Neuronal responses to tactile stimuli and tactile sensations evoked by microstimulation in the human thalamic principal somatic sensory nucleus (ventral caudal)

Abstract

The normal organization and plasticity of the cutaneous core of the thalamic principal somatosensory nucleus (ventral caudal, Vc) have been studied by single-neuron recordings and microstimulation in patients undergoing awake stereotactic operations for essential tremor (ET) without apparent somatic sensory abnormality and in patients with dystonia or chronic pain secondary to major nervous system injury. In patients with ET, most Vc neurons responded to one of the four stimuli, each of which optimally activates one mechanoreceptor type. Sensations evoked by microstimulation were similar to those evoked by the optimal stimulus only among rapidly adapting neurons. In patients with ET, Vc was highly segmented somatotopically, and vibration, movement, pressure, and sharp sensations were usually evoked by microstimulation at separate sites in Vc. In patients with conditions including spinal cord transection, amputation, or dystonia, RFs were mismatched with projected fields more commonly than in patients with ET. The representation of the border of the anesthetic area (e.g., stump) or of the dystonic limb was much larger than that of the same part of the body in patients with ET. This review describes the organization and reorganization of human Vc neuronal activity in nervous system injury and dystonia and then proposes basic mechanisms.

the organization and function of the monkey thalamic principal somatic sensory nucleus (ventral posterior, VP) has often been studied by electrophysiological techniques (Jones and Friedman 1982; Kaas et al. 1984; Loe et al. 1977; Mountcastle and Henneman 1952; Poggio and Mountcastle 1963; Pubols and Pubols 1972; Weng et al. 2000). However, plasticity of VP in response to nervous system injury is less well studied (Garraghty and Kaas 1991; Pollin and Albe-Fessard 1979; Rasmusson 1996a, 1996b). Even less well studied are the normal organization and plasticity of the human thalamic somatic sensory nucleus (ventral caudal, Vc) (Hassler 1959; Hirai and Jones 1989). We now review human studies of single-cell recordings and stimulation at microampere (μA) current levels (microstimulation) in patients undergoing operations for the treatment of movement disorders or chronic pain.

During the mapping of Vc that precedes surgical procedures for the treatment of movement disorders and chronic pain, the function and organization of Vc can be studied in detail. In patients with essential tremor (ET) or dystonia, the anterior and inferior borders of Vc are determined routinely, since they predict the borders of the thalamic nucleus ventral intermediate (Vim) and the nucleus ventral oral posterior (Vop), which are located anterior to Vc. During thalamotomy for ET and dystonia, these nuclei may be lesioned (Kobayashi et al. 2009), or deep brain stimulation (DBS) electrodes may be implanted in these nuclei for the treatment of ET (Bertrand and Lenz 1995; Hua et al. 1998; Kobayashi et al. 2009; Vitek and Lenz 1998). DBS electrodes can also be implanted in Vc to treat medically intractable chronic pain due to nervous system injury (Cruccu et al. 2007; Lenz 2006). Therefore, Vc is thoroughly examined in patients with ET, dystonia, and chronic pain.

In preparation for lesion making or implantation of a DBS, the location of the target is estimated by imaging [magnetic resonance imaging (MRI) or computer-assisted tomography (CT)]. Microelectrode studies are then used to confirm this estimate both by recordings of neuronal activity and by microstimulation. These microelectrode studies are a unique window into the characteristics of Vc in patients with ET who do not demonstrate abnormalities of the cutaneous sensory system, as well as in patients with nervous system injury or dystonia.

Our approach is to first present the technical protocols of this field that are essential for interpretation of these topics, such as optimal stimuli for mechanoreceptors, thalamic anatomy, thalamic nomenclature, and anesthetic considerations. We then describe neural activity in Vc of patients with ET in whom the neuronal responses to mechanical cutaneous stimulation and the sensations evoked by microstimulation are assumed to represent function of normal somatic sensory thalamus (see sections on Normal Somatic Sensory Thalamus later in this review). The somatic sensory thalamus in humans and animals is highly segmented in an orderly fashion by somatotopy and response patterns. The neuronal responses to precisely controlled mechanical cutaneous stimuli replicate those of the different primary afferent mechanoreceptors with a high degree of specificity.

This remarkable degree of organization can be dramatically disrupted in patients with major nervous system injury (amputation or spinal transection) leading to chronic pain and in patients with longstanding dystonic movements. Changes in somatotopy, receptive fields (RFs), and mismatches of RFs vs. projected fields (PFs) compared with control (ET patients) are evidence of reorganization of afferent input to the thalamus and of interactions between the thalamus and cortex, which may characterize injury and activity-dependent plasticity.

Techniques of Thalamic Electrophysiological Mapping in Humans, Monkeys, and Raccoons

We will review the methods used to study human and monkey thalamus because they influence the recordings that are the basis of this review. All human operative studies used quantitative sensory testing to evaluate sensory function and to train patients in the use of a validated questionnaire for describing sensations evoked by microstimulation (Lenz et al. 1993; 1998b), as well as to confirm that each patient with ET was not outside the range of normal.

The thalamic stereotactic procedures described below were carried out under local anesthesia with the Leksell frame. First, the frame coordinates of the anterior commissure and posterior commissure are determined by MRI or CT scans. These coordinates are used in an atlas-based approach to estimate the borders of Vc (Fig. 1, top). During the surgery, these coordinates are corroborated by microelectrode stimulation and recording using standard techniques (Lenz et al. 1988a).

Fig. 1.

Map of RFs for trajectories in the region of Vc in a patient with ET. Top: trajectories in the parasagittal plane13.5 mm lateral to the midline through Vc (anterior to the left). The position of the trajectory is shown relative to nuclear boundaries as predicted radiologically and approximates the nuclear locations. The anterior commissure-posterior commissure (AC-PC) line is indicated by the horizontal line, whereas the oblique lines represent trajectories. Middle: AC-PC line and the 2 trajectories (S1, S2). The sites of cellular recordings sites are indicated by ticks to the right of the trajectory: long ticks indicate cells with RFs, and short ticks indicate cells without RFs. “Balloons” at the end of the ticks indicate a response to deep stimuli (filled circles), cutaneous touch (open squares), or pressure stimuli (filled squares). Triangles above the balloons indicate a rapidly (filled triangle) or slowly adapting response (open triangle). Bottom: RFs are shown for sites where a recording or stimulation, or both, was carried out, indicated by the same number as in B. NR (no response) indicates that the cell had no response to stimulation. [Reproduced with permission from Lenz et al. (1988b).]

Neurons are classified either as cutaneous sensory neurons that responded to manual touch, tap, pressure, and stroking or as deep sensory neurons that responded to squeezing of muscles or ligaments or to passive joint movement, but not to manipulation of skin overlying these structures. Neuronal activity was also studied while patients performed multiple different active movements of the face, arm, and leg. Thereafter, microstimulation was carried out (Ohara and Lenz 2003).

The borders of Vc can be identified by the location of the most anterior and the most inferior cutaneous sensory neurons in the region where deep or cutaneous sensory neurons are the majority (Lee et al. 2005). These borders are located by registering atlas maps (e.g., Hassler 1959) to the MRI images of the thalamus and local landmarks, such as the anterior and posterior commissures (Kobayashi et al. 2009). This process provides an estimate of the location of Vc and of adjacent subnuclei, including ventral caudal portae posteriorly (Vcpor) and ventral caudal parvocellular inferiorly (Vcpc), as shown in the sagittal section in Figs. 1 and 3. These human subnuclei are equivalent to the more detailed nomenclature for subnuclei of VP in old world monkeys, including subnuclei located medially (VPM, representing cranial structures) and laterally (VPL, representing corporal structures), anteriorly (VPLa, representing deep structures), posteriorly (VPLp, representing cutaneous structures), and inferiorly (VPI) (Hirai and Jones 1989). This nomenclature is approximately equivalent to VP (including VPL, VPM, and VPI) in Carpenter's nomenclature, which has been applied to other primate species and raccoons (Parent 1996).

Fig. 3.

Map of RFs and PFs for trajectories in the region of Vc in a patient with spinal cord transection at T8. A: trajectory in 15-mm lateral parasagittal plane (Lat. 15 mm) through Vc that represents the arm. B: trajectory 2 mm lateral to the first (Lat. 17 mm). Top: the position of the trajectory, the AC-PC line, and the recording and stimulation sites. Conventions as in Fig. 1, except as described below. Stimulation sites are shown to the left of the trajectory: long ticks indicate a site with PF (threshold, in μA), and short ticks indicate no PF. Neuronal responses are shown to the right, as in Fig. 1. Bottom: the paired locations of microstimulation and recording sites as identified by the number between each pair. The RF location is shown on the right; NR indicates no neuronal response to peripheral stimulation. The PF location is shown on the left at the threshold (μA) indicated below the representation. In this case, the descriptor tingling was endorsed at all sites. [Reproduced with permission from Lenz et al. (1994a).]

The analysis of thalamic spike trains in some subjects with ET included the response to precisely controlled, calibrated mechanical stimuli. These cutaneous stimuli were identified as optimal stimuli for each class of primary afferent mechanoreceptor (Johnson et al. 2000; Pubols and Pubols 1973). For studies of each neuron with precise stimuli, the hand is held in a malleable plastic cast that fixes the hand in position relative to the stimulator, which was positioned by a multijointed manipulandum. These mechanoreceptors and their optimal stimulus include the Pacinian corpuscle (128-Hz vibration), rapidly adapting (32–64 Hz), slowly adapting type 1 (edge), and slowly adapting type 2 (skin stretch); stretch and edge were applied both parallel and perpendicular to the long axis of the limb.

Microstimulation (200 Hz, biphasic) is carried out at intervals along the electrode trajectory through Vc. If a sensation at any site is evoked at less than 50 μA, then the threshold is established by the method of successive approximation as the current is increased and decreased. Microstimulation at this threshold is then used to determine the extent and location of the evoked sensation (projected field, PF), as well as the quality of the sensation (Lenz et al. 1993). The patient is asked whether the evoked sensation is “something that you might encounter in everyday life”; if so, it can be classified as natural. Thereafter, the patient is asked whether the sensation is perceived deep in the skin, on the surface of the skin, or both on the surface and deep.

The size of PFs and RFs can be approximated from intraoperative description of their locations. Classification of somatotopic organization of Vc followed standard conventions (Lenz and Byl 1999; Lenz et al. 1988b). From medial to lateral, separate parts of the body included intraoral, perioral (Fig. 1), facial, digit 1 (D1), D2, D3, D4, D5, multiple digits, palm/hand, forearm, whole arm, upper arm, wrist, trunk, leg, upper leg, lower leg, pelvis, ankle, foot, and toes (Lenz et al. 1988b). If the RF and PF at a site both include the same part of the body, then that site can be classified as a match (Lenz and Byl 1999).

We will review studies of monkeys and raccoons as animal models of dystonia and major somatic sensory injury carried out as nonsurvival thalamic stereotactic and cortical craniotomy surgeries. Anesthesia was carried out with 1) halothane induction and chloralose or barbiturate maintenance (Byl et al. 1996; Pons et al. 1991; Rasmusson. 1996b), 2) ketamine plus xylazine induction and ketamine maintenance, and 3) induction and maintenance on flurane or pentothal (Warren et al. 1986; Zhang et al. 2001b). These methods producing small changes in VP neuronal response patterns at low doses of barbiturates but greater changes with other agents (Dougherty et al. 1997). Recordings are carried out with standard techniques including handheld devices for search stimuli followed by precisely controlled mechanical stimuli in a small number of studies as reviewed below (Hirai and Jones 1989; Warren et al. 1986; Zhang et al. 2001b). All animals are euthanized with recovery of the brain for histological processing, which produces greater accuracy of the recording site location.

Normal Somatic Sensory Thalamus: Organization of the Somatotopic Maps of Human Vc

Studies of normal somatic sensory thalamus are described in this and the next three sections. These sections describe the somatotopic organization of Vc, the neuronal activity evoked by precisely controlled mechanical stimulation, and the effect of lesions and of sensations evoked by thalamic microstimulation, followed by a summary.

In studies of patients with ET, Vc was defined as the region where the majority of cells respond to nonpainful cutaneous mechanical stimuli; the area corresponds to monkey VP (Hirai and Jones 1989). In Vc, RFs remain relatively constant over distances of up to 6 mm in sagittal planes (Lenz et al. 1988b) (Fig. 1, trajectory S2). However, RFs change markedly over much shorter distances in the medial-lateral direction. These results are consistent with the model where the unit of somatotopic isorepresentation within Vc and monkey is lamellae in the parasagittal plane. The sequence of neuronal cutaneous RFs indicates that medial-lateral lamellae represent intraoral structures, the face, thumb, fingers (radial to ulnar), arm, and leg (Lenz et al. 1988b). Within any lamella of VPL, this medial-lateral somatotopy is consistent with most prior studies of the human thalamus (reviewed by Lenz et al. 1988b).

From dorsal to ventral within the representation of corporal structures in Vc, RFs progress from chest and abdominal wall dorsally to proximal and then distal aspects of the limbs, such as shoulder to fingertips. Within any parasagittal trajectory through Vc from anterior-dorsal to posterior-ventral, characteristic changes in RFs are observed. In particular, RFs and PFs on parts of the body represented laterally within Vc (e.g., Fig. 1, trajectory S2, facial RFs) are found both anterior-dorsal and posterior-ventral to RFs on parts of the body that are represented medially (Fig. 1, trajectory S2, intraoral RFs). This pattern is consistent with different body parts being represented by concave medial lamellae, as in monkeys (Mountcastle 1980; Mountcastle and Henneman 1952).

In Fig. 1, cutaneous cells in Vc are significantly clustered by rapidly adapting responses (filled triangles) along trajectory S1, whereas a small cluster of slowly adapting responses (open triangles) is found at the superior aspect of trajectory S1. Similarly, responses to touch (open squares) are exclusively found in trajectory S2, whereas responses to pressure stimuli (filled squares) are almost exclusively found in trajectory S1 (Lenz et al. 1988b). This clustering by response was found to be statistically significant (median runs tests) and was consistent with reports of the detailed organization of VP in monkeys (Jones and Friedman 1982). This highly segmented organization is also reflected in the thalamic neuronal response to the stimuli that optimally evoke the activity of the classes of peripheral mechanoreceptors.

Normal Somatic Sensory Thalamus: Responses to Optimal Tactile Stimuli in VP and Vc

In VP of anesthetized marmosets and raccoons, neuronal response patterns have been compared with those of the mechanoreceptor classes using precisely controlled mechanical cutaneous stimuli (Hirai and Jones 1989; Warren et al. 1986; Zhang et al. 2001b). Each stimulus produces the optimal response for one class of cutaneous mechanoreceptor, and all were applied during the study of individual thalamic neurons (see Techniques of Thalamic Electrophysiological Mapping in Humans, Monkeys, and Raccoons). Thalamic neuronal responses were found to be similar to those of one of the four classes of peripheral mechanoreceptor (Pubols and Pubols 1973). These results are consistent with studies of neuronal responses to cutaneous stimuli in old world monkeys (Mountcastle et al. 1969; Sinclair et al. 1991). In addition, the response of neurons in human Vc, as well as monkey VP, to handheld stimuli has often been used to identify neurons with slowly adapting or rapidly adapting response patterns, as described above (Jones et al. 1982; Jones and Friedman 1982; Lenz et al. 1988b).

Human thalamic responses to precisely controlled cutaneous mechanoreceptive stimuli have been studied using a technique similar to these studies in animals (Weiss et al. 2009). Figure 2 shows an example of a thalamic Pacinian corpuscle (PC) neuron exhibiting a response to mechanical stimuli similar to that of a PC mechanoreceptor, with a RF on the thumb (Fig. 2A; from Weiss et al. 2009). The poststimulus firing rate for this neuron above the baseline skin indentation was significantly higher for the 128-Hz compared with the 32-Hz vibration stimulus (Fig. 2B), as well as stretch, in both the medial-lateral and proximal-distal directions (Fig. 2C), and edge. In the article from Weiss et al. (2009), seven neurons had a response to 128 Hz that was significantly greater than the response to any other optimal stimulus. The percentage entrainment was not different between 128 and 32 Hz for this cell and, indeed, was different in the case of only two of the PC-like neurons studied. Microstimulation adjacent to the recording sites evoked a deep, vibratory sensation at one body site and a deep, touch/electric/warm sensation at another site.

Fig. 2.

Physiological characteristics of a thalamic neuron with responses like those of PC mechanoreceptors. A: locations of RF and PF. B: responses to indentation and vibration at 128 and 32 Hz by the Chubbuck stimulator. C: entrainment of neuronal firing with stimulus signal and spike trains shown at top and cycle histograms at bottom, with frequencies as indicated. The cycle histograms indicate the spike rate as a function of the phase of the vibratory stimulus along the horizontal axis from 0° to 360°. The entrainment was expressed as a percentage expressing the proportion of the greatest number of impulses in any 180° span of the cycle histogram divided by the sum of impulses in the entire histogram. D: skin stretch in the directions indicated. [Reproduced with permission from Weiss et al. (2009).]

Seven rapidly adapting-like (RA) neurons showed a response to vibratory stimuli at 32 and/or 64 Hz that was significantly greater than the response to any other optimal stimulus. Microstimulation at or adjacent to the thalamic recording sites evoked deep cutaneous vibration sensations at four sites, whereas a surface, sharp, moving sensation was reported at one site and tingling sensations at two sites (Weiss et al. 2009).

Slowly adapting type 1 (SA1) characteristics were found in three neurons. Two neurons had a combined tonic and phasic response to edge stimulation that was significantly greater than the response to any other tested stimulus. For one neuron, the response to edge was greater than the response to any other stimulus except vibration at both 32 and 128 Hz. A difference in the response between medial-lateral and posterior-dorsal stimulus orientations was found for only one neuron. Microstimulation produced an unnatural, surface, and deep vibration sensation in a PF, which overlapped with the RF in the case of two neurons and an electric current sensation in the case of one neuron. Therefore, one of three neurons showed evidence that input arising from SA1 receptors converged with that from PC or RA mechanoreceptors (Johansson et al. 1982).

Slowly adapting type 2 (SA2) characteristics were found for three neurons. All had a significant increase in firing response to stretch in the posterior-dorsal as well as medial-lateral, or both, directions. In the case of one neuron, the response to 32 Hz was not significantly different from that to stretch in either direction and therefore received the label nonselective SA2 neuron. Microstimulation at the site of this neuron evoked a sensation of vibration over a PF that included the RF. Stimulation at another recording site evoked an unnatural, surface, and deep, vibration and electric current sensation.

These results demonstrate that the activity of most Vc neurons responding to mechanical cutaneous stimuli are similar to those of peripheral mechanoreceptors (Johnson et al. 2000). However, it is not clear how these responses are related to the sensations evoked by microstimulation at the recording site or at an adjacent site, except in the case of RA neurons. Therefore, there are common mismatches between neuronal responses to precisely controlled stimuli and sensations evoked by microstimulation at adjacent sites in Vc, although the number of sites is very small. However, a survey of a very large number of stimulation sites reveals that evoked sensations can be similar to the sensations produced by application of the optimal stimuli, as reviewed in the next section.

Normal Somatic Sensory Thalamus: Microstimulation evoked Sensations and Effects of Lesions

Human thalamic microstimulation produced sensations in the mechanical and movement categories at 122 sites among 1,854 sites in 42 thalami; the sensations found were touch (33 sites), pressure (17 sites), sharp (5 sites), vibration (40 sites), and movement (movement through the body or across the skin; 27 sites) (Ohara et al. 2004). Of 122 sites where mechanical and movement sensations were evoked, the PF and adjacent RF matched at 66 (54%) of the sites. The proportion of match was not different between different mechanical and movement descriptors.

The PFs from stimulation near cells with facial vs. intraoral RFs were proportionally different among these descriptors (Ohara et al. 2004). Evoked sensations of touch and vibration were more common than pressure in the case of facial RFs, and pressure was more common than vibration in the case of intraoral RFs. Evoked sensations of sharp were too infrequent to be analyzed. In contrast, descriptors were not significantly associated with the location of the PFs on particular parts of the body (Ohara et al. 2004).

Maps of the location of PFs followed a less specific medial-to-lateral and dorsal-to-ventral progression than those for RFs so that the RFs did not overlap with the PFs at the closest stimulation site in ∼60% of such pairs (Lenz and Byl 1999; Lenz et al. 2001; Ohara and Lenz 2003). When we compared the qualities of sensations between sites where the RF and PF overlapped (match) with those where they did not overlap (mismatch), the sensations were more frequently described as “natural” (52%) at matched sites [31% of sites measured in ET patients, as described in Lenz and Byl (1999)].

The frequent mismatch sites may be related to stimulation of axons that are afferent and efferent to the somatic sensory thalamus and that are located throughout the nucleus and activated at low currents of stimulation (Hirai et al. 1988; Jones 1983; Ranck 1975). A similar mismatch was observed between the sensation evoked by microstimulation and the sensation produced by the optimal stimulus for the cell adjacent to microstimulation site, except in the case of thalamic RA neurons (see Normal Somatic Sensory Thalamus: Responses to Optimal Tactile Stimuli in VP and Vc).

These studies of responses to tactile stimuli and sensations evoked by stimulation in Vc are strong evidence of the involvement of this structure in tactile sensations. However, proof that this structure is involved and that it is essential for these sensations rests on lesion studies, such as that by Kim et al. (2007). In that study, the results of sensory testing with von Frey hairs were reported in four patients with ischemic strokes in the region of Vc, all with a volume of ischemic tissue (in mm³) less than the volume of Vc. All patients studied had impaired tactile sensation, and the impairment was greatest in the subject who had the largest lesion of Vc (Kim et al. 2007). Therefore, the evidence of neuronal activity, microstimulation-evoked sensations, and lesions strongly endorse Vc as a critical thalamic structure in tactile sensation.

Normal Somatic Sensory Thalamus: Summary

These results demonstrate that the map of inputs to human Vc is highly segmented into units of isorepresentation by somatotopy and response properties (Lenz et al. 1988b). The somatotopic map is consistent with the details of the mapping of VP in monkeys including the concave medial lamellae, the dorsal-to-ventral progression down the limbs, and the medial-to-lateral progression of RFs with oral structures medial and legs lateral (Jones et al. 1982; Pubols 1968). Histological evidence of this functional map can be shown by standard Nissl stains or by cytochrome oxidase staining of terminations of fibers from the dorsal column nuclei (Lin et al. 1979; Rausell and Jones 1991a, 1991b).

The organization of PFs in Vc must be related to activation of cortex by the axons and terminations of thalamocortical cells (Lenz et al. 2010). These structures form pathways that have been demonstrated by anatomical studies in nonhuman primates (Friedman and Jones 1981; Jones and Friedman 1982; Manger et al. 1995; Walker 1938, 1982) and by diffusion tensor imaging studies in humans, although the latter have relatively low spatial resolution (Behrens et al. 2003).

Human studies demonstrate that the optimal stimulus for PC mechanoreceptors (vibration at 132 Hz) (Johansson and Vallbo 1983; Johnson 2001) is the same as that for thalamic PC neurons in humans and monkeys (Weiss et al. 2009; Zhang et al. 2001a). Most of the PC neuronal RFs included the palmar surface of a digit consistent with the anatomic location of PC mechanoreceptors (Johansson and Vallbo 1983; Johnson 2001). Others had RFs determined with handheld stimuli that were located on perioral structures or the dorsal surface of the hand, sites at which PC mechanoreceptors are not normally found. These findings suggest that there may be convergence of inputs from mechanoreceptors other than PC (Johansson and Vallbo 1979; Nordin and Hagbarth 1989). These PC neurons may receive input from more than one type of mechanoreceptor, since they have RFs on parts of the body that do not normally contain PC mechanoreceptors (Iggo 1985).

RA neurons were defined by the optimal frequency of their response to vibration at 32 to 64 Hz in both humans and monkeys (Zhang et al. 2001a). The areas of the RFs of most RA neurons were comparable to the RFs of human RA mechanoreceptors, but some had substantially larger areas (Vallbo et al. 1979b). Because the thalamic RA neurons' responses are specific to RA optimal stimuli, but the RFs are larger, they certainly receive input from more than one RA mechanoreceptor.

Human thalamic SA1 and SA2 neurons respond optimally to directionally specific edge and stretch stimuli, respectively, as do the corresponding mechanoreceptors (Vallbo et al. 1979b; Weiss et al. 2009). RFs of the SA1 and SA2 neurons were often larger than those of the corresponding mechanoreceptor type, which is consistent with the convergence of inputs from mechanoreceptors transmitted through the dorsal column nuclei and which may be related to injury dependent plasticity (see Injury Dependent Plasticity: Reorganization of Human Vc after Amputations).

One potential confound of the studies described in above is the effect of anesthesia on results of the animal studies. The similarities of responses to precisely controlled cutaneous stimuli between humans versus monkeys and raccoons are much more striking than the differences, although the human studies were performed under local anesthesia, whereas monkeys and raccoons were all studied under general anesthesia. Neurons in the monkey thalamic VP have decreases in spontaneous and evoked firing in response under barbiturate anesthesia, whereas inhalational agents at relevant concentrations may or may not have decreased evoked responses (Dougherty et al. 1997). Nonetheless, anesthetic effects do not produce apparent differences in the somatotopy or the functional characteristics of mechanoreceptor-like neurons in the thalamus of these species.

Therefore, somatotopic organization of the body is segmented into units of isorepresentation in a discrete pattern that is consistent across individuals in humans (Lenz et al. 1988b) and monkeys (Jones et al. 1982; Jones and Friedman 1982; Lenz et al. 1988b). Similarly, the responses of thalamic neurons in Vc largely replicate those of all classes of peripheral mechanoreceptor fibers in peripheral nerve of humans (Vallbo et al. 1979a, 1979b). Microstimulation may evoke sensations (PFs) in the same somatotopic pattern as neuronal responses (RFs) and sensations, which are like those produced by the optimal stimuli for mechanoreceptors. However, there are frequent mismatches between locations and modalities of evoked neuronal responses and sensations, which may result from the activation of fibers of passage (Hirai et al. 1988). Mismatches may also be due to the divergence of thalamic terminations in cortex or to the interaction of lower and higher order thalamic nuclei (Sherman and Guillery 2001, 2002). Therefore, there is segmentation of somatotopic and mechanoreceptive activity in Vc, which seems to be the basis of cutaneous tactile sensation and sensory performance (Johnson 2001).

Injury Dependent Plasticity: Reorganization in Patients with Spinal Cord Transection

The borders of Vc can be difficult to determine in patients with injuries leading to interruption of somatic sensory input to the brain. Therefore, we broadly defined a “region of Vc” as Vc plus adjacent structures and then segmented this region on the basis of physiological properties. Operationally, Vc was defined as the cellular thalamic region where sensations were evoked at less than 25 μA (Fig. 3, sites 7 to 22 along S3 and sites 5 to 19 along S4). In these patients, the region of Vc was segmented on the basis of RF and PF locations. Areas with RFs and PFs, which represented structures distant from the anesthetic part of the body, were named “spinal control” areas (Fig. 3, S3), whereas those at which the patient had partial or complete sensory loss were termed “border zone/anesthetic” areas (Fig. 3, S4). The region of Vc in patients with ET was referred to as the “control” area.

The region of Vc was explored in patients with pain following spinal cord transection (n = 5), as described by Lenz et al. (1994a). Many cells in the representation of the anesthetic part of the body did not have RFs (Fig. 3, NR sites in S4 vs. S3; Lenz et al. 1994a). Figure 3 shows that the detailed organization of the spinal control area replicates results in ET (Fig. 1), including the concave medial lamellae, since laterally represented structures occur anterior-dorsal and posterior-ventral (S3, D4, and D5) to medially represented structures (D2 and D3).

Border zone/anesthetic areas were characterized by large representations of the border zone of sensory loss compared with corresponding parts of the body in control and spinal control areas. Figure 3 shows the same type of data from a patient after a complete spinal cord transection at T7. The RFs and PFs were well matched in spinal control areas (Fig. 3A, S3), but in border zone/anesthetic areas (Fig. 3B, S4), RF/PF mismatches were common. For example, in Fig. 3B (S4, sites 9 to 20), RFs are located on the chest and abdominal wall above the anesthetic level, whereas PFs are located in the leg (i.e., referred below the anesthetic level due to the spinal transection). In border zone/anesthetic areas, RFs are often found at the border of the anesthetic body part, whereas PFs are referred to anesthetic body parts so that activity in the border zone area produces sensations in the anesthetic area. Therefore, abnormal activity in the border zone area may responsible for dysesthesias referred to anesthetic parts of the body (Kim et al. 2007; Lenz et al. 1994a).

Injury Dependent Plasticity: Reorganization of Human Vc after with Amputations

In patients with amputations [n = 3; described in Lenz et al. (1998a)], neurons with RFs on and adjacent to the stump showed a larger representation (Lenz et al. 1998a) than the corresponding part of the body in patients with ET (Lenz et al. 1988b, 1994a). This is evidence for reorganization of the somatic map of inputs from the limb. Based on PFs, the increased representation of the stump versus that of the same part of the body in controls (Fig. 4B, trajectory S2; Lenz et al. 1998b; Ohara et al. 2004) is evidence for reorganization of the image of the limb, which is embedded in thalamocortical circuits (Jensen and Rasmussen 1994; Jensen et al. 1984).

Fig. 4.

RF and PF maps for trajectories in the region of Vc in a patient with amputation below the knee. Results are from trajectories in the 15.0 (left)- and 16.5-mm lateral parasagittal (right) planes through the thalamus. The lines representing the AC-PC line and trajectory are shown in A and B. C shows paired locations for sites as numbered in B. At all sites, sensations evoked by microstimulation were described as tingling. Other conventions are as in Fig. 3. [Reproduced with permission from Lenz et al. (1998a).]

In patients with the illusion of a phantom limb, it has been reported that sensations in the phantom can be evoked by stimulation of parts of Vc in which RFs are found on the stump (Davis et al. 1998). This latter phenomenon is like the situation in patients with spinal transection who have persistence of the PFs in the anesthetic lower extremities (Fig. 3A, S4), which is perhaps related to the size of the anesthetic part (Woods et al. 2000). These examples of reorganization in humans following injuries are consistent with the results of injuries in other species described below.

Injury Dependent Plasticity: Summary

The segmentation and specificity of thalamic somatosensory anatomy and physiology described above are consistent with the somatotopy of S1 cortex of animals, which is dramatically altered after peripheral nerve injury in many species. Following peripheral nerve injury in these animals, cortical somatotopic maps have shown substantial reorganization relative to the extent of the complete somatotopic map (Kaas et al. 1983; Merzenich et al. 1983). Many neurons did not have RFs, demonstrating that acute reactivation can occur but is relatively limited after section of peripheral nerves (Kaas et al. 1983; Kelahan and Doestch 1984; Rasmusson 1982).

Following long survival after a more extensive peripheral injury (C2–T4 dorsal rhizotomy; Levitt 1985; Sweet 1981), “massive” shifts of 1–2 cm were observed in cortical RF maps (Pons et al. 1991). This is about 10 times larger than shifts observed in the more limited nerve sections, as described above (Kaas et al. 1983; Kelahan and Doestch 1984). There is an increase in the representation in VP of the border zone for the body part that is anesthetic as a result of the nerve section (Garraghty and Kaas 1991). Digit amputation in adult raccoons leads to an increase in the representation of the stump with large RFs that include adjacent digits (Rasmusson. 1996a, 1996b).

These results demonstrate anatomical changes from healthy animals, including decreases in terminations of axons of the dorsal column nuclei, as indicated by thalamic CO staining (Rausell et al. 1992b). In healthy animals, neural elements of the lemniscal somatosensory pathway from the spinal cord (e.g., spinocervical tract; Willis and Coggeshall 2004) to the brain stem (e.g., dorsal column nuclei) and thalamus have different staining properties from those of the spinothalamic pathway. Elements in the lemniscal pathway immunostain for the calcium-binding protein parvalbumin, whereas those in the spinothalamic pathway stain for 28-kDa calbindin (Rausell and Jones 1991b; Rausell et al. 1991, 1992a).

In addition to the changes in dorsal column terminations, some features of injury-induced plasticity may result from the convergence of input arising from several different mechanoreceptor types upon a single thalamic neuron, as in the case of SA2 neurons (see above, Normal Somatic Sensory Thalamus: Responses to Optimal Tactile Stimuli in VP and Vc). Furthermore, more than one descriptor was often endorsed in response to microstimulation at a particular site, and evoked sensations were much less clustered than in the case of responses of thalamic neurons (Lenz et al. 1988b; Ohara et al. 2004). These clusters overlap due, in part, to stimulation of axons that branch or that are adjacent to the cell bodies which have been invoked as the mediators of microstimulation evoked sensations (Sherman and Guillery 2001). In the case of S1 and S2 cortex, there are parallel pathways from the thalamus to both areas, but units of isorepresentation are less specific and segregated in S2 (Burton 1986; Robinson and Burton 1980). Different weighting of these inputs may explain some features of somatotopic reorganization, as discussed below.

The rhizotomized monkeys display loss of parvalbumin cells in the brain stem and thalamus and downregulation of thalamic GABA_A receptors (Rausell et al. 1992b). Inhibitory GABAergic inputs to thalamocortical “relay” neurons are integral to the mechanism of low-threshold calcium spikes crested by characteristic low-threshold spike (LTS) bursts of action potentials that occur in all mammalian species (Domich et al. 1986; Jahnsen and Llinás 1984a, 1984b; Steriade et al. 1997b). In addition, dorsal column inputs to Vc have terminations on relay neurons, local interneurons that inhibit relay neurons, and glomeruli. Glomeruli are complexes of synapses, spines, and dendrites of relay neurons and interneurons (Guillery 1969; Ralston and Ralston 1994). The relay neurons contact 1) neurons in the inhibitory (GABA) network of the thalamic reticular nucleus, which then inhibit relay neurons and local inhibitory interneurons, and 2) cortical neurons and inhibitory interneurons, leading to excitatory input back on thalamic relay cells and interneurons (Sherman and Guillery 2001; Steriade et al. 1997a).

This complex inhibitory circuitry is impacted by changes in GABAergic elements after the major injuries to the peripheral component of the somatic sensory system described above (Woods et al. 2000; Rausell et al. 1992b). Synapses and glomeruli are altered following injuries to the somatic sensory system and may be involved in the abnormal physiology that results (Ralston et al. 1996, 2000), including increased LTS bursting (Lenz et al. 1994a, 1998a) and prolonged dystonia (Kobayashi et al. 2011). Finally, LTS bursting, may be related to the unpleasant dysesthesias that result from these injuries (Lenz et al. 1990; Walton et al. 2010).

Injuries of the central somatic sensory system lead to similar changes in human thalamic organization. For example, the proportions of thalamic cells with RFs that include the anesthetic border zone are increased following spinal cord injuries (Fig. 3). Following dorsal column interruption at the T3–T5 level, monkey VP contains fewer cells with hindlimb RFs and more cells with forelimb RFs (Pollin and Albe-Fessard 1979). Many cells have large RFs and respond to noxious stimuli, which suggest that these activations result from spinothalamic tract input. In monkeys at long survival after thoracic anterolateral cordotomy versus controls, there is increased spontaneous and evoked activity in response to cutaneous stimuli applied to the affected hindlimb (Ralston et al. 2000; Weng et al. 2000). RFs are larger than control and often involve more than one digit; fewer cells respond to noxious stimuli. Therefore, injury to either peripheral nerves or roots or the central nervous system can lead to substantial reorganization of the thalamic somatotopic map.

In patients with spinal cord transection, these mismatches result from reorganized input to the thalamus so that input from the border zone part the body (at the level of spinal transection) is mapped onto the relatively normal thalamocortical representation of the anesthetic part of the body (below the level of spinal transection; Lenz et al. 1989). Therefore, microstimulation at the site of neurons with border zone RFs will often produce sensations referred to the anesthetic part of the body leading to a mismatch. These results suggest that thalamic PF maps form a percept or image that is more stable than the representation based on maps of RFs. Abnormal activity within the border zone anesthetic regions will be reflected by abnormal sensations in the corresponding parts of the body, such as pain and dysesthesias (Lenz et al. 1990, 1998b).

Activity-Dependent Plasticity: Reorganization of Human Vc after Prolonged Dystonia

Dystonia is a movement disorder characterized by “continuous twisting movements and sustained postures” (Fahn 1988). Repetitive movements can be associated with reorganization of neuronal activity in the cutaneous mechanoreceptive somatic sensory pathway (Byl et al. 1996; Nagarajian et al. 1998; Nudo et al. 1996). Recordings of neuronal activity and microstimulation-evoked responses were studied in the Vc of dystonia patients who underwent stereotactic thalamotomy as treatment for this condition (Lenz and Byl 1999). Patients with ET who underwent these procedures served as controls.

In these patients, cells at two adjacent recording sites were said to have consistent RFs if the RF of both sites included any part of the body. The length of a trajectory with consistent RFs was defined as the distance along the electrode trajectory over which each RF included the same part of the body. The trajectory length was chosen for the part of the body with the maximal length of the trajectory with a consistent RF. The same protocol was applied to the analysis of consistent PFs. Of course, the distance along a trajectory over which RFs or PFs are consistent is longer for body parts with larger representations (Lenz et al. 1994a). This protocol for estimating size of representations is arbitrary but is applied in all subject groups (Lenz et al. 1988b, 1994a, 1998a).

The trajectory length was determined for the part of the body that maximized the length of the trajectory with a consistent RF. The length of the trajectory with consistent PFs was significantly longer in patients with dystonia than in the controls. The proportion of RFs including multiple parts of the body was greater in dystonia patients (29%) than in patients with ET (11%).

Finally, there was a substantial mismatch between RFs and PFs in patients with dystonia. In Fig. 5, each of the neuronal RFs and the PFs at a site was coded according to body parts (Fig. 5, inset), and all were plotted against each other (Fig. 5). The PF of a stimulation site where a cell was not recorded was paired with the RF of the closest recorded cell. For any RF and PF evoked at the same or an adjacent location in Vc (RF/PF pair), we plotted the part of the body involved in the RF of the recorded cell against the part of the body that was closest in the PF (Fig. 5, A, ET, and B, dystonia). Therefore, points on the diagonal indicate RF/PF pairs that are matched. Sites with matching were significantly less common in patients with dystonia (32%) than in those with ET (58%). These results demonstrate that mismatches of the RF and PF occur more commonly in patients with dystonia than in controls and are functionally significant evidence of activity-dependent plasticity.

Fig. 5.

Matches in neuronal RFs and PFs in patients with ET (A) and in patients with dystonia (B). See text for details. [Reproduced with permission from Lenz and Byl (1999).]

Activity-Dependent Plasticity: Summary

To our knowledge there are no studies of activity-dependent plasticity in monkey VP, although there are such studies in the corresponding cortical area S1. In monkey S1, plasticity has been studied in putative dystonic movements induced by repetition of a motor task involving rapid opening and closing of a hand grip (Byl et al. 1996; Hallett 1995). These monkeys developed hand cramps, disordered motor coordination, and posturing reminiscent of dystonia. Representations of the hand surface were reorganized in cortical area 3b (Byl et al. 1996). The cutaneous RFs extended across multiple digits and the whole hand. This somatotopic organization is dramatically different from the normal representation of the hand, which is defined by small, ordered, discrete RFs (Kaas 1991; Merzenich et al. 1987).

Activity-dependent alterations in RF maps have also been observed in cortex of monkeys trained to perform a number of different sensory tasks. For example, a skin locus such as the tip of a single finger has been stimulated at different frequencies while a monkey carries out a frequency discrimination task (Recanzone et al. 1992). This task produced changes in the representation of the hand in S1 cortex, whereas a similar result was found in human S1 cortex, as mapped by high-resolution electroencephalography (Braun et al. 2000). Cortical plasticity induced by these tasks can be characterized by unusually small RFs and by enlarged representations of the body part that is involved in the behavior (Jenkins et al. 1990; Recanzone et al. 1992). Finally, reorganization of the motor cortical representations of responses to sensory stimulation or motor outputs to the periphery can result from learned repetitive motor tasks (Nudo et al. 1992, 1996).

Injury- and Activity-Dependent Plasticity: Common Mechanisms across Vc and Other Nuclei

Studies of Vc in patients with nervous system injury and dystonia demonstrate large changes in the perceptual image of the body, as revealed by the mapping of PFs (Lenz et al. 1994a, 1998a; Lenz and Byl 1999). In amputees, the area of Vc at which microstimulation evokes sensations in the stump or phantom is larger than the corresponding part of the body in controls with ET (see above and Fig. 4). RF/PF mismatches are more common in patients with dystonia and spinal transections than in patients with ET (Lenz et al. 1994a). In major injuries to the nervous system there is marked reorganization of the map of sensory inputs (RFs) in Vc or VP but less marked reorganization of the map of sensations evoked by microstimulation in Vc, leading to RF/PF mismatches. Similar changes occur in the motor thalamus as described below. In both sensory and motor systems, these activity-dependent changes in cortical organization may involve thalamocortical and corticothalamic pathways, which form networks that jointly process inputs or outputs, or both (Churchland and Sejnowski 1992; Destexhe and Sejnowski 2001).

One feature of this network may be activity in the corticothalamic pathway, which increases the membrane conductance noise in thalamic relay neurons (Debay et al. 2004; Wolfart et al. 2005). This increased variance leads to spike trains composed of both LTS bursts and single spikes (as in Kim et al. 2009) and to more linear thalamic relay cell stimulus response functions (Debay et al. 2004; Wolfart et al. 2005). This kind of firing has been observed in the response of neurons in Vc to cutaneous somatic sensory stimuli and is consistent with the sensations produced by microelectrode pulses delivered with the same temporal pattern (Lenz et al. 2004; Patel et al. 2006). Therefore, this change in thalamic neuronal function as a result of cortical activity may well be critical factor in plastic changes in both thalamic and cortical function.

Both injury- and activity-dependent reorganization are associated with increases in thalamic LTS bursting, which may be a mediator of plasticity (Kobayashi et al. 2013; Lenz et al. 1998a). The significance of LTS bursting in reorganization is suggested by its presence in sensory and motor nuclei in patients with dystonia, regardless of whether dystonia is due to an organic or psychogenic etiology (Kobayashi et al. 2011, 2013). Mixed LTS and single spikes in a train may be the result of traffic in thalamocortical assemblies, which lead to the functional changes described above (Bal et al. 2000; Debay et al. 2004; Wolfart et al. 2005).

Both thalamocortical assembly and LTS-related mechanisms may apply to plasticity in thalamic nuclei more generally, such as the motor nuclei (Vim, Vop). In sensory and motor nuclei, mismatches seem to occur because RF maps are changeable but PFs and motor outputs (Cheney and Fetz 1984) are relatively stable, where motor outputs are measured by spectral cross-correlation of neuronal and electromyographic (EMG) activities (Kobayashi et al. 2011, 2013). These mismatches may contribute to movement disorders including dystonia (Kobayashi et al. 2011; Lenz et al. 1998c, 1999) and tremor (Kiss et al. 2003; Lenz et al. 1994b). In dystonia, EMG activation often spreads to muscles that are not normally involved in the planned movement, which may be due to the mismatches in the motor thalamus. Therefore, the injury- and activity-dependent plasticity of thalamic function and organization that is so clear in Vc may also occur in other thalamic nuclei and may contribute to a range of motor and sensory disorders. These mechanisms provide possible avenues for therapy by manipulation of GABAergic neurotransmission and thalamic low-threshold calcium spikes (Kim et al. 2009; Kobayashi et al. 2013).

GRANTS

Some of the results reviewed in this article were supported by grants (to F. A. Lenz) from Eli Lilly Corporation, National Institute of Neurological Disorders and Stroke Grant NS38493, and the Johns Hopkins Neurosurgical Pain Research Institute.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.-C.S., J.-H.C., J.D.G., I.G., N.W., S.O., and F.A.L. interpreted results of experiments; A.-C.S. and F.A.L. drafted manuscript; A.-C.S., J.-H.C., J.D.G., I.G., N.W., S.O., and F.A.L. edited and revised manuscript; F.A.L. conception and design of research; F.A.L. approved final version of manuscript.