Medullary serotonergic neurones and adjacent neurones that express neurokinin-1 receptors are both involved in chemoreception in vivo

Abstract

Neurokinin-1 receptor (NK1R)-expressing neurones that are involved in chemoreception at the retrotrapezoid nucleus (Nattie & Li, 2002b) are also prominent at locations that contain medullary serotonergic neurones, which are chemosensitive in vitro. In medullary regions containing both types, we evaluated their role in central chemoreception by specific cell killing. We injected (2×100 nl) (a) substance P–saporin (SP-SAP; 1μm) to kill NK1R-expressing neurones, (b) a novel conjugate of a monoclonal antibody to the serotonin transporter (SERT) and saporin (anti-SERT-SAP; 1μm) to kill serotonergic neurones, or (c) SP-SAP and anti-SERT-SAP together to kill both types. Controls received IgG-SAP injections (1μm). There was no double-labelling of NK1R-immunoreactive (ir) and tryptophan-hydroxylase (TPOH)-ir neurones. Cell (somatic profile) counts showed that NK1R-ir neurones in the SP-SAP group were reduced by 31%; TPOH-ir neurones in the anti-SERT-SAP group by 28%; and NK1R-ir and TPOH-ir neurones, respectively, in the combined lesion group by 55% and 31% (P < 0.001; two-way ANOVA; P < 0.05, Tukey's post hoc test). The treatments had no significant effect on sleep/wake time, body temperature, or oxygen consumption but all three reduced the ventilatory response to 7% inspired CO₂ in wakefulness and sleep by a similar amount. SP-SAP treatment decreased the averaged CO₂ responses (3, 7 and 14 days after lesions) in wakefulness and sleep by 21% and 16%, anti-SERT-SAP decreased the responses by 15% and 18%, and the combined treatment decreased the responses by 12% and 12% (P < 0.001; two-way ANOVA; P < 0.05, Tukey's post hoc test). We conclude that separate populations of serotonergic and adjacent NK1R-expressing neurones in the medulla are both involved in central chemoreception in vivo.

An increase in CO₂ or H⁺ in blood or brain stimulates breathing. This occurs largely by means of central chemoreceptors, which are widely distributed within the brainstem (Loeschcke, 1982; Coates et al. 1993; Forster et al. 1997; Li et al. 1999; Nattie, 1998, 1999, 2000, 2001; Ballantyne & Scheid, 2001; H. Wang et al. 2001; Nattie & Li, 2001, 2002a, b; Richerson et al. 2001; Okada et al. 2002; Ribas-Salgueiro et al. 2003; Feldman et al. 2003). Central chemoreceptor sites have been identified by breathing responses to focal acidic stimulation in vivo. They include; (a) the caudal nucleus tractus solitarius (NTS) (b) the locus coeruleus (LC) (c) the rostral aspect of the ventral respiratory group (d) regions lying just beneath the ventral medullary surface in rostral (the RTN and adjacent parapyramidal and marginal glial regions) and caudal locations, and (e) the medullary serotonergic cell group, which is the subject of this study.

To study the function of single chemoreceptor sites in an unanaesthetized in vivo model we have applied a dual strategy. We examine the effects of (1) focal acidosis on breathing, or (2) focal cell specific lesions on the response to systemically applied CO₂, the CO₂ response. Focal CO₂ stimulation at the retrotrapezoid nucleus (RTN) region stimulated breathing in wakefulness (Li et al. 1999), at the caudal NTS it stimulated breathing in both wakefulness and NREM sleep (Nattie & Li, 2002a), and at the medullary raphé it stimulated breathing only in NREM sleep (Nattie & Li, 2001). The state of arousal affects responses to focal stimulation (Nattie, 1998, 1999, 2000, 2001; Feldman et al. 2003). Non-specific chemoreceptor disruption at the ventral medullary surface and other sites by means of cooling, lesions produced by excitatory amino acid neurotoxins, and inhibition produced by muscimol dialysis decreased the response to systemic hypercapnia (Loeschcke, 1982; Berger & Cooney, 1982; Budzinska et al. 1985; Akilesh et al. 1997; Forster et al. 1997; Nattie & Li, 2000; Messier et al. 2002). But many types of neurones were affected in these studies.

As a first attempt to examine in vivo the role of a specific type of neurone in central chemoreception we used (Nattie & Li, 2002b) a conjugate of the ribosomal toxin saporin (SAP) and substance P (SP) the natural ligand for the NK1R (Wiley & Lappi, 1997). SP-SAP has been used successfully to specifically kill NK1R-expressing neurones involved in spinal cord pain responses (Mantyh et al. 1997) and in the generation of the normal respiratory rhythm in rats (Gray et al. 1999, 2001; Wang et al. 2002). We hypothesized that NK1R-expressing neurones are involved in chemoreception based on the similar distributions in the rat brainstem of NK1R immunoreactivity (Nakaya et al. 1994) and of central chemoreceptor sites (Nattie, 2000, 2001). SP-SAP injections bilaterally into one chemoreceptor site, the RTN and adjacent parapyramidal (Ppy) regions, destroyed 40–47% of NK1R-ir neurones and processes and produced both hypoventilation and a decrease in the CO₂ response in both sleep and wakefulness. Unilateral destruction of 47% of NK1R-ir also decreased the CO₂ response in both sleep and wakefulness. We concluded that NK1R-ir neurones or processes in the RTN–Ppy region are involved in central chemoreception and provide a tonic drive to breathe.

In this study we focus on a second chemoreceptor site, the region of the medulla containing serotonergic neurones. This site is of particular interest for a number of reasons. It contains both NK1R-expressing neurones and serotonergic neurones, which are likely to be separate populations (Léger et al. 2002). Serotonergic neurones are chemosensitive in vitro (Wang et al. 1998; Richerson et al. 2001; H. Wang et al. 2001) and are closely apposed to ventral medullary arteries (Bradley et al. 2002), a location that would be an advantage for neurones whose role is to sense blood CO₂ levels. Focal CO₂ stimulation of the rostral aspect of the medullary serotonergic cell group increases breathing in sleep (Nattie & Li, 2001). Large non-specific lesions of much of the region of the medullary serotonergic neurone distribution (Dreshaj et al. 1998) and muscimol dialysis focally in the rostral aspect of this distribution in newborn piglets (Messier et al. 2002) decrease the CO₂ response. Chemical destruction of serotonergic neurones by systemic administration of 5,7-dihydroxytryptamine (DHT) in newborn rats induces, in the adult, hypoventilation (Olson et al. 1979; Mueller et al. 1984) and a decreased CO₂ response (Mueller et al. 1984). Medullary serotonergic neurones may be involved in the pathogenesis of the sudden infant death syndrome (SIDS). There are serotonergic binding abnormalities in many SIDS cases (Panigrahy et al. 2000; Kinney et al. 2001) and there is an association of SIDS with a polymorphism in a promotor for the serotonin transport protein (SERT) gene (Weese-Mayer et al. 2003).

Here we use SP-SAP together with a novel conjugate, anti-SERT-SAP (anti-SERT antibody with saporin), to specifically kill (a) NK1R-ir neurones, (b) serotonergic neurones, or (c) both NK1R-ir and serotonergic neurones together. We find that these are separate neuronal populations and that destruction of a similar amount of either cell type alone or both together alters the breathing response to elevated levels of CO₂ by a similar amount in adult rats in vivo during wakefulness and sleep.

Methods

General

All procedures were within the guidelines of the National Institutes of Health for animal use and care and were approved by the Dartmouth College Institutional Animal Use and Care Committee. A total of 51 rats were used. Thirty male Sprague-Dawley rats (280–350g) were anaesthetized with ketamine (100mgkg⁻¹i.m.) and xylazine (20mgkg⁻¹, i.p.). Hindlimb withdrawal and corneal reflexes were tested to monitor the depth of anaesthesia. Responses resulted in administration of supplemental anaesthesia in the form of a quarter of the initial dose. The skull and a portion of the abdomen were shaved and the skin cleansed with betadine and alcohol. The head was placed into a Kopf stereotaxic holder and two EEG electrodes were screwed into the right side of the skull. One was over frontal cortex (2mm anterior to bregma and 2mm lateral to midline), another over parietal cortex (3.5mm posterior to bregma, 2mm lateral) and an earth lead placed between the two 3mm lateral to midline. For the EMG, a pair of wire electrodes was threaded through the nuchal muscles. A sterile telemetry temperature probe (TA-F20, Data Sciences, St Paul, MN, USA) was placed in the abdominal cavity. After recovery and control data collection, each rat was re-anaesthetized and received two microinjections into the midline of the rostral ventral medulla. The skin on the top of the skull was incised and one hole about 3mm long in the rostral–caudal dimension was drilled into the skull. The coordinates for the placement were 2.2–2.5mm and 3.3–3.6mm caudal from lambda in the midline with the bite bar at 9.5–10.1mm so that the top of the skull was level (Paxinos & Watson, 1982). The microinjections were made by using a 0.5μl Hamilton microsyringe with a 28 gauge needle inserted 10.3–10.6mm below the dorsal surface of the skull. Each microinjection lasted at least 5min and the needle remained in position for another 3–5min before removal. The wound was sutured.

Protocol

The rats were housed in a room with a light, rest period from midnight to noon and a dark, active period from noon to midnight. Food and water were available ad libitum. All the experiments were performed between 9a.m. and 4p.m. After 4–7 days recovery from the EEG/EMG implantation, baseline ventilation V̇_E, tidal volume (V_T), frequency (f), body temperature, and oxygen consumption were measured at least twice in each rat while breathing room air and 7% CO₂ in air during sleep and wakefulness. Each rat then received two injections (100 nl each, 1μm in artificial cerebrospinal fluid) (Nattie & Li, 2002b) of one of the following agents, mouse IgG–saporin (IgG-SAP), substance P–SAP (SP-SAP), anti-SERT-SAP, or SP-SAP and anti-SERT-SAP combined (Advanced Targeting Systems, San Diego, CA, USA). The anti-SERT-SAP is a novel molecule. The antibody to the serotonin transporter was made to amino acids 376–388 of the rat sequence of the serotonin re-uptake transporter. This is a segment of the fourth extracellular loop. Immunization of mice, fusion, screening, isolation of monoclones and purification were as previously described (Harlow & Lane, 1988). Clones were also screened for ability to deliver saporin as described (Weltman et al. 1987; Kohls & Lappi, 2000). A single clone, 4A2.2, was used in these experiments for staining and immunotoxin synthesis. Immunotoxin was synthesized as described in Picklo et al. (1995). The anti-SERT-SAP conjugate had an ED₅₀ of 543 pm against RBL-2H3 cells that express SERT but no effect at 10nm on PC12 cells, which do not (M. D. Kohls, unpublished observations). Ventilatory measurements were made on days 3, 6–8 and 13–15 after the injections. Sleep cycling was measured during 3 h of air breathing prior to and 8–13 days after the lesion.

Data analysis

We determined sleep and wakefulness using EEG and EMG electrode signals, the fast Fourier transform (FFT) of the EEG signal analysed in 3.6s epochs at delta (0.3–5Hz), theta (6–9Hz), and sigma (10–17Hz) frequency bands, and behavioural observations. We used criteria modified from those of Bennington et al. (1994) and Trachsel et al. (1988) as previously described (Nattie & Li, 2001, 2002a, b). In our case, we measured breathing only during quiet wakefulness. In active wakefulness the activity of the rat in the plethysmograph prevents reliable measurement of breathing. We analysed the amount of time the rat spent in NREM and REM sleep and in wakefulness during each experiment. Ventilation was measured using a flow-through whole body plethysmography as described by Pappenheimer (1977) and Jacky (1978). The volume of the plethysmograph was 7.6 l with a 3.5 l top to protect the head pedestal. The plethysmograph was connected by a high resistance leak to a similarly sized reference chamber. The inflow gas for the plethysmograph chamber was humidified and controlled by a flow meter at a minimum of 1.4 l min⁻¹ to prevent rebreathing of exhaled gas. The outflow was matched to the inflow via a flow meter connected to a vacuum system. Approximately 100ml min⁻¹ of outflow gas served O₂ and CO₂ analysers (Applied Electrochemistry). We measured chamber pressure by transducer and calibrated the plethysmograph with multiple 0.3ml injections. We measured chamber temperature by thermometer and rat body temperature by telemetry. For calculation of ventilatory data, we used the DataPac 2000 system to determine the pressure deflections and the respiratory cycle time for each of 100–300 breaths at defined sleep and wake periods breathing air or 7% CO₂. Sighs, sniffing and recording artifacts were edited from analysis. These data were exported to Sigmaplot 4.0 (Jandel Scientific software) with f, V_T and V̇_E per 100 g of body weight calculated for each breath using plethysmograph and body temperatures for that time period. We obtained two to four measurement periods for NREM sleep and wakefulness as a baseline prior to 7% CO₂ exposure and one or two periods during the 30min of CO₂ exposure. In our four experimental groups, we compared V̇_E, V_T and f breathing air or 7% CO₂ in NREM sleep or wakefulness on days 3, 6–8, and 13–15 to the baseline data for that treatment in each rat. We defined the ‘CO₂ response’ as the ΔV̇_E breathing 7% CO₂ minus that in air breathing. The change in this CO₂ response was then expressed as a percentage of the baseline value for each rat in each treatment group on each measurement day. Oxygen consumption V̇_O2 was calculated by the Fick Principle using the difference in O₂ content between inspired and expired gas and the flow rate through the plethysmograph and normalized to ml (g body weight)⁻¹ h⁻¹. We monitored CO₂ content of the outflow gas continuously to ensure that no build-up of CO₂ took place within the chamber.

Anatomy

We performed anatomical analysis on the brains from the rats in these four groups but technical problems resulted in unreliable staining for TPOH-ir. We then repeated the injection portion of the protocol in 21 additional rats; five controls with no injection, four with IgG-SAP injections, four with SP-SAP, four with anti-SERT-SAP, and four with SP-SAP and anti-SERT-SAP. After 14 days, the rats were anaesthetized with ketamine and pentobarbital, then transcardially perfused with 200ml saline followed by 300–500ml of chilled 4% paraformaldehyde (4% in 0.1 m phosphate buffer, pH 7.4). The brain was removed and postfixed overnight in 4% paraformaldehyde at 4°C, then cryoprotected for 48 h in 30% sucrose. The brains were sectioned at 30μm thickness on a Reichert-Jung cryostat. All immunohistochemical procedures were performed by using free-floating sections at room temperature. We used 0.1m phosphate buffer (PB) for NK1R-ir and 0.1m Tris buffer for TPOH-ir quenched by 3% H₂O₂ and then blocked with 5% normal goat serum (NGS). The sections were incubated with either a rabbit polyclonal antibody against the NK1R (1 : 10 000, Advanced Targeting Systems, San Diego, CA, USA) or a mouse monoclonal antibody against TPOH (1 : 10 000, Sigma) for 48h at 4°C followed by a biotinylated goat antirabbit or antimouse IgG overnight at 4°C (1 : 500, Vector Laboratories, Burlingame, CA, USA). An avidin–biotin–horseradish peroxidase procedure with diaminobenzidine was used to visualize NK1R and TPOH staining. These sections were used for all cell counts. Double staining was performed in three additional rats. The primary antibodies are the same as above, except with lower dilution (1: 2000). The following secondary antibodies diluted at 1 : 200 were used: goat antimouse IgG conjugated to Cy2 and goat antirabbit IgG conjugated to Cy3 (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA). The two-step sequential methods was used, with first NK1 then TPOH. All the sections were mounted and dehydrated with graded alcohol (25–100% EtOH), cleared with xylene and cover-slipped.

To estimate the total number of TPOH-ir and NK1R-ir neurones in the medullary area of interest we averaged the counts of somatic profiles obtained from one to four 30μm sections within every 360μm segment for 11 such segments. Each rat contributed one value for cell (somatic profiles) counts for each of the 11 360-μm segments. These extended from the most rostral aspect of the medullary raphé caudally for some 3960μm (Fig. 5). In this manner we estimated the total number of TPOH-ir and NK1R-ir neurones in the medulla for each rat.

Figure 5. The distribution of TPOH-ir and NK1R-ir cell counts along the rostral to caudal length of the medullary region of interest.

Top, mean ±s.e.m. NK1R-ir cell counts for each 360μm segment for Control (•; N= 9; 5 No Injection and 4 IgG-SAP) and for SP-SAP (○; N= 4). The line within the axes at bottom shows the rostral to caudal location of the facial nucleus. The lines at the bottom outside the axes show the rostral-to-caudal locations of sections with noticeable tissue disruption from probe placement in the SP-SAP group, i.e. the injection sites. Two-way ANOVA showed a significant effect of treatment (P < 0.001) and of distance (rostral–caudal location) (P < 0.001) but no significant interaction. Tukey's post hoc test showed that the ‘no injection’ and IgG-SAP control data did not differ, and hence we pooled them. SP-SAP and the ‘BOTH’ (not shown) groups differed significant from either the ‘no injection’ or the IgG-SAP control groups (P < 0.05) but not from each other. Bottom, mean ±s.e.m. TPOH-ir cell counts for each 360μm segment for Control (•; N= 9; 5 ‘no injection’ and 4 IgG-SAP) and for anti-SERT-SAP (○; N= 4). The line within the axes at bottom shows the rostral to caudal location of the facial nucleus. The lines at the bottom outside of the axes show the rostral-to-caudal locations of sections with noticeable tissue disruption from probe placement in the anti-SERT-SAP group, i.e. the injection sites. Two-way ANOVA showed a significant effect of treatment (P < 0.001) and of tistance (rostral–caudal location) (P < 0.001) but no significant interaction. Tukey's post hoc test showed that the ‘no injection’ and IgG-SAP control data did not differ, and hence we pooled them. The anti-SERT-SAP and ‘BOTH’ (not shown) groups differed significant from either the ‘no injection’ or the IgG-SAP control groups (P < 0.05) but not from each other.

To evaluate whether our medullary injections produced distant effects we counted TPOH-ir neurones in a portion of the dorsal raphé and NK1R-ir neurones in a specifically defined segment of the rostral ventrolateral medulla. For the former we counted, in five controls and in three rats treated with anti-SERT-SAP, the number of TPOH-ir neurones over 1260μm extending rostrally from the most caudal aspect of the dorsal raphé. For NK1R-ir neurones we applied the approach of W. Wang et al. (2001) in which we counted, in five control and two SP-SAP-treated rats, the number of NK1R-ir neurones in a specific region of the rostral ventrolateral medulla. This region was a rectangular area just below the nucleus ambiguous extending from the caudal pole of the facial nucleus to the rostral aspect of the lateral reticular nucleus (see Fig, 1 in W. Wang et al. 2001).

Figure 1. Anti-SERT antibody staining and anti-SERT-SAP effectiveness in vitro.

The anti-SERT-SAP conjugate specifically targeted serotonergic neurones. A, neurones in the medullary raphé were immunoreactive using the monoclonal anti-SERT antibody. Shown is a confocal image of the medullary raphé in a transverse slice of the rat medulla. B, neuronal somata were not immunoreactive outside of the regions that contain serotonergic neurones. Shown is a confocal image of a transverse slice through the hypoglossal motor nucleus, with punctuate staining consistent with synaptic terminals. There was also intense staining of the ependymal cells lining the 4th ventricle shown at upper right. C, anti-SERT-SAP caused dose-dependent killing of serotonergic neurones. Shown are the numbers of TPOH-ir neurones per coverslip after treatment with different doses of anti-SERT-SAP for either 7 or 10 days (n= 4 coverslips per data point). D, cell killing by anti-SERT-SAP occurred with a delay of 1–4 days. Shown are the numbers of TPOH neurones present after treatment of coverslips with 11–55nm anti-SERT-SAP for different durations of time (n = 6 coverslips per data point).

In vitro experiments

The specificity of the anti-SERT antibody was determined using immunohistochemistry of tissue sections from the rat medulla. Rat brains were perfused using the methods described above. Transverse slices (40μm) were prepared on a vibratome, and then incubated as free floating sections in anti-SERT antibody (1 : 500 in 0.1m phosphate buffer) for 24–48 h at 4°C. They were then visualized on a confocal microscope (Zeiss LSM 410) using an antimouse IgG conjugated to fluorescein.

The anti-SERT-SAP dose that was used for in vivo experiments was based on results from cell culture experiments in which it was found that the anti-SERT-SAP conjugate selectively killed serotonergic neurones. Medullary raphé neurones were cultured using methods previously described (Wang et al. 1998). Briefly, the medullary raphé was microdissected from postnatal day zero to postnatal day 1 (P0–P1) rats. Neurones and glia were dissociated and plated onto glass coverslips and allowed to grow in vitro for at least 17days. Culture wells containing 1ml of culture medium were fed with 10μl of glial-conditioned culture medium containing anti-SERT-SAP to achieve a final concentration of between 2.7nm and 55nm (1 : 200–1 : 4000 dilutions of a 2.4 mg ml⁻¹ stock solution). Control coverslips from sister culture dishes containing a similar density of neurones were fed with an equal volume of glial-conditioned culture medium without anti-SERT-SAP. After 1, 4, 7 or 10 days, cultures were fixed with 4% formalin for 1h. Plates were then processed for immunohistochemistry using the mouse monoclonal antibody against TPOH (1 : 2000) for 24–48h at 4°C. Serotonergic neurones were visualized with the Vectastain Elite ABC kit (Vector Laboratories) using the peroxidase method with diaminobenzidine as the chromogen. After staining, the total number of serotonergic neurones was counted on each plate by an individual unaware of the specific treatment. The effect on non-serotonergic neurones was determined by counting the number of such neurones per high power field (HPF; × 40 objective) within the portion of each plate containing the highest density of neurones (mean values for 4 HPFs per plate.)

Statistics

Counts of somatic cell profiles of TPOH-ir and NK1R-expressing neurones were compared by two-way ANOVA with ‘treatment group’ and ‘distance along the medulla’ as factors. Tukey's post hoc test was used for post hoc comparisons.

In analysis of ventilatory data we used three approaches. First, in unanaesthetized rats there is variability in V̇_E, V_T, and f during both air and 7% CO₂ breathing. Our study design included a control group, which received two injections of IgG-SAP at the same volume and at the same site as in the treatment groups, with measurements made before injection, as in the treatment groups, and at 3, 7 and 14 days following the injections. We performed a one-way repeated-measures ANOVA on these data using Tukey or Dunnett's test for post hoc comparisons. If there is no effect of IgG-SAP we presume that effects observed with the one-way repeated-measures ANOVA following any treatment are the result of that treatment. This approach minimizes effects of interanimal variability and is a sensitive way to detect small treatment effects and show at which times they might occur. Second, we perform a two-way ANOVA on absolute ventilatory values with Treatment and Time as factors using a Tukey test for posthoc comparisons. This analysis directly compares absolute data values among the four groups and allows us to ask if the effects observed in any treatment group, or with time, reflect differences from those observed in the IgG-SAP control group. Third, we define a variable for ‘the CO₂ response’ as the ΔV̇_E comparing 7% CO₂ to air and express for each rat the percentage change in ΔV̇_E at each post injection time. For all four groups we apply to these data a two-way ANOVA with Treatment and Time as factors with Tukey test for posthoc analysis. This approach normalizes each rat's CO₂ response and allows a comparison among all four treatment groups at all three post injection times while minimizing interanimal variability.

For analysis of effects on vigilance state, we applied a one-way ANOVA comparing the effects of treatment on percentage time awake, in NREM sleep or in REM sleep and a two-way ANOVA with Treatment and State as factors. For analysis of the effects of vigilance state on ventilation we performed a two-way ANOVA with State and Time as factors using the Tukey test for post hoc comparisons.

Results

Body weight and normalization

The initial and final mean (s.e.m.) body weights were: for the IgG-SAP group 320 ± 12 and 359 ± 11 g; for the SP-SAP group, 302 ± 8 and 332 ± 12 g; for the anti-SERT-SAP group, 282 ± 24 and 341 ± 12; and for the combined SP-SAP and anti-SERT-SAP group, 340 ± 7 and 376 ± 12. The increases in body weight were not significantly different among the groups. We normalized our ventilatory and oxygen consumption data using body weight values. Over the small increase in body weight observed in our 15 day protocol there was no detectable difference in results if ventilation was normalized by body weight or oxygen consumption with one exception in the IgG-SAP group as discussed below. Neither body temperature nor V̇_O2 expressed as ml O₂100 g⁻¹ h⁻¹ differed significantly among the four groups during this 15-day period.

In vitro experiments

The anti-SERT antibody specifically stained a subset of neurones (Fig. 1A) in regions of the medulla that contain serotonergic neurones. Outside of these regions, there was punctate staining consistent with nerve terminals, but no staining of somata (Fig. 1B). There was also staining of ependymal cells lining the fourth ventricle (Fig. 1B).

Exposure of cultures to anti-SERT-SAP led to killing of serotonergic neurones without an effect on other neurones on the same plates. The destruction of serotonergic neurones occurred in a dose-dependent manner with concentrations greater than 11nm (= 1 : 1000 dilution of 2.4mg ml⁻¹ stock solution) leading to loss of 90–95% of TPOH-ir neurones within 4–7 days after exposure (Fig. 1C). Control coverslips (n= 24) contained a mean (±s.e.m.) of 50 ± 9 neurones. After 7–10 days of exposure to anti-SERT-SAP, the number of TPOH-ir neurones on treated plates was 0.5 ± 0.5 for 22–55nm (n= 4), 4 ± 1 for 11nm (n= 4), and 17 ± 5 for 2.7–5.5nm (n= 4). Cell killing by anti-SERT-SAP was also delayed as expected (Mantyh et al. 1997) with a significant effect present at 4 days after treatment, and maximum killing reached by 7–10 days (Fig. 1D). There was sparing of non-serotonergic neurones on these same plates; with 6.4 ± 3.9 neurones per high power filed (HPF) in control plates (n= 8), 7.4 ± 3.9 neurones per HPF in plates treated with 11nm anti-SERT-SAP (n= 4), and 9.0 ± 6.6 neurones per HPF in plates treated with 22–55nm anti-SERT-SAP (n= 4). The highest concentration of anti-SERT-SAP caused some alteration in the morphology of a subset of non-serotonergic neurones, and caused some glia to detach from the surface of the plates. For in vivo injections we used a 1μm concentration in a volume of 100 nl to obtain a wide killing field.

Anatomy

Medullary serotonergic neurones extend from a level about 800–900μm rostral to the rostral pole of the facial nucleus caudally for about 4–5mm ending at the beginning of the spinal cord. We include all TPOH-ir neurones seen in each cross section (Fig. 2A–D). In this paper we refer to the medullary serotonergic neurones as a single cell population and do not use the anatomical designations of raphé pallidus, obscurus, and magnus. We include extra-raphé TPOH-ir neurones, ones that are outside of classic midline raphé nuclei, e.g. in the nucleus paragigantocellularis lateralis or parapyramidal region.

Figure 2. Examples of TPOH-ir and NK1R-ir neuronal staining.

A, typical plump, oval-shaped TPOH-ir neurones in raphé pallidus. B, narrow, elongated TPOH-ir neurones that extend along the ventral medullary surface. Bar = 100μm and applies to A and B. C, midline raphé TPOH-ir neurones that extend over the pyramids to the parapyramidal region at the level of the caudal facial nucleus. D, midline cluster of TPOH-ir neurones just dorsal to the pyramids at the level of the caudal facial nucleus. E, NK1R-ir neurones extending over the pyramids at the level of the caudal facial nucleus. F, NK1R-ir neurones clustered at the parapyramidal region just caudal to the facial nucleus. Bar = 100μm and applies to C–F.

For this study we examine NK1R-ir neurones present only within our selected boundaries for the medullary serotonergic neurones. NK1R-ir neurones are present at many other sites throughout the medulla (Nakaya et al. 1994; Nattie & Li, 2002b; Gray et al. 2001; W. Wang et al. 2001). Typical NK1R-ir neurones found in the region of interest are shown in Figs 2(E and F). The distribution of NK1R-ir in the medulla also includes neuronal processes, which can form a complex reticular-like pattern as shown in the RTN/Ppy region (Nattie & Li, 2002b). Here we count only NK1R-ir neurones; we make no special effort to measure the extent of labelling present on neuronal processes. The number of NK1R-ir neurones along the rostral to caudal length of the region examined changed in parallel with the number of TPOH-ir neurones (Fig. 5) but at every rostral to caudal level there were many more TPOH-ir neurones.

Given the intermingling of TPOH-ir and NK1R-ir neurones and the hypothesis that both might be involved in central chemoreception, we performed double labelling of the same neurones using goat antirabbit IgG conjugated to Cy3 as the secondary antibody for NK1R and goat antimouse IgG conjugated to Cy2 as the secondary antibody for TPOH. It was important to stain first for NK1R-ir then for TPOH-ir. The reverse order resulted in a non-specific cross-reaction. In three rats we counted 1188, 1211 and 1036 TPOH-ir neurones and 110, 107 and 148 NK1R-ir neurones, respectively, finding a total of three with double labelling. Figure 3(A–C) shows at a level 200μm caudal to the caudal pole of the facial nucleus an example of a single section showing TPOH-ir cells (A) NK1R-ir (B) cells, and both together (C). There was no double labelling. We conclude that TPOH-ir neurones in the medullary serotonergic group do not express NK1Rs, that they are separate cell populations.

Figure 3. TPOH-ir and NK1R-ir neurones are separate populations in the medullary raphé.

A, TPOH-ir neurones 200μm caudal to the facial nucleus just lateral to the midline above the pyramids. B, NK1R-ir neurones in the same section. Bar = 100μm and applies to all three sections. C, TPOH-ir and NK1R-ir neurones shown together in the same section. There are no double-labelled cells. Extensive counts in 3 rats verified this finding (see text).

To evaluate the location and amount of specific cell loss caused by our treatments we used separate rats for the anatomical versus physiological studies. Technical problems with TPOH-ir staining in the original animals forced this choice. The locations of IgG-SAP, SP-SAP, anti-SERT-SAP, and SP-SAP plus anti-SERT-SAP injections in the rats used for the physiological experiments were the same as in the rats used for subsequent anatomy experiments.

The mean ±s.e.m. of the total cell counts of NK1R-ir (top) and TPOH-ir (bottom) neurones for each of the five groups are shown on Fig. 4. The stars refer to significant differences determined by two-way ANOVA analysis of cell counts along the rostral to caudal length of the area of interest using Treatment and Distance as factors. Figure 5 displays these cell counts for each 360μm segment along the rostral to caudal length of the medulla. The top panel shows the NK1R-ir results with the pooled control data (No Injection controls, N= 5, plus IgG-SAP controls, N= 4) and the SP-SAP data (N= 4); the bottom panel shows the TPOH-ir results for the pooled controls and the anti-SERT-SAP group (N= 4). The figure also shows the rostral to caudal location of the facial nucleus as a convenient landmark and the locations of the injections in each of the 4 SP-SAP (top) and anti-SERT-SAP (bottom) injected rats. For NK1R-ir neurone counts, two-way ANOVA using Treatment and Distance as factors showed a significant effect for each (P < 0.001) but no significant interaction. Tukey post hoc analysis showed that there was no difference between the No Injection and IgG-SAP controls thus allowing them to be pooled. The number of NK1R-ir cells was significantly less in the SP-SAP and combined anti-SERT-SAP plus SP-SAP groups compared to either the No Injection or the IgG-SAP control groups (P < 0.05, Tukey multiple comparisons). For the TPOH-ir neurone counts, two-way ANOVA using Treatment and Distance as factors showed a significant effect for each (P < 0.001) but no significant interaction. Tukey post hoc analysis showed that there was no difference between the No Injection and IgG-SAP controls thus allowing them to be pooled. The number of TPOH-ir cells was significantly less in the anti-SERT-SAP and combined anti-SERT-SAP plus SP-SAP groups compared to either the No Injection or the IgG-SAP control groups (P < 0.05, Tukey multiple comparisons).

Figure 4. Estimates of total counts of TPOH-ir and NK1R-ir neurones in the medulla of all four experimental groups plus a group of five uninjected rats.

The top panel shows mean ±s.e.m. NK1R-ir total cell counts, the bottom shows the same for TPOH-ir cells. The groups are: Control, 5 rats that received no injections; IgG, 4 rats injected with IgG-SAP; SERT, 4 rats injected with anti-SERT-SAP; SP, 4 rats injected with SP-SAP; BOTH, 4 rats injected with both anti-SERT-SAP and SP-SAP together. In each rat, cell (somatic profile) counts from 1–3 30-μm sections were averaged to yield a single value used to estimate the number of cells in each of 11 360-μm segments from the most rostral section that contained 5 or more TPOH-ir neurones caudally for 3960μm. These values were averaged to yield a single total cell count estimate for each rat. We applied a two-way ANOVA with treatment and distance as factors using data from all 11 segments, as shown graphically in Fig. 5. Treatments different from control are marked with the asterisks (P < 0.001, two-way ANOVA; P < 0.05, Tukey's post hoc test). These values differed from both the ‘no injection’ and the IgG-SAP controls. Note that we actually counted somatic profiles to represent neurones.

The number of NK1R-ir neurones in a specific region of the rostral ventrolateral medulla, as defined by Wang et al. (2001b), was unaffected by SP-SAP injections. In five controls (N= 3 No Injection rats and N= 2 IgG-SAP injected rats) the mean ±s.e.m. value was 50.6 ± 5.3 while in the two SP-SAP injected rats it was 57.6 ± 7.1. The number of TPOH-ir cells counted in a restricted caudal-most portion of the dorsal raphe was, in four controls (N= 2 No Injection rats plus N= 2 IgG-SAP rats) 142 ± 18 cells and in three anti-SERT-SAP treated rats, 204 ± 53.

Physiology

We measured the amount of time spent in wakefulness, NREM and REM sleep during a 3 h continuous air breathing period from 9 AM to noon (the last three h of the light, resting period in our midnight to noon imposed diurnal cycle) during the final few days post injection. The IgG-SAP group (N= 5) spent 38 ± 3 (mean ±s.e.m.)% of this time awake, 54 ± 2% in NREM sleep, and 7 ± 1% in REM sleep. The anti-SERT-SAP group (N= 6) spent 46 ± 4% awake, 46 ± 4% in NREM sleep, and 7 ± 1% in REM sleep. The SP-SAP group (N= 5) spent 42 ± 3% awake, 54 ± 4% in NREM sleep, and 3 ± 1% in REM. The combined SP-SAP and anti-SERT-SAP group (N= 5) spent 48 ± 5% awake, 49 ± 4% in NREM sleep, and 3 ± 1% in REM sleep. There is a tendency for all three treatment groups to be awake more but this is not statistically significant (one-way ANOVA comparing the effects of treatment on percentage time awake, in NREM sleep or in REM sleep; two-way ANOVA with Treatment and State as factors).

IgG-SAP injection effects on ventilation

In the control group, which received two injections of IgG-SAP, there was no substantial effect on breathing or on the CO₂ response in wakefulness or sleep (Fig. 6). V̇_E in wakefulness was significantly decreased (P < 0.01, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test at 14 days versus baseline) but the magnitude of this effect was small. Examination of V̇_O2 showed that it too was decreased at this 14 day time point such that the V̇_E/V̇_O2 ratio was unchanged. Breathing in air and in 7% CO₂ was less in NREM sleep than in wakefulness (P < 0.02, two-way ANOVA with state and time as factors, P < 0.05 Tukey post hoc for state).

Figure 6. Injection of IgG-SAP into the medullary raphé has no major effect on breathing or on the CO₂ response.

Mean ±s.e.m. values of ventilation breathing air or 7% CO₂ awake (•) or asleep (NREM; ○) in the pretreatment control period and at 3, 7, and 14 days after two injections at day = 0 of IgG-SAP into the ventral medulla. The asterisk marks the 14 day awake room-air-breathing level, which is slightly but significantly different from baseline (P < 0.01, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test; see text).

SP-SAP injection effects on ventilation

The SP-SAP injections had no effect on air breathing in wakefulness or sleep (Fig. 7). During 7% CO₂, V̇_E was significantly decreased in wakefulness (P<0.03, one-way repeated-measures ANOVA; P<0.05, Dunnett's post hoc test of 3, 7 and 14 day values versus baseline). The effect was entirely due to a decrease in V_T (P<0.02, one-way repeated-measures ANOVA; P<0.05, Dunnett's post hoc test of 3, 7 and 14 day values versus baseline). During 7% CO₂ in NREM sleep there was an overall significant decrease in V̇_E (P= 0.05, one-way repeated-measures ANOVA). Posthoc analysis using a Bonferroni modification showed significance at 14 days compared to baseline (P<0.05). Breathing in air and in 7% CO₂ was less in NREM sleep than in wakefulness (P<0.05, two-way ANOVA with state and time as factors, P<0.05 Tukey post hoc test for state).

Figure 7. Specific killing of medullary raphé NK1R-expressing neurones by SP-SAP decreases the CO₂ response in wakefulness and sleep.

Mean ±s.e.m. values of ventilation breathing air or 7% CO₂ awake (•) or asleep (NREM; ○) in the pretreatment control period and at 3, 7, and 14 days after two injections at day = 0 of SP-SAP into the ventral medulla. The asterisks mark the 3, 7 and 14 day awake 7%-CO₂-breathing levels, which are significantly different from baseline (P < 0.03, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test). The 7% CO₂ values in sleep are also decreased but the one-way repeated-measures ANOVA shows a P value of 0.05. No post hoc tests were performed. There was no significant effect on room air breathing in sleep or wakefulness.

Anti-SERT-SAP injection effects on ventilation

The anti-SERT-SAP injections had no effect on air breathing in wakefulness or sleep (Fig. 8). During 7% CO₂, V̇_E was significantly decreased in wakefulness (P < 0.02, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test of 3, 7 and 14 day values versus baseline) and sleep (P < 0.001, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test of 3, 7 and 14 day values versus baseline). In wakefulness, the effect was due to a decrease in V_T and f (both P < 0.01, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test of 7 and 14 day values versus baseline). In NREM sleep the effect was due to a decrease in V_T (P < 0.01, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test of 7 and 14 day values versus baseline). Breathing in air and in 7% CO₂ was less in NREM sleep than in wakefulness (P < 0.01, two-way ANOVA with state and time as factors, P < 0.05 Tukey's post hoc test for state).

Figure 8. Specific killing of medullary serotonergic neurones decreases the CO₂ response in wakefulness and sleep.

Mean ± s.e.m. values of ventilation breathing air or 7% CO₂ awake (•) or asleep (NREM; ○) in the pretreatment control period and at 3, 7, and 14 days after two injections at day = 0 of anti-SERT-SAP into the ventral medulla. The asterisks mark the 3, 7 and 14 day awake and asleep 7%-CO₂-breathing levels, which are significantly different from baseline (P < 0.02 and P < 0.001, respectively, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test). There was no significant effect on room air breathing in sleep or wakefulness.

Combined SP-SAP and anti-SERT-SAP injection effects on ventilation

The combined SP-SAP and anti-SERT-SAP injections had no effect on air breathing in wakefulness (Fig. 9). In sleep, V̇_E was significantly decreased (P<0.001, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test for day 7 and 14 values versus baseline). During 7% CO₂, V̇_E was significantly decreased in wakefulness and sleep (both P < 0.001, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test of 3, 7 and 14 day values versus baseline). In wakefulness, the effect was due to a decrease in V_T (P < 0.05, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test of 3, 7 and 14 day values versus baseline). In NREM sleep the effect was due to a decrease in V_T (P < 0.005, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test of 7 and 14 day values versus baseline). Breathing in air and in 7% CO₂ was less in NREM sleep than in wakefulness (P < 0.002, two-way ANOVA with state and time as factors, P < 0.05 Tukey's post hoc test for state).

Figure 9. Specific killing of both medullary serotonergic and adjacent NK1R-expressing neurones decreases the CO₂ response in wakefulness and sleep and decreases room air breathing in sleep.

Mean ± s.e.m. values of ventilation breathing air or 7% CO₂ awake (•) or asleep (NREM; ○) in the pretreatment control period and at 3, 7, and 14 days after two injections at day = 0 of both SP-SAP and anti-SERT-SAP into the ventral medulla. The asterisks mark the 3, 7 and 14 day awake and asleep 7%-CO₂-breathing levels, which are significantly different from baseline (P < 0.001, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test) and the 7 and 14 day sleep room-air-breathing levels, which are significantly different from baseline (P < 0.003, one-way repeated-measures ANOVA; P < 0.05, Dunnett's post hoc test).

Comparison among groups

We also applied a two-way ANOVA to examine among the four groups the effects of treatment and time on the absolute V̇_E values (a) while breathing room air and (b) while breathing 7% CO₂. This approach does not minimize interanimal variability but allows a comparison of responses among the different treatment groups to those in the IgG-SAP control group. During air breathing in wakefulness there is a significant effect of treatment (P<0.05) and of time (P<0.05) but no post hoc significance. During air breathing in NREM sleep there was no significant treatment or time effect. During 7% CO₂ breathing awake there were significant treatment (P < 0.001) and time effects (P < 0.001) but no significant interaction. Tukey's posthoc test showed the SP-SAP, anti-SERT-SAP, and the combined SP-SAP and anti-SERT-SAP treatment groups being different from the IgG-SAP control (P<0.05) and times 3, 7, and 14 days as different from the preinjection control day (P < 0.05). This analysis supports that of the one-way repeated-measures ANOVA. During 7% CO₂ breathing in NREM sleep there was no significant treatment effect but there was a time effect (P<0.001). Tukey's post hoc test showed times 7 and 14 days as different from the preinjection control day (P < 0.05). There was not a significant interaction. Here the between-group analysis does not support the results of the one-way repeated-measures ANOVA. By inspection, the reason for this is that the IgG-SAP control group had, before any treatment, a larger NREM sleep effect on the V̇_E while breathing 7% CO₂ than did any of the other three groups before treatment.

Normalized effects on the CO₂ response in wakefulness and sleep

To normalize the treatment effects on the CO₂ response and minimize any initial interanimal variability and to allow an intergroup comparison we first defined this response as ΔV̇_E breathing 7% CO₂versus breathing air for each measurement day for each rat. We then compared ΔV̇_E at days 3, 7 and 14 after treatment to that of the pretreatment baseline period, expressing it as the percentage change in the CO₂ response for each rat in each of the four groups. The mean results (+s.e.m.) for days 7 and 14 after injection are shown in Fig. 10. In wakefulness and in sleep there was a significant decrease in the CO₂ response in the SP-SAP, anti-SERT-SAP, and combined SP-SAP and anti-SERT-SAP groups compared to the IgG-SAP control group (P < 0.001, one-way ANOVA; P < 0.05, Tukey post hocversus IgG-SAP control). There was no significant effect of time after lesion nor was there a significant interaction. There was no significant difference between SP-SAP and anti-SERT-SAP or between anti-SERT-SAP and combined SP-SAP and anti-SERT-SAP.

Figure 10. All three treatments decrease the normalized CO₂ response in wakefulness and sleep by a similar amount.

Top, mean ±s.e.m. values in each treatment group and in the IgG-SAP control group for the percentage decrease in the ‘CO₂ response’ in sleep for the 7 and 14 day measurement periods as compared in each rat with the preinjection control value. Each rat contributed one value to these data. In all three treatment groups, the decrease in the CO₂ response at 3 (not shown), 7, and 14 days was significantly less than in the IgG-SAP control group. (P < 0.001 for treatment, two-way ANOVA with treatment and time as factors; P < 0.05, Tukey's post hoc test. There were no significant time or interaction effects.) The values in the three treatment groups were not significantly different from each other. Bottom, same as top but for awake responses. Again, in all three treatment groups, the decrease in the CO₂ response at 3 (not shown), 7 and 14 days was significantly less than in the IgG-SAP control group (P < 0.001 for treatment, two-way ANOVA with treatment and time as factors; P < 0.05, Tukey's post hoc test. There were no significant time or interaction effects).

Discussion

Major findings

We found that medullary serotonergic and adjacent NK1R-expressing neurones are separate cell populations. Specific killing of medullary serotonergic neurones, or of adjacent NK1R-expressing neurones, reduced the ventilatory response to systemic CO₂ in wakefulness and in NREM sleep. Specific killing of both cell types simultaneously did not result in a greater effect.

Analysis

We used three analytic approaches for the ventilatory data; (1) a one-way repeated-measures ANOVA applied to absolute ventilatory values for each group separately, (2) a two-way ANOVA comparing the overall response in terms of absolute ventilatory values among groups, and (3) a two-way ANOVA using values of ΔV̇_E, the ‘CO₂ response’, to compare normalized percentage changes among all groups. In wakefulness, all three indicated that these treatments did not affect air breathing but did decrease the CO₂ response. In NREM sleep, there was a small but significant decrease in V̇_E during air breathing at 7 and 14 days after combined SP-SAP and anti-SERT-SAP treatment, uncovered only by the repeated-measures one-way ANOVA. Further experiments will have to examine whether or not this is a real treatment effect. With respect to the CO₂ response in sleep, both the one-way repeated-measures ANOVA and the two-way group comparison of normalized ΔV̇_E indicated that all three treatments decreased the CO₂ response. The two-way ANOVA of absolute V̇_E values at 7% CO₂ did not show a treatment effect. By inspection, the reason for this is that the IgG-SAP control group had, before any treatment, a larger NREM sleep effect on the V̇_E while breathing 7% CO₂ than did any of the other three groups before treatment. We conclude, based on the two approaches that normalize initial value variability, that the three treatments decreased the CO₂ response in NREM sleep.

Anatomy of medullary serotonergic and associated NK1R-expressing neurones

The serotonergic neurones in the rat medulla are distributed in long columns beginning rostral to the facial nucleus and extending caudally to the very end of the medulla. They are located in the raphé nuclei, i.e. raphé magnus, obscurus and pallidus, and in extra-raphé sites like the parapyramidal region and the nucleus paragigantocellularis lateralis. We evaluate our lesions within the entire medullary serotonergic system.

NK1R-expressing neurones are intermingled among these serotonergic neurones and, in addition, are distributed widely within the brainstem (Nakaya et al. 1994). They are large with prominent and extensive processes (Figs 2E and F and 3B). In some locations, e.g. the RTN–Ppy region, there is an extensive reticular-like network of NK1R-ir processes (Nattie & Li, 2002b); in others, e.g. the pre-Bötzinger region, there is a cluster of NK1R-ir soma (Gray et al. 2001; W. Wang et al. 2001). Here we focus on NK1R-expressing neuronal soma present within the medullary serotonergic neurone region of interest.

In this region of interest, the TPOH-ir neurones do not express NK1 receptors. We found essentially no double labelling of NK1R-ir and TPOH-ir neurones in an extensive examination of the entire medullary serotonergic cell group, which verifies an earlier report by Léger et al. (2002). While NK1 receptors are not present on serotonergic neurones, they are present on many other brainstem neurones. These include: noradrenergic neurones of the locus coeruleus (Chen et al. 2000), GABA-ergic neurones in periaqueductal grey, dorsal raphé (Ma & Bleasdale, 2002), and ventrolateral medulla (W. Wang et al. 2001), glutamatergic neurones in the pre-Bötzinger complex (Guyenet et al. 2002) and dorsal raphe (Liu et al. 2002), and undefined neurones in the NTS (Ljungdahl et al. 1978). A small number of C1 cells also express NK1-receptors (Makeham et al. 2001). NK1R-ir is prominent in central chemoreceptor regions, which include the locus ceruleus, RTN, parapyramidal region, medullary raphé, caudal NTS, caudal ventral medulla, pre-Bötzinger complex, and rostral nucleus ambiguus (Nakaya et al. 1994; Nattie & Li, 2002b).

Cell specific lesions and chemoreception

SP-SAP has been previously used to specifically kill NK1R-expressing neurones or processes that mark a functionally important anatomical site. For example, in the pre-Bötzinger complex, which contains a dense NK1R-ir cell group (Gray et al. 1999, 2001; Wang et al. 2002), selective and substantial bilateral lesions (loss of > 75% NK1R-ir neurones) produced a dramatic respiratory phenotype with irregular breathing, hypoventilation, and a reduced response to increased CO₂.

In this study, the SP-SAP injections focally killed 31% of the NK1R-expressing neurones in the region of the medullary serotonergic neurone distribution. Two lines of evidence support the conclusion that the lesions are focal. First, caudal to the preponderance of injection sites (Fig. 5) the numbers of NK1R-ir neurones were similar in control and treated rats. Second, SP-SAP injections did not decrease the numbers of NK1R-ir neurones in a specifically defined region of the rostral ventrolateral medulla (W. Wang et al. 2001). The 31% loss of NK1R-ir neurones reduced the ventilatory response to systemic CO₂, averaged over the postinjection period, by 21% in wakefulness and 16% in NREM sleep. In prior experiments, slightly greater destruction of NK1R-ir neurones and processes in the RTN/Ppy region (Nattie & Li, 2002b) reduced the CO₂ response in wakefulness and sleep by 17–22% after bilateral lesions and by 28–30% after unilateral lesions. We conclude that approximately similar amounts of destruction of NK1R immunoreactivity in the RTN/Ppy region or in the medullary serotonergic system produce similar effects on chemoreception. NK1R-expressing neurones at both sites are involved in central chemoreception. Bilateral destruction in the RTN/Ppy region also resulted in hypoventilation during air breathing indicating the presence of a tonic drive for breathing from NK1R-expressing neurones in this region. While we did not observe hypoventilation following SP-SAP injection in the medullary serotonergic system, it is possible that greater loss of NK1R-expressing neurones would uncover a tonic drive emanating from the region of the medullary serotonergic system as well.

We used a novel approach to selectively kill serotonergic neurones. Lappi and colleagues conjugated an antibody to the external portion of the serotonin transport protein with the cell toxin, saporin, in order to produce selective destruction of serotonergic neurones. We tested this conjugate using serotonergic cells in culture and found specific killing of TPOH-ir neurones within days by exposure to an 11nm concentration with more efficacious killing at 2 × and 5 × greater doses. From these data we chose a dose of 1μm for in vivo injection. This proved useful in that two 100 nl injections of this concentration killed 28% of medullary serotonergic neurones. That our injections had no effect on TPOH-ir neurones located in the caudal aspect of the dorsal raphe argues against any distant effects of our focal injections. But the presence of SERT protein on axons and terminals indicates that some of the medullary neurones may have been killed by anti-SERT-SAP access via these sites.

The 28% loss of medullary serotonergic neurones by our anti-SERT-SAP injections significantly reduced the ventilatory response to systemic CO₂, averaged over the postinjection period, by 15% in wakefulness and 18% in NREM sleep. There was no effect on air breathing. Other studies have examined the role of the medullary raphé in chemoreception. Extensive non-specific cell killing in the ventral midline medulla of decerebrate piglets substantially decreased the responses to CO₂ of both phrenic and hypoglossal nerve activities (Dreshaj et al. 1998). Focal non-specific neuronal inhibition in the rostral aspect of the medullary serotonergic distribution of unanaesthetized newborn piglets by dialysis of muscimol, a GABA-A receptor agonist, decreased the breathing response to CO₂ by 17% in wakefulness (this treatment also disrupted sleep cycling) (Messier et al. 2002). Both studies suggest that neurones, of unknown phenotype, in the region of the medullary serotonergic system are involved in central chemoreception.

More specific inhibition of medullary serotonergic neurones was produced by focal microdialysis of the serotonin 1A receptor agonist, 8-OH DPAT, in unanaesthetized newborn piglets. This treatment, designed to inhibit serotonergic neurones that express the serotonin 1 A auto-receptor, decreased the CO₂ response in wakefulness (sleep cycling was disrupted) in an age-dependent manner. Linear regression analysis showed a significant negative correlation (P<0.001) between the percent change in the CO₂ response with piglet age over a 3–16 days of age range (Messier et al. 2004). In a previous experiment, Mueller et al. (1984) injected the serotonin chemical toxin 5,7 DHT in 3-day-old rats and measured chemoreception in adult rats under anaesthesia. An approximately 90% reduction in tissue serotonin levels was associated with a reduction in the response to CO₂ by approximately 50%. Our anti-SERT-SAP data provide specific evidence for involvement of serotonergic neurones in the medullary serotonergic system of the adult rat in chemoreception in both sleep and wakefulness. The data from these 8 OH-DPAT and 5, 7 DHT experiments provide further support for a role for medullary serotonergic neurones in chemoreception. Based on all of these data we suggest that the serotonergic neurones may account for from 20 to 50% of the total ventilatory response to CO₂.

Other ventral chemoreceptor sites can contribute importantly to the ventilatory response to CO₂. These include the RTN/Ppy region, caudal ventrolateral medulla, caudal NTS, LC, and rostral VRG and they do not involve serotonergic neurones (Forster et al. 1997; Nattie, 1998, 1999, 2000, 2001; Okada et al. 2002; Feldman et al. 2003; Ribas-Salgueiro et al. 2003). The relative and specific roles of these different sites in normal chemoreceptor physiology remain to be fully defined. Responses to moderate intensity focal stimulation by CO₂ dialysis at RTN/Ppy, medullary raphé, and caudal NTS varied in sleep and wakefulness suggesting a state dependence in the function of some central chemoreceptor locations (Nattie, 1998, 1999, 2000, 2001; Feldman et al. 2003). In addition, the peripheral chemoreceptors at the carotid bodies are intact in all of these experiments and contribute significantly to the ventilatory responses to systemic hypercapnia. It is possible that some of the reduction in the CO₂ response observed after these lesions could be attributed to loss of modulation of this peripheral reflex by serotonergic or NK1R-expressing neurones.

How might serotonergic and NK1R-expressing neurones interact in chemoreception?

In the experiment with combined injections of both anti-SERT-SAP and SP-SAP we expected to observe an additive effect, that the CO₂ response would be reduced by the sum of the two individual lesions. The amount of cell killing in the combined toxin experiments was comparable to that in the individual toxin experiments. Yet the ventilatory response to systemic CO₂ was reduced by similar amounts. We conclude that both types of neurones are required for a normal response.

While there are no in vitro data to support a role of NK1R-expressing neurones in chemoreception per sé there are ample such data for serotonergic neurones. Richerson and colleagues have shown that serotonergic neurones in brain slices and in tissue culture are highly sensitive to focally applied CO₂ at physiological levels (Richerson et al. 2001; W. Wang et al. 2001) and that 73% of medullary TPOH-ir neurones are chemosensitive in culture (W. Wang et al. 2001). Many serotonergic neurones are located in close association with medullary arteries, which would aid in faithfully sensing arterial CO₂ (Bradley et al. 2002).

Based on the observation that serotonergic neurones are themselves chemosensitive, we put forth two explanations for our observation that simultaneous killing of both NK1R-expressing and TPOH-ir neurones has no greater effect on the CO₂ response than does killing of either population alone.

1. Serotonergic and NK1R-expressing neurones interact during the process of chemoreception. NK1R-expressing neurones may provide a necessary excitatory modulation that allows expression of the full breathing response to stimulation of serotonergic chemosensitive neurones by CO₂. Such an interaction involving NK1R-expressing neurones, the serotonergic system, and sensory sensitivity has been described. For example; NK1R-expressing lamina I dorsal horn neurones provide an excitatory input to medullary serotonergic neurones, which then affect excitability of deeper dorsal horn sensory neurones (Suzuki et al. 2002). In a second example, dorsal raphe NK1R-expressing glutamatergic neurones (Liu et al. 2002) provide an excitatory input to dorsal raphe serotonergic neurones, which can then inhibit other serotonergic neurones via 5HT_1A receptors (Valentino et al. 2003).

2. NK1R-expressing neurones simply respond to substance P released by CO₂ sensitive serotonergic neurones. There is evidence that 64% of medullary raphé neurones, which express c-fos after 1 h of exposure to 12% CO₂, also express the preprotachykinin A gene mRNA-an indicator of the presence of substance P (Pete et al. 2002). In this case, chemosensitive serotonergic neurones would excite NK1R-expressing neurones, which would then stimulate breathing. The serotonergic and NK1R-expressing neurones would act in series. Substance P is known to have direct effects on breathing. Its injection into the pre-Bötzinger complex in vitro (Gray et al. 2001) or into the RTN region in vivo stimulates respiratory output while application of an NK1R antagonist in the ventrolateral medulla inhibits breathing and the response to hypercapnia (Chen et al. 1988, 1990). SP containing processes, likely originating in the medullary raphé, synapse at respiratory neurones in nucleus ambiguus (Holtman, 1988). However, there are known serotonergic receptors on respiratory neurones at these sites so serotonergic neurones could affect their function by release of serotonin as well as substance P.

The medullary raphé, sleep and body temperature regulation

Disruption of the medullary raphé or rostral ventrolateral medulla seems to promote wakefulness. Studies with SP-SAP induced lesions of NK1R-ir neurones and processes in the RTN–Ppy region of adult rat (Nattie & Li, 2002b), muscimol inhibition of RTN (Curran et al. 2001; Darnall et al. 2001) and medullary raphe neurones (Messier et al. 2002) of newborn piglet, and 8-OH-DPAT inhibition of medullary serotonergic neurones of newborn piglets (Messier, A. Li & E. E. Nattie, unpublished observations) have all shown a significant tendency for sleep cycle disruption with less NREM sleep and more wakefulness. However, in this experiment, we did not observe a statistically significant effect on the amount of time spent in NREM or REM sleep or wakefulness among the four treatment groups, although there was a tendency for sleep disruption. It may be that lesions that develop over days as compared to acute inhibition allow a return to a more normal sleep cycle.

Medullary serotonergic neurones have been implicated in body temperature regulation via effects on brown fat metabolism and cutaneous blood flow (Cano et al. 2003). We observed no significant difference in body temperature or resting oxygen consumption in our lesioned animals. This is surprising in that room temperature is in the 22–24°C range, which is below the thermoneutral zone for the rat and we might have expected to see some body temperature effect. Perhaps our lesions did not kill a sufficient number of temperature-related neurones.

Summary and significance

NK1R-expressing neurones are involved in sensory processing (Mantyh et al. 1997) and in the regulation of blood pressure and breathing (Li & Guyenet, 1997; Chen et al. 1988, 1990; Urbanski et al. 1989; Seagard et al. 2000; Riley et al. 2002; Gray et al. 2001). As shown in this and in our prior study (Nattie & Li, 2002b), NK1R-expressing neurones are also important in central chemoreception.

In contrast, the activity of medullary serotonergic neurones, studied in vivo by Jacobs and colleagues (Jacobs et al. 2002), seems most reliably correlated with motor acts, especially repetitive motor acts, and with arousal state, their firing being highest in wakefulness. Azmitia (1999) suggested that serotonergic neurones are important in brain tissue homeostasis while Lovick (1997) and Jacobs et al. (2002) have suggested that serotonergic neurones help to coordinate autonomic activity with motor output. Serotonergic neurones are known to be involved with pain processing (Mason, 2001), the regulation of breathing (Lalley, 1986; Holtman et al. 1987), upper airway tone (Haxhui et al. 2001; Horner, 2001), body temperature regulation (Cano et al. 2003), and possible glucose sensing (Maekawa et al. 2000).

Richerson et al. (2001) has proposed from in vitro studies of slices and single neurones in culture that medullary serotonergic neurones are primary CO₂ chemosensors. The results presented here support the hypothesis that medullary serotonergic neurones are involved in the ventilatory response to an increase in CO₂. The overall importance of the serotonergic chemoreceptors in vivo is unclear, given the presence of other central and peripheral chemoreceptor sites, but the data suggest that they contribute substantially. Our results also point to an adjacent non-serotonergic population of NK1R-expressing neurones that are important for full expression of the chemoreception response in vivo. We suggest that these two different cell types interact in some as yet undetermined manner to affect chemoreception in vivo.

These serotonergic neurones may represent the neurobiological substrate for a subset of cases of the Sudden Infant Death Syndrome (SIDS) that express abnormalities in serotonergic receptor binding (Panigrahy et al. 2000; Kinney et al. 2001). We do not know how developmental defects in these neurones might lead to SIDS. One hypothesis is that these neurones participate in a number of homeostatic processes, including chemoreception, that when disrupted could in the appropriate circumstance contribute to sudden death.