Cancer tissue engineering—new perspectives in understanding the biology of solid tumours—a critical review

Tumour models: comprehension versus complexity

The search for cancer models has started in the second half of the last century and it is still in progress (Figure 1A–B). Traditional in vitro systems are 2D, but they offer the appealing advantage to the scientist, to be highly reproducible and responsive to drugs and radiations^[6]. However, this model has revealed to suffer from a scarce predictability (Figure 1B)^[7]. This is due to a number of reasons, whose deep understanding parallels the ongoing achievements in cancer biology, making 2D models insufficient. Basically, the lack of reliability seems to be associated to three main aspects as follows: cell sources, model dimensionality, and microenvironment complexity^[7,12]. It has to be reminded that in vitro expansion and passaging of cells is known to produce phenotype selection and eventually, alteration with time^[2,15]. This surely makes primary tumour cells preferable to long passaged and immortalised cell lines. However, beside mere cancer cells, as entities of action, the whole cancer microenvironment has recently shown a strong relevance in the comprehension of carcinogenesis and thus, in therapeutic success^[16]. The tumour is a markedly variegated-3D tissue structure, comprising several cell types, exerting mutual support throughout the secretion of specific soluble factors and ECM molecules, including vascularisation (Figure 1A). Considering this, the very first cellular selection is performed during cell isolation from a tumour biopsy, as it involves native ECM disaggregation and culture selection of fast replicating and plastic-adaptive cells, to the detriment of cancer supporting cells. An additional concern related to cell source, which has been pointed out in the last few years, relies on cancer stem cells (CSCs) and their pivotal role in tumour eradication^[17,18]. CSCs have been described as tumourigenic cells, which show stemness features, present in a tumour tissue at some concentration^[18]. Such cells have been addressed as a distinct population of the cancerous tissue, but capable of long-term delivery of differentiated progenies of diverse cancer cell types. Therefore, CSCs have been invoked as the main cause of tumour relapse and metastasis^[18]. In this respect, failure of traditional therapies could be explained with a wrong-cell targeting, because the differentiated cells are the most represented in tumours. Basic problem ever afflicting stem cell recognition and targeting, is the lack of specific surface antigens, which makes their direct identification usually tricky^[19]. This is due to the undifferentiated nature of any stem cells and renders a panel of markers necessary to circumscribe, although not to strictly identify, the cell population of interest. On the other hand, sometimes differences between CSCs and normal stem cells have not been well-identified. Therefore, any CSC-targeted therapies are hypothesised to potentially affect normal stem cells and to be detrimental for patients^[20]. Although a CSC selection and 2D culture is possible for a therapeutic screening, other issues still remain unsolved. They include drug delivery, efficiency and selectivity, as well as tumour self-protecting mechanisms involving cell-cell and cell-ECM interactions. The understanding of all these characteristics involves the availability of complex tumour models able to contain diffusion gradients and to mimic the tumour microenvironment. We still need, essentially, biomimetic 3D models of cancer^{[1,3,10,11,12]}.

Figure 1: Schematic picture of (A) a tumour and (B) tumour models, with their main characteristics. Some important aspects were identified and qualitatively scored according to the findings of the published literature and to our personal experience. They include model-inherent features, such as model type and reproducibility, and some model biomimetic capabilities. 2D, two-dimensional; 3D, three-dimensional.

The most widely used 3D (complex) model of cancer biology is typically an animal model (Fig. 1B). In vitro-selected human cancer cells, are typically, injected in a nude animal as a host (xenograft) and grown to form tumour masses and metastases^[6]. Although animal models have appeared very promising, they have resulted, in the end, as a poor predictive^[7,12]. This can be explained with model-inherent reasons as follows: the immune system of the animals is compromised in order to host human cells, so it cannot be a part of the therapeutic screening, the life span of the animal (usually mice) is usually shorter than the relapse time of tumour in humans, and in the tumour microenvironment, the vascularisation and supporting cell infiltration (e.g., fibroblasts) are of animal origin, while the tumour cells are of human origin; this ‘chimerism’ can cause a completely unpredicted response to therapies^[16]. To overcome these limits, advanced animal models, experimentally laborious, have been developed for some cancer types^[12]. Nevertheless, there are still constitutive anomalies affecting the use of animals in the study of human diseases, such as ethics, cost-effectiveness and a general lack of predictability. However, presumably because of some anthropomorphic perception of the animal model, clinicians typically take a favourable look at the employment of in vivo tests for their research, and accept with difficulty, to put efforts for the improvement of in vitro models, which are conversely reproducible and ethically sustainable.

In time, in vitro models of cancer have started evolving towards the third dimension^[8,9,21]. Simple 3D in vitro models used by scientists include spheroid formation and gel (usually collagen-derived) embedding of tumour cells (Figure 1B)^[9]. Spheroids are culture artefacts leading, for some transformed-cell types, to an induced cell aggregation in the form of compact spheres with diameters ranging from 20–1,000 µm^[8]. For their nature, spheroids partially mimic the tumour microenvironment as follows: they show secretion of tumour ECM, 3D cell-cell interactions, diffusion gradients and increased chemoresistance, while phenotype diversity is missing^[12]. Moreover, spheroid-based assays generally lack accuracy due to several difficulties in the management of these cell aggregates. With the attempt to improve 3D models, cancer cells have also been embedded in biologic hydrogels, which should mimic the primary ECM of tissues. However, such gels usually show insufficient porosity to obtain long-term cell survival and proper tumour ECM deposition. Moreover, spatial distribution of cells in the gel is often not uniform, thus resulting in poor consistent models^[9].

Recently, microfluidics circuits have been developed to make a further step towards 3D cultures in cancer^[22]. Yet, when macroscopically relevant dimensions (higher than 1 mm^[3]) are achieved, nutrient diffusions and cell survival remain problematic^[14]. To solve these challenges, microfluidic well systems, with the capacity of controlling nutrient perfusion, have been developed and used alone or in combination with hydrogels^[22].

Different from xenograft, spheroids and gel embedding, TE models can potentially offer all the fundamental achievements to cancer studies obtained so far for the regeneration of normal tissues as follows high standardisation of assays, multiple cell-type interaction, tailourable architecture allowing spontaneous 3D cell disposition and ECM synthesis, mechanical properties matching those of the tissue and tuneable diffusion profiles, thus appearing, in the end, as potentially elective models for the regeneration of 3D tumours (Figure 1B)^{[1,3,10,11,12,23]}.

Engineered tumours: achievements and perspective

A TE model of cancer should be a bottom-up 3D reconstruction of the tissue, using selected cells (CSCs or tumour cell mixtures), derived from primary cultures or from tissues, thus retracing the schematic diagram shown in Figure 2^[23]. For each tumour type, suitable scaffold architecture should be identified, ideally which is able to match the topographic and mechanical aspects of the native tissues^{[1,3,10,11,12,23]}.

Figure 2: Flowsheet of TE cancer models. This example is rendered with images of a study related to human pancreatic ductal adencocarcinoma (hPDAC) performed in our laboratories. Consequentially, single images/image groups show the following: light micrograph of hPDAC morphology (haematoxylin and eosin stain); light micrograph of isolated primary hPDAC cells (PP244); scanning electron microscopy (SEM) micrograph of scaffold inner structures (3D fiber deposition via Bioplotter®, sponge via emulsion and freeze-drying, microfibers via electrospinning); hPDAC cell/scaffold constructs under different viability assays (colour gradients of the scaffold surfaces highlight spatial localisation of viable cells) and light micrograph of a spongy construct (haematoxylin and eosin stain) showing presence of 3D hPDAC cell clusters within the scaffold pores, whose morphology mimics that of native PDAC (see the tumour biopsy micrograph). 2D, two-dimensional; 3D, three-dimensional.

The current state-of-the-art about the development of in vitro 3D-biomimetic model for some important tumours is reported in Table 1. An overview was given of relevant studies involving the interaction of biomaterials and tumour cells to generate 3D cancerous constructs in vitro^{[1,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39]}. A special focus was finally given to pancreatic, breast and prostate cancers, as such topics already account for a number of published studies about the 3D interaction of cancer cells and biomaterials.

Table 1 Tumour cell/biomaterial models for different cancer types.