Projections of Lamprey Spinal Neurons Determined by the Retrograde Axonal Transport of Horseradish Peroxidase D. TANG A N D M. E. SELZER School of Medicine, University of Pennsyluania, Philadelphia, Pennsylvania 19104

ABSTRACT

The spinal cords of larval sea lampreys (Petromyzon marinus) and adult river lampreys (Ichthyomyzon unicuspis) were injected with horseradish peroxidase through a transection 1 cm caudal to the last gill. Some animals also had a spinal hemisection 1 cm caudal to the injection. After recovery periods of 1 to 52 days, the spinal cords were treated with diaminobenzidine and hydrogen peroxide, and the projections of various cell types determined in wholemount slides. From these observations the following conclusions were drawn. Most dorsal cells (primary sensory cells) are bipolar with a long rostral projection and a short caudal projection of no more than 5-10 mm. Both processes travel in the ipsilateral dorsal column. Their peripheral processes enter the dorsal roots as branches of their central axons. Some dorsal cells send processes out three or more dorsal roots both rostral and caudal to the cell body. Myotomal motoneurons have characteristic locations in the medial gray column and send prominent transversely oriented dendrites into the lateral columns. A few motoneurons are unusually large. In addition to giant interneurons the majority of smaller rostrally projecting interneurons also have decussating axons. A recently described cell type, t h e oblique bipolar cell, appears to have a n exclusively crossed rostral projection. Although most edge cells project rostrally, as many as 20%may have a caudal projection or both rostral and caudal projections. Edge cells project equally to the ipsilateral and contralateral spinal hemicord, but their processes do not extend more than about 18 mm in sea lamprey larvae and 37 mm in adult river lampreys. Lateral cells project exclusively to the ipsilateral caudal hemicord. A few cells which resemble lateral cells in location and in possessing large lateral dendrites, project rostrally. However, these have atypical morphologic features which probably distinguish them from true lateral cells. Thus far, regardless of cell type, all decussating axons seem to pass ventral to the central canal, while decussating medial dendrites pass dorsally.

The spinal cord of the sea lamprey contains several types of large neurons which are distinct enough in shape and location to be identified in the living isolated cord (Rovainen, '67b, '74). This feature makes the lamprey spinal cord a favorable preparation for the study of neuronal organization and synaptic transmission in the vertebrate central nervous system. In the present study we have combined the technique of the retrograde transport of horseradish peroxidase (HRP) (LaVail et al., '73) with a method of examinJ. COMP. NEUR. (1979)188: 629-646.

ing the lamprey spinal cord in wholemount (Selzer, '791, in order to define further the axonal projections and cellular morphologies of several types of spinal neurons. METHODS

Injection of HRP Larval sea lampreys (Petromyzon marinus) 9-14cm long and adult river lampreys (Zchthyomyzon unicuspis) were anesthetized in tricaine methanesulfonate (MS 222 Finquel, Ayerst) 1:4,000, pinned in a dissecting dish and

629

630

D. TANG AND M. E. SELZER

packed on the sides in ice. The spinal cord was exposed through a dorsal incision about 1 cm caudal to the last gill, and completely transected under microscopic vision. One microliter of either a 10%or 20% solution of HRP (type VI, Sigma) in a physiologic lamprey solution (Selzer, '78) was injected into the transection through a Hamilton Syringe. The wound was allowed to air dry a t 4°C for 3 hours, during which the animals remained alive and had slow respiratory contractions of the gills. The animals were then transferred to aerated fresh water tanks a t 12"C, where they were kept in the dark without feeding for 1to 52 days. At the time of HRP injection some animals had a spinal cord hemisection through a second incision 1 cm caudal to the injection site. This procedure should limit labeling caudal to the hemisection to those neurons which have crossed rostral axonal projections. Contralatera1 to the hemisection neurons would be stained only if they had uncrossed rostral projections. The accuracy of the hemisections varied by as much as one-third the width of a hemicord. Thus the medial extent of the cuts ranged from just lateral to the medial border of the ipsilateral gray column to just lateral to the medial border of the contralateral gray column. Accuracy was generally better in adult animals because their spinal cords are much wider than those of larvae. These variations were taken into account in interpreting the results of HRP injections. Wherever possible conclusions were drawn from the most accurately hemisected specimens. While transected spinal cords of larval lampreys are capable of functional regeneration, it is unlikely that axons could have regenerated through the hemisection as far as the HRP injection site even after 52 days (Selzer, '78). In still other animals no spinal injection was made, but motoneurons were labeled by injection of 0.5-2 p1 of a 10%solution of HRP into segmental muscles, or into the base of the dorsal fin. Their spinal cords were examined for HRP labeling a t 12 days. Occasionally a dorsal cell (primary sensory neuron) was also labeled by these injections.

Histology Cells labeled with HRP were examined in spinal cord wholemounts. The brain and spinal cord were exposed ventrally by removal of the underlying part of the cartilaginous brain case and the attached notocord as described

previously (Selzer, '79). The exposed spinal cord was incubated a t room temperature for 10 minutes in a 0.1% solution of collagenase (Sigma type I) in lamprey saline to facilitate the later removal of the meninx primitiva from the spinal cord. This improved the final appearance of the wholemount by removing the meningeal blood vessels, which contain endogenous peroxidase activity, and possibly also by improving penetration by the diaminobenzidine histochemical reagent. Following collagenase treatment, the tissue was fixed' for 90 minutes a t 4°C with 2%glutaraldehyde in 0.1 M phosphate buffer, pH 7.4. Fixative was removed by three 20-minute washings in cold lamprey solution. During the final washing, the ventral part of the meninx was gently stripped away from the brain and spinal cord, which were then removed from the spinal canal and brain case, and pinned, dorsal side up, to a strip of Sylgard (Dow Chemical). The dorsal meningeal covering was then also stripped away, and the isolated neuraxis was washed overnight in lamprey solution at 4°C. In most cases, especially in larvae, it was not possible to remove completely the dorsal meninx without excessive damage to the cord. HRP was demonstrated by incubation a t room temperature with a 0.1% solution of 3,3I-diaminobenzidine hydrochloride (DAB) in 0.1 M phosphate buffer, pH 7.4.After 10 minutes a 30% H,O, solution was added to make a final H,O, concentration of 0.3%.After 15 minutes the incubation solution of DAB and H,O, was changed, and the tissue incubated for an additional 15 minutes. At this time the tissue was visibly tinted with brown reaction product, which was especially intense a t the injection site. The stained tissue was rinsed twice for 10 minutes in cold lamprey solution, dehydrated with 5-minute changes each of 70%,95%,and 100% (3 changes) ethanol. The tissue was cleared for 30 minutes in methylbenzoate and permanently mounted on glass slides with Permount (Fisher) for bright-field and darkfield microscopy. Despite the thickness of the spinal cord (about 250 p after dehydration), the wholemounting had little effect on the integrity of the neurons, as judged by comparing these wholemounts to short pieces of HRPstained spinal cord examined in a well slide. RESULTS

General characteristics of staining Two types of staining were observed following injection of HRP into a spinal cord tran-

LAMPREY SPINAL NEURONS

section. The first was a homogeneous distribution of reaction product within neurons (e.g., figs. 1B,E) and axons up to 2 cm from the injection site. This presumably reflected the intra-axonal diffusion of HRP. Large axons were labeled for equal distances in either direction (fig. 1B). These were the Muller and Mauthner fibers whose cell bodies lie in the reticular formation in the floor of the third and fourth ventricles and the intervening isthmus (Rovainen, '67a; Rovainen et al., '73). This type of staining was maximally intense in giant axons the first two days following injection. The stain spread over longer distances and became less intense a t the injection site during the next seven to ten days. After this time the injection site became progressively paler, without any additional increase in range of axonal staining. Swelling and closing of large axons adjacent to the transection site could be seen in less than two days. However, unequivocal evidence of Wallerian degeneration was not observed until about 12 days. This took the form of beadlike swellings and narrowings in smaller and medium caliber axons, especially in the lateral columns caudal to the transection (fig. 1A) and in the dorsal columns rostrally. In the giant reticulospinal axons (Muller and Mauthner axons) degenerative changes seemed to take longer to appear. Distal segments became vacuolated and showed more gradual tapering and widening than did smaller axons. HRP-filled segments of giant axons were seen caudal to the injection as late as five weeks after transection. In some preparations these large axons could be traced rostrally to their cell bodies in the brain. Observations on such preparations demonstrated that over lengths of several centimeters these axons are unbranched, confirming the observations of Rovainen et al. ('73) in serially sectioned specimens. This is consistent with the observations of Bertolini ('64) and of Smith et al. ('70)that synapses are formed by these axons en passage. The second type of staining consisted of grains of HRP (fig. 1D) which probably had been transported retrogradely from nerve terminals. The grains were often best seen in dark field (figs. 4A,E). The earliest labeling occurred a t three days and was seen as small collections of grains clustered near the nuclei of giant interneurons. Other cells, such as dorsal cells, began to show labeling about two days later. A peculiar characteristic of this labeling was that neurons a t all spinal segments

631

were labeled simultaneously, and this labeling increased in density over the next three weeks. Thus it was not possible to calculate a rate of retrograde transport based on a progressive increase in the maximum distance from the injection site a t which cells were labeled. The density of staining was quite variable from animal to animal, even a t identical recovery times. Therefore the above observations are qualitative and were made from selected animals with the best labeling. They were not quantified further by counting numbers of grains per cell, or counting the number of cells labeled per spinal segment. Labeling and morphological characteristics of individual neuron types Dorsal cells These are primary sensory neurons (Rovainen, '67b; Martin and Wickelgren, '71) with round, most often bipolar shapes (fig. lB), which send processes into the ipsilateral dorsal columns both rostrally and caudally. In sea lampreys each spinal cord contains about 1,800 dorsal cells (Selzer, '79) arranged in irregular rows on either side of the central canal. Dorsal cells were labeled with HRP both by diffusion (fig. 1B) and by retrograde axonal transport (fig. 1D).An interesting feature of these cells is that some of them send processes out more than one dorsal root both rostra1 and caudal to the soma (fig. 1C).This is consistent with the findings of Rovainen ('67b) that occasionally these neurons could be activated by electrical stimulation of either of two dorsal roots. Martin and Wickelgren ('71) have shown that a few of these neurons have two separate cutaneous receptive fields. We traced processes of single dorsal cells to as many as three dorsal roots. The dorsal root processes of dorsal cells generally arose as branches of the axons rather than directly from the soma. While most axon branches of dorsal cells projected to the ipsilateral half of the spinal cord (fig. lB), occasionally a branch could be seen crossing the midline. This may account for the fact that in accurately hemisected specimens, many dorsal cells in the more caudal spinal segments were labeled with HRP ipsilateral to the hemisection, even though no dorsal cells were so labeled in segments of spinal cord immediately caudal to the hemisection. The most accurately hemisected specimen was an adult river lamprey which had no dorsal cells stained ipsilaterally in the first 15.5mm caudal to the hemisection.

632

D. TANG AND M. E.

SELZER

Fig. 1 Sensory and motor neurons labeled with HRP. A. Beaded appearance of HRP-filled axon undergoing Wallerian degeneration in the lateral fasciculus of 1. unicuspis 22 days after transection. Beading is produced by areas of swelling and narrowing compared with the size of the parent axon (not shown). B. Dorsal cell (dc) in hemisected P. marinus stained by diffusion. The axon of another dorsal cell (arrow) is also seen, and demonstrates t h e appearance of axon branching. The large dark longitudinal profile at the bottom (and others not in focus) is a Miiller fiber filled by diffusion. Note t h e absence of labeled dorsal cells or giant axons a t t h e tip of t h e figure, which is ipsilateral to the hemisection. C. Dorsal cell axons (arrows) in P. marinus stained by spinal cord injection of HRP seen leaving a dorsal root (dr). One axon branch is rostra1 to i t s soma while the other leaves caudal to its soma. D. Dorsal cells (dc) labeled with grains of HRP. E. Motoneurons labeled by diffusion in P. marinus. The axons of many of these can be traced to the ventral root (vr) when the tissue is examined with the microscope, although only a few are apparent in this montage. The large neuron labeled mn sends i t s axon out t h e ventral root, and therefore is also presumed to be a motoneuron. Scale is 50 p in all but C.

LAMPREY SPINAL NEURONS

The average length of a segment in this animal was 2.5 mm. Thus no cells were labeled in the first six to seven ipsilateral segments. Ten dorsal cells were labeled in the first six to seven contralateral segments. For purposes of comparison the spinal cord caudal to the hemisection was divided into three 45 mm lengths. In the rostralmost third only 4 of 28 labeled dorsal cells (14.3% were ipsilateral to the hemisection. In the middle third 35 of 80 (43.8%) were ipsilateral. In the caudalmost third 20 of 56 labeled dorsal cells (35.7%)were ipsilateral. Since most dorsal cell outlines can be distinguished in unstained wholemounts by the variations in the refractile properties of the cord it was possible to estimate the proportions of dorsal cells which were labeled with HRP. While the percentage of stained cells varied greatly from preparation to preparation, in a few favorable preparations of larval sea lamprey cords almost all dorsal cells caudal t o an injection were labeled. Figure 2A is a diagram illustrating the mean dorsal cell densities along the length of the spinal cord of a sea lamprey larva. The mean cell densities caudal to the injection are fairly uniform for different lengths of 5 mm and average 15.7/mm. Since in this animal the average length of a segment was about 0.96 mm this comes to 16.4 dorsal cells per segment or 8.2 per unilateral segment. This is close to the average value of 7.9 dorsal cells per unilateral segment observed in toluidine blue stained wholemounts, and 8.9 counted in serial plastic sections (Selzer, '79). In figure 2B the dorsal cell densities are recorded for a well injected adult river lamprey, with an average segment length of 2.64 mm. The mean cell density caudal to the injection is 2.0/mm or 2.6 dorsal cells per unilateral segment. We counted dorsal cells in a toluidine blue stained wholemount of an Zchthyomyzon spinal cord and found an average of about 8.5 per unilateral segment. Thus in the specimen of figure 2B only about a third of dorsal cells caudal to the injection were visibly labeled. The reason for this is not clear since, as in P. marinus, the majority of dorsal cells in more caudal segments of 1. unicuspis can be antridromically activated by stimulating the rostral cord (see below). Labeling of dorsal cells rostral t o an injection was more restricted than caudal to the injection. In figure 2A for the first 5 mm rostral to the injection the density of labeled dorsal cells was 13.4/mm, or 85%of the average

633

dorsal cell density caudal to the transection. In the second 5 mm the dorsal cell density was only 7.2/mm, and no labeled dorsal cells were observed beyond 10 mm. In fact the average maximal distance a t which dorsal cells were labeled rostrally was 5.2 mm in 11 larvae. In the adult river lamprey of figure 2B the most rostral labeled dorsal cell was 13 mm from the injection. It would seem from the above data that dorsal cells have very long rostrally projecting axons, which travel a t least to the rostral 2 cm of the cord. This is consistent with previously reported physiologic data in which 75-90X of dorsal cells could be antidromically activated by stimulating the rostral spinal cord in Petromyzon (Rovainen, '67b; Selzer, '78) or the brainstem in Larnpetru (Martin and Bowsher, '77). We have repeated these observations in sea lamprey larvae placing stimulating electrodes a t the level of the last gill so as to include the territory encompassed by the caudal segments in the present study; 87.3%of 55 dorsal cells caudal to the stimulating electrode could be antidromically activated. Twenty-four of twenty-six dorsal cells (92.3%) within 3 cm of the stimulating electrode and 24 of 29 (82.8%)between 3 and 6 cm could be activated antidromically. For technical reasons the proportion of dorsal cells with long rostral projections may be even greater. (We have not made a systematic study of the proportion of dorsal cells which can be antidromically activated in Zchthyomyzon. However, it is our impression that the proportion is similar t o that in Petromyzon.) The majority of dorsal cells would also seem to be bipolar, rather than unipolar, with relatively short (6 mm or less) caudal processes. Motoneurons Within one or two mm of an HRP injection into a spinal transection many small neurons, whose elongate cell bodies were located mostly in the medial portion of the gray column, were stained by diffusion. They had prominent long lateral dendrites and shorter medial dendrites oriented transversely t o the longitudinal axis of the spinal cord. Many of them had axons which could be traced out nearby ventral roots (fig. lE), and were therefore probably motoneurons. Their appearance in P. marinus was similar to that of two fast green injected physiologically identified body wall motoneurons illustrated by Rovainen and Birnberger ('71). A few of these cells were considerably larger

634

D. TANG AND M. E. SELZER

-

5 mm

Fig. 2 Mean bilateral densities of HRP-labeled dorsal cells listed as cells per mm in 5 mm lengths along the spinal cord. A. Larval sea lamprey; average segment length 0.96 mm. B. Adult river lamprey; average segment length 2.64 mm. Dorsal cell densities for both species average 16-17 per bilateral segment. In the specimen of B only about one-third of visible dorsal cells caudal to the transection are labeled, while most of these cells are labeled in the somewhat better stained preparation in A. This does not represent a species difference in the degree of rostral projection of dorsal cells since in both species the majority of dorsal cells can be activated antidromically by stimulating the rostral cord. Note that cells are not labeled more than 1 cm rostral to the injection in A and 1.5 cm in B, suggesting that caudal processes of dorsal cells are short in comparison with rostral processes.

than the rest (cell labeled mn in fig. 1E).Interestingly Tretjakoff (’09)believed that each ventral root contained one particularly large axon, and called some large neurons motoneurons. Injections of HRP into body wall muscle and base of the dorsal fin resulted in labeling of myotomal and fin motoneurons by retrograde transport. However, the degree of morphologic detail achieved was insufficient to reveal consistent differences between these two types of motoneurons, which were likened by Rovainen and Birnberger (‘71) to the type I and type I1 motoneurons described by Tretjakoff (‘09).The level of intraspinal injections in the present study was rostral to the dorsal fins. The fact that all of the many motoneurons stained in this way resemble the type I

motoneurons supports the conclusion of Rovainen and Birnberger that the latter are myotomal motoneurons. In adult river lampreys motoneurons were less surely identified near the site of spinal cord injection than those in sea lamprey larvae because it was more difficult to follow their axons out the ventral roots. Nevertheless motoneurons could still be identified by retrograde transport after injection of HRP into body wall muscle or into the base of the dorsal fin (fig. 3). They were somewhat larger and less transversely oriented than those in the larvae. However their somata were located mostly in the medial half of the gray columns and their main dendrites extended laterally. As in larval P. marinus, spinal cords of adult I.

Fig. 3 Motoneurons in I. unzcuspis labeled with reaction product following intramuscular injections of HRP. A. An exceptionally large myotomal motoneuron located between the gill region and the rostra1 limit of the dorsal fin, traced with a Nikon drawing attachement. The axon (arrow) can be followed out the ventral root. Insert is a photograph of the same cell a t a single focal plane. B. A more typical myotomal motoneuron in the same part of the spinal cord. C. A myotomal motoneuron in the middle of the region of the dorsal fin (caudal half of body). D.E. Fin motoneurons, located about 2 cm rostral to the cell in C. All scales 50 p . lat indicates lateral.

636

D. TANG AND M. E. SELZER

Fig. 4 Rostrally projecting neuron. A. A giant interneuron (pi) ipsilateral to a rostral hemisection in P. marinus is labeled with grains of HRP and photographed in dark field. A neighboring dorsal cell (dc) is also labeled, as are two smaller rostrally projecting neurons (rn), one just to the left of the giant interneuron and one on the contralateral hemicord (bottom). Rostral is left. B. Outline of an unstained giant interneuron can be seen (arrows) in the contralateral hemicord of the same animal as in A. This shows t h a t giant interneurons (pi) have crossed rostral projections. C. A rostrally projecting trident-shaped cell (tc) located about 5 mm caudal to an HRP injection i n I. unacuspis. Note t h e characteristic decussating medial dendrite (top), and t h e nearby dorsal cell (dc) slightly out of focus. Rostral is left. D. A rostrally projecting interneuron (rn) in P. marinus which is somewhat smaller and located more rostrally than typical giant interneurons. However, with careful focusing, its axon (arrow) can be traced rostrally across the midline ventral to the central canal, and i t s large medial dendrite can also be followed across the dorsal midline. Thus t h e distinction between this cell and giant interneurons is tenuous. cc, central canal. E. An oblique bipolar cell (ob) in P. marinus. Rostral is right. Scale 50 p except in E.

unicuspis contained a few particularly large myotomal motoneurons. In figure 3A one such motoneuron was densely filled with HRP by diffusion through its large axon, which could be followed into the ventral root. This was rather unusual, and most motoneurons were stained with grains of reaction product. In these preparations we were again unable to detect consistent differences between myotomal (figs. 3B,C) and fin (figs. 3D,E) motoneurons.

Giant interneurons These are second order sensory cells (Rovainen, '74) which are located in the caudal half of the spinal cord and are characterized morphologically by their large size, their large medial dendrites and their crossed rostral projecting axons (Rovainen, '67b). In seven HRPinjected larvae without hemisection the mean number of labeled giant interneurons was 17.6 -t 1.9 SD. This is similar to counts of giant in-

637

LAMPREY SPINAL NEURONS TABLE 1

Rostralprojections of neurons in hemisected spinal cords. Numbers are cell counts from the one best hemisected andstained larva (P.marinus) and adult (I. unicuspis) Axon projection Cell type Uncrossed

Giant interneuron P. marinus I. unieuspis

Crossed

3

8

3

9

Rostrally projecting interneurons P. marinus 1. unicuspis

65 69

301

Edge cells P. marinus I. unicuspis

24 17

142

32

18

Total

2, crossed

11 12

73 15

207 370

69 81

56 35

57 51

Cells are listed a s projecting contralaterally if they are located caudal and ipsilateral to the hemisection and vice versa. The percent crossed represents the percentage of counted cells which were ipsilateral to the hemisection. and probably underestimates thc actual proportion of all such neurons which have crossed projections since many neurons which have both crossed and uncrossed projections would be counted under the uncrossed column

terneurons made in toluidine blue stained ing interneurons contralateral to the hemisecwholemounts (Selzer, '79). Their appearance tion was greatest for the most caudal part of in HRP-injected wholemounts is shown in the spinal cord and lowest immediately caudal figure 4A. Because of their large size giant in- to the hemisection. For example, in the river terneurons, like dorsal cells, could be seen lamprey specimen cited above, only 25 of 259 even without HRP labeling. An example of an labeled interneurons (9.7%)in the first 45 mm unlabeled giant interneuron contralateral to a of spinal cord caudal to the hemisection were hemisection is seen in figure 4B. However, in contralateral (fig. 5B). In the next 45 mm, 22 the most accurately hemisected sea lamprey of 70 labeled neurons (31.4%)were contralatlarvae and river lamprey adults some giant in- eral. In the caudalmost 45 mm, 22 of 41 laterneurons were labeled contralateral to the beled neurons (53.7%) were contralateral. hemisection (table 1). All ipsilateral giant in- These data also show that, unlike dorsal cells, terneurons seemed to have been stained. Thus total numbers of labeled rostrally projecting while all giant interneurons project to the interneurons decreased greatly with distance contralateral rostra1 cord, some may also pro- from the injection. This suggests that many ject ipsilaterally. Alternate interpretations rostrally projecting interneurons may have relatively short axons. Nevertheless some of must also be considered (DISCUSSION). these neurons were fairly large and, while not Other rostrally projecting interneurons as large as conventional giant interneurons, In addition to giant interneurons, many did have prominent medial dendrites and smaller neurons were labeled caudal to an crossed rostrally projecting axons which could HRP injection. In the larval spinal cord with be followed for long distances when filled by the best hemisection and with the best uptake diffusion with HRP. An example of such a cell of HRP 207 such cells were counted, of which is seen in figure 4D. As with giant inter65 were located contralateral to the hemisec- neurons, these axons often traveled long distion and 142 ipsilateral. Thus a t least 69%of tances without branching. Also as with giant the rostrally projecting neurons would seem interneurons, whenever they could be followed to project contralaterally. In the best river across the midline, medial dendrites passed lamprey hemisection 301 of 370 rostrally pro- dorsal to the central canal and axons ventral jecting neurons were ipsilateral to the hemi- to the canal. However, unlike giant intersection. Thus a t least 81%of these cells would neurons they were not limited in location t o seem to project contralaterally (figs. 4D, 5; the caudal half of the spinal cord. Among the many morphological types of table 1). The percentage of rostrally project-

638

D. TANG AND M. E. SELZER

neurons observed in toluidine blue wholemounts, two were previously mentioned as having distinctive shapes. These were called trident-shaped cells and obliquely oriented bipolar cells (Selzer, '79). Trident cells (fig. 4C) are found mostly in the rostral half of the spinal cord and were described as triangular in shape having three prominent dendrites or processes arising from the medial base of the cell. While cells of this description were seen prominently in toluidine blue stained wholemounts of adult river lampreys, they were less clearly identifiable in sea lamprey larvae or adults. In HRP injected river lampreys they were less clearly divisible from a group of prominent neurons, generally triangular in appearance with the medial base giving rise to either two or three processes. One was a medial dendrite which always crossed the midline dorsal to the central canal. The others pointed rostrally and/or caudally. These cells were most often found caudal to an injection. Thus their axons project rostrally. Since their axons are narrow they were not followed for long distances, and it is not clear whether the cells project ipsilaterally, contralaterally, or both. Neither was it possible to identify clearly trident cells among profiles of neurons filled with grains of HRP, rather than by diffusion. Thus it is not clear how far rostrally their axons project. Trident cells are clearly not motoneurons since their processes did not enter ventral roots. Thus they would seem to fall into a general category of rostrally projecting interneurons, although they are far from identified as a distinct cell type. Oblique bipolar cells (fig. 4E) on the other hand, were morphologically much more homogeneous. They were clearly identified, even when filled only by grains of HRP, as elongate, spindle-shaped neurons located in the most medial parts of the gray matter at all levels of the spinal cord. The more medial process (dendrite) pointed either rostrally or caudally. When the medial dendrite crossed the midline it passed dorsal t o the central canal. These cells invariably projected rostrally for long distances in the contralateral hemicord, although their axons were not identified. They seemed not t o have ipsilateral rostral projections (fig. 5B). While their function is still unknown their characteristic shape and crossed rostral projection make them attractive candidates for identification as a distinct cell type. Edge cells These cells, previously described as large

rostrally projecting interneurons located lateral to the spinal gray, have been only partly characterized physiologically (Rovainen, '74). Recently they have been shown to be quite heterogeneous in size, shape and position within the lateral white matter (Selzer, '79). About 80'X of edge cells counted in sea lamprey larvae and 75% in river lampreys were caudal to the injection and thus projected rostrally (figs. 6A,C,D; table 2). However, some edge cells did project caudally, and in many instances an axon could be seen dividing into a caudal and a rostrally projecting branch. Caudally projecting edge cells have not been found by physiologic methods (Rovainen, '741, but in several instances the caudal projection was confirmed by following the axon in diffusion filled HRP-stained edge cells (fig. 6B). In such cells it was also possible to see both ipsilateral and contralateral projections. Edge cells labeled caudal to a hemisection were about evenly distributed ipsilateral and contralateral to the hemisection in both P. marinus and I. unicuspis (table 1).These findings are consistent with previous physiological findings that edge cells were about equally divided between those which could be activated antidromically by stimulating either the ipsilateral or contralateral rostral hemicord (Rovainen, '74). Also consistent with Rovainen's physiological findings is the fact that edge cells were not labeled further than 22 mm caudal to an HRP injection in larvae (mean furthest distance measured in 6 larvae with average cord lengths of 96 mm was 18 mm), and 37 mm in a 185 mm adult river lamprey cord. Thus edge cells seem to have only short rostral projections. In every instance in which an edge cell axon could be traced crossing the midline it passed ventral to the central canal, close to the ventral surface of the cord. Lateral cells These large neurons of the rostral half of the spinal cord are characterized by their lateral position in the spinal gray and their largc lateral dendrite extending to the edge of the spinal cord. Rovainen ('74) found that most of these could be activated antidromically by stimulation of the ipsilateral caudal hemicord. It was suggested that those few with rostral projections might be a different cell type. The appearance of typical HRP-stained lateral cells is illustrated in figure 7A for sea lamprey larvae and figure 7B for adult river lampreys. The appearance was quite stereotyped, and the caudally projecting axon

Fig. 5 Hemisected spinal cord in I. unicuspis. A. Appearance of the spinal cord a t the site of a hemisection 16 mm caudal to an HRP injection, 22 days after injection. Rostral is right. B. Higher magnification of the same spinal cord 4 mm caudal to the hemisection. Note the presence of labeled rostrally projecting neurons ( r d , one of which is a n oblique bipolar cell (ob), ipsilateral to t h e hemisection, suggesting t h a t they have decussating axons. There are no such neurons in t h e contralateral hemicord, which does have many large evenly stained (by intra-axonal diffusion) axons. Dotted line shows location of central canal. Rostral is right.

640

D. TANG AND M. E. SELZER

Fig. 6 Edge cells. A. A rostrally projecting edge cell (ec) in P. marinus. Its axon can be followed (arrows) as it crosses t h e midline and turns rostrally. Rostra1 is left. B. Less typically an edge cell (ec) has a caudally and medially directed axon (arrows). This axon crossed the midline and could be followed for 4 mm to the area of injection. I t also gave off a very fine branch which was followed rostrally for 2 mm (not shown). A lateral cell (lc) and other caudally projecting neurons (cn) are also labeled. I. unicuspis; rostral is left. C. Appearance of labeled edge cell (ec) in dark field. Lateral dendritic brush a t top extends to the edge of the spinal cord. P. marinus. D. A rostrally projecting edge cell (ec) in 1. unicuspis.

could often be followed for long distances in the ipsilateral lateral white column. None crossed the midline. In seven favorably injected larvae 88 cells were counted which were large, laterally located and had prominent lateral dendrites. Of these, 83 (94%)were rostral to the injection. Thus their axons projected caudally. In a favorably stained adult river lamprey 12 of 16 (75%)of such cells projected caudally (table 2). However, i t was noticed that none of the rostrally projecting cells in either the larvae or the river lamprey, had the stereotyped appearance shown in figures 7Aand B. Instead they had a variety of morphological peculiarities such as a flattened medial margin in the soma and an extra wide lateral dendrite, giving a triangular appearance (fig. 7C) or a double lateral dendrite (fig. 7D). Moreover, these cells were slightly smaller than most of the caudally projecting

lateral cells, and their axons could not be traced for long distances, suggesting that they may be of small caliber or branched. Thus the rostrally projecting "lateral cells" may indeed be a separate species of cell from the true lateral cell. DISCUSSION

The retrograde transport of HRP has been widely used to map the locations of afferent neurons to many parts of the vertebrate nervous system (reviewed by LaVail, '75), though less widely in non-mammalian vertebrates including birds (LaVail and LaVail, '74; Karten and Finger, '76; Streit and Reubi, '77), reptiles (Donkelaar, '76), amphibians (Mensah, '74), and teleost fish (Finger, '75, '76; Luiten, '75; Karten and Finger, '76; Ito and Kishida, '77). In cyclostomes its use has been confined to the demonstration of cranial motoneurons in

LAMPREY SPINAL NEURONS

64 1

642

D. TANG AND M. E. SELZER TABLE 2

Rostral and caudal projections of edge cells and lateral cells counted in s e w n larval and one adult (I. unicusDis) 1amDrevsoinal cord

(P.m a r i n u s )

Axon projection Cell type

Edge cell P. marinus I. unicuspis Lateral cell P. marinus I. unicuspis

Rostral

Caudal

Total

41 rostral

355 86

87 29

442 115

80 75

83 12

(88) (16)

(5) (4)

(6) (25)

Cells labeled caudal to the HRP injection were listed a s projecting rostrally and vice versa. Rostrally projecting lateral cells had atypical morphological features (see text), and are listed in parentheses. Only cells within 10 mm of the injection were counted in order to avoid ambiguity in identification ofcells (especially lateral cells) which were stained only by grains of HRP and not also by diffusion.

lampreys (Homma, '78; Finger and Rovainen, '78). The application of the retrograde transport technique to a wholemount preparation of lamprey spinal cord allows sampling of large numbers of cells and the identification of cellular morphological details which would not be apparent in cross sections. Thus the atypical morphologies of rostrally projecting neurons which look like lateral cells in the living transilluminated cord and in cross sections (Rovainen, '74) were detected only with the HRP-stained wholemount technique. It has yet to be established whether these atypical, rostrally projecting lateral cells have patterns of synaptic activation, which differ from those of the more typical caudally projecting ones. The wholemount technique has also allowed the visualization of some axons which, because of their small calibers or tortuous paths, could not easily be followed in serial sections. Thus the staining of some cells with HRP close to the injection site might not necessarily represent uptake of HRP by their cut intraspinal axons. HRP-stained motoneurons might be confused with rostrally projecting neurons if viewed in cross-section. However, the wholemount technique facilitates the identification of their axons entering the ventral root, and suggests that they were stained by uptake of HRP through damaged ventral roots or motor nerves near the injection site, or through non-axonal uptake. Several reports have suggested that HRP may be taken up by dendrites or soma and transported anterogradely (Lynch et al., '73, '74; Scalia and Colman, '73; Reperant, '75). Similarly the staining of edge cells rostral to an injection might not necessarily imply a caudal projection.

However, the ability to follow the axon for several mm in its caudal projection removes any doubt that some edge cells do project caudally. In addition the presence of grains of HRP within edge cells rostral to an injection suggests that they were stained by retrograde axonal transport. Perhaps the most striking finding of the present study is the fact that the majority of rostrally projecting interneurons have crossed axons. When the larval spinal cord was hemisected caudal to an injection of HRP, 69% of labeled neurons within the spinal gray caudal to the hemisection were ipsilateral to the hemisection, demonstrating that they had crossed rostral projections. In the adult river lamprey 81% of these cells had crossed axons. In fact these figures may underestimate the prevalence of decussation since some or all of the cells labeled contralateral to the hemisection might also have crossed axon branches. These conclusions are based on two assumptions. First that the hemisections were adequate, and second that extracellular diffusion of HRP from the injection site as far as the hemisection wasmot great enough to result in labeling of neurons whose axons had been cut by the hemisection. The validity of these assumptions might be questioned because labeling occurred in some giant interneurons contralateral to the hemisection and in some dorsal cells ipsilaterally. Previous studies have suggested that giant interneurons project only contralaterally and dorsal cells ipsilaterally (Rovainen, '67b, '74; Rovainen et al., '73). However, inadequacy of hemisections cannot explain these observations because the apparently anomalous labeling of giant interneurons and dorsal cells was seen even when

LAMPREY SPINAL NEURONS

the “hemisection” extended to the medial margin of the contralateral gray column. In addition, several observations suggest that extracellular diffusion of HRP did not significantly affect the results: (1)ipsilateral dorsal cells were not labeled in the first several mm caudal to the best hemisections; (2) for rostrally projecting interneurons the percentage of cells within ipsilateral projections increased in the more caudal spinal segments; and (3) stereotypic lateral cells were never labeled caudal to a hemisection. These observations are best explained by the notion that the longer an axon the more the possibility for branching and the greater the probability of bilateral projection. A less attractive explanation would be that dorsal cells and rostrally projecting interneurons in the caudal segments have different patterns of projection from those in the rostral segments. Finally, it is possible that HRP transported anterogradely from cells in the region of the injection found its way into cells which do not project to the injection site. However we are not aware of any evidence to suggest that this can occur to a degree sufficient to result in light microscopic staining. The existence of a gradient of cell sizes among neurons with general appearances similar to giant interneurons and the variability in maps of giant interneurons in larval sea lamprey spinal cords (Selzer, ’79)leaves open the possibility that the giant interneuron is not an absolutely identifiable cell type. It is now apparent that the possession of crossed rostrally projecting axons is not a distinguishing feature of giant interneurons, since the majority of rostrally projecting interneurons have crossed axons. Neither do the dorsal decussation of medial dendrites and ventral decussation of the axons distinguish giant interneurons from other rostrally projecting interneurons. In describing the projections of cell types it should be made clear that the present experiments do not distinguish between neurons which project synaptically to the region of the HRP injection and those whose axons pass through this region without forming synapses. This is because the injection was made through a spinal cord transection, rather than a non-traumatic injection, and it has been shown that cut axons take up and transport HRP (Kristensson and Olsson, ’74; DeVito et al., ’74). Several limitations of this study should be

643

mentioned. First, the injection sites were restricted to the rostral third of the spinal cord because it was expected that many neurons of interest would either project rostrally, or would be located far rostrally enough so that their caudal projections would be intercepted by the injection. However, many types of neurons at more caudal levels of the spinal cord with either short local projections or caudal projections, were not studied. Second, the reduced visual resolution intrinsic to a wholemount preparation has limited the detail which can be seen in cells labeled by retrograde transport, rather than by intra-axonal diffusion. Identification of such cells has been limited to (1) oblique bipolar cells because of their unusual shape, (2) giant interneurons because of their size, which also permits visualization of cellular details such as prominent medial dendrites, even when inadequately filled by grains of HRP, (3) edge cells because of their characteristic location in the lateral white columns, and (4) dorsal cells because of their large round cell bodies and medial location. Distinctions among most rostrally projecting interneurons could not be made. This was particularly disappointing because it may have prevented more definite identification of the trident shaped cells as a distinct cell type based on unique axonal projection patterns. Nor was it possible to distinguish between myotomal motoneurons and fin motoneurons on the basis of their HRP filled profiles in muscle-injected preparations. Third, hemisections were made only caudal to HRP injections, so that, except for cases in which HRP-filled axons could be followed for long distances (e.g., lateral cells), the lateralities of the caudal projections of neurons was not studied. Finally, we were unsuccessful in using HRP labeling of cells such as dorsal cells, which have long rostrally projecting axons, to calculate the rate of retrograde axoplasmic transport. Labeling of cells did not progress centrifugally from the injection site, but became progressively more intense a t all levels of the spinal cord over a period of 20 or more days. Among the possible explanations for the gradual simultaneous appearance of HRP grains a t all levels of the spinal cord are: (1)there might be several different transport mechanisms, with the fastest components unable to provide enough labeling to allow cell visualization, (2) neurons with long axons might have faster rates of transport than ones with

644

D. TANG AND M. E. SELZER

short axons, (3) the rate of cell labeling might be limited by continuous uptake of HRP from the extracellular space, or by a process inserted between uptake of HRP and its retrograde transport. At the neuromuscular junction HRP is taken up by endocytosis and much of i t is accumulated in cisterns (Heuser and Reese, '73).The mechanisms involved in channeling the HRP into the retrograde transport system are not known, but this channeling might be gradual, thus obscuring a relatively rapid rate of transport. In the chick calculations of the rate of retrograde transport based on HRP studies have suggested a rate of a t least 72-84 mm per day (LaVail and LaVail, '72, '74). Rates in poikilotherms are less well established, but much slower rates (1-3 mm/ day) may be inferred from studies on teleosts (Luiten, '75). Based on the time of earliest appearance of grains in giant interneurons a maximum rate of transport (including any initial delays) would be approximately 20-30 mm/day. ACKNOWLEDGMENTS

We are grateful to Karen Wellerstein for excellent technical assistance, and Dr. C. M. Rovainen for helpful comments on an earlier draft of this paper. This project was supported by NIH Academic Career Development Award 5K07 NS11083 to Michael E. Selzer, and NIH Grants NS14257 and NS14837. LITERATURE CITED Bertolini, B. 1964 Ultrastructure of the spinal cord of the lamprey. J. Ultrastruct. Res., 11: 1-24. DeVito, J. L., K. W. Clausing and 0. A. Smith 1974 Uptake and transport of horseradish peroxidase by cut end of the vagus nerve. Brain Res., 82: 269-271. Donkelaar, H. J. ten 1976 Descending pathways from the hrainstem to the spinal cord in some reptiles. I. Origin. J. Comp. Neur., 167: 421-442. Finger, T. E. 1975 The distribution of the olfactory tracts in the bullhead catfish Ictalurus nebulosus. J. Comp. Neur., 161: 125-142. 1976 Gustatory pathways in t h e bullhead catfish. I. Connections of the anterior ganglion. J. Comp. Neur., 165: 513-523. Finger. T. E., and C. M. Rovainen 1978 Retrograde HRP labeling of t h e oculomotoneurons in adult lampreys. Brain Res., 154: 123-127. Heuser, J. E., and T. S. Reese 1973 Evidence for recycling of synaptic vescicle membrane during transmitter release a t the frog neuromuscular junction. J. Cell Biol.. 57: 315-344. Homma, S. 1978 Organization of the trigeminal motor nucleus before and after metamorphosis. Brain Res., 140: 33-42. Ito, H., and R. Kishida 1977 Tectal afferent neurons identified by the retrograde HRP method in the carp telencephalon. Brain Res., 130: 142-145.

JSarten, H. J., and T. E. Finger 1976 A direct thalamo-cerebellar pathway in pigeon and catfish. Brain Res., 102: 335-338. Kristensson, K., and Y. Olsson 1974 Retrograde transport of horseradish peroxidase in transected axons. I. Time relationships between transport and induction of chromatolysis. Brain Res., 79: 101-109. LaVail, J. H. 1975 The retrograde transport method. Fed. Proc., 34: 1618-1624. LaVail, J. H., and M. M. LaVail 1972 Retrograde axonal transport in the central nervous system. Science, 176: 1416-1417. - 1974 The retrograde intra-axonal transport of horseradish peroxidase in the chick visual system: a light and electron microscopic study. J. Comp. Neur., 157: 303-357. LaVail, G. H., K. R. Winston and A. Tish 1973 A method based on retrograde intra-axonal transport of protein for the identification of cell bodies of origin of axons terminating within the CNS. Brain Res., 58: 470-477. Luiten, P. G. M. 1975 The horseradish peroxidase technique applied to t h e teleostean nervous system. Brain Res., 89: 181-186. Lynch, G., C. Gall, P. Mensah and C. W. Cotman 1974 Horseradish peroxidase histochemistry: a new method for tracing efferent projections in the central nervous system. Brain Res., 65: 373-380. Lynch, G. S., R. L. Smith, P. Mensah and C. W. Cotman 1973 Tracing t h e dentate gyrus mossy fiber system with horseradish peroxidase histochemistry. Exp. Neurol.. 40: 516-524. Martin, A. R., and W. 0. Wickelgren 1971 Sensory cells in the spinal cord of the sea lamprey. J. Physiol., 212: 65-83. Martin, R. J.,and D. Bowsher 1977 An electrophysiological investigation of t h e projection of t h e intramedullary primary afferent cells of the lamprey ammocoete. Neurosci. Letters, 5: 39-43. Mensah, P. 1974 Lateral Funiculus of the Amphibian Spinal Cord. Ph.D. thesis. University of California, Irvine. R e g r a n t , J. 1975 The orthograde transport of horseradish peroxidase in the visual system. Brain Res., 85: 307-312. Rovainen, C. M. 1967a Physiological and anatomical studies on large neurons of central nervous system of sea lamprey (Petromyzon marinus). I. Miiller and Mauthner cells. J. Neurophysiol., 30: 1000-1023. 1967h Physiological and anatomical studies on large neurons of central nervous system of sea lamprey Petromyzon marinus). 11. Dorsal cells and giant interneurons. J. Neurophysiol., 30: 1024-1042. 1974 Synaptic interactions of identified nerve cells in the spinal cord of the sea lamprey. J. Comp. Neur., 154: 189-206. Rovainen, C. M., and K. L. Birnberger 1971 Identification and properties of motoneurons to fin muscle of the sea lamprey. J. Neurophysiol., 34: 974-982. Rovainen, C. M., P. A. Johnson, E. A. Roach and J. A. Mankovsky 1973 Projections of individual axon8 in lamprey spinal cord determined by tracings through serial sections. J. Camp. Neur., 149: 193-202.$ Scalia, F., and D. R. Colman 1974 Aspects of the central projection of the optic nerve in the frog as revealed by anterograde migration of horseradish peroxidase. Brain Res., 79: 496-504. Selzer, M. E. 1978 Mechanisms of functional recovery and regeneration after spinal cord transection in larval sea lamprey. J. Physiol., 277: 395-408.

LAMPREY SPINAL NEURONS

1979 Variability in maps of identified neurons in the sea lamprey spinal cord examined by a wholemount technique. Brain Res., 163: 181-193. Smith, D. S., U. Jarlfors and R. Beranek 1970 The organization of synaptic axoplasm in the lamprey Petromyzon marinus) central nervous system. J. Cell Biol., 46: 199-219.

645

Streit, P., and J. C. Reubi 1977 A new and sensitive staining method for axonally transported horseradish peroxidase (HRP) in the pigeon visual system. Brain Res., 126: 530-537. Tretjakoff, D. 1909 Das Nervensystem von Ammocoetes. 1. Das Ruckenmark. Arch. Mikr. Anat., 73; 607-680.