Collateral projections of neurons in laminae I, III, and IV of rat spinal cord to thalamus, periaqueductal gray matter, and lateral parabrachial area

Projection neurons in lamina I, together with those in laminae III–IV that express the neurokinin 1 receptor (NK1r), form a major route through which nociceptive information reaches the brain. Axons of these cells innervate various targets, including thalamus, periaqueductal gray matter (PAG), and lateral parabrachial area (LPb), and many cells project to more than one target. The aims of this study were to quantify projections from cervical enlargement to PAG and LPb, to determine the proportion of spinothalamic neurons at lumbar and cervical levels that were labelled from PAG and LPb, and to investigate morphological differences between projection populations. The C7 segment contained fewer lamina I spinoparabrachial cells than L4, but a similar number of spino-PAG cells. Virtually all spinothalamic lamina I neurons at both levels were labelled from LPb and between one-third and one-half from PAG. This suggests that significant numbers project to all three targets. Spinothalamic lamina I neurons differed from those labelled only from LPb in that they were generally larger, were more often multipolar, and (in cervical enlargement) had stronger NK1r immunoreactivity. Most lamina III/IV NK1r cells at both levels projected to LPb, but few were labelled from PAG. The great majority of these cells in C7 and over one-fourth of those in L4 were spinothalamic, and at each level some projected to both thalamus and LPb. These results confirm that neurons in these laminae have extensive collateral projections and suggest that different neuronal subpopulations in lamina I have characteristic patterns of supraspinal projection. J. Comp. Neurol. 515:629–646, 2009.

The superficial dorsal horn of the spinal cord (laminae I and II; Rexed, 1952) receives a major input from nociceptive primary afferents, many of which contain substance P. This acts on the neurokinin 1 receptor (NK1r), which is found on a significant proportion of neurons in lamina I as well as on some of those in laminae III-VI (Bleazard et al., 1994;Liu et al., 1994;Nakaya et al., 1994;Brown et al., 1995;Littlewood et al., 1995). In the dorsal horn, the NK1r is thought to be restricted to cells that respond to noxious stimuli (Henry, 1976;Salter and Henry, 1991). Lamina I contains numerous projection neurons, most of which express the NK1r (Ding et al., 1995;Marshall et al., 1996;Li et al., 1998;Todd et al., 2000;Spike et al., 2003). There is also a population of large NK1r-immunoreactive projection neurons in laminae III and IV with prominent dorsal dendrites that enter lamina I (Marshall et al., 1996;Naim et al., 1997;Todd et al., 2000). Cells belonging to both of these populations send predominantly contralateral projections to several brain regions, including the lateral parabrachial area (LPb), periaqueductal gray matter (PAG), and thalamus.
The spinothalamic tract is an important pathway, providing direct spinal input to the thalamus, from which information is transmitted to cortical areas that underlie pain perception. However, we recently reported that there are only ϳ15 lamina I spinothalamic neurons per side in the L4 segment of the rat (Al-Khater et al., 2008). Because there are around 400 lamina I projection neurons on each side in this segment (Spike et al., 2003), the proportion projecting to thalamus is only 4%. In contrast, spinothalamic lamina I neurons were much more be retrogradely labelled from LPb and determine whether spinoparabrachial neurons that project to thalamus differ morphologically from those that do not, 3) to quantify collateral projections from spinothalamic neurons to the PAG, and 4) to establish whether any neurons project to all three of these targets. The other four rats (experiments PAG1-4) were anesthetized with isofluorane and placed in a stereotaxic frame, after which anesthetic was administered through a mask attached to the frame. Each of these rats received two injections of 100 nl 4% Fluorogold targeted on the caudal thalamus on the left (as described above) and 100 nl 1% CTb (Sigma) into the PAG on the left side. Tissue from two of these animals (PAG1-2) had also been used in the study of gephyrin-coated lamina I spinothalamic neurons in L5 (as described above). The PoT nucleus was targeted for the thalamic injections, because this is a major termination zone for projections from the superficial dorsal horn (Gauriau and Bernard, 2004a), and we have shown that injections of tracer into this region label virtually all spinothalamic neurons in lamina I, together with those in laminae III and IV that express the NK1r (Al-Khater et al., 2008). All injections were made through glass micropipettes, which were left in place for 5 minutes after the completion of each injection to minimize leakage of tracer back up the track. In all cases, a different pipette was used for each tracer. The animals made an uneventful recovery from anesthesia. After a survival period of 3 days, they were reanesthetized with pentobarbitone (300 mg i.p.) and perfused through the heart with a fixative that contained 4% freshly depolymerized formaldehyde. Lumbar and cervical spinal cord segments were removed and stored in fixative for 24 hours, and the brain was cryoprotected in 30% sucrose in fixative overnight. cases, the spread of tracer from the injection sites was plotted onto drawings of the thalamus and brainstem (Paxinos and Watson, 2005), and representative examples were photographed.

MATERIALS AND METHODS Surgery
The C7 and L4 spinal cord segments from all animals were initially notched on the left so that the two sides could subsequently be distinguished and were then cut into 60-mthick transverse sections with a Vibratome. The sections were incubated free-floating at 4°C for 3 days in guinea pig anti-Fluorogold (Protos Biotech Corp., New York, NY; 1:500), goat anti-CTb (1:5,000), and rabbit anti-NK1r (Sigma-Aldrich; 1:10,000) and then overnight in species-specific secondary antibodies that were raised in donkey and conjugated either to Alexa 488 (Invitrogen, Paisley, United Kingdom; 1:500) or to rhodamine red or Cy5 (Jackson Immunoresearch, West Grove, PA; 1:100). The sections were mounted in serial order in antifade medium and stored at -20°C.
The C6 and L5 segments from all seven rats and the C8 and L3 segments from experiments Pb1-3 were also notched on the left side and were cut into 60-m-thick horizontal sections. Those from C8 and L3 were processed in the same way as described above, whereas sections from C6 and L5 were incubated for 3 days with guinea pig anti-Fluorogold, goat anti-CTb and mouse monoclonal antibody against gephyrin (mAb 7a; Synaptic Systems, Gö ttingen, Germany; 1:1,000) and then overnight in fluorescent secondary antibodies (as described above). Sections were mounted and stored at -20°C.

Antibody characterization
The NK1r antibody (catalog No. S8305) was raised in rabbit against a peptide corresponding to amino acids 393-407 at the C-terminus of the rat NK1r, which was conjugated to keyhole limpet hemocyanin. The antibody recognises a 46-kDa band in Western blots of rat brain extracts, and this staining is specifically abolished by preabsorption of the antibody with the immunizing peptide (manufacturer's specification). It has been shown that there is no immunostaining with this antibody in sections of medulla and cervical spinal cord from mice in which the NK1r has been deleted (NK1 -/-), whereas staining is present in sections from wild-type mice (Ptak et al., 2002).
The mouse monoclonal antibody against gephyrin was generated against an extract of rat spinal cord synaptic membranes (Pfeiffer et al., 1984) and has been extensively characterized and shown with Western blots to bind to a 93-kDa peripheral membrane protein (gephyrin) in extracts of rat brain membranes (Becker et al., 1989;Kirsch and Betz, 1993).
Goat (catalog No. 703) and guinea pig (catalog No. NM101) polyclonal antibodies were raised against CTb and Fluorogold, respectively. Specificity of each of these antibodies was shown by the lack of staining in regions of the CNS that did not contain neurons that had taken up and transported the tracer and by the presence of immunostaining in populations of neurons that are known to project to the injection sites. The specificity of the Fluorogold antibody was also directly confirmed by comparing Fluorogold fluorescence (observed with an UV filter set) with that for anti-Fluorogold in individual neurons. In all cases examined, there was a perfect match between the two types of fluorescence.

Confocal microscopy and analysis
All analysis of lamina I neurons and of the large NK1rimmunoreactive cells in laminae III and IV was performed on the right (contralateral) dorsal horn. Transverse sections from C7 and L4 segments of all seven rats were used to analyze the numbers of retrogradely labelled lamina I neurons that contained one or both tracers. In each case, 10 or 20 sections (an alternate series) were scanned sequentially (to avoid fluorescent bleed-through) with a confocal microscope (Bio-Rad Radiance 2100; Bio-Rad, Hemel Hempstead, United Kingdom) through dry (؋20) and oil-immersion (؋40) lenses. Darkfield microscopy was used to distinguish laminar boundaries, and retrogradely labelled cells were judged to be in lamina I if they were very close to the dorsal border of the dorsal horn or lay dorsal to the dark band identified as lamina II with darkfield microscopy. To correct for the overcounting that results from the presence of transected cells at the section surfaces, cells were only included in the sample if their nucleus (identified as a filling defect) was entirely contained within the Vibratome section or if part of the nucleus was present in the first optical section in the z-series (corresponding to the top of the Vibratome section); they were excluded if part of the nucleus was present in the last optical section (Spike et al., 2003;Al-Khater et al., 2008). Spinothalamic lamina I neurons are infrequent in the lumbar enlargement (Al-Khater et al., 2008), so we examined 20 sections through the L4 segment to count these cells (and to determine whether they were doublelabelled). We also used 20 sections to quantify spinal neurons labelled from PAG in the L4 segment, whereas 10 sections were used for the other parts of this analysis. In this way, the mean number of retrogradely labelled lamina I cells containing CTb, Fluorogold, or both tracers per 600 m (C7 for each injection site, L4 for LPb injections) or 1,200 m (L4 for thalamic and PAG injections) was determined for each experiment. The presence or absence of NK1r was also noted for each retrogradely labelled lamina I cell. To compare the mediolateral distribution of spinothalamic and spinoparabrachial neurons in lamina I, the locations of single-or double-labelled neurons in the sections analyzed were plotted onto an outline drawing of the dorsal horn with Neurolucida for Confocal software (MicroBrightField Inc., Colchester, VT). Lamina I was divided into three equal parts (medial, middle, and lateral), and the numbers of spinoparabrachial and spinothalamic neurons in each part were counted. For this analysis, spinothalamic neurons in the L4 segment were analyzed on 20 sections, whereas those in C7, together with spinoparabrachial neurons in both segments, were analyzed on 10 sections.
In all seven animals, the complete series of sections from C7 and L4 were also used to count the number of NK1rimmunoreactive neurons that had cell bodies in laminae III or IV and dorsal dendrites that could be followed into laminae I or II and to determine the proportion of these cells that was retrogradely labelled with one or both tracers. Darkfield microscopy was used to ensure that all cells had their somata ventral to lamina II. Sections were initially viewed with epifluorescence through a ؋20 lens to identify NK1rimmunoreactive cell bodies in laminae III or IV. In most cases, it was possible to determine with epifluorescence microscopy whether the dendrites of these cells entered the superficial dorsal horn (laminae I or II). However, in some cases (particularly when dendrites had to be followed into serial sections), it was necessary to scan with the confocal microscope. This was also used to determine whether the cells were retrogradely labelled with CTb and/or Fluorogold, to measure the distance between the cell body and the overlying dorsal white matter, and to define the cells' mediolateral location. We have previously reported that, in the lumbar enlargement, the proportion of these cells retrogradely labelled from the thalamus was higher for those that were medially located. For the L4 segments, we therefore plotted the locations of all of these cells with Neurolucida for Confocal onto an outline of the dorsal horn, drew a vertical line midway through the mediolateral extent of lamina III, and divided the cells into two groups, those in the medial and those in the lateral halves of the dorsal horn.
For experiments Pb1-3, all of the horizontal sections through the C8 and L3 segments that contained retrogradely labelled lamina I neurons were scanned sequentially through dry (؋20) and oil-immersion (؋40) lenses to reveal NK1r, CTb, and Fluorogold. These scans were used to compare the morphology and NK1r expression of neurons that were retrogradely labelled from both thalamus and LPb with those of neurons labelled only from LPb. For C8, we obtained confocal image stacks (1 m z-separation) through cell bodies and dendritic trees of all retrogradely labelled neurons. A much lower number of Fluorogold-labelled cells was present in L3. Therefore, in this segment, only regions that contained Fluorogold-labelled cells were scanned, but all of the retrogradely labelled cells in these scans were analyzed. Cells were excluded from the sample if they were so close to one surface of the Vibratome section that substantial parts of their proximal dendrites (and/or cell bodies) were not present in the section, because this made it impossible to allocate them to one of the three morphological classes. Drawings of the cell bodies and proximal dendrites of all of the retrogradely labelled neurons included in the sample were made with Neurolucida for Confocal software, and the presence or absence of CTb and Fluorogold was recorded for each cell. The drawings were used to analyze neuronal morphology. For each cell in the sample, morphology was assessed independently by two observers who were blind to the projection target(s), and cells were provisionally allocated to one of the following classes: fusiform, multipolar, or pyramidal (Zhang et al., 1996;Zhang and Craig, 1997). In cases of initial disagreement, the cells were reexamined and, where possible, allocated to one of these classes. A small number of cells could not be assigned to one of the three groups because they showed features that were transitional between two of the three morphological classes, and a few could not be allocated to any of these classes because of their atypical appearance (Zhang et al., 1996;Zhang and Craig, 1997). These cells were defined as "unclassified." The maximal cross-sectional area of the soma of all cells was measured from projected confocal images with Neurolucida for Confocal (Puská r et al., 2001; Polgá r et al., 2002). Strength of NK1r immunoreactivity on the plasma membrane was also recorded; because of the variation in staining intensity at different depths of the Vibratome section, a scoring system was used (Spike et al., 2003), and each cell was assigned a score ranging from 4 (strong) to 1 (very weak) or 0 (negative). Because the sample of spinothalamic neurons in the L3 segments was small (a total of 21 cells in the 3 Pb experiments), we also analyzed morphology and soma size of spinothalamic neurons from the L5 segments of these experiments. To allow unbiased analysis of morphology, the spinothalamic cells in this segment, together with a sample of cells labelled only from LPb, were drawn and assessed by two observers blind to their projection targets, as described above. However, only the spinothalamic cells from L5 were included in the analysis.
For all seven rats, horizontal sections through the C6 and L5 segments were examined with an oil-immersion (؋40) lens to allow identification of the large gephyrin-coated lamina I cells, and these were then scanned sequentially to reveal gephyrin, CTb, and Fluorogold. These scans were used to determine the proportion of the gephyrin-coated cells in the C6 segment that was retrogradely labelled from thalamus, LPb, or PAG and the proportion of cells of this type in L5 that were labelled from PAG. Figures were composed with Adobe Photoshop (version 7.0). In some cases, image brightness and contrast were adjusted by using the Levels setting.

Injection sites
The spread of tracers in experiments Pb1-3 was briefly reported in our previous paper (Polgá r et al., 2008), and photomicrographs of the injection sites from one case were illustrated. The drawings in Figure 1 show the spread of Fluorogold and CTb in a series of coronal sections from these experiments. These indicate that the PoT was completely filled with Fluorogold in each case, with spread into surrounding areas (other thalamic, anterior pretectal, and deep mesencephalic nuclei) but not into hypothalamus, PAG, or LPb. CTb injections filled the LPb with variable spread into medial parabrachial, cuneiform, and Kö lliker-Fuse nuclei. In one case (Pb2), there was spread into the extreme caudal part of the ventrolateral PAG.
The extent of spread of Fluorogold in experiments PAG1-4 ( Fig. 2) was similar to that described above. The CTb in these experiments spread between interaural 0.4 and 2.4 mm and was largely contained within the PAG, except in PAG1 and PAG3, in which there was some spread dorsally into the superior colliculus (Fig. 2). The dorsolateral, lateral, and ventrolateral columns of the PAG were filled to a variable extent in these experiments. Examples of injection sites from the PAG experiments are shown in Figure 3.

Analysis of lamina I projection neurons in transverse sections
This part of the study was performed on sections from the C7 and L4 segments. Neurons that were retrogradely labelled from each injection site (Fluorogold-positive from thalamus and CTb-positive from LPb or PAG) were found in lamina I and throughout the deeper laminae (III-VIII, X). The distributions of retrogradely labelled cells were generally similar to those reported for each of these targets in previous studies in the rat  In the Pb experiments, we found that all of the Fluorogold-positive (spinothalamic) lamina I neurons in C7 were also labelled with CTb and that the great majority (93%) of those in L4 were CTb labelled (Table 1). Doublelabelled cells made up 45% of lamina I neurons labelled from LPb in C7 but only 6% of those in L4 (Table 1). In the PAG series, many of the lamina I spinothalamic cells (47%) in C7 were also labelled from the PAG, and these cells represented 58% of the spino-PAG population (Table 2). In the L4 segment, 33% of lamina I spinothalamic neurons were labelled from PAG (corresponding to 9% of spino-PAG cells; Table 2). Examples of single-and double-labelled lamina I neurons are illustrated in Figure 4.
In experiments Pb1-3, retrogradely labelled cells positive for one or both tracers were present throughout the mediolateral extent of lamina I but were concentrated in its middle one-third. The numbers of spinothalamic cells and of cells labelled only from LPb that were present in the medial, middle, and lateral parts of lamina I in these three experiments are shown in Table 3 and examples of their distribution are illustrated in Figure 5. For each of these regions, there was no difference in the proportion of cells labelled from thalamus or only from LPb in either segment ( 2 test, P ‫؍‬ 0.29 for C7, P ‫؍‬ 0.18 for L4).

Morphology, soma size, and NK1r expression of lamina I projection neurons in horizontal sections
This analysis was carried out on the C8 and L3 segments of experiments Pb1-3. In addition, for the morphological analy-   Although neurons belonging to each morphological type were seen within each projection population in both segments (Table 4), the proportions belonging to each type differed significantly between spinothalamic neurons and neurons labelled only from LPb ( 2 test, P < 0.05 for C8, P < 0.001 for L3/L5). Spinothalamic neurons were more often multipolar and less often fusiform than those labelled only from the LPb. This was particularly evident for the lumbar segments, where 32 of 55 (58%) spinothalamic neurons were classified as multipolar. Examples of neurons belonging to different morphological classes are illustrated in Figure 6.
The cross-sectional areas of cell bodies of spinothalamic neurons and those labelled only from LPb are shown in Figure  7. The soma areas of spinothalamic neurons ranged from 142 to 1,132 m 2 (median 434, n ‫؍‬ 142) for those in C8 and from 173 to 1,350 m 2 (median 380, n ‫؍‬ 55) for those in L3/L5, compared with 159 -1,016 m 2 (median 354, n ‫؍‬ 205) and 149 -1,230 m 2 (median 313, n ‫؍‬ 184) for neurons labelled only from LPb in C8 and L3, respectively. These differences were highly significant (Mann-Whitney rank sum test, P < 0.001 for both segments). Because we have recently reported that large gephyrin-coated lamina I neurons (which are either nonimmunoreactive or very weakly immunoreactive for the NK1r) make up ϳ20% of the spinothalamic population in the L5 segment (Polgá r et al., 2008), we also analyzed soma size for the cells that were assigned a strength of 2-4 for NK1r immunoreactivity in the C8 and L3 segments. For these cells, areas of spinothalamic neurons were 253-1,132 m 2 (median 477, n ‫؍‬ 108) and 274 -577 m 2 (median 398, n ‫؍‬ 14) for C8 and L3, respectively. Corresponding values for spinoparabrachial neurons that were not labelled from thalamus were 241-1,016 m 2 (median 374, n ‫؍‬ 137) for C8 and 163-1,230 m 2 (median 319, n ‫؍‬ 115) for L3. The differences between the two projection populations were still highly significant (Mann-Whitney rank sum test, P < 0.005 for both segments).
Results of the analysis of NK1r expression are summarized in Table 5, and examples of immunostaining are shown in Figure 6q-u. NK1r immunoreactivity was detected on 82% and 81% of spinothalamic neurons in C8 and L3 and on 79% and 72% of spinoparabrachial neurons in these segments, respectively. Neurons with NK1r scores of 0 (negative) to 4 (strong) were found within each projection population in both segments. However, in C8, the strength of NK1r expression was significantly higher among spinothalamic neurons than among neurons labelled only from LPb (Mann-Whitney rank sum test, P < 0.05). Within this segment, NK1r immunoreac-  Table 3 was performed on 20 sections.

tivity was scored 3 or 4 in 59% of spinothalamic neurons, but only in 43% of the other spinoparabrachial cells. There was no significant difference in NK1r strength among the two projection populations in L3 (Mann-Whitney rank sum test, P ‫؍‬ 0.9). It has been reported that relatively few NK1r-immunoreactive spinothalamic cells are pyramidal (Yu et al., 2005)
, so we analyzed NK1r expression among pyramidal spinothalamic neurons in C8. We found that 76% of these cells were NK1r immunoreactive and that 64% of them were scored 3 or 4 for NK1r strength. Figure 8.

Lamina III/IV NK1r-immunoreactive projection neurons
The analysis of these cells was carried out on complete sets of serial transverse sections from the C7 and L4 segments, and results are shown in Tables Figure  9. The proportion of these cells labelled from PAG was much lower in both C7 (6%) and L4 (8%), although the numbers labelled from thalamus were similar to those seen in the Pb series (89% and 28%, respectively, Table 7). In both segments, 4 -5% of these cells were double-labelled. When results from all seven experiments were combined, the proportion of these cells labelled from thalamus was 83% for C7 and 28% for L4.

In the Pb experiments, most of the lamina III/IV NK1r cells in C7 were retrogradely labelled from thalamus (75%) or from LPb (87%), and 64% were labelled from both targets. In the L4 segment, 29% of these cells were labelled from thalamus and 64% from LPb, with 14% double-labelled (Table 6). Examples of double-labelled cells in both segments are shown in
In the L4 segment, the proportion of lamina III/IV NK1r cells in the medial half of the dorsal horn retrogradely labelled from the thalamus was 35/72 (data pooled from all seven experiments), whereas the corresponding proportion for those in the lateral half was 14/98. These differed significantly ( 2 test, P < 0.001).

DISCUSSION
The main findings of this study are that 1) the number of lamina I neurons retrogradely labelled from LPb is considerably lower in C7 than in L4, although numbers labelled from PAG in these segments are similar; 2) >95% of spinothalamic lamina I neurons in both enlargements are labelled from LPb, and between one-third and one-half are also labelled from PAG; 3) lamina I spinothalamic neurons differ from other neurons labelled from LPb in morphology, soma size, and strength of NK1r expression; and 4) many lamina III/IV NK1rimmunoreactive projection neurons in cervical enlargement and some of those in lumbar enlargement are labelled from both thalamus and LPb.

Technical considerations
As in all retrograde tracing studies, it is necessary to consider the possibility of uptake of tracer by fibers passing near the injection sites. Axons from the spinal cord are located ventrally in the medulla (Lund and Webster, 1967;Mehler, 1979). In the pons, axons of lamina I spinoparabrachial neurons, which probably represent collateral branches (McMahon and Wall, 1985), ascend dorsally near the lateral aspect of the brainstem at a rostrocaudal level close to interaural 0, to enter the lateral parabrachial area. These give rise to a substantial plexus that occupies the lateral crescent, dorsal lateral, superior lateral, and external lateral nuclei of the LPb, together with a smaller projection to the Kö lliker-Fuse nucleus (Slugg and Light, 1994;Bernard et al., 1995;Feil and Herbert, 1995). From here, axons travel dorsomedially into PAG and arborize extensively within its caudal part (Bernard et al., 1995;Feil and Herbert, 1995;Keay et al., 1997). Since injections of CTb into rostral PAG and surrounding regions label very few lamina I neurons (Lima and Coimbra, 1989), it is unlikely that spinothalamic axons reach the thalamus by passing rostrally from PAG. Anterograde tracing studies have not reported a rostral continuation of axons that have entered LPb from the superficial dorsal horn, apart from those that pass into the PAG, and it appears that the parent axons of lamina I spinothalamic tract neurons are located ϳ500 m lateral to the external lateral nucleus of the LPb (J.F. Bernard, unpublished observations).
Because the projection from superficial dorsal horn occupies most of the mediolateral extent of LPb and extends to its lateral edge, we aimed to fill the entire LPb with tracer in experiments Pb1-3. In Pb2, there was some extension of CTb lateral to LPb, and it is possible that in this case tracer was taken up by some spinothalamic axons. However, it is unlikely that this had a significant effect on the results, as CTb did not spread lateral to the LPb in the region occupied by these axons in experiments Pb1 and Pb3, and both the numbers of spinoparabrachial cells and the proportion of spinothalamic neurons labelled with CTb were consistent among the three experiments. The projection from superficial dorsal horn to PAG travels through the rostral part of the parabrachial area, and it is possible that axons of some lamina I cells project to PAG without arborizing in LPb. If this is the case, and if these cells were labelled through uptake of CTb into their axons within LPb, this would result in an overestimate of the proportion of spinothalamic neurons that projected to LPb. However, because many fewer cells are labelled from the PAG than from LPb, this is unlikely to have had a major effect on our results. In Pb1-3, there was some spread of CTb into other structures, such as the Kö lliker-Fuse, medial parabrachial, and cuneiform nuclei. However, the Kö lliker-Fuse nucleus is considered to be part of the parabrachial complex (Saper, 1995), while medial parabrachial and cuneiform nuclei apparently receive only a sparse input from the superficial dorsal horn (Bernard et al., 1995;Feil and Herbert, 1995).
It is very unlikely that CTb was taken up by spinothalamic axons in experiments PAG1-4, since these axons are located at a considerable distance lateral to the PAG. Although there was spread of tracer dorsally into the overlying part of the superior colliculus in PAG1 and PAG3, this is unlikely to have affected the results, as there is no evidence of a significant projection from superficial dorsal horn to this region (Beitz, 1982;Mené trey et al., 1982). Gauriau and Bernard (2004a)

Differences in lamina I projections between cervical and lumbar enlargements
The results of the present study suggest that, compared with L4, C7 contains fewer spinoparabrachial cells, a similar number of spino-PAG cells, and many more spinothalamic cells in lamina I. The number of lamina I spinothalamic neurons has been found to be higher in cervical than in lumbar enlargements in rat, cat, and monkey (Kevetter and Willis, 1983;Harmann et al., 1988;Burstein et al., 1990a;Dado and Giesler, 1990;Zhang et al., 1996;Zhang and Craig, 1997;Kobayashi, 1998;Klop et al., 2005;Al-Khater et al., 2008). In part, this presumably reflects the greater representation of forelimb compared with hindlimb in somatosensory cortex (Kaas, 1983;Remple et al., 2003). However, the difference appears to be considerably greater in the rat, since Zhang and Craig (1997) found approximately twice as many lamina I spinothalamic cells in C6 -8 as in L5-7 in the monkey, and the present results indicate that the difference between C7 and L4 in rat is more than fourfold. The difference between cervical and lumbar enlargements may be less dramatic for neurons that project to the ventral posterolateral (VPL) nucleus and thus convey information to primary somatosensory (S1) cortex, since some lamina I spinothalamic neurons in rat cervical enlargement project only to PoT or surrounding regions (Zhang and Giesler, 2005), while it is not known whether this is the case for those in the lumbar enlargement. Zhang et al. (2006) reported that cervical lamina I neurons projecting to VPL often had large receptive fields, covering several digits and extending proximally on the limb. If this is the case for lumbar cells, it may explain why so few are needed to provide input to S1 cortex from the entire dermatome. Nonetheless, the larger number of spinothalamic lamina I cells in cervical segments presumably results in more accurate stimulus localisation in forelimb compared with hindlimb. Nociceptive information from the hindlimb is also thought to reach the brain through short-fiber multisynaptic pathways (Basbaum, 1973), and these presumably supplement direct pathways, such as the spinothalamic tract.
It is unlikely that we failed to label a significant number of spinoparabrachial cells in C7, because projections from superficial dorsal horn of lumbar and cervical enlargements terminate in similar areas of LPb (Slugg and Light, 1994;Bernard et al., 1995;Feil and Herbert, 1995;Saper, 1995), which were included in our injection sites. In addition, we found that virtually all lamina I spinothalamic neurons in C7 were labelled from LPb. If we had missed a significant number of spinoparabrachial cells, we would have expected a much larger number of cells retrogradely labelled only from thalamus. The lower number of spinoparabrachial cells in C7 is not due to a difference in the size of lamina I, as the mediolateral extent of the dorsal horn is similar in C7 and L4. The difference may reflect relative sizes of dermatomes, because the L4 dermatome is considerably larger than that of C7 (Takahashi and Nakajima, 1996). The main outputs from regions of LPb innervated by lamina I neurons are the amygdala and hypothalamus, which are thought to play a role in affective and autonomic aspects of pain (Bernard et al., 1996). Although the major targets of PoT are secondary somatosensory (S2) and insular cortices, PoT also projects to the amygdala (Ottersen and Ben Ari, 1979; Gauriau and Bernard, 2004b), which will presumably receive a much smaller input through the spinothalamic tract from the lumbar enlargement. In addition, the number of spinohypothalamic lamina I neurons is considerably lower in lumbar than in cervical segments (Burstein et al., 1990b). The larger number of spinoparabrachial cells in lumbar cord may therefore compensate for the reduced input that this region provides to both amygdala and hypothalamus through spinothalamic and spinohypothalamic tracts. The similarity in numbers of spino-PAG cells in C7 and L4 presumably reflects an equivalent importance of the two segments in activating antinociceptive and other coping mechanisms.
We have previously provided evidence that injection of tracer into LPb labels ϳ85% of all lamina I projection neurons in L4 of the rat and that these include >95% of those that project to PAG (Spike et al., 2003). The lower number of cells labelled from LPb in C7 therefore suggests that there are many fewer lamina I projection neurons per segment in cervical enlargement. Consistently with this, we have found that virtually all spino-PAG neurons in C7 are also labelled from LPb (A.J. Todd, unpublished observations). Although, this cannot be confirmed until numbers and collateral projections of cervical lamina I cells labelled from other targets (e.g., caudal ventrolateral medulla and nucleus of solitary tract) have been determined, the present results suggest that there is a dramatic difference between the two enlargements in the

Collateral projections of spinothalamic neurons
The proportions of spinothalamic cells that were also labelled from LPb in the present study were somewhat higher than those reported by Hylden et al. (1989), presumably because they injected fluorescent beads (which give very restricted injection sites) into LPb and therefore did not label all spinoparabrachial cells. However, Hylden et al. reported that 31% of lumbar spinoparabrachial lamina I neurons projected to thalamus, and this is difficult to explain in light of previous estimates of the numbers of spinothalamic and spinoparabrachial lamina I neurons in the lumbar enlargement (see above). Our finding that relatively few (6%) spinoparabrachial cells in this region project to thalamus is consistent with the report by McMahon and Wall (1985), who examined axons of 13 lamina I projection neurons in rat lumbar enlargement and found that all of them gave collaterals to midbrain (including PAG and an area that corresponds to LPb), whereas none could be activated from more rostral areas.
Previous studies in rat have indicated that some lamina I cells project to both thalamus and PAG. Liu (1986) observed a few double-labelled neurons in lamina I but did not state the proportion of labelled cells that contained both tracers. Harmann et al. (1988) reported that lamina I neurons labelled from both sites made up only ϳ10% of spino-PAG and <5% of spinothalamic neurons. In the present study, we found a much higher degree of double labelling in C7 (58% of spino-PAG cells and 47% of spinothalamic cells) and a higher proportion of spinothalamic cells labelled from PAG (33%) in L4. These discrepancies are probably due to differences in the injection sites.
Our finding that virtually all spinothalamic lamina I neurons were labelled from LPb and that a significant proportion were labelled from PAG suggests that nearly half of the spinothalamic lamina I cells in C7 and almost one-third of those in L4 send collateral branches to both LPb and PAG. This indicates that individual lamina I neurons may contribute to several different functions, including stimulus localization and affective aspects of pain (through projections to thalamus and LPb) as well as activation of descending modulatory influences (through projections to PAG).
We previously reported that 85% of large lamina III/IV NK1rimmunoreactive cells in C6, and 17% of those in L5, were spinothalamic, and these appear to project to PoT but not to VPL (Al-Khater et al., 2008). Our present finding for C7 is very similar to that for C6, although the percentage for L4 (28%) is somewhat higher than we had observed in L5. Most of those in C7 were labelled from both thalamus and LPb, whereas, in L4, approximately half of those projecting to thalamus were labelled from LPb. Our results for both C7 and L4 indicate that PAG is not a major target for axons of these cells. As in our previous study, we found that the proportion belonging to the spinothalamic tract in lumbar enlargement was significantly higher for those in the medial part of the dorsal horn, which receives inputs from the distal part of the hindlimb, and this may reflect the relative importance of distal hindlimb input to S2 cortex.

Differences between projection populations in lamina I
Our results indicate that spinothalamic lamina I cells differed from other neurons labelled from LPb in terms of morphology, soma size, and (in the C8 segment) strength of NK1r expression. Several studies have examined morphology of lamina I spinothalamic cells, but these have not revealed a consistent pattern. For rat, Lima and Coimbra (1988) reported that these cells belonged to pyramidal or flattened (multipolar) classes, whereas Yu et al. (2005) found that fusiform cells were most numerous, particularly in the cervical enlargement. Consistently with our previous study (Spike et al., 2003), we found that the proportion of spinoparabrachial lamina I cells in L3 that belonged to each morphological class was similar, and a comparable result was found in C8. However, spinothalamic lamina I cells differed significantly from other neurons labelled from LPb in that more were multipolar, particularly in the lumbar region, where multipolar cells made up 58% of the population. One reason for the high proportion of multipolar spinothalamic cells in lumbar enlargement is that large gephyrin-coated cells (most of which are multipolar) make up ϳ20% of the lamina I spinothalamic population at this level (Polgá r et al., 2008). These are also present in the cervical enlargement but are greatly outnumbered by other spinothalamic cells and cannot therefore account for the relatively high proportion of multipolar cells in this region. Spinothalamic lamina I cells also had significantly larger somata than other cells labelled from LPb. Again, this is due in part to the presence of large gephyrin-coated neurons within this population, but, when we excluded cells that were negative or very weak for NK1r (which should account for all of the gephyrin-coated cells), we still found a significant difference in soma size. Electrophysiological studies in the rat suggest that spinothalamic lamina I neurons generally have larger receptive fields than those of spinoparabrachial neurons ( , who analyzed 53 spinoparabrachial neurons in lumbar enlargement, reported that these had receptive fields restricted to one or two digits, and the present results suggest that few of these would have belonged to the spinothalamic tract. Because primary afferent input to the dorsal horn is somatotopically organized, receptive field size of lamina I neurons is probably correlated with the extent of their dendritic trees. It is therefore likely that the difference in cell body size is at least partially accounted for by the need for spinothalamic neurons to support larger dendritic trees. In C8, the strength of NK1r immunoreactivity was significantly higher among spinothalamic neurons, and we have previously observed that lumbar lamina I cells with strong NK1r immunoreactivity are under-represented among the population that projects to PAG (Spike et al., 2003). Taken together, these results suggest that subpopulations of lamina I cells differ in their pattern of supraspinal projection. Han et al. (1998) reported that there was a close relationship between morphology and function for lamina I neurons in cat spinal cord, with pyramidal cells responding only to innocuous cooling and those in the other two morphological classes being activated by noxious stimuli. Consistently with this suggestion, Yu et al. (2005) and Almarestani et al. (2007) reported that pyramidal cells projecting to thalamus or LPb in the rat were seldom NK1r immunoreactive. However, although cooling-sensitive lamina I spinothalamic neurons have been identified in the rat, these cells also responded to noxious mechanical stimuli (Zhang et al., 2006). If these are included among the pyramidal class, they may correspond to some of the NK1r-positive pyramidal spinothalamic neurons seen in the present study. However, it is unlikely that all pyramidal lamina I neurons in the rat respond to innocuous cooling, as 23-30% of lumbar spinoparabrachial cells are pyramidal (Spike et al., 2003;Almarestani et al., 2007;present study), and it has been reported that lamina I spinoparabrachial neurons are not activated by innocuous cooling (Bester et al., 2000). Nonetheless, we have observed functional differences related to morphology in lamina I of rat spinal cord; NK1rpositive multipolar projection neurons were more likely to express Fos in response to noxious cold stimuli than those belonging to other morphological classes (Todd et al., 2005). However, we have also found differences among multipolar lamina I cells: