INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

THE RESPIRATORY PUMP muscles, which include the diaphragm and abdominal muscles, participate in generating a large number of behaviors and protective reflexes (11, 12, 21). These muscles contract out of phase during breathing and in phase during a number of behaviors such as emesis (26) and postural adjustments (15) and in response to vestibular stimulation (30, 34, 40). A number of studies have considered the neural circuitry that regulates respiratory muscle discharges during these behaviors. All brain stem premotor neurons that regulate respiratory activity were initially thought to be confined to two columns located in the medulla, the so-called brain stem respiratory groups (10-12). However, neurons in these areas appeared to be silent during sneezing (32) and emesis (1, 29) and were shown not to play a critical role in relaying vestibular signals to respiratory motoneurons (34, 39, 41). Furthermore, recent experiments employing the transneuronal transport of pseudorabies virus (PRV) from respiratory muscles demonstrated that, in addition to cells in the respiratory groups, neurons in the medial medullary reticular formation (MRF) provide inputs to inspiratory (3, 42) and expiratory (2-4) motoneurons. Some of these MRF neurons were shown to send collateralized projections to both inspiratory and expiratory motor pools, unlike neurons located in the medullary respiratory groups that made connections with either inspiratory or expiratory motoneurons (3).

Physiological studies in cats (30) and ferrets (35) demonstrated that inactivation of MRF neurons with the use of muscimol or lidocaine increased the excitability of diaphragm and abdominal motoneurons. These data suggest that at least a subset of MRF neurons that influence respiratory muscle activity is inhibitory. This conclusion is supported by prior anatomic studies in rats (9, 18), rabbits (5), and cats (17) that demonstrated that GABAergic neurons in the medial MRF have projections to the spinal cord. However, these findings do not rule out the possibility that excitatory premotor respiratory neurons also exist in the MRF. If these cells were silent except during particular behaviors, then inactivation of this subpopulation of MRF neurons would not alter tonic respiratory muscle activity. To better understand the role of MRF neurons in respiratory control, it is necessary to determine the neurochemical phenotypes of neurons in this region that provide inputs to respiratory motoneurons. For this purpose, a dual-fluorescence immunohistochemical approach was employed that combined PRV injections into respiratory muscles to identify premotor neurons with the detection of neurochemical markers for GABAergic and glutamatergic neurons. Our hypothesis is that both GABAergic and glutamatergic MRF neurons participate in respiratory control.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Immunohistochemical procedures for detecting GABAergic and glutamatergic neurons were optimized at the onset of this study through the use of tissue archived from our laboratory's previous experiments in ferrets (2-4, 42). Subsequently, PRV was injected into respiratory muscles of four male ferrets (weight range of 1.5-2.0 kg) obtained from Marshall Farms (North Rose, NY). All procedures conformed to the National Institutes of Health "Guide for the Care and Use of Laboratory Animals" [DHEW Publication No. (NIH) 85-23, Revised 1985, Office of Science and Health Reports, DRR/NIH, Bethesda, MD 20205] and were approved by the University of Pittsburgh's Institutional Animal Care and Use Committee.

Recombinant PRV. The characteristics of the two recombinants of the Bartha strain of PRV used in this study, PRV-Bablu and PRV-152, have been published elsewhere (3, 23, 36). Both viruses were the generous gift of Dr. Lynn Enquist (Princeton University, NJ). PRV-Bablu expresses -galactosidase (-gal), and PRV-152 expresses enhanced green fluorescent protein (EGFP), under the gG and cytomegalovirus immediate early gene promoters, respectively. Both recombinants of PRV were grown in pig kidney (PK15) cells and were adjusted to a final concentration of 1 × 10⁸ plaque-forming units/ml.

Surgical procedures. After animals were maintained for at least 1 wk in the animal housing facility for acclimatization, PRV recombinants were injected into the diaphragm and rectus abdominis (RA), according to the following procedures. Surgery was performed by using aseptic procedures in a dedicated operating suite. For this purpose, animals were anesthetized by an intramuscular injection of a mixture of ketamine (25 mg/kg) and xylazine (2.5 mg/kg), which was supplemented with 0.5-1% isoflurane vaporized in O₂ to maintain areflexia. Procedures for injection of viruses were similar to those employed in our laboratory's prior studies (2-4, 42) and thus will be described only briefly. A midline incision was performed through linea alba, and the viscera were retracted to expose the ventral surface of the left diaphragm. PRV-Bablu (100 µl) was injected at multiple sites into the costal and crural regions of the diaphragm by using a 10-µl Hamilton syringe equipped with a 26-gauge needle. Similar procedures were used to inject PRV-152 (60 µl) beneath the conjunctive sheath enveloping the dorsal portion of left RA, within 3 cm of the diaphragm. On completion of injections, incisions were sutured closed, and animals were maintained under BioSafety level II conditions until the end of the experiments. A postinoculation survival period of 5 days was used for all animals, because previous studies showed that this survival period is the minimum necessary after the injection of PRV into respiratory pump muscles to produce transneuronal labeling of a substantial number of brain stem neurons (2, 3, 42). After this survival time, animals were deeply anesthetized with the use of an intramuscular injection of ketamine (35 mg/kg) and xylazine (2 mg/kg) and were perfused transcardially with 1 liter of 0.15 M NaCl followed by 2 liters of 4% paraformaldehyde-lysine-periodate fixative (27). The brain and spinal cord segments C₁-L₄ were subsequently removed, postfixed for 4-5 h in 4% paraformaldehyde-lysine-periodate, and then cryoprotected by incubation for 2 days in 30% aqueous sucrose at 4°C.

Tissue processing. As a first step in these studies, we used tissue archived from prior experiments (2-4, 42) to determine the most effective commercially available antibodies for detecting glutamatergic and GABAergic neurons, as well as the concentration of these antibodies that resulted in the optimal signal-to-noise ratio. We found that GABAergic neurons could be effectively detected by using a rabbit polyclonal primary antibody to the 67-kDa isoform of glutamic acid decarboxylase (GAD; the GABA synthetic enzyme) obtained from Chemicon (Temecula, CA). GAD₆₇ is responsible for the synthesis of >90% of GABA in the central nervous system (see Ref. 37 for review), and thus ascertaining the distribution of this enzyme is an effective method of localizing GABAergic neurons. We also determined that glutamatergic neurons could be effectively localized by using a rabbit polyclonal anti-glutamate antibody obtained from Chemicon, or a mouse monoclonal anti-glutamate antibody obtained from Advanced Targeting Systems (San Diego, CA).

Immunocytochemical procedures. One tissue set from each animal was processed by using a carbocyanine double fluorescence method described in detail elsewhere (3, 4) to determine the distribution of neurons infected by injection of PRV-Bablu and PRV-152 into the diaphragm and RA, respectively. This analysis was used to confirm that infected neurons had a similar distribution as was reported in prior studies (2-4, 42). Briefly, sections were placed for 2 days at 4°C in a primary antibody solution containing mouse anti--gal (1:5,000; Sigma Chemical, St. Louis, MO) and rabbit anti-EGFP (1:1,000; Molecular Probes, Eugene, OR) to localize neurons infected with PRV-Bablu and PRV-152, respectively. Tissue was then incubated for 2 h at room temperature in solution containing donkey anti-mouse secondary antibody conjugated to the CY3 carbocyanine (1:500; Jackson ImmunoResearch Laboratories, West Groves, PA) and goat anti-rabbit secondary antibody conjugated to Bodipy (1:300; Molecular Probes).

Controls for nonspecific labeling. The specificity of all antibodies used in this study has been tested previously by ELISA, by immunoblotting, and with immunohistochemical procedures (7, 13, 14, 16, 22, 24, 25). As additional controls in the present study, we substituted either normal rabbit serum or normal mouse serum for the primary antisera and also omitted primary antibodies. Both treatments eliminated immunolabeling.

Tissue analysis. Analyses were performed on sections spaced 200 µm apart. This spacing between sections was previously shown to be suitable for mapping the distribution of transneuronally infected neurons after the injection of PRV into the diaphragm (3, 42) or RA (2-4). Immunoreacted tissue sections were examined and photographed by using a Zeiss Axioplan photomicroscope equipped with epifluorescence and filters that selectively excited Bodipy or CY3 and with a filter that allowed for the excitation of both fluorophors. As a convention, the red fluorescence of CY3 was used to identify cells infected after injection of PRV into respiratory muscles, whereas the green fluorescence of Bodipy was used to identify neurons expressing GAD₆₇-LI or glutamate-LI. Neurons labeled with both fluorophors (i.e., virus-infected cells that expressed GAD₆₇-LI or glutamate-LI) appeared yellow. Colocalization of PRV infection with either GAD₆₇-LI or glutamate-LI was ascertained by using a ×40 objective. Electronic photographs of MRF neurons were obtained by using a Hamamatsu digital camera (Hamamatsu Photonics, Hamamatsu, Japan) and a Simple-32 PCI image analysis system (Compix, Lake Oswego, OR). Digital images were prepared for publication by using Adobe Systems (San Jose, CA) Photoshop software. Individual images were adjusted for size, brightness, and contrast, but color balance was not altered.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Specificity of the immunohistochemical methods employed. A number of previous studies have confirmed that the antibodies to GAD₆₇ and glutamate employed in the present experiments produce specific labeling of GABAergic and glutamatergic neurons, respectively (7, 13, 14, 16, 22, 24, 25). As additional controls, we conducted immunoperoxidase analyses of the labeling of neurons in the cerebellar cortex whose phenotype is established. Figure 1A illustrates GAD-labeled Purkinje cells as well as inhibitory interneurons in the cerebellar cortex. This localization duplicates the cellular localization for GAD previously reported (31). The patterns of labeling achieved with glutamate immunohistochemical localizations are illustrated in Fig. 1, B and C. Note that granule cells are densely stained, whereas they did not express GAD immunoreactivity. This pattern conforms to that documented in prior immunocytochemical localizations of glutamate (20, 24). A very few Purkinje cells expressed a low level of glutamate-LI, presumably reflecting an intracellular pool of glutamate employed for the synthesis of GABA or glutamate-containing proteins. All immunoreactivity was eliminated when primary antibodies were omitted (Fig. 1D) or when normal rabbit serum (Fig. 1E) or normal mouse serum (Fig. 1F) were substituted for primary antibodies.

View larger version (109K): [in this window] [in a new window]   Fig. 1.   Immunoperoxidase analyses of the labeling of neurons in the cerebellar cortex, to confirm the specificity of the immunohistochemical methods employed in this study. A: labeling of Purkinje cells and presumed inhibitory neurons in the cerebellar cortex achieved by using antibody to glutamic acid decarboxylase (GAD)67. B and C: labeling of granule cells achieved by using either rabbit polyclonal anti-glutamate (B) or mouse monoclonal anti-glutamate (C) antisera. D: labeling was absent when the tissue-processing procedures used for A were duplicated, except for the omission of the primary antibody to GAD67. E and F: similarly, labeling was absent when normal rabbit serum was substituted for rabbit polyclonal anti-glutamate antibody (E) or normal mouse serum was substituted for mouse monoclonal anti-glutamate antibody (F). G, cerebellar granular cell layer; M, molecular layer; P, Purkinje cell layer. Scale bars = 50 µm.

Distribution of GAD₆₇-LI and glutamate-LI in the MRF. Neurons that were immunoreactive for the synthetic enzyme for GABA (GAD₆₇) and the amino acid glutamate had overlapping distributions in the MRF. However, GAD₆₇-LI cells were highest in concentration in the rostral portion of the MRF, in particular in the ventral and ventrolateral parts of this region near the junction separating the magnocellular and gigantocellular tegmental fields from the lateral tegmental field. In contrast, glutamate-LI cells were most densely distributed in the medial MRF. The soma diameters of GAD₆₇-LI cells were typically between 25 and 60 µm, whereas the soma diameters of glutamate-LI neurons ranged from 30-70 µm.

Immunoreactivity of MRF neurons providing inputs to respiratory muscles. The distribution of neurons infected by injection of PRV-Bablu into the diaphragm and PRV-152 into RA appeared to be identical to that reported in previous studies (2-4, 42). Viral immunoreactivity was heaviest in portions of the MRF between 2 and 3.5 mm rostral to the obex, and thus sections from this region were the focus of dual-labeling analyses to detect the presence of PRV and neurochemical markers for GABA and glutamate. Figure 2 shows an example of a neuron that was immunoreactive for both PRV-Bablu and GAD₆₇, whereas Fig. 3 depicts the distribution of MRF neurons in one animal that were infected by injection of PRV-Bablu into the diaphragm and that were and were not also positive for GAD₆₇. The location of the neuron illustrated in Fig. 2 is indicated in Fig. 3 by an arrow. Table 1 shows the number of MRF neurons infected by injection of PRV-Bablu into the diaphragm that were and were not double labeled for the presence of GAD₆₇. In every animal, only a minority of PRV-Bablu-infected neurons in the MRF were positive for GAD₆₇; the percentage of double-labeled neurons ranged from 0 to 31% of the population, and the mean was 18 ± 13 (SD)%.

View larger version (33K): [in this window] [in a new window]   Fig. 2.   Example of a medullary reticular formation (MRF) neuron that is immunoreactive for both pseudorabies virus (PRV) injected into the diaphragm and for GAD67. A: the neuron is shown under illumination that excited both the red CY3 fluorochrome (which labeled neurons infected by injection of PRV-Bablu into the diaphragm) and the green Bodipy fluorochrome (which labeled neurons with GAD67 immunoreactivity). Because the neuron was labeled with both fluorochromes, it appeared yellow. A1 and A2: the same cell is illustrated under illumination that excited only 1 of the 2 fluorochromes. The location of this neuron is indicated in Fig. 3. Scale bar = 100 µm.

View larger version (15K): [in this window] [in a new window]   Fig. 3.   Camera lucida drawings of transverse sections showing the locations of MRF neurons infected after PRV was injected into the diaphragm that were () and were not () immunoreactive for GAD67. The approximate distance of each section from the obex is indicated to the left of each panel. The no. of labeled neurons in this animal (animal 3) is also designated in Table 1. An arrow indicates the location of the neuron illustrated in Fig. 2. Scale bar = 5 mm. FTG, gigantocellular tegmental field; FTL, lateral tegmental field; FTM, magnocellular tegmental field; IO, inferior olive; P, pyramidal tract; PH, nucleus prepositus hypoglossi; Rpa, nucleus raphe pallidus; SM, medial nucleus of the solitary tract; V4, 4th ventricle; XII, hypoglossal nucleus; VN, vestibular nuclei; Rm, nucleus raphe magnus.

View larger version (82K): [in this window] [in a new window]   Fig. 4.   Examples of MRF neurons that were infected after PRV was injected into the diaphragm (A) or rectus abdominis (RA; B) and showed glutamate-like immunoreactivity. A and B illustrate neurons under illumination that excited both the red CY3 and the green Bodipy fluorochromes; A1, A2, B1, and B2 depict the neuron under illumination that excited only 1 of the fluorophors. The locations of these neurons are indicated in Fig. 5. Scale bars = 100 µm.

View larger version (23K): [in this window] [in a new window]   Fig. 5.   Camera lucida drawings of transverse sections showing the locations of MRF neurons infected after PRV was injected into the diaphragm or RA that were or were not immunoreactive for glutamate. Although neurons infected after injection of PRV into the diaphragm and RA are indicated on the same panels, immunohistochemistry to visualize these neurons was conducted on separate bins of tissue from the same animal. Other data obtained from this animal are presented in Fig. 3, and Table 1 indicates the total no. of labeled neurons in this case (animal 3). The approximate distance of each section from the obex is indicated on the left of each panel. Arrows indicate the location of the neurons illustrated in Fig. 4, A and B. Scale bar = 5 mm. Abbreviations are as defined in Fig. 3 legend.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The major finding of this study is that at least two subpopulations of MRF neurons provide inputs to diaphragm motoneurons, one GABAergic and presumably inhibitory and the other glutamatergic and presumably excitatory. Furthermore, we demonstrated that a larger fraction of these premotor neurons expressed glutamate-LI than GAD-LI. The results also showed that a large proportion of MRF neurons that provide inputs to motoneurons innervating the abdominal muscle RA are glutamatergic, although technical limitations prevented the determination of whether MRF neurons that are premotor to RA express GAD-LI. However, because many MRF neurons have collateralized influences on both diaphragm and abdominal motoneurons (3), it seems likely that a subpopulation of MRF neurons that affect RA motoneuron excitability is GABAergic. Collectively, these data suggest that the MRF plays a multifaceted role in respiratory regulation, in that both excitatory and inhibitory neurons in this region that are presumably active at different times provide inputs to respiratory motoneurons.

The findings of this study are surprising, considering that previous physiological studies in the cat (30) and ferret (35) reported that inactivation of the medial MRF using muscimol or lidocaine induced a significant increase in diaphragm and abdominal muscle activity. Furthermore, inactivation of the MRF resulted in an augmentation of respiratory muscle responses to vestibular nerve stimulation (30). There are several potential explanations for the observations that chemical lesions of the MRF increase respiratory muscle activity and excitability, despite the fact that most MRF neurons that project to respiratory motoneurons are glutamatergic. One possibility is that only the inhibitory MRF premotor respiratory neurons are typically spontaneously active, so that inactivation of the more numerous excitatory neurons had little effect on the discharges of respiratory muscles. A second possibility is that, in addition to the premotor neurons that are infected at short latency after injection of PRV into respiratory muscles, the MRF contains inhibitory interneurons that indirectly influence respiratory muscle activity. If such connections are present, then inactivation of the MRF might result in an increase in respiratory muscle activity, despite the fact that a majority of MRF neurons providing direct inputs to respiratory motoneurons are excitatory. Further experiments will be required to reconcile prior physiological data with anatomic data reported in the present study.

The relative physiological roles of glutamatergic and GABAergic projections from the MRF to diaphragm and abdominal motoneurons are also yet to be determined, although previous studies provide some insights. There are presently no data to suggest that MRF neurons display respiratory-related activity or contribute to respiratory rhythmogenesis. However, a large fraction of MRF neurons receive and are excited by signals from the inner ear (6, 33), and thus peripheral lesions that eliminate vestibular inputs would be expected to reduce MRF neuronal activity. Nonetheless, removal of labyrinthine signals results in a tonic increase in respiratory muscle discharges (8), which suggests that inhibitory MRF neurons may receive vestibular inputs such that peripheral vestibular lesions result in a disinhibition of respiratory motoneurons. Data from the present study, in combination with findings from other studies (8, 30, 35), are thereby consistent with the notion that inhibitory MRF neurons are spontaneously active and serve to adjust respiratory muscle activity in accordance with an animal's posture. In contrast, excitatory MRF neurons may only be recruited to activate respiratory motoneurons in a coordinated manner during particular behaviors such as emesis, which is eliminated after chemical lesions of the medial MRF (28). Excitatory MRF neurons could also participate in producing coughing and sneezing, as well as voluntary contractions of respiratory muscles.

Whereas our results show that most MRF premotor respiratory neurons either contain a high concentration of glutamate or express the synthetic enzyme for GABA, we have not demonstrated that the neurons release GABA or glutamate as neurotransmitters. In particular, because glutamate is a precursor for GABA and is a structural component of many proteins, the possibility arises that neurons expressing glutamate-LI utilize this amino acid for the synthesis of GABA or large amounts of glutamate-containing proteins. For example, in our control experiments, we noted that an occasional Purkinje cell expressed a low level of glutamate immunoreactivity. Such labeling was infrequent and light and thus was not deemed to be a major detriment to our analysis. Nonetheless, further physiological studies will be required to confirm the conclusion that the majority of MRF premotor respiratory neurons are glutamatergic.

Although the present data suggest that the large majority of MRF premotor respiratory neurons are glutamatergic or GABAergic, it seems likely that others utilize additional neurotransmitters. Both glycinergic (19) and cholinergic (17) neurons have been reported in the MRF, although it is not known whether these neurons project to respiratory motoneurons. If it is discovered that the MRF provides a large number of neurotransmitter-specific inputs to respiratory motoneurons, it could be concluded that the MRF plays a diversity of roles in respiratory regulation. Furthermore, such a finding would indicate that the MRF is capable of modulating and modifying respiratory motoneuron activity in a variety of different ways and thus could provide the neural substrate for a large assortment of responses produced through the contraction of respiratory muscles.