The use of animal models in diabetes research

Abstract

Diabetes is a disease characterized by a relative or absolute lack of insulin, leading to hyperglycaemia. There are two main types of diabetes: type 1 diabetes and type 2 diabetes. Type 1 diabetes is due to an autoimmune destruction of the insulin-producing pancreatic beta cells, and type 2 diabetes is caused by insulin resistance coupled by a failure of the beta cell to compensate. Animal models for type 1 diabetes range from animals with spontaneously developing autoimmune diabetes to chemical ablation of the pancreatic beta cells. Type 2 diabetes is modelled in both obese and non-obese animal models with varying degrees of insulin resistance and beta cell failure. This review outlines some of the models currently used in diabetes research. In addition, the use of transgenic and knock-out mouse models is discussed. Ideally, more than one animal model should be used to represent the diversity seen in human diabetic patients.

LINKED ARTICLES

Animal Models

This paper is the latest in a series of publications on the use of animal models in pharmacology research. Readers might be interested in the previous papers.

Robinson V (2009). Less is more: reducing the reliance on animal models for nausea and vomiting research.

Holmes AM, Rudd JA, Tattersall FD, Aziz Q, Andrews PLR (2009). Opportunities for the replacement of animals in the study of nausea and vomiting.

Giacomotto J and Ségalat L (2010). High-throughput screening and small animal models, where are we?

McGrath JC, Drummond GB, McLachlan EM, Kilkenny C, Wainwright CL (2010). Guidelines for reporting experiments involving animals: the ARRIVE guidelines.

Kilkenny C, Browne W, Cuthill IC, Emerson M, Altman DG (2010). The ARRIVE guidelines.

Emerson M (2010). Refinement, reduction and replacement approaches to in vivo cardiovascular research.

Berge O-G (2011). Predictive validity of behavioural animal models for chronic pain.

Vickers SP, Jackson HC and Cheetham SC (2011). The utility of animal models to evaluate novel anti-obesity agents.

Percie du Sert N, Holmes AM, Wallis R, Andrews PLR (2012). Predicting the emetic liability of novel chemical entities: a comparative study.

The complete series including future publications, as they occur, can be found at http://onlinelibrary.wiley.com/journal/10.1111/(ISSN)1476-5381/homepage/animal_models.htm.

Introduction

Diabetes mellitus is a chronic disease that is characterized by a relative or absolute lack of insulin, resulting in hyperglycaemia. Chronic hyperglycaemia can lead to a variety of complications such as neuropathy, nephropathy and retinopathy and increased risk of cardiovascular disease. Recent figures suggest the worldwide prevalence of diabetes is 9.2% in women and 9.8% in men, with approximately 347 million people suffering from the disease worldwide in 2008 (Danaei et al., 2011). There are several different classifications of diabetes, the most common being type 1 and type 2 diabetes.

Type 1 diabetes is an autoimmune disease leading to the destruction of the insulin-producing pancreatic beta cells in the islets of Langerhans. Type 1 diabetes is most commonly diagnosed in children and young adults, and by the time of diagnosis, patients have very little endogenous insulin production. Insulin therefore has to be replaced by regular subcutaneous injections, and blood glucose levels must be frequently monitored to manage the risk of hypoglycaemia. Concordance of the disease in identical twins is around 27% (Hyttinen et al., 2003), indicating that although there is a genetic influence, environmental factors play a role in disease development. Indeed, most newly diagnosed patients have no first-degree relatives with the disease. Incidence of type 1 diabetes ranges up to 100-fold, depending on the country and is estimated to be approximately 15–20 per 100 000 in the United Kingdom (Patterson et al., 2009). The incidence in Europe is increasing, with a predicted doubling of children under 5 developing the disease by 2020 (Patterson et al., 2009).

Type 2 diabetes is the most common type of diabetes with prevalence in the United Kingdom of around 4%. It is most commonly diagnosed in middle-aged adults, although more recently the age of onset is decreasing with increasing levels of obesity (Pinhas-Hamiel and Zeitler, 2005). Indeed, although development of the disease shows high hereditability, the risk increases proportionally with body mass index (Lehtovirta et al., 2010). Type 2 diabetes is associated with insulin resistance, and a lack of appropriate compensation by the beta cells leads to a relative insulin deficiency. Insulin resistance can be improved by weight reduction and exercise (Solomon et al., 2008). If lifestyle intervention fails, there are a variety of drugs available to treat type 2 diabetes (Krentz et al., 2008), which can be divided into five main classes: drugs that stimulate insulin production from the beta cells (e.g. sulphonylureas), drugs that reduce hepatic glucose production (e.g. biguanides), drugs that delay carbohydrate uptake in the gut (e.g. α-glucosidase inhibitors), drugs that improve insulin action (e.g. thiazolidinediones) or drugs targeting the GLP-1 axis (e.g. GLP-1 receptor agonists or DPP-4 inhibitors).

As endocrine disorders, type 1 and type 2 diabetes represent quite complex diseases where different bodily systems are involved. Thus, animal models should be chosen carefully, depending on what aspect of the disease is being investigated. In this review, specific models of type 1 and type 2 diabetes are discussed. In addition, models of beta cell regeneration are outlined, and the use of knock-out and transgenic models in diabetes research is considered.

Animal models of type 1 diabetes

The main characteristic of type 1 diabetes is an autoimmune destruction of the pancreatic beta cells, leading to lack of insulin production. In animal models, this deficiency in insulin production is achieved by a variety of different mechanisms, ranging from chemical ablation of the beta cells to breeding rodents that spontaneously develop autoimmune diabetes. Some of the most commonly used models of type 1 diabetes are outlined in Table 1.

Table 1.

Summary of rodent models of type 1 diabetes

Induction mechanism	Model	Main features	Possible uses
Chemical Induction	High dose streptozotocin	Simple model of hyperglycaemia.	New formulations of insulin
	Alloxan		Transplantation models.
	Multiple low dose streptozotocin	Model of induced insulitis.	Treatments that may prevent beta cell death
Spontaneous autoimmune	NOD mice	Beta cell destruction due to an autoimmune process	Understanding genetics of type 1 diabetes
	BB rats		Understanding mechanism of type 1 diabetes
	LEW.1AR1/-iddm rats		Treatments that may prevent beta cell death
			Treatments that may manipulate autoimmune process
Genetically induced	AKITA mice	Beta cell destruction due to ER stress. Insulin dependent.	New formulations of insulin
			Transplantation models.
			Treatments to prevent ER stress
			(could also be used in type 2 diabetes research)
Virally-induced	Coxsackie B virus	Beta cell destruction induced by viral infection of beta cells	Establish potential role of viruses in the development of type 1 diabetes
	Encephalomyocarditis virus
	Kilham rat virus
	LCMV under insulin promoter

Chemically induced type 1 diabetes

In chemically induced models of type 1 diabetes, a high percentage of the endogenous beta cells are destroyed, and thus, there is little endogenous insulin production, leading to hyperglycaemia and weight loss. Chemically induced diabetes not only provides a simple and relatively cheap model of diabetes in rodents but can also be used in higher animals (Dufrane et al., 2006). Diabetes is usually induced around 5–7 days prior to the start of the experiment to ensure stable hyperglycaemia. Two main compounds are used to induce diabetes: streptozotocin (STZ) or alloxan. Due to their similarity in structure to glucose (Bansal et al., 1980), glucose can compete with alloxan and STZ, and thus, fasting animals tend to be more susceptible. Both alloxan and STZ are relatively unstable, and the solutions should ideally be made immediately prior to injection.

Chemically induced diabetes is appropriate to use when testing drugs or therapies where the main mechanism of action is lowering blood glucose in a non-beta-cell-dependent manner; for example to test new formulations of insulin (Jederstrom et al., 2005; Sheshala et al., 2009). This is also an appropriate model for testing transplantation therapies where the end point is lowering of blood glucose (Jansson et al., 1995; Makhlouf et al., 2003; Deeds et al., 2011). At the end of the experiment, any ‘cured’ animal should be nephrectomized of its graft bearing kidney and reversal to hyperglycaemia observed to rule out regeneration of the endogenous beta cells (Baeyens et al., 2005; Rackham et al., 2011). In addition, the endogenous pancreas can be removed for histological examination for insulin-positive cells or for insulin content to be measured (Rackham et al., 2011), although it should be noted that anatomical presence of beta cells is not necessarily correlated to beta cell function (Kargar and Ktorza, 2008).

One disadvantage with chemically inducing diabetes is that the chemicals can be toxic at other organs of the body. It should also be noted that changes in P450 isozymes in the liver, kidney, lung, intestines, testis and brain have been reported after administration of STZ or alloxan, and thus, this should be considered when drugs are being tested in these models (Lee et al., 2010).

STZ

STZ [2-deoxy-2-(3-(methyl-3-nitrosoureido)-d-glucopyranose] is synthesized by Streptomycetes achromogenes. After i.p. or i.v. administration, it enters the pancreatic beta cell through the Glut-2 transporter and causes alkylation of the DNA (Szkudelski, 2001). Subsequent activation of PARP leads to NAD⁺ depletion, a reduction in cellular ATP and subsequent inhibition of insulin production (Sandler and Swenne, 1983). In addition, STZ is a source of free radicals that can also contribute to DNA damage and subsequent cell death. STZ tends to be administered as a single high dose or as multiple low doses.

High-dose STZ

The dose for a single high dose in mice ranges from 100 to 200 mg·kg⁻¹ (Srinivasan and Ramarao, 2007; Dekel et al., 2009), depending on the mouse strain (Hayashi et al., 2006), and in rats 35–65 mg·kg⁻¹ (Srinivasan and Ramarao, 2007). This leads to a rapid ablation of the beta cells and hyperglycaemia. It should be noted, however, that it has been suggested that regeneration of the pancreatic islets can occur after STZ treatment; and thus sufficient controls should be in place to determine that any improvement in glycaemia is not due to spontaneous regeneration of endogenous beta cells (Grossman et al., 2010). High-dose STZ is often used in transplantation models, where islets (Deeds et al., 2011) or putative stem cells (Song et al., 2009) are transplanted under the kidney capsule. It should be noted that STZ has recently been shown to cause lymphopenia and a relative increase in T-regulatory cells (Muller et al., 2011), which could interfere with the interpretation of studies involving immune tolerance to transplants.

Multiple low-dose STZ

STZ can be administered as multiple low doses over 5 days to induce insulitis in mice (Like and Rossini, 1976; Wang and Gleichmann, 1998) or rats (Lukic et al., 1998). Doses range from 20 to 40 mg·kg⁻¹ per day, depending on the species and strain. A reduction in islet number and volume is apparent with concomitant reduction in insulin secretion capacity (Bonnevie-Nielsen et al., 1981). Macrophages are the first cells to infiltrate the islets, and the development of diabetes is dependent on cytokine production (Lukic et al., 1998). Diabetes develops even in the absence of T and B cells, and therefore, it does not model the human disease as closely as spontaneous models of autoimmunity (Reddy et al., 1995). Therapies targeting cytokines (Sandberg et al., 1994) and NO (Flodstrom et al., 1999) tend to be successful in reducing diabetes development in this model, indicating their role in the beta cell destruction.

Alloxan

The diabetic effect of alloxan (2,4,5,6-tetraoxypyrimidine; 5,6-dioxyuracil) is mainly attributed to rapid uptake by the beta cells and the formation of free radicals, which beta cells have poor defence mechanisms to (Nerup et al., 1994). Alloxan is reduced to dialuric acid and then re-oxidized back to alloxan, creating a redox cycle for the generation of superoxide radicals that undergo dismutation to form hydrogen peroxide and thereafter highly reactive hydroxyl radicals that cause fragmentation of beta cell DNA (Szkudelski, 2001). Alloxan is also taken up by the liver, but it has better protection to reactive oxygen species (Malaisse et al., 1982; Mathews and Leiter, 1999) and therefore is not as susceptible to damage. Other mechanisms of beta cell damage by alloxan include oxidation of essential –SH groups, especially that of glucokinase (im Walde et al., 2002) and disturbances in intracellular calcium homeostasis (Kim et al., 1994). Doses in mice range from 50 to 200 mg·kg⁻¹ and in rats from 40 to200 mg·kg⁻¹, depending on the strain and the route of administration with i.p and s.c. administration requiring up to three times as high a dose as the i.v. route (Szkudelski, 2001). A dose of 100 mg·kg⁻¹ has been used to create a long-term diabetes models in rabbits (Wang et al., 2010). It should be noted that alloxan has a narrow diabetogenic dose, and even light overdosing can cause general toxicity, especially to the kidney (Szkudelski, 2001).

Spontaneous autoimmune models of type 1 diabetes

The most commonly used autoimmune models of type 1 diabetes are the non-obese diabetic (NOD) mouse and the Biobreeding (BB) rat (Yang and Santamaria, 2006). In addition, another rat model of autoimmune type 1 the LEW.1AR1/Ztm-iddm rat was described in 2001 (Lenzen et al., 2001). However, the NOD mouse still dominates the literature as the autoimmune model of choice.

NOD mice

The NOD mouse was developed at the Shionogi Research Laboratories in Osaka, Japan in 1974 (Hanafusa et al., 1994). NOD mice develop insulitis at around 3–4 weeks of age. In this pre-diabetic stage, the pancreatic islets are infiltrated by predominately CD4⁺ and CD8⁺ lymphocytes, although B cells and NK cells are also present (Yoon and Jun, 2001).

The insulitis causes beta cell destruction, but the onset of overt diabetes is usually not until apparent until approximately 90% of pancreatic insulin is lost at around 10–14 weeks, although diabetes can develop up to 30 weeks of age. Diabetes is more prevalent in the females with an incidence in females, ranging from 60% to 90% in most colonies, whereas the incidence in males ranges from 10% to 30% in most colonies (Pozzilli et al., 1993; Hanafusa et al., 1994). When these mice become overtly diabetic, they rapidly lose weight and require insulin treatment. The MHC class 2 in NOD mice share structural similarities to that in humans, which may confer resistance or susceptibility to the disease in both NOD mice and humans (Todd and Wicker, 2001; Wicker et al., 2005). This parallel in type 1 diabetes genes between NOD mice and humans has been extremely useful in dissecting some mechanisms and pathways behind type 1 diabetes (Yang and Santamaria, 2006; Driver et al., 2011). Thus, these mice are potentially suitable for testing therapies in which modulation of the autoimmune response is being targeted. It should however been noted that there are a number of drugs that are effective in NOD mice, which were then shown to be ineffective in humans (von Herrath and Nepom, 2009). One of the major issues is the time point of intervention. Many drugs that have been shown to be successful in NOD mice were administered early on, and it has been suggested that is relatively easy to prevent diabetes in young NOD mouse (Roep, 2007). Another difficulty in the translation of therapies tested in NOD mice is that whereas the pancreas of the NOD can be removed for examination after a study, there is a lack of biomarkers in the peripheral blood in humans that could be used to verify the success of the intervention (von Herrath and Nepom, 2005). There are also problems in translating dosing from the NOD mouse to humans (von Herrath and Nepom, 2005).

Development of diabetes is the NOD mouse is negatively associated with microbial exposure; thus, the mice therefore should be kept in specific pathogen-free (SPF) conditions to maintain diabetes incidence. Due to the gender differences, unpredictability of disease onset and the requirement for SPF conditions, these mice are expensive to maintain as a model of type 1 diabetes compared with chemical induction of diabetes.

A more predictable and accelerated onset can be achieved in NOD mice by injecting the mice with cyclophosphamide (Caquard et al., 2010). In addition, adoptive transfer can prove useful, where T cells from diabetic NOD mice are injected into non-diabetic recipient mice, causing the recipient mouse to develop diabetes (Christianson et al., 1993). Also, NOD mice can be used in a recurrence of autoimmunity model, where syngeneic islets from young non-diabetic NOD mice are transplanted into diabetic NOD mice (Rydgren et al., 2007). The graft is rapidly destroyed by autoimmune mechanisms. All three of these models allow the timing of the autoimmune response to be controlled. The NOD mouse is often used in intervention studies in attempts to prevent or delay the onset of the autoimmune disease (Montane et al., 2011; Tai et al., 2011; Lee et al., 2011b).

Strategies to improve the NOD model include using specific genetic manipulation of NOD mice (Yang and Santamaria, 2003) or the creation of humanized mouse models with components of the human immune system (King et al., 2008; Niens et al., 2011). Despite the limitations of the NOD mouse, it is still used extensively as it does represent many aspects of the human disease and is a model that has helped identify many of the genetic and signalling pathways that can lead to type 1 diabetes.

BB rats

BB rats were derived from outbred Wistar rats. Spontaneous autoimmune diabetes in a Canadian colony was first identified in 1974 and lead to the creation of two founder colonies from which all substrains have derived, one inbred (BBDP/Wor) and one outbred (BBdp) (Mordes et al., 2004). Diabetes resistant BB rats have also been bred to act as controls.

BB rats usually develop diabetes just after puberty and have similar incidence in males and females. Around 90% of rats develop diabetes between 8 and 16 weeks of age. The diabetic phenotype is quite severe, and the rats require insulin therapy for survival. Although the animals have insulitis with the presence of T cells, B cells, macrophages and NK cells, the animals are lymphopenic with a severe reduction in CD4⁺ T cells and a near absence of CD8⁺ T cells (Mordes et al., 2004). Lymphopenia is not a characteristic of type 1 diabetes in humans or NOD mice (Mordes et al., 2004) and is seen to be a disadvantage in using the BB as a model of type 1 diabetes in humans. Also, in contrast to NOD mice, the insulitis is not preceded by peri-insulitis. However, the model has been valuable in elucidating more about the genetics of type 1 diabetes (Wallis et al., 2009), and it has been suggested that it may be the preferable small animal model for islet transplantation tolerance induction (Mordes et al., 2004). In addition, BB rats have been used in intervention studies (Hartoft-Nielsen et al., 2009; Holmberg et al., 2011) and studies of diabetic neuropathy (Zhang et al., 2007).

LEW.1AR1/-iddm rats

This rat model of type 1 diabetes arose spontaneously in a colony of congenic Lewis rats with a defined MHC haplotype (LEW.1AR1), which were being bred at the Institute of Laboratory Animal Science of Hannover Medical School (Ztm). These rats exhibit insulitis, and overt diabetes manifests at around 8–9 weeks. Originally, the incidence of diabetes was approximately 20% (Lenzen et al., 2001); however, with further inbreeding of diabetic rats, the incidence increased to around 60% with equal incidence in both genders (Jorns et al., 2005). The animals exhibit a pre-diabetic period with islet infiltration approximately a week before animals become hyperglycaemic. This relatively short pre-diabetic period allows for effective analysis of different stages of the immune cell infiltration (Jorns et al., 2005). In contrast to the NOD mouse and BB rat, the LEW-iddm rat does not exhibit other autoimmune diseases. It also survives well after the onset of overt diabetes and thus can be used to study diabetic complications (Mathews, 2005). However, most studies in this rat model so far have been investigating the mechanisms involved in the development of diabetes (Jorns et al., 2004; 2005; Peschke et al., 2011) and intervention studies (Arndt et al., 2009; Jorns et al., 2010).

Genetically induced insulin-dependent diabetes

AKITA mice

The AKITA mouse was derived in Akita, Japan from a C57BL/6NSlc mouse with a spontaneous mutation in the insulin 2 gene preventing correct processing of pro-insulin. This causes an overload of misfolded proteins and subsequent ER stress. This results in a severe insulin-dependent diabetes starting from 3 to 4 weeks of age, which is characterized by hyperglycaemia, hypoinsulinaemia, polyuria and polydipsia. Untreated homozygotes rarely survive longer than 12 weeks. The lack of beta cell mass in this model makes it an alternative to streptozotocin-treated mice in transplantation studies (Mathews et al., 2002). It has also been used as a model of type 1 diabetic macrovascular disease (Zhou et al., 2011) and neuropathy (Drel et al., 2011). In addition, this model is commonly used to study potential alleviators of ER stress in the islets and in this respect models some of the pathology of type 2 diabetes (Chen et al., 2011).

Virus-induced models of diabetes

Viruses have been implicated in the pathogenesis of type 1 diabetes (van der Werf et al., 2007). Therefore, several animal models have used viruses to initiate beta cell destruction. The destruction can be either due to direct infection of the beta cell or initiation of an autoimmune response against the beta cell (Jun and Yoon, 2003). Viruses used to induce diabetes in animal models include coxsackie B virus (Yoon et al., 1986; Kang et al., 1994; Jaidane et al., 2009), encephalomyocarditis virus (Craighead and McLane, 1968; Baek and Yoon, 1991; Shimada and Maruyama, 2004) and Kilham rat virus (Guberski et al., 1991; Ellerman et al., 1996).

In addition, a transgenic virus model has been described in which a defined viral antigen (the nucleoprotein or glycoprotein) of lymphocytic choriomeningitis virus (LCMV) is expressed under the rat insulin promoter (von Herrath et al., 1997). These mice do not spontaneously develop any signs of beta cell destruction, but if the mice are then injected with LCMV, the immune response cross-reacts with the antigen expressed in the beta cells, leading to beta cell destruction.

However, the virus-induced model can be complicated as the outcome is dependent on replication levels of the virus as well as timing of the infection. Indeed, it has been shown that viruses can both induce autoimmunity as well as prevent it, depending on the conditions (von Herrath et al., 2011). Although some cases of human type 1 diabetes have been linked to viruses (van der Werf et al., 2007; Richardson et al., 2009), it is unclear to what extent viruses are involved in the pathogenesis of type 1 diabetes.

Non-rodent models of type 1 diabetes

In addition to the extensively studied rodent models of type 1 diabetes, several large animal models have been developed. In large animal models, spontaneous diabetes is relatively rare and unpredictable in onset, and thus, induced models of type 1 diabetes are required. The most common method of inducing insulin dependence in large models is either by pancreatectomy or STZ.

Pancreatectomy

Pancreatectomy has been used to induce hyperglycaemia in pigs (Morel et al., 1991; Mellert et al., 1998), dogs (Fisher et al., 2001) and primates (He et al., 2011). When carried out by a skilled surgeon, this model is a reliable method to induce hyperglycaemia. However, this is very invasive surgery for the animal, increases the chances of hypoglycaemia and also leads to pancreatic exocrine deficiency in the animal (He et al., 2011). However, pancreatectomy in pigs followed by autotransplantation of the isolated islets (Emamaullee et al., 2009) is a reasonably accurate model of autotransplantation of islets in humans (Matsumoto, 2011).

Chemical ablation of beta cells in large animals

There is interspecies variation in the beta cell toxicity of alloxan (Tyrberg et al., 2001) and STZ (Eizirik et al., 1994; Dufrane et al., 2006), which may be due to differences in expression in GLUT-2 (Dufrane et al., 2006). It has been reported that while a dose of 50 mg·kg⁻¹ can produce irreversible diabetes in the rat and Cynomolgus monkey, pigs require a higher dose (150 mg·kg⁻¹) and despite this, a partial correction to hyperglycaemia was seen in pigs 4 weeks after STZ injection (Dufrane et al., 2006). Increasing the dose to 200 mg·kg⁻¹ in pigs leads to renal and hepatic toxicity, suggesting a narrow window of efficacy. It should be noted that other studies have successfully used 150 mg·kg⁻¹ STZ in pigs (Grussner et al., 1993; Jensen-Waern et al., 2009), thus underlining the difficulties in establishing a STZ-induced model of diabetes in larger animals.

Some models in higher animals combine a partial pancreatectomy with STZ treatment, thus reducing the dose of STZ (Wise et al., 1985; He et al., 2011). In addition, a multiple low-dose STZ model has been described in primates (Wei et al., 2011).

Animal models of type 2 diabetes

Type 2 diabetes is characterized by insulin resistance and the inability of the beta cell to sufficiently compensate. Therefore, animal models of type 2 diabetes tend to include models of insulin resistance and/or models of beta cell failure. Many animal models of type 2 diabetes are obese, reflecting the human condition where obesity is closely linked to type 2 diabetes development. Some of the most commonly used models for type 2 diabetes are outlined in Table 2.

Table 2.

Summary of rodent models of type 2 diabetes

Induction mechanism	Model	Main features	Possible uses
Obese models (monogenic)	Lepob/ob mice	Obesity-induced hyperglycaemia	Treatments to improve insulin resistance
	Leprdb/db mice		Treatments to improve beta cell function
	ZDF Rats
Obese models (polygenic)	KK mice	Obesity-induced hyperglycaemia	Treatments to improve insulin resistance
	OLETF rat		Treatments to improve beta cell function
	NZO mice		Some models show diabetic complications
	TallyHo/Jng mice
	NoncNZO10/LtJ mice
Induced obesity	High fat feeding (mice or rats)	Obesity-induced hyperglycaemia	Treatments to improve insulin resistance
	Desert gerbil		Treatments to improve beta cell function
	Nile grass rat		Treatments to prevent diet-induced obesity
Non-obese models	GK rat	Hyperglycaemia induced by insufficient beta cell function/mass	Treatments to improve beta cell function
			Treatments to improve beta cell survival
Genetically induced models of beta cell dysfunction	hIAPP mice	Amyloid deposition in islets	Treatments to prevent amyloid deposition
			Treatments to improve beta cell survival
	AKITA mice	Beta cell destruction due to ER stress.	Treatments to prevent ER stress
			Treatments to improve beta cell survival

Obese models of type 2 diabetes

As type 2 diabetes is closely linked to obesity, most of the current animal models of type 2 diabetes are obese. Obesity can be the result of naturally occurring mutations or genetic manipulation. Alternatively, obesity can be induced by high fat feeding.

Monogenic models of obesity

Although obesity in humans is rarely caused by a monogenic mutation, monogenic models of obesity are commonly used in type 2 diabetes research. The most widely used monogenic models of obesity are defective in leptin signalling. Leptin induces satiety, and thus, a lack of functional leptin in these animals causes hyperphagia and subsequent obesity. These models include the Lep^ob/ob mouse, which is deficient in leptin and the Lepr^db/db mouse and Zucker Diabetic Fatty rat, which are deficient in the leptin receptor. These models are often used to test new therapies for type 2 diabetes (Yoshida et al., 2010; Gault et al., 2011; Park et al., 2011).

Lep^ob/ob mouse

The Lep^ob/ob mouse is a model of severe obesity and derives from a spontaneous mutation discovered in an outbred colony at Jackson Laboratory in 1949. The phenotype was bred into C57BL/6 mice, but it was not until 1994 the mutated protein was identified as leptin (Zhang et al., 1994). The weight increase starts at 2 weeks of age, and the mice develop hyperinsulinaemia. By 4 weeks, hyperglycaemia is apparent, with blood glucose concentrations continuing to rise, peaking at 3–5 months, after which they fall as the mouse becomes older (Lindstrom, 2007). Other metabolic aberrations include hyperlipidaemia, a disturbance in temperature regulation and lower physical activity (Lindstrom, 2007). In addition, these mice are infertile (Chehab et al., 1996).

The pancreatic islet volume is dramatically increased in these mice (Bock et al., 2003). Although there are some abnormalities in insulin release (Lavine et al., 1977), islets maintain insulin secretion, and the lack of complete beta cell failure in this model means diabetes is not particular severe and indeed not completely representative of human type 2 diabetes. It should be noted that on the C57Bl/KS background, a much more severe diabetes develops with regression of islets and early mortality (Coleman, 1978).

Lepr^db/db mice

The Lepr^db/db mouse originated from the Jackson Laboratory (Hummel et al., 1966) and is due to an autosomal recessive mutation in the leptin receptor (Chen et al., 1996). These animals are hyperphagic, obese, hyperinsulinaemic and hyperglycaemic. Obesity is evident from 3–4 weeks of age with hyperinsulinaemia becoming apparent at around 2 weeks of age and hyperglycaemia developing at 4–8 weeks. The most commonly used background used is on the C57BLKS/J, and they develop ketosis after a few months of age and have a relative short lifespan (Srinivasan and Ramarao, 2007).

Zucker fatty rats and Zucker diabetic fatty rats

The Zucker Fatty rats were discovered in 1961 after a cross of Merck M-strain and Sherman rats. They have a mutated leptin receptor (Phillips et al., 1996) that induces hyperphagia, and the rats become obese at around 4 weeks of age. These rats also are hyperinsulinaemic, hyperlipidaemic and hypertensive. They have impaired glucose tolerance (Srinivasan and Ramarao, 2007).

A mutation in this strain lead to the derivation of a substrain with a diabetogenic phenotype: the inbred Zucker Diabetic Fatty Rats (ZDF). These rats are less obese than the Zucker fatty rats but have more severe insulin resistance, which they are unable to compensate for due to increased apoptosis levels in the beta cells (Pick et al., 1998). This is characterized by initial hyperinsulinaemia at around 8 weeks of age followed by decreased insulin levels (Shibata et al., 2000). Diabetes usually develops at around 8–10 weeks in males, but females do not develop overt diabetes (Srinivasan and Ramarao, 2007). These rats also show signs of diabetic complications (Shibata et al., 2000).

Polygenic models of obesity

Polygenic models of obesity may provide a more accurate model of the human condition. A variety of different polygenic mouse models of obesity, glucose intolerance and diabetes exist, allowing a variety of genotypes and susceptibilities to be studied. However, unlike the monogenic models, there are no wild-type controls. In addition, the male sex bias is more extreme in these models (Leiter, 2009). These polygenic models have been used in a wide variety of studies that have aimed to reverse the symptoms of type 2 diabetes (Chen et al., 2009; Fukaya et al., 2009; Guo et al., 2010; Mochizuki et al., 2011; Yoshinari and Igarashi, 2011), understand more about the interplay of obesity and glucose homeostasis (Kluth et al., 2011) (Jurgens et al., 2007) or study diabetic complications (Cheng et al., 2007; Fang et al., 2010; Buck et al., 2011; Lee et al., 2011a).

KK mice

KK mice are a mildly obese and hyperleptinaemic strain derived from wild-derived ddY mice in Japan by Kondo in 1957 (Clee and Attie, 2007). They develop severe hyperinsulinaemia and demonstrate insulin resistance in both muscle and adipose tissue. The pancreatic islets are hypertrophic and degranulated. This mouse strain also shows signs of diabetic nephropathy (Ikeda, 1994).

A derivative of this strain is the KK-A^Y mice, which was created by introducing the yellow obese A^Y gene into the KK strain (Chakraborty et al., 2009). This model develops maturity-obesity and has more severe hyperinsulinaemia and more prominent changes in the pancreatic islets. This is due to the ectopic expression of the agouti protein antagonizing the melanocortin receptor 4 (MCR4) in the hypothalamus.

OLETF rats

The Otsuka Long-Evans Tokushima Fat rat (OLETF) was derived from a spontaneously diabetic rat discovered in 1984 in an outbred colony of Long Evans Rats. Selective breeding at the Tokushima Research Institute lead to the OLETF strain that has mild obesity and late onset hyperglycaemia (after 18 weeks). Diabetes is inherited by the males. The pancreatic islets undergo three stages of histological change. At an early stage (6–20 weeks old), cellular infiltration and degeneration is seen. This is followed by a stage of hyperplasia between 20 and 40 weeks. The final stage is characterized by islets becoming fibrotic and become replaced by connective tissue (Kawano et al., 1994). These rats also exhibit renal complications (Lee et al., 2011a).

New Zealand Obese (NZO) mice

The NZO mouse is a polygenic model of obesity, created by selective breeding. It is hyperphagic and obese, which may be a consequence of leptin resistance as these mice are hyperleptinaemic by 9–12 weeks of age (Leiter and Reifsnyder, 2004). Indeed, they are resistant to peripheral leptin administration but sensitive to centrally administrated leptin (Halaas et al., 1997), indicating a defect in leptin transport across the blood–brain barrier. These mice are also hyperinsulinaemic, which stems from hepatic insulin resistance from an early age, which seems to result from impaired regulation of hepatic fructose-1,6-bisphosphatase (Andrikopoulos et al., 1993). Blood glucose concentrations are elevated, and they show impaired glucose tolerance, which worsens with age and approximately 50% of males develop diabetes (Haskell et al., 2002). Islets are hyperplastic and hypertrophic at 3–6 months of age, but beta cell loss occurs at later time points and there are signs of ‘latent autoimmune diabetes of adults’ (Junger et al., 2002).

TallyHo/Jng mice

The TallyHo mouse is a naturally occurring model of obesity and type 2 diabetes derived from selective breeding of mice that spontaneously developed hyperglycaemia and hyperinsulinaemia in an outbred colony of Theiler Original mice (Kim et al., 2005). In these mice, adiposity is increased, and plasma triglycerides, cholesterol and free fatty acid levels are elevated. Hyperglycaemia is limited to male mice, which develops as early as between 10 and 14 weeks of age. The pancreatic islets are hypertrophied and degranulated, and hyperinsulinaemia is evident. The TallyHo mouse has not yet been completely characterized for diabetic complications (Leiter, 2009), although a recent study has used this model to study diabetic wound healing (Buck et al., 2011).

NoncNZO10/LtJ mice

This strain was created by combining independent diabetes risk-conferring quantitative trait loci from two unrelated strains of NZO mice with nonobese nondiabetic mice (NON/Lt) (Cho et al., 2007). These mice develop liver and skeletal muscle insulin resistance at 8 weeks before developing chronic hyperglycaemia from about 12 weeks (Leiter, 2009). Islet mass initially increases before beta cell loss occurs. Diabetic nephropathy has been observed in some males aged about one year (Leiter, 2009), and it has also been suggested that this model is suitable for studies in diabetic wound healing (Fang et al., 2010)

High fat feeding

The model of high fat feeding to C57BL/6 mice was first described in 1988 (Surwit et al., 1988). High fat feeding can lead to obesity, hyperinsulinaemia and altered glucose homeostasis due to insufficient compensation by the islets (Winzell and Ahren, 2004). Normal chow (on a caloric basis usually around 26% protein, 63% carbohydrate and 11% fat) is exchanged for a diet where the number of calories from fat is increased substantially (around 58% of energy derived from fat). The amount of food eaten should be monitored to ensure that the mice do not compensate by eating less. It has been shown that high-fat-fed mice can weigh more than chow-fed controls within a week of starting the high-fat diet (Winzell and Ahren, 2004), although typically mice are fed the high-fat diet for several weeks to induce a more pronounced weight gain. The weight gain is associated with insulin resistance, and lack of beta cell compensation leads to impaired glucose tolerance.

Since the obesity is induced by environmental manipulation rather than genes, it is thought to model the human situation more accurately than genetic models of obesity-induced diabetes. High fat feeding is often used in transgenic or knock-out models, which may not show an overt diabetic phenotype under normal conditions, but when the beta cells are ‘pushed’, the gene may be shown to be of importance. It should be noted that the background strain of the mice can determine the susceptibility to diet-induced metabolic changes, and thus, effects could be missed if a more resistant strain is used (Surwit et al., 1995; Bachmanov et al., 2001; Almind and Kahn, 2004). It has also been reported that there is heterogeneity of the response to high fat feeding within the inbred C57BL/6 strain, indicating that differential responses to a high-fat diet are not purely genetic (Burcelin et al., 2002).

Other diet-induced rodent models of type 2 diabetes

Although rats and mice are the most commonly used models for studies of type 2 diabetes, other rodents have also been identified as useful models. These include the desert gerbil and the newly described Nile grass rat, both of which tend to develop obesity in captivity.

Desert gerbil

The desert gerbil (Psammomys obesus) was originally discovered to develop diabetes in captivity in the 1960s. The diabetes ranged from mild hyperglycaemia with hyperinsulinaemia to severe hyperglycaemia with hypoinsulinaemia and ketoacidosis. Indeed, four stages were subsequently identified in the Jerusalem colony: stage A (normal glycaemic normoinsulinaemic), stage B (normoglycaemic hyperinsulinaemic), stage C (hyperglycaemic and hyperinsulinaemic) and stage D (hyperglycaemic insulinopenic). Progression from stage A to stage C can be prevented with food restriction, but there is no recovery from stage D (Shafrir et al., 2006). This animal is not hyperphagic, but when high-energy nutrition is made available in a laboratory setting, obesity, hyperinsulinaemia and subsequently diabetes develop (Ziv et al., 1999). Due to its poor adaption to excess nutrition, it has been suggested that the Psammomys represents an ideal model of the ‘thrifty gene’ effect and could be used for studying populations where insulin resistance and metabolic syndrome is common after a rapid evolution from scarcity to nutritional abundance. Researchers have used these animals in studies that aim to prevent nutritionally induced diabetes (Mack et al., 2008; Bodvarsdottir et al., 2010; Vedtofte et al., 2010).

Nile grass rat

The Nile grass rat (Arvicanthis niloticus) has recently been suggested as a model for metabolic syndrome (Noda et al., 2010). Most of these animals spontaneously develop obesity, dyslipidaemia and hyperglycaemia by one year of age when kept on a normal chow diet in captivity. They show other signs of diabetes and metabolic syndrome such as reduced beta cell mass, atherosclerosis and liver steatosis.

Nonobese models of type 2 diabetes

Not all type 2 diabetes patients are obese, and thus, it is important that lean animal models of type 2 diabetes are also studied. These include models that have beta cell inadequacy, which is what ultimately leads to overt type 2 diabetes in humans (Weir et al., 2009).

Goto–Kakizaki rats

Goto–Kakizaki (GK) rats were created by a Japanese group by repetitive breeding of Wistar rats with the poorest glucose tolerance (Goto et al., 1976). This leads to the development of a lean model of type 2 diabetes, which is characterized by glucose intolerance and defective glucose-induced insulin secretion. The development of insulin resistance does not seem to be the main initiator of hyperglycaemia in this model, and the defective glucose metabolism is regarded to be due to aberrant beta cell mass (Portha et al., 2001) and/or function (Ostenson and Efendic, 2007).

However, islet morphology and metabolism seem to differ between differing colonies of these rats. In some colonies (Stockholm and Dallas), the volume and density of beta cells is similar to controls; thus the hyperglycaemia seems to result from insulin secretory defects. However, in the Paris colony of GK rats, it has been reported that there is a reduction in beta cell mass (Ostenson and Efendic, 2007). GK rats have been used in studies ranging from investigations of beta cell dysfunction in type 2 diabetes (Movassat et al., 2007; Ehses et al., 2009; Portha et al., 2009; Dolz et al., 2011; Giroix et al., 2011) to diabetic complications (Liepinsh et al., 2009; Carneiro et al., 2010; Okada et al., 2010).

hIAPP mice

A characteristic of type 2 diabetes is the formation of amyloid within the islet tissue, which derives from islet amyloid polypeptide (IAPP). Rodent IAPP is not amyloidgenic, and thus, rodents normally do not model this aspect of the disease (Hoppener et al., 1994). However, transgenic mice have been created to express human IAPP (hIAPP) under the insulin promoter (Matveyenko and Butler, 2006), which can form amyloid within the islets. A variety of hIAPP models have been created, and it has been demonstrated that increasing the expression of hIAPP increases beta cell toxicity (Matveyenko and Butler, 2006). In addition, replicating beta cells are more susceptible to hIAPP toxicity, and thus, beta cell adaption to increased insulin demand in this model is restricted (Matveyenko et al., 2009).

Non-rodent models of type 2 diabetes

Larger animals have also been utilized in type 2 diabetes research. Type 2 diabetes in cats resembles the human condition in several aspects, including clinical, physiological and pathological aspects. Some characteristics common to humans include that type 2 diabetes in cats develops in middle age, is associated with obesity and insulin resistance, and subsequent beta cell loss occurs (O’Brien, 2002). In addition, cats are one of the few species other than humans and macaques that form amyloid in islets, making them a good model for studying islet amyloidosis (Henson and O’Brien, 2006). Old-world non-human primates can also develop type 2 diabetes, which has similarities to the human condition and thus be useful as a model (Wagner et al., 2006). In addition, several strains of pigs have a phenotype that resembles type 2 diabetes (Bellinger et al., 2006). A novel model of obesity and mild type 2 diabetes has recently been established in the dog (Ionut et al., 2010) by combining a high-fat diet with STZ.

Models of beta cell function

The beta cell plays a central role in the development of both type 1 and type 2 diabetes as well as playing a key role in less common classifications of diabetes such as maturity onset diabetes of the young (MODY), gestational diabetes, neonatal diabetes and other beta cell syndromes such as hyperinsulinism. Therefore, models of beta cell function are highly relevant in understanding pathways that can lead to the inability of beta cells to secrete appropriate amounts of insulin. Such models are often genetically manipulated, such as mutations of Kir6.2 to study K_ATP channel function (Girard et al., 2009) or mutations in glucose kinase to understand the function of the glucose sensor in beta cells (Fenner et al., 2011). A role for serotonin in the expansion of islets in pregnancy has recently been elucidated by studying the islets of mice lacking the serotonin receptor Htr2b (Kim et al., 2010). Studies such as these can increase our knowledge of beta cell function and its role in a variety of conditions. However, it should be pointed out that the same mutation in humans can lead to different symptoms in mice as recently shown by Hugill et al., where a mutation in Kcnj11 (encoding a subunit of the K_ATP channel) caused hypersecretion of insulin and hypoglycaemia in their patient, but glucose intolerance and reduced insulin secretion in mice (Hugill et al., 2010). However, this may prove useful in understanding the transition from hyperinsulinism of infancy (HI) to diabetes in some patients (Hugill et al., 2010).

Models of beta cell regeneration

Both type 1 and type 2 diabetes can be regarded as conditions when beta cell mass is insufficient to cope with demand for insulin. Therefore, a lot of effort has focused on understanding how beta cell mass is regulated. In these models, beta cell mass tends to be partially or nearly completely ablated by a variety of means, and the mechanisms of beta cell mass renewal are investigated.

Pancreas injury models

Pancreas injury models are often used in studies investigating the regenerative capacity of beta cells or beta cell progenitors. These models of injury include pancreatectomy and duct ligation and due to technical difficulties are more frequently carried out in rats than in mice. Sixty percent pancreatectomy does not lead to an increase in blood glucose concentrations, and there is only a moderate increase in beta cell mass (Leahy et al., 1988). However, a 90% pancreatectomy in the rat leads to moderate hyperglycaemia followed by extensive regeneration of the pancreas (Bonner-Weir et al., 1983). By 60 h after pancreatectomy, duct-enriched areas appear. By 4 weeks, the endocrine portion of the pancreas has increased by eightfold and the exocrine portion by sixfold.

Ligation of the tail portion of the pancreas, which accounts for 50–60% of the pancreas, leads to a significant decrease in mass of this part of the pancreas (Wang et al., 1995). The acinar tissue in the tail portion is replaced by small ductal structures by day 3, which is associated with a fibrotic and inflammatory reaction. By day 5, the tissue predominately consists of ductal tissue and small islets in connective tissue. In the first week after the duct ligation, the beta cell mass in the tail portion nearly doubles, making it an ideal model to study beta cell regeneration. In this model, blood glucose levels in the animal do not rise; indeed, for 2 weeks after the duct ligation, blood glucose levels were significantly lower (Wang et al., 1995). A disadvantage of these models is the invasiveness, which makes them technically difficult and a rather extreme model so any regeneration seen is not physiological. However, they have lead to useful information on the regenerative capacity of the pancreas, at least in rodents.

Neonatal STZ administration

STZ administration to neonatal rats (2 days old) induces a regeneration and/or type 2 diabetes model in the adults (Portha et al., 1974). In this model, a peak of hyperglycaemia is seen 2 days after STZ administration (100 mg·kg⁻¹, i.v. or i.p.), which is followed by regeneration of beta cells and normoglycaemia by day 10. However, hyperglycaemia returns by 6 weeks, which is thought to be due to inadequate beta cells mass and beta cell dysfunction (Bonner-Weir et al., 1981). Therefore, in the later phase, this model can be used to study type 2 diabetes (Tourrel et al., 2001).

Regeneration after ablation of beta cells using genetic approaches

It has been noted that genetic ablation of beta cells can be followed by extensive beta cell regeneration. However, complete beta cell ablation from birth is not compatible with life as indicated by insulin knock-out mice (Duvillie et al., 1997) and pdx-1 knock-out mice (Jonsson et al., 1994; Duvillie et al., 1997). Therefore, genetically modified mice have been created to allow ablation of beta cells in adult mice.

In these models of beta cell ablation, the gene itself does not induce beta cell death unless a specific compound is injected to activate the gene. This allows temporal control of the beta cell death. Interestingly, these models often show recovery with extensive beta cell replication and thus can be used to study beta cell regeneration (Herrera et al., 1994; Nir et al., 2007; Cano et al., 2008; Wang et al., 2008). Some examples are outlined below.

Doxycycline-induced expression of diphtheria toxin in beta cells

In these mice, a reverse tetracycline-dependent transactivator (rtTA) is expressed in beta cells (Insulin-rtTA; in which rtTA expression is driven by 9.5 kb of the 5′ flanking region of the rat insulin II gene). In addition, diphtheria toxin A is expressed under an rtTA responsive promoter. Administration of doxycycline to the mice causes rtTA to induce the expression of diphtheria toxin A, causing beta cell apoptosis (Nir et al., 2007). Approximately 80% of beta cells were lost, and the mice became overtly diabetic. However, on doxycycline withdrawal, beta cell regeneration was evident with beta cell mass increasing and mice reverting to normoglycaemia.

Diphtheria toxin receptor-RIP mice

In this model, the beta cells have been genetically modified to express the diphtheria toxin receptor under the rat insulin promoter. Thus, diphtheria toxin can be administered and will selectively ablate the insulin producing cells as mouse cells do not normally express the DT receptor, and thus, other cells are not susceptible to damage (Buch et al., 2005). Using this method, >99% of the beta cells are ablated. There is extensive beta cell regeneration after ablation, and by 10 months, there is between a 10- and 44-fold increase in beta cells. However, this corresponds to just 4–17% beta cell mass compared with non-treated control mice (Herrera et al., 1994).

PANIC-ATTAC mice

The PANIC-ATTAC (pancreatic islet b-cell apoptosis through targeted activation of caspase 8) mouse is a model for inducible and reversible ablation of beta cells. A mutated FK506 binding protein (FKBP) that is fused to caspase 8 is expressed under the insulin promoter. The expression of the FKBP-caspase 8 protein in its monomeric form does not give rise to a phenotype, but dimerization, by administration of a dimerizer compound, leads to apoptosis. Up to a 90% reduction in pancreatic insulin content can be observed with histology, also demonstrating a marked reduction in beta cells. The mice become hyperglycaemic, but there is a recovery, and by 8 weeks, the mice return to normoglycaemia, although still show signs of impaired glucose-stimulated insulin secretion (Wang et al., 2008).

Knock-out and transgenic mice in diabetes research

Transgenic mice have been used to create specific models of type 1 and type 2 diabetes, including hIAPP mice, humanized mice with aspects of the human immune system and mice allowing conditional ablation of beta cells, as outlined above. Beta cells expressing fluorescent proteins can also provide elegant methods of tracking beta cells for use in diabetes research (Hara et al., 2003).

In addition, knock-out and transgenic mice have become powerful tools in elucidating the influence of specific genes in glucose metabolism and the pathogenesis of diabetes. This includes understanding which transcription factors are involved in pancreas development (Habener et al., 2005) and elucidation of insulin signalling pathways (Kahn, 2003; Wang and Jin, 2009). Tissue-specific knockouts have proven to be particularly useful in studying insulin signalling (Neubauer and Kulkarni, 2006) as the global insulin receptor knock-out is non-viable (Accili et al., 1996).

As mentioned previously with diet-induced metabolic aberrations, caution should be used when interpreting data as different background strains have differing susceptibilities to obesity and alterations in blood glucose homeostasis. This could potentially lead to the effect of a gene being missed or alternatively the relative importance of a gene being overestimated. An example of the impact of background strain of the mouse has been demonstrated using mice heterozygous for the insulin receptor knockout and heterozygous for the insulin receptor substrate-1 (IRS-1) knockout. When the background is C57BL/6, 85% of the mice are overtly diabetic by 6 months, whereas in 129Sv and DBA mice the incidence of overt diabetes is much lower at 2% and 64% respectively (Kulkarni et al., 2003).

Another consideration with knock-out and transgenic mice in diabetes research is that pancreatic promoters such as the rat insulin promoter (RIP), Ngn3 and Pdx-1 can be expressed at low levels in the hypothalamus (Song et al., 2010). Indeed, it has been suggested that RIP-Cre mice per se have disturbed glucose tolerance (Lee et al., 2006), although another study did not see any metabolic aberration in their RIP-Cre mice (Fex et al., 2007). It was suggested this was due to genetic differences as their mice had been rigorously backcrossed onto a C57BL/6 background. Nevertheless, this underlines the importance of including control RIP-Cre mice in experiments.

It is likely that more knock-out models of interest for the study of diabetes will be created due to the efforts of the International Knockout Mouse Consortium (IKMC), which aims to mutate all protein-coding genes in the mouse (http://www.knockoutmouse.org/). In addition, the Mouse Genome Informatics website is a rich source of information that provides an international database resource for the laboratory mouse, providing integrated genetic, genomic and biological data to facilitate the study of human health and disease (http://www.informatics.jax.org/).

Modelling genome-wide association studies genes in mice

Recently, genome-wide association studies (GWAS) in humans have led to a number of susceptibility loci in the pathogenesis of diabetes and obesity to be identified. Such studies have been instrumental in identifying susceptibility genes in the disease. but their function has not always been clear. Mouse models have allowed mechanistic studies to be carried out to elucidate the function of genes identified in GWAS such as FTO (Church et al., 2009; 2010) and the SLC30A8 gene that encodes the zinc transporter (ZnT8) (Wijesekara et al., 2010). The use of mouse models in the interpretation of human GWAS in type 2 diabetes and obesity has recently been elegantly reviewed by Cox and Church (2011).

End-points to study in animal models of diabetes

When testing therapies in animal models of diabetes, the most common end-point of measurement is blood glucose concentrations. It should be pointed out that different species tend to have different blood glucose concentrations than humans, and thus, definitions for diabetes in humans should not necessarily be applied to animals. For example, mice tend to have higher blood glucose concentrations than humans, and it has been suggested that a non-fasting blood glucose concentration over 250 mg·dL⁻¹ (13.8 mM) or preferably a chronic elevation over 300 mg·dL⁻¹ (16.7 mM) is appropriate to consider a mouse diabetic (Leiter, 2009). Normal mice fasted for 16 h during the entire dark period when they usually feed and usually have blood glucose of between 50 and 100 mg·dL⁻¹ (2.8–5.6 mM), whereas mice with type 2 diabetes will have fasting blood glucose levels of around 150–300 mg·dL⁻¹ (8.3–16.7 mM).

Detection of glucose in the urine can also be measured as a sign of diabetes. However, in a complex disease such as diabetes, other end-points should be investigated. Other end-points will depend on the putative mechanism of the drug and the model being used. In models of type 1 diabetes, it is critical that the animals are also weighed to ensure that decreased blood glucose concentrations are not associated with weight loss. This indicates that decreases in blood glucose concentrations are not due to toxic effects of the therapy and possible cessation of eating. However, it should be noted that in models of type 2 diabetes, the mechanism of the drug lowering blood glucose levels may include weight loss (Knudsen, 2010). Glucose tolerance tests are often used to investigate beta cell function. This can allow impaired glucose tolerance to be identified, which is generally regarded as a pre-diabetic state. This is often done after an overnight fast, although it should be noted that it has been suggested that such a prolonged fast may be inappropriate in mice as it induces a metabolic stress and enhances insulin action (McGuinness et al., 2009), and thus, a 6 h fast may be preferable. There are no clear definitions of impaired glucose tolerance for rodents, but in normal nondiabetic mice, an IPGTT reaches a peak at 15–30 min; and by 120 min, the blood glucose should be close to the baseline value (Leiter, 2009). Serum insulin or c-peptide levels can be measured to indicate beta cell function, although high insulin levels can indirectly indicate insulin resistance. An insulin tolerance test can be carried out as an approximate measure of insulin resistance, or a more elegant hyperinsulinaemic–euglycaemic clamp can be carried out (Declercq et al., 2010). It should be noted that surrogate measures of insulin sensitivity such as the homeostasis model index of insulin resistance (HOMA-IR) can be used in rodents, although species specific adjustments may need to be made (Mather, 2009).

Pancreas histology can be used to study the effects of a therapy on the islets, which may be particularly relevant to study in interventions on insulitis (Tian et al., 2010). Whole pancreas insulin content can also be measured to indicate beta cell mass, although ideally morphometric analysis is preferable (Montanya and Tellez, 2009). Islets can also be isolated and insulin secretion experiments carried out ex vivo (Szollosi et al., 2010).

The time course of the disease should also be carefully considered when considering end-points of a study. Some models of type 2 diabetes show beta cell expansion and hyperinsulinaemia prior to subsequent beta cell failure, and the stage of disease may affect the parameters that are being measured. It should also be noted that in humans, type 2 diabetes tends to present later in life, and thus, the use of older mice when studying this condition should be considered.

Choosing an appropriate animal model for diabetes research

A variety of animal models of type 1 and type 2 diabetes are described above, each with their own characteristics. There are several different purposes that these models of diabetes could be used for including pharmacological testing, studies of genetics and understanding disease mechanisms. The choice of model will depend on the purpose of the study. For example, in the case of pharmacological testing, the putative mechanism of the drug being tested will be instrumental in choosing an appropriate animal model.

In type 1 diabetes, the main determinant in choosing an animal model is whether a model of autoimmunity is required. The timing and predictability of onset is also variable in different models of type 1 diabetes.

In type 2 diabetes, it is important to consider the mechanisms underlying the hyperglycaemia and whether this is relevant to your study. These mechanisms can include insulin resistance and/or beta cell failure. Indeed, to determine whether a drug intervention can improve symptoms in any given model may depend on whether beta cells have failed. Animal models of type 2 diabetes can be divided into those that are obese and those that are nonobese. The majority of type 2 diabetes models are obese, by either genetic or dietary means. However, this usually comes with a variety of associated pathologies such as dyslipidaemia and artherosclerosis. Although these co-morbidities are common in some humans with type 2 diabetes, it only represents a portion of the diabetic population. Also, it should be noted that not all animal models of diabetes and strains develop diabetic complications (e.g. the C57BL/6 strain is relatively resistant to nephropathy)(Brosius III et al., 2009a), so care should be taken in choosing an appropriate model if the end-point of the study is to investigate diabetic complications such as nephropathy or neuropathy (Breyer et al., 2005; Sullivan et al., 2007; 2008; Brosius III et al., 2009b).

Strain and species differences should also be carefully considered when choosing a model as different species and background strains have different susceptibilities to diabetes and treatments. Ideally, more than one species or strain should be investigated. Gender should also be taken into account (Franconi et al., 2008), with many models described above having a gender bias (e.g. NOD, NZO and TallyHo mice; OLETF, Zucker Diabetic rats), which does not exist in humans. In addition, many knock-out and transgenic models of diabetes show a gender bias (Franconi et al., 2008). It has been suggested that in some cases this is due to the effects of sex hormones (Inada et al., 2007), although the exact mechanism of gender bias has not been elucidated. Indeed, sex hormone effects can be contradictory in different mouse models with, for example, male gonadectomy protecting against diabetes in some models while being ineffective or increasing incidence in other models (Franconi et al., 2008). Indeed, gender bias could also involve mitochondria and stress responses (Franconi et al., 2008). Care should also be taken when using knock-out and transgenic mice to ensure that potential hypothalamic expression is not affecting the phenotype, and relevant controls should be included.

Models also differ in their physiological relevance. with some models more closely resembling disease development than others. Some models such as those of pancreas regeneration are rather extreme, and it remains to be elucidated whether the mechanisms of beta cell expansion in these models can play a role in humans. Indeed, when choosing a model for either type 1 or type 2 diabetes, it is highly desirable that a variety of different models are used to represent the diversity seen in human diabetic patients.

Glossary

NOD

non-obese diabetic

SPF

specific pathogen-free

STZ

streptozotocin

Conflict of interest

None.