Hypothermia in mouse is caused by adenosine A₁ and A₃ receptor agonists and AMP via three distinct mechanisms

Abstract

Small mammals have the ability to enter torpor, a hypothermic, hypometabolic state, allowing impressive energy conservation. Administration of adenosine or adenosine 5'-monophosphate (AMP) can trigger a hypothermic, torpor-like state. We investigated the mechanisms for hypothermia using telemetric monitoring of body temperature in wild type and receptor knock out (Adora1^−/−, Adora3^−/−) mice. Confirming prior data, stimulation of the A₃ adenosine receptor (AR) induced hypothermia via peripheral mast cell degranulation, histamine release, and activation of central histamine H₁ receptors. In contrast, A₁AR agonists and AMP both acted centrally to cause hypothermia. Commonly used, selective A₁AR agonists, including N⁶-cyclopentyladenosine (CPA), N⁶-cyclohexyladenosine (CHA), and MRS5474, caused hypothermia via both A₁AR and A₃AR when given intraperitoneally. Intracerebroventricular dosing, low peripheral doses of Cl-ENBA [(±)-5'-chloro-5'-deoxy-N⁶-endo-norbornyladenosine], or using Adora3^−/− mice allowed selective stimulation of A₁AR. AMP-stimulated hypothermia can occur independently of A₁AR, A₃AR, and mast cells. A₁AR and A₃AR agonists and AMP cause regulated hypothermia that was characterized by a drop in total energy expenditure, physical inactivity, and preference for cooler environmental temperatures, indicating a reduced body temperature set point. Neither A₁AR nor A₃AR were required for fasting-induced torpor. A₁AR and A₃AR agonists and AMP trigger regulated hypothermia via three distinct mechanisms.

Graphical Abstract

1. Introduction

Mammals are endotherms, typically maintaining a warm core body temperature (Tb) of ~35 °C to 37 °C, depending on circadian fluctuations (Refinetti, 2010). Some small mammals, including mice, when exposed to an inadequate food supply in a quiet, cool environment use torpor to achieve significant energy conservation (Geiser, 2004; Melvin and Andrews, 2009). Torpor is an example of a regulated hypothermia (anapyrexia), which is characterized by a reduced Tb set point, metabolic rate, and physical activity, with Tb falling close to the environmental temperature. In regulated hypothermia multiple physiologic mechanisms are coordinated to cool the body, including vasodilation, decreased physical activity, reduced brown adipose tissue thermogenesis, and seeking a cool environment (Lute et al, 2014). Regulated hypothermia differs from hypothermia caused by cold exposure, where the body attempts, but is unable, to maintain Tb.

Hypothermia can also be caused by a wide variety of drugs and neurotransmitters (Clark and Lipton, 1985), although in most cases it is not documented if there is a reduction in Tb set point. Clinically, hypothermia is routinely employed to minimize tissue damage after hypoxic or ischemic injury (Arrich et al, 2012; Azzopardi et al, 2014). The hypothermia is typically induced with surface cooling. However, pharmacological reduction of the Tb might minimize undesired compensatory mechanisms (e.g., sympathetic activation, shivering). Development of a drug regimen that produces controlled, regulated hypothermia and avoids activating counter-regulatory responses would likely be of clinical utility (Drew et al, 2015; Tupone et al, 2014).

The hypothermic effect of adenosine was first reported in 1931 (Bennet and Drury, 1931). Once multiple adenosine receptor (AR) subtypes were identified, the hypothermia was attributed to action at A₁AR, likely within the brain (Anderson et al, 1994). Peripheral dosing of N⁶-cyclohexyladenosine (CHA; see Fig. S1 for structures) elicited hypothermia, which was lost in A₁AR knock out (Adora1^−/−) mice (Johansson et al, 2001). CHA also caused hypothermia when infused directly into the nucleus of the solitary tract in rats (Tupone et al, 2013). However, some data suggested a contribution from non-A₁AR mechanisms, as only partial attenuation of hypothermia caused by N⁶-R-phenylisopropyladenosine (R-PIA) was seen in Adora1^−/− mice (Yang et al, 2007).

More recently, it has become clear that adenosine agonists also induce hypothermia in mice via A₃AR. A critical observation was that hypothermia caused by R-PIA was attenuated in A₃AR knock out (Adora3^−/−) mice (Yang et al, 2010). In rodents, A₃AR is expressed on immune cells (Borea et al, 2015) and A₃AR agonists trigger mast cell degranulation (Auchampach et al, 1997; Fozard et al, 1996; Salvatore et al, 2000). We have recently shown that A₃AR agonists activate peripheral mast cells releasing histamine, which then acts on central histamine H₁ receptors to lower the Tb set point (Carlin et al, 2016).

Adenosine 5′-monophosphate (AMP) is a proposed natural regulator of torpor and injection of a large dose of AMP (500–3500 mg/kg, i.p.) causes hypothermia (Zhang et al, 2006). Given the dose size, one might consider if some of the effects of AMP are due to its conversion to free adenosine (Rittiner et al, 2012; Swoap et al, 2007). The observation that AMP’s hypothermic effects remain intact in mice lacking any one of the four AR subtypes, indicates that AMP is not acting non-redundantly via a single AR (Daniels et al, 2010). However, AMP-induced hypothermia was reported to be blocked by infusion of an A₁AR antagonist in the pre-optic area (Muzzi et al, 2013). The mechanism by which AMP causes hypothermia is currently unclear.

While studying hypothermia caused by A₃AR agonists, we observed that some nucleoside derivatives commonly used as A₁AR agonists also had activity at the A₃AR. This prompted the current re-examination and comparison of A₁AR agonists, A₃AR agonists, and AMP. Our data suggest that each of these can trigger hypothermia in mice via different mechanisms.

2. Materials and Methods

2.1 Mice

Male C57BL/6J and Kit^W-sh/W-sh mice (Stock #012861) (Grimbaldeston et al, 2005; Nigrovic et al, 2008) were obtained from the Jackson Laboratory. Adora1^−/− mice on a C57BL/6J background were provided by Dr. Jurgen Schnermann (Sun et al, 2001) and genotyped by PCR (Adora1 reverse common primer 5'-ACATGGGGGTTGAACAGAGA, Adora1 forward primer 5'-AGCTGGCTACCGCTACACAT, and Neo forward primer 5'-TCTGGATTCATCGACTGTGG), producing 302 bp wild type and ~900 bp null allele products. Adora3^−/− mice on a C57BL/6 background made by Merck (Salvatore et al, 2000) were provided by Dr. Stephen Tilley and genotyped by PCR (Adora3 reverse common primer 5'-ACTGGCCCATACACAACCTG, Adora3 forward primer 5'-AGACAATGAAATAGACGGTGGTG, and Neo forward primer 5'-ATGGAAGGATTGGAGCTACG), producing 208 bp wild type and ~400 bp null allele products. Mice were singly housed at ~22 °C with a 12:12-h light-dark cycle. Chow (NIH-07, Envigo Inc, Madison, WI) and water were available ad libitum. Mice were studied ≥ 7 days after any operation or prior treatment. Reuse of mice tends to reduce physical activity levels, presumably due to acclimatization. No specific effort was made to acclimatize mice to handling in individual experiments. Studies were approved by the Animal Care and Use Committee of National Institute of Diabetes and Digestive and Kidney Diseases.

2.2 Drugs

The following compounds (vehicle) were purchased from Sigma (St. Louis, MO) or Tocris (Minneapolis, MN): Cl-ENBA, (±)-5'-chloro-5'-deoxy-N⁶-endo-norbornyladenosine (Franchetti et al, 2009; Trivedi et al, 1989) (10% DMSO in saline); CHA, N⁶-cyclohexyladenosine (dissolved in DMSO, then diluted to 10% DMSO with saline); CPA, N⁶-cyclopentyladenosine (saline); pyrilamine (saline); AMP, adenosine 5'-monophosphate (saline); CCPA, 2-chloro-N⁶-cyclopentyladenosine. MRS5474, (1R,2R,3S,5S)-4-(2-chloro-6-((dicyclopropylmethyl)amino)-9H-purin-9-yl)bicyclo[3.1.0]hexane-2,3-diol, (dissolved in DMSO, then diluted with 9 volumes 30% PEG400) was synthesized as described (Tosh et al, 2012b).

2.3 Adenosine Receptor Binding Affinities

Binding affinity for mouse A₁AR, A_2AAR, and A₃ARs was measured as described (Kreckler et al, 2006) using membranes from human embryonic kidney (HEK)-293 cells stably expressing individual recombinant mouse adenosine receptors and using the agonists [¹²⁵I]N⁶-(4-amino-3-iodobenzyl)adenosine-5'-methyluronamide ([¹²⁵I]AB-MECA; A₁AR and A₃AR) and [³H]CGS21680 (A_2AAR) as radioligands. Nonspecific binding was defined using 100 µM adenosine-5'-N-ethylcarboxamide (NECA). K_i values were obtained using the Cheng-Prusoff equation from IC₅₀ values calculated by non-linear regression analysis of specific binding data using GraphPad Prism software (San Diego, CA).

2.4 Central infusions

Mice were anesthetized with ketamine/xylazine (80/10 mg/kg, i.p.). Sterile guide cannulas (5.25 mm, 26 gauge, Plastics One, Roanoke, VA) were unilaterally implanted into the lateral ventricle (coordinates relative to bregma: −0.34 mm anterior, 1.0 mm lateral, and +1.7 mm ventral) and fixed with dental cement (Parkell, Edgewood, NY). Compounds in 5 µl were infused (0.5 µl/min) through a 33 gauge cannula protruding 0.5 mm past the tip of the guide cannula using PE-50 tubing fitted to a 5 µl syringe (Hamilton, Reno, NV). Cannula positions were verified by postmortem histological analysis.

2.5 Body Temperature and Activity Telemetry and Indirect Calorimetry

Tb and activity were measured continuously by telemetry (Starr Life Sciences, Oakmont, PA) using ER4000 energizer/ receivers, G2 E-mitters implanted intraperitoneally, and VitalView software with data collected each minute. Invalid E-mitter data points (defined as a change of ≥ 1 °C in one minute) were replaced by interpolation using flanking data. Any Tb ≤ 24 °C were scored as 24 °C. Unless noted otherwise, the average Tb response was calculated using the first 60 min after agonist dosing. Hypothermia duration is the duration of the interval between dosing and 300 min during which core temperature is <35 °C. Activity is the sum of counts from 10–60 min after A₁AR agonist administration. Inhibitors were dosed 15 min before agonists. Indirect calorimetry was performed as described (Carlin et al, 2016). Experiments were performed at ~22 °C. Occasionally a mouse did not become hypothermic with a treatment that generally caused hypothermia. In these cases, the same mouse was retested ≥ 7 days later and routinely became hypothermic and the second data set was used.

2.6 Temperature Gradient

Environmental temperature preference was measured using an in-house apparatus by continuous video monitoring of the position of the mouse in a 45 cm thermal gradient, nominally 15 °C to 35 °C (Carlin et al, 2016). Data are reported as mean ± SEM. Significance (two-tailed p < 0.05) was determined by t test or ANOVA followed by post hoc Holm-Sidak multiple comparison tests.

3. Results

3.1 CHA induces hypothermia via both A₁AR and A₃AR

We investigated the ability of systemically-dosed adenosine agonists to cause hypothermia. CHA is ~290-fold selective for binding to mouse A₁AR over A₃AR (Table 1). With vehicle treatment, dosing-associated handling caused an increase in Tb and activity that lasted about an hour. Treatment of C57BL/6J mice with CHA (0.02–0.2 mg/kg, i.p.) caused a drop in Tb, accompanied by reduced physical activity (Fig. 1A–E). The Tb and physical activity decrease and the increase in duration of hypothermia were all dose-dependent.

Table 1.

Ligand binding affinity at adenosine receptors.

K_i, nM (or % inhibition at 10 µM)

	A1AR	Mouse A2AAR	A3AR	A1AR	Human A2AAR	A3AR
Agonists
IB-MECAa,b	5.9	~1000	0.087	51	2900	1.8
Cl-IB-MECAa,b	35	~10,000	0.18	220	5400	1.4
MRS5474c,d	3.20±0.05	34±9%	1056±251	50	3950	470
MRS5698e,e	14%	27%	3.1	6%	41%	3.5
MRS5841f,f	15%	1%	11	16%	7%	1.9
MRS5980g,g	38%	7%	36	6%	24%	0.70
CCPAh,i	0.27	988	16	0.83	2270	43
CPAc,i	0.22±0.01	808±89	534±14	2.3	794	43
CHAc,j	2.15±0.37	1695±60	611±12	2.4	1390	73
SPAc,c	1.08±0.04	33±2%	3111±183	7.92±1.55	3778±890	246±21
Cl-ENBAc,k	0.20±0.01	3985±358	2414±325	0.51	1340	1290
CGS21680c,i	193±38	10±2	48±11	289	27	67
Antagonists
MRS1523n,l	5,330	0%	702	>10,000	3660	19
MRS1191c,l	0%	0%	32±3%	>10,000	>10,000	31
SCH442416c,m	765±36	1.27±0.04	6±2%	35%	4.1	67%
DPCPXn,i	1.5	598	0%	3.9	129	3960

Bold indicates a K_i ≥100-fold lower than the other two receptors of the same species. Superscripts indicate the references for literature data (mouse, human). SEMs are included for experimental results.

^a(Ge et al, 2006)

^b(Tchilibon et al, 2005)

^ccurrent results

^d(Tosh et al, 2012b)

^e(Tosh et al, 2012a)

^f(Paoletta et al, 2013)

^g(Tosh et al, 2014)

^h(Alnouri et al, 2015)

ⁱ(Klotz et al, 1998)

^j(Ferkany et al, 1986; Gao et al, 2003; Stone et al, 1988)

^k(Franchetti et al, 2009)

^l(Liang et al, 2010)

^m(Kumar et al, 2011)

ⁿ(Kreckler et al, 2006)

Fig. 1. Systemic CHA causes hypothermia and decreased physical activity through both A₁AR and A₃AR.

(A, B) Tb and physical activity response to the indicated CHA doses injected i.p. into C57BL/6J mice. (C–E) Effect of CHA on average Tb (0–60 min), duration of hypothermia, and physical activity (10–60 min). Data are mean ± SEM, n=5/group; every tenth SEM is shown in A and SEMs were omitted in B for visual clarity. (F,G) Tb response to CHA (0.05 mg/kg, i.p.) or vehicle in C57BL/6J (WT) and Adora1^−/− (KO) mice. (H,I) Tb response to CHA (0.05 mg/kg, i.p.) or vehicle in C57BL/6J (WT) and Adora3^−/− (KO) mice. (J,K) Tb response to CHA (0.05 mg/kg, i.p.) or vehicle in C57BL/6J (WT) and Adora1^−/−;Adora3^−/− (DKO) mice. In F–K, data are mean ± SEM, n=5–10/group in a crossover design; every tenth SEM is shown in F, H, and J; * p<0.05, *** p<0.001.

The in vivo specificity of CHA was examined using Adora1^−/− and Adora3^−/− mice. Hypothermia elicited by CHA (0.05 mg/kg, i.p.) was attenuated in both Adora1^−/− and Adora3^−/− mice, but was completely lost in Adora1^−/−;Adora3^−/− double knock out mice (Fig. 1F–K). These data demonstrate that the hypothermic effect of this low dose of CHA is contributed by agonism at both A₁AR and A₃AR.

3.2 MRS5474 induces hypothermia via A₃AR

Truncated nucleoside MRS5474, which contains a bicyclic substitute for ribose, is a moderately selective, full A₁AR agonist that is well tolerated in vivo (Tosh et al, 2012b) and is ~280-fold selective for binding to mouse A₁AR over A₃AR (Table 1). It caused dose-dependent hypothermia and hypoactivity (Fig. S2A,B and data not shown). Treatment of Adora1^−/− and Adora3^−/− mice with MRS5474 (3 mg/kg, i.p.) demonstrated that the hypothermia was via A₃AR, without a clear contribution from A₁AR (Fig. S2C–F). Since centrally-active dopamine D₂-like receptor agonists cause hypothermia (Nunes et al, 1991), we investigated the effect of raclopride, a D₂-like receptor antagonist. Hypothermia induced by MRS5474 (3 mg/kg, i.p.) was not inhibited by pretreatment with raclopride (1 mg/kg, i.p.; data not shown).

3.3 CPA induces hypothermia via A₃AR more than A₁AR

CPA is another widely used adenosine agonist with better selectivity (~2400-fold) for binding to mouse A₁AR over A₃AR (Table 1). CPA caused a dose-dependent (0.03–3 mg/kg, i.p.) hypothermia and reduction in physical activity in C57BL/6J mice (Fig. S3A,B and data not shown). However, hypothermia induced by CPA (0.3 mg/kg, i.p.) was not significantly attenuated in Adora1^−/− mice and was attenuated but still present in Adora3^−/− mice (Fig. S3C–F). These data suggest that the hypothermic effects of this dose of CPA are via agonism at A₃AR, likely with a lesser contribution from A₁AR.

A related agonist, CCPA, is often used as a more selective A₁AR agonist. However, we found that its binding affinity at the A₃AR would make it even less desirable than CPA for use in mouse (Table 1). In general, there is a need to reexamine the affinities of widely used AR ligand probes across species (Alnouri et al, 2015). Therefore, we compare the affinities of various agonists and antagonists at human and mouse receptors. Some previous studies of adenosine and hypothermia used DPCPX as a selective antagonist of the A₁AR. However, the selectivity ratios of DPCPX vary significantly between species (Table 1 and (Alnouri et al, 2015)).

3.4 In vivo Cl-ENBA is a more selective A₁AR agonist

Cl-ENBA is the most selective A₁AR agonist described, with ~2600-fold selectivity for binding to human A₁AR vs. A₃AR (Franchetti et al, 2009; Trivedi et al, 1989) and ~10,000-fold selectivity for mouse A₁AR over A₃AR (Table 1). Cl-ENBA (0.3–10 mg/kg, i.p.) caused dose-dependent hypothermia and hypoactivity (Fig. 2A,B and data not shown). Cl-ENBA (3 mg/kg, i.p.) hypothermia was not affected by pretreatment with raclopride (1 mg/kg, i.p.; data not shown). It was largely abolished in Adora1^−/− mice (Fig. 2C,D), remained present in Adora3^−/− mice (Fig. 2E,F), and was completely lost in Adora1^−/−;Adora3^−/− mice (Fig. 2G,H). These data (and see 3.5) demonstrate that the hypothermic effects of Cl-ENBA are mainly mediated via A₁AR, but at higher doses there is probably also a contribution from A₃AR.

Fig. 2. Systemic Cl-ENBA acts largely via A₁AR to induce hypothermia.

(A,B) Tb response to the indicated Cl-ENBA dose injected i.p. into C57BL/6J mice. (C,D) Tb response to Cl-ENBA (3 mg/kg, i.p.) or vehicle in C57BL/6J (WT) and Adora1^−/− (KO) mice. (E,F) Tb response to Cl-ENBA (3 mg/kg, i.p.) or vehicle in C57BL/6J (WT) and Adora3^−/− (KO) mice. (G,H) Tb response to Cl-ENBA (3 mg/kg, i.p.) or vehicle in C57BL/6J (WT) and Adora1^−/−;Adora3^−/− (KO) mice. Data are mean ± SEM, n=3–6/group in a crossover design; every tenth SEM is shown in C, E, and G; * p<0.05*** p<0.001.

3.5 A₁AR agonist-induced hypothermia occurs via central sites, while A₃AR agonist-mediated hypothermia requires mast cells

The poor in vivo A₁AR selectivity of systemic CHA and CPA prompted us to re-examine the role of mast cells and histamine in A₁AR and A₃AR agonist-induced hypothermia. Hypothermia from A₃AR agonists is caused by histamine release from activated mast cells (Carlin et al, 2016), while a role for mast cells in A₁AR agonist-induced hypothermia has not been reported. Hypothermia from systemic Cl-ENBA at a dose of 1 mg/kg was largely intact in mast cell-deficient Kit^W-sh/W-sh mice (Fig. 3A,B), while the hypothermia with 3 mg/kg was significantly attenuated (Fig. 3C,D). In contrast, the hypothermic effect of MRS5474, at a dose that is selective for A₃AR, was abolished in Kit^W-sh/W-sh mice (Fig. 3E,F). Pyrilamine, an histamine H₁R antagonist, inhibited hypothermia from this dose, but did not block A₁AR agonist-induced hypothermia (Fig. S4A–D). These results show that, in contrast to A₃AR-mediated hypothermia, A₁AR hypothermia does not require mast cells or H₁R, and suggest that Cl-ENBA at 1 mg/kg i.p., but not 3 mg/kg, is relatively selective for A₁AR.

Fig. 3. A₁AR-mediated hypothermia is partially intact and A₃AR-mediated hypothermia is abolished in Kit^W-sh/W-sh mice.

(A,B) Tb response to Cl-ENBA (1 mg/kg, i.p.) in Kit^W-sh/W-sh mice. (C,D) Tb response to Cl-ENBA (3 mg/kg, i.p.) in Kit^W-sh/W-sh mice. (E,F) Tb response to MRS5474 (3 mg/kg, i.p.) in Kit^W-sh/W-sh mice. Data are mean ± SEM, n=4–6/group in a crossover design; every tenth SEM is shown in A, C, and E; ** p<0.01.

We next examined the site of A₁AR agonist action to induce hypothermia. Cl-ENBA infused intracerebroventricularly (i.c.v.) at 3.4 µg/mouse (~0.12 mg/kg) caused a drop in Tb comparable to 1 mg/kg given systemically (Fig. 4A,B). The effect of i.c.v. Cl-ENBA was greatly diminished in Adora1^−/− mice (Fig. 4C,D). These results indicate that Cl-ENBA is acting via central A₁AR.

Fig. 4. Cl-ENBA causes hypothermia via central A₁AR.

(A,B) Tb response to Cl-ENBA (3.4 µg/mouse) or vehicle injected i.c.v. into C57BL/6J. (C,D) Tb response to Cl-ENBA (15 µg/mouse) or vehicle injected i.c.v. into C57BL/6J or Adora1^−/− mice. Data are mean ± SEM, n=3–6/group; every tenth SEM is shown in A, C; * p<0.05.

To avoid confounding due to the modest in vivo receptor specificity of the A₁AR agonists, we used Adora3^−/− mice to investigate mechanistic aspects of A₁AR-driven hypothermia. Cl-ENBA administration reduced metabolic rate and physical activity before the Tb nadir (Fig. 5A–C). Additionally, Adora3^−/− mice treated with Cl-ENBA had a modest preference for a cooler place in a thermal gradient (Fig. 5D–G). These results suggest that there is a reduced Tb set point and coordinated physiologic response to an A₁AR agonist.

Fig. 5. Systemic Cl-ENBA-induced hypothermia is accompanied by a reduced metabolic rate and preference for a cooler environment.

The effect of Cl-ENBA (3 mg/kg, i.p.) or vehicle in Adora3^−/− mice on (A) total energy expenditure (TEE), (B) Tb, and (C) physical activity. The TEE falls before Tb (nadir ~50 min vs ~100 min for Tb). Data are mean ± SEM, n=5/group. Position (D,E) and activity (F,G) of Adora3^−/− mice treated with Cl-ENBA (3 mg/kg, i.p.) or vehicle and placed in a thermal gradient. Mean ± SEM, n=6/group, *p<0.05, ***p<0.001.

3.6 Neither A₁AR nor A₃AR is necessary for fasting-induced torpor

To better understand the role of adenosine in thermal physiology, baseline Tb and physical activity in light and dark phase were measured in Adora1^−/−, Adora3^−/−, and Adora1^−/−;Adora3^−/− mice. There were no clear differences between littermate control and Adora1^−/− or Adora3^−/− mice or Adora1^−/−;Adora3^−/− and age-matched C57BL/6J mice (Fig. 6A–F). A 24-hour fast elicited torpor in both Adora1^−/− and Adora3^−/− mice (Fig. 6G–H). In a test for receptor redundancy between A₁AR and A₃AR, Adora1^−/−;Adora3^−/− double knockout mice were found to enter fasting-induced hypothermia (Fig. 6I). These data demonstrate that neither A₁AR nor A₃AR, individually or in combination, is required for fasting-induced torpor.

Fig. 6. A₁AR or A₃AR is not required for fasting-induced hypothermia.

Baseline Tb (A) and activity (B) in WT (n=6) and Adora1^−/− (n=4) littermates during the light or dark phase. Baseline Tb (C) and activity (D) in WT (n=5) and Adora3^−/− (n=4) littermates during the light or dark phase. Baseline Tb (E) and activity (F) in C57BL/6J (n=6) and Adora1^−/−;Adora3^−/− (n=3) mice. during the light or dark phase. Individual traces of Tb in (G) WT and Adora1^−/− littermates, (H) WT and Adora3^−/− littermates, and (I) C57BL/6J and Adora1^−/−;Adora3^−/− mice during a 24 hour fast. White and black bars indicate light and dark phases, respectively. Data are mean ± SEM.

3.7 AMP does not require A₁AR or A₃AR to induce hypothermia and can act centrally

Large AMP doses (500–3500 mg/kg, i.p.) induce torpor (Zhang et al, 2006). We studied a modest dose (100 mg/kg, i.p.) to minimize possible effects of impurities or breakdown products. The AMP-induced hypothermia was intact in Adora1^−/− and Adora3^−/− mice (Fig. 7A–D). It was also intact in Adora1^−/−;Adora3^−/− double knockout and partially intact in Kit^W-sh/W-sh mice (Fig. 7E–H). Pretreatment of Adora1^−/− mice with the brain-penetrant H₁R antagonist pyrilamine (10 mg/kg, i.p.) did not ablate the hypothermia (data not shown). Similarly, pretreatment with raclopride (2 mg/kg, i.p.) did not block the hypothermia (data not shown). These data demonstrate that the hypothermic effects of AMP do not require mast cells, A₁AR, A₃AR, H₁R, D₂R, or the combination of A₁AR/A₃AR.

Fig. 7. Systemic AMP causes hypothermia independent of A₁AR, A₃AR, and mast cells.

(A,B) Tb response to AMP (100 mg/kg, i.p.) in C57BL/6J (WT) and Adora1^−/− (KO) mice (n=6/group). (C,D) Tb response to AMP (100 mg/kg, i.p.) in C57BL/6J (WT) or Adora3^−/− mice (n=4/group). (E,F) Tb response to AMP (100 mg/kg, i.p.) in C57BL/6J and Adora1^−/−;Adora3^−/− mice (n=5–6/group). (G,H) Tb response to AMP (100 mg/kg, i.p.) in C57BL/6J (WT) or Kit^W-sh/Wsh mice (n=7–8/group). Data are mean ± SEM; every tenth SEM is shown in A, C, E, and G; * p<0.05, ** p<0.01, *** p<0.001.

To investigate the site of action where AMP acts to cause hypothermia, AMP was infused i.c.v. into C57BL/6J mice. A low i.c.v. dose (100 µg/mouse, ~3.3 mg/kg) induced a drop in Tb that was comparable to a 30-fold larger (100 mg/kg) systemic dose and the hypothermia was not attenuated in Adora1^−/− mice (Fig. 8A,B).

Fig. 8. Centrally-administered AMP induces hypothermia.

(A,B) Tb response to 100 µg (~3.3 mg/kg) AMP injected i.c.v. into C57BL/6J mice or Adora1^−/− (n=4/group). Data are mean ± SEM; every tenth SEM is shown; * p<0.05, ** p<0.01.

Mice treated with AMP (100 mg/kg, i.p.) demonstrated a drop in metabolic rate that preceded the hypothermia (Fig. 9A-C). We did not observe a clear change in ambient temperature preference at this dose (not shown). With a higher dose (300 mg/kg, i.p.), the mice generally moved to a cooler region of a thermal gradient and became immobile (Fig. 9D–G). When AMP was administered centrally (100 µg/mouse, ~3.3 mg/kg, i.c.v.), there was a similar movement toward cooler environmental temperature but without a significant drop in activity (Fig. 9H–K). These findings are consistent with AMP acting in the brain to reduce the Tb set point.

Fig. 9. AMP-induced hypothermia is accompanied by a reduced metabolic rate and preference for a cooler environment.

The effect of AMP (100 mg/kg, i.p.) or vehicle in C57BL/6J mice on (A) total energy expenditure (TEE), (B) Tb, and (C) physical activity in a calorimetry chamber. The TEE falls before Tb (nadir ~40 min vs ~60 min). Data are mean ± SEM, n=5/group. Position (D,E) and activity (F,G) of C57BL/6J mice treated with AMP (300 mg/kg, i.p.) or vehicle and placed in a thermal gradient. Position (H,I) and activity (J,K) of C57BL/6J mice treated with AMP (100 µg, i.c.v.) or vehicle and placed in a thermal gradient. Mean ± SEM, n=5–6/group, *p<0.05, ***p<0.001

4. Discussion

We demonstrate at least three distinct adenosine-related routes to trigger hypothermia in mice (Fig. S5): 1) activation of central A₁AR, 2) peripheral mast cell activation by A₃AR agonists, releasing histamine, which stimulates histamine H₁ receptors, and 3) a central AMP-mediated mechanism. Additionally, we report that some agonists commonly used as A₁AR probes also activate A₃AR in vivo.

4.1 Suboptimal in vivo A₁AR agonist selectivity

Much understanding of adenosine physiology has been acquired with selective ligands (Fredholm, 2014) (Table 1). CHA and CPA were developed to be selective for A₁AR vs “A₂AR” (Moos et al, 1985). The subsequent identification of four ARs and the generation of receptor-ablated animals now permit improved dissection of the specific receptors mediating adenosine actions. To our initial surprise, CHA, MRS5474, and CPA all produced hypothermia via A₃AR, despite their much higher affinity for A₁AR over A₃AR. This is presumably due to higher drug exposure at peripheral mast cell A₃AR, with poor availability at brain A₁AR. This is striking, with a 2000-fold advantage in binding affinity being nullified by pharmacokinetic considerations. We did not measure peripheral A₁AR effects of these agonists (eg., on heart rate, blood pressure, or kidney function (Sun et al, 2001; Vallon et al, 2006)) for comparison with the peripheral mast cell A₃AR effect.

This observation suggests a reexamination of prior in vivo studies using CHA and CPA as A₁AR probes. In mice, CHA is typically used at i.p. doses of 0.2 to 1 mg/kg, with ED₅₀s in locomotor assays of 0.3 to 1.7 mg/kg (Akula and Kulkarni, 2014; Heffner et al, 1989; Zarrindast et al, 1993). Doses of 0.1 to 1 mg/kg are reported in other species (Jinka et al, 2010; Jinka et al, 2015; Olson et al, 2013). We demonstrate that both A₁AR and A₃AR contribute to hypothermia at a CHA dose of 0.05 mg/kg. Similarly, CPA is typically used in mice at doses of 0.2 to 1 mg/kg i.p., with ED₅₀s in locomotor assays of 1.0 to 1.7 mg/kg (El Yacoubi et al, 2000; Heffner et al, 1989; Listos et al, 2011; Von Lubitz et al, 1993). Comparable doses (0.1 to 1 mg/kg) are used in rats (Heurteaux et al, 1995; O'Neill et al, 2014; Ramos-Zepeda and Herrero, 2013). Our results with 0.3 mg/kg show that both A₁AR and A₃AR contribute to CHA-induced hypothermia. Thus, commonly used i.p. doses of both CPA and CHA stimulate both A₁AR and A₃AR and some effects attributed to A₁AR may actually be due completely, or in part, to A₃AR. This is of particular concern when the pharmacodynamic effect requires brain-penetration for an A₁AR ligand but only peripheral exposure for A₃AR. A non-brain penetrant agonist, SPA, was found to be highly selective for the mouse A₁AR (~3000-fold vs A₃AR) and could be useful in future in vivo studies.

There have been clinical trials of A₁AR agonists for conditions such as diabetic foot pain, diabetes, glaucoma, and cardiac arrhythmias, many of which have failed (Elzein and Zablocki, 2008). Other envisioned therapeutic applications of A₁AR agonists are based on reported neuroprotective, antiseizure, anti-nociceptive, sleep promoting, and antidepressant effects (Tosh et al, 2012b). The fact that some in vivo actions of agonists, that appear selective in in vitro receptor assays, also might include A₃AR effects, should be considered in future development of such agonists. A₃AR agonists display some similar effects, such as neuroprotection and antinociception (Janes et al, 2016), which might contribute to efficacy of nominal A₁AR agonists and would be difficult to separate mechanistically in clinical trials. The A₃AR is expressed at relatively low levels in the mouse hippocampus, cerebellum and medulla (Janes et al, 2016; Yaar et al, 2005). Thus, although we found no effects of centrally administered A₃AR agonists on Tb, there may be effects of this receptor on CNS function.

Cl-ENBA is a more selective A₁AR agonist than CPA, CHA, and MRS5474 in the mouse for binding (Franchetti et al, 2009; Trivedi et al, 1989) (Table 1) and in vivo hypothermia. At 1 mg/kg i.p., Cl-ENBA was largely selective for A₁AR, although at 3 mg/kg some of the effect was contributed by A₃AR. Cl-ENBA is anti-nociceptive in the mouse at 0.5 mg/kg (Luongo et al, 2012), suggesting that this effect is via A₁AR. Cl-ENBA’s greater selectivity makes it a preferred agonist for investigating in vivo A₁AR physiology, at least in mice.

The Adora1^−/− and Adora3^−/− mice were essential for assessing in vivo ligand selectivity. In the absence of sufficiently selective A₁AR agonists, using Adora3^−/− mice eliminates confounding A₃AR effects. Adenosine agonist affinities and selectivities can vary profoundly among species (Table 1) (Alnouri et al, 2015). Adenosine physiology can also vary by species--for example, A₃AR is not present in some human mast cells (Rudich et al, 2012). In summary, the results particularly suggest caution when studying in vivo effects of A₁AR when using i.p. dosing.

4.2 A₃AR Hypothermia

A₃AR agonists activate peripheral mast cells, releasing histamine, which then acts on central histamine H₁ receptors to lower the Tb set point (Carlin et al, 2016). Here we examined the hypothermic effects of MRS5474 (Tosh et al, 2012b) and found that at 3 mg/kg i.p., MRS5474 is acting via A₃AR. The hypothermia is lost in Kit^W-sh/W-sh mice, confirming that A₃AR agonism is acting via mast cells. Since histamine is reported to both increase and decrease Tb (Tabarean, 2016), further study is required.

The kinetics of hypothermia induction via A₃AR and A₁AR are similar, despite the peripheral vs central sites of agonist action. The likely explanation is that heat loss (and thus hypothermia) occurs slowly compared to the rapid increase in circulating histamine and quick neural transmission.

4.3 A₁AR and regulated hypothermia

The poor in vivo A₁AR selectivity of peripherally-dosed agonists raises concern that some physiology attributed to A₁AR might actually be due to A₃AR. For hypothermia, the lack of a Tb effect by brain A₃AR agonist indicates that studies using central agonist dosing are probably studying A₁AR effects. Thus, the hypothermia elicited by local CHA delivery to the anterior hypothalamus of Syrian hamsters (Shintani et al, 2005) or to the nucleus of the solitary tract of rats (Tupone et al, 2013) remain plausibly attributed to A₁AR. A recent rat study used brain-penetrant CHA to produce A₁AR hypothermia combined with a non-brain-penetrant antagonist (8-(p-sulfophenyl)theophylline) to block peripheral A₁AR-mediated bradycardia (Jinka et al, 2015). We interpret this experiment as additionally having hypothermia from peripheral CHA activation of rat mast cell A₃AR (Fozard et al, 1996), which may not be blocked by 8-(p-sulfophenyl) theophylline (Gao et al, 2001).

4.4 AMP Hypothermia

Circadian fluctuation of plasma AMP concentration is a proposed endogenous regulator of torpor in rodents (Zhang et al, 2006). Large exogenous doses of AMP (500 to 3500 mg/kg) cause torpor and have protective effects in various injury models (Miao et al, 2015; Miao et al, 2012; Tao et al, 2011; Wang et al, 2014). There is no clearly identified receptor for AMP (although AMP has been proposed to be an A₁AR ligand (Rittiner et al, 2012)). In our studies, AMP-induced hypothermia did not require mast cells, A₁AR, A₃AR, H₁R, D₂R-like, or a combination of A₁AR/A₃AR. Other nucleotides, such as ATP, ADP and cyclic-AMP, have hypothermic effects when injected into the hypothalamic region in cats (Dascombe and Milton, 1975), although the mechanisms have not been defined.

Proposed mechanisms for AMP-induced hypometabolism are that AMP uptake into erythrocytes causes increased 2,3-bisphosphoglycerate levels and reduced oxygen transport (Daniels et al, 2010) and/or that the AMP activates an intracellular energy depletion sensor, AMP-kinase (Melvin et al, 2009). The observation that metabolic rate reduction precedes the drop in Tb (Daniels et al, 2010) and our results that relatively small amounts of i.c.v. AMP elicit hypothermia and AMP-treated mice chose a cool environment, all indicate that AMP acts centrally to reduce the Tb set point, suggesting direct or indirect action at the hypothalamus. Muzzi et al. have shown that AMP injected into the mouse preoptic area induces hypothermia and that AMP reduces firing activity of preoptic area neurons in a DPCPX-sensitive manner (Muzzi et al, 2013). These observations and the Adora1−/− data can be reconciled by hypothesizing that AMP causes hypothermia via both A₁AR-dependent and A₁AR-independent mechanisms.

4.5 Fasting-induced hypothermia does not require A₁AR or A₃AR

Mice have a large surface area/volume ratio and therefore a disproportionally high energetic cost of staying warm, so they enter torpor to conserve energy (Abreu-Vieira et al, 2015; Hudson and Scott, 1979; Swoap, 2008). It is likely that the control of torpor is complex, integrating information on nutritional status, energy needs, and environmental situation. For example, leptin deficiency sensitizes to induction of torpor, but replacement of leptin in adipose-deficient mice does not prevent fasting-induced torpor (Gavrilova et al, 1999; Himms-Hagen, 1985).

There is much support for A₁AR having a central role in regulating torpor, including daily torpor in mice (Iliff and Swoap, 2012), seasonal torpor in arctic ground squirrel (Jinka et al, 2011), and hibernation in Syrian hamsters (Tamura et al, 2005). These conclusions are based on pharmacologic manipulation, showing torpor induction with agonist and exit with antagonist treatment. However, the observation that mice lacking A₁AR (or A₃AR, or both) enter torpor upon fasting demonstrates that A₁AR is not required. However, due to the intrinsic variability of fasting-induced hypothermia and the size of our dataset, we cannot rule out quantitative or probabilistic effects of A₁AR and/or A₃AR on this process. This illustrates the utility of the knock out mice, highlights the limitations of purely pharmacologic approaches, and suggests searching for the other regulators of torpor-related states. Presumably, like leptin, A₁AR signaling is one of multiple inputs that are integrated to cause entry into (or exit from) torpor.

4.6 Limitations

Hypothermia can vary with inter-experiment differences, particularly in duration (with larger standard errors at longer times after dosing). Possible contributing factors include mouse age, weight, and adiposity, acclimation to handling, season, environmental temperature, and the noise/activity level in the animal facility. To assure that this variability does not compromise the conclusions, we focus on the all-or-none effects and have not attempted to quantitate the fraction of a mixed effect that is due to A₁AR versus A₃AR.

While knockout mice are excellent probes of in vivo ligand specificity, compensation for receptor loss may occur. No compensatory changes in Adora1−/− or Adora3−/− mice have been described (Johansson et al, 2001; Salvatore et al, 2000), but it is possible that undetected compensatory changes do occur. Importantly, there may also be redundant sites of action, for example AMP may act at both A₁AR and other site(s).

4.7 Common themes in A₁AR, A₃AR, and AMP Hypothermia

A₁AR and A₃AR agonists and AMP all trigger a regulated hypothermia that is not inhibited by D₂R antagonist. Yet the three act in different places in the body, with different ligands and ligand affinities. Extracellular adenosine (and AMP) can be thought of as indicators of physiologically unfavorable situations or ‘extreme physiology’ (Fredholm, 2014). It would be advantageous for the body to be able to broadly sense many forms of pathology and funnel the signals to common response mechanisms such as hypothermia. The ability to go into torpor is a widely conserved behavior that reduces injury, conserves energy, and increases survival. Understanding the neural regulation of hypothermia will identify signaling pathways and may lead to improved pharmacologic approaches to hypothermia induction and maintenance in clinical practice.

Highlights.

1. Some commonly used A₁AR agonists are not selective; they act at A₃AR

2. A₁AR, A₃AR, and AMP are three distinct adenosine-related inducers of hypothermia

3. Neither A₁AR nor A₃AR is required for fasting-induced torpor

4. AMP can cause hypothermia via the brain, independent of A₁AR and A₃AR

Abbreviations

i.p.

intraperitoneal

i.c.v.

intracerebroventricular

AxAR

adenosine x receptor