Preclinical rodent models of physical inactivity-induced muscle insulin resistance: challenges and solutions

Abstract

Physical inactivity influences the development of muscle insulin resistance yet is far less understood than diet-induced muscle insulin resistance. Progress in understanding the mechanisms of physical inactivity-induced insulin resistance is limited by a lack of an appropriate preclinical model of muscle insulin resistance. Here, we discuss differences between diet and physical inactivity-induced insulin resistance, the advantages and disadvantages of the available rodent inactivity models to study insulin resistance, and our current understanding of the mechanisms of muscle insulin resistance derived from such preclinical inactivity designs. The burgeoning rise of health complications emanating from metabolic disease presents an alarming issue with mounting costs for health care and a reduced quality of life. There exists a pressing need for more complete understanding of mechanisms behind the development and progression of metabolic dysfunction. Since lifestyle modifications such as poor diet and lack of physical activity are primary catalysts of metabolic dysfunction, rodent models have been formed to explore mechanisms behind these issues. Particularly, the use of a high-fat diet has been pervasive and has been an instrumental model to gain insight into mechanisms underlying diet-induced insulin resistance (IR). However, physical inactivity (and to some extent muscle disuse) is an often overlooked and much less frequently studied lifestyle modification, which some have contended is the primary contributor in the initial development of muscle IR. In this mini-review we highlight some of the key differences between diet- and physical inactivity-induced development of muscle IR and propose reasons for the sparse volume of academic research into physical inactivity-induced IR including infrequent use of clearly translatable rodent physical inactivity models.

INTRODUCTION

The main causes of obesity and metabolic diseases are attributable to lifestyle, namely the balance of physical activity and energy intake (1). Physical activity is a major player behind the prevention and treatment for obesity (2) and its metabolic comorbidities (3–6). Indeed, the negative consequences of physical inactivity are many (3), and health impairments can occur independent of obesity or adiposity. The United States Department of Health and Human Services estimates that ∼80% of Americans do not meet physical activity requirements and that this level of physical inactivity amounts to ∼$117 billion in healthcare costs (7). Furthermore, these costs increase with worsening levels of physical inactivity (8). Using PubMed searches for key terms (Fig. 1), one can see that physical inactivity is an often overlooked and much less frequently studied lifestyle modification in comparison to a high-fat diet (HFD), particularly in rodents (Fig. 1, A and B) (9). Unfortunately, the mechanisms of physical inactivity-induced insulin resistance (IR) have been minimally studied in rodents due to several less recognized barriers to its investigation (discussed below). As a result, PubMed searches again reveal there is an abundance of very extreme rodent models of muscle disuse (Fig. 1C), but fewer experimental models that mimic human conditions of physical inactivity and that of insulin resistance (IR) (distinctions between disuse and physical inactivity are presented in the next section). A majority of the rodent models that broadly represent situations that induce physical inactivity were purposely designed to study structural and functional changes of muscle with disuse like muscle atrophy or weakness, but not metabolic changes (i.e., insulin resistance). In fact, a 2005 review by Morey-Holton et al. (10) (summarizing the rodent hindlimb unloading literature at that time), reported that only 23 out of 1113 publications covered the topic of energy metabolism. This pattern has not changed much 15 years later. Comparatively, very few models of physical inactivity and/or muscle disuse in rodents have focused on IR as a primary outcome, either in the whole animal or specifically in the muscle. Thus, it is no surprise that PubMed searches also reveal a greater amount of research on physical inactivity-induced IR that has been conducted in clinical compared to preclinical settings (Fig. 1D). Although clinical studies are extremely valuable, they are extremely expensive to conduct and potentially invasive with muscle biopsies and clamp studies. Even so, it is rare or ethically impossible to focus beyond blood and muscle in many clinical trials. Rodent studies do provide the option of examining the physiology of other tissues (e.g. liver, visceral fat, heart, kidneys, and brain). Additionally, it is more challenging to obtain mechanistic information of causation with clinical studies. Improved preclinical representations of physical inactivity-induced IR in rodents would be less expensive and beneficial to unravel molecular mechanisms such as with genetic manipulation.

Figure 1.

PubMed searches for number of hits per year for “high-fat diet rodents” compared to “physical inactivity rodents” and “disuse OR unloading OR immobilization OR suspension and rodents” (A); “high-fat diet rats” vs. “high-fat diet mice” vs. “high-fat diet clinical trial” (B); “disuse OR unloading OR immobilization OR suspension and Rat” vs. “disuse OR unloading OR immobilization OR suspension AND mice” vs. “disuse OR unloading OR immobilization OR suspension – clinical Trial” vs. “small cage OR detraining OR wheel-lock OR physical inactivity AND rodent” (C); “physical inactivity rats” vs. “physical inactivity mice” vs. “physical inactivity clinical trial” (D). HFD = high-fat diet, PIA = physical inactivity.

EXTREME YET LESS CLINICALLY RELEVANT RODENT INACTIVITY MODELS

Most of what is known regarding physical inactivity in mice has been garnered from studies of muscle disuse using the hindlimb unloading (HU), cast or staple immobilization (HI), and denervation (DEN) models. These models are considered extreme versions of physical inactivity in which there is no load bearing and typically focused on hindlimb muscle-isolated effects. The benefit of these models, at least in the case of HI and DEN, would be the ability to utilize a unilateral design to examine localized responses with the other limb serving as an internal control. Nonetheless, these models often fail to recapitulate the full spectrum of whole-body effects of physical inactivity observed in humans. Moreover, these models typically force the rodent to use their forelimbs for locomotion, thereby impacting overall physical activity. Although we broadly classify HU, HI, and DEN in a “muscle disuse” subset of physical inactivity, there are some important distinctions to make between these three muscle disuse models. Hindlimb immobilization induces isometric contractions during the first few days (11), whereas DEN is characterized with more complete cessation of muscle contraction. Hindlimb unloading results in reductions of mechanical loading as well, however, unrestricted range of motion is one of the distinguishing factors of the HU model resulting in periodic “free-wheeling” of the hindlimbs over the time course of disuse. Moreover, hindlimb unloading results in loss of lean fat mass in both mice and rats (12–14), which is not consistent in humans where fat gain is more representative during physical inactivity (15). These (rapid hindlimb movement and fat loss) might be some possible reasons why HU sometimes results in increased muscle insulin sensitivity, whereas HI and DEN result in muscle IR (Table 1).

Table 1.

Comparison of rodent physical inactivity and disuse models on muscle atrophy and insulin sensitivity at the muscle and whole body level

Rodent Inactivity Model	Description	Muscle Atrophy	Insulin Sensitivity
			Muscle Specific	Whole Body (Insulin or glucose tolerance test)
Spaceflight	Exposure of rodents to microgravity. Rare, expensive, confounders of stress, radiation, etc.	↓	↑ SL, ↓FT (16)	?
Hindlimb unloading	Suspending a rodent by their tail for a period of time to unload the hindlimbs. Designed as a spaceflight analog	↓	↑1 day to 5 wk (16–24);↓6 h to 1 day (25, 26)	↓(27); ↑(13, 28, 29)
Hindlimb cast or staple immobilization	Local immobilization of a hindlimb, typically with a cast or a staple	↓↓	↓(30, 31)	?
Denervation	Block of muscle contraction by interrupting the nerve signal to the muscle by physical or chemical means	↓↓	↓(19, 25, 32)	?
Small cage	Restricting physical activity by reducing the living space by varying degrees	↓(33) ↕ST only; ↕ST, ↓FT (34); ↓both (35)	↓14 days (33); ↓8 days (36)	↓7 days (35); ↔14 dys (33, 35); ↓8 days (36); ↔19 wk (37)
Wheel lock	Making the standard sedentary laboratory rodent physically active with voluntary wheel running and then locking the wheel to induce physical inactivity	↔(9)	↓R(38), M?	Insulin ↑(39); 7 days small cage, ↔(40)
Detraining	Making the standard sedentary laboratory rodent physically active with forced treadmill training and then stopping the training	?	↓∼2 days (41, 42)	?
Fasting	The reduced amount of food will have secondary cause of making a mouse less active	↓	↑(43, 44)	↑(43, 44)

M, mouse; R, rat; ST, slow-twitch muscle; FT, fast-twitch muscle.

LESS EXTREME YET MORE TRANSLATABLE RODENT INACTIVITY MODELS

Other models of physical inactivity that are more representative of the whole body phenotypic effects of human physical inactivity have been used, albeit much less frequently and with reduced attention on muscle IR. These encompass rodent models of detraining where mice or rats are forced to train using treadmill or swimming exercises or voluntarily trained with running wheels for a period of time before the training stimulus is removed. With the running wheels, the rodent is placed in a new standard cage without a wheel or kept in the cage with a running wheel, but the wheel is locked (wheel lock). Other methods to modify the size of the rodent cage have been used. Increasing the size of the cage can induce greater physical activity, whereas decreasing the size of the cage may reduce physical activity. Although this small cage approach is not a new method for use with rats (33–35, 40, 45–48), its use has been re-established in a few recent studies yet now adapted to mice and in the context of muscle adaptation and IR (36, 37). Additional studies using the small mouse cage model over short- and long-term periods and improved assessments of whole body and muscle IR (hyperinsulinemic-euglycemic clamps, muscle 2-deoxy-D-glucose (2DG) uptake) are needed. A very recent study has even demonstrated that prohibiting climbing is another effective method to induce physical inactivity phenotypes (37). Combining detraining with some of the above methods (HU, small cage) has also been considered to further accelerate physical inactivity-induced effects (36, 40, 49).

DIFFERENCES BETWEEN DIET- AND INACTIVITY-INDUCED IR

High-fat diet studies in rodents have been thoroughly used to inform the mechanisms of inactivity-induced IR. Although there may be some overlap at certain points of disease progression between the lifestyle factors of diet and physical activity, there are drastic differences in the early development of IR between the scenarios. Skeletal muscle is the initial and primary site of IR following physical inactivity and disuse (50), unlike HFD where the liver or another tissue (i.e., fat) may be the main organ initially influenced (50). First, at the onset of the exposure, inactivity-induced skeletal muscle IR is more potent (i.e., greater magnitude of change) than diet-induced IR in skeletal muscle (51). Second, disuse and physical inactivity-induced muscle IR occurs before or even in the absence of hepatic IR (52, 53), whereas under HFD conditions, hepatic impairments almost always occur before muscle IR (54) (Fig. 2). Third, the development of inactivity-induced IR (as determined by insulin-stimulated glucose uptake) occurs more rapidly than IR induced by HFD. Some studies have shown that skeletal muscle IR is prominent within at least 3 h of denervation (rat) (32), 4 h of hindlimb unloading (rat) (25), 6 h of hindlimb immobilization (mice) (30), 1 day of hindlimb unloading (rat) (26) or immobilization (mice) (31), 40–42 h of detraining (rat) (41, 42), and 2 days of running wheel-lock (rat) (5) in rodents. Similarly, with human studies, inactivity-induced IR is rapidly seen as soon as 1 day of sitting (53), 3 days of bed rest (55, 56), and 1 week of reduced ambulation (57). In contrast, IR with HFD treatment becomes more evident in muscle after ∼ 3 weeks of HFD exposure (54). This general pattern of a delay in HFD-induced muscle IR (versus physical inactivity- induced IR) is apparent across many species (58). Even though excess lipids in circulation will induce muscle IR when rapidly infused to supraphysiological levels (59, 60), the physiological irrelevance of that observation does provide strong evidence of an early developmental role of muscle IR with HFD. Finally, a fourth distinction is that during the initial stages of IR development, HFD-induced IR may be mediated by circulating factors, whereas physical inactivity- and disuse-induced IR is not (61).

Figure 2.

High-fat diet results in insulin resistance in liver prior to skeletal muscle. In contrast, physical inactivity or disuse results in insulin resistance to skeletal muscle and later may develop into liver insulin resistance. Numbers denote order of tissue insulin resistance. Graphical illustration was created by Diana Lim.

MECHANISMS RELATED TO THE EARLY EVENTS OF PHYSICAL INACTIVITY- AND/OR DISUSE-INDUCED MUSCLE IR IN RODENTS

Most of what is known regarding the mechanisms underlying disuse-induced IR arises from work in the 1980-1990s using rodent models of DEN and some limited studies using HI or HU. Importantly, these mechanisms have not been examined in more translationally relevant physical inactivity rodent models. Following disuse, the rapid and nearly immediate (within hours) decrease in insulin sensitivity occurs concurrently with inactivation of glycogen synthase (62) but is unrelated to GLUT4 expression (63), glucocorticoids, adenosine, prostaglandins, circulating factors, insulin binding to its receptor (64, 65), changes in hexokinase activity (66), distal glycogenolysis (66, 67), or elevated glycogen content (61, 66). Of note, prolonged disuse leads to a decrease in GLUT4 expression, possibly worsening the severity of disuse-induced IR (33, 68–70). Nonetheless, the rapid induction of IR following disuse is likely due to a defect in intracellular signaling (71), albeit one that is downstream of the receptor tyrosine kinase (72). Since the effect of DEN is rapid, some have suggested that proteolysis or an early post-receptor defect (64, 65) may be linked to the initial IR following DEN (61). The literature has majorly focused on the early development (within hours to a few days) of disuse-induced IR and points to inactivity-induced dysregulation of insulin signaling downstream of the insulin receptor. However, it is unknown if these same potential mechanisms are observed with longer term disuse (1–2 days versus 1–2 weeks) as IR worsens weeks later when adiposity develops and other tissues develop IR (15).

Although IR induced by physical inactivity, disuse, and HFD may have the same end point, the path to that end point is divergent. Overfeeding (HFD) and increased contractile activity (i.e., electrical stimulation) independently increase long chain fatty acids in muscle, yet have an opposite effect on insulin sensitivity (73). Whereas lipid transport is decreased during muscle disuse (74), yet both disuse and HFD typically induce IR. At face value this is paradoxical during the condition of contractile activity (physical activity), since lipid influx and specifically that of saturated free fatty acids like palmitate are known triggers of muscle IR. In fact, palmitate transport is increased with increased contractile activity, but decreased during denervation, a phenomenon that correlates with the abundance of membrane-bound fatty acid transport proteins (74). Thus, HFD and lack of contractile activity results in increased IR, but likely through different mechanisms. These mechanisms warrant further investigation with more translatable models of physical inactivity.

CURRENT BARRIERS WITH RODENT MODELS OF PHYSICAL INACTIVITY

Methodological Limitations

To compare the rodent models of inactivity-induced IR to the phenotype observed in human studies of inactivity, it is important to use similar methods to assess glucose metabolism and IR (58). In humans (and rodents), the gold standard method to assess IR is the hyperinsulinemic-euglycemic clamp, which has the ability to not only assess the whole-body insulin sensitivity but also contributions from the glucose disposal and appearance rates from central and peripheral tissues (58). Tissue-specific glucose uptake in humans may also be directly assessed by the use of stable isotope and tissue biopsies, particular in skeletal muscle. Many rodent inactivity studies discussed above utilize the ex vivo glucose uptake assay as their primary outcome for skeletal muscle insulin sensitivity. Glucose tolerance tests, insulin tolerance tests, and plasma glucose and insulin assessments were most commonly used. No rodent studies have yet used the hyperinsulinemic-euglycemic clamp to report inactivity-induced IR. Although hyperinsulinemic-euglycemic clamps would be optimal to study IR as a result of physical inactivity, there are likely reasons why this method has not been used to date. One of which is that this method requires a stressful catheter implantation surgery before the clamp procedure that is often followed by a bout of forced inactivity during recovery from the surgery. Thus, any result due to the intervention is likely to be confounded due to the insult of the surgery/recovery period. Although the 2DG method is an excellent way to assess changes in insulin sensitivity that occur in rodent skeletal muscle, these data are not directly numerically comparable to data collected with the clamp method. Thus, the most relevant comparison is with the rodent ex vivo muscle glucose uptake data to the stable-isotope/muscle biopsy glucose uptake data with the clamp procedure in humans. An oral glucose tolerance test (or in the case of rodents through an i.p. mediated glucose tolerance test), although useful to assess whole-body glucose handling and more readily performed, does not provide information on organ-specific responses. Overall, our ability to compare rodent and human inactivity studies are inundated by differing methods of assessment (58).

IR in Nonmuscle Tissue Depots

It is also not possible to make a conclusive statement in rodent disuse models whether IR occurs primarily in the skeletal muscle, or whether it also occurs in other tissues. Unlike in human studies of physical inactivity, IR appears to be primarily driven by skeletal muscle (50). Several reports suggest that heart, muscle, liver, bone, and visceral fat tissues reduce in size with HU (12–14). This is likely due to reduced rates of protein and adipose tissue triglyceride turnover typifying a negative net energy balance (12), instead of the positive energy balance observed in human models of physical inactivity. One might infer that reduced tissue mass in rodent inactivity models would promote an overall reduced ability to clear circulating glucose because the reduced tissue masses would have a reduced sum of peripheral tissue glucose uptake. Part of this problem is because few studies have examined non-muscle tissue (liver, heart, and adipose) insulin resistance responses following disuse (12, 27, 75, 76). One report suggests greater gluconeogenesis in the liver with HU (76). Another study demonstrates reduced mitochondrial content and fat oxidation capacity in the liver with locking voluntary running wheels (27). More studies are needed to examine non-muscle tissue insulin resistance with rodent models of physical inactivity. Therefore, it remains to be seen whether rodent inactivity models promote IR across various tissue depots and their contributions.

Obvious but Often Neglected Differences between Rodent and Human Metabolism and Physical Activity Status

When examining localized effects of disuse between rodents and humans, an important observation is the magnitude in IR responsiveness between rodents and humans. For example, limb immobilization in humans results in muscle IR, but to a smaller magnitude compared to rodent hindlimb immobilization (77). There are several factors that may explain the divergent IR responses of humans and rodents using models of physical inactivity (mostly disuse). First, rodents (especially mice) have a higher insulin responsiveness (i.e., the capacity to respond to insulin) than humans (58). Second, liver metabolism accounts for half of the metabolic rate in a mouse, which is a much higher contribution compared to humans (78). Third, rodents are not inactive in the wild (79) as they are in the laboratory where there is no survival advantage for them to be active (4). So often, it is more difficult to induce a change in insulin resistance unless it is isolated to a limb and extreme (denervation and immobilization). In fact, some have postulated that control rodents in the laboratory are “metabolically morbid,” because they quickly become insulin-resistant, overweight, hypertensive, and are at greater risk of premature death (80). This is likely due to their imposed sedentary state and lack of adequate environmental stimuli (9, 80). Because of the high glucose metabolism and insulin responsiveness in rodents and their reduced physical activity status (in case of laboratory rodents), there is high potential to mask subtle effects of inactivity.

FUTURE DIRECTIONS

We find ourselves with the untenable task of forcing inactivity on a sedentary rodent. Data from step reduction studies in humans offers some insight as to why this has been a failed endeavor. Krogh-Madsen et al. (81) showed that change in steps was not correlated with the change in insulin sensitivity (hyperinsulinemic-euglycemic clamp) following reduced activity in young men. We show similar findings (r = 0.151, P = 0.64) in older adults (82). Together, this suggests that the magnitude of the decrease in physical activity levels after a certain point is not proportional to a continued change in insulin sensitivity. Thus, it is interesting to speculate that the inactivity “threshold” may have already been attained in the sedentary laboratory rodent. Yet, even with these limitations, general patterns with rodent research have been initially impactful to provide general advancement to the field. For example, work by Kump and Booth (38) inspired the human step reduction study in Pederson’s (81) laboratory highlighted above. Both the rodent and human studies just mentioned agreed in showing that insulin signaling was impaired upon the onset of inactivity. We highlight that there are areas where the rodent and human model do agree, in general, but if we are to advance the field further we need more translatable rodent models. Recent developments by various laboratories hold promising models to examine the mechanisms of inactivity-induced muscle IR. Methods to normalize the rodent physical activity levels back to their intrinsically active state with wheel running (5), wheel running combined with wheel-lock (55), small cage housing (33, 45), or HU (49) may open up promising opportunities to future research regarding the mechanisms of inactivity-induced muscle IR. However, these models have not yet been validated with the hyperinsulemic-euglycemic clamp, and it would appear there is a limited window (in the first 1–2 weeks) to capture the effect of physical inactivity before the rapid metabolism of the rodent compensates in a new homeostasis and effects of inactivity are no longer apparent (83). Alternatively, genetic modifications of behavior (84) to alter physical activity may also provide a novel way to manipulate rodent physical activity and to more mechanistically address the development of inactivity-induced IR. These models are promising, but still much work and validation must be conducted to find an appropriate model with translational applicability.

GRANTS

This work was supported by National Institutes of Health Grants R01AG050781, R01DK107397, and F32AR072481.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

P.T.R. and M.J.D. conceived and designed research; P.T.R. and M.J.D. performed experiments; P.T.R., C.E.P., and M.J.D. analyzed data; P.T.R., J.M.M., C.E.P., K.F., and M.J.D. interpreted results of experiments; P.T.R. and M.J.D. prepared figures; P.T.R., J.M.M., K.F., and M.J.D. drafted manuscript; P.T.R., J.M.M., C.E.P., K.F., and M.J.D. edited and revised manuscript; P.T.R., J.M.M., C.E.P., K.F., and M.J.D. approved final version of manuscript.