www.sciencemag.org/cgi/content/full/science.1190721/DC1

Supporting Online Material for

Astrocytes Control Breathing Through pH-Dependent Release of ATP

Alexander V. Gourine,* Vitaliy Kasymov, Nephtali Marina, Feige Tang, Melina F. Figueiredo, Samantha Lane, Anja G. Teschemacher, K. Michael Spyer,

Karl Deisseroth, Sergey Kasparov*

*To whom correspondence should be addressed. E-mail: a.gourine@ucl.ac.uk (A.V.G.);

sergey.kasparov@bristol.ac.uk (S.K.)

Published 14 July 2010 on Science Express DOI: 10.1126/science.1190721

This PDF file includes:

Materials and Methods

SOM Text

Figs. S1 to S26

Table S1

References Other Supporting Online Material for this manuscript includes the following: (available at www.sciencemag.org/cgi/content/full/science.1190721/DC1)

Movies S1 to S5

SUPPORTING ONLINE MATERIAL

Astrocytes Control Breathing Through pH-dependent Release of ATP

AV Gourine, V. Kasimov, N. Marina, F. Tang, M.F. Figueiredo, S. Lane, A.G.Teschemacher, K.M. Spyer, K. Deisseroth, S. Kasparov

1. Materials and Methods

All experiments were performed on Sprague-Dawley rats in accordance with the UnitedKingdom Animals (Scientific Procedures) Act of 1986.

1.1 Experimental models

In vivo preparations. The rats were anesthetized with pentobarbitone sodium (initial dose60 mg kg-1, i.p.; then 5-10 mg kg–1h–1, i.v.) Adequate anesthesia was ensured bymaintaining stable levels of arterial blood pressure (ABP), heart and central respiratoryrate showing lack of responses to a paw pinch. The femoral artery and vein werecannulated for measurement of ABP and administration of anesthetic, respectively. Thetrachea was cannulated, the animal was vagotomized and ventilated with a mixture of50% oxygen and 50% nitrogen using a positive pressure ventilator with a tidal volume of~2 ml and a ventilator frequency similar to the resting respiratory rate (~60 strokes min-

1). The animal was then placed in a stereotaxic frame, and the ventral surface of thebrainstem (VS) was exposed as described previously (S13). Activity of the phrenic nervewas recorded as an indicator of central respiratory drive. The signal was amplified (x20,000), filtered (500-1500 Hz), rectified and smoothed ( = 50 ms). For imagingexperiments the animal was paralyzed with gallamine triethiodide (50 mg kg-1, i.v.; then10 mg kg-1h-1, i.v.). PO2, PCO2 and pH of the arterial blood were measured every 1-2 h.End-tidal levels of CO2 were monitored on-line using a fast-response CO2 analyzer andkept at a designated level by altering tidal volume and pump frequency. In all theexperiments PO2 in the arterial blood was kept at >100 mmHg to reduce the drive fromthe peripheral chemoreceptors. The body temperature was maintained at 37.0 0.2°C.

Acute brainstem slices. Animals were killed by halothane inhalation overdose and thebrainstem was quickly removed and placed in chilled (~4°C) artificial cerebrospinal fluid(aCSF; 124 mM NaCl, 3 mM KCl, 2 mM CaCl2, 26 mM NaHCO3, 1.25 mM NaH2PO4, 1 mMMgSO4, 10 mM D-glucose saturated with 95% O2/ 5% CO2, pH 7.4) with an additional 10mM Mg2+. The medulla was isolated and sequences of transverse (250 m) or horizontal(400 m) slices were cut. Once cut, the slices were stored before use in aCSF saturatedwith 95% O2 and 5% CO2 (pH 7.4) at room temperature. Recordings were made in a flowchamber at 35-37°C from the slices placed on an elevated grid to permit access of aCSFfrom both sides of the slice.

Organotypic cultured slices. Rat pups (8–10 days old) were killed with halothaneinhalation overdose and transferred into a laminar flow hood. All procedures wereperformed aseptically. The brainstem was quickly removed and bathed in ice-cold Hank’sBalanced Salt Solution (HBSS) without Ca2+ with added 20 mM glucose (total 25.6 mM),

2

10 mM MgCl2, 1 mM HEPES, 1 mM kynurenic acid, 0.005% phenol red, 100 U ml-1

penicillin, and 0.1 mg ml-1 streptomycin. The medulla was isolated and a sequence oftransverse (250 m) slices was cut. Slices were inspected under a low magnificationdissecting microscope and sections (2-3 per animal) containing the facial nucleus (astandard anatomical landmark for RTN area) were selected for plating. For theexperiments on C1 neurons 3 sections caudal to the facial nucleus were selected.Millicell-CM organotypic culture membrane inserts were used for culturing which werecovered with 1.1 ml plating medium containing 50% Optimem-1, 25% fetal bovine serum(FBS), 21.5% HBSS; 25 mM glucose, 100 U ml-1 penicillin, and 0.1 mg ml-1 streptomycin.After 3 days, the plating medium was removed and DMEM medium containing 10% FBS,2 mM L-Glutamine, 100 U ml-1 penicillin, and 0.1 mg ml-1 streptomycin was added andsubsequently replaced twice a week. Viral vectors (AVV-sGFAP-Case12, AVV-sGFAP-ChR2(H134R)-Katushka1.3, AVV-PRSx8-DsRed2, AVV-PRSx8-EGFP or AVV-SuperI-PRSx8-TN-XXL) were added to the medium at the time of slice culture preparation at5x108–5x1010 transducing units ml-1. Experiments were performed after 7-10 days ofincubation.

Dissociated astrocyte culture. Primary astrocyte cultures were prepared from cerebralcortices, cerebellum and brainstem tissues of rat pups (P2-3) as described (S31). Toprepare cultures of VS astrocytes, 500-m-thick brainstem slices were cut. Slices wereinspected under a low magnification dissecting microscope and sections containing thefacial nucleus were selected. A scalpel microblade was used to dissect the most ventralsurface layer (300 m thick) ~ 0.5-2 mm lateral from the midline. These tissue cuts from2-3 animals were used for primary astrocyte culture preparation. After isolation the cellswere plated on poly-D-lysine-coated coverslips and AVV-sGFAP-Case12 was added to theincubation medium.

1.2 Viral gene transfer in vivo

Young adult rats (~200 g) were anesthetized (ketamine, 60 mg kg-1 and medetomidine,250 mg kg-1; i.m.) and placed in a stereotaxic frame. Ventral regions of the brainstemwere targeted either unilaterally or bilaterally with microinjections of viral vectors (AVV-sGFAP-ChR2(H134R)-Katushka1.3 or AVV-sGFAP-Case12, respectively) using thefollowing coordinates from bregma: X, 1.8 mm; Y, -10.5 mm; Z, -9 mm. The secondmicroinjection was delivered to each side of the medulla 1 mm caudally (Y, -11.5 mm).After microinjections the wound was sutured and anesthesia was reversed withatipemazole (1 mg kg-1). Immediate postoperative care was given and animals were leftto recover for 7-10 days before the experiments to ensure a high level of transgeneexpression.

1.3 Optical imaging

[Ca2+]i responses in VS astrocytes were visualized using a genetically encoded Ca2+

indicator Case12 (S45) expressed in the brainstem via an adenoviral vector withenhanced glial fibrillatory acidic protein (GFAP) promoter (S26). Specificity of thispromoter was confirmed using anti-GFAP immunostaining (fig. S1) and patch clamprecordings of the transduced cells (fig. S19). Simultaneous imaging of pH- and ATP-evoked responses in the same VS astrocytes using Case12 and conventional Ca2+

indicator Fura-2 confirmed Ca2+ sensitivity and dynamic range of Case12 (fig. S2). Notethat Case12 fluorescence declines towards the end of the pH challenge relative to that ofFura-2. This is a result of transient quenching of Case12 fluorescence caused by

3

acidification of the intracellular milieu. As with all other cyclically permutated GFP-basedCa2+ sensors, Case12 fluorescence quenches when pH decreases (S45). Therefore, inmost cases cells imaged with Case12 display a reduction in baseline fluorescence clearlyevident at the end of the pH stimulus or when the pH-evoked Ca2+ responses are blockedin zero Ca2+ media or in the presence of apyrase or bafilomycin A (Fig. 2b and 2d; fig.S10). These decreases in Case12 fluorescence are fully reversible.

In vivo Ca2+ imaging using Case12 (Fig. 1a) was performed using a Leica fluorescencemicroscope and MiCam02 high-resolution camera. Brainstem surface transduced withAVV-sGFAP-Case12 was continuously superfused (4 ml min-1) with HEPES-bufferedsolution (HBS; 137 mM NaCl, 5.4 mM KCl, 0.25 mM Na2HPO4, 0.44 mM KH2PO4, 1.3 mMCaCl2, 1.0 mM MgSO4, 4.2 mM NaHCO3, 10 mM HEPES; pH 7.4-7.45) and bulk [Ca2+]i

responses to a 0.2 unit pH decrease were recorded. As expected, a 0.2 unit decrease inpH on the VS of the brainstem always increased central respiratory drive (fig. S20). Alarge area of the brainstem was imaged at low magnification (Fig. 1a), therefore, ROIs inthis case encompass multiple astrocytes which cannot be individually resolved underthese imaging conditions. Hence the dynamic responses of single cells such as shown inFig. 1b or Fig. 1e are not visible.

Confocal imaging of pH-evoked Ca2+ responses in individual VS astrocytes transducedwith AVV-sGFAP-Case12 was performed in brainstem slices of adult rats (Fig. 1b) andorganotypic brainstem slice cultures (Fig. 1e). For imaging, a section of the membranewith the organotypic slice was cut out and placed on an elevated grid in a flow chamber(volume 2 ml). Acute horizontal slices containing ventral brainstem areas were preparedas described above. To visualize VS blood vessels, acute brainstem slices transduced withAVV-sGFAP-Case12 were incubated in aCSF containing TRITC labeled lectin (10 mg ml-1)and 0.7% Pluronic F127 for 30 min at 37°C. Recordings were performed at 35-37°C inHBS. The rate of perfusion was 4 ml min-1. Images were obtained using an uprightconfocal microscope with 40x water immersion objective. The 488 nm Argon laser linewas used to excite Case12 fluorescence which was measured using a bandpass filter505–550nm. Illumination intensity was kept to a minimum (at 0.5–0.7% of laseroutput).

For [Ca2+]i monitoring in RTN neurons using a genetically encoded indicator TN-XXL,organotypic slices were transduced with AVV-SuperI-PRSx8-TN-XXL as described above.The mechanism of Ca2+ detection by TN-XXL is based on Förster resonance energytransfer (FRET) between two genetically encoded fluorescent proteins, CFP and YFP.Binding of Ca2+ alters the protein conformation leading to an increase in YFP/CFPfluorescence ratio (S30). To enhance TN-XXL expression we used the recently developedIRES-based two step transcriptional amplification strategy (S2; S26). Cells were imagedusing a confocal microscope. For excitation the 456 nm line of argon laser was used,while emission was sampled using 480-520 nm and 535-590 nm filters for CFP and YFPrespectively. At the end of the recordings KCl (40 mM) was applied as a positive controlto estimate a maximal [Ca2+]i increase during neuronal depolarization.

There is a consensus that fluctuations in PCO2 are detected by the medullarychemosensors predominantly indirectly via changes in [H+] following the reaction of CO2

with water (S27; S40). Thus, to elicit chemosensory responses the same HEPES-bufferedsolution (composition is given above) with pH adjusted to 7.2 or 7.0 was applied. pHmeasurements on the VS of the brainstem slice in our perfusion system demonstrated a

4

0.15 unit pH decrease within 30 s from the moment when the solution with lower pH wasexpected to reach the chamber.

For low magnification CCD imaging experiments (fig. S9) transverse medullary slices cutfrom the brains of 15 day-old rats were incubated in aCSF containing 10 M of themembrane-permeant Ca2+ sensitive dye Rhod-2 AM with the addition of 0.05% PluronicF127 for 1 h in the dark at 37°C. Ca2+ imaging was performed using a fluorescencemicroscope, equipped with a high-resolution CCD camera. The excitation filter was520/20 nm, the emission filter was a 590 nm long-pass filter, and the dichroic mirror cut-off at 580 nm. The analogue of respiratory acidosis was induced by perfusing thechamber with aCSF (124 mM NaCl, 3 mM KCl, 2 mM CaCl2, 26 mM NaHCO3, 1.25 mMNaH2PO4, 1 mM MgSO4, 10 mM D-glucose) saturated with a gas mixture containing 7-8%CO2 (with the balance being O2) to reduce pH by ~0.2-0.3 units.

Fluorescence measurements in dissociated cultures of VS astrocytes (fig S5) wereperformed using an epifluorescence inverted microscope equipped with a CCD camera,with a 20x oil immersion objective. Case12 fluorescence was excited at 490 nm andemitted fluorescence at 510-530 nm was collected.

To visualize influx of Ca2+ ions, astrocytes were loaded with Fura-2 AM and incubated inthe medium containing 50 M Mn2+. As the affinity of Fura-2 to Mn2+ is much higher thanto Ca2+, and since the Fura-2-Mn2+ complexes are non-fluorescent at all wavelengths, adecrease of Fura-2 fluorescence would indicate an influx of Mn2+ via Ca2+-permeablechannels (S32). Data were recorded at excitation wavelength at 360nm (Ca2+-insensitiveisobestic point) and emission signal at 510nm (3 s time interval).

1.4 Measurement of ATP release

ATP release (Fig. 2a) was measured using enzyme-based biosensors as previouslydescribed (S13). The sensors were aligned with the pyramidal tracts and placed in directcontact with the VS of the brainstem 0.2-0.5 mm lateral from the pyramidal tracts. Thesensitive part of the sensor was ~2 mm in length placed over a significant portion of thebrainstem surface. ATP and control sensors were placed in equivalent positions on theslice as judged by the distance from the pyramidal tracts and the remaining nerve roots.

A luciferase assay was used to determine whether optogenetic activation of astrocytesinduces release of ATP. Organotypic slices of rat pups were prepared and transduced withAVV-sGFAP-ChR2(H134R)-Katushka1.3 as described above. After 7 days in culture, sliceswere placed into a tissue chamber and superfused with HBS at 34oC at 0.5 ml min-1. ATPlevels in the outflow were measured before and after the stimulation using aluminometer and ATP Bioluminescent assay kit. To activate astrocytes, ventral aspects ofthe slice were illuminated with blue light (445 nm, 1 min, ~7 mW mm-2).

1.5 Generation and validation of AVV-sGFAP-ChR2(H134R)-Katushka1.3

AVV-sGFAP-ChR2(H134R)-Katushka1.3 was produced by homologous recombination aspreviously described (S6; S26). Adenoviral vector was chosen because astrocytes arehighly susceptible to transduction by these vectors (S6). A compact transcriptionallyamplified version of GFAP promoter - GfaABC1D was used to target astrocytes (S26). Thelayout of the vector is shown on Fig. 3e. Because the functionality of the ChR2 is critically

5

dependent on the level of expression we employed a two-step transcriptionalamplification strategy (TSTA)(S26). TSTA relies on a positive feedback loop build into thevector which operates by co-expression of a chimeric transcriptional enhancer Gal4-NFB. In this construct GfaABC1D operates in a bi-directional manner driving expressionof both ChR2(H134R)-Katushka1.3 fusion and the enhancer. This results in a strongenhancement of GFAP activity which is weak otherwise. Transcriptional specificity of thissystem was demonstrated by previous studies (S24; S26) and further confirmed here(fig. S1 and S19).

Katushka1.3 as fluorescent marker was chosen because it can be excited by green oryellow laser light (thus avoiding activation of ChR2) and has an emission peak at ~630nm (S44; fig. S14b). This allowed identification of the cells expressing the construct (figS14), monitoring Ca2+ responses with Rhod-2AM (peak emission at ~575 nm) using ayellow 561 nm laser line of a confocal microscope, and activation of ChR2 with anadditional beam of blue light (fig S21).

Validation and initial characterization of the new vector were performed in primarycultures of the rat astrocytes transduced with AVV-sGFAP-ChR2(H134R)-Katushka1.3and loaded with Rhod-2 AM (fig. S14a). The presence of both fluorophores in the samecell was ascertained using the spectral scan with 561 nm excitation (fig. S14b). In orderto excite ChR2(H134R) with blue light (470 nm) and simultaneously image Rhod-2, thelight path of the SP2 microscope was modified (fig. S21). Flashing blue light (LED 470nm wavelength; 0.3-0.5 mW/mm2; 20/20 ms cycle) triggered robust increases in [Ca2+]i

in all transduced astrocytes (Fig. 3f, fig. S15, S17). The profile of the response dependedon the level of ChR2 expression, intensity and regime of stimulation. At low levels ofboth, increases in [Ca2+]i were readily reversible, while high level of transgene expressionin >70% of astrocytes was associated with sustained Ca2+ responses lasting for morethan 15 min (not shown). Similar results were obtained in acute slices of the adult ratbrainstem transduced with AVV-sGFAP-ChR2(H134R)-Katushka1.3 (fig. S16). Preliminaryexperiments indicated that 20/20 ms duty cycle of stimulation induces Ca2+ responseswhich are more rapid and better reversible in comparison to those induced by acontinuous light.

1.6 Patch clamp recording of RTN neurons and astrocytes, optogenetic activation ofastrocytes in cultured slices

The primary population of central respiratory chemoreceptors is believed to be locatednear the VS, in particular, in the overlapping retrotrapezoid nucleus/parafacial respiratorygroup (RTN/pFRG) (S34; S38). RTN neurons were identified by green or red fluorescencefollowing transduction of organotypic slices with AVV-PRSx8-EGFP (Fig. 3a) or AVV-PRSx8-DsRed2 (Fig. 3g), respectively. PRSx8 is active in RTN neurons because theycharacteristically express Phox2 – the key transcriptional activator of the PRSx8promoter (S1; S20). Organotypic slice cultures containing RTN were prepared asdescribed above. Slices were placed into a recording chamber mounted under an uprightconfocal microscope and patch clamp recordings were made from fluorescent cellslocated near the VS (Fig. 3a, 3g). Slices were continuously superfused with HBS asdescribed above. Neuronal activity was sampled and processed using a 1401 interfaceand CED Spike 2 software.

6

In the experiments involving light stimulation of astrocytes expressing ChR2, slices wereco-transduced with AVV-sGFAP-ChR2(H134R)-Katushka1.3 and AVV-PRSx8-DsRed2 (Fig.3g). RTN neurons were visualized via DsRed2 because illumination of slices with bluelight to identify EGFP-expressing neurons would inevitably excite astrocytes expressingChR2. Therefore, we had to rely on cell size and morphology to distinguish between redfluorescent neurons and astrocytes in search of a suitable target for patching while usingonly yellow 561 nm light for excitation (Fig. 3g). Once patched, cells could be easilyidentified based on their electrophysiological characteristics. After a stable baseline hadbeen recorded, an area surrounding the recording electrode was stimulated by flashing470 nm light (20/20 ms duty cycle, 0.5-2 min). Light-activation of astrocytes evokedclear-cut depolarizations of all tested RTN neurons (n = 13; Fig 3h – left trace). Theseastrocyte-evoked depolarizations of RTN neurons in some cases were very long lasting,with recovery time ranging 4-70 min (mean 28 9 min). Therefore, for the experimentswith ATP receptor blockade we patched additional 8 RTN neurons where light stimulationwas performed twice, first in the presence of MRS2179, and second, after 1 hourrecovery and MRS2179 wash-out (Fig. 3h, right trace).

To assess whether activation of RTN neurons is a stimulus which may trigger astrocyticCa2+ excitation, RTN neurons were transduced to express DsRed2 and the adjacentastrocytes were transduced to express Case12 (fig. S7). RTN neurons (n=11) wereactivated by positive current pulses delivered via a patch pipette to generate spiking at arate of 5–10 Hz for 2 min. Fluorescence in 83 adjacent Case12-expressing VS astrocytesfrom 11 separate experiments was monitored in time-lapse mode (fig. S7).

1.7 Light stimulation of VS astrocytes in vivo

Animals in which the VS had been transduced with AVV-sGFAP-ChR2(H134R)-Katushka1.3 were anesthetized and instrumented as described above. The VS wasexposed using ventral approach and illuminated unilaterally with blue light (445 nmwavelength; 7 mW mm-2, 20/20 ms duty cycle) focused with a miniature objective over adesignated area of the VS. To determine whether activation of VS astrocytes can triggerphrenic nerve activity from apnea, the latter was induced by mechanicallyhyperventilating the animal so that PCO2 in the arterial blood and end-tidal level of CO2

were below the apneic threshold (~30 mmHg and 2%, respectively). At the end of theexperiments the rats were intracardially perfused with 4% paraformaldehyde, the brainswere removed and stored in the same fixative overnight at 4oC. Medulla oblongata wasisolated and a sequence of transverse (30 m) slices was cut and processed forimmunohistochemical detection of Phox2b expression. Tissue was incubated in rabbitanti-Phox2b antibody for 24 h at 4°C (1:800, generous gift of JF Brunet, Ecole NormaleSupérieure, Paris, France) followed by a biotinylated goat anti-rabbit (1:500) antibody for1 hour. Immunostaining was revealed after incubation with Fluorescein avidin DCS(1:250) for 1 hour.

1.8 Data analysis

Built-in analysis software tools were used to analyze the results of the imagingexperiments. Changes in Case12 and Rhod-2 fluorescence of individual cells or regions ofinterest were normalized and expressed as F/F0, where F0 is the initial fluorescence.Records from the patch clamp recordings and in vivo experiments were analyzed off-lineusing CED Spike 2 software. Astrocytic and neuronal Ca2+ responses, changes in

7

neuronal membrane potential and respiratory responses evoked by decreases in pH oroptogenetic activation of astrocytes in the absence and presence of test drugs werecompared using Student’s paired or unpaired t tests, as appropriate. A value of p < 0.05was considered to be significant.

8

2. Supporting Text: Extended Discussion

2.1 Central Respiratory Chemosensitivity

The nature of the central respiratory CO2 chemosensory mechanism has been underintense scrutiny for the past half century (S10; S15; S27; S40). It is a fundamentalhomeostatic mechanism because CO2 provides the major drive to breathe in mammals.Indeed, the mammalian central nervous circuitry that generates respiratory activity issilent in the absence of CO2 and requires CO2 to operate (see for example S39). ArterialPCO2 is also monitored by the peripheral chemoreceptors of the carotid and aortic bodies,however, up to 80% of the ventilatory responses to CO2 are preserved after theirdenervation (S17), indicating that the essential respiratory chemosensors are locatedwithin the CNS.

The anatomical location of the sensors, their cellular identity and the molecularmechanisms underlying detection of rising levels of PCO2/[H

+] remain controversial (S15;S36; S42). The primary population of CO2 chemoreceptors is believed to be located nearthe VS of the brainstem, in particular, in the overlapping retrotrapezoidnucleus/parafacial respiratory group (RTN/pFRG) (S34; S38). In addition, other regionsof the brainstem such as the medullary raphé nuclei, the pontine locus coeruleus andsome others have been proposed to contain respiratory chemoreceptors (S36; S40;S42). Indeed, all these structures contain neurons that are activated by increases inPCO2/[H

+] in vitro (S40). However, whether these “chemoresponsive” neurons are truefunctional respiratory chemosensors is as yet unresolved.

2.2 Astrocytes as Central Respiratory Chemosensors

According to our data, glial cells may play an important role as primary centralrespiratory chemosensors. Astrocytes which reside at and in close proximity to the VSare capable of sensing fluctuations in arterial PCO2 via changes in pH and impart thesechanges into a modified pattern of breathing via ATP-mediated activation of“chemoreceptor” neurons.

Acidification-evoked Ca2+ excitation of VS astrocytes and its propagation is highly unlikelyto be initiated by local neurons. TTX and muscimol were used to minimize neuronalinfluences and failed to affect pH-evoked [Ca2+]i responses in VS astrocytes in any of thepreparations (Fig. 1d, fig S6). Direct recordings from RTN neurons (the only knownchemosensitive neurons in this area, see below) demonstrated, as expected, theireffective silencing in the presence of these compounds (complete cessation of actionpotentials driven by current pulses 2-3 minutes after application of TTX (n=4),hyperpolarisation, decrease in input resistance and disappearance of EPSP afterapplication of muscimol (n=5), data not shown). Thus, astrocytic responses to adecrease in pH must have been abolished or at least strongly reduced in the presence ofTTX and muscimol if they were driven by neurons, but they were not. In addition, wedirectly demonstrate here that selective activation of RTN neurons by current injectionvia a patch pipette does not trigger significant [Ca2+]i elevations even in the immediatelyadjacent astrocytes (fig S7). Together these results effectively eliminate neurons, andRTN neurons specifically, as the primary pH sensors responsible for triggering [Ca2+]i

elevations in local astrocytes.

9

Strong pH-evoked astrocytic [Ca2+]i responses also remain in dissociated primary cellcultures prepared from the ventral regions of the medulla oblongata (fig. S5), but not incortical astrocytes (fig. S8a). These results, taken together with the data fromorganotypic brainstem slices (Fig. 1e), also exclude a contribution of micro-vasculature toastrocytic pH-sensitivity because after a few days in vitro vascular elements degeneratecompletely.

High pH-sensitivity appears to be a distinctive feature of astrocytes which reside at andin a close proximity to the VS and is not shared by astrocytes from at least one otherregion of the medulla oblongata. In our previous studies we failed to detect anysignificant release of ATP in response to CO2 elsewhere in the brainstem, apart from theVS (S13). In this study astrocytes from the dorsal nucleus of the solitary tract wereunresponsive to acidification under the same experimental conditions (fig S8b) whichtriggered robust Ca2+ responses in VS astrocytes. Cortical astrocytes in primary culturesalso appear to be insensitive to this stimulus (fig S8a).

2.3 ATP Release in Response to Chemosensory Stimulation

Previously, using ATP biosensors placed in a direct contact with the VS pia mater inanesthetized rats we demonstrated that an increase in inspired CO2 triggers animmediate release of ATP from the ventral medulla oblongata (S13). We now show that a0.2 unit decrease in pH from its normal value of 7.4 is a sufficient stimulus to trigger ATPrelease (Fig 2a) which is largely responsible for propagation of astrocytic Ca2+ excitation.Indeed, the effect of exogenously applied ATP on [Ca2+]i in astrocytes residing in themarginal surface layer and immediately adjacent to it is very similar to the effect oflowering external pH (Fig. 2d, fig. S22).

Several putative mechanisms of ATP release have been demonstrated in a variety of glialpreparations, including exocytosis and the opening of gap junction hemichannels (S3;S5; S46; S51). Based on our pharmacological analysis we conclude that in response toacidification, VS astrocytes release ATP mainly through exocytosis (from conventional orunconventional compartments, see S16), while functional hemichannels, includingconnexins and pannexins, do not appear to be involved to a major extent.

2.4 ATP Receptors Responsible for Propagation of Astrocytic Ca2+ excitation

In the presence of the ATP-hydrolyzing enzyme apyrase, pH-evoked Ca2+ excitation in VSastrocytes was reversibly abolished (Fig. 2b). Three different ATP receptor antagonists:MRS 2179 (3 M), PPADS (5 M) or TNP-ATP (20 nM) all markedly reduced (by about80%) acidification-induced [Ca2+]i signals in these astrocytes (Fig. 2e). MRS2179, widelyconsidered as selective antagonist of metabotropic P2Y1 receptors (IC50 = 0.3 M), isalso effective at rat P2X1 (IC50 = 1.1 M) and able to inhibit P2X3 (IC50 = 12.9 M)receptors (S4; S21). PPADS is a broad spectrum antagonist of many ionotropic P2X andmetabotropic P2Y receptors (S41), while TNP-ATP is a high-affinity (nanomolar range)antagonist of P2X receptors, particularly potent at rat P2X1 (IC50 = 6 nM), P2X3 (IC50 =0.9 nM) and heteromeric P2X2/3 receptor assemblies (IC50 = 7 nM) (S50). Thispharmacological profile suggests the involvement of ionotropic P2X receptors, likely tocontain the P2X1 receptor subunit. Indeed, expression of P2X1 receptor subunit inastrocytes has been demonstrated recently (S22).

10

The involvement of ionotropic P2X receptors is also supported by the fact thatacidification- as well as ATP-induced [Ca2+]i responses in astrocytes are absent in Ca2+-free medium (with an addition of 2 mM Mg2+ and 1 mM EGTA to chelate any residualCa2+) (figs. S10, S22). Furthermore, VS astrocytes loaded with Fura-2 AM and incubatedin the medium containing 50 M Mn2+ showed an immediate decrease in fluorescence inresponse to ATP (n=8 slices; fig S23). As the affinity of Fura to Mn2+ is much higher thanto Ca2+, and since the Fura-Mn2+ complexes are non-fluorescent, these observationsdirectly demonstrate opening of Ca2+-permeable channels.

2.5 Dynamics of Acidification-Induced [Ca2+]i Responses in VS Astrocytes

In the in vitro preparations used in this study (acute slices, organotypic cultured slicesand dissociated cell cultures), astrocytes imaged using Case12 responded to acidificationwith the increases in [Ca2+]i which in many cases were not sustained for the duration ofthe stimulation. It is well known, however, that activation of chemosensitive brainstemneurons and respiratory activity is maintained during and for a short period after thetermination of the hypercapnic stimulus, therefore, such a discrepancy warrants adiscussion.

In order to confirm Case12 specificity and sensitivity we simultaneously imaged pH- andATP-evoked Ca2+ responses in the same VS astrocytes using Case12 and conventionalCa2+ indicator Fura-2 and found that changes in Case12 fluorescence faithfully followchanges in Fura-2 fluorescence (fig. S2). However, these experiments also demonstratedthat towards the end of the pH stimulation period, Case12 fluorescence decreases.Intrinsic pH-sensitivity of Case12 is responsible for the reduction of Case12 fluorescencein experiments where pH is lowered for lengthy periods of time. Case12, as all other Ca2+

sensors based on cyclically permutated GFP, is sensitive to acidification and this becomesnoticeable at pH 7.2 and below (S45). When HEPES buffered solution with lower pH isapplied, decrease in intracellular pH is not immediate, thus allowing detection of cellular[Ca2+]i responses. If the stimulus is strong and applied for a longer period of time Case12fluorescence decreases, as evident from many traces at the end of the pH stimulus orwhen the pH-evoked Ca2+ responses are blocked in zero Ca2+ media or in the presence ofapyrase or bafilomycin A (Fig. 2b and 2d; fig. S10). When a conventional indicator (e.g.Rhod-2) is used, acidification-induced increases in [Ca2+]i on the VS are always sustainedfor the duration of the pH stimulation episode (fig. S6b).

This is further confirmed by the in vivo experiments demonstrating that the lateralaspects of the VS, corresponding to the anatomical location of the RTN, display sustainedastrocytic [Ca2+]i responses to changes in pH (Fig. 1a). Moreover, amperometric ATPbiosensors detect an increase in ATP release from the VS in response to acidificationwhich is also maintained during the whole period of a decreased pH (Fig. 2a). Thissuggests that the VS astrocytic network is activated for the duration of pH stimulationand confers this activation onto local “chemoreceptor” neurons which drive the increasein respiratory activity (see detailed discussion below). It is also conceivable, however,that on a longer time scale, other, possibly Ca2+ independent, mechanisms of ATP releasecould be engaged (S16)

11

2.6 The “Initial Chemosensory Event“

Together the data reported and discussed here suggest very strongly that (i) VSastrocytes release ATP in response to a decrease in pH and (ii) [Ca2+]i responsesdetected in VS astrocytes are largely mediated by prior release of ATP. The most logicalconclusion from these observations is that astrocytes release ATP in response toacidification and ATP amplifies and propagates this signal by acting on the same andadjacent astrocytes as well as local chemoreceptor neurons to produce an enhancementin the respiratory activity. Indeed, essentially the same sequence of events can bemimicked by optogenetic activation of astrocytes using ChR2 which brings about the keyphysiological outcome of central chemoreception – an increase in respiratory activity.

However, the key initial mechanism linking a decrease in pH with [Ca2+]i elevation and(presumably) exocytotic ATP release as yet remains obscure. In search of this “primarymechanism” we tested pharmacologically a possible involvement of several plausible pH-sensitive mechanisms, including pH-sensitive K+, TRPV and TRPP channels (Table S1).None of the agents used and known to interfere with a range of these channels had anyeffect on pH-evoked [Ca2+]i responses in VS astrocytes (fig. S24 and Table S1). Our dataalso do not support the idea that membrane depolarization of astrocytes is the key initialevent, since depolarization of VS astrocytes by >20 mV (significantly more in comparisonto depolarizations reported to be evoked by a pH decrease (~5 mV) (S43) via a patchpipette had no significant effect on [Ca2+]i in the depolarized cell or neighboringastrocytes (fig. S25). Thus, the key initial pH chemosensory transduction event remainsunknown and requires a wider screening of other potential targets.

2.7 ATP mediates pH-evoked Responses of Brainstem Respiratory Neurons and RTNChemoreceptor Neurons

ATP release is responsible not only for the spread of Ca2+ excitation among VS astrocytes– it also imparts this astrocytic activation to the respiratory network, leading to amodified pattern of breathing. Indeed, robust astrocytic Ca2+ responses were detected inthe areas located just beneath, and in a close proximity to, the brainstem respiratoryrhythm and pattern generator. The stimulatory effects of ATP on the activity of therhythm-generating circuits and individual medullary respiratory neurons have beendemonstrated in several studies from our laboratory and others (S12; S13; S19; S28;S29; S49). Blockade of ATP receptors following microinjections of P2 antagonists into theventral medulla or by direct application of the antagonists onto the VS markedly reducesCO2-evoked increases in breathing (S13; S48). At a single cell level, blockade of ATPreceptors abolishes CO2-evoked excitation of individual medullary inspiratory neurons(S49). Interestingly, a recent study has demonstrated involvement of astroglia inpurinergic modulation of the activity of the preBötzinger complex, an area directlyinvolved in respiratory rhythmogenesis (S18).

In this study we have also demonstrated that pH-evoked responses of Phox2b-expressing chemoreceptor neurons located in the RTN are largely mediated by priorrelease of ATP. RTN, which is adjacent to the VS, has been advocated to play a crucialrole in central respiratory CO2/pH chemosensitivity (S7; S15). It contains a distinctpopulation of Phox2b-expressing pH-sensitive neurons which either reside within themarginal glial layer or have extensive projections to it (Mulkey et al., 2004). Phox2b-expressing RTN neurons project to the brainstem respiratory network and their activation

12

leads to an increase in respiratory activity (S1). In this study we specifically targetedthese neurons using AVV-PRSx8-EGFP, ADD-PRSx8-DsRed2 (to identify RTN neurons forpatch clamp recordings) or AVV-SuperI-PRSx8-TN-XXL (to monitor their [Ca2+]i).Although MRS2179 had no effect on acidification-induced responses of 2 cells (whichwere also insensitive to applied ATP; not shown), in the majority (85%) of Phox2b-expressing neurons located near the VS pH-evoked depolarizations were inhibited whenATP actions were blocked in the presence of MRS2179 or apyrase (Fig. 3a, 3b; fig. S13).MRS2179 also effectively abolished [Ca2+]i responses to decreases in pH in 9 additionalRTN neurons imaged using TN-XXL (Fig. 3c and 3d). A previous study (S33) has failed toreveal such a role for ATP and, therefore, such a discrepancy warrants a detaileddiscussion.

Mulkey et al (S33) did not specifically target Phox2b-expressing cells – recordings weremade from a sample of neurons located within the anatomical region of RTN in acutetransverse slices prepared from neonatal rats. In our study RTN neurons were identifiedusing adenoviral vectors AVV-PRSx8-EGFP or AVV-SuperI-PRSx8-TN-XXL (Phox2b is thekey transcriptional activator of the PRSx8 promoter) in organotypic brainstem slices cutspecifically at the level of RTN. Patch clamp or [Ca2+]i recordings were made fromfluorescent cells located near the ventral edge of the slice immediately ventral to thefacial nucleus. In our sample ventrally located neurons expressing EGFP or TN-XXLdisplayed profound pH-evoked responses. In electrophysiological experiments fourneurons were successfully loaded with Rhod-2 via the patch pipette and two of thesecells were found to contain galanin immunoreactivity (fig. S26) – a peptide found inabout 50% of all RTN chemosensitive neurons (S47). In addition, in a separateexperiment we targeted EGFP-positive neurons in an anatomical location caudal to thefacial nucleus (corresponding to the C1 group of Phox-2b-positive cells) and found thatthey are insensitive to changes in pH (n = 9; fig. S12) confirming results of a previousreport (S23). Thus, we conclude with a high degree of confidence that the units recordedin our experiments are representative of the population of VS chemoreceptor Phox-2b-expressing RTN neurons.

Electrical properties of units recorded in our experiments could be slightly different fromthose reported by Mulkey et al (S33; S34), due to differences in preparation, perfusionmedia and recording configurations. Mulkey et al (S33; S34) utilized a “loose patch”mode, which is semi-extracellular, i.e. without full dialysis of the cell with the pipettesolution. Reported RTN neurons displayed slow spontaneous discharge (S33; S34) whichin neurons recorded in our experiments often (although not always, see fig. S7) ceasedafter a few minutes as the cells became dialyzed and slightly hyperpolarized (quitecommon for whole cell recordings). However, depolarizations of RTN neurons in responseto changes in pH in our experiments are very similar in magnitude to those reported byRitucci et al (S43) from RTN neurons in acute slices of neonatal rats.

There is another methodological reason to explain why the results of Mulkey et al. (S33)are not directly comparable with the data presented in our study. That study wasconducted at room temperature where sensitivity of glial “ATP-releasing mechanism” tochanges in pH would be expected to be significantly reduced. As in the Mulkey et al study(S33) ATP release was not demonstrated in conjunction with the effect of an antagonist(or the lack of such an effect) these data remain inconclusive. In support of the datapresented here, Mulkey et al (S33) demonstrated that the RTN neurons are directlyexcited by ATP through P2Y receptors (MRS2179 is a potent antagonist of P2Y1 receptors,

13

see above) and, therefore, are fully equipped to sense changes in local levels of ATPreleased in response to pH change by astrocytic networks in which these neurons areembedded (Fig 4g). At the same time, our results do not exclude the possibility ofintrinsic pH-sensitivity of one or more types of neurons located elsewhere in thebrainstem.

2.8 Activation of Astrocytes using ChR2: Ca2+ Responses and Physiological Effects

In order to determine the functional significance of astrocytic Ca2+ excitation onbreathing we have expressed a H134R mutant of ChR2 (which is more potent than theparent ChR2 protein (S14) selectively in astrocytes under the control of an enhancedcompact GFAP promoter (S26). Ca2+ permeability of ChR2 expressed in Xenopus oocytesand HEK293 cells was documented previously (S25; S35). In our experiments werecorded precisely timed depolarizations (18.8 2.1 mV, n = 8; fig. S19) and strongCa2+ signals in astrocytes expressing ChR2 in response to blue light at physiologicallevels of extracellular Ca2+ (Fig. 3f; fig. S15-S17).

For several reasons ChR2-induced Ca2+ signals in astrocytes cannot be considered as anexact replica of the physiological “Ca2+ excitation” of these cells. Astrocytes do notoperate using fast Na+-mediated action potentials and do not possess high density ofvoltage-gated Ca2+ channels. Indeed, although in astrocytes depolarization of the cellmembrane exactly follows the light stimulation (fig. S19), changes in membranepotential per se do not appear to trigger increases in [Ca2+]i because depolarization ofpatched astrocytes from the ventral regions of the brainstem by >20 mV failed to elicitsignificant increases in [Ca2+]i in the same and adjacent astrocytes (fig S25).

When external Ca2+ is removed, Ca2+ excitation of astrocytes in response to ChR2activation with light is markedly reduced, indicating an influx of Ca2+ from theextracellular milieu (fig S17). However, even after 30 min in zero Ca2+ media theseresponses were not completely abolished (fig S17) suggesting that ChR2 activation inthese cells also recruits Ca2+ from the internal stores via an as yet unknown mechanism.

We demonstrate here directly that activation of astrocytes expressing ChR2 using lightalso results in the release of ATP (fig. S18). In patch clamp experiments, activation ofthe adjacent Phox2b-expressing RTN neurons is prevented in conditions of ATP receptorblockade (Fig. 3h, 3i). Importantly, in in vivo experiments we found that selectivestimulation of ventral brainstem surface astrocytes using this approach activatesbreathing in an ATP-dependent manner and is also sufficiently powerful to induce strongrespiratory activity even in those animals kept below apneic threshold by mechanicalhyperventilation (Fig. 4a, 4c).

One distinctive feature of ChR2 activation in astrocytes is the longevity of the evokedCa2+ responses – in many cases [Ca2+]i elevations in astrocytes significantly outlasted theduration of the light stimulation period (Fig. 3f; fig S15, S16). The profile of these Ca2+

responses clearly depends on a combination of the level of expression of ChR2 and theintensity of light stimulation which is relatively easy to control in vitro but difficult in vivo.With lower level of expression and light intensity the Ca2+ responses following ChR2stimulation tended to be quickly reversible, while stronger expression typically led to thelengthening of the response so that it could last for many minutes (e.g. fig. S16). It isalso possible that ATP-induced ATP release could amplify and prolong these responses

14

under certain circumstances. The exact mechanisms responsible for long Ca2+ responsesare currently under investigation and are beyond the scope of this study. Mostimportantly, we demonstrate here that the duration of light-induced Ca2+ responses inastrocytes expressing ChR2 matches long-lasting responses in the adjacent Phox2b-expressing neurons (see above) and the prolonged episodes of increased respiratoryactivity in in vivo rats following light stimulation of VS astrocytes (Fig. 4b). However, theuse of light of moderate intensity to drive VS astrocytic excitations allows more precisecontrol of respiratory activity as shown on Fig 4c and movie S5.

2.9 Summary

More than 30 years ago Fukuda et al (S11) reported depolarization of brainstem surface“silent cells” in response to a decrease in pH. A few subsequent reports confirmed thatsurface glial cells do indeed respond to chemosensory challenges (S37; S43) and thatpharmacological manipulation of glial function near the ventral medulla oblongata leadsto changes in ventilation (S8; S9). The current study capitalizes on the previous evidenceand demonstrates that astrocytes which reside at and in close proximity to the VS arehighly chemosensitive and respond to increases in [H+] within a physiological range withrobust elevations in [Ca2+]i. Mimicking acidification-induced Ca2+ responses by selectivelight-induced activation of these astrocytes induces depolarization of Phox2b-expressingRTN chemoreceptor neurons in an ATP-dependent manner and enhances centralrespiratory drive. This illustrates a potentially pivotal role of CNS astrocytes in afundamental physiological mechanism – the central respiratory chemoreceptor reflex –and strengthens the case for the active involvement of astrocytes in information-processing within the brain.

Acknowledgements

We gratefully acknowledge B-H. Liu and J. Hewinson for their help with the preparation ofviral vectors, J-F. Brunet (Département de Biologie, Ecole Normale Supérieure, Paris,France) for providing us with the Phox2b antibody, O. Griesbeck (Martinsried, Germany)for providing TN-XXL clone, and B. King for expert advice on the pharmacology of P2receptors.

15

SUPPORTING FIGURES

Fig. S1 Vector used to visualize Ca2+ responses in astrocytes. (A) Layout of the AVV-sGFAP-Case12 expression cassette. This vector expresses Case12 under the control of ashortened version of GFAP promoter, GfaABC1D. mCMV operating in the antisenseorientation drives expression of a chimeric transcriptional activator Gal4p65 while thespecificity of expression in both directions is determined by GfaABC1D. (B) Case12expression in dissociated culture, organotypic brainstem slices and following transductionof the adult rat ventral brainstem areas with AVV-sGFAP-Case12 in vivo. Note thatCase12 concentrates mainly in the somata while GFAP accumulates in the astrocyticprocesses. Therefore, they mainly co-localize in cell processes and on rare occasionswithin the cell bodies. Bar = 20 m.

16

Fig. S2 Simultaneous imaging of pH- and ATP-evoked responses in the same ventralsurface astrocytes using Case12 and conventional Ca2+ indicator Fura-2 to confirm Ca2+

sensitivity and dynamic range of Case12 (dissociated VS astrocyte culture). Note thatfollowing application of the acidic solution Case12 fluorescence is reduced due to itsreversible quenching. This is a common feature of all sensors based on cyclicallypermutated green fluorescent protein.

17

Fig. S3 (A) Ventral medullary astrocytes visualized by Case12 fluorescence. Arrowspoint at glia limitans. (B) Penetrating arteriole enwrapped by processes of astrocytesexpressing Case12. Coronal brainstem sections (50 m) from 2 individual rats used for invivo imaging experiments (see Fig 1a).

Fig. S4 A representative example of pH-evoked [Ca2+]i oscillations in VS astrocytes oforganotypic brainstem slice culture transduced with AVV-sGFAP-Case12. Astrocytesimaged in vitro in some cases (~10%) responded to a decrease in pH with [Ca2+]i

oscillations with gradually diminishing amplitude lasting for the whole duration of thestimulus. Traces illustrate pH-evoked changes in Case12 fluorescence of six individualastrocytes.

VS

20 μm 20 μm

VS

VS

20 μm 20 μm

VS

A B

18

Fig. S5 Primary culture of VS astrocytestransduced with AVV-sGFAP-Case12. A decreasein pH from 7.4 to 7.2 induces rapid transientincreases in [Ca2+]i as determined by changes inCase12 fluorescence.

Fig. S6 (A) TTX (1 M) does notprevent acidification-induced Ca2+

excitation of VS astrocytes. [Ca2+]i

responses were measured aschanges in Case12 fluorescence infour individual ventral surfaceastrocytes (slice preparation of anadult rat). (B) Extracellularacidification (decrease in pH from7.4 to 7.0), in the presence of TTX(1 M) and muscimol (100 M)induces rapid, biphasic andsustained for the duration of thepH stimulation episode [Ca2+]i

signals in Rhod-2 AM loaded VScells. These cells are likely to beastrocytes as we have directlyconfirmed TTX- and muscimol-induced neuronal silencing (seetext for details). Slice preparationof an adult rat (n=5).

pH 7.2pH 7.4

50 μm

A

B

19

Fig. S7 Activation of RTN neurons by injection of positive current fails to triggersignificant [Ca2+]i elevations in adjacent astrocytes. For these experiments RTN neuroneswere targeted with AVV-PRSx8-DsRed2 to express red fluorescence and astrocytes weretransduced with AVV-sGFAP-Case12. Using an SP2 confocal microscope, RTN neuroneswere identified and recorded in whole cell mode. Case12 fluorescence was monitoredusing the 488 nm laser line. Fluorescence intensities of astrocytes (upper image)adjacent to the patched neuron were monitored while the RTN neuron was induced to fireaction potentials at 5–10 Hz for ~2 minutes by depolarizing current pulses delivered viathe patch pipette (lower trace). Activation of the RTN neuron did not produce anysubstantial [Ca2+]i elevations in adjacent astrocytes. Some cells exhibited spontaneouschanges in [Ca2+]i which are not timed with the activity of the neuron. Note that theneuron is not visible on this image because it is expressing DsRed2, while the system istuned to monitor Case12 (green fluorescence, the image is pseudo-colored).

20

Fig. S8 Astrocytes of the cerebral cortex (A, dissociated cell culture) and the dorsalmedullary nucleus of the solitary tract (B, slice preparation of the adult rat brainstemtransduced with AVV-sGFAP-Case12 10 days prior to the experiment) are insensitive toacidification. Astrocytes were identified by Case12 fluorescence. Decrease in pH from 7.4to 7.2 led to a reduction in Case12 signal (Case12 fluorescence quenches when pHdecreases) in both preparations. Application of ATP (100 M) evoked profound elevationsin [Ca2+]i in cortical and NTS astrocytes, an effect similar to that elicited by ATP in VSastrocytes.

Fig. S9 Low magnification Rhod-2 AM imaging ina coronal slice of the rat brainstem illustratingthat CO2-evoked decrease in pH at constant[HCO3

-] evokes robust [Ca2+]i responsesoriginating and propagating near the ventralsurface (see also associated movie S4).

pH 7.15

VS

pH 7.4

300 μm

pH 7.15

VS

pH 7.4

300 μm

21

Fig. S10 Acidification-evoked [Ca2+]i responses in VS astrocytes depend on extracellularCa2+. Traces illustrate pH-induced responses of six individual ventral brainstem surfaceastrocytes in organotypic brainstem slice in control conditions and in the absence ofexternal Ca2+.

Fig. S11 (A, B) Blockade of pannexin andconnexin hemichannels with carbenoxolone(CBX, 10 M) and lanthanum (La3+, 100M) had no effect on [Ca2+]i responsesevoked in VS astrocytes by extracellularacidification (slice preparation of an adultrat). CBX is a broad-spectrum gap junctionhemichannel blocker which at 10 Mconcentration inhibits pannexin, but notconnexin hemichannels. La3+ is a connexinhemichannel blocker that does not affectgap junctions when applied externally. (C)CBX at concentration 100 M which blocksconnexin hemichannels as well as gapjunctions, reduces acidification-induced[Ca2+]i elevations in VS astrocytes by 43%(n=18 cells in 4 slices; p=0.0013).

22

Fig. S12 Catecholaminergic C1 neurons are insensitive to extracellular acidification.Since the PRSx8 promoter which was used to target RTN neurons is also active incatecholaminergic neurons located caudally to RTN we tested whether these cells are pH-sensitive under the same experimental conditions. Slice cultures were made usingsections caudal to RTN and AVV-PRSx8-EGFP-targeted neurons were patched andchallenged with the pH stimulus following the same protocol used to study RTN neurons(see Fig 3, main text). R – resistance tests using current pulses.

Fig. S13 ATP mediates responses of VS chemoreceptor neurons to decreases in pH.Time-condensed record of pH-evoked changes in the membrane potential of an RTNneuron in the absence and presence of ATP hydrolyzing enzyme apyrase (25 U ml-1). AP -action potentials (truncated).

23

Fig. S14 (A) Primary astrocytes expressing ChR2(H134R)-Katushka1.3. (B) Spectrumof fluorescence emission of five individual astrocytes transduced with AVV-sGFAP-ChR2(H134R)-Katushka1.3 and loaded with Ca2+ indicator Rhod-2 AM. 561 nm laser lightwas used for excitation. Peaks of fluorescence of the two fluorophores are clearlyseparated allowing Ca2+ imaging using Rhod-2 within the 570-610 nm band.

0

20

40

60

80

100

120

140

160

600 630 660

Rhod-2 Katushka 1.3

Wave length (nm)

FI (AU)

5700

20

40

60

80

100

120

140

160

600 630 660

Rhod-2 Katushka 1.3

Wave length (nm)

FI (AU)

570

20 μm20 μmA

B

24

Fig. S15 Repetitive stimulations of astrocytes lead to a build-up of the [Ca2+]i response.The same cells were stimulated 3 times using 20-sec long bursts of blue light (20/20 msduty cycle). Stronger stimulation and high levels of ChR2(H134R) expression generated[Ca2+]i elevations which lasted for many minutes.

Fig. S16 Lasting Ca2+ responses triggered by flashing blue light in astrocytes transducedwith AVV-sGFAP-ChR2(H134R)-Katushka1.3 and loaded with Ca2+ indicator Rhod-2 AM.In this example rats were microinjected with AVV-sGFAP-ChR2(H134R)-Katushka1.3.Acute brainstem slices were prepared 7 days later, astrocytes were loaded with Rhod-2and imaged using a confocal microscope.

25

Fig. S17 ChR2-induced [Ca2+]i elevation in astrocytes partially depends on extracellularCa2+. Cultured astrocytes showing moderate level of ChR2(H134) expression were loadedwith Rhod-2 AM. Different areas of the same dish were stimulated. In 2 mM externalCa2+, flashing blue light induced profound [Ca2+]i responses in all the transducedastrocytes. These responses were reduced following 10 min of incubation in zero Ca2+

media. After 30 min of incubation in zero Ca2+ media, only small [Ca2+]i increases wereelicited by light. This suggests that activation of ChR2 in astrocytes could also recruitCa2+ from the internal stores.

26

Fig. S18 Release of ATP followingillumination of VS areas in organotypicbrainstem slices (n=6) transduced withAVV-sGFAP-ChR2(H134R)-Katushka1.3 asevident from a significant increase in ATPconcentration in superfusate (p<0.05).RLU – relative luminescence units.

Fig. S19 Patch clamprecording of an astrocyteexpressing ChR2(H134R)-Katushka1.3 showingimmediate depolarization inresponse to blue light. Notethat the membranepotential faithfully followsthe light pulses.

Fig. S20 Raw data showing an increase in central respiratory drive in response to 0.2unit pH decrease on the VS in in vivo imaging experiments illustrated in Fig 1a(anaesthetized and artificially ventilated rat transduced with AVV-sGFAP-Case12). IPNA –integrated phrenic nerve activity.

27

Fig. S21 Optical path for confocal Ca2+ imaging and light activation of ChR2. Leica SP2upright confocal microscope was modified as following: (i) the standard Hg lamp wasreplaced by a LED-based illumination system; (ii) an additional dichroic mirror wasplaced into the light path with the cut-off at 495 nm. This mirror reflected the light fromthe 470 nm LED onto the specimen while being transparent for both, the yellow 561 nmlaser and the emitted red fluorescence. Thus, activation of ChR2 was possible whilesampling Rhod-2 fluorescence with minimal interference.

5 mW 470 nmLaser diode

Computer

1401

Scanhead

Microscope body

TTLLaser561 nm

Detector560-600 nm

Additionaldichroicmirror highpass 500 nm

Water-immersionobjective

EEmmiitttteedd CCaa22++

--

rreellaatteedd

fflluuoorreesscceennccee

RRhhoodd--22--

ffiilllleedd cceellll

ChR2 eexxcciittaattiioonn

28

Fig. S22 (A) Application of ATP mimics the effect of acidification on [Ca2+]i in VSastrocytes. Traces illustrate changes in Case12 fluorescence of four individual VSastrocytes in organotypic brainstem slice culture transduced with AVV-sGFAP-Case12.Two images illustrate ATP-evoked increases in Case12 fluorescence in VS astrocytes. (B)The dependence of ATP-evoked [Ca2+]i responses in VS astrocytes on extracellular Ca2+

(organotypic brainstem slice culture).

A

B

29

Fig. S23 ATP opens Ca2+-permeable channels in VS astrocytes as measured by Mn2+-induced quenching of Fura-2 fluorescence. Experiments were performed in the presenceof 2 mM Ca2+. Addition of Mn2+ (50 μM) resulted in a minimal basal quenching of Fura-2 fluorescence signal. Application of ATP (100 μM) evoked an immediate opening of Ca2+-permeable channels and rapid quench of Fura-2 fluorescence. Digitonin (10 M) wasadded at the end of the recording to permeabilize cell membranes and obtain amaximum response.

Fig. S24 Summary data showing lack of an effect of various inhibitors of K+ and TRPchannels on acidification-induced Ca2+ responses in VS astrocytes expressed aspercentage of the initial response (horizontal brainstem slice preparations of adult ratstransduced with AVV-sGFAP-Case12). Numbers of individual astrocytes sampled from 3-5separate experiments are given in parentheses. 4-AP, 4-Aminopyridine; TEA,tetraethylammonium. Concentrations of individual drugs are given in Table S1.

30

Fig. S25 Depolarization of VS astrocytes by positive current injections fails to triggersignificant [Ca2+]i elevations in the recorded and adjacent astrocytes. For theseexperiments astrocytes were targeted with AVV-sGFAP-Case12, some of them (arrow)were patched and recorded in whole cell mode. Case12 fluorescence in the same andadjacent astrocytes was monitored using the 488 nm laser line (image on the left,pseudocolor). Fluorescence intensities of astrocytes (upper traces) adjacent to thepatched astrocyte (arrow) were monitored while the cell was depolarized by as much as100 mV by positive current injection via a patch pipette (not visible on this image as it isnot fluorescent, position indicated by white arrow). Even with this non-physiologicaldepolarization only a few adjacent astrocytes responded with mild increases in [Ca2+ ]i.When a weaker stimulus was used (~20 mV) no Ca2+ responses were observed (5patched astrocytes, responses were monitored in 42 adjacent cells, p>0.1, data notshown).

Fig. S26 Galanin-immunoreactivity of apH-sensitive RTN neuron identified byEGFP-expression in the organotypic sliceculture and loaded with Rhod-2.

31

Supporting Table S1

Compound Action ConcentrationEffect onresting [Ca2+]i

in VSastrocytes

Effect onpH-evoked[Ca2+]i

responses

4-Amino-pyridine

Blocker of voltage-dependent K+-channels

0.5-1 mM Increases[Ca2+]i

no

AMG9810 Potent TRPV1receptor antagonist

30 nM no no

Amiloride Na+ channel blocker,TRPP3 channelblocker

0.5 mM no no

Halothane Volatile anesthetic;inhibits NMDAreceptors, ActivatesGABAA receptors andselect 2-pore K+

channels

2 mM no no

Isoflurane Volatile anesthetic;inhibits NMDAreceptors, ActivatesGABAA receptors andselect 2-pore K+

channels

2 mM induces [Ca2+]i

oscillationsno

RutheniumRed

Blocker of TRPVchannels

30 M no no

Tetraethyl-ammonium

Non-selective K+

channel blocker25 mM no no

32

Supporting References

S1. Abbott SB, Stornetta RL, Fortuna MG, Depuy SD, West GH, Harris TE and GuyenetPG. Photostimulation of retrotrapezoid nucleus Phox2b-expressing neurons in vivoproduces long-lasting activation of breathing in rats. J Neurosci 29: 5806-5819,2009.

S2. Benzekhroufa K, Liu BH, Teschemacher AG and Kasparov S. Targeting centralserotonergic neurons with lentiviral vectors based on a transcriptional amplificationstrategy. Gene Ther 16: 681-688, 2009.

S3. Bowser DN and Khakh BS. Vesicular ATP is the predominant cause of intercellularcalcium waves in astrocytes. J Gen Physiol 129: 485-491, 2007.

S4. Brown SG, King BF, Kim YC, Jang SY, Burnstock G and Jacobson KA. Activity ofnovel adenine nucleotide derivatives as agonists and antagonists at recombinantrat P2X receptors. Drug Development Research 49: 253-259, 2000.

S5. Coco S, Calegari F, Pravettoni E, Pozzi D, Taverna E, Rosa P, Matteoli M andVerderio C. Storage and release of ATP from astrocytes in culture. J. Biol. Chem.278: 1354-1362, 2003.

S6. Duale H, Kasparov S, Paton JFR and Teschemacher AG. Differences intransductional tropism of adenoviral and lentiviral vectors in the rat brainstem.Exp Physiol 90: 71-78, 2005.

S7. Dubreuil V, Ramanantsoa N, Trochet D, Vaubourg V, Amiel J, Gallego J, Brunet JFand Goridis C. A human mutation in Phox2b causes lack of CO2 chemosensitivity,fatal central apnea, and specific loss of parafacial neurons. Proc Natl Acad Sci U SA 105: 1067-1072, 2008.

S8. Erlichman JS, Hewitt A, Damon TL, Hart M, Kurascz J, Li A and Leiter JC. Inhibitionof monocarboxylate transporter 2 in the retrotrapezoid nucleus in rats: a test ofthe astrocyte-neuron lactate-shuttle hypothesis. J Neurosci 28: 4888-4896, 2008.

S9. Erlichman JS, Li A and Nattie EE. Ventilatory effects of glial dysfunction in a ratbrain stem chemoreceptor region. J Appl Physiol 85: 1599-1604, 1998.

S10. Feldman JL, Mitchell GS and Nattie EE. Breathing: rhythmicity, plasticity,chemosensitivity. Annu Rev Neurosci 26: 239-266, 2003.

S11. Fukuda Y, Honda Y, Schlafke ME and Loeschcke HH. Effect of H+ on the membranepotential of silent cells in the ventral and dorsal surface layers of the rat medulla invitro. Pflugers Archive 376: 229-235, 1978.

33

S12. Gourine AV, Atkinson L, Deuchars J and Spyer KM. Purinergic signalling in themedullary mechanisms of respiratory control in the rat: respiratory neuronesexpress the P2X2 receptor subunit. J Physiol 552: 197-211, 2003.

S13. Gourine AV, Llaudet E, Dale N and Spyer KM. ATP is a mediator of chemosensorytransduction in the central nervous system. Nature 436: 108-111, 2005.

S14. Gradinaru V, Thompson KR, Zhang F, Mogri M, Kay K, Schneider MB andDeisseroth K. Targeting and readout strategies for fast optical neural control invitro and in vivo. J Neurosci 27: 14231-14238, 2007.

S15. Guyenet PG. The 2008 Carl Ludwig Lecture: retrotrapezoid nucleus, CO2

homeostasis, and breathing automaticity. J Appl Physiol 105: 404-416, 2008.

S16. Hamilton NB and Attwell D. Do astrocytes really exocytose neurotransmitters? NatRev Neurosci 11: 227-238, 2010.

S17. Heeringa J, Berkenbosch A, de GJ and Olievier CN. Relative contribution of centraland peripheral chemoreceptors to the ventilatory response to CO2 duringhyperoxia. Respir Physiol 37: 365-379, 1979.

S18. Huxtable AG, Zwicker JD, Alvares TS, Ruangkittisakul A, Fang X, Hahn LB, Possede CE, Baker GB, Ballanyi K and Funk GD. Glia contribute to the purinergicmodulation of inspiratory rhythm-generating networks. J Neurosci 30: 3947-3958,2010.

S19. Huxtable AG, Zwicker JD, Poon BY, Pagliardini S, Vrouwe SQ, Greer JJ and FunkGD. Tripartite Purinergic Modulation of Central Respiratory Networks duringPerinatal Development: The Influence of ATP, Ectonucleotidases, and ATPMetabolites. J Neurosci 29: 14713-14725, 2009.

S20. Hwang DY, Carlezon WA, Jr., Isacson O and Kim KS. A high-efficiency syntheticpromoter that drives transgene expression selectively in noradrenergic neurons.Human Gene Therapy 12: 1731-1740, 2001.

S21. Jacobson KA, Hoffmann C, Kim YC, Camaioni E, Nandanan E, Jang SY, Guo DP, JiXD, von Kugelgen I, Moro S, Ziganshin AU, Rychkov A, King BF, Brown SG,Wildman SS, Burnstock G, Boyer JL, Mohanram A and Harden TK. Molecularrecognition in P2 receptors: Ligand development aided by molecular modeling andmutagenesis. Nucleotides and Their Receptors in the Nervous Systems 120: 119-132, 1999.

S22. Lalo U, Pankratov Y, Wichert SP, Rossner MJ, North RA, Kirchhoff F andVerkhratsky A. P2X1 and P2X5 subunits form the functional P2X receptor in mousecortical astrocytes. J Neurosci 28: 5473-5480, 2008.

S23. Lazarenko RM, Milner TA, Depuy SD, Stornetta RL, West GH, Kievits JA, Bayliss DAand Guyenet PG. Acid sensitivity and ultrastructure of the retrotrapezoid nucleus inPhox2b-EGFP transgenic mice. J Comp Neurol 517: 69-86, 2009.

34

S24. Lee Y, Messing A, Su M and Brenner M. GFAP promoter elements required forregion-specific and astrocyte-specific expression. Glia 56: 481-493, 2008.

S25. Lin JY, Lin MZ, Steinbach P and Tsien RY. Characterization of engineeredchannelrhodopsin variants with improved properties and kinetics. Biophys J 96:1803-1814, 2009.

S26. Liu B, Paton JF and Kasparov S. Viral vectors based on bidirectional cell-specificmammalian promoters and transcriptional amplification strategy for use in vitroand in vivo. BMC Biotechnol 8: 49, 2008.

S27. Loeschcke HH. Central chemosensitivity and the reaction theory. J Physiol 332: 1-24, 1982.

S28. Lorier AR, Huxtable AG, Robinson DM, Lipski J, Housley GD and Funk GD. P2Y1

receptor modulation of the pre-Botzinger complex inspiratory rhythm generatingnetwork in vitro. J Neurosci 27: 993-1005, 2007.

S29. Lorier AR, Lipski J, Housley GD, Greer JJ and Funk GD. ATP sensitivity ofpreBotzinger complex neurones in neonatal rat in vitro: mechanism underlying aP2 receptor-mediated increase in inspiratory frequency. J Physiol 586: 1429-1446,2008.

S30. Mank M, Santos AF, Direnberger S, Mrsic-Flogel TD, Hofer SB, Stein V, Hendel T,Reiff DF, Levelt C, Borst A, Bonhoeffer T, Hubener M and Griesbeck O. Agenetically encoded calcium indicator for chronic in vivo two-photon imaging. NatMethods 5: 805-811, 2008.

S31. Marriott DR, Hirst WD and Ljungberg MC. Astrocytes. In: Neural cell culture - Apractical approach, edited by Cohen J and Wilkin GP. Oxford: Oxford UniversityPress, 1995, p. 85-96.

S32. Merritt JE, Jacob R and Hallam TJ. Use of manganese to discriminate betweencalcium influx and mobilization from internal stores in stimulated humanneutrophils. J Biol Chem 264: 1522-1527, 1989.

S33. Mulkey DK, Mistry AM, Guyenet PG and Bayliss DA. Purinergic P2 receptorsmodulate excitability but do not mediate pH sensitivity of RTN respiratorychemoreceptors. J Neurosci 26: 7230-7233, 2006.

S34. Mulkey DK, Stornetta RL, Weston MC, Simmons JR, Parker A, Bayliss DA andGuyenet PG. Respiratory control by ventral surface chemoreceptor neurons in rats.Nat Neurosci 7: 1360-1369, 2004.

S35. Nagel G, Szellas T, Huhn W, Kateriya S, Adeishvili N, Berthold P, Ollig D,Hegemann P, Bamberg E, Nagel G, Szellas T, Huhn W, Kateriya S, Adeishvili N,Berthold P, Ollig D, Hegemann P and Bamberg E. Channelrhodopsin-2, a directlylight-gated cation-selective membrane channel. Proc Natl Acad Sci U S A 100:13940-13945, 2003.

35

S36. Nattie E and Li A. Central chemoreception is a complex system function thatinvolves multiple brain stem sites. J Appl Physiol 106: 1464-1466, 2009.

S37. Okada Y, Chen Z, Jiang W, Kuwana S and Eldridge FL. Anatomical arrangement ofhypercapnia-activated cells in the superficial ventral medulla of rats. J Appl Physiol93: 427-439, 2002.

S38. Onimaru H, Ikeda K and Kawakami K. CO2-sensitive preinspiratory neurons of theparafacial respiratory group express Phox2b in the neonatal rat. J Neurosci 28:12845-12850, 2008.

S39. Phillipson EA, Duffin J and Cooper JD. Critical dependence of respiratoryrhythmicity on metabolic CO2 load. J Appl Physiol 50: 45-54, 1981.

S40. Putnam RW, Filosa JA and Ritucci NA. Cellular mechanisms involved in CO2 andacid signaling in chemosensitive neurons. Am J Physiol Cell Physiol 287: C1493-C1526, 2004.

S41. Ralevic V and Burnstock G. Receptors for purines and pyrimidines. Pharmacol Rev50: 413-492, 1998.

S42. Richerson GB. Serotonergic neurons as carbon dioxide sensors that maintain pHhomeostasis. Nat Rev Neurosci 5: 449-461, 2004.

S43. Ritucci NA, Erlichman JS, Leiter JC and Putnam RW. Response of membranepotential and intracellular pH to hypercapnia in neurons and astrocytes from ratretrotrapezoid nucleus. Am J Physiol 289: R851-R861, 2005.

S44. Shcherbo D, Murphy CS, Ermakova GV, Solovieva EA, Chepurnykh TV, ShcheglovAS, Verkhusha VV, Pletnev VZ, Hazelwood KL, Roche PM, Lukyanov S, ZaraiskyAG, Davidson MW and Chudakov DM. Far-red fluorescent tags for protein imagingin living tissues. Biochem J 418: 567-574, 2009.

S45. Souslova EA, Belousov VV, Lock JG, Strömblad S, Kasparov S, Bolshakov AP,Pinelis VG, Lukyanov S, Mayr LM and Chudakov DM. Single fluorescent protein-based Ca2+ sensors with increased dynamic range. BMC Biotechnol 7: 37, 2007.

S46. Spray DC, Ye ZC and Ransom BR. Functional connexin "hemichannels": a criticalappraisal. Glia 54: 758-773, 2006.

S47. Stornetta RL, Spirovski D, Moreira TS, Takakura AC, West GH, Gwilt JM, PilowskyPM and Guyenet PG. Galanin is a selective marker of the retrotrapezoid nucleus inrats. J Comp Neurol 512: 373-383, 2009.

S48. Thomas T, Ralevic V, Gadd CA and Spyer KM. Central CO2 chemoreception: Amechanism involving P2 purinoceptors localized in the ventrolateral medulla of theanaesthetized rat. J Physiol 517: 899-905, 1999.

S49. Thomas T and Spyer KM. ATP as a mediator of mammalian central CO2

chemoreception. J Physiol 523 Pt 2: 441-447, 2000.

36

S50. Virginio C, Robertson G, Surprenant A and North RA. Trinitrophenyl-substitutednucleotides are potent antagonists selective for P2X1, P2X3, and heteromeric P2X2/3

receptors. Mol Pharmacol 53: 969-973, 1998.

S51. Zhang Z, Chen G, Zhou W, Song A, Xu T, Luo Q, Wang W, Gu XS and Duan S.Regulated ATP release from astrocytes through lysosome exocytosis. Nat Cell Biol9: 945-953, 2007.