All layers matter: Innovative three-dimensional epithelium-stroma-endothelium intestinal model for reliable permeability outcomes

Abstract

Drug development is an ever-growing field, increasingly requesting reliable in vitro tools to speed up early screening phases, reducing the need for animal experiments. In oral delivery, understanding the absorption pattern of a new drug in the small intestine is paramount. Classical two-dimensional (2D) in vitro models are generally too simplistic and do not accurately represent native tissues. The main goal of this work was to develop an advanced three-dimensional (3D) in vitro intestinal model to test absorption in a more reliable manner, by better mimicking the native environment. The 3D model is composed of a collagen-based stromal layer with embedded fibroblasts mimicking the intestinal lamina propria and providing support for the epithelium, composed of enterocytes and mucus-secreting cells. An endothelial layer, surrogating the absorptive capillary network, is also present. The cellular crosstalk between the different cells present in the model is unveiled, disclosing key players, namely those involved in the contraction of collagen by fibroblasts. The developed 3D model presents lower levels of P-glycoprotein (P-gp) and Multidrug Resistance Protein 2 (MRP2) efflux transporters, which are normally overexpressed in traditional Caco-2 models, and are paramount in the absorption of many compounds. This, allied with transepithelial electrical resistance (TEER) values closer to physiological ranges, leads to improved and more reliable permeability outcomes, which are observed when comparing our results with in vivo data.

Introduction

The field of drug development is in constant growth, challenged by the needs of a world that is always facing new diseases and requesting for effective tools to fight them. In the case of oral delivery, which is the preferable route for drug administration, understanding the absorption potential of a drug at an early stage of its development takes a major role, since without effective absorption, the drug is not able to reach the site of action and fulfill its purpose.

Ineffective intestinal absorption and metabolic instability are responsible for 50% of drug failure in clinical studies [[1], [2], [3]]. This shows that in vitro screening platforms can take a major role in preclinical drug development and help reducing, or even replacing, animal experiments, which are more laborious, expensive and raise ethical concerns [4]. Thus, results provided by these platforms must be reliable. The gold standard of in vitro intestinal permeability assays remains the Caco-2 model, which presents several drawbacks, as the presence of a tighter barrier than what is observed in vivo, the up-regulation of P-gp and other transporters, the lack of a mucus layer, and the fact that it only represents one cell type and one layer of the intestinal mucosa [5,6]. There have been some improvements in the Caco-2 model, such as the co-incorporation of HT29-MTX cells that reduce barrier tightness and provide the mucus component. More recently, the development of improved 3D models has also been achieved, but the 2D models are still the most widely used [[7], [8], [9], [10], [11], [12], [13], [14], [15]].

In 2D models, cells are seeded on top of semipermeable membranes, which are made of hard plastic and do not recapitulate the 3D environment that exists in vivo. This environment plays a crucial role in the maintenance of cell functions and polarity, and the presence of extracellular matrix (ECM) strongly influences cell behavior [9,[16], [17], [18], [19], [20]]. Indeed, it has been demonstrated that ECM composition can alter intestinal epithelial proliferation and differentiation [21,22]. While in 2D, mechanical and biochemical cues, as well as cell-cell and cell-matrix communication are lost, in 3D these interactions can be recapitulated, helping to understand key events, such as proliferation, migration and apoptosis, which are determined by the cellular context [18]. There is no doubt that reproducing the 3D environment is crucial when the goal is to mimic native tissue. In fact, the last decades of research in the field have shown that culturing cells in 3D can bridge the gap between the traditional 2D cultures and the native tissue [18].

In order to be biologically relevant, a model of the gastrointestinal tract should include, besides the epithelial layer, ECM with stromal cells to mimic the complex interactions between the cells and the surrounding tissues [23]. The endothelium, which is present in our model, also has an impact in the behavior of the surrounding cells, taking part of the cellular crosstalk that occurs.

When it comes to the development of in vitro models for permeability drug screening, mimicking the native environment is important, as it can affect the outcomes of absorption. In fact, some of the disadvantages of the Caco-2 model, that are observed when cells are cultured in 2D in semipermeable plastic membranes, may be surpassed when cells are cultured in 3D. The tightness of the barrier is one of them, since it has been demonstrated that when Caco-2 cells are seeded on top of more physiological substrates, the TEER is reduced, which can be due to a lower expression of tight junctions, relating more to what is observed in vivo, which can greatly affect the permeability of paracellular compounds [9,10,12,24,25]. Besides, in terms of intestinal drug transporters, it has been observed that seeding the cells in more physiological matrices can lead to an expression more similar to in vivo levels, which can have a major impact in the permeability outcomes, especially when referring to important efflux transporters, as P-gp and MRP2 [10,25,26]. In fact, while 2D Caco-2 models can have some success at predicting permeability data for passively absorbed drugs, in the case of carrier-mediated compounds they have been less successful, showing significantly different correlation between in vivo and in vitro results, mainly in terms of the expression of intestinal transporters [27].

The main goal of the present study was to develop an advanced 3D intestinal in vitro model that could better mimic the native small intestine. This was achieved by providing a biological 3D support and include cells that could have an impact in cellular crosstalk and in the obtained results, by adding complexity to the model. Still, the goal was to keep it simple enough, so that it could be easily reproduced by others, in order to replace the conventional 2D models.

The model proposed in this work is comprised of a collagen layer with embedded human fibroblasts representing the intestinal lamina propria and giving the 3D support to the intestinal epithelium. The latter is composed by Caco-2 and HT29-MTX cells, which mimic the enterocytes and goblet cells, respectively. An endothelial layer was added underneath the lamina propria, mimicking the barrier posed by the blood vessels, which can also affect permeability outcomes. To our knowledge, this is the first attempt of addition of an endothelial layer in a 3D intestinal in vitro model to understand its impact in the absorption process.

In our previous work, we had proposed a 3D model composed of a collagen layer with human intestinal fibroblasts (HIF) embedded and epithelial cells on top, but contraction of the collagen by the fibroblasts was observed, which limited the set of studies performed [9]. In the present study, the model was optimized to avoid contraction, and the secretion of factors that may lead to the enhancement of the contractile ability of fibroblasts was evaluated.

Furthermore, the 3D model was characterized in terms of the expression of drug transporters, intestinal markers, tight junctions and MUC2, to understand if it would, in fact, better mimic the in vivo environment. Since the rationale for the development of this model laid on the hypothesis that obtaining a more physiological model would render better permeability outcomes, the absorption of different model drugs was evaluated and compared to the conventional 2D models, as well as in vivo data.