Vesicular Glutamate Transporter 2 Is Required for Central Respiratory Rhythm Generation But Not for Locomotor Central Pattern Generation

Glutamatergic excitatory neurotransmission is dependent on glutamate release from presynaptic vesicles loaded by three members of the solute carrier family, Slc17a6–8, which function as vesicular glutamate transporters (VGLUTs). Here, we show that VGLUT2 (Slc17a6) is required for life ex utero. Vglut2 null mutant mice die immediately after birth because of the absence of respiratory behavior. Investigations at embryonic stages revealed that neural circuits in the location of the pre-Bötzinger (PBC) inspiratory rhythm generator failed to become active. However, neurons with bursting pacemaker properties and anatomical integrity of the PBC area were preserved. Vesicles at asymmetric synapses were fewer and malformed in the Vglut2 null mutant hindbrain, probably causing the complete disruption of AMPA/kainate receptor-mediated synaptic activity in mutant PBC cells. The functional deficit results from an inability of PBC neurons to achieve synchronous activation. In contrast to respiratory rhythm generation, the locomotor central pattern generator of Vglut2 null mutant mice displayed normal rhythmic and coordinated activity, suggesting differences in their operating principles. Hence, the present study identifies VGLUT2-mediated signaling as an obligatory component of the developing respiratory rhythm generator.


Introduction
Central pattern generators (CPGs) are defined as neuronal circuits capable of producing a rhythmic and coordinated output without the influence of sensory input. The respiratory and locomotor neuronal circuits are two of the better characterized CPGs, although much work remains to fully understand how these networks operate. Glutamatergic neurons are involved in most neuronal circuits of the nervous system and considerable efforts have been made to study the role of glutamate receptors in nervous system signaling using a variety of approaches. It has been difficult to pinpoint the role of glutamatergic neurons in neuronal circuits because of the complexity of glutamate-mediated signaling and the variety of receptors triggered by glutamate. In addition, glutamate is an amino acid used by every cell, which has hampered identification of glutamatergic neurons. The discovery of the vesicular glutamate transporter (VGLUT) proteins, which transport glutamate into presynaptic vesicles (Bellocchio et al., 2000;Takamori et al., 2000Takamori et al., , 2001Fremeau et al., 2001;Herzog et al., 2001), has enabled a novel approach to identify and interfere with presynaptic release of glutamate. Recently, mice with a targeted deletion of the Vglut1 gene were unexpectedly demonstrated to be viable, although they displayed progressive neurodegeneration and died prematurely (Fremeau et al., 2004;Schuske and Jorgensen, 2004;Wojcik et al., 2004).
The pre-Bötzinger complex (PBC) is an inspiratory oscillator network within the respiratory CPG primarily pacing the respiratory rhythm (Smith et al., 1991;Koshiya and Smith, 1999;Feldman and Del Negro, 2006). It is located in the rostroventrolateral medulla and relies on excitatory glutamatergic connections (Funk et al., 1993;Rekling and Feldman, 1998). PBC interneurons discharge through activation of NMDA and non-NMDA glutamate receptors (Funk et al., 1993(Funk et al., , 1997 and are immunopositive for glutamate receptor subunits (Robinson and Ellenberger, 1997;Paarmann et al., 2005). Neurokinin-1-receptor (NK1R)-expressing PBC neurons, essential for maintenance of normal breathing in vivo (Gray et al., 2001;McKay et al., 2005), have been shown to contain Vglut2 mRNA and to receive VGLUT2-positive terminals (Gray et al., 1999;Guyenet et al., 2002;Stornetta et al., 2003a,b) but also have GABAergic and glycinergic terminals (Liu et al., 2002). The locomotor CPG is located in the ventral part of the lumbar spinal cord (Kjaerulff and Kiehn, 1996) and controls rhythmically alternating activity between the left and right sides and flexor-extensor muscles. In vivo, locomotor activity is likely initiated by glutamatergic descending pathways from the brain (Mori et al., 2001). In the in vitro spinal cord preparation, fictive locomotion can be induced by bath application of neurotransmitters or their agonists, like dopamine, 5-HT, and NMDA (Jiang et al., 1999;Cazalets et al., 2000;Nishimaru et al., 2000). Data mainly derived from studies in lamprey and mouse have suggested that the locomotor CPG in both developing and adult vertebrates is built on inhibitory components and excitatory glutamatergic components (Buchanan and Grillner, 1987;Grillner, 1997Grillner, , 2003Kiehn, 2006).
Here, we have genetically eliminated presynaptic release of glutamate mediated by VGLUT2 in the mouse and present evidence that VGLUT2 is necessary for life ex utero. We further demonstrate that respiratory-related rhythm generation in the PBC fails to initiate during embryonic development leading to absence of respiration. Targeted deletion of Vglut2 has no notable consequence on the anatomy of the brainstem but affects the number and morphology of excitatory synaptic vesicles in the brainstem. As a probable consequence, we have observed a complete loss of AMPA/kainate-mediated excitatory synaptic processing in neurons of the PBC area of Vglut2 f/f;PCre mutants. This results in an absence of population synchronized activity. In contrast, activity of the spinal locomotor CPG is spared by the mutation, suggesting that VGLUT2 is specifically required in the respiratory CPG. This Vglut2 f/f;PCre mutant mouse represents the first genetic model showing complete uncompensated functional impairment of the respiratory CPG.

Materials and Methods
Generation of Vglut2 null animals. The Vglut2 targeting construct was built by conventional and Red-ET recombination cloning (Zhang et al., 2000). In the construct, exons 4, 5, and 6 were flanked by an upstream loxP site and downstream by an FRT-flanked neo cassette followed by a second loxP site. A dual prokaryotic/eukaryotic promoter enabling kanamycin resistance selection in bacteria and neomycin resistance in embryonic stem (ES) cells preceded the neo cassette. ES cells (Sv129/R1) were electroporated with the targeting construct and the modified allele was detected by standard PCR using oligos that recognized the 5Ј-loxP site, the neo cassette, and the 3Ј-loxP site. Positive clones were also confirmed by Southern blot analysis of the 3Ј end using a probe directed against the neo cassette after digestion with BstEII. Two ES clones were selected for blastocyst (C57BL/6) injection, which produced a large number of chimeric mice. These were bred with C57BL/6 mice to generate heterozygous mice carrying one floxed allele (Vglut2 f/ϩ ), which were then intercrossed to produce homozygous mice (Vglut2 f/f ). Vglut2 f/f mice were crossed to PGK-Cre mice (Lallemand et al., 1998) to generate null mutant mice (Vglut2 f/f;PCre ). The neo cassette was removed by crossing Vglut2 f/ϩ mice to Deleter-FlpE mice (Rodriguez et al., 2000). As controls, littermates with at least one wild-type (wt) allele were used. All experiments involving live animals were performed in accordance with ethical guidelines defined by the French Agricultural Ministry and the European Union Council Directive for the Care and Use of Laboratory Animals (number 2889) and have been approved by the appropriate local Swedish ethical committee (C156/4).
Genotyping by PCR. For genotyping of mice, 1-2 mm of tail was incubated in 75 l of buffer consisting of 25 mM NaOH and 200 M EDTA at 96°C for 45 min and placed on ice for a few minutes. Tris-HCl (40 mM), pH 8.0, was then added to a final volume of 150 l, after which the preparations were subjected to PCR analysis. Oligos a and b (see Fig. 1 A) detects a 150 bp fragment in the wt allele and a 184 bp band in the floxed allele. Oligos a and c (see Fig. 1 A) detects a 327 bp fragment in the null allele and no band in the wt and floxed allele, respectively. A separate Cre-specific PCR was run to verify the presence of the PGK-Cre allele. In addition, Neo and Deleter-FlpE specific PCRs were used to genotype Neo-excised mice.
Preparation of tissue and histology. Animals were mated overnight, and females were checked for vaginal plug the next morning. In the morning of plug, embryos were staged as embryonic day 0.5 (E0.5). Embryos were collected at E12.5, E15.5, E16.5, and E18.5. In the morning of E19, pups were born. Newborn pups are denoted as postnatal day 0 (P0). For dissection of embryos, pregnant females were killed by cervical dislocation and embryos were removed. E12.5 embryos were kept intact, whereas older embryos and newborn pups were decapitated and the skin was removed. Tails were collected for genotyping. The animals were fixed in zinc-formalin (Richard-Allan Scientific, Kalamazoo, MI) for 18 -24 h at room temperature before dehydration and paraffin infusion (Tissue Tek vacuum infiltration processor; Miles Scientific, Elkhart, IN). Sections (7 m thick) were cut on Microm microtome, attached on Superfrost slides (Menzel-Gläser, Braunschweig, Germany) and stored at 4°C until usage. Slides were deparaffinized in X-tra solve (Medite Histotechnic, Burgdorf, Germany) and rehydrated in ethanol/water before subsequent treatments. For hematoxylin (Richard-Allan Scientific) and eosin (Sigma, St. Louis, MO) histological staining, slides were incubated in hematoxylin 20 -30 s, rinsed in water, 0.1% ammonium hydroxide, and water again before incubation in eosin 20 -30 s followed by water. Slides were then dehydrated in ethanol/water, infused in X-tra solve, and mounted.
Imaging. Fluorescent and bright-field stainings were analyzed using an Olympus (Tokyo, Japan) microscope with an Optigrid system (Thales, Fairport, NY) and Volocity software (Improvision, Lexington, MA); fluorescent stainings were also analyzed by confocal microscopy using Zeiss (Oberkochen, Germany) LSM 510 META system and bright-field stainings on a MZ16F dissection microscope with DFC300FX camera and FireCam software (Leica, Nussloch, Germany). Captured images were auto-leveled using Adobe Photoshop software.
Electron microscopy. For electron microscopy (EM), E18.5 embryos were dissected and brains were rapidly placed in 2.5% glutaraldehyde (Sigma). After 1 h of fixation, brains were cut in 1-mm-thick coronal sections using a brain matrix, and from these slices, round pieces of 1 mm in diameter were punched out from the following regions: olfactory bulb, cortex/hippocampus border, thalamus, striatum, medulla, and pons. The punches (from two control brains and two null mutant brains) were mounted in Epon plastic, sectioned at 1 m, and stained in toluidine blue, after which the punches were appropriately cut in one-third to allow ultrathin sectioning into 50 nm slices (three slices per region) for EM analysis. The slices were contrast stained with 5% uranyl acetate and lead citrate for analysis on a Hitachi H-700 EM (Hitachi Scientific Instruments, Nissei Sangyo, Tokyo, Japan). Synaptic structures were analyzed morphologically in all brain areas. Synapses with both presynaptic and postsynaptic membrane thickening were defined as symmetric, whereas those with only a postsynaptic thickening were defined as asymmetric. In the control brains, symmetric vesicles were either round or displayed a flattened appearance, whereas asymmetric vesicles were always round. In the null mutant, symmetric vesicles looked as in the control, whereas the asymmetric vesicles frequently displayed an irregular shape (see Results). The medulla was selected for synaptic vesicle quantification. Three consecutive medullary slices were analyzed per animal, and values given are averages per genotype Ϯ SEM. The number of vesicles per synapse was subdivided into groups of 1, 2-8, 9 -18, 19 -25, and Ͼ25 vesicles per synapse. Photographs were taken at 40,000ϫ magnification.
In vivo measurement of the ventilation. Pregnant mice were killed by cervical dislocation at E18.5, and fetuses surgically delivered from uterine horns were transferred into a ventilation recording chamber. Ventilation was monitored using a barometric method (Bartlett and Tenney, 1970). The plethysmograph chamber (15 ml) equipped with a temperature sensor, was connected through a slow leak to a reference chamber of the same volume. The pressure difference between the two chambers was measured with a differential pressure transducer (DP-103-10; Validyne, Northridge, CA) connected to sine wave demodulator (CD15; Validyne). The spirogram was stored on a personal computer (PC) (Chatonnet et al., 2002). Calibrations were made during each recording session by injecting 5 l of air into the experimental chamber with a Hamilton syringe. Mean values for the ventilatory frequency were calculated from averaging 100 breaths.
Brainstem electrophysiology. Pregnant mice were killed by cervical dislocation on E16.5 to E18.5. Isolated whole hindbrain preparations and transverse slices containing the PBC were prepared according to previously published protocols (Thoby-Brisson et al., 2005). Embryos were excised from the uterus and kept in oxygenated artificial CSF (aCSF) at room temperature until the electrophysiological recording session. aCSF composition was as follows (in mM): 128 NaCl, 8 KCl, 1.5 CaCl 2 , 1 MgSO 4 , 24 NaHCO 3 , 0.5 Na 2 HPO 4 , 30 glucose, pH 7.4. One transverse 450-m-thick slice was prepared from each embryo using a vibratome, and then transferred to a recording chamber continuously superfused with oxygenated aCSF, at 30°C. Hypoglossal nerve root activity in whole hindbrain preparations and local population activity in slices were recorded using glass micropipettes suction electrodes (150 m tip diameter). The micropipettes filled with aCSF were connected through silver wires to a high-gain AC amplifier (7P511; Grass, West Warwick, RI), filtered (bandwidth, 3 Hz to 3 kHz), integrated using an electronic filter (Neurolog System; time constant, 100 ms), recorded on a computer via a digitizing interface (Digidata 1322A; Molecular Devices, Foster City, CA), and analyzed with the pClamp9 software (Molecular Devices). Whole-cell patch-clamp neuronal recordings were performed under visual control using differential interference contrast and infrared video microscopy, an Axoclamp2A amplifier (Molecular Devices), a digitizing interface (Digidata 1322A; Molecular Devices), and the software program pClamp9 (Molecular Devices). Patch electrodes (resistance, 4 -6 M⍀) were pulled from borosilicate glass tubes (GC 150TF; Clark, Pangbourn, UK) and filled with a high chloride internal solution containing the following (in mM): 123 K-gluconic acid, 21 KCl, 0.5 EGTA, 3 MgCl 2 , 10 HEPES, pH 7.2. Ongoing synaptic currents were recorded at a holding potential of Ϫ70 mV. The inward direction of synaptic currents mediated by GABA A and glycine receptors was imposed by our recording conditions to mimic the immature transmembrane chloride gradient enforcing, at the embryonic time of recording, a depolarizing action of GABA (Thoby-Brisson et al., 2005). Drugs were obtained from Sigma, dissolved in aCSF, and bath applied for 10 -15 min at the following final concentrations (in M): 0.1 for substance P (SP), 0.3 for D-Ala2, MePhe4,Gly-ol5-enkephalin (DAMGO), 10 for 6-cyano-7nitroquinoxaline-2,3-dione (CNQX), 10 for bicuculline, 5 for strychnine, and 1000 for NMDA. A pressure pulse (50 ms; 0.5 bar) was applied to a patch pipette filled with AMPA (1 mM) diluted in aCSF, positioned right below the surface of the slice over the PBC area to test the presence of orthodromically evoked excitatory responses in the contralateral PBC area. A concentric bipolar stimulation electrode (Phymep, Paris, France) was positioned at the surface of the slice on the PBC area and at a ventral level on the midline to stimulate commissural fibers connecting the PBCs bilaterally. Values are given as means Ϯ SEM.
Calcium imaging. Whole hindbrain and slices were incubated for 45 min in oxygenated aCSF containing the cell-permeable calcium indicator dye Calcium Green-1AM (10 M; Invitrogen). Whole hindbrain preparations were positioned in the recording chamber with the ventral side up to allow optical accessibility of the facial motor nuclei. After a 30 min recovery period in the recording chamber to wash out the dye, a standard epifluorescent illumination system on an E-600-FN upright microscope (Nikon, Tokyo, Japan) equipped with a fluorescein filter block was used to excite the dye and capture the emitted light. Fluorescence images were captured with a cooled CCD camera (Coolsnap HQ; Photometrics, Tucson, AZ) using an exposure time of 100 ms in overlapping mode (simultaneous exposure and readout) during periods of 30 -120 s and analyzed using MetaMorph software (Universal Imaging, West Chester, PA). The average intensity in a region of interest was calculated for each frame, and changes in fluorescence were normalized to their initial value by expression as the ratio of changes in fluorescence to initial fluorescence (⌬F/F ).
Spinal cord electrophysiology. The functional output of the locomotor CPG was studied using extracellular electrophysiology on an in vitro preparation of the spinal cord (for review, see Whelan, 2003). The embryos were decapitated just before dissection and mounted ventral side up in a dissection chamber. After filling the chamber with ice-cold oxygenated (95% O 2 plus 5% CO 2 ) dissection buffer (in mM: 128 NaCl, 4.69 KCl, 25 NaHCO 3 , 1.18 KH 2 PO 4 , 3.5 MgSO 4 , 0.25 CaCl 2 , 22 D-glucose), the animals were eviscerated and the spinal cord was exposed by ventral laminectomy. The spinal cord was dissected out to preserve the integrity of the ventral roots, pinned down in a perfusion chamber, and perfused with oxygenated aCSF (in mM: 128 NaCl, 4.69 KCl, 25 NaHCO 3 , 1.18 KH 2 PO 4 , 1.25 MgSO 4 , 2.5 CaCl 2 , 22 D-glucose) at room temperature. Embryos without reflex response were dissected first. The spinal cords of the litter mates were then dissected and put in a chamber filled with aCSF, continuously oxygenated, until recordings were started. Suction electrodes were attached to left and right lumbar (L)2 and L5 ventral roots and the spinal cord was equilibrated at least 30 min before application of neurotransmitter chemicals. A combination of NMDA (0 -10 M), serotonin (5-HT) (3-15 M), and dopamine (0 -50 M) were added to the perfusing aCSF to induce stable locomotor-like output. All chemicals were obtained from Sigma. The signals were amplified 10,000 times and bandpass filtered 10 Hz to 5 kHz (model 1700; A-M Systems, Carlsborg, WA). The signals were A/D converted (Digidata 1322A; Molecular Devices) before being recorded on a PC (Axoscope 9.0.2.05) for later offline analysis. Data analysis of the recorded signals was performed as described previously (Kjaerulff and Kiehn, 1996); the onset of each burst was determined directly from recordings without rectifying the signals. The intrasegmental left/right alternations were analyzed in either L2 or L5 roots, and the intersegmental L2/L5 alternations were analyzed on either the right or left side in each animal. Only traces showing a stable  Vglut2 is expressed early in the embryo and is essential for normal development. A-C, Vglut2 f/f;PCre embryos (*) appear arrested in development and can be distinguished from control littermates already at E16.5 by their hunched posture (A). At E18.5 (B), their smaller size (weights: control, 1.31 Ϯ 0.03 g, n ϭ 49; Vglut2 f/f;PCre , 1.20 Ϯ 0.03 g, n ϭ 20; p ϭ 0.0269) is easily distinguished at P0. C, The pups display cyanosis and die directly after birth. D, E, Vglut2 in situ hybridization of E12.5 embryos reveals strong mRNA expression in distinct brain regions (D), such as the ventral and dorsal diencephalon and ventral mesencephalon (arrows), brainstem, spinal cord, and dorsal root ganglia (close-up in E, marked with arrows) and also along the telencephalic and fourth ventricles and the aqueduct. F, G, Vglut2 continues to be strongly expressed in the brainstem (F ) and the thalamus during late development (G). H, I, Hematoxylin/eosin staining of control (H ) and Vglut2 f/f;PCre embryos (I ) coronal sections at E18.5 shows grossly normal brain histology of the null mutant compared with control, except a few animals (n ϭ 4/9, E18.5 and P0) that displayed enlarged appearance of ventricles (e.g., third ventricle) (inset in I ). J, K, VGLUT2-positive nerve terminals (red) are readily detected by immunofluorescence (nuclear costaining with DAPI in blue) in control internal organs at E18.5, as shown here in transverse sections of the muscular wall of the esophagus (J ), whereas these nerve terminals are undetectable in the null mutants ( rhythmic activity were analyzed for coordination coupling. Vector points were derived from 15 random locomotor cycles. Each point represents the endpoint of a vector, the direction given by the mean phase value, and the length given by the focus of phase values around the mean. A phase value of 0.5 reflects alternating activity, and 0.0 reflects synchronous activity.

Vglut2 f/f;PCre mutants die at birth
We generated a Vglut2 deficient allele by replacing part of the Vglut2 gene with a targeting construct by homologous recombination in ES cells (Fig. 1 A). Successful targeting produced a modified allele with a loxP site preceding exons 4, 5, and 6, and a neomycine selection cassette flanked by Frt sites, followed by a second loxP site. We confirmed the correct targeting event in the ES cells and in the animals by Southern blotting (data not shown) and a PCR strategy ( Fig. 1 B). Vglut2 f/f mice carrying the entire construct were viable, fertile, and subsequently bred to PGK-Cre mice (Lallemand et al., 1998) to delete the targeted region including the neo cassette. The region flanked by the loxP sites contains three exons coding for a substantial part of the protein including three and a half transmembrane domains and two large extracellular loops (Fig. 1C). A transcript containing the 3Ј end of the mRNA was detectable by in situ hybridization, albeit at very low levels compared with control tissue (Fig. 1 D, E). We failed to detect VGLUT2 protein by immunohistochemistry (Fig. 1C,F ), whereas we found normal levels of the closely related VGLUT1 protein in the brain ( Fig. 1 H, I ). This suggests that removal of the targeted region severely affects transcription efficiency and that there is no detectable VGLUT2 protein product in null mutant mice.
None of the mice homozygous for the null allele (Vglut2 f/f;PCre ) survived at birth (n ϭ 20). Litters from embryonic stages E12.5 to E18.5 all showed a normal Mendelian ratio of genotypes, indicating that lethality occurred at birth. Observation of delivering females revealed that Vglut2 f/f;PCre pups were born completely immobile with a beating heart that rapidly ceased as they became cyanotic (Fig. 2C). Of 20 null mutant pups dissected out at E18.5, none had a functional toe-pinch reflex, whereas 44 of 49 tested heterozygote or wt siblings responded with limb movements. Vglut2 mRNA is expressed in developing dorsal root ganglia (Fig. 2 D, E) containing cell bodies of sensory neurons, and dysfunction of these sensory neurons provides one explanation for the absence of a reflex response. Furthermore, Vglut2 f/f;PCre embryos were hunched, significantly smaller (Fig. 2 A, B), and weighed 1.20 Ϯ 0.03 g, which was 8% less than littermate controls (control, n ϭ 49; Vglut2 f/ f;PCre mutants, n ϭ 20; p ϭ 0.0269). Because of the early expression of Vglut2 both in the nervous system (Fig. 2 D-G), and in peripheral organs (Fig. 2 J, K ), we speculated that the early dysfunction of glutamate signaling mediated by VGLUT2 could potentially lead to developmental defects. We therefore analyzed whether such defects were present in E18.5 embryos and P0 pups.
Comparing Vglut2 f/f;PCre E18.5 embryos with littermate controls, the gross morphology of the brain appeared normal (Fig.  2 H, I ). However, in four of nine E18.5 and P0 Vglut2 f/f;PCre mice, the third ventricle appeared enlarged and reminded us more of an embryo at stage E16.5 (Fig. 2 I, inset). We then analyzed embryos at E12.5 (control, n ϭ 2; Vglut2 f/f;PCre , n ϭ 2), E15.5 (control, n ϭ 2; Vglut2 f/f;PCre , n ϭ 3), and E16.5 (control, n ϭ 1; Vglut2 f/f;PCre , n ϭ 1), but we did not detect any defects with regard to brain morphology or proliferation (data not shown). In contrast, in Vglut2 f/f;PCre neonates, lung alveoli were considerably reduced in size, indicating that the lungs had never been inflated (Fig.  2 L-O). and right (r) facial motor nuclei (white outlines 7m,l and 7m,r in C). E, Corresponding fluorescence changes (⌬F/F ) after loading whole hindbrain isolated preparation with Calcium Green-1AM are shown superposed to the hypoglossal nerve (12n) integrated neurogram (bottom trace). Note rhythmic fluorescence changes of facial motoneurons phased to the hypoglossal nerve root electrical activity. F, Photomicrograph of a whole-mount hindbrain preparation double immunostained for NK1R (red) and the motorneuron marker Islet1,2 (green) demonstrates that the position of the right facial motor nucleus matches the location of the fluorescence changes presented in D. G, Bath application of 10 M CNQX to a wt whole-mount hindbrain preparation suppressed both rhythmic calcium changes over the facial motor nuclei (G, H ) and accompanying electrical activity of the 12n (H, bottom trace). I, The Vglut2 f/f;PCre mutants showed normal hindbrain morphology (I ) but never displayed calcium transients over the facial motor nuclei (J ) nor electrical hypoglossal motor activity (K ).

Complete absence of rhythmic activity in the PBC of
Vglut2 f/f;PCre mutants The cyanosis, noninflated lung morphology, and immediate postnatal death prompted a functional investigation of ventilation and respiratory generating circuits in the brainstem of the embryo. In vivo measurement of the ventilation of wt (n ϭ 3), heterozygous (n ϭ 6), and Vglut2 f/f;PCre homozygous mutants (n ϭ 4) were performed on fetuses surgically delivered at E18.5 (from two mice). Wild-type and heterozygous siblings (nine of nine) immediately initiated breathing and showed a stable breathing behavior ϳ10 min after delivery at a frequency of 139 Ϯ 12 breath ⅐ min Ϫ1 (Fig. 3A). In contrast, Vglut2 f/f;PCre mutant littermates failed to show any respiratory behavior in a 3-5 min postdelivery period when they were placed in a plethysmographic chamber (Fig. 3B).
Breathing movements in rodent embryos arise before birth and respiratory-related rhythmic activity in the mouse is first detected at E15.5 in en bloc brainstem and slice preparations (Suzue and Shinoda, 1999;Kobayashi et al., 2001;Thoby-Brisson et al., 2005). Suspecting that altered lung morphology could have resulted from impaired central respiratory command, we first used E16.5 brainstem en bloc preparations to check the presence of rhythmic activity on cranial motor nerve outputs in wt (n ϭ 2), heterozygous (n ϭ 6), and Vglut2 f/f;PCre mutants (n ϭ 4). Electrophysiological recording from the hypoglossal nerve root (12n) while optically recording changes in fluorescence occurring over the facial motor nuclei (7m) demonstrated bilateral rhythmic changes in fluorescence over the 7m synchronized with rhythmic bursts of activity of the 12n. Rhythmic activities emerging from wt and heterozygous mutant preparations were similar, with a combined average frequency for the rhythm of 9.5 Ϯ 1.3 bursts ⅐ min Ϫ1 (Fig.  3C-E). Application of the AMPA/kainate receptor antagonist CNQX (10 M; n ϭ 4) suppressed both 12n electrical and 7m optically detected activities (Fig. 3F-H). In control conditions, recordings performed on Vglut2 f/f;PCre mutant preparations revealed the absence of both spontaneous electrical activity of the 12n and accompanying changes in fluorescence over the facial motor nuclei areas (Fig. 3I-K). Hence, the Vglut2 f/ f;PCre mutants showed a complete impairment of respiratory-related rhythmic motor output.
We next examined premotor rhythm generation at the level of the PBC on another set of embryos. We recorded the neuronal activity within the PBC from brainstem transverse slices. The spontaneous activity of the PBC cellular population was recorded from 11 slices from either wt or heterozygous embryos through calcium imaging and extracellular electrophysiological recordings at E16.5, ϳ24 h after it becomes first active (Thoby-Brisson et al., 2005). In all cases, a spontaneous rhythmic activity could be recorded and visualized through associated changes in fluorescence in the PBC area of the slice (Fig.  4 A-C) (Thoby-Brisson et al., 2005). Bath application of the -opioid agonist DAMGO (0.3 M) first slowed and then reversibly fully suppressed the PBC rhythm (data not shown), whereas SP (0.1 M) increased the rhythm frequency from 8.0 Ϯ 1.4 burst ⅐ min Ϫ1 (n ϭ 5) to 14.5 Ϯ 2.5 burst ⅐ min Ϫ1 (Fig. 4 D). Furthermore, blockade of AMPA/kainate receptors with bath application of CNQX (10 M) abolished the rhythm (n ϭ 5) [see below and also Thoby-Brisson et al. (2005)]. In contrast, no activity could be recorded from corresponding seven of seven slices (E15.5, n ϭ 2; E16.5, n ϭ 2; E17.5, n ϭ 1; E18.5, n ϭ 2) from Vglut2 f/f;PCre mice (Fig. 4 E-G). Prolonged (15 min) applications of SP (0.1 M; n ϭ 3) (Fig. 4 H) or NMDA (5 M; n ϭ 2) (data not shown) failed to induce rhythmic activity.

The PBCs in Vglut2 mutants have contralateral projections but no longer communicate over the midline
Commissural synaptic relationships across ventrolaterally located PBC areas in E16.5 transverse slices from wt and Vglut2 f/f;PCre mutants were investigated by electrical and chemical stimulation. In wt slices, unilateral electrical stimulation of the PBC (Fig.  5A-C) resulted in evoked bilateral bursts mixed with spontaneously occurring rhythmic bursts. The same stimulus in Vglut2 f/f;PCre mutant slices instead evoked a calcium response restricted ispilaterally in the vicinity of the stimulation site (Fig. 5J-L). When the stimulating electrode was placed in a medial position to stimulate PBC axonal connections crossing the midline ( Koshiya and Smith, 1999), it resulted in bilateral evoked calcium responses of the PBCs in wt slices (Fig. 5D-F ). In Vglut2 f/f;PCre mutant slices, midline stimulation could also elicit a calcium response in the expected location of the PBC (Fig. 5M-O). However, the amplitude of this response was reduced ϳ10-fold and resembled the response evoked in wt slices after blockade of AMPA/kainate receptors with CNQX ( Fig. 5G-I ). Finally, orthodromic recruitment of the contralateral PBC through commissural interneurons was assessed by chemical stimulation. Local pressure application of a glutamate agonist or GABA onto the PBC on one side is known to induce bilateral calcium variation in wt E15.5 PBC slices (Thoby-Brisson et al., 2005). Here, pressure application of the glutamate agonist AMPA (1 mM; 50 ms; 0.5 bar) on the PBC on one side resulted in bilateral PBC calcium responses in wt slices (n ϭ 3) (data not shown). However, the same AMPA application in Vglut2 f/f;PCre slices (n ϭ 2) ( Fig. 5P-R) elicited a calcium response limited to the injection site, demonstrating that local AMPA responsive cells were unable to synaptically propagate their excitation contralaterally. Hence, in the mutant, the small-amplitude PBC-evoked responses to electrical stimulation from the midline probably arose from antidromic activation of preserved commissural PBC neurons. These experiments indicate a conserved albeit nonfunctional PBC-like commissural connectivity in the Vglut2 f/f;PCre mutant.

The brainstem nuclei involved in respiration appear morphologically normal
We next characterized the structural and cellular organization of several nuclei located in the brainstem that are known to be part of the central respiratory network. Embryo sections at E16.5 and E18.5 were analyzed by in situ hybridization and immunohistochemistry. At E18.5, the locations, distributions, and the size of cellular populations of upper airway motorneurons in the hypoglossal nucleus (XII), the vagal nucleus (X), and nucleus ambiguous (NA) as well as phrenic motoneurons (MN) in the cervical spinal cord were similar in wt and Vglut2 f/f;PCre embryos as shown by VAChT mRNA expression (Fig.  6C-H ). In Figure 6, D and G, the area of the PBC is boxed to show its location; however, this is a VAChTnegative nucleus. Phox2b mRNA expression (Pattyn et al., 2000b) was used to localize the locus ceruleus (LC), the A1/C1 nucleus, and the nucleus of the solitary tract (NTS) in sections at E18.5 (Fig. 6 A shows control sections; Vglut2 f/f;PCre sections are not shown). Adjacent sections were then processed for immunoreactivity of VGLUT2 and nuclei-specific markers. VGLUT2positive nerve terminals are seen closely surrounding TH-expressing noradrenergic neurons in the LC (Fig. 6 I), Phox2b/THexpressing neurons in the noradrenergic/adrenergic A1/C1, and Phox2b-expressing neurons in the NTS (Fig. 6O). In sections covering the PBC area, VGLUT2-positive terminals were detected on NK1R-positive PBC neurons (Guyenet et al., 2002;Liu et al., 2003;Stornetta et al., 2003a,b) (Fig. 6 R). We then examined how loss of Vglut2 expression affects these nuclei. The neurons of the LC had a normal distribution and number of cells in Vglut2 f/ f;PCre E18.5 embryos (Fig. 6 J, K ), which was true also for the A1/C1 nucleus and the NTS (Fig.  6 M, N, P, Q). Medial in the NTS, the THexpressing noradrenergic/adrenergic A2/C2 nucleus could be detected in both control and Vglut2 f/f;PCre brains ( Fig.  6 P, Q) (data not shown). Finally, the organization of NK1R/islet1,2-immunopositive visceral motorneurons in the nucleus ambiguous and NK1R-positive PBC neurons appeared similar in control and Vglut2 f/f;PCre embryos at both E16.5 (Fig. 6S,T ) and E18.5 (data not shown). Together, these data indicate that components of the respiratory network in Vglut2 f/ f;PCre embryos are present and have normal locations and size, suggesting little impact on brainstem structures and expression of genetic markers.

Fast, excitatory synaptic events are abolished in Vglut2 f/f;PCre mutants
We next analyzed how lack of VGLUT2 affected the activity of individual neurons of the PBC area. The ongoing synaptic activity present in neurons of the PBC area of wt and Vglut2 f/f;PCre mutants was examined on transverse slices. Whole-cell voltage-clamp experiments, using a high chloride concentration (26 mM) intrapipette solution, indicated that the spontaneous synaptic activity in wt preparation was composed of fast and slow decaying inward synaptic currents (Fig. 7A-D) (15 cells from two E18.5 slices). Slow but not fast synaptic currents were abolished in the presence of bicuculline and strychnine (Fig. 7D), indicating that they were likely mediated by GABA A and glycinergic receptors. In all cases, fast synaptic events resistant to bicuculline and strychnine (Fig. 7D) were abolished by additional application of CNQX (data not shown). Hence, the fast synaptic events were mediated by AMPA/kainate receptors. Interestingly, the spontaneous synaptic activity recorded in neurons of the PBC area in Vglut2 f/f;PCre slices (12 cells from two slices) revealed the selective absence of fast synaptic events (Fig. 7E-G). Additional application of bicuculline and strychnine suppressed all remaining synaptic activity in Vglut2 f/f;PCre mutant preparations (Fig. 7H ). Hence, the absence of population synchronous rhythmic activity in the expected location of the PBC in Vglut2 f/f;PCre mutants is associated to the lack of AMPA/kainate receptor-mediated synaptic processing.
Neurons with intrinsic bursting properties are preserved in the PBC area in Vglut2 f/f;PCre mutants Rhythm generation in the PBC may also rely on a small proportion (5-25%) of cells capable of intrinsic bursting (Del Negro et  Pena et al., 2004;Pagliardini et al., 2005;Thoby-Brisson et al., 2005). Such neurons remain active when glutamatergic synaptic transmission is blocked by CNQX (Smith et al., 1991;Koshiya and Smith, 1999;Thoby-Brisson and Ramirez, 2001;Johnson et al., 2002;Thoby-Brisson et al., 2005). We recorded the spontaneous calcium changes occurring in individual neurons of the PBC area in Vglut2 f/f;PCre mutant transverse slices (Fig. 8 A, B). Continuous 1 min samplings of the activities from 46 neurons (from two E16.5 slices) indicated that the majority of neurons (34) presented spontaneous transient calcium changes and that the remainder of neurons were silent. The active neurons showed a large cell-to-cell variability in both the frequencies (range, 1-15 calcium events per minute) and amplitudes (⌬F/F, 1-5%) of calcium events. In contrast to what is observed in wt slices (Thoby-Brisson et al., 2005), we failed to detect occurrence of phased calcium events between different cells. However, among the active cellular population, three cells (ϳ5%) displayed spontaneous periodic calcium changes (Fig. 8C). When substance P (0.1 M) was bath applied, individual cells tended to show more calcium events, although this did not result in emergence of synchronous calcium events across the cellular population. Among cells with periodic bursting activity, substance P application resulted in an increased burst frequency in two of three cases (Fig. 8 D, cell 8), whereas the remaining cells lost all calcium fluctuations. Interestingly, application of substance P also transformed the spontaneous activity in one cell from nonperiodic to periodic (Fig. 8, compare the activity of cell 3 in C and D). Substance P, probably acting on NK1 receptors present on neurons in the PBC area of Vglut2 f/f;PCre mutants ( Fig. 6S,T ), depolarized individual neurons in appropriate voltage domains in which the periodic busting pattern can either emerge, increase in frequency, or can no longer be maintained. In support of this, we detected voltage-dependent bursting properties in 1 of 12 neurons by whole-cell recordings (Fig. 8 E, F ). Collectively, these data strongly suggest that the Vglut2 f/f;PCre mutant PBC area contain a small set of independently active neurons capable of intrinsic bursting.

Synaptic vesicles are redistributed and malformed in asymmetric synapses
We analyzed several brain regions by EM to determine the influence of VGLUT2 deficiency on synaptic morphology. In the pons, cortex, and hippocampus of control embryos, the asymmetric synapses contained round vesicles, whereas in the Vglut2 f/f;PCre embryos, the vesicles had an elongated, drop-like, or irregular shape (data not shown). A closer analysis of the distribution and morphology of synaptic vesicles was per-formed at the level of the medulla in the hypoglossal motor region showing strong immunoreactivity for VGLUT2 (Fig.  9A). Also, in this region, we found oddly shaped vesicles in asymmetric synapses from Vglut2 f/f;PCre embryos ( Fig. 9B-D) compared with control littermates (Fig. 9E). Quantification of the number of vesicles in each synapse revealed that symmetric synapses had a similar distribution of vesicles in wt and Vglut2 f/f;PCre embryos (Fig. 9 F, G,J ). In contrast, asymmetric synapses from Vglut2 f/f;PCre embryos had a higher number of synapses containing only one vesicle than had littermate controls ( Fig. 9 H, I,K ). The average number of vesicles in symmetric synapses was similar between the genotypes with 5.5 Ϯ 0.6 in controls (n ϭ 167 synapses) and 6.4 Ϯ 0.7 (n ϭ 172) in Vglut2 f/f;PCre embryos. In contrast, the average number of vesicles in asymmetric synapses in Vglut2 f/f;PCre embryos (3.7 Ϯ 0.4; n ϭ 108) was significantly smaller (t test, p ϭ 0.003) compared with control embryos (8.3 Ϯ 1.1; n ϭ 99). Because asymmetric synapses are generally excitatory (Kandel et al., 1991), our data demonstrate that, in excitatory terminals of Vglut2 f/f;PCre embryos, synaptic vesicles are often malformed and are reduced in number with overrepresented terminals comprising only one vesicle. In the Vglut2 f/f;Pcre mutant neurons (E-H ), almost all synaptic events were slow (E, F ); fast synaptic events were found to be extremely scarce (Ͻ3%; see the histogram in G) resulting in a modal distribution of the decay time constants of all synaptic events (n ϭ 460) in the slow event range (white triangle above histogram). Furthermore, blocking of GABA A and glycine receptors suppressed all slow events (H ). All current measurements were performed at a holding potential of Ϫ70 mV.

Vglut2 deficiency spares the function of the locomotor CPG in vitro
To evaluate the influence of Vglut2 deficiency in other CPG neurocircuits, we investigated functionality of the locomotor CPG for which excitatory neurons have been shown to be necessary components (Buchanan and Grillner, 1987;Grillner, 1997Grillner, , 2003. Vglut2 is expressed already at stage E12.5 in the spinal cord (Fig. 10 A). After birth, Vglut2 but not Vglut1, is abundantly expressed (Fig. 10 B, C) in the ventral aspect of the spinal cord, which is known to be critical for locomotor rhythm generation (Kjaerulff and Kiehn, 1996). We therefore expected to detect major deficiencies in in vitro measurements of fictive locomotion. To induce fictive locomotion in dissected spinal cords from E18.5 fetuses, we added a combination of NMDA, serotonin, and dopamine (see also Materials and Methods). Surprisingly, the Vglut2 f/f;PCre spinal cords showed in all seven preparations tested, a pattern of rhythmic left-right alternation as well as coordination of the lumbar level 2 and 5 activity with a periodicity indistinguishable from that recorded from control (n ϭ 11) preparations (Fig. 10D-G). This is in stark contrast to the situation found in the respiratory CPG, suggesting a dispensable role for VGLUT2 in the locomotor CPG.

Discussion
Here, we present the first functional study of VGLUT2 by conditional deletion in mice and demonstrate that VGLUT2 signaling is an essential component of the developing respiratory rhythm generator. VGLUT2-deficient mice fail to initiate synchronized rhythmic activity in the PBC and do not breathe at birth. Vglut2 f/f;Pcre mutants have fewer and malformed synaptic vesicles at asymmetric but not symmetric synapses in the hindbrain. This results in a selective impairment of AMPA/kainate receptor-mediated synaptic currents in neurons of the PBC, which in turn prevents synchronized PBC population activity. We here suggest that VGLUT2 is necessary for PBC cells to function as a network with synchronized and rhythmic activity.

Vglut2 f/f;Pcre mutants fail to generate a respiratory rhythm
The central rhythmic activity of the PBC beginning on E15.5 in the mouse and E16.5 in the rat is the first sign of respiratory-related activity in developing rodents (Pagliardini et al., 2003;Thoby-Brisson et al., 2005) and is initiated shortly before the start of respiratory-like movements (Ezure, 1990). Spontaneous PBC area rhythmic activities in the Vglut2 f/f;Pcre mutant mouse was not observed at E15.5. Later on, at E16.5, central rhythm generation was missing at a stage when respiratory movements would normally be present. Fetal development of the lung is strongly influenced by mechanical forces arising from fluid inflation and fetal breathing movements (Alcorn et al., 1977;Tschumperlin and Drazen, 2006). The defective lungs found in Vglut2 f/f;Pcre mutants are consistent with absence of breathing movements. Therefore, the absence of ventilation in Vglut2 f/f;Pcre mutants observed in E18.5 fetuses in vivo and the cyanotic appearance of born mutants probably results from a deficient respiratory motor drive that has been dysfunctional throughout prenatal development. It would be interesting to investigate whether Vglut2 f/f;Pcre mutant fetuses also have other impaired primordial movements such as head flexions, hiccup, and startle movements that are set before onset of fetal breathing movements (de Vries et al., 1982;Suzue and Shinoda, 1999). Along the same line, it is presently unknown whether other brainstem networks such as those involved in regulation of swallowing, cardiovascular function, or sleep may also be functionally impaired in Vglut2 f/f;Pcre mutants. We conclude that VGLUT2-mediated signaling is crucial for the functional onset of the PBC and ventilation at birth.

The VGLUT2 deficiency does not alter the anatomy of central respiratory areas
Vglut2 f/f;Pcre mutants were here shown to develop in the absence of gross anatomical defects of the brainstem hosting central respiratory areas. The location and anatomical layout of the NTS, . At a hyperpolarized potential of Ϫ64 mV, this neuron is inactive; at its resting membrane potential (Vrest., Ϫ55 mV), it generates bursts of action potential (one such burst indicated by the arrowhead is shown with a faster time base in F ); and when further depolarized to Ϫ50 mV, the bursts show increased duration and are generated at a higher frequency. A1/C1, A2/C2, LC, dorsal motor nuclei of the vagus nerve, hypoglossal motor nuclei as well as the phrenic motorneuronal column in the cervical spinal cord appeared normal at E18.5 and P0. Phox2b and/or TH expression were unaltered in identified neuronal groups, suggesting that normal morphogenetic development occurs within the hindbrain of Vglut2 f/f;Pcre mutants. In Vglut2 f/f;PCre mutants, the PBC area adjacent to normally developed A1/C1 nuclei hosted candidate PBC neurons characterized by NK1R expression ventrally located with respect to normally located Islet1,2/NK1R-positive NA visceral motorneurons (Gray et al., 1999). Furthermore, PBC commissural fibers as well as GABAergic/glycinergic inhibitory synapses are present in Vglut2 null mutants. Together, these results demonstrate that the anatomical integrity of the hindbrain and of the PBC in particular is preserved in Vglut2 f/f;Pcre mutants.
Breathing defects that result from abnormal development of brainstem respiratory circuits have previously been described in genetically altered mice. In several mutants such as Mash1 (Dauger et al., 2004), Phox2a (Morin et al., 1997;Viemari et al., 2004), Phox2b (Pattyn et al., 2000a), Rnx (Shirasawa et al., 2000), Hoxa1 (del Toro et al., 2001), Krox20 (Jacquin et al., 1996), Ret (Viemari et al., 2005), and MafB (Blanchi et al., 2003), life-threatening apneas or irregularities of the respiratory rhythm have been demonstrated, but so far, none has shown a complete extinction of respiratory activity. These mutants preserve the capacity to generate inspiratory bursts even if their occurrence may be considerably reduced in frequency (Blanchi et al., 2003). Hence, mechanisms underlying the production of synchronized inspiratory bursts appear particularly robust and allow respiratory movements despite abnormal development of respiratory circuits. Interestingly, the Vglut2 f/f;PCre mutant is the only genetic model to date featuring both spared anatomical integrity and complete absence of synchronized activities in the PBC, throughout the prenatal period from E15.5 when these activities normally first appear, to P0 when they begin to control the breathing behavior. Intriguingly, our data point to VGLUT2-mediated activity having little influence on PBC circuit formation; however, this issue will need additional examination.

Impaired vesicular transport of glutamate prevents emergence of synchronized neural activities in the PBC
We have here found that the proportion of neurons in the PBC area showing voltage-dependent bursting pacemaker properties is comparable (ϳ10%) between Vglut2 f/f;PCre mutants and CNQX-treated wt preparations (Thoby-Brisson et al., 2005). It is therefore likely that VGLUT2-mediated neurotransmission is dispensable for activation of the intrinsic membrane currents underlying intrinsic bursting of individual cells. Other synaptic inputs, including immature depolarizing chloride currents, exist in the Vglut2 f/f;PCre mutants to activate bursting. Also, burstgenerating conductances are redundant. At least two have been identified, a persistent Na ϩ current (I NaP ) (Smith et al., 1991;Del Negro et al., 2002;Johnson et al., 2002), and a Ca 2ϩ -activated nonspecific and voltage-insensitive cation current (I CAN ) (Pena et al., 2004;Del Negro et al., 2005). Others probably exist because blockade of both I NaP and I CAN does not irremediably abolish the PBC ability to generate bursting activity (Del Negro et al., 2005). Active pacemaker neurons are unable to transmit and synchronize activation in the absence of VGLUT2. Therefore, because the , whereas the control vesicles show a regular, circular shape (E). The white arrows indicate vesicle borders. F-I, Low magnification of synaptic structures containing presynaptic vesicles shows that symmetrical synapses, defined by thickened presynaptic and postsynaptic membranes, do not differ between control (F ) and Vglut2 f/f;PCre (G), whereas asymmetric synapses, with thickened postsynaptic membrane in both genotypes, contain round vesicles in the control (H ) and irregularly shaped vesicles in the Vglut2 f/f;PCre (I ). J, K, Quantification of the number of vesicles per synapse, subdivided into groups of 1, 2-8, 9 -18, 19 -25, or Ͼ25 vesicles per synapse, in symmetrical (J ) and asymmetrical (K ) synapses in brainstem of two control and two Vglut2 f/f;PCre E18.5 embryos. In symmetrical synapses, the resulting profiles do not differ significantly between controls and mutants; both have similar number of synapses in each subgroup. In contrast, the profiles of asymmetrical synapses differ in that 30 -50% of synapses contain one vesicle only in the Vglut2 f/f;PCre brain, whereas no control synapses belong to this subgroup. Instead, control synapses contain a larger number of vesicles (8% have 19 -25 vesicles and up to 13% have Ͼ25 vesicles), whereas the corresponding number in the mutant for both these subgroups is zero. Scale bars: A, 155 m; C, 50 nm; H, 175 nm.
bursting process persists in neurons of the PBC area of VGLUT2 mutants despite the lack of breathing rhythm, our view on respiratory rhythm generation extends the recently proposed "group pacemaker hypothesis" (Feldman and Del Negro, 2006), which posits that inspirations result from recurrent excitations that preexist and activate membrane pacemaker properties. We propose the "synchronized group pacemaker hypothesis," in which the process to generate inspirations is a VGLUT2-sensitive glutamatergic synchronization of PBC bursts that may be produced by the various (and somewhat VGLUT2-independent) synaptic/intrinsic interactions within the group pacemaker.
Our hypothesis is built on the assumption that most PBC pacemaker neurons are glutamatergic, which still requires additional experimental support. It also remains poorly understood why in vivo, pharmacological blockade of AMPA/kainate receptors does not entirely eliminate breathing in neonatal mice (Borday et al., 1998). A simple explanation would be that pharmacology in vivo is unable to entirely disrupt glutamatergic transmission, whereas the present genetic inactivation does. The disruption of genes encoding glutamate NMDA receptor subunits (Forrest et al., 1994;Li et al., 1994;Sakimura et al., 1995;Ebralidze et al., 1996;Kadotani et al., 1996;Kutsuwada et al., 1996;Funk et al., 1997;Poon et al., 2000;Miyamoto et al., 2002), non-NMDA receptors (Jia et al., 1996(Jia et al., , 2001Pekhletski et al., 1996;Huettner, 2001;Gerlai et al., 2002;Morishima et al., 2005), glutamate-synthesizing enzyme (Masson et al., 2006), or of glutamate fate-determining selector genes (Cheng et al., 2004) failed to result in a complete collapse of the respiratory activity. Hence, genetically impacting the glutamate neuron itself provides a powerful tool to interfere with the activity of neural circuits. Because the genetically targeted allele of Vglut2 used here can be conditionally deleted, future genetic dissection of excitatory components in many neural circuits is now possible.
Many neuronal populations express Vglut2 mRNA in the spinal locomotor CPG area (Kullander et al., 2003;Kiehn, 2006). These include populations of interneurons with ipsilateral or contralateral projections onto motorneurons as well as Hb9/Vglut2 double-positive ipsilaterally projecting interneurons hypothesized to be elements of the locomotor CPG Butt et al., 2005;Wilson et al., 2005). Therefore, preservation of the rhythm generation, left/right alternation, and extensor/flexor coordination in VGLUT2-deficient mice is an unexpected finding. Additional experiments are required to investigate possible compensatory mechanisms including regulatory expression of other VGLUTs (Miyazaki et al., 2003). However, such compensations do not take place in the brainstem respiratory CPG, emphasizing the necessity of VGLUT2-mediated signaling in the respiratory CPG.
In conclusion, we have shown that, when the functional maturation of glutamatergic synchronization in the PBC is missing during development, the resulting anomaly cannot be compensated by later plasticity. In contrast to spinal cord locomotor rhythm generation, the PBC activity and synchronization is dependent on VGLUT2-mediated signaling. Therefore, the present study identifies VGLUT2-mediated signaling as an obligatory component of the developing respiratory rhythm generator.  11) and Vglut2 f/f;PCre (n ϭ 7) mice was not statistically different ( p ϭ 0.0680); the average was calculated from the mean period of 15 random locomotor cycles from each mouse. Error bars indicate SEM.