**1. Introduction**

Translational medicine is an interdisciplinary branch of biomedicine, which aims to translate results from preclinical (in vitro and in vivo) research to the clinical setting, the so-called "from bench to bedside" path [1]. Its main objective is to improve prevention, diagnosis, and treatment strategies for human diseases by filling the gap between the preclinical and the clinical setting. For fruitful translational research, it is fundamental that the discovery process, and the relative experimental design, is immediately aimed at its application in clinical practice [1,2].

In this context, the correct choice of the experimental model and the study design are a fundamental part of the translational plan, and they must be made carefully. In oncology, the use of animal models represents a still-irreplaceable step after in vitro studies. Indeed, they are an essential source of in vivo information, which validates in vitro results and improves their translational value [3–5]. Rodents, particularly mice, are the most used in vivo oncological models, thanks to their high genetic homology with humans, their easy genetic manipulability, and their fast reproductive cycle. Hence, their use allows for concluding studies in a relatively short time [3–7].

In vivo preclinical imaging makes possible noninvasive longitudinal studies; thus, achieving a reduction in the biological variability and a substantial decrease in the number of animals while maintaining the statistical power of the data [8–10]. Most imaging techniques for small animals are already the same used in clinical settings, i.e., computed tomography (CT), positron emission

tomography (PET), high-frequency ultrasonography (HFUS) and magnetic resonance imaging (MRI), and hence, are readily translatable. In oncological research, these imaging modalities are useful in studying drug biodistribution, monitoring treatment response, and identifying new molecular targets and biomarkers for early tumor detection [3,11–15]. Preclinical scanners have to deal with smaller "subjects" with higher metabolic and physiological rates. Hence, dedicated scanners are designed to meet these needs in terms of higher spatial and temporal resolutions but still adopt the same physical and technological principles of their clinical counterparts, which lead to more direct and faster translatability of findings [15].

In this perspective, the MRI can be considered the least invasive and most comprehensive technique, from an anatomo-functional point of view, in the clinical setting. The preclinical MR scanners have to, as mentioned deal with a smaller field of view and higher spatial resolution, assured by the use of higher magnetic fields; e.g., up to 11.4 T. Nonetheless, novel sequences able to show specific pathophysiological features in oncologic disease might be directly translated to the clinic [9,16,17].

This work aimed to review the most recent scientific literature to highlight the translatability of MRI from preclinical animal studies to clinical oncology. In particular, we focused on thyroid, breast, and prostate cancer rodent models, according to our previous and ongoing research lines. Indeed, such histotypes display a significant impact both on clinical oncology and in preclinical research. Moreover, to underline the versatility of the MRI techniques, some preclinical studies on human patients have been reported, as well.

#### **2. Animal Models in Oncology**

Four broad categories of oncological small animal models are used: xenografts, orthotopic, patient-derived tumor xenografts (PDXs), and genetically engineered mouse models (GEMMs). As for spontaneous human cancer, chemical and radiation-induced murine models have also been developed, and they seem to behave more closely like human tumors [18,19]. Finally, specific considerations will be made about metastatic models. For most animal models, general anesthesia is usually needed, either to eliminate the stress and the pain linked to a surgical procedure, or to obtain immobility of the mouse. The other procedures are usually executable with physical restraint by personnel well-trained in handling animals [20].

#### *2.1. Xenograft and Orthotopic Models*

Subcutaneous and orthotopic xenograft models are based on cancer cell lines directly derived from human tumors. Such cell lines belong from primary, lymph nodes, or metastatic tumors; they are immortalized and usually well-defined from the genetic point of view [21,22]. Hence, they allow extremely reproducible experiments for growth rate, metastatic potential, histopathologic homogeneity, and biological behavior. For their growth in the host species, i.e., the mouse, the host itself has to be immunocompromised [3,22]. Depending on the cell line, various degrees of immune system "knock-out" might be necessary [23,24]. In general, these models summarize well, as mentioned, the histopathological features present in the original tumoral tissue, but the use of immunodeficient mice may hinder the study of the tumor–host immune interaction. Subcutaneous xenografts are more easily reproducible compared to orthotopic implantation, which usually requires invasive surgical procedures. However, xenografts are not representative of the original tumor in its native environment [21,25]. On the other hand, the use of orthotopic models reestablishes the interactions between the tumor and its origin organ and might recapitulate the metastatic behavior with sufficient penetrance and reproducibility. Nonetheless, a limit of orthotopic models, besides the lack of interaction with the host's immune system, is that they do not allow modeling the pre-neoplastic process since the cell line already has completely tumorous genetic and biological behavior [18,25,26]. Depending on the murine strain selected, these models might be less expensive than GEMMs [25,26]. These animal models, for their reproducibility, seem to be ideal for preclinical and translational studies concerning

therapies for tumor growth and angiogenesis inhibition. Furthermore, they are useful for studying the pathogenetic aspects of cancer vascular invasion and metastasis [26–28].

Both xenograft and orthotopic models are used to replicate thyroid cancer. In the xenograft models, selected cell lines are injected into the subcutaneous tissue of immunodeficient mice. The selection of the appropriate cell line is important not only to fulfill the experimental aims but also because it has been demonstrated that among 40 thyroid cell lines, almost 40% are cross-contaminated or misidentified [29,30]. Hence, genetic evaluation of the chosen cell lines should always be performed [30,31]. The orthotopic model is more complex and requires microsurgery to inject tumor cells directly into the murine thyroid gland [25,26,32]. Recently, an HFUS-guided injection procedure has been presented and resulted in less invasiveness compared to the standard surgical procedure [33]. Such a procedure was easily reproducible for the orthotopic implantation of thyroid carcinoma cells and allowed for continuous monitoring of the disease, for a more extended period and with less contamination of other neck structures by the carcinomatous cells compared to the surgical approach [33]. In thyroid cancer research, the orthotopic model allows the understanding of the molecular and cellular mechanisms of thyroid cancer pathogenesis, improving the evaluation of new therapeutic compounds, and testing the effects of therapeutic interventions on both the primary tumor and metastases [34,35].

The xenograft and orthotopic mice models are widely used for investigating breast cancer. These are commonly produced by injecting subcutaneous tumor cells in the flank or orthotopically in the mammary fat pad of immunocompromised mice. The xenograft models allow one to study, in vivo, the tumor environment, and tumor growth, but the absence of an intact immune system profoundly affects tumor development, in particular, its early stages, and the progression of the metastatic process. Orthotopic models of breast cancer provide a more favorable microenvironment, but there are critical differences between human and mouse mammary stroma [36]. Commonly, these models involve athymic nude, severe combined immunodeficiency (SCID) and non-obese diabetic (NOD) SCID mice, which show genetic and immunological differences that mainly influence the induction and dissemination of metastases to organs. Indeed, natural killer (NK) cells and the remaining innate immune cells in nude and SCID mice probably contribute to the reduction of tumor engraftment, growth, and metastases in these models. On the other hand, NOD SCID and particularly NOD SCID gamma (NSG) represent a better background to study breast cancer metastases due to the lack of NK cells, and thus they could be considered ideal for the study of anti-metastatic treatments [37–39].

Prostatic cancer models are usually xenografts, in which tumor cells are inoculated subcutaneously, and the cellular lines mostly used are either androgen-independent (PC-3 and DU-145) or androgenresponsive (LNCaP) cells. Such a set of cell lines reproduces various features of the naturally occurring disease [40,41]. Orthotopic prostate tumor models are relevant to study the growth and metastasis of prostate cancer. Generally, tumor cells are injected into the prostates of immunodeficient mice; however, there is a low incidence of tumor formation due to technical difficulties that can be overcome by intratesticular inoculation [41–43].

The xenograft models of colon cancer are rarely used [44]. The orthotopic model represents an accurate reproduction of the human pathology and can be generated by the implantation of tumor cells, via a surgical celiotomy, under the cecal serosa. Alternatively, a model generated by trans-anal rectal injection has been developed, and it may allow a more accurate investigation of the inflammatory and immune responses, without the influence of previously-used abdominal surgery. The orthotopic colon cancer model is useful in inducing local tumor growth but is limited by a low rate of lymph node and hepatic metastases [45,46].

#### *2.2. Patient-Derived Xenografts (PDX)*

The PDX models are based on portions of primary tumors, are collected by biopsy or surgical excisions, and are transferred from the patient in immunodeficient mice by subcutaneous or orthotopic implantation. Since these models are derived directly from the patient's tumor, they have a great translational potential for the study and development of anticancer therapies in the perspective of personalized medicine. The main advantages of PDX are that they duplicate precisely, the original clinical tumor compared to long-established cell lines, and hence, they preserve the molecular, genetic, and histological heterogeneity of naturally occurring cancers [47–49]. These models have been shown to have a predictive value in the clinical outcome and are useful for studying novel or alternative drugs, in order to identify biomarkers for early predictive responses to the treatment, and thus, to guide therapeutic approaches in clinical patients. Furthermore, when this model is generated by an orthotopic implant, it often reproduces, as mentioned, the same metastatic process of the human disease. All these features give to PDX models, the ability to study metastases and the genetic evolution of a tumor [47–49]. There are some challenges concerning PDX models. Occasionally, to obtain a reasonable growth rate, they are propagated between different hosts; i.e., the graft is removed from one mouse and implanted in another one. During this passage, the stromal components of the mouse may become dominant over the human one; hence, limiting the translatability of results of the tested therapies. Moreover, the interaction of the tumor with the immunocompromised hosts does not allow studying the interaction with the immune system during the anticancer treatment, eventually hiding crucial therapeutic mechanisms [47,49,50]. Last but not least, the limitation of this model is linked to long bureaucratic waiting periods when obtaining the necessary authorization for animal experimentation. Only big governmental projects might be prepared from the perspective of a truly personalized medicine approach [48].

Concerning thyroid cancer, PDX mouse models accurately reproduce the tumor microenvironment and the biochemical interaction between tumor cells and stromal components, as previously mentioned [47]. However, particularly for medullary thyroid carcinomas, there are serious difficulties in generating a murine PDX model due to the slow tumor growth [51].

The PDX models of human breast cancer are gaining high relevance for preclinical evaluations of experimental therapeutics. In general, cells derived from patients are transplanted into a mammary fat pad of immunocompromised mice, particularly in NSG mice that are commonly used and better suited for PDX generation for their high engraftment rates [52,53].

The PDX models are used to highlight various aspects of prostate cancer in preclinical research. In particular, they are used to test individual responses to therapy, to clarify the angiogenetic pathways, and to identify new antimetastatic treatments. The tumor/human stroma interaction can support tumor growth and improve the onset of metastases resembling the naturally occurring disease. However, there are several issues during the orthotopic implantation, related to the size of the fragment and the choice of the implantation site, which have led to lower availability of PDX models of prostate cancer compared to other histotypes [41,54].

The PDX of colon cancer is generally generated using cultured primary cells derived from a patient's colon tumor, which are injected orthotopically into the cecal serosa. The direct transplantation of colon tumor tissue is being discarded, due to the bacterial contamination and the consequent septic shock in the immunocompromised host. Moreover, a significant amount of tissue is required, which can hardly be replaced in the event of engraftment failure. Nonetheless, PDX models of colon cancer allow the monitoring of tumor growth, metastatic evolution, and the pharmacological responses of individual patients [55,56].

#### *2.3. Genetically Engineered Mouse Models (GEMMs)*

The GEMMs play a significant role in cancer research, and they can be considered the most advanced animal models recapitulating human pathology [21,57]. The GEMMs develop tumors in situ, according to the gene(s) activated and/or deleted in immunocompetent, genetically manipulated, mice; the tumor growth in such models can be either spontaneous or molecularly-induced. These characteristics are extremely useful in studying the complex processes of carcinogenesis and also allow evaluating the interactions between tumor cells and the immune system in a more "natural" microenvironment [18,21]. These models better reflect human disease in terms of biological behavior,

histology, and genetic accuracy, but they also show some disadvantages, such as variable penetrance, heterogeneous growth, and tumor latency, and they are expensive [4,13,15,21,57].

Transgenic thyroid cancer mouse models are widely used to study the numerous mutations already systematically identified in human specimens [25,32]. The GEMMs models represent a way to study carcinogenesis, tumor progression, and metastatization, and hence, responses to therapeutics. The use of these models of animals with intact immune systems, as well as the ability to histologically reproduce human tumors, allows the evaluation of different drugs at different stages of the disease during tumoral development. However, these models are not useful for evaluating antimetastatic therapies since they do not always reproduce the specific pathways of human disease [32,57]. They are also useful for the study of unconventional therapies, such as micro-RNA, small interfering-RNA, or immunotherapies, during the early events of tumor transformation and progression, and as a promising approach to customizing anticancer therapy [58].

The GEMMs models of breast cancer display many features of the human pathology, although they do not entirely recapitulate all its aspects, due to the complexity and heterogeneity of breast cancers. However, these models are invaluable for investigating the biology and pathogenesis of breast cancer, and they are especially helpful for elucidating the mechanisms that regulate the initiation and progression of this disease, as well as for understanding the genes involved in these phases [36,59]. Transgenic mouse models of breast cancer are developed in immunocompetent hosts, and therefore, they allow evaluating the cancer/host immune interaction. Some of these models spontaneously metastasize to the lungs, whereas the incidence of bone metastases is very low. Furthermore, there are some selective tissue promoters, such as the whey acidic protein and the mouse mammary tumor virus, which may be directed to the oncogenic expression of the mammary gland [36,60].

In the study of prostate cancer, GEMMs reproduce, in depth, its different stages, and their use has provided valuable information on tumor initiation and progression. These transgenic models have been divided into different categories. The first generation of GEMMs utilized probasin to produce the probasin-large T antigen transgenic mouse (LADY) and the transgenic adenocarcinoma of the mouse prostate (TRAMP) models, which both show an aggressive phenotype, often with metastatic bone involvement. On the other hand, the second generation GEMMs have integrated the single molecular changes observed in human diseases, such as the loss of phosphatase and TENsin homolog (PTEN) and Myc over-expression, but they seem to be unable to reproduce the different stages of prostate cancer [54,61].

A variety of GEMMs have been generated as models for colon cancer to understand how common, coexisting mutations cooperate in a natural environment. These transgenic models are useful for studying pathogenesis and testing potential therapeutic agents, although they do not show phenotypes similar to human diseases. Moreover, these models are not practical for studying colon cancer metastases [44,62].
