Research Article

The TRPM2 channel is a hypothalamic heat sensor that limits fever and can drive hypothermia

See allHide authors and affiliations

Science  23 Sep 2016:
Vol. 353, Issue 6306, pp. 1393-1398
DOI: 10.1126/science.aaf7537


Body temperature homeostasis is critical for survival and requires precise regulation by the nervous system. The hypothalamus serves as the principal thermostat that detects and regulates internal temperature. We demonstrate that the ion channel TRPM2 [of the transient receptor potential (TRP) channel family] is a temperature sensor in a subpopulation of hypothalamic neurons. TRPM2 limits the fever response and may detect increased temperatures to prevent overheating. Furthermore, chemogenetic activation and inhibition of hypothalamic TRPM2-expressing neurons in vivo decreased and increased body temperature, respectively. Such manipulation may allow analysis of the beneficial effects of altered body temperature on diverse disease states. Identification of a functional role for TRP channels in monitoring internal body temperature should promote further analysis of molecular mechanisms governing thermoregulation and foster the genetic dissection of hypothalamic circuits involved with temperature homeostasis.

Core body temperature (Tcore) is normally maintained very accurately in mammals within a narrow range around 37°C and serves as an important vital sign that is routinely monitored in hospitalized patients. Pathological conditions such as infections and systemic inflammation cause fever (1, 2), an increase in the body’s temperature that is thought to be mediated by prostaglandin E2 (PGE2)–induced inhibition of warm-sensitive neurons (WSNs) in the hypothalamic thermoregulatory center (3). Adverse drug-induced fluctuations in Tcore can have severe and even fatal consequences (4). Conversely, clinically controlled modulation of Tcore has beneficial effects and can prevent tissue damage, promote trauma recovery (5, 6), or help decrease obesity (7).

The preoptic area (POA) of the hypothalamus serves as a thermostat, integrating temperature information to orchestrate the autonomous and behavioral adaptations needed to achieve thermal homeostasis (3, 8). POA neurons receive and integrate information from peripheral temperature sensors located in the skin, spinal cord, and viscera. Local POA warming and cooling experiments (913) and genetic manipulations (14) have shown that this region of the central nervous system also detects local brain temperature changes, which have a strong impact on Tcore. In fact, temperature changes in the POA can exert a dominant effect on Tcore and can override lower-priority temperature inputs from peripheral temperature sensors, such as those located in the skin (15, 16). Warming of the POA induces heat dissipation mechanisms such as cutaneous vasodilation, evaporative heat loss mechanisms, and behavioral adaptations, whereas cooling of the POA triggers heat conservation and thermogenesis.

POA neurons exhibiting pronounced temperature sensitivity have been described in in vivo models and in vitro preparations (1719). WSNs enhance their firing rate upon warming, a process that is thought to relay temperature information to peripheral organs to promote heat loss (3). Conversely, a drop in deep brain temperature inhibits WSNs, a process that correlates with induction of thermogenesis and heat gain.

However, the genetic identity of WSNs and the molecular basis of their temperature sensitivity have remained elusive. Our work defined a population of WSNs that controls temperature homeostasis and identified a transient receptor potential (TRP) ion channel that subserves thermodetection in hypothalamic neurons.

TRPM2 is a heat sensor in a subset of POA neurons

To identify thermosensory molecules in the thermoregulatory center of the hypothalamus, we established calcium imaging as a means to monitor warming-induced responses of primary POA neuronal cultures (fig. S1, A to F). We applied a stringent criterion to identify WSNs by fura-2–mediated calcium imaging. We only counted neurons as WSNs if their relative increase in signal on receiving a thermal stimulus (up to 45°C) was larger than the mean fura-2 signal plus 5 times the standard deviation (5SD) of thermally unresponsive human embryonic kidney–293 cells subjected to the same temperature stimulus (fig. S1, E and F). This procedure allowed us to account and correct for the small unspecific temperature effect on the fluorescent fura-2 signal (20) (see fig. S1F and table S1 for details). We found that 16.3 ± 4.7% (mean ± SEM; n = 6 mice) of cultured POA neurons responded to warming, a fraction that is in the same range (10 to 16%) observed in electrophysiological studies of rodent POA cultures (21, 22) and slightly lower than WSN percentages (20 to 40%) observed in POA slice preparations (17, 23). Extracellular calcium appeared to be the source of the temperature-triggered increase in intracellular concentrations of free calcium, because chelation of external calcium abolished the response, whereas depletion of intracellular calcium stores did not have any effect on warm sensitivity (fig. S1, D and G).

Several ion channels are thermosensitive, exemplified by a steep temperature dependence (Q10 coefficient) of their opening probabilities. In particular, cationic TRP ion channels have been identified as thermosensors in the peripheral nervous system and function in the detection of ambient temperature changes (2427).

To identify candidate channels mediating the observed warm sensitivity in neuronal POA cultures, we tested various inhibitors and agonists of TRPs and other thermosensitive ion channels (fig. S1, H to L). The calcium response appeared to be mediated by direct Ca2+ flux through a thermosensitive ion channel and not an indirect readout of voltage-gated channels, because nifedipine and tetrodotoxin, which are blockers of voltage-gated calcium and sodium channels, respectively, had only a small effect on the responses to increased temperature (fig. S1, H, J, K, and L). 2-aminoethoxydiphenyl borate (2-APB), a potent inhibitor of several ion channels and activator of TRPV-type channels (2830), abolished the temperature response of POA neurons in a reversible manner (Fig. 1A).

Fig. 1 TRPM2 mediates heat responses in a subset of POA neurons.

(A) Cultured mouse POA neurons were examined for calcium responses [black trace; plotted are non-normalized fura-2 fluorescence intensity ratios after excitation at 340 and 380 nm (F340/F380)] to repetitive increases in temperature (red trace). 2-APB (30 μM, black bar) reversibly blocked thermal responses in POA neurons. (B) Cartoon depicting a coronal section of a mouse brain (Bregma level anterior-posterior, 0.14 mm). (C) Immunostaining of a representative mouse brain section corresponding to the cartoon in (B), using antibodies against TRPM2 (cyan) and neurofilament medium (NFM, magenta). Scale bar, 1 mm. (D) H2O2 (1 mM, black bar) sensitized POA neurons to respond to lower activation temperatures (40°C), as assessed by fura-2–based calcium imaging (black trace). (E) Differential interference contrast (DIC) image (top) and fura-2 ratio images before (middle; RT, room temperature) and during (bottom) a temperature stimulus of cultured POA neurons. Arrowheads indicate WSNs. (F) Current-voltage (I-V) relationships of voltage-clamped WSNs shown in (E) in response to a voltage ramp shown at the top, in the absence (blue) or presence (black) of ADPR (100 μM) or at a temperature stimulus of 45°C (red). (G and H) POA cultures from Trpm2+/+ (G) and Trpm2−/− (H) mice were examined by calcium imaging for responses to repetitive thermal stimuli (upper red traces) in the presence or absence of 2-APB (30 μM, black bar). In Trpm2+/+ cultures, both WSNs (red) and non-WSNs (light gray) could be detected, whereas WSNs were absent in cultures of Trpm2−/− mice. (I) Population analysis of thermal responses of POA neurons from Trpm2+/+ and Trpm2−/− mice as shown in (G) and (H). Plotted are the peak responses of individual neurons at the second (without 2-APB) and third (with 2-APB) temperature stimulus. POA neurons of both genotypes had similar responses to a high-potassium solution (100 mM KCl, not shown). WSNs are colored in red and non-WSNs in gray. The dashed line demarcates the cutoff for warm sensitivity as defined in fig. S1F and table S1. Throughout, error bars indicate SEM.

Among known temperature-responsive ion channels, TRPC5, TRPM2, TRPM3, TRPM8, TRPV1, TRPV2, TRPV3, and STIM1-ORAI1 are sensitive to 2-APB (fig. S1M) (2426). Because TRPC5 and TRPM8 are cold-sensitive channels (3133), we eliminated them from our list of candidates. The STIM1-ORAI1 channel complex responds to cooling subsequent to a heat stimulus (34), a characteristic that differs from the POA responses described here.

The TRPV1 to -3 channels are activated rather than inhibited by 2-APB (28, 35), and capsaicin, the cognate agonist of heat-sensitive TRPV1, did not have any effect on POA neurons (fig. S1N), suggesting that TRPV1 to -3 do not serve as temperature sensors in POA neurons. Moreover, these channels are sensitive to the pore blocker ruthenium red (36), a substance that had only a subtle effect on the temperature sensitivity of POA neurons (fig. S1, I and L).

TRPM3, one of the two remaining candidates, appeared to be expressed in the POA, albeit in low amounts, as assessed by in situ hybridization (fig. S1O). However, the TRPM3 agonist pregnenolone sulfate did not produce calcium responses in neuronal POA cultures (fig. S1N). It did so in dorsal root ganglia sensory neurons that express the receptor (37), which we used as a positive control.

Immunohistochemistry and in situ hybridization for TRPM2 showed prominent labeling within the mouse POA (Fig. 1, B and C, and fig. S3, B and C). H2O2, a sensitizer of TRPM2 channel activity (38), shifted the warm sensitivity of POA cultures to a lower activation temperature (Fig. 1D). Moreover, adenosine diphosphate ribose (ADPR), an intracellular agonist of TRPM2 (39, 40), induced large currents in POA neurons that were identified by calcium imaging to be warm-sensitive (Fig. 1, E and F, and fig. S1, P to S). The ADPR-induced current showed a TRPM2-characteristic linear current-voltage relationship (39, 40). Additionally, a similar current was induced in these neurons by increased temperatures (Fig. 1F).

Thus, TRPM2 may mediate responses to temperature increases in the POA. To test this, we used cultures obtained from Trpm2−/− mice (41) and compared their thermal responses with those from wild-type (Trpm2+/+) littermate controls. The magnitude of the response of POA neurons to increased temperature was decreased in the absence of TRPM2, and these neurons were indistinguishable from warm-insensitive (non-WSN) Trpm2+/+ POA neurons (Fig. 1, G and H). Using our stringent definition of warm sensitivity (fig. S1F), no responses to increased temperatures were detected in cultures obtained from Trpm2−/− mice (Fig. 1I).

The activation temperature of cultured WSNs—in the absence of a sensitizer such as H2O2 or any other reactive oxygen species—was in the range of 45°C (fig. S2A). Although this temperature is similar to that previously reported for TRPM2 activation (38), it exceeds any physiologically relevant deep brain temperature, even when compared with extreme forms of fever and hyperthermia (42). However, similar to other thermosensitive TRP channels, the activation temperature for TRPM2 appears not to be a fixed threshold temperature. Rather, it ranges from 33° to 47°C, depending on the cellular context (38, 43). We therefore tested whether the activation temperature of POA neurons shifted to lower, more physiological temperatures when calcium imaging was performed on acutely obtained brain slices from 7- to 8-week-old mice. In this experimental setting, the neurons largely retained their native cellular environment and had attained maturity.

To this end, we injected virus particles into the mouse POA for Cre-dependent expression of the virally encoded intracellular calcium indicator GCaMP6, which shows increased fluorescence at 500 to 530 nm upon an increase in cytosolic calcium levels (44). To test temperature sensitivity in POA neurons, we used the neuron-specific Nestin-Cre mice (45) that express Cre recombinase, and thus GCaMP6, in POA neurons. Subsequent temperature stimulation of brain slices prepared from these animals revealed calcium responses in POA neurons at 38 ± 1°C (Fig. 2, A and B). Moreover, POA neurons in slices derived from mice lacking TRPM2 (Nestin-Cre Trpm2−/−) had attenuated responses to the same temperature stimulus [mean relative heat responses ± SEM: Nestin-Cre Trpm2−/−, 0.111 ± 0.033 (n = 78); Nestin-Cre Trpm2-WT, 0.341 ± 0.040 (n = 79); P < 0.0001, two-tailed Mann-Whitney test; Trpm2-WT signifies Trpm2+/+ and Trpm2+/- mice] (Fig. 2, A and C, and fig. S2B). General excitability, as assessed by high potassium stimulation, was indistinguishable in the presence or absence of TRPM2 (fig. S2C). These results demonstrate that preoptic TRPM2 mediates responses to heat stress at temperatures of (or exceeding) 38°C.

Fig. 2 TRPM2 mediates heat responses in POA brain slices and modulates fever temperature in vivo.

(A) Acute POA slice preparations from Nestin-Cre mice with (Trpm2-WT; upper panel) or without (Trpm2−/−; lower panel) functional TRPM2 were consecutively subjected to a temperature stimulus of 38 ± 1°C (red) and a high-potassium stimulus (HiK, black). Neuronal activity was recorded by means of a Cre-induced, virally encoded GCaMP6 calcium sensor that was stereotactically introduced into the POA 2 weeks before slice preparation. A nonlinear pseudocolor scale (bottom) indicates relative calcium responses in individual neurons. Numbering on the left refers to individual neurons. The black bar indicates the time scale. (B) Average relative calcium responses of exemplary GCaMP6+ WSNs (red; n = 5) and non-WSNs (black; n = 3) from Nestin-Cre mice to repetitive temperature stimuli of up to 40°C (top, red trace). (C) Population analysis of cellular responses shown in (A), demonstrating that relative heat-induced calcium signals are reduced in POA slices from mice lacking functional TRPM2 (***P < 0.0001, two-tailed Mann-Whitney test). Blue bars show the mean and SEM. (D) Cartoon depicting the experimental paradigm used to examine the fever response by thermotelemetrical measurements of Tcore upon preoptic injection of PGE2 into the POA. (E) Enhanced PGE2-induced fever in Trpm2−/− mice (black trace; n = 9) compared with littermate Trpm2+/+ controls (red trace; n = 10). Mice were briefly anesthetized to allow stereotactic injection of PGE2 (or vehicle control), resulting in a transient drop of Tcore for all conditions and genotypes. *P < 0.05, two-way analysis of variance (ANOVA) (F1,318 = 40.78) followed by post hoc Fisher’s least significant difference (LSD) test.

Preoptic TRPM2 receptors modulate the magnitude of fever temperature

Fever is a cardinal response to infection and systemic inflammation. The POA, as part of its thermoregulatory function, is the key brain structure in controlling the increased body temperature in fever (2, 3). The TRPM2-mediated activation of POA neurons at stimulation temperatures slightly above the physiological set point of 37°C (sometimes referred to as the balance point), might allow hypothalamic TRPM2 to detect and modulate inflammatory fever temperature.

As a consequence of infection and downstream of the resulting immune response, PGE2 is the final mediator that triggers fever by acting directly on POA neurons (2, 46). To address a putative fever-modulatory role for TRPM2 directly in the POA, we microinjected PGE2 into this brain area in mice (Fig. 2D and fig. S2D), an established fever model (4648). Telemetrically measured Tcore revealed similar diurnal temperatures in the presence and absence of TRPM2 at normothermic (nonfever) conditions (fig. S2E). We found that injection of a low dose of PGE2 (0.4 nmol per mouse) produced a mild fever response that was similar in both genotypes (fig. S2, F and I). However, a high dose of PGE2 (4 nmol per mouse) caused an increased fever temperature in Trpm2−/− mice compared with that of Trpm2+/+ littermate controls (Fig. 2E and fig. S2I). In the absence of functional TRPM2 protein, the maximal fever temperature reached 40.4 ± 0.2°C, compared with 39.6 ± 0.2°C in the presence of the receptor (n = 9 Trpm2−/− and 10 Trpm2+/+ mice; P < 0.001, two-tailed unpaired t test).

We also tested the effects of the interleukins IL-1β and IL-6, which have been shown to synergistically induce fever at central sites (49). Again, significantly higher fever was detected in Trpm2−/− mice only for the stronger of the two stimuli, when both interleukins were co-injected into the POA, but not when IL-1β was injected alone (fig. S2, G to I). Together, these results indicate that preoptic TRPM2 limits the magnitude of the temperature increase in strong fever responses, presumably to prevent excessive temperature increases and tissue damage.

Trpm2-expressing POA neurons are part of a circuit controlling body temperature homeostasis

Our results indicated that Trpm2-expressing (Trpm2+) neurons might be part of a thermoregulatory circuit and that TRPM2-driven activity might counteract body heating and perhaps promote peripheral heat loss at normothermic conditions as well. In the absence of specific agents to modulate TRPM2 function in vivo, we generated Trpm2-2A-Cre knock-in mice (referred to hereafter as Trpm2-Cre mice) that express Cre recombinase exclusively in Trpm2+ cells (fig. S3A). We thereby gained genetic access to Trpm2+ POA neurons and ascertained the potential thermoregulatory role of this specific cell population in unrestrained, conscious mice. We first assessed proper Cre expression in Trpm2+ neurons, using in situ hybridization and Cre reporter mice. Cre was expressed in the POA of Trpm2-Cre mice, and its expression pattern recapitulated that of native Trpm2 transcripts (fig. S3, B to F). Two-color in situ hybridization analysis of the POA revealed essentially complete cellular overlap of Trpm2 and Cre (fig. S3F), with 97% of Trpm2+ neurons (335 out of 344 cells) expressing the Cre-induced reporter.

We then tested whether modulating the activity of Trpm2+ POA neurons influenced body temperature in freely moving mice. We chose a chemogenetic, DREADD (designer receptors exclusively activated by designer drugs) receptor–assisted approach that enabled us to remotely turn the activity of Trpm2+ neurons on and off. DREADDs are engineered heteromeric guanine nucleotide–binding protein (G protein)–coupled receptors that are activated by the otherwise inert drug clozapine-N-oxide (CNO) and couple to different G protein signaling cascades that either activate (Gq-DREADD) or inhibit (Gi-DREADD) neuronal activity (50). We injected viral constructs encoding Cre-dependent (“flexed”) DREADD receptor expression cassettes (51) into the POA of Trpm2-Cre mice (Fig. 3A). This allowed the specific activation or inhibition of Trpm2-Cre+ POA neurons through systemic application of CNO. No expression of DREADDs occurred in the absence of Cre recombinase.

Fig. 3 Control of core body temperature by Trpm2+ POA neuron activity.

(A) Virally encoded, Cre-inducible DREADD-receptor–cherry fusion proteins (left panel) were bilaterally injected into the POA of Trpm2-Cre mice and visualized by fluorescent microscopy (right panel). The red arrows indicate the orientation of the DREADD–cherry cDNA reading frame; i.p., (injected) intraperitoneally. (B) Illustration detailing the experimental paradigm for examining the impact of Trpm2+ neuronal activity on Tcore. (C to E) CNO (0.3 mg/kg; red) or vehicle (saline; black) was intraperitoneally injected into mice infected with Gq-DREADD [(C) and (E)] or control virus (D), and Tcore and activity [green trace in (E)] were measured telemetrically. Arrows indicate the time of injection. a.u., arbitrary units. (F) IR thermal imaging of Trpm2-Cre mice preoptically expressing Gq-DREADD in Trpm2+ neurons, before (0 min) or at indicated time points after CNO injection, revealed pronounced cutaneous heat dissipation through tail vasodilation (arrows). (G) Change in tail temperature over time [(arrows in (F)] in Trpm2-Cre mice infected with Gq-DREADD (red), saline control (black), or Gi-DREADD (green) upon CNO treatment. (H) CNO-mediated inhibition of Trpm2+ neurons in Gi-DREADD–infected Trpm2-Cre mice showed an oppositely directed effect compared with that in Gq-DREADD–infected animals (C) and resulted in a significant increase in Tcore. The arrow indicates the time of injection. Telemetry data [(C), (D), and (H)]: Gq-DREADD F1,6 = 75.6, ***P < 0.0001; control virus F1,6 = 1.55; Gi-DREADD F1,6 = 38.5, **P < 0.01, ***P < 0.0001. IR imaging (G): Gq-DREADD F1,5 = 9.3, *P < 0.05, ***P < 0.001; Gi-DREADD F1,4 = 5.2; two-way ANOVA followed by post hoc Fisher’s LSD test (n = 4 mice per group). P values indicate significance of the post hoc test.

To verify the capacity of CNO to exert control over Trpm2+ neurons in vitro, we co-infected POA neurons by stereotactic injection of Cre-inducible DREADD receptors and GCaMP6 viral constructs into Trpm2-Cre mice. In brain slices derived from these mice, the respective DREADD receptor and GCaMP6 are coexpressed in Trpm2+ neurons. Calcium imaging of POA neurons expressing Gq-DREADD or Gi-DREADD together with GCaMP6 confirmed that CNO application activated or inhibited Trpm2+ neurons, respectively (fig. S3G).

We implanted telemetric temperature probes in the peritoneal cavity of DREADD-virus–injected Trpm2-Cre mice to permit recording of Tcore in conscious, unrestrained animals (Fig. 3B). Diurnal body temperature regulation of DREADD-infected Trpm2-Cre mice in the absence of CNO was similar to that of control animals (fig. S3H). In contrast, CNO-mediated activation of Trpm2+ POA neurons resulted in a sustained decrease in body temperature (Fig. 3, C to E, and fig. S3I). Tcore dropped to 27.4 ± 0.6°C when Trpm2+ neurons were activated in Gq-DREADD–infected mice but not in control-infected or saline-injected mice.

The temperature decrease correlated with a reduction in activity of the animals and was also reflected by a temperature drop of the body shell, as visualized by infrared (IR) imaging (Fig. 3F and movie S1). The hypothermic state lasted several hours (0.3 mg/kg of CNO induced hypothermia for 12.5 ± 2.4 hours) and could be induced repetitively without causing any apparent long-term adverse effects (Fig. 3E).

The animal’s tail, an important thermal effector organ controlling heat dissipation (3, 52), showed an increase in surface temperature of 3.7 ± 0.5°C that coincided with the overall heat loss (Fig. 3, F and G; fig. S3J; and movie S1), demonstrating that Trpm2+ POA neurons drive hypothermia by triggering cutaneous vasodilation. Additionally, stimulation of Trpm2+ neurons prevented activation of interscapular brown adipose tissue (BAT; fig. S3K), which normally mediates thermogenesis when the body cools.

Conversely, Gi-DREADD receptor–mediated inhibition of Trpm2+ POA neurons resulted in a significant temperature increase, reaching a Tcore of up to 39.1 ± 0.1°C (Fig. 3H and fig. S3I), suggesting that under normothermic conditions, Trpm2+ neurons in the POA are tonically active. This is in agreement with previous electrophysiological recordings that show tonic ongoing activity of WSNs (1719). Gi-DREADD–mediated neuronal inhibition had no significant effect on tail vasomotor tone at ambient temperature (~20°C), likely because the prominent vasoconstriction that prevails under these conditions (53) precludes detection of any further reduction of the IR signal (Fig. 3G). These results demonstrate that Trpm2+ neurons in the POA are part of a circuit that orchestrates thermal effector processes to modulate Tcore homeostasis.

Excitatory Trpm2+ neurons drive hypothermia

WSNs in the POA have been proposed to be γ-aminobutyric acid (GABA)–releasing neurons that exert their effect largely by inhibiting downstream components of the thermoregulatory circuit (3). We therefore used two-color in situ hybridization on Trpm2-Cre reporter mice to ascertain whether Trpm2+ neurons are of inhibitory or excitatory origin. We used in situ probes for transcripts encoding the vesicular glutamate transporter 2 (VGLUT2) and glutamate decarboxylase 1 (GAD1), which are excitatory and inhibitory markers, respectively (54) (Fig. 4A).

Fig. 4 Stimulation of preoptic Vglut2+ neurons recapitulates Trpm2+ neuron–driven hypothermia and activates Crh+ neurons in the PVH.

(A) Representative two-color in situ hybridizations of coronal brain sections from Trpm2-Cre mice crossed with a Cre reporter expressing tdTomato (tdT) with probes detecting tdT (red) and Gad1 or Vglut2 (green). The right panels show close-up views indicated by the boxes in the left panels. (B to D) Vglut2-Cre mice and Vgat-Cre mice were bilaterally injected with Gq-DREADD virus and subjected to the same experimental paradigm as shown in Fig. 3B. Arrows indicate the time of injection. CNO-mediated activation of Vglut2-Cre mice [(B) and (D)], but not Vgat-Cre mice (C), recapitulated Trpm2+ neuron–driven hypothermia both in magnitude (B) and kinetics (D). Vglut2-Cre F1,6 = 86.08, ***P < 0.0001; Vgat-Cre F1,4 = 0.44; two-way ANOVA followed by post hoc Fisher’s LSD test (n = 4 Vglut2-Cre and 3 Vgat-Cre mice). (E) Representative fluorescent images of brain sections from Vglut2-Cre and Trpm2-Cre mice preoptically expressing Gq-DREADD–cherry fusion protein (Fig. 3A), showing cherry+ fibers projecting to the PVH. (F) Two-color fluorescent in situ hybridizations of brain sections from Vglut2-Cre mice preoptically expressing Gq-DREADD, using probes to detect Fos (green) and Crh or Trh (red). The animals were injected with CNO 30 minutes before preparation of the tissue sections. Neuronal activation of Crh+ neurons but not Trh+ neurons in the PVH was detected. (G) Model depicting Trpm2+ POA neurons and their bidirectional effect on Tcore. Heat (and reactive oxygen species, ROS) activate TRPM2 to depolarize WSNs, a process that can be mimicked by exogenous Gq-DREADD activation. Increased WSN activity (black traces) triggers an excitatory pathway that promotes heat loss and hypothermia, possibly through a circuit involving Crh+ neurons in the PVH. Conversely, inhibiting Trpm2+ neurons results in an elevation of Tcore. AC, anterior commissure. P values indicate significance of the post hoc test. All scale bars, 100 μm.

By this analysis, 30 ± 1% of Trpm2+ POA neurons were excitatory (Vglut2+) neurons, whereas 57 ± 7% coexpressed the inhibitory marker Gad1 (mean ± SEM; n = 3 mice) (Fig. 4A and fig. S4A). We therefore tested whether activation of the excitatory or inhibitory neuronal population within the POA recapitulated the strong thermoregulatory phenotype observed on activation of Trpm2+ neurons (Fig. 3, C to G). Vgat-Cre (54) and Vglut2-Cre (55) mice have been used previously to label inhibitory and excitatory hypothalamic neurons, respectively. Thus, Tcore recordings of Gq-DREADD–injected Vgat-Cre and Vglut2-Cre mice were acquired, and viral expression in the respective POA neurons was later verified (fig. S4, B and C). CNO activation of Vglut2+ neurons but not Vgat+ neurons reproduced the prolonged drop in Tcore (Fig. 4, B to D), indicating that excitatory, Trpm2+ neurons mediate the strong thermoregulatory effect. Vglut2+ neuron–driven hypothermia could be induced repetitively and was well tolerated by the animals (Fig. 4D), similar to observations for in vivo chemogenetic activation of preoptic Trpm2+ neurons (Fig. 3E). In situ hybridization revealed that Cre recombinase in the Vglut2-Cre line was expressed only in ∼20% of all Vglut2+ POA neurons (fig. S4D). This indicates that activation of a subpopulation of Vglut2+ POA neurons, making up only ∼6% of all preoptic Trpm2+ neurons (fig. S4A), is sufficient to mediate the strong hypothermic effect (Fig. 4, B and D).

Thermoregulatory signals arising from the POA are channeled to different peripheral effector organs through discrete hypothalamic and extrahypothalamic output nuclei (3). Given the overlapping expression pattern of Vglut2 and Trpm2 in the POA and a similar capacity of both neuron groups to trigger hypothermia on activation, we reasoned that axons of both neuron cohorts might converge on the same downstream output nucleus. Projections of Cre-expressing POA neurons were labeled with fluorescent reporters that were introduced together with the DREADD expression cassette by viral infection of the POA area (Fig. 3A). POA infections of Trpm2-Cre and Vglut2-Cre mice resulted in prominent labeling of fibers terminating in the paraventricular nucleus of the hypothalamus (PVH) (Fig. 4E), a region not previously identified as a major thermoregulatory outlet (3). Moreover, CNO administration in Trpm2-Cre and Vglut2-Cre mice harboring Gq-DREADD–infected POA neurons resulted in DREADD-specific expression of Fos in the PVH, demonstrating that Trpm2+ Vglut2+ POA neurons can excite PVH neurons (fig. S4E). Corticotropin-releasing hormone–positive (Crh+) neurons in the PVH are part of the hypothalamic-pituitary-adrenal (HPA) stress response system, which is also activated in response to fever (56). We found that DREADD-mediated activation of Vglut2+ POA neurons specifically excites parvocellular Crh+ neurons but not neighboring thyrotropin-releasing hormone–positive (Trh+) neurons or magnocellular vasopressin+ and oxytocin+ PVH neurons (Fig. 4F and fig. S4F). Of all Fos+ neurons, we found that 88.4 ± 1.5% (mean ± SEM; n = 3 mice) are Crh+.

Collectively, these results indicate that Trpm2+ Vglut2+ POA neurons detect increased body temperature and initiate thermoregulatory defense mechanisms, putatively through activation of stress-responsive Crh+ neurons in the PVH (Fig. 4G).


Explanations of the molecular mechanisms mediating thermal detection in preoptic WSNs have remained speculative (57, 58). We identified TRPM2 as a thermal sensor in a subset of POA neurons that are part of a circuit controlling Tcore.

Preoptic TRPM2 was activated at temperatures above the physiological set point of 37°C, arguing for a function of this receptor at conditions of elevated core temperatures and heat stress. Our results demonstrate a role for preoptic TRPM2 in regulating the magnitude of the fever response, suggesting that heightened temperatures, potentially together with reactive oxygen species that are produced under these conditions (59), activate the channel to promote heat loss and prevent overheating.

Fever temperatures activate the immune system to modulate inflammatory responses (60). TRPM2 is expressed in the immune system (61), suggesting that fever-activated TRPM2 might modulate immune cell function. TRPM2 activity has been shown to inhibit and counteract inflammatory (pyrogenic) signaling in phagocytes (62), paralleling our results and emphasizing a convergent TRPM2-based mechanism used by both the immune and nervous systems to limit the extent of an inflammatory response and the mediation of antipyresis.

We observed a TRPM2-mediated reduction in Tcore only when strong fever responses were induced, suggesting that TRPM2 limits the upper fever range and acts as an “emergency break” to prevent overheating and tissue damage. Alternatively, it is possible that TRPM2 plays a more prominent role in fever regulation, but that this function is masked by compensatory mechanisms triggered in Trpm2−/− mice that ubiquitously lack this ion channel throughout development.

The current model of thermoregulation predicts that the action potential firing rate of WSNs in the POA underlies Tcore homeostasis and controls peripheral autonomous thermal responses (3). We found that chemogenetic activation and inhibition of Trpm2+ WSNs in mice results in hypo- and hyperthermia, respectively, providing direct in vivo evidence for the validity of this hypothesis.

Twenty to 40% of POA neurons are sensitive to temperature and respond to stimuli around 37°C in electrophysiological experiments (17). Using calcium imaging, we found that only around 16% of neurons responded to a temperature increase. It is possible that our approach only allowed the detection of neurons that robustly responded to a thermal stimulus. Buffering of intracellular free calcium by the calcium indicators used in our experiments is one possible explanation for the reduced sensitivity and might explain the lower percentage of responsive neurons. Our data indicate that the TRPM2 receptor is unlikely to account for temperature sensitivity at normothermia but rather detects heat stress above 37°C. Accordingly, in POA brain slices, some thermal responses were still detectable in the absence of TRPM2, implying that additional mechanisms mediate preoptic responses to warming. Given the relatively broad expression of TRPM2 in the POA, it is plausible that the receptor covers detection of the upper, pathological temperature range in many or even all WSNs.

Previous single-cell transcriptome analysis has shown that preoptic WSNs are unexpectedly heterogeneous (63). Nevertheless, WSNs have been found to largely be inhibitory (GABAergic) neurons, which is in agreement with our histochemical analysis (summarized in fig. S4A). However, because the hypothermic phenotype that we observed in Trpm2-Cre mice was recapitulated by activating the smaller subpopulation of excitatory Vglut2+ neurons but not by activating inhibitory (Vgat+) neurons, we focused on the excitatory branch of the Trpm2+ population. We found that preoptic Vglut2+ neurons project to and excite Crh+ neurons in the PVH, a hypothalamic cell population that is not primarily associated with thermoregulation (3), but rather controls stress reactions including fever responses by activation of the HPA axis. Excitation of PVH neurons inhibits PGE2-induced fever (47). Moreover, corticosterone release by HPA axis activation has been found to counteract and reduce fever responses (64, 65). These findings are in agreement with our data. The TRPM2 receptor may curtail PGE2-triggered fever and induce hypothermia by mediating monosynaptic excitation of Crh+ neurons in the PVH through an excitatory thermoregulatory pathway that connects the POA with the HPA axis.

Our study delineates a genetic framework for dissecting the central pathways controlling temperature homeostasis and provides a way to remotely control core body temperature in conscious, unrestrained mice by chemogenetic manipulation of the hypothalamic thermostat. These hypo- and hyperthermic model systems may help in exploring the effects of altered core body temperature on diverse processes such as trauma recovery, immune modulation, energy expenditure, obesity, and longevity.

Supplementary Materials

Materials and Methods

Figs. S1 to S4

Table S1

References (6676)

Movie S1

  • * Present Address: Max Delbrück Center for Molecular Medicine, Robert Rössle Straße 10, 13125 Berlin, Germany.

References and Notes

  1. Acknowledgments: We thank Y. Mori for Trpm2−/− mice and TRPM2 antisera; B. Lowell for Vgat-Cre mice; members of R. Kuner’s laboratory, the Nikon Imaging Center, and V. Grinevich for sharing reagents, equipment, and expertise; A. von Seggern and C. Steinmeyer-Stannek for technical support; K. Schrenk-Siemens, W. de Lima, and P. Bengtson for help with calcium imaging; and C. Bohlen and A. Littlewood-Evans for criticism and reading of the manuscript. All data necessary to understand the study and its conclusions are in the body of the manuscript and in the supplementary materials. All primary data are archived on secure servers at the University Hospital Heidelberg. All data will be made available upon request. This work was supported by the European Research Council (grant ERC-2011-StG-280565 to J.S.), the Alexander von Humboldt Foundation (German Federal Ministry for Education and Research, BMBF; fellowship to G.B.K. and grant to J.S.), and the German Research Foundation (grants as part of the Collaborative Research Center SFB-1158 to P.H. and J.S.). Additional support came from the CellNetworks Cluster of Excellence. J.S., together with K.S., H.Wa., and G.B.K., conceived the project and designed the experiments. K.S. designed and conducted the cellular screen and analysis; J.P. designed and performed electrophysiological experiments; G.B.K. conceived the slice physiology and fever experiments; and H.Wa. conceived and conducted the chemogenetic experiments and, together with H.We., P.H., and F.d.C.R., generated Trpm2-Cre mice. H.We. and H.Wa. conducted histochemical analysis. J.S. wrote the paper. All authors commented on and approved the paper.
View Abstract

Stay Connected to Science

Navigate This Article