Distinct, parallel pathways link the medial mammillary bodies to the anterior thalamus in macaque monkeys

Fluorescent tracer injections

Six adult rhesus monkeys received injections of two fluorescent tracers, Dy (Keizer et al., 1983) or Fb (Kuypers et al., 1980), with different tracers being placed into different thalamic sites in the same animal. The animals were tranquilized with ketamine hydrochloride (10–15 mg/kg, intramuscular injection) and surgical anesthesia was maintained with gas (isoflurane 1–4%, to effect). They were placed in a specialized head holder and, following aseptic procedures, dorsal bone and dural flaps were opened to expose the midline. The wall of the left hemisphere was then gently retracted, and a 5–10-mm portion of the corpus callosum and the underlying fornix were split longitudinally to expose the thalamic midline. An injection of either Fb (Sigma) or Dy (Sigma) was made through a 5-µL Hamilton syringe fitted with a 28-gauge needle. Both tracers were injected as 3% suspensions in distilled water. In three cases (BRh3, BRh4, BRh6), a fiber pathway (fornix or extreme capsule) was cut unilaterally immediately prior to the injections (Table 1).

To minimize the use of animals, some monkeys received two (BRh2), three (BRh5, BRh6) or even four (BRh3, BRh4) injections of the two fluorescent tracers, each in different thalamic sites (Table 1). In BRh2, single injections of the two different tracers (Dy, Fb) were directed at either the anterior thalamic nuclei (Fb) or nucleus medialis dorsalis (Dy) in opposite hemispheres. In animals BRh3 and BRh4, separate injections were placed in the anterior thalamic nuclei and nucleus medialis dorsalis in the same hemisphere, but using different tracers (Fb and Dy). This procedure was then repeated in the opposite hemisphere, ensuring that the anterior thalamic nuclei were injected with Fb in one hemisphere and Dy in the other hemisphere (Table 1). In case BRh5, bilateral injections (one Fb, the other Dy) were placed in nucleus lateralis dorsalis in opposite hemispheres, while a further injection of Fb was placed in the anterior thalamic nuclei in the same hemisphere as that with the Dy injection. Finally, in BRh6, two injections (one Fb, the other Dy) were placed in nucleus lateralis dorsalis in opposite hemispheres, while an injection of Fb was placed in the midline and so involved the most medial portion of medialis dorsalis in both hemispheres (Table 1).

Three of the rhesus monkeys (BRh3, BRh4 and BRh6) had additional surgical procedures immediately before the tracer injections, i.e. as part of the same surgery (Table 1). In all three cases this involved unilateral removal of the fornix and, in the other hemisphere, a unilateral cut in the ventral amygdalofugal pathway/extreme capsule. Details of these concurrent surgical procedures have been provided previously (Aggleton et al., 2005; Saunders et al., 2005). The cut in the ventral amygdalofugal pathway was lateral and dorsal to the amygdala, and so did not interfere with any structures or tracts in the present study. Even though the fornix lesions will have removed hippocampal inputs to the anterior thalamic nuclei and the mammillary bodies, the surgery is most unlikely to have affected direct projections from the mammillary bodies to the anterior thalamus as it was completed just prior to tracer injection, i.e. in the same surgery. Furthermore, transection of the fornix in rhesus monkeys does not result in a long-term loss of neurons within the mammillary bodies, despite evidence of atrophy due to a loss of neuropil (Loftus et al., 2000).

After a postoperative period of between 5 and 10 days the animals were deeply anesthetized with sodium pentobarbital. They were then perfused intracardially with saline followed by approximately 2 L of 4–6% paraformaldehyde in 0.1 m cacodylate buffer (pH 7.4). The brains were then removed and placed in a series of cryoprotectant solutions consisting of first 10% and then 20% glycerol in 0.1 m cacodylate buffer with 2% dimethylsulfoxide and 4–6% paraformaldehyde (pH 7.4, 4 °C). Four−6 days after the perfusion the brains were rapidly frozen by immersion in −75 °C isopentane, and then cut at 40 µm in the coronal plane on a freezing microtome (Rosene et al., 1986). Three, one-in-10 series of sections were mounted immediately onto subbed slides, dried and stored in the dark at 4 °C.

HRP injections

A single injection of HRP was made into the medial part of the thalamus in six adult cynomolgus monkeys weighing 3.5–6.8 kg at the time of surgery. The surgical procedures used to visualize the dorsal thalamus closely matched those for the fluorescent injections. All injections were then made under visual guidance. In three cases (ACy1, ACy2, ACy26) the injection tract was in or very close to the midline at the level of the anterior thalamic nuclei. In the remaining three cases (ACy3, ACy5, ACy7) the HRP injections were targeted at the magnocellular part of nucleus medialis dorsalis.

Four cases received a single injection of 40% HRP (Sigma, type IV) with a 1-µL Hamilton syringe. The volume injected ranged from 0.13 µL (ACy2) to 0.15 µL (ACy3, ACy5) to 0.22 µL (ACy1). In animal ACy7 a 10% solution of HRP in Tris buffer was delivered iontophoretically through a glass pipette, while in ACy26 the injection consisted of 0.09 µL of a 4% solution of HRP (Sigma, type VI) conjugated with wheat germ agglutin.

Two days after surgery the monkeys were deeply anesthetized with Nembutal and perfused intracardially with 0.9% saline followed by a solution of 1% paraformaldehyde and 1.25% glutaraldehyde in 0.1 m phosphate buffer (pH 7.2). The brains were stored in 30% sucrose buffer at 4 °C for 3–4 days and then cut in 50-µm coronal sections. A one-in-five series was then treated according to the protocols of either Hardy & Heimer (1977) or Mesulam (1978). Alternate sections were dehydrated, counterstained with thionine and coverslipped, while the remaining sections were just dehydrated and coverslipped.

Anatomical designations

The primate mammillary bodies are located at the base of the brain in the posterior hypothalamus (Fig. 1A) and comprise two main nuclear groups, the medial and lateral mamillary nuclei (Rose, 1939; Veazey et al., 1982a). By far the larger of the two is the medial mammillary nucleus, which is surrounded by a dense, fiber capsule (Fig. 1B–D). The major afferent tract is the postcommissural fornix, which can be traced running into the rostral part of the medial mammillary nucleus. The mammillothalamic tract, the primary efferent tract, exits from the dorsal half of the same nucleus (Fig. 1A). Using the descriptions of the cynomolgus monkey brain by Veazey et al. (1982a), three cytoarchitectonic divisions can be observed within the medial mammillary nucleus: pars medialis, pars basalis and pars lateralis (Fig. 1B and 1D, see also Rose, 1939). Pars medialis is distinctive as it contains the largest neurons within the medial mammillary nucleus, while the cells in pars basalis are among the smallest in the posterior hypothalamus (Fig. 1B and D). The third component, pars lateralis, is the largest subdivision and occupies the lateral half of the medial nucleus. The medial and central parts of pars lateralis are distinguished from pars medialis by the lower density of cells, mainly as a consequence of the large number of fibers in this more lateral part of the medial nucleus. As a consequence there is not a sharp border with pars medialis. In the caudal half of the medial mammillary nucleus the more dorsal cells in pars lateralis form a distinctive ‘cap’ where they appear more densely aggregated (Fig. 1). This cap is sometimes more visible in myelin stained sections due to a reduction in fiber density (Fig. 1C).

The other nuclear group, the lateral mammillary nucleus, lies lateral and ventral to the medial mammillary nucleus (Fig. 1B). The lateral nucleus, which is shorter in the anterior–posterior plane than the medial nucleus (Fig. 1A), contains relatively large cells that stain intensely for Nissl substance. These cells are uniformly distributed, and so no cytoarchitectonic subdivisions have been described within the primate lateral nucleus. Although a third mammillary nucleus, nucleus intercalatus, has been reported by some authors (Veazey et al., 1982a), there is disagreement over its existence (Crosby & Showers, 1969; Nauta & Haymaker, 1969). This region, which lies lateral to pars lateralis of the medial mammillary nucleus, is indistinct as it is perforated by many fibers. While the study by Veazey et al. (1982a) recognized a separate nucleus intercalatus, it was rarely visible in the present study and so is omitted from descriptions of the data.

The nomenclature and borders of the thalamic nuclei, including those of the midline nuclei, are all taken from Olszewski (1952), as this monograph provides the most complete cytoarchitectonic description of all of the thalamic nuclei in the macaque brain. The anterior thalamic nuclei comprise three major nuclei (anterior medialis, and anterior dorsalis anterior ventralis). Immediately caudal to nucleus anterior dorsalis lies nucleus lateralis dorsalis, which is often grouped with the anterior thalamic nuclei (Bentivoglio et al., 1993). For this reason, nucleus lateralis dorsalis also received injections of retrograde tracers.

Injections in nucleus anterior medialis

Very striking evidence of a topographic projection from the medial mammillary nucleus was found in case BRh2 after an injection of Fb placed in the mid anterior–posterior level of nucleus anterior medialis just lateral to the midline (Fig. 2). There was an extremely clear dorsoventral gradient of retrograde label within the medial mammillary nucleus (Fig. 4). While the dorsal third of the medial mammillary nucleus contained a high density of labeled cells, there was only a diffuse presence throughout the rest of the medial mammillary nucleus. This more dense label was largely confined to pars lateralis, and was especially obvious in the dorsal cap of pars lateralis (4, 5). Very light Fb label was also present in the dorsal half of the medial mammillary nucleus in the contralateral hemisphere. Finally, a few Fb-labeled cells were spread diffusely across the lateral nucleus of the mamillary bodies in both hemispheres.

All cases in which the tracer injection was centered in the nucleus anterior medialis resulted in label in the pars lateralis of the medial mammillary nucleus with a clear dorsoventral gradient in the density of labeled cells. In case BRh4, the Dy injection (Fig. 2) was centered in the caudal part of the nucleus anterior medialis, resulting in a number of labeled cells throughout the medial mammillary nucleus on the ipsilateral side, with the exception of pars basalis, which contained few if any cells. As with the more rostral injection there was a clear gradient as the density of label increased appreciably in the dorsal half of the nucleus (4, 5). The Dy label in the dorsal cap within pars lateralis was especially dense (4, 5). Very small numbers of Dy-labeled cells were present in the lateral nucleus of the mamillary bodies in both hemispheres.

Similar patterns of label came from two cases with HRP injections (ACy1, ACy2) that involved nucleus anterior medialis (Fig. 2). In case ACy2, the needle tract was in the midline portion of rostral nucleus anterior medialis, but also involved nucleus centralis densocellularis. Near mirror image patterns of mammillary body label were present in both hemispheres, presumably reflecting the spread of HRP into medial nucleus anterior medialis in both hemispheres (Fig. 2). The very large majority of retrograde label in ACy2 was in the pars lateralis of the medial mammillary nucleus. In the rostral third of the medial mammillary nucleus this label was grouped in a central region so that it involved the lateral parts of pars medialis and medial part of pars lateralis (Fig. 4). At mid anterior–posterior levels pars lateralis contained a great many retrogradely labeled cells that were present almost throughout this subdivision. The dorsal portion of pars medialis also contained many labeled cells, which contrasted with the rest of pars medialis and pars basalis as they contained extremely few labeled cells. At slightly more caudal levels the dense retrograde label was almost completely confined to the dorsal portion of pars lateralis, completely filling the dorsal cap. Because some sections in case ACy2 were counterstained for Nissl, it was possible to confirm that every cell contained HRP retrograde label in the dorsal cap region of pars lateralis. At this more caudal level there were only a handful of labeled cells distributed across pars medialis. The lateral nucleus of the mammillary bodies contained a scattering of labeled cells in both hemispheres. These cells appeared to be evenly distributed across the lateral nucleus, except for a lack of label at the most caudal portion of this nucleus.

The needle placement in ACy1 was very similar to ACy2 as it crossed centralis densocellularis and midline anterior thalamic nuclei (Fig. 2). The HRP injection was, however, just a little more caudal and a little deeper. Although a much greater amount of HRP was injected into the thalamus, the pattern of retrograde label within the mammillary bodies indicated a restricted region of uptake. Apart from the most rostral level of the mammillary bodies, where labeled cells were scattered evenly across the medial mammillary nucleus, the dense retrograde label was almost entirely confined to pars lateralis (Fig. 4). Here, the label was very heavily concentrated in the lateral and dorsal parts of pars lateralis. This label filled much of the dorsal cap in the caudal half of the structure, only avoiding its most medial part. Occasional labeled cells were also found in pars basalis and pars medialis within the medial mammillary nucleus, but these were very few in number. The lateral nucleus of the mammillary bodies contained a scattering of labeled cells, with no obvious topography.

While case BRh3 had an injection of Fb centered in the nucleus anterior medialis, the injection tract clearly severed the mammillothalamic tract (Fig. 2). The fibers of the mammillothalamic tract were densely filled, and in contrast to the localized label within the pars lateralis there was dense label throughout the medial mammillary nucleus. The lateral nucleus of the mammillary bodies nucleus also had numerous labeled cells throughout (Fig. 6).

Injections in nucleus AV

In case BRh4 an injection of Fb was centered in the mid anterior–posterior level of nucleus anterior ventralis, and is likely to have extended into adjacent parts of nuclei anterior dorsalis and anterior medialis (Fig. 2). There were large numbers of labeled cells distributed throughout the medial mammillary nucleus (5, 6). The lowest density was in the pars basalis, and the highest density was in the more dorsal portions of pars medialis. Labeled cells were found across pars lateralis, and in contrast to the pattern observed after the nucleus anterior medialis injections there was no obvious increase or decrease of label in the dorsal cap. Small numbers of Fb-labeled cells were found evenly distributed across the lateral nucleus of the mammillary bodies in both hemispheres (5, 6). A very similar pattern of retrograde label was found across the medial mammillary nucleus in case BRh3, which had a Dy injection involving the more medial parts of nucleus anterior ventralis (5, 6).

In case BRh5 an injection of Fb was made in the caudal part of anterior ventralis, but it also encroached into caudal nucleus anterior medialis and some of caudal nucleus anterior dorsalis (Fig. 2). Considerable label was found in all three subregions of the medial mammillary nucleus (5, 7). At rostral levels, pars basalis contained the fewest cells, while pars lateralis and medialis contained very similar densities of labeled cells. At mid anterior–posterior levels label was present across all subregions, but there was a denser band across the ventral and central parts of pars lateralis, along with dorsal pars medialis (5, 7). As a consequence the dorsal cap showed a decrease in the density of label. At the most caudal levels very dense packing of labeled cells was observed across the medial mammillary nucleus. Scattered cells were found across the lateral mammillary nucleus in both hemispheres. In addition, a light scattering of Fb-labeled cells was found across the medial mammillary nucleus in the contralateral hemisphere. This label was most evident close to the periphery of the nucleus, i.e. near to its outer edges, and so included all three subregions.

Injections in the rostral midline thalamic nuclei

In one case (ACy26) an injection of HRP was centered in the rostral thalamic midline just ventral to nucleus anterior medialis, so that the tract passed through the latter nucleus (Fig. 2). The needle tip was placed in ventral nucleus centralis latocellularis, and so the core of the injection involved nucleus reuniens along with some of nucleus anterior medialis. Large numbers of HRP-retrogradely labeled cells were observed in both the medial mammillary nucleus and lateral nucleus of the mammillary bodies (5, 7). A dense aggregation of labeled cells was restricted to pars lateralis of the medial mammillary nucleus. These labeled cells were, for the most part, restricted to the ventral, central and lateral parts of pars lateralis (5, 7). This aggregation was present throughout the anterior–posterior extent of the nucleus. Little or no label was observed in the dorsal cap of pars lateralis. Compared with other cases, larger numbers of labeled cells were scattered across the lateral nucleus of the mammillary bodies, with a slight increase in the medial half of this nucleus. The same patterns of mammillary body label were observed in both hemispheres (Fig. 7).

Injections in nucleus lateralis dorsalis

Two animals received bilateral injections of fluorescent tracers centered in nucleus lateralis dorsalis (BRh5, BRh6; Fig. 3). BRh5 had Dy in the left and Fb in the right nucleus lateralis dorsalis, as well as Fb in nucleus anterior ventralis (Table 1). Potential mammillary inputs to nucleus lateralis dorsalis can therefore only be inferred from the Dy injection in BRh5. While no Dy-positive cells were seen in the medial mammillary nucleus, a few Dy-positive cells were present in the lateral nucleus of the mammillary bodies.

BRh6 had bilateral injections of Dy in nucleus lateralis dorsalis, and an injection of Fb along the midline involving the midline thalamic nuclei and the nucleus medialis dorsalis (Table 1). Once again, only a very few Dy cells were found in lateral nucleus of the mammillary bodies and none in medial mammillary nucleus.

Injections in MD and/or caudal midline nuclei

Animals with injections placed in nucleus medialis dorsalis helped to clarify the findings from the two cases with multiple tracer injections (BRh3, BRh4) involving the anterior thalamic nuclei and nucleus medialis dorsalis (Table 1, 2, 3). Three cynomolgus monkeys had single injections of HRP centered in magnocellular nucleus medialis dorsalis (ACy3, ACy5, ACy7). There was variable medial spread across the midline, but little evidence of rostral spread into the anterior thalamic nuclei (Fig. 3). Additional information came from three cases with fluorescent tracer injections (BRh1, BRh2, BRh6), where nucleus medialis dorsalis was the only target for one specific tracer. In none of these six cases was there evidence of a projection from the medial mammillary nucleus. In view of the placement of these injections (2, 3), it can be concluded that the medial mammillary nucleus does not project to either nucleus medialis dorsalis or nucleus centralis intermedialis, the midline thalamic nucleus running between the left and right nucleus medialis dorsalis at this level.

The findings for the lateral nucleus of the mammillary bodies nucleus were not so consistent after the injections involving nucleus medialis dorsalis. No labeled cells were observed in the lateral nucleus of the mammillary bodies nucleus in cases ACy7 (HRP), BRh1 (Fb) or BRh2 (Dy). In case BRh6, with a Fb injection into midline thalamus and nucleus medialis dorsalis, a couple of cells were present in lateral nucleus of the mammillary bodies. Consistent with this finding, an extremely small number of HRP-positive cells were seen in the lateral nucleus of the mammillary bodies in cases ACy3 and ACy5. A common feature of the injections in these three cases was the greater involvement of the midline thalamic nuclei at the level of the rostral half of nucleus medialis dorsalis. This pattern of results suggested the presence of an extremely light projection to one or more midline thalamic nuclei, either at the level of nucleus medialis dorsalis or immediately rostral to this nucleus (e.g. centralis intermedialis, centralis superior or centralis densocellularis).