Animal models of colorectal peritoneal metastasis

Abstract Colorectal cancer remains an important cause of mortality worldwide. The presence of peritoneal carcinomatosis (PC) causes significant symptoms and is notoriously difficult to treat. Therefore, informative preclinical research into the mechanisms and possible novel treatment options of colorectal PC is essential in order to improve the prognostic outlook in these patients. Several syngeneic and xenograft animal models of colorectal PC were established, studying a wide range of experimental procedures and substances. Regrettably, more sophisticated models such as those giving rise to spontaneous PC or involving genetically engineered mice are lacking. Here, we provide an overview of all reported colorectal PC animal models and briefly discuss their use, strengths, and limitations.


Introduction
With an annual worldwide mortality rate of over half a million, colorectal cancer (CRC) remains a major cause of cancer related mortality [1]. Since malignant disease ultimately causes death by distant organ invasion, the unravelling of molecular mechanisms underlying hematogenous and lymphatic metastasis is a topic of intensive research activity [2]. In parallel, the introduction of targeted biological agents has met with considerable survival prolongation in patients with metastatic disease [3]. On the other hand, intraperitoneally located tumors may be at the origin of locoregional peritoneal spread. Although often coexisting with systemic disease, it is increasingly realised that colorectal tumor dissemination within the peritoneal cavity may represent a separate phenotypic and molecular entity. Established peritoneal carcinomatosis (PC) from CRC is much less responsive to systemic therapy and causes considerable morbidity in affected patients. Synchronous peritoneal metastases are found at the time of surgery with curative intent in about five to six percent of patients, and are more frequently observed in right sided cancers [4]. Peritoneal carcinomatosis is present in 25-30 % of patients with recurrent or metastatic colorectal cancer; in approximately 3 % isolated peritoneal disease without systemic spread is observed [5,6]. Recognition of the causes and mechanisms of peritoneal metastasis may contribute to strategies to effectively prevent the development of PC in colorectal cancer. Moreover, in a small group of patients with low volume peritoneal disease, a locoregional treatment strategy combining surgery with intracavitary cytotoxic therapy has been shown to improve outcome [7]. The concept of intraperitoneal (IP) drug delivery in itself not entirely new. The earliest IP "drug therapy" was reported in 1744 by the English surgeon Christopher Warrick, who, apparently with great success, injected a mixture of 'Bristol water' and 'claret' (a Bordeaux wine) in the peritoneal cavity of a woman suffering from intractable ascites [8]. Intraperitoneal adjuvant chemotherapy has been extensively studied in stage III epithelial ovarian cancer, where it was found to be superior over intravenous (IV) chemotherapy alone in large randomized trials [9]. In patients with PC from appendiceal or colorectal origin, the combination of cytoreductive surgery (CRS) and intraoperative hyperthermic intraperitoneal chemoperfusion (HIPEC) has witnessed an impressive rise in clinical application over the past years [10]. In parallel, innovative pharmaceutical platforms such as targeted agents, nano-sized medicine and drug eluting beads have the potential to further increase the appeal of locoregional drug delivery.
Preclinical animal based research remains an essential tool in the unraveling of the pathophysiology of the metastatic cascade, and the therapeutic insights gained therefrom. Here, we provide a systematic overview of the animal models that have been used to study various aspects of peritoneal metastasis from colorectal origin.

Methods
A systematic search (completed 25/12/2015) was performed using Web of Science with the following keywords: [periton* and meta* and colo* and (animal or mice or mouse or rat or rodent or rabbit)]. Eligible studies reported on animal experiments involving, in part or exclusively, the establishment of colorectal peritoneal carcinomatosis by IP introduction of a colorectal cancer cell line or tissue fragment. Only papers published as full text were eligible. The resulting abstracts were scrutinized and those deemed to fit the criteria were retrieved as full text papers. Additional studies were searched for in the reference lists of these papers, and in the citing studies.

Results
The selection process (Figure 1) resulted in a total of 164 included papers, the details of which are summarized in Table 1. The large majority of studies used syngeneic rodent cell lines (usually CT26, MC38, or CC531) injected in the peritoneal cavity of immunocompetent mice or rats. Human colon cancer cells were xenogafted IP in athymic nude or BALB/c mice in 46 studies, and in athymic nude rats in two. Only a small minority of studies used SCID mice, patient derived xenografts, or tansgene animals.

Research questions and topics
From the wide variety of reported research topics, only a small minority addressed fundamental mechanisms of the peritoneal metastatic cascade. Experimental designs and questions include: -Mechanisms and prevention of port site metastasis after laparoscopy -Activity of IP chemotherapy, heparin, anti-adhesive products, gene therapy, photodynamic therapy, immunotherapy, or radioimmunotherapy -Evaluation of novel pharmaceutical formulations and carriers for IP delivery -Evaluation of novel optical, fluorescence, or radioactivity based imaging techniques for diagnosis and staging of PC

Choice of cell line and animal model Syngeneic models
Syngeneic or allograft models use cells or tissue derived from the same genetic background. The recipient  animals have a normal immunity, and the resulting IP tumors therefore display a more representative microenvironment. On the other hand, these colon tumors are chemically induced and are not representative of the genetic and molecular heterogeneity of human cancers. Obviously, use of syngeneic models is the preferred approach for the study of cancer immunotherapy.
In immunocompetent mice, all published studies have used either the CT26 (colon tumor 26) or MC38 cell line, which are syngeneic to the BALB/c and C57BL/6 mouse, respectively. Both cell lines were developed in 1975 by exposing mice to repeated intrarectal applications of Nnitroso-N-methylurethane (NMU) or 1,2-dimethylhydrazine dihydrochloride (DMH) [11]. CT26 is a rapid-growing grade IV carcinoma that is easily implanted and readily metastasizes; it shares molecular features with aggressive, undifferentiated, refractory human colorectal carcinoma cells [12]. The MC-38 murine colon tumor is a grade III adenocarcinoma [11]. Both cell lines cause widespread PC two to three weeks after IP injection.
In immunocompetent rats, the most commonly cited model is the syngeneic CC531 cell line in the WAG (Wistar Albino Glaxo) or WAG/Rij rat. Tumor CC531 is a DMH-induced, transplantable adenocarcinoma exhibiting weak immunogenicity and which has been widely used in metastasis research [13]. Upon IP injection, the CC531 cell line causes widespread carcinomatosis and haemorrhagic ascites after three weeks [14]. In Fischer F344 rats, the spontaneously metastatic RCN-9 syngeneic cell line was established by subcutaneous administration of DMH [15]. Other syngeneic, chemically induced rat colon cancer models include the BN7005-H1D2 cell line in the Brown Norway rat, DHD/K12/TRb in the BD IX rat, and RCC2 in the Fischer F344 rat.

Xenograft models
Xenograft models involve the transplantation of human cancer cells or tissue to immunodeficient animals. Nude mice (athymic nude and BALB/c nude) and the athymic nude rat have a biallelic mutation of the FOXN1 gene (which in humans encodes the Forkhead box protein N1), leading to an athymic state and the hairless phenotype. These animals are unable to generate mature T lymphocytes and the related adaptive immune response. Severe combined immunodeficiency (SCID) mice carry a homozygous mutation of a gene coding for Prkdc, an enzyme involved in DNA repair, resulting in absent or atypical T and B lymphocytes. Non obese diabetic (NOD) SCID mice have, in addition, deficient natural killer (NK) cell function. The disadvantages of xenografted models are higher costs due to isolation requirements, the fact that the stromal component of the tumors is rodent, that the hosts are immunodeficient, and that most of the tumor lines were developed using early technology. Also, a striking feature of xenografted tumors is early and extensive necrosis, which may hamper efficacy and imaging studies. In addition, use of a "standard" cell line can result in a population that is not truly representative of the original tumor and may therefore respond differently to therapy compared to. In fact, the use xenograft models has been debated due to their low ability to predict clinical response [16]. The colon cancer cell lines that were used in xenografted PC models include HCT116, LS174T, and HT29.

Patient derived xenografts (PDX)
In order to overcome the most important drawback of xenograft models, i. e. the loss of genetic and morphological heterogeneity of the original tumor, patient derived xenografts (PDX) were developed [17]. These models consist of patient derived cancer cells or tissues transplanted in immunodeficient animals. PDX models have a long latency period and low engraftment rate, and are therefore very costly to maintain. They are ideally suited for testing novel and "personalized" cancer therapeutics. In the field of colorectal peritoneal metastasis, three studies reported the use of PDX. Kotanagi et al. obtained colorectal PM tissue fragments from a patient with stage IV right sided colon cancer [18]. Intraperitoneal injection of a single cell suspension resulted in poorly differentiated PC in four out of five SCID mice. Flatmark and coworkers implanted tumor fragments originating from mucinous colonic or appendiceal cancer in BALB/c nude mice [19]. Mice developed mucinous ascites and widespread mucinous implants; after several passages the ascites component became more prominent. The histological and molecular properties of the engrafted tumors closely resembled those of the originating clinical material. Tumor tissue fragments from an ovarian metastasis in a stage IV colon cancer patient was transplanted IP in NOD-SCID mice by Navarro-Alvarez et al. [20] The resulting xenografts were used to identify and characterize a novel tumor-initiating cell (NANK).

Genetically engineered mouse models
Genetically engineered mice (GEM) including transgenic, knock-out, knock-in, and their intercrosses have not been used in the study of colorectal peritoneal metastasis. Only one author describes the use of mice expressing human CEA as a transgene [21].

Large animal models
Larger animals are rarely used in PC research. Apart from the cost and handling issues, colorectal syngeneic or xenograft models are unavailable in large animals. In rabbits, a non-colorectal PC model based on the V×2 cell line is available. The V×2 cell line is derived from the Shope papilloma virus (family Papovaviridae), an oncogenic DNA virus, transmitted by biting arthropods and causing hyperkeratotic skin lesions resulting in malignant transformation in the rabbit [22]. A 'gastric' peritoneal carcinomatosis model based on the V×2 cell line was proposed by Tang et al. [23,24] The authors simulated gastric cancer with early stage PC in New Zealand white rabbits (Oryctolagus cuniculus) by transmural injection of V×2 cells in the stomach. Turner and coworkers succeeded in engrafting human colon cancer cells (LS174T) in cyclosporine treated sheep by subperitoneal injection [25]. Tumors grew at all sites within three weeks, and were used to study the biodistribution of a radiolabelled antibody. The use of a pig model was reported by Hewett, who studied the pneumoperitoneum induced movement of colon cancer cells immediately after IP instillation [26].

Establishment of experimental PC
An orthotopic PC model is easily established by IP injection of cancer cells, which results in widespread and progressive carcinomatosis, leading to cachexia, hemorrhagic ascites, and death of the animal. The efficacy (engraftment or take rate) and speed of this process depend on the number of cells injected, virulence of the cell line used, and immunocompetence of the host. Although this model is orthotopic, the metastatic process and its underlying biology are different from spontaneous PC arising from a primary colon cancer. Cespedes and coworkers established a primary colon cancer model by submucosal injection of HCT116 cells in the colon of nude mice, and observed the development of PC in 100 % of the animals [27]. Using this model, the same group showed that use of a colon cancer cell line overexpressing Snail1, which decreases E-cadherin, completely blocked spontaneous PC [28]. Similarly, Puig et al. injected patient derived colon cancer cell lines into the cecal wall of NOD-SCID mice and observed spontaneous PC when cell lines were used originating from cancer with a mucinous differentiation [29]. The disadvantage of the IP injection and spontaneous PC models is that the resulting tumor load is difficult to quantify. Also, their very small size precludes detailed physiological or drug penetration study at the individual tumor level. We recently established a colorectal PC model consisting of two isolated peritoneal nodules, which develop upon subperitoneal injection of HT29 cells in Matrigel.™ [30] This model allowed assessment of tumor tissue interstitial fluid pressure, oxygenation, platinum penetration, and growth delay ( Figure 2).

Experimental endpoints Extent and distribution of PC
Most authors have quantified the extent of experimental PC by a scoring system based on the number and/or size of peritoneal implants, similar to the peritoneal cancer index (PCI) that is clinically used. Use of such a score is difficult when the tumor forms a confluent mass or film rather than isolated nodules. Others have used the total weight or volume (as determined by water displacement) of the tumor mass, ascites presence and volume, or the metastatic pattern as endpoints. Alternatively, the extent of microscopic disease has been studied on resected omental tissue, peritoneal biopsies, or omental lysates using (immune)histology or PCR. The above methods require invasive procedures. Several authors have quantified PC load at different time points using optical (fluorescence or bioluminescence) techniques based on cancer cell lines transfected with a green or red fluorophore, or with the firefly luciferase gene. Alternatively, cells may be labeled immediately before injection with quantum dots or other reporters [31]. These techniques are sensitive and fast, and allow reproducible quantification using a variety of image processing methods. Some authors have used bioluminescence of organ and tissue lysates in order to quantify tumor growth.

Survival
In studies investigating novel therapies of colorectal cancer, survival is an important endpoint. Since advanced PC causes considerable animal suffering, care should be taken to sacrifice the animals whenever a predefined humane endpoint is reached. Actuarial (rather than actual) survival is usually calculated, and comparisons made with the log rank test or the Cox model.

Other endpoints
Various other endpoints were reported. Some authors have analysed the pO 2 , VEGF concentration, or immune response of tumor associated ascites. Others have imaged PC distribution using optical techniques (Figure 3), or have analysed the biodistribution of isotope labelled tracers in tissue or in the whole animal.

Conclusions and recommendations
Colorectal peritoneal metastasis remains little studied in preclinical models, when compared to ovarian cancer or liver metastasis research. Standardized, reproducible syngeneic and xenograft colorectal PC models are available in rodents. The choice of a specific model is dictated by the aim of the study. Technical models involving IP chemoperfusion or laparoscopy are easier in a rat model. Tumor physiology, pharmacokinetics, and growth delay are better studied in isolated peritoneal tumors established by peritoneal implantation of tissue fragments or subperitoneal injection. Very few genetically modified mouse models have been reported in PM research. With the advent of sophisticated genome editing tools such as CRISPR-Cas9 (clustered regularly interspaced short palindromic repeats associated nuclease 9), the use of genetically engineered models is expected to gain in importance in the near future.
Author contributions: All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission. Research funding: None declared. Employment or leadership: None declared. Honorarium: None declared. Competing interests: The funding organization(s) played no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the report for publication.