Comparison with in situ hybridization and immunocytochemical studies
We have simultaneously investigated the expression of 15 glutamate receptor genes plus five splice or editing variants in the mouse. The analysis included all four AMPA receptor subunits including flip and flop splice variants as well as the R and Q editing variants of GluR2, all five KA receptor subunits, and six NMDA receptor subunits. This is the first time that the expression of all known functional glutamate receptor subunits was investigated in parallel in the brainstem.
The expression of GluR1 in the brainstem has previously been investigated (Petralia and Wenthold, 1992; Bahr et al., 1996; Williams et al., 1996; Ambalavanar et al., 1998) in immunocytochemical studies indicating weak to moderate expression of GluR1 in the PBC, the NH, and the NTS and strong expression of GluR1 in the IO. These findings are confirmed by the detection of GluR1 in all four nuclei in the present study (see Table 1). In contrast to GluR1, the expression of GluR2 in identified brainstem nuclei has not yet been studied in detail. GluR2 has been described as being weakly expressed in the lower brainstem of the adult rat (Pellegrini-Giampietro et al., 1991; Petralia et al., 1997). Using an antibody specific for both GluR2 and GluR3, moderate immunoreactivity has been detected in the PBC and in the NH. In contrast to the NH and the IO, in the NTS differences between the species have been encountered. Strong immunoreactivity for GluR2/3 has been found in the cat but only moderate signals in the rat and none at all in human tissue (Petralia and Wenthold, 1992; Williams et al., 1996; Ambalavanar et al., 1998). Our findings of strong GluR2 expression in all four nuclei (see Table 1) suggest that the immunoreactivity in the immunocytochemical studies mentioned is at least partially caused by the presence of GluR2 and not exclusively by GluR3. Similar to GluR2, weak expression of GluR3 in the brainstem has been revealed in an in situ hybridization study (Pellegrini-Giampietro et al., 1991). Our findings indicate the presence of GluR3 in all four nuclei, most prominently in the IO and the PBC (see Table 1). Compared with GluR2 and GluR3, much more is known about the expression of GluR4 in the brainstem. GluR4 is expressed in the brainstem of postnatal day 7 mice (Bettler et al., 1990) and adult rats (Bahr et al., 1996). Expression of GluR4 has been described in the PBC, the NH, the NTS, and the IO (Petralia and Wenthold, 1992; Williams et al., 1996; Ambalavanar et al., 1998). Because the present study revealed strong expression of GluR4 in all four nuclei investigated (see Table 1), qualitatively our findings are in agreement with the data obtained from the rat, cat, and human.
The expression of flip and flop splice variants of the AMPA receptors has been investigated in earlier studies in most parts of the brain but not in the brainstem (Sommer et al., 1990; Monyer et al., 1991; Tölle et al., 1995). Flip splice variants are prominent throughout early postnatal brain development. In contrast, flop splice variants appear later in development, being still less prominent than flip variants at postnatal day 8 (Monyer et al., 1991). Assuming that similar developmental changes occur in the rat and the mouse, the strong predominance of the flop over the flip splice variants in the IO seen in the present study (see Table 1) is quite extraordinary. In contrast to the IO, no preferential splicing of the AMPA receptor subunits was detected in the NTS, the PBC, and the NH in tissue slices (see Table 1). For the AMPA receptor GluR2 editing variants exist in addition to the splice variants. Low levels of the Q variant of GluR2 have been shown to be present in fetal but not in adult brain (Burnashev et al., 1992; Paschen and Djuricic, 1995). In the present study, single respiratory-related neurons of the PBC and the NH motoneurons almost exclusively contained the R editing variant but not the Q editing variant of GluR2 (see Table 3); this fits well with previous findings in the rat.
Compared with the AMPA receptor subunits, much less is known about the expression of KA receptor subunits in the PBC, the NH, the NTS, and the IO. In our study, we observed GluR5 expression in all nuclei analyzed, being most prominent in the NTS. This fits well with the strong expression of GluR5 in the brainstem in embryonic day 12 to postnatal day 12 mice as shown by in situ hybridization (Bettler et al., 1990). More attention has been paid to GluR6 than to GluR5. Weak hybridization signals were found in several mouse brainstem nuclei (Egebjerg et al., 1991), and immunoreactivity was detected in young rats in all four nuclei by an antibody recognizing both GluR6 and GluR7 (Petralia et al., 1994a). Because the present study revealed the presence of GluR6 mRNA in all four nuclei, our findings suggest that the immunoreactivity described by Petralia et al. (1994a) was due at least partially to the presence of the GluR6 protein. Similar to GluR6, the expression of GluR7 has not yet been extensively studied in the brainstem. GluR7 is expressed in the entire CNS in embryonic day 15 rats and in the brainstem of adult rats (Lomeli et al., 1992). This fits well with the detection of GluR7 in the PBC demonstrated in our study. Like GluR7, weak expression of KA1 RNA in the brainstem of adult and embryonic rats has been described (Herb et al., 1992; Wisden and Seeburg, 1993). This might partly be due to the presence of KA1 in the NTS and the IO, but also in the NH, as shown in the present study. In contrast to the other KA receptor subunits, the expression of KA2 has been previously studied in identified brainstem nuclei. Weak expression of KA2 was detected in the PBC, the NH, the NTS, and the IO of rats (Petralia et al., 1994a). In this study, the reported expression of KA2 in the PBC and the NH could not be confirmed in mice (see Table 1). The same primers used for KA2 in the present study have been successfully used by others to detect both KA1 and KA2 at the single-cell level in the rat (Ruano et al., 1995); furthermore, the KA2 primers have no mismatches with the mouse KA2 sequence, and the detection limit of KA2 is similar to that of other glutamate receptor subunits (see Table 4). Therefore, the difference in KA2 expression observed in this study as compared with the study by Petralia et al. (1994a) might be due to species differences or the different developmental stages used [5-8-day-old mice in the present study versus rats of 120-250 g in the study by Petralia et al. (1994a)].
Table 4. Comparison of efficiency of RT-PCR for different glutamate receptor subunits
The expression of NMDAR1 in the PBC, the NH, the NTS, and the IO has been investigated in prior studies that revealed the presence of NMDAR1 both at the RNA and at the protein level (Petralia et al., 1994b; Watanabe et al., 1994; Ambalavanar et al., 1998). The detection of NMDAR1 at the RNA level was confirmed in the present study. Like NMDAR1, NR2A is expressed in the brainstem. The expression of NR2A in mouse brainstem is known to start between postnatal day 0 and 21 (Watanabe et al., 1992). NR2A was detectable in the PBC, the NH, the NTS, and the IO (Watanabe et al., 1994). These findings were confirmed in the present study, in which strong expression of NR2A was shown in all four nuclei. This suggests that the immunoreactivity found with an antibody specific for both NR2A and NR2B (Petralia et al., 1994c) in young rats in all four nuclei is at least partially caused by the presence of NR2A. The expression of NR2B in the mouse brainstem starts before NR2A expression, but between postnatal day 1 and 21 a developmental shift in the expression pattern occurs, and NR2B is largely replaced by NR2A (Watanabe et al., 1992). In the NTS, the IO, and the NH, but not in the PBC, in postnatal day 21 mice weak expression of NR2B was described by Watanabe et al. (1994). Because in our study strong NR2B expression was demonstrated in postnatal day 5-8 mice in all four nuclei, our findings are in agreement with an incomplete developmental switch from NR2B to NR2A between postnatal day 0 and 21 in the brainstem. In contrast to NR2A and NR2B, NR2C reportedly is not expressed in the PBC, the NH, the NTS, and the IO (Watanabe et al., 1994). In the present study, however, NR2C expression was observed in all four nuclei (see Table 1). The differences between the two studies might be caused by a higher sensitivity of the RT-PCR technique compared with in situ hybridization. The presence of NR2D in the brainstem has been shown at the RNA level (Nakanishi, 1992; Watanabe et al., 1992; Monyer et al., 1994) and also at the protein level (Wenzel et al., 1996; Laurie et al., 1997). Nevertheless, Watanabe et al. (1994) failed to detect NR2D in the PBC, NH, NTS, and IO of mice. In contrast, in our study NR2D turned out to be very prominent in all four nuclei (see Table 1). These differences again might be caused by different sensitivities of the methods used. Compared with the other NMDA receptor subunits, much less is known about the modulatory subunit NR3A. The expression of NR3A in the brainstem has been mentioned in one study (Ciabarra et al., 1995). NR3A expression is high in mouse in the first 2 weeks after birth but subsequently decreases (Sucher et al., 1995). This accords well with the detection of NR3A in all four nuclei from 5-8-day-old mice in the present study.
Glutamate receptor subunits in nonneuronal cells
Glutamate receptor subunits have been detected in glial cells of the postnatal cortical layer I and of the corpus callosum (Monyer et al., 1994), in tanycytes (Molnar et al., 1993; Petralia et al., 1994a), pituitary gland cells (Petralia et al., 1994a, b), and Bergmann glial cells (Monyer et al., 1991). Comparing the data in the present study obtained from tissue slices of the PBC (Table 1) with the data from single neurons of the PBC (Table 2), an obvious difference is the strong preferential splicing of the flip variants of the AMPA receptors, which is only seen in single neurons but not in tissue slices. This observation is most likely explained by the expression of flop variants in nonneuronal cells located within the PBC.
Additional differences have been observed between glutamate receptor expression data obtained from PBC tissue slices and single respiratory-related interneurons harvested from the PBC. Because at the single-cell level detection of GluR5, KA1, and the NMDA receptor subunits (except NR2D) was rather poor (see Table 2) compared with the tissue slices (see Table 1), these differences might reflect an overall lower detection rate for the single-cell RT-PCR as compared with whole-tissue RT-PCR. This might be due to less template material, a slightly less efficient RT-PCR technique (Table 4), or different cell types in the area of the PBC expressing different glutamate receptor subunits. For NR2D, but not for NMDAR1, based on investigation of both NMDAR1 and NR2D at the RNA and protein level with development, a posttranscriptional mechanism regulating the amount of NR2D protein has been postulated (Wenzel et al., 1996). Therefore, the observed excess of NR2D over NMDAR1 at the RNA level might not necessarily lead to excess amounts of NR2D over NMDAR1 proteins.
Functional implications of glutamate receptor subunit expression pattern
Functional NRs are not necessary for the prenatal development of the circuits that generate the respiratory rhythm (Funk et al., 1997). However, the respiratory network continues to develop postnatally (Ramirez et al., 1996, 1997). Also, critical periods for NR-mediated activity-dependent development have been demonstrated postnatally in many systems [visual cortex (Shatz, 1990), trigeminal system (Li et al., 1994), and spinal cord motoneurons (Kalb and Hockfield, 1990)]. Because NR3A is thought to be involved in the development of synaptic elements (Das et al., 1998), NR3A in postnatal day 5-8 mouse brainstem nuclei might serve to down-regulate the NR function in these nuclei and thereby reduce the number of newly formed spines.
Because the mice used in this study were 5-8 days old, only the R editing variant of GluR2 is expected to be present (Burnashev et al., 1992; Paschen and Djuricic, 1995). This has been confirmed for the PBC and the NH at the single-cell level (see Table 3). Only the R editing variant, but not the Q editing variant, of GluR2 confers Ca2+ impermeability to AMPA receptors (Hume et al., 1991; Verdoorn et al., 1991). Therefore, the abundant expression of edited GluR2 in all four nuclei suggests formation of AMPA receptors with low Ca2+ permeabilities. For this reason the AMPA receptors of the respiratory-related interneurons of the area of the PBC are unlikely to be highly Ca2+-permeable and thus may contribute only small amounts of calcium influx to activity-related Ca2+ oscillations in these neurons (Frermann et al., 1999).
Selective vulnerability of motoneurons due to excitotoxic events is associated with motoneuron diseases such as amyotrophic lateral sclerosis in humans (DePaul et al., 1988; Rothstein and Kuncl, 1995), and hypoglossal motoneurons are among the motoneuron populations most notably affected (Reiner et al., 1995). Several mechanisms have been discussed to explain this selective vulnerability. These include lack of calcium homeostasis resulting from rapid calcium signaling (Lips and Keller, 1998; Palecek et al., 1999) and/or weak expression of Ca2+-binding proteins with concomitant low endogenous Ca2+ buffering capacity (Reiner et al., 1995; Shaw and Ince, 1997; Lips and Keller, 1998), as well as selective vulnerability of axonal neurofilaments resulting from long-distance axonal projections in motoneurons (Morrison et al., 1998). One particularly attractive hypothesis for the cause of this differential vulnerability is the expression of an extraordinary collection of Ca2+-permeable ion channels, including Ca2+-permeable AMPA receptors, which leads to significant amounts of calcium influx on small membrane depolarizations (Pellegrini-Giampietro et al., 1997; Williams et al., 1997). In fact, expression of Ca2+-permeable AMPA receptors lacking GluR2 has been described as enhancing neuronal susceptibility to glutamate toxicity (Choi and Rothman, 1990; Bennett et al., 1996; Shaw et al., 1999). In contrast to this view, we found that the edited GluR2 variant is strongly expressed in NH motoneurons (see Table 3). Accordingly, these neurons are unlikely to possess a significant population of highly Ca2+-permeable AMPA receptors. Thus, our data suggest that mechanisms other than the expression of AMPA receptors lacking GluR2 might be responsible for Ca2+-mediated selective damage of NH motoneurons.