Experimental Murine Models for Colorectal Cancer Research

Simple Summary In its early stages, colorectal cancer (CRC) is a localized tumor, but when it metastasizes, it has dramatic consequences. Murine models in CRC research are important tools for advancing the knowledge in diagnostic and treatment of this pathology. The present review aims to provide a variety of murine models in CRC research describing their particular advantages and drawbacks. Abstract Colorectal cancer (CRC) is the third most prevalent malignancy worldwide and in both sexes. Numerous animal models for CRC have been established to study its biology, namely carcinogen-induced models (CIMs) and genetically engineered mouse models (GEMMs). CIMs are valuable for assessing colitis-related carcinogenesis and studying chemoprevention. On the other hand, CRC GEMMs have proven to be useful for evaluating the tumor microenvironment and systemic immune responses, which have contributed to the discovery of novel therapeutic approaches. Although metastatic disease can be induced by orthotopic injection of CRC cell lines, the resulting models are not representative of the full genetic diversity of the disease due to the limited number of cell lines suitable for this purpose. On the other hand, patient-derived xenografts (PDX) are the most reliable for preclinical drug development due to their ability to retain pathological and molecular characteristics. In this review, the authors discuss the various murine CRC models with a focus on their clinical relevance, benefits, and drawbacks. From all models discussed, murine CRC models will continue to be an important tool in advancing our understanding and treatment of this disease, but additional research is required to find a model that can correctly reflect the pathophysiology of CRC.


Introduction
Colorectal cancer (CRC) is the third most common malignancy and the second most deadly cancer worldwide. It is estimated that the number of cases in 2020 has reached c. 1.9 million with some 0.9 million deaths worldwide [1]. If there is no advancement in early detection and efficient therapies for late-stage CRC, this significant public health burden is anticipated to grow [2].
Animal models have become the foundation of CRC basic research, enabling the study of the disease's pathogenesis and validating new therapies (Figure 1). The murine models show a resource with enormous promise since they allow researchers to simultaneously observe and control a complex disease such as CRC. The advantages of murine models include their low cost, manageability, short gestation period, anatomical resemblance to humans and ease of genetic manipulation. For research on carcinogenesis, quick tumor development, the ability to analyze the adenoma-carcinoma sequence, the use of transgenic, knock-out and knock-in animals are additional benefits [23][24][25]. Various murine models have been reported, but each one has limitations [26]. Additionally, there is animal-to-animal variation in the development of CRC in various murine models. Despite this, animal models have developed into a crucial tool for better understanding the impact of genetic changes on the disease process [27].
This review provides an overview of the most used CRC murine models, describing their particular benefits and drawbacks.

Materials and Methods
An extensive evaluation of the literature on murine models of colon carcinogenesis was conducted by scanning the PubMed database of the National Library of Medicine. MeSH phrases such as "colorectal cancer", "animal model", "chemoprevention", "colon- This review provides an overview of the most used CRC murine models, describing their particular benefits and drawbacks.

Materials and Methods
An extensive evaluation of the literature on murine models of colon carcinogenesis was conducted by scanning the PubMed database of the National Library of Medicine.
MeSH phrases such as "colorectal cancer", "animal model", "chemoprevention", "coloncarcinogenesis", "min-mice", "colorectal cancer", "xenograft", "heterotopic model", "orthotopic model", "metastatic model", "patient-derived tumor", "genetically engineered murine models", and "transplant murine model" were used in the MEDLINE search. By using the aforementioned search terms, 9063 papers were initially identified from 1940 until the present. All of the 650 pertinent publications were carefully examined and studied and 273 of the manuscripts dealing with the murine models for CRC were included in this study. To help the reader, a full list of abbreviations used is presented in Abbreviations.
The administration of these carcinogens is possible via ad libitum feeding, oral gavage (o.g.), intraperitoneal (i.p.), subcutaneous (s.c.) or intramuscular (i.m.), injections or intrarectal (i.r.). These chemical carcinogens can be administered alone or in combination. Some carcinogens require biotransformation to cause cancer, while others do not. In contrast to direct agents, indirect carcinogens are delivered in an inactive state and acquire carcinogenic activity only after being biotransformed into their active form in the liver [29]. Intestinal mutations are the key to the success of the CIMs. Tumors from these animals randomly develop genetic and pathological similarities to human CRC [30]. Typically, this model takes a considerable amount of time to establish the type of cancer, because these tumors progress slowly from normal cells to adenocarcinoma/carcinoma [28]. In addition, the incidence of CRC development is influenced by the rodents' gender, age, and genetic background. Additionally, the intestinal flora, nutrition and immunological condition of rodents can interfere with the metabolism of carcinogenic chemicals, consequently affecting their effective local concentration. Relevant research has been summarized in Table 1.
The general characteristics of the murine models and the mechanisms by which these carcinogens cause CRC are addressed in the following section.

1,2-Dimethylhydrazine (DMH)
DMH and azoxymethane (AOM), a metabolite of DMH, are procarcinogens that must undergo metabolic activation to produce DNA-reactive by-products. The methylation of guanine at position N-7 in DNA is the starting point for the mutagenic activity of the alkylating agents DMH and AOM. By providing a proton, the alkylated guanine is coupled with thymidine rather than cytosine, changing the bases [31]. DNA mutations result from further replication, mismatching of guanine to thymine and cytosine to adenine, and other events. Different metabolic enzymes, such as xenobiotic-metabolizing enzymes, process these procarcinogenic chemicals' N-oxidation and N-hydroxylation steps, resulting in the creation of the final carcinogen, MAM [31], a reactive metabolite of DMH and AOM [29].
There are different routes that DMH can be administered such as o.g., s.c., i.p., and i.r. [29]. In Table 1 several examples using DMH are listed, including the animal species, dose injected, latency time and main characteristics of developed tumor.
The formation of tumors in the colon can be induced by DMH at a wide range of dosing levels from 2 mg to 200 mg/kg b.w. (a single injection to 30 injections) and with a latency period from 8 h to a maximum of 78 weeks (critical for the development of a tumor in the colon) (Table 1). Recently, routes of administrations and dosages were standardized (15 and 20 mg/kg b.w.) depending on the experimental study [32].
Despite the fact that most experimental CRC research has been conducted in murine animals, the high frequency of tumors in the lower part of the colon and the histopathological evidence of multiple adenomas and subsequent progression of adenocarcinoma validate the importance of DMH-induced models in the pathogenesis of CRC [31].
Although DMH-induced colon tumors in rodents are comparable to human colon tumors, this model has some drawbacks, such as the requirement for numerous DMH injections to cause tumors [27], the presence of a latency phase lasting at least six months and the absence of hepatic metastases up until this point [29] with non-transgenic rodents.
DMH and Colitis-Associated CRC (CAC) Models CAC is a consequence of inflammatory bowel disease (IBD) with a poor prognosis since it is frequently identified in advanced stages with local development or metastases. CAC has different molecular processes than polyp-induced sporadic CRC (sCRC), which is more prevalent [33]. Although a full understanding of IBD pathogenesis is unclear, the predominant pathological finding is characterized by persistent inflammatory processes at the local site [34]. Previous research has demonstrated that local immunological processes during chronic inflammation are distinct in ulcerative colitis (UC) and Crohn's disease (CD) [34,35]. CD is characterized by the presence of activated T helper type 1 (Th1) cells in the intestine, as well as high expression of interferon gamma (IFN-γ) and tumor necrosis factor alpha (TNF-α) [34,35]. On the other hand, enhanced expression of cytokines produced from T helper type 2 (Th2) cells is frequently observed in UC [34,35].
The most common chemical agents to generate CAC models are dextran sulphate sodium (DSS), which promotes a Th2 response that is similar to UC, and 2,4,6-rinitrobenzenesulfonic acid (TNBS), which promotes a Th1 response that is similar to CD [36,37]. Some examples are listed in Table 1. For instance, male Balb/c mice received a single dose i.p. of DMH (20 mg/kg b.w.), and one week after, DSS (3%) was given in drinking water for 1 week followed by normal drinking water for next 2 weeks (three alternate cycles of DSS), and at week 20, mice were euthanized [38]. In this study, upregulation of all the signaling events in DMH/DSS mice indicates the development of an aggressive and invasive carcinoma, mainly driven by nuclear factor kappa B (NF-κB) and also the loss of goblet cells that result in decreased production of goblet cell specific mucin such as MUC2, which are associated with poor prognosis in CRC. The reduction in both acidic and neutral mucins in DMH/DSS-treated mice suggested the advanced stage of tumor progression with carcinogen treatment [38].
In a DMH/TNBS model, male Wistar rats received twice every week four doses of DMH (40 mg/kg b.w) and two weeks after receiving DMH, 10 mg of TNBS mixed in 0.25 mL of 50% ethanol (v/v) was administered i.r. in the middle colon to generate acute inflammation and expedite the carcinogenesis process [39]. By week 25, the rats were euthanized and showed tumors with low grade dysplasia (60%) and high grade dysplasia (40%), a tumor multiplicity of 3.50 ± 1.72, and aberrant crypt foci (ACF) [39].

Azoxymethane (AOM)
Given its benefits over the original compound, AOM has been utilized more frequently than DMH in the induction of CRC [40]. It is suggested that AOM induces colon carcinogenesis more effectively than other carcinogens due to its greater stability, with enhanced potency being one of its advantages [36]. Similar to DMH, AOM is a procarcinogen that is activated in the liver by N-oxidation via cytochrome P450 2E1, producing proinflammatory metabolites such as methylazoxymethanol and methyl-diazoxide; however, it seems that AOM acts like a more efficient carcinogen [41]. A drawback of the AOM approach includes the fact that this procarcinogen is more expensive than DMH, highly toxic, and has stringent shipping requirements, despite its widespread use [33].
AOM promotes CRC in rodents when injected i.r., s.c. or i.p. as described throughout the literature; however, the preferred form of administration of this procarcinogen is i.p. The distribution of small intestine and colon tumors (mostly in the distal colon) is comparable to that reported in the human colon [42]. AOM-induced tumors share the same histological and histochemical characteristics as human cancers, being classed as adenomas and adenocarcinomas [43]. Rosenberg et al. found that mice with different genetic backgrounds have distinct sensitivity to AOM, with A/J and SRJ/R mice having great sensitivity (8 weeks after AOM treatment), while C57B/L6 and Balb/c mice have moderate sensitivity [27]. The tumor morphology was similar among these two mouse strains, while metastases or invasion was not observed even in the mouse strain with high AOM sensitivity. Histopathological progression varies among these two mouse strains, with tumor crypts composed of closely packed cells with indistinct cell borders and infiltration of neoplastic cells into the muscular wall of the distal colon. This model is suitable for studying early-stage but not late-stage or metastatic CRC [43].
As a result of their high similarity to human CRC, AOM-induced CRC tumors and carcinogenesis in murine animal models are widely used to evaluate novel chemopreventive and therapeutic strategies and to provide new insights into the risk factors and pathophysiological mechanisms of human CRC [44].

AOM and Colitis-Associated CRC (CAC) Models
The combination of AOM and DSS is a reliable method for inducing CRC in mice, replicating the pathogenesis of CAC, and is thus a highly reproducible acute or chronic model of intestinal colonic inflammation [36,45]. The standard protocol for DSS-induced colitis in rodents involves addition of DSS to drinking water at a concentration of 2-10% (3-5 cycles). This model is also helpful for gaining insight into the innate immune mechanisms of UC [36]. Table 1 also lists some examples. A disadvantage of this model is that it has a relatively low propensity of developing metastases [46]. Furthermore, the AOM/DSS model's immunological characteristics may not reflect those of typical human CRC cases, making it inappropriate for preclinical research [28].
Although TNBS model is a readily inducible, fast, robust and highly reproducible CAC model [36], it is not frequently used in association with AOM. Nevertheless, Xiao et al. used this model with 8-week-old C57BL/6 mice. The authors used AOM (10 mg/kg b.w., i.p.) followed by 2.5 mg i.r. of TNBS (150 µL with 50% EtOH) [47]. The results showed that the mice displayed extensive inflammatory, dysplasia, or carcinoma lesions all over entire mucosa with numerous ulcers and edema of submucosal and muscle layers. In addition, an overexpression of proinflammatory cytokines TNF-α and IL-12 were also observed.
Variations in susceptibility to TNBS-induced colitis among mouse strains demand optimization of the concentration of TNBS. Individual adjustment of inflammatory agent doses is necessary. Although the experimental protocols are straightforward in concept, individual variances in the intestinal microbiota between animal facilities and the genetic variety of mice strains necessitate pilot studies to adjust the dosages of TNBS and DSS [36,37]. Additionally, substantial variation between batches of these chemicals is possible.

Heterocyclic Amines (HCAs)
In rodents, HCAs, such as IQ and PhIP, are mutagens that are created when meat and fish are broiler-cooked [41,48]. Creatinine, amino acids and carbohydrates found in meat and fish are the precursors of IQ-type HCAs. For IQ to transform into its ultimate carcinogen, liver microsome metabolic activity is required [49]. Cytochrome P450s in the liver convert an amino group to a hydroxyamino group, which activates HCAs, and these are further triggered, resulting in DNA adducts that cause cancer [50]. IQ and PhIP have attracted a significant amount of interest due to their multitarget organ specificity, as they induce cancer in the stomach, colon, mammary gland and prostate of rodents [27,51].
Dietary treatment of IQ and PhIP to rodents for 52 weeks resulted in a low incidence of tumors (5-28%); however, feeding rodents PhIP for 104 weeks was associated with a high incidence of colon tumors (43-55%) and significant toxicity [52]. Malignancy type and onset time appear to be highly dependent on the experimental model; however, in all models, spontaneous tumor induction by PhIP alone requires lengthy periods of time, typically between 52-82 weeks [53]. Nevertheless, when PhIP is paired with DSS, colon adenomas and adenocarcinomas arise in rodents, and the duration of tumor initiation is decreased to 6-24 weeks, depending on the animal strain [53]. Additionally, to stimulate tumor growth, PhIP can be used alone or in conjunction with AOM and DMH.
This model is utilized more frequently in chemoprevention research since this carcinogen may be ingested by humans [53]. As early as 10 weeks after PhIP/DSS administration, histopathologic and biochemical analyses revealed tubular adenocarcinomas with substantial overexpression of critical proteins in the Wnt signaling pathway (e.g., β-catenin), cell regulation (e.g., c-Myc and cyclin D1), and inflammation (e.g., iNOS and COX-2), thus providing one advantage of this model when studying the overexpression of the proteins above mentioned [54,55].

Aromatic Amines
The first intestinal cancer was chemically induced in mice fed with the polyaromatic hydrocarbon 3,2-dimethyl-4-aminobiphenyl (DMAB) or methylcholanthrene [56,57]. The DMAB model has two main disadvantages: a) multiple injections are generally required; and b) low specificity since tumors may be induced in other tissues such as adenocarcinoma of mammary glands, urothelial carcinoma of the bladder, sarcoma, lymphoma, among others [41,52]. In rodent models, DMAB is less efficient than the series of compounds formed from DMH or AOM on a molar equivalent basis [27].
Several studies were made with DMAB and some of them are listed in Table 1. The murine model that stands out was performed by Reddy et al. with male F344 rats who received weekly s.c. injections (50 mg/kg b.w.) for 20 weeks [58]. The rats were autopsied after the last injection of DMAB and the histopathological findings revealed adenocarcinomas. This study found that diets containing wheat bran and citrus fiber reduce the risk for DMAB-induced intestinal cancer and that the protection against CRC depends on the type of fiber. DMAB reacts with DNA through the formation of DNA adducts via N-hydroxylation, O-acetylation, and hydrolysis by cytochrome P450 [59].
To cause cancer, heterocyclic and aromatic amines undergo metabolic activation by the enzymes NAT1 and NAT2. A high frequency of NAT1 and NAT2 polymorphisms makes humans prone to cancer when exposed to aromatic and heterocyclic carcinogens [60]. F-344 and WKY rats received a s.c. injection of DMAB (50 or 100 mg/kg b.w. in peanut oil) in weeks 1 and 2 and were euthanized at week 10, demonstrating a higher frequency of ACF in colonic tissue of rapid acetylator (NAT2) genotype, thus showing it to be a risk factor in aromatic amine-induced colon carcinogenesis [61].
In another study, male F344 rats (5-weeks old) received a s.c. injection of DMAB (100 mg/kg b.w. in peanut oil), and 48 h after, the rats were euthanized and DMAB-derived adducts were analyzed in colon and liver [59]. This study investigated the chemoprotective effect of celecoxib on levels of DMAB-derived adducts in the target organ colon and the non-target liver. DMAB shares structural similarities with mutagens found in well-done meat and has been demonstrated to be affected by dietary fat and fiber in F344 rats. DNA adduct development is vital, but additional elements such as cell proliferation are required for tissue susceptibility to tumor development. Comparable quantities of adducts were found in the liver and colon; however, in this model, cancer only occurs in the colon, and celecoxib showed a related dose-dependent decrease in DMAB-derived DNA adducts.

Alkylating Substances
N-methyl-N-nitro-N-nitrosoguanidine (MNNG) and methylnitrosourea (MNU) are direct alkylating substances, i.e., they do not need metabolic activation. MNU or MNNG injected intrarectally has been shown to cause CRC in rat studies [62,63]. For instance, i.r. treatment with MNNG (1-3 mg/week b.w.) for 20 weeks caused CRC at the injection site in male F344 rats (57% adenomas and 43% adenocarcinomas) [29,41,49]. MNU, despite developing CRC when administered i.r., may also induce thymic lymphoma and lung malignancies, which can be fatal [64]. It has been demonstrated that the development of DNA adducts and abnormal crypt foci are the results of i.r. administration of MNU [65]. CRC induced by MNU or MNNG contains Kras (5-30%) and Apc (6%) mutations; nevertheless, the complete molecular profile of mutations caused by these carcinogens is still unknown [66][67][68]. Since i.r. administration of MNNG and MNU selectively produces tumors in the distal colon and rectum, these models have been widely utilized to test the therapeutic effects of various drugs for CRC management [29].
The primary drawback of the alkylnitrosamines models is the difficulty in administering a precise dose of the carcinogens per rectum [27,29,41]. Additionally, the animals must remain inverted for one minute following administration to prevent the carcinogens from returning to the anus [27,29].
Furthermore, to clarify the carcinogenic factors (such as the type of carcinogen and the duration of exposure) that determine whether CRC develops from an adenoma, occurs de novo, or in the absence of an adenoma, Endo et al. performed a comparative histopathological study of DMH (30 mg/kg b.w., i.p., once a week) and MNNG (10 mg/kg b.w., i.r., 3 times a week) for 3 or 15 weeks in male F344 rats [66]. In the 3-week group, low grade dysplasia coexisted with 71% of DMH-induced carcinomas and 82% of MNNG-induced carcinomas, while this was observed in only 10% of DMH-induced and 27% of MNNGinduced carcinomas in the 15-week groups, and the low-grade dysplasia predominated from the initial period of tumor occurrence. The study also investigated whether or not mutations in the Kras and Apc genes were linked to these carcinogenesis patterns. No Kras mutations were found in tumors that had been exposed for 3 weeks. Nevertheless, this mutation was found in 57% of DMH-induced tumors and 13% of MNNG-induced tumors in the 15-week group. Only 6% of tumors had mutations in an area of Apc that is similar to the human mutation cluster area. These findings provide further evidence that the patterns of carcinogenesis in the rat colon are time-dependent and that Kras mutations contribute partly to a subset of the patterns. Table 1 summarizes some murine models induced by chemical carcinogens. Male Wistar rats 40 mg/kg b.w., i.p./wk 10 wks Wnt signalling pathway (e.g., β-catenin and p53), cell regulation (e.g., c-Myc and cyclin D1), inflammation (e.g., IL-6, ROS and COX-2) and alterations of bacterial enzymes (e.g., β-glucuronidase and β-glucosidase) [
Mutations in two or more of those genes are frequently associated with the malignant phenotype of CRC. Each of these genes plays a unique role in the formation of CRC tumors [109].There are many genetically engineered murine models (GEMMs) that are used in CRC research ( Table 2). The most common model is the mouse with a mutated Kras gene, which is found in 86% of human CRC [110,111]. This model is used to study the development and progression of CRC, as well as the effectiveness of potential treatments. Other GEMMs that are used in CRC research include mice with mutations in the Apc gene, which is found in 90% of human CRC [112], and mice with mutations in the p53 gene, which is found in 60% of human CRC [113].
All these models are used to study the role of these genes in the development and progression of CRC, as well as the effectiveness of potential treatments.
GEMMs are also frequently used to study CRC in addition to the CRC CIMs, TMMs, and metastatic models.
The general characteristics of the APC mutant murine model and other transgenic animals for the investigation of CRC are covered in the following section.

Adenomatous Polyposis Mouse Models (APMM)
Carcinogenesis is a multi-step genetic process. It was shown that the mutation of the Apc gene is the first step in the carcinogenesis process of human CRC. The Apc gene controls a wide variety of cellular processes, including β-catenin levels, cytoskeleton organization, cell cycle regulation, apoptosis and adhesion [114]. Germline mutations in this gene are related to familial adenomatous polyposis (FAP) and CRC tumors [115].
Adenomatous Polyposis Mouse Models (APMM) are useful tools for the pre-clinical assessment of CRC therapies. These models involve mice with a genetic mutation causing them to develop multiple adenomas in the colon and rectum, similar to humans [43,116]. The tumors range from early neoplasia to advanced adenocarcinomas, allowing for realistic assessment of CRC treatments via targeted therapies, chemotherapy or combination therapies. These models allow a direct comparison between pre-clinical observations and clinical trial results by providing an in vivo platform which accurately replicates human tumor characteristics [117].
Apc Min/+ model applications have progressed most in the areas of CRC-tumor prevention and treatment using primarily chemical and pharmaceutical strategies [115]. The histological analysis of Apc mutation-induced tumors of the colon revealed that they are benign adenomas, making this model appropriate for investigating the premalignant rather than malignant phases of CRC. Nevertheless, tumor malignancy increases, and latency time shortens when AOM or other carcinogenic compound is administered [117,119,120]. Despite the fact that additional Apc-targeting murine models have been developed (Apc ∆716 , Apc ∆14 , Apc 1638N , among others), Apc Min/+ continues to be the most widely employed transgenic murine model of CRC [43].
Cre-loxP is a site-specific recombinase technology used to carry out deletions, insertions, translocations, and inversions at specific sites in DNA. Cre-loxP-mediated murine models were developed to allow the tissue-specific and conditional knock-out of tumor suppressor genes or activation of oncogenes, respectively, overcoming the obstacles of frequent embryonic lethality caused by germline knock-outs of tumor suppressors [117]. Deleting Apc in Lgr5+ ISCs mice using Cre-LoxP results in the rapid development of intestinal adenomas. This fact suggests that Lgr5+ ISCs are the cells of origin for intestinal cancer.
Colon adenocarcinoma can develop in mice that carry a combination of Apc Min/+ and Smad −/− or Apc ∆716/+ and Smad4 +/− ; however, Apc Min/+ and Apc ∆716 alone only induce the formation of adenomas and not invasive tumors [30,43]. Researchers are now able to regulate the timing and/or location of Apc deletions due to Cre-loxP technology. Cre recombinase deletions of Apc are restricted to the epithelial cells lining the gastrointestinal tract when expressed from tissue-specific promoters such as the Fabpland Villinpromoters [42,121]. To investigate the role of p53 in colon tumor invasion, Cre-LoxP was used to create Apc fl/+ p53 fl/+ and Apc fl/+ p53 R172H/+ mice, which displayed 25% and 100% stroma invasion, respectively [30,122]. Accordingly, these findings demonstrate both p53 tumor-suppressive function and the varied effects of p53 at various mutant loci [30,122]. Together, these results highlight the Apc Min/+ mouse model's significance in the investigation of colon tumorigenesis by establishing the pioneering function of the Apc mutation in the emergence of CRC.
The main disadvantages associated with the use of APMM are cost and complexity. These models require specific genetic modifications which can be costly to obtain, and they may not be amenable to certain therapies and treatments due to the inherent variation between individual mouse strains. Additionally, due to the complexity of the disease and the need to replicate multiple tumor characteristics, these models are difficult to design and set up compared to other pre-clinical models [117].

Hereditary Nonpolyposis Colon Cancer Mouse Models (HNPCC)
Germline pathogenic variants in DNA MMR genes cause hereditary non-polyposis CRC, also known as Lynch syndrome (LS), which is one of the most common cancer predisposition syndromes [123]. LS is caused by mutations in the DNA MMR genes Mlh1, Msh2, Msh3, Msh6, Pms1, and Pms2 (MutL Homolog 1; MutS Homolog 2, 3, and 6; and Post-Meiotic Segregation Increased 1 and 2; respectively) alone or in combination with a germline mutation in the Apc tumor-suppressor gene [116,124].
DNA MMR gene mutations cause chromosomal instability, while Apc mutations cause microsatellite instability (MSI). CRC, and LS in particular, can be affected by MSI in the promoter regions of the Apc, TGF-βRIII, and Bax genes [125]. Patients who have a hereditary predisposition to CRC (such as those with FAP or LS) often have severe genetic defects due to germ line mutations in tumor-suppressor genes and DNA MMR genes [126]. Incidence rates for LS are about as high as those for other subtypes of cancer combined (1 to 5%); however, sCRC also exhibits somatic mutations in these genes [126,127].
HNPCC murine models are useful tools for pre-clinical assessment of CRC therapies. These models involve the generation of mice that possess a genetic mutation known as the MMR defect, which is responsible for LS in humans. This genetic mutation, combined with other environmental factors, causes these mice to develop multiple malignant tumors in the colon and rectum, similar to the human condition. HNPCC murine models provide researchers with an in vivo platform to study the underlying biology and pathways associated with LS, as well as evaluate potential therapies.
Examples of HNPCC murine models include Mlh1, Mlh3, Msh2, Msh6, and Pms2 knockouts [128]. The Mlh1 and Msh2 knockouts involve the generation of mice lacking either the Mlh1 or Msh2 gene, resulting in a defective MMR system which leads to multiple tumors in the colorectum [129]. Similar to Mlh1and Msh2-deficient mice, the loss of both Msh3 and Msh6 increases gastrointestinal tumors at a much younger age, whereas Msh3 loss does not increase cancer susceptibility until later in life [130].
There are also multiple transgenic murine models that have been developed for LS, including the Kras G12D+ and Pms2 transgenic animal models. The Kras G12D+ expresses a constitutively active mutant form of the Kras gene, leading to the formation of numerous malignant tumors in the colorectum and with high lymph node metastases [30,134]. The Pms2 transgenic mouse model overexpresses the Pms2 gene and results in the formation of multiple tumors in the colorectum. In Biswas et al. [135] study, an increased intestinal polyp formation of ≈4.5-fold was observed in Pms2 ki/ki mice with heterozygous APC mutation (chain-termination mutation in the 15th exon, Apc +/− ) compared to Apc +/− or Pms2 ki/+ ;Apc +/− mice. Accumulated MSI is an indicative sign of MMR deficiency, and this was also demonstrated in Pms2 ki/ki intestinal adenomas.
The use of HNPCC murine models in CRC research has numerous advantages. These models provide an in vivo platform to study the biological pathways associated with LS and the formation of tumors. Additionally, they can be used to evaluate potential therapies in pre-clinical studies, as well as gain insight into the development of resistant tumor subtypes.
They can also be used to better understand drug delivery mechanisms and to assess response to therapeutics. Other advantages include the ability to assess gene expression profiles, identify new biomarkers, and gain insight into the molecular mechanisms of cancer progression. Moreover, these models can provide a platform for testing novel combinations of targeted therapies and immunotherapies, as well as studying the mechanisms of drug resistance.
Finally, they can also be used to assess recommendations of dietary modification, environmental factors, and lifestyle habits that may influence cancer progression; however, these models are expensive and time-consuming to create, and there can be variability in the phenotypes of the mice due to environmental factors. Additionally, the results obtained from these models might not be directly applicable to humans due to differences in biological pathways and genetic backgrounds. Thus, there is still a limited understanding of the role of genetic mutations in the development of LS, making it difficult to accurately predict the outcome of treatments using these models. Adenoma formation in the distal rectum in most of the Apc 580S homozygotes within 4 weeks after infection by rectal infusion with recombinant adenoviruses encoding the Cre recombinase.
In total, 50% of animals show invasive adenocarcinoma after 1 year without lymphatic or distant metastases.
Useful for studying the mechanism of CRC development and to test therapeutic or chemopreventative agents.
Only effective in Apc 580S/580S mice and not Apc 580S/+ , an outcome that reflects the poor ability of the approach to influence the proliferating cells at the crypt base. [137] CAC; APC 580S/+ Adenomatous lesions in the distal colon; DSS treatment increased the incidence and number of tumors, and this occurred predominantly in distal colon.
Mimics the tissue and cellular environment of heritable cancers such as FAP and LS.
Use tissue specific promoters in intestinal mucosa to target gene knockout; Some invasive adenocarcinomas seen can be used to target specific tumor-suppressor or oncogenes.

BRAF-V600E
Promotes rapid serrated tumor development and progression and assesses the role of Smad4 in early-stage serrated tumorigenesis Oncogenic β-catenin mutations (combinations of Ctnnb1, Braf, and Smad4) drive rapid serrated dysplasia formation.

Transplant and Metastatic Murine Models
In contrast to xenogeneic grafts, syngeneic tumor transplantation is characterized by the engraftment of tumor tissue or cancer cell lines inside the same murine strain. Herein, it is also possible to discriminate between orthotopic and heterotopic models.
The three cornerstones of Transplant Metastatic Models (TMMs) are host, xenograft, and methods of transplantation. The host has undergone numerous alterations from the first generation to the fourth generation over the course of many years of continual development and growth. Specifically, immunocompetent, genetically immunodeficient, new combination immunodeficient, and humanized murine models make up the first, second, third, and fourth generations of hosts, respectively.
Metastasis the leading cause of death in patients with CRC [155]; therefore, it is essential to employ murine models to replicate the clinical characteristics to study the underlying mechanism and find an effective treatment for. Although significant CRC murine models have been developed, animals that can develop metastatic characteristics remain scarce.
In the following section, several transplant and metastases murine models will be discussed.

Transplant Murine Models (TMMs)
TMMs has been extensively used in several areas of CRC therapy as a bridge to clinical use. Each TMM has distinct qualities, yet each model also has limitations. For the creation of TMMs, cell-derived xenografts (CDX), patient-derived xenografts (PDX), and/or patient-derived organoids xenografts (PDOX) are generally employed. Tumor materials are mostly transplanted into the host by s.c., intrasplenic, or orthotopic pathways to generate the TMMs. For instance, CDX models do not accurately represent patients' treatment responses, which has a relatively poor clinical approval rate for cancer drugs (less than 15%) [156]. Through extensive pharmacological testing, s.c. transplantation models have little prognostic value for human clinical response [157]. The orthotopic PDX models, in contrast, continue to exhibit the highest concordance of treatment responses across human patients and murine models, hence confirming their use as the best screening platform for the assessment of CRC anticancer drugs [158].
At the present, the creation of biomarkers, pharmacological testing, and surgical modeling are the main applications of TMMs. Figure 2 summarizes the main characteristics of TMMs.

Cell-Derived Xenografts (CDX) Models
From cell line banks all across the world, more than 100 cell lines have been established as CRC cell lines. To find underlying mechanisms and develop cancer drugs, CDX models were created by implanting cancer cell lines into immunodeficient animals [158]. When human CRC cells are s.c. injected into an immunocompromised mouse, the injection site frequently develops a tumor. The conventionally naked (athymic) and severe combined immunodeficient (SCID) mouse strains, which lack T lymphocytes or both B and T lymphocytes, respectively, are frequently utilized. In contrast to SCID mice, NOD/SCID mice also lack NK cells [159].
This model is commonly used in CRC research and DD due to its low level of technical expertise, ease with which tumor growth can be observed, affordable cost of maintaining colonies, high yield production and tolerable tumor latency [160]; however, due to the loss of original inheritance and the absence of pertinent tumor microenvironment (TME) components during in vitro passage, this model is unable to reproduce the tumor genetic heterogeneity of the initial tumor [161]. Additionally, repeated passaging with enrichment for particular subclones may result in changes at the genetic and epigenetic levels [162].
The C57BL/6-derived adenocarcinoma line MC38 and the Balb/c-derived lymphoma line CT26 are commonly used in studies of syngeneic s.c. transplantation [160]. Due to their high mutation rates, these cells have been proven to be effective preclinical models of human tumors; however, the success of orthotopic engraftment of MC38 tumors has varied greatly between studies, in contrast to the high fidelity with which CT26 cells have been implanted [163].

Cell-Derived Xenografts (CDX) Models
From cell line banks all across the world, more than 100 cell lines have been established as CRC cell lines. To find underlying mechanisms and develop cancer drugs, CDX models were created by implanting cancer cell lines into immunodeficient animals [158]. When human CRC cells are s.c. injected into an immunocompromised mouse, the injection site frequently develops a tumor. The conventionally naked (athymic) and severe combined immunodeficient (SCID) mouse strains, which lack T lymphocytes or both B and T lymphocytes, respectively, are frequently utilized. In contrast to SCID mice, NOD/SCID mice also lack NK cells [159].
This model is commonly used in CRC research and DD due to its low level of technical expertise, ease with which tumor growth can be observed, affordable cost of maintaining colonies, high yield production and tolerable tumor latency [160]; however, due to the loss of original inheritance and the absence of pertinent tumor microenvironment (TME) components during in vitro passage, this model is unable to reproduce the tumor genetic heterogeneity of the initial tumor [161]. Additionally, repeated passaging with enrichment for particular subclones may result in changes at the genetic and epigenetic levels [162].
Subcutaneous CRC cell line xenograft has several advantages and drawbacks. The main benefits of injecting cells into immune deficient mice are their low cost, rapid tumor growth, well-characterized cell lines, ease of genetic manipulation and model accessibility [116]. However, besides representing the disease at an advanced stage it has undergone significant clonal selection and rarely metastasizes [116].
In orthotopic xenografts of CRC cell lines, the injection of cells into intestinal serosa of immune deficient mice also has their benefits and disadvantages. The benefits of this model are similar to the s.c. model with the difference being a more natural microenvironment for CRC cells and that some cell lines metastasize to the liver (e.g., HCT-116 or HT-29). The disadvantages are similar as in s.c. model with the exception of a surgical requirement to implant cells [116]. Orthotopic implantation of human CRC cells such as HT-29, SW620, HCT-116, and SW480 is commonly performed using immunodeficient humanized mice, such as severe combined immunodeficient, Rag1, or nude mice; however, the inability to study adaptive immunity, cytotoxic T cells, or checkpoint blockade therapies is a major drawback of these models for immuno-oncology studies [163].
Syngraft/Isograft models involve transferring 1-2 mm mouse tumor fragments or mouse tumor cell lines to a genetically identical inbred, immune-competent mouse. These models have strengths such as no species mismatch between tumor and stromal cells, and an intact immune system that enables immunotherapeutic studies; however, they are labor intensive and time consuming and do not use human cell lines, making them the main weaknesses [116].
In order to expand our understanding of tumor biology and better identify innovative therapeutics for cancer treatment, PDX models have been developed in order to get around the limitations of the CDX model [172,[198][199][200].

Patient-Derived Xenograft (PDX) Models
The PDX model is a murine tumor model created by grafting human tumors onto immunodeficient mice. It has been demonstrated to be a useful tool for studying the biology of tumors and assessing the effectiveness of anticancer agents in a variety of tumor types [201,202]. Drug evaluation outcomes are most comparable to clinical situations, and the type of immunodeficient mice used and the delivery method affect the rate of engraftment [201,202]. Additionally, it has been demonstrated that PDX models preserve the heterogeneity of the underlying tumor in CRC [203][204][205][206]. To facilitate engraftment, monitoring and resecting the tumor s.c. implantation is the most common [13,156,207]; however, this model was created using tumor tissue cells that were extracted through enzymatic digestion in multiple investigations [208].
In other studies, researchers created CRC PDX8models orthotopically with endogenous metastases that can travel to the lungs and liver similarly to a patient's main tumor. To the best of our knowledge, metastases to other organs do not occur in s.c. engraftment [24,199,204,208]. The highest rate of engraftment (60-100%) in CRC is seen in PDX models using Balb/c nude mice (100%) as hosts. Nevertheless, NSG and NOD/SCID mice are also frequently used [206,209]. Surgical specimens are the most frequently employed original source because the amount of initial tumor material has a significant impact on the success of PDX engraftment [210].
There are important gene mutations including Kras, Braf (v-Raf murine sarcoma viral oncogene homolog B), and PIK3CA of the primary tumor that are still present in PDX CRC models, in addition to gene expression, copy number alterations, and MSI [156].
PDX is rich in stromal component compared to 2D cultivated cancer cell lines, which may be advantageous for research on the interactions between cancer cells and TME. It has been demonstrated that PDX maintains the global gene-expression patterns, mutational status, metastatic potentials, histological differentiation, and histopathological subtypes of the human donor tumor [156,210,211].
Numerous studies have used the CRC PDX model to assess the effectiveness of immunotherapy drugs and other systemic chemotherapeutic agents [212][213][214], as well as to identify drugs and biomarkers [166,198], produce cell lines [215,216], create colospheric structures [217], and learn more about tumor biology [218]. As a result, it can be used to create individualized cancer treatments; however, there are certain restrictions on using PDX, such as a lengthy engraftment phase, usually lasting 4-8 months. Therefore, employing the PDX model to offer patients with fast drug screening findings can be difficult [219]. Further, the poor cryopreservation and reanimation efficiency of PDX raises the possibility of losing valuable tumor samples [220].
Recent research has shown that the humanized mouse model is useful for studying the human immune system and evaluating the response to immunotherapeutic drugs. But there are ethical concerns with conducting experiments on humanized mice, such as the need for a tissue bank consisting of various types of human leukocytes and human hematopoietic cells obtained from bone marrow, umbilical cord blood and fetal organs [221][222][223]. Chimera studies are often considered unethical due to the crossing of species barriers [224].
PDX orthotopic models provide a strong framework for investigating the biology of metastases and therapeutic response in CRC. Nevertheless, this model has a limited capacity and reproducibility due to the technical ability required for orthotopic implantation. In recent years, it has also been possible to create PDXs that sustain tumorigenicity in mice by using fluid from malignant ascites or pleural effusions, circulating tumor cells (CTCs), or both [204,[225][226][227]. Ex vivo cultivated CTCs have been demonstrated to retain their tumorigenic capacity in CRC [228]. As a result, CTC-derived PDX models show potential for the analysis of tumor genomic evolution and the assessment of tumor responses to new therapeutics.

Patient-Derived Organoid Xenograft (PDOX) Models
It is known that PDX model is a time-consuming and relatively expensive task but the patient-derived organoids (PDOs) offer a potential solution for that issues. PDOs are developed from isolated organ progenitor cells or patient stem cells collections that originate clusters of 3D cultivated multicellular aggregates [13,229]. Evidence suggests that PDOs maintain both the tissue functions and the properties of the parent matrix. Furthermore, PDOs faithfully mimic in vivo tissues during homeostasis and diseases such as CRC. This model is simple to maintain, and it can genetically editable. It has been demonstrated that PDOs retain the genetic, transcriptomic and histological traits of the parental tumors [229].
On the other hand, traditional PDOs frequently only contain cancer cells and lack TME constituents, including fibroblasts, endothelial cells, and immune cells, among others. As a result, efforts are still being made to recreate the parental tumors' microenvironment by including TME components into organoids [13,230,231]. PDOs have some drawbacks, including having a protracted culture cycle, a single-cell source, variable culture conditions, and being time-consuming and a technically difficult model [219]; however, being the most recent source of xenografts with exceptional fidelity and adaptability in the CRC model, PDOs are still seen as a major advance in the study of cancer biology and therapeutic response, after the PDX model [231].
PDOX murine models have emerged as a useful tool for pre-clinical evaluation of CRC therapies, surpassing most of the limitations of CDX, PDX and PDO models. This model involves the generation of a PDOX cell line from a patients' tumor, which is then transplanted into a humanized mouse [222] or a mouse with an immunodeficient background. This model has been shown to accurately replicate both the histology and biology of the patients' tumor and to retain driver mutations present in the original tumor. This mouse background is preferred, as it eliminates any potential rejection of the tumor cells due to its lack of an adaptive immune system while still retaining the necessary cell signaling pathways to support the growth of the tumor cells [232].
A key advantage of the PDOX model is its ability to accurately replicate both the histology and biology of the patients' tumor, as well as retain driver mutations present in the original tumor. This allows for realistic assessment of drug efficacy and toxicities in an in vivo model that closely resembles the clinical situation. Another advantage is PDOX high success rate of production from primary CRC tissue (up to 90%). PDOX implantation provides a solid foundation for more accurate CRC murine models due to the tumor formation rate of 60% and 100% for implantation into the colon wall and cecal, respectively [13]. In addition, the technique allows researchers to study drug resistance and cancer metastases, which are two major factors in determining the overall efficacy of treatments. Moreover, PDOX models have been employed to evaluate the potential combination therapy strategies to treat CRC, by testing the efficacy of individual therapies in combination with each other. Finally, this PDOX model is suitable for pre-clinical evaluation of CRC therapies, providing researchers with a platform to effectively analyze and test [233].
The main drawbacks of PDOX murine models in CRC are the complexity and cost associated with setting up the model. As it requires reprogramming the patient's tumor and transplanting it into a mouse background, the procedure can be quite costly. Nevertheless, when comparing against PDX models, PDOX are less costly to establish and have a high throughput [231]. Additionally, as this procedure is relatively new, there is still a lack of validation and standardization protocols regarding PDOX technology. Finally, as these models are based on individual patients, they may not be suitable for providing generalizable results across different patient populations.

Metastases Models
The key factor contributing to CRC patients' high mortality rate is distant metastases [234,235]. Therefore, revealing biomarkers that predict drug response and identifying patients that are most likely to benefit from a specific treatment is crucial.
Animal models should be carefully chosen to closely resemble the molecular, histopathological and etiological features of the donor tumors [236]. Studies on metastases have made extensive use of PDX, GEMMs, and PDOX models (Table 3) [237]. Xenograft models are ideal for testing new therapies, but they remove the protective role of the immune system in disease development. It is possible to avoid some or all the steps necessary for metastases formation by injecting tumor cells directly into the cecal or colonic wall, or even into the bloodstream. GEMMs are widely used in studies of carcinogenic progression and the mechanisms of individual cancer-related genes, but they can be costly and have a long latency.
Research groups have developed orthotopic CRC PDX models, which preserve the TME necessary to investigate tumor cells with metastatic potential [204,216,238,239]. In Rashidi et al. [240] study, all mice implanted with a tumor had liver metastases within 10 days, and lymph nodes draining to the liver showed metastases 19 days after implantation; however, the main drawbacks of metastatic PDX models are that it is time-consuming, technically difficult and high-cost [219].  GEMMs of CRC can spread to the liver, as demonstrated by a study [244] in which LSL-KRAS G12V /APC flox/flox mice and an AdenoCre were injected into the colon. This causes the activation of oncogenic KRAS G12V , loss of Apc tumor suppressor, development of sCRC, and liver metastases [244,258].
GEMMs are useful for studying the role of individual genetic mutations in the carcinogenic process, promoting tumor progression and liver metastases while limiting overgrowth of cancer cells. They better depict the dynamics between tumor cells and their microenvironments throughout tumor progression than TMMs [259]. Furthermore, they are useful for assessing the earliest stages of tumor development [24,259]; however, mutations in genes can cause embryonic lethality, developmental defects, or sterility prior to the development of liver metastases in GEMMs. Thus, due to the slow progression and low incidence of metastases in GEMMs, it can be challenging to assess therapeutic responses [24,28,136].
In vivo manipulation of 3D CRC organoids has recently been described, with cecal or colonic implantation of these structures [236,237]. CRC and liver metastases models have been established by transplanting PDOXs with multiple cancer-related mutations into the colon, renal capsule, and spleen of mice [236,237]; however, neither tumor invasion into the muscularis propria nor tumor extravasation into the circulation through the colon serosa can be studied in these ectopic transplantation models [43,260]. To examine primary cancers and liver metastases, some research groups used orthotopic transplantation to place PDOXs into the colonic or rectal mucosa of NRGA-immunodeficient mice (PDOXwE), which was then subcultured in Balb/c mice [261,262]. Using gene-editing techniques, PDOXs can be engineered for desired mutations, a process that is much quicker than creating germline GEMMs.
PDOXs can also have fluorescent labels and other features added to them with relative ease [13]. Through xenotransplantation into the kidney subcapsules of immunodeficient (NOG) mice, Fujii et al. generated matched pairs of primary and metastatic organoids from CRC patients [256]. Organoids derived from CRC metastases in this model were more able to metastasize than their primary tumor counterparts. Orthotopic implantation of tumor organoids is preferable to the germline GEMM for pre-clinical study because it ensures that all mice in the same experiment have tumors of the same number and similar volume. These benefits ensure that the orthotopic CRC model based on organoids will be the most sought-after and useful model for future preclinical research [28].
CRISPR-Cas9 technology, for example, has added flexibility to genomic editing and has been heralded as a highly effective tool for achieving metastatic disease, particularly when combined with CRC organoids [136,262].

Meeting the Criteria for a Successful Murine Model for Colorectal Cancer Investigation
Previous murine models for CRC were compared in terms of key criteria to assess their potential for a successful CRC investigation and their potential translation into preclinical and clinical settings. This general comparison is depicted in Table 4, and it was based on authors' opinion.

Future Perspectives
The response of the tumor to anticancer drugs is highly variable, making it essential to comprehend the role of a heterogeneous TME in order to effectively manage treatment. For effective therapy management, understanding the role of a heterogeneous TME is essential.
It is known that highly translational cancer models are becoming increasingly important in the field of precision medicine.
The use of animal models always following the 3R principles (replacement, reduction, and refinement) is crucial but other models can and should be also used to complement that information. This indicates that in vitro and in silico models must also be improved in parallel to in vivo models. Organs-on-chips (OoCs) are one example of a key tool for this purpose. OoC platforms recreate key features of the TME in vitro [175]. An OoC model combines 2D and 3D cell-culture advancements with artificial organs that mimics the most typical sites of metastatic spread [263]. In addition, Guinney et al. highlight the importance of computational models in cancer science and the potential of bioinformatics research [264]. The relationship between treatment response and molecular subtypes has been partially shown by retrospective analysis of clinical trial samples [265].
A strategy to integrate and analyze the vast amount of data is required due to the constantly expanding knowledge of cancer pathways and their interaction on the one hand, and the rising interindividual complexity of tumors on the other [266,267]. Data mining, pattern recognition, machine learning, and network approaches are some examples of in silico models/techniques that can predict the behavior of virtual CRC cells [268], identify new biomarkers [269], identify unknown driver mutations [270], reveal genetic patterns linked to survival [271], and identify potential compounds [272,273]; however, they still have a long way to go before they can accurately forecast how a new compound will affect a patients' treatment response [273].

Conclusions
Murine models of CRC have yielded insights into pathogenesis mechanisms, tools for drug discovery, validation of novel therapeutic targets and a predictive platform for testing new preventative and therapeutic strategies. CIMs, GEMMs, TMMs and metastatic models have been used to study various aspects of CRC.
In this search, we observed that there is still not only one model to fully address this disease. In case of CIMs, these models are a powerful tool for understanding the development and progression of the CRC. Additionally, Cre-LoxP technology and TMMs have been used to advance our understanding of CRC but are limited in their ability to accurately reflect the complexity of the human immune system. PDOX models are challenging to use to investigate the impact of immune TME components on cancer progression, but more research is needed to confirm these facts.
To sum up, there is still a need for a model that can correctly reflect the pathophysiology of CRC. In the meanwhile, murine CRC models will continue to be an important tool in advancing our understanding and treatment of this disease.