Control of hypothalamic orexin neuronsby acid and CO2Rhı̂annan H. Williams*, Lise T. Jensen†, Alex Verkhratsky*‡, Lars Fugger†§, and Denis Burdakov¶�

*Faculty of Life Sciences, University of Manchester, Manchester M13 9PT, England; †Department of Clinical Immunology, Aarhus University Hospital,DK-8200 Aarhus C, Denmark; ‡Institute of Experimental Medicine, Academy of Sciences of the Czech Republic, Videnska 1083, 142 20 Prague, CzechRepublic; §Medical Research Council Human Immunology Unit and Department of Clinical Neurology, Weatherall Institute of Molecular Medicine,University of Oxford, Oxford OX3 9DS, England; and ¶Department of Pharmacology, University of Cambridge, Tennis Court Road, CambridgeCB2 1PD, England

Edited by Richard D. Palmiter, University of Washington School of Medicine, Seattle, WA, and approved May 8, 2007 (received for review March 22, 2007)

Hypothalamic orexin/hypocretin neurons recently emerged as keyorchestrators of brain states and adaptive behaviors. They arecritical for normal stimulation of wakefulness and breathing:Orexin loss causes narcolepsy and compromises vital ventilatoryadaptations. However, it is unclear how orexin neurons generateappropriate adjustments in their activity during changes in phys-iological circumstances. Extracellular levels of acid and CO2 arefundamental physicochemical signals controlling wakefulness andbreathing, but their effects on the firing of orexin neurons areunknown. Here we show that the spontaneous firing rate ofidentified orexin neurons is profoundly affected by physiologicalfluctuations in ambient levels of H� and CO2. These responsesresemble those of known chemosensory neurons both qualita-tively (acidification is excitatory, alkalinization is inhibitory) andquantitatively (�100% change in firing rate per 0.1 unit change inpHe). Evoked firing of orexin cells is similarly modified by physio-logically relevant changes in pHe: Acidification increases intrinsicexcitability, whereas alkalinization depresses it. The effects of pHe

involve acid-induced closure of leak-like K� channels in the orexincell membrane. These results suggest a new mechanism of howorexin/hypocretin networks generate homeostatically appropriatefiring patterns.

arousal � hypocretin � hypothalamus � pH � breathing

Even small changes in extracellular levels of protons ([H�]e)are fatal to mammalian cells and tissues. The brain rapidly

counteracts such changes with finely tuned adaptations inbreathing, behavioral arousal, and aversive panic responses(1–4). Breathing controls [H�]e through CO2 removal (in thebody, H� � HCO3

� 7 CO2 � H2O), while behavioral arousalensures that any obstructions to breathing can be effectivelyremoved. For example, respiratory acidosis produced by sleepapnea causes awakening and stimulates breathing, which nor-malizes [H�]e. Failure of such reflexes is thought to contributeto fatal conditions such as sudden infant death syndrome (5, 6).

To elicit appropriate changes in breathing and arousal, neu-rons need to translate changes in [H�]e into appropriatelycoordinated responses. In many neurons, [H�]e depresses elec-trical excitability (i.e., acidosis is inhibitory and alkalosis isexcitatory) (7). However, the opposite type of cellular responseis thought to be important for the vital adaptive reflexesmentioned earlier: a steep increase in firing rate by physiologicalacidification and a steep decrease in firing upon alkalinization(4). Such specialized chemosensory firing responses are tradi-tionally associated with neurons of the brainstem, a key centerin the control of arousal and breathing (1, 3, 4). The ionic basisof these responses is controversial and has been proposedto involve ion currents mediated by GABAA Cl� channels,Ca2� channels, inwardly rectifying K� channels, voltage-gatedK� channels, and leak-like K� channels (4). However, all thesechannels are also expressed in many nonchemosensory neurons(8), and thus their presence alone does not signify whether aneuron is chemosensory. To draw this conclusion, it is necessary

to perform direct measurements of firing rate and membranepotential at different levels of [H�]e.

Although until now most work on neuronal chemosensing hasfocused on the brainstem, historical evidence indicates that thehypothalamus is also vital for maintaining normal breathing andarousal (9, 10). Hypothalamic neurons responsible for thesefunctions have only recently been identified and found to containpeptide transmitters called orexins/hypocretins (11, 12). Orexin-containing neurons (orexin neurons) are located in the lateralhypothalamus (LH), but project widely throughout the brain,where orexins act on two specific G protein-coupled receptors(13). Both arousal and breathing centers receive stimulatoryinputs from orexin neurons (14–16). Orexin neurons attractedgreat attention when they proved to be targets in narcolepsy,where their destruction causes irresistible daytime sleepiness,disturbed night-time sleep, and cataplexy (17–19). The activity oforexin neurons is also essential for adaptive exploratory re-sponses to food shortage (20). Furthermore, recent data showthat orexin knockout severely compromises the ability to in-crease breathing during acidosis (16).

These studies established that orexin neurons are essential forappropriate adjustments of brain and behavioral states to theinternal and external environments (13). However, the regula-tion of orexin neurons is much less understood. In particular, itis unknown whether the firing of orexin neurons is sensitive tophysiologically relevant changes in fundamental homeostaticsignals such as [H�] and [CO2]. To document such responses andcompare them to known chemosensory neurons, we performedelectrophysiological recordings from orexin neurons identifiedby targeted expression of GFP in transgenic mice. We show thatorexin cell firing is potently stimulated by CO2 and H� througha mechanism involving acid-induced inhibition of postsynapticleak-like K� channels. We find that this enables orexin cell firingto encode small physiological changes in pH with high sensitivity,comparable to that of classical chemosensory neurons of thebrainstem. These results identify a new mechanism by whichhypothalamic orchestrators of adaptive behavior generate ho-meostatically appropriate patterns of activity.

ResultsFiring Responses of LH Orexin Cells to Physiological Fluctuations inpHe. Living orexin neurons are difficult to identify because theyexhibit few defining anatomical features and are intermixed withmany other types of cells in the LH. To overcome this problem,

Author contributions: D.B. designed research; R.H.W. and D.B. performed research; L.T.J.,A.V., and L.F. contributed new reagents/analytic tools; R.H.W. and D.B. analyzed data; andD.B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Abbreviation: LH, lateral hypothalamus.

�To whom correspondence should be addressed. E-mail: dib22@cam.ac.uk.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0702676104 PNAS � June 19, 2007 � vol. 104 � no. 25 � 10685–10690

NEU

ROSC

IEN

CE

we used brain slices from transgenic mice that selectively expressGFP in orexin neurons, allowing precise identification of livingorexin cells for electrophysiological recordings (see Methods).

In the intact brain in vivo, the interstitial pH is generally �7.1to 7.25, but can fluctuate between 6.9 and 7.4, for example, inresponse to hypo- and hyperventilation, respectively (21). Wefirst mimicked the natural shifts in extracellular pH (pHe) bychanging bath [CO2] (see Methods), which in the body controls[H�]e via the reaction H� � HCO3

�7 CO2 � H2O. Switchingfrom 10% to 5% CO2 (which creates pHe of 7.02 and 7.25,respectively) reversibly suppressed the spontaneous firing oforexin cells by �4-fold, hyperpolarized their membrane poten-tial, and increased their membrane conductance (Fig. 1A; thestatistics are given in the figure legends). To confirm that this wasdue to changes in [H�] rather than [CO2] or [HCO3

�], we alsovaried pHe without changing the latter two parameters, usingbicarbonate-free solutions buffered with the artificial H� bufferHepes (see Methods). Switching pHe from 6.9 to 7.4 withHepes-buffered solutions suppressed the spontaneous firing oforexin cells �10-fold, induced membrane hyperpolarization, andincreased membrane conductance (Fig. 1B). Hepes-buffered

solutions were used in subsequent experiments for better pHcontrol (22).

Firing Responses of LH Nonorexin Cells to Physiological Fluctuationsin pHe. To examine the specificity of the [H�]e effects, we alsoperformed recordings from other LH neurons. We chose cellsthat were near orexin neurons, but did not display GFP fluo-rescence and did not possess the defining electrical fingerprintsof orexin neurons [electrical fingerprinting involving hyperpo-larizing current injections was routinely performed as in ourprevious studies (23, 24)]. In contrast to orexin neurons, thefiring and membrane resistance of these nonorexin cells waseither unaffected by a pHe switch from 6.9 to 7.4 (see Fig. 1C)or accelerated by this alkalinizing switch (n � 3, data not shown).These data imply that the changes in orexin cell firing onphysiological changes in pHe (Fig. 1 A and B) are not a generalcharacteristic of LH neurons, but represent a more specificproperty of orexin cells.

Steepness of the Relationship Between pHe and Orexin Cell Firing. Theslope of the relationship between spontaneous firing rate andextracellular pH is considered of critical importance for classi-fying neurons as chemosensory (4). To investigate this parameterin orexin neurons, we exposed them to different levels of pHeand measured steady-state firing rates at each of these levels(Fig. 2). pHe shifts that were much smaller than the extremechanges used in Fig. 1 were also effective in changing orexin cellfiring: Even switching pHe from 7.15 to 7 increased the firing rate�4-fold in some cells (e.g., Fig. 2A). The same cells were ableto integrate both small and more extreme physiological changesin extracellular acidity (Fig. 2 A). Within the physiological rangeof pHe (6.9–7.4), the firing rate was steeply dependent on pHe,

Fig. 1. Effects of H� and CO2 on orexin and nonorexin cells of the LH.Current-clamp whole-cell recordings with K-gluconate intracellular solu-tion. To monitor membrane conductance, cells were injected with periodichyperpolarizing current pulses. The size of the resulting downward voltagedeflections is inversely proportional to membrane conductance. (A) Effectof CO2-induced changes in pHe on an orexin neuron. Firing rate was 0.8 �0.3 Hz in 5% CO2 and 3.7 � 0.4 Hz in 10% CO2 (n � 5; P � 0.01, two-tailedt test). (B) Effect of Hepes-buffered changes in pHe on an orexin neuron.Firing rate was 4 � 0.6 Hz at pH 6.9 and 0.5 � 0.2 Hz at pH 7.4 (n � 15; P �0.001, two-tailed t test). (C) No effect of changes in pHe on firing of anonorexin neuron.

Fig. 2. Sensitivity of orexin cell firing to extracellular [H�] (pHe). Current-clamp whole-cell recordings with K-gluconate intracellular solution. (A) Effectof small and large pHe changes on spontaneous firing in the same cell. (B)Mean steady-state firing rate plotted against pHe. Each point represents morethan three cells.

10686 � www.pnas.org�cgi�doi�10.1073�pnas.0702676104 Williams et al.

decreasing from �4 Hz at pHe of 6.9 to near zero at pHe of 7.4(Fig. 2B). The slope of the steep part of the firing-pHe curve wasclose to 10 Hz per unit change in pH (Fig. 2B), which iscomparable to values observed in chemosensitive cells of thebrainstem (4). We also calculated the percent change in firingrate per 0.1 unit change in pHe (see Methods), which is thoughtto be a better comparative measure of chemosensitivity than theabsolute changes in firing rate (4). In orexin neurons, theincrease in firing rate per 0.1 pH unit acidification was 108% (seeDiscussion).

Effects of Physiological pHe Changes on Evoked Firing and IntrinsicExcitability. In the behaving animal, the firing of orexin neuronsis likely to be modulated by an array of neurotransmitters,hormones, and nutrients that act by triggering membrane cur-rents (25). It is therefore of interest to define how the sensitivityof orexin cell firing to current inputs (intrinsic excitability) isaffected by pHe. We measured excitability as the number ofaction potentials induced by a fixed-amplitude current pulse(2-sec duration) from the same baseline membrane potential(�70 mV). The baseline potential was imposed by offsettingpHe-induced Vm changes by sustained current injection (26).Current amplitude was chosen that evoked 10 to 15 actionpotentials at pHe of 7.1 (typically this was 30 pA). Afterextracellular alkalinization from pH 7.1 to 7.4, there was asignificant decrease in the number of evoked action potentials(Fig. 3). This decrease was reversible on return to pHe of 7.1 (Fig.3). These data suggest that, like the spontaneous firing rate, theintrinsic excitability of orexin neurons is increased by acidifica-tion and depressed by alkalinization.

Effects of Physiological pHe Changes on Membrane Currents. Toaddress the mechanism underlying H�-sensing by orexin neu-rons, we combined intracellular ion substitutions with measure-ments of the equilibrium potential of the H�-modulated con-ductance. In Figs. 1–3, we used a low [Cl�] intracellular solution(K-gluconate; see Methods), which meant that, at typical mem-brane potentials of orexin cells, only K� and Cl� conductanceswere hyperpolarizing, and all other ion channels were depolar-izing (27). Therefore, the hyperpolarizing conductance activatedby rises in pHe (i.e., a fall in [H�]e) (Fig. 1) must have beenmediated by K� or Cl� channels. We next switched to a high[Cl�] intracellular solution (K-chloride), which sets ECl to 0 mV,

thereby making K� channels the only ion channels that canhyperpolarize resting potentials of orexin cells (27). Neither thismanipulation (Fig. 4A) nor blocking action potential-dependentsynaptic communication with tetrodotoxin (Fig. 4B) affected themembrane potential and membrane conductance responses toH�, implying that postsynaptic K�, and not Cl�, channels areresponsible. In direct confirmation of this conclusion, the H�

e-modulated current had a mean reversal potential of �93 � 4 mV(n � 14; mean � SEM), close to that of K� (theoretical EK ��97 mV). This current was activated by physiological alkalin-ization and inhibited by acidification (Fig. 5A).

Several ion channels implicated in chemosensing have recentlybeen found in orexin neurons (27, 29). Among these are ‘‘leak’’K� channels (K2P channels) (27). We next directly examined theeffects of pHe on membrane currents using whole-cell voltage-clamp recordings (Fig. 5) to test whether acid-inhibited leak K�

channels are involved in firing and membrane potential re-sponses of orexin cells to extracellular H�. We found strongsupport for this idea: The net H�-inhibited current in orexinneurons exhibited (i) current-voltage rectification well describedby the Goldman–Hodgkin–Katz current equation (Fig. 5 A andB), and (ii) leak-like rapid activation and minimal time-dependent decay in response to voltage steps (Fig. 5 C and D).Because these two properties, together with inhibition by H�, arenot features of any known ion channels apart from acid-sensitiveK2P channels (28), our findings are strong evidence that theymediate the K� currents involved in H� sensing in orexinneurons.

DiscussionExtracellular pH is a fundamental signal regulating breathingand behavioral arousal. Although the activity of hypothalamicorexin neurons is essential for normal adaptive increases inarousal and respiration (20, 16), it was unknown whether orexincell firing is sensitive to physiological changes in pHe. Here weshow that the spontaneous firing rate of identified orexinneurons is profoundly affected by physiological f luctuations inacid and CO2 levels (Figs. 1 and 2). Crucially, these firingresponses resemble those displayed by classical chemosensoryneurons both qualitatively (acidification is excitatory, alkalin-ization is inhibitory) and quantitatively (�100% change in firingrate per 0.1 unit change in pHe). Evoked firing of orexin cells is

Fig. 3. Effects of extracellular pH on evoked firing of orexin cells. Current-clamp whole-cell recordings with K-gluconate intracellular solution. The cellwas set to the same baseline of �70 mV (see Methods). The time scale isexpanded in the lower part to show responses to 30-pA current pulses (shownschematically below the traces) (n � 4; 12 � 1 spikes at pHe 7.1 vs. 3.3 � 0.6spikes at pHe 7.4; P � 0.005, two-tailed t test).

Fig. 4. Effects of intracellular ion substitutions and action potential block-ade on H� responses of orexin cells. Current-clamp whole-cell recordings withK-chloride intracellular solution. (A) Effects of pHe changes on firing andmembrane potential. (B) Effects of pHe changes in the presence of tetrodo-toxin (0.7 �M). The amplitude of downward deflections is inversely propor-tional to membrane conductance (see Fig. 1).

Williams et al. PNAS � June 19, 2007 � vol. 104 � no. 25 � 10687

NEU

ROSC

IEN

CE

also modulated by physiologically relevant pHe: Acidificationincreases intrinsic excitability, whereas alkalinization depressesit (Fig. 3). The effects of pHe are caused by changes in postsyn-aptic conductance, but Cl� channels are not essential for theseresponses (Fig. 4). Instead, our data imply that acid-inhibitedleak-like K� channels in the orexin cell membrane are involved(Fig. 5). These results provide an insight into how behaviorallydefined hypothalamic networks generate homeostatically appro-priate patterns of activity.

Orexin Neurons as Potential Polymodal Chemosensors. After therecent realization of the behavioral importance of orexin neu-rons (13), much work became focused on stimuli controllingtheir activity. This work revealed that the firing of orexin cells isregulated by diverse hormones and neurotransmitters (20, 30).Furthermore, orexin neurons were found to act as specializedelectrical sensors of glucose, translating rises in ambient levels ofthe sugar into reductions in their firing rate (20, 31). Our newdata show that orexin cell firing is also specialized to encodephysiological levels of acid and CO2. This ability to senseendocrine, neural, metabolic, and physicochemical signals isreminiscent of the classical polymodal chemosensory organs,such as the carotid body (32, 33). Combined with their ability toorchestrate diverse brain states (see below), this places orexincells in a pivotal position to guard body homeostasis throughappropriate behavioral corrections. The relative role of eachtype of input to orexin cells during different circumstancesand stages of development is an important subject for futureinvestigation.

In the field of chemosensing, our study supports an emergingconcept that different chemosensory systems operate at differ-

ent levels in the brain to link pH with breathing and arousal. Inparticular, there appear to be chemosensors tightly linked to thecontrol of breathing, e.g., neurons of the retrotrapezoid nucleusof the brainstem (34) and, in parallel, systems that link pH toboth breathing and arousal, such as serotonin and noradrenalineneurons of the brainstem (3, 4) and orexin neurons (this study).We estimated the chemosensitivity of orexin cell firing to be�100% rate increase per 0.1 pH unit. In the brainstem, thisparameter varies from �100% for regions with the highestdegree of chemosensitivity, such as the medullary raphe, to�15% for regions with the lowest chemosensitivity, such asthe locus coeruleus (4). In terms of sensitivity of firing to pHe,orexin cells thus appear to be among the most sensitive knownchemosensors.

Implications for Brain States and Adaptive Behaviors. We proposethat the acid/CO2-induced changes in orexin cell firing, reportedhere, represent a previously overlooked physiological pathwaycontributing to fundamental reflexes such as increased wake-fulness and breathing during sleep apnea. This sensing pathwaycould provide a possible explanation for the recent observationthat orexin knockout reduces the breathing responses to CO2 by�50% (16). Orexin cells project to respiratory centers in thebrainstem and spinal cord neurons where injections of orexinstimulate breathing (15), although the underlying mechanismsremain to be established in detail. With regard to behavioralarousal and alertness, it is well established that the activity oforexin neurons stimulates wakefulness by providing excitatoryinputs to most classical arousal centers as well as directly to thecortex (13, 35, 36). In addition, orexins may also generate anxietyand stresslike states (37, 38, 39) probably due to the reciprocalexcitatory interactions with neurons containing corticotropin-releasing factor (40, 41). This pathway is of particular interest inthe context of this study because high CO2 levels are well knownto induce anxiety responses, which are thought to be mediatedin part by CO2/H�-sensitive neurons of the locus coeruleus (4).However, as noted above, the chemosensitivity of locus coer-uleus neurons appears to be �6-fold lower than that of orexinneurons. The orexin system may thus ensure a more gradedtransition to panic states during rising CO2 levels, possibly incooperation with the serotonin system of the brainstem (3).Finally, it is noteworthy that orexin neurons have many addi-tional central projections that exert critical influences on appe-tite and reward/addiction (11, 42, 43). Our findings may thusrepresent a new cellular link between acid-base balance andthese behavioral drives.

MethodsPreparation of Brain Slices from Transgenic Mice. Mice were madetransgenic for enhanced GFP (eGFP) under the control ofprepro-orexin promoter as described in our previous study (27).This procedure results in very specific targeting of eGFP only toorexin neurons (27). Procedures involving animals were carriedout in accordance with the Animals Scientific Procedures Act of1986 (U.K.). Mice were maintained on a 12-h light:dark cycle(lights on at 0800) and had free access to food and water.Animals (14–20 days postnatal) were killed during their subjec-tive night (the light phase), and coronal slices (250–300 �mthick) containing the lateral hypothalamus were prepared aspreviously described (27).

Extracellular and Intracellular Solutions. For experiments involvingchanges in [CO2] (Fig. 1 A), we used bicarbonate-buffered bathsolution, which contained 125 mM NaCl, 2.5 mM KCl, 2 mMMgCl2, 1.2 mM NaH2PO4, and 26 mM NaHCO3. Bubbling thissolution with 5% CO2 (95% N2) or 10% CO2 (90% N2) gave pHevalues of 7.25 and 7.02, respectively (at 25°C). In the rest of thepaper, we changed pHe using Hepes-buffered bath solutions

Fig. 5. Ionic basis of H� responses of orexin cells. Voltage-clamp whole-cellrecordings with K-chloride intracellular solution and bath tetrodotoxin (0.7�M). (A) Effect of pHe changes on membrane current-voltage relationship ofthe cell shown in Fig. 4B. The current-voltage relationship was obtained usingslow voltage ramps (see Methods). (B) Current-voltage relationship of the netcurrent (n � 4; data are means � SEM) inhibited by acid (i.e., current in pH 7.4minus current in pH 6.9). The line through the points is a fit of the Goldman–Hodgkin–Katz equation to the data (see Methods). (C) Effects of pHe onmembrane current obtained with voltage-clamp steps (shown schematicallybelow the traces). (D) Time course of activation and inactivation of the netacid-inhibited current (pH 7.4–6.9) obtained by a step to �60 mV (from �100mV). The trace is representative of results obtained in five cells.

10688 � www.pnas.org�cgi�doi�10.1073�pnas.0702676104 Williams et al.

based on Hopwood and Trapp (22). These Hepes-bufferedsolutions were bubbled with 100% O2 and contained 118 mMNaCl, 3 mM KCl, 1 mM MgCl2, 1.5 mM CaCl2, 25 mM Hepes,and 1 mM glucose (pH adjusted with NaOH) (22). We alsocarried out experiments where ACSF was buffered with bothbicarbonate and Hepes (27), and changes in its pH were createdby addition of NaOH or HCl. The effects of pHe on firing andcurrents of orexin cells using this solution were similar to thoseshown in Figs. 1–3 (n � 15, data not shown). Intracellular(pipette) solutions were based on either K-gluconate or K-chloride. K-gluconate intracellular solution (used in Figs. 1–3)contained 120 mM K-gluconate, 10 mM KCl, 0.1 mM EGTA, 10mM Hepes, 4 mM K2ATP, 1 mM Na2ATP, and 2 mM MgCl2(pH � 7.3 with KOH). K-chloride solution (used in Figs. 4 and5) contained 130 mM KCl, 0.1 mM EGTA, 10 mM Hepes, 5 mMK2ATP, 1 mM NaCl, and 2 mM MgCl2 (pH 7.3) with KOH.Tetrodotoxin was obtained from Tocris Cookson Ltd. (Avon-mouth, U.K.). All other chemicals were from Sigma–Aldrich(Poole, Dorset, U.K.).

Electrophysiology and Data Analysis. For patch-clamp recordings,living orexin-eGFP neurons were visualized in brain slices usingan Olympus BX50WI upright microscope equipped with infra-red gradient contrast optics, a mercury lamp, and filters forvisualizing eGFP-containing cells (Olympus, Tokyo, Japan)(27). Whole-cell voltage- and current-clamp recordings fromorexin-eGFP neurons were performed using an EPC-9 amplifier(HEKA, Lambrecht, Germany) at 25°C as previously described(27). Control experiments showed that physiological alkaliniza-tion of orexin cells (going from pHe 6.9–7.4) is hyperpolarizingand inhibitory, whereas acidification (going from pHe 7.4–6.9)is depolarizing and excitatory at both 25°C (n � 10 cells) and35°C (n � 4 cells). Patch pipettes were pulled from borosilicateglass and had tip resistances of 2–3 M� when filled with theK-chloride pipette solution. Series resistances were in the rangeof 5–8 M� and were not compensated. Data were sampled andfiltered using Pulse/Pulsefit software (HEKA) and analyzed withPulsefit and Origin (Microcal, Northampton, MA) software.

To monitor membrane resistance together with changes inmembrane potential (Figs. 1 and 4B), cells were injected every10–20 sec with a fixed-amplitude (10–40 pA, 1-sec duration)

square pulse of hyperpolarizing current (24). The amplitude ofthe resulting downward deflections in membrane potential isproportional to membrane resistance (Ohm’s law: resistance �voltage divided by current), i.e., inversely proportional to mem-brane conductance (conductance � current divided by voltage �1 over resistance).

Current-voltage relationships (Fig. 5 A and B) were obtainedby performing voltage-clamp ramps from 0 mV to �140 mV ata rate of 0.1 mV/msec ramp, which is sufficiently slow to allowleak-like K� currents to reach steady-state at each potential (27,44). In Fig. 5B, the net current-voltage relationship was fittedwith the Goldman–Hodgkin–Katz current equation (8) in thefollowing form:

I � PKz2VF2

RTK� i � K�oexp��zFV /RT�

1 � exp��zFV /RT�

Here I is current, V is membrane potential, z is the charge ofa potassium ion (�1), [K�]i is the pipette K� concentration,[K�]o is the ACSF K� concentration, T is temperature (298 K,corresponding to 25°C), and R and F have their usual meanings(8). PK (a constant reflecting the K� permeability of themembrane) was the only free parameter during fits. The value ofPK used to obtain the fit shown in Fig. 5B was 0.00686. Valuesare presented as means � SEM.

To calculate chemosensitivity as percent firing change per 0.1unit change in pHe, we adapted the method described in Putnamet al. (4) and used the following equation:

% change in FR �100�FRA � FRALK�/�FRALK�

pH/0.1.

Here FR is firing rate, FRA is FR in acidic solution, FRALK isFR in alkaline solution, and pH is the change in pH. The valueof % change in FR given in Results (108%) was calculated usingFRA and FRALK of 3.7 Hz and 1 Hz, respectively, which weretaken from data in Fig. 2B at pHe of 7 and 7.25, respectively( pH � 0.25).

This work was supported by the Royal Society (D.B.), the Biotechnologyand Biological Sciences Research Council (studentship for R.H.W.), andthe Karen Elise Jensen Foundation of Denmark (L.F. and L.T.J.).

1. Nattie E (1999) Prog Neurobiol 59:299–331.2. Phillipson E, Bowes G (1986) in Handbook of Physiology. The Respiratory System

(American Physiological Society, Bethesda), pp 649–689.3. Richerson GB (2004) Nat Rev Neurosci 5:449–461.4. Putnam RW, Filosa JA, Ritucci NA (2004) Am J Physiol 287:C1493–C1526.5. Hunt CE (1992) in Respiratory Control Disorders in Infants and Children, eds

Beckerman RC, Brouillette RT, Hunt CE (Williams & Wilkins, Baltimore), pp190–211.

6. Richerson GB (1997) Neuroscientist 3:3–7.7. Chesler M (2003) Physiol Rev 83:1183–1221.8. Hille B (2001) Ion Channels of Excitable Membranes (Sinauer, Sunder-

land, MA).9. Redgate ES, Gellhorn E (1958) Am J Physiol 193:189–194.

10. Saper CB, Chou TC, Scammell TE (2001) Trends Neurosci 24:726–731.11. Sakurai T, Amemiya A, Ishii M, Matsuzaki I, Chemelli RM, Tanaka H,

Williams SC, Richardson JA, Kozlowski GP, Wilson S, et al. (1998) Cell92:573–585.

12. de Lecea L, Kilduff TS, Peyron C, Gao X, Foye PE, Danielson PE, FukuharaC, Battenberg EL, Gautvik VT, Bartlett FS II, et al. (1998) Proc Natl Acad SciUSA 95:322–327.

13. Sakurai T (2007) Nat Rev Neurosci 8:171–181.14. Sutcliffe JG, de Lecea L (2002) Nat Rev Neurosci 3:339–349.15. Young JK, Wu M, Manaye KF, Kc P, Allard JS, Mack SO, Haxhiu MA (2005)

J Appl Physiol 98:1387–1395.16. Nakamura A, Zhang W, Yanagisawa M, Fukuda Y, Kuwaki T (2007) J Appl

Physiol 102:247–248.17. Lin L, Faraco J, Li R, Kadotani H, Rogers W, Lin X, Qiu X, de Jong PJ, Nishino

S, Mignot E (1999) Cell 98:365–376.

18. Chemelli RM, Willie JT, Sinton CM, Elmquist JK, Scammell T, Lee C,Richardson JA, Williams SC, Xiong Y, Kisanuki Y, et al. (1999) Cell 98:437–451.

19. Siegel JM, Boehmer LN (2006) Nat Clin Pract Neurol 2:548–556.20. Yamanaka A, Beuckmann CT, Willie JT, Hara J, Tsujino N, Mieda M,

Tominaga M, Yagami K, Sugiyama F, Goto K, et al. (2003) Neuron 38:701–713.21. Kintner DB, Anderson MK, Fitzpatrick JH, Jr, Sailor KA, Gilboe DD (2000)

Neurochem Res 25:1385–1396.22. Hopwood SE, Trapp S (2005) J Physiol 568:145–154.23. Burdakov D, Alexopoulos H, Vincent A, Ashcroft FM (2004) Eur J Neurosci

20:3281–3285.24. Burdakov D, Gerasimenko O, Verkhratsky A (2005) J Neurosci 25:2429–2433.25. Burdakov D (2004) Neuroscientist 10:286–291.26. Rosenkranz JA, Johnston D (2006) J Neurosci 26:3229–3244.27. Burdakov D, Jensen LT, Alexopoulos H, Williams RH, Fearon IM, O’Kelly I,

Gerasimenko O, Fugger L, Verkhratsky A (2006) Neuron 50:711–722.28. Bayliss DA, Sirois JE, Talley EM (2003) Mol Interv 3:205–319.29. Li Y, van den Pol AN (2005) J Neurosci 25:173–183.30. Li Y, Gao XB, Sakurai T, van den Pol AN (2002) Neuron 36:1169–1181.31. Burdakov D (2007) Pflugers Arch 454:19–27.32. Lopez-Barneo J (2003) Curr Opin Neurobiol 13:493–499.33. Kumar P, Bin-Jaliah I (2007) Respir Physiol Neurobiol 157:12–21.34. Mulkey DK, Stornetta RL, Weston MC, Simmons JR, Parker A, Bayliss DA,

Guyenet PG (2004) Nat Neurosci 7:1360–1369.35. Peyron C, Tighe DK, van den Pol AN, de Lecea L, Heller HC, Sutcliffe JG,

Kilduff TS (1998) J Neurosci 18:9996–10015.36. Lambe EK, Aghajanian GK (2003) Neuron 40:139–150.37. Georgescu D, Zachariou V, Barrot M, Mieda M, Willie JT, Eisch AJ,

Yanagisawa M, Nestler EJ, DiLeone RJ (2003) J Neurosci 23:3106–3111.

Williams et al. PNAS � June 19, 2007 � vol. 104 � no. 25 � 10689

NEU

ROSC

IEN

CE

38. Kayaba Y, Nakamura A, Kasuya Y, Ohuchi T, Yanagisawa M, Komuro I, FukudaY, Kuwaki T (2003) Am J Physiol 285:R581–R593.

39. Suzuki M, Beuckmann CT, Shikata K, Ogura H, Sawai T (2005) Brain Res1044:116–121.

40. Winsky-Sommerer R, Yamanaka A, Diano S, Borok E, Roberts AJ, Sakurai T,Kilduff TS, Horvath TL, de Lecea L (2004) J Neurosci 24:11439–11448.

41. Sakamoto F, Yamada S, Ueta Y (2004) Regul Pept 118:183–191.42. Boutrel B, Kenny PJ, Specio SE, Martin-Fardon R, Markou A, Koob GF, de

Lecea L (2005) Proc Natl Acad Sci USA 102:19168–19173.43. Harris GC, Aston-Jones G (2006) Trends Neurosci 29:571–577.44. Meuth SG, Budde T, Kanyshkova T, Broicher T, Munsch T, Pape HC (2003)

J Neurosci 23:6460–6469.

10690 � www.pnas.org�cgi�doi�10.1073�pnas.0702676104 Williams et al.