes of the human adult GI tract such as pH,peristaltics, and secretion of GI enzymes (pepsin, lipase and pan-creatic juice) and bile (fresh pig bile) after the intake of a glass ofwater or a standard high-fat, high-calorie meal according to FDAguidance for industry (FDA, 2002). The experimental settings werethe same as described by Brouwers et al. (2011), except we per-formed our experiments without dosing any compound. Luminaljejunum samples were collected over time up to 6 h resulting inso-called TIM-water and TIM-meal matrices. Luminal sampleswere stored in sub-portions at �20 �C until use.

2.6. Viability and integrity studies with porcine intestinal tissue

Viability of porcine jejunal tissue segments mounted in InTES-Tine™ was assessed by measuring different viability markers dur-ing incubation time. After the equilibration described inSection 2.3, the apical KRB buffer was used to remove the mucusbuild-up and was aspirated off. Afterwards, 1 mL of fresh, pre-warmed KRB buffer (37 �C) containing the compounds of interestwas added to the mucosal compartment (t = 0) and the basolateralKRB buffer was replaced by 7.5 mL fresh, pre-warmed KRB buffer

J. Westerhout et al. / European Journal of Pharmaceutical Sciences 63 (2014) 167–177 171

(37 �C) containing 0.1% BSA. The intestinal segments were incu-bated up to 120 min (unless stated otherwise) at 37 �C in an incu-bator that provides 90% O2, 5% CO2 and a high humidity. During theincubation the InTESTine™ system was gently shaken on a rockerplatform at 60 rpm. To ascertain the integrity of the intestinal tis-sue segments, TEER values were measured before and after theexperiment using a the Millicell-ERS epithelial volt-ohmmeter(Millipore, Billerica, Massachusetts, United States). The integrityof the intestinal barrier was considered unaffected when TEER val-ues remained stable lower than 100 O cm2, in agreement withreported TEER values across porcine jejunal tissue of approxi-mately 70 O cm2 (Moeser et al., 2007). The integrity of the intesti-nal mucosa barrier during incubation was further assessed bymeasuring the leakage of the high molecular weight membraneand paracellular transport marker fluorescein isothiocyanate–dex-tran, MW 4000 (FD4, 50 lM) over time to the basolateral compart-ment, as described in Section 2.9. The intestinal barrier wasconsidered intact when the FD4 leakage was less than 0.5%. Possi-ble cellular membrane damage due to the presence of active prote-ases was determined by measuring the leakage of the stablecytosolic enzyme lactate dehydrogenase (LDH) to both the mucosaland serosal compartment (Section 2.9). The membrane epitheliallayer was considered intact when LDH leakage to the mucosalcompartment was less than 10%. Furthermore, as a marker of epi-thelium barrier integrity, linearity of transport over 3 h wasassessed measuring mucosal to serosal transport of [3H]mannitol(10 lM) (as a marker for passive paracellular transport), and[14C]caffeine (10 lM; as a marker for passive transcellular trans-port). Metabolic activity of jejunal segments was also determined.Porcine intestinal tissue expresses relatively high levels of CYP3AmRNA (Thörn et al., 2011), and testosterone, a known substratefor cytochrome CYP3A, was chosen as model compound. Tissueswere incubated with a mixture of [14C]testosterone and non-labelled testosterone at a total concentration of 250 lM and a totalradioactivity of 0.05 lCi/ml for 60, 120, and 180 min, after whichmedia samples (both apical and basolateral) and intestinal tissuewere collected and stored at �20 �C until analysis. Finally, as amarker for active transport, the transport of [14C]PhIP, a knownsubstrate for breast cancer resistance protein (BCRP), across jejunalsegments was also determined. Porcine intestinal tissue is knownto express a 86% homologue of BCRP (Eisenblatter and Galla,2002). Active transport of [14C]PhIP (5 lM) was determined inthe absence and presence of the BCRP inhibitor elacridar (10 and50 lM; van Herwaarden et al., 2003; Schutte et al., 2006).

2.7. Permeability studies with porcine intestinal tissue

For permeability studies, porcine intestinal tissue segments weprimarily focused on jejunal absorption of 12 compounds, as thejejunum is considered as the most important region in intestinalabsorption for a large set of compounds. In addition, the regionalabsorption of 4 compounds (atenolol, caffeine, cimetidine andmannitol) across duodenum, jejunum, and ileum was also deter-mined (N = up to 4 experiments). After the pre-incubation (asdescribed in Section 2.3), the mucosal KRB buffer was used toremove the mucus build-up and was aspirated off. Afterwards,1 mL of fresh, pre-warmed (37 �C) dose solutions, either containingthe compound of interest (at a concentration of 10 lM, except for[14C] PhIP (5 lM)), with radioactivity levels of approximately2 kBq/mL for the 14C compounds and 20 kBq/mL for the 3H com-pounds) and 50 lM FD4 as a membrane integrity marker wasadded to the apical compartment (t = 0) and the basolateral KRBbuffer was replaced by 7.5 ml fresh, pre-warmed KRB buffer(37 �C). For the permeability studies with biorelevant samples1 mL of pre-warmed (37 �C) TIM-water matrix, or TIM-mealmatrix, spiked with a mixture of [3H]mannitol (10 lM),

[14C]caffeine (10 lM) and FD4 (50 lM), was added to the apicalcompartment (t = 0) and the basolateral KRB buffer was replacedby 7.5 mL pre-warmed KRB buffer (37 �C). The InTESTine™ systemwas then incubated on a rocker platform at 60 rpm at 37 �C in anincubator that provides approximately 90% O2, 5% CO2 and a highhumidity. Transport was measured by collecting samples fromthe apical (100 lL) and basolateral side (2 mL) after various timepoints (e.g. after 45 and 105 min (permeability studies), or after60 and 120 min (biorelevant studies). At the first time point thebasolateral KRB buffer was replaced by 7.5 mL pre-warmed KRBbuffer (37 �C) containing 0.1% BSA, whereas the apical sampleremained 900 lL. After the incubation, the mucosa was dis-mounted, washed with ice-cold KRB in order to remove non-specif-ically bound test compound, dissolved in KOH (1.5 M in 20%ethanol) and the compound accumulated in the mucosa was deter-mined. Samples were analyzed as described in Section 2.9 andrecovery of the test compound was calculated. In each permeabil-ity study, parallel incubations were included to monitor the per-meability of mannitol, caffeine and leakage of FD4. Permeabilitystudies were only approved when the FD4 leakage was less than0.5%, the ratio of Papp caffeine/Papp mannitol was higher than 3,and recovery of the test compounds was more than 95%.

2.8. Permeability studies with Caco-2 and Caco-2/HT29-MTX cells

Caco-2 cell monolayers and co-cultures of Caco-2/HT29MTXcells in Transwell� filter inserts were apically exposed to 500 lLpre-warmed (37 �C) dose solutions, either containing the com-pound of interest (at a concentration of 10 lM, with radioactivitylevels of approximately 2 kBq/mL for the 14C compounds and20 kBq/mL for the 3H compounds) and 50 lM FD4 as an integritymarker. For the permeability studies with biorelevant samples500 lL TIM-water matrix, or TIM-meal matrix were spiked with[3H]mannitol (10 lM), [14C]caffeine (10 lM) and FD4 (50 lM)and dosed to each apical compartment. Each basolateral compart-ment was filled with 1.8 mL KRB. All cultures were incubated on arocker platform device (approximately 60 rpm) in a humidifiedincubator containing 5% CO2 in air at 37 �C. The apparent perme-ability of mannitol and caffeine, and the leakage of FD4 from theapical to basolateral compartment were determined. Transportwas measured by taking 500 lL samples from the basolateral com-partment and 200 lL samples from the apical compartment after30, 60 and 120 min. After sampling, the original volume wasrestored by adding 500 lL or 200 lL of transport medium to thecorresponding compartment. The concentration of [3H]mannitoland [14C]caffeine and the amount of FD4 in culture media wasdetermined as described in Section 2.9. Permeability studies wereonly approved when the FD4 leakage was less than 1.0%.

2.9. Analysis of FD4, LDH, and test compounds

Samples from both the apical and basolateral compartmentscontaining FD4 were analyzed using a FLUOstar OPTIMA fluores-cence spectrometer (BMG Labtech, Ortenberg, Germany) at excita-tion wavelength 490 nm and emission wavelength 520 nm. TotalLDH levels in the exposed area of intestinal tissue were determinedby homogenizing tissue in ice-cold KRB with a Potter–Elvehjem-type PTFE pestle tissue grinder (Braun, Melsungen, Germany) for5 min at 200 rpm. LDH activity in homogenized tissue, as well asin samples from the apical and basolateral compartments wasdetermined using a LDH kit (Sigma Aldrich, Zwijndrecht, The Neth-erlands) and the amount of leakage into the incubation media wasexpressed as a percentage of the total LDH present in the exposedtissue.

Radioactive labeled compounds were measured on a Tri-Carb3100TR Liquid Scintillation counter (LSC, Perkin Elmer, Boston,

172 J. Westerhout et al. / European Journal of Pharmaceutical Sciences 63 (2014) 167–177

Massachusetts, United States) after adding scintillation liquid(Ultima Gold, Perkin Elmer Inc., Boston, Massachusetts, UnitedStates) to samples of the Caco-2, Caco-2/HT29-MTX and InTES-Tine™ experiments.

Ranitidine concentrations were determined using high-pressureliquid chromatography (HPLC) with ultraviolet (UV) detection at230 nm. An Inertsil ODS-3 column (250 mm � 4.6 mm, 5 lm) (GLSciences Inc., Tokyo, Japan) was used with a mixture of demineral-ized water/acetonitrile (93:7, v/v) with 0.1% formic acid (v/v) asmobile phase. Samples were injected directly into the HPLC sys-tem. The flow rate was 1.0 mL/min.

Testosterone and its main metabolites androstenedione, 6b-hydroxytestosterone and 2b-hydroxytestosterone were deter-mined as described before (van de Kerkhof et al., 2006) using HPLCwith UV detection at 242 nm, using 11b-hydroxytestosterone asinternal standard. Media and tissue samples were thawed, and tis-sue segments homogenized with a sonicator. Samples were spiked11b-hydroxytestosterone, extracted with dichloromethane, evapo-rated, and finally dissolved in 130 lL 50% methanol, after which100 lL was injected into the HPLC system. An Inertsil ODS-2 col-umn (250 mm � 4.6 mm, 5 lm) (GL Sciences Inc., Tokyo, Japan)was used at a temperature of 35 �C. A flow rate was 1.0 mL/minwas used with a gradient of demineralized water/methanol(75:25, v/v, solvent A) and demineralized water/methanol/acetoni-trile (30:64:6 (v/v) as mobile phase. Solvent B was increased line-arly from 20% to 31% over 8 min, then 31% to 48% over 12 min, keptat 48% over 10 min, followed from 48% to 58% over 18 min, fol-lowed from 58% to 59% in 1 min, and finally followed from 59%to 100% over 7 min. Protein levels in tissue homogenate were mea-sured using the Bradford method (BioRad, Hercules, CA).

2.10. Calculations

The apparent permeability values (Papp values, cm/s) represent-ing the apical (mucosal) to basolateral (serosal) permeability of thetest compound were calculated according to Eq. (1).

Papp ¼dQ=dt

SA � CApi;0ð1Þ

Here, dQ/dt indicates the appearance rate of the test compound atthe basolateral side over time, SA is the surface area of the exposurearea and CApi,0 is the initial donor concentration of the compound.

Recovery of each compound was determined at the end of theexperiment by calculating the remaining amount of compound inthe apical and basolateral compartment in relation to the initialamount in the donor compartment. In case recovery was below95%, the apparent permeability value total (Papp,total values, cm/s)was determined in which the time dependent accumulation ofthe test compound in the tissue was added to apparent permeabil-ity values as described earlier (Rozehnal et al., 2012).

Fig. 2. Mucosal to serosal transport of [3H]mannitol and [14C]caffeine over 180 minacross porcine jejunal tissue segments. Data represent the mean ± SD (n = 4).

2.11. Data analysis

In order to identify potential correlations between the Papp val-ues determined with porcine intestinal tissue and Papp valuesdetermined with human intestinal tissue, or Papp values deter-mined with Caco-2 cells and Papp values determined with humanintestinal tissue, the Papp values were plotted and different regres-sion analyses of the resulting x,y-scatterplot were performed. As ameasurement of the goodness of fit, the R2 (R2

Pig/Human or R2Caco-2/

Human) is given. To investigate the applicability domain of bothin vitro systems for the prediction of human intestinal absorptionthe compounds for which the regression analysis works best wereidentified.

3. Results

3.1. Viability, barrier integrity, and intestinal metabolism of porcineintestinal mucosa

To study the viability, barrier integrity and intestinal metabo-lism of porcine intestinal segments as mounted in the InTESTine™system, different parameters were analyzed. The transepithelialelectrical resistance (TEER) of porcine jejunal tissue was58 ± 7 O cm2 and remained stable during incubation of 120 minreaching 62 ± 12 O cm2 (n = 4). For comparison, TEER values mea-sured in Caco-2 and Caco-2/HT29-MTX cultures were405.2 ± 18.0 O cm2, and 302.0 ± 11.0 O cm2, respectively. Leakageof FD4 through jejunal tissue segments used in the permeabilitystudies was determined to be on average 0.17 ± 0.11% (N = 32),which remained below the pre-set criterion of 0.5%. For compari-son, FD4 leakage for Caco-2 monolayers and Caco-2/HT29-MTXco-cultures used in permeability studies was determined to be0.55 ± 0.06% and 0.47 ± 0.02%, respectively. Leakage of intracellularLDH from the jejunal tissue segments was found to be in total6.8 ± 1.1% after 120 min, of which 3.3 ± 0.5% was found in themucosal compartment and 3.5 ± 0.8% was found in the basolateralcompartment, which remained below the pre-set criterion of LDHleakage < 10%. Besides the above mentioned epithelium integritymarkers, markers for permeability were also evaluated. The muco-sal to the serosal transport across porcine jejunal tissue segmentsof the known non-carrier mediated transported compounds, man-nitol and caffeine, showed a lag time of approximately 45–60 min,where after transport remained linear between t = 60 andt = 180 min (R2 > 0.99), indicating no loss of intestinal epithelialintegrity (Fig. 2).

Porcine jejunal segments were also determined to be metabol-ically active using a mixture of both radiolabelled and non-radiola-belled testosterone. Since the metabolites of radiolabelled [14C]-testosterone are also radiolabelled, when the Papp,total is calculatedon the basis of radioactivity, it also includes the metabolites. Incontrast, the Papp,total value for non-labelled testosterone was calcu-lated upon the transport of only the parent compound. As a result,the Papp,total value of radiolabelled and non-radiolabelled testoster-one differed (22.0 ± 3.74 versus 14.2 ± 2.96 � 10�6 cm/s, respec-tively), indicating metabolic activity. Formation of the mainmetabolites, androstenedione and 6b-hydroxytestosterone wasfound to be linear up to t = 240 min (R2 = 0.99), indicating no lossof metabolic capacity (Fig. 3).

Finally, active transport processes were studied in porcineintestinal tissue. The mucosal to serosal transport of the knownBCRP substrate [14C]PhIP was clearly affected by the co-exposurewith the BCRP inhibitor elacridar, resulting in an slightly enhanced

Fig. 3. Sum of apical, basolateral and intracellular formation of testosteronemetabolites (androstenendione and 6b-testosterone) over 240 min in porcinejejunal tissue. Data represent the mean ± SD (n = 4) per mg protein.

Fig. 4. Apparent permeability (Papp and Papp,total) of the known BCRP substrate[14C]PhIP in the presence of the BCRP inhibitor elacridar (0, 10 and 50 lM) asdetermined in porcine jejunal tissue. Data represent the mean ± SD (n = 4).

J. Westerhout et al. / European Journal of Pharmaceutical Sciences 63 (2014) 167–177 173

Papp, but clearly enhanced intracellular accumulation, as indicatedby the Papp,total (Fig. 4). All together, we concluded that porcineintestinal tissue mounted in InTESTine™ was considered viablewith an intact and functional intestinal barrier within the time per-iod tested (120 min).

3.2. Permeability studies

The permeability of a selected set of 12 compounds with differ-ent chemical characteristics (Table 2) was measured across porcinejejunal tissue mounted in the InTESTine™ system and Caco-2 cellmonolayers in Transwells�. The Papp values as determined withintestinal tissue from either pigs or minipigs did not differ signifi-cantly for the majority of compounds and were therefore averaged.However, only for atenolol and mannitol (both transported para-cellularly) a significant lower transport (p < 0.05) was observed inminipigs, when compared to pigs (Papp values 1.25 ± 0.45 and2.92 ± 0.20 � 10�6 cm/s, respectively for atenolol and 2.89 ± 0.42and 4.54 ± 0.39 � 10�6 cm/s, respectively for mannitol). For ateno-lol and mannitol we have therefore used the permeability dataobtained with both pigs and minipigs separately for further calcu-lations. Papp values as determined with porcine intestinal tissuehad a smaller range as compared to Caco-2 cells and human intes-tinal tissue mounted in the Ussing chamber system, ranging from0.31 ± 0.12 � 10�6 cm/s (digoxin) to 18.0 ± 1.35 � 10�6 cm/s (caf-feine) (Table 2). Recovery of all compounds based upon mediaalone were > 95%, except for the highly lipophilic compounds pro-panolol and verapamil. These compounds appeared to accumulatein the intestinal tissue, which was reflected in very low Papp values.Drug accumulation in the tissue was therefore determined by firstremoving the mucus build-up and dissolving the segments in KOH.By including the drug accumulation as suggested by Rozehnal et al.

(2012), the Papp,total value was 25.1 ± 6.36 � 10�6 cm/s for propran-olol and 37.4 ± 9.09 � 10�6 cm/s for verapamil, comparable to datafound with human intestinal tissue (Table 2).

Literature data of Papp values as determined with human intesti-nal tissue mounted in the Ussing chamber system were compared toPapp values as determined with porcine intestinal tissue mounted inthe InTESTine™ system and literature Papp values as determinedwith Caco-2 cells. Regression analysis of the data showed that thecorrelation between human intestinal tissue Papp values and porcineintestinal tissue Papp values or Caco-2 Papp values was best describedby means of a power function. In general, a reasonable correlationwas found between the Caco-2 Papp values and the Papp values deter-mined for human intestinal tissue mounted in the Ussing chambersystem (Human Papp ¼ 4:49� Caco2 P0:57

app , R2 = 0.61, for the 29 com-pounds; Human Papp ¼ 4:30� Caco2 P0:67

app , R2 = 0.71 for own data ofselected compounds). The correlation between Papp,total values asdetermined for the 12 selected compounds with porcine intestinaltissue and human intestinal tissue was slightly better with an R2 of0.75 (Human Papp ¼ 2:72� Porcine P0:74

app ). However, due to the smalldataset the goodness of fit is highly sensitive to outliers. In general,when directly comparing the human intestine Papp values to the por-cine intestine Papp values or Caco-2 Papp values (i.e., assuming a linearcorrelation) for the selected compounds, it appears that porcineintestinal tissue Papp values are more comparable to human Papp val-ues (9 out of 12 are within 3-fold difference), compared to Caco-2Papp values (4 out of 12 are within 3-fold difference). It can be seenthat for certain paracellularly transported compounds (atenolol,ranitidine, mannitol and cimetidine) porcine Papp values are within3-fold difference of human Papp values, whereas the Caco-2 Papp val-ues are beyond 3-fold difference. For the transcellularly transportedcompounds, it appears that the Caco-2 Papp values are equally com-parable to human Papp values as porcine Papp values. However, sinceCaco-2 monolayer lacks villus structures, thereby limiting the effec-tive surface area when compared to ex vivo intestinal tissue, it is notjustified to expect a linear correlation between Caco-2 Papp valuesand human Papp values.

To evaluate whether porcine intestinal tissue can be used tostudy the window of absorption, the regional permeability of aten-olol, mannitol, cimetidine, and caffeine across duodenum, jejunum,and ileum was investigated. As can be seen in Table 3 for thehydrophilic compounds the small intestine paracellular (mannitol)and transcellular (cimetidine and caffeine) permeability increasessignificantly (p < 0.05) from the duodenum to the ileum. For themoderately lipophilic compound atenolol, the paracellular perme-ability remained stable from duodenum to ileum. Leakage of FD4 inall segments remained below 0.5%.

3.3. Permeability studies in presence of biorelevant matrices

Exposure of Caco-2 cells to TIM-water and TIM-meal matricesclearly enhanced the permeability of mannitol and leakage ofFD4 as compared to KRB, indicating loss of intestinal barrier integ-rity due to the cytotoxic effects of (biorelevant components in the)luminal TIM matrices (Fig. 5A). Co-culturing with the mucussecretingCaco-2/HT29-MTX cell line did not reduce the cytotoxiceffect of TIM-water and TIM-meal matrices on the barrier integrity,as indicated by similar results as compared to those of the singleculture Caco-2 cells (Fig. 5B).

Exposure of ex vivo porcine jejunal tissue segments to TIM-water and TIM-meal matrices, however, did not affect the perme-ability of mannitol and leakage of FD4, indicating no (cytotoxic)effect of TIM samples on the barrier integrity of porcine intestinaltissue (Fig. 5C). Exposure of all three systems to TIM-water andTIM-meal matrices reduced the permeability of caffeine, indicatingan inhibitory effect of (components in) TIM matrices on the trans-cellular permeability across the apical membrane.

Table 2Apparent permeability (Papp) of selected compounds as determined in porcine intestinal tissue and own batch of Caco-2 cells, compared to reported Papp values in humanintestinal tissue. Porcine en Caco-2 Papp data represent means ± SEM from at least three independent experiments (N indicates the number of donors) performed at least intriplicate (n indicates the number of replicates).

Compound BCSa MW logP Papp/Papp,total value (�10�6 cm/s), from apical to basolateral compartment, based upon

Porcine intestinal tissueb Caco-2b Human intestinal tissuec

PEG 4000 III 3500–4500 �1.4 1.52 ± 0.14 (N = 3) 0.84 ± 0.48 (n = 5) 0.88 ± 0.66Digoxin III 780.94 2.37 0.31 ± 0.12 (N = 4) 0.40 ± 0.02 (n = 3) 1.32 ± 0.12Quinidine I 324.42 2.51 0.54 ± 0.13 (N = 3) 21.8 ± 1.40d 3.36 ± 0.56Atenolol III 266.34 0.43 1.25 ± 0.45 (N = 4 minipigs) 0.38 ± 0.09 (n = 5) 3.47 ± 0.65

2.92 ± 0.20 (N = 4 pigs)Ranitidine III 314.4 0.98 2.22 ± 0.30 (N = 3) 0.32 ± 0.09 (n = 3) 5.50Mannitol III 182.17 �3.73 2.89 ± 0.42 (N = 11 minipigs) 1.04 ± 0.06 (n = 5) 5.56 ± 2.01

4.54 ± 0.39 (N = 14 pigs)Cimetidine III 252.34 �0.11 3.98 ± 0.61 (N = 6) 1.90 ± 0.11 (n = 3) 7.22 ± 3.48Testosterone I 288.42 3.37 0.76 ± 0.22/ 48.4 ± 9.78d 15.1

14.2 ± 2.96 (n = 4)Creatinine III 113.12 0.22 3.18 ± 0.56 (N = 3) 1.55d 20.2 ± 7.84Propranolol I 259.34 2.58 0.42 ± 0.12/ 18.0 ± 0.80 (n = 5) 34.7 ± 2.75

25.1 ± 6.36 (N = 6)Verapamil I 454.6 5.04 2.37 ± 0.53/ 17.2 ± 3.63 (n = 5) 36.0 ± 10.0

37.4 ± 9.09 (N = 9)Antipyrine I 188.23 1.22 17.5 ± 2.60 (N = 7) 29.7 ± 1.73 (n = 5) 38.3 ± 11.5Caffeine I 194.19 �0.55 17.3 ± 1.18 (N = 25) 26.3 ± 1.56 (n = 5) N.A.

N.A. = Not availablea BCS class indication based upon Papp values in Caco-2 data (Amidon et al., 1995).b Porcine intestinal tissue Papp values and Caco-2 Papp values generated as described in methods section.c Human intestinal tissue Papp values obtained from Haslam et al., 2011; Rozehnal et al., 2012; Sjöberg et al., 2013. Data were averaged when more than one source was

available.d Literature data.

174 J. Westerhout et al. / European Journal of Pharmaceutical Sciences 63 (2014) 167–177

4. Discussion

In the current study we evaluated the applicability of porcineintestinal tissue as an alternative for human intestinal tissue forpermeability studies, since porcine intestinal tissue is highly com-parable to human intestinal tissue based on the omnivore charac-teristics of both humans and pigs (Patterson et al., 2008; Walterset al., 2011; Groenen et al., 2012). Especially for gastrointestinalprocesses (e.g. transit time, pH, and surface area) pigs appear tobe the most analogous to humans when compared to other preclin-ical animal species (DeSesso and Williams, 2008). Moreover, as thedifferences between domestic pigs and minipigs are more relatedto growth rates and size at maturation rather than anatomical,physiological, and functional differences (Swindle et al., 2012),healthy intestinal tissue from both pigs and minipigs can be used.Currently, only limited information is available on the differencesin overall GI expression levels of active transport systems and CYPsof domestic pigs and minipigs (Sjögren et al., 2014). Furthermore,information on the regional GI transporter and CYP expression pat-terns in domestic pigs and minipigs is lacking in literature and istherefore currently under investigation.

With respect to the intestinal permeability data, we observeddifferences for the paracellularly transported compounds atenololand mannitol, indicating that the paracellular route is somewhatrestricted in minipigs, compared to the domestic pigs. This couldbe caused by a difference in age, feeding and/or health status andremains to be investigated. A major advantage of using porcineintestinal tissue for permeability studies is the possibility to inves-tigate regional differences, since the availability of duodenal, jeju-nal, ileal or colon tissue is not restricted, as is the case for humanintestinal tissue. Yet, a major disadvantage of using ex vivo (eitherhuman or animal) tissue for permeability studies is the relativelyshort viability time of the tissue segments. Following our protocols,using KRB as incubation buffer, porcine intestinal tissue mountedin the InTESTine™ system remained viable with an intact intestinalbarrier function for at least up to 120 min, as indicated by stableTEER values, low FD4 and LDH leakage, linear passive transport

of compounds over time, linear rate of testosterone metabolism,and functional active transport. We have shown that testosteronemetabolism is linear up to t = 240. We found that androstenedionewas the main metabolite formed, whereas the formation of 6b-hydroxytestosterone was approximately 7% of the total. Otherknown metabolites 2b-hydroxytestosterone and 16a-hydroxytes-tosterone (van de Kerkhof et al., 2006) could not be detected.

The viability of the porcine intestinal tissue in InTESTine™ wasfurther confirmed by metabolic activity and active transport. It haspreviously been shown for Caco-2 cells that inhibition of BCRP byexposure to 10 lM elacridar (GF120918) in the apical medium,resulted in a 4-fold increase in Papp values of PhIP to the serosalcompartment (Schutte et al., 2006). Using porcine intestinal tissue,a 2-fold increase in Papp values of PhIP in the presence of 50 lMelacridar was detected, with no difference in Papp values measuredat 10 lM elacridar. But when intracellular accumulation of PhIPwas included in the calculation of Papp,total, a clear inhibition ofBCRP transport of PhIP to the luminal side was observed resultingin increased Papp,total values up to 2-fold at 10 lM elacridar and upto 3.5-fold at 50 lM elacridar. This indicates that the functionalityof active transport at the intestinal barrier cannot necessarily beproven by an increase in Papp values after exposure to specificinhibitors, since the intestinal tissue segments consist of multiplelayers of cells. It is therefore suggested to use Papp,total values as atrue indication of active transport functionality.

The use of porcine intestinal tissue allowed us to investigateregional differences in intestinal permeability within donors. As afirst demonstrator, we determined the permeability of atenolol,mannitol, cimetidine, and caffeine from duodenum to ileum. Forthe moderately lipophilic compound atenolol, the permeabilityremained stable from duodenum to ileum. For the moderatelyhydrophilic compounds cimetidine and caffeine and the stronghydrophilic compound mannitol we observed a significant(p < 0.05) increase in permeability from duodenum to ileum. Cur-rently, only limited data is reported for differences in regionalabsorption in humans, and data indicate that human intestinalregional absorption is highly dependent on the compound used

Table 3Apparent permeability (Papp) of atenolol, mannitol, cimetidine and caffeine across porcine duodenal, jejunal, and ileal tissue mounted in InTESTine™. Data represent means ± SEMfrom at least two independent experiments (two donors) performed at least in triplicate. N indicates the number of donors.

Compound Papp value (�10�6 cm/s), from apical to basolateral compartment, determined for

Duodenum Jejunum Ileum

Atenolol 2.24 ± 0.34 (N = 2) 2.41 ± 1.33 (N = 2) 2.71 ± 0.55 (N = 2)Mannitol 2.75 ± 0.35 (N = 4) 4.00 ± 0.97 (N = 4) 4.71 ± 0.48 (N = 4)�

Cimetidine 2.54 ± 0.48 (N = 4) 4.07 ± 0.95 (N = 4) 5.89 ± 0.47 (N = 4)�

Caffeine 10.5 ± 1.65 (N = 4) 15.4 ± 2.19 (N = 4) 17.0 ± 1.64 (N = 4)�

� Significantly different from duodenum (p < 0.05).

Fig. 5. Apparent permeability values (Papp) of mannitol and caffeine, and leakage of FD4 (%) from spiked KRB, TIM-water and TIM-meal matrices across (A) Caco-2 cellmonolayers, (B) Caco-2/HT29-MTX co-cultures and (C) ex vivo porcine jejunal tissue segments. The leakage of FD4 from spiked TIM-meal samples across Caco-2/HT29-MTXco-cultures could not be determined due to quenching of fluorescence. FD4 leakage from spiked 4-fold diluted TIM-meal matrices across Caco-2/HT29-MTX indicated highsimilarities to leakage across Caco-2 cells (results not shown).

J. Westerhout et al. / European Journal of Pharmaceutical Sciences 63 (2014) 167–177 175

(Sjöberg et al., 2013). We are currently investigating regional dif-ferences in intestinal permeability for more compounds and alsoinclude different parts of the colon, to characterize the windowof absorption. Further research is needed to elucidate whether dif-ferences in permeability are due to e.g. differences in villi-length,mucosal thickness and/or differences in passive and/or activetransport.

In general, permeability studies are performed with the com-pound of interest dissolved in KRB buffer, rather than in the formu-lation that is used for dosing in humans. Furthermore, some of thedrugs are intended to be taken orally with a glass of water, whileothers are best dosed during a meal. As a result, the oral bioavail-ability can be influenced by the formulation of the drug or by thepresence (or lack thereof) of food. In order to be able to investigatethe effect of food–drug or excipient–drug interactions on intestinalabsorption in vitro, biorelevant luminal samples (TIM-water andTIM-meal) were combined with Caco-2 cells, Caco-2/HT29-MTXco-cultures and porcine intestinal tissue segments mounted inInTESTine™. The loss of barrier integrity due to (components inthe) luminal TIM samples to Caco-2 cells and Caco-2/HT29-MTXco-cultures was in line with previous findings by Ingels et al.(2002), who used FaSSIF and FeSSIF buffers on Caco-2 cells. Highertoxicity was found after exposure to TIM-meal samples as com-pared to TIM-water samples, probably caused by the higheramount of bile and pancreatic enzymes and/or food componentsin the TIM-meal samples. Dilution of the TIM samples was requiredto prevent cytotoxicity in Caco-2 cells and Caco-2/HT29-MTX co-cultures. However, this reduces the biological relevance of thesamples drastically. Patel and colleagues (2006) later modifiedthe composition of the FaSSIF and FeSSIF buffers for increased com-patibility with Caco-2 cells. In contrast to the Caco-2 cells, porcineintestinal tissue was much less affected by biorelevant matrices.

Moreover, a direct exposure of porcine jejunal tissue segments toundiluted biorelevant luminal TIM samples, even after the diges-tion of a high fat meal, did not affect the integrity of the intestinalbarrier. The mucus layer of porcine intestinal tissue is probablyproviding a protective layer against luminal material. Previousstudies by Hilgendorf et al. (2000), Mahler et al. (2009) andWalter et al. (1996) have indicated that the mucus layer in Caco-2/HT29-MTX co-cultures ranges only from 2 to 10 lm. In contrast,Varum et al. (2012) indicated that the mucus layer in porcine ileumis over 50 lm thick. While we have performed no measurements ofthe mucus thickness, we also observed a substantial mucus pro-duction in the porcine intestinal tissue segments during incuba-tion. The variety of cells and the presence of the mucus layer inthe ex vivo intestinal tissue segments make them a physiologicallymore relevant in vitro model for the human intestinal barrier ascompared to single-cell cultured Caco-2 cells.

Overall, the Papp values of the selected set of compounds mea-sured with porcine jejunal tissue or Caco-2 cells show comparablecorrelations with Papp values measured with human jejunal tissue.It can be hypothesized that the permeability of especially paracel-lularly transported compounds across tissue of porcine jejunal tis-sue is more directly comparable to the permeability across humanjejunal tissue than is the permeability across Caco-2 cell monolay-ers, Ultimately, it is the goal to be able to predict human oral bio-availability on the basis of in vitro absorption studies. While goodcorrelations between the permeability data as determined witheither human intestinal tissue mounted in the Ussing chamber sys-tem or Caco-2 cells and human oral bioavailability values can beobtained for the 29 compounds mentioned in Table 1 (R2 = 0.76and 0.71, respectively, results not shown), our dataset of 12 com-pounds is too small to identify the correlation between porcineintestinal tissue Papp values and human oral bioavailability data.

176 J. Westerhout et al. / European Journal of Pharmaceutical Sciences 63 (2014) 167–177

It is clear that with a larger data set and better knowledge of theeffect of protein expression levels of membrane embedded trans-porters and expression levels of the metabolizing enzymes on per-meability, these processes can be scaled to the human situation inorder to improve the translation of the obtained Papp data andeventually to human oral bioavailability.

Currently, we are expanding our data set and investigating thedifferences in the absolute expression levels of different transport-ers, such as P-glycoprotein, BCRP and multidrug resistance-associ-ated proteins, as well as metabolizing enzymes, such ascytochrome P450, UDP-glucuronyltransferase and glutathione S-transferases, between Caco-2 cells, porcine intestinal tissue andhuman intestinal tissue. With the use of physiologically-basedpharmacokinetic models this information will be integrated toaim for a prediction of human oral bioavailability for a broadergroup of compounds based on in vitro absorption studies.

5. Conclusions

The results presented indicate that porcine jejunal intestinal tis-sue mounted in the InTESTine™ system provides a good alternativefor human intestinal tissue in Ussing chambers, to assess the intes-tinal permeability in human. Moreover, we have shown for the firsttime a unique combination of in vitro tools studying the permeabil-ity of compounds in the presence of undiluted biorelevant samples,without loss of integrity. It should be noted that further character-izing of porcine intestinal tissue with respect to metabolism andactive transport capacity is needed, as well as testing with addi-tional drugs, food and excipient matrices in order to better under-stand the processes of absorption in the pig to increase thepredictability of the fraction absorbed in man.

Acknowledgement

We acknowledge Bas Swildens and Louis van den Boom fromthe University of Utrecht (The Netherlands) and Marcel Wijnandsfrom TNO Triskelion BV (The Netherlands) for providing us porcineintestinal tissue. We especially thank Angelique Speulman-Saat,Harm Jansen, Irene Nooijen Freek Schrander, and Wendy Borstfor their excellent technical assistance with the Caco-2 and porcineintestinal tissue experiments. The work described was funded bythe Dutch Government: Ministry of Economic Affairs, the Provenceand Municipality of Utrecht. The work is performed as part of theSLiM project (PID101063) in close collaboration with the Univer-sity of Applied Sciences Utrecht (The Netherlands) and the Univer-sity Utrecht (The Netherlands)

References

Amidon, G.L., Lennernäs, H., Shah, V.P., Crison, J.R., 1995. A theoretical basis for abiopharmaceutic drug classification: the correlation of in vitro drug productdissolution and in vivo bioavailability. Pharm. Res. 12, 413–420.

Artursson, P., Karlsson, J., 1991. Correlation between oral drug absorption inhumans and apparent drug permeability coefficients in human intestinalepithelial (Caco-2) cells. Biochem. Biophys. Res. Com. 175, 880–885.

Béduneau, A., Tempesta, C., Fimbel, S., Pellequer, Y., Jannin, V., Demarne, F.,Lamprecht, A., 2014. A tunable Caco-2/HT29-MTX co-culture model mimickingvariable permeabilities of the human intestine obtained by an original seedingprocedure. Eur. J. Pharmaceut. Biopharmaceut.. http://dx.doi.org/10.1016/j.ejpb.2014.03.017.

Brouwers, J., Tack, J., Augustijns, P., 2007. In vitro behavior of a phosphate esterprodrug of amprenavir in human intestinal fluids and in the Caco-2 system:illustration of intraluminal supersaturation. Int. J. Pharmaceut. 336, 302–309.

Brouwers, J., Anneveld, B., Goudappel, G.J., Duchateau, G., Annaert, P., Augustijns, P.,Zeijdner, E., 2011. Food-dependent disintegration of immediate releasefosamprenavir tablets: In vitro evaluation using magnetic resonance imagingand a dynamic gastrointestinal system. Eur. J. Pharm. Biopharm. 77, 313–319.

Chen, X.M., Elisia, I., Kitts, D.D., 2010. Defining conditions for the co-culture of Caco-2 and HT29-MTX cells using Taguchi design. J. Pharmacol. Toxicol. Meth. 61,334–342.

Dahan, A., Miller, J.M., Hoffman, A., Amidon, G.E., Amidon, G.L., 2010. The solubility–permeability interplay in using cyclodextrins as pharmaceutical solubilizers:mechanistic modeling and application to progesterone. J. Pharm. Sci. 99, 2739–2749.

DeSesso, J.M., Williams, A.L., 2008. Contrasting the gastrointestinal tracts of mammals:factors that influence absorption. Ann. Rev. Med. Chem. 43, 353–371.

Eisenblatter, T., Galla, H.J., 2002. A new multidrug resistance protein at the blood–brain barrier. Biochem. Biophys. Res. Commun. 293, 1273–1278.

FDA Guidance for Industry, 2000. Waiver of In Vivo Bioavailability andBioequivalence Studies for Immediate Release Solid Oral Dosage Forms basedon a Biopharmaceutics Classification System. <http://www.fda.gov/cder/guidance/index.htm>.

FDA Guidance for Industry, 2002. Food Effect Bioavailability and Fed BioequivalenceStudies. <http://www.fda.gov/cder/guidance/index.htm>.

Fischer, S.M., Buckley, S.T., Kirchmeyer, W., Fricker, G., Brandl, M., 2012. Applicationof simulated intestinal fluid on the phospholipid vesicle-based drug permeationassay. Int. J. Pharmaceut. 422, 52–58.

Grès, M.-C., Julian, B., Bourrié, M., Meunier, V., Roques, C., Berger, M., Boulenc, X.,Berger, Y., Fabre, G., 1998. Correlation between oral drug absorption in humans,and apparent drug permeability in TC-7 cells, a human epithelial intestinal cellline: comparison with the parental Caco-2 cell line. Pharm. Res. 15, 726–733.

Groenen, M.A.M., Archibald, A.L., Uenishi, H., Tuggle, C.K., Takeuchi, Y., Rothschild,M.F., et al., 2012. Analyses of pig genomes provide insight into demography andevolution. Nature 491, 393–398.

Haslam, I.S., O’Reilly, D.A., Sherlock, D.J., Kauser, A., Womack, C., Coleman, T., 2011.Pancreatoduodenectomy as a source of human small intestine for Ussingchamber investigations and comparative studies with rat tissue. Biopharm.Drug Dispos. 32, 210–221.

Hilgendorf, C., Spahn-Langguth, H., Regårdh, C.G., Lipka, E., Amidon, G.L., Langguth,P., 2000. Caco-2 versus Caco-2/HT29-MTX co-cultured cell lines: permeabilitiesvia diffusion, inside- and outside-directed carrier-mediated transport. J. Pharm.Sci. 89, 63–75.

Ingels, F., Deferme, S., Destexhe, D., Oth, M., Van den Mooter, G., Augustijns, P., 2002.Simulated intestinal fluid as transport medium in the Caco-2 cell culture model.Int. J. Pharm. 31, 183–192.

Kataoka, M., Itsubata, S., Masaoka, Y., Sakuma, S., Yamashita, S., 2011. In vitrodissolution/permeation system to predict the oral absorption of poorly water-soluble drugs: effect of food and dose strength on it. Biol. Pharm. Bull. 34, 401–407.

Lennernäs, H., 2007. Animal data: the contributions of the Ussing Chamber andperfusion systems to predicting human oral drug delivery in vivo. Adv. Drug Del.Rev. 59, 1103–1120.

Lennernäs, H., Palm, K., Fagerholm, U., Artursson, P., 1996. Comparison betweenactive and passive drug transport in human intestinal epithelial (Caco-2) cellsin vitro and human jejunum in vivo. Int. J. Pharm. 177, 103–115.

Lentz, K.A., 2008. Current methods for predicting human food effect. AAPS J. 10,282–288.

Madara, J.L., Trier, J.S., 1987. Functional morphology of the mucosa of the smallintestine. In: Johnson, R. (Ed.), Physiology of the Gastrointestinal Tract, seconded. Raven Press, New York, pp. 1209–1249.

Mahler, G.J., Shuler, M.L., Glahn, R.P., 2009. Characterization of Caco-2 and HT29-MTX cocultures in an in vitro digestion/cell culture model used to predict ironbioavailability. J. Nutr. Biochem. 20, 494–502.

Minekus, M., Marteau, P., Havenaar, R., Huis in’t Veld, J.H.J., 1995. A multicompartmental dynamic computer-controlled model simulating the stomachand small intestine. Altern. Lab. Anim. 23, 197–209.

Moeser, A.J., Vander Klok, C., Ryan, K.A., Wooten, J.G., Little, D., Cook, V.L., Blikslager,A.T., 2007. Stress signaling pathways activated by weaning mediate intestinaldysfunction in the pig. Am. J. Physiol. Gastrointest. Liver Physiol. 292, G173–G181.

Musther, H., Olivares-Morales, A., Hatley, O.J.D., Liu, B., Rostami Hodjegan, A., 2014.Animal versus human oral drug bioavailability: do they correlate? Eur. J. Pharm.Sci. 57, 280–291.

Parrot, N., Lave, T., 2008. Application of physiologically based absorption models indrug discovery and development. Mol. Pharm. 5, 760–775.

Patel, N., Forbes, B., Eskola, S., Murray, J., 2006. Use of simulated intestinal fluidswith Caco-2 cells and rat ileum. Drug Dev. Ind. Pharm. 32, 151–161.

Patterson, J.K., Gen Lei, X., Miller, D.D., 2008. The pig as an experimental model forelucidating the mechanisms governing dietary influence on mineral absorption.Exp. Biol. Med. 233, 651–664.

Polentarutti, B.I., Peterson, A.L., Sjöberg, A.K., Anderberg, E.K., Utter, L.M., Ungell,A.L., 1999. Evaluation of viability of excised rat intestinal segments in the usingchamber; investigation of morphology, electrical parameters, and permeabilitycharacteristics. Pharm. Res. 16, 446–454.

Rozehnal, V., Nakai, D., Hoepner, U., Fischer, T., Kamiyama, E., Takahasi, M., Yasuda,S., Mueller, J., 2012. Human small intestinal and colonic tissue mounted in theUssing chamber as a tool for characterizing the intestinal absorption of drugs.Eur. J. Pharm. Sci. 46, 367–373.

Schutte, M.A., Freidig, A.P., van de Sandt, J.A.M., Alink, G.M., Rietjens, I.M.C.M.,Groten, J.P., 2006. An in vitro and in silico study on the flavonoid-mediatedmodulation of the transport of 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) through Caco-2 monolayers. Tox. Appl. Pharm. 217, 204–215.

Sjöberg, Å., Lutz, M., Tannergren, C., Wingolf, C., Borde, A., Ungell, A.-L., 2013.Comprehensive study on regional human intestinal permeability and predictionof fraction absorbed of drugs using the Ussing chamber technique. Eur. J. Pharm.Sci. 48, 166–180.

J. Westerhout et al. / European Journal of Pharmaceutical Sciences 63 (2014) 167–177 177

Sjögren, E., Abrahamsson, B., Augustijns, P., Becker, D., Bolger, M.B., Brewster, M.,Brouwers, J., Flanagan, T., Harwood, M., Heinen, C., Holm, R., Juretschke, H.P.,Kubbinga, M., Lindahl, A., Lukacova, V., Münster, U., Neuhoff, S., Nguyen, M.A.,van Peer, A., Reppas, C., Hodjegan, A.R., Tannergren, C., Weitschies, W., Wilson,C., Zane, P., Lennernäs, H., Langguth, P., 2014. In vivo methods for drugabsorption – comparative physiologies, model selection, correlations within vitro methods (IVIVC), and applications for formulation/API/excipientcharacterization including food effects. Eur. J. Pharm. Sci. 57, 99–151.

Swindle, M.M., Makin, A., Herron, A.J., Clubb, F.J., Frazier, K.S., 2012. Swine as modelsin biomedical research and toxicology testing. Vet. Pathol. 49, 344–356.

Thörn, H.A., Lundahl, A., Schrickx, J.A., Dickinson, P.A., Lennernäs, H., 2011. Drugmetabolism of CYP3A4, CYP2C9 and CYP2D6 substrates in pigs and humans.Eur. J. Pharm. Sci. 43, 89–98.

Ungell, A.L., Karlsson, J., 2003. Cell cultures in drug discovery: an industrialperspective. In: van de Waterbeemd, H., Lennernäs, H., Artursson, P. (Eds.), DrugBioavailability: Estimation of Solubility, permeability, Absorption andBioavailbility. Wiley-VCH, Weinheim, pp. 90–131.

van de Kerkhof, E.G., Ungell, A.L., Sjöberg, A.K., de Jager, M.H., Hilgendorf, C., deGraaf, I.A., Groothuis, G.M., 2006. Innovative methods to study human intestinal

drug metabolism in vitro: precision-cut slices compared with ussing chamberpreparations. Drug Metab. Dispos. 34, 1893–1902.

van Herwaarden, A.E., Jonker, J.W., Wagenaar, E., Brinkhuis, R.F., Schellens, J.H.M.,Beijnen, J.H., Schinkel, A.H., 2003. The breast cancer resistance protein (Bcrp1/Abcg2) restricts exposure to the dietary carcinogen 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine. Cancer Res. 1, 6447–6452.

Varum, F.J., Veiga, F., Sousa, J.S., Basit, A.W., 2012. Mucus thickness in thegastrointestinal tract of laboratory animals. J. Pharm. Pharmacol. 64, 218–227.

Walter, E., Janich, S., Roessler, B.J., Hilfinger, J.M., Amidon, G.L., 1996. HT29-MTX/Caco-2 cocultures as an in vitro model for the intestinal epithelium: in vitro –in vivo correlation with permeability data from rats and humans. J. Pharm. Sci.4, 1070–1076.

Walters, E.M., Agca, Y., Ganjam, V., Evans, T., 2011. Animal models got you puzzled?:Think pig. Ann. N. Y. Acad. Sci. 1245, 63–64.

Yazdanian, M., Glynn, S.L., Wright, J.L., Hawi, A., 1998. Correlating partitioning andCaco-2 cell permeability of structurally diverse small molecular weightcompounds. Pharm. Res. 15, 1490–1494.

Yee, S., 1997. In vitro permeability across Caco-2 cells (colonic) can predict in vivo(small intestinal) absorption in man – fact or myth. Pharm. Res. 14, 763–766.